COMMENT pubs.acs.org/est
Environmental Science & Technology Presents the 2011 Excellence in Review Awards
I
n recognition of the contribution that reviewers provide to the scientific community and the publication of scholarly research, ES&T Editor Jerald L. Schnoor and the associate editors of the journal are proud to recognize the following reviewers for their significant contributions to the journal over the past year. A journal exists on the goodwill and expertise of its reviewer corps. Our editors rely on high-quality, constructive, and timely reviews from the more than 5000 reviewers in our database to ensure the excellence of papers that appear in ES&T. We realize that reviewing is a time-consuming job with no tangible reward. In recognition of this, and in order to express our gratitude to some of our best reviewers, ES&T established an annual “Excellence in Review” award in 2003 to honor reviewers who have consistently provided both scholarly and timely reviews. This year, we introduce our first “Super Reviewer” award to acknowledge and express our appreciation to three previous winners who continue to provide large numbers of high quality reviews. We salute all of the individuals below.
Alison M. Cupples Michigan State University, East Lansing, Michigan http://www.egr.msu.edu/∼cupplesa/
’ SUPER REVIEWER AWARD
Arpad Horvath University of California, Berkeley, California http://www.ce.berkeley.edu/∼horvath/
David J. Ehresman 3M Company, St Paul, Minnesota Steven J. Eisenreich Joint Research Centre, European Commission, Brussels, Belgium http://jrc.ec.europa.eu Claudia Gunsch Duke University, Durham, North Carolina http://gunsch.pratt.duke.edu/ Stuart Harrad University of Birmingham, Birmingham, United Kingdom http://www.gees.bham.ac.uk/staff/harradsj.shtml
Bruce E. Logan Penn State University, University Park, Pennsylvania www.engr.psu.edu/ce/enve/logan/
Xia Huang Tsinghua University, Beijing, China http://www.tsinghua.edu.cn/publish/enven/index.html
Dionysios (Dion) D. Dionysiou University of Cincinnati, Cincinnati, Ohio http://www.eng.uc.edu/dept_cee/people/faculty/dionysiou
Arturo A. Keller University of California, Santa Barbara, California www.bren.ucsb.edu/∼keller
Jason C. White The Connecticut Agricultural Experiment Station, New Haven, Connecticut http://www.ct.gov/caes/cwp/view.asp?a = 2812&q = 345092
Tamar Kohn Ecole Polytechnique Federale de Lausanne, Switzerland http://lce.epfl.ch Jung-Hwan Kwon Ajou University, Suwon, Republic of Korea http://www.ajou.ac.kr/∼jhkwon
’ EXCELLENCE IN REVIEW AWARD Souhail Al-Abed USEPA, National Risk Management Research Laboratory, Cincinnati, Ohio www.epa.gov
Rainer Lohmann University of Rhode Island, Narragansett, Rhode Island http://www.gso.uri.edu/users/lohmann
Hans Peter Arp Norwegian Geotechnical Institute, Norway http://www.ngi.no/no/
Shaily Mahendra University of California, Los Angeles, California http://www.cee.ucla.edu/faculty/mahendra/profile
Laurie S. Balistrieri US Geological Survey, Seattle, Washington http://minerals.usgs.gov/west/spokane/laurie.htm
Jonathan W. Martin University of Alberta, Edmonton, Canada http://lmp.med.ualberta.ca/Admin/Faculty/Pages/default. aspx?P = 38
Patrick Brezonik University of Minnesota, Minneapolis, Minnesota http://www.ce.umn.edu/directory/faculty/brezonik.html r 2011 American Chemical Society
Published: October 28, 2011 9113
dx.doi.org/10.1021/es202956u | Environ. Sci. Technol. 2011, 45, 9113–9114
Environmental Science & Technology
COMMENT
Andrew A. Meharg University of Aberdeen, Aberdeen, Scotland http://www.abdn.ac.uk/biologicalsci/staff/details/a.meharg Hans W. Paerl University of North Carolina, Chapel Hill, North Carolina http://www.unc.edu/ims/paerllab/ Gene Parkin University of Iowa, Iowa City, Iowa http://www.engineering.uiowa.edu/faculty-staff/profiledirectory/cee/parkin_g.php Alessandro Piccolo Universita di Napoli Federico II, Portici, Italy http://www.suprahumic.unina.it Amy Pruden Virginia Tech, Blacksburg, Virginia http://www.cee.vt.edu/people/pruden.html Debra R. Reinhart University of Central Florida, Orlando, Florida http://www.cece.ucf.edu/people/reinhart/ John Sumpter Brunel University, London, United Kingdom http://www.brunel.ac.uk/about/acad/ife Hideshige Takada Tokyo University of Agriculture and Technology, Tokyo, Japan http://www.pelletwatch.org/ Ronald F. Turco Purdue University, West Lafayette, Indiana http://www.ag.purdue.edu/agry/Pages/rturco.aspx Clemens von Sonntag Max Planck Institute for Bioinorganic Chemistry, Muelheim an der Ruhr and University of Duisburg-Essen, Essen, Germany http://www.mpibac.mpg.de/bac/index_en.php Eddy Y. Zeng Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou, China http://sourcedb.cas.cn/sourcedb_gig_cas/yw/rckyw/ 200907/t20090710_2057308.html Jerald L. Schnoor Editor ES&T, University of Iowa
9114
dx.doi.org/10.1021/es202956u |Environ. Sci. Technol. 2011, 45, 9113–9114
COMMENT pubs.acs.org/est
Eawag at 75
F
ew institutions and none outside the U.S. have been associated with ES&T as closely as Eawag, the Swiss Federal Institute of Aquatic Science and Technology. This shared history includes the service of Prof. Alexander (Sascha) Zehnder (Eawag Director, 1993 2004) as Senior Associate Editor for Europe and that of numerous Eawag researchers (W. Giger, J. Hering, A. Johnson, L. Sigg, R. Schwarzenbach, U. von Gunten) as Associate Editors and/or members of the Editorial Advisory Board. This history reflects the common interests of ES&T and Eawag, which celebrates its 75th anniversary this year. Founded in 1936 as an Advisory Center for Wastewater Treatment and Drinking Water Supply at the Swiss Federal Institute of Technology (ETH) in Zurich, Eawag moved to its present location in D€ubendorf in 1970, concurrent with the arrival of its new Director, Professor Werner Stumm. ES&T began publishing in 1967, and its early years were also critical transition years for Eawag under Stumm’s leadership. Under Stumm, Eawag’s central focus on the fundamental understanding of the processes determining water quality in technical systems and in aquatic ecosystems was established. This focus on chemical, physical and biological processes was and is necessarily supported by advances in analytical and modeling capabilities. The strong science base developed under Stumm allowed Eawag to contribute to solving many of the acute and visible problems of environmental degradation that were widespread in industrialized countries, including Switzerland. In addition, Eawag initiated its 30-year tradition of research for development and capacity building in developing countries (see photo of Eawag researcher testing well water for arsenic with children in Sumatra).
natural sciences and engineering at Eawag. This was paralleled by the explicit inclusion of environmental policy analysis in ES&T beginning in 1995. The planning, design and construction of Eawag’s “zero-energy” building, the Forum Chriesbach, was initiated during Zehnder’s term and completed under his successor Ueli Bundi. Eawag’s research on No-Mix technology (featured as the cover article of ES&T issue 9 in 2001) is the basis of one of the many energy- and resource-conserving features incorporated in the Forum Chriesbach. Eawag’s emphasis on fundamental process understanding, the integration of engineering, natural and social sciences, and the implementation of research advances into practice are critical elements to solve urgent environmental problems worldwide. It is one of the great challenges of our time to meet societal needs for natural resources while preserving the capacity of the environment to provide essential ecosystem services. Research institutions, like Eawag, and prominent environmental journals, like ES&T, have an important role to play in ensuring a sustainable future. In addition to their primary roles in performing and disseminating research, such institutions can and should seek to build broader awareness of environmental issues, address the need to preserve environmental data and make them widely accessible, support collaboration and cooperation within the environmental science community, and help to identify and promote solutions to the pressing environmental problems of our time. Congratulations to Eawag for 75 years of such dedication and service to a quality environment. Janet G. Hering Director, Eawag Jerald L. Schnoor Editor
’ AUTHOR INFORMATION Corresponding Author
[email protected].
Lenny Winkel, Eawag Arsenic-sampling at a drinking fountain in Sumatra.
The overlapping interests of Eawag and ES&T provided the motivation for Stumm’s successor, Prof. Alexander Zehnder, to establish the ES&T European office at Eawag in 2000. Zehnder also emphasized the integration of the social sciences with the r 2011 American Chemical Society
Published: September 30, 2011 9115
dx.doi.org/10.1021/es203291e | Environ. Sci. Technol. 2011, 45, 9115–9115
VIEWPOINT pubs.acs.org/est
Legitimate Conditions for Climate Engineering Richard Owen University of Exeter, U.K. decades of knowledge and incorporated into trials of efficacy and safety. We do not have this for the emerging science of climate engineering and are therefore compelled to proceed under conditions of ignorance. The response is that we should establish strong research governance processes, developing and then employing tests of efficacy and safety before any decision to deploy (i.e., proceed with caution) in the same way we have built up understanding of pharmaceutical efficacy and safety over time and incorporated this into the tests required of medicines before use. This is to be recommended.
O
n September 13th scientists announced preparations were underway for the first UK field trial of climate engineering feasibility.1 The proposed trial will be modest: it will pump water through a 1 km high balloon-tethered hose, to assess the feasibility of reflective particle injection high into the atmosphere, mimicking the temperature-reducing effects of volcanic eruptions. But it has stimulated considerable debate about whether research in this controversial field should be undertaken at all, and if so the conditions under which it is acceptable to proceed. Responding, the President of the UK’s Royal Society, Paul Nurse, replied that there should be research on both the efficacy and safety of geoengineering:2 “One would not take a medicine that had not been rigorously tested to make sure that it worked and was safe. But, if there was a risk of disease, one would research possible treatments and, once the effects were established, one would take the medicine if needed and appropriate. Similarly we need controlled testing of any technologies that might be used in the future”. His comments, and specifically this analogy to pharmaceuticals, raise important questions concerning the conditions under which we decide to deploy controversial technologies such as solar radiation management. Pharmaceuticals indeed go through a rigorous testing process before they are authorized for use (“data before market”), but this is because we know the harmful effects to look for and there are well-established test methods to quantify these, built up over r 2011 American Chemical Society
’ THE LIMITS OF KNOWLEDGE There is however an important caveat to this approach. Despite our best intentions, it may only be once deployment has actually occurred that any nasty surprises surface. The history of nasty surprises is long, from CFCs to asbestos.3 Indeed surprises such as thalidomide were a major driver of regulation for pharmaceuticals, which in turn strives to ensure these effects do not occur again. But this happens after the fact. Regulation is often blind to that which it has not encountered before. Such unanticipated effects might not emerge for solar radiation management, but this will always be a gamble for which the probabilities can never be known, a point acknowledged by the Royal Society in 2009. The unintended side effects of many well-intentioned innovations have not been predicted. Here there is an analogy with pharmaceuticals: despite tests, who could have predicted that the birth control pill would cause environmental endocrine disruption?4 The argument is that research can help us rule out the technology, on the basis of efficacy, safety or both. But what happens if it is not ruled out? What if, after careful consideration of risks and feasibility, solar radiation management becomes a serious option? Who then would be prepared to place a bet for which the stakes can never be fully known? Perhaps the seriousness of climate change would make deployment a gamble worth taking. But who would make that decision? Who would have the authority to make a (possibly intergenerational) commitment to solar radiation management? Who would decide that conventional attempts at carbon management and climate change mitigation had proved insufficient or unsuccessful? Who would negotiate the distribution of impacts across the globe (beneficial or otherwise, known or unknown) that might result? Who would compensate those who suffer for the collective good? What are the conditions for such planetary technological gambles? Received: September 21, 2011 Accepted: September 28, 2011 Published: October 10, 2011 9116
dx.doi.org/10.1021/es2033185 | Environ. Sci. Technol. 2011, 45, 9116–9117
Environmental Science & Technology
VIEWPOINT
In his essay on the Imperative of Responsibility5 Jonas wrote “One would not deny the statesman the right to risk his nation’s existence for its future if really ultimate issues are at stake. It is in this manner that awesome but morally defensible decisions about war and peace come about when, for future’s sake, the stake is the future itself’. He added that ‘this should never happen because of the enticement of a wonderful future but only under the threat of a terrible future”. The supreme “malum” justifies a collective wager. This has been the catalyst for many technological wagers in the past, of which the push for mass production of penicillin in World War Two is arguably one. Would the prevention of a terrible future (a “climate change tipping point” for example) be a legitimate condition for a collective gamble on a geoengineering solution? Perhaps, but this presupposes that this condition has been collectively arrived at, and that there is a mechanism for this to be achieved. This does not currently exist. It is particularly important for approaches such as solar radiation management which may have impacts that may be trans-national and unequally distributed in nature. It is here that Nurse’s statement “one would take the medicine if needed and appropriate” becomes critical. Who will decide there is a need, that climate engineering is an “appropriate medicine”? The conditions for making such a technological wager legitimate, democratic, and equitable must be explicit. For we are naive to assume the decision to administer a medicine by a physician is based on efficacy and safety alone. And we will pay the price if we fail to acknowledge that there are limits to knowledge and ignore the lessons of history.
’ REFERENCES (1) http://www.nerc.ac.uk/press/releases/2011/22-spice.asp (accessed June 10, 2011). (2) http://www.guardian.co.uk/environment/2011/sep/08/geoengineering-research-royal-society (accessed June 10, 2011). (3) http://www.eea.europa.eu/publications/environmental_issue_ report_2001_22 (accessed June 10, 2011). (4) Jobling S., Owen R. Ethinyl oestradiol: Bitter pill for the precautionary principle. In: Late Lessons from Early Warnings II; European Environment Agency, (in press). (5) Jonas, H. The Imperative of Responsibility; University of Chicago Press, Chicago, 1984.
9117
dx.doi.org/10.1021/es2033185 |Environ. Sci. Technol. 2011, 45, 9116–9117
VIEWPOINT pubs.acs.org/est
Green Power A Modest Experiment Michael R. Piggott* Chemical Engineering and Applied Chemistry, University of Toronto
A
t the present time, the Ontario government is encouraging people to invest in solar and wind power generation. They do this by subsidizing any excess power produced and fed into the grid at a rate many times the cost of power generated from big conventional plants. The Ontario Power Authority is offering feed-in tariff contracts at between 45 and 80.2 cents to companies building new solar power generating facilities, 13.5 cents on landbased wind farms, 19 cents on off-shore wind farms, and between 10.4 and 19.5 cents on biogas projects.1 3 The experiment described herein is a test of the economics of domestic production from a small and easily affordable system, on the shore of Lake Huron, in the light of these figures. Two solar panels, with a rated capacity of 80 W each, were mounted on the roof of a shed, which was located in an exposed spot. They faced approximately south and were about 23° to the horizontal. Having only 100 foot wide lot, a tall tower was not practical for a wind generator, since it would have to be guyed sufficiently to withstand gale force winds from any direction. So a 900 W wind generator was mounted between two cedar tree trunks, about 35 feet tall, set two feet apart in an approximately east west direction, and another cedar trunk, about 20 feet tall, set at 12 feet north of the westmost 35 foot cedar. They were braced by connecting them to each other, and to the shed. The cedar posts were embedded in reinforced concrete, resting on the bedrock, about two feet beneath. The wind generator was attached to the top of a 2 1/2 in. diameter galvanized steel pipe, about 24 feet long. The pipe was in two halves joined at the middle by a specially shaped hollow rod. The assembly was pivoted in the middle by a one in. diameter hardened steel rod. This steel rod went right through the support cedars, with the 2 1/2 in. pipe assembly in the middle. The generator was fixed to one end of the assembly, and balanced by steel slugs affixed inside the other end. Thus the generator could be swung down to connect the electrical conductors, and swung up again for wind generation of electrical power. (And swung down again for service.) This structure withstood gale force winds (sustained winds of 63 84 km/h) on many occasions. A gust of 162 km/h was r 2011 American Chemical Society
recorded on the anemometer in May 2010 and another, of 165km/h, in November of that year. The three phase electrical power produced by the generator was regulated and rectified by a special “charge controller”, and was stored in deep cycle lead-acid batteries. The power produced was dissipated as heat. An automatic switching system (Schmidt trigger) was designed and constructed to turn the 115 V heater on when the battery voltage reached 12.8 V, and turn it off again when the voltage fell to 11.4 V. The time that the heater was turned on was logged. With the aid of an anemometer, the power output of the generator was plotted as a function of wind velocity, averaged over about 12 h. This showed that a 16km/h wind was required before any power was generated. Thereafter, the power increased approximately linearly to 340 W at 40km/h. By extrapolation of the straight line, it was predicted that a 50km/h wind was required to produce 500 W. But above 30km/h wind velocity, the generator was designed to fold, so it was avoiding the full force of the wind. The manufacturer’s design figure of 900 W at 45km/h thus appears to be very optimistic, as reported elsewhere for these devices.4 (Note that the power was fed through an inverter to produce 120 V AC from 12 V DC. This conversion was about 90% efficient. So 900 W from the generator produces about 810 W of useable AC power.) The power produced by the solar panels working in tandem with the wind generator was logged for 12 months, from January 1 2010 to January 31 2011, see Figure 1. (April 2010 did not yield reliable results, due to instrument problems.) While there are significant variations, the overall mean useable power produced was 28.3 W. The solar panels only contributed a significant amount of power in the summer. They were covered in snow in December, January, February, and part of March. The monthly averages conceal much day-to-day variation. For instance, in June 2010 the power varied between 2 and 120 W. The out-of pocket cost of the setup was about 7450 Canadian dollars. The wind generator cost $2933, the solar panels $1333, the batteries $575, the inverter $580, and the concrete and tower hardware about $1150. Heavy cable was needed for the wind generator charge controller battery connections, and 115/230 V cable was needed to connect the system to the house. Hidden costs include the cedar tree trunks, the design and building of the Schmidt trigger from its basic components, and the value of the excavation, etc. work associated with the construction of the wind generator support. The current (March 2011) cost of domestic electricity in the household concerned with the project was 14.82 cents per Received: September 21, 2011 Accepted: September 26, 2011 Published: October 05, 2011 9118
dx.doi.org/10.1021/es2033229 | Environ. Sci. Technol. 2011, 45, 9118–9119
Environmental Science & Technology
VIEWPOINT
Figure 1. Monthly average power produced by the combined efforts of the solar panels and wind generator. Overall average for 12 months was 28.3 W. (Allowing for the 90% efficiency of the inverter, that amounts to 31.4 W at 12 V DC.).
kilowatt hour. $7450 would, at that rate buy 50,247 kWh. The solar/wind power assembly, working at an average rate of 30 kW would take about 190 years to produce the same amount of electricity. While there is no figure given for the feed-in tariff for solar/ wind domestic generation, we could take an average between 80.2c (solar) and 13.5c (wind farm), that is, 46.85c, and see how long the power generated would have to be fed into the grid to recover $7450 at that rate. We thus make the same calculation as in the previous paragraph, replacing 14.82c with 46.85c. This comes to about 60 years. Thus wind/solar power is not a good investment. In addition to spending $7430, there was a lot of time spent designing and constructing the system. All for about 30 W.
’ AUTHOR INFORMATION Corresponding Author
*Phone 519-534-5592; e-mail
[email protected].
’ REFERENCES (1) Corcoran, T.; National Post (Toronto) September 30th 2010. (2) Source Watch: Comparative Electrical Generation Costs; California Regulatory Agencies data for May 2008. (3) Ontario Power Authority FIT Power Price Schedule, August 13th 2010. (4) Consumer Reports, October 2011, page 30.
9119
dx.doi.org/10.1021/es2033229 |Environ. Sci. Technol. 2011, 45, 9118–9119
VIEWPOINT pubs.acs.org/est
The Challenge of Exposure Correction for Polar Passive Samplers— The PRC and the POCIS Christopher Harman,†,* Ian John Allan,† and Patrick Steven B€auerlein‡ †
Norwegian Institute for Water Research (NIVA), Oslo Centre for Interdisciplinary Environmental and Social Research (CIENS), Gaustadalleen 21, NO-0349, Oslo, Norway ‡ KWR Watercycle Research Institute, P.O. Box 1072, 3430 BB, Nieuwegein, The Netherlands
P
assive sampling devices such as the polar organic chemical integrative sampler (POCIS)1 may have much to offer in response to the challenge of measuring low and fluctuating concentrations of polar compounds of interest in aquatic environments. For example they have recently been used to obtain illicit drug monitoring data that would not have been practically and financially possible to achieve using other sampling methods.2 One of the biggest challenges facing the quantitative use of such samplers is the lack of a method to correct for in situ exposure conditions (water flow rates, temperature, pH etc.) which are known to affect uptake rates. This issue has been elegantly overcome for hydrophobic passive samplers by the use of so-called performance reference compounds (PRCs).3 These analytically noninterfering substances are spiked into samplers prior to deployment and as their dissipation follows first order kinetics analogous to uptake they can be used to estimate sampling rates (Rs) of target compounds in situ. This is not currently the case for polar passive samplers. Despite a comprehensive consideration of the processes involved in chemical uptake in the original description of POCIS,1 most of the subsequent work has concentrated on making measurements rather than trying to understanding the mechanisms involved. r 2011 American Chemical Society
Thus theories from hydrophobic passive sampling have been applied for the polar samplers despite the fundamentally different processes of two-way isotropic exchange in hydrophobic samplers and sorption phenomena in polar samplers. Such an approach might be risky, especially with the range of interactions that need to be taken into account when considering polar compounds, particularly those containing N and O; H-bonding between water and solute, between water and sorbent, between solute and sorbent, van der Waals interactions etc. The majority of published POCIS studies (including our own) have made measurements directly in the environment using Rs derived from laboratory calibrations. Of these the static renewal type experiment dominates, which although likely to be the most reproducible between studies, is the one most unlikely to be replicated under field conditions. This means that although results are often presented as time-weighted average concentrations, in reality, due to the absence of a dedicated PRC approach (or equivalent), they are at best semiquantitative. For exactly these reasons investigations have been undertaken to develop a PRC approach for POCIS, built around the observation that some compounds could be released again after uptake.4 This is perhaps unsurprising as it was earlier shown that OASIS HLB exhibits both the properties of an adsorbent, and a contribution of some partitioning mechanism.5 All compounds that are suitable to be POCIS PRCs, that is, have the ability to desorp (or poor sorption) are likely to have relatively similar retention mechanisms for the sorbent. This means finding suitable PRCs for compounds which do not behave in this way, that is, those which more strongly bound might be problematic. The question thus arises can one correct for the other? As one objective of the PRC approach is to use several compounds to correct for a suite of similar ones, this is an important question. This may be plausible if the change in Rs due to a reduction in the water boundary layer for example, is equally represented for weakly bound PRCs on the way out of the sampler and strongly bound analytes on the way in. However, the situation is further complicated by an apparent increasing interaction with the polyethersulphone membrane used in POCIS, with increasing target compound hydrophobicity and the importance of apolar moieties generally for retention on OASIS HLB.5 Thus (assumed) weakly bound and rapidly equilibrating compounds that appear suitable for use as POCIS PRCs are unlikely to be Received: September 26, 2011 Accepted: September 27, 2011 Published: October 06, 2011 9120
dx.doi.org/10.1021/es2033789 | Environ. Sci. Technol. 2011, 45, 9120–9121
Environmental Science & Technology
VIEWPOINT
able to be used to correct for those with different interactions. Additionally as sorption is a competitive process, it is plausible that compounds with strong interactions can replace those with weak ones including PRCs? If this is the case then higher PRC dissipations may be seen, and thus higher Rs estimated, with increasing concentrations of competitive substances in exposure water. As a result even if it can be shown in laboratory experiments that the PRC approach works for certain compounds and exposure scenarios, these may not necessarily be applied to all compounds and all exposures. A point Mazzella and co-workers4 are careful to point out, but one which appears to be increasingly overlooked, based on presentations at the recent passive sampling workshop and symposium (Krakow, Poland). Other external correction methods may be applicable, such as using hydrophobic samplers as a “surrogate PRC approach”, or in situ calibration.2 However these approaches are also not without their challenges. Additionally the use of different materials, for example silicone rubber, which appears to be able to sample several groups of medium polar compounds, is worth investigating. There are many gaps in our knowledge concerning polar samplers and the processes involved in accumulation. For example the first detailed studies considering transport through the PES membrane are only just emerging. In summary we need to address some of these basic questions before we can hope to use polar passive samplers in a truly quantitative way. Even accepting that OASIS HLB may exhibit partitioning properties for some compounds, we should not correct sorption phenomena which we understand poorly with desorption phenomena (PRCs) we understand even less. The use of PRCs for certain exposure situations may be possible, but an all-encompassing, robust approach for POCIS and similar types of samplers appears unlikely. We suggest that, currently, polar passive samplers are a useful screening tool, which may be used to estimate not calculate the order of water concentrations.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +47 22 18 51 00; fax: +47 22 18 52 00; e-mail:
[email protected].
’ REFERENCES (1) Alvarez, D.; Petty, J.; Huckins, J.; Jones-Lepp, T.; Getting, D.; Goddard, J.; Manahan, S. Development of a passive, in situ, integrative sampler for hydrophilic organic contaminants in aquatic environments. Environ. Toxicol. Chem. 2004, 23, 1640–1648. (2) Harman, C.; Reid, M.; Thomas, K. V. In situ calibration of a passive sampling device for selected illicit drugs and their metabolites in wastewater, and subsequent year-long assessment of community drug usage. Environ. Sci. Technol. 2011, 45, 6233–6234. (3) Booij, K.; Sleiderink, H.; Smedes, F. Calibrating the uptake kinetics of semipermeable membrane devices using exposure standards. Environ. Toxicol. Chem. 1998, 17, 1236–1245. (4) Mazzella, N.; Lissalde, S.; Moreira, S.; Delmas, F.; Mazellier, P.; Huckins, J. N. Evaluation of the use of performance reference compounds in an oasis-HLB adsorbent based passive sampler for improving water concentration estimates of polar herbicides in freshwater. Environ. Sci. Technol. 2010, 44 (5), 1713–1719. (5) Dias, N. C.; Poole, C. F. Mechanistic study of the sorption properties of OASIS HLB and its use in solid phase extraction. Chromatographia 2002, 56, 269–275. 9121
dx.doi.org/10.1021/es2033789 |Environ. Sci. Technol. 2011, 45, 9120–9121
CRITICAL REVIEW pubs.acs.org/est
Connecting the Dots: Responses of Coastal Ecosystems to Changing Nutrient Concentrations Jacob Carstensen,*,† María Sanchez-Camacho,‡ Carlos M. Duarte,‡,§ Dorte Krause-Jensen,† and Nuria Marba‡ †
Department of Bioscience, Aarhus University, Frederiksborgvej 399, DK-4000 Roskilde, Denmark Department of Global Change Research, IMEDEA (CSIC-UIB), Instituto Mediterraneo de Estudios Avanzados, Miquel Marques 21, 07190 Esporles (Illes Balears), Spain § The UWA Oceans Institute, University of Western Australia, 35 Stirling Highway, Crawley 6009, Australia ‡
bS Supporting Information ABSTRACT: Empirical relationships between phytoplankton biomass and nutrient concentrations established across a wide range of different ecosystems constitute fundamental quantitative tools for predicting effects of nutrient management plans. Nutrient management plans based on such relationships, mostly established over trends of increasing rather than decreasing nutrient concentrations, assume full reversibility of coastal eutrophication. Monitoring data from 28 ecosystems located in four well-studied regions were analyzed to study the generality of chlorophyll a versus nutrient relationships and their applicability for ecosystem management. We demonstrate significant differences across regions as well as between specific coastal ecosystems within regions in the response of chlorophyll a to changing nitrogen concentrations. We also show that the chlorophyll a versus nitrogen relationships over time constitute convoluted trajectories rather than simple unique relationships. The ratio of chlorophyll a to total nitrogen almost doubled over the last 3040 years across all regions. The uniformity of these trends, or shifting baselines, suggest they may result from large-scale changes, possibly associated with global climate change and increasing human stress on coastal ecosystems. Ecosystem management must, therefore, develop adaptation strategies to face shifting baselines and maintain ecosystem services at a sustainable level rather than striving to restore an ecosystem state of the past.
’ INTRODUCTION Increased nutrient inputs to coastal ecosystems, derived from the rapid rise in fertilizer use in agriculture, production of manure from farm animals, domestic sewage, and atmospheric deposition associated with fossil fuel combustion,14 have led to the global spread of coastal eutrophication since the late 1970s. Realization of the negative effects of eutrophication, involving the loss of value of coastal ecosystem services,1,5,6 prompted efforts, initiated in the late 1980s, to reduce nutrient inputs. The result was expected to be a phase of oligotrophication with decreasing primary production7 which would reverse the effects of eutrophication and return coastal ecosystems to an earlier state.4,8,9 However, recent analyses have provided evidence that reduced nutrient inputs often fail to fully reverse the trajectories of ecosystems during eutrophication and have challenged the assumption that oligotrophication drives coastal ecosystems back to their original condition.10 The expectation that reduced nutrient inputs would reverse eutrophication effects originated from predictions derived from broad-scale relationships between chlorophyll a concentration (Chla), as an indicator of algal biomass, and nutrient concentrations across coastal ecosystems.1113 These relationships were r 2011 American Chemical Society
comparable to those developed in the 1970s for lake ecosystems1416 and confirmed experimentally (e.g., 1719). Yet, the empirical basis supporting use of the general relationship to predict oligotrophication responses was lacking, as all case studies and experimental tests to the 1990s reflected ecosystem responses to addition of nutrients, rather than to their removal. Hence, the use of relationships between Chla and nutrient concentrations to predict the response of coastal ecosystems to oligotrophication rests on the assumption that eutrophication is a fully reversible process involving a single path identical for eutrophication and oligotrophication trajectories. This fundamental tenet underlying coastal ecosystem management has not been sufficiently tested to date, but the current availability of dozens of cases of individual ecosystems undergoing eutrophication and subsequent oligotrophication now allows such tests to be conducted. Empirical relationships between Chla and nutrient concentrations, presented as general static relationships, have been based Received: December 1, 2010 Accepted: September 29, 2011 Revised: September 23, 2011 Published: September 29, 2011 9122
dx.doi.org/10.1021/es202351y | Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology
CRITICAL REVIEW
Table 1. Surface Samples of Water Quality Data Used in the Analysisa no. of observations region
no. of coastal ecosystems
no. of annual means
years
TN
TP
Chla
TN
TP
Chla
source www.chesapeakebay.net
Chesapeake Bay
7
19842006
17133
17465
21427
149
149
159
Denmark coast
10
19772006
12409
12354
12211
256
257
257
www.dmu.dk
Tampa Bay
4
19772006b
12825
17889
17885
104
120
120
www.tbeptech.org
Wadden Sea (Dutch part)
7
19772006c
3828
3878
3323
203
200
204
www.waterbase.nl
a
Data were downloaded from public monitoring databases with long time series (>20 years) of coastal ecosystems from four regions. b No TN data before 1981. c TN and TP calculated as sum of measured particulate and filtered fractions between 1991 and 1996.
on data from many different ecosystems encompassing wide ranges of Chla and nutrient concentrations.12,13,20 However, the broad, order-of-magnitude variability characteristic of these relationships may not represent only random variability, but may partially derive from diverse and idiosyncratic responses within individual ecosystems and/or changes in the nature of these relationships over time. Indeed, our conceptual understanding of ecosystem functioning has evolved from the initial expectation of a uniform response to accommodate a diversity of responses to nutrient inputs.8 Increased availability of long-term time series describing the responses of coastal ecosystems to changes in nutrient concentrations now makes it possible to connect the dots in Chlanutrient relationships to examine the trajectories of individual ecosystems over time10 as well as to examine variability in trajectories among ecosystems. Such analyses may provide an improved basis to derive expectations on the possible response of individual coastal ecosystems to increases as well as to decreases in nutrient inputs. Here we use long-term monitoring data to explore the existence of general patterns in the relationship between Chla and nutrient concentrations in 28 coastal ecosystems from four different regions where eutrophication has led to the implementation of management plans to reduce nutrient inputs. We aim to examine whether the relationship between Chla and nutrient concentration follows similar pathways through periods of eutrophication and oligotrophication. We do so by deconstructing general Chlanutrient relationship to examine the variability among regions and individual ecosystems, as well as the variability in these relationships over time. Data Sources and Processing. Water quality data were obtained from four public monitoring databases in Europe and North America (Table 1), constituting some of the most comprehensive and longterm data sets in the world. Although some of the databases had observations prior to 1977, the analysis was restricted to 1977 2006 for consistency across regions. The compiled database included over 45 000 observations of nutrient as well as chlorophyll a concentrations from the surface layer. Nutrient and Chla concentrations, the most widely used indicators of nutrient-driven eutrophication in the literature, were used as proxies of organic enrichment in coastal ecosystems.1 Sampling stations with salinity <6 were discarded from the analysis to focus on mesohaline and polyhaline coastal ecosystems. For each coastal ecosystem (n = 28) within the four regions, annual means of total nitrogen (TN), total phosphorus (TP), and Chla were computed from a general linear model9 after employing a log-transformation to normalize the concentrations. logðconc:Þ ¼ stationi þ yearj þ monthk þ eijkl
ð1Þ
Stationi described spatial differences between monitoring stations within the ecosystem, yearj described the interannual variation, and monthk described the seasonal pattern. Annual means were chosen, as opposed to seasonal means, to achieve comparable values across all regions, i.e., unbiased by seasonal differences between regions. Observations of Chla with zero concentration (n = 47) were replaced by 0.1 μg L1, an adequately small concentration allowing for log-transformation. The model 1 accounted for the heterogeneity in time and space of the monitoring data by separating variations into spatial (stationi) and temporal (yearj and monthk) components. Comparable yearly estimates were computed for each ecosystem by back-transforming (i.e., using the exponential function) the marginal means of the factor yearj in model 1 to represent the geometric mean of all stations in each coastal ecosystem over all months (JanuaryDecember). This resulted in 712, 726, and 740 annual mean values for TN, TP, and Chla, respectively, almost evenly distributed across ecosystems (Table 1). Weights to be used in the subsequent analyses of mean values were computed as the inverse variance of the mean estimates. Analysis of ChlaNutrient Relationships. General linear models of weighted annual means of Chla vs TN and TP were computed for all yearly means combined (“global model”), for each of the four regions specifically (“regional models”) and for individual ecosystems within each of the regions. The number of observations used in the regressions was equal to or less than the total number of means (Table 1), because there were years without matching Chla and TN or TP concentrations. The established relationships were compared to those reported from other studies in the literature.12,13,20 We present the general global and regional models as functions of TN as well as TP to illustrate the relationship to both major nutrients. The models for individual ecosystems were presented for TN only, as nitrogen is reported to be the main nutrient limiting primary production in the four studied regions, although phosphorus limitation may occur in spring and in oligohaline waters (Chesapeake Bay: 21,22; Denmark coast:23,24; Tampa Bay: 25; Dutch Wadden Sea: 26,27). However, the results may, to some extent, apply to TP as well since TN and TP are commonly correlated in coastal ecosystems. The generality of the regional relationships across time and space was examined by analyzing the residuals for differences among ecosystems within regions and time (interannual variation), because if the regional relationships were indeed generic to all coastal ecosystems within regions and constant over time there would be no systematic variations in the residuals. The residuals were analyzed for differences among ecosystems within regions by means of analysis of variance (ANOVA), and changes in the residuals over time were investigated using linear regression and Generalized Additive Model (GAM) for a linear 9123
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology
CRITICAL REVIEW
Figure 1. Relationships between annual mean Chla and annual mean TN (left panel, A and C) and TP (right panel, B and D) concentrations for all data combined (“Global model”, A and B) and for the four regions separately (“Regional models”, C and D). Solid black lines show the fitted loglog regression equations from this study, with associated statistics listed in the plots, and the “Global model” is compared to relationships (A and B) developed elsewhere (Florida coast,12 Finnish estuaries,20 global13).
and smoothed trend, respectively. The ChlaTN relationship was further spatially broken down by estimating the relationships for each individual ecosystem in the four regions (28 regression equations). Finally, we explored if time trajectories of individual ecosystems conformed to the general power relationship (linear relationship on the loglog scale). This was done by connecting the dots representing the annual mean concentrations of Chla and TN in individual ecosystems to form trajectories describing the temporal dynamics during periods of eutrophication and oligotrophication. A semiparametric GAM was employed to investigate the time trajectory of the ChlaTN relationship, using a quadratic LOESS smoother to describe and test the significance (likelihood ratio test) of departures in time from a linear parametric ChlaTN relationship, i.e., analyzing if years were randomly distributed around the linear ChlaTN relationship or systematically deviating over time. Weights for the Chla means could not be employed by the GAM method (in SAS version 9.2). For plotting the time trajectory, a smooth trend for TN was modeled using time as a nonparametric effect in GAM, and Chla predictions were obtained by scoring the semiparametric GAM with the smooth TN trend.
’ RESULTS Chla and nutrient concentrations varied broadly across systems (Figure 1) and over time. The average TN:TP molar ratio varied among regions with levels indicating combined N and P limitation to strong N limitation from 7.6 in Tampa Bay, 19.1 in the Wadden Sea, 26.9 in Danish coastal waters, to 38.8 in Chesapeake Bay (Figure 2). Log-transformed TN and TP were significantly correlated for all data combined (r = 0.58, p < 0.0001, n = 707) and for all regions separately (Chesapeake Bay: r = 0.40, p < 0.0001, n = 149; Denmark coast: r = 0.80, p < 0.0001, n = 256; Tampa Bay: r = 0.52, p < 0.0001, n = 104; Wadden Sea: r = 0.72, p < 0.0001, n = 198). Despite indications of a stronger N than P limitation (Figure 2) and the significant correlations between TN and TP, ”global” and “regional” models are presented both for TN and TP. Chla concentration was scaled as the 0.82 power of TN across ecosystems (Figure 1A). The power exponent was somewhat lower than that of previously reported relationships (Chla ∼ TN1.38 in 12; Chla ∼ TN1.13 in 20), although the Chla values predicted for individual TN concentrations were comparable to those delivered by these relationships. A weaker relationship was 9124
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology
CRITICAL REVIEW
Table 2. Analysis of Variance for Differences between Ecosystems within Each Region Taking Differences in TN Levels into Accounta region
F
p
142
4.89
<0.0001
246
23.73
<0.0001
4
100
14.00
<0.0001
7
155
29.46
<0.0001
dfecosystem
dferror
Chesapeake Bay
7
Denmark coast
10
Tampa Bay Wadden Sea a
Observations are residuals (Figure S1) from the region-specific regressions of log(Chla) vs. log(TN) (Figure 1C), weighted by the inverse variance of the annual Chla means.
Figure 2. Box and whiskers plot of the TN:TP molar ratio for the four regions compared to thresholds for N deficient (TN:TP < 20) and P deficient (TN:TP > 50) phytoplankton growth reported by Guildford and Hecky.11 Boxes show lower and upper quartiles with median (line) and mean (square) inside the box. Whiskers mark the 95% confidence intervals, crosses are the 1st and 99th percentiles, and dash symbols show minimum and maximum values.
obtained for Chla in relation to TP (lower R2) scaling as the 0.184 power of TP (Figure 1B), well below those found in previous studies (Chla ∼ TP1.17 in 12; Chla ∼ TP0.71 in 13; Chla ∼ TP1.26 in 20), consistent with indications that N is the primary limiting nutrient in the ecosystems investigated here. However, Chla concentrations ranged about 1 order of magnitude for any given TN or TP concentration (Figure 1), also comparable to previous analyses.13 Similar to this study, Hoyer et al.12 used annual means, whereas Meeuwig et al.20 used means over the productive season and Smith13 used a combined data set of both annual and seasonal means. Spatial Partitioning of the ChlaNutrient Relationship. The relationship between Chla and TN differed significantly among regions (F6,698 = 93.45, p < 0.0001), with power exponents ranging from 0.68 to 0.92 (Figure 1C). Significant differences among regions were also found for the ChlaTP relationship (F6,680 = 178.98, p < 0.0001), although the power exponents were consistently lower than for the ChlaTN relationships ranging from 0.24 to 0.65 (Figure 1D). TN was better at describing variations in Chla for Chesapeake Bay and Denmark coast, whereas TP gave a higher R2 value for the Wadden Sea and Tampa Bay. ChlaTN relationships were also variable within regions, as significant variation among individual coastal ecosystems within regions was observed for the residuals (Table 2, Figure S1) derived from regional ChlaTN relationships (Figure 1C). Further analysis of linear relationships between Chla and TN for individual coastal ecosystems showed a broad range of relationships, spanning from lack of significant relationships at some sites with very low scaling exponent (e.g., 0.28 for Århus Bay, Denmark, p = 0.5766) to very high scaling exponent (e.g., 1.30 for Køge Bay, Denmark, p = 0.0197) within any one region (Table 3, Figure 3). Indeed, over half of the ChlaTN relationships describing interannual changes at individual coastal areas were not statistically significant, even though some of these coastal areas experienced large changes in TN concentrations (Table 3, Figure 3). Whereas only 11 out of 28 ecosystems displayed a significant linear relationship
between Chla and TN (Table 3), this number is still significantly higher than the 1 or 2 significant relationships that would be expected by chance alone in the absence of an overall relationship between Chla and TN. Temporal ChlaTN Trajectories. The large fraction of coastal ecosystems experiencing no significant relationship between Chla and TN despite large changes in TN suggests that the ChlaTN trajectories of individual sites are complex and more convoluted than expected from the assumption of a simple power scaling of Chla to TN. Indeed this was confirmed by visual inspection of the trajectories of individual ecosystems (Figure 4) and, more formally, by statistical tests (using GAM) of the departures from a power relationship. Actually, 18 out of 28 of the coastal areas investigated (64%) had significant time departures from a simple power scaling of Chla to TN (Table 3). Most (17 out of 28) of the trajectories for coastal ecosystems deviated from a simple and monotonic increase or decrease in Chla with increasing or decreasing TN predicted from power relationships (Figure 4, Figure S2). The smoothed trajectories of the individual coastal ecosystems generally showed an initial decline in Chla with decreasing TN followed by stabilization or even an increase with further reductions in TN (Figure 4). In fact, 43% (12 out of 28) of the trajectories ended with a higher Chla concentration and more than half of these were from Chesapeake Bay, where decreases in TN were generally smaller (Figure 4, Figure S2). This finding is in sharp contrast with the expected simple response (according to the global and regional models) in Chla concentration with decreasing TN concentration, which was observed only for 29% of the coastal areas investigated (8 of 28 areas, Table 3, decreasing linear slopes in Figure 4). The examination of these trajectories (Figure 4) suggests that the yield of Chla for any given TN concentration has increased in coastal ecosystems since the onset of nutrient abatement in the late 1980s in the majority of the watersheds.9,10,22,25 A shift in the yield of Chla for any given TN should be reflected in an increase in the residuals of power Chla to TN relationships. Indeed, an analysis of residuals from the regional Chla vs TN power relationships showed that the residuals were not randomly distributed over time but rather shifted significantly from the mid 1970s to present (Figure 5). In particular, there was a significant linear trend toward increasing residuals for the Chesapeake Bay, Danish coastal areas, and the Dutch Wadden Sea, whereas Tampa Bay showed an overall decreasing trend driven by large positive residuals in the first 3 years with data (Figure 5C). Closer inspection, using a nonparametric regression model (GAM), revealed that (1) residuals increased from 1977 to the mid 1980s in Danish coastal areas and the Wadden Sea, (2) residuals declined between the mid 1980s and the early 1990s in all regions investigated, and 9125
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology
CRITICAL REVIEW
Table 3. Regressions between log(Chla) and log(TN) for Different Coastal Areas and Trajectory Analyses for Temporal Departures from Linear Relationshipa linear regression log(Chla)log(TN) region
area
trajectory analysis
short name
no. of years
relationship
RMSE
p
RMSE
χ2
p
Chesapeake Bay
Choptank
CHO
23
1.44 + 0.87x
0.24
0.0034
0.17
19.31
<0.0001
Chesapeake Bay
James River
JMS
23
1.37 + 0.12x
0.22
0.6646
0.18
7.61
0.0236
Chesapeake Bay Chesapeake Bay
Mainstem mesohaline Mainstem polyhaline
MH PH
18 18
0.93 + 0.79x 2.36 + 1.22x
0.14 0.22
0.0303 0.0409
0.13 0.19
6.46 8.87
0.0412 0.0124
Chesapeake Bay
Patuxent
PAX
21
0.30 + 0.54x
0.25
0.2233
0.21
7.50
0.0237
Chesapeake Bay
Potomac
POT
23
0.16 + 0.50x
0.19
0.0484
0.17
8.21
0.0175
Chesapeake Bay
Rappahannock
RPP
23
1.19 + 0.23x
0.26
0.5519
0.17
15.54
0.0005
Denmark coast
Flensborg Fjord
FLF
21
0.46 + 0.22x
0.17
0.2019
0.14
11.98
0.0025
Denmark coast
Horsens Fjord
HOF
26
1.85 + 0.17x
0.24
0.4580
0.31
3.24
0.1920
Denmark coast
Køge Bay
KØB
21
3.71 + 1.30x
0.15
0.0002
0.13
7.88
0.0197
Denmark coast Denmark coast
Limfjorden Nissum Bredning
LIM NIB
25 25
0.84 + 0.61x 0.62 + 0.20x
0.28 0.24
0.0374 0.2645
0.24 0.26
9.93 5.00
0.0067 0.0802
Denmark coast
Odense Fjord
ODF
30
0.85 + 0.12x
0.20
0.4680
0.26
21.57
<0.0001
Denmark coast
Roskilde Fjord
ROF
29
0.11 + 0.37x
0.17
0.0679
0.23
1.01
0.5893
Denmark coast
Skive Fjord
SKF
25
2.67 + 1.09x
0.20
<0.0001
0.20
17.21
0.0002
Denmark coast
Sydfynske Øhav
SFØ
29
0.78 + 0.03x
0.21
0.9148
0.16
22.36
<0.0001
Denmark coast
Århus Bay
ÅRB
25
1.61 + 0.28x
0.23
0.5766
0.22
5.76
0.0554
Tampa Bay
Hillsborough Bay
HB
26
0.50 + 0.73x
0.25
0.0046
0.19
21.81
<0.0001
Tampa Bay Tampa Bay
Lower Tampa Bay Middle Tampa Bay
LTB MTB
26 26
0.43 + 0.24x 0.25 + 0.56x
0.23 0.28
0.1956 0.0387
0.20 0.19
11.30 35.61
0.0038 <0.0001
Tampa Bay
Old Tampa Bay
OTB
26
1.74 + 0.08x
0.20
0.6882
0.15
23.56
<0.0001
Wadden Sea
Dantziggat
DG
26
0.10 + 0.56x
0.27
0.0880
0.24
13.43
0.0012
Wadden Sea
Doove Balg
DB
25
1.19 + 0.25x
0.29
0.4630
0.30
3.16
0.2063
Wadden Sea
Huibertgat
HG
29
1.05 + 0.64x
0.18
<0.0000
0.19
1.93
0.3872
Wadden Sea
Lauwers
LW
29
0.03 + 0.52x
0.23
0.0875
0.24
0.71
0.7062
Wadden Sea
Marsdiep
MD
30
0.52 + 0.31x
0.18
0.0583
0.17
2.90
0.2276
Wadden Sea Wadden Sea
Vliestroom Zoutkamperlaag
VS ZK
29 29
0.78 + 0.29x 0.44 + 0.64x
0.20 0.26
0.1102 0.0486
0.19 0.22
3.84 5.98
0.1499 0.0515
a
Linear relationship, residual mean square error (RMSE, weighted) and significance level (p) refer to the linear regression of log(Chla) vs. log(TN). Temporal departures from the linear relationship were investigated by means of GAM; RMSE was calculated as the deviance per observation (unweighted), χ2 and p derive from likelihood ratio test statistics. Number of observations is equal to the number of years. Significant relationships (p < 0.05) are highlighted in bold.
(3) residuals increased during the past decade in all four regions (GAM models in Figure 5). The remarkable and, to some extent, consistent trends toward shifting Chla residuals across regions corresponds to an increase in the intercept of the power relationship between Chla and TN over time, implying that these relationships may systematically underestimate Chla expected for a specific reduction in TN into the future. An analysis of the Chla:TN ratio across systems showed similar trends, as the mean ratio has increased consistently since the mid 1970s to present for three of the four regions (Figure 5). The data series revealed a 2-fold range in the Chla:TN ratios and suggests nearly a doubling of the average Chla per unit of TN over this 30-year period. All four regions have experienced decreasing concentrations of total phosphorus (TP) during the study period due to point source abatement. The trends in the residuals of the ChlaTN relationship over time were correlated to the decreasing trends in TP taking spatial variations between areas into account (Table 4), albeit negatively for all regions except Tampa Bay. However, Tampa Bay, where the watershed is highly urbanized and contains large
deposits of phosphate rock,25 had the highest TP concentrations and the lowest TN:TP ratios of all four regions (Figure 2). Thus, the increasing trend in the residuals could not be causally linked to patterns of potential phosphorus limitation.
’ DISCUSSION Widespread evidence of coastal eutrophication starting in the 1970s led scientists to develop predictive frameworks to guide nutrient abatement plans, building on early experiences from freshwater management.14,15 Our mechanistic understanding of eutrophication processes has evolved in parallel from simple nutrient-driven relationship to complex models involving various trophic levels and ecosystem-specific features.8 Yet, the belief that phytoplankton biomass should decline with decreasing nutrient inputs, and thus concentrations, in a manner similar to the way it increased remains a pervasive principle in the management of coastal ecosystem (e.g., 12). The results presented challenge this belief, with important implications for nutrient management and the evaluation of its effectiveness. 9126
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology
Figure 3. Linear regressions between log(Chla) and log(TN) for individual ecosystems (n = 28) in four regions. Annual means of Chla and TN used for the regressions are shown by gray dots.
Figure 4. Smoothed time trajectories, obtained by means of GAM, of the log(Chla) vs log(TN) relationships for the 28 ecosystems representing the four regions. Open and filled symbols mark the start and end of the trajectories, and the annual means used for computing the trajectories are shown by gray dots. Linear relationships are shown for ecosystems (n = 10), where time departures from a stationary power relationship between Chla and TN was not significant (Table 3). A close-up of the trajectories for the different regions can be found in Figure S2.
Nutrient inputs from land and atmosphere can be controlled by appropriate management measures, but nutrient inputs from sediments28 and exchanges over the open boundary29 can also be considerable. Except for systems with strong light limitation increasing inputs of nutrients enhances phytoplankton growth and consequently biomass,8,30 but total inputs from all nutrient sources are difficult to estimate, resulting in poor relationships between nutrient loading from land and phytoplankton production.31
CRITICAL REVIEW
Total nutrient concentrations, representing a balance between inputs and losses, constitute good general predictors for phytoplankton biomass31 and can be linked to nutrient input from land through system-specific relations (e.g., 32). Therefore, a general relationship describing the effect of nutrient enrichment on phytoplankton biomass would be expected for total nutrient concentrations. Should we Expect a Simple, General ChlaNutrient Relationship? Recognition of the prevalent role of nutrients, particularly nitrogen, in limiting phytoplankton abundance in marine ecosystems3335 drove efforts to derive a general relationship allowing the prediction of chlorophyll a from nitrogen concentrations. Power functions of various kinds were used to describe the relationship between Chla and nutrients as such models are quite flexible and easy to estimate by linear regression on a loglog scale. Chlanutrient relationships are expected to eventually flatten out because of density-dependent processes, such as self-shading and allelopathy.36 This implies a slope <1 on the loglog scale as found in the global and regional models presented here (Figure 1), as well as in models presented by Nixon et al.,37 who reported a scaling exponent of 0.72. However, some studies reported slopes >1,12,13,20 suggesting that other mechanisms enhancing Chla covaried positively with nutrient levels in the ecosystems studied. Carstensen and Henriksen32 found that the slope decreased from 0.92 to 0.53 when a systemspecific intercept was included in the relationship, also indicating a positive covariation between TN levels and other mechanisms favoring high Chla concentrations. Our results, combined with those from Carstensen and Henriksen32 and Nixon et al.,37 suggest that Chla should scale to TN with a scaling exponent between 0.5 and 1. Nitrogen is generally believed to be the main limiting nutrient in coastal ecosystems although some systems may also display phosphorus limitation34 and the debate of nitrogen versus phosphorus limitation is still ongoing.35 In our study Chla correlated with both TN and TP, which were intercorrelated, but TN was a better predictor of Chla for Chesapeake Bay and Denmark coast, while TP was a better predictor for Tampa Bay and Wadden Sea (Figure 1C and D), although both of these regions have TN:TP ratios consistent with nitrogen limitation (Figure 2). TP levels in Meeuwig et al.20 and Hoyer et al.12 were generally low compared to TP values in our study, despite similar ranges for Chla, suggesting that phosphorus limitation could be more pronounced in the systems they studied. Our study confirms the reported predominant nitrogen limitation of the regions examined here, consistent with previous reports for these regions.22,23,25,27 Although general relationships between Chla and nutrients were indeed derived, e.g., by Smith13 and here (Figure 1), these were all based on data pooled from many different ecosystems covering several orders of magnitude for both Chla and nutrients. Thus, in principle, these relationships could be driven by differences among ecosystems rather than by a common response pattern. There is overwhelming evidence that system-specific attributes modulate the response of phytoplankton to nutrient enrichment.8 Monbet38 documented significant differences in Chla responses (almost by a factor of 10) to nitrogen levels between micro- and macrotidal estuaries, because increased tidal mixing enhances light limitation by increased vertical mixing and resuspension of sediments. Moreover, increased vertical mixing also enhances grazing pressure on phytoplankton by benthic grazers. Petersen et al.39 reported a 56 fold decrease in Chla, whereas nutrient 9127
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology
CRITICAL REVIEW
Figure 5. Trends in residuals (annual means with 95% confidence interval of the mean value) from region-specific regressions (Figure 1C) between log(Chla) and log(TN) investigated by linear regression and a nonparametric GAM with statistics listed. For comparison the ratio between annual mean Chla and TN is also shown.
Table 4. Analysis of Variance from Analyzing Residuals from Region-Specific Relationships (log(Chla)log(TN) Relationship in Figure 1C) vs. log(TP) with an Intercept Specific to the Coastal Ecosystem, i.e., Residual = αecosystem + β log(TP)a TP-dependent slope ecosystem-specific intercept region
β
F
p
dfecosystem
F
p
Chesapeake Bay 141 0.30
4.13 0.0440
7
5.36 <0.0001
Denmark coast 245 0.07
2.32 0.1289
10
24.07 <0.0001
99 0.24 13.46 0.0004 184 0.21 10.45 0.0015
4 7
17.42 <0.0001 31.14 <0.0001
Tampa Bay Wadden Sea a
dferror
Residuals were weighted by the inverse variance of the annual Chla means.
levels changed much less, following a regime shift from a bottom-up to a top-down controlled coastal ecosystem. Very shallow coastal ecosystems dominated by macrophytes also respond to changes in nutrient inputs differently from phytoplanktondominated ones.37 Differences in the labile fraction of N and P would similarly lead to deviations from a general relationship based on total nutrients32 as would differences in the degree of N versus P limitation. Moreover, latitudinal and temperature differences affect the turnover of nutrients and the annual mean Chla for a given nutrient concentration. Evidently, many factors besides nutrient enrichment could lead to changes in Chla and thus account for the order-of-magnitude variability around the general relationship between Chla and nutrients. Hence, nutrient concentrations set an order-of-magnitude range for the annual average Chla concentration for coastal ecosystems, but do not
allow prediction of Chla trajectories in response to changes in nutrient concentrations within these boundaries. Are Spatial Differences Important? The results presented here clearly demonstrate that the general relationship between Chla and TN conceals important diversity in the nature of the relationship and the underlying response of Chla to changing nitrogen concentrations in individual coastal areas. Chesapeake Bay had the highest Chla relative to TN, followed by Tampa Bay, Wadden Sea, and the Denmark coast. Several ecosystem features could potentially account for these regional differences in the ChlaTN relationships. All the studied regions are classified as microtidal but the tidal range in the Wadden Sea (∼ 1.52 m) is higher than for the other regions (<1 m).38 Tampa Bay is a relatively shallow (∼ 4 m) and wind-exposed ecosystem where average Secchi depths range from 1 m in Hillsborough Bay to 2.53 m in Lower Tampa Bay,25 suggesting potential light limitation. High abundances of filter feeders characterize the Danish coastal sites that are mostly well-mixed to the bottom,23 and a relatively large fraction of TN is refractory due to mixing with Baltic Sea water.32 Additional features could be added as candidates to account for the reported regional differences in ChlaTN relationships. Whereas the general relationships help delineate the Chla concentration expected for a particular TN concentration, this only applies to the mean Chla to be expected for a universe of coastal ecosystems with similar TN concentrations, but not to any one individual ecosystem in particular. The key consideration in using this general relationship is that the error in the prediction of Chla can only be assumed to be randomly distributed within the broad order-of-magnitude error bounds about the predicted Chla for such hypothetical population of coastal ecosystems with similar TN. Yet, this does not hold for any one 9128
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology individual coastal ecosystem. Hence, the general relationship is not only invalid to predict the particular value of Chla to be expected for a particular change, increase or decrease, in TN for a particular coastal ecosystem, but is also invalid to predict even the ensuing direction of change (increasing, decreasing, or no change). Indeed, individual systems showed the full range of TNChla relationships, including positive, negative ones, and no relationship, stressing the need to develop system-specific predictions. Do ChlaTN Relationships Change over Time? The order of magnitude error characteristic of the general ChlaTN relationship includes differences in the Chla level at a particular TN concentration among individual ecosystems, but also includes dynamic changes in the yield of Chla for a particular TN concentration over time. More than half of the 28 ecosystems had significant time departures from a static power relationship between Chla and TN (Table 3), implying that the residuals of the relationship were not randomly distributed over time. Indeed, the stabilization or increase of Chla over time despite decreasing TN concentrations observed in most ecosystems (Figure 4) is likely to be the main reason for the low number of significant linear relationships between Chla and TN across the ecosystems studied (39% in Table 3). The complex trajectories of Chla concentration with changes in TN concentration resemble those reported by Duarte et al.,10 but these complex trajectories do not only result from idiosyncratic responses to nutrient reduction but from shifts over time in the yield of Chla for any one TN concentration. The shift in Chla for any given TN concentration over time (assessed by the time trend in the residuals from the ChlaTN relationship as well as the Chla:TN ratio, Figure 5) could not be explained from corresponding changes in TP with time. If the functional relationship between Chla and TN flattens out with a diminishing Chla yield per unit nitrogen as TN increases, as implied by a power relationship Chla = a TNb where b < 1, then an increase in the Chla:TN ratio (Chla:TN = a TNb‑1 and b < 1) is expected with decreasing TN concentrations over time (Figure 6). However, the estimated power relationships for the 4 regions (Figure 1C) could only account for increases in the Chla: TN ratio of <11% for Chesapeake Bay, < 32% for Denmark coast, < 5% for Tampa Bay, and <6% for Wadden Sea (Figure 6), much lower than the observed increases in the Chla:TN (Figure 5). Even power relationships with a lower exponent (e.g., b = 0.5) or no relationship at all (b = 0) could not fully account for the almost doubling of the Chla:TN ratio over time observed here (Figure 6). Thus, the trends in Chla:TN cannot be explained as an artifact from overestimating the exponent in the ChlaTN relationship, and reflect, therefore, a functional change in the ecosystem, consistent within but also among regions. The increase in the yield of Chla for any given TN concentration over time cannot be accounted for by system-specific attributes that did not change over time, such as the tidal mixing regime.38 Hence, the actual yield for any TN concentration depends on other factors, including the taxonomic composition of the community, the turnover of nitrogen in water and sediments, the partitioning of nitrogen in pools of different availability to phytoplankton, and the role of other limiting factors, including other nutrients and light availability, loss factors, including grazers and diseases, and stresses (e.g., high temperature or high UV radiation). Does the Increase in Chla Yield Per Unit Nitrogen Signal a Global Change? A higher-order process, acting across large scales, is required to account for the almost consistent shift in the yield of Chla per unit TN across regions. Indeed, the baselines
CRITICAL REVIEW
Figure 6. Predicted increases in the Chla:TN ratio for declining TN levels (equation inserted) computed from the estimated power relationships for the 4 regions as well as for a square root relationship (b = 0.5) and no relationship (b = 0). Thick lines show estimated reduction ranges in TN during the study period for the different ecosystems (Chesapeake Bay: 17% to 29%, Denmark coast: 957%, Tampa Bay: 2344%, Wadden Sea: 1645%).
determining the structure and function of coastal ecosystems are shifting rapidly, and this may affect the response of phytoplankton to nutrient concentrations in multiple ways.10 Some possible explanations for this shift are described in the following. Increasing CO2 concentrations and water temperature, and decreasing pH may alter phytoplankton communities directly at many levels, from physiological processes to nutrient requirements and community structure. For instance, experimental and comparative analyses (e.g., 4042) show that warmer temperatures lead to reduced mean cell size and a shift from diatoms to cyanobacteria,43 which could reduce the removal of microalgal biomass by grazers under a warmer climate (e.g., 42). Increasing genus richness stimulates the resource use efficiency of the phytoplankton community.44 Large-scale changes in the food web structure and function of coastal ecosystems derived from overfishing and excess harvest of filter feeders, both global phenomena, erode ecosystem buffers and triggers increased phytoplankton biomass enhancing the vulnerability of coastal ecosystems to eutrophication (e.g., 4547). For example, Casini et al.48 showed that cod overfishing in the Baltic Sea led to decreases in zooplankton through trophic cascades, alleviating the grazing pressure on phytoplankton. Eutrophication and overfishing may lead to proliferation of jellyfish altering the food-web through increasing predation on protists and thus reducing the pelagic grazing pressure on phytoplankton.49 The parallel trend toward increase in Chla yield per unit nitrogen over the past decade in all regions examined here could be the result of the major shift in the baselines for the functioning of coastal ecosystems resulting from the combined effects of climate change, overfishing, and, possibly, other components of global change. Whereas the trajectories of individual ecosystems in ChlaTN plots appear idiosyncratic, the coherent trend in the residuals and the general increase in the yield of Chla per unit nitrogen over time indicate that there is a common, underlying 9129
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology shift in ecosystem baselines affecting the behavior of individual ecosystems, rendering the relationship between Chla and TN dynamic over time. Ecosystem Management with Shifting Baselines. The departure of individual ecosystems from the general relationship between Chla and TN derived from global comparative analyses calls for a reconsideration of the predictive tools underpinning management actions. Trends in the yield of Chla per unit nitrogen over time results in a bias in the prediction of Chla concentration as the intercept of the relationship between Chla and TN concentration shifts over time. Hence, regression equations produced during the eutrophication phase in the late 1980s generally underestimated the Chla concentration corresponding to more recent and lower TN concentrations obtained during the oligotrophication phase. Moreover, the apparent idiosyncrasy of Chla vs TN relationships among systems, reflected in a range of slopes and intercepts of Chla vs TN regressions among coastal areas (Table 3, Figure 4), could partially result from shifting Chla to TN ratios over time. The shift in the functional relationship between Chla and TN over time demonstrated here helps explain the reported failure to revert eutrophied coastal ecosystems to their previous state following reduction of nutrient inputs.10 This calls for a reconsideration of our conceptual model of eutrophication to face the complexity of the operating factors, differences between coastal ecosystems,8 regime shifts, and shifting baselines.10 Persistent eutrophication may sustain elevated Chla through feedback mechanisms significantly altering ecosystem functions, potentially leading to hysteresis in the pressureresponse relationship.50,51 Feedbacks associated with eutrophication-enhanced hypoxia could lead to hysteresis through disrupting benthic food webs, alleviating benthic grazing pressure on phytoplankton, and stimulating nitrogen-fixing cyanobacteria blooms through the release of sediment phosphate.5254 In fact, Chesapeake Bay and the Danish coastal sites have experienced seasonal hypoxia of increasing extent throughout the study period22,23 that could explain changes in the Chla:TN ratios for these two regions. Feedback mechanisms enhanced through persistent eutrophication may be potentially reversible over time and must be considered together with shifting baselines in predicting Chla responses to reduced nutrient inputs. Whereas the lack of reduction in Chla concentrations following nutrient reductions demonstrated here may be disturbing to managers, our analysis contains important lessons that can help improve the design and assessment of managerial actions. The effectiveness of responses in Chla concentration following nutrient reduction plans should not be assessed relative to the Chla concentration at the time nutrient reduction was initiated but relative to the Chla concentration the ecosystem would support if nutrients had not been controlled. This rationale is illustrated in Figure 7, which considers what would be the trajectory of a hypothetical coastal area under two scenarios: (1) a “do nothing” scenario, where nutrient concentrations do not change over time and where the ecosystem will exhibit a trajectory that departs from the general regression line to occupy a position of increased Chla due to the increase in the yield of chlorophyll per unit nitrogen over time, and (2) a “nutrient reduction” scenario, where the ecosystem will follow a trajectory that leads to an increase in Chla relative to the initial state, due to the increase in the yield of chlorophyll per unit nitrogen, but that in fact represents a reduction in the realized Chla concentration relative to the “do nothing” scenario (Figure 7). The key message underlying this conceptual model is that shifting baselines imply
CRITICAL REVIEW
Figure 7. Conceptual model demonstrating the implications of shifts in the yield of Chlorophyll a per unit total nitrogen in coastal ecosystems for the evaluation of the outcome of managerial actions to reduce nitrogen concentrations. The possible outcomes of two alternative strategies, “do nothing” and “nutrient reduction”, are shown.
that the present state of the system is not an adequate reference to evaluate the effectiveness of nutrient reduction plans, as the future status of the ecosystem would differ from that at present under a “do nothing” scenario. Moreover, the confidence of Chla predictions for nutrient reduction scenarios will weaken, as the uncertainty of predicting the shifting baselines has to be included as well. Trajectories of recovery are even more complex in the potential presence of hysteresis responses,10,55 and declining Chla could result from small changes in nutrient inputs if a proper functioning of the coastal ecosystem is reestablished. Thus, despite observed increases in Chla concentrations it is still important to stress that nutrient reductions do release pressure on the ecosystem and improve conditions relative to what these would have been under a “do nothing” scenario. Provided the importance of changing baselines for the setting and evaluation of actions to reverse eutrophication, it is fundamental that our understanding of the causes of such shift in baselines improves to allow forecasting the trajectories of individual coastal ecosystems. A better understanding of the dynamics of coastal ecosystems forced by both changes in nutrient inputs, derived from factors operating at the basin scale, and shifting baselines derived from forces operating at various scales is fundamental to achieve this goal. This requires a research agenda that faces the complex interactions, operating at multiple levels, controlling the variability in the yield of chlorophyll per unit nitrogen over time. The mechanisms leading to shifts in the yield of chlorophyll per unit nitrogen must be investigated under a range of scales, from physiological experiments, to investigate the responses to increasing CO2 and temperature, to mesocosm scales where effects from changes in food webs can be investigated. These experiments should be supplemented by modeling efforts, specifically addressing shifting baselines and regime shifts (e.g., 56), to synthesize experimental results at the ecosystem scale and improve predictions of nutrient management. This research agenda must also include research on trajectories at the ecosystem level involving the investigation of replicated experimental coastal ecosystems, subject to deliberate manipulation of nutrients, over decadal time scales such as those included in this study. This approach is comparable to long-term whole-ecosystem experiments conducted at the Experimental Lakes Area in 9130
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology Canada57 or the Hubbard Brook watersheds,58 but should involve longer time scales and improved replication. Only these largescale experiments, conducted at the ecosystem level and sustained over long time scales, hold the power, when supported by mechanistic understanding, to resolve the forces driving complex trajectories of coastal ecosystems through time necessary to gain the capacity to forecast them with sufficient accuracy, a precondition for effective management. The results presented here confirm that the expectation that the response of coastal ecosystems to increasing or decreasing nutrient concentrations can be predicted from a single general relationship is unsupported. In particular, the results presented provide evidence for idiosyncratic ChlaTN relationships for individual ecosystems and suggest that much of the variability in the trajectories and ChlaTN relationships among individual systems derives from consistent shifts in the yield of chlorophyll per unit nitrogen. These shifts, derived from large-scale forcing likely associated with global change, imply that future Chla concentration in coastal areas cannot be predicted from current Chlanutrient relationships. These results indicate that ecological sciences must progress to face uncertainty and shifting baselines and be able to operate with relative, rather than absolute, predictions and targets. The challenge of our times governed by global change rests in that, as the French poet Paul Valery put it, “the future is no longer what it used to be”.
’ ASSOCIATED CONTENT
bS
Supporting Information. Figures S1 and S2. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +45 46301345; fax: +45 46301114; e-mail:
[email protected].
’ ACKNOWLEDGMENT This research is a contribution to the Thresholds Integrated Project (contract FP6-003933-2) and WISER (contract FP7226273), funded by the European Commission. D.K.-J. was also funded by the Danish Agency for Science, Technology and Innovation. We thank Jens W€urgler Hansen, Timo Tamminen, Holly Greening, Jim Cloern, and Scott Nixon for comments on earlier versions of the paper as well as Emilio Agustí for encouragement, challenging insights, and discussion. ’ REFERENCES (1) Nixon, S. W. Coastal marine eutrophication - A definition, social causes, and future concerns. Ophelia 1995, 41, 199–219. (2) Vitousek, P. M.; Mooney, H. A.; Lubchenco, J.; Melillo, J. M. Human domination of Earth’s ecosystems. Science 1997, 277, 494–499. (3) Vidal, M.; Duarte, C. M.; Sanchez, M. C. Coastal eutrophication research in Europe: Progress and imbalances. Mar. Pollut. Bull. 1999, 38, 851–854. (4) Boesch, D. F. Challenges and opportunities for science in reducing nutrient over-enrichment of coastal ecosystems. Estuaries 2002, 25, 886–900. (5) Duarte, C. M. Submerged aquatic vegetation in relation to different nutrient regimes. Ophelia 1995, 41, 87–112. (6) Costanza, R.; d’Arge, R.; de Groot, R.; Farber, S.; Grasso, M.; Hannon, B.; Limburg, K.; Naeem, S.; O’Neill, R. V.; Paruelo, J.; Raskin,
CRITICAL REVIEW
R. G.; Sutton, P.; van den Belt, M. The value of the world’s ecosystem services and natural capital. Nature 1997, 387, 253–260. (7) Nixon, S. W. Eutrophication and the macroscope. Hydrobiologia 2009, 629, 5–19. (8) Cloern, J. E. Our evolving conceptual model of the coastal eutrophication problem. Mar. Ecol.: Prog. Ser. 2001, 210, 223–253. (9) Carstensen, J.; Conley, D. J.; Andersen, J. H.; Aertebjerg, G. Coastal eutrophication and trend reversal: A Danish case study. Limnol. Oceanogr. 2006, 51, 398–408. (10) Duarte, C. M.; Conley, D. J.; Carstensen, J.; Sanchez-Camacho, M. Return to Neverland: Shifting baselines affect eutrophication restoration targets. Estuaries Coasts 2009, 32, 29–36. (11) Guildford, S. J.; Hecky, R. E. Total nitrogen, total phosphorus, and nutrient limitation in lakes and oceans: Is there a common relationship? Limnol Oceanogr. 2000, 45, 1213–1223. (12) Hoyer, M. V.; Frazer, T. K.; Notestein, S. K.; Canfield, D. E., Jr. Nutrient, chlorophyll, and water clarity relationships in Florida’s nearshore coastal waters with comparisons to freshwater lakes. Can. J. Fish. Aquat. Sci. 2002, 59, 1024–1031. (13) Smith, V. Responses of estuarine and coastal marine phytoplankton to nitrogen and phosphorus enrichment. Limnol. Oceanogr. 2006, 51, 377–384. (14) Sakamoto, M. Primary production by phytoplankton community in some Japanese lakes and its dependence on lake depth. Arch. Hydrobiol. 1966, 62, 1–28. (15) Vollenweider, R. A. Scientific Fundamentals of the Eutrophication of Lakes and Flowing Waters with Particular Reference to Nitrogen and Phosphorus As Factors of Eutrophication; Organisation for Economic Cooperation and Development Technical Report DA S/SCI/68.27.250; OECD: Paris, 1968. (16) Dillon, P. J.; Rigler, F. H. The phosphoruschlorophyll relationship in lakes. Limnol. Oceanogr. 1974, 19, 767–773. (17) Nixon, S. W.; Pilson, M. E. Q.; Oviatt, C. A.; Donoghay, P.; Sullivan, B.; Seitzinger, S.; Rudnick, D.; Frithsen, J. Eutrophication of a coastal marine ecosystem - An experimental study using MERL microsoms. In Flows of Energy and Materials in Marine Ecosystems; Fasham, M. J. R., Ed.; Theory and Practice; Plenum Press: New York, 1984; pp 105135. (18) Nixon, S. W.; Oviatt, C. A.; Frithsen, J.; Sullivan, B. Nutrients and the productivity of estuarine and coastal marine ecosystems. J. Limnol. Soc. South Africa. 1986, 12, 43–71. (19) Olsen, Y.; Agustí, S.; Andersen, T.; Duarte, C. M.; Gasol, J. M.; Gismervik, I.; Heiskanen, A.-S.; Hoell, E.; Kuuppo, R.; Lignell, R.; Reinertsen, H.; Sommer, U.; Stibor, H.; Tamminen, T.; Vadstein, O.; Vaque, D.; Vidal, M. A comparative study of responses in plankton food web structure and function in contrasting European coastal waters exposed to experimental nutrient addition. Limnol. Oceanogr. 2006, 51, 488–503. (20) Meeuwig, J. J.; Kauppila, P.; Pitk€anen, H. Predicting coastal eutrophication in the Baltic: A limnological approach. Can. J. Fish. Aquat. Sci. 2000, 57, 844–855. (21) Fisher, T. R.; Gustafsson, G. A.; Sellner, K.; Lacouture, R.; Haas, L. W.; Wetzel, R. L.; Magnien, R.; Everitt, D.; Michaels, B.; Karrh, R. Spatial and temporal variation of resource limitation in Chesapeake Bay. Mar. Biol. 1999, 133, 763–778. (22) Kemp, W. M.; Boynton, W. R.; Adolf, J. E.; Boesch, D. F.; Boicourt, W. C.; Brush, G.; Cornwell, J. C.; Fisher, T. R.; Glibert, P. M.; Hagy, J. D.; Harding, L. W.; Houde, E. D.; Kimmel, D. G.; Miller, W. D.; Newell, R. I. E.; Roman, M. R.; Smith, E. M.; Stevenson, J. C. Eutrophication of Chesapeake Bay: Historical trends and ecological interactions. Mar. Ecol.: Prog. Ser. 2005, 303, 1–29. (23) Conley, D. J.; Kaas, H.; Møhlenberg, F.; Rasmussen, B.; Windolf, J. Characteristics of Danish estuaries. Estuaries 2000, 23, 820–837. (24) Carstensen, J.; Henriksen, P.; Heiskanen, A.-S. Summer algal blooms in shallow estuaries: Definition, mechanisms, and link to eutrophication. Limnol. Oceanogr. 2007, 52, 370–384. (25) Greening, H.; Janicki, A. Toward reversal of eutrophic conditions in a subtropical estuary: water quality and seagrass response to 9131
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
Environmental Science & Technology nitrogen loading reductions in Tampa Bay, USA. Environ. Manage. 2006, 38, 163–178. (26) Philippart, C. J. M.; Cadee, G. C.; van Raaphorst, W.; Riegman, R. Long-term phytoplanktonnutrient interactions in a shallow coastal sea: Algal community structure, nutrient budgets, and denitrification potential. Limnol. Oceanogr. 2000, 45, 131–144. (27) Colijn, F.; Cadee, G. C. Is phytoplankton growth in the Wadden Sea light or nitrogen limited? J Sea Res. 2003, 49, 83–93. (28) Jørgensen, B. B. Material flux in the sediment. In Eutrophication in Coastal Marine Ecosystems; Jørgensen, B. B., Richardson, K., Eds.; American Geophysical Union: Washington, DC, 1996; pp 115135. (29) Mackas, D. L.; Harrison, P. J. Nitrogenous nutrient sources and sinks in the Juan de Fuca Strait/Strait of Georgia/Puget Sound estuarine system: Assessing the potential for eutrophication. Estuar. Coast Shelf. Sci. 1997, 44, 1–21. (30) Carstensen, J.; Conley, D. J.; M€uller-Karulis, B. Spatial and temporal resolution of carbon fluxes in a shallow coastal ecosystem. Mar. Ecol.: Prog. Ser. 2003, 252, 35–50. (31) Borum, J. Shallow waters and land/sea boundaries. In Eutrophication in Coastal Marine Ecosystems; Jørgensen, B. B., Richardson, K., Eds.; American Geophysical Union: Washington, DC, 1996; pp 179203. (32) Carstensen, J.; Henriksen, P. Phytoplankton biomass response to nitrogen inputs: a method for WFD boundary setting applied to Danish coastal waters. Hydrobiologia 2009, 633, 137–149. (33) Downing, J. A. Marine nitrogen:phosphorus stoichiometry and the global N:P cycle. Biogeochemistry 1997, 37, 237–252. (34) Howarth, R. W.; Marino, R. Nitrogen as the limiting nutrient for eutrophication in coastal marine ecosystems: Evolving views over 3 decades. Limnol. Oceanogr. 2006, 51, 364–376. (35) Conley, D. J.; Paerl, H. W.; Howarth, R. W.; Boesch, D. F.; Seitzinger, S. P.; Havens, K. E.; Lancelot, C.; Likens, G. E. Controlling Eutrophication: Nitrogen and Phosphorus. Science 2009, 323, 1014–1015. (36) Prairie, Y.-T.; Duarte, C. M.; Kalff, J. Unifying nutrient-chlorophyll relationships in lakes. Can. J. Fish. Aquat. Sci. 1989, 46, 1176–1182. (37) Nixon, S. W.; Buckley, B.; Granger, S.; Bintz, J. Responses of very shallow marine ecosystems to nutrient enrichment. Hum. Ecol. Risk. Assess. 2001, 7, 1457–1481. (38) Monbet, Y. Control of phytoplankton biomass in estuaries: A comparative analysis of microtidal and macrotidal estuaries. Estuaries 1992, 15, 563–571. (39) Petersen, J. K.; Hansen, J. W.; Laursen, M. B.; Clausen, P.; Carstensen, J.; Conley, D. J. Regime shift in a coastal marine ecosystem. Ecol. Appl. 2008, 18, 497–510. (40) Agawin, N. S. R.; Duarte, C. M.; Agustí, S. Growth and abundance of Synechococcus sp. in a Mediterranean Bay: Seasonality and relationship with temperature. Mar. Ecol.: Prog. Ser. 1998, 170, 45–53. (41) Agawin, N. S. R.; Duarte, C. M.; Agustí, S. Nutrient and temperature control of the contribution of picoplankton to phytoplankton biomass and production. Limnol. Oceanogr. 2000, 45, 591–600. (42) Sommer, U.; Lengfellner, K. Climate change and the timing, magnitude, and composition of the phytoplankton spring bloom. Global Change Biol. 2008, 14, 1199–1208. (43) Paerl, H. W.; Huisman, J. Blooms like it hot. Science 2008, 320, 57–58. (44) Ptacnik, R.; Solimini, A. G.; Andersen, T.; Tamminen, T.; Brettum, P.; Lepist€o, L.; Willen, E.; Rekolainen, S. Diversity predicts stability and resource use efficiency in natural phytoplankton communities. Proc. Natl. Acad. Sci., U.S.A. 2008, 105, 5134–5138. (45) Officer, C.; Smayda, T.; Mann, R. Benthic filter feeding: A natural eutrophication control. Mar. Ecol.: Prog. Ser. 1982, 9, 203–210. (46) Daskalov, G. M.; Grishin, A. N.; Rodionov, S.; Mihneva, V. Trophic cascades triggered by overfishing reveal possible mechanisms of ecosystem regime shifts. Proc. Natl. Acad. Sci., U.S.A. 2007, 104, 10518–10523. (47) Heck, K. L.; Valentine, J. F. The primacy of top-down effects in shallow benthic ecosystems. Estuaries Coasts 2007, 30, 371–381. (48) Casini, M.; L€ovgren, J.; Hjelm, J.; Cardinale, M.; Molinero, J.-C.; Kornilovs, G. Multi-level trophic cascades in a heavily exploited open marine ecosystem. Proc. R. Soc. B 2008, 275, 1793–1801.
CRITICAL REVIEW
(49) Richardson, A. J.; Bakun, A.; Hays, G. C.; Gibbons, M. J. The jellyfish joyride: Causes, consequences and management responses to a more gelatinous future. Trends Ecol. Evol. 2009, 24, 312–322. (50) Scheffer, M.; Carpenter, S.; Foley, J. A.; Folke, C.; Walker, B. Catastrophic shifts in ecosystems. Nature 2001, 413, 591–596. (51) Scheffer, M.; Carpenter, S. Catastrophic regime shifts in ecosystems: Linking theory to observation. Trends Ecol. Evol. 2003, 18, 648–656. (52) Vahtera, E.; Conley, D. J.; Gustafsson, B. G.; Kuosa, H.; Pitk€anen, H.; Savchuk, O. P.; Tamminen, T.; Viitasalo, M.; Voss, M.; Wasmund, N.; Wulff, F. Internal ecosystem feedbacks enhance nitrogenfixing cyanobacteria blooms and complicate management in the Baltic Sea. Ambio 2007, 36, 186–193. (53) Diaz, R. J.; Rosenberg, R. Spreading dead zones and consequences for marine ecosystems. Science 2008, 321, 926–929. (54) Conley, D. J.; Bj€orck, S.; Bonsdorff, E.; Carstensen, J.; Destouni, G.; Gustafsson, B. G.; Hietanen, S.; Kortekaas, M.; Kuosa, H.; Meier, H. E. M.; M€uller-Karulis, B.; Nordberg, K.; Norkko, A.; N€urnberg, G.; Pitk€anen, H.; Rabalais, N. N.; Rosenberg, R.; Savchuk, O. P.; Slomp, C. P.; Voss, M.; Wulff, F.; Zillen, L. Hypoxia-related processes in the Baltic Sea. Environ. Sci. Technol. 2009, 43, 3412–3420. (55) Kemp, W. M.; Testa, J. M.; Conley, D. J.; Gilbert, D.; Hagy, J. D. Temporal responses of coastal hypoxia to nutrient loading and physical controls. Biogeosciences 2009, 6, 2985–3008. (56) Liu, Y.; Evans, M A.; Scavia, D. Gulf of Mexico hypoxia: Exploring increasing sensitivity to nitrogen loads. Environ. Sci. Technol. 2010, 44, 5836–5841. (57) Schindler, D. W. Eutrophication and recovery in experimental lakes: Implications for lake management. Science 1974, 184, 897–899. (58) Likens, G. E.; Bormann, F. H.; Johnson, N. M.; Fisher, D. W.; Pierce, R. S. Effects of forest cutting and herbicide treatment on nutrient budgets in the Hubbard Brook watershed ecosystem. Ecol. Monogr. 1970, 40, 23–47.
9132
dx.doi.org/10.1021/es202351y |Environ. Sci. Technol. 2011, 45, 9122–9132
POLICY ANALYSIS pubs.acs.org/est
Life Cycle Assessment of Potential Biojet Fuel Production in the United States Datu B. Agusdinata,*,† Fu Zhao,‡ Klein Ileleji,§ and Dan DeLaurentis|| System-of-Systems Laboratory, College of Engineering, ‡School of Mechanical Engineering and Division of Environmental and Ecological Engineering, §School of Agricultural and Biological Engineering, and School of Aeronautics and Astronautics, Purdue University 701 West Stadium Avenue, West Lafayette, Indiana 47907, United States )
†
bS Supporting Information ABSTRACT: The objective of this paper is to reveal to what degree biobased jet fuels (biojet) can reduce greenhouse gas (GHG) emissions from the U.S. aviation sector. A model of the supply and demand chain of biojet involving farmers, biorefineries, airlines, and policymakers is developed by considering factors that drive the decisions of actors (i.e., decision-makers and stakeholders) in the life cycle stages. Two kinds of feedstock are considered: oil-producing feedstock (i.e., camelina and algae) and lignocellulosic biomass (i.e., corn stover, switchgrass, and short rotation woody crops). By factoring in farmer/feedstock producer and biorefinery profitability requirements and risk attitudes, land availability and suitability, as well as a time delay and technological learning factor, a more realistic estimate of the level of biojet supply and emissions reduction can be developed under different oil price assumptions. Factors that drive biojet GHG emissions and unit production costs from each feedstock are identified and quantified. Overall, this study finds that at likely adoption rates biojet alone would not be sufficient to achieve the aviation emissions reduction target. In 2050, under high oil price scenario assumption, GHG emissions can be reduced to a level ranging from 55 to 92%, with a median value of 74%, compared to the 2005 baseline level.
1. INTRODUCTION The aviation sector, facing mounting pressures to reduce its greenhouse gas (GHG) emissions, has begun to consider the use of biobased jet fuels (henceforth called biojet) as alternatives to fossil fuels (henceforth called petrojet).1,2 One of the goals that has been put forward is to achieve carbon neutral growth by 2020 and reduce the GHG emissions by 50% compared to the 2005 baseline level by 2050.3 In the United States, the aviation sector is responsible for about 11% of the total transportation GHG emissions.4 A recent study has shown that biojet on a unit energy basis can potentially lower GHG emissions by up to 85% when compared to petrojet.5 For biojet to achieve its GHG emission reduction potential, technical and economic hurdles must be overcome. Alternative jet fuels must have characteristics sufficiently similar to current petrojet regardless of the feedstock and refining process (i.e., be a “drop-in” fuel). However, the biojet produced by current refinery processes does not contain aromatic compounds, which account for up to 25% of petrojet by volume and are needed for proper lubrication and sealing.6 This, along with the requirement to meet fuel density specifications for aviation r 2011 American Chemical Society
fuel, requires that biojet be blended with petrojet. Currently, a 50% 50% blend by volume between biojet and petrojet fuel is the norm for meeting fuel property and performance specifications and is thus used in this paper as the upper threshold for blending.7 Although life cycle assessment (LCA) provides a sound basis to evaluate the overall environmental impacts (including GHG emissions) of biofuels,8 traditionally LCA studies have focused mostly on the environmental performance of technology options and largely left out the economic aspect of the system in question9 or at most include economic performance as a separate part.10 12 To reveal the achievable environmental benefits of emerging technologies such as biofuels, the economic motives of actors (i.e., decision-makers and stakeholders) involved along the life cycle stages have to be considered along with technical advances. There have been increasing interests and efforts on Received: September 7, 2010 Accepted: September 29, 2011 Revised: September 26, 2011 Published: September 29, 2011 9133
dx.doi.org/10.1021/es202148g | Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology developing LCA methodology along this line (e.g., 13, 14), but to our best knowledge there has been no study conducted which focuses on the GHG emissions reduction potential of biojet. The objective of this paper is to reveal the extent to which biojet can reduce U.S. aviation GHG emissions through the consideration of the role and perspective of relevant actors. Decisions made by actors based on their motives, interests, and responses to incentives determine whether a policy objective can be achieved.15 This study looks at solely the GHG emissions within the U.S. domestic context with regard to the air transportation market and feedstock production. There are currently two biojet production technologies available: (1) hydrotreating/hydrocracking process which uses vegetable oil as the feedstock,16 and (2) gasification followed by Fischer Tropsch synthesis and syncrude upgrading, which uses lignocellulosic feedstocks17 (see Supporting Information, SI). The following feedstock options are considered in this study: (1) oil-producing feedstock: camelina (as representative of low-input oilseeds), and algae, and (2) lignocellulosic biomass: short rotation woody crops (SRWCs), corn stover (as representative of agriculture residue), and switchgrass (as representative of herbaceous energy crops). Out of concerns to the disruption on food production and effects of land-use change, soybean is excluded even though it is an important feedstock for biodiesel.18,19
2. METHODOLOGIES The model developed in this paper contains two major elements. The first element calculates the costs and GHG emissions associated with different life cycle stages for different feedstock and production pathways. The second element is a forecast model to determine the actual biojet supply and demand, and hence the life cycle GHG emissions of the U.S. aviation industry. The feedstock and process parameter values that are used as the baseline case in this study are specified in Tables S1 S9, SI. An uncertainty analysis is then conducted using the probability distribution specified for certain parameters (Tables S12 S17, SI). 2.1. Biojet Production Cost and GHG Emissions at Different Life Cycle Stages. Because this study focuses on system level
interaction among the actors involved, U.S. average data are used when developing the life cycle inventory. For feedstock considered, average yield is used over the regions suitable for growing the feedstock. For biorefinery, effects of location are not considered. That is, it is assumed that the biorefinery will require same capital and operating costs and have same conversion efficiency no matter where it is located. We use the data entries in Ecoinvent v.2.0 (especially the RER processes) as the bases for life cycle inventory development. This is based on the technology similarity between U.S. and EU. To better reflect U.S. scenarios, we modify the Ecoinvent unit processes (“U”) by replacing the original electricity input with U.S. power grid mix.20 Given the limited coverage of USLCI database, this has been a common approach adopted by researchers.21,22 System expansion method is adopted to account for coproducts in terms of revenues and emission credits. For feedstock production, the functional unit is defined as per ton of feedstock delivered to a biorefinery, as this is also the unit used by actors to make transactions. Similarly, for fuel production, the functional unit is defined as per gallon of biojet delivered. In this study, climate change is the only environmental impact considered and it is measured by kg
POLICY ANALYSIS
CO2 equivalent. The IPCC AR4 (2007) version is used to convert CO2, CH4, and N2O into CO2e23 for time horizon 100 year. Emissions of N2O from nitrogen fertilizers applied during cultivation are included. Using both direct and indirect emission factors from IPCC AR4, 1.325% of applied N in fertilizer is emitted as N2O nitrogen. To predict feedstock and biojet production cost over the time period considered, the current total cost is first broken down into subcategories such as land rent, fertilizer, capital equipment, fuel, and labor. Table S1 lists the models used to forecast price change. General inflation is assumed for other costs involved. Camelina. Although camelina (Camelina sativa) is not the most common oilseed grown in the U.S., it is considered here because camelina-based jet fuel has been actually produced, albeit in small quantity and tested using airplane of commercial airlines.7 Field tests suggest that seeding camelina requires no till or minimum till. Drilling lightly or broadcasting with roller harrow is appropriate for planting. Harvesting can be done using combining (direct or swathed) with 6/64 to 3/64 slotted screens, which are the same for harvesting alfalfa. Camelina yield in marginal land (i.e., Conservation Reserve Program, CRP, land) has not yet been documented, so is highly uncertain. So, based on expert consultation,24 it is assumed that camelina yield in marginal land falls between 50 and 70% of that observed in field tests.25,26 A study by Montana State University Agriculture Extension provides a detailed cost breakdown of camelina cultivation and those data are adopted here.27 Since agronomic and crop production improvement strategies are just beginning to be applied, it is expected that camelina yield could increase significantly over years. This increase is assumed to be achieved through scientific improvements instead of increased fertilizer usage.28 Although camelina meal contains high levels of Omega-3 fatty acids, it also has glucosinolate content which limits its use as animal feed.29 So, it is assumed that most of the camelina meal will be used as fuel to replace mill residue. As a conservative assumption, it will be sold at mill residue price of $20/ton.30 The farming and preprocessing (oil extraction) parameters for camelina are given in Tables S2 and S3. Algae. Currently, there are mainly two approaches to grow algae: open pond and photobioreactors. Photobioreactors offer higher productivity and less evaporative water loss, but require significantly larger capital investment and operating cost. A couple of recent studies suggest that biodiesel derived from algae cultivated in photobioreactors can have GHG emissions up to ten times higher than that from open-pond cultivation.10,31,32 Also, the production cost of algae lipid in the photobioreactors case is 60 100% higher than that in the case of open-pond production. Therefore, in this study it is assumed that open-pond systems will be used for algae cultivation. Lipid will be extracted through mechanical pressing and spent biomass will be digested for methane generation. The EPA report Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis lists the achievable algae yield, lipid content, lipid production cost, and GHG emissions for three cases: (1) a base case which corresponds to a reasonable but still challenging target for the near future (the year 2022); (2) an aggressive case which assumes identification of a strain with near optimal growth rates and lipid content; and (3) a maximum case which represents the near theoretical maximum based on photosynthetic efficiencies.10 These predictions, along with the material/energy input data, are adopted in this study. For the current scenario, data are extrapolated using algae yield 9134
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology and lipid content reported throughout the literature.33 The cultivation parameters for algae are given in Table S4. Corn Stover. To avoid soil erosion, it is assumed that up to 50% of corn stover can be collected for biojet production. It is also assumed that annual corn stover yield is the same as the grain yield. Thus, the future yield of corn stover can be estimated by linearly extrapolating corn grain yield data between 1985 and 2009.10 Removing corn stover will require replenishing the soil with nutrients. The amount of N/P/K required for every ton of corn stover removed is assumed to be constant. N2O credit due to corn stover removal is included by assuming 0.45% N content in corn stover and 1.25% N in N2O avoided per unit of N in stover removed. Material inputs and cost data are derived from an EPA report10 (all data are averaged for three different farm sizes. i.e., 200, 400, and 800 acres).The plant size in the EPA report is 4000 ton/day while the plant in this study is 2000 ton/day. This will affect the transportation cost from satellite storage to the plant, which is adjusted accordingly. In the base case, the farmer profit margin is assumed to be 10% of the production cost. It should be noted that although the on-farm product cost suggested by the EPA report is largely the same as reported in previous studies, there are significant differences reported on the cost of transportation (from farm edge to satellite storage and from satellite storage to plant) and storage.34 The total delivered cost in the EPA report is in agreement with some more recent publications thus is believed to be a more realistic estimate.35 The cultivation parameters for corn stover are given in Table S5. Switchgrass. According to Lee et al.,36 the average switchgrass yield in the U.S. CRP land varies between 0.85 and 3.6 ton/acre. For material inputs and cost details, a recent study conducted by Iowa State University Agriculture Extension is adopted.37 To reduce fertilizer usage, switchgrass will be harvested once a year. After establishment, the stands can produce for the next 10 years. It has been reported that switchgrass yield responds to N application rate linearly with a maximum yield achieved at 112 kg N/ha. It is assumed that the yield increase in the future will come from improved species, field management, and timing of fertilizer application so the N application rate can be kept the same. Also, although switchgrass yield does not respond to P and K, they are needed to maintain soil nutrient level. The amount of P and K needed is calculated based on a per ton residue removed basis. For storage, the Iowa State study cited a higher storage cost than in the EPA report ($16.67/ton vs $8.89/ton) mainly due to a higher building cost. For comparison purposes, the same EPA number is used as in the corn stover case. The cultivation parameters for switchgrass are given in Table S6. Short Rotation Woody Crops (SRWCs). As assumed in the case of switchgrass, CRP land or other low productivity cropland will be used for growing SRWCs. Since there is very little data, if anything, published on SRWC yield on CRP, the yield on natural forest is used as a surrogate, which ranges from 1 to 3.8 ton/ha.38 Based on the maximum observed annual yield, it is expected that the yield will be doubled in 2030.39 It is assumed that yield increase is the result of scientific improvements rather than increased fertilizer application. That is, the amount of fertilizer applied per acre will not change over the years. Compared with switchgrass, SRWCs require much less chemical inputs. The production cost is dominated by land rent, stumpage, and harvesting. The cultivation parameters for SRWCs are given in Table S7. Production Process. Hydrotreating/Hydrocracking Based Conversion. The developer of the hydrotreating/hydrocracking technology suggests that for a plant of 350 000 m3/year diesel
POLICY ANALYSIS
production capacity, the typical Inside Battery Limits (ISBL) erected capital cost estimates are between 60 and 80 million.40 It is assumed that a biojet plant of the same size will have similar ISBL erected cost. Total capital costs are usually assumed to be twice the ISBL capital costs.41 Recently, the DOE provided a $241 million loan guarantee to support the construction of a 137million gallon per year renewable diesel facility using the technology.42 The capital cost investment is estimated from this amount of capital. Because the considered plant has a lower capacity, an economies of scale factor of 0.8 is assumed.43 For diesel production, Kalnes et al. gives the range of yields for diesel and coproducts. The low end of diesel yield is adopted since biojet yield is lower than that of diesel.22 The biorefinery parameters for oil-producing feedstock are given in Table S8. Gasification and FT (Fischer Tropsch)-Synthesis. It is assumed that a plant size of 2000 ton/day is a good balance between the economies of scale and transportation cost at the current feedstock yield. Swanson et al. investigated the production cost of biofuel production via gasification and FT synthesis for the nth plant.44 The data for the case of low temperature fluidized bed gasifier is used here due to its higher technology readiness. For the same plant size, EPA report gives the production cost for the year 2022 based on improved plant availability, reduced capital cost, and higher fuel yield. It should be noted that diesel is the target product in both reports and is used as a surrogate for biojet. The biorefinery parameters for lignocellulosic feedstock are given in Table S9. 2.2. Forecast Model. To evaluate actors’ decisions and their impacts on achievable GHG emissions reduction, a forecast model is employed as shown in Figure 1. The model presents simplified biojet supply and demand logic influenced by actor’s decisions, regulatory and land constraints, as well as the cost, technology, and dynamics. The model is implemented using the Matlab software. Policymakers. From the policymaker’s point of view, although this study focuses on the U.S. context, U.S. domestic policies on climate change are partly conditioned to the results of international negotiations. The U.S., for instance, would not commit to a large GHG emission cut without similar commitments from all other major emitting countries.45 Policymakers drive the biojet demand by setting policy goals to reduce GHG emissions. Based on the carbon reduction goal, a “committed” emissions reduction trajectory is established. The discrepancy between the committed reduction trajectory and the emissions due to demand growth creates demand for biojets over time (Figure S2, SI). In the model, three oil and petrojet price scenarios are adopted. The scenarios were extrapolated from the U.S. Energy Information Administration data46 and are designated as LowOil, ReferenceOil, and HighOil (Figure S3, SI). Airlines. For airlines, the decision to use biojet is mainly influenced by the degree of savings in carbon costs. To influence this decision, policymakers have leverage on establishing a carbon marketplace to regulate the CO2 price. The earliest possible certification of biojet standard is assumed to be 2013.5 The aggregate U. S. airline fleet has shown a robust trend of improved fuel efficiency.47 The efficiency is captured by what is known as payload fuel energy efficiency (PFEE) (kg-km/MJoule), which measures payload (i.e., passenger and aircraft belly cargo) and flight range with respect to one unit energy of fuel.48 Air travel demand is commonly represented as revenue-passenger-kilometer (RPK), which equals the number of passengers multiplied by the flight distance, a counterpart 9135
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology
POLICY ANALYSIS
Figure 1. Flowchart for biojet demand and supply algorithm based on actors’ decision criteria, policy drivers, technological learning, time variable, and land constraint.
of the vehicle-miles-traveled (VMT) measure for road transport. A simple compound growth model of RPK is implemented in this study. The PFEE and RPK projections are used to estimate the overall jet fuel consumptions and carbon emissions (more is described in SI). Biorefineries. For biorefineries, the decision to build a refinery plant depends on whether such an undertaking provides viable return on investment. Thus, the net present value (NPV) and internal rate of return (IRR) are used as decision criteria. If the NPV and IRR exceed the threshold value and there is enough biojet demand, the biorefinery plant will be built, creating demand for feedstocks. The potential market share of biojet is modeled based on the logic that because all biojet fuels are supposed to be delivered as “drop in” fuels, the market share of each fuel source would be mainly determined by its relative unit production cost.49 When the supply of a particular feedstock is limited by factors such as land availability, its actual market share will be lower than its
potential. In this case, a land allocation approach similar to that of Smeets et al.50 is employed: the market share gap will be filled by other unconstrained and economically viable feedstocks, proportionally to their potential market share (eq S5, SI). Feedstock Producers/Farmers. Farmers/feedstock producers will satisfy the demand only if they get a certain profit margin from producing the feedstock. A threshold value of 10% is set, above which the feedstock will be produced, provided that there is demand and land available.51 Applied to the feedstock production costs, the profit margin in effect will determine the price that biorefineries would need to pay for the feedstock. To forecast feedstock cost, the cost at the plant gate is broken down into contributions from land rent, fertilizer/herbicide cost, farm labor cost, fuel cost, and other costs (e.g., farm machinery, seed, harvest, and transport). For algae, an open-pond system (raceway type) is assumed, which is deemed more economically viable than a photobioreactor system. 9136
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology
POLICY ANALYSIS
Figure 2. Comparison of the 2013 unit life cycle GHG emissions estimates and their range of uncertainties for different feedstocks. The unit emissions are broken down into major constituents.
Land Availability and Feedstock Land Competition. Another key factor influencing biojet supply is land availability for feedstock. To minimize the impact of biofuel production (including biojet) on the food supply, only abandoned agricultural land in the U.S. is considered.10 Due to its climatic and soil conditions and considering feedstock characteristics, the same patch of land can be shared among multiple feedstocks or is only suitable to a certain feedstock. The land suitability of the feedstock has been estimated by the U.S. Department of Energy for switchgrass and SRWCs52 and by the U.S. Department of Agriculture for camelina, which is based on natural habitat observation.53 A complete state by state distribution of land availability and suitability for the three feedstocks is given in Table S11. For corn stover, 75 million dry ton/year is now available and 169.7 million dry ton/year is available in the long term.54 A large-scale algal cultivation is subject to the availability of saline groundwater, solar radiation, and large stationary sources of CO2.55 The competition among feedstock for land (i.e., between camelina, switchgrass, and SRWCs) was dealt with based on
their profit-making potential, as implied by the market share equation (eq S5 SI). The competition also considers revenues from soil organic carbon (SOC) sequestration as an important new income source for farmers.56 A sequestration of 800 kg C/ha/year is assumed for switchgrass56 and 1860 kg C/ha/ year is assumed for SRWCs.57 There is no evidence for carbon sequestration for camelina in the literature. A carbon price of $27 per metric ton CO2 based on EPA’s projection is adopted.58 It should be noted that there is also competition for feedstock (thus land) to support production of biofuels used by different transportation sectors (e.g., ground transportation, aviation, ocean transportation). However, this is beyond the scope of this paper. About 11% of the total land available for biofuel production was allocated to biojet in this study. This fraction resembles the share of energy consumed by air transportation relative to the whole transportation sector.4 Experience/Learning Curve Effect. As the biojet production accumulates over time, the production costs will decline due to learning (i.e., experience curve). For the baseline case, a progress 9137
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology
POLICY ANALYSIS
Figure 3. Unit production cost of biojet comprising farming and refining components and the uncertainty range in 2013.
ratio of 81% for refinery cost reductions was assumed, which is derived from experiences within the bioethanol industry.59 In other words, when the accumulated biojet production doubles, the refinery costs will decline by 19%. Time Delay and Risk Attitude. Time delays create inertia in the biojet production between demand signal and actual supply.60 A time delay factors in the cultivation and biorefinery plant construction time. The length of time delay was modeled as a function of fuel price based on the assumption that the higher the fuel price, the more risks the actors are willing to take. This risk-seeking attitude will manifest by, for example, starting the plant construction earlier in anticipation of future demand.61 For the baseline case, a total of 3, 2, and 1 year delays are assumed for the LowOil, ReferenceOil, and HighOil scenarios, respectively. Furthermore, actors’ more risk-seeking attitude as the price of oil increases is reflected also in how they select the decision criteria threshold value. First, the IRR threshold is assumed to decrease from 15% in the LowOil to 10% and 7.5% in the ReferenceOil and High Oil, respectively. Second, in the LowOil scenario, a biorefinery plant will be built only if the biojet demand equals or exceeds the plant design capacity. This criterion is relaxed in the ReferenceOil and HighOil to become 0.9 and 0.75, respectively. In the sensitivity analysis, the value of these parameters is varied.
3. RESULTS AND ANALYSIS 3.1. GHG Emissions of Biojet. Figure 2 presents the estimated unit GHG emissions of biojet, in g CO2 e/MJoule, produced from the five feedstocks in 2013. The GHG emissions are broken down into eight components and for each of these categories, the uncertainty range from the baseline case is estimated (SI provides the specification of uncertain variables). Feedstock cultivation is largely responsible for GHG emissions, particularly in biojet from oil-containing feedstocks. The amount can reach as high as 90% of the total in the case of algae. For camelina, the use of fertilizer contributes about 70% of the total emissions. The land use effect due to soil organic carbon (SOC) sequestration (for switchgrass and SRWCs) has considerable contribution and uncertainty. Overall, all feedstocks have a lower unit of GHG emissions compared to the standard unit emission reference for petrojet, which is 85 g CO2e/MJ fuel.62 A sensitivity analysis on the emissions drivers confirms that yield is the most important factor influencing emissions for most feedstock (details in SI). The second most important factor is the efficiency of fertilizer usage by the plant (in kg feedstock/kg fertilizer). Moreover, the change in SOC sequestration has a more profound effect in switchgrass (9% reduction in emissions for every 1% SOC increase) than in SRWCs (3% reduction for every 1% SOC increase). 9138
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology
POLICY ANALYSIS
Figure 4. Comparison of biojet economic performance in the HighOil price scenario (baseline case). The drop in production cost is due to learning curve effect and consequently results in a rise in NPV and IRR.
3.2. Biojet Unit Production Costs. Figure 3 presents the breakdown of unit production cost per gallon of biojet for each feedstock in 2013. Camelina and corn stover have the lowest total unit cost whereas algae has the highest. The unit production cost is divided into farming and refining process to identify the contribution of each subprocess. The farming costs range between 45% (for corn stover) and 92% (for algae) of the total unit production costs. For camelina, the major cost driver is land costs. For algae, about 40% of the unit cost is due to capital costs. The analysis also shows that revenues obtained from refinery coproducts have a bigger influence than those from carbon sequestration in lowering the production costs. A sensitivity analysis performed illuminates some cost drivers. For all feedstocks, it is of no surprise that yield has the major influence in the unit cost. For camelina, for instance, a 1% increase in yield will reduce unit cost by almost 1%, whereas for switchgrass the reduction is about 0.4%. The second major cost driver is capital cost, which represents the capital needed for the acquisition of the equipment and machinery. Evolution of Biojet Economic Performance: Unit Production Cost, NPV, and IRR. To illustrate the economic performance of
feedstocks, Figure 4 shows the evolution of biojet unit production cost, net present value (NPV), and internal rate of return (IRR) of biojet in the HighOil price scenario (figures for the other scenarios are given in the SI). In terms of unit production cost, biojet from camelina and corn stover can compete with petrojet from the beginning of the time frame (2013) (Figure 4a). Switchgrass and SRWCs-derived biojet becomes competitive after 2020 whereas biojet from algae does so after 2040. By contrast, in the LowOil scenario, the unit production cost of all five feedstocks is higher than the price of petrojet, so none of them is viable for production (Figure S5, SI). The second criterion for biojet production is NPV (Figure 4b). In the HighOil scenario, the NPV figures are positive at year 2013 for all feedstocks except for SRWCs and algae, which will break even (NPV g 0) around 2025 and 2040, respectively. In the ReferenceOil scenario, only the production of biojet from camelina and corn stover is expected to result in a positive NPV region (Figure S6, SI). The threshold value for the IRR criterion is set at 15%, 10%, and 7.5% for the LowOil, ReferenceOil, and HighOil scenarios, respectively, serving as a higher investment barrier compared to 9139
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology
POLICY ANALYSIS
Figure 5. Emissions trajectories under the three oil price scenarios (baseline case). None of emissions trajectories achieve the 2050 reduction target.
the positive NPV criterion. In the HighOil scenario, there is a progression in time for feedstock to meet biorefinery’s criteria to be a viable biojet source. For example, camelina-derived biojet will become viable at year 2013 and algae by 2040. In the ReferenceOil scenario, camelina-based biojet is the only economically viable feedstock, which occurs around 2022. None of the feedstocks is viable in the LowOil scenario Consequently, biojet from a mix of feedstocks manifests at different time frame in the HighOil scenario. The supply mix constitutes corn stover, camelina, and switchgrass starting from 2022 onward when there is enough demand to justify a full industrial production (Figure S8c, SI). Biojet from SRWCs will be produced starting 2030 and algae-derived fuel after 2040. By contrast, in the ReferenceOil the biojet supply is dominated solely by camelina (Figure S8b, SI). There are several observations on the dynamics of system behavior. First, supply does not become available until 2022 in the HighOil scenario due to the demand and time delay. Second, the drop in production cost trend is caused mainly by the reduction in refining costs due to experience/learning rate effect. The decline in unit production costs results in a jump in the NPV and IRR figures. The algae-derived biojet unit price also benefits from camelina learning rate since both feedstocks share the same biorefinery facility. 3.3. GHG Emissions Trajectories. Figure 5 shows the emissions trajectories under the three oil price scenarios using the baseline case set up. As a business-as-usual case is the emission trajectory generated without biojet use and a 2% annual demand growth. Due to the continuous improvement in the aircraft payload fuel energy efficiency, PFEE, the emissions grow at an annual average rate of around 0.67% but the emissions in 2050 could be 28% higher than the 2005 baseline level, widely missing the target of 50% below the 2005 baseline level. With biojet options, the realized emissions level in 2050 is 71% of the 2005 baseline level for the HighOil scenario. In the LowOil and ReferenceOil scenario, the emissions level is 128% (i.e., the same as the case without biojet) and 124%, respectively. In the HighOil price scenario, emissions begin to rise around 2038 after downward trend due to the constraint of the 50 50 blend requirement with petrojet (i.e., maximum biojet demand equals 50% of the total jet fuel consumption). By contrast, in the
ReferenceOil scenario, the supply of camelina based biojet is limited and cannot keep pace with the growth of air travel demand. In all three scenarios, the 2050 reduction target cannot be achieved. The assessment of uncertainty in the GHG emissions estimate is presented using a box plot at a 5-year interval (Figure 6). The plot applies for HighOil scenario and is generated from the 10 000 data points sampled using the Latin-Hypercube sampling method.63 In 2050, the median (i.e., 50th percentile, Q2) of the emissions is about 74% of the 2005 baseline level. In this scenario, the lowest emission level attainable (i.e., minimum value of 55% of the 2005 baseline level) is still above the 2050 reduction target. The condition is much worse in the ReferenceOil and LowOil scenarios, in which the minimum emission level is 120% and 128% of the 2005 baseline level, respectively (Figure S9, SI). The sensitivity analysis on aviation GHG emissions shows the extent of factors’ influence in achieving the reduction target (Figure S12, SI). The most significant factor is the fraction land dedicated for biojet production, followed by oil price.
4. DISCUSSION A commercial biojet fuel production system in the U.S. does not yet exist and thus offers opportunities for policy makers to influence its evolution to achieve the desired impacts. This study reveals some insights and implications for policy design. First, feedstock viability is conditional on two major factors: oil price and land availability. The lignocellulosic biomass based biojet (i.e., corn stover, switchgrass, and short rotation woody crops) only becomes viable when the oil price is high (i.e., the HighOil scenario). In this condition, its supply potential is more than four times larger than that of oil-producing feedstock (i.e., camelina and algae). When the oil price is lower (i.e., the ReferenceOil scenario), camelina is viable but its supply is constrained by suitability of land on which it can grow. Because policy makers may not intend to favor certain feedstock prematurely, they will need to consider the likelihood of oil price evolution. Second, to avoid potential competition with food production, the study considers the use of marginal lands for cultivating camelina, switchgrass, and SRWCs. The low productivity of this type of land, which can reach as low as half of that of cropland, 9140
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology
POLICY ANALYSIS
Figure 6. Uncertainty range of the evolution of aviation GHG emissions under the HighOil price scenario, given in a box plot depicting the minimum, quartile's, and maximum value.
can be detrimental for feedstock viability. A policy aiming to improve the productivity through a development of special feedstock variety, for example, should therefore become a priority. Third, it is also evident that biojet alone is not sufficient to achieve the 2050 GHG emission reduction target. Consequently, other measures are needed, including a steeper improvement in the fuel efficiency of the U.S. aircraft fleet than the current trend shows and more fuel efficient operational procedures. This result confirms the current projections, in which biojet will be responsible for a large share of reduction. Lastly, the result shows that the 50 50 blend requirement can significantly hamper the attainment of the policy goal. The relaxation of this requirement will be enabled by improvements in areas such as development of additives for improving biojet density and improving aircraft fuel tank to prevent leakage due to lack of aromatics compounds in biojet. The scope of the work may underestimate the amount of potential biojet supply as well as feedstock mix and hence GHG emissions impact. This work focuses only on U.S. production capacity and is limited to the consideration of marginal lands. It therefore excludes supply that may come from countries such as Canada and Brazil and a possibility that farmers may actually cultivate crop lands, resulting in higher yields. First generation feedstocks such as soybean cannot be ruled out completely. Due to its importance, different oil price scenarios such as price spikes and oscillations may result in different actor decision behaviors. Further work will need to address these issues. It should also include a further study of different incentive schemes that can be targeted to actors and feedstocks. The methodology presented in this paper can inform the design of incentives that are more aligned with actors’ interest. Also, the questions of “who should pay” and “how the costs and payoffs should be shared” encapsulate the central policy problem. This equity issue needs to be addressed with vigor commensurate with the technical evaluation (e.g., 64).
’ ASSOCIATED CONTENT
bS
Supporting Information. Parameter specifications and additional model descriptions and results. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (765) 494-0418; fax: (765) 494-0307; e-mail: bagusdin@ purdue.edu.
’ REFERENCES (1) Kim, B. Y.; Fleming, G. G.; Lee, J. J.; Waitz, I. A.; Clarke, J. P.; Balasubramanian, S.; Malwitz, A.; Klima, K.; Locke, M.; Holsclaw, C. A.; Maurice, L. Q.; Gupta, M. L. System for assessing aviation’s global emissions (SAGE), Part 1: Model description and inventory results. Transp. Res., Part D-Transp. Environ. 2007, 12 (5), 325–346. (2) ATAG. Beginner’s Guide to Aviation Biofuels; Air Transport Action Group (ATAG), 2009. http://www.enviro.aero/Biofuels.aspx [accessed February 5, 2010]. (3) Butterworth-Hayes, P. Environmental regulations fly high and wide. Aerosp. Am. 2010, 4–6. (4) FAA. Aviation & Emissions: A Primer; Federal Aviation Administration, Office of Environment and Energy: Washington, DC, 2005. (5) Hileman, J. I.; Wong, H. M.; Ortiz, D.; Brown, N.; Maurice, L.; Rumizen, M., The feasibility and potential environmental benefits of alternative fuels for commercial aviation. In 26th International Congress of the Aeronautical Sciences (ICAS), Anchorage, AK, 2008. (6) Chevron. Aviation Fuels Technical Review; 2006. (7) Rahmes, T.; Kinder, J.; Crenfeldt, G. Sustainable Bio-Derived Synthetic Paraffinic Kerosene (Bio-SPK) Jet Fuel Flights and Engine Tests Program Results. In 9th AIAA Aviation Technology, Integration, and Operations Conference (ATIO) and Aircraft Noise and Emissions Reduction Symposium (ANERS), Hilton Head, SC, 2009. (8) Huo, H.; Wang, M.; Bloyd, C.; Putsche, V. Life-Cycle Assessment of Energy and Greenhouse Gas Effects of Soybean-Derived Biodiesel and Renewable Fuels; Technical Report, DOE ANL/ESD/08-2, Argonne National Laboratory; 2008. (9) Maclean, H.; Lave, L. Life cycle assessment of automobile/fuel options. Environ. Sci. Technol. 2003, 37 (23), 5445–5452. (10) U.S. EPA. Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis; Washington, DC, 2010. (11) Kou, N. N.; Zhao, F. Effect of multiple-feedstock strategy on the economic and environmental performance of thermochemical ethanol production under extreme weather conditions. Biomass Bioenergy 2011, 35 (1), 608–616. (12) Hang, Y.; Qu, M.; Zhao, F. Economical and environmental assessment of an optimized solar cooling system for a medium-sized benchmark office building in Los Angeles, California. Renewable Energy 2011, 36 (2), 648–658. 9141
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology (13) Grimes-Casey, H. G.; Seager, T. P.; Theise, T. L.; Powers, S. E. A game theory framework for cooperative management of refillable and disposable bottle lifecycles. J. Cleaner Prod. 2007, 15 (17), 1618–1627. (14) Davis, C.; Nikolic, I.; Dijkema, G. P. J. Integration of Life Cycle Assessment Into Agent-Based Modeling. J. Ind. Ecol. 2009, 13 (2), 306–325. (15) Walker, W. E. Policy Analysis: A Systematic Approach to Supporting Policymaking in the Public Sector. J. Multi-Criteria Decis. Anal. 2000, 9, 11–27. (16) Kinder, J. D.; Rahmes, T. Evaluation of Bio-Derived Synthetic Paraffinic Kerosene (Bio-SPK); The Boeing Company Sustainable Biofuels Research & Technology Program, 2009. (17) IATA. 2nd Generation Biomass Conversion Efficiency; International Air Transport Association: Montreal, 2010. (18) Hileman, J. I.; Ortiz, D. S.; Bartis, J. T.; Wong, H. M.; Donohoo, P. E.; Weiss, M. A.; Waitz, I. A. Near-Term Feasibility of Alternative Jet Fuels, Technical Report; Rand & Partner, 2009. (19) Williams, P. R. D.; Inman, D.; Aden, A.; Heath, G. A. Environmental and Sustainability Factors Associated With Next-Generation Biofuels in the US: What Do We Really Know? Environ. Sci. Technol. 2009, 43 (13), 4763–4775. (20) Ecoinvent. www.ecoinvent.org. (21) Hsu, D. D.; Inman, D.; Heath, G. A.; Wolfrum, E. J.; Mann, M. K.; Aden, A. Life Cycle Environmental Impacts of Selected US Ethanol Production and Use Pathways in 2022. Environ. Sci. Technol. 2010, 44 (13), 5289–5297. (22) Kalnes, T. N.; Koers, K. P.; Marker, T.; Shonnard, D. R. A Technoeconomic and Environmental Life Cycle Comparison of Green Diesel to Biodiesel and Syndiesel. Environ. Prog. Sustain. Energy 2009, 28 (1), 111–120. (23) Intergovernmental Panel on Climate Change (IPCC). Climate Change 2007: Synthesis Report; Geneva, 2007. (24) Johnson, K., Department of Agronomy, Purdue University, 2011. (25) McVay, K. A.; Lamb, P. F. Camelina Production in Montana; Montana State University Extension: Bozeman, MT, March 2008. (26) Ehrensing, D. T.; Guy, S. O. Camelina; Oregon State University Extension Service, 2008. (27) Johnson, D. Crops for Biodiesel Research in Montana: Camelina; Montana Department of Environmental Quality, Biodiesel Production Education Workshop: Livingston, MT, 2007. (28) Shonnard, D. R.; Williams, L.; Kalnes, T. N. Camelina-Derived Jet Fuel and Diesel: Sustainable Advanced Biofuels. Environ. Prog. Sustain. Energy 2010, 29 (3), 382–392. (29) Pekel, A. Y.; Patterson, P. H.; Hulet, R. M.; Acar, N.; Cravener, T. L.; Dowler, D. B.; Hunter, J. M. Dietary camelina meal versus flaxseed with and without supplemental copper for broiler chickens: Live performance and processing yield. Poultry Sci. 2009, 88 (11), 2392– 2398. (30) U. S. Environmental Protection Agency. Combined Heat and Power Partnership, Biomass Combined Heat and Power Catalog of Technologies; Washington, DC, 2007. (31) Jorquera, O.; Kiperstok, A.; Sales, E. A.; Embirucu, M.; Ghirardi, M. L. Comparative energy life-cycle analyses of microalgal biomass production in open ponds and photobioreactors. Bioresour. Technol. 2010, 101 (4), 1406–1413. (32) Stephenson, A. L.; Kazamia, E.; Dennis, J. S.; Howe, C. J.; Scott, S. A.; Smith, A. G. Life-Cycle Assessment of Potential Algal Biodiesel Production in the United Kingdom: A Comparison of Raceways and AirLift Tubular Bioreactors. Energy Fuels 2010, 24, 4062–4077. (33) Sheehan, J.; Dunahay, T.; Benemann, J.; Roessler, P. A Look Back at the U.S. Department of Energy’s Aquatic Species Program— Biodiesel from Algae; NREL/TP-580-24190; 1998. (34) Brechbill, S. C.; Tyner, W. E.; Ileleji, K. E. The Economics of Biomass Collection and Transportation and Its Supply to Indiana Cellulosic and Electric Utility Facilities. Bioenergy Res. 2011, 4 (2), 141–152. (35) Kazi, F. K.; Fortman, J. A.; Anex, R. P.; Hsu, D. D.; Aden, A.; Dutta, A.; Kothandaraman, G. Techno-economic comparison of process
POLICY ANALYSIS
technologies for biochemical ethanol production from corn stover. Fuel 2010, 89, S20–S28. (36) Lee, D. K.; Owens, V. N.; Doolittle, J. J. Switchgrass and soil carbon sequestration response to ammonium nitrate, manure, and harvest frequency on conservation reserve program land. Agron. J. 2007, 99 (2), 462–468. (37) ISU Ag Extension. Estimated Costs for Production, Storage and Transportation of Switchgrass; Iowa State University: Ames, IA, 2007. (38) Wright, L. L.; Hughes, E. E. United-States Carbon Offset Potential Using Biomass Energy-Systems. Water, Air Soil Pollut. 1993, 70 (1 4), 483–497. (39) Mann, M. K.; Spath, P. L. Life Cycle Assessment of a Biomass Gasification Combined-Cycle System; DOE TP-430-23076; Washington, DC, 1997. (40) Marker, T. L. Opportunities for Biorenewables in Oil Refineries; Final Technical Report, DOE GO15085, 2005. (41) Toman, M. J. G.; Lempert R. J. Impacts on U.S. Energy Expenditures and Greenhouse-Gas Emissions of Increasing RenewableEnergy Use; Technical Report TR-384-1; RAND Corporation, 2008. (42) DOE. Department of Energy Offers First Conditional Commitment for a Loan Guarantee for Advanced Biofuels Plant, 2011. http://www.energy.gov/9991.htm [accessed January 2011]. (43) Agusdinata, D. B. Exploratory Analysis to Support Real Options Analysis: An Example from Electricity Infrastructure Investment. IEEE International Conference on Systems, Man and Cybernetics, Waikoloa, Hawaii, 2005; Waikoloa, Hawaii, 2005. (44) Swanson, R. M.; Satrio, J. A.; Brown, R. C.; Platon, A.; Hsu, D. D. Techno-Economic Analysis of Biofuels Production Based on Gasification; NREL/TP-6A20-46587; 2010. (45) Committee on America's Climate Choices, National Research Council, America’s Climate Choices; The National Academies Press: Washington, DC, 2011. (46) EIA. Annual Energy Outlook 2010; U.S. Energy Information Administration: Washington, DC, 2010. (47) Lee, J.; Lukachko, S.; Waitz, I.; Schafer, A. Historical and future trends in aircraft performance, cost, and emissions. Ann. Rev. Energy Environ. 2001, 26, 167–200. (48) Hileman, J. I.; Katz, J. B.; Mantilla, J. G.; Fleming, G., Payload Fuel Energy Efficiency as a Metric for Aviation Environmental Performance. In 26th International Congress of the Aeronautical Sciences, Anchorage, AK, 2008. (49) Boyd, D. W.; Phillips, R. L.; Regulinski, S. G. A Model of Technology Selection by Cost Minimizing Producers. Manage. Sci. 1982, 28 (4), 418–424. (50) Smeets, E. M. W.; Faaij, A. P. C.; Lewandowski, I. M.; Turkenburg, W. C. A bottom-up assessment and review of global bio-energy potentials to 2050. Prog. Energy Combust. Sci. 2007, 33 (1), 56–106. (51) USDA. Structure and Finances of U.S. Farms: 2005 Family Farm Report; Economic Research Service-USDA: Washington, DC, 2005. (52) DOE. Breaking the Biological Barriers to Cellulosic Ethanol: A Joint Research Agenda. A Research Roadmap Resulting from the Biomass to Biofuels Workshop Sponsored by the U.S. Department of Energy, 2005. (53) USDA. http://plants.usda.gov/. (54) Perlack, R. D. W.; Turhollow, L. L.; Anthony, F.; Graham, R. L.; Stokes, B. J.; Erbach, D. C. Biomass As Feedstock for a Bioenergy and Bioproducts Industry: The Technical Feasibility of a Billion-Ton Annual Supply; U.S. Department of Energy and U.S. Department of Agriculture: Washington, DC, 2005. (55) Pienkos, P. T.; Darzins, A. The promise and challenges of microalgal-derived biofuels. Biofuels, Bioprod. Biorefin. 2009, 3 (4), 431–440. (56) Lemus, R.; Lal, R. Bioenergy crops and carbon sequestration. Crit. Rev. Plant Sci. 2005, 24 (1), 1–21. (57) Updegraff, K.; Baughman, M. J.; Taff, S. J. Environmental benefits of cropland conversion to hybrid poplar: Economic and policy considerations. Biomass Bioenergy 2004, 27 (5), 411–428. (58) U.S. Environmental Protection Agency. EPA Analysis of the American Clean Energy and Security Act of 2009 H.R. 2454 in the 111th 9142
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
Environmental Science & Technology
POLICY ANALYSIS
Congress, June 23, 2009. http://www.epa.gov/climatechange/economics/pdfs/HR2454_Analysis.pdf. (59) Bake, J. D. V.; Junginger, M.; Faaij, A.; Poot, T.; Walter, A. Explaining the experience curve: Cost reductions of Brazilian ethanol from sugarcane. Biomass Bioenergy 2009, 33 (4), 644–658. (60) Sterman, J. D. Business Dynamics: Systems Thinking and Modeling for a Complex World; Irwin/McGraw-Hill, 2000. (61) Neuhoff, K. Investment Decisions under Climate Policy Uncertainty; Electricity Policy Research Group, University of Cambridge, 2007. (62) Wong, H. M. Life-cycle Assessment of Greenhouse Gas Emissions from Alternative Jet Fuels; MIT, 2008. (63) Helton, J. C.; Davis, F. J. Latin hypercube sampling and the propagation of uncertainty in analyses of complex systems. Reliab. Eng. Syst. Saf. 2003, 81 (1), 23–69. (64) Agusdinata, D. B.; Delaurentis, D. A. Addressing equity issue in multi-actor policymaking via a system-of-systems approach: Aviation emissions reduction case study. J. Syst. Sci. Syst. Eng. 2011, 20 (1), 1–24.
9143
dx.doi.org/10.1021/es202148g |Environ. Sci. Technol. 2011, 45, 9133–9143
POLICY ANALYSIS pubs.acs.org/est
A “Carbonizing Dragon”: China’s Fast Growing CO2 Emissions Revisited Jan C. Minx,†,‡,* Giovanni Baiocchi,§ Glen P. Peters,|| Christopher L. Weber,^,( Dabo Guan,z,Δ,# and Klaus Hubacek3 †
)
Department Economics of Climate Change, Department Sustainable Engineering, Technical University Berlin, D-10623 Berlin, Germany ‡ Potsdam Institute for Climate Impact Research (PIK), D-14412 Potsdam, Germany § Norwich Business School, University of East Anglia, Norwich Research Park, NR4 7TJ Norwich, United Kingdom Center for International Climate and Environmental Research Oslo (CICERO), PB 1129 Blindern, Oslo, Norway ^ Science and Technology Policy Institute, Washington, D.C. 20010, United States ( Department for Civil and Environmental Engineering, Carnegie Mellon University, Pittsburgh, PA15213, United States z Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang, 110016, China Δ School of Earth and Environment, University of Leeds, Leeds, LS2 9JT, United Kingdom # St Edmund's College, University of Cambridge, Cambridge, CB3 0BN, United Kingdom. 3 Department of Geography, University of Maryland, College Park, Maryland 20742, United States
bS Supporting Information ABSTRACT: China’s annual CO2 emissions grew by around 4 billion tonnes between 1992 and 2007. More than 70% of this increase occurred between 2002 and 2007. While growing export demand contributed more than 50% to the CO2 emission growth between 2002 and 2005, capital investments have been responsible for 61% of emission growth in China between 2005 and 2007. We use structural decomposition analysis to identify the drivers for China’s emission growth between 1992 and 2007, with special focus on the period 2002 to 2007 when growth was most rapid. In contrast to previous analysis, we find that efficiency improvements have largely offset additional CO2 emissions from increased final consumption between 2002 and 2007. The strong increases in emissions growth between 2002 and 2007 are instead explained by structural change in China’s economy, which has newly emerged as the third major emission driver. This structural change is mainly the result of capital investments, in particular, the growing prominence of construction services and their carbon intensive supply chain. By closing the model for capital investment, we can now show that the majority of emissions embodied in capital investment are utilized for domestic household and government consumption (35 49% and 19 36%, respectively) with smaller amounts for the production of exports (21 31%). Urbanization and the associated changes in lifestyle are shown to be more important than other socio-demographic drivers like the decreasing household size or growing population. We argue that mitigation efforts will depend on the future development of these key drivers, particularly capital investments which dictate future mitigation costs.
’ INTRODUCTION China recently became the world’s largest consumer of energy1 and emitter of CO2,2 overtaking the U.S. earlier than expected.3 A growing body of literature has analyzed these trends in energy use and CO2 emissions in China.2,4 8 A key challenge to understanding these trends is the joint influence of several technological and socioeconomic drivers, which affect levels of CO2 emissions and energy use in China in different directions and to varying degrees. Given the rapid pace of development in China, monitoring the trends in emission drivers over time is important. China’s annual industrial CO2 emissions have grown by a factor of 2.7 (166%), that is almost 3992 million tonnes (Mt), between 1992 and 2007, r 2011 American Chemical Society
but more than 70% (2815 Mt) of this emission growth has occurred between 2002 and 2007. Hence, the rate of growth of CO2 emissions in China has increased (Figure S2, Table S10 in the Supporting Information (SI)), whereas the average annual emission growth rate was 5% for the period 1992 2002, it was 16% for the period 2002 2007. Extrapolation from 2002 to the year 2007 of the linear trend in annual emissions over the Received: May 2, 2011 Accepted: September 2, 2011 Revised: September 2, 2011 Published: September 02, 2011 9144
dx.doi.org/10.1021/es201497m | Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology Table 1.
POLICY ANALYSIS
Decomposition Factors Considered in This Analysis. A Comprehensive List Is Provided in SI Table S-3
abbreviation
name
definition the use of fuel type k as a share of total fuel use
Emix
fuel mix
ei
fuel intensity
TJ of total fuel used per unit Renminbi (RMB) of total economic output
f
process CO2 intensity
tonnes of industrial process CO2 emission used per RMB of total economic output
fp
CO2 intensity
tonnes of industrial (energy and process) CO2 emissions per RMB of total economic output
L
production structure
the direct and indirect input requirements throughout the Chinese economy to produce one unit of final demand
yl
final demand levels
the sum of all sector final demands measured in RMB
yc
final demand composition
final demand share of sector i in total final demand (yl)
yh u
per household final demand levels sum of all household final demand per household measured in RMB urbanisation share of urban population in China
hs
household size
households per person in China (inverse definition)
F
population size
size of population in China
1992 2002 period yields an increment of only 649 Mt of CO2 emissions, so that the 2166 Mt above the projected value can be considered “additional CO2 emission growth” compared to the earlier periods in our data (see Figure S4, SI). Hence, almost 80% (2166 Mt) of the 2815 Mt CO2 emission growth between 2002 and 2007 is considered “additional” here. This additional emission growth is not the sole reflection of the higher GDP growth rates (see Table S13, SI) achieved between 2002 and 2007. The CO2 intensity per unit GDP has also started to increase between 2002 and 2007 (Figure S6, Table S15, SI). This reverses a long-term trend of a decreasing CO2 intensity of GDP observed for the period 1992 2002. China has become a “carbonizing dragon” over recent years, but none of the available literature (e.g., refs 4,9), to our knowledge, has so far provided an in-depth understanding of the driving forces. Instead most of the recent literature on China has focused on the structural analysis of China’s emissions for the year 2007,10 12 China’s carbon emissions associated with international trade,9,13 17 emission patterns and trade at the provincial level, 18 20 as well as the supply chain of high emitting sectors such as buildings or construction.21,22 Peters et al.23 used structural decomposition analysis (SDA) to quantify the contribution of a set of emission drivers to changes in industrial CO2emissions in China between 1992 and 2002. The results show that CO2 emission increases from growing domestic and foreign final demands outweighed emission reductions from efficiency improvements. Other factors contributed insignificantly to CO2 emission changes in China over this period. In this paper we extend our earlier analysis by (1) updating the available evidence on emission drivers and trends in China through the inclusion of the most recent input-output data for the year 2007; (2) providing specific analysis for newly emerging trends not apparent in earlier studies; (3) providing more detailed decompositions to account for a larger variety of drivers such as process emissions, energy mix, urbanization, and so on, Table 1; (4) treating capital investments as input to production to demonstrate the importance of allocation issues in emission accounting.
’ METHODS AND DATA Our calculations use environmental input-output (EIO) analysis, which is a commonly accepted way of quantifying sectoral CO2 emissions and other environmental interventions under full consideration of the supply chain. The methodological details have been described elsewhere 24,25 and a complete description of the model is provided in the SI. Structural decomposition analysis (SDA) uses EIO to estimate the relative contribution of
changes in a set of emission drivers over time—such as carbon intensity, energy mix, production structure, final demand level, and composition and demographic factors—to overall changes in CO2 emissions,23,26 28 Table 1. Each of these factors is analyzed with regards to their respective contribution to changes in CO2 emissions over a given time period by keeping all other emission drivers constant. This ceteris paribus interpretation applies to all SDA interpretations undertaken in this paper without permanently referring to it. A challenge for any structural decomposition exercise is dealing with the existence of equally acceptable decomposition forms for the same changes.27 29 To resolve the issue, we compute the average of all possible decompositions, as has become common practice.9,23,28 30 For reference we report the standard errors and ranges for all our results in the SI and discuss sensitivities. We use three data sets throughout this study: time-series inputoutput tables (IOTs), corresponding sectoral energy and CO2 emission statistics and capital investment data. We obtained the IOTs from the Chinese National Bureau of Statistics (NBS) for 1992 with 118 sectors,31 1997 with 122 sectors,32 2002 with 124 sectors 33and 2007 with 135 sectors.34 As we are interested in Chinese emissions only, we remove imports from the IOT and only consider domestic supply chains. To facilitate comparisons we aggregated all tables to 95 sectors and converted the IOTs from current into constant prices using the double deflation method.35 Deflators were compiled based on the price data provided from Chinese Statistics Yearbooks of the various years.36 We documented our energy and emission data set previously for the years 1992 2002.23,37 We apply the same method to extend this data set to the year 2007 using information from China’s Energy Statistical Yearbook 2008.38 The complete data set consists of 18 types of fuel, heat, and electricity consumption in physical units and is available at a 44 sector level which we mapped to the 95 sectors of the IOTs. CO2 emissions from industrial processes are included in our analysis as well. For one part of our analysis we close the model for capital investment using the augmentation method in the absence of a detailed capital flow matrix for China,39 that is, we treat capital as an input to production rather than final demand. We take the required capital investment data from the China Statistical Yearbook 2008.40 The data available for 2007 showed investment in both construction and equipment by economic sector, whereas in years before 2007, data were only available for construction. Thus given no better data, for years before 2007 we assume that investment in equipment is proportional to investment in construction projects based on the 2007 ratio. 9145
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology
Figure 1. Cumulative changes in industrial CO2 emissions in China 1992 2007 by emission determinants (in million tonnes of CO2 emissions per year). The graph highlights the emergence of production structure changes (dash-dotted blue line) as the third major emission driver explaining observed increases in annual CO2 emissions (solid gray line) between 2002 and 2007 beside the two historically important drivers of fuel intensity of industrial activities (short-dash red line) and final demand level growth (dash-dotted magenta line) explaining observed increases in annual CO2 emissions (solid gray line) between 2002 and 2007. The underlying models are outlined in Table S4 in the SI.
’ RESULTS Structural Decomposition—General. Figure 1 (Table S5, SI) presents the overall results of the SDA and identifies the drivers behind the growth in CO2 emissions. It is derived by merging results from two separate decompositions: one focusing on changes in fossil fuel based CO2 emissions and one focusing on changes in CO2 emissions from industrial processes. Model and variable definitions are provided in the SI (Table S3 and Table S4, SI). Focusing on the entire period 1992 2007 additional CO2 emissions from growing consumption (yl) continue to outstrip emission reductions from efficiency gains (f) as in Peters et al.23 Reductions in the CO2 intensity (f) of production activities decreased emissions in China by 3737 Mt of CO2 between 1992 and 2007. Increases in final demand levels (yl) added 5421 Mt of CO2 emissions. While CO2 emission increases from growing consumption outpaced emission reductions from efficiency improvements by a factor two between 1992 and 2002,23 this is no longer the case when we only consider emission changes for the most recent period (2002 2007). The 2509 Mt rise in annual CO2 emissions from growing consumption (yl) was largely offset by CO2 emission savings from reductions in the CO2 intensity (f) of production activities ( 2284 Mt CO2). Hence, efficiency savings have been catching up with consumption growth in terms of CO2 emissions in recent years (2002 2007) despite the higher GDP growth rate than before. This could be seen as evidence that China’s efforts to improve production efficiencies as detailed in the 11th Five Year Plan for National Economic and Social Development41 have shown some success.
POLICY ANALYSIS
The increase in CO2 emission growth in China between 2002 and 2007 is explained by the emergence of production structure changes as a new major emission driver. Even though final demand growth (yl) remains the largest individual emission determinant for the period 2002 to 2007, production structure changes (L) cause a CO2 emission increase of a magnitude similar to that of the “additional” CO2 emission growth (108% or 2344 Mt of 2166 Mt CO2) estimated for that period. Our analysis therefore suggests that the steep rise in CO2 emissions in China between 2002 and 2007 is the result of structural changes in China’s economy toward more carbon intensive activities. Instead of focusing on overall changes in the CO2 intensity of production (f), we show contributions of fuel mix (Emix), fuel intensity (ei), and process CO2 intensity (fp) changes in Figure 1.9,23,42 This analysis reveals that almost all emission savings in China were linked to improvements in the energy efficiency of production (ei). CO2 savings from reductions in process CO2 emission intensity (fp) were negligible between 1992 and 2002, but have gained some relevance between 2002 and 2007 ( 583 Mt CO2). The industrial energy mix (Emix) remained stable and did not contribute to any significant CO2 emission changes between 1992 and 2007. Even though 1.5 GW of new capacity has been added on average to the Chinese electricity grid through the equivalent of building new or extending existing “large” coal power plants each week since 2005 to meet growing energy demand,43 this has not greatly affected the industrial energy mix due to considerable investments of the Chinese government into other energy technologies including renewables and nuclear. In the remainder of the paper we analyze emission drivers in each of three major final demand categories (capital, exports, and households) individually. Capital Investment. Emissions released to satisfy the growing demand for capital investment amounted to an annual increase of 1884 Mt CO2 between 1992 and 2007, of which 74% (1388 Mt CO2) occurred between 2002 and 2007. In absolute terms, this is the largest contribution to CO2 emission growth in China among all final demand categories (Table S10, SI). Most of the capital investment related CO2 emission changes in China can be attributed to changes in the production structure (L) and growth in final demand levels (yl,cap), Figure 2 (Table S6, SI). While production structure changes (L) contributed to reductions of CO2 emissions between 1992 and 2002, they were the largest component (1442 Mt) of capital investment related CO2 emission growth over the period 2002 to 2007. In fact, this contribution is also larger than the CO2 emission increase attributable to the growth in the level (yl,cap) of capital investments (1204 Mt CO2). It outstrips the joint emission savings from changes in the composition yc,cap of capital investments ( 219 Mt CO2) and reductions in the CO2 intensity (f) of production processes ( 1039 Mt CO2). Altogether, more than 60% of the additional CO2 emission growth estimated for the period 2002 and 2007 can be attributed to production structure changes (L) associated with final demands for capital investments. From a consumption perspective, 68% (944 Mt CO2) of China’s CO2 emission growth associated with capital investments from 2002 to 2007 are related to increased final demands for construction services, 22% (305 Mt) to heavy manufacturing sectors such as iron and steel or aluminum production and the remaining 10% to other sectors (139 Mt), Figure 2. The contribution of the construction sector as the dominant driver of change varies in importance across individual emission determinants: while between 63% and 78% of the CO2 emission 9146
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology
POLICY ANALYSIS
Figure 2. Stacked bar chart showing components of capital investment related CO2 emission changes in China between 2002 and 2007 (in million tonnes of CO2 emissions). The horizontal axis identifies major sector final demands for capital investments (sector of consumption) by emission determinant. The vertical axis shows changes in CO2 emissions. Negative changes are represented as bars stacking below zero. Production sectors releasing the CO2 in the supply chain of sector final demands are identified as colored segments of the stacks. The graph highlights production structure changes as the dominant factor associated with emission growth from capital investments as well as the importance of the construction sector and its supply chain for understanding recent CO2 emissions changes in China. The underlying model is outlined in Table S4 in the SI.
changes associated with the CO2 intensity (f), final demand composition (yc,cap) and final demand level (yl,cap) components can be attributed to construction demands, this number is 97% (1402 of 1442 Mt CO2) for production structure changes (L). The importance of construction is directly related to its carbon intensive supply chain, mainly associated with cement manufacturing (46% of total CO2 emissions), iron and steel production (20% of total CO2 emissions) and electricity demand (30% of total CO2 emissions). Hence, growing demand for construction from infrastructure investments—building of roads, houses, and other infrastructure—are the major driver of the carbonization (increase in CO2 emissions per unit of GDP) of China’s economy and explain the majority of the recent increases in CO2 emission growth observed between 2002 and 2007. Exports. While final demands for capital investments show the largest CO2 emission growth in China between 1992 and 2007 in absolute terms, export final demands have grown most rapidly compared to all other final demands (225% increase). In 1992 emissions from the production of goods and services destined for exports represented 408 Mt CO2 (17% of total emissions), whereas in 2007 export related emissions accounted for 1732 Mt CO2 (27% of total emissions). Hence, emissions from the production of exports in China in 2007 alone were greater than the total emissions from any country other than the U.S. During the period 1992 to 2007 the average emission growth associated with export production in China was 225% compared to 130% for capital
investments. Export-related emission growth peaked during the period 2002 2005, when it accounted for 50% of the overall growth in China’s CO2 emissions.9 However, this importance of exports for CO2 emission growth has been declining since, accounting only for 18% of the total emission changes between 2005 and 2007 (Table S10, SI). Between 2002 and 2007 the total CO2 emission growth from the production of export products amounted to about 985 Mt. Figure 3 (Table 7, SI) attributes the emission growth to a set of emission determinants as well as to producing and consuming sectors. Export related CO2 emission increases are almost entirely attributable to final demand level growth (yl,exp). In fact, for the period 2002 2007 exports have overtaken capital investments and contribute the largest share to CO2 emission growth associated with growing final demand levels across all final demand categories (yl). Overall, growing exports levels (yl,exp) contributed an additional 1257 Mt of CO2 emissions between 2002 and 2007. This is 46% of total CO2 emission increases from final demand level growth observed for that period across all final demand categories. In difference to the trends observed for capital investments, export final demand only made a relatively minor contribution (10% or 215 Mt) to emission growth related to changes in the production structure of China’s economy (L). Exports are not only growing, their composition (yc,exp) is also changing. Even though the resulting changes in CO2 emissions are comparatively small (approximately 10% (103 Mt CO2) of 9147
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology
POLICY ANALYSIS
Figure 3. Stacked bar chart of components of export related CO2 emission changes in China between 2002 and 2007. On the horizontal axis major sectors of final demand are identified, whereas on the vertical axis the cumulative changes in emissions released to meet that demand by different sectors of a specific supply chain are identified as colored segments of the stacks. Negative changes are represented as bars stacking below zero. The graph shows the importance of the final demand level component of CO2 emission increases associated with growth in exports. The underlying model is outlined in Table S4 in the SI.
the 985 Mt export related CO2 emission increase between 2002 and 2007), a brief analysis is instructive for better understanding China’s economic development path. Compared to previous periods demand for China’s typical export products such as toys, textiles or electrical devices remained strong, but there has been particularly strong growth in both energy-intensive and manufactured products of higher quality. Between 2002 and 2007 emissions associated with the export of “primary iron and steel manufacturing” increased 10-fold, while “metal products” doubled. For example, the export of rolled steel grew from 5.5 million tons in 2002 to 62.7 million tons in 2007 (11 times). Billet and crude forgings grew from 1.3 to 6.4 million tons (five times):38 other sectors expanding faster than in previous years include “household electric appliances”, “motor vehicles”, and “cement and asbestos materials” which all grew by over 500% between 2002 and 2007. This diversification of China’s export base may suggest a changing role of China in the global economy from a provider of cheap, labor intensive manufacturing toward high-end and more capital-intensive products. While CO2 emission increases attributable to changes in the domestic supply chains (L) of light and heavy manufacturing goods were very similar (428 Mt and 481 Mt respectively), variations were larger across other emission determinants. Light manufacturing, for example, contributed 57% to emission reductions from decreases in the CO2 intensity of sectors (f) and 49% to emission increases from growth in final demand (yl,exp), while this was 33% in both cases for heavy manufacturing. Even though
the sales of light manufacturing products to other countries are still overall the most rapidly expanding export segment, this does not directly translate into CO2 emission trends, because the respective domestic supply chains of these products are decarbonizing more strongly. Household Consumption and Lifestyles. Household consumption was the single most important final demand category in 1992 with a share of 40% (including direct and indirect emissions) in China’s CO2 emissions. In 2007 households were responsible for only 25% (1611 Mt) of total CO2 emissions. This is comparable in size to the export related emission share (27%). Between 2002 and 2007 households only had a minor role in the CO2 emission growth (12% or 340 Mt). This trend is distinct from evidence provided for many developed countries as well as the global average, where emissions from household consumption usually make up the largest share across final demand categories and growing household consumption is the dominant final demand force behind CO2 emission growth.44 50 However, these aggregate figures hide important differences between rural and urban households: while the direct and indirect CO2 emissions from rural household consumption fell by 55 Mt between 1992 and 2007, emissions from urban household consumption grew by 710 Mt CO2 over the same period. Key emission drivers behind changes in households’ CO2 emissions are typically population, income level, consumption patterns, and household structure.20,51 53 In China urbanization is another important socio-economic process that needs to be 9148
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology
POLICY ANALYSIS
Figure 4. Determinants of changes in industrial energy related CO2 emissions from household consumption in China 1992 2007 (in million tonnes of CO2 emissions per year). Urbanization and consumption growth together with production structure changes are driving CO2 emission growth from household final demand rather than socio-demographic factors such as household and population size. The underlying model is outlined in Table S4 in the SI.
considered. Migration to urban growth centers led to an 85% increase in the urban and a 14% decrease in the rural population (Table S16, SI). An emission increase of 339 Mt CO2 and an emission decrease of 101 Mt associated with urban and rural household consumption, respectively, can be attributed to urbanization (u), Figure 4 (Table S8, SI). The overall net emission growth from urbanization (u) of 238 Mt of CO2 emissions is larger than the CO2 emission growth from the 13% overall increase in the Chinese population (F) and the contribution of decreasing household size (hs) in rural and urban China. This highlights systematic differences between rural and urban lifestyles, which also manifest in differences in the composition and level of household consumption. Per capita household spending (in constant, 1997 producer prices) increased for both rural and urban households between 1992 and 2007. These increases in per household final spending levels were the dominant drivers of CO2 emission growth associated with household consumption (yh): 529 Mt for urban and 147 Mt CO2 emissions for rural households. Our data generally supports other evidence suggesting that the rapid growth of the Chinese economy has been accompanied by an increase in economic wealth across the Chinese population and a sharp decrease in poverty.54,55 However, it also highlights an important second trend: the growing inequality in China, here exemplified by the differences between rural and urban living in China. Per capita household consumption was 2.3 times larger for urban than for rural households in 1992 and 3.6 times larger in 2007. The key aspect of urbanization (u) in the Chinese context is that migration is often associated with higher incomes and hence
wealthier urban lifestyles. This lifestyle effect dominates potential carbon savings associated with the reaping of economies of scale associated with life in cities.56 58 If inequalities between urban and rural life continue to grow as observed between 1992 and 2007, the overall CO2 emissions increase associated with a switching in lifestyle will also continue to grow. Revisiting Results—treating Capital Investments As Inputs to Production. So far we have shown the distinct relationship between CO2 emission drivers and different final demand categories and their importance for understanding the increased CO2 emission growth in China between 2002 and 2007. In this section we highlight the importance of modeling assumptions in discussions of China’s CO2 emission growth. While the importance of capital investments for understanding trends in China’s CO2 emissions has been increasingly recognized,9,23 the question of how capital goods are treated has been largely neglected. In the previous sections we chose to treat capital investments as a domestic final demand. This is the default procedure in the input-output literature.9,11 13,23 However, capital goods could be equally considered an input to production. In this case CO2 emissions from capital investments arise within the domestic supply chain and are assigned to household, government and export final demands. How capital related emissions are accounted for is a policy relevant issue: For instance, one could argue that standard estimates of China’s emissions embodied in exports are underestimated since they do not include the emissions embedded in the factories and equipment used to produce the exported goods. The results in Table 2 are derived from an input-output model that treats capital investment as an input to production (see SI for 9149
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology
POLICY ANALYSIS
methodological details). We find that a majority of emissions embodied in capital investment are utilized for domestic household and governmental consumption (35 49% and 19 36%, respectively) with smaller amounts for the production of exports (21 31%), Figure 5 (Table S9, SI). This can be explained by an examination of the different types of capital flows in different years, which tend to be concentrated in domestic sectors such as electricity generation and coal mining (4 20% of total RMB invested between years), public administration (3 8%), education and health care (3 5%), road and rail transport (5 13%), and telecommunications (2 19%). Smaller amounts have been invested in export sectors in different years, though the sum of these small values does add up to appreciable amounts in total, as seen in Table 2. This means the major share of capital investments are induced by domestic final demand. Figure 5 shows how the closure of model for capital investment changes the overall perception of Chinese emissions, contrasting with Figure S2 (SI) where capital is treated as a final demand (like in the rest of the analysis presented). Household consumption clearly becomes more important, responsible for nearly half of emissions in all years. In the case of government consumption capital induced CO2 emissions make the larger share compared to emissions not induced by capital investments. Export related CO2 emissions still grow fast, but the growth is less pronounced in relative terms compared to a situation where capital is treated as final demand. Exports are now responsible for 37% of Chinese emissions (2.5 Gt CO2) in 2007, a large increase from only 27% and 28% in 1997 and 2002, respectively. Table 2. Capital Investment Emissions Re-Allocated to Final Demand Categories through a Partial Closure of Input-Output Model (Mt CO2 and % of Total Yearly Capital Emissions) households
government
exports
1992
347 (42%)
295 (36%)
178 (22%)
1997
609 (49%)
240 (19%)
387 (31%)
2002
462 (35%)
572 (44%)
270 (21%)
2007
1320 (49%)
627 (23%)
742 (28%)
’ DISCUSSION China’s annual CO2 emissions have grown by 3992 Mt between 1992 and 2007, but more than 70% of this emission growth has occurred between 2002 and 2007. Unlike many industrialized countries, China’s emission growth is mainly accruing in capital investment (47%) and exports (33%), rather than household (16%) and government consumption (3%). While previous analyses9,13 highlighted exports to explain more than 50% of the sharp emission increase between 2002 and 2005, our new data shows that capital investment seem to have taken the lead in this emission race, being responsible for 61% of the emission growth between 2005 and 2007 compared to only 18% in the case of exports. CO2 emission growth in China is no longer the sole result of a race between efficiency gains and rapidly growing consumption:23 while efficiency savings (f) are catching up with consumption growth (yl) in terms of CO2 emissions, changes in the production structure (L) have emerged as the third major emission driver. This more recent development can fully account for the sharp increase in China’s emission growth between 2002 and 2007. It is also the major driver behind the observed increase in the CO2 intensity of GDP over this period due to the growing dominance of the carbon intensive supply chain of construction in China’s economy. Hence, with the recent shift of China’s economic growth away from export (external) demand and toward capital investments, production is shifting toward energy and pollution intensive electricity, iron, cement and steel production and away from light industries like toys and electric appliances. Even though we find a large range in our results across equally acceptable decomposition forms, the observed effects remain prominent in all cases and are consistent with observed economic trends. We provide an extensive discussion of result sensitivities in the SI and also address potential dependency problems associated with SDA.28 Production structure changes (L) as a significant emission driver in China were first observed by Guan et al.9 for the period 2002 to 2005, but not further analyzed. Since then this effect has considerably gained in strength according to our model. Furthermore, while Guan et al.9 argue that fast growing exports as the main driver of these production structure changes, our data
Figure 5. CO2 Emissions in China allocated to final demands for the period 1992 2007 after capital closure of the model. The bracketed term ‘capital’ indicates the emission component induced in the supply chain by capital investments. The figure highlights that the majority of capital induced CO2 emissions are for domestic and not export consumption. 9150
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology suggests that more than 60% of the changes are associated with growing capital investments and the rapid expansion of the construction industry and its supply chain in recent years. Exports are, in fact, the smallest individual contributor across final demand categories (households, government, capital, export). Zhang4 does not find changes in the production structure to be an important emission determinant for the period 1991 to 2006. The analysis by Zhang for the years 2003 2006 is based on annual input-output data derived by updating the input-output table for the year 2002 based on a RAS methodology.25 Not finding structural changes to be an important emission driver for the post 2002 period may point toward the potential risk of missing important trends when analysis for fast growing economies such as China is based on nonsurvey tables. The energy and carbon intensive nature of capital investment might be hard to avoid. China as an emerging economy is currently building up its infrastructure. The high levels of CO2 emissions from capital investment might therefore only be of temporary nature. This is particularly true if the immense buildup in stocks of recyclable energy-intensive goods like steel increase the proportion of recycled goods in future years.59 If these investments meet long-term environmental objectives, then short-term emission spikes may be justified in the longrun. At the same time these developments raise concerns about the creation of path dependencies through capacity extensions in key sectors of the construction supply chain and infrastructure lock-in (see also ref 60). For example, China currently adds more than 2 billion square meters of floor area in buildings every year—faster than anywhere else in the world. While more stringent building standards have been put in place by the government,61 large parts of the new building stock put in place at the moment may not be constructed in an energy efficient way62 nor for a long lifetime. The resulting lock-in effects are largely avoidable because energy-efficient new-built can be created at negative or little extra costs (e.g., ref 63). Similarly, strict land-use policies and transport infrastructure development, particularly in urban areas, are economically sensible and directly benefit the quality of life of millions, because unnecessary pollution—including CO 2 emissions—is avoided. 64,65 CO2 emissions from the production of exports in China are arguably more challenging to address as they are intertwined in the consumption patterns of other countries. First, China has become the largest exporter in the world. Regular monitoring of international trade related emissions is needed to quantify if the CO2 emission growth in China offsets reductions in developed country emissions, as this could limit the effectiveness of the international climate change regime as established under the United Nations Framework Convention on Climate Change.66 68 Second, given China’s current role as the world’s manufacturing powerhouse, the idea of “exports reliefs” has been introduced into international climate change discussions, that is, continued transfer of emissions from developed to developing countries could be compensated with more stringent targets in developed countries.69,70 Among others this would require a discussion on how export related CO2 emission should be calculated—an issue that has been largely neglected so far. In this analysis we show that export related CO2 emission are 43% higher (742 Mt), if the capital requirements in the domestic supply chain are taken into account. Along similar lines, Dietzenbacher et al.71 identifies a non-negligible bias associated with the treatment of processing trade, which could lead to a considerable overestimation (38% reported in study) of export related CO2 emissions. Third, border tax adjustments (BTAs) on Chinese
POLICY ANALYSIS
exports or similar trade sanctions could be used to incentivize a low carbon transition in China. While BTAs are often proposed for carbon intensive products such as iron and steel, our analysis shows that such a focus on a few carbon intensive products could be of limited effectiveness:72 despite a recent growth of emissions from carbon intensive exports, a large emission share is still triggered by a high number of less carbon intensive goods exported such as computers, toys etc., which might not be affected by such a trade measure. A careful evaluation of such measures and their effects on CO2 emissions needs to be undertaken and balanced against the potential misuse by importing countries for protecting domestic industries from competition. There is still a widespread belief that government action to reduce China’s global, regional and local environmental impacts should focus on household related issues such as encouraging greener purchasing decisions or decreasing household formation rates.73 While this overlooks the important role of capital investments and exports as drivers of growth and emissions in China,74 our analysis provides evidence that in relation to households there are more pressing issues: China is urbanizing at a considerable pace and this process is a more important emission determinant than socio-demographic pressures such as household formation rates or population growth. The net emission increases associated with a transition from a rural to an urban lifestyle offsets the carbon saving potential of urban life.56,58,75 While consumption growth still remains the most important emission driver, it is deeply connected with urbanization. First, people are changing from a low consuming to a high consuming lifestyle due to increased incomes. Second, given the rapidly growing inequalities between rural and urban consumption, this effect is strengthening over time. A detailed analysis of the effects of growing inequalities on emissions and quality of life in China remains an interesting research avenue for the future given the growing inequalities in Chinese society.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional technical details about the model, data, and methodology; additional tables and figures for other decomposition factors; sensitivity analysis of methodology. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank four anonymous reviewers for useful comments, which have helped to improve the quality of the manuscript. We are further indebted to Jan Steckel at the Potsdam Institute for Climate Impact Research and the staff at the Department for the Economics of Climate Change at Technische Universit€at Berlin, in particular Felix Creutzig, for comments on and fruitful discussions of earlier versions of the manuscript. Any mistakes remain the sole responsibility of the authors. ’ REFERENCES (1) International Energy Agency. World Energy Outlook 2010; International Energy Agency: Paris, 2010. 9151
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology (2) Gregg, J. S.; Andres, R. J.; Marland, G. China: Emissions pattern of the world leader in CO2 emissions from fossil fuel consumption and cement production. Geophys. Res. Lett. 2008, 35 (8), L08806. (3) International Energy Agency. World Energy Outlook 2006; International Energy Agency: Paris, 2006. (4) Zhang, Y. Structural decomposition analysis of sources of decarbonizing economic development in China; 1992 2006. Ecol. Econ. 2009, 68 (8 9), 2399–2405. (5) Lin, J.; Zhou, N.; Levine, M.; Fridley, D. Taking out 1 billion tons of CO2: The magic of China’s 11th Five-Year Plan? Energy Policy 2008, 36 (3), 954–970 (6) Donglan, Z.; Dequn, Z.; Peng, Z. Driving forces of residential CO2 emissions in urban and rural China: An index decomposition analysis. Energy Policy 2010, 38 (7), 3377–3383. (7) Zhang, M.; Mu, H.; Ning, Y.; Song, Y. Decomposition of energyrelated CO2 emission over 1991 2006 in China. Ecol. Econ. 2009, 68 (7), 2122–2128. (8) Steckel, J. C.; Jakob, M.; Marschinski, R.; Luderer, G. From carbonization to decarbonization? Past trends and future scenarios for China’s CO2 emissions. Energy Policy 2011, 39 (6), 3443–3455. (9) Guan, D.; Peters, G. P.; Weber, C. L.; Hubacek, K. Journey to world top emitter: An analysis of the driving forces of China’s recent CO2 surge. Geophys. Res. Lett. 2009, 36, 1–5. (10) Chen, G. Q.; Chen, Z. M. Carbon emissions and resources use by Chinese economy 2007: A 135-sector inventory and input-output embodiment. Commun. Nonlinear Sci. Numer. Simul. 2010, 15 (11), 3647–3732. (11) Chen, G. Q.; Zhang, B. Greenhouse gas emissions in China 2007: Inventory and input-output analysis. Energy Policy 2010, 38 (10), 6180–6193. (12) Zhang, B.; Chen, G. Q. Methane emissions by Chinese economy: Inventory and embodiment analysis. Energy Policy 2010, 38 (8), 4304–4316. (13) Weber, C. L.; Peters, G. P.; Guan, D.; Hubacek, K. The contribution of Chinese exports to climate change. Energy Policy 2008, 36 (9), 3572–3577. (14) Su, B.; Huang, H. C.; Ang, B. W.; Zhou, P. Input-output analysis of CO2 emissions embodied in trade: The effects of sector aggregation. Energy Econ. 2010, 32 (1), 166–175. (15) Guo, J.; Zou, L.-L.; Wei, Y.-M. Impact of inter-sectoral trade on national and global CO2 emissions: An empirical analysis of China and US. Energy Policy 2010, 38 (3), 1389–1397. (16) Yunfeng, Y.; Laike, Y. China’s foreign trade and climate change: A case study of CO2 emissions. Energy Policy 2010, 38 (1), 350–356. (17) Liu, X.; Ishikawa, M.; Wang, C.; Dong, Y.; Liu, W. Analyses of CO2 emissions embodied in Japan-China trade. Energy Policy 2010, 38 (3), 1510–1518. (18) Guan, D.; Hubacek, K. China can offer domestic emission capand-trade in post 2012. Environ. Sci. Technol. 2010, 44 (14), 5327–5327. (19) Su, B.; Ang, B. W. Input-output analysis of CO2 emissions embodied in trade: The effects of spatial aggregation. Ecological Economics 2010, 70 (1), 10–18. (20) Arvesen, A.; Liu, J.; Hertwich, E. G. Energy Cost of Living and Associated Pollution for Beijing Residents. J. Ind. Ecol. 2010, 14 (6), 890–901. (21) Chang, Y.; Ries, R. J.; Wang, Y. The embodied energy and environmental emissions of construction projects in China: An economic input-output LCA model. Energy Policy 2010, 38 (11), 6597–6603. (22) Chen, G. Q.; Chen, H.; Chen, Z. M.; Zhang, B.; Shao, L.; Guo, S.; Zhou, S. Y.; Jiang, M. M. Low-carbon building assessment and multiscale input-output analysis. Commun. Nonlinear Sci. Numer. Simul. 2011, 16 (1), 583–595. (23) Peters, G. P.; Weber, C. L.; Guan, D.; Hubacek, K. China’s Growing CO2 Emissions-A Race between Increasing Consumption and Efficiency Gains. Environ. Sci. Technol. 2007, 41 (17), 5939–5944. (24) Leontief, W. Input-Output Economics; Oxford University Press: New York, NY, 1986. (25) Miller, R. E.; Blair, P. D. Input Output Analysis: Foundations and Extensions, 2nd ed. ed.; Cambridge University Press: Cambridge, 2009.
POLICY ANALYSIS
(26) Hoekstra, R.; van den Bergh, J. C. J. M. Structural decomposition analysis of physical flows in the economy. Environ. Resour. Econ. 2002, 23, 357–378. (27) Hoekstra, R.; van den Bergh, J. C. J. M. Comparing structural and index decomposition analysis. Energy Economics 2003, 25 (1), 39–64. (28) Dietzenbacher, E.; Los, B. Structural decomposition techniques: sense and sensitivity. Econ. Syst. Res. 1998, 10 (4), 307–323. (29) Seibel, S. Decomposition Analysis of Carbon-Dioxide Changes— Conceptual Framework and Empirical Results; Eurostat: Luxembourg, 2003. (30) Baiocchi, G.; Minx, J. C. Understanding Changes in the UK’s CO2 Emissions: A Global Perspective. Environ. Sci. Technol. 2010, 44 (4), 1177–1184. (31) National Bureau of Statistics. 1992 Input-Output Table of China; Statistics Press: Beijing, 1996. (32) National Bureau of Statistics. 1997 Input-Output Table of China; Statistical Press: Beijing, 1999. (33) National Bureau of Statistics. 2002 Input-Output Table of China; Statistics Press: Beijing, 2006. (34) National Bureau of Statistics. 2007 Input-Output Table of China; Statistical Press: Beijing, 2009. (35) United Nations Department for Economic and Social Affairs Statistics Division. Handbook of Input-Output Table Compilation and Analysis; United Nations: New York, 1999. (36) National Bureau of Statistics. China Statistical Yearbooks 1993 2008; Statistical Press: Beijing, 1993 2008. (37) Peters, G. P.; Weber, C. L.; Liu, J. Construction of a Chinese Energy and Emissions Inventory; Norwegian University of Science and Technology: Trondheim, 2006. (38) National Bureau of Statistics. Energy Statistical Yerabook of China; Statistical Press: Beijing, 2009. (39) Lenzen, M.; Treloar, G. Endogenising capital: a comparison of two methods. J. Appl. Input-Output Anal. 2005, 10, 1–11. (40) National Bureau of Statistics. China Statistical Yearbook 2008; Statistical Press: Beijing, 2009. (41) NDRC. Overview of the 11th Five Year Plan for National Economic and Social Development; National Development and Reform Commission (NDRC): Beijing, 2006. (42) Guan, D.; Hubacek, K.; Weber, C. L.; Peters, G. P.; Reiner, D. M. The drivers of Chinese CO2 emissions from 1980 2030. Global Environ. Change 2008, 18, 626–634. (43) Analysis and Forecast of National Electricity Suppy & Demand and Economic Situation 2008-2009; China Electrity Council, 2009. (44) Minx, J. C.; Wiedmann, T.; Wood, R.; Peters, G.; Lenzen, M.; Owen, A.; Scott, K.; Barrett, J.; Hubacek, K.; Baiocchi, G.; Paul, A.; Dawkins, E.; Briggs, J.; Guan, D.; Suh, S.; Ackerman, F. Input-output analysis and carbon footprinting: An overview of UK applications. Econ. Syst. Res. 2009, 21 (3), 187–216. (45) Wiedmann, T.; Wood, R.; Minx, J. C.; Lenzen, M.; Guan, D.; Harris, R. A carbon footprint time series of the UK—Results from a multi-region input-output model. Econ. Syst. Res. 2010, 22 (1), 19–42. (46) Weber, C. L.; Matthews, H. S. Quantifying the global and distributional aspects of American household carbon footprint. Ecol. Econ. 2008, 66 (2 3), 379–391. (47) Sanchez-Ch oliz, J.; Duarte, R.; Mainar, A. Environmental impact of household activity in Spain. Ecol. Econ. 2007, 62 (2), 308–318. (48) Kondo, Y.; Moriguchi, Y. CO2 emissions in Japan: Influences of imports and exports. Applied Energy 1998, 59 (2 3), 163–174. (49) Hertwich, E. G.; Peters, G. P. Carbon footprint of nations: A global, trade-linked analysis. Environ. Sci. Technol. 2009, 43 (16), 6414–6420. (50) Wood, R.; Dey, C. J. Australia’s carbon footprint. Econ. Syst. Res. 2009, 21 (3), 243–266. (51) Hertwich, E. G. Life cycle approaches to sustainable consumption: A critical review. Environ. Sci. Technol. 2005, 39 (13), 4673– 4684. (52) Baiocchi, G.; Minx, J. C.; Hubacek, K. The impact of social factors and consumer behavior on CO2 emissions in the UK: A panel regression based on input-output and geo-demographic consumer segmentation data. J. Ind. Ecol. 2010, 14 (1), 50–72. 9152
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
Environmental Science & Technology
POLICY ANALYSIS
(53) Hubacek, K.; Guan, D.; Barrett, J.; Wiedmann, T. Environmental implications of urbanization and lifestyle change in China: Ecological and water footprints. J. Clean. Prod. 2009, 17 (14), 1241–1248. (54) Chen, S.; Ravallion, M. Absolute poverty measures for the developing world, 1981-2004. Proceedings of the Nat. Acad. Sci. 2007, 104 (43), 16757–16762. (55) Ravallion, M.; Chen, S. China’s (uneven) progress against poverty. J. Dev. Econ. 2007, 82 (1), 1–42. (56) Newman, P. The environmental impact of cities. Environ. Urbanization 2006, 18 (2), 275–295. (57) Rees, W. E. Ecological footprints and appropriated carrying capacity: what urban economics leaves out. Environ. Urbanization 1992, 4 (2), 121–130. (58) Brunner, P. H. Reshaping urban metabolism. J. Ind. Ecol. 2007, 11 (2), 11–13. (59) Wang, T.; M€uller, D. B.; Graedel, T. E. Forging the anthropogenic iron cycle. Environ. Sci. Technol. 2007, 41 (14), 5120–5129. (60) Davis, S. J.; Caldeira, K.; Matthews, H. D. Future CO2 emissions and climate change from existing energy infrastructure. Science 2010, 329 (5997), 1330–1333. (61) ZhongXiang, Z. Is it fair to treat China as a Christmas tree to hang everybody's complaints? Putting its own energy saving into perspective. Energy Econ. 2010, 32 (Supplement 1), S47 S56. (62) Li, J.; Colombier, M.; Graud, P.-N. Grappling with climate change in the built environment in China. J. Energy Eng. 2010, 27, 27–31. (63) Pathways to a Low Carbon Economy: Version 2 of the Global Greenhouse Gas Abatement Cost Curves; McKinsey & Company: New York, 2009. (64) Creutzig, F.; Thomas, A.; Kammen, D. M.; Deakin, E., Transport Demand Management in Beijing, China: Progress and Challenges. In Low Carbon Transport in Asia: Capturing Climate and Development Co-Benefits; Zusman, E., Srinivasan, A., Dhakal, S., Eds.; Earthscan: London, 2011. (65) Creutzig, F. Climate change mitigation co-benefits of feasible transport demand policies in Beijing. Transp. Res. D 2009, 14, 120–131. (66) Peters, G.; Hertwich, E. CO2 Embodied in international trade with implications for global climate change policy. Environ. Sci. Technol. 2008, 42 (5), 1401–1407. (67) Davis, S. J.; Caldeira, K. Consumption-based accounting of CO2 emissions. Proc. Natl. Acad. Sci. U. S. A. 2010, 107 (12), 5687–5692. (68) Peters, G. P.; Minx, J. C.; Weber, C. L.; Edenhofer, O. Growth in emission transfers via international trade from 1990 to 2008. Proc. Natl. Acad. Sci. 2011, 108 (21), 8903–8908. (69) BBC News China seeks export carbon relief http://news.bbc. co.uk/1/hi/7947438.stm (accessed January 30, 2011), (70) Peters; Edgar G. Hertwich, G. P., Trading Kyoto. 2008, (0804). (71) Dietzenbacher, E.; Pei, J.; Yang, C., The Environmental Pains and Economics Gains of Outsourcing to China. In 17th Conference of the International Input-Output Association, Sao Paulo, 2009. (72) Sinden, G.; Peters, G.; Minx, J. C.; Weber, C. International flows of embodied CO2 with an application to aluminium and the EU ETS. Clim. Policy 2011, 11 (5), 1226–1245. (73) Liu, J. China’s road to sustainability. Science 2010, 328 (5974), 50. (74) Peters, G. P.; Guan, D.; Hubacek, K.; Minx, J. C.; Weber, C. Effects of China's Economic Growth. Science 2010, 328 (5980), 824–825. (75) Kenworthy, J. R.; Laube, F. B. Automobile dependence in cities: an international comparison of urban transport and land use patterns with implications for sustainability. Environ. Impact Assess. Rev. 1996, 16 (4 6), 279–308.
9153
dx.doi.org/10.1021/es201497m |Environ. Sci. Technol. 2011, 45, 9144–9153
POLICY ANALYSIS pubs.acs.org/est
Are Reductions in Industrial Organic Contaminants Emissions in Rich Countries Achieved Partly by Export of Toxic Wastes? Knut Breivik,*,†,‡ Rosalinda Gioia,§ Paromita Chakraborty,|| Gan Zhang,|| and Kevin C. Jones§,|| †
Norwegian Institute for Air Research, P.O. Box 100, NO-2027 Kjeller, Norway Department of Chemistry, University of Oslo, P.O. Box 1033, NO-0315 Oslo, Norway § Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, U.K. Guangzhou Institute of Geochemistry, The Chinese Academy of Sciences, Guangzhou 510640, China
)
‡
ABSTRACT: Recent studies show that PCB (polychlorinated biphenyl) air concentrations remain surprisingly high in parts of Africa and Asia. These are regions where PCBs were never extensively used, but which are implicated as recipients of obsolete products and wastes containing PCBs and other industrial organic contaminants, such as halogenated flame retardants (HFRs). We hypothesize that there may be different trends in emissions across the globe, whereby emissions of some industrial organic contaminants may be decreasing faster in former use regions (due to emission reductions combined with uncontrolled export), at the expense of regions receiving these substances as obsolete products and wastes. We conclude that the potential for detrimental effects on the environment and human health due to long-range transport by air, water, or wastes should be of equal concern when managing and regulating industrial organic contaminants. This calls for a better integration of life-cycle approaches in the management and regulation of industrial organic contaminants in order to protect environmental and human health on a global scale. Yet, little remains known about the amounts of industrial organic contaminants exported outside former use regions as different types of wastes because of the often illicit nature of these operations.
’ INTRODUCTION PCBs are industrial organic contaminants identified as toxic, bioaccumulative, persistent, and subject to long-range transport (LRT) with global scale significance. PCBs and other persistent organic pollutants (POPs) fulfilling these criteria are now regulated under the Stockholm Convention in order to protect environmental and human health from these harmful substances.1 PCBs may additionally be formed by de novo synthesis in various combustion processes, but these emissions are assumed less significant with respect to the overall global mass balance of these chemicals.2 4 Global production of PCBs peaked around 1970 and ceased in 1993 when the last factory closed in Russia. About half of the known historical production of PCBs took place at factories in the United States (∼48%), followed by Europe (∼33%), Russia (∼13%), Japan (∼5%) and China (<1%), while ∼97% of the cumulative historical usage of PCBs took place in the northern hemisphere, mainly at midlatitudes.5 Consequently, the existing paradigm has been that the major emission regions of PCBs have been industrialized countries in the northern hemisphere6 where these chemicals were mainly manufactured and used,5 and where significant reductions in atmospheric concentrations are now typically observed.7 10 Recent studies are reporting overall halving times in air from about 5 years in the U.K.9 and up to 17 years in the Great Lakes region.7 Almost 70% of the PCBs produced were used in electrical equipment (capacitors and transformers) with typical life-spans from 10 to 30 years, which implies significant delays in r 2011 American Chemical Society
disposal time trends as compared to global production trends.5 PCBs are now approaching the end of their life-cycle in the “anthroposphere” (man-made products). A key question now is whether emission reductions seen in industrialized regions will be followed by significant and immediate reductions in environmental burdens even in remote areas, such as the Arctic. This is also a difficult question to answer as several factors have the potential to mitigate the effect of primary emission reductions on atmospheric concentrations of PCBs. One is the buffering effect caused by secondary emissions.11 13 Another hot topic has been the potential impacts of climate change, leading to possible enhanced volatilization of PCBs from both primary and secondary sources.14 However, it is also prudent to ask whether there remain any significant primary sources and source regions on a global scale which are yet to be identified and controlled. Almost two decades ago Iwata and co-workers6 suggested that the observed pattern of PCBs in air and seawater was likely indicating reduction of PCBs in highly contaminated areas in developed nations along with the expansion of PCBs toward tropical regions. Here we discuss more recent studies which demonstrate major atmospheric emissions of PCBs and other industrial organic contaminants associated with the end-of-life-cycle, often Received: July 6, 2011 Accepted: September 29, 2011 Revised: September 26, 2011 Published: September 29, 2011 9154
dx.doi.org/10.1021/es202320c | Environ. Sci. Technol. 2011, 45, 9154–9160
Environmental Science & Technology
Figure 1. Concentrations of Σ7PCBs (in log pg/m3) in ambient air at e-waste sites in China,22,28,29 major cities9,33 36 and European background sites.38 Data for Σ7PCBs in Guangzhou35 are estimated as detailed in the text.
outside regions of former production and major use. We then discuss whether there has been a shift in primary emission regions of PCBs on a global scale, paralleled by significant reductions in atmospheric burdens within former use regions. We close by identifying and discussing some of the key research needs and potential implications for global control strategies of industrial organic contaminants.
’ MAJOR EMISSIONS FROM RECYCLING AND WASTES The often illegal export of obsolete products and wastes containing PCBs from former use regions has been occurring for decades (e.g., 15,16). Recognition of a toxic trade with hazardous substances led to the adoption of the Basel Convention which aims to control the trans-boundary movement of hazardous wastes and their disposal, and which entered into force in 1992. However, the impacts in terms of elevated atmospheric emissions at recycling and waste sites in subtropical and tropical regions have only more recently been studied. Evidence for major atmospheric emissions of PCBs associated with the end-of-lifecycle of these chemicals is now well documented from elevated concentrations of PCBs observed at e-waste recycling sites in China (e.g., 17,18). Although various definitions of e-waste exist19,20 it is used here in broad terms as any discarded electrical and electronic equipment, although not all e-waste will contain PCBs. Thus, the composition of e-waste and other wastes containing PCBs could be expected to vary significantly from site to site. Procedures used in the recycling of e-waste at these sites are considered primitive without adequate measures of protecting environmental and human health18,21 as recently reviewed by Tsydenova and Bengtsson.21 Techniques involve melting and open burning of the e-waste to recover precious metals, but inevitably also make PCBs and other semivolatile organic substances prone to volatilization. In Figure 1, we compare recent studies reporting concentrations of Σ7PCBs (28, 52, 101, 118, 138, 153, and 180) in air from Chinese e-waste sites and contrast these with measurements from urban source regions, both within and outside former use regions. Xing et al.22 report data for Σ7PCBs from Guiyu, a major e-waste recycling center in the Guangdong Province in southeastern China, based on 18 samples of which 6 were collected from an open burning site, 10 from residential areas, and 2 control samples (local background) in late April 2006. The total air concentrations
POLICY ANALYSIS
(gas + particulate) of Σ7PCBs from these 3 sites were 98.52 ( 96.28 ng/m3, 1.45 ( 1.16 ng/m3, and 0.93 ( 0.13 ng/m3, respectively. The extremely high values of PCBs observed at the open burning site indicate significant potential for emissions from open burning, in line with past findings. 23 Indeed, open burning of waste containing PCBs has also previously been implicated as a key release pathway of PCBs to the atmosphere.4,24 27 The elevated emissions of PCBs associated with open burning are attributed to a combination of elevated temperatures, limited turbulence, and short residence time, indicating that PCBs are literally escaping the flames through volatilization.26,27 From extrapolation of experimental data, the relative amounts emitted into air have been estimated to range from 10%4,26 up to 20%.25 Elevated air concentrations of PCBs have also been reported from Taizhou in 200528 and 2007,29 an e-waste dismantling area in Zhejiang province of southeast China where transformers and capacitors had been recycled for two decades,30 and where mechanical shredding is a common recycling technique employed at the workshops.31 In October 2005, Li et al.28 measured an average concentration of Σ7PCBs of 7.1 ng/m3 (range 4.2 11.2 ng/m3) in 12 samples from 3 sites. In January 2007, Han et al.29 reported average air concentrations of Σ7PCBs of 5.64 ng/ m3 (range 2.6 15.0 ng/m3) near an e-waste dismantling area, which was more than 50 times higher than their reference urban site. According to Han and co-workers PCB air concentrations have declined in recent years, attributed to a ban by the local government on dismantling of products containing PCBs.29,32 To put the recent studies from e-waste sites in China into context, we have included a comparison with data from major cities in Asia, the United States, and Europe (U.K.), along with European background air in Figure 1. All samples were collected within the same recent 5-year period (2004 2008) in an attempt to provide a snapshot. The mean air concentrations at the Chinese e-waste sites were found to be consistently higher than in selected major cities in India (Delhi in 2006 2007;33 Mumbai, Bangalore and Kolkata in 2006 34 ), China (Guangzhou) in 2004,35 U.S. (Chicago) in 2006 2007,36 and U.K. (London, Manchester) in 2005 2008.9 Dump sites,33 e-waste, and ship-breaking activities34 have previously been implicated as potential sources for the elevated air concentrations of PCBs occasionally measured in India using passive air samplers (PAS). The mean air concentrations of Σ7PCBs over a 2-year period at two agricultural sites near Delhi were 547 pg/m3 (range 166 1364 pg/m3) and 523 pg/m3 (range 193 1158 pg/m3) (N = 10 at each site) according to Pozo et al.,33 while air concentrations of Σ7PCBs in Mumbai, Bangalore, and Kolkata of 253, 243, and 239 pg/m3 have been reported by Zhang et al.34 (Figure 1). According to Zhang et al.,34 India is the world’s largest ship-breaking nation in terms of volume and Mumbai is among the areas where this takes place. India is also generating ∼150 000 tons of e-waste per year, with Mumbai, Delhi, and Bangalore producing ∼7%, ∼6%, and ∼3% of this amount, respectively. Elevated concentrations of PCBs in human breast milk have furthermore been reported from mothers living close to an open dumping site in Kolkata.37 According to Widmer et al.,19 Mumbai and Delhi additionally have known e-waste recycling sites, and such recycling is suspected to take place in Guangzhou as well. The air concentration of Σ7PCBs in Guangzhou (519 pg/m3, range 94 1510 pg/m3) as included in Figure 1 was estimated from data on Σ64PCBs across Guangzhou (935 pg/m3, range 170 2720 pg/m3, N = 32)35 by first assuming 9155
dx.doi.org/10.1021/es202320c |Environ. Sci. Technol. 2011, 45, 9154–9160
Environmental Science & Technology that Σ64PCBs contributes 31% (100 64/209) of Σ209PCBs. Second, the relative contribution of Σ7PCBs to Σ209PCBs was assumed to be 17%. The latter estimate was derived from an inventory of total historical production of Σ7PCBs which made up ∼18% of the total production of Σ209PCBs5 and empirical air measurements from a comprehensive study of Σ209PCBs across 37 sites in Chicago which reports a mean contribution of ∼16% for Σ7PCBs.36 The mean air concentration of Σ7PCBs across Chicago (133 pg/m3, range 12 873 pg/m3, N = 184)36 is rather similar to the mean air concentrations of Σ7PCBs in London (94 pg/m3; range 25 251 pg/m3, N = 16) and Manchester (90 pg/m3, range 31 228 pg/m3, N = 16) derived from quarterly samples.9 In comparison, the mean concentration of Σ7PCBs in air at 86 European background sites across 34 different European countries in 2006 derived on the basis of PAS was only 17 pg/m3 (range 2 121 pg/m3).38 The difference in air concentrations recorded in these Asian cities, relative to major cities in U.K. and U.S. are particularly noteworthy, given that the estimated cumulative use of PCBs per capita in China, India, U.K., and U.S. is ∼0.007, ∼0.02, ∼0.5, and ∼2 kg/capita.5 However, we believe the main reason for the elevated concentrations in Asian cities could be that China and India both have known e-waste recycling sites and main ports where e-waste is received and dispatched.19
’ GLOBAL MOVEMENT OF WASTES CONTAINING PCBS E-waste is certainly not only a problem for China and India, but considered to be representative of a global issue.17,19 Most of the e-waste produced globally is from Europe, the U.S., and Australasia.20 The total annual e-waste productions in EU-27 and the U.S. have been estimated to be 8.3 9.1 and 2.6 million tonnes, respectively.20 With higher labor costs and tougher environmental regulations in these regions, a large proportion of wastes are not processed in the source country but rather exported elsewhere.20 For example, it has previously been estimated that 50 80% of e-waste collected in the U.S. for recycling is exported to developing countries.39,40 Because the U.S. has not ratified the Basel Convention, this may not necessarily be illegal.39,40 Reuse is also highlighted as a source of electronic equipment to developing countries which accept donations of obsolete equipment from rich countries, but may end up as e-waste in recipient countries.20 China receives about 70% of all e-waste being exported,20 and is claimed to be the world’s largest importer and recycler of e-waste.17,41 Ni et al.17 identify two key reasons why the magnitude of the problem has become particularly large in China. Following rapid economic developments in domestic China, there is an increasing demand for electronic equipment which inevitably also leads to large amounts of domestic e-waste. Second, illegal import of e-waste is also a major issue in China. As the Chinese government is trying to reduce import of e-waste,18,41 it is expected that substantial amounts will end up in other countries in the region.18 Significant amounts of e-waste are also exported to other countries in Asia (India, Pakistan, Vietnam, The Philippines, Malaysia) and African countries (e.g., Nigeria and Ghana).20,42 However, Robinson20 points out that the actual mass and amounts of e-waste being exported are impossible to quantify accurately because of the “semi-clandestine” nature of these operations. This also illustrates why trans-boundary movement of wastes has not been adequately addressed in global emission inventories of PCBs thus far.5 The global generation of e-waste is expected to increase in the future,17,18 and possibly more significantly in
POLICY ANALYSIS
Eastern Europe and Latin America as well as China, because of economic growth and technological developments in these areas.20 In addition to use of PCBs in electrical equipment, it is well established that these chemicals were additionally used for a wide range of nominally closed systems and open applications including (but not limited to) hydraulic and heat transfer fluids, paints, sealants, plasticizers, and carbonless copy paper.4 Thus, e-waste represents only one among several types of obsolete products and wastes with a potential for trans-boundary movement. For example, PCBs were used on ships in the past and concerns have been expressed about potential releases from ship dismantling activities in developing regions.43,44 Furthermore, PCBs were also extensively used in small capacitors in cars in the past11,16 which may ultimately be discarded in regions with more informal recycling and disposal procedures.
’ ELEVATED AIR CONCENTRATIONS OF PCBS IN AFRICA In parallel to the reductions in emissions of PCBs in use regions and the high levels recorded at e-waste sites in China, recent studies have also recorded surprisingly high levels of PCBs far away from use regions. Examples are the elevated air concentrations of PCBs which have been recorded off west Africa on cruises on board the research vessel R/V Polarstern in 2001,45 2005,46 and 2008.44 The highest values of Σ7PCBs were 200 pg/m3 in 200546 and 190 pg/m3 in 200844 in samples collected about 400 km from the coast of West Africa. The most recent study concluded that some of the measured air concentrations of PCBs, which were in line with major cities in U.S. and U.K. (Figure 1), were simply far too high to be rationalized by historical usage of PCBs in Africa.44 Nor could these levels be explained by emissions due to fires, based on satellite data and parallel air measurements of polycyclic aromatic hydrocarbons (PAHs) which are byproducts of incomplete combustion. The authors argued that the sources could be releases of PCBs from contaminated waste.44 Data on PCBs in air from the African continent still remain scarce. However, Klanova et al.47 reported monthly resolved air data for Σ7PCBs derived on the basis of PAS from 26 sites across 15 countries on the African continent in 2008. Samples that exceeded 20 ng Σ7PCBs per sample, corresponding to a monthly average air concentration of ∼100 Σ7PCBs pg/m3 or more, comparable to recent data from cities in U.K. and U.S. (Figure 1), were reported from mostly urban/industrial sites for about half of these countries (South Africa, Senegal, Kenya, Egypt, Democratic Republic of Congo, Ghana, Mali, and Sudan).47 Similar air concentrations in excess of 100 Σ7PCBs pg/m3 were also found at sites on the Ivory Coast and in Gambia in 2008.44 ’ GREATER POTENTIAL FOR EMISSIONS IN SUBTROPICAL AND TROPICAL REGIONS The elevated temperatures encountered in subtropical and tropical regions in southeastern Asia and Africa are very different from those in former use regions at mid-latitudes. This has profound implications for any attempt to assess atmospheric emissions of semivolatile organic contaminants as volatilization is expected to be a key process by which PCBs are emitted to air, which is strongly temperature dependent.4 For example, it is estimated in the context of climate change that an increase in air temperature of 1 °C will increase the volatility of compounds like PCBs by about 10 15%.14 For comparison, we have mapped the 9156
dx.doi.org/10.1021/es202320c |Environ. Sci. Technol. 2011, 45, 9154–9160
Environmental Science & Technology
Figure 2. Map of the annual average global volatility of PCB-153 (in Pa), illustrating the enhanced potential for volatilization of semivolatile organic contaminants in subtropical and tropical regions.
global volatility of one arbitrary indicator PCB (PCB-153) in Figure 2, as derived on the basis of long-term annual average surface air temperatures and physical chemical properties of PCBs from Li et al.48 Among the localities specifically included in Figure 1, the lowest volatility of PCB-153 is predicted for Manchester (∼0.0001 Pa). Relative to Manchester, the volatility of PCB-153 at the other sites would increase by 12% (London), 62% (Chicago), 280% (Taizhou), 499% (Guiyu), 526% (Guangzhou), 850% (Bangalore), 896% (Mumbai), 1041% (Kolkata), and 1329% (Delhi). The consequence of exporting semivolatile substances like PCBs toward subtropical and tropical regions in terms of enhancing the potential for atmospheric emissions through volatilization is therefore significant. Disturbingly, the areas showing elevated potential for volatilization of semivolatile organic substances largely coincide with many of the major e-waste countries and regions where informal recycling and disposal practices are believed to occur. Following possible recycling, most of the e-waste ultimately ends up in landfills and dumps.20 According to Weber et al.,49 many developing countries and countries with economies in transition rely heavily on uncontrolled landfilling, often combined with open burning. Clearly, the potential for atmospheric emissions and releases to other environmental media from such sites are substantial compared to contained and designated landfills in developed countries at mid-latitudes.24,49 51 The situation may be further exacerbated by the more limited storage capacity of environmental surface media within tropical regions in retaining these chemicals. For example, PCBs sorbed to soils with a low content of soil organic matter (SOM) in a hot area are predicted to be more prone to re-emissions than soils with a high content of SOM in a cold region.52
’ OTHER INDUSTRIAL ORGANIC CONTAMINANTS OF RELEVANCE Additional evidence exists for e-waste sites in China being major hot-spots for a much wider range of other contaminants (e.g., 17,18,21,53,54). Additional industrial organic contaminants often contained in e-waste are various halogenated flame retardants (HFRs), such as polybrominated diphenyl ethers (PBDEs), polybrominated biphenyls (PBBs), tetrabromobisphenol-A (TBBP-A),
POLICY ANALYSIS
polybrominated phenols (PBPs), and Dechlorane Plus (DP). Some of these HFRs are also being regulated/banned internationally; others may be in the future. Much of the discussion herein would thus apply for a wider range of industrial organic contaminants. For example, several studies have reported significantly elevated levels of PBDEs at e-waste sites,28,55,56 and we refer to the recent review by Tsydenova and Bengtsson21 for a similar comparison as presented in Figure 1. PBDEs are additive flame retardants used in textiles, plastics, and electrical and electronic equipment,57 and two commercial mixtures of PBDEs (octa-BDE and penta-BDE) were recently regulated under the Stockholm Convention on POPs. In 2001, global usage of PBDEs mainly occurred in America (49%), Asia (37%), and Europe (12%) with only minor use (2%) in other parts of the world.58 Consequently, there are similarities in the main areas of global production and use for both PCBs and PBDEs, albeit with a more significant influence from some Asian countries in the case of PBDEs and some other HFRs,58 presumably reflecting the rapid economic developments in parts of Asia which occurred between peak PCB and PBDE production. However, there are also significant differences in the timing of peak production, which may have occurred about three decades later in the case of PBDEs, and in the timing of various control measures. Additionally, there is an anticipated time lag between peak production and disposal reflected by the typical lifespan of products containing these chemicals.5,59 For PBDEs, the typical lifespan of products which will end up as e-waste, such as mobile phones, computers, and television sets, only range from about 2 to 5 years.20 Taken together, this suggests that the timing when significant amounts of PBDEs and PCBs are entering the end of their life-cycle in the “anthroposphere” may be much closer than expected from their temporal trends in production.
’ IMPLICATIONS FOR RESEARCH AND CONTROL STRATEGIES On the basis of the previous discussion, we hypothesize that there may be different trends in emissions across the globe, whereby emissions of PCBs and possibly also some HFRs may be decreasing faster in former use regions (due to emission reductions combined with uncontrolled export), at the expense of regions receiving these industrial organic contaminants as obsolete products and wastes. In the case of PCBs, this can be seen as a transition which started already decades ago.6 An intriguing feature associated with translocation of wastes to other areas is that it may coincide with transport from colder to warmer regions, which by itself would enhance the potential for emissions of semivolatile substances (“anti-cold-trapping effect”). Further research is therefore needed to understand the global transport and emissions associated with these chemicals as contained in obsolete products and wastes, as it has obvious and immediate implications for understanding occurrence, transport, and fate of these chemicals in the global environment. While significant efforts and achievements have been made by the scientific community to understand and predict LRT of such chemicals by air and water (e.g., 13,60 63) we caution that the global sources and fate of these chemicals still cannot be fully rationalized (nor controlled) without an understanding of emissions due to “LRT” by products and wastes. Although an extensive literature has emerged on e-waste in recent years, there is an urgent need for further studies to understand transboundary movement of other types of wastes containing these chemicals and to establish reliable inventories of industrial 9157
dx.doi.org/10.1021/es202320c |Environ. Sci. Technol. 2011, 45, 9154–9160
Environmental Science & Technology organic contaminants in developing regions and countries with economies in transition. We note that significant efforts are now invested to promote national inventories of PCBs under the Stockholm Convention through the PCB Elimination Network which may provide more data in the future. As indicated, our results may have important implications for studies aiming to interpret long-term trends of industrial organic contaminants in the environment and its food-chains. To further test our hypothesis of a potential shift in source regions, it would be informative to determine, compare, and contrast long-term time trends of industrial organic contaminants in various parts of the globe, discriminating between trends in regions believed to have been mainly affected by historical production/use versus recycling/wastes as well as remote areas subject to LRT by air and water, such as the Arctic. For example, it is worth noting that a study on human breast milk from Ghana now reports significant increases of both PCBs and PBDEs from 2004 to 2009, with levels of PCBs indicating a potential health risk for children.64 The authors attribute exposure from toxic wastes originating from industrial countries as a plausible explanation.64 This temporal trend is in striking contrast to long-term trends in Sweden, for which levels of PCBs and most PBDEs in human breast milk have declined since the 1970s65 and 1995,66 respectively. Among the critical research needs are thus further studies to identify major sources and source regions of PCBs outside former use regions where relevant information remains scarce, such as in West Africa. Spatially and temporally resolved air measurements combined with inverse modeling techniques may be one strategy to gain further insights into the location and magnitude of these emissions and to monitor future developments. Highly elevated air concentrations of HFRs have now recently also been reported on yet another cruise off the west coast of Africa;67 a plausible explanation suggested for the high levels recorded could be electronic waste exported from use regions like Europe and U.S. to Africa.67 It would therefore be informative to cover a wider range of industrial organic contaminants outside regions of major production and use to further assess whether sources and source regions associated with these industrial chemicals show similarities or not. There is clearly also a need for future studies in regions which are implicated as recipients of e-waste, but for which little or no data on environmental concentrations is available thus far. Similarly, one may anticipate a more rapid decline in air concentrations from urban regions in former use areas which have experienced significant emissions in the past, compared to remote sites within the same region which may be relatively more affected by sustained long-range atmospheric transport. There are clues in existing long-term monitoring data for PCBs from northwestern Europe that this could be occurring. The time for a 50% decline in air concentrations for Σ7PCBs averaged 4.7 ( 1.6 years for all congeners from six rural and urban monitoring sites in the U.K. from 1991 to 2008.9 In comparison, the amounts sequestered by PAS deployed along a latitudinal transect of background sites from U.K. to northern Norway only declined with the average 50% decline time of 8.4 ( 3.2 years from 1994 to 2008.10 The authors of this study argue that primary emissions continue to control the underlying trends observed at these remote sites.10 If relevant global international agreements like the Basel, Rotterdam, and Stockholm Convention on POPs are to be successful and efficient in protecting environmental and human health from these hazardous substances, rational control strategies should target the emission sources which are believed to be
POLICY ANALYSIS
most significant on a global scale. Even if only a small proportion of the total inventory of PCBs and other semivolatile organic contaminants reaches subtropical and tropical regions, the fraction actually emitted into the atmosphere could be significant. Thus, further efforts to mitigate export of obsolete products and wastes containing semivolatile organic contaminants from colder to warmer regions may by itself be a prudent measure to mitigate atmospheric emissions beyond the enhanced emissions that are expected from informal recycling and disposal strategies in the regions discussed. Although Article 6 of the Stockholm Convention on POPs includes provisions on how to deal with stockpiles and wastes, the results presented herein indicate that regulatory efforts may have been more oriented toward protecting environmental and human health in remote regions subject to LRT by air and water, rather than “LRT” by wastes. As the divergent trends of PCBs and PBDEs in human breast milk between Sweden65,66 and Ghana64 illustrate, regulations focused on eliminating or restricting production and use of POPs in industrialized regions alone may not be sufficient to protect environmental and human health on a global scale. In fact, regulating industrial POPs by introducing expensive control measures with regard to the management of obsolete products, stockpiles, and wastes may unintentionally have contributed to create a market for export of wastes from industrialized countries, which in turn may have resulted in the elevated emissions now observed outside regions of former production and major use. The potential for detrimental effects on environmental and human health due to LRT by air, water, wastes or by any other means should be of equal concern when managing and regulating POP-like chemicals. A better integration of life-cycle approaches/considerations in industrial organic contaminant management and regulation is therefore needed to protect environmental and human health from industrial POPs on a global scale. In a broader context, this study also inevitably leads to the more difficult question of how much of the reduction in environmental burdens of PCBs and other industrial organic contaminants seen in rich countries may have been due to uncontrolled export of toxic wastes to developing regions, rather than domestic emission reductions achieved through environmentally sound waste management.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (+47) 63 89 80 00; fax: (+47) 63 89 80 50; e-mail:
[email protected].
’ ACKNOWLEDGMENT We thank the Research Council of Norway (183437/S30) as well as the U.K. Department of Environment, Food and Rural Affairs for financial support. KCJ wants to thank the support from Chinese Academy of Sciences for the Visiting Professorship for Senior Scientists. We also thank Wenliang Han, Tom Harner, Karla Pozo, Ming H. Wong, Guanhua Xing, and Qinghua Zhang who provided us with data from studies in Asia, and Sabine Eckhardt for preparing Figure 2. ’ REFERENCES (1) UNEP. The Stockholm Convention on Persistent Organic Pollutants (POPs); 2001. (2) Tanabe, S.; Kannan, N.; Subramanian, A.; Watanabe, S.; Tatsukawa, R. Highly toxic coplanar PCBs - Occurence, source, 9158
dx.doi.org/10.1021/es202320c |Environ. Sci. Technol. 2011, 45, 9154–9160
Environmental Science & Technology persistency and toxic implications to wildlife and humans. Environ. Pollut. 1987, 47 (2), 147–163. (3) Takasuga, T.; Senthilkumar, K.; Matsumura, T.; Shiozaki, K.; Sakai, S. I. Isotope dilution analysis of polychlorinated biphenyls (PCBs) in transformer oil and global commercial PCB formulations by high resolution gas chromatography-high resolution mass spectrometry. Chemosphere 2006, 62 (3), 469–484. (4) Breivik, K.; Sweetman, A.; Pacyna, J. M.; Jones, K. C. Towards a global historical emission inventory for selected PCB congeners - a mass balance approach 2 Emissions. Sci. Total Environ. 2002, 290 (1 3), 199–224. (5) Breivik, K.; Sweetman, A.; Pacyna, J. M.; Jones, K. C. Towards a global historical emission inventory for selected PCB congeners - A mass balance approach 3 An update. Sci. Total Environ. 2007, 377 (2 3), 296–307. (6) Iwata, H.; Tanabe, S.; Sakal, N.; Tatsukawa, R. Distribution of persistent organochlorines in the oceanic and surface seawater and the role of ocean on their global transport and fate. Environ. Sci. Technol. 1993, 27 (6), 1080–1098. (7) Venier, M.; Hites, R. A. Time Trend Analysis of Atmospheric POPs Concentrations in the Great Lakes Region Since 1990. Environ. Sci. Technol. 2010, 44 (21), 8050–8055. (8) Sun, P.; Ilora; Basu; Blanchard, P.; Brice, K. A.; Hites, R. A. Temporal and spatial trends of atmospheric polychlorinated biphenyl concentrations near the Great Lakes. Environ. Sci. Technol. 2007, 41 (4), 1131–1136. (9) Schuster, J. K.; Gioia, R.; Sweetman, A. J.; Jones, K. C. Temporal Trends and Controlling Factors for Polychlorinated Biphenyls in the UK Atmosphere (1991 2008). Environ. Sci. Technol. 2010, 44 (21), 8068–8074. (10) Schuster, J. K.; Gioia, R.; Breivik, K.; Steinnes, E.; Scheringer, M.; Jones, K. C. Trends in European Background Air Reflect Reductions in Primary Emissions of PCBs and PBDEs. Environ. Sci. Technol. 2011, 44 (17), 6760–6766. (11) Harrad, S. J.; Sewart, A. P.; Alcock, R.; Boumphrey, R.; Burnett, V.; Duartedavidson, R.; Halsall, C.; Sanders, G.; Waterhouse, K.; Wild, S. R.; Jones, K. C. Polychlorinated Biphenyls (PCBs) in the British Environment - sinks, sources and temporal trends. Environ. Pollut. 1994, 85 (2), 131–146. (12) Li, Y. F.; Harner, T.; Liu, L. Y.; Zhang, Z.; Ren, N. Q.; Jia, H. L.; Ma, J. M.; Sverko, E. Polychlorinated Biphenyls in Global Air and Surface Soil: Distributions, Air-Soil Exchange, and Fractionation Effect. Environ. Sci. Technol. 2010, 44 (8), 2784–2790. (13) Wania, F.; Mackay, D. Global fractionation and cold condensation of low volatility organochlorine compounds in polar regions. Ambio 1993, 22 (1), 10–18. (14) UNEP/AMAP. Climate change and POPs: Predicting the impacts. Report of the UNEP/AMAP Expert Group; 2011. (15) Vir, A. K. Toxic trade with Africa. Environ. Sci. Technol. 1989, 23 (1), 23–25. (16) Cummins, J. E. Extinction: The PCB threat to marine mammals. Ecologist 1988, 18 (6), 193–195. (17) Ni, H. G.; Zeng, H.; Tao, S.; Zeng, E. Y. Environmental and human exposure to persistent halogenated compounds derived from e-waste in China. Environ. Toxicol. Chem. 2010, 29 (6), 1237– 1247. (18) Wong, M. H.; Wu, S. C.; Deng, W. J.; Yu, X. Z.; Luo, Q.; Leung, A. O. W.; Wong, C. S. C.; Luksemburg, W. J.; Wong, A. S. Export of toxic chemicals - A review of the case of uncontrolled electronic-waste recycling. Environ. Pollut. 2007, 149 (2), 131–140. (19) Widmer, R.; Oswald-Krapf, H.; Sinha-Khetriwal, D.; Schnellmann, M.; Boni, H. Global perspectives on e-waste. Environ. Impact Assess. Rev. 2005, 25 (5), 436–458. (20) Robinson, B. H. E-waste: An assessment of global production and environmental impacts. Sci. Total Environ. 2009, 408 (2), 183–191. (21) Tsydenova, O.; Bengtsson, M. Chemical hazards associated with treatment of waste electrical and electronic equipment. Waste Manage. 2011, 31 (1), 45–58.
POLICY ANALYSIS
(22) Xing, G. H.; Chan, J. K. Y.; Leung, A. O. W.; Wu, S. C.; Wong, M. H. Environmental impact and human exposure to PCBs in Guiyu, an electronic waste recycling site in China. Environ. Int. 2009, 35 (1), 76–82. (23) Gonzalez, M. J.; Jimenez, B.; Fernandez, M.; Hernandez, L. M. PCBs, PCDDs and PCDFs in soil samples from uncontrolled burning of waste electrical material for metal relamation. Toxicol. Environ. Chem. 1991, 33 (3 4), 169–179. (24) Ruokojarvi, P.; Ruuskanen, J.; Ettala, M.; Rahkonen, P.; Tarhanen, J. Formation of polyaromatic hydrocarbons and polychlorinated organiccompounds in municipal waste landfill fires. Chemosphere 1995, 31 (8), 3899–3908. (25) Nisbet, I.; Sarofim, A. F. Rates and routes of transport of PCBs in the environment. Environ. Health Perspect. 1972, 1, 21–38. (26) Ahling, B.; Lindskog, A. Thermal destruction of PCB and hexachlorobenzene. Sci. Total Environ. 1978, 10, 51–59. (27) Breivik, K.; Alcock, R.; Li, Y. F.; Bailey, R. E.; Fiedler, H.; Pacyna, J. M. Primary sources of selected POPs: Regional and global scale emission inventories. Environ. Pollut. 2004, 128 (1 2), 3–16. (28) Li, Y. M.; Jiang, G. B.; Wang, Y. W.; Wang, P.; Zhang, Q. H. Concentrations, profiles and gas-particle partitioning of PCDD/Fs, PCBs and PBDEs in the ambient air of an E-waste dismantling area, southeast China. Chin. Sci. Bull. 2008, 53 (4), 521–528. (29) Han, W. L.; Feng, J. L.; Gu, Z. P.; Wu, M. H.; Sheng, G. Y.; Fu, J. M. Polychlorinated biphenyls in the atmosphere of Taizhou, a major e-waste dismantling area in China. J. Environ. Sci. 2010, 22 (4), 589–597. (30) Xing, G. H.; Wu, S. C.; Wong, M. H. Dietary exposure to PCBs based on food consumption survey and food basket analysis at Taizhou, China - The World’s major site for recycling transformers. Chemosphere 2010, 81 (10), 1239–1244. (31) Xing, G. H.; Liang, Y.; Chen, L. X.; Wu, S. C.; Wong, M. H. Exposure to PCBs, through inhalation, dermal contact and dust ingestion at Taizhou, China - A major site for recycling transformers. Chemosphere 2011, 83 (4), 605–611. (32) Bi, X. H.; Chu, S. G.; Meng, Q. Y.; Xu, X. B. Movement and retention of polychlorinated biphenyls in a paddy field of WenTai area in China. Agric. Ecosyst. Environ. 2002, 89 (3), 241–252. (33) Pozo, K.; Harner, T.; Lee, S. C.; Sinha, R. K.; Sengupta, B.; Loewen, M.; Geethalakshmi, V.; Kannan, K.; Volpi, V. Assessing seasonal and spatial trends of persistent organic pollutants (POPs) in Indian agricultural regions using PUF disk passive air samplers. Environ. Pollut. 2011, 159 (2), 646–653. (34) Zhang, G.; Chakraborty, P.; Li, J.; Sampathkumar, P.; Balasubramanian, T.; Kathiresan, K.; Takahashi, S.; Subramanian, A.; Tanabe, S.; Jones, K. C. Passive Atmospheric Sampling of Organochlorine Pesticides, Polychlorinated Biphenyls, and Polybrominated Diphenyl Ethers in Urban, Rural, and Wetland Sites along the Coastal Length of India. Environ. Sci. Technol. 2008, 42 (22), 8218–8223. (35) Chen, L. G.; Peng, X. C.; Huang, Y. M.; Xu, Z. C.; Mai, B. X.; Sheng, G. Y.; Fu, J. M.; Wang, X. H. Polychlorinated Biphenyls in the Atmosphere of an Urban City: Levels, Distribution, and Emissions. Arch. Environ. Contam. Toxicol. 2009, 57 (3), 437–446. (36) Hu, D. F.; Lehmler, H. J.; Martinez, A.; Wang, K.; Hornbuckle, K. C. Atmospheric PCB congeners across Chicago. Atmos. Environ. 2010, 44 (12), 1550–1557. (37) Someya, M.; Ohtake, M.; Kunisue, T.; Subramanian, A.; Takahashi, S.; Chakraborty, P.; Ramachandran, R.; Tanabe, S. Persistent organic pollutants in breast milk of mothers residing around an open dumping site in Kolkata, India: Specific dioxin-like PCB levels and fish as a potential source. Environ. Int. 2010, 36 (1), 27–35. (38) Halse, A. K.; Schlabach, M.; Eckhardt, S.; Sweetman, A.; Jones, K. C.; Breivik, K. Spatial variability of POPs in European background air. Atmos. Chem. Phys. 2011, 11, 1549–1564. (39) Wang, T. E-waste creates hot spots for POPs. Environ. Sci. Technol. 2007, 41 (8), 2655–2656. (40) Kahhat, R.; Kim, J.; Xu, M.; Allenby, B.; Williams, E.; Zhang, P. Exploring e-waste management systems in the United States. Resour. Conserv. Recycl. 2008, 52 (7), 955–964. 9159
dx.doi.org/10.1021/es202320c |Environ. Sci. Technol. 2011, 45, 9154–9160
Environmental Science & Technology (41) Yu, J. L.; Williams, E.; Ju, M. T.; Shao, C. F. Managing e-waste in China: Policies, pilot projects and alternative approaches. Resour. Conserv. Recycl. 2010, 54 (11), 991–999. (42) Frazzoli, C.; Orisakwe, O. E.; Dragone, R.; Mantovani, A. Diagnostic health risk assessment of electronic waste on the general population in developing countries’ scenarios. Environ. Impact Assess. Rev. 2010, 30 (6), 388–399. (43) Hossain, M. S.; Chowdhury, S. R.; Jabbar, S. M. A.; Saifullah, S. M.; Rahman, M. A. Occupational health hazards of ship scrapping workers at Chittagong coastal zone, Bangladesh. Chiang Mai J. Sci. 2008, 35 (2), 370–381. (44) Gioia, R.; Eckhardt, S.; Breivik, K.; Jaward, F. M.; Prieto, A.; Nizzetto, L.; Jones, K. C. Evidence for major emissions of PCBs in the West African region. Environ. Sci. Technol. 2011, 45, 1349–1355. (45) Jaward, F. M.; Barber, J. L.; Booij, K.; Dachs, J.; Lohmann, R.; Jones, K. C. Evidence for dynamic air-water coupling and cycling of persistent organic pollutants over the open Atlantic Ocean. Environ. Sci. Technol. 2004, 38 (9), 2617–2625. (46) Gioia, R.; Nizzetto, L.; Lohmann, R.; Dachs, J.; Temme, C.; Jones, K. C. Polychlorinated Biphenyls (PCBs) in air and seawater of the Atlantic Ocean: Sources, trends and processes. Environ. Sci. Technol. 2008, 42 (5), 1416–1422. (47) Klanova, J.; Cupr, P.; Holoubek, I.; Boruvkova, J.; Pribylova, P.; Kares, R.; Tomsej, T.; Ocelka, T. Monitoring of persistent organic pollutants in Africa. Part 1: Passive air sampling across the continent in 2008. J. Environ. Monit. 2009, 11 (11), 1952–1963. (48) Li, N. Q.; Wania, F.; Lei, Y. D.; Daly, G. L. A comprehensive and critical compilation, evaluation, and selection of physical-chemical property data for selected polychlorinated biphenyls. J. Phys. Chem. Ref. Data 2003, 32 (4), 1545–1590. (49) Weber, R.; Watson, A.; Forter, M.; Oliaei, F. Persistent organic pollutants and landfills - a review of past experiences and future challenges. Waste Manage. Res. 2011, 29 (1), 107–121. (50) Lewis, R. G.; Martin, B. E.; Sgontz, D. L.; Howes, J. E. Measurement of fugitive atmospheric emissions of polychlorinatedbiphenyls from hazardous-waste landfills. Environ. Sci. Technol. 1985, 19 (10), 986–991. (51) Persson, N. J.; Pettersen, H.; Ishaq, R.; Axelman, J.; Bandh, C.; Broman, D.; Zebuhr, Y.; Hammar, T. Polychlorinated biphenyls in polysulfide sealants - Occurrence and emission from a landfill station. Environ. Pollut. 2005, 138 (1), 18–27. (52) Valle, M. D.; Jurado, E.; Dachs, J.; Sweetman, A. J.; Jones, K. C. The maximum reservoir capacity of soils for persistent organic pollutants: Implications for global cycling. Environ. Pollut. 2005, 134 (1), 153–164. (53) Chen, S.-J.; Tian, M.; Wang, J.; Shi, T.; Luo, Y.; Luo, X.-J.; Mai, B.-X. Dechlorane Plus (DP) in air and plants at an electronic waste (e-waste) site in South China. Environ. Pollut. 2011, 159 (5), 1290–1296. (54) Tian, M.; Chen, S.-J.; Wang, J.; Zheng, X.; Luo, X.-J.; Mai, B.-X. Brominated Flame Retardants in the Atmosphere of E-waste and Rural Sites in Southern China: Seasonal Variation, Temperature Dependence, and Air-Particle Partitioning. Environ. Sci. Technol. 2011, 45, 8819–8825. (55) Tian, M.; Chen, S.-J.; Wang, J.; Shi, T.; Luo, X.-J.; Mai, B.-X. Atmospheric Deposition of Halogenated Flame Retardants at Urban, E-Waste, and Rural Locations in Southern China. Environ. Sci. Technol. 2011, 45 (11), 4696–4701. (56) Bi, X. H.; Thomas, G. O.; Jones, K. C.; Qu, W. Y.; Sheng, G. Y.; Martin, F. L.; Fu, J. M. Exposure of electronics dismantling workers to polybrominated diphenyl ethers, polychlorinated biphenyls, and organochlorine pesticides in South China. Environ. Sci. Technol. 2007, 41 (16), 5647–5653. (57) Darnerud, P. O.; Eriksen, G. S.; Johannesson, T.; Larsen, P. B.; Viluksela, M. Polybrominated diphenyl ethers: Occurrence, dietary exposure, and toxicology. Environ. Health Perspect. 2001, 109, 49–68. (58) Law, R. J.; Allchin, C. R.; de Boer, J.; Covaci, A.; Herzke, D.; Lepom, P.; Morris, S.; Tronczynski, J.; de Wit, C. A. Levels and trends of brominated flame retardants in the European environment. Chemosphere 2006, 64 (2), 187–208.
POLICY ANALYSIS
(59) Prevedouros, K.; Jones, K. C.; Sweetman, A. J. Estimation of the production, consumption, and atmospheric emissions of pentabrominated diphenyl ether in Europe between 1970 and 2000. Environ. Sci. Technol. 2004, 38 (12), 3224–3231. (60) Scheringer, M. Characterization of the environmental distribution behavior of organic chemicals by means of persistence and spatial range. Environ. Sci. Technol. 1997, 31 (10), 2891–2897. (61) Bennett, D. H.; McKone, T. E.; Matthies, M.; Kastenberg, W. E. General formulation of characteristic travel distance for semivolatile organic chemicals in a multimedia environment. Environ. Sci. Technol. 1998, 32 (24), 4023–4030. (62) Hollander, A.; Scheringer, M.; Shatalov, V.; Mantseva, E.; Sweetman, A.; Roemer, M.; Baart, A.; Suzuki, N.; Wegmann, F.; van de Meent, D. Estimating overall persistence and long-range transport potential of persistent organic pollutants: A comparison of seven multimedia mass balance models and atmospheric transport models. J. Environ. Monit. 2008, 10 (10), 1139–1147. (63) Scheringer, M. Long-range transport of organic chemicals in the environment. Environ. Toxicol. Chem. 2009, 28 (4), 677–690. (64) Asante, K. A.; Adu-Kumi, S.; Nakahiro, K.; Takahashi, S.; Isobe, T.; Sudaryanto, A.; Devanathan, G.; Clarke, E.; Ansa-Asare, O. D.; Dapaah-Siakwan, S.; Tanabe, S. Human exposure to PCBs, PBDEs and HBCDs in Ghana: Temporal variation, sources of exposure and estimation of daily intakes by infants. Environ. Int. 2011, 37 (5), 921–928. (65) Noren, K.; Meironyte, D. Certain organochlorine and organobromine contaminants in Swedish human milk in perspective of past 20 30 years. Chemosphere 2000, 40, 1111–1123. (66) Fangstrom, B.; Athanassiadis, L.; Odsjo, T.; Noren, K.; Bergman, A. Temporal trends of polybrominated diphenyl ethers and hexabromocyclododecane in milk from Stockholm mothers, 1980 2004. Mol. Nutr. Food Res. 2008, 52 (2), 187–193. (67) Xie, Z.; M€uller, A.; Ahrens, L.; Sturm, R.; Ebinghaus, R. Brominated flame retardants in seawater and atmosphere of the Atlantic and the Southern Ocean. Environ. Sci. Technol. 2011, 45 (5), 1820–1826.
9160
dx.doi.org/10.1021/es202320c |Environ. Sci. Technol. 2011, 45, 9154–9160
POLICY ANALYSIS pubs.acs.org/est
China’s Functioning Market for Sulfur Dioxide Scrubbing Technologies Yuan Xu†,‡,* † ‡
Industrial Performance Center, Massachusetts Institute of Technology, Cambridge, Massachusetts, 02139, United States Department of Geography and Resource Management and Institute of Environment, Energy and Sustainability, The Chinese University of Hong Kong, Shatin, New Territories, Hong Kong, China ABSTRACT: Countries’ differing positions in technology transfer have been a barrier in climate negotiations. Developed countries want market-based solutions with effective protection of intellectual property rights, whereas developing countries look for external support and nonmarket solutions. Although China has shared common negotiation positions with other developing countries, it has actually relied heavily on markets to acquire foreign technologies. This paper systematically examines the case of sulfur dioxide (SO2) scrubbing technologies, with first-hand information from the author’s field interviews, to explain why such a functioning market for technology could emerge in China. Existing studies focus mainly on technology suppliers or licensors and this paper adds to the understanding of consumers or licensees. Two factors should have made major contributions to the market’s emergence: (i) the huge size of the Chinese market of SO2 scrubbers, and (ii) the knowhow and maturity of the technologies. Market-based solutions of technology transfer might help large developing countries like China and India to efficiently acquire mature environmental technologies and satisfy their rapid development.
1. INTRODUCTION Developing countries are performing a more and more important role in the international cause of climate mitigation. Over the years 2000 2009, developing countries outside of the Organization for Economic Co-operation and Development (OECD) witnessed their share of global CO2 emissions from energy consumption increasing from 45% to 58%.1 China alone was responsible for 25% in 2009, up from 12% in 2000.1 Functioning markets for transferring technologies to developing countries not only are important for their economic development and upgrading along the value chain, but also have critical implications for the domestic and global environment. Due to their huge and steadily growing emissions, how fast and effective developing countries—especially those large ones like China—adopt environmentally friendly technologies is a key determinant for future climate change. Technology transfer from developed to developing countries has long been recognized as a key measure in addressing CO2 mitigation.2 One important method of technology transfer is through technology licensing. With available markets for technologies, a technology owner could choose between licensing its product or directly investing in the client country and a firm that needs technology could either license-in or innovate indigenously.3 5 In international negotiation on transferring low-carbon technologies from developed to developing countries, developed countries generally argue for market-based solutions and adequate protection of intellectual property rights (IPR), whereas developing countries often demand nonmarket solutions at lower than market rates.6 The differing positions become an obstacle to the agreement of new and effective climate treaties.6 r 2011 American Chemical Society
Even in developed countries—as Gans and Stern argue—an effective market for technology is difficult to establish because it often fails to satisfy the three criteria of effective market design as specified by Roth that successful marketplaces must be “thick, uncongested and safe”.7,8 The Roth criteria were proposed to fix broken markets or build new ones if they are missing, which could be especially useful for environmental protection as market failure is often the cause. First, an efficient market requires many potential buyers and sellers, or market thickness, to enhance the chances of effective matching. However, many ideas are not independent but reliant on other complementary ideas and assets to achieve their full value, with notable examples in low-carbon technologies.9 This problem makes the licensing of a single idea less desirable. If the ideas belong to different entities, ineffective co-ordination could limit the willingness of potential buyers and sellers to participate in the market. Second, the market should overcome Roth’s “congestion” criterion, whereby buyers and sellers should be able to negotiate with a number of possible trading partners and have sufficient time to make effective selections. In a congested market, competition is not sufficient and the price does not reach market equilibrium. Because necessary information disclosure for buyers to assess a technology’s value might lead to unwanted diffusion, the information is often kept secret between buyers and sellers to constrain open market competition, thus failing the “congestion” criterion. Received: July 7, 2011 Accepted: September 29, 2011 Revised: September 20, 2011 Published: September 29, 2011 9161
dx.doi.org/10.1021/es202333t | Environ. Sci. Technol. 2011, 45, 9161–9167
Environmental Science & Technology Third, market transactions should be “safe”; that is, conducted in good faith and with safeguards that allow the expression of real intention and information and result in mutual satisfaction. A drawback on this point is that, after licensors have disclosed information, licensees might be able to exploit it independently, without signing licensing contracts, creating issues over misuse of intellectual property. Despite unfavorable conditions, the global market for technology has been significant, amounting to about US$35 to US $50 billion in the mid-1990s5 and roughly US$100 billion in 2002.10 However only a small portion—less than one-third for the United States—of technological transactions were between unaffiliated organizations and thus true market transactions.10,11 Most cross-border technology licensing happens among developed countries and that from developed to developing countries is much rarer.10 Product markets in most developing countries are not large enough to attract many potential technology licensors. Developing countries generally lag behind developed countries in human and technological capacities which enable them to effectively absorb licensed foreign technologies and exploit their full value.12 Additionally, effective IPR protection could help address the problems of unauthorized use of intellectual property,7 but developing countries often do not have welldeveloped systems of IPR protection and thus are placed in relatively disadvantageous positions in creating an attractive market for technology.13 However, large developing countries like China and India are able to access foreign low-carbon technologies, although not those at the cutting edge.6,14 China’s rapid development of many industries has roots in the importation of foreign technologies, including, for example, wind turbines,14 large hydro-electric turbines15 and high-speed railways.16 An especially prominent case is that of sulfur dioxide (SO2) scrubbers, or Flue Gas Desulfurization facilities, which remove SO2, a major pollutant, at large point sources such as coal-fired power plants. SO2 scrubbing technologies have been commercially deployed since the mid-1970s, mainly in developed countries. Up until 1998 (expressed in terms of generating capacity of power stations thus equipped), the pace of deployment was about 10,000 MW per year in the world and 4000 MW per year in the United States.17 Many international companies had established their technological and engineering reputations in this field. China began to significantly deploy SO2 scrubbers about three decades later than developed countries, with an approximate deployment rate of 100 000 MW per year.18,19 Because of their high SO2 removal efficiencies—generally over 90% with wet-type technologies—SO2 scrubbers became the most vital technology in achieving China’s goal of 10% reduction in SO2 emissions in the 11th Five-Year Plan (2006 2010).20,21 Among the 576 000 MW of SO2 scrubbers in China at the end of 2010, over 90% were installed by Chinese companies using licensed foreign technologies.19 Major Chinese companies universally licensed foreign technologies and relied heavily upon them. Conversely, fewer than 5% were installed by foreign companies or under joint ventures.19 Domestic companies dominated the market, in spite of their initial lack of proven technologies and experience. China’s reliance on markets to acquire new technologies is quite different from the general position of developing countries in climate negotiations.6 As the largest developing country with rapid economic growth, China stands between developed and averagely developing countries. Its large markets are attractive to most technology licensors wishing to commercially exploit their
POLICY ANALYSIS
innovations. The growth of China’s human and technological capacity over the past few decades enables it to effectively take up imported technologies. In 2007, China had 1.4 million people engaged in research; about the same as the United States.22 As an indicator of China’s capacity for assimilating foreign technologies, its ratio of research and development spending to imports of technology increased about 10-fold in a decade up to 2005.23 On the other hand, China still appears among developing countries in terms of IPR protection, being ranked 63rd out of 125 economies.13 These features could act both for and against the establishment of a sound market to license-in foreign technologies. With the three Roth criteria of effective market design as a theoretical framework, this paper systematically explains why a properly functioning market for technology could emerge in China. Particular attention will be paid to understanding why large, developing countries like China differ from small ones in accessing foreign technologies and how the obstacle of inadequate IPR protection is overcome. An in-depth case study on SO2 scrubbing technologies is conducted with first-hand information from the author’s field interviews. Existing literature on the subject focuses sharply on technology licensors but to a much lesser extent on licensees.10 This paper’s analysis could contribute to the knowledge gap and help design markets for technology more effectively in the future, especially for low-carbon and other environmental technologies where the government plays a central role in creating the demand. After the introduction of data and methods in Section 2, Sections 3, 4, and 5 analyze how the Roth criteria were addressed on market thickness, congestion and safety respectively. Section 6 discusses potential implications.
2. DATA AND METHODOLOGY Publicly accessible information is utilized in this paper, including statistical data published by the Chinese government, along with annual reports of companies listed on stock markets and company information in the public domain. However, the information alone does not make possible a deep understanding of the strategies and situations of technology licensors and licensees and accordingly how the functioning market for SO2 scrubbing technologies emerged in China. Another, more important, data source is the author’s field interviews conducted with SO2 scrubber companies from March to May 2010. The interviewed companies include both foreign companies as technology licensors and Chinese clients as licensees. The two licensors are both based in the United States, each with a small representative office in Beijing, but their licensing strategies are notably different. Their Chinese licensees include several major companies. Publicly available information and field interviews are used to describe a Japanese firm with a different licensing strategy. Also, the interviewed Chinese licensees include two in each of the three following categories; state-owned, university-established and nonstate. “State-owned” firms refers to those controlled by state-owned power corporations, which could have faced less fierce competition to win SO2 scrubber projects because of their special “internal” relationship. Indigenous SO2 scrubbing technologies had been developed by a few research institutes and universities to directly transfer their human and technological capabilities to state-owned and university-established firms. Nonstate firms could behave differently due to their relative lack of such initial capabilities. In addition, though most of China’s major firms rely heavily on licensed technologies, some concentrate on applying their own. The six 9162
dx.doi.org/10.1021/es202333t |Environ. Sci. Technol. 2011, 45, 9161–9167
Environmental Science & Technology Chinese firms all licensed-in foreign technologies and one nonstate firm used its own technology in most projects. Interview questions were specified for each firm with different foci for licensors and licensees. Licensors were interviewed on: (i) The choice between technology licensing and direct investment. (ii) Licensing contracts, including competition with other licensors, contract negotiation and implementation and selection of licensees. Licensees were interviewed on (i) The choice between indigenous innovation and technology licensing. (ii) Licensing contracts, including the importance of patents and acquired knowhow, assimilation of licensed technologies, selection of licensors, payment of royalties and the evolution of SO2 scrubber costs and profit margins. As well as SO2 scrubber firms, a coal-fired power plant in Hong Kong was visited in April 2010 because it was equipped with one of the few SO2 scrubbers that were built by Chinese firms outside of mainland China. The interview explored how the contract was initiated and negotiated. The respective roles of the Chinese firm and its technology licensor, an American firm, were examined.
3. MARKET THICKNESS The Chinese market for SO2 scrubbing technologies satisfied the most important Roth criterion of an effective market market thickness with enough sellers and buyers.8 Multiple sellers from the United States, Europe, and Japan actively licensed out their technologies.18,24 In addition, in the Chinese market up to 2010, 16 companies—all Chinese—had completed at least 10 000 MW of SO2 scrubbers, all using licensed-in foreign technologies.19,24 After several decades of commercial deployment in developed countries, many firms had acquired complete technology packages. The value of one technology package is independent of others to enable licensors and licensees to hold one-to-one negotiations. 3.1. Why Did Foreign Firms Become Licensors? When deciding whether to license-out technologies or set up direct subsidiaries in developing countries or even just do nothing, firms from developed countries need to compare the expected profits of each market option. The dominant business reality in the market is technology licensing. For major foreign firms, the option to do nothing was rarely attractive because of the temptation of the huge emergent Chinese market. The two interviewed American firms reported that the Chinese government put no restrictions on allowing foreign firms to bid for SO2 scrubber projects, but many foreign firms did not expect that they would earn significant profits by establishing subsidiaries or joint ventures in China. One major American firm expected the Chinese market to peak for only a few years before it began shrinking; this expectation proved prescient.19 The initial investment of capital and human resources to establish a subsidiary in China would therefore only be of temporary benefit. The company’s past experience in other countries suggested that direct investment could not be freely withdrawn and, accordingly, it was not justified in this particular Chinese market. In addition, the lack of adequate human resources also constrained some foreign firms from choosing direct investment, particularly due to the revived U.S. market for SO2 scrubbers.1 Furthermore, their costs would hardly give them any competitive advantage. In 2000, before the emergence of a
POLICY ANALYSIS
significant indigenous SO2 scrubber industry, average capital costs in China were roughly equivalent to those in the United States but, since then, have dropped dramatically by over 80%, especially as U.S. costs have increased.1,24,25 Because of the existence of many technology providers, no foreign firm was able to prevent others from licensing technologies to China.18 Technologies therefore could not be used as a barrier to exclude Chinese firms from competing. With more and more Chinese companies entering the market, fierce competition seemed unavoidable and the contract value of SO2 scrubbers would rapidly diminish, together with profit margins.19 In addition, China’s coal-fired power plants relied on SO2 scrubber firms for engineering, procurement and construction (EPC). The EPC contracts had fixed values and SO2 scrubber firms were responsible for budget control. Foreign companies generally trail behind Chinese ones in terms of relationship networks and their understanding of the Chinese market. Interviews revealed that foreign firms were particularly concerned about financial risks associated with the construction phase. The decision of technology licensing involves the revenue effect (i.e., payments received from licensing) and rent-dissipation effect (i.e., revenue loss due to a new or strengthened competitor in the product market).26 A stronger revenue effect promotes the decision to license, whereas a stronger rentdissipation effect discourages licensing. If the downstream operations of a company are small or the downstream market is in fierce competition, the rent-dissipation effect will be limited and technology licensing becomes more likely.10 Indeed, the Chinese downstream SO2 scrubber market was newly created and in fierce competition.19 In addition, market evolution also demonstrated that the rent-dissipation effect should be minimal. Among all the foreign companies, the examined Japanese firm ought to be the best prepared for the Chinese market. It owns more Chinese patents on flue gas desulfurization than any other company27 and, between the late 1980s to 1990s, has won contracts to install China’s first-ever commercial wet SO2 scrubbers (four units of 360 MW capacity).28 However, up to the end of 2010, its technology was only applied to a further 3300 MW, with the final project in 2006.29 Interviews in China revealed that many foreign firms generally licensed design software together with other knowhow in order to enable their Chinese licensees to compete independently, but this Japanese firm was reluctant to hand over design software and wanted to participate more actively. Thus the technology transfer of knowhow was not complete. The decision could have been influenced by the expectedly significant rent-dissipation effect due to potentially high rents as a result of its favorable position in granted patents. However, partly because the relationship made them slower in responding to the market and hampered their competitiveness, its Chinese licensees decided instead to do business with other technology licensors. For example, according to the annual reports from a company listed on the Shanghai Stock Exchange—Jiulong Electric, the holding company of Yuanda Environmental Protection Engineering—although $1.1 million was paid to the Japanese firm as the up-front lump-sum fee, just two years later it decided to sign another licensing contract with a European firm and gave up the Japanese technology (Table 1). Even with the tight control of technology licensing, the Japanese firm earned little profit or rent from the Chinese market, an indication of a small rent-dissipation effect. The existence of many technology licensors diminished the rent-dissipation effect because no single licensor had significant market power. 9163
dx.doi.org/10.1021/es202333t |Environ. Sci. Technol. 2011, 45, 9161–9167
Environmental Science & Technology
POLICY ANALYSIS
Table 1. Up-Front Lump-Sum Fees of SO2 Scrubbing Technology Licenses (The Chinese Licensees Here Are All Listed on Stock Markets and the Data Are from Their Annual Reports) Chinese licensee
a
country origin of the foreign licensor
lump-sum feea
year
technology type
Wuhan Kaidi
Germany
$277,304
1998
dry type
Fujian Longjing Wuhan Kaidi
Germany United States
$3,989,234 $652,118
2001 2002
wet type circulation fluidized bed wet type
Zhejiang Feida
United States
$1,250,000
2002
wet type
Jiulong Electric
Japan
$1,126,563
2002
wet type
Jiulong Electric
Austria
$1,423,765
2004
wet type
Insigma Technology
France
$1,200,000
2004
wet type
Exchange rates on December 31, 2010 are used: 1 US$ = 6.62 RMB = 0.75 Euro.
In contrast, the revenue effect was significant. Technology licensing only requires a small office in China to monitor licensees and to ‘service’ the partnership. For example, each of the two American firms interviewed has an office in Beijing with about five staff members, whereas their licensees are in charge of contracts worth several hundred million dollars annually. The initial cost in transferring technologies was covered by up-front lump-sum fees paid by licensees (Table 1). The commercial success of licensees will result in considerable royalties to the licensor if the contracts are honored. After the knowhow and trade secrets were transferred, the intellectual property rights were at risk of misuse or infringement, possibly with the royalties not being fully paid. Despite this, most foreign firms decided to take this risk in order to avoid the much greater risk inherent in direct investment. 3.2. Why Did Chinese Firms Become Licensees? For a potential licensee, the technology could either be developed internally or acquired externally. Favorable conditions created the demand for foreign technologies in the Chinese market. The “Not Invented Here” syndrome—that internally developed technologies are preferred—was found to be a barrier to technology licensing10 but it does not seem to be deeply rooted in China. Secondary innovation based on imported technologies, coupled with original and integrated innovation, have been established as three cornerstones of China’s indigenous innovation strategy.30 With regard to the installation of SO2 scrubbers, China stipulated in tendering documents that established technologies were required. As late as 2005, bidders were clearly asked to specify a foreign technology provider that had installed SO2 scrubbers of the same or greater scale.31 Interviews also confirmed the general requirement for foreign, commercialized technologies in the early years when almost no Chinese companies had any proven experience. This requirement was relaxed only in more recent years after many firms in the market had completed enough projects. Although the state-owned, university-established, and nonstate Chinese firms had different technological backgrounds, they generally did not display a “Not Invented Here” syndrome attitude and foreign technologies were often favored. From as early as the 1970s, China had, through its own research and development on SO2 scrubbers, built up vital capabilities to assimilate imported technology.32 From the mid1970s to the mid-1980s, China appraised several technologies, although on scales that were at least 1 or 2 orders of magnitude smaller than any commercial project. For example, a 300 MW unit corresponds to a flue gas flow rate of about 1 000 000 N m3/ hour (cubic meter at standard temperature and pressure per hour), while the largest Chinese experiment at the time had a
flow rate of 70 000 N m3/hour.32 From the mid-1980s to 2000, foreign technologies were demonstrated on a commercial scale.28,32 In 2000, having resulted in a considerable fund of domestic human and technological capability, foreign technologies were officially recognized as the basis for further development of SO2 scrubber technologies in China.25 China’s absorptive capacities were effectively distributed to all major firms including nonstate ones through a free labor market of engineers and managers. China’s rapid deployment of SO2 scrubbers also helped tilt the balance toward technology licensing. Time was a serious constraint, given the sharp annual increase in the nation’s coal-fired generating capacity. Peak demand occurred around 2002 and 2003 but, at that time, few domestic firms were capable of designing SO2 scrubbers. The sudden appearance of a huge market led to the creation of many new firms and the reorientation of existing ones from other industries. Because few firms had any prior experience and the market was large enough to accommodate many, most—except those owned by coal-fired power corporations—were placed on a more or less equal footing. Firms would achieve distinction if they could establish engineering and management teams and develop their technological capability faster than others. Another time constraint was the short period from the issue of tendering documents to completion of the bidding process; this typically lasted only one to four weeks. Additionally, the design process could not take more than a few months if the construction was to begin on schedule. Successful companies had to respond quickly and provide acceptable quality. When demand for SO2 scrubbers started to surge, domestic technologies were generally not able to satisfy the time constraints because of their immaturity. Domestic research and development generated “naked” technologies, to quote the word of one interviewee. Demonstration projects on a commercial scale should be followed by multiple projects to make the technology mature and ready for wide commercial deployment. The commercialization of these “naked” technologies would require at least a few years, plus significant financial resources and the willingness of coal-fired power plants to take risks by trying them. The expected short-term peak in China’s scrubber market diminished the potential return on investment in indigenous technology. The easy prospect of licensing foreign technologies also reduced the incentive to take risks with indigenous innovations. All the major Chinese firms in the market licensed foreign technologies, in order to acquire and substantiate their technological capabilities. No clear difference could be found among state-owned, university-established and nonstate firms. Even the nonstate firm that mainly applied its own technology had to initially license from abroad. 9164
dx.doi.org/10.1021/es202333t |Environ. Sci. Technol. 2011, 45, 9161–9167
Environmental Science & Technology As tacit knowledge cannot be so easily transferred as codified knowledge, knowhow played a positive role in establishing a sound market for technology. The contractual acquisition of knowhow presents more problems than licensing patents.5 However, in a developing country like China with poor IPR protection, the licensing of patents might be unnecessary in the absence of knowhow as the knowledge contained in the patents have already entered the public domain. Chinese firms have generally chosen to legally license, rather than to illegally acquire SO2 scrubbing technologies. Legal licensing secures a complete package including systematic training, technical documentation and trade secrets in a relatively short time scale, without exposing the licensees to legal disputes. One feasible solution is to recruit experts from foreign companies but the legal risks are not insignificant and the received technologies may not be complete because it would be difficult to recruit an entire team. It would also take much longer for the acquiring firms to comprehend a technology by this means than they would through technology licensing. The associated costs would not be low either, because foreign experts generally have to be paid considerably more than standard Chinese salaries. Furthermore, illegal acquisition does not provide a technological guarantee from a trusted provider, while this guarantee was stipulated by coal-fired power plants in their tendering documents.
4. MARKET CONGESTION The Chinese market also met the second Roth criterion on the lack of congestion. Interviews revealed that, although the negotiation of technology licensing was generally bilateral, without disclosing information to third parties, licensors and licensees often negotiated with several entities on the other side at the same time for most suitable licensing contracts. Competition took place between foreign firms for licensing to especially promising Chinese firms that were expected to win many projects and return significant revenues from royalties. Those potential licensees were mainly established by coal-fired power producers. Interviews showed that financial payments were the most critical aspect in negotiating licenses although other aspects were also important, such as the suitability of technologies and the scope of licenses. The willingness to accept lower up-front lump-sum fees and lower royalty rates made a licensor more competitive. A few Chinese firms on the stock market published information on up-front lump-sum fees (Table 1). The information could have helped to benchmark a price for technology licensing for later contracts. After the significant variance of early contracts, the up-front lump-sum fee stabilized to be around $1.2 million for wet scrubbers (Table 1). The existence of many potential licensees enabled licensors to design their strategies to maximize profit. At least three clear strategies emerged among three licensors. First, a major American firm licensed to only two Chinese firms and built up longterm partnerships through full technical support. One license was restricted to the licensee’s home province for a certain period and the other covered the whole of mainland China. The licensees had a near-monopoly to use the specific technology in their assigned market territories. Second, another significant American firm had about eight licensees in China; the strategy was to increase the market share of its technology as well as its royalties, but the licensees were still selected so as to prevent unqualified ones from ruining the technology’s reputation. Third, as
POLICY ANALYSIS
mentioned above, a Japanese firm licensed its technology to a few Chinese firms but, unlike the two American firms, refused to transfer design software. The two American firms had their technologies widely applied but the Japanese technology was abandoned without much deployment. From the perspective of the level of royalties, the two American strategies were clear winners. Generally, certain essential information needs to be disclosed for a licensee to estimate the value of a technology but such disclosure could shrink the potential monopoly and its revenue potential.7 Compared with cutting-edge technologies, mature technologies such as SO2 scrubbing are much more easily assessed. The market for scrubbers was well-defined with its dominant application in coal-fired power plants and broadly understood price range. Scrubbing technologies had been widely and commercially deployed; licensees could acquire sufficient technological knowledge by visiting actual projects. The existence of significant knowhow facilitates more information disclosure, because there is a much smaller decline in value for knowhow than for codified knowledge.
5. MARKET SAFETY IPR protection is recognized as a key means to ensure market safety and satisfy the third Roth criterion.7 In China, this criterion was met in different ways. One concern is that licensors might not transfer technologies completely after receiving payments.5 In the case of SO2 scrubbing technology, royalties dominate the revenue stream in technology licensing and effectively deter such a moral hazard. By way of example, an American firm charged one licensee $652,118 as the up-front lump-sum fee (Table 1): interviews discovered that a license’s approximate royalty rate should be 2% of SO2 scrubber contract values. Between 2004 and 2010, the firm’s income from royalties was nearly 40 times as much as the up-front lump-sum fee (the licensee completed 34 900-MW wet SO2 scrubbers in that period19 and the national average contract value was about $35/kW24). From another perspective, as demonstrated in the above case of the Japanese licensor, a licensee’s loss was limited to approximately the up-front lumpsum fee when the technology transfer was not satisfactory. In addition, if a licensor gained a bad reputation, this could limit its future business opportunities in the huge and rapidly growing Chinese market. After technologies are transferred, another concern arises on whether licensees pay royalties honestly. Both licensors and licensees reported in interviews that major Chinese firms were still paying royalties regularly. Also, several expiring licenses had been renewed, indicating a good record of royalty payments. As a preventative measure, design software was encrypted and only specially prepared computers could install it with annual reregistration. Several interviewees in the Chinese firms said that, after a few years, they had figured out what was inside the black box but still chose to pay royalties. It was not very difficult to keep track of licensees. The huge size of SO2 scrubbers often made local news and the Ministry of Environmental Protection annually published details of every SO2 scrubber and its contractor.19 Besides, a good partnership with licensors suits the long-term interests of licensees. Technological sophistication has increased step by step in the Chinese SO2 scrubber market as reflected in the unit scales: the 300-MW scale was dominant before 2005, but after 2006 the 600-MW scale became crucial with 31 scrubbers already 9165
dx.doi.org/10.1021/es202333t |Environ. Sci. Technol. 2011, 45, 9161–9167
Environmental Science & Technology at 1000 MW capacity or greater.19 Every significant increase in scale indicates a new technical advance. Accordingly, the licensing of scrubbing technologies was a continuous operation and not a one-off process. Good partnerships, strengthened by honest royalty payments, could also help licensees expand into new markets through future technology licensing. In addition, a partnership may generate business opportunities for both sides. For example, when a large coal-fired power plant in Hong Kong decided to install SO2 scrubbers, it first approached several international firms, including one from the United States. But the American firm was fully committed in the domestic market and was not willing to take the financial risk of an EPC project in Hong Kong. Its Chinese licensee was introduced and finally won the contract. Royalty rates may decrease over time to reduce the costs of honoring licensing contracts. For example, one license divided the 10-year contract period into three phases with declining royalty rates. In several other cases, the royalty rate was renegotiated when competition in the market became much too fierce to significantly shrink the profit margin. Excessively high royalty rates could damage licensees’ competitiveness. The final result might be a reduced income from royalties and an increased risk of no payment being made at all. The renegotiation strengthened the partnerships between licensors and licensees and thus worked for the interests of both sides. In one licensing contract signed in 1998, the level of royalties was originally associated with the volume of flue gases. Because China’s capital costs of installing SO2 scrubbers had dropped substantially since then,24,25 the royalty rate would increase significantly as a percentage of the contract value. Renegotiation took place to lower the royalty rate. The partnership remained strong with both the licensor and licensee maintaining market success. Lawsuits, particularly those resolved outside China, are also a deterrent to potential infringement. For example, Insigma Technology is a Chinese company listed on the Shanghai Stock Exchange and it releases information regularly. It signed a technology licensing contract with a French firm in December 2004 (Table 1). However, in April 2006, Insigma declared that it would cancel the contract and thereafter stop using the licensed technology. Royalties were paid for six projects in 2005 and 2006 with a total capacity of 7450 MW.33 The company later signed a new contract with an Italian firm in September 2006, which was for one year and was to be automatically renewed if no objections were received from either side. The fee for royalties was a fixed sum of h20,000 ($26,600) for every project regardless of the contract value.33 The French firm later sued Insigma in Singapore (where disputes should be resolved according to the licensing contract). The court made a decision in February 2010 and Insigma was ordered to pay compensation of $2,085,737 for the loss of royalties in 2005 and $24,566,684 for the loss afterward.33 The lawsuit may have helped deter other significant licensees from not honoring their licensing contracts.
6. DISCUSSION Issues over technology transfer have become obstacles to effective treaties in international climate negotiations. Developed countries argue for market-based solutions and the strengthening of intellectual property rights (IPR) protection, while developing countries want external support and nonmarket solutions. However, though China negotiated as a developing country, it makes extensive use of international technology markets to rapidly build
POLICY ANALYSIS
up its technological strength. Foreign licensors received largely honest royalty payments to earn significant profits. Through a case study on China’s market for SO2 scrubbing technologies with first-hand information from field interviews, this paper explains why a functioning market for technology emerged in China. China’s large market size and the maturity of available technologies should be the most critical factors meeting the three Roth criteria of effective market design: respectively market thickness, lack of congestion and market safety. Because the size of the Chinese market for SO2 scrubbers as a downstream market for the technologies is far greater than any other country, major foreign SO2 scrubber firms, as potential licensors, could hardly overlook the potential business opportunities. The large market and low technological barriers facilitated by technology licensing have created many domestic firms as potential licensees. The three types of Chinese firms—state-owned, university-established, and nonstate—did not show significantly different behavior in the market for technology. Fierce competition drove down costs and diminished expected profit from direct investment, but revenue from technology licensing was significant. The rentdissipation effect was overwhelmed by the revenue effect of technology licensing, which accordingly became a dominant choice of the foreign firms. As a large country, China has a strong capacity to absorb new technology due to its previous research and development and this capacity was effectively distributed to all three types of firms through a free labor market. In order to meet time constraints and technological requirements, major Chinese firms universally licensed-in foreign technologies to quickly build up technological strength. Licensors and licensees held multiple bilateral negotiations simultaneously to help solve the market congestion problem. Furthermore, the safety of technology licensing also benefited from China’s large market size. As a result of the large market, there were significant revenues from royalties which encouraged licensors to transfer complete packages of technologies. The market for SO2 scrubbers at every unit scale was substantial and the unit scales escalated over time to require continuous technological support from licensors. Such dynamism favored long-term partnerships between licensors and licensees for their mutual benefit and fostered honest royalty payments. SO2 scrubbing technologies were commercialized long before their wide deployment in China. Personal and corporate expertise, or knowhow as tacit knowledge, was a vital part of the technology package. Acquiring knowhow raises costs and contracting problems, but given the inadequate standard of IPR protection in China, technology licensing becomes necessary in order to acquire complete packages of technologies. Many foreign companies had become independent technology holders and a potential licensee only needs to negotiate with one licensor for a complete technology package. The maturity and wide deployment of SO2 scrubbing technologies also enabled a fairly accurate estimation of the technology’s value to facilitate market transactions. The disclosure of the necessary information for value assessment in negotiations caused fewer problems because knowhow could not be easily acquired. An effective market for cutting-edge technologies is understandably more difficult to establish. It is probable that not many organizations have acquired intellectual property as potential licensors. The value of a particular cutting-edge technology is harder to assess and the accumulation of knowhow may still be in progress with a consequently high price of the final product which will limit its deployment. These unfavorable conditions 9166
dx.doi.org/10.1021/es202333t |Environ. Sci. Technol. 2011, 45, 9161–9167
Environmental Science & Technology discourage the emergence of potential licensees. Information disclosure to facilitate licensing will also raise more concerns on the part of technology owners. As a result, the Roth criteria of effective market design will be harder to meet for cutting-edge than for mature technologies. The functioning market for technology should not be confined to China or to SO2 scrubbing technologies. Other large developing countries, such as India, also have potentially large markets through which to spawn many domestic operators and fierce competition. Many other low-carbon and pollution-control technologies have been commercialized with much knowhow. A caveat is that these large developing countries may not necessarily always have large domestic markets for pollution mitigation. These are partly determined by government policies and not just by the overall sizes of their economies. Their abilities to take on board foreign technologies might not be consistently strong. However, there is great potential for large developing countries to make use of markets for technology to build their industrial prowess with mature technologies. Though external financial support could further help these countries, their current strategies at international negotiations could be revised to better satisfy their urgent need of mature technologies for rapid development and environmental cleanup. China and India in particular may want to use climate negotiations as a platform to establish more effective markets to acquire technologies from developed countries.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +852-39436647; e-mail:
[email protected].
’ ACKNOWLEDGMENT I especially thank Edward Steinfeld for his critical guidance and support. I also thank Robert Socolow, Robert Williams, Nicolas Lefevre, Richard Lester, Kin-Che Lam, Yongqin Chen, David Wilmshurst, a number of interviewees, three reviewers and the editor for valuable comments and information sharing. Funding was provided by the Industrial Performance Center at Massachusetts Institute of Technology and a grant from Hong Kong Research Grant Council (code: C001-2021063). ’ REFERENCES (1) U.S. Energy Information Administration, Official Energy Statistics from the U.S. Government. http://www.eia.doe.gov/ (accessed May 18, 2011). (2) United Nations, United Nations Framework Convention on Climate Change,1992. (3) Arora, A.; Fosfuri, A.; Gambardella, A. Markets for technology and their implications for corporate strategy. Ind. Corporate Change 2001, 10 (2), 419–451. (4) Teece, D. J. Capturing value from technological innovation— Integration, strategic partnering, and licensing decisions. Interfaces 1988, 18 (3), 46–61. (5) Arora, A.; Fosfuri, A.; Gambardella, A., Markets for Technology: The Economics of Innovation and Corporate Strategy; MIT Press: Cambridge, Mass., 2001; p xi. (6) Ockwell, D. G.; Haum, R.; Mallett, A.; Watson, J. Intellectual property rights and low carbon technology transfer: Conflicting discourses of diffusion and development. Global Environ. Change 2010, 20 (4), 729–738.
POLICY ANALYSIS
(7) Gans, J. S.; Stern, S. Is there a market for ideas? Ind. Corporate Change 2010, 19 (3), 805–837. (8) Roth, A. E. What have we learned from market design? Econ. J. 2008, 118 (527), 285–310. (9) Harvey, I., Intellectual Property Rights: The Catalyst to Deliver Low Carbon Technologies; The Climate Group: London, U.K., 2008. (10) Arora, A.; Gambardella, A. Ideas for rent: an overview of markets for technology. Ind. Corporate Change 2010, 19 (3), 775–803. (11) Saggi, K. Trade, foreign direct investment, and international technology transfer: A survey. World Bank Res. Obs. 2002, 17 (2), 191–235. (12) Metz, B.; Turkson, J. K.; Intergovernmental Panel on Climate Change. Working Group III; Methodological and Technological Issues in Technology Transfer; Cambridge University Press: Cambridge, UK, 2000; p 466. (13) Strokova, V. International Property Rights Index—2010 Report; Americans for Tax Reform Foundation/Property Rights Alliance: Washington, DC, 2010. (14) Lewis, J. I. Technology acquisition and innovation in the developing world: Wind turbine development in China and India. Stud. Comp. Int. Dev. 2007, 42 (3 4), 208–232. (15) Liang, W. Power equipment of the gigantic three gorges project. In Proceedings of the Fifth International Conference on Electrical Machines and Systems, 2001, 2001, 2001; Vol. 1, pp 676 678. (16) Chan, L.; Aldhaban, F. Technology transfer to china: with case studies in the high-speed rail industry. In PICMET 2009 Proceedings, August 2 6, 2009, Portland, OR, 2009; pp 2858 2867. (17) Srivastava, R. K.; Jozewicz, W.; Singer, C. SO2 scrubbing technologies: A review. Environ. Prog. 2001, 20 (4), 219–228. (18) Xu, Y.; Williams, R. H.; Socolow, R. H. China’s rapid deployment of SO2 scrubbers. Energy Environ. Sci. 2009, 2, 459–465. (19) Ministry of Environmental Protection. China’s Capacities of Water Treatment Plants, SO2 Scrubbers and SCR Systems at Coal Power Plants: Beijing, China, 2011. (20) Xu, Y. Improvements in the operation of SO2 scrubbers in China’s coal power plants. Environ. Sci. Technol. 2011, 45 (2), 380–385. (21) Xu, Y. The use of a goal for SO2 mitigation planning and management in China’s 11th Five-Year Plan. J. Environ. Plann. Manage. 2011, 54 (6), 769–783. (22) National Science Board. Science and Engineering Indicators 2010; National Science Foundation: Arlington, VA, 2010; Vol. NSB 10 01. (23) OECD. Reviews of Innovation Policy—China; OECD Publishing: Paris, France, 2008. (24) Xu, F.; Yi, B.; Zhuang, D.; Yang, M.; Yan, J.; Yan, Z. Survey report on the construction and operation of SO2 scrubbers at coal power plants in the 10th Five-Year Plan. In The Fourth Conference on Flue Gas Desulfurization Technologies: Beijing, China, 2006. (25) Key Planning Points on Flue Gas Desulfurization Technologies and Their Domestic Production (2000 2010); National Economic and Trade Commission: Beijing, China, 2000. (26) Arora, A.; Fosfuri, A. Licensing the market for technology. J. Econ. Behav. Organ. 2003, 52 (2), 277–295. (27) State Intellectual Property Office, Database of patents granted in China. http://www.sipo.gov.cn/sipo2008/zljs/ (accessed July 2, 2010). (28) Gu, X. A summary of installing SO2 scrubbers in Luohuang power plant and the impact on the sociery and environment. In Annual conference of the Chinese Association of Science, Hainan, China, 2004. (29) Mitsubishi Heavy Industries, Delivery record. http://www.mhi. co.jp/en/products/pdf/delivery_record.pdf (accessed January 27, 2011). (30) State Council. The Outline of National Science and Technology Development Plan in the Middle and Long term, 2006. (31) Guizhou Qiandong Power Station, Tendering Document for the Flue Gas Desulfurization island. http://www.in-en.com/power/ html/power-2006200604145584.html (accessed January 27, 2011). (32) Shu, H. SO2 emission control for coal fired power plant. Electr. Equip.t 2003, 4 (4), 4–8. (33) Sina Finance, Information on Insigma Technology Co., Ltd. http://finance.sina.com.cn (accessed February 9, 2011). 9167
dx.doi.org/10.1021/es202333t |Environ. Sci. Technol. 2011, 45, 9161–9167
ARTICLE pubs.acs.org/est
Characterization of Solvent-Extractable Organics in Urban Aerosols Based on Mass Spectrum Analysis and Hygroscopic Growth Measurement Toshiyuki Mihara† and Michihiro Mochida*,‡,§ †
Department of Earth and Environmental Science, Graduate School of Environmental Studies, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan ‡ Institute for Advanced Research, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan
bS Supporting Information ABSTRACT: To characterize atmospheric particulate organics with respect to polarity, aerosol samples collected on filters in the urban area of Nagoya, Japan, in 2009 were extracted using water, methanol, and ethyl acetate. The extracts were atomized and analyzed using a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) and a hygroscopicity tandem differential mobility analyzer. The atmospheric concentrations of the extracted organics were determined using phthalic acid as a reference material. Comparison of the organic carbon concentrations measured using a carbon analyzer and the HR-ToF-AMS suggests that organics extracted with water (WSOM) and ethyl acetate (EASOM) or those extracted with methanol (MSOM) comprise the greater part of total organics. The oxygencarbon ratios (O/C) of the extracted organics varied: 0.510.75 (WSOM), 0.370.48 (MSOM), and 0.270.33 (EASOM). In the ion-group analysis, WSOM, MSOM, and EASOM were clearly characterized by the different fractions of the CH and CO2 groups. On the basis of the hygroscopic growth measurements of the extracts, k of organics at 90% relative humidity (korg) were estimated. Positive correlation of korg with O/C (r 0.70) was found for MSOM and EASOM, but no clear correlation was found for WSOM.
1. INTRODUCTION Atmospheric particulate organics are a major aerosol component, accounting for 1070% of the total aerosol mass.1 For this reason, their characterization is important to elucidate the role of aerosols in visibility, climate, and human health.1,2 Particulate organics comprise myriad species, among which only about 1020% in a bulk sample are identifiable on a molecular level.1,3 Chemical characterization of a wider fraction of organics therefore remains as an important task to elucidate the key roles of atmospheric aerosol particles, such as their role as cloud condensation nuclei (CCN). Evidence of the influence of organic aerosol aging on changes in chemical and physical properties has been reported.47 Therefore, characterization of unidentified and identified organics in view of aging is important.6,7 Chemical characterization of a broader range of particulate organics has been undertaken using several approaches with assessments of functional groups, degree of oxidation, volatility, and solubility in solvents.2,411 Russell et al. investigated the functional groups of organics using FTIR spectroscopy, reporting that the CCN activity correlated with the oxygencarbon ratios (O/C).8 Polidori et al. used FTIR spectroscopy to determine the organic mass (OM) to organic carbon (OC) ratios of organics fractionated by extraction with solvents having different polarities.2 Zhang et al. reported the characterization of organics based on results of factor analysis using an aerosol mass spectrometer (AMS).11 These recent studies have provided information that is useful to clarify the behavior of particulate organics in the atmosphere. More detailed characterization is r 2011 American Chemical Society
necessary to clarify the characteristics of organics further and to elucidate how those characteristics affect the particle properties related to atmospheric processes. For this study, using multiple solvents with different polarities, we extracted organics from ambient particles collected on filters in the urban area of Nagoya, Japan. We investigated them using a high-resolution time-of-flight aerosol mass spectrometer (HRToF-AMS) and a hygroscopicity tandem differential mobility analyzer (HTDMA). The chemical compositions of the extracts were determined. Furthermore, elemental analysis of organics and ion-series analysis combined with ion-group analysis using high-resolution (HR) mass spectra of organics were conducted to assess the characteristics of the chemical structures. Although AMS is commonly used in online aerosol measurements,12 results of this study demonstrate its high applicability to off-line analysis of atmospheric organic aerosol components. In addition, we assessed the hygroscopicity of solvent-extractable organics in light of AMS-derived characteristics.
2. EXPERIMENTAL SECTION 2.1. Aerosol Sampling. Eight aerosol samples were collected on quartz fiber filters (25 20 cm) for 7296 h using a Received: April 14, 2011 Accepted: August 30, 2011 Revised: August 9, 2011 Published: August 30, 2011 9168
dx.doi.org/10.1021/es201271w | Environ. Sci. Technol. 2011, 45, 9168–9174
Environmental Science & Technology
ARTICLE
Figure 1. Schematic representation of different organic fractions. WSOM, MSOM, and EASOM are the organic fractions extracted from a filter sample. WS-EASOM is the organic fraction extracted from EASOM with water. Blue boxes show the organic fractions measured directly using HR-ToF-AMS. White boxes show the organic fractions calculated from the differences of two measured fractions. Gray box shows EOM, which is similar to MSOM (see section S8 in the Supporting Information).
high-volume air sampler (ca. 1100 L min1, model 120B; Kimoto Electric Co. Ltd.) equipped with a cascade impactor (50% cutoff diameter, 0.95 μm, TE-230; Tisch Environmental, Inc.) on the balcony (10 m agl) of a building at Nagoya University in Nagoya, Japan (35°090 N and 136°580 E) during spring and summer of 2009. Filter punches (9.1 cm2 3) were ultrasonicated with 3 g of a solvent three times (10 min each) to prepare the extract. The solvents used were water (Fluka/Sigma-Aldrich Corp.), methanol (Wako Pure Chemical Inds. Ltd.), and ethyl acetate (Wako), whose solvent polarity indices are, respectively, 9.0, 6.6, and 4.3.2 We also prepared a water-soluble (WS) fraction in ethyl acetate extract (WS-EASM) by concentrating an ethyl acetate extract (EASM) using a rotary evaporator, drying it under a N2 flow, and dissolving the WS fraction in the residue with 2.5 g of water. Definitions of different organic fractions are presented in Figure 1. All definitions and abbreviations of organic fractions are summarized in section S1 in the Supporting Information. Aerosols were generated from each extract (3 mL) using an atomizer (see section S2, Supporting Information). The aerosol was passed through a prehumidifier to allow for possible compaction of the particles because some atomized aerosol particles might be aspherical or porous in a dry condition. The solvents in the aerosols were removed using diffusion-dryer-type scrubbers filled with (1) activated carbon (Wako) mixed with silica gel (only used for HTDMA measurements of aerosols from extracts with organic solvents), (2) silica gel (Wako), and (3) molecular sieves (13X/4A; Supelco and Sigma-Aldrich). We expected the removal of the vapor of organic solvents by activated carbon and of water by silica gel and molecular sieves. The dry aerosol (<10% RH) was then used for HR-ToF-AMS and HTDMA analyses. 2.2. Chemical Composition Measurements. The HR-ToFAMS was operated in the V-mode,13 which was used to measure all HR mass spectra in this study (see section S2, Supporting Information, for details). The collected data were then analyzed using ToF-AMS data analysis software (Squirrel, Pika and Apes, http://cires.colorado.edu/jimenez-group/ToFAMSResources/ ToFSoftware/). The unit mass spectra were used to quantify the concentrations of organics and inorganic salts. To quantify the organic CHO+ and CO2+ signals accurately, the equations in the fragment table of air for 15NN+ (m/z 29) and CO2+ (m/z 44) were modified based on the blank measurements under the condition of atomizing organic-free water (see section S5, Supporting Information). The signal intensities from the CO, HO, and O groups (CO+, H2O+, HO+, and O+) in the particulate organics were estimated according to the procedure described by
Aiken et al.14 The atmospheric concentrations of solvent-extractable organics and inorganic salts were quantified through additional HR-ToF-AMS analyses: A known amount of phthalic acid (PA) was mixed with the extract in the solvent, with subsequent determination of the relative abundance of the extract and PA from the ToF-AMS spectrum of the mixture (see section S5, Supporting Information). The chemical stability of organics in methanol (MSOM) and ethyl acetate (EASOM) against possible reactions (e.g., transesterification) was acceptable for mass spectrum analyses, as inferred from experiments using deuterated solvents (d solvents) (see section S6, Supporting Information). The effects of solvent vapor for the mass spectra of MSOM and EASOM were found to be negligible from measurements during atomization of extract-free organic solvents (see section S5, Supporting Information). However, this study did not assess the possible effects of solvents on hygroscopicity. The mass concentrations of organics and mass spectra of the water-insoluble fraction in EASOM (WI-EASOM) were calculated by subtracting those of WS-EA-soluble organic matter (WS-EASOM) from those of EASOM, respectively (see Figure 1 and section S7, Supporting Information). Similarly, the mass concentrations of organics and mass spectra of WS-ethyl acetate insoluble organic matter (WS-EAISOM) were calculated by subtracting those of WS-EASOM from those of the water-soluble organic matter (WSOM). The results for WI-EASOM should be regarded with caution because of the possible uncertainty: negative values of the mass concentrations of the CHO group, whose absolute values were much less than the mass concentrations of total organics, were substituted with zero for ion-series analysis (see section S8, Supporting Information). 2.3. Elemental, Ion-Series, and Ion-Group Analyses. Elemental ratios (O/C, H/C) of organics and OM/OC were calculated as described by Aiken et al.14,15 OM/OC was calculated from molar quantities of elements of oxygen (MO), carbon (MC), and hydrogen (MH) as OM 12MC þ 16MO þ MH ¼ OC 12MC
ð1Þ
Nitrogen and sulfur were not considered because their relative abundances in organics in ambient aerosols are presumably low. (N/C and S/C in Riverside were on average only about 0.015 and about 0.001, respectively.16 The averaged mass fraction of nitrogen-enriched organic aerosol accounted for only 5.8% of the AMS-derived organic aerosol mass in New York.)17 9169
dx.doi.org/10.1021/es201271w |Environ. Sci. Technol. 2011, 45, 9168–9174
Environmental Science & Technology
ARTICLE
The degree of oxidation and presence of cyclic or unsaturated bonds or terpene derivative structures were assessed using ionseries analysis for the HR mass spectra.18 This analysis is based on the classification of AMS signals according to the delta (Δ) value as Δ ¼ m=z 14n þ 1 ðn is integerÞ
ð2Þ
where m/z denotes the mass-to-charge ratio. Organics with cyclic or unsaturated bonds or terpene derivative structures yield Δ values that are less than or equal to 0, whereas oxygenated organics yield Δ greater than or equal to 2.10,18 Ion-group analysis using HR mass spectra also provides information related to the degree of oxidation. Signal intensities of ions from organics were classified into several groups based on the elemental species in the ions: CH, CHO, CHN, and CHON groups and CO, OH, and O groups. Drewnick et al.,19 for example, demonstrated the usefulness of ion-series analysis to characterize the organics originated from different sources (traffic-related and photochemically generated aerosols). We combined ion-series analysis with ion-group analysis to characterize the organics further, based on different ion groups in the same nominal masses (e.g., the ions at m/z 29 could include C2H5+ (CH group, associated with hydrocarbon-like organics) and CHO+ (CHO group, associated with oxygenated organics)). Elemental ratios of WI-EASOM were calculated from those of WS-EASOM and EASOM (see section S7, Supporting Information). The ion-series analyses of WI-EASOM were performed using the mass spectra calculated by subtracting the mass spectra of WS-EASOM from those of EASOM (see section S7, Supporting Information). 2.4. Hygroscopic Growth Measurement. The hygroscopicity of particles generated from the extracts was measured using HTDMA (see section S2, Supporting Information). Dry particles of 200 nm in mobility diameter (<10% RH) were classified in the first DMA. Then the humidity conditioner, the second DMA, and the CPC were used to measure the hygroscopic growth. In dehumidification mode, the particles were prehumidified (about 95% RH) before introduction into the humidity conditioner. The RH was scanned every 190 min (130 min for up-scan from about 10% to about 96% RH and 60 min to dry the system). The hygroscopic growth factor g(RH), which is defined here as the ratio of the mobility diameter under the humidified condition to that in the dry (<10% RH) condition, was determined as described in an earlier report by Mochida et al. (including auxiliary materials).20 2.5. Measurements of Inorganic Ions and Organic Carbon. Inorganic ions (NH4+, SO42, and NO3) in the filter samples were quantified using two different methods: one using an ion chromatograph (ICS-1000; Dionex Corp.) and the other using AMS (section S5, Supporting Information). Total organic carbon (TOC) in the filter samples was quantified using thermaloptical transmittance (TOT) and thermaloptical reflectance (TOR) techniques using a Lab OC-EC aerosol analyzer (Sunset Laboratory Inc.) and a Thermal/Optical Carbon Analyzer (model 2001; Desert Research Institute) with the IMPROVE temperature program9 (see section S4, Supporting Information).
3. RESULTS AND DISCUSSION 3.1. Aerosol Chemical Composition. Figure 2 presents the AMS-derived atmospheric concentrations of nonrefractory components. The concentrations of WSOM, MSOM, and EASOM from eight filter samples were within the respective ranges of
Figure 2. Atmospheric concentrations of inorganic salts and WSOM, MSOM, and EASOM. Bars corresponding to one sampling period show, from left to right, WSM, MSM, and EASM.
2.05.3, 2.67.2, and 1.45.0 μg m3. The concentrations of WS-EASOM, WI-EASOM, and WS-EAISOM were, respectively, in the ranges of 0.82.9, 0.32.1, and 0.42.4 μg m3. Organics, which were extracted most efficiently using methanol, constituted the major component in all extracts. Concentrations of sulfate, nitrate, and ammonium were determined, respectively, from water extracts (WSM) as 1.03.5, 0.21.3, and 0.51.7 μg m3. The trend of the correlations between sulfate in WSM (which is a proxy of the aging of aerosols, considering the lifetime of SO2 (e.g., 26 h in Tokyo in summer21)) and WSOM, MSOM, and EASOM (r: 0.63, 0.35, and 0.40, respectively) suggests that organics in aged aerosols were more likely to be extracted with more polar solvents than organics in less aged aerosols. Although inorganic salts were extracted substantially with methanol (49 117% with respect to the AMS-derived WSM), they were not extracted efficiently with ethyl acetate. To confirm the validity of the atmospheric concentrations of chemical species calculated using the AMS data, the AMS-derived concentrations of inorganic ions were compared with the concentrations measured using ion chromatography (IC) (see section S5, Supporting Information). The AMS-derived and IC-derived concentrations agreed well for nitrate and ammonium. The mass concentration of the AMS-derived sulfate was slightly lower than that of the IC-derived sulfate; no correction was made for further analysis explained later. 3.2. Comparison of OC Concentrations Measured Using a Carbon Analyzer and HR-ToF-AMS. Concentrations of TOC determined from eight filter samples using the thermaloptical method were compared with those of the solvent-extractable OC measured using HR-ToF-AMS. Figure 3 shows that TOCs were 1.13.5 μg C m3, whereas the sums of OC in WI-EASOM (WIEASOC) and WSOM (WSOC) and in MSOM (MSOC) were, respectively, in the ranges of 1.13.8 and 1.33.6 μg C m3. The MSOC and sum of the WSOC and WI-EASOC (EOC) were close to those of the TOC (r2: 0.38 and 0.60, respectively; MSOC/TOC and EOC/TOC were 1.09 ( 0.30 and 1.00 ( 0.23 (mean ( SD), respectively), suggesting that most organics were extracted with methanol or the combination of water and ethyl acetate. If nonextractable organics are neglected, 9170
dx.doi.org/10.1021/es201271w |Environ. Sci. Technol. 2011, 45, 9168–9174
Environmental Science & Technology
Figure 3. Atmospheric concentrations of solvent-extractable OC and TOC measured using the thermaloptical method. The AMS-derived OM was converted to OC using OM/OC obtained from elemental analysis. WI-EASOC and WS-EAISOC were calculated indirectly. Ionization efficiency (IE) of the WI-EASOC was assumed to be 0.65 times the common IE for organics (1.4), considering the low ionization efficiency of hydrocarbons.32 Similarly, the MSOC was assumed to contain the WI-EASOC with low IE. The fraction was taken from the ratio of the WI-EASOC to the sum of the WSOC and the WI-EASOC. For TOC, the average values from different optical correction methods and different instruments are presented with the ranges shown by the bars with caps.
the atmospheric concentrations of OM can be estimated as 2.37.4 μg m3. The AMS-derived OC concentrations of some samples were larger than the TOC concentrations, suggesting some errors in the AMS measurements used for quantification of OM, OM/ OC, or in the TOC measurements. However, at least on average, the method of solvent extraction and subsequent AMS measurement must be capable of quantifying the total organics with sufficient accuracy. The similarity of the mean WSOC/OC measured in this study (0.74 ( 0.14 (mean ( SD)) and those in urban background aerosols (mean of PM2.5 in Hong Kong in summer 0.69 ( 0.09 (mean ( SD); annual mean of PM1 in Helsinki 0.56)22,23 is consistent with this inference, although widely various ratios in other locations have also been reported (mean of PM1 in Tokyo in summer, 0.35; mean of PM2.5 in southeastern United States in summer, 0.440.64).24,25 3.3. Elemental ratios of solvent-extractable organics. Figure 4 presents plots of H/C against O/C for WSOM, MSOM, and EASOM. Plots for WS-EASOM and indirectly calculated WI-EASOM are also presented. Plots for WS-EAISOM are not shown because its large error is suggested by its large variation. The mean O/C of MSOM and the sum of WSOM and WIEASOM (EOM) were 0.44 and 0.42, respectively, which are presumably similar to that of total organics (see section 3.2 and section S8, Supporting Information). The ratios are similar to that of urban organic aerosols in Mexico City (mean O/C 0.41), as determined from a real-time AMS measurement.14 The O/C of WSOM and WS-EASOM was higher than those of MSOM and EASOM extracted with less polar solvents (methanol and ethyl acetate, respectively). Conversely, the H/C of EASOM and WI-EASOM was higher than those of MSOM
ARTICLE
Figure 4. Elemental ratios of H/C against O/C determined from the HR mass spectra. Ratios of WI-EASOM and total organics were calculated indirectly. Bars with caps show the ranges of the ratios in which the intersections correspond to the mean values.
and EASOM. Jimenez et al.4 explained the evolution of organic aerosol based on PMF-derived semivolatile OOA (SV-OOA) and low-volatility OOA (LV-OOA). Ng et al.26 reported the respective mean O/C of LV-OOA and SV-OOA as 0.73 ( 0.14 and 0.35 ( 0.14 (mean ( SD). Figure 4 shows that the O/C of WSOM was similar to the ratio of LV-OOA. The O/C of the extracted organics except for WSOM was in most cases in the range of the ratio of SV-OOA, but the O/C of WI-EASOM was similar to the ratio of hydrocarbon-like organic aerosol (HOA).14 WS-EASOM did not follow the trend from EASOM to WSOM, perhaps because WS-EASOM is rich in cyclic or unsaturated bonds and because it has lower H/C. 3.4. Characterization of Solvent-Extractable Organics from Ion-Series and Ion-Group Analyses. Figure 5 presents ion-series and ion-group analyses for organics in different fractions (see section S8, Supporting Information). In all extracts, the ion signal intensities were dominant between Δ from 2 to 5. As Figure 5ac shows, the ion group of CO2, which originates from a highly oxygenated fraction like carboxyl groups, was extracted as larger fractions with more polar solvents (32%, 17%, and 13% for WSOM, MSOM, and EASOM, respectively). This result suggests that the oxygenated fraction is present abundantly in organics in aged aerosols because the order of the percentages of the CO2 group is the same as the order of the preference in extraction of organics in aged aerosols for preparation of WSOM, MSOM, and EASOM (section 3.1). By contrast, the CH group, which might originate from an aliphatic and/or aromatic hydrocarbon fraction mainly in primary organics, was extracted as smaller fractions with more polar solvents (35%, 46%, and 54%, respectively, for WSOM, MSOM, and EASOM). Bahreini et al. reported that the signal intensity at Δ e 0 is a proxy for cyclic or unsaturated bonds or terpene derivative structures.10 In Figure 5ac, the signal intensities at Δ e 0 are higher for the organics extracted with less polar solvents (29%, 32%, and 38%, respectively, for WSOM, MSOM, and EASOM). This result implies that cyclic or unsaturated or terpene derivative structures were present as larger fractions in organics extracted with less polar solvents (e.g., in association with humic-like substance 9171
dx.doi.org/10.1021/es201271w |Environ. Sci. Technol. 2011, 45, 9168–9174
Environmental Science & Technology
Figure 5. Results of ion-series and ion-group analyses of HR mass spectra obtained from (a) WSOM, (b) MSOM, (c) EASOM, (d) WSEASOM, and (e) WI-EASOM. The result in e was estimated from those in c and d. Contributions of different ion-groups are shown by different colors. Bars with caps show ranges of values.
(HULIS) with cyclic and conjugated bonds27). The result is explained by the greater influence of the oxidation of organics along with aging on the extracts with more polar solvents. However, loss of cyclic or unsaturated or terpene derivative structures by aging is not evident, given that the ratios of the CH group at Δ e 0 (the major ion group at Δ e 0) to all the CH groups were comparable for WSOM, MSOM, and EASOM (0.61 ( 0.07, 0.62 ( 0.03, and 0.48 ( 0.16 (mean ( SD), respectively). WI-EASOM (Figure 5e) was characterized by the greater fraction of the CH group (66%) than that of WS-EASOM (Figure 5d), as expected from its insolubility in water. This result suggests the similarity of WI-EASOM to hydrocarbon. 3.5. Connection between the Hygroscopicity of SolventExtractable Organics and AMS-Derived Chemical Characteristics. Figure 6 presents the RH dependence of the ratios of aerosol water volume Vw(RH) divided by the dry particle volume Vdry for WSM, methanol, and ethyl acetate-soluble matter (MSM and EASM, respectively), as derived from HTDMA measurements. As predicted from the aerosol inorganic model (AIM) for the inorganic salt fractions,28 differences of Vw(RH)/Vdry between humidification and dehumidification modes were observed at <80% RH, although the Vw(RH)/Vdry in both modes
ARTICLE
Figure 6. Mean aerosol water volume Vw(RH) divided by the dry particle volume Vdry of each extract as a function of RH in humidification (solid diamond) and dehumidification (open diamond) modes, as calculated from HTDMA data. Solid and dashed lines, respectively, show the humidification-mode and dehumidification-mode Vw,inorg(RH)/Vdry calculated using the AIM model, where Vw,inorg(RH) is the volume of water retained by inorganic salts. The dehumidificationmode Vw,inorg(RH)/Vdry was calculated by preventing formation of solid phases in the AIM model. The mean Vw(RH)/Vdry with 5% RH intervals were calculated using interpolation of g(RH) near the RH. Bars with short and long caps, respectively, show the ranges of values in humidification and dehumidification modes.
agree well at >80% RH. At 90% RH, the Vw(RH)/Vdry were in the ranges of 1.713.28, 1.372.92, and 0.120.66 for WSM, MSM, and EASM, respectively, as derived from g(RH) in the ranges of 1.391.62, 1.331.58, and 1.041.18, respectively. The relation between the O/C and the hygroscopicity k of organics (korg) at 90% RH is presented in Figure 7 (see section S9, Supporting Information, for calculation of korg). Here, a nitrate loss of 43.3 vol % in the experimental system is assumed, as estimated from the measured size change of NH4NO3 between the first and the second DMAs in the HTDMA and extrapolation of the loss to the other part of the sample aerosol line. (The nitrate loss is not considered for measurements using the HR-ToF-AMS.) The korg of the MSOM and the EASOM with low inorganic salt fractions showed moderate hygroscopicity: 0.030.27 and 0.010.08, respectively. The average value of korg of the MSOM was 0.1, which is presumably similar to that of total organics (see section 3.2 and section S8, Supporting Information). The korg of the combination of MSOM and 9172
dx.doi.org/10.1021/es201271w |Environ. Sci. Technol. 2011, 45, 9168–9174
Environmental Science & Technology
ARTICLE
found to be correlated with the O/C. The presented fractionation method might be useful to extend the analyses further, for example, to assess the effect of organics on particle CCN activity and to examine the optical properties of brown carbon-containing unsaturated bonds.
’ ASSOCIATED CONTENT
bS
Supporting Information. Definitions and abbreviations of organic fractions, instrumental setup, sampling artifacts, and extraction efficiency during analysis, blank levels of quartz fiber filters, method used to quantify organics and inorganic salts, assessment of the stability of organics in the solvents, derivation of OM, and elemental ratios of WI-EASOM, ion-series and iongroup analyses for organics, estimation of the aerosol water content, and calculation of hygroscopicity k. This material is available free of charge via the Internet at http://pubs.acs.org. Figure 7. Relation between the O/C and the korg of WSOM, MSOM, and EASOM. korg estimated with evaporative losses of nitrate of 43.3 and 0100 vol % in the HTDMA are presented, respectively, with the markers and bars with caps. Circles and squares, respectively, denote results for experiments in humidification and dehumidification modes. WSOM and WS-EASOM were excluded from analysis of the correlation coefficient and the regression line (solid line). The dotted line represents the regression line (0.3 < O/C < 0.6) reported by Chang et al.29
EASOM positively correlate with the O/C (r = 0.70). The slope of the regression line (0.20) is similar to the value of 0.29 reported by Chang et al.29 The korg of WS-EASOM with no inorganic salts was estimated as 0.030.21 on the assumption that korg of the WI-EASOM is zero. Although the plots for WSEASOM are close to the regression line, the plots of korg of WSOM were scattered (from 0.10 to 0.25). The values of korg of WSOM with high inorganic salts collected in spring were negative even if complete evaporation of nitrate was assumed, but the plots for WSOM with low inorganic salt fractions were close to the regression line for MSOM and EASOM. Although Huffman et al. reported some volatilities of nitrate and organics in atmospheric aerosols,30 large decreases in particle diameters between the first and the second DMAs in the HTDMA due to evaporative losses were not evident for the extracts; under dry conditions (<10% RH), the growth factors of the extracts were nearly identical to those of (NH4)2SO4, evaporation of which is not expected. (The growth factors of WSOM, MSOM, and EASOM were, respectively, 1.00, 1.01, and 0.99 times of that of (NH4)2SO4.) Hence, large evaporative losses of inorganic salts and/or organics might not occur in the HTDMA measurements, although the evaporative losses without compaction of particles cannot be fully ruled out because the particles may not be perfectly spherical or nonporous under dry conditions. Although measurement biases and other biases (e.g., that by particle asphericity) are possible causes, nonadditivity of the volume of water retained by organics and inorganic salts,31 which possibly leads to the negative korg, is not ruled out. This point remains unresolved; it must be studied using more precise measurements in the future. Organics fractionated based on the solubility to solvents were found to show different characteristics such as degrees of oxidation and amounts of cyclic or unsaturated bonds or terpene derivative structures. The results suggest that organics with different degrees of oxidation were mixed in the urban air of Nagoya. Furthermore, the hygroscopicity of the extracted organics was
’ AUTHOR INFORMATION Corresponding Author
*Phone/fax: +81-52-788-6157. E-mail:
[email protected]. Present Addresses §
Department of Earth and Environmental Science, Graduate School of Environmental Studies, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan.
’ ACKNOWLEDGMENT We thank Murata Keisokuki Service Co. Ltd. for help in the analyses of inorganic ions and OC/EC, and we appreciate the help of Tokyo Dylec Corp. in the analysis of OC/EC. This study was supported in part by a Grant-in-Aid for Young Scientists (S) (20671001). ’ REFERENCES (1) Turpin, B. J.; Saxena, P.; Andrews, E. Measuring and simulating particulate organics in the atmosphere: problems and prospects. Atmos. Environ. 2000, 34, 2983–3013. (2) Polidori, A.; Turpin, B. J.; Davidson, C.; Rodenburg, L. A.; Maimone, F. Organic PM2.5: Fractionation by polarity, FTIR spectroscopy, and OM/OC ratio for the Pittsburgh aerosol. Aerosol Sci. Technol. 2008, 42, 233–246. (3) Rogge, W. F.; Mazruek, M. A.; Hildemann, L. M.; Cass, G. R.; Simoneit, B. R. Quantification of urban organic aerosols at a molecular level: identification, abundance and seasonal variation. Atmos. Environ. Part A 1993, 27 (8), 1309–1330. (4) Jimenez, J. L.; Canagaratna, M. R.; Donahue, N. M.; Prevot, A. S. H.; Zhang, Q.; Kroll, J. H.; DeCarlo, P. F.; Allan, J. D.; Coe, H.; Ng, N. L.; Aiken, A. C.; Docherty, K. S.; Ulbrich, I. M.; Grieshop, A. P.; Robinson, A. L.; Duplissy, J.; Smith, J. D.; Wilson, K. R.; Lanz, V. A.; Hueglin, C.; Sun, Y. L.; Tian, J.; Laaksonen, A.; Raatikainen, T.; Rautiainen, J.; Vaattovaara, P.; Ehn, M.; Kulmala, M.; Tomlinson, J. M.; Collins, D. R.; Cubison, M. J.; Dunlea, E. J.; Huffman, J. A.; Onasch, T. B.; Alfarra, M. R.; Williams, P. I.; Bower, K.; Kondo, Y.; Schneider, J.; Drewnick, F.; Borrmann, S.; Weimer, S.; Demerjian, K.; Salcedo, D.; Cottrell, L.; Griffin, R.; Takami, A.; Miyoshi, T.; Hatakeyama, S.; Shimono, A.; Sun, J. Y.; Zhang, Y. M.; Dzepina, K.; Kimmel, J. R.; Sueper, D.; Jayne, J. T.; Herndon, S. C.; Trimborn, A. M.; Williams, L. R.; Wood, E. C.; Middlebrook, A. M.; Kolb, C. E.; Baltensperger, U.; Worsnop, D. R. Evolution of organic aerosols in the atmosphere. Science 2009, 326, 1525–1529. (5) Duplissy, J.; DeCarlo, P. F.; Dommen, J.; Alfarra, M. R.; Metzger, A.; Barmpadimos, I.; Prevot, A. S. H.; Weingartner, E.; Tritscher, T.; 9173
dx.doi.org/10.1021/es201271w |Environ. Sci. Technol. 2011, 45, 9168–9174
Environmental Science & Technology Gysel, M.; Aiken, A. C.; Jimenez, J. L.; Canagaratna, M. R.; Worsnop, D. R.; Collins, D. R.; Tomlinson, J.; Baltensperger, U. Relating hygroscopicity and composition of organic aerosol particulate matter. Atmos. Chem. Phys. 2011, 11 (3), 1155–1165. (6) Sun, Y. L.; Zhang, Q.; Anastasio, C.; Sun, J. Insights into secondary organic aerosol formed via aqueous-phase reactions of phenolic compounds based on high resolution mass spectrometry. Atmos. Chem. Phys. 2010, 10 (10), 4809–4822. (7) Sun, Y.; Zhang, Q.; Zheng, M.; Ding, X.; Edgerton, E. S.; Wang, X. Characterization and source apportionment of water-soluble organic matter in atmospheric fine particles (PM2.5) with highresolution aerosol mass spectrometry and GCMS. Environ. Sci. Technol. 2011, 45 (11), 4854–4861. (8) Russell, L. M.; Takahama, S.; Liu, S.; Hawkins, L. N.; Covert, D. S.; Quinn, P. K.; Bates, T. S. Oxygenated fraction and mass of organic aerosol from direct emission and atmospheric processing measured on the R/V Ronald Brown during TEXAQS/GoMACCS 2006. J. Geophys. Res. 2009, 114, D00F05; doi: 10.1029/200JD011275. (9) Chow, J. C.; Watson, J. G.; Pritchett, L. C.; Pierson, W. R.; Frazier, C. A.; Purcell, R. G. The DRI thermal/optical reflectance carbon analysis system: description, evaluation and applications in U.S. Air quality studies. Atmos. Environ., Part A 1993, 27, 1185–1201. (10) Bahreini, R.; Keywood, M. D.; Ng, N. L.; Varutbangkul, V.; Gao, S.; Flagan, R. C.; Seinfeld, J. H.; Worsnop, D. R.; Jimenez, J. L. Measurements of secondary organic aerosol from oxidation of cycloalkenes, terpenes, and m-xylene using an aerodyne aerosol mass spectrometer. Environ. Sci. Technol. 2005, 39, 5674–5688. (11) Zhang, Q.; Alfarra, M. R.; Worsnop, D. R.; Allan, J. D.; Coe, H.; Canagaratna, M. R.; Jimenez, J. L. Deconvolution and quantification of hydrocarbon-like and oxygenated organic aerosols based on aerosol mass spectrometry. Environ. Sci. Technol. 2005, 39, 4938–4952. (12) Dunlea, E. J.; DeCarlo, P. F.; Aiken, A. C.; Kimmel, J. R.; Peltier, R. E.; Weber, R. J.; Tomlinson, J.; Collins, D. R.; Shinozuka, Y.; McNaughton, C. S.; Howell, S. G.; Clarke, A. D.; Emmons, L. K.; Apel, E. C.; Pfister, G. G.; van Donkelaar, A.; Martin, R. V.; Millet, D. B.; Heald, C. L.; Jimenez, J. L. Evolution of Asian aerosols during transpacific transport in INTEX-B. Atmos. Chem. Phys. 2009, 9, 7257–7287. (13) DeCarlo, P. F.; Kimmel, J. R.; Trimborn, A.; Northway, M. J.; Jayne, J. T.; Aiken, A. C.; Gonin, M.; Fuhrer, K.; Horvath, T.; Docherty, K. S.; Worsnop, D. R.; Jimenez, J. L. Field-deployable, high-resolution, time-of-flight aerosol mass spectrometer. Anal. Chem. 2006, 78, 8281–8289. (14) Aiken, A. C.; DeCarlo, P. F.; Kroll, J. H.; Worsnop, D. R.; Huffman, J. A.; Docherty, K. S.; Ulbrich, I. M.; Mohr, C.; Kimmel, J. R.; Sueper, D.; Sun, Y.; Zhang, Q.; Trimborn, A.; Northway, M.; Ziemann, P. J.; Canagaratna, M. R.; Onasch, T. B.; Alfarra, M. R.; Prev^ot, A. S. H.; Dommen, J.; Duplissy, J.; Metzger, A.; Baltensperger, U.; Jimenez, J. L. O/C and OM/OC ratios of primary, secondary, and ambient organic aerosols with high-resolution time-of-flight aerosol mass spectrometry. Environ. Sci. Technol. 2008, 42, 4478–4485. (15) Aiken, A. C.; DeCarlo, P. F.; Jimenez, J. L. Elemental analysis of organic species with electron ionization high-resolution mass spectrometry. Anal. Chem. 2007, 79, 8350–8358. (16) Docherty, K. S.; Aiken, A. C.; Huffman, J. A.; Ulbrich, I. M.; DeCarlo, P. F.; Sueper, D.; Worsnop, D. R.; Snyder, D. C.; Grover, B. D.; Eatough, D. J.; Goldstein, A. H.; Ziemann, P. J.; Jimenez, J. L. The 2005 Study of Organic Aerosols at Riverside (SOAR-1): instrumental intercomparisons and fine particle composition. Atmos. Chem. Phys. Discuss. 2011, 11, 6301–6362. (17) Sun, Y.-L.; Zhang, Q.; Schwab, J. J.; Demerjian, K. L.; Chen, W.-N.; Bae, M.-S.; Hung, H.-M; Hogrefe, O.; Frank, B.; Rattigan, O. V.; Lin, Y.-C. Characterization of the sources and processes of organic and inorganic aerosols in New York city with a high-resolution time-of-flight aerosol mass spectrometer. Atmos. Chem. Phys. 2011, 11, 1581–1602. (18) McLafferty, F.; Turecek, F. Interpretation of Mass Spectra; University Science Books: Sausalito, CA, 1993. (19) Drewnick, F.; Schwab, J. J.; Jayne, J. T.; Canagarantna, M.; Worsnop, D. R.; Demerjian, K. L. Measurement of ambient aerosol
ARTICLE
composition during the PMTACS-NY 2001 using an aerosol mass spectrometer. Aerosol Sci. Technol. 2004, 38 (S1), 92–103. (20) Mochida, M.; Nishita-Hara, C.; Kitamori, Y.; Aggarwal, S. G.; Kawamura, K.; Miura, K.; Takami, A. Size-segregated measurements of cloud condensation nucleus activity and hygroscopic growth for aerosols at Cape Hedo, Japan in spring 2008. J. Geophys. Res. 2010, 115, D21207; doi: 10.1029/2009JD013216. (21) Miyakawa, T.; Takegawa, N.; Kondo, Y. Removal of sulfur dioxide and formation of sulfate aerosol in Tokyo. J. Geophys. Res. 2007, 112, D13209; doi: 10.1029/2006JD007896. (22) Ho, K. F.; Lee, S. C.; Cao, J. J.; Li, Y. S.; Chow, J. C.; Watson, J. G.; Fung, K. Variability of organic and elemental carbon, water soluble organic carbon, and isotopes in Hong Kong. Atmos. Chem. Phys. 2006, 6, 4569–4576. (23) Saarikoski, S.; Timonen, H.; Saarnio, K.; Aurela, M.; J€arvi, L.; Keronen, P.; Kerminen, V.-M.; Hillamo, R. Sources of organic carbon in fine particulate matter in northern European urban air. Atmos. Chem. Phys. 2008, 8, 6281–6295. (24) Miyazaki, Y.; Kondo, Y.; Takegawa, N.; Komazaki, Y.; Kawamura, K.; Mochida, M.; Okuzawa, K.; Weber, R. J. Time-resolved measurements of water-soluble organic carbon in Tokyo. J. Geophys. Res. 2006, 111, D23206; doi: 10.1029/2006JD007125. (25) Weber, R. J.; Sullivan, A. P.; Peltier, R. E.; Russell, A.; Yan, B.; Zheng, M.; de Gouw, J.; Warneke, C.; Brock, C.; Holloway, J. S.; Atras, E. L.; Edgerton, E. A study of secondary organic aerosol formation in the anthropogenic-influenced southern United States. J. Geophys. Res. 2007, 112, D13302; doi: 10.1029/2007JD008408. (26) Ng, N. L.; Canagaratna, M. R.; Zhang, Q.; Jimenez, J. L.; Tian, J.; Ulbrich, I. M.; Kroll, J. H.; Docherty, K. S.; Chhabra, P. S.; Bahreini, R.; Murphy, S. M.; Seinfeld, J. H.; Hildebrandt, L.; Donahue, N. M.; DeCarlo, P. F.; Lanz, V. A.; Prev^ot, A. S. H.; Dinar, E.; Rudich, Y.; Worsnop, D. R. Organic aerosol components observed in Northern Hemispheric datasets from Aerosol Mass Spectrometry. Atmos. Chem. Phys. 2010, 10, 4625–4641. (27) Graber, E. R.; Rudich, Y. Atmospheric HULIS: How humic-like are they? A comprehensive and critical review. Atmos. Chem. Phys. 2006, 6, 729–753. (28) Clegg, S. L.; Brimblecombe, P.; Wexler, A. S. Thermodynamic model of the system H+NH4+SO42NO3H2O at tropospheric temperatures. J. Phys. Chem. A 1998, 102, 2137–2154. (29) Chang, R. Y.-W.; Slowik, J. G.; Shantz, N. C.; Vlasenko, A.; Liggio, J.; Sjostedt, S. J.; Leaitch, W. R.; Abbatt, J. P. D. The hygroscopicity parameter (k) of ambient organic aerosol at a field site subject to biogenic and anthropogenic influences: relationship to degree of aerosol oxidation. Atmos. Chem. Phys. 2010, 10, 5047–5064. (30) Huffman, J. A.; Docherty, K. S.; Aiken, A. C.; Cubison, M. J.; Ulbrich, I. M.; DeCarlo, P. F.; Super, D.; Jayne, J. T.; Worsnop, D. R.; Ziemann, P. J.; Jimenez, J. L. Chemically-resolved aerosol volatility measurements from two megacity field studies. Atmos. Chem. Phys. 2009, 9, 7161–7182. (31) Chan, M. N.; Chan, C. K. Hygroscopic properties of two model humic-like substances and their mixtures with inorganics of atmospheric importance. Environ. Sci. Technol. 2003, 37, 5109–5115. (32) Jimenez, J. L.; Jayne, J. T.; Shi, Q.; Kolb, C. E.; Worsnop, D. R.; Yourshaw, I.; Seinfeld, J. H.; Flagan, R. C.; Zhang, X.; Smith, K. A.; Morris, J. W.; Davidovits, P. Ambient aerosol sampling using the aerodyne aerosol mass spectrometer. J. Geophys. Res. 2003, 108 (D7), 8425; doi: 10.1029/2001JD001213.
9174
dx.doi.org/10.1021/es201271w |Environ. Sci. Technol. 2011, 45, 9168–9174
ARTICLE pubs.acs.org/est
Accumulation of Microplastic on Shorelines Woldwide: Sources and Sinks Mark Anthony Browne,*,†,‡,§ Phillip Crump,z Stewart J. Niven,§,|| Emma Teuten,§ Andrew Tonkin,z Tamara Galloway,^ and Richard Thompson§ †
School of Biology & Environmental Sciences, University College Dublin, Science Centre West, Belfield, Dublin 4, Ireland Centre for Research on the Ecological Impacts of Coastal Cities, A11 School of Biological Sciences, University of Sydney, NSW 2006, Australia § Marine Biology & Ecology Research Group, School of Marine Science & Engineering, University of Plymouth, Plymouth PL4 8AA, United Kingdom z School of Geography, Earth & Environmental Sciences, University of Plymouth, Plymouth PL4 8AA, United Kingdom Waters Canada, Ontario, Canada ^ School of Biosciences, College of Life & Environmental Sciences, University of Exeter, Exeter EX4 4PS, United Kingdom
)
‡
ABSTRACT: Plastic debris <1 mm (defined here as microplastic) is accumulating in marine habitats. Ingestion of microplastic provides a potential pathway for the transfer of pollutants, monomers, and plastic-additives to organisms with uncertain consequences for their health. Here, we show that microplastic contaminates the shorelines at 18 sites worldwide representing six continents from the poles to the equator, with more material in densely populated areas, but no clear relationship between the abundance of miocroplastics and the mean size-distribution of natural particulates. An important source of microplastic appears to be through sewage contaminated by fibers from washing clothes. Forensic evaluation of microplastic from sediments showed that the proportions of polyester and acrylic fibers used in clothing resembled those found in habitats that receive sewage-discharges and sewage-effluent itself. Experiments sampling wastewater from domestic washing machines demonstrated that a single garment can produce >1900 fibers per wash. This suggests that a large proportion of microplastic fibers found in the marine environment may be derived from sewage as a consequence of washing of clothes. As the human population grows and people use more synthetic textiles, contamination of habitats and animals by microplastic is likely to increase.
’ INTRODUCTION We use >240 million tonnes of plastic each year1 and discarded ‘end-of-life’ plastic accumulates, particularly in marine habitats,1 where contamination stretches from shorelines2 to the open-ocean3 5 and deep-sea.6 Degradation into smaller pieces means particles <1 mm (defined here as microplastic 2,7,8 ) are accumulating in habitats,1 outnumbering larger debris.7 Once ingested by animals, there is evidence that microplastic can be taken up and stored by tissues and cells, providing a possible pathway for accumulation of hydrophobic organic contaminants sorbed from seawater, and constituent monomers and plasticadditives, with probable negative consequences for health.9 16 Over the last 50 years the global population-density of humans has increased 250% from 19 to 48 individuals per square km,17 during this time the abundance of micrometer-sized fragments of acrylic, polyethylene, polypropylene, polyamide, and polyester have increased in surface waters of the northeast Atlantic Ocean.1 This debris now contaminates sandy, estuarine, and subtidal habitats in the United Kingdom,1,6 Singapore,18 and India.19 Despite these isolated reports, the global extent of contamination by microplastic is largely unknown. This has prompted the United Nations, Group of Experts on Scientific Aspects of r 2011 American Chemical Society
Marine Environmental Protection, International Oceanographic Commision,14 European Union,15 Royal Society,3 and National Oceanic and Atmospheric Administration (USA)16 to all identify the need to improve our understanding about how widespread microplastic contamination is, where it accumulates, and the source of this material. If spatial patterns of microplastic result primarily from the transportation of natural particulates by currents of water, shores that accumulate smaller-sized particles of sediment should accumulate more microplastic. Alternatively, spatial patterns may be influenced by sources of microplastic; with more material along shorelines adjacent to densely populated areas which already have a greater abundance of larger items of debris20 and receive millions of tonnes of sewage each year21 which has also been shown to contain microplastic.22 26 Although larger debris is removed in sewage treatment plants, filters are not specifically designed to retain microplastic and terrestrial soils that have received sewage sludge do contain Received: May 27, 2011 Accepted: September 6, 2011 Revised: August 27, 2011 Published: September 06, 2011 9175
dx.doi.org/10.1021/es201811s | Environ. Sci. Technol. 2011, 45, 9175–9179
Environmental Science & Technology microplastic fibers.27 In the UK alone, over 11 km3 of water is discharged into inland waters, estuaries, and the sea each year21 from treatment plants. Certain subtidal marine sites may, however, contain large quanitites of microplastic in their sediments because for nearly 30 years, a quarter of UK sewage sludge was dumped at 13 designated marine disposal-sites around the coast, until this practice was stopped in 1998 through The Urban Waste Water Treatment Regulations 1994.21,22 Since substantial quantities of sewage sludge and effluent are discarded to the sea, there is considerable potential for microplastic to accumulate in aquatic habitats, especially in densely populated countries. To manage the environmental problems of microplastic it is important to understand and target the major pathways of microplastic into habitats with mitigation-measures. While sewage waste provides one potential route for entry of microplastics, others have been identified including fragmentation of larger items, introduction of small particles that are used as abrasives in cleaning products, and spillage of plastic powders and pellets. Forensic techniques that compare the size, shape, and type of polymers28 may provide useful insights into the sources of the microplastic. For instance, if the material originated from fragmentation, the frequency-distribution of sizes of plastic debris would be skewed to smaller irrgeular fragments from the major types of macroplastic (e.g., polyethylene, polystyrene, polypropylene) found in habitats.7 If, however, scrubbers in cleaning products were more important, we would expect most of the material to consist of fragments and spheres of polyethylene. These sources do not, however, account for the occurrence of microplastic fibers in sludge and effluent taken from sewage treatment works26 and soil from terrestrial habitats where sewage sludge had been applied, the source of which is more likely explained by fibers shed from clothes/textiles during washing.27 Work is therefore needed to gather forensic information about the number, type of polymer and shape, to assess the likelihood of microplastic entering marine habitats through this possible pathway. Here, we investigate the spatial extent of microplastic across the shores of six continents to examine whether spatial patterns relate to its sources or sinks. We test the following hypotheses that there will be more microplastic in habitats that accumulate smaller particles of sediment (hypothesis 1) and in areas with larger population-densities of humans (hypothesis 2). Based on forensic analyses of the material we then tested the hypotheses that sediment collected from sewage-disposal sites contains more microplastic than reference sites (hypothesis 3), that microplastic found on the shoreline will resemble microplastic found in subtidal sewage disposal sites, sewage-effluent discharged from treatment works, and wastewater from washing clothes using washing machines (hypothesis 4).
’ MATERIALS AND METHODS Global Sampling of Sediment from Shores. Samples of sediment were collected from sandy beaches in Australia (Port Douglas; 1629S, 14528E; Busselton Beach 3339S, 11519E), Japan (Kyushu 3224N, 13139E), Oman, United Arab Emirates (Dubai 2517N, 5518E), Chile (Vina Del Mar 3256S, 7132W; Punta Arenas 5308S, 7053W), Philippines (Malapascua Island 0118N, 01103E), Portugal (Faro 3659N, 0757W), Azores (Ponta Delgado 3744N, 2534W), USA (Virginia 3656N, 7614W; 3657N, 7614W; California 3550N, 11823W), South Africa (Western Cape 3306S, 1757E), Mozambique
ARTICLE
(Pemba 1901S, 3601E), and the United Kingdom (Sennon Cove 5004N, 0541W) from 2004 to 2007. During collection (and in subsequent sections), cotton clothing was worn rather than synthetic items (such as fleeces) to avoid contamination by plastic fibers. Samples were collected by working down-wind to the particular part of the highest strandline deposited by the previous tide. Sediment was sampled to a depth of 1 cm deep using established techniques.7 As the sampling was opportunistic, the sampling design was unable to remove possible confounding due to intrinsic dfferences in the tidal range and position of the strandline that will vary spatially and temporally on the shores. The extraction and identification of microplastic, including the analysis of sediment particle-size, was done using established methods.1,7 Microplastic debris was extracted from a 50 mL subsample of sedimentary material using a filtered, saturated solution of sodium chloride to separate particles of microplastic from sediments. This involved three sequential extractions using the saline solution and identifying the microplastic using Transmittance FT- IR and a spectral database of synthetic polymers (Bruker I26933 Synthetic fibres ATRlibrary). Marine Sewage Disposal and Reference Sites. In 2008 and 2009, samples of sediment (n = 5) were haphazardly collected from each reference (Plymouth 5014N, 0410W and Tyne 5506N, 0118W) and sewage-sludge disposal site (Plymouth 5014N, 0418W; Tyne 5503N, 0117W) using van Veen grabs deployed from a boat. The surface 5 10 cm of sediment of each sample was placed into precleaned 500 mL aluminum foil containers and microplastic extracted as before. During collection, cotton clothing was worn rather than synthetic items to avoid contamination by plastic fibers. Sewage Effluent. Microplastic was extracted from effluent discharged (n = 5) by two sewage treatment plants. Precleaned glass bottles (750 mL) with metal caps were used to collect effluent from discharges from Tertiary-level Sewage Treatment Plants at West Hornsby and Hornsby Heights (NSW, Australia) in 2010. Effluent was filtered and microplastic counted as before but without additional saline water and standardized to give the amount of microlastic per liter of effluent. Washing Machine Effluent. Because the proportions of synthetic fibers found in marine sediments and sewage resembled those used for textiles, we counted the number of fibers discharged into wastewater from using domestic washing machines used to launder clothing. To estimate the number of fibers entering wastewater from washing clothes, 3 different frontloading washing machines (Bosch WAE24468GB, John Lewis JLWM1203 and Siemens Extra Lasse XL 1000) were used (40 C, 600 R.P.M.) with and without cloth (polyester blankets, fleeces, shirts). Detergent and conditioner were not used because these blocked the filter-papers. Cross-contamination was minimized (<33 fibers) at the start of the experiment and in between washes, by running washing-machines at 90 C, 600 R.P.M for 3 cycles without clothes. Effluent was filtered and microplastic counted.1,7
’ RESULTS AND DISCUSSON Eighteen shores across six continents were contaminated with microplastic (Figure 1), and so we investigated whether spatial patterns relate to its sources or sinks. The abundance of microplastic per sample ranged from 2 (Australia) to 31 (Portugal, U.K.) fibers per 250 mL of sediment (Figure 2A), consisting of 9176
dx.doi.org/10.1021/es201811s |Environ. Sci. Technol. 2011, 45, 9175–9179
Environmental Science & Technology
Figure 1. Examples of Fourier transform infrared spectra of microplastic and corresponding reference material from ATR spectral database, vertical axis represents transmission in standard optical density units. (Bruker Optics ATR-Polymer Library - a Collection of Synthetic Fibres, Copyright 2004 Bruker Optic GmbH).
polyester (56%), acrylic (23%), polypropylene (7%), polyethylene (6%), and polyamide fibers (3%). There was more microplastic in densely populated areas24 with a significant relationship between its abundance and human population-density (Linear Regression, F1,16 = 8.36, P < 0.05, n = 18, r2 = 0.34; Figure 2B), but no clear relationship with the mean-size of natural particulates (Spearman Rank rho = 0.39, n = 18, P > 0.05). As a consequence we explored the importance of sewage-disposal as a source of microplastic to marine habitats (Figure 2C). Despite sewage not being added for more than a decade, disposal-sites still contained >250% more microplastic than reference sites (2 Factor ANOVA, F1,16 = 4.50, n = 5, P < 0.05), mainly fibers of
ARTICLE
polyester (78%) and acrylic (22%). To further examine the role of sewage as a source, microplastic was extracted from effluent discharged by sewage treatment plants and compared with sediments from disposal-site. Effluents contained, on average, one particle of microplastic per liter. As expected, polyester (67%) and acrylic (17%) fibers dominated, including polyamide (16%), showing proportions of polyester and acrylic fibers in sewage-effluent resembled microplastic contaminating sediments from shores and disposal-sites. This suggests these microplastic fibers were mainly derived from sewage via washingclothes,26,27 rather than fragmentation1,4,5,7,13 15,18,23 or cleaningproducts.2,7,11,14,16,23 25 Because proportions of polyester fibers found in marine sediments and sewage resembled those used for textiles (78% polyester, 9% polyamide, 7% polypropylene, 5% acrylic),29 we counted the number of fibers discharged into wastewater from using washing blankets, fleeces, and shirts (all polyester). Here we show a garment can shed >1900 fibers per wash. All garments released >100 fibers per liter of effluent, with > 180% more from fleeces (Figure 2E), demonstrating that using washing machines may, indirectly, add considerable numbers of microplastic fibers to marine habitats. Because people wear more clothes during the winter than in the summer30 and washing machine usage in households is 700% greater in the winter,31 we would expect more fibers to enter sewage treatment during the winter. Research is therefore needed to assess seasonal changes in the abundance of plastic fibers in sewage effluent and sludge. In our study it was not possible to use detergent and conditioners because they blocked the filter-papers and prevented us from fitering the samples of effluent, so work is needed to investigate the effect of detergent and conditioner on the quantities of fibers in effluent. Our work provides new insights into the sources, sinks, and pathway of microplastic into habitats. We show polyester, acrylic, polypropylene, polyethylene, and polyamide fibers contaminate shores on a global-scale, with more in densely populated areas and habitats that received sewage. Work is now needed to establish the generality of the relationship with populationdensity at smaller spatial scales, including freshwater and terrestrial habitats where sewage is also discharged. One source of these fibers of microplastic appears to be the disposal of sewage contaminated with fibers from washing clothes because these textiles contain >170% more synthetic than natural fibers29 (e.g., cotton, wool, silk). The quantity of microplastic in sewage and natural habitats is, however, likely to be much greater. Brightly coloured fibers are easily distinguished from natural particulates, but microplastic from cleaning products and fragmentation will be discoloured by biofilms and resemble natural particulates, so better methods are required. In the future microplastic contamination is likely to increase as populations of humans are predicted to double in the next 40 years and further concentrate in large coastal cities17 that will discharge larger volumes of sewage into marine habitats. To tackle this problem, designers of clothing and washing machines should consider the need to reduce the release of fibers into wastewater and research is needed to develop methods for removing microplastic from sewage. One means of mitigation may be ultrafiltration because fewer fibers have been found downstream from a sewage treatment plant that use this process as opposed to one that did not.26 Work is urgently needed to determine if microplastic can transfer from the environment and accumulate in food-webs through ingestion. In humans, inhaled microplastic fibers are taken up by the lung tissues and can become associated with tumors,32 while 9177
dx.doi.org/10.1021/es201811s |Environ. Sci. Technol. 2011, 45, 9175–9179
Environmental Science & Technology
ARTICLE
Figure 2. (A) Global extent of microplastic in sediments from 18 sandy shores and identified as plastic by Fourier transform infrared spectrometry. The size of filled-circles represents number of microplastic particles found. (B) Relationship between population-density and number of microplastic particles in sediment from sandy beaches. (C) Number of particles of microplastic in sediments from sewage disposal-sites and reference-sites at two locations in U.K. (D) Number of polystester fibers discharged into wastewater from using washing-machines with blankets, fleeces, and shirts (all polyester).
dispersive dyes from polyester and acrylic fibers have been shown to cause dermatitis.33 Research is therefore needed to determine if ingested fibers are taken up by the tissues of the gut and release monomers (e.g., ethylene glycol, dimethyl terephthalate, propenenitrile, acrylonitrile, acrylonitrile, vinyl chloride, vinylidene chloride, vinyl bromide), dispersive dyes, mordants (e.g., aluminum, chromium, copper, potassium, tin), 34 plasticisers from manufacture and sorbed contaminants from sewage (e.g., organotin,35 nonylphenol,36 and Triclosan.37 The bioavailability of these chemicals is likely to be greater from fibers of polyester and acrylic, compared to the more hydrophobic microplastics (e.g., polyethylene, polypropylene) that have more heterogenic atoms. In conclusion, our study shows the importance of testing hypothesis to improve our understanding about the sources and sinks of microplastic in habitats. Such experimental approaches are vital if we are to target the pathways of microplastic into habitats with effective mitigation-measures that reduce contamination by microplastic.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +353 (0) 870 916 484. Fax: +353 (0) 1 716 1152. E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank the crews of Aquatay (University of Plymouth) and RV Bernica (University of Newcastle), and CLJ Frid, A. Foggo, A. Dissanayake, A. Silva, M. Day, P. Fredrickson, A. Rubio for assisting with sampling and A. Richardson, D. Barnes for comments on the manuscript. Work was funded by Leverhulme Trust
(grant F/00/568/C), EICC (University of Sydney) and Hornsby Shire Council.
’ REFERENCES (1) Thompson, R. C.; Olsen, Y.; Mitchell, R. P.; Davis, A.; Rowland, S. J.; John, A. W. G.; McGonigle, D.; Russell, A. E. Lost at Sea: Where Is All the Plastic? Science 2004, 304, 5672, 838; DOI: 10.1126/ science.1094559 (2) Browne, M. A.; Dissanayake, A.; Galloway, T. S.; Lowe, D. M.; Thompson, R. C. Ingested microscopic plastic translocates to the circulatory system of the mussel, Mytilus edulis (L.) Environ. Sci. Technol. 2008, 42, 5026-5031; DOI: 10.1021/es800249a (3) Thompson, R. C.; Moore, C. J.; vom Saal, F. S.; Swan, S. H. Plastics, the environment and human health: current consensus and future trends. Philos. Trans. R. Soc., B 2009, 364, 2153-2166; DOI: 10.1098/rstb.2009.0053 (4) Lavender Law, K.; Moret-Ferguson, S.; Maximenko, N. A.; Proskurowski, G.; Peacock, E. E.; Hafner, J.; Reddy, C. M. Plastic accumulation in the North Atlantic Subtropical Gyre. Science 2010, 329, 1185-1188; DOI: 10.1126/science.1192321 (5) Barnes, D. K. A.; Galgani, F.; Thompson, R. C.; Barlaz, M. Accumulation and fragmentation of plastic debris in global environments. Philos. Trans. R. Soc., B 2009, 364, 1985 1998; DOI: 10.1098/rstb.2008.0205 (6) Galgani, F.; F. Leaute, J. P.; Moguedet, P.; Souplet, A.; Verin, Y.; Carpentier, A.; Goraguere, H.; Latrouitee, D.; Andralf, B.; Cadiou, Y.; Mahe, J. C.; Poulard, J. C.; Nerisson, P. Litter on the sea floor along European coasts. Mar. Pollut. Bull. 2000, 40, 516 527; DOI: 10.1016/ S0025-326X(99)00234-9. (7) Browne, M. A.; Galloway, T. S.; Thompson, R. C. Spatial patterns of plastic debris along estuarine shorelines. Environ. Sci. Technol. 2010, 44, 3404 3409; DOI: 10.1021/es903784e. (8) Browne, M. A.; Galloway, T. S.; Thompson, R. C. Microplastic an emerging contaminant of potential concern? Integr. Environ. Assess. Manage. 2007, 3, 559 561; DOI: 10.1002/ieam.5630030412. 9178
dx.doi.org/10.1021/es201811s |Environ. Sci. Technol. 2011, 45, 9175–9179
Environmental Science & Technology (9) Mato, Y.; Isobe, T.; Takada, H.; Kahnehiro, H.; Ohtake, C.; Kaminuma, T. Plastic resin pellets as a transport medium for toxic chemicals in the marine environment. Environ. Sci. Technol. 2001, 35, 308-324; DOI: 10.1021/es0010498. (10) Teuten, E. L.; Rowland, S. J.; Galloway, T. S.; Thompson, R. C. Potential for plastics to transport hydrophobic contaminants. Environ. Sci. Technol. 2007, 41, 7759-7764; DOI: 10.1021/es071737s. (11) Teuten, E. L.; Saquing, J. M.; Knappe, D. R.; Barlaz, M. A.; Jonsson, S.; Bj€orn, A.; Rowland, S. J.; Thompson, R. C.; Galloway, T. S.; Yamashita, R.; Ochi, D.; Watanuki, Y.; Moore, C.; Viet, P. H.; Tana, T. S.; Prudente, M.; Boonyatumanond, R.; Zakaria, M. P.; Akkhavong, K.; Ogata, Y.; Hirai, H.; Iwasa, S.; Mizukawa, K.; Hagino, Y.; Imamura, A.; Saha, M.; Takada, H. Transport and release of chemicals from plastics to the environment and to wildlife. Philos. Trans. R. Soc., B 2009, 364, 2027 2045; DOI: 10.1098/rstb.2008.0284. (12) Gouin, T.; Roche, N.; Lohmann, R.; Hodges, G. A. thermodynamic approach for assessing the environmental exposure of chemicals absorbed to microplastic. Environ. Sci. Technol. 2011, 45, 1466 1472; DOI: 10.1021/es1032025. (13) Lang, I.; Galloway, T. S.; Depledge, M.; Bowman, R.; Melzer, D. Association of urinary bisphenol A concentration with medical disorders and laboratory abnormalities in adults. J. Am. Med. Assoc. 2008, 300, 1303-1310; DOI: 10.1001/jama.300.11.1303. (14) Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection. Proc. GESAMP International Workshop on plastic particles as a vector in transporting persistent, bio-accumulating and toxic substances in the oceans, UNESCO-IOC, Paris, 2010. (15) Galgani, F.; Fleet, D.; van Franeker, F.; Katsanevakis, S.; Maes, T.; Mouat, J.; Oosterbaan, L.; Poitou, I.; Hanke, G.; Thompson, R. C.; Amato, E.; Birkun, A.; Janssen, C. Marine strategy framework directive task group 10 report on marine litter, European Union, Luxembourg, 2010. (16) National Oceanographic and Atmosphere Administration. Proceedings of the International Research Workshop on the Occurrence, Effects and Fate of Microplastic Marine Debris, NOAA, Silver Spring, 2008. (17) United Nations, World Population Prospects: The 2008 Revision Population Database, New York, 2008. www.esa.un.org/unpp (accessed October 1, 2008). (18) Ng, K. L.; Obbard, J. P. Prevalence of microplastics in Singapore’s coastal marine environment. Mar. Pollut. Bull. 2006, 52, 761-767; DOI: 10.1016/j.marpolbul.2005.11.017. (19) Reddy, M. S.; Basha, S.; Adimurthy, S.; Ramachandraiah, G. Description of the small plastics fragments in marine sediments along the Alang-Sosiya ship-breaking yard, India. Estuarine, Coastal Shelf Sci. 2006, 68, 656 660; DOI: 10.1016/j.ecss.2006.03.018/ (20) Barnes, D. K. A. Remote Islands reveal rapid rise of Southern Hemisphere, sea debris. Sci. World J. 2005, 5, 915 921; DOI: 10.1100/ tsw.2005.120/. (21) Centre for Environment, Fisheries and Science. Monitoring and surveillance of non-radioactive contaminants in the aquatic environment and activities regulating the disposal of wastes at sea, CEFAS, Lowestoft, 1997. (22) British Government, Urban Waste Water Treatment (England and Wales) Regulations 1994, London, 1994.http://www.legislation.gov. uk/uksi/1994/2841/regulation/9/made (accessed month day, year). (23) Gregory, M. R. Plastic ‘scrubbers’ in hand-cleansers: a further (and minor) source for marine pollution identified. Mar. Pollut. Bull. 1996, 32, 867 871; DOI: 10.1016/S0025-326X(96)00047-1. (24) Zitko, V.; Hanlon, M. Another source of pollution by plastics: skin cleaners with plastic scrubbers. Mar. Pollut. Bull. 1991, 22, 41 42; DOI: 10.1016/0025-326X(91)90444-W. (25) Fendall, L. S.; Sewell, M. A. Contributing to marine pollution by washing your face: Microplastics in facial cleansers. Mar. Pollut. Bull. 2009, 58, 1225 1228; DOI: 10.1016/j.marpolbul.2009.04.025. (26) Habib, B.; Locke, D. C.; Cannone, L. J. Synthetic fibers as indicators of municipal sewage sludge, sludge products and sewage treatment plant effluents. Water, Air, Soil Pollut. 1996, 103, 1 8; DOI: 10.1023/A:1004908110793.
ARTICLE
(27) Zubris, K. A. V.; Richards, B. K. Synthetic fibers as an indicator of land application of sludge. Environ. Pollut. 2005, 138, 201 211; DOI: 10.1016/j.envpol.2005.04.013. (28) De Wael, K.; Lepot, J.; Gason, F.; Gilbert, B. In search of blood —detection of minute particles using spectroscopic methods. Forensic Sci. Int. 2008, 180, 37-42; DOI: 10.1016/j.forsciint.2008.06.013. (29) Oerlikon. The Fiber Year 2008/09: A world-Survey on Textile and Nonwovens Industry, Oerlikon, Swizerland, 2009. (30) Erlandson, T. M.; Cena, K.; de Dear, R. Gender differences and non-thermal factors in thermal comfort of office occupants in a hot-arid climate. Els. Erg. B. S., 2005, 3, 263-268; doi:10.1016/S1572347X(05)80043-6. (31) Takuma, Y.; Inoue, H.; Nagano, F.; Ozaki, A.; Takaguchi, H.; Watanabe, T. Detailed research for energy consumption of residences in Northern Kyushu, Japan. Energ. Buildings 2006, 38, 1349-1355; DOI: 10.1016/j.enbuild.2006.04.010. (32) Pauly, J. L.; Stegmeier, S. J.; Allaart, H. A.; Cheney, R. T.; Zhang P, J.; Mayer, A. G.; Streck R. J. Inhaled cellulosic and plastic fibers found in human lung tissue. Cancer Epidemiol., Biomarkers Prev. 1998, 7, 419428; DOI: 10.1136/tc.11.suppl_1.i51. (33) Pratt, M.; Taraska, V. Disperse blue dyes 106 and 124 are common causes of textile dermatitis and should serve as screening allergens for this condition. Am. J. Contact. Dermat. 2000, 11, 30-41; DOI: 10.1016/S1046-199X(00)90030-7. (34) Nielson, K. J. Interior textiles: fabrics, application, and historic style, John Wiley & Sons, 2007; p 512. (35) Fent, K. Organotin compounds in municipal wastewater and sewage sludge: contamination, fate in treatment process and ecotoxicological consequences. Sci. Total Environ. 1996, 185, 151-159; DOI: 10.1016/0048-9697(95)05048-5. (36) Ferguson, P. L.; Iden, C. R.; Brownawell, B. J. Distribution and fate of neutral alkylphenol ethoxylate metabolites in a sewage-impacted urban estuary. Environ. Sci. Technol. 2001, 35, 2428-2435; DOI: 10.1021/es001871b. (37) Aguera, A.; Fernandez-Alba, A. R.; Piedra, L.; Mezcua, M.; Gomez, M. J. Evaluation of triclosan and biphenylol in marine sediments and urban wastewaters by pressurized liquid extraction and solid phase extraction followed by gas chromatography mass spectrometry and liquid chromatography mass spectrometry. Anal. Chim. Acta 2003, 480, 193-205; DOI: 10.1016/S0003-2670(03)00040-0.
9179
dx.doi.org/10.1021/es201811s |Environ. Sci. Technol. 2011, 45, 9175–9179
ARTICLE pubs.acs.org/est
Formation of Nanocolloidal Metacinnabar in Mercury-DOM-Sulfide Systems Chase A. Gerbig,*,‡ Christopher S. Kim,§ John P. Stegemeier,§ Joseph N. Ryan,‡ and George R. Aiken† ‡
Department of Civil, Environmental, and Architectural Engineering, University of Colorado, 428 UCB, Boulder, Colorado 80309, United States § School of Earth and Environmental Sciences, Chapman University, One University Drive, Orange, California 92866, United States † U.S. Geological Survey, 3215 Marine Street, Suite E127, Boulder, Colorado 80303, United States
bS Supporting Information ABSTRACT: Direct determination of mercury (Hg) speciation in sulfidecontaining environments is confounded by low mercury concentrations and poor analytical sensitivity. Here we report the results of experiments designed to assess mercury speciation at environmentally relevant ratios of mercury to dissolved organic matter (DOM) (i.e., <4 nmol Hg (mg DOM)1) by combining solid phase extraction using C18 resin with extended X-ray absorption fine structure (EXAFS) spectroscopy. Aqueous Hg(II) and a DOM isolate were equilibrated in the presence and absence of 100 μM total sulfide. In the absence of sulfide, mercury adsorption to the resin increased as the Hg:DOM ratio decreased and as the strength of Hg-DOM binding increased. EXAFS analysis indicated that in the absence of sulfide, mercury bonds with an average of 2.4 ( 0.2 sulfur atoms with a bond length typical of mercury-organic thiol ligands (2.35 Å). In the presence of sulfide, mercury showed greater affinity for the C18 resin, and its chromatographic behavior was independent of Hg:DOM ratio. EXAFS analysis showed mercurysulfur bonds with a longer interatomic distance (2.512.53 Å) similar to the mercurysulfur bond distance in metacinnabar (2.53 Å) regardless of the Hg:DOM ratio. For all samples containing sulfide, the sulfur coordination number was below the ideal four-coordinate structure of metacinnabar. At a low Hg:DOM ratio where strong binding DOM sites may control mercury speciation (1.9 nmol mg1) mercury was coordinated by 2.3 ( 0.2 sulfur atoms, and the coordination number rose with increasing Hg:DOM ratio. The less-than-ideal coordination numbers indicate metacinnabar-like species on the nanometer scale, and the positive correlation between Hg:DOM ratio and sulfur coordination number suggests progressively increasing particle size or crystalline order with increasing abundance of mercury with respect to DOM. In DOM-containing sulfidic systems nanocolloidal metacinnabar-like species may form, and these species need to be considered when addressing mercury biogeochemistry.
’ INTRODUCTION Predicting the fate and transport of soft, chalcophilic metals in the environment depends in part on metal speciation in the presence of sulfide and dissolved organic matter (DOM). The speciation of mercury (Hg) is of particular concern because of the potential formation of methylmercury (especially in sulfatereducing systems 1) and bioaccumulation in aquatic food chains.2 Studies of other metals have identified nanocolloidal metalsulfide minerals in sulfide-containing systems, including ZnS(s) in biofilms3 and at microbial interfaces,4 and CuS(s) in experimentally flooded wetlands5 and experimental systems containing DOM.6 Colloidal mercury-sulfide minerals, particularly metacinnabar (β-HgS(s)), the low-temperature polymorph of HgS(s), have been observed in experimental systems79 and at mining10 and contaminated field sites11 but not in natural sulfate-reducing environments with relatively low mercury concentrations and no point-source contamination.12 Efforts to thermodynamically model the speciation of Hg(II) primarily focus on Hg-DOM complexes in the absence of sulfide r 2011 American Chemical Society
and Hg-sulfide complexes in the absence of DOM. Provided the mercury concentration is sufficiently low, DOM exhibits a high affinity for Hg(II), dominating mercury speciation in typical oxic surface waters.1318 The high strength of Hg-DOM interactions at low Hg:DOM ratios, coupled with directly observed mercurysoil organic matter binding sites,1921 suggests DOM binding sites are thiol-like in nature, although the mercury coordination environment has never been directly observed in aquatic DOM as it has been in soil organic matter. Studies of mercury speciation with sulfide in the absence of DOM show rapid precipitation of metacinnabar22 and a number of dissolved mercury-sulfide complexes (e.g., HgHS2, Hg(HS)20, HgS22, HgS0), of which neutrally charged complexes have been hypothesized to be the most important for methylation.2328 Received: June 1, 2011 Accepted: August 31, 2011 Revised: August 23, 2011 Published: August 31, 2011 9180
dx.doi.org/10.1021/es201837h | Environ. Sci. Technol. 2011, 45, 9180–9187
Environmental Science & Technology Thermodynamic models that suggest mercury-sulfide complexes dominate mercury speciation at low mercury concentrations do not compare well with empirical observations of colloidal HgS(s) stabilized by DOM in experimental systems. In sulfide- and DOM-containing systems with a mercury concentration of 50 μM, metacinnabar particles were observed as particles or aggregates of less than 100 nm in diameter. At 50 nM Hg, the particles or aggregates, if present, were too small to remove via conventional centrifugation.7 Similar work using ultracentrifugation has demonstrated removal of mercury particles from solutions with concentrations as low as 1 nM Hg, although the removed mercury was only definitely characterized as metacinnabar-like at 10 μM Hg.8 The metacinnabar particles formed in the presence of DOM, sulfide, and relatively high concentrations of total mercury (i.e., >10 μM) become coated with DOM, which increases electrostatic repulsion and prevents aggregation and bulk precipitation of metacinnabar.79 The direct observation of DOM-stabilized metacinnabar particles is limited to studies conducted at mercury concentrations far in excess of most natural systems, where only the weakest DOM binding sites are relevant for mercury speciation.13 Speciation calculations, however, suggest that DOM-stabilized HgS(s) may also be present at common environmental levels of mercury, DOM, and sulfide.9 The goal of this study was to empirically determine mercury speciation in DOM-containing solutions with and without free aqueous sulfide at Hg:DOM ratios and total mercury concentrations that are lower than previously studied and span a range of Hg-DOM binding strengths. We adopted a solid phase extraction (SPE) method previously used to determine Hg-DOM binding constants to concentrate hydrophobic mercury species18,29 and applied this method over a wide range of Hg:DOM ratios. The speciation of mercury concentrated by SPE was subsequently examined with extended X-ray absorption fine structure (EXAFS) spectroscopy for samples of selected Hg:DOM ratios. The results presented in this paper provide direct insight into the nature of the Hg-DOM bond and on the role of DOM in mercury speciation in sulfidic environments.
’ METHODS DOM Isolation. Whole water was collected from the F1 site (26°210 3500 N, 80°220 1400 W) in the Florida Everglades, filtered through a 0.3 μm glass fiber filter, acidified to pH 2 with HCl, and passed through a column of Amberlite XAD-8 resin according to the method of Aiken et al.30 The hydrophobic acid fraction (HPoA; comprised of humic and fulvic acids) was retained on the XAD-8 resin and eluted with 0.1 N NaOH. The eluate was hydrogen-saturated, desalted, freeze-dried, and stored for later use. This DOM isolate has been used in several studies of mercuryorganic matter interactions.7,13,14,31 Information on the DOM source and characterization is available elsewhere.12,32 Experimental Solutions. Two identical sets of experimental solutions were prepared a set for experiments only involving SPE and a set for SPE followed by EXAFS analysis of mercury on the resin. Experimental solutions for both sets were prepared in deionized water (g18.0 MΩ cm resistivity) and contained 0.01 M NaH2PO4, enough NaClO4 to bring the ionic strength to 0.1 M (as calculated by Visual MINTEQ33) and an appropriate amount of 0.1 M NaOH to bring the pH to 6.5 ( 0.1. DOM stock solution was prepared daily, filtered (0.45 μm Supor membranes), and added to the experimental solutions to yield a
ARTICLE
DOM concentration of approximately 10 mg L1 for all SPE and most SPE-EXAFS experiments (measured range 8.611.3 mg DOM L1). Some of the SPE-EXAFS experiments were conducted at approximately 50 mg DOM L1. Appropriate volumes of Hg(II) stock solution (Hg(NO3)2 in 10% HNO3) were spiked into the experimental solutions to achieve mercury concentrations ranging from 0.35 nM to 1.4 μM. The range of mercury and DOM concentrations allowed some experiments to be conducted at a Hg:DOM ratio at or below 4 nmol Hg (mg DOM)1, the ratio at which all strong binding DOM sites become saturated and weak-binding sites begin to also bind mercury.13 Sulfidecontaining solutions were prepared in an oxygen-free glovebox. Sodium sulfide (Na2S 3 9H2O; washed before use) stock solution was prepared daily and added to experimental solutions to bring the total sulfide concentration to 100 μM. Solution bottles were wrapped with aluminum foil to prevent photoreactions and allowed to mix at room temperature on a shaker table rotating at 150 rpm. Solutions were equilibrated for 2024 h, which has been shown elsewhere to give sufficient time for Hg-DOM equilibration34,35 and Hg-DOM-sulfide equilibration.9,36 Containers for solution/stock preparation and sampling were glass with Teflon-lined caps cleaned in a solution of 10% HNO3 and 10% HCl (trace metal-grade) for at least 24 h and baked at 400 °C for 4 h. Solid Phase Extraction. The SPE portion of the experiments was carried out on glass columns (10 cm length, 0.9 cm diameter; Spectrum Chromatography) packed with 0.500 g of C18 resin (Supelclean ENVI-18, Spectrapor). The column fittings and lines were Teflon, except for the pump tubing, which was polyvinylchloride. Resin-free columns and tubing were cleaned with a mixture of 10% HNO3 and 10% HCl and rinsed repeatedly with deionized water. Clean resin was prepared in the column by suspending resin in methanol and rinsing (20 min per rinse at 4 mL min1) with deionized water followed by 5 mM HCl, repeating once, and concluding with deionized water. The loss of mercury to a resin-free column (<5%) and contamination from a resin-filled column (<0.03 nM) were sufficiently small to be ignored in the subsequent SPE experiments, but there was some DOC contamination from resin-filled columns (<5 mg C L1; presumably methanol). Cleaned and resin-filled columns were loaded with approximately 1 L of experimental solution for SPE experiments and 2 L of solution for SPE-EXAFS experiments. Experimental solutions were pumped through the cleaned resin-filled columns at a flow rate of 4.0 ( 0.2 mL min1. After expunging the first 2 mL of solution out of each column, the remaining loaded volume was collected as effluent fractions for chemical analyses. Resin was harvested from the column following solution loading and was stored under an oxygen-free atmosphere for sulfide-containing experiments until EXAFS analysis. Mercury recovery from the SPE experiments, including mercury in effluent fractions and mercury adsorbed to the resin, was greater than 90% of the total mercury loaded. Error in the SPE of mercury was related to errors in mercury measurements (described in next section) and depended on the amount of mercury passing through the resin. At high retentions (>90%) the error was less than 1% retained mercury, and at lower retentions (∼60%) the error was approximately 4% retained mercury. Sample Analysis. Dissolved organic carbon (DOC) concentrations were determined using a total organic carbon analyzer (OI Analytical Model 700). DOM concentrations were calculated based on DOC measurements and the carbon content of 9181
dx.doi.org/10.1021/es201837h |Environ. Sci. Technol. 2011, 45, 9180–9187
Environmental Science & Technology the Everglades F1 HPoA isolate (52.2% C by mass). Measurements of ultraviolet and visible light absorbance at wavelengths ranging from 254 to 412 nm were made using a UVvisible spectrophotometer (Agilent model 8453) with a 1 or 5 cm path length quartz cuvette. Total aqueous mercury concentrations in initial and effluent samples from the SPE were determined by cold vapor atomic fluorescence spectroscopy using a Millennium Merlin mercury analyzer according to EPA Method 245.7. Analytical mercury stocks were prepared from National Institute of Standards and Technology (NIST) standard reference material 3133. Mercury standards and most samples were oxidized with 1% (v/v) KBr/ KBrO3 solution. High DOM and sulfide-containing samples were oxidized with 2% (v/v) KBr/KBrO3 solution to ensure sufficient residual oxidant to preserve mercury after oxidation of organic matter and sulfide species. Acceptable recovery of standards was 80120% with less than 20% relative difference in duplicate measurements. Typical recovery was 90110% with less than 10% relative difference. The detection limit for any given run was always below 0.013 nM Hg based on three standard deviations of seven replicates of a sample with a concentration one-half of the lowest standard. Solid phase mercury concentrations on the harvested chromatography resin were measured on a DMA-80 direct mercury analyzer (Milestone Inc.) by thermal decomposition of the sample, catalytic conversion to elemental mercury, amalgamation, and atomic absorption. Calibration was done with a series of standard reference materials obtained from NIST and Environment Canada. Acceptable recovery of the reference materials was 80120%. Extended X-ray Absorption Fine Structure Spectroscopy. Resin samples were prepared for EXAFS by loading 2 L of the experimental solutions outlined in SI Table 1 onto C18 resin. Two liters of solution were necessary to maximize the amount of mercury loaded onto the resin due to the relatively high concentration threshold (approximately 40 ppm Hg) needed to collect viable EXAFS spectra. The top third of the resin in the column was removed from the column and used for EXAFS analysis because solid phase mercury analysis indicated it was more concentrated than the resin in the bottom two-thirds of the column. EXAFS data were collected on wiggler beamline 112 at the Stanford Synchrotron Radiation Lightsource using a Si(220) monochromator crystal in the j = 90° crystal orientation. Mercury LIII-edge EXAFS spectra were collected using an aluminum coldfinger liquid nitrogen cryostat (77 K) to minimize thermal vibration and improve the quality of the spectra from low mercury concentration samples. The resin samples were loaded into aluminum holders in an oxygen-free environment, enclosed in Kapton tape, and quickly transferred to the liquid nitrogen cryostat to minimize exposure to oxygen. Spectra were collected on a 32-element high-throughput germanium detector in fluorescence-yield mode. Gallium filters were used to minimize interference from inelastic scattering. HgCl2 was used as an internal standard for energy calibration of each spectrum collected. Multiple scans (1322) were collected for each sample, energy-corrected using the calibration standard, deadtime-corrected for potential loss of signal due to finite photon detection times, and averaged together. After background subtraction, the data were converted to k-space with a k3-weighting and Fouriertransformed. The EXAFS spectra were fit over a k-range of 2.0 9.5 Å1 using phase and amplitude functions from model singleshell scattering paths generated in SIXPack37 using Feff6l.38
ARTICLE
HgC, HgO, and HgS models (the only realistic first shell interactions in Hg-DOM-sulfide systems) were created and constrained based on the results of the first shell fitting of the resin samples. Mixed interactions were attempted (i.e., HgO and HgS), but single atom interactions consistently proved to be better fits. Given the limited energy range over which spectra were resolvable, only first shell fitting was successfully completed for each resin sample. The scale factor (S0) was fixed at 0.9 for all samples, and the DebyeWaller factor (σ2), which serves as a measure of thermal vibration and static disorder around mercury in the sample, was first allowed to float for all fits; the average DebyeWaller factor for all samples (0.007 Å2) was selected and final fits fixed at this value in order to directly compare fitting results between samples.
’ RESULTS DOM Solid Phase Extraction. The absorption of ultraviolet and visible (UVvis) light was used to track DOM adsorption to the C18 resin because small amounts of methanol contamination in effluent fractions led to erratic DOC measurements. DOM retention by the resin was consistent regardless of mercury concentration (0.35 nM1.4 μM) or the presence or absence of sulfide (Figure 1a). DOM retention decreased as the volume of loaded solution increased (SI Figure 1). The fraction of UVvis-absorbing components retained by the resin increased with increasing wavelength but was generally low less than 35%. The UVvis absorbance of DOM at 254 nm correlates well with the aromaticity of the organic matter,39 although more conjugated molecules are expected to absorb at 412 nm. These data indicate that the more conjugated organic molecules are also somewhat more hydrophobic and preferentially adsorb to the resin. Mercury Solid Phase Extraction. Retention of mercury by C18 resin was a function of mercury concentration and the presence or absence of sulfide (Figure 1b), unlike the retention of DOM. Mercury adsorption to the column did not change substantially through the course of loading up to 1 L of sulfidefree solution (SI Figure 2). The overall efficiency of mercury adsorption from sulfide-free solutions was dependent on the mercury concentration in the loading solution (Figure 1b). At 5.6 nM Hg and below, the retention of mercury was 8591%. At 39 nM Hg and above, retention dropped to 4861%. Retention curves for mercury in systems containing 100 μM total sulfide were distinctly different from those without sulfide (SI Figure 2). The retention of mercury from a sulfide-containing solution with 0.45 nM Hg increased as the total volume loaded increased. In contrast, at mercury concentrations from 1.5 to 490 nM, the mercury adsorption was consistently high (Figure 1b, SI Figure 2). Based on these chromatography results, the mercury species formed in the presence of sulfide at higher mercury concentrations are slightly more hydrophobic (>99% retention at 490 nM) than those formed at lower mercury concentrations (95% retention at 1.5 nM) and substantially more hydrophobic than those formed in the absence of sulfide. For all cases, greater than 60% of the mercury retained by the resin was present in the top one-third of the column based on solid phase analysis. The 0.45 nM Hg and sulfide solution resulted in uncommon chromatographic behavior the retention of mercury increased as the volume of solution loaded onto the resin increased. Such behavior indicates that the sorbent phase becomes more favorable for the sorption of the compound in solution as the amount 9182
dx.doi.org/10.1021/es201837h |Environ. Sci. Technol. 2011, 45, 9180–9187
Environmental Science & Technology
ARTICLE
Figure 2. Mercury retention and DOM retention (as measured by the UV absorbance of DOM components that absorb at 254 nm) for a system without DOM preloading of the resin and a system with DOM preloading of the resin. The system without preloading (0.45 nM Hg, 10.6 mg DOM L1, 100 μM sulfide) was run as a standard chromatography experiment with the mercury-containing solution started at a loaded volume of 0 mL. The preloaded system consisted of DOM preloading (9.8 mg DOM L1, 100 μM sulfide) up to a volume of 428 mL (dashed vertical line) at which time an identical solution equilibrated with mercury (0.40 nM) was loaded on to the resin.
Figure 1. (a) The average fraction of ultraviolet and visible lightabsorbing DOM components retained by the C18 resin as a function of absorbing wavelength for approximately 1 L of eight sulfide-free and five sulfide-containing solutions. Error bars represent 95% confidence intervals for all mercury concentrations. (b) The fraction of total mercury retained on C18 resin for all experiments with DOM (8.611.3 mg DOM L1) and with and without sulfide as a function of total mercury concentration.
of the sorbed compound increases. In the mercury-DOM-sulfide systems in this study, two components are accumulating on the resin—mercury and DOM—and either could be responsible for the increased retention of mercury with loaded volume. Either the adsorption of mercury from solution could promote the sorption of more mercury, which could potentially lead to the formation of mercury species on the resin which are not present in solution, or the adsorption of DOM from solution could promote the adsorption of more DOM along with the bound mercury species. To determine which mechanism was responsible for the increasing mercury retention with increased loading, we compared the retention of mercury from the 0.45 nM Hg, 100 μM sulfide, 10.6 mg DOM L1 solution with the retention of mercury after the resin was preloaded with DOM (Figure 2). A mercury-free preloading solution (9.8 mg DOM L1, 100 μM sulfide, 428 mL) was loaded onto C18 resin and followed with an identical solution that also contained 0.40 nM Hg. The DOM retention was identical in both systems as indicated by the retention of UV254 absorbing components (Figure 2). After preloading the resin with DOM, mercury retention was initially very high (>97%), and the retention did not increase with increased loading volume as observed in the system without preloading. We interpret the difference in mercury retention to mean that mercurymercury interactions were not
driving mercury retention because the DOM-preloaded system showed high mercury retention at the beginning of mercury loading. Had mercury retention increased with volume after DOM preloading, there would have been evidence for mercury mercury interactions, which would have brought into question whether the mercury species on the resin are present in solution. Instead, we surmise that DOM served as a bridge between the mercury species in solution and the resin, and an abundance of DOM on the resin increased mercury affinity for the sorbent phase. Extended X-ray Absorption Fine Structure Spectroscopy. Experimental and fitted mercury LIII-edge EXAFS spectra and the Fourier transforms corresponding to the conditions outlined in SI Table 1 are shown in Figure 3. The EXAFS spectra of the three sulfide-containing systems (Figure 3b, 3c, and 3d) are in phase with one another and out of phase with the sulfidefree sample (Figure 3a). This corresponds with the alignment of the primary Fourier transform features of the sulfide-containing samples (indicated in Figure 3 by a vertical line) and the misalignment of the sulfide-containing samples with the sulfide-free sample. The spectra for samples with added sulfide (3b, 3c, and 3d) were best modeled by a mercurysulfur bond in the first shell with a HgS interatomic distance of 2.512.53 Å ((0.010.02 Å, depending on the sample; Figure 3). The mean sulfur coordination number for the sulfide-containing samples increased with increasing Hg:DOM ratio from 2.3 ( 0.2 sulfur atoms at 1.6 nmol Hg (mg DOM)1 to 3.3 ( 0.2 sulfur atoms at 34 nmol Hg (mg DOM)1. The spectra for the sample without added sulfide (3a) was also best modeled by mercurysulfur bonds in the first shell, despite the absence of added sulfide in the sulfidefree system. The Hg-DOM interaction was fit with a significantly shorter HgS distance of 2.35 ( 0.01 Å and a coordination number of 2.4 ( 0.2 sulfur atoms. 9183
dx.doi.org/10.1021/es201837h |Environ. Sci. Technol. 2011, 45, 9180–9187
Environmental Science & Technology
ARTICLE
Figure 3. k3-weighted mercury LIII-edge EXAFS, Fourier transforms, and fitting results for collected spectra (solid) and fits (dashed) for the four SPEEXAFS samples: (a) sulfide-free at a Hg:DOM ratio of 4.0 nmol Hg (mg DOM)1; (b, c, and d) 100 μM total sulfide and Hg:DOM ratios of 1.9, 4.9, and 34 nmol Hg (mg DOM)1, respectively. Solution chemistries are summarized in SI Table 1. All spectra are best fit by a HgS interaction. The sulfur coordination number (CN) and average bond distance (R) are noted for each sample with the 95% confidence interval ((2σ). The DebyeWaller factor (σ2) was fixed at 0.007 Å2 for all four fits.
’ DISCUSSION Mercury-Dissolved Organic Matter Interactions. In the absence of sulfide and at sufficiently low Hg:DOM ratios, we hypothesized that Hg-DOM binding would be dominated by mercurysulfur interactions because (1) Hg-DOM binding studies have measured large stability constants consistent with thiol-like sites13,15 and (2) Hg-soil organic matter (SOM) studies using EXAFS spectroscopy have detected HgS bonds at low Hg: SOM ratios.20,40 We observed 2.4 coordinating sulfur atoms at 2.35 Å, which is consistent with observations of a 23 sulfur coordination environment in soil organic matter and soil humic acid as detected by X-ray spectroscopy21,40 and pH titrations.19 The HgS interatomic distance is in good agreement with twocoordinate mercury binding environments observed for model thiolates41 and represents the first known direct observation of mercury binding environments in aquatic DOM. Sulfur is a relatively minor element in DOM (1.7 wt % in the isolate used in this study), and the proportion of sulfur that is actually involved in metal binding is low (<2% of the total sulfur in this isolate based on 2.4 atoms/site and a binding capacity of 4 nmol (mg DOM)1). Multiple sulfur atoms per site suggests the possibility that these sites may be (1) of biological origin (e.g., dithiols in protein residues), (2) the result of abiotic sulfide incorporation into DOM, or (3) the result of multiple DOM molecules coordinating a mercury atom. The concentration requirement, or detection limit, of EXAFS restricted identification of the Hg-DOM binding environment to a Hg:DOM ratio of 4.0 nmol Hg (mg DOM)1, which is the strong binding capacity of the DOM isolate.13 Hg:DOM ratios in most environmental settings are typically a few orders of magnitude lower than this strong binding capacity. The chromatographic data suggest we can extrapolate information gained at the Hg:DOM ratio of 4.0 nmol Hg (mg DOM)1 to lower and more environmentally relevant Hg:DOM ratios. The sulfide-free data in Figure 1b show high retention of mercury (>85%) at mercury concentrations below 5.6 nM Hg and lower retention of mercury (<62%) at mercury concentrations above 39 nM Hg.
When normalized to the DOM content of each system (all had approximately 10 mg DOM L1), the transition observed between 5.6 and 39 nM Hg corresponds to a transition between 0.67 and 4.6 nmol Hg (mg DOM)1. Below the 4 nmol Hg (mg DOM)1 strong binding capacity of the DOM, the Hg-DOM complexes are significantly more hydrophobic with respect to the C18 resin than they are above the strong binding capacity. The transition from more to less hydrophobic complexes as the Hg:DOM ratio increases past 4 nmol Hg (mg DOM)1 corresponds to the transition from thiol-like Hg-DOM binding strengths to carboxyllike Hg-DOM binding strengths.13 Both types of complexes are significantly more hydrophobic than the mercury-free DOM, which is only retained at about 20% as measured by the retention of UV254 absorbing components (Figure 1a). The sulfur dominated binding environment observed with EXAFS at 4 nmol Hg (mg DOM)1 is likely only present in a small subset of DOM molecules and that subset is more hydrophobic than other portions of the DOM pool. The chromatography data shown in Figure 1b, coupled with our understanding of DOM binding strengths, suggests that the small number of DOM molecules involved in the directly observed sulfur dominated mercury binding at 4 nmol (mg DOM)1 persist at lower, more environmentally significant Hg:DOM ratios where EXAFS was not possible. Mercury-Dissolved Organic Matter-Sulfide Interactions. The EXAFS spectra from all three sulfide-containing samples were best fit with a HgS scattering interaction at an interatomic distance of 2.512.53 Å (Figure 3). The HgS interatomic distance from all three samples agrees, within uncertainty, with the 2.53 Å HgS distance in crystalline metacinnabar.22 The observed EXAFS spectra are not consistent with cinnabar (which has six coordinating sulfur atoms at three distinct distances22), polymeric HgS species that exhibit two sulfur atoms at a shorter distance of 2.30 Å,26 neutrally charged complexes (i.e., HgS(aq) and HgHSOH(aq)) with a single HgS interaction at less than 2.40 Å,24 nor the Hg-DOM interaction described previously. In addition, simultaneous fits of a HgS scattering path at 2.53 Å indicative of metacinnabar and a HgS scattering path at 2.35 Å 9184
dx.doi.org/10.1021/es201837h |Environ. Sci. Technol. 2011, 45, 9180–9187
Environmental Science & Technology indicative of Hg-DOM complexes showed no significant DOM binding in systems that contained sulfide. The HgS bond distance is independent of Hg:DOM ratio for the three sulfide-containing samples and matches well with metacinnabar, but the sulfur coordination numbers are all lower than the four-coordinate structure of crystalline metacinnabar. The modeled coordination numbers may be explained by imperfectly ordered crystal structures or nanosized HgS(s) particles where under-coordinated mercury atoms on the particle surface comprise a large percentage of all mercury atoms in the phase. The disorder in the particles may even be greater than the coordination number implies because the DebyeWaller factor was fixed in the EXAFS modeling, which implicitly assumes that changes in the spectra were related to changes in coordination number and not the degree of disorder. The modeled coordination number increases with increasing Hg:DOM ratio (Figure 3, samples b, c, and d), which suggests that the Hg:DOM ratio is an important factor in dictating the size or crystalline order of the metacinnabar-like species. Metacinnabar-like species formed at the lowest Hg:DOM ratio resemble the initial phases of metacinnabar crystallization characterized by under-coordinated mercury atoms, whereas the metacinnabar-like species formed at the highest Hg:DOM ratio resembles a structure approaching that of bulk crystalline metacinnabar.22 The interaction of DOM with particle surfaces and subsequent control of particle aggregation has been documented for HgS(s)79 and other metal sulfides and metal oxides (e.g., refs 6 and 42), although the formation of HgS(s) has never been directly observed at the mercury concentrations and Hg:DOM ratios at which mercury is interacting with the strongest DOM binding sites. The strong DOM binding sites are not strong enough to prevent the formation of metacinnabar, but the metacinnabar that forms when mercury speciation is dominated by thiol sites is smaller or less ordered than metacinnabar formed at higher Hg:DOM ratios. Our results show that the portion of DOM interacting with the surface of HgS(s) and preventing growth is more hydrophobic with respect to fractionation on C18 resin than the DOM that remains in solution. The majority of the DOM in our SPE experiments passed through the column, although the portions of the DOM that absorb light at higher wavelengths, which presumably represents greater conjugation, are retained on the resin to a greater degree than the bulk DOM (Figure 1). Additionally, the hydrophobic fraction retained by the resin favors adsorption of metacinnabar-like mercury species (Figure 2). These results are supported by previously observed preferential adsorption of aromatic DOM molecules to colloidal metacinnabar particles.7 EXAFS-derived coordination numbers can be used to estimate particle sizes when they are significantly lower than the coordination number of the bulk phase and the particles are well ordered.43,44 The less-than-ideal sulfur coordination numbers for the three metacinnabar-like samples observed in this study indicate the particles are on the nanometer scale. Referred to as the termination effect, this phenomenon arises when under-coordinated atoms on a particle surface make up a significant fraction of the atoms in the particle. The abundance of under-coordinated atoms drives down the average coordination number for particles in the diameter range of tens of nanometers or less. If we assume that the HgS(s) is perfectly crystalline, the less-than-ideal coordination numbers of the mercurysulfur bonding observed in this study point to particles that are less than 20 nm in diameter, with the smallest (and most under-coordinated) particles as small as just a few nanometers in diameter. In a previous study
ARTICLE
designed to assess the importance of DOM in inhibiting the precipitation of metacinnabar,7 metacinnabar colloids decreased in sized with decreasing mercury concentration. The minimum mercury concentration for direct characterization, 50 μM, led to metacinnabar particles or aggregates less than 100 nm as determined by centrifugation. We have now identified evidence for smaller metacinnabar-like nanoparticles at 50 nM Hg. Our results are also consistent with observations at 10 μM Hg of poorly crystalline and under-coordinated HgS(s) particles that are on the nanometer scale.8 The results presented here contrast with the conclusions of octanolwater partitioning studies which suggest neutrally charged species (e.g., HgS0 and Hg(SH)2) dominate mercury speciation in the presence of sulfide in natural environments.23 DOM significantly alters the octanol partitioning of HgS species when mercury concentrations are as low as 0.1 nM,36 and the partitioning of amorphous metacinnabar-like nanoparticles to octanol has been demonstrated at a mercury concentration of 3 μM.9 Our study bridges the concentration divide by empirically observing a metacinnabar-like species at an intermediate mercury concentration, which chromatography suggests is present at even lower mercury concentrations than could be directly observed with EXAFS. Mercury-sulfide speciation modeling predicts metacinnabar will form at the intermediate and high mercury concentrations used in this study. However, the speciation modeling is ambiguous at the low concentrations in this study below about 4 nM Hg because on the uncertainty in thermodynamic constants (constants for the modeling reproduced in SI Table 2). If metacinnabar is not formed below 4 nM, then speciation modeling predicts hydrophilic complexes (primarily HgHS2) will dominate in these systems (quantitative speciation presented in SI Figure 3). Hydrophilic complexes will not be retained by the C18 resin. Our data show the mercury species in sulfidic systems are consistently retained at high levels by the resin, which indicates that the HgS(s) observed at higher mercury concentrations also dominates at lower, more environmentally relevant concentrations. Modeling efforts elsewhere have shown that uncertainty in thermodynamic constants, particularly the metacinnabar solubility product, makes mercury speciation difficult to predict at environmentally relevant concentrations.9 Our results provide empirical evidence that a discrete inorganic metacinnabar-like phase is stabilized by dissolved organic matter at mercury concentrations and Hg:DOM ratios that are more representative of natural systems. Environmental Implications. Conventional filtration methods are insufficient to diagnose the presence or absence of nanosized particles in the environment; however, the potential exists to use a chromatographic approach to detect the presence of mercury-containing nanoparticles. Hsu-Kim and Sedlak45 noted the adsorption of mercury species to C18 resin when a wastewater effluent sample was exposed to sulfide. As that study and another9 have noted, some of those mercury species are not labile to a strong competing ligand, such as glutathione, whereas dissolved complexes with organic matter are labile. Now that direct observation of mercury speciation has identified metacinnabar (or at least, a metacinnabar-like phase) as a potential source of that nonlabile portion at lower mercury concentrations approaching environmentally relevant concentrations, the potential exists to identify similar species in natural anoxic waters with a chromatographic approach. Knowledge of the speciation of mercury is paramount in assessing the extent and kinetics of the biologically driven conversion of 9185
dx.doi.org/10.1021/es201837h |Environ. Sci. Technol. 2011, 45, 9180–9187
Environmental Science & Technology mercury into methylmercury. While this study does not attempt to determine the role of speciation in methylation, it provides evidence that sulfate-reducing microbes, which typically reside in sulfide- and organic matter-rich environments, are likely to be exposed to disordered, nanoparticulate metacinnabar stabilized by dissolved organic matter. Our results are consistent with the observation of poorly crystalline, nanometer-scale HgS(s) particles in a mercury contaminated site11 and illuminate the role of DOM in HgS(s) formation and stabilization. The mechanism of that stabilization, the rate of nanoparticle growth and aggregation, and the role DOM-coated nanoparticulate metacinnabar plays in methylation are critical areas for further research. In addition, the thermodynamics of a nanoparticulate phase are not necessarily well represented by thermodynamic constants of the bulk phase,46 and thus mercury speciation models may need to account for disordered nanoparticulate HgS(s).
’ ASSOCIATED CONTENT
bS
Supporting Information. Table of the solution composition for the EXAFS samples, a table of thermodynamic constants used to calculate possible mercury speciation, a figure of typical SPE data for DOC and UVvis-absorbing components, a figure of typical mercury SPE, and a figure of modeled mercury speciation. This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (585) 704-8167. E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank J. Moreau for his assistance securing X-ray spectroscopy beamtime. We also thank H. Hsu-Kim at Duke University and B. McCleskey at the U.S. Geological Survey for their critical reviews of the manuscript. This research was supported by the National Science Foundation, grant #EAR-0447386, and the U.S. Geological Survey's Priority Ecosystem Science Program. Portions of this research were carried out at the Stanford Synchrotron Radiation Lightsource, a Directorate of SLAC National Accelerator Laboratory and an Office of Science User Facility operated for the U.S. Department of Energy Office of Science by Stanford University. The use of trade names in this report is for identification purposes only and does not constitute endorsement by the U.S. Geological Survey. ’ REFERENCES (1) Compeau, G. C.; Bartha, R. Sulfate-reducing bacteria - principal methylators of mercury in anoxic estuarine sediment. Appl. Environ. Microbiol. 1985, 50 (2), 498–502. (2) Benoit, J. M.; Gilmour, C. C.; Heyes, A.; Mason, R. P.; Miller, C. L., Geochemical and biological controls over methylmercury production and degradation in aquatic ecosystems. In Biogeochemistry of Environmentally Important Trace Elements, ACS Symposium Series v. 835, Cai, Y., Braids, O. C., Eds.; American Chemical Society: WA, 2003; Vol. 835, pp 262297. (3) Labrenz, M.; Druschel, G. K.; Thomsen-Ebert, T.; Gilbert, B.; Welch, S. A.; Kemner, K. M.; Logan, G. A.; Summons, R. E.; Stasio, G. D.; Bond, P. L.; Lai, B.; Kelly, S. D.; Banfield, J. F. Formation of sphalerite (ZnS) deposits in natural biofilms of sulfate-reducing bacteria. Science 2000, 290 (5497), 1744–1747.
ARTICLE
(4) Moreau, J. W.; Weber, P. K.; Martin, M. C.; Gilbert, B.; Hutcheon, I. D.; Banfield, J. F. Extracellular proteins limit the dispersal of biogenic nanoparticles. Science 2007, 316 (5831), 1600–1603. (5) Weber, F. A.; Voegelin, A.; Kaegi, R.; Kretzschmar, R. Contaminant mobilization by metallic copper and metal sulphide colloids in flooded soil. Nat. Geosci. 2009, 2 (4), 267–271. (6) Horzempa, L. M.; Helz, G. R. Controls on the stability of sulfide sols: Colloidal covellite as an example. Geochim. Cosmochim. Acta 1979, 43 (10), 1645–1650. (7) Ravichandran, M.; Aiken, G. R.; Ryan, J. N.; Reddy, M. M. Inhibition of precipitation and aggregation of metacinnabar (mercuric sulfide) by dissolved organic matter isolated from the Florida Everglades. Environ. Sci. Technol. 1999, 33 (9), 1418–1423. (8) Slowey, A. J. Rate of formation and dissolution of mercury sulfide nanoparticles: The dual role of natural organic matter. Geochim. Cosmochim. Acta 2010, 74 (16), 4693–4708. (9) Deonarine, A.; Hsu-Kim, H. Precipitation of mercuric sulfide nanoparticles in NOM-containing water: Implications for the natural environment. Environ. Sci. Technol. 2009, 43 (7), 2368–2373. (10) Slowey, A. J.; Johnson, S. B.; Rytuba, J. J.; Brown, G. E. Role of organic acids in promoting colloidal transport of mercury from mine tailings. Environ. Sci. Technol. 2005, 39 (20), 7869–7874. (11) Barnett, M. O.; Harris, L. A.; Turner, R. R.; Stevenson, R. J.; Henson, T. J.; Melton, R. C.; Hoffman, D. P. Formation of mercuric sulfide in soil. Environ. Sci. Technol. 1997, 31 (11), 3037–3043. (12) Gilmour, C. C.; Riedel, G. S.; Ederington, M. C.; Bell, J. T.; Benoit, J. M.; Gill, G. A.; Stordal, M. C. Methylmercury concentrations and production rates across a trophic gradient in the northern Everglades. Biogeochemistry 1998, 40 (23), 327–345. (13) Haitzer, M.; Aiken, G. R.; Ryan, J. N. Binding of mercury(II) to dissolved organic matter: The role of the mercury-to-DOM concentration ratio. Environ. Sci. Technol. 2002, 36 (16), 3564–3570. (14) Haitzer, M.; Aiken, G. R.; Ryan, J. N. Binding of mercury(II) to aquatic humic substances: Influence of pH and source of humic substances. Environ. Sci. Technol. 2003, 37 (11), 2436–2441. (15) Lamborg, C. H.; Tseng, C. M.; Fitzgerald, W. F.; Balcom, P. H.; Hammerschmidt, C. R. Determination of the mercury complexation characteristics of dissolved organic matter in natural waters with “reducible Hg” titrations. Environ. Sci. Technol. 2003, 37 (15), 3316– 3322. (16) Benoit, J. M.; Mason, R. P.; Gilmour, C. C.; Aiken, G. R. Constants for mercury binding by dissolved organic matter isolates from the Florida Everglades. Geochim. Cosmochim. Acta 2001, 65 (24), 4445–4451. (17) Han, S. H.; Gill, G. A. Determination of mercury complexation in coastal and estuarine waters using competitive ligand exchange method. Environ. Sci. Technol. 2005, 39 (17), 6607–6615. (18) Hsu, H.; Sedlak, D. L. Strong Hg(II) complexation in municipal wastewater effluent and surface waters. Environ. Sci. Technol. 2003, 37 (12), 2743–2749. (19) Khwaja, A. R.; Bloom, P. R.; Brezonik, P. L. Binding constants of divalent mercury (Hg2+) in soil humic acids and soil organic matter. Environ. Sci. Technol. 2006, 40 (3), 844–849. (20) Skyllberg, U.; Xia, K.; Bloom, P. R.; Nater, E. A.; Bleam, W. F. Binding of mercury(II) to reduced sulfur in soil organic matter along upland-peat soil transects. J. Environ. Qual. 2000, 29 (3), 855–865. (21) Skyllberg, U.; Bloom, P. R.; Qian, J.; Lin, C. M.; Bleam, W. F. Complexation of mercury(II) in soil organic matter: EXAFS evidence for linear two-coordination with reduced sulfur groups. Environ. Sci. Technol. 2006, 40 (13), 4174–4180. (22) Charnock, J. M.; Moyes, L. N.; Pattrick, R. A. D.; Mosselmans, J. F. W.; Vaughan, D. J.; Livens, F. R. The structural evolution of mercury sulfide precipitate: an XAS and XRD study. Am. Mineral. 2003, 88 (89), 1197–1203. (23) Benoit, J. M.; Mason, R. P.; Gilmour, C. C. Estimation of mercury-sulfide speciation in sediment pore waters using octanol-water partitioning and implications for availability to methylating bacteria. Environ. Toxicol. Chem. 1999, 18 (10), 2138–2141. 9186
dx.doi.org/10.1021/es201837h |Environ. Sci. Technol. 2011, 45, 9180–9187
Environmental Science & Technology (24) Tossell, J. A. Calculation of the structures, stabilities, and properties of mercury sulfide species in aqueous solution. J. Phys. Chem. A 2001, 105 (5), 935–941. (25) Paquette, K. E.; Helz, G. R. Inorganic speciation of mercury in sulfidic waters: The importance of zero-valent sulfur. Environ. Sci. Technol. 1997, 31 (7), 2148–2153. (26) Bell, A. M. T.; Charnock, J. M.; Helz, G. R.; Lennie, A. R.; Livens, F. R.; Mosselmans, J. F. W.; Pattrick, R. A. D.; Vaughan, D. J. Evidence for dissolved polymeric mercury(II)-sulfur complexes? Chem. Geol. 2007, 243 (12), 122–127. (27) Lennie, A. R.; Charnock, J. M.; Pattrick, R. A. D. Structure of mercury(II)-sulfur complexes by EXAFS spectroscopic measurements. Chem. Geol. 2003, 199 (34), 199–207. (28) Benoit, J. M.; Gilmour, C. C.; Mason, R. P.; Heyes, A. Sulfide controls on mercury speciation and bioavailability to methylating bacteria in sediment pore waters. Environ. Sci. Technol. 1999, 33 (6), 951–957. (29) Black, F. J.; Bruland, K. W.; Flegal, A. R. Competing ligand exchange-solid phase extraction method for the determination of the complexation of dissolved inorganic mercury(II) in natural waters. Anal. Chim. Acta 2007, 598 (2), 318–333. (30) Aiken, G. R.; McKnight, D. M.; Thorn, K. A.; Thurman, E. M. Isolation of hydrophilic organic-acids from water using nonionic macroporous resins. Org. Geochem. 1992, 18 (4), 567–573. (31) Waples, J. S.; Nagy, K. L.; Aiken, G. R.; Ryan, J. N. Dissolution of cinnabar (HgS) in the presence of natural organic matter. Geochim. Cosmochim. Acta 2005, 69 (6), 1575–1588. (32) Ravichandran, M.; Aiken, G. R.; Reddy, M. M.; Ryan, J. N. Enhanced dissolution of cinnabar (mercuric sulfide) by dissolved organic matter isolated from the Florida Everglades. Environ. Sci. Technol. 1998, 32 (21), 3305–3311. (33) Gustafsson, J. P. Visual MINTEQ, version 3.0; Stockholm, Sweden, 2007. http://www2.lwr.kth.se/English/OurSoftware/vminteq/ (accessed March 1, 2011). (34) Gasper, J. D.; Aiken, G. R.; Ryan, J. N. A critical review of three methods used for the measurement of mercury (Hg2+)-dissolved organic matter stability constants. Appl. Geochem. 2007, 22 (8), 1583–1597. (35) Miller, C. L.; Southworth, G.; Brooks, S.; Liang, L.; Gu, B. Kinetic controls on the complexation between mercury and dissolved organic matter in a contaminated environment. Environ. Sci. Technol. 2009, 43 (22), 8548–8553. (36) Miller, C. L.; Mason, R. P.; Gilmour, C. C.; Heyes, A. Influence of dissolved organic matter on the complexation of mercury under sulfidic conditions. Environ. Toxicol. Chem. 2007, 26 (4), 624–633. (37) Webb, S. M. SIXpack: a graphical user interface for XAS analysis using IFEFFIT. Phys. Scr. 2005, 2005 (T115), 1011. (38) Rehr, J. J.; Mustre de Leon, J.; Zabinsky, S. I.; Albers, R. C. Theoretical x-ray absorption fine structure standards. J. Am. Chem. Soc. 1991, 113 (14), 5135–5140. (39) Weishaar, J. L.; Aiken, G. R.; Bergamaschi, B. A.; Fram, M. S.; Fujii, R.; Mopper, K. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environ. Sci. Technol. 2003, 37 (20), 4702–4708. (40) Nagy, K. L; Manceau, A.; Gasper, J. D.; Ryan, J. N.; Aiken, G. R. Metallothionein-like multinuclear clusters of mercury(II) and sulfur in peat. Environ. Sci. Technol. 2011. (41) Manceau, A.; Nagy, K. L. Relationships between Hg(II)-S bond distance and Hg(II) coordination in thiolates. Dalton Trans. 2008, 11, 1421–1425. (42) Kodama, H.; Schnitzer, M. Effect of fulvic acid on the crystallization of Fe(III) oxides. Geoderma 1977, 19 (4), 279–291. (43) Calvin, S.; Miller, M. M.; Goswami, R.; Cheng, S. F.; Mulvaney, S. P.; Whitman, L. J.; Harris, V. G. Determination of crystallite size in a magnetic nanocomposite using extended x-ray absorption fine structure. J. Appl. Phys. 2003, 94 (1), 778–783. (44) Frenkel, A. Solving the 3D structure of nanoparticles. Z. Kristallogr. 2007, 222, 605–611. (45) Hsu-Kim, H.; Sedlak, D. L. Similarities between inorganic sulfide and the strong Hg(II) - complexing ligands in municipal wastewater effluent. Environ. Sci. Technol. 2005, 39 (11), 4035–4041.
ARTICLE
(46) Gilbert, B.; Banfield, J. F. Molecular-scale processes involving nanoparticulate minerals in biogeochemical systems. Rev. Mineral. Geochem. 2005, 59 (1), 109–155.
9187
dx.doi.org/10.1021/es201837h |Environ. Sci. Technol. 2011, 45, 9180–9187
ARTICLE pubs.acs.org/est
Anomalous Plutonium Isotopic Ratios in Sediments of Lake Qinghai from the Qinghai-Tibetan Plateau, China Fengchang Wu,*,†,^ Jian Zheng,*,§,^ Haiqing Liao,† Masatoshi Yamada,§ and Guojiang Wan‡ †
State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing, 100012, China § Office of Biospheric Assessment for Waste Disposal, National Institute of Radiological Sciences, 4-9-1 Anagawa, Inage, Chiba 263-8555, Japan ‡ State Key Laboratory of Environmental Geochemistry, Institute of Geochemistry of the Chinese Academy of Sciences, Guiyang 550002, China
bS Supporting Information ABSTRACT: The vertical profiles of 239+240Pu and 137Cs activities and 240Pu/239Pu isotopic ratios are determined for three sediment cores of Lake Qinghai from the Qinghai-Tibetan Plateau, China, and compared with those in sediments of another three lakes (Lakes Bosten, Sugan, and Shuangta), the only existing ones closest to Lop Nor area, China’s nuclear weapons test site in the northwestern part of the country. The mean inventory of 47.7 ( 18.7 MBq km2 for 239+240Pu activity in Lake Qinghai is comparable to the average value of global fallout expected at the same latitude, yet the mean inventory of 1112.0 ( 78.0 MBq km2 for 137Cs is slightly lower than that of global fallout. Anomalously low 240Pu/239Pu isotopic ratios (0.0380.125) were found in the 36.5 cm deep sediment layers, indicating the trace Pu input from early nuclear weapons research activities at Atomic City in the lake’s watershed during the 195060s. Model calculation indicated that the Pu input accounted for approximately 516% of the total Pu inventory. The observation of low 240Pu/239Pu ratio in the deep sediment layer provided a new time marker for recent sediment dating in the lake and around the area. The results are of great significance to the further understanding of sources, records, and environmental impacts of global and regional nuclear activities in the environment and provide important chronological information for further studies on the water eutrophication process and climatic change, and reconstruction of pollution history of organic contaminants and heavy metals in the watershed of Lake Qinghai.
’ INTRODUCTION It is of great importance to obtain information on accurate dating of recent lacustrine sediments for environmental studies, such as retrospectively investigating the variation of nutrients, reconstructing the pollution history, and elucidating the variation process of biological productivities. There are mainly two kinds of dating methods for recent lake sediment chronology. One is the 210Pb dating using unsupported 210Pb derived from the atmospheric fallout of 222Rn daughters to calculate the sedimentation rate or accumulation rate.1 That is, however, sustained by the assumptions that either sediment surface activity or flux to the sediment surface is relatively constant.2,3 The other is the use of fallout radionuclides, such as 137Cs, 239+240Pu, and 241Am, in the sediment to obtain chronological information. These radionuclides were assumed to be derived from stratospheric fallout due to the atmospheric testing of nuclear weapons.36 Considering the relatively short half-life of 30.2 years, the sediment dating using 137Cs will be impractical because more than 60% of the original inventory of global fallout 137Cs has decayed.2 The use of global fallout radionuclides for sediment dating is based on the fact that the obvious peak of activity of these r 2011 American Chemical Society
radionuclides can be found in sediments as a time marker for the years 19631964 due to a series of large-scale atmospheric nuclear tests conducted by the former Soviet Union in 19611962.7 A weak activity peak of radionuclides signed to the year of 1953 may be recorded in well-preserved sediment cores due to atmospheric nuclear tests in the early 1950s.8 Deposition of global/local fallout radionuclides derived from nuclear incidents can also contribute reliable time markers to sediment dating. For example, the deposition of 137Cs and Pu resulted from the nuclear plant explosion at Chernobyl in lake sediments provided a time maker of 1986 in some European countries.2,9,10 Moreover, atmospheric fallout of 238Pu was deposited in the Antarctic sediments and ice cores as a result of the burn up of a US satellite, which contained about 1 kg of 238 Pu in auxiliary nuclear power, in 1964.11,12 In addition to activities of radionuclides, the unique atomic ratio of Pu isotopes Received: July 6, 2011 Accepted: September 28, 2011 Revised: September 23, 2011 Published: September 28, 2011 9188
dx.doi.org/10.1021/es202315c | Environ. Sci. Technol. 2011, 45, 9188–9194
Environmental Science & Technology
ARTICLE
Table 1. Typical Plutonium Isotopic Ratios of Various Sources in the Environment 240
Pu/239Pu atom ratio
global fallout
0.178 ( 0.019 (030N)
241
Pu/239Pu activity ratio
238
Pu/239+240Pu activity ratio
2.7
0.025
references 15, 16
0.180 ( 0.014 (3071N) weapon-grade plutonium
0.010.07
5.813
0.0150.42
17, 18
reactor-grade plutonium
0.240.80
>130
>2.12.9
1921
Chernobyl accident
0.310.40
135213
0.490.56
20, 22
Pacific proving ground
0.3060.36
27a
0.0010.014
13, 23
Nagasaki atomic bomb
0.0280.037
1.21b
0.074
24, 25
Thule hydrogen bomb debris Semipalatinsk nuclear test site
0.0551 0.030.05
0.87c
0.0161 0.00640.085
26 27, 28
a 241 +240
Pu/239+240Pu activity ratio, reference date of 241Pu: 19521954. b 241Pu/239+240Pu activity ratio, reference date of 241Pu: August 1945. c 241Pu/239 Pu activity ratio, reference date of 241Pu: October 2001.
and activity ratio of 239+240Pu/137Cs are also useful for the source identifications and dating in the aquatic sediments, especially in lakes close to nuclear testing sites.6,13,14 In particular, the isotopic composition of Pu is characteristic for various Pu sources (Table 1),13,1528 and therefore accepted as a good indicator for identifying Pu sources in the environment. However, information on the distribution, inventory, and sources of fallout radionuclides in China is very limited, which severely hampered the study of recent lake sediment dating needed in China for the reconstruction of the pollution history of organic pollutants and heavy metals. It is known that Chinese atmospheric nuclear tests were conducted at the Lop Nor nuclear test site in northwestern China since 1964. However, the early nuclear weapon research and development activities performed at Atomic City in Lake Qinghai’s watershed remain unknown.29,30 This research center, located in Haibei Prefecture, about 20 km northeast of Lake Qinghai was constructed in 1958 for China’s independent development and assembly of nuclear weapons.29 It was here that China’s first atomic and hydrogen bombs were designed and developed in the 1950s and 1960s. It was completely shut down in 1987 following an official environmental safety evaluation and was officially declassified in 1995. It is now open to the public as a National Patriotism Education Demonstration Base. Since studies are very limited, it is not clear whether there were sediment records of radionuclides resulting from the early nuclear activities in Lake Qinghai. If so, the deposited Pu may bear a unique isotope composition different from that of global fallout source Pu, thus provide a new tool for recent sediment dating in the area. In the present work, in order to illustrate the possible influence of global fallout, the early nuclear activities conducted in the watershed, and close-in fallout from Lop Nor in Lake Qinghai, three sediment cores in the lake are collected, and vertical profiles of 239+240Pu and 137Cs activities and 240Pu/239Pu isotopic ratios are investigated and compared with those of another three lakes (Lakes Bosten, Sugan, and Shuangta), the only existing ones closest to Lop Nor area, China’s nuclear weapon test site. For the first time, the Pu input from the early nuclear research and development activities in Atomic City with a 240Pu/239Pu atom ratio well below typical global atmospheric fallout levels was found. The unique 240Pu/239Pu isotopic ratios and the vertical profiles of 239+240Pu and 137Cs activities in the sediments were found to be important for the further understanding of possible environmental impact of regional nuclear activities and useful for recent sediment dating which is needed for investigation of pollution histories in the lake and its watershed.
Figure 1. Map showing the locations of Lake Qinghai, Lake Bosten, Lake Sugan, and Lake Shuangta, and sampling sites in Lake Qinghai and Atomic City.
’ EXPERIMENTAL SECTION Sample Collection. Lake Qinghai (99360 100160 E,
36320 37150 N) is the largest inland salt lake in China with a lake surface area of 4278 km2, a watershed area of 16 570 km2, a lake surface height of 3193 m above sea level, a maximum depth of 26.5 m, a water pH value of 9.2 on the average, a mineralization degree of 12.3 g L1, and a salinity of 14.2 %.31,32 Lake Qinghai is situated on the Qinghai-Tibetan Plateau in Qinghai Province, China, and is about 1000 km away from Lop Nor, China’s nuclear weapons test site, together with locations of Lake Bosten, 9189
dx.doi.org/10.1021/es202315c |Environ. Sci. Technol. 2011, 45, 9188–9194
Environmental Science & Technology
ARTICLE
Figure 2. Vertical profiles of 239+240Pu and 137Cs activities and 240Pu/239Pu atom ratios in the sediments of Lake Qinghai.
Shuangta, and Sugan. China’s nuclear weapons research and development center is located in Haibei Prefecture, about 20 km northeast of Lake Qinghai in the watershed (Figure 1). Three sediment cores were sampled in Lake Qinghai in 2006 (Figure 1). Three sediment cores were collected in the center area of Lakes Bosten, Sugan, and Shuangta in 20062007, respectively. Each sediment core was sectioned at 0.5-cm intervals in the field. The sediment samples were weighed immediately after collection. They were kept cool until brought to a laboratory and were then freeze-dried, weighed, and ground to 120 mesh for further analyses of 137Cs and 239+240Pu activities and 240Pu/239Pu isotopic ratios. Analytical Procedure. 137Cs activity was determined using gamma-spectrometry on a Canberra S-100 multichannel spectrometer with a GC5019 HP Ge coaxial detector (efficiency 50%) at the Institute of Geochemistry of the Chinese Academy of Sciences. The 137Cs peak used to determine its activity was 661.6 KeV. Liquid standard (Catalog No. 7137) supplied by the Institute of Atomic Energy of the Chinese Academy of Sciences, was used. For 137Cs activity measurements, the relative standard deviation was within 10%.33 Activities of 239Pu and 240Pu were determined at National Institute of Radiological Sciences, Japan. Pu isotope separation and analytical methods were described in our previous reports.34,35 Briefly, 0.52.0 g samples were digested using HNO3 to extract Pu isotopes. AG 1-X8 and AG MP1-M anion resins were used in the separation and purification process. The analytical procedure was characterized by 75% Pu chemical recovery and a U decontamination factor of 105. 239Pu and 240Pu were analyzed using a sector-field ICP-MS (Element 2,
Finngan, Germany) combined with a high efficiency sample introduction system (APEX-Q). Sediment standard reference materials IAEA-368 (marine sediment standards, International Atomic Energy Agency) and SRM-4354 (freshwater lake sediment standards, American National Standards Institute of Technology) were used for analytical method validation. The results obtained for the two reference materials clearly show the suitability of the analytical procedure for the analysis of sediment samples (Table 1 in Supporting Information).
’ RESULTS AND DISCUSSION Vertical Profiles of Pu Isotopes and
137
Cs in Sediments. The vertical profiles of Pu and Cs activities in sediments of Lake Qinghai are characterized by a single-peak distribution pattern (Figure 2). The peak values of 239+240Pu and 137Cs activities appear at the mass depths of 0.459 and 0.716 g cm2 in 2006QH-2 and 2006QH-3, respectively. 239+240Pu activities are as high as 4.74 and 6.99 mBq g1, and 137Cs activities are 94.62 and 112.58 mBq g1, respectively. These vertical profiles are consistent with those found in other lakes in China3,36 and are also similar to those in other lacustrine and marine sediments.3739 This suggests that the single peak of 239+240Pu and 137Cs activities in the sediments corresponds to the 19631964 global fallout. In the sediment from the surface to the layer of peak deposition in 1964, 240Pu/239Pu isotopic ratios range from 0.169 ( 0.007 to 0.228 ( 0.042 (Figure 2; Table 2 in Supporting Information); the inventory-weighted mean 240Pu/239Pu atom ratios are 0.181, 0.175, and 0.170 for 2006QH-1, 2006QH-2, and 239+240
9190
137
dx.doi.org/10.1021/es202315c |Environ. Sci. Technol. 2011, 45, 9188–9194
Environmental Science & Technology
ARTICLE
Figure 3. Vertical profiles of 239+240Pu and 137Cs activities and 240Pu/239Pu atom ratios in the sediments of Lake Bosten, Lake Sugan, and Lake Shuangta (redrawn from Wu et al.6,41).
2006QH-3, respectively. These values are close to 0.180 ( 0.014, the value reported for the global atmospheric fallout.16 This provides further isotopic evidence for the source of 239+240Pu and 137 Cs in the surface sediments, i.e., they originated from global atmospheric fallout. However, in the deep sediment layers below the 1964 peak (below mass depth of 1.60 g 3 cm2 in 2006QH-1, 0.46 g 3 cm2 in 2006QH-2, and 0.72 g 3 cm2 in 2006QH-3, respectively), 240Pu/239Pu ratios range from 0.038 to 0.159, which are significantly lower than those in the surface sediments, and they also decrease as depth increases. Previous studies have shown that 240Pu/239Pu ratios from nuclear weapon-grade materials ranged from 0.01 to 0.07,17 those from nuclear reactors ranged from 0.24 to 0.80, and the global atmospheric fallout prior to 1963 was slightly higher than 0.18.13,21,40 Therefore, the low 240 Pu/239Pu atom ratios in the deep sediment layers may suggest the possible occurrence of weapon-grade Pu input in the lake, which may not originate from the global atmospheric fallout. We also analyzed 241Pu/239Pu activity ratio in sediment cores of 2006QH-2 and 2006QH-3. Although the obtained 241Pu/239Pu activity ratios have relatively high RSD (2568%) due to the low activity of 241Pu, they ranged from 2.35 to 3.04, close to the value of 2.7 for global fallout,15,16 in the sediment layers above the 1964
maximum deposition peak while low values of 0.55 to 2.08 were seen in the layers below the 1964 peak (Table 2 in Supporting Information). The plot of 240Pu/239Pu atom ratio vs 241Pu/239Pu activity ratio for 2006QH-2 and 2006QH-3 sediment cores (Figure 1 in Supporting Information) indicated the mixing of local source Pu input and the global fallout source in Lake Qinghai. Another possibility for the low 240Pu/239Pu atom ratios in Lake Qinghai is the influence of China’s nuclear weapons tests. China’s nuclear weapons test site, where the first nuclear test took place on October 16, 1964, is Lop Nor which is 1000 km from Lake Qinghai.29 240Pu/239Pu atom ratios in the sediments of Lakes Bosten, Sugan, and Shuangta near Lop Nor are generally within the global atmospheric fallout range (Figure 3) except at two incident mass-depth layers of 2.83 g 3 cm2 in 07Bosten 102 (Lake Bosten) and 0.90 g cm2 in 2007SG-2 (Lake Sugan) with ratios of 0.080 ( 0.016 and 0.103 ( 0.009, respectively.6,41 Those two layers are located at or above the 239+240Pu peak layer. The ratios are significantly lower than the global fallout value of 0.180 but slightly higher than the nuclear weapon-grade characteristic value (0.010.07), showing the limited influence of China’s atmospheric nuclear weapon tests. As the observed 240 Pu/239Pu atom ratios in the deep layers of Lake Qinghai 9191
dx.doi.org/10.1021/es202315c |Environ. Sci. Technol. 2011, 45, 9188–9194
Environmental Science & Technology
Figure 4. Inventories of 239+240Pu and 137Cs in the sediments of Lake Qinghai (note: 137Cs activity was corrected back to July 1998; for comparison, the average values of global atmospheric fallout in 3040N latitude bands are also provided).
sediments are even lower than those found in Lake Bosten and Lake Sugan, and the sediment layer of low 240Pu/239Pu atom ratios were below the 1964 peak layer, the possibility of the influence of China’s atmospheric nuclear tests can be excluded. This further suggests that the low Pu ratios observed in the deep layers in Lake Qinghai originated from the early nuclear weapon research and development activities, e.g., trial tests and/or waste discharges made within Atomic City in the watershed. The Net 239+240Pu Inventories from Atomic City. Inventories of 239+240Pu and 137Cs (decay corrected to July 1998) in the sediments of Lake Qinghai are shown in Figure 4. 239+240Pu inventories are 68.4 ( 2.7, 31.9 ( 0.9, and 42.8 ( 0.9 MBq km2 in 2006QH-1, 2006QH-2, and 2006QH-3, respectively, with an average of 47.7 ( 18.7 MBq km2, which is slightly higher than the average value of global atmospheric fallout expected at the same latitude (42 MBq km2).15 Inventories of 137Cs in 2006QH-2 and 2006QH-3 are 933.9 ( 81.1 and 1112.0 ( 78.0 MBq km2, respectively, significantly lower than the average value of global atmospheric fallout at the same latitude (1923 MBq km2, decay corrected to July 1998).42 This may be related to the geographical location of the lake in the area with dry weather, low annual precipitation, and strong prevailing winds, which resulted in the low atmospheric particle settlement.43 A low 137Cs inventory in sediments is also observed in Lake Sugan, where 239+240Pu inventory is lower than the average value of global atmospheric fallout at this latitude (Figure 2 in Supporting Information). The 239+240Pu/137Cs inventory ratios in the sediments of the four studied lakes are higher than the value of global atmospheric fallout (0.029 ( 0.003, decay corrected back to July 1998) (Figure 3 in Supporting Information),44 and the highest value comes from Lake Qinghai although the other three lakes are closer to Lop Nor. In 2006QH-2 and 2006QH-3, 239+240Pu/137Cs ratios are 0.034 and 0.038, respectively. The differential deposition of Pu and 137Cs, the so-called fractionation effect that fallout particles bearing mainly Pu were removed from the nuclear cloud at earlier times or at shorter distances than fallout particles bearing mainly 137Cs, has been reported by Simon et al.,45 and they found that 239+240Pu/137Cs ratio decreased with increasing distance to the nuclear weapons test site in the Marshall Islands. Thus, if the four lakes under discussion were all influenced by nuclear weapons test activities in Lop Nor, 239+240Pu/137Cs ratios in the far-away Lake Qinghai should be lower than those in the near-by Lakes Bosten, Sugan, and
ARTICLE
Figure 5. The relative contributions of 239+240Pu inventories from both global fallout and local input origins.
Shuangta. However, the present observations are just the opposite. This observation further supports above discussion that Lake Qinghai did not receive direct close-in fallout Pu from Lop Nor. The higher 239+240Pu/137Cs ratios in the lake can be attributed to the trace, but unique, Pu input from activities in Atomic City. The two-component mixing model 11,46 was used to estimate contributions of 239+240Pu input from Atomic City to the total Pu inventory, taking 0.18 as the 240Pu/239Pu ratio of global fallout origin and the minimum ratio of 0.038 in the sediments as the local input origin. The contributions are 5.1%, 15.7%, and 9.2% in 2006QH-1, 2006QH-2, and 2006QH-3, respectively, and the 239+240Pu inventories are 3.38, 5.02, and 3.92 MBq 3 km2, with an average of 4.11 ( 0.84 MBq 3 km2 (Figure 5). Based on the lake area, the total 239+240Pu inventory is approximately estimated to be 17.6 ( 3.6 GBq in the lake. During the period from 1964 to 1980, 22 atmospheric nuclear weapons tests were conducted at Lop Nor. Because little information on the Chinese nuclear weapons tests and related nuclear activities is available, the environmental impact, in particular, the possible radioactive contamination in northwestern China, has been a great concern. This study for the first time reveals that Lake Qinghai did not receive significant direct close-in fallout Pu from the Chinese atmospheric nuclear weapons tests at Lop Nor. The Pu input from the early nuclear activities in Atomic City is only 516% of the total Pu inventory, which is close to the expected Pu inventory from global fallout; therefore, the radiation effect on the local population would be expected to be negligible. Dating Sediments. Although the local Pu input recorded in the lake is very small, the unique and significantly low 240Pu/239Pu ratios in the deep sediment layers provide a new indicator for the dating of recent sediments, making it possible to estimate more accurately the sedimentation rate. The lowest 240Pu/239Pu ratios are 0.125 ( 0.018, 0.051 ( 0.019, and 0.038 ( 0.018 in the vertical profiles of 2006QH-1, 2006QH-2, and 2006QH-3, respectively (Table 2 in Supporting Information). These low ratios appear at mass depths of 2.52, 1.25, and 1.77 g cm2 in 2006QH-1, 2006QH-2, and 2006QH-3, respectively (Figure 2). These depths were assumed to correspond to the initial period of nuclear weapon research and development activities in Atomic City in 1958, i.e., the date for the initial local-origin Pu input in Lake Qinghai, and from them the sedimentation rates are calculated as 0.143, 0.099, and 0.125 g cm2 a1 in 2006QH-1, 2006QH-2, and 2006QH-3 between 1958 and 1964. They are almost 9 times the 9192
dx.doi.org/10.1021/es202315c |Environ. Sci. Technol. 2011, 45, 9188–9194
Environmental Science & Technology
ARTICLE
Table 2. Sedimentation rates for Sediment Cores in Different Time-Scales, Together with the Contribution of Pu from Atomic City sediment cores 2006QH-1
time-scale
sedimentation
Pu from Atomic
rates g cm‑2 a‑1
City (MBq km‑2)
19642007
0.032
0.0
19581964
0.143
3.49
2006QH-2
19642007
0.011
0.0
2006QH-3
19581964 19642007
0.099 0.017
5.02 0.0
19581964
0.125
3.92
sedimentation rate from 1964 to the present (0.032, 0.011, and 0.017 g cm2 a1), suggesting a dramatic impact from unprecedented large-scale human activities in the watershed during the 195060s (Table 2). Therefore, the present results suggest that the sole use of 137Cs or 239+240Pu activity profiles for the estimation of recent sedimentation in Lake Qinghai is not reliable, where strong human activities have occurred in the recent decades. If the vertical profiles of 240Pu/239Pu ratios are taken into consideration, more detailed chronological evidence can be obtained for investigating historical nuclear activities in modern lake sediments in China, for reconstructing the pollution history of organic pollutants and heavy metals and for studying environmental changes. It is difficult to resolve the local source from the global fallout and evaluate the influence of local nuclear activities in areas close to nuclear research and/or test sites if only activities of radionuclides were measured. In the present work, at least two sources of plutonium could be identified in Lake Qinghai by using the vertical profiles of 239+240Pu and 137Cs activities, and 240Pu/239Pu isotopic ratios in sediment cores, combined with those in sediments of another three lakes (Lakes Bosten, Sugan, and Shuangta). For the first time, the existence of trace Pu contamination from the early nuclear weapons research and development activities in Atomic City was found in Lake Qinghai. However, from the calculation of Pu inventory of this local source, the radiation effect on the local population can be considered to be negligible. Because of its unique Pu isotopic composition, the trace Pu input recorded in sediments provides important chronological information for further studies on the water eutrophication process and climatic change, and reconstruction of pollution history of organic contaminants and heavy metals in the watershed of Lake Qinghai.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Tel: 0081-43-206-4634; fax: 0081-43-255-0721; e-mail:
[email protected] (J.Z.);
[email protected] (F.W.). Author Contributions ^
F. C. Wu and J. Zheng contributed equally to this study.
’ ACKNOWLEDGMENT This work was jointly supported by the exploratory research fund, National Institute of Radiological Sciences, Japan, Chinese National Basic Research Program (2008CB418200), Natural Science Foundation of China (40903037), and the Committee of National Defense Science & Technology of China (2008-124) and has been partly supported by the Agency for Natural Resources and Energy, the Ministry of Economy, Trade and Industry (METI), Japan. ’ REFERENCES
(1) Goldberg, E. D. Geochronology with 210Pb. In radioactive dating; Vienna: IAEA, 1963; pp 121131. (2) Ketterer, M. E.; Hafer, K. M.; Jones, V. J.; Appleby, P. G. Rapid dating of recent sediments in Loch Ness: inductively coupled plasma mass spectrometric measurements of global fallout plutonium. Sci. Total Environ. 2004, 322, 221–229. (3) Zheng, J.; Wu, F. C.; Yamada, M.; Liao, H. Q.; Liu, C. Q.; Wan, G. J. Global fallout Pu recorded in lacustrine sediments in Lake Hongfeng, SW China. Environ. Pollut. 2008, 152, 314–321. (4) Krishnaswamy, S.; Lal, D.; Martin, J. M.; Meybeck, D. M. Geochronology of lake sediments. Earth Planet. Sci. Lett. 1971, 15, 407–414. (5) Zheng, J.; Liao, H. Q.; Wu, F. C.; Yamada, M.; Fu, P. Q.; Liu, C. Q.; Wan, G. J. Vertical distributions of 239+240Pu activity and 240 Pu/239Pu atom ratio in sediment core of Lake Chenghai, SW China. J. Radioanal. Nucl. Chem. 2008, 275, 37–42. (6) Wu, F. C.; Zheng, J.; Liao, H. Q.; Yamada, M. Vertical distributions of plutonium and 137Cs in lacustrine sediments in northwestern China: quantifying sediment accumulation rates and source identifications. Environ. Sci. Technol. 2010, 44, 2911–2917. (7) Pennington, W.; Cambray, R. S.; Fisher, E. M. Observation on lake sediments using fallout 137Cs as a tracer. Nature 1973, 242, 324–326. (8) Robbins, J. A.; Edgington, D. N. Determination of recent sedimentation rates in Lake Michigan using Pb-210 and Cs-137. Geochim. Cosmochim. Acta 1975, 39, 285–304. (9) Santschi, P. H.; Bollhalder, S.; Farrenkothen, K.; Lueck, A.; Zingg, S.; Sturm, M. Chernobyl radionuclides in the environment: tracers for the tight coupling between atmospheric, terrestrial, and aquatic geochemical processes. Environ. Sci. Technol. 1988, 22, 510–516. (10) Olivier, S.; Bajo, S.; Fifield, L. K.; Gaggeler, H.; Papina, T.; Santschi, P. H.; Schotterer, U.; Schwikowski, M.; Wacker, L. Plutonium from global fallout recorded in an ice core from the Belukha Glacier, Siberian Altai. Environ. Sci. Technol. 2004, 38, 6507–6512. (11) Hardy, E. P.; Krey, P. W.; Volchok, H. L. Plutonium fallout in Utah. In USAEC-Report HASL-257, 1972, pp 1-95. (12) MacKenzie, A. B.; Stewart, A.; Cook, G. T.; Mitchell, L.; Ellet, D. J.; Griffiths, C. R. Manmade and natural radionuclides in north east Atlantic shelf and slope sediments: Implications for rates of sedimentary processes and for contaminant dispersion. Sci. Total Environ. 2006, 369, 256–272. (13) Koide, M.; Bertine, K. K.; Chow, T. J.; Goldberg, E. D. The 240 Pu/239Pu ratio, a potential geochronometer. Earth Planet. Sci. Lett. 1985, 72, 1–8. (14) Ketterer, M.; Watson, B.; Matisoff, G.; Wilson, C. Rapid dating of recent aquatic sediments using Pu activities and 240Pu/239Pu as determined by quadrupole inductively coupled plasma mass spectrometry. Environ. Sci. Technol. 2002, 36, 1307–1311. (15) UNSCEAR. Ionizing Radiation: Sources and biological effects. In United Nations Scientific Committee on the Effects of Atomic Radiation. 1982 Report to the General Assembly, New York, 1982, pp 223238. (16) Kelley, J. M.; Bond, L. A.; Beasley, T. M. Global distribution of Pu isotopes and 237Np. Sci. Total Environ. 1999, 237/238, 483–500. (17) Micholas, N. J.; Coop, K. L.; Estep, R. J. Capability and Limitation Study of DDT Passive-Active Neutron Waste Assay Instrument, Los Alamos National Laboratory, LA-12237-MS, 1992.
9193
dx.doi.org/10.1021/es202315c |Environ. Sci. Technol. 2011, 45, 9188–9194
Environmental Science & Technology (18) Yamamoto, M.; Tsumura, A.; Katayama, Y.; Tsukatani, T. Plutonium isotopic composition in soil from the former Semipalatinsk nuclear test site. Radiochim. Acta 1996, 72, 209–215. (19) Boulyga, S. F.; Becker, J. S. Isotope analysis of uranium and plutonium using ICP-MS and estimation of burn-up of spent uranium in contaminated environmental samples. J. Anal. At. Spectrom. 2002, 17, 1143–1147. (20) Ketterer, M. E.; Hafer, K. M.; Mietelski, J. W. Resolving Chernobyl vs. global fallout contributions in soil from Poland using plutonium atom ratios measured by inductively coupled plasma mass spectrometry. J. Environ. Radioact. 2004, 73, 183–201. (21) Warneke, T.; Croudace, I. W.; Warwick, P. E.; Taylor, R. N. A new ground-level fallout record of uranium and plutonium isotopes for northern temperate latitudes. Earth Planet. Sci. Lett. 2002, 203, 1047–1057. (22) Varga, Z. Origin and release date assessment of environmental plutonium by isotopic composition. Anal. Bioanal. Chem. 2007, 389, 725–732. (23) Muramatsu, Y.; Hamilton, T.; Uchida, S.; Tagami, K.; Yoshida, S.; Robinson, W. Measurement of 240Pu/239Pu isotopic ratios in soils from the Marshall Islands using ICP-MS. Sci. Total Environ. 2001, 278, 151–159. (24) Saito-Kokubo, Y.; Yasuda, K.; Magara, M.; Miyamoto, Y.; Sakurai, S.; Usuda, S.; Yamazaki, H.; Yoshikawa, S.; Nagaoka, S.; Mitamura, M.; Inoue, J.; Murakami, A. Depositional records of plutonium and 137Cs released from Nagasaki atomic bomb in sediment of Nishiyama reservoir at Nagasaki. J. Environ. Radioact. 2008, 99, 211–217. (25) Yamamoto, M.; Komura, K.; Sakanoue, M. Discrimination of the plutonium due to atomic explosion in 1945 from global fallout plutonium in Nagasaki soil. J. Radiat. Res. 1983, 24, 250–258. (26) Eriksson, M.; Lindahl, P.; Roos, P.; Dahlgaard, H.; Holm, E. U, Pu, and Am nuclear signatures of the Thule hydrogen bomb debris. Environ. Sci. Technol. 2008, 42, 4717–4722. (27) Yamamoto, M.; Hoshi, M.; Tanaka, J.; Sekerbaev, A. K.; Gusev, B. I. Plutonium isotopes and 137Cs in the surrounding areas of the former Soviet Union’s Semipalatinsk nuclear test site. J. Radioanal. Nucl. Chem. 1999, 242, 63–74. (28) Yamamoto, M.; Hoshi, M.; Tanaka, J.; Sakaguchi, A.; Apsalikov, K. N.; Gusev, B. I. Distributions of Pu isotopes and 137Cs in soil from Semipalatinsk nuclear test site detonations throughout southern districts. J. Radioanal. Nucl. Chem. 2004, 261, 19–36. (29) Norris, R. S.; Burrows, A. S.; Fieldhouse, R. W. Nuclear Weapons Databook, Vol. V: British, French and Chinese Nuclear Weapons; Westview Press: Boulder, 1994; pp 333336. (30) Wright, D.; Gronlund, L. Estimating China’s production of plutonium for weapons. Sci. Global Secur. 2003, 11, 61–80. (31) Lanzhou Institute of Geology. A comprehensive investigation report of Lake Qinghai; Beijing Scientific Press: Beijing, 1979; pp 1270 (in Chinese). (32) Li, J. G.; Philp, R. P.; Pu, F.; Allen, J. Long-chain alkenones in Lake Qinghai sediments. Geochim. Cosmochim. Acta 1996, 60, 235–241. (33) Wan, G. J.; Chen, J. A.; Wu, F. C.; Xu, S. Q.; Bai, Z. G.; Wan, E. Y.; Huang, R. G.; Yeager, K. M.; Santschi, P. H. Coupling between 210 Pbex and organic matter in sediments of a nutrient-enriched lake: an example from Lake Chenghai, China. Chem. Geol. 2005, 224, 223–236. (34) Zheng, J.; Yamada, M. Inductively coupled plasma sector field mass spectrometry with a high-efficiency sample introduction system for the determination of Pu isotopes in settling particles at femtogram levels. Talanta 2006, 69, 1246–1253. (35) Liao, H. Q.; Zheng, J.; Wu, F. C.; Yamada, M.; Tan, M. G.; Chen, J. M. Precise determination of plutonium isotopes in freshwater lake sediment by ICP-SFMS after separation using ion-exchange chromatography. Appl. Radiat. Isot. 2008, 66, 1138–1145. (36) Huh, C. A.; Chu, K. S.; Wei, C. L.; Liew, P. M. Lead-210 and plutonium fallout in Taiwan as recorded at a subalpine lake. J. Asian Earth Sci. 1996, 14, 373–376. (37) Shen, J.; Zhang, E. L.; Xia, W. L. Records from lake sediments of the Lake Qinghai to mirror climatic and environmental changes of the past about 1000 years. Quat. Sci. 2001, 21, 508–513 (in Chinese) (http://www.dsjyj.com.cn).
ARTICLE
(38) Xu, H.; Ai, L.; Tan, L. C.; An, Z. S. Geochronology of a surface core in the northern basin of Lake Qinghai: Evidence from 210Pb and 137 Cs radionuclides. Chin. J. Geochem. 2006, 25, 301–306. (39) Zeng, Y.; Zhang, X. B.; Zhou, W. J.; Qi, Y. Q. On the source of radioisotope 137Cs in the surface sediments of Lake Qinghai. J. Lake Sci. 2007, 19, 516–521 (in Chinese). (40) Taylor, R. N.; Warneke, T.; Andrew, M. J.; Croudace, I. W.; Warwick, P. E.; Nesbitt, R. W. Plutonium isotope ratio analysis at femtogram to nanogram levels by multicollector ICP-MS. J. Anal. At. Spectrom. 2001, 16, 279–284. (41) Wu, F. C.; Zheng, J.; Liao, H. Q.; Yamada, M. Distribution of artificial radionuclides in lacustrine sediments in China. Radiat. Prot. Dosim. 2011, 146, 291–294. (42) UNSCEAR. Sources and effects of ionizing radiation. In United Nations Scientific Committee on the Effects of Atomic Radiation. 1977 Report to the General Assembly, New York, 1977, pp 3944. (43) Ren, T.; Zhang, S.; Li, Y.; Zhong, Z.; Su, Q.; Xu, C.; Tang, X. Methodology of retrospective investigation on external dose of the downwind area in Jiuquan region, China. Radiat. Prot. Dosim. 1998, 77, 25–28. (44) Hodge, V.; Smith, C.; Whiting, J. Radiocesium and plutonium still together in “background” soils after more than thirty years. Chemosphere 1996, 32, 2067–2075. (45) Simon, S. L.; Graham, J. C.; Borchert, A. W. Concentrations and spatial distribution of plutonium in the terrestrial environment of the Marshall Islands. Sci. Total Environ. 1999, 229, 21–39. (46) Krey, P. W.; Hardy, E. P.; Pachucki, C.; Rourke, F.; Coluzza, J.; Benson, W. K. Mass isotopic composition of global fallout plutonium in soil. Transuranium Nuclides in the Environment; International Atomic Energy Agency: Vienna, Austria, 1976; pp 671678.
9194
dx.doi.org/10.1021/es202315c |Environ. Sci. Technol. 2011, 45, 9188–9194
ARTICLE pubs.acs.org/est
Standard Gibbs Free Energies of Reactions of Ozone with Free Radicals in Aqueous Solution: Quantum-Chemical Calculations Sergej Naumov*,† and Clemens von Sonntag*,‡,§ †
Leibniz-Institut f€ur Oberfl€achenmodifizierung (IOM), Permoserstrasse 15, D-04318 Leipzig, Germany Max-Planck-Institut f€ur Bioanorganische Chemie, Stiftstrasse 3436, P.O. Box 101365, D-45470 M€ulheim an der Ruhr, Germany § Instrumentelle Analytische Chemie, Universit€at Duisburg-Essen, Universit€atsstrasse 5, 45117 Essen, Germany ‡
bS Supporting Information ABSTRACT: Free radicals are common intermediates in the chemistry of ozone in aqueous solution. Their reactions with ozone have been probed by calculating the standard Gibbs free energies of such reactions using density functional theory (Jaguar 7.6 program). O2 reacts fast and irreversibly only with simple carbon-centered radicals. In contrast, ozone also reacts irreversibly with conjugated carbon-centered radicals such as bisallylic (hydroxycylohexadienyl) radicals, with conjugated carbon/ oxygen-centered radicals such as phenoxyl radicals, and even with nitrogen- oxygen-, sulfur-, and halogen-centered radicals. In these reactions, further ozone-reactive radicals are generated. Chain reactions may destroy ozone without giving rise to products other than O2. This may be of importance when ozonation is used in pollution control, and reactions of free radicals with ozone have to be taken into account in modeling such processes.
’ INTRODUCTION There is a fast growing interest in ozone reactions in aqueous solution, as ozone is increasingly used in drinking water treatment and for pollution control in wastewater. For a better mechanistic understanding of ongoing primary processes, the fundamentals of ozone reactions have to be understood. Here, quantum-chemical calculations can be of considerable help.17 In ozone reactions, free radicals are often generated, and their reactions play an important role. Formation and reactions of • OH and O2• are commonly taken into account, but peroxyl radicals, nitrogen- and sulfur-centered radicals, phenoxyl radicals, nitroxyl radicals, and even halogen-derived radicals are also formed in these reactions, and their ozone reactions must also be considered. It will be shown that they all react with ozone. This is in contrast to the most common free radical scavenger, O2. The ground state of O2 is a triplet state [3O2, O2(3∑g)], and interaction with other molecules that are typically in their singlet ground states are spin forbidden and thus very slow. However, reactions with free radicals are spin allowed. Yet this reaction is restricted to a rather small number of radical types, and exceptions dominate (see below). The ground state of ozone, in contrast, is a singlet state, and this allows ozone to react fast with a large number of organic and inorganic compounds. It also reacts with many more free radicals than O2 does. Here, we explore the reason for this with the help of quantumchemical calculations based on density functional theory (DFT). For quantum-chemical calculations, ozone is an especially difficult molecule which has long been considered to be a demanding test case for quantum-chemical methods due to multireference problems,8,9 notably with DFT.10 The quality of calculations of ozone interaction with other molecules such as its reaction with r 2011 American Chemical Society
olefins, which proceeds by a 1,3-cycloaddition reaction,11,12 must thus be considered with some doubt. In contrast, this problem falls away in the calculations of the structures and standard Gibbs free energies of formation of ozone adducts to closed-shell molecules and free radicals, since here the multireference problem no longer prevails. For example, the reaction of ozone with benzene and its derivatives leads to a zwitterionic adduct as an intermediate (potential minimum), and there is an excellent correlation of the standard Gibbs free energy of ozone adduct formation with the logarithm of the ozone rate constant.5 The latter varies by 10 orders of magnitude between nitrobenzene and phenolate ion. Moreover, standard Gibbs free energy calculations may lead one to realize unexpected rearrangements of ozone adducts as has been the case with the ozone adducts to hypobromite4 and bisulfide ions.7 Also in a multifaceted system such as the reaction of ozone with nitrite, such calculations substantiate mechanistic suggestions.6 Moreover, DFT calculations allowed correction of the pKa of HO3• (reported at 6.15,13 revised to 8.2, B€uhler, private communication) to a much lower value (pKa ≈ 2.0).14 DFT calculations also contributed in reassessing the mechanism of the decay of O3•, a primary free radical intermediate in many ozone reactions, into •OH.2,14 Standard Gibbs free energies are related to the equilibrium constant K of a given reaction [ΔG° = RT ln(K)]. Thus, endergonic reactions still take place at a substantial rate as long as the reverse reaction is fast. In the linear free energy relationship concept, use is made of the observation that the more exergonic Received: May 27, 2010 Accepted: September 12, 2011 Revised: September 12, 2011 Published: September 13, 2011 9195
dx.doi.org/10.1021/es2018658 | Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology
ARTICLE
Table 1. Reactions of Reducing Free Radicals with Ozone in Aqueous Solution: Compilation of Products, Rate Constants (M1 s1), and Calculated Standard Gibbs Free Energies (ΔG°, kJ mol1) of the Overall Reaction and of RadicalOzone Adducts (if Formed) and Their Decay educts eaq, •
O3
H, O3
CO2•, O3 CO2•, O3 O2•, O3 a
final products O3
•
•
OH, O2 •
reaction
rate constant
overall reaction
adduct formation a
1
3.6 10
(374; pH > 11)
2/3
2 1010 23
411
373 b
10 23
see ref 2
4
315
CO3•, O2
5
370
b
O3•, O2
6
176
b
CO2, O3
1.5 109 23 1.6 109 13
adduct decay 38 [47 14]
Calculated from reported reduction potentials.22 b An adduct cannot be calculated as a potential minimum on the reaction path.
the reaction the higher its rate. For some ozone reactions this concept works perfectly.5 Yet this holds only for a series of similar reactions. With markedly different reactions, such as addition vs electron transfer or H-abstraction, kinetics may overrun thermodynamics and the thermodynamically favored reaction may be barely observed. An example is given in the Supporting Information. With this caveat in mind, it is hoped that standard Gibbs free energies calculations, properly taken into perspective, will provide us with useful information on the reactions of free radicals with ozone.
’ THEORETICAL SECTION DFT calculations were carried out using the B3LYP hybride functional15,16 (Jaguar 7.6 program).17 The molecular geometries of all calculated molecules were optimized in water at the B3LYP/LACV3P**+ level of theory. The LACV3P**+ basis set uses the standard 6-311G**+ basis set for light elements and the LAC pseudopotential18 for third-row and heavier elements (in our case Br). For a given compound (intermediate), various structures were explored at a lower level of theory, and then the best one was calculated at the highest level of theory. The interactions between the molecule and the solvent were evaluated by Jaguar’s PoissonBoltzmann solver (PBF).19 Frequency calculations were done at the same level of theory to characterize the stationary points on the potential surface and to obtain standard Gibbs free energies (G°), which were calculated at a standard temperature of 298.15 K and a pressure of 1 atm using unscaled frequencies. Reaction parameters were calculated as the difference of the standard Gibbs free energies ΔG° between reactants and products. It is difficult to estimate the errors that may be involved in our calculations. In absolute terms, thermochemical data for the reaction CO2• + O3 f CO3• + O2 (ΔG° = 335.8 kJ mol1) and our calculated value (ΔG° = 370 kJ mol1) differ by 34 kJ mol1. Yet, in relative terms, this difference is only 10%, and this agreement is quite good. When the above reaction is compared with an electron transfer reaction, the difference between thermochemical and calculated values is only 5 kJ mol1 (see the Supporting Information). The standard Gibbs free energies of the reaction HO3• f •OH + O2 obtained by Jaguar and G1 differ by 9 kJ mol1 (cf. Table 1). G1 is the most reliable method for calculating standard Gibbs free energies of free radical reactions; that is, the values come closest to those of experiments.1 It is, however, restricted to very small and uncharged molecules/radicals and thus cannot be used here. With circumneutral reactions it seems fair to estimate an error of (10 kJ mol1, and this is, of course, much more than 10% on relative terms.
’ RESULTS AND DISCUSSION There are a number of radicals whose ozone rate constants have been determined. These vary by 7 orders of magnitude and range from 103 M1 s1 to diffusion-controlled (k ≈ 1010 M1 s1) (cf. Tables 15). As mentioned in the Introduction, there is increasing evidence, including experimental evidence,20,21 for ozone adducts as the first intermediate in ozone reactions. Such adducts may also play a role in reactions of ozone with free radicals (R• + O3 f ROOO•). For some of these, a potential minimum is reached when the structures of such adducts are optimized and the standard Gibbs free energy of their formation can be calculated. Such adducts are regarded as intermediates rather than as transition states. In other cases, no potential minimum is reached, and during energy optimization the reaction proceeds to the products. “Adducts” are then only bona fide transition states. In the following, the ozone reactions of the various groups of radicals are discussed. Reactions of Ozone with Reducing Radicals. The oneelectron reduction potential of ozone is +1010 mV (at pH > 11),22 and electron transfer reactions with reducing radicals such as eaq (2870 mV), •H (2400 mV), CO2• (1900 mV), and O2• (330 mV) are feasible. Rate constants and standard Gibbs free energies for the formation of adducts, if formed, and products are compiled in Table 1. The hydrated electron (eaq) adds to ozone, reaction 1, and subsequent decomposition of O3• into O• plus O2 is measurably slow (for the rate and equilibrium constants see ref 24). O3 þ eaq f O3 •
ð1Þ
For the •H atom, an adduct (reaction 2), subsequently decaying into •OH and O2 (reaction 3), is formed (for equilibrium 3 see ref 14). In competition with reaction 3, HO3• may deprotonate since HO3• is a very strong acid (pKa(HO3•) = 2.0).14 O3 þ • H f HOOO• a • OH þ O2
ð2=3Þ CO2•,
In the calculations of the reaction of ozone with an adduct is not observed, and two competing reactions may occur, an O-transfer (reaction 4) and an electron transfer (reaction 5). Electron transfer, although also markedly exergonic, is not energetically favored, the difference from reaction 4 being ΔG° = +55 kJ mol1 (cf. Table 1). This is in good agreement with thermokinetic data that yield ΔG° = +50 kJ mol1 (see the Supporting Information). CO2 • þ O3 f CO3 • þ O2
ð4Þ
CO2 • þ O3 f CO2 þ O3 •
ð5Þ
It is re-emphasized that kinetics may determine the route that is taken and the thermodynamically favored product may be 9196
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology
ARTICLE
Table 2. Reactions of Carbon-Centered Free Radicals and Phenoxyl Radicals with Ozone in Aqueous Solution: Compilation of Products, Rate Constants (M1 s1), and Calculated Standard Gibbs Free Energies (ΔG°, kJ mol1) of the Overall Reaction and of RadicalOzone Adducts (if Formed) and Their Decay
a
An adduct cannot be calculated as a potential minimum on the reaction path.
disfavored when both reactions are of a different reaction type and both are thermodynamically feasible. Examples for free radical reactions have been discussed in ref 25 and for an ozone reaction in ref 1. A further example is given in the Supporting Information. Thus, additional information is often required for assessing the preferred route in competing exergonic reactions. In the reaction of CO2• with ozone, no chain reaction of any importance is observed, and it has been concluded that the electron transfer reaction 5 can be neglected.26 The reduction potential of O2• is the least negative of these four reducing radicals, and the rate constant is an order of magnitude slower than those of eaq and H•. No adduct is observed; the standard Gibbs free energy for the overall reaction 6 is lower than that for the reaction with CO2• (Table 1), but as there is no competing reaction, electron transfer is the only process. O2 • þ O3 f O2 þ O3 •
very fast, 2 109 M1 s1 (Table 2), if the acetate radical, CH2C(O)O, is a good guide.27 Calculations have been carried out for the methyl radical and the acetate radical. For the methyl radical an adduct with a CH3OOO• bond length of 1.58 Å can be calculated, reaction 7, that subsequently decays into methoxyl and O2 (reaction 8).
•
•
ð7=8Þ
With the acetate radical, an adduct as a potential minimum along the reaction path could not be established. The reaction of the acetate radical is very fast (Table 2), and other alkyl radicals may react similarly fast. The high driving force for reaction 7 is compatible with this. In water, the methoxyl radical undergoes a rapid (water-catalyzed) 1,2-H shift (reaction 9, k ≈ 106 s1, ΔG° = 34 kJ mol1).28,29 CH3 O• f • CH2 OH
ð6Þ
Reactions of Ozone with Carbon-Centered Radicals. The reaction of ozone with simple carbon-centered radicals must be
CH3 þ O3 f CH3 OOO• f CH3 O• þ O2
ð9Þ
Moreover, strongly branched alkoxyl radicals undergo a similarly fast β-fragmentation.28 Thus, alkoxyl radicals are too shortlived in water for reactions with ozone to occur, and their reactions with ozone (see below) are not of importance in practice. 9197
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology
ARTICLE
An interesting group of radicals are the hydroxycylohexadienyl radicals that are formed upon •OH attack on benzene and its derivatives (note that in the ozonation of a wastewater 13% of ozone is converted into •OH 30 that reacts predominantly with the aromatic moieties of the organic matter). In contrast to O2 that reacts reversibly with these radicals (see below), ozone must react irreversibly (reaction 10) without a detectable adduct as an intermediate. In this respect, the hydroxycyclohexadienyl radicals are similar to the acetate radical but differ from the methyl radical, where an, albeit very labile, adduct has been calculated (cf. reactions 7/8).
Phenoxyl radicals are commonly written with the free spin at oxygen, but the overwhelming spin density is at carbon at the ortho and para positions of the benzene ring (see Figure S1, Supporting Information). Recombination of two phenoxyl radicals at oxygen is strongly endergonic, but recombination may occur by forming CO and CC linkages3133 (see Figure S2, Supporting Information). Thus, phenoxyl radicals largely behave as carbon-centered radicals. Although ozone addition at oxygen and concomitant O2 loss is exergonic (reaction 11), an addition at carbon (cf. reaction 12) is even more so (Table 2).
Ozone is an electrophilic agent,34 and addition at carbon seems to be more likely. Reaction 12 may hence be preferred over reaction 11. Further O2 loss from the phenylperoxyl radical is markedly endergonic (Table 2), and such peroxyl radicals are indeed well documented.35,36 No adducts have been detected as intermediates in reactions 11 and 12. The high reactivity of phenoxyl radicals with ozone is in contrast to their reaction with O2 (see below), and as phenoxyl radicals are important intermediates in the reactions of ozone with phenols,37,38 their reaction with ozone may be of some relevance. Reactions of Ozone with Oxygen-Centered Radicals. The reaction of ozone with O2• has been discussed above. The reaction with the •OH radical is fast,39 but below diffusion controlled (Table 3). The reaction has been formulated as an O-transfer (reaction 13). An adduct could not be calculated. •
OH þ O3 f HO2 • þ O2
ð13Þ
In water, alkoxyl radicals are too short-lived (see above) for reacting with ozone. Yet reaction 14 is of some interest in comparison with the analogous reaction of •OH (reaction 13). A somewhat lower standard Gibbs free energy has been calculated (Table 3). Again, no adduct is detectable. CH3 O• þ O3 f CH3 OO• þ O2
With the methylperoxyl radical, a short-lived adduct can be calculated (reaction 15) that subsequently releases O2 (reaction 16); cf. Table 3. CH3 OO• þ O3 a CH3 OOOOO• f CH3 O• þ 2O2
With all the other peroxyl radicals, adducts could not be established, and the reactions run through to final products. Nitroxyl radicals also belong to the group of oxygen-centered radicals. The stable nitroxyl radical TEMPO that is often used as an ESR standard reacts with ozone with an observed rate constant kobsd = 1 107 M1 s1.40 In the first step, one molecule of O2 is released (reaction 17) without a detectable adduct as an intermediate. The nitrogen-centered peroxyl radical thus formed is unstable and loses O2 (reaction 18). For standard Gibbs free energies see Table 3.
In its reaction with ozone, the morpholine-derived nitroxyl radical shows the same overall standard Gibbs free energy for the release of two molecules of O2 (Table 3). In contrast to the TEMPO system, an intermediate peroxyl radical cannot be calculated here and must be a transition state rather than a very short-lived intermediate. Reactions of Ozone with Nitrogen- and Sulfur-Centered Radicals. Nitrogen-centered radicals are likely intermediates in the reactions of ozone with primary and secondary amines, and their reactions may contribute to product formation in these systems. An example is shown for the reaction of ozone with morpholine (reactions 19 and 20), for which an •OH yield near 33% has been found (A. Tekle-R€ottering, private communication). The precursor of •OH is O3• (for its decay into •OH, see above), and material balance considerations require that an equivalent amount of aminyl radicals must be formed as well.
Here, we have carried out calculations on the nitrogen-centered radical formed in the reaction of ozone with TEMPO (cf. reaction 18) and in the morpholine system (reactions 19/20). The reaction of ozone with these radicals (reaction 21) is highly exergonic (Table 4) and proceeds without an adduct that is sufficiently stable to be calculated as an intermediate. R 2 N• þ O3 R 2 NO• þ O2
ð14Þ
A number of peroxyl radicals have been generated radiolytically, and their reactions with ozone have been studied.26 From an evaluation of the data, rate constants have been calculated that are compiled in Table 3. The rate constants of the carboncentered peroxyl radicals with ozone center near 104 M1 s1, and that of O3SOO• is an order of magnitude higher (Table 3).
ð15=16Þ
ð21Þ
Rate constants for the reaction of aminyl radicals with ozone are as yet not known, but the exergonicity of this reaction is remarkable (Table 4), and this reaction may be quite fast. Combined with reactions such as reactions 17/18, it would give rise to a chain reaction that consumes ozone without giving rise to products other than O2. Such a cycle has been suggested to play a 9198
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology
ARTICLE
Table 3. Reactions of Oxygen-Centered Free Radicals with Ozone in Aqueous Solution: Compilation of Products, Rate Constants (M1 s1), and Calculated Standard Gibbs Free Energies (ΔG0, kJ mol1) of the Overall Reaction and of RadicalOzone Adducts (if Formed) and Their Decay
a
An adduct cannot be calculated as a potential minimum on the reaction path.
role in the ozone chemistry of diclofenac for explaining why the diclofenac consumption by ozone is so surprisingly low.41 Yet kinetic data that would substantiate this suggestion are still missing. Aminyl radicals may rearrange into the corresponding carboncentered radicals, e.g., reaction 22 (ΔG° = 31 kJ mol1). CH3 CH2 • NH f CH3 • CHNH2
ð22Þ
Such reactions may be slower than the corresponding 1,2-H shift reaction of alkoxyl radicals (ΔG°(ethoxylfhydroxyethyl) = 36 kJ mol1), as the driving force is lower by 5 kJ mol1. Such rearrangements may shorten the above chain reaction.
Thiols react very fast with ozone without giving rise to thiyl radicals (an electron transfer would be endergonic). Yet the thiyl radical/ozone reaction is mechanistically of interest in comparison with the corresponding reaction of O2 discussed below. Reaction 23, without an adduct as a potential minimum on the reaction path, is highly exergonic (Table 4). CH3 S• þ O3 f CH3 SO• þ O2
ð23Þ
Reactions of Ozone with Halogen-Containing Radicals. The reactions of ozone with halide ions are complex (see the Supporting Information) and do not directly give rise to free radicals. Only when the oxidation proceeds to chlorite42 and 9199
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology
ARTICLE
Table 4. Reactions of Nitrogen- and Sulfur-Centered Free Radicals with Ozone in Aqueous Solution: Compilation of Products and Calculated Standard Gibbs Free Energies (ΔG0, kJ mol1) of the Overall Reaction
bromite, 43 ozone-induced free radical formation sets in (cf. reaction 24). ClO2 þ O3 f ClO2 • þ O3 •
ð24Þ
Yet the reaction of •OH with Br that gives rise to Br• is very high (reaction 25, k = 1.1 1010 M1 s1 44), and the •OH route in ozone reactions can contribute to bromate formation in drinking water ozonation.45 •
OH þ Br f Br• þ OH
ð25Þ
•
In neutral solutions, the reaction of Cl with OH gives rise to a three-electron-bonded intermediate that only in acid solution is converted into Cl• (equilibrium 26, K = 0.7 M1 46 and reaction 27). •
OH þ Cl a HOCl•
HOCl• þ Hþ a H2 O þ Cl•
ð26Þ
The Cl2• reaction resembles the reaction of •OH with ozone. In both cases, rate constants are well below the diffusioncontrolled limit (Table 5) despite the high exergonicity of these reactions. The reasons for this are not yet fully understood, but the pronounced electrophilicity of ozone may play a role in preventing a favorable transition state in reactions of nitrogen-, oxygen- and halogen-centered radicals (note that the reactions of TEMPO and peroxyl radicals with ozone are also below diffusion controlled). For the subsequent oxidation of the ClO• radical by ozone, two exergonic reactions may be written. Reaction 32 gives rise to chlorine dioxide, OClO•, but a reaction to its isomer, the chlorineperoxyl radical, ClOO•, is even more exergonic (reaction 33, Table 5). Missing additional information, a decision as to which of them is kinetically favored cannot be made.
ð27Þ
In the presence of Cl, equilibrium 28 (K = 6 104 M1,47 1.4 10 M1 48) has to be taken into account. 5
Cl• þ Cl a Cl2 •
Br• þ Br a Br2 •
ð29Þ
The potential importance of halogen-derived radicals under ozonation conditions justifies having a closer look at their reactions with ozone. For rate constants and calculated standard Gibbs free energies, see Table 5. The •Cl atom undergoes an O-transfer without an adduct as a potential minimum on the reaction path (reaction 30). A rate constant has not been measured. As expected, the reaction becomes less exergonic when Cl• is complexed to Cl (reaction 31, Table 5). Cl• þ O3 f ClO• þ O2
ð30Þ
Cl2 • þ O3 f ClO• þ O2 þ Cl
ð31Þ
ð32Þ
ClO• þ O3 f ClOO• þ O2
ð33Þ
The peroxyl radical ClOO• is highly unstable (equilibrium 34).
ð28Þ
A similar equilibrium can be written for Br• plus Br (equilibrium 29, K = 3.9 105 49), and there must also be an analogous mixed three-electron-bonded intermediate, BrCl•.
ClO• þ O3 f OClO• þ O2
ClOO• a Cl• þ O2
ð34Þ
The species that has been obtained with a potential minimum in the calculations has a ClOO• bond length of 2.331 Å and ClOO• bond length of 1.888 Å. The OdO bond length is practically identical, 1.886 Å. This indicates that the ClOO• species is a kind of van der Waals complex of Cl• with O2. The value of ΔG° = +5 kJ mol1 can thus most likely not be related to the equilibrium constant of reaction 34. This is in contrast to the analogous equilibrium HOOO• a •OH + O2 (ΔG° = 47 kJ mol1).14 Here, HOOO• (1.520 Å) and HOOO• (1.243 Å) bond lengths are quite different from those of the final products, and the HOOO• species thus characterized is not a mere van der Waals complex. The oxidation of chlorine dioxide by ozone (reaction 35) is only moderately fast (Table 5), and this is compatible with the reaction being slightly endergonic. OClO• þ O3 f ClO3 • þ O2 9200
ð35Þ
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology
ARTICLE
Table 5. Reactions of Halogen-Containing Free Radicals with Ozone in Aqueous Solution: Compilation of Products, Rate Constants (M1 s1), and Calculated Standard Gibbs Free Energies (ΔG0, kJ mol1) of the Overall Reaction and of RadicalOzone Adducts (if Formed) and Their Decay educts •
•
reaction
rate constant
overall reaction
adduct formation
204
a
133
a
Cl, O3
ClO , O2
30
Cl2•, O3
ClO•, O2, Cl
31
ClO•, O3
OClO•, O2
32
119
a
•
ClO , O3
ClOO•, O2
33/34
229
+5; see the text
OClO•, O3
ClO3•, O2
35
OClO•, O3
ClO•, 2 O2
2 ClO3
•
(9 ( 0.7) 107 50
(1.05 ( 0.1) 103 42 1.1 103 51
+9
a
36
311
a
adduct decay
O3ClClO3
37
+109
2 ClO3•
ClO4, ClO2+
38/39
40
OClO• + ClO3•
O3ClClO2
41
+93
OClO• + ClO3•
O2ClOClO2
42
OClO• + ClO3•
ClO• + O2 + OClO•
43/44
ClO3• + ClO ClO3• + ClO2
ClO3 + ClO• ClO3 + OClO•
Br•, O3
BrO•, O2
45
149
a
BrO , O3
BrOO•, O2
46
268
a
BrOO•
Br• + O2
47
11
a
BrO•, O3
OBrO•, O2
48
47
a
2 OBrO•
OBrOOBrO
49
+11
2 OBrO•
O2BrOBrO
50
+130
OBrOOBrO OBrO•, O3
2 •Br, 2 O2 BrO• + 2 O2
51 52
282 380
OBrO•, O3
O2BrO•, O2
53/54
+81
+102
21
O2BrO , O3
BrO4•, O2
55/56
+93
+114
21
BrO3•
BrO•, O2
57
506
BrO4•
OBrO•, O2
58
413
•
•
a
final products
a
+29 222
a
144 162
An adduct cannot be calculated as a potential minimum on the reaction path.
An ozone addition to one of the oxygens may give rise via a tetroxide as a transition state to OCl• plus 2O2 (reaction 36). OClO• þ O3 f OCl• þ 2O2
ð36Þ
The overall reaction is strongly exergonic (Table 5). Yet reaction 35 may compete successfully in case there is substantial activation energy for reaction 36. Formation of 1O2 in uncommonly high yields (150%) has been reported for CCl4 solutions.52 The energetics of some of the reactions discussed above would also allow the formation of 1O2, but yields above 100% are difficult to reconcile. The ClO3• radical cannot be further oxidized by ozone and has to decay bimolecularly. There are two potential combination reactions, 37 (ClCl = 2.5 Å) and 38 followed by 39. The high endergonicity of reaction 37 (Table 5) prevents this reaction from taking place, and the bimolecular decay of ClO3• radical must occur via an asymmetrical recombination, reaction 38. Upon trying to calculate the standard Gibbs free energy of this reaction, the program proceeded automatically toward perchlorate ion and ClO2+ (reaction 39, Table 5). ClO2+ reacts with water, yielding chlorate (reaction 40). 2ClO3 • f O3 ClClO3 2ClO3 • f ½O3 ClOClO2 f ClO4 þ ClO2 þ ClO2 þ þ H2 O f ClO3 þ 2Hþ
Perchlorate formation has been intriguing the scientific community for a long time, and this question has been recently addressed again.53 ClO3• has been envisaged as a precursor, but details have remained obscure. At this point it is important to note that perchlorate is not an important product in the reaction of ClO/ClO2 with ozone, and perchlorate formation53 must be a minor side reaction (see the Supporting Information). Since the proposed O-transfer (ClO2 + O3 f ClO3 + O2) has been ruled out in favor of an electron transfer (ClO2 + O3 f OClO• + O3•),42 a route from OClO• to chlorate has to be found. In competition with a bimolecular decay of ClO3•, one may consider a termination with OClO• radicals in case there is a sufficiently high steady-state concentration of OClO• radicals present in equilibrium (note the relatively slow reaction of OClO• with ozone, Table 5). This reaction could eventually give rise to two molecules of chlorate ion in case the dimers were sufficiently long-lived for reacting with water. Yet reaction 41 is markedly endergonic with a long ClCl bond (1.980 Å) (Table 5); that is, reaction 41 is most likely reversible. Reaction 42 is less endergonic but potentially also reversible.
ð37Þ ð38=39Þ ð40Þ
ClO3 • þ OClO• a O3 ClClO2
ð41Þ
ClO3 • þ OClO• a O3 Cl OClO
ð42Þ
If the (exergonic) hydrolysis of these dimers (Cl2O5 + H2O f 2 ClO3 + 2 H+) were fast compared to the reverse reactions, 9201
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology
ARTICLE
41 and 42 (note the ready hydrolysis of Br2O4), preference of chlorate over perchlorate formation could be rationalized. Moreover, electron transfer from hyopchlorite and chlorite to ClO3• that would also give rise to chlorate is highly exergonic (Table 5). To which extent these reactions contribute to chlorate formation has still to be explored. There is also the potential and highly exergonic termination via the oxygens that proceeds directly, with the peroxidic arrangement only as a likely transition state (reaction 43), to OCl• and OClO• plus O2 (reaction 44). ClO3 • þ OClO• ½O3 ClO OClOTS f OCl• þ O2 þ OClO•
reaction sequence 53/54 is markedly endergonic (Table 5). OBrO• þ O3 f BrO• þ 2O2 OBrO• þ O3 f O2 BrOOO• f O2 BrO• þ O2 •
O2 BrO• þ O3 f O3 BrOOO• f O3 BrO• þ O2 •
O2 BrO• f BrO• þ O2
The reaction of the Br atom with ozone is strongly exergonic (reaction 45, Table 5). For the ensuing BrO • one can envisage reactions 46/47 and 48. BrO• þ O3 f BrOO• þ O2
ð46Þ
BrOO• a Br• þ O2
ð47Þ
BrO• þ O3 f OBrO• þ O2
ð48Þ
All three reactions are exergonic (Table 5), and a preference will be determined by kinetics, that is, whether ozone addition at oxygen or at bromine is preferred. In the ozonation of Br-containing waters, there is a second route to OBrO•. There is now strong evidence that the current concept of bromate formation45 has to be revised. The reaction of bromite with ozone does not give rise to bromate by O-transfer but is an electron transfer reaction that yields OBrO• and O3• 43 (for details see the Supporting Information). Formation and decay of OBrO• has been studied by pulse radiolysis,54 and it has been concluded that the bimolecular decay leading to a Br2O4 intermediate (i) is reversible, (ii) decays by reacting with water, and (iii) decays by reacting with OH. There are two conceivable dimers (reactions 49 and 50). 2OBrO• a OBrOOBrO
ð49Þ
2OBrO• a O2 BrOBrO
ð50Þ
While the symmetrical dimerization is mildly endergonic (Table 5) and could account for the reported reversibility, the asymmetrical dimerization is strongly endergonic and is unlikely to occur. There is, however also the possibility that the symmetrical dimer decays according to reaction 51. •
OBrOOBrO f 2Br þ 2O2
ð51Þ
This is the route taken upon prolonged optimization. Yet in water, where a OH-induced component of the decay of OBrO• radicals is observed, this reaction seems not to be kinetically favored despite the high exergonicity (Table 5). The bimolecular decay of OBrO• to Br2O4 is reversible, but Br2O4 reacts fast with water and OH, and it is unlikely that sufficiently high OBrO• concentrations build up for reactions of OBrO• with ozone to become relevant. Yet in case it would, the energetically favored (Table 5) route is reaction 52, while
ð55=56Þ
•
If O2BrO and O3BrO were formed, they would undergo the strongly exergonic 3O2-releasing reactions 57 and 58 (Table 5).
•
ð45Þ
ð53=54Þ
•
An oxidation of O2BrO to O3BrO (reactions 55 and 56) is endergonic (Table 5).
ð43=44Þ
Br• þ O3 f BrO• þ O2
ð52Þ
ð57Þ
O3 BrO• f OBrO• þ 2O2
ð58Þ •
It follows from the above that Br is not oxidized by ozone beyond OBrO•. Reactions of Ozone and Oxygen with Free Radicals: A Comparison. Oxygen is commonly regarded as a good scavenger of free radicals. This is only correct insofar as most free radicals under study are carbon-centered radicals, and any free radical chemistry involving alkyl radicals (k(R3C• + O2) ≈ 2 109 M1 s1) turns into a peroxyl radical chemistry in the presence of even low O2 concentrations.29 However, there are exceptions. Bisallylic radicals such as the hydroxycyclohexadienyl radicals formed upon •OH attack on benzene and its derivatives react already reversibly with O2,36,5557 and the phenoxyl radicals do not react with O2 (k < 103 M1 s1 33,58), unless activated by electrondonating substituents,59 despite the fact that there is a high spin density at carbon (see the Supporting Information). Under ozonation conditions, the O2 concentration typically exceeds the ozone concentration. As reactions of O2 and ozone with alkyl radicals are equally fast, competition is in favor of a reaction with O2. Corrections for a reaction of alkyl radicals with ozone are minor and rarely exceed 10%. The reaction of O2 with the hydroxycyclohexadienyl radicals is reversible.36,5557 This may give them a kinetic advantage in their reaction with ozone. To what extent this is of consequence cannot be predicted, as their rate constant with ozone is not yet known. Oxygen-centered radicals such as •OH, O2•, ROO•, and R2NO• do not react with O2, but they react readily with ozone (see above). This also holds for nitrogen-centered radicals that react with O2 very slowly or not at all.29 This is now confirmed by our calculations (reaction 18 is endergonic). Sulfur-centered radicals such as thiyl radicals react reversibly with O2.36,60 In agreement with this, a standard Gibbs free energy of +15 kJ mol1 has now been calculated for reaction 59. CH3 S• þ O2 a CH3 SOO•
ð59Þ
The 1,2-H shift reaction into a carbon-centered radicals observed with oxygen- and nitrogen-centered radicals, cf. reactions 9 and 22, is very slow here, as the reaction is endergonic (cf. reaction 60, ΔG° = +26 kJ mol1). CH3 CH2 S• f CH3 • CH SH
ð60Þ
The reaction of thiyl radicals with ozone, however, is irreversible (Table 4), and scavenging of thiyl radicals by ozone would be much more efficient than scavenging by O2. The much higher efficiency of ozone in scavenging free radicals as compared to O2 can lead to a much higher ozone demand in oxidation reactions by ozone as is currently believed. 9202
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology Advanced modeling of ozone reactions will have to take such reactions into account. Although there are already a number of rate constants of ozone with free radicals known (cf. Tables 15), a larger set is certainly required for reliable simulations. Here, the most important ones seem to be the nitrogen-centered radicals.
’ ASSOCIATED CONTENT
bS
Supporting Information. Standard Gibbs free energies for the reactions of •OH with phenol and the phenolate ion, spin distribution in the phenoxyl radicals and the standard Gibbs free energies for their recombination, reactions of ozone with chloride and bromide ions and subsequent reactions, and thermochemical calculations. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected] (C.v.S.); sergej.naumov@ iom-leipzig.de (S.N.).
’ REFERENCES (1) Merenyi, G.; Lind, J.; Naumov, S.; von Sonntag, C. The reaction of ozone with the hydroxide ion. Mechanistic considerations based on thermokinetic and quantum-chemical calculations. The role of HO4 in superoxide dismutation. Chem.—Eur. J. 2010, 16, 1372–1377. (2) Merenyi, G.; Lind, J.; Naumov, S.; von Sonntag, C. The reaction of ozone with hydrogen peroxide (peroxone process). A revision of current mechanistic concept based on thermokinetic and quantummechanical considerations. Environ. Sci. Technol. 2010, 44, 3505–3507. (3) Mvula, E.; Naumov, S.; von Sonntag, C. Ozonolysis of lignin models in aqueous solution: Anisole, 1,2-dimethoxybenzene, 1,4-dimethoxybenzene and 1,3,5-trimethoxybenzene. Environ. Sci. Technol. 2009, 43, 6275–6282. (4) Naumov, S.; von Sonntag, C. The reactions of bromide with ozone towards bromate and the hypobromite puzzle: A density functional theory study. Ozone: Sci. Eng. 2008, 30, 339–343. (5) Naumov, S.; von Sonntag, C. Quantum chemical studies on the formation of ozone adducts to aromatic compounds in aqueous solution. Ozone: Sci. Eng. 2010, 32, 61–65. (6) Naumov, S.; Mark, G.; Jarocki, A.; von Sonntag, C. The reactions of nitrite ion with ozone in aqueous solution—New experimental data and quantum-chemical considerations. Ozone: Sci. Eng. 2010, 32, 430–434. (7) Mark, G.; Naumov, S.; von Sonntag, C. The reaction of ozone with bisulfde (HS) in aqueous solution—Mechanistic aspects. Ozone: Sci. Eng. 2011, 33, 37–41. (8) Raghavachari, K.; Trucks, G. W.; Pople, J. A.; Replogle, E. S. Higly correlated systems: Structure, binding energy and harmonic vibrational frequencies of ozone. Chem. Phys. Lett. 1989, 158, 207–212. (9) Watts, J. D.; Bartlett, R. J. Coupled-cluster calculations of structure and vibrational frequencies of ozone: Are triple excitations enough?. J. Chem. Phys. 1998, 108, 2511–2514. (10) Foresman, J. B.; Frisch, A. Exploring Chemistry with Electronic Structure Methods; Gaussian, Inc.: Pittsburgh, PA, 1995. (11) Anglada, J. M.; Crehuet, R.; Bofill, J. M. The ozonolysis of ethylene: A theoretical study of the gas-phase reaction mechanism. Chem.—Eur. J. 1999, 5, 1809–1822. (12) Olzmann, M.; Kraka, E.; Cremer, D.; Gutbrod, R.; Andersson, S. Energetics, kinetics, and product distributions of the reactions of ozone with ethene and 2,3-dimethyl-2-butene. J. Phys. Chem. A 1997, 101, 9421–9429. (13) B€uhler, R. E.; Staehelin, J.; Hoigne, J. Ozone decomposition in water studied by pulse radiolysis. 1. HO2/O2 and HO3/O3 as intermediates. J. Phys. Chem. 1984, 88, 2560–2564.
ARTICLE
(14) Naumov, S.; von Sonntag, C. The reaction of •OH with O2, the decay of O3• and the pKa of HO3• interrrelated questions in aqueous free-radical chemistry. J. Phys. Org. Chem. 2011, 24, 600–602. (15) Becke, A. D. Density-functional thermochemistry. III. The role of exact exchange. J. Chem. Phys. 1993, 98, 5648–5652. (16) Lee, C.; Yang, W.; Parr, R. G. Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. Phys. Rev. B 1988, 37, 785–789. (17) Jaguar 7.6; Schrodinger LLC: New York, 2009. (18) Wadt, W. R.; Hay, P. J. Ab initio effective core potentials for molecular calculations—Potentials for main group elements Na to Bi. J. Chem. Phys. 1985, 82, 284–298. (19) Tannor, D. J. M.; Murphy, B.; Friesner, R.; Sitkoff, R. A.; Nicholls, D.; Ringnalda, A.; Goddard, W. A., III; Honig, B. Accurate first principles calculation of molecular charge distributions and solvation energies from ab initio quantum mechanics and continuum dielectric theory. J. Am. Chem. Soc. 1994, 116, 11875–11882. (20) Mu~ noz, F.; Mvula, E.; Braslavsky, S. E.; von Sonntag, C. Singlet dioxygen formation in ozone reactions in aqueous solution. J. Chem. Soc., Perkin Trans. 2 2001, 1109–1116. (21) Liu, Q.; Schurter, L. M.; Muller, C. E.; Aloisio, S.; Francisco, J. S.; Margerum, D. W. Kinetics and mechanisms of aqueous ozone reactions with bromide, sulfite, hydrogen sulfite, iodide, and nitrite ions. Inorg. Chem. 2001, 40, 4436–4442. (22) Wardman, P. Reduction potentials of one-electron couples involving free radicals in aqueous solution. J. Phys. Chem. Ref. Data 1989, 18, 1637–1755. (23) Sehested, K.; Holcman, J.; Hart, E. J. Rate constants and products of the reactions of e, O2 and H with ozone in aqueous solutions. J. Phys. Chem. 1983, 87, 1951–1954. (24) Elliot, A. J.; McCracken, D. R. Effect of temperature on O• reactions and equilibria: A pulse radiolysis study. Radiat. Phys. Chem. 1989, 33, 69–74. (25) Naumov, S.; von Sonntag, C. The energetics of rearrangement and water elimination reactions in the radiolysis of the DNA bases in aqueous solution (eaq and •OH attack). DFT calculations. Radiat. Res. 2008, 169, 355–363. (26) Lind, J.; Merenyi, G.; Johansson, E.; Brinck, T. The reaction of peroxyl radicals with ozone in water. J. Phys. Chem. A 2003, 107, 676–681. (27) Sehested, K.; Holcman, J.; Bjergbakke, E.; Hart, E. J. Ozone decomposition in aqueous acetate solutions. J. Phys. Chem. 1987, 91, 2359–2361. (28) von Sonntag, C. Free-Radical-Induced DNA Damage and Its Repair. A Chemical Perspective; Springer Verlag: Berlin, Heidelberg, 2006. (29) von Sonntag, C.; Schuchmann, H.-P. Peroxyl radicals in aqueous solution. In Peroxyl Radicals; Alfassi, Z. B., Ed.; Wiley: Chichester, U.K., 1997; pp 173234. (30) N€othe, T.; Fahlenkamp, H.; von Sonntag, C. Ozonation of wastewater: Rate of ozone consumption and hydroxyl radical yield. Environ. Sci. Technol. 2009, 43, 5590–5595. (31) Benn, R.; Dreeskamp, H.; Schuchmann, H.-P.; von Sonntag, C. Photolysis of aromatic ethers. Z. Naturforsch. 1979, 34b, 1002–1009. (32) Ye, M.; Schuler, R. H. The second order combination reactions of phenoxyl radicals. J. Phys. Chem. 1989, 93, 1898–1902. (33) Jin, F.; Leitich, J.; von Sonntag, C. The superoxide radical reacts with tyrosine-derived phenoxyl radicals by addition rather than by electron transfer. J. Chem. Soc., Perkin Trans. 2 1993, 1583–1588. (34) Hoigne, J.; Bader, H. Rate constants of reactions of ozone with organic and inorganic compounds in water—I. Non-dissociating organic compounds. Water Res. 1983, 17, 173–183. (35) Mertens, R.; von Sonntag, C. Determination of the kinetics of vinyl radical reactions by the characteristic visible absorption of vinyl peroxyl radicals. Angew. Chem., Int. Ed. Engl. 1994, 33, 1262–1264. (36) Naumov, S.; von Sonntag, C. UV-visible absorption spectra of alkyl-, vinyl-, aryl- and thiylperoxyl radicals and some related radicals in aqueous solution: A quantum-chemical study. J. Phys. Org. Chem. 2005, 18, 586–594. 9203
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
Environmental Science & Technology (37) Mvula, E.; von Sonntag, C. Ozonolysis of phenols in aqueous solution. Org. Biomol. Chem. 2003, 1, 1749–1756. (38) Ramseier, M. K.; von Gunten, U. Mechanism of phenol ozonization—Kinetics of formation of primary and secondary products. Ozone: Sci. Eng. 2009, 31, 201–215. (39) Sehested, K.; Holcman, J.; Bjergbakke, E.; Hart, E. J. A pulse radiolytic study of the reaction OH + O3 in aqueous medium. J. Phys. Chem. 1984, 88, 4144–4147. (40) Mu~noz, F.; von Sonntag, C. Determination of fast ozone reaction by competition kinetics. J. Chem. Soc., Perkin Trans. 2 2000, 661–664. (41) Sein, M. M.; Zedda, M.; Tuerk, J.; Schmidt, T. C.; Golloch, A.; von Sonntag, C. Oxidation of diclofenac with ozone in aqueous solution. Environ. Sci. Technol. 2008, 42, 6656–6662. (42) Kl€aning, U. K.; Sehested, K.; Holcman, J. Standard Gibbs energy of formation of the hydroxyl radical in aqueous solution. Rate constants for reaction ClO2 + O3 a O3 + ClO2. J. Phys. Chem. 1985, 89, 760–763. (43) Nicoson, J. S.; Wang, L.; Becker, R. H.; Huff, K. E.; Muller, C. E.; Margerum, D. W. Kinetics and mechanism of the ozone/bromite and ozone/chlorite reactions. Inorg. Chem. 2002, 41, 2975–2980. (44) Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (•OH/•O) in aqueous solution. J. Phys. Chem. Ref. Data 1988, 17, 513–886. (45) von Gunten, U. Ozonation of drinking water: Part II. Disinfection and by-product formation. Water Res. 2003, 37, 1469–1487. (46) Stanbury, D. M. Reduction potentials involving inorganic free radicals in aqueous solution. Adv. Inorg. Chem. 1989, 33, 69–138. (47) Buxton, G. V.; Bydder, M.; Salmon, G. A. Reactivity of chlorine atoms in aqueous solution. Part 1: The equilibrium Cl 3 + Cl = Cl2•. J. Chem. Soc., Faraday Trans. 1998, 94, 653–657. (48) Yu, X.-Y.; Barker, J. R. Hydrogen peroxide photolysis in acidic solutions containing chloride ions. I. Chemical mechanism. J. Phys. Chem. A 2003, 107, 1313–1324. (49) Liu, Y.; Pimentel, A. S.; Antoku, Y.; Barker, J. R. Temperaturedependent rate and equilibrium constants for Br•(aq) + Br(aq) a Br2•(aq). J. Phys. Chem. A 2002, 106, 11075–11082. (50) Bielski, B. H. J. A pulse radiolysis study of the reaction of ozone with Cl2• in aqueous solutions. Radiat. Phys. Chem. 1993, 41, 527–530. (51) Hoigne, J.; Bader, H.; Haag, W. R.; Staehelin, J. Rate constants of reactions of ozone with organic and inorganic compounds in water III. Inorganic compounds and radicals. Water Res. 1985, 19, 993–1004. (52) Chertova, Yu. S.; Avzyyanova, E. V.; Timerganzin, K. K.; Khalizov, A. F.; Sherreshovets, V. V.; Imashev, U. B. The formation of singlet molecular oxygen in the interaction of chlorine dioxide with ozone. Russ. J. Phys. Chem. 2000, 74 (Suppl. 3), 473–475. (53) Rao, B.; Anderson, T. A.; Redder, A.; Jackson, W. A. Perchlorate formation by ozone oxidation of aqueous chlorine/oxy-chlorine species: Role of ClxOy radicals. Environ. Sci. Technol. 2010, 44, 2961–2967. (54) Buxton, G. V.; Dainton, F. The radiolysis of aqueous solutions of oxybromine compounds; the spectra and reactions of BrO and BrO2. Proc. R. Soc. London, A 1968, 304, 427–439. (55) Pan, X.-M.; von Sonntag, C. OH-radical-induced oxidation of benzene in the presence of oxygen: R• a RO2• equilibria in aqueous solution. A pulse radiolysis study. Z. Naturforsch. 1990, 45b, 1337–1340. (56) Pan, X.-M.; Schuchmann, M. N.; von Sonntag, C. Oxidation of benzene by the OH radical. A product and pulse radiolysis study in oxygenated aqueous solution. J. Chem. Soc., Perkin Trans. 2 1993, 289–297. (57) Fang, X.; Pan, X.; Rahmann, A.; Schuchmann, H.-P.; von Sonntag, C. Reversibility in the reaction of cyclohexadienyl radicals with oxygen in aqueous solution. Chem.—Eur. J. 1995, 1, 423–429. (58) Hunter, E. P. L.; Desrosiers, M. F.; Simic, M. G. The effect of oxygen, antioxidants and superoxide radical on tyrosine phenoxyl radical dimerization. Free Radical Biol. Med. 1989, 6, 581–585. (59) Wang, D.; Gy€orgy, G.; Hildenbrand, K.; von Sonntag, C. Freeradical-induced oxidation of phloroglucinol—A pulse radiolysis and EPR study. J. Chem. Soc., Perkin Trans. 2 1994, 45–55.
ARTICLE
(60) Zhang, X.; Zhang, N.; Schuchmann, H.-P.; von Sonntag, C. Pulse radiolysis of 2-mercaptoethanol in oxygenated aqueous solution. Generation and reactions of the thiylperoxyl radical. J. Phys. Chem. 1994, 98, 6541–6547.
9204
dx.doi.org/10.1021/es2018658 |Environ. Sci. Technol. 2011, 45, 9195–9204
ARTICLE pubs.acs.org/est
Enhanced Transport of Colloidal Oil Droplets in Saturated and Unsaturated Sand Columns Micheal J. Travis, Amit Gross, and Noam Weisbrod* Department of Environmental Hydrology & Microbiology, Zuckerberg Institute for Water Research, The Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus, Midreshet Ben-Gurion, 84990 Israel
bS Supporting Information ABSTRACT: Colloidal-sized triacylglycerol droplets demonstrated enhanced transport compared to ideal latex colloid spheres in both saturated and unsaturated quartz sand columns. Oil droplets (mean diameter 0.74 ( 0.03 μm, density 0.92 g cm 3, ζ-potential 34 ( 1 mV) were injected simultaneously with latex microsphere colloids (FluoSpheres; density 1.055 g cm 3, diameters 0.02, 0.2, and 1.0 μm, ζ-potentials 16 ( 1, 30 ( 2, and 49 ( 1, respectively) and bromide into natural quartz sand (ζ-potential 63 ( 2 mV) via short-pulse column breakthrough experiments. Tests were conducted under both saturated and unsaturated conditions. Breakthrough of oil droplets preceded bromide and FluoSpheres. Recovery of oil droplets was 20% greater than similarly sized FluoSpheres in the saturated column, and 16% greater in the 0.18 ( 0.01 volumetric water content (VWC) unsaturated column. Higher variability was observed in the 0.14 ( 0.01 VWC column experiments with oil droplet recovery only slightly greater than similarly sized FluoSpheres. The research presents for the first time the direct comparison of colloidal oil droplet transport in porous media with that of other colloids, and demonstrates transport under unsaturated conditions. Based on experimental results and theoretical analyses, we discuss possible mechanisms that lead to the observed enhanced mobility of oil droplets compared to FluoSpheres with similar size and electrostatic properties.
’ INTRODUCTION Triacylglycerol (i.e., food oil) contamination of the environment may occur in various ways such as: soil-based waste disposal of materials from edible oil or other food processing operations;1 bioremediation of recalcitrant organic pollutants in which edible oil is introduced to the subsurface;2,3 and irrigation with wastewater.4 Edible oils are essentially insoluble in water, but may be dispersed in solution as colloidal-sized droplets from a few nanometers to several micrometers in diameter. Emulsion droplets in water may be stabilized by surfactants or finely divided solids,5 and present difficult treatment challenges.6 If solutions containing oil emulsion droplets are applied to, and move through the soil, they can enhance the transport of oil soluble compounds such as pesticides, pharmaceuticals, or environmental estrogens.7 The transport of edible oil emulsion droplets through saturated porous media has been previously published (e.g., refs 2,3,8 11). However, many of these studies focused on relatively large oil droplets (several μm8), or highly concentrated emulsion solutions (e.g., >10% oil3,9). To the best of our knowledge, the direct comparison of dilute oil colloid mobility in porous media to that of latex microsphere “ideal” colloids10 14 has never been reported. Furthermore, the transport of oil emulsion droplets in unsaturated media has not been documented. Transport characteristics of colloids in porous media may be influenced by water content,12 pore velocity,13 colloid r 2011 American Chemical Society
concentration,14 pH,15 and ionic strength.16 Colloid retention is influenced by factors at the interface, collector, and pore scales.17 Filtration theory predicts the transport of colloids through porous media.18 Derjaguin Landau Verwey Overbeck (DLVO) theory quantifies electrostatic energies to predict colloid and collector surfaces interactions.19,20 Colloid colloid and colloid grain surface interactions may be influenced by surface charge heterogeneity,21 the air water interface,22 Lewis acid base interactions,23 and steric contributions.24 The link between colloid transport in unsaturated versus saturated media has been explored mainly for latex microspheres as “ideal” colloids (e.g., refs 17,25 27). Nevertheless, the interactions of physical and chemical processes that govern unsaturated colloid transport and retention are still not well understood.17,28 Multiple interfaces in unsaturated media (e.g., solid solid, solid water, air water, air water solid) increase potential colloid retention through wedging/straining, bridging, or film straining.16,17,29 Furthermore, different colloids have been shown to possess unique transport qualities (e.g., biocolloids30,31 and clay13,32). Received: December 2, 2010 Accepted: September 28, 2011 Revised: September 6, 2011 Published: September 28, 2011 9205
dx.doi.org/10.1021/es104040k | Environ. Sci. Technol. 2011, 45, 9205–9211
Environmental Science & Technology The objective of this research was to determine the transport characteristics of colloidal-sized oil droplets under conditions in saturated and unsaturated natural sand, and to compare with the mobility of “ideal” latex microspheres.
’ MATERIALS AND METHODS Soil Columns and Porous Media. Short-pulse tracer experiments were conducted in soil columns (9.9 cm diameter, 30 cm long) equipped to control and monitor volumetric water content (VWC) and matric pressure. The setup was similar to systems used in previous studies.33,34 A detailed schematic is included in Figure SI-1 of the Supporting Information (SI). The columns were packed with washed and sieved (300 500 μm) natural beach sand from the coastal region of southern Israel, as detailed in SI S1. No clay-minerals or organic matter were detected in the sand. The sand was >99% quartz, with iron and other metals detected as potential oxide coatings by elemental analysis of the grain surfaces (SI Table SI-1). Three separate column packs were used, each for a set of four replicate experiments: (1) saturated; (2) unsaturated “high” water content (0.18 ( 0.01 VWC); and (3) unsaturated “low” water content (0.14 ( 0.01 VWC), so that porosity and pore structure were uniform within each set of experiments. Average pore diameter for the sand was 75 ( 2 μm calculated from capillary rise experiments,35 described in SI S2. Bulk density of the sand was 1.72 ( 0.02 g cm 3, and porosity 0.35 ( 0.007. Saturated hydraulic conductivity, measured by the constant head method, was 0.039 ( 0.001 cm s 1. ζ-potential of the sand grains was 63 ( 2 mV (pH 7), calculated from streaming potential (Anton Paar, Graz, Austria) using the Helmholtz-Smoluchowski equation and the Fairbrother-Mastin approach.36 Soil column and sand characteristics are summarized in SI Table SI-2. Solution was introduced to the upper surface of the column via a rain simulator. Background Solution and Pulse Preparation. All experiments used artificial rainwater (ARW)37 for the background and tracer solutions (SI S3) with ionic strength 0.021 mM, pH 7.2 ( 0.2, and electrical conductivity 183 ( 10 μS cm 1. Tracers included (1) oil droplets (100 mg L 1); (2) 1.0, 0.2, and 0.02 μm diameter FluoSpheres (Invitrogen Corporation, Eugene, OR, at concentrations 1, 5, and 10 mg L 1, respectively); and (3) lithium bromide (40 mg L 1). FluoSpheres are monodisperse, carboxylate-modified latex microspheres impregnated with fluorescent dye. FluoSpheres sizes were selected within the approximate range of oil droplet sizes (described below) in order to enable comparison. Oil droplet preparation is detailed in SI S4. Oil droplets were prepared first in a stable, concentrated emulsion consisting of 45% refined sunflower oil, 2.5% surfactant and 52.5% ARW. The concentrated emulsion was then diluted into the tracer solution. Final surfactant concentration in the tracer solution was 5 mg L 1, <23% of the critical micelle concentration. At this concentration the formation of hemimicelles and admicelles on the sand grain surfaces is not likely to occur.38 Surface tension of the transport solution was measured by the pendant drop method with DataPhysics OCA20 (DataPhysics Instruments GmbH, Filderstadt, Germany). The addition of emulsion to ARW reduced the surface tension by 0.5 mN m 1. This change was not expected to significantly affect solution transport through the column. Average oil droplet diameter was 0.74 ( 0.03 μm, and polydispersivity was 0.29 ( 0.01 (Brookhaven ZetaPlus, Holtsville, NY). Particle size analysis demonstrated a bimodal distribution
ARTICLE
of oil droplets (reflected in the high polydispersivity), with two droplet groupings observed around 0.3 and 0.8 μm (SI Figure SI-4). Filtration of the oil tracer colloids through precision polycarbonate membrane filters (Sterlitech Corp., Kent, WA) confirmed that 97% of droplets were in the range of 0.1 and 2.0 μm in diameter (SI S4). Average diameter of the oil droplet/ FluoSpheres mixture in the tracer solution prior to injection to the column was 0.40 ( 0.01 μm. The ζ-potentials of oil droplets and 1.0, 0.2, and 0.02 μm microspheres in ARW (pH 7.2) were determined from electrophoretic mobility39 using the ZetaPlus, and were 34 ( 1.4, 49 ( 1.3, 30 ( 1.8, and 16 ( 1.0 mV, respectively. Transport Experiments Setup and Analysis. Transport experiments were not conducted under sterile conditions, and the possibility of oil biodegradation was considered. Since the sand was essentially pasteurized by the drying process (i.e., 48 h at 65 °C), the initial experiment for each new column pack was assumed to be clean. Furthermore, the short duration of each transport experiment (tracer contact time in the column <3 h) and immediate analysis minimized biodegradation as a factor. A concern with conducting multiple tests was the potential effect of residual oil on subsequent transport. It was estimated that given the small volume of oil added to the column with each test, there would be no appreciable effect from residual oil on transport in subsequent tests within each series, that is, total of four experiments per column pack. This was verified by close reproducibility in replicate experiments. For saturated experiments, the inflow and outflow were synchronized at 14.6 ( 0.4 mL min 1, yielding an average pore velocity (i.e., Darcy q/porosity) of 0.088 ( 0.002 mm sec 1. Samples were collected for 2 4 pore volumes to capture complete tracer pulse transport. Before each subsequent experiment, g10 pore volumes of ARW were flushed through the column. For the unsaturated experiments, VWC was regulated by suction applied to the bottom of the high and low VWC columns ( 14.0 ( 0.01 and 24 ( 0.02 cm water, respectively) and by the pumping rate (6.1 ( 0.3 and 4.6 ( 0.3 mL min 1, respectively) (SI S1). Average pore velocity in the unsaturated experiments (calculated from Darcy q/VWC), were similar to the saturated experiments at 0.074 ( 0.003 and 0.070 ( 0.004 mm sec 1 for the 0.18 and 0.14 VWC columns, respectively). For the saturated column the pulse volume was 0.38 ( 0.1 normalized column water volumes (VCθ); for the high VWC 0.32 ( 0.1 VCθ, and for the low VWC 0.36 ( 0.1 VCθ. Normalized column water volume, defined by Mishurov et al.,37 was used instead of pore volume as typically reported for saturated experiments. Under fully saturated conditions, VCθ equals pore volume. For unsaturated conditions VCθ equals the average VWC multiplied by total column volume. Column discharge was collected using a CF-1 fraction collector (Spectrum Chromatography, Houston, TX). Samples were analyzed for oil by UV spectrophotometry at 254 nm.40 FluoSpheres concentrations were measured using a Cary Eclipse fluorescence spectrophotometer (Varian, Inc., Walnut Creek, CA) at excitation/emission wavelengths of 540/560, 365/415, and 580/605 nm for the 1.0, 0.2, and 0.02 μm FluoSpheres, respectively. Previous work has proven that these FluoSpheres can be used simultaneously without interference.37,41,42 Bromide was analyzed by phenol red method.40 Arrival time of each tracer was defined as a concentration g0.1% initial tracer concentration. Concentration (C) was normalized by dividing by the initial 9206
dx.doi.org/10.1021/es104040k |Environ. Sci. Technol. 2011, 45, 9205–9211
Environmental Science & Technology
ARTICLE
Figure 2. Dimensionless recovery of bromide, oil droplets, 1.0 and 0.2 μm diameter FluoSpheres from saturated, high volumetric water content (VWC) (0.18), and low VWC (0.14) unsaturated column experiments. Letters a, b, c, represent statistical differences between recoveries within each water content, as determined by ANOVA and the Tukey range test (P < 0.05).
Figure 1. Breakthrough curves for oil droplets, FluoSpheres (1.0 and 0.2 μm diameter) and bromide in (a) saturated sand; (b) saturated sand, pulse arrival time detail; (c) unsaturated sand, (0.18 volumetric water content [VWC]); (d) 0.18 VWC pulse arrival time detail; (e) unsaturated sand, 0.14 VWC; and (f) 0.14 VWC pulse arrival time detail. Note different scales on graphs b, d, and f.
pulse concentration (C0). Total recovery for each tracer was calculated as an integration of concentration and sample volume. Average colloid diameter and ζ-potential of the outflow were determined as described above. Column discharge samples at peak oil concentration were filtered (SI S4) and confirmed that measured oil was present as droplets only. Following completion of each set of experiments, the columns were dissected in 5 cm intervals and dried. Soil samples were extracted for gravimetric O&G analysis40 by accelerated solvent extraction (ASE, Dionex Corp., Sunnyvale, CA) using hexane. Transport Theory. Deposition of oil droplets and FluoSpheres was evaluated using filtration theory developed by Yao et al.18 and modified by Rajagopalan and Tien43 and by Tufenkji and Elimelech.44 The theory predicts particle capture in porous media based on a combined collision efficiency factor (ηT) that incorporates the effects of (1) diffusion; (2) interception; and (3) gravity (detailed description in SI S5). The ηT is dependent on characteristics of the colloid particle, the porous media, and flow. For neutral density particles, the contribution of gravity sedimentation becomes zero. Filtration theory does not facilitate prediction of colloid collision efficiency when colloid density is less than the fluid density. Favorability for colloid attachment to sand grains was calculated using the DLVO theory. Hamaker constants of 5 10 21 and 1 10 20 J were used for the oil droplets and FluoSpheres, respectively.45,46 Despite an underlying assumption of flat, rigid interacting surfaces, the DLVO theory is routinely applied to idealized colloid transport studies, and has also been reported to be appropriate for dilute oil-in-water emulsions systems5,11,47 and other nonrigid colloids such as bacteria.48 Classical DLVO theory presents an idealized system that accounts for van der
Waals and electric double-layer forces in considering colloid attachment. It has been reported that surface roughness such as found on natural sand grain surfaces may affects DLVO profiles.49 Under certain conditions, non-DLVO steric and Lewis acid base interactions may also affect colloid retention in porous media. Steric forces are important in systems with polymers or other large molecules within distances on the order of molecule length, and may aid in colloid sticking or repulsion.50 Steric control of deposition (whether repulsive or attractive) is typical of high ionic strength solutions, (g0.1 M).24 Lewis acid base forces are also short-range, and typically of contribution in low ionic strength solutions such as used in this research.23,51 Potential Deformability of oil emulsion droplets can affect transport properties, compared to hard particles. This phenomena is generally applicable to droplets in the range of several micrometers or larger.52 54
’ RESULTS AND DISCUSSION Breakthrough curves within each set of experiments (i.e., four experiments at each water content) exhibited comparable tracer arrival times ((0.05 VCθ), therefore only one breakthrough curve is shown for each set of experiments (Figure 1). The x-axes of the graphs, VCθ, assume that water is homogeneously distributed across the column. This assumption, however, does not represent differences in water content through the column (SI Figure SI-2), or actual flow paths that develop within the unsaturated column. Therefore, comparison between arrival times in different VWCs is not plausible. Impact of VWC on arrival time was reported by others (e.g., ref 55). Although total bromide recovery was ∼90% in all experiments, peak concentration decreased with reduced VWC. This behavior has been previously demonstrated since reduction in VWC increases dispersivity in porous media.56 All measured oil in outflow samples was confirmed to be droplets, that is, no oil was detected after microfiltration (data not shown). These findings agree with the low solubility of triacylglycerols.5 The transport results for the 0.02 μm FluoSpheres are not presented in this paper as this colloid diameter is about an order of magnitude below the range of oil droplets used in the experiments. Details on oil droplet distribution are presented in the SI. 9207
dx.doi.org/10.1021/es104040k |Environ. Sci. Technol. 2011, 45, 9205–9211
Environmental Science & Technology
Figure 3. Average colloid (oil droplets and FluoSpheres) diameter (μm) in recovered pulse versus column water volumes (VCθ), for representative (1) saturated (circles), (2) unsaturated 0.18 VWC (squares), and (3) unsaturated 0.14 VWC (triangles) column experiments. Filled symbols indicate pulse arrival time, that is, concentration g0.1% pulse concentration. Dashed line represents average colloid diameter of initial pulse.
Saturated Column Transport. Pulse arrival time in saturated media is represented in Figure 1b. Arrival time of oil droplets was observed before other tracers. Bromide recovery was 0.95 ( 0.07 for the saturated experiments. Total oil recovery (0.95 ( 0.06) exceeded that of all FluoSpheres, and was similar to bromide, indicating minimal deposition in the column (Figure 2, SI Table SI-3). Low retention of oil droplets was corroborated by sampling and analysis of the sand after the set of experiments was completed; that is, no oil was detected. Arrival time of oil droplets was followed closely by 1.0 and 0.2 μm FluoSpheres, with recoveries of 0.74 ( 0.04 and 0.69 ( 0.04 for the 1.0 and 0.2 μm FluoSpheres, respectively. Arrival time of bromide followed these colloids. The arrival times and recoveries of the FluoSpheres relative to one another were similar to Shani et al.41 and Mishurov et al.,37 who used the same mix of FluoSpheres, although in different size columns and various flow rates. Earlier arrival of larger colloids has been shown in several studies and explained by the size exclusion process that keeps larger colloids on central streamlines, resulting in selectively faster transport of larger versus smaller colloids.57 Average colloid diameter in the recovered pulse (oil droplets+FluoSpheres) decreased from 0.89 ( 0.12 at tracer arrival time to 0.34 ( 0.01 μm at the end of the breakthrough curve (Figure 3). All experiments consistently demonstrated this phenomenon; for simplicity one result from each set is presented. The results corroborate the understanding that larger colloids are transported faster. Lower total recovery of smaller FluoSpheres is predicted by filtration theory, which calculated increased collision frequency with the collector surface (sand grain) due to a greater diffusion coefficient of the smaller colloids (e.g., 2.4 10 12 and 4.9 10 13 m2 s 1 for the 0.2 and 1.0 μm FluoSpheres, respectively). The ζ-potentials for colloids in the column discharge for one saturated experiment from beginning to end of the pulse breakthrough measured 32 ( 1 mV, with no observable trend (data not shown). This suggests that ζ-potential of the oil droplets is relatively uniform and generally unaffected by transport through the column. Unsaturated Column Transport. Breakthrough curves for the 0.18 and 0.14 VWC unsaturated experiments exhibited greater fluctuation in tracer concentrations compared to the saturated experiments (Figure 1c and e). This is likely due to increased flow
ARTICLE
path tortuosity and dispersion associated with low water content porous media.55 In the 0.18 VWC column, bromide recovery was slightly less than for the saturated experiments (0.88 ( 0.05). As expected, recoveries of oil droplets and FluoSpheres were lower than the saturated experiments at 0.47 ( 0.05, 0.25 ( 0.08, and 0.31 ( 0.10 for the oil droplets, 1.0, and 0.2 μm FluoSpheres, respectively (Figure 2). Reduced recovery of colloids in unsaturated versus saturated media is well documented in the literature, resulting from improved collector efficiency (lower water content reduces transport distance), pore or film straining, and/or retention at the air water interface.16,58,59 In the 0.14 VWC experiments, bromide recovery was 0.91 ( 0.11 and oil droplet recovery was 0.66 ( 0.08. FluoSpheres recoveries were 0.45 ( 0.19 and 0.57 ( 0.08 for the 1.0 and 0.2 μm colloids, respectively (Figure 2). Unexpectedly, the recoveries of oil droplets and all FluoSpheres were higher in the 0.14 VWC compared to the 0.18 VWC experiments. This contradicts understanding of transport processes in unsaturated media.22,25,29 Torkzaban et al.12 reported that in low ionic strength solutions (6 mM), changes in colloid recovery were relatively small even at larger differences in VWC. The difference between average VWCs of 0.18 and 0.14 may be insignificant compared with other factors influencing transport. We suggest that the differences in colloid recoveries may be the result of variation in the column packing (and resulting pore spaces), water film continuities, and/or slight differences in the natural sand used in these experiments. Variation in relative amount of coatings, even small, can dramatically affect colloid recovery.60 With reduced travel distance from the bulk liquid to the sand grain surface under unsaturated conditions, colloid collector collisions are increased and the effect of coatings is likely amplified. However, we expect that the effects would be similar for both oil droplets and FluoSpheres due to similar surface properties and subsequent DLVO predictions. In comparison to the saturated column, a 30 50% reduction in oil droplet recovery was observed under unsaturated conditions. Sand samples collected from the two unsaturated columns showed oil deposition in the upper 5 cm of the column, with a mass of 14 and 11 mg retained in the 0.18 and 0.14 VWC columns, respectively (SI Table SI-3), representing 51 and 66%, respectively, of the oil not accounted for in the column discharge. No oil was detected below 5 cm. This agrees with the findings of Bradford et al.,58 who showed colloid retention greatest in the upper 20% of a column, and relatively low deposition in the remainder of the column. Similar to the saturated experiments, oil droplet recovery under unsaturated conditions exceeded that of FluoSpheres. The decrease in FluoSpheres recovery in unsaturated columns was greatest for the 1.0 μm colloids (39 67%) followed by the 0.2 μm colloids (17 55%). Average droplet size in the column discharge decreased from 0.51 ( 0.09 μm at the beginning to 0.37 ( 0.03 μm at the end of pulse outflow for the 0.18 VWC experiment, and from 0.49 ( 0.09 to 0.34 ( 0.03 μm for the 0.14 VWC experiments (Figure 3). Size fractionation and subsequent decrease in colloid size while colloids of different diameters are migrating through natural flow pathways was observed in other systems.32 The significant difference in diameter of recovered colloids between saturated and unsaturated experiments is likely related to straining processes, discussed below. Colloid Retention Mechanisms. Straining. Sand grain uniformity and the large average pore diameter (75 ( 2 μm) relative to FluoSpheres and oil droplet diameters suggests that mechanical blockage in pores (i.e., straining) in the saturated media 9208
dx.doi.org/10.1021/es104040k |Environ. Sci. Technol. 2011, 45, 9205–9211
Environmental Science & Technology should be negligible.61 Bradford et al.62 reported high colloid recovery (C/C0∼1) when the colloid to grain size ratio was e0.007, which is comparable to the maximum ratio observed in this study. It is not likely that deformability had a major role in the enhanced transport of the oil colloids, since a maximum of 3% by volume of the emulsion droplets were >2 μm in diameter.52 54 In unsaturated media, however, thin-film straining can immobilize colloids when water film thickness approaches the colloid diameter.27,29 Other straining and wedging mechanism also become more prevalent.17 As discussed previously, average particle diameter in the column outflow decreased from beginning to end of pulse recovery. A notable reduction in maximum colloid diameter was observed between saturated and unsaturated experiments. We suggest that larger colloids, observed in saturated column outflow, were captured in the unsaturated columns through straining mechanisms. This was indicated by oil detected in the upper 5 cm of the unsaturated columns, as previously noted. DLVO Interactions. DLVO calculations based on measured ζ-potentials indicated strong primary electrostatic repulsion between the sand grains and oil droplet or FluoSpheres. Assuming homogeneous surfaces, the repulsive force was calculated to extend approximately 30 60 nm from the sand grain surface (SI Figure SI-3). The relatively large separation distance is typical of low ionic strength solutions, and suggests that both steric and Lewis acid base interactions would have minimal impact on colloid deposition. Nevertheless, clearly the natural sand grains do not have homogeneous surfaces and positively charged patches are likely to exist on the surfaces. Therefore, these calculations may not accurately portray the full explanation.60 The presence of nonionic surfactant on the oil droplets rendered them hydrophilic (similar to the FluoSpheres), which minimized potential hydrophobic attraction (i.e., as might be expected for pure oil). Oil emulsion colloids are stabilized by surface active agents, and the colloid properties can vary significantly depending on the stabilizing agents used.5 This research presents transport of one specific type of oil colloid, stabilized by nonionic surfactants. Future research should examine the effects of other surfactants, such as anionic, cationic or zwitterionic. A small secondary energy minimum was calculated (up to 0.5 kT), its magnitude increasing with greater colloid size. The secondary energy minimum represents a zone close to, but separated from the collector surface (and the primary repulsion), which has a net attractive energy toward colloids. This zone has been suggested for possible colloid retention,21,63 even in the presence of strong primary DLVO repulsion. It has also been suggested that the secondary energy minimum may act as a channel in which colloids may be moved along collector surfaces and funneled into constricted areas or stagnant zones where they may be physically captured.12 The magnitude of the secondary minimum for our experimental conditions was small, and less than the thermal energy of colloid particles, which suggests that Brownian motion (diffusion) would prevent retention.64 It must be noted, however, that although the sand ζ-potential used in the DLVO calculations was strongly negative ( 65 ( 2 mV), this value represents an average for the measured sample. As noted, the sand contained oxide coatings (which exhibit positive ζ-potential60), which would be favorable for deposition. Furthermore, the presence of patches are increasingly important under unsaturated conditions, where increased interaction at the solid-water interface leads to increased retention.65 Calculations based on DLVO theory predicted similar repulsive
ARTICLE
Figure 4. Comparison of theoretical total collector efficiency (ηT) for FluoSpheres microspheres (F = 1.055) and neutral density colloids (F = 0.997) based on filtration theory by Yao et al.18 (Yao), Rajagopalan and Tien43 (RT), and Tufenkji and Elimelech44 (TE) at the experimental conditions.
forces between the silica surfaces and the oil droplets or FluoSpheres (SI Figure SI-3). Conversely, when assuming colloids in the presence of positive (+) ζ-potential metal oxide coatings,66 DLVO calculations predicted similar net attractive forces for oil droplets and microspheres (data not shown). The similarity of DLVO calculations for the two colloid types shows that the theory fails to explain the relatively higher transportability of oil droplets compared to FluoSpheres. Filtration Theory. Collision efficiency, ηT, was evaluated for FluoSpheres and neutral-density colloids (i.e., contribution of gravity equals zero) using Yao et al.,18 Rajagopalan & Tien,43 and Tufenkji & Elimelech44 methods (Figure 4). All methods predicted reduced ηT (i.e., greater mobility) with increasing size for FluoSpheres (i.e., up to ∼2 μm) and neutral density colloids (i.e., up to ∼5 μm). No statistical differences were observed between the recovery rates of 0.2 and 1.0 μm FluoSpheres. We note that this is inconsistent with theoretical calculations which would predict that 0.2 um colloids would be more strongly retained. Similar results, however, have been reported in other published studies using this same mixture of latex colloid sizes in sand37,41 and also in fractured rock cores.42 Bradford et al.,58 contrary to pure theoretical expectations, observed that 0.45 and 1.0 μm latex colloids had comparable recoveries from natural saturated sand. This was observed in sands of several different grain sizes. Because of the large pore sizes of these sands they concluded that colloid removal was due to attachment processes (i.e., and not straining). However the actual mechanisms of attachment could not be deduced. The differences in predicted ηT between the two colloid types did not account for the 20% greater recovery observed in the saturated experiments for oil droplets versus FluoSpheres. We emphasize that the oil colloid density (0.92 g cm 3) was less than the water density (0.98 g cm 3), as well as that of the FluoSpheres (1.055 g cm 3). The significance of the colloid density differences is difficult to determine, since the filtration theory is limited to predictions for colloids with neutral or greater density compared to the bulk fluid. None of the collector efficiency equations18,43,44 are able to deal with low density particles (relative to the transport solution), and the potential effect on colloid-collector interactions is theoretically poorly understood. Though Brownian motion dominates transport of colloids in the size range used in this study, gravity and interception are still important contributing factors.18,43,44 It is worth mentioning that 9209
dx.doi.org/10.1021/es104040k |Environ. Sci. Technol. 2011, 45, 9205–9211
Environmental Science & Technology the movement of colloids with lower density than the transport solution is also relevant in other environments.67 In summary, the research demonstrated that oil emulsion colloids experienced enhanced transport properties compared to ideal colloids in both saturated and unsaturated porous media. Theoretical explanations were explored, and the lower density of the oil colloids compared to the transport fluid is suggested to have a potential role in the observed differences in transport between oil and ideal colloids. None of the current filtration theory equations are able to assess the effect of this relative low density on colloid transport. Additional research is required to identify the mechanisms and to quantify their impacts.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional details on materials and methods used and supplemental figures and tables. This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +972 8 6596903; fax: +972 8 6596909; e-mail: weisbrod@ bgu.ac.il.
’ ACKNOWLEDGMENT We acknowledge the Daniel E. Koshland Fund and the Rieger Foundation for funding contributions toward this research. ’ REFERENCES (1) Gonzalez-Vila, F. J.; Verdejo, T.; Del Rio, J. C.; Martin, F. Accumulation of hydrophobic compounds in the soil lipidic and humic fractions as result of a long term land treatment with olive oil mill effluents (alpechin). Chemosphere 1995, 31 (7), 3681–3686. (2) Berge, N. D.; Ramsburg, C. A. Oil-in-water emulsions for encapsulated delivery of reactive iron particles. Environ. Sci. Technol. 2009, 43 (13), 5060–5066. (3) Borden, R. C. Effective distribution of emulsified edible oil for enhanced anaerobic bioremediation. J. Contam. Hydrol. 2007, 94 (1 2), 1–12. (4) Travis, M. J.; Weisbrod, N.; Gross, A. Accumulation of oil and grease in soils irrigated with greywater and their potential role in soil water repellency. Sci. Total Environ. 2008, 394 (1), 68–74. (5) Becher, P. Emulsions: Theory and Practice; Oxford University Press: New York, 2001. (6) Shin, S. H.; Kim, D. S. Studies on the interfacial characterization of O/W emulsion for the optimization of its treatment. Environ. Sci. Technol. 2001, 35 (14), 3040–3047. (7) Khanal, S. K.; Xie, B.; Thompson, M. L.; Sung, S.; Ong, S.-K.; van Leeuwen, J. Fate, transport, and biodegradation of natural estrogens in the environment and engineered systems. Environ. Sci. Technol. 2006, 40 (21), 6537–6546. (8) Soo, H.; Radke, C. J. Flow mechanism of dilute, stable emulsions in porous media. Ind. Eng. Chem. Fund. 1984, 23 (3), 342–347. (9) Coulibaly, K. M.; Long, C. M.; Borden, R. C. Transport of edible oil emulsions in clayey sands: One-dimensional column results and model development. J. Hydrol. Eng. 2006, 11 (3), 230–237. (10) Soma, J.; Papadopoulos, K. D. Deposition of oil-in-water emulsions in sand beds in the presence of cetyltrimethylammonium bromide. Environ. Sci. Technol. 1997, 31 (4), 1040–1045. (11) Soma, J.; Papadopoulos, K. D. Flow of dilute, sub-micron emulsions in granular porous media: Effects of pH and ionic strength. Colloids Surf. 1995, 101 (1), 51–61.
ARTICLE
(12) Torkzaban, S.; Bradford, S. A.; van Genuchten, M. T.; Walker, S. L. Colloid transport in unsaturated porous media: The role of water content and ionic strength on particle straining. J. Contam. Hydrol. 2008, 96 (1 4), 113–127. (13) Compere, F.; Porel, G.; Delay, F. Transport and retention of clay particles in saturated porous media: Influence of ionic strength and pore velocity. J. Contam. Hydrol. 2001, 49 (1 2), 1–21. (14) Zhang, W.; Morales, V. L.; Cakmak, M. E.; Salvucci, A. E.; Geohring, L. D.; Hay, A. G.; Parlange, J.-Y.; Steenhuis, T. S. Colloid transport and retention in unsaturated porous media: Effect of colloid input concentration. Environ. Sci. Technol. 2010, 44 (13), 4965–4972. (15) Walshe, G. E.; Pang, L.; Flury, M.; Close, M. E.; Flintoft, M. Effects of pH, ionic strength, dissolved organic matter, and flow rate on the co-transport of MS2 bacteriophages with kaolinite in gravel aquifer media. Water Res. 2010, 44 (4), 1255–1269. (16) Saiers, J. E.; Lenhart, J. J. Ionic-strength effects on colloid transport and interfacial reactions in partially saturated porous media. Water Resour. Res. 2003, 39 (9), 1256–1268. (17) Bradford, S. A.; Torkzaban, S. Colloid transport and retention in unsaturated porous media: A review of interface-, collector-, and porescale processes and models. Vadose Zone J. 2008, 7 (2), 667–681. (18) Yao, K.-M.; Habibian, M. T.; O’Melia, C. R. Water and waste water filtration: Concepts and applications. Environ. Sci. Technol. 1971, 5 (11), 1105–1112. (19) Derjaguin, B. V.; Landau, L. D. Theory of the stability of strongly charged lyophobic sols and the adhesion of strongly charged particles in solutions of electrolytes. Acta Physicochim. URSS 1941, 14, 733–762. (20) Verwey, E. J. W.; Overbeek, J. T. G. Theory of Stability of Lyophobic Colloids; Elsevier: Amsterdam, 1948. (21) Tufenkji, N.; Elimelech, M. Breakdown of colloid filtration theory: Role of the secondary energy minimum and surface charge heterogeneities. Langmuir 2005, 21 (3), 841–852. (22) Lenhart, J. J.; Saiers, J. E. Transport of silica colloids through unsaturated porous media: Experimental results and model comparisons. Environ. Sci. Technol. 2002, 36 (4), 769–777. (23) van Oss, C. J. Long-range and short-range mechanisms of hydrophobic attraction and hydrophilic repulsion in specific and aspecific interactions. J. Mol. Recognit. 2003, 16 (4), 177–190. (24) Rijnaarts, H. H. M.; Norde, W.; Lyklema, J.; Zehnder, A. J. B. DLVO and steric contributions to bacterial deposition in media of different ionic strengths. Colloids Surf. B 1999, 14 (1 4), 179–195. (25) Wan, J.; Wilson, J. L. Colloid transport in unsaturated porous media. Water Resour. Res. 1994, 30 (4), 857–864. (26) Saiers, J. E.; Lenhart, J. J. Colloid mobilization and transport within unsaturated porous media under transient-flow conditions. Water Resour. Res. 2003, 39 (1), 1019. (27) Chen, G.; Flury, M. Retention of mineral colloids in unsaturated porous media as related to their surface properties. Colloids Surf. A 2005, 256 (2 3), 207–216. (28) DeNovio, N. M.; Saiers, J. E.; Ryan, J. N. Colloid movement in unsaturated porous media: Recent advances and future directions. Vadose Zone J. 2004, 3 (2), 338–351. (29) Wan, J.; Tokunaga, T. K. Film straining of colloids in unsaturated porous media: Conceptual model and experimental testing. Environ. Sci. Technol. 1997, 31 (8), 2413–2420. (30) de Kerchove, A. J.; Elimelech, M. Relevance of electrokinetic theory for “soft” particles to bacterial cells: Implications for bacterial adhesion. Langmuir 2005, 21 (14), 6462–6472. (31) Azeredo, J.; Visser, J.; Oliveira, R. Exopolymers in bacterial adhesion: Interpretation in terms of DLVO and XDLVO theories. Colloids Surf. B 1999, 14 (1 4), 141–148. (32) Zvikelsky, O.; Weisbrod, N.; Dody, A. A comparison of clay colloid and artificial microsphere transport in natural discrete fractures. J. Colloid Interface Sci. 2008, 323 (2), 286–292. (33) Cherrey, K. D.; Flury, M.; Harsh, J. B. Nitrate and colloid transport through coarse Hanford sediments under steady state, variably saturated flow. Water Resour. Res. 2003, 39 (6), 1165. 9210
dx.doi.org/10.1021/es104040k |Environ. Sci. Technol. 2011, 45, 9205–9211
Environmental Science & Technology (34) Weigand, H.; Totsche, K. U. Flow and reactivity effects on dissolved organic matter transport in soil columns. Soil Sci. Soc. Am. J. 1998, 62 (5), 1268–1274. (35) Weisbrod, N.; McGinnis, T.; Rockhold, M. L.; Niemet, M. R.; Selker, J. S. Effective Darcy-scale contact angles in porous media imbibing solutions of various surface tensions. Water Resour. Res. 2009, 45, W00D39. (36) Fairbrother, F.; Mastin, H. Studies in electro-endosmosis. J. Chem. Soc. 1924, 75, 2318. (37) Mishurov, M. M.; Yakirevich, A.; Weisbrod, N. Colloid transport in a heterogeneous partially saturated sand column. Environ. Sci. Technol. 2008, 42 (4), 1066. (38) West, C. C.; Harwell, J. H. Surfactants and subsurface remediation. Environ. Sci. Technol. 1992, 26 (12), 2324–2330. (39) Adamson, A. W.; Gast, A. P. Physical Chemistry of Surfaces, 6th ed.; John Wiley & Sons, Inc.: New York, 1997. (40) APHA; AWWA; WEF, Standard Methods for the Examination of Water and Wastewater, 20th ed.; American Public Health Association: Washington, DC, 1998. (41) Shani, C.; Weisbrod, N.; Yakirevich, A. Colloid transport through saturated sand columns: Influence of physical and chemical surface properties on deposition. Colloids Surf. A 2008, 316 (1 3), 142–150. (42) Zvikelsky, O.; Weisbrod, N., Impact of particle size on colloid transport in discrete fractures. Water Resour. Res. 2006, 42. (43) Rajagopalan, R.; Tien, C. Trajectory analysis of deep-bed filtration with the sphere-in-cell porous media model. AIChE J. 1976, 22 (3), 523–533. (44) Tufenkji, N.; Elimelech, M. Correlation equation for predicting single-collector efficiency in physicochemical filtration in saturated porous media. Environ. Sci. Technol. 2004, 38 (2), 529–536. (45) Binks, B. P.; Lumsdon, S. O. Pickering emulsions stabilized by monodisperse latex particles: Effects of particle size. Langmuir 2001, 17 (15), 4540–4547. (46) Elimelech, M.; O’Melia, C. R. Kinetics of deposition of colloidal particles in porous media. Environ. Sci. Technol. 1990, 24 (10), 1528– 1536. (47) Malmsten, M.; Lindstrom, A.-L.; Warnheim, T. Electrostatic effects on interfacial film formation in emulsion systems. J. Colloid Interface Sci. 1996, 179 (2), 537–543. (48) Redman, J. A.; Grant, S. B.; Olson, T. M.; Estes, M. K. Pathogen filtration, heterogeneity, and the potable reuse of wastewater. Environ. Sci. Technol. 2001, 35 (9), 1798–1805. (49) Hoek, E. M. V.; Agarwal, G. K. Extended DLVO interactions between spherical particles and rough surfaces. J. Colloid Interface Sci. 2006, 298 (1), 50–58. (50) Israelachvili, J. N., Intermolecular and Surface Forces; Academic Press Limited: San Diego, CA, 1992. (51) van Oss, C. J. Acid--base interfacial interactions in aqueous media. Colloids Surf. A 1993, 78, 1–49. (52) Headen, T. F.; Clarke, S. M.; Perdigon, A.; Meeten, G. H.; Sherwood, J. D.; Aston, M. Filtration of deformable emulsion droplets. J. Colloid Interface Sci. 2006, 304 (2), 562–565. (53) Dagastine, R. R.; Manica, R.; Carnie, S. L.; Chan, D. Y. C.; Stevens, G. W.; Grieser, F. Dynamic forces between two deformable oil droplets in water. Sci. Total Environ. 2006, 313, 210–213. (54) Saiki, Y.; Prestidge, C. A.; Horn, R. G. Effects of droplet deformability on emulsion rheology. Colloids Surf. A 2007, 299 (1 3), 65–72. (55) Padilla, I. Y.; Yeh, T. C. J.; Conklin, M. H. The effect of water content on solute transport in unsaturated porous media. Water Resour. Res. 1999, 35 (11), 3303–3313. (56) Nutzmann, G.; Maciejewski, S.; Joswig, K. Estimation of water saturation dependence of dispersion in unsaturated porous media: Experiments and modelling analysis. Adv. Water Res. 2002, 25, 565–576. (57) Keller, A. A.; Auset, M. A review of visualization techniques of biocolloid transport processes at the pore scale under saturated and unsaturated conditions. Adv. Water Res. 2007, 30 (6 7), 1392–1407.
ARTICLE
(58) Bradford, S. A.; Yates, S. R.; Bettahar, M.; Simunek, J. Physical factors affecting the transport and fate of colloids in saturated porous media. Water Resour. Res. 2002, 38 (12), 1327. (59) Zhuang, J.; Qi, J.; Jin, Y. Retention and transport of amphiphilic colloids under unsaturated flow conditions: Effect of particle size and surface property. Environ. Sci. Technol. 2005, 39 (20), 7853–7859. (60) Elimelech, M.; Nagai, M.; Ko, C.-H.; Ryan, J. N. Relative insignificance of mineral grain zeta potential to colloid transport in geochemically heterogeneous porous media. Environ. Sci. Technol. 2000, 34 (11), 2143–2148. (61) Huber, N.; Baumann, T.; Niessner, R. Assessment of colloid filtration in natural porous media by filtration theory. Environ. Sci. Technol. 2000, 34 (17), 3774–3779. (62) Bradford, S. A.; Torkzaban, S.; Walker, S. L. Coupling of physical and chemical mechanisms of colloid straining in saturated porous media. Water Res. 2007, 41 (13), 3012–3024. (63) Redman, J. A.; Walker, S. L.; Elimelech, M. Bacterial adhesion and transport in porous media: Role of the secondary energy minimum. Environ. Sci. Technol. 2004, 38 (6), 1777–1785. (64) O’Melia, C. R.; Hahn, M. W.; Chen, C.-T. Some effects of particle size in separation processes involving colloids. Water Sci. Technol. 1997, 36 (4), 119–126. (65) Chu, Y.; Jin, Y.; Flury, M.; Yates, M. V. Mechanisms of virus removal during transport in unsaturated porous media. Water Resour. Res. 2001, 37 (2), 253–263. (66) Zhuang, J.; Yu, G.-R. Effects of surface coatings on electrochemical properties and contaminant sorption of clay minerals. Chemosphere 2002, 49 (6), 619–628. (67) Magal, E.; Weisbrod, N.; Yechieli, Y.; Walker, S. L.; Yakirevich, A. Colloid transport in porous media: Impact of hyper-saline solutions. Water Res. 2011, 45 (11), 3521–3532.
9211
dx.doi.org/10.1021/es104040k |Environ. Sci. Technol. 2011, 45, 9205–9211
ARTICLE pubs.acs.org/est
Fractionation of Stable Zinc Isotopes in the Zinc Hyperaccumulator Arabidopsis halleri and Nonaccumulator Arabidopsis petraea A. M. Aucour,†,* S. Pichat,‡ M. R. Macnair,§ and P. Oger‡ †
Laboratoire de Geologie de Lyon, Universite Claude Bernard Lyon1, CNRS, 2, rue Rapha€el Dubois, 69622 Villeurbanne Cedex, France Laboratoire de Geologie de Lyon, Ecole Normale Superieure de Lyon, CNRS, 69007 Lyon, France § School of Biosciences, Hatherly laboratories, University of Exeter, Prince of Wales Road, Exeter, EX4 4PS, U.K. ‡
bS Supporting Information ABSTRACT: Zn isotope fractionation may provide new insights into Zn uptake, transport and storage mechanisms in plants. It was investigated here in the Zn hyperaccumulator Arabidopsis halleri and the nonaccumulator A. petraea. Plant growth on hydroponic solution allowed us to measure the isotope fractionation between source Zn (with Zn2+ as dominant form), shoot and root. Zn isotope mass balance yields mean isotope fractionation between plant and source Zn Δ66Znin-source of 0.19 ( 0.20 % in the nonaccumulator and of 0.05 ( 0.12% in the hyperaccumulator. The isotope fractionation between shoot Zn and bulk Zn incorporated (Δ66Znshoot-in) differs between the nonaccumulator and the hyperaccumulator and is function of root-shoot translocation (as given by mass ratio between shoot Zn and bulk plant Zn). The large isotope fractionation associated with sequestration in the root (0.37 %) points to the binding of Zn2+ with a high affinity ligand in the root cell. We conclude that Zn stable isotopes may help to estimate underground and aerial Zn storage in plants and be useful in studying extracellular and cellular mechanisms of sequestration in the root.
’ INTRODUCTION In living organisms, zinc is the most abundant transition metal after iron. It is an essential catalytic component for a wide range of enzymes and plays a structural role in many proteins. Understanding the mechanisms of Zn uptake, transport and storage by plants is of critical importance for pollution and nutrition issues. The ability of plants to deal with either excessive or deficient levels of Zn in the soil are important adaptative responses. These adaptations form the basis for the phytoremediation and phytostabilization of contaminated soils and for the biofortification of crops. Excessive levels of Zn occur in soils contaminated by mining and smelting activities, in agricultural soil treated with sewage sludges and in urban and periurban soils.1 However, Zn deficiency is a widespread limiting factor to crop production.2 While elemental analysis allows one to document the distribution of Zn in soil and plants, one potential way to improve our understanding of uptake and trafficking of Zn in plants is to use stable Zn isotope ratios. The five stable isotopes 64Zn, 66Zn, 67Zn, 68Zn and 70 Zn have average natural abundances of 48.6, 27.90, 4.10, 18.75 and 0.62%, respectively. Recent developments in multiple-collector inductively coupled mass-spectrometry (MC-ICP-MS) have allowed highly precise measurements of the stable isotope composition of zinc.3,4 They have revealed large natural variations in the abundances of the stable zinc isotopes between photosynthetic organisms and their growth medium and between organs of higher plants.58 Plant organs (leaf, stem, root) collected in the field show a r 2011 American Chemical Society
wide δ66Zn range between 0.91% to 0.75%.7 Germination experiments on lentils have shown a systematic depletion in heavy isotopes of 0.34% between leaves and seeds.8 Plant cultures on hydroponic solution have provided evidence for an enrichment in heavy isotopes of 0.08 to 0.18% from solution to root and depletion from root to shoot of 0.13 to 0.26%, the extent of which depends on the species studied (tomato, lettuce, rice).5 The above data indicate that large isotope fractionations are associated with Zn uptake and translocation. The Zn isotope ratios could thus potentially be used as a tracer of uptake, storage and translocation processes in the plant if we can assign a fractionation value to each process. The uptake of Zn has been reported to lead to a depletion in heavy isotopes in the diatom Thallassiosira oceanica.6 The extent of the depletion varied with increasing Zn concentration from 0.2% to 0.8%, corresponding to the switch from high- to low- affinity transport into the cell. A very similar, and better characterized, transport machinery is used by higher plants to incorporate Zn, which could also lead to Zn isotope fractionation. Within the plant, Zn isotope compositions can yield information on the main translocation and storage steps as well as help to estimate translocation fluxes between the different plant Zn compartments.9 Received: March 16, 2011 Accepted: September 1, 2011 Revised: August 1, 2011 Published: September 01, 2011 9212
dx.doi.org/10.1021/es200874x | Environ. Sci. Technol. 2011, 45, 9212–9217
Environmental Science & Technology The Zn hyperaccumulator Arabidopsis halleri and nonaccumulator A. petraea (A. lyrata ssp. petraea) are central models for the study of metal uptake and translocation. The Zn hyperaccumulator A. halleri can grow in the presence of high levels of zinc and translocate exceptionally high amounts of metal to the shoot with leaf Zn concentration as high as 5% dry wt.10 Its sister species A. petraea was isolated from a noncontaminated soil and does not accumulate Zn in its shoot. They are the sister clade to A. thaliana whose genome has been entirely sequenced and from which putative Zn transporters have been characterized.11 Enhanced Zn uptake into root cells is thought to be driven by transporters of the ZIP (zinc-regulated transporter, iron-regulated transporter protein) family.11,1315 Efficient translocation of Zn to the shoot requires active xylem loading and has been shown to depend on enhanced activity of the heavy metal transporting ATPase 4 (HMA4).16 In the root cells, efflux of Zn from the cytosol into the vacuole seems to be catalyzed by the cation diffusion facilitator MTP1 12,18 while the role of different chelators (nicotianamine, phytochelatines) in Zn transport is still a matter of debate.17 Extended X-ray absorption spectroscopy (EXAFS) has identified Zn phosphate as storage forms in the roots of A. halleri and A. petraea, either as cell wall disordered inorganic phosphate or as cellular organic phosphate.19,20 The present study focuses on the use of Zn isotope fractionations in connection with uptake and root-shoot translocation to address Zn transfer mechanisms and fluxes. The growth of the hyperaccumulator A. halleri and nonaccumulator A. petraea under controlled conditions allowed us to (i) investigate the isotope fractionation with uptake in the hyper- and nonaccumulator, (ii) model the Zn isotope fractionation in the root as a function of the extent of translocation to shoot, and (iii) propose constraints on putative sequestrations mechanisms in the root.
’ MATERIAL AND METHODS Experimental Setup and Sample Collection. The provenance of the seeds of A. halleri and A. petraea used in this study has been given previously.10 Seeds were germinated on a mixture of sand and compost in a greenhouse. Two weeks after germination seedlings were transferred to polycarbonate vessels containing 7-l of nutrient solution under aeration with three to six plants per vessel. The nutrient solution consisted of Ca(NO3)2, 0.5 mM; MgSO4, 2 mM; KNO3, 0.5 mM; K2HPO4, 0.1 mM; CuSO4, 0.2 μM; MnCl2, 2 μM; H3BO4 10 μM; MoO3, 0.1 μM; FeEDDHA (Fe(III)-ethylenediamine-di(o-hydroxyphenylacetic acid)), 10 μM. Zinc was added as Zn sulfate. We conducted five experiments with (1) A. petraea at starting Zn concentration of 0. Five μM for two weeks and then at Zn concentration of 10 μM, (2) A. petraea at Zn concentration of 10 μM, (3) and (4) A. halleri at Zn concentration of 10 μM and (5) A. halleri at Zn concentration of 250 μM. The Zn concentration of 10 μM is a nontoxic concentration that permits both species to grow healthily whereas 250 μM is toxic for A. petraea but not for A. halleri.10 The solution was regularly sampled, Zn concentration measured, and the solution renewed to avoid Zn depletion in the growth vessel. Nutrient solution volume variations remained lower than 5% in the course of the experiments. The plants were grown in a climatic growth chamber. For experiments (1), (2), (4), and (5), conditions were 16 h day length, 24 °C/18 °C day/ night temperature, 60% relative humidity. For experiment (3), conditions were continuous light, temperature of 24 °C, 50% relative humidity. Day conditions are 160 μmoles photons m2 s1.
ARTICLE
Figure 1. Zn concentration and δ66Zn in nutrient solutions over time. A. petraea (A.p.): a, b and A. halleri (A.h.): c,d,e. The blue dashed line represents the initial solution concentration, the green one the initial solution δ66Zn. Error bars show 2σ external reproducibility. Solution renewal is indicated by thin vertical dashed lines. The δ66Zn prior to renewal generally matches the initial one. Two exceptions are the solutions largely depleted in Zn at the end of experiments (3) and (4) (with low Zn, A. halleri).
Plants of A. petraea and A. halleri were harvested six and four weeks after their transplantation, respectively. Shoot and root from three individual plants per vessel were separated, washed several times with ultrapure (18.2 MΩ) water in order to release Zn adsorbed on root as performed in previous studies.5,19,21,22 Plant materials were freeze-dried and weighted prior to analysis. Zn Isotope Measurements. Zn was extracted by anion exchange chromatography from nutrient solutions, roots and shoots following a procedure adapted from Moynier et al.8 Concentrations of Zn were measured by quadrupole inductively coupled plasma spectrometry (ICP-MS) on nutrient solutions and on the root and shoot digests prior to Zn purification. The Zn isotope ratios were measured by multicollector inductively coupled plasma mass spectrometry (MC-ICP-MS) using an Nu plasma 500 HR (Nu instrument). Mass spectrometry and data processing procedures followed previously described methods.4,22 The Zn isotopic compositions were expressed as the relative deviation from the Zn Lyon standard JMC 30749 L 3 in %: δx Znsample ¼ ½ðx Zn=64 ZnÞsample =ðx Zn=64 ZnÞstandard 1103 ð1Þ where x = 66, 67, or 68. The total procedural blank was less than 10 ng. The instrumental reproducibility based on repetitive δ66Zn measurements of JMC 30749 L standard over six months was 0.06% (2σ). 9213
dx.doi.org/10.1021/es200874x |Environ. Sci. Technol. 2011, 45, 9212–9217
Environmental Science & Technology
ARTICLE
Measurements of six replicates (including sample digestion and column separation) from the initial nutrient solution yield an external reproducibility of 0.08% (2σ). Duplicates from shoot or root material fell within external reproducibility. We checked that isotope fractionation is mass-dependent for all the samples (δ67Zn = 1.47 δ66Zn, r2 = 0.99, n = 134) and we report here one isotopic ratio 66Zn/64Zn. To express the isotope fractionation between two components A and B, we used Δ66ZnA-B that equals the difference δ66ZnA δ66ZnB.
’ RESULTS Zn Concentration and Isotope Composition in Nutrient Solution. The evolution of the Zn concentration in the nutrient
solution ([Zn]sol.) over time is shown in Figure 1. The end solution concentration before renewal is generally similar to the initial state for A. petraea (Figure 1a, b). In contrast, when grown in 10 μM Zn plants of A. halleri quickly depleted the medium in Zn after the fourth week transplantation (Figure 1c,d), which is why plants of A. halleri were harvested at that stage. When an initial concentration of 250 μM Zn was used, Zn concentration systematically dropped to ca. 200 μM in the end solution before renewal (Figure 1e). The variation in the concentration of other metals (Fe, Cu, Mn, Ca, Mg, K, Mo) between initial and end solutions falls within analytical uncertainty with uptake being quite small relative to the metal stock in solution with the exception of Mo. The pH was 7.3 for initial media. Prior to the renewal of the solutions, it dropped to 6.8 in the low Zn media and to 6.2 in the high Zn media. The theoretical equilibrium pH, Zn speciation and saturation indices for Zn minerals were calculated with the MINTEQA2 program and EDDHA acidity constants and complexation constants from references.23,24 The predicted pH for the nutrient solution at equilibrium with atmospheric CO2 is 7.3 in good agreement with measurements in initial solutions. Free Zn2+ is the major species (84%) and aqueous ZnSO4 a minor species (14%). Undersaturation versus Zn solid phases was found except for hopeite (Zn3(PO3)4,4H2O). The low Zn media are close to saturation versus hopeite (with a saturation index of 0.4 at pH of 7.3 to 0.9 at pH of 6.8). The high Zn media are initially oversaturated (with a saturation index of 4.6); they become close to saturation with a final concentration of 200 μM Zn and a pH of 6.2 (with a saturation index of 0.5). Thus, the systematic loss of Zn and the decrease in pH that were observed in the high Zn media may be due to phosphate precipitation. The isotopic composition of the nutrient solution (δ66Znsol.) over time is shown in Figure 1. The isotopic composition of the initial solution is 0.20 ( 0.08%. The isotopic composition of the nutrient solution prior to renewal generally matches the initial one within analytical uncertainty. Two exceptions were the solutions most depleted in Zn at the end of experiments (3) and (4) that showed a large enrichment in heavy isotopes (Figure 1c,d). For each experiment, the isotope composition of the source of Zn for the plant (δ66Znsource) is calculated as the average isotope composition of the nutrient solution measured at each step. Zn Concentration, Mass and Isotopic Composition in Root and Shoot. The Zn concentrations in root ([Zn]root) and shoot ([Zn]shoot) are given for the five experiments, (1) to (5), in Figure 2a,b. The different plants analyzed show large differences in zinc partitioning between hydroponic solution, root and shoot. The [Zn]root, [Zn]shoot, and [Zn]shoot/[Zn]root ratio (ca. 48
Figure 2. Concentration in Zn (a,b), dry mass (c,d) and Zn mass (e,f) of root and shoot of the nonaccumulator A. petraea and hyperaccumulator A. halleri for the five experiments (1) to (5). The initial Zn concentrations are indicated for each experiment. DW: dry weight.
mg/g, 0.50.6 mg/g, 0.1, respectively) of A. petraea grown at 10 μM matches those reported for this species under similar conditions.10,20 In low Zn medium (10 μM Zn), the hyperaccumulator A. halleri has a [Zn]root close to that of A. petraea, while it has a higher [Zn]shoot and a higher [Zn]shoot/[Zn]root ratio (of 3). The A. halleri plants grown in the high Zn medium (250 μM Zn) show only slight increase in [Zn]shoot but 10-fold-increase in [Zn]root and consequently a low [Zn]shoot/[Zn]root ratio (of 0.4). Our observed [Zn]shoot and [Zn]root are higher than those reported previously in low Zn medium for A. halleri.10,15,19,21 As A. halleri plants grown for several weeks quickly removed Zn from the low Zn medium, the above discrepancy can be explained by our efforts to maintain the Zn concentration constant over the duration of the experiment leading to higher availability and uptake by the plants. Zn partitioning between high Zn medium, root and shoot was in good agreement with a previous study made under similar conditions.19 When looking at Zn transfer in the plant, it is essential to determine precise mass-balances of the element in the different organs.5 Masses of Zn in the root MZn,root and shoot MZn,shoot (Figure 2e,f) were calculated using dry biomasses of root Mroot and of shoot Mshoot (Figure 2c,d) and Zn concentrations of root and shoot (Figure 2a,b). The mass of Zn stored in the root of A. petraea (Figure 2e) is equal to or higher than that exported to the shoot. In contrast, the mass of Zn stored in the root of the hyperaccumulator (Figure 2f) is much smaller than that stored in the shoot with shoot-to-root Zn mass ratios decreasing between low and high Zn media (1015 to 2). The Zn isotope compositions of the root (δ66Znroot) and the shoot (δ66Znshoot) of individual plants are shown, along with those of the Zn source, in Figure 3. For each experiment, the roots of individual plants generally fell in a quite narrow range as did the shoots. Depletion in heavy isotopes is observed from root 9214
dx.doi.org/10.1021/es200874x |Environ. Sci. Technol. 2011, 45, 9212–9217
Environmental Science & Technology
ARTICLE
Figure 4. Zn isotope compositions as a function of fraction of Zn left (f) for nutrient solutions of A. halleri plants under low Zn (10 μM). The curve represents the best fit between δ66Znsol. and ln f using a Rayleigh fractionation model for plant uptake.
older A. halleri plants grown under low Zn conditions (Figure 1c,d). The Zn drop and isotope shift depended on species and plant age and are thus unequivocally related to plant uptake. The solution can be considered as an open system with Zn loss limited to plant uptake. A Rayleigh type mass-balance yields: δ66 Znsol: ¼ δ66 Znsol:, 0 þ δ66 Znin-source ln f
Figure 3. Zn isotope compositions of source Zn (black disks), root (gray disks), and shoot (circles) of individual plants for the five experiments (1) to (5) (A. p.: A. petraea; A. h.: A. halleri; 10 or 250 μM: initial Zn concentration of the nutrient solution). External reproducibility is 0.08% (2σ). Isotope composition estimates for the whole plant δ66Znin (triangles) are based on Zn masses and isotopic compositions of roots and shoots.
to shoot for the two species with average Δ66Znroot-shoot of 0.64 ( 0.22% in A. petraea (2σ; n = 6) and 0.66 ( 0.22% in A. halleri (2σ; n = 9). The two species present large differences of root and shoot δ66Zn values. The roots of A. petraea lie close to that of source zinc while those of A. halleri are enriched in heavy isotopes by 0.4 to 0.8%. Shoots are more largely depleted in heavy isotopes relative to source zinc in A. petraea than in A. halleri. Mean depletion from source zinc to shoot is 0.15 ( 0.14 % in A. halleri and 0.66 ( 0.22 % in A. petraea.
’ DISCUSSION Isotope Fractionation with Zn Uptake. Our experimental setup generally allowed us to keep the isotope composition of the source Zn constant in the different growth experiments (Figure 1). Thus, the isotope fractionation due to uptake can be estimated from the difference between the δ66Zn of the whole plant and that of δ66Znsource. The isotope composition of the whole plant Zn δ66Znin (Figure 3) can be estimated according to the following:
δ66 Znin ðMZn, root þ MZn, shoot Þ ¼ δ66 Znroot MZn, root þ δ66 Znshoot MZn, shoot
ð2Þ
The fractionation between plant and source (Δ66Znin-source) yielded overall depletion in heavy isotope ranging from 0.36 to 0.12% in A. petraea (mean 0.19 ( 0.20%) and from 0.13 to 0% in A. halleri (mean 0.05 ( 0.12%). The fractionation linked to uptake can also be inferred from the isotopic compositions of the nutrient solutions that show a large Zn drop and isotope shift. This was observed only for the
ð3Þ
where δ66Znsol. and δ66Znsol., 0 are the δ66Zn of actual and initial solution respectively. The fraction of Zn left in solution f is given by the ratio between actual and initial Zn concentration in solution (as the volume variation of the solution was of less than 5% in the course of the experiment). The fitting parameters between δ66Znsol. and ln f (Figure 4) yield a depletion in heavy isotopes due to plant uptake (Δ66Znin-source) of 0.15%, which is consistent with the Δ66Znin-source values inferred from the isotopic compositions of plant and source Zn. The isotope fractionation from medium to the plant (Δ66Znin-source) possibly results from three different mechanisms: transportermediated uptake by root cells, adsorption-precipitation on outer root surfaces and advection with the transpiration stream. Advection does not fractionate isotopes and will reduce the putative effects of other processes. Zn adsorbed on root was removed with ultrapure water following previous studies.5,19,21 Most of the Zn adsorbed on outer root surfaces (and cortical cell walls) should be removed after this extraction.25 Zinc phosphate precipitates have been detected by SEM-EDX on the outer root surfaces of A. halleri plants grown on hydroponics. However, these precipitates have only been observed for A. halleri plants growing at higher Zn concentration than ours (500 μmol Zn) and not at lower concentrations. Thus, the contribution of root surface Zn in our study should be very limited. The Zn isotope data is consistent with this assumption. Indeed, adsorption-precipitation on the outer root surfaces cannot explain the large difference in root δ66Zn that is observed between A. halleri and A. petraea plants grown in similar media. Furthermore, Zn of cell wall has been found enriched in heavy isotopes relative to Zn2+ in the external medium.6,26 Adsorptionprecipitation on outer cell surfaces thus cannot account for the overall depletion in heavy isotopes from medium to the plant that is likely due to transport-mediated uptake. Molecular studies have demonstrated that Zn2+ is taken up in the root cell by transporters of the ZIP family.11 The isotope fractionation for uptake (Δ66Znin-source) close for A. petraea and A. halleri confirms that similar transporters operate in the hyperaccumulator and nonaccumulator. 1315 Noticeably, the Δ66Znin-source values match that reported for the highaffinity Zn transporter of the plasma membrane in diatoms, e.g., 0.2%.6 This suggests that the ZIP-transporters active in 9215
dx.doi.org/10.1021/es200874x |Environ. Sci. Technol. 2011, 45, 9212–9217
Environmental Science & Technology
ARTICLE
Figure 5. Δ66Znshoot-in as a function of F, the fraction of Zn translocated to shoot F. F is given by Zn mass-ratio between shoot and whole plant. The curve represents the best fit between Δ66Znshoot-in and ln F. The fit supports Rayleigh fractionation of Zn isotopes in the root and an associated isotope fractionation of 0.37% during root sequestration for both plant species.
Arabidopsis have a high affinity for Zn. In our growth media, theoretical speciation predicts Zn2+ to be the dominant species and not significantly complexed by organic ligand. The depletion in heavy isotopes from the medium to the plant is thus likely due to the interaction of Zn2+ with the ZIP transporters. Isotope Model for Root-Shoot Translocation. While the hyperaccumulator and nonaccumulator present relatively small heavy isotope depletion with uptake (Δ66Znin-source), they largely differ by their shoot and root isotope compositions. The shoot is depleted in heavy isotopes relative to whole plant Zn in the nonaccumulator (with Δ66Znshoot-in values ranging from 0.59 to 0.21%) and only slightly depleted in the hyperaccumulator (with Δ66Znshoot-in values ranging from 0.22 to 0.03%). These two species differ essentially in terms of Zn uptake and root-shoot translocation. Plotting Δ66Znshoot-in as a function of the fraction of Zn translocated to shoot F confirmed the correlation between the two variables (Figure 5). To model the relation between Δ66Znshoot-in and F, we tested a Rayleigh mass-balance describing the loss of Zn in root as Zn moved toward the xylem: Δ66 Znshoot-in ¼ δ66 Znshoot δ66 Znin ¼ Δ66 Znstorage ln F ð5Þ where δ66Znin represents the δ66Zn of Zn incorporated in the plant, Δ66Znstorage the isotope discrimination between sequestered and mobile (free or weakly bound) Zn in the root. The data produced an excellent fit between Δ66Znshoot-in and lnF (Figure 5) thus supporting the scenario of a progressive sequestration of Zn in the root during its radial transfer to the xylem. The fitting parameters yielded a single value for Δ66Znstorage of 0.37% for both species. This indicates the occurrence of a large and consistent isotope fractionation associated with root sequestration in the two species. Our results confirm that the mechanisms of root-shoot translocation in both species may be identical 1218 at the exception of the fluxes involved. Root Sequestration Processes. The chemical form of Zn has been determined by EXAFS in roots of A. halleri and A. petraea grown in media similar to those used here.19 Zn has been found to occur partly as Zn malate in roots of A. petraea and otherwise to be bound to phosphate groups in the roots of A. halleri and A. petraea. Zinc bound to phosphate could be either disordered inorganic phosphate or organic phosphate (phytate) presumably present in cell wall and inside the cell, respectively.19,20 Isotopic fractionation linked to the precipitation of Zn phosphate has not
Figure 6. Zn isotope compositions as a function of fraction of Zn left (f) for nutrient solutions (250 μM Zn) initially oversaturated versus Zn phosphate (hopeite). Hypothetical precipitation curves for closed system isotopic equilibrium (dashed) and for Rayleigh fractionation (bold) are shown for an isotope fractionation of 0.4% between precipitate and solution.
yet been documented. In our experiments, the high Zn nutrient solutions suffered a systematic 20% drop in Zn concentration related to phosphate precipitation. This allowed us to attempt a characterization of the isotope fractionation associated with Zn phosphate precipitation. For the fraction of Zn left in solution f in present experiments, no significant isotope shift is observed between initial and final solution (Figure 2e). Model isotope compositions of the solution were calculated using an isotope discrimination of 0.4 % between Zn precipitated and Zn in solution. The measured data lie significantly above the model curves (Figure 6). Available isotope data thus show that inorganic phosphate precipitation does not account for the isotope fractionation associated with root storage (∼0.4%) and rather support the hypothesis of cellular sequestration. Cellular storage is also indicated by upregulation of vacuolar Zn transporter MTP1 (metal tolerance protein 1) upon exposure to high Zn concentration in roots of A. halleri.12 The large positive isotope fractionation associated with root storage is consistent with isotopic exchange between free (or weakly bound) Zn2+ and Zn bound to a high affinity ligand. For comparison, large positive isotope discrimination of 0.4% has been reported between Zn of high affinity sites of organic macromolecules (humic acids) and free Zn2+ in solution.27 Implications for Biogeochemical Cycling of Zn. Several biogeochemical implications arise from this work. First, monitoring of the isotope ratio in the growth media and isotope measurements in final roots and shoots show an overall depletion in heavy isotopes from the growth medium to the plant (Δ66Znin-source = 0.36 to 0%, n = 15). Theoretical Zn speciation predicts that the free cation Zn2+ is the dominant species in the initial nutrient solutions used. Thus, the observed isotope fractionation is likely due to the biological uptake of Zn and not to Zn complexation in the solution. Consequently, the bioavailable Zn is expected to get enriched in heavy isotopes with increasing uptake. Second, our results confirm the occurrence of a large systematic depletion in heavy isotopes from root to shoot.5 Furthermore, we show that the isotope fractionation between shoot Zn and whole plant Zn (Δ66Znshoot-in) differs between the hyperaccumulator A. halleri and the nonaccumulator A. petraea as a function of the extent of translocation from root to shoot. This indicates that the δ66Zn of shoot and zinc source can be used to estimate the relative sizes of underground and aerial Zn sinks in plants. Third, Zn stable isotope measurements could usefully complement spectroscopic investigations of Zn storage forms. Indeed, previous EXAFS investigations in roots of A. halleri and petraea have demonstrated Zn binding to phosphate,19 but could not differentiate between cellular sequestration in organic phosphate and extracellular precipitation in 9216
dx.doi.org/10.1021/es200874x |Environ. Sci. Technol. 2011, 45, 9212–9217
Environmental Science & Technology inorganic phosphate. Our data show a relatively small isotope fractionation associated with inorganic phosphate precipitation and a large isotope fractionation (∼0.4%) associated with Zn storage in root. Thus, δ66Zn of root and source zinc could help to trace extracellular or cellular mechanisms of sequestration in the root.
’ ASSOCIATED CONTENT
bS
Supporting Information. Dry biomasses, Zn concentration, δ66Zn of shoot and root and isotope fractionations for the A. halleri and A. petraea plants studied (Table S1). This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*author for correspondence (Tel.: 33(0)472448416, Fax: 33(0)4728593, E-mail:
[email protected]).
’ ACKNOWLEDGMENT We thank Chantal Douchet and Philippe Telouk for their assistance in sample preparation and analysis and Deborah Lloyd for her assistance in the plant growth setup. This work was supported by the Program Ecosphere Continentale et C^otiere from CNRS-INSU (France). We also wish to thank the three anonymous reviewers for their comments that greatly help to clarify the present manuscript. ’ REFERENCES (1) Chaney, R. L. Zinc Phytotoxicity. In Zinc in Soil and Plants; Robson, A. D., Ed.; Kluwer Academic Publishers: Dordrecht, 1983, pp 135150. (2) Hacisalihoglu, G.; Kochian, L. V. How do some plants tolerate low levels of soil zinc? Mechanisms of zinc efficiency in crop plants. New Phytol. 2003, 159, 341–350. (3) Marechal, C. N.; Telouk, P.; Albarede, F. Precise analysis of copper and zinc isotopic composition by plasma-source mass spectrometry. Chem. Geol. 1999, 156, 251–273. (4) Albarede, F. The stable isotope geochemistry of copper and zinc. Rev. Mineral. Geochem. 2004, 55, 409–427. (5) Weiss, D. J.; Mason, T. F. D.; Zhao, F. J.; Kirk, G. J. D.; Coles, B. J.; Horstwood, M. S. A. Isotopic discrimination of zinc in higher plants. New Phytol. 2005, 165, 703–710. (6) John, S. G.; Geis, R. W.; Saito, M. A.; Boyle, E. A. Zinc isotope fractionation during high-affinity and low-affinity zinc transport by the marine diatom Thalassiosira oceanica. Limnol. Oceanogr. 2007, 52 (6), 2710–2714. (7) Viers, J.; Oliva, P.; Nonell, A.; Gelabert, A.; Sonke, J. E.; Freydier, R.; Gainville, R.; Dupre, B. Evidence of Zn isotopic fractionation in soilplant system of a tropical pristine watershed (Nsimi, Cameroon). Chem. Geol. 2007, 239, 124–137. (8) Moynier, F., Pichat, S., Pons, M. L., Fike, D., Balter, V., Albarede, F. 2009. Isotopic fractionation and transport mechanisms of Zn in plants. Chem. Geol. 2009, 267, 125-130. (9) von Blanckenburg, F.; von Wiren, N.; Guelke, M.; Weiss, D. J.; Bullen, T. Fractionation of metal stable isotopes by higher plants. Elements 2009, 5, 375–380. (10) Macnair, M.; Bert, V.; Huitson, S. B.; Saumitou-Laprade, P.; Petit, D. Zinc tolerance and hyperaccumulation are genetically independent characters. Proc. R. Soc. London B. 1999, 266, 2175–2179. (11) Colangelo, E. P.; Guerinot, M. L. Put the metal to the petal: Metal uptake and transport throughout plant. Curr. Opin. Plant Biol. 2006, 9, 322–330.
ARTICLE
(12) Dr€ager, D. B.; Desbrosses-Fonrouge, A.-G.; Krach, C.; Chardonnens, A. N.; Meyer, R. C.; Saumitou-Laprade, P.; Kr€amer, U. Two genes encoding Arabidopsis halleri MTP1 metal transport proteins co-segregate with zinc tolerance and account for high MTP1 transcript levels. Plant J. 2004, 39, 425–439. (13) Weber, M.; Harada, E.; Vess, C.; v. Roepenack-Lahaye, E.; Clemens, S. Comparative microarray analysis of Arabidopsis thaliana and Arabidopsis halleri roots identifies nicotianamine synthase, a ZIP transporter and other genes as potential hyperaccumulation factors. Plant J. 2004, 37, 269–281. (14) Filatov, V.; Dowdle, J.; Smirnoff, N.; Ford-Llyod, B.; Newburry, H. J.; Macnair, M. A quantitative trait loci analysis of zinc hyperaccumulation in Arabidopsis halleri. New Phytol. 2007, 174, 580–590. (15) Talke, I. N.; Hanikenne, M.; Kr€amer, U. Zinc-dependent global transcriptional control, transcriptional deregulation, and higher gene copy number for metal of genes homeostasis of the hyperaccumulator Arabidopsis halleri. Plant Physiol. 2006, 142, 148–167. (16) Hanikenne, M.; Talke, I. N.; Haydon, M. J.; Lanz, C.; Nolte, A.; Motte, P.; Kroymann, J.; Weigel, D.; Kr€amer, U. Evolution of metal hyperaccumulation required cis-regulatory changes and triplication of HMA4. Nature 2008, 453, 391–395. (17) Verbruggen, N.; Hermans, C.; Schat, H. Molecular mechanisms of metal hyperaccumulation in plants. New Phytol. 2009, 181, 759–776. (18) Van der Zaal, B. J.; Neuteboom, L. W.; Pinas, J. E.; Chardonnens, A. N.; Schat, H.; Verkleij, J. A. C.; Hookyas, P. J. J. Overexpression of a novel Arabidopsis gene related to putative zinc-transporter genes from animals can lead to enhanced resistance and accumulation. Plant. Physiol. 1999, 119, 1047–1055. (19) Sarret, G.; Saumitou-Laprade, P.; Bert, V.; Proux, O.; Hazemann, J. L.; Traverse, A.; Marcus, M. A.; Manceau, A. Forms of zinc accumulated in hyperaccumulator Arabidopsis halleri. Plant Physiol. 2002, 130, 1815– 1826. (20) Sarret, G.; Willems, G.; Isaure, M. P.; Marcus, M. A.; Fakra, S. C.; Frerot, H.; Pairis, S.; Geoffroy, N.; Manceau, A.; SaumitouLaprade, P. Zinc distribution and speciation in Arabidopsis halleri Arabidopsis lyrata progenies presenting various zinc accumulation capacities. New Phytol. 2009, 184, 581–595. (21) K€upper, H.; Lombi, E.; Zhao, F. J.; McGrath, S. Cellular compartmentation of cadmium and zinc in relation to other elements in the hyperaccumulator Arabidopsis halleri. Planta 2000, 212, 75–84. (22) Pichat, S.; Douchet, C.; Albarede, F. Zinc isotope variation in deep-sea carbonates from the eastern equatorial Pacific over the last 175 ka. Earth Planet. Sci. Lett. 2003, 210, 167–178. (23) Bannochie, C. J.; Martell, A. E. Affinities of racemic and meso forms of N,N0 -Ethylenebis (2-(o-hydroxyphenyl)glycine) for divalent and trivalent metal ions. J. Am. Chem. Soc. 1989, 111, 4735–4642. (24) Yunta, F.; Garcia-Marco, S.; Lucena, J. L.; Gomez-Gallego, M.; Alcazar, R.; Sierra, M. A. Chelating agents related to ethylenediamine bis(2-hydroxyphenyl)acetic acid (EDDHA): synthesis, characterization and equilibrium studies of the free ligands and their Mg2+, Ca2+, Cu2+ and Fe3+ chelates. Inorg. Chem. 2003, 42, 5412–5421. (25) Straczek, A.; Sarret, G.; Manceau, A.; Hinsinger, P.; Geoffroy, N.; Jaillard, B. Zinc distribution and speciation in roots of tobacco exposed to Zn. Environ. Exp. Bot. 2008, 63, 80–90. (26) Gelabert, A.; Prokrovsky, O. S.; Viers, J.; Schott, J.; Boudou, A.; Feurtet-Mazel, A. Interaction between zinc and freshwater and marine diatom species: Surface complexation and Zn isotope fractionation. Geochim. Cosmochim. Acta 2006, 70, 839–857. (27) Jouvin, D.; Louvat, P.; Juillot, F.; Marechal, C.; Benedetti, M. Zinc isotope fractionation: Why organic matters. Environ. Sci. Technol. 2009, 43 (15), 5747–5754.
9217
dx.doi.org/10.1021/es200874x |Environ. Sci. Technol. 2011, 45, 9212–9217
ARTICLE pubs.acs.org/est
Arsenite Oxidation by a Poorly-Crystalline Manganese Oxide. 3. Arsenic and Manganese Desorption Brandon J. Lafferty,*,† Matthew Ginder-Vogel,‡ and Donald L. Sparks Department of Plant and Soil Sciences, Delaware Environmental Institute, University of Delaware, 152 Townsend Hall, Newark, Delaware 19716, United States
bS Supporting Information ABSTRACT: Arsenic (As) mobility in the environment is greatly affected by its oxidation state and the degree to which it is sorbed on metal oxide surfaces. Manganese oxides (Mn oxides) have the ability to decrease overall As mobility both by oxidizing toxic arsenite (AsIII) to less toxic arsenate (AsV), and by sorbing As. However, the effect of competing ions on the mobility of As sorbed on Mn-oxide surfaces is not well understood. In this study, desorption of AsV and AsIII from a poorly crystalline phyllomanganate (δ-MnO2) by two environmentally significant ions is investigated using a stirred-flow technique and X-ray absorption spectroscopy (XAS). AsIII is not observed in solution after desorption under any conditions used in this study, agreeing with previous studies showing As sorbed on Mn-oxides exists only as AsV. However, some AsV is desorbed from the δ-MnO2 surface under all conditions studied, while neither desorptive used in this study completely removes AsV from the δ-MnO2 surface.
’ INTRODUCTION Arsenic (As) is an element with toxic properties, commonly found in the environment. Elevated As levels in soils result from both natural weathering processes and anthropogenic activities such as mining, agriculture, and manufacturing.1 In several locations throughout the world, As contamination of soil and water occurs near human populations, posing a significant threat to human health. Therefore, understanding the chemical reactions controlling As mobility in the environment is critical. Arsenic behavior in the environment is significantly determined by its chemical speciation. Usually, As occurs as one of two inorganic oxyanions: arsenite (AsIII) or arsenate (AsV). Below pH 9, AsIII appears predominately in its fully protonated form (H3AsO3), while at circumneutral pH values, AsV occurs as a mixture of H2AsO4 and HAsO42‑.2 Arsenic speciation also determines its toxicity because AsIII is more toxic than AsV.3 Several Mn-oxides can readily oxidize AsIII to AsV, most notably layered Mnoxides (i.e., phyllomanganates),414 thus Mn-oxide minerals can have a determining effect on As speciation in soils and sediments. In terrestrial environments, As mobility is generally determined by the extent to which it is adsorbed by metal oxides.1521 Phyllomanganates represent one type of metal oxide that exhibit the ability to sorb As.6,11,12,2225 However, sorption of As on Mnoxide surfaces can be quite complex. Interestingly, a higher level of As sorption has been observed when AsIII is reacted with phyllomanganates (oxidation and sorption) compared to reaction of AsV with phyllomanganates (sorption alone).4,6,11,23,25 Also, when AsV is sorbed on Mn-oxide surfaces, it forms a variety of surface complexes.14,2224 r 2011 American Chemical Society
Since Mn-oxides have shown a propensity to oxidize and sorb As in nature, it is important to understand potential mobility of As adsorbed on Mn-oxide surfaces. Because AsV forms multiple surface complexes with Mn-oxides, it is possible that mobility of As sorbed on Mn-oxides varies depending on the surface complexes present. Also, in nature, As coexists with other ions which can compete for sorption sites on Mn-oxide surfaces. Therefore, understanding As mobility in the environment requires understanding the mobility of various As surface complexes, as well as the potential for competing ions to desorb As from Mn-oxide surfaces. To date, few studies have investigated the ability of common environmental ions to desorb As from the surface of Mn-oxides. The purpose of this study is to determine to what extent a cation (Ca2+), and oxyanion (H2PO4/HPO42‑), both common in the environment, are able to desorb As from a poorly crystalline phyllomanganate (δ-MnO2).
’ MATERIALS AND METHODS Stirred Flow Method. Stirred flow experiments were conducted using the same 30 mL reactor and experimental procedures described previously.6 All stirred flow reactions were conducted in a background electrolyte (10 mM NaCl), buffered at pH 7.2 (5 mM 3-(N-morpholino)propanesulfonic acid (MOPS)), Received: April 15, 2011 Accepted: September 28, 2011 Revised: August 9, 2011 Published: September 28, 2011 9218
dx.doi.org/10.1021/es201281u | Environ. Sci. Technol. 2011, 45, 9218–9223
Environmental Science & Technology and had a constant flow rate of 1 mL/min. Also, all experiments were mixed via a magnetic stir bar with a constant rate of stirring (100 rpm). In all desorption reactions, 1 g/L δ-MnO2 was reacted with 100 μM AsIII (i.e., As is oxidized and sorbed) to ensure maximum As sorption on the solid phase, followed immediately by reaction with a desorptive. Oxidation of AsIII (and As sorption) prior to desorption was carried out for 4, 10, or 24 h. The desorptives used were 100 μM calcium chloride (CaCl2, abbreviated Ca2+) in the presence of background electrolyte (10 mM NaCl and 5 mM MOPS), 100 μM sodium phosphate (NaH2PO4, abbreviated PO4) in the presence of background electrolyte, or background electrolyte alone. All solutions were adjusted to pH 7.2 with HCl and NaOH prior to experiments, and background electrolyte was introduced into the stirred flow reactor (containing δ-MnO2) for at least 2 h at a rate of 1 mL/min prior to each experiment. In each reaction, AsIII oxidation was stopped (after 4, 10, or 24 h) by introducing desorbing solution into the reactor, thus beginning the desorption phase of the reaction. Changing influent solution from AsIII solution to desorption solution took less than 5 s, and thus flow of solution into the reactor was effectively constant throughout each experiment (oxidation followed by desorption). A plot showing the full data (AsIII oxidation followed by desorption) from one experiment is presented in the Supporting Information (Figure S1). As, Mn, Ca, and P Analysis. Inorganic arsenic in the stirred flow effluent was analyzed via liquid chromatography inductively coupled plasma mass spectrometry (LC-ICP-MS), and aqueous Mn was analyzed by ICP-MS as described in Lafferty et al.6 Total P and Ca in the reactor effluent were analyzed by ICP-OES to verify the concentration of desorptive present in the stirred-flow reactor (data not shown). EXAFS Analysis. Extended X-ray absorption fine structure (EXAFS) spectroscopic analysis was performed at the National Synchrotron Light Source (Brookhaven National Laboratory) on reacted δ-MnO2 after 24 h of desorption (following reaction with 100 μM AsIII for 4, 10, and 24 h). Detailed descriptions of sample collection, data collection and data processing for samples analyzed using these techniques can be found in the Supporting Information. Sorption and Desorption Calculations. To quantify the total amount of sorbed and desorbed AsV, AsIII, and Mn2+ data were integrated using the “area below curves” tool in SigmaPlot 8 (Systat Software Inc., San Jose, California).
’ RESULTS AND DISCUSSION δ-MnO2 Structure. Mineral structure must be taken into consideration in order to accurately interpret adsorption and desorption data. The δ-MnO2 used in this study is a poorly ordered form of hexagonal birnessite.24 Hexagonal birnessite (and thus δMnO2) has two types of reactive sites: at vacancies within MnIV octahedral sheets (vacancy sites) and at the edges of MnIV octahedral sheets (edge sites).26,27 It has been shown that As reacts primarily with edge sites, rather than vacancy sites of birnessite;14,23,24 therefore, in this study, As desorption is expected to occur at δ-MnO2 edge sites. However, heavy metals tend to sorb strongly at phyllomanganate vacancy sites,2633 as well as react with edge sites of phyllomanganates.31,34 Because of the high affinity of vacancy sites for Mn2+, vacancy sites are likely the primary location of Mn2+ sorption on δ-MnO2, however, once δ-MnO2 vacancy sites begin to fill, Mn2+ is expected to react
ARTICLE
Figure 1. The amount (nmol) of AsIII, AsV, and Mn2+ in stirred flow reactor effluent as well as the amount (nmol) of As sorbed during AsIII oxidation by δ-MnO2, prior to desorption by Ca2+, PO4, or background electrolyte. Vertical dashed lines indicate times for which desorption was initiated.
more with δ-MnO2 edge sites.6,24 Therefore, in these experiments, desorption of Mn2+ is expected to occur at both edge and vacancy sites of δ-MnO2 depending on the extent to which δ-MnO2 vacancy sites are filled with Mn2+. A graphic representation of As and Mn sorption on the δ-MnO2 surface can be found in Figure 4 of Lafferty et al.24 AsIII Oxidation and Sorption. All desorption experiments in this study are preceded by reaction of AsIII with δ-MnO2 for 4, 10, or 24 h. This initial reaction between AsIII and δ-MnO2 results in AsIII oxidation and produces AsV and Mn2+ (eq 1). Subsequently, AsV and Mn2+ produced during AsIII oxidation are sorbed on the δ-MnO2 surface. > MnIV OH þ H3 AsIII O3 ðaqÞ f Mn2þ ðaqÞ þ HAsV O4 2 ðaqÞ þ 3Hþ
ð1Þ
To briefly summarize Lafferty et al.,6,24 the reaction between AsIII and δ-MnO2 in a stirred-flow reactor, under the conditions used in this study, proceeds in two distinct phases. First, an initial reaction phase occurs from 0 to 6.4 h, which includes the period of fastest AsIII oxidation, highest AsV sorption, and no Mn2+ release into solution (Figure 1). A second reaction phase characterized by lower δ-MnO2 reactivity occurs beyond 6.4 h, which includes a second period of (decreased) As sorption, a decrease in AsIII oxidation rate, and the presence of Mn2+ in solution (Figure 1). Decreased δ-MnO2 reactivity in the second phase of this reaction has been attributed to Mn2+ sorption on the δ-MnO2 surface and the subsequent production of MnIII via Mn(II)/(IV) conproportionation at the δ-MnO2 surface.24 In this study, desorption experiments are conducted by stopping the initial reaction between AsIII and δ-MnO2 after 4, 10, or 24 h (Figure 1), and simultaneously beginning desorption by PO4, Ca2+, or background electrolyte alone. The first time point for beginning desorption is after 4 h of reaction between AsIII and δ-MnO2, which coincides with maximum AsV concentration in the stirred-flow reactor effluent, and the end of an initial period of AsV sorption (Figure 1). Between 0 and 4 h of AsIII oxidation, Mn2+ is expected to react primarily with vacancy sites and not edge sites. The second time point for beginning desorption is after 10 h of AsIII oxidation by δ-MnO2, which is near the end of a second period of lesser As sorption, and occurs early in the second, less reactive phase of AsIII oxidation (Figure 1). Between 4 and 10 h, Mn2+ is expected to begin reacting with edge sites, resulting in formation of some MnIII.24 A change in the sorption 9219
dx.doi.org/10.1021/es201281u |Environ. Sci. Technol. 2011, 45, 9218–9223
Environmental Science & Technology
ARTICLE
Table 1. Structural Parameters Derived from Least-Square Fits to Raw k3-Weighted As-EXAFS Spectra for δ-MnO2 after 24 h Desorption by Background Electrolyte (Elec), Calcium Solution (Ca), and Phosphate Solution (PO4)a sample
AsO b
b
AsMn σ
b
r
b
AsMn CN
rb
σ2b
0.005(1)
0.6(4)
3.50(5)
0.005(3)
0.005(2)
0.3(5)
3.51(9)
0.004(7)
time
CN
r
10 h-elec
4.1(2)
1.70(1)
0.003(0)
1.1(3)
3.12(2)
10 h-Ca
4.1(2)
1.70(1)
0.003(0)
0.7(4)
3.14(3)
10 h-PO4 24 h-elec
4.2(2) 4.1(1)
1.70(1) 1.70(1)
0.003(0) 0.003(0)
0.8(3) 0.8(3)
3.16(3) 3.15(2)
0.005(2) 0.005(2)
2b
CN
σ
2b
24 h-Ca
4.1(1)
1.69(0)
0.003(0)
0.9(3)
3.14(2)
0.005(2)
24 h-PO4
4.3(2)
1.69(0)
0.003(0)
0.7(3)
3.16(3)
0.006(3)
b
a Desorption data shown here followed 10 or 24 h of AsIII oxidation. b Coordination number (CN), interatomic distance (r), and DebyeWaller factor (σ2) were obtained by fitting data with theoretical phase and amplitude functions. Estimated errors at 95% confidence interval from the least-squares fit are given in parentheses.
complexes formed between AsV and δ-MnO2 also occurs between 4 and 10 h of AsIII oxidation.24 The last time point for desorption is after 24 h of reaction, when the system is stable within the less reactive phase of the reaction (Figure 1). AsIII Desorption. No AsIII is desorbed from the δ-MnO2 surface in any desorption experiments discussed here. Also, As EXAFS analysis indicates that all As associated with δ-MnO2 after desorption is present as AsV (Table 1 and Figures 2 and 3), which agrees with previous results indicating that As present on phyllomanganate surfaces only occurs as AsV.14,23,24,35 However, it should be noted that there is not a sufficient amount of As remaining on the surface of δ-MnO2 after 4 h of AsIII oxidation followed by 24 h of desorption to measure As using EXAFS analysis. AsV Desorption. Previous studies have shown that As reacts primarily with edge sites of phyllomanganates rather than vacancy sites,14,23,24 therefore AsV desorption in this study is expected to occur at δ-MnO2 edge sites. Of the desorptives used in this study, PO4 is expected to desorb AsV most readily because it is chemically similar to AsV and is known to compete with AsV for sorption sites on metal oxide minerals.12,3638 However, Ca2+ has the potential to react with δ-MnO2 vacancy sites as well as edge sites, therefore, Ca2+ also has the potential to desorb AsV. The background electrolyte (MOPS and NaCl) used in these studies is expected to react weakly with δ-MnO2 edge sites, and thus should not desorb AsV to a large extent. When AsV is desorbed (for 24 h) after 4 h of AsIII oxidation, roughly 67% of AsV sorbed during the 4 h of AsIII oxidation is mobilized from the δ-MnO2 surface by all three desorptives (Figure 2). Because of this, one can infer that the majority of AsV sorbed on the δ-MnO2 surface during the initial phase of high δ-MnO2 reactivity is fairly labile. Conversely, there is a portion of AsV sorbed during the first 4 h of AsIII oxidation that remains immobile on the δ-MnO2 surface, even in the presence of PO4. Previous EXAFS analysis of δ-MnO2 reacted with AsIII under identical experimental conditions revealed that AsV is bound in mononuclear-monodentate as well as binuclear-bidentate complexes on the δ-MnO2 surface during the first 4 h of AsIII oxidation by δ-MnO2.24 Unfortunately, there is not a sufficient amount of As remaining on the surface of δ-MnO2 after 4 h of AsIII oxidation followed by 24 h of desorption to determine the stability of these two complexes by EXAFS analysis. After 10 and 24 h of AsIII oxidation, PO4 is a more efficient desorptive of AsV than Ca2+ or the background electrolyte (Figure 2). Also, the proportion of AsV desorbed by PO4
increases in the 10 and 24 h experiments (Figure 2). It should be noted that after 10 and 24 h of AsIII oxidation by δ-MnO2, two significant changes occur in the speciation of Mn associated with δ-MnO2. First, Mn2+ begins sorbing at edge sites after δ-MnO2 vacancy sites are occupied by sorbed Mn2+ (at ∼6.4 h).6,24 Also, MnIII begins to appear in Mn octahedral layers of δ-MnO2 between 4 and 10 h of AsIII oxidation, and increases between 10 and 24 h.24 An increase in the proportion of AsV desorbed by all desorptives after 10 and 24 h (compared to 4 h) happens concurrently with increased competition from Mn2+ for edge sites, and an increase in MnIII content within the δ-MnO2 structure. Thus, increased AsV desorption in the 10 and 24 h experiments could be the result of direct competition between AsV and Mn2+ for sorption sites or the formation of weaker bonds between AsV and MnIII.39 It is difficult to distinguish between the effects of increased Mn2+ sorption and increased MnIII content at δ-MnO2 edge sites as they occur simultaneously. EXAFS analysis of δ-MnO2 after AsIII oxidation for 10 h and subsequent desorption by Ca2+ and background electrolyte revealed AsMn distances of ∼3.13 Å and ∼3.50 Å (Table 1). These distances correspond to AsV bound to the δ-MnO2 surface in bidentate-binuclear and monodentate-mononuclear complexes, respectively.2224 However, for all other desorption experiments after 10 (PO4) and 24 (Ca2+, PO4, and background electrolyte) hours of AsIII oxidation by δ-MnO2 the only AsMn distances present in EXAFS spectra was ∼3.15 Å (Table 1 and Figure 3), corresponding to a bidentate-binuclear complex between AsV and the δ-MnO2 surface. While, it is tenuous to attribute a specific desorption event with a single adsorption complex, EXAFS data of As sorption complexes before and after desorption seem to indicate that bidentate-mononuclear and monodentate-mononuclear complexes between AsV and the δ-MnO2 surface are less stable than AsV- δ-MnO2 bidentatebinuclear complexes. Mn Desorption. During AsIII oxidation by δ-MnO2, Mn2+ is produced and subsequently sorbed by δ-MnO2 (eq 1 and Figure 1).6,24 Mn2+ tends to initially sorb at δ-MnO2 layer vacancy sites under the conditions used in this study, followed by sorption at δ-MnO2 edge sites as vacancy sites become more occupied.24 Previous studies have indicated that some As sorbed on phyllomanganate surfaces could be bound through a bridging complex through sorbed Mn.12,25 Although As/δ-MnO2 bridging complexes were not seen in previous studies conducted under the experimental conditions used in the reactions described here, it is possible that Mn2+ on δ-MnO2 could facilitate As sorption. 9220
dx.doi.org/10.1021/es201281u |Environ. Sci. Technol. 2011, 45, 9218–9223
Environmental Science & Technology
ARTICLE
Figure 2. AsV (left) and Mn2+ (right) desorbed by Ca2+, PO4, and background electrolyte (10 mM NaCl, 5 mM MOPS) after AsIII oxidation by δ-MnO2. The initial data points on each graph (time = 0 h) correspond to the beginning of desorption (initial AsIII oxidation data not shown). Data shown are first 10 h of 24 h desorption experiments.
Of the desorptives used in this study, Ca2+ is expected to react with δ-MnO2 sorption sites most similarly to Mn2+.40 Some cations could potentially desorb Mn2+ more readily than Ca2+,32,41,42 however Ca2+ is ubiquitous in nature, and thus has a high probability of interacting with Mn2+ sorbed on Mn-oxide surfaces. Desorption of Mn2+ by Na+ (in background electrolyte) is predicted to be negligible because Na+ reacts with δ-MnO2
interlayers differently than Ca2+ or Mn2+, in that Na+ is not expected to bind in triple corner sharing complexes at vacancy sites as is the case with Mn2+ and Ca2+.40 When AsIII is reacted with δ-MnO2 for only 4 h, no Mn2+ is desorbed under the conditions used in this study (data not shown). It is important to note that no Mn2+ appears in the stirred-flow reactor effluent during the first 4 h of AsIII oxidation, and all Mn2+ 9221
dx.doi.org/10.1021/es201281u |Environ. Sci. Technol. 2011, 45, 9218–9223
Environmental Science & Technology
ARTICLE
is a certain amount of As that is not desorbed under any of the conditions studied here. Thus, if As comes in contact with Mnoxides in nature, these minerals could potentially decrease As availability and mobility both by oxidation of AsIII and sorption of AsV. It appears that AsV and Mn2+ desorption potential is intricately linked to the type of reaction site on the δ-MnO2 surface to which each is bound, as well as Mn speciation within the δ-MnO2 structure. This study emphasizes the importance of understanding mineral structures and temporal variability when predicting As mobility in the environment.
’ ASSOCIATED CONTENT
bS
Supporting Information. Supporting Information is provided which includes detailed information about EXAFS analysis, Mn EXAFS fitting results, further Mn desorption discussion, and an example of aqueous data from a full experiment. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (601) 634-3589; fax: (601) 634-4017; e-mail: blafferty@ gmail.com. Figure 3. Fourier transformed As K-edge EXAFS of δ-MnO2 reacted with AsIII (100 μM) in a stirred-flow reactor for 10 and 24 h (10 h start and 24 h start) and desorbed by Ca2+, PO4, and background electrolyte for 24 h following AsIII reaction. XAS data are presented as solid lines and fits are presented as dashed lines (fit data provided in Table 1).
produced during this time is expected to sorb strongly at δ-MnO2 vacancy sites.24 However, after 10 and 24 h of AsIII oxidation, Mn2+ is desorbed by all desorptives studied (Figure 2), indicating that Mn2+ sorbed at δ-MnO2 edge sites is more labile than Mn2+ sorbed at δ-MnO2 vacancy sites. Mn EXAFS analysis of δ-MnO2 revealed no detectable changes in Mn speciation of the solid material after desorption which would appear as a broadening and decrease in the peak height of the 9.25 Å1 peak in the EXAFS spectra 24 (Figures S2A and S2B and Table S1 of the Supporting Information, SI). As predicted, Ca2+ is the most efficient Mn2+ desorptive of those studied. The proportion of Mn2+ desorbed by Ca2+ is greatest after 10 h of AsIII oxidation and decreases slightly after 24 h of AsIII oxidation (Figure 2). This decrease in Mn2+ mobility with increased AsIII oxidation time could potentially be due to increased formation of less mobile MnIII after 10 h of reaction.6,24 Also, desorption by PO4 (with background electrolyte present) is nearly identical to desorption by background electrolyte alone in the 10 and 24 h samples (Figure 2), which suggests that PO4 does not desorb Mn2+ appreciably. Interestingly, increased Mn2+ desorption by Ca2+ compared to other desorptives does not result in an increase in AsV desorption, which provides some evidence that AsV is not bound to the δ-MnO2 surface via a bridging complex through Mn2+. Implications for As Mobility. Phyllomanganates are capable of sorbing AsV, especially during AsIII oxidation. However, in this study, AsV can be desorbed from the δ-MnO2 surface, to some extent, under all conditions studied. Even Na+ (present in background electrolyte) is able to desorb AsV, to some extent, under all conditions studied here, indicating that a portion of As sorbed by Mn-oxides is potentially quite mobile in the environment. Although some sorbed AsV can be desorbed from δ-MnO2, there
Present Addresses †
United States Army Corps of Engineers, Engineer Research and Development Center, 3909 Halls Ferry Rd, Vicksburg, MS 39180. ‡ Calera Corporation, 14600 Winchester Blvd., Los Gatos, CA 95030.
’ ACKNOWLEDGMENT The authors thank Gerald Hendricks and Caroline Golt for laboratory assistance. B.L. is grateful for funding provided by a University of Delaware graduate fellowship and the Donald L. and Joy G. Sparks Graduate Fellowship in Soil Science. This research was funded by United States Department of Agriculture Grant 200535107-16105, National Science Foundation Grant EAR-0544246, and Delaware National Science Foundation EPSCoR Grant EPS-0447610. Use of the National Synchrotron Light Source, Brookhaven National Laboratory, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886. ’ REFERENCES (1) Cullen, W. R.; Reimer, K. J. Arsenic speciation in the environment. Chem. Rev. 1989, 89, 713–764. (2) Sadiq, M. Arsenic chemistry in soils: An overview of thermodynamic predictions and field observations. Water, Air, Soil Pollut. 1997, 93, 117–136. (3) Petrick, J. S.; Ayala-Fierro, F.; Cullen, W. R.; Carter, D. E.; Aposthian, H. V. Monomethylarsonous acid (MMA(III)) is more toxic than arsenite in Chang human hepatocytes. Toxicol. Appl. Pharmacol. 2000, 163, 203–207. (4) Driehaus, W.; Seith, R.; Jekel, M. Oxidation of arsenate (III) with manganese oxides in water treatment. Water Res. 1995, 29, 297–305. (5) Ginder-Vogel, M.; Landrot, G.; Fischel, J. S.; Sparks, D. L. Quantification of rapid environmental redox processes with quickscanning x-ray absorption spectroscopy (Q-XAS). Proc. Natl. Acad. Sci. 2009, 106, 16124–16128. (6) Lafferty, B. J.; Ginder-Vogel, M.; Sparks, D. L. Arsenite oxidation by a poorly crystalline manganese-oxide 1. Stirred-flow experiments. Environ. Sci. Technol. 2010, 44, 8460–8466. 9222
dx.doi.org/10.1021/es201281u |Environ. Sci. Technol. 2011, 45, 9218–9223
Environmental Science & Technology (7) Moore, J. N.; Walker, J. R.; Hayes, T. H. Reaction scheme for the oxidation of As (III) to As (V) by birnessite. Clays Clay Miner. 1990, 38, 549–555. (8) Nesbitt, H.; Canning, G.; Bancroft, G. XPS study of reductive dissolution of 7Å-birnessite by H3AsO3, with constraints on reaction mechanism. Geochim. Cosmochim. Acta 1998, 62, 2097–2110. (9) Oscarson, D.; Huang, P.; Defosse, C.; Herbillon, A. Oxidative power of Mn (IV) and Fe (III) oxides with respect to As (III) in terrestrial and aquatic environments. Nature 1981, 291, 50–51. (10) Oscarson, D.; Huang, P.; Liaw, W. Role of manganese in the oxidation of arsenite by freshwater lake sediments. Clays Clay Miner 1981, 29, 219–225. (11) Oscarson, D.; Huang, P.; Hammer, U.; Liaw, W. Oxidation and sorption of arsenite by manganese dioxide as influenced by surface coatings of iron and aluminum oxides and calcium carbonate. Water, Air, Soil Pollut. 1983, 20, 233–244. (12) Parikh, S. J.; Lafferty, B. J.; Meade, T. G.; Sparks, D. L. Evaluating Environmental Influences on AsIII Oxidation Kinetics by a Poorly Crystalline Mn-Oxide. Environ. Sci. Technol. 2010, 44, 3772–3778. (13) Scott, M. J.; Morgan, J. J. Reactions at oxide surfaces. 1. Oxidation of As (III) by synthetic birnessite. Environ. Sci. Technol. 1995, 29, 1898–1905. (14) Tournassat, C.; Charlet, L.; Bosbach, D.; Manceau, A. Arsenic(III) oxidation by birnessite and precipitation of manganese(II) arsenate. Environ. Sci. Technol. 2002, 36, 493–500. (15) Arai, Y.; Elzinga, E. J.; Sparks, D. L. X-ray absorption spectroscopic investigation of arsenite and arsenate adsorption at the aluminum oxide-water interface. J. Colloid Interface Sci. 2001, 235, 80–88. (16) Dixit, S.; Hering, J. G. Comparison of arsenic(V) and arsenic(III) sorption onto iron oxide minerals: Implications for arsenic mobility. Environ. Sci. Technol. 2003, 37, 4182–4189. (17) Raven, K. P.; Jain, A.; Loeppert, R. H. Arsenite and arsenate adsorption on ferrihydrite: kinetics, equilibrium, and adsorption envelopes. Environ. Sci. Technol. 1998, 32, 344–349. (18) Masue, Y.; Loeppert, R. H.; Kramer, T. A. Arsenate and arsenite adsorption and desorption behavior on coprecipitated aluminum: Iron hydroxides. Environ. Sci. Technol. 2007, 41, 837–842. (19) Anderson, M. A.; Ferguson, J. F.; Gavis, J. Arsenate adsorption on amorphous aluminum hydroxide. J. Colloid Interface Sci. 1976, 54, 391–399. (20) Gupta, S. K.; Chen, K. Y. Arsenic removal by adsorption. J. Water Pollut. Control Fed. 1978, 50, 493–506. (21) Hingston, F. J. In Adsorption of Inorganics at Solid-Liquid Interfaces; Anderson, M. A.; Rubin, A. J., Eds.; Ann Arbor Science: Ann Arbor, MI, 1981; pp 5190. (22) Foster, A. L.; Brown, G. E.; Parks, G. A. X-ray absorption fine structure study of As (V) and Se (IV) sorption complexes on hydrous Mn oxides. Geochim. Cosmochim. Acta 2003, 67, 1937–1953. (23) Manning, B. A.; Fendorf, S. E.; Bostick, B.; Suarez, D. L. Arsenic(III) oxidation and arsenic(V) adsorption reactions on synthetic birnessite. Environ. Sci. Technol. 2002, 36, 976–981. (24) Lafferty, B. J.; Ginder-Vogel, M.; Zhu, M.; Livi, K. J. T.; Sparks, D. L. Arsenite oxidation by a poorly crystalline manganese-oxide. 2. Results from X-ray absorption spectroscopy and X-ray diffraction. Environ. Sci. Technol. 2010, 44, 8467–8472. (25) Tani, Y.; Miyata, N.; Ohashi, M.; Ohnuki, T.; Seyama, H.; Iwahori, K.; Soma, M. Interaction of inorganic arsenic with biogenic manganese oxide produced by a Mn-oxidizing fungus, strain KR212. Environ. Sci. Technol. 2004, 38, 6618–6624. (26) Drits, V. A.; Silvester, E.; Gorshkov, A. I.; Manceau, A. Structure of synthetic monoclinic Na-rich birnessite and hexagonal birnessite: I. Results from X-ray diffraction and selected-area electron diffraction. Am. Mineral. 1997, 82, 946–961. (27) Silvester, E.; Manceau, M.; Drits, V. A. Structure of synthetic monoclinic Na-rich birnessite and hexagonal birnessite: II. Results from chemical studies and EXAFS spectroscopy. Am. Mineral. 1997, 82, 962–978. (28) Marcus, M. A.; Manceau, A.; Kersten, M. Mn, Fe, Zn and As speciation in a fast-growing ferrmanganese marine nodule. Geochim. Cosmochim. Acta 2004, 68, 3125–3136.
ARTICLE
(29) Peacock, C. L.; Sherman, D. M. Sorption of Ni by birnessite: Equilibrium controls on Ni in seawater. Chem. Geol. 2007, 238, 94–106. (30) Manceau, A.; Tommaseo, C.; Rihs, S.; Geoffroy, N.; Chateigner, D.; Schlegel, M.; Tisserand, D.; Marcus, M. A.; Tamura, N.; Chen, Z. S. Natural speciation of Mn, Ni, and Zn at the micrometer scale in a clayey paddy soil using X-ray fluorescence, absorption, and diffraction. Geochim. Cosmochim. Acta 2005, 69, 4007–4034. (31) Manceau, A.; Lanson, M.; Geoffroy, N. Natural speciation of Ni, Zn, Ba, and As in ferromanganese coatings on quartz using X-ray fluorescence, absorption, and diffraction. Geochim. Cosmochim. Acta 2007, 71, 95–128. (32) Toner, B.; Manceau, A.; Webb, S. M.; Sposito, G. Zinc sorption to biogenic hexagonal-birnessite particles within a hydrated bacterial biofilm. Geochim. Cosmochim. Acta 2006, 70, 27–43. (33) Pena, J.; Kwon, K. D.; Refson, K.; Bargar, J. R.; Sposito, G. Mechanisms of nickel sorption by a bacteriogenic birnessite. Geochim. Cosmochim. Acta 2010, 74, 3076–3089. (34) Villalobos, M.; Bargar, J.; Sposito, G. Mechanisms of Pb(II) sorption on a biogenic manganese oxide. Environ. Sci. Technol. 2005, 39, 569–576. (35) Parikh, S. J.; Lafferty, B. J.; Sparks, D. L. An ATR-FTIR spectroscopic approach for measuring rapid kinetics at the mineral/ water interface. J. Colloid Interface Sci. 2008, 320, 177–185. (36) Jackson, B. P.; Miller, W. P. Effectiveness of phosphate and hydroxide for desorption of arsenic and selenium species from iron oxides. Soil Sci. Soc. Am. J. 2000, 64, 1616–1622. (37) Lafferty, B. J.; Loeppert, R. H. Methyl arsenic adsorption and desorption behavior on iron oxides. Environ. Sci. Technol. 2005, 39, 2120–2127. (38) Liu, F.; De Cristofaro, A.; Violante, A. Effect of pH, phosphate and oxalate on the adsorption/desorption of arsenate on/from goethite. Soil Sci. 2001, 166, 197–208. (39) Zhu, M.; Paul, K. W.; Kubicki, J. D.; Sparks, D. L. Quantum chemical study of arsenic (III, V) adsorption on Mn-oxides: Implications for arsenic(III) oxidation. Environ. Sci. Technol. 2009, 43, 6655–6661. (40) Drits, V. A.; Lanson, B.; Gaillot, A. C. Birnessite polytype systematics and identification by powder X-ray diffraction. American mineralogist 2007, 92, 771. (41) Murray, J. W. The interaction of metal ions at the manganese dioxide-solution interface. Geochim. Cosmochim. Acta 1975, 39, 505–519. (42) Tonkin, J. W.; Balistrieri, L. S.; Murray, J. W. Modeling sorption of divalent metal cations on hydrous manganese oxide using the diffuse double layer model. Appl. Geochem. 2004, 19, 29–53.
9223
dx.doi.org/10.1021/es201281u |Environ. Sci. Technol. 2011, 45, 9218–9223
ARTICLE pubs.acs.org/est
Multiple Fluorescence Labeling and Two Dimensional FTIR13C NMR Heterospectral Correlation Spectroscopy to Characterize Extracellular Polymeric Substances in Biofilms Produced during Composting Guang-Hui Yu,‡,† Zhu Tang,‡,† Yang-Chun Xu,† and Qi-Rong Shen*,† †
Jiangsu Key Lab for Organic Solid Waste Utilization, College of Resources and Environmental Sciences, Nanjing Agricultural University, Nanjing 210095, People's Republic of China
bS Supporting Information ABSTRACT: Knowledge on the structure and function of extracellular polymeric substances (EPS) in biofilms is essential for understanding biodegradation processes. Herein, a novel method based on multiple fluorescence labeling and two-dimensional (2D) FTIR13C NMR heterospectral correlation spectroscopy was developed to gain insight on the composition, architecture, and function of EPS in biofilms during composting. Compared to other environmental biofilms, biofilms in the thermophilic (>55 °C) and cooling (mature) stage of composting have distinct characteristics. The results of multiple fluorescence labeling demonstrated that biofilms were distributed in clusters during the thermophilic stage (day 14), and dead cells were detected. In the mature stage (day 26), the biofilm formed a continuous layer with a thickness of approximately 20100 μm around the compost, and recolonization of cells at the surface of the compost was easily observed. Through 2D FTIR13C NMR correlation heterospectral spectroscopy, the following trend in the ease of the degradation of organic compounds was observed: heteropolysaccharides > cellulose > amide I in proteins. And proteins and cellulose showed significantly more degradation than heteropolysaccharides. In summary, the combination of multiple fluorescence labeling and 2D correlation spectroscopy is a promising approach for the characterization of EPS in biofilms.
’ INTRODUCTION Biofilms are well-organized communities of microorganisms embedded in a matrix of extracellular polymeric substances (EPS).14 The composition of EPS is complex and is dependent on the bacterial species and the growth conditions. However, the main constituents of EPS are proteins, polysaccharides, cellulose, and lipids.1,35 In many bioprocesses, the growth of biofilm affects the degradation and conversion of organic compounds.6 Unfortunately, a complete biochemical profile of biofilms is difficult to obtain.5 Therefore, the structure and function of biofilms must be elucidated to obtain a deeper understanding of bioprocesses. Currently, one of the best approaches for the investigation of biofilms in situ is the use of fluorescently labeled lectins.5,7 Many investigators have shown that multiple fluorescence labeling and confocal laser scanning microscopy (CLSM) can be combined to obtain a powerful tool for studying the composition, architecture, and function of biofilm constituents at the microscale.3,4,610 Nevertheless, the identification and quantification of specific biofilm constituents is limited by the availability of fluorescently labeled probes.11 CLSM and various methods based on chemical structural analysis, such as Fourier transform infrared (FTIR) and nuclear magnetic resonance (NMR) spectroscopy, can be combined to provide a comprehensive understanding of biofilm development. In previous studies, changes in the structure of biofilm r 2011 American Chemical Society
constituents have been detected via traditional FTIR1214 or NMR11,14,15 spectroscopy. However, the individual spectral features of FTIR or NMR often overlap because of the extreme heterogeneity of biofilm constituents.16 Two-dimensional (2D) correlation spectroscopy17 can be used to resolve the overlapped peaks problems of traditional FTIR or NMR spectroscopy. By distributing spectral intensity trends within a data set collected as a function of the perturbation sequence (e.g., time, temperature, pressure) over a second dimension, one can get 2D correlation spectroscopy. The main advantages of 2D correlation spectra are as follows: (i) simplification of complex spectra consisting of many overlapped peaks, and enhancement of spectral resolution by spreading peaks over the second dimension; (ii) establishment of unambiguous assignments through correlation of bands; (iii) probing the specific sequencing of spectral intensity changes through asynchronous analysis; (iv) so-called heterospectral correlation, i.e., the investigation of correlation among bands in two different types of spectroscopy; and (v) truly universal applicability of the technique, which is not limited to any type of spectroscopy, or even any form of analytical technique (e.g., chromatography, microscopy, etc).18,19 Received: April 30, 2011 Accepted: September 13, 2011 Revised: August 16, 2011 Published: September 13, 2011 9224
dx.doi.org/10.1021/es201483f | Environ. Sci. Technol. 2011, 45, 9224–9231
Environmental Science & Technology Although the second derivative and peak fitting analysis would also be used to solve the peak overlapping problem and enhance spectral resolution,14,20 both of them could not be applied to probe the specific sequencing of spectral intensity changes and investigate the heterospectral correlation. However, information on the distribution and architecture of biofilms cannot be obtained using this method. To our knowledge, 2D correlation spectroscopy has not previously been applied to investigate the function of biofilms in a bioprocess. Composting is a cheap, efficient, and sustainable treatment for solid organic materials.2123 Until now, composting research has mainly focused on optimization of process parameters, degradtion of organic matter, and assessment of maturity.2125 Few studies have been explored in the structure and function of EPS in biofilms of compost, which is essential for understanding biodegradation processes. Thus, the objectives of the present study were to combine multiple fluorescence labeling and 2D correlation spectroscopy to characterize the composition, architecture and function of biofilms. For this purpose, two piles in a full-scale compost facility were constructed and biofilms were allowed to grow. Compared to other environmental biofilms, those present in compost are expected to have distinct characteristics because of the presence of thermophilic and cooling (mature) stages.
’ MATERIALS AND METHODS Composting Process and Biofilm Sample Collection. Two windows with dimensions of 13 1 1.5 m (length height width) were constructed from a mixture of swine manure and wheat straw. The moisture content, pH, water extractable organic carbon (WSC), and water extractable total nitrogen (WSN) content of the feedstock in the two piles were 66.8 ( 0.1%, 8.0 ( 0.1, 13.6 ( 0.3 mg g1, and 2.0 ( 0.1 mg g1, respectively. Composting was performed under aerobic conditions for 26 days, and the piles were turned when a temperature of 60 °C was attained. During the composting process, 2 kg of representative material was collected on days 0, 2, 6, 10, 14, 18, 22, and 26 of composting and was then divided into two subsamples. The detailed description of sampling could be seen in Tang et al.24 Multiple fluorescent labeling and CLSM were conducted on one subsample of compost, and biofilm was extracted from the other subsample to determine its chemical structure and composition of the biofilm. Briefly, the biofilm was separated from the compost by shaking the samples in deionized water (solid to water ratio of 1:10 w/v) for 24 h on a horizontal shaker at room temperature. The separated biofilm from the fresh compost was filtered through a 0.45-μm polytetrafluoroethylene (PTFE) filter in dead-end membrane filtration tests controlling 30 cmHg vacuum before being freeze-dried at 50 °C for 48 h prior to performing a FTIR and NMR spectral analysis. Multiple Fluorescence Labeling and CLSM Observation. The hydrated compost samples were labeled with fluorescent stains possessing different excitation and emission spectra, and the distribution patterns of proteins, α-polysaccharides, cellulose, total cells, and dead cells were simultaneously visualized according to the method of Chen et al.8 In brief, fluoresceinisothiocyanate (FITC), concanavalin A (Con A), calcofluor white (CW), STYO 63, and SYTOX blue identify proteins, α-polysaccharides, cellulose, total cells, and dead cells, respectively. Specifically, SYTO 63 (20 μM, 100 μL) was added to the
ARTICLE
sample, and the resulting mixture was placed on a shaker table for 30 min. Subsequently, 0.1 mol of NaHCO3 buffer (100 μL) was added to maintain a pH of 9. A solution of FITC (10 g/L, 10 μL) was added, and the mixture was stirred for 1 h. Next, a solution of Con A (250 mg/L, 100 μL) was added to the sample for 30 min, followed by CW (300 mg/L, 100 μL) for 30 min. After each stage of the labeling process, the sample was washed twice with phosphate buffered saline (PBS) solution to remove the extra probe. Finally, a solution of SYTOX blue (2.5 μM, 100 μL) was incubated with the sample for 10 min. The labeled samples were embedded for cryosectioning and were then frozen at 20 °C. Subsequently, 30-μm sections were cut on a cryomicrotome (Cyrotome E, Thermo Shandon Limited, U.K.) and were mounted onto gelatin-coated (0.1% gelatin and 0.01% chromium potassium sulfate) microscopic slides for CLSM (Leica TCS SP2 confocal spectral microscope imaging system, Germany) observation. Four slides were made for each sample. In order to ensure the integrity of each slide, it was important to keep bubbles out of the samples when the labeled samples were embedded for cryosectioning. The samples were imaged using a 20 objective. Detailed information about the sample preparation for the CLSM slides is shown as Figure S1 of the Supporting Information, SI. Three-dimensional reconstructions were obtained with Leica confocal software, and movie files generated from the image stack were saved as uncompressed AVI files. Morphological parameters of the CLSM image were determined using Image J software (NIH, Bethesda, MD, U.S.A.). Analysis of FTIR and Solid-State 13C NMR Spectroscopy. Samples were prepared as a mixture of 1 mg of freeze-dried sample and 100 mg of potassium bromide (KBr, IR grade) and then ground and homogenized to reduce light scatter.26 A subsample was then compressed between two clean, polished iron anvils twice in a hydraulic press at 20 000 psi to form a KBr window. The FTIR spectra were obtained by collecting 200 scans with a Nicolet 370 FTIR spectrometer. Solid-state 13CCPMAS-NMR spectroscopy was conducted on a Bruker AV-400, equipped with a 4-mm wide-bore MAS probe. NMR spectra were obtained by applying the following parameters: rotor spin rate of 13 000 Hz, 1 s recycle time, 1 ms contact time, 20 ms acquisition time, and 4000 scans. Samples were packed in 4-mm zirconia rotors with Kel-F caps. The pulse sequence was applied with a 1H ramp to account for nonhomogeneity of the HartmannHahn condition at high spin rotor rates. Structural carbons determined include the following group shifts: 050 ppm (alkyl), 50112 ppm (alcohol, amine, carbohydrate, ether, methoxyl and acetal), 110145 ppm (aromatic), 145163 ppm (phenolic), 163215 ppm (carboxyl and carbonyl). Chemical shifts were calibrated with adamantine. Analysis of 2D Correlation Spectroscopy. The 2D correlation spectra were produced according to the method of Noda and Ozaki.18 In this study, the composting time was applied as an external perturbation, and a set of time-dependent FTIR or NMR spectra was obtained. Let us consider analytical spectrum I(x, t). The variable x is the index variable representing the FTIR or NMR spectra induced by the perturbation variable t. We intentionally use x instead of the general notation used in conventional 2D correlation equations based on spectral index v. Analytical spectrum I(x, t) at m evenly spaced points in t (between Tmin and Tmax) can be represented as follows: Ij ðxÞ ¼ Iðx, tj Þ, j ¼ 1, 2, 3 3 3 , m 9225
ð1Þ
dx.doi.org/10.1021/es201483f |Environ. Sci. Technol. 2011, 45, 9224–9231
Environmental Science & Technology
ARTICLE
Figure 1. Performance of the composting process.
A set of dynamic spectra is given by the following: ~I ðx, tÞ ¼ Iðx, tj Þ ̅ lðxÞ
ð2Þ
where ̅ l(x) denotes the reference spectrum, which is typically the average spectrum and is expressed as ̅ l(x) = 1/m∑m j = 1I(x, tj). The synchronous correlation intensity can be directly calculated from the following dynamic spectra: ϕðx1 , x2 Þ ¼
1 m ~I j ðx1 Þ~I j ðx2 Þ m 1 j¼1
∑
ð3Þ
Asynchronous correlation can be obtained by the following: Lðx1 , x2 Þ ¼
m 1 m ~I j ðx1 Þ Njk~I j ðx2 Þ m 1 j¼1 k¼1
∑
∑
ð4Þ
The term Njk corresponds to the jth column and the kth raw element of the discrete HilbertNoda transformation matrix, which is defined as follows: 8 > 0 < if j ¼ k 1 ð5Þ Njk ¼ > : πðk jÞ otherwise The intensity of a synchronous correlation spectrum (L(x1, x2)) represents simultaneous changes in two spectral intensities measured at x1 and x2 during the interval between Tmin and Tmax. In contrast, an asynchronous correlation spectrum (j(x1, x2)) includes out-of-phase or sequential changes in spectral intensities measured at x1 and x2.
Figure 2. CLSM images of biofilms in pile 1 after 14 (A) and 26 (B) days of cultivation. The images were obtained with a 20 objective lens: (a) proteins (FITC), green; (b) α-polysaccharides (Con A), light blue; (c) cellulose (CW), blue; (d) total cells (SYTO 63), red; (e) dead cells (SYTO blue), violet; (f) merged image of (a)-(e). Bar = 100 μm.
Prior to 2D analysis, the FTIR or NMR spectra were normalized by summing the absorbance from 4000 to 400 cm1 or 0200 ppm, respectively, and multiplying by 1000. Subsequently, normalized FTIR or NMR spectra were analyzed using principal component analysis (PCA) to reduce the level of noise.27 Finally, 2D correlation spectroscopy was produced using 2Dshige software (Kwansei-Gakuin University, Japan).
’ RESULTS Performance of the Compost. The temperature, moisture content, and pH at various stages of the composting process are shown in Figure 1. Both of the piles attained a plateau value of 65 °C at the second day after composting, indicating that the piles rapidly reached the thermophilic phase. The temperature decreased to approximately 50 °C on the 18th day and then remained constant at approximately 60 °C. The moisture content abruptly declined from 70% on the first day to 30% on the 16th day. Evolution of temperature, moisture content, and fluorescence excitationemission matrix contours of dissolved organic matter (Figure S2 of the SI) indicated that the compost was mature after 18 days, which is consistent with the results of our previous investigations.2224 The pH of the piles climbed rapidly from 8.2 on the first day to 8.6 on the eighth day and remained constant over time. During composting, the pH of both 9226
dx.doi.org/10.1021/es201483f |Environ. Sci. Technol. 2011, 45, 9224–9231
Environmental Science & Technology piles ranged from 8.0 to 8.6, evidencing satisfactory microbial activity.25 The aforementioned results suggested that the characteristics of the piles were typical of those observed during composting. Moreover, the findings of the present study were consistent with those obtained from previous investigations.24,28,29 Architecture and Structure of Biofilms Observed by Multiple Fluorescence Labeling Combined with CLSM. Compared to biofilms observed during the cooling (mature) stage, biofilms in the thermophilic stage are expected to be distinct. Therefore, CLSM images of compost samples were obtained during both stages. For brevity, the CLSM images of pile 2 are provided in Figure S3 of the SI. Figure 2 displays the CLSM images of compost samples from pile 1 during the thermophilic (14th d) and mature stages (26th d), respectively. During the thermophilic stage, pig manure and wheat straw were visually apparent, with the former surrounding the latter. Bright images of the composts revealed that pig manure and wheat straw were present in the compost (Figure S4 of the SI). Proteins (FITC) were predominant in pig manure, whereas cellulose (CW) and αpolysaccharides (Con A) formed a continuous layer on the wheat straw. Total cells (SYTO 63) were primarily detected in wheat straw, while dead cells (SYTOX blue) were nearly ubiquitous in both pig manure and wheat straw. Three-dimensional reconstructions of the composts on the 14th day clearly demonstrated that biofilms in the thermophilic stage were highly dispersed throughout the material and were aggregated into clusters located along the outer of the compost (the movie documents generated from the image stack are provided in the Supporting Information). In addition, the wheat straw displayed the characteristics of lignocellulose, which suggested that most of the wheat straw was not completely degraded. During the mature stage, most of the pig manure was degraded, and only wheat straw was observed in the CLSM images. The fluorescence intensity of proteins, cellulose, and α-polysaccharides in the wheat straw during the mature stage was markedly lower that of the thermophilic stage. Moreover, the structure of wheat straw was loose, indicating that most of the wheat straw was degraded. Compared to the thermophilic stage, a quantity of cells in biofilm was observed during the mature stage. The total cell count in the mature stage was markedly greater than that of the thermophilic stage (14th d). Most of the total cells were distributed within the biofilm surrounding the wheat straw. Alternatively, dead cells were primarily observed in the interior of the wheat straw. These results revealed that cell recolonization occurs during the mature stage of composting. When composts were applied to soil, recolonized cells play an important role in the biological control of plant disease.31 In summary, multicolor fluorescence labeling provides information on the detailed architecture and distribution of biofilms in compost. The architecture and distribution patterns of biofilm constituents are closely related to the degradability of biofilms, which can be observed using 2D heterospectral correlation spectroscopy. Function of Biofilms Investigated by 2D Heterospectral Correlation Spectroscopy. The area-normalized FTIR and NMR spectra of the composts over time were noisy. Because the first two principal components accounted for 96% and 98% of the peaks in the FTIR and NMR spectra, respectively, the PCA noise reduction method was applied to reconstruct less noisy spectra. In the reconstructed spectra, the primary bands were maintained, and the level of noise was reduced (data not shown
ARTICLE
Figure 3. Synchronous and asynchronous 2D FTIR correlation maps generated from the 1800900 cm1 region of the spectra and 2D NMR correlation maps of dissolved organic matter in the two piles over time. Red represents positive correlations; a higher color intensity indicates a stronger positive correlation.
for brevity). All of the FTIR and NMR 2D correlation results presented below were generated from reduced-noise spectra. A synchronous spectrum is a symmetric spectrum with respect to a diagonal line. Correlation peaks include autopeak and crosspeak, which appear at both diagonal and off-diagonal positions, respectively. An autopeak represents the overall susceptibility of the corresponding spectral region to change in spectral intensity as an external perturbation is applied to the system. Crosspeaks represent simultaneous or coincidental changes of spectral intensities observed at two different spectral variables. Such a synchronized change, in turn, suggests the possible existence of a coupled or related origin of the spectral intensity variations. An asynchronous spectrum is antisymmetric with respect to the diagonal line, which has no autopeaks and consists exclusively of crosspeaks located at off-diagonal positions. The sign of an asynchronous cross peak can be either 9227
dx.doi.org/10.1021/es201483f |Environ. Sci. Technol. 2011, 45, 9224–9231
Environmental Science & Technology negative or positive. It provides useful information on the sequential order of events observed by the spectroscopic technique along the external variable. The 1800900 cm1 region of the 2D FTIR correlation spectra was evaluated because this region of the spectra contains bands corresponding to amides, carboxylic acids, esters, and carbohydrates.32 Time-dependent one-dimensional FTIR spectra during composting of the two piles are shown as Figure S5 of the SI for brevity. The synchronous maps of the biofilms (Figure 3) from the two piles were similar, and three major autopeaks were detected at 1650, 1380, and 1080 cm1. The greatest change in intensity was observed in the band located at 1650 and 1380 cm1, followed by the peak at 1080 cm1. The band at 1650 cm1 was attributed to amide I in proteinaceous compounds, the band at 1380 cm1 was assigned to the OH bending vibration of cellulose, and the band at 1080 cm1 was attributed to the CO stretching of polysaccharides or polysaccharide-like substances.12,13,32,33 Polysaccharide-like substances are composed of cellulose and hemicellulose. Cellulose is a homopolysaccharide composed of D-glucose units linked to each other via β-1,4-glucosidic bonds; however, hemicellulose is a heteropolysaccharide composed of different sugar units, i.e, mannans, xylans, arabinans, and galactans.34 In this study, polysaccharide-like substances are referred to both homopolysaccharide and heteropolysaccharide, whereas cellulose is assigned to the homopolysaccharide. The above results suggested that proteins and cellulose degraded at a faster rate than polysaccharides during composting. Moreover, three crosspeaks at (1650 and 1380 cm1), (1650 and 1080 cm1), and (1380 and 1080 cm1) were identified. These crosspeaks were positively correlated, suggesting that proteins, cellulose, and polysaccharides varied/degraded concurrently during composting. Compared to the synchronous maps, the asynchronous maps of the biofilms from the two piles displayed distinctive characteristics (Figure 3). In the map of pile 1, three positive crosspeaks were observed at (1690 and 1650 cm1), (1550 and 1380 cm1), and (1420 and 1380 cm1). Moreover, five negative crosspeaks were observed at (1650 and 1550 cm1), (1650 and 1380 cm1), (1650 and 1110 cm1), (1380 and 1110 cm1), and (1380 and 1250 cm1). However, in the map of pile 2, three positive crosspeaks were detected at (1650 and 1280 cm1), (1400 and 1380 cm1), and (1110 and 1080 cm1). In addition, four negative crosspeaks were observed at (1650 and 1610 cm1), (1650 and 1420 cm1), (1650 and 1110 cm1), and (1380 and 1110 cm1) (Figure 3). According to Noda’s rule,18 the following trend in the degradation of peaks was observed during the composting of piles 1 and 2, respectively: 1550, 1420, 1110 cm1 > 1380 cm1 > 1650 cm1 and 1110 cm1 > 1080, 1420 cm1 >1650 cm1 > 1280 cm1 for piles 1 and 2, respectively. Therefore, organic compounds in piles 1 and 2 were degraded in the following sequence: amide II, heteropolysaccharides > cellulose > amide I and heteropolysaccharides > cellulose > amide I for piles 1 and 2, respectively. In conclusion, cells embedded in the biofilms matrix of compost initially utilize easily degradable heteropolysaccharides. Subsequently, the cells degrade cellulose, followed by proteins. The synchronous map of 2D NMR spectra showed that during composting for the two piles, the greatest degradation of organic compounds was O-alkylated (HCOH) carbons (74 ppm), followed by long chain aliphatic carbons (38 ppm), mirroring the results of 2D FTIR spectra that heteropolysaccharides and
ARTICLE
Figure 4. Synchronous maps obtained via 2D heterospectral correlation analysis of the FTIR and 13C NMR spectra of dissoloved organic matter in the composting piles over time. Red represents positive correlations and blue represents negative correlations; a higher color intensity indicates a stronger positive or negative correlation.
cellulose degradated much more than proteins. Moreover, the asynchronous map of 2D NMR spectra further demonstrated that during composting, long chain aliphatic carbons (38 ppm) degraded prior to O-alkylated (HCOH) carbons (74 ppm). The 2D heterospectral correlation maps were used to examine the covariation between bands in the FTIR and 13C NMR spectra. As shown in Figure 4, the FTIR bands at 1650, 1380, 1080 cm1 were positively correlated with the NMR band at 38 ppm. In addition, a negative correlation between the three FTIR bands and the NMR band at 74 ppm was observed. Lastly, the FTIR bands at 1650 and 1380 cm1 were positively correlated with the NMR band at 168 ppm. These results revealed that proteins, cellulose, and heteropolysaccharides in the biofilms consisted of long chain aliphatic carbons rather than O-alkylated (HCOH) carbons. Moreover, proteins and cellulose in the biofilm also contained carboxyl groups, suggesting that O-alkyl carbons were produced during the degradation of long chain aliphatic compounds (i.e., proteins, cellulose, and heteropolysaccharides). These results are supported by those of previous investigations, which showed that the aromatic process performed by microorganisms occurs predominantly in the water-soluble phase.22,24 The fluorescence EEM data (Figure S2 of the SI) also suggested that the extent of aromatic polycondensation conjugated chromophores content and degree of humification increased with an increase in composting time.
’ DISCUSSION Heterogeneity of the Biofilms in Composts. Fluorescence labeling is a valuable tool for assessing the in situ detection of EPS glycoconjugates in undisturbed and fully hydrated and complex environmental biofilms. Although the previous investigations had shown the microscale heterogeneities of bacteria in biofilms,6 few studies are conducted in the compost system. Quantitative 9228
dx.doi.org/10.1021/es201483f |Environ. Sci. Technol. 2011, 45, 9224–9231
Environmental Science & Technology analysis with Image J software clearly demonstrated that the biofilms were approximately 20100 μm thick (Figure S6 of the SI), suggesting that significant heterogeneity in the biofilm of composts was observed. This result will help to modify the modeling of compost degradation. The individual colonies were observed in the biofilm and penetrated the composts for significant depths. Since the performance of the compost system is closely connected with its characteristic, it is reasonable to suppose that a good biofilm may be developed by adjustment of oxygen and moisture content, which will be beneficial for achieving a good performance of composting. Moreover, the distribution of organic compounds observed by a fluorescence labeling approach could also be applied to explain their degradation patterns. The organic compounds of composts, i.e., proteins, α-polysacchrides, and cellulose, were found to have a distinct distribution pattern, determining that the degradation pattern of them may also be different.1 In this study, during the thermophilic stage, α-polysacchrides had the same distribution pattern with cells, whereas cellulose possessed a similar distribution pattern with cells (Figure 2). However, proteins had a distinct distribution with cells. As a consequence, the heteropolysacchrides and cellulose in compost were degraded prior to proteins (Figure 3). Cells observed during the thermophilic stage were associated with cellulose and α-polysaccharides; thus, these cells were attributed to cellulose- and α-polysaccharides-degrading bacteria rather than protein-degrading bacteria. The results of previous investigations also suggest that the majority of cellulose-degraded bacteria are thermophilic.30 Alternatively, the dead cells were attributable to the poor adaption of mesophilic bacteria to the thermophilic environment. The CLSM observations also revealed that cells were evenly distributed throughout the wheat straw, suggesting that degradation did not occur from the outside. It should be noted that a fluorescence labeling approach depends on the specificity of the selected probes or stains and limits by a lack of understanding of EPS composition and structure.15 Zippel and Neu35 showed that the fluorescence labeling approach is not completely free of uncertainties and the selected probes or stains interact with their target through multiple binding sites increases affinity and specificity, owing to the enormous variety of macromolecules in complex natural microbial biofilms. It has been suggested that determination of “dead cells” by SYTOX blue is questionable. Therefore, investigators should be cautious when they want to apply a fluorescence labeling approach. Degradation of Organic Compounds in Biofilms. Characterizing the chemical and biological changes of organic matter can improve the knowledge of organic matter transformations and maturity assessment during the composting process. In previous investigations, the degradation of organic compounds during composting was often studied by conventional methods. For example, Francou et al. 36 demonstrated that at the thermophilic phase, the hemicellulose fraction varied as cellulose, but with lower contents. Our results also support the codegradation of polysaccharides, cellulose, and proteins during composting by the conventional methods (Figure S7 of the SI), which is consistent with that proteins, cellulose, and heteropolysaccharides varied/degraded concurrently during composting by the synchronous map (Figure 3). Nevertherless, it is difficult to give the sequencing of organic compounds degradation by the conventional methods. Therefore, the findings in this study need to
ARTICLE
be further verified by independent methods in the future investigation. Through 2D FTIR—13C NMR correlation heterospectral spectroscopy, our results for the first time demonstrated that the degradation of organic compounds in biofilm followed the order: heteropolysaccharides > cellulose > amide I in proteins. The degradation sequencing is closely related to the nature of organic compounds. As we all know, cellulose is a semicrystalline polymeric material containing both crystalline and amorphous components, whereas hemicellulose is considered as an amorphous component.34,37 Himmel et al. 37 had shown that crystalline cellulose is resistant to degradation because of the strong interchain hydrogen-bonding network, whereas hemicellulose and amorphous cellulose are readily degradable. M€aki-Arvela et al. 34 also demonstrated that the crystalline structure in cellulose is very stable. Therefore, heteropolysaccharides were degraded prior to cellulose during composting. In this study, the thermophilic phase were attained at the second day for the two piles after composting (Figure 1), in which cellulose—rather than proteins—degraded bacteria should be predominant.30 As a consenquence, proteins were degraded at the last sequencing. However, another investigation (unpublished data) showed that the degradation of organic compounds during composting was related to the distribution of them in materials. Therefore, as for the different materials, the degradation sequencing of organic compounds may be different. Environmental Implications. Although 13C NMR is a powerful approach to investigating functional group variations, it suffers from ambiguities in the structural information it provides. For example, the carbonyl band (CdO) of carboxyl, amide, and aliphatic esters in biofilms all resonate at the same position (around 175 ppm). FTIR can be used to provide an additional view of the functional groups, and can help to resolve the carboxyl, amide, and aliphatic ester contributions to biofilms. Therefore, their complementarity in providing information on the distribution of biofilms functional groups could help to construct a more comprehensive picture of the change in biofilms. The novelty of this study is that we applied, for the first time, two-dimensional correlation spectroscopy to characterize the function of biofilms, which provide many advantages when compared with traditional FTIR or NMR spectroscopy. Knowledge on the composition, architecture, and function of biofilms is essential for understanding biodegradation processes. In the present study, a novel method for the characterization of the composition, architecture, and function of biofilms was developed by combining multiple fluorescence labeling and two-dimensional correlation spectroscopy. Multiple fluorescence labeling supplies structural information on the distribution of biofilm constituents in situ, while two-dimensional correlation spectroscopy provides detailed but locally unresolved information on biofilm constituents. The combination of multiple fluorescence labeling and 2D correlation spectroscopy is a promising approach for the characterization of biofilms. Knowledge on the constituents of biofilm contributes to our understanding of the composting process and provides novel information for engineering applications and scientific research.
’ ASSOCIATED CONTENT
bS
Supporting Information. Detailed descriptions of determination of fluorescence EEM, FTIR, and solid-state 13C NMR spectroscopy; one table listing evolution of Pi,n (%) during
9229
dx.doi.org/10.1021/es201483f |Environ. Sci. Technol. 2011, 45, 9224–9231
Environmental Science & Technology composting achieved from fluorescence regional integrity (FRI) analysis; one figure showing fluorescence EEM contours of composts; two figures showing the CLSM images of biofilms for compost from pile 2 and bright images of composts in pile; one figure showing image analysis results; one figure presenting the degradation of organic matter by the conventional method. This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +86-25-8439 5212; fax: +86-21-8439 5212; e-mail:
[email protected]. Author Contributions ‡
G.H.Y. and Z.T. contributed equally to this work
’ ACKNOWLEDGMENT The work was funded by the National Basic Research Program of China (No. 2011CB100503), the National Natural Science Foundation of China (No. 21007027), Specialized Research Fund for the Doctoral Program of Higher Education (No. 20100097120015), China Postdoctoral Science Foundation (No. 20100481156), the Agricultural Ministry of China (No. 2011-G27 and 201103004), and Key Agricultural Project of Jiangsu Province (SX(2010)220). We would also like to thank three anonymous reviewers for their helpful comments and Dr. David Chadwick from North Wyke Research, U.K. for his careful revision on this manuscript. ’ REFERENCES (1) Yu, G. H.; He, P. J.; Shao, L. M.; Zhu, Y. S. Extracellular proteins, polysaccharides and enzymes impact on sludge aerobic digestion after ultrasonic pretreatment. Water Res. 2008, 42, 1925–1934. (2) Wagner, M.; Ivleva, N. P.; Haisch, C.; Niessner, R. Combined use of confocal laser scanning microscopy (CLSM) and Raman microscopy (RM): Investigations on EPS-matrix. Water Res. 2009, 43, 63–76. (3) Adav, S. S.; Lin, J. C. T.; Yang, Z.; Whiteley, C. G.; Lee, D. J.; Peng, X. F.; Zhang, Z. P. Stereological assessment of extracellular polymeric substances, exo-enzymes, and specific bacterial strains in bioaggregates using fluorescence experiments. Biotechnol. Adv. 2010, 28, 255–280. (4) Dominik, D. M.; Nielsen, J. L.; Nielsen, P. H. Extracellular DNA is abundant and important for microcolony strength in mixed microbial biofilms. Environ. Microbiol. 2010, 13, 710–721. (5) Flemming, H. C.; Neu, T. R.; Wozniak, D. J. The EPS matrix: the “house of biofilm cells. J. Bacteriol. 2007, 189, 7945–7947. (6) Stewart, P. S.; Franklin, M. J. Physiological heterogeneity in biofilms. Nat. Rev. Microbiol. 2008, 6, 199–210. (7) Neu, T. R.; Kuhlicke, U.; Lawrence, J. R. Assessment of fluorochromes for two photon laser scanning microscopy of biofilms. Appl. Environ. Microbiol. 2002, 68, 901–909. (8) Chen, M. Y.; Lee, D. J.; Tay, J. H. Distribution of extracellular polymeric substances in aerobic granules. Appl. Microbiol. Biotechnol. 2007, 73, 1463–1469. (9) Yu, G. H.; Juang, Y. C.; Lee, D. J.; He, P. J.; Shao, L. M. Enhanced aerobic granulation with extracellular polymeric substances (EPS)-free pellets. Bioresour. Technol. 2009, 100, 4611–4615. (10) Yu, G. H.; Lee, D. J.; He, P. J.; Shao, L. M.; Lai, J. Y. Fouling layer with fractionated extracellular polymeric substances of activated sludge. Sep. Sci. Technol. 2010, 45, 993–1002. (11) Garny, K.; Neu, T. R.; Horn, H.; Volke, F.; Manz, B. Combined application of 13C NMR spectroscopy and confocal laser scanning
ARTICLE
microscopy-investigation on biofilm structure and physic-chemical properties. Chem. Eng. Sci. 2010, 65, 4691–4700. (12) Hu, Z. H.; Liu, S. Y.; Yue, Z. B.; Yan, L. F.; Yang, M. T.; Yu, H. Q. Microscale analysis of in vitro anaerobic degradation of lignocellulosic wastes by rumen microorganisms. Environ. Sci. Technol. 2008, 42, 276–281. (13) Cao, B.; Shi, L.; Brown, R. N.; Xiong, Y. J.; Fredrickson, J. K.; Romine, M. F.; Marshall, M. J.; Lipton, M. S.; Beyenal, H. Extracellular polymeric substances from Shewanella sp. HRCR-1 biofilms: Characterization by infrared spectroscopy and proteomics. Environ. Microbiol. 2011, 13, 1018–1031. (14) Yuan, S. J.; Sun, M.; Sheng, G. P.; Li, Y.; Li, W. W.; Yao, R. S.; Yu, H. Q. Identification of key constituents and structure of the extracellular polymeric substances excreted by Bacillus megaterium TF10 for their flocculation capacity. Environ. Sci. Technol. 2011, 45, 1152–1157. (15) Seviour, T.; Lambert, L. K.; Pijuan, M.; Yuan, Z. G. Structural determination of a key exopolysaccharide in mixed culture aerobic sludge granules using NMR spectroscopy. Environ. Sci. Technol. 2010, 44, 8964–8970. (16) Plaza, C.; Senesi, N.; Brunetti, G.; Mondelli, D. Evolution of the fulvic acid fractions during co-composting of olive oil mill wastewater sludge and tree cuttings. Bioresour. Technol. 2007, 98, 1964–1971. (17) Noda, I. Generalized two-dimensional correlation method applicable to infrared, Raman, and other types of spectroscopy. Appl. Spectrosc. 1993, 47, 1329–1336. (18) Noda, I., Ozaki, Y., Eds. Two-Dimensional Correlation Spectroscopy- Applications in Vibrational and Optical Spectroscopy; John Wiley & Sons: England, 2004. (19) Noda, I. Two-dimensional correlation spectroscopy-biannual survey 20072009. J. Mol. Struct. 2010, 974, 3–24. (20) Abdulla, H. A.; Minor, E. C.; Dias, R. F.; Hatcher, P. G. Changes in the compound classes of dissolved organic matter along an estuarine transect: A study using FTIR and 13C NMR. Geochim. Cosmochim. Acta 2010, 74, 3815–3838. (21) Gajalakshmi, S.; Abbasi, S. A. Solid waste management by composting: State of the art. Crit. Rev. Environ. Sci. Technol. 2008, 38, 311–400. (22) Yu, G. H.; Luo, Y. H.; Wu, M. J.; Tang, Z.; Liu, D. Y.; Yang, X. M.; Shen, Q. R. PARAFAC modeling of fluorescence excitationemission spectra for rapid assessment of compost maturity. Bioresour. Technol. 2010, 101, 8244–8251. (23) Yu, G. H.; Wu, M. J.; Luo, Y. H.; Yang, X. M.; Ran, W.; Shen, Q. R. Fluorescence excitation-emission spectroscopy with regional integration analysis for assessment of compost maturity. Waste Manage. 2011, 31, 1729–1736. (24) Tang, Z.; Yu, G. H.; Liu, D. Y.; Xu, D. B.; Shen, Q. R. Different analysis techniques for fluorescence excitation-emission matrix spectroscopy to assess compost maturity. Chemosphere 2010, 82, 1202–1208. (25) Bernal, M. P.; Alburquerque, J. A.; Moral, R. Composting of animal manures and chemical criteria for compost maturity assessment: A review. Bioresour. Technol. 2009, 100, 5444–5453. (26) Yu, G. H.; He, P. J.; Shao, L. M.; Lee, D. J. Enzyme activities in activated sludge flocs. Appl. Microbiol. Biotechnol. 2007, 77, 605–612. (27) Jung, Y. M.; Shin, H. S.; Kim, S. B.; Noda, I. New approach to generalized two-dimensional correlation spectroscopy. 1: Combination of principal component analysis and two-dimensional correlation spectroscopy. Appl. Spectrosc. 2002, 56, 1562–1567. (28) Wang, C. M.; Watson, P.; Michel, M. E.; Hoitink, H. A. J. Assessment of the reliability of the solvitaw maturity test for composted manures. Compost Sci. Util. 2003, 11, 125–143. (29) Bustamante, M. A.; Paredes, C.; Marhuenda-Egea, F. C.; PerezEspinosa, A.; Bernal, M. P.; Moral, R. Co-composting of distillery wastes with animal manures: carbon and nitrogen transformations in the evaluation of compost stability. Chemosphere 2008, 72, 551–557. (30) Bergquist, P. L.; Gibbs, M. D.; Morris, D. D.; Te’o, V. S. J.; Saul, D. J.; Morgan, H. W. Molecular diversity of thermophilic cellulolytic and hemicellulolytic bacteria. FEMS Microb. Ecol. 1999, 28, 99–110. 9230
dx.doi.org/10.1021/es201483f |Environ. Sci. Technol. 2011, 45, 9224–9231
Environmental Science & Technology
ARTICLE
(31) Hoitink, H. A. J.; Stone, A. G.; Han, D. Y. Suppression of plant diseases by composts. HortScience 1997, 32, 184–187. (32) Abdulla, H. A. N.; Minor, E. C.; Hatcher, P. G. Using twodimensional correlations of 13C NMR and FTIR to investigate changes in the chemical composition of dissolved organic matter along an estuarine transect. Environ. Sci. Technol. 2010, 44, 8044–8049. (33) Tandy, S.; Healey, J. R.; Nason, M. A.; Williamson, J. C.; Jones, D. L.; Thain, S. C. FT-IR as an alternative method for measuring chemical properties during composting. Bioresour. Technol. 2010, 101, 5431–5436. (34) M€aki-Arvela, P.; Salmi, T.; Holmbom, B.; Willf€or, S.; Murzin, D. Y. Synthesis of sugars by hydrolysis of hemicelluloses—A review. Chem. Rev. 2011, DOI: 10.1021/cr2000042. (35) Zippel, B.; Neu, T. R. Characterization of glycoconjugates of extracellular polymeric substances in tufa-associated biofilms by using fluorescence lectin-binding analysis. Appl. Environ. Microbiol. 2011, 77, 505–516. (36) Francou, C.; Lineres, M.; Derenne, S.; Le Villio-Poitrenaud, M.; Houot, S. Influence of green waste, biowaste and paper-cardboard initial ratios on organic matter transformations during composting. Bioresour. Technol. 2008, 99, 8926–8934. (37) Himmel, M. E.; Ding, S. Y.; Johnson, D. K.; Adney, W. S.; Nimlos, M. R.; Brady, J. W.; Foust, T. D. Biomass recalcitrance: Engineering plants and enzymes for biofuels production. Science 2007, 315, 804–807.
9231
dx.doi.org/10.1021/es201483f |Environ. Sci. Technol. 2011, 45, 9224–9231
ARTICLE pubs.acs.org/est
Environmentally Persistent Free Radicals (EPFRs)-2. Are Free Hydroxyl Radicals Generated in Aqueous Solutions? Lavrent Khachatryan and Barry Dellinger* Louisiana State University, Department of Chemistry, Baton Rouge, Louisiana 70803, United States
bS Supporting Information ABSTRACT: A chemical spin trap, 5,5-dimethyl-1-pyrroline-Noxide (DMPO), in conjunction with electron paramagnetic resonance (EPR) spectroscopy was employed to measure the production of hydroxyl radical ( 3 OH) in aqueous suspensions of 5% Cu(II)O/silica (3.9% Cu) particles containing environmentally persistent free radicals (EPFRs) of 2-monochlorophenol (2MCP). The results indicate: (1) a significant differences in accumulated DMPOOH adducts between EPFR containing particles and non-EPFR control samples, (2) a strong correlation between the concentration of DMPOOH adducts and EPFRs per gram of particles, and (3) a slow, constant growth of DMPOOH concentration over a period of days in solution containing 50 μg/mL EPFRs particles + DMPO (150 mM) + reagent balanced by 200 μL phosphate buffered (pH = 7.4) saline. However, failure to form secondary radicals using standard scavengers, such as ethanol, dimethylsulfoxide, sodium formate, and sodium azide, suggests free hydroxyl radicals may not have been generated in solution. This suggests surface-bound, rather than free, hydroxyl radicals were generated by a surface catalyzed-redox cycle involving both the EPFRs and Cu(II)O. Toxicological studies clearly indicate these bound free radicals promote various types of cardiovascular and pulmonary disease normally attributed to unbound free radicals; however, the exact chemical mechanism deserves further study in light of the implication of formation of bound, rather than free, hydroxyl radicals.
’ INTRODUCTION Stable and relatively nonreactive ‘‘environmentally persistent free radicals (EPFRs)’’ have recently been demonstrated to form in the postflame and cool-zone regions of combustion systems and other thermal processes.13 These resonance-stabilized radicals, including semiquinones, phenoxyls, and cyclopentadienyls can be formed by the thermal decomposition of molecular precursors including catechols, hydroquinones, and phenols. Association with the surfaces of fine particles imparts additional stabilization to these radicals such that they can persist almost indefinitely in the environment.2,4 A mechanism of chemisorption and electron transfer from the molecular adsorbate to a redox-active transition metal or other receptor is shown through experiment, and supported by molecular orbital calculations, to result in EPFR formation.2,3,5 Both oxygen-centered and carboncentered EPFRs are possible, the exact structure of which can significantly affect their environmental and biological activity.6,7 An important question is whether EPFRs associated with transition metal oxide-containing nanoparticles can red-ox cycle to generate reactive oxygen species (ROS) such as hydroxyl radicals ( 3 OH), superoxide anion-radicals (O2 3 ‑), and hydrogen peroxide (H2O2) in aqueous media? Information about formation and identification of these ROS has been reported recently.8 A chemical spin trap 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) in conjunction with electron paramagnetic resonance (EPR) r 2011 American Chemical Society
spectroscopy was employed to measure the production of ROS in aqueous suspension of laboratory surrogates of particleassociated EPFRs derived from 2-monochlorophenol (2-MCP) by chemisorption on Cu(II)O/silica particles. The concentration of hydroxyl radicals was measured at ∼1 μM for a 140 min incubation of EPFR-containing solution.8 Hydroxyl radical is one of the most aggressive intermediate species responsible for critical tissue damage and oxidative stress.912 However, its high reactivity, and short lifetime may result it in not being able to reach some biological targets. Furthermore, its short half-life, makes direct detection of hydroxyl radical virtually impossible; therefore, indirect detection methods such as EPR, coupled with appropriate spin-trapping agents such as 5,5-dimethyl1-pyrroline-N-oxide (DMPO), has been used.1317 We provide here evidence of in vitro generation of hydroxyl radical by EPFRs produced from adsorption of 2-monochlorophenol at 230 °C (2-MCP-230) on copper oxide catalyst supported by silica nanoparticles, 5% Cu(II)O/silica (3.9% Cu).3,18 Our results suggest hydroxyl radicals generated at the interface of the particle and solution remain associated with the surface of the particle. Received: May 18, 2011 Accepted: September 26, 2011 Revised: September 26, 2011 Published: September 26, 2011 9232
dx.doi.org/10.1021/es201702q | Environ. Sci. Technol. 2011, 45, 9232–9239
Environmental Science & Technology
’ EXPERIMENTAL SECTION Materials. High purity 5,5-dimethyl-1-pyrroline-N-oxide (DMPO, 99%+, GLC) was obtained from ENZO Life Sciences International and used without further purification. Desferrioxamine, (DFO, assay 92.5%+, TLS); diethylenetriaminepentacetic acid, (DETAPAC, 99%+); L-Ascorbic acid (99%+); β-Nicotinamide Adenine Dinucleotide phosphate, (NADPH, assay, g 95%); sodium formate (BioUltra, g 99%); sodium azide (BioXtra); dimethyl sulfoxide, (DMSO, 99.7+%); 2-monochlorophenol, (2-MCP, 99+%); copper nitrate hemipentahydrate (99.9+%), 0.01 M phosphate buffered saline, (PBS; NaCl 0.138M: KCl 0.0027M), were all obtained from Sigma-Aldrich. Hydrogen peroxide and Cab-O-Sil, as silica powder, were obtained from Fluka (assay, 30%) and Cabot (EH-5, 99+%), respectively. EPFR Surrogate Synthesis. Five % CuO/silica (3.9% Cu), particles were prepared by impregnation of Cab-O-Sil powder with 0.1 M solution of copper nitrate hemipentahydrate and calcinated at 450 °C for 12 h.19 The sample was then ground and sieved (mesh size 230, 63 μm). Prior to exposure, the particles were heated in situ in air to 450 °C for 1 h to pretreat the surface. They were then exposed to saturated vapors of 2-MCP at 230 °C using a custom-made vacuum exposure chamber for 5 min. Once exposure was completed, the temperature of the system was cooled to 150 °C for 1 h at 102 Torr. The EPR spectra were then acquired at ambient conditions to confirm the existence of EPFRs. EPR Measurements. EPR spectra were recorded using a Bruker EMX-20/2.7 EPR spectrometer (X-band) with dual cavities, modulation and microwave frequencies 100 kHz and 9.516 GHz, respectively. Typical parameters were as follows: sweep width of 100 G, EPR microwave power of 10 mW, modulation amplitude of 0.8 G, time constant of 40.96 ms, and sweep time of 167.77 s. Values of the g-tensor were calculated using Bruker’s WIN-EPR SimFonia 2.3 program, which allows control of the Bruker EPR spectrometer, data-acquisition, automation routines, tuning, and calibration programs on a Windows-based PC.20 The exact g-values for key spectra were determined by comparison with 2,2-diphenyl-1-picrylhydrazyl (DPPH) standard. ROS Generation Studies. Both control and sample solution suspensions, containing particles without EPFRs (CuO/Silica) and with EPFRs (EPFR/CuO/Silica), respectively, were prepared in similar manners. One mg/mL suspensions of the control, CuO/silica, and sample, EPFR/CuO/silica were prepared in water and saturated with air by bubbling for 5 min. Prior to adding DMPO, the surrogate solutions were sonicated 5 min (Fisher Scientific, FS-20) at 40 W. 0.01 M PBS was used to maintain the pH at 7.4 and balance the final volume at 200 μL. The order of introduction of final solution components to PBS was as follows: particle suspension (10 μL from solution of 1 mg/mL), DMPO (10 μL from a freshly prepared solution of 3 M), reagents (chelators, ascorbic acid, NADPH), and PBS to balance at 200 μL. The final composition of the suspension in most experiments was particles (50 μg/mL), DMPO (150 mM), reagent (200 μL). The solutions were stored in the dark and shaken in touch mode for 30 s using a Vortex Genie 2 (Scientific Industries). Twenty μL of solution was transferred to an EPR capillary tube (i.d. ∼1 mm, o.d. 1.55 mm) and sealed at one end with sealant (Fisherbrand). The capillary was inserted in a 4 mm EPR tube and placed into the EPR resonator.21 The intensities of the EPR spectra of DMPOOH adducts were reported in arbitrary units, DI/N, (double integrated (DI) intensity of the EPR spectrum normalized (N) to account for the conversion time, receiver gain,
ARTICLE
Scheme 1. Hypothesized Red-Ox Cycle of EPFRs (2-Hydroxyphenoxyl) Originating from 2-MCP Molecule Adsorbed on Cu(II) Domain in Biological System8
number of data points and sweep width.20 Each experiment was performed at least twice, and the reported EPR intensities are an average of all spectra obtained for each experiment. Since the interaction chemistry of chelators with the surface of the model particles is unclear, we abstained from the use of chelators such as DFO, DETAPAC, which minimize the iron content in solution. Chelators have been reported to drastically change the reactivity of particles by affecting the redox potential of metals.22 Adsorbed EPFRs may also undergo enhanced extraction in the presence of metal chelators, based on the metal-chelate complex stability. The oxidizing species formed by the Fentontype reactions can also depend on the nature of the iron chelator.2325 Nonuse of chelators in this work was also based on the fact that the buffer, prepared in deionized water and treated with Chelex 100 ion-exchange resin (Bio-Rad Laboratories, Hercules, LA) to remove trace heavy metal contaminants,26 did not significantly impact the spin trapping results.
’ RESULTS AND DISCUSSION Our model of formation of EPFRs, reduced metals, and ROS via chemisorption of molecular precursors on metal centers is summarized in Scheme 1.8 The EPFRs formed from 2-MCP adsorbed on CuO/Silica are o-semiquinone (2-hydroxyphenoxyl) and 2-chlorophenoxyl (latter not shown for clarity).3 These EPFRs may red-ox cycle to generate ROS as depicted.8 The average concentration of EPFRs on Cu(II)O/silica was ∼1017 spins/g and exhibited a singlet, structureless EPR spectrum (g = 2.0042, ΔHp-p = 6.5 G). Undosed Cu(II)O/silica particles, which did not contain EPFRs, were used as controls. To establish the optimal conditions for generating DMPOOH adducts, a series of experiments were initially performed in different solvent solutions (water, dimethylsolfoxide (DMSO), and ethanol (EtOH)) in which the concentrations of spin traps and reductants were varied. Spin Trapping by DMPO. The appearance of the EPR spectrum of DMPOOH adducts is depicted in Figure 1A. The time evaluation of the DMPOOH adducts EPR intensity at different DMPO concentrations is represented in Figure 1B. Further incubation resulted in increasing intensity of the DMPO OH adduct spectrum, and a prominent signal was detected at 180 min. A noisy spectrum was obtained in sample solutions of EPFRs (50 μg/mL) and DMPO (150 mM) in PBS at 2 min of incubation. The 4 lines marked with red asterisks at 2 min incubation correspond to the DMPOOH adduct (hfsc αN = αH = 14.95 G, 9233
dx.doi.org/10.1021/es201702q |Environ. Sci. Technol. 2011, 45, 9232–9239
Environmental Science & Technology
ARTICLE
Figure 1. (A) Evaluation of EPR spectra of DMPOOH adducts as a function of incubation time for a solution of EPFRs (50 (μg/mL)), DMPO (150 mM) in PBS. (B) Time dependence of EPR spectral intensity of DMPOOH adducts as a function of DPMO concentration in solution.
literature data αN = αH = 14.90 G17,27). The other 6 lines are characteristic of a carbon centered species reported in literature and identified as aminoxyl radical formed from the hydroxylamine impurity in DMPO, or formed immediately as high purity DMPO is transferred in an oxygen reach environment.28,29 The EPR intensity of this impurity decreases slowly with time and does not interfere with the measurements of DMPOOH adduct generation. To provide sufficient EPR intensity for convenient analysis, but avoid potential secondary reactions reactions (Decomposition by light, oxidation by dissolved oxygen, reducing/oxidation, dimerization etc.)16 which may occur at high DMPO concentrations, a final DMPO solution concentration of 150 mM was used in all further studies. The results of nonaeration/aeration on both non-EPFR (control) and EPFR particles solutions are presented in Figure 2A,B. The difference in the DMPOOH adduct spectral intensity for the sample and control solutions (calculated from the second line at low magnetic field of their respective 4 line spectra) increased with time, most notably at incubation times >150 min. This difference was larger and occurred at earlier times for the aerated solution. The difference in the nonaerated solution was only ∼50% at 1055 min for the nonaerated solution but was ∼100% for the aerated solution at only 220 min. These results confirm involvement of O2 in the redox cycle generating 3 OH, and all further experiments were performed with aerated samples. Dependence of DMPOOH Adduct Generation on EPFR Concentration. Figure 3A depicts three sets of experimental data for the sample solutions of EPFR-containing particles (50 μg/mL) and DMPO (150 mM) in PBS (total 200 μL) with different initial concentrations of EPFRs (spins/gram). The initial rate of hydroxyl radical generation increased proportionally to the EPFR concentration when it was doubled from 5.56 1016 spins/gram and 1.29 1017 spins/gram. When the EPFR concentration was again approximately doubled to 2.32 1017 spins/gram, there was no increase in the DMPOOH adduct concentration and the concentration actually decreased at longer incubation times. This is probably due to radicalradical recombination at high EPFR concentrations. The dependence of DMPOOH adduct concentration on particle concentration in solution (mg of particle/mL of solution) is depicted in Figure 3B. The intensity of DMPOOH adduct signal exhibited a maximum at a particle concentration of 0.25 mg/mL for concentrations ranging from 0.05 to 1.0 mg/mL. For concentrations >0.25 mg/mL, the DMPOOH concentration decreased, again, probably due to radicalradical annihilation
Figure 2. (A) The time evolution of the EPR signal intensity of DMPOOH for the nonaerated control (red) and sample (black) solutions. (B) The time evaluation of the EPR signal intensity of DMPOOH for the aerated control (red) and sample (black) solutions. The green line represents the time evaluation of the EPR signal intensity of DMPOOH for pure DMPO (150 mM) in 200 μL PBS solution and indicates no statistical increase in DMPOOH concentration.
reactions at higher concentrations. Consequently, very dilute suspensions of 50 μg/mL were used in all subsequent experiments. Dependence of DMPOOH Adduct Generation on EPFR Aging. To determine the lifetimes of EPFRs in aqueous solution and their viability for hydroxyl radical generation, a time evaluation 9234
dx.doi.org/10.1021/es201702q |Environ. Sci. Technol. 2011, 45, 9232–9239
Environmental Science & Technology
ARTICLE
Figure 3. EPR spectral intensity of DMPOOH as a function of EPFR concentration (spins/g) and incubation time. (A) The initial EPFR concentration on the particles was parametrically varied (radicals/g of particle). (B) The initial particle concentration in solution was varied (mg of particle/mL).
Figure 4. EPR spectral intensity of DMPOOH adducts as a function of incubation time for solution allowed to age over 1 to 7 days while exposed to air.
of DMPOOH spectral intensity was performed for particles allowed to age in room air over a period of a week (cf. Figure 4). A stock solution of EPFR-containing particles (50 μg/mL) and DMPO (150 mM) in PBS (total 200 μL) was prepared, and fresh sample solutions were subjected to the spin trapping procedure each day. A marked increase in the DMPOOH adduct intensity was detected on the third day. Concomitantly, the concentration of the EPFRs on the catalyst surface was measured. For the latter, the solutions were subjected to agitation, decanting, and vacuum drying before the residue solid powder was subjected EPR examination. On the first day, the sample exhibited a singlet EPR spectrum (g = 2.0042. ΔHp-p = 6.5 G) which matched well with the initial spectra before aging. This signal remained strong through the third day, before decaying slowly to barely detectable quantities by the seventh day. This coincided with the maximum in hydroxyl radical generation observed on day three. Thus, the observed hydroxyl radical formation is thought to be mediated, rather than catalyzed, and EPFRs are consumed in the mediated process. However, as proposed in Scheme 1, a biological reducing equivalent is necessary to complete the truly catalytic hydroxyl radical generation cycle. Dependence of DMPOOH Adduct Generation on Biological Reducing Equivalents. The generation of hydroxyl radical in biological systems is usually enhanced by the presence of H-donors, e.g., NADPH and ascorbates etc.30,31 Up to 1 mM of NADPH or ascorbic acid was added to the solution of EPFR’s
(50 ug/mL) and DMPO (150 mM) in PBS (cf. Supporting Information). The NADPH induced a small effect; while ascorbic acid significantly increased DMPOOH adduct formation at its lower concentrations (100 μM). However, at higher ascorbic acid concentrations (500 μM), it acted as an effective antioxidant by formation of a characteristic doublet line of ascorbyl radicals. To avoid secondary reactions involving these reductants, their use was minimized in subsequent experiments. Radical Scavengers. Existence of free hydroxyl radicals in biological red-ox systems is usually confirmed using solution, radical scavengers and formation of secondary radicals.32,33 The kinetic competition between DMPO and hydroxyl radical scavengers (ethanol, DMSO, sodium formate, and sodium azide) is used to establish or rule out the presence of free hydroxyl radical. Otherwise, the hydroxyl radical may be bound, or associated with a surface or other molecular species. The effect of scavengers on radical formation were performed in the sample solutions used previously. Inhibition of DMPOOH adduct formation was observed with addition of 10% (v/v) EtOH at 50 min incubation time (cf. Figure 5A). Thirty minutes later, the intensity was decreased by 20% (DI/N = 20) and then slightly increased. The effect of different ethanol concentrations on inhibition of DMPOOH adduct formation is presented in Figure 5B. Hydroxyl radicals in the presence of DMPO and ethanol can undergo the competitive reactions: 3 OH
þ DMPO f DMPO OH
ðreaction1Þ
3 OH
þ CH3 CH2 OH f CH3 3 CHOH þ H2 O ðreaction2Þ
and the new radical CH3 3 CHOH will be trapped by DMPO via: CH3 3 CHOH þ DMPO f DMPO CHðCH3 ÞOH ðreaction3Þ The resulting DMPO-1-hydroxyethyl adduct exhibits a characteristic 6 line spectrum, vide infra.34,35 If this reaction occurs, then the DMPOOH signal should decrease as the DMPO CH(CH3)OH signal increased. This was not observed, suggesting the reaction of ethanol and hydroxyl radical did not occur. DMSO is also a good scavenger of free hydroxyl radicals via reactions similar to Rxns. reaction 1 and reaction 2 for ethanol.34 9235
dx.doi.org/10.1021/es201702q |Environ. Sci. Technol. 2011, 45, 9232–9239
Environmental Science & Technology
ARTICLE
Figure 5. (A) The inhibitory effect of EtOH 10% (v/v) on DMPOOH EPR signal intensity (DI/N) added after 50 min incubation for a sample solution of EPFRs (50 ug/mL) + DMPO (150 mM) + PBS. (B) Time dependence of DMPOOH adduct intensity as a function of ethanol concentration (percent, v/v).
Figure 6. Time evaluation of DMPOOH adduct intensity (DI/N) as a function of DMSO concentration (percent, v/v) in water solution (EPFRs (50 μg/mL) and DMPO (150 mM) in PBS).
At low concentration (2%, v/v), DMSO was a promoter of DMPOOH formation, while at >10% concentration, it completely inhibited formation, (cf. Figure 6). As in case of ethanol, no concomitant formation of the DMPOCH3 adduct, 6-line EPR spectrum was detected. In an attempt to generate observable free hydroxyl radical, 20 mM of hydrogen peroxide was added to both the EtOH- and DMSO-containing solutions.36 Generation of hydroxyl radical by H2O2 has been reported in the presence of Cu(II) ions and CuO micrometer-sized catalyst particles.36,37 The Cu(I), present in our EPFR-containing particles (Scheme 1), should also react with the peroxide via Fenton-type reactions to produce hydroxyl radical. The results are summarized in Figure 7 for the EPR spectra of DMPOOH adduct (blue line). The EPR spectra of DMPOCH(CH3)OH adduct (red line), marked by asterisks, was derived from the solution of EPFRs, DMPO, and ethanol (30%,v/v) + H2O2 (20 mM). The observed EPR spectral parameters (hfsc: αN = 15.1 G, αHβ = 23.1G) agreed well with the literature values (αN = 15.8 G, αHβ = 22.8G34). The EPR spectra of DMPOCH3 adduct (black line), marked by asterisks, was generated from a solution of EPFRs, DMPO and DMSO (10%, v/v) + H2O2 (20 mM). The spectral parameters (hfsc of αN = 16.8 G, αHβ = 23.8G) also agreed well with the literature (αN = 16.4 G, αHβ = 23.4G34). These results suggest free hydroxyl radicals generated by the EPFR particle systems should have been detected in the previously tested solutions if they were present.
Figure 7. EPR spectra of spin adducts generated in different solutions. (1) DMPOOH in solution of EPFRs (50 μg/mL) and DMPO(150 mM) in PBS + H2O2 (20 mM). (2) Mixture of DMPOOH and DMPOCH(CH3)OH in solution of EPFRs (50 μg/mL), DMPO (150 mM), and EtOH (30% v/) in PBS + H2O2 (20 mM). (3) Mixture of DMPOOH and DMPOCH3 in a solution of EPFRs (50 μg/mL), DMPO(150 mM), and DMSO (10% v/v) in PBS + H2O2 (20 mM). The background DMPOOH EPR signal after 5 min without addition of H2O2 was extremely weak (data not shown).
Formate and sodium azide were also used in an attempt to scavenge free hydroxyl radical via the reactions: DMPO
OH þ HCOO f H2 O þ 3 COO sf DMPO COO
ðreaction4Þ 3 OH
DMPO
þ N3 f OH þ 3 N 3 sf DMPO N3 ðreaction5Þ
Neither the characteristic 6-line spectrum of DMPOCOO nor the 12-line spectrum of DMPO-N3 was detected.35 These lines were generated only following addition of H2O2 (20 mM in solution) as a source of free hydroxyl radicals (data not shown). Free versus Bound Hydroxyl Radical. The DMPO spin trapping results indicate hydroxyl radical is being generated. However, the failure of scavenging hydroxyl radicals to form secondary radicals in solution, suggests they are not truly free hydroxyl radicals. Several observations support this contention. 1 . The control solutions, containing Cu(II)O/siica, generate free hydroxyl radical, as evidenced by the scavenger results 9236
dx.doi.org/10.1021/es201702q |Environ. Sci. Technol. 2011, 45, 9232–9239
Environmental Science & Technology
ARTICLE
Figure 8. (A) Time evaluation of DMPOOH EPR spectral intensity (DI/N) as a function of hydrogen peroxide concentration in a sample solution of EPFRs (50 μg/mL) and DMPO (150 mM) in PBS. (B) The minimum amount of hydrogen peroxide (0.25 mM) at which hydroxyl radical is scavenged by 1.7 M ethanol to result in a detectable 6-line spectrum of DMPOCH(CH3)OH adduct (marked with asterisks).
(not shown), only when hydrogen peroxide is added. The detection of DMPOOH adducts without the addition of hydrogen peroxide is likely due to the Cu(II) or Fe(III) impurity, catalyzed nucleophilic addition of water to DMPO.16,17,38 This has been proposed in the literature, but the issue is by no means settled.3942 The nonradical nucleophilic reaction of water has been proposed to be a significant pathway to the formation of DMPOOH radical adducts, even during a Fenton reaction,40,41 i.e., 8090% of the total DMPOOH in 17O-enriched water was due to irondependent nucleophilic addition of water.41 However, the same authors also discuss a water-independent mechanism of DMPOOH formation,41 and how Fe or Cu ioninduced nucleophilic addition of water to DMPO may be significantly suppressed in experiments performed in most common buffers.40 2 . The observed DMPOOH adducts may form due to a secondary mechanism not involving hydroxyl radical trapping.43 DMPOOH adduct formation has been proposed to be the result of conversion of DMPO superoxide adduct (DMPO-OOH).34,4345 However, only 3% of DMPO OOH adduct has been reported to be converted into DMPOOH,34 and the concentration of DMPOOH adduct may be affected only when there is a high concentration of superoxide radicals.46 Other researchers have reported that this conversion does not occur.47 Another secondary mechanism might be oxidation of DMPO through it is cation radical, by addition of water (and elimination of a proton) with ultimate formation of DMPOOH adduct.48 This, as well as all other possible conversions of DMPO in aqueous solution (oxidation by dissolved oxygen, dimerization, reduction/oxidation, etc.), may occur and every specific case must be considered. The differences we have observed in accumulation of DMPO OH adducts between the control and sample solutions (cf. Figure 2) and the direct dependency of the intensity of DMPOOH adducts on EPFR concentration per gram of particle (cf. Figure 3A) are explained by of the activity of EPFRs. Other potential explanations appear to apply to both the samples and controls. 3 Scheme 1 depicts how hydroxyl radicals can be generated by a surface-catalyzed, redox cycle. Our results indicate hydroxyl radical is produced and forms adducts with DMPO, but
the concentration of free hydroxyl radical in solution is too low to be scavenged to form secondary radicals or the rate of reaction with the secondary radicals with DMPO is too slow to be easily detectable. The rate coefficient for reaction of hydroxyl radicals with organics has been reported to be (2.15.7) 109 M1.s1 for reaction 1 and 1.8 109 M1. s1 for reaction 2,49 while the reaction coefficient for secondary 3 CH(CH3)OH radicals, formed by the ethanol scavenger has been reported to be 2 orders of magnitude slower, viz. 4.1 107 M1.s1 for reaction 3.50 Using these rate constants it can be easily established that reaction 3 may compete with reaction 1 at a ratio of concentration of secondary to hydroxyl radicals of ∼100. Thus, it appears the secondary radicals cannot compete with hydroxyl radicals to be trapped by DMPO unless the concentration of 3 CH(CH3)OH is much greater than hydroxyl radical. To determine the minimum concentration of hydroxyl radicals that could be effectively detected using DMPO, additional experiments using hydrogen peroxide to generate hydroxyl radical were performed. The ability of sample solutions to generate DMPO OH adducts at hydrogen peroxide concentrations of 1.5, 0.3, 0.06, and 0.006 mM was determined. The minimum concentration of hydrogen peroxide necessary to generate more DMPOOH adduct than the EPFR-containing particles during the first 100 min is between 0.006 and 0.06 mM (cf. Figure 8A, blue, pink, and black lines). The question is then at what minimum concentration of hydrogen peroxide, assuming the hydroxyl radicals are free in solution, secondary 3 CH(CH3)OH might be generated in detectable quantities. Thus, the previous experiments were repeated in the presence of a large excess (1.7 M) of ethanol scavenger. The characteristic 6 lines of the DMPOCH(CH3)OH adduct 34,35 were barely detectable and were only clearly identifiable when the hydrogen peroxide concentration was increased to 0.25 mM and above (cf. Figure 8B). At < 0.25 mM hydrogen peroxide, the scavenging efficiency of ethanol is not high enough to form detectable secondary radicals. Thus, a more effective spin trap for secondary radicals is needed with a scavenging rate coefficient >4.1 107 M1 3 s1,50 or a more sensitive EPR method, which can detect less than 106 M in solutions,51,52 must be employed. For instance, the detection limit of particle-generated hydroxyl 9237
dx.doi.org/10.1021/es201702q |Environ. Sci. Technol. 2011, 45, 9232–9239
Environmental Science & Technology radicals can be quantified at a concentration of only 50 nM by a fluorescence-based technique using 30 -(p-aminophenyl) fluorescein (APF).53 However, this technique may not have sufficient specificity.44,45,53,54 New EPR spin trapping techniques employing heteroaryl nitrones may be useful since they have been reported to be highly soluble in water, are less sensitive to nucleophilic attack, have long half-lives of the spin adducts, and exhibit high selectivity.55 On the basis of all of these experiments, we believe hydroxyl radicals are generated by the surface-mediated cycle in Scheme 1, with the resulting hydroxyl radicals remaining primarily on the surface such that they cannot be readily scavenged to form secondary organic radicals. This hypothesis is not without experimental or theoretical precedence.5663 For example, oxidizing metal sites have been proposed to form and trap hydroxyl radicals with reactivities similar (but distinguishable) to those of free hydroxyl radical.64,65 The surface bound hydroxyl radicals have even been suggested of being capable of oxidizing of substrates which are oxidized by free hydroxyl radical.56 In our theory, the combination of the surface-bound hydroxyl radical and the reduced metal in the immediate vicinity are responsible for this enhanced activity of the particles.
’ ASSOCIATED CONTENT
bS Supporting Information. Details of “Dependence of DMPOOH adduct generation on biological reducing equivalents”, NADPH and Ascorbic acid, Figures S1 and S2. This material is available free of charge via the Internet at http:// pubs.acs.org. ’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT The authors gratefully acknowledge the partial support of this research from NIEHS as part of the LSU Superfund Center under Superfund Research and Training Program Grant P42ES13648. ’ REFERENCES (1) Dellinger, B.; Pryor, W. A.; Ceuto, R.; Squadrito, G. L.; Hedge, V.; Deutsch, W. A. Role of free radicals in the toxicity of airborne fine particulate matter. Chem. Res. Toxicol. 2001, 14, 1371–1377. (2) Dellinger, B.; Lomnicki, S.; Khachatryan, L.; Maskos, Z.; Hall, R.; Adounkpe, J.; McFerrin., C.; Truong, H. Formation and stabilization of persistent free radicals. Proc. Combust. Inst. 2007, 31, 521–528. (3) Lomnicki, S.; Truong, H.; Vejerano, E.; Dellinger, B. Copper oxide-based model of persistent free radical formation on combustionderived particulate matter. Environ. Sci. Technol. 2008, 42 (13), 4982– 4988. (4) Valavanidis, A.; Iopoulos, N.; Gotsis, G.; Fiotakis, K. Persistent free radicals, heavy metals and PAHs generated in particulate soot emissions and residue ash from controlled combustion of common types of plastic. J. Hazard. Mater. 2008, 156 (13), 277–284. (5) McFerrin, C. A.; Hall, R. W.; Dellinger, B. Ab Initio study of the formation and degradation reactions of chlorinated phenols. J. Mol. Struct. 2009, 902 (13), 5–14. (6) Dellinger, B.; Pryor, W. A.; Cueto, R.; Squadrito, G. L.; Deutsch, W. A. The role of combustion-generated radicals in the toxicity of PM2.5. Proc. Combust. Inst. 2000, 28, 2675–2681.
ARTICLE
(7) Cormier, S. A.; Lomnicki, S.; Backes, W.; Dellinger, B. Origin and health impacts of emissions of toxic by-products and fine particles from combustion and thermal treatment of hazardous wastes and materials. Environ. Health Perspect. 2006, 114, 810–817. (8) Khachatryan, L.; Vejerano, E.; Lomnicki, S.; Dellinger, B. Environmentally Persistent Free Radicals (EPFRs). 1. Generation of Reactive Oxygen Species (ROS) in aqueous solutions. Environ. Sci. Technol. 2011, 45 (19), 8559–8566. (9) Halliwell, B.; Gutteridge, J. M. C. Free Radicals and Metal Ions in Human Disease in Methods in Enzymology. Methods Enzymol. 1990, 186, 1–85. (10) Fridovich, I. Superoxide radical—An endogenous toxicant. Ann. Rev. Pharmacol. Toxicol. 1983, 23, 239–257. (11) Halliwell, B. Oxidative Stress Dis. 2001, 7, 1–16. (12) Zweier, J. L.; Talukder, M. A. H. The role of oxidants and free radicals in reperfusion injury. Cardiovasc. Res. 2006, 70 (2), 181–190. (13) Forshult, S.; Lagercra, C Use of nitroso compounds as scavengers for study of short-lived free radicals in organic reactions. Acta Chem. Scand. 1969, 23 (2), 522–523. (14) Janzen, E. G.; Blackbur, Bj Detection and identification of short-lived free radicals by an electron spin resonance trapping technique. J. Am. Chem. Soc. 1968, 90 (21), 5909–5910. (15) Harbour, J. R.; Chow, V.; Bolton, J. R. Electron-spin resonance study of spin adducts of OH and HO2 radicals with nitrones in ultraviolet photolysis of aqueous hydrogen-peroxide solutions. Can. J. Chem. 1974, 52 (20), 3549–3553. (16) Finkelstein, E.; Rosen, G. M.; Rauckman, E. J. Spin trapping of superoxide and hydroxyl radical—Practical aspects. Arch. Biochem. Biophys. 1980, 200 (1), 1–16. (17) Makino, K.; Hagiwara, T.; Murakami, A. Fundamental-aspects of spin trapping with dmpo. Radiat. Phys. Chem. 1991, 37 (56), 657–665. (18) Truong, H., Copper (II) oxide mediated formation and stabilization of combustion generated persistent free radicals. Dissertation for Ph.D. degree 2007, LSU, LA. (19) Truong, H.; Lomnicki, S.; Dellinger, B. Potential for misidentification of environmentally persistent free radicals as molecular pollutants in particulate matter. Environ. Sci. Technol. 2010, 44 (6), 1933–1939. (20) Eaton, G. R.; Eaton, S. S.; Barr, D. P.; Weber, R. T. Quantitative EPR; Springer/Wien: NewYork, Germany,2010; p 185. (21) Nakagawa, K. Is quartz flat cell useful for the detection of superoxide radicals? J. Act. Oxyg. Free Rad. 1994, 5, 81–85. (22) Fubini, B.; Mollo, L.; Giamello, E. Free radical generation at the solid/liquid interface in iron containing minerals. Free Rad. Res. 1995, 23 (6), 593–614. (23) Tomita, M.; Okuyama, T.; Watanabe, S.; Watanabe, H. Quantitation of the hydroxyl radical adducts of salicylic-acid by micellar electrokinetic capillary chromatography-oxidizing species formed by a Fenton reaction. Arch. Toxicol. 1994, 68 (7), 428–433. (24) Yamazaki, I.; Piette, L. H. EPR spin-trapping study on the oxidizing species formed in the reaction of the ferrous ion with hydrogen-peroxide. J. Am. Chem. Soc. 1991, 113 (20), 7588–7593. (25) Zhu, B. Z.; Har-El, R.; Kitrossky, N.; Chevion, M. New modes of action of desferrioxamine: scavenging of semiquinone radical and stimulation of hydrolysis of tetrachlorohydroquinone. Free Radical Biol. Med. 1998, 24 (5), 880–880. (26) Finkelstein, E.; Rosen, G. M.; Rauckman, E. J. Spin trapping— Kinetics of the reaction of superoxide and hydroxyl radicals with nitrones. J. Am. Chem. Soc. 1980, 102 (15), 4994–4999. (27) Harbour, J. R.; Bolton, J. R. Involvement of hydroxyl radical in destructive photo-oxidation of chlorophylls in vivo and in vitro. Photochem. Photobiol. 1978, 28 (2), 231–234. (28) Buettner, G. R.; Oberley, L. W. Considerations in spin trapping of superoxide and hydroxyl radical in aqueous systems using 5,5-dimethyl1-pyrroline-1-oxide. Biochem. Biophys. Res. Commun. 1978, 83 (1), 69–74. (29) Makino, K.; Imaishi, H.; Morinishi, S.; Hagiwara, T.; Takeuchi, T.; Murakami, A.; Nishi, M. An artifact in the esr-spectrum obtained by spin trapping with dmpo. Free Radical Res. Commun. 1989, 6 (1), 19–28. 9238
dx.doi.org/10.1021/es201702q |Environ. Sci. Technol. 2011, 45, 9232–9239
Environmental Science & Technology (30) Alaghmand, M.; Blough, N. V. Source-dependent variation in hydroxyl radical production by airborne particulate matter. Environ. Sci. Technol. 2007, 41 (7), 2364–2370. (31) Briede, J. J.; De Kok, T. M. C. M.; Hogervorst, J. G. F.; Moonen, E. J. C.; Den Camp, C. L. B. O.; Kleinjans, J. C. S. Development and application of an electron spin resonance spectrometry method for the determination of oxygen free radical formation by particulate matter. Environ. Sci. Technol. 2005, 39 (21), 8420–8426. (32) Yim, M. B.; Chock, P. B.; Stadtman, E. R. Copper, zinc superoxide-dismutase catalyzes hydroxyl radical production from hydrogen-peroxide. Proc. Natl. Acad. Sci. U.S.A. 1990, 87 (13), 5006–5010. (33) Buettner, G. R.; Mason, R. P.Critical Reviewes of Oxidative Stress and Aging: Advances in Basic Science, Diagnostics and Intervention. Chapter 2; Cutler, R. G., Rodrigues, H., Eds.; 2003, 2738 (34) Finkelstein, E.; Rosen, G. M.; Rauckman, E. J. Production of Hydroxyl Radical by Decomposition of Superoxide Spin-Trapped Adducts. Mol. Pharmacol. 1982, 21 (2), 262–265. (35) Zhu, B. Z.; Zhao, H. T.; Kalyanaraman, B.; Frei, B. Metalindependent production of hydroxyl radicals by halogenated quinones and hydrogen peroxide: An ESR spin trapping study. Free Radical Biol. Med. 2002, 32 (5), 465–473. (36) Kim, J. K.; Metcalfe, I. S. Investigation of the generation of hydroxyl radicals and their oxidative role in the presence of heterogeneous copper catalysts. Chemosphere 2007, 69 (5), 689–696. (37) Ozawa, T.; Hanaki, A. The 1st ESR spin-trapping evidence for the formation of hydroxyl radical from the reaction of copper(II) complex with hydrogen-peroxide in aqueous-solution. J. Chem. Soc.-Chem. Commun. 1991, 5, 330–332. (38) Burkitt, M. J.; Tsang, S. Y.; Tam, S. C.; Bremner, I. Generation of 5,5-dimethyl-1-pyrroline N-oxide hydroxyl and scavenger radical adducts from copper/H2O2 mixtures—Effects of metal-ion chelation and the search for high-valent metal-oxygen intermediates. Arch. Biochem. Biophys. 1995, 323 (1), 63–70. (39) Makino, K.; Hagiwara, T.; Hagi, A.; Nishi, M.; Murakami, A. Cautionary note for dmpo spin trapping in the presence of iron-ion. Biochem. Biophys. Res. Commun. 1990, 172 (3), 1073–1080. (40) Hanna, P. M.; Chamulitrat, W.; Mason, R. P. When are metal ion-dependent hydroxyl and alkoxyl radical adducts of 5,5-dimethyl-1pyrroline N-Oxide artifacts. Arch. Biochem. Biophys. 1992, 296 (2), 640–644. (41) Chamulitrat, W.; Iwahashi, H.; Kelman, D. J.; Mason, R. P. Evidence against the 12-21 quartet dmpo spectrum as the radical adduct of the lipid alkoxyl radical. Arch. Biochem. Biophys. 1992, 296 (2), 645–649. (42) Alegria, A. E.; Ferrer, A.; Sepulveda, E. Photochemistry of water-soluble quinones. Production of a water-derived spin adduct. Photochem. Photobiol. 1997, 66 (4), 436–442. (43) Finkelstein, E.; Rosen, G. M.; Rauckman, E. J.; Paxton, J. Spin trapping of superoxide. Mol. Pharmacol. 1979, 16 (2), 676–685. (44) Roubaud, V.; Sankarapandi, S.; Kuppusamy, P.; Tordo, P.; Zweier, J. L. Quantitative measurement of superoxide generation using the spin trap 5-(diethoxyphosphoryl)-5-methyl-1-pyrroline-N-oxide. Anal. Biochem. 1997, 247 (2), 404–411. (45) Bacic, G.; Spasojevic, I.; Secerov, B.; Mojovic, M. Spin-trapping of oxygen free radicals in chemical and biological systems: New traps, radicals and possibilities. Spectrochim. Acta A- Mol. Biomol. Spectrosc. 2008, 69 (5), 1354–1366. (46) Pou, S.; Cohen, M. S.; Britigan, B. E.; Rosen, G. M. Spintrapping and human-neutrophils - limits of detection of hydroxyl radical. J. Biol. Chem. 1989, 264 (21), 12299–12302. (47) Shi, H. L.; Timmins, G.; Monske, M.; Burdick, A.; Kalyanaraman, B.; Liu, Y.; Clement, J. L.; Burchiel, S.; Liu, K. J. Evaluation of spin trapping agents and trapping conditions for detection of cell-generated reactive oxygen species. Arch. Biochem. Biophys. 2005, 437 (1), 59–68. (48) Eberson, L. Adv. Phys. Org. Chem. 1998, 31, 91–141. (49) Dorfman, L. M., and; Adams, G. E., Reactivity of the hydroxyl radical in aqueous solutions. Nat. Stand. Ref. Data Ser., Nat. Bur. Stand. 1973, NSRDS-NBS 46.
ARTICLE
(50) Brustolon, M.; Giamello, E. Electron Paramagnetic Resonance A Practitioner’s Toolkit; Wiley: New Jersey, 2009. (51) Buettner, G. R. Spin trapping—electron-spin-resonance parameters of spin adducts. Free Radical Biol. Med. 1987, 3 (4), 259–303. (52) Harbour, J. R.; Hair, M. L. Transient radicals in heterogeneous systems—detection by spin trapping. Adv. Colloid Interface Sci. 1986, 24 (23), 103–141. (53) Cohn, C. A.; Pedigo, C. E.; Hylton, S. N.; Simon, S. R.; Schoonen, M. A. A. Evaluating the use of 3 0 -(p-Aminophenyl) fluorescein for determining the formation of highly reactive oxygen species in particle suspensions. Geochem. Trans. 2009, 10 (8), doi:10.1186/14674866-10-8. (54) Li, B. B.; Gutierrez, P. L.; Blough, N. V. Trace determination of hydroxyl radical in biological systems. Anal. Chem. 1997, 69 (21), 4295–4302. (55) Barriga, G.; Olea-Azar, C.; Norambuena, E.; Castro, A.; Porcal, W.; Gerpe, A.; Gonzalez, M.; Cerecetto, H. New heteroaryl nitrones with spin trap properties: Identification of a 4-furoxanyl derivative with excellent properties to be used in biological systems. Bioorg. Med. Chem. 2010, 18 (2), 795–802. (56) Tojo, S.; Tachikawa, T.; Fujitsuka, M.; Majima, T. Oxidation processes of aromatic sulfides by hydroxyl radicals in colloidal solution of TiO2 during pulse radiolysis. Chem. Phys. Lett. 2004, 384 (46), 312–316. (57) Hodgson, E. K.; Fridovich, I. Interaction of Bovine Erythrocyte Superoxide-Dismutase with Hydrogen-Peroxide—Inactivation of Enzyme. Biochemistry 1975, 14 (24), 5294–5299. (58) Hodgson, E. K.; Fridovich, I. Interaction of bovine erythrocyte superoxide-dismutase with hydrogen-peroxide—Chemiluminescence and peroxidation. Biochemistry 1975, 14 (24), 5299–5303. (59) Lawless, D.; Serpone, N.; Meisel, D. Role of OH radicals and trapped holes in photocatalysis—A pulse-radiolysis study. J. Phys. Chem. 1991, 95 (13), 5166–5170. (60) Donaldson, K.; Brown, D. M.; Mitchell, C.; Dineva, M.; Beswick, P. H.; Gilmour, P.; MacNee, W. Free radical activity of PM10: Iron-mediated generation of hydroxyl radicals. Environ. Health Perspect. 1997, 105, 1285–1289. (61) Venkatachari, P.; Hopke, P. K.; Brune, W. H.; Ren, X. R.; Lesher, R.; Mao, J. Q.; Mitchel, M. Characterization of wintertime reactive oxygen species concentrations in Flushing, New York. Aerosol Sci. Technol. 2007, 41 (2), 97–111. (62) Chen, X.; Hopke, P. K.; Carter, W. P. L. Secondary organic aerosol from ozonolysis of biogenic volatile organic compounds: Chamber studies of particle and reactive oxygen species formation. Environ. Sci. Technol. 2011, 45 (1), 276–282. (63) Narayanasamy, J.; Kubicki, J. D. Mechanism of hydroxyl radical generation from a silica surface: Molecular orbital calculations. J. Phys. Chem. B 2005, 109 (46), 21796–21807. (64) Lloyd, R. V.; Hanna, P. M.; Mason, R. P. The origin of the hydroxyl radical oxygen in the Fenton reaction. Free Radical Biol. Med. 1997, 22 (5), 885–888. (65) Borda, M. J.; Elsetinow, A. R.; Strongin, D. R.; Schoonen, M. A. A mechanism for the production of hydroxyl radical at surface defect sites on pyrite. Geochim. Cosmochim. Acta 2003, 67 (5), 935–939.
9239
dx.doi.org/10.1021/es201702q |Environ. Sci. Technol. 2011, 45, 9232–9239
ARTICLE pubs.acs.org/est
Adsorption of Aromatic Carboxylate Ions to Black Carbon (Biochar) Is Accompanied by Proton Exchange with Water Jinzhi Ni,† Joseph J. Pignatello,*,‡ and Baoshan Xing§ †
College of Geographical Sciences, Fujian Normal University, Fuzhou 350007 China Department of Environmental Sciences, Connecticut Agricultural Experiment Station, 123 Huntington Street, P.O. Box 1106, New Haven, Connecticut 06504-1106, United States § Department of Plant, Soil and Insect Sciences, University of Massachusetts, Amherst, Massachusetts 01003, United States ‡
bS Supporting Information ABSTRACT: We examined the adsorption of the allelopathic aromatic acids (AA), cinnamic and coumaric, to different charcoals (biochars) as part of a study on bioavailability of natural signaling chemicals in soil. Sorption isotherms in pH 7 buffer, where the AAs are >99% dissociated, are highly nonlinear, give distribution ratios as high as 104.8 L/ kg, and are insensitive to Ca2+ or Mg2+. In unbuffered media, sorption becomes progressively suppressed with loading and is accompanied by release of OH with a stoichiometry approaching 1 at low concentrations, declining to about 0.40.5 as the pH rises. Sorption of cinnamate on graphite as a model for charcoal was roughly comparable on a surface area basis, but released negligible OH. A novel scheme is proposed that explains the pH dependence of adsorption and OH stoichiometry and the graphite results. In a key step, AA undergoes proton exchange with water. To overcome the unfavorable proton exchange free energy, we suggest AA engages in a type of hydrogen bond recognized to be of unusual strength with a surface carboxylate or phenolate group having a comparable pKa. This bond is depicted as [RCO2 3 3 3 H 3 3 3 O-surf]. The same is possible for AA, but results in increased surface charge. The proton exchange pathway appears open to other weak acid adsorbates, including humic substances, on carbonaceous materials.
’ INTRODUCTION The carboxylic acid functional group is abundant in natural soil organic matter and is present in the molecular structures of many natural and synthetic compounds released to soil, including plant exudates, natural signaling chemicals between rhizosphere species, pesticides, and environmental contaminants. Charcoal black carbon is a component of the soil carbon pool as a result of forest fires and deliberate burning practices.1 In addition, interest has emerged in the application of engineered charcoal from biomass waste, known as biochar, to agricultural and forest lands for its potential benefits to soil quality and for its carbon sequestration value.2 Contemplated levels of biochar to croplands and potting soils range from 1 to 10% or more by weight. The effects of natural or added charcoal on chemical and biological processes in the rhizosphere are mostly uncharacterized. A potentially critical property of charcoal with respect to these processes is its surface activity as an adsorbent. The adsorbent strength of charcoal toward organic compounds is a function of the biomass precursor, charring conditions (time and temperature profile, oxygen concentration), degree of postcharring weathering, and other factors that dictate specific surface area, microporosity, and surface chemistry of the final material. Depending on these factors and abundance in soil, charcoal may contribute substantially to sorption, and therefore reduce the physical mobility and biological availability of r 2011 American Chemical Society
contaminants, as well as the above-mentioned natural compound classes. The factors that govern interactions of neutral organic compounds with charcoal and soot are well-known and characterized.35 By contrast, the interactions of charcoals with weak organic acids that undergo dissociation within the normal pH range of most soils—most relevantly, carboxylic acids, phenols, and sulfonamides—have received little attention. Sorption of weak acids in soils is a function of pH, ionic strength, surface charge and charge density, type and concentration of metal ions, and in some cases the structural metal ion. Sorption of the neutral molecule is governed by the weak forces available to neutral compounds including van der Waals, hydrogen bonding, and solvophobic effects. Specific interactions of organoanions with minerals and whole soils that have been identified include (i) anion-exchange at positively charged sites; (ii) repulsion with the developing negative charge on the surface as the pH increases above the point of zero net charge (pzc); (iii) bridging by metal cations; and (iv) when chelation is possible, inner-sphere coordination to structural Received: May 31, 2011 Accepted: September 22, 2011 Revised: September 12, 2011 Published: October 14, 2011 9240
dx.doi.org/10.1021/es201859j | Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology metal ions.610 Sorption of the organoanion may also involve the above-mentioned weak forces and solvophobic effects, depending on the structure of the rest of the molecule, but solvophobic effects are weaker because of the increased water solubility of the anion relative to the neutral molecule. Sorption of the organoanion in some studies is said to be “negligible”, while in others it is found to be appreciable; for example, polychlorophenolate ions sorb significantly to variable-charge soils even at high pH.6 Although soil organic matter (SOM) is known to be important in the binding of weak acids to whole soils, it has been difficult to separate the influence of SOM from the other components. Binding of carboxylic acids and their anions to SOM has also been studied computationally.11,12 The prior literature on adsorption of weak acids to carbonaceous materials is negligible except in regard to activated carbon.1315 It is generally found that adsorption decreases with increasing ionization of the molecule as the pH increases above the pzc of the surface due to charge repulsion between the anion with the increasingly negatively charged surface, and to the reduced solvophobic effect of the anion relative to the molecule. However, the anion appears to have appreciable affinity for carbons even under strongly alkaline condition. M€uller et al.13,14 modeled adsorption of weak organic electrolytes (benzoic acid and p-nitrophenol) from aqueous solution by combining electrochemical, diffuse-double-layer, and normal adsorption thermodynamic models. Their model assumes that the affinity of the molecular and ionized forms for the surface are identical except for the charge attraction or repulsion term acting on the ionized form. Thus, at pH values where the surface is net negatively charged, the organoanion would be excluded from the surface unless the nonelectrostatic interaction energy outweighed the electrostatic repulsion energy. Our study was undertaken to characterize the adsorption of selected aromatic acid (AA) allelochemicals by black carbon as part of a broader study on the influence of biochar addition to agricultural fields on chemical signaling in the rhizosphere. We studied sorption of cinnamic and coumaric acids to commercial biochar prototypes. Allelochemicals are low molecular weight compounds secreted into soil by plant tissues and/or decay of plant residues that influence the interaction of plants with other individuals of the same species, other plant species, microbes, viruses, or insects. Allelochemicals play an important role in agricultural and ecological dynamics.1620 An important class of allelochemicals is the single-ring “phenolic acids” released by many plants that include coumaric, ferulic, caffeic, p-hydroxybenzoic, phenylacetic, salicylic, trans-cinnamic, vanillic, gallic, and syringic acids, among others.19,20 We have identified an important and heretofore unrecognized mechanism of adsorption of organoanions of weak acids on black carbon—namely, proton exchange with water that results in a speciation change on the surface and concomitant release of hydroxide ion into solution. It should be noted that none of the studies above report any change in pH associated with sorption of organoanions.
’ EXPERIMENTAL SECTION
ARTICLE
CQuest) or gently broken up in a mortar and passed through a sieve to obtain the <2 mm-size fraction (Soil Reef). Source information and selected properties of the biochars are listed in Table S1 of the Supporting Information. t-Cinnamic acid (g99%) and p-coumaric acid (g98%) were purchased from Sigma-Aldrich. Their acid dissociation constants (pKa) are 4.4421 and 4.39,22 respectively. Stock solutions were prepared by dissolving them in water containing a stoichiometic amount of NaOH, and readjusting the pH if necessary. Water referred to below as “nanopure” is tap water purified by reverse osmosis followed by treatment in a commercial purifier (Barnstead Nanopure) to achieve resistivity of >18.2 MΩ-cm. Surface/pore analysis was conducted by gas porisimetry on an Autosorb-1 (Quantachrome Instruments., Boynton Beach, FL). The outgas temperature was 200 °C. Gas adsorption isotherms were evaluated with the BrunaurEmmettTeller (N2 isotherm at 77 K; 11 points) or Grand Canonical Monte Carlo Density Functional Theory (CO2 isotherm at 273 K) models using built-in software to calculate surface areas and pore size distribution. Potentiometric Titration of the Biochars. Biochar (0.4 g for Agrichar and 0.5 g for Soil Reef) was prewetted in 5 mL of nanopure water for 48 h at 20 ( 1 °C with end-over-end mixing at 40 rotations per minute (rpm). Then varying amounts of standard HCl or NaOH solution were added to each sample and to a corresponding blank vial containing the water but no biochar. Preboiled water was used for titration in the alkaline region and the vials were degassed with N2 prior to addition of the NaOH through the septum. The pH was measured after 48 h of mixing at 20 ( 1 °C. The nominal initial H+ or OH concentration in the sample was calculated from the pH of its corresponding blank. Sorption Experiments. Sorption isotherms were constructed by placing 40 mg of Agrichar or 100 mg of Soil Reef into a 60-mL polytrifluoroethylene (PTFE)-lined screw cap glass vial, along with 50 mL of nanopure water or 0.05 M phosphate buffer (pH 7.0). A parallel set of controls without biochar was set up. Samples and controls without buffer were degassed with N2. After 48 h prewetting, the pH was measured in three sacrificed samples to establish initial pH, and a stock solution of the AA was adjusted to the average pH of the sacrificed samples. This stock solution was used to spike the samples and corresponding controls. The vials were mixed end-over-end at 40 rpm at 20 ( 1 °C for an additional 48 h. The aqueous phase was then sampled and microfiltered (0.45 μm) to remove any biochar. The AA concentration was determined by high-performance liquid chromatography on a C-18 column (S 5 ODS2; phase Sep, Clwyd, U.K.) eluted with 30:70 (v/v) CH3CN/water containing 20 mM acetic acid (pH 3.2) with monitoring at 270 nm for cinnamic acid and 314 nm for coumaric acid. The sorbed concentration was calculated by material balance. In preliminary experiments 48 h appeared sufficient to reach equilibrium. Whereas true equilibrium is difficult to judge, we make the reasonable assumption that trends in sorption observed over the 48-h contact period are representative of trends in any sorption occurring after that time. Isotherms were fit to the Freundlich model (eq 1) and the Langmuir model (eq 2)
Materials. Biochars were generously provided by different manufacturers: Soil Reef by EcoTechnologies Group, LLC, Berwyn, PA; CQuest by Dynamotive Energy Systems Corp., McLean, VA; and Agrichar by BEST Energies Australia, Somersby, Australia. The samples were used either as-received (Agrichar and
S ¼ K F CN S¼ 9241
ð1Þ
Smax L KL C 1 þ KL C
ð2Þ
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology
Figure 1. Isotherms of (A) cinnamate and (B) coumarate on Agrichar and Soil Reef in phosphate buffer (pH 6.97.0) and fits to two sorption models.
where S and C are the sorbed (mg/kg) and solution (mg/L) concentrations, respectively, N is the Freundlich exponent, is the KF is the Freundlich affinity-capacity parameter, Smax L Langmuir capacity parameter, and KL is the Langmuir affinity parameter. The Freundlich parameters were determined by linear regression of log-transformed data, while the Langmuir parameters were determined by nonlinear regression of untransformed data. In both cases the data were weighted by the dependent variable. The distribution ratio, Kd, is defined as S/C at a specified concentration. Sorption experiments to determine stoichiometry were carried out in the same way except using a higher biochar/water ratio (0.4 g for Agrichar, 1.0 g for Soil Reef, and 3.0 g for graphite per 10 mL). Experiments to determine the influence of metal ions on sorption of AA by Agrichar (40 mg of solids and 50 mL of liquid phase) were conduced in a similar manner except for the addition at the prewetting step of CaCl2 or MgCl2 and NaCl to keep ionic strength equal in all vials. A constant mass of AA was added to each vial.
’ RESULTS AND DISCUSSION In screening tests we measured the reduction in solutionphase concentration of AAs after equilibration with increasing
ARTICLE
biochar concentrations in water initially adjusted to pH 5 or 7 with HCl/NaOH (Figure S1, SI). At pH 5, the fraction of cinnamic and coumaric acids in dissociated form is 78.5% and 80.3%, respectively. At pH 7, cinnamic and coumaric acids are >99.7% dissociated. We found that sorption is greater at pH 5 than 7, follows the order Agrichar > Soil Reef . CQuest, and is slightly greater for cinnamic than coumaric acids in all cases at the tested concentrations. Undoubtedly the weak sorbent property of CQuest in comparison to the others is due to the fast pyrolysis method of production, which leaves the material with significant incompletely charred biopolymer and permeated with a greater amount of tarry residue. Sorption isotherms of cinnamic acid and coumaric acid for Agrichar and Soil Reef in phosphate buffer at pH 6.9 are shown in Figure 1 and the model parameters are listed in Table S2. Isotherms on CQuest were not constructed in view of its poor sorbent ability in the screening tests. The isotherms are highly nonlinear even on log scale. Neither the Freundlich nor the Langmuir models proved universally suitable. The order in sorption intensity regardless of liquid phase concentration is Agrichar > Soil Reef. Sorption intensity follows the order cinnamate > coumarate over most of the tested concentration range; the difference is more pronounced for Soil Reef than Agrichar. The trends displayed in the screening tests and the isotherms have conventional explanations. Sorption is greater at pH 5 due to the greater abundance of the molecular form and the lower negative charge of the surface (see below) compared to pH 7.14 Sorption trends qualitatively with the N2 BET of the biochars listed in Table S1: namely, Agrichar (427 m2/g) > Soil Reef (338 m2/g) . CQuest (0.1 m2/g). Sorption also trends with the CO2 GCMC surface area. The order in sorption intensity between the two AAs is plausibly related to solvophobic effects. The octanol water partition coefficient (Kow) is a commonly used index of solvophobicity. According to SPARC calculator (http://sparc. chem.uga.edu/sparc/; accessed November 17, 2010) the log Kow of the molecular and anionic forms of cinnamic acid are 2.50 and 0.42, respectively, and those of coumaric acid are 1.78 and ∼ 1, respectively, consistent with this conclusion. Sorption of the organoanions, reflected in the Kd at pH 6.9, is remarkably strong, however, a fact that is not well-explained by solvophobic effects alone. Depending on concentration, the log Kd for cinnamate on Agrichar ranges 3.74.2 and on Soil Reef ranges 3.13.8. Likewise, log Kd of coumarate on Agrichar ranges 3.54.8 and on Soil Reef ranges 2.63.9. The Kd values are thus many orders of magnitude greater than the estimated Kow value of the respective organoanion. This finding seems inconsistent with the sorbed species being the free organoanion. Rather, it implicates either a speciation change or a strong specific interaction of the organoanion on the surface. We next determined the effects of up to 0.1 M Ca2+ and Mg2+ on sorption of the AAs at constant mass of AA added and ionic strength (Figure S2, SI). We expected that if the anionic form were sorbing, these metal ions would enhance sorption by serving as a cation bridge between the carboxylate group and a negatively charged surface group, such as a carboxylate or phenolate group (e.g., RCO2 3 3 3 M2+ 3 3 3 O2C-BC). The metal may interact with these anions either by contact or solvent-separated ion pairing.23 Cation bridging is an important mechanism triggering the aggregation of humic molecules into larger colloidal structures (NOM) according to molecular dynamics computations.23 Cation bridging also has been 9242
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology
ARTICLE
Figure 2. Sorption isotherms of cinnamate for biochars comparing buffered (phosphate pH 6.9) and nonbuffered conditions and the accompanying evolution of hydroxide ion concentration. The initial solution composition was 0.005 M CaCl2. The initial nonbuffered pH averaged 7.38 for Agrichar and 7.95 for Soil Reef.
proposed as a mechanism for sorption of carboxylate and phenolate compounds to whole soils,10,6 model soil minerals,9,24 and soil organic matter25 on the basis of physical experiments, as well as to model humic structures on the basis of computations.11,12 Figure S2, however, reveals little, if any, systematic change in sorption induced by Ca2+ and Mg2+. This finding implies that sorption of the AAs is not greatly affected by charged sites under the influence of these metal cations. Figure 2A shows linear-scale plots of the isotherms of cinnamate on Agrichar in phosphate-buffered vs nonbuffered suspensions. At zero concentration of cinnamate the buffered and nonbuffered suspensions had equilibrated during the prewetting stage to a similar pH (6.9 and 7.2, respectively). The isotherms are seen to deviate from one another as AA concentration increases—the nonbuffered samples giving reduced sorption relative to the buffered samples. Moreover, the OH concentration of the nonbuffered solutions increases relative to the buffered solution as loading increases. Because the AA stock solution was adjusted to the approximate initial concentration of the biochar suspension, vials containing just the aqueous phase showed no significant increase in hydroxide ion concentration with increasing cinnamate concentration up to the same levels added (data not shown). Soil Reef showed results qualitatively similar to those for Agrichar, except the isotherm and [OH] data are more scattered (Figure 2B). Taken together, the results show that sorption of AA by biochar is accompanied by the release of hydroxide ion into solution (eq 3), which presumably is the cause of progressive sorption suppression. RCO2 þ BC h ðRCO2 Þ 3 3 3 BC þ OH
ð3Þ
To determine the magnitude of OH release the buffering capacity of the biochar must be taken into account OH þ BC h BC þ H2 O
ð4Þ
At a given pH, the amount of OH released by AA sorption is the observed amount appearing in solution plus the amount consumed by the biochar at the final pH as determined in an independent titration experiment using the same equilibration period (48 h) and temperature (20 °C) as the sorption experiment. The raw titration curves and the curves representing specific uptake of H+ or OH versus pH calculated from the raw titration data are provided in Figures S3 and S4, respectively. The crossover pH—where the pH of the sample is equal to the pH of the blank (see Figure S3)—is 8.07 for Agrichar, 7.96 for Soil Reef, and 6.6 for CQuest. Consumption of OH at any pH above the crossover pH, which represents the biochar’s buffering capacity, follows the order CQuest > Soil Reef > Agrichar. Consumption of H+ at any pH below the crossover pH follows the reverse order. The pH at the pzc is best determined by electrophoretic mobility. The pHpzc for Agrichar is 3.94.3 (Table S1), indicating that the net charge on the surface is negative under the conditions of all sorption experiments of this study. This is likely to be true also for Soil Reef because of the similarity in the crossover pH. Quantification of OH released as a function of AA sorbed (the stoichiometry) required separate experiments using higher biochar/water ratios than used for constructing the isotherms in Figure 2 in order to obtain greater accuracy in the pH change. Figure 3 shows the results of these experiments. Total moles OH generated is the observed moles OH in solution in these sorption experiments plus the moles OH consumed by the biochar at the same pH in the titration experiments, both after 48 h. Moles OH consumed by the biochar at each pH was estimated by curve fitting the titration curve in the alkaline region, shown as the curves in Figure S4, and using the fit for interpolation purposes in the sorption experiment. Figure 3 shows that the stoichiometry between OH and cinnamate sorbed is not constant but decreases with increasing cinnamate loading and/or pH accompanying loading. At the lowest sorbed concentration the OH/cinnamate molar ratio is 9243
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology
ARTICLE
Figure 3. Stoichiometry of hydroxide ion release versus moles cinnamic acid adsorbed for (A) Agrichar (per 0.4 g), (B) Soil Reef (per 1.0 g), and (C) graphite (per 3.0 g). The pH record of the blanks (without biochar) shows that adding AA does not contribute appreciably to the increase in [OH] in solution in the absence of biochar.
approximately 1, while this ratio decreases to about 0.4 (Agrichar) or about 0.5 (Soil Reef). We also tested whether hydroxide is released on adsorption of AA to nonporous powdered graphite, which we found previously to be a good model for black carbon with respect to adsorption of nonionic compounds.26 Sorption and titration experiments were conducted for cinnamate on graphite in the same way as for the biochars. Not surprisingly, sorption of cinnamate was much weaker on graphite than on Agrichar and Soil Reef on a sorbent mass basis (Figure S3, Table S2), the Kd (L/kg) being >300 times smaller than on Agrichar and >70 times smaller than on Soil Reef. However, on a N2BET surface area
basis, adsorption of cinnamate was <4 times smaller than on Agrichar and on a par with Soil Reef. Also not surprisingly, the surface acidity of graphite is far lower compared to Agrichar and Soil Reef (Figure S4). The surface acidity of graphite may be attributed to impurities and/or the presence of a small amount of functionality along defects and edges. The results in Figure 3C show that, although adsorption of cinnamic acid on graphite, like biochar, was accompanied by an increase in pH, the stoichiometry of mole OH released per mole of cinnamate adsorbed was only a tiny fraction of that for the biochars, never exceeding 0.0012. Mechanism. One may reasonably assume that sorption occurs by an adsorptive rather than an absorptive mechanism, but the 9244
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology
ARTICLE
following proposal does not require a distinction. Potential sources of hydroxide release include (i) a spike artifact; (ii) displacement of adsorbed hydroxide ion by AA; and (iii) proton exchange of AA with water. An artifact of the spike procedure is ruled out because the pH of the controls without biochar changed insignificantly with loading (Figure 3), since the pH of the stock solution was adjusted to the same pH as the prewetted biochar. Moreover, the change in ionic strength accompanying spiking (<0.003 M) was negligible. The most plausible form of adsorbed hydroxide is coordination to a surface metal ion associated with the small amount of ash present in these biochars (Table S1). Nevertheless, ligand exchange of AA for hydroxide is deemed unlikely, since previous researchers have found no evidence for inner-sphere coordination of non-chelating carboxylic acid compounds sorbed to minerals: e.g., 2,4-dichlorophenoxyacetic acid (2,4-D) and 2-(4-chloro-2-methylphenoxy)propionic acid adsorbed to α-alumina9 or several different iron oxides,8 and 2,4-D sorbed to variable charge soils.6 The most likely source of hydroxide release, therefore, is proton exchange with water (eq 5) RCO2 þ H2 O h RCO2 H þ OH
ð5Þ
The standard free energy of proton exchange is given by ΔGH0 +exch = RT ln(KaAA/Kaw), where KaAA and Kaw are the acid dissociation constants of AA and water, respectively. The resulting values, 54.5 kJ/mol for cinnamate and 54.8 kJ/mol for coumarate, indicate that proton exchange is rather unfavorable. One might suppose it could be compensated by the increased solvophobicity of the uncharged molecule relative to the anion. This may be tested by considering the hypothetical competition reaction, involving solvophobic forces only, between molecule and anion in eq 6, and assuming a 1:1 free energy relationship between solvophobic-only adsorption and octanolwater partitioning. RCO2 H þ ðRCO2 Þ 3 3 3 BC h RCO2 þ ðRCO2 HÞ 3 3 3 BC
ð6Þ
Kd ðRCO2 HÞ Kd ðRCO2 Þ Kow ðRCO2 HÞ ≈ 17 kJ=mole ≈ RT ln Kow ðRCO2 Þ
ΔΔG0competition ¼ RT ln
The ratio between the experimental Kow for the acid and the calculated Kow for the anion is roughly 103 (see above), which corresponds to only 17 kJ/mol at 298 K for the free energy of reaction 6. Allowing for uncertainty in this number, if this ratio were instead 105, the free energy of reaction 6 would still be only about 28 kJ/mol. Thus, the increased solvophobicity of the adsorbate upon protonation is not by itself great enough to overcome the proton exchange penalty. Consequently, an additional stabilization mechanism on the surface is required to rationalize proton exchange-assisted adsorption. We suggest the extra stabilization energy is gained by AA engaging in a strong hydrogen bond with a surface carboxylate or phenolate group, which are abundant in charcoal:
negative charge-assisted hydrogen bond ()CAHB represented by (D 3 3 3 H 3 3 3 A). This type includes, among others, hydrogen dicarboxylate conjugate pairs [RCO2 3 3 3 H 3 3 3 O2CR] and hydrogen carboxylate/phenolate complexes [RCO2 3 3 3 H 3 3 3 OR] in which ΔpKa is small. The ()CAHB is characterized by equal or nearly equal sharing of the proton and delocalization of the charge between D and A. This gives it considerable covalent character, and in the limiting case makes it a 3-center, 4-electron covalent bond. By contrast, the ordinary H-bond is electrostatic in character. The hydrogen dicarboxylate conjugate pair, in which ΔpKa is zero, ranks among the strongest known H bonds between organic species.27,28 The gas-phase enthalpies of the conjugate pairs of formic, acetic, and 2-methylpropionic acids are 154, 123, and 125 kJ/mol, respectively.29 The enthalpy of the H bond between phenol and acetate, [C6H5O 3 3 3 H 3 3 3 O2CCH3], at 114.8 kJ/mol is slightly lower because ΔpKa is 5.2. It is reasonable to expect that surface carboxyl groups have pKa values above 4 and that surface phenolic groups have pKa values of 10 or below. Thus, we expect the ΔpKa between AA and carboxyl and phenoxyl groups on the charcoal surface to be e ∼5 in most cases, and therefore the enthalpy to range between about 115 and about 154 kJ/mol. These values may be compared with the enthalpy of the gasphase complex between the carboxylate anion and water, RCO2 3 3 3 H2O, which is 5767 kJ/mol depending on R, and between phenoxide and water which is 64.5 kJ/mol.29 The free energy of formation ΔG0()CAHB in water can be estimated from the gas-phase reactions in a straightforward manner by a thermodynamic cycle (see SI).29,30 The resulting ΔG0()CAHB is about 56.2 kJ/mol for the hydrogen dicarboxylate conjugate pair, and about 50.2 kJ/mol for the hydrogen carboxylatephenolate pair. We see from this analysis that the free energy of ()CAHB formation between AA and surface oxyl groups comes close to compensating for the unfavorable proton exchange free energy of 54.8 kJ/mol (eq 5), and exceeds it if the solvophobic contribution from converting AA to AA is included. The lack of O functionality on graphite may rationalize why adsorption of AA is not accompanied by significant OH release. One can see that formation of a ()CAHB at the surface is possible for both AA and AA: the product is the same depending on whether the surface OH group is dissociated (9O) or not (9OH). While adsorption of AA does not demand the “uphill” proton exchange process, AA is less solvophobic than AA and its adsorption places a negative charge on the already negatively charged surface. We propose a model that explains the dependence of overall sorption on pH, the dependence of stoichiometry on pH, and the adsorption of AA on graphite that occurs without the release of hydroxide. The individual ()CAHB reactions of AA and AA with surface H-acceptor and H-donor groups, respectively, may be written Keq7
RCO2 þ H2 O þ 9O a ð9OHO2 CRÞ þ OH Keq7 ¼
½ð9OHO2 CRÞ ½OH ½9O ½RCO2
ð7Þ
Keq8
RCO2 þ 9OH a ð9OHO2 CRÞ
According to Gilli,27,28 the enthalpy of H-bonds (DH 3 3 3 A) is ΔpKa-driven; the strongest are those in which the difference between the pKa of the H-donor (DH h D + H+) and the pK a of the H-acceptor (AH + h A: + H + ) is small. Included in the category of very strong is the
Keq8 ¼ 9245
½ð9OHO2 CRÞ ½9OH½RCO2
ð8Þ
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology
ARTICLE
where the brackets represent concentration in appropriate units, and Keq 7 and Keq 8 are the respective equilibrium constants. Note that eq 7 incorporates the proton exchange in eq 5. The relationship between surface H-acceptor and H-donor sites is given by KaS
9OH a 9O þ Hþ ; ½9O 3 Kaw intr zFΨ=RT ¼ KaS e KaS ¼ ½OH ½9OH
ð9Þ
where KaS is the acid dissociation constant of 9OH at existing charge density, Kintr aS is the intrinsic acid dissociation constant of 9OH groups, Kaw is the acid dissociation constant of water22 (1014), z is the proton valence (+1), F is the Faraday constant, Ψ is the surface potential with respect to bulk solution, R is the gas constant, and T is temperature.13,14,31 Both AA and AA may undergo interactions also with sites other than those represented by reactions 7 and 8. We may refer to these interactions collectively as “solvophobic” adsorption sites. Solvophobic adsorption of AA on the charcoal surface is demonstrated by the significant adsorption of cinnamate ion by graphite, and further supported by the fact that the OH /adsorbate stoichiometry decreases with pH while adsorption remains strong (K d > 1000 L/kg), even four pH units above the pK a. Solvophobic adsorption may be written as RCO2 H þ 9 Kd, AA solvo RCO2 þ 9
Kd, AAsolvo
a
9 3 3 3 RCO2 H ½9 3 3 3 RCO2 H ¼ ½RCO2 H Kd, AA solvo
a
9 3 3 3 RCO2
zFΨ=RT Kd, AA solvo ¼ Kd,non-elec ¼ AA solvo e
ð10Þ
½9 3 3 3 RCO2 ½RCO2
ð11Þ
zFΨ/RT where Kd,nonelec is AA solvo is the nonelectrostatic component and e the electrostatic component of the surface potential operative for AA. The observed, concentration-dependent adsorption distribution ratio may be written S Kd ¼ ¼ ½ð9OHO2 CRÞ reaction7 C þ ½ð9OHO2 CRÞ reaction8 þ ½RCO2 HKd, AA solvo þ ½RCO2 Kd, AA solvo =C ð12Þ
where S and C refer to the total adsorbed and dissolved concentrations,, respectively, including all species. Recognizing that C ≈ [RCO2] above pH 7, and combining eqs 79 and 12 gives ! ½9O 1 Keq7 þ Keq8 3 14 intr zFΨ=RT Kd ¼ ½OH 10 3 KaS e þ
1 Kd, AA solvo 1 þ ½OH 3 10ð14 pKa Þ 3
zFΨ=RT þ Kd,non-elec AA solvo e
ð13Þ
Equation 13 predicts that overall adsorption will decline with rising pH, consistent with observation. The ratio [9-O]/[OH]
in the first term decreases with increasing pH because 9OH groups are incompletely deprotonated by released (or added) hydroxide due to charge buildup on the surface. The second term includes aqueous speciation of the neutral molecule and is inversely related to pH. The third term reflects the growing charge repulsion of AA on the surface (Ψ is negative) as pH rises. In addition, eq 13 predicts the decline in reaction 7 relative to zFΨ/RT , which reaction 8 with pH owing to the term 1/1014 3 Kintr aS e increases with pH because Ψ becomes progressively more negative. The decline in the contribution of reaction 7 relative to reaction 8, in turn, predicts the observed decline in OH/ adsorbate stoichiometry with loading. Another way to look at it is that, with increasing pH, proton exchange becomes energetically more costly at a faster pace than placing a charge on the surface. Interestingly, the OH/adsorbed-cinnamate stoichiometry for graphite increases with loading (Figure 3C). This result may represent a cooperative effect in which adsorbed AA ions act as H-bond acceptor sites for AA; that is, AA and AA form a conjugate pair (RCO2 3 3 3 H 3 3 3 O2CR) on the graphite surface linked by a ()CAHB RCO2 þ H2 O þ RCO2 3 3 3 9 h ðRCO2 HO2 CRÞ 3 3 3 9 þ OH
ð14Þ However, the cooperative effect, if it occurs, is quite small and is unlikely to contribute much to sorption of the aromatic acids on the biochars. The results of this study point to a novel pathway for adsorption of carboxylic acid anions from water to black carbon that involves proton exchange with water and speciation change on the surface accompanied by release of hydroxide into solution. Adsorption through this pathway appears to constitute from ∼40% to ∼100% of total adsorption in the pH range of our experiments. The reaction appears driven by the formation of an especially strong H-bond on the surface, supplemented by the increase in solvophobicity of the adsorbate. Although accurate numbers are unavailable, it is estimated that these processes together contribute ∼70 kJ/mol to offset the ∼55 kJ/mol standard free energy of proton exchange. It is possible that proton exchange-assisted adsorption on natural particles rich in carboxylic and phenolic groups is characteristic of any weak acid that is capable of forming a very strong H-bond by virtue of pKa rough equivalency. Other examples include various kinds of amines and heterocyclic amines (e.g., anilines, azines, azoles), sulfonamides, and phenols with electronegative substituents.28 We have now demonstrated proton exchange-assisted adsorption on Agrichar of the weak acid sulfamethazine, a veterinary antibiotic having a pKa2 of 7.42 for the sulfonamide (SO2NH) group.37 The stroichiometry of OH release ranged from about 1 at pH 9.3 to about 0.4 at pH 10.5. Just as with the aromatic carboxylates, we have proposed that adsorption of sulfamethazine is augmented by a ()CAHB between the sulfonamide moiety and a surface oxyl group: [(R)(R0 SO2)N 3 3 3 H 3 3 3 O-9]. Extension of the sorption edge of weak acids well above the pKa has been noted many times previously for soils6 and activated carbons.1315 We think our proposed mechanism can partly account for this observation because it has the effect of raising the pKa of the acid on the surface. We are aware of no study linking organoanion sorption to protonation, nor of any mention of formation of a ()CAHB with the surface. The closest to come was a study32 finding that aromatic sulfonic acids do not fit a Polanyi-linear solvation free energy model for adsorption of 9246
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology neutral compounds by activated carbon because they are fully ionized (pKa < 1). The authors proposed that sulfonate ions take up protons from solution and adsorb onto the surface. However, they did not provide direct evidence for proton exchange, nor suggest a driving force that could overcome the free energy of proton exchange, which is even more unfavorable than for the AAs. Proton exchange during weak acid sorption has gone unrecognized until now (apparently) for two main reasons. First, most experiments are performed using buffered systems. Second, the sorbent itself serves as a buffer, masking the release of hydroxide to some degree. Indeed, the observed changes in solution pH were smaller for Soil Reef than for Agrichar because Soil Reef contains more O (inferred from the C, H, and ash composition; Table S1) and consumes more OH at any given pH (Figure S4). An important question is whether weak acids undergo proton exchange-assisted sorption to soil organic matter, which is rich in carboxylic and phenolic functionality. Proton exchange in this case may be difficult to observe, however, because of the high buffering capacity of soil organic matter. In a computational study,12 various H-bonds were observed between the carboxylate group of the antimicrobial Ciprofloxacin, and a fully protonated model humic molecule and were concluded to be important to the stability of the complex. Finally, the results of this study have implications for adsorption of dissolved natural organic matter by carbonaceous materials including black carbon,33,34 activated carbon,35 and engineered carbon nanoparticles,36 in which interactions of dissociable groups are likely to be important.
’ ASSOCIATED CONTENT
bS
Supporting Information. Tables and figures on sources and properties of the biochars; sorption model parameters; the results of sorption screening experiments; effect of metal ion concentration, titration curves of the biochars and graphite; sorption isotherm of cinnamate on graphite; and estimation of conjugate bond formation free energy in water. This information is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]; phone: 203-974-8518.
’ ACKNOWLEDGMENT The bulk of this work was performed while J.Z. Ni was a Visiting Scholar at the Connecticut Agricultural Experiment Station through an arrangement with the University of Massachusetts, Amherst, and with funding provided by Fujian Normal University, China. J.Z. Ni thanks these institutions for the opportunity. The authors also thank the Department of Analytical Chemistry at CAES for the emergency use of instrumentation. ’ REFERENCES (1) Skjemstad, J. O.; Reicosky, D. C.; Wilts, A. R.; McGowan, J. C. Charcoal carbon in U.S. agricultural soils. Soil Sci. Soc. Am. J. 2002, 66, 1249–1255. (2) Lehmann, J.; Joseph, S., Eds. Biochar for Environmental Management: Science and Technology; Earthscan Publications Ltd.: Lndon, UK, 2009.
ARTICLE
(3) Cornelissen, G.; Gustafsson, O.; Bucheli, T. D.; Jonker, M. T. O.; Koelmans, A. A.; Van Noort, P. C. M. Extensive sorption of organic compounds to black carbon, coal, and kerogen in sediments and soils: Mechanisms and consequences for distribution, bioaccumulation, and biodegradation. Environ. Sci. Technol. 2005, 39, 6881–6895. (4) Pignatello, J. J. Biophysico-Chemical Processes of Anthropogenic Organic Compounds in Environmental Systems. In IUPAC Series on Biophysico-Chemical Processes in Environmental Systems; Huang, P. M., Ed.; Wiley, 2011; Vol. 3, pp 350. (5) Allen-King, R. M.; Grathwohl, P.; Ball, W. P. New modeling paradigms for the sorption of hydrophobic organic chemicals to heterogeneous carbonaceous matter in soils, sediments, and rocks. Adv. Water Res. 2002, 25, 985–1016. (6) Hyun, S.; Lee, L. S. Hydrophilic and Hydrophobic Sorption of Organic Acids by Variable-Charge Soils: Effect of Chemical Acidity and Acidic Functional Group. Environ. Sci. Technol. 2004, 38, 5413–5419. (7) Hyun, S.; Lee, L. S. Factors Controlling Sorption of Prosulfuron by Variable-Charge Soils and Model Sorbents. J. Environ. Qual. 2004, 33, 1354–1361. (8) Clausen, L.; Fabricius, I. Atrazine, isoproturon, mecoprop, 2,4-D, and bentazone adsorption onto iron oxides. J. Environ. Qual. 2001, 30, 858–869. (9) Clausen, L.; Fabricius, I.; Madsen, L. Adsorption of pesticides onto quartz, calcite, kaolinite, and alumina. J. Environ. Qual. 2001, 30, 846–857. (10) Dubus, I. G.; Barriuso, E.; Calvet, R. Sorption of weak organic acids in soils: clofencet, 2,4-D and salicylic acid. Chemosphere 2001, 45, 767–774. (11) Aquino, A. J. A.; Tunega, D.; Haberhauer, G.; Gerzabek, M. H.; Lischka, H. Interaction of the 2,4-dichlorophenoxyacetic acid herbicide with soil organic matter moieties: A theoretical study. Eur. J. Soil Sci. 2007, 58, 889–899. (12) Aristilde, L.; Sposito, G. Binding of ciprofloxacin by humic substances: A molecular dynamics study. Environ. Toxicol. Chem. 2010, 29, 90–98. (13) M€uller, G.; Radke, C. J.; Prausnitz, J. M. Adsorption of Weak Organic Electrolytes from Aqueous Solution on Activated Carbon. Effect of pH. J. Phys. Chem. 1980, 84, 369–376. (14) M€uller, G.; Radke, C. J.; Prausnitz, J. M. Adsorption of Weak Organic Electrolytes from Dilute Aqueous Solution onto Activated Carbon. Part I. Single-Solute Systems. J. Colloid Interface Sci. 1985, 103, 466–483. (15) Radovic, L. R.; Morena-Castilla, C.; Rivera-Utrilla, J. In Chemistry and Physics of Carbon: A Series of Advances; Radovic, L. R., Ed.; Marcel Dekker, Inc.: New York, 2001; Vol. 27, pp 227405. (16) Putnan, A. R.; Tang, C. S. The Science of Allelopathy; John Wiley & Sons: New York, NY, 1986. (17) Gniazdowska, A.; Bogatek, R. Allelopathic interactions between plants. Multi site action of allelochemicals. Acta Physiol. Plant. 2005, 27, 395–407. (18) Inderjit; Duke, S. O. Ecophysiological aspects of allelopathy. Planta 2003, 217, 529–539. (19) Tharayil, N.; Bhowmik, P. C.; Xing, B. S. Preferential sorption of phenolic phytotoxins to soil: Implications for altering the availability of allelochemicals. J. Agric. Food Chem. 2006, 54, 3033–3040. (20) Tharayil, N.; Bhowmik, P. C.; Xing, B. S. Bioavailability of allelochemicals as affected by companion compounds in soil matrices. J. Agric. Food Chem. 2008, 56, 3706–3713. € (21) Erdemgil, F. Z.; Sanli, S.; Sanli, N.; Ozkan, G.; Barbosa, J.; Guiteras, J.; Beltran, J. L. Determination of pKa values of some hydroxylated benzoic acids in methanolwater binary mixtures by LC methodology and potentiometry. Talanta 2007, 72, 489–496. (22) Lide, D. R., Ed. Handbook of Chemistry and Physics; 85th ed.; CRC Press: Boca Raton, FL, 2004. (23) Iskrenova-Tchoukova, E.; Kalinichev, A. G.; Kirkpatrick, R. J. Metal Cation Complexation with Natural Organic Matter in Aqueous Solutions: Molecular Dynamics Simulations and Potentials of Mean Force. Langmuir 2010, 26, 15909–15919. 9247
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
Environmental Science & Technology
ARTICLE
(24) Nowara, A.; Burhenne, J.; Spiteller, M. Binding of Fluoroquinolone Carboxylic Acid Derivatives to Clay Minerals. J. Agric. Food Chem. 1997, 45, 1459–1463. (25) MacKay, A. A.; Canterbury, B. Oxytetracycline sorption to organic matter by metal-bridging. J. Environ. Qual. 2005, 34, 1964–1971. (26) Zhu, D.; Pignatello, J. J. Characterization of aromatic compound sorptive interactions with black carbon (charcoal) assisted by graphite as a model. Environ. Sci. Technol. 2005, 39, 2033–2041. (27) Gilli, G.; Gilli, P. Towards an unified hydrogen-bond theory. J. Mol. Struct. 2000, 552, 1–15. (28) Gilli, P.; Pretto, L.; Bertolasi, V.; Gilli, G. Predicting HydrogenBond Strengths from Acid-Base Molecular Properties. The pK(a) Slide Rule: Toward the Solution of a Long-Lasting Problem. Acc. Chem. Res. 2009, 42, 33–44. (29) Meot-Ner, M.; Sieck, L. W. The ionic hydrogen bond and ion solvation. 5. OH 3 3 3 O- bonds. Gas-phase solvation and clustering of alkoxide and carboxylate anions. J. Am. Chem. Soc. 1986, 108, 7525–7529. (30) Ouyang, B.; Howard, B. J. High-resolution microwave spectroscopic and ab initio studies of propanoic acid and its hydrates. J. Phys. Chem. A 2008, 112, 8208–8214. (31) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental Organic Chemistry, 2nd ed.; John Wiley & Sons: New York, 2002. (32) Crittenden, J. C.; Sanongraj, S.; Bulloch, J. L.; Hand, D. W.; Rogers, T. N.; Speth, T. F.; Ulmer, M. Correlation of Aqueous-Phase Adsorption Isotherms. Environ. Sci. Technol. 1999, 33, 2926–2933. (33) Koelmans, A. A.; Meulman, B.; Meijer, T.; Jonker, M. T. O. Attenuation of polychlorinated biphenyl sorption to charcoal by humic acids. Environ. Sci. Technol. 2009, 43, 736–742. (34) Pignatello, J. J.; Kwon, S.; Lu, Y. Effect of natural organic substances on the surface and adsorptive properties of environmental black carbon (char): Attenuation of surface activity by humic and fulvic acids. Environ. Sci. Technol. 2006, 40, 7757–7763. (35) Kilduff, J. E. Adsorption of natural organic polyelectrolytes by activated carbon: A size-exclusion chromatography study. Enivron. Sci. Technol. 1996, 30, 1336–1343. (36) Hyung, H.; Fortner, J. D.; Hughes, B.; Kim, J.-H. Natural Organic Matter Stabilizes Carbon Nanotubes in the Aqueous Phase. Environ. Sci. Technol. 2007, 41, 179–184. (37) Teixido, M.; Pignatello, J. J.; Beltran, J. L.; Grenados, M.; Peccia, J. Speciation of the Ionizable Antibiotic Sulfamethazine on Black Carbon (Biochar). Environ. Sci. Technol. in press (doi:10.1021/ es202487h).
9248
dx.doi.org/10.1021/es201859j |Environ. Sci. Technol. 2011, 45, 9240–9248
ARTICLE pubs.acs.org/est
Simulated Sunlight Action Spectra for Inactivation of MS2 and PRD1 Bacteriophages in Clear Water Michael B. Fisher,† David C. Love,†,‡ Rudi Schuech, and Kara L. Nelson* Department of Civil and Environmental Engineering, University of California Berkeley, Berkeley, California 94720-1710, United States
bS Supporting Information ABSTRACT: Action spectra for simulated sunlight were measured in clear water for two viruses: PRD1, a double-stranded DNA bacteriophage, and MS2, a single-stranded RNA bacteriophage. Viruses were diluted into phosphate buffered saline (20 mM PBS, pH 7.5) and exposed for 22 h to simulated sunlight either directly or through one of six glass filters with 50% cutoff wavelengths ranging from 280 to 350 nm. Virus survival was measured using the double agar layer plaque method. Both UVA (320400 nm) and UVB (280320 nm) light were found to contribute to PRD1 inactivation, while only UVB inactivated MS2. A computational model was developed for interpreting these action spectra with 3-nm resolution. Using these methods, we provide detailed estimates of the sensitivity of MS2 and PRD1 to photoinactivation from 285 to 345 nm. The resulting sensitivity coefficients can be combined with solar spectra to estimate inactivation rates in clear water under different sunlight conditions. This approach will be useful for modeling the inactivation of viruses and other microorganisms in sunlit natural and engineered systems.
’ INTRODUCTION The germicidal properties of natural sunlight and artificial light on animal viruses,1 plant viruses,2 bacterial viruses (bacteriophage),3 bacteria,3 and fungi4 are longstanding research topics. The biological response (i.e., persistence, inactivation, mutation) of an organism to ultraviolet (UV) light exposure over a range of wavelengths can be described by photoaction spectra3,5 and biological weighting functions.6 The earliest action spectra were obtained primarily in order to characterize the chemical and biological structures of microorganisms;7 researchers first determined that genes were composed of nucleic acids when action spectra for mutations in corn pollen, fungi, and viruses matched nucleic acid absorbance spectra.810 Because one of the absorption maxima of DNA occurs near 260 nm11 and because of the technical limitations of the light sources traditionally used, most action spectra studies have only utilized wavelengths spanning a portion of the UVC region (190280 nm). Action spectra (or BWFs) for solar UV wavelengths have been determined to quantify the effects of ozone depletion on biological processes, particularly photosynthesis.5,6 These functions were designed to quantify the impacts of changes in irradiance spectra on the relative rates of biological processes, rather than to predict or model the absolute rates of these processes. Understanding the role of sunlight in inactivating viruses and other microorganisms requires characterizing the effects of the different solar wavelengths that reach the surface of the earth, especially in the UVB (280320 nm) and UVA (320400 nm) regions. While action spectra for the loss of culturability of many organisms closely correspond to the absorption spectra of their genetic material in the UVC region, photoaction spectra for r 2011 American Chemical Society
sunlight inactivation in the UVB and UVA regions may be quite different.11,12 Longer wavelengths may damage organisms through a variety of mechanisms including protein damage13,14 and reactions with endogenous and exogenous sensitizers to form potentially harmful reactive oxygen species (ROS).15 Few studies have produced UVB and UVA action spectra for viruses.1618 In addition, out of all the previous studies of natural and simulated sunlight inactivation of viruses, only those of Sinton and colleagues19,20 measured and reported the spectrum of the light used, which is critical information for interpreting results. In this study we developed action spectra for one DNA and one RNA virus in clear water (no exogenous sensitizers) using polychromatic simulated sunlight. We modeled the viruses’ response to simulated sunlight to develop coefficients for estimating the sensitivity of each virus to wavelengths over the 280496 nm range with 3-nm resolution. These spectral sensitivity coefficients can be combined with measured or predicted sunlight intensity spectra to estimate inactivation rates under different sunlight conditions.
’ MATERIALS AND METHODS Viruses. MS2, a single stranded 3.6-kbp RNA bacteriophage, and PRD1, a double stranded 15-kbp DNA bacteriophage, were propagated in E. coli Famp (ATCC no. 700891) and in Received: June 1, 2011 Accepted: September 21, 2011 Revised: September 13, 2011 Published: September 21, 2011 9249
dx.doi.org/10.1021/es201875x | Environ. Sci. Technol. 2011, 45, 9249–9255
Environmental Science & Technology
ARTICLE
Table 1. Linear Regression Coefficients for MS2 and PRD1 for Each of Eight Simulated Sunlight Exposure Conditions MS2 (n = 3)
PRD1 (n = 3)
filter
sig difa
kobs (h1) ( stdev
R2 ( stdev
sig dif
kobs (h1) ( stdev
R2 ( stdev 0.986 ( 0.013
no filter
A
0.148 ( 0.004
0.985 ( 0.005
A
0.475 ( 0.043
f-280b
B
0.107 ( 0.009
0.986 ( 0.011
B
0.407 ( 0.007
0.996 ( 0.001
f-295
C
0.076 ( 0.005
0.979 ( 0.021
B
0.374 ( 0.041
0.992 ( 0.002
f-305
D
0.060 ( 0.003
0.856 ( 0.121
B
0.338 ( 0.038
0.997 ( 0.003
f-320
E
0.009 ( 0.004
0.489 ( 0.136
C
0.199 ( 0.025
0.997 ( 0.002
f-335
E
0.013 ( 0.006
0.808 ( 0.105
C/Dc
0.159 ( 0.005
0.995 ( 0.003
f-345
E
0.005 ( 0.002
0.983 ( 0.022
D
0.107 ( 0.008
0.983 ( 0.022
Dark
E
0.003 ( 0.001
0.634 ( 0.049
E
0.002 ( 0.005
0.429 ( 0.859
p < 0.05. b n = 2 for f-280 for MS2 and PRD1. c PRD1 inactivation with a WG320 filter was similar to the f-335 filter but different than f-345 filter, while the PRD1 inactivation rates with f-335 and f-345 filters were not different from each other. a
Salmonellatyphimurium LT2 (ATCC no. 19585), respectively, by broth enrichment.21 Bacteriophage and hosts were kindly provided by Prof. Mark Sobsey (University of North Carolina). Bacteriophage enrichments were purified as described in the Supporting Information (SI). Previous research suggests that these methods adequately remove broth photosensitizers capable of contributing to indirect inactivation of the phage.22 Bacteriophage plaque assays were performed using the double agar layer (DAL) method23 to titer stocks and to enumerate viruses after exposure to UV light as described in the SI. Action Spectrum Experimental Design. Purified MS2 and PRD1 were diluted together to titers of 106 PFU/mL in 100 mL phosphate buffered saline (PBS, 20 mM, pH 7.5) in 55 100 mm black-painted glass beakers (“reactors”). Samples were stirred magnetically and maintained at 20 °C in a water bath with a recirculating chiller (Thermo Electron). Sample beakers were (i) left uncovered; (ii) covered with 5 5 cm square glass optical cutoff filters [glass filters: Schott WG280 (“f-280”), Schott WG295 (“f-295”), Schott WG305 (“f-305”), Schott WG320 (“f-320”), Schott KG5 (“f-335”), and Kopp 9345 (“f-345”)]; or (iii) covered with aluminum foil for dark controls and exposed to simulated sunlight for 22 h. Subsamples were removed at 0, 2, 4, 6, 8, 12, 22 h and immediately frozen at 80 °C. Experiments were performed in triplicate over three consecutive days (each condition was tested in one reactor each day). The measured biological response to simulated sunlight was loss of culturability (ability to form plaques), which was quantified as described above. Glass filters had 50% transmittance values at wavelengths ranging from 280 to 350 nm (SI Figure S1 A). Filter transmittance spectra were recorded on a Perkin-Elmer Lambda 14 UV visible spectrophotometer (Waltham, MA), and lamp intensity spectra were adjusted for the absorbance of each filter as described in the SI. Solar Simulator. Samples were irradiated using an ozone-free 1000 W Xe arc lamp housed in an Oriel solar simulator (Oriel model # 911941000, Newport Co., Irvine, CA) which was fitted with an Oriel AM 1.5:G:A “global” filter and an atmospheric attenuation filter (Oriel part no. 81017, Newport Co.) to simulate a solar spectrum (SI Figure S1 B, no filter). Spectra were measured using calibrated portable UVvisible spectroradiometers (RPS 200 and RPS 380, International Light, Peabody, MA). Both the simulator configuration and spectral measurements are described in greater detail in the SI. Solar simulator output over the course of the experiments was constant at 277 W/m2summed over 280700 nm.
Model Development. Intensity Spectra. Each of the seven reactors was covered with a different filter (or none), and thus delivered a different light spectrum. The transmittance of each filter multiplied by the solar simulator irradiance at each wavelength is shown in SI Figure S1 B (the raw irradiance data are used for the no filter condition). These spectra represent the intensity to which organisms in the different reactors were exposed and were used as inputs for the numerical model. Inactivation curves. Virus inactivation was modeled using pseudo first-order kinetics (see below); inactivation rate constants (kij) were calculated as the negative slope of the linear regression lines (ln(N/No) vs time) for each combination of virus (i) and reactor (j).These kinetics were found to accurately describe the measured data (Table 1). Inactivation Model. Inactivation rate constants were modeled using eq 1:
kij ¼
Z∞
½Ij ðλÞ Pi ðλÞdλ
ð1Þ
0
Where Ij(λ) is the spectral irradiance of light (in W/m2 3 nm) delivered to reactor j and Pi(λ) is the spectral sensitivity coefficient, or the relative contribution (in m2/W 3 h) of photons at wavelength λ to the inactivation rate of organism i. Because measurements revealed that the solution absorbed less than 1% of light at all wavelengths of interest, and the reflection factor, divergence factor, and Petri factor24 were each found to differ from 1 by less than 5%, the simplifying assumption I(λ) = I0(λ) could be made. To further simplify calculations, spectra were discretized into 3 nm bins (chosen to provide sufficient resolution without requiring excessive computing time). Thus, Ij,w(λ) is the irradiance of light (in W/m2) entering reactor j integrated over a 3 nm wavelength range centered at λ, and Piw(λ) is the average spectral sensitivity coefficient, or the relative contribution (in m2/W 3 h) of irradiance in this range to the inactivation rate of organism i. Equation 1 thus becomes: kij ¼
λupper
∑ ½Ij, w ðλÞ Pwi ðλÞ Δλ
λlower
ð2Þ
This equation is analogous to a biological weighting function (BWF) as described by Cullen and Neale.6 An important difference is that eq 2 specifies a first-order rate constant with units of inverse time, which can be used to model the inactivation rate of viruses in sunlit waters. In contrast, the ratio of biologically 9250
dx.doi.org/10.1021/es201875x |Environ. Sci. Technol. 2011, 45, 9249–9255
Environmental Science & Technology effective to measured radiation described by previous studies5,6 is a relative measure, which is only useful if the reference condition is known. Computational Model. A computational model was developed in MATLAB (The MathWorks, Natick, MA) to characterize action spectra by calculating the spectral sensitivity coefficients Piw(λ) that best fit the measured inactivation rate constants kij (which represented the mean of triplicate inactivation trials) for each virus (i) in each of the seven reactors (j). The model was run 1000 times for each virus using random initial Piw(λ) guesses and the results were analyzed to obtain a single best-fit set of sensitivity coefficients [Piw(λ)] for the vector of central bin wavelengths (λ) for each organism (i). Briefly, the random initial Piw(λ) values were iteratively adjusted using a constrained nonlinear multivariable optimization function to minimize the sum of squared errors between observed and calculated kij values. Model development is described in detail and the MATLAB code is presented in the SI. Several previous studies have produced action spectra by assuming the shape can be described by an exponential or polynomial function.5,6 We chose not to restrict the shape to avoid potentially inaccurate assumptions about the phenomena of interest. An alternative computational method that avoids such assumptions is principle component analysis (PCA).6 We opted for an alternate optimization method designed to be less computationally intensive and facilitate multiple replicate simulations in a short period of time. Sensitivity Analysis. Several methods were employed to characterize the sensitivity of the model output to initial conditions and measured values. For each kij, alternative values were randomly chosen from within one standard deviation of that mean kij value. These alternative kij’ values were used in the computational model to generate 100 sets of alternative spectral sensitivity coefficients, and the median and mean average deviation of these results were plotted. Sensitivity analysis procedures are described in detail in the SI. Back-Testing. The computational approach was back-tested by generating 31 hypothetical sets of spectral sensitivity coefficients, each with a single peak centered at 250, 260, 270,...,550 nm. These hypothetical values were then used to calculate kij values using eq 1. These kij values were inserted into the computational model, which was solved 100 times using random initial values for Piw. The resulting “best fit datum” set of spectral sensitivity coefficients was compared to the hypothetical initial values to give an indication of the computational model’s useful working range with the given filter set and simulated sunlight spectrum. The errors obtained for these fits were plotted as a function of peak center wavelength. The relative standard deviations of light spectrum intensities across the 7 filter conditions were also plotted in the same figure, since increased variation in light conditions was expected to contribute to improved model resolution. Additionally, a 3-peak back-test was also conducted to determine whether multiple peaks disproportionately confounded the computational model. The effects of varying the distance between peaks and the location of the three-peak ensemble were studied. Finally, a back-test was conducted using a monotonically decreasing curve to assess the performance of the model in the absence of any input peaks. Photodamage Spectra. The spectral sensitivity coefficients calculated using the above model were multiplied by the irradiance values from the unfiltered simulated solar spectrum in order to demonstrate the relative contributions of different wavelengths of light to inactivation under typical sunlight conditions.
ARTICLE
Figure 1. Effect of simulated sunlight wavelength on (A) MS2 and PRD1 inactivation rate constant k(m2/MJ) and B) k normalized relative to full simulated sunlight (“no filter”). k’s from Table 1 were plotted against each of six sunlight filters 50% transmittance values in nm. k = ksun kdark and was the mean of n = 3 trials. Full simulated sunlight was represented at 270 nm at our discretion.
These irradiance-weighted spectral sensitivity coefficients are referred to below as “photodamage coefficients” (Di(λ) = I0(λ) Pi(λ)) for convenience. Literature Review. Virus action spectra were reviewed from nearly twenty publications since 1934. Two papers19,20 reported enough data to make direct comparisons with our work, and their data were captured using Graph Grabber software (http://www. quintessa.org/FreeSoftware/GraphGrabber/) and converted to inactivation rate constants (k) with units of m2/MJ (megajoule). Statistical Tests. Statistical tests were performed using MATLAB and Prism (GraphPad Software, La Jolla, CA). ANOVA and Tukey’s post-test were used to compare virus inactivation rate constants within a single virus type, and two-tailed t tests were used to compare between PRD1 and MS2 rate constants within filter treatments.
’ RESULTS Inactivation Rates of MS2 and PRD1. Virus inactivation roughly followed first-order kinetics for each filter condition (SI Figure S2). Within each virus type, there were significant differences among inactivation rate constants for the seven different filter conditions (p < 0.0001, ANOVA). However, Tukey’s post-test showed that MS2 inactivation rate constants for dark controls were not significantly different from k values for samples irradiated in reactors with f-320, f-335, or f-345 filters (Table 1, SI Figure S2). All other filters produced MS2 inactivation rate constants significantly different from the dark control (p < 0.05) and, as expected, filters transmitting increasing amounts of UVB light produced successively (and significantly) faster MS2 inactivation rates (Table 1). The rate constant for PRD1 dark controls was significantly different from that for each irradiated PRD1 sample (p < 0.05) 9251
dx.doi.org/10.1021/es201875x |Environ. Sci. Technol. 2011, 45, 9249–9255
Environmental Science & Technology
Figure 2. Effect of simulated sunlight wavelength on relative inactivation rate k normalized to full simulated sunlight for (A) DNA bacteriophage and (B) RNA bacteriophage in PBS (circles), seawater (squares), and river water (triangles). k = ksun kdark and was the mean of n = 3 trials for MS2 and PRD1 (n = 2 for f-280), and n = 1 F+ RNA and somatic coliphage. Inactivation rates for F+ RNA and somatic coliphage from Sinton and colleagues.19,20
(Table 1). No significant difference (p < 0.05) in PRD1 inactivation rate constants was observed among f-280, f-295, and f-305 filters, although they were each less effective than no filter at inactivating PRD1. Inactivation rates with the f-335 and f-345 filters were not significantly different from each other. MS2 and PRD1 linear regression R2 coefficients were high for most samples and were uniformly high for samples with inactivation rates more than two standard deviations above those of the dark controls (Table 1). There was no difference in the inactivation rates of MS2 and PRD1 in the dark controls (p = 0.7474, twotailed t test). Action Spectra of MS2 and PRD1. The rate of inactivation for both viruses decreased with increasing 50% cutoff filter wavelengths. PRD1 was inactivated faster than MS2 by simulated sunlight under each filter condition tested (p < 0.05 for each test, two-tailed t test) (Figure1A). For filters with 50% cutoff wavelengths in the UVB range, PRD1 was inactivated three to six times faster than MS2 (Figure1A). MS2 was not inactivated when filters with 50% cutoff wavelengths in the UVA range were used, while PRD1 was inactivated. Values of kobs were normalized relative to full simulated sunlight (i.e., no filter treatment) (Figure1B), emphasizing the fact that longer wavelengths contributed more to the inactivation of PRD1 than to that of MS2. Comparisons to Published Action Spectra. Normalized kobs values for MS2 and PRD1 were compared with extrapolated normalized kobs values from the literature.19,20 In our solar simulator, PRD1 had a normalized inactivation rate profile similar to the profiles observed for somatic coliphage naturally present in wastewater as extrapolated from the work conducted by Sinton and colleagues in natural sunlight19,20 (Figure2A).
ARTICLE
Figure 3. Calculated sensitivity coefficients for (A) MS2 and (B) PRD1 from 280 to 500 nm. Sensitivity coefficients (in m2/W 3 h) illustrate the contribution of a given irradiance (W/m2 3 nm) at each wavelength to the observed inactivation rate (1/h). Each figure represents the single best fit model solution.
Both PRD1 and somatic coliphage were sensitive to wavelengths in the UVA range. The relative inactivation rate profiles of F+ RNA coliphage, as extrapolated from the work of Sinton and colleagues,19,20 were similar to each other and both profiles were different from the profile of MS2 (a type of F+ RNA coliphage) observed in our study using simulated sunlight (Figure 2B). UVB wavelengths were alone responsible for the majority of the MS2 inactivation in our work, while Sinton et al.’s findings showed UVA and visible wavelengths were also important to F+ RNA coliphage inactivation. Computational Model for MS2 and PRD1 Inactivation. Spectral sensitivity coefficients (see eq 1) for MS2 and PRD1 were determined from the experimentally measured inactivation rate constants using a novel computational approach. Both MS2 and PRD1 were more sensitive to simulated sunlight at 280 nm (the shortest wavelength measured) than to other wavelengths, with sensitivity coefficients decreasing to a local minimum around 290295 nm (SI Table S1, Figure 3). At longer wavelengths, MS2 and PRD1 had peaks at approximately 305 nm, while PRD1 also exhibited a peak at approximately 350 nm; neither virus was sensitive to visible wavelengths in the PBS solution used for all experiments. A sensitivity analysis indicated that the peaks in virus sensitivity below 300 nm and at 305315 nm were robust to random perturbation, as indicated by the prevalence of nonzero values generated by the analysis in these wavelength regions (SI Figure S3 A, B). The third peak observed for PRD1 at approximately 350 nm was less robust, as an alternate model solution was common, in which the second two peaks were merged into a single peak (SI Figure S3 B). This peak is particularly difficult to verify because the optical cutoff filters used did not provide sufficient resolution at these longer wavelengths (SI Figure S8). 9252
dx.doi.org/10.1021/es201875x |Environ. Sci. Technol. 2011, 45, 9249–9255
Environmental Science & Technology
Figure 4. Calculated irradiance-weighted spectral sensitivity coefficients (photodamage coefficients: Di(λ) = I0(λ) Pi(λ)) for (A) MS2 and (B) PRD1 from 280 to 500 nm. Photodamage coefficients (in nm/h) illustrate the contribution of a given wavelength of typical simulated sunlight to the observed inactivation rate (in 1/h). Each figure represents the product of the unfiltered simulated solar irradiance spectrum and the lowest error single solution for the spectral sensitivity coefficients P(λ) generated by repeated model runs as presented in Figure 3. The total area under each curve is equal to the inactivation rate constant measured for that organism under the no filter condition (see eq 2).
Single-peak back-tests revealed relatively sound model performance at all wavelengths below 450 nm (SI Figure S4), while three-peak back-testing results demonstrated reasonable accuracy from 285 - 345 nm but not outside of that range (SI Figures S5, S6). The “no-peak” back-test produced reasonably accurate results over the 280496 nm range and did not produce significant artifactual peaks (SI Figure S7). It was encouraging that neither the sensitivity analysis nor the back-test validation procedures produced large “false peaks” under the conditions tested, although a small false “daughter” peak occurred next to a larger peak in one 3-peak back-test trial (SI Figure S6 C), and some instances of peaks merging (SI Figure S6 D) or being split in two (SI Figure S6 H) occurred at longer wavelengths (>345 nm). Two additional measures of model robustness, the sum of absolute errors from single peak back-tests and standard deviation of transmittance values among the seven filters were plotted as a function of wavelength (SI Figure S8). Error values were highest above 450 nm with a smaller peak at 290 nm, while transmittance variance was lowest below 300 nm. Irradiance-weighted photodamage spectra show the relative contributions of different wavelengths of light to inactivation under typical sunlight conditions. From SI Table S2 and Figure 4, it can be seen clearly that the sensitivity of PRD1 to longer wavelengths, which are present in much higher intensity in sunlight, results in an overall higher inactivation rate constant compared to MS2. The small peaks in photodamage coefficients observed at 380 nm for MS2 and approximately 420 nm for PRD1 were not found to be significantly different from the baseline in sensitivity analyses, and are estimated to account for less than 5% of inactivation under typical sunlight conditions. Thus, while it is not known whether these peaks are authentic or artifactual, they are likely to be negligible for most applications.
’ DISCUSSION Sensitivity of MS2 and PRD1 to Simulated Sunlight. PRD1 was found to be more sensitive to simulated sunlight than MS2 for all conditions studied, particularly to UVA light, which had
ARTICLE
little effect on MS2. The greater sensitivity of PRD1 is consistent with that phage’s larger genome, and is in agreement with most previous studies,22,25 although Hotze and colleagues26 found that MS2 was more sensitive than PRD1 to UVA light from a fluorescent UVA source. This discrepancy is puzzling, but may be due to differences in experimental methods and in the output spectra of the light sources used. Model-Derived Sensitivity Spectra. Published virus action spectra are typically characterized by peaks around 260 nm where DNA and RNA maximally absorb UV, followed by a steady decline in virus susceptibility up to approximately 300 nm, where most researchers stopped collecting data.27 We studied inactivation and sensitivity from 280 to 500 nm to explore the effects of all likely biocidal sunlight wavelengths on viruses. While the precise mechanisms of inactivation remain unknown, absorption of UVB and UVA photons by the two basic components of viruses, nucleic acids and proteins, may be a critical step in the inactivation of MS2 and PRD1 in PBS.27,28 These excited chromophores may undergo direct photolysis or react in aerobic solutions to form reactive oxygen species that damage other targets.27 A study on the inactivation of MS2 by UVC suggests that nucleic acids may photosensitize damage to proteins;14 such protein-genome interactions might play a similar role in UVBmediated damage. The observation that both MS2 and PRD1 were highly sensitive to the shortest simulated sunlight wavelengths (280 290 nm) is consistent with direct or indirect nucleic acid-sensitized damage. By contrast, the sensitivity peaks identified for both bacteriophages in the 305310 nm region (Figure3), while similar to a ∼313 nm shoulder in the UV sensitivity of T4 bacteriophage,18do not correspond to known peaks for DNA or RNA absorbance or photodamage. These peaks may represent absorbance by and damage to aromatic amino acids (e.g., tryptophan) or other protein components. Light at 254 nm can damage amino acid residues in the protein capsid of MS214 and UVB light might produce similar damage, affecting viruses’ capsid integrity or their ability to attach to, infect, or replicate within a host. While previous MS2 absorbance spectra did not reveal a peak near 305310 nm,29 nor did quantum yield data reveal a peak in that range for many viruses of interest,27 neither approach measured virus inactivation in the 305310 nm region. However, circular dichroism (CD) spectroscopy (a technique that measures protein folding and stability under stress) showed aromatic amino acid activity at 305310 nm for hepatitis C virus.30 Thus, spectra for photochemical activity and/or virus inactivation may differ from absorbance spectra.31 Sunlight absorption by and damage to viral nucleic acids and proteins should be measured in parallel with loss of infectivity to further elucidate the mechanisms of inactivation. The Role of Photosensitizers. We attempted to eliminate all sensitizers from our experimental solutions, whereas Sinton et al.’s work was performed in river water or seawater spiked with 23% (v/v) waste stabilization pond effluent or sewage,19,20 and thus very likely contained significant concentrations of photosensitizers. Our normalized MS2 inactivation rates were far lower than those of Sinton et al.’s F+ RNA coliphage (a family to which MS2 belongs), particularly at longer wavelengths (Figure 2B). Although biological differences may partly explain the variations in spectral response, a more likely explanation is that photons at longer wavelengths were absorbed by photosensitizers in Sinton et al.’s reactors, producing ROS such as singlet oxygen which subsequently damaged the coliphage.15,32,33 Interestingly, the 9253
dx.doi.org/10.1021/es201875x |Environ. Sci. Technol. 2011, 45, 9249–9255
Environmental Science & Technology normalized inactivation rates of PRD1 in our study were in good agreement with the rates for somatic coliphage reported by Sinton and colleagues (Figure 2A). This agreement may be coincidental, or may indicate similar spectral sensitivity of PRD1 and somatic coliphage to sunlight, and may likewise indicate that exogenous photosensitizers did not play a significant role in the inactivation of the latter variety of somatic coliphages. It should be noted that somatic coliphage are a diverse group, and variable response to sunlight has been documented in field isolates.22 Sensitivity Analysis. The results of the sensitivity analysis (SI Figure S3) and model back-testing (SI Figures S4S8) suggest that the computational model produced reasonable estimates of virus sensitivity to simulated sunlight over the 285345 nm range, and that the spectral sensitivity peaks observed at wavelengths <300 nm and between 305 and 310 nm are likely to be genuine, although the magnitudes predicted by the model may not be exact. For PRD1, an apparent peak at approximately 350 nm should be interpreted with some caution, as it falls outside the region over which the model could predict with confidence. It cannot be conclusively determined whether PRD1 has two distinct peaks at 305 and 350 nm (as our unperturbed model indicates) or a single, broader peak at slightly longer wavelengths (as the sensitivity analysis suggests). Nonetheless, the inactivation behavior of PRD1 under typical sunlight conditions would be quite similar for both sensitivity spectra. Future action spectrum experiments with filters providing greater resolution at wavelengths below 285 nm and above 345 nm would increase the power of the current method and its ability to characterize and resolve the sensitivity of viruses over a broader range of sunlight wavelengths. Furthermore, eliminating the atmospheric attenuation filter used in this trial could increase simulated sunlight intensity below 300 nm and increase resolution at the shortest UVB wavelengths. Advantages and Limitations of the Study Design and the Computational Model. Many action spectra have relied on monochrometers to produce narrow bands of light or simple optical filters to produce sharp cutoffs. When observed biological responses are attributed to the desired spectrum (i.e., the central wavelength of a monochromator slit or the wavelengths above a filter’s nominal cutoff), rather than to the entire spectrum transmitted, significant errors may occur. Furthermore, light sources with discrete emission bands such as mercury vapor lamps may introduce artifacts by delivering unrealistically high intensities at wavelengths of low organism sensitivity, while delivering little or no intensity at highly biocidal wavelengths. This study used a xenon arc lamp and optical filters to produce polychromatic light that was similar in intensity and spectral properties to natural sunlight. By taking advantage of the gradual cut-offs of optical filters and by modeling inactivation as a function of actual irradiances reaching target organisms, we were able to resolve detailed viral responses to simulated sunlight wavelengths. However, additional resolution was obtained at the cost of greater uncertainty. Sensitivity coefficients produced by the model are estimates based on optimizations of an under-determined system, rather than direct measurements. This uncertainty could be reduced by using greater numbers of filters with more diverse transmittance values over the wavelength ranges of interest. Furthermore, higher order terms could be included in future models to address possible synergistic effects between wavelengths. Finally, further work is required to assess the effects of using PBS for inactivation experiments. This buffer contains far more phosphate than natural waters and lacks divalent cations,
ARTICLE
and may thus affect the speciation of transition metals and the surface charge and aggregation of viruses in photoinactivation trials. Applications of Action Spectra Findings. Measuring the sensitivity of organisms to polychromatic light is critical for modeling sunlight-mediated inactivation in processes including solar water disinfection, wastewater stabilization pond operation, and the fate of pathogens in recreational waters. In clear waters, eq 2 and the sensitivity coefficients from Figure 3 can be used to estimate MS2 and PRD1 inactivation rate constants for any time of day, season, and latitude for which irradiance spectra can be measured or modeled. Accounting for light attenuation with depth would further allow the impacts of mixing and stratification on sunlight-mediated inactivation to be explored. Similar approaches are widely applied for estimating the photodegradation rates of chemicals in natural waters based on the quantum yield for the transformation of interest;34,35 the sensitivity coefficients for viruses, Pi(λ), are analogous to the product of a wavelength-specific quantum yield and the compound’s molar extinction coefficient. An important next step is to determine the sensitivity coefficients for enteric viruses of interest. For example, Adenovirus type 2 and Poliovirus type 3 were recently found to be significantly inactivated only in the presence of UVB wavelengths, suggesting that these viruses may have action spectra more similar to that of MS2 than PRD1.22 Further work is required to refine, validate, and apply our experimental approach and computational model. It should be noted that in waters with significant exogenous photosensitizers, additional inactivation mechanisms may occur,32 and eq 2 does not account for these. Validation under field conditions is also desirable, for example via focused studies similar to those reported by Boehm and colleagues.36
’ ASSOCIATED CONTENT
bS
Supporting Information. This section presents all supporting figures and tables referenced in the text as well as detailed descriptions of additional methods and the MATLAB code for the computational model. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 510.643.5023; fax: 510.642.7483; e-mail: nelson@ce. berkeley.edu. Present Addresses ‡
Center for a Livable Future, Bloomberg School of Public Health, Johns Hopkins, 615 N. Wolfe Street, W7007 Baltimore, MD 21205-2179.
Author Contributions †
Michael B. Fisher and David C. Love contributed equally to this work.
’ ACKNOWLEDGMENT This work was supported by an NSF CAREER/PECASE award to K.L.N. (BES-0239144) and by the U.C. Berkeley Blum Center for Developing Economies. ’ REFERENCES (1) Hollaender, A.; Oliphant, J. The inactivating effect of monochromatic ultraviolet radiation on influenza virus. J. Bacteriol. 1944, 48 (4), 447–454. 9254
dx.doi.org/10.1021/es201875x |Environ. Sci. Technol. 2011, 45, 9249–9255
Environmental Science & Technology (2) Hollaender, A.; Duggar, B. M. Irradiation of plant viruses and of micro€organisms with monochromatic light: III. Resistance of the virus of typical tobacco mosaic and Escherichia coli to radiation from lambda 3000 to lambda 2250 A. Proc. Natl. Acad. Sci. U.S.A. 1936, 22 (1), 19–24. (3) Gates, F. Results of irradiating Staphylococcus aureus bacteriophage with monochromatic ultraviolet light. J. Exp. Med. 1934, 60 (2), 179–188. (4) Hollaender, A.; Emmons, C. W. Wavelength dependence of mutation production in the ultraviolet with special emphasis on fungi. Cold Spring Harbor Symp. Quant. Biol. 1941, 9, 179–186. (5) Rundel, R. D. Action spectra and estimation of biologically effective UV radiation. Physiol. Plant. 1983, 58, 380–386. (6) Cullen, J. J.; Neale, P. J., Biological weighting functions for describing the effects of ultraviolet radiation on aquatic systems. In Effects of Ozone Depletion on Aquatic Ecosystems; Haeder, D. P., Ed.; R.G. Landes: Austin, TX, 1996; pp 97118. (7) Powell, W.; Setlow, R. The effect of monochromatic ultraviolet radiation on the interfering property of influenza virus. Virology 1956, 2 (3), 337–43. (8) Hollaender, A.; Emmons, C. Wavelength dependence of mutation production in ultraviolet with special emphasis on fungi. Cold Spring Harbor Symp. Quant. Biol. 1941, 9, 179–186. (9) Rivers, T.; Gates, F. Ultraviolet light and vaccinia virus. II. The effect of monochromatic ultraviolet light upon vaccine virus. J. Exp. Med. 1928, 47, 45–49. (10) Stadler, L. J.; Uber, F. M. Genetic effects of ultraviolet radiation in maize. IV. Comparison of monochromatic radiations. Genetics 1942, 27 (1), 84–118. (11) Jagger, J., Solar-UV actions on living cells. 1985. (12) Coohill, T. Action spectra again? Photochem. Photobiol. 1991, 54 (5), 859–870. (13) Eischeid, A. C.; Meyer, J. N.; Linden, K. G. UV disinfection of adenoviruses: Molecular indications of DNA damage efficiency. Appl. Environ. Microbiol. 2009, 75 (1), 23–28. (14) Wigginton, K. R.; Menin, L.; Montoya, J. P.; Kohn, T. Oxidation of virus proteins during UV254 and singlet oxygen mediated inactivation. Environ. Sci. Technol. 2010, 44 (14), 5437–5443. (15) Kohn, T.; Nelson, K. L. Sunlight-mediated inactivation of MS2 coliphage via exogenous singlet oxygen produced by sensitizers in natural waters. Environ. Sci. Technol. 2007, 41 (1), 192–7. (16) Peak, M. J.; Peak, J. G. Action spectra for the ultraviolet and visible light inactivation of phage T7: Effect of host-cell reactivation. Radiat. Res. 1978, 76 (2), 325–30. (17) Ronto, G.; Gaspar, S.; Berces, A. Phage T7 in biological UV dose measurement. J Photochem. Photobiol. B, Biol. 1992, 12 (3), 285–94. (18) Tyrrell, R. M. Solar dosimetry with repair deficient bacterial spores: action spectra, photoproduct measurements and a comparison with other biological systems. Photochem. Photobiol. 1978, 27 (5), 571–9. (19) Sinton, L.; Hall, C.; Lynch, P.; Davies-Colley, R. Sunlight inactivation of fecal indicator bacteria and bacteriophages from waste stabilization pond effluent in fresh and saline waters. Appl. Environ. Microbiol. 2002, 68 (3), 1122–1131. (20) Sinton, L. W.; Finlay, R. K.; Lynch, P. A. Sunlight inactivation of fecal bacteriophages and bacteria in sewage-polluted seawater. Appl. Environ. Microbiol. 1999, 65 (8), 3605–13. (21) Method 1601: Male-Specific (F+) and Somatic Coliphage in Water by Two-Step Enrichment Procedure, 821-R-01-030; U.S. Environmental Protection Agency: Washington, DC, . 2001; p 25. (22) Love, D. C. S.; Andrea, C.; Nelson, Kara L. Human virus and bacteriophage inactivation in clear water by simulated sunlight compared to bacteriophage inactivation at a southern California beach. Environ. Sci. Technol. 2010, 44 (18), 6965–6970. (23) Adams, M. H., Bacteriophages; Interscience Publishers: New York, 1959. (24) Bolton, J. R.; Linden, K. G. Standardization of methods for fluence (UV dose) determination in bench-scale UV experiments. J. Environ. Eng. 2003, 129 (3), 209–216.
ARTICLE
(25) Lytle, C. D.; Sagripanti, J. L. Predicted inactivation of viruses of relevance to biodefense by solar radiation. J. Virol. 2005, 79 (22), 14244–52. (26) Hotze, E. M.; Badireddy, A. R.; Chellam, S.; Wiesner, M. R. Mechanisms of bacteriophage inactivation via singlet oxygen generation in UV illuminated fullerol suspensions. Environ. Sci. Technol. 2009, 43 (17), 6639–45. (27) Rauth, A. The physical state of viral nucleic acid and the sensitivity of viruses to ultraviolet light. Biophys. J. 1965, 5, 257–273. (28) Chen, R. Z.; Craik, S. A.; Bolton, J. R. Comparison of the action spectra and relative DNA absorption spectra of microorganisms: Information important for the determination of germicidal fluence (UV dose) in an ultraviolet disinfection of water. Water Res. 2009, 43, 5087–5096. (29) Johnson, H. R.; Hooker, J. M.; Francis, M. B.; Clark, D. S. Solubilization and stabilization of bacteriophage MS2 in organic solvents. Biotechnol. Bioeng. 2007, 97 (2), 224–34. (30) Kunkel, M.; Watowich, S. J. Biophysical characterization of hepatitis C virus core protein: Implications for interactions within the virus and host. FEBS Lett. 2004, 557 (13), 174–80. (31) Chen, R. Z.; Craik, S. A.; Bolton, J. R. Comparison of the action spectra and relative DNA absorbance spectra of microorganisms: Information important for the determination of germicidal fluence (UV dose) in an ultraviolet disinfection of water. Water Res. 2009, 43 (20), 5087–5096. (32) Kohn, T.; Grandbois, M.; McNeill, K.; Nelson, K. L. Association with natural organic matter enhances the sunlight-mediated inactivation of MS2 coliphage by singlet oxygen. Environ. Sci. Technol. 2007, 41 (13), 4626–4632. (33) Davies-Colley, R. J.; Donnison, A. M.; Speed, D. J.; Ross, C. M.; Nagels, J. W. Inactivation of faecal indicator microorganisms in waste stabilisation ponds: Interactions of environmental factors with sunlight. Water Res. 1999, 33 (5), 1220–1230. (34) Schwartzenbach, R. P. G., P.M.; Imboden, D. M., Environmental Organic Chemistry, 2nd ed.; John Wiley and Sons: Hoboken, NJ, 2003. (35) Zepp, R. G.; Cline, D. M. Rates of direct photolysis in aquatic environment. Environ. Sci. Technol. 1977, 11 (4), 359–366. (36) Boehm, A. B.; Yamahara, K. M.; Love, D. C.; Peterson, B. M.; McNeill, K.; Nelson, K. L. Covariation and photoinactivation of traditional and novel indicator organisms and human viruses at a sewageimpacted marine beach. Environ. Sci. Technol. 2009, 43 (21), 8046–52.
9255
dx.doi.org/10.1021/es201875x |Environ. Sci. Technol. 2011, 45, 9249–9255
ARTICLE pubs.acs.org/est
Development of a Novel Bioelectrochemical Membrane Reactor for Wastewater Treatment Yun-Kun Wang, Guo-Ping Sheng,* Wen-Wei Li, Yu-Xi Huang, Yang-Yang Yu, Raymond J. Zeng, and Han-Qing Yu Department of Chemistry, University of Science & Technology of China, Hefei 230026, China ABSTRACT: A novel bioelectrochemical membrane reactor (BEMR), which takes advantage of a membrane bioreactor (MBR) and microbial fuel cells (MFC), is developed for wastewater treatment and energy recovery. In this system, stainless steel mesh with biofilm formed on it serves as both the cathode and the filtration material. Oxygen reduction reactions are effectively catalyzed by the microorganisms attached on the mesh. The effluent turbidity from the BEMR system was low during most of the operation period, and the chemical oxygen demand and NH4+-N removal efficiencies averaged 92.4% and 95.6%, respectively. With an increase in hydraulic retention time and a decrease in loading rate, the system performance was enhanced. In this BEMR process, a maximum power density of 4.35 W/m3 and a current density of 18.32 A/m3 were obtained at a hydraulic retention time of 150 min and external resister of 100 Ω. The Coulombic efficiency was 8.2%. Though the power density and current density of the BEMR system were not very high, compared with other high-output MFC systems, electricity recovery could be further enhanced through optimizing the operation conditions and BEMR configurations. Results clearly indicate that this innovative system holds great promise for efficient treatment of wastewater and energy recovery.
’ INTRODUCTION In past decades, the membrane bioreactor (MBR) process, which integrates a filtration membrane with a bioreactor,1 has gained worldwide attention and popularity due to its high treatment efficiency, low sludge production, and good effluent quality.2 However, there are still some problems to be solved before its more widespread application, such as high costs of membrane materials, severe membrane fouling, and high energy consumption for aeration. Utilization of cheap coarse-pore mesh, as an alternative to conventional microfiltration/ultrafiltration membrane,3 5 may lower the MBR construction costs, increase the economic viability, and promote the application of such processes.3,6 Expectedly, this process may go a step forward if the energy contained in wastewater can be recovered to partially offset the energy consumption for aeration. Microbial fuel cells (MFCs) have emerged as a promising technology to recover energy from wastewater.7 Exciting progress has been made in MFC construction and electricity generation over the past decade. However, for wastewater treatment, MFCs usually have poor effluent quality and low treatment efficiency because of their limited biomass retention,8 which necessitates a further treatment with additional operational costs. It has been hypothesized that a combined MBR MFC system might offer an attractive option.8 MBR operation may overcome the drawbacks of MFC by improving biomass retention and chemical oxygen demand (COD) removal efficiency, while MFC may generate power to partially offset the energy demand for aeration and filtration in MBR and lower the overall oxygen consumption for COD removal. r 2011 American Chemical Society
However, effective integration of the two reactors is a challenge. It has been proposed that an MFC can be connected with a bioreactor as a pre- or posttreatment step9,10 or be directly immersed into bioreactors to recover electricity from wastewater.11 13 For example, anaerobic bioreactor MFC aerobic bioreactor, or MFCs with various configurations submerged into various bioreactors, have been reported for bioelectricity generation and wastewater treatment. However, these processes simply connect two individual reactors in sequence without truly integrating them. This increases the operation complexity and limits the contribution of MFC to the combined system. In addition, the performance of such combined systems is not always satisfactory.12 Thus, novel hybrid processes with effective integrated configuration are needed, and we expect that a more appropriate integration of MBR and MFC may provide an efficient way to meet the dual goals of wastewater treatment and energy recovery. Therefore, in this work a novel bioelectrochemical membrane reactor (BEMR) was developed for wastewater treatment and energy recovery. The system performance in terms of both power generation and nutrient removal were investigated. The effects of main operating parameters, including hydraulic retention time (HRT), external resistor, and volume Received: June 10, 2011 Accepted: September 20, 2011 Revised: September 17, 2011 Published: October 06, 2011 9256
dx.doi.org/10.1021/es2019803 | Environ. Sci. Technol. 2011, 45, 9256–9261
Environmental Science & Technology
ARTICLE
loading rate (VLR), on the performance of the system were evaluated.
’ MATERIALS AND METHODS BEMR Assembly. The setup of BEMR is illustrated in Figure 1. The tubular cathodic and the anodic chambers were separated by a nonwoven cloth (400 g/m2) supported with a Plexiglas tube, which was perforated to facilitate mass transfer. Before use, the nonwoven cloth was pretreated with poly(tetrafluoroethylene) to prevent water leakage according to previous studies.13,14 The spacing between cathode and Plexiglas tube was 3.9 cm (Figure 1A). Silicone tubes were installed on both ends of the Plexiglas tube to enable a continuous flow. The anodic chamber was filled with granular graphite (Sanye Carbon Co.; 3 5 mm diameter) with a graphite rod (Sanye Carbon Co.; 6 mm diameter) inserted. The total volume of the anodic chamber was 210 mL, while the net effective volume was 109 mL. A stainless steel mesh with pore size 40 μm was used as the cathode without pretreatment (Huayang Ironware Co.). The assembly process of the cathode is shown in Figure 1B. The total surface area of the cathode was 494 cm2. The electrodes were connected to the circuit with copper wires across an external resistance. The copper wires exposed to the solution were sealed with epoxy to avoid metal corrosion in long-time operation. The electrode assembly was submerged in a column-type reactor (50 cm height, 10.4 cm diameter, 2.3 L working volume), which acted as a cathodic chamber (Figure 1A). The reactor was operated in continuousflow mode. Fine air bubbles were supplied for aeration through a dispenser at the reactor bottom, and the dissolved oxygen (DO) concentration was controlled at 4 5 mg/L. Inoculation and Operation Conditions. The anodic chamber was inoculated with 20 mL of concentrated anaerobic sludge from a laboratory-scale upflow anaerobic sludge blanket reactor, with a mixed liquor suspended solid (MLSS) concentration of 10 g/L. Synthetic wastewater was continuously fed into the anodic chamber through a peristaltic pump (Lange Co.). The cathodic chamber was seeded with activated sludge from the secondary clarifier of a municipal wastewater treatment plant in Hefei, China. The initial MLSS concentration in the reactor was 4.5 g/L, and during the operation no excess sludge was discharged. The composition of the synthetic wastewater was: CH3COONa 3 3H2O, 0.64 g/L; NH4Cl, 57 mg/L; K2HPO4 3 3H2O, 22 mg/L; CaCl2, 11.5 mg/L; MgSO4 12 mg/L; and 10 mL of trace element solution. The synthetic wastewater was directly pumped into the anodic chamber, and the effluent from the anode then flew into the cathodic chamber. After filtration through the stainless steel mesh, the effluent was finally discharged from the system. The system performance under various operating conditions was also evaluated, and the experimental design is summarized in Table 1. In the system operation, the permeation flux was kept constant, and the trans-membrane pressure (TMP) across the steel mesh was measured through the difference in water level, which was monitored every 20 min by a pressure transmitter (LD187, Leide Electronic Ltd.). After a long-term operation, the biofilm attached on steel mesh would become thick, and the TMP would increase sharply. Thus, off-line backwashing was periodically carried out to remove the overgrown biofilm. Analysis and Calculations. Dissolved oxygen was monitored with a DO meter (HQ 30d, Hach Co.). COD, NH4+-N, TN and turbidity were measured by APHA standard methods.15
Figure 1. (A) Schematic of BEMR system and (B) procedures for BEMR assembly. (1) Anodic chamber; (2) effluent from anodic chamber; (3) cathodic chamber; (4) hydraulic head; (5) biofilm; (6) stainless steel mesh; (7) nonwoven cloth separator; (8) pressure transmitter.
Electrochemical analysis (cyclic voltammetry, CV) of stainless steel mesh cathode with and without biofilm was performed by use of a CHI660C electrochemical workstation (CH Instruments, Chenhua Instrument Co.) with a three-electrode cell. The stainless steel mesh served as the working electrode, whereas a Ag/ AgCl (3 M KCl; 0.205 V vs standard hydrogen electrode, SHE) electrode and a Pt wire were used as the reference electrode and counterelectrode, respectively. The working electrode with the surface area of 0.64 cm2 was cut from the BEMR after 139 days of operation. CV scanning was conducted between 1.0 and 0.6 V at a scan rate of 1 mV/s. Synthetic wastewater was used as the electrolyte. All the measurements were conducted at ambient temperatures (25 ( 1 °C) with a nitrogen or oxygen saturation time of 30 min. The voltage across resistor was automatically collected every 20 min by use of a data acquisition system (USB2801, ATD Co.). Linear sweep voltammetry was performed to determine polarization curves by use of a CHI660C electrochemical workstation, as reported elsewhere.16,17 First, the circuit of the BEMR was opened for 12 h to measure the open circuit voltage. Then, voltammetry scanning was performed with the anode as the working electrode, the cathode as the counterelectrode, and a Ag/AgCl electrode as reference electrode. The scan range was set between zero and the open-circuit voltage, while the scan rate was set at 1 mV/s. Current (I) was calculated according to Ohm’s law (I = V/R). Power (P) was obtained as P = IV. Coulombic efficiency (CE) was calculated as CE = Cp/Cth 100, where Cp (C) is the total coulombs calculated by integrating the current over time, and Cth is the theoretical amount of coulombs available on the basis of the COD removed in the anodic chamber. The current density and power density were normalized to the total anodic chamber volume. The HRT of the anodic chamber was calculated from the net effective volume of the anodic chamber and the influent flow rate. 9257
dx.doi.org/10.1021/es2019803 |Environ. Sci. Technol. 2011, 45, 9256–9261
Environmental Science & Technology
ARTICLE
Table 1. Experimental Parameters of the System in Operation run
operation time (days)
HRT (min)
external resistance (Ω)
inflow COD (mg/L)
volume loading rate [kg/(m3 3 day)]
1
23 38
14.5
500
439.1 ( 60.2
22.6
2
39 40
14.5
500
253.6 ( 10.9
13.1
3
41 47
14.5
500
126.6 ( 13.4
6.5
4
48 81
29
500
160.7 ( 22.2
4.2
5
82 112
29
500
280.5 ( 35.1
7.2
6
113 139
29
100
295.3 ( 20.4
7.6
7
140 144
150
100
234.2 ( 14.0
1.2
The composition and architecture of the biofilm on the stainless steel were analyzed by confocal laser scanning microscopy (CLSM) (FV1000, Olympus Co., Japan). Fluorescent staining was conducted as described by Chen et al.18 Major fluorescence probes including Syto 63, fluorescein isothiocyanate (FITC), and Calcofluor white (all purchased from Sigma Inc.) were used to simultaneously probe the cells (total and dead), proteins, and βpolysaccharides, respectively.
’ RESULTS AND DISCUSSION Performance of the Bioelectrochemical Membrane Reactor. The initial rejection performance of stainless steel in the
system was assessed in terms of effluent turbidity from the BEMR system. As shown in Figure 2, the effluent turbidity dropped significantly to about 0.8 nephelometric turbidity unit (NTU) within 1 day and remained constant after that, implying the rapid formation of biofilm on the mesh surface. This biofilm effectively rejected the small flocs, leading to a low effluent turbidity. Compared with the conventional activated sludge without membrane, flocs in the MBR process are generally smaller in size, mostly below 60 μm, and over 50% are in a range of 0.1 10 μm.19 In this BEMR system, the cathode mesh showed a good capability to reject the fine sludge flocs in the stable operation, mainly attributed to the formation of three-dimensional biofilm structure on the mesh surface.20 For wastewater treatment systems, it is important to evaluate their COD and nutrient removal efficiencies. Figure 3 reveals that the BEMR system had good COD and NH4+-N removal efficiencies during the operation. The removal efficiencies of COD and NH4+-N in the BEMR system were comparable to those of MBRs using microfiltration or ultrafiltration membranes and coarse mesh.21 23 The average effluent concentration and removal efficiency of COD were 21.5 ( 8.4 mg/L and 92.4%, respectively, while those of NH4+-N were 0.61 ( 0.74 mg/L and 95.6%, respectively. Approximately 5.5 78% of influent COD had been already removed in the anodic compartment and 13.4 86.4% of COD had been oxidized in the cathodic compartment for the BEMR (Table 2). However, only a small portion (0.51 8.2%) of electrons from acetate oxidation was transferred to the electrode. COD removal in the cathodic chamber could be attributed mainly to the respiration of heterotrophic bacteria. NH4+-N was mainly removed in the cathodic compartment. In the BEMR operation, the total nitrogen removal efficiency reached 27.6 60.7%, implying that denitrification occurred in the cathodic chamber.24 The electrochemical performance of the system at varied HRTs and external resistances is shown in Figure 4. The experiments were carried out after more than 20 days of enrichment of electricity-producing microbes. During runs 1 3, the MFC was
Figure 2. Profile of the BEMR system effluent turbidity over operating time.
Figure 3. Removal of (A) COD and (B) NH4+-N by the BEMR system.
operated at a relatively short HRT of 14.5 min, and VLR decreased from 22.6 to 6.5 kg COD/(m3 3 day) in steps. A maximum voltage of 226 mV and a high current density of 2.15 A/m3 were achieved at this stage. A longer HRT of 29 min was adopted during runs 4 6 (days 48 139), and the maximum voltage increased to 515 mV, with the current density further raised to 4.90 A/m3. After the external resister was changed from 500 to 100 Ω in run 6 (days 113 139), the maximum voltage declined to 251 mV, but the current density peaked at 11.96 A/m3. In this BEMR process, a maximum power density of 4.35 W/m3 and a current density of 18.32 A/m3 were obtained at HRT of 150 min and 9258
dx.doi.org/10.1021/es2019803 |Environ. Sci. Technol. 2011, 45, 9256–9261
Environmental Science & Technology
ARTICLE
Table 2. System Performance during Different Experimental Periods run
Coulombic
COD removal efficiency
COD removal efficiency
NH4+-N removal efficiency
TN removal efficiency
efficiency (%)
by anode (%)
by BEMR (%)
by BEMR (%)
by BEMR (%)
1
0.51
9.8 ( 4.7
90.7 ( 6.6
92.7 ( 6.0
29.3 ( 4.3
2
1.79
5.5 ( 4.8
91.9 ( 0.1
95.8 ( 1.2
27.6 ( 6.2
3
2.69
6.6 ( 5.0
82.5 ( 2.4
94.9 ( 0.9
31.2 ( 7.3
4
2.87
19.2 ( 11.6
86.6 ( 2.9
97.8 ( 0.9
37.0 ( 11.2
5
2.31
15.7 ( 15.3
93.6 ( 2.5
95.8 ( 5.4
48.8 ( 14.0
6
4.24
21.9 ( 8.4
93.9 ( 2.7
94.2 ( 4.9
49.5 ( 15.6
7
8.2
78.0 ( 3.8
91.4 ( 4.0
99.6 ( 0.2
60.7 ( 3.0
Figure 4. Variations of voltage and TMP during the long-term operation.
external resister of 100 Ω (run 7). Because of the differences in configurations, substrate type, and loading rate, etc., the power generation of MFC reported in literature varies significantly. Compared with other MFC studies,25 the power density of the BEMR system was not high. When MFCs were submerged into bioreactors for wastewater treatment, a maximum power density of 0.58 16.7 W/m3 was obtained.11 13 These results indicate that the MFC performance in terms of electricity recovery could be further enhanced through optimizing its operating conditions. Factors Affecting BEMR System Performance. During BEMR operation, the effects of HRT, external resistor, and VLR on the system electrochemical performance were evaluated to identify the key factors affecting BEMR performance. HRT affected the electricity generation of the MFC working in continuous mode. With an increase in HRT, the CE, power density, and COD removal efficiency in the anodic chamber increased (Table 2). As the HRT was increased to 2.5 h, about 78% of COD could be removed in the anode chamber and 8.2% of electrons from acetate oxidation were transferred to the electrode, and the maximum power density of 4.35 W/m3 was obtained (run 7). This indicates that a long HRT benefited the growth of electricitygenerating bacteria in the anode chamber and thus resulted in increased COD removal, CE, and power produced. External resistance also influences the electricity generation of the BEMR system. When the external resistance was decreased from 500 to 100 Ω, the current density was doubled (Figure 4), and the COD removal efficiency and CE in the anodic chamber were enhanced (Table 2). These results are in agreement with a previous study.26 This suggests that a low external resistor benefits electron transfer by bacteria to electrode, thus increasing the electrical generation efficiency. The electrochemical performance of the BEMR system was also affected by VLR. When the VLR was decreased, both current
Figure 5. (A) Power output and (B) polarization curves of the biocathode MFC under different operating conditions.
density and CE increased (Figure 4 and Table 2). With a further decrease in VLR to 1.2 kg COD/(m3 3 day), a high CE value of 8.2% could be obtained (Table 2). To further examine the effect of VLR, the polarization curves in runs 4 and 5 were measured (Figure 5). Results show that a higher open-circuit voltage (650 mV) and power density (3.15 W/m3) were obtained at a lower VLR of 4.2 kg COD/(m3 3 day). This indicated that low loading rate might benefit electricity production in the BEMR system. Biofilm Formation and Electrochemical Activity. After operation, a uniform and dense layer of sludge cake was formed on the mesh surface. In this BEMR system, the stainless steel mesh with the biofilm formed on it plays the dual roles of filter and biocathode. It has been demonstrated that different types of stainless steel mesh in terms of wire diameter, pore size, and open area percentage resulted in different MFC performance.27 This was attributed to the fact that the characteristics of stainless steel mesh could affect biofilm development. To get a deep insight into the specific composition of the biofilm, a CLSM imaging analysis was performed. Figure 6 shows that bacteria and biopolyers were abundant in the biofilm. Bacteria (red), β-polysaccharides (blue), and some proteins (green) were present on the mesh surface (Figure 6D). Both bacteria and biopolymers were found to coexist or overlap on the mesh surface. Biopolymers can be easily adsorbed onto the mesh surface and adhered 9259
dx.doi.org/10.1021/es2019803 |Environ. Sci. Technol. 2011, 45, 9256–9261
Environmental Science & Technology
Figure 6. Morphology and CLSM images of stainless steel mesh surfaces. (A) Raw mesh. (B D) CLSM images of the surface of stainless steel mesh after 139 days of operation: (B) phase-contrast photograph; (C) CLSM image of microbial cells (Syto 63); and (D) combined image of proteins (stained with FITC), β-polysaccharides (stained with Calcofluor white), and microbial cells (stained with Syto 63). Red, microbial cells; green, proteins; blue, β-polysaccharides.
Figure 7. Cyclic voltammetry curves for (A) raw stainless steel mesh electrode and (B) stainless steel mesh electrode with biofilm attached.
on bacterial cells, which might facilitate the initial sludge deposition, increase the fouling resistance, and contribute to the irreversible fouling of mesh after long-term operation.28 CV was performed to examine the catalytic capability of the biofilm on cathode surface. As shown in Figure 7A, the midpoint potential of the faradaic current was about 0.42 V. At this potential, the faradaic current was higher in oxygen-saturation solution than that in nitrogen-saturation solution, because of the oxygen
ARTICLE
reduction on the electrode. In order to further ascertain the biofilm-associated electron transfer, CV was performed for a used stainless steel mesh cathode under the same conditions. As shown in Figure 7B, an obvious reduction peak at 0.275 V was found for the steel mesh with biofilm formed. However, the peak disappeared when the solution was purged with nitrogen. The absence of the redox peak implied a direct biofilm-mediated electrode reduction for oxygen.29 The electricity generation performance was closely associated with the formation and detachment of biofilm, which drive catalysis of oxygen reduction. After physical cleaning of the stainless steel mesh, the TMP initially decreased to a low value and then increased slightly to a steady state due to formation of biofilm on the mesh. Accordingly, the output voltage also increased and then became stable. This positive correlation between TMP and output voltage was attributed to that the microorganisms attached to the cathode steel mesh surface could promote oxygen reduction at the cathode and meanwhile increased the TMP.12 However, with further development of biofilm, the TMP continued to increase sharply, but the output voltage dropped slightly (Figure 4). The reason might be that the thickly developed biofilm started to significantly limit oxygen transfer and lower the electricity production. The oxygen concentration decreased rapidly inside the biofilm.30 It was reported that DO was depleted at a depth of 1.5 2.0 mm in biofilm.5 Thus, oxygen transfer at the cathode of this BEMR might also be limited by formation of a thick biofilm. Therefore, removal of the overgrowth biofilm by physical cleaning is necessary. Significance of This Work. Here we developed a novel BEMR system, in which stainless steel mesh with biofilm formed on it were served as the filtration material and cathode, and nonwoven cloth was the separator between electrodes. The oxygen demand for COD removal and excess sludge production could be reduced because of the electricity generation from wastewater and the anaerobic operation of the anode chamber. Furthermore, the system shows excellent COD and NH4+-N removal efficiencies and enables direct harvest of electricity from wastewater. More importantly, compared to other MFC configurations reported,11 13 high-quality effluent was obtained in our study. This is mainly attributed to the biofilm formation on mesh, which not only improved sludge retention but also efficiently catalyzed oxygen reduction. The BEMR system integrates the advantages of both MBR for nutrient removal and MFC for energy recovery, which presents a promising system for wastewater treatment and energy recovery processes. In this study, effects of the main operation parameters on the system performances were evaluated. This would be a useful reference for wide application of the new process. Nevertheless, the electricity generation capacity and energy recovery efficiency of the BEMR system is still limited. To make this novel system more effective and applicable, more studies are required to further elucidate the biofilm-driven catalysis mechanisms and to optimize the electrode materials and MFC construction.
’ AUTHOR INFORMATION Corresponding Author
*Fax: +86-551-3601592; e-mail:
[email protected].
’ ACKNOWLEDGMENT We thank the CAS (KSCX2-YW-G-055), the National HiTechnology Development 863 Program of China (2011AA060907), 9260
dx.doi.org/10.1021/es2019803 |Environ. Sci. Technol. 2011, 45, 9256–9261
Environmental Science & Technology the NSFC (51008290), and the Key Special Program on the S&T for the Pollution Control and Treatment of Water Bodies (2008ZX07010-003) for partial support of this study.
’ REFERENCES (1) Williams, M. D.; Pirbazari, M. Membrane bioreactor process for removing biodegradable organic matter from water. Water Res. 2007, 41, 3880–3893. (2) Wang, Z.; Wu, Z.; Yu, G.; Liu, J.; Zhou, Z. Relationship between sludge characteristics and membrane flux determination in submerged membrane bioreactors. J. Membr. Sci. 2006, 284, 87–94. (3) Fan, B.; Huang, X. Characteristics of a self-forming dynamic membrane coupled with a bioreactor for municipal wastewater treatment. Environ. Sci. Technol. 2002, 36, 5245–5251. (4) Chu, H. Q.; Cao, D. W.; Jin, W.; Dong, B. Z. Characteristics of bio-diatomite dynamic membrane process for municipal wastewater treatment. J. Membr. Sci. 2008, 325, 271–276. (5) Zhou, X. H.; Shi, H. C.; Cai, Q.; He, M.; Wu, Y. X. Function of self-forming dynamic membrane and biokinetic parameters’ determination by microelectrode. Water Res. 2008, 42, 2369–2376. (6) Okamura, D.; Mori, Y.; Hashimoto, T.; Hori, K. Effects of microbial degradation of biofoulants on microfiltration membrane performance in a membrane bioreactor. Environ. Sci. Technol. 2010, 44, 8644–8648. (7) Lovley, D. R. The microbe electric: conversion of organic matter to electricity. Curr. Opin. Biotechnol. 2008, 19, 564–571. (8) Logan, B. E., Microbial fuel cells; Wiley-Interscience: Hoboken, NJ., 2008; pp 149 154. (9) Zhang, B.; Zhao, H.; Zhou, S.; Shi, C.; Wang, C.; Ni, J. A novel UASB-MFC-BAF integrated system for high strength molasses wastewater treatment and bioelectricity generation. Bioresour. Technol. 2009, 100, 5687–5693. (10) Cheng, J.; Zhu, X.; Ni, J.; Borthwick, A. Palm oil mill effluent treatment using a two-stage microbial fuel cells system integrated with immobilized biological aerated filters. Bioresour. Technol. 2010, 101, 2729–2734. (11) Min, B.; Angelidaki, I. Innovative microbial fuel cell for electricity production from anaerobic reactors. J. Power Sources 2008, 180, 641–647. (12) Cha, J.; Choi, S.; Yu, H.; Kim, H.; Kim, C. Directly applicable microbial fuel cells in aeration tank for wastewater treatment. Bioelectrochemistry 2010, 78, 72–79. (13) Liu, X. W.; Wang, Y. P.; Huang, Y. X.; Sun, X. F.; Sheng, G. P.; Zeng, R. J.; Li, F.; Dong, F.; Wang, S. G.; Tong, Z. H.; Yu, H. Q. Integration of a microbial fuel cell with activated sludge process for energy-saving wastewater treatment: Taking a sequencing batch reactor as an example. Biotechnol. Bioeng. 2011, 108, 1260–1267. (14) Logan, B. E.; Cheng, S.; Liu, H. Power densities using different cathode catalysts (Pt and CoTMPP) and polymer binders (Nafion and PTFE) in single chamber microbial fuel cells. Environ. Sci. Technol. 2006, 40, 364–369. (15) APHA. Standard Methods for the Examination of Water and Wastewater, 20th ed.; American Public Health Association: Washington, DC, 1998. (16) Verstraete, W.; Aelterman, P.; Rabaey, K.; Pham, H. T.; Boon, N. Continuous electricity generation at high voltages and currents using stacked microbial fuel cells. Environ. Sci. Technol. 2006, 40, 3388–3394. (17) Logan, B. E.; Hamelers, B.; Rozendal, R. A.; Schrorder, U.; Keller, J.; Freguia, S.; Aelterman, P.; Verstraete, W.; Rabaey, K. Microbial fuel cells: Methodology and technology. Environ. Sci. Technol. 2006, 40, 5181–5192. (18) Chen, M. Y.; Lee, D. J.; Tay, J. H.; Show, K. Y. Staining of extracellular polymeric substances and cells in bioaggregates. Appl. Microbiol. Biotechnol. 2007, 75, 467–474. (19) Zhang, B.; Yamamoto, K.; Ohgaki, S.; Kamiko, N. Floc size distribution and bacterial activities in membrane separation activated
ARTICLE
sludge processes for small-scale wastewater treatment/reclamation. Water Sci. Technol. 1997, 35, 37–44. (20) Lee, J.; Ahn, W. Y.; Lee, C. H. Comparison of the filtration characteristics between attached and suspended growth microorganisms in submerged membrane bioreactor. Water Res. 2001, 35, 2435–2445. (21) Zhang, H. M.; Wang, X. L.; Xiao, J. N.; Yang, F. L.; Zhang, J. Enhanced biological nutrient removal using MUCT-MBR system. Bioresour. Technol. 2009, 100, 1048–1054. (22) Ersu, C. B.; Ong, S. K.; Arslankaya, E.; Lee, Y. W. Impact of solids residence time on biological nutrient removal performance of membrane bioreactor. Water Res. 2010, 44, 3192–3202. (23) Chu, L.; Li, S. Filtration capability and operational characteristics of dynamic membrane bioreactor for municipal wastewater treatment. Sep. Purif. Technol. 2006, 51, 173–179. (24) Virdis, B.; Read, S. T.; Rabaey, K.; Rozendal, R. A.; Yuan, Z.; Keller, J. Biofilm stratification during simultaneous nitrification and denitrification (SND) at a biocathode. Bioresour. Technol. 2011, 102, 334–341. (25) Logan, B. Scaling up microbial fuel cells and other bioelectrochemical systems. Appl. Microbiol. Biotechnol. 2010, 85, 1665–1671. (26) Katuri, K. P.; Scott, K.; Head, I. M.; Picioreanu, C.; Curtis, T. P. Microbial fuel cells meet with external resistance. Bioresour. Technol. 2011, 102, 2758–2766. (27) Zhang, F.; Merrill, M. D.; Tokash, J. C.; Saito, T.; Cheng, S.; Hickner, M. A.; Logan, B. E. Mesh optimization for microbial fuel cell cathodes constructed around stainless steel mesh current collectors. J. Power Sources 2011, 196, 1097–1102. (28) Wang, X. M.; Li, X. Y. Accumulation of biopolymer clusters in a submerged membrane bioreactor and its effect on membrane fouling. Water Res. 2008, 42, 855–862. (29) Marshall, C. W.; May, H. D. Electrochemical evidence of direct electrode reduction by a thermophilic Gram-positive bacterium, Thermincola ferriacetica. Energ Environ Sci. 2009, 2, 699–705. (30) W€asche, S.; Horn, H.; Hempel, D. C. Influence of growth conditions on biofilm development and mass transfer at the bulk/biofilm interface. Water Res. 2002, 36, 4775–4784.
9261
dx.doi.org/10.1021/es2019803 |Environ. Sci. Technol. 2011, 45, 9256–9261
ARTICLE pubs.acs.org/est
Mercury Export from the Yukon River Basin and Potential Response to a Changing Climate Paul F. Schuster,*,† Robert G. Striegl,† George R. Aiken,† David P. Krabbenhoft,‡ John F. Dewild,‡ Kenna Butler,† Ben Kamark,† and Mark Dornblaser† † ‡
U.S. Geological Survey, 3215 Marine Street, Boulder, Colorado 80303, United States U.S. Geological Survey, 8505 Research Way, Middleton, Wisconsin 53562, United States
bS Supporting Information ABSTRACT: We measured mercury (Hg) concentrations and calculated export and yield from the Yukon River Basin (YRB) to quantify Hg flux from a large, permafrost-dominated, high-latitude watershed. Exports of Hg averaged 4400 kg Hg yr 1. The average annual yield for the YRB during the study period was 5.17 μg m 2 yr 1, which is 3 32 times more than Hg yields reported for 8 other major northern hemisphere river basins. The vast majority (90%) of Hg export is associated with particulates. Half of the annual export of Hg occurred during the spring with about 80% of 34 samples exceeding the U.S. EPA Hg standard for adverse chronic effects to biota. Dissolved and particulate organic carbon exports explained 81% and 50%, respectively, of the variance in Hg exports, and both were significantly (p < 0.001) correlated with water discharge. Recent measurements indicate that permafrost contains a substantial reservoir of Hg. Consequently, climate warming will likely accelerate the mobilization of Hg from thawing permafrost increasing the export of organic carbon associated Hg and thus potentially exacerbating the production of bioavailable methylmercury from permafrost-dominated northern river basins.
’ INTRODUCTION Atmospheric deposition of mercury (Hg) and its subsequent export from watersheds is a critical concern to aquatic and terrestrial food webs.1 This concern is of particular significance in northern latitudes where arctic rivers presently deliver substantial amounts of both organic matter and Hg to the Arctic Ocean, where elevated foodweb levels of Hg are particularly problematic.2 Future export of Hg by these systems may potentially be influenced by two factors. First, the landscapes of northern regions are changing as a result of a warming climate.3 Frozen soils in this region contain large stores of both organic matter4 and Hg that may be mobilized from thawing permafrost.5 Second, it is anticipated that atmospheric deposition of Hg in northern regions will increase due to rapidly expanding Eurasian industrialization.6 Thus, it is important to measure the current exports and yields of Hg from northern river basins so that a baseline is established to compare against future conditions. Much of what is known about Hg transport in rivers is derived from studies of small streams and lakes at lower latitudes.7,8 In many of these systems, Hg and organic matter concentrations are strongly correlated,9 and it is generally accepted that interactions of Hg with organic matter in soils and stream waters are major factors controlling Hg export from a watershed.8,10 12 Boreal forests contain approximately one-third of the world’s terrestrial organic carbon, and DOC concentrations and fluxes in northern streams draining boreal forest catchments are generally greater This article not subject to U.S. Copyright. Published 2011 by the American Chemical Society
than those at lower latitudes.13 In these systems, changes in climate are resulting in permafrost thaw14 and increased fire frequency.15,16 In turn, both the biogeochemical activity within soils, which controls the fate of organic matter, and the amount and timing of water discharge (Q) from a watershed are being altered.17,18 Here, we present Hg export data for the Yukon River, a major northern river draining to the Bering Sea and, ultimately, the Arctic Ocean. The Yukon River, the subject of U.S. Geological Survey investigations since 2001 (see Supporting Information), provides a unique opportunity to study the cotransport of Hg and organic matter in a large, high latitude boreal river. It is the longest free-flowing river in the world,19 and with little development along its 3200 km reach, it is considered relatively pristine. In addition, nearly 75% of the Yukon River Basin (YRB) is covered by organic-rich permafrost soils. These exceptional characteristics along with vast reservoirs of sequestered carbon, complex geologic terrain, and naturally high suspended sediment loads provide a unique setting to examine Hg dynamics at a large scale. These data will serve as a baseline by which to compare future conditions, increase our understanding of correlations of Received: June 17, 2011 Accepted: September 12, 2011 Revised: September 8, 2011 Published: September 12, 2011 9262
dx.doi.org/10.1021/es202068b | Environ. Sci. Technol. 2011, 45, 9262–9267
Environmental Science & Technology Hg and organic carbon (OC) at large scales, and establish the basis for utilization of the organic matter concentrations and water yield to be used as proxy measures of Hg yield at large scales.
’ EXPERIMENTAL SECTION The Yukon River is the 19th longest river in the world (3200 km), draining the fourth largest basin (853 500 km2) with the fifth largest average annual discharge (Q) (6500 m3 s 1) in North America (see Supporting Information for an overview of YRB characteristics). The YRB is located in Arctic and Subarctic latitudes (60 68° N) that include most of interior central Alaska and the Yukon Territory of Canada (see Supporting Information, Figure S1). Discharge was monitored continuously for the Yukon River at Pilot Station, Alaska (YRP) by the U.S. Geological Survey (gaging station 15565447; http://waterdata.usgs.gov/ ak/nwis/uv 15565447). In addition, Q was measured during water sampling using an Acoustic Doppler Current Profiler (ADCP).20 Surface water grab samples were collected in the centroid of flow off the bow of a small boat that was headed slowly into the current. For all organic matter sampling, surface water was sampled using the USGS EDI method21 and processed according to USGS protocols.22 Samples were filtered in the field, stored chilled, and shipped for analyses as soon as possible after collection. Discharge at YRP is operationally defined as the outflow of the YRB because (1) this is the furthest downstream site having no tidal influence and (2) the river is confined to a single channel providing the ability to accurately measure Q. River water at this site is composed of surface and groundwater from three general sources: (1) snowmelt and glacial meltwater that has a high sediment load and low dissolved organic carbon (DOC); (2) surface water originating from terrestrial landscapes dominated by permafrost and/or peat and has a high DOC but low sediment load; and (3) water derived from groundwater sources that are low in sediment and DOC but high in dissolved inorganic carbon (DIC). Water samples were collected from the YRP about 100 miles upstream from the Yukon Delta 7 8 times a year, once in March or April (under ice base flow) and then every 2 3 weeks during open water from October 2001 through September 2005.23 During this time, filtered total mercury (FHg) and particulate total mercury (PHg) were measured in 38 samples. Accurate estimation of average exports and yields required measurements of Hg and OC over a range of hydrologic conditions (abbreviations of Hg species are defined in Table 2). The 2001 05 data set captured a robust representation of end member or extreme flow events (see Supporting Information, Figure S2) and also included 4 years in which average flow conditions closely matched the 29 year average (see Supporting Information, Figure S3). As with all remote areas of the world, sampling logistics can be problematic. Thus, not all flow regimes could be sampled (i.e., very few samples between 37 and 82% of the time that Q was equaled or exceeded). Because a main objective of this research was to calculate annual exports as accurately as possible, we concentrated on sampling during high flows. The load (or export) estimating program used in this study (LOADEST)24 requires 12 concentration measurements with half of them as close to peak flow as possible. Thus, most of the measurements are concentrated between 0 and 50% of the time that Q was equaled or exceeded.
ARTICLE
’ ANALYTICAL APPROACH All Hg analyses were performed at the USGS Mercury Research Laboratory (MRL) in Middleton, Wisconsin.25 Dissolved organic carbon (DOC) analyses were performed by the USGS National Research Program (NRP) laboratory in Boulder, Colorado.23,26 POC analyses were performed at the USGS National Water Quality Laboratory and also at the Nutrient Analytical Services Lab (NASL), Chesapeake Bay Laboratory, Maryland, using the same methodologies described in Zimmerman et al.27 Annual exports (also referred to as fluxes) (mass yr 1) and yields (mass m 2 yr 1) of Hg were calculated using two independent approaches. In the first approach, LOADEST was used to calculate daily exports as described in Striegl et al.28 and Dornblaser and Striegl.29,30 Exports were estimated from 38 measurements of Hg concentrations and continuous Q over a wide range of flow conditions between 2001 and 2005 (see Supporting Information, Figure S2). In a second approach, linear regressions were developed between DOC and FHg and POC and PHg. The regression equations were used to estimate FHg and PHg exports and yields based on dissolved and particulate carbon concentrations, respectively. Mercury yields were calculated by dividing export by the watershed area. ’ RESULTS AND DISCUSSION Hydrology. Average annual Hg and OC exports and yields were determined over a range of hydrologic regimes. During the 5 years of study (2001 2005), continuous Q was measured at YRP and compared closely with Q measured over the 29 year long-term record. The study period captured record low and high annual flows and 3 years of average flows that matched closely to the 29 year average (see Supporting Information, Figure S3). Nearly one-third of the total annual Q from the YRB is from snowmelt occurring from early April to late June. Furthermore, 87% of annual water export occurs in less than half the water year (WY) (Oct. 1 through Sept. 30). Therefore, to obtain the best estimates of annual geochemical exports and yields, sample collection was coordinated with maximum flow. Additional samples were collected in the late winter under ice (see Supporting Information, Figure S2, plotted as greater than 80% on the flow duration curve). Because Q is at its lowest during the early spring, samples collected during this time represent mostly base flow (groundwater) for more than 200 days of the water year. Concentrations of Hg. Hg concentrations were dominantly in particulate phase (90%) with annual discharge weighted PHg concentrations averaging 9.6 ng L 1 (n = 34), whereas FHg concentrations averaged 1 ng L 1 (n = 38), with observed increases of 24 and 9 fold, respectively, from winter base flow to peak spring flow (Table 1). Compared to other large Northern hemisphere rivers (Table 2) these relatively high Hg concentrations observed in the YRB suggest contributions from geogenic and/or other terrestrial-based sources in addition to atmospheric Hg deposition. About 80% of the 34 samples measured for THg at the YRP, with a mean value of 21 ng L 1, exceeded the USEPA aquatic life THg standard for adverse chronic effects to biota (12 ng L 1).31 Forty percent of these exceeded values occurred during the spring, in advance of the fish migration and spawning season. Concentrations of filtered (dissolved) methylmercury (FMeHg) were generally below the method detection limit (MDL) of 0.04 ng L 1 (n = 38). The weighted average annual 9263
dx.doi.org/10.1021/es202068b |Environ. Sci. Technol. 2011, 45, 9262–9267
Environmental Science & Technology
ARTICLE
Table 1. Seasonal Exports of Filtered Mercury (FHg) (kg yr 1) and Particulate (PHg) (kg yr 1) and Dissolved Organic Carbon (DOC) (106 kg yr 1) from Five Year Means for the Yukon River at Pilot Station (YRP), 2001 2005, with Coefficients of Variation (CV) from the LOADEST Resultsa FHg
CV
PHg
CV
THg
DOCb
CVb
spring
198
-
1999
-
2205
826
-
summer
131
-
1955
-
2093
550
-
winter
24.0
-
49.0
-
73.9
197
annual exp
353
9.2
4003
12.9
4372
annual avgc
-
-
-
-
-
season
POCb
CVb
TOC
[FHg]
[FMeHg]
[PHg]
[PMeHg]
[THg]
[DOC]
345
-
1171
2.38
<0.04
18.1
0.05
20.5
11.3
385
-
935
1.33
<0.04
18.5
0.05
19.9
5.7
-
17.3
-
214
0.28
<0.04
0.75
<0.01
1.02
2.6
1585
6.0
751
7.2
2318
-
-
-
-
-
-
-
-
-
-
-
0.98
<0.04
9.58
<0.05
10.6
5.08
a
Average seasonal concentrations for this period for mercury, methylmercury, and DOC are in brackets. Mercury species: filtered mercury (FHg), filtered methymercury (FMeHg), particulate mercury (PHg), particulate methylmercury (PMeHg), total mercury (THg). Organic carbon species: dissolved organic carbon (DOC), particulate organic carbon (POC), total organic carbon (TOC). Seasons: Spring, May 1 to June 30; Summer, July 1 to October 31; Winter, November 1 to April 30. b From Striegl et al.28, Table 2, the error, % reported, are CVs. c Annual averages are weighted by season.
concentration of PMeHg was 0.05 ng L 1 (n = 36) with a MDL ranging from <0.01 to <0.06 ng L 1. Thirty six of the 74 MeHg measurements were below the MDL resulting in a high uncertainty for use in export and yield calculations. Therefore, only concentrations of FMeHg and PMeHg are reported in Table 1. The median value for TMeHg/THg ratio was 1%, ranging from 0.1% to 9% (n = 34). This value is much lower than what has been typically been reported in other studies (2 15%).32 Despite the low relative abundance of MeHg in the YRB samples, the ranges and estimated exports of MeHg indicate that Hg methylation is occurring. Other studies in temperate systems have positively correlated the percent wetland cover in watersheds to the export of MeHg.33 However, for the YRB, approximately 25% of which is classified as wetland, no clear relationship between percent wetland and TMeHg yields was observed from YRB subbasins (see Supporting Information, Figure S6). More thorough studies are needed to assess important Hg methylation sites in the YRB. Exports and Yields of Hg. Annual THg export from the YRB averaged 4400 kg yr 1 (92% was in the particulate phase (PHg); Table 1) and yields were 5.2 μg m 2 yr 1. Almost all the THg export (98%) occurred during the spring and summer (May 1 to Oct. 31). Hg exports estimated by LOADEST compared well with discretely measured exports (Figure 1). In addition, the coefficients of variation (CV), calculated as the standard error of prediction of the mean export, also suggest that Hg was generally modeled well for both the dissolved and particulate forms of Hg (Table 1). Overall, these results indicate that LOADEST was effective in estimating Hg exports and yields in the YRB. Correlation of Mercury, Organic Carbon, and Water Export. The strong correlation of Hg to OC in surface waters is widely reported in the scientific literature. However, most study areas are orders of magnitude smaller8 and are generally much less diverse ecosystems than the YRB. Hg concentrations from the YRB at the YRP were significantly correlated to concentrations of OC with DOC explaining 81% of the variability of FHg (p < 0.001) and POC explaining 50% of the variability of PHg (still strongly significant, p < 0.001) (Figure 2). Although exports and yields of OC are generally proportional to water yield24,28 (see Supporting Information, Figure S4; DOC Q correlation is weak but significant, R2 = 0.52, p < 0.001), Hg concentrations from the YRB at YRP were weakly correlated to Q (see Supporting Information, Figure S5). LOADEST uses only Q as an input parameter and does not account for the strong positive correlation of OC and Hg. To assess the performance of LOADEST to estimate Hg exports based only on inputs of Hg concentrations and Q, linear regressions between DOC and FHg
and POC and PHg concentrations were compared to Hg exports estimated by LOADEST (Figure 3). There is disagreement in predicting PHg export during rare extreme flow events. Overall, however, there is good agreement between the two approaches, supporting the idea that both LOADEST and the Hg-OC correlation can be used as predictive tools for Hg export. Extrapolation of the Hg Export Record. Although there are significant correlations among water, OC, and THg exports and good agreement between LOADEST and the linear regression approach (see Supporting Information), it is not yet possible to reconstruct or project annual Hg export from the YRB or other arctic and subarctic basins. Increased atmospheric deposition6 will increase potential sources of Hg available for hydrologic transport from watershed surfaces. However, hydrologic trends throughout the YRB suggest increased infiltration and groundwater contribution to streamflow attributable to permafrost thaw, which will result in increased mineralization or adsorption of DOC17,18 and potentially decreased FHg export. There is also evidence that increased DOC mineralization and soil respiration can lead to increased Hg vapor emissions from soils.34 Conversely, there is evidence that DOC exports are increasing in Eurasian high latitudes experiencing increased precipitation in recent decades.35 More relevant to the Hg export problem is quantification by our study that the vast majority (90%) of THg export is associated with particulates, most likely POC. POC data are sparse and difficult to measure accurately for large rivers, particularly for remote rivers, and the POC database for the Yukon28 and other subarctic and arctic rivers36 is small and mostly recent. Given the importance of POC to Hg transport and to other important biogeochemical processes, more emphasis needs to be placed on its measurement in large river studies. Comparison to Other Large Rivers. The data required to estimate exports and yields of Hg from other large river basins are limited both spatially and temporally. Table 2 compares selected watershed characteristics, water yields, and the average annual concentrations, exports, and yields of FHg and PHg, suspended sediment, and TOC (calculated as the sum of DOC and POC) of the YRB to published results for eight major Northern hemisphere river basins (details and references for the development of Table 2 are in the Supporting Information). THg yields from the YRB (dominantly in the particulate phase) are up to 32 times greater than these other 8 major river basins suggesting five possibilities: (1) the YRB receives higher rates of atmospheric Hg deposition; (2) the YRB receives additional terrestrial Hg, such as from thawing permafrost; (3) all the other rivers in this comparison are significantly impacted by flow control structures 9264
dx.doi.org/10.1021/es202068b |Environ. Sci. Technol. 2011, 45, 9262–9267
Except where noted, all TOC values are the sum of DOC and POC. b Filtered only. c Ranges are reported for watersheds where percent cover was calculated. d Hg measurements include under-ice sampling. Citations in paper.
4372 0.09 15.0 13.1 1.9 38f 2001 05 24 203 29, 35e Yukond
75
12, 14, 21, 22, 25 2.58 106 Yenisei
e
8.50 105
a
2.3
2.8 5.17
72.0
1.8 0.27 4.7 109 5.9 109 700 1993 24
12, 14, 21, 22, 25 2.42 10 Lena
36 66
626
13£
0.3
0.5
0.8
no data
2.3 109 6.1 1010
1.8 1.65
7.3
1.4 0.51
6.7 109 1.8 1010 4000 no data 1993 22
4 18 12, 14, 21, 22, 25 2.57 106 Ob
6
4 11 1 7 10, 11, 22 24 1.32 106 Nelsond St. Lawrence 18 20, 23, 24, 27 1.30 106
78 94
525
6£
1.0
1.7
2.7
1300 1991 17 405
9£
0.6
1.1
1.7
no data
3.5 109 1.7 1010
6.4
0.6 5.1 1.1b 1.4 113 1189 2003 07 1995 96 9 30 112 382
2£ 79f
0.5 0.6
0.4 2.3
0.9 2.9
0.05 no data
1.5 109b 7.4 108 1.8 109 6.9 109
0.09 0.89
73.1
11.2 1.0 0.66 2117 no data 5.0 3.8 1.1 2001 04 18 12, 15 17 Mississippi
3.20 106
12 14, 22, 26 Mackenzie
1.71 10
0
580
62f
0.076 7.2 3.7 2.8 2003 05 18 67
315
79£
3.2 109 3.6 1010
1.1 1.28 1.3 10 1.9 10 2200
0.6 1.4b
9
3.9 108b 1.8 108 2003 07 15 31
6
2£
1.7
0.3
2.0
0.2
b
37
11
0.13
ARTICLE
17
6
2.81 105 10, 11, 22 24 Churchilld
TSS TOC
(μg/m2/yr) (g/m2/yr) (g/m2/yr)
THg TSS
(kg/yr) (kg/yr)
TOC THg TMeHg THg PHg
(ng/L) (ng/L) (ng/L) (ng/L) (kg/yr)
FHg samples
f or sites £ yield (cm/yr) sampling (km3/yr) coverc area (km2)
of Hg water
number of Hg period average
Supporting
average percent
watershed permafrost water export
citations listed in
Information river
annual average yield annual average export annual average concentration
Table 2. Comparison of Water, Total Suspended Sediments (TSS), Organic Carbon (TOC), and Methylmercury (TMeHg), and Filtered and Particulate Forms of Mercury (Hg) in the Yukon River to 8 Other Major Rivers in the Northern Hemispherea
Environmental Science & Technology
Figure 1. Comparison of calculated LOADEST mercury (Hg) export to discretely measured Hg export at the outlet of the Yukon River Basin at Pilot Station, Alaska, 2001 2005: (a) filtered (dissolved) total mercury (FHg); (b) particulate total mercury (PHg).
Figure 2. Relationship between concentrations of mercury (Hg) and organic carbon for samples collected from the outlet of the Yukon River Basin at Pilot Station, Alaska, 2001 2005: (a) filtered (dissolved) total mercury (FHg) and dissolved organic carbon (DOC) (n = 37); (b) particulate total mercury (PHg) and particulate organic carbon (POC) (n = 33).
Figure 3. Comparison of mercury (Hg) export calculated by LOADEST to Hg export calculated from the regression of organic carbon with Hg: (a) filtered (dissolved) Hg (FHg); (b) particulate Hg (PHg). Thick line, linear regression; thin line, 1:1 correlation.
such as dams, levees, and/or locks, which served as PHg traps and thus resulting comparatively lower THg exports; (4) late summer glacial melt waters originating from vast ice fields in the headwaters of the YRB rich in glacial sediments may provide additional sources (both anthropogenic and geogenic) of Hg (the Yukon River is the only river in this comparison with a significant source of glacial melt); (5) the lower yields for the other major rivers reported here may be underestimated as a result of infrequent sampling, which may not have captured maximum exports during high flow periods. However, it should be noted that, even if the yields from the other rivers were doubled, they still would be lower than the YRB estimates. Despite the similarities and differences between the YRB and the eight other river basins in this comparison these results highlight the importance of sampling major rivers frequently throughout the water year, especially during periods of high flow, to gain a better understanding of Hg exports to the world’s oceans. During base flow in the winter and early spring, contributions from 9265
dx.doi.org/10.1021/es202068b |Environ. Sci. Technol. 2011, 45, 9262–9267
Environmental Science & Technology glacial meltwater sources are negligible and Hg export from the YRB is closer to the other rivers in this comparison (Table 1). If the climate continues to warm, Hg exports may initially increase in response to increased wasting of the glacial sources. However, as those glacial PHg sources become exhausted, Hg export from the Yukon River may more closely resemble Hg export from other large arctic rivers such as the Lena. The percentage of permafrost cover across these nine watersheds is highly variable (Table 2) with the Yukon and the Lena having the greatest coverage (75 94%). Preliminary results from analyses of THg in a permafrost core from the YRB5 indicate that permafrost soils contain a substantial reservoir of historically accumulated Hg (see Supporting Information for further details). Therefore, in response to a warming climate, northern watersheds with a high percentage of permafrost cover may experience an increase in THg export in the future. Unlike the glacial PHg source, Hg derived from thawing permafrost would likely have a much greater proportion of FHg due to its partitioning to DOC. The pristine setting of the YRB provides a unique opportunity to measure Hg concentrations, exports, and yields and evaluate terrestrial and atmospheric sources (natural and anthropogenic). The absence of Hg point sources, such as those at lower latitudes, greatly reduces complications associated with the study of Hg dynamics and the response to climate change in an ecosystem. These findings underscore the importance of a better understanding of Hg-OC interactions in largescale northern ecosystems given the vast reservoirs of OC that exist in Arctic region wetlands and permafrost. Twenty percent of the Earth’s land surface is covered by permafrost, which is thawing in the Arctic and Subarctic regions, thereby increasing the potential release of vast reserves of OC-bound Hg. This is particularly noteworthy because dissolved, OC-associated Hg is considered to be an important bioavailable Hg fraction that feeds the methylation process. Continued warming will likely promote or accelerate the mobilization of Hg from permafrost increasing concentrations, exports, yields, and potential Hg methylation in the YRB and other northern region river basins of similar character.
’ ASSOCIATED CONTENT
bS
Supporting Information. Detailed descriptions of the study area, experimental procedures, and analytical approaches. This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 303-541-3052; email:
[email protected].
’ DISCLOSURE Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. ’ ACKNOWLEDGMENT The authors acknowledge numerous colleagues for their contributions to this work. Specifically, we would like to acknowledge Tom Sabin, Shane Olund, and Nicole Herman-Mercer. In addition, thanks to the field staff of the USGS Alaska Science
ARTICLE
Center for their field sample collection efforts. We would also like to thank all reviewers for their valuable suggestions. This work was supported by the USGS Toxic Substances Hydrology Program and the National Research Program.
’ REFERENCES (1) Morel, F. M. M.; Kraepiel, A. M. L.; Amyot, M. The chemical cycle and bioaccumulation of mercury. Annu. Rev. Ecol. Syst. 1998, 29, 543–566. (2) Outridge, P. M.; MacDonald, R. W.; Wang, F.; Stern, G. A.; Dastoor, A. P. A mass balance inventory of mercury in the Arctic Ocean. Environ. Chem. 2008, 5 (2), 89–111. (3) Hinzman, L. D.; et al. Evidence and implications of recent climate change in northern Alaska and other arctic regions. Clim. Change 2005, 72 (3), 251–298. (4) Zimov, S. A.; Schuur, E. A.; Chapin, S. F., III Permafrost and the global carbon budget. Science 2006, 312, 1612–1613. (5) Schuster, P. F.; Krabbenhoft, D. P.; DeWild, J. F.; Sabin, T. G. A paleoenvironmental record of atmospheric mercury deposition in a permafrost core from Northern Alaska. Eos Trans. 2008. AGU, 89 (53), Fall Meet. Suppl., Abstract B13C-0452. (6) Ford, J.; Creceluis, E. Heavy metals and trace elements in Arctic Alaska; point sources vs. long range transport. Arct. Sci. Conf. 2001, 52, 45. (7) Brigham, M. E.; Wentz, D. A.; Aiken, G. R.; Krabbenhoft, D. K. Mercury cycling in stream ecosystems. 1. Water column chemistry and transport. Environ. Sci. Technol. 2009, 43, 2720–2725. (8) Grigal, D. F. Inputs and outputs of mercury from terrestrial watersheds: a review. Environ. Rev. 2002, 10, 1–39. (9) Dittman, J. A.; Shanley, J. B.; Driscoll, C. T.; Aiken, G. R.; Chalmers, A. T.; Towse, J. E.; Selvendiran, P. Mercury dynamics in relation to dissolved organic carbon concentration and quality during high flow events in three northeastern streams. Water Resour. Res. 2010, 46, 7. (10) Shanley, J. B.; Scherbatskoy, T.; Donlon, A.; Keeler, G. Mercury cycling and transport in the Lake Champlain Basin. In Lake Champlain in transition: From research toward restoration, Water Resources Monograph Series 14; Manley, T. O., Manley, P., Eds.; American Geophysical Union: Washington, D.C., 1999; pp 227 299. (11) Hurley, J. P.; Krabbenhoft, D. P.; Cleckner, L. B.; Olson, M. L.; Aiken, G. R.; Rawlik, P. S., Jr. System controls on the aqueous distribution of mercury in the northern Florida Everglades. Biogeochemistry 1998, 40, 293–310. (12) Driscoll, C. T.; Blette, V.; Yan, C.; Scholfield, C. L.; Munson, R.; Holsapple, J. The role of dissolved organic carbon in the chemistry and bioavailability of mercury in remote Adirondack lakes. Water, Air, Soil Pollut. 1995, 80, 499–508. (13) O’Donnell, J. A.; Aiken, J. R.; Kane, E. S.; Jones, J. B. Source water controls on the character and origin of dissolved organic matter in streams of the Yukon River basin, Alaska. J. Geophys. Res 2010, 115, G03025. (14) Jorgenson, M. T.; Shur, Y. L.; Pullman, E. R. Abrupt increase in permafrost degradation in Arctic Alaska. Geophys. Res. Lett. 2006, 33, 2. (15) Kane, E. S.; Harden, J. W.; Kasischke, E. S.; Turetsky, M. R.; Manies, K. L. Soil Drainage and topographic influences on wildfire consumption of soil organic carbon in boreal forests: implications for carbon stability. Am. Geo. Union 2006, 87, 52. (16) Turetsky, M. R.; Harden, J. W.; Friedli, H. R.; Flannigan, M.; Payne, N.; Crock, J.; Radke, L. Wildfires threaten mercury stocks in northern soils. Geophys. Res. Lett. 2006, 33, 16. (17) Walvoord, M. A.; Striegl, R. G. Increased groundwater to stream discharge from permafrost thawing in the Yukon River Basin: potential impacts on lateral export of carbon and nitrogen. Geophys. Res. Lett. 2007, 34, L12402. (18) Striegl, R. G.; Aiken, G. R.; Dornblaser, M. M.; Raymond, P. A.; Wickland, K. P. A decrease in discharge-normalized DOC export by the 9266
dx.doi.org/10.1021/es202068b |Environ. Sci. Technol. 2011, 45, 9262–9267
Environmental Science & Technology
ARTICLE
Yukon River during summer through autumn. Geophys. Res. Lett. 2005, 32 (21), L21413. (19) Nilsson, C.; Reidy, C. A.; Dynesius, M.; Revenga, C. Fragmentation and flow regulation of the world’s large river systems. Science 2005, 308, 405–408. (20) Oberg, K. A.; Morlock, S. E.; Caldwell, W. S. Quality assurance plan for discharge measurements using acoustic Doppler current profilers. Scientific Investigations Report 2005 5183; U.S. Geological Survey, 2005. Available at http://pubs.usgs.gov/sir/2005/5183/ (accessed July 2010). (21) http://water.usgs.gov/owq/FieldManual/chapter4/html/Ch4_ contents.html (22) http://water.usgs.gov/owq/FieldManual/chapter5/pdf/chap5. pdf (23) Schuster, P. F. Water and sediment quality in the Yukon River Basin, Alaska, during water year 2001. Open-File Report 2003 03427; U.S. Geological Survey, 2003. Available at http://water.usgs.gov/pubs/ of/2003/ofr03427/ (accessed August 2010). (24) Runkel, R. L.; Crawford, C. G.; Cohn, T. A. Load estimator (LOADEST): A FORTRAN program for estimating constituent loads in streams and rivers. In U.S. Geological Survey Techniques Methods, Book 4; U.S. Geological Survey: Washington, D.C., 2004; Ch. A5, p 69. (25) http://wi.water.usgs.gov/mercury-lab/ (26) Aiken, G. R. Chloride interference in the analysis of dissolved organic carbon by the wet oxidation method. Environ. Sci. Technol. 1992, 26, 2435–2439. (27) Zimmerman, C. F.; Keefe; Bashe, J. Determination of carbon and nitrogen in sediments and particulates of estuarine/coastal waters using elemental analysis-Method 440.0; U.S. Environmental Protection Agency: Cincinnati, OH, 1997. (28) Striegl, R. G.; Dornblaser, M. M.; Aiken, G. R.; Wickland, K. P.; Raymond, P. A. Carbon export and cycling by the Yukon, Tanana, and Porcupine rivers, Alaska, 2001 2005. Water Resour. Res. 2007, 43, W02411. (29) Dornblaser, M. M.; Striegl, R. G. Nutrient (N, P) loads and yields at multiple scales and subbasin types in the Yukon River basin, Alaska. J. Geophys. Res. 2007, 112, G04S57. (30) Dornblaser, M. M.; Striegl, R. G. Suspended sediment and carbonate transport in The Yukon River Basin, Alaska: fluxes and potential future responses to climate change. Water Resour. Res. 2009, 45, W06411. (31) Alaska water quality criteria manual for toxic and other deleterious organic and inorganic substances, Department of Environmental Conservation, State of Alaska, May 15, 2003. (http://water.epa.gov/ scitech/swguidance/standards/upload/2004_07_01_standards_wqslibrary_ ak_ak_10_toxics_manual.pdf. (32) Shanley, J. B.; Mast, A. M.; Campbell, D. H.; Aiken, G. R.; Krabbenhoft, D. P.; Hunt, R. J.; Walker, J. F.; Schuster, P. F.; Chalmers, A.; Aulenbach, B. T.; Peters, N. E.; Marvin-DiPasquale, M.; Clow, D. W.; Shafer, M. M. Comparison of total mercury and methylmercury cycling at five sites using the small watershed approach. Environ. Pollut. 2008, 154 (1), 143–154. (33) Barringer, J. L.; Risken, M. L.; Szabo, Z.; Reilly, P. A.; Rosman, R.; Bonin, J. L.; Fischer, J. M.; Heckathorn, H. A. Mercury and methylmercury dynamics in a coastal plain watershed, New Jersey, USA. Water, Air, Soil Pollut. 2010, 212 (1 4), 251–273. (34) Wickland, K. P.; Krabbenhoft, D. P.; Olund, S. Evidence for a link between soil respiration and mercury emission from organic soils. Eighth International Conference on Mercury as a Global Pollutant, Madison, WI, August 6 11, 2006. (35) Frey, K. E.; McClelland, J. W. Impacts of permafrost degradation on arctic river biogeochemistry. Hydrol. Processes 2009, 23, 169–182. (36) McClelland, J. W.; Holmes, R. M.; Peterson, B. J.; Amon, R.; Brabets, T.; Cooper, l.; Gibson, J.; Gordeev, V. V.; C. Guay, C.; Milburn, D.; Staples, T.; Raymond, P. A.; Shiklomanov, I.; Striegl, R.; Zhulidov, A.; Gurtovaya, T.; Zimov, S. Development of a pan-Arctic database for river chemistry. EOS 2008, 89, 217–218. 9267
dx.doi.org/10.1021/es202068b |Environ. Sci. Technol. 2011, 45, 9262–9267
ARTICLE pubs.acs.org/est
Long-Range Transported Atmospheric Pollutants in Snowpacks Accumulated at Different Altitudes in the Tatra Mountains (Slovakia) Lourdes Arellano,† Pilar Fernandez,*,† Jolana Tatosova,‡ Evzen Stuchlik,‡ and Joan O. Grimalt† †
Department of Environmental Chemistry, Institute of Environmental Assessment and Water Research (IDÆA-CSIC), Jordi Girona 18, 08034 Barcelona, Catalonia, Spain ‡ Hydrobiological Station, Institute for Environmental Studies, Faculty of Science, Charles University in Prague, P.O.Box 47, CZ-388 01 Blatna, Czech Republic
bS Supporting Information ABSTRACT: Persistent organic pollutants (POPs), including polychlorobiphenyls (PCB), endosulfans, hexachlorocyclohexanes (HCHs), hexachlorobenzene (HCB), polybromodiphenyl ethers (PBDEs), and polycyclic aromatic hydrocarbons (PAHs), were analyzed in snowpack samples collected along an altitudinal gradient (16832634 meters above sea level) in the High Tatra Mountains (Slovakia). All analyzed compounds were found at all altitudes, pointing to their global distribution. The presence of PBDEs, particularly BDE 209, in the snowpack samples is especially relevant, as it reflects the air transport capacity of this low volatile, very hydrophobic pollutant to remote mountain regions. The most abundant compounds at all altitudes were PAHs, with mean values ranging from 90 to 300 ngL1, 1 order of magnitude higher than concentrations of other compounds. PCBs (sum of PCB 28, 52, 101, 118, 153, 138, and 180) and BDE 209 were the dominant organohalogen pollutants, with concentrations from 550 to 1600 pg L1 and from 670 to 2000 pgL1, respectively. Low brominated PBDEs, endosulfans, HCHs and HCB were consistently found in all samples at lower concentrations. The concentrations of these compounds correlated positively with altitude (i.e., negatively with temperature), which is consistent with cold-trapping effects. The regression coefficients were positive and statistically significant (p < 0.05) for all compounds except BDE 209, endosulfan sulfate, HCB and α-HCH. Contrariwise, the concentrations of BDE 209 and endosulfan sulfate exhibited a statistically significant positive correlation with total particle amount, which agrees with long-range atmospheric transport associated to aerosols according to the physical-chemical properties of these compounds. Snow specific surface area, which determines the maximum amount of each organic compound that can be sorbed by snow, proved utile for describing the distribution of the more volatile compounds, namely α-HCB and HCB, in the snowpack.
’ INTRODUCTION Persistent organic pollutants (POPs) encompass a wide range of chemical compounds that differ in origin but that have similar physical-chemical properties and pose serious health concerns for humans and other organisms.1 Their strong resistance to photolytic, chemical and biological transformation enables them to remain in the environment for long periods of time. Furthermore, they have low water solubility and high lipid affinity, which leads to bioaccumulation and biomagnification of them through food chains. International efforts were taken over the last two decades to reduce or eliminate major primary sources of POPs, which culminated in the signing of the Stockholm Convention on POPs in 2001.2 Many POPs are susceptible to long-range atmospheric transport, as indicated by the fact that they have been identified in polar regions where they were never produced nor used. The terms global distillation and selective trapping have been used to describe the mechanisms responsible for the selective enrichment of these compounds in cold areas which results from partition processes controlled by ambient temperature.3,4 Remote mountain regions are usually considered to be the most pristine continental areas in temperate and tropical latitudes. r 2011 American Chemical Society
However, despite their remoteness, these ecosystems may receive significant loads of POPs, which are deposited after long-range atmospheric transport.5,6 As observed in the polar regions, field studies in Europe,7 Western Canada,8,9 South America,10,11 the Tibetan Plateau,12 and other sites, showed that mountains can act as cold-traps for POPs or even for chemicals with limited atmospheric transport (e.g., currently used pesticides).13 Additionally, mountain regions constitute pronounced small-scale temperature gradients. Thus, the distributions of chemical compounds along their altitudinal gradients can provide valuable information on the factors influencing transport and accumulation processes which can be exploited for understanding the mechanisms that operate on the global scale. In midlatitude mountains, snow is important for the deposition and accumulation of long-range transported organic chemicals into soils and freshwater systems.14 First, it constitutes a Received: June 21, 2011 Accepted: September 15, 2011 Revised: August 25, 2011 Published: September 15, 2011 9268
dx.doi.org/10.1021/es202111n | Environ. Sci. Technol. 2011, 45, 9268–9275
Environmental Science & Technology
Figure 1. Map showing the location of the sampling sites used. The yellow contour lines represent elevation above sea level (asl).
significant fraction of the annual precipitation that has better scavenger efficiency of atmospheric contaminants than rain.15,16 Second, the seasonal snowpack can act as a temporary reservoir of these pollutants which are subsequently released during snowmelt. Indeed, snowpack melting may provide a major portion of the annual contaminant influx to mountain ecosystems, an amount that may exceed the contribution from direct atmospheric deposition. As reported elsewhere,14 snow deposition may have synergetic effects with cold temperature that may enhance POP transfer from air to terrestrial surfaces in mountain regions. Finally, seasonal snow cover is also relevant for the airsurface exchange of these chemicals due to its lower storage capacity than that of soil and vegetation.17 Previous reports on the presence of POPs in snow have focused on high latitude areas.18,19 Few studies have addressed the role of mountain snow in subtropical latitudes, and those available are mainly from North America.8,13,20,21 The information on polybromodiphenyl ethers (PBDEs) is even scarcer, since their occurrence in high altitude sites has been examined only in soils, air or organisms22,23 and no data on snow accumulation in mountain regions has yet been published. However their chemical properties suggest that they should also be accumulated in high mountain regions such as other persistent organic pollutants.24,25 Further insight is needed into both the role of snow in trapping and storing these compounds in high mountain regions, and the influence of snow properties (e.g., specific surface area, density and particle content) and environmental properties (e.g., temperature) on the supply of toxicologically relevant contaminants to high mountain ecosystems. To this end, in the present work the concentrations of organochlorine compounds (OCs), PBDEs, and polycyclic aromatic hydrocarbons (PAHs) were examined in seasonal snowpacks taken along an altitudinal gradient (16832634 m asl (meters above sea level)) in the Tatra Mountains (Slovakia). Previous studies have shown significant concentrations of some of these atmospherically transported pollutants in this region.26,27 The compounds included in the study reported here encompass both legacy and currently used pollutants of diverse physicalchemical properties and origins, providing representative cases for description of the environmental processes involved in the observed accumulation patterns.
ARTICLE
’ EXPERIMENTAL SECTION Total snowpack samples were collected in duplicate at six different altitudes along Skalnata valley in the High Tatras (Carpathian Mountains, Slovakia) (Figure 1). Sampling elevations span from 1683 m asl to 2634 m asl, and are situated above the regional timberline, which is defined at 1550 m on the Northern slope and at 1650 m on the Southern slope. The temperature lapse rates are 0.70 °C/100 m in the Northern and 0.67 °C/100 m in the Southern slope,28 involving annual average temperature differences of 6.5 °C between the lowest and the highest sampling points. Sampling was performed in April 2005 at the time of maximum annual snow accumulation, before the onset of spring melting. Representative samples of the whole snow material deposited during the cold season (from October to the sampling date) were obtained by excavating a snow pit down to the ground. The pit wall was subsampled with a stainless-steel cube (volume: 1 L) over the entire vertical profile. Approximately 30 L of snow were collected at each sampling site. The samples were transported to the field laboratory in clean polyfluoralkoxy bags that had been hermetically sealed and placed inside black polyethylene bags for protection from light. The collected snow was transferred to prerinsed stainless-steel containers and left to melt in the dark at room temperature. Snow depth (in water equivalents) and snow density were determined for each sample as described in the Supporting Information (SI). Daily precipitation and temperature data were obtained from the meteorological observatory in Skalnate Pleso (Geophysical Institute of Slovak Academy of Science) situated at 1778 m asl, close to the second sampling point of the altitudinal gradient. Detailed descriptions of the analytical methods for OC, PBDEs, and PAHs, and quality assurance/quality control procedures, are reported in the SI. ’ RESULTS AND DISCUSSION Snow Properties. Snow density, snow depth in water equivalents (SWE), and snow specific surface area (SSA) of the analyzed snowpacks are listed in Table 1. The last of these parameters describes the snow surface area accessible to organic contaminants when present in the atmospheric gas phase, and it is related to the maximum amount of chemical compounds that can be sorbed by the snow.20 SSA varies during snow aging, with reported values of ca. 800 cm2 g1 for fresh snow and ca. 150 cm2 g1 for aged snow.29 Lower SSA values result in lower sorption capacity. SSA reduction during snow aging enhances the volatilization of the accumulated contaminants back to the atmosphere; obviously, this process is most relevant for the most volatile compounds. The observed SSA values in the Tatra samples ranged from 68 to 123 cm2 g1, values which are characteristic of aged snow30 with lower sorption capacity than fresh snow. Lower SSA values were generally observed in the samples collected at higher altitudes. Snow densities ranged from 0.35 to 0.45 kg L1 and SWE values barely varied, ranging from 108 to 118 cm. Concentrations and Deposition Fluxes. The mean concentrations of PAHs, hexachlorocyclohexanes (α- and γ-HCH), hexachlorobenzene (HCB), endosulfans (α- and β-endosulfan, and endosulfan sulfate), polychlorobiphenyls (PCBs, sum of PCB 28, 52, 101, 118, 153, 138, and 180) and PBDEs are summarized in Table 2. PAHs were the most abundant compounds at all altitudes, with mean values ranging from 90 to 300 ng L1, 9269
dx.doi.org/10.1021/es202111n |Environ. Sci. Technol. 2011, 45, 9268–9275
Environmental Science & Technology
ARTICLE
Table 1. Sampling Site Location and Physical Properties of the Snowpacks site
altitude
temp.
SWE
snow density
SSA
particles
latitude (N)
longitude (E)
(m asl)a
(°C)b
(cm)c
(kg L-1)c
(cm2 g -1)c
(mg L-1)c
49°110 22.3000
20°140 11.9000
1683
1.9
108 ( 8
0.35 ( 0.00
123 ( 0.6
0.92 ( 0.04
Skalnate Pleso meteorological st.
49°11 23.76
20°140 3.1800
1787
2.6
107 ( 1
0.40 ( 0.01
97.1 ( 2.9
1.44 ( 0.18
above Skalnate Pleso
49°110 26.3000
20°130 56.5000
below Skalnate Pleso
0
00
1840
2.9
115 ( 10
0.38 ( 0.03
104 ( 18
1.81 ( 0.77
Lomnica Pass
49°11 25.38
20°130 7.5600
2150
5.0
111 ( 3
0.45 ( 0.00
68.0 ( 0.7
1.08 ( 0.15
above Lomnica Pass
49°110 26.3400
20°120 53.5800
2300
6.1
118 ( 3
0.38 ( 0.03
106 ( 14
2.51 ( 0.89
Lomnica Peak
49°110 42.5400
20°120 47.7600
2634
8.3
113 ( 2
0.41 ( 0.00
88.8 ( 2.2
2.48 ( 0.29
0
00
a
Meters above sea level. b Mean temperature from October to April estimated from data measured at the Skalnate Pleso meteorological station assuming temperature lapse rate of 0.68 °C/100 m.28 c Mean value from duplicate samples at each site. SWE, snow water equivalent; SSA, snow specific surface area estimated using the equation reported for alpine snow.29
Table 2. Mean Concentrations of Organic Contaminants Measured in the Tatras Mountains Snowpacksa altitude (m asl)b
PCBsc
HCHsd
endosulfanse
HCB
PBDEsf
BDE 209
Tatra Mountains (This Study) below Skalnate Pleso station 2005
1683
550
30
12
5.8
53
2040
90
observator Skalnate Pleso station
1787
660
40
16
4.4
64
675
115
above Skalnate Pleso
1840
690
26
20
3.4
119
1220
120
Lomnica Pass
2150
1045
34
12
5.8
52
670
130
above Lomnica Pass
2300
795
54
20
9.9
80
1675
120
Lomnica Peak
2634
1630
75
24
6.2
112
1060
305
Other Remote Sites 1998
2240
217
520
10
5.6
Alps (Austria)26
1997
2417
730
1100
BDL
17
Alps (Swirtzerland)26
1997
2519
2200
500
BDL
16
Tatra (Slovakia)26
1998
2000
200
22
7.1
81
Maladeta (Spain)49
2005
18203196
380590
190370
Colle del Lys (Monte Rosa, Alps)56
2003
4250
15.9243
10.7234
Banff National Park (Canada)57
1997
site
Pyrenees (Spain)26
sample year
parks across the western U.S.13 Mt. Xixabangma (the Himalayas)58
20034 2005
Rocky Mountains (Canada)8
1996
Eastern Italian Alps38
2005
4303240 6720
PAHsg
2.65.8
490
360 j
16
9.364h 90340
BDL-400 BDL
501500
3.070
11001400i
170430
490 j
8.997
rural and remote sites
6002150
1598
urban sites
280460
57239
Values in pg L1 except for PAHs (ng L1). Snow concentrations reported from other regions are included for comparison. b Meters above sea level. Sum of seven congeners. d Sum of α-HCH and γ-HCH. e Sum of α-endosulfan, β-endosulfan and endosulfan sulfate. f Sum of all PBDE congeners found except BDE 209. g Sum of 20 PAHs parent compounds (Figure S1). h Sum of 4 congeners. i Sum of 40 congeners. j α-endosulfan. BDL, below detection limit. a c
1 order of magnitude higher than the other compounds. PCBs and BDE 209 were the dominant organohalogen contaminants, with concentrations of 5501600 pg L1 and 6702000 pg L1, respectively. Low-brominated PBDEs, endosulfans, HCHs and HCB were also found in all samples, at lower concentrations (Table 2). The high PAH concentrations were similar to the levels found in urban areas and were consistent with the concentrations of PAH in air (5.15 ng m3 for gas+particulate phase),31 sediments (1318 μg g1 dw),27 lake water (12 ng L1 for dissolved+suspended particles)32 and aquatic organisms (33 ng g1 ww)33 from this region. These levels may reflect the influences of different sources, such as industrial emissions from the Black Triangle and domestic heating (burning of wood and coal).34
The concentrations of PCBs and of HCB fall within the range of values reported in snow from remote regions, whereas the concentrations of past and current use pesticides (HCHs and endosulfans) are lower than those found in previous studies. The annual snow deposition fluxes of these pollutants (expressed in μg m2 yr1) can be calculated by multiplying their concentrations in snow by the SWE, and then normalizing the value to 1 year. The PAH deposition in the snowpack samples ranged from 120 to 590 μg m2 yr1. These values are high relative to those reported in other mountain areas, such as the western U.S. (0.005310 μg m2 yr1)35 or west-central Europe (4.237 μg m2 yr1).26 Similar deposition fluxes were observed for PCBs (1.03.2 μg m2 yr1) and BDE 209 (1.2 3.7 μg m2 yr1); however, the deposition flux values of the 9270
dx.doi.org/10.1021/es202111n |Environ. Sci. Technol. 2011, 45, 9268–9275
Environmental Science & Technology other pollutants were 1 order of magnitude lower: total HCHs (0.040.16 μg m2 yr1); total endosulfans (0.020.05 μg m2 yr1); total BDEs excluding BDE 209 (0.090.24 μg m2 yr1); and HCB (0.010.02 μg m2 yr1). Compound Distributions. Polycyclic Aromatic Hydrocarbons. The relative PAH distributions were rather uniform along the entire snow altitudinal transect (SI, Figure S1). Fluoranthene (Fla) was the most abundant compound, followed by pyrene (Pyr), and phenanthrene. These three compounds, which are mainly present in the gas phase of the atmosphere,36 represented 40% of the total PAH content. Other parent PAHs compounds, such as chrysene + triphenylene (Chr + Tri) (10%), benzo [b+j]fluoranthene (8.0%), benzo[k]fluoranthene (6.9%), benzo[a]pyrene (B[a]P) (3.6%), benzo[e]pyrene B[e]P (5.5%), indeno[1,2,3-cd]pyrene (4.5%), and benzo[ghi]perylene (3.3%), were consistently found at all altitudes. These compounds are generally present in the atmospheric particulate phase.37 The presence of all these compounds in the snow samples confirmed that snow is an efficient scavenger of gas phase and particulate phase PAHs. The observed average distribution was similar to that of PAH mixtures found in other studies from high-mountain areas in Europe26,38 or North America35 and evidenced enhanced deposition of low molecular weight compounds at low temperatures, mainly due to the increase of more volatile PAH content in atmospheric particles, which will be deposited by the falling snow, and the snow scavenging of gas phase compounds. The PAH distributions in these snowpacks were also characterized by the high relative abundance of the reactive PAHs. Pyr, Benz[a]anthracene (B[a]A) and B[a]P are more susceptible to photodegradation than their corresponding isomers, Fla, Chr+Tri and B[e]P, respectively.39 In general, higher ratios of Pyr/(Pyr+Fla), B[a]A/(Chr+Tri+B[a]A) and B[a]P/(B[e]P+ B[a]P) were found in the Tatra snowpacks than those observed in lake waters or soils collected in this region, indicating a higher proportion of the more chemically labile isomers in the former environmental compartment (SI Table S1). These snowpack PAH ratios were close to those found in air samples (particle + gas phase) collected at Skalnate Pleso meteorological station31 (Figure 1: Sampling Site 5). This agreement showed a good preservation of the atmospheric PAH composition in snow after deposition. Furthermore, PAHs in the atmosphere of the Tatra Mountains were found in higher concentrations in winter,31 the season of the highest snow precipitation. Therefore, it was concluded that snowpacks store airborne PAH and preserve the more labile congeners. In the warm season, upon snowmelt, these mixtures are released into freshwater and into soils, representing a significant fraction of the annual PAH influx to this mountain ecosystem. Degradation of the more labile compounds occurs at higher temperature in these compartments, as evidenced by the observed PAH ratios32,40 (SI Table S1). Polybromodiphenyl Ethers. Fourteen PBDE congeners (BDEs 17, 28, 47, 66, 71, 85, 99, 100, 138, 153, 154, 183, 190, and 209) were measured in the snowpack. However, only BDEs 47, 99, 100, 183, and 209 were consistently found in all samples above the detection limit. BDE 209 showed the highest levels and the highest range of variation among all the organohalogen compounds (Table 2). The relatively high concentrations of this extremely low volatile compound in the snowpack reveals its capacity for transport to remote sites. The sum concentration of BDEs 47, 99, 100, and 183 ranged from 52 to 119 pg L1. Similar levels were found for BDE 47 (1748 pg L1) and BDE 99 (1651 pg L1). BDE 100
ARTICLE
Figure 2. Mean concentrations of PBDEs in the snowpack at different altitudes.
concentrations ranged from 5.1 to 13 pg L1, whereas those of BDE 183 ranged from 3.6 to 14 pg L1 (Figure 2). BDE congeners 47, 99, and 100 represent 9599% of the total BDEs in the technical penta-BDE mixture, whereas BDE 183 is the major congener (42%) in the technical octa-BDE mixture;41 both of these mixtures were banned in the European Union in 2004, just before the snow had been sampled. BDE 209 is the dominant compound in the technical deca-BDE mixture (from 92 to 97% of the total BDE content41), which was banned in July 2008.42 Therefore, the PBDE composition in the Tatra snow was consistent with the composition of these potential sources in Europe. The less-brominated compounds can also be formed from the environmental degradation of deca-BDE,43 which may also have contributed to the presence of these compounds in the Tatra snowpacks. Hexachlorocyclohexanes. HCH was used as a commercial insecticide. Past usage of HCHs involved the HCH technical mixture, where α-HCH was the dominant steroisomer with an αHCH/γ-HCH ratio of 47. This mixture was replaced by lindane, which consisted almost entirely on the γ-isomer. Therefore the ratio of α-HCH/γ-HCH is often used for discriminating between technical HCH versus lindane contributions.44 α- and γ-HCH were detected in all snow samples. In all cases, the concentration of γ-HCH (9.148 pg L1) was higher than that of α-HCH (4.434 pg L1), except for in one of the snowpack samples taken at the highest altitude site. The αHCH/γ-HCH ratios ranged from 0.22 to 1.58 which is far below the values measured in the technical HCH mixture. This is consistent with the phasing out of technical HCH in Europe and the subsequent use of lindane. Endosulfans. The mixtures of these insecticides in the snow samples were dominated by endosulfan sulfate (6.2 to 13.8 pg L1), which represented 5070% of this group of compounds in the snowpack. β-endosulfan was less abundant (3.49.1 pg L1) and the α-isomer exhibited the lowest concentrations (0.61 3.9 pg L1). The predominance of endosulfan sulfate and the low concentrations of the α- and β-isomers may indicate past uses or, more likely, contributions from long-range transport of the commercial insecticide (dominated by α- endosulfan (6467%) and β-endosulfan (2932%)) from distant sources. In this sense, endosulfan sulfate was the only compound found in the water column of Lake Ladove, a high mountain lake situated in the Tatra Mountains, whereas the α- and β- isomers were detected in other remote lake waters from Central and Southern Europe.32 The dominance of β-endosulfan over the α-isomer (α/β ratio: 0.140.43), does not coincide with either the composition of the technical mixtures or the composition of the air in the Tatra Mountains, where the α-isomer is generally found in higher 9271
dx.doi.org/10.1021/es202111n |Environ. Sci. Technol. 2011, 45, 9268–9275
Environmental Science & Technology
ARTICLE
Figure 3. Representation of log-transformed snow deposition fluxes of selected organic pollutants (ng m2yr1) versus sampling site altitudes. r = correlation coefficients.
Figure 4. Representation of a) log-transformed ratios of less volatile to more volatile PCBs (∑(153,138,180)/∑ (52,101,118)) and sampling site altitudes; (b) log-transformed ratios of less volatile to more volatile PCBs and snow specific surface area; and (c) log-transformed ratios of low to high molecular weight PAHs and snow specific surface area. r = correlation coefficients.
abundance.45 The higher aqueous solubility and lower Henry’s Law constant of the β-isomer relative to those of α-endosulfan (see SI Table S2) may result in the preferential deposition of the former in both dry and wet processes. This effect, together with the higher volatilization potential of the α-isomer from solid and aqueous surfaces, may explain the dominance of β-endosulfan in the snowpack samples.46 Endosulfan was extensively used in Canada, India, and China. Recently it has been included in the list of chemicals banned under the Stockholm Convection with some specific exemptions.47 Their presence in these snowpacks from the Tatras indicates that they are persistent enough to be transported to remote sites. Contaminant Distribution along the Altitudinal Transect. The coefficients of variation (CVs) for replicate samples taken at each site ranged from 12 to 24% for all compounds except for α-HCH (mean value: 37%) and BDE 209 (up to 90%). The values for PAHs and OCs were similar to those reported in other studies13,48 and may reflect the great heterogeneity of snow. Accordingly, identification of snow concentration differences between sites with statistical significance requires average sample concentration differences of at least 30%.
Organohalogen Compounds. Previous studies indicate that concentrations of semivolatile organic compounds in snow tend to increase with altitude;8,21,49,50 this same trend was observed in the Tatras Mountains samples, in which consistent altitudinal gradients of contaminant concentrations in snow were observed for almost all compounds except HCB, α-HCH, BDE 209 and endosulfan sulfate (Figure 3 and SI Table S2). Regression analysis between log deposition fluxes and altitude showed statistically significant positive correlation coefficients (p < 0.05) in all these cases except for BDE 100 (significance at p < 0.1). Concerning the PCBs, in the sites of highest elevation, the less volatile congeners had accumulated at higher levels than did the more volatile ones (Figure 4a). PCB 101 was the most abundant compound at the lowest sites (16831840 m asl) and PCB 138 was the dominant congener at the highest sites (21502634 m asl). This trend could be related to postdepositional processes, which ultimately determine the chemical burden in the snowpack. Revolatilization of the most volatile compounds during snow aging results in less volatile compounds remaining in the snow until spring melt.20 Moreover, revolatilization is more significant in snow than in other media because the snow-air partition 9272
dx.doi.org/10.1021/es202111n |Environ. Sci. Technol. 2011, 45, 9268–9275
Environmental Science & Technology
Figure 5. Relation between log-transformed deposition of BDE 209 and endosulfan sulfate (ng m2yr1) versus concentration of particles (mgL1) in the snowpack. r = correlation coefficients.
coefficients (Kia) of organic compounds are lower than the corresponding soilair and vegetationair partition coefficients.17 Thus, there is a trend between the SSA of the snowpacks and the ratio of low to high volatility PCBs (∑(153,138,180)/∑(52,101, 118)); however, this trend is not statistically significant (Figure 4b). The differences in SSA between the sampled snowpacks were low, which complicates regression analysis. However, lower SSA values correspond to enhanced selective enrichment of less volatile PCBs, which is consistent with the lower capacity of aged snow to retain the more volatile compounds. Snow aging might also be responsible for the lack of altitudinal correlation observed for HCB and for α-HCH. Both compounds showed higher than expected concentrations in the lowest altitude sites, those with the highest SSA values. Differences between HCH isomers are likely related to the higher volatility and Henry’s Law constant of α-HCH than γ-HCH, which will increase the preferential volatilization of the α-isomer during snow aging (see SI Table S2). Therefore, cold trapping is the dominant factor that influences the distribution of pollutants in the snow accumulated in the studied mountain slope, despite the fact that snow aging partly counteracts this effect in the case of more volatile compounds. The same trapping-altitude trend was also observed for the PBDEs. To the best of our knowledge, this is the first report of statistically significant altitudinal correlations of PBDEs in snow that involve higher concentrations and deposition fluxes at higher elevation. Interestingly, although recent studies on the PBDE content in fish from lakes distributed along an altitudinal transect in the Tatra Mountains did not show any correlation with altitude,22 similar analyses performed on fish from the Pyrenees did show a clear altitude dependence for PBDEs.22,51 The difference observed in these mountain ranges has been attributed to later use of these compounds in the Tatra Mountains than in Pyrenees. The discrepancy between the altitudinal PBDE distributions in snowpacks and fish from the Tatra Mountains may result from differences among these environmental compartments. Snow deposition has a direct scavenging effect on air pollutants and stores the atmospheric deposition accumulated during the cold seasons, when temperatures are low. Contrariwise, chemical accumulation in fish stems from more complex processes during both the cold and warm seasons, and statistically significant temperature dependence may require the combined interaction of concentration differences over long time periods.52 Endosulfan sulfate and BDE 209 did not show a defined altitudinal trend. The deposition flux of each compound showed a statistically significant positive correlation (p < 0.05) with particle concentrations (Figure 5), which is consistent with their low vapor pressure and high octanol-air partition coefficient
ARTICLE
(Koa; see SI Table S2). In the atmosphere these compounds are almost entirely associated to aerosols53 and their long-range transport follows that of particles. Once in the snow, their high Koa keeps them bound to particulate organic matter. Thus, their distribution in the snowpack will be largely controlled by particle abundance.30 Polycyclic Aromatic Hydrocarbons. As observed for organohalogen compounds, good correlations were found between individual PAH deposition fluxes and the altitudes of the sampling sites (Figure 3). Linear regression analyses of log-transformed PAH deposition fluxes and altitudes indicated statistically significant positive correlation coefficients (p < 0.05) (SI Table S2). In addition, a statistically significant positive correlation was found between the ratio of low to high molecular weight PAHs (LMW-PAHs from acenaphthylene to Pyr- and HMW-PAHs from B[a]A to coronene-, respectively), on one hand, and SSA, on the other (p < 0.05, Figure 4c). This result is consistent with the previously mentioned observations with organohalogen compounds and shows that snow aging also involves the preferential loss of the more volatile PAHs. Particle content only showed statistically significant correlations (p < 0.05) for fluorene and acenaphthylene, which contrasts with the important role of particles in the atmospheric transport and fate of PAHs, mainly for HMW compounds.5,54 The relative importance between vapor and particle scavenging by the falling snow is controlled by the relative magnitude of the Koa and the Kia coefficients.30 Despite the relatively high Koa of these compounds, their high Kia values (SI Tables S2) involve a dominance of the vapor over particle scavenging processes due to their relatively strong snow surface sorption. Therefore wet gaseous deposition predominates over particle association.55 Thus, the deposition fluxes of the PAHs found in the Tatra snowpacks correlated well with altitude, but they only correlated well with particle content in a few cases.
’ ASSOCIATED CONTENT
bS
Supporting Information. Further information on analytical methods, mean qualitative PAH distributions in snowpack, PAH isomeric ratios, selected physical-chemical properties of the studied compounds, and regression analysis of the log-transformed deposition of POPs in snow versus the altitude of the sampling site. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +34 93 400 61 22; fax: +34 93 204 59 04; e-mail: pilar.
[email protected]..
’ ACKNOWLEDGMENT We thank P. Alabart, R. Chaler, D. Fanjul, and M. Comesa~ na for their technical assistance in GC and GCMS instrumental analysis. Financial support was provided by the EU Project EUROLIMPACS (GOCE-CT-2003-505540) and GRACCIE (CSD2007-00067). L.A. is thankful for a grant provided jointly by Banco Santander Central Hispano and the CSIC. 9273
dx.doi.org/10.1021/es202111n |Environ. Sci. Technol. 2011, 45, 9268–9275
Environmental Science & Technology
’ REFERENCES (1) Scheringer, M.; Jones, K. C.; Matthies, M.; Simonich, S.; Van De Meent, D. Multimedia partitioning, overall persistence, and long-range transport potential in the context of POPs and PBT chemical assessments. Integr. Environ. Assess. Manage. 2009, 5 (4), 557–576. (2) www.pops.int Stockholm Convention on Persistent Organic Pollutants (POPs)-as amended in 2009 (accessed August 23, 2011). (3) Simonich, S. L.; Hites, R. A. Global distribution of persistent organochlorine compounds. Science 1995, 269, 1851–1854. (4) Wania, F.; Mackay, D. Global fractionation and cold condensation of low volatility organochlorine compounds in polar regions. Ambio 1993, 22, 10–18. (5) Fernandez, P.; Carrera, G.; Grimalt, J. O.; Ventura, M.; Camarero, L.; Catalan, J.; Nickus, U.; Thies, H.; Psenner, R. Factors governing the atmospheric deposition of polycyclic aromatic hydrocarbons to remote areas. Environ. Sci. Technol. 2003, 37, 3261–3267. (6) Carrera, G.; Fernandez, P.; Grimalt, J. O.; Ventura, M.; Camarero, L.; Catalan, J.; Nickus, U.; Thies, H.; Psenner, R. Atmospheric deposition of organochlorine compounds to remote high mountain lakes of Europe. Environ. Sci. Technol. 2002, 36 (11), 2587–2588. (7) Grimalt, J. O.; Fernandez, P.; Berdie, L.; Vilanova, R. M.; Catalan, J.; Psenner, R.; Hofer, R.; Appleby, P. G.; Lien, L.; Rosseland, B. O.; Massabuau, J.-C.; Battarbee, R. W. Selective trapping of organochlorine compounds in mountain lakes of temperate areas. Environ. Sci. Technol. 2001, 35 (13), 2690–2697. (8) Blais, J. M.; Schindler, D. W.; Muir, D. C. G.; Kimpe, L. E.; Donald, D. B.; Rosenberg, B. Accumulation of persistent organochlorine compounds in mountains of western Canada. Nature 1998, 395, 585–588. (9) Davidson, D. A.; Wilkinson, A. C.; Blais, J. M.; Kimpe, L. E.; McDonald, K. M.; Schindler, D. W. Orographic cold-trapping of persistent organic pollutants by vegetation in mountains of western Canada. Environ. Sci. Technol. 2003, 37 (2), 209–215. (10) Grimalt, J. O.; Borghini, F.; Sanchez-Hernandez, J. C.; Barra, R.; Garcia, C. J. T.; Focardi, S. Temperature dependence of the distribution of organochlorine compounds in the mosses of the Andean Mountains. Environ. Sci. Technol. 2004, 38, 5386–5392. (11) Estellano, V. H.; Pozo, K.; Harner, T.; Franken, M.; Zaballa, M. Altitudinal and seasonal variations of persistent organic pollutants in the Bolivian Andes Mountains. Environ. Sci. Technol. 2008, 42 (7), 2528–2534. (12) Wang, P.; Zhang, Q.; Wang, Y.; Wang, T.; Li, X.; Li, Y.; Ding, L.; Jiang, G. Altitude dependence of polychlorinated biphenyls (PCBs) and polybrominated diphenyl ethers (PBDEs) in surface soil from Tibetan Plateau, China. Chemosphere 2009, 76 (11), 1498–1504. (13) Hageman, K. J.; Simonich, S. L.; Campbell, D. H.; Wilson, G. R.; Landers, D. H. Atmospheric deposition of current-use and historic-use pesticides in snow at national parks in the western United States. Environ. Sci. Technol. 2006, 40 (10), 3174–3180. (14) Herbert, B. M. J.; Villa, S.; Halsall, C. J. Chemical interactions with snow: Understanding the behavior and fate of semi-volatile organic compounds in snow. Ecotox. Environ. Saf. 2006, 63 (1), 3–16. (15) Franz, T. P.; Eisenreich, S. J. Snow scavenging of polychlorinated biphenyls and polycyclic aromatic hydrocarbons in Minnesota. Environ. Sci. Technol. 1998, 32 (12), 1771–1778. (16) Lei, Y. D.; Wania, F. Is rain or snow a more efficient scavenger of organic chemicals? Atmos. Environ. 2004, 38 (22), 3557–3571. (17) Stocker, J.; Scheringer, M.; Wegmann, F.; Hungerbuhler, K. Modeling the effect of snow and ice on the global environmental fate and long-range transport potential of semivolatile organic compounds. Environ. Sci. Technol. 2007, 41 (17), 6192–6198. (18) Halsall, C. J. Investigating the occurrence of persistent organic pollutants (POPs) in the arctic: Their atmospheric behaviour and interaction with the seasonal snow pack. Environ. Pollut. 2004, 128 (12), 163–175. (19) Herbert, B. M. J.; Halsall, C. J.; Villa, S.; Jones, K. C.; Kallenborn, R. Rapid changes in PCB and OC pesticide concentrations in Arctic snow. Environ. Sci. Technol. 2005, 39, 2999–3005. (20) Burniston, D. A.; Strachan, W. J. M.; Hoff, J. T.; Wania, F. Changes in surface area and concentrations of semivolatile organic
ARTICLE
contaminants in aging snow. Environ. Sci. Technol. 2007, 41 (14), 4932–4937. (21) Hageman, K. J.; Hafner, W. D.; Campbell, D. H.; Jaffe, D. A.; Landers, D. H.; Simonich, S. M. Variability in pesticide deposition and source contributions to snowpack in western U.S. national parks. Environ. Sci. Technol. 2010, 44 (12), 4452–4458. (22) Gallego, E.; Grimalt, J. O.; Bartrons, M.; Lopez, J. F.; Camarero, L.; Catalan, J.; Stuchlik, E.; Battarbee, R. Altitudinal gradients of PBDEs and PCBs in fish from European high mountain lakes. Environ. Sci. Technol. 2007, 41 (7), 2196–2202. (23) Bartrons, M.; Grimalt, J. O.; Catalan, J. Concentration changes of organochlorine compounds and polybromodiphenyl ethers during metamorphosis of aquatic insects. Environ. Sci. Technol. 2007, 41 (17), 6137–6141. (24) de Wit, C. A.; Alaee, M.; Muir, D. C. G. Levels and trends of brominated flame retardants in the Arctic. Chemosphere 2006, 64 (2), 209–233. (25) Wania, F.; Dugani, C. B. Assessing the long-range transport potential of polybrominated diphenyl ethers: A comparison of four multimedia models. Environ. Toxicol. Chem. 2003, 22 (6), 1252–1261. (26) Carrera, G.; Fernandez, P.; Vilanova, R. M.; Grimalt, J. O. Persistent organic pollutants in snow from European high mountain areas. Atmos. Environ. 2001, 35, 245–254. (27) Fernandez, P.; Vilanova, R. M.; Grimalt, J. O. Sediment fluxes of polycyclic aromatic hydrocarbons in European high altitude mountain lakes. Environ. Sci. Technol. 1999, 33 (21), 3716–3722. (28) Zasadni, J.; Klapyta, P. An attempt to assess the modern and the Little Ice Age climatic snowline altitude in the Tatra Mountains. Landform Anal. 2009, 10, 124–133. (29) Domine, F.; Taillandier, A. S.; Simpson, W. R. A parameterization of the specific surface area of seasonal snow for field use and for models of snowpack evolution. J. Geophys. Res. 2007, 112 (F2), F02031. (30) Daly, G. L.; Wania, F. Simulating the influence of snow on the fate of organic compounds. Environ. Sci. Technol. 2004, 38, 4176–4186. (31) van Drooge, B. L.; Fernandez, P.; Grimalt, J. O.; Stuchlik, E.; García, C. J. T.; Cuevas, E. Atmospheric polycyclic aromatic hydrocarbons in remote European and Atlantic sites located above the boundary mixing layer. Environ. Sci. Pollut. Res. 2010, 17, 1207–1216. (32) Fernandez, P.; Carrera, G.; Grimalt, J. O. Persistent organic pollutants in remote freshwater ecosystems. Aquat. Sci. 2005, 67, 263–273. (33) Vives, I.; Grimalt, J. O.; Fernandez, P.; Rosseland, B. Polycyclic aromatic hydrocarbons in fish from remote and high mountian lakes in Europe and Greenland. Sci. Total Environ. 2004, 324, 67–77. (34) Umlauf, G.; Cristoph, E. H.; Eisenreich, S. J.; Mariani, G.; Paradiz, B.; Vives, I. Seasonality of PCDD/Fs in the ambient air of Malopolska Region, southern Poland. Environ. Sci. Pollut. Res. 2010, 17, 462–469. (35) Usenko, S.; Simonich, S. L. M.; Hageman, K. J.; Schrlau, J. E.; Geiser, L.; Campbell, D. H.; Appleby, P. G.; Landers, D. H. Sources and deposition of polycyclic aromatic hydrocarbons to western U.S. national parks. Environ. Sci. Technol. 2010, 44 (12), 4512–4518. (36) van Drooge, B. L.; Grimalt, J. O.; Torres-García, C.; Cuevas, E. Semivolatile organochlorine compounds in the free troposphere of northeastern Atlantic. Environ. Sci. Technol. 2002, 36 (6), 1155–1161. (37) Aceves, M.; Grimalt, J. O. Seasonally dependent size distributions of aliphatic and polycyclic aromatic hydrocarbons in urban aerosols from densely populated areas. Environ. Sci. Technol. 1993, 27, 2896–2908. (38) Gabrieli, J.; Decet, F.; Luchetta, A.; Valt, M.; Pastore, P.; Barbante, C. Occurrence of PAH in the seasonal snowpack of the eastern Italian Alps. Environ. Pollut. 2010, 158 (10), 3130–3137. (39) Masclet, P.; Mouvier, G.; Nikolaou, K. Relative decay index and sources of polycyclic aromatic hydrocarbons. Atmos. Environ. 1986, 20 (3), 439–446. (40) Quiroz, R.; Grimalt, J. O.; Fernandez, P.; Camarero, L.; Catalan, J.; Stuchlik, E.; Thies, H.; Nichus, U. Polycyclic aromatic hydrocarbons in soils from European high mountain areas. Water, Air, Soil Pollut. 2011, 215, 655–666. (41) LaGuardia, M. J.; Hale, R. C.; Harvey, E. Detailed polybrominated diphenyl ether (PBDE) congener composition of the widely used 9274
dx.doi.org/10.1021/es202111n |Environ. Sci. Technol. 2011, 45, 9268–9275
Environmental Science & Technology
ARTICLE
penta-, octa-, and Deca-PBDE technical flame-retardant mixtures. Environ. Sci. Technol. 2006, 40 (20), 6247–6254. (42) de Wit, C. A.; Herzke, D.; Vorkamp, K. Brominated flame retardants in the Arctic environment—Trends and new candidates. Sci. Total Environ. 2010, 408 (15), 2885–2918. (43) Wang, X.-M.; Ding, X.; Mai, B.-X.; Xie, Z.-Q.; Xiang, C.-H.; Sun, L.-G.; Sheng, G.-Y.; Fu, J.-M.; Zeng, E. Y. Polybrominated diphenyl ethers in airborne particulates collected during a research expedition from the Bohai Sea to the Arctic. Environ. Sci. Technol. 2005, 39 (20), 7803–7809. (44) Jantunen, L. M.; Helm, P. A.; Kylin, H.; Bidleman, T. F. Hexachlorocyclohexanes (HCHs) in the Canadian Archipelago. 2. Airwater gas exchange of a- and g-HCH. Environ. Sci. Technol. 2008, 42 (2), 465–470. (45) van Drooge, B. L.; Grimalt, J. O.; Camarero, L.; Catalan, J.; Stuchlik, E.; Torres-Garcia, J. Atmospheric semivolatile organochlorine compounds in European high mountain areas (central Pyrenees and High Tatras). Environ. Sci. Technol. 2004, 38, 3525–3532. (46) Weber, J.; Halsall, C. J.; Muir, D.; Teixeira, C.; Small, J.; Solomon, K.; Hermanson, M.; Hung, H.; Bidleman, T. Endosulfan, a global pesticide: A review of its fate in the environment and occurrence in the Arctic. Sci. Total Environ. 2010, 408 (15), 2966–2984. (47) http://chm.pops.int/Implementation/NewPOPs/The9newPOPs/tabid/672/language/en-US/Default.aspx (accessed August 2011). (48) Herbert, B. M. J.; Halsall, C. J.; Fitzpatrick, L.; Villa, S.; Jones, K. C.; Thomas, G. O. Use and validation of novel snow samplers for hydrophobic, semi-volatile organic compounds (SVOCs). Chemosphere 2004, 56, 227–235. (49) Grimalt, J. O.; Fernandez, P.; Quiroz, R. Input of organochlorine compounds by snow to European high mountain lakes. Freshwater Biol. 2009, 54, 2533–2542. (50) Fernandez, P.; Grimalt, J. O. On the global distribution of persistent organic pollutants. Chimia 2003, 57 (9), 514–521. (51) Blais, J. M.; Charpentie, S.; Pick, F.; Kimpe, L. E.; Amand, A. S.; Regnault-Roger, C. Mercury, polybrominated diphenyl ether, organochlorine pesticide, and polychlorinated biphenyl concentrations in fish from lakes along an elevation transect in the French Pyrenees. Ecotox. Environ. Saf. 2006, 63 (1), 91–99. (52) Wania, F.; Westgate, J. N. On the mechanism of mountain coldtrapping of organic chemicals. Environ. Sci. Technol. 2008, 42 (24), 9092–9098. (53) Gouin, T.; Thomas, G. O.; Chaemfa, C.; Harner, T.; Mackay, D.; Jones, K. C. Concentrations of decabromodiphenyls ether in air from Southern Ontario: Implications for particle-bound transport. Chemosphere 2006, 64, 256–261. (54) Fernandez, P.; Grimalt, J. O.; Vilanova, R. M. Atmospheric gas particle partitioning of polycyclic aromatic hydrocarbons in high mountain regions of Europe. Environ. Sci. Technol. 2002, 36 (6), 1162–1168. (55) Meyer, T.; Wania, F. Organic contaminant amplification during snowmelt. Water Res. 2008, 42 (89), 1847–1865. (56) Finizio, A.; Villa, S.; Raffaele, F.; Vighi, M. Variation of POP concentrations in fresh-fallen snow and air on an alpine glacier (Monte Rosa). Ecotoxicol. Environ. Saf. 2006, 63 (1), 25–32. (57) Blais, J. M.; Schindler, D. W.; Muir, D. C. G.; Sharp, M.; Donald, D.; Lafreniere, M.; Braekevelt, E.; Strachan, W. M. J. Melting glaciers: A major source of persistent organochlorines to subalpine Bow Lake in Banff National Park, Canada. Ambio 2001, 30 (7), 410–415. (58) Wang, X. P.; Yao, T. D.; Wang, P. L.; Wei, Y.; Tian, L. D. The recent deposition of persistent organic pollutants and mercury to the Dasuopu glacier, Mt. Xixiabangma, central Himalayas. Sci. Total Environ. 2008, 394 (1), 134–143.
9275
dx.doi.org/10.1021/es202111n |Environ. Sci. Technol. 2011, 45, 9268–9275
ARTICLE pubs.acs.org/est
Sorption of Peat Humic Acids to Multi-Walled Carbon Nanotubes Xilong Wang,†,* Liang Shu,† Yanqi Wang,† Bingbing Xu,‡ Yingchen Bai,‡ Shu Tao,† and Baoshan Xing§ †
Laboratory for Earth Surface Processes, College of Urban and Environmental Science, Peking University, Beijing, 100871, China Chinese Research Academy of Environmental Sciences, Beijing, 100012, China § Department of Plant, Soil and Insect Sciences, University of Massachusetts, Amherst, Massachusetts 01003, United States ‡
bS Supporting Information ABSTRACT: Sorption of humic acids (HAs) from a peat soil by multiwalled carbon nanotubes (MWCNTs) was examined in this work. Sorption rate of HAs to MWCNTs was dominantly controlled by their diffusion from liquidMWCNT boundary to MWCNT surfaces. Size exclusion chromatography analysis did not detect preferential sorption of HA fractions to MWCNTs at equilibrium, whereas the components with lower molecular weight in some HA fractions (e.g., HA1) would more preferentially be sorbed to MWCNTs at the initial sorption stage. Equilibrium sorption intensity of HAs by MWCNTs was dependent on their surface area and a sum of meso- and macropore volume. The surface area and sum of meso- and macroporosity-normalized sorption coefficient (Kd) values of a given HA by MWCNTs increased with increasing outer diameter of MWCNTs, because MWCNTs with larger outer diameter were more strongly dispersed by HAs thereby making more sorption sites exposed for HA sorption. Van der Waals interaction between the alkyl components rather than the aromatic ones of HAs with MWCNTs was likely the key driving force for their sorption. This study highlights the sorption rate-controlling step of HAs from a same source to MWCNTs and the major factors affecting their sorption intensity at equilibrium.
’ INTRODUCTION Carbon nanotubes (CNTs) have been attracting increasing interest since their discovery in 1991,1 due primarily to their unique structural, super conductivity, and mechanic properties,2 as well as outstanding thermal and chemical stability.3 CNTs have widely been used as biosensors,4 electronic devices,5 and even supercapacitors.6 They have also been proposed for environmental applications as a potential sorbent to remove organic contaminants due to their relatively high sorption capacity. It was reported that CNTs had much higher removal efficiency for dioxin than activated carbon.7 Our previous study showed that sorption of organic contaminants by CNTs was greatly influenced by the coexisting dissolved organic matter (DOM).8 Studies on the sorption mechanisms of DOM by CNTs are thus necessary and essential for a better understanding of their environmental behaviors. DOM, as a mixture of complex polyelectrolytes, is ubiquitously present in the environment. As reported, sorption of DOM to activated carbon (AC) was regulated by its chemical composition and molecular weight (MW), solution chemistry, as well as surface area, pore size distribution, surface chemistry and pore structure of AC.9 DOM components of smaller molecular size were more favorably sorbed to AC than those with higher MW.9 Kilduff et al.10 used high performance size-exclusion chromatography (HPSEC) to examine the sorption behavior of humic mixtures on AC and reported that HA of smaller molecular size was more preferentially sorbed. It was further r 2011 American Chemical Society
demonstrated that this is a general feature for sorption of polyelectrolytes by AC.9 Hyung et al.11 investigated the effect of sorbent concentration on fractionation of HA components, and they however reported that HA components of higher MW were more preferentially sorbed to the multiwalled carbon nanotube (MWCNTs) at equilibrium as its concentration was increased. It is hypothesized that sorption behavior (e.g., molecular weight selectivity) of HAs by carbonaceous materials is affected by their porosity difference. It was observed that humic acids (HAs) from various sources with higher MW and hydrophobicity had higher sorption by MWCNTs than fulvic acids, and the sorption was linearly correlated with their aromaticity.11 Till now, information concerning the molecular weight selectivity of DOM in its kinetic sorption to CNTs and the key step controlling sorption rate of DOM to CNTs is scarce. Due to extremely complex composition of DOMs, multiple mechanisms such as hydrophobic (nonspecific) interactions, ππ interactions, hydrogen bonding and electrostatic interactions can be involved in their sorption by CNTs, whereas the relative importance of individual mechanisms is DOMdependent. Lin et al.12 reported that the interaction between tannic acid and MWCNTs followed a two-stage sorption model. Received: July 1, 2011 Accepted: September 19, 2011 Revised: September 17, 2011 Published: September 19, 2011 9276
dx.doi.org/10.1021/es202258q | Environ. Sci. Technol. 2011, 45, 9276–9283
Environmental Science & Technology
ARTICLE
Table 1. Solid-State CPMAS 13C NMR Spectrum Data of HAsa distribution of C chemical shift (ppm) (%) sample
6196
96109
109145
145163
163190
190220
aliphatic C (%)
aromatic C (%)
polar C (%)
050
5061
HA1
15.8
5.7
9.6
6.3
26.3
7.8
26.2
2.4
37.4
26.3
57.9
HA2
19.2
4.9
12.4
6.7
22.0
8.7
19.0
7.1
43.2
22.0
58.8
HA3
23.1
4.9
11.3
6.2
20.6
8.3
18.6
6.9
45.5
20.6
56.2
HA4
22.5
7.3
16.8
4.8
18.9
7.6
17.3
4.7
51.4
18.9
58.6
a
Aliphatic C: total aliphatic C region (0109 ppm); Polar C: total polar carbon region (50109 ppm and 145220 ppm); Aromatic carbon region (109145 ppm). The chemical shift for aliphatic C, aromatic C, and polar C assignment is described in Wang et al.15 The experimental error of the integrated 13C NMR data is below 5%.
We hypothesize that sorption of DOM by CNTs is dependent on polarity, chemical composition and MW of DOM as well as the surface area and porosity of CNTs. The driving forces for interactions between HAs with AC and CNTs could be similar since both AC and CNTs are composed of graphene sheets thus similar chemical composition. However, due to distinct arrangement of graphene sheets in AC and CNTs thereby dissimilar surface and structural characteristics, the interaction mechanisms between DOM and these two types of carbonaceous materials could be different. Because only very limited work has been done on DOM sorption by CNTs, importance of the surface area and porosity of CNTs and the chemical composition of DOM in their sorption still remains largely unknown. To better understand the interaction mechanisms between DOM and CNTs, the major objectives of this work were to (1) identify the key process that controls sorption rate of HAs by MWCNTs; (2) examine molecular weight selectivity in kinetic sorption of HAs to MWCNTs, and (3) identify the key driving force and mechanism that regulate sorption of HAs by MWCNTs. To achieve our research aim, humic acid fractions sequentially extracted from a peat soil were used. The advantage of using HAs from a same source was that the difference in sorption characteristics of HAs by MWCNTs was primarily due to their distinct physicochemical properties instead of the source materials. MWCNTs with various diameters were used due to their different surface area and porosity.
’ MATERIALS AND METHODS Sorbates and Sorbents. Multiwalled carbon nanotubes (MWCNTs) of various outer diameters with purity >95% were purchased from Shenzhen Nano-Harbor Co., Ltd., China. Selected properties of the MWCNTs are reported in our recent publication8 and summarized in Table S1 in the Supporting Information (SI). Humic acids were progressively fractionated from a peat soil collected from Amherst, Massachusetts using 0.1 M Na2P4O7 for first to sixth fractions and 0.1 M NaOH for the seventh one. The extracting solution was switched from Na2P4O7 to NaOH mainly because basicity of the latter was stronger, such that HA7 could have quite different properties from other HA fractions extracted earlier. Such an extraction method aimed to get HA fractions with distinct MW, chemical composition (e.g., alkyl and aromatic carbon content) and possibly polarity for investigating the kinetic and equilibrium sorption mechanisms of HAs by MWCNTs as stated above. The first and seventh fractions were respectively marked as HA1 and HA4. There would be little difference in composition between the second and third fractions and between the fourth through
sixth fractions, they were separately combined and labeled as HA2 (second and third fractions) and HA3 (from fourth to sixth fractions), respectively.13 The extracted HAs were treated with 0.1 M HCl/0.3 M HF to remove minerals,14 and then rinsed five times with deionized water, freeze-dried, and finally ground to pass through a 250 μm sieve and stored for characterization and sorption experiments. HA Characterization. Solid-state cross-polarization magic angle-spinning and total-sideband-suppression 13C NMR spectrum of all HAs were obtained using a Bruker DSX-300 spectrometer (Germany) operated at 13C frequency of 75.5 MHz. The NMR running parameters and chemical shift assignments are summarized elsewhere.15 The 13C NMR spectra and the integrated data for all tested HAs are presented in Figure S1 in the SI and Table 1, respectively. Size Exclusion Chromatography Analysis. Molecular weight distribution of both original HA and that remaining in the supernatant of samples at individual time points from kinetic sorption experiments was determined with SEC and compared to probe the molecular weight selectivity of HAs with increasing solute-sorbent contact time. An Agilent 1200 high performance liquid chromatography coupled with a YMC 60 column (300 6 mm i.d.) (YMC Co., Ltd. Japan) and a UV detector was used, and the detecting wavelength was set as 254 nm. The mobile phase was a mixture of 0.03 M NaCl, 0.001 M NaH2PO4 and 0.001 M Na2HPO4 with a flow rate of 0.5 mL/min.16 Equilibrium Sorption of HAs. Sorption isotherms of HAs by MWCNTs were obtained using batch experiments in screw cap vials. To prepare HA stock solutions, 150 mg HA was added to 100 mL 0.1 M NaOH, followed by ultrasonicating the system with a Branson digital sonifier 250 D at 50% amplitude for 10 min, and adjusting pH to 7 using 0.1 M HCl. The solution was then diluted to 1 L to get the HA stock solution with a concentration of 150 mg/L. The background solution contained 200 mg/L NaN3 to inhibit biodegradation. The solid to solution ratio was adjusted to have 2080% HA sorption. Test solutions containing HAs of various concentrations were added to the screw cap vials which contained appropriate amount of preweighed MWCNTs until a minimum headspace was achieved. All vials were mixed on a rotary shaker for 9 days at room temperature, and such a time period was long enough to reach sorption equilibrium. This was verified by the fact that variation of the final solute concentration was below 3% after one additional day shaking in our preliminary test. After mixing, the vials were centrifuged at 5000 rpm for 30 min, followed by filtering the supernatant through 0.45 μm hydrophilic polyethersulfone (PES) syringe. No sorption of HAs by this filter was observed in our preliminary test. Finally, HA concentrations in the supernatant were measured using a UV 9277
dx.doi.org/10.1021/es202258q |Environ. Sci. Technol. 2011, 45, 9276–9283
Environmental Science & Technology
ARTICLE
Figure 1. Sorption kinetics of HAs by MWCNTs. MWCNT20 (0); MWCNT60 (4); MWCNT100 (O).
visible spectrometer following the method used in our previous study.8 All samples along with blanks were run in duplicate. The sorbed amount of HA to MWCNTs was calculated by mass difference, due to its negligible mass loss. HA Sorption Kinetics. Sorption kinetics of HAs by MWCNTs was also performed using batch experiments in screw cap vials. The solid to solution ratio, HA stock solution preparation, sample centrifugation at each time point, and the approach for HA concentration determination were identical to those for the equilibrium sorption systems as stated above. To avoid changing the solid to solution ratio of the kinetic sorption systems in the whole experimental period, duplicate samples with the same condition were prepared for each sampling time point, and at each time point the postsampled samples were removed. The blanks for each time point were run in replicate as well. Kinetic and Equilibrium Sorption Data Fitting. The Lagergren pseudo first- and second-order models and WeberMorris model have widely been used to describe sorption kinetics of sorbates (e.g., DOMs and organic compounds) from liquid to CNTs and activated carbons.17 These models can be grouped into reaction-based models (e.g., Lagergren pseudo first- and second order models) and diffusion-based models (e.g., WeberMorris model), and they were used for kinetic sorption data fitting in this work. Lagergren pseudo first order model: Qt ¼ Qe ð1 eK1 t Þ
ð1Þ
where Qt and Qe are sorbed amount of HAs to MWCNTs at time point t and equilibrium (mg/g), respectively. K1 is the Lagergren rate constant (1/h) and t is time (h). Lagergren pseudo second order model: 1 1 1 1 1 1 1 1 þ þ ¼ ¼ 3 3 3 2 Qt K2 Q e t Qe V0 t Qe
ð2Þ
where K2 refers to the sorption rate constant of the pseudo second order kinetic model (g/mg/h). V0 is the initial sorption rate (mg/g/h). Weber-Morris model: Qt ¼ A þ Ka t 0:5
ð3Þ
where A is the intercept of the vertical axis (mg/g), and Ka is the overall diffusion constant for sorption (mg/(g 3 h0.5)).18 The Langmuir and Freundlich models were employed to fit the equilibrium sorption data of HAs by MWCNTs. Langmuir model: Q ¼
Qm BCe 1 þ BCe
ð4Þ
where Q and Qm are sorbed amount of HAs to MWCNTs at equilibrium and the maximum sorption capacity of HAs (mg/g), respectively. B is the affinity parameter (mL/mg). Ce is the HA concentration in aqueous phase at equilibrium (mg/mL). Freundlich model: Q ¼ K f 3 Ce n
ð5Þ
where Kf is the sorption coefficient ((mg/g)/(mg/mL)n), and n is a constant often used as an indicator of isotherm nonlinearity. Sorption coefficient (Kd) of HAs by MWCNTs was calculated with the equation Kd = Q/Ce.
’ RESULTS AND DISCUSSION HA Sorption Kinetics. The assumption for the Lagergren pseudo first- and second order models is that the difference in sorbed concentration of sorbate at equilibrium (Qe) and that at time t (Qt) is the key driving force for sorption,18 and the sorption capacity is proportional to the number of active sorption sites occupied on the sorbent.17 Three steps are considered in Lagergren kinetic models. They are (1) the sorbate molecules diffuse from liquid phase to liquidsolid boundary; (2) the sorbate molecules move from liquidsolid boundary to solid surfaces; and (3) the sorbate molecules diffuse into the particles.17,19 Diffusion of sorbate molecules can either be an intraparticle, surface diffusion process or both. For our case, intraparticle diffusion of HAs is not applicable because the pore size in between the layers of MWCNTs is not large enough to accommodate HA molecules. However, it is possible that HAs are able to get to a portion of sites in the interstitial spacing between neighboring nanotubes and in the grooves and curved surfaces of the periphery of nanotube bundles.20 Sorption kinetics of HAs by MWCNTs with distinct outer diameters are presented in Figure 1, and the kinetic sorption data were fitted with Lagergren pseudo first- and second order models as well as the diffusion-based Weber-Morris model. It was evident that kinetic sorption data of HAs by MWCNTs were better fitted with the Lagergren pseudo second order model relative to the pseudo first order model as indicated by higher R2 values (Table 2). The Lagergren pseudo second order model-derived sorption rate of HA1 and HA2 by MWCNTs was approximately 10 mg/g/ h and that of HA3 and HA4 was around 20 mg/g/h at 1 h, and they respectively steeply decreased to 2 and around 3 mg/g/h at 3 h. After 1d, all sorption rates dropped to approximately 0.05 mg/g/h. Reduction in sorption rate of HAs by MWCNTs with increasing contact time can be a result of gradual saturation of sorption sites. Diffusion of HA molecules from aqueous phase to the liquidMWCNT boundary should be much faster than other processes for our case, because the Brownian motion of HA molecules in 9278
dx.doi.org/10.1021/es202258q |Environ. Sci. Technol. 2011, 45, 9276–9283
a
1.3 ( 0.2
1.1 ( 0.2
48.3 ( 1.2
46.0 ( 1.4
MWCNT60
MWCNT100
9279
186 ( 35b
125 ( 16
94 ( 10
60.9 ( 1.2b
49.8 ( 1.1
47.0 ( 1.1
MWCNT20
MWCNT60
MWCNT100
1.2 ( 0.1b
1.0 ( 0.2
1.2 ( 0.2
50.6 ( 0.9b
40.0 ( 1.2
40.0 ( 1.5
MWCNT20
MWCNT60
MWCNT100
0.847
0.777
0.919
R 2
0.904
0.830
0.741
R 2
0.683
0.582
0.344
R
2
27.4 ( 1.2
29.6 ( 1.3
40.1 ( 1.4b
A
40.4 ( 2.1
42.4 ( 1.3
54.0 ( 1.4b
Qe
41.4 ( 2.5
41.5 ( 1.6
52.7 ( 1.8b
Qe
0.488
0.579
0.516
R 2
61.4 ( 2.5
61.0 ( 1.4
84.0 ( 1.9b
Qe
0.735
0.873
0.839
R
1.5 ( 0.1
1.4 ( 0.2
1.5 ( 0.2b
0.928
0.864
0.859
46.9 ( 2.6
49.3 ( 2.1
60.8 ( 4.3b
0.030
0.031
0.008
K2
1.7 ( 0.4
1.3 ( 0.3
2.1 ( 0.5b
Kd
R
120 ( 13
124 ( 9
Kd
A
63.6 ( 1.9
63.0 ( 0.9
60 ( 6b
V0
HA3
HA3 2
Qe 85.5 ( 1.8b
WeberMorris model
0.036
0.037
0.035
K2 2
1.0 ( 0.2
1.1 ( 0.2
0.4 ( 0.04b
K1
HA3
HA2
58 ( 12
67 ( 8
103 ( 13b
V0
HA2
Lagergren pseudo second order model
0.5 ( 0.1
0.9 ( 0.2
1.0 ( 0.2b
K1
HA2
K2 was derived from V0 and Qe. b Standard errors of Qe, K1, V0, A, and Kd, respectively.
Kd
0.043
0.050
0.050
A
HA1
V0
Qe
K2
1.5 ( 0.4b
59.2 ( 1.5b
MWCNT20
HA1
K1
Qe
HA1
Lagergren pseudo first order model
Table 2. Parameters for Lagergren Pseudo First- and Second Order As Well As Webermorris Modelsa
0.759
0.673
0.706
R
2
0.920
0.944
0.936
R
2
0.681
0.713
0.857
R
2
66.0 ( 1.3
70.6 ( 3.9
92.8 ( 3.0b
A
80.1 ( 1.8
87.1 ( 1.9
106.5 ( 1.5b
Qe
78.9 ( 2.0
84.6 ( 1.8
104.2 ( 1.6b
Qe
0.050
0.022
0.033
K2
1.5 ( 0.2
1.5 ( 0.5
1.3 ( 0.4b
Kd
HA4
318 ( 66
170 ( 18
379 ( 46b
V0
HA4
1.7 ( 0.3
1.0 ( 0.1
1.6 ( 0.2b
K1
HA4
0.907
0.425
0.490
R2
0.699
0.887
0.847
R2
0.442
0.756
0.690
R2
Environmental Science & Technology ARTICLE
dx.doi.org/10.1021/es202258q |Environ. Sci. Technol. 2011, 45, 9276–9283
Environmental Science & Technology aqueous phase under shaking was fast, which strongly facilitated them to approach the liquid-MWCNT boundary. After approaching the liquid-MWCNT boundary, the HA molecules then moved to the external MWCNT surfaces. This step along with diffusion in the interstitial spacing between neighboring nanotubes and in the grooves and curved surfaces of the periphery of nanotube bundles are hypothesized to control the sorption rate of HA molecules to MWCNTs. The kinetic sorption data showed that sorption rate constant (K2) of HA1 and HA2 was generally higher than that of HA3 and HA4 (Table 2). This may be because HA1 and HA2 had smaller molecular weight than HA3 and HA4 as indicated by the SEC data (SI Figure S2). HA1 and HA2 would thus have lower steric hindrance that may prevent it from approaching and further interacting with MWCNTs. However, we are still unable to conclude whether the diffusion process of HA molecules in pores or that from liquid-MWCNT boundary to MWCNT surfaces dominantly controlled their sorption rate to MWCNTs. Further analysis showed that, the kinetic sorption data of HAs by MWCNTs were also well fitted with the WeberMorris model. As reported, if the regression of Qt against t0.5 is linear and the regression line passes through the origin, then it can be concluded that the diffusion in pores is the rate-controlling step.21 Our WeberMorris model results revealed that Qt of HAs to all MWCNTs was linearly correlated with t0.5 (p e 0.05), and all regression lines had positive intercepts (Figure 1, Table 2), suggesting that diffusion in pores may not be a rate-controlling step for kinetic sorption of HAs to MWCNTs. Therefore, it can be concluded that sorption kinetics of HAs to MWCNTs was regulated by surface-diffusion mechanism and the diffusion rate of HA molecules from liquidMWCNT boundary to MWCNT surfaces dominantly determined their sorption rate. The surfacediffusion mechanism could most likely be pore-size independent, because MWCNT20, MWCNT60 and MWCNT100 had different pore size and porosity whereas they had comparable sorption rate constant (K2) for HA1 and HA2 and the sorption rate constant of HA3 by MWCNT60 and MWCNT100 was comparable as well (Table 2). Preferential Sorption. Previous studies showed that molecular size of DOM was an important factor that affected its sorption by AC.9,10 Sorption of humic substances by activated carbon showed an inverse molecular size dependence.22 Kilduff et al.10 compared sorption of several polyelectrolytes including humic acids extracted from peat and soil, synthesized polymaleic acid, and natural organic matter by activated carbon at equilibrium, and they found that DOM components of smaller molecular size were sorbed to a greater extent when activated carbon concentration was increased. The same phenomenon occurred on sorption of humic acid extracted from Laurentian soil and synthesized polystyrene sulfonate by granular activated carbon.9 Preferential sorption of HA components with low molecular weight to AC was ascribed to the size exclusion effect of pores in its structure.11 Particularly, small HA molecules can access both small and large pores, whereas those of large molecular weight can only access large pores. Different from activated carbons, the larger molecular-sized components of natural organic matter from Suwannee River and the humic and fulvic standards from this river were sorbed to a greater extent by MWCNT at equilibrium when its concentration was increased.11 The difference in molecular weight selectivity for sorption of DOM by these two types of carbonaceous materials at equilibrium could result from their pore structure difference. Activated carbon had
ARTICLE
porous structure which offered a large amount of pores for polyelectrolytes with small molecular size to penetrate into.23 It is most likely that sorption of DOM (e.g., HAs) by MWCNTs is dominated by surface-sorption mechanism, otherwise preferential sorption of DOMs with lower molecular weight should have happened at equilibrium as the concentration of MWCNTs was increased. This is contrary to the previous experimental observations showing that natural organic matter with larger molecular size was preferentially sorbed to MWCNTs at equilibrium when its concentration was increased as stated above.11 The HPSEC data showed that the retention time of HAs followed an order: HA1 > HA2 > HA3 ≈ HA4, suggesting that their molecular weight increased in an order: HA1 < HA2 < HA3 ≈ HA4 (SI Figure S2). HPSEC chromatogram of the residue HA1, HA2, HA3, and HA4 in the supernatant of the HAMWCNT sorption systems at different contact times are presented in SI Figures S3, S4, S5, and S6. It was clear that the components of smaller molecular weight in HA1 were more preferentially sorbed to MWCNT20 in the first 5 h, because the molecular weight of residues remaining in aqueous phase increased with increasing contact time. After 5 h, no molecular weight selectivity was detected. In contrast, more preferential sorption of the smaller molecular size components in HA1 by MWCNT60 and MWCNT100 happened only in the first 1 h contact, and such a time period was much shorter than to MWCNT20 (SI Figure S3). However, no clear molecular weight selectivity was observed for kinetic sorption of HA2, HA3 and HA4 by individual MWCNTs in the whole experimental period (SI Figures S4, S5, and S6). The difference in molecular weight selectivity in sorption of HA1, HA2, HA3, and HA4 by individual MWCNTs revealed that kinetic sorption behavior of HA by MWCNTs was dependent on its chemical composition and molecular size distribution. Since HA1 was the first fraction isolated from the peat soil, it should have had more heterogeneous composition and diverse components that were readily dissolved in Na2P4O7 thus a broader range in molecular weight distribution as compared to other HAs. A portion of HA1 molecules with small molecular size was able to be preferentially sorbed to the sites in the interstitial spacing between neighboring nanotubes and in the grooves and curved surfaces of the periphery of nanotube bundles due to their low steric hindrance.20,24 It was however difficult for HA1 molecules to get into the pores between layers inside the nanotubes due mainly to their molecular size thus the diffusion limitation. As the HA extraction proceeds, its organic matter composition and molecular weight tends to be more homogeneous.14 This can be clearly indicated by the difference in SEC data of all tested HAs. Particularly, HA1 has a relatively flat and broad SEC spectrum, and the spectra of HAs become narrower with progressive extraction, suggesting the more homogeneous composition.25 Meanwhile, the molecular weight of HA molecules tends to be larger (SI Figure S2). Hence, all components in HA2, HA3, and HA4 were uniformly sorbed to individual MWCNTs. It can be concluded that fractionation of components in the tested HAs by MWCNTs at equilibrium may not happen regardless of the molecular weight and composition of HAs and porosity of MWCNTs, whereas the molecular weight selectivity could occur for some fractions (e.g., HA1) at the initial sorption stage and it may be affected by the porosity of MWCNTs (SI Figures S3, S4, S5, and S6). Equilibrium Sorption of HAs. Sorption isotherms of HAs by MWCNTs are presented in Figure 2, and the isotherm data were fitted with Freundlich and Langmuir models (Table 3). Sorption 9280
dx.doi.org/10.1021/es202258q |Environ. Sci. Technol. 2011, 45, 9276–9283
Environmental Science & Technology
ARTICLE
Figure 2. Sorption isotherms of humic acids by MWCNTs. MWCNT20 (0), MWCNT60 (4), MWCNT100 (O).
Table 3. Langmuir and Freundlich Model Parameters for Sorption of HAs by MWCNTsa Langmuir model MWCNTs
HAs
Qm
B b
Freundlich model R
2
b
b
Kd/SA
Kd/Vmeso+macro
0.946 0.854
1340 1160
11 16
3670 7190
62.8 ( 1.4
321 ( 47
0.942
111.6 ( 12.5
0.211 ( 0.033
0.810
1190
20
10400
HA2
60.8 ( 1.7
449 ( 88
0.877
105.3 ( 8.6
0.203 ( 0.024
0.871
1150
9
3150
MWCNT60
45.9 ( 1.0
553 ( 86
0.892
65.9 ( 6.0
0.140 ( 0.026
0.693
867
12
5350
MWCNT100
41.6 ( 1.7
174 ( 16
0.937
100.1 ( 3.6
0.322 ( 0.011
0.986
763
13
6690 4580
MWCNT100 MWCNT20
0.234 ( 0.017 0.187 ( 0.024
Kd
71.7 ( 1.4 59.6 ( 0.7
0.960 0.978
134.7 ( 7.8 101.9 ( 8.9
b
R2
HA1
MWCNT20 MWCNT60
278 ( 33 472 ( 39
n
Kf
84.8 ( 3.4
413 ( 94
0.866
151.7 ( 7.2
0.200 ( 0.013
0.967
1670
13
MWCNT60
65.6 ( 1.4
1558 ( 335
0.784
97.3 ( 2.6
0.125 ( 0.007
0.961
1340
18
8260
MWCNT100 MWCNT20
64.5 ( 1.2 89.8 ( 1.5
2802 ( 575 869 ( 126
0.795 0.881
92.4 ( 2.7 127.3 ( 4.5
0.107 ( 0.007 0.117 ( 0.009
0.947 0.934
1340 1790
23 14
11800 4930
MWCNT20
HA3
HA4
MWCNT60
76.8 ( 1.2
761 ( 118
0.855
103.2 ( 5.5
0.101 ( 0.014
0.826
1530
21
9410
MWCNT100
64.9 ( 1.5
4984 ( 1401
0.671
93.0 ( 3.0
0.104 ( 0.008
0.929
1360
23
12000
; a Qm: mg/g; B: mL/mg. b Standard errors of Qm, B, Kf, and n, respectively. Kd was derived from the equilibirium concentration of HAs at 0.05 mg/mL. coefficients (Kd) of HAs by MWCNTs were calculated from Freundlich model with equilibrium concentration of HAs at 0.05 mg/mL. It was clear that all MWCNTs had nonlinear sorption for HAs due to their heterogeneous sorption sites. Microporefilling was reported to be a key mechanism for sorption of small hydrophobic organic molecules by chars at low solute concentrations.26 However, due to large molecular weight thus high steric hindrance of HAs, it was very difficult for a majority of HA molecules to enter micropores in MWCNTs. Therefore, micropore-filling could not be able to contribute greatly to HA sorption by MWCNTs. Thurman et al.27 used small-angle X-ray scanning technique to determine the molecular size of aquatic HAs, and reported that their molecular size (i.e., radius of gyration) was in 0.473.3 nm. It was further demonstrated that size exclusion in pores in sorbents (i.e., molecular sieve carbon) occurred with average diameter smaller than 1.7 times the sorbate molecule’s second-widest dimension,28 and a tested NOM was reported to be sorbed to pores with size of 310 nm.29 The meso- and macropores in MWCNTs were therefore large enough to accommodate HA molecules. It is hypothesized that equilibrium sorption intensity of HAs by MWCNTs is dependent on their surface area and a sum of meso- and macroporosity. The surface area and sum of meso- and macroporosity of MWCNTs decreased with increasing outer diameter, reflecting the reduction in number of effective sites for HA sorption. Following this sequence, sorption capacities of HAs by MWCNTs predicted with Lagergren pseudo second order kinetic model (Qe) and
Langmuir model (Qm) also decreased, supporting the importance of these two factors in HA sorption. The Kd values for sorption of HA1, HA2, HA3, and HA4 by individual MWCNTs at equilibrium concentration of 0.05 mg/mL ranged in 11601340 mL/g, 7631150, 13401670 mL/g, and 1360 1790 mL/g, respectively (Table 3). Stareka et al.30 reported that sorption of HAs by activated carbon was much higher than that by activated charcoal cloth, and only the meso- and macropores were accessible but the micropores were inaccessible to HAs. However, HA was almost not sorbed to chars even though its concentration was increased from 0.05 to 1 mg/mL, and the same phenomenon also occurred for fulvic acid until its concentration reached 0.2 mg/mL.31 This suggested that char had very low sorption affinity for humic substance, quite different from MWCNTs and ACs although they were all comprised of graphene sheets. The polar moieties in HAs can interact with the O-containing hydrophilic functionalities at the graphene surfaces of MWCNTs through hydrogen bonding,12 thus enhancing their sorption. However, due to relatively low oxygen content of all MWCNTs (00.09%), the difference in hydrogen bonding between MWCNTs of various outer diameters and a given HA was not big enough to interpret their sorption difference. All HAs had comparable polar carbon content (i.e., from 56.2 to 58.8%), which also could not interpret the difference in their sorption by individual MWCNTs (Table 3). Therefore, hydrogen bonding may not be an important driving force for HA sorption to MWCNTs. 9281
dx.doi.org/10.1021/es202258q |Environ. Sci. Technol. 2011, 45, 9276–9283
Environmental Science & Technology The surface area and sum of meso- and macroporositynormalized Kd values of a given HA by MWCNTs increased with an increase in their outer diameter, implying that the MWCNTs with larger outer diameter had higher sorption for HAs on per surface area and porosity basis relative to those of smaller outer diameter. This can be a result that MWCNTs of larger outer diameter were more strongly dispersed by HAs in contrast to those of smaller outer diameter, which was visually observed in our experiment. Likewise, Lin et al.12 also observed that in the presence of 5 mg/L DOM (i.e., tannic acid), dispersibility of MWCNTs was enhanced in an order: MWCNT100 > MWCNT60 > MWCNT40 > MWCNT20. Dispersion increased accessibility of a portion of sorption sites in MWCNTs that previously were inaccessible to DOM molecules, thus increasing their effective surface area and porosity for DOM sorption.32,33 The ππ interaction was reported to be an important mechanism governing sorption of both π-acceptors and π-donors to graphene surfaces of MWCNTs and black carbons.34,35 All tested HAs were rich in aromatic moieties, which were theoretically able to interact with the electron-rich sites at the graphene surfaces of MWCNTs via ππ interactions, thereby enhancing their sorption. Our 13C NMR data showed that the aromatic carbon content of HAs decreased with sequential extraction, with values ranging from 26.3% for HA1 to 18.9% for HA4 (Table 1). However, sorption of HAs by all MWCNTs generally increased from HA1 to HA4 (Table 3), indicating that HA sorption to MWCNTs could not be dominantly regulated by ππ interaction mechanism. Lin and Xing36 investigated roles of aromatic structure and OH substitution in sorption of polar aromatic compounds to MWCNTs using cyclohexanol, phenol, catechol, pyrogallol, 2-phenylphenol and 1-naphthol as sorbates. They reported that sorption of polar aromatics by MWCNTs was more strongly enhanced with increasing number of electron-donating substitution (OH), with an order of phenol (1-OH) < catechol (2-OH) < pyrogallol (3-OH). A previous study also revealed that the electron-attracting group (NO2) on benzene derivatives enhanced their sorption to MWCNTs.37 These studies noted the importance of electron donating- and attracting-substituents on polar aromatics in their sorption by MWCNTs through ππ interactions. Hydrophobic interaction is another mechanism affecting sorption of HOCs and DOM by carbonaceous materials, which always acts simultaneously with other driving forces.34 MWCNTs were extremely hydrophobic as indicated by their very low oxygen contents. HAs had abundant alkyl and aromatic moieties (Table 1), which can interact with the graphitic sheets of MWCNTs via hydrophobic (or nonspecific) interactions thereby facilitating their sorption. There was no clear difference in hydrophobic carbon (a sum of alkyl and aromatic carbon) content of all tested HAs (Table 1).15 The alkyl carbon contents followed an order: HA4 ≈ HA3 > HA2 > HA1, and the aromatic carbon content of HAs decreased with progressive extraction (Table 1). The Kd values of individual HAs by a given MWCNT increased in an order: HA2 < HA1 < HA3 < HA4 (Table 3), implying that the van der Waals force (or nonspecific hydrophobic interaction) between the alkyl components in HAs and MWCNTs was mainly responsible for their sorption. HA2 gave higher Kd values by a given MWCNT in comparison with HA1, due mainly to its larger molecular weight thus higher steric hindrance that more strongly suppressed its sorption (SI Figure S2). Consistent with the findings of HAs, our recent work38 also suggested that sorption of organic chemicals with vast difference
ARTICLE
in hydrophobicity, electron polarizability, aromaticity, polarity and molecular size (i.e., phenanthrene, lindane, and atrazine) by MWCNTs was dominantly regulated by hydrophobic interactions regardless of their chemical structure. Environmental Implications. The diffusion of HA molecules from liquid-MWCNT boundary to MWCNT surfaces was identified to be a rate-limiting step controlling their sorption. This can be a key point for elucidating the mechanisms governing kinetic sorption of natural organic matter (e.g., HAs) to MWCNTs. Dispersion of MWCNTs would increase their effective surface area and porosity, thereby enhancing HA sorption, and such an effect is more pronounced for MWCNTs with larger outer diameter. Sorption of natural organic matter (e.g., HAs) by MWCNTs is driven mainly by the van der Waals force between its alkyl components rather than the aromatic ones with MWCNTs. Such information may help better understand the interaction mechanisms between HAs and MWCNTs, thus better predict the transport and fate of both MWCNTs and the organic contaminants and heavy metals associated with these sorbents in the natural environment.
’ ASSOCIATED CONTENT
bS
Supporting Information. Selected properties of MWCNTs (Table S1); Solid-state 13C NMR spectra of HAs (Figure S1); Size exclusion chromatograms spectra of HAs (Figure S2); Size exclusion chromatograms spectra of the residue HA1 (Figure S3), HA2 (Figure S4), HA3 (Figure S5), and HA4 (Figure S6) in solution after contact with MWCNTs for different times. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (86) 10-62757822; fax: (86) 10-62767921; e-mail:
[email protected].
’ ACKNOWLEDGMENT This study was supported by the National Natural Science Foundation of China (40971246, 40730737, and 40710019001), the Startup Fund for the Peking University 100-Talent Program, the National Basic Research Program (2007CB407301), BARD (IS-4353-10), USDA Hatch program (MAS 00978) and USDA NIFA AFRICGP (2011-67006-30181). ’ REFERENCES (1) Iijima, S. helical microtubules of graphitic carbon. Nature 1991, 354, 56–58. (2) Zhang, Z.; Lieber, C. M. Nanotube structure and electronicproperties probed by scanning-tunneling-microscopy. Appl. Phys. Lett. 1993, 62, 2792–2794. (3) Blase, X.; Rubio, A.; Louie, S. G.; Cohen, M. L. Stability and band-gap constancy of boron-nitride nanotubes. Europhys. Lett. 1994, 28, 335–340. (4) Hrapovic, S.; Liu, Y. L.; Male, K. B.; Luong, J. H. T. Electrochemical biosensing platforms using platinum nanoparticles and carbon nanotubes. Anal. Chem. 2004, 76, 1083–1088. (5) Bachtold, A.; Hadley, P.; Nakanishi, T.; Dekker, C. Logic circuits with carbon nanotube transistors. Science 2001, 294, 1317–1320. (6) Frackowiak, E.; Beguin, F. Carbon materials for the electrochemical storage of energy in capacitors. Carbon 2001, 39, 937–950. 9282
dx.doi.org/10.1021/es202258q |Environ. Sci. Technol. 2011, 45, 9276–9283
Environmental Science & Technology (7) Long, R. Q.; Yang, R. T. Carbon nanotubes as superior sorbent for dioxin removal. J. Am. Chem. Soc. 2001, 123, 2058–2059. (8) Wang, X. L.; Tao, S.; Xing, B. S. Sorption and competition of aromatic compounds and humic acid on multiwalled carbon nanotubes. Environ. Sci. Technol. 2009, 43, 6214–6219. (9) Kilduff, J. E.; Karanfil, T.; Weber, W. J. Competitive interactions among components of humic acids in granular activated carbon adsorption systems: Effects of solution chemistry. Environ. Sci. Technol. 1996, 30, 1344–1351. (10) Kilduff, J. E.; Karanfil, T.; Chin, Y. P.; Weber, W. J. Adsorption of natural organic polyelectrolytes by activated carbon: A size-exclusion chromatography study. Environ. Sci. Technol. 1996, 30, 1336–1343. (11) Hyung, H.; Kim, J. H. Natural organic matter (NOM) adsorption to multi-walled carbon nanotubes: Effect of NOM characteristics and water quality parameters. Environ. Sci. Technol. 2008, 42, 4416–4421. (12) Lin, D. H.; Xing, B. S. Tannic acid adsorption and its role for stabilizing carbon nanotube suspensions. Environ. Sci. Technol. 2008, 42, 5917–5923. (13) Ghosh, S.; Mashayekhi, H.; Bhowmik, P.; Xing, B. S. Colloidal stability of Al2O3 nanoparticles as affected by coating of structurally different humic acids. Langmuir 2010, 26, 873–879. (14) Kang, S. H.; Xing, B. S. Phenanthrene sorption to sequentially extracted soil humic acids and humins. Environ. Sci. Technol. 2005, 39, 134–140. (15) Wang, X. L.; Cook, R.; Tao, S.; Xing, B. S. Sorption of organic contaminants by biopolymers: Role of polarity, structure and domain spatial arrangement. Chemosphere 2007, 66, 1476–1484. (16) Fu, P. Q.; Wu, F. C.; Liu, C. Q.; Wei, Z. Q.; Bai, Y. C.; Liao, H. Q. Spectroscopic characterization and molecular weight distribution of dissolved organic matter in sediment porewaters from Lake Erhai, Southwest China. Biogeochemistry 2006, 81, 179–189. (17) Ho, Y. S.; McKay, G. Pseudo-second order model for sorption processes. Process. Biochem. 1999, 34, 451–465. (18) Shen, X. E.; Shan, X. Q.; Dong, D. M.; Hua, X. Y.; Owens, G. Kinetics and thermodynamics of sorption of nitroaromatic compounds to as-grown and oxidized multiwalled carbon nanotubes. J. Colloid Interface Sci. 2009, 330, 1–8. (19) Ho, Y. S.; Ng, J. C. Y.; Mckay, G. Kinetics of pollutant sorption by biosorbents: Review. Sep. Purif. Method. 2000, 29, 189–232. (20) Agnihotri, S.; Rostam-Abadi, M.; Rood, M. J. Temporal changes in nitrogen adsorption properties of single-walled carbon nanotubes. Carbon 2004, 42, 2699–2710. (21) Svilovic, S.; Rusic, D.; Basic, A. Investigations of different kinetic models of copper ions sorption on zeolite 13X. Desalination 2010, 259, 71–75. (22) Summers, R. S.; Roberts, P. V. Activated carbon adsorption of humic substances. 2. Size exclusion and electrostatic interactions. J. Colloid Interface Sci. 1988, 122, 382–397. (23) Martinez, M. L.; Torres, M. M.; Guzman, C. A.; Maestri, D. M. Preparation and characteristics of activated carbon from olive stones and walnut shells. Ind. Crops Products 2006, 23, 23–28. (24) Wang, X. L.; Lu, J. L.; Xing, B. S. Sorption of organic contaminants by carbon nanotubes: Influence of adsorbed organic matter. Environ. Sci. Technol. 2008, 42, 3207–3212. (25) Wang, L. Y.; Wu, F. C.; Zhang, R. Y.; Li, W.; Liao, H. Q. Characterization of dissolved organic matter fractions from Lake Hongfeng, Southwestern China Plateau. J. Environ. Sci. 2009, 21, 581–588. (26) Wang, X. L.; Xing, B. S. Sorption of organic contaminants by biopolymer-derived chars. Environ. Sci. Technol. 2007, 41, 8342–8348. (27) Thurman, E. M.; Wershaw, R. L.; Malcolm, R. L.; Pinckney, D. J. Molecular size of aquatic humic substances. Org. Geochem. 1982, 4, 27–35. (28) Kasaoka, S.; Sakata, Y.; Tanaka, E; Naitoh, R. Design of molecular sieve carbon. Studies on the adsorption of various dyes in the liquid phase. Int. Chem. Eng. 1989, 29, 734–742. (29) Ebie, K.; Li, F.; Hagishita, T. Effect of pore size distribution of activated carbon on the adsorption of humic substances and trace organic compounds. Water Supply 1995, 13, 65–70.
ARTICLE
(30) Stareka, J.; Zukala, A.; Rathousky, J. Comparison of the adsorption of humic acids from aqueous solutions on active carbon and activated charcoal cloths. Carbon 1994, 32, 207–211. (31) Pignatello, J. J.; Kwon, S.; Lu, Y. F. Effect of natural organic substances on the surface and adsorptive properties of environmental black carbon (char): Attenuation of surface activity by humic and fulvic acids. Environ. Sci. Technol. 2006, 40, 7757–7763. (32) Gai, K.; Shi, B. Y.; Yan, X. M.; Wang, D. S. Effect of dispersion on adsorption of atrazine by aqueous suspensions of fullerenes. Environ. Sci. Technol. 2011, 45, 5959–5965. (33) Li, Q. L.; Snoeyink, V. L.; Marinas, B. J.; Campos, C. Pore blockage effect of NOM on atrazine adsorption kinetics of PAC: The roles of PAC pore size distribution and NOM molecular weight. Water. Res. 2003, 37, 4863–4872. (34) Pan, B.; Xing, B. S. Adsorption mechanisms of organic chemicals on carbon nanotubes. Environ. Sci. Technol. 2008, 42, 9005–9013. (35) Zhu, D. Q.; Hyun, S. H.; Pignatello, J. J.; Lee, L. S. Evidence for pi-pi electron donor-acceptor interactions between pi-donor aromatic compounds and pi-acceptor sites in soil organic matter through pH effects on sorption. Environ. Sci. Technol. 2004, 38, 4361–4368. (36) Lin, D. H.; Xing, B. S. Adsorption of phenolic compounds by carbon nanotubes: Role of aromaticity and substitution of hydroxyl groups. Environ. Sci. Technol. 2008, 42, 7254–7259. (37) Chen, W.; Duan, L.; Zhu, D. Q. Adsorption of polar and nonpolar organic chemicals to carbon nanotubes. Environ. Sci. Technol. 2007, 41, 8295–8300. (38) Wang, X. L.; Liu, Y.; Tao, S.; Xing, B. S. Relative importance of multiple mechanisms in sorption of organic compounds by multiwalled carbon nanotubes. Carbon 2010, 48, 3721–3728.
9283
dx.doi.org/10.1021/es202258q |Environ. Sci. Technol. 2011, 45, 9276–9283
ARTICLE pubs.acs.org/est
Analysis of Nanoparticle Agglomeration in Aqueous Suspensions via Constant-Number Monte Carlo Simulation Haoyang Haven Liu,†,‡ Sirikarn Surawanvijit,†,‡ Robert Rallo,†,§ Gerassimos Orkoulas,‡ and Yoram Cohen*,†,‡ †
Center for the Environmental Implications of Nanotechnology, California NanoSystems Institute, University of California, Los Angeles, Los Angeles, California 90095, United States ‡ Chemical and Biomolecular Engineering Department, University of California, Los Angeles, Los Angeles, California 90095, United States § Departament d’Enginyeria Informatica i Matematiques, Universitat Rovira i Virgili, Av. Paisos Catalans 26, 43007 Tarragona, Catalunya, Spain
bS Supporting Information ABSTRACT: A constant-number direct simulation Monte Carlo (DSMC) model was developed for the analysis of nanoparticle (NP) agglomeration in aqueous suspensions. The modeling approach, based on the “particles in a box” simulation method, considered both particle agglomeration and gravitational settling. Particleparticle agglomeration probability was determined based on the classical DerjaguinLandauVerweyOverbeek (DLVO) theory and considerations of the collision frequency as impacted by Brownian motion. Model predictions were in reasonable agreement with respect to the particle size distribution and average agglomerate size when compared with dynamic light scattering (DLS) measurements for aqueous TiO2, CeO2, and C60 nanoparticle suspensions over a wide range of pH (310) and ionic strength (0.01156 mM). Simulations also demonstrated, in quantitative agreement with DLS measurements, that nanoparticle agglomerate size increased both with ionic strength and as the solution pH approached the isoelectric point (IEP). The present work suggests that the DSMC modeling approach, along with future use of an extended DLVO theory, has the potential for becoming a practical environmental analysis tool for predicting the agglomeration behavior of aqueous nanoparticle suspensions.
’ INTRODUCTION Nanosized materials are increasingly being utilized in various modern industrial products and processes, primarily due to their unique nanoscale properties.14 Engineered nanomaterials (eNMs) are estimated to be components of more than 1000 commercial products,5 and thus there is an increased public concern regarding the potential adverse impacts of exposure to eNMs that may take place in the workplace, during product use and disposal, and in the environment.4,611 The environmental transport and fate of eNMs3,12 as well as their behavior at the bionano interface3,10,13 are affected by their physicochemical properties10 with the particle size being a major factor.2,12 Fundamentally, mass diffusivity, sedimentation velocity, deposition velocity, and attachment efficiency of nanoparticles onto solid and biological surfaces are significantly influenced by their size.1417 As has been shown in numerous studies, nanoparticles in aqueous suspensions, which are the focus of the present study, generally do not exist as stable suspensions of primary nanoparticles.12,1824 This is particularly the case under environmental conditions where the ionic strength of natural water sources is sufficiently high and where adsorption of hydrophobic organics onto the nanoparticles25,26 can both promote rapid nanoparticle agglomeration.1820,23 In recent years, efforts to quantify the agglomeration state of eNMs have intensified.12 However, comprehensive experimental r 2011 American Chemical Society
mapping of eNMs agglomeration for the large number of present and anticipated emerging nanoparticles, over wide ranges of possible environmental water chemistries and nanoparticle properties, is a daunting and possibly impractical task. Thus, there has been a growing interest to explore various approaches18,2729 to better understand and generalize the agglomeration behavior of nanoparticles in aqueous suspensions. The majority of studies on the environmental or toxic impact of nanoparticles have focused on qualitative interpretation of observed agglomeration behavior of nanoparticles via the classical DLVO theory.18,19,21,22 Yet, there are factors (e.g., steric, geometric, hydrodynamic, hydration, magnetic) that can impact nanoparticle agglomeration that are not considered by the classical DLVO theory. Accordingly, extended versions of the DLVO theory have been proposed.3034 It has been generally accepted, however, that classical DLVO theory can provide a reasonable starting point for describing nanoparticle agglomeration in aquatic media under a wider range of environmental conditions12,15,18,28 and even for surface coated nanoparticles.28 Received: June 22, 2011 Accepted: September 14, 2011 Revised: August 29, 2011 Published: September 14, 2011 9284
dx.doi.org/10.1021/es202134p | Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology Quantitative simulation methods to describe agglomeration of nanoparticles can be implemented while accounting for particle particle interactions. 15,3538 Indeed, molecular dynamics (MD) and Brownian dynamics (BD) simulations have been used to study details of nanoparticle agglomeration kinetics and agglomerate morphology.3537 BD simulations of nanoparticle agglomeration in aqueous suspensions have demonstrated that the DLVO theory can provide a reasonable description of nanoparticle agglomeration.39 It has been shown that that BD type simulations can track the temporal change in nanoparticle suspension concentration due to sedimentation.40 Also recent comparison of the MD and BD methods37 revealed that nanoparticle clusters with higher fractal dimensions were predicted with the MD approach. While MD and BD methods provide detailed information regarding particle agglomeration/disagglomeration via tracking of individual particles, they do so at the expense of significant computational resources. Thus, these methods place a limit on the practical number of particles that can be effectively modeled.36,39 Another popular approach is the class of direct simulation Monte Carlo (DSMC) methods which treats the simulation domain as a statistical particle ensemble where each event has a probability of occurrence quantified via frequency functions calculated based on particleparticle interaction energies.36,41,42 The DSMC simulation approach36,38,4146 is a convenient method for describing particle agglomeration based on solving the population balance equations (PBEs) as described by the classical Smoluchowski coagulation equation47,48 or its extension to include disagglomeration43 nucleation and surface growth.36 DSMC methods can be conveniently applied with moderate computational resources, in conjunction with the DLVO approach (or its extension) to quantitatively describe nanoparticle agglomeration in aqueous suspensions. In addition, most laboratory experiments dealing with the fate and transport20,21,24,27,44 or toxicity13,49,50 of nanoparticles in aqueous suspensions are typically conducted over extended periods of time (hours to days). Environmental time scales of interest are also of similar or longer magnitude. Therefore, model simulations need to consider the evolution of the particle size distribution (in aqueous suspension) to its steady-state (or stable) condition given the combined effects of agglomeration and sedimentation. Experimental studies have documented the sedimentation of agglomerated nanoparticles, based on UVvis spectrometry20,21,51,52 and visual observations53,54 for measurement time scales on the order of hours20,21,52 to days.51,53,54 For example, studies with aqueous nanoparticle suspensions (∼10200 mg L1) have demonstrated significant concentration decrease (up to ∼90% in some cases) for TiO221,31 (21 nm) and iron-based52 (50 nm) nanoparticles over a period of up to ∼6 h. In the present study, nanoparticle agglomeration is investigated within the context of the DLVO theory based on a computational constant-number DSMC approach41,42 of “particles in a box” considering Brownian motion and agglomerate sedimentation. In this approach, evolution of the particle size distribution is tracked to its steady state condition with model validation based on present experimental and literature reported dynamic light scattering (DLS) measurements of agglomerates sizes for TiO2, CeO2, and C60 nanoparticles in aqueous suspensions. A series of parametric simulations were also carried out in order to illustrate the influence of various model parameters (e.g., primary nanoparticle diameter, pH) on nanoparticle agglomeration behavior and thus suggest the potential use of the DSMC modeling approach and its potential extension (e.g.,
ARTICLE
modified DLVO), as a useful tool for the study of nanoparticle agglomeration in aqueous suspensions.
’ MATERIALS AND METHODS Evaluation of Nanoparticle Agglomeration. The agglomerate size of nanoparticles in aqueous suspensions was evaluated via a constant-number DSMC approach42 making use of the classical DLVO theory48,55 and accounting for agglomerate sedimentation. The modeling approach focused on enabling comparison of predicted and measured (via DLS) nanoparticle agglomerate sizes (in aqueous suspensions) over a period (24 h) typical in high throughput nanoparticle toxicity screening studies.21,49 Experimental Section. The agglomeration state of TiO2 (Evonik Industries, Parsippany, NJ) and CeO2 (Meliorum Technologies, Rochester, NY) nanoparticles (21 and 15 nm primary diameter, respectively; Supporting Information (SI), Figure S9) was quantified in aqueous suspensions over a pH range of 310 and ionic strength (IS) of 1021 mM. Aqueous suspensions of the commercial nanoparticles were prepared using 18 MΩ cm ultrapure D.I. water with pH adjustment using HCl, NaOH and NaHCO3 and NaCl for ionic strength adjustment (ACS grade, Fisher Scientific, Waltham, MA). Nanoparticle stock suspensions were first prepared at a concentration of 1000 mg L1 by adding (over a ∼5 s period) 10 mg of nanoparticles to a stirred 10 mL volume of water (in a 40 mL vial) previously adjusted to pH of 3 or 10; stirring was accomplished using a Teflon coated 15.8 mm (L) x 8 mm (D) magnetic stir bar and stirrer (Magnestir S8290, Scientific Products, McGaw Park, IL). The resulting suspension was immediately sonicated for 30 min at 23 ( 0.5 C in a temperature controlled sonication bath (Bransonic 2510, Branson, Danbury, CT, 0.75 gallon, 130 W @ 40 Hz). The sonication bath temperature was maintained ((0.5 C) by circulating water through a copper coil tube (0.635 cm inside diameter), submerged in the sonication bath, from a constant temperature water circulator (NESLAB RTE-111, Thermo Scientific, Waltham, MA). Maintaining a constant temperature in the sonication bath was essential since, in the absence of temperature control, significant temperature rise was observed (up to ∼20 C over a 30 min period) as a consequence of the sonication process. It is noted that a sonication period longer than about 30 min did not improve nanoparticle dispersion. After sonication, 0.4 mL of the suspension was withdrawn and added to 19.6 mL of the same pH adjusted water in a 40 mL glass vial (resulting in 20 mg L1 suspension) with additional 5 min sonication also at 23 C. Nanoparticle suspensions at pH 8 were prepared by a similar procedure, but with 0.4 mL of a 1000 mg L1 pH 10 stock suspension added to a 19.6 mL water at pH 8 and 2.4 104 mM NaHCO3. Immediately after 5 min sonication of the 20 mg L1 (20 mL) suspension, 3 mL sample was transferred to a 4.5 mL cuvette for DLS analysis of the particle size distribution. Particle size distribution (PSD) measurements for the different nanoparticle suspensions were obtained via DLS measurements (ZetaSizer Nano S90, Malvern Instruments, Worcestershire, U.K.) over a 24 h period. This DLS instrument utilizes a horizontal entry laser beam of ∼40 μm in diameter, and detection angle of 90 with measurement reliability down to ∼0.25 mg L1.56 In the present study, the nanoparticle concentration decreased from its initial value of 20 mg L1 to 117 mg L1 (depending on the pH and ionic strength) and thus all DLS measurements were at concentrations above the minimum detection limit. The mass concentrations 9285
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology
ARTICLE
the added mass contributed by the above step of particle replenishment (SI, Figure S1). Particles can move out of the box due to gravitational sedimentation. A particle that has settled out of the box is replaced by a particle selected from the size distribution of the particles that have settled. For particles that have traveled out of the box (top or sides) by Brownian motion, periodic boundary conditions15 are invoked to replace the particles (i.e., the same removed particle is placed back into the box). Model Equations. The simulation is initialized by first distributing N nanoparticles in a simulation box of volume V [m3] satisfying the initial particle mass concentration, Co [mg L1]: F N 4 3 π r V ¼ 3 ð1Þ C i¼1 3 3 i
∑
Figure 1. Flow diagram of the Monte Carlo particles in a box simulation. The open and filled circles indicate the points in the algorithm at which the particle size distributions are determined for the particles in suspension and the settled particles, respectively.
of the nanoparticles in the stock solutions and in samples used for DLS measurements were determined via elemental analysis of the suspension, by inductively coupled plasma mass spectrometry (ICP-MS) following a previously published protocol.57 Finally, zeta potential for the nanoparticles was measured (Brookhaven ZetaPAL, Brookhaven Instruments, Holtsville, NY) for the above nanoparticle suspensions at ionic strength of 0.1 mM. Simulation Algorithm. A constant-number DSMC model of particles in a box42 was developed to describe nanoparticle agglomeration in suspensions while accounting for agglomerate sedimentation. The accuracy of the DSMC simulations varies with the number of √ particles, N, in the simulation domain (i.e., box) with the error 1/ N.45 Therefore, the number of particles was kept constant42 and the nanoparticle mass concentration was preserved through simulation box expansion.41 The simulation algorithm is described in Figure 1 with details of the working equations provided in Table S1 (Supporting Information (SI)). The simulation approach (Figure 1) consists of initially distributing N nanoparticles (NP) in a simulation box (SI, Figure S1) of dimensions that are consistent with the desired nanoparticle mass concentration. The agglomeration frequencies for each of the N2 pairs of particles are calculated following the approach of the Smoluchowski coagulation equation.55 Subsequently, a pair of particles is selected for agglomeration based on a probability density function constructed from the agglomeration frequencies (SI, Figure S2), and the size and position are then computed for the agglomerated particle pair. Only pairwise particleparticle interactions are considered48,55 and the agglomeration events assumed to be irreversible.18,25 The time step for each consecutive agglomeration event is determined based on the inverse of agglomeration frequency (eq 7).38,41,42 Each agglomeration event reduces the total number of particles (i.e., freely suspended primary particles and agglomerates) by one. Therefore, following each agglomeration event, a particle is added to the box in order to preserve the total number of particles in the box. The added particle is selected by sampling from the particle size distribution of the particles in the simulation box. In order to preserve the mass concentration of nanoparticles, the simulation box is expanded to accommodate
where F [g m3] is the nanoparticle density, C is the mass concentration [mg L1], and ri [m] is the radius of particle i. The initial nanoparticles in the simulation box (SI, Figure S1) are positioned at locations (xi,yi,zi) = (V)1/3 3 [Px,Py,Pz], where the random numbers Px, Py, and Pz are uniformly distributed in [0,1], while disallowing two particles to occupy the same space. ^ij, [no. of agglomeration events The agglomeration frequency N per m3 s] between two particles i and j is determined based on the Smoluchowski coagulation theory where47,55,58 ^ij ¼ kij 3 ni 3 nj N
ð2Þ
where ni and nj are the number concentrations of particles i and j, respectively. The agglomeration frequency function, kij, is expressed as: kij ¼
βij
ð3Þ
Wij
in which βij is the collision frequency function which arises due to Brownian motion42,48 given by ! 2kT 1 1 ðri þ rj Þ þ ð4Þ βij ¼ 3μ ri rj where ri and rj are the effective radii [m] (eq 6) of particles i and j, respectively, k is Boltzmann’s constant [k = 1.38 1023 J K1], T is temperature [K], and μ is the viscosity of the medium [Pa s]. The stability ratio, Wij [dimensionless], is the inverse of the fraction of successful collisions for particle pair i,j, given as48,55 no: of collisions no: of collisions that result in agglomeration ΦT, ij Z ∞ exp kT ¼ 23 ds 2 s 2
Wij ¼
ð5Þ
in which s = 2R/(ri+rj), where R is the distance between the center of the particles [m], and ΦT,ij is the total interaction energy [J] between particle pair i,j. There are various approaches to quantifying ΦT,ij;31,55 however, in the present work the classical DLVO theory48,55,58 (SI, Table S1) was adopted as a first order analysis, without a loss of generality of the simulation approach, given its success in describing surface interactions in colloidal systems.48,55,58 Recent studies have also suggested that the DLVO approach is a reasonable basis for quantifying the agglomeration of nanoparticles.12,15,18 According to the classical DLVO theory, the total interaction energy for a particle pair is the 9286
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology
ARTICLE
sum of the van der Waals attraction energy (as a function of the Hamaker constant AH [J], particle sizes ri and rj [m], and separation distance between the particle pair R [m])48,58 and the electrical double layer repulsion energy (being a function of the surface zeta potential of nanoparticles in solution ζ [V], solution ionic strength IS [M], ri, rj, and R) as summarized in Table S1 (SI). Other model parameters include the primary nanoparticle diameter dp [m], initial nanoparticle concentration Co [mg L1], agglomerate fractal dimension df [dimensionless],47 and primary particle density F [g m3]. Once the agglomeration frequency function is determined (eq 3) for all particle pairs, a pair of nanoparticles (i,j) is selected to agglomerate following the sampling approach proposed by Kruis et al.,45 where the probability for the selected pair is taken to be proportional to the relative magnitude of its agglomeration frequency function (SI, Figure S2). In this approach, the kij values are added sequentially for the list of all particle pairs until the kij sum exceeds the value of a sampled random number R in N [0, ΣN i = 1Σj = 1kij]. When the above criterion is satisfied, the particle pair (i,j) corresponding to the last added kij is selected for agglomeration. The mass of this newly formed agglomerate, mk, is the linear sum of the agglomerated particle pair, that is, mk = mi+ mj. The effective diameter of this agglomerate, de, is calculated as:59 f de ¼ dp 3 n1=d p
ð6Þ
where np is the number of primary particles in the agglomerate, and df is the fractal dimension for the specific agglomeration regime.47 In order to maintain the total particle count in the simulation box, the lost particle (due to agglomeration) is replenished by sampling from the existing particle size distribution. The simulation box is then expanded to a new volume (eq 1) in order to maintain the mass concentration. The box is expanded horizontally (in the x,y plane; SI, Figure S1) in order to maintain consistency with experimental DLS measurements in which the thickness of the laser beam is fixed. Following box expansion (Figure 1), the time step [s] to the next agglomeration event is estimated based on the inverse of the average agglomeration frequency function
over all N2 particle pairs:42 Δt ¼
2 C 3 N 3 Ækij æ
ð7Þ
Diffusion and settling distances due to Brownian diffusion and Stokes’ sedimentation, respectively, are subsequently determined for the above time step (eq 7) for all particles in the box. The vertical distance traveled by a particle i is estimated as Δzi ¼ zS, i þ zB, i
ð8Þ
where Δz < 0 represents net downward movement. The gravitational settling distance for particle i is given as zS,i = vS,i 3 Δt, where vS,i is its Stokes’ settling velocity given by14 2ðFp Ff Þ 3 ð1 ϕi Þ 3 g 3 ri2 ð9Þ vS, i ¼ 93μ in which Fp is the particle density (g L1), Ff is the fluid density, g is the gravitational constant [m3 kg1 s2] and ϕ is the particle agglomerate porosity (i.e., ϕ = 1 (rm/re)3, where rm is the mass equivalent radius of the particle (rm = ((3mi)/(4Fπ))1/3, [m]), and re is its effective radius (eq 6). The Brownian diffusion distance, zB,i (positive or negative for upward or downward movement, respectively) is estimated assuming that the distance
traveled vertically is determined based on a random number sampled from a normal distribution with μ = 0 and s = (2 3 Di 3 Δt)1/2 47,55 such that zB,i ∼ N(0,2 3 Di 3 Δt), in which Di is the Brownian diffusivity [m2 s1] determined from47,55 Di = (kT)/(6πμri). Following the above approach, particles for which zi+Δzi e 0 (where zi is the particle vertical position at time ti) are considered to have settled out of the box, thereby decreasing the number of particles in the box. This necessitates the introduction of new particles into the box in order to preserve the total particle count. The size of each replenishing particle is determined by sampling from the size distribution of the last N particles that have settled, given the expectation that particles can enter the box via its top face via sedimentation. The vertical positioning of each replenishing particle in the simulation box is assigned as zi = Lz+Pz 3 Δzi, where Lz is the box height, and Pz is a uniformly distributed random number in [0,1]. Δzi is the vertical distance that would be traveled due to the combined effect of sedimentation and Brownian motion, calculated as described above, for the same agglomeration time step (eq 7). The above sampling is repeated until the condition, 0 < zi e Lz, for placing a particle inside the box is met and the mass concentration in the box is recalculated. Finally, it is noted that particles can diffuse out of the vertical sides of the simulation box due to Brownian motion. However, in this case, one can invoke a periodic boundary condition specifying that a particle that diffuses out of a given vertical face would be reintroduced from the opposite face. A similar reasoning is applied to particles that diffuse out of the box through the top (horizontal) face. Simulations. Simulations were performed for TiO2 and CeO2 nanoparticles of 21 and 15 nm primary particle diameters, respectively, at initial concentration of 20 mg L1, matching the 24 h period of DLS measurements (SI, Table S2). Simulation conditions also matched the pH range of 310, and ionic strength of 0.020.4 mM at 23 C for the present DLS measurements. For comparison with reported literature data for TiO2 nanoparticle suspensions, additional simulations were also carried out for an initial concentration range of 40 50 mg L1, pH of 310.4, and IS of 0.01 12.5 mM.21,22,24 Simulations were also carried for aqueous C60 nanoparticle suspensions (primary size of 80 and 168 nm), pH 5.5, 7 and ionic strength of 10156 mM. The basic fundamental model parameters included: particle primary size, zeta potential, Hamaker constant, solution ionic strength and temperature. The first three parameters were obtained from either independent measurements or from literature reported data (See SI, Table S2), while the latter two were the conditions as specified in the particular experiments. Therefore, all model simulations that are compared with experimental data are a priori predictions. Simulations were carried out with a number of particles in the box ranging up to 10 000. Different numbers of repeated simulations (for the same particle number) were carried out in order to evaluate the optimally reasonable number of particles needed to achieve convergence of the simulation results. Simulations were carried out on a cluster of 20 Intel Xeon Quad-Core processors at 2.23.0 GHz with 176 GB of total RAM. Simulation CPU times on this cluster ranged from as short as 100 s to as along as 15 days for 500 and 10 000 simulated nanoparticles, respectively.
’ RESULTS AND DISCUSSION Convergence of Simulations. Simulations of nanoparticle agglomeration were first carried out to determine the number of particles necessary to achieve a convergent solution. A series of 9287
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology simulations with 50010 000 nanoparticles in the simulation box indicated, consistent with the existing DSMC literature, that the use of 5000 particles (e.g., Figure 2 and SI, Figure S3) was sufficient for reaching accurate simulation results.36,38 However, given that the simulations were seeded with random numbers (Figure 1; SI, Figure S2), there were statistically measurable variations in the results for repeated simulations (e.g., SI, Figure S4). It is found that above five replicate simulations and with 5000 (or more) particles in the simulation box, the change in the predicted average agglomerate size was less than 0.1% (e.g., SI, Figure S4b) and the standard deviation of the predicted average agglomerate size over 10 replicate simulations was typically below 1% (Figure 2; SI, Figures S3). Accordingly, all results in the present work are presented as the average over 10 simulation replicates for 5000 particles in the simulation box. Nanoparticle Agglomeration. As nanoparticles agglomerate in aqueous suspensions, gravitational sedimentation can take place thereby altering the size distribution of the particles remaining in suspension. For example, nanoparticle agglomerates of size 100 and 250 nm can sediment a distance equivalent to a typical width of a DLS laser beam (∼0.04 mm) in a period of 7 and 1 h, respectively (SI, Figure S5). Thus, if nanoparticle agglomerates reach the above size range or greater, settling due to gravity has to be considered, especially for applications and
Figure 2. Mean agglomerate diameter based on the average of 10 simulations as a function of the number of particles in the simulation box. The vertical bars represent one standard deviation of the mean particle size over 10 simulation replicates. Simulation conditions: ζ = 40 mV, IS = 0.1 mM, dp = 21 nm, AH = 42 zJ.
ARTICLE
studies of extended duration. Of particular interest in the present study are predictions of agglomerate sizes as determined by DLS measurements in the presence of particle settling. The occurrence of sedimentation can be inferred from a decrease in the photon count rate in DLS measurements. This is illustrated, for example, in Figure S6 (SI) revealing ∼18% and ∼30% photon count rate decrease in DLS measurements over a 24 h period for TiO2 (pH 8 and 10, IS = 0.37 mM) and CeO2 (pH 8, IS = 0.37 mM) nanoparticles, respectively. The simulations revealed a rapid rate of agglomeration with an agglomerate size that increases with time when sedimentation is not considered in the model. However, upon inclusion of sedimentation a “stable” (i.e., time independent) agglomerate size is reached (in suspension) after several hours. This behavior is illustrated in Figure 3 for CeO2 and TiO2 nanoparticles of 15 and 21 nm primary particle diameters, respectively, at pH 8 at low ionic strength of 0.065 mM. As a result of sedimentation, the mass concentration of the TiO2 and CeO2 nanoparticles in suspension decreased by 1050% for the present range of simulation conditions (SI, Table S2) as was verified via ICP measurements; however the average particle size in suspension remained essentially at steady state (after ∼15 h, Figure 3). Such behavior is reached once the rate of sedimentation is balanced by the rate of nanoparticle agglomeration. The predicted average agglomerate size is in remarkable agreement with the present DLS measurements (Figure 3) with a prediction absolute error of 1% and 0.4% for TiO2 and CeO2 nanoparticles for the solution conditions indicated in Figure 3. However, it is noted that sonication of the suspensions, prepared from commercial nanoparticle powders, could only breakup the nanoparticle agglomerates to their “stable” suspension particle size. Therefore, the region of agglomerate evolution could not be traced via DLS and only the stable nanoparticle agglomerate size could be monitored (Figure 3). Comparison of MC model predictions of TiO2 nanoparticle agglomerate size with the present series of DLS measurements and published literature data21,22,24,60,61 along with comparisons of predictions with literature data for C60 nanoparticles are provided in Figure 4 and Table S2 (SI). Only literature data for agglomerate sizes below 1 μm were considered as this size approaches the limitations of DLS measurements56 as well as the DLVO theory.55 Reasonable agreement of predictions with DLS data (present and literature data) were obtained with an absolute error in the range of 0.625.3% and ∼10.8% average absolute
Figure 3. Evolution of CeO2 (left) and TiO2 (right) agglomerate diameter based on simulation with and without sedimentation, based on average of 10 simulation instances with 5000 particles in the simulation box. Simulation conditions: pH 8, ζCeO2 = 24.5 mV, ζTio2 = 29 mV, IS = 0.065 mM, AH,CeO2 = 21 zJ, AH,TiO2 = 42 zJ. 9288
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology error for the 26 measurements depicted in Figure 4 (also SI, Table S2). Notwithstanding the success of the constant-number DSMC model predictions (Figures 3 and 4; SI, Table S1) of nanoparticle agglomeration, broader assessment of the suitability of the present simulation approach for predicting short-time agglomeration kinetics would necessitate DLS measurements with aqueous suspensions in which the nanoparticles are initially near their primary diameter. Dependence of Nanoparticle Agglomeration on Model Parameters (dp, ζ, IS, AH). According to the DLVO theory48,55 nanoparticle agglomeration depends primarily on the nanoparticle primary diameter (dp), Hamaker constant (AH), ionic strength (IS), and surface electrical potential (estimated from the zeta potential, ζ). The zeta potential varies with solution pH and there is a pH at which the isoelectric point (IEP) is reached where the net surface charge is zero (i.e., ζ = 0 mV). Particle agglomeration is expected to be most significant near the IEP as the electrical double layer diminishes and the attractive energy due to van der Waals forces becomes dominant. As the solution pH deviates away from the IEP, the electrical double layer thickness increases as does the surface charge (either positive or negative; Figure 5a) and thus greater particle repulsion and smaller agglomerates are expected. The above behavior is
Figure 4. Comparison between simulation and experimental results for average agglomerate diameter for TiO2,21,22,24 CeO2, and C6060,61 nanoparticles. For TiO2, dp,TiO2 = 5 (Δ), 15 (O), and 21 (), 0) nm, for CeO2, dp,CeO2 = 15 (9) nm, and for C60, dp,C60 = 80 (1) nm, 168 (b) nm. IS = 0.01156 mM, pH 310.4, ζ = 4542 mV.
ARTICLE
demonstrated in Figure 5a for simulations of TiO2 nanoparticle agglomerates revealing agreement with measured DLS data within an absolute average error of 2.8%. The predictions are consistent with experimental observations of the maximum agglomerate particle size (400 nm) at the IEP and decreasing size away from the IEP (by about a factor of 3 for the measurements at the lowest and highest pH levels). As shown in Figure 5b, nanoparticle agglomeration can also increase with increased ionic strength due to suppression of the electrical double layer. The typical reported dependence of particle agglomeration on ionic strength22 is illustrated in the simulation results shown in Figure 5b for TiO2 nanoparticles (dp = 21 nm) at ζ = 20 mV and 40 mV. In this example, the critical coagulation concentrations at the above two conditions are estimated at ∼0.08 mM and ∼2.89 mM at which the agglomerate size is ∼805 nm and ∼563 nm, respectively. Although the predicted rise in agglomerate size at high ionic strength portrays the expected trend, the deviation of the calculated average agglomerate size from the mean is noteworthy. Considerably more than 10 simulation instances with N > 5000 would have been necessary (at the expense of considerable CPU time of ∼4 weeks) to achieve an accurate stationary solution, since the size fluctuation is large at high IS due to a significant degree of settling. According to the DLVO theory, agglomerate sizes would also increase with increasing magnitude of the Hamaker constant, as this would imply increasing van der Waals attraction energy (SI, Table S1, eq S1). For the above TiO2 example, the simulations reveal a linear dependence of agglomerate nanoparticle size on the Hamaker constant (AH) demonstrating an agglomerate size increase from 110 nm to 550 nm with AH increase from 10 zJ to 90 zJ (SI, Figure S7), corresponding to the range of literature reported AH values for the anatase62,63 and rutile63,64 forms of TiO2. The simulations revealed that the agglomerate size of nanoparticles in aqueous suspension increases with decreasing primary particle diameter (Figure 6). This result should not be surprising since van der Waals interactions increase with decreasing particle size (SI, eq S1), the collision frequency is more pronounced with smaller particles (eq 4) and electrostatic repulsion increases with increased particle size (SI, eqs S2 and S3). As a result, smaller primary nanoparticles will form larger agglomerates. It is noted, however, that DLS measurements are only indicative of the size of particles remaining in suspensions and do not provide a measure of the true distribution of all
Figure 5. Average agglomerate diameter of nanoparticle aggregates after 24 h as a function of (left) pH levels (at IS = 0.037 mM) and (right) ionic strength. Simulation conditions: AH = 42 zJ, dp = 21 nm, Co = 20 mg L1, temperature = 23 C. Note: the vertical bars represent one standard deviation over 10 simulation replicates. 9289
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology
ARTICLE
’ AUTHOR INFORMATION Corresponding Author
*E-mail: [email protected].
’ ACKNOWLEDGMENT This work was supported, in part, by the National Science Foundation and the Environmental Protection Agency under Cooperative Agreement Number DBI 0830117, the UCLA Water Technology Research Center and the California Department of Water Resources. Any opinions, findings, conclusions or recommendations expressed herein are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or the Environmental Protection Agency. This work has not been subjected to an EPA peer and policy review. We also thank Zhaoxia Ji for assistance in obtaining the TEM images. Figure 6. Dependence of average agglomerate diameter on the primary nanoparticle diameter. Simulation condition: IS = 0.5 mM, Co = 20 mg L1, T = 23 C.
agglomerates that may have formed. Accordingly one would expect that, as a result of agglomeration and sedimentation, the nanoparticle size distribution in suspension (as determined by DLS) will reveal an increasing tail of smaller size agglomerates with increasing primary particle diameter (Figure 6; SI, Figure S8). The above behavior should not be taken as a universal representation of nanoparticle agglomeration as one must be cautious with the limitation of the DLVO theory to nanoparticles with kr , 1 (e.g., corresponding to ri , 10 nm at IS = 1 mM; note that k is the inverse Debye length, SI, Table S1). Moreover, it is also noted that the present application of the DLVO theory, as well as the simple application of gravitational settling, does not consider the impact of the details of agglomerate geometry and morphology. Nonetheless, the present analysis suggests that interpretation of nanoparticle behavior in environmental aquatic media and potential toxic outcomes due to exposure to nanoparticles must carefully consider not only the particle size distribution but also the experimental protocols used to determine such size distributions. In summary, the present constant-number DSMC approach of simulating nanoparticle agglomeration in aqueous suspensions demonstrated that classical DLVO theory can provide reasonably accurate predictions of the average nanoparticle agglomerate size as well as the particle size distribution over a wide range of solution pH (310) and ionic strength (0.01156 mM; SI, Table S2). Extension of the present approach using the extended DLVO theory is presently underway in order to explore a wide range of nanoparticle types aqueous solution chemistries of environmental interest.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information is included in the supporting information regarding figures of the box expansion approach, particle pair selection method, added figures of experimental and simulation results, calculation of sedimentation distances, TEM images of nanoparticles, list of DLVO working equations, and table of experimental and simulation conditions. This material is available free of charge via the Internet at http://pubs.acs.org/.
’ REFERENCES (1) Guo, Z.; Tan, L. Fundamentals and Applications of Nanomaterials, 1st ed.; Artech House Publishers: Norwood, MA, 2009. (2) Klaine, S. J.; Alvarez, P. J. J.; Batley, G. E.; Fernandes, T. F.; Handy, R. D.; Lyon, D. Y.; Mahendra, S.; McLaughlin, M. J.; Lead, J. R. Nanomaterials in the environment: Behavior, fate, bioavailability, and effects. Environ. Toxicol. Chem. 2008, 27 (9), 1825–1851. (3) Handy, R. D.; Owen, R.; Valsami-Jones, E. The ecotoxicology of nanoparticles and nanomaterials: Current status, knowledge gaps, challenges, and future needs. Ecotoxicology 2008, 17 (5), 315–325. (4) Wiesner, M. R.; Lowry, G. V.; Alvarez, P.; Dionysiou, D.; Biswas, P. Assessing the risks of manufactured nanomaterials. Environ. Sci. Technol. 2006, 40 (14), 4336–4345. (5) The Project on Emerging Nanotechnologies: Consumer Products Inventory (Woodrow Wilson International Center). http://www. nanotechproject.org/inventories/consumer/ (accessed 8/29/2011). (6) Farre, M.; Gajda-Schrantz, K.; Kantiani, L.; Barcelo, D. Ecotoxicity and analysis of nanomaterials in the aquatic environment. Anal. Bioanal. Chem. 2009, 393 (1), 81–95. (7) Barnard, A. S. Computational strategies for predicting the potential risks associated with nanotechnology. Nanoscale 2009, 1 (1), 89–95. (8) Stone, V.; Nowack, B.; Baun, A.; van den Brink, N.; von der Kammer, F.; Dusinska, M.; Handy, R.; Hankin, S.; Hassell€ov, M.; Joner, E.; Fernandes, T. F. Nanomaterials for environmental studies: Classification, reference material issues, and strategies for physico-chemical characterisation. Sci. Total Environ. 2010, 408 (7), 1745–1754. (9) Kahru, A.; Dubourguier, H.-C. From ecotoxicology to nanoecotoxicology. Toxicology 2010, 269 (23), 105–119. (10) Biswas, P.; Wu, C. Y. Critical Review: Nanoparticles and the environment. J. Air Waste Manage. Assoc. 2005, 55 (6), 708–746. (11) Colvin, V. L. The potential environmental impact of engineered nanomaterials. Nat. Biotechnol. 2003, 21 (10), 1166–1170. (12) Petosa, A. R.; Jaisi, D. P.; Quevedo, I. R.; Elimelech, M.; Tufenkji, N. Aggregation and deposition of engineered nanomaterials in aquatic environments: Role of physicochemical interactions. Environ. Sci. Technol. 2010, 44 (17), 6532–6549. (13) Long, T. C.; Saleh, N.; Tilton, R. D.; Lowry, G. V.; Veronesi, B. Titanium dioxide (P25) produces reactive oxygen species in immortalized brain microglia (BV2): Implications for nanoparticle neurotoxicity. Environ. Sci. Technol. 2006, 40 (14), 4346–4352. (14) Kajihara, M. Settling velocity and porosity of large suspended particle. J. Oceanogr. 1971, 27 (4), 158–162. (15) Elimelech, M. Particle Deposition and Aggregation: Measurement, Modelling, And Simulation. Butterworth-Heinemann: Boston, 1995; p xv. (16) Areepitak, T.; Ren, J. Model simulations of particle aggregation effect on colloid exchange between streams and streambeds. Environ. Sci. Technol. 2011null-null. 9290
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology (17) Sharma, V. K. Aggregation and toxicity of titanium dioxide nanoparticles in aquatic environment—A review. J. Environ. Sci. Health, Part A: Environ. Sci. Eng. 2009, 44 (14), 1485–1495. (18) Zhang, Y.; Chen, Y.; Westerhoff, P.; Hristovski, K.; Crittenden, J. C. Stability of commercial metal oxide nanoparticles in water. Water Res. 2008, 42 (89), 2204–2212. (19) Guzman, K. A. D.; Finnegan, M. P.; Banfield, J. F. Influence of sufface potential on aggregation and transport of titania nanoparticles. Environ. Sci. Technol. 2006, 40 (24), 7688–7693. (20) Keller, A. A.; Wang, H. T.; Zhou, D. X.; Lenihan, H. S.; Cherr, G.; Cardinale, B. J.; Miller, R.; Ji, Z. X. Stability and aggregation of metal oxide nanoparticles in natural aqueous matrices. Environ. Sci. Technol. 2010, 44 (6), 1962–1967. (21) Ji, Z.; Jin, X.; George, S.; Xia, T.; Meng, H.; Wang, X.; Suarez, E.; Zhang, H.; Hoek, E. M. V.; Godwin, H.; Nel, A. E.; Zink, J. I. Dispersion and stability optimization of TiO2 nanoparticles in cell culture media. Environ. Sci. Technol. 2010, 44 (19), 7309–7314. (22) Jiang, J. K.; Oberdorster, G.; Biswas, P. Characterization of size, surface charge, and agglomeration state of nanoparticle dispersions for toxicological studies. J. Nanopart. Res. 2009, 11 (1), 77–89. (23) Baveye, P.; Laba, M. Aggregation and toxicology of titanium dioxide nanoparticles. Environ. Health Perspect. 2008, 116 (4), A152–A152. (24) French, R. A.; Jacobson, A. R.; Kim, B.; Isley, S. L.; Penn, R. L.; Baveye, P. C. Influence of Ionic strength, pH, and cation valence on aggregation kinetics of titanium dioxide nanoparticles. Environ. Sci. Technol. 2009, 43 (5), 1354–1359. (25) Baalousha, M. Aggregation and disaggregation of iron oxide nanoparticles: Influence of particle concentration, pH and natural organic matter. Sci. Total Environ. 2009, 407 (6), 2093–2101. (26) Thio, B. J. R.; Zhou, D. X.; Keller, A. A. Influence of natural organic matter on the aggregation and deposition of titanium dioxide nanoparticles. J. Hazard. Mater. 2011, 189 (12), 556–563. (27) He, Y. T.; Wan, J. M.; Tokunaga, T. Kinetic stability of hematite nanoparticles: The effect of particle sizes. J. Nanopart. Res. 2008, 10 (2), 321–332. (28) Huynh, K. A.; Chen, K. L. Aggregation kinetics of citrate and polyvinylpyrrolidone coated silver nanoparticles in monovalent and divalent electrolyte solutions. Environ. Sci. Technol. 2011, 45 (13), 5564–5571. (29) Buettner, K. M.; Rinciog, C. I.; Mylon, S. E. Aggregation kinetics of cerium oxide nanoparticles in monovalent and divalent electrolytes. Colloids Surf., A 2010, 366 (13), 74–79. (30) Bos, R.; van der Mei, H. C.; Busscher, H. J. Physico-chemistry of initial microbial adhesive interactions its mechanisms and methods for study. FEMS Microbiol. Rev. 1999, 23 (2), 179–230. (31) Grasso, D.; Subramaniam, K.; Butkus, M.; Strevett, K.; Bergendahl, J. A review of non-DLVO interactions in environmental colloidal systems. Rev. Environ. Sci. Bio/Technol. 2002, 1 (1), 17–38. (32) Hermansson, M. The DLVO theory in microbial adhesion. Colloids Surf., B 1999, 14 (14), 105–119. (33) Hong, Y. S.; Honda, R. J.; Myung, N. V.; Walker, S. L. Transport of iron-based nanoparticles: Role of magnetic properties. Environ. Sci. Technol. 2009, 43 (23), 8834–8839. (34) Honig, E. P.; Roeberse., Gj; Wiersema, P. H. Effect of hydrodynamic interaction on coagulation rate of hydrophobic colloids. J. Colloid Interface Sci. 1971, 36 (1), 97–109. (35) Alimohammadi, M.; Fichthorn, K. A. Molecular dynamics simulation of the aggregation of titanium dioxide nanocrystals: Preferential alignment. Nano Lett. 2009, 9 (12), 4198–4203. (36) Maisels, A.; Kruis, F. E.; Fissan, H. Direct simulation Monte Carlo for simultaneous nucleation, coagulation, and surface growth in dispersed systems. Chem. Eng. Sci. 2004, 59 (11), 2231–2239. (37) Markutsya, S.; Subramaniam, S.; Vigil, R. D.; Fox, R. O. On Brownian dynamics simulation of nanoparticle aggregation. Ind. Eng. Chem. Res. 2008, 47 (10), 3338–3345. (38) Zhao, H.; Maisels, A.; Matsoukas, T.; Zheng, C. Analysis of four Monte Carlo methods for the solution of population balances in dispersed systems. Powder Technology 2007, 173 (1), 38–50.
ARTICLE
(39) Peng, Z. B.; Doroodchi, E.; Evans, G. DEM simulation of aggregation of suspended nanoparticles. Powder Technol. 2010, 204 (1), 91–102. (40) Jiang, W. T.; Ding, G. L.; Peng, H.; Hu, H. T. Modeling of nanoparticles’ aggregation and sedimentation in nanofluid. Curr. Appl. Phys. 2010, 10 (3), 934–941. (41) Lin, Y. L.; Lee, K.; Matsoukas, T. Solution of the population balance equation using constant-number Monte Carlo. Chem. Eng. Sci. 2002, 57 (12), 2241–2252. (42) Smith, M.; Matsoukas, T. Constant-number Monte Carlo simulation of population balances. Chem. Eng. Sci. 1998, 53 (9), 1777–1786. (43) Lee, K.; Matsoukas, T. Simultaneous coagulation and break-up using constant-N Monte Carlo. Powder Technol. 2000, 110 (12), 82–89. (44) Kim, T.; Lee, C. H.; Joo, S. W.; Lee, K. Kinetics of gold nanoparticle aggregation: Experiments and modeling. J. Colloid Interface Sci. 2008, 318 (2), 238–243. (45) Kruis, F. E.; Maisels, A.; Fissan, H. Direct simulation Monte Carlo method for particle coagulation and aggregation. AIChE J. 2000, 46 (9), 1735–1742. (46) Zhao, H. B.; Zheng, C. G.; Xu, M. H. Multi-Monte Carlo method for particle coagulation: Description and validation. Appl. Math. Comput. 2005, 167 (2), 1383–1399. (47) Friedlander, S. K. Smoke, Dust, And Haze: Fundamentals of Aerosol Dynamics, 2nd ed.; Oxford University Press: New York, 2000; p xx. (48) Kruyt, H. R. Colloid Science; Elsevier: New York, 1949. (49) Liu, R.; Rallo, R.; George, S.; Ji, Z. X.; Nair, S.; Nel, A. E.; Cohen, Y. Classification nanoSAR development for cytotoxicity of metal oxide nanoparticles. Small 2011, 7 (8), 1118–1126. (50) Rallo, R.; France, B.; Liu, R.; Nair, S.; George, S.; Damoiseaux, R.; Giralt, F.; Nel, A.; Bradley, K.; Cohen, Y. Self-organizing map analysis of toxicity-related cell signaling pathways for metal and metal oxide nanoparticles. Environ. Sci. Technol. 2011, 45 (4), 1695–1702. (51) Allouni, Z. E.; Cimpan, M. R.; Høl, P. J.; Skodvin, T.; Gjerdet, N. R. Agglomeration and sedimentation of TiO2 nanoparticles in cell culture medium. Colloids Surf., B 2009, 68 (1), 83–87. (52) Tiraferri, A.; Chen, K. L.; Sethi, R.; Elimelech, M. Reduced aggregation and sedimentation of zero-valent iron nanoparticles in the presence of guar gum. J. Colloid Interface Sci. 2008, 324 (12), 71–79. (53) Fedele, L.; Colla, L.; Bobbo, S.; Barison, S.; Agresti, F. Experimental stability analysis of different water-based nanofluids. Nanoscale Res. Lett. 2011, 6 (1), 300. (54) Li, X. F.; Zhu, D. S.; Wang, X. J. Evaluation on dispersion behavior of the aqueous copper nano-suspensions. J. Colloid Interface Sci. 2007, 310 (2), 456–463. (55) Hunter, R. J. Foundations of Colloid Science, 2nd ed.; Oxford University Press: New York, 2001; p xii. (56) Filella, M.; Zhang, J. W.; Newman, M. E.; Buffle, J. Analytical applications of photon correlation spectroscopy for size distribution measurements of natural colloidal suspensions: Capabilities and limitations. Colloids Surf., A 1997, 120 (13), 27–46. (57) Chen, K. J.; Wolahan, S. M.; Wang, H.; Hsu, C. H.; Chang, H. W.; Durazo, A.; Hwang, L. P.; Garcia, M. A.; Jiang, Z. K.; Wu, L.; Lin, Y. Y.; Tseng, H. R. A small MRI contrast agent library of gadolinium(III)-encapsulated supramolecular nanoparticles for improved relaxivity and sensitivity. Biomaterials 2011, 32 (8), 2160–2165. (58) Vold, R. D.; Vold, M. J. Colloid and Interface Chemistry; Addison-Wesley: Reading, MA, 1983; p xxv. (59) Schwarzer, H. C.; Peukert, W. Prediction of aggregation kinetics based on surface properties of nanoparticles. Chem. Eng. Sci. 2005, 60 (1), 11–25. (60) Brant, J.; Lecoanet, H.; Wiesner, M. R. Aggregation and deposition characteristics of fullerene nanoparticles in aqueous systems. J. Nanopart. Res. 2005, 7 (45), 545–553. (61) Chen, K. L.; Elimelech, M. Relating colloidal stability of fullerene (C60) nanoparticles to nanoparticle charge and electrokinetic properties. Environ. Sci. Technol. 2009, 43 (19), 7270–7276. (62) Gomez-Merino, A. L.; Rubio-Hernandez, F. J.; VelazquezNavarro, J. F.; Galindo-Rosales, F. J.; Fortes-Quesada, P. The Hamaker 9291
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
Environmental Science & Technology
ARTICLE
constant of anatase aqueous suspensions. J. Colloid Interface Sci. 2007, 316 (2), 451–456. (63) Visser, J. On Hamaker constants: A comparison between Hamaker constants and Lifshitz-van der Waals constants. Adv. Colloid Interface Sci. 1972, 3 (4), 331–363. (64) Ackler, H. D.; French, R. H.; Chiang, Y. M. Comparisons of Hamaker constants for ceramic systems with intervening vacuum or water: From force laws and physical properties. J. Colloid Interface Sci. 1996, 179 (2), 460–469.
9292
dx.doi.org/10.1021/es202134p |Environ. Sci. Technol. 2011, 45, 9284–9292
ARTICLE pubs.acs.org/est
Impact Assessment of Ammonia Emissions on Inorganic Aerosols in East China Using Response Surface Modeling Technique Shuxiao Wang,*,† Jia Xing,† Carey Jang,‡ Yun Zhu,§ Joshua S. Fu,|| and Jiming Hao† †
)
School of Environment, and State Key Joint Laboratory of Environment Simulation and Pollution Control, Tsinghua University, Beijing 100084, P. R. China ‡ U.S. Environmental Protection Agency, Research Triangle Park, North Carolina 27711, United States § School of Environmental Science and Engineering, South China University of Technology, Guangzhou 510006, P. R. China Department of Civil and Environmental Engineering, University of Tennessee, Knoxville, Tennessee 37996, United States
bS Supporting Information ABSTRACT: Ammonia (NH3) is one important precursor of inorganic fine particles; however, knowledge of the impacts of NH3 emissions on aerosol formation in China is very limited. In this study, we have developed China’s NH3 emission inventory for 2005 and applied the Response Surface Modeling (RSM) technique upon a widely used regional air quality model, the Community Multi-Scale Air Quality Model (CMAQ). The purpose was to analyze the impacts of NH3 emissions on fine particles for January, April, July, and October over east China, especially those most developed regions including the North China Plain (NCP), Yangtze River delta (YRD), and the Pearl River delta (PRD). The results indicate that NH3 emissions contribute to 811% of PM2.5 concentrations in these three regions, comparable with the contributions of SO2 (911%) and NOx (511%) emissions. However, NH3, SO2, and NOx emissions present significant nonlinear impacts; the PM2.5 responses to their emissions increase when more control efforts are taken mainly because of the transition between NH3-rich and NH3-poor conditions. Nitrate aerosol (NO3) concentration is more sensitive to NOx emissions in NCP and YRD because of the abundant NH3 emissions in the two regions, but it is equally or even more sensitive to NH3 emissions in the PRD. In high NO3 pollution areas such as NCP and YRD, NH3 is sufficiently abundant to neutralize extra nitric acid produced by an additional 25% of NOx emissions. The 90% increase of NH3 emissions during 19902005 resulted in about 5060% increases of NO3 and SO42‑ aerosol concentrations. If no control measures are taken for NH3 emissions, NO3 will be further enhanced in the future. Control of NH3 emissions in winter, spring, and fall will benefit PM2.5 reduction for most regions. However, to improve regional air quality and avoid exacerbating the acidity of aerosols, a more effective pathway is to adopt a multipollutant strategy to control NH3 emissions in parallel with current SO2 and NOx controls in China.
’ INTRODUCTION The importance of ammonia (NH3) in contributing to secondary inorganic aerosols (SIA, i.e., sulfate (SO42), nitrate (NO3), and ammonium (NH4+)) has been well documented in recent studies. Excess NH3 provides a weak base, which allows a larger aqueous uptake of sulfur dioxide (SO2) to be oxidized and, at the same time, also affects the effective cloud SO2 oxidation rate due to strong pH-dependent oxidation by ozone (O3).1,2 In the presence of NH3, NO3 is formed by the gas-to-particle conversion process from nitric acid (HNO3) which was first produced through a photochemical reaction as nitrogen dioxide (NO2) and hydroxyl radical (•OH). Multisensitivity studies for European countries and the United States29 have been conducted using air quality models (AQMs) to explore the response of inorganic fine particles to emission changes of SO2, nitrogen oxides (NOx = NO + NO2), NH3, or nonmethane volatile r 2011 American Chemical Society
organic compounds (NMVOC). Derwent et al.9 used a moving air parcel trajectory model to estimate the mass concentrations of PM components for a rural location in the southern UK, and found that PM mass concentrations are nonlinear with PM precursor emissions, and suggested that abatement of NH3 emissions should be considered to obtain the largest PM2.5 reduction. Tsimpidi et al.2 applied a three-dimensional chemical transport model (PMCAMx) to investigate the changes in PM2.5 concentrations responding to changes of SO2 and NH3 emissions in the eastern United States, and indicated that coupled reductions of SO2 and NH3 emissions are more effective than the Received: June 30, 2011 Accepted: September 22, 2011 Revised: September 8, 2011 Published: September 22, 2011 9293
dx.doi.org/10.1021/es2022347 | Environ. Sci. Technol. 2011, 45, 9293–9300
Environmental Science & Technology
Figure 1. Map of the CMAQ/RSM modeling domain and spatial distributions of NH3 emissions.
control of individual pollutants. Pinder et al.6 conducted a series of PMCAMx simulations to estimate the cost-effectiveness and uncertainty of NH3 emission reductions on inorganic aerosols in the eastern United States and found that many currently available NH3 control technologies were cost-effective compared to SO2 and NOx. China, as the most populated country in the world, has significant agricultural activities which release large amounts of NH3 to the atmosphere. Enhanced concentrations of NH3 over the Beijing area in northeast China have been first detected in space-based nadir viewing measurements that penetrate into the lower atmosphere.10 The North China Plain (NCP), as shown in Figure 1, is one of the areas with the highest NH3 column density retrieved from infrared satellite observations.11 National NH3 emissions in China are estimated to be 1214 Tg for year 2000 and 1316 Tg for year 2005,1214 and account for 3055% of total Asia NH3 emissions.12,15,16 SO2 emissions have become better controlled in China.17 National emissions of SO2 were required by the government to be reduced 10% by 2010 compared to the level in 2005. However, such reduction of SO2 may adversely affect PM2.5, because it will lead to an increase in aerosol nitrate in the regions where air quality is more acidic.5,18,19 Additionally, in terms of acidification effects, Zhao et al.20 indicated that the benefits of SO2 reductions by 10% in China during 2005 to 2010 would almost be negated by the increase of NOx and NH3 emissions. Xing et al.18 suggested NH3 emission control should be considered to reduce the total nitrogen deposition in the future. Undoubtedly, NH3 is one of the most important pollutants which should receive attention; however, modeling studies to understand the impacts of NH3 emission on fine particles in China are quite limited. In this paper, we conducted 3-D air quality simulations in conjunction with the Response Surface Modeling (RSM) technique to investigate sensitivities of the PM components to changes of their precursor emissions, including SO2, NOx, NH3, NMVOC, and primary particles, in east China. Nonlinear impacts of NH3 emissions on SIA have been evaluated, and a more effective NH3 emission control pathway is recommended.
’ METHODOLOGY The processes involved are the establishment of emission inventories, selection of air quality modeling domain and
ARTICLE
configuration, development, and validation of the emission control/air quality response prediction using RSM methodology. Emission Inventory. Emissions of SO2, NOx, PM10, PM2.5, black carbon (BC), organic carbon (OC), NH3, and NMVOC were calculated based on the framework of the GAINS-Asia model.21 The general method and some improvements used to develop the China regional emission inventory are described in our previous papers.22 In 2005, NH3 emissions from livestock farming, N-fertilizer application, N-fertilizer production, and human excreta are estimated to be 7.16, 8.35, 0.17, and 0.87 Tg, respectively. The first two are the most important NH3 contributors; they account for 43% and 50% of total emissions, respectively. Urea, ammonium bicarbonate (ABC), and other fertilizers account for 56%, 22%, and 22% of the N-fertilizer used in China. The consumption of N-fertilizer has been increasing in the past 15 years. In 2010, the consumption of ammonia fertilizer was 26.4% higher than that in 2005. Large variations presented in the geographical distribution are shown in Figure 1. The North China Plain, including Henan, Shandong, Hebei, and Jiangsu Provinces, contribute approximately 33% of national emissions, with an emission intensity as high as 9.0 t km2, 4 times above the national average level (i.e., 1.7 t km2). NH3 emissions have strong seasonal variations since the related agricultural activities and emission factors (i.e., N-volatilization rate) are significantly affected by the meteorological conditions.12,14,23,24 Highest NH3 emissions occur during JuneAugust because of more favorable meteorological conditions (i.e., higher temperature) for NH3 volatilization and intensive agricultural activities. In this study, the monthly NH3 emissions in January, April, July, and October are estimated as 2.9%, 4.2%, 18.3%, and 7.5% of annual emissions, respectively. MM5/CMAQ Modeling. The air quality model used in this study is the Model-3/Community Multiscale Air Quality (CMAQ) modeling system (ver. 4.7), developed by U.S. EPA,25 which has been tested, evaluated, and applied in China.2631 A one-way nested technique is employed in this study. Modeling domain 1 covers almost all of China with a 36 36 km horizontal grid resolution and generates the boundary conditions for nested domain at 12 12 km resolution over East China. The three most developed regions, North China Plain (NCP), Yangtze River delta (YRD), and Pearl River delta (PRD), have been chosen as the target areas, as shown in Figure 1 and Table S1. The target period is January, April, July, and October in 2005. A complete description of CMAQ configuration, meteorological, emission, and initial and boundary condition inputs used for this analysis are described in Xing et al.18,33 The Aerosol Optical Depth (AOD), NO2 and SO2 column density, as well as the ground concentrations of SO2, NO2, PM10, PM2.5, and its chemical components simulated by this modeling system have been validated through comparison with observations of satellite retrievals and surface monitoring data. Response Surface Modeling (RSM) Technique. A real-time emission control/air quality response tool, i.e., RSM, was developed at the U.S. EPA and applied to a number of air quality policy analyses and assessments.32 RSM uses advanced statistical techniques to characterize the relationships between model outputs (i.e., air quality responses) and input parameters (i.e., emission changes) in a highly efficient manner. Table 1 gives the sampling method and numbers of simulations used in this RSM application. Following the principle of RSM development as discussed in our previous paper,33 the responses of PM concentrations to 9294
dx.doi.org/10.1021/es2022347 |Environ. Sci. Technol. 2011, 45, 9293–9300
Environmental Science & Technology
ARTICLE
Table 1. Sample Methods and Key Parameters Involved during PM RSM Development RSM case
variable number
sample method
sample number
LHS-30-a
total-NOx, total-SO2
Latin hypercube sampling
30
LHS-30-b
total-NOx, total-NH3
Latin hypercube sampling
30
LHS-30-c
total-SO2, total-NH3
Latin hypercube sampling
30
LHS-30-d
total-NOx, total-NMVOC
Latin hypercube sampling
HSS-100
total-NOx, SO2, NH3, NMVOC, and PM
Hammersley quasi-random sequence sample
the changes of the total emissions of SO2, NOx, NH3, NMVOC, and PM over east China have been calculated. We define “emission ratios” as the ratio of the changed emissions compared to the baseline emissions. The emissions of each pollutant change from 0 to 200%, which means the emission ratios are from 0 to 2. We used 100 random emission control scenarios generated by Hammersley quasi-random Sequence Sample (HSS) method to establish the emission-based prediction model (HSS-100). In this study, RSM surface (emissions control and corresponding concentration change) prediction system is statistically generalized by MPerK (MATLAB Parametric Empirical Kriging) program followed Maximum Likelihood EstimationExperimental Best Linear Unbiased Predictors (MLE-EBLUPs).34 Such control/response prediction system (i.e., HSS-100) has been validated through “leave-one-out cross validation” (LOOCV) (see Table S2), “out of sample” validation (see Table S3) and 2-D isopleths validation (see Figures S1 and S2). These results indicate that the HSS-100 predictions have good accuracy compared with CMAQ simulations. The stability of RSM with high dimensions (i.e., HSS-100) has been confirmed through its comparison with the one with low dimensions (i.e., LHS-30).
’ RESULTS AND DISCUSSION PM2.5 Sensitivity to NH3 Emissions. Following other sensitivity studies,35,36 we defined the PM2.5 sensitivity as the change ratio of PM2.5 concentration change to the change ratio of emissions, to evaluate the control effects of each pollutant,
SXa ¼
ΔC=C ðC Ca Þ=C ¼ ΔEX =EX 1a
ð1Þ
where SXa is the PM2.5 sensitivity to pollutant X (i.e., SO2, NOx, NH3, NMVOC, and PM) at its emission ratio a; Ca is the concentration of PM2.5 when the emission ratio of X is a; C* is the baseline concentration of PM2.5 (when emission ratio of X is 1). Figure 2 gives the comparison of PM2.5 sensitivities to the emissions of each pollutant (i.e., SO2, NOx, NH3, NMVOC, and PM) in the three target regions. The SIA accounts for about 2050% of PM2.5 concentrations in three regions, which is consistent with observations.37 The PM2.5 sensitivities to PM emissions are about the same in various control levels. However, NH3, SO2, and NOx present significant nonlinear impacts; the PM2.5 sensitivities to their emissions get larger when more control efforts are taken, because of the transition between NH3-rich and NH3-poor conditions, the transition between NOx-limited and VOC-limited for ozone chemistry regimes and other thermodynamic effect and etc. The PM2.5 response to NH3 emissions is comparable with that of SO2 and NOx, and it is larger under higher control levels. Under 50% control level, NH3, NOx, and SO2 emissions reduce 7.9%, 10.8%, and 10.4% of
30 100
PM2.5 concentrations in NCP; 10.7%, 7.7%, and 8.9% in YRD; 9.9%, 5.2%, and 10.8% in PRD; and 10.7%, 10.2%, and 11.4% in east China. Nonlinear Impacts of NH3 Emission on SO42 and NO3 Aerosol. The reaction mechanism of atmospheric chemistry is given in Figure S3. Using the Beijing urban site as an example, the nonlinear response of SO42 and NO3 aerosol concentrations to the emission changes of precursors, is given in Figure 3. For SO42 concentration, the dominating contributor is SO2 emissions (Figure 3a, c). NH3 emissions slightly enhance the SO42 concentrations under NH3-poor condition, because NH3 provides a weak base condition to uptake more SO2 and also enhances the cloud SO2 oxidation rate by O3. But no effects are found under NH3-rich condition for both January and July (Figure 3c). Lower NOx emissions (an emission ratio of 0.2 0.4 in January and 0.70.9 in July, higher in summer due to stronger atmospheric oxidation capacity than in winter) and suitable NOx/NMVOC ratios benefit SO42‑ generation (Figure 3b, d). The hydroxyl radical is the key reactive species in both homogeneous (SO2 + •OH) or aqueous-phase paths of SO42 formation. Both NOx and NMVOC could be involved in •OH removals during the generation of NO3 and RO2/ HO2, therefore suitable NOx/NMVOC ratios will enhance the generation of ozone, the major source of the hydroxyl radical. In NOx-rich conditions, the SO42 sensitivity to NOx emissions is negative. The results are opposite in NOx- poor conditions. For NO3 concentration, NOx emissions are the dominating contributor. However, NH3 emissions are very important under NH3-poor conditions (as shown in Figure 3b), because NH3 reacts preferentially with H2SO4 instead of HNO3. The sensitivities of NO3 concentration to NOx and NH3 emissions (under baseline, i.e., emission ratio =1) are relatively larger in summer than those in winter, because NO3 is very volatile in the summer (due to high temperature) and, thus, the equilibrium moves dominantly toward the gas-phase HNO3 instead of particle-phase NH4NO3. SO2 emissions slightly benefit NO3 formation under NH3-rich conditions, especially at lower SO2 emissions level (Figure 3c). This is caused by the thermodynamic effect.2 The increase of NH4+ and SO42 ions will decrease the NH4NO3 equilibrium constant, shifting its partitioning toward the particulate phase.38 However, when NH3 is insufficient, SO2 emissions inhibit NO3 formation due to its competition with NH3. NMVOC emission slightly contributes NO3 pollution under NOx-rich condition in both January and July, and NOx emission slightly inhibits NO3 formation under NOx-rich condition in January (Figure 3d). Identification of NH3-Rich/-Poor Condition. Indicators for PM chemistry such as the degree of sulfate neutralization (DSN), gas ratio (GR), and adjusted gas ratio (AdjGR) could be used to identify the NH3-poor, -neutral, or -rich condition, then to 9295
dx.doi.org/10.1021/es2022347 |Environ. Sci. Technol. 2011, 45, 9293–9300
Environmental Science & Technology
ARTICLE
Figure 2. PM2.5 concentration sensitivity to the stepped control of individual pollutants (PM2.5 sensitivity = change ratio of PM2.5/change ratio of emission; all values are monthly average in January, April, July, and October in 2005).
determine the sensitivity of NO3 to precursors’ emissions.39 Their definitions are given as follows: DSN ¼
GR ¼
½NHþ 4 ½NO3 ½SO4
2 ð½NH3 þ ½NHþ ½TA 2 ½TS 4 Þ 2 ½SO4 ¼ ½TN ½NO3 þ ½HNO3
ð2Þ
ð3Þ
AdjGR ¼
½NH3 þ ½NO ½TA DSN ½TS 3 ¼ ½TN ½NO þ ½HNO 3 3
ð4Þ
where [TA], [TN], and [TS] are the total molar concentration of ammonia ([NH3] + [NH+4 ]), nitrate ([NO 3 ] + [HNO3]), and sulfate ([SO2 4 ]), respectively. From RSM results, not only the NH3-rich/-poor condition under baseline scenario but also that under certain emission 9296
dx.doi.org/10.1021/es2022347 |Environ. Sci. Technol. 2011, 45, 9293–9300
Environmental Science & Technology
ARTICLE
Figure 3. 2-D Isopleths of SO42 and NO3 response to the emission changes of NOx, SO2, NH3, and NMVOC in Beijing, monthly average, 2005 (μg/m3).
control scenarios can be determined.30 The NH3-poor condition means the total available ammonia (gaseous ammonia + aerosol phase ammonium) is insufficient to charge-balance difference the remaining of other anions and cations,40 with the result that small perturbations in the ammonia emissions may have a significant effect on particle mass.41 Based on this principle, we defined an indicator—“Flex Ratio (FR)”—to identify the NH3-poor/-rich condition. As shown in Figure S4, under baseline NOx emissions (i.e., NOx emission ratio = 1), along with the decrease of NH3 emission ratio from 2.0 to 0, the NO3 slightly increases at first, but it gets sharply increased after the transition point (i.e., Flex Ratio). In the isopleths of NO3 response to NOx/NH3 emission changes predicted by RSM, the Flex Ratio is defined as the NH3 emission ratio at the flex NO3 concentrations under baseline NOx emissions (see Figure S4). When the FR is larger than the current NH3 emission ratio (in baseline = 1), the sensitivity of the NO3 concentration to NH3 emissions is more than that to NOx emissions, which indicate NH3-poor condition (see Table S4). In contrast, when the FR is less than 1, the NO3 concentration is more sensitive to NOx emissions instead of NH3 emissions, which indicates a NH3-rich condition, and the value (1 FR) reflects the ratio of free NH3 which could neutralize extra nitric acid produced by additional increases of NOx emissions. The spatial distributions of NO3 concentrations and GR are given in Figure S5. NO3 concentrations are found higher in January and lower in July, since higher temperature benefits NO3 evaporation and stronger atmospheric oxidation capacity favors converting S(IV) to S(VI), then enhancing the NH3
competition between SO42 and NO3 in July. Values of GR indicate NH3-rich, neutral, and poor conditions.39 The spatial distributions of GR value suggest that most of the polluted areas are located in NH3-rich conditions in all months (i.e., GR > 1). The FR over east China is shown in Figure 4. The FR derived from RSM gives consistent results, and the FR values in heavy NO3 pollution areas are mainly below 0.8. On an average annual basis, NCP and YRD are mainly located in NH3-rich conditions (FR is 0.60.7 and 0.81.0, respectively), therefore NO3 is more sensitive to NOx emissions, but PRD is located in NH3-poor conditions (FR is 1.01.5) and NO3 in PRD is more sensitive to NH3 emissions. The FR is around 0.8 in high NO3 areas, indicating NH3 is sufficiently abundant to satisfy an additional 25% (= 1/0.8 1) increase of NOx emissions to generate NO3. Impacts of NH3 Emission Increase on SO42 and NO3 Aerosols. Previous studies on the emission trends in China indicate the NH3 emissions have been growing along with other precursors. According to these results, the emission trends for each pollutant during 19902005 could be fitted by parameterized quadratic functions, as shown in Figure 5a. NOx is the fastest growing pollutant, increasing over 100% from 1990 to 2005. SO2 emissions have increased by 30% during the same period. The NH3 and NMVOC emissions in 2005 are about 90% increased from that in 1990. The growth trends of SO42 and NO3 concentrations driven by the increases of the emissions during 19902005 have been calculated by RSM. The results are given in Figure 5 9297
dx.doi.org/10.1021/es2022347 |Environ. Sci. Technol. 2011, 45, 9293–9300
Environmental Science & Technology
ARTICLE
Figure 4. Flex ratio in January, April, July, and October, 2005 (FR < 1 suggests NH3-rich condition; FR > 1 suggests NH3-poor condition).
Figure 5. Historical and future growth of emissions impacts on SO42 and NO3 (average of 4 months, in east China).
(in a 4-month average). As seen in Figure 5, the base scenario reflects the impacts of all five pollutants emission simultaneous changes with SO42 and NO3 concentration. In addition, a series of hypothetical scenarios has been conducted to evaluate the impacts of each pollutant emission change on SO42 and NO3 concentrations. In each hypothetical scenario, one pollutant is held at the 1990 level (i.e., no increases during 19902005) and the rest are kept the same as the base scenario. In the baseline, the NO3 and SO42 concentrations increase by 150% and 20%, respectively. It is obvious that the growth of NOx
and SO2 emissions are the dominant factor to enhance NO3 and SO42, respectively. Significant impacts could also be seen from the growth of NH3 emissions. About 5060% increases of NO3 and SO42 are caused by the growth of NH3 emissions. The growths of NMVOC and SO2 emissions have no significant impacts on NO3, while the growth of NOx hasnegative impacts on SO42 formation, possibly due to its influence on •OH as discussed in the previous section, especially during wintertime. Emissions of air pollutants and their projections have been changing significantly in recent years. The satellite data have 9298
dx.doi.org/10.1021/es2022347 |Environ. Sci. Technol. 2011, 45, 9293–9300
Environmental Science & Technology shown that NO2 increase in East Asia has been growing much faster than previous projections. Therefore, it is important to understand how China’s air pollutant emission change will affect the regional air quality in the future. Alternative scenarios for future SO2, NOx, and NMVOC emissions 18 were developed using forecasts of energy consumption and emission control strategies based on emissions in 2005, and on recent development plans for key industries in China, as shown in Figure 5b and c. In the reference scenario, which is based on the current control legislations and implementation status, i.e., REF scenario, the emissions of all pollutants are increasing from 2005 to 2030. In 2030, NO3 and SO42 will increase significantly, by 50% and 10%, respectively. In 2030, the NH3 emissions will increase by 20%, which may cause 15% and 4% increase of NO3 and SO42, respectively. In the policy scenario, which is based on the improvement of energy efficiencies and strict environmental legislation, i.e., PC2 scenario, though NOx emissions will be better controlled in 2030, the increase of NH3 emissions will enhance NO3 by 10%. The decrease of SO2 emissions leads to significant reduction of SO42, while the growth of NH3 will slightly improve SO42 by 2%. This implies future potential control of NH3 is important, especially for NO3 reduction. NH3 Impacts on the Acidity of Aerosols. The major concern about the potential negative impacts of NH3 control is the enhancement of aerosol acidity. In this study, we select the DSN as the indicator of the acidity of aerosols. When the DSN is less than 2, SO42‑ is insufficiently neutralized and the aerosol is more likely to be acid. The NH3 emissions level resulting in DSN less than 2 are calculated from RSM. Its spatial distributions over four months are given in Figure S6. High NH3 emissions are beneficial to the formation of NO3. Over the polluted areas such as NCP and YRD which have the highest NH3 emission intensities,14 the values are 0.81 in January, April, and October, but higher than 1 in July. This indicates the acidity of aerosols is more sensitive to NH3 emissions in summer than in other seasons, mainly because of the high evaporation of NO3 in summer and the stronger atmospheric oxidation capacity which converts S(IV) to S(VI) and enhances the NH3 competition between SO42 and NO3 in July. Therefore, the acidity of aerosols is more sensitive to NH3 emissions in the summer than in other seasons.
’ ASSOCIATED CONTENT
bS
Supporting Information. This information is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +86-10-62771466; fax: +86-10-62773650; e-mail: shxwang@ tsinghua.edu.cn.
’ ACKNOWLEDGMENT The study was financially supported by Natural Science Foundation of China (20921140409), MEP’s Special Funds for Research on Public Welfares (201009001), and the U.S. EPA. We thank Dr. Thomas J. Santner and Dr. Gang Han at The Ohio State University for their help using the MperK program and Satoru Chatani from Toyota Central R&D Laboratories for aid
ARTICLE
with emission processing. We appreciate that Dr. Chuck Freed helped improve the language of the paper.
’ REFERENCES (1) Makar, P. A.; Moran, M. D.; Zheng, Q.; Cousineau, S.; Sassi, M.; Duhamel, A.; Besner, M.; Davignon, D.; Crevier, L.-P.; Bouchet, V. S. Modelling the impacts of ammonia emissions reductions on North American air quality. Atmos. Chem. Phys. 2009, 9, 7183–7212, DOI: 10.5194/acp-9-7183-2009. (2) Tsimpidi, A. P.; Karydis, V. A.; Pandis, S. N. Response of Inorganic Fine Particulate Matter to Emission Changes of Sulfur Dioxide and Ammonia: The Eastern United States as a Case Study. J. Air Waste Manage. Assoc. 2007, 57, 1489–1498, DOI: 10.3155/ 1047-3289.57.12.1489. (3) Nguyen, K.; Dabdub, D. NOx and VOC Control and Its Effects on the Formation of Aerosols. Aerosol Sci. Technol. 2002, 36, 560–572. (4) Mueller, S. F.; Bailey, E. M.; Kelsoe, J. J. Geographic Sensitivity of Fine Particle Mass to Emissions of SO2 and NOx. Environ. Sci. Technol. 2004, 38, 570–580. (5) Blanchard, C. L.; Tanenbaum, S.; Hidy, G. M. Effects of Sulfur Dioxide and Oxides of Nitrogen Emission Reductions on Fine Particulate Matter Mass Concentrations: Regional Comparisons. J. Air Waste Manage. Assoc. 2007, 57, 1337–1350, DOI: 10.3155/1047-3289.57.11.1337. (6) Pinder, R. W.; Adams, P. J.; Pandis, S. N. Ammonia Emission Controls as a Cost-Effective Strategy for Reducing Atmospheric Particulate Matter in the Eastern United States. Environ. Sci. Technol. 2007, 41, 380–386. (7) Tsimpidi, A. P.; Karydis, V. A.; Pandis, S. N. Response of Fine Particulate Matter to Emission Changes of Oxides of Nitrogen and Anthropogenic Volatile Organic Compounds in the Eastern United States. J. Air Waste Manage. Assoc. 2008, 58, 1463–1473, DOI: 10.3155/1047-3289.58.11.1463. (8) Redington, A. L.; Derwent, R. G.; Witham, C. S.; Manning, A. J. Sensitivity of modelled sulphate and nitrate aerosol to cloud, pH and ammonia emissions. Atmos. Environ. 2009, 43, 3227–3234. (9) Derwent, R.; Witham, C.; Redington, A.; Jenkin, M.; Stedman, J.; Yardley, R.; Hayman, G. Particulate matter at a rural location in southern England during 2006: Model sensitivities to precursor emissions. Atmos. Environ. 2009, 43, 689–696. (10) Beer, R.; Shephard, M. W.; Kulawik, S. S.; Clough, S. A.; Eldering, A.; Bowman, K. W.; Sander, S. P.; Fisher, B. M.; Payne, V. H.; Luo, M. Z.; Osterman, G. B.; Worden, J. R. First satellite observations of lower tropospheric ammonia and methanol. Geophys. Res. Lett. 2008, 35, L09801, DOI: 10.1029/2008GL033642. (11) Clarisse, L.; Clerbaux, C.; Dentener, F.; Hurtmans, D.; Coheur, P. F. Global ammonia distribution derived from infrared satellite observations. Nat. Geosci. 2009, 2, 479–483, DOI: 10.1038/ngeo551. (12) Streets, D. G.; Bond, T. C.; Carmichael, G. R.; Fernandes, S. D.; Fu, Q.; He, D.; Klimont, Z.; Nelson, S. M.; Tsai, N. Y.; Wang, M. Q.; Woo, J. H.; Yarber, K. F. An inventory of gaseous and primary aerosol emissions in Asia in the year 2000. J. Geophys. Res. 2003, 108, 8809, DOI: 10.1029/2002JD003093. (13) Wang, S. W.; Liao, Q. J. H.; Hu, Y. T.; Yan, X. Y. A Preliminary Inventory of NH3-N Emission and Its Temporal and Spatial Distribution of China. Chin. J. Agro-Environ. Sci. 2009, 28 (3), 619–629. (14) Dong, W. X.; Xing, J.; Wang, S. X. Temporal and Spatial Distribution of Anthropogenic Ammonia Emissions in China: 1994 2006. Chin. J. Environ. Sci. 2010, 31 (7), 1457–1463. (15) Zhao, D.; Wang, A. Emission of anthropogenic ammonia emission in Asia. Atmos. Environ. 1994, 28, 689–694. (16) Yamaji, K.; Ohara, T.; Akimoto, H. Regional-specific emission inventory for NH3, N2O, and CH4 via animal farming in South, Southeast, and East Asia. Atmos. Environ. 2004, 38, 7111–7121. (17) Lu, Z.; Streets, D. G.; Zhang, Q.; Wang, S.; Carmichael, G. R.; Cheng, Y. F.; Wei, C.; Chin, M.; Diehl, T.; Tan, Q. Sulfur dioxide emissions in China and sulfur trends in East Asia since 2000. Atmos. Chem. Phys. 2010, 10, 6311–6331. 9299
dx.doi.org/10.1021/es2022347 |Environ. Sci. Technol. 2011, 45, 9293–9300
Environmental Science & Technology (18) Xing, J.; Wang, S. X.; Chatani, S.; Zhang, C. Y.; Wei, W.; Klimont, Z.; Cofala, J.; Amann, M.; Hao, J. M. Projections of Air Pollutant Emissions and its Impacts on Regional Air Quality in China in 2020. Atmos. Chem. Phys. 2011, 11, 3119–3136, DOI: 10.5194/ acp-11-3119-2011. (19) West, J. J.; Ansari, A. S.; Pandis, S. N. Marginal PM2.5: Nonlinear aerosol mass response to sulfate reductions in the eastern United States. J. Air Waste Manage. Assoc. 1999, 49, 1415–1424. (20) Zhao, Y.; Duan, L.; Xing, J.; Larssen, T.; Nielsen, C. P.; Hao, J. M. Soil Acidification in China: Is Controlling SO2 Emissions Enough? Environ. Sci. Technol. 2009, 43, 8021–8026. (21) Amann, M.; Bertok, I.; Borken, J.; Chambers, A.; Cofala, J.; Dentener, F.; Heyes, C.; Hoglund, L.; Klimont, Z.; Purohit, P.; Rafaj, P.; Sch€opp, W.; Toth, G.; Wagner, F.; Winiwarter, W. A Tool to Combat Air Pollution and Climate Change Simultaneously; Methodology report; International Institute for Applied Systems Analysis (IIASA): Laxenburg, Austria, 2008. (22) Klimont, Z.; Cofala, J.; Xing, J.; Wei, W.; Zhang, C.; Wang, S.; Kejun, J.; Bhandari, P.; Mathur, R.; Purohit, P.; Rafaj, P.; Chambers, A.; Amann, M. Projections of SO2, NOx and carbonaceous aerosols emissions in Asia. Tellus, Ser. B 2009, 61, 602–617. (23) Zhang, M. S.; Luan, S. J. Application of NARSES in Evaluation of Ammonia Emission from Nitrogen Fertilizer Application in Planting System in China. Chin. J. Anhui Agric. Sci. 2009, 37 (8), 3583–3586. (24) Cao, G. L.; An, X. Q.; Zhou, C. H.; Ren, Y. Q.; Tu, J. Emission inventory of air pollutants in China. Chin. Environ. Sci. 2010, 30 (7), 900–906. (25) Byun, D. W.; Schere, L. K. Review of the governing equations, computational algorithms and other components of the models-3 Community Multiscale Air Quality (CMAQ) Modeling System. Appl. Mech. Rev. 2006, 59 (2), 51–77. (26) Zhang, M.; Uno, I.; Zhang, R.; Han, Z.; Wang, Z.; Pu, Y. Evaluation of the Models-3 Community Multi-scale Air Quality (CMAQ) modeling system with observations obtained during the TRACE-P experiment: comparison of ozone and its related species. Atmos. Environ. 2006, 40 (26), 4874–4882. (27) Streets, D. G.; Fu, J. S.; Jang, C.; Hao, J.; He, K.; Tang, X.; Zhang, Y.; Li, Z.; Zhang, Q.; Wang, L.; Wang, B.; Yu, C. Air quality during the 2008 Beijing Olympic games. Atmos. Environ. 2007, 41 (3), 480–492. (28) Fu, J. S.; Jang, C. J.; Streets, D. G.; Li, Z.; Kwok, R.; Park, R.; Hang, Z. MICS-Asia II: Evaluating gaseous pollutants in East Asia using an advanced modeling system: Models-3/CMAQ System. Atmos. Environ. 2008, 42 (15), 3571–3583. (29) Fu, J. S.; Streets, D. G.; Jang, C. J.; Hao, J.; He, K.; Wang, L.; Zhang, Q. Modeling Regional/Urban Ozone and Particulate Matter in Beijing, China. J. Air Waste Manage. Assoc. 2009, 59, 37–44. (30) Wang, L.; Carey, C. J.; Zhang, Y.; Wang, K.; Zhang, Q.; Streets, D. G.; Fu, J.; Lei, Y.; Schreifels, J.; He, K.; Hao, J.; Lam, Y. F.; Lin, J.; Meskhidze, N.; Voorhees, S.; Evarts, D.; Phillips, S. Assessment of air quality benefits from national air pollution control policies in China. Part II: Evaluation of air quality predictions and air quality benefits assessment. Atmos. Environ. 2010, 44, 3449–3457. (31) Li, L.; Chen, C. H.; Fu, J. S.; Huang, C.; Streets, D. G.; Huang, H. Y.; Zhang, G. F.; Wang, Y. J.; Jang, C. J.; Wang, H. L.; Chen, Y. R.; Fu, J. M. Air quality and emissions in the Yangtze River Delta, China. Atmos. Chem. Phys. 2011, 11, 1621–1639. (32) U.S. Environmental Protection Agency. Technical Support Document for the Proposed PM NAAQS Rule: Response Surface Modeling; Office of Air Quality Planning and Standards: Research Triangle Park, NC, 2006. (33) Xing, J.; Wang, S. X.; Jang, C.; Zhu, Y.; Hao, J. M. Nonlinear Response of Ozone to Precursor Emission Changes in China: a Modeling Study using Response Surface Methodology. Atmos. Chem. Phys. 2011, 11, 5027–5044, DOI: 10.5194/acp-11-5027-2011. (34) Santner, T. J.; Williams, B. J.; Notz, W. The Design and Analysis of Computer Experiments; Springer Verlag: New York, 2003. (35) Yarwood, G.; Wilson, G.; Morris, R. Development of The CAMx Particulate Source Apportionment Technology (Psat), final report; ENVIRON International Corporation, 2005.
ARTICLE
(36) Koo, B.; Wilson, G. M.; Morris, R. E.; Dunker, A. M.; Yarwood, G. Comparison of Source Apportionment and Sensitivity Analysis in a Particulate Matter Air Quality Model. Environ. Sci. Technol. 2009, 43, 6669–6675. (37) Chan, C. K.; Yao, X. H. Air pollution in mega cities in China. Atmos. Environ. 2008, 42, 1–42. (38) Seinfeld, J. H.; Pandis, S. N. Atmospheric Chemistry and Physics: From Air Pollution to Climate Change; John Wiley and Sons, Inc.: New York, 2006. (39) Zhang, Y.; Wen, X. Y.; Wang, K.; Vijayaraghavan, K.; Jacobson, M. Z. Probing into Regional O3 and PM Pollution in the U.S., Part II. An Examination of Formation Mechanisms through a Process Analysis Technique and Sensitivity Study. J. Geophys. Res. 2009, 114(D22305), DOI: 10.1029/2009JD011900. (40) Blanchard, C. L.; Roth, P. M.; Tanenbaum, S. J.; Ziman, S. D.; Seinfeld, J. H. The use of ambient measurements to identify which precursor species limit aerosol formation. J. Air Waste Manage. Assoc. 2000, 50, 2073–2084. (41) Makar, P. A.; Moran, M. D.; Zheng, Q.; Cousineau, S.; Sassi, M.; Duhamel, A.; Besner, M.; Davignon, D.; Crevier, L.-P.; Bouchet, V. S. Modelling the impacts of ammonia emissions reductions on North American air quality. Atmos. Chem. Phys. 2009, 9, 7183–7212, DOI: 10.5194/acp-9-7183-2009.
9300
dx.doi.org/10.1021/es2022347 |Environ. Sci. Technol. 2011, 45, 9293–9300
ARTICLE pubs.acs.org/est
Plasma-TiO2 Catalytic Method for High-Efficiency Remediation of p-Nitrophenol Contaminated Soil in Pulsed Discharge Tie Cheng Wang,† Na Lu,†,‡ Jie Li,†,‡,* and Yan Wu†,‡ † ‡
Institute of Electrostatics and Special Power, Dalian University of Technology, Dalian 116024, PR China Key Laboratory of Industrial Ecology and Environmental Engineering, Ministry of Education of the People’s Republic of China, Dalian 116024, PR China
bS Supporting Information ABSTRACT: Nonthermal discharge plasma and TiO2 photocatalysis are two techniques capable of organic pollutants removal in soil. In the present study, a pulsed discharge plasma-TiO2 catalytic (PDPTC) technique by combining the two means, where catalysis of TiO2 is driven by the pulsed discharge plasma, is proposed to investigate the remediation of p-nitrophenol (PNP) contaminated soil. The experimental results showed that 88.8% of PNP was removed within 10 min of treatment in PDPTC system and enhancing pulse discharge voltage was favorable for PNP degradation. The mineralization of PNP and intermediates generated during PDPTC treatment was followed by UV-vis spectra, denitrification, total organic carbon (TOC), and COx selectivity analyses. Compared with plasma alone system, the enhancement effects on PNP degradation and mineralization were attributed to more amounts of chemically active species (e.g., O3 and H2O2) produced in the PDPTC system. The main intermediates were identified as hydroquinone, benzoquinone, catechol, phenol, benzo[d][1, 2, 3]trioxole, acetic acid, formic acid, NO2, NO3, and oxalic acid. The evolution of the main intermediates with treatment time suggested the enhancement effect of the PDPTC system. A possible pathway of PNP degradation in soil in such a system was proposed.
’ INTRODUCTION Nitrophenols are toxic and biorefractory organic compounds, used extensively as raw materials and intermediates in the production of explosives, pharmaceuticals, pesticides, pigments, dyes, wood preservatives, and rubber chemicals.1 Three nitrophenols (2-nitrophenol, 4-nitrophenol, and 2, 4-dinitrophenol) have been listed in the USEPA list of priority pollutants.2 Nitrophenols were released to soil as fugitive emissions during their production and use, causing serious health hazards. Taking China as an example, some contaminated lands, where nitrophenols extensively exist, have been left at the center of the city after some chemical factories migrated to the suburb in the industrial rearrangement. These lands are of high economic values because of their locations, and hence, need to be remedied rapidly. Several technologies such as physical method,3 chemical method,4 bioremediation,5 electrokinetic remediation,6 thermal technology,7 and photocatalysis8 have been employed to remediate organic pollutants contaminated soils. With the increasing strictness of industrial standard and the increment of economic values of lands, high-efficient and rapid soil remediation method is becoming a necessity. In this case, the conventional remediation technologies would not meet the requirement of high-efficient and rapid r 2011 American Chemical Society
remediation due to the drawbacks such as second pollution and time-consuming. In our previous study, pulsed discharge plasma technology, one of the advanced oxidation processes, has been employed to remediate pentachlorophenol contaminated soil, and great performance of soil remediation was obtained in a short remediation period.9 In pulsed discharge plasma process, discharge energy is released in forms of high energy electrons, strong electric field, and UV light radiation, etc., some of which can excite gases in plasma region to generate chemically active species. However, the utilization efficiency of the discharge energy is still needed to be enhanced in order to satisfy the requirement of practical application. Catalysis is suggested to be viable by introducing the active catalyst into the discharge plasma soil remediation system. Anatase TiO2, an economic and photosensitive semiconductor material with a band gap of about 3.2 eV, can be excited by strong electric field and UV light radiation to generate pairs of electrons and holes, resulting in more numerous chemically Received: April 26, 2011 Accepted: September 16, 2011 Revised: September 15, 2011 Published: September 16, 2011 9301
dx.doi.org/10.1021/es2014314 | Environ. Sci. Technol. 2011, 45, 9301–9307
Environmental Science & Technology
ARTICLE
Figure 1. Effect of pulsed discharge voltage on PNP degradation in PDPTC system.
Figure 2. Evolution of UV-vis absorption spectra of PNP with treatment time.
active species generation.10 Recently, the combination of nonthermal discharge plasma with TiO2 photocatalysis for pollutant removal has drawn great attention.1113 Previous studies presented that the introduction of TiO2 into discharge plasma system could enhance pollutants removal and promote energy efficiency.11 However, relevant research was mostly focused on wastewater treatment and little has been reported on soil remediation. In this work, a pulsed discharge plasma-TiO2 catalytic (PDPTC) technique is proposed to enhance the remediation of p-nitrophenol (PNP) contaminated soil. The study was focused on exploring the enhancement of PNP degradation in this PDPTC system, and the variances of the main intermediates between PDPTC system and plasma alone system were compared to evaluate the enhanced behavior. Possible mechanisms of such enhancement were discussed by analyzing the variances in the amounts of chemically active species. A possible pathway of PNP degradation in such a system was proposed.
Extraction and Analysis. After discharge treatment, PNP in soil was extracted immediately, and the extraction procedure was described in SI S4. The extractions produced average recoveries of 90.195.3%. PNP concentration, total organic carbon (TOC), intermediates, O3 and H2O2 concentrations, and CO2 and CO were analyzed and the details were shown in SI S5. The COx selectivity, CO2 conversion, and denitrification efficiency were defined in SI S6. All experiments were conducted in duplicates.
’ MATERIALS AND METHODS Materials. PNP was used in the study, and its detailed introduction was presented in S1 of the Supporting Information (SI). Soil samples were collected from a suburb of Dalian, China. The details were presented in SI S2. The original PNP concentration in the soil was 800 mg kg1. TiO2 (Degussa, P25) (BET area =50 m2 g1) was used as the catalyst. Treatment of Contaminated Soil. The schematic diagram of the experimental apparatus was illustrated in Figure S1 in the SI, which was similar with our previous work.9 The details of the reactor were showed in SI S3. The pulse frequency and pulseforming capacitance Cp were 100 Hz and 200 pF, respectively, and the input energies per pulse were 0.016, 0.020, and 0.023 J at pulse voltage of 16, 18, and 20 kV, respectively. The experiments were all conducted at 20 kV unless special illustration. In each experiment, a certain amount of TiO2 was added into PNP contaminated soil and then homogenized. TiO2 amount in the soil was 2 wt %. The soil sample (approximately 2.0 g) was spread on the ground electrode with a thickness of about 1.3 mm. Prior to discharge treatment, the moisture content of soil was adjusted to 10% with deionized water. Air was injected for one side of the reactor and out from the other side with the flow rate of 0.5 L min1.
’ RESULTS AND DISCUSSION PNP Degradation in PDPTC System. Figure 1 showed the effect of pulse discharge voltage on PNP degradation in soil in PDPTC system. On the one hand, the introduction of TiO2 enhanced PNP removal in soil. At pulsed discharge voltage of 20 kV, PNP degradation efficiency reached 88.8% after 10 min of discharge treatment in the PDPTC system, which was 78.1% in plasma alone system. In the case of TiO2 catalyst, conduction band electrons and holes (h+) would be generated when TiO2 was irradiated with high energy input, and the photogenerated electrons and holes could react with PNP directly or indirectly, increasing PNP degradation. Hoffmann14 has reported that the conduction band electrons, holes (h+) and the reactive oxygen species such as •OH radicals and superoxide radicals generated on the illuminated catalyst could promote pollutant removal. On the other hand, increase in pulsed discharge voltage greatly enhanced PNP degradation efficiency in the PDPTC system. For example, at discharge voltage of 16 kV, only 65.2% of PNP was removed after 10 min of discharge treatment, while it increased to 88.8% at 20 kV. More energy is injected into the reactor when the discharge voltage increases, and then more plasma channels with strong energy are very effective to generate more amounts of chemically active species, and therefore PNP degradation efficiency is enhanced; meanwhile, the effects including high energy electrons, strong electric field and UV light radiation etc would also become stronger at higher discharge voltage. In this case, more conduction band electrons and holes (h+) would be generated, and then the formation of chemically active species was accelerated, which promoted PNP removal. In addition, the energy efficiencies for PNP removal in the PDPTC system after 10 min of discharge treatment were 3.90, 3.84, and 3.70 g kWh1 at pulse voltages of 16, 18, and 20 kV, respectively, as presented in SI Table S1. Considering PNP degradation and energy efficiency comprehensively, 20 kV was used in the following experiments. 9302
dx.doi.org/10.1021/es2014314 |Environ. Sci. Technol. 2011, 45, 9301–9307
Environmental Science & Technology
Figure 3. Evolution of NO2 and NO3 with treatment time.
Control experiments with O3 and H2O2 addition in the absence of plasma were conducted, and the results were presented in SI Table S2. Herein, method of ozonation experiment was the same as our previous study.9 The results in SI Table S2 suggested that O3 played a decisive role for PNP degradation, and H2O2 also played a certain role. Mineralization of PNP in PDPTC System. Mineralization of PNP in the PDPTC system was studied by changes of UVvis spectra, NO3 and NO2 formation, TOC removal, and CO2 and CO generation. UV-vis Absorption Spectra. The change of the UVvis absorption spectra of PNP during degradation process in the PDPTC system was shown in Figure 2. The absorption peak at 400 nm disappeared quickly with the increase of treatment time, indicating that PNP was removed gradually. Formation of NO2 and NO3. Since there is a nitro-group in PNP molecule, tracing changes of nitrogen forms is an approach to evaluate the degree of PNP degradation. To our knowledge, the nitro-group can be converted into NO3 and NO2 ions.15,16 Therefore, NO3 and NO2 ions were both monitored during PNP degradation in soil, and as control experiments, the formation of NO2 and NO3 were also analyzed in clean soil (uncontaminated soil) during pulsed discharge process. The concentrations of NO3 and NO2 released from PNP were calculated by subtracting the concentrations in clean soil from those in contaminated soil, respectively. The evolution of NO3 and NO2 concentrations released from PNP with treatment time was shown in Figure 3, and their concentrations generated in clean soil during pulsed discharge plasma were presented in SI Figure S2. The NO2 concentration in the PDPTC system declined after an initial increase in Figure 3, probably due to its further oxidation to other nitrogen forms, whereas the NO3 concentration increased gradually with treatment time. Similar trends were also presented in plasma alone system. Moreover, the NO3 concentration increased slowly in the initial 10 min, and then the rate increased gradually. These results indicated that the NO2 mainly resulted from PNP degradation, and NO2 was formed first when the nitrogen-tocarbon single bond (—N—C—) of the PNP was broken down, and then it was oxidized into NO3. Active species reacted rapidly with nitrophenol to produce NO2, and the NO2 concentration quickly reached a maximum and then decreased rapidly, and during the process it was oxidized into NO3.15,16 More importantly, as shown in Figure 3, more amounts of NO3 and NO2 were formed in the PDPTC system than in plasma alone system in the initial 10 min, whereas higher NO3 and lower NO2 concentrations occurred in the PDPTC system after
ARTICLE
Figure 4. Evolution of TOC removal with treatment time.
Figure 5. Changes of COx selectivity with treatment time in PDPTC system.
30 min of treatment, compared with those in plasma alone system. These results suggested that more amounts of NO2 were oxidized into NO3 in the PDPTC system, which was attributed to the intense oxidation environment caused by TiO2 catalyst. The enhanced oxidation environment in the PDPTC system could also be further confirmed by denitrification efficiency. The evolution of denitrification efficiency of PNP with treatment time was presented in SI Figure S3. It was found that 81.3% of denitrification efficiency was achieved after 30 min of treatment in the PDPTC system, and there was a 20% rise as compared with that in plasma alone system. TOC Removal. The TOC values have been related to the total concentration of organic compounds, and the decrease of TOC with treatment time can reflect the degree of mineralization. Therefore, the TOC removal during PNP degradation process in the PDPTC system was shown in Figure 4. TOC removal efficiency achieved 55.1% in the PDPTC system after 30 min of treatment, compared with that of 42.9% in plasma alone system. These results demonstrated that more PNP and intermediates were mineralized to smaller organic molecules or to CO2 in the PDPTC system. Generation of CO2 and CO in Offgas. The changes of UVvis absorption spectra, the formation of NO3 and NO2, and the TOC removal only reflect indirectly the mineralization extent of PNP, whereas the generation of CO2 and CO can reflect directly its mineralization. Therefore the generation of CO2 and CO during PNP degradation was measured. Figure 5 presented the evolution of COx selectivity with treatment time during PNP degradation. With the treatment time continued, CO2 selectivity increased and CO selectivity decreased in the 9303
dx.doi.org/10.1021/es2014314 |Environ. Sci. Technol. 2011, 45, 9301–9307
Environmental Science & Technology
ARTICLE
Table 1. Comparison of O3 and H2O2 Concentrations and Their Energy Costs with/Without TiO2 in Plasma Process uncontaminated soil
contaminated soil
CO3 (mgL-1)
CH2O2 (mmol L1)
O3 energy costs (mgkJ-1)
H2O2 energy costs (mgkJ-1)
CO3 (mgL-1)
CH2O2 (mmol L1)
without TiO2
42
0.038
0.76
0.014
23
0.022
with TiO2
57
0.062
1.03
0.022
30
0.041
•OH þ •OH f H2 O2
ð4Þ
O2 þ ecb f O2 •
ð5Þ
O2 • þ 2H2 O f H2 O2 þ 2OH þ O2
ð6Þ
On the other hand, O3 concentration in the PDPTC system was always higher than that in plasma alone system, as shown in Table 1. O2 can be cleaved into single atomic oxygen radical anion (O•) by gas phase electrical pulses, and it can also be converted to superoxide radical anion (O2•) by the effect of highly energized electrons (e) and ecb on the surface of TiO2. Subsequently, O• and O2• are transformed into more chemically active species (O and O2) by hvb+ on the surface of TiO2, resulting in more amounts of O3 production through the following possible reaction pathways:10,20,21 Figure 6. Evolution of main intermediates with treatment time.
O2 þ e f O2 •
ð7Þ
PDPTC system. 96.8% of CO2 selectivity was obtained in the PDPTC system after 20 min of treatment, with 3.2% of CO selectivity. These results suggested that CO formed in the reaction was further oxidized into CO2. More importantly, higher CO2 selectivity and lower CO selectivity occurred in the PDPTC system than in plasma alone system. The intense oxidative environment in the PDPTC system was the reason for the enhanced CO2 selectivity and suppressed CO selectivity. Formation of Active Species. To explore the enhancement mechanisms of PNP degradation in the PDPTC system, the variances of chemically active species such as O3 and H2O2 were analyzed in clean and contaminated soils, respectively, and the results were presented in Table 1. Herein the discharge time was 20 min. As shown in Table 1, on the one hand, more amounts of H2O2 were generated in the PDPTC system both in clean soil and contaminated one, compared with those in plasma alone system. When an electron on the valence band (VB) of TiO2 absorbs some energy higher than the band gap between the VB and the conduction band (CB), it will be promoted to the CB and thus an electronhole pair (ecbhvb+) is formed. The holes will oxidize either H2O molecule or OH anions to form •OH. The electrons on the CB react with O2 to generate superoxide radical anion (O2). Then, the O2 will react with H2O to produce •OH. •OH can react with each other to form H2O2. Therefore, more amounts of H2O2 were generated through the following possible reaction pathways:1719
O2 • þ hvb þ f O2
ð8Þ
O• þ hvb þ f O
ð9Þ
TiO2 þ hv sf ecb þ hvb þ
energy
ð1Þ
hvb þ þ H2 O f Hþ þ •OH
ð2Þ
hvb þ þ OH f •OH
ð3Þ
O þ O2 f O3
ð10Þ
As mentioned above, it is believed that the PDPTC system is effective to generate more chemically active species (such as O3 and H2O2), leading to the enhancement of PNP degradation in soil. Possible Degradation Pathways. The intermediates of PNP degradation in soil in the PDPTC system were analyzed using HPLC, HPLC/MS and IC. They mainly included hydroquinone, benzoquinone, catechol, phenol, benzo[d][1, 2, 3]trioxole, acetic acid, formic acid, NO2, NO3, and oxalic acid. Similar results were also reported by Oturan,22 where hydroquinone and benzoquinone were two main intermediates during PNP degradation by Fenton method. Hydroquinone, benzoquinone, phenol, formic acid, and oxalic acid were also detected as intermediates during PNP degradation by ozonation, and NO2 group could be easily removed from the aromatic ring in the process.23 The evolution of some aromatic intermediates with treatment time in the PDPTC system and plasma alone system were analyzed, as depicted in Figure 6. Maximum concentrations of hydroquinone, benzoquinone and catechol in the PDPTC system were all lower than those in the plasma alone system. Hydroquinone reached the maximum concentration earlier in the PDPTC system, whereas benzoquinone and catechol reached the maximum concentrations almost at the same time in the two discharge systems. Phenol was generated earlier in the PDPTC system. Besides that, these intermediates in each reaction system all encountered further degradation with the 9304
dx.doi.org/10.1021/es2014314 |Environ. Sci. Technol. 2011, 45, 9301–9307
Environmental Science & Technology
ARTICLE
Figure 7. Possible degradation pathways of PNP in soil in PDPTC system.
treatment process went on, and lower concentrations occurred in the PDPTC system finally. The results presented that greater degradation performance occurred in the PDPTC system, which was attributed to the enhanced generation of chemically active species by the effect of TiO2 catalyst. On the other hand, from the sequence of accumulation of aromatic intermediates, phenol appeared later than hydroquinone, benzoquinone and catechol. Therefore, it could be concluded that hydroquinone, benzoquinone and catechol were generated more easily than phenol during PNP degradation. The differences in evolution of intermediates were due to the intensive oxidation environment in the PDPTC system. On the one hand, more amounts of H2O2 and O3 generated in PDPTC system could increase the formation of •OH radicals through trapping of photogenerated electrons, and they also interfered in recombination of electrons and positive holes due to the electrophilic properties of H2O2 and O3.24 On the other hand, active species generated in discharge plasma could act directly on the active sites of TiO2, and then accelerate TiO2 to trigger reactions, and meanwhile, the strong electric field in discharge plasma could inhibit the recombination of electrons and holes on TiO2 surface.25,26 Based on the intermediates and their evolution with treatment time, and the roles of O3, H2O2, and •OH radicals played in the present study, possible degradation pathways of PNP in soil were proposed in Figure 7. The patterns of intermediates indicated
that hydroxylation was the main oxidation pathway. Hydroxylated intermediates could result from electrophilic attack on PNP by O3 and •OH radicals. Aromatic ring of PNP contains two substituents, OH and NO2. The OH is electron-donating and an ortho- and para-director, while NO2 is electron-withdrawing and meta-director. •OH radicals preferentially attack the ortho- or para-position with respect to the OH group due to the electrophilic nature. The •OH radicals may eliminate nitrous acid from PNP to yield 1,4-benzosemiquinone as an intermediate, which subsequently disproportionates into hydroquinone and benzoquinone. Similar results were reported by Liu et al.27 and Suarez et al.28 The possibility of a direct attack of •OH radicals at the position carrying the NO2 group also exists, with resultant hydroquinone formation.29 In addition, the •OH radicals may attack the NO2 group due to the relatively long length of C—N bond in PNP molecule, which is the longest bond and would be potential to be attacked to form phenol,30 and then hydroquinone, benzoquinone and catechol would be generated by further oxidation of phenol. Further reactions of these intermediates with •OH radicals lead to ring cleavage and formation of aliphatic compounds. O3 reacts with organic pollutants through nucleophilic, electrophilic and cyclo-addition reactions.31,32 The nucleophilic and electrophilic attacks of O3 on PNP proceeds preferentially on the ortho- and para-positions with respect to the OH group to yield 9305
dx.doi.org/10.1021/es2014314 |Environ. Sci. Technol. 2011, 45, 9301–9307
Environmental Science & Technology the same hydroxylated intermediates as in the case of •OH radicals attack.23 Furthermore, the cyclo-addition reactions of O3 may cause the addition of O3 to PNP molecule structure, and finally form cyclo-addition intermediates,33 such as benzo[d][1, 2, 3]trioxole in the present study. Further attacks to these intermediates by O3 would lead to ring-cleavage of the aromatic ring to form acetic acid, formic acid, and oxalic acid. The formation of NO2 is the result of the denitrification of PNP. In the process, the NO2 concentration quickly reaches a maximum and then decreases rapidly to be oxidized into NO3. This study opens a possible way to improve soil remediation in pulsed discharge plasma through TiO2 catalyst.
’ ASSOCIATED CONTENT
bS Supporting Information. Text S1S6 include introduction of PNP and other reagents, details of the soil sample, reactor introduction, extraction procedure, analysis methods, COx selectivity, and denitrification efficiency. Figures S1S3 include the reactor system, nitrite and nitrate concentrations in clean soil, and denitrification efficiency. Table S1 presents the energy efficiency within 10 min of discharge treatment at different discharge voltages. Table S2 presents the comparison of PNP degradation in soil by pulsed discharge plasma, ozonation, and H2O2 oxidation. This material is available free of charge via the Internet at http://pubs.acs.org. ’ AUTHOR INFORMATION Corresponding Author
*Phone: +86-411-84708576; fax: +86-411-84709869; e-mail: [email protected].
’ ACKNOWLEDGMENT We thank the National Natural Science Foundation, P.R. China (Project No. 40901150), the Ministry of Science and Technology, P.R. China (Project No. 2008AA06Z308), and Program for Liaoning Excellent Talents in University, China (Project No. 2009R09) for their financial support to this research. ’ REFERENCES (1) Uberoi, V.; Bhattacharya, S. K. Toxicity and degradability of nitrophenols in anaerobic systems. Water Environ. Res. 1997, 69 (2), 146156; DOI: 10.2175/106143097X125290. (2) USEPA; http://www.scorecard.org. 2002. (3) Khodadoust, A. P.; Bagchi, R.; Suidan, M. T.; Brenner, R. C.; Sellers, N. G. Removal of PAHs from highly contaminated soils found at prior manufactured gas operations. J. Hazard. Mater. 2000, 80 (13), 159174; DOI: 10.1016/S0304-3894(00)00286-7. (4) Liao, C. J.; Chung, T. L.; Chen, W. L.; Kuo, S. L. Treatment of pentachlorophenol-contaminated soil using nano-scale zero-valent iron with hydrogen peroxide. J. Mol. Catal. A: Chem. 2007, 265 (12), 189194; DOI: 10.1016/j.molcata.2006.09.050. (5) Lamar, R. T.; Evans, J. W.; Glaser, J. A. Solid-phase treatment of a pentachlorophenol-contaminated soil using lignin-degrading fungi. Environ. Sci. Technol. 1993, 27 (12), 25662571; DOI: 10.1021/ es00048a039. (6) Zhang, S. P.; Rusling, J. F. Dechlorination of polychlorinated biphenyls on soils and clay by electrolysis in a biocontinuous microemulsion. Environ. Sci. Technol. 1995, 29 (5), 11951199;DOI: 10.1021/es00005a009.
ARTICLE
(7) Acharya, P.; Ives, P. Incineration at bayou bounfouca remediation project. Waste Manage. 1994, 14 (1), 1326; DOI: 10.1016/0956053X(94)90017-5. (8) Balmer, M. E.; Goss, K. U.; Schwarzenbach, R. P. Photolytic transformation of organic pollutants on soil surfaces-An experimental approach. Environ. Sci. Technol. 2000, 34 (7), 12401245; DOI: 10.1021/es990910k. (9) Wang, T. C.; Lu, N.; Li, J.; Wu, Y. Evaluation of the potential of pentachlorophenol degradation in soil by pulsed corona discharge plasma from soil characteristics. Environ. Sci. Technol. 2010, 44 (8), 31053110; DOI: 10.1021/es903527w. (10) Mills, A.; LeHunte, S. An overview of semiconductor photocatalysis. J. Photochem. Photobiol., A 1997, 108 (1), 135; DOI: 10.1016/S1010-6030(97)00118-4. (11) Hao, X. L.; Zhou, M. H.; Lei, L. C. Non-thermal plasma-induced photocatalytic degradation of 4-chlorophenol in water. J. Hazard. Mater. 2007, 141 (3), 475482; DOI: 10.1016/j.jhazmat.2006.07.012. (12) Lukes, P.; Clupek, M.; Sunka, P.; Peterka, F.; Sano, T.; Negishi, N.; Matsuzawa, S.; Takeuchi, K. Degradation of phenol by underwater pulsed corona discharge in combination with TiO2 photocatalysis. Res. Chem. Intermed. 2005, 31 (46), 285294; DOI: 10.1163/ 1568567053956734. (13) Maroulf-Khelifa, K.; Abdelmalek, F.; Khelifa, A.; Addou, A. TiO2-assisted degradation of a perfluorinated surfactant in aqueous solutions treated by gliding arc discharge. Chemosphere 2008, 70 (11), 19952001; DOI: 10.1016/j.chemosphere.2007.09.030. (14) Hoffmann, M. R.; Martin, S. T.; Choi, W. Y.; Bahnemann, D. W. Environmental applications of semiconductor photocatalysis. Chem. Rev. 1995, 95 (1), 6996; DOI: 10.1021/cr00033a004. (15) Tang, Q.; Lin, S.; Jiang, W. J.; Lim, T. M. Gas phase dielectric barrier discharge induced reactive species degradation of 2, 4-dinitrophenol. Chem. Eng. J. 2009, 153 (13), 94100; DOI: 10.1016/j. cej.2009.06.022. (16) Wang, K. H.; Hsieh, Y. H.; Chen, L. J. The heterogeneous photocatalytic degradation, intermediates and mineralization for the aqueous solution of cresols and nitrophenols. J. Hazard. Mater. 1998, 59 (23), 251260; DOI: 10.1016/S0304-3894(97)00151-9. (17) Zhang, J. L.; Xu, H. S.; Chen, H. J.; Anpo, M. Study on the formation of H2O2 on TiO2 photocatalysts and their activity for the photocatalytic degradation of X-GL dye. Res. Chem. Intermed. 2003, 29 (79), 839848;DOI: 10.1163/156856703322601843. (18) Sato, M.; Ohgiyama, T.; Clements, J. S. Formation of chemical species and their effects on microorganisms using a pulsed high-voltage discharge in water. IEEE Trans. Ind. Appl. 1996, 32 (1), 106112; DOI: 10.1109/28.485820. (19) Pichat, P.; Disdier, J.; Hoang-Van, C.; Mas, D.; Goutailler, G.; Gaysse, C. Purification/deoderization of indoor air and gaseous effluents by TiO2 photocatalysis. Catal. Today 2000, 63 (9), 363369; DOI: 10.1016/S0920-5861(00)00480-6. (20) Simek, M.; Clupek, M. Efficiency of ozone production by pulsed positive corona discharge in synthetic air. J. Phys. D: Appl. Phys. 2002, 35 (11), 11711175; DOI: 10.1088/0022-3727/35/11/311. (21) Ghezzar, M. R.; Abdelmalek, F.; Belhadj, M.; Benderdouche, N.; Addou, A. Gliding arc plasma assisted photocatalytic degradation of anthraquinonic acid green 25 in solution with TiO2. Appl. Catal., B 2007, 72 (34), 304313; DOI: 10.1016/j.apcatb.2006.11.008. (22) Oturan, M. A.; Peiroten, J.; Chartrin, P.; Acher, A. J. Complete destruction of p-nitrophenol in aqueous medium by electro-Fenton method. Environ. Sci. Technol. 2000, 34 (16), 34743479; DOI: 10.1021/es990901b. (23) Shi, H. X.; Xu, X. W.; Xu, X. H.; Wang, D. H.; Wang, Q. D. Mechanistic study of ozonation of p-nitrophenol in aqueous solution. J. Environ. Sci. 2005, 17 (6), 926–929. (24) Logemann, F. P.; Annee, J. H. J. Water treatment with a fixed bed catalytic ozonation process. Water Sci. Technol. 1997, 35 (4), 353360; DOI: 10.1016/S0273-1223(97)00045-0. (25) Sano, T.; Negishi, N.; Sakai, E.; Matsuzawa, S. Contributions of photocatalytic/catalytic activities of TiO2 and gamma-Al2O3 in 9306
dx.doi.org/10.1021/es2014314 |Environ. Sci. Technol. 2011, 45, 9301–9307
Environmental Science & Technology
ARTICLE
non-thermal plasma on oxidation of acetaldehyde and CO. J. Mol. Catal. A: Chem. 2006, 245 (12), 235241; DOI: 10.1016/j.molcata. 2005.10.002. (26) Chavadej, S.; Kiatubolpaiboon, W.; Rangsunvigit, P.; Sreethawong, T. A Combined multistage corona discharge and catalytic system for gaseous benzene removal. J. Mol. Catal. A: Chem. 2007, 263 (12), 128136; DOI: 10.1016/j.molcata.2006.08.061. (27) Liu, Y. J.; Wang, D. G.; Sun, B.; Zhu, X. M. Aqueous 4-nitrophenol decomposition and hydrogen peroxide formation induced by contact glow discharge electrolysis. J. Hazard. Mater. 2010, 181 (13), 10101015; DOI: 10.1061/(ASCE)WR.1943-5452.0000099. (28) Suarez, C.; Louys, F.; Gunther, K.; Eiben, K. OH-radical induced denitration of nitrophenols. Tetrahedron Lett. 1970, 11 (8), 575–578. (29) Di Paola, A.; Augugliaro, V.; Palmisano, L.; Pantaleo, G.; Savinov, E. Heterogeneous photocatalytic degradation of nitrophenols. J. Photochem. Photobiol., A 2003, 155 (13), 207214; DOI: 10.1016/ S1010-6030(02)00390-8. (30) Dai, Q. Z.; Lei, L. C.; Zhang, X. W. Enhanced degradation of organic wastewater containing p-nitrophenol by a novel wet electrocatalytic oxidation process: Parameter optimization and degradation mechanism. Sep. Purif. Technol. 2008, 61 (2), 123129; DOI: 10.1016/j. seppur.2007.10.006. (31) Hong, P. K. A.; Zeng, Y. Degradation of pentachlorophenol by ozonation and biodegradability of intermediates. Water Res. 2002, 36 (17), 42434254; DOI: 10.1016/S0043-1354(02)00144-6. (32) Benitez, F. J.; Acero, J. L.; Real, F. J.; Garcia, J. Kinetics of photodegradation and ozonation of pentachlorophenol. Chemosphere 2003, 51 (8), 651662; DOI: 10.1016/S0045-6535(03)00153-X. (33) Lukes, P.; Locke, B. R. Degradation of substituted phenols in a hybrid gas-liquid electrical discharge reactor. Ind. Eng. Chem. Res. 2005, 44 (9), 29212930; DOI: 10.1021/ie0491342.
9307
dx.doi.org/10.1021/es2014314 |Environ. Sci. Technol. 2011, 45, 9301–9307
ARTICLE pubs.acs.org/est
Influence of pH on the Formation of Sulfate and Hydroxyl Radicals in the UV/Peroxymonosulfate System Ying-Hong Guan,† Jun Ma,*,†,‡ Xu-Chun Li,† Jing-Yun Fang,† and Li-Wei Chen† † ‡
State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology, Harbin, People's Republic of China National Engineering Research Center of Urban Water Resources, Harbin Institute of Technology, People's Republic of China
bS Supporting Information ABSTRACT: The influence of pH on the degradation of refractory organics (benzoic acid, BA) in UV(254 nm)/Peroxymonosulfate (UV/PMS) system was investigated. The degradation of BA was significantly enhanced at the pH range of 811, which could not be explained only by the generally accepted theory that SO4•‑ was converted to HO• at higher pH. A hypothesis was proposed that the rate of PMS photolysis into HO• and SO4•‑ increased with pH. The hypothesis was evidenced by the measured increase of apparent-molar absorption coefficient of PMS (εPMS, 13.8149.5 M1 3 cm1) and photolysis rate of PMS with pH, and further proved by the increased quasi-stationary concentrations of both HO• and SO4•‑ at the pH range of 810. The formation of HO• and SO4•‑ in the UV/PMS system was confirmed mainly from the cooperation of the photolysis of PMS, the decay of peroxomonosulfate radical (SO5•‑) and the conversion of SO4•‑ to HO• by simulation and experimental results. Additionally, the apparent quantum yield for SO4•‑ in the UV/PMS system was calculated as 0.52 ( 0.01 at pH 7. The conclusions above as well as the general kinetic expressions given might provide some references for the UV/PMS applications.
’ INTRODUCTION Increasing attention has been paid to the sulfate radical (SO4•‑) due to its high efficiency of mineralization of organic pollutants and its efficient removal of halogen-substituted pollutants.1,2 SO4•‑ is a strong oxidant with a redox potential of 2.5 3.1 V,3 which is similar to hydroxyl radical (HO•) with a redox potential of 1.82.7 V.4 Peroxodisulfate (PDS) activated by UV, heat, base, or transition metals is commonly used to generate SO4•‑ and has been widely studied.58 The activation of peroxymonosulfate (Oxone, PMS) is also an efficient source of SO4•‑. Recently, many studies related to the activation of PMS have focused on transition metals, among which Co(II) was found to be the best activator.8 But the adverse effects of Co(II) on human health need to be considered. Supported cobalt catalysts were used as heterogeneous activators of PMS to reduce the concentration of free cobalt ion.9 Meanwhile, PMS activated by iron or UV irradiation was considered as an environmentally friendly and applicable technology.10,11 PMS irradiated by UV was proposed to generate HO• and SO4•‑ through the cleavage of the peroxo bond.12 The production of SO4•‑ from PMS irradiated by a pulsed laser at λ = 248 nm was verified by its characteristic absorption at λ = 445 nm.13 The molar absorption coefficients of PMS at λ = 248 nm and at λ = 254 nm available in the form of HSO5 were 19.11 M1 3 cm1 and 12.5 or 14 M1 3 cm1, respectively.10,13 It was rational to expect that PMS irradiated by a low pressure Hg lamp (λ = 254 nm), could result in the production of SO4•‑ and HO•. r 2011 American Chemical Society
SO4•‑ reacted with HO to form HO• with a rate constant of 6.5 107 M1 3 s1 at alkaline pH.14 It was reported that the degradation of nitrobenzene (NB) was enhanced as pH increased from 7 to 12 in the thermally activated PDS system, which was due to the conversion of SO4•‑ to HO•.15 The conversion also contributed to the increased rate of butylated hydroxyanisole decay as pH increased from 3 to 11 in the UV/ PDS system.5 Meanwhile, the degradation efficiencies of acetic acid and iopromide decreased at the pH range of 911 in the UV/PDS system, due to the conversion of SO4•‑ to HO•.1,16 Hence, controlling pH could be considered as one approach to manipulate the degradation of pollutant in the SO4•‑-existing system. However, the mechanism on the pH-dependent degradation of pollutant was rarely studied in UV/PMS system. The aim of this study was to investigate the formation of HO• and SO4•‑ versus pH, which was the fundamental reason for the variation of pollutant degradation rate versus pH, and to present the main factors affecting the formation of HO• and SO4•‑. Benzoic acid (BA) and nitrobenzene (NB) were selected as the probe compounds to investigate the formation of HO• and SO4•‑ versus pH in the UV/PMS system. Received: May 21, 2011 Accepted: September 23, 2011 Revised: September 15, 2011 Published: October 14, 2011 9308
dx.doi.org/10.1021/es2017363 | Environ. Sci. Technol. 2011, 45, 9308–9314
Environmental Science & Technology
ARTICLE
’ EXPERIMENTAL SECTION Materials. Potassium peroxymonosulfate (2KHSO5 3 KHSO4 3 K2SO4 available as Oxone, PMS), potassium peroxodisulfate (PDS), benzoic acid, nitrobenzene, sodium phosphate monobasic monohydrate, sodium phosphate dibasic, sodium tetraborate decahydrate, boric acid, and 5,5-dimethyl-1-pyrrolidine Noxide (DMPO) were of ACS reagent grade and purchased from Sigma-Aldrich, Inc. Phosphoric acid and methanol of HLPC grade were purchased from Dima-Tech Inc. and Thermo Fisher Scientific Inc., respectively. Tert-butyl-alcohol (TBA) was of guaranteed reagent grade and purchased from Tianjin Chemical Reagent Co., Ltd., China. Catalase (from bovine liver) was purchased from Tokyo Kasei Kogyo Co., Ltd. Other reagents used were of analytical-reagent grade and purchased from Sinopharm Chemical Reagent Co., Ltd., China. All of the chemicals were used as received without further purification. All solutions were prepared in 18.2 MΩ 3 cm Milli-Q-water produced on a Milli-Q Biocel water system. Experimental Procedure. All of the photochemical experiments were performed in a 1.3 L cylindrical borosilicate glass vessel (1.2 L samples, 3.5 cm path length). The optical path length (b) was determined to be 4.30 ( 0.05 cm by measuring the photolysis rate of H2O2.17 A low-pressure mercury UV lamp (Heraeus, GPH 135T5 L/4, 6 W nameplate output at 254 nm) was placed in the center of the cylindrical vessel axially along the length of the reactor. The incident radiation intensity of UV lamp (I0) was about 1.5 106 Einstein 3 s1 measured by the method of iodide-iodate chemical actinometer.18 Phosphate/borate buffer with the concentration of 2 mM was used to adjust pH value from 6 to 12 and the pH value during the reaction was measured to be in the range of predetermined pH value (0.1. Samples were withdrawn at predetermined time intervals and quenched using excessive sodium nitrite. Most experiments were conducted in triplicate, at ambient temperature (20 ( 2 °C). Error bars were based on the results of replicate experiments. Details concerning the experimental procedure are provided in Text S1 of the Supporting Information, SI. Analytical Method. The concentrations of BA and NB were determined by high performance liquid chromatography 1525
RQ SC ¼
equipped with a Waters 717 autosampler and a Waters 2487 dual λ detector. Separate column used was a Waters symmetry C18 column (4.6 mm 150 mm, 5 μm particle size). An eluent of water (pH 3, adjusted by phosphoric acid) and methanol (55:45, v/v %) was used to separate BA and its products at a flow rate of 1.0 mL/min. The concentration of BA was quantified at λ = 227 nm. The concentration of NB was analyzed at λ = 263 nm with an eluent of water and methanol (55:45, v/v %). The pH value was measured by a pH meter (PHS-3C, Shanghai Precision & Scientific instrument Co., Ltd.). The solution of PMS was prepared as needed and standardized using iodometric titration.19 The absorption spectra of PMS solution at different pH values were carried out on a Varian Cary 300 spectrometer. EPR experiments were performed on a Bruker A200 spectrometer with DMPO as a spin-trapping agent. Detailed parameters can be seen in SI Text S2. Kinetic Rate Expressions. Table 1 summarizes the photochemical and chemical reactions in the UV/PMS system along with their rate constants. On the basis of the reactions in Table 1, radicals consumed by radical collision induced by HO• and SO4•‑ could be neglected in the UV/PMS system in the presence of about 10 μM BA. When 10 mM TBA was added, less than 1% of HO• reacted with BA. Hence, the degradation of BA by HO• was negligible. The kinetic expression of BA degradation in the UV/ PMS system with the addition of TBA could be expressed as eq 1, based on the pseudosteady state assumption:
dcBA ¼ k12 ½SO• 4 SS cBA þ kd, BA cBA dt ¼ kS0, BA cBA þ kd, BA cBA
ð1Þ
where [SO4•‑]ss is defined as the quasi-stationary concentration of SO4•‑, kS0,BA is defined as k12[SO4•‑]ss, kd,BA is the first-order rate constant of the direct photolysis of BA. Assuming that HO• and SO4•‑ were formed from the photolysis of PMS, the decay of peroxomonosulfate radical (SO5•‑) and the conversion of SO4•‑ to HO•, the relative quasi-stationary concentration of SO4•‑ (RQSC) could be derived as eq 2 (details can be seen in SI Text S3).
½SO• 1 10A 10pH 10pKa1 4 SS ¼ cPMS b εHSO5 þ pKa1 ε 2 ϕI0 =V A 10pKa1 þ 10pH 10 þ 10pH SO5
!
! 11 10pH 10pKa1 k2 pKa1 þ k cPMS 3 6 10 þ 10pH 10pKa1 þ 10pH ! 1 10pH 10pKa1 pH 14 k2 pKa1 cPMS þ k16 cTBA Þ þ k12 cBA þ k17 cTBA þ k5 10 þ k3 pKa1 6 10 þ 10pH 10 þ 10pH
k11 cBA þ k16 cTBA þ
ðk12 cBA þ k17 cTBA þ k5 10pH 14 Þðk11 cBA
ð2Þ The variations of kS0,BA and RQSC (based on eqs 1 and 2) with pH were compared to testify the hypothesis that HO• and SO4•‑ were formed from the cooperation of the photolysis of PMS, the decay of SO5•‑ and the conversion of SO4•‑ to HO•. In the UV/PMS system, NB and BA were used simultaneously as probe compounds to indicate the variations of the formation rates of HO• and SO4•‑ (FHO 3 and FSO43 ‑) with pH. FHO 3 and FSO43 ‑ were defined as eqs 3 and 4. On the basis of the pseudosteady
state assumption that the quantity of radicals formed was equal to the quantity of radicals consumed, the sum of FHO 3 and FSO43 ‑ (Ftotal) in the UV/PMS system could be derived as eq 5 (details can be seen in SI Text S4).
9309
FHO• ¼ PPMS þ QHO•
ð3Þ
FSO4 • ¼ PPMS þ TSO4 • QSO4 •
ð4Þ
dx.doi.org/10.1021/es2017363 |Environ. Sci. Technol. 2011, 45, 9308–9314
Environmental Science & Technology
ARTICLE
Table 1. Principal Reactions in the UV/PMS System no.
references
1
HSO5 =SO25 sf SO•4 þ HO•
r = ϕI0bεPMScPMS(1 10A)/(AV)a
20
2
HO• þ HSO5 f SO•5 þ H2 O
k2 = 1.7 107
21
k3 = 2.1 10
21
hv
•
3
HO þ
4
SO•4
5
SO•4
SO25
þ
HO þ
f
HSO5
þ HO
•
6
a
rate constants (M1 3 s1)
reactions
HSO4
•
SO•5 f
þ HO
SO•5 •
þ
f HO þ SO•4
f
HSO4
9
k4 < 10
5
SO24
21
k5 = 6.5 10
þ H2 O
•
7
14
k6 = 6.9 105
4
7
HO þ HO f H2 O2
k7 = 5.5 109
4
8
SO•4 þ SO•4 f S2 O28
k8 = 3.1 10
3
•
SO•4
9
HO þ
10a
SO•5
10b
SO•5
11
in the presence of BA HO• þ C6 H5 COO f product
k11 = 5.9 109
4
12
SO•4
k12 = 1.2 10
3
k13 ≈ 4 107
4
•
f
HSO5
8
þ
SO•5
þ
SO•5
f
SO•4
f
S2 O28
þ C6 H5 COO þ C6 H5 COO
k9 = 1 10
10
þ
SO•4
þ O2
k10 = 1 10
22
8
23
þ O2
f product
9
f product
13
O
14
in the presence of NB HO• þ C6 H5 NO2 f product
k14 = 3.9 109
4
15
SO•4
k15 e 106
3
16
in the presence of TBA HO• þ ðCH3 Þ3 COH f product
k16 = 6.0 108
4
17
SO•4
5
k17 = 4.0 10
3
18
in the presence of methanol HO• þ CH3 OH f product
k18 = 9.7 108
4
19
SO•4
k19 = 3.2 10
3
20
HSO5
pKa1 = 9.4
24
pKa2 = 11.9
4
pKa3 = 4.2
25
•
þ C6 H5 NO2 f product
þ ðCH3 Þ3 COH f product
þ CH3 OH f product S
SO25
•
þ H
þ H
6
þ
þ
21
HO S O
22
C6 H5 COOH S C6 H5 COO þ H þ
A is the absorbance of solution.
Ftotal ¼ FHO• þ FSO4 • ¼ RBA þ RNB ! 10pH 10pKa1 k2 þ pKa1 k3 cPMS 10pKa1 þ 10pH 10 þ 10pH
þ
k14 cNB
RNB
ð5Þ PPMS ¼
þ
’ RESULTS AND DISCUSSION
1 1 RBA þ RNB 2 2
1 12
where FHO 3 and FSO43 ‑ are the formation rates of HO• and SO4•‑, PPMS is the rate of PMS photolysis into HO• and SO4•‑, QHO 3 and QSO43 ‑ are the production rate of HO• and the consumption rate of SO4•‑ through the conversion of SO4•‑ to HO•, TSO43 ‑ is the production rate of SO4•‑ from the decay of SO5•‑, Ftotal is the sum of the formation rates of HO• and SO4•‑, RNB and RBA are the consumption rates of NB and BA.
! 10pH 10pKa1 k2 þ pKa1 k3 cPMS 10pKa1 þ 10pH 10 þ 10pH k14 cNB
RNB
ð6Þ
Effect of pH on BA Degradation. BA was used as a recalcitrant organic compound to investigate the decontamination effect in the UV/PMS system at the pH range of 612. The degradation of BA was found to be enhanced significantly as pH increased from 8 to 11 (Figure 1a). Pseudofirst-order kinetic model fitted very well to the degradation of BA within the initial 9310
dx.doi.org/10.1021/es2017363 |Environ. Sci. Technol. 2011, 45, 9308–9314
Environmental Science & Technology
ARTICLE
Figure 2. The εPMS, speciation of PMS, and decomposition of PMS in the UV/PMS system in the presence of TBA at different pH (condition: [PMS] = 1 mM as 1/2 Oxone; [TBA] = 100 mM; irradiation time: 10 min; error bar represents a confidence interval with a confidence of 0.95).
Figure 1. (a) Degradation of BA in the UV/PMS system at different pH. Inset indicates kinetics of BA degradation versus pH. (b) Pseudofirst-order rate constants versus pH in the UV/PMS system. (conditions: [BA] = 9.90 μM; [PMS] = 100 μM as 1/2 Oxone; error bar represents a confidence interval with a confidence of 0.95).
1.5 min (Inset of Figure 1a), though the BA degradation in the UV/PMS system was complex and might not follow a pseudofirst-order kinetic model for longer experimental time in theory. By fitting each pH series for pseudofirst-order kinetics, pseudofirst-order rate constants (k0,BA) were obtained and plotted as the function of pH (Figure 1b). k0,BA kept almost invariant at the pH range from 6 to 8, increased significantly from 8 to 11, and dropped from 11 to 12. No obvious degradation of BA by PMS was observed in the investigated time scale. The degradation of BA by direct photolysis of UV, with kd,BA of (5.80 ( 0.31) 105 s1, could be negligible compared with that achieved by UV/PMS (SI Figure S3). This indicated that UV and PMS should have a synergistic effect on the degradation of BA. PMS was reported to produce HO• and SO4•‑ by pulsed laser irradiation at λ = 248 nm.13 The formation of HO• and SO4•‑ was also checked in the present system (λ = 254 nm). Radical-scavenging (TBA and methanol) experimental results and EPR spectra (SI Figures S3 and S4) indicated that both HO• and SO4•‑ contributed to BA degradation (details can be seen in SI Text S5). SO4•‑ could react with HO to form HO• at higher pH.14 In the present study, both HO• and SO4•‑ were mainly consumed by BA at neutral pH and HO• was also consumed by PMS forming SO5•‑ at basic pH. Meanwhile, SO5•‑ was reported to partly decay into SO4•‑.23 However, the conversion of SO4•‑ to HO• and the decay of SO5•‑ to SO4•‑ would not result in the increased efficiency of BA degradation as pH increased, which was not in accordance with the experimental results (Figure 1a,b) that a sharp increase was observed at the pH range of 811. Apparent-Molar Absorption Coefficient (εPMS) and Speciation of PMS versus pH. The absorption spectra curve of PMS shift to right as pH increased from 8 to 11 (SI Figure S5). The apparent-molar absorption coefficient (εPMS) was calculated
from the absorbance of PMS at λ = 254 nm and shown in Figure 2. The εPMS increased with pH from 13.8 to 149.5 M1 3 cm1 in the pH range of 612. It was similar to the variation of εH2O2 that increased from 19.6 to 229 M1 3 cm1 as H2O2 dissociated into HO2.17 The εPMS of 13.8 M1 3 cm1 (existing in HSO5 form) obtained in this study was also in accordance with the value of 14 M1 3 cm1 reported previously.13 The parent acid of PMS, H2SO5, was reported to be a strong acid as sulfuric acid.26 Hence, the speciation of PMS at studied pH range was calculated based on its second pKa (Table 1) and shown in Figure 2. PMS mainly exists in its monoanion form (HSO5) at the pH range of 68 and its dianion form (SO52‑) at pH g 11. The εPMS correlated closely with the speciation of PMS. Assuming the quantum yield was constant for dissociated and undissociated species, it was reasonable to infer that the photolysis rate of PMS should increase significantly with pH near its second pKa and the increased photolysis rate of PMS into HO• and SO4•‑ would contribute to the sharply enhanced degradation of BA at pH around 9.4. Decomposition of PMS versus pH. HO• was reported to induce the acceleration of PMS decomposition.21 Thus, the decomposition of PMS in the UV/PMS system might consist of three parts: the photolysis by UV irradiation, decomposition by HO• or other radicals attack, and spontaneous decomposition. PMS was unstable and decomposed to H2O2 at basic pH.7 Catalase was used to quench H2O2 if produced.27 The results showed that no observable quantity of H2O2 was produced and the spontaneous decomposition of PMS was found to be negligible (e3%) under the given condition (details can be seen in SI Text S6). TBA was reported to be a good radical scavenger to reduce the decomposition of H2O2 by radical attack.28 Also, TBA was selected to reduce the decomposition of PMS by radical attack rather than methanol in the present study (see SI Text S6 for details). Therefore, the decomposition of PMS in the UV/ PMS system with the addition of TBA was used to indicate the photolysis of PMS, although slight decomposition of PMS by radical attack might exist. Figure 2 shows that C/C0,PMS decreased as pH increased and a sharp decrease was observed at pH around 9.4. This indicated that a sharp increase of photolysis of PMS was obtained at pH around its second pKa, which strengthened the inference above. Quasi-stationary concentration of SO4•‑ (QSCSO43 ‑) versus pH. In order to investigate the variation of QSCSO43 ‑ with 9311
dx.doi.org/10.1021/es2017363 |Environ. Sci. Technol. 2011, 45, 9308–9314
Environmental Science & Technology
Figure 3. (a) kS0,BA and RQSC versus pH in the UV/PMS system with the addition of TBA (conditions: [BA] = 9.90 μM; [PMS ] = 100 μM as 1/2 Oxone; [TBA] = 10 mM; error bar represents a confidence interval with a confidence of 0.95). (b) k0,NB versus pH in the UV/PMS system (conditions: [NB] = 18.07 μM; [BA] = 9.90 μM; [PMS] = 100 μM as 1/2 Oxone; error bar represents a confidence interval with a confidence of 0.95).
pH, kS0,BA of BA degradation in the presence of TBA was obtained based on eq 1 by fitting each pH series for pseudofirst-order kinetics (SI Figure S7). Figure 3a shows that kS0,BA was almost invariant at the pH range of 68, increased sharply from 8 to 10, and decreased obviously from 10 to 12. Hence, QSCSO43 ‑ varied in the same way as kS0,BA did over the corresponding pH range. Quasi-stationary Concentration of HO• (QSCHO 3 ) versus pH. Reaction between SO4•‑ and NB was quite slow (Table 1), which was also proven in SI Figure S8 (detail can be seen in SI Text S7). The first-order rate constant of direct photolysis of NB (kd, 5 1 s and the degradation of NB NB) was (1.8531 ( 0.0992) 10 by UV could be negligible compared with that achieved by UV/ PMS. NB was thus selected as the probe compound to indicate the variation of QSCHO 3 with pH in the UV/PMS system in the presence of BA. Figure 3b shows that the pseudofirst-order rate constant (k0,NB) of NB degradation was almost invariant at the pH range of 68 and increased with pH from 8 to 12. The dissociation of HO• into O•‑ became obvious at pH g 11, which would result in the change of apparent second-order rate constant (the weighted average of the rate constants of HO• and O•‑ with NB). Assuming that the rate constant of the reaction between NB and O•‑ was no more than that of HO• and NB,4 the conclusion could be derived that the QSCHO 3 (the sum of quasi-stationary concentrations of HO• and O•‑) kept almost invariant at the pH range of 68 and increased with pH from 8 to 12. Role of the Photolysis of PMS, the Decay of SO5•‑ and the Conversion of SO4•‑ to HO• in the Formation of HO• and SO4•‑. The almost invariant QSCHO 3 and QSCSO43 ‑ with pH increasing from 6 to 8, were in accordance with the speciation of PMS, the variation of εPMS, and the almost unchanged C/C0, PMS at the same pH range (Figure 2). The increase of both QSCHO 3 and QSCSO43 ‑ with pH from 8 to 10 confirmed the
ARTICLE
conjecture that the rate of PMS photolysis into HO• and SO4•‑ increased with pH around its second pKa. Theoretically, the production of SO4•‑ from the photolysis of PMS should keep increasing with pH from 10 to 11 and remain stable at pH > 11 according to the variation of εPMS. Besides, the decay of SO5•‑ to SO4•‑ would also contribute to the increase of QSCSO43 ‑. However, QSCSO43 ‑ decreased as pH increased from 10 to 12 (Figure 3a). It might result from the conversion of SO4•‑ to HO•, which became apparent at pH > 9.3 (more than 10% SO4•‑ converted to HO•) when the initial concentration of BA was 10 μM. RQSC was then deduced from eq 2 and plotted in Figure 3a, based on a simplified hypothesis that HO• and SO4•‑ were formed mainly from the photolysis of PMS, the decay of SO5•‑ and the conversion of SO4•‑ to HO•. The simulation results of RQSC showed the same trends with the variation of kS0,BA as pH varied. But the ratios of maximum value to minimum value of kS0,BA and simulated RQSC were not consistent. It might be due to the omission of the competition of intermediate products for radicals in the simulation model for simplification. This would lead to a higher estimated value of RQSC than the actual value and the difference was especially obvious around pH 10 where BA was degraded quickly. Besides, some intermediate radical products such as superoxide radical and semiquinone might promote the decomposition of PMS and the production of radicals by one electron transfer,3,2932 which made the reactions in the studied system more complicated. However, the simplified simulation results of RQSC showed the same trends with the experimental results as pH changed, which confirmed that the formation of HO• and SO4•‑ was mainly due to the photolysis of PMS, the decay of SO5•‑ and the conversion of SO4•‑ to HO• in the UV/PMS system. It could also be obtained from Figure 3a that the apparent quantum yield of SO4•‑ (ϕSO43 ‑) at pH 7 was 0.52 ( 0.01 in the present system (λ = 254 nm). The apparent quantum yield for both HO• and SO4•‑ was estimated to be 1.04 based on the assumption that HO• and SO4•‑ were produced equally by the photolysis of PMS, which was close to the value of 1.0 for HO• from UV/H2O2 at λ = 254 nm.17 The ϕSO43 ‑ from UV/PMS previously reported was available at λ = 248 nm and it was reported to be 0.12.13 The large difference of ϕSO43 ‑ in the UV/ PMS system between the present and previous study might be due to the SO4•‑ sink through the conversion of SO4•‑ to HO•, which was taken into consideration in the present study but omitted in the previous study. Furthermore, QSCHO 3 and QSCSO43 ‑ were calculated to be in the magnitude of 1013 1012 M for SO4•‑ and 10141013 M for HO•, which indicated that the omission of the radical combination reactions in the UV/PMS system in the presence of BA was reasonable (details see SI Text S8). PPMS versus εPMS. On the basis of the hypothesis that the formation of HO• and SO4•‑ was mainly due to the photolysis of PMS, the decay of SO5•‑ and the conversion of SO4•‑ to HO•, the rate of PMS photolysis into HO• and SO4•‑ (PPMS) could be derived as eq 6. Then, the calculation results of PPMS would change in proportion to εPMS in the UV/PMS system (εNB = 6200 M1 3 cm1 and εBA = 760 M1 3 cm1 in the form of benzoate ion). As shown in Figure 4, PPMS was in direct proportion to εPMS with R2 = 0.9992, which further strengthened the hypothesis that the formation of HO• and SO4•‑ mainly resulted from the photolysis of PMS, the decay of SO5•‑ and the conversion of SO4•‑ to HO• in the UV/PMS system. Furthermore, the ϕSO43 ‑ from the photolysis of PMS was estimated to be 9312
dx.doi.org/10.1021/es2017363 |Environ. Sci. Technol. 2011, 45, 9308–9314
Environmental Science & Technology
ARTICLE dcSO• 5 ¼ dt
! 10pH 10pKa1 k2 þ pKa1 k3 cHO• cPMS 2k10 c2SO• 5 10pKa1 þ 10pH 10 þ 10pH
ð9Þ ðεPMS cPMS þ ∑ εi ci Þb
PPMS
Figure 4. PPMS and k0,BA versus εPMS in the UV/PMS system (error bar represents a confidence interval with a confidence of 0.95).
0.35 from the slope of the fitted line in Figure 4, which was smaller than the value of 0.52 ( 0.01 for ϕSO43 ‑ obtained above at pH 7. The faster decomposition of PMS and the more quantity of intermediates produced at higher pH might lead to the lower value of ϕSO43 ‑ obtained by the fitted line in Figure 4. k0,BA versus εPMS. On the basis of the pseudosteady state assumption, k0,BA should increase in proportion to εPMS, since absorbed quanta of irradiation could be simplified to be proportional to εPMS (the maximum absorbance of solution was about 0.06 and the maximum error induced by simplification was less than 8%). But k0,BA did not increase as assumed. As shown in Figure 4, the increase of k0,BA slowed down obviously as εPMS g 118 M1 3 cm1 (Figure 4). It might be due to the increased consumption of HO• by PMS as εPMS increased because the rate constant of the reaction between HO• and PMS was reported to increase as HSO5 dissociated to SO52‑ (Table 1). Meanwhile, the decay of SO5•‑ to SO4•‑ might contributed to the increase of k0,BA. Considering the two factors simultaneously, the termination reaction of SO5•‑ (reaction 10b) might be the main reason for the nonproportional increase of k0,BA with εPMS.33 Besides, the lower rate constant of the reaction between O•‑ and BA (Table 1) was an important reason for the significant decrease of k0,BA when εPMS was more than 146 M1 3 cm1 (pH g 11). The formation of HO• and SO4•‑ in the UV/PMS system was mainly due to the photolysis of PMS, the decay of SO5•‑ and the conversion of SO4•‑ to HO•. However, the latter was actually influenced by probe compounds (BA in the present system). Thus, it was necessary to give general kinetic expressions (eqs 710) to depict the variation of the concentrations of HO• and SO4•‑ under neutral or basic pH (pH e 11) in the application to natural water. The general kinetic expressions were examined by simulation results of the pseudofirst-order rate constant of BA degradation (k0,BA) in the UV/PMS system. The simulation result were basically coincident with the experimental results as pH changed as shown in SI Figure S11. dcHO• ¼ PPMS þ k5 cSO• 10pH 14 4 dt
∑i ki, HO ci cHO •
•
dcSO• 10 4 ¼ PPMS þ k10 c2SO• k5 cSO• 10pH 14 4 5 6 dt ki, SO• ci cSO• 4 4
∑i
i
∑i εi ci Þ
Þ
ð10Þ
where PPMS is the rate of PMS photolysis into HO• and SO4•‑, I0 is the incident radiation intensity, ϕ is the apparent quantum yield of PMS photolysis into HO• and SO4•‑ (0.435 is used in the simulation of BA degradation for it is the average of 0.52 and 0.35 obtained in this study), ci is the concentration of probe compound i, εi is the molar absorption coefficient of probe compound i, ki,HO 3 and ki,SO43 ‑ are the second-order rate constants of probe compound i with HO• and SO4•‑. In summary, the formation of HO• and SO4•‑ in the UV/PMS system was increased with pH at the pH range of 810. It was quite different from the increased formation of HO• and the decreased formation of SO4•‑ with the increase of pH at the corresponding pH range in the UV/PDS system. Therefore, PMS could be more suitable for the application to enhance the degradation of organic matter under basic condition. Comparing with H2O2, PMS is easier to be transported and more effective in the degradation of some kinds of organics,2 although not as environmentally friendly as H2O2 and a bit more expensive than H2O2.33 The conclusions derived from the present study, as well as the general kinetic expressions (eqs 710) would provide some references for practical applications.
’ ASSOCIATED CONTENT
bS Supporting Information. Additional texts and figures. This material is available free of charge via the Internet at http:// pubs.acs.org. ’ AUTHOR INFORMATION Corresponding Author
*Phone: 86-451-86283010; fax: 86-451-86282292; e-mail: majun@ hit.edu.cn.
’ ACKNOWLEDGMENT The support from the Natural Science Foundation of China (No. 50821002), the 863 high tech. scheme (No. 2009AA06Z310), the special S&T project on water treatment and control of pollution (2009ZX07424-005, 2009ZX07424006), and SKLUWRE (No. 2010DX10) are greatly appreciated. ’ REFERENCES
!
10pH 10pKa1 k2 þ pKa1 k3 cHO• cPMS 10pKa1 þ 10pH 10 þ 10pH
I0 ϕεPMS cPMS ð1 10 ¼ V ðεPMS cPMS þ
ð7Þ
ð8Þ
(1) Criquet, J.; Leitner, N. K. V. Degradation of acetic acid with sulfate radical generated by persulfate ions photolysis. Chemosphere 2009, 77 (2), 194–200, DOI: 10.1016/j.chemosphere.2009.07.040. (2) Hori, H.; Yamamoto, A.; Hayakawa, E.; Taniyasu, S.; Yamashita, N.; Kutsuna, S. Efficient decomposition of environmentally persistent perfluorocarboxylic acids by use of persulfate as a photochemical oxidant. Environ. Sci. Technol. 2005, 39 (7), 2383–2388, DOI: 10.1021/es0484754. (3) Neta, P.; Huie, R. E.; Ross, A. B. Rate constants for reactions of inorganic radicals in aqueous solution. J. Phys. Chem. Ref. Data 1988, 17 (3), 1027–1284, DOI: 10.1063/1.555808. 9313
dx.doi.org/10.1021/es2017363 |Environ. Sci. Technol. 2011, 45, 9308–9314
Environmental Science & Technology (4) Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (•OH/•O) in aqueous solution. J. Phys. Chem. Ref. Data 1988, 17 (2), 513–886, DOI: 10.1063/1.555805. (5) Lau, T. K.; Chu, W.; Graham, N. J. D. The aqueous degradation of butylated hydroxyanisole by UV/S2O82‑: study of reaction mechanisms via dimerization and mineralization. Environ. Sci. Technol. 2007, 41 (2), 613–619, DOI: 10.1021/es061385a. (6) Waldemer, R. H.; Tratnyek, P. G.; Johnson, R. L.; Nurmi, J. T. Oxidation of chlorinated ethenes by heat-activated persulfate: kinetics and products. Environ. Sci. Technol. 2007, 41 (3), 1010–1015, DOI: 10.1021/es062237m. (7) Furman, O. S.; Teel, A. L.; Watts, R. J. Mechanism of base activation of persulfate. Environ. Sci. Technol. 2010, 44 (16), 6423–6428, DOI: 10.1021/es1013714. (8) Anipsitakis, G. P.; Dionysiou, D. D. Radical generation by the interaction of transition metals with common oxidants. Environ. Sci. Technol. 2004, 38 (13), 3705–3712, DOI: 10.1021/es035121o. (9) Yang, Q. J.; Choi, H.; Chen, Y. J.; Dionysiou, D. D. Heterogeneous activation of peroxymonosulfate by supported cobalt catalysts for the degradation of 2,4-dichlorophenol in water: The effect of support, cobalt precursor, and UV radiation. Appl. Catal., B 2008, 77 (34), 300–307, DOI: 10.1016/j.apcatb.2007.07.020. (10) Rivas, J.; Gimeno, O.; Borralho, T.; Carbajo, M. UV-C photolysis of endocrine disruptors. The influence of inorganic peroxides. J. Hazard. Mater. 2010, 174 (13), 393–397, DOI: 10.1016/j. jhazmat.2009.09.065. (11) Rastogi, A.; Al-Abed, S. R.; Dionysiou, D. D. Effect of inorganic, synthetic and naturally occurring chelating agents on Fe(II) mediated advanced oxidation of chlorophenols. Water Res. 2009, 43 (3), 684–694, DOI: 10.1016/j.watres.2008.10.045. (12) Anipsitakis, G. P.; Dionysiou, D. D. Transition metal/UVbased advanced oxidation technologies for water decontamination. Appl. Catal., B 2004, 54 (3), 155–163, DOI: 10.1016/j.apcatb.2004.05.025. (13) Herrmann, H. On the photolysis of simple anions and neutral molecules as sources of O/OH, SOx and Cl in aqueous solution. Phys. Chem. Chem. Phys. 2007, 9 (30), 3935–3964, DOI: 10.1039/b618565g. (14) Hayon, E.; Treinin, A.; Wilf, J. Electronic spectra, photochemistry, and autoxidation mechanism of the sulfite-bisulfite-pyrosulfite systems. SO2, SO3, SO4, and SO5 radicals. J. Am. Chem. Soc. 1972, 94 (1), 47–57, DOI: 10.1021/ja00756a009. (15) Liang, C. J.; Su, H. W. Identification of sulfate and hydroxyl radicals in thermally activated persulfate. Ind. Eng. Chem. Res. 2009, 48 (11), 5558–5562, DOI: 10.1021/Ie9002848. (16) Chan, T. W.; Graham, N. J. D.; Chu, W. Degradation of iopromide by combined UV irradiation and peroxydisulfate. J. Hazard. Mater. 2010, 181 (13), 508–513, DOI: 10.1016/j.jhazmat.2010.05.043. (17) Baxendale, J. H.; Wilson, J. A. The photolysis of hydrogen peroxide at high light intensities. Trans. Faraday Soc. 1957, 53, 344–356, DOI: 10.1039/TF9575300344. (18) Rahn, R. O.; Stefan, M. I.; Bolton, J. R.; Goren, E.; Shaw, P. S.; Lykke, K. R. Quantum yield of the iodide-iodate chemical actinometer: dependence on wavelength and concentration. Photochem. Photobiol. 2003, 78 (2), 146–152, DOI: 10.1562/0031-8655(2003). (19) Ball, R. E.; Edwards, J. O.; Haggett, M. L.; Jones, P. A kinetic and isotopic study of the decomposition of monoperoxyphthalic acid. J. Am. Chem. Soc. 1967, 89 (10), 2331–2333, DOI: 10.1021/ja00986a015. (20) Crittenden, J. C.; Hu, S. M.; Hand, D. W.; Green, S. A. A kinetic model for H2O2/UV process in a completely mixed batch reactor. Water Res. 1999, 33 (10), 2315–2328, DOI: 10.1016/S0043-1354(98) 00448-5. (21) Maruthamuthu, P.; Neta, P. Radiolytic chain decomposition of peroxomonophosphoric and peroxomonosulfuric acids. J. Phys. Chem. 1977, 81 (10), 937–940, DOI: 10.1021/j100525a001. (22) Klaning, U. K.; Sehested, K.; Appelman, E. H. Laser flash photolysis and pulse radiolysis of aqueous solutions of the fluoroxysulfate ion, SO4F. Inorg. Chem. 1991, 30 (18), 3582–3584, DOI: 10.1021/ ic00018a040.
ARTICLE
(23) Huie, R. E.; Clifton, C. L.; Altstein, N. A pulse radiolysis and flash photolysis study of the radicals SO2, SO3, SO4, and SO5. Int. J. Radiat. Appl. Instrum. C Radiat. Phys. Chem. 1989, 33 (4), 361–370, DOI: 10.1016/1359-0197(89)90034-9. (24) Rani, S. K.; Easwaramoorthy, D.; Bilal, I. M.; Palanichamy, M. Studies on Mn(II)-catalyzed oxidation of alpha-amino acids by peroxomonosulphate in alkaline medium-deamination and decarboxylation: a kinetic approach. Appl. Catal., A 2009, 369 (12), 1–7, DOI: 10.1016/j. apcata.2009.07.048. (25) Tao, L.; Han, J.; Tao, F. M. Correlations and predictions of carboxylic acid pka values using intermolecular structure and properties of hydrogen-bonded complexes. J. Phys. Chem. A 2008, 112 (4), 775–782, DOI: 10.1021/jp710291c. (26) Lawrance, G. A.; Ward, C. B. Kinetics of oxidation of manganese(II) by peroxomonosulfuric acid in aqueous acidic solution. Transition Met. Chem. 1985, 10 (7), 258–261, DOI: 10.1007/ bf00621082. (27) Liu, W. J.; Andrews, S. A.; Stefan, M. I.; Bolton, J. R. Optimal methods for quenching H2O2 residuals prior to UFC testing. Water Res. 2003, 37 (15), 3697–3703, DOI: 10.1016/s0043-1354(03)00264-1. (28) Popov, E.; Mametkuliyev, M.; Santoro, D.; Liberti, L.; Eloranta, J. Kinetics of UV-H2O2 advanced oxidation in the presence of alcohols: The role of carbon centered radicals. Environ. Sci. Technol. 2010, 44 (20), 7827–7832, DOI: 10.1021/es101959y. (29) Peiro, A. M.; Ayllon, J. A.; Peral, J.; Domenech, X. TIO2photocatalyzed degradation of phenol and ortho-substituted phenolic compounds. Appl. Catal., B 2001, 30 (34), 359–373, DOI: 10.1016/ S0926-3373(00)00248-4. (30) Anipsitakis, G. P.; Dionysiou, D. D.; Gonzalez, M. A. Cobaltmediated activation of peroxymonosulfate and sulfate radical attack on phenolic compounds. Implications of chloride ions. Environ. Sci. Technol. 2006, 40 (3), 1000–1007, DOI: 10.1021/Es050634b. (31) Weinstein, J.; Bielski, B. H. J. Kinetics of the interaction of perhydroxyl and superoxide radicals with hydrogen peroxide. The Haber-Weiss reaction. J. Am. Chem. Soc. 1979, 101 (1), 58–62, DOI: 10.1021/ja00495a010. (32) Koppenol, W. H.; Butler, J. Energetics of interconversion reactions of oxyradicals. Adv. Free Radic. Biol. Med. 1985, 1 (1), 91–131, DOI: 10.1016/8755-9668(85)90005-5. (33) Anipsitakis, G. P.; Dionysiou, D. D. Degradation of organic contaminants in water with sulfate radicals generated by the conjunction of peroxymonosulfate with cobalt. Environ. Sci. Technol. 2003, 37 (20), 4790–4797, DOI: 10.1021/Es0263792.
9314
dx.doi.org/10.1021/es2017363 |Environ. Sci. Technol. 2011, 45, 9308–9314
ARTICLE pubs.acs.org/est
Randomized Intervention Study of Solar Disinfection of Drinking Water in the Prevention of Dysentery in Kenyan Children Aged under 5 Years Martella du Preez,† Ronan M. Conroy,‡ Sophie Ligondo,§ James Hennessy,§ Michael Elmore-Meegan,§ Allan Soita,§ and Kevin G. McGuigan*,|| †
Natural Resources and the Environment, CSIR, P.O. Box 395, Pretoria, South Africa Division of Population Health Sciences, Royal College of Surgeons in Ireland, 123 St Stephens Green, Dublin 2, Ireland § ICROSS, P.O. Box 507, Kenya, Ngong Hills, Kenya Department of Physiology & Medical Physics, Royal College of Surgeons in Ireland, 123 St Stephens Green, Dublin 2, Ireland
)
‡
bS Supporting Information ABSTRACT: We report the results of a randomized controlled intervention study (September 2007 to March 2009) investigating the effect of solar disinfection (SODIS) of drinking water on the incidence of dysentery, nondysentery diarrhea, and anthropometric measurements of height and weight among children of age 6 months to 5 years living in peri-urban and rural communities in Nakuru, Kenya. We compared 555 children in 404 households using SODIS with 534 children in 361 households with no intervention. Dysentery was recorded using a pictorial diary. Incidence rate ratios (IRR) for both number of days and episodes of dysentery and nondysentery diarrhea were significantly (P < 0.001) reduced by use of solar disinfection: dysentery days IRR = 0.56 (95% CI 0.40 to 0.79); dysentery episodes IRR = 0.55 (95% CI 0.42 to 0.73); nondysentery days IRR = 0.70 (95% CI 0.59 to 0.84); nondysentery episodes IRR = 0.73 (95% CI 0.63 to 0.84). Anthropometry measurements of weight and height showed median height-for-age was significantly increased in those on SODIS, corresponding to an average of 0.8 cm over a 1-year period over the group as a whole (95% CI 0.7 to 1.6 cm, P = 0.031). Median weight-for-age was higher in those on SODIS, corresponding to a 0.23 kg difference in weight over the same period; however, the confidence interval spanned zero and the effect fell short of statistical significance (95% CI 0.02 to 0.47 kg, P = 0.068). SODIS and control households did not differ in the microbial quality of their untreated household water over the follow-up period (P = 0.119), but E. coli concentrations in SODIS bottles were significantly lower than those in storage containers over all follow-up visits (P < 0.001). This is the first trial to show evidence of the effect of SODIS on childhood anthropometry, compared with children in the control group and should alleviate concerns expressed by some commentators that the lower rates of dysentery associated with SODIS are the product of biased reporting rather than reflective of genuinely decreased incidence.
’ INTRODUCTION Although a preventable and treatable disease, nearly 1.8 million children under 5 years of age die from diarrhea each year.1 The World Health Organization estimates that in 94% of cases diarrhea is preventable by increasing the availability of clean water and improving sanitation and hygiene.1 Diarrheal disease is strongly linked to fecal contamination. Contamination can occur at source or within the storage container during transport or storage.2 Recontamination may also occur if the drinking utensils are not subject to a regular hygiene regimen.3,4 The prohibitive cost of universally supplying piped water has made household water treatment (HWT) an attractive alternative worldwide. Reviews of the effectiveness of HWT methods5 7 have confirmed that in-home interventions, such as filtration 8,9 chlorination,10,11 a combination of flocculation and chlorination,12,13 and solar disinfection 14 18 can reduce the incidence of diarrhea substantially. r 2011 American Chemical Society
The fundamental principles of one of the simplest and cheapest HWT, solar disinfection (SODIS), were first discussed in 1877 by Downes and Blunt.19 Acra and his colleagues from the American University of Beirut laid the foundations of current research on SODIS with their work on solar irradiation of water and oral rehydration solutions in 1980.14,20,21 More recent laboratory studies have consistently shown that exposing water to sunlight results in significant reduction in bacterial contamination.22 26 However, there are still relatively few controlled field trials to show that this reduction in bacterial levels translates into a reduction in risk of disease in people. Initial trials in Received: June 2, 2011 Accepted: September 21, 2011 Revised: September 13, 2011 Published: September 21, 2011 9315
dx.doi.org/10.1021/es2018835 | Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology Kenyan children reported that solar disinfection was associated with a significant reduction in the risk of diarrheal disease in children aged 5 and under16 and in older children,15 and a further study reported a significant reduction of risk of cholera in children.17 A study by Rose and his colleagues in India in children under 527 showed a significant reduction in risk which occurred despite 86% of the children drinking water other than the solar disinfected water. Rai and co-workers showed a reduction of childhood diarrhea by approximately 76% in an urban population of 136 children under age 5 in North Eastern India.28 A recent trial of solar disinfection in Bolivia by M€ausezahl and colleagues in a setting of very low compliance (32%) failed to show a statistically significant reduction in diarrheal disease, although a reduction in diarrhea was observed for both the test and control communities.29 A study of SODIS in a South African peri-urban environment by du Preez and co-workers in 200930 also reported low compliance levels. Dysentery incidence rates were, however, lower in those drinking solar disinfected water (incidence rate ratio 0.64, 95% CI 0.39 1.0, P = 0.071) but not statistically significantly so. Compared with the control, only participants with higher motivation (defined as adhering to the study protocol at least 75% of the time) achieved a significant reduction in dysentery (incidence rate ratio 0.36, 95% CI 0.16 0.81, P = 0.014). There was no significant reduction in risk at lower levels of motivation. These two studies underline the importance of participant motivation in translating the bactericidal effects of SODIS into health gains. The published research has also some deficiencies. All published trials to date have been carried out on children; there are no trials of the effect of solar disinfection in populations of adults at high risk of water-borne diseases, such as the elderly or those with compromised immune function. Previous Kenyan trials were all carried out in populations drinking heavily contaminated water with high levels of disease risk.31 Furthermore, since the control group participants in these three studies stored their SODIS water indoors in lidded SODIS bottles and refrained from consuming drinking water normally stored in-house, the effect of this improved storage may have caused an underestimation of the true benefit of solar disinfection. Importantly, the previous trial methodology did not allow for the differentiation between dysentery, which has serious health consequences, and nondysentery diarrhea. This is an important weakness, as Wright and his colleagues reported that dysentery in children in rural South Africa and Zimbabwe is associated with faecal contamination of source water, while nondysentery diarrhea was uncorrelated with water quality.32,33 The present trial was one of a series of trials which were carried out in South Africa, Zimbabwe, Kenya, and Cambodia as part of the EU funded SODISWATER project.34 It aimed to address some of the deficiencies of earlier research by distinguishing between dysentery and nondysentery diarrhea in a setting of moderate, rather than severe fecal contamination of drinking water. In the SODISWATER randomized intervention study of 12 month duration among a large population (n = 927) of children under age 5 years in rural Cambodia, McGuigan and coworkers35 have reported high compliance (>90%) and reduced incidence of dysentery, with an incidence rate ratio (IRR) of 0.50 (95% CI 0.27 to 0.93, p = 0.029). SODIS also had a protective effect against nondysentery diarrhea, with an IRR of 0.37 (95% CI 0.29 to 0.48, p < 0.001).
ARTICLE
’ METHODS Participant Selection. Participants were recruited in August and September 2007 from six areas in the Nakuru District of Kenya. Three of these areas (Bondeni, Lanet, and Kaptembwa) are urban slum townships in the city of Nakuru, while three (Mogotio, Salgaa, and Wanyororo) are poor rural areas. The urban locations were supplied almost exclusively by standpipes provided by the Nakuru Water Sanitation Services Company. The Company uses conventional water treatment methods to treat ground- and surface water (personal communication, ICROSS, 2010). In the rural locations, water sources were more variable. Only Salgaa was partly supplied by standpipes (54 of 97 households), while the other rural areas used a mix of river (20.7% of households) borehole water, both protected (4.7%) and unprotected (9.1%) and a small number of miscellaneous sources (see Table 1 of the Supporting Information). Sample Size. Sample size was estimated based on comparison of two Poisson event rates in the presence of significant clustering. Since neither the underlying rates of dysentery nor the strength of clustering effects within households were known, we carried out a series of calculations based on rates of 1 to 10 days of dysentery per year and on different degrees of clustering effects. The projected sample of 1000 children was chosen as offering a 90% power to detect a 10% reduction in risk where the underlying rate was 5 episodes per child per year and clustering effects were strong (rho = 0.2). The sample provided more than 90% power to detect a 20% reduction in incidence for all rates of 2 episodes per child per year or greater Randomization. After obtaining ethical approval from the Kenya Medical Research Institute households were identified using local information provided by health workers operating in the areas. Eligible households stored water in containers in-house, did not have a drinking water tap in the house or yard, and had at least one child (but not more than 5) between 6 months and 5 years old residing in the house. Field workers located the households on foot and recorded their addresses. A study area acronym and house number, linked to the Global Positioning System (GPS) coordinates of the household, was allocated to each household. The addresses and coordinates constituted the sample frame of households. Random numbers between zero and one were generated and allocated to the households. If the random number allocated to a household was less than 0.5 the household was randomized to the test group. If the allocated number was above 0.5 the house was randomized to the control group. Field workers were unaware of how the numbers were allocated. Sampling Issues. The decision to use multistage (cluster) sampling method used in the study was a pragmatic one. No regional sample frame exists which would have allowed the identification of eligible households. The identification of eligible households within villages thus entailed a sampling procedure which selected villages and, within these villages, recruited households. There are two significant sources of clustering within the data: at village level, shared environmental factors such as water and sanitation as well as sociodemographic factors will cause households from each village to resemble each other. Furthermore, recruitment of more than one child per household generates further clustering within the data, since children within the same household will share environmental factors affecting health to a greater extent than children from different households within the same village. This required the use of robust (Huber-White) 9316
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology variance estimation in order to correct for the statistical effects of clustering on estimates of precision. Presurvey. Details of the study and what would be expected of each household and the children during the study were provided verbally and in writing in the local language to parents or carers. Written informed consent was obtained from the head of the household or the carer. Household selection, during which participants were trained at home to complete diarrhea diaries and the use of SODIS was undertaken by field workers that are well trained in aspects of community work and data collection. In addition a field manual provided clear instructions on all the procedures executed during the field study. The presurvey was completed three months prior to the start of the main survey. Household information with regard basic hygiene and water use practices and sanitation were also collected (see Table 1 of the Supporting Information). Field data were captured using handheld computers and scanned barcodes to link records. The data were downloaded into a database and checked for completeness and consistency before analysis. Two 2-L PET bottles were provided for each child in the intervention group. Carers of children in the intervention group were instructed to fill one bottle and place it in full, unobscured sunlight for a minimum of 6 h every day. In practice most bottles were exposed for longer than 6 h since parents or guardians usually placed the bottle outdoors early in the morning and brought it in at the end of the day. Consequently children in the intervention group drank from a bottle which had been exposed to sunlight on the previous day. Treated water was consumed on the day after exposure. To minimize the possibility of regrowth of partially inactivated bacteria carers were instructed to store the water for a maximum of 48 h. Carers were advised that, where possible, children in the intervention group should drink disinfected water directly from the SODIS bottle rather than from a cup or other container which might have presented a risk of recontamination of the water. Children in the control group were not provided with SODIS bottles and instead were instructed to maintain their usual practices. Anthropometry. Formal anthropometric standardization to determine the precision and accuracy of each person taking height and weight measurements was not conducted because anthropometry was not the main focus of the study. However, field supervisors, who took the measurements, attended a weeklong training session in South Africa during which the use of the equipment (standard adult digital battery operated weighing scales, stature meter and rollameter) was demonstrated and practiced. Special attention was given to the basic anthropometric principles such as calibration of the scales, accuracy when taking measurements, measuring techniques, and ensuring that correct data were recorded. Babies weighing less than 10 to 15 kg were weighed in the arms of the mother or carer. The weight of the babies was calculated in the laboratory. Older children were weighed standing unsupported on the adult scale. In either case the child was shoeless, wearing only a minimum of light clothing. A plank was used as a smooth horizontal position for the scales, stature meter, and the rollameter. The stature meter was always set up against a sturdy vertical wall or door frame. Attention was given to the position of the feet, knees, and position of the head of the subjects when using either the stature or rollameter (see Figures 1 and 2 of the Supporting Information). A manual provided detailed illustrated information and instructions on conducting anthropometry. An initial pilot scale study in South
ARTICLE
Africa and second pilot scale study undertaken in Kenya provided further practical sessions. Health Outcome. The primary health outcome of the study was days on which the child had dysentery, defined according to Baqui, as any loose stool which contained blood or mucus.36 A dysentery diarrhea day was defined as a single day in which one or more stools contained either blood or mucus. One or more consecutive dysentery diarrhea days occurring followed by three consecutive days on which neither dysentery nor nondysentery diarrhea occurred constituted a dysentery episode. Nondysentery diarrhea was defined as three or more loose or watery stools on the same day without blood or mucus, while an episode was characterized by three consecutive days on which neither dysentery nor nondysentery diarrhea occurred. Nondysentery diarrhea days and dysentery diarrhea days were recorded daily using pictorial diaries developed by Gundry and colleagues,33 which record the number and consistency of the child’s stools. Diarrheal incidence was recorded daily for both control and test children for 17 months. Monitoring. Three monitoring visits to determine the microbial water quality and anthropometry were undertaken (July 2008, October 2008, and January 2009). Each visit included all households, but carers and children were often absent from their homes. This was particularly so during holidays when children are sent away to live with relatives or grandparents. Attempts to obtain data from these households included additional follow-up visits, but the long distances between the study areas made this an expensive and unfeasible procedure. As a result we were unable to collect data from every household for each of the three visits. Between monitoring visits trained field staff visited participating households every two weeks to collect diarrheal diaries. The diaries were checked for discrepancies and corrected when possible. Problems raised by the participants were resolved during these visits. Compliance was measured from the collection of the pictorial diarrhea diaries and by recording the responses of household caregivers during monitoring visits, every three months. On these occasions caregivers were asked (i) whether they were using SODIS and (ii) whether it was possible to collect a water sample from the SODIS bottle that was in use. Between monitoring visits field staff regularly reminded the SODIS group about the technique and inquired if they were still using it. Water from the storage containers and SODIS bottles were collected in commercially available 100 mL sample bottles containing sodium thiosulfate to neutralize any residual chlorine in the water. Samples were transported on ice and analyzed on the same day using the Colilert-18 Quantitray, most probable number (MPN) method37 to quantify E. coli. The maximum possible count obtainable using the 0 200 cell forming units per 100 mL Quanti-tray is >200.5 and the minimum <1. Statistical Methods. Data were analyzed with Stata/SE, Release 11. Stata’s robust variance estimation routines for clustered data, implemented in the svy procedures, were used to adjust for the effects of the multistage sample design, with children sampled within houses, and stratified by village (6 units). Initial analysis confirmed that incidence rates of dysentery were overdispersed, making a Poisson regression inappropriate. Generalized negative binomial regression was used to calculate incidence rate ratios. Generalized negative binomial regression allows for variation in disease rates between individuals who have the same risk factor profile and allows this variation to be modeled as a function of predictor variables. 9317
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology
ARTICLE
Figure 1. Flow diagram showing the course of the intervention.
E. coli data were transformed to a log scale and analyzed using interval regression. This allows values of zero to be analyzed as representing <1 colony forming unit (CFU) and values above the upper threshold of the system (>200.5) to be analyzed as representing a value greater than the threshold. The advantage of this method is that it allows the presence of values which are interval-censored (not known precisely but known to lie in a defined interval). This method overcomes the unreasonable assumptions frequently made when analyzing water quality data that a) values which are greater than the highest readable value are equal to that value and b) that samples which show no CFUs indicate that the water from which the sample was taken contains zero E. coli. Interval regression also overcomes the problem of expressing readings of zero on a logarithmic scale, since such values are simply expressed as being less than 1, rather than truly zero. Analysis of faecal coliform concentrations also used robust variance estimation to adjust for clustering of data within households. Anthropometry measures were analyzed by using fractional polynomial quantile regression to estimate the fiftieth and tenth centiles of weight-for-age and height-for-age. This approach was adopted rather than using z-scores based on means and standard deviations on the grounds that such an approach assumes a normal distribution of data which is unlikely in socially deprived conditions, and the use of published norms may fail to take into account the empirical growth trajectories of the study population. By using the extensive data available, we were able to construct empirical growth curves that best fitted the study population. The number of months drinking SODIS water was taken as the predictor variable to indicate the effect of SODIS on weight for age. Since the time on SODIS is zero for all measures taken from control children, the regression estimates the effect of duration of
SODIS use on anthropometric variables. We chose to estimate growth curves directly from the data because anthropometry norms developed for one population may be inappropriate for another.
’ RESULTS There were 765 households, with 404 (53%) randomized to solar disinfection. The types of water sources (combining the spring, dug well protected and unprotected and the canal and other; see Table 1 of the Supporting Information) used by the test and control groups were not statistically significantly different (χ2 = 4.394, df = 4, P = 0.355). Almost all (92%) had access to a toilet, of which the majority (89%) were pit latrines. Only 7% of participants had access to a flush toilet. A median of 15 people shared a toilet. Sharing of toilets was evident in both the periurban areas where residents were provided with public toilets and the rural areas where one toilet in a yard was often shared with neighboring households. Hygiene levels determined by hand washing at critical times were high and not statistically significantly different between the test and control group for example, before eating (χ2 = 0.017, df = 1, P = 0.897), before preparing food, (χ2 = 1.410, df = 1, P = 0.234), after changing a baby’s nappy (χ2 = 1.737, df = 1, P = 0.1875), and after using the toilet (χ2 = 2.000, df = 1, P = 0.157). Loss to Follow-up. Six of the participating children died during the study period. Of these, three were in the intervention and three in the control groups. Cause of death is unknown for two of the three children in each group. One child in the intervention group died of diarrhea and one child in the control group died of pneumonia. Post election violence caused displacement of households in February 2008 resulting in a temporary loss of 444 children directly after the upheaval. 9318
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology
ARTICLE
Table 1. Unadjusted Annual Rates of Dysentery and Nondysentery Diarrhoea Days and Episodes dysentery
nondysentery diarrhea
Table 2. Incidence Rate Ratios for Dysentery and Nondysentery Days and Episodes with Estimates Adjusted for Water Source, Study Area, and Child Age end point
group
incidence rate ratio
95% Ci
sig
days
episodes
days
episodes
control
5.20
2.02
10.89
4.75
dysentery days dysentery episodes
0.56 0.55
0.40 to 0.79 <0.001 0.42 to 0.73 <0.001
test (SODIS)
3.34
1.31
8.07
3.65
nondysentery diarrhea days
0.70
0.59 to 0.84 <0.001
nondysentery diarrhea episodes
0.73
0.63 to 0.84 <0.001
Many households returned to their homes once the violence ended, and at the end of the study 24 children in the test group and 20 in the control group were completely lost to follow-up. The displacement of a large proportion of the study sample meant that many would have changed water source, sanitation facilities, and living conditions on several occasions during the study, making use of baseline data on these variables inappropriate for analysis of disease rates. A flow diagram showing the course of the intervention is provided in Figure 1. The total number of children randomized was 1089, with 555 (51%) randomized to solar disinfection. The intervention and control groups did not differ with respect to either age (P = 0.980, t test) or sex (P = 0.744, Chi-squared test). Median duration of diarrhea recording was 11 months; 9% of participants had less than 3 months, 75% had 17 months or less, and 15% had 17 or 18 months. Intervention and control groups did not differ in the number of days of diarrhea data recorded (t test, P = 0.492). Table 1 shows the unadjusted annual rates of dysentery- and nondysentery days and episodes for the test and control group of children. Rates of dysentery were related to age. Compared with children under 1 year, children aged 1 had an incidence rate ratio of 0.72 (P = 0.109) for days of dysentery, children aged 2 an incidence rate of 0.52 (P = 0.001), children aged 3 an incidence rate of 0.45 (P < 0.001), and children aged 4 a rate of 0.29 (P < 0.001). Likewise, nondysentery diarrhea incidence fell with age. Compared with children under 1 year, children aged 1 had an incidence rate ratio of 0.68 (P = 0.003) for days of nondysentery diarrhea, children aged 2 an incidence rate of 0.67 (P = 0.002), children aged 3 an incidence rate of 0.52 (P < 0.001), and children aged 4 a rate of 0.47 (P < 0.001). Children drinking water from standpipes were at somewhat lower risk of dysentery with an incidence rate ratio 0.77 for days of dysentery, but the associated confidence interval was wide (0.41 1.4). However, water source was adjusted for in the analysis of the effect of SODIS in view of the absolute effect size. Table 2 shows the incidence rate ratios for each end point with estimates adjusted for water source (standpipe versus other water source), study area (entered as 5 dummy variables), and child age in whole years (4 dummy variables). Dispersion was parametrized by study area (5 dummy variables). All diarrhea end points were significantly reduced by use of solar disinfection, with reductions of roughly 50% in the incidence of dysentery and approximately 30% in the incidence of nondysentery diarrhea. We used interval regression to compare E. coli concentrations in storage and SODIS bottle water in intervention households. E. coli concentrations were transformed to log10 units as described in the Methods section before analysis. There were data available on 516 households at visit 1 (24 to 34 weeks from trial start), and 468 households at both visit 2 (34 to 40 weeks) and visit 3 (47 to 55 weeks). 50% of samples from stored household water had 10 CFU/100 mL or less; however, 23% had 100 CFU mL or more. We compared storage water quality in control
and SODIS households using interval regression. E. coli concentrations did not differ over the follow-up period between control and SODIS households (z = 1.56, P = 0.119). Water samples from SODIS bottles showed lower concentrations of E. coli at each visit, as shown in Table 3. The use of interval regression requires transformation of the data to logarithmic units. The values for storage water, back-transformed, correspond to geometric mean values of 8.4, 5.1, and 3.2 CFU/ 100 mL at the three time points, while the geometric mean values for the SODIS bottle samples represent 0.3, 0.2, and 0.1 CFU/ 100 mL, respectively. The coefficients for the difference correspond to the ratio between the storage and SODIS samples, which are 31, 30, and 22 at the three follow-ups, respectively. Height and weight measurements were available on 656 children at visit 1, 653 at visit 2, and 632 at visit 3. There was no significant difference between SODIS and control groups in the numbers of measurements made at each visit (Chi-squared test, P = 0.972). We examined height-for-age and weight-for-age by modeling the effects of age on each parameter as a two-term fractional polynomial, having verified that no significant improvement in fit was obtained by modeling age as three parameters. The effect of SODIS was modeled by converting the length of time on SODIS to a fraction of a year, allowing calculation of the effect of a year on SODIS. Median weight in the children in the control group at age 1 was 10.0 kg, 25th percentile 9.2, 75th percentile 11.1 kg. At age 5, median weight was 16.4 kg, 25th percentile 14.9, 75th percentile 17.8 kg. Median height-for-age was significantly increased in those on SODIS, corresponding to an average of 0.8 cm over a 1-year period over the group as a whole (95% CI 0.7 to 1.6 cm, P = 0.031). Although median weight-for-age was similarly higher in those on SODIS, corresponding to a 0.23 kg difference in weight after a year on SODIS, the confidence interval spanned zero and the effect fell short of statistical significance (95% CI 0.02 to 0.47 kg, P = 0.068). An examination of weight for height revealed no significant effect of length of time on SODIS (P = 0.351).
’ DISCUSSION This study adds to the evidence of the effectiveness of solar disinfection as a public health measure to reduce the risk of childhood dysentery. SODIS was associated with a reduction of 44% in the incidence of dysentery days and a reduction of 30% in nondysentery diarrhea days. The use of diarrheal diaries allows dysentery and nondysentery diarrhea to be analyzed separately. In addition, the use of diaries allows analysis using either days or episodes. This is important, as the mortality risk associated with diarrheal disease in children is associated with dysentery rather than nondysentery diarrhea.38 Bloody diarrhea in children is a sign of intestinal infection caused by invasive enteric bacteria often associated with. Shigella. Of the Shigella species pathogenic to man, Shigella boydii, S. flexneri, S. sonnei, and S. dysenteriae type 1, 9319
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology
ARTICLE
Table 3. E. coli Concentrations (log10 cfu/100 mL units) in Untreated Water Storage Containers and SODIS Bottles at Each Follow-up Visita visit 1 2
a
E. coli, storage container water
E. coli, SODIS bottle
difference (95% CI)
(log10 cfu/100 mL)
(log10 cfu/ 100 mL)
(log10 cfu/ 100 mL)
0.923 0.707
0.562 0.770
1.48 ( 1.73 to 1.48 ( 1.70 to
1.24) 1.26)
sig <0.001 <0.001
3
0.501
0.847
1.35 ( 1.62 to
1.08)
<0.001
all visits
0.723
0.727
1.45 ( 1.60 to
1.30)
<0.001
Mean levels and confidence intervals calculated by interval regression.
S. flexneri is the main cause for endemic shigellosis in developing countries39 S. dysenteriae type1 is associated with epidemic and endemic shigellosis40 Most endemic shigellosis occurs in children between 6 months and 3 years of age.41,42 Shigellosis also causes loss of important micronutrients for example zinc43 and vitamin A,44 contributing to the nutritional deterioration of children and consequently adversely affect growth.45 The etiology of diarrheal illness among the study population during this intervention is not certain. However, several of the authors conducted a study of prevalence of pathogens in the stools of rural Maasai children under age 5 years within the current study area in the mid 1990s.46 The most common pathogens isolated from the 70 samples at that time were Giardia lamblia (31%), Entamoeba histolytica (23%), enteropathogenic Eschericia coli (13%) Strongyloides stercoralis (4%), Blastocystis hominis (3%), and Cryptosporidium sp (3%). Although all samples were screened for Campylobacter and rotavirus, neither pathogen was detected. While no information was available regarding shigellosis within this population, the presence of dysentery causing pathogen E. hystolytica indicated that amoebic dysentery would not be unexpected. Co-infection with enteropathogens was common with two, three, and four species detected in 18.6%, 1.4%, and 1.4%, respectively, of the same samples studied. However, no pathogens were isolated from 47.2% of the samples. More recently a four year laboratory-based surveillance for bloody diarrhea at five clinics in Western Kenya analyzed 451 stools for the presence of Shigella. Shigella was the most common pathogen, 198 (44%) of the isolates of which 97 (22%) were S. flexneri, 41 (9%) S. dysenteriae type 1, 13 (3%) S. boydii, and 8 (2%) S. sonnei. Campylobacter (33 isolates), nontyphoidal Samonella (15 isolates), and a single Vibrio cholera O1 were also isolated. Shigella was the main cause of bloody diarrhea, and the most common isolate was S. flexneri.47 Other studies conducted in African countries reported similar results.48,49 Although much evidence exists on how improving the water quality at the point-of-use dramatically improves water quality and subsequently reduces diarrhea as much as 40% at household levels,50 53 Schmidt and Cairncross6 concluded that the true effect size, for specifically home water treatment interventions is strongly biased. Confirmation of bias seems apparent in published blinded home-based water quality trials that failed to show any significant effect on diarrheal disease reduction.54 56 Schmidt and Cairncross6 suggested that household water treatment intervention studies should either be blinded or include, as the primary outcome measure, an objective outcome such as mortality, weight gain, or growth. These types of measurements cannot easily be influenced by bias and therefore have the ability to show whether the effect size can truly be attributed to the intervention or not. This trial addresses the concern that
reported associations between SODIS and diarrheal disease may be due to biased reporting on the part of participants, due to the unblinded nature of the trials. The fact that while SODIS and control communities did not differ in their source water quality, but water samples taken from SODIS bottles had lower bacterial levels, provides plausible support for the association being causal. More importantly, however, this is the first trial to show evidence of the effects of SODIS on childhood anthropometry, with a statistically significant difference in the height-for-age of children on SODIS (0.8 cm 95% CI 0.7 cm to 1.6 cm, P = 0.031) compared with children in the control group. These findings should go some way toward alleviating the concern expressed by some commentators that the lower rates of dysentery associated with SODIS are the product of biased reporting by parents rather than reflective of genuinely decreased incidence. The failure to find a significant difference in weight for age may suggest that the observed effect in height may be a false positive finding. However, the considerably greater measurement error entailed in measuring child weight compared with child height may have meant that the study had greater statistical power to detect effects on height than on weight. We should add that the recommendation made by some critics that SODIS be tested in a double-blind fashion reveals a lack of awareness of the conditions in which the communities live who most would benefit from household water treatment. The essential simplicity of SODIS, and the low labor cost, would be lost in the complex organization of a double-blind trial with daily deliveries of anonymized bottles of water, making the trial of questionable external validity, even supposing the resources and organization could be mustered. The gain in height shown in this study is consistent with an effect on child health mediated through the significant reductions in reported dysentery and nondysentery diarrhea. While we cannot rule out observer bias in the measuring of children, the effects demonstrated here are small and unlikely to be detectable by field staff in individual children. Indeed, field staff reported in feedback sessions that they were disappointed by the lack of any apparent difference in the growth of children in the SODIS arm, making biased reporting less likely. Compliance or the motivation to adhere in a sustainable way to the protocols set for water quality interventions is greatly influenced by human behavior and has been problematic in previous SODIS studies.28,29 Furthermore, assessing compliance in trials of this sort is problematic. Observing whether participants have SODIS bottles on their roof suffers from the problem that once field staff appear in a community, bottles appear rapidly on roofs on which they had not been previously. This form of courtesy bias also affects self-reported use of SODIS. More significantly, households that are more compliant with the trial protocol are likely to differ from less compliant households in other health behaviors, potentially biasing any analysis of the 9320
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology effect of compliance. In a previous report, we used compliance with diarrheal data recording as a measure of participant motivation, which allowed comparison between intervention and control groups adjusted for the potential bias associated with degree of compliance. We reported that participants who were more than 75% compliant with diarrhea recording were likely to benefit from SODIS, but below this level there was no statistically significant benefit.30 In the present study, participants maintained good compliance in spite of very difficult circumstances created by political violence that broke out after elections in December 2007. However, compliance was partly dependent upon external factors, due to the displacement of a significant proportion of the study population, and the data available do not allow us to replicate our previous analysis which used completeness of diarrhea recording as a proxy for protocol compliance.. Compliance is driven by socio-economic circumstances, belief, education, the perceived need and benefits of an intervention, the general opinion of the community, compatibility of the intervention to existing values and past experiences, the degree to which results are visible57 study design and the procedures followed to implement and communicate an intervention.58,59 Different cultural and geographical settings naturally elicit different responses to a new idea such as a water quality intervention. In addition unexpected external factors, for example political upheaval, as experienced in Kenya or an infectious disease outbreak can derail adherence to any water quality intervention. Factors that contributed to the good compliance in this study are the need for clean water, poverty (UNICEF estimates that in Kenya, gross national income per capita in 2009 was US$770, and 20% of the population are below the international poverty line of US$1.25 per day60), and subsequently the inability to pay for medical treatment. Anecdotal information confirmed that SODIS water bottles were sought after items and carers believed that their children had fewer incidences of diarrhea reducing costs for medical treatment. In addition the study was extremely well managed by well informed, educated, and positive field coordinators, and participant motivation was probably enhanced by the remarkable role that the field staff played in providing aid during the ethnic violence and its aftermath, despite personal danger. These characteristics have been shown to be important aspects for the success of a water quality intervention.58 The present study adds to the growing literature on the utility of SODIS in the reduction of risk of diarrheal disease. The unanswered questions now are the factors which affect adherence to SODIS, and the comparative merits of SODIS and other household water treatment methods such as filtration and chlorination, and, most importantly, the question of how to introduce household water treatment into communities in such a way as to make sustained changes to behavior.
’ ASSOCIATED CONTENT
bS
Supporting Information. Table 1 and Figures 1 and 2. This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +353 1 4022207. E-mail: [email protected].
ARTICLE
’ ACKNOWLEDGMENT The authors must record our praise and admiration for the extraordinary work accomplished by the members of the SODIS field work team under difficult and dangerous conditions during the period of postelection violence in Kenya from December 2007 to February 2008. The team consisted of Danny Ngwiri, Nyakoboke Oriere, Geoffrey Njoroge, Amos Omare, Lillian Atieno, Mary Waruguru, Pamela Auma, Jessica Kesui, David Okinja, Christine Mokua, Peter Mbugua, Anthony Gikonyo, Elizabeth Maina, Maurice Oketch, Stanley Muchangi, Peris Wambui, and Masaai Kipruto. Despite the fact that several of them had been forced to flee their homes and live in refugee camps, they elected to keep data collection to schedule and gave humanitarian assistance to members of the communities on both sides of the conflict. We would also like to thank the study communities in Nakuru for their participation and support for the project. This research was funded by the European Union (contract FP6-INCO-CT-06-031650). The authors have no proprietary, professional, financial, or other personal interest of any nature or kind in any product, service, and/or company that could be construed as influencing the position presented in, or the review of, this work. ClinicalTrials.gov Registration: NCT01306383. ’ REFERENCES (1) WHO. Combating waterborne disease at the household level; World Health Organisation: Geneva, 2007; pp 1 35. http://www.who.int/ household_water/advocacy/combating_disease.pdf (accessed month day, year). (2) Wright, J. A.; Gundry, S. W.; Conroy, R. M.; Wood, D.; du Preez, M.; Ferro-Luzzi, A.; Genthe, B.; Kirimi, M.; Moyo, S.; Mutisi, C.; Ndamba, J.; Potgieter, N. Defining episodes of diarrhoea: Results from a three-country study in Sub-Saharan Africa. J. Health, Popul. Nutr. 2006, 24 (1), 8–16. (3) Rufener, S.; Maeusezahl, D.; Mosler, H.-J.; Weingartner, R. Drinking water quality between source and point-of-consumption drinking cups as a high potential recontamination risk: A field study in Bolivia. J. Health, Popul. Nutr. 2010, 28 (1), 34–41. (4) Levy, K.; Nelson, K. L.; Hubbard, A.; Eisenberg, J. N. S. Following the water: a controlled study of drinking water storage in northern coastal Ecuador. Environ. Health Perspect. 2008, 116, 1533–1540. (5) Clasen, T.; Menon, S. Microbiological performance of common water treatment devices for household use in India. Int. J. Environ. Health Res. 2007, 17 (2), 83–93. (6) Schmidt, W. P.; Cairncross, S. Household water treatment in poor populations: Is there enough evidence for scaling up now? Environ Sci. Technol. 2009, 43 (4), 986–992. (7) Sobesy, M. Managing water in the home: accelerated health gains from improved water supply. Household Water Treatment & Storage 2004 [cited 2005 Nov 2005]. Available from http://www.who.int/water_ sanitation_health/dwq/wsh0207/en/index.html (accessed month day, year). (8) Hunter, P. R. Household Water Treatment in Developing Countries: Comparing Different Intervention Types Using MetaRegression. Environ. Sci. Technol. 2009, 43 (23), 8991–8997. (9) Lantagne, D.; Meierhofer, R.; Allgood, G.; McGuigan, K.; Quick, R. Comment on “Point of Use Household Drinking Water Filtration: A Practical, Effective Solution for Providing Sustained Access to Safe Drinking Water in the Developing World”. Environ. Sci. Technol. 2009, 43 (3), 968–969. (10) Betancourt, W. Q.; Rose, J. B. Drinking water treatment processes for removal of Cryptosporidium and Giardia. Vet. Parasitol. 2004, 126 (1 2), 219–34. 9321
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology (11) Hirata, T.; Chikuma, D.; Shimura, A.; Hashimoto, A.; Motoyama, N.; Takahashi, K.; Moniwa, T.; Kaneko, M.; Saito, S.; Maede, S. Effects of ozonation and chlorination on viability and infectivity of Cryptosporidium parvum oocysts. Water Sci. Technol. 2000, 41, 39–46. (12) Huertas, A.; Barbeau, B.; Desjardins, C.; Galarza, A.; Figueroa, M. A.; Toranzos, G. A. Evaluation of Bacillus subtilis and coliphage MS2 as indicators of advanced water treatment efficiency. Water Sci. Technol. 2003, 47 (3), 255–9. (13) Souter, P. F.; Cruickshank, G. D.; Tankerville, M. Z.; Keswick, B. H.; Ellis, B. D.; Langworthy, D. E.; Metz, K. A.; Appleby, M. R.; Hamilton, N.; Jones, A. L.; Perry, J. D. Evaluation of a new water treatment for point-of-use household applications to remove microorganisms and arsenic from drinking water. J. Water Health 2003, 1 (2), 73–84. (14) Acra, A.; Raffoul, Z.; Karahagopian, Y. Solar disinfection of drinking water and oral rehydration solutions: Guidelines for household application in developing countries; 1984. (15) Conroy, R. M.; Elmore-Meegan, M.; Joyce, T.; McGuigan, K. G.; Barnes, J. Solar disinfection of drinking water and diarrhoea in Maasai children: a controlled field trial. The Lancet 1996, 348 (9043), 1695–7. (16) Conroy, R. M.; Meegan, M. E.; Joyce, T.; McGuigan, K.; Barnes, J. Solar disinfection of water reduces diarrhoeal disease: an update. Arch. Dis. Child. 1999, 81 (4), 337–8. (17) Conroy, R. M.; Meegan, M. E.; Joyce, T.; McGuigan, K.; Barnes, J. Solar disinfection of drinking water protects against cholera in children under 6 years of age. Arch. Dis. Child. 2001, 85 (4), 293–5. (18) Joyce, T. M.; McGuigan, K. G.; Elmore-Meegan, M.; Conroy, R. M. Inactivation of fecal bacteria in drinking water by solar heating. Appl. Environ. Microbiol. 1996, 62 (2), 399–402. (19) Downes, A.; Blunt, T. P. Researches on the effect of light upon bacteria and other organisms. Proceedings of the Royal Society; 1877; Vol. 28, pp 488 500. (20) Acra, A.; Jurdi, M.; Mu’Allem, H.; Karahagopian, Y.; Raffoul, Z. Sunlight as disinfectant. The Lancet 1989, 1 (8632), 280. (21) Acra, A.; Karahagopian, Y.; Raffoul, Z.; Dajani, R. Disinfection of oral rehydration solutions by sunlight. The Lancet 1980, 2 (8206), 1257–8. (22) Berney, M.; Weilenmann, H. U.; Egli, T. Flow-cytometric study of vital cellular functions in Escherichia coli during solar disinfection (SODIS). Microbiology 2006, 152 (Pt 6), 1719–29. (23) Boyle, M.; Sichel, C.; Fernandez-Iba~nez, P.; Arias-Quiroz, G.; Iriarte-Pu~na, M.; McGuigan, K. Identifying the bactericidal limits of Solar Disinfection (SODIS) of water under real sunlight conditions. Appl. Environ. Microbiol. 2008, 74 (10), 2997–3001. (24) McGuigan, K. G.; Joyce, T. M.; Conroy, R. M.; Gillespie, J. B.; Elmore-Meegan, M. Solar disinfection of drinking water contained in transparent plastic bottles: characterizing the bacterial inactivation process. J. Appl. Microbiol. 1998, 84 (6), 1138–48. (25) McGuigan, K. G.; Mendez-Hermida, F.; Castro-Hermida, J. A.; Ares-Mazas, E.; Kehoe, S. C.; Boyle, M.; Sichel, C.; Fernandez-Iba~ nez, P.; Meyer, B. P.; Ramalingham, S.; Meyer, E. A. Batch solar disinfection (SODIS) inactivates oocysts of Cryptosporidium parvum and cysts of Giardia muris in drinking water. J. Appl. Microbiol. 2006, 101 (2), 453–463. (26) Wegelin, M.; Canonica, S.; Mechsner, K.; Fleischmann, T.; Pesaro, F.; Metzler, A. Solar water disinfection: scope of the process and analysis of radiation experiments. J. Water SRT - Aqua 1994, 43, 154–169. (27) Rose, A.; Roy, S.; Abraham, V.; Holmgren, G.; George, K.; Balraj, V.; Abraham, S.; Muliyil, J.; Joseph, A.; Kang, G. Solar disinfection of water for diarrhoeal prevention in southern India. Arch. Dis. Child. 2006, 91 (2), 139–41. (28) Rai, B. B.; Pal, R.; Kar, S.; Tsering, D. C. Solar Disinfection Improves Drinking Water Quality to Prevent Diarrhea in Under-Five Children in Sikkim. India J. Global Infect. Dis. 2010, 2 (3), 221–225. (29) M€ausezahl, D.; Christen, A.; Pacheco, G. D.; Tellez, F. A.; Iriarte, M.; Zapata, M. E.; Cevallos, M.; Hattendor, f.J.; Cattaneo, M. D.; Arnold, B.; Smith, T. A.; Colford, J. M., Jr. Solar drinking water disinfection (SODIS) to reduce childhood diarrhoea in rural Bolivia: a cluster-randomized, controlled trial. PLoS Med. 2009, 6 (8), e1000125.
ARTICLE
(30) du Preez, M.; McGuigan, K. G.; Conroy, R. M. Solar disinfection of drinking water (SODIS) in the prevention of dysentery in South African children aged under 5 years: the role of participant motivation. Environ. Sci. Technol. 2010, 44 (22), 8744–8749. (31) Amin, M. T.; Han, M. Roof-harvested rainwater for potable purposes: application of solar disinfection (SODIS) and limitations. Water Sci. Technol. 2009, 60 (2), 419–431. (32) Wright, J. A.; Gundry, S. W.; Conroy, R. M. Household drinking water in developing countries: a systematic review of microbiological contamination between source and point-of-use. Trop. Med. Int. Health 2004, 9, 106–117. (33) Gundry, S. W.; Wright, J. A.; Conroy, R. M.; du Preez, M. D.; Genthe, B.; Moyo, S.; Mutisi, C.; Potgieter, N. Child dysentery in the Limpopo Valley: a cohort study of water, sanitation and hygiene risk factors. J. Water Health 2009, 7, 259–66. (34) SODISWATER, Solar Disinfection of Drinking Water for Use in Developing Countries or in Emergency Situations (EU Grant no. FP6-INCO-CT-2006-031650). European Union: Kenya, South Africa, Zimbabwe, 2006. (35) McGuigan, K. G.; Samaiyar, P.; du Preez, M.; Conroy, R. M. A high compliance randomised controlled field trial of solar disinfection (SODIS) of drinking water and its impact on childhood diarrhoea in rural Cambodia. Environ. Sci. Technol., 2011. In press. DOI http://dx. doi.org/10.1021/es201313x. (36) Baqui, A. H.; Black, R. E.; Yunus, M.; Hoque, A. R.; Chowdhury, H. R.; Sack, R. B. Methodological issues in diarrhoeal diseases epidemiology: Definition of diarrhoeal episodes. Int. J. Epidemiol. 1991, 20 (4), 1057–1063. (37) Covert, T. C.; Shadix, L. C.; Rice, E. W.; Haines, J. R.; Freyberg, R. W. Evaluation of the auto-analysis Colilert test for detection and enumeration of total coliforms. Appl. Environ. Microbiol. 1989, 55 (10), 2443–2447. (38) Kotloff, K. L.; J. P. Winicoff, I. B.; Clemens, J. D.; Swerdlow, D. L.; , Sansonetti, P. J. Global burden of Shigella infection: implications for vaccine development and implementation of control strategies. Bulletin of the World Health Organisation; 1999; Vol. 77, pp 651-666. (39) WHO. The management of bloody diarrhoea in young children WHO/CCD 1994 28/8/2011]; Available from: https:// apps.who.int/chd/publications/cdd/bloody_d.htm (accessed month day, year). (40) Bennish, M. L.; Wojtyniak, B. J. Mortality due to shigellosis: community and hospital data. Rev. Infect. Dis. 1991, 13 (4), 245–251. (41) Ahmad-Clemens, J.; Rao, M. R.; Sack, D.; Khan, M. R.; Haque, E. Community based evaluation of breast feeding on the risk of microbiologically confirmed or clinically presumptive shigellosis in Bangladeshi children. Pediatrics 1992, 90, 406–411. (42) Kahn, M. U.; Chandra, R. N.; Islam, R.; Huq, I.; Stoll, B. Fourteen years of Shigellosis in Dhaka: an epidemiological analysis. Int. J. Epidemiol. 1985, 14, 607–613. (43) Castillo-Duran, C.; Vial, P.; Uauy, R. Trace mineral balance during acute diarrhoea in infants. J. Pediatr. 1988, 113, 452–457. (44) Mitra, A.; Alvarez, J. O.; Guay-Woodford, L. G.; Fuchs, G. J.; Wahed, M. A.; Stephenson, C. B. Urinary retinol excretion and kidney function in children with shigellosis. Am. J. Clin. Nutr. 1998, 68, 1095–1103. (45) Bennish, M. L.; Salam, M. A.; Wahed, M. A. Enteric protein loss during shigellosis. Am. J. Gastroenterol. 1993, 88 (1), 53–57. (46) Joyce, T.; McGuigan, K. G.; Elmore-Meegan, M.; Conroy, R. M. Prevalence of enteropathogens in stools of rural Maasai children under five years of age in the Maasailand region of the Kenyan Rift Valley. East African Med. J. 1996, 73 (1), 59–62. (47) Brooks, J. T.; Shapiro, R. L.; Kumar, L.; Wells, J. G.; PhillipsHoward, P. A.; Shi, Y.-P.; Vulule, J. M.; Hoekstra, R. M.; Mintz, E.; Slutsker, L. Epidemiology of sporadic bloody diarrhea in rural Western Kenya. Am. J. Trop. Med. Hyg. 2003, 68 (6), 671–677. (48) Ronsmans, C.; Bennish, M. L.; Wierzba, T. Diagnosis and management of dysentery by community health workers. Lancet 1988, 2 (8610), 552–555. 9322
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
Environmental Science & Technology
ARTICLE
(49) Adeleye, I. A.; Adetosoye, A. I. Antomicrobial resistance patterns and plasmid survey of Salmonella and Shigella isolated in Ibadan, Nigeria. East African Med. J. 1993, 70, 259–262. (50) Arnold, B. F.; Colford, J. M., Jr. Treating water with chlorine at point-of-use to improve water quality and reduce child diarrhea in developing countries: a systematic review and meta-analysis. Am. J. Trop. Med. Hyg. 2007, 76 (2), 354–64. (51) Clasen, T. F.; Brown, J.; Collin, S.; Suntura, O.; Cairncross, S. Reducing diarrhea through the use of household-based ceramic water filters: a randomized, controlled trial in rural Bolivia. Am. J. Trop. Med. Hyg. 2004, 70 (6), 651–7. (52) Clasen, T. F.; Cairncross, S. Household water management: refining the dominant paradigm. Trop. Med. Int. Health 2004, 9 (2), 187–91. (53) Fewtrell, L.; Kaufmann, R. B.; Kay, D.; Enanoria, W.; Haller, L.; Colford, J. M., Jr. Water, sanitation, and hygiene interventions to reduce diarrhoea in less developed countries: a systematic review and metaanalysis. Lancet Infect. Dis. 2005, 5 (1), 42–52. (54) Boisson, S.; Kiyombo, M.; Sthreshley, L.; Tumba, S.; Makambo, J.; Clasen, T. Field assessment of a novel household-based water filtration device: a randomised, placebo-controlled trial in the Democratic Republic of Congo. PLoS One 2010, 5 (9), e12613. (55) Jain, S.; Sahanoon, O. K.; Blanton, E.; Schmitz, A.; Wannemuehler, K. A.; Hoekstra, R. M.; Quick, R. E. Sodium dichloroisocyanurate tablets for routine treatment of household drinking water in peri-urban Ghana: a randomized controlled trail. Am. J. Trop. Med. Hyg. 2010, 82, 16–22. (56) Kirchoff, L. V.; McClelland, K. E.; Do Carmo Pinho, M.; Araujo, J. G.; De Sousa, M. A.; Guerrant, R. L. Feasibility and efficiency of in-home water chlorination in rural North-east Brazil. J. Hyg. (Camb) 1985, 94, 173–180. (57) Diffusion of innovations, 5th ed.; Roger, E. M., Ed.; Free Press: New York, 2003. (58) Meierhofer, R.; Landolt, G. Factors supporting the sustained use of solar water disinfection: Experiences from a global promotion and dissemination programme. Desalination 2009, 248, 144–151. (59) Tamas, A. Mosler, H.-J. SODIS Promotion Investigating the behaviour change process. In Disinfection;Atlanta, USA, 2009. (60) UNICEF. http://www.unicef.org/infobycountry/kenya_statistics. html#79 (accessed month day, year).
9323
dx.doi.org/10.1021/es2018835 |Environ. Sci. Technol. 2011, 45, 9315–9323
ARTICLE pubs.acs.org/est
Size-Tunable Hydrothermal Synthesis of SnS2 Nanocrystals with High Performance in Visible Light-Driven Photocatalytic Reduction of Aqueous Cr(VI) Yong Cai Zhang,*,† Jing Li,† Ming Zhang,† and Dionysios D. Dionysiou‡ †
Key Laboratory of Environmental Material and Environmental Engineering of Jiangsu Province, College of Chemistry and Chemical Engineering, Yangzhou University, Yangzhou 225002, China ‡ Environmental Engineering and Science Program, 705 Engineering Research Center, University of Cincinnati, Cincinnati, Ohio 45221-0012, United States
bS Supporting Information ABSTRACT: SnS2 nanocrystals with adjustable sizes were synthesized via a hydrothermal method from the aqueous solution of common and inexpensive SnCl4 3 5H2O, thioacetamide and citric acid, simply by varying the reaction temperature and reaction time. The structures, BrunauerEmmettTeller (BET) specific surface areas and optical properties of the resultant SnS2 nanocrystals were characterized by X-ray diffraction, transmission electron microscopy, N2 adsorption/desorption isotherms, and UVvis diffuse reflectance spectra. Besides, their photocatalytic properties were tested for the reduction of aqueous Cr(VI) under visible light (λ > 420 nm) irradiation. It was found that the photocatalytic activities of SnS2 nanocrystals in aqueous suspension depended on their synthesis conditions. The product synthesized under suitable hydrothermal conditions (for example, at 150 °C for 12 h) not only showed high visible light-driven photocatalytic activity in the reduction of aqueous Cr(VI), but also showed good photocatalytic stability. Our photocatalytic results suggested that SnS2 nanocrystals are a promising photocatalyst in the efficient utilization of solar energy for the treatment of Cr(VI)-containing wastewater.
’ INTRODUCTION Cr(VI) is a frequent contaminant in the wastewaters arising from industrial processes such as leather tanning, paint making, electroplating and chromate production, etc. It is toxic to most organisms, and has been classified as carcinogenic and mutagenic.110 Due to its high toxicity and high mobility in water, Cr(VI) has been included in the list of priority pollutants and its concentration in drinking water has been regulated by many countries. For instance, the allowable limit of Cr(VI) in drinking water in China is 0.05 mg/L. Therefore, how to economically and efficiently treat the Cr(VI)-containing wastewaters has attracted intense interest from both academic and industrial societies.110 A common method of treating Cr(VI) in water is to convert it into Cr(III), which is considered as a nontoxic and essential trace metal in human nutrition.110 Furthermore, Cr(III) can be precipitated as Cr(OH)3 in neutral or alkaline solutions (KspQ(Cr(OH)3) = 6.3 1031) and removed as a solid waste.110 However, the conventional chemical reduction methods require massive use of reducing agents such as ferrous sulfate, sodium hydrogensulfite, sodium pyrosulfite, hydrazine hydrate, or sulfur dioxide, etc. which are not cost-effective. Compared with the conventional chemical reduction methods, the semiconductor photocatalytic reduction of aqueous Cr(VI) has some obvious advantages, such as simple r 2011 American Chemical Society
operation, ambient conditions, low cost, high efficiency, reusability, direct use of infinite, clean and safe natural solar energy, and no use and no release of other unwanted chemicals.110 Consequently, the semiconductor photocatalytic reduction method is widely regarded as a promising way in treating aqueous Cr(VI).110 TiO2 is undoubtedly the most studied semiconductor photocatalyst so far, by virtue of its low cost, high activity for many photocatalytic reactions, excellent chemical and photochemical stability, and good biocompatibility.110 However, due to its wide band gap (3.2 eV), TiO2 cannot be activated by the visible light, which accounts for about 46% of the total solar energy.11 In order to make full use of solar energy, it is desirable to develop new visible light-responsive semiconductor photocatalysts. The semiconducting metal sulfides usually have light-absorbing capabilities in the visible and short-wavelength near-infrared regions, which enable them to work as a class of promising sensitizers for wide band gap semiconductors or visible lightdriven photocatalysts.1217 Among them, CdS, which has a band Received: June 14, 2011 Accepted: October 4, 2011 Revised: September 7, 2011 Published: October 04, 2011 9324
dx.doi.org/10.1021/es202012b | Environ. Sci. Technol. 2011, 45, 9324–9331
Environmental Science & Technology
ARTICLE
Table 1. Abbreviated Names and Properties of the SnS2 Products Synthesized under Different Reaction Conditionsa
a
BET SA (m2/g)
Eg (eV)
DAA (%)
k (min1)
103.8
2.32
4.6
0.0129
87.6
2.30
5.8
0.0198
1434 (T); 14 (S)
82.4
2.27
10.4
0.0394
18
1535 (T);16 (S)
81.3
2.27
9.9
0.0301
150
24
1541 (T); 17 (S)
76.6
2.25
8.0
0.0205
170
12
2142 (T); 21 (S)
73.1
2.22
9.3
0.0252
name
T (°C)
t (h)
SnS2-(a)
130
12
SnS2-(b)
150
6
1024 (T); 10 (S)
SnS2-(c)
150
12
SnS2-(d)
150
SnS2-(e) SnS2-(f)
size (nm) 717 (T); 7 (S)
SA = surface area; DAA = dark adsorption amount for Cr(VI); k = the photocatalytic reaction rate constants obtained using the pseudo-first-order model; T and S denote the sizes derived from TEM and Scherrer formula, respectively.
gap of about 2.4 eV, is currently the focus of significant attention.1315 However, CdS itself is detrimental to human health and the environment due to its high toxicity. SnS2 is a CdI2-tpye layered semiconductor with a band gap of about 2.2 eV,16 which is a little smaller than that of CdS. It is innocuous, chemically stable in acid or neutral aqueous solution, and so has the potential to be an efficient visible light-driven photocatalyst.12,16,17 Although SnS2 is superior to CdS in terms of lower toxicity and wider spectral response (or higher photocatalytic activity), there is still no report about its use as photocatalyst in the reduction of aqueous Cr(VI) by far. The hydrothermal method is a versatile wet chemical process that has been widely used to prepare semiconductor nanomaterials.1719 It not only enables the preparation of highly crystalline products at low temperatures (generally below 200 °C), but also has the ability of controlling the morphology and size of the resultant products.1719 Herein, we propose a simple and practical hydrothermal route for the size-tunable synthesis of SnS2 nanocrystals at 130170 °C for 624 h, using common and inexpensive SnCl4 3 5H2O, thioacetamide and citric acid as the reactants and water as the solvent. The structures, BET specific surface areas and optical properties of the resultant SnS2 nanocrystals are characterized by X-ray diffraction, transmission electron microscopy, N2 adsorption/desorption isotherms and UVvis diffuse reflectance spectra, and their photocatalytic properties are evaluated in aqueous suspension for the reduction of Cr(VI) under visible light (λ > 420 nm) irradiation.
’ EXPERIMENTAL SSECTION All the reagents used were of analytical grade and purchased from Sinopharm Chemical Reagent Co., Ltd. Synthesis of SnS2 Nanocrystals. Forty mL of aqueous solution containing 5 mmol SnCl4 3 5H2O and equimolar citric acid was first prepared in Teflon-lined stainless steel autoclaves of 50 mL capacity, and subsequently 10 mmol thioacetamide was added to the autoclaves with stirring. The autoclaves were sealed and heated at 130170 °C for 624 h, then cooled naturally to room temperature. The as-formed yellow precipitates were filtered, washed with deionized water, and dried in vacuum at 100 °C for 4 h. For the convenience of description, the SnS2 products synthesized under different reaction conditions were hereinafter named as “SnS2-(a)”, “SnS2-(b)”, and “SnS2-(c)”, etc., as shown in Table 1. Characterization. X-ray diffraction (XRD) patterns of the obtained products were recorded on a German Bruker AXS D8 ADVANCE X-ray diffractometer under the conditions of
Figure 1. XRD patterns of SnS2-(af) and SnS2-AP (the product collected after the fifth reuse cycle of SnS2-(c) in photocatalysis).
generator voltage = 40 kV; generator current = 200 mA; divergence slit = 1.0 mm; scan speed = 6.0 degree/min; Cu Kα (λ = 1.5406 Å); graphite diffracted-beam monochromator; dynamic scintillation counter detector; and polyethylene holder. Transmission electron microscopy (TEM) images were taken on a Holland Philips Tecnai-12 transmission electron microscopy. BET surface areas were measured on an American Micromeritics Instrument Corporation TriStar II 3020 surface area and porosity analyzer. UVvis diffuse reflectance spectra were measured on a Japan Shimadzu UV-3101PC ultravioletvisible-near-infrared spectrophotometer, using BaSO4 as reference. High-resolution transmission electron microscopy (HRTEM) images were taken on an American FEI Tecnai G2 F30 S-TWIN fieldemission transmission electron microscopy. X-ray photoelectron spectroscopy (XPS) measurements were conducted on an American Thermo-VG Scientific ESCALAB 250 XPS system with Al Kα radiation as the exciting source, where the binding energies were calibrated by referencing C 1s peak (284.6 eV) to reduce the sample charging effect. Photocatalytic Tests. The photocatalytic experiments were carried out in a homemade photochemical reactor (Supporting Information Figure S1), which includes mainly four parts: light source system including a 250 W Xe lamp, (λ > 420 nm) cutoff filters and cooling attachments such as electric fan; reactor (two-layer Pyrex glass bottles of 400 mL capacity, the space between the two layers is filled with circulating water to cool the reactor); magnetic stirrer; and temperature controller. In each experiment, the distance between the Xe lamp and the reactor was set to be 5.8 cm, and the reaction temperature was 20 °C. Prior to illumination, 300 mL of 50 mg/L K2Cr2O7 aqueous solution containing 0.3 g of photocatalyst was magnetically stirred in the dark for 1 h. During illumination, about 4 mL of suspension was taken from the reactor at a scheduled interval 9325
dx.doi.org/10.1021/es202012b |Environ. Sci. Technol. 2011, 45, 9324–9331
Environmental Science & Technology
ARTICLE
Figure 2. TEM images of SnS2-(af).
and centrifuged to separate the photocatalyst. The Cr(VI) content in the supernatant solution was determined colorimetrically at 540 nm using the diphenylcarbazide method with a detection limit of 0.005 mg/L.20 The measured absorbance intensities at different illumination times were transformed to the reduction ratio of Cr(VI), which is calculated using the following expression: reduction ratio of CrðVIÞ ¼ ðA0 At Þ=A0 100% where A0 and At are the absorbance intensities when illuminated for 0 (that is, just after the dark adsorption) and t min, respectively.
’ RESULTS AND DISCUSSION Structural Characterization. The XRD patterns of the assynthesized SnS2-(af) are shown in Figure 1. The high background intensities of the XRD patterns may suggest the
existence of amorphous materials in SnS2-(af). All the products displayed only the characteristic XRD peaks of hexagonal phase SnS 2 (JCPDS card no. 23-677). Nevertheless, when the reaction temperature or the reaction time was increased (while the other reaction conditions remained the same), the XRD peaks of the resultant SnS2 products generally became stronger and sharper, suggesting the increase of their crystal sizes. Using the well-known Scherrer formula based on the half-widths of (001) peak in their XRD patterns, the crystal sizes of SnS2-(af) were calculated and presented in Table 1. The particle sizes of SnS2-(af) were further determined from their TEM images (Figure 2), and the obtained results are given in Table 1. As can be seen from Figure 2(af) and Table 1, all the as-synthesized SnS2 products comprised nanoparticles. Although there was some difference between the particle sizes derived from TEM and XRD (Scherrer formula) due to the shape (such as nanoparticles or nanoplates) and aggregation effects of the 9326
dx.doi.org/10.1021/es202012b |Environ. Sci. Technol. 2011, 45, 9324–9331
Environmental Science & Technology
ARTICLE
Figure 4. UVvis diffuse reflectance spectra of SnS2-(af). Figure 3. N2 adsorption/desorption isotherms of SnS2-(af).
as-synthesized SnS2 nanoparticles, the size-changing trends inferred from these two methods are in agreement: the sizes of SnS2 nanoparticles generally increased with the increase of the reaction temperature and reaction time, while the other synthesis conditions were kept constant. These phenomena may be explained by the effects of the reaction temperature and reaction time on the nucleation and crystal growth processes of SnS2 under the hydrothermal conditions, as shown in the following section of Formation mechanism. Formation Mechanism. The current hydrothermal synthesis of SnS2 nanocrystals was carried out using SnCl4 3 5H2O, thioacetamide and citric acid as the reactants and water as the solvent. The Sn4+ cations released from the dissociation of SnCl4 3 5H2O in water (Reaction 1) will combine with citric acid (CA) to form the [Sn(CA)]4+ complex (Reaction 2).21 The formation and dissociation reactions of [Sn(CA)]4+ are reversible (Reaction 2). Meanwhile, thioacetamide (CH3CSNH2) hydrolyzes in the aqueous solution to yield H2S (Reaction 3), which further dissociates to generate S2 anions (Reaction 4 and 5).22 The dissociation/hydrolysis reactions of [Sn(CA)]4+ and CH3CSNH2 can proceed faster at higher temperatures.21,22 SnCl4 3 5H2 O ¼ Sn4þ þ 4Cl þ 5H2 O
ð1Þ
Sn4þ þ CA h ½SnðCAÞ4þ
ð2Þ
CH3 CSNH2 þ H2 O ¼ CH3 CONH2 þ H2 S
ð3Þ
H2 S þ H2 O h H3 Oþ þ HS
ð4Þ
HS þ H2 O h H3 Oþ þ S2
ð5Þ
4+
2
The released Sn and S ions will combine to form SnS2 precipitate once the ionic product of Sn4+ and S2 exceeds the solubility product of SnS2, as expressed by Reaction 6: Sn4þ þ 2S2 ¼ SnS2 V
ð6Þ
But, [Sn(CA)]4+ and CH3CSNH2 tend to release Sn4+ and S2 ions gradually, rather than entirely at once, through the Reactions 25. So, the nucleation rate of SnS2 is controlled by the releasing rates of Sn4+ and S2 ions, and the number of SnS2 crystal nucleus formed in the initial stage (for example, during the temperature rise from room temperature to 130 °C) is limited. Accordingly, there are still large amounts of unreacted Sn4+ and S2 ions in the reaction solution in the forms of [Sn(CA)]4+ and CH3CSNH2, respectively. When the reaction temperature is increased or the reaction time is prolonged, the
Figure 5. Plots of (F(R∞)hν)2 versus (hν) for obtaining the band gaps of SnS2-(af).
successively released Sn4+ and S2 ions will further combine with each other to form SnS2. Because the energy of nucleation required in the heterogeneous phases is lower than that in the homogeneous phase,23,24 the later nucleation of SnS2 should be mainly on the surface of preformed SnS2 crystal nucleus, making contribution to the crystal growth process of SnS2. In addition, the increase of temperature and time also helps to the crystal growth of SnS2 via the Ostwald ripening process, where larger crystals grow at the expense of smaller ones.25,26 Thus, the size of SnS2 nanocrystals becomes bigger with the increase of the reaction temperature and reaction time. BET Surface Area. Figure 3 shows the N2 adsorption/ desorption isotherms of SnS2-(af). All the products displayed type IV isotherms with hysteresis loops at relative pressure (P/P0) between 0.4 and 1.0, indicative of their mesoporous feature.2731 The mesoporous structures of SnS2-(af) were considered to be formed from the aggregation of their primary particles.2731 The values of the BET specific surface areas of SnS2-(af) are listed in Table 1, which exhibited the order of SnS2-(a) > SnS2-(b) > SnS2-(c) > SnS2-(d) > SnS2-(e) > SnS2-(f). Optical Characterization. The UVvis diffuse reflectance spectra of SnS2-(af) were measured and converted into the absorption spectra (Figure 4) using the KubelkaMunk function:3235 FðR ∞ Þ ¼ ð1 R ∞ Þ2 =2R ∞ ¼ α=S R ∞ ¼ R sample =R BaSO4 where F(R∞), R, α and S are the KubelkaMunk function, reflectance, absorption coefficient and scattering coefficient, respectively. According to a previous study,17 the band gaps 9327
dx.doi.org/10.1021/es202012b |Environ. Sci. Technol. 2011, 45, 9324–9331
Environmental Science & Technology
Figure 6. Photocatalytic activities of SnS2-(af) and P25 TiO2 in the reduction of aqueous Cr(VI) under visible light (λ > 420 nm) irradiation.
ARTICLE
Figure 7. Photocatalytic performances of SnS2-(c) in the first five reuse cycles.
(Eg) of SnS2-(af) were determined based on the theory of optical absorption for direct band gap semiconductors: αhν ¼ Bðhν Eg Þ1=2 where hν and B are discrete photon energy and a constant related to the material, respectively. The value of α can be calculated from the diffuse reflectance data using the Kubelka Munk function. But, for the diffused reflectance spectra, the KubelkaMunk function can be used instead of α for estimating the optical absorption edge energy.3235 So, the curves of (F(R∞)hν)2 versus (hν) for SnS2-(af) are plotted in Figure 5. By extrapolating the linear portion of the (F(R∞)hν)2 versus (hν) curves to F(R∞) = 0, the Eg values of SnS2-(af) were estimated to be 2.222.32 eV (Table 1). Photocatalytic Tests. Photocatalytic Activities. Figure 6 shows the photocatalytic activities of SnS2-(af) and P25 TiO2 in the reduction of aqueous Cr(VI) under visible light (λ > 420 nm) irradiation. As can be seen from Figure 6, in the presence of P25 TiO2 or in the absence of any photocatalyst, the reduction of Cr(VI) hardly occurred under visible light (λ > 420 nm) irradiation for 120 min. Instead, the reduction of Cr(VI) proceeded quite rapidly in the presence of SnS2-(af). Nevertheless, the photocatalytic activities of SnS2-(af) differed greatly, indicating that the synthesis conditions of SnS2 nanocrystals played an important role in their photocatalytic activities. The photocatalytic activities of SnS2-(af) followed the order of SnS2-(c) > SnS2-(d) > SnS2-(f) > SnS2-(e) > SnS2-(b) > SnS2-(a). For instance, when irradiated for 120 min, the reduction ratios of Cr(VI) over SnS2-(c), SnS2-(d), SnS2-(f), SnS2-(e), SnS2-(b), and SnS2-(a) were 99.6%, 97.5%, 96.0%, 91.2%, 90.4%, and 77.9%, respectively. Moreover, the photocatalytic reaction rate constants (k) in the presence of SnS2-(c), SnS2-(d), SnS2-(f), SnS2-(e), SnS2-(b), and SnS2-(a) were in turn 0.0394, 0.0301, 0.0252, 0.0205, 0.0198, and 0.0129 min1 (Supporting Information Figure S2 and Table 1), obtained using the pseudo-first-order model as expressed by3641 lnðC0 =CÞ ¼ kt The difference in the photocatalytic activities of SnS2-(af) was most likely a result of the combined action of many factors, such as particle size, specific surface area, adsorption capacity for Cr(VI), band gap, morphology, composition, crystallinity, crystal defects, and dispersibility, etc. Since almost all of the aforementioned factors were strongly coupled, it was difficult to characterize the specific function and influence of a single parameter in the
Figure 8. HRTEM image of SnS2-AP. The fringe interval of 0.316 nm in this image is consistent with the interplanar spacing of (100) crystal planes of hexagonal phase SnS2.
photocatalytic activity of SnS2 nanocrystals. But, there was a direct correlation between the dark adsorption amounts for Cr(VI) and the photocatalytic activities of SnS2-(af) (that is, both the photocatalytic activities and the dark adsorption amounts for Cr(VI) of SnS2-(af) followed the same order of SnS2-(c) > SnS2-(d) > SnS2-(f) > SnS2-(e) > SnS2-(b) > SnS2-(a), as shown in Table 1), suggesting that the adsorption capacities for Cr(VI) of SnS2 nanocrystals should play a predominant role in their photocatalytic activities. Because the photocatalytic reactions are commonly believed to occur on the surface of the photocatalyst, the larger adsorption amounts of Cr(VI) onto SnS2 nanocrystals may contribute to the faster reduction rate of Cr(VI).41,42 Photocatalytic Stability. Since the stability of sulfide photocatalysts has always been a concern, it is important to investigate the stability and reusability of the as-synthesized SnS2 nanocrystals in photocatalytic reduction of aqueous Cr(VI). So, in the current work, SnS2-(c) was recycled for five times in the same photocatalytic reactions. After each reuse cycle which lasted for 120 min, the photocatalyst was separated from the aqueous suspension by filtration, washed with 1 mol/L HNO3 aqueous solution (to reduce the amount of greenish Cr(OH)3 deposited on the surface of SnS2-(c)) and deionized water, dried in vacuum at 100 °C for 4 h, and weighed for the next reuse cycle. Taking into account the mass loss of photocatalyst during each reuse cycle, the fourth reuse cycle must be conducted twice in order to accumulate enough sample for the fifth reuse cycle, the third reuse cycle must be conducted twice in order to accumulate enough sample for the fourth reuse cycle, and so on. Figure 7 shows the photocatalytic performance of SnS2-(c) in the first five 9328
dx.doi.org/10.1021/es202012b |Environ. Sci. Technol. 2011, 45, 9324–9331
Environmental Science & Technology
ARTICLE
Figure 9. XPS spectra of SnS2-AP and SnS2-(c).
reuse cycles. Apparently, the photocatalytic activity of SnS2-(c) deteriorated with the increase in the number of reuse cycle, but only very slightly. Even in the fifth reuse cycle of SnS2-(c), the reduction ratio of Cr(VI) can still reach 97% after visible light irradiation for 120 min. The product collected after the fifth reuse cycle of SnS2-(c) in photocatalysis (which was hereinafter called SnS2-AP for the convenience of description) was further characterized by means of XRD, HRTEM, and XPS. Both the XRD pattern (Figure 1 SnS2-AP) and the HRTEM image (Figure 8) of SnS2-AP demonstrated that it was still hexagonal phase SnS2. The survey XPS spectrum of SnS2-AP (Figure 9) showed the presence of Sn and S components, as well as Cr, C, and O contaminants. From the high resolution XPS spectra of Sn 3d and S 2p core levels (Figure 9), it can be seen that the binding energies of Sn 3d and S 2p of SnS2-AP were nearly the same as those of SnS2-(c), for instance, the binding energies of Sn 3d5/2 and S 2p3/2 of SnS2-AP and SnS2-(c) were 486.61 and 486.65 eV, 161.68, and 161.74 eV, respectively. Furthermore, the binding energies of Sn 3d5/2 and S 2p3/2 of SnS2-AP and SnS2-(c) were all consistent with the reference data of Sn4+ and S2 in SnS2.11,43,44 Besides, the binding energy of Cr 2p3/2 was observed at 577.36 eV (Figure 9), which corresponded to Cr(III) in Cr(OH)3.45,46 The formation of Cr(OH)3 on the surface of SnS2-AP can be due to the hydrolysis-precipitation of Cr(III) cations, which were generated from the photocatalytic reduction of adsorbed Cr(VI). Unfortunately, the deposition of Cr(OH)3 on the surface of SnS2-(c) was likely to occupy some photocatalytic active sites of the latter, and accordingly decreased slightly the photocatalytic activity of SnS2-(c) during its reuse.46
’ ASSOCIATED CONTENT
bS
Supporting Information. Figure S1 and Figure S2. This material is available free of charge via the Internet at http://pubs. acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 086 0514 87962581; fax: 086 0514 87975244; e-mail: [email protected].
’ ACKNOWLEDGMENT Thanks to the China Postdoctoral Science Foundation funded project, Jiangsu Planned Projects for Postdoctoral Research Funds, National Natural Science Foundation of China (50873085), and Natural Science Foundation of Jiangsu Province (08KJB150019). ’ REFERENCES (1) Testa, J. J.; Grela, M. A.; Litter, M. I. Heterogeneous photocatalytic reduction of chromium(VI) over TiO2 particles in the presence of oxalate: Involvement of Cr(V) species. Environ. Sci. Technol. 2004, 38 (5), 15891594; DOI: 10.1021/es0346532. (2) Kleiman, A.; Marquez, A.; Vera, M. L.; Meichtry, J. M.; Litter, M. I. Photocatalytic activity of TiO2 thin films deposited by cathodic arc. Appl. Catal., B 2011, 101 (34), 676681; DOI: 10.1016/j.apcatb.2010.11.009. (3) Vinu, R.; Madras, G. Kinetics of simultaneous photocatalytic degradation of phenolic compounds and reduction of metal ions with nano-TiO2. Environ. Sci. Technol. 2008, 42 (3), 913919; DOI: 10.1021/es0720457. (4) Mu, R.; Xu, Z.; Li, L.; Shao, Y.; Wan, H.; Zheng, S. On the photocatalytic properties of elongated TiO2 nanoparticles for phenol degradation and Cr(VI) reduction. J. Hazard. Mater. 2010, 176 (13), 495502; DOI: 10.1016/j.jhazmat.2009.11.057. (5) Yu, H.; Chen, S.; Quan, X.; Zhao, H.; Zhang, Y. Fabrication of a TiO2BDD heterojunction and its application as a photocatalyst for the simultaneous oxidation of an azo dye and reduction of Cr(VI). Environ. Sci. Technol. 2008, 42 (10), 37913796; DOI: 10.1021/es702948e. (6) Luo, S.; Xiao, Y.; Yang, L.; Liu, C.; Su, F.; Li, Y. Simultaneous detoxification of hexavalent chromium and acid orange 7 by a novel Au/ TiO2 heterojunction composite nanotube arrays. Sep. Purif. Technol. 2011, 79 (1), 8591; DOI: 10.1016/j.seppur.2011.03.019. (7) Yang, L.; Xiao, Y.; Liu, S.; Li, Y.; Cai, Q.; Luo, S. Photocatalytic reduction of Cr(VI) on WO3 doped long TiO2 nanotube arrays in the 9329
dx.doi.org/10.1021/es202012b |Environ. Sci. Technol. 2011, 45, 9324–9331
Environmental Science & Technology presence of citric acid. Appl. Catal., B 2010, 94 (12), 142149; DO:I 10.1016/j.apcatb.2009.11.002. (8) Sun, B.; Reddy, E. P.; Smirniotis, P. G. Visible light Cr(VI) reduction and organic chemical oxidation by TiO2 photocatalysis. Environ. Sci. Technol. 2005, 39 (16), 62516259; DOI: 10.1021/es0480872. (9) Kim, G.; Choi, W. Charge-transfer surface complex of EDTATiO2 and its effect on photocatalysis under visible light. Appl. Catal., B 2010, 100 (12), 7783; DOI: 10.1016/j.apcatb.2010.07.014. (10) Wang, N.; Zhu, L.; Deng, K.; She, Y.; Yu, Y.; Tang, H. Visible light photocatalytic reduction of Cr(VI) on TiO2 in situ modified with small molecular weight organic acids. Appl. Catal., B 2010, 95 (34), 400407; DOI: 10.1016/j.apcatb.2010.01.019. (11) Zhang, Y. C.; Du, Z. N.; Li, K. W.; Zhang, M.; Dionysiou, D. D. High-performance visible-light-driven SnS2/SnO2 nanocomposite photocatalyst prepared via in situ hydrothermal oxidation of SnS2 nanoparticles. ACS Appl. Mater. Interfaces 2011, 3 (5), 15281537; DOI: 10.1021/am200102y. (12) Yang, C.; Wang, W.; Shan, Z.; Huang, F. Preparation and photocatalytic activity of high-efficiency visible-light-responsive photocatalyst SnSx/TiO2. J. Solid State Chem. 2009, 182 (4), 807812; DOI: 10.1016/j.jssc.2008.12.018. (13) Xie, Y.; Ali, G.; Yoo, S. H.; Cho, S. O. Sonication-assisted synthesis of CdS quantum-dot-sensitized TiO2 nanotube arrays with enhanced photoelectrochemical and photocatalytic activity. ACS Appl. Mater. Interfaces 2010, 2 (10), 29102914; DOI: 10.1021/am100605a. (14) Li, W.; Li, D.; Meng, S.; Chen, W.; Fu, X.; Shao, Y. Novel approach to enhance photosensitized degradation of rhodamine B under visible light irradiation by the ZnxCd1‑xS/TiO2 nanocomposites. Environ. Sci. Technol. 2011, 45 (7), 29872993; DOI: 10.1021/es103041f. (15) Guo, Y.; Wang, L.; Yang, L.; Zhang, J.; Jiang, L.; Ma, X. Optical and photocatalytic properties of arginine-stabilized cadmium sulfide quantum dots. Mater. Lett. 2011, 65 (3), 486489; DOI: 10.1016/j. matlet.2010.10.057. (16) Zhang, Y. C.; Du, Z. N.; Li, S. Y.; Zhang, M. Novel synthesis and high visible light photocatalytic activity of SnS2 nanoflakes from SnCl2 3 2H2O and S powders. Appl. Catal., B 2010, 95 (12), 153159; DOI: 10.1016/j.apcatb.2009.12.022. (17) Zhang, Y. C.; Du, Z. N.; Li, K. W.; Zhang, M. Size-controlled hydrothermal synthesis of SnS2 nanoparticles with high performance in visible light-driven photocatalytic degradation of aqueous methyl orange. Sep. Purif. Technol. 2011, 81(1), 101107; DOI: 10.1016/j. seppur.2011.07.016. (18) Lei, Y.; Song, S.; Fan, W.; Xing, Y.; Zhang, H. Facile synthesis and assemblies of flowerlike SnS2 and In3+-doped SnS2: Hierarchical structures and their enhanced photocatalytic property. J. Phys. Chem. C 2009, 113 (4), 12801285; DOI: 10.1021/jp8079974. (19) Wang, C.; Zhang, H.; Li, F.; Zhu, L. Degradation and mineralization of bisphenol A by mesoporous Bi2WO6 under simulated solar light irradiation. Environ. Sci. Technol. 2010, 44 (17), 68436848; DOI: 10.1021/es101890w. (20) Idris, A.; Hassan, N.; Rashid, R.; Ngomsik, A. F. Kinetic and regeneration studies of photocatalytic magnetic separable beads for chromium (VI) reduction under sunlight. J. Hazard. Mater. 2011, 186 (1), 629635; DOI: 10.1016/j:jhazmat.2010.11.101. (21) Uchiyama, H.; Shirai, Y.; Kozuka, H. Formation of spherical SnO2 particles consisting of nanocrystals from aqueous solution of SnCl4 containing citric acid via hydrothermal process. J. Cryst. Growth 2011, 319 (1), 7078; DOI: 10.1016/j.jcrysgro.2011.02.002. (22) Reddy, N. K.; Devika, M.; Ahsanulhaq, Q. Growth of orthorhombic SnS nanobox structures on seeded substrates. Cryst. Growth Des. 2010, 10 (11), 47694772; DOI: 10.1021/cg100621d. (23) Tsukimura, K.; Suzuki, M.; Suzuki, Y.; Murakami, T. Kinetic theory of crystallization of nanoparticles. Cryst. Growth Des. 2010, 10 (8), 35963607; DOI: 10.1021/cg100488t. (24) Vekilov, P. G. Nucleation. Cryst. Growth Des. 2010, 10 (12), 50075019; DOI: 10.1021/cg1011633. (25) He, H.; Gao, C. Supraparamagnetic, conductive, and processable multifunctional graphene nanosheets coated with high-density
ARTICLE
Fe3O4 nanoparticles. ACS Appl. Mater. Interfaces 2010, 2 (11), 32013210; DOI: 10.1021/am.100673g. (26) Jimenez, J. A.; Sendova, M.; Sendova-Vassileva, M. Real-time monitoring of plasmonic evolution in thick Ag:SiO2 films: Nanocomposite optical tuning. ACS Appl. Mater. Interfaces 2011, 3 (2), 447454; DOI: 10.1021/am101021a. (27) Guo, C.; Ge, M.; Liu, L.; Gao, G.; Feng, Y.; Wang, Y. Directed synthesis of mesoporous TiO2 microspheres: Catalysts and their photocatalysis for bisphenol A degradation. Environ. Sci. Technol. 2010, 44 (1), 419425; DOI: 10.1021/est9019854. (28) Wang, Y.; Shi, J. C.; Cao, J. L.; Sun, G.; Zhang, Z. Y. Synthesis of Co3O4 nanoparticles via the CTAB-assisted method. Mater. Lett. 2011, 65 (2), 222224; DOI: 10.1016/j.matlet.2010.09.090. (29) Guo, G.; Huang, J. Preparation of mesoporous tantalum oxide and its enhanced photocatalytic activity. Mater. Lett. 2011, 65 (1), 6466; DOI: 10.1016/j.matlet.2010.09.027. (30) Araujo, P. Z.; Luca, V.; Bozzano, P. B.; Bianchi, H. L.; Soler-Illia, G. J. A. A. Aerosol-assisted production of mesoporous titania microspheres with enhanced photocatalytic activity: The basis of an improved process. ACS Appl. Mater. Interfaces 2010, 2 (6), 16631673; DOI: 10.1021/am100188q. (31) Feng, Y.; Li, L.; Ge, M.; Guo, C.; Wang, J.; Liu, L. Improved catalytic capability of mesoporous TiO2 microspheres and photodecomposition of toluene. ACS Appl. Mater. Interfaces 2010, 2 (11), 31343140; DOI: 10.1021/am100620f. (32) Wodka, D.; Biela nska, E.; Socha, R. P.; Wodka, M. E.; Gurgul, J.; Nowak, P. Photocatalytic activity of titanium dioxide modified by silver nanoparticles. ACS Appl. Mater. Interfaces 2010, 2 (7), 19451953; DOI: 10.1021/am1002684. (33) Zhang, L.; Cao, X. F.; Chen, X. T.; Xue, Z. L. BiOBr hierarchical microspheres: Microwave-assisted solvothermal synthesis, strong adsorption and excellent photocatalytic properties. J. Colloid Interface Sci. 2011, 354 (2), 630636; DOI: 10.1016/j.jcis2010.11.042. (34) Spadavecchia, F.; Cappelletti, G.; Ardizzone, S.; Bianchi, C. L.; Cappelli, S.; Oliva, C. Solar photoactivity of nano-N-TiO2 from tertiary amine: role of defects and paramagnetic species. Appl. Catal., B 2010, 96 (34), 314322; DOI: 10.1016/j.apcatb.2010.02.027. (35) Liao, Y.; Que, W.; Zhong, P.; Zhang, J.; He, Y. A facile method to crystallize amorphous anodized TiO2 nanotubes at low temperature. ACS Appl. Mater. Interfaces 2011, 3 (7), 28002804; DOI: 10.1021/ am200685s. (36) Wang, L.; Wang, N.; Zhu, L.; Yu, H.; Tang, H. Photocatalytic reduction of Cr(VI) over different TiO2 photocatalysts and the effects of dissolved organic species. J. Hazard. Mater. 2008, 152 (1), 9399; DOI: 10.1016/j.jhazmat.2007.06.063. (37) Waldmann, N. S.; Paz, Y. Photocatalytic reduction of Cr(VI) by titanium dioxide coupled to functionalized CNTs: An example of counterproductive charge separation. J. Phys. Chem. C 2010, 114 (44), 1894618952; DOI: 10.1021/jp105925g. (38) Parida, K. M.; Sahu, N. Visible light induced photocatalytic activity of rare earth titania nanocomposites. J. Mol. Catal. A: Chem. 2008, 287 (12), 151158; DOI: 10.1016/j.molcata.2008.02.028. (39) Nasrallah, N.; Kebir, M.; Koudri, Z.; Trari, M. Photocatalytic reduction of Cr(VI) on the novel hetero-system CuFe2O4/CdS. J. Hazard. Mater. 2011, 185 (23), 13981404; DOI: 10.1016/j. jhazmat.2010.10.061. (40) Qamar, M.; Gondal, M. A.; Yamani, Z. H. Laser-induced efficient reduction of Cr(VI) catalyzed by ZnO nanoparticles. J. Hazard. Mater. 2011, 187 (13) 258263; DOI: 10.1016/j.jhazmat.2011.01.007. (41) Jiang, F.; Zheng, Z.; Xu, Z.; Zheng, S.; Guo, Z.; Chen, L. Aqueous Cr(VI) photo-reduction catalyzed by TiO2 and sulfated TiO2. J. Hazard. Mater. 2006, B134 (13), 94103; DOI: 10.1016/j.jhazmat.2005.10.041. (42) Gherbi, R.; Nasrallah, N.; Amrane, A.; Maachi, R.; Trari, M. Photocatalytic reduction of Cr(VI) on the new hetero-system CuAl2O4/ TiO2. J. Hazard. Mater. 2011, 186 (23), 11241130; DOI: 10.1016/j. jhazmat.2010.11.105. (43) Ma, D.; Zhou, H.; Zhang, J.; Qian, Y. Controlled synthesis and possible formation mechanism of leaf-shaped SnS2 nanocrystals. Mater. 9330
dx.doi.org/10.1021/es202012b |Environ. Sci. Technol. 2011, 45, 9324–9331
Environmental Science & Technology
ARTICLE
Chem. Phys. 2008, 111 (23), 391395; DOI: 10.1016/j.matchemphys. 2008.04.035. (44) Yang, Q.; Tang, K.; Wang, C.; Zhang, D.; Qian, Y. The synthesis of SnS2 nanoflakes from tetrabutyltin precursor. J. Solid State Chem. 2002, 164 (1), 106109; DOI: 10.1006/jssc.2001.9453. (45) Chenthamarakshan, C. R.; Rajeshwar, K.; Wolfrum, E. J. Heterogeneous photocatalytic reduction of Cr(VI) in UV-irradiated titania suspensions: Effect of protons, ammonium ions, and other interfacial aspects. Langmuir 2000, 16 (6), 27152721; DOI: 10.1021/la9911483. (46) Wang, N.; Xu, Y.; Zhu, L.; Shen, X.; Tang, H. Reconsideration to the deactivation of TiO2 catalyst during simultaneous photocatalytic reduction of Cr(VI) and oxidation of salicylic acid. J. Photochem. Photobiol. A. 2009, 201 (23), 121127; DOI: 10.1016/j.photochem. 2008.10.002.
9331
dx.doi.org/10.1021/es202012b |Environ. Sci. Technol. 2011, 45, 9324–9331
ARTICLE pubs.acs.org/est
Advanced Oxidation Process Based on the Cr(III)/Cr(VI) Redox Cycle Alok D. Bokare and Wonyong Choi* School of Environmental Science and Engineering, Pohang University of Science and Technology (POSTECH), Pohang 790-784, Korea
bS Supporting Information ABSTRACT: Oxidative degradation of aqueous organic pollutants, using 4-chlorophenol (4-CP) as a main model substrate, was achieved with the concurrent H2O2-mediated transformation of Cr(III) to Cr(VI). The Fenton-like oxidation of 4-CP is initiated by the reaction between the aquo-complex of Cr(III) and H2O2, which generates HO• along with the stepwise oxidation of Cr(III) to Cr(VI). The Cr(III)/H2O2 system is inactive in acidic condition, but exhibits maximum oxidative capacity at neutral and near-alkaline pH. Since we previously reported that Cr(VI) can also activate H2O2 to efficiently generate HO•, the dual role of H2O2 as an oxidant of Cr(III) and a reductant of Cr(VI) can be utilized to establish a redox cycle of Cr(III)Cr(VI)Cr(III). As a result, HO• can be generated using both Cr(III)/H2O2 and Cr(VI)/H2O2 reactions, either concurrently or sequentially. The formation of HO• was confirmed by monitoring the production of p-hydroxybenzoic acid from [benzoic acid + HO•] as a probe reaction and by quenching the degradation of 4-CP in the presence of methanol as a HO• scavenger. The oxidation rate of 4-CP in the Cr(III)/H2O2 solution was highly influenced by pH, which is ascribed to the hydrolysis of CrIII(H2O)n into CrIII(H2O)n‑m(OH)m and the subsequent condensation to oligomers. The present study proposes that the Cr(III)/H2O2 combined with Cr(VI)/H2O2 process is a viable advanced oxidation process that operates over a wide pH range using the reusable redox cycle of Cr(III) and Cr(VI).
’ INTRODUCTION Advanced oxidation processes (AOPs) using hydrogen peroxide (H2O2) as a precursor of hydroxyl radical (HO•) have emerged as efficient technologies for the rapid destruction of recalcitrant organic pollutants.1 The nonselective reactivity of HO• toward most organic pollutants with near diffusion-limited bimolecular rate constants (108109 M1 s1), combined with the easy availability (million metric ton-scale), low price (ca. 1.0 $/kg of 100% H2O2), and environmentally benign nature of H2O2, facilitates large-scale applications.2 The success of H2O2based oxidation processes depends critically on the choice of reagent used to enhance the formation of HO• from H2O2 decomposition. Transition metal ions (Fe2+, Fe3+, Cu2+) have been extensively used in classical or modified Fenton (photoFenton and electro-Fenton) and Fenton-like reactions for the oxidation of various organic contaminants.38 However, the fact that the active metal species is consumed as a reagent and lost through precipitation severely limits the process efficiency. As a result, the continuous supply of metal reagent is needed to sustain the activation of H2O2, which causes the problem of metal sludge. Heterogeneous transition metal catalysts may provide an alternative solution for such problems but suffer from mass transfer limitation and metal leaching.9 Therefore, the ideal process of the metal-induced decomposition of H2O2 should require the regeneration of the active metal species through a redox cycle. To achieve this objective, the redox states of the involved metal species should be stable over a wide r 2011 American Chemical Society
pH range. Compared to iron and copper, chromium exists in a wider range of oxidation states (from 2 to +6), with the trivalent [Cr(III)] and hexavalent [Cr(VI) or chromate] species commonly found in water. Being an oxyanion, chromate is completely soluble over the entire pH range.10 However, due to its extreme toxicity and carcinogenecity, any deliberate addition of Cr(VI) as a reagent into wastewaters is not sensible, even if various physicochemical or biological post-treatments11 can easily remove it from aqueous solution. Trivalent chromium [Cr(III)], on the other hand, is the most thermodynamically stable oxidation state of chromium, kinetically inert, and significantly less toxic. Although the two chromium species (CrVI vs CrIII) are characterized by different chemical behavior, bioavailability, and toxicity,12 they are readily interconverted in aqueous solution. Cr(VI) is a strong oxidant [E0(HCrO4/Cr3+ = 1.35 VNHE)]13 and reacts rapidly with numerous reducing agents (like Fe0, Fe2+, S2‑, and natural organic matter) to form Cr(III).14 On the other hand, Cr(III) is thermodynamically stable under reducing conditions and is oxidized to Cr(VI) by Mn(III,IV) (hydr)oxides15 or photo-oxidized by FeOH2+.16 H2O2 alone can interconvert Cr(III) and Cr(VI) into each other because of its Received: June 25, 2011 Accepted: September 22, 2011 Revised: September 21, 2011 Published: October 11, 2011 9332
dx.doi.org/10.1021/es2021704 | Environ. Sci. Technol. 2011, 45, 9332–9338
Environmental Science & Technology
ARTICLE
Scheme 1. Schematic Illustration of HO• Generation from H2O2 using Cr(III)Cr(VI) Redox Cyclea
a
The numbered paths indicate the following: (1) hydrolysis of Cr(III)aquocomplex; (2) Fenton-like oxidation of Cr(III) to Cr(VI) by H2O2; (3) Cr(VI)-mediated decomposition of H2O2 via dissociation of Cr(V)peroxo complex (demonstrated in the previous work20); (4) regeneration of Cr(III) by H2O2-mediated reduction of Cr(VI).
ability to act as both an oxidizing agent [E0(H2O2/H2O) = 1.77 V] and a reducing agent [E0(O2/H2O2) = 0.68 V].17 The pe-pH relationship of Cr(VI)/Cr(III) and O2/H2O2 couples indicates that H2O2 can oxidize Cr(III) at pH > 8 and reduce Cr(VI) at lower pH.18 Since the reducing strength of H2O2 strongly increases with decreasing pH, the H2O2-induced reduction of Cr(VI) to Cr(III) at pH < 3 is used for removing chromate from wastewaters.19 In our previous work,20 we demonstrated that Cr(VI) can also activate H2O2 and generate HO• for the oxidative degradation of aqueous organic pollutants although the toxicity of Cr(VI) limits practical applications only to the degradation of organics in chromate-contaminated wastewaters. The oxidation mechanism involves the formation of tetraperoxochromate(V) complex and works over a wide range of pH 311. However, the previous method of H2O2 activation can utilize only Cr(VI), not Cr(III). The present study successfully demonstrates that Cr(III) can also activate H2O2 to generate HO• along with the stepwise oxidation of Cr(III) to Cr(VI). This enables a redox cycling of Cr(III)/Cr(VI) by H2O2 that serves as both an oxidant of Cr(III) and a reductant of Cr(VI). As a result, a new AOP that generates HO• repeatedly based on the Cr(III)/Cr(VI) redox cycle is developed (see Scheme 1). We also demonstrate that the Cr(III)/Cr(VI) redox transformation can be easily manipulated by H2O2 in pH-controlled reactions and H2O2 serves the dual roles of a precursor of HO• and an oxidant/reductant of Cr(III)/ Cr(VI). Through this reversible chromium catalytic cycle coupled with the decomposition of H2O2, the hydroxyl radicalmediated degradation of organic compounds can be achieved in repeated cycles in a single batch reactor.
’ EXPERIMENTAL SECTION Chemicals and Materials. Chemicals that were used as received in this study included chromium(III) nitrate (Sigma), sodium chromate (Sigma), hydrogen peroxide (30%, Kanto), 4-chlorophenol (4-CP, Sigma), phenol (Aldrich), aniline (Aldrich), nitrobenzene (Aldrich), benzoic acid (Aldrich), p-hydroxy
Figure 1. (a) Effect of initial pH on the degradation of 4-CP in the Cr(III)/H2O2 system. [4-CP]0 = 100 μM, [Cr(III)]0 = 2 mM, and [H2O2]0 = 20 mM. (b) Degradation of 4-CP and the concurrent generation of chloride and chromate (Cr(VI)) under the condition of (a) and pHi = 7. The degradation of 4-CP in the presence of Cr(III) only (no H2O2) or H2O2 only (no Cr(III)) is denoted by open triangle (r) and closed triangle (2), respectively.
benzoic acid (p-HBA, Aldrich), β-cyclodextrin hydrate (Aldrich), methanol (Daejung), and acetonitrile (Merck). All solutions were prepared in ultrapure water (18 MΩ cm) obtained from a Barnstead purification system. Procedure and Analytical Methods. The reactions were carried out in 50-mL glass beakers stirred on a magnetic stirrer. An aliquot of stock solution of 4-CP (or other substrate, 1 mM) was added to make a desired concentration (typically 0.1 mM), and then the reaction was initiated by the sequential addition of Cr(III) and H2O2. Unless otherwise mentioned, [Cr(III)] was fixed at 2 mM. The initial pH (pHi) of the solution was adjusted to a desired value with 1 N NaOH standard solution. Sample aliquots (1 mL) were withdrawn at regular time intervals from the reactor and injected into 4-mL glass vials containing 50 μL of sodium sulfite (Na2SO3, 2 M) to quench residual H2O2. All experiments were carried out in triplicate for a given condition. Quantitative analysis of substrates was done using a highperformance liquid chromatograph (HPLC Agilent 1100) equipped with a C-18 column (Agilent Zorbax 300SB) and a diode-array detector. The eluent compositions were as follows: (a) 0.1% phosphoric acid aqueous solution and acetonitrile 9333
dx.doi.org/10.1021/es2021704 |Environ. Sci. Technol. 2011, 45, 9332–9338
Environmental Science & Technology
ARTICLE
Figure 2. Degradation of phenol, nitrobenzene, and aniline (separate single-component experiments) in the presence of Cr(III) and H2O2 ([Cr(III)]0 = 2 mM, [H2O2]0 = 20 mM, [substrate]0 = 100 μM, and pHi = 7).
(80:20 v/v) for 4-CP, (b) water, acetonitrile, and acetic acid (78:20:2 v/v) for phenol, (c) water and methanol (50:50 v/v) for nitrobenzene and aniline, and (d) 0.1% phosphoric acid aqueous solution and acetonitrile (85:15 v/v) for benzoic acid. Quantification of ionic intermediates/products was performed using an ion chromatograph (IC, Dionex DX-120) equipped with Dionex IonPac AS-14 column and a conductivity detector. The eluent composition was 3.5 mM Na2CO3 + 1 mM NaHCO3. Cr(VI) concentration was determined using a modified diphenylcarbazide (DPC) method. H2O2 interferes in the Cr(VI) determination by the standard DPC method21 due to its ability to rapidly reduce Cr(VI) to Cr(III) in acidic solution. In the modified method,22 2-mL sample aliqouts were quenched by sequential addition of 0.5 mL of DPC in acetone (20 g/L) followed by the addition of 0.2 mL of 9 M H2SO4. The absorbance of the colored Cr-DPC complex was analyzed spectrophotometrically at 540 nm within 5 min of the color development. Total organic carbon (TOC) was measured using a TOC analyzer (TOCVCSH, Shimadzu). Various species of Cr(III) aquo-complexes were analyzed by forming their inclusion complexes with β-cyclodextrin (β-CD), which were then determined by matrix-assisted laser desorption and ionization time-of-flight mass spectrometry (MALDI-TOF MS, Bruker REFLEX III). The inclusion complexes were prepared by mixing Cr(III) and β-CD at 1:1 molar ratio and adjusting the pH to the desired value with 1 N NaOH. The matrix used for the MALDI-TOF experiments was α-cyano-4hydroxycinnamic acid (CHCA, Aldrich) dissolved in acetone at 80 mg/mL. The CHCA and Cr(III) + β-CD solutions were mixed at 4:1 ratio (matrix:analyte, v/v), and the mixed solution was dropped onto the MALDI plate and air-dried.
’ RESULTS AND DISCUSSION Oxidation in Cr(III)/H2O2 System. To evaluate the oxidative capacity of the Cr(III)/H2O2 system, 4-CP degradation in aqueous solution was investigated at different pHi under airequilibrated conditions. As shown in Figure 1a, 4-CP degradation was completely inhibited at pH 3, but increased with increasing
Figure 3. (a) Comparison of the time profiles of p-HBA formation during the oxidation of benzoic acid (BA) in the Cr(III)/H2O2 system. [BA]0 = 10 mM, [Cr(III)]0 = 2 mM, and [H2O2]0 = 20 mM. (b) Effect of methanol (OH radical scavenger) on the degradation of 4-CP in the Cr(III)/H2O2 system. [4-CP]0 = 100 μM, [Cr(III)]0 = 2 mM, [H2O2]0 = 20 mM, [CH3OH]0 = 100 mM, and pHi = 7.
pH leading to complete degradation in 6 h at pH 7. However, with further increase in pHi, the 4-CP removal rate decreased. The complete absence of 4-CP oxidation in acidic condition (pH < 4) can be attributed to the kinetic inertness of the Cr(III) aquocompexes. At pH < 4, Cr(III) exists as the hexaaquo complex [Cr(H2O)6]3+ (pKa = 4, see Supporting Information Figure S1) and its reaction with any organic or inorganic species (H2O2 in the present case) involves the substitution of the coordinated water molecules by the reactant. However, the extremely low frequency of water exchange within Cr(III)-aquocomplex (∼106.3 s1 or half-life ∼40 h)23 makes the aquo-complex substitutionally inert and unreactive toward H2O2 in the present experimental time scale. For comparison, the corresponding water exchange frequencies for Al(III) and Fe(III) are 100.8 s1 and 103.5 s1, respectively.23 However, when pHi increased to 7, a complete removal of 4-CP was obtained. Cr(III) at neutral pH is neither an oxidant nor a reductant and exists solely as insoluble Cr(OH)3(s). To confirm whether 4-CP was removed from aqueous solution 9334
dx.doi.org/10.1021/es2021704 |Environ. Sci. Technol. 2011, 45, 9332–9338
Environmental Science & Technology
ARTICLE
through simple adsorption on Cr(OH)3, control degradation experiments were carried out only in the presence of Cr(III) at pH 7 (without H2O2). The removal of 4-CP was not observed at all in the absence of H2O2 (Figure 1b), which ruled out the possibility of 4-CP removal by adsorption. Moreover, the concurrent production of chloride ions (Figure 1b) accounted for 90% of the removed 4-CP. This implies that a reaction between Cr(III) and H2O2 generates a reactive species that is responsible for the degradation of 4-CP. TOC was also reduced by 33((4)% in 6 h reaction, which indicates that some fraction of 4-CP was actually mineralized. The minor deficit in chloride mass balance (∼10%) may be ascribed to the generation of chlorinated intermediates that were not determined in this work. This clearly demonstrates that 4-CP can be oxidatively degraded at neutral pH in the Cr(III)/H2O2 system. Oxidation of other organic pollutants such as phenol [34((2)% TOC reduction], nitrobenzene [35((1)% TOC reduction], and aniline [25((4)% TOC reduction] was also successfully achieved at neutral pH (Figure 2). To ascertain whether 4-CP is oxidized by hydroxyl radicals generated through Cr(III)-mediated decomposition of H2O2, the oxidative conversion of benzoic acid (BA) to p-hydroxybenzoic acid (p-HBA) was used as a probe reaction. The formation of p-HBA from the reaction of (BA + HO•) has been used as an indirect method to detect HO• formation.20,24 Figure 3a shows the production of p-HBA from BA in the Cr(III)/H2O2 system at different pHi. The formation of p-HBA is completely inhibited at pHi 3 but increases with increasing pH, which is similar to the pH-dependent behavior of 4-CP degradation (see Figure 1). This indicates that HO• generated from the reaction of Cr(III) and H2O2 is initiated only at pH g 5, which corroborates the electron paramagnetic resonance (EPR) study of Shi et al.25 who reported the formation of HO• in neutral (pH = 7.2) solution of Cr(III)/H2O2 but not under acidic (pH = 3) condition. Furthermore, the addition of methanol as a hydroxyl radical scavenger completely inhibited the degradation of 4-CP at neutral pH (Figure 3b). This confirms that H2O2 acts as a precursor of hydroxyl radicals, which are primarily responsible for 4-CP oxidation in the presence of Cr(III). The reaction between Cr(III) and H2O2 system generates HO• through a Fenton-like reaction with the simultaneous formation of intermediate Cr(IV) species.26 CrðIIIÞ þ H2 O2 f CrðIVÞ þ HO• þ OH
ð1Þ
•
Cr(IV) immediately generates another HO from H2O2 (reaction 2)27 or undergoes disproportionation to generate Cr(V) species (reaction 3).27 Cr(V) induces another Fentonlike reaction (reaction 4)28 with further generation of HO•. CrðIVÞ þ H2 O2 f CrðVÞ þ HO• þ OH
ð2Þ
2CrðIVÞ f CrðVÞ þ CrðIIIÞ
ð3Þ
CrðVÞ þ H2 O2 f CrðVIÞ þ HO• þ OH
ð4Þ
The generation of Cr(VI) during 4-CP oxidation via stepwise oxidation of Cr(III) is shown in Figure 1b. Thus, the Cr(III)/ H2O2 system generates HO• via a series of Fenton-like reactions involving intermediate Cr(IV) and Cr(V) species,29 leading to the transformation of Cr(III) into Cr(VI). However, it should be realized that the actual oxidation chemistry can be more complex. It may be possible that the intermediate Cr(IV) and
Table 1. Proposed Chemical Structures of Cr(III)β-CD Inclusion Complexes Identified by MALDI-TOF at Different pHi m/z
7
1272.6
[Cr2(μOH)2(H2O)8]4+β-CD
1311.7
[Cr2(μOH)2(H2O)7(OH)]3+β-CDNa+
1273.8
[Cr2(μOH)2(H2O)8]4+β-CD
1312.2
[Cr2(μOH)2(H2O)7(OH)]3+β-CDNa+
1332.6 1358.4
[Cr2(μOH)2(H2O)6(OH)2]2+β-CDNa+ [Cr3(μOH)4(H2O)9]5+β-CD
1359.2
[Cr3(μOH)4(H2O)9]5+β-CD
9
11 a
proposed complex compositiona
pHi
Na β-CD (m/z = 1158) adducts were observed in all samples. At pHi e 5, only the Na+ adducts were detected. +
Cr(V) species take part in the direct oxidation of 4-CP and its intermediates, which should complicate the overall redox chemistry. The formation of a stable Cr(V)-complex is possible in the presence of organic substrates with ligand groups (e.g., hydroxycarboxylate and 1,2-diol moieties in natural organic matters).30 Moreover, EPR studies have also suggested the formation of stable Cr(V)-peroxo complexes in the Cr(VI)/H2O2 system.31,32 Because these intermediate chromium complexes should influence the oxidation kinetics and mechanisms, the proposed path in Scheme 1 should be taken as a simplified representation of the complex redox process in the Cr(III)Cr(VI)H2O2 system. pH-Dependent Speciation of Cr(III) and the Reactivity for H2O2 Activation. The degradation rate of 4-CP increases with pH above 3 but subsequently decreases beyond neutral pH condition (see Figure 1). When pH increases toward near neutral and alkaline values, the hexaaquo ions are hydrolyzed to hydroxocomplexes (reaction 5). ½CrðH2 OÞ6 3þ f ½CrðH2 OÞ5 ðOHÞ2þ þ Hþ
ð5Þ
This Cr(III) monomeric hydroxo-complex subsequently undergoes hydrolytic condensation to form polynuclear complexes like dimer, trimer, and higher oligomers containing μ-hydroxo bridges between adjacent chromium atoms (see Supporting Information Figure S2). These soluble hydroxo-complexes are eventually polymerized and precipitated as Cr(OH)3(s). However, the hydrolytic conversion of Cr(III)-oligomers into solid Cr(OH)3 occurs sufficiently slowly (>1 yr) to permit separation and isolation of a series of individual oligomers up to a hexamer.33,34 This means that oligomers formed through Cr(III) hydrolysis should be long-lived enough to react with H2O2. The pH-dependent formation of Cr(III) oligomers, however, strongly influences the reaction kinetics with H2O2. Rao et al.35 demonstrated that the reaction rate constant (k) obtained with isolated oligomers decreases as oligomerization proceeds (kdimer > ktrimer). The rate constant for unseparated oligomers in solution (analogous to the present study) was 2 orders of magnitude slower than for the isolated dimer. Thus, the increase in 4-CP oxidation with increasing pH from 3 to 7 indicates that Cr(III) hydrolysis (or oligomerization) initiates the decomposition of H2O2 to generate HO•. On the other hand, the decrease in oxidation efficiency in alkaline condition suggests the formation of higher oligomers, which are less reactive toward H2O2 decomposition.35 To confirm the presence of Cr(III) oligomers and determine the degree of oligomerization at different pH, β-CD was used as a complexing ligand to form inclusion complexes with the 9335
dx.doi.org/10.1021/es2021704 |Environ. Sci. Technol. 2011, 45, 9332–9338
Environmental Science & Technology oligomers. β-CD is a torus-shaped cyclic oligosaccharide with an internal hydrophobic cavity and contains seven α-D-glucopyranose units linked together by α-1,4-glycosidic linkages. It has been extensively used for molecular recognition and as metal ion receptor in hostguest systems.36 Using the oxygen atoms on adjacent pyranose rings as a diol ligand (see Supporting Information, Figure S3), β-CD can form highly stable inclusion complexes with binuclear hydroxy-bridged metal structures (structurally similar to Cr(III) oligomers).37,38 This complexation property of β-CD can be used to isolate different Cr(III) oligomers by using the internal cavity as an efficient trapping site. Table 1 shows the MALDI-TOF analysis of Cr(III) aqueous solutions containing β-CD at different pH. At pH e 5, only the Na+ and K+ adducts of β-CD were detected at m/z = 1158 and 1174, respectively. The absence of β-CD complexes with Cr(III) at pH e 5 can be attributed to the weak interaction between unmodified β-CD and the metal ion. Transition metal cations can form stable complexes only with surface functionalized-βCD, wherein, the metal ion is complexed with ligands situated outside the CD cavity.37 Thus, Cr(III) monomeric species cannot form inclusion complexes with unmodified β-CD and, hence, were not detected in the MALDI-TOF analysis. However, with increasing pH, the degree of oligomerization increased along with the hydrolysis of Cr(III), which is evident from the detection of β-CD inclusion complexes with dimer [Cr2(μOH)2(H2O)8]4+, trimer [Cr3(μOH)4(H2O)9]5+, and intermediate hydroxo-bridged species (see Table 1). At pH 11, only the trimer complex was detected, which indicates that the dimer was involved in formation of higher oligomers (most probably tetramer).39 The absence of MALDI-TOF peaks corresponding to tetramer and/or higher oligomers may be attributed to their larger molecular size, which cannot fit inside the β-CD internal cavity. The MALDI-TOF analysis confirms that Cr(III) oligomerization is initiated at pH > 5 and the subsequent pH increase leads to the formation of higher oligomers. The reactivity of these different Cr(III) hydrolytic species can be correlated to the pH-dependent 4-CP oxidation behavior. In Fenton-like reactions involving H2O2 and transition metal complexes, the oxidation of the metal center is promoted by the formation of a metalhydroperoxo complex intermediate (via ligand exchange), followed by the homolytic cleavage of the peroxo bond to generate HO•.40 In the case of Cr(III)-mediated Fenton-like reaction, the rate of water exchange in Cr(H2O)63+ is too slow (∼106.3 s1) to form a hydroperoxo complex, which explains the absence of 4-CP oxidation at pH 3 (see Figure 1). However, at pH 5, the monohydroxy complex (H2O)5CrOH2+ (pKa = 6.1)34 is the dominant Cr(III) species (see Supporting Information, Figure S1, S2). The complex of (H2O)5CrOH2+ is 75 times more reactive in water exchange reaction41,42 and 605500 times more reactive in anion complexation reaction,43 compared to Cr(H2O)63+. Thus, the hydrolysis of the Cr(III) aquocomplex at pH 5 can facilitate the peroxo ligand substitution, and initiate HO• generation for the oxidation of 4-CP. At pH > 5, the (H2O)5CrOH2+ species can be sequentially deprotonated, and then oligomerized. The rapid oligomerization competes with and mostly dominates stepwise deprotonation reactions.42 At pH 7, the deprotonated dimer [Cr2(μOH)2(H2O)6(OH)]3+ was isolated and identified by MALDI-TOF analysis (see Table 1). The water-exchange rate of this deprotonated dimer is 27 times (for cis form) or 70 times (for trans form) higher compared to the (H2O)5CrOH2+ species.44 Thus, the
ARTICLE
Figure 4. (a) Repeated cycles of 4-CP degradation and the concurrent generation of Cr(VI) in the presence of Cr(III) and H2O2 (pHi = 7). At the end of each cycle, an aliquot of HClO4 (1 N, 1 mL) was added to regenerate Cr(III) (at the point of 1), then 4-CP (3 mM, 0.9 mL) and H2O2 (20 mM) were replenished (at the point of 2), and finally the pH of the solution was readjusted to 7 before initiating the next degradation cycle. (b) The initial cycle of 4-CP degradation ([Cr(III)]0 = 2 mM, [H2O2]0 = 20 mM, pHi = 7) and the subsequent cycle of 4-CP degradation without the regeneration of Cr(III) and without pH readjustment. The time profile of pH change is shown together.
peroxo ligand exchange reaction (hence the generation of HO•) will be enhanced when raising pH from 5 to 7, which is consistent with the faster oxidation of 4-CP at pH 7 than pH 5 (see Figure 1a and Figure 3a). When pH is further increased to alkaline values (pH 9), the OH-ligands are bridged through condensation. As a result, the concentration of oligomers containing nonbridging OH groups (species A and B in Supporting Information, Figure S4) will decrease with retarding the water exchange reaction,41 which subsequently leads to retardation of peroxo complexation and suppression of OH-radical mediated oxidation (see Figure 1). Furthermore, at pH 11, the complete absence of any species with nonbridging OH groups combined with the formation of higher oligomers (trimer and possibly tetramer) lowers the Cr(III) reactivity toward H2O2. Thus, Cr(III) species coordinated with OH ligands are required to catalyze the peroxo complexation mechanism for the generation of HO•. Cr(III)/Cr(VI) Redox Process as AOP. All practical applications of metal-catalyzed Fenton and Fenton-like AOPs are severely 9336
dx.doi.org/10.1021/es2021704 |Environ. Sci. Technol. 2011, 45, 9332–9338
Environmental Science & Technology
Figure 5. Effect of solution aging time on the degradation of 4-CP in the Cr(III)/H2O2 system. [Cr(III)]0 = 2 mM, [H2O2]0 = 20 mM, and pHi = 7.
limited by the fact that the precipitation of metal ions limits working conditions to acidic region,9 and prevents the reuse of the catalyst. The Cr(III)-mediated activation of H2O2 shows the maximum oxidation capacity at neutral and near-alkaline pH. Some hydrolytic oligomers are formed at neutral and alkaline pH and they remain soluble without precipitation and can activate H2O2 with less efficiency. This will reduce the need of adding a large amount of metal salt to compensate for the catalyst loss, and subsequently prevent the problem of sludge disposal. Furthermore, the resulting oxidation product, Cr(VI), is soluble over the entire pH range.10 However, the extreme toxicity of Cr(VI) is a major concern and its complete removal from the treated wastewater is essential. Cr(VI) can be reduced to Cr(III) by using H2O2 as a reductant in acidic condition with the concurrent generation of OH radicals.19,20 Therefore, Cr(III) species can be easily regenerated by simply decreasing pH to acidic values in the presence of H2O2. In this way, we can exploit the pH-dependent dual role of H2O2 as Cr(III) oxidant and Cr(VI) reductant to establish a cyclic redox transformation of chromium along with the generation of HO radicals. This makes the Cr(III)/ Cr(VI)/H2O2 system a new AOP based on the redox cycle of chromium species without the loss of active metal species. We have successfully established the process viability by sustaining the repeated cycles of 4-CP removal at neutral pH using Cr(III) regenerated from Cr(VI) prior to oxidation (Figure 4a). However, the inhibition of 4-CP oxidation under acidic condition requires the pH to be raised back to neutral before each successive oxidation cycle. This repeated addition of acid and base will increase the total ionic strength of the solution and the overall treatment cost. To minimize the salinity increase induced by repeated pH adjustments, Cr(VI)/H2O220 as well as Cr(III)/H2O2 process should be concurrently utilized to generate HO•. That is, as Cr(III) is depleted along with the generation of HO• in the Cr(III)/H2O2 process, the accompanying Cr(VI) can also generate HO• at the same time through the Cr(VI)/H2O2 process. It should be noted that the Cr(VI)/H2O2 and Cr(III)/H2O2 processes have complementary pH-dependence: the former favored at acidic pH but the latter inhibited in the acidic condition. We verified this dual process by achieving consecutive cycles of 4-CP oxidation without regenerating Cr(III) and without pH
ARTICLE
readjustment (Figure 4b). Because the pH decreased to around 4 after the first cycle, the Cr(III)/H2O2 reaction is inhibited in the second cycle. However, the Cr(VI)-induced activation of H2O2 is more efficient at acidic pH20 and hence 4-CP oxidation was achieved through the Cr(VI)/H2O2 process in the second cycle. That is, both Cr(III)/H2O2 and Cr(VI)/H2O2 can be utilized as an AOP that generates HO• and both processes can work either concurrently or sequentially (Scheme 1). The combination of Cr(VI)/H2O2 and Cr(III)/H2O2 processes should reduce the cost for pH adjustments. However, such sequential processes without additional pH adjustments cannot be efficiently sustained for further cycles since the pH gradually converges to a relatively higher pH (>5), at which the Cr(VI)/H2O2 process is very slow. Anyway, despite the additional cost needed for sequential pH adjustments, the Cr(III)/Cr(VI)/H2O2 process works over a wide pH range by recycling the active metal species and provides an advantage over the classical Fenton or Fentonlike process, which needs a strict acidic condition to prevent the iron loss by precipitation. Finally, it should be mentioned that the aging of Cr(III) solution is also an important factor to be considered. Figure 5 shows that the degradation of 4-CP was significantly retarded when using Cr(III) solutions aged at ambient temperature for 1 month. The oligomerization process is dependent on not only pH but also the aging time. Generally, the reactivity of Cr(III) aqueous solutions decreases with the aging time, which is attributed to the transformation of soluble Cr(III) hydrolytic species (monomer and oligomer) into insoluble polynuclear species and/or amorphous chromium oxyhydroxides.45 In the present study, although the 4-CP oxidation efficiency decreased as expected, a significant concentration of 4-CP could be still removed even after 30-day aging period. This AOP based on the redox cycle of Cr(III)/Cr(VI) with H2O2 can be versatilely applied to the degradation of organics through the generation of HO• in chromium-contaminated wastewaters regardless of the oxidation state of the chromium species. However, the practical implementation of this chromium-based AOP and its effectiveness will be influenced by various constituents in wastewaters, which may interfere with the catalytic cycle of Cr(III)/Cr(VI). More thorough studies are required to understand such interfering effects.
’ ASSOCIATED CONTENT
bS
Supporting Information. The pH-dependent speciation of aqueous Cr(III) and H2O2, schematic representation of Cr(III) hydrolysis and oligomerization reactions, chemical structure of β-cyclodextrin and comparative MALDI-TOF spectra at pHi = 7 and pHi = 9. This information is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +82-54-279-2283; fax: +82-54-279-8299; e-mail: wchoi@ postech.edu.
’ ACKNOWLEDGMENT This work was supported by KOSEF NRL program (R0A-2008000-20068-0), KOSEF EPB center (Grant R11-2008-052-02002), 9337
dx.doi.org/10.1021/es2021704 |Environ. Sci. Technol. 2011, 45, 9332–9338
Environmental Science & Technology and KCAP (Sogang Univ.) funded by MEST through NRF (NRF-2009-C1AAA001-2009-0093879).
’ REFERENCES (1) Gogate, P. R.; Pandit, A. B. A review of imperative technologies for wastewater treatment. I: Oxidation technologies at ambient conditions. Adv. Environ. Res. 2004, 8, 501–551. (2) M€agerlein, W.; Dreisbach, C.; Hugl, H.; Tse, M. K.; Klawonn, M.; Bhor, S.; Beller, M. Homogeneous and heterogeneous ruthenium catalysts in the synthesis of fine chemicals. Catal. Today 2007, 121, 140–150. (3) Pignatello, J. J.; Oliveros, E.; MacKay, A. Advanced oxidation processes for organic contaminant destruction based on the Fenton reaction and related chemistry. Crit. Rev. Environ. Sci. Technol. 2006, 36, 1–84. (4) Laine, D. F.; Cheng, I. F. The destruction of organic pollutants under mild reaction conditions: A review. Microchem. J. 2007, 85, 183–193. (5) Nam, S.; Renganathan, V.; Tratnyek, P. G. Substituent effects on azo dye oxidation by the FeIII-EDTA-H2O2 system. Chemosphere 2001, 45, 59–65. (6) Collins, T. J. TAML oxidant activators: A new approach to the activation of hydrogen peroxide for environmentally significant problems. Acc. Chem. Res. 2002, 35, 782–90. (7) Lin, T. Y.; Wu, C. H. Activation of hydrogen peroxide in copper(II)/amino acid/H2O2 systems: Effects of pH and copper speciation. J. Catal. 2005, 232, 117–126. (8) Shah, V.; Verma, P.; Stopka, P.; Gabriel, J.; Baldrian, P.; Nerud, F. Decolorization of dyes with copper(II)/organic acid/hydrogen peroxide systems. Appl. Catal., B 2003, 46, 287–292. (9) Pestunova, O. P.; Ogorodnikova, O. L.; Parmon, V. N. Studies on the phenol wet peroxide oxidation in the presence of solid catalysts. Chem. Sustain. Dev. 2003, 11, 227–232. (10) Beverskog, B.; Puigdomenech, I. Revised Pourbaix diagrams for chromium at 25300 °C. Corros. Sci. 1997, 39, 43–57. (11) Owlad, M.; Aroua, M. K.; Daud, W. A. W.; Baroutian, S. Removal of hexavalent chromium-contaminated water and wastewater: A review. Water, Air Soil Pollut. 2009, 200, 59–77. (12) Wang, W. X.; Griscom, S. B.; Fisher, N. S. Bioavailability of Cr(III) and Cr(VI) to marine mussels from solute and particulate pathways. Environ. Sci. Technol. 1997, 31, 603–611. (13) CRC Handbook of Chemistry and Physics, 85th ed.; CRC Press: Boca Raton, FL, 2004. (14) Rai, D.; Eary, L. E.; Zacharia, J. M. Environmental chemistry of chromium. Sci. Total Environ. 1989, 86, 15–23. (15) Fendorf, S. E.; Zasoski, R. J. Chromium(III) oxidation by deltaMnO2 (I) Characterization. Environ. Sci. Technol. 1992, 26, 79–85. (16) Zhang, H.; Bartlett, R. J. Light-induced oxidation of aqueous chromium(III) in the presence of iron(III). Environ. Sci. Technol. 1999, 33, 588–594. (17) Cotton, F. A.; Wilkinson, G.; Murillo, C. A.; Bochmann, M. Advanced Inorganic Chemistry; Wiley-Interscience: New York, 1999. (18) Richard, F. C.; Bourg, A. C. M. Aqueous geochemistry of chromium: A review. Water Res. 1991, 25, 807–816. (19) Pettine, M.; Campanella, L.; Millero, F. Reduction of hexavalent chromium by H2O2 in acidic solutions. Environ. Sci. Technol. 2002, 36, 901–907. (20) Bokare, A. D.; Choi, W. Chromate-induced activation of hydrogen peroxide for oxidative degradation of aqueous organic pollutants. Environ. Sci. Technol. 2010, 44, 7232–7237. (21) APHA, AWWA, WEF. Standard Methods for the Examination of Water and Wastewater, 20th ed; APHA: Washington, DC, 1988; pp 3-653-68. (22) Pettine, M.; La Noce, T.; Liberatori, A.; Loreti, L. Hydrogen peroxide interference in the determination of chromium(VI) by the diphenylcarbazide method. Anal. Chim. Acta 1988, 209, 315–319. (23) Hunt, J. P. Metal Ions in Aqueous Solutions; Benjamin, Inc: New York, 1963. (24) Joo, S. H.; Feitz, A. Z.; Sedlak, D. L.; Waite, T. D. Quantification of the oxidizing capacity of nanoparticulate zero-valent iron. Environ. Sci. Technol. 2005, 39, 1263–1268.
ARTICLE
(25) Shi, X.; Mao, Y.; Knapton, A. D.; Ding, M.; Rojanasakul, Y.; Gannett, P. M.; Dalal, N. S.; Liu, K. Reaction of Cr(VI) with ascorbate and hydrogen peroxide generates hydroxyl radicals and causes DNA damage: Role of a Cr(IV)-mediated Fenton-like reaction. Carcinogenesis 1994, 15, 2475–2478. (26) Tsou, T.-C.; Yang, J.-L. Formation of reactive oxygen species and DNA strand breakage during interaction of chromium(III) and hydrogen peroxide in vitro: Evidence for a chromium(III)-mediated Fenton-like reaction. Chem.-Biol. Interact. 1996, 102, 133–153. (27) Haight, G. P., Jr.; Huang, T. J.; Shakhashiri, B. Z. Reactions of Cr(IV). J. Inorg. Nucl. Chem. 1971, 33, 2169–2176. (28) Shi, X.; Dalal, N. S. ESR spin trapping detection of hydroxyl radicals in the reactions of Cr(V) complexes with hydrogen peroxide. Free Radical Res. Commun. 1990, 10, 17–26. (29) Shi, X.; Dalal, N. S.; Kasprzak, K. S. Generation of free radicals from hydrogen peroxide and lipid hydroperoxides in the presence of Cr(III). Arch. Biochem. Biophys. 1993, 302, 294–299. (30) Codd, R.; Dillon, C. T.; Levina, A.; Lay, P. A. Studies on the genotoxicity of chromium: From the test tube to the cell. Coord. Chem. Rev. 2001, 216, 537–582. (31) Zhang, L.; Lay, P. A. EPR spectroscopic studies on the formation of chromium(V) peroxo complexes in the reaction of chromium(VI) with hydrogen peroxide. Inorg. Chem. 1998, 37, 1729–1733. (32) Perez-Benito, J. F.; Arias, C. A kinetic study of the chromium(VI)-hydrogen peroxide reaction. Role of the diperoxochromate(VI) intermediates. J. Phys. Chem. A 1997, 101, 4726–4733. (33) St€unzi, H.; Spiccia, L.; Rotzinger, F. P.; Marty, W. Early stages of the hydrolysis of chromium(III) in aqueous solution. 4. Stability constants of the hydrolytic dimer, trimer, and tetramer at 25 °C and I = 1.0 M. Inorg. Chem. 1989, 28, 66–71. (34) St€unzi, H.; Marty, W. Early stages of the hydrolysis of chromium(III) in aqueous solution. 1. Characterization of a tetrameric species. Inorg. Chem. 1983, 22, 2145–2150. (35) Rao, L.; Zhang, Z.; Friese, J. I.; Ritherdon, B.; Clark, S. B.; Hess, N. J.; Rai, D. Oligomerization of chromium(III) and its impact on the oxidation of chromium(III) by hydrogen peroxide in alkaline solutions. J. Chem. Soc., Dalton Trans. 2002, 267–274. (36) Szejtli, J. Introduction and general overview of cyclodextrin chemistry. Chem. Rev. 1998, 98, 1743–1753. (37) Norkus, E. Metal ion complexes with native cyclodextrins. J. Inclusion Phenom. Macrocyclic Chem. 2009, 65, 237–248. (38) McNamara, M.; Russell, N. R. FT-IR and Raman spectra of a series of metallo-β-cyclodextrin complexes. J. Inclusion Phenom. Mol. Recogn. Chem. 1991, 10, 485–495. (39) St€unzi, H.; Rotzinger, F. P.; Marty, W. Early stages of the hydrolysis of chromium(III) in aqueous solution. 2. Kinetics and mechanism of the interconversion between two tetrameric species. Inorg. Chem. 1984, 23, 2160–2164. (40) Ensing, B.; Buda, F.; Baerends, E. J. Fenton-like chemistry in water: Oxidation catalysis by Fe(III) and H2O2. J. Phys. Chem. A 2003, 107, 5722–5731. (41) Lay, P. A. Recent developments on the mechanisms of substitution reactions of octahedral coordination complexes. Coord. Chem. Rev. 1991, 110, 213–233. (42) Xu, F.-C.; Krouse, R.; Swaddle, T. W. Conjugate base pathway for water exchange on aqueous chromium(III): Variable-pressure and temperature kinetic study. Inorg. Chem. 1985, 24, 267–270. (43) Espenson, J. H. Formation rates of monosubstituted chromium(III) complexes in aqueous solution. Inorg. Chem. 1969, 8, 1554–1556. (44) Crimp, S. J.; Spiccia, L.; Krouse, H. R.; Swaddle, T. W. Early stages of the hydrolysis of chromium(III) in aqueous solution. 9. Kinetics of water exchange on the hydrolytic dimer. Inorg. Chem. 1994, 33, 465–470. (45) Pettine, M.; Gennari, F.; Campanella, L.; Millero, F. J. The effect of organic compounds in the oxidation kinetics of Cr(III) by H2O2. Geochim. Cosmochim. Acta 2008, 72, 5692–5707.
9338
dx.doi.org/10.1021/es2021704 |Environ. Sci. Technol. 2011, 45, 9332–9338
ARTICLE pubs.acs.org/est
Catalytic Ozonation of Oxalate with a Cerium Supported Palladium Oxide: An Efficient Degradation Not Relying on Hydroxyl Radical Oxidation Tao Zhang,† Weiwei Li,‡ and Jean-Philippe Croue*,† †
Water Desalination and Reuse Center (WDRC), King Abdullah University of Science and Technology (KAUST), Thuwal 4700, Kingdom of Saudi Arabia ‡ Research Center for Eco-Environmental Sciences (RCEES), Chinese Academy of Sciences, Beijing 100085, China
bS Supporting Information ABSTRACT: The cerium supported palladium oxide (PdO/ CeO2) at a low palladium loading was found very effective in catalytic ozonation of oxalate, a probe compound that is difficult to be efficiently degraded in water with hydroxyl radical oxidation and one of the major byproducts in ozonation of organic matter. The oxalate was degraded into CO2 during the catalytic ozonation. The molar ratio of oxalate degraded to ozone consumption increased with increasing catalyst dose and decreasing ozone dosage and pH under the conditions of this study. The maximum molar ratio reached around 1, meaning that the catalyst was highly active and selective for oxalate degradation in water. The catalytic ozonation, which showed relatively stable activity, does not promote hydroxyl radical generation from ozone. Analysis with ATR-FTIR and in situ Raman spectroscopy revealed that 1) oxalate was adsorbed on CeO2 of the catalyst forming surface complexes, and 2) O3 was adsorbed on PdO of the catalyst and further decomposed to surface atomic oxygen (*O), surface peroxide (*O2), and O2 gas in sequence. The results indicate that the high activity of the catalyst is related to the synergetic function of PdO and CeO2 in that the surface atomic oxygen readily reacts with the surface cerium-oxalate complex. This kind of catalytic ozonation would be potentially effective for the degradation of polar refractory organic pollutants and hydrophilic natural organic matter.
’ INTRODUCTION Catalytic ozonation with metal oxides is a potential process to enhance the degradation of recalcitrant organics in water. It is generally accepted that catalytic ozonation with metal oxides can be ascribed to either of the two pathways: ozone decomposition on the catalyst surface generating hydroxyl radicals and direct ozone oxidation of surface metalorganic complexes.1 In the past decade, many studies were conducted on catalytic ozonation that follows the hydroxyl radical pathway. Successful works were accomplished on the preparation of efficient catalyst accelerating hydroxyl radical production from ozone and the identification of active sites involved in this process.24 With regard to catalytic ozonation relying on surface complexes, a lot of publications refer to the degradation of organic acids with mono- or hybrid- metal oxides as catalysts.57 However, minor achievements were made in this domain of research because of the lack of highly efficient catalysts as well as the difficulties in identifying active sites and active oxidant species, information that is essential to understand the mechanism and optimize the catalyst preparation.1 Hydroxyl radical has high reaction rate constants with almost all organics in water. However, hydroxyl radical oxidation is not effective to degrade aliphatic hydrophilic compounds that contain carbonyl or carboxylic groups, e.g., the products formed during the ozonation of natural organic matter.8 This is because the consumption of hydroxyl radical by bicarbonates/carbonates r 2011 American Chemical Society
(k = 8.5 106/3.9 108 M1 s1) and ozone (k = 1 1082 109 M1 s1) are usually faster than its reactions with the saturated compounds.8 From this point, catalytic ozonation through a complexation pathway would be more selective and efficient for the degradation of saturated hydrophilic organics, if highly active catalyst can be prepared. Oxalate is one of the major byproducts in ozonation or advanced oxidation of natural organic matter and organic pollutants.9 Its degradation in reaction with both molecular ozone (k e 0.04 M1 s1) and hydroxyl radical (k = 7.7 106 M1 s1) are relatively low.10 Therefore, it is usually used as a probe compound to study catalytic ozonation that follows surface complexation pathway. MnO2, Fe2O3, Co3O4, and NiO had been found effective in the catalytic ozonation of oxalic acid but the removal rates at normal pHs in a reasonable reaction time were too low and the ozone doses were too high for practical uses.5,7,11,12 There are still no data on ozone consumption for oxalate degradation which is critical to evaluate the catalysis efficiency. No evidence shows active sites and active oxidant species that are responsible for oxalate degradation, the information Received: June 28, 2011 Accepted: October 4, 2011 Revised: October 1, 2011 Published: October 04, 2011 9339
dx.doi.org/10.1021/es202209j | Environ. Sci. Technol. 2011, 45, 9339–9346
Environmental Science & Technology that is needed to better understand the reaction pathway and tailor cost-effective catalysts. PdO supported on ceria-zirconium oxide has been used successfully as a three-way catalyst for automobile-exhausted gases to reduce NO and oxidize CO and hydrocarbons.13 With regard to catalytic ozonation of organic compounds in water with supported PdO, there is still no report. PdO had been found effective in decomposing aqueous ozone,14 but intermediates in this process are unknown. CeO2 can reduce bromate formation during ozonation, while it is inactive in promoting ozone decomposition and organic compound degradation.15 In this study, the effectiveness and efficiency of a cerium supported palladium oxide (PdO/CeO2) in catalytic ozonation was tested using oxalate as a probe compound. In order to elucidate the efficient pathway for catalytic ozonation not relying on hydroxyl radical, surface active sites of the catalyst and surface oxygen intermediates formed from ozone decomposition were investigated with ATR-FTIR and in situ Raman spectroscopy.
’ EXPERIMENTAL SECTION Metal Oxides Preparation. The support CeO2 was synthesized with a urea-hydrothermal method. Ce(NO3)3 3 6H2O and urea were dissolved in distilled water with a molar ratio of 1:3. The mixture was heated at 140 °C for 5 h. After filtration and repeated washing, the precipitate was dried at 120 °C for 2 h and calcined at 450 °C for 4 h. PdO/CeO2 was prepared by impregnating the CeO2 with Pd(NO3)2 aqueous solution with incipient wetness. The impregnated oxide was dried at 60 °C and finally calcined in air at 550 °C for 2 h. Palladium mass proportion of the PdO/CeO2 was measured to be 3.8%. PdO particle was prepared by direct calcination of the dried Pd(NO3)2 at 550 °C for 2 h. Characterization. BET surface area of the metal oxides was determined on a Micromeritics ASAP2000 analyzer. pHpzc (pH at which the surface is zero-charged) was determined with acidbase titration. Average particle size was measured on a Mastersizer 2000 laser particle size analyzer. STEM (scanning transmission electron microscopy) pictures taken on a Titan 80300 transmission electronic microscope were used to characterize the dispersion of PdO on CeO2. The major characteristics of the metal oxides used in this study are listed in Table S1 (Supporting Information). Experimental Procedure. Batch Reaction. Milli-Q water at room temperature (21°C) was continuously bubbled with gaseous ozone produced with an ozone generator (3S-A5, Tonglin Technology) from dried oxygen gas. The aqueous ozone concentration was analyzed continuously with an ultraviolet spectrometer (Hach 500) at 258 nm (molar absorbance coefficient = 3000 M1 cm1) until it reached a steady state. The steady ozone concentration in water was controlled by adjusting the electric current of the ozone generator. Predetermined volume of ozone stock solution was quickly mixed with tetraborate buffered oxalate solution in a glass reactor. In the case of catalytic ozonation, the catalyst (mostly at a dose of 150 mg L1 unless specified) was also instantly introduced into the reaction solution. Then, the reactor was sealed and magnetically stirred. Samples were taken at each time-point and then filtered through 0.45 μm acetate-fiber syringe filters and purged with pure N2 to remove residual ozone in water. The filtration had no impact on the oxalate concentration. In order to reduce the impact of filtration on aqueous ozone, 50 mL of the ozone stock solution was pressed through the filter before sample filtration for ozone
ARTICLE
analysis. Then, the filtration had nearly no impact on the aqueous ozone concentration when the filter had been pretreated in this way. Semicontinuous Reaction. Gaseous ozone (concentration = 21 mg O3 L1, flow rate = 3.0 L min1) was continuously introduced into a 500 mL reaction solution (0.2 mM oxalate in 10 mM tetra-borate buffer) in a glass vessel. After the first 10 min, 10 mL of reaction solution was withdrawn from the reaction vessel with a syringe. Subsequently, the same volume of oxalate stock solution (10 mM, buffered with 10 mM tetraborate) was immediately added into the reactor. The same operation was repeated 30 times. The samples were filtered with 0.45 μm acetate fiber filters and then purged with pure N2 to remove residual ozone. Analysis. Ozone concentration in reaction solution was directly determined on the UV spectrometer at 258 nm after the filtration, using the tetraborate buffer (10 mM and pH 6.5) as zero background, because oxalate had no detectable absorbance at this wavelength. Oxalate was analyzed on a Dionex ICS-1600 IC equipped with an AS-9 column. The mobile phase was 9 mM Na2CO3 at a flow rate of 1.0 mL min1. TOC was measured on a Teledyne Tekmar TOC Fusion analyzer. Atrazine used as a probe compound of hydroxyl radical in this study was determined on a Waters HPLC equipped with a Symmetry C-18 column at a UV wavelength of 220 nm. The mobile phase was isocratic H2O/acetonitrile at a volume ratio of 3/7 and a flow rate of 1.0 mL min1. Dissolved palladium and cerium ions were determined on an ICP-MS (Agilent 7500) with detection limits of 9.4 104 and 1.7 102 μg L1 for the two cations, respectively. Palladium content of the catalyst was also determined by the ICP-MS after digestion of the catalyst with HCl + HNO3 (v/v = 3:1) and HF in sequence. A Perkin Elmer FTIR spectrometer (Spectrum 100) equipped with a Universal ATR accessory was used to characterize the catalyst surface in presence or absence of ozone and oxalate. In the study of aqueous ozone adsorption, the suspension of metal oxide (750 mg L1) was continuously bubbled with gaseous ozone (concentration = 21 mg L1, flow rate = 3.0 L min1) for over 10 min in a 25 mL glass tube which was cooled with ice. The suspension was quickly dropped on the ZnSe crystal of the ATR accessory with a glass pipet, covered with a stainless lid, and scanned in the range of 8004000 cm1 at a resolution of 4 cm1. In the study of oxalate adsorption, 0.2 mM oxalate in 10 mM tetraborate buffer of pH 6.5 was mixed with metal oxide particles (150 mg L1) and stirred for over 2 h. After settling, the particles were dropped on the ATR crystal and analyzed at the same conditions as that described above. In situ Raman spectra were taken on a confocal microscopic Raman spectrometer (Aramis, Horiba Jobin Yvon) with a 9 mW 633 nm laser light irradiation. Before analysis, the metal oxides were pressed into slices of 12 mm in thickness and 13 mm in diameter. The slice was then stuck on a microscope slide with a double-sided tape, wetted with Milli-Q water, and then blown with gaseous ozone (concentration = 1.5 mg L1, flow rate = 1.4 L min1) which was humidified through a gas washer. The slice was scanned from 300 to 1200 cm1 at a resolution of 1 cm1 and a duration time of 100 s under the humidified ozone gas.
’ RESULTS AND DISCUSSION Effectiveness. Figure 1A shows the decrease in oxalate concentration in batch reaction mode during catalytic ozonation 9340
dx.doi.org/10.1021/es202209j |Environ. Sci. Technol. 2011, 45, 9339–9346
Environmental Science & Technology
ARTICLE
Figure 1. Oxalate removal in catalytic ozonation and adsorption alone with the single and the binary oxides (A) and ozone decomposition in the catalytic ozonation (B). Experimental conditions: ozone dose = 0.18 mM, initial oxalate concentration = 0.2 mM, oxide dose = 150 mg L1, T = 21 °C, 10 mM tetraborate buffered pH = 6.5.
Figure 2. Effect of catalyst dose (a- 0 mg L1, b- 70 mg L1, c- 150 mg L1, and d- 250 mg L1) on oxalate removal (A) and the effective ozone consumption ratio (B) in the catalytic ozonation. Experimental conditions: ozone dose = 0.18 mM, initial oxalate concentration = 0.2 mM, T = 21 °C, 10 mM tetraborate buffered pH = 6.5.
and adsorption alone with CeO2, PdO, and PdO/CeO2. Oxalate was quite stable during ozonation alone (absence of catalyst). Its concentration decreased less than 0.014 mM in 12 min at the ozone dose of 0.18 mM. Oxalate loss in O3/CeO2 approximated the sum of that removed during ozonation alone and that in CeO2 adsorption alone, meaning that CeO2 has no activity in oxalate degradation. PdO showed nearly no adsorption for oxalate. However, it promoted oxalate removal during catalytic ozonation to 0.037 mM. Even PdO mass on the PdO/CeO2 was less than 4%, the binary oxide showed a much higher efficiency than PdO. Oxalate loss reached to 0.08 mM in O3/(PdO/CeO2), which can largely be attributed to degradation because the oxalate loss in PdO/CeO2 adsorption alone was only 0.014 mM. After catalytic ozonation with PdO/CeO2, the presence of cerium and palladium ions in the reaction solution were examined. The cerium ion concentration was below the detection limit. The concentration of palladium ion was 0.24 μg L1, which was 4.2 105 times the one of the solid PdO dose. The suspension was filtered with a 0.45 μm filter and ozonated again without the presence of PdO/CeO2. No further oxalate degradation was observed (not shown). Therefore, the oxalate degradation is due to heterogeneous catalysis but not due to catalysis effect of trace palladium ion in water. It was reported that effective catalytic oxalate degradation with several typical metal oxides were achieved only at much lower pHs (e.g., 4.1 and 3.2 for MnO2, 2.5 for Fe2O3 and Co2O3, and 2.4 for NiO) and continuous ozone introduction during the reaction over 30 min.5,7,11,12 This binary oxide seems to be more efficient for practical catalytic ozonation reaction.
The ozone decay recorded during ozonation in the presence or in the absence of the oxides exhibited a similar pattern as the oxalate removal profile (Figure 1B). Consistent with previous results,15 CeO2 nearly did not improve ozone decomposition. The ozone depletion in water was promoted by PdO and to a higher rate by PdO/CeO2. PdO seems to be an active metal oxide in oxalate degradation and ozone decomposition. The activity was significantly improved by CeO2 support. The disperse effect of CeO2 for small PdO crystals on the surface (i.e., the effect against PdO particle agglomeration) (Figure S1, Supporting Information) as well as the adsorption of oxalate on CeO2 possibly contributed to the high activity of the PdO/CeO2. Effect of Catalyst Dose. The increase of PdO/CeO2 dose from 0 to 250 mg L1 significantly accelerated oxalate removal (Figure 2) as well as ozone decomposition (Figure S2A, Supporting Information). The oxalate removal in the presence of ozone (i.e., catalytic ozonation) was much more significant than that in the absence of ozone (i.e., adsorption alone) (Figure S2B, Supporting Information). Moreover, ozonation did not improve the absorbability of the catalyst for oxalate (Figure S2B, Supporting Information). Therefore, the high oxalate removal in the catalytic ozonation was due to degradation but not adsorption. It is interesting that a further loss of oxalate concentration by 0.02 mM was observed during the catalytic ozonation with 250 mg L1 of PdO/CeO2 when ozone in water had been completely consumed after 12 min. Since oxalate concentration in adsorption alone reached equilibrium within 3 min at this PdO/CeO2 dose, the further oxalate loss might be due to the degradation of adsorbed 9341
dx.doi.org/10.1021/es202209j |Environ. Sci. Technol. 2011, 45, 9339–9346
Environmental Science & Technology
Figure 3. Effect of ozone dose (a- 0.05 mM, b- 0.13 mM, c- 0.18 mM) on oxalate removal (A) and the effective ozone consumption ratio (B) in the catalytic ozonation. Experimental conditions: initial oxalate concentration = 0.2 mM, catalyst dose = 150 mg L1, T = 21 °C, 10 mM tetraborate buffered pH = 6.5.
oxalate by adsorbed ozone or new oxidant species and a further adsorption. The loss of oxalate during the catalytic ozonation with different catalyst doses compensated well for TOC loss (Figure S3, Supporting Information). It means that the oxalate was oxidized directly into CO2 with no formation of other stable organic intermediates, consistent with the results found in homogeneous catalytic ozonation of oxalate with Co2+.10 The effective ozone consumption ratio for oxalate degradation was calculated as the oxalate degradation (total removal subtracted by the removal in adsorption alone) per mole of ozone consumed. Figure 2B shows that this ratio substantially increased from 0.05 to 0.5 in 25 min reaction as the catalyst dose increased from 0 to 250 mg L1. Effect of Ozone Dose. Over half of oxalate removal always occurred within the first 3 min of reaction in O3/(PdO/CeO2) (shown in Figure 1A and 2A). It was expected that the increase of ozone dose will improve oxalate removal in this initial reaction phase. However, the increase of ozone dose from 0.05 to 0.18 mM reduced oxalate removal rate (Figure 3A). The positive effect of increasing ozone dose on oxalate degradation rate was observed only after the initial phase. Because aqueous ozone concentration was relatively high in the initial phase compared with the following phase (Figure S4, Supporting Information), the result suggests that molecular ozone might not be the oxidant species directly responsible for the oxalate degradation. It is likely that several oxygen species can be formed in sequence from ozone decomposition on the catalyst surface. Ozone might compete with oxalate to react with some of the new oxygen species that are effective for oxalate degradation, thus leading to
ARTICLE
Figure 4. Effect of pH on oxalate removal in ozonation alone (solid symbols) and catalytic ozonation (open symbols) (A) and the effective ozone consumption ratio (B). Experimental conditions: ozone dose = 0.09 mM, initial oxalate concentration = 0.1 mM, catalyst dose = 150 mg L1, T = 21 °C, 10 mM tetraborate buffer and diluted HNO3 adjusted pH, 2.5 mM bicarbonate was also used for pH 7.8.
the negative effect on oxalate degradation in the initial phase when ozone dose was raised. At the dose of 0.05 mM, ozone in water was totally consumed in 5 min (Figure S4, Supporting Information). However, oxalate was further removed by 0.007 mM thereafter, which is similar to that observed in Figure 2A. The effective ozone consumption ratio for oxalate degradation increased from 0.5 to around 1.0 as the ozone dose decreased from 0.18 to 0.05 mM (Figure 3B). It may indicate that ozone reacts with some of the new oxidant species that are effective for oxalate degradation, thus reducing the effective ozone consumption ratio at high ozone doses. Effect of pH. Owing to ozone reactions with OH and some dissociating organics, ozone decomposition in water and the hydroxyl radical generation can be significantly accelerated at elevated pHs.8,16,17 Ozone decomposition at pH 9.2 was so fast that there was nearly no difference in residual ozone between ozonation alone and catalytic ozonation (Figure S5, Supporting Information). The fast ozone decomposition at this pH definitely accelerated hydroxyl radical generation. However, only about 5% of oxalate was removed during ozonation alone and catalytic ozonation (Figure 4A). The low efficiency of hydroxyl radical oxidation can be ascribed to its much higher reaction rate with molecular ozone (1 1082 109 M1 s1)8 than oxalate oxidation (7.7 106 M1 s1).10 The disappearance of catalytic effect at this pH can be ascribed to the fast ozone decomposition in water, which reduced significantly the chances of ozonecatalyst surface interaction. As the pH was decreased, oxalate was removed by 25%, 40%, 70%, and 98% during catalytic ozonation at pH 7.8, 7.0, 6.5, and 4.2, respectively (Figure 4A). In parallel, the ozone decomposition 9342
dx.doi.org/10.1021/es202209j |Environ. Sci. Technol. 2011, 45, 9339–9346
Environmental Science & Technology rate was also enhanced by the catalyst with that at pH 4.2 was slightly higher than that at pH 6.5 (Figure S5, Supporting Information). A similar observation was described during the catalytic ozonation of oxalate with Co2+.10 Oxalate adsorption was promoted by the decrease of pH. About 2%, 9%, and 28% oxalate was removed in adsorption alone at pH 7.8, 6.5 and 4.2, respectively (not shown). As the pHpzc of PdO/CeO2 is 5.4, it became more positively charged at lower pH, improving oxalate adsorption through electrostatic attraction. Therefore, the improvement of oxalate removal at a lower pH is related to enhanced oxalate adsorption on the catalyst which might further promote surface oxalate degradation during catalytic ozonation. The effective ozone consumption ratio in catalytic ozonation also increased from 0.06 to 0.78 as the pH decreased from 9.2 to 4.2 (Figure 4B). The low ratio at alkaline pH possibly is due to 1) fast ozone depletion in water which reduced the chances for ozone-catalyst interaction and 2) activity decrease of the catalyst caused by the negative effect of high pH on oxalate adsorption. No Hydroxyl Radical Generation. The introduction of t-BuOH into the reaction solution showed nearly no influence on ozone decomposition and oxalate degradation in the catalytic ozonation (Figure S6A and S6B, Supporting Information). It is clear that hydroxyl radical is not involved in the catalytic ozone decomposition and oxalate degradation. Atrazine (kO3 = 6 M1 s1, k•OH = 3 109 M1 s1)8 was used as an additional probe compound for the catalytic ozonation (Figure S7, Supporting Information). Degradation rates of atrazine at trace level during catalytic ozonation in the presence or the absence of oxalate were similar. Pines and Reckhow observed that hydroxyl radical was a byproduct of oxalate degradation in catalytic ozonation with cobalt ion.10 However, no hydroxyl radical was generated from the catalytic oxalate degradation here, indicating that the degradation pathways involved in the two processes are different. The atrazine degradation rate during catalytic ozonation was much lower than during ozonation alone, proving again that hydroxyl radical was not generated from the catalytic ozone decomposition. Stable Activity. Oxalate stock solution was intermittently added into the reactor by 30 times in the semicontinuous catalytic ozonation to test the stability of the catalyst. The amount of oxalate added into the reaction solution at the end of each 10 min reaction can increase oxalate concentration by 0.2 mM. In most cases, nearly all of the oxalate added was degraded within the 10 min reaction (Figure S8, Supporting Information). No decrease of the oxalate removal was observed with the increase of reaction runs in the semicontinuous reaction. Therefore, the activity of PdO/CeO2 remained relatively stable for our experimental conditions. Active Sites, Oxidant Species, and Reaction Pathway. Figure 5A-C shows ATR-FTIR spectra obtained from PdO/ CeO2, CeO2, and PdO in water with and without the presence of ozone and oxalate, respectively. New double reflectance bands with weak intensities appeared for PdO/CeO2 (2086 and 2048 cm1) and PdO (2040 and 2014 cm1) in contact with aqueous ozone. According to refs 1820, these bands can be assigned to distorted molecular ozone vibrations with interaction of ozone with surface metal sites. There was no adsorbed ozone features for CeO2 in contact with aqueous ozone. These results indicate that PdO is the component that adsorbs ozone. The reflectance of adsorbed ozone on the catalyst shifted to higher wavenumbers as compared with PdO alone, suggesting that the CeO2-supported PdO has a stronger binding or distortion effect
ARTICLE
Figure 5. ATR-FTIR spectra of PdO/CeO2 (A), CeO2 (B), and PdO (C) in contact with water (curve a), aqueous ozone (curve b), and aqueous oxalate (curve c).
for adsorbed ozone molecule. A stronger distortion can make the adsorbed ozone more unstable against dissociation.18 Therefore, the faster ozone decomposition on the catalyst than that on PdO alone (shown in Figure 1B) can be related to the higher affinity of the catalyst for ozone molecule. New reflectance bands at 1426 and 1432 cm1 appeared for PdO/CeO2 and CeO2 in contact with oxalate solution, respectively. Oxalate was reported to have two characteristic IR bands in the region of 16501550 and 1400 cm1.21 However, oxalate in water only showed the apparent 1400 cm1 peak with the ATR-FTIR (Figure S9, Supporting Information), probably because pure water that was used for a background scan also had a reflectance around 1600 cm1. The 1400 cm1 band which is due to CdO stretch shifted to higher wavenumbers in the presence of CeO2 and PdO/CeO2. According to Marley et al.,21 this result indicates that oxalate was adsorbed onto cerium sites of the oxides forming surface complexes. Although oxalate can be adsorbed on CeO2 through surface complexation, CeO2 alone had nearly no activity for oxalate 9343
dx.doi.org/10.1021/es202209j |Environ. Sci. Technol. 2011, 45, 9339–9346
Environmental Science & Technology
ARTICLE
The lattice CeIII content increase can stabilize surface metal ions when their oxidation states increase.28 It is likely that here the lattice oxygen vacancy formation in the CeO2 support is due to the stabilization of the surface palladium ions at the strongly oxidative atmosphere. Two new broad peaks appeared at 828 and 937 cm1. They are characteristic features of surface peroxide (*O2) and surface atomic oxygen (*O), respectively.19,2224 When the ozone flow was cut off, the peaks slowly decreased (Figure S10, Supporting Information), indicating the catalyst can recover itself. The Raman spectra of CeO2 with and without the presence of ozone were nearly the same (Figure 6B). No intensity increase at 556 cm1 was observed in the presence of ozone, meaning that the oxygen vacancies cannot be formed when ozone contacts with CeO2 alone. The features of surface palladium oxide, surface peroxide, and atomic oxygen were all observed for PdO in contact with ozone (Figure 6C). However, the peak wavenumbers of surface PdO (630 cm1) and atomic oxygen (912 cm1) were lower than the values observed for PdO/CeO2 and higher for peroxide (837 cm1). This result indicates that there is a strong interaction between the CeO2 support and the palladium ion which led to the changes of binding strength between palladium and these surface oxygen-containing species. Our results clearly showed that PdO on the catalyst surface is the active site in inducing the decomposition of ozone into the intermediate oxygen species. Since ozone can adsorb onto PdO (shown in Figure 5A and 5C), it is possible that the adsorbed ozone further decomposed to surface atomic oxygen and gaseous O2 (eq 1). Another ozone molecule would react with the surface atomic oxygen forming a surface peroxide species and a gaseous O2 (eq 2). Because the peak of the surface peroxide disappeared when ozone was removed, the surface peroxide would further decompose to gaseous O2 (eq 3). Such a catalytic ozone decomposition process under the mimic aqueous condition is consistent with gaseous ozone decomposition on manganese oxides19,22 Figure 6. Raman spectra of PdO/CeO2 (A), CeO2 (B), and PdO (C) with and without the presence of ozone.
degradation during ozonation (shown in Figure 1A). Therefore, direct oxidation of the cerium-oxalate complex by molecular ozone is still not the catalytic ozonation pathway. It is likely that the adsorbed ozone molecule can produce more active surface oxygen species that are quite selective for the complex degradation. In situ Raman spectroscopy had been used to characterize intermediate oxygen species formed on manganese oxides in contact with gaseous ozone.19,22 It was applied here to get insights on ozone decomposition on the catalyst. Figure 6A-C shows Raman spectra of PdO/CeO2, CeO2, and PdO respectively with and without the presence of ozone. Ozone itself had no Raman signal in this spectrum range. The intensities of the peak at 556 cm1 and the shoulder at 644 cm1 increased significantly for PdO/CeO2 in contact with ozone (Figure 6A). The 556 cm1 feature arises from lattice oxygen vacancies in CeO2.23,24 The 644 cm1 is ascribed to new palladium oxide species.25,26 CeO2 has a special property of changing oxidation state of lattice Ce between CeIII and CeIV through oxygen release and storage.27 The intensity increase of the 556 cm1 peak means that lattice CeIII content in the CeO2 support increased.
Pd þ O3 f PdO þ O2
ð1Þ
PdO þ O3 f PdO2 þ O2
ð2Þ
PdO2 f Pd þ O2
ð3Þ
The Raman feature of surface atomic oxygen is attributed to the stretches of metaloxygen double bond MedO.29 The higher PddO wavenumber observed on the PdO/CeO2 indicates that the CeO2 support led to stronger affinity of palladium for the atomic oxygen. Such a kind of palladium-atomic oxygen interaction on the catalyst in comparison with that on PdO probably would increase the stability of the atomic oxygen and consequently reduce the rate of its reaction with ozone which forms surface peroxide. Because the oxidation potential of atomic oxygen (2.43 V in water) is much higher than that of peroxide (1.35 V in protonated form),30 the surface atomic oxygen is likely the active oxygen species in the catalytic ozonation. The maximum effective ozone consumption ratio of the catalyst was observed to be around 1 for oxalate degradation under our experimental conditions (shown in Figure 3B). This result reinforces the hypothesis that the surface atomic oxygen is the major effective oxidant in the catalytic ozonation. If surface peroxide was effective for oxalate degradation, the maximum effective ozone consumption ratio would be around 0.5 (2 mols 9344
dx.doi.org/10.1021/es202209j |Environ. Sci. Technol. 2011, 45, 9339–9346
Environmental Science & Technology
ARTICLE
of O3 needed to produce surface peroxide). This point is also supported by the fact that high ozone concentration showed negative effect on oxalate degradation in the initial phase (shown in Figure 3A). Ozone in excess may react with surface atomic oxygen to produce surface peroxide. Our reaction mechanism hypothesis agrees with the fact that further oxalate degradation is observed after complete consumption of aqueous ozone (Figure 2A and 3A) because the adsorbed ozone on the catalyst decomposes to form atomic oxygen and it can survive for some time in absence of aqueous ozone (Figure S10, Supporting Information). It has been reported that oxidative decarboxylation of CeIVcarboxylates occurs under heating conditions in water through one-electron transfer forming alkyl radical and two-electron transfer forming carbonium ion (eqs 4 and 5)31,32 CeIV ðO2 CRÞ3 þ f CeIII ðO2 CRÞ2 þ þ CO2 þ R •
ð4Þ
R • þ CeIV f CeIII þ R þ
ð5Þ
In the surface cerium-oxalate complex, oxalate partially donates its electron density to CeIV. Because atomic oxygen has a higher oxidation potential than CeIV (1.70 V), it is likely that the surface atomic oxygen extracts one or two electrons from the surface cerium-oxalate complex, thus initiating oxidative decarboxylation under mild conditions without reducing surface CeIV. The high efficiency of the catalytic ozonation with PdO/CeO2 in oxalate degradation can then be related to a synergetic effect between PdO and the support: 1) PdO acts as the active site to decompose ozone to surface atomic oxygen, and 2) the support CeO2 activates oxalate in reaction with the surface atomic oxygen through forming surface complexes. In addition, the CeO2 support also promotes the adhesion and decomposition of ozone on PdO. The strong affinity probably increased the stability of the surface atomic oxygen against the fast reaction with ozone forming peroxide. This work shows potentially effective degradation for polar refractory organics with composite metal oxide-assisted ozonation. It would be also promising for the degradation of hydrophilic natural organic matter (NOM) with high carboxylic contents known as refractory to conventional water treatments and one of the precursors of disinfection byproduct.33 Future research in following aspects might be necessary for the preparation of active catalysts and the application: 1) the influence of coordination affinity for target compounds on its degradation efficiency when another metal oxide is combined with CeO2 as a hybrid support, 2) activity changes when other metal oxides that can decompose ozone are supported on CeO2 as ozone active sites, and 3) degradation of polar refractory compounds and hydrophilic NOM components that cannot be effectively removed in hydroxyl radical oxidation.
’ ASSOCIATED CONTENT
bS
Supporting Information. One table and ten figures. This material is available free of charge via the Internet at http://pubs. acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: + 966 (0) 2 808 2984. E-mail: [email protected].
’ ACKNOWLEDGMENT We want to thank Dr. Yang Yang and Mr. Qingxiao Wang of Imaging and Characterization Laboratory of KAUST for their help in performing Raman and STEM analysis and Dr. Cyril Aubry, Ms. Tong Zhan, and Dr. Min Yoon of WDRC of KAUST in TEM, ICP-MS, and IC analysis. We also want to thank our anonymous reviewers for their valuable comments to improve this work. ’ REFERENCES (1) Nawrocki, J.; Kasprzyk-Horden, B. The efficiency and mechanisms of catalytic ozonation. Appl. Catal., B 2010, 99, 27–42. (2) Zhang, T.; Li, C.; Ma, J.; Tian, H.; Qiang, Z. Surface hydroxyl groups of synthetic a-FeOOH in promoting •OH generation from aqueous ozone: Property and activity relationship. Appl. Catal., B 2008, 82, 131–137. (3) Yang, L.; Hu, C.; Nie, Y.; Qu, J. Catalytic ozonation of selected pharmaceuticals over mesoporous alumina-supported manganese oxide. Environ. Sci. Technol. 2009, 43, 2525–2529. (4) Zhao, L.; Sun, Z.; Ma, J. Novel relationship between hydroxyl radical initiation and surface group of ceramic honeycomb supported metals for the catalytic ozonation of nitrobenzene in aqueous solution. Environ. Sci. Technol. 2009, 43, 4157–4163. (5) Andreozzi, R.; Insola, A.; Caprio, V.; Marotta, R.; Tufano, V. The use of manganese dioxide as a heterogeneous catalyst for oxalic acid ozonation in aqueous solution. Appl. Catal., A 1996, 138, 75–81. (6) Delanoe, F.; Acedo, B.; Vel Leitner, N. K.; Legube, B. Relationship between the structure of Ru/CeO2 catalysts and their activity in the catalytic ozonation of succinic acid aqueous solutions. Appl. Catal., B 2001, 29, 315–325. (7) Beltran, F. J.; Rivas, F. J.; Montero-de-Espinosa, R. Iron type catalysts for the ozonation of oxalic acid in water. Water Res. 2005, 39, 3553–3564. (8) von Gunten, U. Ozonation of drinking water: part I.Oxidation kinetics and product formation. Water Res. 2003, 37, 1443–1467. (9) Hammes, F.; Salhi, E.; Koster, O.; Kaiser, H. P.; Egli, T.; von Gunten, U. Mechanistic and kinetic evaluation of organic disinfection by-product and assimilable organic carbon (AOC) formation during the ozonation of drinking water. Water Res. 2006, 40, 2275–2286. (10) Pines, D. S.; Reckhow, D. A. Effect of dissolved cobalt(II) on the ozonation of oxalic acid. Environ. Sci. Technol. 2002, 36, 4046–4051. (11) Beltran, F. J.; Rivas, F. J.; Montero-de-Espinosa, R. Ozoneenhanced oxidation of oxalic acid in water with cobalt catalysts. 2. Heterogeneous catalytic ozonation. Ind. Eng. Chem. Res. 2003, 42, 3218– 3224. (12) Avramescu, S. M.; Bradu, C.; Udrea, I.; Mihalache, N.; Ruta, F. Degradation of oxalic acid from aqueous solutions by ozonation in presence of Ni/Al2O3 catalysts. Catal. Commun. 2008, 9, 2386–2391. (13) Jen, H. W.; Graham, G. W.; Chun, W.; McCabe, R. W.; Cuif, J. P.; Deutsch, S. E.; Touret, O. Characterization of model automotive exhaust catalysts: Pd on ceria and ceriazirconia supports. Catal. Today 1999, 50, 309–328. (14) Lin, J.; Kawai, A.; Nakajima, T. Effective catalysts for decomposition of aqueous ozone. Appl. Catal., B 2002, 39, 157–165. (15) Zhang, T.; Chen, W.; Ma, J.; Qiang, Z. Minimizing bromate formation with cerium dioxide during ozonation of bromide-containing water. Water Res. 2008, 42, 3651–3658. (16) Hoigne, J.; Bader, H. Rate constants of reactions of ozone with organic and inorganic compounds in water. II. Dissociating organic compounds. Water Res. 1983, 17, 185–94. (17) Xiong, F.; Croue, J. P.; Legube, B. Long-term ozone consumption by aquatic fulvic acids acting as precursors of radical chain reactions. Environ. Sci. Technol. 1992, 26, 1059–1064. (18) Bulanin, K. M.; Lavalley, J. C.; Tsyganenko, A. A. IR spectra of adsorbed ozone. Colloids Surf., A 1995, 101, 153–158. 9345
dx.doi.org/10.1021/es202209j |Environ. Sci. Technol. 2011, 45, 9339–9346
Environmental Science & Technology
ARTICLE
(19) Radhakrishman, R.; Oyama, S. T. Ozone decomposition over manganese oxide supported on ZrO2 and TiO2: A kinetic study using in situ laser Raman spectroscopy. J. Catal. 2001, 199, 282–290. (20) Zeng, Y.; Liu, Z.; Qin, Z.; Liu, H. Infrared study on adsorption of O3 at SnO2 surface. Spectr. Spectral Anal. 2008, 28, 1035–1038. (21) Marley, N. A.; Bennett, P.; Janecky, D. R.; Gaffney, J. S. Spectroscopic evidence for organic diacid complexation with dissolved silica in aqueous systems I. Oxalic acid. Org. Geochem. 1989, 14, 525–528. (22) Li, W.; Gibbs, G. V.; Ted Oyama, S. Mechanism of ozone decomposition on a manganese oxide catalyst. 1. In situ Raman spectroscopy and Ab initio molecular orbital calculations. J. Am. Chem. Soc. 1998, 120, 9041–9046. (23) Wu, Z.; Li, M.; Howe, J.; Meyer, H. M.; Overbury, S. H. Probing defect sites on CeO2 nanocrystals with well-defined surface planes by Raman spectroscopy and O2 adsorption. Langmuir 2010, 26, 16595– 16606. (24) Vindigni, F.; Manzoli, M.; Damin, A.; Tabakova, T.; Zecchina, A. Surface and inner defects in Au/CeO2 WGS catalysts: Relation between Raman properties, reactivity and morphology. Chem.—Eur. J. 2011, 17, 4356–4361. (25) Otto, K.; Hubbard, C. P.; Weber, W. H.; Graham, G. W. Raman spectroscopy of palladium oxide on r-alumina applicable to automotive catalysts. Appl. Catal., B 1992, 1, 317–327. (26) Demoulin, O.; Navez, M.; Gaigneaux, E. M.; Ruiz, P.; Mamede, A. S.; Grangerb, P.; Payen, E. Operando resonance Raman spectroscopic characterization of the oxidation state of palladium in Pd/g-Al2O3 catalysts during the combustion of methane. Phys. Chem. Chem. Phys. 2003, 5, 4394–4401. (27) Yao, H. C.; Yu Yao, Y. F. Ceria in automotive exhaust catalysis I. Oxygen storage. J. Catal. 1984, 86, 254–265. (28) Mayernick, A. D.; Janik, M. J. Methane oxidation on Pd-Ceria: A DFT study of the mechanism over PdxCe1‑xO2, Pd, and PdO. J. Catal. 2011, 278, 16–25. (29) Che, M.; Tench, A. J. Characterization and reactivity of mononuclear oxygen species on oxide surfaces. Adv. Catal. 1982, 31, 78–128. (30) Bharara, M. S.; Atwood, D. A. Oxygen: Inorganic Chemistry. Encyclopedia of Inorganic Chemistry; John Wiley & Sons: New York, 2006. (31) Sheldon, R. A.; Kochi, J. K. Photochemical and thermal reduction of cerium (IV) carboxylates: Formation and oxidation of alkyl radicals. J. Am. Chem. Soc. 1968, 90, 6688–6698. (32) Serguchev, Y. A.; Beletskaya, I. P. Oxidative decarboxylation of carboxylic acids. Russ. Chem. Rev. 1980, 49, 1119–1134. (33) Dickenson, E. R. V.; Summers, R. S.; Croue, J. P.; Gallard, H. Haloacetic acid and trihalomethane formation from the chlorination and bromination of aliphatic β-dicarbonyl acid model compounds. Environ. Sci. Technol. 2008, 42, 3226–3233.
9346
dx.doi.org/10.1021/es202209j |Environ. Sci. Technol. 2011, 45, 9339–9346
ARTICLE pubs.acs.org/est
Odorous Compounds in Municipal Wastewater Effluent and Potable Water Reuse Systems Eva Agus,† Mong Hoo Lim,‡ Lifeng Zhang,‡ and David L. Sedlak†,* † ‡
Department of Civil and Environmental Engineering, University of California, Berkeley, California 94720, United States PUB, Singapore’s National Water Agency, 228231, Singapore
bS Supporting Information ABSTRACT: The presence of effluent-derived compounds with low odor thresholds can compromise the aesthetics of drinking water. The potent odorants 2,4,6-trichloroanisole and geosmin dominated the profile of odorous compounds in wastewater effluent with concentrations up to 2 orders of magnitude above their threshold values. Additional odorous compounds (e.g., vanillin, methylnaphthalenes, 2-pyrrolidone) also were identified in wastewater effluent by gas chromatography coupled with mass-spectrometry and olfactometry detection. Full-scale advanced treatment plants equipped with reverse osmosis membranes decreased odorant concentrations considerably, but several compounds were still present at concentrations above their odor thresholds after treatment. Other advanced treatment processes, including ozonation followed by biological activated carbon and UV/H2O2 also removed effluentderived odorants. However, no single treatment technology alone was able to reduce all odorant concentrations below their odor threshold values. To avoid the presence of odorous compounds in drinking water derived from wastewater effluent, it is necessary to apply multiple barriers during advanced treatment or to dilute wastewater effluent with water from other sources.
’ INTRODUCTION In many regions facing freshwater scarcity, municipal wastewater effluent constitutes a considerable part of the potable water supply. Over the past two decades, the practice of subjecting wastewater effluent to advanced treatment—including reverse osmosis, activated carbon adsorption and chemical oxidation— has become more commonplace. The even more widespread practice of obtaining potable water supplies from effluentimpacted surface waters is also growing as population pressures place further stress on freshwater supplies. Despite the increasing importance of potable water reuse and intensified attention being given to wastewater-derived trace organic contaminants, little effort has been directed at compounds that could cause taste and odor problems in drinking water. Previous research has demonstrated that potent odorants in lakes, rivers and water distribution systems 1 6 frequently result in consumer complaints. Odorous compounds in drinking water have often been attributed to algae or bacteria in the source water or fungi in biofilms on pipe surfaces (see Supporting Information (SI) Table S1). For example, geosmin and 2-methylisoborneol have been identified as the sources of earthy odors in numerous surface waters 6 8 while the musty odor of 2,4,6trichloroanisole has been detected in rivers and water distribution systems.3,4,7 Due to the potency of these odorants, sensitive r 2011 American Chemical Society
analytical methods with gas chromatography coupled with mass spectrometry or olfactometry are often needed to identify 9 11 and quantify these compounds in drinking water supplies.12,13 Municipal wastewater effluent also contains odorants but most previous studies on wastewater-derived odors have focused on nuisance air pollution produced by wastewater treatment processes (e.g., reduced sulfides in sludge thickening).14 16 These studies have been useful in the assessment of commonly applied control measures, such as biofilters, activated carbon, and chemical oxidants,17 but they have not provided insight into the potential for wastewater-derived odorants to compromise potable water supplies. Through experience, engineers have learned that it is often necessary to use activated carbon during drinking water treatment to minimize taste and odor issues in effluent-impacted sources but few attempts have been made to quantify the wastewater-derived compounds responsible for taste and odors. To assess the occurrence and fate of odorants in potable water reuse systems, analytical techniques developed by researchers Received: July 26, 2011 Accepted: September 27, 2011 Revised: September 19, 2011 Published: October 11, 2011 9347
dx.doi.org/10.1021/es202594z | Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology studying taste and odors in drinking water and the food and beverage industry were applied to reclaimed water systems. Quantitative analysis of known potent odorants was accomplished by gas chromatography/mass spectrometry (GC/MS) while other compounds were analyzed by GC/MS-Olfactometry (GC/MS-Olf) and flavor profile analysis (FPA). To characterize the occurrence and fate of odorants, samples were collected at different stages of treatment from six full-scale advanced treatment plants. The removal of the most potent odorants was then evaluated in pilot- and bench-scale studies of different treatment processes under controlled conditions.
’ MATERIALS AND METHODS Chemical Standards. 2-Methylisoborneol, 2,3,4-trichloroanisole and 2,4,6-tribromoanisole were purchased from Dr. Ehrenstorfer Gmbh (Augsburg, Germany). 2-Bromophenol, 2,6dibromophenol, 2,4,6-tribromophenol, 2,4,6-trichlorophenol, 2,4,6-trichloroanisole, 2,3,6-trichloroanisole, β-ionone, and iodoform were purchased from Aldrich (St Quentin Fallavier, France) and Sigma-Aldrich (Saint Louis, MI). Deuterated surrogate standards (d5-geosmin and d5 2,4,6-trichloroanisole) were purchased from Cambridge Isotopes (Andover, MA). All other solvents and reagents were purchased at the highest level of purity available from Sigma-Aldrich and Merck KGaA (Darmstadt, Germany). Ultrapure deionized water (R g 18.2 MΩ-cm) was produced in-house with a Milli-Q purification system. Sample Collection. Samples were collected from six full-scale potable water reuse systems between September 2009 and February 2011 (SI Table S2). The plants had design capacities ranging from 60 to 200 ML d 1. Five rounds of bimonthly samples were collected at Plants A D while Plants E and F were sampled twice. All six advanced treatment plants received effluent from municipal wastewater treatment plants employing secondary biological treatment. In full-scale Plants A-D, incoming nitrified effluent was chlorinated with an initial concentration of approximately 2 mg/L Cl2 prior to microfiltration and reverse osmosis. The chlorine contact time between oxidant addition and the dechlorination point upstream of the reverse osmosis membrane was approximately 30 min. Plants E and F employed similar pretreatment trains except the wastewater entering the advanced treatment plants was not nitrified. After reverse osmosis, ultraviolet (UV) disinfection was employed at Plants A D at fluence values of approximately 80 mJ/cm2. UV/H2O2 was employed at Plants E and F with a fluence of approximately 500 mJ/cm2 and an initial H2O2 concentration of approximately 5 mg/L. In Plant A, ozonation (2 mg/L dose, 10 min contact time) was applied to a portion of the water after UV disinfection. Samples were also collected at a pilot plant treating denitrified municipal wastewater effluent with biological activated carbon filter (BAC) as detailed in Reungoat (2010).18 Pilot plant samples were collected during February and April 2010 before and after passage of the water through three different treatment columns: BAC without ozonation, ozonation followed by BAC, and ozonation followed by sand filtration. Before it was applied to the columns, wastewater effluent was ozonated (2 mg/L initial concentration) and subjected to coagulation, flocculation and aeration. For the two columns employing ozonation, an initial concentration of 5 mg/L O3 and a 15 min contact time was employed.
ARTICLE
All samples were collected in 1 L amber glass bottles with minimal headspace, shipped in iced coolers with overnight express service and extracted within 48 h of receipt. Samples were stored at 4 °C and were filtered (0.45 μm) prior to extraction. Field blanks, matrix spike samples and duplicates were included for analysis in all sampling rounds. Benchscale Experiments. Benchscale experiments were performed to assess the treatment efficacy of UV, UV/H2O2, chlorination, and chloramination. Secondary wastewater effluent or reverse osmosis permeate samples collected from Plants A and C were amended with target odorants at concentration approximately ten times higher than their lowest reported odor thresholds. Concentrated spiking solutions contained methanol because a number of commercial standards were only available in this solvent. Less than 50 μL of methanol was added to each 4 L sample prepared for the bench-scale experiments. Under these conditions, the steady-state concentrations of OH• are estimated to be reduced by methanol by approximately 90% and 20% in reverse osmosis permeate and secondary effluent, respectively (see SI). UV and UV/H2O2 treatments were assessed in a tubular stainless steel flow reactor (2.6 L, 15 cm o.d.) with helical internal baffles. Other than a 10-cm segment of Tygon tubing attached to the peristaltic pump, steel tubing was used to minimize losses of odorants via sorption. No loss of compounds was observed in control experiments without UV light. The reactor was equipped with two Puritec immersible low-pressure UV lamps (OSRAM, Munich, Germany) installed laterally in the center of the reactor. UV fluence was estimated from the average hydraulic residence time and photometer reading taken at quartz portholes located along the reactor. H2O2 was quantified in water flowing in and out of the reactor by KMnO4 titration.19 For chlorination and chloramination experiments, secondary effluent samples were dosed in 1-L amber glass bottles at initial concentrations of 5 and 15 mg/L as Cl2 typically applied in effluent chlorination with contact times up to 120 min. Free chlorine was added from a standardized stock solution of sodium hypochlorite. Premixed chloramine dosing solutions were made fresh daily by slowly adding sodium hypochlorite with NH4Cl at elevated pH.20 Free chlorine and monochloramine were determined using DPD colorimetric kits with a Hach DR 3800 spectrophotometer (Loveland, CO). Controls without free chlorine and chloramine indicated negligible losses of compounds. Experiments were carried out in triplicate. At the end of the experiments, excess oxidant was quenched by sodium bisulfite. Analytical Methods. Solid phase extraction of 0.45 μmfiltered samples was perfomed using a hydrophobic/hydrophilic polymeric resin (Oasis-HLB by Waters) conditioned with 5 mL methanol, 5 mL dichloromethane and 10 mL Milli-Q water. Sample pH values were adjusted to 4 5 with HCl to ensure that the weakly acidic bromophenols (pKa 7 9) and weakly basic methoxypyrazines (pKa ∼3) were present in their neutral forms. Samples were amended with 5 ng of d5-geosmin and d5 2,4,6trichloroanisole prior to extraction. Analytes were eluted from the cartridge with 10 mL dichloromethane. A sample preconcentration factor of 1000 yielded optimal instrument sensitivity while minimizing loss of the most volatile analytes. Sample extracts were concentrated to a final volume of 500 μL using a 40 °C circulating water bath and a gentle stream of ultrapure N2. Analysis was carried out with an Agilent 7890A series GC system with flow equally split between a mass spectrometer and 9348
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology an olfactory detector port (ODP). The 5975C series mass spectral detector (Agilent, Santa Clara, CA) was operated in selected ion monitoring (SIM) mode with chromatographic conditions as described in Zhang et al. (2006).12 Olfactometry was conducted with a Gerstel ODP3 (M€ulheim an der Ruhr, Germany). Sample from Plants E and F were analyzed using a Quattro micro GC triple quadrupole tandem mass spectrometer (Waters, Milford, MA) under similar chromatographic conditions. Olfactometry and flavor profile analysis (FPA) were also employed to identify other odorous compounds as described elsewhere.21 Briefly, olfactory analysis was carried out for 15 min beginning one minute after the solvent peak while, simultaneously, mass spectra were collected in full-scan mode between m/z 40 to 550. Each sample was analyzed by three members of a team of eight analysts who had been trained using reference standards and blind testing. Peak intensities of odorous compounds were classified on a scale of 0 to 4, with 4 being the strongest odor intensity. Only peaks eliciting a response of 3 (moderate intensity) or greater in 75% of the secondary effluent samples were evaluated further. Odor descriptors were categorized according to the wastewater odor wheel.22 Compounds associated with the most frequently detected odors were identified using several tools. Mass spectra were compared with the NIST mass spectral library (Agilent, Santa Clara, CA). Odor descriptions and retention times also were compared with data for compounds reported in peer-reviewed publications and public databases. Finally, compounds identified by these screening methods were compared with mass spectra, reference times and olfactometry data obtained from reference standards. Whole sample odor was assessed by sensory panels taken from the eight trained analysts using the flavor profile analysis method described in Standard Method 2170B.23
’ RESULTS AND DISCUSSION Odorous Compounds in Municipal Wastewater Effluent.
Twelve of the 15 target odorants were detected at least once in secondary effluent at concentrations up to approximately 100 ng/L (SI Table S3). The median concentrations of 2-methylisoborneol (2MIB, 11 ng/L), geosmin (27 ng/L), 2,6-dibromophenol (26DBP, 2.8 ng/L) and 2,4,6-trichloroanisole (246TCA, 9.5 ng/L) in secondary effluent were between 2 and 100 times higher than their respective odor thresholds. Another notable odorant, 2,4,6-tribromoanisole (246TBA) was detected in 40% of the secondary effluent samples at concentrations up to 6.6 ng/L. To express the concentration of odorants relative to their odor intensity, the measured concentrations were divided by the lowest reported odor thresholds (SI Table S1). This ratio, referred to as the relative odor intensity, indicates that the compounds of greatest concern detected in secondary effluent were 2,4,6-trichloroanisole and geosmin (Figure 1). The characteristic earthy and musty odors of these compounds were repeatedly detected during flavor profile analysis of secondary effluent. 2,4,6-trichloroanisole and geosmin were detected during olfactometry as strong odors—consistently scoring between 3 (moderate) and 4 (strong) during olfactometry runs—at retention times corresponding to those observed for authentic standards. The relative concentrations of the dominant target odorants in secondary effluent exhibited considerable intraplant variability
ARTICLE
Figure 1. Relative odor intensity (ROI) of common odor compounds detected in secondary effluent from municipal wastewater treatment plants.
Figure 2. Intraplant variability of common odor compounds in secondary effluent. Standard deviation was not calculated for locations E, F, and G because only two rounds of sampling were performed.
(Figure 2). 2,4,6-trichloroanisole was the dominant odorant at Plants A, B, F, and G while geosmin contributed significantly to the overall odor at Plants B, C, and D. Geosmin was the dominant odorant at Plant E, which was the only treatment plant employing a trickling filter. The intraplant variability may have been influenced by precursor concentrations in the raw sewage or by the microbial community in the biological treatment systems. Primary effluent samples collected between November 2009 and June 2010 indicated that biological wastewater treatment was a potential source for geosmin and 2,4,6-trichloroanisole (SI Table S3). In surface water supplies, geosmin is produced by a wide variety of microbes which also are commonly found in activated sludge, including cyanobacteria, actinomycetes,7 actinobacteria,24 and anabaena.25 Odors attributed to 2-methylisoborneol and geosmin have been reported in effluent from activated sludge plants treating wastes from pulp mills.2 Biological wastewater treatment was the main source of 2,4,6trichloroanisole. While primary effluent samples rarely contained the odorant (median concentration <0.38 ng/L), the compound was present in secondary effluent at a median concentration of 9.5 ng/L. 2,4,6-trichlorophenol, a potential precursor to 2,4,6trichloroanisole, was consistently detected in primary effluent. The decrease in concentration of 2,4,6-trichlorophenol during biological wastewater treatment in Plants A D was correlated with the concentration of 2,4,6-trichloroanisole detected in the secondary effluent (r2 = 0.851, SI Figure S1). 9349
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology Previous research has demonstrated that halophenols can be converted into haloanisoles in rivers 2,7 and in drinking water distribution systems.3 Fungi that biomethylate halophenols in biofilms of water distribution systems 3,7 are also present in many activated sludge microbial communities.26 To test the hypothesis that halophenols served as precursors for haloanisoles during biological wastewater treatment, batch activated sludge experiments were conducted using 13C-labeled 2,4,6-trichlorophenol and 2,4,6-tribromophenol (SI Figure S2). During a 24 h incubation period, a molar yield of 5% was observed for conversion of halophenols into their respective haloanisoles, which is consistent with observations from the full-scale municipal treatment systems. While we did not identify microbes responsible for halophenol methylation, it is evident that haloanisoles were formed during biological wastewater treatment process. The concentrations of brominated compounds such as 2,6dibromophenol, 2,4,6-tribromoanisole, and 2,4,6-tribromophenol in secondary effluent were correlated with effluent conductivity. Highest concentrations of brominated compounds were detected in Plant D, E, and F (conductivity 800 1800 μS/cm, <500 μS/cm in other treatment plants). Converting conductivity to Br concentration by assuming salt composition identical to seawater, we estimated Br concentrations of 160 to 1000 μg/L in Plant D, E, and F which is the range where brominated disinfection byproducts start to become important in chlorinated water.27,28 The water supply for Plant D includes a tidally influenced river and desalinated seawater. Water supplies at Plants E and F include a local aquifer with known seawater intrusions and bromide-rich imported water. On the basis of these results, we surmise that the halophenols may be formed when chlorine is used during sewage treatment or in household applications. In addition to the earthy/musty odors from geosmin, 2-methylisoborneol and 2,4,6-trichloroanisole, odors classified as rancid, sulfide, soapy, and fishy also were detected frequently during flavor profile analysis of secondary effluent. Characterization of secondary effluent by olfactometry yielded 15 odorous compounds which were consistently present at intensities comparable to geosmin and 2,4,6-trichloroanisole. During GC-MS/Olf, the earthy/musty odors characteristic of 2-methylisoborneol, geosmin and 2,4,6-trichloroanisole were detected at the expected retention times, confirming the identity of these three compounds. Other compounds identified during olfactometry of secondary effluent by comparison of authentic standards included 2-pyrrolidinone, methylnaphthalene isomers, hydroxyvanillin and vanillin.21 Methylnaphthalenes have recently been identified as an off-flavor in popular breakfast cereal29 and vanillin has been reported in influent and effluent from wastewater treatment plants.30 Fate of Odorous Compounds during Reverse Osmosis Treatment. Reverse osmosis treatment resulted in substantial reductions in the concentrations of potent odorants in full-scale treatment plants. Concentration of odorants quantified by GC/ MS decreased by 78 to 97%, depending on the compound (SI Table S3). For compounds analyzed by GC/olfactometry, intensity scores decreased by 1 to 3 intensity units (e.g., from an average score of 3.7 to 3.0 for 2,4,6-trichloroanisole; from 3.7 to 0.7 for xylene isomers) depending on the abundance and threshold of the compounds. Although a relatively high average rejection of 86% was observed for 2,4,6-trichloroanisole, the compound and its musty odor were still observed during GC-MS, flavor profile analysis and GC/Olf due to its abundance in
ARTICLE
feedwater and its extremely low odor threshold. Quantitative analysis indicated that the concentration of 2,4,6-trichloroanisole was 10 70 times higher than the odor threshold after reverse osmosis. For geosmin, the permeate contained concentrations close to the odor threshold. Several of the compounds identified by olfactometry—including 2-pyrrolidinone, methylnaphthalenes, vanillin, and hydroxyvanillin—also were detected in the reverse osmosis permeate. In the full-scan chromatograms, peak areas for these compounds decreased by 30 80% after reverse osmosis. Previous research indicates that low-molecular weight and neutral compounds are often not completely removed during reverse osmosis treatment.31 33 For example, removal of N-nitrosodimethylamine (NDMA, MW = 74 Da) ranged from 10 to 50% in full-scale plants with thin film composite membranes34 and 25 to 55% in benchscale systems with composite polyamide membranes.35 Monitoring of odorants before and after reverse osmosis at the treatment plants indicated average rejection of 90% and 95% for 2-methylisoborneol (MW = 168 Da) and geosmin (MW = 182 Da), respectively. Meanwhile, 2,4,6-trichloroanisole (MW = 212 Da) and 2,6-dibromophenol (MW = 252 Da) only decreased by an average of 86% and 76%, respectively. The lower removal efficiency for 2,6-dibromophenol was consistent with previous studies indicating that the phenolic moiety can enhance passage of organic compounds through reverse osmosis membranes.32 Other neutral low-molecular weight compounds detected by GC-Olfactometry (e.g., 2-pyrrolidinone, MW = 85 Da and methylnaphthalenes, MW = 108 Da) did not exhibit evidence of substantial removal during reverse osmosis treatment. Fate of Odorous Compounds during Oxidative Treatment. Oxidants used for disinfection or in advanced oxidation processes have the potential to remove odors from water.36 38 Transformation reactions produce changes in molecular structures that alter the affinity of the compounds for olfactory receptors. In some cases, such as oxidation of sulfides, oxidation eliminates the odor of the compounds. To assess the potential for removal of odorous compounds during disinfection or oxidative treatment, chlorination, ozonation, and an UV/H2O2 advanced oxidation process were evaluated (SI Table S4). Chlorine and chloramine are capable of transforming many organic compounds under the conditions employed in drinking water treatment.39 Based on the apparent reaction rate constants under circumneutral pH conditions, free chlorine (as HOCl or OCl ) should oxidize certain compounds with carbonyl groups (i.e., β-ionone, nonadienal, lactones), and to a lesser extent, phenolic groups (i.e., halophenols, vanillins). Free chlorine is not expected to be strong enough to oxidize alcohols (i.e., geosmin, 2-methylisoborneol) or haloanisoles. Chloramine is a weaker oxidant than free chlorine and could potentially react with compounds that are oxidized by chlorine at a slower rate. Samples collected from full-scale treatment plants that applied chlorine prior to reverse osmosis did not exhibit significant removal of the five frequently detected odorants (p < 0.05). However, in approximately 25% of the samples from advanced treatment plants where effluent was denitrified prior to free chlorine addition (Plants A D), the concentration of bromophenols increased. Bench-scale chlorination experiments conducted with free chlorine and NH2Cl at doses ranging from 50 to 1800 mg/L 3 min confirmed these observations (SI Table S4). Only bromophenols, β-ionone, and nonadienal were transformed during free chlorine treatment. During chloramine bench-scale experiments, only nonadienal exhibited a measurable 9350
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology decrease in concentration (by 40%) after a dose of 1800 mg/ L 3 min. The flavor profile panel reported that odors of free chlorine or chloramines masked the odors of other odorants in the wastewater effluent. GC-Olfactometry of treatment plant and benchscale experiment samples also indicated that chlorination or chloramination did not lower the odor intensity of odorous compounds in wastewater effluent. Poor removal of odorous compounds is also expected for UV treatment at recommended germicidal doses (60 100 mJ/cm2), as practiced at Plants A D. UV treatment has previously been documented to be ineffective in the removal geosmin and 2-methylisoborneol, even at doses up to 30 times higher than the germicidal dose.36 Odorous compounds with conjugated bonds (i.e., halophenols, haloanisoles, β-ionone, and nonadienal) might be more reactive during UV treatment. Furthermore, indirect photolysis enhanced by effluent organic matter might also contribute to removal of odorous compounds.40 Full-scale UV disinfection at Plants A D, at a dose of 80 mJ/ cm2, applied to permeate containing 2,6-dibromophenol, geosmin and 2,4,6-trichloroanisole did not produce detectable decreases in the concentrations of odorous compounds (p < 0.05). Similarly, flavor profile analysis and GC-olfactometry results did not show loss of any of the dominant odorants in the permeate during UV disinfection. To further evaluate the potential of UV treatment to remove odorants, wastewater effluent and reverse osmosis permeate spiked with target compounds were subjected to UV irradiation at up to 20 times the germicidal dose. As expected, the concentrations of halophenols, haloanisoles, β-ionone, and nonadienal decreased by >70% in reverse osmosis permeate after a fluence of 1000 mJ/cm2 (Figure 3). Slightly faster removal of these compounds was observed when UV treatment was conducted in secondary effluent. For 2-methylisoborneol and geosmin, removal by direct UV photolysis in permeate was minimal (<5% loss at fluence values of 1000 mJ/ cm2) while the concentration of 2,4,6-trichloroanisole decreased by approximately 20% at the same fluence. Slightly higher removal of 2-methylisoborneol, geosmin, and 2,4,6-trichloroanisole were observed in secondary effluent, presumable due to indirect photolysis. During GC-Olfactometry, the intensity of strong odor peaks (mean intensity 4) only decreased to moderate-strong levels (mean intensity 3.4) after application of
ARTICLE
1000 mJ/cm2 for a few odorants (including 5-hydroxyvanillin and vanillin). All other odorants were unaffected by UV treatment. Addition of H2O2 to the UV reactor should increase the removal of odorants through the production of hydroxyl radicals. In a previous study with geosmin and 2-methylisoborneol, UV treatment in ultrapure water resulted in a decrease of approximately 10% for both odorants. Due to the presence of methanol in the spiking solution, the rates of contaminant disappearance observed in the bench-scale studies are slower than those expected in the treatment plant, especially in RO permeate. After addition of H2O2, concentration of the compounds decreased by more than 70% at the same UV fluence.36 At plants that employed UV/H2O2 treatment (E and F), only 2-methylisoborneol and 2,4,6-trichloroanisole were detected after reverse osmosis, at concentrations up to 1.9 ng/L and 3.4 ng/L, respectively. These compounds were not detected above their method detection limits after UV/H2O2 treatment. During benchscale experiments with reverse osmosis permeate (Figure 4), nearly complete (>95%) removal of all odor compounds was observed at fluence of 1000 mJ/cm2 and an initial H2O2 concentration of 10 mg/L. As predicted by the 4-fold increase in the OH• sink terms in secondary effluent (SI Table S7), the removal of odorants was noticeably slower in secondary effluent relative to reverse osmosis permeate. GC/Olfactometry results indicated that UV/H2O2 treatment was effective in reducing the concentration of most odorant compounds below their threshold levels. For potent odorants in wastewater effluent, the intensity score decreased by at least 2 intensity units (e.g., from a mean of 3.3 to 0.3 for 2,4,6trichloroanisole, Table 1). Previous research has demonstrated the removal of odorous compounds 37,38 and halophenols41 during ozonation. Only βionone, 2,6-(E,Z)-nonadienal and halophenolate anions (present at high pH) react quickly with O3 [kO3 >104 M 1s 1].38 Geosmin, 2-methylisoborneol, and haloanisoles are transformed during ozonation mostly by OH•, making the process less effective in wastewater effluent where more OH• scavengers are present. Ozonation at Plant A (initial O3 concentration 2 mg/L, contact time 10 min) was applied on reverse osmosis permeate
Figure 3. UV treatment of odor compounds observed during benchscale experiment of spiked secondary effluent and reverse osmosis permeate at fluence 0 2000 mJ/cm2. Initial concentration Co = 50 ng/L. 9351
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology
ARTICLE
Figure 4. UV/H2O2 treatment of odor compounds observed during benchscale experiment of spiked secondary effluent and reverse osmosis permeate at UV fluence 0 2000 mJ/cm2 and 10 mg/L H2O2 dose.
containing geosmin, 2,4,6-trichloroanisole and 2,6-dibromophenol at concentrations up to 50 times the respective odor thresholds. Under these conditions, ozonation decreased the concentrations of odorants to levels below their GC-MS detection limits. The strong earthy/musty odors present in the permeate (intensity >3) were not reported by panelists in flavor profile analysis or GC-Olfactometry with the exception of 2-pyrrolidinone, which was present at a weak intensity (∼1). At the biofilter pilot plant (Plant G), preozonation (5 mg/L, 15 min) was applied to wastewater effluent that contained geosmin, 2-methylisoborneol, 2,4,6-trichloroanisole, 2,3,4-trichloroanisole, 2,4,6-tribromoanisole at concentrations up to 50 times higher than the respective odor thresholds. Under these conditions, the concentration of 2-methylisoborneol decreased by between 60 and 90% and the haloanisole concentrations decreased by approximately 40%. The odors of geosmin, 2-pyrrolidinone and lactones were still detected by the panelists during GC-olfactometry of the ozonated effluent. Fate of Odorous Compounds during Activated Carbon Treatment. Historically, granular and powder activated carbon have been used to eliminate taste and odor caused by geosmin and 2-methylisoborneol.42,43 Other odorous compounds identified in wastewater effluent generally have a similar or higher affinity for activated carbon to geosmin and 2-methylisoborneol, indicating a high potential for removal. BAC has previously been shown to remove a variety of pharmaceuticals with log Kow values above 318 with better removal observed for more readily biodegradable and hydrophobic compounds. At the BAC pilot treatment system, 2,4,6-trichloroanisole, 2-methylisoborneol and geosmin as well as 10 other odorants were detected by olfactometry in the column influent. Without ozone pretreatment (SI Table S6), BAC treatment reduced the concentration of geosmin (51 and 61%) and 2-methylisoborneol (60 and 53%). It also reduced the concentration of 2,4,6trichloroanisole from about 4 ng/L to below the method detection limit (<0.22 ng/L). When ozone was applied prior to the biofilter in Plant G, complete removal of all odor compounds (>95%) was observed. No significant odor was detected during GC-olfactometry of samples from the outlet of biofilter pretreated with ozone, while at least eight odorants (including 2-pyrrolidone, methylnaphthalene isomers, and alkyl acids) were still detected at weak intensity in BAC samples without ozonation.
Dilution and Volatilization of Odorous Compounds in Surface Waters. In many situations, secondary effluent is
discharged to surface waters that serve as potable water supplies. As indicated previously, at least 15 odorants are typically present in secondary effluent at concentrations above their odor thresholds. The dilution of secondary effluent with water free from odorous compounds could eliminate aesthetic problems downstream of the outfalls. For example, effluent containing 10 ng/L of 2,4,6-trichloroanisole (i.e., the median concentration detected in effluent samples) would need to be diluted until effluent accounted for less than 1% of the total flow before the concenontration of the compound in the source water would no longer exceed the odor threshold. Application of flavor profile analysis to diluted wastewater effluent from Plants A and C (11 and 27 ng/L 2,4,6-trichloroanisole, respectively) indicated that a weak earthy/musty odor could still be detected by panelists when effluent accounted for 3% of the sample volume. At this dilution factor, odors of 2,4,6-trichloroanisole and geosmin (intensity 2.0 3.0) were confirmed by GC-Olfactometry. In addition, weak odors at retention times corresponding to those of 2-pyrrolidinone and vanillin were detected in the diluted effluents. Assuming little removal downstream of treatment plant, the odorous compounds could pose aesthetic problems for many downstream water supplies. Volatilization of odorants during storage or downstream transport could reduce the concentrations of odorous compounds. Previous research has yielded predictive models for the fate of volatile organic compounds in rivers based on a twofilm model with or without turbulence.44 Similarly, a fugacitybased model has been developed to predict volatilization potential in reservoirs.45 In both models, the Henry’s Law constant (KH) is an indicator of volatilization potential (SI Table S5) with actual volatilization rates dependent on site-specific characteristics such as water and wind velocity, depth, temperature,44 hydraulic residence time, surface area and mixing.45 Assuming conditions typically encountered in rivers, compounds with KH > 101 Pa m3/mol are predicted to exhibit a decrease of approximately an order of magnitude during 25 km flow downstream in a river and a decrease of approximately 2 orders of magnitude during an 18-month storage period in a reservoir. Among the odorous compounds detected in wastewater effluent, the haloanisoles, crotyl mercaptan and 2,6-dibromophenol have the 9352
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology
ARTICLE
Table 1. Key GC-MS/Olfactometry Odor Peaks Detected in RO-Ozone, RO, UV/Peroxide and Ozone-BAC Treatment Trains
potential to undergo substantial losses through volatilization in surface waters (i.e., KH > 101 Pa m3/mol). However, 2-MIB, geosmin, 2-pyrrolidinone, vanillin, and hydroxyvanillin are unlikely to be substantially affected by volatilization. There are other potential mechanisms through which odorants might be attenuated in surface waters. For example, biotransformation and phototransformation of pharmaceuticals occurred with half-lives of approximately one week in the Trinity River.46 Limited information is available on the potential for odorants identified in wastewater effluent to undergo attenuation under similar mechanisms. For geosmin and 2-methylisoborneol, microbial transformation has been observed in reservoirs.8 Additional research is needed to make accurate predictions of
the potential for these compounds to undergo biotransformation and photolysis in surface waters.
’ IMPLICATIONS A suite of odorous compounds are present in wastewater effluent at concentrations well above their odor thresholds. While the presence of these compounds does not imply a health risk, their presence has the potential to pose challenges to potable water supplies. For surface waters that receive municipal wastewater effluent, substantial dilution coupled with long residence times are needed to reduce odorant concentrations to values below odor thresholds. Volatilization during storage or transit 9353
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology might be sufficient to remove haloanisoles but it will not remove less volatile odorants, such as geosmin, 2-pyrrolidone and hydroxyvanillin. To remove these odorants, downstream drinking water treatment plants may need to use activated carbon or an advanced oxidation process. Advanced treatment of secondary effluent with multiple treatment barriers—as practiced in most potable water reuse systems—is needed to reduce the concentrations of odorants to values below threshold levels. Reverse osmosis is effective in removing odorants but several may be present at concentrations above their odor thresholds in the permeate. Ozonation or UV/H2O2 can eliminate these odors from the permeate. Advanced oxidation processes (i.e., UV/H2O2) or ozonation coupled with biological activated carbon also may provide a means for removing odorous compounds even in systems that do not employ reverse osmosis. A summary of data from two full-scale advanced wastewater treatment plants and one pilot plant (Table 1) illustrates the ways in which GC-MS/Olfactometry of effluent coupled with GC/MS quantification of specific contaminants can be used to study the fate of odorants. As indicate by the olfactometry intensity scores, 2,4,6-trichloroanisole (RT = 17.0 min) and geosmin (RT = 18.5 min) are among the most persistent odorants in advanced treatment systems and can be used as indicators47 of other odors thereby avoiding the need for labor-intensive olfactometry studies. After advanced treatment is completed, any remaining compounds can be identified and quantified using the approach described above.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional figures, tables, calculations and method details are provided. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (510) 643-0256; e-mail: [email protected].
’ ACKNOWLEDGMENT We thank the PUB, Singapore’s National Water Agency for financial support. We are also grateful to PUB staff—especially Mr. Qinglin Lu and Ms. Xiaoqing Qian—for their sampling, quantitative and sensory analysis assistance. We thank Dr. Julien Reungoat at University of Queensland (Australia), Mr. Patrick Versluis at Orange County Water District and Mr. Gregg Oelker at West Basin Water Management District for field sample collection. ’ REFERENCES (1) Izaguirre, G.; Hwang, J.; Krasner, S. W. Geosmin and 2-methylisoborneol from cyanobacteria in three water supply systems. App. Environ. Microbiol. 1982, 43, 708–714. (2) Brownlee, B. G.; MacInnis, G. A.; Noton, L. R. Chlorinated anisoles and veratroles in a Canadian river receiving bleached kraft pulp mill effluent: Identification, distribution and olfactory evaluation. Environ. Sci. Technol. 1993, 27, 2450–2455. (3) Karlsson, S.; Kaugare, S.; Grimvall, A.; Boren, H.; Savenhed, R. Formation of 2,4,6-trichlorophenol and 2,4,6-trichloroanisole during treatment and distribution of drinking water. Water Sci. Technol. 1995, 31, 99–103.
ARTICLE
(4) Piriou, P.; Malleret, L.; Bruchet, A.; Kiene, L. Trichloroanisole kinetics and musty tastes in drinking water distribution systems. Water Sci. Technol.: Water Supply 2001, 1, 11–18. (5) Watson, S. Aquatic taste and odor: A primary signal of drinking water integrity. J. Toxicol. Environ. Health, Part A 2004, 67, 1779–1795. (6) Peter, A.; K€ oster, O.; Schildknecht, A.; Von Gunten, U. Occurrence of dissolved and particle-bound taste and odor compounds in Swiss lake waters. Water Res. 2009, 43, 2191–2200. (7) Jensen, S. E.; Anders, C. L.; Goatcher, L. J.; Perley, T.; Kenefick, S.; Hrudey, S. E. Actinomycetes as a factor in odor problems affecting drinking water from the North Saskatchewan River. Water Res. 1994, 28, 1393–1401. (8) Westerhoff, P.; Rodriguez-Hernandez, M.; Baker, L.; Sommerfeld, M. Seasonal occurrence and degradation of 2-methylisoborneol in water supply reservoirs. Water Res. 2005, 39, 4899–4912. (9) Young, W. H.; Horth, H.; Crane, R.; Ogden, T.; Arnott, M. Taste and odour threshold concentrations of potential potable water contaminants. Water Res. 1996, 30, 331–340. (10) Whitfield, F. B. Chemistry of off-flavours in marine organisms. Water Sci. Technol. 1988, 20, 63–74. (11) Díaz, A.; Ventura, F.; Galceran, M. T. Determination of odorous mixed chloro-bromoanisoles in water by solid-phase microextraction and gas chromatography mass detection. J Chromatogr, A 2005, 1064, 97–106. (12) Zhang, L.; Hu, R.; Yang, Z. Routine analysis of off-flavor compounds in water at sub-part-per-trillion level by large-volume injection GC/MS with programmable temperature vaporizing inlet. Water Res. 2006, 40, 699–709. (13) Salemi, A.; Lacorte, S.; Bagheri, H.; Barcel o, D. Automated trace determination of earthy-musty odorous compounds in water samples by on-line purge-and-trap gas chromatography mass spectrometry. J. Chromatogr., A 2006, 1136, 170–175. (14) Lambert, D. D.; Beaman, A. L.; Winter, P. Olfactometric characterisation of sludge odours. Water Sci. Technol. 2000, 41, 49–55. (15) Gostelow, P.; Parsons, S. A.; Stuetz, R. M. Odour measurements for sewage treatment works. Water Res. 2001, 35, 579–597. (16) Kim, K. H.; Park, S. Y. A comparative analysis of malodor samples between direct (olfactometry) and indirect (instrumental) methods. Atmos. Environ. 2008, 42, 5061–5070. (17) Harshman, V.; Barnette, T. Wastewater Odor Control: An Evaluation of Technologies. Water Eng. Manage. 2000, 147, 34–46. (18) Reungoat, J.; Macova, M.; Escher, B. I.; Carswell, S.; Mueller, J. F.; Keller, J. Removal of micropollutants and reduction of biological activity in a full-scale reclamation plant using ozonation and activated carbon filtration. Water Res. 2010, 44, 625–637. (19) Klassen, N.; Marchington, D; McGowan, H. H2O2 determination by the I3-method and by KMnO4 titration. Anal. Chem. 1994, 66, 2921–2925. (20) Mitch, W. A.; Sedlak, D. L. Formation of N-nitrosodimethylamine (NDMA) from dimethylamine during chlorination. Environ. Sci. Technol. 2002, 36, 588–595. (21) Agus, E.; Sedlak, D. L. Application of gas chromatography with mass spectrometer and olfactory detectors (GC-MS/Olfactometry) to identify odor compounds in municipal wastewater effluent and advanced water treatment. In Preparation. (22) Burlingame, G. A.; Suffet, I. H.; Khiari, D.; Bruchet, A. L. Development of an odor wheel classification scheme for wastewater. Water Sci. Technol. 2004, 49, 201–209. (23) APHA, WEF. Standard Methods for the Examination of Water and Wastewater, 19th ed.; American Public Health Association: Washington, DC, 1997 (24) Klausen, C.; Nicolaisen, M. H.; Strobel, B. W.; Warnecke, F.; Nielsen, J. L.; Jørgensen, N. O. Abundance of actinobacteria and production of geosmin and 2-methylisoborneol in Danish streams and fish ponds. FEMS Microbiol. Ecol. 2005, 52, 265–278. (25) Saadoun, I. M. K.; Schrader, K. K.; Blevins, W. T. Environmental and nutritional factors affecting geosmin synthesis by Anabaena SP. Water Res. 2001, 35, 1209–1218. 9354
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
Environmental Science & Technology
ARTICLE
(26) Bux, F.; Kasan, H. C. A microbiological survey of 10 activatedsludge plants. Water SA 1994, 20, 61–72. (27) Sun, Y. X.; Wu, Q. Y.; Hu, H. Y.; Tian, J. Effect of bromide on the formation of disinfection by-products during wastewater chlorination. Water Res. 2009, 43, 2391–2398. (28) Hua, G. H.; Reckhow, D. A.; Kim, J. S. Effect of bromide and iodide ions on the formation and speciation of disinfection byproducts during chlorination. Environ. Sci. Technol. 2006, 40, 3050–3056. (29) Schor, E. Hydrocarbons in cereal stoke new debate over food safety. In New York Times. Published: July 13, 2010. (30) Trenholm, R. A.; Vanderford, B. J.; Drewes, J. E.; Snyder, S. A. Determination of household chemicals using gas chromatography and liquid chromatography with tandem mass spectrometry. J Chromatogr., A 2008, 1190, 253–262. (31) Bellona, C.; Drewes, J. E.; Xu, P.; Amy, G. Factors affecting the rejection of organic solutes during NF/RO treatment—A literature review. Water Res. 2004, 38, 2795–2809. (32) Schafer, A. I.; Nghiem, L. D.; Waite, T. D. Removal of the natural hormone estrone from aqueous solutions using nanofiltration and reverse osmosis. Environ. Sci. Technol. 2003, 37, 182–188. (33) Agus, E.; Sedlak, D. L. Formation and fate of chlorination byproducts in reverse osmosis desalination systems. Water Res. 2010, 44, 1616–1626. (34) West Basin Municipal Water District. Investigation of N-itrosodimethylamine (NDMA) Fate and Transport; WateReuse Foundation: Alexandria, VA2006 (35) Plumlee, M. H.; Lopez-Mesas, M.; Heidlberger, A.; Ishida, K. P.; Reinhard, M. N-nitrosodimethylamine (NDMA) removal by reverse osmosis and UV treatment and analysis via LC-MS/MS. Wat. Res. 2008, 42, 347–355. (36) Rosenfeldt, E.; Melcher, B.; Linden, K. UV and UV/H2O2 treatment of methylisoborneol (MIB) and geosmin in water. J. Water Supply: Res. Technol. 2005, 54, 423–434. (37) Pei, P.; Westerhoff, P.; Nalinakumari, B. Kinetics of MIB and geosmin during ozonation. Ozone: Sci. Eng. 2006, 28, 277–286. (38) Peter, A.; Von Gunten, U. Oxidation kinetics of selected taste and odor compounds during ozonation of drinking water. Environ. Sci. Technol. 2007, 41, 626–631. (39) Deborde, M.; Von Gunten, U. Reactions of chlorine with inorganic and organic compounds during water treatment—Kinetics and mechanisms: A critical review. Water Res. 2008, 42, 13–51. (40) Pereira, V. J.; Weinberg, H. S.; Linden, K. G.; Singer, P. C. UV Degradation kinetics and modeling of pharmaceutical compounds in laboratory grade and surface water via direct and indirect photolysis at 254 nm. Environ. Sci. Technol. 2007, 41, 1682–1688. (41) Benitez, F. J.; Beltran-Heredia, J.; Acero, J. L.; Rubio, F. J. Rate constants for the reactions of ozone with chlorophenols in aqueous solutions. J. Hazard Mater. 2000, 79, 271–285. (42) Chen, G.; Dussert, B.; Suffet, I. Evaluation of granular activated carbons for removal of methylisoborneol to below odor threshold concentration in drinking water. Water Res. 1997, 31, 1155–1163. (43) Cook, D.; Newcombe, G.; Sztajnbok, P. The application of powdered activated carbon for MIB and geosmin removal: Predicting PAC doses in four raw waters. Water Res. 2001, 35, 1325–1333. (44) Rathbun, R. E. Transport, Behavior and Fate of Volatile Organic Compounds in Streams, Professional Paper 1589; United States Geological Survey: Washington, DC, 1998 (45) Hawker, D. W.; Cumming, J. L.; Neale, P. A.; Bartkow, M. E.; Escher, B. I. A screening level fate model of organic contaminants from advanced water treatment in a potable water supply reservoir. Water Res. 2011, 45, 768–780. (46) Fono, L. J.; Kolodziej, E. P.; Sedlak, D. L. Attenuation of wastewater-derived contaminants in an effluent-dominated river. Environ. Sci. Technol. 2006, 40, 7257–7262. (47) Dickenson, E. R. V.; Drewes, J. E.; Sedlak, D. L.; Wert, E. C.; Snyder, S. A. Applying surrogates and indicators to assess removal efficiency of trace organic chemicals during chemical oxidation of wastewaters. Environ. Sci. Technol. 2009, 43, 6242–6247. 9355
dx.doi.org/10.1021/es202594z |Environ. Sci. Technol. 2011, 45, 9347–9355
ARTICLE pubs.acs.org/est
Effects of Copper Nanoparticles Exposure in the Mussel Mytilus galloprovincialis T^ania Gomes,† Jose P. Pinheiro,‡ Ibon Cancio,§ Catarina G. Pereira,† Catia Cardoso,† and Maria Jo~ao Bebianno†,* †
CIMA, Faculty of Science and Technology, University of Algarve, Campus de Gambelas, 8005-139 Faro, Portugal CBME, Faculty of Science and Technology, University of Algarve, Campus de Gambelas, 8005-139 Faro, Portugal § Dept. Zoology & Animal Cell Biology, Scholl of Science and Technology, University of the Basque Country, E-48080 Bilbao, Spain ‡
ABSTRACT: CuO NPs are widely used in various industrial and commercial applications. However, little is known about their potential toxicity or fate in the environment. In this study the effects of copper nanoparticles were investigated in the gills of mussels Mytilus galloprovincialis, comparative to Cu2+. Mussels were exposed to 10 μgCu 3 L1 of CuO NPs and Cu2+ for 15 days, and biomarkers of oxidative stress, metal exposure and neurotoxicity evaluated. Results show that mussels accumulated copper in gills and responded differently to CuO NPs and Cu2+, suggesting distinct modes of action. CuO NPs induced oxidative stress in mussels by overwhelming gills antioxidant defense system, while for Cu2+ enzymatic activities remained unchanged or increased. CuO NPs and Cu2+ originated lipid peroxidation in mussels despite different antioxidant efficiency. Moreover, an induction of MT was detected throughout the exposure in mussels exposed to nano and ionic Cu, more evident in CuO NPs exposure. Neurotoxic effects reflected as AChE inhibition were only detected at the end of the exposure period for both forms of copper. In overall, these findings show that filter-feeding organisms are significant targets for nanoparticle exposure and need to be included when evaluating the overall toxicological impact of nanoparticles in the aquatic environment.
’ INTRODUCTION Nanotechnology is a rapid growing field that comprises the research and development of particles <100 nm. As nanotechnology start to come on line with larger scale production and increasing applications, it is inevitable that nanomaterials and their byproducts end up in the aquatic environment where they can induce short and long-term effects in aquatic organisms.1,2 Copper is an essential metal, with a role as a cofactor in numerous enzymes (cytochrome oxidase, superoxide dismutase, among others) that is toxic when present in higher concentrations than those necessary for organisms.36 Soluble forms of Cu have been extensively investigated on its bioavailability and effects in aquatic organisms (e.g., refs 46). In the nanoform, copper is increasingly used in various applications such as air and liquid filtration, wood preservation, bioactive coatings, and coatings on integrated circuits and batteries and thermal and electrical conductivity. Additionally, these Cu nanoparticles are applied in several products as inks, skin products, and textiles mainly due to their bactericide properties.711 The toxicity of copper oxide nanoparticles (CuO NPs) was assessed in several test organisms, namely, bacteria (Vibrio fischeri, Escherichia coli, Staphylococcus aureus, Bacillus subtilis), protozoa (Tetrahymena thermophila), crustaceans (Daphnia magna, Thamnocephalus platyurus, Daphnia pulex, and Ceriodaphnia dubia), algae (Pseudokirchneriella r 2011 American Chemical Society
subcapitata), and zebrafish (Danio rerio), showing a cytotoxic effect in all these species.710,1216 However, there is a severe lack of information on the potential effects of CuO NPs in bivalve species, as well as its behavior in aqueous environments. Despite the rapid emerging literature on the production of reactive oxygen species (ROS) and oxidative stress as main effects of NPs exposure, its mechanisms of toxicity need further clarification in invertebrate species.1,2,17,18 Recent studies have suggested that oxidative stress may be the cause of CuO NPs cytotoxicity in bacteria, daphnids, zebrafish, as well as in human lung cells.7,9,11,14,15 Biomarkers that have been used as early warning signals of the presence of contaminants in aquatic environments are important tools to assess the toxic effects of nanomaterials in aquatic organisms.1,4,5,9 Filter-feeding molluscs such as Mytilus sp. are a target group for the uptake of nanoparticles present in the aquatic environment. They have been widely used in the assessment of water quality due to their ability to accumulate conventional contaminants in the dissolved or the suspended form.1,2,7,19 Due to their filter feeding habits, bivalves Received: March 22, 2011 Accepted: September 27, 2011 Revised: September 26, 2011 Published: September 27, 2011 9356
dx.doi.org/10.1021/es200955s | Environ. Sci. Technol. 2011, 45, 9356–9362
Environmental Science & Technology gill epithelium is the main interface between the organism and the surrounding environment, being the primary pathway of exposure to environmental contaminants. In bivalves exposed to nanoparticles, gills seem to be the first targeted organ, either by direct passage or particle uptake (e.g., refs 1,2,7,14). Therefore, in this work, the toxic effects of CuO NPs in gills of mussels Mytilus galloprovincialis were evaluated using as end points biomarkers of oxidative stress (antioxidant enzymes SOD, CAT, and GPX, and lipid peroxidation), metal exposure (MT), and neurotoxicity (AChE). These effects were compared with mussels exposed to ionic Cu since the mode of action of Cu accumulation in this species is well understood.
’ EXPERIMENTAL SECTION Nanoparticles Characterization. Copper oxide nanoparticles (<50 nm) stock solution was prepared in ultrapure water, sonicated for 30 min and kept in constant shaking to reach a concentration of 10 μgCu 3 L1. Ionic copper stock solution (Cu2+) was prepared identically but not sonicated. The particles size was characterized using transmission electron microscopy (TEM) and dynamic light scattering (DLS). For TEM analysis, CuO NPs were diluted in ultrapure water and sonicated to keep the particles in solution and avoid aggregation. A drop of the dilution at 32 ppm was allowed to dry on a Ni grid cover and examined at 80 KV. The range of particles sizes was determined through analysis of 250 NPs randomly selected. Images were recorded using a JEOL JEM-1230 TEM equipped with a digital camera model 785 ES1000W Erlangshen CCD. Additionally, particle size and agglomerates, as well as behavior in natural seawater during 12 h were followed using DLS. The hydrodynamic radii of the nanoparticles were determined using an ALV apparatus with Ar ion lased (514.5 nm). Diluted particle dispersions (100 μg 3 L1 CuO NPs) were measured at 90° and intensity fluctuations analyzed automatically and in a single run by an ALV-7000 digital correlator. The temperature was controlled (20 ( 0.1 °C) using a Haake Phoenix-II heater/circulator with a C30P cooling bath, with Haake Sil 180 mineral oil. The temperature was read directly from the decalin bath using a Platinum Pt100 temperature sensor. Laboratory Exposure. Mussels Mytilus galloprovincialis (61.7 ( 8.4 mm) were collected in South of Portugal and acclimated during 7 days in natural seawater at constant temperature and aeration. Afterward, fifty mussels were placed in tanks filled with seawater in a triplicate design (around 2.5 mussels/L) and exposed to 10 μgCu 3 L1 of CuO NPs and Cu2+ along with a control group kept in clean seawater, for a period of 15 days. The copper concentration selected was environmentally relevant.5,20 Water was changed every 12 h (to avoid nanoparticles aggregation) with redosing after each change. Temperature (17.8 ( 1.1 °C), salinity (36.3 ( 0.2), oxygen saturation (97.8 ( 4.9%), and pH (7.8 ( 0.1) were measured daily. Mussels were collected from controls, CuO NPs and Cu2+ in the beginning of the experiment and after 3, 7, and 15 days of exposure. No mortality was detected during the exposure period. After sampling, gills were dissected and immediately frozen in liquid nitrogen and stored at 80 °C until further use. Metal Analysis. Copper concentrations were determined in water samples from CuO NPs and Cu2+ exposures after a period of 12 h before water renewal and redosing. Total copper concentrations from both exposures were determined after acid digestion with 2% nitric acid (HNO3), while dissolved copper from CuO NPs exposure was determined after water filtration (0.02 μm
ARTICLE
filter, Anotop 25, Whatman) and acid digestion.14 Cu in all water samples and on dried (80 °C) gills of mussels after wet digested with HNO3 were analyzed by graphite furnace atomic absorption spectrometry (AAS AAnalyst 800, Perkin-Elmer). Quality assurance was checked using a standard reference material (Lobster Hepatopancreas) provided by the National Research Council, Canada—TORT II. The mean ( standard deviation (106.8 ( 2.5 μg 3 g1) was similar to the certificated value (106.0 ( 10.0 μg 3 g1). Quality assurance was checked using a standard reference material (Lobster Hepatopancreas) provided by the National Research Council, Canada—TORT II. The mean ( standard deviation (106.8 ( 2.5 μg 3 g1) was similar to the certificated value (106.0 ( 10.0 μg 3 g1). Enzymatic Activities. Superoxide dismutase, catalase and glutathione peroxidase activities were measured in the gills cytosolic fraction. Superoxide dismutase activity (SOD) was determined by the absorption of the reduction of cytochrome c by the xanthine oxidase/hypoxanthine system at a wavelength of 550 nm.21 Catalase activity (CAT) was a result of the decrease of the absorbance at 240 nm due to hydrogen peroxide consumption, using a molar extinction coefficient of 40 M1 cm1.22 Total glutathione peroxidase (GPX) was measured following NADPH oxidation at 340 nm in the presence of excess glutathione reductase, reduced glutathione and cumene hydroperoxide as substrate.23 Metallothioneins. Gills were homogenized in three volumes of Tris-HCl buffer (0.02 M, pH 8.6) and centrifuged at 30 000g for 45 min (4 °C). The supernatant was separated from the pellet, and two aliquots were used for lipid peroxidation and total protein determination. The remaining supernatant was heattreated at 80 °C and recentrifuged at 30 000g for 45 min (4 °C). An aliquot of the heat-treated cytosol was used for the quantification of MT concentration by differential pulse polarography.24 Acetylcholinesterase. Gills were homogenized on ice in five volumes of a Tris-HCl buffer (100 mM, pH 8.0) containing 10% Triton and centrifuged at 12 000g for 30 min (4 °C). This colorimetric method is based on the coupled enzyme reaction of acetylthiocholine as the specific substrate for AChE and 5,50 dithio-bis-2-nitrobenzoate as an indicator for the enzyme reaction at 450 nm.25 Lipid Peroxidation. Lipid peroxidation (LPO) was assessed by determining malondialdehyde (MDA) and 4-hydroxyalkenals (4-HNE) concentrations upon the decomposition by polyunsaturated fatty acid peroxides using malondialdehyde bis-(tetrametoxypropan) as a standard.26 Total Protein Concentration. Total protein content of gills was measured by the Lowry method27 using Folin’s Reagent and Bovine Serum Albumin (BSA) as a standard. Statistical Analysis. The data obtained was tested using oneway analysis of variance (ANOVA) or the KruskalWallis One Way Analysis of Variance on Ranks. If significant, pairwise multiplecomparison procedures were conducted, using the Tukey test or the Dunn’s method. Linear regression was also applied, to verify existing relationships between variables. Statistical significance was set at p < 0.05 and analyses were performed using SigmaPlot10. Principal Component Analysis (PCA) was applied to evaluate the relationship between copper concentrations, antioxidant enzymes activities, MT concentrations, AChE activity and LPO levels in the gills of control and exposed mussels along the period of exposure. Computations were performed using XLStat2009. 9357
dx.doi.org/10.1021/es200955s |Environ. Sci. Technol. 2011, 45, 9356–9362
Environmental Science & Technology
ARTICLE
Figure 1. (A) Transmission electron microscopic image of CuO nanoparticles at 32 ppm in Milli-Q water. (B) Particle size distribution histogram of CuO NPs obtained from TEM images. (C) Copper concentrations in gills of mussels M. galloprovincialis from controls and exposed to CuO NPs and Cu2+ for 15 days in a dry weight tissue basis (average ( Std). Capital and lower letters represent statistical differences between treatments in each exposure day and for each treatment during the exposure duration, respectively (p < 0.05).
’ RESULTS AND DISCUSSION To our knowledge this is the first study that focused on the effects of CuO NPs in the gills of M. galloprovincialis. The nanoparticles used are spherical in shape and not strongly aggregated, with a mean diameter of 31 ( 10 nm (Figure 1A and B). The particle size distribution of CuO NPs obtained by DLS showed polydisperse aggregates (polydispersity index between 0.26 and 0.48) characterized by single particles with sizes from 30 to 40 nm to aggregates ranging from 238 to 338 nm. The higher size of CuO NPs suspended in seawater obtained by DLS compared to TEM is due to the propensity of these particles to aggregate in aqueous state. This finding is supported by other studies that used CuO NPs, some of which from the same manufacturer.7,1315,28 As the number of CuO NPs applications increase, it is likely that they will end up in the environment, and in significant quantities. However, emissions of NPs to the aquatic environment are difficult to detect and quantify, and no available data exists on CuO NPs. In our study, more than 50% of the nominal concentration of 10 μgCu 3 L1 added in the nano or ionic form was removed from the water column after the 12 h exposure (53% for CuO NPs and 66% for Cu2+). The lost of this amount of copper may be due either to the presence of the mussels, copper dissolution or nanoparticles aggregation and sedimentation.13,14 Of the total Cu concentration (4.8 ( 01 μgCu 3 L1) obtained from the CuO NPs exposure, less than 1% of the initial
added dose is present in the dissolved form, indicating that most of the Cu present in solution is in the nanoparticulate form. Other authors using CuO NPs also showed lower dissolution from nanocopper, suggesting that Cu toxicity is mainly due to CuO NPs.7,1315 The bioaccumulation of NPs in invertebrates provide valuable knowledge on NPs bioavailability and allow more realistic toxicity information.1,2,29 Data about internal exposure concentrations and accumulation of NPs in various tissues on chronic exposure of aquatic organisms is practically inexistent.2,29 In this study, the exposure to CuO NPs resulted in a significant accumulation of copper in mussel gills with time (9.8 ( 1.9 to 12.5 ( 1.4 μg 3 g1 dw, Figure 1C). In mussels exposed to Cu2+, accumulation only occurred in the first week (16.9 ( 2.4 and 15.5 ( 2.4 μg 3 g1 dw, p < 0.05), followed by a decrease at the end of the experiment, to levels similar to control (p > 0.05, Figure 1C). This decrease is indicative of the elimination rate of Cu in bivalves through detoxification processes,3,19 whereas in those exposed to CuO NPs the elimination rate is slower than its accumulation. Mussels accumulated more copper from Cu2+ than CuO NPs in the first week of exposure, suggesting a higher copper bioavailability from Cu2+. Mussel gills are a target organ for nanoparticles exposure, being more sensitive to metal dissociation from NPs than its internalization,1,2,7,14,30 nevertheless, no distinction was made between dissolved and incorporated 9358
dx.doi.org/10.1021/es200955s |Environ. Sci. Technol. 2011, 45, 9356–9362
Environmental Science & Technology
Figure 2. Superoxide dismutase (A), catalase (B), and glutathione peroxidase (C) activities in gills of mussels M. galloprovincialis from control and exposed to CuO NPs and Cu2+ for 15 days (average ( Std). Capital and lower letters represent statistical differences between treatments in each day of exposure and for each treatment during the exposure duration, respectively (p < 0.05).
copper particles. Several studies have shown copper accumulation in bivalves tissues (e.g., refs 4,5,31), however, no data exists on the accumulation of copper from NPs exposure. Chemical reactivity, as well as specific surface characteristics confers nanomaterials the capacity to generate ROS by mere interaction with subcellular structures and by directing its reactivity to subcellular compartments. In the case of metal nanoparticles, the physical contact between cells and particles may cause changes in the vicinity of the contact area and increase the dissolution of metals or generate extracellular ROS.7,9,17 Copper, being a redox active metal, has the capacity to produce ROS through Fenton-type reactions leading to the production of oxyradicals that activate/inhibit several antioxidant enzymes.39 The activities of SOD, CAT and GPX were used along with lipid peroxidation to assess the oxidative status of mussel gills exposed to CuO NPs and Cu2+ (Figures 2 and 3). SOD, CAT and GPX activities changed after exposure to CuO NPs, showing that these NPs have also potent redox properties with the capacity to generate ROS (Figure 2). In CuO NPs exposed mussels, SOD activity increased linearly (7.5 U 3 mg1prot 3 d1, r = 0.99, p < 0.05) in the first 7 days, indicative of the formation of superoxide anions. CAT was only induced after 3 days of NPs exposure (43%) while GPX activity remained unchanged and similar to unexposed mussels (p > 0.05). The induction of GPX after a week of NPs exposure
ARTICLE
Figure 3. Metallothionein concentrations (A), inhibition of acetycholinesterase activity (B) and lipid peroxidation (C) in gills of mussels M. galloprovincialis from controls and exposed to CuO NPs and Cu2+ for 15 days (average ( Std). Capital and lower letters represent statistical differences between treatments in each exposure day and for each treatment during the exposure duration, respectively (p < 0.05). Asterisks represent statistical differences between control and exposed mussels (p < 0.05).
(15.8 ( 3.1 to 21.3 ( 1.7 nmol 3 min 3 mg1prot) suggests the detoxification of hydroperoxides possibly associated with increased levels of hydroxyl radicals originated by CuO NPs, whereas at the beginning SOD and CAT levels may have been sufficient to counteract the overproduction of ROS. The SOD and CAT similar antioxidant efficiencies were supported by the PCA analysis (Figure 4A) that shows a significant correlation in the first week of exposure. After two weeks, both SOD and CAT activities decreased (38 and 33% of inhibition, p < 0.05) in mussels exposed to CuO NPs, whereas GPX continued to increase. These inhibitory effects suggest an overproduction of ROS that could have led to the degeneration of the enzymes. These ROS can be available to react with Cu2+ from CuO NPs dissolution, leading to the formation of hydroxyl radicals generated from H2O2 under Cu+ exposure through the Fenton and Haber Weiss reactions, possibly leading to SOD and CAT inactivation.5,6 These data are in line with recent observations that show that CuO NPs cytotoxicity is mediated by oxidative stress, altering the antioxidant capacity of cells against ROS. In human lung epithelial cells, CuO NPs (80 μg 3 cm2, 4 h, 30 nm) blocked the 9359
dx.doi.org/10.1021/es200955s |Environ. Sci. Technol. 2011, 45, 9356–9362
Environmental Science & Technology
Figure 4. Principal component analysis (PCA) of copper accumulation and the battery of biomarkers in gills of mussels M. galloprovincialis from controls and exposed to CuO NPs and Cu2+ for 15 days. A, PC1 vs PC2; B, PC1 vs PC3.
antioxidant defenses by inhibiting CAT and GR activities and increasing GPX or SOD and CAT activities after exposure to 10, 25, and 50 μg 3 mL1 for 24 h (52.5 ( 10.2 nm).11,28,32 In bivalves, the only existing data on antioxidant efficiency are of Cu2+ exposure. Mussels exposed to Cu2+ showed different antioxidant responses (Figure 2) with the enzymatic activities unchanged or increased (Figure 2). SOD activity was activated during the whole experiment (171% increase by day 15) resulting in the formation of superoxide radicals. CAT activity only increased after 3 days of exposure (36%) and remained unchanged from day 7 until the end of the experiment, at levels similar to controls (p > 0.05). As mentioned above, this result can be associated with the involvement of Cu in Fenton and Haber Weiss reactions, leaving no substrate available for CAT activation, or to the induction of other components of the antioxidant defense system.5,6 Like for CAT, GPX activity was induced in the first 3 days of exposure (25.0 ( 1.7 nmol 3 min 3 mg1prot, p < 0.05) remaining unchanged until the end of the experiment, always higher than that in control. This increase in GPX activity suggests a further detoxification of ROS combined with the action of MT; either by ROS scavenging (day 7) or Cu detoxification (day 15), justifying CAT unaltered activities. The PCA analysis shows a clear association between GPX activity and Cu2+-exposed mussels, validating the enhancement of this enzyme activity to neutralize ROS (Figure 4A). Similar results were detected in mussels exposed to 60 μgCu 3 L17 for 3 weeks and in the clam R. decussatus exposed to 0.5 and 2.5 μgCu 3 L1 Cu for 3 days.5
ARTICLE
Metallothioneins are low-molecular weight cysteine-rich proteins induced by metals that can also act as an oxygen species scavengers, participating in antioxidant processes protecting cells from oxidative stress.19,20,31 Although information on MT behavior upon exposure to CuO NPs is nonexistent, the role of MT in ionic/soluble Cu detoxification mechanisms is well understood in bivalves, either by controlling its intracellular availability or by detoxifying excessive metal concentrations.4,5,20,31,33 In mussels exposed to CuO NPs, MT increased linearly with time of exposure, with an induction rate of 0.3 mg 3 g1prot 3 d1 (r = 0.99, p < 0.05), reflecting not only the role of this protein in Cu homeostasis and detoxification (Figure 3A), but also a possible involvement in gills antioxidant defense system that can explain the absence of SOD and CAT responses (day 15). Only two studies addressed the role of MT in bivalve species: in C. virginica exposed to silver nanoparticles (16 μg 3 L1-1.6 ng 3 L1, 15 ( 6 nm) an increase in MT expression was associated with silver metabolism or to the increase of oxyradicals and in C. fluminea exposed to gold nanoparticles (1.6 1031.6 105 Au NP/cell, 10 nm) to protect cells against gold-induced oxidative stress.34,35 In mussels exposed to Cu2+, MT levels also increased in the first week of exposure with a lower induction rate (0.2 mg 3 g1prot 3 d1, r = 0.99, p < 0.05) when compared to CuO NPs (Figure 3A), denoting its importance in Cu metabolism, as also seen by the close association between Cu concentrations and MT in the PCA (Figure 4). Contrarily to the response for CuO NPs, MT decreased in the gills of mussels exposed to Cu2+ at the end of the experiment (6.7 mg 3 g1prot), suggesting a role of MT in copper detoxification, which is in agreement with the copper accumulation results in mussel gills (Figure 1C). Cu can bind to MT to form insoluble CuMT complexes that precipitate into lysosomes and are eliminated by exocytosis.3,20,31 Similar results were detected in R. decussatus31 and Crassostrea gigas20 exposed to 50 μgCu 3 L1 and 0.55 μgCu 3 L1, respectively. Acetylcholinesterase is a biomarker of exposure to organophosphorus pesticides that can also be inhibited by a diverse range of metals, including copper.4,5,33 A dose-dependent decrease of this enzyme after Cu2+exposure is well established in bivalve species, as in R. decussatus (75 μgCu 3 L1, 5 days)5 and mussels (40 μg 3 L1 and 60 μg 3 L1, 1 and 3 weeks).33,4 In this study, inhibition of AChE was observed in CuO NPs and Cu2+ exposed mussels (Figure 3B) only at the end of the experiment, with a 34% and 53% inhibition, respectively (p < 0.05), also confirmed by the PCA (Figure 4). The high affinity of Cu to sulfur donor groups can cause AChE inhibition by binding to its thiol residues, as in MT.5 These results confirm the specificity of AChE response to Cu exposure, either in the nano or ionic form. The neurotoxic effects of nanoparticles in M. edulis exposed to 1 mg 3 L1 Fe NPs (590 nm, 12 h) showed no significant differences in AChE activity.36 Nevertheless, one study showed that AChE has the potential to be used as a biomarker for CuO NPs (25 nm), because of its strong AChE inhibition (76%) and low median inhibitory concentration (4 mg 3 L1).37 Significant variations of enzymatic activities exist between control and Cu-exposed mussels throughout the experimental period suggesting that gills responded differently to both forms of copper (Figure 2). The overall PCA analysis (Figure 4) indicates a clear separation between control and Cu-exposed mussels. Unexposed mussels, as well as those exposed to Cu2+ are closely associated at different times of exposure (day 3, 7, and 15) showing similar biomarker tendency. As for CuO NPs exposed mussels, a clear separation of the sampling periods 9360
dx.doi.org/10.1021/es200955s |Environ. Sci. Technol. 2011, 45, 9356–9362
Environmental Science & Technology occurred, suggesting a marked different behavior between mussel gills response with time of exposure. Failure of antioxidant defenses to counteract ROS produced by both forms of Cu either by being inhibited or overwhelmed can interrupt the balance between the antioxidant/prooxidant system in mussels leading to oxidative damage of biomolecules.46 One of the best known effects of excess Cu is the peroxidative damage to membrane lipids, triggered by the reaction of lipid radicals and oxygen to form peroxyl radicals that can alter membrane fluidity and permeability or attack other intracellular molecules.46 Despite different antioxidant efficiency, LPO increased linearly with time in mussels exposed to CuO NPs and Cu2+ (Figure 3C), with induction rates of 36.8 nmol 3 g1 prot 3 d1 (r = 0.99; p < 0.05) and 49.7 nmol 3 g1prot 3 d1 (r = 0.97, p < 0.05), respectively. In the first three days of CuO NPs exposure, SOD and CAT activities proved to be antioxidant efficient and prevent deleterious effects in lipids of cellular membranes, confirmed by the relative proximity of these mussels to the control group in the PCA analysis (Figure 4A). In the remaining period, CuO NPs seems to continuously increase ROS production activating the combined action of antioxidant defenses (SOD, CAT, GPX, and MT) until a point where the antioxidant capacity was overwhelmed causing SOD and CAT inactivation and a continuous MT and GPX increase. Although GPX and MT can remove most of the ROS by increasing its activities, they cannot compete with hydroxyl radicals’ generation via the Fenton reaction thereby causing an increase in LPO levels. In mussels exposed to Cu2+, antioxidant enzymes were activated during the whole exposure period (except CAT) along with an increase in MT levels leading to a detoxification process by the end of the exposure, nevertheless, not enough to prevent LPO. These results are in agreement with the PCA that shows a clear association between copper concentrations in gills and LPO levels, as well as with MT and GPX (Figure 4). In human cells and E. coli exposed to CuO NPs (3050 nm) their toxicity was related to oxidative stress, mediated by lipid peroxidation, oxidative lesions and increase of intracellular ROS11,13,15,28,32 Evidence that LPO occurs after Cu exposure was also observed in several bivalve species, as clams and mussels.46 Altogether, our results support the conclusion that oxidative stress is a significant mechanism of toxicity for CuO NPs7,9,11,15,32,38 and that its mode of action appears distinct from Cu2+. In other aquatic organisms (V. fisheri, D. magna, T. platyurus, P. subcapitata, T. thermophila) CuO NPs (∼30 nm) showed a higher toxicity when compared to its ionic/soluble form, associated with Cu ions dissolution.9,10,12 Nevertheless, the dissolution of Cu ions do not fully explain the toxicity of CuO NPs in zebrafish,7,14 human cell cultures,11,29,32 or daphnids15 exposed to particles with similar size (3050 nm), where other mechanisms derived from the particle effect had to be considered (e.g., oxidative stress due to ROS formation). In our study, a combination of the particle effect and ions dissolution can account for the differences in the toxic effects exerted by CuO NPs along the exposure period. Mussel gills can be taking up dissolved Cu released from the particles combined with a cellular uptake of nanoparticles aggregates. CuO NPs can pass the cellular membrane, enter inside the cell, dissolve rapidly and release high concentrations of ions sufficient to disrupt Cu homeostasis and generate radicals.1,14,28,38 This NPs mechanism of toxicity named “Trojan horse-type mechanism” was identified in cell cultures.38,39 The increasing copper concentrations in mussel gills can be indicative of an increasing rate of exposure that leads to a continuous release of Cu from the NPs. The reaction time of gills cells is slower than the particle dissolution and uptake leading to enzymatic
ARTICLE
breakdown and to a continuous increase in MT levels, whereas in mussels exposed to Cu2+, this metal is eliminated more rapidly via MT detoxification pathway. Another fraction of the CuO NPs can be taken up by endocytosis and their toxicological response controlled by surface processes (ROS, adsorption).1,2,38 The presence of CuO NPs aggregates in suspension (as seen by DLS) facilitate a continuous source of NPs that can either be dissolved or incorporated, leading to a continuous ROS generation (intra and/or extracellular), that increases with time of exposure. A correlation between formation of larger aggregates and biomarker responses with increasing time of exposure was suggested in M. galloprovincialis exposed to nano carbon black, C60 fullerenes, nano-TiO2 and nano-SiO2.40 A more efficient and rapid capture and ingestion of NPs in aggregated form was also observed in mussels and oysters exposed to polystyrene NPs when compared to those in suspension.29 As for M. edulis, NPs from glass wool and Fe are taken up by gills epithelial cells as pathways of uptake by diffusion or by endocytosis, independently of the size of the aggregates.37,41 Aggregation has a crucial role in nanoparticles toxicity, and the cumulative effects of the dissociation of metal ions, size and surface-area properties of these particles cannot be discarded and need further clarification in CuO NPs mechanisms.1,2,29,35,40 Despite the information given by acute experiments, they do not provide complete information about the interactions of nanomaterials with classical test species and there is a need to direct research toward invertebrate tests using long-term exposure to better understand NPs toxicity mechanisms.2,9,13,15 As for CuO NPs, most of the data available concerns acute toxicity across a wide spectrum of aquatic species,710,1216 and this study is one of the first to address long-term effects of these NPs in this species. Overall our results show that mussels represent a target for environmental exposure to nanoparticles where exposure duration may be a contributing factor in NPs mediated toxicity. In summary, long-term exposure to CuO NPs cause oxidative stress in gills of mussels as evidenced by the breakdown of the antioxidant defense system and lipid peroxidation, as well as acetylcholinesterase inhibition and metallothionein induction. Nevertheless the underlying mechanisms associated with biomarkers responses are still uncertain, and the observed oxidative stress may due to an association between the nanoparticle effect and the dissociation of copper ions from the nanoparticles. Future research is required to understand the mechanisms of CuO NPs toxicity in aquatic organisms, where the uptake and accumulation of CuO NPs in other mussel tissues should be considered, as well as the importance of bioavailability and particle aggregation for long periods of time.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (+351) 289800100; fax: (+351) 289800069; e-mail: [email protected].
’ ACKNOWLEDGMENT This research was supported by a Foundation of Science and Technology PhD Grant (SFRH/BD/41605/2007). ’ REFERENCES (1) Moore, M. N. Do nanoparticles present ecotoxicological risks for the health of the aquatic environment? Environ Int. 2006, 32, 967–976. 9361
dx.doi.org/10.1021/es200955s |Environ. Sci. Technol. 2011, 45, 9356–9362
Environmental Science & Technology (2) Baun, A.; Hartmann, N. B.; Grieger, K.; Kusk, K. O. Ecotoxicity of engineered nanoparticles to aquatic invertebrates: A brief review and recommendations for future toxicity testing. Ecotoxicol. 2008, 17, 387–395. (3) da Silva, J. J. R. F.; Williams, R. J. P. The Biological Chemistry of the Elements: The Inorganic Chemistry of Live, 2nd ed.; Oxford University Press: New York, 2001. (4) Regoli, F.; Principato, G. Glutathione, glutathione-dependent and antioxidant enzymes in mussel, Mytilus galloprovincialis, exposed to metals under field and laboratory conditions: Implications for the use of biochemical biomarkers. Aquat. Toxicol. 1995, 31, 143–164. (5) Bebianno, M. J.; Geret, F.; Hoarau, P.; Serafim, M. A.; Coelho, M. R.; Gnassia-Barelli, M.; Romeo, M. Biomarkers in Ruditapes decussatus: A potential biondicator species. Biomarkers 2004, 9 (45), 305–330. (6) Maria, V. L.; M. J. Bebianno, M. J. Antioxidant and lipid peroxidation responses in Mytilus galloprovincialis exposed to mixtures of benzo(a)pyrene and copper. Comp. Biochem. Pharmacol. C 2011, 154 (1), 56–63. (7) Griffitt, R. J.; Weil, R.; Hyndman, K. A.; Denslow, N. D.; Powers, K.; Taylor, D.; Barber, S. D. Exposure to copper nanoparticles causes gill injury and acute lethality in zebrafish (Danio rerio). Environ. Sci. Technol. 2007, 41, 8178–8186. (8) Yoon, K. Y.; Byeon, J. H.; Park, J. H.; Hwang, J. Susceptibility constants of Escherichia coli and Bacillus subtilis to silver and copper nanoparticles. Sci. Total Environ. 2007, 373, 572–575. (9) Heinlaan, M.; Ivask, A.; Blinova, I.; Dubourguier, H. C.; Kahru, A. Toxicity of nanosized and bulk ZnO, CuO and TiO2 to bacteria Vibrio fischeri and crustaceans Daphnia magna and Thamnocephalus platyurus. Chemosphere 2008, 71, 1308–1316. (10) Aruoja, V.; Dubourguier, H. C.; Kasemets, K.; Kahru, A. Toxicity of nanoparticles of CuO, ZnO and TiO2 to microalgae Pseudokirchneriella subcapitata. Sci. Total Environ. 2009, 407 (4), 1461–1468. (11) Fahmy, B.; Cormier, S. A. Copper oxide nanoparticles induce oxidative stress and cytotoxicity in airway epithelial cells. Toxicol. In Vitro 2009, 23, 1365–1371. (12) Mortimer, M.; Kasemets, K.; Kahru, A. Toxicity of ZnO and CuO nanoparticles to ciliated protozoa Tetrahymena thermophila. Toxicol. 2010, 269 (23), 182–189. (13) Griffitt, R. J.; Luo, J.; Gao, J.; Bonzongo, J. C.; Barber, D. S. Effects of particle composition and species on toxicity of metallic nanomaterials in aquatic organisms. Environ. Toxicol. Chem. 2008, 27 (9), 1972–1978. (14) Griffitt, R. J.; Hyndman, K.; Denslow, N. D.; Barber, D. S. Comparison of molecular and histological changes in zebrafish gills exposed to metallic nanoparticles. Toxicol. Sci. 2009, 107 (2), 404–415. (15) Heinlaan, M.; Kahru, A.; Kasemets, K.; Arbeille, B.; Prensier, G.; Dubourgier, H. C. Changes in the Daphnia magna midgut upon ingestion of copper oxide nanoparticles: A transmission electron microscopy study. Water Res. 2011, 45, 179–190. (16) Ruparelia, J. P.; Chatterjee, A. K.; Duttagupta, S. P.; Mukherji, S. Strain specificity in antimicrobial activity of silver and copper nanoparticles. Acta Biomat. 2008, 4 (3), 707–716. (17) Unfried, K.; Albrecht, C.; Klotz, L.; Mikecz, A. V.; GretherBeck, S.; Schins, R. P. F. Cellular responses to nanoparticles: Target structures and mechanisms. Nanotoxicol. 2007, 1 (1), 52–71. (18) Xia, T.; Kovochich, M.; Brant, J.; Hotze, M.; Sempf, J.; Oberley, T.; Sioutas, C.; Yeh, J. I.; Wiesner, M. R.; Nel, A. E. Comparison of the abilities of ambient and manufactured nanoparticles to induce cellular toxicity according to an oxidative stress paradigm. Nano Lett. 2006, 6 (8), 1794–1807. (19) Langston, W. J.; et al. Metal handling strategies in molluscs. In Metal metabolism in Aquatic Environments; Langston, W. J., Bebianno, M. J., Eds.; Kluwer Academic Publishers: 1998; 219283. (20) Damiens, G.; Mouneyrac, C.; Quiniou, F.; His, E.; Gnassia-Barelli, M.; Romeo, M. Metal bioaccumulation and metallothionein concentrations in larvae of Crassostrea gigas. Environ. Pollut. 2006, 140, 492–499. (21) McCord, J. M.; Fridovich, I. Superoxide dismutase: An enzymatic function for erythrocuprein (hemocuprein). J. Biol. Chem. 1969, 244 (22), 6049–6955. (22) Greenwald, R. A. Handbook of Methods for Oxygen Radical Research; CRC Press: Boca Raton, FL 1985.
ARTICLE
(23) Lawrence, R. A.; Burk, R. F. Glutathione peroxidase activity in selenium-deficient rat liver. Biochem. Biophys. Res. Commun. 1976, 71, 952–958. (24) Bebianno, M. J.; Langston, W. J. Quantification of metallothioneins in marine invertebrates using differential pulse polarography. Port. Electrochim. Acta 1989, 7, 59–64. (25) Ellman, G. L.; Courtney, K. O.; Anders, V.; Featherstone, R. M. A new and rapid colorimetric determination of acetylcholinesterase activity. Biochem. Pharmacol. 1961, 7, 88–95. (26) Erdelmeier, I.; Gerard-Monnier, D.; Yadan, J. C.; J. Acudiere, J. Reactions of N-methyl-2-phenylindole with malondialdehyde and 4-hydroxyalkenals. Mechanistic aspects of the colorimetric assay of lipid peroxidation. Chem. Res. Toxicol. 1998, 11, 1184–1194. (27) Lowry, O. H.; Rosenbrough, N. J.; Farr, A. L.; Randall, R. J. Protein measurement with the Folin phenol reagent. J. Biol. Chem. 1951, 193, 265–275. (28) Karlsson, H. L.; Cronholm, P.; Gustafsson, J.; M€ oller, L. Copper oxide nanoparticles are highly toxic: A comparison between metal oxide nanoparticles and carbon nanotubes. Chem. Res. Toxicol. 2008, 21, 1726–1732. (29) Ward, J. E.; Kach, D. J. Marine aggregates facilitate ingestion of nanoparticles by suspension-feeding bivalves. Mar. Environ. Res. 2009, 68, 137–142. (30) Peyrot, C.; Gagnon, C.; Gagne, F.; Willkinson, K. J.; Turcotte, P.; Sauve, S. Effects of cadmium telluride quantum dots on cadmium bioaccumulation and metallothionein production to the freshwater mussel, Elliptio complanata. Comp. Biochem. Physiol. C 2009, 150, 246–251. (31) Serafim, A.; Bebianno, M. J. Metallothionein role in the kinetic model of copper accumulation and elimination in the clam Ruditapes decussatus. Environ. Res. 2009, 109, 390–399. (32) Ahamed, M.; Siddiqui, M. A.; Akhtar, M. J.; Ahmad, I.; Pant, A. B. Genotoxic potential of copper oxide nanoparticles in human lung epithelial cells. Biochem. Biophys. Res. Co. 2010, 396, 578–583. (33) Lehtonen, K. K.; Leini€o, S. Effects of exposure to copper and malathion on metallothionein levels and acetylcholinesterase activity of the mussel Mytilus edulis and the clam Macoma balthica from the Northern Baltic Sea. Bull. Environ. Contam. Toxicol. 2003, 71, 489–496. (34) Renault, S.; Baudrimont, M.; Mesmer- Dudons, N.; Gonzalez, P.; Mornet, S.; Brisson, A. Impacts of gold nanoparticle exposure on two freshwater species: A phytoplanktonic alga (Scenedesmus subspicatus) and a benthic bivalve (Corbicula fluminea). Gold Bullet. 2008, 41 (2), 116–126. (35) Ringwood, A. H.; McCarthy, M.; Bates, T. C.; Carroll, D. L. The effects of silver nanoparticles on oyster embryos. Mar. Environ. Res. 2009, 69 (1), 549–551. (36) Kadar, E.; Lowe, D. M.; Sole, M.; Fisher, A. S.; Jha, A. N.; Readman, J. W.; Hutchinson, T. H. Uptake and biological responses to nano-Fe versus soluble FeCl3 in excised mussel gills. Anal. Bioanal. Chem. 2010, 396, 657–666. (37) Wang, Z.; Zhao, J.; Li, F.; Gao, D.; Xing, B. Adsoprtion and inhibition of acetylcholinesterase by different nanoparticles. Chemosphere 2009, 77 (1), 67–73. (38) Studer, A. M.; Limbach, L. K.; Duc, L. V.; Krumeich, F.; Athanassiou, E. K.; Gerber, L. C.; Moch, H.; Stark, W. J. Nanoparticle cytotoxicity depends on intracellular solubility: Comparison of stabilized copper metal and degradable copper oxide nanoparticles. Toxicol. Lett. 1995, 197, 169–174. (39) Limbach, L. K.; Wick, P.; Manser, P.; Grass, R. N.; Bruinink, A.; Stark, W. J. Exposure of engineered nanoparticles to human lung epithelial cells: Influence of chemical composition and catalytic activity on oxidative stress. Environ. Sci. Technol. 2007, 41, 4158–4163. (40) Canesi, L.; Fabbri, R.; Vallotto, D.; Marcomini, A.; Pojana, G. Biomarkers in Mytilus galloprovincialis exposed to suspensions of selected nanoparticles (Nano carbon black, C60 fullerene, Nano-TiO2, Nano-SiO2). Aquat. Toxicol. 2010, 100, 168–177. (41) Koehler, A.; Marx, U.; Broeg, K.; Bahns, S.; Bressling, J. Effects of nanoparticles in Mytilus edulis gills and hepatopancreas—A new threat to marine life? Mar Environ. Res. 2008, 66 (1), 12–14. 9362
dx.doi.org/10.1021/es200955s |Environ. Sci. Technol. 2011, 45, 9356–9362
ARTICLE pubs.acs.org/est
Perchlorate Uptake in Spinach As Related to Perchlorate, Nitrate, And Chloride Concentrations in Irrigation Water Wonsook Ha,*,†,§ Donald L. Suarez,† and Scott M. Lesch‡ † ‡
U.S. Salinity Laboratory, USDA-ARS, 450 W. Big Springs Rd., Riverside, California 92507, United States Statistical Consulting Collaboratory, University of California-Riverside, Riverside, California 92507, United States ABSTRACT: Several studies have reported on the detection of perchlorate (ClO4) in edible leafy vegetables irrigated with Colorado River water. However, there is no information on spinach as related to ClO4 in irrigation water nor on the effect of other anions on ClO4 uptake. A greenhouse ClO4 uptake experiment using spinach was conducted to investigate the impact of presence of chloride (Cl) and nitrate (NO3) on ClO4 uptake under controlled conditions. We examined three concentrations of ClO4, 40, 220, and 400 nmolc/L (nanomoles of charge per liter of solution), three concentrations of Cl, 2.5, 13.75, and 25 mmolc/L, and NO3 at 2, 11, and 20 mmolc/L. The results revealed that ClO4 was taken up the most when NO3 and Cl were lowest in concentration in irrigation water. More ClO4 was detected in spinach leaves than that in the root tissue. Relative to lettuces, spinach accumulated more ClO4 in the plant tissue. Perchlorate was accumulated in spinach leaves more than reported for outer leaves of lettuce at 40 nmolc/L of ClO4 in irrigation water. The results also provided evidence that spinach selectively took up ClO4 relative to Cl. We developed a predictive model to describe the ClO4 concentration in spinach as related to the Cl, NO3, and ClO4 concentration in irrigation water.
’ INTRODUCTION Perchlorate (ClO4) salt is used as an oxidizing agent in rocket propellants and explosives.1 Perchlorate has been detected in various water sources, both surface and groundwater, as well as in wine, beverages, baby formula, breast milk, and leafy vegetables.26 Perchlorate has been found in ground and surface water in 35 states in the U.S. Currently, a drinking water standard for ClO4 has been set by U.S. Environmental Protection Agency7 and a few states have established advisory levels (for example, 5 μg/L in New York, 6 μg/L as a maximum contaminant level or public health goal in California, and 14 μg/L in Arizona, ref 8 and 9). Perchlorate salts are very soluble in water. Once dissolved, the ClO4 anion is chemically very stable having a +7 oxidation state and persisting in the environment because of high activation energy necessary for reduction.10 The main health concern for ClO4 ion is that it substitutes iodine (similar charge and ionic radius) and thus interrupts thyroid iodine uptake in human beings11 resulting in subsequent hormone disruption and potential perturbations of metabolic activities.12 Perchlorate in water is of concern due to impact on ecosystems13,14 and an additional pathway for humans’ intake via accumulation in vegetables from irrigation water.1517 Elevated concentrations of ClO4 have been detected in various groundwater sources, related to the release of ammonium ClO4 by military operations, the aerospace industry, and among others. Perchlorate in Colorado River water has been related to ClO4 contamination by the ClO4 salt manufacturing plant previously located near the Las Vegas wash in Nevada.10,1822 The fresh vegetable industry relies on Colorado River water for r 2011 American Chemical Society
irrigation in the lower Colorado River regions of California and Arizona. Use of Colorado River water thus caused elevated ClO4 concentrations in vegetables.20,23 More recently, installation of a treatment plant on Las Vegas wash has subsequently reduced the ClO4 concentration of the Colorado River.24 Spinach has above-ground parts consumed by humans; which makes it a good choice to study ClO4 uptake. The interaction between salts and ClO4 in edible plants when ClO4 is taken up by plant is not fully investigated. Leaf chloride (Cl) declined from 4.37 to 2.43% Cl as NO3 increased from 3 to 15 mmolc/ L in wheat (Triticum aestivum) leaves.25 Net uptake rate of NO3 in Plantago maritima L. was reduced by 23, 33, and 51% at 50, 100, and 200 mol/m3 NaCl, respectively.26 Tan and others14 reported that the uptake of ClO4 in smartweed (Polygonum spp.) was not greatly affected by the presence of NO3-N, SO42‑, PO43‑, or Cl in 500 mg L1 solution. However, ClO4 uptake in three different types of lettuce as independently affected by NO3, SO42‑, Cl, pH, and HCO3 was evaluated by Seyfferth et al.27 They concluded that increasing solution NO3 markedly decreased ClO4 uptake but observed no severe effect of Cl on ClO4 uptake in lettuce leaves. The combined effect of NO3 and Cl ions on ClO4 uptake has not been examined although both NO3 and Cl are present at varying concentrations in the soilwater during the crop growing season. Also, the accumulation pattern of ClO4 in root Received: March 25, 2011 Accepted: September 22, 2011 Revised: August 23, 2011 Published: September 22, 2011 9363
dx.doi.org/10.1021/es2010094 | Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology tissues has not been thoroughly investigated under the presence of Cl and NO3 salts in irrigation water. To date, the research focus has been on lettuce and there are almost no data on ClO4 accumulation of other leafy vegetables such as spinach, as related to ClO4 concentration in irrigation water, and no information on the effects of ions that may potentially inhibit ClO4 uptake in these other leafy vegetables. The objectives of this study are (1) investigate the uptake of ClO4 by spinach as related to ClO4 in irrigation water, (2) evaluate the effect of NO3 and Cl anions on the uptake of ClO4 by spinach, (3) examine the physiological effect of ClO4 uptake by measuring ClO4 concentration in both leaf and root parts and determine the pattern of translocation of ClO4 within plant materials, and (4) develop predictive equations to represent ClO4 uptake in spinach as related to ClO4, NO3, and Cl in irrigation water.
’ MATERIALS AND METHODS 1. Greenhouse Experiment. The experiment was conducted in 30 sand tanks at the greenhouse facility in U.S. Salinity Laboratory, Riverside, CA. Washed sand (average bulk density: 1.4 Mg/m3) was contained in sand tanks (1.2 0.6 0.5 m depth each). After filling up the water reservoir with deionized (DI) water, the water was circulated through the sand several times to ensure equilibration of the saturation within the sand media. The electric conductivity (EC) was monitored before initiating irrigation to ensure low EC in the DI water. Sorption of ClO4 onto the surface of the sand particles and container was determined to be negligible and thus is not considered further. There were 10 different combinations of irrigation water treatments. Each treatment was replicated three times (Table 1). Three randomly selected sand tanks were irrigated with each water composition during the experiment. We utilized three different concentrations of ClO4, 40, 220, and 400 nmolc/L, (nanomoles of charge per liter of solution; approximately 4, 22, and 40 μg/L, respectively), Cl at 2.5, 13.75, and 25 mmolc/L, and NO3 at 2, 11, and 20 mmolc/L. Half Hoagland’s solution (plant nutrient solution) was prepared and the concentration in the irrigation reservoir in mmolc/L was: 0.17 KH2PO4, 0.75 MgSO4 3 7H2O, 2.0 KNO3, 0.25 CaSO4. We utilized DI water and prepared the solutions with reagent grade salts. The pH of the irrigation water ranged between 7.7 and 8.5. The experiment was designed as a classic 23 factorial design, with a center point and one additional nonstandard point set to high ClO4, medium NO3, and medium Cl level. Seeds of hybrid spinach (Spinacia oleracea L., cv. “Space”) were purchased from Johnny’s seeds (Winslow, ME) and planted in November 2007. The plants were irrigated in the sand tanks twice a day at 0900 and 1300 h, saturating the sand, and ensuring a uniform root zone solution composition. Irrigation time was 45 min per event. After each irrigation event, the nutrient solution drained back into the 890 L reservoirs below the sand tanks for a subsequent reuse. The ion concentrations in the reservoirs were constantly maintained by supplementing the nutrients every other week, back to the initial nutrient levels. Water loss by evapotranspiration (ET) was replenished by adding DI water back to the reservoirs to maintain constant volumes and osmotic potentials in each reservoir. Spinach was grown for 71 days under controlled greenhouse conditions which are 41% of relative humidity and 18 and 15 °C of day and nighttime temperatures, respectively. Spinach leaves
ARTICLE
Table 1. Initial Concentrations of ClO4 (nmolc/L), Cl (mmolc/L), and NO3 (mmolc/L) in Irrigation Water reservoir number
ClO4
Cl
NO3
number of replications
1
40
2.5
2
3
2 3
400 40
2.5 25
2 2
3 3
4
400
25
2
3
5
40
2.5
20
3
6
400
2.5
20
3
7
40
25
20
3
8
400
25
20
3
9
220
13.75
11
3
10
400
13.75
11
3
and root tissue samples were harvested from each sand tank at the end of the experiment. 2. Plant Tissue Processing and Perchlorate Extraction. The plant tissue extraction procedure from Seyfferth and Parker28 was utilized, but weight of plant and volume of water added were modified. The leaves were frozen right after harvesting and stored in the freezer. Approximately 25.0 g of frozen plant sample was weighed and 80.0 mL of DI water was added to plant material for grinding. All standard solutions were made with at least 17.8 MΩ water. After grinding samples, plant material was transferred in 250.0 mL HDPE Nalgene bottle (Nalge Nunc International, Rochester, NY) for 4 h of shaking to release any remaining ClO4 to solution. Samples were centrifuged at 5400 RCF (relative centrifugal force) for 1 h. We filtered approximately 30.0 mL of supernatant using 0.2 μm cellulose NO3 membrane filters (Whatman International Ltd., Maidstone, England). We took approximately 3.0 mL of filtered aliquot and passed it through a preconditioned ENVI-18 SPE cartridge, discarding the first 1.0 mL of sample and collecting 2.0 mL of liquid sample in the glass tube for ClO4 analysis. Perchlorate standard solutions were made from reagent grade sodium perchlorate (NaClO4, Aldrich Chemical Co., Inc., Milwaukee, WI) having density of 2.0 g/cm3 and molecular weight of 122.44 g/mol.
3. ANALYSIS 3.1. Perchlorate Analysis of Plant Samples and Irrigation Water. Perchlorate was analyzed using an Agilent 1100 series
high performance liquid chromatography/mass spectrometry (HPLC/MS). Detailed method for ClO4 analysis utilized for this analysis can be obtained by Snyder et al.29 and U.S. EPA Method 6850. The HPLC settings are briefly discussed as follows. Analytical column: M.IX.MSD1 for LC/MS by MetrohmPeak; autosampler injection volume: 10.0 μL; HPLC pump— flow rate: 0.7 mL/min, mobile phase: 30% of 50.0 mM ammonium formate (NH4COOH), and 25 mM ammonium carbonate ((NH4)2CO3) mixture +70% of acetonitrile (CH3CN). These parameter combinations resulted in elution of the perchorate in approximately 16 min, with a total run time of 18 min. Mass spectrometer parameters are, ionization mode: Electrospray (API-ES); polarity: Negative; Spray chamber—drying gas flow: 12.0 L/min, nebulizer pressure: 35 psig, drying gas temperature: 250 °C; SIM parameters—SIM ion: 99.0, fragmentor: 70 V, gain: 1.0 EMV, dwell time: 290 ms, % relative dwell: 100.0; capillary voltage; 3500 Vcap. The response variable of interest in this study is the concentration of ClO4 in the fresh weight 9364
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology
ARTICLE
plant tissue samples (μg/kg FW). The method detection limit (MDL) of HPLC/MS for ClO4 was determined to be 0.5 μg/L in plant extract, which was equivalent to 1.6 μg of ClO4/kg FW of plant tissue. As shown in Table 2, some of spinach root samples were below the 1.6 μg/kg of detection limit (left-censored in the statistical analyses). All ClO4 concentrations in spinach leaves were above the detection limit. 3.2. Nitrate and Chloride Analysis of Plant Samples and Irrigation Water. The filtered samples after centrifugation (approximately 30.0 mL) were utilized for NO3 and Cl analysis of plant extracts. Nitrate in plant slurry was measured by UV spectrometry method30 and Cl was determined by coulometric-amperometric titration method.31 4. Statistical Methodology. The following statistical analysis was conducted to examine the factors controlling ClO4 uptake in spinach and to develop equations relating ClO4, NO3, and Cl concentrations in irrigation water to ClO4 plant tissue concentration. The following linear factorial model (with 2-way interaction) was fit to both the natural log transformed leaf and root tissue data: lnðClO 4 : accumÞ ¼ β0 þ β1 lnðClO4 Þ þ β2 lnðNO3 Þ þ β3 lnðCl Þ þ β12 lnðClO 4 Þ lnðNO3 Þ þ β13 lnðClO4 Þ lnðCl Þ þ β23 lnðNO 3 Þ lnðCl Þ þ ε
ð1Þ
In eq 1, the ε error term represents an independently, identically and normally distributed error component and the various β parameters quantify the primary (first order) and two-way interaction terms.32 Positive parameter estimates in this model imply that the log ClO4 concentrations in the plant tissue increase as the log transformed ClO4, NO3, and/or Cl water concentrations Table 2. Number of Left-Censored Spinach Root Tissue for Perchlorate Measurements (I.E., Measurements Below the 1.6μg/kg FW Method Detection Limit of ClO4) treatmenta
low ClO4 , low NO3 , low Cl
spinach roots
0
high ClO4, low NO3, low Cl
0
low ClO4, low NO3, high Cl
1
high ClO4, low NO3, high Cl
0
low ClO4, high NO3, low Cl high ClO4, high NO3, low Cl
1 0
low ClO4, high NO3, high Cl
3
high ClO4, high NO3, high Cl
0
mid ClO4, mid NO3, mid Cl
0
high ClO4, mid NO3, mid Cl
0
Low, mid, and high ClO4 [ppb] represent 4, 22, and 40, respectively; Low, mid, and high NO3 [mmolc/L] represent 2, 11, and 20, respectively.; Low, mid, and high Cl [mmolc/L] represent 2.5, 13.75, and 25, respectively. a
increase, while negative estimates imply that the log ClO4 concentrations decrease as these water concentration levels increase. For the spinach leaves data (where all samples were above the detection limit and thus no censoring occurred), eq 1 was estimated using standard linear modeling techniques.32 For the left-censored spinach root data, eq 1 was estimated using maximum likelihood techniques.33 All model estimation was performed using the GLM and LIFETEST procedures in SAS.34 Based on the p-values associated with the estimated parameters (Table 3), reduced forms of eq 1 were also fit to each plant tissue data set. These reduced models were estimated by removing all nonsignificant parameter estimates from the linear factorial model (at the 0.05 significance level). Goodness-of-fit (GOF) tests were calculated to assess the adequacy of each fitted equation. For the complete (i.e., noncensored) leaves data sets, traditional lack-of-fit (LOF) F-tests were computed.32,35 For the left-censored root data sets, asymptotic GOF tests were computed by calculating the log-likelihood (LL) score differences between the reduced and saturated models and then comparing these 2 LL scores to Chi-square distributions with the appropriate degrees of freedom. The primary goal in each analysis was to identify a parsimonious linear factorial model that fully described how the changing irrigation water ClO4, NO3, and Cl concentrations influenced the plant tissue ClO4 concentrations.
’ RESULTS AND DISCUSSION The sand tank environment was hypothesized to potentially cause ClO4 degradation by bacteria in the root zone.14 Because of this consideration, various researchers utilized an aerated hydroponic system for the laboratory scale plant uptake experiment to minimize the rhizosphere degradation effect on ClO4 uptake (refs 14,27,36, etc.). However, commercial leafy vegetables have been grown primarily in soil under field environments. Our experiment was designed to evaluate the combined effect of NO3 and Cl on ClO4 uptake in spinach in a controlled sand tank environment which both more closely reflects field conditions and yet enables accurate monitoring of root zone ClO4 concentrations. The concentration of ClO4 in reservoirs was monitored and maintained at the constant concentrations (4, 22, and 40 μg/L) throughout the experiment as indicated in the Materials and Methods section. We found no evidence of decrease in ClO4 related to a soil process, suggesting that, as expected, our rhizosphere was highly aerobic and ClO4 degradation did not need to be further considered in our experiment. ET losses were approximately equal within the range of 1.9 and 2.8 cm of water for all treatments. Low Cl treatments had ET loss of 2.23 (average of four treatments) and high Cl reservoirs showed 2.55 cm of water, while water losses by ET in low and high NO3 treatment reservoirs were 2.38 and 2.4 cm of water, respectively. The interactive effects of the three independent variables on ClO4 in plant parts can best be evaluated by a multivariate
Table 3. Summary Statistics of RMSE and Parameter P-Values: Full Factorial Models type III p-values for individual parameter estimates data set
RMSE
β1
β2
β3
β12
β13
β23
spinach: leaves
0.373
<0.001
0.002
0.154
0.899
0.327
0.402
spinach: roots
0.710
0.235
0.027
0.087
0.170
0.046
0.377
9365
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology
ARTICLE
Figure 1. Perchlorate content in spinach leaves as related to irrigation water nitrate and chloride at two perchlorate concentrations. Parts a and b represent the perchlorate content when the chloride concentration is constant at 2.5 and 25 mmolc /L, respectively. Parts c and d show the perchlorate content when nitrate concentration is constant at 2 and 20 mmolc /L, respectively. The error bars indicate one standard deviation of the mean where n = 3.
statistical analysis. However, for understanding the impact of the various variables it appears useful to examine a subset of the data consisting of end member concentrations of ClO4 while holding the other two variables constant, for both high and low concentrations of NO3 and Cl. 1. Perchlorate Uptake As Related to NO3 and Cl. Perchlorate content in spinach leaves was dramatically greater than literature reports for iceberg and butterhead lettuce leaves,27 and appeared to be the highest vegetable ClO4 accumulator. However, ClO4 in spinach leaves was lower than that for the forage crop alfalfa, grown in sand.23 As shown in Figure 1, the mean concentration of 3.20 mg/kg of ClO4 was obtained when NO3 and Cl was low. As expected, ClO4 in the leaves increased with increased ClO4 in solution. Increased NO3 in irrigation water suppressed ClO4 uptake under both high and low Cl (Figure 1a and Figure 1b, respectively). Increased Cl had no effect on ClO4 uptake under low NO3, as shown in Figure 1c, and only a small reduction under high NO3 (Figure 1d) was observed. Spinach appeared to have a different ClO4 uptake mechanism compared to iceberg and butterhead lettuce. The highest ClO4 concentrations in spinach leaves were obtained when the NO3 level was low (2 mmolc/L). Also, Cl had a smaller effect on ClO4 uptake in spinach as compared to lettuce (Ha and Suarez, in preparation). As shown in Figure 2, the data for spinach roots indicated that ClO4 accumulation was
greatly suppressed by elevated NO3 but only slightly affected by elevated Cl in irrigation water. 2. Statistical Analysis. Table 4 showed the pertinent model summary statistics and reduced factorial model parameter estimates for two spinach data sets. The only statistically significant parameter estimates in the reduced factorial model associated with the leaf data were the main ln(ClO4) and ln(NO3) effects. The positive ln(ClO4) parameter estimate implied that increased ClO4 concentrations in irrigation water translating into higher ClO4 accumulation levels in the leaf tissue. Likewise, the negative ln(NO3) parameter estimate implied that as the NO3 level increased, the ClO4 accumulation level in the plant tissue decreased. Interestingly, the ln(Cl) main effect was not statistically significant in this model, suggesting that the ln(Cl) anion levels did not influence ClO4 accumulation in spinach leaves. A somewhat more complex factorial model needed to be employed to adequately describe the ln(ClO4) accumulation pattern in the spinach root tissue samples. With respect to main effects, the log ClO4 accumulation level increased as the ln(ClO4) anion concentration level increased, and the accumulation level decreased as both the ln(NO3) and ln(Cl) anion concentration levels increased. Additionally, the positive ln(ClO4) x ln(Cl) parameter estimate implied that the ln(ClO4) accumulation rate attributed specifically to the 9366
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology
ARTICLE
Figure 2. Perchlorate content in spinach roots as related to irrigation water nitrate and chloride at two perchlorate concentrations. Parts a and b represent the perchlorate content when the chloride concentration is constant at 2.5 and 25 mmolc /L, respectively. Parts c and d show perchlorate content when nitrate concentration is constant at 2 and 20 mmolc /L, respectively. The error bars indicate one standard deviation of the mean where n = 3.
Table 4. Reduced Factorial Model Summary Statistics and Parameter Estimates (With Associated Standard Errors): Spinach Data Setsa model statistics
leaves
roots
RMSE
0.373
0.725
r (correlation)
0.972
0.814
GOF p-value
0.754
0.276
intercept (std.error) 1(std.error)
5.107 (0.21) 1.034 (0.06)
2.291 (0.79) 0.730 (0.26)
2(std.error)
0.747 (0.06)
0.596 (0.13)
3 (std.error)
ne
0.746 (0.34)
12(std.error)
ne
ne
13(std.error)
ne
0.212 (0.11)
23(std.error)
ne
ne
parameter estimates
a
Note: ne = not estimated.
ln(ClO4) anion concentration level increases as the ln(Cl) concentration level rose. In the spinach tissue samples, the root-mean-square error (RMSE) estimate (unit: μg/L) for the root model is about two times bigger than the leaf RMSE estimate (0.725 versus 0.373). This result suggests that the relative variation in ln(ClO4)
accumulation in the roots is greater than the relative variation in the leaves. Additionally, both spinach models also exhibit nonsignificant GOF test statistics. These results indicate that these fitted factorial models adequately describe the leaf and root tissue sample data collected from the spinach crop. 3. Ion Uptake and Translocation. Competition between ions, for the example of NO3 and Cl during plant uptake process, has been known to be significant for crop production.37 Competition between NO3 and Cl on ClO4 uptake in higher plants has not still been extensively investigated. In order to compare ion uptake of different anions and evaluate ion specific mechanisms, ratios of concentrations in the plant to the concentrations in irrigation water [bioconcentration factor (BCF)] were calculated. In this instance we calculated ratios of ClO4 and Cl in the plant leaves and roots to the ion concentrations in irrigation water. The relative uptake of spinach is shown in Figures 3 and 4, which has ratios that are expressed in μg/kg FW divided by μg/L (ppb) for ClO4 and g/kg FW divided by g/L for Cl. Student’s paired t test with two-tailed distribution was conducted with ClO4/ClO4 and Cl/Cl ratio data. Although it was known that plant roots did not appear to take up ions selectively without having specific transporters for specific ions,27,37 our results clearly indicated that there was a different uptake pattern of ClO4 and Cl in roots among the different anion concentrations in irrigation water. Seyfferth et al.27 also cited Marschner37 for the statement that Cl and 9367
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology
ARTICLE
Figure 3. Perchlorate and chloride concentration in spinach leaves divided by concentration in solution. Low, mid, and high indicate ClO4 concentrations of 40, 220, and 400 nmolc/L, respectively. Error bars represent one standard deviation of the mean where n = 3.
NO3 competed each other for Cl uptake in Barley plants. They commented that the ClO4 uptake mechanism was affected by NO3 in plant tissue as a result of sharing a common anion transport mechanism in higher plants. Perchlorate uptake in iceberg and butterhead lettuce was affected by NO3, as reported earlier27 from their hydrophonic growth chamber system. Based on ref 27 the BCF values for crisp head were calculated approximately between 11.0 and 20.0 with the NO3 ranges between 4 and 12 mM, and between 11.6 and 14.0 with the Cl ranges between 5 and 15 mM. In contrast, the BCF values in this study ranged from 16.6 shown in Figure 3c at high NO3 concentrations to 102.1 shown in Figure 3b at low NO3 concentrations. The reported BCF for alfalfa forage crop was approximately 360.23 The ratio of leaf/solution concentration of ClO4 and Cl is presented in Figure 3. These data all showed that ClO4 was preferentially accumulated relative to Cl under all conditions. Under low Cl and NO3 concentrations, the accumulation of
ClO4 relative to solution concentration was about 7 times higher than that for Cl (Figure 3a). These data are in contrast to lettuce data, in which the Cl concentration ratio is roughly comparable but the ClO4 ratio is much lower. Based on the ratio data it is clear that spinach leaves accumulate more ClO4 compared to that in lettuce leaves (data not shown). Increasing solution Cl concentration suppressed the spinach leaf/solution Cl ratio and resulted in a slight increase in ClO4 leaf/solution ratio when ClO4 was low and almost no change at higher ClO4 (Figure 3b). These data are in contrast to lettuce data where increased Cl suppressed the ClO4 ratio (data not shown). In contrast to Cl, NO3 drastically suppressed ClO4 ratios (Figure 3c). These data demonstrate that spinach has a different uptake or translocation mechanism for ClO4 and increased Cl does not affect ClO4 accumulation in the leaves. Interestingly, Cl uptake in spinach leaves was not high when compared to Cl uptake in lettuce leaves (data not shown). 9368
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology
ARTICLE
Figure 4. Perchlorate and chloride concentration in spinach roots divided by concentration in solution. Low, mid, and high indicate ClO4 concentrations of 40, 220, and 400 nmolc/L, respectively. Error bars represent one standard deviation of the mean where n = 3.
In contrast to the leaf data, spinach roots showed a high root/ solution Cl ratio and a very low root/solution ClO4 ratio (Figure 4a) under low NO3 and Cl in solution. As Cl increased in solution, the Cl root/solution ratio decreased as observed in leaves, and the ClO4 ratio is increased in low ClO4 but similar in magnitude for both high and low ClO4 levels (compare Figure 4a and Figure 4b). Increased solution NO3 suppressed Cl and ClO4 uptake in spinach root (compare Figure 4c with Figure 4a). Based on these data it appears that transport of ClO4 and Cl in the spinach plant is vastly different, with ClO4 transport being similar in the leaves and Cl transport being restricted to the leaves as Cl increases in irrigation water. Chloride uptake by spinach (Spinacia oleracea L.) leaves and roots was earlier investigated by Speer and Kaiser.38 In their study, spinach was treated in 100 mmol/L NaCl solution in a growth chamber for 10 days. Another set of experiment required a few stepwise increments of 100 mmol/L NaCl solution to reach at the final concentration of 300 mmol/L NaCl in solution to
evaluate Na+ and Cl distribution between symplastic and apoplastic space of leaves for 17 days. The results of the first experiment of Speer and Kaiser38 showed that the Cl concentration in spinach leaves reached a relatively low pseudosteady state after the fourth day of the experiment. Slightly less NaCl was accumulated in spinach roots compared to spinach leaves in four days and spinach roots ended up accumulating more NaCl in 10 days. However, the total Cl concentration of leaves and roots resulted in means with error bars indicating that the standard deviations were very close to each other. Also, it was noted by Speer and Kaiser38 that spinach accrued more Cl in symplast space of leaves than in apoplasm. Our experimental results also revealed that there was a small Cl uptake ratio reduction in spinach roots when Cl concentration is high (25 mmolc/L) when NO3 concentration increased (compare Figure 4b with Figure 4d). This phenomenon was also examined by Glass and Siddiqi39 with their experiment using barley plants. Marschner37 mentioned Glass and Siddiqi39’s work where the inhibition of Cl uptake with an increase in 9369
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology NO3 in solution appeared to result from a negative feedback from NO3 stored in the vacuoles of root cells and Cl influx at the plasma membrane. The Cl concentration in barley shoots was greatly reduced in the presence of NO3 in solution. Although there were no experimental results of barley roots reported by Glass and Siddiqi,39 their barley shoot results looked similar to our Cl uptake experimental results of spinach leaves. The following findings were obtained: (1) The ln(ClO4) leaf accumulation shows a small increase as the ln(ClO4) anion concentration in irrigation water increases and ln(ClO4) leaf accumulation decreases as the ln(NO3) and ln(Cl) anion concentrations increase. (2) There are few statistically significant anion interactions in the leaf ln(ClO4) accumulation models. In contrast, the root ln(ClO4) accumulation model is more complex, exhibiting more statistically significant anion interaction parameter estimates, although the main effect trends are consistent across both the root and leaf tissue samples. (3) Transport of ClO4 and Cl in the spinach plant is largely different between leaves and roots; ClO4 translocation is constant in the leaves and Cl transport is being restricted to the leaves when Cl increases in irrigation water. The mass balance of ClO4 was not estimated due to the missing information of total weight of spinach harvest, thus this would be a limitation of this study.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 806-356-5717; e-mail: [email protected]. Present Addresses §
Conservation and Production Research Laboratory, USDAARS, 2300 Experiment Station Rd., Bushland, TX 79012
’ ACKNOWLEDGMENT This study was funded by U.S. Department of Agriculture under the ARS postdoctoral program. We are especially grateful to Dr. Wei Zheng for assistance in the HPLC/MS analysis of ClO4 and Ms. Stephanie Stasiuk for her assistance in both preparation of plant samples and analysis of ClO4 during the experiment. The anonymous reviewers’ comments were valuable and appreciated. Mention of company names or products is for the benefit of the reader and does not imply endorsements, guarantee, or preferential treatment by the USDA or its agents. USDA is an equal opportunity provider and employer. ’ REFERENCES (1) Bradford., C. M.; Park, J.-W.; Rinchard, J.; Anderson, T. A.; Liu, F.; Theodorakis, C. W. Uptake and elimination of perchlorate in eastern mosquitofish. Chemosphere. 2006, 63, 1591–1597. (2) Herman, D. C.; Frankenberger, W. T. Microbial-mediated reduction of perchlorate in groundwater. J. Environ. Qual. 1998, 4, 750–754. (3) Ellington, J. J.; Evans, J. J. Determination of perchlorate at partsper-billion levels in plants by ion chromatography. J. Chromatogr., A 2000, 2, 193–199. (4) Urbansky, E. T.; Brown, S. K.; Magnuson, M. L.; Kelty, C. A. Perchlorate levels in samples of sodium nitrate fertilizer derived from Chilean caliche. Environ. Pollut. 2001, 112, 299–302. (5) Kirk, A. B.; Martinelango, P. K.; Tian, K.; Dutta, A.; Smith, E. E.; Dasgupta, P. K. Perchlorate and iodide in dairy and breast milk. Environ. Sci. Technol. 2005, 7, 2011–2017.
ARTICLE
(6) El Aribi, H.; Le Blanc, Y. J. C.; Antonsen, S.; Sakuma, T. Analysis of perchlorate in foods and beverages by ion chromatography coupled with tandem mass spectrometry (IC-ESI-MS/MS). Anal. Chim. Acta 2006, 1, 39–47. (7) U.S. EPA, CCL and Regulatory Determinations Home, 2011. http:// water.epa.gov/scitech/drinkingwater/dws/ccl/index.cfm (accessed May 20, 2011). (8) U.S. EPA. Perchlorate Treatment Technology Update, EPA 542-R05-015; U.S. Environmental Protection Agency: Washington, DC, 2005. (9) Zhang, P.; Avudzega, D. M.; Bowman, R. S. Removal of perchlorate from contaminated waters using surfactant-modified zeolite. J. Environ. Qual. 2007, 36, 1069–1075. (10) Urbansky, E. T. Perchlorate as an environmental contaminant. Environ. Sci. Pollut. Res. 2002, 9, 187–192. (11) Nzengung, V. A.; Wang, C.; Harvey, G. Plant-mediated transformation of perchlorate into chloride. Environ. Sci. Technol. 1999, 33, 1470–1478. (12) Saito, K.; Yamamoto, K.; Takai, T.; Yoshida, S. Inhibition of iodide accumulation by perchlorate and thiocyanate in a model of the thyroid iodide transport system. Acta Endrocrinol. 1983, 104, 456–461. (13) Urbansky, E. T.; Magnuson, M. L.; Kelty, C. A.; Brown, S. K. Perchlorate uptake by salt cedar (Tamarix ramosissima) in the Las Vegas Wash riparian ecosystem. Sci. Total Environ. 2000, 256, 227–232. (14) Tan, K.; Anderson, T. A.; Jackson, W. A. Uptake and exudation behavior of perchlorate in smartweed. Int. J. Phytorem. 2006, 8, 13–24. (15) Susarla, S.; Collette, T. W.; Garrison, A. W.; Wolfe, N. L.; McCutcheon, S. C. Perchlorate identification in fertilizers. Environ. Sci. Technol. 1999, 33, 3469–3472. (16) Sanchez, C. A.; Blount, B. C.; Valentin-Blasini, L.; Lesch, S. M.; Krieger, R. I. Perchlorate in the feed-dairy continuum of the southwestern United States. J. Agric. Food Chem. 2008, 56, 5443–5450. (17) Parker, D. R. Perchlorate in the environment: the emerging emphasis on natural occurrence. Environ. Chem. 2009, 6, 10–27. (18) Urbansky, E. T. Quantitation of perchlorate ion: Practices and advances applied to the analysis of common matrices. Crit. Rev. Anal. Chem. 2000, 30, 311–343. (19) Hutchinson, S. L. A Study on the Accumulation of Perchlorate in Young Head Lettuce, EPA report/ 600/R-03/003; U.S. Environmental Protection Agency: Research Triangle Park, NC, 2004. (20) Sanchez, C. A.; Krieger, R. I.; Khandaker, N.; Moore, R. C.; Holts, K. C.; Neidel, L. L. Accumulation and perchlorate exposure potential of lettuce produced in the lower Colorado River region. J. Agric. Food Chem. 2005, 53, 5479–5486. (21) Sanchez, C. A.; Krieger, R. I.; Khandaker, N. R.; ValentinBlasini, L.; Blount, B. C. Potential perchlorate exposure from Citrus sp. Irrigated with contaminated water. Anal. Chim. Acta 2006, 567, 33–38. (22) Holdren, G. C.; Kelly, K.; Weghorst, P. Evaluation of potential impacts of perchlorate in the Colorado River on the Salton sea, California. Hydrobiologia. 2008, 604, 173–179. (23) Jackson, W. A.; Joseph, P.; Laxman, P.; Tan, K.; Smith, P. N.; Yu, L.; Anderson, T. A. Perchlorate accumulation in forage and edible vegetation. J. Agric. Food Chem. 2005, 53, 369–373. (24) NDEP, Nevada Division of Environmental Protection, 2009. http://ndep.nv.gov/ (accessed August 1, 2009). (25) Bernal, C. T.; Bingham, F. T. Salt tolerance of Mexican wheat: I. Effect of NO3 and NaCl on mineral nutrition, growth, and grain production of four wheats. Soil Sci. Soc. Am. Proc. 1973, 37, 711–715. (26) Rubinigg, M.; Posthumus, F. S.; Elzenga, J. T. M.; Stulen, I. Effect of NaCl salinity on nitrate uptake in Plantago maritima L. Phyton (Austria). 2005, 45, 295–302. (27) Seyfferth, A. L.; Henderson, M. K.; Parker, D. R. Effects of common soil anions and pH on the uptake and accumulation of perchlorate in lettuce. Plant Soil. 2008, 302, 139–148. (28) Seyfferth, A. L.; Parker, D. R. Determination of low levels of perchlorate in lettuce and spinach using ion chromatography-electrospray 9370
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
Environmental Science & Technology
ARTICLE
ionization mass spectrometry (IC-ESI-MS). J. Agric. Food Chem. 2006, 54, 2012–2017. (29) Snyder, S. A.; Vanderford, B. J.; Rexing, D. J. Trace analysis of bromate, chlorate, iodate, and perchlorate in natural and bottled waters. Environ. Sci. Technol. 2005, 39, 4586–4593. (30) Cawse, P. A. The determination of nitrate in soil solutions by ultraviolet spectrophotometry. Analyst. 1967, 92, 311–315. (31) Rhoades, J. D., Soluble salts. In Methods of Soil Analysis, Part 2. Chemical and Microbiological Properties, 2nd ed.; In Page, A. L., Ed.; American Society of Agronomy: Madison, WI, 1982; pp 161179. (32) Montgomery, D. C. Design and Analysis of Experiments, 5th, ed.; John Wiley: New York, NY. 2002. (33) Lawless, J. E. Statistical Models and Methods for Lifetime Data; John Wiley: New York, NY. 1982. (34) SAS Institute Inc. SAS/STAT User’s Guide, Ver. 8; SAS Institute Inc.: Cary, NC, 1999. (35) Myers, R. H. Classical and Modern Regression with Applications; Duxbury Press: Boston, MA., 1986. (36) Seyfferth, A. L.; Parker, D. R. Effects of genotype and transpiration rate on the uptake and accumulation of perchlorate (ClO4) in lettuce. Environ. Sci. Technol. 2007, 41, 3361–3367. (37) Marschner, H. Mineral Nutrition of Higher Plants, 2nd, ed.; Academic press: San Diego, CA., 1995. (38) Speer, M.; Kaiser, W. M. Ion relations of symplastic and apoplastic space in leaves from Spinacia oleracea L. and Pisum sativum L. under salinity. Plant Physiol. 1991, 97, 990–997. (39) Glass, A. D. M.; Siddiqi, M. Y. Nitrate inhibition of chloride influx in barley: Implications for a proposed chloride homeostat. J. Exp. Bot. 1985, 36, 556–566.
9371
dx.doi.org/10.1021/es2010094 |Environ. Sci. Technol. 2011, 45, 9363–9371
ARTICLE pubs.acs.org/est
Widespread Occurrence of Bisphenol A in Paper and Paper Products: Implications for Human Exposure Chunyang Liao and Kurunthachalam Kannan* Wadsworth Center, New York State Department of Health, and Department of Environmental Health Sciences, School of Public Health, State University of New York at Albany, Empire State Plaza, P.O. Box 509, Albany, New York 12201-0509, United States
bS Supporting Information ABSTRACT: Bisphenol A (BPA) is used in a variety of consumer products, including some paper products, particularly thermal receipt papers, for which it is used as a color developer. Nevertheless, little is known about the magnitude of BPA contamination or human exposure to BPA as a result of contact with paper and paper products. In this study, concentrations of BPA were determined in 15 types of paper products (n = 202), including thermal receipts, flyers, magazines, tickets, mailing envelopes, newspapers, food contact papers, food cartons, airplane boarding passes, luggage tags, printing papers, business cards, napkins, paper towels, and toilet paper, collected from several cities in the USA. Thermal receipt papers also were collected from Japan, Korea, and Vietnam. BPA was found in 94% of thermal receipt papers (n = 103) at concentrations ranging from below the limit of quantitation (LOQ, 1 ng/g) to 13.9 mg/g (geometric mean: 0.211 mg/g). The majority (81%) of other paper products (n = 99) contained BPA at concentrations ranging from below the LOQ to 14.4 μg/g (geometric mean: 0.016 μg/g). Whereas thermal receipt papers contained the highest concentrations of BPA (milligram-per-gram), some paper products, including napkins and toilet paper, made from recycled papers contained microgram-per-gram concentrations of BPA. Contamination during the paper recycling process is a source of BPA in paper products. Daily intake (DI) of BPA through dermal absorption was estimated based on the measured BPA concentrations and handling frequency of paper products. The daily intake of BPA (calculated from median concentrations) through dermal absorption from handling of papers was 17.5 and 1300 ng/day for the general population and occupationally exposed individuals, respectively; these values are minor compared with exposure through diet. Among paper products, thermal receipt papers contributed to the majority (>98%) of the exposures.
’ INTRODUCTION Bisphenol A (BPA) is an endocrine-disrupting chemical and has been implicated in a wide variety of adverse health outcomes in humans.16 Due to its toxicity and widespread human exposure, BPA has received the attention of regulatory agencies across the globe.1 An oral reference dose (RfD) for BPA of 50 μg/kg body weight (bw)/day has been established by the United States Environmental Protection Agency (USEPA) and the European Food Safety Authority (EFSA).7,8 Nevertheless, some studies have reported that BPA can stimulate cellular responses and toxic effects at exposure doses below the currently recommended RfD.13 The low-dose exposures and toxicity of BPA are subjects of debate.6,9 Produced in quantities of over 8 billion pounds each year worldwide, BPA is one of the most widely used chemicals, as the base chemical in the manufacture of polycarbonate plastics and the resin lining of food and beverage cans.10,11 BPA can leach out of products, through the hydrolysis of ester bonds linking BPA monomers, under acidic or basic conditions.12 BPA has been reported to occur in various environmental matrices, including air, water, sewage sludge, soil, dust, foodstuffs, and soft drinks.2,3,1315 r 2011 American Chemical Society
Biomonitoring studies have reported widespread occurrence of BPA at ng/g or ng/mL levels in human tissues and fluids, including urine, blood (maternal blood, cord blood, and fetal blood), and other bodily fluids (amniotic fluid, follicle fluid, saliva, and breast milk).13,1621 There are multiple sources that contribute to human exposure to BPA.3,9 Diet, however, appears to be a major source of human exposure to BPA.18,2224 The U.S. National Toxicology Program reported that oral exposure from foods of adults in the USA to BPA ranged from 0.008 to 1.5 μg/kg bw/day.22 On the basis of the leaching levels from consumer products and consumption of canned foods, BPA exposures have been estimated to range from <1 to 5 μg/kg bw/day.3,21 However, recent studies indicate that human exposure doses to BPA have been underestimated because the contribution of nondietary sources to BPA exposure is not well understood. Nonfood sources that are presumed to be Received: July 20, 2011 Accepted: September 23, 2011 Revised: September 12, 2011 Published: September 23, 2011 9372
dx.doi.org/10.1021/es202507f | Environ. Sci. Technol. 2011, 45, 9372–9379
Environmental Science & Technology important sources of exposure include inhalation and dermal absorption, especially in occupational settings.13 BPA is present in some cigarette filters 25 and there exits an association between smoking and urinary BPA concentrations.26 A joint expert meeting held in 2010 by the Food and Agriculture Organization and the World Health Organization estimated inhalation exposure to BPA at 0.003 μg/kg bw/day and soil/dust ingestion at 0.0001 to 0.03 μg/kg bw/day for the general population.27 Additional sources of human exposure to BPA include thermal papers and paper currencies,28,29 medical devices,30 dental sealants,31 and printing ink.32 The thermal papers used in direct thermal printing process typically consist of two layers: the base paper (a standard paper formulation) and the thermal sensitive layer.28 The thermal sensitive layer typically consists of three components: the thermochromic dye, a weakly acidic color developer (traditionally BPA), and a solvent (generally a longchain aliphatic compound such as fatty acid, amide, alcohol). When the thermal sensitive layer is heated above the melting point of the solvent by a printing stylus, BPA (the developer) interacts with the thermochromic dye, donating protons which open the rings of the dye and increase the conjugation of the system, resulting in a color.28 Large amounts of BPA have been used in the production of thermal receipt papers due to its efficacy, availability, and low cost.28 Very high levels of BPA, of up to 3 to 22 mg/g, have been reported in thermal receipt papers.28,29 Thermal receipt papers are produced in large quantities for use in cash register receipts, luggage tags, and tickets (bus/train and lottery).28 It has been reported that approximately 30% of thermal papers enter the paper recycling streams.33 Recycling of thermal paper can introduce BPA into the cycle of paper production.34 BPA can be absorbed into human skin during the handling of thermal receipt papers.29 Because most people contact with thermal receipt papers and paper products on a daily basis, exposure of humans to BPA through dermal contact is expected.35 Two approaches have been used to derive estimates of daily intake of BPA. First, by summation or aggregation of the amount of BPA measured in various sources of exposure (e.g., food, food packaging, air, water, dust, paper products) for the estimation of total exposure (“aggregate sources” method).36 Second, on the basis of biomonitoring data, such as the urinary BPA concentrations to estimate the total exposure that reflects all sources of exposure, both known and unknown (“back calculation” method).17 Since BPA is biotransformed in the body, some uncertainties exist in the “back calculation” method of BPA exposure assessment.17 Nevertheless, both of these approaches are dependent on various assumptions, such as body weight, skin surface area, amount of food or beverage consumed, daily volume of urine output, or ability of a single measurement to characterize general exposures. Although a few studies have measured BPA in thermal receipt papers,28,29 those studies involved small sample sizes (n < 25). Furthermore, exposure to BPA from the handling of papers was not estimated previously. In addition, BPA levels in other types of paper and paper products, such as flyers, magazines, tickets, mailing envelopes, newspapers, food contact papers (e.g., paper cups, plates), food cartons, napkins, kitchen rolls (i.e., paper towels), and toilet paper, that are frequently used in our daily lives, and the potential contribution of these paper types to BPA exposure remain unknown. In this study, we measured BPA concentrations in 202 samples, representing several types of papers and paper products, with the aim of establishing baseline
ARTICLE
concentrations in paper products and estimating potential exposure doses to BPA from the handling of papers.
’ MATERIALS AND METHODS Sampling. Thermal receipt paper samples (n = 103) were collected from 58 locations, including supermarkets, grocery stores, banks, public libraries, gas stations, restaurants, and fast food restaurants in Albany, New York City, and Buffalo (New York), Boston (Massachusetts), Chicago (Illinois), Weston (Vermont), and Charlotte (North Carolina), in the USA in 2010 and 2011, and from retail stores in Matsuyama and Tokyo, Japan, Incheon, Korea, and Hanoi, Vietnam in 2010 and 2011. Other paper and paper products (n = 99) were grouped into 14 categories: flyers (e.g., advertisement brochures, store coupons, gift cards, bus schedule), magazines, tickets (e.g., train and bus tickets), mailing envelopes, newspapers, food contact papers (e.g., fast-food wrappers, paper cups, paper plates), food cartons (e.g., pizza paperboards, food buckets, snack food boxes), airplane boarding passes, airplane luggage tags, printing paper (i.e., regular copy paper), business cards, facial tissue (referred to as napkins in this study), paper towels (or kitchen rolls), and toilet paper. Most of the paper products were made from recycled paper.32,34,37,38 These samples were collected mainly in Albany, New York, USA, and a few samples were collected in New York City and Buffalo (New York) and Boston (Massachusetts) in 2010 and 2011. Samples were individually wrapped in polyethylene bags and stored in a freezer at 20 °C until analysis. Analysis. Paper and paper products were analyzed for BPA by following a method similar to that described earlier, with some modifications.15 For thermal receipt papers, a circular spot (19 mm diameter) was taken in the middle of each receipt using a punch (Uchida Corp., Torrance, CA). After weighing the spot accurately (∼ 0.0172 g), samples were cut into small pieces and extracted with methanol three times. Further details of the analysis are given in the Supporting Information. An Applied Biosystems API 2000 electrospray triple quadrupole mass spectrometer (ESI-MS/MS; Applied Biosystems, Foster City, CA) coupled with an Agilent 1100 Series HPLC system (Agilent Technologies Inc., Santa Clara, CA), consisting of a binary pump and an autosampler, was used for the measurement of BPA. An analytical column (Betasil C18, 100 2.1 mm column; Thermo Electron Corporation, Waltham, MA), connected in series with a Javelin guard column (Betasil C18, 20 2.1 mm; Thermo Electron Corporation), was used for analysis. Data were acquired using multiple reaction monitoring for the transitions of 227 > 212 for BPA (97%; Sigma-Aldrich, St. Louis, MO) and 239 > 224 for 13C12BPA (99%; Cambridge Isotope Laboratories, Andover, MA). Quality Assurance and Quality Control (QA/QC). For each batch of 20 samples analyzed, a procedural blank, a spiked blank, a pair of matrix spiked samples, and duplicate samples were processed. The procedural blank, containing water in place of paper sample, was analyzed by passage through the entire analytical procedure as a check for interferences or laboratory contamination. BPA was not detected in procedural blanks or sample containers (i.e., polypropylene tube). The recoveries of BPA from spiked blanks and spiked matrices were 100 ( 15% and 101 ( 17% (mean ( SD), respectively. The relative standard deviation (RSD) of replicate analysis of samples was <10%. Eight thermal receipt samples and 10 samples from other paper types were randomly selected, and a fourth extraction was carried out 9373
dx.doi.org/10.1021/es202507f |Environ. Sci. Technol. 2011, 45, 9372–9379
Environmental Science & Technology
ARTICLE
with 3 mL of methanol (after the first three extractions) to confirm that extraction of BPA from paper samples was complete. BPA was detected in the fourth extraction, at concentrations ranging from 12.5 to 23.3 ng/mL, in thermal receipt samples. In comparison with the extremely high concentrations of BPA detected in thermal receipt samples (on the order of milligram-per-gram), the residual BPA found in the fourth extraction was only 0.03 to 0.07% of the total concentration in the thermal receipt samples. For other paper types, BPA was not detected in the fourth extraction. Instrumental drift in sensitivity was checked by duplicate injection of samples and by analyzing a continuing calibration check standard after every ten samples. The calibration of BPA standard injected at concentrations ranging from 0.05 to 100 ng/mL showed good linearity (r > 0.99, n = 10). The limit of quantitation (LOQ) was 1 ng/g. Quantification was made using the isotope-dilution method. Prior to the analysis of samples, recovery and reproducibility of the method were verified by spiking known concentrations of BPA into selected paper samples. Because thermal receipt papers contained high concentrations of BPA, these samples were stored and analyzed in separate batches from other paper types. Estimation of Daily Intake. Skin uptake/dermal absorption from handling of paper products (especially thermal paper receipts) can be a pathway of human exposure to BPA. In thermal papers, BPA exists as a free monomer that is mobile and transferable to objects with which it comes in contact.29,35 The extent of human exposure to BPA through handling of paper products remains unknown.29,35,39,40 Biedermann et al. 29 tested the transfer of BPA from thermal receipt paper to human skin and reported that holding a thermal paper with 15.2 g BPA/kg (mean value) for 5 s resulted in the transfer of 1636 ng BPA to the surface of the hand. On the basis of this information, we calculated a paper-to-skin transfer rate of BPA as k = 1636 ng/(15.2 g/kg 3 5s) = 21 522 ng/s. It has been reported that 27% of BPA found on skin surface penetrates and reaches the bloodstream within 2 h.29 We estimated human exposure to BPA based on the measured concentrations and frequency of handling of paper products. We assumed that the general population handles thermal receipts twice a day. The frequency of use and handling time are probably different for various paper types. Some paper types, including magazines, newspapers, napkins, paper towels (or kitchen roll), and toilet papers may be handled more frequently. We assumed that the general population handles these five paper types ten times a day and the remaining paper types five times a day. Individuals who work at cash registers (e.g., cashiers, bank tellers), however, can handle receipts more frequently on a daily basis. For occupational exposures, we assumed that individuals handle thermal receipts 150 times a day, and handle other paper types at rates similar to the general population. Based on the geometric mean, median, fifth and 95th percentile concentrations of BPA measured in our paper samples, we estimated the daily intake (EDI; ng/day) of BPA as shown in eq 1: EDI ¼ k C HF HT AF=106
ð1Þ
where k is paper-to-skin transfer coefficient of BPA (calculated as 21522.4 ng/s); C is the concentration of BPA in paper samples (μg/g); HF is handling frequency (times/day; for thermal receipt paper, 2 and 150 times/day for the general population and occupationally exposed individuals, respectively; for other paper types, 5 times/day for flyer, ticket, mailing envelope, food
contact paper, food carton, airplane boarding pass, airplane luggage tag, printing paper, and business card, and 10 times/ day for magazine, newspapers, napkin, paper towel, and toilet paper for both the general population and occupationally exposed individuals); HT is handling time of paper and is assumed to be 5 s for each handling; and AF is the absorption fraction of BPA by skin, which is 27%.29 For example, if the median BPA concentration in thermal receipt is 0.299 mg/g (= 299 μg/g), the estimated median daily intakes of BPA, from handling of thermal receipts, by the occupationally exposed individual can be calculated as follows: EDI ¼ 21522:4ng=s 299μg=g 150times=day 5s=time 27%=106 ¼ 21:5224ng=s 0:299 150times=day 5s=time27% ¼ 1303ng=day
Statistical Analysis. Geometric mean (GM), median, and concentration ranges were used to describe the results. Concentrations below the LOQ were substituted with a value equal to the LOQ divided by the square root of 2 for the calculation of GM. Differences between groups were tested by a one-way ANOVA with the Tukey test.
’ RESULTS AND DISCUSSION Thermal Receipt Papers. BPA was detected in 94% of thermal receipt paper samples (n = 103) at concentrations ranging from below LOQ to 13.9 mg/g with a GM of 0.211 mg/g (Table 1). Of the 103 receipt samples analyzed, 73 (71%) were collected from Albany, New York. BPA concentrations in receipts from Albany were in the range of 0.005 to 9.38 mg/g with a GM of 0.341 mg/g (Figure 1). These values are slightly lower than those found in samples collected from other cities in the USA, including New York City, Buffalo, Boston, Chicago, Weston, and Charlotte (GM: 0.496 mg/g; one-way ANOVA, p > 0.05), but significantly lower than the concentrations found in samples from Incheon, Korea (GM: 1.56 mg/g; p < 0.01) and Hanoi, Vietnam (GM: 6.32 mg/g; p < 0.05). No significant difference was found in the concentrations of BPA in thermal receipt papers collected between Korea and Vietnam (Figure 1). Interestingly, BPA was not detected in any of the six thermal receipt paper samples collected from several stores in Matsuyama and Tokyo, Japan (Table 1). This is attributed to a phase-out of BPA usage in thermal receipt papers in Japan in 2001.41 Although BPA was reported to have been replaced with Bisphenol S by a major manufacturer of thermal receipt papers in the USA, BPA is still widely present in thermal receipt papers from the USA in 20102011. Some receipt papers claimed to be “BPA-free” (specifically printed on the receipt papers), but all of these receipt papers contained hundreds of μg/g levels of BPA (GM: 217 μg/g). A few earlier studies have reported BPA concentrations in thermal receipt papers.28,29 Eight of 10 blank (unprinted) thermal receipt papers collected from retail stores in Wilmington, Massachusetts, USA, contained BPA at concentrations ranging from
dx.doi.org/10.1021/es202507f |Environ. Sci. Technol. 2011, 45, 9372–9379
Environmental Science & Technology
ARTICLE
Table 1. BPA Concentrations in Thermal Receipt Papers (mg/g) From the USA, Japan, Korea, and Vietnam and in Other Types of Papers (μg/g) from Albany, New York, USA BPA concentration in thermal receipt papers (mg/g) site
n
a
b
GM
fifth percentile
median
95th percentile
Albany, USA
73
0.341
0.0069
0.269
other cities, USA
10
0.496
0.0144
3.26
Japanc Korea
6 11
0.0000007 1.56
0.0000007 0.0166
0.0000007 5.89
0.0000007 9.79
3
6.32
6.17
6.29
103
0.211
0.0005
0.299
Vietnam all thermal receipt papers
8.46 12.3
range
detection ratio (%)
0.00489.38
100
0.013113.9
100
0 100
6.51
6.156.53
100
9.31
94
BPA concentration in several papers and paper products (μg/g) paper type
a
n
GM
fifth percentile
median
95th percentile
range
detection ratio (%)
flyers
24
0.0230
0.0007
0.0226
0.346
88
magazines
5
0.0193
0.0083
0.0167
0.0520
0.00830.055
100
tickets
4
0.676
0.134
0.388
0.11814.4
100
mailing envelopes newspapers
5 8
0.0015 0.151
0.0007 0.0315
0.0007 0.277
12.3 0.0130 0.656
40 100
food contact papers
12
0.0038
0.0007
0.0014
0.248
50
food cartons
7
0.0134
0.0007
0.0471
0.0547
71
airplane boarding passes
4
0.0510
0.0132
0.0451
0.364
0.01240.415
100
airplane luggage tags
4
0.0031
0.0007
0.0007
0.212
25
printing paper
3
0.0197
0.0181
0.0203
0.0211
0.01780.0212
100
business cards
6
0.0238
0.0080
0.0279
0.0859
0.00750.100
100
napkins kitchen rolls (paper towels)
8 3
0.0110 0.0014
0.0009 0.0008
0.0034 0.0018
2.38 0.0022
88 67
toilet paper
6
0.0092
0.0019
0.0028
0.164
0.00180.180
100
all papers
99
0.0160
0.0007
0.0178
0.600
81
n: number of samples. b GM: geometric mean. c not detectable. d LOQ: 1 ng/g. e 1 out of 4 samples was detectable.
reported from Switzerland (Figure 1; p < 0.001). Further, 100% of the thermal receipt papers collected from the USA contained BPA. It should be noted that our LOQ (1 ng/g) was two to 4 orders of magnitude lower than the LOQs reported in the two earlier studies (26 μg/g and 0.5 μg/g). The earlier studies reported the occurrence of BPA in 8085% of the thermal receipt papers collected from grocery stores.28,29 The concentration/distribution of BPA in several portions of thermal receipt papers was compared. Six thermal receipt samples were randomly selected, and five 19 mm circular spots were taken from each receipt: upper left corner (denoted as ULC), lower left corner (LLC), middle (M), upper right corner (URC), and lower right corner (LRC). No remarkable difference in the concentrations of BPA was found among the five spots taken within each receipt paper (Figure S1; Supporting Information). BPA has been used for the elastification of phenolic resins that are used as binding agents in printing inks. Thus, printing ink is considered a potential source of BPA in papers.32 To examine potential sources of BPA on receipts, we determined BPA concentrations in two blank (unprinted) receipt-paper rolls, purchased from an office products supply store in Albany, New York. Of the two receipt-paper rolls, one was thermal and the other was regular ink-printing paper. The blank (unprinted) thermal receipt paper was analyzed in triplicate in two batches, and the concentration of BPA was found to be 8.88 ( 1.39 mg/g (mean ( SD), which was greater than the mean concentration of
BPA (3.06 ( 3.65 mg/g) found for all printed thermal receipt samples combined. The regular (i.e., non-thermal) ink-paper receipt was also analyzed in triplicate, and the concentration of BPA was found to be 0.050 ( 0.005 μg/g, which was lower than the concentration of BPA (0.26 ( 1.48 μg/g) found in nonthermal papers analyzed in this study. Our results suggest that BPA contamination in thermal receipt papers originates mainly from the use/coating of BPA on the thermal paper as a color developer rather than from printing ink.28 Paper and Paper Products. Because BPA in thermal receipt paper is not chemically bound, it can easily be transferred from thermal paper to other objects, including various types of papers.28,29 BPA contamination in paper products can arise from recycling of thermal receipt paper, along with other papers, and from the use of BPA for the elastification of phenolic resins in printing inks.28,29 Recycled paper is used in the production of a wide range of paper products, from toilet papers, paper towels, newspapers to cartons for snack foods and cardboard boxes.32,34,37,38 Recycling of thermal papers along with other papers can increase the risk of human exposure to BPA via crosscontamination of foods stored in recycled paper products. BPA has been found in recycled paper towels and other food-contact papers at higher levels than is found in virgin papers.32 In this study, 14 types of paper and paper products contained BPA, in 81% of the samples analyzed. The overall BPA concentrations in the 99 paper samples ranged from < LOQ to 14.4 μg/g (GM: 0.016 μg/g). BPA concentrations were similar among the 13 9375
dx.doi.org/10.1021/es202507f |Environ. Sci. Technol. 2011, 45, 9372–9379
Environmental Science & Technology
Figure 1. Comparison of BPA concentrations in thermal receipt papers from different countries. The box plot shows fifth (lower whisker), 25th (bottom edge of the box), 75th (top edge of the box), and 95th (upper whisker) percentiles. The lower and upper stars represent 1st and 99th percentiles, respectively. The arithmetic mean and median concentrations are given as the open square and the line within the box, respectively. The dots are outliers. The number in parentheses is the sample number for each group. In x-axis, “other cities, USA” includes New York City and Buffalo in New York, Boston in Massachusetts, Chicago in Illinois, Weston in Vermont, and Charlotte in North Carolina; “two cities, Japan” includes Matsuyama and Tokyo. The two plots on the right were based on published values, and the number within parentheses adjacent to x-axis is the reference number.
paper types, except for tickets, which contained elevated concentrations (p < 0.05). Usually, people place tickets aside thermal receipts in their wallets and tickets can be contaminated by BPA via contacting with thermal receipts. One outlier value was also found in a ticket sample (14.4 μg/g, Figure S2, Supporting Information), which elevated the overall mean concentration of BPA in tickets. BPA concentrations in paper products were three to 4 orders of magnitude lower than the concentrations found in thermal receipt papers (Figure S2, Supporting Information). BPA was detected in 100% of magazine papers, tickets, newspapers, airplane boarding passes, printing paper, business cards, and toilet papers (Table 1). Most of these products were made from recycled paper.32,34,37,38 Our results indicate that exposure of humans to BPA via paper products is ubiquitous because these papers are frequently used on a daily basis. The highest concentration of BPA was found in tickets (GM: 0.676 μg/g; Table 1), followed by newspapers (0.151 μg/g) and flyers (0.023 μg/g). BPA concentrations in ticket samples were 2 orders of magnitude higher than the concentrations found in mailing envelopes (0.0015 μg/g) and kitchen rolls (i.e., paper towels) (0.0014 μg/g). It is likely that paper products made from recycled papers contain BPA due to contamination arising from the recycling process, as explained above. Few studies have reported the occurrence of BPA in paper and paper products.32,34,37,38 Vinggaard et al. 32 collected 20 brands of paper towels, nine of which were made from recycled paper and the remainder from virgin pulp, from retail stores in Copenhagen, Denmark. BPA was detected at concentrations of 0.6 to 24.1 μg/g (GM: 3.9 μg/g) in recycled paper products, and the concentrations in virgin papers were below 0.04 μg/g, with the
ARTICLE
exception of one sample that contained 0.1 μg BPA/g. The concentrations of BPA reported for paper towels from Denmark were two to 3 orders of magnitude higher than the concentrations found in our paper towels (range: < LOQ to 0.0022; GM: 0.0014 μg/g) and toilet paper (range: 0.0018 to 0.180; GM: 0.0092 μg/g; Table 1). Gehring et al. 34 collected three kinds of toilet papers made from 100% recycled paper and municipal waste paper (including flyers, magazines, and newspapers) from local supermarkets in Dresden, Germany. BPA was found in all types of paper samples, at concentrations ranging from 3.2 to 46.1 μg/g (GM: 18.9 μg/g) for toilet paper and from 0.09 to 5.1 μg/g (1.24 μg/g) for other waste paper types. Ozaki et al.37 collected 28 paper and cardboard products in Osaka, Japan, which have been used as food containers (e.g., cereal boxes), 12 of which were made from recycled paper and the remainder from virgin pulp. BPA was present in both virgin and recycled paper samples (67%), at a concentration range of 160 000 tons per year.45 On the basis of the median and 95th percentile concentrations of BPA found in thermal receipt papers (0.299 and 9.31 mg/g) analyzed in our study (Table 1), and the estimation that 30% of the thermal paper enters recycling processes of municipal wastepaper (i.e., 70% of the thermal paper is released into the environment),33 we calculated that from 33.5 (based on median) to 1040 (based on 95th percentile) tons of BPA are released into the environment in the USA and Canada through the disposal of thermal receipt papers. This is an underestimate because it does not take into account the production estimates for thermal papers from other manufactures and imports from other countries. Many paper products, such as paper towels, facial tissue, advertisement brochures (flyers), magazines, and newspapers, are made from recycled paper. The occurrence of BPA in various paper products can depend on the proportion of wastepaper introduced into the production process.32,34,37 Many types of 9376
dx.doi.org/10.1021/es202507f |Environ. Sci. Technol. 2011, 45, 9372–9379
Environmental Science & Technology
ARTICLE
Table 2. Estimated Daily Intake (ng/day) of BPA, via Handling of Papers, by the General Population and Occupationally Exposed Individuals general population paper type thermal receipts
a
GM 12.3
fifth percentile 0.0279
median 17.4
occupational exposure 95th percentile 541
GM 921
fifth percentile 2.09
median 1303
95th percentile 40590
flyers
0.0033
0.0001
0.0033
0.0503
0.0033
0.0001
0.0033
0.0503
magazines tickets
0.0056 0.0982
0.0024 0.0194
0.0048 0.0564
0.0151 1.79
0.0056 0.0982
0.0024 0.0194
0.0048 0.0564
0.0151 1.79
mailing envelopes
0.0002
0.0001
0.0001
0.0019
0.0002
0.0001
0.0001
0.0019
newspapers
0.0438
0.0091
0.0803
0.1906
0.0438
0.0091
0.0803
0.1906
food contact papers
0.0006
0.0001
0.0002
0.0360
0.0006
0.0001
0.0002
0.0360
food cartons
0.0019
0.0001
0.0068
0.0080
0.0019
0.0001
0.0068
0.0080
airplane boarding passes
0.0074
0.0019
0.0065
0.0528
0.0074
0.0019
0.0065
0.0528
airplane luggage tags
0.0004
0.0001
0.0001
0.0308
0.0004
0.0001
0.0001
0.0308
printing paper business cards
0.0029 0.0035
0.0026 0.0012
0.0029 0.0041
0.0031 0.0125
0.0029 0.0035
0.0026 0.0012
0.0029 0.0041
0.0031 0.0125
napkins
0.0032
0.0003
0.0010
0.693
0.0032
0.0003
0.0010
0.693
kitchen rolls (paper towels)
0.0004
0.0002
0.0005
0.0006
0.0004
0.0002
0.0005
0.0006
toilet paper
0.0027
0.0005
0.0008
0.0477
0.0027
0.0005
0.0008
total exposure (∑EDI)a
12.5
exposure percentage from receipt (%)
98.6
0.066 42.2
17.5
544
99.0
99.5
921 100.0
2.13 98.2
1303 100.0
0.0477 40593 100.0
Rounded values.
paper and paper products are widely used as food packaging cartons and cardboard boxes, which are in direct contact with foodstuffs and can contribute to contamination of foods. When the food is heated (e.g., microwaved) with the paper product, this may accelerate the migration of BPA to the food.32 Considering that paper products are placed in indoor environments, indoor dust 15,23,46 can be an additional source of BPA contamination in paper products and vice versa. Exposure of Humans to BPA via Handling of Papers. The estimated fifth percentile, median, and 95th percentile daily intakes of BPA through dermal absorption, from handling of thermal receipt papers, by the general population were 0.028, 17.4, and 541 ng/day, respectively; the corresponding values for occupationally exposed individuals were 2.09, 1303 and 40 590 ng/day (Table 2). Among the other paper types, the highest daily BPA intakes (calculated from fifth percentile, median and 95th percentile concentrations in papers) by the general population and occupationally exposed individuals were from the handling of tickets (0.019, 0.056, and 1.79 ng/day) and newspapers (0.009, 0.080, and 0.191 ng/day). The estimated median and 95th percentile values for the total daily intakes of BPA (∑EDI) from all 15 types of paper products analyzed in this study were 17.5 and 544 ng/day for the general population, and 1303 and 40 593 ng/day for occupationally exposed individuals, respectively (Table 2). It should be noted that the EDI values for BPA via handling of 14 types of papers (except for thermal receipt paper) by the general population and occupationally exposed individuals were same in our exposure assessments. The EDI values for BPA from the handling of paper products were 10-fold higher than those (8.44 and 43.5 ng/day) reported from dust ingestion in the USA.15 Despite the high concentrations of BPA found in several types of paper products (e.g., tickets, newspapers, napkins), thermal receipt paper contributed to the preponderance of the daily BPA exposures (>98%; Table 2), with an exception of the fifth percentile value for the general population (42%;
Table 2). This is mainly because thermal receipt papers contain extremely high concentration of BPA (Figure 1). For the calculation of intakes adjusted for body weight (bw), the values were divided by a nominal body weight of 80 kg for adults. The estimated median daily intake values of BPA were 0.219 and 16.3 ng/kg bw/day for the general population and occupationally exposed individuals, respectively. Several previous studies have estimated BPA exposures in the range of 0.008 to 5 μg/kg bw/day for adults in the USA.3,21,22 Assuming a total BPA intake value of 1 μg/kg bw/day, the dermal contact with paper contributes 1.6% (based on median) to 51% (based on the 95th percentile) of the total intake for occupationally exposed individuals. The EDI values calculated for BPA from paper products were several orders of magnitude lower than the RfD of 50 μg/kg bw/day (based on oral toxicity) established by the USEPA and the EFSA.7,8 There are several uncertainties in our estimate of BPA exposures from paper products. A doseresponse relationship for the transfer coefficient of BPA from paper products to skin has not yet been established. The EDI can be higher when paper products are handled with wet or greasy fingers or by hand-tomouth contact.29 The duration and frequency of exposures can be much higher than what was used in our calculations. Inhalation during the handling of papers can be an important source of BPA exposures, and this was not estimated. Transfer of BPA from food-cartons to foods can be another source of human exposure. A recent study indicated that human exposure to BPA from unknown sources is much higher than what was previously assumed because many sources have still not been identified or characterized.24 Studies have also reported that BPA at doses as low as tens to hundreds of ng/kg bw/day can cause adverse endocrine disruptive effects,9,27 suggesting the significance of identifying all potential sources of human exposures. Our study fills an important knowledge gap by identifying paper products as a source of BPA exposure in the general population. 9377
dx.doi.org/10.1021/es202507f |Environ. Sci. Technol. 2011, 45, 9372–9379
Environmental Science & Technology In summary, BPA was detected in most thermal receipt paper samples at very high concentrations. BPA concentrations in thermal receipt papers from the USA were lower than the concentrations found in samples from Korea and Vietnam. BPA was not found in thermal receipt papers from Japan. On the basis of the annual consumption of thermal paper in the USA and Canada, and median concentrations of BPA measured in thermal receipt papers (0.299 mg/g), we estimated that approximately 33.5 tons of BPA are released into the environment through discharge of thermal paper in the USA and Canada every year. BPA was found in other types of paper and paper products (such as newspapers, toilet paper, napkins, paper towels) that are frequently used on a daily basis. The estimated median daily intake values of BPA through dermal absorption from touching of paper products were 0.219 and 16.3 ng/kg bw/day for the general population and occupationally exposed individuals, respectively. The transfer of BPA from paper products into air and food, and subsequent human exposures, need to be examined in future studies.
’ ASSOCIATED CONTENT
bS
Supporting Information. Details of the analytical method and a figure showing comparison of BPA in different portions of thermal receipt papers and a figure showing distribution of BPA in paper products. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 1-518-474-0015; fax: 1-518-473-2895; e-mail: kkannan@ wadsworth.org.
’ REFERENCES (1) Vandenberg, L. N.; Chahoud, I.; Heindel, J. J.; Padmanabhan, V.; Paumgartten, F. J.; Schoenfelder, G. Urinary, circulating, and tissue biomonitoring studies indicate widespread exposure to bisphenol A. Environ. Health Perspect. 2010, 118, 1055–1070. (2) Vandenberg, L. N.; Maffini, M. V.; Sonnenschein, C.; Rubin, B. S.; Soto, A. M. Bisphenol-A and the great divide: A review of controversies in the field of endocrine disruption. Endocr. Rev. 2009, 30 (1), 75–95. (3) Vandenberg, L. N.; Hauser, R.; Marcus, M.; Olea, N.; Welshons, W. V. Human exposure to bisphenol A (BPA). Reprod. Toxicol. 2007, 24, 139–177. (4) Hengstler, J. G.; Foth, H.; Gebel, T.; Kramer, P. J.; Lilienblum, W.; Schweinfurth, H.; V€olkel, W.; Wollin, K. M.; Gundert-Remy, U. Critical evaluation of key evidence on the human health hazards of exposure to bisphenol A. Crit. Rev. Toxicol. 2011, 41 (4), 263–291. (5) Golub, M. S.; Wu, K. L.; Kaufman, F. L.; Li, L. H.; MoranMessen, F.; Zeise, L.; Alexeeff, G. V.; Donald, J. M. Bisphenol A: Developmental toxicity from early prenatal exposure. Birth. Defects Res. B Dev. Reprod. Toxicol. 2010, 89 (6), 441–466. (6) Goodman, J. E.; Witorsch, R. J.; McConnell, E. E.; Sipes, I. G.; Slayton, T. M.; Yu, C. J.; Franz, A. M.; Rhomberg, L. R. Weight-ofevidence evaluation of reproductive and developmental effects of low doses of bisphenol A. Crit. Rev. Toxicol. 2009, 39 (1), 1–75. (7) USEPA (United States Environmental Protection Agency). Childspecific exposure factors handbook, EPA/600/R-06/096F, Sep 2008, http://www.epa.gov/ncea; National Center for Environmental Assessment, Office of Research and Development: Washington, DC, 2008. (8) EFSA (European Food Safety Authority). Opinion of the Scientific Panel on food additives, flavourings, processing aids, and
ARTICLE
materials in contact with food on a request from the commission related to bisphenol A question number EFSA-Q-2005100. EFSA J. 2006, 428, 175. (9) vom Saal, F. S.; Hughes, C. An extensive new literature concerning low-dose effects of bisphenol A shows the need for a new risk assessment. Environ. Health Perspect. 2005, 113, 926–933. (10) Bailin, P. D.; Byrne, M.; Lewis, S.; Liroff, R. Public awareness drives market for safer alternatives: Bisphenol A market analysis report. http://www.iehn.org/publications.reports.bpa.php, 2008. (11) Alonso-Magdalena, P.; Ropero, A. B.; Soriano, S.; Quesada, I.; Nadal, A. Bisphenol-A: A new diabetogenic factor?. Hormones (Athens) 2010, 9 (2), 118–126. (12) Lim, D. S.; Kwack, S. J.; Kim, K. B.; Kim, H. S.; Lee, B. M. Potential risk of bisphenol A migration from polycarbonate containers after heating, boiling, and microwaving. J. Toxicol. Environ. Health A. 2009, 72 (2122), 1285–1291. (13) Tsai, W. T. Human health risk on environmental exposure to bisphenol-A: A review. J. Environ. Sci. Health, Part C: Environ. Carcinog. Ecotoxicol. Rev. 2006, 24 (2), 225–255. (14) Clarke, B. O.; Smith, S. R. Review of ’emerging’ organic contaminants in biosolids and assessment of international research priorities for the agricultural use of biosolids. Environ. Int. 2011, 37 (1), 226–247. (15) Loganathan, S. N.; Kannan, K. Occurrence of Bisphenol A in indoor dust from two locations in the Eastern United States and implications for human exposures. Arch. Environ. Contam. Toxicol. 2011, 61, 68–73. (16) Inoue, K.; Yamaguchi, A.; Wada, M.; Yoshimura, Y.; Makino, T.; Nakazawa, H. Quantitative detection of bisphenol A and bisphenol A diglycidyl ether metabolites in human plasma by liquid chromatographyelectrospray mass spectrometry. J. Chromatogr. B 2001, 765, 121–126. (17) Dekant, W.; V€olkel, W. Human exposure to bisphenol A by biomonitoring: Methods, results and assessment of environmental exposures. Toxicol. Appl. Pharmacol. 2008, 228, 114–134. (18) Calafat, A. M.; Ye, X.; Wong, L. Y.; Reidy, J. A.; Needham, L. L. Exposure of the U.S. population to bisphenol A and 4-tertiary-octylphenol: 20032004. Environ. Health Perspect. 2008, 116, 39–44. (19) Padmanabhan, V.; Siefert, K.; Ransom, S.; Johnson, T.; Pinkerton, J.; Anderson, L.; Tao, L.; Kannan, K. Maternal bisphenol-A levels at delivery: A looming problem?. J. Perinatol. 2008, 28, 258–263. (20) He, Y.; Miao, M.; Herrinton, L. J.; Wu, C.; Yuan, W.; Zhou, Z.; Li, D. K. Bisphenol A levels in blood and urine in a Chinese population and the personal factors affecting the levels. Environ. Res. 2009, 109 (5), 629–633. (21) von Goetz, N.; Wormuth, M.; Scheringer, M.; Hungerb€uhler, K. Bisphenol A: How the most relevant exposure sources contribute to total consumer exposure. Risk Anal. 2010, 30 (3), 473–487. (22) National Toxicology Program, U.S. Department of Health and Human Services (20071126). “CERHR Expert Panel Report for Bisphenol A” (PDF). Archived from the original on 20080218. http://web.archive.org/web/20080218195117/http://cerhr.niehs. nih.gov /chemicals/bisphenol/BPAFinalEPVF112607.pdf. Retrieved 20080418. (23) Wilson, N. K.; Chuang, J. C.; Morgan, M. K.; Lordo, R. A.; Sheldon, L. S. An observational study of the potential exposures of preschool children to pentachlorophenol, bisphenol-A, and nonylphenol at home and daycare. Environ. Res. 2007, 103, 9–20. (24) Taylor, J. A.; vom Saal, F. S.; Welshons, W. V.; Drury, B.; Rottinghaus, G.; Hunt, P. A.; Vandevoort, C. A. Similarity of Bisphenol A pharmacokinetics in Rhesus monkeys and mice: Relevance for human exposure. Environ. Health Perspect. 2011, 119 (4) 422430. (25) Jackson, W. J.; Darnell, W. R. Process for foaming cellulose acetate rod. US Patent 4,507,256, filed 26 May 1983, and issued 26 March 1985. (26) Braun, J. M.; Kalkbrenner, A. E.; Calafat, A. M.; Bernert, J. T.; Ye, X.; Silva, M. J.; Barr, D. B.; Sathyanarayana, S.; Lanphear, B. P. Variability and predictors of urinary bisphenol A concentrations during pregnancy. Environ. Health Perspect. 2011, 119 (1), 131–137. 9378
dx.doi.org/10.1021/es202507f |Environ. Sci. Technol. 2011, 45, 9372–9379
Environmental Science & Technology (27) FAO/WHO. Joint FAO/WHO Expert Meeting to Review Toxicological and Health Aspects of Bisphenol A. 2010 [cited November 25, 2010]; Available from: http://www.who.int/foodsafety/chem/ chemicals/BPA_Summary2010.pdf. (28) Mendum, T.; Stoler, E.; VanBenschoten, H.; Warner, J. C. Concentration of bisphenol A in thermal paper. Green Chem. Lett. Rev. 2010, 1–6iFirst article.. (29) Biedermann, S.; Tschudin, P.; Grob, K. Transfer of bisphenol A from thermal printer paper to the skin. Anal. Bioanal. Chem. 2010, 398 (1), 571–576. (30) Calafat, A. M.; Weuve, J.; Ye, X.; Jia, L. T.; Hu, H.; Ringer, S.; Huttner, K.; Hauser, R. Exposure to bisphenol A and other phenols in neonatal intensive care unit premature infants. Environ. Health Perspect. 2009, 117 (4), 639–644. (31) Joskow, R.; Barr, D. B.; Barr, J. R.; Calafat, A. M.; Needham, L. L.; Rubin, C. Exposure to bisphenol A from bis-glycidyl dimethacrylate-based dental sealants. J. Am. Dent. Assoc. 2006, 137 (3), 353–362. (32) Vinggaard, A. M.; K€orner, W.; Lund, K. H.; Bolz, U.; Petersen, J. H. Identification and quantification of estrogenic compounds in recycled and virgin paper for household use as determined by an in vitro yeast estrogen screen and chemical analysis. Chem. Res. Toxicol. 2000, 13 (12), 1214–1222. (33) European Commission-Joint Research Centre. European Union Risk Assessment Report, 4,40 -Isopropylidenediphenol (BisphenolA). 2008 [cited April 8, 2011]; Available from: http://ecb.jrc.ec.europa. eu/documents/Existing-Chemicals/RISK_ASSESSMENT /ADDENDUM/bisphenola_add_325.pdf. (34) Gehring, M.; Tennhardt, L.; Vogel, D.; Weltin, D.; Bilitewski, B. Bisphenol A Contamination of Wastepaper, Cellulose and Recycled Paper Products. In Waste Management and the Environment II. WIT Transactions on Ecology and the Environment; Brebbia, C. A., Kungulos, S., Popov, V., Itoh, H., Eds.;Southampton, Boston: WIT Press, 2004; Vol. 78, 294300. (35) Zalko, D.; Jacques, C.; Duplan, H.; Bruel, S.; Perdu, E. Viable skin efficiently absorbs and metabolizes bisphenol A. Chemosphere 2011, 82 (3), 424–430. (36) von Goetz, N.; Wormuth, M.; Scheringer, M.; Hungerb€uhler, K. Bisphenol A: How the most relevant exposure sources contribute to total consumer exposure. Risk Anal. 2010, 30 (3), 473–487. (37) Ozaki, A.; Yamaguchi, Y.; Fujita, T.; Kuroda, K.; Endo, G. Chemical analysis and genotoxicological safety assessment of paper and paperboard used for food packaging. Food Chem. Toxicol. 2004, 42 (8), 1323–1337. (38) Lopez-Espinosa, M. J.; Granada, A.; Araque, P.; Molina-Molina, J. M.; Puertollano, M. C.; Rivas, A.; Fernandez, M.; Cerrillo, I.; OleaSerrano, M. F.; Lopez, C.; Olea, N. Oestrogenicity of paper and cardboard extracts used as food containers. Food. Addit. Contam. 2007, 24 (1), 95–102. (39) Kaddar, N.; Harthe, C.; Dechaud, H.; Mappus, E.; Pugeat, M. Cutaneous penetration of bisphenol A in pig skin. J. Toxicol. Environ. Health A. 2008, 71 (8), 471–473. (40) Mørck, T. J.; Sorda, G.; Bechi, N.; Rasmussen, B. S.; Nielsen, J. B.; Ietta, F.; Rytting, E.; Mathiesen, L.; Paulesu, L.; Knudsen, L. E. Placental transport and in vitro effects of Bisphenol A. Reprod. Toxicol. 2010, 30 (1), 131–137. (41) National Institute of Technology and Evaluation. Summary of the Interim Report - Bisphenol A. National Institute of Technology and Evaluation - Japan. 2003 [cited April 8, 2011]; Available from: http:// www.safe.nite.go.jp/risk/pdf/interimreport_summary_bpa.pdf. (42) Raloff, J. Receipts a large—and largely ignored—source of BPA. ScienceNews.org. 2010 [cited April 18, 2011]; Available from: http://www.sciencenews.org/view/generic/id/61764 /title/Receipts_ a_large_%E2%80%94_and_largely_ignored_%E2%80% 94_source_of_BPA. (43) Appleton. Nation’s Largest Maker of Thermal Receipt Paper Does Not Use BPA. AppletonIdeas.com. 2010 [cited April 18, 2011]; Available from: http://www. appletonideas.com/pdf/Appleton% 20BPA%20free%20news%20release.7.27.2010.pdf.
ARTICLE
(44) Fukazawa, H.; Hoshino, K.; Shiozawa, T.; Matsushita, H.; Terao, Y. Identification and quantification of chlorinated bisphenol A in wastewater from wastepaper recycling plants. Chemosphere 2001, 44 (5), 973–979. (45) Appleton Papers Inc. Securities and exchange commission. Washington, D.C. 20549. 2002 [cited April 18, 2011]; Available from: http://www.appletonideas.com/pdf/200310k.pdf. (46) Rudel, R. A.; Camann, D. E.; Spengler, J. D.; Korn, L. R.; Brody, J. G. Phthalates, alkylphenols, pesticides, polybrominated diphenyl ethers, and other endocrine-disrupting compounds in indoor air and dust. Environ. Sci. Technol. 2003, 37, 4543–4555.
9379
dx.doi.org/10.1021/es202507f |Environ. Sci. Technol. 2011, 45, 9372–9379
ARTICLE pubs.acs.org/est
Environmental Impact of Pyrolysis of Mixed WEEE Plastics Part 1: Experimental Pyrolysis Data Sue M. Alston,*,† Allan D. Clark,‡ J. Cris Arnold,† and Bridget K. Stein§ †
Materials Research Centre, Swansea University, Singleton Park, Swansea SA2 8PP, U.K. ITEM Wales Ltd, 11 Heol Morlais, Llannon, Carmarthenshire SA14 6BD, U.K. § EPSRC National Mass Spectrometry Service Centre, Swansea SA2 8PP, U.K. ‡
bS Supporting Information ABSTRACT: Growth in waste electrical and electronic equipment (WEEE) is posing increasing problems of waste management, partly resulting from its plastic content. WEEE plastics include a range of polymers, some of which can be sorted and extracted for recycling. However a nonrecyclable fraction remains containing a mixture of polymers contaminated with other materials, and pyrolysis is a potential means of recovering the energy content of this. In preparation for a life cycle assessment of this option, described in part 2 of this paper set, data were collected from trials using experimental pyrolysis equipment representative of a continuous commercial process operated at 800 °C. The feedstock contained acrylonitrile-butadiene-styrene and high impact polystyrene with high levels of additives, and dense polymers including polyvinylchloride, polycarbonate, polyphenylene oxide, and polymethyl methacrylate. On average 39% was converted to gases, 36% to oils, and 25% remained as residue. About 35% of the gas was methane and 42% carbon monoxide, plus other hydrocarbons, oxygen and carbon dioxide. The oils were almost all aromatic, forming a similar mixture to fuel oil. The residue was mainly carbon with inorganic compounds from the plastic additives and most of the chlorine from the feedstock. The results showed that the process produced around 70% of the original plastic weight as potential fuel.
’ INTRODUCTION The growth in waste electrical and electronic equipment (WEEE) is posing an increasing problem of waste management. The EU was estimated to dispose of 9.5 million tonnes of WEEE in 2008, and this was forecast to rise to 12.3 million tonnes by 2020.1,2 Historically the only WEEE recycled on a large scale in the UK has been large household appliances (e.g., washing machines, refrigerators), either because they contain ozone depleting substances or for their scrap metal value. The remainder has been untreated and mostly sent to landfill.3 However landfill space is becoming scarcer, and there is a need to make use of products’ embodied energy both to decrease dependency on natural sources of fuel, and to reduce the climate change implications of further energy use. Typically WEEE contains 20 25% plastics.4 A range of different polymers are in use; an example of the breakdown from actually collected WEEE is acrylonitrile-butadiene-styrene (ABS) 30%, high impact polystyrene (HIPS) 25%, polycarbonate (PC) 10%, PC/ABS 9%, polypropylene (PP) 8%, polyphenylene ether(PPE)/HIPS 7%, polyvinyl chloride (PVC) 3%, polystyrene (PS) 3%, polyamide (PA) 3%, polybutylene terephthalate (PBT) 2%.5 Recent improvements in sorting methods are enabling an increasing proportion of these to be extracted for mechanical recycling. Density sorting is most commonly used r 2011 American Chemical Society
and can separate light polymers (e.g., PP, polyethylene (PE)) and medium ones (e.g., ABS, HIPS without significant additives). However, this leaves a mixed “heavy fraction” containing less common denser polymers, those with high additive content (including brominated flame retardants (BFRs)) and contaminants. The complexity of the mix and small quantity of individual polymers mean that this is not currently recycled. A possible alternative to landfill for treatment of this heavy fraction is pyrolysis, giving breakdown products which can be used as fuels in place of gas, diesel or fuel oils. Many studies of pyrolysis of relevant polymers have been published, showing a wide range of gas and liquid yields. Pyrolysis of styrenic polymers such as ABS, HIPS, and styrene-acrylonitrile (SAN) has been reported to produce benzene, toluene, styrene, ethylbenzene, 2-propenenitrile, naphthalene, and a range of other aromatic compounds; in the case of acrylonitriles the aromatics may have attached nitrile functional groups and quinoline may be produced.6 9 The gas yields found varied considerably, from negligible8 to 26%6 at 400 500 °C, and increasing to over 40% Received: May 16, 2011 Accepted: September 22, 2011 Revised: September 21, 2011 Published: September 22, 2011 9380
dx.doi.org/10.1021/es201664h | Environ. Sci. Technol. 2011, 45, 9380–9385
Environmental Science & Technology
ARTICLE
Figure 1. Schematic of experimental rig.
at 875 °C.7 Results for polycarbonate at a slower temperature ramp showed that the oxygen content in the polymer led to phenol compounds in the oils and carbon dioxide in the gas.10,11 The presence of brominated flame retardants in the feedstock could lead to brominated organic compounds in the oils and antimony tribromide in both the oils and residue.12,8,9 PVC undergoes two stages of decomposition during slow heating, with the formation of hydrogen chloride at around 300 °C followed by the production of aromatics from the remaining polyene backbone at about 450 °C. If PVC is mixed with other polymers then the hydrogen chloride can react with other breakdown products to form chlorinated organic compounds.13 Polyolefins have been found to produce higher proportions of gas, ranging from 34 to 57%, with oils containing linear hydrocarbons as well as some benzene, toluene and naphthalene.14 16 The addition of 20% PE or PP to HIPS was found to increase the gas proportion from less than 1% to around 8%.17,18 The products from WEEE plastics of specific types (e.g., wire cladding, table phones, mobile phones) reflected the polymers involved in each case,19 with the trials being carried out at a slow ramp rate and resulting in relatively low gas levels. The environmental effect of using pyrolysis to deal with WEEE plastics can only be assessed if data is available for the specific process being considered, since published research clearly indicates a significant dependence of pyrolysis products on the plastic mix, temperature and ramp rate. A commercial process is likely to be continuous, in order to maximize productivity, meaning that feedstock would be introduced to a preheated chamber and undergo very rapid heating. It would also operate at a temperature sufficiently high to avoid dioxin formation, probably above 600 °C. In preparation for carrying out a life cycle assessment (LCA) to establish the environmental impact, experimental work was undertaken to identify the pyrolysis products from such a process. The equipment design was based on a pilot plant developed and patented by ITEM Technology Solutions Ltd.20 which had been used for plastics from agricultural and municipal waste. This study looked specifically at the
potential for use of this process for WEEE plastics remaining after density sorting. This paper forms the first part of a two part paper set, with part 2 describing the LCA study based on the data.
’ MATERIALS AND METHODS Materials. The trial feedstock was obtained from Axion Polymers (Salford, UK) and consisted of the “heavy fraction” of mixed WEEE plastics after shredding and density sorting. This fraction was expected to contain all polymers heavier than ABS and HIPS, together with those containing significant amounts of higher density additives, including BFRs. Two batches of shredded material were provided, in the form of granules from 1 mm to 10 mm in size. Pyrolysis Process. The experimental equipment was designed to be as representative as possible of the ITEM Technology Solutions Ltd. process. The design is shown in the schematic in Figure 1. Feed material was ground in a ball mill to produce granules less than 4 mm in size, then introduced to the rig through two gate valves, between which an argon purge was fed. When released from the second valve, the material fell into a “boat”, which was then pushed into the pyrolysis cylinder, rotated to deposit the material, then retracted. The pyrolysis cylinder itself was rotated during operation to allow the feed material to reach temperature as quickly as possible. The pyrolysis products then passed through a filter, maintained at 400 °C, to remove dust and soot. Two flasks containing glass balls, kept cool with ice, were used to condense out volatiles, and then the gases passed through distilled water to dissolve any soluble gases such as hydrogen chloride, hydrogen bromide, hydrogen cyanide, or ammonia. Gas sampling was carried out via a T-piece and valve after the last flask. A pressure gauge at the entry to the filter and a valve at the system outlet were used to control to around 20 mbar above atmospheric pressure to reduce risk of air ingress. Repeated “boatfuls” of feed material were processed until a steady flow of 9381
dx.doi.org/10.1021/es201664h |Environ. Sci. Technol. 2011, 45, 9380–9385
Environmental Science & Technology
ARTICLE
Figure 2. Comparison of residues with ground polymers and ash for (a) Batch 1 and (b) Batch 2 (note different scales).
pyrolysis gases reached the final outlet, at which point samples were taken. The residence time was estimated as around 6 min. Analytical Methods. In order to get an indication of the difference between the two batches of plastic supplied, a basic density sorting exercise was carried out by dropping a random sample of granules from each batch into salt solution at a specific gravity of 1.1, determined using a hydrometer. The density level was selected so that polyolefins and ABS and HIPS granules without significant additives would float, but those with BFRs, plus heavier polymers such as PVC, would sink. It was found that 60% of the second batch sank, compared to only 16% of the first batch. This indicated that the sorting method had been tuned between the batches so as to retain more of the higher quality, lighter material for mechanical recycling. Polymer type analysis of individual granules was carried out using Fourier transform infrared spectroscopy (FTIR), after compression molding to provide a suitable surface. This used a Perkin-Elmer Spectrum One FT-IR spectrometer in reflectance mode. A random sample of 211 granules from Batch 1 contained 46.3% ABS/SAN, 40.9% HIPS, 1.5% PVC, 1.2% PP, and 10.1% other; a more limited but visually representative sample of 25 granules from Batch 2 contained 12% ABS/SAN, 20% HIPS, 40% PVC and other chlorinated polymers, and 28% other. The increased percentage of chlorinated and “other” polymers in Batch 2 correlated with the higher proportion of denser granules in this batch. The elemental content of the plastics was identified using energy dispersive X-ray analysis (EDX) with a Jeol JSM-35C scanning electron microscope linked to Oxford Instruments Link ISIS software. Samples were prepared from a random selection of granules by two methods, chosen so as to obtain results representative of the bulk of the plastics not just the surface. These were, first, to grind the granules to a powder, and second, to heat them in air at 700 °C until all organic content had been removed to produce an ash. In each case the resulting powder was scattered on to an adhesive disk 1 cm in diameter and an average taken from five scans, each covering a different area of around 1 mm2. In the case of the ground polymer each scan covered about 100 different particle surfaces, mostly from the inside of original granules. The ashing process entirely removed the original granule structure to give a much finer powder. There is the possibility of some additional oxygen being incorporated onto these newly created surfaces during their preparation, but
the oxygen content from EDX was not a significant factor in the LCA. The results are shown as part of Figure 2. Batch 2 showed a significantly higher percentage of chlorine, corresponding to the higher proportion of PVC, and also of calcium; calcium carbonate is commonly used as filler for PVC, sometimes with a calcium stearate coating. Generally the ground polymer and ash results showed a similar pattern. Where differences did occur the reasons for these were not entirely clear, although some elements other than carbon might have been lost during the ashing process. Bromine was not found in the ground polymer but EDX analysis of individual granules confirmed its presence as found in the ash, suggesting that the concentration in the overall mix was too low to be detected. The amount found in the ash was taken to be a reasonable indication of the amount in the polymer; the results from individual granules were for a smaller sample and for the surface only so could not be treated as quantitative. The pyrolysis results were based on four trials. One was carried out on the first batch of material and three on the second batch of material. All used a pyrolyser temperature of 800 °C. The gases from each trial were collected in a gas collection bottle and subsequently extracted through a septum seal for analysis. A range of compounds was expected, including light hydrocarbon gases, vapor from heavier hydrocarbons, and oxygen- and nitrogen-containing gases. No single analysis method was suitable for the whole of this range and so a range of methods was used. Immediate analysis for C1 C5 compounds was carried out using a Philips PU 4400-19 chromatograph with a flame ionization detector (GC/FID), using a wide bore (0.53 mm inside diameter) 50 m Restek RT-Alumina PLOT (porous layer open tubular) column with a 6 μm film thickness. This was calibrated by running a known mixture of C1 C4 gases before each analysis. A temperature program of 40 °C for 6 min, then a ramp at 5 °C/min to 200 °C and hold for 30 min, was used. At the same time samples of 20 30 μL were extracted from the gas collection bottle and passed through a thermal desorption tube for later analysis (TD/GC/MS). Tenax was used as an adsorbent, with an approximate volatility range from C6 to C26. Analysis was carried out using a Markes International Ultra TD/ Unity desorption unit, connected to an Agilent 6890N GC and Agilent 5973 MS with electron impact ionization. A 30 m 0.25 mm 0.25 μm HP5-MS GC column was used, with a temperature program ramping from 40 to 200 °C at 5 °C/minute then holding for 8 min. Direct injection of 2 μL samples of collected 9382
dx.doi.org/10.1021/es201664h |Environ. Sci. Technol. 2011, 45, 9380–9385
Environmental Science & Technology
ARTICLE
Table 1. Overall Percentage Weight Breakdown of Pyrolysis Products Batch 1 permanent gases methane
30.4
Batch 2
Batch 2
average
std dev
45.5
4.8
9.6
17.1
2.0
11.7
21.0
2.4
ethane
1.2
1.2
0.1
ethene
4.6
4.6
0.7
propene
0.1
0.2
0.04
C4 hydrocarbons
0.1
0.1
0.1
oxygen
1.1
0.6
0.2
carbon dioxide sulfur dioxide
1.4 0.5
0.7 0.02
0.3 0.02
carbon monoxide
propenenitrile oils + tars
0.1 46.1
0.02 27.8
0.03 4.7
benzene
23.7
13.3
4.3
toluene
5.3
2.6
1.5
styrene
1.0
0.7
0.6
chlorobenzenes
0.2
0.01
0.02
other < c8 naphthalene
0.2 3.8
0 4.9
0 2.3
benzo- and naphthalene carbonitrile
0.6
0.2
0.1
other 2 ring aromatic
1.4
1.4
1.0
3 ring aromatic
5.2
2.1
1.4
4 ring aromatic
3.2
1.9
1.2
5 ring aromatic
0.3
0.2
0.1
alcohols, aldehydes, esters
0.9
0.4
0.3
other c8+ residues
0.3
0.1
0.1
23.5
26.7
total potential fuel
73.4
72.0
1.0
gases into a GC/MS was used to give an indication of the levels of carbon dioxide, oxygen and nitrogen oxides. This used an Agilent 7890A GC with a 30 m 250 μm 0.25 μm HP5-MS column, connected to an Agilent 5975C MS system using electron impact ionization. Finally a Finnigan MAT 95 XP double-focusing mass spectrometer with electron impact ionization was used to quantify the amount of carbon monoxide and check for the presence of ammonia. A direct injection of a 10 μL sample from the gas collection bottle was made. The above two instruments were also used to analyze the oils, using both electron impact (EI) and chemical ionization (CI). In this case the GC temperature program comprised a hold at 30 °C for 1 min, then a ramp at 30 °C/min to 300 °C and a final hold for 5 min. The contents of the water flask were investigated in two ways. FTIR in reflectance mode as described above was used to look for the presence of CtN triple bonds indicating the presence of HCN. Samples dissolved in methanol were analyzed using a Waters ZQ4000 low resolution single quadrupole mass spectrometer with electrospray ionization, and spectra collected for both negative and positive ions. Neither of these methods indicated any significant dissolved compounds. The elemental breakdown of the residues collected from the pyrolyser and filter was obtained using EDX analysis, after the residues had been ground to a fine powder. As above, five scans were carried out for each sample from different areas of powder.
’ RESULTS AND DISCUSSION Gas and Oil Analysis. Analysis of the gases and oils was strongly interdependent since, as well as permanent gases, the collected gases contained vapor from the oils, the proportions depending on the volatility of individual components in the oils and the effectiveness of the condensing flasks. The composition of the oils was therefore obtained by adding the amounts of volatile compounds from the gas analysis to the results from the oil analysis. Quantification of oils and gases is given in Table 1. GC/FID analysis of the gases showed 99.5 molar % of the detectable gases to be methane, ethane, ethene, propene, benzene, toluene and styrene. Around 93 molar % of the compounds detected by TD/GC/MS analysis were benzene and toluene (percentages based on peak area). The remainder included linear, cyclic and aromatic hydrocarbons, alcohols, acids, aldehydes and nitriles, with little difference between the two batches. Both also indicated the presence of carbon dioxide and sulfur dioxide. Halogenated compounds found were small percentages of chloro- and dichloro-benzene. Direct injection GC/MS analysis of the gases showed approximately equal molar concentrations of oxygen and carbon dioxide but no nitrogen oxides. High resolution mass spectrometry showed the molar percentage of carbon monoxide to be 70% of that of methane. A check was also made for ammonia but none was found. The four sets of gas analysis data each provided relative percentages of the compounds identified by the corresponding method. In order to combine them to cover all the gases found, some common compounds were required. Benzene, toluene, and styrene were identified from the GC/FID, TD/GC/MS, and GC/MS results and were used to match these, whereas methane enabled the accurate mass data to be matched to the GC/FID results. The data were then converted to weight percentages to give an overall breakdown. The condensed “oils” produced by the pyrolysis process were yellow/brown in color and included both a fairly low viscosity liquid and sticky tar. Since the GC/FID and TD/GC/MS results had shown the presence of benzene and toluene, a solvent which would allow these to be identified was required for GC/MS analysis. Ethanol together with EI was used and approximate quantification carried out taking account of the lower solubility of some of the compounds in ethanol. Samples were also dissolved in toluene to provide further information on higher molecular weight compounds and analysis was carried out using both EI and CI. The range of compounds found was the same for both batches of material. Almost all were aromatic or polyaromatic hydrocarbons (PAHs), some with attached functional groups. The separate analyses using EI, positive ion CI and negative ion CI indicated the presence of different isomers and functional groups.21,22 High resolution MS was used to identify the nitrogen-containing functional groups as nitrile groups, which would be expected given the acrylonitrile component of the feedstock. No indication of nitro compounds was found. The composition of the oils was similar to coal tar or fuel oil. Residue Analysis. The pyrolyser and filter residues from each trial were analyzed separately using EDX. The levels of iron found in the residues were higher than could be explained by the content in the original polymers, which was thought to be due to some corrosion of the stainless steel pyrolyser and filter casing by the chlorine content of the pyrolysis gases. It was therefore 9383
dx.doi.org/10.1021/es201664h |Environ. Sci. Technol. 2011, 45, 9380–9385
Environmental Science & Technology assumed that the level of iron equaled that found in the ash from the relevant plastic batch. For each of the trials, the percentage of each element found in the combined residues was converted into a percentage of the original plastic feedstock weight. These figures are given in Figure 2 (note the different scales for Batch 1 and Batch 2) and compared with those from the ground polymers and ash for the main elements. The agreement was found to be quite good, with the closeness of the chlorine content from polymer to residues supporting the lack of chlorinated compounds in the oils and gases. The error bars show the standard errors for the results. The higher variability of the ground polymer results compared to the ash suggests that this was a less successful method of homogenization than the ashing. Some difference between the ground polymer and ash figures might also be expected as a result of the smaller sample of the ground polymer giving a less representative mix. Bromine was not detected in the residue and this is discussed below. Figure 2 shows that the higher level of chlorine in Batch 2 corresponded to a higher level of calcium. A sample of pyrolyser residue from Batch 2 was dissolved in water and the EDX analysis was then repeated. The original percentages (by weight of residue) of 11.4% chlorine and 6.7% calcium were reduced to 1.4% and 3.3% respectively, indicating that the chlorine in the residue was largely in the form of water-soluble calcium chloride. Overall Analysis. The weights of gases, oils and residue, and the composition of gases and oils, are summarized in Table 1. A small amount of heavy oils condensed in the system pipework before they reached the condensing flasks, estimated as 2% of the feedstock weight. The weight of gases (including volatiles) was calculated as the difference between the feedstock weight and the total weight of residues and condensed tars and oils. These weights were then converted to percentages of the original feedstock. The figure for potential fuel, at more than 70% of the original feedstock weight, is made up of the oils and tars, and the permanent gases other than oxygen, carbon dioxide, sulfur dioxide and nitriles. The residues are not included as fuel since these include a variety of metals and halogens which would require additional treatment. Batch 2 produced considerably more gas than Batch 1, a result which would correspond to the difference in plastic types. The aromatic rings in the structure of styrenic polymers are unlikely to break down at the temperatures considered, so contributing to the production of oils. The linear compounds within these polymers (the acrylonitrile backbone or butadiene), and similarly the linear backbone of PVC, may either form aromatic rings or break up into gases, depending on processing conditions. The increased amount of PVC in Batch 2 would have provided a greater chance of producing gases. The results for the three trials on Batch 2 were reasonably consistent. An approximate calculation of the amount of oxygen in the gases from Batch 1 indicated that there was around 6% (by weight of original feedstock) more than could be explained by the estimated amount in the polymers. This was most likely to have come from air ingress to, or residual air in, the experimental rig, despite the attempts to make the rig airtight and maintain positive pressure during the trials. The presence of oxygen might be expected to have led to water in the pyrolysis products. Identification or quantification of this was problematic due to the presence of water in the condensation system; however a proportion would be expected to condense along with the oils, which was not detected. The presence of oxygen but absence of
ARTICLE
water could have been a result of oxygen ingress occurring after the pyrolysis chamber. However, the possible presence of a small amount of water was allowed for in the LCA within the uncertainty ranges for quantity and calorific value of oils. In a fully sealed production process this possibility should be reduced and a lower proportion of carbon monoxide and other oxygencontaining compounds could be achieved. The proportion of gases from both feedstocks was significantly higher than had been found in most other work using comparable plastics. The closest was for fast pyrolysis of HIPS at temperatures above 800 °C,7 whereas WEEE-type mixtures at temperatures from 500 to 600 °C and slow temperature increases gave at most 12% gas. It therefore seems likely that the combination of fast pyrolysis and a higher temperature were the reason for the high gas proportion. The oils found in these trials showed some difference from other reported work, as might be expected given both the variety of mixed WEEE plastics and differences in process conditions. Although all trials with similar plastic mixes reported mostly aromatics, with some substituted nitrile and halogen groups, most of these were single ring and some also reported quite high levels of phenols. The trials described here showed PAHs with two or more rings to constitute 30 40% of the oils. Since these were not part of the structure of any of the original polymers, they must have been formed during the pyrolysis process as a result of the combination of temperature and residence time. Similarly it seems likely that the absence of phenols but significant level of carbon monoxide also reflected the process conditions. One of the key concerns in any thermal treatment of plastics containing chlorine or bromine is the creation of halogenated organic compounds and the potential for polyhalogenated dibenzodioxin creation, as well as the problem of corrosion caused by hydrogen chloride or hydrogen bromide in the system. In these trials the bulk of the chlorine measured in the feed material (mostly in the form of PVC plus a small amount included in flame retardants) remained in the pyrolysis residue, largely as calcium chloride. This appears to be a serendipitous result of calcium compounds being a common filler in PVC, and fits with results from other research.23,24 Some chlorine was also “trapped” by the steel of the filter housing even when stainless steel was used. Chlorinated aromatic compounds were found at a very low level in the gases from some trials, which would be expected since it was unlikely that the mechanisms just described would remove all of the chlorine. Only a very small amount of chlorine and no bromine were found in the water from the trials, suggesting that any HCl or HBr that was formed reacted with other elements or compounds in the pyrolyser or filter. The fate of the bromine detected in the original samples was not established as it could not be detected in any of the pyrolysis products, possibly because it was at too low a level (as was found for the ground polymer). Since the control of halogens is a significant factor in the use of WEEE plastics for pyrolysis, for commercial use a method would be required to ensure that halogens were removed from the gases and oils. Therefore for the LCA, the amount of bromine found in the polymer analysis was assumed to be present in the residue, while up to legal limits of polyhalogenated dibenzodioxins were allowed for in the gases and oils. Removal of halogens could involve the addition of an excess of a calcium containing substance such as lime to the feed, or the use of a catalyst after the pyrolysis chamber to remove halogens from organic compounds. 9384
dx.doi.org/10.1021/es201664h |Environ. Sci. Technol. 2011, 45, 9380–9385
Environmental Science & Technology These trials provided the bulk of data required for an LCA of the pyrolysis process. Results which were not established, such as the exact breakdown of the oils and the fate of bromine, were taken account of by uncertainty analysis in the LCA allowing for a realistic range of possible values.
’ ASSOCIATED CONTENT
bS
Supporting Information. Detailed breakdown of gas and oil analysis. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (44) 1792 602009; e-mail: [email protected].
’ ACKNOWLEDGMENT We thank ITEM Recycling Ltd for the design and building of the experimental rig, Axion Polymers for the supply of WEEE plastics, and the European Social Fund for financial assistance for this work. ’ REFERENCES (1) Waste Electrical and Electronical Equipment, Data 2008 (Tonnes, Updated 17 June 2011), WEEE Key Statistics and Data; Eurostat; European Commission: Brussels, 2011; http://epp.eurostat.ec.europa. eu/portal/page/portal/waste/data/wastestreams/weee. (2) Summary of the Impact Assessment on WEEE, SEC(2008) 2934; European Commission: Brussels, 2008; http://eur-lex.europa. eu/LexUriServ/LexUriServ.do?uri=SEC:2008:2934:FIN:EN:PDF. (3) Electronic Waste, Postnote number 291; UK Parliamentary Office of Science and Technology: London, UK, 2007; http://www.parliament.uk/documents/post/postpn291.pdf. (4) Dalrymple, I.; Wright N.; Kellner R.; Bains N.; Geraghty K.; Goosey M.; Lightfoot L. An integrated approach to electronic waste (WEEE) recycling. Circuit World. 2007, 33(2), 52; DOI: 10.1108/ 03056120710750256. (5) Develop a Process to Separate Brominated Flame Retardants from WEEE Polymers Final Report; PLA-037(2006); UK Waste & Resources Action Programme: Banbury, UK, 2007; http://www.wrap.org.uk/ recycling_industry/publications/develop_a_process__1.html. (6) Shun, D.; Dal-Hee, B.; Sung-Ho, C.; Keun-Hee, H. Bench-Scale Fluidized Bed Pyrolysis of Waste ABS Resin. In Proceedings of ISES 1997 Solar World Congress: Taejon, Korea, August 24 30, 1997. (7) Karaduman A.; S-ims-ek E. H.; C-ic-ek B.; Bilgesu A. Y. Flash pyrolysis of polystyrene wastes in a free-fall reactor under vacuum. J. Anal. Appl. Pyrol. 2001, 60, 179; DOI: 10.1016/S0165-2370(00) 00169-8. (8) Hall W. J.; Williams P. T. Pyrolysis of brominated feedstock plastic in a fluidised bed reactor. J. Anal. Appl. Pyrol. 2006, 77, 75; DOI: 10.1016/j.jaap.2006.01.006. (9) Hall W. J.; Williams P. T. Fast pyrolysis of halogenated plastics recovered from waste computers. Energ. Fuel. 2006, 20, 1536; DOI: 10.1021/ef060088n. (10) Day M.; Cooney J. D.; Touchette-Barrette C.; Sheehan S. E. Pyrolysis of mixed plastics used in the electronics industry. J. Anal. Appl. Pyrol. 1999, 52, 199 ;DOI: 10.1016/S0165-2370(99)00045-5. (11) Chiu S.; Chen S.; Tsai C. Effect of metal chlorides on thermal degradation of (waste) polycarbonate. Waste Manage. 2006, 26, 252; DOI: 10.1016/j.wasman.2005.03.003. (12) Bhaskar T.; Hall W. J.; Mitan N. M. M.; Muto A.; Williams P. T.; Sakata Y. Controlled pyrolysis of polyethylene/polypropylene/ polystyrene mixed plastics with high impact polystyrene containing
ARTICLE
flame retardant: Effect of decabromo diphenylethane (DDE). Polym. Degrad. Stab. 2007, 92, 211; DOI: 10.1016/j.polymdegradstab. 2006.11.011. (13) Brebu M.; Bhaskar T.; Murai K.; Muto A.; Sakata Y.; Uddin M. A. The individual and cumulative effect of brominated flame retardant and polyvinylchloride (PVC) on thermal degradation of acrylonitrile butadiene styrene (ABS) copolymer. Chemosphere. 2004, 56, 433; DOI 10.1016/j.chemosphere.2004.04.002 (14) Hernandez, MdR.; García, A. N.; Marcilla A. Study of the gases obtained in thermal and catalytic flash pyrolysis of HDPE in a fluidized bed reactor. J. Anal. Appl. Pyrol. 2005, 73, 314; DOI: 10.1016/j. jaap.2005.03.001. (15) Sodero S. F.; Berruti F.; Behie L. A. Ultrapyrolytic cracking of polyethylene—A high yield recycling method. Chem. Eng. Sci. 1996, 51, 2805; DOI: 10.1016/0009-2509(96)00156-X. (16) Alston S. A comparison of the environmental impact of pyrolysis and other treatments for waste electronic equipment and plastics. Ph.D. thesis, Swansea University, UK, 2009. (17) Hall W. J.; Miskolczi N.; Onwudili J.; Williams P. T. Thermal processing of toxic flame-retarded polymers using a waste fluidized catalytic cracker (FCC) catalyst. Energ. Fuel. 2008, 22, 1691; DOI: 10.1021/ef800043g. (18) Hall W. J.; Mitan N. M. M.; Bhaskar T.; Muto A.; Sakata Y.; Williams P. T. The co-pyrolysis of flame retarded high impact polystyrene and polyolefins. J. Anal. Appl. Pyrol. 2007, 80, 406; DOI: 10.1016/j.jaap.2007.05.002 (19) de Marco I.; Caballero B. M.; Chomon M. J.; Laresgoiti M. F.; Torres A.; Fernandez G.; Arnaiz S. Pyrolysis of electrical and electronic wastes. J. Anal. Appl. Pyrol. 2008, 82, 179; DOI: 10.1016/j.jaap. 2008.03.011. (20) U.K. Patent number GB2441721A; ITEM Technology Solutions Ltd (GB): Treorchy, UK. (21) Riahi K.; Sellier N. Separation of isomeric polycyclic aromatic hydrocarbons by GC-MS: Differentiation between isomers by positive chemical ionization with ammonia and dimethyl ether as reagent gases. Chromatographia. 1998, 47, 309; DOI: 10.1007/BF02466537 (22) Hilpert L. R.; Byrd G. D.; Vogt C. R. Selectivity of negative ion chemical ionization mass spectrometry for benzo[a]pyrene. Anal. Chem. 1984, 56, 1842; DOI: 10.1021/ac00275a019. (23) Hall W. J.; Williams P. T. Analysis of products from the pyrolysis of plastics recovered from the commercial scale recycling of waste electrical and electronic equipment. J. Anal. Appl. Pyrol. 2007, 79, 375; DOI: 10.1016/j.jaap.2006.10.006. (24) PVC Recovery Options—Concept for Environmental and Economic System Analysis; Kreissig J.; Baitz M.; Schmid J.; Kleine-M€ollhoff P.; Mersiowsky I.; Vinyl 2010: Brussels, 2003; http://www.pvc.org/MediaCentre/Documents-Library/PVC-Recovery-Options-Final-Report.
9385
dx.doi.org/10.1021/es201664h |Environ. Sci. Technol. 2011, 45, 9380–9385
ARTICLE pubs.acs.org/est
Environmental Impact of Pyrolysis of Mixed WEEE Plastics Part 2: Life Cycle Assessment Sue M. Alston* and J. Cris Arnold Materials Research Centre, Swansea University, Singleton Park, Swansea SA2 8PP, U.K.
bS Supporting Information ABSTRACT: Waste electrical and electronic equipment (WEEE) contains up to 25% plastics. Extraction of higher quality fractions for recycling leaves a mix of plastic types contaminated with other materials, requiring the least environmentally harmful disposal route. Data from trials of pyrolysis, described in part 1 of this paper set, were used in a life cycle assessment of the treatment of WEEE plastics. Various levels of recycling of the sorted fraction were considered, and pyrolysis was compared with incineration (with energy recovery) and landfill for disposal of the remainder. Increased recycling gave reduced environmental impact in almost all categories considered, although inefficient recycling decreased that benefit. Significant differences between pyrolysis, incineration and landfill were seen in climate change impacts, carbon sent to landfill, resources saved, and radiation. There was no overall “best” option. Landfill had the least short-term impact on climate change so could be a temporary means of sequestering carbon. Incineration left almost no carbon to landfill, but produced the most greenhouse gases. Pyrolysis or incineration saved most resources, with the balance depending on the source of electricity replaced by incineration. Pyrolysis emerged as a strong compromise candidate since the gases and oils produced could be used as fuels and so provided significant resource saving without high impact on climate change or landfill space.
’ INTRODUCTION The increasing growth in waste electrical and electronic equipment (WEEE),1 together with rising recycling and recovery targets, means that recovery of the metal fraction of WEEE is no longer sufficient and there is a need to also make use of the plastic fraction. Some plastic parts can easily be extracted from WEEE during dismantling, and can then be mechanically recycled. The remainder of the equipment is shredded and sorted. A proportion of the sorted plastic is of sufficiently high quality for mechanical recycling but the rest must be disposed of in some other way. Currently the most likely disposal options are landfill or incineration, with incineration plants incorporating energy recovery through generation of electricity and possibly heat. A possible alternative is pyrolysis, where the waste is heated without oxygen so that it breaks down into oils and gases, leaving a residue containing carbon and inorganic content from the original waste. Previous work has shown that more than 70% of the feedstock can be converted to oils and gases suitable for use as fuels,2 and the volume of remaining residue is considerably less than that of the original waste. Decisions on waste treatment tend to be based on legal requirements and economic factors; however the policy driving these should be guided by knowledge about the environmental impact of the various options. Although an old technology in principle, the commercial use of pyrolysis for waste treatment is still at the developmental stage. The economics of the process are r 2011 American Chemical Society
affected by external factors such as changing prices for alternative waste treatment methods and energy sources, but can be improved by optimization of the technology. Consequently each proposed process is different and produces different outputs. Knowledge of the environmental impact compared to other waste treatments could also affect the drive to improve performance. With these aims in mind, the outputs measured during work on a pilot plant developed and patented by ITEM Technology Solutions Ltd.,3 described in part 1 of this paper set,2 are assessed here via life cycle assessment (LCA) studies. LCA studies of pyrolysis of plastics, and in particular WEEE plastics, are limited and have studied rather different pyrolysis processes. A study by the UK Waste and Resources Action Programme 4 on treatment of mixed waste plastics included two types of pyrolysis, one for cracking polyolefins and the other producing diesel. In comparison to recycling and incineration alternatives, these scored highly for primary energy consumption, human toxicity, and ozone depletion impacts but poorly for global warming potential, solid waste, and abiotic depletion. Huisman et al.5 found that, for plastic dominated WEEE and using an aggregate score, it was worthwhile to recycle the plastic housing rather than incinerate it. However another study6 found Received: May 16, 2011 Accepted: September 22, 2011 Revised: September 21, 2011 Published: September 22, 2011 9386
dx.doi.org/10.1021/es2016654 | Environ. Sci. Technol. 2011, 45, 9386–9392
Environmental Science & Technology
Figure 1. Process flow for WEEE treatment.
that mechanical recycling could be less beneficial than other methods if the resulting recycled material was of low quality. Current WEEE plastics can contain brominated flame retardants (BFRs) which limits their direct recycling since many of these are now banned in the EU. Freegard et al.7 identified that the “Creasolv” solvent extraction method of recovering BFR-free plastics was potentially commercially viable in the UK at a throughput of 10 000 tonne/year. This combined with mechanical recycling of the resulting BFR-free plastic was found to have a significant net environmental benefit compared to landfill and incineration with energy recovery. A study of plastic packaging waste 8 suggested that for this particular mix, feedstock recycling through pyrolysis could give a comparable greenhouse effect benefit to a high level of mechanical recycling. Two studies9,10 on municipal solid waste also showed potential advantages to pyrolysis. In one the replacement of 55% recycling, 45% landfill with 17% recycling, 83% micropyrolysis gave significant extra benefit in respiratory inorganics and fossil fuels and small benefits in other categories. Azapagic10 compared large scale incineration with small scale pyrolysis and gasification, both with energy recovery. With a functional unit of amount of waste treated, incineration was slightly more beneficial in all categories except global warming where it was significantly worse. However if the functional unit was the amount of electricity produced, pyrolysis was more beneficial in all categories. The specific example of a small-scale pyrolysis process used to treat WEEE plastics is compared here with the alternatives of more or less recycling, incineration with energy recovery, and landfill.
’ MATERIALS AND METHODS Goal and Functional Unit. The goal was defined as “To compare the environmental impact of pyrolysis as a means of
ARTICLE
disposal of WEEE plastics with the alternatives of mechanical recycling, incineration and landfill.” The WEEE entering the system consisted of a full range of material including metals, plastics, glass, wood, batteries, capacitors, etc. However the aim of this study was to look specifically at treatment of plastics. The functional unit in this case was therefore taken to be the weight of WEEE containing 1 kg of plastic. Plastics were divided into three categories: a light fraction (less dense than water, typically polypropylene, suitable for mechanical recycling), a medium fraction (mainly styrenic polymers suitable for mechanical recycling) and a remaining heavy fraction containing plastics with high levels of additives, dense polymers found in small quantities, and contaminants. System Boundaries. The start point of the study was taken to be a designated WEEE collection facility (typically a civic amenity site). The impact of all waste treatment and disposal activities was included, through to final disposal of fractions that could not be recovered. The effect of recovery of energy and plastics was taken account of by including the impacts avoided as a result of not having to produce these from other sources. The infrastructure required for the various activities was included to provide consistency with the background data used. Process Flow. The process flow for the WEEE treatment route is shown in Figure 1. A certain level of “de-pollution” is required by the EU WEEE regulations, for example, removal of batteries, LCD backlights etc. This leads to a minimum level of dismantling which must always be carried out, and will lead to a certain amount of recovered metals and plastics. The depolluted WEEE can then be shredded and metals removed. The residues, mostly plastics with some contaminants, can either be sent to a plastics recycling facility or for pyrolysis, incineration or landfill. Rather than simply depolluting the WEEE, a greater level of dismantling can be carried out with more segregated parts removed and sent directly for recycling. The remainder can then be shredded, after which it is assumed it will go to a recycling facility. After removal of metals, the remainder is sorted, recyclable fractions are washed, dried, and extruded to pellets, and the remainder goes for pyrolysis, incineration or landfill. Analysis was carried out using Simapro (version 7) software from Pre Consultants. Allocation. The separation and recycling of metals was not included within the system boundary since the environmental benefits of these processes are very large and would have swamped the impacts of the treatment of plastics which was the area of interest. For example recycling of 1 kg of aluminum typically saves 5 10 kg CO2-eq of greenhouse gases compared to 1.5 2 kg for 1 kg of plastics.11 This meant that the environmental burdens associated with separating metal for recycling, that is, dismantling, shredding, and metal separation, needed to be excluded. Allocation of burdens was made using material mass. Inventory Data. The “ecoinvent” database from the Swiss Centre for Life Cycle Inventories was used for background data, including transport, energy sources, and production of virgin materials. Data applicable to the UK were used where available, and where this was not the case either European or Swiss data were used. Ecoinvent data for incineration and landfill were also used, with long-term being taken as infinite time. Typical WEEE composition was taken to be 45% metal, 23% plastic, 20% CRT components, 3.9% glass, 3.9% wood/fiber, 2% circuit boards and cables, 1.6% copper fines and 0.6% batteries.12 The CRT components, circuit boards and cables were assumed 9387
dx.doi.org/10.1021/es2016654 |Environ. Sci. Technol. 2011, 45, 9386–9392
Environmental Science & Technology
ARTICLE
Table 1. Environmental Impact Categories Used category
characterization unit
from method
normalization quantity
weighting factor
Human Health climate change
DALY
EI 99
2.382 10
3
0.4
ozone layer depletion
DALY
EI 99
2.193 10
4
0.1049
radiation
DALY
EI 99
2.99 10
respiratory organics
DALY
EI 99
6.978 10
respiratory inorganics
DALY
EI 99
0.01067
carcinogens
DALY
EI 99
1.994 10
human toxicity
kg 1,4-DB eq
CML
19936
0.0048
freshwater aquatic ecotoxicity
kg 1,4-DB eq
CML
1329
0.0048
marine aquatic ecotoxicity
kg 1,4-DB eq
CML
298686
0.0048
terrestrial ecotoxicity
kg 1,4-DB eq
CML
124.1
0.0048
acidification
kg SO2 eq
CML
71.89
0.0727
eutrophication
kg PO4 eq
CML
32.81
0.0557
land use
PDF m2yr
EI 99
3944
0.0182
carbon deposit
kg C
ES 06
116
0.0364
kg Sb eq
CML
39.05
0.0819
5
0.0727 5
0.0606
3
0.0048
0.0727
Ecosystem Quality
Resources abiotic depletion
to be removed during the depollution stage. The proportion of each material which could be reused or removed in whole parts for recycling was based on the transfer coefficients for manual treatment in the ecoinvent processes. The plastic was estimated to be made up of a 10% light fraction, a 60% medium fraction and a remaining 30% heavy fraction, based on figures from Freegard et al.7 The elemental composition of the heavy fraction was obtained from measurements carried out on samples obtained from Axion Polymers.13 These were combined with ecoinvent data to estimate the composition of the light and medium fractions. Most data for the operation of mechanical recycling processes were obtained from a range of manufacturers and an average taken. These included power, weight, size, air, water, salts (for density sorting), chemicals and emissions. Ecoinvent modules were used to estimate emissions from shredding and metal separation processes, and as a basis for extrusion of recycled plastic, assuming that the process would suffer 2.5% losses and that, in the case of ABS and HIPS, 0.5% of emissions would escape to the atmosphere. Plant infrastructure was taken as a proportion of the ecoinvent infrastructure module for a mechanical treatment plant based on equipment weight. The proportion of a particular fraction (light or medium) recovered by a sorting process was set to 77%. Data from the most effective process for removal of BFRs, “Creasolv”, was not publicly available, so data for steam and toluene used in the “Centrevap” process were used as representative.7 Most of the data for material inputs to and outputs from the pyrolysis process were obtained from experimental work on a pilot rig.2,13 Of the gases produced, 0.1% was assumed to escape and be emitted to air; the remainder substituted for fuels. Operational data for feeding and pyrolyzing the feedstock were obtained from design calculations for a production facility, including heat input of 6.8 MJ per kg feedstock.
Transport distances between a WEEE treatment center, a plastic recycling plant, and pyrolysis, incineration and landfill facilities were obtained from published data4 or estimated. It was estimated that 10% of material was transported by train, with wagons fully loaded and no return cost. The remaining transport was by lorry, fully loaded but with an empty return journey. Avoided Impacts. Recycling of plastics was considered to avoid the use of virgin polymer with a “substitution factor” λ, that is, One kg of recycled plastic replaced λ (<1) kg of virgin plastic, since it could potentially have worse properties. The nearest producers of ABS and HIPS are in mainland Europe rather than the UK, so it was assumed that transport would be avoided by UK-based recycling plants supplying UK-based parts manufacturers. Calculation of avoided burdens due to the production of fuel by the pyrolysis process was carried out on the basis of replacing fuel of equivalent calorific value, using lower heating value (LHV). LHV values and the measured weight fraction for each component of the permanent gases were used to produce a combined figure, and the gases then assumed to replace natural gas with calorific value equal to this. The oils produced were split into three parts. The first comprised the measured weight fractions of benzene, toluene, and styrene, with an LHV of 40 MJ/kg. The second, the heaviest 10% of the oils, were suitable to replace heavy fuel oil (HFO) with an LHV of 41.1 MJ/kg. The third, remaining, part was suitable to replace light fuel oil (LFO) with an LHV of 40.1 MJ/kg. The calculated LHV figures were then reduced to allow for up to 50% of the oils to have attached nitrile and/or hydroxyl functional groups. Uncertainties in the exact composition of the oils were taken account of in the LCA sensitivity analysis, although the effect of this uncertainty on calorific value was small. The total heat content of the saved gases and oils was calculated as 24.5, 27.0, and 41.1 MJ per kg feedstock for the three plastic fractions: heavy, medium, and light, respectively. If only the heavy plastics fraction was pyrolyzed, the heat 9388
dx.doi.org/10.1021/es2016654 |Environ. Sci. Technol. 2011, 45, 9386–9392
Environmental Science & Technology
ARTICLE
Figure 2. Significant impacts for treatment scenarios with varying recycling levels and quality.
required as input to the process would be approximately 25% of that produced. Incineration provided electricity at 11% and heat at 23.9% of the LHV of “plastics less water content”, calculated from suggested average figures for MSW from ecoinvent. No compensation was made for energy generation from landfill gas since the amount of methane produced by plastics is very small. Impact Assessment. Characterization factors extracted from three established methods were used: Eco-Indicator ’99 (EI 99), CML 2001 (CML), and Ecological Scarcity 2006 (ES 06). Categories were chosen to be as comprehensive as possible without duplication of effects, and a long-term approach was taken. In order to assess the relative significance of categories, normalization was carried out using the annual impact of one Western European in each category. Weighting factors were derived from the Ecological Scarcity “distance to target” approach. This resulted in the final selection of categories and factors shown in Table 1.13 The normalized and weighted results were used to identify the most significant categories for further investigation. However comparisons between treatment processes were assessed within categories only and not amalgamated. Sensitivity Analysis. The majority of sensitivity analysis was carried out using a “Monte Carlo” method to translate probability distributions for the input data into corresponding distributions for the resulting impacts. The main areas of uncertainty in the modeling of the pyrolysis, incineration and landfill processes were the chemical composition of the plastics, and the details of the pyrolysis process. Variation in the pyrolysis data was taken account of for the heat and electricity required to run the process, the percentage of gas which escaped to air, SOx emissions, NOx emissions, the LHV of the oils and the proportions of gases and oils produced. Also the possibility of dioxin emissions up to the maximum legally allowed was included, since the study of the pyrolysis process had been unable to identify the fate of bromine in the feedstock, and small differences in the tracking of chlorine could be significant in this
context. The final results also included uncertainties in the incoming material mix and the inputs to the recycling processes.
’ RESULTS AND DISCUSSION Key Impacts. Initial analysis was carried out on each individual disposal process, that is, pyrolysis, incineration with energy recovery, and landfill, for each of the plastic fractions, that is, light, medium, and heavy. The categories were compared to identify the most significant impacts. After normalization, these were marine and freshwater aquatic toxicity, carcinogens, and carbon deposit. The effect of weighting was substantial and added the categories of climate change, abiotic depletion, eutrophication and radiation. This highlighted the difference that can be made by using a target driven approach rather than one related to current emissions. A process that has a relatively small impact compared to the current situation can none the less be significant in comparison to a target which requires large reductions in emissions. These eight categories were included in further investigations. Recycling. Three recycling scenarios were considered. In “most” recycling, all WEEE was dismantled as far as possible, all shredded material was sorted for recycling, and all plastics with brominated flame retardants (BFRs) were recovered using solvent processes. Sorting was carried out using an identification process. This is not yet a realistic option (not all processes are commercially available yet) but was used to represent the “best” case. In “least” recycling, no dismantling was carried out, and plastic parts removed during depollution went for disposal rather than recycling. After depollution, all WEEE was shredded and the nonmetal fraction went for disposal. “Typical” recycling fell between the two, with 50% of WEEE being dismantled, while of the remainder 50% of the plastic recovered during depollution was sent for recycling. Of the shredded plastics, 50% were sorted for recycling while the rest went for disposal. These percentages were intended to represent averages across the industry; individual plants would be likely to be closer to either “most” or “least”. BFR plastics were not recovered, and density sorting was used. 9389
dx.doi.org/10.1021/es2016654 |Environ. Sci. Technol. 2011, 45, 9386–9392
Environmental Science & Technology
ARTICLE
Figure 3. Breakdown of significant impacts by process for three disposal methods, “typical” recycling scenario.
Figure 2 shows the mean impacts from Monte Carlo analysis for the three scenarios, with material for disposal split equally between pyrolysis, incineration and landfill. The electricity produced from incineration was assumed to replace that from the standard UK mix. Figure 2 also includes the effect of sorting efficiency and quality of recycled material (represented by the substitution rate for virgin polymer) on the impact of the “most” recycling scenario. “Average quality” used the inventory data given above. “Low quality” reduced both the efficiency and the substitution factor to 0.7, while “high quality” increased both to 1.0. Positive figures show an environmental burden, while negative figures show a benefit, so that the best case scenario would show negative bars. In almost all categories the impact reduced as the amount of recycling increased. The exception was the radiation impact, which was driven by the use of nuclear generated electricity. Since recycling used electricity, this impact was smaller in the least recycling scenario, and in addition the fraction of disposal by incineration produced electricity which further reduced the impact. However these effects were small compared to the benefits of recycling in other categories. It can also be seen that the benefits of more recycling can be significantly reduced if that recycling is poor quality. Comparison of Disposal Methods. The next study considered the effects of the residual disposal method for the “typical” recycling scenario. The impacts on carcinogens, freshwater aquatic ecotoxicity and marine aquatic ecotoxicity were found to be affected very little by the disposal process. These were driven by the long-term leaching of elements in the plastic additives from landfill. Analysis of the feedstock, ash from combustion of the feedstock, and the pyrolysis residues all showed similar quantities of these elements,2 indicating that they remain in incinerator ash and pyrolysis residue which goes to landfill. The fate of chlorine is of particular interest, with the pyrolysis study showing that the bulk of this was found in the residue, mostly as calcium chloride. In the case of disposal by landfill or incineration the chlorine also ends up in landfill either directly, or indirectly in fly ash, suggesting little difference between the impacts of the three processes. In fact the ecoinvent landfill model and the impact methods considered indicate that
chlorine, if not already in the form of chlorides, would react with other material to become chlorides before any leaching took place and that these would not have an adverse environmental impact. A possible exception might be for chlorinated organic compounds. A pyrolysis process in normal operation would be designed to avoid these in the emissions or residue. However the potential for polyhalogenated dibenzodioxins was included up to the legal maximum for emissions to air and within the residue. These were found to contribute at most 1% to the carcinogen and ecotoxicity categories. Eutrophication and carbon deposit showed similar patterns since both were driven by the amount of carbon ending up in landfill. The inclusion of chemical oxygen demand from carbon in the eutrophication category is part of the CML method, and is not included in the EcoIndicator 99 method. A check was made on the effect of removing this. The resulting impact became very small and continued to be highest for landfill. The significant categories in comparing disposal methods were therefore climate change, radiation, carbon deposit and abiotic depletion. The stacked bars in Figure 3 show these broken down by process for each disposal method, for the “typical” recycling scenario. The single bars show the total for all processes, normalized within the category. The error bars on these are the standard deviation of the Monte Carlo distribution. The climate change impact of the recycling processes, that is, shredding and polymer processing, was lower than the benefits of the corresponding saved materials giving a net advantage for the material recycled. Disposal of the remainder to landfill had little effect on this, giving an overall climate change benefit. Pyrolysis reduced the overall climate change benefit because the effect of the energy inputs was greater than that of the saved fuels. Incineration had the greatest impact since the process is a high carbon dioxide emitter and this was not outweighed by the benefit of the saved electricity. The radiation impact was driven by nuclear electricity generation in the UK mix. In the case of incineration, electricity is generated without any radiation. When this replaces part of the UK mix generation, there will be a reduced radiation impact. The net burden from the recycling processes reflected the fact that, 9390
dx.doi.org/10.1021/es2016654 |Environ. Sci. Technol. 2011, 45, 9386–9392
Environmental Science & Technology
ARTICLE
Table 2. Comparison of Disposal Processes for Main Impact Categories for “Least” Recycling Level impact category
incineration
pyrolysis
landfill
climate change
short-term
240% to 400%
100%
40%
climate change radiation
long-term
170% to 280% 220% to 35%
100% 100%
130% 46%
14%
100%
350%
40% to 170%
100%
20%
carbon deposit abiotic depletion
while recycling was assumed to take place in the UK, the data for the virgin plastics which were saved had a low radiation impact. Carbon deposit was least for incineration where almost all carbon was converted to carbon dioxide, slightly higher for pyrolysis since the residue contained a proportion of carbon, and most when the nonrecycled plastics were landfilled directly. The recycling led to a negative impact for abiotic depletion for all disposal routes, and pyrolysis and incineration provided an additional benefit. Sensitivity of Comparison of Disposal Methods. The error bars in Figure 3 give an indication of the robustness of conclusions developed regarding the difference between disposal methods. In the cases of climate change and radiation, there were significant overlaps between the distributions for pyrolysis and landfill. The carbon deposit distributions were quite distinct. Pyrolysis and incineration gave similar distributions for abiotic depletion with landfill clearly separated. As the amount of recycling was reduced, the distributions of the climate change, radiation and carbon deposit impacts all became well spaced, indicating that the difference between disposal methods was significant despite uncertainties in data. For abiotic depletion, the distributions for pyrolysis and incineration still had considerable overlap suggesting that the impacts would be comparable. With increased recycling there was considerably more overlap since the disposal options contributed to less of the overall impact. In this case the effects of recycling efficiency (sorting efficiency, substitution factor and recyclability of plastic mix) were found to be more significant than differences between disposal processes. Alternative Scenarios. Two alternative scenarios were also assessed varying the avoided energy from incineration. These were (1) Electricity replacing that from gas, and no heat recovery. (2) Electricity replacing that from coal, with heat recovery. These were found to have a significant effect on the results for radiation and abiotic depletion. The radiation saving from replacing the UK electricity mix was lost and the impact for incineration became similar to that for landfill. In the case of abiotic depletion, the impacts of pyrolysis and incineration were no longer so similar. The original scenario had a mean impact of 14 mpts if the disposal method was pyrolysis and 16 mpts if it was incineration. For (1) the impact of incineration became less beneficial at 6 mpts (since it was now replacing a less detrimental option and not saving heat), while for (2) it became more beneficial at 24 mpts. The models used in ecoinvent for landfill assume that all carbon which is not released during short-term decomposition will end up in long-term groundwater causing oxygen depletion. No account is taken of the fate of the carbon once it reaches the sea. However it is likely that some percentage of it is released as carbon dioxide to the atmosphere. An alternative analysis was therefore carried out using a proportion of 50% to give an idea of
the effect. No change was made to the oxygen depletion effect since the conversion of carbon to carbon dioxide would still use oxygen. This only affected the climate change impact, for which it was sufficient to reverse the relative impacts of landfill and pyrolysis. Summary Results. Table 2 summarizes these results showing the differences between the three disposal methods when recycling levels were low. The figures are given in relation to pyrolysis as 100% and the range for incineration covers the three possible options for electricity substitution. From this it is clear that the decision as to which disposal method to use depends on which impacts are given priority. Disposal to landfill had the least impact on climate change in the short term (100 years). Its main impact was in terms of requirement for landfill space with few other detrimental effects. It could therefore be considered as a temporary means of sequestering carbon. Incineration was the most effective method of reducing carbon deposits, that is, effectively reducing landfill space, but had the worst climate change impact. It could also be most effective in reducing radiation if the electricity it generated replaced nuclear power. Resource saving would be best achieved through pyrolysis or incineration. The balance would depend on the source of electricity being replaced by incineration. Decisions on which process to use would therefore be heavily influenced by policy requirements. However pyrolysis emerged as a strong compromise candidate, since it avoided the extremes of greenhouse gas emissions or requirement for landfill space, while giving significantly greater resource savings than landfill and a comparable level to those achieved by incineration.
’ ASSOCIATED CONTENT
bS
Supporting Information. Data for material composition, process and other inputs; calculation of avoided burdens; graphical examples of key impacts, comparison of disposal methods and alternative scenarios. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (44) 1792 602009; e-mail: [email protected].
’ ACKNOWLEDGMENT We thank the European Social Fund for financial assistance for this work. ’ REFERENCES (1) 2008 Review of Directive 2002/96 on Waste Electrical and Electronic Equipment (WEEE), Study No. 07010401/2006/442493/ETU/G4; United Nations University: Bonn, Germany, 2007; www.ec.europa.eu/ environment/waste/weee/pdf/final_rep_unu.pdf. (2) Alston, S.; Clark, A. D.; Arnold, J. C.; Stein, B. K. Environmental impact of pyrolysis of mixed WEEE plastics part 1: Experimental pyrolysis data Environ. Sci. Technol. 2011, DOI: 10.1021.es201664h. (3) UK Patent number GB2441721A. ITEM Technology Solutions Ltd (GB): Treorchy, UK. (4) LCA of Management Options for Mixed Waste Plastics, Project MDP017; The Waste & Resources Action Programme: Banbury, Oxon, 2008; www.wrap.org.uk/downloads/LCA_of_Management_Options_ for_Mixed_Waste_Plastics.4980bb9f.5497.pdf. 9391
dx.doi.org/10.1021/es2016654 |Environ. Sci. Technol. 2011, 45, 9386–9392
Environmental Science & Technology
ARTICLE
(5) Huisman, J.; Stevels, A. L. N.; Stobbe, I. Eco-efficiency considerations on the end-of-life of consumer electronic products. IEEE Trans. Electron. Packag. Manuf. 2004, 7 (1), 9-25; DOI: 10.1109/ TEPM.2004.832214. (6) Final Report Eco-Efficiency of Electrical and Electronic Equipment (WEEE) End-of-Life-Options, Report 3/2005; Plastics Europe Deutschland: Frankfurt, 2005; www.plasticseurope.org/Document/final-reporteco-efficiency-of-electrical-and-electronic-equipment-weee-end-of-lifeoptions.aspx. (7) Develop a Process to Separate Brominated Flame Retardants from WEEE Polymers—Final Report, Project PLA-037; The Waste & Resources Action Programme: Banbury, Oxon, 2006; www.wrap.org.uk/ downloads/BrominatedWithAppendices.29705853.fdea5b6a.3712.pdf. (8) Comparison of Feedstock Recycling and Alternative Treatment Methods for Household Plastic Waste, Report OR 04.03; Østfold Research Foundation: Krakerøy, Norway, 2003; www.ostfoldforskning.no/uploads/dokumenter/publikasjoner/123.pdf. (9) di Maria, F.; Fantozzi, F. Life cycle assessment of waste to energy micro-pyrolysis system: Case study for an Italian town. Int. J. Energy Res. 2004, 28 (5), 449-461; DOI 10.1002/er.977 (10) Azapagic, A. Energy from Municipal Solid Waste: Large-scale incineration or small-scale pyrolysis? Environ. Eng. Manage. J. 2007, 6 (5), 337–346. (11) Environmental Benefits of Recycling—An International Review of Life Cycle Comparisons for Key Materials in the UK Recycling Sector; The Waste & Resources Action Programme: Banbury, Oxon, 2006; http://www.wrap.org.uk/downloads/Recycling_LCA_Report_Sept_ 2006_-_Final.f1a3d917.2838.pdf. (12) An Integrated Approach to WEEE Recycling, Dissemination presentation; WEEE-Tech Project: UK, 2006. (13) Alston, S. A comparison of the environmental impact of pyrolysis and other treatments for waste electronic equipment and plastics. Ph.D. thesis, Swansea University, UK, 2009.
9392
dx.doi.org/10.1021/es2016654 |Environ. Sci. Technol. 2011, 45, 9386–9392
ARTICLE pubs.acs.org/est
Life Cycle Greenhouse Gas Emissions of Current Oil Sands Technologies: GHOST Model Development and Illustrative Application Alex D. Charpentier,† Oyeshola Kofoworola,† Joule A. Bergerson,‡ and Heather L. MacLean*,†,§ †
Department of Civil Engineering, University of Toronto, 35 St. George Street, Toronto, Ontario, Canada M5S 1A4 ISEEE Energy and Environmental Systems Group, Center for Environmental Engineering Research and Education, Department of Chemical and Petroleum Engineering, University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada T2N 1N4 § Department of Chemical Engineering and Applied Chemistry, School of Public Policy and Governance, University of Toronto, Toronto, Ontario, Canada M5S 1A4 ‡
bS Supporting Information ABSTRACT: A life cycle-based model, GHOST (GreenHouse gas emissions of current Oil Sands Technologies), which quantifies emissions associated with production of diluted bitumen and synthetic crude oil (SCO) is developed. GHOST has the potential to analyze a large set of process configurations, is based on confidential oil sands project operating data, and reports ranges of resulting emissions, improvements over prior studies, which primarily included a limited set of indirect activities, utilized theoretical design data, and reported point estimates. GHOST is demonstrated through application to a major oil sands process, steam-assisted gravity drainage (SAGD). The variability in potential performance of SAGD technologies results in wide ranges of “well-to-refinery entrance gate” emissions (comprising direct and indirect emissions): 1841 g CO2eq/MJ SCO, 918 g CO2eq/MJ dilbit, and 1324 g CO2eq/MJ synbit. The primary contributor to SAGD’s emissions is the combustion of natural gas to produce process steam, making a project’s steam-to-oil ratio the most critical parameter in determining GHG performance. The demonstration (a) illustrates that a broad range of technology options, operating conditions, and resulting emissions exist among current oil sands operations, even when considering a single extraction technology, and (b) provides guidance about the feasibility of lowering SAGD project emissions.
’ INTRODUCTION Alberta’s oil sands, an amalgamation of bitumen, sand, clay, and water, are the third largest oil reserves in the world (175.2 billion bbls) behind Saudi Arabia and Venezuela.1 Bitumen is a heavy, highly viscous form of petroleum that does not flow at reservoir conditions. Its recovery, extraction, upgrading, and refining to products such as gasoline require largely fossil-energy inputs, leading to emissions of greenhouse gases (GHG) and other negative environmental impacts. Recent legislation in Alberta has set GHG intensity reduction targets that apply to the oil sands industry.2 Further, several jurisdictions have or are considering following California in enacting low carbon fuel standards,3 which require a reduction in the average carbon intensity, on a life cycle (LC) basis, of transportation fuels sold in the jurisdiction. These regulations require accurate assessments of the LC emissions of oil sands-derived fuels to define carbon intensity values for these fuel production pathways.4 Life cycle assessments (LCAs) are critical for targeting and benchmarking improvements in process and supply chain performance, and r 2011 American Chemical Society
comparing the performance of existing and emerging technologies (see ref 5 for information on LCA). Bitumen is produced through two techniques: surface mining and in situ recovery. Surface mining accesses shallow reserves (generally less than 65 m to the top of the oil sands zone6) through open-pit mines, where oil sands are recovered and then transported to an extraction facility for separation of the bitumen using hot water, and finally diluted for transport to an upgrader. In situ production from underground reservoirs can use cold or thermal technologies. The latter involve steam injection into the reservoir to reduce the viscosity of bitumen, extraction of the bitumen underground (in situ), and conveyance of the bitumen to the surface through wells. For additional information on production technologies see ref 7. In 2009, all surface mined Received: November 21, 2010 Accepted: September 15, 2011 Revised: July 13, 2011 Published: September 15, 2011 9393
dx.doi.org/10.1021/es103912m | Environ. Sci. Technol. 2011, 45, 9393–9404
Environmental Science & Technology bitumen and a small portion of in situ production were upgraded to synthetic crude oil (SCO) and shipped to a refinery, while the remaining in situ produced bitumen was blended with a diluent and shipped directly to a refinery able to process diluted bitumen.8 The above activities are shown in the context of the “well-towheel” stages associated with production of a final product (e.g., gasoline) from bitumen and its combustion in a vehicle in Figure S1 in Supporting Information. Up to year-end 2009, the oil sands industry had produced 7 billion bbl of crude bitumen; 65% using surface mining and 35% using in situ techniques.8 However, ref 6 estimates just 20% of reserves are surface mineable whereas the remainder must be recovered through in situ techniques. One source forecasts the production of both nonupgraded bitumen and SCO to more than double in the next decade,9 however, forecasts vary and often conflict. In our review of 13 studies (published up to 2008) reporting GHG emissions associated with oil sands production we found inconsistencies in the emissions intensities reported, both for oil sands and conventional crude oil pathways, and called for further research with transparent study boundaries and consistent assumptions, and a need for improved data in the public realm.10 Two more recent studies11,12 focus on emissions associated with the production of transportation fuels in the U.S. from local and imported conventional crudes. In contrast to most studies, refs 11 and 12, which are studies that were contracted by the Alberta Government, present widely varying emissions for the production of gasoline from different conventional crude oils. The studies also assessed emissions of several oil sands pathways using publicly available data for extraction and production activities, and proprietary refining models for that aspect of the LC. Whereas the results of a couple of other studies (e.g., 13, 14) published since 10 add to the emissions data set, they do not address two of the limitations discussed in 10: reliance on publicly available oil sands data (which have been limited and often of low quality as discussed in the next paragraph) and reporting results as point estimates. Few LCA software models include oil sands production activities. The U.S. Department of Energy and Natural Resources Canada have spreadsheet models that do so 15,16 (see 10 for a review of these models). LCA software (e.g., 17,18 may be utilized as platforms to examine oil sands but the analyst would have to develop modules and provide process, parameter, and input data for any oil sands-specific activities as the software do not include these components. Studies of oil sands GHG emissions have been based primarily on publicly available data as a result of the challenges of obtaining actual operating data due to their proprietary nature. However, care must be taken in interpreting publicly available data as some are not representative of actual operations. For example, project performance data published in Environmental Impact Assessments are generally based on projections and/or design stage modeling, and 19 reported that actual cumulative steam-to-oil ratios (SOR refers to the cold water equivalent volume of steam required to produce one volume unit of oil and is a measure of the efficiency of oil production) were up to four times greater than projected targets for thermal in situ recovery operations in Alberta. Moreover, the majority of studies have reported emissions as point estimates for each of mining and in situ production, which is not sufficient as each project is unique (e.g., different reservoirs require different process techniques and produce bitumen and SCO with distinct characteristics). To more accurately depict the oil sands industry’s emissions
ARTICLE
performance, a range of LC results from a representative set of projects based on actual operating data (and a consideration of the range of potential performance using existing technologies) is needed. The objective of this research is to provide more detailed estimates of the GHG emissions associated with current major oil sands production pathways. To attain this objective, a LC model (GreenHouse gas emissions of currently operating Oil Sands Technologies—GHOST) capable of assessing the variability in performance of recovery, extraction, and upgrading technologies is developed. GHOST’s structure and input data are informed by technical experts, confidential operating data collected from the industry, and, when necessary, the confidential data augmented with publicly available data whose source and reliability were verified. As with any LC-based tool, GHOST is most useful for comparing the relative magnitude of emissions rather than absolute values. This paper (1) documents the development of GHOST including model structure and data collection; (2) provides an illustrative application of GHOST to a major in situ extraction technology, steam-assisted gravity drainage (SAGD) with and without upgrading; and, (3) identifies GHG-intensive processes and the sensitivity of emissions to input parameters through scenario and sensitivity analyses. The research will be extended in ref 20 which employs GHOST to examine the two other major oil sands recovery and extraction technologies (surface mining and Cyclic Steam Stimulation/CSS).
’ METHOD The purpose of GHOST is to represent a more complete and transparent range, compared to prior studies, of GHG emissions that may result from currently operating oil sands technologies. GHOST utilizes a process-based LC approach to quantify the emissions, and includes the LC activities associated with the three primary bitumen recovery and extraction technologies (SAGD, surface mining, and CSS) and two of the three major upgrading technologies (delayed coking and hydrocracking). Technologies not included are fluid coking upgrading and gasification of heavy byproducts (e.g., coke, asphaltenes) to satisfy project energy demands. These technologies are each employed by only one operating project6 but should be explored in future work (see Supporting Information). To complete the model, LC activities associated with production of diluent as well as the transport of diluent, diluted bitumen, and SCO are included. User Interface. GHOST allows the user to define a project by selecting from several options. The first is a choice among the different recovery and extraction technologies (SAGD, surface mining, and CSS). The second is a choice of whether or not the bitumen produced will be upgraded, and if so, through delayed coking and/or hydrocracking. Other options include transport distances, type and proportion of diluent utilized, whether cogeneration is/is not utilized, as well as each LC stage’s input parameters (Table 1). GHOST’s input parameters are defined with ranges of values compiled through an inventory process (described in the section, GHOST Model Input Parameters and Data and in Supporting Information), but the model also specifies default values based on their source, quality, and frequency of occurrence in the data series. The user is able to modify any of the input parameters and can run GHOST for a specific project, or examine the performance of a selected technology, using the data series and ranges provided in the model. The model is primarily designed to 9394
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
9395
upgrading
diluted bitumen
transport of
diluent
and extraction
surface mining recovery
(SAGD and CSS)
in situ recovery and extraction
m3/m3 SCO m3/m3 SCO m3/m3 SCO
natural gase
process gas
hydrogen
bit* km)
Wh/ (m3 diluted
piping energy
requirements
km
transport distance
volume of diluent
% volume
natural gas SMR EF
hydrogen volume/SMR hydrogen/ natural gas volume ratio *
combustion EF
process gas volume * natural gas
combustion EF
natural gas volume * natural gas
as collected in the inventory
nature of diluent
as collected in the inventory
kg CO2eq/m3 bit
fugitive hydrocarbons
diesel volume * diesel combustion EF
gas combustion EF
kg CO2eq/m3 bit
kWh/m3 bit
electricity from the grid
ER&EI
EUPI
ETR1
EDB
ratio * natural gas fuel cycle EF
hydrogen volume/SMR hydrogen/natural gas volume
natural gas fuel cycle EF
volume natural gas *
grid EF + feedstock fuel cycle EF)
distance * (Alberta power
feedstock fuel cycle EF]) piping energy requirement *
1,000 km * [Alberta power grid EF +
piping energy requirement *
(diluent fuel cycle EF +
diluent volume *
(Alberta power grid EF + feedstock fuel cycle EF)
amount of power used *
diesel volume * diesel fuel cycle EF
gas fuel cycle EF
natural gas volume * natural
used * (Alberta power grid EF + feedstock fuel cycle EF)
natural gas volume * natural
if no cogeneration: amount of power
gas fuel cycle EF
(2) natural gas volume * natural
boiler efficiency/natural gas HHV
volume required based on the SORb,c: SORd * water/steam enthalpy change/
(1) calculation of the natural gas
indirect emissions calculation
gas use calculations above
flared hydrocarbons
L/m3 bit
diesel
EUPd
as collected in the inventory
m3/m3 bit
kg CO2eq/m3 bit
fugitive hydrocarbons
natural gas
as collected in the inventory
kg CO2eq/m3 bit
flared hydrocarbons
ER&EI
part of Etotal
if cogeneration: included in natural
combustion EF kWh/m3 bit
solution gas volume * natural gas
combustion EF
(2) natural gas volume * natural gas
efficiency/natural gas HHV
required based on the SORb,c: SORd * water/steam enthalpy change/boiler
(1) calculation of the natural gas volume
direct emissions calculation
electricity from the grid
ER&Ed
ER&Ed
part of Etotal
m3/m3 bit
inventory unit
solution gas
SOR
inventory parameter
Table 1. GHOST Model Parameters and GHG Emissions Calculationsa
Environmental Science & Technology ARTICLE
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
as collected in the inventory
km
Wh/ (m3 SCO* km)
kg CO2eq/m3 SCO
kg CO2eq/m3 SCO
transport distance
piping energy requirements
flared hydrocarbons
fugitive hydrocarbons
transport of SCO
feedstock fuel cycle EF)
CSS: cyclic steam stimulation; EF: emissions factor; HHV: higher heating value; SAGD: steam-assisted gravity drainage; SCO: synthetic crude oil; SMR: steam methane reforming; SOR: steam-to-oil ratio; bit: bitumen. For emissions factors see Table S1, Supporting Information. b For Cogeneration Case, the volume of natural gas also depends on the cogeneration elements’ efficiencies and the exhaust gas’ energy content. c Instantaneous SOR (iSOR) for SAGD and cumulative SOR (cSOR) for CSS. This difference of approach is due to the nature and the quality of the data that could be collected for both technologies. d Wet SOR (in contrast to dry SOR) as described in Supporting Information. e Natural gas for steam (and eventually electricity) generation, not for hydrogen production.
a
as collected in the inventory
L/m3 SCO make up diluent
kWh/m SCO the grid
3
electricity from
Table 1. Continued
inventory parameter
inventory unit
diluent volume * diluent fuel cycle EF
part of Etotal
direct emissions calculation
ETR2
(Alberta power grid EF +
piping energy requirement * distance *
(Alberta power grid EF + feedstock fuel cycle EF)
amount of power used *
part of Etotal
indirect emissions calculation
Environmental Science & Technology
ARTICLE
be run with annual average data and to calculate average, steadystate emissions. However, to assess unique emissions scenarios (e.g., during project startup), the user would run the model using input data specific to those situations. GHOST Boundary and Structure. GHOST assesses the “wellto-refinery entrance gate” (WTR) GHG emissions associated with the production of diluted bitumen and SCO from current oil sands technologies. The stages of bitumen recovery and extraction, dilution, transport, and the optional step of upgrading the bitumen to SCO are considered (see Figure 1). GHOST does not account for emissions associated with land use change, construction/decommissioning of facilities, transport vehicle manufacture, or site reclamation. Both direct and indirect emissions associated with WTR activities are included and defined as follows: Direct emissions: Emissions released on-site at the oil sands project during the operation phase (e.g., emissions associated with the combustion of natural gas for steam production), Indirect emissions: Emissions associated with the supply chains of inputs into the operation (e.g., emissions associated with electricity produced off-site but consumed by the project). The emissions included in the calculation of WTR emissions (ETotal) are shown in eq 1[units are g CO2equivalent(eq)/MJ of product entering refinery (diluted bitumen or SCO)]: ETotal ¼ ðER&Ed þ ER&Ei Þ þ EDB þ ETR1 þ ðEUPd þ EUPi Þ þ ETR2
ð1Þ
where ER&Ed: direct emissions associated with bitumen recovery and extraction ER&Ei: indirect emissions associated with bitumen recovery and extraction EDB: emissions associated with diluent blending (all indirect, from diluent supply chain) ETR1: emissions associated with diluted bitumen transport to upgrader or refinery (currently all indirect, from grid electricity generation) EUPd: direct emissions associated with upgrading EUPi: indirect emissions associated with upgrading ETR2: emissions associated with SCO transport to a refinery (currently all indirect, from grid electricity generation) The emissions are calculated based on energy inputs to the processes and their respective emissions factors. Fugitive emissions and emissions associated with flaring, (pipeline) transport, and upstream activities associated with diluent, natural gas, and electricity production are also included. The parameters used to estimate the emissions for each LC stage and the associated direct and indirect emissions calculations are summarized in Table 1 and emissions factors are reported in Table S2. The following section demonstrates how the calculations are carried out for the specific LC activities within in situ recovery and extraction. Details for other pathways and stages can be found in Supporting Information and ref 20. Emissions associated with in situ bitumen recovery and extraction are calculated based on five key parameters: SOR (and associated calculation of required volume of natural gas), solution gas (a byproduct of bitumen production) volume, electricity use, flared hydrocarbons, and fugitive hydrocarbons; and relevant emissions factors (Table 1). The direct emissions associated with natural gas use are calculated in a two step 9396
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
Environmental Science & Technology
ARTICLE
Figure 1. Flowchart of the processes included within boundary of GHOST model for SAGD recovery and extraction, upgrading, and transportation activities for the No Cogeneration Case. Steam is generated using boilers and electricity purchased from the grid.
process: (1) the volume of natural gas required by the process is calculated using the SOR, the boiler feedwater temperature, the water-to-steam enthalpy change, and the system’s HHV efficiency (derivation in Supporting Information); (2) the direct emissions associated with natural gas use are then calculated as the product of the volume of natural gas used and the emissions factor for natural gas combustion. A small fraction (16% of the total gas requirement in SAGD projects based on data collected) of the above natural gas requirement can be displaced by the coproduced solution gas. Direct emissions associated with solution gas are calculated as the product of the volume of solution gas and the emissions factor. Default values for the density, HHV, and emissions factor for solution gas are assumed equal to those of natural gas with justification provided in Supporting Information. Two electricity and steam generation cases that reflect industry practice are included in GHOST: (1) No Cogeneration Case: utilizes a large-scale on-site natural gas industrial boiler and electricity purchased from the Alberta grid, and (2) Cogeneration Case: utilizes an on-site steam and electricity cogeneration facility. If the electricity is purchased from the grid then it is considered indirect (calculations described below). Electricity emissions are considered direct if the electricity is produced onsite (Cogeneration Case). In this case, an emissions factor is applied to the entire cogeneration unit (i.e., total natural gas consumed to produce both steam and electricity is multiplied by an emissions factor that represents the efficiency of the cogeneration unit). The calculation of emissions is more complicated if surplus electricity is generated by the cogeneration unit and sold to the grid, since one of several allocation methods must be chosen and applied (see Supporting Information). In addition, there is the potential for offset credits as the electricity sold would have a lower emissions-intensity than the primarily coal-based Alberta grid.21 Applying these different methods can lead to very different resulting emissions.21 Flared and fugitive emissions are not calculated, they are collected as part of the inventory and added to the calculated direct emissions to yield ER&Ed, the first term in eq 1.
The indirect emissions associated with natural gas use are calculated as the product of the volume of natural gas (described above) and the emissions factor for the natural gas fuel cycle (encompassing activities from natural gas extraction through distribution to the oil sands project). There are no indirect emissions associated with solution gas because it is a byproduct of bitumen production. For the No Cogeneration Case, emissions associated with electricity use are indirect (no emissions result at the point of use) and result from the product of total electricity use and the emissions factor associated with average Alberta grid electricity production. All indirect emissions associated with this LC stage are summed to produce ER&Ei. The model follows a similar process for the other stages as detailed in Table 1. SOR has been shown in prior studies to be the largest determinant of GHG emissions of thermal in situ projects 10 and is a key parameter in determining emissions for in situ pathways in GHOST as described above. Two measures of SOR exist: instantaneous SOR (iSOR) and cumulative SOR (cSOR). The iSOR is typically reported on a per day basis while cSOR is defined over a specified time frame (either a year or over a project’s lifetime)22 and calculated as total steam injected divided by total bitumen produced over that period. For the calculation of LC emissions, using the cSOR over a project lifetime is most relevant, because, for example, the iSOR of a SAGD project will vary during its lifetime. Depending on where a well pair is in the cycle of start up to blow down, iSOR could over- or under-state the average steam required per barrel of bitumen produced from that well. At the start of the project, the iSOR is “infinite” (steam injected but no bitumen produced), then the iSOR decreases as the bitumen production increases and a near “steady state” is achieved. The iSOR increases as the well pair nears completion. It is assumed that the well pair will be shut down when the iSOR increases to uneconomic limits. However, economic limits will differ by company. For additional information on SOR, see the Model Demonstration section and Supporting Information. GHOST Model Input Parameters and Data. GHOST is based on a set of input parameters (see Table 1 and Supporting 9397
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
Environmental Science & Technology Information). For each parameter, a range of values was defined through an iterative process (see Supporting Information). Data in GHOST’s inventory are annual averages that encompass seasonal effects on energy inputs and emissions. An aim of GHOST is to characterize oil sands operations accurately with a reasonable set of parameters to facilitate the model’s use. This difficult balance required a detailed investigation of the importance of each parameter and extensive consultation with industry, government, and academic experts. The ranges are primarily informed by confidential operating data obtained from industry, but to provide a more robust data set where they were limited, the data are augmented with publicly available data7,2331 as well as data provided through expert elicitation (see Supporting Information). Although the data ranges are not exhaustive, they endeavor to represent the variation in potential performance of current recovery, extraction, or upgrading technologies. Although 65% of bitumen has been produced through surface mining up to year-end 2009,8 there are only four operating projects, each with unique attributes based on reservoir characteristics, time since start-up, operating procedures, etc. Thermal in situ operations have produced lesser volumes of bitumen but through a larger number of projects (19 active projects consisting of 13 SAGD and six CSS).3234 As with the surface mining projects, there are unique features of each project. This context, combined with not all operating projects being willing to provide data to support this research, necessitated that we supplement the confidential industry data, and in addition, limited the type of uncertainty analysis feasible with the data set (e.g., the sample sizes and information provided by the data were not sufficient to support a Monte Carlo analysis). Multiple methods were used to evaluate GHOST to determine that it provides consistent/robust results throughout the ranges of input parameters, technologies, and operating conditions (see Supporting Information). Model Demonstration. GHOST is run utilizing the ranges of parameter values compiled for each of the WTR LC stages of the SAGD pathways (Figures 1 and S3 Supporting Information). A key difference between the pathways is the type of product that enters the refinery, which is linked to whether or not the bitumen is upgraded. The pathways that include upgrading result in SCO entering the refinery whereas the diluted bitumen pathways result in either dilbit (if a diluent such as naphtha or natural gas condensate is added) or synbit (if SCO is added as the diluent) entering the refinery. Each of these intermediate products will require different amounts of processing and therefore will result in different emissions at the refinery stage. The broad ranges of parameter values (Table 2) illustrate differences in steam and electricity requirements, natural gas and solution gas use, etc., among SAGD projects. SAGD Recovery and Extraction. SAGD uses a pair of horizontal wells—the upper to inject steam and the lower to collect the heated bitumen flowing by gravity.7 As noted earlier, SOR and associated combustion of natural gas to produce steam to heat the bitumen is a large contributor to overall energy requirements of thermal in situ recovery and extraction processes. Although cSOR over a SAGD project lifetime would be most relevant for calculation of LC GHG emissions, SAGD projects have not been operating long enough to gather sufficient historical data and perspective on cSOR. As a result, in this demonstration of GHOST, emissions intensities are calculated based on a comprehensive data set of iSORs reported by industry for all projects operating in January 2009.35 The data set includes
ARTICLE
projects with wells operating at the beginning, midpoint, and end of their lifetimes. Emissions intensities associated with the full range of iSOR (i.e., 2.15.4) are assessed and reported, but the discussion focuses on results for iSORs of 2.23.3, which accounted for 85% of the bitumen produced. Electricity is consumed in SAGD to operate pumps, to treat process water, etc. In the demonstration, both the No Cogeneration and Cogeneration Cases are examined and in the latter case it is assumed that the cogeneration unit meets all steam and electricity needs and that no electricity is purchased from or sold to the grid (see Supporting Information). Transport. The range of electricity intensity for pumps and controls for pipeline operation is shown in Table 2. For the pathways without upgrading a transport distance of 3000 km is assumed (bitumen extraction to refineries in Petroleum Administration for Defense District (PADD) II, which is comprised of Midwest U.S. states). PADD II was selected as it was the recipient of the majority (70%) of heavy oil exports from Canada in 2010.36 For the pathways with upgrading, a 500 km distance is assumed between bitumen extraction and upgrading and 2500 km between upgrading and refining. Upgrading. The primary purpose of the optional upgrading activity is to separate the light components of bitumen, and convert the heavy components into refinable products. Delayed coking breaks bitumen’s long carbon chains (cracks the bitumen) and pyrolizes it into gas, liquid products, and coke. Hydrocracking also cracks the bitumen but adds hydrogen atoms to the newly formed hydrocarbon molecules. Pathways comprising SAGD recovery and extraction combined with each of the upgrading technologies are included in the model demonstration.
’ GHOST DEMONSTRATION RESULTS The WTR emissions calculated by GHOST for SAGD, comprising both direct and indirect emissions, are 8.718.5 g CO2eq/MJ dilbit, 12.623.8 g CO2eq/MJ synbit, and 18.140.9 g CO2eq/ MJ SCO. The ranges of results are calculated by summing all of the lower and then all of the upper emissions estimates from each LC stage. Appropriate qualifications need to be made in interpreting these results as both these estimates and their aggregation are unlikely to reflect the reality of any one project. However, these WTR results provide a basis for discussing the potential performance of SAGD projects and are an improvement over prior reporting of point estimates based on publicly available data. Table 3 provides a breakdown of the activities contributing to these emissions, with the ranges of results characterizing the input parameters’ variability. Whereas the WTR emissions for dilbit are lower than for SCO, dilbit is a lower-value product and will likely result in higher refinery-related emissions than will SCO. An oil sands operator must consider relevant trade-offs when deciding whether or not to upgrade their bitumen. The availability and price of different diluents will also influence the diluent choice. Considering all of the pathways, there is a wide range in the contribution of direct emissions [direct emissions represent 5464% (dilbit pathway), 2433% (synbit), and 6769% (SCO)]. Indirect emissions represent a much larger fraction of the synbit pathway’s emissions due primarily to the GHG intensity of the diluent (SCO in this case) production. When upgrading is employed, it makes a contribution to the WTR emissions similar to that of the extraction/recovery stage (3541% and 4657% for upgrading and extraction/recovery, respectively) whereas 9398
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
Environmental Science & Technology
ARTICLE
Table 2. GHOST Model Input Inventory: Ranges for SAGD Recovery and Extraction, Upgrading, and Transporta SAGD recovery and extraction
range 2.23.3b
instantaneous steam-to-oil ratio (iSOR - dry) electricity used by the process (kWh/m3 bitumen)
45120
coproduced solution gas used by the process (m3/m3 bitumen)
112
flared hydrocarbon emissions (kg CO2eq/m3 bitumen) fugitive methane emissions (kg CO2eq/m3 bitumen)
0.10.6 0.31.0
boiler or cogen feedwater temperature (°C)
100200
(1) no cogeneration case efficiency: boiler ηB
8085%
(2) cogeneration case efficiency: gas turbine ηGT
3035%
efficiency: HRSG exhaust heat recovery ηHR
5060%
efficiency: HRSG direct firing duct burners ηDB total electricity produced (kWh/m3 bitumen)
95% 3003000 range delayed coking
hydrocracking
SCO/bitumen ratio (m SCO/m bitumen)
0.780.90
0.951.05
electricity used by the process (kWh/m3 SCO)
4070
85130
coproduced process gas used by the process (m3/m3 SCO) hydrogen used by the process (m3/m3 SCO)
55115 65200c
25115 75200c
make-up diluentd (L/m3 SCO)
530
upgrading 3
3
flared hydrocarbons emissions (kg CO2eq/m3 SCO)
510
fugitive methane emissions (kg CO2eq/m3 SCO)
02
(1) no cogeneration case efficiency: boiler ηB
8085%
total gas used by the processe (m3/m3 SCO)
95115
(2) cogeneration case efficiency: gas turbine ηGT
3035%
efficiency: HRSG exhaust heat recovery ηHR
5060%
efficiency: HRSG direct firing duct burners ηDB
95%
total electricity produced (kWh/m3 SCO)
2202200
transport
55115
4004000 range
3
electricity required for pipeline pumping (Wh/(m .km))
1565
a
SAGD: steam-assisted gravity drainage; HRSG: heat recovery steam generator; SCO: synthetic crude oil. b 85% of SAGD bitumen was produced within this range of iSOR, while the full range is 2.15.4.35 c No consensus in data collected on hydrogen requirement upper bound. d The diluent input to the upgrader is not totally recovered; makeup diluent must be purchased (or produced) so that the amount of diluent shipped back to the bitumen extraction plant equals the diluent demand. e Total gas for steam generation = natural gas + process gas; if the amount of process gas is sufficient, the amount of purchased natural gas is 0.
the transportation stage contributes the remaining 725%. Following is a discussion of the breakdown of WTR emissions by LC stage. Bitumen Recovery and Extraction. The emissions calculated by GHOST for SAGD recovery and extraction range from 9.0 to 16.1 g CO2eq/MJ bitumen. Direct emissions are responsible for 7493% of total emissions from this stage. Combustion of natural gas and solution gas (direct) for steam production (and electricity generation in the Cogeneration Case) clearly dominates both direct and total emissions from bitumen recovery and extraction, accounting for 83% (low) and 74% (high) of total emissions for the No Cogeneration Case (92% (low) and 88% (high) for the Cogeneration Case). The emissions resulting from combustion are driven largely by the iSOR of the project. The full range of iSOR (2.15.4 rather than 2.23.3) is explored in the sensitivity analysis and expands the range of recovery and
extraction emissions to 8.624.6 g CO2eq/MJ bitumen. However, we consider these lower and upper iSORs to be less reflective of performance of existing technologies when recovering a large fraction of the bitumen. SAGD projects operating at iSORs lower than 2.2 are outstanding and nontypical of bitumen and reservoir characteristics in Alberta’s oil sands regions. At the other end of the range, companies will likely decide to discontinue projects operating for substantial periods of time at high iSORs due to unattractive economics and high GHG emissions. Whereas on-site natural gas combustion dominates the recovery and extraction stage, several other contributors need to be included in emissions estimates. These include direct flared and fugitive emissions, which are relatively small in the GHOST data set (<0.03 g CO2eq/MJ bitumen) but could become significant in situations of upset operation. In addition, indirect emissions from the natural gas fuel cycle can be substantial when the iSOR 9399
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
b
fugitive emissions
9400
upgrading (g CO2eq/MJ SCO)
total: transport 500 km diluent recycled delayed coking: no cogeneration
(g C02eq/MJ SCO) upgrading (g CO2eq/MJ SCO)
6.8 4.5 5.8
total: hydrocracking - no cogeneration total: hydrocracking upgrading - cogeneration
no cogeneration total: delayed coking upgrading cogeneration
0
make-up diluent
6.4
0.13 0
1.5
hydrogen productiond fugitive emissions
0 4.1
electricity generation (grid) electricity and steam generationd flaring
1.5
steam generation
hydrogen productiond
2.8
1.3
hydrogen productiond d
0 5.4
electricity and steam generationd
hydrogen productiond electricity generation (grid)
1.3
steam generationd
0 0 4.9
electricity for pumping stations
0 0
0
diluent production electricity for pumping stations
total: delayed coking upgrading -
common elements
hydrocracking: cogeneration
hydrocracking: no cogeneration
delayed coking: cogeneration
500 km: dilbit -diluent recycled
(diluent not recycled)
total: transport 3000 km synbit
3000 km: synbit - diluent not recycled
(diluent not recycled)
total: transport 3000 km dilbit
0
3000 km: dilbit - diluent not recycled 0 0
8.4
total: SAGD cogeneration electricity for pumping stations
7.9
total: SAGD no cogeneration diluent production
0.007
flaring
common elements
8.3 0.002
electricity and steam generation
cogeneration
7.9 0
low
steam generation electricity generation (grid)
process
no cogeneration
scenario
diluted bitumen transport to upgrader
(g CO2eq/MJ diluted bitumen)c
diluted bitumen transport to refinery
SAGD recovery and extraction (g CO2/MJ bitumen)
life cycle stage
b
high
12.1
10.3
11.1
10.3
0
0.05
0.27
4.1
0 7.7
4.1
5.9
4.1
6.7
0
4.1
0 5.9
0
0
0
0
0 0
0
12.3
11.9
0.025
0.015
12.2
11.9 0
direct
0.7
2.6
0.5
1.5
0.01
0
0
0.7
2.1
0.6
0.5
1.0
0.5
0.4
0.4
7.9
1.1
6.8
1.1 1.5
0.4
0.6
1.6
0
0
0.6
0.6 1.0
low
3.4
6.4
3.1
4.7
0.07
0
0
3.3
3.2
3.1
3.0
1.7
2.9
1.6
1.6
15.3
4.6
10.7
4.7 5.6
1.0
1.6
4.2
0
0
1.6
1.5 2.7
high
indirect
6.4
7.1
7.4
7.8
0.01
0
0.13
6.3
2.1
4.9
7.2
1.0
6.7
0.4
0.4
7.9
1.1
6.8
1.1 1.5
0.4
9.0
9.5
0.007
0.002
8.9
8.5 1.0
low
total
15.5
16.7
14.2
15.0
0.07
0.05
0.27
15.1
3.2
13.1
13.8
1.7
12.9
1.6
1.6
15.3
4.6
10.7
4.7 5.6
1.0
13.8
16.1
0.025
0.015
13.8
13.4 2.7
high
Table 3. Well-to-Refinery Entrance Gate (WTR) GHG Emissions Associated with SAGD Recovery and Extraction, Upgrading, and Transportation: Ranges of Results from GHOST (g CO2eq MJ1) Corresponding to iSORs of 2.23.3a
Environmental Science & Technology ARTICLE
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
9401
low
high
low
low
8.7
0.9
0.9
hydrocracking + transport 2500 km
delayed coking + transport 2500 km 18.1
4.0
4.0
high
SAGD + dilution + transport 500 km +
0.9
0.9
12.6 19.0
0
0
SAGD + dilution + transport 3000 km SAGD + dilution + transport 500 km +
0
0
SAGD + dilution + transport 3000 km
electricity for pumping stations
synbit (g CO2eq/MJ synbit) SCO (g CO2eq/MJ SCO)
total: transport 2500 km
2500 km
process
indirect
dilbit (g CO2eq/MJ dilbit)
(WTR) results summary
well-to-refinery entrance gate
(g CO2eq/MJ SCO)
SCO transport to refinery
scenario
direct
total
40.9
23.8 39.2
18.5
4.0
4.0
high
SAGD: steam assisted gravity drainage; SCO: synthetic crude oil. Emissions totals represent ranges of emissions and are based on summing all minimum or all maximum emissions results. Note that totals may not add due to rounding. b With the exception of SAGD Recovery and Extraction and Upstream Diluent Production and Diluted Bitumen Transport to Refinery, results are expressed on MJ SCO basis to facilitate the aggregation of WTR results. The disaggregated results do not sum to the WTR result due to the different units used for the different life cycle stages. The SAGD Recovery and Extraction results can be converted to other units using the following ratios: 0.80 (g CO2eq/MJ dilbit)/(g CO2eq/MJ bitumen); 0.52 (g CO2eq/MJ synbit)/(g CO2eq/MJ bitumen); and 1.15 (g CO2eq/MJ SCO)/ (g CO2eq/MJ bitumen). These conversions result in the following ranges of emissions for SAGD Recovery and Extraction: 7.212.9 g CO2eq/MJ dilbit, 4.78.5 g CO2eq/MJ synbit, and 10.418.7 g CO2eq/MJ SCO. c Results presented on/MJ dilbit or/MJ synbit basis as relevant. Dilbit volume blend: 75% bitumen and 25% diluent (naphtha or condensate). Synbit volume blend: 50% bitumen and 50% SCO. Synbit diluent production assumes an emissions factor range of 510770 kg CO2eq/m3 SCO. Make-up diluent fuel cycle emissions accounted for in Upgrading process. d Indirect emissions from Steam Generation, Electricity and Steam Generation, and Hydrogen Production are emissions from the natural gas supply chain and therefore reported jointly. Low-end results assume use of shallow gas; high-end results assume use of conventional natural gas from Southeastern Alberta.
a
b
life cycle stage
Table 3. Continued
Environmental Science & Technology ARTICLE
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
Environmental Science & Technology is high (1.5 g CO2eq/MJ bitumen at an iSOR of 3.3). SAGD recovery and extraction emissions ranges for the No Cogeneration and Cogeneration Cases are similar due to the emissions allocation method (see Supporting Information) employed in GHOST. Given the similarity in the results and the model and parameter uncertainty and variability in our model we cannot conclude that there is a difference in emissions between these options. However, the emissions difference between the two cases widens if the iSOR increases (scale effect), if different allocation methods are employed, and/or if surplus electricity is produced, sold to the electricity grid in Alberta, and provided a credit for offsetting more GHG intensive electricity in the Province.21 Upgrading. Upgrading emissions range from 7.8 to 15.0 g CO2eq/MJ SCO for the No Cogeneration Case and 7.414.2 g CO2eq/MJ SCO for the Cogeneration Case for delayed coking and 7.116.7 (No Cogeneration Case) and 6.415.5 (Cogeneration Case) for hydrocracking. As noted above, given the uncertainty and variability in the model, we cannot conclude that there is a difference between the emissions of the two upgrading technologies even though they are fundamentally different technologies. Reference 37 reviewed studies reporting emissions related to the two technologies and found conflicting results in previous studies. The decision between the technologies is driven by economics, availability, and expected price of fuel to produce hydrogen vs the potential liability/benefit of producing byproduct coke. Upgrading emissions are also dominated by the combustion of both natural gas and process gas for steam and electricity generation as well as the use of natural gas for hydrogen production (direct emissions from these activities are responsible for 6090% of total upgrading emissions, depending on the scenario). The variation in emissions associated with hydrogen production (Table 3) results from the wide range of hydrogen requirements compiled in the inventory (related to the degree of upgrading pursued) and how efficiently the hydrogen is produced. Consequently, the indirect emissions from the natural gas fuel cycle can also account for a large fraction of the total upgrading emissions (up to 22%). Insights for the other contributors to upgrading emissions are similar to those discussed above for bitumen recovery and extraction. Transportation Including Diluent Production. Transport emissions in GHOST are driven by the electricity required to pump the product to the upgrader and/or the refinery. This activity contributes a relatively small to moderate fraction of WTR emissions (725%). Also related to the transport, however, are the upstream emissions resulting from the production of the diluent (discussed above), which is required to decrease the viscosity of the product, allowing it to flow in a pipeline.
’ SENSITIVITY ANALYSIS A sensitivity analysis is completed for a WTR pathway comprised of SAGD recovery and extraction, delayed coking, with the No Cogeneration Case and the relevant transport activities. This pathway was selected to evaluate the impact of individual parameters on the model results based on delayed coking being a more common technology than hydrocracking, and the similarity between the Cogeneration/No Cogeneration Case results. An Example Scenario consisting of a set of default values provided in GHOST that are “typical” or “representative” of operating conditions based on our review of a set of currently operating projects and discussion with experts is used as the base case (see Table S3, Supporting Information for these values). The
ARTICLE
Example Scenario does not reflect an average of the industry nor any one operating project. The sensitivity of the WTR emissions to each of the model input parameters is examined through varying each individual parameter from its low to high values shown in Table 2. The WTR emissions are by far most sensitive to the iSOR, reinforcing the message that this SAGD parameter most significantly affects the process’ GHG emissions (Figure S4, Supporting Information). The Example Scenario results in WTR emissions of 27.9 g CO2eq/MJ SCO, with lower and upper iSOR values resulting in emissions of 25.4 and 41.8 g CO2eq/MJ (9% decrease and 50% increase), respectively. The WTR emissions are also sensitive to the amount of hydrogen consumed in upgrading and the amount of electricity required for pumping the crudes to the refinery. The ranges considered for each parameter result in an approximately 5 g CO2eq/MJ change in WTR emissions. Several of the parameters in the sensitivity analysis have a lesser impact on WTR emissions (e.g., boiler efficiency and feedwater temperature, upgrading conversion volume ratio of bitumen to SCO). However, all of the parameters in the analysis should be carefully accounted for in in situ GHG emissions modeling as the parameters other than SOR can contribute a significant portion to total WTR emissions (up to 76%). In addition, actual project values may fall outside the ranges examined, emphasizing the importance of accounting for the full set of parameters in a WTR analysis.
’ DISCUSSION GHOST improves upon the LC framework and data available for evaluating GHG emissions associated with currently operating oil sands technologies. The model is primarily based on actual operating and process performance data obtained from industry and reviewed with academic, industry, and government experts. The model can be run for a reasonably comprehensive set of oil sands pathway options, using the model’s default data/ranges or user input data, resulting in ranges of model outputs. Users of the model can compare different technologies and pathways using consistent boundaries and have the detail and transparency required to understand the influential parameters/activities as well as explain differences between pathways. These represent advances over prior models that principally have the capability of analyzing only a subset of the technology choices available in GHOST, utilize theoretical design data, and present results as point estimates. Evidence of these refinements is the inclusion of a larger set of WTR activities, broader ranges of input data, and emissions estimates that represent a more extensive range of technology choices and operating conditions than have been previously reported. Limitations of the model and analysis that will be addressed in future work include the following: only WTR activities are included; model input parameters are not explicitly linked with reservoir conditions, nor are they directly linked with the resulting product’s properties; the ranges of results are generated utilizing combinations of “low” and “high” input values for all parameters. The application of GHOST shows that a wide range of technology options, operating conditions, and resulting emissions exist among current oil sands operations, even when considering a single recovery and extraction technology (SAGD). In addition, some SAGD performance could fall outside of the ranges reported in our analysis. On-site combustion of natural gas is the dominant emissions contributor, although the combined 9402
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
Environmental Science & Technology contribution of other activities to WTR emissions is considerable and these activities should not be excluded from analyses. We demonstrate that in reporting WTR emissions associated with diluted bitumen and SCO, ranges of results should be reported as potential performance varies substantially. Understanding the contribution of each activity to WTR emissions can indicate potential opportunities for lowering emissions if combined with a determination of the parameters/activities over which the operators have control. Initiatives that result in lower SORs for the SAGD pathways offer the most obvious potential for reducing GHG emissions intensity. SOR is determined primarily by reservoir conditions (e.g., oil saturation, permeability, porosity), however, design and placement of wells, heat integration, the decision to stop operating a well/project based on performance, and a decision to replace natural gas used in the process with low/zero carbon content energy sources, are all within the purview of the operator. Replacing natural gas with low/zero carbon content energy sources for generating the steam used in bitumen recovery would have the most impact in reducing SAGD emissions, followed by replacing natural gas with a low/zero carbon source of steam, electricity, and hydrogen in the upgrading stage. These changes, whose applicability and economics are beyond the scope of this research, may be unlikely to be implemented in existing projects but have potential in future projects. An example of the types of trade-offs involved in such a transition can be found in ref 38. Based on the assumption that industry will continue to rely on fossil fuelbased energy inputs, the development and usage of carbon capture and storage (CCS) could reduce the GHG emissions of oil sands projects (ref 39 discusses the applicability of CCS in the oil sands). Alternatively, steam solvent processes have the potential to make progress in reducing SORs and consequently emissions, but these processes are still under development and it is uncertain how much makeup solvent will be required (which would impact emissions).22 Another option to facilitate lower SORs is the development of the best reservoirs and the delayed development of those of lesser quality until technology improvements have been achieved. How electricity-related emissions are treated, particularly in cases utilizing cogeneration, can significantly impact the emissions associated with oil sands processes. Finally, lesser reductions in emissions could be obtained through improving operating protocols, process integration, and ensuring operations do not result in significant flare/fugitive emissions. Energy costs are the single largest operating cost for any project, therefore motivating operators to continually improve efficiency and thereby reduce emissions. GHOST can be employed to examine the life cycle activities and resulting emissions associated with the production of oil sandsderived products, providing stakeholders with insights on emissions reduction potential, and when combined with economic analysis, could prioritize cost-effective emissions reductions.
’ ASSOCIATED CONTENT
bS
Supporting Information. Flowcharts and assumptions, data inventory and emissions calculation methods, boiler/ cogeneration details, product characteristics, calculation details, model evaluation, and sensitivity analysis. This material is available free of charge via the Internet at http://pubs.acs.org.
ARTICLE
’ AUTHOR INFORMATION Corresponding Author
*Phone: (416) 946 5056; e-mail: [email protected].
’ ACKNOWLEDGMENT We thank Alberta Innovates-Energy and Environment Solutions, Natural Resources Canada, AUTO21 NCE, and five oil sands companies for financial support of the LCA of Oil Sands Technologies project; Jessica Abella and Ganesh Doluweera (University of Calgary); Sarah Jordaan and David Keith(Harvard University); and Jennifer McKellar and Sylvia Sleep (University of Toronto). ’ REFERENCES (1) U.S. Department of Energy. International Energy Statistics; Energy Information Administration, 2011; http://www.eia.gov/cfapps/ ipdbproject/iedindex3.cfm?tid=5&pid=57&aid=6&cid=CA,SA,VE,& syid=2011&eyid=2011&unit=BB. (2) Climate Change and Emissions Management - Specified Gas Emitters Regulation. Alberta Regulation 139/2007; Alberta Government: Edmonton, AB, 2007; http://www.canlii.org/en/ab/laws/regu/alta-reg139-2007/latest/alta-reg-139-2007.html. (3) Sperling, D.; Yeh, S. Low Carbon Fuel Standards. Issues Sci. Technol. 2009, 2, 57–66. (4) Final Statement of Reasons. State of California Air Resources Board: Sacramento, CA, 2009; http://www.arb.ca.gov/regact/2009/ lcfs09/lcfsfsor.pdf. (5) Guinee, J. B.; Gorree, M.; Heijungs, R.; Huppes, G.; Kleijn, R.; de Koning, A.; van Oers, L.; Wegener Sleeswijk, A.; Suh, S.; Udo de Haes, H. A.; de Bruijn, H.; van Duin, R.; Huijbregts, M. A. J. Handbook on Life Cycle Assessment: Operational Guide to the ISO Standards; Kluwer Academic Publishers: Dordrecht, Netherlands, 2002. http://cml.leiden. edu/research/industrialecology/researchprojects/finished/new-dutchlca-guide.html. (6) Alberta’s Energy Reserves 2009 and Supply/Demand Outlook 20102019; Serial Publication ST98-2010; ISSN 1910-4235; Alberta Energy Resources Conservation Board: Calgary, AB, 2010; http://www. ercb.ca/docs/products/STs/st98_2010.pdf. (7) Oil Sands Technology Roadmap, Unlocking the Potential; Alberta Chamber of Resources: Edmonton, AB, 2004; http://www.arc.ab.ca/ documents/Oil%20Sands%20Technology%20Roadmap.pdf. (8) Dunbar, R. B. Canada’s Oil Sands A World-Scale Hydrocarbon Resource; Strategy West Inc.: Calgary, AB, 2010; http://www.strategywest.com/downloads/StratWest_OilSands_2010.pdf. (9) Canada’s Energy Future, Infrastructure Changes and Challenges to 2020; National Energy Board: Edmonton, AB, 2009; http://www.neb. gc.ca/clf-nsi/rnrgynfmtn/nrgyrprt/nrgyftr/2009/ nfrstrctrchngchllng2010/nfrstrctrchngchllng2010-eng.pdf. (10) Charpentier, A. D.; Bergerson, J. A.; MacLean, H. L. Understanding the Canadian oil sands industry’s greenhouse gas emissions. Environ. Res. Lett. 2009, 4, 1–11. (11) Life Cycle Assessment Comparison of North American and Imported Crudes; Jacobs Consultancy and Life Cycle Assoc. for the Alberta Energy Research Institute: Chicago, IL, 2009; http://eipa.alberta.ca/ media/39640/life%20cycle%20analysis%20jacobs%20final%20report.pdf. (12) Comparison of North American and Imported Crude Oil Lifecycle GHG emissions; TIAX LLC for the Alberta Energy Research Institute: Cupertino, CA, 2009; http://eipa.alberta.ca/media/39643/life%20cycle%20analysis%20tiax%20final%20report.pdf. (13) An Evaluation of the Extraction, Transport, and Refining of Imported Crude Oils and the Impact on Life Cycle Greenhouse Gas Emissions; DOE/ NETL-2009/1362; National Energy Technology Laboratory, 2009; http:// www.netl.doe.gov/energy-analyses/pubs/PetrRefGHGEmiss_ImportSourceSpecific1.pdf. 9403
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
Environmental Science & Technology (14) Oil Sands, Greenhouse Gases, and US Oil Supply: Getting the Numbers Right; Special Report; IHS CERA Inc.: Cambridge, MA, 2010; http://www2.ihscera.com/docs/Oil_Sands_Energy_Dialogue_0810.pdf. (15) Wang, M. Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) Model, Version 1.8.d.1; Center for Transportation Research, Argonne National Laboratory: Argonne, IL, 2010; http://greet.es.anl.gov/. (16) Natural Resources Canada. GHGenius Version 3.20; Natural Resources Canada: Ottawa, ON, 2011; http://www.ghgenius.ca. (17) PE International. GaBi, Version 4.4; PE International: LeinfeldenEchterdingen, Germany, 2011; http://gabi-software.com. (18) PRe Consultants. SimaPro, Version 7.2; PRe Consultants: Amersfoort, The Netherlands, 2010; http://www.pre.nl/. (19) Gates, I. D.; Adams, J.; Larter, S. The impact of oil viscosity heterogeneity on the production characteristics of tar sand and heavy oil reservoirs. Part II: Intelligent, geotailored recovery processes in compositionally graded reservoirs. J. Can. Pet. Technol. 2008, 47 (9), 40–49. (20) Kofoworola, O.; Charpentier, A. D.; Bergerson, J. A.; Sleep, S.; MacLean, H. L. Life cycle greenhouse gas emissions of current oil sands technologies: surface mining and CSS applications. Environ. Sci. Technol. 2011 Under revision. (21) Doluweera, G. H.; Jordaan, S. M.; Moore, M. C.; Keith, D. W.; Bergerson, J. A. Evaluating the role of cogeneration for carbon management in Alberta. Energy Policy 2011, in press. (22) Bitumen Recovery Technology: A Review of Long-Term R&D Opportunities; LENEF Consulting Ltd.: Calgary, AB, 2005; http:// www.ptac.org/links/dl/BitumenRecoveryTechnology.pdf. (23) Alberta’s Energy Reserves 2008 and Supply/Demand Outlook 20092018; Serial Publication ST98-2009; ISSN 1910-4235; Energy Resources and Conservation Board: Calgary, AB, 2009; http://www. ercb.ca/docs/products/STs/st98-2009.pdf. (24) Alberta Mineable Oil Sands Plant Statistics - Annual 2007; Statistics Series 43; Energy Resources and Conservation Board: Calgary, AB, 2008. (25) Application for Approval of the Scotford Upgrader Project; Shell Canada Ltd.: Calgary, AB, 1998. (26) Application for Approval of the Muskeg River Mine Project; Shell Canada Ltd.: Calgary, AB, 1997. (27) Natural Resources Canada. GHGenius, Version 3.14b; Natural Resources Canada: Ottawa, ON, 2009; http://www.ghgenius.ca. (28) Wang, M. Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) Model, Version 1.8b; Center for Transportation Research, Argonne National Laboratory: Argonne, IL, 2008; http://greet.es.anl.gov/. (29) Purchased Natural Gas Use by the Canadian Oil Sands Industry; Strategy West Inc. for the Canadian Association of Petroleum Producers: Calgary, AB, 2007; http://www.capp.ca/getdoc.aspx?DocID=119713. (30) Oil Sands Co-Generation Potential - Survey Report; Athabasca Regional Issues Working Group, Co-Generation/Transmission Committee: Fort McMurray, AB, 2008. (31) Canada’s Oil Sands: Opportunities and Challenges to 2015 - an Update; National Energy Board: Calgary, AB, 2006; http://www.neb.gc. ca/clf-nsi/rnrgynfmtn/nrgyrprt/lsnd/pprtntsndchllngs20152006/ pprtntsndchllngs20152006-eng.pdf. (32) Alberta Mineable Oil Sands Plant Statistics, Monthly Supplement; March 2011; ST 39; Energy Resources Conservation Board: Calgary, AB, 2011. (33) Alberta Oil Sands Industry: Quarterly Update; Spring 2011; Reporting on the period Dec. 14, 2010 to Mar. 4, 2011; Government of Alberta, 2011; http://www.albertacanada.com/documents/AOSID_QuarterlyUpdate_.pdf. (34) AccuLOGS. AccuMap Software; Data provided under license by IHS Inc., its subsidiary and affiliated companies, 2011. (35) AccuLOGS. AccuMap Software; Data provided under license by IHS Inc., its subsidiary and affiliated companies, 2009. (36) National Energy Board. Total Crude Oil Exports by Destination Annual; National Energy Board, 2011; http://www.neb-one.gc.ca/clf-nsi/ rnrgynfmtn/sttstc/crdlndptrlmprdct/ttlcrdlxprtdstntn-eng.html.
ARTICLE
(37) Appendix B. ICF Report: Life Cycle Greenhouse Gas Emissions of Petroleum Products from WCSB Oil Sands Crudes Compared with Reference Crudes; ICF International, 2011; http://www.keystonepipeline-xl.state. gov/clientsite/keystonexl.nsf/SDEIS_Appendix%20B_ICF%20Report.pdf?OpenFileResource. (38) McKellar, J. M.; Bergerson, J. A.; MacLean, H. L. Replacing natural gas in Alberta’s Oil Sands: Trade-offs Associated with alternative fossil fuels. Energy Fuels 2010, 24, 1687–1695. (39) Bergerson, J.; Keith, D. The truth about dirty oil: Is CCS the answer? Environ Sci. Technol. 2010, 44 (16), 6010–6015.
9404
dx.doi.org/10.1021/es103912m |Environ. Sci. Technol. 2011, 45, 9393–9404
ARTICLE pubs.acs.org/est
Greenhouse Gas Emissions from Operating Reserves Used to Backup Large-Scale Wind Power Matthias Fripp* Environmental Change Institute, University of Oxford, South Parks Road, Oxford, OX2 3QY, United Kingdom
bS Supporting Information ABSTRACT: Wind farms provide electricity with no direct emissions. However, their output cannot be forecasted perfectly, even a short time ahead. Consequently, power systems with large amounts of wind power may need to keep extra fossil-fired generators turned on and ready to provide power if wind farm output drops unexpectedly. In this work, I introduce a new model for estimating the uncertainty in short-term wind power forecasts, and how this uncertainty varies as wind power is aggregated over larger regions. I then use this model to estimate the reserve requirements in order to compensate for wind forecast errors to a 99.999% level of reliability, and an upper limit on the amount of carbon dioxide that would be emitted if natural gas power plants are used for this purpose. I find that for regions larger than 500 km across, operating reserves will undo 6% or less of the greenhouse gas emission savings that would otherwise be expected from wind power.
1. INTRODUCTION Forecasts of electricity production from wind farms are uncertain, even an hour or less ahead. Conventional power plants require time to start up and begin delivering power to the electric grid, so system operators must turn on extra reserve capacity in advance, to provide power if the output from wind farms falls below the forecast. Alternatively, wind production could rise above the forecast, leaving additional conventional capacity online but unused. Both types of online, idle power plants could consume significant amounts of fuel, causing substantial greenhouse gas emissions, and reversing some of the emission savings that would otherwise be expected from wind power. Several studies have concluded that little or no additional reserves would be needed to integrate wind power meeting 1030% of loads in large power systems.16 However, most of these studies have not looked closely at the possibility of large, very rare errors in wind power forecasts. Some assess the uncertainty of wind forecasts made during a 13 year period,13 which is not long enough to achieve the 1-in-10-year reliability level often sought by the power industry. Others 46 assume that wind power forecast errors can be modeled with a Gaussian distribution, using the standard deviation of errors during a 13 year period. However, wind forecast errors often follow a thickertailed, non-Gaussian distribution,7 so this approach can underestimate the likelihood of large errors. In this work, I introduce a new model for estimating the distribution of errors in short-term, regional wind power forecasts. This model could be used to improve large-scale wind integration studies, but in this work, I use it for a simpler example, estimating the emissions that would occur if gas-fired reserves were used to compensate directly for 99.999% of wind forecast errors in regions of various sizes. This example serves to set an upper limit on the emissions that could occur when firming up wind power using natural gas power plants. Power systems already carry significant reserves to compensate for errors in electricity load forecasts or unexpected r 2011 American Chemical Society
power plant outages. Wind forecast errors are mostly independent of these other sources of uncertainty, and many power systems already have substantial surplus reserves. Consequently, if wind farms provide a small to moderate share of a system’s power, existing reserves will be able to compensate for most wind forecast errors, and only a small amount of additional reserves will be required—much less than if dedicated reserves were used to backup the wind separately. However, if wind is added in much larger amounts, then its uncertainty could begin to dominate over the other sources, and the incremental reserve requirements would asymptotically approach (but never exceed) the reserves that would be required to backup wind separately.1,8 Katzenstein and Apt 9 estimated the emissions from backing up wind farms in this way using natural gas plants, but they assumed that wind farms would be backed up 100% by natural gas power plants at all times. Their approach makes no use of wind power forecasts or geographic aggregation, and implicitly assumes that wind power output could drop to zero at any time. Mills et al. 10 pointed out that some backup plants could be turned off entirely when wind power forecasts indicate that they are not needed. However, they did not investigate how closely operating reserves could be tailored to the wind forecast without risking shortfalls in electricity production. The power system model introduced in the second half of this work investigates this possibility further. The example given here focuses on two types of natural gas power plants with startup times of an hour or less, and correspondingly short forecast horizons. This could represent a power system with large shares of gas and wind power (but no hydroelectric or demand-response capability), or a tranche of power provided by gas and wind in a system that contains other, Received: February 7, 2011 Accepted: July 28, 2011 Revised: June 14, 2011 Published: July 28, 2011 9405
dx.doi.org/10.1021/es200417b | Environ. Sci. Technol. 2011, 45, 9405–9412
Environmental Science & Technology less-flexible resources. Large scale wind may be difficult to develop in power systems with large shares of coal or nuclear power, partly because those resources are not economically or technically suited to ramping up and down to complement wind, and partly because power systems seeking low emissions may switch from coal to gas power at the same time as they adopt large shares of wind power. The models shown here could be readily extended to apply to power systems with additional technologies and forecast horizons (e.g., coal plants committed 24 h or more in advance).
2. WIND FORECAST ERROR SIMULATION MODEL Because there is a delay between the time when a conventional power plant is committed (given the order to startup) and when it can be dispatched (deliver power on-demand), power system operators must commit plants based on an advance forecast of the amount of power that will be needed. They must also commit extra capacity in case this forecast is incorrect. In the power system model presented below, combined cycle natural gas plants (CC capacity) are committed based on a regional wind forecast 60 min before power is needed (t60 min), and then simple cycle combustion turbines (CT capacity) are committed at t20 min. Consequently, the performance of the power system depends on two random variables: the changes in the regional wind forecast from time t60 to t20 and from t20 to t. The first of these variables determines how much CT capacity must be committed at t20, to “true up” for errors in the t60 forecast. The second variable determines whether the capacity committed at time t20 is sufficient to prevent a generation shortfall at time t. The wind forecast error model described below simulates samples from the joint distribution of these two random variables at multiple independent time steps. It first specifies the probability distributions for forecast errors at hypothetical wind farm sites, and the correlation between those sites, based on statistical data from 10 existing wind farms. Then it randomly samples from these distributions to simulate the forecast errors that could occur at the hypothetical wind farms. In Section 3, these simulated forecast errors are used to set appropriate reserve margins for wind dispersed across various-sized geographic regions, and to estimate the carbon dioxide emissions when natural gas plants are used to correct for the wind forecast errors. 2.1. Wind Model Definitions. As discussed above, the key variables governing operation of the power system model below are the changes in the wind forecast from time t60 to t20 and from t20 to t. I identify these random variables as Δt,40 and Δt,20, respectively: Δt, 40 Ft, 20 Ft, 60 Δt, 20 Wt Ft, 20 where Ft,h is the forecast of total output from all the wind farms in a region at time t (expressed in minutes), made h minutes earlier, and Wt is the total output from all wind farms at time t. Both Ft,h and Wt are expressed as “capacity factors,” i.e., fractions relative to the total rated capacity (maximum possible output) of all wind farms in the region. The regional forecast Ft,h and regional power output Wt are defined as the weighted sum of forecasts (fi,t,h) of power output
ARTICLE
(wi,t) at individual wind farms: Ft, h
∑i fi, t, h zi = ∑i zi
Wt
∑i wi, t zi = ∑i zi
where i indexes over all wind farms in the power system and zi is the rated capacity (maximum possible power output) of wind farm i (in MW). It is also helpful to define forecast error variables for individual sites, analogous to the variables used for the whole region: δi, t, 40 fi, t, 20 fi, t, 60 δi, t, 20 wi, t fi, t, 20 With these definitions, the region-wide forecast errors are simply the weighted sum of forecast errors at individual wind farms: Δt, 40 ¼
∑i δi, t, 40 zi = ∑i zi
ð1Þ
Δt, 20 ¼
∑i δi, t, 20 zi = ∑i zi
ð2Þ
For this work, I assume that the individual wind farm forecasts fi,t,h are based on “persistence.” This is the simplest possible forecasting method, assuming simply that wind power output will stay constant at the current level. Although other methods may be able to improve on persistence forecasts (especially for horizons longer than a few hours), this method is always available. Consequently, it provides an upper bound on forecast uncertainty. (The model could also be used with other forecasting methods.) Persistence forecasting gives the following relations: fi, t, h ¼ wi, th δi, t, 40 ¼ fi, t, 20 fi, t, 60 ¼ wi, t20 wi, t60
ð3Þ
δi, t, 20 ¼ wi, t fi, t, 20 ¼ wi, t wi, t20
ð4Þ
In this case, Δt,40 simply shows how much the wind changes between 60 min before the time when electricity is needed and 20 min before. Then Δt,20 shows the change in wind power output during the last 20 min before electricity is needed. The sum of the two shows the change in wind power during the full hour before electricity is needed. 2.2. Simulation of Wind Power Forecast Errors. The wind error model simulates samples from the joint distribution of Δt,40 and Δt,20, for arbitrary collections of hypothetical wind farms. This simulation begins by assigning marginal probability distributions for δi,t,40 and δi,t,20 to individual hypothetical wind farms, based on historical data from 10 real wind farms. Correlated random draws from these distributions are then used to represent simultaneous errors of each type at each site. Finally, the errors at individual wind farms are combined to create samples of region-wide forecast errors (Δt,40 and Δt,20). These steps are repeated for many independent time steps, t. The core of this method is the creation of correlated random draws from the probability distributions for the hypothetical wind farms. For each time step t, these are generated in two steps: First (in a system with n hypothetical wind farms), 2n variables are generated with uniform marginal distributions between 0 and 1, 9406
dx.doi.org/10.1021/es200417b |Environ. Sci. Technol. 2011, 45, 9405–9412
Environmental Science & Technology
ARTICLE
and a correlation matrix Rs. Next, these uniform variables are transformed to match the assigned distributions of δi,t,40 and δi,t,20 for each wind farm. The second step is done via inverse transform sampling, in which the uniform variables are treated as ranks of the wind farm error distributions, and used to select quantiles from those distributions. The transformed variables have the assigned marginal probability distributions, and by definition their ranks have a correlation matrix equal to Rs (in other words, their Spearman’s rank correlation is equal to Rs). In this simulation, Rs is assigned a priori to match the rank correlation expected for the hypothetical wind farms, based on the distance between them. Specifically, the simulation proceeds via these steps: Gather statistics from real wind farms:
rs(δi,40, δj,20)), based on the distance between the two wind farms. The expected rank correlations are derived from the rank correlations at the real wind farms via linear interpolation among the values collected in step 3. This is equivalent to choosing rank correlations for sites d km apart by reading the value for distance d from one of the straight lines shown in Figure SI.2 of the SI. (7) The tables from step 6 are assembled into a single large matrix, Rs, showing the expected rank correlation between every pair of wind farms, for both error statistics, i.e., the expected correlation matrix for the vector [F1,40(δ1,40), ..., Fn,40(δn,40) | F1,20(δ1,20), ..., Fn,20(δn,20)], for wind farms 1...n. (8) Rs is not guaranteed to be a valid correlation matrix (in particular, positive definite), so it is adjusted using the algorithm proposed by Higham.11 This adjustment produces a matrix very similar to the original one, with around 0.999 correlation between the terms of the two matrices. Generate data with the specified statistical properties:
(1) Empirical values of δm,t,40 and δm,t,20 are calculated for each of ten active wind farms, using eqs 3 and 4 with power production data for each minute of the year from July 2008 through June 2009. These calculations exclude periods when the wind farm was curtailed or data was unavailable. (More details on the wind farms and curtailments are given in the Supporting Information.) (2) The marginal empirical cumulative density functions of δm,40 and δm,20 are tabulated for each of the 10 wind farms (indexed by m). These functions, F^m,40(x) and F^m,20(x), show the fraction of the time when the forecast error δm,40 or δm,20 for each wind farm m is less than or equal to x, where x can range from 1 to 1. The inverse of these functions will later be used to convert uniform variables into variables with the appropriate distribution. (3) The following information is tabulated for each pair of wind farms (m and n) chosen from among the 10 wind farms (55 unique pairs, including pairing each wind farm with itself): (a) distance between the two wind farms in kilometers (dm,n), (b) Spearman’s rank correlation between δm,40 and δn,40 (rs(δm,40, δn,40)), (c) Spearman’s rank correlation between δm,20 and δn,20 (rs(δm,20, δn,20)), and (d) Spearman’s rank correlation between δm,40 and δn,20 (rs(δm,40, δn,20)).
(9) The methods described by Li and Hammond 12 are used to simulate random variables with uniform marginal distributions on [0, 1], and the correlation structure defined in step 7. There are 2n such variables: u1,t,40, ..., un,t,40, u1, t,20, ..., un,t,20. These variables are simulated repeatedly for many time steps; 1 e t e 107 for each scenario reported in this work. (10) Forecast errors for each site i are calculated from the simulated uniform variables using the inverse cumulative density function assigned to each wind farm for each type of error: δi, t, 20 Fi, 20 1 ðui, t, 20 Þ δi, t, 40 Fi, 40 1 ðui, t, 40 Þ The simulated errors at each virtual wind farm have the same distribution as the corresponding real wind farm. Each pair of virtual wind farms also has the rank correlation that would be expected based on their distance apart. This correlation applies when comparing the δ40 errors at different sites, the δ20 errors at different sites, or the δ40 and δ20 errors at the same or different sites. (11) The simulated errors at all wind farms are then summed to simulate the total error at all wind farms:
The rank correlations are calculated based on all periods when data were available for both of the paired wind farms. These distances and rank correlations are shown in Figure SI.2 of the SI. Define the statistical properties for hypothetical wind farms (4) A set of “virtual” wind farms is defined for analysis. Each wind farm is described by its rated output zi and location. These can correspond to existing or proposed wind farms in a power system. (5) Each virtual wind farm i is assigned randomly to have the same marginal distributions of δ40 and δ20 as one of the ten real wind farms m:
Δt, 40 ¼ Δt, 20 ¼
∑i δi, t, 40 zi = ∑i zi ∑i δi, t, 20 zi = ∑i zi
The simulated variables Δ*t,40 and Δ*t,20 have the joint distribution that would be expected for Δt,40 and Δt,20 for the full set of virtual wind farms, based on their relative locations and the expected correlation between the 40-min and 20-min error statistics at each wind farm (as derived from the 10 real wind farms). (It should be noted that the simulated variables for different time steps t do not represent a chronological sequence; they simply show forecast errors that could occur during the hour before any randomly chosen minute.)
Fi, 40 ðxÞ F^m, 40 ðxÞ Fi, 20 ðxÞ F^m, 20 ðxÞ (6) Tables are made showing the expected rank correlation between the error variables at every possible pair of virtual wind farms i and j (rs(δi,40, δj,40), rs(δi,20, δj,20) and 9407
dx.doi.org/10.1021/es200417b |Environ. Sci. Technol. 2011, 45, 9405–9412
Environmental Science & Technology
ARTICLE
Table 1. Operating Parameters and Variable Names for Natural Gas Power Plants combined cycle (CC) simple cycle (CT) committed capacity
cCC
cCT
idle capacity generating capacity
rCC gCC
rCT gCT
startup time (minutes) minimum generation
model, I simulated Δt,40 and Δt,20 for all wind farms in Texas in 20062009, and compared the result to the actual in aggregate power output during this period.13 This simulation used virtual wind farms identical in size and location to the actual Texas wind farms, and the virtual wind portfolio was updated as the real wind portfolio expanded from 1840 MW to 7892 MW during the 3-year period. For this simulation, only data from the six NextEra wind farms outside of Texas were used to initialize the simulation model. Figure 1 compares kernel probability densities for the simulated and observed values of Δt,40 and Δt,20 during this period. The simulated values are close to the observed values, but slightly overestimate the likelihood of rare, extreme changes in wind power (the outer edges of the distribution). The green traces show the closest match that could be achieved using Gaussian or Laplacian distributions. Both of these distributions underestimate the likelihood of extreme events. As discussed in the Supporting Information, the model tends to overestimate the likelihood of extreme events in the following 20 min when Δt,40 is near zero, or when Δt,40 is strongly positive or negative. Two traits make this model a good choice for studying power system reserve requirements. (1) It tends to predict extreme events as often or more often than they actually occur, while generic models (e.g., Gaussian or Laplacian distributions) tend to underestimate the likelihood of extreme events. (2) This model works with no advance knowledge of the Texas wind farms other than their location and size. Gaussian or Laplacian models must be parametrized using the standard distribution and correlation of Δt,40 and Δt,20, which are generally not known in advance. It is also unclear in advance whether the distribution of errors in a given region will be more nearly Gaussian or Laplacian. Gaussian models are valid if a power system faces independent errors from many different wind farms; however, they tend to underestimate the risk in real power systems, where forecast errors from different wind farms may be strongly correlated. In the simulation described in Section 4, small regions were found to have nearly Laplacian distributions, while larger regions had more Gaussian distributions.
3. POWER SYSTEM OPERATION 3.1. Commitment and Dispatch of Operating Reserves. Section 2 presented a statistical model of the error in forecasting wind power in a large power system. In order to assess the effect
20 mCC = 35%
vr,CC = 57
vr,CT = 109
vg,CC = 382
vg,CT = 573
(% of committed capacity)
Figure 1. Simulated and observed probability densities of 40-min (Δ40) and 20-min (Δ20) changes in wind power production in Texas, 20062008. Red traces show observed values and blue traces show the simulated distribution. Green traces show Laplacian (dotted) and Gaussian (dashed) distributions with the same mean and standard deviation as the observations.
2.3. Wind Model Validation. To validate the wind variability
60 mCC = 35%
idle-mode emission rate (kg CO2 per MWh) generation-mode emission rate (kg CO2 per MWh)
of these errors on operation of the power system, we must identify how these forecasts could be used to decide which other (conventional) power plants to use and when. Fossil power plants need time to start up. Consequently, decisions about which plants will be used to serve loads at time t must be made before the wind power output for time t is known accurately. For this work, I identify two generic types of fossil power plant: combined cycle natural gas turbines (CC) and simple cycle natural gas combustion turbines (CT). I treat each of these resources as a homogeneous block of “capacity.” At any given time, parts of the CC and CT capacity can be in either a “committed” (turned on and running) or “off” mode. The committed power plants can be in two submodes: either idle (acting as spinning reserves) or generating power. The variables used to indicate the amount of capacity in each online mode are listed in the first three rows of Table 1. Power plants require advance notice to switch from uncommitted to committed mode. Once capacity has been committed, it can be switched between idle and generating mode very quickly. However, due to engineering and environmental restrictions, a minimum percentage of the committed capacity must always be kept in generating mode. All committed capacity generates some greenhouse gas emissions, and these are higher if the capacity is actively generating electricity. The last four rows of Table 1 show the startup time, minimum generation level, idle-mode emissions and generating-mode emissions assumed for this study, as well as the names of variables used to refer to these factors. Startup times are conservative estimates for CC plants after an 8-h shutdown,15 or for typical CT plants.16 Emissions for CC and CT capacity in generation mode come from the profiles used in the U.S. National Energy Modeling System.17 Idle-mode emissions are assumed to be 15% of the generatingmode emissions for CC plants and 19% for CT plants, in keeping with the part-load efficiency curves shown in Kim 18 for state-ofthe-art plants with inlet guide vane control. A key trade-off comes from the fact that CC capacity generates lower emissions when online, but requires longer startup times than CT capacity. This work does not consider ramp rate limitations of fossil power plants, other than the time required to startup. Katzenstein and Apt 14 show ramp rates around 40%/min for CT and 2.5%/min for CC (relative to nameplate capacity). These are much higher than the largest changes in wind power for any but the smallest regions studied in section 4 (e.g., spinning reserve requirements of 10% of wind capacity to cover 20-min forecast errors). In this work, I estimate the emissions that could be generated if CC and CT capacity are used in concert with wind power to reliably meet a time-varying electricity load Lt. 9408
dx.doi.org/10.1021/es200417b |Environ. Sci. Technol. 2011, 45, 9405–9412
Environmental Science & Technology
ARTICLE
In order to estimate the emissions due to firming up wind power separately from load and other power system contingencies, I assume that the load Lt is known in advance, and that natural gas power plants are perfectly reliable. In reality, a common pool of reserves would be used to manage all of these risks, reducing the amount needed for each one (since the three sources of uncertainty often offset each other). In power systems where the uncertainty of wind power is small compared to the uncertainty in loads or the largest possible conventional plant outage, much less reserves will need to be added to manage the uncertainty in wind power. However, in systems with very large amounts of wind power, wind output could become the dominant risk factor, and the wind-specific reserve requirements would asymptotically approach the values reported here, although they may always remain well below. Consequently, the assumptions used in this work are highly conservative for power systems with small amounts of wind power, but remain safe for power systems with any amount of wind power. I assume a simple algorithm is used for commitment and dispatch of the fossil capacity, based on a rolling assessment of available resources and risk. At time t60 min, system operators commit enough CC capacity to meet the expected load at time t (Lt Ft,60), plus an extra spinning reserve margin s60, which does not change over time. That is, cCC, t ¼ ðLt Ft, 60 Þ þ s60
ð5Þ
where cCC,t is the amount of CC capacity committed at time t60 to come online at time t, and s60 is the hour-ahead spinning reserve margin. Equations 1, 2, and 5 can be combined to express cCC,t in terms of the random variables Lt, Wt, Δt,40 and Δt,20, and the operating parameter s60: cCC, t ¼ Lt Wt þ Δt, 40 þ Δt, 20 þ s60
ð6Þ
At time t60, operators also ensure that they have an additional nonspinning reserve margin n60, consisting of power plants that are not turned on yet but could be committed at time t20 if needed to serve loads at time t. For this work, these are assumed to be existing CT plants that are available for commitment but not yet turned on. The sum of the parameters s60 and n60 must be high enough to compensate for extreme drops in wind power over the following 60 min; their values are discussed further below. Then, at time t20, system operators check the latest wind power forecast and, if needed, commit CT capacity (cCT,t) to cover the electricity load (Lt), net of the new wind power forecast
Rðs60 , s20 , n60 Þ ¼
∑t
(
1 0
if Lt Wt > cCC, t þ cCT, t otherwise
∑t 1
¼ min
max
Δ40 þ s20 s60 0
n60
ð7Þ
The second line uses eqs 2 and 6 to express cCT,t in terms of the operating parameters s20 and s60, and the random variable Δ40. Equation 7 indicates that generally at time t20, system operators will commit enough CT generators to compensate for any change in the wind forecast since the CC generators were committed (Δ40), as well as a reserve margin to cover a possible drop in wind power between time t20 and t (s20), less the spinning reserves already committed at time t60 (s40). This strategy reduces emissions by taking a “wait and see” approach to generator commitment, making final decisions about spinning reserves at the latest possible forecast horizon (20 min in this example), when the uncertainty is relatively low compared to earlier forecast horizons. However, when the final commitment decision is made, the system always commits enough generators to compensate for the worst-case error that could occur by the time power is needed. (This model could be generalized to cover multiple forecast horizons and power technologies such as coal, hydro or load-response; however, for simplicity I only investigate a two-technology scenario in this paper.) 3.2. Risk of Power Shortfall. It is now possible to estimate the probability of running short of wind power during any randomly selected minute due to errors in the wind forecasts made 20 and 60 min before. The power system will run short of power during minute t if and only if the load minus wind power production is greater than the amount of CC and CT generation committed at t60 and t20, respectively. So the average risk of power shortfall, as a function of the operating parameters s60, s20, and n60 is as follows:
if s60 þ Δ40 þ Δ20
> > :0
otherwise
Note that this does not consider the possibility of running short of power because too little generation capacity has been built or because of unexpected outages at nonwind power plants. It focuses only on operational decisions in a system with sufficient resources available. 3.3. Emissions Impacts of Forecast Errors. In order to estimate the emissions that come from using fossil plants to
(
8 > > > > <1
∑t > > ¼
(Ft,20), plus a constant 20-min spinning reserve margin s20, set high enough to compensate for the largest foreseeable drop in wind power over the next 20 min. The amount of CT committed at time t20 must meet two constraints: it cannot be more than the nonspinning reserves contracted at time t60 (cCT,t e n60), and it cannot be less than zero (cCT,t g 0). Consequently, the amount of CT capacity committed at time t20 to be online by time t is as follows: ( Lt Ft, 20 cCC, t þ s20 max 0 cCT, t ¼ min n60
8 ( )9 > < max Δ40 þ s20 s60 , ,> = 0 þ min <0 > > : n60 ;
∑t 1
ð8Þ
compensate for wind forecast errors, we must know how often gas capacity will be used to generate electricity and how often it will be committed but idling. For this work, I assume that dispatch follows a simple set of rules, in order: (1) all CT capacity committed for time t must generate at least its minimum required level of power (column 5 of Table 1) (For simplicity I 9409
dx.doi.org/10.1021/es200417b |Environ. Sci. Technol. 2011, 45, 9405–9412
Environmental Science & Technology
ARTICLE
assume that loads are always high enough to keep CC production above its minimum level. This assumption does not affect reliability, since wind can be curtailed in order to maintain CC generation above its minimum level. However, it does omit the emissions that occur during occasional periods when gas power is kept on by the minimum-generation constraint.); (2) committed CC capacity will be dispatched (switched from idle to generation mode) to satisfy the remaining load, until all committed CC capacity is exhausted; (3) CT capacity will be dispatched to satisfy any remaining load, up to the total amount of committed CT capacity. This gives the following equations for power generation by CC and CT plants at time t. ( Lt Wt gCT, t gCC, t ¼ min cCC, t ( Lt Wt gCT, t ¼ min ð9Þ Lt Wt þ s60 þ Δ40, t þ Δ20, t
( gCT, t ¼ max
min
mCT cCT, t
( ¼ max
Lt Wt cCC, t cCT, t
min
Δ40, t Δ20, t s60 cCT, t
mCT cCT, t
E¼
∑t νg, CC gCC, t
ð10Þ
rCC, t ¼ cCC, t gCC, t
ð11Þ
rCT, t ¼ cCT, t gCT, t
ð12Þ
It is now possible to estimate the emissions associated with firming up wind power for use in the electricity system. These “excess emissions” can be defined as the difference between the emissions when running a combination of wind, CC and CT plants using the rules above, and those expected when running a CC plant alone to meet the loads. If wind power directly displaced CC operation, then the emission savings would be equal to the wind power output, multiplied by the generation emission rate of CC plants. However, in reality some of these emission savings are reversed when the system commits additional CC or CT capacity to compensate for the risk of falling wind power, or dispatches inefficient CT capacity because insufficient CC capacity was committed in advance. Excess emissions, as a fraction of the expected emission savings due to wind power are calculated as follows:
þ νg, CT gCT, t þ νr, CC rCC, t þ νr, CT rCT, t ðLt Wt Þνg, CC νg, CC
≈
The second version of each equation is found using eq 6. We can also estimate the amount of generation capacity that is online but idle (not actively generating power) at each time t:
∑t ðνg, CT νg, CC ÞgCT, t
∑t Wt
þ νr, CC maxðgCT, t þ Δ40, t þ Δ20, t þ s60 , 0Þ þ νr, CT ðcCT, t gCT, t Þ
The second version of this equation is obtained via eqs 6, 9, 11 and 12, but with a simplification that ignores the emissions “savings” that could occur when the system fails to commit enough CC and CT capacity to satisfy the load unmet by wind. The full derivation of the second equation is shown in the Supporting Information. Noting the definitions of cCT,t and gCT,t in eqs 7 and 10, it is clear that in this formulation the excess emissions, E, depend only on the emission coefficients for each power plant in each mode (vg/r,CC/CT from Table 1), the operating reserve targets (s60, s20, and n60), the random forecast errors (Δt,40 and Δt,20), and the average power production from the wind farms in the region (ΣtWt/Σt1). In particular, this model does not depend on the specific level of load or wind at any time, because reserve targets are assumed to be constant at all times and the model ignores the emission consequences of (rarely) shedding load due to undercommitment of CC and CT or curtailing wind due to overcommitment of CC. Before proceeding, it should be noted that this model focuses on the reduction of emissions from the ongoing operation of power plants. It does not address the effect of wind power on the frequency of starts and stops for natural gas power plants. Although this model makes commit/decommit decisions for CC and CT capacity every minute, the actual operation of these plants does not vary sharply up and down. Generally, CC plants
νg, CC
∑t Wt
ð13Þ
are scheduled to follow the hour-to-hour changes in wind, and CT plants are only committed when there are unusually large declines in wind power output. This model also does not address the fact that power plants must be committed in discrete quantities—if the system needs an additional 50 MW committed at any point, then it must commit a full generator unit, which may be on the order of 200400 MW. On average, an extra half of a generator would be expected to be committed at any time. However, this overcommitment should be the same whether or not wind power is included in the system.
4. RESULTS I used the model described above to estimate the emissions from wind farms dispersed across regions of various sizes. This analysis proceeded as follows: (1) Define several sets of virtual wind farms whose forecast errors will be simulated. For this work, I use the locations and sizes of potential eastern-U.S. wind farms modeled in the Eastern Wind Integration and Transmission study.5 I perform the analysis nine times, using increasingly large subregions, as shown by the black rings in Figure 2. (2) Use the statistical model to simulate 107 instances of Δt,40 and Δt,20 for each of the modeled regions (equivalent to 19 years of wind errors, reported every minute). 9410
dx.doi.org/10.1021/es200417b |Environ. Sci. Technol. 2011, 45, 9405–9412
Environmental Science & Technology
Figure 2. Location of virtual wind farms used for this study. Blue dots indicate wind farm locations (gradated by size), and black rings show the subsets simulated for this study (regions with diameters of 100, 200, 500, 1000, ..., 3500 km).
Figure 3. (a) Reserves used to firm wind power and (b) and emissions due to these reserves, relative to direct emission savings of wind power (both for regions 1003500 km across).
(3) Set the nonspinning reserve margin n60 equal to 100% of wind capacity in the region; set the spinning reserve margins s60 and s20 equal to each other, and adjust them until the risk R for the region (eq 8) is 0.000009 (just below 0.00001). Setting s60 and s20 equal to each other generally achieves the lowest possible emissions for any level of risk, ensuring that enough CC generators are turned on 60 min ahead to meet the final 20-min spinning reserve margin (setting s60 higher than s20 tends to result in commitment of unneeded CC capacity, and setting it lower leads to commitment of lessefficient CT instead of CC). (4) Freeze s60 and s20 and adjust n60 down until the risk level R is 0.00001. This corresponds to about 50 min of generation shortfall every 10 years, which is similar to the reliability goals of existing power systems. This provides a full set of reserve targets, which can be used to assess the emissions that may occur as a result of dynamically firming up wind power with natural gas plants while maintaining a satisfactory risk level. (5) Estimate the excess emission rate E for each subregion using eq 13 with the simulated values from step 2 and the operating parameters from step 4.
ARTICLE
Figure 3(a) shows the nonspinning and spinning reserves allocated for each subregion. For power systems with wind farms dispersed across a 500 km or larger area, it should be possible to firm up wind with 10% or less spinning reserves and 20% or less nonspinning reserves (both relative to the total rated capacity of the wind farms). (These are both upper limits, based on the conservative assumptions described above). It is also possible to integrate wind using less nonspinning CT capacity if the 60-min spinning reserves target is raised, but this strategy will increase emissions somewhat. The optimal balance between spinning and nonspinning reserves depends on the amount of CT capacity already available and the relative cost of fuel and emissions vs CT capacity. It should also be noted that wind power is generally added to systems that already have enough conventional plants to meet their peak electricity loads, so there is no need to build additional reserves specifically to complement wind. However, wind development may be constrained in systems that have limited flexible generation resources (e.g., heavy dependence on coal or nuclear power). Figure 3(b) shows the emissions from the operation of reserve plants to compensate for the uncertainty of wind power forecasts. These are shown as a fraction relative to the emission savings that would occur if wind power directly displaced CC generation with no need for operating reserves. Alternatively, the plot can be interpreted as the net emission rate of firmed-up wind power when compared to a CC plant. When wind power is dispersed across regions larger than about 500 km across, 6% or less of their emission savings are reversed by the reserve power operations. This is much less than the 20% reported by Katzenstein and Apt 9 for full-time, 100% backup of individual wind farms. It should also be noted that this is an upper limit on the expected emissions; emissions will be lower if the uncertainty of wind forecasts is small compared to other uncertainties in the system, or if better wind forecasts are used, or if operating reserves are provided by lower-emission or faster-starting resources (e.g., hydropower or load response, which create no direct emissions).
5. DISCUSSION This work introduces a new model to estimate the range of possible errors in forecasts of wind power production in regions of various sizes. An example is also given of using this model with a simplified power system model to estimate the amount of greenhouse gas emissions that could be released when natural gas power plants are assigned to compensate for 99.999% of wind forecast errors. This power system model uses “worst-case” assumptions—that dedicated reserves are used for wind power separately from other sources of uncertainty in the system—and so it estimates an upper limit on the emissions that could occur. Emissions are likely to be significantly lower in power systems where wind is not the dominant source of uncertainty. Two key conclusions emerge from this work: (1) The larger the area over which wind farms are dispersed, the lower the burden is likely to be for firming the supply of wind power; and (2) For large regions (500 km or more across), the emissions due to firming of wind power are likely to be low, less than 6% of the emissions that would have been generated if CC 9411
dx.doi.org/10.1021/es200417b |Environ. Sci. Technol. 2011, 45, 9405–9412
Environmental Science & Technology plants had provided the power instead of wind farms. These emissions are due primarily to running excess CC plants (expending energy to keep idle capacity online), and to a much smaller degree to running excess CT plants or generating power from inefficient (but fast-starting) CT plants instead of more efficient CCGT plants. These conclusions imply that in large power systems, producing power in wind farms will cause at least 94% less CO2 emissions than producing the same amount of power with natural gas. However, to make best use of wind power, more flexible market rules may be needed to take advantage of the fastest-start capabilities of reserve power plants, e.g., rolling 5-min markets for reserves instead of hourahead or day-ahead. The later decisions can be delayed about committing spinning reserves, the fewer will need to be committed, and the lower costs and emissions will be. This work has not addressed several other options for providing operating reserves with potentially lower emissions than the approach studied here. Other possibilities may become more attractive as wind makes up a larger share of the power system: addition of superpeaking capability to CC or CT plants, so they can quickly provide more power without waiting for another plant to start up; taking advantage of the fact that CC and CT plants begin to supply power within a few minutes of startup instructions, well before they reach full power; short-term demand-response (e.g., cycling air conditioners or refrigerators off for a few minutes at a time),19 and zero-emission reserve sources such as hydro power, flywheels, or ultracapacitors. The unit commitment method presented in this work is mostly constrained by the start time of the fastest units in the system, so the addition of even small amounts of fast-start capacity could make a large difference. Finally, it should be noted that this model represented a simplified power system with natural gas capacity at least equal to the wind capacity. Different results are likely to be found in systems with different resource mixes: coal plants have higher emissions than gas but longer unit commitment delays; hydrobased systems can integrate wind more easily, but may have fewer emissions to avoid. This work has also assumed that ample transmission capacity is available to move reserves throughout the region; in reality there may be trade-offs between the cost of transmission upgrades, vs losses on long-distance lines, vs reductions in reserve requirements.
’ ASSOCIATED CONTENT
bS
Supporting Information. Details on the existing wind farms used to parametrize this model; a figure showing the rank correlation between wind farm forecast errors as a function of distance; additional validation figures and data for the wind forecast error model; a discussion of the idle- and generation-mode emission coefficients; a derivation of the excess emissions equation; and the Matlab code for the models. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail: [email protected].
ARTICLE
NextEra Energy Resources staff for wind farm data and industry insight, especially Skelly Holmbeck, Mark Mango, Mark Ahlstrom, Peter Wybierala, and Matt Schafer. NextEra Energy Resources funded this research, but the findings were not subject to review or approval by the company.
’ REFERENCES (1) CRA SPP WITF Wind Integration Study; Charles River Associates for Southwest Power Pool: Boston, MA, January 4, 2010. (2) NYISO Growing Wind: Final Report of the NYISO 2010 Wind Generation Study; New York Independent System Operator: Rensselaer, NY, September, 2010. (3) GE Energy Western Wind and Solar Integration Study; National Renewable Energy Laboratory: Golden, CO, May, 2010; p 535. (4) EWIS European Wind Integration Study; European Wind Integration Study: Brussels, March 31, 2010. (5) EnerNex Corporation Eastern Wind Integration and Transmission Study; National Renewable Energy Laboratory: Golden, CO, January, 2010; p 241. (6) Porter, K.; IAP Team Intermittency Analysis Project: Final report; CEC-5002007081; California Energy Commission: Sacramento, CA, July, 2007; p 71. (7) Holttinen, H.; Meibom, P.; Orths, A.; Hulle, F. v.; Lange, B.; O’Malley, M.; Pierik, J.; Ummels, B.; Tande, J. O.; Estanqueiro, A.; Matos, M.; Ricardo, J.; Gomez, E.; S€oder, L.; Strbac, G.; Shakoor, A.; Smith, J. C.; Milligan, M.; Ela, E. Design and operation of power systems with large amounts of wind power; Final report, IEA WIND Task 25, Phase one 20062008; VTT Tiedotteita Research Notes 2493; VTT Technical Research Centre of Finland: 2008. (8) Weber, C. Adequate intraday market design to enable the integration of wind energy into the European power systems. Energy Policy 2010, 38 (7), 3155–3163. (9) Katzenstein, W.; Apt, J. Air emissions due to wind and solar power. Environ. Sci. Technol. 2008, 43 (2), 253–258. (10) Mills, A.; Wiser, R.; Milligan, M.; O’Malley, M. Comment on “Air emissions due to wind and solar power.” Environ. Sci. Technol. 2009, 43 (15), 6106–6107. (11) Higham, N. J. Computing the nearest correlation matrix--a problem from finance. IMA J. Numer. Anal. 2002, 22 (3), 329–343. (12) Li, S. T.; Hammond, J. L. Generation of pseudorandom numbers with specified univariate distributions and correlation coefficients. IEEE Trans. Syst. Man Cybernet. 1975, SMC-5 (5), 557–561. (13) Wan, Y.-h. Summary Report of Wind Farm Data: September 2008; National Renewable Energy Laboratory: Golden, CO, May, 2009. (14) Katzenstein, W.; Apt, J. Air emissions due to wind and solar power: Supporting information. Environ. Sci. Technol. 2008, 43 (2), 253–258. (15) Kehlhofer, R.; Rukes, B.; Hannemann, F.; Stirnimann, F., Combined-Cycle Gas and Steam Turbine Power Plants. 3rd ed. ed.; PennWell Books: Tulsa, OK, 2009. (16) Angello, L. Advanced Monitoring to Improve Combustion Turbine/ Combined Cycle CT/(CC) Reliability, Availability and Maintainability (RAM); Semi-Annual Report Agreement Number DE-FC2601NT41233; Electric Power Research Institute: Palo Alto, CA, April 1September 30, 2004. (17) EIA Assumptions to the Annual Energy Outlook 2010; DOE/EIA0554(2010); Energy Information Administration, U.S. Department of Energy: Washington, DC, April 9, 2010. (18) Kim, T. S. Comparative analysis on the part load performance of combined cycle plants considering design performance and power control strategy. Energy 2004, 29 (1), 71–85. (19) Kirby, B. J. Spinning Reserve from Responsive Loads; ORNL/ TM-2003/19; Oak Ridge National Laboratory: Oak Ridge, TN, March, 2003.
’ ACKNOWLEDGMENT Thanks to Mike Slattery, Nick Eyre, and three anonymous reviewers for insightful questions and advice. Thanks to several 9412
dx.doi.org/10.1021/es200417b |Environ. Sci. Technol. 2011, 45, 9405–9412
ARTICLE pubs.acs.org/est
Carbon Sequestration Kinetic and Storage Capacity of Ultramafic Mining Waste Julie Pronost,† Georges Beaudoin,*,† Joniel Tremblay,† Faïc-al Larachi,‡ Josee Duchesne,† Rejean Hebert,† and Marc Constantin† †
Departement de geologie et genie geologique & GEOTOP, Universite Laval, 1065 avenue de la Medecine, Quebec (QC), Canada G1V 0A6 ‡ Departement de genie chimique, Universite Laval, 1065 avenue de la Medecine, Quebec (QC), Canada G1V 0A6
bS Supporting Information ABSTRACT: Mineral carbonation of ultramafic rocks provides an environmentally safe and permanent solution for CO2 sequestration. In order to assess the carbonation potential of ultramafic waste material produced by industrial processing, we designed a laboratory-scale method, using a modified eudiometer, to measure continuous CO2 consumption in samples at atmospheric pressure and near ambient temperature. The eudiometer allows monitoring the CO2 partial pressure during mineral carbonation reactions. The maximum amount of carbonation and the reaction rate of different samples were measured in a range of experimental conditions: humidity from dry to submerged, temperatures of 21 and 33 °C, and the proportion of CO2 in the air from 4.4 to 33.6 mol %. The most reactive samples contained ca. 8 wt % CO2 after carbonation. The modal proportion of brucite in the mining residue is the main parameter determining maximum storage capacity of CO2. The reaction rate depends primarily on the proportion of CO2 in the gas mixture and secondarily on parameters controlling the diffusion of CO2 in the sample, such as relative saturation of water in pore space. Nesquehonite was the dominant carbonate for reactions at 21 °C, whereas dypingite was most common at 33 °C.
’ INTRODUCTION The increasing concentration of greenhouse gases in the atmosphere, such as methane and carbon dioxide, has led to the development of several mitigation strategies to reduce anthropogenic impact on climate.1 Sequestration of CO2 by reaction with Ca or Mg-rich natural minerals has been suggested as an environmentally safe and permanent method for storage of CO2.26 Due to their high content in Ca and Mg, mafic and ultramafic rocks are the most reactive rocks for CO2 capture and storage. They are found in the Earth’s crust as greenstone belts, ophiolites, volcanic and intrusive rocks, and they are abundant enough to potentially store the carbon that would be produced by combustion of the world’s known coal reserves.7 Carbonation of these rocks is a thermodynamically favored exothermic reaction that occurs naturally during weathering and which exerts a first order control of atmospheric CO2 concentration over geological time scales.8 However, the kinetics of the reaction at Earth’s surface conditions is too slow to be suitable for industrial processes. Several studies showed that the reactivity of olivine and serpentine is enhanced by thermal and mechanical activation,4,9,10 but this treatment is energy intensive and would lead to storage cost of ∼54 USD/tonne of CO2, which is not economically viable.11 Chemical activation using acids and bases was also studied.12 It should however be noted that, even r 2011 American Chemical Society
if mineral carbonation is a costly process, it has benefits compared to storage in geological formations because the carbon is stored in stable, environmentally benign minerals that do not require long-term monitoring. As the price of open-pit mining and crushing for ultramafic rocks is low (45 USD/ton13), partial carbonation at ambient conditions could prove to be a cost-effective option compared to complete carbonation at high temperature and pressure, to offset the required energy intensive or chemical pretreatment with associated environmental and capital costs. Olivine, though less naturally available, displays high reaction rates,14,15 but the preferential mobilization of Mg in aqueous solution leads to the formation of a silica-rich passivating layer coating the olivine grains thus hindering the carbonation reaction.1618 Experiments of gassolid carbonation of chrysotile at atmospheric pressure show that the atoms of Mg that react are those from the outside, brucitic, layer of the chrysotile structure that are not bound to Si atoms, which remain unreacted in the chrysotile lattice.19
Received: May 17, 2011 Accepted: September 15, 2011 Published: September 15, 2011 9413
dx.doi.org/10.1021/es203063a | Environ. Sci. Technol. 2011, 45, 9413–9420
Environmental Science & Technology Unlike previous experimental studies of mineral carbonation, that were mostly performed on pure phases such as olivine,4,14,18,20 serpentine,21,22 orthopyroxene,23 brucite,10,24 the present study assesses the mineral carbonation potential of ultramafic mining residues from two different deposits: 1) samples from various stages mineral processing pilot tests from the Dumont Nickel deposit (Amos, Canada) and 2) samples from chrysotile milling residue from the Black Lake mine (Thetford Mines, Canada). The surface of ultramafic mining waste piles has been shown to undergo passive mineral carbonation by reaction with meteoric water and atmospheric CO2.26,27 The mineral carbonation reaction also occurs within the mining waste piles as shown by warm, CO2-depleted, air venting at the surface of the piles.28 The estimation of CO2 uptake by analysis of the carbonated product after reaction does not yield insight about the evolution of the reaction rate and dynamics throughout the experiment. In previous studies, real time monitoring has been performed using in situ synchrotron X-ray diffraction (XRD) and Raman spectrometry,10,25 running a set of parallel or sequential experiments that are stopped at different stages of the reaction.11,24 Here we present an original method that allows real time monitoring of CO2 consumption by means of a modified eudiometer, in order to determine the critical parameters controlling the carbonation of ultramafic residues and the kinetics of the reaction. Based on manometric principles, the technique enables following the carbonation kinetics of several samples at the same time via the CO2 partial pressure decrease over periods of several weeks. Samples of ultramafic mining waste from the two sites were studied under variable conditions of CO2 partial pressure, sample size, relative humidity, and temperature, to determine optimal conditions for mineral carbonation. The consumption of gaseous CO2, the amount of C captured, and the mineralogy of the carbonate mineral products are used to understand the reaction and derive the rate of the mineral carbonation reaction under various experimental conditions.
’ ANALYTICAL METHODS All the samples have been characterized before and after carbonation. XRD analyses have been performed at Universite Laval, using a Siemens D5000 diffractometer with Cu Kα radiation. Scans were taken for 2θ ranging from 1° to 65° with steps of 0.02°/s. A JEOL-840A scanning electron microscope (SEM) equipped with energy dispersive X-ray spectroscopy (EDS) was used for imaging and semiquantitative major element analysis. Whole-rock major elements were analyzed by X-ray fluorescence (XRF) by Activation Laboratories (Ancaster Canada). For all the oxides the detection limit is 0.01 wt % except for MnO (0.001 wt %). Analysis of a serpentinite reference material UB-N indicates that accuracy is better than 2% for elements with concentrations higher than 1 wt %. Carbon content was measured by Activation Laboratories (Ancaster Canada) using standard infrared (IR) and IR low-level carbon (LLC) instruments. In both cases the analyses indicate the total carbon content, without any discrimination between inorganic or organic carbon. Low-level carbon analysis allows a detection limit of 0.004 wt % CO2 and is suitable for samples with less than 1 wt % CO2. Reproducibility has been tested by analyzing selected samples in triplicate. The SRM UB-N (0.39% ( 0.08 wt % CO2) yielded a LLC average of 0.5 (
ARTICLE
Figure 1. Schematic representation of the experimental device. (A, B) At instant time t of test (reactive sample) and control (corundum sample) eudiometers: instant atmospheric pressure, Pa(t), instant test and control headspace volumes, Vt(t), Vc(t), and corresponding rising heights of glycerol, dt(t), dc(t). (C) (Identical) initial states in test and control eudiometers: atmospheric pressure, Pa(0), test and control headspace volume, V0, and rising height of glycerol, d0.
0.09 wt % CO2, which indicates good accuracy and precision. Standard infrared analysis has a detection limit of 0.04 wt % CO2 and is appropriate for samples containing more than 1 wt % CO2. Reproducibility of randomly selected duplicates is better than 2%. Analysis of SRM SY-4 (3.5 ( 0.1 wt % CO2) yielded an average of 3.9 ( 0.05 wt % CO2, indicating good precision and accuracy. Amos samples were prepared by wet sorting in a pilot plant. For each sample, an aliquot of process water has been filtered in order to remove fine particles in suspension and has been analyzed for major elements by Exova (Quebec) by inductively coupled plasma atomic emission spectrometry (ICP-AES). The detection limit is between 0.5 and 0.001 ppm depending on the element. Two standard solutions were analyzed at the beginning and at the end of each analytical series and the reproducibility is better than 3.3% and the accuracy better than (10%. The acidity of the process water has also been measured with a pHmeter (Model 415, Denver Instrument) calibrated before each use with three buffer solutions with pH of 4, 7, and 12.
’ EXPERIMENTAL PROCEDURE Protocol for Carbonation Monitoring. The CO2 instantaneous uptake by Mg-rich milling waste material is monitored using a controlled temperature volumetric technique. The experiments are set in a controlled-temperature room at 20.7 ( 0.7 °C or 33.3 ( 2.7 °C. The eudiometer is a device designed to measure gas volumes,29,30 which has been modified to measure CO2 uptake during periods of time up to three weeks. The setup is sketched in Figure 1, which highlights some of the variables influencing the carbonation reaction monitoring. All the glassware is PYREX. The test (reactive sample) and control (corundum sample) eudiometers comprise a small beaker, and a 500 mL graduated cylinder lowered on it until its open-end, immersed in a glycerol container, delineates a reaction headspace volume (Figure 1A,B). The air trapped in both eudiometers is partially pumped out, and a determined amount of CO2 is injected until the air mixture reaches a desired composition. The test and control beakers contain weighted samples (typically 5 g) of, respectively, milling waste rock and inert corundum that were humidified with distilled water. 9414
dx.doi.org/10.1021/es203063a |Environ. Sci. Technol. 2011, 45, 9413–9420
Environmental Science & Technology
ARTICLE
Figure 2. Volume change in eudiometers during carbonation experiments. Dashed line marks a week. A) Measured atmospheric pressure. B) Gas volume change in control eudiometer and in test eudiometers for four samples entirely submerged in distilled water and one sample partially saturated with distilled water. C) Gas volume of reacted CO2 after correction for atmospheric pressure. D) Quantity (mmole, filled symbol) of reacted CO2, for several CO2 injections inside the eudiometer. The molar fraction of CO2 (open symbol) records CO2 consumption and vertical steps show time of injection.
Instantaneous gas volumes, Vt(t) and Vc(t), and atmospheric pressure, Pa(t), are registered during the course of carbonation experiments. However, processes other than CO2 consumption can affect the gas volumes in the device. CO2 leakage via physical absorption and then desorption through glycerol is marginal.30 Also, dynamic changes in the eudiometers are very slow so that a hydrostatic correction of the headspace pressures is reasonably accurate. Hence, spurious deviations in volumes are likely to result only from fluctuations of atmospheric pressure. They are captured in the volumes measured as Vc(t) (Figure 1B), for a fixed surface A equal to the graduated tube cross-section, which deviate from the initial state (V0) according to the following expression (symbols are explained in the caption of Figure 1), based on hydrostatic and volume conservation principles A V0 þ δ0 Vc ðtÞ ¼ Pa ðtÞ Fg A 2Fg ffi# sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 V0 V0 þ δ0 þ 4Fg ðPa ð0ÞFgδ0Þ þ Pa ðtÞ Fg A A
The control eudiometer allows systematic compensation of the external pressure disturbances over the extended observational periods characterizing the measurements in the test eudiometer (Figure 2). Consequently, the running gas volumes, Vt(t), in the test eudiometers can be synchronously corrected to reflect solely the actual conversion of carbon dioxide which is given by X ¼
Vc ðtÞ Vt ðtÞ V0 Fg ðVc ðtÞ þ Vt ðtÞÞ þ δ0 þ Pa ðtÞ Fg RTnc0 A A
The method’s sensitivity is such that volume changes as lows as 1 mL can be measured, which correspond to 1.8 mg (or 41 μmol) of CO2 captured during carbonation. After consumption of the initial CO2 load (nc0), known CO2 amounts can be reinjected using a glass tube connected to a gas cylinder. To assess the materials maximum carbonation capacity, replenishment of CO2 can be repeated until carbonation of the solid sample ceased (Figure 2D). The maximum carbonation capacity accounts only for the added carbon after the materials are carbonated in the eudiometer regardless of their content in 9415
dx.doi.org/10.1021/es203063a |Environ. Sci. Technol. 2011, 45, 9413–9420
Environmental Science & Technology native carbonates, a clear advantage over other methods, such as TGA that measure the total carbonate content regardless of their origin. Samples Characterization. The milling residues from Thetford Mines have been sampled on the mining heap and have been crushed to 1 mm (80% passing). XRD analyses indicated the presence of serpentine (dominant chrysotile + lizardite) as well as accessory brucite. Magnetite, chromite, chlorite, and phlogopite have been detected in small amounts. Some fine grained deposits observed on the heap display a small concentration of hydromagnesite, but our samples, sampled at the head of the conveyor from the processing plant, were unweathered. The BET specific surface area of Thetford Mines samples ranged from 5.17 to 10.17 m2/g (n = 10). The composition of Thetford Mines sample is very close to the theoretical composition of pure serpentine with Mg# (Mg/(Fe+Mg)) of 95 (Supporting Information Figure S1). The slight deviation toward higher MgO and loss of ignition (LOI) and lower SiO2 can be ascribed to the presence of brucite. The proportion of brucite over serpentine can be estimated from the relative amount of SiO2 and MgO. Using this method, the estimated amount of brucite is low (1.8 wt %), and it contains less than 3 wt % of the total MgO of the samples. The Amos samples were taken at three steps of the pilot plant concentration process. The fraction produced by the defibering process (fluff) has a grain size of 20 mesh (80% under 840 μm), the fraction produced after desliming (slimes) has a grain size of 100 mesh (80% under 150 μm), and the fraction left after flotation (final tail) has a grain size of 150 mesh (80% under 40 μm). The BET specific surface area for samples from the processing steps were similar, ranging from 9.03 to 11.49 m2/g. XRD patterns of the 12 samples reveal that they are mostly composed of serpentine (chrysotile and lizardite) and brucite. Chlorite and magnetite occur as minor components. The Amos samples plot on a mixing line between serpentine Mg#86 and brucite (Supporting Information Figure S1). The relative proportions of these phases are calculated from the relative amounts of SiO2 and MgO, which yields proportion of brucite ranging from 10 to 15 wt %, such that brucite contains between 18 and 26 wt % of the total MgO of the samples. Spontaneous formation of coalingite (Mg10Fe3+2(CO3)(OH)24 3 2H2O) has been observed on exploration drill core in Amos. The amount of CO2 already present in the samples prior to experimental carbonation was measured by LLC. The fluff fraction of each sample has the lowest carbon content (av. 0.30 wt % CO2), the slimes carbon contents are slightly higher (av. 0.49 wt % CO2), and the final tail fractions contain markedly more carbon (av. 0.88 wt % CO2). In addition to spontaneous carbonation during sorting, it is possible that a small amount of carbonates were present in the rocks before processing and became relatively enriched in the final tail fractions as a consequence of the sorting process. The final tail and slime fraction samples are separated by a wet sorting process and were delivered submerged in water. Acidity measurements for the four final tail fractions fall in a narrow range of basic pH, from 9.61 to 9.66. Process water from three of the four slime fractions have pH in the range 8.38 to 9.17, whereas sample S197B is more acidic (6.18), presumably because of the dissolution of sulfides. Chemical analyses reveal that, among the 29 analyzed elements, only Al, Ca, Mg, K, Si, and Na have significantly higher contents than detection level in all samples. The process water of slime samples tends to have higher content of dissolved ions. The relationship is particularly notable for Si and Fe. However, it should be noted that the concentration
ARTICLE
Figure 3. Comparison of CO2 content in carbonated samples according to eudiometer observations and infrared analysis. Here, only the samples that have reached their ultimate carbonation capacity are shown. Legend displays brucite content of Thetford Mines and Amos in wt%. Submerged samples contain 5992 wt % H2O, whereas the wet samples contain 1652 wt % H2O.
in dissolved Mg, comprised between 16 and 31 ppm, is unaffected by the pH.
’ RESULTS AND DISCUSSION Eudiometer-Monitoring of Carbonation Kinetics. The simple construction of the eudiometers from Pyrex glass, without joints or valves, ensures that no leakage occurs in the instrument over periods of weeks, as tested with empty eudiometers. Tests were carried out by loading the test eudiometer beaker with water only (distilled water or process water) and with dry samples. These experiments yielded undetectable, to low CO2 uptake (0 to 0.34 wt % CO2 in the final product). These tests demonstrate that CO2 is not diffusing through glycerol during the duration of the experiments. Dried solid samples were humidified with distilled water (31 to 92 wt % H2O). Samples with more than 6070 wt % H2O formed slurries where the sample was totally submerged. Atmospheric pressure variation (Figure 2A) causes volume change in the control eudiometer and test eudiometers (Figure 2B), which allows computing the volume change related to reaction with a sample as in Figure 2C. Figure 2D shows an example where a sample reacts with a series of batch of CO2, injected into the eudiometer after either the CO2 was entirely consumed or after the sample became unreactive with respect to CO2. The maximum carbon capture capacity is reached when injection of new CO2 produces no measurable reaction. Characterization of Reaction Products. The samples that have reached maximum capacity after repeated carbonation experiments were dried, and their CO2 content was measured by LLC analysis. In Figure 3, the results of these analyses are compared with the amount of consumed carbon calculated from measurement with the eudiometers. The LLC data were corrected after subtracting the contribution from carbon pre-existing in the samples (0.30 to 0.88% CO2). The two methods yield consistent results, albeit somewhat higher CO2 values are obtained using the eudiometer, likely a result of subtraction of initial carbon from the LLC measurement. XRD analyses identified nesquehonite (Mg(HCO3)(OH) 3 2H2O) as the dominant product of carbonation for the experimental series carried out at 21 °C (Supporting Information 9416
dx.doi.org/10.1021/es203063a |Environ. Sci. Technol. 2011, 45, 9413–9420
Environmental Science & Technology
Figure 4. Secondary electron photomicrographs of carbonate minerals formed by mineral carbonation: (a) numerous flakes, one of them being pointed by an arrow, have grown within a network of chrysotile fibers, (b) flat agglomerate that has probably developed between two grains, (c) elongated crystal in the typical habit of nesquehonite,31 and (d) massive crystal with cracks probably from dehydration after sample removal or in the SEM.
Figure S2), whereas dypingite (Mg5(CO3)4(OH)2 3 5H2O) prevails in samples carbonated at 33 °C. Nesquehonite has previously been identified as the most common product of brucite carbonation under high CO2 pressure.24 SEM observation in secondary electron mode and EDS analysis reveal different carbonates characterized by flakes, agglomerates, massive crystals, and well crystallized prisms (Figure 4). It should be noted that carbonates cover only a small fraction of the total grain surface and that there is no obvious control by sample minerals or habitus. Hence product coating of the reactive surface is not a limiting feature for the reaction. Stability of the carbonated phases has been tested by leaving some samples, partially saturated, in their eudiometers during 18 days after they reached their maximum carbonation capacity. These samples did not release or consume any CO2. Six other postcarbonated samples were kept at atmospheric conditions during 6 to 9 months. XRD analyses indicated that all the samples contained the same carbonates that they did at the end of the experiment. As hydrated carbonates are metastable,32 the carbonation reaction in eudiometers appears to be kinetically controlled.24
’ PARAMETERS DETERMINING THE TOTAL AMOUNT OF CARBONATION AND THE REACTION RATE The content of brucite entwined with serpentine appears to be the major parameter controlling the maximum amount of carbon a sample can store (Figure 3). The brucite-rich Amos milling residues has maximum carbonate uptake up to 9 wt % CO2 (8.6% CO2 by IR), whereas it barely reaches 2 wt % in the brucite-poor samples from Thetford Mines. The preferential consumption of brucite is confirmed by XRD analyses (Supporting Information Figure S2). Out of the 42 carbonated samples from Amos, brucite is not detected in 35 of the samples after reaction though it was a major component of all the samples before the experiments. The percentage of Mg that has been mobilized in the formation of
ARTICLE
carbonates is less than 2 wt % for Thetford Mines samples and is typically in the range of 1520 wt % for Amos samples. In both cases, it is less or equal to the amount of Mg stored in brucite. It is thus possible that serpentine has not been involved in the reaction, or at least, this might have been to a very limited extent. There is no influence of the grain size (0.04 to 0.84 mm) among the Amos samples, which have similar specific surface area (911 m2/g), but the coarser grain size of the Thetford Mines samples (1 mm) and their lower specific surface area (510 m2/g) is perhaps one cause for their lower maximum carbon uptake, in addition to their lower content in brucite. The reaction of CO2 with pure brucite and pure chrysotile is shown in Figure 5A. In that experiment, both brucite and chrysotile initially react at a similar rate, but reaction with chrysotile slows after ∼2 h whereas brucite captures 2.8 times more CO2 until it stops reacting after ∼15 h. Brucite has depleted CO2 from the gas mixture, whereas chrysotile is in contact with a gas mixture with 25% CO2. Experimental data of reacted CO2 have been fitted to appropriate polynomial or exponential equations to study reaction rates for the first injection stage. The derivative yields the reaction rate in the course of carbonation and is used to estimate the influence of different parameters controlling the reaction. The initial reaction rate of chrysotile in the presence of high concentration of CO2 (3334 mol %) is near 2.1 mmol CO2/h, and the rate of reaction decreases linearly with the concentration of CO2 in the gas mixture, except at low reaction rates (Figure 5A). The reaction rate of chrysotile decreases rapidly as it becomes unreactive after 5 h (Figure 5A). Brucite has a lower initial reaction rate (∼1.5 mmol CO2/h), but the rate of reaction displays a shallow slope with concentration of CO2 (Figure 5B). The control of the CO2 concentration in the gas mixture is further shown by reaction of the fluff fraction of sample 184F (Figure 6A), which shows that the initial reaction rate is proportional to the initial concentration of CO2 in the gas mixture. Figure 6A also shows that the sample had a slightly lower rate of reaction at a lower temperature (21 vs 33 °C), but in our test conditions, temperature (21 and 33 °C) has no significant effect on the reaction rate. The maximum rate of reaction of the samples is about 0.14 mmol CO2/h, and the rate displays a linear decrease with CO2 content in the gas mixture (Figure 6B). The initial reaction rates define a near unity partial order for CO2 (Figure 6A). In general, samples that were submerged displayed lower reaction rates in comparison to unsaturated samples, likely because of slower diffusion of CO2 in the water layer above the sample surface and in the pore space water of saturated samples. In unsaturated samples, CO2 diffuses in pore space filled with the gas mixture, and where water is wetting the grains surfaces. Two of the submerged samples display high reaction rates. One of them contained 60 wt % H2O and was barely submerged, whereas the other contained 78 wt % H2O. The potential causes for the higher reaction rates are unknown, but we speculate that undetected trace carbonate minerals acted as seeds for crystallization. A lower rate of reaction is observed for larger amounts of solid material (Supporting Information Table S3). Because the sample holders have the same cross-section area, the weight of the sample determines the height of the sample in the holder. The rate of reaction is not affected by the sample weight because of the fixed surface area of the sample exposed to the gas mixture. The sample grains closer to the upper surface react first, sometimes forming a cemented crust, whereas particles deeper in the sample holder undergo a delay induced by the diffusion of CO2 in the pore space. 9417
dx.doi.org/10.1021/es203063a |Environ. Sci. Technol. 2011, 45, 9413–9420
Environmental Science & Technology
ARTICLE
Figure 5. A) Comparison of the quantity of CO2 (mmole, filled symbols) reacted with brucite (red) and chrysotile (green) as a function of time (h). Reaction of CO2 with the samples decreases the molar fraction of CO2 (open symbols) in the gas mixture. B) Rate of reaction of brucite and chrysotile versus the CO2 content of the gas mixture. Experimental data are superposed on the exponential fit curve.
The CO2 concentrations dealt with in the eudiometer kinetic tests by far outweigh those typically encountered under field conditions. In addition, the time scale for conducting the reaction experiments lasted up to 17 days (Figure 2). Under the assumption of stagnant gas conditions (Sherwood number Shgas = 2 and DCO2 ca. 106 m2/s), the gas-side mass transfer coefficients would be of the order of 105 m/s. This is tantamount to the lower-limit gas-side mass transfer coefficient since, in reality, the atmospheric pressure fluctuations act as gentle convective mixers helping to achieve faster homogenization of the CO2 composition in the eudiometer headspace. Therefore, it is highly unlikely that the eudiometer-measured chemical responses will be prone to mass transfer retardation effects from the gas phase, and one can safely assume the measurements reflect true intrinsic gassolid kinetics.
The results of these experiments are in accord with gassolid carbonation experiments which indicate that CO2 reacts preferentially with the Mg(OH)2, brucitic, layer of the chrysotile structure, leaving the internal silica-rich layers largely unreacted.19 Figure 5 is interpreted to show initial reaction of CO2 with the external brucitic layer of chrysotile after which, the chrysotile reacts slowly whereas brucite reaction depletes the gas mixture in CO2, such that the lower concentration of CO2 slows the reaction until new CO2 is injected (Figure 2D). The initial higher rate of reaction for chrysotile is likely related to the its specific surface areas (∼14.4 m2/g, 19) compared to that of brucite (0.2 m2/g, 50100 um size fraction33). The Amos samples are estimated to contain 1015% brucite (Supporting Information Figure S2), and the fluff samples of Figure 6 display a rate of reaction approximately 10% of that of pure brucite 9418
dx.doi.org/10.1021/es203063a |Environ. Sci. Technol. 2011, 45, 9413–9420
Environmental Science & Technology
ARTICLE
Figure 6. Experimental results for unsaturated fluff fractions of sample 184F with ∼50 wt % H2O at 21 and 33 °C. A) Rate of reaction versus time for different CO2 contents in the gas mixture and at temperatures of 21 and 33 °C. The CO2 partial order (n) is based on the initial reaction rates. B) Rate of reaction versus the CO2 content of the gas mixture.
(Figure 5), suggesting that the rate of reaction in the mine waste samples studied is largely controlled by the abundance of brucite in the sample. In southern Quebec only, the amount of ultramafic milling and mining residues stored in heaps from chrysotile mining is estimated to 2 Gt, which could store up to ca. 700 Mt C.30 New mining projects, such as Dumont Nickel near Amos, Canada, can estimate the potential for CO2 capture and storage in mine waste to offset their expected greenhouse gas emissions. The worldwide amount of variably serpentinized peridotites exceeds the
requirement to store the excess of atmospheric CO2, including future emissions.4 As the carbonation potential of ultramafic material is variable and depends on the brucite content, tests in eudiometers can help identify the most reactive materials and determine the optimal conditions for carbonation.
’ ASSOCIATED CONTENT
bS
Supporting Information. Tables providing whole-rock chemical composition of samples, pH of process water, and
9419
dx.doi.org/10.1021/es203063a |Environ. Sci. Technol. 2011, 45, 9413–9420
Environmental Science & Technology parameters and results of carbonation experiments. Also included are additional figures showing chemical and experimental data. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 1-418-656-3141. Fax: 1-418-656-7339. E-mail: beaudoin@ ggl.ulaval.ca.
’ ACKNOWLEDGMENT This research has been funded by a Natural Science and Engineering Research Council of Canada Discovery grant to G. Beaudoin and by Royal Nickel Corporation. M. Plante is gratefully acknowledged for his help with experiments and ingenuity solving experimental problems. ’ REFERENCES (1) Intergovernmental Panel on Climate Change 2007 Synthesis Report of the IPCC Fourth Assessment Report: Summary for Policy Makers. 2007. Available at http://www.ipcc.ch/pdf/assessment-report/ ar4/syr/ar4_syr_spm.pdf (accessed February 5, 2009). (2) Seifritz, W. CO2 disposal by means of silicates. Nature (London, U.K.) 1990, 345, 486. (3) IPCC Special Report on Carbon Dioxide Capture and Storage; Metz, B., Davidson, O., de Coninck, H., Loos, M., Meyer, L., Eds.; Cambridge University Press: New York, 2005. (4) Lackner, K.; Wendt, C.; Butt, D.; Joyce, E.; Sharp, D. Carbon dioxide disposal in carbonate minerals. Energy 1995, 20, 4802. (5) Lackner, K. S. A guide to CO2 sequestration. Science (Washington, DC, U.S.) 2003, 300, 1677–1678. (6) Sundquist, E. T. The global carbon dioxide budget. Science (Washington, DC, U.S.) 1993, 259, 934–941. (7) Goldberg, P.; Chen, Z. Y.; O’Connor, W; Walters, R.; Ziock, H. CO2 Mineral Sequestration Studies. Presented at First National Conference on Carbon Sequestration, Washington, DC, May 1417, 2001. (8) Berner, R. A.; Kothavala, Z. GEOCARB III: a revised model of atmospheric CO2 over Phanerozoic time. Am. J. Sci. 2001, 301, 182–204. (9) O’Connor, W. K.; Dahlin, D. C.; Rush, G. E.; Gerdemann, S. J.; Penner, L. R. Energy and economic consideration for ex situ aqueous mineral carbonation. In Proceeding of the 29th International Technical Conference on Coal Utilization & Fuel System 2004, 71. (10) McKelvy, M. J.; Chizmeshya, A. V. G.; Diefenbahcer, J.; Bearat, H.; Wolf, G. Exploration of the role of heat activation in enhancing serpentine carbon sequestration reactions. Environ. Sci. Technol. 2004, 38, 6897–6903. (11) Gerdemann, S. J.; O’Connor, W. K.; Dahlin, D. C.; Penner, L. R.; Rush, G. E. Ex situ aqueous mineral carbonation. Environ. Sci. Technol. 2007, 19, 95–101. (12) Maroto-Valer, M. M.; Fauth, D. J.; Kuchta, M. E.; Zhang, Y.; Andresen, J. M. Activation of magnesium rich minerals as carbonation feedstock material for CO2 sequestration. Fuel Process. Technol. 2005, 86, 1627. (13) O’Connor, W. K.; Walters, R. P.; Dahlin, D. C.; Rush, G. E.; Nilsen, D. N.; Turner, In Proceedings of the 26th International Technical Conference on Coal Utilization & Fuel Systems 2001, 765. (14) H€anchen, M.; Prigiobbe, V.; Storti, G.; Seward, T. M.; Mazzotti, M. Dissolution kinetics of fosteritic olivine at 90150°C including effects of the presence of CO2. Geochim. Cosmochim. Acta 2006, 70, 4403–4416. (15) H€anchen, M.; Prigiobbe, V.; Baciocchi, R.; Mazzotti, M. Precipitation in the Mg-carbonate system - effects of temperature and CO2 pressure. Chem. Eng. Sci. 2008, 63, 1012–1028.
ARTICLE
(16) Bearat, H.; McKelvy, M. J.; Chizmeshya, A. V. G.; Gormley, D.; Nunez, R.; Carpenter, R. W.; Squires, K.; Wolf, G. H. Carbon sequestration via aqueous olivine mineral carbonation: role of passivating layer formation. Environ. Sci. Technol. 2006, 40, 4802–4808. (17) Andreani, M.; Luquot, L.; Gouze, P.; Godard, M.; Hoise, E.; Gibert, B. Experimental study of carbon sequestration reactions controlled by the percolation of CO2-rich brine through peridotites. Environ. Sci. Technol. 2009, 43, 1226–1231. (18) Jarvis, K.; Carpenter, R. W.; Windman, T.; Kim, Y.; Nunez, R.; Alawneh, F. Reaction mechanisms for enhancing mineral sequestration of CO2. Environ. Sci. Technol. 2009, 43, 6314–6319. (19) Larachi, F.; Daldoul, I.; Beaudoin, G. Fixation of CO2 by chrysotile in low-pressure dry and moist carbonation: Ex-situ and insitu characterizations. Geochim. Cosmochim. Acta 2010, 74, 3051–3075. (20) Prigiobbe, V.; H€anchen, M.; Werner, M.; Baciocchi, R.; Mazzotti, M. Mineral carbonation process for CO2 sequestration. Energy Procedia 2009, 1, 4885–4890. (21) Fagerlund, J.; Teir, S.; Nduagu, E.; Zevenhoven, R. Carbonation of magnesium silicate mineral using a pressurised gas/solid process. Energy Procedia 2009, 1, 4907–4914. (22) Teir, S.; Eloneva, S.; Fogelholm, C.-J.; Zevenhoven, R. Fixation of carbon dioxide by producing hydromagnesite from serpentinite. Appl. Energy 2009, 86, 214–218. (23) Dufaut, F.; Martinez, I.; Shilobreeva, S. Experimental study of Mg-rich silicates carbonation at 400 and 500 °C and 1 kbar. Chem. Geol. 2009, 265, 79–87. (24) Zhao, L.; Sang, L.; Chen, J.; Ji, J.; Teng, H. H. Aqueous carbonation of natural brucite: relevance to CO2 sequestration. Environ. Sci. Technol. 2010, 44, 406–411. (25) Wolf, G. H.; Chizmeshya, A. V. G.; Diefenbahcer, J.; McKelvy, M. J. In situ observation of CO2 sequestration reactions using a novel microreaction system. Environ. Sci. Technol. 2004, 38, 932–936. (26) Beaudoin, G.; Hebert, R.; Constantin, M.; Duchesne, J.; Cecchi E.; Huot, F.; Vigneau, S.; Fiola, R. Spontaneous carbonation of serpentine in milling and mining waste, southern Quebec and Italy. In Proceedings of Accelerated Carbonation for Environmental and Materials Engineering (ACEME2008) 2008, Rome, Italy, pp 7382. (27) Wilson, S. A.; Dipple, G. M.; Power, I. M.; Thom, J. M.; Anderson, R. G.; Raudsepp, M.; Gabites, J. E. ; Southam, G. Carbon Dioxide Fixation within Mine Wastes of Ultramafic-Hosted Ore Deposits: Examples from the Clinton Creek and Cassiar Chrysotile Deposits, Canada: Economic Geology, 2009, Vol. 104, pp 95-112. (28) Beaudoin, G.; Pronost, J.; Marcouiller, S.; Hebert, R.; Constantin, M.; Duchesne, J.; Lemieux, J.-M.; Molson, J. W.; Larachi, F.; Klein, M.; Maldague, X. First discovery of CO2-depleted warm air vents in chrysotile milling waste: surface evidence for natural carbon sequestration at depth. GSA Annual meeting Program with Abstracts 2010, 425, #178136. (29) Landriani, M. Ricerche fisiche intorno alla salubrita dell’aria; Milan, Italy, 1775. (30) Maries, A. Measurement of gas consumption during accelerated carbonation of Portland cement mortar. In Proceedings of Accelerated Carbonation for Environmental and Materials Engineering (ACEME2008) 2008, Rome, Italy, pp 131-138. (31) Genth, F. A.; Penfield, S. L. On lansfordite, nesquehonite, a new mineral, and pseudomorphs of nesquehonite after lansfordite. Am. J. Sci. 1890, 39 (230), 121–137. (32) Marini, L. Geological Sequestration of Carbon Dioxide, Vol. 11: Thermodynamics, Kinetics, and Reaction Path Modeling; Elsevier (Developments in Geochemistry): Amsterdam, The Netherlands, 2007. (33) Pokrovsky, O. S.; Schott, J. Experimental study of brucite dissolution and precipitation in aqueous solutions: Surface speciation and chemical affinity control. Geochim. Cosmochim. Acta 2004, 68, 31–475.
9420
dx.doi.org/10.1021/es203063a |Environ. Sci. Technol. 2011, 45, 9413–9420
ARTICLE pubs.acs.org/est
Experimental Study on the Reuse of Spent Rapidly Hydrated Sorbent for Circulating Fluidized Bed Flue Gas Desulfurization Yuan Li, Kai Zheng, and Changfu You* Key Laboratory for Thermal Science and Power Engineering of Ministry of Education, Department of Thermal Engineering, Tsinghua University, Beijing100084, China
bS Supporting Information ABSTRACT: Rapidly hydrated sorbent, prepared by rapidly hydrating adhesive carrier particles and lime, is a highly effective sorbent for moderate temperature circulating fluidized bed flue gas desulfurization (CFB-FGD) process. The residence time of fine calcium-containing particles in CFB reactors increases by adhering on the surface of larger adhesive carrier particles, which contributes to higher sorbent calcium conversion ratio. The circulation ash of CFB boilers (α-adhesive carrier particles) and the spent sorbent (β and γ-adhesive carrier particles) were used as adhesive carrier particles for producing the rapidly hydrated sorbent. Particle physical characteristic analysis, abrasion characteristics in fluidized bed and desulfurization characteristics in TGA and CFB-FGD systems were investigated for various types of rapidly hydrated sorbent (α, β, and γ-sorbent). The adhesion ability of γ-sorbent was 50.1% higher than that of α-sorbent. The abrasion ratio of β and γ-sorbent was 16.7% lower than that of α-sorbent. The desulfurization abilities of the three sorbent in TGA were almost same. The desulfurization efficiency in the CFB-FGD system was up to 95% at the bed temperature of 750 °C for the β-sorbent.
1. INTRODUCTION Dry flue gas desulfurization has the advantages of low investment, little water consumption, little equipment corrosion and small area requirements with dry powder byproduct. The main disadvantage is low gassolid sulfuration reaction activity, leading to a low calcium conversion ratio and low desulfurization efficiency.1,2 Thus, it is very important to develop an easily prepared and economical sorbent with high sulfur reaction activity for industrial application of dry flue gas desulfurization. Hou et al.3 developed a moderate temperature dry flue gas desulfurization technique using circulating fluidized bed system. The sorbent was prepared by rapidly hydrating the mixture of coal fly ash and lime by the mass ratio of 4:1 for 2 h at ambient temperature and then drying for 1 h at 150 °C.4,5 Larger ash particles were used as adhesive carrier particles, which carried the fine calcium-containing particles in the CFB reactor. Compared with other calcium-based sorbent modified by coal fly ash,610 the hydration time and temperature are significantly reduced for rapidly hydrated sorbent. The desulfurization efficiency for a pilot-scale CFB-FGD system was 6783% at the bed temperature of 600800 °C and the Ca/S ratio of 2.0.11 However, the particle abrasion of rapidly hydrated sorbent in CFB reactor reduced the residence time of the fine calcium-containing particles, which reduced the calcium conversion ratio of the sorbent. Lee12 used the circulation ash from CFB boilers as the adhesive carrier particles instead of coal fly ash and reduced the mass ratio of coal fly ash and lime to 2:1. Experimental results showed that the circulation ash enhanced the adhesion between the fine calcium-containing particles and the adhesive carrier r 2011 American Chemical Society
particles, reduced the sorbent abrasion ratio and increased the sorbent desulfurization ability in TGA. However, the source of the circulation ash is inconvenient, which would limit the application of this sorbent. One potential solution is to reuse the spent sorbent as adhesive carrier particles to prepare the rapidly hydrated sorbent. Particle physical characteristic analysis, abrasion experiment and desulfurization experiment in TGA were conducted for the rapidly hydrated sorbent prepared by the spent sorbent and lime; desulfurization experiments on a pilot-scale CFB-FGD system was conducted to investigate the desulfurization ability of the rapidly hydrated sorbent in the CFB-FGD system.
2. EXPERIMENTAL SECTIONS 2.1. Sorbent Preparation and Desulfurization Experiment on the CFB-FGD System. The lime used in the experiment was
from Laishui, Hebei in China. The circulation ash, which was circulated in the CFB boiler, was from the thermal power plant of Tsinghua University. Adhesive carrier particles were mixed with lime to prepare the rapidly hydrated sorbent. The sorbent preparation steps were as follows.4,12 The lime, adhesive carrier particles, and water were mixed in a hydration mixer with continuously stirring at ambient Received: May 23, 2011 Accepted: September 19, 2011 Revised: September 18, 2011 Published: September 19, 2011 9421
dx.doi.org/10.1021/es2017285 | Environ. Sci. Technol. 2011, 45, 9421–9426
Environmental Science & Technology
Figure 1. Particle size distributions from a laser diffraction instrument (Malvern Mastersizer 2000).
temperature to produce sorbent slurry. The water/solid mass ratio was 1:1 and the mass ratio of adhesive carrier particles and lime was 2:1.The hydration time was 2 h. Then, the sorbent slurry was dried in an infrared desiccator to a water content less than 10%. The drying time was 1.5 h at a drying temperature of 150 °C. Moderate temperature desulfurization experiment was conducted on the pilot-scale CFB-FGD system. The pilot-scale CFB-FGD system (Supporting Information (SI) Figure S1) mainly includes the sorbent preparation subsystem, the flue gas generator subsystem and the CFB reactor. Detailed descriptions of the system were given in refs 3,11. Flue gas generated by the oil burner was mixed with a small amount of air to produce 600800 °C simulated flue gas. SO2 was added to the flue gas before the CFB reactor. The CFB reactor riser was 6 m high with a diameter of 0.305 m and a flue gas flow rate of 300 N m3/h. The flue gas passed through the CFB reactor and reacted with the sorbent and then went through the cyclone separator and the bag filter before emitting from the stack. The sorbent particles collected in the cyclone separator were fed back into the reactor for further circulation or drained out of the system. The O2, CO2, and SO2 concentrations in the flue gas were measured online at the CFB reactor inlet and outlet by a PS3400 type gas analyzer. The desulfurization efficiency was directly calculated from the inlet and outlet SO2 concentrations. First, rapidly hydrated sorbent (α-sorbent) was prepared by the circulation ash (α-adhesive carrier particles) and lime. The α-sorbent was used in CFB-FGD desulfurization experiments and the first spent sorbent (β-adhesive carrier particles) was collected. Second, β-sorbent was prepared by β-adhesive carrier particles and lime. The β-sorbent was used in CFB-FGD desulfurization experiments and the second spent sorbent
ARTICLE
(γ-adhesive carrier particles) was collected. Third, γ-sorbent was prepared by γ-adhesive carrier particles and lime. The relationship among the adhesive carrier particles and sorbent were shown in SI Figure S2. The experimental conditions: bed temperature, T = 600800 °C; bed superficial velocity, U0 = 2.5 m/s; inlet SO2 concentration, CSO2‑in = 1500 ppm; inlet CO2 concentration, CCO2‑in = 10%; inlet O2 concentration, CO2‑in = 10%; bed pressure drop, ΔP = 1000 Pa; Ca/S ratio, Ca/S = 1.5 or 1.0. 2.2. Abrasion Experiment. The abrasion of rapidly hydrated sorbent had significant influence on the sorbent desulfurization performance.12,13 A fluidized particle abrasion test bed (SI, Figure S3) was developed to simulate the sorbent abrasion characteristics in a CFB.14,15 The sorbent used for the abrasion experiment was first calcined for 30 min at 750 °C to simulate moderate temperature conditions. The fan (250W) pushed the air at ambient conditions into the bed (inner diameter 90 mm and height 1100 mm) to fluidize the precalcined rapidly hydrated sorbent. The large particles remained in the bed while the fine particles were entrained in the cyclone separator (d99 = 40 μm). Some fine particles from the fluidized bed were collected by the cyclone separator while other fine particles were collected by the bag filter (efficiency 99%). The test began with 200 g of sorbent being put into the bubbling fluidized bed with an air flow rate of 5 N m3/h. Some minutes later, the fan was stopped and the particles in the cyclone separator and the bag filter were measured. The total abrasion time was set to be 60 min.12 An abrasion ratio, Ra, was defined to reflect the sorbent mass change in the fluidized abrasion test bed. Ra ¼
mc þ mb 100% m0
ð1Þ
where mc represents the mass of particles collected by the cyclone separator, mb represents the mass of particles collected by the bag filter, and m0 is the initial mass of calcined sorbent. 2.3. Desulfurization Experiment in TGA. First, about 10 mg of sorbent was preheated to 750 °C in TGA under N2 atmosphere. After the sorbent weight became stable, the reaction gas was introduced into the TGA at a gas flow rate of 100 mL/min for a reaction time of 80 min. The reaction gas consisted of SO2 (1500 ppm), O2 (5%), and N2 (balance gas). CO2 was not added to the reaction gas because CO2 does not obviously affect the sorbent desulfurization ability at 750 °C since almost all desulfurization reaction products were CaSO4 at this condition.16 The sorbent desulfurization ability in TGA, Rd (mg/g), was represented by the mass of SO2 absorbed by per gram of sorbent. Since the desulfurization reaction product was CaSO4 for this condition, CaO(solid) + SO2(gas) + 1/2O2(gas) f CaSO4(solid), Rd can be calculated by eq 2. Rd ¼
ðmi m0 Þ 64 1000 80 m0
ð2Þ
where mi represents the sorbent mass at a given time (mg) and m0 represents the initial mass of precalcined sorbent (mg). 2.4. Particle Characterization. A laser diffraction instrument (Malvern Mastersizer 2000) was used to evaluate the particle size distribution of the samples. A scanning electron microscope (SEM: KYKY-2800) was used to observe the particle surface micrographs. The specific surface areas and pore volume distribution were measured by the nitrogen adsorption method 9422
dx.doi.org/10.1021/es2017285 |Environ. Sci. Technol. 2011, 45, 9421–9426
Environmental Science & Technology
ARTICLE
Figure 2. Typical SEM (KYKY-2800) images of various particles.
using an ASAP2010 type BET analyzer and AUTOSCAN-33 type mercury porosimetry. A Bruker D8 Advance X-ray diffractometer (XRD) was used to characterize the sample components.
3. RESULTS AND DISCUSSION 3.1. Particle Physical Characteristics. The particle size distributions of various particles are shown in Figure 1. The particle size distributions of the three adhesive carrier particles were similar with an average diameter of about 250 μm and the Ca(OH)2 was smaller than 10 μm. Compared with the corresponding adhesive carrier particles, the volume ratio of the particles smaller than 10 μm increased significantly for the sorbent, indicating that part of the fine calcium-containing particles did not adhere to the adhesive carrier particles. Typical SEM images of various particles are shown in Figure 2. The three adhesive carrier particles had irregular, rough and porous surfaces. Many fine calcium-containing particles were adhered to the surfaces of the three adhesive carrier particles. The XRD results are shown in Figure 3. α-adhesive carrier particles mainly contain SiO2(q, Quartz), Al6Si2O13 (Mullite) and little CaAl2Si2O8 (a, Anorthite). β-adhesive and γ-adhesive carrier particles have similar components, including CaSO4 (s, Anhydrite), little CaCO3 (c, Calcite), and CaO (e, Lime). α-sorbent mainly contains Ca(OH)2 (p, Portlandite), SiO2+Al6Si2O13 (q), CaCO3 (c) and little cementitious components. Cementitious components, such as CaAl2Si2O8 3 4H2O (g, hydrated anorthite), were the products of pozzolanic reaction.1719 β-sorbent and γ-sorbent also have similar components, including CaSO4 (s) and more cementitious components (CaAl2Si2O8 3 4H2O, et al.). The main reason was that CaSO4 can promote the pozzolanic reaction between lime and adhesive carrier particles.19 The specific surface areas and pore volumes of various particles are shown in Table 1. Figure 4 shows the sorbent pore volume distributions. The measurement uncertainties of these results were 5%. The specific surface area of α-adhesive carrier particles was larger than that of β and γ-adhesive carrier particles and coal
fly ash. The main reason was that β and γ-adhesive carrier particles had been abraded in the CFB reactor which decreased the specific surface area of the particles. The pore volumes and pore volume distributions of α, β, and γ-adhesive carrier particles were similar and larger than those of coal fly ash. α, β, and γ-adhesive carrier particles had similar porous structures which were not significantly influenced by particle abrasion while the coal fly ash particles were more compact. The specific surface areas and pore volumes of α, β, and γ-sorbent were also similar and larger than those of coal fly ash sorbent, which could improve the desulfurization ability. 3.2. Adhesion Characteristic of Adhesive Carrier Particles. VPM10 represents the volume ratio of particles smaller than 10 μm to the total volume of particles, which can be obtained from the particle size distribution. ΔVPM10, the increase of VPM10 compared with the sorbent and the adhesive carrier particles was calculated, which is listed in Table 2. Since the lime particles were smaller than 10 μm, the lime particles that did not adhere to the adhesive carrier particles led to ΔVPM10. The adhesion characteristic of the adhesive particles was represented by ΔVPM10. The ΔVPM10 of γ-sorbent compared with γ-adhesive carrier particles was smallest, indicating that γ-adhesive carrier particles had the best adhesion characteristics. Compared with α-adhesive carrier particles, the adhesion characteristic of γ-adhesive carrier particles increased 50.1%, which indicated that the adhesion characteristic of adhesive carrier particles would increase when the spent sorbent was used as adhesive carrier particles. As shown in Table 1 and Figure 4, the specific surface of α-adhesive carrier particles was larger than those of the other three adhesive carrier particles. The pore volume and pore volume distribution of coal fly ash was smaller than those of α, β, and γ-adhesive carrier particles. The adhesion characteristic of coal fly ash (ΔVPM10 = 36.57%)13 was much worse than that of the other three adhesive carrier particles, indicating that the pore structure of the adhesive carrier particles had significant influence on the adhesion characteristic of the adhesive carrier particles while the influence of the specific surface area was negligible. 9423
dx.doi.org/10.1021/es2017285 |Environ. Sci. Technol. 2011, 45, 9421–9426
Environmental Science & Technology
ARTICLE
Figure 3. X-ray diffraction (Bruker D8 Advance) results of various particles.
Table 1. Particles Specific Surface Area and Pore Volume by the Nitrogen Adsorption Method at 77 K α-adhesive carrier particles
β-adhesive carrier particles
γ-adhesive carrier particles
Coal fly ash
specific surface area, m2/g
8.10
2.96
3.05
2.19
pore volume, cm3/g
0.0136
0.0121
0.0128
0.00491 coal fly ash sorbent
α-sorbent
β-sorbent
γ-sorbent
specific surface area, m2/g
17.5
15.0
16.7
5.60
pore volume, cm3/g
0.0540
0.0644
0.0495
0.0391
XRD results in Figure 3 show that β and γ-adhesive carrier particles had large amount of CaSO4 (Figure 5-b,c), which promoted the pozzolanic reaction and increased the production of cementitious components (Figure 5-e, f, CaAl2Si2O8 3 4H2O et al.).20 These cementitious components enhanced the adhesion force between the fine calcium-containing particles and the
adhesive carrier particles. Thus, the adhesion characteristic of β and γ-adhesive carrier particles increased. 3.3. Abrasion Resistant Characteristic of Sorbent. Figure 5 shows the abrasion test results of the three sorbent. The abrasion ratios of β and γ-sorbent were similar in 60 min (∼8.5%), which were lower than that of α-sorbent (10.2%). The sorbent abrasion 9424
dx.doi.org/10.1021/es2017285 |Environ. Sci. Technol. 2011, 45, 9421–9426
Environmental Science & Technology
ARTICLE
Figure 5. Abrasion ratios of various sorbent.
Figure 4. Pore volume distributions of various particles by the nitrogen adsorption method at 77 K.
Table 2. VPM10 Content of Various Particles adhesive carrier α-adhesive carrier β-adhesive carrier γ-adhesive carrier particles particles particles particles VPM10/%
1.33
2.27
sorbent
α-sorbent
β-sorbent
γ-sorbent
VPM10/%
13.2
9.76
5.92
ΔVPM10/%
11.9
7.49
5.92
0.00
resistant characteristic increased 16.7% due to the spent sorbent reuse as adhesive carrier particles. The reason was that the improved adhesion characteristic of β and γ-adhesive carrier particles enhanced the adhesion force between the fine calciumcontaining particles and the adhesive carrier particles, which was good for improving the abrasion resistant characteristic of the sorbent, increasing the residence time of fine calcium-containing particles in CFB and increasing the calcium conversion ratio of the sorbent. 3.4. Desulfurization Ability in TGA. The desulfurization abilities of α, β, and γ-adhesive carrier particles and sorbent in TGA are shown in Figure 6-a. The desulfurization abilities of the lime in various sorbent were calculated by deducting the desulfurization abilities of adhesive carrier particles from that of the sorbent, as shown in Figure 6-b. The initial reaction rate (010 min) of α-sorbent, which mainly was influenced by the specific surface area of the sorbent, was larger than those of β and γ-sorbent. The desulfurization abilities of the three sorbent were all about 313 mg/g in 80 min. The desulfurization ability of α-adhesive carrier particles was almost zero, while that of the β and γ-adhesive carrier particles was about 30 mg/g due to the existence of CaCO3 (c) and CaO (e) in the components of β and γ-adhesive carrier particles (Figure 3-b, c). The desulfurization abilities of the lime in α, β, and γ-sorbent were about 900 mg (SO2)/g (lime), while that of coal fly ash
Figure 6. Sorbent desulfurization abilities in TGA.
sorbent was 725 mg (SO2)/g (lime). This was because the pore volumes and pore volume distributions of the α, β, and γsorbent, especially the medium pore (250 nm) distribution that had significant influence on the sulfur reaction,20 were similar and larger than those of the coal fly ash sorbent, as shown in Figure 4-b. The sorbent desulfurization ability in TGA, where the abrasion phenomena would be negligible, provides ideal capacity for the sorbent desulfurization ability. Desulfurization experiments in the pilot-scale CFB-FGD system could provide more actual desulfurization ability of the sorbent. 3.5. Desulfurization Efficiency of the CFB-FGD System. The desulfurization efficiency at various bed temperatures for the CFB-FGD system is shown in Figure 7. The desulfurization efficiency in the CFB-FGD system increased from 68% to 83% when the α-sorbent was used instead of the coal fly ash sorbent at the bed temperature of 750 °C and the Ca/S ratio of 1.5. The corresponding sorbent calcium conversion ratio increased from 45% to 55%, confirming that the circulation ash, used as adhesive carrier particles instead of coal fly ash, can 9425
dx.doi.org/10.1021/es2017285 |Environ. Sci. Technol. 2011, 45, 9421–9426
Environmental Science & Technology
ARTICLE
’ ACKNOWLEDGMENT This research was supported by the Special Funds for Major State Basic Research Projects (No. 2006CB200305). ’ REFERENCES
Figure 7. Desulfurization efficiency of CFB-FGD system with various sorbent (U = 2.5 m/s, Fbed = 16.7 kg/m3, CSO2 = 1500 ppm, CCO2 = 10%, CO2 = 10%).
increase the sorbent calcium conversion ratio. The desulfurization efficiency of the CFB-FGD system with β-sorbent was about 10% higher than that with the α-sorbent. The calcium conversion ratio of β-sorbent achieved 80% at the bed temperature of 750 °C and the Ca/S ratio of 1.0. The reason was that the adhesion characteristic of β-adhesive carrier particles and the abrasion resistant characteristic of β-sorbent were improved, which significantly increased the residence time of the fine calciumcontaining particles in CFB. By comparing the particle components collected by the bagfilter in Figures 5-g and 5-h, the bag-filter particles from αsorbent contained large amount of Ca(OH)2 (p), while there was nondetectable diffraction of Ca(OH)2 in the bag-filter particles for β-sorbent. This phenomenon proved that the residence time of the fine calcium-containing particles of α-sorbent in CFB was relatively shorter than that of β-sorbent; the fine calcium-containing particles were quickly collected by the bag filter after the sorbent was fed in and could not react with SO2 effectively.
’ ASSOCIATED CONTENT
bS
Supporting Information. Schematic of the pilot-scale CFB-FGD system (Figure S1); Relationship between α,β,γadhesive carrier particles and α,β,γ-sorbent (Figure S2); Schematic of the fluidized abrasion test bed (Figure S3). This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +86-10-62785669; fax:+86-10-62770209; e-mail: youcf@ tsinghua.edu.cn.
(1) Matsushima, N.; Nishioka, Y. L.; Nishioka, M.; Sadakata, M.; Qi, H. Y.; Xu, X. C. Novel dry-desulfurization process using Ca(OH)2/fly ash sorbent in a circulating fluidized bed. Environ. Sci. Technol. 2004, 38 (24), 6867–6874. (2) Pandey, R. A.; Biswas, R.; Chakrabrti, T.; Devotta, S. Flue gas desulfurization: Physicochemical and biotechnological approaches. Crit. Rev. Environ. Sci. Technol. 2005, 35 (6), 571–622. (3) Hou, B.; Qi, H. Y.; You, C. F. Dry desulfurization in a circulating fluidized bed (CFB) with chain reactions at moderate temperatures. Energy Fuels 2005, 19 (1), 73–78. (4) Zhang, J.; Zhao, S. W.; You, C. F.; Qi, H. Y.; Chen, C. H. Rapid hydration preparation of calcium-based sorbent made from lime and fly ash. Ind. Eng. Chem. Res. 2007, 46 (16), 5340–5345. (5) Zhang, J.; You, C. F.; Zhao, S. W.; Chen, C. H.; Qi, H. Y. Characteristics and reactivity of rapidly hydrated sorbent for semidry flue gas desulfurization. Environ. Sci. Technol. 2008, 42 (5), 1705–1710. (6) Ali, Al-S.; Hitoki, M. Comparative reactivity of treated FBC-and PCC-fly ash for SO2 removal. Can. J. Chem. Eng. 1995, 73 (11), 678–685. (7) Paolo, D. Investigation of the SO2 adsorption properties of Ca(OH)2-fly ash systems. Fuel 1996, 75 (6), 713–716. (8) Ishizuka, T.; Tsuchiai, H.; Murayama, T.; Tanaka, T.; Hattori, H. Preparation of active absorbent for dry-type flue gas desulfurization from calcium oxide, coal fly ash, and gypsum. Ind. Eng. Chem. Res. 2000, 39 (5), 1390–1396. (9) Fernandez, J.; Renedo, M. J. Kinetic modelling of the hydrothermal reaction of fly ash, Ca(OH)2 and CaSO4 in the preparation of desulfurant sorbents. Fuel 2004, 83 (45), 525–532. (10) Lee, K. T.; Mohamed, A. R.; Bhatia, S.; Chu, K H. Removal of Sulfur Dioxide by fly ash/CaO/CaSO4 sorbents. J. Chem. Eng. 2005, 114 (13), 171–177. (11) Zhang, J.; You, C. F.; Qi, H. Y.; Hou, B.; Chen, C. H.; Xu, X. C. Effect of operating parameters and reactor structure on moderate temperature dry desulfurization. Environ. Sci. Technol. 2006, 40 (13), 4300–4305. (12) Li, Y.; You, C. F.; Song, C. X. Adhesive carrier particles for rapidly hydrated sorbent for moderate-temperature dry flue gas desulfurization. Environ. Sci. Technol. 2010, 44, 4692–4696. (13) Li, Y.; Song, C. X.; You, C. F. Experimental study on abrasion characteristics of rapidly hydrated sorbent for moderate temperature dry flue gas desulfurization. Energy Fuels 2010, 24, 1682–1686. (14) Chu, C. Y.; Hsueh, K. W.; Hwang, S. J. Sulfation and abrasion of calcium sorbent in a bubbling fluidized bed. J. Hazard. Mater. 2000, B80, 119–133. (15) Chu, C. Y.; Hwang, S. J. Abrasion and sulfation of calcium sorbent and solids circulation rate in an internally circulating fluidized bed. Powder Technol. 2002, 127, 185–195. (16) Hou, B.; Qi, H. Y.; You, C. F.; Fan, B. G.; Xu, X. C. The efficiency of CO2 and NO to dry desulfurization process at medium temperature. J. Eng. Thermo. 2004, 25 (6), 1061–1064. (17) Matinez, J. C.; Izquierdo, J. F.; Cunill, F.; Tejero, J.; Querol, J. Reactivation of fly ash and Ca(OH)2 mixtures for SO2 removal of flue gas. Ind. Eng. Chem. Res. 1991, 30 (9), 2143–2147. (18) Tsuchiai, H.; Ishizuka, T.; Ueno, T.; Hattori, H.; Kita, H. Highly active absorbent for SO2 removal prepared from coal fly ash. Ind. Eng. Chem. Res. 1995, 34 (4), 1404–1411. (19) Tsuchiai, H.; Ishizuka, T.; Nakamura, H.; Ueno, T.; Hattori, H. Study of flue desulfurization absorbent prepared from coal fly ash: Effects of the composition of the absorbent on the activity. Ind. Eng. Chem. Res. 1996, 35 (7), 2322–2326. (20) Renedo, M. J.; Fernandez, J.; Garea, A.; Ayerbe, A.; Irabien, J. A. Microstructural changes in the desulfurization reaction at low temperature. Ind. Eng. Chem. Res. 1999, 38 (4), 1384–1390. 9426
dx.doi.org/10.1021/es2017285 |Environ. Sci. Technol. 2011, 45, 9421–9426
ARTICLE pubs.acs.org/est
Effect of Fuels and Domestic Heating Appliance Types on Emission Factors of Selected Organic Pollutants Michal Syc,*,† Jirí Horak,‡ Frantisek Hopan,‡ Kamil Krpec,‡ Tomas Tomsej,§ Tomas Ocelka,§ and Vladimír Pekarek† †
Environmental Process Engineering Laboratory, Institute of Chemical Process Fundamentals, Academy of Sciences of the Czech Republic, v. v. i., Rozvojova 135/2, 165 02 Prague 6 Suchdol, Czech Republic ‡ Energy Research Center, Technical University of Ostrava, Innovation for Efficiency and Environment, 17.listopadu 15/2172, 708 33 Ostrava, Czech Republic § Department of Hygienic Laboratories, Institute of Public Health Ostrava, Partyzanske namestí 7, 702 00 Ostrava, Czech Republic
bS Supporting Information ABSTRACT: This study reports on the first complex data set of emission factors (EFs) of selected pollutants from combustion of five fuel types (lignite, bituminous coal, spruce, beech, and maize) in six different domestic heating appliances of various combustion designs. The effect of fuel as well as the effect of boiler type was studied. In total, 46 combustion runs were performed, during which numerous EFs were measured, including the EFs of particulate matter (PM), carbon monoxide, polyaromatic hydrocarbons (PAH), hexachlorobenzene (HxCBz), polychlorinated dibenzo-p-dioxins and furans (PCDD/F), etc. The highest EFs of nonchlorinated pollutants were measured for old-type boilers with over-fire and under-fire designs and with manual stoking and natural draft. Emissions of the above-mentioned pollutants from modern-type boilers (automatic, downdraft) were 10 times lower or more. The decisive factor for emission rate of nonchlorinated pollutants was the type of appliance; the type of fuel plays only a minor role. Emissions of chlorinated pollutants were proportional mainly to the chlorine content in fuel, but the type of appliance also influenced the rate of emissions significantly. Surprisingly, higher EFs of PCDD/F from combustion of chlorinated bituminous coal were observed for modern-type boilers (downdraft, automatic) than for old-type ones. On the other hand, when bituminous coal was burned, higher emissions of HxCBz were found for old-type boilers than for modern-type ones.
’ INTRODUCTION Emissions from domestic heating appliances significantly add to total environmental pollution, which was proved by seasonal changes of polycyclic aromatic hydrocarbons (PAH) levels obtained by long-term monitoring programs.1 The participation of particular sources in total pollution varies depending on the emission inventory source.2 Breivik et al.2 have found out that, in Europe, 20 45% of PAH emissions and 18 30% of the emissions of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/F) come from domestic combustion. Quass et al.3 even assert that 30 35% of PCDD/Fs are produced in domestic combustion. Emission inventories use emission factors (EFs) to calculate the participation of individual sources in total emissions. However, the reported EF values of PCDD/Fs and other pollutants from domestic combustion differ widely. The EFs of PCDD/F, polychlorinated biphenyls (PCB), hexachlorobenzene (HxCBz), PAH, and particulate matter (PM) have been determined for a fireplace and a woodstove for oak and pine fuels.4 The resulting PCDD/F EFs were 19.1 87.7 ng international toxicity equivalents (I-TEQ)/GJ depending on fuel and combustion facility types. A significant effect of the combustion facility r 2011 American Chemical Society
age on the EF values has been reported, as well as the effect of the in- or stationary phase of the combustion period.5 For the tested facilities, the PCDD/F EFs varied from 100 to 700 ng World Health Organization toxicity equivalents (WHO-TEQ)/GJ in the case of woody biomass. Even though the effect of boiler operating conditions on pollutant formation has been investigated previously, the analyzed pollutants included only the PAH of persistent organic pollutants (POPs).6 Wevers et al.7 reported no effect of the combustion period or fuel age on the EFs from five tested stoves. The resulting EF values of PCDD/F were 2 89 ng I-TEQ/kg (i.e., approximately 100 5000 ng I-TEQ/GJ); the effect of the facility type was not investigated. Chimney emissions from 30 households were sampled for determination of real PCDD/F emission factors.8 The tested combustion facilities included stoves and boilers with a wide range of thermal input and facility age. The measured emission factors of PCDD/F were Received: May 27, 2011 Accepted: September 20, 2011 Revised: August 1, 2011 Published: September 20, 2011 9427
dx.doi.org/10.1021/es2017945 | Environ. Sci. Technol. 2011, 45, 9427–9434
Environmental Science & Technology
ARTICLE
Table 1. Ultimate and Proximate Analyses of Tested Fuelsa fuel properties
a
lignite (LI)
bituminous coal (BC)
beech logs (BL)
maize straw pellets (MP)
spruce logs (SL)
ash (wt %)
5.8
3.2
0.9
17.6
combustibles (wt %)
94.2
96.8
99.1
82.4
1.2 98.8
LHVb (MJ/kg)
26.3
32.8
17.3
14.7
18.1
carbon (wt %)
64.6
79.9
45.5
39.3
50.5
hydrogen (wt %)
5.28
4.45
5.65
4.92
5.98
nitrogen (wt %)
0.902
1.03
0.101
0.881
0.409
oxygen (wt %)
22.6
10.7
47.6
37.2
41.9
sulfur (wt %) chlorine (mg/kg)
0.854 40.0
0.720 1 620
0.243 58.0
0.144 1 170
0.0710 60.0
All on dry basis. b Lower heating value.
within 2 4500 ng I-TEQ/GJ depending on a tested facility, fuel, and/or an inappropriate operation of the facility. However, isokinetic sampling was not possible due to a low flue gas velocity in the chimney. Moreover, the EFs with respect to the boiler design and age were observed, but the study concentrated mainly on other pollutants than POPs.9 The EFs of PCDD/F, PCB, polychlorinated naphthalenes (PCN), PAH and PM for wood and coal burning were monitored in simulated open-fire domestic facilities that eliminated the effect of the combustion facility.10 The emissions of volatile organic compounds (VOC) and PAH for birch wood combustion in a wood stove were also quantified.11 Launhardt and Thoma12 published a complex study about the emissions of PCDD/F, polychlorinated phenols (PCPh), polychlorinated benzenes (PCBz), and PAH during combustion of five herbaceous and woody fuels in a modern automatic boiler. A possible effect of different dilution systems and filter media used for sampling on the EFs has been described as well.13 Numerous studies, besides the above-mentioned, have been conducted on this topic with varying results.14,15 Moreover, some studies mentioned the effect of facility design, but this effect has not been studied in detail yet. The majority of the studies particularly focus on the fuel effect. Therefore, this study focuses on the determination of both effects, the effect of the facility as well as the fuel. The EFs of some major pollutants [CO, total organic carbon (TOC), PM] and selected organic pollutants (PCDD/F, PCB, PAH, PCPh, PCBz) were determined for combustion of five different fuels in six combustion facilities with various combustion designs. The general aim of this study was to obtain complex and comparable data allowing identification of the above-mentioned influences. Furthermore, the obtained data enable us to particularize the emission inventories of the studied POPs and the estimation of emissions of new POPs such as pentachlorobenzene (PeCBz).
’ EXPERIMENTAL SECTION Fuels. Fuels were chosen according to their consumption in
domestic heating. Lignite (LI) is used in domestic combustion mainly in the Czech Republic and Poland.16 Bituminous coal (BC) from Poland was chosen due to its high Cl content and because of its high consumption therein. Three biomass fuels were tested as examples of the currently favored solid fuels for domestic heating. Beech logs (BL) were chosen for experiments as a hardwood sample, spruce logs (SL) as a softwood species, and maize straw pellets (MP) as a high Cl content herbaceous species. The results of the ultimate and proximate analyses of the
Table 2. Summary of Performed Runsa fuel
run
FFR
W
output
T
O2
(kg/h)
(wt %)
(kW)
(°C)
(vol. %)
B1, Over-Fire Boiler LI
1
8.1
27.5
21.6
235
11.3
LI
2 3
4.8
27.5
13.6
196
16.3
BC
4 6
2.9
2.41
16.1
234
12.5
BL
7 9
6.9
9.58
19.3
222
10.9
LI
10 11
5.7
22.6
182
10.7
BC
12 13
3.2
2.41
18.6
204
11.4
BL
14 16
6.2
9.58
18.2
163
12.5
LI
17 19
5.9
27.5
23.9
256
LI
20 22
3.7
26.4
15.4
167
12.0
26.4
13.3
B2, Under-Fire Boiler 27.5
B3, Automatic Under-Fire Boiler with Screw Conveyor
LI
23 25
1.9
BC
26 28
4.0
MP
29 31
5.1
LI
32 34
7.7
BL
35 37
9.8
SL
38 40
7.2
BL
41 43
8.7
BL
44 46
2.3
2.83 10.3
8.0
127
25.5
193
13.8
161
9.81
7.81 12.8
B4, Downdraft Boiler 27.5 9.58
33.4
305
28.9
260
10.6
6.83
23.1
224
10.5
29.7
261
B5, Downdraft Boiler 10.0 5.86
6.60
S6, S-Draft-type Stove 9.58
6.5
302
15.0
a
FFR, fuel feeding rate; W, fuel water content; output, measured according to EN 303-5; T, temperature at chimney outlet; O2, average O2 in chimney outlet.
tested fuels are shown in Table 1. The water content in the tested fuels is displayed in Table 2. Tested Facilities. The tested facilities represent the main designs of boilers used for domestic heating.17 For comparison, a classic S-draft type stove was used. Schemes of the tested facilities are shown in Figure 1; details about the facilities are given below. Boiler 1 is a hot water over-fire boiler with manual stoking and natural draft (see Figure 1a). The whole fuel batch is combusted at one time, and the operation of the facility is periodical. Primary air (P) is blown under the water-cooled fixed grate (1) through an automatic draft-regulating damper in the ash pit door 9428
dx.doi.org/10.1021/es2017945 |Environ. Sci. Technol. 2011, 45, 9427–9434
Environmental Science & Technology
ARTICLE
Figure 1. Schemes of tested combustion facilities.
(see Figure 1a). A secondary air (S) inlet into the gas combustion zone is in the fuel feeding door and can be manually regulated with a damper. The recommended fuels are coke, bituminous coal, and wood logs; lignite is also possible. Boiler 2 is an under-fire boiler with natural draft and manual stoking (see Figure 1b). The boiler can be divided into three parts: a fuel storage (1), a combustion chamber (2), and a gas flow chamber (3). Devolatilization and partial combustion occurs in a small part of fuel in the bottom of the fuel storage, while main combustion takes place in the follow-up combustion chamber. Primary air (P) is supplied through a damper in the fuel feeding door. Secondary air (S) is led through a channel to the combustion chamber; tertiary air is supplied sidewards to the combustion chamber as well. Rotary grates (4) are placed below the fuel storage and the combustion chamber. The recommended fuel is lignite, but other solid fuels can be used as well. Boiler 3 is a modern under-fire boiler (see Figure 1c) with forced draft and automatic stoking by a screw conveyor (1). The upper part of the boiler is a lamellate heat exchanger (2). The lower part is a combustion chamber formed by an iron grate (3), a ceramic heat reflector (4), a retort for fuel feeding (5), and an air mixing system (6). Primary air (P) is supplied by a fan (7) to the air mixing system. There is an ash chamber (8) situated under the combustion chamber. The recommended fuels are lignite and biomass pellets. Other solid fuels with required granulometry can be combusted as well. Boiler 4 is a modern downdraft boiler with manual stoking and forced draft by a draw-off fan (see Figure 1d). The boiler consists of two chambers; the upper one is for fuel storage (1) and the lower one is a combustion chamber (2). The chambers are divided by a special rotating burner (4). Primary air (P) is supplied to the combustion chamber from above through the batch of fuel and a special cast-iron grate (3). Secondary air (S) is supplied to the grate. The recommended fuels
are lignite, but wood logs and other solid fuels can be used as well. Boiler 5 is a modern downdraft boiler with manual stoking and forced draft by a draw-off fan. It has a similar construction to boiler 4 with larger chambers. It is for wood combustion only and has a stationary fire-clay grate. The recommended fuel is wood logs. Stove 6 is a modern S-draft stove with grate (1) and periodical combustion operation (see Figure 1e). Combustion Runs. Parameters of the realized combustion runs are summarized in Table 2. Thirty-nine combustion tests were performed according to the producer’s instructions (i.e., steady-state regime) and/or European standard EN 303-5. All combustion runs started with fuel ignition. After the ignition, the combustion process was conducted so that the nominal output of the boiler was reached immediately. The facility was operated under a steady-state regime for at least 2 h. Then, sampling of flue gases for off-line analyzed pollutants commenced. Boiler B1 and stove S6 are batch-operated facilities with combustion periods, therefore sampling of the flue gases started synchronously with the combustion period start and finished with the period end. The fuel batch for the measured combustion period was added onto a thin layer of burning fuel. Hence, the obtained EFs are not cold-start values. Boilers B2, B4, and B5 operate on fuel batch principles, however, due to their design there is no strictly binding combustion period. The B3 facility operates quasi-continuously. The sampling time for these four facilities was chosen to be about 6 h. The majority of the runs were triplicate in order to obtain robust and representative data. The “memory effect” has significant impact on the formation levels of some POPs. Therefore, the testing facility was mechanically cleaned up after triplicate runs with the same fuel and the same combustion facility, which effectively suppressed the effect of ash deposition inside the testing facility. 9429
dx.doi.org/10.1021/es2017945 |Environ. Sci. Technol. 2011, 45, 9427–9434
Environmental Science & Technology The stoking period and the size of the fuel batch influenced the emissions of the facilities operating with a combustion period, that is, B1 and S6. Both parameters were set in accordance with EN 303-5 (except runs 2 and 3) as follows: run 1 (one LI batch of 25 kg), runs 2 3 (stoking period 0.5 h, fuel batch 2.4 kg of LI), runs 4 6 (one 11.6 kg weight batch of BC), runs 7 9 (stoking period ca. 2 h, fuel batch 13.8 kg of BL), and runs 44 46 (stoking period ca. 45 min, fuel batch 1.8 kg). The effect of lower than nominal output was observed in B3. Runs 17 19 were performed with the nominal output of 23.9 kW (fuel feeding rate 5.9 kg/h, fan output 90%, screw conveyor on/ off 12/20 s). Runs 20 22 were performed with approximately 65% of the nominal output, that is, 15.4 kW (fuel feeding rate 3.7 kg/h, fan output 45%, screw conveyor on/off 6/23 s). Runs 23 25 were performed with the output of 8.0 kW, which is 33% of the nominal boiler output (fuel feeding rate 1.9 kg/h, fan output 7%, screw conveyor on/off 5/40 s). The design of the whole runs and the sampling was the same as mentioned above for the steady-state runs: the runs started with fuel ignition, subsequently the boiler was operated at intermittent state for ca. 1 2 h, then sampling of flue gases of off-line analyzed pollutants began. Testing Facility. The boilers and the stove were tested at a domestic combustion testing facility consisting of a balance, the tested boiler or stove, an isolated chimney system exhausting to a dilution tunnel hood, a dilution tunnel, and a fan. The testing facility was constructed on the principle of EPA Test Method 5G. More details and the scheme of the facility have been already published elsewhere.18,19 Analytical Procedures. Continuous measurements of CO and total organic carbon (TOC) were performed in the isolated chimney. Measurements of CO2 and O2 were performed simultaneously in the isolated chimney and the dilution tunnel. O2, CO, and CO2 were measured by an Advance Optima multigas analyzer in accordance with EN 15058 (CO), ISO 10396 (standard for sampling). The content of TOC was analyzed by a Multi-FID 100 (EN 12619). The particulate matter content was determined in the dilution tunnel in accordance with ISO 9096. Flue gases for determination of organic compounds were sampled isokinetically in the dilution tunnel according to EN 1948 (filtration condensation method). The determination of PCDD/F and DL-PCB was based on liquid liquid/Soxhlet extraction, followed by multistep cleanup (silica gel/alumina/ carbon) and gas chromatography (GC)/high-resolution mass spectrometry (HRMS) and GC/tandem mass spectrometry (MS/MS) analysis in according with EN 1948. For analysis of PAH and PCBz, aliquots of raw extracts were taken and analyzed after cleanup by HPLC with postcolumn fluorescence derivatization (FLD) and GC/MS/MS. The detailed analytical procedure can be found in Supporting Information. Calculation. Levels of some of the analytes were below detection limits; in such cases the detection limit values were used as a representative. The presented emission factors were related to fuel energy, that is, the EFs are based on the real lower heating values of the given fuels.
’ RESULTS AND DISCUSSION The results are subdivided into three groups. The first group consists of CO, TOC, PAH, and PM, i.e. nonchlorinated products of incomplete combustion. In the second group, there are PCBz as the main precursors of PCDD/F formation. And finally,
ARTICLE
the third part of discussion deals with the results of PCB and PCDD/F. Carbon Monoxide, Total Organic Carbon, Polycyclic Aromatic Hydrocarbons, and Particulate Matter. The obtained emission factors of PM, CO, TOC, and PAHs are depicted in Figure 2. A strong effect of boiler design on the EFs can be made out from the results. The most common indicator of combustion quality is the CO level. In both old-type boilers B1 and B2, the EFs of CO almost reached over 4 kg/GJ; the highest value of nearly 8 kg/GJ was observed during run 1. Similar values in the range 2.1 4.3 kg of CO/GJ had been measured for old-type boilers previously.8 The EFs of CO reached values in units of kilograms per gigajoule even for the modern boilers during combustion runs 29 31, 35 37, and 38 40, when inappropriate or substitute fuels were used. The results confirm that the excess amount of combustion air in flue gases deteriorates the quality of combustion and causes an increase in CO.9 The above-mentioned effect took place in modern boilers more significantly than old-type ones. The decrease of flue gas temperature in the chimney inlet with increased O2 was observed (see Table 2); this implies that deterioration of combustion was caused by a decrease of combustion temperature and/or by shortening of residence time in the postcombustion zones. The excess amount of combustion air (and O2) can be caused by use of substitute fuels for which the boilers were not designed, especially when they differ in energy density and devolatilization rate. Boiler B4 (concretely the combustion chamber size and the burner type) was designed for lignite combustion. Hence, in the case of BL combustion (runs 35 37), the EFs of CO increased 33 times in comparison with LI combustion (runs 32 34), and O2 concentration in flue gases rose from ca. 6.8 to 10.6 vol %. These high EFs are also given by the ratio of primary/secondary combustion air, which was set up for LI combustion on B4. The influence of primary to secondary air ratio is obvious also from the comparison of CO values obtained during runs 38 40 and 41 43 in B5. To minimize CO emissions, the change of the above-mentioned ratio is necessary in the case of substitution of hardwood for softwood. CO increase was also observed during runs 29 31 for B3; it was caused by deterioration of combustion due to formation of ash sinters on the retort. This had been reported previously for straw combustion in automatic residential boilers.5 The highest EFs of PM were measured for the over-fire boiler (B1), which seems to be logical due to its primitive design without any postcombustion zones or zones where particles can be separated from flue gases by means of gravitation (see Figure 1). Similarly, high EFs were found for the old-type boiler (B2) with the under-fire concept of combustion. The EFs of PM for B1 and B2 were higher by approximately 1 order of magnitude compared to the other four tested facilities. The lowest PM EFs were observed for both downdraft boilers (B4 and B5), provided the recommended fuels were burned. Among other things, the quality of combustion influenced the rate of PM emissions, because the PM is also formed by unburnt carbon residues. The influence of combustion quality was obvious for B4 where, surprisingly, 3.5 times lower emissions were measured from LI (runs 32 34) than from BL (35 37). The effect of ash content in fuel appeared mainly for both old-type boilers, more significantly for B1 than B2. The fuel ash content seems to be low in the case of modern boilers; the PM EFs are mainly predetermined by boiler design. Generally, the effect of a boiler on PM emissions is indisputable from our results, which is remarkable, as the effect of fireplaces or/and stoves on PM emissions had been considered small previously.20 9430
dx.doi.org/10.1021/es2017945 |Environ. Sci. Technol. 2011, 45, 9427–9434
Environmental Science & Technology
ARTICLE
Figure 2. Mean values of emission factors of CO, PM, TOC, and PAH with standard deviations. PAH is the sum of 10 polyaromatic hydrocarbons: fluoranthene, pyrene, benzo[a]anthracene, chrysene, benzo[b]fluoranthene, benzo[k]fluoranthene, benzo[a]pyrene, benzo[g,h,i]perylene, dibenzo[a, h]anthracene, and indeno[1,2,3-cd]pyrene.
Large differences between boilers were observed for PAH emissions. The EF values of PAH from B1 and B2 were higher by 1 2 orders of magnitude than those from the modern boilers. Contrary to the CO EFs, the PAH emission from S6 was comparable with that from the modern boilers. The lowest PAH EFs were observed for the modern automatic boiler, followed by both downdraft boilers. Inappropriate or substitute fuel combustion (runs 29 31, 35 37, and 38 40) in modern-type boilers caused approximately 10-fold increase of the PAH emission, but the EFs were still 1 order of magnitude lower than the EFs from the old-type boilers. The highest EFs of nonchlorinated pollutants were measured for the old-type boilers B1 and B2. The whole fuel batch burns at once in B1; that is, the boiler operates periodically depending on the stoking period. Therefore, the fuel batch size and the stoking period affected the quality of the combustion process and the emission levels significantly. This is evident from comparison of run 1 with runs 2 3. Lignite (ca. 25 kg) for the whole combustion period (set up according to EN 303-5) was added all at once for run 1. At first, the lignite devolatilization proceeded. The O2 in flue gases was decreasing to ca. 1.5 vol %. The CO was rising over 10 vol % during the first 25 min and stayed at the maximum values for the next 10 min. Subsequently, significant oxidation of the produced gases started and the CO values decreased. The cause
of such high emission values lies in the combination of the boiler design and the specific lignite properties. The characteristic features of B1 are a very poor quality of combustion given by poor mixing of the combustion air with local oxygen deficiency, a low combustion temperature (between 400 and 800 °C),21 and a nonhomogenous temperature field (so-called “cold wall effect”). The above-mentioned disadvantages were further multiplied by slow devolatilization of lignite.22 Moreover, the observed high values were measured during “hot starts”; the cold-start emission from B1 would have been even worse. Shortening of the stoking period to 0.5 h and changing of the fuel batch size to 2.4 kg led to 2 4 times lower EFs. Steadier operation and lower temperature fluctuation were the cause of that. Emissions from the combustion of BL and BC in B1 (runs 7 9 and 4 6) were 2 or more times lower than during run 1; however, the whole fuel batch was added at once as well. The above-mentioned disadvantages of B1 keep on holding true, but the lower EFs were caused by faster devolatilization of these fuels, so the fuel batch started to burn rapidly. The fuel devolatilization and combustion are taking place separately in the under-fire boiler B2. The drawbacks of B1, such as poor mixing of gases, a heterogeneous temperature field, and bad control of combustion, can be attributed to B2 as well. On the other hand, the effect of fuel devolatilization is partially suppressed by the boiler design. Hence, the EFs from LI were 9431
dx.doi.org/10.1021/es2017945 |Environ. Sci. Technol. 2011, 45, 9427–9434
Environmental Science & Technology
ARTICLE
Figure 3. Mean values of emission factors of PeCBz and HxCBz with standard deviations. Runs 1 3 of PCBz were not analyzed due to matrix effects. Run 32 34 were based on only two values.
comparable to the other fuels. The obtained EFs from B2 were at comparable or slightly lower levels than the emissions from B1. The lowest EFs were found for the modern automatic boiler B3 with well-managed combustion by means of quasi-continuous fuel stoking by a screw conveyor and an air forced draft. The temperature of combustion was also sufficiently high (950 1000 °C was measured in the flame under the heat reflector, 600 800 °C was measured in the inlet of the heat exchanger). It resulted into the lowest levels of nonchlorinated products of incomplete combustion. Similarly, B4 and B5 are of modern design and do not feature any of the old-type boilers’ drawbacks. The temperature in the combustion chamber reached 800 1000 °C; in the fuel storage chamber the temperatures ranged 200 400 °C, so a predried and partially devolatilized fuel was coming onto the grate. Polychlorinated Benzenes. The bituminous coal (BC) combustion in the old-type boilers (B1 and B2, runs 4 6 and 12 13) produced at least 2 orders of magnitude higher levels of HxCBz than the other tested fuel boiler combinations did. The high HxCBz EFs were given by high chlorine content in BC and also by the boiler design because the HxCBz was an absolutely prevailing homologue of PCBz. Conditions shifting competitive chlorination/ dechlorination reactions toward chlorination occurred in both oldtype boilers. High emissions of PCBz were also observed during BC combustion in the modern B3 (runs 26 28), but the homologue profile was quite opposite with domination of TeCBz. Therefore, HxCBz EFs more than 100 times lower than those from B1 were found. High PCBz EFs could be expected from MP combustion because of fuel chlorine content above 0.1 wt %, but surprisingly the emissions were comparable to the other fuels except for BC. On the contrary, a higher EF of HxCBz was observed unexpectedly during SL combustion in the downdraft boiler B5 (runs 38 40). A higher HxCBz EF was also found for BL for the B5 boiler. To promote chlorination toward higher chlorinated homologues is probably a characteristic feature of B5 because HxCBz was the dominant PCBz species there. Homologue profiles shifted toward the higherchlorinated ones were found also for PCDD/F on the B5. Generally, the EF of HxCBz from beech logs combustion was in the range 0.575 10.2 μg/GJ, at least 10 times higher than the previously found values for woody biomass.5 Comparable HxCBz EFs were measured during oak combustion in a woodstove.4 For lignite combustion, the HxCBz EFs were in the range 0.524 0.852
μg/GJ, that is, lower values than from woody biomass combustion. Unfortunately, the results from “the worst” boiler B1 are missing due to matrix effects during analyses. In the case of PeCBz, the highest EFs were also found for BC combustion. The EF values were in the range 96.8 230 μg/GJ depending on the boiler’s design. For the remaining fuels and boilers, the EFs of PeCBz were between 0.0890 and 1.88 μg/GJ. Generally, the fuel determines mainly the total amount of emitted PCBz, but the homologue profile can be strongly affected by the boiler design. It is obvious mostly from the results of the BC combustion. In the case of B1 and B2, the HxCBz formed approximately 90% of the PCBz, while for B3 it formed only approximately 7%. The resulting EFs of PeCBz and HxCBz are shown in Figure 3. Polychlorinated Biphenyls, Dibenzo-p-dioxins and Dibenzofurans. The EFs of PCB and PCDD/F are summarized in Figure 4. Similarly to their precursors, the highest PCDD/F EFs were found for the two high-chlorine fuels (BC and MP). Much lower EFs were found for lignite and woody biomass fuels. As expected, the chlorine content in fuel was decisive for the PCDD/ F emission rate. Moreover, during BC combustion the EF values varied approximately 16 times depending on the boiler, hence the influence of the boiler design was evident as well. Surprisingly, the two highest values of EFs of I-TEQ PCDD/F were measured for the modern B3 when BC and MP were combusted. The old-type boilers B1 and B2 (runs 4 6 and 12 13) emitted 7 and 16 times lower PCDD/F than the modern boiler B3 (runs 26 28) during BC combustion. The EFs of I-TEQ PCDD/F from LI combustion varied from 3.16 to 60.7 ng/GJ depending on the boiler: much lower EFs were found for the modern-type boilers than for the old-type ones (B1 and B2). A similar trend was found for the EFs from BL combustion. However, the width of EFs range was shorter: 3.48 to 24.7 ng I-TEQ (PCDD/F)/GJ. Surprisingly, the same value of I-TEQ PCDD/F EFs was found during run 1 and runs 2 3 in the B1, that is, the runs with a different fuel stoking period. The EFs of PAH, CO, and PM were 2 4 times higher in run 1, but the I-TEQ PCDD/F were the same and WHO-TEQ PCB were even lower in run 1. It means that the much worse combustion in B1 did not increase the emission of I-TEQ PCDD/F. Very interesting is the comparison of the results from both downdraft boilers. The BL combustion in B4 (primary designed for lignite) produced 2 times higher EFs 9432
dx.doi.org/10.1021/es2017945 |Environ. Sci. Technol. 2011, 45, 9427–9434
Environmental Science & Technology
ARTICLE
Figure 4. Emission factors of PCB and PCDD/F. PCB is the sum of PCBs 77, 81, 105, 114, 118, 123, 126, 156, 157, 167, 169, 170, 180, and 189. PCDD/ F is the sum of tetra- to octa-CDD/F. TEQ values were determined according to EN 1948.
than that in B5 (primary designed for wood). Again, it is welldocumented that the mere dimension of the combustion chamber can substantially influence the emissions because of the different energy density of each fuel and thus the different change of the temperature gradient inside the combustion chamber and the postcombustion zones. Very intriguing is the comparison of EFs for BC combustion with other fuels because of the opposite trend of boiler influence on the I-TEQ PCDD/F levels. For BC combustion, the highest EF of I-TEQ PCDD/F was found for B3 and the lowest for B1, whereas for lignite combustion the converse was found. No special effect of boilers or fuels on the homologue profiles of PCDD/F was found, with the notable exception of B5. For B5, a shift of PCDD/F homologue profiles toward higher-chlorinated ones was observed. The emission of WHO-TEQ PCB showed the same trends depending on the fuel and boiler as the emission of I-TEQ PCDD/F. The PCB contributed 5.3% to the total TEQ values on average, which was a 10 times higher contribution than found for fireplaces and the stove.4
’ ASSOCIATED CONTENT
bS
Supporting Information. Detailed analytical procedure for organic compounds, and eight tables of emission factors of PM, CO, TOC, PAH, PCBz, PCPh, PCB, and PCDD/F related
to fuel mass and also to fuel energy content (LHV). This material is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Telephone: + 420 220 390 261; fax: +420 220 920 661; e-mail: [email protected].
’ ACKNOWLEDGMENT We gratefully acknowledge the financial support of the Ministry of the Environment of the Czech Republic (Project SP/1A2/116/07). ’ REFERENCES (1) Prevedouros, K.; Brorstr€ om-Lunden, E.; Halsall, C. J.; Jones, K. C.; Lee, R. G. M.; Sweetman, A. J. Seasonal and long-term trends in atmospheric PAH concentrations: evidence and implications. Environ. Pollut. 2004, 128, 17–27. (2) Breivik, K.; Vestreng, V.; Rozovskaya, O.; Pacyna, J. M. Atmospheric emissions of some POPs in Europe: a discussion of existing inventories and data needs. Environ. Sci. Policy. 2006, 9, 663–674. (3) Quass, U.; Fermann, M.; Br€oker, G. The European Dioxin Air Emission Inventory Project;Final Results. Chemosphere 2004, 54, 1319–1327. 9433
dx.doi.org/10.1021/es2017945 |Environ. Sci. Technol. 2011, 45, 9427–9434
Environmental Science & Technology
ARTICLE
(4) Gullett, B. K.; Touati, A.; Hays, M. D. PCDD/F, PCB, HxCBz, PAH, and PM emission factors for fireplace and woodstove combustion in the San Francisco Bay region. Environ. Sci. Technol. 2003, 37, 1758– 1765. (5) Hedman, B.; Naslund, M.; Marklund, S. Emission of PCDD/F, PCB, and HCB from combustion of firewood and pellets in residental stoves and boilers. Environ. Sci. Technol. 2006, 40, 4968–4675. (6) Bignal, K. L.; Langrudge, S.; Zhou, J. L. Release of polycyclic aromatic hydrocarbons, carbon monoxide and particulate matter from biomass combustion in a wood-fired boiler under varying boiler conditions. Atmos. Environ. 2008, 42, 8863–8871. (7) Wevers, M.; De Fre, R.; Vanermen, G. PCDD/F and PAH emissions from domestic heating appliances with solid fuel. Organohalogen Compd. 2003, 63, 21–24. (8) H€ubner, C.; Boos, R.; Prey, T. In-field measurments of PCDD/F emissions from domestic heating appliances for solid fuels. Chemosphere 2005, 58, 367–372. (9) Johansson, L. S.; Leckner, B.; Gustavsson, L.; Cooper, D.; Tullin, C.; Potter, A. Emission characteristics of modern and old-type residential boilers fired with wood logs and wood pellets. Atmos. Environ. 2004, 38, 4183–4195. (10) Lee, R. G. M.; Coleman, P.; Jones, J. L.; Lohmann, R. Emission factors and importance of PCDD/Fs, PCBs, PCNs, PAHs and PM10 from the domestic burning of coal and wood in the U.K. Environ. Sci. Technol. 2005, 39, 1436–1447. (11) Hedberg, E.; Kristensson, A.; Ohlsson, M.; Johansson, C.; Johansson, P. A.; Swietlicki, E.; Vesely, V.; Wideqvist, U.; Westerholm, R. Chemical and physical characterization of emission from birch wood combustion in a wood stove. Atmos. Environ. 2002, 36, 4823–4837. (12) Launhardt, T.; Thoma, H. Investigation on organic pollutants from a domestic heating system using various solid biofuels. Chemosphere 2000, 40, 1149–1157. (13) Kinsey, J. S.; Kariher, P. H.; Dong, Y. Evaluation of methods for the physical characterization of the fine particle emissions from two residential wood combustion appliances. Atmos. Environ. 2009, 43, 4959–4967. (14) Lavric, E. D.; Konnov, A. A.; De Ruyck, J. Dioxin levels in wood combustion: A review. Biomass Bioenergy 2004, 26, 115–145. (15) Ravindra, K.; Sokhi, R.; Van Grieken, R. Atmospheric polycyclic aromatic hydrocarbons: Source attribution, emission factors and regulation. Atmos. Environ. 2008, 42, 2895–2921. (16) Junninen, H.; Mønster, J.; Rey, M.; Cancelinha, J.; Douglas, K.; et al. Quantifying the impact of residential heating on the urban air quality in a typical European coal combustion region. Environ. Sci. Technol. 2009, 43, 7964–7970. (17) van Loo, S.; Koppejan, J. The Handbook of Biomass Combustion & Co-firing; Earthscan: London, 2008. (18) Horak, J.; Hopan, F.; Krpec, K.; Dej, D.; Kubacka, M.; Pekarek, V.; Syc, M.; Ocelka, T.; Tomsej, T.; Machalek, P. Determination of emission factors for combusting solid fuels in residential combustion appliances. Organohalogen Compd. 2008, 70, 2470–2473. (19) Horak, J.; Hopan, F.; Syc, M.; Machalek, P.; Krpec, K.; Ocelka, T.; Tomsej, T. Estimation of selected pollutants emission from solid fuels combustion in small appliances. Chem. Listy 2011, 105 (11). (20) McDonald, D. J.; Zielinska, B.; Fujita, E. M.; Sagebiel, J. C.; Chow, J. C.; Watson, J. G. Fine particle and gaseous emission rates from residential wood combustion. Environ. Sci. Technol. 2000, 34, 2080– 2091. (21) Kubica, K.; Paradiz, B.; Dilara, P. Small combustion installations: Techniques, emissions and measures for emission reduction. Joint Research Centre Scientific and Technical Reports, EUR 23214 EN, 2007. (22) Kastanski, E.; Vamvuka, D.; Grammelis, P.; Kakaras, E. Thermogravimetric studies of the behavior of lignite-biomass blends during devolatilization. Fuel Process. Technol. 2002, 77 78, 159–166.
9434
dx.doi.org/10.1021/es2017945 |Environ. Sci. Technol. 2011, 45, 9427–9434
ARTICLE pubs.acs.org/est
CO2 Saturation, Distribution and Seismic Response in Two-Dimensional Permeability Model Hamid Behzadi, Vladimir Alvarado,*,† and Subhashis Mallick‡ † ‡
Chemical and Petroleum Engineering Department, University of Wyoming, Laramie, Wyoming 82071, United States Geology and Geophysics Department, University of Wyoming, Laramie, Wyoming 82071, United States
bS Supporting Information ABSTRACT: Carbon dioxide capture and storage (CCS) has been actively researched as a strategy to mitigate CO2 emissions into the atmosphere. The three components in CCS are monitoring, verification, and accounting (MVA). Seismic monitoring technologies can meet the requirements of MVA, but they require a quantitative relationships between multiphase saturation distributions and wave propagation elastic properties. One of the main obstacles for quantitative MVA activities arises from the nature of the saturation distribution, typically classified anywhere from homogeneous to patchy. The emerging saturation distribution, in turn, regulates the relationship between compressional velocity and saturation. In this work, we carry out multiphase flow simulations in a 2-D aquifer model with a log-normal absolute permeability distribution and a capillary pressure function parametrized by permeability. The heterogeneity level is tuned by assigning the value of the DykstraParson (DP) coefficient, which sets the variance of the log-normal horizontal permeability distribution in the entire domain. Vertical permeability is a 10th of the horizontal value in each gridcell. We show that despite apparent differences in saturation distribution among different realizations, CO2 trapping and the Vp-Sw Rock Physics relationship are mostly functions of the DP coefficient. When the results are compared with the well accepted limits, GassmannWood (homogeneous) (A Text Book of Sound; G. Bell and Suns LTD: London, 1941) and GassmannHill (patchy) models, the Vp-Sw relationship never reaches the upper bound, that is, patchy model curve, even at the highest heterogeneity level in the model.
’ INTRODUCTION Carbon dioxide capture and storage (CCS) has been proposed as a viable approach to mitigate CO2 emissions from large point sources such as coal fired power and chemical processing plants.1,2 Monitoring, verification, and accounting (MVA) of CO2 are three components in CCS. The primary objectives of MVA is to identify and quantify CO2 migration in geological media and help to identify any leakage of stored CO2. These objectives highlight the importance of multiphase flow and its controlling properties, namely relative permeability and capillary pressure, which govern how phases migrate in subsurface porous media and consequently how rock elastic properties change during storage operations. Hysteresis of multiphase flow functions plays a significant role in CCS. In addition, hysteresis trapping can be enhanced if, for example, CO2 is injected at the bottom of the formation. In this approach, once CO2 is injected at the bottom of an aquifer, it will spontaneously migrate upward, leading to consecutive drainage and imbibition processes. Moreover, as CO2 moves up, it dissolves in water and is stored. These two storage mechanisms are characterized by being fast, typically in the scale of years, as opposed to hundreds of years or more for more permanent storage. Saadatpour3 showed that local capillary trapping may significantly contribute to phase distribution and trapping. A region r 2011 American Chemical Society
with high entry capillary pressure may act as a barrier to drainage by CO2 and redirect the flow laterally. Therefore, this mechanism is associated with heterogeneous systems. The redirection of flow increases the odds of both hydraulic and dissolution trapping. In the first part of this paper, we consider hysteresis and local capillary trapping effects on trapping, and consequently on saturation distribution. This will impact elastic properties, as shown in the second part of this work. Mineral trapping is not included in this work, but it should become important after hundreds of years. The second part of this study focuses on seismic monitoring. Monitoring is an essential tool to design and fine-tune improved oil recovery, optimize CO2 storage and mitigate leakage risks. To accomplish this, it becomes necessary to map saturation and pressure throughout the formation over the time span of interest. Time-lapse seismic has been utilized to map these changes mostly qualitatively.47 A quantitative interpretation of saturation changes may be possible, if the case-specific velocitysaturation relationship for a system is known. In practice, there are two main velocity-saturation (Vp-Sw) models based on two extremes of phase-distribution type: uniform Received: June 12, 2011 Accepted: September 22, 2011 Revised: August 22, 2011 Published: September 22, 2011 9435
dx.doi.org/10.1021/es201969a | Environ. Sci. Technol. 2011, 45, 9435–9441
Environmental Science & Technology and patchy. In GassmannWood8 model (uniform), the size of patches is smaller than the fluid diffusion length. In contrast, if the size of patches is larger than the fluid diffusion length, then there is no pressure communication between gas pockets during one wave cycle. In this latter case, the elastic wet rock properties can be estimated using GassmannHill equation.9 These two bounds are also named low and high frequency limits. However, in reality, the system-specific velocity-saturation relationship is in the intermediate frequency domain. There are very few velocity-saturation models available in the open literature. White’s model10 is limited to regular distribution with simple geometry, whereas Muller11 provides the velocitysaturation relationship for partially saturated rock. Muller assumes that the distribution of fluid patches form a two-phase, self-affine mono fractal,11 not considering geology and multiphase flow. Muller then introduces this fractal saturation map into Toms’ model12 to compute elastic rock properties and consequently Vp-Sw relationship as a function of the Hurst exponent, ν, which controls the persistence or antipersistence of the random fractal. In this model, each cell is saturated with either a brine or a gas, but not both. In our study, saturation map is the response of multiphase flow functions to cells properties of the static model. In summary, we first build a geological model with randomly uncorrelated permeability distribution and carefully allocate rockfluid properties, that is, relative permeability and capillary pressure. An equation of state is used to predict thermodynamic fluid properties. The saturation and elastic rock properties are simulated on fine-scale models and then upscaled to a coarser one using Backus’s method,13 in order to derive velocity-saturation relationships.
’ MATERIALS AND METHODS There are four main CO2 storage mechanisms, namely structural, hysteresis, dissolution, and mineral trapping. However, local heterogeneity may result in significant contrast in static and dynamic properties, as to produce preferential paths as well as local barriers to flow in heterogeneous systems, a process coined local capillary trapping.3 This trapping mechanism is different from hysteresis trapping in terms of scale and magnitude. The simulations are carried out in a compositional mode, which more realistically represents solubility of CO2 in brine. It also allows us to track a more realistic thermodynamical behavior, and hence it captures density changes accurately. Structural Model. Local capillary trapping effects on buoyancy-driven migration is a central motivation of this work. The model consists of a 2D aquifer with a source of CO2 at the bottom. In addition, the system boundaries, namely top, bottom and sides represent no-flow conditions, which should enhance countercurrent flow. The model has three main sections: source at the bottom where CO2 is initially located, main reservoir body in the middle, at very fine resolution of 1 m-X 0.3 m-Z cells and a sink at top, where CO2 migrates. Fine-scale models are of interest to capture the fine details of multiphase flow, trapping and velocity-saturation (Vp-Sw) relationship. Grid size influences trapping and Vp-Sw relationship. For instance, if grid size is coarse, then buoyancy may overcome capillary entry pressure. Petrophysical Model. Permeability and the corresponding capillary pressure are assigned on per grid-cell basis, while porosity and relative permeability are constant throughout the domain. The distribution of horizontal permeability is randomly
ARTICLE
uncorrelated according to a log-normal Gaussian distribution. The DykstraParson coefficient (DP) parametrizes the variance of the log-normal distribution conveniently. A value of DP = 0 means that the variance is zero. The larger the value of DP, the larger the variance and therefore the system is more heterogeneous. Vertical permeability is assigned to be 10 times smaller than the grid-cell horizontal permeability. The base case is generated with DykstraParson coefficient 0.7 (DP = 1 e(σK)) and a permeability mean of 200 md. Leveret J-function is utilized to parametrize capillary pressure as a function of permeability. sffiffiffi Pc k ð1Þ JðSw Þ ¼ σ cos θ ϕ where θ is the contact angle, and k and ϕ are the absolute permeability and porosity, respectively. Chequet14 showed that contact angle variation of 4050° and 1525° for quartz and mica respectively, can occur as pressure changes. Chequet15 and Chalbaud16 also showed that pressure affects IFT. However, wettability and interfacial tension (IFT) are assumed constant. Hysteresis is limited to the relative permeability model, similar to Killough’s model.17 A maximum gas residual saturation of 0.286 is used in all simulations. Capillary pressure and relative permeability are presented in Figure S1, Supporting Information (SI). Compressional-Wave Formulation. Compressional wave velocity in the matrix is a function of bulk and shear moduli, and density of the medium. Compressional or longitudinal wave or P-wave, Vp, can be estimated in partially saturated rock through Gassmann’s equation, eq 2,18 where Ksat, μ, and Fsat denote the saturated-bulk, shear moduli and density of the saturated rock. Based on this equation, bulk modulus is function of saturations of phases present in the pores, while shear modulus is not, therefore, μsat = μ (μ = 8.4 GPa). vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u uKsat 4 μ t 3 ð2Þ Vp ¼ Fsat Gassmann equation relates saturated bulk modulus to fluid-bulk, dry-bulk, matrix and porosity denoted by Kf, Kd (Kd = 8.3 GPa) Km (Km = 38 GPa) and ϕ, as follows: Kd 2 1 Km Ksat ¼ Kd þ ð3Þ ϕ 1 ϕ Kd þ 2 Kf Km Km We assume ideal mixing, therefore, mixing properties equal pure phase properties weighted by saturation. GassmannWood,8 eq 4, and GassmannHill,19 eq 5, represent the lower and upper bounds of mixing, i.e. uniform and patchy.9 The uniform case describes a mixing level where all pores have the same fraction of phases while patchy corresponds to situation where some pores are saturated with CO2 and the others saturated by water (no hydraulic connectivity between gas pockets in a single wave cycle). The mixing is somewhere between these two bounds, which might be estimated by Brie’s model.20
9436
1 Sw 1 Sw ¼ þ Kf Kw KCO2
ð4Þ
Kf ¼ Sw Kw þ ð1 Sw Þ KCO2
ð5Þ
Kf ¼ ðSw Þe Kw þ ð1 Sw Þe KCO2
ð6Þ
dx.doi.org/10.1021/es201969a |Environ. Sci. Technol. 2011, 45, 9435–9441
Environmental Science & Technology
ARTICLE
where Sw, Kw (Kw = 3 GPa) and KCO2 (KCO2 = 0.046 GPa) are water saturation, and water and CO2 moduli. Brie’s formula, eq 6, becomes Hill’s model, eq 5, when e = 1 and approaches Wood’s model, eq 4, when e is high. The saturated-bulk density, Fsat, can be estimated through eqs 7 and 8. In this study, the density of each phase is calculated using an equation of state, which provides better estimates of density. Fsat ¼ Fd þ ϕ Ff
ð7Þ
Ff ¼ Sw Fw þ ð1 Sw Þ FCO2
ð8Þ
where Ff, Fd, Fw, and FCO2 are fluid, dry-bulk, water and CO2 density. Compressional Wave Velocity Upscaling. In the numerical simulation, properties are homogeneous at grid scale. Therefore, GassmannWood equation can be used to compute velocity as function of saturation, bulk and shear moduli. If the simulation is a seismic scale then velocity-saturation follows GassmannWood equation. Therefore, use of the finer scale and its consequent upscaling is necessary to calculate Vp-Sw relation. Here, the simulation resolution is 1 m-X 0.3 m-Z and Backus13 method has been utilized to upscale data to the seismic scale. Bakcus13 technique involves averaging the elastic moduli and bulk density of thin layers/grids into average properties similar to coarser thick layer/grid. This method has been successfully used for upscaling well logs to seismic wavelength.21 Sequential Backus averaging assumes a normal incidence ray path. That is, the stack of thin layers is horizontal and wavefront is normal to bedding, which is the case in this particular simulation. However, upscaling should be executed carefully because it may not preserve exact Vp-Sw relationship as Mukerji and Mavko22 showed. Backus’s method can produce variability of Vp-Sw relationship with a good estimate if Backus number B, eq 9, is less than a third.23 Liner and Fei23 results show that the simulated wave-field is similar for the original model and the averaged one if B is less than a third (scattered limit). Under this limit, scattered field and transmitted field will be preserved intact. B¼
f L0 minðVs Þ
ð9Þ
where f, L0 , and min(Vs) are the seismic frequency, averaging length and minimum shear velocity after upscaling. Here, we assumed the seismic frequency to be 35 Hz.
’ RESULTS The evolution of the CO2 saturation is depicted in Figure 1. Comparison between Figure 1b and Figure 1d shows that local capillary trapping does not keep CO2 at high saturation for a long period of time, though it stabilizes it at a still significant values. The CO2 saturation becomes essentially stationary after 40 years, but the simulation time was extended to 140 years for completeness. Saturation distribution changes from realization to realization as fluid flows (Figure 3). The rest of this section is organized as follow: first heterogeneous and homogeneous system are compared, and then accessibility ratio and trapping saturation among different realizations are compared. The Vp-Sw relationship as a function of time, specific realization and heterogeneity level are presented in the last part. Accessibility ratio is defined here as the ratio of CO2 accessed porous area to the total area, which is equivalent to sweep efficiency in the oil recovery context.
Figure 1. CO2 upward migration snapshots for DP = 0.7 at different times, simulation start time (a), 7 years (b), 40 years (c), and 140 years (d).
Homogenous vs Heterogenous. Heterogeneity increases CO2 upward migration time 1.5 to 3 fold. The maximum average gas saturation in the body of the homogeneous system happens after 8 years, which is much sooner than that of heterogeneous systems, e.g. 26, 28, and 33 years for realizations 1, 2, and 3. Figure 2 shows how the saturation for three different realizations (DP = 0.7) increases at early time and then stabilizes at the end of simulation. These results are consistent with Flett’s observations.24 He constructed different geostatistical models honoring the same distribution. In the models, porosity and permeability are subject to change. Flett concluded that heterogeneity can serve as an additional containment mechanism, although he did not consider capillary pressure heterogeneity. Heterogeneity results in more fragmented migration and lower accessibility ratio. The modeled accessibility ratio and trapped saturation are 100% and 9% in the homogeneous model, while they are 65% and 20% in the heterogeneous model, respectively. The accessibility ratio and gas trapping for different DP value are shown in Table 1. As expected, the average hydraulic trapping increases as DP increases, while the accessibility ratio reduces. The multiplication of these two parameters tells us that the total hydraulic trapping is higher in the heterogeneous system, therefore, the higher the heterogeneity, the more hydraulic trapped gas there will be the system.
total hydraulic trapping ¼ average hydraulic trapping accessibility ratio The effect of capillary pressure heterogeneity is often overlooked in the literature. Figure S2, SI, shows the contribution of capillary pressure heterogeneity on total hydraulic trapping versus time as well as accessibility ratio . It is observed that capillary 9437
dx.doi.org/10.1021/es201969a |Environ. Sci. Technol. 2011, 45, 9435–9441
Environmental Science & Technology
ARTICLE
Figure 2. Average gas saturation as a function of time for three realizations at DP = 0.7, and a homogeneous mode. Gas saturation increases at early time and stabilizes at the end.
Table 1. Accessibility Ratio and Trapped Gas Saturation for Different Heterogeneity Level after 140 Years DP 0
accessibility ratio (%) 100
average trapped gas saturation (%) 9.0
0.3
97.8
9.8
0.5
80.9
12.3
0.7
63.6
19.5
0.9
42.2
30.3
pressure heterogeneity may increase the total hydraulic trapping up to 21%. Trapping saturation and Accessibility ratio. Saturation maps vary from realization to realization as shown in Figure 3. However, trapping saturation and accessibility ratio are the same. To show this, DP values of 0.3 0.5, 0.7, and 0.9 are used to generate a wide range of heterogeneity in the model. Trapping saturation and accessibility ratio are plotted as functions of the DP in Figure 4. As the error bar shows, trapping and accessibility do not vary much from realization to realization at a given value of DP. The standard deviation (error bar size) of the accessibility ratio is relatively high for DP = 0.9, perhaps because of the small system size. Vp-Sw Relationship. Elastic rock properties at each fine scale cell are computed based on Wood model and then upscaled using Backus method (B = (1/3), upscaled grids are 56 times larger in Z direction). Finally, Vp and Sw are plotted for each individual coarse grid. Figure 5 shows the Vp-Sw trend for realizations 1, 2, and 3 after 5 years when DP = 0.7. Although various saturation
Figure 3. Gas saturation distribution after 26 years for three different realizations, DP = 0.7.
distribution maps can be seen for these realizations (Figure 3), they follow similar Vp-Sw trend. Additional realizations results were consistent with this observed response. Therefore, as trapping and accessibility ratio, velocity-saturation relationship is similar among different realizations for the same DP value. The Vp-Sw relationship becomes more GassmannWood’s equation-like as time progresses, mainly because the saturation distribution becomes more uniform with time (Figure S3, SI). The velocity vs saturation is plotted for years 2013, 2015, 2020, 2040, 2060, and 2150 (simulations start in year 2010). Results for 9438
dx.doi.org/10.1021/es201969a |Environ. Sci. Technol. 2011, 45, 9435–9441
Environmental Science & Technology
ARTICLE
Figure 4. Trapped gas saturation and accessibility ratio vs time for various levels of heterogeneity; these values are calculated for different realizations and their standard deviations are plotted as error bars.
Figure 5. Vp-Sw relationship for DP = 0.7 in realizations of 1, 2, and 3, after 5 years.
Figure 6. Fitted Brie’s model to different Vp-Sw relationship data sets.
years 2013, 2015, and 2020 are naturally grouped, while those for 2060 and 2150 appear to fall in the different category. The 2040s relationship sits between the two aforementioned groups. It can be noticed, as expected, that the Vp-Sw relationship for DP = 0.3 is very close to GassmannWood model as opposed to DP = 0.9, which is in between these two bounds. The velocity-saturation relationship for high heterogeneity (DP = 0.9) does not follow GassmannHill model, which is consistent with observed field evidence. The experimental evidence shows that the velocitysaturation relationship is close to the lower bound, Gassmann Wood, despite the existence of patches.11 Brie’s model may not be appropriate for prediction of velocitysaturation relationship. The model is fitted to all four DP sets of data and Brie’s exponent is estimated. It appears that the Brie’
model cannot predict well between upper and lower boundary frequency. Muller also recommended to employ Biot’s equation of dynamic poroelasticity rather than Brie’s experimental model.11 Figure 6 shows how Brie’s model fits to different VpSw data sets. Comparison with Other Results. Lebedev et al.25 utilized CT scan to observe patchy saturation and its effect on ultrasonic velocity. He also compared his observation with numerical simulation results of wave propagation in 2D using a finite-difference solver for Biot’s equations of dynamic poroelasticity. Although the patchy saturation distribution derived from CT scan was not used in the 2D numerical simulation, the results captured the overall behavior of the measured velocity-saturation relationship. Muller’s 11showed that the shape of the Vp-Sw relationship relates to the size and shape of patches. In Muller’s study, ν controlled 9439
dx.doi.org/10.1021/es201969a |Environ. Sci. Technol. 2011, 45, 9435–9441
Environmental Science & Technology the anisotropic scaling of a random fractal. The greater ν is, the fewer, but bigger the patchy clusters are. Therefore, for the same saturation, the higher ν is, the more the patches are observed (closer to GassmannHill bound). Our results cover both Lebedev’s and Muller’s results. Based on our results, the velocitysaturation relationship can be shown to differ for different heterogeneity levels, which explains Muller’s conclusions. However, for different realizations of similar heterogeneity, this relationship is essentially identical, which explains Lebedev’s conclusion. In this sense, our modeling efforts help to place apparently opposite results in a coherent context. Limitation. For the sake of simulation numerical stability, the dissolution of CO2 in water is not considered, and therefore, the mixture modulus equals CO2 and water moduli weighted by volume fraction. However, the solubility of CO2 in water can be significant as Gilfillan et al.26 and Behzadi27 showed, and can alter the bulk modulus noticeably at different equilibrium stages, as shown in Vanorio et al.’s laboratory study,28 which means that bulk modulus dynamically changes along with encroaching CO2 flooding front, and this alteration is also valid for matrix itself (reaction with rock). Therefore, CO2-brine-rock interaction not only changes porosity, permeability and density, but also baseline properties for Gassmann fluid substitution scheme. GassmannWood’s equation may not hold at very low saturation and low frequency for liquidgas system where there is phase conversion under pressure oscillation characteristics of the acoustic wave. In this limit, LandauLifshitz’s method might be used, instead. Wood’s model may result in optimistic approximation of velocity and consequently underestimation of gas saturation.29,30 Employing LandauLifshitz’s method, thermodynamically equilibrated, as opposed to Wood’s method, frozen, causes even better detectability of gas fingerprint at very low saturation. LandauLifshitz’s method Shifts down the lower boundary and increases the envelop volume between two boundaries. This increase in the envelop volume helps clearer effect of time and DP on Vp-Sw relationships, that is, separates more Vp-Sw relationship of DP = 0.5 from that of DP = 0.7.
’ DISCUSSION This study sheds light on how heterogeneity affects trapping, storage and velocity-saturation relationship. This is important to quantitatively enhance subsurface CO2 monitoring. This study also shows that limited number of stochastic realizations might be sufficient, at least in terms of producing trapping, storage, VpSw relationship estimates. In this study, capillary pressure was parametrized with respect to permeability. The results show that absolute permeability, and hence capillary pressure heterogeneity significantly affect trapping. Moreover, both static heterogeneity, namely absolute permeability heterogeneity, and dynamic heterogeneity, associated with multiphase flow function distributions, can improve hydraulic trapping, but may reduce accessibility ratio and dissolution trapping. The aforementioned heterogeneity types, that is, static and dynamic, significantly constraint CO2 migration toward and beneath a caprock, consequently reducing the risk of CO2 leakage. This outcome, as shown in the results, denotes the importance of properly incorporating this flow controls, often neglected in modeling exercises. Results for multiple realizations at the value of the Dykstra Parson coefficient clearly show that total hydraulic trapping and accessibility ratio are weak functions of the specific realization of given level of heterogeneity in the random model. Even more interesting in
ARTICLE
the fact that the Rock Physics response of interest for seismic monitoring, for example, the velocity-saturation relation, is mainly a function of the heterogeneity level (DP), a varies insignificantly from realization to realization at the same value of the DykstraParson coefficient. This means that in order to quantify (bookkeep) CO2 in storage sites by means of time-lapse seismic, it is considerably more important to know the nature of the distribution of heterogeneity and not so much the specific details.
’ ASSOCIATED CONTENT
bS
Supporting Information. Relative permeability and capillary pressure, Variation of velocity-saturation through elapsed time and effect of capillary pressure heterogeneity are presented in Figures S1, S2 and S3 respectively. This information is available free of charge via the Internet at http://pubs.acs.org/
’ AUTHOR INFORMATION Corresponding Author
*E-mail: [email protected].
’ ACKNOWLEDGMENT We acknowledge CMG, Ltd. for providing academic licenses of the software suite. We thank the Enhanced Oil Recovery Institute at the University of Wyoming for financial assistance. Financial support was provided by DOE through grant DEFE0001160. ’ REFERENCES (1) Martinsen, D.; Linssen, J.; Markewitz, P.; V€ogele, S. CCS: A future CO2 mitigation option for Germany? A bottom-up approach. Energy Policy 2007, 35, 2110–2120. (2) Odenberger, O.; Johnsson, F. The role of CCS in the European electricity supply system. Energy Procedia 2009, 1, 4273–4280. (3) Saadatpour, E.; Bryant, S. L.; Sepehrnoori, K. New trapping mechanism in carbon sequestration. Transp. Porous Media 2010, 82, 3–17. (4) M€uller, N.; Ramakrishnan, T.; Boyda, A.; Sakruai, S. Time-lapse carbon dioxide monitoring with pulsed neutron logging. Int. J. Greenhouse Gas Control 2007, 1, 456–472. (5) Chadwick, R. A.; Noy, D.; Arts, R.; Eiken, O. Latest time-lapse seismic data from Sleipner yield new insights into CO2 plume development. Energy Procedia 2009, 1, 2103–2110. (6) Urosevic, M.; Pevzner, R.; Shulakova, V.; Kepic, A.; Caspari, E.; Sharma, S. Seismic monitoring of CO2 injection into a depleted gas reservoir Otway Basin Pilot Project, Australia. Energy Procedia 2011, 4 3550–3557. (7) Preston, C.; Monea, M.; Jazrawi, W.; Brown, K.; Whittaker, S.; White, D.; Law, D.; Chalaturnyk, R.; Rostron, B. IEA GHG Weyburn CO2 monitoring and storage project. Fuel Process. Technol. 2005, 1547–1568. (8) Wood. A Text Book of Sound; G. Bell and Suns LTD: London, 1941. (9) Mavko, G.; Mukerji, T. Bounds on low-frequency seismic velocities in partially saturated rocks. Geophysics 1998, 63, 918–924. (10) White, J. Computed seismic speeds and attenuation in rocks with partial gas saturation. Geophysics 1975, 40, 224–232. (11) Muller, T. M.; Toms-Stewart, J.; Wenzlau, F. Velocity-saturation relation for partially saturated rocks with fractal pore fluid distribution. Geophys. Res. Lett. 2008, 35, L09306. (12) Toms, J.; Muller, T. M.; Curevich, B. Seismic attenuation in porous rocks with random patchy saturation. Geophys. Prospect. 2007, 55, 671–678. 9440
dx.doi.org/10.1021/es201969a |Environ. Sci. Technol. 2011, 45, 9435–9441
Environmental Science & Technology
ARTICLE
(13) Backus, G. E. Long-wave elastic anisotropy produced by horizontal layering. J. Geophys. Res. 1962, 67, 4427. (14) Chiquet, P.; Broseta, D.; Thibeau, S. Wettability alteration of caprock minerals by carbon dioxide. Geofluids 2007, 7, 112–122. (15) Chiquet, P.; Daridon, J. L.; Broseta, D. e. a. CO2/water interfacial tensions under pressure and temperature conditions of CO2 geological storage. Energy Convers. Manage. 2007, 48, 736–744. (16) Chalbaud, C.; Robin, M.; Lombard, J. M.; et al. Interfacial tension measurements and wettability evaluation for geological CO2 storage. Adv. Water Resour. 2009, 32, 98–109. (17) Kllough, J. E. Reservoir simulation with history-dependent saturation functions. SPE J. 1976, 16. € ber die Elastizit€at por€oser Medien: in Z€urich. (18) Gassmann, F. U Vierteljahrsschr. Naturforsch. Ges. Zuerich 1951, 96, 1–23. (19) Hill, R. Elastic properties of reinforced solids: Some theoretical principles. J. Mech. Phys. Solids 1963, 11, 357–372. (20) Brie, A.Pampuri, F.Marsala, A. F.Meazza, O.Shear Sonic Interpretation in Gas-Bearing Sands; SPE Technical Paper 30595; Society of Petroleum Engineers: Dallas, TX, 1995. (21) Lindsay, R.Koughnet, V. R.Sequential backus averaging: Upscaling well logs to seismic wavelengths. The Leading Edge, 2001 (22) Mukerji, T.; Mavko, G. Assessing uncertainty in rock physics interpretations: The pitfalls of ignoring variability. In 6th International Conference and Exposition on Petroleum Geophysics, 2006, (23) Liner, C.; Fei, T. The Backus Number. The Leading Edge 2007, 26, 420–426. (24) Flett, M.; Gurton, R.; Weir, G. Heterogeneous saline formations for carbon dioxide disposal: impact of varying heterogeneity on containment and trapping. J. Petrol. Sci. Eng. 2007, 57, 106–118. (25) Lebedev, M.; Toms-Stewart, J.; Clennell, B.; Pervukhina, M.; Shulakova, V.; Paterson, L.; Muller, T. M.; Gurevich, B.; Wenzlau, F. Direct laboratory observation of patchy saturation and its effects on ultrasonic velocities. The Leading Edge 2009, 28, 24–27. (26) Gilfillan, S. M. V.; Sherwood Lollar, B.; Holland, G.; Blagburn, D.; Stevens, S.; Schoell, M.; Cassidy, M.; Ding, Z.; Zhou, Z.; LacrampeCouloume, G.; Ballentine, C. J. Solubility trapping in formation water as dominant CO2 sink in natural gas fields. Nature 2009, 458, 614–618. (27) Behzadi, H. Comparison of Chemical and Hysteresis CO2 Trapping in the Nugget Formation 2, SPE Technical Paper 133463; Society of Petroleum Engineers: Florence, Italy, 2010, (28) Vanorio, T.; Mavko, G.; Vialle, S.; Spratt, K. The rock physics basis for 4D seismic monitoring of CO2 fate. The Leading Edge 2010, 29, 156–162. (29) Broseta, D.; Khalid, P.; Nichita, D.; Galliero, G.; FavrettoCristini, N.; Blanco, J. A new look at the seismic properties of low gassaturated reservoirs. In 71st EAGE Conference & Exhibition Held in Amsterdam, 2009, (30) Nichita, D.; Khalid, P.; Broseta, D. Calculation of isentropic compressibility and sound velocity in two-phase fluids. Fluid Phase Equilib. 2009, 291, 95–102.
9441
dx.doi.org/10.1021/es201969a |Environ. Sci. Technol. 2011, 45, 9435–9441
ARTICLE pubs.acs.org/est
Mercury Emissions from Biomass Burning in China Xin Huang,† Mengmeng Li,† Hans R. Friedli,‡ Yu Song,†,* Di Chang,† and Lei Zhu† †
State Key Joint Laboratory of Environmental Simulation and Pollution Control, Department of Environmental Science, Peking University, Beijing 100871, China ‡ National Center for Atmospheric Research, Boulder, Colorado 80307, United States
bS Supporting Information ABSTRACT: Biomass burning covers open fires (forest and grassland fires, crop residue burning in fields, etc.) and biofuel combustion (crop residues and wood, etc., used as fuel). As a large agricultural country, China may produce large quantities of mercury emissions from biomass burning. A new mercury emission inventory in China is needed because previous studies reflected outdated biomass burning with coarse resolution. Moreover, these studies often adopted the emission factors (mass of emitted species per mass of biomass burned) measured in North America. In this study, the mercury emissions from biomass burning in China (excluding small islands in the South China Sea) were estimated, using recently measured mercury concentrations in various biomes in China as emission factors. Emissions from crop residues and fuelwood were estimated based on annual reports distributed by provincial government. Emissions from forest and grassland fires were calculated by combining moderate resolution imaging spectroradiometer (MODIS) burned area product with combustion efficiency (ratio of fuel consumption to total available fuels) considering fuel moisture. The average annual emission from biomass burning was 27 (range from 15.1 to 39.9) Mg/year. This inventory has high spatial resolution (1 km) and covers a long period (20002007), making it useful for air quality modeling.
1. INTRODUCTION Mercury is a toxic and persistent environmental pollutant. Atmospheric mercury is deposited in various pathways into the ground and water. Some of the mercury is transformed into methyl mercury, which bioaccumulates and biomagnifies in food webs, resulting in increased concentrations in higher organisms. Mercury remains an important subject of global pollution control efforts because of its toxicity and its involvement in the atmosphere-biosphere biogeochemical cycle and long-range transport. Mercury emissions from biomass burning have recently received increasing attention due to their potentially significant contribution to the atmospheric mercury budget and particularly their impact on the global mercury cycle.1 Earlier studies have estimated that the average global annual mercury emitted from biomass burning for 19972006 was 675 ( 240 Mg/year. This is equivalent to 8% of all currently known anthropogenic and natural mercury emission sources for the same period.2 During the biomass burning process, mercury can be remobilized and reemitted into the air. Biomass burning therefore accelerates emission and deposition cycles of mercury between terrestrial ecosystems and the atmosphere. China is rich in mercury mineral resources; moreover, as a developing country, the amount of coal consumed in China approaches approximately 28% of the world’s total consumption. Previous work has shown that atmospheric mercury emissions from nonferrous metal smelting and coal combustion in China could be the highest in the world.3 However, as a large agricultural r 2011 American Chemical Society
country, the mercury emissions from biomass burning in China could also be significant. The majority of the population lives in rural areas and biofuel (including crop residues, fuelwood, and animal dung) is an important energy source. Moreover, open burning, such as forest fires, grassland fires, and burning of field crop residues in rural areas also release mercury.4 A few attempts have been made to estimate the mercury emission from biomass burning to the atmosphere in China. Streets et al. estimated that 19 Mg of mercury was released from biomass burning in China during 1999, including grassland burning, forest burning, biofuel combustion and agricultural residues burning.4 This inventory was developed on the basis of a variety of statistical data for the 1950s-1990s. The data employed in that work are outdated, as the burning activities in China have changed significantly during the last two decades, especially forest fires. Friedli et al. have evaluated that 7 ( 2 Mg mercury was released yearly from biomass burning other than biofuel combustion in Central Asia during 19972006.2 Another limitation is that the emission factors adopted in these two studies were measured in North America, which might be inappropriate for China. China is known for high levels of atmospheric mercury from coal combustion and nonferrous metals smelting, which results in high deposition of Received: June 29, 2011 Accepted: September 27, 2011 Revised: September 21, 2011 Published: September 27, 2011 9442
dx.doi.org/10.1021/es202224e | Environ. Sci. Technol. 2011, 45, 9442–9448
Environmental Science & Technology
9443
L and s in parentheses refer to leaves and stems, respectively. Values in parentheses are the ranges of EFs.
7.89.8(s),9 3741 841394(l), 9221(s),5 70130(l), 1121(s),45 3054(s),46 122(s) 47 319 (841394)5
24 24
3.044 crop residue
grass
b a
15883(l), 6233(s), 24,40 3438 40 40
35 (9221)5
60,1 4272,6 157(s),14 848(l)14 6, 183943 2
2135(l)
40 26
796(l),36 18122(l), 668(s),37 3145(l), 1737(s),38 2332(l),39 100 (21204)26 5.035 forest and fuelwood
40 (1776)26
China leaves stems biomass of stem/leaf
Table 1. Emission Factors Used for Different Biomass Categories
Emission Factors. Mercury emission was estimated using the product of the amount of biomass burned and emission factors (EFs). Previous researches adopted the EF values measured in North America to estimate the mercury emissions in China because native measurements of EFs in China were not available. Several studies showed that mercury contained in vegetation (live, dead, coniferous, and deciduous) was essentially completely released in burns. The speciation of the emitted mercury is primarily in the form of gaseous elementary mercury (GEM) for both dry and green fuels.6,9 Mercury concentrations in vegetation and soil are treated as the EFs in some estimates of mercury emissions from biomass burning.68 Fortunately, a few recent studies reported the mercury concentrations in various biomes in China. Because the majority of forest fires in China occur in the northeast regions,10 mercury concentration measurements carried out in this area were chosen as EFs for Chinese forest fires. Since more crop residues were burned in south China,11 we chose the experimental mercury concentration results from the southern part of the country as the EFs for crop residue burning. The biomass ratio of stems/leaf and EFs are averaged from the reported results, as listed in Table 1. The emissions from stem and leaf burning should be calculated separately because the mercury content in different tissues may be very different. Mercury concentrations are much higher in leaf tissue than in the stems. This can be attributed to the fact that nearly all of the mercury in both herbaceous plants and woody plants was absorbed from the atmosphere by stomata on the surface of the foliage. This is the predominant pathway by which mercury accumulates in plants.12 Recent studies found concentrations of ambient mercury in China that are much higher than in other countries.13 As shown in Table 1, EFs for biomass burning in China are greater than the values in other countries. Results from recent investigations of Hg concentrations across 14 forest site in the United States shows that average Hg concentrations in stems and foliage are 15 and 40 ng/g respectively.14 The concentrations measured in China for forest leaves are, on average, 11% higher than the
biomass type
2. METHODOLOGY AND DATA SOURCE
EFs used in this study (ng/g)b
comparison of total mercury concentration (ng/g)a
other countries
mercury into the local terrestrial ecosystem and absorption by vegetation.5 Third, both the results have 1° 1° resolution, which could be too coarse for air quality modeling. A new mercury emission inventory with fine resolution to reflect recent biomass burning emission in China is needed, especially for atmospheric simulation. In this study, the mercury emissions from biomass burning in China (excluding small islands in the South China Sea) during 20002007 were estimated. We adopted the newly available MODIS burned area product (MCD45A1) at a 500 m resolution to calculate the emissions from forest and grassland fires. Because the burned area products from remote sensors with medium resolution often miss the crop burning in fields due to its small size, we used the official statistics data at the provincial level to estimate the mercury emissions from crop residues burning in fields and biofuel combustion in homes. The mercury concentrations for different parts of biomass (leaf and stem) measured in recent studies in China were used as surrogates for emission factors.68 Finally, the inventory is allocated spatially at 1 km resolution by using the Global Land Cover Data set 2000 (GLC-2000), MODIS thermal anomalies/fire products (MOD14A2 and MYD14A2) and rural population density.
19133(l), 2427(s),9 20,6 38190,41 0130,42
ARTICLE
dx.doi.org/10.1021/es202224e |Environ. Sci. Technol. 2011, 45, 9442–9448
Environmental Science & Technology corresponding values in the United States and those for crop stems are 60138% higher. In addition, approximately 37 times more mercury was found in China’s grasslands than that in Africa’s grasslands. The reason for this could be that Chinese mineral resources are rich in mercury, the control measures during coal burning, smelting and mining are poor, and atmospheric deposition is correspondingly high. Finally, the accumulated mercury is re-emitted into the atmosphere during biomass burning process. Forest and Grassland Fires. Mercury resides both in the above-ground biomes and organic soil. Nevertheless, histosols is very sparse in China and its emission is relatively small and could be neglected.15 The amounts of burned biomass from forests and grasslands were computed based on the 500 m MCD45A1 burned area product, fuel load and combustion efficiency.16 The MCD45A1 product was validated for forest burning in China by comparing MCD45A1 results with the fire-affected forest area recorded by the National Forestry Bureau, by month and province from 2001 to 2006.17 Fuel loads for individual provinces were assigned by vegetation types. Impacts of fuel type and fuel moisture were taken into account in calculating combustion efficiency.18 More details can be found in Song et al.’s paper.19 In-field Burning and In-home Combustion of Crop Residues. Crop residues, including residues from rice, wheat, corn, coarse cereals, cotton, legumes, peanut, or rape, are often burned for household energy and used as fertilization in the field. The emissions from this burning were found to be significant.20 Crop fires in the fields in China were often missed by MODIS due to their small size. In this study, the provincial amounts of crop residues burned in fields and in homes as fuel are estimated by multiplying the total crop residues, the percentage of field/ domestic burning of crop residues and the combustion efficiency. The total crop residues is the product of crop productions at the provincial level distributed by the government and the residue/ crop ratio (listed in Supporting Information (SI) Table S1).11 The percentage of crop residues that were burned in fields or in homes as fuel were adopted from a large-scale investigation on the usage of crop residues in different provinces, which provides crop-specific percentages of domestic and field burning (see SI Table S2 and S3).20 Combustion efficiencies (SI Table S1) are specified by crop type. Much more of the crop residues were used as fuel in densely populated rural areas; this is because these populations have less income and poor access to other energy sources. In the developed regions, the crop residues are more likely to be burned in the fields. Fuelwood Combustion. As another important energy source, fuelwood is widely distributed and available in many remote mountainous areas of China. For instance, more than 10 Tg fuelwood is consumed every year in Guangxi and Hunan. In this study, the amount of fuelwood combustion at the provincial level was estimated on the basis of fuelwood consumption and combustion efficiency.10 The combustion efficiency of fuelwood was assumed to be 87%.21 Spatial Allocation. The aim of this study is to produce a mercury emission inventory with fine resolution. Although the amount of crop residues burnt in fields in China could not be reflected accurately in burned area products (MCD45A1) because of their small size, they could be located by MODIS fire counts data. 22 We selected 20022007 MOD14A2 and MYD14A2 products (MODIS Thermal Anomalies/Fire 8-Day 1 km L3 global products from satellites Terra and Aqua, when Terra passes over China at 10:30 while Aqua is at 14:30 local time,
ARTICLE
Table 2. Mercury Emissions (Mg) from Biomass Burning in China during 2000-2007
year
total
field
crop
crop
residues as fuel
forest
grassland
residues
fuelwood
2000
24.2
0.01
0.00
4.3
14.2
5.7
2001
25.5
0.03
0.00
4.3
14.3
6.9
2002
26.9
0.02
0.00
4.2
14.7
8.0
2003
26.2
0.11
0.01
4.0
13.9
8.2
2004
28.4
0.08
0.01
4.4
15.4
8.5
2005
27.7
0.03
0.01
4.5
15.9
7.3
2006
28.4
0.03
0.00
4.6
16.1
7.6
2007 average
28.7 27.0
0.04 0.04
0.00 0.00
4.7 4.4
16.6 15.2
7.3 7.4
0.1%
0.0%
16.2%
percentage
100%
56.3%
27.4%
respectively) to determine the open fire frequencies and spatial distribution on a 1 km grid.23 Only open fires in the land cover classes defined as “Farm” and “Mosaic of cropping” in the GLC2000 land cover data set (also at 1 km resolution) were identified as crops burning in fields. The mercury emission in the i-th zone (Ei) was calculated using the following equation: Ei ¼
FCi Ek FCk
where FCi is the fire count in i-th zone, FCk is the total fire count in province k and Ek is the total estimated mercury emission from crop residues burning in fields in province k. Similarly, the provincial level emissions from in-home combustion of crop residue and fuelwood combustion are allocated by using the rural population density in 1 km zones.24 The forest and grass fires emission data were derived with 500 m spatial resolution. Finally, all the emissions were aggregated at 1 km resolution.
3. RESULTS AND DISCUSSION Comparisons to Other Studies. The total annual mercury emissions from biomass burning in China during 20002007 are listed in Table 2, and range from 24.2 Mg (2000) to 28.7 Mg (2007). The average annual total mercury emission is 27.0 Mg. The amount of burned biomass, combustion efficiency and EFs were the sources that caused the uncertainties in the estimation. Given the larger presumed uncertainties of statistics for untracked energy use, the probability of the value used for burned biomass amount was assumed to have a normal distribution with a coefficient of variation (CV) of 30%.25 The typical uncertainty of the EF is 50%.5,26 We ran 20000 Monte Carlo simulations to estimate the range of fire emissions with a 90% confidence interval. The estimated emission range is 15.139.9 Mg/year. The emissions from crop residues combusted in homes is the biggest contributor to the total mercury emissions (56%), followed by fuelwood combustion (27%) and field burning of crop residues (16%). Emissions from forest fires are relatively small, and grassland fire emission are negligible. Previous studies also found that crop residues and fuelwood were the dominant gaseous pollutants out of all biomass burning in China.27 Streets et al. estimated the total mercury emission from biomass burning in 1999 as 19.2 Mg, somewhat lower than our 9444
dx.doi.org/10.1021/es202224e |Environ. Sci. Technol. 2011, 45, 9442–9448
Environmental Science & Technology result, 27.0 Mg.3 The biggest disparity is in the biofuel emission estimates. We estimated 7.4 Mg of emissions came from fuelwood and 15.2 Mg came from crop residue, for a total of 22.5 Mg of mercury from biofuel sources. Street et al. estimated only 8.3 Mg of mercury was released from biofuel sources in 1999.3 The two biofuel amounts (the sum of fuelwood and crop residues) were comparable: 413 Tg in Street et al.’s paper and 349 Tg (the sum of fuelwood and crop residues) in our research. The fuelwood amount in the year of 2000, 185 Tg, was close to the 177 Tg in 2000 reported by Yan et al.,28 and the biofuel amount from crop residues, 164 Tg, was close to the 106 Tg reported by Zhang et al. in 2004.29 Thus, the high EFs in our study are probably the reason for the disparity. We adopted the recent measurements in China, 40 ng/g for crop stems and 100 ng/g for leaves, while Streets et al used the EF value of only 20 ng/g. The other disparities between the two studies were the emissions from forest fires and grass fires. The forest fire emission in our estimation was 0.04 Mg, much lower than the 2.8 Mg reported in Streets et al.’s research. Streets et al. adopted 113 ng/ g as the emission factor, comparable to ours (Table 1); however, they used government records of the 1950s1990s as the burned areas, whereas we chose to use the 20002007 MODIS burned area products. The recent forest fires in China were captured very well, as confirmed by comparing MCD45A1 results with the monthly, provincial fire-affected forest area recorded by the government from 2001 to 2006.17 Recent study also confirmed that forest fire activities have decreased drastically over the last two decades due to law enforcement.15 For grassland burning, Streets et al. estimated 4.17 Mg mercury emission, while the grass burning emission in our results was small enough to be neglected. That study used a higher EF for grass, 80 ng/g. Another reason for the disparity is that the assumed grass burned areas are larger in their study. They assumed a uniform burned fraction over all of China equal to that in Mongolia (3.0%), while we derived grass burned areas again from the MCD45A1 product. The recent records showed that only 0.6% of total grassland was burned in China during 20002008.30 The mercury emissions both from forest and grassland fires may therefore be overestimated by Streets et al. The field crop burning contributed 3.9 Mg mercury in Streets et al.’s paper, close to the 4.4 Mg in our results. There was a difference in EF values: Streets et al. used 37 ng/g, and we used 35 ng/g for stems and 319 ng/g for leaves. Streets et al. estimated the ratio of crop residues burned in the field with a single value of 17% for every kind of crops over the whole country; however, in our estimates, the provincial-level and crop-specific percentages used were based on surveys from 2000. From our method, an average of 6.6% of crop residues was burned in field. China accounts for approximately 4% of the global mercury emissions from biomass burning, 675 Mg as estimated by Friedli et al.2 In contrast to the larger contributors to mercury emissions, such as the United States with 44 Mg yearly,8 the majority of the emissions from China were from crop residues combustion (72%, including burning in fields and home fuel use), whereas the corresponding percentage in the United States was only 3%, the rest is mostly caused by forest fires. The total atmospheric mercury contributed from biomass burning in China was only 5% of that from anthropogenic activities (536 Mg) mercury, most of which was emitted from coal combustion and nonferrous metals smelting.3 However, as proposed previously, biomass burning could play an important role in mercury exchange between terrestrial ecosystems and the atmosphere.
ARTICLE
Temporal Variations. The total mercury emission from biomass burning ranged from 24 Mg in 2000 to 29 Mg in 2007. The annual variations could be attributed largely to the crop residues combustion from in-field and in-home use. The amount of crop residues used for domestic burning depends on the rural population and crop residues yield. The reduction of straw-based forage and the rising price of fossil fuel in recent years also promoted the usage of crop residues. The amount of crop residues burned in fields dropped from 4.3 Mg in 2000 to 4.0 Mg in 2003, and increased in 20042006. Fuelwood consumption, another important contributor to mercury emissions, increased rapidly during 20002004 and fell slightly in 20052007. This fall could be caused by the government vigorously advocating the use of renewable energy sources and energy-saving measures in rural areas in recent years. In addition, an increasingly strict regulatory system of forest protection played an important role in the control of fuelwood consumption. Forest and grassland fires made only a very small contribution to the total amount of biomass burning in China. The forest and grassland fires correlate with both natural causes (lightning, low precipitation, high temperature, etc.) and human causes (land clearing for agriculture or habitation, etc.). The emissions in 2003 are by far the highest (115 kg), and the minimum emissions occurred in the year 2000 (6 kg). Correspondingly, these two years exhibit the greatest and the lowest amounts of burned areas in the MCD45A1 product. The large forest and grassland fires in 2003 may have been influenced by El Ni~no and an abnormal climate.31 Spatial Distribution. The year 2006 was selected to demonstrate the spatial patterns, which are shown in Figure 1 and SI Figures S1S3. In general, there is a high emission zone located in central China, which includes Sichuan, Hubei, Henan, Shandong, Jiangsu, and Anhui Provinces (see SI Table S4). Higher mercury emission was caused by larger rural population and their reliance on crop residue combustion. These provinces accounts for 42% of the total mercury emissions in 2006. This region is an important agricultural zone in China. The amounts of crop residues in these provinces are high. Moreover, the populations are also high; dense population needs more energy to consume. However, the economic income in such rural areas is often low. The forest cover is often lower in the agricultural areas, with less wood to be used as biofuel, and thus, crop residues become their most important energy source (SI Figure S2). The straw from wheat, corn, rice, and cotton can be burned easily. The percentage of these four crop residues used as domestic biofuel is 59% in Sichuan, 53% in Hubei and 45% in Jiangsu, as examples, which is much higher than the average of 24% for China as a whole.20 Compared with the other provinces, the portion of crop residues burned in the fields in central China was lower. The Southwestern provinces also play a significant role in mercury emissions from biomass burning, especially Hunan, Guangxi, Guizhou and Yunnan Provinces. Total mercury deposition fluxes measured in Guizhou range from 336 to 2340 g/km2/ y, much higher than average value for East Asia (<50 g/km2/y).32 The reason is the high mercury content of raw coal in this province and the relatively large amount of uncontrolled coal combustion. This area contributed 17% of the total mercury emissions from biomass burning. Most of the people there live in mountainous areas. Unlike the agricultural provinces, the crop residues amounts are often much lower. The rural people here depend on wood combustion for energy. Fortunately, the forest coverage is high, approximately 43%, which is 25% higher than 9445
dx.doi.org/10.1021/es202224e |Environ. Sci. Technol. 2011, 45, 9442–9448
Environmental Science & Technology
ARTICLE
Figure 1. Mercury Emissions from Biomass Burning for the year 2006 (1 Km 1 Km).
the average level for the whole country.33 More than 50% of total mercury came from fuelwood combustion, as shown in SI Figure S2. In 2006, 12 Tg of fuelwood and 5 Tg of crop residues were burned as domestic fuel in Guangxi province. Another important contributor is northeast China, including Liaoning, Jilin and Heilongjiang Province. The deposition flux measured in Jilin is approximately 320 g/km2/y, which can be attributed to the presence of nonferrous metal smelters.34 This area contributed 14% of the total mercury emission from biomass burning. As this area is an important producer of winter crops and has dense boreal forest areas, the amounts of emissions from fuelwood and crop consumption were also large. People need more energy for heating over the long winter. Moreover, it should be mentioned that the MCD45A1 product indicates that northern China was the largest contributor to the total burned forest areas (averaging 62% of the forest) during the period of 20002006. The western region of China has the lowest emissions because of low population density and crop yields. However, high cotton and wheat yields in western Xinxiang lead to some higher local mercury emission from crop residue combustion. In summary, China has a unique bioenergy consumption pattern which results in regional differences in mercury emissions. The mercury concentration in vegetation is very site-dependent and species-dependent. Since China is known for high levels of atmospheric mercury and correspondingly high deposition to vegetation, area specific measurement of Hg concentration values for various biomes in China are very desirable and necessary in future research. We also note that the mercury emission estimation from soil during open fires is needed.
’ ASSOCIATED CONTENT
bS
Supporting Information. Tables for crop-specific dry weight ratios of production to residue, combustion efficiency and percentages of crop residues burning in field and domestic burning as fuel in China used in the emission calculations, as mentioned in Section 2. Summary table for the average annual mercury emission from biomass burning at provincial-level in China, as described in Section 3. This information is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: + 86-10-62755343; e-mail: [email protected].
’ ACKNOWLEDGMENT This study was funded by National Natural Science Foundation of China (No. 40975088, 70821140353, and 70773001) and the Public Welfare Projects for Environmental Protection (200809122). ’ REFERENCES (1) Sigler, J. M.; Lee, X.; Munger, W. Emission and long-range transport of gaseous mercury from a large-scale Canadian boreal forest fire. Environ. Sci. Technol. 2003, 37 (19), 4343–4347. (2) Friedli, H. R.; Arellano, A. F.; Cinnirella, S.; Pirrone, N. Initial estimates of mercury emissions to the atmosphere from global biomass burning. Environ. Sci. Technol. 2009, 43 (10), 3507–3513. 9446
dx.doi.org/10.1021/es202224e |Environ. Sci. Technol. 2011, 45, 9442–9448
Environmental Science & Technology (3) Streets, D. G.; Hao, J. M.; Wu, Y.; Jiang, J. K.; Chan, M.; Tian, H. Z.; Feng, X. B. Anthropogenic mercury emissions in China. Atmos. Environ. 2005, 39 (40), 7789–7806. (4) Streets, D. G.; Yarber, K. F.; Woo, J. H.; Carmichael, G. R. Biomass burning in Asia: Annual and seasonal estimates and atmospheric emissions. Global Biogeochem. Cycles 2003, 17 (4), 1099. (5) Feng, X. B.; Li, G. H.; Qiu, G. L. A preliminary study on mercury contamination to the environment from artisanal zinc smelting using indigenous methods in Hezhang County, Guizhou, China: Part 2. Mercury contaminations to soil and crop. Sci. Total Environ. 2006, 368 (1), 47–55. (6) Friedli, H. R.; Radke, L. F.; Lu, J. Y.; Banic, C. M.; Leaitch, W. R.; MacPherson, J. I. Mercury emissions from burning of biomass from temperate North American forests: laboratory and airborne measurements. Atmos. Environ. 2003, 37 (2), 253–267. (7) Michelazzo, P. A. M.; Fostier, A. H.; Magarelli, G.; Santos, J. C.; de Carvalho, J. A., Mercury emissions from forest burning in southern Amazon. Geophys. Res. Lett. 2010, 37. (8) Wiedinmyer, C.; Friedli, H. Mercury emission estimates from fires: An initial inventory for the United States. Environ. Sci. Technol. 2007, 41 (23), 8092–8098. (9) Obrist, D.; Moosmuller, H.; Schurmann, R.; Chen, L. W. A.; Kreidenweis, S. M. Particulate-phase and gaseous elemental mercury emissions during biomass combustion: Controlling factors and correlation with particulate matter emissions. Environ. Sci. Technol. 2008, 42 (3), 721–727. (10) Ed.ial Committee of the Yearbook of Rural Energy Yearbook of China, Rural Energy Yearbook of China (20002007); China Agriculture Press: Beijing, China, 20012008. (11) National Bureau of Statistics of China. China Rural Statistical Yearbook (20002007); China Statistics Press: Beijing, China, 20012008. (12) Ericksen, J. A.; Gustin, M. S.; Schorran, D. E.; Johnson, D. W.; Lindberg, S. E.; Coleman, J. S. Accumulation of atmospheric mercury in forest foliage. Atmos. Environ. 2003, 37 (12), 1613–1622. (13) Fang, G. C.; Wu, Y. S.; Chang, T. H. Comparison of atmospheric mercury (Hg) among Korea, Japan, China and Taiwan during 20002008. J. Hazard Mater. 2009, 162 (23), 607–615. (14) Obrist, D.; Johnson, D. W.; Lindberg, S. E.; Luo, Y.; Hararuk, O.; Bracho, R.; Battles, J. J.; Dail, D. B.; Edmonds, R. L.; Monson, R. K.; Ollinger, S. V.; Pallardy, S. G.; Pregitzer, K. S.; Todd, D. E. Mercury distribution across 14 U.S. Forests. Part I: Spatial patterns of concentrations in biomass, litter, and soils. Environ. Sci. Technol. 2011, 45 (9), 3974–3981. (15) Song, Y., A new emission inventory for nonagricultural open res in Asia from 2000 to 2009. Environ. Res. Lett. 2010. (16) Seiler, W.; Crutzen, P. J. Estimates of gross and net fluxes of carbon between the biosphere and the atmosphere from biomass burning. Clim. Change 1980, 2 (3), 207–247. (17) Chang, D.; Song, Y., Comparison of L3JRC and MODIS global burned area products from 2000 to 2007. J. Geophys. Res., [Atmos.] 2009, 114(D16). (18) Ito, A.; Akimoto, H. Seasonal and interannual variations in CO and BC emissions from open biomass burning in Southern Africa during 19982005. Global Biogeochem. Cycles 2007, 21 (2), 2011. (19) Song, Y.; Liu, B.; Miao, W. J.; Chang, D.; Zhang, Y. H. Spatiotemporal variation in nonagricultural open fire emissions in China from 2000 to 2007. Global Biogeochem. Cycles 2009, 23, 2008. (20) Gao, X.; Ma, W. Analysis on the current status of utilization of crop straw in china. J. Huazhong Agric. Univ. 2002, 21 (3), 242–247. (21) Feng, T. T.; Cheng, S. K.; Min, Q. W.; Li, W. Productive use of bioenergy for rural household in ecological fragile area, Panam County, Tibet in China: The case of the residential biogas model. Renewable Sustainable Energy Rev. 2009, 13 (8), 2070–2078. (22) van der Werf, G. R.; Randerson, J. T.; Giglio, L.; Collatz, G. J.; Kasibhatla, P. S.; Arellano, A. F. Interannual variability in global biomass burning emissions from 1997 to 2004. Atmos. Chem. Phys. 2006, 6, 3423–3441.
ARTICLE
(23) Giglio, L. MODIS Collection 4 Active Fire Product User’s Guide Version 2.2; Science Systems and Applications, Inc.: Lanham, MD2005. (24) China’s population distribution database. http://www.resdc.cn/ ua/download/POP-GDP/cnpop2003.htm (accessed March 1, 2011). (25) Zhao, Y.; Nielsen, C. P.; Lei, Y.; McElroy, M. B.; Hao, J. Quantifying the uncertainties of a bottom-up emission inventory of anthropogenic atmospheric pollutants in China. Atmos. Chem. Phys. 2011, 11 (5), 2295–2308. (26) Zheng, D.; Wang, Q.; Li, Z. Comparative study of mercury pollution in two different typical cities of northern China: coal-combustion and industrial cities. Earth Environ. 2007, 35 (3), 273–278. (27) Yevich, R.; Logan, J. A. An assessment of biofuel use and burning of agricultural waste in the developing world. Global Biogeochem. Cycles 2003, 17 (4), 1095. (28) Yan, X. Y.; Ohara, T.; Akimoto, H. Bottom-up estimate of biomass burning in mainland China. Atmos. Environ. 2006, 40 (27), 5262–5273. (29) Zhang, H. F.; Ye, X. N.; Cheng, T. T.; Chen, J. M.; Yang, X.; Wang, L.; Zhang, R. Y. A laboratory study of agricultural crop residue combustion in China: Emission factors and emission inventory. Atmos. Environ. 2008, 42 (36), 8432–8441. (30) Editorial Committee of the Yearbook of China. China Animal Industry Yearbook (20002007); China Agriculture Press: Beijing, China, 20012008. (31) van der Werf, G. R.; Randerson, J. T.; Collatz, G. J.; Giglio, L.; Kasibhatla, P. S.; Arellano, A. F.; Olsen, S. C.; Kasischke, E. S. Continental-scale partitioning of fire emissions during the 1997 to 2001 El Nino/La Nina period. Science 2004, 303 (5654), 73–76. (32) Tan, H.; He, J. L.; Liang, L.; Lazoff, S.; Sommer, J.; Xiao, Z. F.; Lindqvist, O. Atmospheric mercury deposition in Guizhou, China. Sci. Total Environ. 2000, 259 (13), 223–230. (33) China Forestry Department. China Forestry Statistical Yearbook (2008); China Statistics Press: Beijing, China, 2009. (34) Fang, F. M.; Wang, Q. C.; Li, J. F. Urban environmental mercury in Changchun, a metropolitan city in Northeastern China: Source, cycle, and fate. Sci. Total Environ. 2004, 330 (13), 159–170. (35) Wang, Q. the Genus Pinus; Jiangsu Science and Technology Publishing House: Nanjing, 1994. (36) Zhang, L.; Zhou, Z. Mercury pollution characteristics of plants and soils in different functional areas in Qingdao city. Ecol. Environ. 2008, 17 (2), 802–806. (37) Li, Z.; Wang, Q. Characters of main plants polluted by mercury in Changchun City. J. Grad. Sch. Chin. Acad. Sci. 2003, 20 (4), 477–481. (38) Liu, R.; Wang, Q.; Li, X.; Ma, Z.; Fang, F. Mercury in the peat bog ecosystem in Xiaoxing’an Mountain in China. Chin. J. Environ. Sci. 2002, 23 (4), 102–106. (39) Fu, X. W.; Feng, X. B.; Zhu, W. Z.; Rothenberg, S.; Yao, H.; Zhang, H. Elevated atmospheric deposition and dynamics of mercury in a remote upland forest of southwestern China. Environ. Pollut. 2010, 158 (6), 2324–2333. (40) Zhang, z.; Wang, Q.; Zheng, D. Mercury biogeoehemistry in the soil-plant-insect system in Huludao City. Acta Sci. Circumstantiae 2008, 28 (10), 2118–2124. (41) Friedli, H. R.; Radke, L. F.; Prescott, R.; Hobbs, P. V.; Sinha, P. Mercury emissions from the August 2001 wildfires in Washington State and an agricultural waste fire in Oregon and atmospheric mercury budget estimates. Global Biogeochem. Cycles 2003, 17 (2), 1039. (42) Harden, J. W.; Neff, J. C.; Sandberg, D. V.; Turetsky, M. R.; Ottmar, R.; Gleixner, G.; Fries, T. L.; Manies, K. L., Chemistry of burning the forest floor during the FROSTFIRE experimental burn, interior Alaska, 1999. Global Biogeochem. Cycles 2004, 18, (3). (43) Engle, M. A.; Gustin, M. S.; Johnson, D. W.; Murphy, J. F.; Miller, W. W.; Walker, R. F.; Wright, J.; Markee, M. Mercury distribution in two Sierran forest and one desert sagebrush steppe ecosystems and the effects of fire. Sci. Total Environ. 2006, 367 (1), 222–233. (44) Ma, Y.; Yang, B.; Yang, W. Comparison of different feed corn straw yield and nutritional quality. Chin. Livest. Poul. Breed. 2007, 3, 88–89. 9447
dx.doi.org/10.1021/es202224e |Environ. Sci. Technol. 2011, 45, 9442–9448
Environmental Science & Technology
ARTICLE
(45) Meng, B.; Feng, X. B.; Qiu, G. L.; Cai, Y.; Wang, D. Y.; Li, P.; Shang, L. H.; Sommar, J. Distribution patterns of inorganic mercury and methylmercury in tissues of rice (Oryza sativa L.) plants and possible bioaccumulation pathways. J. Agric. Food Chem. 2010, 58 (8), 4951–4958. (46) Liu, W. X.; Shen, L. F.; Liu, J. W.; Wang, Y. W.; Li, S. R. Uptake of toxic heavy metals by rice (Oryza sativa L.) cultivated in the agricultural soil near zhengzhou city, People’S Republic of China. Bull. Environ. Contam. Toxicol. 2007, 79 (2), 209–213. (47) Rothenberg, S. E.; Du, X.; Zhu, Y. G.; Jay, J. A. The impact of sewage irrigation on the uptake of mercury in corn plants (Zea mays) from suburban Beijing. Environ. Pollut. 2007, 149 (2), 246–251.
9448
dx.doi.org/10.1021/es202224e |Environ. Sci. Technol. 2011, 45, 9442–9448
ARTICLE pubs.acs.org/est
Quantitative Measurement of Direct Nitrous Oxide Emissions from Microalgae Cultivation Kelly D. Fagerstone,† Jason C. Quinn,† Thomas H. Bradley,† Susan K. De Long,‡ and Anthony J. Marchese*,† † ‡
Department of Mechanical Engineering, Colorado State University, Fort Collins, Colorado, 80523-1374 United States Department of Civil and Environmental Engineering, Colorado State University, Fort Collins, Colorado, 80523-1372 United States
bS Supporting Information ABSTRACT: Although numerous lifecycle assessments (LCA) of microalgae-based biofuels have suggested net reductions of greenhouse gas emissions, limited experimental data exist on direct emissions from microalgae cultivation systems. For example, nitrous oxide (N2O) is a potent greenhouse gas that has been detected from microalgae cultivation. However, little quantitative experimental data exist on direct N2O emissions from microalgae cultivation, which has inhibited LCA performed to date. In this study, microalgae species Nannochloropsis salina was cultivated with diurnal light dark cycling using a nitrate nitrogen source. Gaseous N2O emissions were quantitatively measured using Fourier transform infrared spectrometry. Under a nitrogen headspace (photobioreactor simulation), the reactors exhibited elevated N2O emissions during dark periods, and reduced N2O emissions during light periods. Under air headspace conditions (open pond simulation), N2O emissions were negligible during both light and dark periods. Results show that N2O production was induced by anoxic conditions when nitrate was present, suggesting that N2O was produced by denitrifying bacteria within the culture. The presence of denitrifying bacteria was verified through PCR-based detection of norB genes and antibiotic treatments, the latter of which substantially reduced N2O emissions. Application of these results to LCA and strategies for growth management to reduce N2O emissions are discussed.
1. INTRODUCTION Microalgae are currently under consideration as a nextgeneration feedstock for the production of biofuels. Compared to first generation biofuel feedstocks, microalgae are characterized by higher solar energy yield, year-round cultivation, the ability to grow in lower quality or brackish water, and the use of less and lower-quality land.1 6 Microalgae feedstock cultivation can be integrated with waste streams such as wastewater treatment facilities and CO2 generating processes such as combustion power plants, to provide critical nutrients that enable accelerated growth, while improving the environmental impact of the process.7 10 These advantages, coupled with the potential for microalgae to meet mandated alternative fuel goals, have led to a renewed interest in microalgae biofuels. However, the lifecycle effects of replacing traditional transportation fuel sources with microalgae biofuels must be critically evaluated. Numerous life cycle assessments (LCA) have been performed to evaluate the sustainability of microalgae biofuels.9,11 17 However, due to the current immaturity of various components of the microalgal value chain, all microalgae LCA analyses contain simplifying assumptions that have yet to be verified experimentally. For example, direct nitrous oxide (N2O) emissions from microalgae cultivation could have a substantial impact on predicted green house gas (GHG) emissions due to the high global r 2011 American Chemical Society
warming potential of N2O (298 CO2eq).18 The microalgae biofuel LCAs performed to date have typically ignored or modeled N2O emissions using a set of broad assumptions. LCA direct emissions assumptions have been based on data from the National Renewable Energy Lab Aquatic Species Program,19 other aquaculture and oceanographic literature,20 22 or the Intergovernmental Panel on Climate Change (IPCC) guidelines for terrestrial crop GHG emissions,23 none of which are entirely applicable to the direct N2O emissions from microalgae cultivation for biofuels. Some experimental studies have observed N2O emissions from microalgae cultivation, but none have provided quantification in such a way that results can be translated to LCA-relevant metrics. Weathers20 detected N2O emissions from axenic cultures of one class of green algae and suggested that N2O emissions from green algae when nitrite was present could be an intermediate byproduct of the denitrification pathway. However, no work was done to verify the production mechanism or quantify N2O emissions. More recently, Florez-Leiva et al.21 Received: July 25, 2011 Accepted: September 22, 2011 Revised: September 21, 2011 Published: September 22, 2011 9449
dx.doi.org/10.1021/es202573f | Environ. Sci. Technol. 2011, 45, 9449–9456
Environmental Science & Technology detected N2O emissions from microalgae cultivation in a fullscale open raceway pond (ORP) using urea as a nitrogen source and hypothesized that oxic NH4 oxidation was the most likely N2O production pathway. Again, no attempt to measure the total quantity of N2O emissions over the microalgae cultivation cycle was documented. Neither of these experimental studies present data quantifying the direct N2O emissions from microalgae cultivation in a way that is directly applicable for use in LCA. Although the cultivation of microalgae cultures has been shown to generate direct N2O emissions, quantitative data did not exist to enable previous microalgae LCA to include the netGHG effects of N2O emissions during cultivation.11 17,20,21 For example, Batan et al.9 considered direct N2O emissions from cultivation by postulating that N2O emissions were generated by denitrification. Their LCA included the energy consumption required to maintain oxygen levels during periods with nitrate present, thereby minimizing denitrification and resultant N2O emissions, because anoxic denitrification has been shown to be a major cause of nitrogen loss in terrestrial crop N2O emissions.24 However, no data existed to validate their claim that oxic conditions suppress the direct emission of N2O in microalgae cultivation. A better understanding of microalgae N2O emissions processes and quantities would not only lead to more accurate microalgae LCA, but would also be beneficial for development of microalgae cultivation practices that can minimize direct N2O emissions. The objectives of this study were (i) to quantify N2O production under conditions that are representative of photobioreactors (PBR) and ORP with diurnal light dark cycling to inform N2O emissions accounting in LCA, (ii) to develop an improved understanding of N2O production mechanisms during microalgae cultivation, and (iii) to ultimately provide GHG-reducing cultivation management strategies. To meet these objectives, N2O production was quantitatively measured under conditions of varying microalgal culture oxygen concentration, light, and nitrate concentration to identify conditions that induce and repress N2O production. The role of bacteria on N2O emissions was investigated because full-scale microalgae cultures will not be maintained axenic. Thus, the N2O production mechanism was investigated using molecular biological techniques (i.e., PCR assays) targeting bacterial genes involved in N2O production. The discussion focuses on the implications of these experimental results for LCA of microalgae-based biofuels and the management of growth systems to reduce direct N2O emissions.
2. METHODS AND MATERIALS The following sections detail the materials and methods used for microalgae growth and data collection, including the cultivation techniques used to simulate PBR and ORP conditions, quantitative N2O measurements using Fourier transform infrared (FTIR) spectroscopy, and detection of genes involved in N2O production. 2.1. Cultivation Techniques. In the present study, a microalgae cultivation system was developed and a test protocol was established to simulate, at bench-scale, the growth conditions in full-scale closed PBR and ORP systems. The bench-scale growth system consisted of three identical 1-L Erlenmeyer flasks attached to a PolyScience shaking thermal bath. The system was operated at 140 rpm with an eccentricity setting of 9 with the thermal basin temperature maintained at 23 °C. Illumination of the system was supplied using a Sun Systems Yield Master II Classic (Sunlight Supply, Vancouver, WA) with a 1000-W
ARTICLE
daylight metal halide grow lamp selected for its accurate representation of solar photosynthetic active radiation (PAR). Lighting was operated on a 16 h light, 8 h dark period. The light intensity was measured to be 120 μmole m 2 s 1 at the top of the flasks and 90 μmole m 2 s 1 at the microalgae level using a Heinz Walz US-SQS/L spherical PAR sensor (Heinz Walz, Effeltrich, Germany) connected to a LI-COR L1-250A light meter (LI-COR, Lincoln, NE). As detailed below, sampling was performed in 8-h increments with microalgae growth, nitrate concentration, culture pH, dissolved oxygen (DO), and N 2 O emissions being measured over the course of replicate 4-day batches. Industrial-scale PBR and ORP growth conditions were simulated at laboratory scale by cultivating microalgae under a nitrogenfilled headspace or an air-filled headspace, respectively. The inert (nitrogen gas) headspace was used to simulate a closed PBR as the majority of full-scale PBR growth systems have little or no gaseous headspace. To simulate the growth conditions of a wellmixed ORP, the headspace of the reactors was filled with air. It should also be noted that the air-filled headspace is representative of a closed PBR sparged at night with air or a CO2-air mixture to maintain DO concentrations. 2.1.1. Organism, Culture Media, and Inoculation. Microalgae species Nannochloropsis salina (1776) was obtained by Solix BioSystems from the Provasoli-Guillard National Center for Culture of Marine Phytoplankton. A nutrient-rich growth media was produced by modifying f/2 growth media to a salinity of 20 g L 1 and adding 10 mM NO3 L 1, 7.9 mM PO4 L 1, and 1 mL L 1 Guillard trace metals.25 The nutrient-rich growth media was filtered using a 0.2-μm absolute filter prior to inoculation. Cultures were inoculated at 1 g L 1 and typically harvested at 2 g L 1 over a 4-day batch. Inoculum was concentrated by centrifuging the harvested culture at 1000g for 20 min, the supernatant was discarded, and the inoculum was resuspended with fresh nutrient media. 2.1.2. Monitoring and Control of pH. The culture pH was monitored twice daily with a YSI pH 100 temperature and pH probe. For analysis purposes, the gaseous species emitted from the microalgae culture were allowed to accumulate in the headspace above the microalgae culture on an 8-h time scale, which presented challenges for maintaining pH and supplying carbon. Upon inoculation, 25 mM sodium bicarbonate was supplied and an additional 5 mM was added daily to provide carbon to the microalgae, which has been shown to be an effective alternate to sparging CO2.26,27 The utilization of carbon from the sodium bicarbonate leads to a rise in culture pH due to the accumulation of OH in the media.26 To maintain a culture pH of 7.5 ( 0.5, the culture media was buffered with the addition of 25 mM HEPES free acid (ShineGene Molecular Biotech, Inc., Shanghai, China); HEPES is an organic zwitterionic buffer commonly used in microalgae growth studies. When required, additional pH control was conducted via the addition of hydrochloric acid to lower the pH to within HEPES buffering range. In this study, growth rates utilizing this method of pH regulation were similar to growth rates determined with conventional CO2 pH control (data not shown). 2.1.3. Growth, Nitrate, and DO Monitoring. Growth and nitrate monitoring in the culture medium were performed daily, at the beginning of each light cycle. Optical density (OD) measurements at 750 nm were performed using a Hach DR5000 spectrophotometer and dry mass was calculated based on a predetermined dry mass to OD correlation coefficient. A nitrate assay was performed using a Hach DR 5000 spectrophotometer. DO concentrations were 9450
dx.doi.org/10.1021/es202573f |Environ. Sci. Technol. 2011, 45, 9449–9456
Environmental Science & Technology
ARTICLE
Figure 1. Average N2O emissions, culture density, and nitrate (from nitrate fertilizer) in growth media for three replicates of the air-filled headspace. The width of the N2O bar represents the headspace gas accumulation period. Error bars represent the standard deviation (n = 9).
measured every 15 min over a 24-h growth cycle using a Hach HQ40d meter with luminescent dissolved oxygen probe. More details on sample preparation and measurement techniques are presented in the Supporting Information (SI). 2.1.4. Antibiotic Growth Conditions. To evaluate the role of denitrifying bacteria in N2O production, microalgae were cultured with and without antibiotics under a nitrogen headspace using the methods described above. Two flasks were treated with antibiotics and a third was used as a control. As adapted from Andersen,28 100 mg L 1 of penicillin g potassium salt and 25 mg L 1 streptomycin sulfate salt were added to the f/2 media at inoculation followed by a booster of 25 mg L 1 penicillin and 5 mg L 1 streptomycin twice daily for 4 days. 2.2. N2O Monitoring. FTIR spectrometry has been shown to be a fast and precise N2O measurement technique,29 and is widely used in a variety of fields including low concentration atmospheric science measurements,30 32 high concentration combustion studies,33 35 and other various applications.36,37 Therefore, an FTIR was used to quantitatively measure the accumulated mole fraction of N2O in the gaseous headspace above the microalgae cultures. Sampling was performed at 8-h increments to concentrate headspace N2O levels to detectible levels. Headspace gases were sampled using a 140-mL polypropylene syringe (Kendall Monoject, 140 cc LL, Mansfield, MA) with a one-way stopcock. The sample gas was scrubbed with Drierite (10 20 mesh, with Indicator, Acros Organics, Geel, Belgium) and Ascarite II (CO2 absorbent, 20 30 mesh, capacity: 40 50%, Acros Organics, Geel, Belgium) to remove water vapor and CO2, respectively.38,39 After sampling, the flask headspace was purged for 5 min with either air or nitrogen gas depending on the desired simulated growth system configuration (PBR vs ORP). For both configurations, the scrubbing tubes were purged with nitrogen gas. Headspace gas samples were analyzed with a Thermo/Nicolet Magna-IR 560 ESP FTIR spectrometer equipped with a 2-m gas cell and a liquid nitrogen cooled MCT-A detector with ZnSe windows. The gas samples were analyzed using Nicolet OMNIC ESP software. All gas samples were processed within 24 h to ensure sample stability.39 41 To prepare for each analysis, the FTIR cell was purged with nitrogen gas for 10 min and evacuated to a consistent pressure with a vacuum pump. A constant volume of headspace exhaust gas was then injected and analyzed over 128 cycles using the N2O spectra between 2170 cm 1 and 2250 cm 1.42
Theoretical calculations based on the IPCC standard for direct N 2 O emissions from terrestrial crops (1% reduction of NO 3 N to N 2 O N), 23 and a concentration of 105 mg NO3 N L 1, suggested the accumulation of N2O at 36 ppmv (N2O N = 22.9 ppmv) for the experimental configuration and test protocol described above. Therefore, a calibration curve (n = 10) for N2O was developed by diluting a 60 ppmv calibration gas with air. The resultant calibration curve was linear from 0 to 60 ppmv (R2 value of 0.9992) with the FTIR manufacturer certifying a calibration range of 5 1000 ppmv. For greater precision at lower N2O levels, a second linear curve (n = 6) was fit from 0 to 12 ppmv (R2 value of 0.9988) which was used for the air-filled headspace (ORP conditions). Atmospheric N2O levels were subtracted out of both calibration curves and all N2O experimental results presented as N2O concentrations above atmospheric levels. For N2O emissions greater than 60 ppmv, the N2O mole fraction was estimated by the linear extrapolation of the 0 60 ppmv calibration curve. N2O has been shown to maintain a linear relationship over a wide range of concentrations,29,43 and the error associated with extrapolation is expected to be minor (explicit details on the extrapolation are presented in the SI). 2.3. DNA Extractions and PCR Conditions. Prokaryotic DNA was extracted from the experimental cultures, and the norB gene (encoding for nitric oxide reductase, which catalyzes the reduction of NO to N2O)44 48 was analyzed to further elucidate the role of denitrifying bacteria in N2O emissions. Microalgae were removed from the growth media by filtration with an absolute 1.2-μm filter. The filtrate was then centrifuged at 12 000g for 15 min to pellet bacteria. The pellet was resuspended in phosphate buffered saline, and DNA was extracted using an UltraClean Microbial DNA Isolation Kit (MoBio Inc., Carlsbad, CA) following the manufacturer’s instructions. DNA was quantified using the Quanti-iT dsDNA Assay Kit (Broad Range Q33130, Molecular Probes, Invitrogen, Carlsbad, CA). PCR amplification was performed using three specific primer sets to target the norB genes.49,50 PCR assay conditions were adapted from Braker et al.49 and Lee et al.50 PCR reactions were performed in 50-μL reactions and consisted of 1 PCR buffer (New England Biolabs, Ipswich, MA), 40 pmol of each primer, 200 μM deoxynucleoside triphosphate, 1.5 mM MgSO4, 2 units 9451
dx.doi.org/10.1021/es202573f |Environ. Sci. Technol. 2011, 45, 9449–9456
Environmental Science & Technology
ARTICLE
Figure 2. Average N2O emissions, culture density, and nitrate (from nitrate fertilizer) in growth media for three replicates of the nitrogen-filled headspace. The width of the N2O bar represents the headspace gas accumulation period. Error bars represent the standard deviation (n = 9).
Taq polymerase (New England Biolabs, Ipswich, MA), 400 ng of bovine serum albumin (BSA), and 70 ng of template DNA. The following amplification program was used: initial denaturation for 5 min at 95 °C, 35 cycles of denaturation for 30 s at 95 °C, primer annealing for 45 s at 54 °C, primer extension for 45 s at 68 °C, with a final extension of 7 min at 68 °C. PCR products (10 μL) were analyzed using 1% (wt/vol) agarose gel in 1 Tris-acetate-EDTA (TAE) buffer and visualized with UV excitation.51 Additional details are presented in the SI.
3. RESULTS AND DISCUSSION The bench-scale growth system and test protocols described above were used to simulate the large-scale cultivation of microalgae in PBR and ORP architectures for the purpose of characterizing direct N2O emissions from these systems. Results from three replicate batches operated in biological triplicate in each architecture are presented. PCR assays were used to detect the denitrifying norB gene in bacteria within the microalgal media. The discussion section focuses on the implications for these results on microalgae biofuels LCA. 3.1. Direct N2O emissions. The following sections detail the environmental conditions and N2O emissions for the two growth architectures represented: a well-mixed ORP (air-filled headspace) and a closed PBR (nitrogen-filled headspace). All N2O results are presented as concentrations above atmospheric levels. 3.1.1. Air-Filled Headspace: Growth, Nitrate, and N2O Results. In this set of experiments, the microalgae cultures were grown with a closed headspace and purged with air every 8 h to represent a well-mixed ORP. The results indicate an average growth rate of 0.24 g L 1 day 1 with the nitrate decreasing from 106 mg NO3 N L 1 to complete uptake by the microalgae within 3 days, Figure 1. Upon depletion of the nitrate, the N2O emissions decrease to zero (0.0 ppmv ( 0.0 ppmv, n = 9). Analysis of the data shows that the cumulative N2O mass emissions over the light periods for the days with nitrate totaled 6.51 10 6 kg N2O N (kg N input) 1. Over the same time frame with nitrate, the total cumulative N2O mass emissions for the dark periods was 1.53 10 5 kg N2O N (kg N input) 1. A two-tailed distribution, two-sample unequal variance Student’s t test indicated that the N2O emissions for the light periods were
significantly lower than the emissions for the dark periods when nitrate was present (P value of 0.0015). The cumulative N2O emissions for light periods with nitrate and dark periods with nitrate also show a statistically significant difference when compared to the periods without nitrate (P values of 0.0001 and 0.0000, respectively). There was no significant difference in total N2O production between the light and dark periods when no nitrate was present (P value of 0.306). The total mass of direct N2O produced from N applied for the air-filled headspace was calculated to be 2.35 10 5 kg N2O N (kg N input) 1 ( 1.97 10 5 kg N2O N (kg N input) 1 (or 0.002% ( 0.002%, n = 9). By comparison, this emission factor is substantially less than the IPCC standard for terrestrial crop N2O emissions, which utilizes an emission factor for N2O emissions from the nitrogen applied of 0.01 kg N2O N (kg N input) 1 (or 1%).23 3.1.2. Nitrogen-Filled Headspace: Growth, Nitrate, and N2O Results. For the nitrogen-filled headspace (simulated PBR) experiments, the average growth rate was nearly identical to that of the air-filled headspace at 0.23 g L 1, and the nitrate was consumed in just over 3 days, Figure 2. The N2O emissions for the nitrogen-filled headspace exhibited trends similar to the emissions of the air-filled headspace but were several orders of magnitude higher. The trend of increased N2O production over dark periods was also more pronounced in the nitrogen-filled headspace in comparison to the air-filled headspace. Cumulative N2O emissions for the dark periods with nitrate present totaled 3.12 10 3 kg N2O N (kg N input) 1, which is substantially higher than the 1.53 10 5 kg N2O N (kg N input) 1 measured for the air-filled headspace under the same conditions. The cumulative N2O emissions during the dark periods with nitrate were also much higher than those measured for the light periods with nitrate present, which totaled 4.38 10 4 kg N2O N (kg N input) 1. Once the nitrate was depleted, the N2O concentrations dropped below the detection limit (0.3 ppmv). For the nitrogen-filled headspace, the total mass of N2O produced from nitrate applied as fertilizer was calculated to be 3.90 10 3 kg N2O N (kg N input) 1 ( 2.82 10 3 kg N2O N (kg N input) 1 (or 0.390% ( 0.282%, n = 9), which is roughly half that of the 0.01 kg N2O N (kg N input) 1 (or 1%) IPCC standard used for terrestrial crops.23 9452
dx.doi.org/10.1021/es202573f |Environ. Sci. Technol. 2011, 45, 9449–9456
Environmental Science & Technology
ARTICLE
Figure 3. DO concentrations over 24 h for both air and nitrogen headspace conditions. The upper limit for anoxic denitrification (0.2 ppmv) is shown as the thin dashed line.52
A two-tailed distribution, two-sample unequal variance Student’s t test shows that the cumulative N2O emissions for the light periods are significantly less than for the dark periods when nitrate is present in the growth media (P value of 0.0006). There is also a statistically significant difference in N2O production between the light periods with nitrate and the light periods without nitrate (P value of 0.0016), as well as for the dark periods with nitrate and the dark periods without nitrate (P value of 0.0002). There was no significant difference in N2O production between the light and dark periods with no nitrate present (P value of 0.387). The laboratory simulated well-mixed ORP (air-filled headspace) produced significantly less N2O than the simulated closed PBR (nitrogen-filled headspace) for both the light and dark periods when nitrate was present in the growth media, resulting in P values of 0.0023 and 0.0002, respectively. The results also show that once the nitrate is consumed, N2O emissions drop below detection limits, indicating that N2O production is the greatest for PBRs while nitrate is present. The experimentation was designed to represent the classic operation of a closed PBR and a well-mixed ORP. However, it is acknowledged that not all cultivation systems operate classically. 3.1.3. Large-Scale N2O Emissions. The LCA done by Batan et al.9 indicates that an increase in GHG emissions of 0.392 g CO2-eq MJ 1 would have a 1% impact on the overall GHG emissions. This 1% impact was chosen to correspond to the minimum GHG contribution that would have a significant impact on the overall LCA. The N2O mole fraction per flask per 8-h sample period that corresponds to this 1% impact level was calculated to be 1.34 ppmv in the headspace of the experimental system presented herein. Repeating these calculations with the N2O mole fractions measured in this study over a 5-day growth period, the air-filled headspace (simulated ORP conditions) results suggest a modest net GHG emissions increase of 0.149% due to N2O production. However, for the nitrogen-filled headspace (simulated PBR conditions), over the same 5-day growth period, direct N2O emissions from cultivation would dramatically increase the net GHG emissions by 16.6%. If the maximum N2O concentrations observed in this study (372 ppmv) were applied
to these calculations, the resulting life-cycle GHG emissions would increase by 74.5%, rendering microalgae ineffectual as a feedstock for advanced biofuels based on DOE environmental goals. 3.2. Verification of N2O Production Route: Denitrifying Bacteria. 3.2.1. Dissolved Oxygen Concentrations. The experimental results indicate that a low culture oxygen level may lead to high N2O emissions. It was therefore hypothesized that the N2O emissions observed in these studies are a byproduct of the anoxic bacterial denitrification pathway. To test this hypothesis, the DO contents for both the air-filled and nitrogen-filled headspaces were monitored over 24 h; Figure 3 shows that the air-filled headspace maintained an oxygen-rich environment for the entirety of the experiment, with the lowest oxygen concentration of 4.26 mg L 1 occurring at the end of the dark period, while the nitrogen-filled headspace experienced partially anoxic conditions. During the light periods of the nitrogen headspace study, the microalgae cultures maintained oxic conditions. This is explained by the fact that during the light periods, the microalgae is photosynthetically active and producing oxygen, but over the dark periods the microalgae are respiring and consuming oxygen.52,53 It can be seen that when the lights are turned off, the DO concentration for the nitrogen-filled headspace decreases rapidly to below the detection limit (0.1 mg L 1) and the culture remains anoxic for the remainder of the dark period. Previous studies have reported that maintaining an oxygen level of greater than 0.2 ppmv will inhibit the reduction of nitrate by some species of bacteria, indicating that bacterial denitrification is possible over the dark periods of the nitrogen-filled headspace.52 This result can be seen in Figure 3 with the limit for denitrification (shown by the dotted line) clearly above the DO of the nitrogenfilled headspace, indicating that the microalgae cultures are within the range of possible denitrification for the majority of the dark period. The DO concentration measurements for the nitrogen-filled headspace show that the culture becomes anoxic at the same time that the N2O emissions peak, supporting the hypothesis that the N2O production is from denitrifying bacteria present in the culture media. However, N2O emissions were still observed under oxic conditions. It is possible that N2O detected during 9453
dx.doi.org/10.1021/es202573f |Environ. Sci. Technol. 2011, 45, 9449–9456
Environmental Science & Technology oxic periods may have been a carryover effect of dissolved N2O generated during the anoxic dark period transferring to the gas phase during the following light period. This explanation may account for the elevated N2O emissions observed over the oxic light period on the fourth day of the nitrogen-filled headspace run (Figure 2). Also, denitrification can occur at bulk liquid DO levels above 0.2 ppmv if bacteria-formed flocs exist in biofilms where oxygen transport limitations allow bacteria to experience lower DO than in the surrounding bulk liquid. This may be an explanation for the N2O emissions observed at the higher DO levels of the light periods. The production of N2O observed under oxic conditions could be accounted for by the presence of microbial nitrifiers or other N2O production mechanisms. To further test the denitrifying bacteria hypothesis, the microalgae cultures were grown in the presence of antibiotics and tests were performed to detect bacterial genes involved in N2O production. These results are detailed in the following sections. 3.2.2. Detection of Genes Involved in N2O Production. The bacterial denitrification pathway reduces nitrate (NO3 ) to nitrogen gas through a multistep process, and N2O is an intermediate in this pathway that is produced by reduction of NO to N2O. This step is catalyzed by nitric oxide reductase, which is encoded by the norB gene.24,54 56 Therefore, to investigate the hypothesis that denitrifying bacteria are producing N2O emissions, PCR assays were used to determine if the norB gene was present in the experimental cultures under investigation. The norB gene was detected indicating that denitrifying bacteria with the ability to reduce NO to N2O were present in the culture. Results are presented in SI. No gene detection was performed for the nitrification pathway. 3.2.3. Antibiotic Cultivation Results. Broad spectrum antibiotics are known to inhibit the growth of bacteria, and it has been shown that penicillin g and streptomycin will effectively kill both gram-positive and gram-negative bacteria while having a minimal effect on the microalgae cultures.28 To determine whether inactivating bacteria present would lead to a reduction in N2O production, these broad spectrum antibiotics were added to the microalgae culture media and N2O concentrations were measured in the headspace gases using the same nitrogen headspace test protocol described above. The microalgal growth rate and nitrate depletion rate were consistent between the cultures treated with antibiotics and control culture without antibiotics and the rates were also comparable to the previous experimental runs. The N2O production in the antibiotic treated cultures was significantly less than that of the untreated flask with an average N2O reduction of 78.9% over the 4-day experiment. A complete reduction in N2O emissions was not expected because the antibiotics take time to inhibit growth of the bacteria in the culture media, and some bacterial strains may have been resistant to the antibiotics used. However, these results strongly support the hypothesis that N2O production by denitrifying bacteria is a primary contributor to N2O production in microalgae cultures. It is important to note that bacterial nitrification was not investigated and may also contribute to N2O production; more experimentation would be required. 3.3. Conditions and Control Techniques for N2O Production. The results of this study indicate that microalgae cultures produce significant levels of N2O under anoxic conditions in the presence of nitrate. In the microalgae growth stage, nitrate is supplied in the form of dissolved fertilizer at the beginning of the batch growth process. The uptake rate of nitrogen by the
ARTICLE
microalgae is a light dependent process and nitrate was depleted within 3 days. 10,57,58 During photosynthetic active periods the microalgae produce oxygen and therefore grow in an oxic environment. 52 Accordingly, there is a small window in the microalgae cultivation process for the production of N2O by denitrifying bacteria. This window consists of the first few dark periods after inoculation of the culture when nitrate is present in the growth media and anoxic conditions exist. Direct N2O emissions can be reduced or eliminated by understanding the mechanism for the formation of N2O from nitrogen-based fertilizers and by creating an environment where the reduction of nitrate to N2O is not favorable during the cultivation of microalgae. Several possible methods for controlling N2O production exist. Steps could be taken to prevent bacterial contamination in microalgae cultures to effectively remove the pathway for N2O production. However, it would be impractical to maintain axenic cultures in a large-scale cultivation system due to the risk of contamination, high cost of antibiotics, and potential for antibiotics to reduce microalgae growth rates or lead to the spread of antibiotic resistant bacteria.28 Maintaining a DO level during the dark periods of greater than 0.2 ppmv has been reported to inhibit the reduction of nitrogen by some bacteria by inhibiting the production or activity of enzymes involved in denitrification.59,60 An oxygen level well above this threshold for denitrification can be achieved by cultivating under 24 h light, sparging air through the culture, or maintaining an oxygen rich headspace while mixing the culture. To maintain the oxygen level in a PBR, a sparge system could be operated at night to generate an oxic environment while nitrate is present, thereby eliminating the potential for denitrification and the production of N2O. In the case of controlling emissions for an ORP growth system, there are challenges present such as environmental contamination with denitrifying bacteria, development of anoxic regions due to incomplete mixing, and a buildup of sludge and decaying material at the bottom of the pond. Therefore, to ensure minimal N2O emissions, an ORP must be thoroughly mixed at night to avoid the development of anoxic zones. 3.4. Implications for LCA Evaluations. Based on the data presented in this study, future microalgae LCAs must consider the potential for direct N2O emissions from the cultivation stage. Direct N2O emissions can be included in the LCA, which will effectively increase the GHGs; this is the only option for an ORP that does not thoroughly mix the microalgae culture. Alternatively, the energy required to fully mix an ORP or sparge PBRs with air at night could be included, as in Batan et al.9 Based on the results from this study, the recommended direct N2O emissions to be included in future LCAs are 0.002% of the nitrogen fertilizer applied to a fully oxic culture (such as a well mixed ORP or PBR sparged with air) and 0.390% for a microalgae culture that will be anoxic during dark periods (including a closed PBR with no sparging), a considerable difference from the 1% IPCC standard used for terrestrial crops.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional details including schematic diagrams, PCR primer sequences, calibration curves, detailed calculations, the denitrification pathway, and photographs of the experimental apparatus. This information is available free of charge via the Internet at http://pubs.acs.org.
9454
dx.doi.org/10.1021/es202573f |Environ. Sci. Technol. 2011, 45, 9449–9456
Environmental Science & Technology
’ AUTHOR INFORMATION Corresponding Author
*E-mail: [email protected]; phone: (970) 491-2328; fax: (970) 491-3827.
’ ACKNOWLEDGMENT This work was supported by the Department of Energy through the National Alliance for Advanced Biofuels and Bioproducts. We also gratefully acknowledge the support of Solix BioSystems who provided access to facilities and microalgae cultures. ’ REFERENCES (1) Dismukes, G. C.; Carrieri, D.; Bennette, N.; Ananyev, G. M.; Posewitz, M. C. Aquatic phototrophs: Efficient alternatives to landbased crops for biofuels. Curr. Opin. Biotechnol. 2008, 19 (3), 235–240. (2) Brown, L. M.; Zeiler, K. G. Aquatic biomass and carbon-dioxide trapping. Energy Convers. Manage. 1993, 34 (9 11), 1005–1013. (3) Li, Y.; Horsman, M.; Wu, N.; Lan, C. Q.; Dubois-Calero, N. Biofuels from microalgae. Biotechnol. Prog. 2008, 24 (4), 815–820. (4) Raja, R.; Hemaiswarya, S.; Kumar, N. A.; Sridhar, S.; Rengasamy, R. A perspective on the biotechnological potential of microalgae. Crit. Rev. Microbiol. 2008, 34 (2), 77–88. (5) Posten, C.; Schaub, G. Microalgae and terrestrial biomass as source for fuels-a process view. J. Biotechnol. 2009, 142 (1), 64–69. (6) Williams, P. R. D.; Inman, D.; Aden, A.; Heath, G. A. Environmental and sustainability factors associated with next-generation biofuels in the US: What do we really know? Environ Sci. Technol. 2009, 43 (13), 4763–4775. (7) Schenk, P. M.; Thomas-Hall, S. R.; Stephens, E.; Marx, U. C.; Mussgnug, J. H.; Posten, C.; Kruse, O.; Hankamer, B. Second generation biofuels: High-efficiency microalgae for biodiesel production. BioEnergy Res. 2008, 1 (1), 20–43. (8) Wijffels, R. H.; Barbosa, M. J. An outlook on microalgal biofuels. Science 2010, 329 (5993), 796–799. (9) Batan, L.; Quinn, J.; Willson, B.; Bradley, T. Net energy and greenhouse gas emission evaluation of biodiesel derived from microalgae. Environ. Sci. Technol. 2010, 44, 7975–7980. (10) Yamaberi, K.; Takagi, M.; Yoshida, T. Nitrogen depletion for intracellular triglyceride accumulation to enhance liquefaction yield of marine microalgal cells into a fuel oil. J. Mar. Biotechnol. 1998, 6 (1), 44–48. (11) Hirano, A.; Hon-Nami, K.; Kunito, S.; Hada, M.; Ogushi, Y. Temperature effect on continuous gasification of microalgal biomass: Theoretical yield of methanol production and its energy balance. Catal. Today 1998, 45 (1 4), 399–404. (12) Aresta, M.; Dibenedetto, A.; Barberio, G. Utilization of macroalgae for enhanced CO2 fixation and biofuels production: Development of a computing software for an LCA study. Fuel Process. Technol. 2005, 1679–1693. (13) Lardon, L.; Helias, A.; Sialve, B.; Stayer, J. P.; Bernard, O. Lifecycle assessment of biodiesel production from microalgae. Environ. Sci. Technol. 2009, 43 (17), 6475–6481. (14) Clarens, A. F.; Resurreccion, E. P.; White, M. A.; Colosi, L. M. Environmental life cycle comparison of algae to other bioenergy feedstocks. Environ. Sci. Technol. 2010, 44 (5), 1813–1819. (15) Luo, D.; Hu, Z.; Choi, D. G.; Thomas, V. M.; Realff, M. J.; Chance, R. R. Life cycle energy and greenhouse gas emissions for an ethanol production process based on blue-green algae. Environ. Sci. Technol. 2010, 44 (22), 8670–8677. (16) Stephenson, A. L.; Kazamia, E.; Dennis, J. S.; Howe, C. J.; Scott, S. A.; Smith, A. G. Life-cycle assessment of potential algal biodiesel production in the United Kingdom: A comparison of raceways and airlift tubular bioreactors. Energy Fuels 2010, 24 (7), 4062–4077.
ARTICLE
(17) Campbell, P. K.; Beer, T.; Batten, D. Life cycle assessment of biodiesel production from microalgae in ponds. Bioresour. Technol. 2011, 102 (1), 50–56. (18) Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B., Tignor, M., Miller, H. L. Eds., Climate Change 2007: Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, U.K. and New York, 2007. (19) Sheehan, J.; Dunahay, T.; Benemann, J.; Roessler, P. A Look Back at the U.S. Department of Energy’s Aquatic Species Program: Biodiesel from Algae; NREL Report TP-580-24190; National Renewable Energy Laboratory, 1998. (20) Weathers, P. J. N2O evolution by green algae. Appl. Environ. Microbiol. 1984, 48 (6), 1251–1253. (21) Florez-Leiva, L.; Tarifeno, E.; Cornejo, M.; Kiene, R.; Farías, L. High production of nitrous oxide (N2O), methane (CH4) and dimethylsulphoniopropionate (DMSP) in a massive marine phytoplankton culture. Biogeosci. Discuss. 2010, 7 (5). (22) Greenwell, H. C.; Laurens, L. M. L.; Shields, R. J.; Lovitt, R. W.; Flynn, K. J. Placing microalgae on the biofuels priority list: A review of the technological challenges. J. R. Soc. Interface 2010, 7 (46), 703–726. (23) IPCC. 2006 IPCC Guidelines for National Greenhouse Gas Inventories; National Greenhouse Gas Inventories Programme: Japan, 2006; Vol. 4. (24) Jones, C. M.; Stres, B.; Rosenquist, M.; Hallin, S. Phylogenetic analysis of nitrite, nitric oxide, and nitrous oxide respiratory enzymes reveal a complex evolutionary history for denitrification. Mol. Biol. Evol. 2008, 25 (9), 1955–1966. (25) Morel, F.; Rueter, J.; Anderson, D.; Guillard, R. Aquil: A chemically defined phytoplankton culture medium for trace metal studies. J. Phycol. 1979, 15, 135–141. (26) Yeh, K.-L.; Chang, J.-S.; Chen, W.-m. Effect of light supply and carbon source on cell growth and cellular composition of a newly isolated microalga Chlorella vulgaris esp-31. Eng. Life Sci. 2010, 10 (3), 201–208. (27) Altenburger, R.; Kr€uger, J.; Eisentr€ager, A. Proposing a pH stabilised nutrient medium for algal growth bioassays. Chemosphere 2010, 78 (7), 864–870. (28) Andersen, R. A. Algal Culturing Techniques; Elsevier, 2005. (29) Kagann, R. H. Infrared absorption intensities for N2O. J. Mol. Spectrosc. 1982, 95 (2), 297–305. (30) Esler, M. B.; Griffith, D. W. T.; Wilson, S. R.; Steele, L. P. Precision trace gas analysis by FT-IR spectroscopy. 1. Simultaneous analysis of CO2, CH4, N2O, and CO in air. Anal. Chem. 2000, 72 (1), 206–215. (31) Mohn, J.; Zeeman, M. J.; Werner, R. A.; Eugster, W.; Emmenegger, L. Continuous field measurements of δ13C-CO2 and trace gases by FTIR spectroscopy. Isot. Environ. Health Stud. 2008, 44, 241–251. (32) Esler, M. B.; Griffith, D. W. T.; Turatti, F.; Wilson, S. R.; Rahn, T.; Zhang, H. N2O concentration and flux measurements and complete isotopic analysis by FTIR spectroscopy. Chemosphere: Global Change Sci. 2000, 2 (3 4), 445–454. (33) Allen, M. T. The thermal decomposition of nitrous oxide and its reaction with hydrogen, carbon monoxide and methane. Princeton University, 1996. (34) Allen, M. T.; Yetter, R. A.; Dryer, F. L. Hydrogen/nitrous oxide kinetics - implications of the NxHy species. Combust. Flame 1998, 112 (3), 302–311. (35) Allen, M. T.; Yetter, R. A.; Dryer, F. L. High pressure studies of moist carbon monoxide/nitrous oxide kinetics. Combust. Flame 1997, 109 (3), 449–470. (36) Rivallan, M.; Ricchiardi, G.; Bordiga, S.; Zecchina, A. Adsorption and reactivity of nitrogen oxides (NO2, NO, N2O) on Fe-zeolites. J. Catal. 2009, 264 (2), 104–116. (37) Maiella, P. G.; Schoppelrei, J. W.; Brill, T. B. Spectroscopy of hydrothermal reactions. Part XI: Infrared absorptivity of CO2 and N2O in water at elevated temperature and pressure. Appl. Spectrosc. 1999, 53 (3), 351–355. 9455
dx.doi.org/10.1021/es202573f |Environ. Sci. Technol. 2011, 45, 9449–9456
Environmental Science & Technology (38) Ryden, J. C.; Lund, L. J.; Focht, D. D. Direct in-field measurement of nitrous oxide flux from soils. Soil Sci. Soc. Am. J. 1978, 42 (5), 731–737. (39) Mosier, A. R.; Mack, L. Gas chromatographic system for precise, rapid analysis of nitrous oxide. Soil Sci. Soc. Am. J. 1980, 44 (5), 1121–1123. (40) Virkaj€arvi, P.; Maljanen, M.; Saarij€arvi, K.; Haapala, J.; Martikainen, P. J. N2O emissions from boreal grass and grass - clover pasture soils. Agric. Ecosyst. Environ. 2010, 137 (1 2), 59–67. (41) Maljanen, M.; Liikanen, A.; Silvola, J.; Martikainen, P. J. Measuring N2O emissions from organic soils by closed chamber or soil/snow N2O gradient methods. Eur. J. Soil Sci. 2003, 54 (3), 625–631. (42) Pouchert, C. J., The Aldrich Library of FT-IR Spectra, 1st ed.; Aldrich Chemical Company, Inc.: Milwaukee, WI, 1989. (43) Callomon, H. J.; McKean, D. C.; Thompson, H. W. Intensities of vibration bands. III. Nitrous oxide. Proc. R. Soc. London, Ser. A 1951, 208 (1094), 332–341. (44) Kotzerke, A.; Klemer, S.; Kleineidam, K.; Horn, M.; Drake, H.; Schloter, M.; Wilke, B.-M. Manure contaminated with the antibiotic sulfadiazine impairs the abundance of nirK- and nirS-type denitrifiers in the gut of the earthworm Eisenia fetida. Biol. Fertil. Soils 2010, 46 (4), 415–418. (45) Golterman, H., Denitrification in the nitrogen cycle; Plenum Press: New York, 1985. (46) Horn, M. A.; Drake, H. L.; Schramm, A. Nitrous oxide reductase genes (nosZ) of denitrifying microbial populations in soil and the earthworm gut are phylogenetically similar. Appl. Environ. Microbiol. 2006, 72 (2), 1019–1026. (47) Delwiche, C. C. Denitrification, Nitrification, and Atmospheric Nitrous Oxide; John Wiley & Sons: New York, 1981. (48) Bothe, H. Biology of the Nitrogen Cycle, 1st ed.; Elservier: Amsterdam, Netherlands, 2007. (49) Braker, G.; Tiedje, J. M. Nitric oxide reductase (norB) genes from pure cultures and environmental samples. Appl. Environ. Microbiol. 2003, 69 (6), 3476–3483. (50) Lee, S.-W.; Im, J.; DiSpirito, A.; Bodrossy, L.; Barcelona, M.; Semrau, J. Effect of nutrient and selective inhibitor amendments on methane oxidation, nitrous oxide production, and key gene presence and expression in landfill cover soils: Characterization of the role of methanotrophs, nitrifiers, and denitrifiers. Appl. Microbiol. Biotechnol. 2009, 85 (2), 389–403. (51) Maniatis, T.; Fritsch, E. F.; Sambrook, J. Molecular Cloning: A Laboratory Manual; Cold Spring Harbor Laboratory: Cold Spring Harbor, NY, 1982. (52) Skerman, V. B. D.; Macrae, I. C. The influence of oxygen on the reduction of nitrate by adapted cells of Pseudomonas denitrificans. Can. J. Microbiol. 1957, 3 (2), 215–230. (53) Jannasch, H. W. Denitrification as influenced by photosynthetic oxygen production. J. Gen. Microbiol. 1960, 23 (1), 55–63. (54) Hino, T.; Matsumoto, Y.; Nagano, S.; Sugimoto, H.; Fukumori, Y.; Murata, T.; Iwata, S.; Shiro, Y. Structural basis of biological N2O generation by bacterial nitric oxide reductase. Science 2010, 330 (6011), 1666–1670. (55) Kandeler, E.; Deiglmayr, K.; Tscherko, D.; Bru, D.; Philippot, L. Abundance of narG, nirS, nirK, and nosZ genes of denitrifying bacteria during primary successions of a glacier foreland. Appl. Environ. Microbiol. 2006, 72 (9), 5957–5962. (56) Chon, K.; Chang, J.-S.; Lee, E.; Lee, J.; Ryu, J.; Cho, J. Abundance of denitrifying genes coding for nitrate (narG), nitrite (nirS), and nitrous oxide (nosZ) reductases in estuarine versus wastewater effluent-fed constructed wetlands. Ecol. Eng. 2009, 37 (1), 64–69. (57) Flynn, K. J.; Davidson, K.; Leftley, J. W. Carbon-nitrogen relations during batch growth of Nannochloropsis oculata (Eustigmatophyceae) under alternating light and dark. J. Appl. Phycol. 1993, 5 (4), 465–475. (58) Takagi, M.; Watanabe, K.; Yamaberi, K.; Yoshida, T. Limited feeding of potassium nitrate for intracellular lipid and triglyceride accumulation of Nannochloris sp UTEX LB1999. Appl. Microbiol. Biotechnol. 2000, 54 (1), 112–117.
ARTICLE
(59) Skerman, V. B. D.; Macrae, I. C. The influence of oxygen availability on the degree of nitrate reduction by pseudomonasdenitrificans. Can. J. Microbiol. 1957, 3 (3), 505–530. (60) Sacks, L. E.; Barker, H. A. The influence of oxygen on nitrate and nitrite reduction. J. Bacteriol. 1949, 58 (1), 11–22.
9456
dx.doi.org/10.1021/es202573f |Environ. Sci. Technol. 2011, 45, 9449–9456
ARTICLE pubs.acs.org/est
Interstitial Incorporation of Plutonium into a Low-Dimensional Potassium Borate Shuao Wang,† Juan Diwu,† Antonio Simonetti,† Corwin H. Booth,‡ and Thomas E. Albrecht-Schmitt*,† †
Departments of Civil Engineering and Geological Sciences and of Chemistry and Biochemistry, University of Notre Dame, Notre Dame, Indiana 46556, United States ‡ Chemical Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States
bS Supporting Information ABSTRACT: The molten boric acid flux reaction of PuBr3 with KBO2 at 200 °C results in the formation of large light-yellow crystals of K[B5O7(OH)2] 3 H2O:Pu4+. Singlecrystal X-ray diffraction experiments on the Pu-doped K[B5O7(OH)2] 3 H2O demonstrate two features: (1) K[B5O7(OH)2] 3 H2O:Pu4+ adopts a one-dimensional borate chain structure with void spaces between the chains. (2) The doping plutonium atoms do not reside on the potassium sites. The latter are not fully occupied. Both laserablation inductively coupled plasma mass spectrometry (LA-ICP-MS) and energydispersive spectrometry analyses indicate that plutonium atoms are uniformly distributed in crystals of K[B5O7(OH)2] 3 H2O with an atomic K:Pu ratio of approximately 65:1 measured by LA-ICP-MS. UV vis NIR spectra taken from both freshly made and one day old crystals show that the plutonium present within the crystals is predominantly characterized by Pu(IV). A small amount of Pu(III) is also present initially, but slowly oxidized to Pu(IV) via interaction with oxygen in the air. X-ray absorption near-edge structure and extended X-ray absorption fine structure spectroscopic measurements confirm that plutonium is mainly present as a form similar to that of a PuO2 cluster. The combined results suggest that the clusters containing Pu(IV) ions are uniformly distributed in the void spaces between the borate chains.
’ INTRODUCTION Nuclear weapon testing and plutonium production during the cold war, and recent events such as the catastrophe at the Fukushima Daiichi nuclear plant in Japan, resulted in the contamination of large areas of oceans, groundwater, soils, and sediments by actinides, such as uranium and plutonium, along with their fission and decay products. Thus, possible migration of actinides within contaminated sites becomes an important environmental concern.1 3 Generally, the incorporation of actinides into natural materials retards their transport in the environment. Knowledge of the incorporation mechanisms of actinides into different types of natural materials is therefore required both for predicting the migration of radionuclides at repositories and at contaminated sites and for designing/manufacturing suitable radioactive waste forms.4 6 During the past two decades, significant efforts have been made to better understand the incorporation of early actinide elements (namely, Th, U, Np, and Pu) into a variety of mineral phases, such as calcite,7 15 zircon,4,16,17 monazite,4,17,18 zirconolite,4,18 22 perovskite,4,22 garnet,4 pyrochlore,4,23 25 brucite,26 and several other systems.27 30 Particularly for transuranium elements (Np and Pu), it was determined that these can also be incorporated into a series of uranium- or thorium-based minerals such as brannerite,4 ianthinite,31 and schoephite, becquerelite, compreignacite, and boltwoodite 32 or even synthetic phases such as Th2 x/2AnIVx/2(PO4)2(HPO4) 3 H2O (An = U, Np, Pu),33 r 2011 American Chemical Society
ThSiO4,34 and Ba3(UO2)2(HPO4)2(PO4)2.35 The majority of these studies focus on examining incorporation mechanisms based on the presence of a suitable crystallographic site in the lattice for actinide units to reside in the incorporated materials; i.e., the actinide units substitute into lattice sites that are occupied by other cations that have similar crystal chemistry. This commonly occurring mechanism is termed “substitutional incorporation”. For example, numerous studies indicate that trace amounts of uranium, present as both U(VI) and U(IV), can be incorporated into natural calcite.7 14 U(IV) was found to have a stable location on the sites of divalent calcium,7 while U(VI) occupies mainly calcium sites along with other possible defects or disordered sites.8,9 For cases of plutonium incorporation, Pu(IV) substitutes into calcium sites in zirconolites,19 and Pu(III) can enter the structure of monazite in the position of La(III).18 Furthermore, Pu(VI) can readily occur in many structures containing U(VI) by substituting onto the U(VI) sites.31,32 In light of all these observations, one could conclude that the incorporation of actinides requires suitable crystallographic sites in the lattice of materials. We have recently undertaken a study involving the preparation, structural elucidation, and physicochemical property Received: August 12, 2011 Accepted: September 20, 2011 Revised: September 16, 2011 Published: September 20, 2011 9457
dx.doi.org/10.1021/es2028247 | Environ. Sci. Technol. 2011, 45, 9457–9463
Environmental Science & Technology measurements of actinide borates36 47 with the aim of uncovering the solubility-limiting phases of actinides in some repositories such as the Waste Isolation Pilot Plant (WIPP), which is near Carlsbad, NM. A similar repository is being considered in Germany. In this salt deposit the concentration of borate species in intergranular brines can be as high as 166 ppm.48 Studies of the complexation of Nd(III) by borates in solution have been performed indicating that borate is the primary complexant in WIPP for trivalent cations, such as Pu(III).49 Furthermore, WIPP is self-sealing, and once closed will be saturated with hydrogen and methane, making it a highly reducing environment that will favor lower oxidation states for plutonium. In an attempt to prepare a plutonium(IV) borate compound that might serve as a model for plutonium(IV) borates in WIPP, a potassium borate that is doped with Pu(IV) was discovered. This compound, with a formula of K[B5O7(OH)2] 3 H2O, contains no suitable lattice sites for plutonium substitution. However, it was determined that Pu(IV) was capable of entering the structure of K[B5O7(OH)2] 3 H2O with an atomic ratio K:Pu around 65:1, as measured by laserablation inductively coupled plasma mass spectrometry (LAICP-MS). We have used single-crystal X-ray diffraction, X-ray absorption near-edge structure (XANES) spectroscopy, and extended X-ray absorption fine structure (EXAFS) spectroscopy to demonstrate that Pu(IV) can enter the interstitial spaces between neighboring borate chains while removing some lattice potassium to maintain charge balance. Interstitial incorporation suggests an approach for understanding how actinides can enter natural materials, although interstitial incorporation was found in a few other, nonactinide-bearing systems,50 52 as well as in NZP (sodium zirconium phosphate) type materials.53 There are several objectives for this work. First, we demonstrate that plutonium can incorporate into interstitial sites in naturally occurring phases that have low-dimensional networks. Second, we show that while plutonium can be detected visually, as well as with a variety of spectroscopic techniques, its presence can only be inferred from X-ray diffraction data. Finally, we address the broader environmental implications of interstitial incorporation of actinides into low-dimensional minerals.
’ EXPERIMENTAL SECTION Syntheses. Specialized facilities and procedures are needed for this work. All free-flowing solids were handled within negative-pressure gloveboxes, and products were examined when coated with either water or Krytox oil and water. 242PuO2 (99.98% isotopic purity, Oak Ridge National Laboratory, t1/2 = 3.76 105 y) was used as received. While the plutonium is of very high isotopic purity, there are trace amounts of 238Pu, 240 Pu, 241Pu, 244Pu, and 241Am. The majority of the radioactivity comes from the 241Pu even though it represents only 0.008% of the plutonium. Caution: 242Pu still represents a serious health risk owing to its α and γ emission! This isotope was selected because of its long half-life, which increased the longevity of the crystals. A 0.365 M stock solution of plutonium(VI)-242 nitrate was prepared by first digesting PuO2 in 8 M HNO3 for 3 days at 200 °C in an autoclave. The solution was then reduced to a moist residue and redissolved in water. This solution was then ozonated for approximately 5 h to ensure complete oxidation of the plutonium to +6. An aliquot containing 5 mg of plutonium(VI) was taken from the above stock solution and reduced to a residue. A 50 μL volume of concentrated HBr was added to this residue, resulting
ARTICLE
in the immediate formation of bromine gas and a red solution (the red color is from the dissolved bromine, which masks the color of Pu(III)). The red solution was reduced to a purple-black residue at 130 °C. The residue was redissolved in 30 μL of argonsparged water, producing a navy blue-purple solution characteristic of Pu(III). A droplet of Pu(III) was then transferred to a PTFE autoclave liner. A large excess of boric acid (47 mg) and KBO2 (7.8 mg) were then added directly to the droplet containing Pu(III). The mixture was then sealed in an autoclave and heated at 200 °C for three days followed by slow cooling to room temperature over a period of two days. The autoclave was then opened, and boiling water was added to dissolve the excess boric acid. The wash solution was almost colorless, which indicates that an insignificant amount of Pu was released upon washing. Plutonium in solution was not detected by UV vis spectroscopy. Yellow crystals of K[B5O7(OH)2] 3 H2O:Pu4+ with a tablet habit were then isolated as a pure product. Crystallographic Studies. A crystal of K[B5O7(OH)2] 3 H2O: Pu4+ was mounted on a CryoLoop with Krytox oil and optically aligned on a Bruker APEXII Quazar X-ray diffractometer using a digital camera. Initial intensity measurements were performed using an IμS X-ray source, a 30 W microfocused sealed tube (Mo Kα, λ = 0.71073 Å) with high-brilliance and high-performance focusing Quazar multilayer optics. Standard APEXII software was used for determination of the unit cells and data collection control. The intensities of reflections of a sphere were collected by a combination of four sets of exposures. Each set had a different j angle for the crystal, and each exposure covered a range of 0.5° in ω. A total of 1464 frames were collected with an exposure time per frame of 30 s. SAINT software was used for data integration, including Lorentz and polarization corrections. Semiempirical absorption corrections were applied using the program SCALE (SADABS).54 The crystal structure of K[B5O7(OH)2] 3 H2O:Pu4+ was solved by direct methods. All atoms were refined anisotropically. The site occupancy factor refinement on the potassium sites showed that the sites are not fully occupied with approximately 2.0% defects. However, this number is not reliable. Instead the occupancy of the K sites was fixed on the basis of the ICP-MS data (vide infra). Also, the substitutional disorder refinement on the potassium sites based on both K and Pu atoms failed to provide a reasonable occupancy, indicating the doping Pu atoms do not occupy K sites. However, it was still not possible to locate the doping Pu atoms in the final difference Fourier map, which indicates Pu atoms are highly disordered within the void space between the borate chains. LA-ICP-MS Analysis. Laser-ablation analysis of nine different crystals of K[B5O7(OH)2] 3 H2O:Pu4+ was conducted using a ThermoFinnigan high-resolution magnetic sector Element2 ICP-MS instrument coupled to a UP213 Nd:YAG laser-ablation system (New Wave Research). Selected crystals were fixed on 1 in. round glass slides with double-sided tape. Individual analyses consisted of 60 s measurement of background ion signals followed by a 60 s interval of measurement of ion signals (11B, 39K, 242Pu) subsequent to the start of lasering. Each analysis represents a total of 93 scans (93 runs 1 pass) with a sample (dwell) time of 0.01 s with 20 samples per ion signal peak. Analyses were conducted in medium mass resolution mode (resolution = mass/peak width ≈ 4000) to eliminate possible spectral interferences. The ablated particles were transported from the ablation cell to the ICP-MS instrument using He carrier gas at a flow rate of 0.7 L/min. Crystals were ablated using a range of spot sizes between 40 and 55 μm, 9458
dx.doi.org/10.1021/es2028247 |Environ. Sci. Technol. 2011, 45, 9457–9463
Environmental Science & Technology a repetition rate of 2 Hz, and 70% power output corresponding to an energy density of 12 15 J/cm2. Using these ablation conditions, the depth of penetration of the laser is between 5 and 15 μm.55,56 Scanning Electron Microscopy and Energy-Dispersive Spectrometry (SEM/EDS) Analysis. SEM/EDS images and data were collected using a LEO EVO 50 with an Oxford INCA energy-dispersive spectrometer. The energy of the electron beam was set at 29.02 kV, and the spectrum acquisition time was 120 s. All of the data were calibrated with standards, and all EDS results are provided in the Supporting Information. Plutonium in the K[B5O7(OH)2] 3 H2O:Pu4+ was easily identified in the energydispersive spectrum (Supporting Information). UV Vis NIR Absorption Spectroscopy. UV vis NIR data were acquired from both freshly made and one day old crystals of K[B5O7(OH)2] 3 H2O:Pu4+ using a Craic Technologies microspectrophotometer. Crystals were placed on quartz slides under Krytox oil, and the data were collected from 500 to 1400 nm. The exposure time was optimized automatically by the Craic software. X-ray Absorption Spectroscopy. A sample of crystalline K[B5O7(OH)2] 3 H2O:Pu4+ was measured at the Pu LIII edge on beamline 11-2 at the Stanford Synchrotron Radiation Lightsource (SSRL) using a Si(220) double monochromator (j=0°) detuned by 50% to avoid unwanted harmonics. Data were collected in fluorescence mode using a multielement Ge detector and corrected for dead time. The data were also corrected for selfabsorption57 assuming 5% of the K atoms are replaced by Pu. The EXAFS data were reduced and fit using the RSXAP analysis suite in r-space58 using standard procedures.59 The XANES data are shown in Figure 5, and both the EXAFS data and fit are shown in Figure 6. EXAFS fits utilize scattering amplitudes and phases calculated with the FEFF8.1 code60 based on a PuO2 fluorite structure.
’ RESULTS AND DISCUSSION Syntheses. The pure phase of K[B5O7(OH)2] 3 H2O:Pu4+
can be synthesized by the molten boric acid flux reaction of PuBr3 with KBO2 at 200 °C. Apparently, Pu(III) was slowly oxidized to Pu(IV) by O2 in the atmosphere during the reaction. We found that a similar reaction in the absence of KBO2 in the starting materials will result in the formation of a plutonium(IV) borate. Thus, the presence of KBO2 will kinetically eliminate the formation of plutonium(IV) borates. K[B5O7(OH)2] 3 H2O:Pu4+ forms large yellow crystals with a tablet habit and are shown in Figure 1. The crystals can be cut, and the interior shows the same color as the surface, which excludes the case of surface absorption of Pu(IV) on the crystals of K[B5O7(OH)2] 3 H2O. Given that pure, nondoped K[B5O7(OH)2] 3 H2O crystals are expected to be colorless, the yellow coloration indicates high levels of plutonium incorporation. Crystallographic Studies. The synthesis and crystal structure of K[B5O7(OH)2] 3 H2O were recently reported by Zhang et al.61 The structure of K[B5O7(OH)2] 3 H2O contains a series of double-helical one-demensional borate chains shown in Figure 2a. The potassium atoms reside in the void space within the doublehelical borate chains. There are larger void spaces present between the neighboring double chains (Figure 2b). The vertical distance for these void spaces along the a axis is approximately 3.9 Å, which is compatible for trapping the 8-coordinated Pu(IV)O 8 cubic polyhedra (Figure 2b). It should be noted
ARTICLE
Figure 1. Photo of crystals of K[B5O7(OH)2] 3 H2O:Pu4+.
Figure 2. (a) Skeletal structure of double-helical one-dimensional borate chains in K[B5O7(OH)2] 3 H2O. (b) Cartoon of the incorporation mechanism of Pu(IV) into crystals of K[B5O7(OH)2] 3 H2O. Potassium atoms are shown in purple, borate units (both BO3, triangles, and BO4, tetrahedra) are shown in blue, and plutonium atoms are shown as yellow polyhedra.
Table 1. Crystallographic Information for K[B5O7(OH)2] 3 H2O,61 K[B5O7(OH)2] 3 H2O:Pu4+, and Larderellite (NH4[B5O7(OH)2] 3 H2O)62 K[B5O7(OH)2] 3 H2O
K[B5O7(OH)2] 3 H2O:Pu4+
larderellite
color
colorless
yellow
colorless
space group
P21/c
P21/c
P21/c
a (Å)
9.4824(8)
9.4619(12)
9.47(1)
b (Å)
7.5180(6)
7.4691(10)
7.63(1)
c (Å)
11.4154(6)
11.3740(15)
11.65(1)
β (deg)
97.277(3)
97.412(1)
97.08(25)
V (Å3)
807.2(1)
797.1(2)
835.37
ref
60
this work
61
that although K[B5O7(OH)2] 3 H2O does not have an exact natural mineral analogue and can only be synthesized in the laboratory, the structure of larderellite, with the formula of NH4[B5O7(OH)2] 3 H2O,62 is isotypic with K[B5O7(OH)2] 3 H2O (Table 1). By incorporating Pu(IV), the unit cell parameters for K[B5O7(OH)2] 3 H2O:Pu4+ are slightly smaller than those for K[B5O7(OH)2] 3 H2O reported by Yang et al. (Table 1). Approximate 2.0% defects were shown on K1 sites on the basis of the site occupancy factor refinement. However, we have fixed the 9459
dx.doi.org/10.1021/es2028247 |Environ. Sci. Technol. 2011, 45, 9457–9463
Environmental Science & Technology
ARTICLE
Figure 3. Illustration of a typical time-resolved spectrum (ion signal vs time) for an LA-ICP-MS analysis of a single K[B5O7(OH)2] 3 H2O:Pu4+ crystal. The time interval between 0 and ∼60 s represents the background measurement, whereas the subsequent interval represents measurement of ion signals after the start of lasering.
Figure 4. UV vis NIR absorption spectra taken from both freshly made (black) and one day old (red) crystals of K[B5O7(OH)2] 3 H2O: Pu4+.
occupancy to that of the ICP-MS data, which is far more reliable, although the occupancy refinement and ICP-MS data only differ by 0.5%. More importantly, the substitutional disorder refinement on K1 sites based on both K and Pu atoms fails to give a reasonable second free variable for occupancy determination. Together, these facts indicate that the doping Pu atoms are not occupying the K sites. This conclusion is reasonable from the point of view that there are large differences between Pu(IV) and K+ in both coordination geometry and ionic radius.63 Thus, the doping by Pu(IV) is only possible by allowing it to reside in the void spaces between the neighboring double borate chains (Figure 2b). As a consequence of Pu(IV) entering the structure, some of the K+ cations are lost to maintain charge balance, which results in the 2.0% defects on the K sites. LA-ICP-MS Studies. Figure 3 demonstrates a typical timeresolved spectrum (i.e., ion signal measured in counts per second (cps) vs time (s)) for an LA-ICP-MS analysis of a single K[B5O7(OH)2] 3 H2O:Pu4+ crystal in medium mass resolution mode. It is evident from Figure 3 that 242Pu is incorporated into the crystal (and not solely onto the crystal’s surface) since its ion signal measured during the lasering interval is concomitant with those recorded for 11B and 39K and importantly does not decrease rapidly with time; the latter feature is consistent for atoms incorporated solely within the surface of the crystal (∼first micrometer of depth). Moreover, repeated analysis (n = 9) by LA-ICP-MS of individual K[B5O7(OH)2] 3 H2O:Pu4+ crystals yielded an average, calculated K:Pu “mass” ratio of 3.37 ( 0.7. However, a more accurate assessment of the K:Pu ratio should take into account the lower ionization efficiency or lower “ion yield” (ion signal counts per second/concentration unit) of K versus that for Pu in a plasma environment. In general, the ion yield for elements analyzed by an ICP-MS instrument is mass dependent, with “heavier” elements (>80 amu 1) recording higher values compared to those for lighter elements. The differential ion yield for K versus Pu was assessed in medium mass resolution via the measurement of four solutions, each containing a 1:1 mass ratio of K:U at variable concentrations (∼10 50 ppb). For ease of sample preparation, uranium was used as a valid proxy for Pu given their similar atomic masses.
The average, measured K:U mass ratio obtained for the solutions was 1:3.12. Hence, this result when applied to the laser ablation values yields an average normalized: K:Pu mass ratio of ∼10.5:1 and a corresponding atomic ratio for K:Pu of ∼65:1 for the K[B5O7(OH)2] 3 H2O:Pu4+ crystals. UV Vis NIR Absorption Spectroscopy. By possessing complicated f electron configurations, both Pu(IV) and Pu(III) are known to produce a series of weak, Laporte-forbidden f f transitions in the vis NIR region in both solution and the solid state.64,65 In solution, the spectrum of Pu(IV) consists of characteristic transitions such as the peaks near 477, 660, 800, and 1080 nm.64 For Pu(III), the most important transition that can be used for distinguishing it from Pu(IV) is at 920 nm.64 The UV vis NIR absorption spectra taken from both freshly made and one day old crystals of K[B5O7(OH)2] 3 H2O:Pu4+ are shown in Figure 4. It is found that the freshly made crystals contain predominantly Pu(IV), whereas a very small amount of Pu(III) is also present on the basis of the weak peak at 926 nm. In the spectrum of one day old crystals, the 926 peak disappears, which indicates that the tiny amount of remaining Pu(III) incorporated into the crystals is rapidly oxidized by ambient O2 in the air. Nevertheless, the results indicate that both Pu(III) and Pu(IV) are capable of entering the structure of K[B5O7(OH)2] 3 H2O. It should be noted that the incorporation of Pu(III) into K[B5O7(OH)2] 3 H2O still does not prevent the oxidation of Pu(III). This oxidation contrasts sharply with the fact that many Pu(III) compounds are not air-sensitive.46,66 Apparently, Pu(III) can be further stabilized by the lattice energy in most Pu(III) compounds, while the lattice energy is lacking for Pu(III) in crystals of K[B5O7(OH)2] 3 H2O since plutonium atoms are highly disordered in the void space. XANES and EXAFS Studies. The XANES data shown in Figure 5 are clearly consistent with a Pu(IV) state for the K[B5O7(OH)2] 3 H2O:Pu4+ sample, as indicated by the consistent position of the white line when compared to data from PuO2. Moreover, the first EXAFS oscillation (peak at ∼18100 eV) is practically identical to that of PuO2. The increased peak height of the data indicates differences in the long-range structure, for instance, due to particle or cluster size, which tends to increase the amplitude of the white line for smaller clusters.67 9460
dx.doi.org/10.1021/es2028247 |Environ. Sci. Technol. 2011, 45, 9457–9463
Environmental Science & Technology
ARTICLE
Table 2. Fit Results for Pu LIII-Edge EXAFS Data on K[B5O7(OH)2] 3 H2O:Pu4+ at T = 30 Ka N Pu O
8
Pu Pu ΔE0
3.9(9)
S02 a
R (%)
σ2 (Å2)
R (Å)
0.010(1)
2.325(8)
0.0027(5) 11.0(9)
3.808(4)
1.14(14) 12.7
The fit range is between 1.8 and 3.9 Å. The k3-weighted data are transformed between 2.5 and 15.0 Å 1 and are Gaussian narrowed by 0.3 Å 1. The degree of freedom of the fit is estimated to be 11.7.68 Error estimates use a Monte Carlo method.69 The amplitude reduction factor for these fits is determined to be S02 = 1.14 ( 0.14, and the edge shift is ΔE0 = 11.0 ( 0.9 eV.
Figure 5. Pu LIII-edge XANES of K[B5O7(OH)2] 3 H2O:Pu4+ at 30 K, together with that of PuO2 collected at room temperature. Data in this figure and Figure 6 are displayed with error bars estimated by collecting several scans.
Figure 6. Pu LIII-edge EXAFS data from K[B5O7(OH)2] 3 H2O:Pu4+ at 30 K in both (a) k-space and (b) r-space. The Fourier transform (FT) in (b) shows the amplitude (outer envelope) and the real part (modulating line) of the transform and is taken between 2.5 and 15.0 Å 1, Gaussian narrowed by 0.3 Å 1. The fit shown in (b) is between 1.8 and 3.9 Å. Fit results are summarized in Table 2.
Some of this difference could be due to the difference in the measurement temperatures between the borate and the PuO2 sample. The EXAFS data in Figure 6 are also quite similar to those expected from PuO2 (not shown in the figure), although the Pu Pu peak at ∼3.7 Å in the spectra (corresponding to a Pu Pu
distance of about 3.8 Å) is actually less than half the expected amplitude. The fit results listed in Table 2 clearly indicate the similarity between the measured local structure and that of PuO2, as the number of Pu Pu neighbors is 3.9 ( 0.9, as opposed to the 12 Pu Pu neighbors in PuO2, the measured Pu O bond length of 2.325 ( 0.8 Å is typical for a Pu(IV) compound,66 and the amplitude reduction factor S02 is reasonable given an oxygen coordination of 8 around the Pu atoms. It should be noted, however, that the estimated error in S02 allows for 9 Pu O nearest neighbors. These data are not compatible with a significant number of Pu atoms occupying the K position in the structure of K[B5O7(OH)2] 3 H2O. In that structure, the K O near-neighbor shell has 10 neighbors with bond lengths between 2.76 and 3.22 Å, which would generate an EXAFS peak approximately 15% of the amplitude centered about 0.4 Å further in the Fourier transform of Figure 6. Even if one allows for a local rearrangement of O around Pu when Pu occupies a K site, one would not expect a Pu Pu pair at 3.8 Å in this structure, where the K K nearest distance is about 5.0 Å with only two neighbors. These data are therefore more consistent with PuO2 clusters, rather than a random distribution of Pu on K positions within the K[B5O7(OH)2] 3 H2O structure. Implications for the Migration of Radionuclides at Repositories. It is well-known that the release of radionuclides at repositories is retarded by the incorporation of radionuclides into natural materials. However, predictions of actinide incorporation based solely on lattice substitution mechanisms may be a limited way of predicting uptake by minerals. With our observation of incorporation occurring in K[B5O7(OH)2] 3 H2O:Pu4+, many natural materials that lack suitable sites for lattice substitution but that possess void spaces of suitable dimensions may also show the capacity for trapping certain radionuclides. Furthermore, studies focused on the incorporation of actinides into borate minerals are scarce. Borate minerals are known to have a very complicated and rich structural chemistry, and countless structural topologies have been observed in the borate mineral family.70,71 A variety of borate minerals, including larderellite (natural analouge of K[B5O7(OH)2] 3 H2O), as well as other one-dimensional materials, such as vimsite, colemanite, calciborite, hydroboracite, probertite, ezcurrite, kaliborite, kernite, aristarainite, etc.,71 are expected to form in the WIPP that can provide a series of structures containing void space with a wide range of dimensions. These void spaces could trap actinides, which would mitigate their transport. 9461
dx.doi.org/10.1021/es2028247 |Environ. Sci. Technol. 2011, 45, 9457–9463
Environmental Science & Technology
’ ASSOCIATED CONTENT
bS
Supporting Information. EDS spectrum and X-ray file (CIF) for K[B5O7(OH)2] 3 H2O:Pu4+. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail: [email protected]; phone: (574) 631-1872; fax: (574) 631-9236.
’ ACKNOWLEDGMENT We are grateful for support provided by the U.S. Department of Energy (DOE), Subsurface Biogeochemical Research Program, under Grant ER64804. Work at Lawrence Berkeley National Laboratory was supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. DOE under Contract DE-AC02-05CH11231. ’ REFERENCES (1) Novikov, A. P.; Kalmykov, S. N.; Utsunomiya, S.; Ewing, R. C.; Horreard, F.; Merkulov, A.; Clark, S. B.; Tkachev, V. V.; Myasoedov, B. F. Colloid transport of plutonium in the Far-Field of the Mayak Production Association, Russia. Science 2006, 314, 638–641. (2) Pockley, P. Clean-up strategy at Australian nuclear site called into question. Nature 2000, 404, 797–797. (3) Dai, M.; Kelly, J. M.; Buesseler, K. O. Sources and migration of plutonium in groundwater at the Savannah River Site. Environ. Sci. Technol. 2002, 36, 3690–3699. (4) Lumpkin, G. R. Ceramic waste forms for actinides. Elements 2006, 2, 365–372. (5) Ewing, R. C. Nuclear waste forms for actinides. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 3432–3439. (6) Donald, I. W.; Metcalfe, B. L.; Taylor, R. N. J. The immobilization of high level radioactive wastes using ceramics and glasses. J. Mater. Sci. 1997, 32, 5851–5887. (7) Sturchio, N. C.; Antonio, M. R.; Soderholm, L.; Sutton, S. R.; Brannon, J. C. Tetravalent uranium in calcite. Science 1998, 281, 971–973. (8) Kelly, S. D.; Troy Rasbury, E.; Chattopadhyay, S.; Jeremy Kropf, A.; Kemner, K. M. Evidence of a stable uranyl site in ancient organic-rich calcite. Environ. Sci. Technol. 2006, 40, 2262–2268. (9) Kelly, S. D.; Newville, M. G.; Cheng, L.; Kemner, K. M.; Sutton, S. R.; Fenter, P.; Sturchio, N. C.; Spotl, C. Uranyl incorporation in natural calcite. Environ. Sci. Technol. 2003, 37, 1284–1287. (10) Wang, Z.; Zachara, J. M.; Mckinley, J. P.; Smith, S. C. Cryogenic laser induced U(VI) fluorescence studies of a U(VI) substitued natural calcite: Implications to U(VI) speciation in contaminated Hanford sediments. Environ. Sci. Technol. 2005, 39, 2561–2659. (11) Heberling, F.; Denecke, M. A.; Bosbach, D. Neptunium(V) coprecipitation with calcite. Environ. Sci. Technol. 2008, 42, 471–476. (12) Reederm, R. J.; Elzinga, E. J.; Drew Tait, C.; Rector, K. D.; Donohoe, R. J.; Morris, D. E. Site-specific incorporation of uranyl carbonate species at the calcite surface. Geochim. Cosmochim. Acta 2004, 68, 4799–4808. (13) Valle-Fuentes, F.-J.; Garcia-Guinea, J.; Cremades, A.; Correcher, V.; Sanchez-Moral, S.; Gonzalez- Martin, R.; Sanchez-Munoz, L.; Lopez-Arce, P. Low-magnesium uranium-calcite with high degree of crystallinity and gigantic luminescence emission. Appl. Radiat. Isot. 2007, 65, 147–154. (14) Arai, Y.; Marcus, M. A.; Tamura, N.; Davis, J. A.; Zachara, J. M. Spectroscopic evidence for uranium bearing precipitates in Vadose Zone sediments at the Hanford 300-area. Environ. Sci. Technol. 2007, 41, 4633–4639.
ARTICLE
(15) Marques Fernandes, M.; Stumpf, T.; Rabung, T.; Bosbach, D.; Fanghanel, Th. Incorporation of trivalent actinides into calcite: A time resolved laser fluorescence spectroscopy (TRLFS) study. Geochim. Cosmochim. Acta 2008, 72, 464–474. (16) Ewing, R. C.; Lutze, W.; Weber, W. J. Zircon: A host-phase for the disposal of weapons plutonium. J. Mater. Res. 1995, 10, 243–246. (17) Meldrum, A.; Boatner, L. A.; Ewing, R. C. Displacive radiation effects in the monazite- and zircon-structure orthophosphates. Phys. Rev. B 1997, 56, 13805–13814. (18) Deschanels, X.; Picot, V.; Glorieux, B.; Jorion, F.; Peuget, S.; Roudil, D.; Jegou, C.; Broudic, V.; Cachia, J. N.; Advocat, T.; Den Auwer, C.; Fillet, C.; Coutures, J. P.; Hennig, C.; Scheinost, A. Plutonium incorporation in phosphate and titanate ceramics for minor actinide containment. J. Nucl. Mater. 2006, 352, 223–240. (19) Vance, E. R.; Ball, C. J.; Day, R. A.; Smith, K. L.; Blackford, M. G.; Begg, B. D.; Angel, P. J. Actinide and rare earth incorporation into zirconolite. J. Alloys Compd. 1994, 213/214, 406–409. (20) Clinard, F. W., Jr; Rohr, D. L.; Roof, R. B. Structural damage in a self-irradiated zirconolite-based ceramic. Nucl. Instrum. Methods Phys. Res., B 1984, 1, 581–586. (21) Weber, W. J.; Wald, J. W.; Matzke, H. Effects of self-radiation damage in Cm-doped Gd2Ti2O7 and CaZrTi2O7. J. Nucl. Mater. 1986, 138, 196–209. (22) Begg, B. D.; Vance, E. R.; Conradson, S. D. The incorporation of plutonium and neptunium in zirconolite and perovskite. J. Alloys Compd. 1998, 271 273, 221–226. (23) Nastren, C.; Jardin, R.; Somers, J.; Walter, M.; Brendebach, B. Actinide incorporation in a zirconia based pyrochlore (Nd1.8An0.2)Zr2O7+x (An = Th, U, Np, Pu, Am). J. Solid State Chem. 2009, 182, 1–7. (24) Ewing, R. C.; Weber, W. J.; Lian, J. Nuclear waste disposalpyrochlore (A2B2O7): Nuclear waste form for the immobilization of plutonium and “minor” actinides. J. Appl. Phys. 2004, 95, 5949–5971. (25) Strachan, D. M.; Scheele, R. D.; Buck, E, C.; Icenhower, J. P.; Kozelisky, A. E.; Sell, R. L.; Elovich, R. J.; Buchmiller, W. C. Radiation damage effects in candidate titanates for Pu disposition: Pyrochlore. J. Nucl. Mater. 2005, 345, 109–135. (26) Douglas Farr, J.; Neu, M. P.; Schulze, R. K.; Honeyman, B. D. Plutonium uptake by brucite and hydroxylated periclase. J. Alloys Compd. 2007, 444 445, 553–539. (27) Meis, C.; Gale, J. D.; Boyer, L.; Carpena, J.; Gosset, D. Theoretical study of Pu and Cs incorporation in a mono-silicate neodymium fluoroapatite Ca9Nd(SiO4)(PO4)5F2. J. Phys. Chem. A 2000, 104, 5380–5387. (28) Yamazaki, S.; Yamashita, T.; Matsui, T.; Nagasaki, T. Thermal expansion and solubility limits of plutonium-doped lanthanum zirconates. J. Nucl. Mater. 2001, 294, 183–187. (29) Vance, E. R.; Ball, C. J.; Begg, B. D.; Carter, M. L.; Day, R. A.; Thorogood, G. J. Pu, U, and Hf incorporation in Gd silicate apatite. J. Am. Ceram. Soc. 2003, 86, 1223–1225. (30) Holiday, K.; Hartmann, T.; Mulcahy, S. R.; Czerwinski, K. Synthesis and characterization of zirconia-magnesia inert matrix fuel: plutonium studies. J. Nucl. Mater. 2010, 402, 81–86. (31) Burns, P. C.; Finch, R. J.; Hawthorne, F. C.; Miller, M. L.; Ewing, R. C. The crystal structure of ianthinite, [U24+(UO2)4O6(OH)4(H2O)4](H2O)5: A possible phase for Pu4+ incorporation during the oxidation of spent nuclear fuel. J. Nucl. Mater. 1997, 249, 199–206. (32) Burns, P. C.; Ewing, R. C.; Miller, M. L. Incorporation mechanisms of actinide elements into the structures of U6+ phases formed during the oxidation of spent nuclear fuel. J. Nucl. Mater. 1997, 245, 1–9. (33) Dacheux, N.; Grandjean, S.; Rousselle, J.; Clavier, N. Hydrothermal method of preparation of actinide(IV) phosphate hydrogenphosphate hydrates and study of their conversion into actinide(IV) phosphate diphosphate solid solutions. Inorg. Chem. 2007, 46, 10390– 10399. (34) Krupa, J. C.; Carnall, W. T. Electronic structure of U4+, Np4+, and Pu4+ doped into ThSiO4 single crystal. J. Chem. Phys. 1993, 99, 8577–8584. 9462
dx.doi.org/10.1021/es2028247 |Environ. Sci. Technol. 2011, 45, 9457–9463
Environmental Science & Technology (35) Wu, S.; Chen, F.; Simonetti, A.; Albrecht-Schmitt, T. E. Incorporation of neptunium(V) and iodate into a uranyl phosphate: Implications for mitigating the release of 237Np and 129I in repositories. Environ. Sci. Technol. 2010, 44, 3192–3196. (36) Wang, S.; Alekseev, E. V.; Diwu, J.; Casey, W. H.; Phillips, B. L.; Depmeier, W.; Albrecht-Schmitt, T. E. NDTB-1: A supertetrahedral cationic framework that removes TcO4 from solution. Angew. Chem., Int. Ed. 2010, 49, 1057–1060. (37) Yu, P.; Wang, S.; Alekseev, E. V.; Depmeier, W.; AlbrechtSchmitt, T. E.; Phillips, B.; Casey, W. Technetium-99 MAS NMR spectroscopy of a cationic framework material that traps TcO4 ions. Angew. Chem., Int. Ed. 2010, 49, 5975–5977. (38) Wang, S.; Alekseev, E. V.; Ling, J.; Liu, G.; Depmeier, W.; Albrecht-Schmitt, T. E. Polarity and chirality in uranyl borates: Insights into understanding the vitrification of nuclear waste and the development of nonlinear optical materials. Chem. Mater. 2010, 22, 2155–2163. (39) Wang, S.; Alekseev, E. V.; Stritzinger, J. T.; Depmeier, W.; Albrecht-Schmitt, T. E. How are centrosymmetric and noncentrosymmetric structures achieved in uranyl borates? Inorg Chem. 2010, 49, 2948–2953. (40) Wang, S.; Alekseev, E. V.; Stritzinger, J. T.; Depmeier, W.; Albrecht-Schmitt, T. E. Crystal chemistry of the potassium and rubidium uranyl borate families derived from boric acid fluxes. Inorg. Chem. 2010, 49, 6690–6696. (41) Wang, S.; Alekseev, E. V.; Stritzinger, J. T.; Liu, G.; Depmeier, W.; Albrecht-Schmitt, T. E. Structure-property relationships in lithium, silver, and cesium uranyl borates. Chem. Mater. 2010, 22, 5983–5991. (42) Wang, S.; Alekseev, E. V.; Ling, J.; Skanthakumar, S.; SoderholmL.; Depmeier, W.; Albrecht-Schmitt, T. E. Neptunium diverges sharply from uranium and plutonium in crystalline borate matrixes: insights into the complex behavior of the early actinides relevant to nuclear waste storage. Angew. Chem., Int. Ed. 2010, 49, 1263–1266. (43) Wang, S.; Alekseev, E. V.; Depmeier, W.; Albrecht-Schmitt, T. E. Further insights into intermediate- and mixed-valency in neptunium oxoanion compounds: Structure and absorption spectroscopy of K2[(NpO2)3B10O16(OH)2(NO3)2]. Chem. Commun. 2010, 46, 3955–3957. (44) Wang, S.; Alekseev, E. V.; Miller, H. M.; Depmeier, W.; Albrecht-Schmitt, T. E. Boronic acid flux synthesis and crystal growth of uranium and neptunium boronates and borates: A low-temperature route to the first neptunium(V) borate. Inorg. Chem. 2010, 49, 9755–9757. (45) Wang, S.; Villa, E. M.; Diwu, J.; Alekseev, E. V.; Depmeier, W.; Albrecht-Schmitt, T. E. Role of anions and reaction conditions in the preparation of uranium(VI), neptunium(VI), and plutonium(VI) borates. Inorg. Chem. 2011, 50, 2527–2533. (46) Wang, S.; Alekseev, E. V.; Depmeier, W.; Albrecht-Schmitt, T. E. Surprising coordination for plutonium in the first plutonium(III) borate. Inorg. Chem. 2011, 50, 2079–2081. (47) Wang, S.; Alekseev, E. V.; Diwu, J.; Miller, H. M.; Oliver, A.; Liu, G.; Depmeier, W.; Albrecht-Schmitt, T. E. Functionalization of borate networks by the incorporation of fluoride: syntheses, crystal structures, and nonlinear optical properties of novel actinide fluoroborates. Chem. Mater. 2011, 23, 2931–2939. (48) Snider, A. C. Verification of the Definition of Generic Weep Brine and the Development of a Recipe for This Brine; ERMS 527505, 2003. (49) Borkowski, M.; Richmann, M.; Reed, D. T.; Xiong, Y. Complexation of Nd(III) with tetraborate ion and its effect on actinide(III) solubility in WIPP brine. Radiochim. Acta 2010, 98, 577–582. (50) Chu, S. N. G.; Logan, R. A.; Geva, M.; Ha, N. T.; Karlicek, R. F. Substitutional, interstitial, and neutral zinc incorporation into InP grown by atmospheric pressure metalorganic vapor phase epitaxy. J. Appl. Phys. 1996, 80, 3221–3227. (51) Lehmann, G. Interstitial incorporation of di- and trivalent cobalt in quartz. J. Phys. Chem. Solids 1969, 30, 395–399. (52) Wu, S.; Wang, S.; Simonetti, A.; Chen, F.; Albrecht-Schmitt, T. E. Incorporation of iodate into uranyl borates and its implication for the immobilization of 129I in nuclear waste repositories. Radiochim. Acta 2011, 99, 1–7.
ARTICLE
(53) Bykov, D, M.; Orlova, A. I.; Tomilin, S. V.; Lizin, A. A.; Lukinykh, A. N. Americium and plutonium in trigonal phosphates (NZP type) Am1/3[Zr2(PO4)3] and Pu1/4[Zr2(PO4)3]. Radiochemsitry 2006, 48, 234–239. (54) Sheldrick, G. M. SADABS 2001, program for absorption correction using SMART CCD based on the method of Blessing: Blessing, R. H. Acta Crystallogr. 1995, A51, 33-38. (55) Simonetti, A.; Heaman, L. M.; Chacko, T.; Banerjee, N. R. Insitu petrographic thin section U-Pb dating of zircon, monazite, and titanite using laser ablation-MC-ICP-MS. Int. J. Mass Spectrom. 2006, 253, 87–97. (56) Simonetti, A.; Heaman, L. M.; Chacko, T. Use of discretedynode secondary electron multipliers with Faradays—A ‘reduced volume’ approach for in-situ U-Pb dating of accessory minerals within petrographic thin section by LA-MC-ICP-MS. In 2008 V.M. Goldschmidt Laser Ablation Short Course 40; Sylvester, P., Ed.; Mineralogical Association of Canada: Vancouver, BC, Canada, 2008; pp 241 264. (57) Booth, C. H.; Bridges, F. Improved self-absorption correction for fluorescence measurements of extended X-ray absorption finestructure. Phys. Scr. 2005, T115, 202–204. (58) Hayes, T. M.; Boyce, J. B Extended X-ray absorption fine structure. Solid State Physics; Academic: New York, 1982; Vol. 37, pp 173 351. (59) Li, G. G.; Bridges, F.; Booth, C. H. X-ray-absorption finestructure standards: a comparison of experiment and theory. Phys. Rev. B 1995, 52, 6332–6348. (60) Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Real space multiple scattering calculation of XANES. Phys. Rev. B 1998, 58, 7565–7576. (61) Zhang, H.-X.; Zhang, J.; Zheng, S.-T.; Yang, G.-Y. Two new potassium borates, K4B10O15(OH)4 with stepped chain and KB5O7(OH)2 3 H2O with double helical chain. Cryst. Growth Des. 2005, 5, 157–161. (62) Merlino, S.; Sartori, F. The crystal structure of lardellite, NH4B5O7(OH)2 3 H2O. Acta Crystallogr. 1969, B25, 2264–2270. (63) Shannon, R. D. Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallogr. 1976, A32, 751–767. (64) Lee, M. H.; Park, Y. J.; Kim, W. H. Absorption spectroscopic properties for Pu(III, IV and VI) in nitric and hydrochloric acid media. J. Radioanal. Nucl. Chem. 2007, 273, 375–382. (65) Liu, G.; Beitz, J. V. Spectra and electronic structures of free actinide atoms. In The Chemistry of the Actinide and Transactinide Elements; Morss, L. R., Edelstein, N. M., Fuger, J., Eds.; Springer: Dordrecht, The Netherlands, 2006; Vol. 4, Chapter 16, pp 2013 2111. (66) Clark, D. L.; Hecker, S. S.; Jarvinen, G. D.; Neu, M. P. Plutonium. In The Chemistry of the Actinide and Transactinide Elements; Morss, L. R., Edelstein, N. M., Fuger, J., Eds.; Springer: Dordrecht, The Netherlands, 2006; Vol. 2, Chapter7, pp 813 1264. (67) Rothe, J.; Walther, C.; Denecke, M. A.; Fanghanel, T. XAFS and LIBD investigation of the formation and structure of colloidal Pu(IV) hydrolysis products. Inorg. Chem. 2004, 43, 4708–4718. (68) Stern, E. A.; Livins, P.; Zhang, Z. Thermal vibration and melting from a local perspective. Phys. Rev. B 1991, 43, 8850–8860. (69) Hu, Y.-J.; Booth, C. H. Predetermining acceptable noise limits of EXAFS spectra in the limit of stochastic noise. J. Phys. Conf. Ser. 2009, 190, 012028. (70) Burns, P. C.; Grice, J. D.; Hawthorne, F. C. Borate minerals: I. Polyhedral clusters and fundamental building blocks. Can. Mineral. 1995, 33, 1131–1151. (71) Grice, J. D.; Burns, P. C.; Hawthorne, F. C. Borate minerals: II. A hierarchy of structures based upon the borate fundamental building block. Can. Mineral. 1999, 37, 731–762.
9463
dx.doi.org/10.1021/es2028247 |Environ. Sci. Technol. 2011, 45, 9457–9463
CORRESPONDENCE/REBUTTAL pubs.acs.org/est
Comment on “High Levels of Bisphenol A in Paper Currencies from Several Countries, and Implications for Dermal Exposure”
I
am writing regarding the paper by Liao and Kannan (“High Levels of Bisphenol A in Paper Currencies from Several Countries, and Implications for Dermal Exposure”).1 The paper reports that the estimated daily intake (EDI) values calculated for bisphenol A (BPA) from paper currencies were “several orders of magnitude” lower than the oral reference dose of 50 μg/kg bw/ day established by the United States Environmental Protection Agency and the European Food Safety Authority. The highest occupational (worst case) exposure was for paper currency from Brazil (their Table 2) where the mean EDI value was 21 ng/day. For a standard 60 kg body weight person, the worst case EDI is in fact over 140 thousand fold lower than the oral reference dose. This comparison does not take into account the lower rates of uptake of BPA from dermal versus oral exposures. See for instance ref 2. The authors conclude their paper by citing the 2010 Joint Food and Agriculture Organization (FAO)/World Health Organization (WHO) Expert Meeting on BPA3 as the source of studies reporting “adverse endocrine disruptive effects of BPA at doses as low as a few tens to a few hundreds of ng/kg bw/day,” concentrations which still exceed human exposure to BPA from paper currency by factors of 30 or more even in the worst case (Assuming 10 ng/kg bw/d as lowest effective dose and standard 60 kg bw person. Worst case EPI = 21 ng/day = 0.35 ng/kg bw/d. MOE = 10/0.35 = 28.6). The FAO/WHO Report itself does not include these studies among the low dose studies it considered as suitable for hazard characterization of BPA. Even among those studies, the report concludes that “there is considerable uncertainty regarding the validity and relevance of these observations” and “it would be premature to conclude that these evaluations provide a realistic estimate of the human health risk, given the uncertainties.” In short, the FAO/WHO Report does not support the authors’ claims of health effects at low doses. The FAO/WHO Report indicates that considerably more research will be required before studies claiming low dose effects of BPA can be accepted as valid for human health risk. This continues to be the view of regulatory authorities and independent research institutes worldwide, including most recently the Research Institute of Science for Safety and Sustainability of Japan.4 In the meantime, the current study demonstrates that even with the worst case BPA exposure level - occupational exposure to the currency with the highest level of BPA - exposure was 140-thousand fold lower than the oral reference dose, which in turn is based on multiple safety factors for lifetime exposure. The large margins of exposure (MOEs) that can be calculated between the oral reference dose and the BPA exposure level from currency in this new study firmly contradict claims of health risks from BPA in paper currency, as asserted by various activist groups.
’ AUTHOR INFORMATION Corresponding Author
*E-mail: [email protected].
’ REFERENCES (1) Liao, C.; Kannan, K. High levels of bisphenol A in paper currencies for several countries, and implications for dermal exposures. Environ. Sci. Technol. 2011, 45, 6761–6768. (2) Biedermann, S; Tschudin, P.; Grob, K. Transfer of bisphenol A from thermal printer paper to the skin. Anal. Bioanal. Chem. 2010, 398 (1), 571–576. (3) FAO (Food and Agriculture Organization)/WHO (World Health Organization). Joint FAO/WHO Expert Meeting to Review Toxicological and Health Aspects of Bisphenol A. 2010. http://www. who.int/foodsafety/chem/chemicals/BPA_Summary2010.pdf (accessed August 16, 2011). (4) Kawasaki, H.; Kazaoui, R. Updated hazard assessment of bisphenol A. In The Research Institute of Science for Safety and Sustainability (RISS); National Institute of Advanced Industrial Science and Technology (AIST): Japan July 2011; Available from: http://www.aist-riss.jp/ main/modules/product/inde.phd?content_id=73&ml_lang=en.
John E. Heinze* Environmental Health Research Foundation r 2011 American Chemical Society
Published: September 28, 2011 9464
dx.doi.org/10.1021/es203169y | Environ. Sci. Technol. 2011, 45, 9464–9464
CORRESPONDENCE/REBUTTAL pubs.acs.org/est
Reply to Comment on “High Levels of Bisphenol A in Paper Currencies from Several Countries, and Implications for Dermal Exposure”
W
e appreciate the opportunity to respond to the comment of John E. Heinze1 on our recently published article2 that reported ubiquitous occurrence of high levels of bisphenol A (BPA) in paper currencies and concomitant dermal exposure through handling of currencies. The major intent of our publication2 was to report the widespread occurrence of BPA in paper currencies at concentrations on the order of several micrograms per gram (“partsper-million” level); this is important because the sources of human exposure to BPA are still not well characterized. The 2010 Joint Food and Agriculture Organization (FAO) and World Health Organization (WHO) expert meeting 3 identified that human exposure to BPA from nonfood sources is not wellknown. Evaluation of the sources of human exposure is essential if we are to develop strategies to mitigate exposures. In our study, we found that the mean concentrations of BPA in U.S. dollar bills (as an example) were several fold higher than the concentrations reported for house dust4 and foodstuffs (including canned foods; as shown in ref 3) from the U.S. This is a significant finding because prior to this study it was not known whether paper currencies contained BPA at concentrations greater than those found in house dust and foodstuffs. Heinze claims that our citation of the 2010 FAO/WHO report3 does not support our statement on health effects of BPA at low exposure doses. One of the major conclusions of the FAO/WHO expert meeting (see ref 3) on BPA was “some emerging new end-points (sex-specific neurodevelopment, anxiety, preneoplastic changes in mammary glands and prostate in rats, impaired sperm parameters) in a few studies show associations at lower levels”. Our intent was to convey the information that there are studies reporting adverse effects of BPA at low exposure doses (tens to hundreds of nanograms per kilogram body weight per day), and therefore the exposure levels measured from paper currencies should not be ignored. In fact, there are studies that show effects of BPA at doses below the current reference dose of 50 μg/kg bw/day (for reviews see refs 5 7). Nevertheless, considerable controversies surround the issue of low dose exposures and the reference dose. We believe that exposure to BPA from currencies can augment the total daily exposures; several such sources of BPA exposures still remain to be characterized. Heinze calculated the margin of exposure (MOE) by comparing the reference dose and the dermal intake from currencies, but ignored to indicate that the intake from currencies only represents “dermal pathway” and does not take into account of “inhalation pathway”. Because BPA is available as a powdery film on the surface of paper currencies, inhalation exposure can be significant. As indicated in our paper,2 our exposure assessment is limited to dermal uptake and does not include inhalation from handling of currencies (we did not analyze BPA released into air while handling paper currencies and therefore we did not report inhalation exposures). Furthermore, our exposure assessment involved several assumptions and uncertainties (which is intrinsic in all exposure and risk assessments). Overall, our r 2011 American Chemical Society
exposure assessment is an underestimate of the total exposure from currencies. The MOE of 140 thousand estimated by Heinze is an overestimate and his statement on health risk of BPA from currencies is premature and inconclusive. The 2010 FAO/WHO report concluded as “establishing a safe exposure level for BPA continues to be hampered by a lack of data from experimental animal studies that are suitable for risk assessment”. Furthermore, (contrary to what Heinze had indicated on his comment) we included a dermal permeation factor of 0.27 in our exposure assessment (absorption fraction: 27%, as we described clearly in our paper) and our exposure doses depict the lower dermal uptake rates, because only a fraction of BPA deposited on skin enters the blood circulation. Human exposure to BPA is a contentious public health issue and considerable controversies surround the issue of reference dose and low dose toxicity.5 7 The 2010 FAO/WHO report concluded that “...given the uncertainties, these findings should drive the direction of future research with the objective of reducing this uncertainty”. Our study provides valuable data for reducing one of the uncertainties in BPA exposure assessments. Chunyang Liao and Kurunthachalam Kannan* Wadsworth Center, New York State Department of Health, and Department of Environmental Health Sciences, School of Public Health, State University of New York at Albany, Empire State Plaza, P.O. Box 509, Albany, New York 12201-0509, United States
’ AUTHOR INFORMATION Corresponding Author
*Phone: 1-518-474-0015; fax: 1-518-473-2895; e-mail: kkannan@ wadsworth.org.
’ REFERENCES (1) Heinze, J. Comment on “high levels of bisphenol A in paper currencies from several countries, and implications for dermal exposure”. Environ. Sci. Technol. 2011, 45, DOI: dx.doi.org/10.1021/es203169y. (2) Liao, C.; Kannan, K. High levels of bisphenol A in paper currencies for several countries, and implications for dermal exposures. Environ. Sci. Technol. 2011, 45, 6761–6768. (3) FAO/WHO. Joint FAO/WHO Expert Meeting to Review Toxicological and Health Aspects of Bisphenol A. 2010. http://www. who.int/foodsafety/chem/chemicals/BPA_Summary2010.pdf (accessed September 20, 2011). (4) Loganathan, S. N.; Kannan, K. Occurrence of bisphenol A in indoor dust from two locations in the eastern United States and implications for human exposures. Arch. Environ. Contam. Toxicol. 2011, 61, 68–73. (5) Vandenberg, L. N.; Maffini, M. V.; Sonnenschein, C.; Rubin, B. S.; Soto, A. M. Bisphenol A and the great divide: A review of controversies in the field of endocrine disruption. Endocr. Rev. 2009, 30, 75–95. Published: September 28, 2011 9465
dx.doi.org/10.1021/es203380e | Environ. Sci. Technol. 2011, 45, 9465–9466
Environmental Science & Technology
CORRESPONDENCE/REBUTTAL
(6) vom Saal, F. S.; Hughes, C. An extensive new literature concerning low-dose effects of bisphenol A shows the need for a new risk assessment. Environ. Health Perspect. 2005, 113, 926–933. (7) Goodman, J. E.; Witorsch, R. J.; McConnell, E. E.; Sipes, I. G.; Slayton, T. M.; Yu, C. J.; Franz, A. M.; Rhomberg, L. R. Weight-ofevidence evaluation of reproductive and developmental effects of low doses of bisphenol A. Crit. Rev. Toxicol. 2009, 39, 1–75.
9466
dx.doi.org/10.1021/es203380e |Environ. Sci. Technol. 2011, 45, 9465–9466
ADDITION/CORRECTION pubs.acs.org/est
Correction to “Historically and Currently Used Dechloranes in the Sediments of the Great Lakes” An Li Environmental and Occupational Health Sciences, School of Public Health, University of Illinois at Chicago, 2121 West Taylor Sreet, Room 304, Chicago, Illinois 60612, United States Due to a mistake in unit conversion, the numbers for latitude and longitude in Supporting Information Table S2 are inaccurate. The correct Supporting Information is posted here and with the original article.
’ ASSOCIATED CONTENT
bS
Supporting Information. This material is available free of charge via the Internet at http://pubs.acs.org.
Published: October 04, 2011 9467
dx.doi.org/10.1021/es203210v | Environ. Sci. Technol. 2011, 45, 9467–9467
