Chemical Physics of Nanostructured Semiconductors
Editors: Alexander I. Kokorin and Detlef W. Bahnemann
/// vsP/// Utrecht
Boston, 2003
VSP BV P.O. Box 346 3700 AH Zeist The Netherlands
Tel: +31 30 692 5790 Fax: +3 1 30 693 208 1
[email protected] www.vsppub.com
0VSP BV 2003 First published in 2003 ISBN 90-6764-382-3
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner.
Printed in The Netherlands by Ridderprint bv, Ridderkerk.
Chemical Physics ofNanosfructured Semiconductors, pp. xi-xv A.1. Kokorin and D.W. Bahnemann (Eds.) 0 VSP 2003.
About the Authors Alonso-Vante, Nicolils: was born in Oaxaca, Mexico and educated in Mexico-city. He received his Dr. Xs Sc. (1984) from the Louis Pasteur University of Strasbourg, France. He then joined the research group of Prof. H. Tributsch at the Hahn-Meitner-Institut in Berlin, Germany, as an Alexander von Humboldt fellow. Thereafter, he continued as a senior scientist. He is presently Professor of Chemistry at the Chemistry faculty of the University of Poitiers, France. His research interest has concentrated, these late years, on material science research for (photo)electrocata-lysis, and the in situ investigation of interfacial processes combining spectroscopy with electrochemical techniques. E-mail:
Bahnemann, Detlef W.: studied chemistry at the TU Berlin, Germany, where he received his Ph.D. in 1981. From 1981 to 1988 he worked as a Senior Scientist at the Hahn-Meitner-Institute (HMI) Berlin with Prof. Arnim Henglein. He joined the group of Prof. M. R. Hoffmann at the California Institute of Technology in Pasadena, USA as a Visiting Associate (1985-1987). In 1988-2002 he was a Department Head at the Institute for Solar Energy Research (ISFH) in Hannover, FRG. Since June 2002 he became an Academic Director at the Institute for Technical Chemistry of the Hannover University where he is responsible for the research field of Photochemistry and Nanotechnology. Prof. Dr. Bahnemann is Honorary Visiting Prof. at the Robert-Gordon Univ. in Aberdeen (UK), Lecturer for Physical Chemistry at the Carl-vonOssietzky University in Oldenburg, FRG. His research interests include Free Radical Chemistry, Fast Reaction Kinetics, Photocatalysis and Inorganic Nanomaterials. E-mail: Bavykin, Dmitry V.: is a Ph.D. researcher in the Laboratory of photocatalysis on semiconductors at the Boreskov Institute of Catalysis, Novosibirsk, Russia. The title of his PhD thesis (1998): “Luminescent and photocatalytic properties of CdS nanocolloids”. Area of his interests is the photophysical-photochemical properties of nanosized sulfide semiconductors, including synthesis of particles with definite size and surface properties, their characterisation; the study of the photoexcited states dynamics, relaxation in quantum dots by the luminescence and flash photolysis measurements; studies of the interfacial charge transfer from colloidal semiconductor particles by the steady state photolysis, luminescence quenching method. E-mail: Dillert, Ralf: studied chemistry at the Technical University Braunschweig and received his Dr. rer. nat. in 1988. He worked as a scientist at the Gesellschaft fur Biotechnologische Forschung mbH (GBF) at the Institute of Physical Chemistry of the TU Braunschweig, and at the Institut fur Solarenergieforschung GmbH (ISFH) Hannover, FRG. He was a lecturer for wastewater treatment at the University of Applied Science in Flensburg. In 1986 he founded EcoTRANSfair Gesellschaft fur Umwelt und Gesundheit mbH, an environmental service
xii
About the authors
company, and is actually its managing director. His research interests are chemical technologies in water and wastewater treatment and especially photocatalysis. E-mail: Kokorin, Alexander I.: was born in 1947. Was graduated as a biophysicist in 1970; Ph.D. (Candidate of Sciences) in 1974; D.Sc. degree (Doctor of Sciences) in physical chemistry - in 1992. At present: Principal Researcher and Deputy Head of the Division of Kinetics and Catalysis, N. Semenov Institute of Chemical Physics of Russian Academy of Sciences, Moscow, Russia. Area of research interests: chemical methods of solar energy conversion; chemical physics of organized molecular systems, including nanosized oxide semiconductors doped with transition metal ions, and polymer-metal complexes; the study of their structure, absorptive, catalytic, photocatalytic and photoelectrochemical properties. EPR spectroscopy and spin-spin interaction between paramagnetics. He is the author and co-author of more than 170 publications, including two books and several reviews and book chapters. E-mail: Lindgren, Torbjorn: was born in 1972, received a M.Sc. (chemical engineering) in 1999 at Chalmers University of Technology, Gtiteborg, Sweden. He continued education in Uppsala University, Sweden as a Ph.D. student with Sten-Eric Lindquist and Anders Hagfeldt as supervisors. His Ph.D. project is concerned with direct solar induced water splitting on semiconductors. E-mail: Lindquist, Sten-Eric: Professor of the Department of Physical Chemistry, Uppsala University (UU), Sweden. In 2001 was retired. As professor emeritus he is currently active doing research in the fields of electrochemistry and photoelelctrochemistry in UU. His main research field is solar energy, in particular - photoelectrochemical (PEC) cells for electrical power or hydrogen generation, and also in the field of redoxpolymer wired enzyme based amperometric biosensors. He has been the leader of a number of national research programs and the Swedish coordinator for several European JOULE programs including a research program on dye-sensitized solar
About the authors
...
Xlll
cells within Angstrom Solar Center in Uppsala. He is at present working within the framework of the Swedish research program “Sustainable Energy Conversion with Hydrogen as a Storage Agent”, studying PEC systems with the aim of producing H2 from water with solar light. He is also engage in studies of photocatalysis at gas/semiconductor interfaces at nanostructured thin films. E-mail: <sten-eric .Lindquist @ fki.uu.se>
Martyanov, Igor N.: is a Ph.D. researcher and recently worked in the laboratory of photocatalysis on semiconductors at the Boreskov Institute of Catalysis, SB U S , Novosibirsk, Russia. The title of his PhD Thesis (1998) was: “Kinetics of photocatalytic redox reactions of organic molecules in semiconductor suspensions (CdS and TiO2)”. Areas of his interests: kinetics of photocatalytic reactions in liquid phase at deep conversion; the influence of the surfactants. Parmon, Valentin N.: Professor, the Academician of U S (from 1997), Director of the Boreskov Institute of Catalysis (BIC), Siberian Branch of Russian Academy of Sciences, Novosibirsk, Russia. He was born in 1948 and graduated from the Moscow Physical-Technical Institute in 1972 as a chemical physicist. He received Ph.D. in 1975 at NSemenov Institute of Chemical Physics (Moscow). In 1977 he organized the Laboratory of Catalytic Methods of Solar Energy Conversion in BIC. He is the author or co-author of more than 500 publications, including 5 monographs in chemical kinetics and catalysis, photochemistry and radiation chemistry, chemical radiospectroscopy and physical chemistry of energy production, as well as renewable energetics and transfer of new technologies to industry. He is a Chairman of the Russian Scientific Council on Catalysis. E-mail: <[email protected]> Poluektov, Oleg G.: he started his research career in the NSemenov Institute of Chemical Physics RAS, Moscow, in 1983, after receiving a Ph.D. in physics from the Moscow Institute of Physics and Technology. In 1991-1996 he was a visiting scientist in the Physical Department of the University of Leiden, The Netherlands, and in 1997-1998 a visiting scientist in the Chemistry Division at Argonne National Laboratory (CD ANL),USA. In 2000 he joined the CD ANL as a research scientist. His specialization is magnetic resonance and especially very high frequency EPR spectroscopy, where he contributed significantly in the development and application of new technique. His research interests have covered a broad range of physics and chemistry: the molecular dynamics and structure of glass, crystal, liquid crystal, polymer and biopolymer systems; the mechanisms and kinetics of chemical reactions in liquid phase. His current research is focused on the relation of function, structure, and dynamics in photosynthetic reaction center proteins. He is the author of more than 70 scientific articles and book chapters. E-mail: <[email protected]>
xiv
About the authors
Rajh, Tijana: is a physical chemist in the Chemistry Division of Argonne National Laboratory, USA. She received her Ph.D. from Belgrade University while working at the Vinca Research Institute in Belgrade on the synthesis and characterization of 11-VI semiconductor quantum dots. In 1984-85, Dr. Rajh was a visiting scientist at the Solar Energy Research Institute (now known as NREL), Golden, Colorado, USA. In 1987 she worked at the Hebrew University in Jerusalem as a visiting scientist on the investigation of quantization effects in the initial stages of semiconductor particle growth. Currently, she studies the encapsulation of metal oxide semiconductor nanoparticles with organic ligands which result in strong electronic coupling, and allowing the electronic linking of the nanoparticle into molecular circuits. These systems can be applied for remediation of various organic and inorganic pollutants. She is the author of more than 45 scientific articles and book chapters. E-mail: Robertson, Peter K. J.: B.Sc., D.Phil., Eng. On completing his D.Phi1. in Chemistry at the University of Ulster (1989) he joined the Faraday Centre in Carlow, Ireland, where he was involved in electrochemical research on bulk electrolysis processes and environmental treatment systems. In 1991, he moved to the Industrial Research and Technology Unit in Northern Ireland as a Higher Scientific Officer in the Material ScienceDnorganic Chemistry Section, where he managed a range of research projects on photocatalytic and electrochemical waste treatment for industry; he was involved in environmental assessments of contaminated land sites. He joined the School of Applied Sciences at R. Gordon University in 1995 where he was involved in research in environmental science and technology and photocatalysis. From 2000, he is a Professor at the Chair of Energy and Environmental Engineering in the School of Engnieering at RGU. His research interests focus on advanced oxidation technologies for water treatment, specifically by-products generated by the offshore oil and gas industry and toxic compounds in drinking water, and also in the development of sensor technologies for marine and fresh water environments. E-mail:
About the authors
xv
Thurnauer, Marion C.: is a Senior Scientist and Division Director of the Chemistry Division of Argonne National Laboratory, USA. She received her B.A., M.S., and Ph.D. (1974) degrees from the University of Chicago. She has authored over 100 publications including journal articles and book chapters, primarily on the subject of solar energy conversion in natural and artificial photosynthesis. Her research interests include the photochemical energy conversion and photocatalysis in natural and artificial photosynthetic systems; the development and application of time resolved magnetic resonance techniques to study photochemical charge separation; and electron spin polarization as observed in photosynthetic and model systems. Dr. Thurnauer was honored with many American and International Awards. Her recent honor was receiving the 2002 F. P. Garvan - J. M. Olin Medal Award from the National American Chemical Society. E-mail: <[email protected]> Vayssieres, Lionel: born in 1968, received a B.Sc. in physical chemistry in 1989 and a Ph.D. in chemistry in 1995 on the thermodynamic control of metal oxide nanoparticles at the University Pierre et Marie Curie in Paris, France with Jacques Livage. He joined Uppsala University, Sweden as a researcher for 5 years at the department of Physical-Chemistry with Sten-Eric Lindquist and, with Hans Siegbahn and Joseph Nordgren at the Physics department, developing novel metal oxide nanomaterials for photovoltaic devices and characterizing their electronic structure by x-ray spectroscopies. He worked as a visiting researcher with Adam Heller at the chemical engineering department of the University of Texas at Austin, USA and at the department of Biochemistry at Stellenbosch University, South Africa with Pieter Swart on bionanocomposite materials. Recently he joined the Texas Materials Institute at the University of Texas at Austin as a visiting scientist. He is the author of 20 scientific articles, and performed many invited lectures and seminars. Wang, Heli: he has worked at the National Renewable Energy Laboratory (NREL) in the United States. He received a Ph.D. in corrosion science and materials chemistry from the Helsinki University of Technology in Finland. From 1998, he had worked with nanostructured semiconductors of metal oxides at the Department of Physical Chemistry, Uppsala University, Sweden. His research work has been in materials, electrochemistry, photoelectrochemistry, as well as fuel cell components. E-mail:
Chemical Physics of Nanostructured Semiconductors, pp. ix-x A.I. Kokorin and D.W. Bahnemann (Eds.) 0 VSP 2003.
Preface The idea to write this book arose during the participation in several Workshops on Quantum Solar Energy Conversion, organized in the Alps by the European Society for Quantum Solar Energy Conversion (ESQSEC). Deep and detailed discussions on chemistry, chemical physics, photoelectrochemistry, photophysics, photocatalysis and possible applications of nanostructured semiconductor materials have shown the increasing interest in the matter by scientists representing various research areas as well as industrial enterprises. Indeed, solar energy conversion and chemical methods for its realization became very popular again after the “great jump” of renewable energy sources between the middle of the 1970s and the beginning of the 1980s. Several excellent books have been published since these years; for example, Energy Resources through Photochemistry and Catalysis, by M. Gratzel (1983), Photocatalysis: Fundamentals and Applications, by N. Serpone and E. Pelizzetti (1989), Photoelectrochemical Conversion of Solar Energy, by Yu. V. Pleskov (1990), Photochemical Conversion and Storage of Solar Energy, by E. Pelizzetti and M. Schiavello (199 1) and Photocatalytic Purification and Treatment of Water and Air, by D. F. Ollis and H. Al-Ekabi (1993). Nevertheless, in these books no attempt was made to approach this research area from the point of view of classical chemical physics. When we decided to edit this book we aimed for three main goals: a) to generate an adequate scope of the modem trends and data obtained during the last years in the area of chemical physics of nanostructured materials, in particular, nanocrystalline semiconductors; b) to select an equal mix of scientists from Western and Eastern countries, all of them experts in their respective research areas, and c) to present to the international scientific community many interesting and important results which have been obtained by former Soviet Union researchers, but are not well known because they had originally been published in Russian books and journals. Many different experimental methodologies, each with special utility, are currently being used for the determination of structure, of physicochemical properties and of functionality in the area of solar energy conversion. The editors invited the authors to contribute reviews from the field of their own expertise, including specific and typical examples from their own experimental data (mainly). It will be left to the readers to decide whether or not our attempt has been successful.
X
Preface
Acknowledgements for the financial support are given at the end of each chapter. We are also thankful to Prof. Dr. Tjeerd Schaafsma (President of the ESQSEC) and Prof. Dr. Helmut Tributsch, who continuously supported our idea to write this book, and to Mr. Ilja A. Kokorin for technical assistance in the preparation of the book. We hope this book will be interesting and useful for scientists working in the area of semiconductor nanotechnology, photoelectrochemistry, photocatalysis, photochemistry of water and air purification, as well as for graduate and post-graduate students who are planning to join these research areas.
Alexander I. Kokorin Moscow, Russia
Detlef W. Bahnemann Hannover, Germany
Contents Preface
ix
About the authors
xi
Chapter 1. Charge Separation in Titanium Oxide Nanocrystalline Semiconductors Revialed by Magnetic Resonance (T. Rajh, 0. G. Poluektov and M. C. Thurnauer) 1.1. Introduction 1.2. Basic Mechanisms of Semiconductor-AssistedPhotocatalysis 1.2.1. Energy Band Structure of Ti02 Nanoparticles 1.2.2.Photogeneration of Charge Pairs and Intrinsic Properties of Semiconductors 1.2.3. Space Charge Layer and Band Banding 1.3. Charge Separation and Carrier Trapping in Ti02 Nanoparticles 1.3.1. Nature of Trapping Sites 1.3.2. Hole Trapping Sites 1.3.3. Hole Trapping Sites 1.3.4. Charge Separation in the Surface-ModifiedTi02 Colloids 1.4. Kinetic and Mechanism of the Charge Separation in Ti02 Colloid Nanoparticles as Studied by Time-Resolved EPR Technique 1.5. Summary References
23 30 31
Chapter 2. Kinetic Peculiarities of Photocatalytic Reactions on CdS Nanocolloids (D. V. Bavykin, E. N. Savinov, I. N. Martyanov and V. N. Parmon)
35
2.1. Introduction 2.2. Synthesis of CdS Colloids 2.2.1. Thermodynamic Calculation of Equilibrium Size of CdS Colloidal Particles in the Presence of Cadmium Components 2.2.2. Method of CdS Preparation 2.3. Reaction of Interfacial Electron Transfer as the First Stage of Redox Photocatalytic Processes 2.3.1. Photobleaching Relaxation of Colloids 2.3.2. Physical Causes Resulting in the Observable Kinetic Peculiarities of Photobleaching Relaxation 2.3.3. Effect of CdS Adsorption Properties on the Kinetics of Its Photobleaching Relaxation in the Presence of Various Electron Acceptors 2.4. Luminescence Quenching of CdS Colloids 2.4.1. Regularities of Luminescence Quenching of Colloidal CdS Particles 2.4.2. Effect of Surface Properties of Ultradispersed CdS on Regularities of Luminescence Quenching 2.4.2.a. Luminescence Quenching of Colloidal CdS by Quenchers of Various Nature
4 5 6 6 7 12 14
35 35 36 39 40 40 45 48 51 52
59 59
vi
Contents
2.4.2.b. Effect of the Surface Charge of Colloidal CdS on the Efficiency of its Luminescence Quenching by Various Quenchers 2.4.3. Effect of Excitation Wavelength on the Luminescence Spectrum of CdS/Cu,S 2.4.4. Conclusion 2.5. Kinetics of Photocatalytic Reactions at Stationary Illumination 2.5.1. Photoreduction of Methyl Orange on CdS Colloids in Deep Conversion 2.5.1.a. Kinetic Peculiarities of Photocatalytic Processes on Ultradispersed CdS Colloids at Stationary Illumination 2.5.1.b. Semiquantitative Description of the Kinetics of Photocatalytic Processes on CdS Colloids in Terms of Adsorption-Desorption Processes in the System 2.5.l.c. Analysis of Kinetic Regularities of the System under Study 2.5.2. Photoreduction of Methylviologen and Phosphotungstic Acid on CdS Colloids; Effect of Surface Charge 2.5 -3.Effect of Excitation Light Wavelength on the Rate of Photocatalytic Reaction 2.6. Conclusion References
Chapter 3. Photooxidation of Water at Hematite Electrodes (T. Lindgren, L. Vayssieres, H. Wang and S.-E. Lindquist) 3.1. Introduction 3.2. General Background of Photooxidation of Water 3.3. Structural and Physical Properties of Hematite 3.3.1. Crystal Structure 3.3.2. Electronic Structure 3.3.3. Electrical Properties 3.3.4. Flat-band Potential 3.4. Photoresponse of Hematite Materials 3.4.1. Colloidal Solutions of Hematite 3.4.2. Hematite Single Crystal Materials 3.4.2.a. Hematite Single Crystal 3.4.2.b. Doped Hematite Single Crystal 3.4.2.c. Mixed Hematite Single Crystal 3.4.3. Polycrystalline Hematite Materials 3.4.3.a. Polycrystalline Hematite 3.4.3.b. Polycrystalline Hematite Thin Film 3.4.3.c. Polycrystalline Doped Hematite 3.4.3.d. Mixed Polycrystalline Oxides of Hematite 3.4.4. Nanostructured Thin Films of Hematite 3.4.4.a. Nanostructured Films of Spherical Particles of Hematite 3.4.4.b. Hematite Nanorods 3.5. Conclusion and Future Scopes References
61 62 64 65 65 66 69 72 77 80 80 80 83 84 85 89 89 90 92 92 95 96 97 97 97 98 98 98 98 100 102 102 102 104 106 107
Contents
Chapter 4. Photoelectrochemistryof Nanocrystalline Aggregates of Cyanine Dyes on the Semiconductor Electrodes (D. V. Sviridov) 4.1. Introduction 4.2. Spectroscopic Properties of Aggregated Cyanine dyes 4.3. Effect of Aggregate Formation upon Electronic Energy Levels of Cyanine Dyes (CDs) 4.4. Aggregates of Cyanine Dyes as the Sensitizing Agents in the Photoelectrochemical Systems 4.5. PhotoelectrochemicalSpectral Sensitization by Partially Aggregated CDs 4.6. Spectral Sensitization of Semiconductor by Mixture of J-aggregated CDs 4.7. PhotoelectrochemicalBehaviour of Thin Films of Aggregated CDs 4.8. Conclusions References
Chapter 5. Physico-Chemical Properties of Novel Nanocrystalline Ruthenium Based Chalcogenide Materials (N. Alonso-Vunte) 5.1. Introduction 5.2. Clusters or Particles 5.2.1 Size Comparison 5.2.2. Metal and Semiconductor Nanoparticles 5.3. A Non-aqueous Chemical Route of Synthesis 5.3.1. Synthesis Ideals 5.3.2. The Chemical Route 5.3.3. Morphology and Stoichiometry 5.4. Electrocatalysis and Photoelectrocatalysis 5.4.1. Thin to Ultra Thin Layers of Chalcogenide Materials 5.4.2. Electrocatalysis via Thin and Ultra Thin Layers of Chalcogenide Materials 5.4.3. (Photo)Electrocatalysison Nanostructured Ti02 Surface Modified via Chalcogenide Materials 5.4.4. Implications of Molecular Oxygen for Photooxidation Process 5.5. Summary References
Chapter 6. Metal Nanoparticles on Semiconductor Surfaces: Electrochemistry and Photocatalysis (A. 1. Kulak) 6.1. Introduction 6.2. Dependence of Nanoparticles Morphology and Behavior in Electron Transfer Processes on the Mean of Metal Nanophase Deposition onto Semiconductor Surface (ScS) 6.2.1. Contact Deposition of Metal Nanoparticles onto ScS 6.2.2. Galvanic Deposition of Metal Nanoparticles onto ScS 6.2.3. Photocatalytic Formation of Metal Nanophase on Semiconductor Surfaces
vii
111 111 111 114 116 120 125 127 132 132
135 135 136 136 138 139 139 140 140 142 142 144 145 147 150 150
153 153 155 155 158 161
viii
Contents
6.3. Formation of Electronic Surface States in Semiconductor Band Gap as a Result of Deposition of Metal Particles on Semiconductor Surface 6.4. Electrocatalytic Activity of Semiconductor Electrodes Modified by Surface-Deposited Metal Nanophase 6.5. Impedometric Investigation of Titania Electrodes Surface-Modified with Metal Particles 6.6. Interaction of Metal Nanoparticles with the Associates of Donor Defects in Wide-Band-Gap n-type Semiconductors 6.7. Conclusion References
Chapter 7. Photocatalysis: Electron Transfer Serving the Environment ( D . W. Bahnemann, R. Dillert and P. K. J. Robertson) 7.1. Introduction 7.2. Primary Processes upon Band Gap Irradiation of Semiconductor Particles 7.3. Chemical Nature of Trapped Charge Carriers 7.4. Fate of Trapped Charge Carriers 7.4.1. Recombination Kinetics 7.4.2. Charge Transfer Kinetics 7.4.2.a. Interfacial Electron Transfer 7.4.2.b. Direct Interfacial Hole Transfer 7.4.2.c. Hole Transfer through the Intermediate Formation of Hydroxyl Radicals 7.5. Conclusions References
Chapter 8. Electron Spin Resonance of Nanostructured Oxide Semiconductors (A. I. Kokorin) List of General Symbols 8.1. Introduction 8.2. EPR Signals of Oxide Semiconductors 8.3. EPR of Small Molecules Adsorbed on the Semiconductor Surface 8.3.1. Oxygen Radicals 8.3.2. N,O, Radicals 8.4. Structural Aspects in the Study of Nanocrystalline Materials 8.4.1, The Measurement of Local Concentration of Paramagnetic Centers 8.4.2. Regularities and Peculiarities of Spatial Distribution of PCs 8.5. Vanadium Ions “Behaviour” i d o n Oxide Semiconductors 8.5.1. Titanium Dioxide 8.5.1.a. Substitutional and Interstitial Centers 8.5.1.b. Surface Doping 8.5.2. Other Oxides 8.6. Other Paramagnetic Dopants 8.7. Identification of Adsorbed Gun+ Ions at the Ti02 Surface and ElectroChemical Behaviour of Copper-Modified Electrodes 8.8. General Conclusions References
166 17 1 174 177 179 180 183 183 184 187 189 189 191 191 192 195 199 200 203 203 204 205 209 209 21 1 215 219 22 1 224 225 227 230 236 23 8 242 252 253
Chemical Physics of Nanostructured Semiconductors, pp, 1-34 A.I. Kokorin and D.W. Bahnemann (Eds.) 0 VSP 2003.
CHAPTER 1
Charge Separation in Titanium Oxide Nanocrystalline Semiconductors Revealed by Magnetic Resonance Tijana Rajh, Oleg G. Poluektov and Marion C. Thurnauer Chemistry Division, Argonne National Laboratory, Argonne IL 60439, USA Keywords: Kinetics, mechanism, charge separation, Titanium dioxide, nanocrystals, photocatalysis, semiconductors, Magnetic resonance
1.1. Introduction Semiconductor photocatalysis using nanoparticlate Ti02 has proven to be a promising technology for use in photocatalytic reactions, in the cleanup of water contaminated with hazardous industrial by-products [ 1-61, or as a photoactive material in nanocrystalline solar cells [7-141. Titanium dioxide, in particular, could be the catalyst of choice for a large variety of applications because it is cheap, non-toxic, and has redox properties which are favorable both for oxidation of many organic pollutants and for reduction of a number of metal ions or organics in aqueous solution. Due to the large oxidation potential of photogenerated holes, oxidative degradation of the most resistant pollutants by photocatalysis can be achieved. To date, the primary focus of Ti02-assisted photocatalysis studies has been on the destruction of organic pollutants in solutions using photocatalytic oxidation. Numerous reports of organic pollutant degradation with mineralization to organic products in several cases have been published [ 15-19]. The common feature of the photocatalytic process always involves reactions via surface trapped photogenerated holes. To maintain neutrality, both photogenerated electrons as well as positive holes have to be consumed efficiently in redox processes on the semiconductor surface. Usually photogenerated electrons were consumed in the reaction with dissolved oxygen in order to achieve fast oxidation of organic compounds. However, organic pollutants are often accompanied by heavy metal ion contaminants that can be reduced by photogenerated electrons into their less toxic, nonsoluble metallic form. TiOz-assisted photoreductive catalysis was found to be useful in the removal of certain heavy metals including mercury, silver, platinum, palladium, rhodium, and gold via their reduction followed by deposition at the catalyst surface [20-221 or photoreduction of nitroaromatic compounds [23-261. The use of photogenerated electrons for deposition of metal layers on
2
Charge Separation in Ti02 Revealed by EPR
TiOz nanocrystalline substrate have new potential applications such as photolithography (nanowriting) - formation of conducting patterns in chips or integrated circuits. In these systems a photochemical reaction is limited to the size of the semiconductor particle. Thus, the metallic conductor size can be obtained in the nanometer size regime [27]. Although TiOz is very effective from an energetic point of view, it is relatively inefficient. The main energy loss in all investigated particulate systems is due to the recombination of charges generated in the illumination of semiconductor particles which is manifest as relatively low efficiency of pollutant decomposition. Therefore, the main focus of future research for the application of semiconductor assisted photocatalysis is to improve separation of charges and at the same time preserve or improve their redox properties. Understanding the events that take place in the early stages of photocatalytic reactions is necessary step for the design and optimization of an efficient artificial photocatalytic system. These primary events involve electron transfer reactions with the creation of paramagnetic species. For this reason Electron Paramagnetic Resonance (EPR) is an important technique for the study of the transient species and examining mechanisms of light-induced sequential electron transfer leading to stabilized charge pair formation [27]. On one hand, EPR spectroscopy allows unique identification of the paramagnetic species involved in the charge separation. On the other hand, interconversion between species and the kinetics of the chargetransfer reaction can be recorded in the wide time range (from ns to days) by monitoring the intensity of the EPR signals, thus providing a detailed description of the charge-separation reaction.
1.2. Basic Mechanisms of Semiconductor-AssistedPhotocatalysis 1.2.1. Energy Band Structure of Ti02 Nanoparticles
Titanium dioxide exists in three crystal modifications, anatase, brookite, and rutile. In bulk crystal rutile is the most thermodynamically stable form. Very few laboratories have been able to synthesize anatase single crystals. However, in the nanocrystalline particle regime the anatase crystal structure is dominant. For this reason, the anatase structure mainly will be discussed below. Anatase crystal modification is a tetragonally distorted octahedral structure in which every titanium atom is surrounded by six oxygen atoms in an elongated octahedron geometry (DZd).Due to the resultant crystal field, the 3d levels of Ti4+that form the conduction band of TiOz are split into 3tZgand 3e, sub-levels. The unequal length of the six cationic ligand bonds to titanium produces a splitting of the tzg and eg orbitals into two subsets (Fig. 1.1). Symmetry considerations show that the p orbitals of the central Ti atom mix with d orbitals of neighboring Ti atoms and 3d-4p hybridized subbands form the conduction bands of anatase TiOz. This is exhibited in the near edge structure of X-ray Absorption Near Edge Structure (XANES) spectrum of T i 0 2 [28]. The lowest band of Ti02 is formed by the bl and e subbands that have mainly x character and can overlap in p-type fashion with oxygen p i orbitals having anionic character. The result is a series of bands in reciprocal space with the maximum of the valence band and
Chapter I ; T. Rajh, et al.
Oh
D2h (rutile)
D2d (IMtasB)
3
u
55
49M1
4965
4970
4975
4980
Fig. 1.1.Energy level diagram of the LUMO of a [TiOs]*-cluster with Oh,Dzh(ruti1e) and Dzd(anatase) symmetry in conjunction with (b) experimental K-edge XANES spectra.
minimum of the conduction band in the center of the Brillouine zone (BZ). However, this direct transition F3-r1is dipole forbidden. (The HOMO is formed as a mixture of oxygen px and pr orbitals that correspond to bl in DU symmetry, and the LUMO is formed from the lowest of 3t, orbitals that correspond to bl in D2d symmetry, and therefore dipole forbidden.) The first allowed transitions are the indirect transitions from the X point at the edge of the BZ to the I? point in the center of BZ. The two lowest transitions are X1-Fl observed with light absorption with an energy of 3.026 eV and perpendicular to the crystalline c axes and X2-rl observed with light absorption with an energy of 3.06 eV and parallel to the crystalline c axes. However, the transition r3- X1 of 3.19 eV is allowed both with parallel and perpendicular light accounting for the increase of the transition probability at this energy corresponding to the band gap energy of anatase TiOz. The curvature of the bands at these extremes is very small indicating large effective masses that are determined by different authors to range from 5m, to 30m, [29, 301. The precise value of the electron effective mass in anatase remains elusive primarily because of difficulty in the preparation of anatase single crystals. Because of the large effective masses, however, the quantization effect in titania particles is negligible for particle sizes > 15 A, and therefore unimportant. The other significant feature of Ti02 particles is that the curvature of the wavefunctions on the top of the valence band is larger than the curvature of the bands in the conduction band. This is different from most semiconductors and indicates that the effective mass of the hole is smaller than the effective mass of the electrons. In the formation of the exciton pairs, a dominant process in small semiconductor particles, it will lead to the orbiting of the photogenerated holes around the heavier photogenerated electrons. This suggests that in small titania particles, photogenerated holes would more likely be trapped at the surface of the particle, while electrons would first find trapping sites in the particle interior.
4
Charge Separation in Ti02 Revealed by EPR
1.2.2. Photogeneration of Charge Pairs and Intrinsic Properties of Semiconductors
Semiconductor particles behave as microelectrochemical cells. Absorption of light energy greater than the band gap of semiconductor materials generates conduction band electrons and valence band holes. Semiconductor particles are light harvesting units offering distinct advantages in the heterogeneous photocatalyzed process: (1) high absorption cross section of the incident photons; (2) fast carrier diffusion; and (3) suitable redox levels of the valence and conduction band edges that can yield high efficiencies in converting light energy to useful redox events. Metal oxides have band structures that are characterized by the existence of an energy gap that separates the highest occupied energy levels (valence band) from the lowest unoccupied energy levels (conduction band). For example, excitation of Ti02 with W light with energy greater than the band gap (> 3.2 eV) promotes electrons from the valence into the conduction band and generates electron-hole pairs that can be exploited in various processes at the particle interface:
where e,<are conduction band (CB) electrons and hvb+ are valence band (VB) holes. Photogenerated carriers migrate to the particle surface and participate in reduction and oxidation processes at the surface. The thermodynamic limit for the reaction that can be carried out with the photogenerated charge carriers is given by the position of the band edges (Le., flat band potential). For example, if a reduction of a particular species (A) in the adjacent solution is to occur, the conduction band of the semiconductor must be more negative than the relevant redox level, while the oxidation of a particular species @) by valence band holes requires that the valence band be more positive than the relevant redox potential (Fig. 1.2). Thus, the relative position of the band edges in a given semiconductor determines their redox functioning.
Fig. 1.2. Salient features of the electronic structures of semiconductors.
Chapter 1; T. Rajh, et al.
5
The threshold energy needed for light absorption (Le., the band-gap energy) and the relative positions of the band edges (i.e., the flat band potential) are intrinsic material properties of each semiconductor. 1.2.3. Space Charge Layer and Band Banding
When a semiconductor is brought into contact with an electrolyte a migration of charge carriers occurs until the Fermi level of the semiconductor is equilibrated with the chemical potential of the electrolyte [31]. This results in the formation of a depleted space charge layer in the surface region of a semiconductor and a Helmholtz double layer in the electrolyte adjacent to the semiconductor surface (Fig. 1.3). The depletion of the surface region creates a barrier for further transfer of electrons to the electrolyte which is manifest by band banding in the space charge layer. Photons absorbed in the depletion layer produce electron hole pairs that separate under the influence of the electric field.
Semiconductor
Bulk SC - n o electroiyte
(b)
Bulk SC - equilibrium condition m dark
Nanocrystalline SC- equilibrium in dark
Fig. 1.3. Space charge layer formation in a bulk and nanocrystalline semiconductor particle with band gap E, and flat band potential of Ufbin equilibrium with a solution redox system for which Fermi level is E+ Vbis the band banding caused by depletion of charge carriers.
The space charge region in the crystalline solid phase is fairly large, approximately lOOOA, while the Helmholtz double region together with the diffuse layer is 100 A [32]. However, when the size of a semiconductor approaches the size of the space charge layer the small particle is depleted almost completely of charge carriers and band banding is negligible. The particles are too small to develop a space charge layer and, in this case, the potential difference resulting from transfer of a charge from a semiconductor to electrolyte has to drop within the Helmholtz layer (neglecting diffuse layer contributions). As a consequence, the position of the band edges of semiconductor particles will shift with the shift of the Fermi level and will not enhance charge separation [33].
6
Charge Separation in Ti02 Revealed by EPR
1.3. Charge Separation and Carrier Trapping in Ti02 Nanoparticles 1.3.1. Nature of Trapping Sites
When the size of the semiconductor nanoparticles is in the nanometer regime, their photoelectrochemical properties may not be the same as those of bulk material. The particle diameter can be smaller than the thickness of the space charge layer and in that case the details of charge separation may not be the same as in a compact semiconductor electrode. When the metal oxide particles are in the nanocrystalline regime, the large fraction of the atoms that constitute the nanoparticle are located at the surface with significantly altered electrochemical properties. In the colloidal solution, due to the weaker covalent bonding of surface atoms with solvent molecules compared to the bonding within the lattice, the energy levels of the surface species are found in the mid-gap region. This decreases the reducing and oxidizing abilities of the surface atoms [34]. Therefore, once the electrodhole pairs are generated in the conductiodvalence bands, in the absence of an electric field produced by band banding, the loss of excess energy through localization on the trapping sites is more favorable than their direct chemical reaction with redox couples dissolved in the adjacent solution [27]. Due to the small size of the particles, the surface trapping sites of electrons and holes are not physically separated to distances required for stabilization of charge separation. They experience a strong electrostatic attraction leading to fast recombination. Understanding the properties of the surface states is therefore especially important in the interpretation of photoelectrochemical behavior of particulate semiconductors. In order to be able to block the surface trapping sites and convert them into coupling agents for interfacial electron transfer, one must understand the nature of trapping sites at the molecular level. The primary photochemical events that occur in colloidal TiOz particles upon band gap irradiation is the production of holes and electrons with the subsequent reactions they may undergo, such as recombination, trapping, and reactions with adsorbed molecules on the surface of the particles. It was suggested that interfacial electron transfer in Ti02 colloids occurs via surface Ti atoms which are coordinated with solvent molecules [35]. Meanwhile, the hole transfer occurs via surface oxygen atoms covalently linked to surface titanium atoms [36, 371. The structure of trapped electron and hole centers has attracted much attention during the last two decades. Both trapping sites have been investigated using several spectroscopic techniques such as flash photolysis [38, 391, pulse radiolysis [40-421, and EPR techniques [36, 37, 43-46]. While pulse radiolysis and optical spectroscopy enables investigation of the kinetics of both hole trapping and reactions of trapped holes, EPR spectroscopy provides detailed information of the molecular environment and the electronic structure of paramagnetic intermediates in electron transfer reactions. Below we review the last two decades of the study of charge separation and stabilization processes in nanoparticles of TiOz under light excitation revealed by the EPR technique.
Chapter 1; T. Rajh, et al.
7
1.3.2. Electron Trapping sites
The first identification of the electron centers formed under irradiation of colloidal Ti02 was reported by R.F. Howe and M.Gratze1 [44]. Pure colloidal solutions and aqueous solutions with additions of polyvinyl alcohol, iodine anion, acetate or methanol as a hole scavenger were studied. The authors found that depending upon the system, two different types of EPR spectra can be measured. All g-values of these spectra are less than 2, which is an indication that the signal arise from the electron centers. The first signal that was observed upon low temperature UV illumination in hydrated anatase, had narrow lines, and did not depend upon pH. The g-components of an axially symmetric tensor (gll= 1.957, gi = 1.988) were similar to previously reported centers in the niobium, antimony, or tantalum doped polycrystalline anatase [43,47,48]. This signal was described as an electron center trapped by a Ti4' ion, thus forming a Ti3' electron center. The narrow linewidth (around 5 G) and insensitivity to the deuteration of the colloid surface indicates that this center occupies a single site, and is located in the interior of an anatase particle. On warming to room temperature this signal decayed either through recombination or reduction of H20 at the surface. The authors [44, 491 tentatively assigned this center to the interstitial ion Ti3+,although they acknowledged that there is not much evidence in favor of the interstitial position as compared to substitutional. The concentration of the centers formed was low and the maximum measured number of Ti3' sites is of the same order as the number of colloid particle. Thus on average there is only one interstitial Ti3' ion produced per colloid particles. This observation explains the low intensity of the EPR signal in light illuminated polycrystalline anatase powders. With a decrease in the size of the particles the concentration of the particles increases and so does the EPR signal. The second type of EPR signal is observed in the presence of scavengers and has an intense broad (linewidth of around 100 G), asymmetric line, whose g-tensor parameters depends slightly upon the solution pH and scavenger molecules. The dependence of the signal on the properties of colloidal solution suggests that this signal can be attributed to the Ti3+ion sites on the surface of the particles. The values of the g-factors for these surface Ti3' particles are significantly lower than those usually found in solid TiO2. The only exception is a signal with g = 1.92 and linewidth of 100 G found in platinized Ti02 [50], which was attributed to the Ti3' ions formed in the vicinity of the platinum clusters. A deuteration experiment did not lead to the expected decrease of the linewidth. On the basis of this experiment the unusually broad linewidth of the surface Ti3+was explained by the distribution of the surface Ti3' sites with slightly different geometries, surroundings, and g-tensor components. Based on the EPR experiment the authors [44,49] presented the following description of the electron trapping process in colloidal TiO2. Band-gap irradiation produces hole-electron pairs, most of which recombine in the absence of the hole scavengers. Upon low temperature irradiation just a few electrons are trapped in the interior of the particles, producing interstitial Ti3' ions. The concentration of these ions is equal or less than the number of particles. In the presence of a hole scavenger many more electrons are trapped on the surface Ti4+, thus producing octahedral distorted Ti3' surface sites. These surface sites are not detected in neutral and alkaline solutions. Both surface and lattice centers are stable at room temperature in the presence of hole scavengers.
Charge Separation in Ti02 Revealed by EPR
8 Table 1.1.
g-values of the electron centers in the Ti02 nanoparticles System
gl
82
g3
Ref.
Ti3+interstitial, colloidal Ti" in Ti02 dopped with Sb
1.957
1.988
1.988
44
1.959
1.989
1.989
48
Ti3+in Ti02 dopped with Nb
1.962
1.992
1.992
47
Ti3' in hydrated anatase
1.960
1.990
1.990
49
Ti3' in heat treated Ti02
1.961
1.992
1.992
60
Ti3+in Ti02 before heat treatment Ti3+in surface-modified Ti02 Ti3+in surface-modified Ti02 Ti(H20)63+in frozen solution Ti3+on the surface in colloidal Ti02
1.957
1.990
1.990
60
1.96 1
1.988
1.988
57
1.958
1.988
1.988
57
1.892
1.892
1.988
75
1.885
1.925
1.925
44
Ti3+on the surface in colloidal Ti02
1.885
1.930
1.930
44
Ti3+on the surface in colloidal Ti02
1.880
1.945
1.945
44
Ti3+surface in cysteine-modified Ti02 Ti3+(surface) in cysteine-modified TiOz
1.958
46
1.934
46
Colloidal T i 0 2particles doped with transition metals, Fe, Mo, and V, have been studied by EPR [51-531. It was demonstrated that the dopant ions can participate in different ways in the low temperature electron and hole generation and trapping processes. Both hole and electron trapping by the dopant ions have been observed. Irradiation of the aqueous Fe-doped colloids causes the growth of Ti3+ signals, including a signal due to the aqueous Ti3+(HzO), complex with reversed values of the g-tensor (see Table 1.1). These changes where attributed to the inhibition of the hole-electron recombinations by Fe3+ions. Vanadium doping of colloids causes a similar inhibition. The decrease of the V5* signal upon irradiation was associated with the hole or electron trapping. Interstitial Mo& ions produced durin preparation of the colloid are extremely effective and irreversible electron traps, while Mo', on the other hand, is a reversible hole trap. Similar electron centers trapped on the surface or interior Ti3+ions were also observed in TiOz particles having modified surfaces [26, 36, 46, 54-57], anchored Ti02 particles supported on porous Vycor glass [58] or incorporated into silica gel pores by impregnation [59]. A small change in the g-values of the axially symmetrical g-tensor (previously identified as interstitial interior Ti3+ ions) upon the sample heating was reported [60] (gll= 1.957, g l = 1.990 for untreated sample and gll= 1.961, gl= 1.992 for sample heated at 700°C for 5 h). Authors attribute the first signal to the photogenerated electron trapped on the surface Ti3+ions, and the second one to the inner Ti3' ions.
Chapter 1; T. Rajh, et al.
9
The same type of signals were found in TiOz colloid particles, whose surface was modified with vitamin C [57]. Two typical electron centers in the unmodified TiOz particles were reported- broad surface and narrow interior. After surface modification with vitamin C the surface signal disappeared, instead two well resolved in the parallel orientation signals can be observed (gill = 1.9615, gIl = 1.9885 and 81'1 = 1.9581, gZL= 1.988). The signal with a parallel component at higher field was identified with an interstitial internal Ti3' ion. Meanwhile the narrow component at g = 1.9615 tentatively was associated with a signal motionally narrowed by an electron hopping from one center to another. A similar signal has been observed before from partially reduced rutile Ti02 and attributed to an electron loosely bound to an interstitial titanium ion or completely delocalized electron [61,62].
structure I
structure I1
structure I11
Fig. 1.4.Coordination of surface Ti atoms in TiOz nanoparticle. Dotted lines denote bonding to Ti02
lattice. The presence of a surface modifier also affects the electron surface trapping sites. In [46] it was shown that modification of the surface with cysteine leads to the two distinct surface electron trapping sites. In this case two asymmetric EPR signals with g = 1.958, and g = 1.934 are recorded. The same type of signals have been reported previously for Ti02 colloids in the presence of methanol [37]. In the latter case it was shown that these two signals belong to two distinct species which decay with different kinetics at different temperatures. While the g-factor of the interstitial Ti3+centers is not affected when the TiOz surface is modified with cysteine, the g-factor for surface Ti3' centers changes. This was explained as the result of the change in the axial crystal field that becomes stronger than in the OH coordinated structure. Two surface structures that contribute to the surface Ti3+ signal after surface modification have been proposed. The first signal, having the EPR signal at g = 1.934, was attributed to the case of one cysteine molecule coordinated to the surface Ti ion (Fig. 1.4, Structure 11) , and the second (g = 1.958) with two cysteine molecules coordinated to Ti atom (Fig. 1.4, Structure 111). The fate of the trapped electron with an increase in temperature was followed in work [46] for cysteine-modified TiOz aqueous colloid with lead ion added into solutions. It was shown that in this case lead is chelated with the carboxyl and mercapto groups of cysteine, and
10
Charge Separation in Ti02 Revealed by EPR
in the Ti02/cysteine/Pb2+ system cysteine acts as a bridging bidentate ligand. The low temperature EPR signal for trapped electrons did not change after the addition of lead ions, and the electrons were trapped on lattice and/or surface Ti atoms (Fig. 1.4) according to the equation:
I
I
ecb-+ (Ti02),Ti(IV)-O-CO-CH(NH2)CH~SPb-+
I
I
Ti(III)~att02(Ti0~),-ITi(IV)-O-CO-CH(NH~)CH~SPb or
I
I
(TiO2),Ti(III),~O-CO-CH(NH2)CH2SPb At 120 K, however, the relative intensity of the signal for lattice type Ti3' trapped electrons decreased, and surface trapped Ti electrons with g = 1.958 (associated with Structure 111) and g = 1.934 (associated with Structure 11) increased (Fig. 1.5): Ti(III)latt -+ Ti(III)sd
(1.3)
When the temperature was raised to 200 K the signal for the surface trapped electrons at the sites chelated with two cysteine molecules (Fig. 1.4) disappeared, while the signal for surface trapped electrons at the sites chelated with one cysteine molecule increased four times relative to the trapped hole signal:
These results indicate that the transition from surface state I11 to surface state I1 is thermodynamically favorable. The redox potential of the trapped site coordinated with one carboxyl groups is less negative than the redox potential of the site coordinated with two carboxyl groups simultaneously. On this basis the energy level diagram of the electron trapped sites was proposed (Fig. 6). Such temperature transformations are typical for Pb doped and pure aqueous colloidal solutions. When the temperature is raised to room temperature for Pb ion doped solution, all photogenerated electrons are scavenged by metal ions,
2Ti(III)1a~t0~(TiO~),Ti(IV)-O-CO-CH(NH~)CH~SPb -+
11
Chapter I ; T. Rajh, et al.
I
I
2(TiO~),Ti(III),~O-CO-CH(NH~)CH2SPb + 2(TiO2),Ti(IV)-O2C-CH(NH2)CH2SH + Pbo and precipitation of metallic lead is observed.
30 0 K+8 K
31 00.00
3200.00
3300.00
3400.00
Magnetic Field (C auss) Fig. 1.5. EPR spectra of degassed aqueous Ti02 colloids (0.2 M) in the presence of (0.1 M) cysteine and Pb2*ions (0.05 M) illuminated with 308 nm excimer laser at 77 K, recorded at different
temperatures indicated at the figure. Formation of metallic lead was identified by a Pb containin dark brownish-gray precipitate formed after steady state illumination of the TiOdcysteinePb system. It should be noted that P b 0 2 could not be reduced directly to metallic Pb because addition of sodium borohydride to the solution of Pb2' ions in the presence of Ti02 does not produce metallic lead.
f+
12
Charge Separation in Ti02 Revealed by EPR
The possibility of forming PbS was ruled out by the following experiments. Formation of PbS requires generation of HS- from cysteine. Illumination of cysteine-modified Ti02 did not lead to the reduction of cysteine which would result in H S formation, but led to the accumulation of trapped electrons having a broad optical absorption band at=,A 700-800 nm.
J
hv
Fig. 1.6. Energy level diagram of the observed electron trapping sites
In conclusion, in spite of the variety of EPR signals, the nature of the trapped electron in the Ti02 nanoparticles is well understood. There are two general types of traps - internal, having a narrow axially symmetric EPR signal, and surface with broad EPR lines. Reported gvalues of the trapped electron signals are summarized in Table 1.1. Magnetic resonance parameters of the internal, interstitial Ti3’ ions slightly vary due to the different delocalization of the unpaired electron density and symmetry of the local surroundings (presence of vacancies and impurities in the nearest coordination sphere). The same is true for the surface electron trap. In this case, g-values and the linewidth of the surface Ti3’ ions mainly depend upon surface modification. Conduction band electrons have not been recorded with EPR techniques probably owing to the fast recombination and trapping mechanism. 1.3.3.Hole Trapping Sites
Many publications are devoted to the EPR study of the hole trapped centers in Ti02 nanoparticles after low temperature irradiation [36,37,44-46,49,63-721. The interpretations of the observed EPR signals and assignments of the photogenerated hole species are less clear than electron traps, and several contradictions can be found in the literature. As it was shown recently [36,37,46,54,55,57] the main reason for the controversy is that holes are localized in the surface region of the nanoparticles and the structure of the hole centers strongly depends upon surface treatment and modification.
Chapter 1; T. Rajh, et al.
13
The studies of the radicals formed in aqueous colloids of Ti02, which was carried out with the use of an indirect spin-trapping technique, indicated the presence of OH' and HO; radicals [73, 741. Studies of the hole centers in hydrated anatase Ti02 nanoparticles were done in several labs [49,63,64]. While the authors of [63] reported observation of the H02 and 0 or 0;. species generated upon illumination of hydrate anatase in 02,Anpo et al. [64]for the same system observed a spectrum, which was assigned to the OH' radical. On the other hand Howe and Gratzel [49], observed similar EPR spectra with an asymmetric g-tensor: 2.002, 2.012, 2.016. This signal was assigned to the hole, which is localized on the oxygen anion not on the surface, but in the immediate subsurface layer. The hole was trapped in the following proposed structure: Ti4'0' Ti4+OH. Recently Micic et al. [37] reconsidered the assignment of the hole traps in the Ti02 colloids of anatase and rutile powders. The sensitivity of the EPR spectra was improved by preparing aqueous colloidal solutions with a large surface area owing to the smaller particle size and, as a consequence, larger concentration of the particles. For identification of the surface centers, colloidal systems were prepared in isotopically exchanged water: heavy water DO2 and " 0 enriched water H1702. It was demonstrated that holes produced by band gap irradiation of Ti02 colloids move from the oxygen lattice to the surface and are trapped directly on oxygen atoms bound to surface Ti4+ ions. Trap holes in anatase and rutile systems exhibit similar properties. The results obtained with Ti02 colloids prepared with 170enriched water confirms the assignment of trapped holes as oxygen surface anion radicals covalently bound to Ti4+ions in the form: Ti4+02Ti4+0-*.The intensity of the EPR signal from holes is very sensitive to hydration and total surface area. This signal completely disappears with the addition of the hole scavengers that are strongly bound to the surface, such as polyvinyl alcohol or vitamin K1. This was the first direct demonstration of the influence of surface modification on the charge transfer reaction and hole stabilization in the Ti02 colloidal nanoparticles. Table 1.2. g-values of the hole trap centers in the Ti02 nanoparticles. System
Ref., comments
gl
gz
Hydrated TiOl AN (Degussa P-25)
2.007
2.014
2.024
63, recorded at 12 K
Hydrated TiOz AN (Degussa P-25)
2.004
2.016
2.028
63, recorded at 77 K
Hydrated Ti02 AN particles
2.002
2.012
2.016
49, rec. at 4.2&77 K
Hydrated Ti02 AN particles
2.003
2.0146
2.0146
64, recorded at 77 K
Hydrated Ti02 AN (Degussa P-25)
2.007
2.014
2.024
36, recorded at 6 K
A1 doped Ti02
2.0034
2.0261
2.0297
66, recorded at 10 K
A1 doped Ti02
2.003
2.0189
2.0197
66, recorded at 30 K
g3
AN Ti02 before heat treatment
2.004
2.014
2.018
60, recorded at 77 K
AN Ti02 after heat treatment
2.004
2.018
2.030
60, recorded at 77 K
AN is anatase
14
Charge Separation in T i 0 2Revealed by EPR
The work of Nakaoka and Nosaka [60], in which anatase Ti02 powder was treated by heating at various temperatures in the air, should be mentioned. Two different signals in the region of the g-values corresponding to hole centers were detected in pretreated samples and samples heated at 700OC. On the basis of the comparison of g-tensors, the spectrum in the sample before heating was assigned to the subsurface Ti4+0' Ti4+OH structure, and after heating to the surface Ti4+02Ti4+ structure (see Table 1.2). 0 - O
1.3.4. Charge Separation in the Surface-Modijied Ti02 Colloids The approach that takes advantage of the presence of surface defect sites and converts them into coupling sites for selective bindings of photodegradable substrates is based on surface modification of nanocrystalline Ti02. Photogenerated electrons and holes lose significant energy in the trapping processes at the surface and cannot be used to carry out effective redox chemistry. By blocking the surface defect sites with appropriate surface modifiers one can enhance the redox properties of photogenerated charges, and, at the same time, enhance the rate of photodegradation by increasing the local concentration of pollutants on the particle surface. Metal oxide colloids have been effectively coupled with multifunctional ligands containing carboxyl groups that bind to the surface of nanoparticles [46].One can rationally design optimal photocatalysts by tailoring functional groups for selective adsorption of specific
rrapped holes
2
I
r1.987
trapped electrons
IIK-8
2.022
IIK-8K
2.008
d I I K - 8K
3200.00
3300 00
3 4 0 0 00
M a g n e t i c F i e l d (Gauss)
Chapter 1; T. Rajh, et al.
15
T i 0 Zleysteine
n/J---thiolac tate -ir r adiate d
3200 00
3300.00
3400 00
Magnetic Field (Gauss)
3500 00
Fig. 1.7. Surface modification of 45 8, TiOz colloids with different mercapto-carboxylic acids. EPR spectra of degassed aqueous TiOz colloids (0.3 M) illuminated with 308 rn excimer laser (a) in the presence of different surface modifiers illuminated at 77K and recorded at 8K (b) the same samples recorded at 150 K; at the bottom irradiation of pure TLA acid in N 2 0 leads to formation of sulfur centered radical dimmer of TLA.
pollutants. Extensive work on surface modification for efficient charge separation resulting in the removal of heavy metal ions has recently been performed [54, 5 5 , 761. The surface of colloidal Ti02 were modified with a series of bidentate and tridentate compounds having mercapto, carboxyl, and amino groups in different relative positions and appendant hydrocarbon chain lengths. EPR results have indicated principles for design of an optimal surface modifier for the reduction of heavy metal ions such as lead or cadmium [ 5 5 ] . Surface modifiers are effectively linked to the Ti02 surface via a carboxyl group. EPR results indicated that upon illumination of cysteine and S-methyl cysteine the carboxyl group acts as the primary trapping site for photogenerated holes (Fig. 1.7) at low temperatures (4-77 K). The holes are trapped on cysteine as a carboxyl radical (gx = 2.022 and g, = g, = 2.004) [77-791, while the electrons are trapped on the Ti02 particle. In cysteine-modified TiOl the charge separation distance increases further with increasing temperatures, due to the existence of the appending mercapto group.
16
Charge Separation in Ti02 Revealed by EPR
The formation of a sulfur centered radical at 150 K was observed. The increased separation distance prevents recombination of trapped charges and enhances the lifetime of trapped electrons. In the presence of metal ions the signal for trapped electrons disappears indicating their reaction with metal ions reducing them to the metallic state. However, when intraparticle charge transfer in the surface modifier is prevented by blocking the mercapto group which acts as the hole trap (S-methyl cysteine), stabilization of charge separation is not achieved and precipitation of lead is much less effective (Fig. 1.7). Stabilization of the charge pairs was also achieved by using modifiers in which mercapto groups chelate Ti surface atoms concomitantly with the carboxyl groups. This bidentate binding of surface modifiers such as mercaptoacetic acid (MAA) and thiolactic acid (TLA) restructures undercoordinated defect sites in nanocrystalline titania [801. EPR spectra reveal that at low temperatures (4-77K) photogenerated holes in MAA-modified colloids are trapped at the carboxyl group with the signal having g, = 2.022 and g, = g, = 2.004. In TLAmodified colloids, however, the transfer of the holes within the surface modifier molecule occurs even at low temperatures (77K). The EPR spectrum of TLA-modified T i 0 2 now is composed of five lines with hyperfine coupling of aH= 22 G. Because the signal for trapped electrons is still present indicating that electrons cannot be involved in the reduction of TLA, we attribute this radical to a =CH*CH2 radical [81], the product of oxidation of TLA. The sulfur radical of TLA would have the EPR spectrum either at g 2.17 in RS. form or g, = 2.052, g, = 2.021 and g, = 2.001 in RSa-SR form [82] none of which were observed upon illumination of surface-modified TiOl. These results show that once the mercapto group is bound to the Ti surface atom it does not act as a trap for photogenerated holes. Instead, the hydrocarbon group in a position a to the mercapto group which is also the group furthest from the particle surface, becomes the hole-trapping site. The oxidation potential of this radical is more positive than the corresponding sulfur centered radical and therefore can be scavenged with a variety of electron donating compounds. In MAA-modified colloids in which a side hydrocarbon group does not exist, the charge transfer ends at the carboxyl group and the photogenerated electrons at 150 K recombine with holes trapped close to the particle surface. Additionally investigation of the primary trap for photogenerated holes in TLA by illumination at 4.2 K has been carried out (Fig. 1.8) [83]. At helium temperatures a very rich hyperfine structure for photogenerated holes was observed. This spectrum was deconvoluted into a quartet representing a radical species that survives up to 30 K and a quintet representing a species that is present up to 200 K, when it probably disappears in a radical recombination reaction. The quartet signal with 20 G hyperfine coupling fits the signal that can be associated with the holes located at the backbone carbon of thiolactic acid (=CCH3), while the quintet with a hyperfine coupling of 22 G corresponds to the holes trapped on the appending hyrocarbon group in a position, the same signal observed at elevated temperatures (=CH.CH2). As the distance between charges is enhanced, charge separation becomes more efficient and more charges are available for reactions at the colloid surface. These EPR studies have shown that the surface modifier must contain a carboxyl group to bind to the colloid surface and at the same time to bind to the metal ions. The surface modifier has to have a hole trap that enhances photogenerated charge pair separation distance. A mercapto group that is in an a position relative to a carboxyl group enhances adsorption of
-
Chapter 1; T.Rajh, et al.
17
Fig. 1.8. EPR spectra of thiolactic acid modified Ti02 colloids illuminated with 308 nm excimer laser at 4.3 K and annealed at different temperatures indicated at the figure
both surface modifiers and heavy metals to small particle Ti02 colloids. In these systems, side hydrocarbon groups such as the -CH3 group in thiolactic acid provide a trapping site for holes that can be used for the design of systems in which the hydrophobic aliphatic or aromatic part of a surface modifier will be selectively used for oxidation of organic compounds. Using electrochemical methods we have also found that surface derivatives modify the redox properties of TiOz particles and the redox properties of metal ions [ S I . The crucial parameter for effective removal of heavy metal ions is the trade off between enhanced redox properties of Ti02 by surface modification and enhanced redox potential of chelated metal ions. EPR studies were also performed to determine the principles for the design of optimal surface modifiers for deposition of metal layers on a Ti02 nanocrystalline substrate having the potential applications in the formation of conducting patterns in integrated circuits [27]. The mechanism of charge separation that leads to reduction of copper and silver ions was investigated by EPR spectroscopy (Fig. 1.9). Different surface modifiers containing the carboxyl, phosphono, and amino groups were used. Illumination (355 nm) of alanine-modified Ti02 colloids at 10 K in the presence of copper ions leads again to the formation of signals associated with trapped electrons (see above) and a signal associated with holes trapped on the carboxyl groups (gx = 2.022 and g, = g, = 2.004). This result suggests that the carboxyl group is again participating in the collaborative binding of alanine to the Ti02 surface.
18
Charge Separation in Ti02 Revealed by EPR
Fig. 1.9. EPR spectra of Ti02 colloids modified with alanine (upper) and amino phosphono propionic acid (lower) in the presence of copper ions illuminated with 355 nm laser at 10 K and recorded at different temperatures indicated at the figure.
Chapter 1; T. Rajh, et al.
19
The g-factor for the surface trapped electrons was found to be g = 1.924, the same as in the unmodified TiOz colloids. These trapping sites are not significantly affected by adsorption of alanine, probably because of the low surface coverage of alanine. However, in the presence of copper, heating of the sample to room temperature resulted in the disappearance of the signal for trapped electrons. Under the same conditions, the reduction of copper ions to a metallic state was confirmed using X-ray absorption spectroscopies (XAS) [27]. When a carboxyl group in alanine was replaced with a phosphono group, much stronger adsorption of surface modifiers was observed. However EPR results show more hindered hole transfer to the phosphono group. Photogenerated holes in this system were primarily trapped as a symmetrical oxygen centered radical on the Ti02 surface (gx = g, = 2.029, g, = 2.007, Fig. 1.9), suggesting that hole transfer to the linking phosphono group is an activated process. Upon raising the temperature to 77 K, the surface trapped radical possibly is transformed to an oxygen centered radical on the phosphono group. The same radical was previously obtained by illumination of semiconductor colloids stabilized with trioctyl phosphine oxide [84].These results indicate weaker coupling of the phosphono groups to surface titanium atoms as compared to carboxyl groups from alanine.
3100.00
3200.00
3300.00
3400.00
3500.00
Magnetic Field (G ) Fig. 1.10. EPR spectra of TiOz colloids modified with alanine in the presence of methanol and metal ions illuminated with 355 nm laser at 77 K and recorded at different temperatures.
20
Charge Separation in Ti02 Revealed by EPR
In order to protect the surface modifier (cysteine, thiolactic acid or alanine) against oxidation, a sacrificial electron donor that can also enhance the reduction yield of metal ions can be introduced into the solution. We investigated the surface-modified TiOz in the presence of methanol. Methanol is a known current doubling agent [85] and can be easily oxidized by photogenerated holes. In the presence of methanol, the EPR signal of the surface-modified Ti02 colloid at 8.9 K (Fig. 1.10) is composed of a partially obscured set of three lines with separation of -18 G, and a set of two lines with 136 G separation (arising from the methanol radical (CH20(H)) and formyl radical (CHO), respectively). The signals associated with trapped electrons are those due to Ti3' in the bulk lattice (g = 1.988), and of Ti3+ at the surface with g = 1.924. Thus, the holes are transferred to adsorbed methanol which becomesis oxidized rather than the surface modifier (similar signals were obtained in cysteine, thiolactate and alaninemodified colloids). The large negative potential of the methanol radical (-0.95 V vs. NHE) [86] induces electron injection into colloidal Ti02 at 120 K with formation of surface trapped electrons and formaldehyde. Consequently, the yield of electrons is doubled (Fig. 1.10). This spectrum
Fig. 1.11. Schematic presentation of copper deposition on TiOz nanoparticles in the presence of methanol as current doubling agent.
disappears at room temperature indicating the reduction of metal ions as confirmed in XAS measurements. Based on these results, the following mechanism has been proposed (Fig. 1.11). Effective coupling of a carboxyl group to the surface Ti atoms was also demonstrated in the work of Konovalova et al. [56]. In this work surface modification of TiOz nanoparticles with carotenoids containing terminal -C02H groups was found to bind strongly to the nanocrystalline surface. Full electron transfer from carotenoid to the Ti02 nanoparticle in the ground state was demonstrated. The EPR spectra (77 K) prior to irradiation exhibit the signal of the carotenoid radical cation with g = 2.0028 and a broad line with g < 2 due to the electrons trapped on the Ti02 nanoparticles. The optical absorption spectra of TiOz nanoparticles modified by these carotenoid carboxylic acids show a broad absorption band with a maximum near 650 nm that is characteristic of the absorption of surface trapped electrons. Other carotenoid/TiOz systems do not produce any EPR signals in the absence of irradiation. However, illumination (400-600 nm) of aldoxime carotenoid-modified Ti02 colloids
Chapter 1; T. Rajh, et aZ.
21
(-NOH linking group) with light energy less than the Ti02 band gap at 77 K leads to the formation of the carotenoid radical cation with additional signals attributed to trapped electrons. Thus, EPR measurements demonstrate charge separation on the carotenoid-modified TiOz surface. Trapped electrons can reduce the acceptor molecules such as 2,5-dichloro-l,4-benzoquinone, nitrobenzene, and oxygen. The EPR signals associated with trapped electrons disappeared as a result of the reaction with the acceptors, and new EPR peaks, assigned to radical anions of the acceptor molecules, are observed. In this process TiOz nanoparticles act as mediators in the reduction sensitized by carotenoids. This EPR study has resulted in the design of alternative nanocrystalline photovoltaic cells photosentisized with carotenoids [ 871. The effects of surface modification of nanocrystalline Ti02 with specific chelating agents on photocatalytic degradation of nitrobenzene (NB)was investigated to design a selective and effective photocatalyst for removal of nitroaromatic compounds from contaminated waste streams [26]. The coordination sphere of the surface titanium atoms is incomplete and thus exhibits high affinity to oxygen containing ligands to form chelating structures. Three compounds were investigated to enhance adsorption of NB:a long chain carboxylic acid (lauryl sulfate) to make the surface of the TiOz particles hydrophobic; an amino acid (L-arginine) with a high affinity for hydrogen bonding and electron donating properties; and a benzene derivative (salicylic acid) that may form n-n donor-acceptor complexes. Arginine, lauryl sulfate, and salicylic acid were found to bind to Ti02 via their oxygen containing functional groups. The NB degradation on unmodified Ti02 follows both an oxidative and reductive mechanism but is completely altered to a reductive pathway over arginine-modified Ti02 and this reduction is enhanced upon addition of methanol. Arginine improved the coupling between NB and Ti02, and facilitated the transfer of photogenerated electrons from the T i 0 2 conduction band to the adsorbed NB. Modification with salicylic acid provided the greatest enhancement of NB adsorption over unmodified Ti02. However, only arginine-modified Ti02 resulted in enhanced photodecomposition of NB when compared to unmodified TiO2. The initial quantum yield for photodegradation of NB over arginine-modified TiOl was enhanced to @idt= 0.28 compared to the one obtained for Degussa P25 of @*, = 0.15. These results indicate that surface modification of nanocrystalline T i 0 2 with electron donating chelating agents is an effective route to enhance photodecomposition of nitroaromatic compounds. Surface modification can also result in a charge transfer complex with small particle TiOz colloids. In that case the optical absorption threshold is shifted to the visible range of the light spectrum, i.e. improved optical properties for solar energy conversion [55-57,881. These composite systems have a core of Ti02 and a shell of organic modifiers and exhibit hybrid properties that differ from the properties of both constituents. Using EPR we have shown that excitation of these composite systems results in charge separatio, electrons localized in the Ti02 core, and holes localized in the organic shell. It was found that nanosize Ti02 particles experience an adjustment in the coordination geometry of the Ti atoms near the particle surface from octahedral to square-pyramidal in order to accommodate the large surface curvature [57]. X-ray absorption near edge structure reveals that surface modification with enediol ligands (ascorbate, ortho-hydroxy cyclobutene dione, catechol, etc.) restores the pre-edge features of octahedrally coordinated Ti in the anatase crystal environment. Specific binding of the enediol modifiers to surface “corner defects”
22
Charge Separation in Ti02 Revealed by EPR
induces a 1.6 eV shift in the onset of absorption, compared to unmodified nanocrystallites. The red shift of the onset of absorption is explained by a charge transfer (CT) mechanism, as with CT salts [57].
Fig. 1.12. Light-induced EPR (X-band) spectra of degassed aqueous Ti02 colloids (0.3 M) modified with (top) ascorbic acid (0.1 M) or (bottom) dopamine, irradiated with white light (cut off filter 400 nm) at 4.2 K.
These newly developed systems have an important feature in that charge pairs are instantaneously separated into two phases following photoexcitation. Electron paramagnetic
resonance spectroscopy (EPR) was used to obtain a molecular understanding of the origin of the charge transfer complex and the corresponding electron accepting and donating sites. Contrary to optical measurements, EPR spectroscopy has the ability to unambiguously identify the species involved in the charge separation processes. It was found using EPR that excitation with visible light, A > 500 nm of the charge transfer complex of 45 A Ti02 nanoparticles with ascorbic acid, resulted in low temperature reversible charge separation in which an electron was excited from a donor to the acceptor site. This was manifest by the appearance of two signals in the continuous light photoinduced CW EPR spectrum, one for lattice trapped electrons (g = 1.988 and AHpp= 2.5 Gauss g = 1.958) and a signal at g = 2.004 with small hyperfine coupling
Chapter 1; T. Rajh, et al.
23
(Fig. 1.12). This latter signal has linewidth and g-value consistent with a carbon-centered radical. As electrons are involved in the reduction of Ti, this signal can be reasonably assigned to the reversible trap for photogenerated holes. The EPR signal at g = 2.004 was dependent on the ligands used for surface modification. In the case of modification with ascorbate ligands AHppwas 11 Gauss. A similar but broader signal with g = 2.004, AHpp=16 Gauss was obtained when dopamine was used as surface modifier. The linewidth decreased to AHpp= 10 Gauss upon using deuterated dopamine (ring-D2 or 2,2-D2). As the signal was dependent on the nature of the surface modifier and photogenerated electrons were found to be involved in the reduction of Ti, this signal can be assigned to the reversible trap for photogenerated holes. The fact that the trapping is reversible, suggests that oxidation is not followed by proton loss forming an allylic radical. The number of holes trapped on the surface modifier was rough1 equal to the number of trapped electrons and for the ascorbate modifier was found to be 4x10 Y to lx1015 spins cm-' in systems containing l O I 4 to 5 ~ 1 0 Ti02 ' ~ particles cm-' as determined by comparison to a calibrated sample. Under the same conditions, no EPR signals were observed from 20 8 Ti02 particles modified with ascorbate, probably because of the fast recombination rate due to the short charge separation distance (10 8). 1.4. Kinetic and Mechanism of the Charge Separation in Ti02 Colloid Nanoparticles as Studied by Time-Resolved EPR Technique.
Application of EPR spectroscopy provides not only data on the structure of the charge separated states in the Ti02 nanoparticles, but it can be a valuable source of information on the mechanism and kinetics of photoinduced electron-transfer reactions in such particles. Martino et al. [89] reported the first application of the time-resolved Fourier Transform EPR (FTEPR) in the study of room temperature formation of free radicals after laser excitation of the TiOl colloid. High spectral and time resolution of the technique allows authors to identify free radicals formed and to deduce the time constant of the electron-transfer reaction. Colloidal solution of Ti02 in ethanol containing methyl viologen (MV2') and coumarin 343 dye together with MV2' have been studied. In both cases the reduction of MV" was observed. The kinetics indicate that the trapped electrons responsible for MV2' reduction are generated by excitation of adsorbed dye molecules, which leads to electron injection into the conduction band of the semiconductor nanopartcles. Dye modification of the Ti02 surface strongly increases the free radical yield. However, the rate of the electron-transfer from semiconductor particles to acceptor in solution is strongly attenuated as the coverage of the surface approaches saturation level. These results provide the first direct experimental confirmation of the suggestion [go] that the efficiency of Gratzel photovoltic cells depends on the degree of coverage of the TiOz anode by an insulating layer which inhibits the direct return of the electron, injected to the conduction band, to I i present in solution. The mechanism of the charge separation in the colloidal Ti02 particles was studied at low temperature by direct time-resolved and light-modulated EPR techniques [91]. The recombination kinetics in the nanocrystalline semiconductor particles usually is very fast, on
24
Charge Separation in Ti02 Revealed by EPR
the order of picoseconds [38]. The efficiency of charge separation is low [89] and can not be observed by the time-resolved EPR techniques, unless the charge separation is enhanced by reaction with adsorbed species. As described above enhanced charge separation and improved optical properties of nanocrystalline TiOz that involve photoinduced interfacial electron transfer from surface modifiers into the conduction band of nanocrystalline Ti02 particles have been reported [57]. These systems have an important feature that charge pairs are instantaneously separated into two phases, the holes on the donating organic modifier and the electrons in the conduction band of TiOz. The charge separation is reversible in these systems at low temperatures. Surface modification with bidentate ortho- substituted hydroxylated electron donating ligands allows observation of electron spin polarization phenomena in continuous wave time- resolved EPR experiments [91]. The obtained data reveals the mechanism of the charge separation process. This is the first direct observation of transient species involved in the early stages of charge pair formation in surface-modified nanocrystalline TiOz using timeresolved EPR techniques. In contrast to the optical measurements that give extensive kinetic information on transient species, EPR spectroscopy can unambiguously identify paramagnetic species involved in the charge separation processes, monitor their molecular environment and spin dynamics. Using conventional continuous wave EPR spectroscopy two li ht-induced reversible signals attributed to oxidized donors and reduced acceptors in 45 TiOZ nanoparticles modified with ascorbic acid or dopamine were obtained at helium temperatures. As discussed above, the first signal ( g i = 1.988, g', = 1.961 and gZ1= 1.958, with a line width AHpp= 2.5 Gauss) is characteristic of a radical in which the unpaired electron occupies the d-orbitals of lattice Ti atoms [57] having a strong component of angular momentum due to the spin orbit coupling. No EPR signals associated with the surface components were observed indicating that both ascorbate and dopamine ligands raise the energy of the surface trapping sites. The EPR signal at g=2.004 was dependent on the ligands used for surface modification and was assigned to the reversible trap for photogenerated holes. By applying time-resolved (TR) EPR techniques (insert Fig. 1.13), the formation of the initial radical intermediate has been observed. The EPR spectrum obtained one microsecond after the laser excitation at helium temperatures (Fig. 1.13) is composed of two emissive lines and one absorptive line and the overall spectrum exhibits excess emission. The emissive line at g = 2.004 and absorptive line at g = 1.988 are observed at the same g values as the signals associated with the holes on the ascorbate modifier and electrons in the TiOz lattice, respectively. The algebraic average (gave= (gh+ g,)/2) of the low-field and high-field lines is very nearly equal to the central emissive line at g=1.995, which was not observed in continuous light measurements. This suggests that the origin of the signal comes from the exchange interaction in electrons and holes, otherwise localized Ti d-orbitals (electrons) and carbon centered 71: orbitals of the organic modifier (hole). The EPR spectrum of photoexcited 45 A Ti02 modified with dopamine (Fig. 1.14) shows similar properties, except that the linewidth of the signal attributed to the dopamine radical cation is larger, as observed in corresponding continuous light EPR measurements. Because of the broader signal for trapped holes, the two emissive signals overlap. However, the shoulder in the spectrum has a g-value of 1.995, the same value as observed in ascorbate-
x
Chapter I ; T. Rajh, et al.
25
modified TiOz. Again, the algebraic average of the low-field and high-field lines is very nearly equal to the central emissive line at g = 1.995. hv I
I
IA
\I
3300.00
Emission
3320.0vv
TR g=1.995
Fig. 1.13. Comparison between integrated continuous light-induced (upper trace) and time-resolved pulsed laser-induced (lower trace) EPR spectra from 45A Ti02 (0.3M) modified with ascorbic acid (0.08 M). The lower trace was obtained with a 550 nm laser (laser intensity 10 rnJ per pulse, 10 ns pulse duration, 20 scans), 1 ps after the laser pulse. Both spectra were recorded at 8 K. Insert: schematic presentation of events in time-resolved direct detection. The spectrum is taken at the time T after the laser pulse for each magnetic field using gate integrators. Magnetic field (H) is not
modulated. These observations suggest that the polarization features presented in Fig. 1.13 and 1.14 can be explained by the following (see Fig. 1.14). Light absorption in Ti02 yields the formation of conduction band electrons and holes that form singlet exciton states. Strong spin orbit coupling of Ti facilitates intersystem crossing to the triplet exciton. The triplet excitons are longer lived and have the potential to undergo further chemical transformations. The paramagnetic species thus produced are trapped holes and electrons, which have not recombined due to their triplet spin character. The trapped holes and electrons have very different environments, different g factors, which foster S-To Ag mixing in the radical pairs [92-971 giving rise to the emission in the higher g-factor hole and the absorption in the lower g-factor trapped electron signals [98]. There is
Charge Separation in Ti02 Revealed by EPR
26
Emission
hv ->exciton
TR
s
ISC ->excitonT
Fig. 1.14. Comparison between integrated continuous light-induced (upper trace) and time-resolved pulsed laser-induced (lower trace) EPR spectra from 45A TiOz (0.3M) modified with dopamine (0.08 M).The lower trace was obtained with a 550 nm laser (laser intensity 10 mJ per pulse, 10 ns pulse duration, 20 scans), 1 ps after the laser pulse. Both spectra were recorded at 8 K. Right section shows how triplet radical pair mechanism of CIDEP in addition to fast exchange can contribute to the observed polarized spectrum.
Chapter I ; T. Rajh, et al.
27
also an apparent overall emission character to the spectrum which could arise from the triplet polarization induced in the triplet exciton which is the precursor of the charge separated trapped holes and electrons. Some radical pairs also exhibit exchange effects giving rise to the EPR line at the midpoint between the lines of non-interacting radicals. This fast exchange feature was previously reported for analogous corehhell quantum dots systems using optically detected magnetic resonance. Lifshitz et al. [99] have suggested that the fast exchange signal features may be a consequence of the substantial electron-hole coupling in the exciton. The separation distance for exciton interaction in anatase was previously found to be -15 8, [loo].
n
/ vA
g
R , = Holes on modifier
A
I 3260 00
I
1 3280 00
I
I 3300 00
I
I 3320 00
I
1 3340 00
I
I 3360 00
Magnetic Field ( G a u s s ) Fig, 1.15. Integrated LFM EPR spectru? of dopamine- modified (upper trace) and 6-pamitate ascorbic acid modified (lower trace) 42A TiOz Insert: Schematic presentations of events in LFM EPR. Both magnetic field (H) and light (hv) are modulated with frequencies a f i e l d and a l i g h t , respectively.
A subset of electron-hole radical pairs exhibits features of Spin Correlated Radical Pair (CRRP) electron spin polarization mechanism [ 1011 which can be observed at somewhat longer times via lighvfield modulated (LFM) EPR measurements. This technique is only sensitive to the light dependent part of the EPR spectrum on the time scale of the light modulation frequency (millisecond regime, insert Fig. 1.15). Using LFh4 EPR it was observed that both the transitions of the holes localized on the surface modifier and electrons localized on the TiOz
Charge Separation in Ti02 Revealed by EPR
28
lattice were split into antisense doublets (two sets of absorptive (A) and emissive (E)lines, Fig. 1.15). The electron spin polarized spectrum obtained with ascorbate used as the surface modifier in conjunction with the energy level diagram for triplet born CRPP dominated charge pair polarization is shown in Fig. 1.16.
I
I
T-triplet S -singlet
R
I 3260.00
I
I 3280.00
I
I 3300.00
I
I 3320.00
I
I 3340.00
I
I 3360.00
Magnetic Field (Gauss)
Fig. 1.16. Integrated LFM EPR spectrum of ascorbate-modified 42A TiOz in conjunction with energy level diagram for triplet born CRPP dominated charge pair polarization.Light was modulated with frequency 0.5 KHz.
The low-field signal associated with photogenerated holes varies for different surface modifiers; the presence of hyperfine interactions in the samples modified with 6-palmitate ascorbic acid (6-PAA) have induced further splitting of the absorptive and emissive signals, while in dopamine-modified samples the signal was significantly broadened (Fig. 1.15). On the other hand, the signal associated with photogenerated electrons trapped in Ti02 lattice shows the same features regardless of the surface modifier. The A/E/A/E pattern of the photogenerated EPR signal reflects the existence of weak spin-spin interaction between photogenerated holes (on organic modifier) and electrons (in T i 0 2 lattice) [101-1031. This weak interaction and presence of CRRP mimics the characteristics of the EPR signal observed for charge separated state P'Q- (where P is chlorophyll electron donor and Q is quinone electron acceptor) in natural photosynthetic systems [ 1041 and has been replicated in a molecular donor-acceptor model system.
Chapter I ; T. Rajh, et al.
29
The theoretical modeling of the spin polarized LFM EPR spectrum of Ti02 nanoparticles modified with ascorbic acid has been carried out with a general analytical treatment of spin-correlated radical pair EPR spectra having weakly coupled spins [105]. According to the treatment each spin polarized signal consists of three independent contributions. The first contribution is determined by the exchange interaction, second -by the electron dipole-dipole interaction, and the third - by the contribution of the thermalized spectra. Individual equilibrium spectra of the hole stabilized on the ascorbic acid radical and trapped = (1.988, 1.988, electron were simulated with axial g-tensors: &ole = (2.004,2.004, 2.000), gelectr. 1.958) according to the data experimentally determined earlier [57]. Some anisotropy of the linewidths originating from unresolved hyperfine interaction was assumed to improve the fit: = (4.3 Gs, 4.3 Gs, 5.6 Gs), = (2.5 Gs, 2.5 Gs, 7.0 Gs). This set of individual magnetic parameters was used to simulate the equilibrium, exchange derivative, and dipole derivative lines for both radicals according to the previously formulated approach, formulated in [106]. The best fit (shown on Fig. 1.17) was simulated without any contribution of the equilibrium spectrum and with the dipole (HD)contribution dominating over that of exchange (H,): HD/HJ= -15. One very important conclusion from the analyses is that the presence of spin polarization assumes a nonvanishing angular average . With enhanced coupling between the donating and accepting site attained, one can expect that the rate of forward electron transfer is enhanced. In the Ti02 system with a modified surface the strong electronic coupling between ortho-substituted hydroxylated aromatic ligands and metal oxide particles leads to the instantaneous forward electron transfer from ligand to semiconductor nanoparticles. On the other hand, charge separation upon electron trapping leads to formation of radical pairs with negligible coupling. The magnitude of electron interaction is smaller than lo4 meV, which is four orders of magnitude smaller than the exciton binding
K
Charge Separation in Ti02 Revealed by EPR
30
1
3280
' 3300
3320
'
3340
'
3360
3380
'
3100
Magnetic Field (G)
Fig. 1.17. Simulated (solid line) and experimental (dotted line) spectra of radical pairs in colloidal TiOz
modified with ascorbic acid. The g-tensors were taken as an axially simmetric with: g(R'+)=(2.004,2.004,2.000), g(R") = (1.988, 1.988, 1.958)according to the data experimentally determined earlier [57]. Anisotropic line width was obtained from the simulation of the equilibrium spectrum: AH(R'+) = (4.3Gs, 4.3 Gs, 5.6 Gs),AH(R'-) = (2.5 Gs, 2.5 Gs,7.0 Gs). The best fit of the polarized spectrum was simulated without any contribution of the equilibrium spectrum and with the dipole (H,) contribution dominating over the exchange (HJ) one: H a J -15. Good fit can be obtained only when HDsmaller than 3.5G. 15:
energy in bulk TiOl [105]. Thus, small exchange interaction J may stabilize the charge separated state resulting in prolonged lifetimes. A long lifetime of the fully charge separated state significantly enhances the probability of reaction with chemical species present in the surrounding environment, Le. more efficient chemistry. In this way a very important feature of the highly optimized photosynthetic system in natural photosynthesis is replicated - fast charge separation rates and slow charge recombination rates. These findings suggest that the investigated surface modifier ligands can act as conductive leads that may allow electronic linking of the Ti02 nanoparticle into molecular circuits. When covalently linked (wired) to electron donating moieties, photoinduced electron transfer of stabilized radical pairs can further extend the lifetime of charge separation. Summary The EPR technique is a valuable tool for gaining a better understanding of the mechanism and kinetics of photoinduced electron-transfer reactions in colloidal TiOl systems. There are two main features of the EPR technique which are indispensable in these studies.
Chapter 1; T. Rajh, et al.
31
First, due to the specificity of the EPR spectra, the paramagnetic species, involved in the charge separation, can be uniquely identified. Second, interconversion between species and kinetics of the charge-transfer reaction can be recorded in the wide time range (from ns to days) by monitoring the intensity and changes of the EPR signals. It is worth mentioning that in the past two decades many new modern methods of EPR spectroscopy have been developed. Among them are multifrequency EPR spectroscopy, multidimensional pulsed Electron Spin Echo spectroscopy, pulsed double resonance techniques (ENDOR and ELDOR), and high frequency EPR spectroscopy. In the near future application of the modern methods of the EPR spectroscopy can be anticipated to provide more detailed characteristics of the electron transfer processes in nanoparticles. Most of the efforts of researchers, working in the field of semiconductor colloids, are directed to the chemical modification of the system to make photoproduction reactions more efficient. EPR spectroscopy will certainly contribute to the development of new nanoscale materials which can effectively undergo charge separation for wide application in photocatalysis and solar energy conversion.
Acknowledgement This work was supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences under contract W-3 1-109-Eng-38. The authors acknowledge the insights gained form discussions with A. Trifunac. REFERENCES
10.
Gerisher H. and Heller A. J. Electrochem. SOC, 139, No. 1, 113-118 (1992). Photocatalysis -- Fundamentals and Applications, N. Serpone and E. Pelizzetti (Eds.), John Willey & Sons, New York (1989). Fox M. A. Top. Curr. Chem., 142, NO. 1,72-99 (1987). Ollis D. F. Environ. Sci. Technol., 19, No. 6,480-484 (1985). Photocatalytic Purification and Treatment of Water and Air, D. F. Ollis and H. Al-Ekabi (Eds.), Elsevier, Amsterdam (1993). Kormann C., Bahnemann D. W. and M. R. Hoffmann, Environ. Sci. Technol., 25, No. 3,494-500 (199 1). O’Regan B. and Gratzel M. Nature, 353, No. 6346,737-740 (1991). Rehm J. M., McLendon G. L., Nagasawa Y., Yoshihara K., Moser J. and Gratzel M. J. Phys. Chem., 100, No. 23, 9577-9578 (1996). Nasr C., Liu D., Hotchandani S . and Kamat P. J. Phys. Chem., 100, No. 26, 11054-11061 (1996). Burfeindt B., Hannappel T., Stork W. and Willig F. J. Phys. Chem., 100, No. 41, 16463-16465
11. 12.
Mayer G. J. J. Chem Education, 74, No. 6, 652-656 (1997). Martini I., Hodak J., Hartland G. V. and Kamat P. V. J. Chem. Phys., 107, No. 19, 8064-8072
1. 2.
3. 4.
5. 6. 7. 8. 9.
(1996). (1997).
13. 14.
Huang S. Y., Schlichthorl G., Nozik A. J., Gratzel M. and Frank A. J. J. Phys. Chem. B, 101, No. 14,2576-2582 (1997). Ellingson R. J., AsbuIy J. B., Ferrere S . , Ghosh H. N., Sprague J. R., Lian T. Q.and Nozik A. J. J. Phys. Chem. B, 102, No. 34,6455-6458 (1998).
Charge Separation in Ti02 Revealed by EPR
32 15. 16. 17.
Pelizzetti E., Borgarello M., Minero C., Pramauro E., Borgarello E. and Serpone N. Chemosphere, 17, NO.3,499-5 10 (1988). Matthews R. W. WaterRes., 25, No. 10, 1169-1176 (1991). Carey J. H., Lawrence J. and Tosine H. M. Bull. Environ. Contam. Toxicol., 16,No. 6, 697-701 (1976).
23.
Sclafani A., Palmisano L. and Schiavello M. J. Phys. Chem., 94, No.2, 829-832 (1990). Serpone N. Res. Chem. Zntenned., 20, No. 9, 953-992 (1994) and references within. Domhech J. and AndrCs M. Gauetta Chimica Ztaliana, 117, No. 8,495-498 (1987). Serpone N., Ah-You Y. K., Harris R., Hidaka H., Pelizzetti E. and Tran T. P. Solar Energy, 39, NO. 6,49 1-498 (1987). Prairie M. R., Pacheco J. and Evans L. R. Solar Detoxification of Water Containing Chlorinated solvents and Heavy Metals via Ti02 Photocatalysis, Prepared for the 1992 ASME Intl. Solar Energy Conf., April, SAND91-1285C, Sandia National Laboratory, Albuquerque, NM (1992). Hoffmann M. R., Martin S. T., Choi W. and Bahnemann D. W. Chem. Rev., 95, No. 1, 69-96
24.
Schmelling D. C., Gray K. A. and Kamat P. V. Environ. Sci. Technol., 30, No. 8, 2547-2555
25. 26.
38.
Piccinini P., Minero C., Vincenti M. and Pelizzetti E. Catal. Today, 39, No. 3, 187-195 (1997). Makarova 0. V., Rajh T., Thurnauer M. C., Martin A., Kemme P. A. and Cropek D. Environ. Sci. Technol., 34, No. 22,4797-2803 (2000). Rajh T., Nedeljkovic J., Chen L. X., Tiede D. M. and Thurnauer M. C. J. Adv. Oxid. Technol., 3, NO. 3, 292-298 (1998). Wu Z. Y., Ouvrard G., Gressier P. and Natoli C. R. Phys. Rev. B, 55, No. 16, 10382-10389 (1997). Pascual J., Camassel J. and Mathieu H. Phys. Rev B, 18, No. 10, 5606-5614 (1978). Boudreaux D. S., Williams F. and Nozik A. J. J. Appl. Phys., 51, No. 4,2158-2163 (1980). Nozik A. J. Ann. Rev. Phys. Chern., 29, 189-222 (1978). Morrison S. R., Electrochemistry at Semiconductor and Oxidized Metal Electrodes, p. 56, Plenum Press, New York (1980). Hagfeldt A. and Gratzel M. Chem. Rev., 95, No. 1,49-68 (1995). Morrison S. R., Electrochemistry at Semiconductor and Oxidized Metal Electrodes, p. 12, Plenum Press, NewYork (1980). Kolle U., Moser J. and Gratzel M. Znorg. Chem., 24, No. 14,2253-2258 (1985). Micic 0. I., Zhang Y., Cromack K. R., Trifunac A. D. and Thurnauer M. C. J. Phys. Chem., 97, NO.28,7277-7283 (1993). Micic 0. I., Zhang Y., Cromack K. R., Trifunac A. D. and Thurnauer M. C. J. Phys. Chem., 97, NO. 50, 13284-13288 (1993). Rothenberger G., Moser J., Gratzel M., Serpone N. and Sharma D. K. J. Am. Chem. SOC.,107, No.
39. 40.
Bahnemann D., Henglein A., Lilie J. and Spanhel L. J. Phys. Chem., 88, No. 4,709-71 1 (1984). Dimitrijevic N. M., Savic D., Micic 0. I. and Nozik A. J. J. Phys. Chem., 88, No. 19, 4278-4283
41.
Micic 0. I., Rajh T. and Comor M. V., in: Electrochemicstry in Colloids and Dispersions, R. A . Mackay and J. Texter (Eds.), p. 457, VCH Publ., New York (1992). Lawless D., Serpone N. and Meisel D. J. Phys. Chem., 95, No. 13,5166-5170 (1991). Meriaudeau P., Che M. and Jorgensen C. K. Chem. Phys. Lett., 5, No. 3, 131-133 (1970). Howe R. F. and Gratzel M. J. Phys. Chem., 89, No. 21,4495-4499 (1985). Gonzales-Elipe A. R., Soria J. J. and Munuera G. Chem. Phys. Lett., 57, NO. 2,265-268 (1978).
18. 19. 20. 21. 22.
(1995). (1996).
27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37.
26,8054-8059 (1985).
(1984).
42. 43. 44. 45.
Chapter I ; T. Rajh, et al.
46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82.
33
Rajh T., Ostafin A. E., Micic 0. I., Tiede D. M. and Thumauer M. C. J. Phys. Chem., 100, No. 11, 4538-4545 (1996). Kiwi J., Suss J. T. and Szopiro S. Chern. Phys. Lett., 106, No. 1,2, 135-138 (1984). Che M., Gravelle P. C. and Meriaudeau P. C. R. Acad. Sci. Paris C, 268, No. 9,768-771 (1969). Howe R. F. and Gratzel M. J. Phys. Chem., 91, No. 14, 3906-3909 (1987). Huizinga T. and Prins R. J. Phys. Chem., 85, No. 15,2156-2158 (1981). Moser J., Gratzel M. and Gallay R. Helv. Chim. Acta, 70, No. 6, 596-604 (1987). Gratzel M. and Howe R. F. J. Phys. Chem., 94, No. 6,2566-2572 (1990). Soria J., Conesa J. C., Augugliaro V., Palmisano L., Schiavello M. and Sclafani A. J. Phys. Chem., 95, NO. 1,274-282 (1991). Rajh T., Tiede D. M. and Thumauer M. C. J. Non-Cryst. Solids, 207, No. 2, 815-820 (1996). Rajh T., Tiede D. M. and Thumauer M. C. Acta Chem. Scand., 51,610-618 (1997). Konovalova T. A., Kispert L. D. and Konovalov V. V. J. Phys. Chem. B, 103, No. 22,4672-4677 (1999). Rajh T., Nedeljkovic J. M., Chen L. X., Poluektov 0. and Thurnauer M. C. J. Phys. Chem. B, 103, NO. 18, 3515-3519 (1999). Anpo M., Shima T., Fujii T., Suzuki S. and Che M. Chem. Lett., Chem. SOC.Japan, No. 10, 1997 (1987). Ahn S. W. and Kevan L. J. Chem. SOC.,Farad. Trans., 94, No. 20,3147-3153 (1998). Nakaoka Y. and Nosaka Y . J. Photochem. Photobiol. A: Chem., 110, No. 3,299-305 (1997). Kerssen J. and Volger J. Physica, 69, No. 2,535-561 (1973). Yagiu E. and Hasiguti R. R. J. Phys. SOC.Japan, 43, No. 6, 1998-2005 (1977). Gonzales-Elipe A. R., Munuera G. and Soria J. J. J.Chern.Soc., Faraday Trans. I , 75, 748-761 (1979). Anpo M., Shima T. and Kubokawa Y. Chem. Lett., No. 12, 1799-1802 (1985). Gratzel M. and Howe R. F. J. Phys. Chem., 94, No. 6,2566-2572 (1987). Zwingwl D. Solid State Commun., 20, No. 4,397-400 (1976). Aundaithai M. and Kutty T. R. N. Mater.Res.Bull., 23, No. 11, 1675-1683 (1988). Anpo M., Shima T., Kodama S. and Kubokawa Y. J. Phys. Chem., 91, No. 16,4305-4310 (1987). Nikisha V. V., Shelimov B. N. and Kazanskii V. B. Kineticu i Kataliz, 15, No. 3, 676-680 (1974) (in Russian). Meriaudeau P. and Verdine J. C. J. Chem. Soc., Faraday Trans. II,72, No. 2,472-480 (1976). Lumpov A. I., Mikheikin I. D., Chuvylkin N. D., Zhidomirov G. M. and Kazanskii V. B. Kinetica i Kataliz, 17, No. 3, 803-805 (1976) (in Russian). Berdnikov V. M. and Schastnev P. V. Kinetica i Kataliz, 16, No. 1, 83-91 (1975) (in Russian). Jaeger C. D. and Bard A. J. J. Phys. Chem., 83, No. 24, 3146-3152 (1979). Harbour J. R., Tramp J. and Hair M. L. Can. J. Chem., 63, No. 1,204-208 (1985). Wilson R. C. and Meyers R. J. J. Chem. Phys., 64,No. 5,2208-2211 (1979). Chen L. X., Rajh T., Micic O., Wang Z., Tiede D. M. and Thurnauer M. C. Nucl. Znstr. Meth. Phys. Res. B, 133,No. 1-4, 8-11 (1997). Box H. C., Freund H. G., Lilga K. T. and Budzinski E. E. J. Phys. Chem., 74, No. 1,40-52 (1970). Box H. C., Budzinski E. E. and Lilga K. T. J. Chem. Phys., 57, No. 10,4295-4298 (1972). Kou W. W. H. and Box H. C. J. Chem. Phys., 64,No. 7,3060-3062 (1976). Chen L. X., Rajh T., Wang Z. and Thumauer M. C. J. Phys. Chem., 101, No. 50, 10688-10697 (1998). Ayscough P. B. and Evans H. E. J. Phys. Chem., 68, No. 10,3066-3068 (1964). Nelson D. J., Petersen R. L. and Symons M. C. R. J. Chem. SOC.,Perkin Tr. II, No. 15,2005-2015 (1977).
34 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99.
100. 101. 102. 103. 104. 105. 106.
Charge Separation in Ti02 Revealed by EPR
Rajh T. and Thumauer M. C., unpublished work. Rajh T., Micic 0. I. and Thumauer M. C., unpublished work. Nogami G. and Kennedy J. H. J. Electrochem. Soc., 136, No. 9,2583-2588 (1989). Breintkamp M., Henglein A. and Lilie J. Ber. Bunsenges., Phys. Chem., 80, NO. 10, 973-979 (1976). Gao F. G., Bard A. J. and Kispert L. D. J. Photochem. Photobiol., 130, No. 1,49-56 (2000). Rajh T., Thumauer M. C., Thyagarajan P. and Tiede D. M. J. Phys. Chem., 103, NO. 12, 21722177 (1999). Martino D. M., van Willigen H. and Spitler M. T. J. Phys. Chem. B, 101, No. 44, 8914-8919 (1997). Nazzeruddin M. K., Kay A., Rodicio I., Humphry-Baker R., Muller E., Liska P., Vlachopoulos N. and Gratzel M. J. Am. Chem. SOC.,115, No. 14, 6382-6390 (1993). Rajh T., Poluektov 0. G., Dubinski A. A,, Wiederrecht G., Trifunac A. D. and Thumauer M. C., Chem. Phys. Lett., 344, No. 1,2,31-39 (2001). Fessenden R. W. and Schuler R. H. J. Chem. Phys., 39, No. 9,2147-2195 (1963). Eiben K. and Fessenden R. W. J. Phys. Chem., 75, No. 9, 1186-1201 (1971). Atkins P. W., Buchanan L. C., Gurd R. C., McLauchlan K. A. and Simpson A. F. Chem. Comm., NO.9, 513-514 (1970). Atkins P. W. and McLauchlan K. A., in: Chemically Induced Magnetic Polarization, A. R. Lepley and G. L. Closs (Eds.), p. 41, John Wiley & Sons, New York (1973). Trifunac A. D. and Thumauer M. C., in Time Domain Electron Spin Resonance, L. Kevan and R. N. Schwartz (Eds.), p. 107, John Wiley & Sons, New York, (1979). Trifunac A. D. and Avery A. C. Chem. Phys. LRtt., 27, No. 1, 141-143 (1974). Trifunac A. D. and Thumauer M. C. J. Phys. Chem., 62, No. 12,4889-4895 (1975). Lifshitz E., Porteanu H., Glozman A., Weller H., Pflughoefft M. and Echymuller A. J. Phys. Chem. B, 103, No. 33,6870-6875 (1999). Tang H., Prasad K., Sanjines R., Schmid P. E. and Levy F. J. Appl. Phys., 75, No. 4, 2042-2047 (1994). . . Thurmauer M. C., Katz J. J. and Norris J. R. Proc. Natl. Acad. Sci. USA, 72, No. 9, 3270-3274 (1975). Thumauer M. C. and Noms J. R. Chem. Phys. Lett., 76, No. 3,557-561 (1980). Noms J. R., Moms A. L., Thumauer M. C. and Tang J, J. Chem. Phys., 92, No. 7, 4239-4249 (1990). Snyder S. and Thurnauer M. C., in: The Photosynthetic Reaction Center, J. Deisenhofer and J. R. Noms (Eds.), Academic Press, , Vol. 11, p. 285 (1993), and references therein. Pascual J., Camassel J. and Mathieu H. Phys. Rev. B, 18, No. 10, 5606-561 1 (1978). Dubinski A. A., Perekhodsev G. D., Poluektov 0. G . ,Rajh T. and Thumauer M. C., J. Phys. Chem B, 106, NO. 5, 938-944 (2002).
Chemical Physics of Nanostructured Semiconductors, pp. 35-82 A.I. Kokorin and D.W. Bahnemann (Eds.) 0 VSP 2003.
CHAPTER 2
Kinetic Peculiarities of Photocatalitic Reactions on CdS Nanocolloids D. V. Bavykin, E. N. Savinov, I. N. Martyanov and V. N. Parmon Boreskov Institute of Catalysis, Novosibirsk, Russia Kevwords: Kinetics, photocatalysis, Cadmium suljide, nanoparticles, colloids, electron transfer, photoreduction
2.1. Introduction Kinetic studies of photoreactions on semiconductor nanoparticles are important for both science and practice. Of scientific interest are the so-called quantum size effects, which are most pronounced on these particles: shifting the edge of adsorption band, participation of <> electrons in the reactions and recombination, dependence of the quantum yield of luminescence and reactions on the excitation wavelength, etc. In one way or another all these phenomena affect the features of photocatalytic reactions. At present photocatalysis on semiconductors is widely used for practical purposes, mainly for the removal of organic contamination from water and air. The most efficient commercial semiconductor photocatalysts (mainly the Ti02 photocatalysts) have primary particles of size 10-20 nm, i.e., they consist of nanoparticles. Results of studying the photoprocesses on semiconductor particles (even of different nature) are used to explain the regularities of photocatalytic processes. This indicates the practical significance of these processes. In this work the authors summarize their own studies of photoprocesses on CdS colloids with particles of various size. In these studies, attention was given precisely to photocatalytic reactions on CdS, the photocatalytic reactions on TiOz were considered concurrently with the reported ones. In most cases photocatalytic reactions on semiconductors are the redox reactions. So of special interest was to study the regularities of reactions of interfacial transfer of photoexcited electron by the pulse photolysis and luminescence quenching methods. Many interesting phenomena were found while studying the model photocatalytic reactions by the method of stationary photolysis, i.e., under the conditions of real photocatalysis. The authors hope that results of their studies will be of interest to the experts in both nanotechnologies and photocatalysis. 2.2. Synthesis of CdS Colloids Colloidal solutions of semiconductor particles are of great interest mainly as photocatalysts of various processes. Of special interest are the so-called Q-particles of semiconductors (for CdS with the size 2R c 50 A), some of their properties being considerably different from
36
Kinetic Peculiarities of Photocatalytic Reactions
those of bulk semiconductors. However, the lack of simple and reliable methods of reproducible preparation of semiconductor nanoparticles restrains the comprehensive experimental investigation of their physico-chemical properties. Unfortunately, there are no complete and consistent quantitative theories describing the growth of colloidal particles in the preparation process; generally, only semiquantitative concepts are used which assume the size of colloidal particles to depend mainly on the number of their growth sites at the initial moment of the colloid formation. It is well known [ 11 that high supersaturation of the parent solution and restraint of the forming particles growth rate are needed to obtain the highly dispersed difficultly soluble substances. Typically, these conditions are reached via mixing a rather highly concentrated solution of one substance with very diluted solution of another substance, with which the first one forms a difficultly soluble compound. High concentration of the first substance provides high superaturation of the solution and high rate of nuclei formation, while low con-centration of the second substance restrains the growth rate of the formed nuclei. In this chapter we consider the feasibility of easily controlled and reproducible synthesis of CdS colloids. To provide control and restrain the growth rate of the CdS nanoparticles, we used the complex salt of a colloid-forming component (Cd2+)instead of its diluted solution; actually, in this case the rate of colloid growth may be limited by the decay rate of the initial cadmium complex.
2 2 . 1 . Thermodynamic Calculation of Equilibrium Size of CdS Colloidal Particles in the Presence of Cadmium Components Upon the formation of CdS colloids, at the point when the colloidal particles stop their growth and the system reaches the quasiequilibrium state, the total amount of cadmium is distributed over CdS particles and the aqueous phase of the solution, where it appears both as activated Cd2+ion and as the compounds with complexing admixtures. One may assume the thermodynamic equilibrium is reached between the different possible forms of cadmium occurrence. Assuming the presence of only one complexing agent L, which participates in a stepwise complexing, we may describe the above equilibrium by a system of chemical equations:
[ Cd2+ + L
@
CdL2+
K
1
where ti are the chemical variables of the indicated reactions equal to the variation of the initial concentration of the reaction components divided by the stoichiometric coefficient; Ki and SP are the stability constants and solubility product of respective compounds. Note that solubility product of dispersed particle depends on its size [2]:
Chapter 2, D. V. Bavykin et al.
37
where SP, is the solubility product of bulk particle (for CdS at 298 K SP, = 1.6.10-28M2 [3]), CJ is the excessive surface energy (surface tension) at the solution-particle interface, i7 is the partial molar volume of the phase composing the particle (for CdS i7 = 30.1 cm3/mol), R is the particle radius, Rr is the universal gas constant, T is the temperature (hereinafter taken as 298 K). For the case when the cadmium salt, sulfide ion and complexing agent L are the initial compounds for colloid formation, the following set of equations is valid:
where [CdO], [Lo] and [So] are the initial concentrations of the cadmium salt, complexing agent and sulfide anion, N is the concentration of the colloidal particles. The above set of equations may be simplified by some assumptions not changing the equilibrium qualitatively. Indeed, let us assume that the whole complexing process is described by the addition of a single ligand and [LO], [So] >> [CdO]. In this case, the set of equations (2) reduces to a single equation interrelating the particles radius R and the stability constant K1of the complex: 2.a.B ~. .
4 - a
3
x
N a-
e
R3
+ [l + [L]. K,].-.eSP,
5
iiTp
= [CdO]
7
(3)
[SI
where [L] and [SI are the equilibrium concentrations that, under our assumptions, are approximately equal to the initial concentrations [Lo] and [So] of the ligand and sulfide anion. Specifying a reasonable N value and substituting in Eq. (3) the value CJ = 1.09 t 0.24 J/m2 estimated from the data of [4], one may determine by Eq. (3) the equilibrium size of the colloidal particle, which appears to be dependent on the stability constant of the complex. A detailed analysis of this calculation is reported elsewhere 121. The CJ value was estimated as follows. According to the data of [4], the range of solubility product (SPcdS) values was found from the condition of dissolving the cadmium sulfide particles of size 2R = 25 A by the added Na2EDTA and concurrent stability of these particles to alkalization: [CdEDTA]
Kl .[EDTA]
SP,
<-
[SI
< SPWOW2 [0Hl2 '
38
Kinetic Peculiarities of Photocatalytic Reactions
where K1 = 5.1016 M-I [9] is the stability constant of cadmium complex with the ethylenediaminetetracetate anion, SPcd(oHp= 2.l0l4 M3 [3] is the solubility product of cadmium hydroxide, [Cd EDTA], [S2-], [OH] and [EDTA] are the equilibrium concentrations of the appropriate reagents. Using the concentration values corresponding to experimental conditions in [4],we obtained 2.10” MZc SPcds < 2.lO-l’ M2. Then, by Eq. (l), the o range was estimated for the calculated SP value: 0.85 c o c 1.32 (J/m2). Note that, in general terms, the estimated o value is nearly an order of magnitude greater than the known typical o values for the aqueous phase - inorganic solid interface. Nevertheless, this experimentally determined value will be used in our further calculations. Unfortunately, the calculation scheme under consideration does not allow an a priori determination of the colloidal particles equilibrium concentration, or the nuclei concentration N, which is controlled most likely by the kinetics of the colloidal particles growth rather than by thermodynamics of the process. Nevertheless, even a simplified calculation scheme shows that (1) thermodynamically stable colloidal particles of the minimum acceptable sizes may form only in the presence of complexing agents characterized by rather low values of the stability constant; (2) in the presence of highly concentrated strong complexing agents, thermodynamically stable are only the colloids with the particle size no less than that specified by the value of the stability constants of substances present in the solution (see Fig. 2.1). So, this allows to answer whether it is possible for colloidal particles of a specified size to exist in the presence of a certain amount of complexing agent with the known stability constant.
Fig. 2.1. The calculated dependence of the upper limit of stability constant K1making possible the existence of CdS particle of radius R. Calculation was made for various CT values at the fixed values of [Lo]= [So] = 5.10” M,[CdO] = 5.104 M.o (J/mz)= 0.2 (1); 0.6(2), 1.1 (3); 2.0(4).
Chapter 2, D. V. Bavykin et a1
39
2.2.2. Method of CdS Preparation Typically, CdS colloids were prepared by the following method. 10 ml of solution with M CdC12, 2-10” M stabilizing surface active substance, and the required amount of complexing admixtures was prepared in a 100 ml vessel at room temperature. 10 ml of 2.lo3 M Na2S was added to the solution under vigorous stirring. The color of the produced colloidal CdS solution depended on.the type and amount of admixtures and varied from yellow to colorless. The produced colloidal solutions were stable, in some cases for more than six months. Fig. 2.2 shows the adsorption spectra of a colloidal CdS solution prepared by the above method in the presence of cadmium complexones of various nature. The position of the colloids adsorption band indicates that the equilibrium size of the colloidal particles decreases as the stability constant of the complex increases. This may relate to the fact that it is precisely the decay rate of the cadmium complex that determines the number of nuclei N and, hence, the size of the forming particles. This is supported by the fact that with the fixed initial (before the addition of the sulfide anion and after the addition of the ligand) concentration of activated Cd2+ (lg[CdL]/[Cd2’]) = const and [Cd’] = const), for complexones of various nature, the sizes of colloidal particles differ: the stronger the initial complex, the smaller the particle size. DI
I
. 1.8
- 1.6 . 1.4 . 1.2 . 1.0 .0.8 .0.6
. 0.4
- 0.2 300
400
Fig. 2.2. Adsorption spectra of colloidal CdS (5.10-4M) obtained at the addition of thioglycerol TG to 10” M solution of CdC12 to provide concentrations: 1) 5.10’ M, 2) 5 1 0 3 M, 3) 2S.10-4 M, 4) 2.5.10” M. Room temperature. [So] = 2.10” M.
In terms of the preparation, some organic sulfur-containing compounds appeared to be the best complexing admixtures allowing to control the size of CdS colloidal particles. Note that the colloids, produced in the presence of sulfur-containing compounds, exhibit a much higher stability compared to those prepared in the presence of ligands which contain no sulfur. It seems interesting that the cadmium sulfide colloids produced by the addition of thioglycerol (see Fig. 2.2) can be dried in air to form solid films capable of further reversible dissolving. In this case, the size of colloidal particles after drying and dissolving remains approximately constant.
40
Kinetic Peculiarities of Photocatalytic Reactions
2.3. Reaction of Interfacial Electron Transfer as the First Stage of Redox Photocatalytic Processes 2.3.1. Photobleaching Relaxation of Colloids Shifting the edge of optical adsorption to the ultraviolet region at loading an excess electron on the nanoparticles is a well known manifestation of the quantum size effect in the nanoparticle optical characteristics. This results in shift of the absorption spectrum near the edge of absorption. When the excess electron emerges at the colloidal particle due to its illumination, the effect is called photobleaching. There are several explanations of this effect in literature. Photogenerated auxiliary current carriers in an ultradispersed semiconductor particle generate large local electric fields. In principle, the presence of an electric field inside the ultradispersed semiconductor particle should increase the exciton energy [5], which shifts the absorption band to the ultraviolet region. On the other hand, it was shown in [6] that in the presence of an surface trapped electron-hole pair, the exciton energy shifts to the cuedn region only by ca. 15 meV, with the oscillation strength decreasing by 90% for CdS clusters with 2R = 50 A. The Burstein-Moss model [7, 8, 91 presents the third explanation of shifting the absorption spectrum of ultradisperse semiconductors to the ultraviolet region. According to this model, when an excess electron occupies a lower vacant state in the conduction band, a greater energy is needed for excitation of the next electron to a higher state (band filling). This leads to shifting the band edge of semiconductor adsorption to the ultraviolet region. In this case, a change of optical density AD is taken to be proportional to the concentration of excess electrons n [lo]. However, a more thorough consideration [ l l ] shows that the relation between the observed signal AD and the electrons amount n may be more complicated and vary with experimental conditions. Indeed, in a general case, the amplitude of the observed signal is determined by the dependence of the level densities on the level energies, by the dependence of the semiconductor absorption coefficient on the light wavelength (or frequency), by the number of trapped electrons. The conventional relationships applied to the CdS absorption properties are as follows:
p ( E ) = B(E - E,
y2
I
where a is the absorption coefficient at frequency o,oh is the threshold frequency, p(E) is the density of the energy level in the conduction band, E is the level energy, E, is the energy of the conduction band bottom. According to [7], a change in oh caused by the electron trapping in concentration n comprises
Here y is a coefficient which depends on the effective electron mass in the conduction band.
Chapter 2, D.V. Bavykin et a1.
41
By the above equations one may readily find a relation between the concentration n of the excess electrons and the change in the sample absorption which reflects the variation of a at the oh changing: n-
In our case, the measurements were done mainly at the colloid adsorption edge. In this region, for CdS it seems more correctly to approximate a(o)by a linear function
and thus Aa- A ' . y . nu3- PAD, n
- AD)^^ .
(2.6)
Relation (2.6) corresponds to the case when the shape of the absorption spectrum near the bandedge is determined by the presence of size distribution in the region of variation bandgap with varying particle size. In this case, relation (2.3) may be taken as valid. This is true for the case when the long wavelength bandedge of absorption is caused by a wide distribution of particle sizes. If the a(o)dependence in the region of absorption edge is caused by dope absorption, a relation between AD and n may be more complicated since the function p(E) for the doped levels will differ from (2.2), and relation between AD and n will differ from (2.4) and (2.6). In our opinion, in a general case, the question for relation between the AD and n can be solved only via direct experiment. For this purpose, we studied concurrently two series of samples with the same CdS colloid solutions, 2R = 10 nm, stabilized by PAA an L-c steine, as an electron donor. The first series of samples contained methylviologen MV" as an electron acceptor, the second one contained no MV''. All the samples in these experiments were saturated by the argon to remove dissolved oxygen. The number of trapped and then transfered to methylviologen electrons was identified with the amount of reduced methylviologen. Fig. 2.3 presents a correlation between the electron concentration in the sample and the absorption change just after a flash, with the flash energy being varied. It follows from this correlation that n (AD)x,where x 0.8. Note, that it will be shown below any of x value in the range of 0.8-1.5 makes practically no changes in the shape of the kinetic curves and in the resulting conclusions. Fig. 2.4 shows the typical kinetic curves of the electron concentration in the samples calculated at x = 1.5. Here one may see initial (immediately after the flash) changes of optical density ADo of the samples of 1=1 cm thickness. Fig. 2.5 shows the same kinetic curves (dots) for the sample with 2R-10 nm obtained at a higher flash energy (ADo (1 = 1 cm) 0.057) and calculated for = 0.8 and x = 1.5. The kinetic curves exhibit fast decay at the initial part and very long cctailw after that. Figs. 2.6 and 2.7 present the relative rate W of electron concentration decay vs. the relative concentration of electrons in the sample calculated at different x values for the kinetic curves from Fig. 2.4. For CdS colloid with 2R 10 nm, the dependence of W on n is fitted by a function:
-
-
-
x
-
Kinetic Peculiarities of Photocatalytic Reactions
42
W- n regardless of
. ,S.
x, where q is an experimental constant. I
Fig. 2.3. Correlation between the electron concentration n in the sample and the absorption change just after a flash AD. The n was calculated fromparallel experiments using MV=5. 10-3M as electron acceptor. The nf and ADf is the electron concentration and differential absorption at maximum energy of pump light. The pump light is at h<400nm using glass cut-off filter UFS-6, intensity of pumps is varied. 2R-lOnm, [CdS] = 2.5. lo4 M, [L-cys] = 2.5.10-* M, [PAA] = M, pH=6. Cell length 1 is 1 sm. T = 2OoC.
In the CdS colloid with 2R c 5 nm, the rate at the initial part of the kinetic deviates from relation of type (2.7), however, at n/no c 0.4 ( x = 1.5)or n/no c 0.7 (x = 0.8), the electron concentration dependence of the rate is also fitted by an equation similar to (2.7). Fig. 2.8 shows the same dependencies for the kinetic curve from Fig. 2.5 calculated for x = 0.8. In this case,
w
I
fj. @'' n .
(2.8)
where 0, q' are experimental constants. The following equation of kinetic curve corresponds to relation of type (2.8):
n -=1--.
1
ln(1 + q'Wot) = 1- A.ln(l t Bet),
q'n,
n0
0 0
where Wo = p * e
.
In Fig. 2.5 the solid lines show the time dependence of n/no calculated by Eq. (2.9) with optimized coefficients A and B. These data indicate also that the x values are noncritical for the functional form of the kinetic curves described by equations of (2.7) and (2.8) type.
Chapter 2, D. V. Bavykin et al.
in I,"
43
L
I
aD'=2, 6.10" I
1.0
-
20
1,o
L
A
GO
00
loo
120(JJSl
2
0.4
0.8
1.2
1.6
20
2.4(mS
-
Fig. 2.4. Kinetic curves of electron transfer reaction for CdS colloid (1) - 2R 5nm, hobs = 450nm (2) - 2R IOnm, kobs = 470nm. [CdS] = 2.5. lo4 M, [L-cys] = 2.5.10-2 M, [PAA] = M, [O,] 10-4MpH 6. Cell length 1 is 1 sm. T = 2OoC. hex< 400nm
-
-
I
0.5
1.0
1.5
tlm/S)
-
Fig. 2.5. Kinetic curves of the interfacial electron transfer reaction from CdS 2R 10nm. The electron concentration was calculated using formula n = ADx,o - = 0.8, - x = 1.5.hobs = 470 nm. [CdS] = 2.5. lo4 M, [L-cys] = 2.5.10-'M, [PAA] = M, [02] 10-4M pH 6. Cell length 1 is 1 cm. T = 2OoC. hex<400 nm. Dots are the experimental data. Line is a fit by formula (2.9)
x
-
Note that kinetic curves obtained for the same colloid with 2R 10 nm, presented in Figs. 2.4 and 2.5, were recorded at varying the flash energy and accordingly ADO.The ADO difference exceeding an order of magnitude corresponds approximately to a similar change of the electron concentration in the sample. For the kinetic curves in Fig. 4, the electron
Kinetic Peculiarities of Photocatalytic Reactions
44
-
Fig. 2.6. The rate of the interfacial electron transfer reaction from CdS 2R 10 nm. as a function (1) - 2R of the electron concentration, which was calculated using formula n = 5nm, hob$= 450 nm, (2) - 2R 10 nm, hobs= 470 nm. [CdS] = 2.5. lo4 M, [L-cys] = 2.510M, [O,] lo4 M pH 6. Cell length 1 is 1 cm. T = 2OOC. hex< 400 nm M, [PAA] =
-
-
'
7 6
2 1
0.2
0.4
0.6 nlno
0.8
-
Fig. 2.7. The rate of the interfacial electron transfer reaction from CdS 2R 10 nm. as a function of the electron concentration, which was calculated using formula n = ADo.*.(1) - 2R 5 nm, hobs= 450 nm, (2) - 2R 10 nm, hobs= 470 nm. [CdS] = 2.5. lo4 M, [L-cys] = 2.5-10-* M, [O,] 104M pH 6. Cell length 1 is 1 cm. T = 2OoC.hex<400 nm M, [PAA] =
-
-
-
concentrations measured with MV2+appeared to be 2.6.10" M and 5.10-7M in the CdS colloids with 2R < 5 nm and 2R 10 nm respectively. According to these data, 27 excessive electrons are photoexcited per one CdS nanoparticle with 2R 10 nm, which provides the volume and surface concentrations of electrons in the particle 5.4-1019cm-3 and 8.7.10" cmS. In the colloids with 2R < 5 nm, the respective values are: 13.5 per the ~, cm-'. particle and 2.5.10m~ m - 2.1.10'3
-
-
Chapter 2, D.V. Bavykin et al.
45
5
-5 -4 -3
4
z3
5:!
-2 -1 -0 - -1
31 0 -1
02
0.4
9.6 n/no
0.8
5
1.0
-
Fig. 2.8. The rate of the interfacial electron transfer reaction from CdS 2R 10 nm. as a function of the electron concentration,which was calculated using formula n = ADo,’. 2R lOnm, hobs = 470 nm. [CdS] = 2.5. lo4 M, [L-cys] = 2.5.10-* M, [PAA] = M, [O,] 104M pH 6. Cell length 1 is 1 cm. T = 2OoC.he,<400 nm
-
-
2.3.2.Physical Causes Resulting in the Observable Kinetic Peculiarities of Photobleaching Relaxation The main reason of appearance of exponential multiplier eqnin Eqs. (2.7) and (2.8) for the rate of interfacial electron transfer from semiconductor particles to dissolved oxygen molecules or protons seems to the electric charge of these electrons. This causes considerable changes in the potential of the nanoparticles double layer, which, in turn, affects directly the rate of electron transfer [12-151. Actually, in a semiconductor particle with 2R-10 nm, 27 excess electrons would cause the following change of the double layer potential (Acp): Acp 27e/Cl 700-70 mV ,where c1 is the electric capacity of the particle double layer (c1 = cgc- S = (6-30)-IO-’’ F, where cgc- (2-20). IO4 F/cm’ are the typical values of the double layer capacity, S 3.10-” cm’ is the overall surface area of a particle). In the CdS nanoparticles with 2R < 5 nm, the change of the double layer potential can be even greater, according to the similar estimation. Remind, that kT at room temperature is only 25 mV. So, a several hundred millivolt variation in Acp can change the rate constant of the interfacial electron transfer by several orders of magnitude, since Acp directly enters the apparent activation energy of the reaction. It seems rather difficult to explain the appearance or disappearance of the electron concentration before exponent in expressions for reaction rate (2.7) and (2.8). As noted above, the reaction rate is directly proportional to the electron concentration (2.7) at low initial electron concentrations. At higher concentrations, the direct proportionality between the reaction rate and electron concentration disappears, and the reaction rate is described by Eq. (2.8). In the colloid with 2R e 5 nm (Figs. 2.6 and 2.7), at the starting part of the kinetic, the reaction rate depends on n more weakly than it is predicted by Eq. (2.7), while at n/no 0.5 it approaches the dependence similar to (2.7). Such switching occurs at the surface electron concentration cm-’. Note, for the colloid with 2R 10 nm, the initial surface electron concentration of 8.7.10’’ cm-’ < 1013cm-’ corresponds to the kinetic curve in Fig. 2.4, while n > l O I 3 cm-’ for all points of the kinetic curves presented in Figs. 2.5 and 2.8, since no 9.1013 cm-’.
-
-
-
-
-
-
Kinetic Peculiarities of Photocatalytic Reactions
46
The switching between dependencies (2.7) and (2.8) at the surface electron concentration ca. l O I 3 cm-’ may be explained as follows. According to the interpretation of the mass action law, the total probability of two particles A and B to enter the surface reaction is the product of the probability of particle B to appear in the reaction space S* around particle A, which equals CB.S* (where CBis the surface concentration of particles B), and the probability of this pair to react for time period t - P(t) [16]: P(t).C,
*s* *
(2.10)
At CB.S* > 1, the concentration of particles B is so high that particle B locates in the reaction space around particle A with probability ca. 1. At such large values of B concentration (CB> 1/S*), the reaction rate becomes independent of the B concentration. In our case, electrons play the role of particles B, while molecules of oxygen, protons or other added electron acceptor play the role of particles A. The absence of linear dependence of the reaction rate on the electron concentration (Eq. (2.8)) implies that for the O2 molecule or H+ ions located at any point of the nanoparticle surface, at the distance of electron transfer there are one or more excess electrons trapped by the surface. As noted above, the switching from dependence (2.7) to (2.8) occurs at n l O I 3 ern-', which corresponds to the cm2 reaction surface of the reaction between electron and electron acceptor. For this case the distance of electron transfer is 1.7 nm. The assumption on the electric charge effect of excess electrons on the rate constant of their interfacial transfer is supported by an evident similarity of these semiconductor colloidal systems with metal colloids, for which effect of the <> of electrons captured by the particle is well known and agrees with the <microelectrode theory,,. Moreover, lunetic curves similar to those we found for CdS colloids were observed previously for silver colloids in ref. [17], where the particles charge q was shown to decrease by the law
-
-
The formal bulk concentrations of electrons captured by CdS particles -lo2’ cm1.3are close to the electron concentration in a metal. In this case, the semiconductor degradation or its metallization is implied [7]. In [18], we generated such excess electrons in a semiconductor colloidal CdS nanoparticles coated with thioglycerol by a short (- 5 ps) flash of light. In this case, photogenerated holes were rapidly (for the time less than the resolution of our devices) transferred to thioglycerol molecules. Fig. 2.9 shows the photobleaching spectra of colloidal CdS prepared in an excess of the cadmium ions with the particles of variable size. Photobleaching spectra were obtained by approximation of kinetic curves to the zero time point. One can see that as the CdS particles decrease in size, the photobleaching spectrum of colloidal CdS shifts to the short wavelength region. This is because the location of the maximum of the semiconductor photobleaching spectrum corresponds rather accurately to the location of the edge of semiconductor colloid absorption. Due to the quantum size effect, the edge of the nanocolloid adsorption shifts to the ultraviolet region as the nanoparticle size decreases; thus, the photobleaching spectra are expected to behave in a similar manner.
47
Chapter 2, D.V. Bavykin et al.
400
420
440
460
460
500
520
540
Wavelength, nm
Fig. 2.9. Photobleaching spectra of a CdS colloid obtained in the excess of Cd2*ions with the particles of various sizes at the zero time point. [CdS] = lo4 M, [SDS] = 2.10-3 M, [TG] = 5.10-3M. Illumination at h < 360 nm (UFS-l), Cell length 1 is 10 cm, T = 2OoC.
1E-4
i 100
1000
10000
T ime, )IS
Fig. 2.10. Photobleaching relaxation kinetic curves at the bandedge of absorption spectrum of a CdS colloid obtained in the excess of CdZ+ions with the particles of various sizes. [CdS] = lo4 M, [DCH] = 2.10-3 M, [TG] = 5.10-3 M. Illumination at h < 360 nm (UFS-l),Cell length 1 is 10 cm, T = 2OoC.
48
Kinetic Peculiarities of Photocatalytic Reactions
The rate of the photobleaching relaxation of ultradispersed CdS, and hence the rate of the electron interfacial transfer from CdS to the surrounded media (finally, to protons yielding the hydrogen) appeared to depend on the size of the colloidal particles (see Fig. 2.10). The photobleaching relaxation rate increases as the size of the CdS semiconductor particles decreases. Such behavior may be caused by the increasing of reductive potential of photoexcited electron with decreasing size of semiconductor nanocolloids. In this case, according to the modern concepts of electron interfacial transfer reaction [19], the rate of electron transfer to the surrounded media should increase. 2.3.3. Effect of the CdS Adsorption Properties on the Kinetics of its Photo-bleaching Relaxation in the Presence of Various Electron Acceptors
To study the regularities of photoexcited electron relaxation in the reaction of the electron transfer by the method of flash photolysis in microsecond timescale, we had to change the electron acceptor concentration in a liquid phase. The ability of the acceptor molecules to adsorb at the surface of the semiconductor colloidal particle was found to determine the character of changes in the photobleaching relaxation kinetic curves.
0.01
0.00
0
1000
2000
3000
4000
Time, ps
Fig. 2.11. Photobleaching relaxation kinetic curves of colloidal CdS with excess of Cd2+ions at 470 nm wavelength after the addition of MV and PW12.The concentration of the added M; PW12:0,4.10-', 8.10-', acceptor along the arrows: MV: 0, 2.10-', 5.10-', 8.10-', 3.10-7M. [CdS] = lo4 M, [SDS] = 2.10-3 M, [TG] = 5.10-3M. Illumination at A c 360 nm (UFS-l), Cell length 1 is 10 cm., T = 2OoC.
Fig. 2.11 shows the photobleaching relaxation kinetic curves of colloidal CdS obtained in an excess of the cadmium ions (the positive surface charge of the colloid) at adding of two different electron acceptors. One may see that the addition of negatively
Chapter 2, D. V. Bavykin et al.
49
charged ions of PWlz phosphotungstic acid, which are well adsorbed on such colloids, results in a decrease of the initial degree of bleaching, with no changes in the photobleaching relaxation rate and the shape of the kinetic dependence. Such behavior may be explained by complete adsorption of the added PWlzat the surface of the CdS nanoparticles and its fast reduction (for the time less than the resolution of our equipment); the observed kinetics of the photobleaching relaxation is determined by the transfer of remaining electrons to oxygen or protons in surrounded media. The addition of positively charged methylviologen ions, which are poorly adsorbed on the colloid under consideration, changes the kinetic dependence of the photobleaching relaxation. The kinetic curves obtained do not obey the logarithmic law (Eq. (2.9)), and are expressed by an exponential dependence on time, (2.1 1) where kffis an effective rate constant of the photobleaching relaxation, and AD0 is photobleaching at the zero time point. In this case, bffincreaseslinearly with increasing the concentration of the added methylviologen (see Fig. 2.12). This evidences that the electron transfer to the molecules of added nonadsorbed methylviologen MV is limited by the diffusion of its molecules to the surface of colloidal CdS (at the experimental concentration of MV).
2,oxiO' 4 , o ~ i o " ~ , O X I O - ' 8 , o ~ i o " i , o ~ 1 0 " MV"
Fig. 2.12. The apparent rate constant of the interfacial electron transfer from the CdS particles prepared with the excess of Cd** ions as a function of the methyl-viologen concentration. M, [SDS] = 2 ~ l O M, - ~ [TG] = 5.10-3M. Illumination at h c 360 nm (UFS-l), [CdS] = Cell length 1 is 10 cm., T = 2OoC.
A kinetic study on the photobleaching relaxation at the addition of PWlz and MV to colloidal CdS prepared in an excess of the sulfide ions, when the colloidal particles have the negatively charged surface, shows the opposite results as compared to those obtained
50
Kinetic Peculiarities of Photocatalytic Reactions
for colloidal CdS prepared with the cadmium ions excess. Thus, the addition of methylviologen cations leads to a decrease in the initial degree of the colloidal CdS photobleaching, while the rate constant of the photobleaching relaxation does not change. However, the addition of PWlz anions up to a certain concentration causes no changes in the kinetics of photobleaching relaxation of colloidal CdS produced in the excess of sulfide ions. After the addition of a certain amount of PWlZrcolloid photobleaching disappears sharply. This seems to be the result of the effect of ccrecharging,, the surface of colloidal CdS, which is discussed below. Indeed, the negatively charged PWlz ions don't adsorb at the CdS surface rich in the sulfide ions. However, addition of PWlz may change pH of the solution and stimulate the adsorption of H+ ions at the surface of CdS particle, changing its charge. Changing of CdS surface charge would result in the adsorption of phosphotungstic acid and a rapid (for the time less than the resolution of our devices) transfer of photoexcited electrons to the PWlz anions. This leads to the fast decay of the ultradispersed CdS photobleaching.
-0,Ol 0'0°
F9t A
-0302
-0,03 0,oo
-0,Ol
a -0,02 -0,03 200
400
600
o'ooll
-0,Ol
-0,02
I
I
I
200
400
600
Time, ps Fig. 2.13. Photobleaching relaxation kinetic curves of colloidal CdS with an excess of CdZ+ions at various temperatures and at the addition of various electron acceptors: a) without acceptors; b) MV, lod M; c) PW12, 3.104 M. The temperature variation along the arrows: 20, 30,40, 50,60,and 70OC. [CdS] = lo4 M, [SDS] = 2.10-3M, [TG]= lo-' M.
To verify the effect of the ions adsorption on the regularities of photoexcitation relaxation, we studied the temperature effect on the kinetics of the ultradispersed CdS photobleaching relaxation at the addition of electron acceptors of various nature. Fig. 2.13 presents the kinetic curves of the colloidal CdS photobleaching relaxation prepared with an excess of cadmium ions at different temperatures and at the addition of different
Chapter 2, D.V. Bavykin et al.
51
electron acceptors (MV and PW12). One may see that without the addition of electron acceptors the temperature variation causes no changes in the photobleaching relaxation kinetics, while at the addition of both acceptors increasing temperature increases the rate of photobleaching relaxation. Two factors seem to determine such behavior. First, elevating the temperature may increase the rate constant of electron transfer to the acceptor. Second, at elevating the temperature, the amount of adsorbed thioglycerol may decrease faster than that of the adsorbed acceptor, which increases the fraction of the CdS surface occupied by the acceptor. Unfortunately, the actual value of the changes observed at temperature elevation is negligible compared to experimental errors, which makes difficult a quantitative analysis of the temperature effect on the kinetics of the photobleaching relaxation.
2.4. Luminescence Quenching of the CdS Colloids Studies on luminescence of CdS colloids provide useful knowledge on the energy and nature of recombination sites of charge carriers in the colloidal particles. The regularities of the colloid photoluminescence quenching provide the information on the dynamics of electrons and holes in semiconductor particles as well as on the kinetics of interfacial electron transfer. Of a particular interest are studies on the luminescence of colloidal solutions of the so-called Q-semiconductors, their properties depending on the size of semiconductor particles due to the quantum size effects. One may expect a very pronounced effect of the surface properties of semiconductor particle on its luminescence peculiarities. This relates to the fact that in semiconductor nanoparticles, the sites of trapping the current carriers locate mainly at the surface. In addition, a very high ratio of the particle surface area to its volume is typical for nanoparticles. In other words, in such particles, a considerable fraction of atoms locates at the surface. E.g., the 20 8, CdS particles contain ca. 170 sulfur and cadmium atoms, 110 of them (67%) locating at the surface. In this connection, the luminescence properties of such particles are to be strongly depending on the nature and contents of adsorbates and on the structure and composition of the particle surface. This could explain numerous contradictions in the literature on luminescence properties of ultradispersed semiconductors. Actually, the composition and the state of the semiconductor nanocolloid surface depend strongly on the particle prehistory: like the preparation method and the way of various admixtures adding during the particle synthesis. Luminescence measurements performed by various authors use the semiconductors prepared by various methods. Nevertheless, some regularities in the semiconductor colloid luminescence are common for all authors. In semiconductor nanoparticles, the luminescence spectrum shifts to the ultraviolet region as the particle size decreases. This phenomenon is a manifestation of the quantum size effect in the nanoparticles luminescence properties. That is, with decreasing the size of nanocolloids, an energy increase is observed not only between the bottom of conductivity band and the top of valence band, but also between various sites of radiative recombination of trapped carriers. This effect was observed with the following semiconductor nanocolloids: CdS [20], CdSe [21], Cd3As2 [22], InP [23], and ZnO 1243. Generally, several luminescence bands are distinguished in the spectra of semiconductor nanocolloids. First, comparatively narrow luminescences band with the energy of several meV less than the band gap [25,26]. This band is commonly observed at low temperatures.
52
Kinetic Peculiarities of Photocatalytic Reactions
Such short wavelength luminescence of ultradispersed semiconductors is usually assigned to the radiative recombination of delocalized exciton [27, 281 or free (not captured) electrons and holes. The short wavelength band has been studied most thoroughly on isolated nanocrystals at low temperatures (ca. 20 K) in ref. [29]. It was shown that, without the broadening produced by a distribution in particles size, this short wavelength band is split into the quartet of bands. The main band (with the highest intensity), two IOWfrequency satellites (the main one and the first harmonic component from longitudinal optical phonons), and a satellite in the high-energy range (biexciton). The properties of these luminescence bands depend only slightly on the characteristics of the colloidal particle surface. Studies on the luminescence of ultradispersed semiconductors at room temperature in various media (solutions, matrices, films) reveal a group of broad long wavelength luminescence bands which are commonly assigned to radiative recombination of the photoexcited charge carriers captured at deep traps [30, 311. The nature of these recombination sites is still unclear. For example, numerous hypotheses concerning CdS are reported in the literature [32]: cadmium vacancy, sulfur vacancy, interstitial cadmium, interstitial sulfur. The only item in explanation of the nature of these long wavelength luminescence bands all the authors agree on is that the sites of radiative recombination of photoexcited charge carriers locate at the surface of colloidal particle [33]. This chapter presents results of the studies on photoluminescence spectra of CdS Q-colloids in the size range of 2R = 20-100 8,as well as on the luminescence quenching of Q-CdS by various quenchers under variation of the colloidal particle sizes. 2.4. I . Regularities of Luminescence Quenching of Colloidal CdS Particles
In most cases, quenching of luminescence of CdS colloids is determined by the reactions of interfacial electron transfer involving either electron or hole. So, study of this process is a convenient method for establishing the regularities of key steps of redox photocatalytic reactions over CdS. In addition, using of various luminescence quenchers (anions, cations, and neutral molecules) allows to reveal the nature of electron capture sites at the CdS surface. Analysis of the dependencies of CdS colloidal solutions luminescence spectra (with varying particle size) on the methylviologen concentration (Fig. 2.14) allows to conclude that the efficiency of luminescence quenching raises with increasing wavelength and particle size. For a quantitative interpretation of the results obtained, let us consider the processes that may occur with excited electron in a colloidal semiconductor particle under stationary illumination. First process is the bulk recombination of nonequilibrium charge carriers (radiative and nonradiative). Since in our experiments the luminescence is quenched completely, one may assume that a fraction of radiative recombination in the particle volume is negligible. Thus, we may take into account only the nonradiative recombination in the bulk proceeding with the rate kb,n*V*e.h, where kb,nis the rate constant of nonradiative recombination in the bulk, V is the volume of colloidal particle, e and h are the stationary concentrations of electrons and holes in colloidal particle. Note also that, according to [34], it seems reasonable to assume the constant concentration of electrons and holes inside the particle along the radius under stationary illumination for the particles of the size order of several hundred angstroms.
Chapter 2, D. V. Bavykin et al.
I.
53
3,008
1 " " " '
0,025
2.006
3,004
3.002
1.000 1.0012
).OW9
LOW6
).OW3
4 b
5 b
6 b
7b
&
).OMx) 6b
760
Wmdengttr, nm
Fig. 2.14. Luminescence spectra of the Q-CdS colloidal solutions with 2R = 20 A, hexc=345 nm (a), 2R = 23 A, he, = 375 nm (b), 2R = 34 A, he,, = 380 nm (c), 2R = 44 A, he, = 400 nm (d) after the addition of the methylviologen. The concentrations of the added MV along the arrows: a) 0, lo", 5.10", lo4, 2.104, 5.104, 2.10-3 M; b) 0, 5.10-7, lo", 2.10", 5.10-6, lo-', 2.10-7, 5.10-', lo4, 2.10", 2.lO-', 4.10-', 8.10-', 1.5.104, 2.104, 3.104, 4.104 M; c) 0, 2.10-7, 5.10-7, lo", 5.10" M. [CdS] = 5.104 M, T = 23°C. 5.10", lo-' M; d) 0,
Another way of disappearance of nonequilibrium charge carriers is their recombination at the particle surface (radiative with the rate constant k,,r,and nonradiative Of basic importance is the question of whether the surface with the rate constant k,,,-,). recombination sites are the sites of the quencher adsorption. In other words, is the quencher adsorption able to result in disappearance of the surface recombination sites. With positive answer, the expression for the surface recombination rate should be written as (ks,T+ k,,,).S.(l - O,).e.h, where S is the particle surface area, and 0,is the surface fraction occupied by the quencher (electron acceptor). Otherwise, the latter multiplier (1 - 0,) should be excluded. Further we will consider the both cases (1 and 2), compare them with experimental data and choose the case providing a better description of the phenomena observed. The third way of the nonequilibrium electron disappearance is their transfer to the quencher molecules. The expression for the rate of quenching could be written as (b,Red&O,-e + b,[email protected]),where kQ,Redand kQ,oxare the heterogeneous rate constants of
Kinetic Peculiarities of Photocatalytic Reactions
54
electron transfer to acceptor and hole transfer to electron donor, respectively; @d is the surface fraction occupied by the electron donor. Under stationary illumination, the rate of appearance of nonequilibrium charge carriers is equal to the rate of their disappearance. Thus, for case 1, the following equation holds true: (2.12) while for case 2 (2.13) Here, R is the particle radius, is the intensity of the incident radiation, a is the light absorption coefficient of the CdS particle. In addition, under stationary conditions, the rate of the electron disappearance is equal to the rate of the hole disappearance: kQ,Red
- 0, e = kQ,ox' 0 , h . *
(2.14)
Equation (2.14) allows to exclude the hole concentration h from Eqs. (2.12) and (2.13).The resulting square equations are solved with respect to the electron concentration e. Therewith, for case 1 we obtain: a'
I+-+ e=
b'(1-0,)
0,
-1
0, 9
(a'+b'.(l- 0,)) * kQ,Red
For case 2 it holds true
e=
/+:-
1
, where
a ' kQ,Red
Using the approach similar to the deduction of the Stern-Folmer's equation 1351, one may easily show that quantum yield CP of the luminescence under stationary regime for case 1 may be expressed as
Chapter 2, D. V. Bavykin et al.
55
(2.15)
while for case 2, as
(2.16) where 0 ' is the quantum yield of the luminescence with no quencher addition. Dependence of 0,on the quencher concentration in a solution may be determined, to a first approximation, by the Langmuir adsorption isotherm: 0, =
A .K , * C, l+K,.N, '
(2.17)
where CQ is the concentration of a quencher in the solution, Kads is the adsorption equilibrium constant of a quencher at the CdS surface. An approximation of the experimental data by the expressions obtained by substitution of Eq.(2.17) into Eqs. (2.15) and (2.16) are presented in Fig. 2.15.
5 -
4 -
Y
3 -
4 .
0 .
e
2 -
1 -
00
1
2
3
4
5
Concentration of MV, IO' M
Fig. 2.15. The Stern-Folmer's presentation of the dependence of the luminescence quantum yields of the Q-CdS colloidal solutions with 2R = 23 A on the methylviologen concentration at h,, = 440 nm. Dots are the experimental data, curve (a) is fitted by Eq. (2.15) with a' = 0.11 1, b' = 2.973, Kads= lo5;curve 0) is fitted by Eq. (2.16) with a' = 0.111, Kads= 3 . 6 2 ~ 1 0curve ~ ; (c) is fitted by Eq.(19) with a' = 6.36, Kads= 6.07.103.The CdS concentration is 5104 M, T = 23OC.
One may see that within the limits of experimental error, the both expressions describe adequately the quencher concentration dependencies of the luminescence quantum yields.
Kinetic Peculiarities of Photocatalytic Reactions
56
However, the expressions derived from Eqs. (2.15) and (2.16) are rather cumbersome, inconvenient for experimental application, and low informative. So further we use a simplified expression to describe the concentration dependencies of the luminescence derived from expressions (2.12) and (2.13) under the assumption that the recombination rates are linear in the current carriers concentration, Le., the addends kb,n*h. e and (ksr,r+ksr,n). h. e are substituted by kb,,’. e and (ksr,r’+k,,,,’). e, respectively. Linear recombination in the current concentration does often occur at a minor level of generation of the carriers, when recombination rate is limited by the rate of capturing one of nonequilibrium carriers by recombination site [7]. The appropriate expressions are as follows. For case 1: R
cpo --
cp
kQp e d + kb,n 1=
’
7
R kbrn * - + ck$-,r -t k$-,n 1
(2.18) ’K adsaCQ
For case 2:
One may see that in Eq. (2.18) @‘/adepends linearly on the concentration of the quencher, which is inconsistent with the experiment (see Fig. 2.15). Therefore, in further description of experimental data we use Eq. (2.19), which corresponds to the case when the recombination sites are not identical to the adsorption sites. In this case, A and Kadsare the varied <
57
Chapter 2, D.V. Bavykin et al.
8
I
.
1
.
I
.
I
.
1
4
,
n
A
K
.5. '
5,
- 14
4
7 ,
A
A K
tmd420nm 74763 t 02848 58506 t 3934 tmd480nm 1 0 1 8 6 t 1180 51060 f 9994
.
n
A
,
K A
'
1- 1 '
A K
tmd440nm 63.514 t 06128 60664 t 1345 !xnd500nm 22323t1357 31484 t 3069
A
.
- 12 - IO
7
e e
0 .
4 -
1 P
3-
I
2-
B
1 -
0:
-8
I-:
-6
A 0
a
:: 1 : ip A
.
-4
- a n
12
-2
-0
7
tmd560nrn
10 A
8
bndMOnm
I
1 14E5 t 4214E4
7
i
0 .
e
6
4
J
T
2
0
Concentration of MV, M Fig. 2.16. The Stern-Folmer's presentation of the dependence of the luminescence quantum yields of the Q-CdS colloidal solutions with the particles of different size on the methylviologen concentration. The size of the particles: a) 2R t: 20 A, b) 2R -- 23 A, c) 2R = 44 A. The CdS concentration is 5.104 M, T = 23'C. Dimensionality of Kadsis M-'.
Indeed, the greater the deepness of recombination site energy level, the less the difference in the energies of the given level and the quencher molecule level, hence, under the assumption of activation electron transfer occurring in the inverted region, the lower the activation energy of such transfer. However the dependence of the parameter A on the deepness of the energy of the trapped carriers could be related with the dependence of the surface recombination constants (k,,, + k,,,) on the deepness of recombination sites. According to the ref. [39] the radiative lifetime of trapped carriers is longer for the deeper traps. This could be interpreted as tunneling recombination of trapped carriers [40].The deep electron trap is the electron trap with a large distance to the nearest occupied hole trap (due to the Coulomb interaction) and the shallow trap is the electron trap with small
58
Next Page
Kinetic Peculiarities of Photocatalytic Reactions 2K,A5 0 4 5 40
35
30
25
20
14
12
10
8
6
0,02
0,04
0,03
~
0,o
1/2R,A-'
Fig. 2.17. The calculated constant Kadsof methylviologen adsorption on the CdS colloids vs. the size of the colloidal particles. The Kadsvalue is obtained by approximation of experimental data on the luminescence quenching to Eq.(2.19). Dimensionality of Kadsis M-'.
- 3
- 2
- 1
/
-0- 2R = 20A -0- 2R 23A E
-A- 2R -V-
2R = 44A
- -1
V
0,2
0,4
-0
= 34A
0,6
0,8
1,0
1,2
The deepness of the trapped carriers, eV
Fig. 18. The parameter A as a function of the deepness (in energetic scale) of the level of
w.
recombination sites for CdS particles. Parameter A is calculated from (2.19).A is proportional to the ratio between the rate of interfacial electron transfer to the rate of trapped carriers recombination. The location of energy level of recombination site is measured from the bottom of conductivity band.
Previous Page
Chapter 2, D. V. Bavykin et al.
59
distance to the nearest occupied hole trap. The tunneling motion to the larger distance requires longer time. Note that this reasoning on the dependence of parameter A on the location depth of the recombination site has only a qualitative character. The actual samples of CdS have always a particle size distribution. Both the quencher adsorption constant and the luminescence emission spectrum depend on the particle size. This introduces a certain error into the A value obtained from fitting by the Eq. (2.19). The error should be particularly great for the ‘deep’ and ‘shallow’ energy levels of the surface recombination site. 2.4.2. Effect of Surface Properties of Ultradispersed CdS on Regularities of Luminescence Quenching
Note that there are alternative considerations of luminescence quenching not taking into account the adsorption of quencher molecules on the surface of nanoparticles [41, 421. They evaluate nonlinear equations, which are the same to equation (2.19). However below we show that in aqueous solutions it is necessary include the surface properties of CdS and adsorption of electron acceptor molecules. As noted in the previous section, the adsorption constant of the quencher molecules at the surface of a luminescent colloidal particle explicitly enters the expression for the luminescence quenching. Therefore, the adsorption-desorption equilibriums established in a colloidal solution should have a considerable effect on the regularities of the luminescence quenching and, hence, on the regularities of interfacial electron transfer. 2.4.2.a. Luminescence Quenching of Colloidal CdS by Quenchers of Various Nature
Methylviologen is a bication and thus it is well adsorbed at the surface of negatively charged particles. Now consider the regularities of the luminescence quenching by other kinds of electron acceptors, anions, exemplified by heteropolyacids. At the addition of phosphotungstic acid PW12to colloidal CdS, a step dependence of the luminescence intensity on the amount of introduced PW12 is observed: up to the PW12 concentration 2.10-4M, the luminescence intensity is practically constant and drops sharply as the concentration increases up to 4.104 M (Fig. 2.19a). Therewith, the pH values change approximately from 9 to 6. However, when H2S04 is being added gradually to the colloid with Z104 M PW12 (Fig. 2.19b), the relative intensity of luminescence also decreases with the same changes in pH of the solution. This result may be explained by
Kinetic Peculiarities of Photocatalytic Reactions
60
4 400
500
600
700
800
4&3
660
i
Wavelength, n m
Fig. 2.19. Changes in the luminescence spectrum of Q-CdS colloidal solutions under the addition of PWI2. (a) Successive addition to give the concentrations: 2-10-5,lo4, 2.104, 4.104 M; (b) successive addition of PW12to the point of 2.104 M, followed by the addition of H2SO4 to give 2.5.104, 5.104, 7.5.104 M. CdS concentration is 5104 M, A,,, = 350 nm, T = 23'C.
effect of the electron acceptor adsorption on the rate of photocatalytic redox processes over semiconductor colloids. We carried out a series of experiments on the luminescence quenching of colloidal CdS by quenchers of different nature (see Table 2.1). Table 2.1. Adsorption constants Kadsfor quenchers of various nature and parameters of electron transfer A obtained from FQ.(2.19).In the experiment, [CdS] = 5.104 M, 2R = 23 A, T = 23'C, A,, = 360 nm, A,,,, = 500 nm Quencher
Kads
(M-')
A
65 f 60
15 f 12
8200 f 6900
0.15 f 0.02
1200 f 300
11 f 2
2000 f 500
8.5 f 1.3
23 f 250
0.35 f 3
One may see that cations (MV2+,BV2+)are characterized by high values of both the adsorption constant and the rate of the luminescence quenching, while anions (K3[Cr(SCN),]) are characterized by a low value of the adsorption constant and a high
Chapter 2, D. V. Bavykin et al.
61
value of parameter A. The substances which are inactive in the reactions of electron transfer (Et4NCl and Pyr), possess low values of parameter A. PWlzhas low values A.Kad,
To prove the above statement on the determining effect of electric charge of both the CdS colloidal particle surface and the quencher molecule on the adsorption of these molecules from aqueous solution, we have modified the surface of colloidal CdS during its preparation. The most efficient method of such modification consists in changing the surface charge of the colloidal particle via the preparation of nonstoichiometric colloid. In this case, the surface charge is determined by the charge of excessive ion (either S-' or Cd"). The experimental values of adsorption constant Kadsfor the quencher molecule at the surface of ultradispersed colloidal CdS prepared by two different methods as well as the values of the electron transfer parameter A are listed in Table 2.2. One may see that the efficient luminescence quenching occurs only when the surface of the colloidal particle and the quencher molecule have opposite charges. If the charges of the colloidal particle surface and quencher molecule have the same sign, it is difficult to determine a precise value of adsorption constant Kadsand electron transfer parameter A due to a significant calculation error caused by minor changes of relative intensity of the luminescence at the addition of the quencher. In other words, when the charges of the luminescent particle surface and quencher ion have the same sign, no significant luminescence quenching occurs. It seems interesting that at the luminescence quenching of colloidal CdS by the PWlz ions, the dependence of adsorption constant on the size of colloidal particle is lacking (see Table 2.2). On the other hand, as noted in Fig. 2.17, the methylviologen adsorption constant at the surface of colloidal CdS depends on the particle size. This difference may relate to the different size of methylviologen and phosphotungstic acid molecules. Actually, the size of the phosphotungstic acid molecules in aqueous solution is 5-6 A [MI, which is comparable with the size of a colloidal particle. Therefore, the suggested equation (2.19) reflects adequately a variation in the properties of luminescence quenchers and describes well the experimental data as a whole. In addition, the data suggest that for the colloids obtained with an excess of the sulfide ions, the quencher adsorption site is represented most likely by the negatively charged surface sulfur atoms, on which the cations adsorb readily while the anions adsorb poorly. Actually, in this case, the dominant surface defects are either the vacancies of the cadmium ions, or
Kinetic Peculiarities of Photocatalytic Reactions
62
the interstitial sulfur. According to the literature [32,45], the both kinds of surface defects may serve as recombination sites of photoexcited charge carriers. However, while the surface cadmium vacancy may be potentially substituted by the quencher cations, the latter are expected to have no effect on the interstitial sulfur. Under the assumption that the quencher adsorption does not result in disappearance of recombination sites of charge carriers, the interstitial sulfur seems to be a more probable recombination site. Similarly, one may assume the surface positively charged cadmium ions, which are well adsorbing the anionic electron acceptors, to be the surface recombination sites for the CdS colloids prepared in the excess of cadmium ions. Table 2.2.
The quencher adsorption constant Kads and parameter of electron transfer A obtained via Eq. (2.19) from the data on the luminescence quenching of colloidal CdS by two methods (method 1 -with the excess of S2- ions, method 2 - with the excess of Cd2+ions) by quenchers of various nature Quencher
*
Method Of CdS preparation
hobs(nm) 2 ~ ( h A
Kads
A. Kads
MVZ+
1 (NSC)
480
20
10
510
MV2+
1 (NSC)
500
23
22
3100
MV2+
1 (NSC)
550
26
7
11000
MV2+
1 (NSC)
560
34
7
1o5
MV2+
1 (NSC)
560
44
1
%lo5
MV2+
2 (PSC)
550
21.5
<<1
MV2+
2 (PSC)
550
24
<<1
MV2+
2 (PSC)
550
26.5
<<1
MV2+
2 (PSC)
550
32
<<1
PW12-*
1 (NSC)
550
23
<<1
pwl;-
2(PSC)
500
21
0.88
18000
pwl;-
2(PSC)
500
22.5
0.8
18000
pw12-
2(PSC)
550
26.5
1.5
16000
Under experimental conditions (pH 5-7, T = 23OC), phosphotungstic acid in aqueous solution is mainly in a completely dissociated form (PW120403-)[43].NSC means negative surface charge; PSC means positive surface charge.
2.4.3. Effect of Excitation Wavelength on the Luminescence Spectrum of CdS/Cu,S Substitution of the lattice cadmium ions in a CdS colloidal particle by the ions of another metal is often accompanied by the formation of the so-called <> particles CdS/Me,S,. Such particles are readily produced via the substitution of cadmium ions by other ions if only their sulfide are less soluble compared to the cadmium sulfide. Our studies on the luminescence properties of such particles and regularities of their luminescence
Chapter 2, D. V. Bavykin et al.
63
quenching were reported in detail in [46]. Here we note only a remarkable feature of such nanosystems, the effect of excitation wavelength of CdS/Cu,S colloids on their luminescence spectrum.
0,Oj
, 400
,
, 500
,
~
600
,
, 700
,
,I 800
Wavelergtb nm
Fig. 2.20. Luminescence spectra of colloidal CdS/Cu,S obtained at two different excitation wavelengths 1) 380 nm, 2) 410 nm. The luminescence intensity of the both spectra is divided by the actual optical density of the sample at the excitation wavelengths. [CdS] = 5104 M, [TG] = 2.5-10-*M, [SDS] = lo-' M, [CUI= 1% mole, 2R = 26.5 A, T = 20'C.
Fig. 2.20 shows the luminescence spectra of colloidal CdS/Cu,S (2R = 26.5 A) measured at two different excitation wavelengths. To normalize the luminescence spectra to the amount of light absorbed the luminescence intensity of each spectrum is divided by the optical density of the sample at the appropriate excitation wavelength. The luminescence spectra of the colloid are seen to be shifting to the long wavelength region as the wavelength of the excitation light increases. According to the spectra of the colloidal CdS/Cu,S absorption (2R = 26.5 A, Cu = 1% mole), at the 410 nm wavelength, ca. 10% of the light is absorbed by the Cu,S phase, while at 380 nm wavelength, only ca. 4% of the light is absorbed by the Cu,S phase. One can suppose that the shift of the luminescence spectrum of CdS/Cu,S results from the difference between quantities of the light quanta absorbed by Cu,S at different excitation wavelengths and the shift of luminescence spectra due to the Cu,S phase luminescence. However this assumption contradicts with that the pure Cu,S colloids has no luminescence spectrum at all. Note, that under an increase of the temperature the difference between the two spectra at two excitation wavelengths is reduced (see Fig. 2.21). This means that the processes of the excited electron capture on the surface recombination sites is activated. This observation seems to prove the existence of the <
Kinetic Peculiarities of Photocatalytic Reactions
64
r 00400-04
I 5M) 600 700 800
p1
Wuvelength, nm Fig. 2.21. Normalized luminescence spectra of colloidal CdSICu,S at different excitation wavelengths and different temperatures: 1) 350 nm, 2) 400 nm. [CdS] = 5-104 M, [TG] = 2.5.10-2 M, [SDS] = lO-’M, CU= 1%, 2R -- 26.5 A.
2.4.4. Conclusion
The above data reveal evident distinctions in the efficiency of the quenching the different luminescence bands of the same colloid, as well as distinctions in the efficiency of the luminescence quenching of Q-colloids of various size. The regularities of the CdS luminescence quenching by the methylviologen can be described adequately by the expressions similar to the Stern-Folmer’s equation, taking into account the shape of the isotherm adsorption for the quencher molecules at the surface of the CdS particles. We suggested an expression for the efficiency of such quenching under the assumption that the recombination rate depends linearly on the concentration of nonequilibrium current carriers in a semiconductor particle. The validity of this assumption is supported by a good agreement between the experimental and calculated data. Moreover, the calculated values of bS and parameter A have a clear physical meaning which is consistent with the previous concepts on the nature of quenching of the semiconductor particle excitation. This hypothesis is additionally supported by the data of [47],indicating that the kinetics of photoluminescence decay of CdS colloids at a flash excitation can be also described under the assumption of linear (with respect to the concentration of current carriers) rate of the current carriers recombination.
Chapter 2, D. V. Bavykin et al.
65
Equation (2.19) allows to determine the particle size dependence of Kad,. Good agreement of Eq. (2.19) to the experiment suggests that for colloidal CdS particles, the sites of recombination and adsorption are not identical, at least, for the studied quenchers. Of importance is the value of parameter A, which is equal to the ratio between the probability of electron transfer to a quencher and the probability of disappearance in a recombination process. The value of this parameter depends on the energy of the luminescence quantum for the semiconductor colloids. Therewith, the higher the energy of recombination site level, the lower the value of parameter A, which formally corresponds to either decreasing rate of electron transfer or increasing the rate of recombination of trapped carriers. The ability of a luminescence quencher molecule to adsorb on the surface of ultradispersed colloidal semiconductor in aqueous solutions was shown to depend mainly on the charge of the colloidal particle surface and the charge of quencher ions. The luminescence spectrum of colloidal CdS/Cu,S depends on the wavelength of excitation light and shifts to the caed>>region as the excitation light wavelength increases.
2.5. Kinetics of Photocatalytic Reactions at Stationary Illumination 2.5.1. Photoreduction of Methyl Orange on CdS Colloids in Deep Conversion
The reduction of organic dye methyl orange (MO) over CdS colloids with the particles size d = 2R 5 nm has appeared to be a convenient reaction for detail studying the kinetics of photocatalytic processes. This dye is readily reducible with no dimers formation. The MO adsorption spectrum in the pH range of 10-12, practically does not change. This allows simplifying the interpretation of the experiments on redox transformations of MO and considering the reaction of photostimulated reduction of MO as a model one. Sodium sulfide was used as the electron donor in the reaction under study. In aqueous solution with pH 10-12, the sulfide ion occurs mainly as HS- anions [48]. Photochemical processes in the presence of CdS colloids are caused by the generation of nonequilibrium holes (h) and electrons (e) in CdS:
-
CdS
hv
>
h+e,
In the course of the photoinduced process, HS- ion is oxidized by holes to give elementary sulfur HS-+2h-+ S+H+. Further, the formed sulfur is removed from the surface of CdS colloidal particle due to reaction with the sulfite ion added [49]
s +so:- + s,o:As result, in this reaction, two CdS nonequilibrium electrons reduce methyl orange to a noncolored MOHz leucoform MO+ 2e + 2H'
+
MOH, ,
66
Kinetic Peculiarities of Photocatalytic Reactions
which has no adsorption at the wavelength h = 500 nm. The protonation of the MO reduced form occurs very fast immediately after the electron transfer [50]. In the literature there is no description of a stable one-electron reduced form of MO. So further we will assume that under the action of light, the photocatalyst provides transfer of one electron only, while the two-electron MO reduction and the two-electron HS- oxidation occur as an overall processes in the rate not limiting steps. Therefore, in the context of the assumptions indicated, the overall reaction of the process under study may be written as follows: MO + HS-
+ ,9032- + H+
hv, CdS
>
MOH, +S,Oi-
The estimates of the molar concentration of colloidal particles in the solution and the amount of light quanta adsorbed by each colloidal particle in a unit time are essential for the further discussion. Taking the volume of the colloidal particle with 2R 5 nm as 100 nm3, CdS density 4.8 g/cm3, CdS molar weight 144.4 g/mol, at the CdS concentration in solution [CdS] = 0.4.104 M, molar concentration K of colloidal particles can be estimated as K = 2.2.10-' M. Under the indicated conditions, the amount v of light quanta adsorbed by colloidal particle per second at the intensity of irradiating light b 3 mW/cm2= 0.8.10-*Einstein-s' 'ern-' may be easily found as:
-
-
(2.20) Here, AI4 is the amount of light quanta adsorbed by the CdS colloid, and K 4 . L is the amount of colloidal particles entering the light beam. 2.5.1.a. Kinetic Peculiarities of Photocatalytic Processes on Ultradisperse CdS Colloids at Stationary Illumination
A typical experimentally observed kinetic curve of the MO reduction is composed of three sections: section AB with the fast initial decay of the MO concentration (see Fig. 2.22), a linear section BC extending to the point of complete decomposition of the dye, and final section CH with fast decrease in the process rate. This kinetic dependence takes a more clear, typical stepped configuration at turning from coordinates <> to coordinates ccquantum yield cp - current MO concentration>>using the formulas for evaluation of quantum yield: (2.21) where S = 1 cmz is the cross section of incident light, V = 2 ml is the volume of sample, L = 1 cm is the length of optical path, = 6 . 103 1. M ' . sm-' absorption coefficient, AI intensity of absorbed by CdS light. Fig. 2.23 presents such kinetic
67
Chapter 2, D. V. Bavykin et al.
t(min) Fig. 2.22. A typical kinetic curve of MO photoreduction in the presence of CdS colloid in coordinates (optical density D, time t) in the system with Na2S as electron donor. [CdS] = 0.4.10-3 M, [PAA] = 2.4~10-~ M, [Na2SIo= M, [Na2S0310= M, [MOlo = M, IO = 0.8.10-'Einstein.s-'.cm-'. Optical density was measured at the wavelength h = 500 nm.
00
0,2
0.4
0,6
0.8
1,0
1,2
1,4
1,6
[MO] .l04(M) Fig. 2.23. Initial quantum yield cpo vs. the MO initial concentration (dotted line), and typical
kinetic curves (solid lines) obtained at different values of the MO initial concentration in the system with Na2S as electron donor. [CdS] = 0.4.10"M, [PAA] = 2.4.10-3 M, M, Io = 0.8.10-'Einstein.s-'.cm-2. [Na2SIo= l o 2 M, [Na2S0310= M. In these coordinates, the change in the curve for initial concentration [ M o l o = M to [MO] = 0.6.10-4 M is methyl orange concentration from the initial [Mol0 = accompanied by a two-fold change of the reaction quantum yield from the initial cpo = 0.03 to the stationary cpst = 0.015 value, then by the quantum yield stabilization at the level qst = 0.015 for the MO concentrations ranging from 0.6.10-4M to 0.1.10'4M , and finally, by a sharp decrease in the system quantum yield to cp = 0 at the MO dye concentration decreasing from 0.1.10-4M to 0.
Kinetic Peculiarities of Photocatalytic Reactions
68
In our study, for all applied intensities of the irradiating light in the range 0.2.10-8 Einstein.s-'.cmp2 c Io c 1.6.10-8 Einstein.s-'.cm-2, the experimental-ly obtained initial reaction rate WOwas found to be proportional to the intensity IOof irradiating light (Fig. 2.24), Le., the initial quantum yield of the reaction cpo is independent of this intensity.
Fig. 2.24. Initial rate Wo of methyl orange reduction vs. intensity Io of irradiating light with h = 365 nm. [CdS] = 0.4.10-3 M, [PAA] = 2.4.10-3 M, [NazSlo= 10" M, [NazSO3I0= M, [Mol0 = M.
1
I
2yo
o,o+
0,O
"
"
0,2
0,4
"
0,6
"
0,8
"
'
1,o
[MO] 104(M)
Fig. 2.25. Initial quantum yield cpo vs. the MO concentration in the system with KzC2O4 as electron M,IO = 0.8.10-8 donor. [CdS] = 0.4.10"M, [PAA] = 2.4.103M, [KzCz04]0= Einstein. s-'.cm-'.
A comparison of the experimentally obtained dependence of the initial quantum yield cpo on the MO concentration (Fig. 2.23) with the shape of the MO adsorption isotherm over an aqueous suspension reveals a qualitative similarity in the form of indicated curves.
Chapter 2, D. V. Bavykin et al.
69
A characteristic increase of q o after the horizontal section is apparently more pronounced when the potassium oxalate K2C204is used as the electron donor instead of the sulfide ions (Fig. 2.25). A qualitative similarity of the adsorption isotherms and the MO concentration dependence on the initial quantum yield indicates that the adsorbed dye molecules take part in the reaction. Note that all kinetic curves attain the same value of the stationary quantum yield qstregardless of the initial MO concentration (Fig. 2.23). Therewith, the q s ~ qratio o depends on the nature of polymeric surfactant used for stabilization of CdS colloid. With PAA, this ratio equals ca. 0.5, and 0.6 with PVS. An interesting feature of the reaction under study is the occurrence of some slow relaxation processes. If the illumination of the sample is terminated when its kinetics reaches the stationary section characterized by qst (at point 1 in Fig. 2.26), and recommenced after an hour, the process quantum yield exceeds qstimmediately after resumption of illumination, but later tends to qstagain.
0,O
0,2
0,4
0,6
0,8
1,0
[MO]. l04(M)
Fig. 2.26. Effect of illumination termination on the brutto-reactionquantum yield cp. Point 1 corresponds to the termination of sample illumination;point 2 corresponds to the resumption of illumination after 1 hour. [CdS] = 0.4.10” M, [PAA] = 2.4.10-3M, [Na2SIo= lo-’ M, [Na2S0& = lo-’ M, I, = 0.8.10-* Einstein.s-’.cm-’.
The experiments with noncolloidal CdS suspensions revealed that the latter feature of the systems under study as well as the establishing of the stationary quantum yields of the process qstare observed only in the presence of a macromolecular surfactant. One may assume that slow conformational transformations in the surfactant macromolecules may affect considerably the adsorption-desorption equilibria at the surface of the semiconductor particles under consideration and thus affect the course of redox processes generated by these particles under the action of light. We present below an attempt in a semiquantitative description of the observed processes. 2.5.1.b. Semiquantitative Description of the Kinetics of Photocatalytic Processes on CdS Colloids in Terms of Adsorption-DesorptionProcesses in the System
For the description of photocatalytic action of semiconductor particles in the reactions of electron transfer from a certain donor to acceptor, it is convenient to recognize three states of photocatalyst particle:
70
Kinetic Peculiarities of Photocatalytic Reactions
state 1, ground state. state 2, photoexcited state. 0 state 3, state with trapped carriers at the surface of the CdS particle Reaction (2.20) accompanied the 3+1 states transition, its probability described by the quantum yield cp of the reaction. Consider the possible ways of the transition from state 3 to state 1 in order to determine the dependence of the quantum yield on the ions concentration of acceptor Aad (in our case, MOad)or donor Dad(in our case, HS,:) at the surface of a colloidal particle. Let us assume the state 3 could decay by three channels: 1. by the recombination of an excessive electron and excessive hole with the first order effective rate constant k, (hereinafter the dimensionality of effective rate constants of type k is s-'); 2. by the initial oxidation of the donor by an excessive hole with the rate constant kD, followed by the reduction of either the MO molecule by an excessive electron with the rate constant klM0,or the side acceptor (e.g., 02,HzO, H') with the rate constant VS; 3. by the initial oxidation of MO and the side acceptor with characteristic rate constants kMo and k,, respectively, followed by oxidation of electron donor by hole with the rate constant k'D. In this case, the kinetic of the colloidal particle transition from state 3 to state 1 is described by the following set of equations (2.22): 0 0
dP = ph-, * (kMo+ k,) - ph * kb dt (2.22)
dpD - ph-,
dt
k, + ph * kb
In these equations, Ph-e, Pe, and Ph are the probabilities of detecting, respectively, the electron-hole pair, the hole, or the electron in the semiconductor particle at time t; PMOand ODare the probabilities of detecting, respectively, the reduced acceptor and oxidized donor at the surface of the colloidal particle at the same time moment; kx = k, + kD + kMo + k,. At the initial time t = 0, it may be assumed that Ph-e = 1, Pe = 0, Ph = 0, PMO= oi and PD = 0. At termination of the process generated by one light quantum, i.e., at t >> 10s, one may observe Phe = 0, Pe = 0, Ph = 0, PMO = APMo,and PD = APD. The equations could be solved via successive calculation of the values: Pe-h(t) from the first equation, Pe(t) and Ph(t) from the second and third equations, PMO(t) and PD(t)from the forth and fifth ones. The integration of probability PMO(t) over t from 0 to 00 gives the value of theoretical surface quantum yield cpT, which is determined as the ratio of the
Chapter 2, D. V. Bavykin et al.
71
probability to detect the reduced MO ion to the initial number of electron-hole pairs at the surface of the semiconductor particle. This value is specified by expression (2.23)
Here, the first term of the sum reflects the probability of the MO reduction by the second channel of the process (see above), while and the second term reflects the probability of the MO reduction by the third channel. Since, as indicated above, qint= 1, expression (2.23) reflects the total theoretical quantum yield of the process under study on a single semiconductor particle. To find the theoretical quantum yield of the process for the whole sample, one should average cpT over all colloid particles. Under the assumption that this averaging does not change considerably the form of function (2.23) and instead of effective constants k one may use their averaged values over all colloidal particles, we find:
1 rp' =-.
1+ "
K,,MO,,
-1
(2.24)
Here, "KMO, "KD, "KMo, and ''K'Mo are the heterogeneous rate constants of appropriate reactions. Expression (24) takes into account that the observed quantum yield cp of the reaction is rather low (cp = 0.03); this allows to take kz = k,; the reaction rate is taken to be proportional to the concentration of adsorbed reagents:
Expression (2.24) gives at least a qualitative description of the experimentally observed dependence of the initial quantum yield cpo of reaction on the concentration of MO acceptor. Indeed, the number (concentration) of dye ions MOad adsorbed at the surface of a colloidal particle before the light illumination, is a function of the MO concentration in a the solution and follows the dye adsorption isotherm at the same surface. Since MOad increases with increasing MO concentration, at surpassing of a certain concentration [MO]', the ratio kj/"K,,MO, can appear to be much less that unity. If the Dadvalue is fixed, the second term of sum (24) becomes constant, and hence (2.25) P
Thus, theoretical quantum yield cpoT at [Molo > [MO]', with an accuracy of an additive constant and a numerical multiplier, follows actually the MO adsorption isotherm over the CdS colloid.
72
Kinetic Peculiarities of Photocatalytic Reactions
Note, that at [MO] = 0 the qoT value also equals zero. Therefore, the [Molo dependencies of qo,qoT,and MOadare expected to be qualitatively similar. 2.5.l.c. Analysis of Kinetic Regularities of the System under Study
As noted above, upon oxidation of the hydrosulfide anion, elementary sulfur forms at the surface of the colloidal particle:
HS; +2h+S,
+Ht
A fraction of elementary sulfur atoms is easily accessible for the sulfite anions from the solution and is rapidly removed from the surface in the course of reaction:
s, +so;-+s,o,2followed by fast adsorption of a new hydrosulfide anion. However, another fraction of the sulfur atoms may appear to be <> (<>) by one or several links of the stabilizing polymer, e.g., PAA. Such <> of the formed sulfur may consist in the surfactant adsorption over the sulfur atoms s a d from near the surface space. The latter process creates steric obstacles for the SO3*- anion to approach the Sad particle, Le., for the removal of s a d from the semiconductor surface. Actually, it means a passivation (decrease) of the <<working>> (active) surface of the colloidal particle, which finally may decrease the reaction quantum yield. If the number of PAA blocking segments is limited and the surface area that can be <> by all these segments being less than the working surface area of colloidal particle, a decrease in the quantum yield at the initial part of kinetic dependence will occur to a certain qst.The value of qstis determined by some stationary state of the working surface not <> by PAA. This agrees with experimental data. Naturally, the fraction of the surface area blocked by the polymer surfactant, and hence the value qSl/qodepend on the nature of the surfactant used. A slow dark relaxation can be naturally explained within the suggested model by the unblocking the sulfur atoms shaded from interaction with SO;- during the some <> process, followed by sulfur removal from the colloidal particle surface in the form of thiosulfate anion and adsorption of a new hydrosulfide anion. A large size of the surfactant polymer molecules (e.g., the PAA molecule consists of ca. lo5 monomers units) and interaction between the segments of these molecules decrease considerably the conformational mobility of both the whole molecule and single segment of its chain. This may increase the characteristic time of sulfur atoms and other particles adsorption-desorption to 1 hour and more, i.e., to the time typical for the observed dark relaxation process. Consider Fig. 2.27 for a semiquantitative description of the initial section of the experimental kinetic curve. One can see from this figure that the kinetic dependencies in coordinates
Chapter 2, D.V. Bavykin et a1.
73
adsorption layer at the colloidal particle surface are a function of the MO concentration in the sample and do not depend explicitly on the time required for attaining this MO concentration. Obviously, at the given light intensity from the studied range of I~,,c I, c I, and fixed initial concentrations of reagents [reaglo, the content of adsorption layer at the colloidal particle surface during MO reduction depends on the current MO concentration in the sample [MO] in two cases. First, in the case of fast (as compared to the rate of the overall reaction) desorption of the reaction products from the colloidal particle surface, when further decrease of the typical time of establishing the adsorption-desorption equilibrium has no effect on the cp([reagIo, [MO]) dependence. Second, in the case of slow (as compared to the rate of overall reaction) desorption of the reagents, when a further increase of the time of adsorption-desorption process also has no effect on the cp( [pearlo, [MO]) dependence.
Fig. 2.27. The rate of methyl orange reduction vs. the MO current concentration at various values of light intensity Io in the system with NazS as electron donor. [CdS] = 0.4.10-3M, [PAA] M. 1) IO = 0.2.10-' = 2.4.10'3M, [Na2S]o = 10" M, [NazS0310= lo-' M, [Mol0 = Einstein.s-'.cm-'; 2) Io = 0.6.10-' Einstein.s-'.cm-'; 3) I, = 1.6.10' Einstein.s".cm".
In the first case, the chemical reaction practically does not disturb the adsorptiondesorption equilibrium between the colloidal particle adsorption layer and the liquid phase. The composition of the adsorption layer during the MO reduction may be expressed in terms of the reagents initial concentrations and the current [MO] value. In the second case, the reaction volume of the colloidal particle is limited by its adsorption layer, which composition during MO reduction may be found from the known [reagIo and [MO]. In the intermediate case, the current composition of the colloidal particle adsorption layer and hence the cp([pearIo, [MO]) value will explicitly depend on Io. The aforesaid may be schematically presented as a time scale hierarchy, where the characteristic times of overall T,,,, and establishment of the adsorption-desorption equilibrium T,d&s are indicated. Let us estimate these times for the reaction system under study.
Kinetic Peculiarities of Photocatalytic Reactions
74
Define T , as ~ ~the~ time, when a dramatic change in the adsorption layer composition takes place in the absence of the adsorption-desorption processes. Therewith, easily estimated by dividing the total number of the reagent ions (MOad, HS, ) adsorbed at the colloidal particle surface by the total rate W = cp-v of photoinduced process. Tahng into account that cp 0.03, and the v value for the range of intensity IAn 0.2. einst. s-’sm-2 c Io
-
-
-
-
~ f- z ~ ~ ~
Estimate also the time
PAA
of establishing the adsorption-desorption
equilibrium for low-molecular components of the system: MO, HS-, and SO3’-. The limiting adsorption rate Wad of noncharged particles on an arbitrary colloidal particle with a low-molecular component, e.g., with the MO ions, is given by the known formula:
where rk, rMO, XK,and XMO are the sizes and diffusion coefficients of colloidal particle and MO ions, respectively. Assuming rk + rMO = rk = 2 nm, XK + XMO = XMO = l o 5 cm2/s and [MO] = l o 4 M, we find that per 1 s each colloidal particle undergoes more than lo’ collisions with the MO ions. Assuming the adhesion coefficient at the collision equal to unity and taking into account that at the stationary state about 100 MO anions are adsorbed at the colloidal particle surface, we find the characteristic time of changing the adsorption layer compo-sition of the colloidal particle surface due to adsorption: Tad 100/105 10” s. Note, that the variation in the concentration of the MO at the initial part of the kinetic from [MO] = 1 0 4 Mto [MO] = 0.8.10m4M- A[MO] = 0.2.10-4Mcorresponds to the MO amount initially adsorbed at the colloidal particle surface: [K]-MO,d = 2.lO-’ M . 100 = 2-10” M. In this case, one may see from the experimental data obtained (see, e.g., Fig. 2.23) that at the initial step of the reaction, the reaction quantum yield decreases only two-fold. Therefore, the desorption rate of the low-molecular components can not be much lower than the rate of the reaction proceeding on the CdS particle. Since the reaction quantum yield does not change with the varying the light intensity, the desorption of the lowmolecular components is a much faster than the redox transformations at the photocatalyst surface. Thus, the hierarchy of characteristic times in the system under study may be expressed as
-
TF- 10-3s, TC-<< TE- 70s <::T
-
- 500s << z;Fdes - 4000s
Thus, the adsorption layer at the colloidal particle surface is represented by the segments (areas) of at least two types, which differ significantly in the time of establishing the adsorption-desorption equilibrium with respect to the characteristic time of photocatalytic overall reaction. A part of the surface of particle is not coated with the PAA molecules, and for this part
Chapter 2, D. V. Bavykin et al.
75
At this surface fraction, the photocatalytic reaction does not disturb the adsorptiondesorption equilibrium. Another part of the surface adsorption layer is created during the reaction due to blocking the forming elementary sulfur by PAA segments. For this surface fraction -z:Zdes-4000s >> 7:, -500s (case 2). The adsorption layer composition at the segments of various types in the studied range of reaction time will be again determined by the initial concentrations of reagents [reaglo and the current concentration [MO] of methyl orange, and not depend explicitly on the light intensity and hence the time of attaining the specified MO concentration. Thus, the cp dependence on [MO] will remain unchanged for any light intensity I,, in the range of Ifin e I,, c I-, which is supported by similarity of the lunetic curves in Fig. 2.27. Within the model of a semiconductor particle with two surfaces of different (< and >) adsorption layer types, one may quantitatively describe the initial part of kinetic curves. Denote the total surface area of the colloidal particle as Q, and the surface area of the colloidal particle blocked by PAA as QpAA. Let us call the colloidal particle surface area free of PAA as the <<working>> surface and denote it as Qw = Q - QPAA. Denote the area of the <<working>> surface before the illumination as Q", Some special experiments have revealed that the MOHz and SZO3'- products formed at the photocatalytic reaction have only a slight influence on the adsorption layer composition free of PAA. Indeed, a complete restoration of EM010 = l o 4 M after which the illumination was switched off, keeping the sample in darkness for 3 hours, addition of the oxidized MO form to give the initial concentration [Mol0 = M, and further restoration under similar conditions showed that the repeated restoration process changed the form of the MO concentration dependence of the reaction quantum yield only slightly. This supports the conclusion that the reaction products compete weakly with the starting reagents for the reaction sites at the CdS surface. In this case, the amount of reagent ions adsorbed at the surface of a single colloidal particle in the course of the photocatalytic process MOldnads,Dldnadswill be determined by both the current concentrations of reagents and the value of the photocatalyst particle surface not blocked by PAA. Assuming the values MOkinads, Dldnadsand the effective rate constant k,' to be proportional to Q , we find: (2.26)
where
$!n
is the restoration rate of side acceptor at the working surface area Q,.
Substitution of MOkinads, Dkinads and
$!n
into (2.24) gives (2.27)
where [reag] is the current concentration of reagents, [reag]~ is the reagents concentration at the reaction onset, Qw([reag]) is the current area of working surface, QwO([reag])is the working surface area of the colloidal particle at the moment of the reaction onset, cpT([reag])is the reaction quantum yield for the arbitrary degrees of the reagent conversion,
76
Kinetic Peculiarities of Photocatalytic Reactions
cpTo([reag]) is the reaction quantum yield for the working surface area of colloidal particle SZwo and the current concentration of the reagents. The expression derived demonstrates that a decrease in the quantum yield during the photocatalytic reaction (i.e., at decreasing [reag]) which is caused by two cofactors. The first one reflects a decay of the quantum yield due to descreasing the total concentration of the reagents (mainly of MO) in the sample, while the second one reflects a decrease of the colloidal particle <<working>> surface area due to a partial blocking of the surface by the PAA macromolecules. The dependence of the initial quantum yield of the photocatalytic reaction on the MO concentration was measured experimentally (see Fig. 2.23) and may be used as a function of cpoT in expression (2.27). Consider the variation of SZ,,, during the reaction. During the reaction, elementary sulfur forms at the surface of the colloidal particle. Denote the fraction of the newly formed sulfur atoms which will be <
the formed sulfur atoms, k,[S032-] is the rate of where kpAA(PAA)'is the rate of <> the reaction between sulfur and S032-anion. Here (PAA)' is the average number, calculated with respect to a single colloidal particle, of the PAA segments capable of blocking the forming sulfur, kPAAis the rate constant of the blocking. At high SO?-concentrations, the rate of blocking the forming sulfur with surfactant molecules may appear to be much lower as compared to the rate of the reacting with S032-.In this case, the fraction of the blocked sulfur atoms is small and proportional to the (PAA)' value:
At the steady state the number of the sulfur atoms produced in the reaction course equals the number of the reduced MO molecules: K.dSad = K*(-dMOad)= -dMO . Hence. dSad = -d(MO/K)=d([MO]/[K]) ,
(2.28)
where MO and K are the total number of the MO molecules and colloidal particles in the system, [MO] and [K] are the MO and colloidal particles concentrations averaged over the sample volume, the differentially small values dSad and dMOad are recognized as the changes in the number of the adsorbed sulfur atoms and MO ions per one colloidal particle, on the average. Taking the surface area blocked by the PAA molecules at closing>>one adsorbed sulfur atom equal to o,it can be easily found that at the formation of dSadsulfur atoms, the working surface area SZ, of the colloidal particle is decreased by the value
Chapter 2, D.V. Bavykin et al.
77 (2.29)
Here, R t is a certain stationary value of the working area of colloidal surface. Integration of (2.29) gives: (2.30) Thus, the working surface area of the colloidal particle decays exponentially with the current MO concentration in the sample decreasing from 52: to The result obtained agrees with the experiment. Consider, e.g., the kinetic dependence of the quantum yield for the system with [Mol0 = 1 0 4 Mand [NazS]o= lo-’ M, [Na2S0310= lo-’ M (Fig. 2.23). Since the value cpo([MO]) depends slightly on the MO concentration in the range 0.5.10-4 M e [MO] e l o 4 M (see Fig. 2.23), according to expression (2.27), the variation in the quantum yield at this section is caused by changing the working surface area of the colloidal particle and, according to (2.30), depends exponentially on the difference ([Molo - [MO]). This conclusion is supported by an appropriate analysis of the shape of the initial section of the curve shown in Fig. 2.23. Thus, the data of above study reveal a pronounced effect of adsorption-desorption processes and, in particular, their dynamic characteristics, on the occurrence of photocatalytic redox reactions, which proceed at stationary illumination over colloidal semiconductors, stabilized by surfactant molecules. The observed linear dependence of the rate of the studied model photocatalytic reaction on the light intensity indicates an apparent single-quantum nature of photocatalytic processes sensitized by colloidal semiconductors. Similarity of the dependence of the reaction quantum yield on the methylviologen concentration in the solution and the MO adsorption isotherm on the CdS aqueous suspension demonstrates the dominating effect of adsorbed MO molecules on the process rate. The observed dynamic regularities of the process at continuous illumination of the system allow assumption on existence of at least two segments at the colloidal particle surface: the surface segment which is free of stabilizing surfactant molecules and is characterized by a short (e< 70 s) time of establishing the adsorption-desorption equilibrium with the liquid phase; and the surface segment which is blocked by the surfactant molecules and is characterized by large (ca. 1 hour) time of establishing the adsorption-desorption equilibriums with the liquid phase. In the course of photocatalytic reaction, the surfactant molecules provide, most likely, a reversible bloclung of photocatalyst surface, which considerably decreases the process quantum yield.
at
I
2.5.2. Photoreduction of Methylviologen and Phosphotungstic Acid on CdS Colloids, the
Eflect of Surface Charge
The surface properties of ultradisperse semiconductor CdS which are determined in during its preparation, were shown to make a decisive influence on the regularities of interfacial transfer of photoexcited electron. In this section, we consider the effect of surface properties of ultradisperse CdS on the regularities of photoreduction of various substances under stationary illumination of CdS.
Kinetic Peculiarities of Photocatalytic Reactions
78
As the reducible reagents, we have chose the methylviologen and phosphotungstic acid (cationic and anionic reagents), which were used in the experiments on the luminescence quenching and photoexcitation relaxation. The adsorption ability of the reagents also appeared to have a pronounced effect on the feasibility of the photocatalytic reaction on ultradisperse CdS. In particular, on colloidal CdS prepared in an excess of the sulfide ions (i.e. the particles with the negatively charged surface), photoreduction of methylviologen bication proceeds efficiently (see Fig. 2.28), with no reduction of phosphotungstic acid anions. Conversely, on colloidal CdS produced in the excess of cadmium ions (with positively charged surface), photoreduction of methylviologen cations practically does not occur, while photoreduction of phosphotungstic acid anions proceeds with a considerable rate (see Fig. 2.29). According to the adsorption spectra of reduced PWlz [51] (see Fig. 2.29), initially an one electron reduced form of PWI2 .-is formed, and then further reduction of PWlz b; two electrons occurs (the absorption spectrum shifts to the longer wavelength region).
D
Wwelength, nm
Fig. 2.28. Adsorption spectra of MV photoreduced at CdS. The spectra were recorded in 10 min at 2OoC. Illumination with DRSh-500; UFS-2 and SZS-20 glass cut-off filters. [CdS] = 5.10-4M, [PAA] = l o 3 M, [TG] = 5.10-4M, [MV] = 2.5.10-4 M.
Since the surface properties of the colloid have a strong influence on the photoreduction kinetics, it seemed interesting to elucidate the effect of the added surface-active substances on the kinetic regularities of the reactions photosensitized by semiconductor colloid. The surfactant molecules are known to concentrate near the surface of the colloidal particle, so they may affect strongly the kinetics of photocatalytic reactions proceeding at the particles surface [52,53]. Fig. 2.30 presents the temperature dependencies of the initial rate of the MV photoreduction over colloidal CdS prepared at the excess of the sulfide ions. These dependencies were obtained at the addition of different amounts of PAA. One may see that both the initial rate and the observed activation energy of the methylviologen photoreduction do not depend, within the experimental error, on the concentration of
Chapter 2, D. V. Bavykin et al.
79
admixed PAA. In other words, the addition of the surfactant has no effect on the methylviologen ability to adsorb at the C d S surface. Probably, these data agree with the ideas that the admixed surfactant can not compete with thioglycerol in the adsorption at the surface of ultradisperse CdS.
'
I
'
,
500
400
'
,
'
600
,
700
'
,
'
800
Wwelength. nm
Fig. 2.29. Adsorption spectra of PWlz photoreduced at CdS. The spectra were recorded in 10 min at 2OoC. Illumination with DRSh-500; UFS-2 and SZS-20 glass cut-off filters. [CdS] = 5.10-4 M, [DCH] = 10" M, [TG] = 5.10" M, [PW1,] = 2.5*10'4M.
3,6-
3,3
3
Uiithwt PAA
E,=8.25+1.75 k J W
0
PAA3.10?
E,=9.75*1.47 k J m
A
PAA lo-?
E,=8.189.86 kJhrol
-
C
-J
3,O-
2,7
I
I
I
Fig. 2.30. Temperature dependence of the initial rate of methylviologen photoreduction over colloidal CdS with an excess of S2- at the addition of variable PAA amounts. [CdS] = 5.10-3 M, [TG] = 5.10-2M, [MV] = 2.5.104 M. Dimensionality of W is M.s-'.
80
Kinetic Peculiarities of Photocatalytic Reactions
The results of this section reveal that relaxation of photoexcited states of ultradispersed CdS in processes of interfacial charge transfer strongly depends on CdS particles adsorption properties and ability of substances, involved in redox transformations, to adsorb at the CdS surface. The adsorption properties of ultradispersed CdS can be varied during its preparation. 2.5.3. Effect of Excitation Light Wavelength on the Rate of Photocatalytic Reaction
It was found that quantum yield q(h) of the reaction of methyl orange MO reduction on CdS is monotonically increase with an increase of the wavelength of excitation light. A considerable monotonous growth of q(h) at h increasing and approaching the adsorption edge was observed also for other electron acceptors: chrysoidin and methylviologen. This fact can not be explained by the sensibilization by dye, since methylviologen initially has no adsorption band in the region under study. The observed phenomena may be caused, e.g., by increasing recombination rate of auxiliary charge carriers at their energy growth, and possible participation of nonthermalized <> electrons in the reaction of interfacial electron transfer.
2.6. Conclusion The data reported in the work indicate that CdS nanocolloids are very mobile systems with many interesting properties, which may be found in various time ranges when studying the photoprocesses. Such properties involve changing the equilibrium size of CdS particles at the addition to the solution of Cd2+complexones - electron donors like EDTA; the effect of nonequilibrium electron charge on the rate of interfacial electron transfer; a dramatic effect of the surface charge of colloidal particle on the efficiency of photoluminescence quenching (charge transfer); the dependence of the efficiency of photoluminescence quenching on the wavelength of emitted by colloidal particle light; the occurrence of slow relaxation processes in photocatalytic reactions on CdS colloids in the presence of surfactants in the solution; the effect of wavelength of excitation light on the photoreaction quantum yield. Interpretation of some of these regularities may be debatable. However, these regularities can be shown in the experiment, while rigorous description of the kinetics of photoprocesses on semiconductor nanoparticles is still a challenging task. Studying the photoprocesses on semiconductor nanoparticles remains a fruitful and very interesting field. REFERENCES 1 . Schukin E. D., Pertsov A. V. and Amelina E. A., Colloidal Chemistry, Vyschaya Shkola, Moscow (1992) (in Russian). 2. KarapetjantsM. H., Chemical thermodynamic, Nauka, Moscow, (1975) (in Russian). 3. Hand-book of chemistry, B. P. Niclolsky (Ed.), pp. 119-166, Khimija, Moscow (1964) (in
Russian). 4. Vucenilovic M.I., Vukelic N. and Rajh T. J. Photochem. and Photobiol., 42,No. 1, 157-161 (1988). 5. Bahnemann D.W., Kormann C. and Hoffmann M. R. J. Phys. Chem., 91,3789-3798 (1987).
Chapter 2, D. V. Bavykin et al.
81
6. Wang Y. and Herron N. J. Phys. Chem., 95,525-532 (1991). 7. Gurevich Yu. A. and Pleskov Yu. V., Photoelectrochemistry of Semiconductors, Nauka, Moscow (1983) (in Russian). 8. Moss T. S.Proc. Phys. SOC.B (London),67, 775 (1954). 9. Burstein B. E. Phys. Rev., 93, 632 (1954). 10. Grenn A. and Mills A. J. Photochem. Photobiol A: Chem., 64, 211-221 (1992). 1 1 . Savinov E. N., Photocatalysis of redox reactions in aqueous solutions with dispersed metals and semiconductors, D.Sc. Thesis, Novosibirsk (1993) (in Russian). 12. Dounghong D., Ramsden J. and Gratzel M. J. Am. Chem. SOC.,104,2977-3001 (1982). 13. Moser J. and Gratzel M. J. Am. Chem. SOC.,105,6547-6555 (1983). 14. Mulvaney P., Swayamcunathan V., Grieser F. and Meisel D. J. Phys. Chem., 92, 6732-6740 (1988). 15. Kamat P. V., Dimitrijevic N. M. and Nozik A. J. J. Phys. Chem., 93,2873-2875 (1989). 16. Vedeneev V . I., Lebedev Ya. S. and Entelis S. G., Lectures in Chemical Kinetics, MPTI, Dolgoprudny (1974) (in Russian). 17. Nanadovic M. T., Micic 0. I. and Adzic R. R. J. Chem. SOC.Far. Trans., 76, No. 4, 1065-1069 (1982). 18. Bavykin D. V., Savinov E. N. and Parmon V. N. J. Photochem and Photobiol A: Chem., 130, No. 1,57-61 (2000). 19. Siders P. and Marcus R. A. J. Am. Chem. Soc., 103,741-747 (1981). 20. Stramel R. D., Nakamura T. and Thomas J. K. Chem. Phys. Lett., 130, No. 5,423-425 (1986). 21. Blanton S.A., Degestani A., Lin P. C. and Guyot-Sionnest P. Chem. Phys. Lett., 229, No. 2,317322 (1994). 22. Fojtik A. and Weller H. Chem. Phys. Lett., 120, No. 6,552-554 (1985). 23. Micic 0. I., Cheong H.M., Fu H., Zunger A., Sprague J. R., Mascarenhas A. and Nozik A. J. J. Phys. Chem. B, 101,4904-4912 (1997). 24. Hotchandani S.and Kamat P. V. J. Phys. Chem., 96, 6834- (1992). 25. Anpo M. and Kubokawa Y. J. Phys. Chem., 88,5556-5560 (1984). 26. Hiramoto M., Hashimoto K. and Sakata T. Chem. Phys. Lett., 133, No. 5,440-444 (1987). 27. Wang Y., in: Photochemical Conversion and Storage of Solar Energy, E. Pelizzetti and M. Schiavello (Eds.),pp. 295-305, Kluwer, Dordrecht (1991). 28. Tian Y., Wu C. and Fendler J. H. J. Phys. Chem., 98,4913-4918 (1994). 29. Tittel J., Gohde W., Koberling F., Basche Th., Kornowski A,, Weller H. and Eychmuller A. J. Phys. Chem., 101, No. 2,313-316 (1997). 30. Becker W. G. and Bard A. J. J. Phys. Chem., 87,4888-4893 (1983). 31. Ferrer I. J. and Salvador P. Chem. Phys. Lett., 141, No. 5,399-404 (1987). 32. Aven M. and Prener J. S.,Physics and chemistry of II-VI Compounds, Amsterdam (1967). 33. Rossetti R. and Brus L. J. Phys. Chem., 86,4470-4472 (1982). 34. Gerischer H., Conditions for efJicientphotocatalytic activity of TiOzparticles, in: Photocatalytic Purification and Treatment of Water and Air, D. F. Ollis and H. Al-Ekabi (Eds.), Elsevier, Amsterdam, (1993). 35. Rabek J., Experimental methods in photochemistry and photophysics, Mir, MOSCOW (1985) (in Russian).
82
Kinetic Peculiarities of Photocatalytic Reactions
36. Rips I., Klafter J. and Jortner J. Solvation dynamics and solvent-controlled electron transfer, in: Photochemical Energy Convertion, J . R. Noms and D. Meisel (Eds.), Elsevier, Amsterdam, (1989). 37. Tachiya M. J. Phys. Chem., 97, 5911-5916 (1993). 38. Siders P. and Marcus R. A. J. Am. Chem. SOC.,103,748-752 (1981). 39. Chestnoy N., Harris T. D., Hull R. and Bms L. E. J. Phys. Chem., 90, 3393-3399 (1986). 40. Nemec P. and Maly P. J. Appl. Phys., 87, 3, 3342-3348 (2000). 41. Christy F. Landes, Markus Braun, and Mostafa A. El-Sayed, J. Phys. Chem. B, 105, 10554-10558 (2001). 42. Landes C., Burda C., Braun M. and El-Sayed M. A. J. Phys. Chem. B, 105, 2981-2986 (2001). 43. Kozhevnikov I. B., Acid and Basic catalysis, Novosibirsk State University (1991) (in Russian). 44. Harmalker P. S.and Pope M. J. Phys. Chem., 82, No. 26,2823-2825 (1978). 45. Ramsden J. J. and Gratzel M. J. Chem. SOC.,Faraday Trans., 80, 1,919-933 (1984). 46. Bavykin D. V., Savinov E. N. and Parmon V. N. Lungmuir, No. 6,4722-4727 (1999). 47. Gmzdkov Yu. A., Savinov E. N. and Parmon V. N. Chem. Phys., 7, No. 1, (1988). 48. Sharlo G., Methods of Analytical Chemistry, Khimiya, Moscow (1965) (in Russian). 49. Nekrasov B. V., Fundamentals of General Physics, Vol. 1-3, Khimiya, Moscow (1973) (in Russian). 50. Gruzdkov Yu. A., Ph.D. Thesis, Institute of Catalysis, Novosibirsk (1987) (in Russian). 51. Savinov E. N., Catalytic and photocatalytic properties of heteropolyacids of 12'h series in the reaction of hydrogen extraction from aqueous and water-alcohol solutions, Ph.D. Thesis, Institute of Catalysis, Novosibirsk (1983) (in Russian). 52. Savinov E. N., Nagomy E. N., Martyanov I. N. and Parmon V. N. Chem. Phys., 14, No. 4,78-86 (1995). 53. Martyanov I. N., Kinetics of photocatalytic redox reactions of organic molecules on suspensions of CdS and Ti02 semiconductors, Ph.D. Thesis. Institute of Catalysis, Novosibirsk (1998) (in Russian).
Chemical Physics of Nanostructured Semiconductors, p p . 83-1 10 A.I. Kokorin and D.W. Bahnemann (Eds.) Q VSP 2003.
CHAPTER 3
Photo-oxidation of Water at Hematite Electrodes Torbjorn Lindgren, Lionel Vayssieres, Heli Wang and Sten-Eric Lindquist Department of Physical Chemistry, Uppsala University, Uppsala, Sweden
Kevwords: Photoelectrochemistry, photocatalysis, Nanostructured electrodes, Hematite, thin films, Water oxidation List of Abbrevations
CB EE IPCE NHE PEC SC SE SEI
uv VB
conduction band electrolytelelectrode incident photon-to-current conversion efficiency the normal hydrogen electrode photoelectrochemical semiconductor substrate/electrode illumination semiconductor/electrolyteinterface ultra violet valence band
List of Symbols
E,, h hv I
J
quantum yield dielectric constant the permittivity of free space absorption coefficient frequency wavelength absorption capacitance of the space charge layer applied potential conduction band edge band gap of the semiconductor standard electrode potential valence band edge Planck constant light illumination energy light intensity photocurrent
Photo-oxidation of Water at Hematite Electrodes
84
Boltzmann constant donor density light power intensity elementary charge surface area temperature flatband potential onset potential
3.1. Introduction Today, the major part of the energy consumed comes from chemical energy stored in fossil fuels. These fossil reserves are being rapidly depleted and the combustion has led to a severe pollution and disturbance of the environment. It is generally accepted that C 0 2 (besides CH4, NO, and FCH) to a great extent contributes to the greenhouse effect. Due to the climate change, likely caused by the greenhouse effect, a growing interest has developed for new alternative energy sources not based on fossil fuel. Solar energy is a candidate of interest. Although solar energy has many advantages, it is an abundant resource and it is renewable, it is neither permanent nor constant in intensity over the world. Therefore, a suitable energy carrier for storage and transport is likely needed. Such an energy carrier could be dihydrogen. An energy system consisting of solar energy in combination with dihydrogen as an energy carrier would be an environmental friendly and attractive energy carrier for the future. Accordingly, the search of methods to produce dihydrogen using sunlight as energy source, for instance by direct photo-oxidation of water, has been a long-term research goal. In the quest for suitable semiconductor anode materials for direct photo-oxidation of water as well as for solar energy conversion, hematite (aFe203)has been a candidate of interest considering the following reasons: (i) (ii) (iii) (iv) (v) (vi)
band gap of 2.0-2.3 eV [l-111; chemical stability in aqueous solutions in a wide pH range [12, 131; stability towards photocorrosion [ 121; thermodynamic stable crystallo-graphic phase [ 141. thermal stability [ 151; abundance of the material and low cost synthesis.
Different procedures have been established; chemical vapor deposition [5, 161, powder sintering combined with pressing [ l , 2, 6, 7, 12, 17-25], sputtering [9, 261, flux and melt grown [lo, 27, 281, chemical deposition [ l l , 14, 29, 301, sol-gel techni-ques [31], mechano-chemical pro-cessing [321, forced hydrolysis [33,34], spray pyrolysis [35-411, and thermal and hydro- thermal oxidation [3,4, 8, 10, 17,42-581. However, hematite exhibits several disadvantages. For instance, like many n-type oxides, the flat-band level of &Fez03 is unsuitable for direct photoelectrolysis of water since an external bias is needed to generate dihydrogen at the counter electrode, which reduces to overall efficiency [4, 8, 591. Hematite was investigated intensively in the 1970’s but is nowadays often not even mentioned in many reviews about potential semiconductor
Chapter 3; T. Lindgren, et al.
85
materials for water splitting applications. The main reason is that hematite, independent of the methods of preparation, has shown a poor photoresponse due to the following facts: (i) (ii) (iii) (iii)
high resistivity [2,5, 17, 18,431, as high as lo6 SZcm for single crystals [5]; low electron mobility, 0.01[60] to 0.1 cm2/V.s[5, 171; short hole diffusion length, 2-4 nm [6]; high rate of charge carrier recombination [6,29,44,46, 611.
Numerous attempts to increase the photoresponse and the quantum yield of hematite have been performed by: (i)
(ii) (iii) (iv)
varying the doping density or changing the dopants to decrease the bulk resistivity [ l , 2,5-7, 10, 12, 18-24,26,37,61-631; using thin films and/or nanosized particles and colloidals to match the hole diffusion length and increase the surface area [3, 4, 11, 14, 29, 30, 35-38, 44, 511; heat treatment and modification of the surface [18-21, 611; mixing with other oxides like Ti02 to form complex oxides [16, 19, 24, 27, 28,45,50].
However, most of these attempts were found unsuccessful. This is mainly due to the fact that the mechanisms and processes responsible for the low photoresponse of hematite are far from being clearly understood. A deeper understanding of the semiconducting properties of this material is needed, such as its surface structure and a better description of the hematite/electrolyte interface. Extensive knowledge is also required to understand the photoelectro- chemical behavior of hematite and the kinetics involved in the photo-oxidation of water. Finally, for effective oxidation of water an efficient catalyst must most probably be coated onto the surface of hematite. In this review, a literature survey of the water oxidation reaction at iron oxide photoelectrodes is reported. 3.2. General Background of Photo oxidation of Water The first report of direct photoconversion of water into dihydrogen and dioxygen was studied on Ti02 (n-type) electrodes by Fujishima and Honda [64]. Although the overall efficiency of the system was low, their results initiated the interest in photoelectrochemical cells for conversion of solar energy to chemical energy (or to electrical energy). Solar-induced photo-oxidation of liquid water to gaseous dihydrogen and dioxygen: H 2 0 + H2 + 0.5 02,(AG = +237.7 Hmol'')
(3.1)
at semiconductor electrodes is a possible practical application for dihydrogen generation and it has been extensively studied [59, 65-72]. Photo-assisted generation of dihydrogen and dioxygen from water can be achieved in various ways [70]. In a photoelectrochemical (PEC) system, both excited electrons in the conduction band (CB) and holes in the valence band (VB) directly participate in the water splitting process, as illustrated in Fig 3.1.
Photo-oxidation of Water at Hematite Electrodes
86
I
anode
cathode
Fig. 3.1. A photoelectrochemical (PEC)cell for photo assisted water splitting.
For an n-type semiconductor, the semiconductor electrode acts as a photoanode. Accordingly, when an n-type semiconductor electrode is under band gap illumination, photogenerated holes in the VB will oxidize water into O2 and electrons in the CB will move towards the counter electrode to reduce water into H2. In the case of a p-type semiconductor, the holes in the VB are the majority charge carrier and the semiconductor acts as a photocathode. Accordingly, the photo process at p-type semiconductor is just opposite to that of n-type. In order to successfully split water upon irradiation, the quasi Fermi level of holes p E has ~ to be located at a lower energy level than the electrochemical potential of dioxygen evolution (H20/02), and the quasi Fermi level of electrons "EF has to be positioned at a higher energy level than the electrochemical potential of dihydrogen evolution (H2/H+), see Fig. 3.1. The energy difference between hydrogen evolution and oxygen evolution is 1.23 eV and that is accordingly the theoretical energy needed for direct splitting of water. If the position of the energy levels of the quasi Fermi levels are not fulfilled an external bias (Ebias)has to be applied in order to induce the photo oxidation process. For long-term 0 2 and Hz generation, the electrodes must also be chemically stable and resistant to photo-corrosion. The kinetics and mechanisms behind oxidation of water by photo-generated holes are not fully understood and vary most probably with interfacial properties of the semiconductor [73]. Four adsorbed photons and two molecules of H20 are involved to yield one molecule of gaseous dioxygen and four protons. hv
+ a-FezO3 + 4 h+(a-Fe203)+ 4 e-
4 h+(a-Fe203)+ 2 H20 + 4 H+ + O2
(3.2a) (3.2b)
This is, by nature, a very complex reaction. The elementary steps in the detailed mechanism are far from known. However, the kinetics can be improved by surface modification. For instance by surface coating with an efficient catalyst for dioxygen evolution, e.g. Ru02 [74], or increasing the surface area, which facilitate a lower current density which in
Chapter 3; T. Lindgren, et al.
87
turndecrease the overpotential [75].The reduction of water into Hz is a kinetically simpler process compared to the four-electron redox process to evolve 02. It is generally accepted that three major processes limit the photoelectrochemical current in semiconductors after a bandgap excitation [76]. These processes are schematically illustrated in the band diagram shown in Fig. 3.2. The bold arrows show the desired processes for efficient water splitting PEC cell after a bandgap excitation: the transport of electrons to the back contact, the transfer of the hole to the semiconductor surface and the oxidation of water at the semiconductor/electrolyte interface. The three major limiting processes are a) bulk recombination via bandgap states, or b) directly electron loss to holes in the valence band (eventually followed by emission of light), and c) surface recombination. These three limiting factors are general for all light absorbing semiconductor materials. If a proper electron scavenger is present (in Fig. 3.2 represented by 02),electrons can also be lost from the conduction band (process d). In addition, photocorrosion of the semiconductor itself and dissolution reactions can also occur (process f). This leads to a degradation of the electrode, which is a major stability problem in some PEC systems, e.g. InP, CdSe and n GaAs [77].
Fig. 3.2. Schematic representation of the different pathways for the photogenerated electron-hole pair for a planar n type semiconductor in electrolyte solution.
The photoelectrochemical properties of different semiconductor materials have been widely reported, for single crystal, polycrystalline, as well as for nanostructured materials. In the literature various methods for measuring the efficiency are found. The most common is the IPCEh (incident photon-to-current conversion efficiency) or quantum efficiency, which is defined as hc J p h J IPCE =-I. 4 PIA
(3.3)
where J p h ,is~ the photocurrent in AmpBre, PA is the light power intensity in Watt, A.is the wavelength in nanometer and h, c, and q have their usual meaning of Planck’s constant,
88
Photo-oxidation of Water at Hematite Electrodes
speed of light in vacuum and the elementary charge. The IPCE plotted versus the wavelength is called an action spectrum. Another common and scientifically more relevant parameter is the quantum yield, $A or APCEk (absorbed photon to current efficiency) where the optical absorption (An) of the film is taken into account: (3.4) Also frequently used is the photocurrent response at a given potential and light condition (intensity and wavelength). The photocurrent obtained with illumination of the ntype hematite used as anodes are directly proportional to the amount of dihydrogen that can be produced by reduction of protons in the electrolyte. Also commonly used is the overall efficiency, 17 where the relation between the energy content of dihydrogen produced in relation to the incident irradiation energy is reported [38]. If a potential is applied (Ebins)then the electrical power input at the given potential should be withdrawn from the total power output (Le. the energy content of the dihydrogen produced). ?=
total power output - electrical power input - j p (Eonset - ~ light power input IO
)
b
i
~
~
(3.5)
where j p is the current density, E,,,, the onset potential, Eappliedthe applied potential and IO the light power input. For any practical applications, an overall efficiency of at least 10% [66] is required to compete with conventional ways of producing dihydrogen, for instance by steam reforming of natural gas. Theoretically, it should be possible to design a device with an overall efficiency of 10% or higher. The peak efficiency for photovoltaic cells would occur for a single bandgap photoconverter in the range 1.4 to 1.6 eV [78] where an efficiency of up to 31% would be achieved. For a tandem device, where two light harvesting semiconductors operate in series to cover a broad range of the solar spectrum, the theoretical upper limit is around 41%. If energy should be stored in a chemical fuel (e.g. dihydrogen or a hydrocarbon like CH4 or C~HSOH), further energy losses must be encountered. Memming, see reference [79] and references therein, calculated the theoretical conversion efficiency (one photoactive electrode and no overpotentials) for photoelectrochemical water splitting to about 27%, which is not far from the maximum theoretical efficiency of 31% for conversion of sunlight into electrical energy. For splitting of water with sunlight, the maximum realizable efficiencies, taking into consideration all possible losses, have been estimated to be about 10% for a single-bandgap and 16% for a two-bandgap system [80]. However, stability problems, both chemically and low stability towards photocorrosion, as well as high costs have so far prevented the commercial breakthrough of photoelectrochemical (PEC) devices for direct conversion of solar energy to dihydrogen. The report by Fujishima and Honda [64] of direct photoconversion of water into dihydrogen and dioxygen was performed at Ti02 (n-type) electrodes with a chemical bias. A disadvantage of the Ti02 system was the wide band gap of 3.2 eV, where only the UV light contributes to the photo process. After the first successful photoelectrochemical cell
Chapter 3; T. Lindgren, et al.
89
by Fujishima and Honda a large number of semiconductor materials, like InP [77, 811, SrTiO3 [82, 831, Fez03 [ l , 4, 9, 20,22, 25, 29, 35-38, 44, 841, W03 [85-881 and GaN [89, 901 have been investigated for H2 and 0 2 generation in order to improve the efficiency of the PEC cell. As for photoelectrolysis of water at p-type semiconductors, Heller and co-workers studies of InP electrodes should be mentioned [77, 811. A solar to chemical conversion efficiency as high as 12% was reported for this system [77, 811, with relatively low stability. Combinations of a photoanode (n-type) and a photocathode (p-type) of doped hematite have been studied [ 1, 20, 221. External bias was not needed for photo-oxidation of water. Both H2 and O2 were detected with this PEC diode assembly. However, the overall conversion efficiency was reported as low as 0.05%[ 11. Recently, the most interesting findings in the field of direct solar induced water splitting is the reports by Turner et a1 [91] and by Grtitzel and co-workers [92]. Both groups used multiple-junction systems, yet based on completely different materials. Turner and coworker used a highly efficient solid-state photovoltaic cell integrated to a photoelectrochemical device consisting of p-GaInP2 [91]. Efficiencies of 12% were obtained but the cost of the system is at present too high to compete with conventional renewable techniques. The tandem device described by Gratzel et a1 [92] was based on nanocrystalline W03 biased by dye-sensitized nanocrystalline Ti02 to absorb transmitted light from W03. The overall solar light to chemical conversion efficiency of this device was reported to be 4.5%. Arakawa et a1 [93] demonstrated that doping of indium-tantalum-oxide with nickel yields a series of photocatalysts, Inl-,Ni,Ta04 (x=O-0.2), which induces direct splitting of water into stoichiometric amounts of oxygen and hydrogen under visible light irradiation with a quantum yield of about 0.66 %. The result is interesting, despite the low quantum yields, since it points out that metal oxides indeed could be used for the purpose of direct splitting water under visible light irradiation. Another interesting work is the recent report by Licht et al [72, 75, 941. Although the system they studied was not a strict photoelectrochemical one, since the photovoltaic system was separated from the water electrolyser, their study is of general interest for the water oxidation field. The photovoltaic cell was connected to a water splitter catalyst system of considerably larger area than the solar cell. With this design, it was possible to combine a high solar cell efficiently with a low photocurrent density over the electrolyzer Ciph = 0.44 mA/cm2), which minimized the overpotential needed for water oxidation. An overall efficiency as high as 18.3% was obtained.
3.3. Structural and Physical Properties of Hematite 3.3.1. Crystal Structure
Hematite crystallizes in the trigonal crystal system, space group R$ I D&, and is isostructural with corundum (a-Al203) see Fig. 3.3. The unit cell can be described as rhombohedral with three equal axes a = 5.43 A and an angle between edges a = 55'18' containing two formula unit (Z = 2), or as hexagonal with a = 5.038, and c = 13.75 8, (Z = 6). The lattice is built on a hexagonal close packed (HCP) array of dioxygen with four of
90
Photo-oxidation of Water at Hematite Electrodes
every six available octahedral sites around 0 atoms occupied with Fe [95]. The octahedral and tetrahedral sites are above and below one another in a HCP lattice, the tetrahedral sites remaining empty. Octahedra are sharing faces along a threefold axis and are distorted to trigonal antiprisms because of the Fe-Fe repulsion occurring across one shared face and not the others.
Fig. 3.3. Crystal structure of hematite (a-Fe203).
This yields to a very dense structure (Le. high dioxygen packing index), showing a high polarisability and a high refractive index. The main intrinsic defect in hematite is oxygen vacancies, which in turn increase the n-type behavior by donated electrons to the conduction band [ 171. High temperature treatment increases the oxygen vacancy concentration [ 17, 201. However, the oxygen deficiency at elevated temperatures (TL15OO0C) may result in the formation of a surface region of magnetite, Fe304 [25] It has been suggested that the presence of Fe304in the surface layer decrease the photoresponse due to extensive recombination of holes and electrons in the highly conductive surface region [25]. The intrinsic doping density varies with preparation and doping densities of about 1.1OI6cm-3for single crystal [9] and as high as 2.2.10'' for pyrolytical synthesized hematite have been reported [381. 3.3.2. Electronic Structure
It was found by ultraviolet photoemission spectroscopy (UPS) that the valence band of well-ordered, nearly stoichiometric hematite consists of overlapping 0 2 p and Fe 3d orbitals which give rise to a complicated structure about 10 eV wide [96]. Gardner et al [17] observed optical absorption peaks for polycrystalline hematite at 2.4, 3.2, and a strong broad absorption band centered at 5.8 eV. Molecular orbital studies explained these energy levels as Fe 3d 3 3d, 0 2p Fe 3d, and 0 2p Fe 3d transitions, respectively [95]. Similar energy levels where obtained with ellipsometry for both single crystals and polycrystalline films, see ref. [95] and references therein. A more recent study based on computational analysis of the density of electronic states, using the ab initio periodic
+
+
Chapter 3; T. Lindgren, et al.
91
unrestricted Hartree-Fock approach, showed that the bandgap is of p-d rather than d-d type [97]. Accordingly the authors suggested that hematite should be classified as a charge transfer rather than a Mott-Hubbard insulator where the conduction and upper valence edge are of the same character. Matsumoto [98] tried to summarize the work done on the electronic structure of iron oxides and concluded that the about 2 eV in the bandgap of the hematite semiconductor measured in photoelectrochemistry indeed is based on the 3d band transition between the Fe3+ions, which supports a Mott-Hubbard insulator. It has been noticed that the optical absorption of hematite has a much wider spectrum than the photocurrent response [6, 441. It was suggested that more than one optical transition occur in hematite, but only one results in the creation of an electron-hole pair where the hole is capable of evolving dioxygen from water [6].The absorption peak around 2.4 eV is located slightly high but in agreement with the bandgap measured for the semiconductor. The bandgap of hematite determined by various methods is found to be in the range of 1.9 2.3 eV (Table 3.1). The energy of the bandgap allows absorption of all the UV light and the blue part of the visible region, which gives hematite its brownish red color.
-
Table 3.1. Bandgap of hematite materials in eV Bandgap, eV
Material
Characteristics
Reference
1.9
Not mentioned
Anodic oxide, 3 nm thick
10
2.0
Amorphous
RF sputtered, 1 pm
9
2.0
Single crystal
Hydrothermal synthesis
8
2.0
Polycrystalline
Size 140 nm
3
2.1
Single crystal
Ti-doped
7
2.1
Single crystal
Zr-doped
5
2.1
Polycrystalline
Thermal grown, 3pm
10
2.2
Polycrystalline
Nanorod-may 3 nm @ x 500 nm length
11
2.2
Single crystal
Flux grown from an Li2M04melt
9
2.2
Polycrystalline
Size 100 nm
4
2.2
Polycrystalline
Undoped
6
2.2
Polycrystalline
Si- and Ge-doped
2
2.3
Polycrystalline
Mg- and Si-doped
1
Photo-oxidation of Water at Hematite Electrodes
92
The bandgap (E,) of a semiconductor is usually determined by means of optical absorption. The following equation gives a satisfactory description of the absorption behavior near the threshold [85]:
where a is the optical absorption coefficient, h the Planck’s constant, v the frequency of light, A an constant, and E, the bandgap energy, n = 0.5 for a direct transition and n = 2 for an indirect one. Hematite has been suggested to have an indirect bandgap [5, 61. However, there is also reported a direct bandgap for hematite [ l l ] . In the latter report, hematite nanorods were studied and it was stated that the quantum size effects due to the aspect ration of the single nanorod (3 nm in diameter x 500 nm in length) might be the reason for the unusual behavior. The bandgap of a semiconductor can also be determined photo-electrochemically [ l , 2, 5, 7, 81, which is based on the fact that the wavelength corresponding to the onset of photocurrent agrees well with the optical absorption edge. For colloids and powders, diffuse reflectance spectroscopy method has been used to characterize the hematite bandgap [4].
3.3.3. Electrical Properties In general, it is accepted that recombination of electrons and holes, trapping of electrons by oxygen deficiency sites and a low mobility of the holes, cause a low conductivity and accordingly a low photoresponse for hematite. Electron mobility in the range 0.01[60] to 0.1 cm2/V.s [ 171 has been reported. In the latter case, it was found that the electron mobility was independent of donor concentration. More recently, an electron mobility of about 0.1 cm2N.s has been measured with doped single crystals and the mobility was also here independent of donor concentration [5]. A diffusion length of holes has been determined to be only of 2-4 nm [6], which is about 100 times lower than many other (III-V) oxides. 3.3.4. Flat-band Potential of Hematite
The flat-band potential (U,) is a very important parameter for a semiconductor in contact with an electrolyte. U, is directly related to the conduction band (CB) level for an n-type semiconductor or the valence band (VB) level for a p-type semiconductor. A classic method to determine U, is by means of Mott-Schotky plot [99] according to:
--1 C&
-
2
.(U-U,
EE~~N,S’
kT
--)
(3.7)
4
where C, is the capacitance of the space charge layer, E the relative dielectric constant, EO the permittivity of free space, q the elementary charge, No the donor concentration, S the surface area of the electrode, U the electrode potential, U, the flat-band potential, k the Boltzmann’s constant and T the absolute temperature. The term kT/q is 25.7 mV at room temperature. Plot of I / Csfzvs. U (Mott-Schottky plot) will give a straight line where the intercept with the potential axis corresponds to the flatband potential (U’) and from the slope of the line the donor concentration, (No) can be determined. The value of C,,can be
Chapter 3; T. Lindgren, et al.
93
determined experimentally as a function of potential (U) from impedance measurements. McCann et a1 [ 1001studied the impedance responses of a a-Fe203/ 1 M NaOH / Pt cell and the general equivalent circuit was determined. There are assumptions, which must be satisfied in the application of the Mott-Schottky plot. For instance, the semiconductor must contain a space charge region. In nanostructured materials there is likely no space charge layer formed [loll and therefore may the Mott-Schottky plots show dispersion and low coefficients of correlation [U]. Table 3.2. pH dependence of on-set potential, U , , and flat-band potential, Up vs. NHE for various hematite materials PH
uom
sc*’
v PC*’
ut& v
sc*’
2.7 0.44
4 6.3
6.7
0.24 0.46
Ref.
PC*’ 0.29
Nanostructured
Not mentioned
44
0.18
Sintered and pressed pellets
0.1 M KzS04, Buffered
7 102
0.04
Sintered and pressed pellets
Complex
**)
0.14
7 102
7
0.04
8.4
-0.07
Thermally grown
1M NaN03, borate buffer
10
8.4
-0.02
Anodic oxide
1M NaN03, borate buffer
10
8.6
0.36
61
102
-0.17 -0.20
Sintered and pressed pellets
0.1 M KzS04 Buffered
7
10.7
-0.31
Nanostructured
Not mentioned
44
12
-0.41
Nanostructured
0.01 M NaOH
44
-0.2 1
Firing of iron foil
0.1 M NaOH
45
Nanostructured
0.1 M NaOH
44
Sintered and pressed pellets
Complex
7
10
0.04
13
-0.06
13
-0.46
14
-0.16
14 *)
Electrolyte
0.14
0.47
6.7
Material
-0.12
-0.36 -0.43
**)
102
SC - single crystal; PC - polycrystalline; **) 0.1 M K2S04 adjusted with NaOH.
The location of U ’ can also be obtained by studying of the potential dependence of the photocurrent. Under monochromatic illumination for wavelengths close to the band
94
Photo-oxidation of Water at Hematite Electrodes
gap, the following relation applies [28, 851:
u-ufi-(JphB)2
(3.8)
where B is a function of wavelength and light intensity and Jph the photocurrent. For semiconductor colloidal particles pulsed radiolysis techniques [ 1031 have also been used to derive the flat-band potential. Due to the simplicity to determine the onset potential ( Uon)experimentally, it has been widely accepted to substitute U, with Uon.Ideally, U,, should coincide with U, [61]. In most n-type semiconductors, however, U,, is shifted a few tenths of a Volt, to more anodic potentials than U, [104]. This represents the situation when sufficient band bending has been established in the semiconductor space charge layer, which enables an efficient hole transfer to the reduced species in the solution. Table 3.2. shows that data collected from the literature for onset and flatband potentials for various pH generally are related. However, there are some discrepancies, especially for the onset potential of nanosized hematite in pH 13 solution. One may conclude that the onset and flat band potential are dependent on the crystal structure, surface morphology and electrolyte medium. The data listed in Table 3.2. are plotted and shown in Fig. 3.4.
0.6
w
0.4
1 0
A
*
cri
0.0
.. . 0
AA M
0
0
I
0
E
3
-0.6
~ , . , . , . , , , . , . l , , , ~ . l , , , , ’ 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4
PH Fig. 3.4. pH dependence of on-set potential, U,,and flat-band potential, U, vs. NHE for various hematite materials as collected from Table 2. The solid line, with a Nernstian slope of -59 mV/pH-unit, shows the potential for hydrogen generation as a function of pH. Ebim stands for the bias, which at least is needed at pH 10 to photogenerate hydrogen at the polycrystalline hematite electrode described by McGregor et a1 [7]. Uo, A U ,single crystal, 0 U,, AU’, polycrystalline.
It can be concluded that the onset and flat band have a close to Nemstian shift with pH. Generally, the pH dependencies of onset and flat band for single crystals have higher slopes. Since a change of pH also changes the electrode potential of the H2D-I’ couple according to:
Chapter 3; T. Lindgren, et al.
95
it is generally accepted that it is not possible to directly generate hydrogen on hematite electrodes. For instance, McGregor et al [7] determined a flatband potential for polycrystalline hematite at pH 10 to -0.20 V vs. NHE, which is 0.39 V positive of the potential for hydrogen evolution. Accordingly, a bias of 0.39 V is at least needed (likely more due to energy barriers in the system) to generate hydrogen. See Fig. 3.3. Matsumoto [98] showed that there is a linear relationship between bandgap and band edges for oxide semiconductors. The relationship between the conduction band edge (E,), the valence band edge (E,,) and the bandgap (E,) was described as:
E, (V vs. NHE) = 1.23 - E, (eV)/2
(3.10a)
E, (V vs. NHE) = 1.23 + E, (eV)/2
(3.10b)
These equations indicated that a bandgap higher than about 2.46 eV is necessary for direct splitting of water without bias voltage. A bandgap energy of 2.46 eV is double the energy theoretically needed for direct splitting water and corresponds to an irradiation wavelength of about 500 nm, which means that a large part of the visible range could not be utilized for direct water oxidation. However, Arakawa et al. [93] showed that nickel doped Inl.,Ni,Ta04 (x = 0-0.2) with a bandgap energy around 2.25 eV indeed split water upon visible light (A > 420 nm) irradiation so there seems to be exceptions from the equations proposed by Matsumoto. 3.4. Photoresponse of Hematite Materials.
Poor charge transport, due to charge recombination, together with a low heterogeneous rate constant for hole transfer from the valence band of hematite to the hydroxide ion seem to cause the low quantum yield for hematite when used for the aim of direct water oxidation [18, 21, 36, 43, 44, 611. The oxidation of water by holes is a very slow process, with a heterogeneous rate constant for hole transfer from the valence band of hematite to OH' being 0.1 1 c d s ; compared with lo3 lo4 c d s for W 0 3 and Ti02 [61]. K. Itoh and J. M. Bockris [36] found a similar value for the hole-water oxidation transferrate constant at hematite thin film electrodes. Such a slow process accumulates holes at the valence band of hematite, which will either recombine or induce competing reaction at the hematite surface. Recently, it was shown that transient absorption decay for hematite nanoparticles was very fast, 70% of the transient absorption disappeared within 8 ps and no measurable transient absorption remained beyond 100 ps [43]. This represented a much faster decay than many other semiconductors, which is consistent with the observed poor charge transfer properties in hematite. It should be mentioned that this decay was independent of the excitation power, which suggests alternative relaxation mechanisms compared to those observed for Ti02 and ZnO for instance [43]. Since the relaxation was independent of pump power, probe wavelength, pH and surface treatment the fast decay was interpreted to be due to intrinsic mid-bandgap states and trap states rather than surface defects. This is in agreement with earlier investigations [ a ] .
-
-
96
Photo-oxidation of Water at Hematite Electrodes
Hematite has a good chemical stability in aqueous solutions over a broad pH range [12, 131. Ishikawa and co-workers [13] studied the solubility of hematite in LiOH, NaOH and KOH solutions. An excess of regular powder a-FezO3 and 20 cm3 of alkali hydroxide solution were put in a water bath shaker for 5 hours. The solubility increased with increasing temperature (studied between 303-349°K) and alkali hydroxide concentration, although it remained very low. The solubility was less than l . l ~ l O M - ~ Fe(m> in 10.1 M NaOH at 349°K. Most of the photoelectrochemical studies of hematite have been performed in sodium hydroxide solutions with a concentration less than 1.0 M. In those cases the dissolution in darkness should be negligible. However, when used for photoelectrochemical purposes the degradation becomes more complicated and the current density as well as the photon flux has to be taken into account. It was stated by Kennedy et al [12] that hematite likely was more stable toward photocorrosion than wider bandgap materials like W 0 3 and that its stability towards photocorrosion was comparable to Ti02. However, in the presence of chloride ions the photocorrosion accelerates and degradation was noticed during visible-light irradiation of hematite colloidal solution at pH 0.3 [105]. Apart from photoelectrochemical and electrochemical systems solar generated dihydrogen can also be produced by biochemical and thermochemical methods [106]. For the latter, non-stoichiometric highly solar energy absorbing iron oxide, Fel.,O, as well as nickel and manganese ferrites have been used and studied [ 107-1091. Semiconductor photocatalysis has been successfully applied in degradation of pollutants e.g. SO2 and phenolic waste [69]. It is mainly TiO2, which have been used and studied for this purpose, but there are reports where hematite has been used for photodegradation [52-54, 56, 57, 110, 1111. Hematite has also been studied as a gas sensing material [40]. In this section a more detailed description of the photoresponse for the various hematite materials, which have been studied for the aim of photooxidation. Due to the different ways of presenting the efficiency, a direct comparison of the photoelectrochemical properties of the materials when used for water oxidation is somewhat difficult. 3.4.1. Colloidal Solutions of Hematite A high quantum yield (more than 80% in UV light) was observed by Moser and Gratzel [84] in photoinduced oxidation of iodide on 120 nm-sized iron colloids in suspension in strongly acidic environment. They concluded that the advantage of employing Fe203-sols is that the dimension can be reduced to a size where practically all the photoinduced charge carriers attain the particle surface before recombination can occur. It was stated that such systems could be suitable to become the 02-producing part in a complete water cleavage system. The work was continued by Gratzel et al [112] who found that H2 gas was evolved, despite the unfavorable position of the conduction band, in aqueous solution on irradiation of 80-120 nm colloidal hematite with a solar simulator (100 mW/cm2). The colloidal suspension was obtained after 24 hours hydrolysis of FeC13 at 125°C. It is interesting to notice that H2 was formed under illumination of the colloidal Fez03 particles even in the absence of nobel-metal catalysts and sacrificial electron donors. It was suggested that the small size of the colloidal Fez03 particles induced a shift in the flat-band position towards more negative potentials. However, Nozik and co-workers [ 113, 1141 determined the flatband potential for particles in the size range of 50 to 350 8, to -0.35 V (vs. NHE) at pH 11, which is in close agreement with the value derived for the bulk
Chapter 3; T.Lindgren, et al.
97
material. It should also be mentioned that contrary to the study by Gratzel and co-workers [ 1121 it was found in a photochemical study of colloidal hematite of similar dimensions to those used by Gratzel and co-workers [47] that without an electron relay (MV”) no dihydrogen gas was produced at all upon irradiation. Suspensions of hematite have also been used and studied for other aims than photooxidation of water, e.g. catalytic oxidation of sulphur dioxide in aqueous solutions [52]. Aqueous dispersion of semiconductor particles could be an easy and attractive way to photooxidise water, but they have the drawback that dihydrogen and dioxygen are produced simultaneously in the same suspension. Apart from the separation problem the two produced gases may create a pathway for back reactions that reduces the yield of the overall photo-oxidation process. The latter obstacle can partly be avoided by addition of Na2C03, which was successfully shown by Arakawa et a1 [115]. 3.4.2. Hematite Single Crystal Materials 3.4.2.a. Hematite Single Crystal
The quantum efficiency of the single crystal hematite electrodes [9] in 0.1 M NaOH aqueous solution polarized at 0.64 V vs. NHE was less than 0.005 at 380 nm and was decreasing with increasing wavelength of monochromic light. These single crystals were obtained by flux growth from a Li2Mo04flux melt and had a doping density of 1 ~ 1 0 ’ ~ cm-3 and a band gap around 2.2 eV. Single crystals of hematite grown by chemical vapor deposition (CVD), were prepared by Launay and Horowitz [5]. However, they reported that the crystals of undoped hematite were not suitable for photoelectrochemical purposes due to their high resistivity, lo6SZcm. 3.4.2.b. Doped Hematite Single Crystal
An investigation of 2-doped single crystal hematite grown by CVD at 1300 K was reported by Launay and Horowitz [5]. The quantum efficiency in 1M NaOH at 0.64 V were about 0.12,0.16 and 0.21 for doping densities of 1 . 5 ~ 1 0 ’3~., 5 ~ 1 0and ’ ~ 6.5~10’~cm”, respectively. It is interesting to note that the quantum efficiency increased with the square root of the reciprocal donor concentration. The resistivity, measured by four point probe technique, varied with the doping density and was found in the range of 75-4 Qcm for the samples mentioned. The bandgap was determined photoelectrochemically to 2.1 eV. It was stated that the stability towards photocorrosion was good in alkaline solution. The corrosion current was lower than 0.1% of the total photocurrent. In order to reduce the bias needed for photo-oxidation of water at hematite electrodes single crystals of YFe03 were flux and melt grown [28]. The bandgap was determined to 2.58 eV for YFe03 and the flat-band in 1 M NaOH was located at -0.22 V vs. NHE. Low photocurrent densities (1 pA/cm2 at 0.74 V) and accordingly low IPCEvalues (less than 0.01) was related to the high resistivity (around 40 k!2. cm) of the material in combination with trap states. The results indicate that this material is not a suitable candidate for photoassisted water splitting in the visible region of the solar spectrum. The results obtained by doping with 2 are promising [5]. It can be concluded that single crystals of hematite generally exhibit low current response and high resistivity, which make them a less attractive candidate for photo-
98
Photo-oxidation of Water at Hematite Electrodes
electrochemical purposes. However, the PEC response of hematite single crystals can be improved by doping [5, 91. Preparation of single crystal of hematite, like single crystal of many other materials, can be complicated and expensive. A polycrystalline material is usually more suitable for industrial production. 3.4.2.c. Mixed Hematite Single Crystal
Single crystals of corundum iron titanates, FeTi03 (ilmenite), FeZTiO4 and FezTi05, were synthesized by heating in vacuum to 1450°C for 15 min the appropriate stoichiometric mixtures of Fe, FezO3, and Ti02 [27]. The single crystals were then grown from the melt under an argon atmosphere by the Czochralski technique [ 1161. Bandgaps for these materials were 2.16 eV for FeTi03, 2.12 eV for Fe2Ti04and 2.18 eV for FezTiO5 similar to the value of hematite. The flatband potential at pH 14 (1 M NaOH) was in the range of 0.34 to 0.74 V, indicating that a larger bias is required to split water at these materials than at hematite. Resistivities for the samples were determined by four probe techniques and were typically of the order of 10-100 Qcm. The best result among these materials was recorded for FeTi03. However, the photocurrent and the IPCE values were low (pA/cm2). A linear relationships of iphoto a light intensity for all three materials up to 1200 mW/cmz where found. The authors interpreted this as if the intrinsic bulk properties of the semiconductor rather than the electrode kinetics at the interface, were limiting the photocurrent. 3.4.3. Polycrystalline Hematite Materials 3.4.3.a. Polycrystalline hematite
In photo-assisted water splitting, high purity hematite anodes resulted in a very low quantum efficiency of 0.03 [ l ] at a wavelength of 380 nm and a bias of 1.25 V vs. NHE in 0.01M NaOH, or even no photoresponse at all [2, 191. Those electrodes were prepared by pressing high purity a-FezO3 powder, the pellets were then sintered at high temperature (T 21000°C). Although the photoresponse of plain polycrystalline hematite has been poor, doping and thin films of poly-crystalline hematite have proven to be efficient ways to decrease recombinations of photogenerated charges and increase the photoresponse, see below. 3.4.3.b. Polycrystalline Hematite Thin Films
To reduce the loss of recombination, one method suggested by Itoh and Bockris [35, 361 was to use very thin film anodes. In this way, the space charge layer can be adjusted to be of the same thickness as the film. Thus recombinations are restricted. Stacking and interconnecting a number of thin film electrodes after each other can compensate the absorption loss due to the small electrode thickness [35]. Thin film iron oxide electrodes were easily made by spray pyrolysis of a 0.1 M FeC13 ethanolic solution containing 0.1 M HC1 onto hot SnO2-coated glass (35OoC, in air) using NZas carrier gas. It was mentioned that approximately 40A of Fe203film was deposited by one spraying; 1-2 min intervals were taken between each spraying to avoid excess cooling of the substrate [36]. Nothing was mentioned about the form of iron oxide obtained by spray pyrolysis. However, considering the preparation procedure one might conclude that hematite was
Chapter 3; T. Lindgren, et al.
99
formed. The thickness of the films was controlled by observing interference colors. Their study showed a photocurrent of 0.3 mA/cm2 for a 60 nm thick film at pH 13 at 0.84 V versus NHE and 1 sun from a solar simulator. Ten stacked electrodes gave 1.6 mA/cm2 under the same condition. This can be compared with a photocurrent density of about 0.5 mA/cm2 for a single anode of thickness 600 nm at similar conditions [35, 361. This represents a significant contribution for the application of hematite. The spray pyrolised thin film electrode of hematite has been further investigated by Khan et al with undoped and iodine-doped [37, 381 iron oxide thin films. For the undoped iron oxide [37, 381 (for the doped spray pyrolised iron oxide see next paragraph) a similar deposition technique as described above was used. However, O2 was used instead of nitrogen as carrier gas. Raman spectroscopic analysis indicated the formation of a hematite film. It was concluded that the rates of photoelectrochemical oxidation of water at these spray-pyrolised thin film electrodes were dependent on the spray time, substrate temperature, solvent composition in the spray solution, and the concentration of the spray solution. Using the optimum conditions, a maximum photocurrent density for the hematite was as high as 3.7 mA/cm*. This was obtained in 1 M NaOH solution at a bias of 0.7 V vs. SCE and at a light intensity of 500 W/m2. The photoresponse corresponded to an overall efficiency (light energy to chemical energy) of 1.84%. The study showed an optimum film thickness which was related to the balance between light absorption by the film and recombination losses due to reduction in the electric field gradient in thicker films. There was no information given about the film thickness in this specific paper. However, referring to the earlier investigation of the same group [117] where a growth rate of 100 d m i n was assumed, it may be estimated that the optimum film thickness in the latest described investigation was in the range 50-100 nm. It was neither stated which form the formed iron oxide had. Considering the preparation temperature and the bandgap of the obtained film it should be hematite. A substrate temperature of 35OoC, a total spray time of 60 s and a spray solution of 0.11 M FeC13 in 100% ethanol was suggested as the optimum parameters. The duration of each spray was kept at 10 s, with a 5-min interval to maintain a constant substrate temperature. The distance between the substrate and the spray nozzle was not mentioned, neither was the spray rate (ml/min) and carrier gas. It should be ' (measured at frequency mentioned that the donor density (Nd) was as high as 2 . 2 ~ 1 0 ~cm-3 of 1000 Hz) [38]. This reduces the resistivity of the film and makes the space charge layer very thin. The authors consequently suggested that high donor density of the pyrolytically synthesized films may be one reason behind the high photocurrent density due to its low resistivity. The bandgap for the spray-pyrolysed thin films of hematite was found to be 1.97-2.05 eV [37, 381 regardless of the doping density. There are others reports on hematite prepared by spray-pyrolysis [39, 401. Although no photoelectrochemical measurements were done these reports are interesting due to the findings and the rigorously material characterizations of the prepared films. Wang et a1 [40] reported successful formation of a highly nanocrystalline a-FezO3 thin films successfully onto Si(ll1) substrates. AFM studies confirmed that the porous film consisted of spherical particles of about 60nm diameter having columnar ordered structure and preferred orientation [ 1041. In the preparation, an aqueous ethanol solution of iron acetylacetonate (0.01 M) was chosen as the precursor. Air was used as carrier gas and typical substrate temperature was around 380°C or higher to obtain the spherical nanoporous structure. The film thickness was around 300 nm. The thin films showed good sensitivity to CH4 and it was suggested that the spray pyrolysed nanocrystalline hematite
100
Photo-oxidation of Water at Hematite Electrodes
films could be promising thin film gas-sensing material. Considering the earlier successful results with spray pyrolised hematite [35-381, it would be interesting to also study the photoelectrochemical properties of the material synthesized by Wang and co-workers [40]. 3.4.3.12.Polycrystalline Doped Hematite
The most frequently used substances for n-doping of hematite are oxides of silicon and titanium. Photo-oxidation of water at Si- and Ti-doped hematite has accordingly been studied extensively [ l , 6, 7, 12, 18, 20, 21, 611. In general, the conductivity is improved by these dopants. However, there seems to be a limit for the improvement by doping. For example, concentrations over 0.5 atom percent of titanium in hematite have been stated to have no effect on the resistivity [7]. Interesting studies of silicon [12, 18, 211 and titanium [6, 211 doped polycrystalline hematite were performed by Kennedy and co-workers. The electrodes were prepared by mixing of a-FezO3 with reagent grade SiOz, or TiOz, pressing and sintering. Sintering was carried out in air at 1250-1350°C from 4 to 24 hours followed by quenching to room temperature. Photoelectrical properties of the doped polycrystalline electrodes were studied. For irradiated silicon doped polycrystalline iron oxide electrodes (Eg-2.2 eV) in HzSO4 and in NaOH over a broad pH range photogenerated dioxygen was collected and a maximum of 100% faradaic collection efficiency [123. This means that all of the photocurrent was utilized for dioxygen evolution at the hematite electrode, this was the case even at high incident light intensities. The photocurrent densities were in the mA/cmz region at a light intensity of 700 mW/cmz. However, after the period of the experiment (80 h) in acid environment traces of Fe(m> were found in the electrolyte which indicate degradation of the electrode. It should be mentioned that low photocurrent and low quantum efficiency for Ti-doped hematite have also been reported [61]. Hematite electrodes doped with Si, Ge, Sn or Pb have been reported by Kennedy et al [2]. Pressed pellets were sintered at 1250-1370°C for 3-40 hours. They showed high photocurrent in 0.1 M NaOH at 400 nm with 4-18 mW/cm2 light intensity. Photocurrent densities at 0.7 V versus NHE were about 160, 115, 110, 90 and 5 pA/cm2 for a-Fez03 electrodes doped with 2% Si, 0.002% Sn, 0.008-0.05% Pb, 0.05% Ge and 1% Ti respectively. The bandgap was similar of all group IV-doped electrodes and was around 2.2 eV. Similarly pressed and sintered iron oxide, doped with 0.1% GeO under 150 mW/cm2 polychromatic illumination in 1 M NaOH, showed a net photocurrent densities as high as 1 mA/cm2 [19]. The resistivity of FeZ-xGex03varied from lo6 SZcm for x = 0 to about 0.1 Qcm for x = 0.05. The solubility limit of GeOz in Fez03was reported to be 5 mol%. The investigations of Fez' as dopant for hematite are interesting due to the ease of preparation [23, 24, 261. Iron oxide pellets were prepared from high purity Fe304, a-FezO3, y-FezO3 and CrzO3 powders. After mixing the powders with B O z to the desired composition the mixtures were pressed and sintered at various temperatures (1300-1600°C) during 24 hr. A doping level of 10'' cm-3 corresponded approximately to 1% Fez+in the structure. The bandgap energy was evaluated to 1.9 eV. The electrode yielded a photocurrent density of around 0.4 mA/cm2 (normalized to 1 W lamp output) in a borate buffer (pH 8.4, 0.6 V versus SCE) at a wavelength of 350 nm [23]. Increased photocurrent with doping concentration was reported. No upper limit for the doping concentration was mentioned. It was stated that the chemical stability of the iron oxide samples in aqueous electrolyte of 5 M HCl decreased dramatically with higher doping densities than 1021cmS3.
Chapter 3; T.Lindgren, et al.
101
However, the stability of Fe2+-dopedhematite could be significantly improved by addition of CrzO3 to the iron oxide samples [23]. The amount of added Cr2O3 was typically of 40%. Thin sputtered Fez03 film electrodes (a RF-magnetron sputtering system was used with a Fez03 target) doped by means of partial reduction treatment at cathodic potentials were prepared by Virtanen and co-workers [26]. The photocurrent behavior of the Fe3'oxides was strongly determined by the presence of Fez+in the oxide. It was shown that the photocurrents were strongly increased after the reduction treatment, as a consequence of increased doping density. A reduction treatment of -0.6 V vs. SCE for one hour gave a photocurrent of 190 pA/cm2 (at a wavelength of 520 nm) measured at 200 mV versus SCE in borate buffer, pH 8.4. The photocurrents were normalized to 1 W lamp power output. The bandgap energies of the electrodes were between 1.8 to 1.9 eV. The authors did not comment on the stability of the material after the reduction treatment nor the doping density was mentioned. The results indicate that hematite doped with Fez+may be prepared from magnetite, Fe304. The p/n assembly described by Somorjai and co-workers [l, 20, 221 was operative only when light created a photopotential large enough. It was reported 111 that the maximum quantum efficiency at 380 nm at a bias of 1.25 V for undoped hematite at pH 12 was 0.03, while a 10% Si-doped n-type hematite electrode resulted in a ten-fold increase in quantum efficiency under the same conditions. The Si-doped n-type iron oxide disk electrodes were prepared by mixing fine powders of a-Fe2O3 and SiOz so that 0 c Si/Si + Fe 5 20 at. % [20]. Similarly, p-type iron oxide disks were produced by mixing powders of a-FezO3 and MgO so that 0 c M g N g + Fe I 20 at. %. The mixtures were pressed into pellets, heated in air at 134O-139O0C and rapidly cooled. The resulting mixed iron oxide disks had resistivities in the range lo3 -lo4Qcm. Both electrodes were found to be indirect bandgap semiconductors, with bandgap of approximately 2.3 eV. The 5% Mg-doped hematite gave at the same wavelengths an IPCE of 0.03 at 0.25 V [l]. No precise illumination conditions were mentioned. Both dihydrogen and dioxygen gases were produced with the unbiased device in 0.01 M NaOH, at the Fermi level of 0.75 V [ l , 221. The photocurrent density measured at a light intensity of 35 mW/cm was stabilized at 15 pA/cm2 for two weeks [ l , 221. The device worked also in 0.1 M NaZSO4solution with a photocurrent density of 5-10 pA/cmZ [20]. Also here the exact illumination condition was not given. Additional work by Somorjai and co-workers on the preparation and the properties of the Mg-doped p-type hematite was reported [25]. It is an interesting work since it is one of the very few articles concerning p-type hematite. Other workers have found p-type behavior of iron oxides, but no photoelectrochemical measurements were reported [9, 171. Sanchez and co-workers [19] tried to synthesize Mg-doped hematite by mixing with metallic Mg or MgO with following firing. However, their attempt was not successful. Iodine-doped hematite has also been studied. The iodine-doped thin films of iron oxide were obtained by spray pyrolysis. The condition for the preparation was not described in detail, apart that a mixture of an 80% ethanolic solution of 0.01 M iodine and 0.1 M FeC13 was used. Undoped films of 50 nm film thickness showed a maximum photocurrent density of 1.1 mA/cmz, while a 100 nm thick iodine-doped films had a maximum photocurrent density of 5 mA/cm2 at 0.82 V vs. NHE at pH 13 [37]. These measurements were performed with a xenon lamp with a light irradiation of only 150 mW/cm2. At the same condition, it was suggested by model calculation that an optimized stack of five iodine-doped hematite electrodes was expected to yield a photocurrent density of 15 mA/cm2.
102
Photo-oxidationof Water at Hematite Electrodes
It can be concluded that doped polycrystalline hematite indeed can be interesting for photoelectrolysis of water. The studied area has been intensively researched and some of the findings represent promising photoelectrochemical properties. In combination with the ease of preparation and good stability towards photocorrosion this make doped polycrystalline electrodes of hematite an interesting material for direct conversion of solar energy into dihydrogen. 3.4.3.d. Mixed Polycrystalline Oxides of Hematite
To improve the photoelectrochemical properties of hematite, mixing with other oxides, like a thin layer of TiOz deposit on polycrystalline hematite has been investigated [45]. The heterojunction electrodes were made by deposition of TiOz with CVD technique on hematite substrates prepared by flame oxidize iron metal. It was reported that holes generated in hematite did participate in the reaction at the TiOz/electrolyte interfacial, though an external bias was still required. The onset potential of the heterojunction moved cathodically to -0.46 V, towards the same value as that of TiOz on Pt in 0.1 M NaOH [45]. This is interesting since the cathodic shift of the onset makes dihydrogen evolution possible. However, the photocurrent remained unchanged at potentials over 0.14 V. No data for the bandgap, the resistivity or the stability of the material were mentioned. More complex oxides have also been investigated, like CdFez04 and PbFelz019 [SI. The former was prepared by solid-state reaction of the respective oxides and the latter was grown from a PbO-PbFz flux. The bandgap was 2.3 eV for both materials. Up in 0.2 M NaOH were 0.8 and 1.0 V, respectively. The photocurrent density of PbFelzO19was good, well established in the mA/cmZregion for an applied potential of 0.75 V vs. NHE. A 100 W Xe lamp was used, but the incident light intensity was not mentioned. The photocurrent density of CdFezO4 was considerable less, which was attributed to the higher resistivity, 500 SZcm compared to less than 10 SZcm for PbFelZOl9. 3.4.4. Nanostructured Thin Films of Hematite 3.4.4.a. Nanostmctured Films of Spherical Particles of Hematite
In nanosized particle film electrodes, photogenerated holes can be rapidly transferred to the semiconductor/electrolyteinterface and there be captured by the redox species in the electrolyte. In this way, the recombination losses can be diminished. This is of great importance for semiconductors like hematite with a very short hole diffusion length (2-4 nm). Another advantage is the large internal surface area, which characterize nanostructured semiconductor film electrodes. The latter decreases the current density per unit area of semiconductor / electrolyte interface. In general, few studies have been conducted on the photoelectrochemical properties of nanosized Fez03. Miyoshi et a1 [58] studied the PEC response of hematite nanoparticles (1.8 - 11.8 nm determined by TEM) prepared in Nafion and attached to a glassy carbon electrode. The photoelectrochemical experiments were carried out in aqueous solution of 0.1 M NazS04 containing 0.1 M sodium tartrate as a sacrificial reagent (pH 7). The photoresponse was in the PA-scale with bias. Interesting was the study of the band gap energy. The maximum size quantization achieved was 1.2 eV with positive shifts in the valence bands and negative shifts in the conduction bands with almost equal rate. Quantum confinement was also reported for nanostructured hematite incorporated in sodium mont-
Chapter 3; T. Lindgren, et al.
103
morillonite clay interlayers for the aim of photodecomposition of acetic acid [ 1181. The nanosized hematite particles in the clay interlayers was determined to be around 7 8, and they exhibited a bandgap ca. 0.28 eV greater than bulk hematite. Bjijrksten et a1 [44] designed and studied a hematite nanostructured electrode. The synthesis of the nanostructured electrode was clearly stated in the article and based on solgel deposition where the colloidal Fez03 was prepared by hydrolysis of FeC13. Unfortunately, investigation showed that the photoresponse of the nanostructured hematite electrode was very low. Illumination through the substrate/electrode interface (SEillumination) [ 1191 yielded an IPCEsE I 0.8% at a wavelength around 350 nm in 0.1 M NaOH aqueous solution with the potential being fixed at 0.4 V vs. SCE [44]. At the same conditions, illumination through the electrolyte/electrode interface (EE-illumination) [ 1191 gave an IPCEEEI 0.008%. The better efficiency for backside illumination is typical for nanostructured electrodes where the electron transport in the film is described by means of diffusion [ l o l l and where the current generation is most efficient close to the back contact [120]. The pronounced difference in IPCE values under SE and EE illumination showed that the collection of photogenerated electrons across the film is very inefficient. In a porous film consisting of interconnected nanometer sized semiconductor particles the effective surface area can be enhanced 1000-fold [ 1211. Therefore, nanostructured electrodes can be good for unravel the surface phenomena. By scrutinizing the effect of iodine on the performance of the electrode it was concluded that the effect of surface states was small in the nanostructured hematite electrode. It was stated that the bulk and grain boundary recombination remained dominant. This is in consistency with the report from Cherepy et aZ[43]. A large number of researcher groups are currently involved in work on nanostructured iron oxides films [16, 31-34,40,48,49, 118, 122-1291,Apart from hematite (~-FezO3),the iron oxide magnetite (Fe304) is also prepared and investigated in the nanoregime for magnetic studies [42, 130-1331. Recently, 30 nm thin films of Fe2O3-TiOZwere prepared on Pt substrates by coldwall low-pressure metallorganic chemical vapor deposition (MOCVD) [ 161. Iron(m> acetylacetonate [Fe(OzC5H7)3]and tetraisopropoxytitanium [Ti(0-i-C3H7)4]were used as vapor sources and heated at 130" and 30"C, respectively. Dioxygen was used as a reaction gas and nitrogen was used as a carrier gas. The thicknesses of the films were small, adjusted to around 30 nm by monitoring the deposition time. A maximum quantum yield (defined as percentage of injected electrons per absorbed photon) of 1.1% in 0.5 M HzS04 (pH 0.4) at 1.2 V was obtained for amorphous electrodes with a mole ratio corresponding to 0.2 Fe-0.8 Ti [16]. However, undoped polycrystalline Fez03 showed higher quantum yield (1.6%) under the same condition. Another finding was that the band gaps of films with different composition changed linearly with the fraction of Ti cation, from 1.9 eV for pure Fez03 to 3.1 eV for pure Ti02 [ 161. Although nanostructured films as well as colloidal solution of hematite have improved the understanding of the photoelectrochemical properties and its limiting factors intrinsic bulk recombinations are still the controlling factor of the photocurrent. However, with a deeper understanding of the fundamental properties of hematite and new synthesizing methods, there is still hope for improvements of the nanostructured materials in the application for photoelectrochemical water splitting. One promising approach is the purpose-designed electrodes of hematite nanorods.
Photo-oxidation of Water at Hematite Electrodes
104
3.4.4.b. Hematite Nanorods
Another work in the nano regime field is the development of hematite nanorods at our department [ l l , 14, 29, 301. Anisotropic crystallites elongated along c axis, e.g., where the dielectric constant is the highest and where the diameter of such nanorods matches the minority carrier diffusion length, reported to be 2-4 nm [30]. For thin film composed of such rod, photogenerated holes have a short distance to travel to reach the electrode/electrolyte interfacial region at the same time the excited electrons can move towards the back contact without passing any grain boundaries, see Fig. 3.5. Thus, the nanorod design facilitates a more efficient charge carrier separation, which should reduce recombinations of photoinduced charge carriers. Accordingly, IPCE measurements were promising. The IPCEsE and IPCEEE at 350 nm in a three-electrode system, with 0.1 M KI in water (pH 6.8) as electrolyte were determined to 13% and 5 %, respectively [ l l ] . Interestingly, in a two-electrode set-up with 0.5 M LiI + 50 mM 12 in ethylene carbonate/propylene carbonate (5050 by weight) as electrolyte a respectable IPCEsE value of 56 % at 340 nm was reported.
I
I
I
I
Fig. 3.5. Schematic representation of the electron transport through (a) spherical particles and (b) nanorods.
Considering there are no grain boundaries present in the material, the quite low photoresponse may be due to i) intrinsic recombinations centers located in mid bandgap states or ii) slow interfacial kinetics at the nanostructured semiconductor/electrolyte interface. To reveal to influence of the charge transfer kinetics on the photoresponse at hematite nanorods IPCEsE was studied as a function of the light intensity in different electrolytes. In an aqueous solution of 0.1 M NaOH (Fig. 3.5a) the highest light intensity yields the lowest IPCE~E,which indicates that most of the produced electron-hole pair recombine. In contrast, in a solution of ethylene carbonate/propylene carbonate (5050 by weight) containing 0.5 M r and 0.5 mM I2 in which r is a well-known hole scavenger [ 111, IPCEsE shows very small dependency with light intensity (Fig. 3.5b). This indicates that the material is able to handle the photogenerated electron-hole pairs fairly well as long as the hole kinetics at the interface is fast. The same result was obtained in a three-electrode set-up and purging the electrolyte in the 3-electrode vessel with N2 or bubbling with 0 2 did not affect the results.
Chapter 3; T. Lindgren, et al.
105
Since the presence or absence of O2 did not alter the photoresponse, it was concluded that the loss of conduction band electrons to dioxygen also in the 2-electrode experiment was negligible. Consequently, the oxidation of water by photogenerated holes causes the strong dependency of the photoresponse on light intensity. The slow water oxidation kinetics at the hematite interface favours the recombination of the photogenerated charges. Experiments on nanoporous Ti02 electrodes showed no dependency between IPCEsE and light intensity in various electrolytes, which confirm the fast oxidation kinetics at the Ti02 semiconductor/electrolyte interface [ 1341. It was also observed that the hematite photocurrent was higher, and a more linear relationship between photocurrent and intensity was observed when r was present as a hole scavenger compared to the experiment when only water was used. Indeed the experiments with 13-/I-indicated that hematite was capable of handling a fairly high rate of photogenerated electrons, as long as the holes are quickly withdrawn at the nanostructured semiconductor/ electrolyte interface (SEI).
400
350
450
500
550
600
650
700
Wavelength I nm
I
204A
i
15
“i
A
25 mWcm” 1 mwcm“
0
v
0.3 mWcm”
u %0
o
0.1 mwcm”
A
b)
E 5 0
350
400
450
500
L 550
600
650
700
Wavelength I nm Fig. 3.6. Action spectra (2-electrode set-up). Illumination from back-side (SE) in a) aqueous 0.1 M NaOH and in b) 0.5 M I-, 0.5 mM I2 in ethylene carbonate/propylene carbonate (5050% by weight).
106
Photo-oxidation of Water at Hematite Electrodes
It was concluded that by combining a hematite nanorod electrode with a suitable water oxidation catalysts, for instance platinum or ruthenium dioxide, the photoelectrochemical activity for direct water splitting applications should increase by a factor of 20. The promising initial study of hematite nanorod initiated a study of hematite nanorods for the aim of water oxidation [29]. IPCEsEand IPCEEEin a three-electrode set-up were determined at 350 nm to 2.3 % and 1.0 %, respectively. The fact that IPCEsE is twice the IPCEEE reveals that the collection of the photogenerated electrons across the hematite film is poor. Nevertheless, the ratio in the quoted investigation [29] is significantly smaller compared to the nanosized isotropic ‘random walk’ hematite particle system reported by Bjiirksten et a1 [44]. The photocurrent density at an intensity of 1 sun was in the pA/cm2 scale in 0.1 M NaOH.
3.5. Conclusion and Future Scopes Hydrogen production from water using hematite and solar energy is still in an experimental stage. Bulk recombination and slow kinetics at the semiconductor/electrolyte interface are the major causes to the low photoresponse for hematite electrodes in aqueous solutions. Due to different ways to report the photoresponse it is difficult to directly compare different studies. In summary it can be said that most attempts to increase the photoresponse has failed or shown little improvement. However, there are some exceptions. Some studies of doped hematite were promising and yielded photocurrents in or just below the mA/cm2region [2, 18, 19,23, 371. The studies of nanostructured hematite are interesting [ l l , 14,29, 30, 43, 441, mainly because they contribute to the understanding of the physical reasons for the poor charge transport properties of hematite, e.g. revealing the low influence of surface states [43, 441 and the importance of fast interfacial kinetics of the hole-water oxidation rate (291. However, it is important to keep in mind that there is no band bending in nanostructured hematite. Therefore, the results from studies of nanostructured hematite materials cannot be considered as valid for hematite materials in general. The most promising studies for water oxidation on hematite are the studies of spray pyrolysed nanometer thin films of hematite [35-381. High efficiency combined with ease of fabrication and low cost of the material make these films very interesting. Conversion efficiencies as high as 4.9% were reached at moderate bias [38]. Since hematite absorbs a great part of the incident solar irradiation energy a similar tandem set-up as described by Gratzel et a1 [92, 1351 would be at least as efficient for hematite as for WO3. The nanocrystalline dye-sensitised Ti02 cell could also be substituted with a long wavelength absorbing p-type material to work in combination with the hematite electrode in a pln assembly. Especially if the hematite is combined with a top layer of a cheap and efficient water oxidation catalyst the possibility for an efficient photoelectrochemical cell could be at hand.
Acknowledgment We like to express our acknowledgement to Swedish National Energy Administration for the financial support. Ph.D. J. A. Turner, NREL, Golden, USA and Ph.D. J. He and Ph.D. G. Boschloo, at the department of Physical Chemistry at Uppsala University are gratefully acknowledged for valuable discussions.
Chapter 3; T. Lindgren, et al.
107
REFERENCES 1. 2. 3. 4.
5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.
Turner J. E., Hendewerk M., Parmeter J., Neiman D. and Somojai G. A. J. Electrochem. SOC.,131,NO. 8, 1777-1783 (1984). Kennedy J. H., Anderman M. and Shinar R. J. Electrochem. Soc., 128, No. 11, 2371-2373 (1981). Leland J. K. and Bard A. J. J. Phys. Chem., 91, No. 19,5076-5083 (1987). Kiwi J. and Gratzel M. J. Chem. Soc., Faraday Trans., 83, No. 1,1101-1108 (1987). Launay J. C. and Horowitz G. J. Crystal Growth, 57, No. 1, 118-124 (1982). Kennedy J. H. and Frese K. W. J. Electrochem. Soc., 125, No. 5,709-714 (1978). McGregor K. G., Calvin M. and Otvos J. W. J. Appl. Phys., 50, No. 1,369-373 (1979). Kung H. H., Jarrett H. S., Sleight A. W. and Ferretti A. J. Appl. Phys., 48, No. 6, 2463-2469 (1977). Redon A. M., Vigneron J., Heindl R., Sella C., Martin C. and Dalbera J. P. Sol. Energy Mat. Sol. Cells, 3, 179-186 (1981). Wilhelm S.M., Yun K. S., Ballenger L. W. and Hackerman N. J. Electrochem. Soc., 126, No. 3,419-424 (1979). Beermann N., Vayssieres L., Lindquist S.-E. and Hagfeldt A. J. Electrochemical Soc., 147, NO. 7,2456-2461 (2000). Kennedy J. H. and Anderman M. J. Electrochem. Soc., 130, No. 4,848-852 (1983). Ishikawa K., Yoshioka T., Sato,\ T. and Okuwaki A. Hydrometallurgy, 45, No. 1, 129-135 (1997). Vayssieres L., Beermann N., Lindquist S.-E. and Hagfeldt A. Chem. ofMat., 13, No. 2, 233235 (2001). Cornel1 R. M. and Schwertmann U., The Iron Oxides, VCH, New York (1996). Kim H., Hara N. and Sugimoto K. J. Electrochemical Soc., 3, No. 3,955-960 (1999). GardnerR. F. G., Sweett F. andTannerD. W. J. Phys. Chem. Solids,24, 1183-1196 (1963). Kennedy J. H., Shinar R. and Ziegler J. P. J. Electrochem. Soc., 127, No. 10, 2307-2309 (1980). Sanchez H. L., Steinfink, H. and White H. S. J. Solid State Chem., 41, No. 1,90-96 (1982). Leygraf C., Hendewerk M. and Somorjai G. A. J. Phys. Chem., 86, No. 23,4484-4485 (1982). Kennedy R. S.a. J. H. J. Electrochem. Soc., 130, No. 4, 860-863 (1983). Turner J. E., Hendewerk M. and Somojai G. A. Chem. Phys. Lett., 105, No. 6, 581-585 (1984). Schmuki P., Buchler M., Virtanen S., Bohni H., Muller R. and Gauckler L. J. J. Electrochem. SOC.,142, NO. 10,3336-3342 (1995). Khader M. M., Lichtin N. N., Vurens G. H., Salmeron M. and Somojai G. A. Langrnuir, 3, NO. 2, 303-304 (1987). Leygraf C., Hendewerk M. and Somorjai G. J. Solid State Chem., 48, 357-367 (1983). Virtanen S.,Schmuki P., Bohni H., Vuoristo P. and Mantyla T. J. Electrochem. Soc., 142, No. 9,3067-3072 (1995). Ginley D. S.and Butler M. A. J. Appl. Phys., 48, No. 5 , 2019-2021 (1977). Butler M. A., Ginley D. S.and Eibschutz M. J. Appl. Phys., 48, No. 7, 3070-3072 (1977). Lindgren T., Wang H., Beermann N., Vayssieres L., Hagfeldt A. and Lindquist S.-E. Sol. Energy Mat. Sol. Cells, 71,231-243 (2002). Vayssieres L., Hagfeldt A. and Lindquist S.-E. Pure Appl. Chem., 72, No. 1,47-52 (2000). Armelao L., Granozzi G., Tondello E., Colombo P., Principi G., Lottici P. P. and Antonioli G. J. Non-Crystalline Sol., 1928~193,435-438(1995). Ding J., Tsuzuki T. and McCormick P. G. Nanostructured Mater., 8, No. 6,739-747 (1997). Li Q, and Wei Y. Mat. Res. Bull., 33, No. 5 , 779-782 (1998). Yu W. and Hui L. Mat. Res. Bull., 34, No. 8, 1227-1231 (1999). Itoh K. and Bockris J. 0. M. J. Appl. Phys., 56, NO. 3, 874-876 (1984). Itoh K. and Bockris J. 0.M. J. Electrochem. SOC.,131,No. 6,1266-1271 (1984).
108 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49.
50. 51. 52. 53. 54. 55.
56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75.
Photo-oxidation of Water at Hematite Electrodes
Khan S.U. M. and Zhou Z. Y. J. Electroanal.Chem., 357,407-420 (1993). Khan S.U. M. and Akikusa J. J. Phys. Chem. B, 103, No. 34,7184-7189 (1999). Yamaguchi I., Manabe T., Kumagai T., Kondo W. and Mizuta S. Thin Solid Films, 365, No. 1, 36-42 (2000). Wang S., Wang W., Jiao Z., Liu J. and Qian Y. Sensors and Actuators B, 69,22-27 (2000). Itoh K., Nakao M. and Honda K. J. Appl. Phys., 57, No. 12,5493-5498 (1985). Meldrum F. C., Kotov N. A. and Fendler J. H. J. Phys. Chem., 98, No. 1 7 , 4 5 0 6 4 1 0 (1994). Cherepy N. J., Liston D. B., Lovejoy J. A., Deng H. and Zhang J. Z. J. Phys. Chem. B, 102, NO. 5 , 770-776 (1998). Bjorksten U., Moser J. and Gratzel M. Chem. Mater., 6, No. 6, 858-863 (1994). Liou F.-T. and Yang C. Y. J. Electrochem. SOC.,129, No. 2,342-344 (1982). Iwanski P., Cuman J. S., Gissler W. and Memming R. J. Electrochem. SOC.,128, No. 10, 2 128-2133 (1981). Chatterjee S., Sarkar S. and Bhattacharyya S.N. J. Photochem. Photobiol. A: Chem., 72, No. 1, 183-187 (1993). Kan S.,Yu S., Peng X., Zhang X., Li D., Xiao L., Zou G. and Li T. J. Coll. Zntelface Sci., 178,673-680 (1996). Huo L., Li W., Lu L., Cui H., Xi S., Wang J., Zhao B., Shen Y. and Lu Z. Chem. Mater., 12, NO. 3,790-794 (2000). Candea R. M. Electrochemica Acta, 26, No. 12, 1797-1801 (1981). Candea R. M. Electrochemica Acta, 26, No. 12,1803-1808 (1981). Faust B. C., Hoffmann M. R. and Bahnemann D. W. J. Phys. Chem., 93, No. 17, 6371-6381 (1989). Kormann C., Bahnemann D. W. and Hoffmann M. R. J. Photochem. Photobiol., A: Chemistry, 48, NO. 1, 161-169 (1989). Pulgarin C. and Kiwi J. Lungmuir, 11, No. 2,519-526 (1995). Gurunathan K. and Maruthamuthu P. Znt. J. Hydrogen Energy, 20, No. 4,287-295 (1995). Ansari A., Peral J., Domenech X., Rodriguez-Clemente R. A. and Roig E. M. J. Photochem. Photobiol., A: Chemistry, 87, No. 1, 121-125 (1995). Bandara J., Mielczarski J. A. and Kiwi J. Lungmuir, 15, No. 22,7680-7687 (1999). Miyoshi H., Tanaka K., Uchida H., Yoneyama H., Mon H. and Sakata T. J. Electroanal. Chem., 295,71-78 (1990). Nozik A. J. and Memming R. J. Phys. Chem., 100, No. 31,13061-13078 (1996). Morin F. J. Phys. Rev., 93, No. 6, 1195-1199 (1954). Dare-Edwards M. P., Goodenough J. B., Hamnett A. and Trevellick P. R. J. Chem. SOC., Faraday Trans., 79,2027-2041 (1983). Harriman A. and West M. A. Photogeneration of Hydrogen, Academic Press, London (1982). Shakhnazaryan G. E., Sarkisyan A. G., Arutyunyan V. M., Arakelyan V. M., Turner J. A. and Vartanyan R. S. Rus. J. Electrochem., 30, No. 5,610-614 (1994). Fujishima A. and Honda K. Nature, 238, 37-38 (1972). Turner J. A. Science, 285, No. 5427, 687-689 (1999). Gratzel M. Cattech., 3, No. 1,4-17 (1999). Bard A. J. and Fox M. A. Acc. Chem. Res., 28, No. 3, 141-145 (1995). Takata T., Tanaka A., Hara M., Kondo J. N. and Domen K. Catal. Today, 44, No. 1, 17-26 (1998). Tryk D. A., Fujishima A. and Honda K. Electrochimica Acta, 45,2363-2376 (2000). Bolton J. R. Solar Energy Mater. Solar Cells, 38, No. 3,543-554 (1995). Yagi M. and Kaneko M. Chem. Rev., 101, No. 1,21-35 (2001). Licht S. J. Phys. Chem. B, 105, No. 27,6281-6294 (2001). Kinoshita K., Electrochemical Oxygen Technology, Wiley & Sons, New York (1992). Gurunathan K. and Maruthamuthu P. Int. J. Hydrogen Energy, 20, No. 4,287-295 (1995). Licht S., Wang B., Mukerji S.,Soga T., Umeno M. and Tributsch H. J. Phys. Chem. B, 104, NO. 38, 8920-8924 (2000).
Chapter 3; T. Lindgren, et al.
109
Khan S. U. M. and Bockris J. 0. M. J. Phys. Chem., 88, No. 12,2504-515 (1984). Heller A. Acc. Chem. Res., 14, No. 2, 154-162 (1981). Green M. A., Operating Principles Solar Cells: Technology and System Applications, Univ. of New South Wales, Kensington (1992). 79. Memming R., Semiconductor Electrochemistry, Wiley-VCH, Weinheim (2001). Bolton J. R., Strickler S. J. and Connolly J. S . Nature, 316, No. 6028,495-500 (1985). 80. 81. Heller A. and Vadimsky R. G. Phys. Rev. Lett., 46, No. 17, 1153-1156 (1981). 82. Kumar A., Santangelo P. G. and Lewis N. S . J. Phys. Chem., 96, No. 2,834-842 (1992). 83. Aspnes D. E. and Heller A. J. Phys. Chem., 87, No. 24,4919-4929 (1983). 84. Moser J. and Gratzel M. Helv. Chim. Acta, 65, No. 141, 1436-1444 (1982). 85. Butler M. A. J. Appl. Phys., 48, No. 5 , 1914-1920 (1977). Wang H., Lindgren T., He J., Hagfeldt A. and Lindquist S.-E. J. Phys. Chem. B, 104, No. 24, 86. 5686-5696 (2000). 87. Santato C., Ulmann M. and Augustynski J. J. Phys. Chem. B, 105, No. 5,936-940 (2001). 88. Santato C., Ulmann M. and Augustynski J. Con$ Abstract, IPS-2000, Snowmass, CO, USA (2000). 89. Kocha S. S., Peterson M. W., Arent D. J., Redwing J. M., Tischler M. A. and Turner J. A. J. Electrochemical SOC.,142, No. 12, L238-L240 (1995). 90. Khaselev O., Beach J., Bansal A., Collins R. T., Turner J. A. and Redwin J. M. Proc. 196th Meeting Electrochem. SOC.,Hawaii, 99-2, Abst. No. 1423 (1999). 91. Khaselev 0. and Turner J. A. Science, 280,425-427 (1998). 92. Augustynski J., Calzafem G., Courvoisier J. C., Gratzel M. and Ulmann M. Proc. 10th Int. Conf: Photochem. Stor. Sol. Energy, Interlaken, Switzerland, 229 (1994). 93. Zou Z., Ye J., Sayama K. and Arakawa H. Nature, 414,625-627 (2001). 94. Licht S., Wang B., Mukerji S . , Soga T., Umeno M. and Tributsch H. Int. J. Hydrogen Energy, 26,653-659 (2001). 95. Debnath N. C. and Anderson A. B. J. Electrochem. SOC.,129, No. 10,2169-2174 (1982). 96. Kurtz R. L. and Henrich V. E. Phys. Rev. B, 36, No. 6,3413-3421 (1987). 97. Catti M., Valerio G. and Dovesi R. Phys. Rev. B, 51, No. 12,7441-7450 (1995). Matsumoto Y. J. Solid State Chem., 126,227-234 (1996). 98. Bard A. J. and Faulkner L. R., Electrochemical Methods: Fundamentals and Applications, 99. Wiley & Sons, New York (2001). 100. McCann J. F. and Badwal S . P. S. J. Electrochem. SOC.,129, No. 3,551-559 (1982). 101. Sodergren S., Hagfeldt A., Olsson J. and Lindquist S.-E. J. Phys. Chem., 98, No. 21, 55525556 (1994). 102. Quinn R. K., Nasby R. D. and Baughman R. J. J. Mater. Res. Bull., 11, 1011 (1976). 103. Dimitrijevic N. M., Savic D., Micic 0. I. and Nozik A. J. J. Phys. Chem., 88, No. 19, 42784283 (1984). 104. Lindquist S.-E. and Vidarsson H. J. Mol. Cat., 38, No. 1, 131-155 (1986). 105. Gratzel M., Kiwi J., Morrison C. L., Davidson R. S . and Tseung A. C. C. J. Chem. SOC., Faraday Trans., 81, 1883-1890 (1985). 106. Battaglia V. and Newman J. J. Electrochem. SOC.,142, No. 5 , 1423-1430 (1995). 107. Ehrensberger K., Frei A., Kuhn P., Oswald H. R. and Hug P. Solid State lonics, 78, 151-160 (1995). 108. Tamaura Y., Hasegawa N., Kojima M., Sano T. and Tsuji M. Int. J. Hydrogen Energy, 22, No. 6,567-573 (1997). 109. Tamaura Y., Kojima M., Sano T., Ueda Y., Hasegawa N. and Tsuji M. Int. J. Hydrogen Energy, 23, No. 12, 1185-1191 (1998). 110. Herrera F., Lopez A., Mascolo G., Albers P. and Kiwi J. Appl. Cat. B: Environmental, 29, 147-162 (2001). 111. HenrichD. S. T. a. V. E. J. Phys. Chem. B, 105, No. 18,3872-3877 (2001). 112. Kiwi J., Suss J. T. and Szapiro S . Chem. Phys. Lett., 106, No. 1-2, 135-138 (1984). 76. 77. 78.
110 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135.
Photo-oxidation of Water at Hematite Electrodes
Nedeljkovic J. M., Nenadovic M. T., Micic 0. I. and Nozik A. J. J. Phys. Chem., 90, No. 1, 12-13 (1986). Nozik A. J., Williams F., Nenadovic M. T., Rajh T. and Micic 0. I. J. Phys. Chem., 89, No. 3, 397-399 (1985). Sayama K. and Arakawa H. J. Chem. Soc., Faraday Trans., 93, No. 8, 1647-1654 (1997). Baughman D. S. G. a. R. J. Mater. Res. Bull., 11, 1539 (1976). Saeki I., Okushi N., Konno H. and Furuichi R. J. Electrochem. SOC.,143, No. 7, 2226-2230 (1996). Miyoshi H. and Yoneyama H. J. Chem. Soc., Faraday Trans. 1 , 8 5 , No. 7, 1873-1880 (1989). Finnstrom B. and Tegner L. J. Electrochem. SOC.,130, No. 2,351-358 (1983). Lindquist S.-E., Hagfeldt A., Bjorksten U. and Lindquist S.-E. Sol. Energy Mat. Sol. Cells, 27, 293-294 (1992). Hagfeldt A. and Gratzel M. Chem. Rev., 95, No. 1,49-68 (1995). Cannas C., Gatteschi D., Musinu A., Piccaluga G. and Sangregorio C. J. Phys. Chem. B, 102, NO. 40,7721-7726 (1998). Okada H., Sakata K. and Kunitake T. Chem. Mater. Comm., 2, No. 2,89-91 (1990). Ennas G., Musinu A., Piccaluga G., Zedda D., Gatteschi D., Sangregorio C., Stanger J. L., Concas G. and Spano G. Chem. Mater., 10, No. 2,495-502 (1998). Gao M., Peng X. and Shen J. Thin Solid Films, 248, 106-109 (1994). Kang Y. S., Risbud S.,Rabolt J. F. and Stroeve P. Chem. Mater., 8, No. 9,2209-221 1 (1996). Du H., Cao Y., Bai Y., Zhang P., Qian X., Wang D., Li T. and Tang X. J. Phys. Chem., 102, NO. 13,2329-2332 (1998). Chen Q.W., Qian Y. T., Qian H., Chen Z. Y., Wu W. B. and Zhang Y. H. Mater. Res. Bull., 30, NO.4,443-446 (1995). Schimanke G. and Martin M. Solid State lonics, 136-137, 1235-1240 (2000). Zhao X. K., Heme P. J. and Fendler J. H. J. Phys. Chem., 93, No. 2,908-916 (1989). Zhao X. K., Xu S. and Fendler J. H. J. Phys. Chem., 94, No. 6,2573-2581 (1990). Okada H., Sakata K. and Kunitake T. Chem. Mater., 2, No. 2, 89-91 (1990). Cannas C., Concas G., Gatteschi D., Falqui A., Musinu A., Piccaluga G., Sangregorio C. and Spano G. Phys. Chem. Chem. Phys., 3,832-838 (2001). Lindgren T. and Lindquist S.-E. unpublished results. Gratzel M. and Augustynski J. Intern. patent applic., Intern. publication No. WO 01102624 A1
Chemical Physics ofNanostructured Semiconductors, pp. 111-1 34 A.I. Kokorin and D.W. Bahnemann (Eds.) 0VSP 2003.
CHAPTER 4
Photoelectrochemistry of Nanocrystalline Aggregates of Cyanine Dyes on the Semiconductor Electrodes Dmitry V, Sviridov Institute for Physico-Chemical Problems, Belarussian State University, Minsk, Belarus
Kevwords: Photoelectrochemistry, Cyanine dyes, nanocrystallites
4.1. Introduction Spontaneous aggregation of cyanine dyes in solution and at liquidsolid interfaces is perhaps the most prominent example of molecular self-organization. The optical properties of aggregates of cyanine dyes differ radically from those of single molecules, being strongly dependent on a number of associated dye molecules and on the aggregate’s structural features; the aggregates of cyanine dyes may be thus considered as the model systems for studies on the Q-sized molecular nanocrystals. The strong interactions existing between tightly packed cyanine dye molecules in the aggregate not only dramatically alter the spectroscopic and redox properties of a dye but also result in a coupling of optical transitions on different molecules similar to molecular crystals. Through this coupling, an optical excitation on a particular molecule can be effectively transferred to other molecules in the aggregate, i. e. the excitation becomes delocalized. The delocalization of exciton over an aggregate considerably affects the photosensitization behaviour of aggregated cyanine dyes adsorbed onto the surface of semiconductor substrate, ensuring the light-harvesting and concentration of excitation energy at the senstization centres. The present chapter concentrates on the photoelectrochemistry of aggregates of cyanine dyes adsorbed at the semiconductor electrodes with special emphasis given to the difference in the photosensitization action of polymer aggregate species and monomeric dye.
4.2. Spectroscopic Properties of Aggregated Cyanine Dyes The phenomenon of aggregation of cyanine dyes, that manifests itself with the appearance of new absorption bands at high concentrations or upon adsorption, was discovered independently in the late 1930s by G. Scheibe and E. Jelley [ 1,2]. According to their spectroscopic properties, the J-aggregates with unusually sharp absorption peak below the monomer transition band (so-called J-band) and H-aggregates showing hypsochromically shifted H-band are distinguished [3] (Fig. 4.1).
112
Nanocrystalline Aggregates of Cyanine Dyes
c q i ! i ! / ,o,-~z~ H-aggregate
J-aggregate
J; * I
CI
CI
RI
!\ ! !
R’ I
!
450
500
550
I
600
Wavelength, nm Fig. 4.1. The absorption spectra of thiacarbocyanine dye (R = C2H2, R’= C4H8S03-)in monomeric (M) state (dye concentration C = 2x10“ M, water-methanol mixture) and in J-aggregated state ( C = 5 ~ 1 0M, - ~0.1 M KCl aqueous solution); the J-aggregate formation occurs predominantly at the walls of optical cuvette. The absorption spectrum of homologous dye with R = CH3 and R’ = C4H8S03-, exhibiting spontaneous H-aggregation in aqueous electrolyte (C = 2 ~ 1 M, 0 ~ solution is cooled to 5OC; adapted from Ref. [4]);a sharp band at 507 nm corresponds to a dimer. The figure also shows possible arrangements of dye molecules in the J-aggregate and in the “stacked” H-aggregate (both in the adsorbed state).
These shifts result from the fact that electromagnetic intermolecular interactions in the aggregate couple optical transitions on different molecules [4-61, whereas the small linewidth inherent in J-aggregates can be explained by an exchange narrowing effect. In the field of photography, J-aggregated cyanine dyes have found an extensive practical use as the sensitizers permitting to expand the sensibility of silver halide materials to entire visible region [7, 81; for this application, the narrow linewidths of J- aggregates make them particular useful for colour photography, since they make photographic emulsion sensitive to specialized narrow wavelength regions. The tendency to form aggregates is mostly pronounced in the case of “compact” symmetrical mono- and trimethyne dyes [9, 10, 31, their aggregation being generally favoured by large substitutes; by contrast, merocyanines and non-planar pentamethynes having a lot of isomers show a poor propensity for aggregation [3, 111. Notwithstanding to the fact that the formation of J- and H-aggregates consisting of large arrays of densely packed chromophores is generally facilitated by the adsorption of dye molecules [lo], the
Chapter 4; D. V. Sviridov
113
process of aggregation is primarily affected by the peculiarities of dye structure and the adsorption regime rather than by nature of the solid surface. Cyanine dye molecules aggregate by stacking on top of each other. They are said to be perfectly stacked when the dye molecules overlap completely, Le., a stacking angle, which defines the angle between the transition dipole and the molecular axis of the aggregate, is of 90’. In principle, such arrangement of dye molecules is to provide a maximum interaction between the dye molecules and especially their n-electrons; however, due to steric hindrance problems, the stacking angle is to be less then 90’ even in “sandwich”-type dimers. In the large aggregates the stacking angle is dictated by interactions between all molecules assembled into an aggregate and corresponds to the maximum gain in free energy upon aggregation [ 121. The peculiarities of the structure of aggregate have a profound effect on its optical properties because a sign of excitonic interaction is dependent on the relative orientation of dye molecules comprising the aggregate. According to theoretical calculations, the absorption maximum is red-shifted (that corresponds to the negative coupling) when the stacking angle is 54’ or less, whereas the aggregate structures characterized by stacking angle greater than 54’ give positive coupling and, correspondingly, a blue-shifted absorption band [6]. The spectral shift from the monomer absorption increases with increasing the size of aggregate and exhibits almost complete saturation in case of polymeric aggregate species consisting of several tens of dye molecules [4]. Theoretical approximations of aggregate wavelength shifts relative to molecular orientations [ 13-15,6], together with the structural data stemming from the analysis of X-ray diffraction and microscopic reflection spectra of single crystals of cyanine dyes [16-181 and the results of polarization absorption measurements of dye aggregates formed on silver halide sheet crystals [19,20], allow to propose several structural models for dye aggregates [13, 14, 16, 211. The J-aggregates on surfaces have been assumed to have a two-dimensional structure [14], and the experiments on large AgBr microcrystals dyed with cyanine dyes [22] have confirmed this assumption for silver halide surfaces. The experimental values of spectral shift in the case of J-aggregates was successfully explained assuming a brick stone work structure [5, 14, 231; direct evidence for the brickstone arrangement was obtained by high resolution scanning tunnelling microscopy [24]. The brickstone model also provides an explanation of the well-known fact that the aggregation is favoured by mesosubstituents [3] which prevent dye molecules from being piled up with a large slip angle [25, 261 and can be tightly packed within the structure of aggregate [23]. Through the strong coupling of optical transitions of the individual dye molecules an optical excitation can be transferred to other molecules in the aggregate that leads to the formation of a delocalized excitonic state [5, 27, 281. However, due to strong coupling to phonons [29], these excitonic states at room temperature are only delocalized over -20 molecules at most [30, 311, i.e. the coherence domain can be smaller then the aggregate size which may amount several thousand molecules [29]. The aggregate size can be thus characterized using two different quantities: (i) the physical size, i e . , the number of monomer units and (ii) the effective aggregate size or the number of coherently coupled dye molecules. The excitons in the aggregates are believed to migrate on a large distance [32] through an incoherent energy transfer mechanism [33], which is restricted by the presence of aggregate edges and structural defects capable of behaving as the traps [34]. The exciton migration along dye aggregates substantially affects their photo-physical properties; for instance, the mobility of excitons is responsible for highly efficient
114
Nanocrystalline Aggregates of Cyanine Dyes
dynamic quenching of fluorescence of aggregated cyanine dyes (one molecule of pyrocatechol provides quenching of 1031O6 cyanine dye molecules associated into aggregates [35]). The delocalized nature of the J-aggregate excited state also plays a significant role in determining the photosensitization behaviour of aggregated cyanine dyes affecting the efficiency of electron injection from the excited aggregates to the conduction band of the sensitized semiconductor substrate. The investigations of the dynamics of excited state of J-aggregated cyanine dyes adsorbed onto silver halide microcrystals with the use of fluorescence lifetime measurements in a picosecond time domain have revealed that, as a general rule, the rate of electron injection increases with aggregate size [36, 371. This is consistent with a model in which the great exciton range in large aggregates increases the probability of exciton locating at a strongly couple site or a defect at silver halide surface [37, 381. Apart from the effects originated from the ability of the mobile Jaggregate exciton to sample more of the sensitized substrate surface, the multiplicity of states in the J-aggregate bands may also enhance the electronic coupling between the Jaggregate excited state and silver halide conduction band states, thus increasing the rate of electron transfer. These factors can provide a significant increase in rate constant of electron injection despite the fact that the increase in the J-aggregate size is to be accompanied with some decrease in the energy difference between the electron in the donor (the excited J-aggregate) and acceptor (the silver halide conduction band). In circumstances where the interaction between the excited aggregate and silver halide surface is insignificantly affected by the exciton mobility, the rate constant of electron injection exhibits some decrease with the aggregate size [39, 401 because in this situation the efficiency of charge exchange between the excited aggregate and the semiconductor becomes governed by the number of terminal dye molecules ( i e . , the exciton traps) which are the suitable places for the electron injection [40]. However, a large disorder typical of dye aggregates grown at silver halide grains leads to a strong enhancement in the rate of non-radiative relaxation that in turn causes the net photosensitzation efficiency to decrease with increasing aggregate size regardless of the trend in the charge transfer rate with aggregate size [26, 37,401. 4.3. Effect of Aggregate Formation upon Electronic Energy Levels of Cyanine Dyes
Much of our understanding of the relationship between cyanine dye energetics and the ability to photosensitize semiconductor substrates is based on electrochemical potential data. One-electron oxidation and reduction potentials (Eoxand Ered,respectively) have been shown to be linearly related to HOMO and LUMO levels of a dye and thus provide a measure of the expected electron affinity and ionization energy of a given dye [41-441. The electrochemical potentials, which are typically measured for monomeric dye dissolved in a nonaqueous solution, enable one to construct the energy level diagram providing an adequate description of charge exchange between the semiconductor substrate and the excited monomer dye adsorbed onto its surface [45,46]. Notwithstanding to the fact that the effect of aggregate formation upon the electronic energy levels of sensitizing dyes should be important, the energy level estimates used to confer the electron donating and hole trapping properties of different aggregated dyes serving as the sensitizers of silver halide photographic systems also relay on electrochemical data obtained for the unaggregated dye
Chapter 4; D. V. Sviridov
115
in solution [47, 481. However, the investigations of the perturbations in the diffuse reflectance spectra accompanying the treatment of dyed photographic emulsion in the redox buffer solutions have evidenced that E,, of J-aggregated neopentylenethiadicarbo-cyanine dye is more positive (by ca. 70 mV) than that of corresponding monomer [49]; due to decrease in the energy of optical transition upon J-aggregation, the energy of the excited state is to be lower than that of monomer by a much larger amount. The analysis of photographic data also allowed a conclusion that the lowering of the energy of the excited electrons when passing from monomer of cyanine dye to its aggregate is a general tendency [50, 511. Since the redox potentials of dye aggregates adsorbed onto platinum electrode exhibit a broad distribution centred at the formal oxidation potential E,,, the direct information on the ionization energy of the aggregated cyanines can be obtained, even in case of irreversibly oxidizing dyes, from the potential dependence of the fractional degree of dye oxidation, 8 , derived from the potential-step electrochemical measurements [52, 531. The analogue of Nernst equation describing the oxidation process can be written as:
e
RT
where E is the electrode potential, 6 is the coefficient representing the heterogeneity of redox sites in the adsorbed dye layer, the other symbols have their traditional meanings. A Gaussian distribution of redox potentials for the adsorbed molecules will modified the slope of a ln(l/e-1) vs. E plot in a following manner [54]:
6 = erf ( 2 0 )
(4.2)
where cs is the standard deviation of redox potentials. The experimental values of the slope for J-aggregated cyanine dyes have been shown to be markedly larger than the 59 mV/decade slope predicted by the classical Nernst equation for a one-electron process [52] that points to a broad distribution of redox sites; thus, for 3,3’-di-y-sulphopropyl-9-ethyl4,5,4’,5’-dibenzothiacarbocyanine, the coefficient 6 amounts 0.33 that corresponds to a Gaussian distribution of redox potential for the adsorbed molecules with the standard deviation of 117 mV. The intercept of ln(l/8-1) vs. E linear dependence with the potential axis yields the formal potential E,,, at the centre of distribution [52]. It is seen from Fig. 4.2 that both anodic (Dye I) and cathodic (Dye 11) shifts of the redox potential can be induced by aggregation. Taking into account that, in principle, the aggregation should cause a stabilization of the dye’s ground state [55], the observed variations in E,, can be attributed to the difference in the interaction between aggregated dye and electrolyte. In the presence of C1- and Br - ions, E,, of the aggregated cyanine dyes (unlike E,, of monomeric ones) becomes less positive, with E,, versus halide ion concentration exhibiting Nerstian-like behaviour (Fig. 4.2). Consequently, by choosing an appropriate electrolyte it is possible to change the position of electronic energy levels of the adsorbed photosensitizing aggregates.
Nanocrystalline Aggregates of Cyanine Dyes
116
,--.I'
I. I' I'
0.7 0.8
I I
I
I
I
I
I
Fig. 4.2. The effect of aggregation on the redox levels of Dye I and Dye 11.
4.4. Aggregates of Cyanine Dyes as the Sensitizing Agents in the Photoelectrochemical Systems
The adsorption of cyanine dyes onto the surface of semiconductor electrode, as a rule, is accompanied with the spontaneous (complete or partial) aggregation. The size of the aggregates formed and their structure are determined mainly by intermolecular interactions and can be varied to some extent by changing the absorption conditions (dye concentration, adsorption temperature, salt concentration in the supporting electrolyte, electrolyte pH, etc.) [39, 56, 571. The coadsorption of cyanine dyes with bis-quaternary salts of pyridine derivatives (PD), which have been initially developed as the promoters of dye aggregation in silver halide photographic systems [58, 591, offer some further possibilities for controlling the aggregation occurring at the surface of semiconductor electrode [53,601. It should be noted that the PDs exert different effect on the aggregation of cyanine dyes in silver halide emulsion and at the surface of oxide semiconductor (cf. Refs. [53] and [59]), the mechanism of their activity still remaining controversial [61]. For cyanine dyes which do not prone to aggregate spontaneously, the coadsorption with PDs allows one to generate different mixtures of monomers and aggregates on the electrode surface (Fig. 4.3, 4). A characteristic feature of aggregates of cyanine dyes, which distinguishes them from monomers, is the possibility of charge separation within aggregates. Owing to this, the aggregated cyanine dyes adsorbed on the surface of wide bandgap semiconductors of n-type have a property of generating (depending on the external bias) not only an anodic photocurrent, which is due to the injection of photoelectrons into the semiconductor electrode, but also a cathodic photocurrent resulting from the irreversible reduction of electron-accepting species (e.g., molecular oxygen) accompanied with the extraction of
Chapter 4; D. V. Sviridov
. 0.6
-
117
mwpcl q-(cH2)1rNp 2Br'
c I-
I
Et
CHI
J
CHi
PD I 1
Fig. 4.3. Photocurrent action spectra for W 0 3 electrode sensitized by Dye 11: in monomeric form (_---__-)., aggregated by coprecipitation with PD I (-) and PD I1 ). Electrode potential: +0.6V. Here and further, all potentials are referred to Ag/AgCl, Cl-(sat.) - ~cm-2.Electrolyte: 0.25 M Na2S04. electrode. Light intensity: 5 ~ 1 0 W (-.-.-.-.-e-
1-
I
PD 111
I
0.4 c-
c
2 L 2
8
c
0
0.2
c
a 0.0
Wavelength, nm
Fig. 4.4. Photocurrent action spectra for W 0 3 electrode sensitized by thiacarbocyanine dye in monomeric form (dashed line); aggregated by coprecipitation with PD I11 (solid line). Electrode potential: +0.6V. Electrolyte: 0.25 M Na2S04. electrons from the conduction band of semiconductor substrate [53,60, 62-69]. Since the latter process competes with the electron injection, a transition from the anodic
Nanocrystalline Aggregates of Cyanine Dyes
118
photocurrent to the cathodic one occurs in the vicinity of the flatband potential (Em),where the trapping of majority charge carriers from the conduction band by oxidized dye species produced during the course of sensitized photoreduction of oxygen becomes possible. By virtue of the fact that aggregates alone can contribute to the generation of the cathodic photocurrent (Fig. 4.4),the comparison of spectral distribution of the anodic photocurrent (caused by both monomeric dye and polymeric aggregate species) with that of the cathodic photocurrent allows identifying the M-, H-, and J-bands in the action spectrum in the case of partial or non-uniform aggregation of sensitizing dye [53,64, 691. The efficiency of generation of anodic and cathodic sensitized photocurrents may differ substantially for the aggregates of different cyanine dyes, in some instances (e.g., for Dye m) nothing but cathodic photocurrent being observed in a wide potential range (Fig. 4.5);in this case the anodic photocurrent is suppressed and the cathodic photocurrent onset potential approaches the oxidation potential of aggregated Dye III. On the other hand, the incorporation of the Dye 111 J-aggregates into porous oxide host through the adsorption of aggregated dye onto colloidal titania used for fabrication of nanostructurated film, followed by the annealing of the obtained electrode at 220°C results in the appearance of the anodic photocurrent (Fig. 4.5).
-15
I . l . l . l . l . r
-0.4
0.0
0.4
-10
-0.2
0.0
0.2
0.4
Electrode potential, V vs. Ag/AgCI
Fig. 4.5. Photocurrent - potential dependence for J-aggregated Dye I11 and Dye IV adsorbed onto the surface of nanostructurated W 0 3 electrode (curves 2,4) and incorporated in the W 0 3 film (curves 1, 3). Electrolyte: 0.25 M Na2S04. Light intensity: 2 mW cm-*.
The splitting of the J-band in the photocurrent action spectrum observed in this case (Fig. 4.6)points to the fact that the a buildup of compactness of the particulate oxide
Chapter 4; D. V. Sviridov
119
film is accompanied with a squeezing and partial fracture of the incorporated aggregates (the collapsing of aggregates in the oxide matrix leads to the formation of smaller aggregates and not the monomeric species that is evidenced by the formation of similar doublets in the action spectra of both anodic and cathodic photocurrents - Fig. 4.6).
0.5
2
1 b
4 0.0
-1
450
525
600
675
750
Wwelength, nm Fig. 4.6. Photocurrent action spectra (curves 1,2) and absorption spectrum (curve 3) of aggregated Dye 111embedded in the nanostructurated Ti02film; electrolyte: 0.25 M Na2S0,; light W ern.'; electrode potential: +0.3 V (curve l), -0.2 V (curve 2). Absorption intensity: spectrum of dyed Ti02colloid (50 mmolfl of titania + 1 mmol/l of Dye 111) used for fabrication of volume-sensitized nanostructurated TiOz electrode. Dye I11 exhibits Jaggregation upon adsorption at Ti02particles.
Low efficient generation of anodic photocurrent by the adsorbed aggregates of DyeIII can be attributed to the presence of bulky substituents (alkyl chains with -SO3groups at their end) which not only provide an oriented adsorption but also have an impact on the reaction coordinate for electron transfer from the exited dye aggregate to the conduction band. Contrastingly, the analogous Dye IV having no large substitutes produces in the aggregated state both anodic and cathodic photocurrents (Fig. 4.5). The same tendency can be traced for other pairs of homologs with and without alkylsulphonate moietes [63, 641. In the presence of large substituents the rate of electron injection diminishes and becomes lower than the rate of photoinduced charge transfer to molecular oxygen as well as the rate of intermolecular conversion that results in the vanishing of the sensitized anodic photocurrent. The alkylsulphonate groups, however, do not hamper the charge transfer from the conduction band to the relatively long-lived dye cation-radicals produced in a consequence of the oxygen photoreduction that allows the aggregates of cyanine dyes to induce sensitized cathodic photocurrent at the n-type semiconductor electrode. On the other hand, once dye aggregates are embedded in the nanostructurated oxide host, the conditions for electron injection will become more favourable permitting the analogous Dye III and
120
Nanocrystalline Aggregates of Cyanine Dyes
Dye IV to operate in a similar manner generating anodic current under positive bias and cathodic photocurrent under negative bias. In case of dye-sensitized photographic emulsions, the effect of bulky substituents on the charge exchange between excited dye aggregates and emulsion grains is much less pronounced as against semoconductor oxides due to a strong tendency toward formation of epitaxial aggregates of cyanine dyes at the surface of silver halide grains [22, 701. The orientation of J-aggregates at the surface of AgBr microcrystals can be suggested to be responsible for the four-fold increase in the rate of electron injection observed when coming from cubic crystals to octahedral ones [37] because the higher surface concentration of bromide ions characteristic of AgBr octahedra [71] may provide more favourable orientation of the adsorbed dye aggregates. It should be emphasized that the results of the photoelectrochemical investigations can be extended to the case of photographic systems only to some degree. In dye-sensitized photographic emulsions, a latent image, which is constituted by silver nanoclusters formed due to reduction of the interstitial silver ions by the injected photoelectrons, suffers partial degradation as a consequence of its interaction with photoproduced holes Dye' (so-called effect of self-desensitization). Diffusion of photoholes within aggregates adsorbed on silver halide can vastly enhance the parasitic oxidation of latent image centres unless these centres are isolated from dye by selective adsorption of organic substances [61]; as the result, the suppression of these desensitization processes and the enhancement of the efficiency of electron injection into emulsion grains appear to be equally important to the attainment of high photographic sensitivity in the sensitized spectral range [61,72]. The photoinduced reduction of molecular oxygen by the excited aggregates of cyanine dyes, which was discussed above, also leads to the formation of Dye' species, being thus desensitizing in nature. In particular, this provides an explanation of highly-effective desensitization of photographic emulsions by molecular oxygen observed within the Jbands of aggregated cyanine dyes (e.g., J-aggregated Dye II [73]).
4.5. Photoelectrochemical Spectral Sensitization by Partially Aggregated Cyanine Dyes The energy diagram in Fig. 4.2 evidences that the partial aggregation of cyanine dyes should result in a splitting of a ground term of sensitizer, this splitting being dependent on the ionic composition of the contacting electrolyte. The observed difference in oxidation potentials of cyanine dye monomers and aggregates is a driving force for charge exchange between aggregated and unaggregated dye forms during the course of sensitizing process. It is apparent that the lateral charge exchange between the components of partially aggregated sensitizer is to be favoured by the mobility of photoholes produced in the aggregates. The information about the photoinduced charge transfer between aggregates and monomeric dye species coexisting at the surface of semiconductor electrode can be obtained using the selective photooxidation of aggregates or monomers. Thus, the extended monochromatic irradiation of positively-biased semiconductor electrode sensitized with partially aggregated imidacarbocyanine Dye I within monomer absorption region leads to a decrease of sensitized photoresponse predominantly within M-band (Fig. 4.7); as the result, the J-band in the photocurrent action spectrum becomes more pronounced upon exposure. By contrast, the excitation of photooxidation reaction in the region of aggregate absorption is accom-panied
Chapter 4; D. V. Sviridov
121
with the rapid degradation of M-band. The observed metamorphoses of photocurrent action spectra for partially aggregated cyanine dye are evidence that the photoexcitation of aggregated Dye I induces the oxidation of its monomer. On the other hand, in contact with 2.5 M LiCl the photooxidation of aggregated Dye I does not induce the oxidation of monomolecular form of this dye that correlates well with the transformations of energy diagram induced by C1- ions (Fig. 4.2).
9 1
I
I
PD IV
_ _ - - - -J- ,,,
M
E=+0.6 V
0.45
$0.401
a
1
’
1 -1
n
,
,
500
550
\
0.35 600
650
Wavelength, n m
Fig. 4.7. Photocurrent action spectra for W 0 3 electrode sensitized by Dye I (dashed line) and Dye I coprecipitated with PD IV (solid line). Spectral distribution of the relative variation in photocurrent Aiph/iph = 1 - iph(t)/iph(0),where iph(0)and iph(t) are measured before and after illumination of the positively-biased Dye I:PD IV-sensitized W 0 3 electrode ( E = 0.6 V) at 550 nm for 10 min. Electrolyte: 0.25 M Na2S04.
The selective photooxidation experiments not only provide an information about the direction of lateral charge transfer between monomer and aggregate during the course of photosensitization process but also pennit better identification of aggregates. Thus, in the case of thiacarbocyanine Dye II aggregated through the coprecipitation with PD IV, the changes in the photocurrent spectrum produced by the monochromatic illumination in the absorption band of monomer or aggregate are consistent with the induced oxidation of dye aggregates accompanying the photoexcitation of monomers, whereas a synchronous degradation of red- and blue-shifted bands in the action spectrum (Fig. 4.8) suggests that the “herribone”-structured aggregates with two allowed transitions [131 are formed at the electrode surface rather than a mixture of H- and J-aggregates.
Nanocrystalline Aggregates of Cyanine Dyes
122
4
I ’
C
‘I
E
5 0.2 0 B 0
c
n
‘ ‘I \,
\ i
t-?&4
0.0
1
E=+0.6 V
E=-0.1 V
Wavelength, nm
Fig. 4.8. Photocurrent action spectra for WO3 electrode sensitized by Dye 11: in monomeric form (dashed line); partially aggregated by coprecipitation with PD IV (solid line). Spectral distribution of the relative variation in photocurrent upon illumination of positivelybiased electrode ( E = 0.6 V) at 560 nm for 10 min. Electrolyte: 0.25 M Na2S04.
Talung into account that the photooxidized cyanine dye molecules, being unstable cation-radicals, tend to undergo irreversible conversions, the rate constant of charge exchange between monomers and aggregates can be obtained from the measurements of the lifetime, 7, of the photooxidized dye molecules at the surface of dye-sensitized semiconductor electrode in contact with the indifferent electrolyte containing no redox species. For this purpose, the potential of illuminated sensitized electrode is modulated rectangular between the potential El, sufficiently positive to provide the dye photooxidation with significant rate, and potential E2 close to Em, at which effective electrochemical reduction of photooxidized dye can occur [53, 74, 751. Under these conditions, the dependence of the relative variation in the sensitized photocurrent, &ph/iph, on the modulation frequency, v, is governed by the following equation [74]: (4.3)
where iph(0) and iph(t) are the photocurrents measured at the potential El before and after extended illumination under modulated polarization (Fig. 4.9), r D is the surface concentration of dye, s is the rate of dye photooxidation, t is the illumination period while the potential is modulated. At sufficiently high modulation frequency the dye phooxidized at El has a chance to undergo complete reduction before its degradation and the photocurrent measured before and after illumination remains the same. As the modulation frequency decreases, the &iph/iph decreases and exhibits saturation at a level corresponding to the relative photocurrent decay under illumination at constant polarization.
Chapter 4; D.V. Sviridov
123
.-'.r.cnO .1n
w
0.1
1
10
-time
Fig. 4.9. (a, top) The GiPh/iphvs. V 1dependence for W 0 3 electrode sensitized by Dye 11: in monomeric form (m); partially aggregated by coprecipitationwith PD IV (0).The excitation wavelength: 560 nm. f = 20 s. The total surface concentration of Dye 11: lo-' mol ern-'. Electrolyte: 0.25 M
Na2S04.(b, bottom) The potential-timeprogramme and corresponding photocurrent-time curves used for z evaluation. Hatched areas indicate the exposure periods. Using Eq. (4.3), the value of z can be found from frequency dependence of Gi,h/i,h by the curve-fitting method. The lifetime value of 0.12 k 0.02 s was determined at 560 nm for oxidized monomeric Dye II, almost the same (within the error limits) value of z being obtained at 630 nm for photooxidized J-aggregated Dye 11 (the aggregates were formed by coprecipitation with PD IV). On the other hand, in the presence of dye aggregates there occurs a decrease in the lifetime of photooxidized monomer, the extent of this decrease being the greater, the higher the concentration of aggregated dye. The photooxidation of Dye 11 monomer thus occurs at lower concentration of Dye' species that is consistent with the supersensitization of Dye 11 monomers by adjacent aggregates. The rate constant of supersensitization, k,,, may be evaluated from the lifetime measurements for monomer contacting with aggregates as follows:
where zo is the lifetime of oxidized monomeric Dye I1 in the absence of aggregates, Tss is the acting concentration of aggregated Dye II involved in the reaction of charge exchange with monomeric Dye II. Using the rssvalues obtained from photocoulometric measure-
124
Nanocrystalline Aggregates of Cyanine Dyes
ments, the k,, was found to be (l-2)x109 cm2mol-'s-', this value being almost not dependent on the monomer-to-aggregate mole ratio [53]. In a similar manner the rate constant of supersensitization by the reducing agents (e.g., hydroquinone, QH2) may be determined from the dependence of z on the QH2 concentration by relation
Thus obtained k,, value is of ca. 4x104 M's-' for both monomeric and J-aggregated Dye II [53]. A comparison of the efficiency of self-supersensitization and supersensitization by QH2 shows that the association of only 15% of the adsorbed DyeII molecules into aggregates results in the supersensitization effect similar to that produced by104 M QH2. In the presence of hydroquinone, the quasi-steady-state photocurrent sensitized by partially aggregated Dye II will be governed by the following relations:
(4.7)
where
e560 -630 zph ,zph are the photocurrents generated under the illumination at 560 and 680 nm
(in the M-band and in the J-band, respectively); k l , k2, k3, k4, k5 are the rate constants describing the deactivation of the excited state, electron injection, recombination, supersensitization by QH2, supersensitization of monomeric Dye II by aggregated one. A double reciprocal plots:
where iph(0), iph(=) are the photocurrent values at [QH2]= 0 and [QH2] + 00, respectively, yield a straight line (Fig. 4.10), while the inverse of the slope gives Stern-Volmer constant, Ksv. The difference of the reciprocal of Stern-Volmer constants for monomer ( K:,60 ) and aggregate ( K,6,3' ) yields an estimate of k5/k4 which is in fair agreement with the results of life-time measurements; this allows a conclusion that the recombination rate is little affected by aggregation. Hence the perturbations in the photocurrent action spectrum observed in the presence of hydroquinone (Fig. 4.10) are caused by the suppression of charge exchange between the monomeric and polymeric forms of partially aggregated
Chapter 4; D. V. Sviridov
125
sensitizer in the presence of the excess of reluctant, not by the difference in the efficiency of supersensitization of monomeric and aggregated dye by hydroquinone. I
I
Wavelength, nm
Fig. 4.10. Photocurrent spectra with (curve 1) and without (curve 2) M hydroquinone (QH2)for Dye I1 partially aggregated at the surface of W 0 3 electrode with the use of PD I11 and the spectral distribution of the relative variation in photocurrent upon insertion of QHz in the electrolyte (curve 3). The inset shows the Stern-Volmer plots at 560 and 630 nm.
4.6. Spectral Sensitization of Semiconductor by Mixture of J-aggregated Cyanine Dyes
The effect of lateral charge transfer has a dramatic impact not only on the photosensitization behaviour of non-uniformly aggregated cyanine dyes but also on the efficiency of sensitization of semiconductor electrode by coprecipitated aggregated dyes, unless these dyes form amalgamation-like mixed aggregates. For the photoinduced interaggregate charge transfer to be effective, the aggregated dyes involved in the lateral charge exchange should differ by the value of E,, that allows the aggregates of one dye to behave as the traps for photoholes produced upon excitation of aggregates of another dye. As for the partially aggregated cyanine dyes, the information about the direction and the dynamics of charge transfer between coprecepitated aggregates of different cyanine dyes can be derived from the metamorphoses of the photocurrent action spectrum produced by the extended monochromatic illumination of the sensitized semiconductor electrode within the absorption band of one of aggregated dyes [60,691. Thus, for example, the excitation of photooxitation reaction within the absorption band of Dye V J-aggregates induces the oxidation of coprecipitated aggregates of Dye VI that manifests itself in a rapid degradation of the peak in the action spectrum corresponding to J-aggregated Dye VI (Fig. 4.11); by contrast, the J-aggregates of Dye V remain almost intact during photooxidation of Jaggregated Dye VI. The observed alterations in the shape of photocur-
Nanocrystalline Aggregates of Cyanine Dyes
126
rent action spectrum points to the fact that photoholes produced in J-aggregates of Dye V upon illumination are effectively captured by J-aggregates of Dye VI which is in a good agreement with the energetic diagram for binary dye mixture constructed on the basis of electrochemical data (Fig. 4.11). Since the peaks in the spectrum of photocurrent sensitized by Dye V + Dye VI mixture are only partially overlapped, the rate constant of the interaggregate charge transfer can be deduced from the time dependence of photocurrent measured at 580 nm (Le., beyond the Dye V photosensitization region) during the course of the extended illumination of the sensitized electrode at 535 nm [60].Thus estimated rate constant of the induced oxidation of J-aggregated Dye VI in Dye V + Dye VI mixture amounts IO7 cm2mol-'s-', with nothing more than 25% of adsorbed Dye VI molecules being involved into the lateral interaggregate charge transfer as evidenced by the photocoulometric measurements [60]. I
0.6
1
'
I
-
I
"
I
'
I
'
I
'
1
-
"Eu 6 5.
0.3
c
0
500
520
540
560
580
600
620
Wavelength, nm
Fig. 4.11. Photocurrent action spectra for W 0 3 electrode sensitized by Dye V + Dye VI mixture and spectral distribution of the relative variation in photocurrent upon illumination of positively-biased electrode ( E = 0.6 V) at 535 and 580 nm for 75 s. r D y e v(J): rDyeV[(J) = 1:2.5.According to the photocoulometric measurements, ca. 25% of the adsorbed Dye VI molecules are involved in the lateral charge exchange with J-aggregated Dye V. Electrolyte: 0.25 M Na2S04.Sensitizing dyes are aggregated with the use of PD 11.
Chapter 4; D. V. Sviridov
127
Thus obtained rate constant of interaggregate charge transfer is less by two orders of magnitude than the rate constant of photoinduced charge transfer between monomers and aggregates of cyanine dyes adsorbed on the semiconductor surface, i. e., the selfsupersensitization effect resulted from the supersensitization action of one dye form (monoor polymolecular) with respect to the other appears to be less pronounced in the case of a mixture of dye aggregates. In the photographic systems which are extremely sensitive to any change in the kinetics of the reactions responsible for the formation or degradation of latent image, the interaggregate charge exchange that promotes photohole transport in the adsorbed dye layer makes it possible the manifestation of both supersensitization effects [76] and specific desensitization ones [61]. 4.7. Photoelectrochemical Behaviour of Thin Films of Aggregated Cyanine Dyes
The possibility of separation of photoinduced charges within aggregates of cyanine dyes and the efficient interaggregate charge exchange furnish a pronounced photoelectrochemical activity for the thin films built of dye aggregates. The intrinsic photoelectrochemical properties of cyanine dyes in a condensed state were investigated using thin films of merocyanines [77-791 and carbocyanines [80, 63, 681 fabricated via vacuum evaporation [77, 791 as well as by spraying [79] or casting [80, 63, 681 of non-aqueous solution onto platinum and IT0 electrodes. The soaking of thus obtained films in aqueous solution may result in the molecular rearrangements leading to the appearance of the pronounced Jband in the absorption spectrum [77]; in some cases, the formation of J-aggregates in the film can also be induced by heat treatment [77]. High permeability of dye layer to electrolyte has a strong impact on the mechanism of photocurrent generation. The films of cyanine dyes function as the specific photogalvanic cells in which the collection of photogenerated charge carriers occurs due to the redox activity of dye layer. In these circumstances, only irreversible photoelectrochemica1 reactions proceed effectively; thus, the merocyanine dyes characterized by negative reduction potentials are capable of photoreduction of proton to molecular hydrogen [78], whereas the photoelectrochemical reduction of molecular oxygen under cathodic bias is typical for carbocyanines [80]. By virtue of the fact that the charge transport in the cyanine dye film occurs due to redox conductivity, the onset potential of the cathodic photocurrent, Ezn,agrees closely with the oxidation potential of aggregated dye, the change in the E,, in the presence of halide ions leading to the corresponding shift of Ezn [80]. High permeability of dye film makes it possible a direct reduction of molecular oxygen at the back electrode that results in the lowering of cathodic photocurrent under high negative biases (Fig. 4.12). For the same reason, the photopotential up, appears to be dependent on the exchange current for oxygen reduction on the metal back contact, so that
(4.10) where io is the exchange current, iphis the photocurrent generated under illumination. Thus, for the Dye III film, u p , increases over an order of magnitude as one goes from a platinum
Nanocrystalline Aggregates of Cyanine Dyes
128
back electrode (Uph= 10 mV at light intensity of lo4 W cm-') to a titanium one (Uph= 120 mV) . 0.0
0.0
-0.1
-0.1
a
a
g
f
f
c-
c
K
K
2 L
? L
a -0.2 s
=I
-0.2g c
0
0
c
.c L
n -0.3
I
500
.
,
550
.
,
600
.
,
650
Wavelength, nm
-0.3 0.2
0.3
0.4
Electrode potential, V
Fig. 4.12. (a) Photocurrent action spectra for Dye I11 layer at Pt electrode (curve 1); for Dye IV layer before (curve 2) and after (curve 3) heating at 120 "C.Electrode potential: +0.25 V. Electrolyte: 0.25 M Na2S04.(b) Photocurrent-potentialdependence for Dye I11 layer (curve 4,5) and Dye IV layer (curve 6) at Pt electrode; Dye IV was aggregated by heating at 120 OC.Electrolyte: 0.25 M Na2S04(curves 4,6);2.5 M NaBr (curve 5). Excitation wavelength: 640nm (curves 4, 5); 620 nm (curve 6).
The exciton migration within aggregates of cyanine dyes and the possibility of oxygen diffusion into the porous dye film result in a bulk generation of photocurrent [80]. Photoholes produced due to the oxidation of excitons by molecular oxygen diffuse to the back contact. The diffusion coefficient of charge carriers in dye layer (D,) can be evaluated from the potential-step chronoamperometric measurements in the indifferent electrolyte. Considering dye film as a thin-layer cell, the current vs. time dependence can be described as follows [81]:
(4.1 1)
where C, is the concentration of redox sites in the dye film, d is the film thickness. For very short times, where the current-time behaviour follows that for semi-infinite diffusion, the plot i vs. t"2 is linear and shows a zero intercept, the slope of this plot yielding CcDc"2[81]. In the case of thin film of Dye III(Fig 4.13), the diffusion coefficient was estimated to be of ca. cm2 s-l, essentially independent of the ionic composition of the contacting electrolyte. Taking into account that the characteristic time 2, = d2/D, can be obtained by fitting Eq. 4.11 to the current decay curve (Fig.4.13), the limiting current through the dye
Chapter 4; D. V. Sviridov
0.0
0.5
1.o
1.5
2.0
129
2.5
+-‘I?, S - l R
Fig. 4.13. Current-time curve (in form i vs. t-ln) for potential-step oxidation of Dye 111 layer on a platinum electrode. Electrolyte: 0.25 M Na2S04.
0.4
1
1
10.2
0.0
5 -0.4 +-
c
2 L
3 0
0 0
c)
c
a
-0.8
Fig. 4.14. Photocurrent-potential dependence for W03/Dye 111heterojunction (curves 1-3), Ti02/ Dye 111 heterojunction (curves 4-7), Ti02/Dye IV heterojunction (curve 8) and WO$Dye IV heterojunction (curve 9). The surface of Ti02 electrode is modified via photocatalytic deposition of Ag nanoparticles prior to deposition of Dye I11 layer (curve 7). Dye IV was aggregated by heating at 120 O C . Electrolyte: 0.25 M Na2S04, pH 6.05 (curves 2,5,7-9); 0.25 M Na2S04 + H2SO4, pH 1.55 (curves 3,4); 2.5 NaBr (curves 3, 6). The excitation wavelength: 640 nm W ern-'. The inset shows the pH depen(curves 1-7); 620 nm (curves 8,9). Light intensity: dence of I?,,, for Ti02/Dye 111 heterojunction and pH dependence of Em for TiOz electrode.
Nanocrystalline Aggregates of Cyanine Dyes
130
film can be evaluated using CcD,"2 and d2/Dcvalues (i.e., without invokmg the data on the film thickness which is hard to be measured accurately):
ifi, = nFCcDcd-' = nFCcD~/2rc-1'2= 8 mA cm1.2
(4.12)
The photoelectrochemical activity inherent in thin films of aggregated cyanine dyes permits them to act as the spectral sensitizers of wide bandgap semiconductors [69]. It is seen from Fig. 4.14 that the photoelectrochemical behaviour of "semiconductor/dye film" heterojunctions fabricated by deposition of -200 nm-thick films of cyanine dyes on the surface of TiOz and W 0 3 electrodes, bears close similarity to that of semiconductor electrodes sensitized by the adsorption of dye aggregates. Thus, both anodic and cathodic photocurrents can be generated under actinic illumination, the efficiency of the photoanodic and photocathodic processes and the potential at which photocurrent changes its direction being dependent on dye and semiconductor substrate [69]. 0.2
T
0.1
0
P
t '
=2
0.0
2 0 0
t
-0.1
-0.2 I
I
I
550
600
Wavelength, nm
650
Fig. 4.15. Photocurrent action spectra for ITO/W03/Dye IV heterojunction under illumination through the transparent back contact (curves 1,2,5,7);through the solution (curves 3 , 4 , 6 , 8 ) . Dye IV was aggregated by heating at 12OOC (curves 1,3,7,8).Electrolyte: 0.25M Na2S04.
The action spectra of W03/Dye III heterojunction shown in Fig. 4.15 evidence that the anodic photocurrent generated under illumination through the transparent back contact is higher than under illumination through the dye/electrolyte interface. The generation of anodic photocurrent thus occurs at the semiconductor/dye boundary due to the lack of mechanism of electron transport in the film of cyanine dye, Le., only the aggregates contacting with the semiconductor surface can contribute into anodic photoresponse. Contrastingly, in the case of films of cyanine dyes there exists a possibility of effective transport of photogenerated holes due to translation of electron vacancies within aggregates accompanied with an interaggregate charge exchange; consequently, in the absence of
Chapter 4; D. V. Sviridov
131
rectifying back contact, the cyanine dyes in a condensed state exhibit the photoresponse of p-type and generate nothing but cathodic photocurrent. The large alkylsulphonate groups located outside the confines of dye aggregate hamper the charge injection into the conduction band; as the result, the thiacarbocyanine Dye 111 and Dye IV of analogous structure, which form spectroscopically similar J-aggregates characterized with very close oxidation potentials, differ as to their photoelectrochemica1 behaviour in the “semiconductor/dye film” heterojunction: W03/Dye 111 electrode generates cathodic photocurrent and no other, whereas W03/Dye 111 electrode is able to generate both anodic and cathodic photocurrents (Fig. 4.14). Bcause of high permeability of dye film to electrolyte the potential drop in it is nil and the process of electron injection into oxide component is potential-independent over a wide potential range (Fig. 4.14). At the vicinity of flatband potential of semiconductor oxide, where the extraction of electrons from conduction band becomes possible due to the decrease in band bending, the photocurrent changes its direction into cathodic one (Fig. 4.14). The generation of cathodic photocurrent, which is due to reduction of molecular oxygen, occurs not only at the “dye filmlelectrolyte” boundary but also in the bulk of the porous dye film. The efficiency of the cathodic photoelectrochemical process is governed by the kinetics of charge transfer through “semiconductor/dye film” junction, whereas a gradual decrease of the cathodic photocurrent observed at high negative biases (Fig. 4.14) is due to the direct reduction of oxygen at the oxide substrate. The extraction of electrons from the conduction band of semiconductor, which occurs in the potential range where some band bending still retains, is facilitated by the clusterization of donor defects typical of heavily doped Ti02 and W03 used (Nd= 1019 donors ~ m -resulting ~) in formation of channels for almost unimpeded charge transfer from conduction band to dye layer. Modification of Ti02 surface by deposition of silver nanoparticles resulting in a radical improvement of the permeability of the Schottky barrier due to the formation of “shallow” surface states in the forbidden zone [82] leads to further enhancement of the cathodic photocurrent (Fig. 4.14). Due to the substantial difference between the flatband potentials of Ti02 (EFB = -0.67 V vs. Ag/AgCl,Cl- (sat.) at pH 6.05 [83]) and W03 (Em = -0.18 V [84]), the cathodic photocurrent onset potential in case of TiOz/Dye VI electrode appears to be negatively shifted as compared to ECon observed for W o n y e I11 electrode; in the latter case the photocurrent onset potential approaches the oxidation potential of aggregated Dye 111, being determined by the intrinsic photoelectrochemical properties of a dye layer deposited onto the oxide surface. Correspondingly, the negative shift of the ground term of aggregated dye in the presence of Br- ions in the contacting solution is accompanied with the cathodic shift of Conin case of W03/Dye I11 electrode, whereas the changes in the position of energy levels of the semiconductor by varying pH of the solution leave Po” unaffected - Fig. 4.14 (the increased cathodic photocurrent at low pH values results from higher efficiency of O2reduction in acid medium). By contrast, the value of Con in case of T i O D y e IV electrode, which is governed by the polarization dependence of the efficiency of recombination of electrons from conduction band with holes produced in the dye layer under illumination, appears to be insensitive to the variations in the oxidation potential of aggregated dye and exhibits the change with pH in parallel with the flatband potential of Ti02 (Fig. 4.14). Hence, being a component of a heterojunction, the thin films of cyanine dyes generate anodic photocurrent through the electron injection mechanism which resembles in many ways the conventional mechanism of sensitization of semiconductor electrodes by monomolecular dyes, whereas the generation of cathodic photocurrent
132
Nanocrystalline Aggregates of Cyanine Dyes
occurring via “recombination” mechanism also involves the diffusion collection of charge carriers produced in the dye layer under illumination. 4.8. Conclusions
The photoelectrochemical studies on the photosensitization behaviour of aggregated cyanine dyes adsorbed onto surface of semiconductor electrode provide a direct information on the processes governing the fate of molecular exciton confined, depending on the aggregate structure, to one or two dimensions and on the kinetics of secondary reactions involving photoproduced oxidized dye states. The association of dye molecules in the aggregates favours the photosensitized reduction of molecular oxygen, the latter process being in competition with the injection of photoelectrons into semiconductor that makes it possible an effective generation of dye-sensitized cathodic photocurrent at the semiconductor of n-type under negative bias. Due to redox activity inherent in the cyanine dyes in a condensed state, not only aggregates adsorbed directly on the semiconductor surface but also multilayer films built of dye aggregates can participate in the generation of cathodic photocurrent. The aggregation-induced changes in redox potential and the mobility of photoholes within aggregates result in an effective lateral charge transfer between monomers and aggregates (or between coadsorbed aggregates of different cyanine dyes) during the course of photosensitization process, i. e. one component of nanoheterogeneous photosensitizer can function as a supersensitizer with respect to the other. REFERENCES 1. Scheibe G. Angew. Chem., 49, No. 30,563 (1936); 50, No. 11,212-219 (1937). 2. Jelley E. E. Nature, 138, No. 3502, 1009-1010 (1936); 139, No. 3519,631-632 (1937). 3. Stunner D. M. andHeseltine D. W. In: The Theory o f t h e Photographic Process, 4th ed., T. H. James (Ed.), pp. 194-234, MacMillan, New York (1977). 4. Emerson E. S., Conlin M. A,, Rosenoff E. A,, Norland K. S., Rodriguez H., Chin D. and Bird G. R. J.Phys.Chem., 71, No. 8,2396-2403 ((1967). 5. Czikkely V., Forserling H. D. andKuhnH. Chem. Phys. Lett., 6, No. 1, 11-14 (1970). 6. Scherer P. 0. J., Molecular aggregate spectra. In: J-aggregates, T. Kobayashi (Ed.), pp. 95-1 10, World Sci. Publ, Singapure (1996). 7. Tani T., Photographic Sensitivity: Theory and Mechanisms, Oxford Series on Optical and Imaging Science, Oxford University Press, NY, USA (1995). 8. West W. and Gilman P. B. In: The Theory of the Photographic Process, 4th ed., T. H. James (Ed.), pp. 251-290, MacMillan, New York (1977). 9. Brooker L. G., White F. L., Heseltine D. W., Keyes G. H., Dent S. G., Jr. and Van Lare E. J. J. Photogr. Sci., 1, No. 1, 173-183 (1953). 10. Natanson S. V. and Lifshits E. B. Uspekhi Nauchn. Fotogr., 17, No. 1,23-42 (1976) (in Russian). 11. West W., Carroll B. H. and Whitcomb D. J. Phys. Chem., 56, No. 9, 1054-1067 (1952). 12. Herz A. H. Photogr. Sci. Eng., 18, No. 3, 323-335 (1974). 13. Reich C. Photogr. Sci. Eng., 18, No. 3, 335-339 (1974). 14. Biicher B. and Kuhn H. Chem. Phys. Lett., 6, No. 3, 183-188 (1970). 15. Dietz F. J. Signal AM, 1, No. 3, 157-180; No. 4,237-252 (1973). 16. Smith D. L. Photogr. Sci. Eng., 18, No. 3,309-322 (1974). 17. Potenza J. and Mastropaolo D. Acta Crystallogr. B, 30, No. 10,2353-2359 (1974). 18. Sano N. and Tanaka J. Bull. Chem. SOC.Jpn., 59, No. 3, 843-851 (1986). 19. Saunders V. I. and Love11 S. P. Photogr. Sci. Eng., 24, No. 4, 171-176 (1980).
Chapter 4; D. V. Sviridov
20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54.
55. 56. 57.
133
Yacynych A. M., Mark H. B., Jr. and Giles C. H. J. Phys. Chem., 80, No. 8,839-842 (1976). Rosenoff A. E., Walworth V. K. and Bird G.R. Photogr. Sci. Eng., 14, No. 5,328-335 (1970). Maskasky J. E. Langmuir, 7, No. 2,407-421 (1991). Kuhn H. and Kuhn C. Chromophore coupling effects. In: J-aggregates, T. Kobayashi (Ed.), pp. 1-40, World Sci. Publ., Singapure (1996). Van Kempen H., Janssens G., Gemtsen J., Deroover G., Callant P., Vandenbroucke D. and Dekeyzer R. In: Proc. Int. Symp. Advanced Characterization Techniques for Nanostructurated Materials, pp. 26-29, UIA, Antwerp, Belgium, (1998). Frei E. A. Photogr. Korr., 105, No. 1,5-15 (1969). Yonezawa Y.,Mtibius D. andKuhn H. Ber. Bunsenges. Phys. Chem., 90, No. 12, 1183-1188 (1987). Kasha M. Radiat. Res., 20, No. 1,55-71 (1963). Briggs J. S. and Herzenberg A. Mol. Phys., 21, No. 5,865-879 (1971). Spano F. C., Kuklinski J. R. and Mukamel S . J. Chem. Phys., 94, No. 11,7534-7544 (1991). Van Burgel M., Wiersma D. A. and Duppen K. J. Chem. Phys., 102, No. 1,20-33 (1995). TanakaT. and Matsubara T. J. Imaging. Sci. Technol., 37, No. 6,585-588 (1993). Higgins D. A. and Barbara P. F. J. Phys. Chem., 99, No. 1, 3-7 (1995). Sundstrtim V., Gillbro T., Gadonas R. A. and Piskarskas A. J. Chem. Phys., 89, No. 5,2754-2762 (1988). Knapp E. W. Chem. Phys., 85, No. 1,73-82 (1984). Katheder F. Kolloid Z., 92, No. 3,299-324 (1940). Muenter A. A., Brumbaugh D. V., Apolito J., Horn L. A., Spano F. C. and Mukamel S. J. Phys. Chem., 96, No. 7,2783-2790 (1992). Lanzafame J. M., Muenter A. A. and Brumbaugh D. V. Chem. Phys., 210, No. 1-2,79-89 (1996). Ryabuk B. N., Yusupov R. G. and Khairutdinov R. F. Khimicheskaya Fizika, 7,No. 8, 1110-1113 (1988) (in Russian). Tani T., Suzumoto T., Kemnitz K. andYoshihara K. J. Phys. Chem., 96, No. 7, 2778-2783 (1992). Tani T. J-aggregates in spectral sensitization of photographic materials. In: J-aggregates, T. Kobayashi (Ed.), pp. 209-228, World Sci. Publ, Singapure (1996). Stanienda A. Natunviss., 47, No. 15,512-513 (1960). Gouverneur L., Leroy G. and Zador I. Electrochim. Acta, 19, No. 2,212-225 (1974). Tani T., Kikuchi S. and Honda K. Photogr. Sci. Eng., 12, No. 2 , 8 0 4 9 (1968). Large R. F. In: Photographic Sensitivity. R. J. Cox (Ed.), pp. 241-263 Academic Press, London & NY, (1973). Memming R. Progr. Surf: Sci., 17, 7-74 (1984). Ryan M. A. and Spitler M. T. J. Imaging. Sci., 33, No. 1,46-49 (1989). Leubner I. Photogr. Sci. Eng., 20, No. 1,61-68 (1976). Gilman P. B. Pure Appl. Chem., 49, No. 3,357-377 (1977). Lenhard J. R. and Hein B. R. J. Phys. Chem., 100, No. 43,17287-17296 (1996). Seki K., Araki T., Ito E., Oichi K., Narioka S., Yokoyama T., Ohta T., Watanabe S., Matsunaga A. and Tani T. J. Imaging. Sci. Technol., 37, No. 6,589-597 (1993). Gilman P. B. Photogr. Sci. Eng., 12, No. 2,230-233 (1968). Sviridov D. V. and Kulak A. I. Zh. Nauch. Prikl. Fotogr. Kinematogr., 32, No. 6, 452-455 (1987) (in Russian). Sviridov D. V. and Kulak A. I. New. J. Chem., 15, No. 7,539-544 (1991). Albery W. J., Boutelle M. G., Colby P. J. and Hillman A. R. J. Electroanal. Chem., 133, No. 1, 135-145 (1982). Hochstrasser R. M. and Kasha M. J. Photochem. Photobiol., 3, No. 4,317-331 (1964). Gerischer H. Faraday Discuss. Chem. Soc., 58,219-236 (1974). Natoli L. M., Ryan M. A. and Spitler M. T. J. Phys. Chem., 89, No. 8, 1448-1453 (1985).
134
Nanociystalline Aggregates of Cyanine Dyes
58. Lifshits E. B., Kirenskaya L. I., Shagalova D. I., Kudryavskaya N. V., Ushanov G. G., Jashukova L. N., Kurkina L. G. and Varlygina U. M. Organ. Veshchestva I ikh Primenenie v Khim-Photo Prom., pp. 166-176, Publ. of Gosniichimfotoproekt (1984) (in Russian). 59. Shapiro B. I. Zh. Nauch. Prikl. Fotogr. Kinematogr., 37, No. 2, 139-154 (1992) (in Russian). 60. Sviridov D. V., Shapiro B. I. and Kulak A. I. J. Photochem. Photobiol. A: Chem., 67, No. 3, 377383 (1992). 61. Shapiro B. I. J. In$ Rec. Mat., 19, No. 1-2, 105-120 (1991). 62. Hada H., Yonezawa Y. and Inaba H. Ber. Bunsenges. Phys. Chem., 85, No. 5,425-430 (1981). 63. Sviridov D. V., Kulak A. I. and Shapiro B. I. J. Int Rec. Mat., 15, No. 4,243-250 (1987). 64. Sviridov D. V. and Kulak A. I. Khimia Vysokih Energii, 21, No. 4,343-347 (1987) (in Russian). 65. Hada H. and Yonezawa Y. Synth. Met., 18, No. 1-3,791-796 (1987). 66. Sviridov D. V., Kulak A. I. and Shapiro B. I. Khimia Vysokih Energii, 24, No. 2, 151-155 (1990) (in Russian). 68. Yonezawa Y., Hanawa R. and Hada H. J. Imaging. Sci, 34, No. 6,249-255 (1990). 69. Sviridov D. V. Sci. Appl. Photo., 39. No. 5,463-475 (1998). 70. Jeunieau L., Verbouwe W., Rousseau E., Van der Auweraer M. and Nagy J. B. Lungmuir, 16, NO. 4, 1602-1611 (2000). 71. Kolesnikov L. V., Guzenko A. F., Zvidentsova N. S., Dzubenko F. A. and Breslav Yu. A. Zh. Nauch. Prikl. Fotogr. Kinematogr., 36, No. 5, 360-366 (1991) (in Russian). 72. James T. H. Advances in Photochem., 13,329-425 (1986). 73. Ferguson P. M., Babcock T. A. and James T. H. Photogr. Sci. Eng., 19, No. 5,266-272 (1975). 74. Matsumura M., Mitsuda K. and Tsubomura H. J. Phys. Chem., 87, No. 25,5248-5251 (1983). 75. Sviridov D. V. and Kulak A. I. J. If: Rec. Mat., 18, No. 1, 3-13 (1990). 76. Collier S. S. Photogr. Sci. Eng., 18, No. 4,430-440 (1974). 77. Mizutani F., Iijima S., Sasaki K. and Shimura Y. Ber. Bunsenges. Phys. Chem., 86, No. 10, 907912 (1982). 78. Mizutani F., Yoshiura M. and Iriyama K. J. Chem. Soc., Chem. Comm., No. 9,393-394 (1980). 79. Chamberlain G. A. and Malpas R. E. Faraday Discuss. Chem. SOC.,70,299-310 (1980). 80. Sviridov D. V. and Kulak A. I. Elektrokhimiya, 23, No. 6, 856-859 (1987) (in Russian). 81. OglesbyD. M., Omany S. H. andReilly C. N.Ana1. Chem., 37, No. 11, 1312-1316 (1965). 82. Kulak A. I., Kokorin A. I. and Sviridov D. V. J. Mater. Res., 16, No. 8,2357-2361 (2001). 83. Talapin D. V., Sviridov D. V. and Kulak A. I. J. Electroanal. Chem., 489, No. 1,28-37 (2000). 84. Sviridov D. V. and Kulak A. I. Thin Solid Films, 198, No, 1/2, 191-198 (1991).
Chemical Physics ofNanosiruciured Semiconductors, pp. 135-1 5 2 A.I. Kokorin and D.W. Bahnemann (Eds.)
0VSP 2003.
CHAPTER 5
Physico-Chemical Properties of Novel Nanocrystalline Ruthenium Based Chalcogenide Materials Nicolas Alonso-Vante Laboratory of Electrocatatalysis, University of Poitiers, France Kevwords: Cluster, photocatalysis, Chalcogenide, nanostructure, Oxygen reduction 5.1. Introduction
The ultrafine grain size of metal and semiconductor based materials (c 50 nm) has created an enormous interest in virtue of the optical, electrical, chemical and magnetic properties it conveys to them [ 13. Several chemical routes to fabricate these nanoscale materials, as for example, the coprecipitation [2] the precursor concept [3] and the colloidal approaches [3,4], have been employed. Material transformation through reactions using chemical precursors can readily produce nanostructures in bulk quantities. One of the targets in tailoring novel materials is the way of arranging individual transition metals to form a cluster-like one. Recently, an alternative process has been developed which takes advantage of the reactivity of the transition metal carbonyls in solvents to tailor ruthenium based chalcogenide cluster-like materials [ 5 ] . Using this latter approach, nanoscale semiconducting MoSe2 material has also been synthesized [6]. Furthermore, the promotion of photoinduced charge separation to enhance the efficiency of photocatalytic reactions has been, for example, performed using noble metals such as platinum or gold deposited on semiconductor nanoclusters [7]. The development of heterogeneous photocatalysis, based on nanosized, nanostructured, or nanoparticulated semiconductor materials has now been placed in forefront [8]. Very recently we have synthesized novel materials based on transition metal compounds (within cluster or framework structures) in colloidal form [ 9 ] . Some characteristics of this type of compounds can briefly be outlined: ( i ) they have a high monodispersitivity; (ii) the presence of a chemical stabilizer, during synthesis, does not perturb the chemical nature of the novel compound, therefore, a similar stoichiometry (Ru,Se,, in which x =: 2 and y = 1) as well as a similar atomic co-ordination distance can be obtained, as revealed by RBS [lo] and EXAFS [I11 measurements, respectively; and (iii) it is easy to deposit this colloidal material, e.g. by dipping, on conducting (e.g. glassy carbon, Sn02:F) [ 111 and semiconducting substrates (e.g. TiOz). These characteristics allowed us to verify selectivity and electrocatalytic activity towards the oxygen reduction. A property which is interesting in the development of selective cathodes for fuel cells. In this chapter the progress in the synthesis of cluster based nanostructured material will be stressed. Furthermore, its physico-chemical properties as well as its application in energy converting systems: catalytic (photo)sensitization of surface reactions on a nanostructured large band gap semiconductor (Ti02) will be discussed.
136
Ruthenium Based Chalcogenide Materials
5.2. Clusters or Particles 5.2.1. Size Comparison
Metal clusters have been considered as model systems for the study of catalytic reactions [12]. Clusters are made of a few atoms to a few hundred atoms. The progression in size (Fig. 6.1) varies from atoms through clusters and nanoparticles to the bulk regime. The cluster size lies at the interface between atomic distance and the so-called nanoparticles, colloids and bulk metal. The lack of boundaries (marked by a question mark) in the low limit size for crystallinity, in the sense of Bragg reflections (presence of a long range order), for bulk may depend on the nature of the metal, alloy or semiconductor. Material bulk properties are already present in the first atomic layers. Nevertheless, from an X-ray point of view, the transition from nanoparticles (with nanocristalline domains) to particles in the micron size regime is observed by the coalescence of the former ones induced e.g., by an annealing treatment under an atmosphere (inert or not). The intensity and the broad Bragg peaks of nanoparticles (see e.g., Fig. 5.6) change in a characteristic way. As for example, after a treatment under hydrogen at 200°C the first reflections of ruthenium particles (with a size of ca. 8 nm) are already well resolved r131.
10
10
.
E lo
c
10
10
Fig. 5.1. Scaling in materials from atoms to the bulk crystallinity limit.
137
Chapter 5; N. Alonso-Vante
In organometallic chemistry, however, it is generally considered that the metalmetal bond length (e.g., &+,= 0.268 nm for Ru (hcp)) in cluster compounds such as MxLy (where M is a transition-metal and L is a ligand) exceeds this length by 15% [14] because of the different electron donating capabilities of the ligands. The term cluster was initially coined to describe a compound in which a group of two or more metal-metal bonds are involved. This is the ascription for the family of molecular clusters. However, materials based on colloidal metal particles may contain from four to several hundreds of metal atoms in their skeleton. Therefore, nanoparticles or clusters such as Au55 [15] or Pd561 [16] with dimensions of up to 4 nm, can be considered as a bridge between molecular clusters [14] and clusters forming metal particles 11171. Therefore, an overlapping in size between traditional cluster compounds (e.g. transition-metal carbonyls), nanoparticles and colloids exists. In the author's view the overlapping or bridge, among clusters, nanoparticles and colloids, presented in Fig. 5.1, might explain the ambiguous use of the different terms found in the literature. C
C
b
a
v
Fig. 5.2. Three examples of crystal structure of clusters : (a) (b) Mo6Se8and (c) Ru, (x = 13).
Ru3(C0)12;
Fig. 5.2 shows examples of clusters of a molecular, a chalcogenide and a metallic material. The fEst one: tris-ruthenium dodeca-carbonyl (Ru3(CO)lZ),in which the metal skeletal geometry is based on metal triangles, is the chemical precursor used to manufacture our material: Ru,Sey (see section 5.3). The second example depicts a structure in which one can recognize the clustering center (octahedral) based on molybdenum atoms surrounded by chalcogenes, the so- called Chevrel-Sergent phase ModXs (X = Se, S or Te) [18]. This unit closely resembles [M0~Cls]~'[19].In these two examples (Figures 5.2a, b) one can appreciate the complex three-dimensional networks in which the metal centers (clusters) are embedded. In contrast to R U ~ ( C Oand ) ~M ~ o a s , one can also imagine a small
138
Ruthenium Based Chalcogenide Materials
nanoparticle (metallic cluster) constituted of 13 ruthenium atoms ( R u ~ ~see ) , Fig. 5 . 2 ~with a diameter of ca. 0.7 nm. This picture is the result of X-ray spectra analysis using the Debye function Analysis (DFA) method for clusters of different size [13]. The stability of suchclusters can be related via an inorganic passivation with the oxygen leading to oxidelike cluster materials. This behavior has been reported for, e.g., S O z [20] and InP [21]. The material in a nanoscale dimension provides an important surface to volume ratio. The change in the surface energy induces the atomic distances of the surface atoms to be contracted. As a result, in this type of nanomaterials, large disorder and higher catalytic activity can be expected [22]. The former phenomenon is clearly observed in the X-ray diagram which delivers broader diffraction peaks [l, 6, 13, 221. 5.2.2. Metal and Semiconductor Nanoparticles
Once the crystal comprises a few tens of atoms, there is an identifiable core, in which the local bonding geometry closely resembles that of bulk solids. As mentioned in the precedent section, nanoparticles constitute the bridge between the atomic-molecular and the extended crystalline limits. In simple metals, the Fermi level, EF, lies in the center of a band. The relevant energy spacing is still very small and above a few K, the electrical and
Fig. 5.3. Schematic diagrammes of density of states in a metal (a) and in a semiconductor (b). From left to right the electrons begin to form discrete energy levels.
Chapter 5; N. Alonso-Vante
139
optical properties more closely resemble those of a continuum even at relatively small sizes. In other words, the electrons begin to form discrete energy levels, that is, an intermediate case between extended structures (electrons are delocalized in 3-D states) and molecules (fully localized bonding electrons), Fig. 5.3(a). In semiconductors, however, the Fermi level lies between two bands, so that the edges of the bands dominate the low-energy optical and electrical behavior, Fig. 5.3(b). As established in the literature [23-251, the optical behavior of semiconductors depends strongly on the size even for those as large as lo4 atoms. As exposed below, our cluster material possesses a metallic character, with an electronic behavior that appears to be modulated by the chalcogen coordination.
5.3. A Non-Aqueous Chemical Route of Synthesis 5.3.1. Synthesis Ideals The kinetic principles leading to a desired morphological structure from chemical reactions are not well understood. The challenge is then, to understand these principles in order to control the morphology leading to highly desired properties. In this connection, the synthesis of transition metal chalcogenide based materials, in mild conditions (T I 4 7 3 K), made it possible to increase the surface to volume (SN)ratio. The essential characteristics of these materials, in powder and in the colloidal form is their high dispersiveness in the nanoscale range [26, 271, and the simplicity to manufacture them [9]. Tailoring the material is also possible [5] using X = S , Te, Se. This original method of synthesis of electrocatalytic materials has been reproduced by other workers [28, 291 with the aim of dispersing them onto conducting substrates, such as Vulcan XC-72 carbon, for the fabrication of technical electrodes used as cathodes in fuel cells. Another type of substrates, such as conducting polymers, has also been used for the fabrication of composite electrodes [30]. Their electroactivity with respect to the molecular oxygen reduction reaction and the chemical stability are higher than those of Chevrel phase materials [31]. Moreover, in spite of the facile way of synthesis, the route of the chemical reaction has not yet been well established. It is necessary to know this route in order to design, at the molecular level, more active novel compounds, cheaper and more selective than platinum (nowadays the best catalysts for fuel cells). This is why, one of the fields of research in our group is the understanding of this chemical route in mild conditions.
5.3.2. The Chemical Route Fig. 5.4 depicts some results obtained in the first stages (high nuclearity complexes formation) of the synthesis in xylene solvent which leads to the formation of nanostructured powders, Ru,Sey, from tris-ruthenium dodeca-carbonyl ( R u ~ ( C O )and ~ ~ )elemental selenium dissolved in an organic solvent (xylene). After 40 minutes of reaction, 13C-NMR spectrum (Fig. 5.4 (c)) puts in evidence the formation of a new polynuclear chemical precursor with a Selenium ) ~ ~ ) . takes part in the coordination chemical shift 6 of 198.89 ppm (i.e., R U ~ S ~ ~ ( C O sphere. The peak intensity with the chemical shift of 199.67 ppm, corresponds to the initial chemical precursor which decreases as a function of the synthesis reaction time (Fig. 5.4(a)). Other chemical shifts (with minor peak intensities) on both sides of the 13C-NMR spectrum, which put in evidence the complex interplay of the reaction, are also observed.
140
Ruthenium Based Chalcogenide Materials
Ru,(CO),,
+ Se + Ru,Se,(CO),, + ...... + RuxSey
Fig. 5.4. The chemical route for the synthesis of Ru,Sey. The initial chemical precursors dissolved in xylene: (a) Ru3(CO)12 and (b) elemental selenium (b). 13CNMR spectrum of a solution after 40 mn reaction time. The chemical shifts: 199.67 ppm belongs to (a); and 198.89 O ) ~ non-stoichiometric ~. reaction (below) describe ppm to (c) R U ~ S ~ ~ ( C The
schematically the complex chemical interplay leading to cluster-likecompound: Ru,Sey. These complexes are difficult to identify. The compound R U ~ S ~ Z ( C(Fig. O ) ~5.4 ~ d) evolves continually with the reaction time either to another selenium-containing complex: Ru$3ez(CO)g (the identification of this latter is due to the appearance of a band in the infrared spectrum at 2050 cm-') [32] or directly to the final product: RuxSey.The loss of the carbonyl groups is obtained by keeping the boiling temperature of the solvent in refluxing conditions under argon or under nitrogen during 20 h of reaction. The intermediates, ruthenium based complexes with high nuclearity, formed following this chemical route have already been synthesized in the past by other synthesis method of organometallic chemistry [33, 341. However, this chemical route allows the identification of the main chemical precursor: R U ~ S ~ ~ ( Cwhich O ) ~ ~leads , to the novel cluster based metallic compound embedded in selenium: RuxSey. It turns out that this method of fabricating R U ~ S ~ ~ ( is C more O ) ~ facile ~ than the one previously reported by Layer et al. via pyrolisis at 185°C under vacuum of Ru3(CO)lZ and PhSeSePh during 19 h [33]. The addition of a chemical stabilizer (octadecanthiol) at the beginning of the reaction does not perturb the chemical route. The product (RuxSey),this time in colloidal solution [ 9 ] , is fully stable and can be applied to coat different kinds of substrates (see section 5.4). 5.3.3. Morphology and Stoichiometry Figures 5 (a) and (b) show electron micrographs of the Ru,Sey particles in powder form, Fig. 5.5(a) and in colloidal form, Fig. 5.5(b). The generated particle size in both cases is ca.2 nm. It is, however, interesting that the colloidal route delivers particles with a narrow size distribution. After multiple analysis by EDX performed with transmission electron microscopy (TEM), and/or via Rut-herford backscattering spectroscopy (RBS) we concluded that the stoichiometry of the RuxSeycompound corresponds to x = 2 and y E 1. This is another experimental evidence that the "real" chemical precursor is the intermediate
Chapter 5; N . Alonso-Vante
141
Fig. 5.5. Transmission electron microscopy picture of Ru,Sey cluster-like materials. (a) in form of powder; (b) from a colloidal solution (xylene).
complex, Ru4Se2(CO)11, generated during the reaction synthesis. X-ray analysis has been carried out on powder samples. 600 500
I
1
II
II
A
I
1
30
40
II
II
n
c
20
II
RuxSey RuxS ey
I
I
I
60 70 2 theta I degree 50
J
4
-:
I
1
80
90
Fig. 5.6. XRD spectrum profile of a Ru,Sey powder. Trace amounts of unreacted elemental selenium are indicated by (*) [ 111.
As an example, Fig. 5.6 depicts a typical diffraction spectrum. It is evident that long range order does not exist in our chalcogenide samples. However, the broad diffractrogram peak centered at 28 = 42.5" has the characteristic of a nanodivided ruthenium metal [22]. This points out that the active center in this chalcogenide materials is essentially of metallic nature. The material, either in powder or colloidal form, was analyzed by the EXAFS technique [ I l l . The local range order of this technique allowed for some structural determination of our samples. Thus, for example, the co-ordination distances for ruthenium-selenium and ruthenium-ruthenium are: R(Ru-s)= 2.43 8, y R(Ru-Ru)= 2.64 A, respectively. The metal-metal co-ordination distance is of the same order of magnitude as that of well known cluster based materials such as the Chevrel phase [35, 371, cf. Fig. 5.2b. This testifies that the used chemical route leads to the formation of cluster-like materials.
142
Ruthenium Based Chalcogenide Materials
5.4. Electrocatalysis and Photoelectrocatalysis 5.4.1. Thin to Ultra Thin Layers of Chalcogenide Materials
To understand chemical and (photo)electrochemical processes induced by clusterlike chalcogenide materials, we proceeded to deposit either directly (dipping any substrate in the reactor) or indirectly, i.e., via coating from a prepared colloidal solution. Using the former approach, for example and taking a glassy carbon (GC) as a substrate, we get the material deposited in form of an inhomogeneous layer as seen with scanning electron microscopy (SEM) as depicted in Fig. 5.7(a).
Fig. 5.7. Scanning electron micrograph of thin layers of Ru,Sey deposited orto a glassy carbon
suface during the synthesis: the layer morphology as prepared (a), and after an ,;,,,rochemica1 treatment (b) in argon or oxygen saturated 0.5MH 2 S Q electrolyte. At the beginning these layers, deposited during synthesis, display three dimensional arrays of interconnected particles. The size of the largest quasi spherical particles is ca. 400 nm. However, Fig. 5.7(b) shows that the outermost particles peel-off during an electrochemical treatment in 0.5 M HzS04 leaving a so-called thin layer formed by particles with strong adherence to the substrate. The typical thickness of such layers is of some tenths of micrometers. The layer shown in Fig. 5.7(b) has a thickness of 0.2 ym. RBS spectroscopy revealed that such layers are porous [31], thus, such layers are in fact an ensemble of even smaller particles in the nanoscale range as revealed by TEM, cf. Fig. 5.5(a). Ultra thin layers can be obtained using the colloidal solution [9] by dipping a substrate from seconds to hours. A thermal treatment is required in order to eliminate the chemical stabilizer (octadecanthiol). In this way various substrates have been used to do electrochemical and/or photoelectrochemical studies. Fig. 5.8 shows an atomic force microscopy picture of the resulting material onto a highly oriented pyrolytic graphite (HOPG) substrate, after thermal annealing. This surface does not allow the formation of an homogeneous film. Rather formation of patches (see the 2D picture) consisting of cohesion of various nanoparticles with an average height roughly between 3 to 4 nm, i.e., about two orders of magnitude thinner than those films obtained via in situ synthesis are formed.
Chapter 5; N. Alonso-Vante
143
Fig. 5.8. Atomic force microscopy picture of Ru,Sey deposited by dipping onto HOPG surface after annealing in argon at 210°C to eliminate the stabilizer (octadecanthiol). (a) 3D representation gives an idea of the nanoparticle agglomeration, and (b) 2D the islands formed due to the agglomeration process during annealing.
Fig. 5.9. Scanning tunneling picture of Ru,Sey deposited by dipping onto Sn02:F (FTO) surface after annealing in argon at 210°C to eliminate the stabilizer (octadecanthiol). (a) The top view representation gives an idea of the nanoparticle distribution, (b) the profile (see straight line in (a)) before and (c) after deposition of nanoparticles .
144
Ruthenium Based Chalcogenide Materials
Another substrate of interest, employed for the in situ grazing incidence EXAFS study [I 11, was the fluorine-doped Sn02 ( R O ) . This surface was modified using the same depositing procedure as on the HOPG substrate. One can recognize the deposited particle distribution on the top view (Fig. 5.9a) of the picture generated via STM. Fig, 5.9b shows the profile of Fig. 5.9a. Again, we observe that a very small amount is deposited forming a very thin non-compact layer. A similar structure is obtained using other substrates, such as the nanostructured TiOz. The chalcogenide nanoparticles can work as catalytic centers on Ti02 substrates. The reaction used as a probe for these modified surfaces is the molecular oxygen reduction in acid medium. 5.4.2. Electrocatalysis via Thin and ultra Thin Layers of Chalcogenide Materials
The novel cluster-like chalcogenide material Ru,Sey deposited in thin [5, 26, 31, 363 and ultra-thin layers [9, 111 or in powder form embedded in a polymer matrix [30] was found to be an efficient catalyst for the molecular oxygen reduction in acid medium. Fig. 5.10 summarizes the current-potential Cj-E) characteristics as a function of the substrate’s nature. First of all, one can appreciate that similar activities are obtained from materials synthesized in powder or in colloidal form when deposited onto GC (Fig. 5.10, compare curves (1) and (2)). For the sake of comparison, the j-E characteristic generated on the naked GC substrate for the electrochemical process is contrasted in curve (5). 10
10
10 111
U E
10
-’
10 .z
10
10
4.2
0
0.2
0.4
0.6
0.8
1
E(SHE)AI
Fig. 5.10. The electrochemical activity (Tafel plots) of surface modified by Ru,Sey towards the molecular oxygen reduction in 0.5MH2S04. (1) layer deposited during the synthesis of powder on glassy carbon (E); (2) ultra thin layer deposited via colloidal solution on glassy (4) ultra thin carbon; (3) ultra thin layer deposited via colloidal solution on Sn02:F (FTO); layer deposited via colloidal solution on TiOz (anatase) supported on FTO; ( 5 ) the electrochemical response of naked GC. All systems (2 to 4) were annealed in argon at 210°C to eliminate the stabilizer (octadecanthiol). TL: Thin layer; UTL: Ultra thin layer of Ru,Sey.
Chapter 5; N. Alonso-Vante
145
The extrapolation of the straight line of the naked GC to an applied electrode potential in the activation region (e.g., 0.8 V/SHE) indicates that the reaction of the molecular oxygen reduction (formation of water) by surface modification of GC is enhanced by 4.8 orders of magnitude. Taking the 0.8 V/SHJ3 value as a reference for activation, within the limits of experimental error, the activity on Sn02:F (curve (3)) and on nanostructured Ti02 (curve (4)) is 0.9 and 2.8 orders of magnitude lower than that measured on modified glassy carbon (curves (1) and (2)). This aspect puts in evidence the semiconducting nature of degenerate Sn02:F (Eg = 3.4 eV) and undoped Ti02 (Eg = 3.2 eV, anatase) substrates. Although Ti02 was deposited onto Sn02:F substrates, the j-E characteristic clearly shows that the process is mostly bypassed onto the Ti02 nanoparticles surface. The activity of the Ti02 naked substrate towards the oxygen reduction in acid can be also represented by the naked GC substrate (see curve (5)). This reaction, which proceeds with a high overpotential on naked Ti02 (same as for naked GC), is not as well favored in an acid medium as it is in an alkaline electrolyte [38]. The product of the reaction in acid is mainly a peroxide species (2 electrons charge transfer: 0 2 + 2H' + 2e- 3 H202). What curve (4) in Fig. 5.10 shows is an effective scavenging of conduction electrons via cluster-like nanoparticles of RuxSeyleading directly to the formation of water ( 0 2 + 4H+ + 2e- H20) [31]. The scavenging of conduction electrons via surface states (dark process) commences to be available from ca. 0.7 V/SHE (at a current density of lo4 mA/cm2). Some energetic considerations should be taken into account. The position of the flat band energy (Etb) of anatase films corresponds, according to considerations of Rothenberg et al. [39], to -0.18V/SHE (pH 0.3). Therefore, the presence of cluster-like material modifying the anatase nanoparticles surfaces works as an effective electron scavenger relay (catalysts) of charges already available at ca. 0.88V below the Fermi level. These charges are conveyed to reduce the molecular oxygen via 4 electrons. This is necessary for the improvement of energetic efficiency in energy converter devices (see section 5.4.3). The charges, although available (and/or stored), are not effectively conveyed to the oxygen molecules, in the absence of the cluster-like catalyst. The consequence of this phenomenon is that the presence of O2 in the electrolyte favors the recombination process (e.g., lowering of photopotential) of nanostructured Ti02 [40].
+
5.4.3.(Photo)Electrocatalysison Nanostructured Ti02 Surface Modified via Chalcogenide Materials The condition for an efficient light-induced charge separation is the presence of a depletion layer at the interface. In the earlier days of photoelectrochemistry [41-421, the electrodeposition of transition metals (e.g. Ru, Pt,..) was employed to enhance interfacial charge transfer process, for example to promote photoelectrocatalytic processes. Nowadays, the use of nanostructured semiconductor materials has permitted the development of a variety of approaches. Systems based on large band gap sensitization, via dyes [43-481, chalcogenides [49-541, and noble metals [7, 551, represent, for example, some strategies. Therefore, to take advantage of the material properties in the nanoscale range, described above, this section describes results regarding the photoelectrochemical and photoelectrocatalytic properties of the FTO/Ti02/RuxSysystem. The material (RuxSey),which modifies the electrochemical behavior of substrates presented in Fig. 5.10, in the colloidal solution has an absorption spectrum shown in Fig.
Ruthenium Based Chalcogenide Materials
146 Absorbance 2.5
I
i
i
i
i
i
2
1.5
I
0.5
0 300
700
500
900
Wavelength (nm)
Fig. 5.11. Absorption spectra of 1.71~10”M Ru,Sey colloid (xylene) (l), and of a diluted colloidal solution (0.32 M) of Ti02 (P25-Degussa)( 2 ) .Optical path = 1 111111.
Iphp
a.u.
400
500
600
700
Wavelength (nm)
Fig. 5.12. Action spectra of naked and Ru,Se,-surface modified Ti02 in argon purged 0.5 M H2S04 electrolyte; applied potential was 0.75 V/SH
5.11. The solution has a black color and is very stable with time, even if exposed to air. The observed spectrum is rather broad with a maximum peak centered at 351 nm. Its morphology is depicted in Fig. 5.5(b). Such featureless spectrum has also been observed with reduced Pd colloid solution [ 5 6 ] . Comparing the optical absorption of TiOz colloidal diluted solutions, prepared from P25 Degussa Ru,Se, ultra thin layers can produce charge
147
Chapter 5; N. Alonso-Vante
separation under illumination. The optical absorption spectrum of diluted Ti02 solution resembles the photocurrent action spectrum reported by Hagfeldt et al. [57]. The recorded action spectrum before and after surface modification, in argon purged electrolyte, of the system: FTO~i02/RuxSd0.5MH2S04 is shown in Fig. 5.12. Thus, as expected, on the modified electrode surface the onset of the photocurrent is shifted to lower energies (E = 1.77 eV) in comparison to the band gap excitation of Ti02 (Eg = 3.2 eV). Because of the absence of an efficient electron donor (e.g., H2Q), the separated photogenerated charges (holes) must react with water. This leads to a decrease of the photocurrent. Nevertheless, the action spectrum of our system is in contrast to recently reported results on gold modified Ti02 nanostructure surface [55]. The authors of this report did not observe a shift of the photocurrent onset towards longer wavelengths, and therefore put in evidence the fact that Au nanoparticles is only favoring charge separation generated via band to band excitation. Moreover, although not as efficient as Ru-based dye compounds [44], our Ru,Sey nanoparticles deposited as a colloid capture some lower energy photons which in turn can be injected to the conduction band of Ti02 (sensitization effect). On the other hand, as discussed in the section 5.4.2, the RuxSey nanoparticles deposited onto Ti02 also favor the charge transfer (dark process) of charges stored in surface state to enhance catalysis of reduction of molecular oxygen. 5.4.4. Implications of Molecular Oxygen for Photooxidation Process.
The presence of dissolved molecular oxygen favors recombination of photogenerated charge carriers in Ti02 particles [40]. In agreement with this, a typical currentpotential characteristic of Ti02 thin layers (particle size about 10 nm) is shown in Fig. 5.13.
c?
5 U
0.30 0.20
E
0.10
-0.10 O.OO
1
i
1
0
/
/ /
I I
--
0.2
0.4 0.6 E(SHE)/V
- - i/mAcm-2 (Ar)
0.8
1
Fig. 5.13. Current-potential characteristic of a nanostructured lTOiTi02 thin layer in 0.5 M H2S04.The photocurrent was generated via UV illumination from a Xe lamp (150 W) in argon purged (dashed line) and in oxygen saturated (full line) electrolyte. The dark current remained the same in both atmosphere.
In an oxygen saturated electrolyte, the photocurrent (e. g., at 0.3 V/SHE), generated via Xe lamp (150 W), is ca. 28% less than the photocurrent generated in an argon
148
Ruthenium Based Chalcogenide Materials
saturated electrolyte. The curve in darkness remains unaffected by the presence or not of oxygen. A similar behavior is observed on cluster-like material modified semiconductor surfaces (n-Ti02/Ru,Sey). In this interface the capture of the majority carrier is enhanced. The current onset (in darkness) can be strongly modified (shifted towards positive potential, cf. Fig. 5.10 (curve 4)) by increasing the concentration of adsorbed catalyst. The investigation of this phenomenon as a function of photogenerated carriers for the photooxydation process in under way.
0.70 : 0.60
7
-
TiO,/Ru xSey(N1)
-
-
- - Ti02/RuxSey(0,)
0.50 :
2 L
0.40
\
0.30 :
0.20 : 0.10
I
0.00
0
50
100
150
200
250
tls
0.40
-
i
4
0.35 -
0.30
-
0.25
0
50
150
100
200
250
tls Fig. 5.14. (a) the current in darkness and photocurrent response as a function of time under UV illumination of FTO/Ti02 and FTO/TiOz/Ru,Sey systems toward the oxidation of 0.5 M formic acid under nitrogen and under oxygen saturated electrolyte (0.5 M HzS04). (b) the photoinduced change in the mass signal ( d e = 32) for oxygen issued from FTO/TiOZ interface (full line); and FTO/Ti02/Ru,Sey (dashed line).
Chapter 5; N. Alonso-Vante
I
9
b I I
4
I
\
E
v
149
/
I
I
0.40
d o'20 0.00
I1 ' 0
"
"
50
"
"
I
3
"
100
"
150
-
"
"
I
200
"
' ' j 250
tls Fig. 5.14. (c) the photoinduced change in the mass signal ( d e = 44)for carbon dioxide issued from FTO/Ti02 interface (full line); and FTO/Ti02/Ru,Se, (dashed line). Arrow (1) refers to the initial UV illumination on FTO/TiO2at open circuit potential; arrow (2) refers to the initial UV illumination on FTO/Ti02/RuxSey at open circuit potential; and arrow (3) refers to the application of a bias of 0.3 V/SHE under UV illumination for both systems.
For the degradation of organic molecules [58-631, one key issue is to convey efficiently the photogenerated electrons to the surface adsorbed oxygen species. That is, efficient photooxidation process may take place in the presence of dissolved molecular oxygen. The system based on Ti02 surface modification via cluster-like materials seems to fulfill this requirement. Therefore, the interface: n-Ti02/Ru,Sey was exposed to formic acid (HCOOH). In order to detect the photogenerated volatile products the coupling of mass spectrometry (DEMS) was necessary [64-671. Figure 5.14 summarizes the photoresponse (current versus time) and the resulting masses issued from Ti02 electrodes before and after modification by cluster-like RuxSeyin 0.5 M H2S04 + 0.5 M HCOOH (nitrogen and oxygen saturated solution). As noted in the paragraph above, the magnitude of the current under UV illumination, Fig. 5.14(a), decreases by the presence of molecular oxygen (cf. Fig. 5.13) either on the non-modified or on the modified TiOz surface. This photocurrent is further diminished by the presence of the adsorbed RuxSey particles, whose magnitude is again influenced by the presence of oxygen. The mass signals ( d e = 32 for oxygen, and d e = 44 for COz) are illustrative of the effect produced by the UV-illumination under open circuit potential. Arrow (1) in Fig. 5.14 indicates the start of UV-illumination for the non-modified surface; arrow (2) for the modified one; and arrow (3) for both under applied 0.3 V/SHE. Since mass signals are relative, the important aspect is the observation of the change induced by photogenerated charge carriers. As indicated by Fig. 5.14(b) and (c), both surfaces are able to photooxidize HCOOH to COz in the presence of oxygen. However, the consumption of this latter at n-TiOz/RuxSeyis more effective with the concomitant enhancement of the C 0 2 signal. Further, the applied bias of 0.3 V/SHE (see arrow (3)) does not enhance the efficiency of the modified surface, but the non-modified one. This indicates that under UV-illumination
150
Ruthenium Based Chalcogenide Materials
the modified interface builds a potential near 0.3 V in which the magnitude of photoanodic and cathodic currents are the same. Although the current generated by UV is about 50% smaller in comparison to the non-modified one (cf. Fig. 5.14(a)), it seems that both reactions (HCOOH + 2h' 3 C 0 2 + 2H'; and 1/202 + 2H' + 2e- 3 H20) are being performed in an efficient way. This model reaction puts in evidence the performance of a photocatalytic system via modification of nanostructured Ti02 surfaces. The reaction pathway of photogenerated holes (for destruction of pollutants) can be improved if the counterpart (electrons) are being conveyed via a catalyst. In this connection, more work, on such modified interfaces, is being performed using monochromatic light as a probe for spatial resolution.
5.5. Summary Significant progress has been achieved in the reserach of materials in the nanometer scale range. This has been testified by numerous reviews in the field and recently by Shipway et a1 [68].The organizational effect of particles or clusters in such scale opens a variety of interesting phenomena for physics and chemistry. However, the target for an efficient photocatalytic process remains in the catalyst materials based on a multi-centered transition metal. A compromise should be found to take advantage of the dynamic structural-electronic changes in response to the number of photoinduced electrons (holes). Therefore, our contribution is an attempt to understand the complex interplay between material design and the physico-chemical aspects that such materials may bring out. Ruthenium, as well as other transition metals (such as Mo, Fe), presents interesting properties in electrocatalysis when the atoms are arranged in such a way as to fulfill the proper structure that lead to a novel compound. Ru,Sey is just an example among many of compounds which can be tailored by means of the method presented in this chapter. We demonstrated that this cluster-like material can be synthesized in colloidal form using the proper stabilizer compound. This method led to the manufacture of a particle with a narrow distribution. Furthermore, its catalytic properties toward the molecular oxygen reduction were tested on different substrates to do either electrocatalysis andor photoelectrocatalysis.
Acknowledgements I would like to thank my students and coworkers. The author specially acknowledges Dr. P. Bogdanoff (Hahn-Meitner-Institut-Berlin) for his help in DEMS measurements. REFERENCES Nanomaterials: Synthesis, properties and applications, A. S . Edelstein and R. C. Cammarata (Eds.), Institute of Physics Publishing, Bristol, Philadelphia (1998). 2. Watanabe M., Uchida M. and Motto S. J. Electroanal. Chem., 229, No. 1-2,395-406 (1987). 3. Bonnemann H., Brijoux W., Brinkmann R., Dinjus T., Joussen B. and Korall B. Angew. Chem., 1.
4. 5. 6.
Znt. Ed. Engl., 30, No. 10, 1312-1314 (1991). Henglein A. Ber. Bunsenges. Phys. Chem., 99, No. 7,903-913 (1995). Alonso-Vante N. J. Chim. Phys., 93, No. 4,702-710 (1996). Bastide S., Boucher A.-C., Uvy-Cltment C. and Alonso-Vante N. Electrochem. Solid State Lett., 3, NO. 9,450-451 (2000).
Chapter 5; N. Alonso-Vante
151
7. Kamat P. V. and Shanghavi B. J. Phys. Chem. B, 101, No. 39,7675-7679 (1997). 8. IPS-12, Photochemical Conversion and Storage of Solar Energy, Book of Abstracts, H. Tributsch and F. Willig (Eds.), Berlin, (1998). 9. Alonso-Vante N. German patent DE 196 44 628 A l . 10. Alonso-Vante N. In: Catalysis and Electrocatalysis at Nanoparticle Surfaces, A. Wiekowski, K. Vayenas and E. Savinova (Eds.), Marcel & Dekker, in press. 1 1 . Alonso-Vante N., Fieber-Erdmann M., Borthen P., Strehblow H. and Holub-Krappe E. Electrochim. Acta, 45, No. 25-26,4227-4236 (2000). 12. Johnson B. F. G., Gallop M. A. and Roberts Y. V. J. Mol. Catal., 86, No. 1,51-69 (1994). 13. Vogel W., Le Rhun V., Gamier E. and Alonso-Vante N. J. Phys. Chem. B., 105, No, 22, 52385243 (2001). 14. Mingo M. and Wales D., Introduction to cluster chemistry, Prentice Hall Int. Ed., Charlottesville, Virginia (1990). 15. Schmid G., Giebel U., Huster W. and Schwenck A. Inorg. Chim. Acta, 85, No. 1,97-102 (1984). 16. Vargaftik M. N., Moiseev I. I., Kochubey D. I. and Zamaraev K. I. Faraday Discuss. Chem. SOC., 92, 13-29 (1991). 17. Bradley J. S . , in: Cluster and Colloids, G. Schmid (Ed.), VCH Publishers Inc., New York, Chapter 6 (1994). 18. Chevrel R. and Sergent M., in: Superconductivity of Ternary Compounds, 0. Fischer and M. B. Maple (Eds.), p. 25, Springer Verlag, Heidelberg, New York (1982). 19. Greenwood N. N. and Earnshaw A. Chemistry of the Elements, 2"d Ed., Butterworth-Heinemann, Oxford, (1998). 20. Schuppler S., Friedman S . L., Marcus M. A., Adler D. L. and Brus L. E. Phys. Rev. Lett., 72, No. 16,2648-265 1 (1994). 21. Guzelian A., Bowen-Katari J. E., Banin U., Alivisatos A. P. and Heath J. R. J. Phys. Chem., 100, NO. 17,7212-7219 (1996). 22. Le Rhun V., Gamier E., Pronier S. and Alonso-Vante N. Electrochem. Comm., 2, No. 7,475-479 (2000). 23. Brus L. E. J. Phys. Chem., 79, No. 11,5566-5571 (1983). 24. Henglein A. Ber. Bunsenges. Phys. Chem., 101, No. 11, 1562-1572 (1997). 25. Alivisatos A. P. J. Phys. Chem., 100, No. 31, 13226-13239 (1996). 26. Alonso-Vante N., Giersig M. and Tributsch H. J. Electrochem. SOC.,138, No. 2, 639-640 (1991). 27. Alonso-Vante N. Forschungsverbund Sonnenenergie, Themen 94/95,15 (1995). 28. Romero T., Rivera R., Solorza 0. and Sebastian P. J. Znt. J. Hydrogen, 23 No. 11, 1041-1044 (1998). 29. Trapp V., Christensen P. and Hamnett A. J. Chem. SOC.,Faraday Tr., 92, No. 21, 4311-4319 (1996). 30. Alonso-Vante N., Cattarin S. and Musiani M. J. Electroanal. Chem., 481, No. 2, 200-207 (2000). 31. Solorza-Feria et al. Electrochim. Acta, 39, No. 11, 1647-1653 (1994). 32. Le Rhun V., Ph.D. Thesis, University of Poitiers, (2001). 33. Layer T. M., Lewis J., Martin A., Raithby P.R. and Wong W.-T. J. Chem. SOC.,Dalton Trans., 3411-3417 (1992). 34. Johnson B. F. G., Layer T. M., Lewis J., Martin A. and Raithby P. R. J. Organomet. Chem., 429, NO. 3, C41-C45 (1992). 35. Alonso-Vante N., Schubert B. and Tributsch H. J. Catal., 112, No. 3,384-391 (1988). 36. Alonso-Vante N. and Tributsch H., in: Electrochemistry of Novel Materials, Ph. Ross and J. Lipkowski (Eds.), Vol. 3, pp. 1-63, VCH Publishers, New York (1994). 37. Alonso-Vante N., Schubert B. and Tributsch H. Materials Chem. Phys., 22,281-307 (1989). 38. Alonso-Vante N. Solar Energy Mat Solar Cells, 31, No. 4, 509-524 (1994). 39. Rothenberg G.,Fitzmaurice D. and Gratzel M. J. Phys. Chem., 96, No. 14, 5983-5986 (1992). 40. Hagfeldt A,, Lindstrom H., Sodergren S . and Lindquist S.-E. J. Electroanal. Chem., 381,No. 1, 39-46 (1995).
152
Ruthenium Based Chalcogenide Materials
41. Aharon-Shalom E. and Heller A. J. Electrochem. SOC., 129,2865-2866 (1982). 42. Heller A., Aharon-Shalom E., Bonner W. A. and Miller B. J. Am. Chem. SOC.,104, No. 25,69426948 (1982). 43. ORegan B. and Grltzel M. Nature, 353,737-740 (1991). 44. Nazeeruddin M., Kay A,, Rodicio I., Humphry-Baker R., Muller E., Liska P., Vlachopoulos N. and Gratzel M. J. Am. Chem. SOC.,115, No. 14,6382-6390 (1993). 45. Alonso-Vante N., Nierengarten J.-F. and Sauvage J.-P. J. Chem. SOC.Dalton Trans., No. 11, 1649-1654 (1994). 46. Bahadur L. and Pandey J. P. J. Appl. Electrochem., 22, No. 9,883-887 (1992). 47. Bedja I., Hotchandani S., Carpentier R., Fessenden R. W. and Kamat P. V. J. Appl Phys., 75, No. 10, 5444-5446 (1994). 48. Bedja I., Kamat P. V., Hua X., Lappin A. G. and Hotchandani S . Lungmuir, 13, No. 8,2398-2403 (1997). 49. Alonso-Vante N. Solar Energy Mat. Solar Cells, 31, No. 4,509-524 (1994). 50. Ashokkumar M., Kudo A,, Saito N. and Sakata T. Chem. Phys. Lett., 229, No. 4,383-388 (1994). 51. Fang J. H., Wu J. W., Lu X. M., Shen Y. C. and Lu Z. H. Chem. Phys. Lett., 270, No. 1, 145-151 (1997). 52. Vogel R., Hoyer P. and Weller H. J. Phys. Chem., 98, No. 12,3183-3188 (1994). 53. Liu D. and Kamat P. V. J. Electroanal. Chem., 347, No. 1-2,451-456 (1993). 54. Vinodgopal K., Bedja I. and Kamat P. Chem. Mater., 8 , No. 8,2180-2187 (1996). 55. Chandrasekharan and Kamat P. V. J. Phys. Chem., 104, No. 46, 10851-10857 (2000). 56. Henglein A. J. Phys. Chem., 104, No. 29,6683-6685 (2000). 57. Hagfeldt A. and Gratzel M. Chem. Rev., 95, No. 1,49-68 (1995). 58. Bahnemann D., Bockelmann D. and Goslich R. Solar Ener. Mat., 24, No. 1-4,561-583 (1991). 59. Hoffmann M. R., Martin S. T., Choi W. Y. and Bahnemann D. W. Chem. Rev., 95, No. 1,69-96 (1995). 60. Trillas M., Peral J. and Domenech X. Appl. Catal. B, 3, No. 1,45-53 (1993). 61. Tsai S . J. and Cheng S . Catal. Today, 33, No. 1-3,227-237 (1997). 62. Zang L., Liu C. Y. and Ren X. M. J. Chem. SOC., Faraday Trans., 91, No. 5,917-923 (1995). 63. Perez M., Torrades F., Garcia-Hortal J. A., Domenech X. and Peral J. , 109, No. 3, 281-286 (1997). 64. Bogdanoff P. and Alonso-Vante N. Ber. Bunsenges. Phys. Chem., 97, No. 7,940-942 (1993). 65. Bogdanoff P. and Alonso-Vante N. J. Electroanal. Chem., 379, No. 1-2,415-421 (1994). 66. Zhang X. G., Murakami Y., Yahikozawa K. and Takasu Y. Electrochim. Acta, 42, No. 2,223-227 (1997). 67. Bogdanoff P., Friebe P. and Alonso-Vante N. J. Electrochem. SOC.,145, No. 2, 576-582 (1998). 68. Shipway A. N., Katz E. and Willner I. ChemPhysChem, 1, No. 1, 18-52 (2000).
Chemical Physics of Nanostructured Semiconductors, pp. 153-1 82 A.I. Kokorin and D.W. Bahnemann (Eds.) 0 VSP 2003.
CHAPTER 6
Metal Nanoparticles on Semiconductor Surfaces: Electrochemistry and Photocatalysis Anatoly I. Kulak Institute of General and Inorganic Chemistry, National Academy of Sciences, Belarus Keywords: metal nanoparticles, nanomaterials, photoelectrochemistry, photocatalysis, electrocatalysis, semiconductor, titanium dioxide.
6.1. Introduction The formation of nano-structured metal contacts on a semiconductor surface is one of the promising means of improving semiconductor electrodes used in the photoelectrochemical solar cells, photocatalysts for the treatment and disinfections of waste water and air, and highly selective electrocatalysts. The advantages introduced by the surface modifycation of a semiconductor with metal nanoparticles are associated mainly with their highly pronounced electrocatalytic properties, which are especially potent in the case of noble metal particles as well as with their ability to influence the separation of photogenerated charges. The elevation of the electrocatalytic activity of semiconducting photoelectrodes (and, accordingly, the reduction of the overvoltage at the electrodes in photoelectrochemical solar cells) represents one of the conventional ways for improving their efficiency and increasing the semiconductor stability through its “kinetic protection”. To do this, numerous researches have been conducted on the formation of metal (Au, Ag, Pt, Pd, Cu, Ru, Bi, etc.) particles on the surface of semiconductors (n-GaP [l],p-WSez [2], p-MoSz [3], p-InP [4-lo], p-Si [ll-141, n-Si [15-191, n-GaAs, [20], p-GaAs [21,22], n-CdTe [23], Ti02 [24], etc.). It has been found that one of the main problems on this way is the necessity of a compromise between two factors. One of them is the substantial enhancement of the electron transfer rate constant owing to the electrocatalytic effect of metal particles. The second problem is the minimization of rising the rate of recombination with the participation of electronic surface states induced by metal ions, ad-atoms, or atom clusters in a semiconductor band gap. Semiconductor photoelectrodes modified by the deposited metal particles has been widely used in various photoelectrochemical processes including hydrogen and oxygen evolution and charge-exchange reactions with the participation of redox-pairs in regenerative solar cells. The deposition of metal particles possessing electrocatalytic activity can also be used for increasing the efficiency of the photoelectrochemical reactions, which are hardly feasible with the use of the naked semiconductor photoelectrodes. For instance, the modification of p-Si photocathode with small (20-200 nm) particles of Cu, Ag,
154
Metal Nanoparticles on Semiconductor Surfaces
and Au makes possible the synthesis of methane, ethylene, and carbon monoxide by way of C 0 2 photoelectrochemical reduction [ 141. In parallel with the employment of metal particles for the surface modification of semiconductor electrodes used in the photoelectrochemical cells, many efforts have been devoted to the similar activation of semiconductor photocatalysts applied for the systems of photocatalytic water treatment, photodecomposition of toxic organic pollutants, and photodestruction of pathogenic microorganisms. To take an illustration, the use of platinized Ti02 photoelectrodes provides a reasonable way for the photodecomposition of liquid water [25, 261. The loading of Ti02 photocatalysts with platinum enables one to perform the photosynthesis of methane from the acetic acid (photo Kolbe reaction) [27] and of ammonia from the azide-ions and water [28] as well as the oxidation of hydrocarbons [29] and interaction of active carbon with water [30, 311. It also gives the possibility to increase substantially the rate of trichlorobenzene photodecomposition [32]. The deposition of noble metals (Pt, Pd, Rh) and nickel onto the Ti02 surface leads to the specific initiation of the reactions involving hydrogen as a reactant (e.g. alkane-deuterium isotope exchange [33-351) or a reaction product (e.g. dehydrogenation of alcohols [36]). The Ti02-Au photocatalyst made as a nanocomposite consisted of nano-Ti02 cores and Au nanoparticle shells has exhibited a great efficiency in the interphase charge transfer during the process of S C N ion photodecomposition [37]. The silver-loaded titania photocatalysts possess high activity in the photocatalytic decomposition of ozone [38], photoreduction of various thiols [391, photodestruction of 1,4-dichlorobenzene [40], dehydrogenation and oxidation of alcohols (e.g. of 2-propanole [41,42]), decoloration of textile waste water [43], photokilling of bacteria [44], and others. For additional detailed information on this topic, one can use a number of review articles and books [e.g. 45-48]. In spite of a great number of investigations aimed at the preparation of photocatalysts and photoelectrodes based on the semiconductors surface-modified with metal nanoparticles, many factors influencing the photoelectrochemical processes under consideration are not yet clearly understood. Among them are: the role of electronic surface (interfacial) states and Schottky barriers at semiconductor / metal nanoparticle interface, the relationship between the efficiency of photoinduced processes and the size of metal particles, the mechanism of the modifying action of such nanoparticles, the influence of the concentration of electronic and other defects in a semiconductor matrix on the peculiarities of metal nanophase formation under different conditions of deposition process (in particular, under different shifts of the electrochemical surface potential from its equilibrium value), etc. Unfortunately, it is very difficult to draw any conclusions regarding the above problems only from the literature data since the different groups of investigators often used the essentially differing approaches and conditions for the preparation of photoelectrodes and photocatalysts as well as for the measurements of their functional characteristics and investigations of the metal nanophase morphology. Moreover, in a number of early studies on the modification of a semiconductor surface with small metal particles, there is at all no data on the sizes of these particles. Taking this into account, the present paper will describe the state of the art of semiconductor surface modification by metal nanoparticles with special emphasis on the experimental results obtained by the author and his coworkers [49591 with the use of well-defined thin-film Ti02 electrodes prepared and studied in the comparable conditions.
Chapter 6; A.I. Kulak
155
6.2. Dependence of Nanoparticles Morphology and Behavior in Electron Transfer Processes on the Mean of Metal Nanophase Deposition onto Semiconductor Surface 6.2.1. Contact Deposition of Metal Nanoparticles onto Semiconductor Surfaces It has long been known that the immersion of chemically active semiconductors, such as silicon, germanium, A"'BV compounds, into the solution, which contains the ions of metals (Au, Ag, Hg, Cu, etc.) with the electrochemical potential more positive than that of a semiconductor, leads to the formation of island metal deposit on their surface [60-621. In this case, the surface concentration of deposited metal particles as well as their size and other parameters of metal micro- or nanophase depend on numerous factors such as the concentration of metal ions in the solution, the duration of deposition process, the porosity of the surface passivation layer, and the chemical properties of a metal and a semiconductor being in use. On the other hand, one can also observe less evident phenomenon consisting in a spontaneous deposition of metal nanophase under the treatment of chemically inert semiconductors (such as titanium dioxide) with the solutions of readily reduced metal ions in the absence of the external bias, illumination, and reducing agents. During this process, metal particles are formed by the interaction of metal ions with donor defects or their associates in a semiconductor. In principle, this process may proceed both through the direct chemical reaction of donor centers with metal ions and through the electrochemical mechanism with a spatial separation of the sites of donor center oxidation and metal particle growth. With titanium dioxide and, probably, with many other semiconductors, the electrochemical mechanism of the spontaneous metal nanophase deposition seems to be more credible than pure chemical. It is evidenced by the observation that the resulted metal particles have a relatively large size and are positioned at the intervals much more (16-40 times) than the particle diameter [49,51]. In particular, under the immersion of Ti02 electrode characterized by the concentration of ionized donors (Nd) equal to cm-3into 10-5-103M water solutions of Ag' ions, the deposition of Ag particles takes place in a dark on the electrode surface. The average particle size increases from 0.8-0.9 to 9-10 nm at the prolongation of the time of electrode contact with a solution from 1 to 10 min. In the case of Ti02 electrode with Nd = 10l8~ m -all~ other , factors being equal, the average size of Ag particles rises from 0.7 to 3.2 nm (Fig. 6.1). Under more prolonged exposure of the electrodes to the solutions of Ag* ions, the size of particles as well as their surface concentration rise moderately. It is characteristic that, in both cases, only a small portion of the electrode surface appears to be covered with silver (no more than 2-6 % in the case of TiOz electrode with Nd = 1019cm"). The quantity of deposited silver (in atoms/cm2) on the surface of TiOz electrode is readily determined with knowledge of the current passed through the solution during the anodic oxidation of Ag particles at a linear potential sweep. At rather high concentrations of donor centers in the initial titanium dioxide (N&1 019cm-3),a detectable background current is observed on the potentiodynamic i,V-curves obtained for these electrodes in indifferent electrolytes in the absence of redox additives. This current is attributed to the oxidation of donor impurities in a semiconductor. After the treatment of the electrode with the solution of Ag' ions, the background current dies out almost completely and, instead of it, the well-pronounced peak corresponding to the anodic oxidation of Ag particles appears (Fig. 6.2).
Metal Nanoparticles on Semiconductor Surfaces
156
a N, = 10" cm" t = 1 min
N
2.
s?
dt
d N, = 10' cm" t = 10 min
N, = io" cm" t = 10 min
1
8 d, nm
9
1
N, = 10" cm" t = 1 min
.
4 '
____/
b
0
0
1
2
3
7 4
d. nm
Fig. 6.1. Size distribution of Ag particles on thin-film Ti02 electrode. The particles were deposited under the treatment of TiOz electrode with 0.001 M Ag2S04 for 1 min (a, b) and 10 min (c, d). The concentration of ionized donors in Ti02 (Nd)is equal to 10'' cm-' (a, c) and 10l8cm-3(b, d).
I
I
0.0
0.2
0.4
0.6
0.8
Electrode potential,V (vsAgfigCI)
Fig. 6.2. Potentiodynamic current-potential curves of oxidation of silver particles deposited on thin-film TiOz electrode. The curves have been obtained in deaerated 0.1 M KCSN + 0.5 M KzSO4 solution (curve 1), 0.1 M &Fe(CN), + 0.5 M KzSO4 (curve 2), 0.5 M KOH (curve 3), and 0.5 M KzSO4 (curve 4). The potential sweep rate is 40 mV/s.
157
Chapter 6; A.I. Kulak
The diminishing of the background current is associated with the consumption of donor impurities in the reaction with Ag' ions as well as with the re-distribution of the potential drop from the free (uncovered) TiOz surface to the surface of Ag particles. High chemical reactivity of metal nanoparticles and a large ratio of surface atoms to the total number of atoms in metal nanophase hinder passivation processes, and a complete oxidation of Ag particles with the average size of 1-3 nm proceeds at pH 5-7 even in the absence of depassivators. It should be noted that, when employing ordinary electrochemical potentiostatic setup for the determination of the surface concentration of Ag particles, the sensitivity of this method is no worse than 1013atoms/cm2, i.e. about 0.01 of monolayer coating. Taking into consideration that the average size of Ag particles deposited with this technique lies in the range 1 to 10 nm, one can estimate from the peaks of anodic current on the potentiodynamic i,V-curves that the surface concentration of nanoparticles is about 109-10'2 cmS2.It should be borne in mind that, when the average size of Ag particles exceeds 4-5 nm, the above method becomes less accurate because of the appearance of the passivation processes. The influence of these processes can be diminished by the introduction of depassivating additives (halide, rodanide, and sulfide ions) to the electrolyte (Fig. 6 . 2 ) provided that particular precautions (very careful removal of dissolved oxygen, etc.) are taken in order to prevent the loss of the chemically unstable smallest-sized fraction.
0.4
0.8
1.2
0.4
1.2
Electrode pterjial,V (vsAgpgCl) Fig. 6.3. Potentiodynamic current-potential curves of oxidation in 0.05 M HCI of Pd particles obtained by the following methods: 1, 5-8 - combined pyrolysis of PdClzand titanium resinate; 2 - adsorption of Pd2+ions on Ti02 film; 3 - vacuum deposition of Pd; 4 - galvanic deposition of Pd on TiO2. All samples except 5 were annealed in hydrogen at 500OC. The g/cmz. dV/dt = 20 mV/s. content of Pd: 1-5 - lo-' ; 6 - 3 ~ 1 0;-7~, 8 -
Using this approach, we have examined [64]the properties and concentration of electroactive (being able to take part in the electrode processes) Pd nanoparticles formed by the different methods: (i) combined pyrolysis of palladium chloride and titanium resinate
158
Metal Nanoparticles on Semiconductor Surfaces
(that is easily melting salts of resin acids - colophony derivatives), (ii) vacuum deposition of Pd onto the Ti02 surface, and (iii) adsorption of Pd2+ions by the Ti02 surface (Fig. 6.3). In the course of this investigation it was shown that combined thermodecomposition of titanium resinate and PdC12, which brought about a concurrent formation of titania phase and Pd nanoparticles incorporated into it, makes possible to obtain smaller-sized and more monodispersed Pd particles in comparison with those obtained by other methods (Table 6.1). Table 6.1.
Surface concentration (A',,), the ranges of particles Pd sizes (d) and average sizes of Pd particles (d,,) for TiOz-Pdfilms synthesized by combined pyrolysis of palladium chloride and titanium resinate (RTi) on air at 45OoC (1 h) and then annealed in hydrogen at 5OO0C and 8OO0Cfor 1 h, by vacuum deposition of Pd and by adsorption of Pd2' ions on TiOz film obtained by pyrolysis of titanium resinate Method for obtaining TiOz-Pd film Combined pyrolysis of PdClZand RTi Vacuum deposition of Pd Adsorption of Pd2+ ions on TiOz film
5OO0C d,nm
d,,nm
2-6
800°C
N,,cmm2
D, nm
d,, nm
N,, cm-'
4.5
8.6.10"
2-15
8
1.2.10"
2-20
9
3.3.10"
2-50
14
l.o*loll
2-120
17
4.4.10"
4-160
36
2.0*1010
6.2.2. Galvanic Deposition of Metal Nanoparticles onto Semiconductor Surfaces In a number of previous investigations, the electrodeposition of metals onto the semiconducting substrates has been studied [65-671. Among other things it was shown that the actual overpotential required for the deposition of metal (Pb, Ag, or Pd) onto the semiconductor substrate (ZnO, CdS, n-Gap, and Sn02) depended on the relative position of the redox potential of the metal with respect to the flat band potential of the semiconductor while the nucleation behavior of the metal was influenced by the differences in work function of the depositing metal and the substrate [ 6 5 ] .The results of another comparative study devoted to the electrodeposition of Pt films onto semiconductor (n-InP, n-GaAs) and glassy carbon supports suggest that the nucleation depends on the properties of the semiconductor while the growth depends only on the overpotential between the metal clusters and the solution [66]. In the processes of metal electrodeposition onto the semiconductor surfaces the electronic surface states play an important role as has been shown under the investigation of the electrodeposition of Pb on n-Ge [67]. In the course of our investigations on the electrochemical behavior of thin-film titanium dioxide electrodes it has been shown that, with cathodically biased TiOz electrodes being in a contact with the solutions of readily reduced metal ions, even a slight shift of the electrode potential in relation to its stationary value causes the growth of metal nanoparticles. During this process, the predominate growth of the currently available particles takes place, and the appearance of fresh species can be stimulated by the increase of the applied bias. At the deposition of Ag, Cu, Pd, Pt, and other readily reduced metal particles on the TiOz electrodes in the potentiostatic conditions, the increase of the cathodic current with time is observed. The current-time dependencies are often linear in a wide
Chapter 6;A.I. Kulak
159
8
6 2'
'44
za
2
t
Y
C
20 600
3 o 305 (d
u
0' " " " " 0 2 0 4 0 6 0 8 0
dE/dt, mV/s
Time, min
Fig. 6.4. Cathodic deposition of silver particles onto titanium-dioxide electrodes: a) time dependence of the surface concentration of silver particles in the potentiostatic conditions at +0.4V (vs. sat. AglAgCl); b) typical view of cyclic polarization curve obtained under cathodic de osition of Ag; c)current-time dependences for rotating disk Ti02 electrodes with Nd 10 c m 3 (curves 1,3,4) and 10'' cm-3 (curve 2); rotation rate was 900 (curve l), 300 (curve 2,3) and 180 (curve 4) mid'; d) typical current vs. electrode potential sweep rate curve at a fixed potential.
P -
160
Metal Nanoparticles on Semiconductor Surfaces
range of potentials and currents. This linearity results from the combination of two factors: the decay of new particle formation with time (the concentration of particles is proportional to the square root of time) (Fig. 6.4) and the diffusion-limited growth of the existing species [49, 681. Such current-time dependencies determine the peculiar type of potentiodynamic polarization i,V-curves, the characteristic feature of which is that the currents at the reverse potential sweep are essentially higher than those observed at the direct sweep; moreover, they govern a non-typical strong influence of the potential sweep rate on the cathodic current values (Fig. 6.4). Obviously, it is impossible to obtain the stationary potentiostatic i,V-curves for such processes because just after attaining the potential of beginning metal deposition a progressive linear increase of the cathodic current with time is observed, and the quazi-stationary current value is attained only after the complete covering of semiconductor surface with a metal layer. Characteristic of this process is practically unhindered electron change between the metal particles growing under cathodic bias and the semiconductor c-band as evidenced by a very small value of cathodic overvoltage (approximately the same as for the metal electrodes). This is not an ordinary fact because the process of metal deposition occurs at the electrode potentials (0.43-0.20 V), which are substantially higher than the flat-band potential of TiOz (-0.4 - -0.6 V); that is, the deposition of metals proceeds with a high (0.60.8 V) Schottky barrier at the TiOz/electrolyte interface. The existence of a distinct Schottky barrier on the surface of TiOz electrode free of the metal particles is confirmed by the data of impedance measurements (the characteristic shape of Mott-Schottky curves [51]) and by the high photovoltage (V, 0.5-0.6 V) obtained under UV illumination of the metal-particle-modified Ti02 electrode being in a contact with indifferent electrolyte. On the other hand, under the contact of this electrode with the electrolyte containing Ag' ions, the interaction (Le. the charge exchange) between Ag particles on the electrode surface and Ag' ions in the solution is very efficient so that the action of UV irradiation in the open circuit conditions causes only a slight cathodic shift of the electrode potential (Fig. 6.5). Evidently this shift corresponds with the establishment of the equilibrium between two half-reactions involved in the photocatalytic process. The first half-reaction is the deposition of silver onto the semiconductor surface (the cathodic partial process with the participation of the majority charge carriers, which is analogous to the dark one). The second one is the attendant anodic partial process related to the generation and consumption of photoholes. Under illumination of the electrode, the parameters of the electronic surface states may slightly change; moreover, in the case of the low-doped semiconductors, the intense illumination may offer a substantial increase of the concentration of the majority charge carriers owing to the photogeneration of charges. Hence some differences may exist between the kinetic characteristics of the dark reaction of metal ion reduction and the cathodic half-reaction involved in the total photocatalytic process, in spite of the participation of majority charge carriers in the both processes. Nevertheless, according to our experimental data [50, 511, the influence of a light on the parameters of the cathodic half-reaction is actually insignificant in many cases, among them in the process of the photocatalytic deposition of Cu and noble metal particles onto the surface of titanium dioxide and other wide-band semiconductors. Taking this into account, the dark cathodic part of the summary polarization curve can be reasonably used for the simulation of the reduction half-reaction involved in the total photocatalytic process. The above features of the cathodic growth of metal particles on the semiconductor Ti02 electrode surface are best consistent with the behavior of titanium dioxide characterized by
Chapter 6;A. I. Kulak
161
-
-
G] -0.8
-
0.0
l
0.5
.
I
1.0
.
1
1.5
.
1
20
.
I
25
.
l
30
.
l
35
.
I
4.0
.
4.5
Fig. 6.5. Time dependence of TiOz electrode potential under UV illumination in 0.2 M aqueous solution of K2S04 without additives (curves 2,3) and in the presence of 0.05 M Ag' ions ' ~ 1,2) and 5 x (curves 1,4). Concentration of ionized donors in TiOz was 1019~ m(curves 10" cm-3(curves 3,4).
the moderate or slightly increased concentration of ionized donors (Nd= 10'8-10'9 ~ m ' ~In ). this case, the electrochemical and photocatalytic behavior of the TiOz electrode modified by the metal nanoparticles can be simulated by the electric circuit composed of two parallelconnected independent parts: a set of metal nanoelectrodes connected directly with a current terminal, on the one hand, and a semiconductor surface free of metal particles and characterized by the existence of a space charge region at the boundary with electrolyte, on the other hand. With substantially decreased doping of titanium dioxide (up to Nd = 10'6-10'7cm3), the deposited Ag particles offer the essentially less values of the exchange current under the contact with the electrolyte contained Ag' ions as well as the less surface concentration of the metal nanophase (calculated in atoms/cm-2). As a result, the influence of metal nanoparticles is reduced mainly to the modification of the structure of existing electronic surface states rather than to the introduction of new ones that, in particular, enables a high open-circuit photovoltage to be generated under UV illumination (Fig. 6.5). We should include into the electric circuit simulating such electrode system an additional resistant or "diode" element, which will characterize the difficulties in contacting metal nanoparticles and current terminal, as well as an impedance element (more pronounced than in the previous case), which will represent the parameters of electronic surface states on the free semiconductor surface and that covered by metal nanoparticles. 6.2.3. Photocatalytic Formation of Metal Nanophase on Semiconductor Sufaces Photocatalytic and photoelectrochemical deposition of metals onto semiconductor surfaces has been of considerable interest for improving the characteristics of both the
Metal Nanoparticles on Semiconductor Surfaces
162
semiconducting photoelectrodes of the solar cells and the heterogeneous photocatalysts [4951, 69-75]. It has also been successfully used as a method for removing the toxic metal im urities (Hg, Pb, Cd, Ag, etc.) from wastewater [76-821 and concentrating of Ag' and Hg ions for analytic purposes [83]. The photoselective variant of this process has been investigated for the potential application in the fabrication of microelectronic circuits and devices [84, 851 and for the development of novel silver-halide and non-silver imaging systems [50, 51, 86, 871. The mechanism and kinetics of the process of small metal particles deposition onto the semiconductor surfaces have been extensively studied by many authors with the use of silver ion photoreduction on the titania surface as a model reaction [50,51,70,80]. In these works consideration has been given to the photogeneration of the electron-hole pairs in titanium dioxide with subsequent interaction of photogenerated holes with adsorbed water molecules, OH- ions, or especially added hole-accepting species and trapping of the electrons with the surface states followed by the reduction of Ag' ions. We have established that for more refined understanding of this process it is necessary also to take into account the role of the electronic surface states induced by the chemisorbed Ag' ions and growing Ag nanoclusters, the characteristic features of the interaction of silver ions and particles with donor defects and their associates in the oxide, the changes in the structure of a double electric layer localized at the sites of metal particle deposition, the variations of the electrode electrocatalytic properties originated from the appearence of nano- and microphase, etc. [49-51,56-58]. Since a comprehensive analysis of the above-mentioned deposition of small metal particles onto the semiconductor supports has been previously given in a number of papers, we shall not describe these processes in details but restrict our consideration to somewhat different way for the formation of metal nanoparticles on the surface of semiconductors, namely, to the photoinduced formation of metal nanoclusters resulted immediately from the photolysis of the photosensitive semiconductor matrix. The photoinduced formation of metal nanophase on semiconductor surface can be effected not only by way of heterogeneous photoreduction of adsorbed metal ions but also as a result of the direct interaction of photogenerated charge carriers with a semiconductor matrix. One of the classical examples of photochemical solid-phase generation of small metal clusters and particles is the process of silver halides (AgHal) photolysis, which has been much studied in the course of investigations on the formation of latent image centers in classical silver halide photography [88]. During the process of photolysis, in contrast to the heterogeneous photocatalytic deposition, metal nanophase may appear not only on the surface but also in the balk of a semiconductor. Its formation depends crucially on the mechanism and efficiency of metal atom (or ion) transfer to the growing cluster over the surface or in the interior of the solid matrix. In particular, to account for the possibility of silver cluster formation in AgHal, the classical Gurney-Mott theory takes into consideration the necessity of the following repeated alternative processes: the capture of electron by the Ag, cluster, the generation of the negatively charged Ag,- particle and subsequent transfer of Ag' ion to it with the formation of the larger cluster Ag,+l. Although a wealth of evidence on the regularities of Ag phase formation and its growth in AgHal matrix is now available, the major part of these data is adapted to the problems of the photographic imaging system, so it is rather complicated to use these data in the consideration of the photochemical and electrocatalytic behavior of the nanostructured semiconductor systems. In relation with this topic, we (together with Poznyak) have investigated the formation of metal nanophase on semiconductor electrodes and the electrochemical
T+
'
Chapter 6;A. I. Kulak
163
behavior of the resulting electrodes in the case of bismuth oxyhalides (BiOHal - BiOCl, BiOBr, BiOI) [89, 901. In their photochemical properties and behavior in photographic recording systems operated in a “direct darkening” regime [91] bismuth oxyhalides are quite similar to silver halides, although they do not exhibit the unique photographic sensitivity inherent in AgHal photomaterials as a result of their photodevelopment. In parallel with these features, bismuth oxyhalides offer the distinct semiconductor properties and highly pronounced photoelectrochemical activity, being at the same time less favorable to slow and not easily reversible ionic processes. Because of the combination of such properties these compounds are convenient as the electrodes for the electro- and photoelectrochemical systems [89,90,921.
1.5 1.0,.
0.8
20
I
1
1.5
25
20
25
30 ,
35 I
30
35
4.0 4.5 , , I , I
4.0
4.5
5.0 ,
,
5.0
hv, eV Fig. 6.6. Spectra of photocurrent (top) and of diffusion reflection (middle), and spectral dependences of ( q h c ~ ) ’(bottom) ~ for BiOI (l), BiOBr (2) and BiOI (3) electrodes at 0.1 V (1) and 0.4 V (2,3).
In our investigation we have used BiOHal electrodes obtained by the exhaustive anodic oxidation of a bismuth layer with a thickness of about 200 nm on a platinum substrate in aqueous solutions of potassium halides (KHal) using the method previously reported in full details [93]. The resulted BiOHal films exhibited porous structure (according to the BET data, the surface area was 47,25, and 6 m2/g for BiOCl, BiOBr, and
164
Metal Nanoparticles on Semiconductor Sugaces
BiOI, correspondingly) with the crystallite size ranged from 10 to 20 nm. From the diffuse reflection spectra and spectral dependencies of photocurrent (Fig. 6.6), it is evident that the band gap energies (Eg) for bismuth halides are 3.5 eV for BiOCl, 2.9 eV for BiOBr, and 1.9 eV for BiOI. Under the illumination of BiOHal electrodes with an actinic light in the conditions of potentiostatic polarization, the cathodic photocurrent is generated in indifferent electrolytes saturated with oxygen, with the maximum quantum yield in the range of 0.3-0.4 for BiOC1,0.07-0.09 for BiOBr, and 0.01 for BiOI [89]. Under the open circuit conditions, the illumination of BiOHal electrodes with a full spectrum of mercury quartz lamp causes the shift of the electrode potential to the anodic direction (typical for p-type semiconductors). In this case, in the indifferent electrolytes (neutral or acidified KHal solutions) saturated with oxygen, the high positive potential values are attained for BiOCl and BiOBr electrodes (Table 6.2). Table 6.2. Photoelectrochemical characteristics of BiOHal electrodes in 0.2M neutral and acidified KHal solutions saturated with oxygen Electrode Solution Potential of cathodic-to-anodic photo-current transi-tion, E,, (V) Open-circuit photovoltage, V,, (V) Maximum quantum yield of cathodic photocurrent
BiOCl KCl PH 6
KCl+HCl PH 2
BiOBr
BiOI
KBr PH 6
KBr+HBr PH 2
KI PH 6
0.6520.05
0.97i0.02
0.5OiO.05
0.78iO.02
0.25iO.02
0.3OiO.02
0.6220.03
0.24iO.02
0.46k0.03
0.06iO.01
0.025k0.005
0.08k0.01
O.lOiO.02
0.35iO.05
0.008~0.002
In the absence of the dissolved oxygen, the illumination leads to the gradual darkening of BiOCl and BiOBr electrode surface due to the accumulation of the photoreduced bismuth particles. At the contact with air, the electrode surface is clarified owing to the oxidation of Bi clusters, which possesses a high chemical reactivity. In deaerated atmosphere these Bi clusters can be oxidized under anodic polarization that is evident from the peaks of anodic current on the potentiodynamic current-potential curves in a high anodic potential region (0.8-1.1 V for BiOCl and 0.6-0.8 V for BiOBr; peak 111, Fig. 6.7). In contrast with this, the oxidation of Bi particles electro- or vacuum-deposited on the Pt substrate occurs at the essentially less potentials (in this case there are two peaks of current on the i,E-curves: peak I at -0.1-0 V and peak I1 at 0.2-0.4 V). The anodic oxidation of Bi particles vacuum-deposited onto the surface of BiOHal films is characterized by the appearance of all three peaks with peak I11 dominating over peaks I and I1 (Fig. 6.7). Peak I11 on the i,E-curves can originate from the oxidation of Bi particles separated from the semiconductor matrix by the space charge region in BiOHal film, hence, its position correlates with a flat band potential (0.95 V for BiOCl and 0.75 V for BiOBr in KHal/Hhal solutions, pH 2 [89]). The appearance of peaks I1 and I11 on the i,E-curves is also characterristic for the anodic oxidation of Bi particles obtained by the illumination of BiOHal films
165
Chapter 6; A.I. Kulak
on the Pt support under cathodic bias at E c Efb (Fig. 6.7). In this case, peak I1 is obviously connected with the oxidation of Bi particles localized in the immediate vicinity of the contact of BiOHal film with the Pt substrate, and peak I11 is of the same nature as for the the Bi particles vacuum-deposited onto the surface of BiOHal matrix.
I
I
I
I
f
Fig. 6.7. Potentiodynamic current-potential curves of oxidation of Bi particles obtained by: 1 vacuum deposition of Bi onto BiOCl; 2 -photolysis of BiOCl under applied bias; 3 photolysis of BiOCl without applied bias; 4 - photolysis of BiOBr under applied bias; 5 photolysis of BiOBr without applied bias; 6 - vacuum deposition of Bi onto Pt support
A distinguishing feature of this process is that the total anodic charge (calculated by the integration of peaks I1 and 111) consumed for the oxidation of Bi particles, which have been obtained by the illumination of BiOHal films under external bias (in the photoelectrochemical regime), is proportional to the value of cathodic charge transferred with a photocurrent. In contrast to this, during the process of BiOHal photoreduction in the absence of external bias (in the photochemical regime), Bi particles are initially accumulated during 5-7 min and then gradually disappear. This is obvious from the decrease of BiOHal darkening and from diminishing anodic charge (calculated from the peak I11 area) consumed for the oxidation of the photogenerated Bi particles (Fig. 6.8).
166
Metal Nanoparticles on Semiconductor Surfaces
As would be expected, the accumulation of Bi particles under an external bias leads to the gradual decrease of the photocurrent density owing to the rise of the recombination rate. It is obvious that such photocurrent decay cannot be detected during conducting the photoelectrochemical process in deaerated KHal solutions because the generated Bi atoms and nanoparticles are quickly oxidized by the dissolved oxygen with a regeneration of BiOHal.
0
5
10
15
7. nin
Fig. 6.8. Accumulation of the products of BiOCl film photolysis in 0.2 M KCl + HCl (pH 2) without applied bias (curve 1) and under applied bias (E = 0.0 V vs.AglAgC1) (curve 2); q, - the value of charge transferred by anodic current under the oxidation of photolytic Bi deposited onto BiOCl.
It is worth noting that in the processes of the direct photochemical or photoelectrochemical synthesis of Bi nanophase being considered, under the conditions of profound photolysis (at the pronounced darkening), the Bi particles range up to 2-5 nm according to the data of transmission electron microscopy. It is clear that under less exposure Bi particles may have essentially less size (the circumstantial evidence is that such particles possess high chemical reactivity and are very unstable on exposure to the air). It is possible to visualize them by the deposition of other metals (e.g. Ag) from the especial solutions employed for the autocatalytic chemical deposition of metals (so called “physical developers”) [9 11. In conclusion, the process of the photoinduced formation of Bi nanoclusters in the photosensitive BiOHal matrix has been considered here as an example, which illustrates the possibility of the direct photochemical formation of metal nanophase as well as the ways for the regulation of this process and identification of metal nanophase using electrochemical techniques. 6.3. Formation of Electronic Surface States in Semiconductor Band Gap as a Result of Deposition of Metal Particles on Semiconductor Surface Electrocatalytic, photocatalytic, and photoelectrochemical behavior of a semiconductor modified by the deposition of metal nanoparticles depends strongly on the
Chapter 6;A. I. Kulak
167
parameters of electronic surface states induced by the metal particles in the band-gap region on the semiconductor surface [49-511. The appearance of such surface states may manifest itself as the additional photoelectrochemical response beyond the region of the fundamental absorption of a semiconductor [94, 951 as well as in the electroluminescence spectra, etc., but the most convenient way for their identification and estimating their parameters is the employment of the electrolyte electroreflectance (EER) spectroscopy.
"r
2r
05
L
1
a1v
1.5
p
3a
0.5
0.0
Fig. 6.9. EER spectra registered at various electrode potentials for titanium dioxide electrodes modified by the deposition of different metal nanoparticles (Cu, Pd, Ag). Concentration of ionized donors in TiOz was ( 1 - 5 ) ~ l O~' ~m -Electrolyte ~. - 0.5 M KzS04.
168
Metal Nanoparticles on Semiconductor Surfaces
Fig. 6.9. (continuation) EER spectra of titanium dioxide (Nd 10’’ ~ m -modified ~) by the deposition of Ag nanoparticles.
In the EER method, the relative variation of the optical reflection coefficient ( A m ) of the electrode surface caused by the low-frequency harmonic modulation of the electrode potential is used as the informational signal [96-981. The high sensitivity of EER technique (it is possible to detect the AWR values of about IO”-lO-’) allow us to identify the impurity electronic levels in a semiconductor band gap at their very small concentrations and, therefore, to control the actual surface energy states rather than those distributed in the sub-surface region [49-571. As is seen from Fig. 6.9, the deposition of noble metal (Ag, Pt, Pd) and Cu particles on the surface of Ti02 electrodes gives rise to the EER peaks in the sub-band-gap region. These signals indicate the creation of the surface electronic states, the population of which may undergo perturbations under the modulation of the electrode potential. These peaks are the most pronounced at cathodically biased electrodes under circumstances where Fermi level of a semiconductor falls within the region of the energy distribution of the electronic surface states (Fig. 6.9). The energies of these electronic states depend on the properties of a semiconductor as well as on the nature of a metal and the average size of metal particles. For instance, the EER si nal from the Ag-modified Ti02 electrode with the concentration of ionized donors Nd = 10’ cm shifts to the low-energy region (from 2.50 to 2.38 eV) when the average size of Ag particles increases from 0.8 to 1.8 nm (Fig. 6.10). This corresponds to the increase of the depth of energy levels, formed by the Ag nanoparticles, in relation to the edge of the semiconductor c-band and through this - to diminishing their electric interaction with the c-band. In the limiting case, when the average size of Ag particles rises to 30-50 nm, they become “quazi-isolated” and do not contribute to the EER response. For the comparable-sized metal particles, the diminishing of donor concentration in Ti02 leads to somewhat increasing energy of the surface electronic states, which become more “shallow” in relation to the conduction band. Concurrent with this, the peak half-width decreases (for instance, from 0.38-0.41 eV for the Ag-modified Ti02
5 -
Chapter 6;A.I. Kulak
169
electrode with Nd = 10'' cm-3to 0.21 eV for that with Nd= 5x10" cm"). The origin of such shift of the surface energy states is not clearly understood being additionally masked by a great influence of the donor defect sub-system on the morphology and other properties of depositing metal particles.
Fig. 6.10. EER spectra of TiOz (Nd 10'' ~ m -electrode ~, potential -0.8 V) modified by Ag nanoparticles with different size distribution illustrated by histograms (the numbers on spectra correspond the numbers on histograms).
Thus, discussing the influence of the concentration of donors in a semiconductor matrix on the parameters of electronic surface states, we can only make the inference that there exists a certain tendency towards the increase of the interaction between metal nanophase and semiconductor matrix with lowering a matrix doping. In a similar manner, the conditions of deposition process (such as variations of surface potentials from its equilibrium value, the energy flow and/or charge carriers density, characteristics of the ionic adsorption and mass-transfer, surface diffusion coefficient for metal ad-atoms, etc.) strongly influence the size distribution of the resulting species. This plague the direct correlation of the electronic surface state parameters for the metal particles obtained by the different methods so that only general tendencies can be revealed. In particular, when the surface concentrations of metal nanophase deposited by different methods are comparable, the nanoparticles obtained galvanostatically (under small cathodic bias) induce, as a rule, the highest-energy levels in the semiconductor matrix (i.e. the most "shallow" impurity levels with respect to the c-band edge). The vacuum-deposited metal particles form the least-energy levels, and species obtained photocatalytically under open circuit conditions - the levels possessing intermediate energies. It is clear that, in spite of the similar surface concentrations of deposited particles, the size distribution of these species can vary essentially for different deposition methods (Fig. 6.11), and even a similarity of the average particle sizes does not yet ensure a resemblance of the properties of metal nanophase.
170
Metal Nanoparticles on Semiconductor Surfaces
It is especially difficult to compare the energies of electronic states formed during the deposition of different metals. On the whole, according to the “depth” of the electronic states in relation to the c-band edge, the metals can by arranged in the following order: Ag
5
10
15
20
1.5
2.0
drnn
0.0
0.5
1.0
d m Fig. 6.11. Size distribution of Ag nanoparticles photodeposited onto TiOz electrode in different conditions: 1 - light intensity 1 mW/cmz, illumination time 100 s; 2 - 10 mW/cmz, 10 s; 3 100 mW/cm2, 1 s; 4 - pulsed UV irradiation, 40 W/cm*, 2 111s.
It should be noted that all above-mentioned results have been obtained using polycrystalline titanium dioxide (anatase, rutile) [49, 511; on the whole, the same regularities are observed during the control experiments with the monocrystalline rutile. When going from poly- to nanocrystalline Ti02 obtained by zol-gel method, the EER spectrum of the oxide substantially changes [53]. In contrary to this, the EER response attributed to the electronic surface states induced, for instance, by the deposition of Cu nanoparticles on the nanocrystalline Ti02 appears to be very similar to that obtained in the case of the same deposition on the polycrystalline TiOz matrix (Fig. 6.12). In summary, the EER investigations suggest that the electroactive (i.e. being able to change their population under varied electrode potential) electronic states in the Ti02 band gap are formed during the deposition of Ag, Pt, Pd, and Cu nanoparticles on the surface of poly- and nanocrystalline semiconductor oxide. It is interesting that the energies
171
Chapter 6;A.I. Kulak
of the electronic surface states in a semiconductor band gap with respect to the band gap edge substantially decreases with increasing the size of the deposited metal particles, all other factors being equal. In this condition the impurity levels in a semiconductor band gap becomes more "deep"; therefore, there is very likely that they may act as recombination sites in the course of the photochemical and photoelectrochemical processes. On the other hand, the electronic surface states formed by the smallest-sized particles and characterized by the high energies are located on the energetic diagram near the edge of conduction band and are in a state of electronic equilibrium with this band. Hence, it is very probable that such particles would function as modifying agent varying the electrocatalytic properties of a semiconductor in dark processes. , . , . , . , . ,
--0.3
V
__ -______ -0.4 V .............. 0.5 V --0.6 V ,.I'
a
............. ........
,/"
,//' ',
-10- --0.3 V . - -- - - - - -0.4 V -8 ................ 0.5 V -0.6 V
-2-6 -
.
O
,I ,/
,./ ,/' ............. /,/'
,?
,
I-.
.....
/,'
/::
.....
.. ... . . ... .. . . . ;:.'
//
54 --4: 2-
*.
...........
.......... ..........
F
2',."'
.
-
/'
// 4
.b
x
.I,"
,
,
,
I
,
~
,
Fig. 6.12. EER spectra of untreated (a) and copper modified (b) nanostructured TiOz film electrodes under the negative polarization in 0.1 M acetic buffer solution (pH 6).
6.4. Electrocatalytic Activity of Semiconductor Electrodes Modified by SurfaceDeposited Metal Nanophase
It is well known that non-degenerated wide-band semiconductors tend to have low electrocatalytic activity. In particular, in the dark conditions under the electrode potentials higher than the flat-band potential (Efb), Ti02 electrodes with ordinary doping ( N d = 10'' -
172
Metal Nanoparticles on Semiconductor Sugaces
lo'' ~ m - are ~ )characterized by the extremely small current densities (0.1-10 pA/cm2) even in the presence of such strong reducing agents as sodium borohydride, sodium hypophosphite, and formaldehyde (Fig. 6.13). The deposition of metal nanoparticles (Pt, Pd, Ag, Au, Cu, Co, Ni) leads to the multiple increase of the anodic current density with the peculiarities of this electrocatalytic effect being substantially different for the noble metals and for Cu, Ni, Co, and Cd. For instance, after the cathodic deposition of rather small amounts of Cu nanoparticles ( 1 ~ 1 0-' ~5 ~ 1 0atoms/cm2) '~ on the surface of Ti02 electrode with Nd = 1019 ~ m - the ~ , typical view of i,V-curves for the resulting electrode in the process of the anodic oxidation of B H i ions (curves 4-6, Fig. 6.13) differs considerably from those obtained with the use of metal Cu electrode (curve 1) and initial T i 0 2 electrode (curves 7,8). The appearance of two peaks of current on the i,V-curve for metal Cu electrode is attributed to the oxidation of Cu and increasing the electrocatalytic activity of the electrode caused by the resulted oxidized species as well as to the following depassivation and passivation processes. In the case of this electrode, the inherent anodic current densities are several orders of magnitude greater than in the case of Cu-modified T i 0 2 electrode, in spite of the fact that the total surface area of Cu particles on the latter electrode is no more than 2-5 fold higher than the specific surface area S of metal Cu electrode. IO
,
. ,
,
.
,
,
. , .
.
,
5 10 L 6
.1.0
0.0
Electrode potential (vs. AglAgCI), V
0.1
Electrode potential (vs. AgIAgCI), V
Fig. 6.13. Anodic current vs. potential curves for the process of BHi ions oxidation on the bulk Cu electrode (curve 1; for comparison see curve 2 registered in the same conditions without BHi ions), on the initial Ti02 electrodes (curve 7 for Ti02 with Nd = lo-'' ~ m -curve ~ ; 8 for TiOz
with Nd 10'' cmS3)and on the Ti02electrodes surface modified with different concentration of Cu (curve 3 - 10l8atoms/cm2,curves 4 3 - 10l6atoms/cm2,curve 6 - lo'' atoms/cm2).The values of Nd for Ti02 were 10" cm-3(curve 5) and id9cm-3(curves 3,4,6). Curve 9 was obtained with the use of represented electrical circuit modeling the system 'Ti02 - Cu particles - electrolyte" (D - solid-state Schottky diode; R - electrical resistor; WE, RE and CE - working, reference and counter electrodes, correspondingly). Electrolyte: 0.1 M NaBH4 + 0.1 M NaOH. The potential sweep rate is 5 mV/s.
Chapter 6;A.I. Kulak
173
On examination of the electrocatalytic activity of metal nanoparticles in dark electrode processes, of significant interest is the appearance of the limiting current on the i,V curve for the Cu-modified Ti02 electrode, especially taking into account that it does not depend on the stirring of electrolyte and cannot be considered as a consequence of any diffusion limitation caused by the electrolyte solution. At the same time, this limiting current is very sensitive to the heating of the electrolyte as well as to the IR illumination of the electrode. In contrast to Cu particles, noble metal particles deposited onto the surface of Ti02 electrode exhibit strong electrocatalytic effect. With similar concentrations of deposited metals, the current values are 50-200 times higher in the case of the noble metals than for the Cu-modified Ti02 electrode. Moreover, with noble metals deposited, the above mentioned current limitation does not appear on the i,V-curves (Fig. 6.14). The current values decrease in the region of high electrode potentials where even the bulk metal electrodes composed of the same metals undergo passivation; besides this, under high anodic current densities the usual diffusion limitations begin to play role. The electrochemical properties of Ti02 electrodes modified by the Ni, Co, Cd, Bi, and Pb particles are intermediate between those of Pt-modified and Cu-modified ones and depend strongly on the surface concentration of metal nanophase and on the nature of reducing agent. For instance, the electrocatalytic action of Ni nanoparticles approaches that of Pt and Pd species in the oxidation processes limited by the dehydrogenation stage.
1 ' " " " " " 1
-1.2
-0.8 -0.4 0.0 0.4 0.8 Electrode potem,V (vsAgpgC1)
Fig. 6.14. Anodic current vs. potential curves for the process of BHi ions oxidation on the TiOz electrodes surface modified by Pd: 5x10'' atoms/cm' - curve 1; lo'' atoms/cm2-curve 2; Ag: lo'' atoms/cm2- curve 3; and by successive electrodeposition of Ag atoms/cm2)and Cu (lo" atomslcm') -curve 4.The values of Nd for TiOz were lo'* cm-3(curve 5 ) and 10'' cm-3 (curves 3,4,6). Electrolyte: 0.1 M NaBH4+ 0.1 M NaOH. The potential sweep rate is 5 mV/s.
To understand the above peculiarities of the electrocatalytic activity of metal nanophase in dark oxidation processes on the Ti02 electrodes, one should take into account the differences in the inherent electrocatalytic properties of deposited metals, on the one hand, and the data on the electronic states formed by the nanoparticles of these metals in a band gap of Ti02 electrode, on the other hand. A peculiar shape of i,V-curves obtained for
174
Metal Nanoparticles on Semiconductor Surfaces
the Cu-modified Ti02 electrodes and the appearance of a limiting current on these curves are attributed to the formation and properties of Schottky barrier at the Ti02/Cu interface [49, 991. The “deep” (with respect to the c-band edge) electronic surface states formed by the Cu nanoparticles are unable to increase the permeability of Schottky barrier to the tunneling electrons. So, probably, this barrier restricts the electron transfer from the Ti02 conduction band to the electronic levels of metal nanophase. Since such a barrier is not completely perfect, there always exist leaking currents. This permit the increased electrocatalytic activity of the Cu-modified Ti02 electrodes in comparison with the naked ones but the attained increase of activity is essentially less pronounced than in the case of TiOz electrodes modified by the deposition of noble metals particles. In the latter case, the “shallow” (in relation to the c-band edge) electronic surface states formed by the metal nanoparticles, when their concentration is sufficiently high, provide the typical “mild puncture” of Schottky barrier, and it does not thereafter influence the process of electron transfer. The nanoparticles of metal fall in the immediate electric contact with the semiconductor matrix, then the electrochemical behavior of the electrode is determined only by the electrocatalytic properties of these particles. From the above reasoning one could expect that the pre-deposition of small amounts of noble metals on the Ti02 surface in a form of the intermediate sub-layer, which can induce the electroactive electronic surface states in the Ti02band gap, may enhance the electrocatalytic effect of subsequently deposited Cu particles. Actually, the photocatalytic ’~ which on its own only deposition of silver particles in amount of 5 ~ 1 0 atoms/cm-2, slightly increases the electrocatalytic activity of Ti02 electrode, leads to 2-3-fold enhancement of the electrocatalytic activity of Cu particles subsequently deposited in a relatively high concentration ( 10’6-10’7atoms/cm-2) [52]. 6.5. Impedometric Investigation of Titania Electrodes Surface-Modified with Metal Particles The occurrence of small metal particles on the surface of semiconductor electrodes affects the relationships between the electrochemical capacity and the electrode potential and frequency. This is mainly due to the appearance of both the capacity of electronic surface (interfacial) states (C,,*) induced by the small metal particles in a semiconductor band gap and the electrochemical capacity associated with a double electric layer at the contact of metal particle surface and electrolyte. The latter factor is well-pronounced only in the condition of effective electron exchange between metal particles and semiconductor bands, as shown, for instance, for the noble metal particles cathodically-deposited onto the surface of film TiOz electrodes. In the simplest case, the electric circuit simulated such electrode system can be represented by two parallel-connected parts, one of them corresponding to free semiconductor surface contacted with electrolyte, and other being associated with the surface of metal particles. If especial precautions in order to rule out the presence of reducing metal ions in the solution are not taken, the substantial changes in the electronic states on the uncovered semiconductor surface represented by the corresponding changes in the model circuit may take place. Evidently, the adequate description of electrode impedance in the case of the actual contact between the surface-modified semiconductor and electrolyte represents a very complicated problem owing to the appearance of some hardly measured parameters
175
Chapter 6; A.I. Kulak
associated with the particle size effects and the contributions of surface and interfacial electronic states. Because of this, we shall restrict our consideration to the simplest situation, when metal particles act as a set of metal microelectrodes electrically connected with the bulk of a semiconductor without the formation of Schottky barrier at the metal/ semiconductor interface, in condition that the average size of these particles is sufficiently large for leaving the properties of the contact unaffected.
-0.6
-0.3
0.0
0.3
0.6
Electrode potential,V Fig. 6.15. Mott-Schottky dependences for non-modified (curves 1-5) and modified by Ag nanoparticles (curves 6-10) TiOz electrodes. The frequency of alternative current was 10 KHz (1, 6); 1 KHz (2, 7); 700 Hz (3, 8); 500 Hz (4, 9); 100 Hz (5, 10). Electrolyte: 0.25 M Na2S04.
The Mott-Schottky plot obtained experimentally for the Ag-modified Ti02 electrode, which satisfy the above requirements, differs from that for the initial electrode by the slope value, with an insignificant shift of the point obtained after extrapolating the plot to the electrode potential axis (Fig. 6.15). Since for the realization of such electrode system we have used a semiconductor characterized by the high concentration of ionized donors, under consideration of Mott-Schottky dependence it is worthwhile to take account of the Helmholtz layer capacity (CH)placed in series with the space charge capacity [loo]:
C2= CH
-2
+ 2[V - Vfb - kT/e]/&&,eNd
As a first approximation, the CH value is assumed to be independent from the electrode potential. When small metal particles are present on a semiconductor surface, the capacity of Helmgoltz layer at the metal particles is connected in parallel with the "-C,,-CH-'' segment and makes an additive contribution to the total electrochemical capacity:
Metal Nanoparticles on Semiconductor Surfaces
176
C = mCM
+ n{CH-2 + 2[V - V fi - k T l e ] l ~ & , e N ~ } - “ ~
(6.2)
where m and n are correspondingly the portions of the electrode surface covered by metal particles and free of them. The results of the numerical calculations made with the use of the above equation suggest that the introduction of the parallel-connected capacity of the metal particles into the model circuit of the electrode system leads to the changes in a slope of Mott-Schottky plots corresponding to the formal increase of Nd value for a given semiconductor (Fig. 6.16). Thus a correction derived from the capacity of metal microelectrodes on a semiconductor surface should be introduced into the commonly accepted determination of Nd value from the slope of Mott-Schottky curves: Nd
* = {me,
-2
+n{CH +2[V-Vy, - k T l e ] l & ~ , e N ~ } - ~ ’ ~ } ~ -2
Nd
n{CH +2[V-Vf, - k T / e ] I & & , e N d } - ” 2 } 3
The above ratio depends certainly on the electrode potential. This dependence becomes more pronounced with the increase of both m value and the difference between the electrode potential and Efb(Fig. 6.15,6.16).
0,4
0,6
E-E,, V Fig. 6.16.Equivalent circuit representing the electrode system ‘Ti02I Ag nanoparticles I electrolyte” and calculated Mott-Schottky curves for different portions of the electrode surface covered b metal particles (m): 0 (curve 1); 0.02 (curve 2); 0.2 (curve 3). CH = 10 pF/cm2; CM = 20 pF/cm! Nd = 1 . 5 ~ 1 0~’ ~m - ~ .
For the more accurate description of the Mott-Schottky dependences of semiconductor electrodes modified with small metal particles, it is reasonable to take into induced by the account the contribution of the capacity of electronic surface states (C,,*)
Chapter 6;A. I. Kulak
177
chemisorptiodreduction of ions and formation of metal ad-atoms and clusters. Taking into account the charge of these electronic states (Q,) one can draw inference about the shift of the extrapolation point of the Mott-Schottky plot by the following value: 2
E’= E@ t k T l e + Q , , I C , - E E , , N ~ / ~ C , At the same time, in the case of the above-considered system “Ti02 - metal nanoparticles”, the role of surface states is negligible at sufficiently high frequencies of alternative current (25-8 KHz) and anodic bias more than 4-5 kT/e over &.The analysis of Mott-Schottky plots (Fig. 6.16) shows that the appearance of the capacity of Helmholtz layer at the metal nanoparticles actually causes the marked changes in the slope of these curves. Accordingly, the N d values calculated from the equation (6.1) appear to be substantially higher than Nd value of the initial Ti02. To take an illustration, the Nd value calculated by the equation (1) from the Mott-Scottky plots for the Ti02 electrode modified by Ag particles with a size 15 nm at a surface concentration of about 10l2cmP2is 6.2N.3 times higher than that of the naked Ti02. This correlates well with the value of the ratio Nd*/Nd= 5.9N.2, which has been calculated by the equation (6.3) with consideration for the total surface area of metal particles determined independently on the basis of the electronic microscopy data [51,101].
6.6. Interaction of Metal Nanoparticles with the Associates of Donor Defects in Wide-Band-Gap n-type Semiconductors The above-mentioned peculiarities of the metal nanoparticle influence on the electrochemical behavior of wide-band-gap semiconductor oxides can be put up into use for clarifying the properties of the initial (uncovered with metal) semiconductor electrodes in the case when a segregation of dopant centers followed by the formation of donor aggregates takes place in the moderately and highly doped semiconductors. This situation is fairly common for the wide-band-gap polar semiconductors such as titanium dioxide [4951, 1021. In these semiconductors, the donor centers predominantly occur in a form of their associates. Depending on the semiconductor pre-history and on the concentration of donor centers, the donor associates may have different sizes (from few nanometers to tens micrometers, in the latter case these aggregates can be already considered as the inclusions of microphase such as Magneli phases or Ti203). At sufficiently high concentrations of donor associates, they can overlap and form the continuous donor clusters, which may act as percolation channels ensuring, in parallel with zone charge transfer, another way of transfer through the bulk of a semiconductor to the electrode I reaction medium (electrolyte) interface [51]. Such continuous donor clusters are characterized by ohmic or hopping type of conductivity in charge transfer processes since they exhibit the properties of degenerated semiconductors or metals and do not form both the intergrain Schottky barriers and the barrier at the cluster I electrolyte interface. The formation of these clusters leads to the appearance of the additional channel for the electron transfer and causes the increase of leaking currents through the currently available Schottky barrier at the electrode I electrolyte interface [51]. On the other hand, when the donor clusters are sufficiently large and represent the microphase inclusions, it is inconceivable that the potential barriers may appear at the interface between donor clusters and semiconductor matrix in which they are formed. In this case, the charge streams flowing through the semiconductor bands and
178
Metal Nanoparticles on Semiconductor Surfaces
through the donor clusters will be completely independent. In actual practice, for instance, in dark anodic processes on the thin-film TiO2, W03, Biz03 and other electrodes, the role of the continuous donor clusters appears to be very large in some cases [49-511. In particular, a dark anodic current flowing under the electrode potentials much higher than Efbproves to be significantly attributed to the electron transfer along the continuous donor clusters rather then to the tunneling through the space charge region in the oxide conduction band. This conclusion is confirmed by numerous experimental facts, e.g. the essentially more high values of anodic currents than those calculated with regard to the tunneling efficiency (see, for example, the results [ 1031 obtained for the ZnO electrode), the characteristic shape of i,V-curves, and much more strong dependency of the dark current value from the chemical properties of the electrolyte components than one would expect [511. The detailed discussion of the roles the continuous donor clusters play and the mechanisms of their functioning is beyond the scope of this paper; therefore we shall restrict our consideration to the brief inspection of their influence on the formation and properties of the deposited metal nanoparticles. It is clear that during the contact deposition of metal (in the dark open circuit conditions) the metal particles are formed on the Ti02 surface at the sites of donor location and, to the most extent, at the continuous donor clusters. The accessible deepness of donor centers “extraction” remains to be relatively small (probably, no more than several oxide lattice constants) because of its limitation by the low diffusion of oxygen, which is necessary for the oxidation of donor centers. To explain the experimentally observed appearance of a rather small concentration of relatively big Ag particles on the Ti02 electrodes, account must be given to the possibility of the lateral electron transfer from the neighboring donor centers, that is the electrochemical mechanism being of widespread occurrence in the processes of the chemical deposition of metals. In any case, metal nanoparticles deposited via the interaction of semiconductor donor centers with soluble metal ions prove to be localized at the sites of the electrode surface exposure of donor centers including continuous donor clusters. In a similar manner, during the process of the existing metal particles growth and the deposition of new species using cathodically biased electrode in a solution of metal ions, the growing metal phase will be also localized at the sites of the surface exposure of the continuous donor centers. The reason for this is that namely these sites possess substantially enhanced electrocatalytic activity in comparison with the stoichiometric oxide surface and exhibit the properties of current channels non-restricted by the Schottky barrier at the interface with the electrolyte. Actually, a peculiar “decoration” of the sites of donor centers accumulation and donor clusters localization by the metal nanoparticles takes place in the dark processes of metal particle deposition onto the surface of the chemically inert wide-band-gap oxides. The increased electrocatalytic activity of the wide-band-gap semiconductor electrodes resulted from the deposition of metal nanoparticles on their surface may be also regarded as a kind of such “decoration”. On the other hand, to eliminate the electrocatalytic activity of the continuous donor clusters, use could be made of the blockage of their surface by the deposition of dielectric polymers such as poly-o-phenylenediamine [ 1041. It has been shown that the cycling of the anodic potential of WO3 films in the solution of o-phenylenediamine leads to a drastic decrease in the dark current caused by the selective deposition of dielectric polymer (Fig.6.17) This decrease is also accompanied by the changes in impedance characteristics and in other parameters [104].
Chapter 6;A. I. Kulak
179
.%
9‘El
B
V
Fig. 6.17. Cyclic voltammograms of o-phenylenediamine (lo-’ M) oxidation for W 0 3 thermaltreated (350’C) anodic films @) and smooth platinum electrode (c): first sweep (curves 1) and repeated sweep (curves 2); scan rate was 80 mV/cmz.The left picture shows a schematic representation of the morphology of thermal-treated anodic W 0 3 film: tungsten support, highly defective oxide (including the continuous donor clusters),moderately doped oxide (non-shaded region), poly-o-phenylenediamine deposits.
With respect to its methodology, this approach is closely related to the previously proposed improvement of performance of n-WSe2 electrodes by electrochemical polymerization of o-phenylenediamine at surface imperfections [ 1051. By and large these results, together with the above-considered experimental data, suggest that of crucial importance for the processes of electron transfer through the wide-band-gap-oxide semiconductor / electrolyte interface are the local electrode processes on the surface of “microelectrodes”, i.e. continuous donor cluster exposures, rather than tunneling of electrons through the semiconductor space charge region.
6.7. Conclusion The surface concentration, size distribution and other properties of metal nanoparticles formed in a dark on the surface of the inert wide-band-gap semiconducting oxides under contact, photocatalytic, or photoelectrochemical deposition depend substantially on the concentration, bulk distribution, and energy characteristics of donor defects in the initial semiconductor substrate. As a rule, the necessary condition for the formation of the smallest-sized particles in the highest surface concentration is the maximum shift of the surface potential of semiconducting matrix from its equilibrium value during metal deposition. This is part of the reason for the experimentally observed fact that the particles formed in the condition of photocatalytic deposition are characterized by less average size and cover superior portion of surface than those obtained under cathodic deposition, all other factors being equal. The formation of electronic surface states in a semiconductor band gap by metal nanoparticles is the major factor that determine the efficiency of electron exchange between metal particles and a semiconductor matrix. It also influences the efficiency of electro-
Metal Nanoparticles on Semiconductor Surfaces
180
catalytic process as a whole. According to the data of electrolyte electroreflectance spectroscopy, Ag, Pd, and Pt nanoparticles induce the formation of “shallow” (with respect to the c-band edge) surface states in a band gap of TiOz, which provide nearunhinderedelectron exchange between metal particles and semiconductor c-band. As the size of metal particles increases, the surface state levels in TiOz become more “deep” in relation to the edge of c-band. The changing of energy levels and of other surface state characteristics is one of the promising ways for the effective control of electrocatalytic, photocatalytic, and photoelectrochemical properties of semiconductors loaded with metal nanoparticles. To take an illustration, the deposition of additional intermediate sub-layer composed of Ag or Pt nanoparticles onto the surface of TiOz electrode enables a drastic increase in the electrocatalytic activity of Cu particles, which form sufficiently “deep” surface states in TiOz band gap, and leads to the minimization of Schottky barrier at the interface between semiconductor and metal particles.
Acknowledgements The author wish to thank his colleagues Prof. D. V. Sviridov, Dr. S . K. Poznyak and Dr. E. A. Streltsov for essential contributions to developing this work and for very helpful discussions. The help of Dr. T. I. Kulak with the preparation of this paper is also gratefully acknowledged. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.
Nakato Y., Abe K. and Tsubomura H. Ber. Bunsenges. Phys. Chem., 80, No. 10, 1002-1007 (1976). Kautek W., Gobrecht J. and Gerischer H. Ber. Bunsenges. Phys. Chem., 84, 1034-1040 (1980). Szklarczyk M. and Bockris J. O’M. Znt. J. Hydrogen Energy, 9 , No. 10, 831-834 (1984). Heller A. and Vadimsky R. G. Phys. Rev. Lett., 46, No. 17, 1153-1156 (1981). Heller A., Aharon-Shalom E., Bonner W. A. and Miller B. J. Am. Chem. Soc., 104, No. 25, 6942-6948 (1982). Szklarczyk M. and Bockris J. O’M. J. Phys. Chem., 88, No. 22,5241-5245 (1984). Chandra N., Wheeler B. L. and Bard A. J. J. Phys. Chem., 89, No. 23,5037-5040 (1985). Kuhne H.-M. and Schefold J. Ber. Bunsenges. Phys. Chem., 92, 1430-1435 (1988). Kuhne H.-M. and Schefold J. J. Electrochem. SOC., 137, No. 2,568-575 (1990). Schefold J. and Kuhne H.-M. J. Electroanal. Chem., 300,211-233 (1991). Dominey R. N., Lewis N. S . , Bruce J. A., Bookbinder D. C. and Wrighton M. J. Am. Chem. Soc., 104, NO. 2,467-482 (1982). Szklarczyk M. and Bockris J. O’M. J. Phys. Chem., 88, No. 9, 1808-1815 (1984). Szklarczyk M., Bockris J. O’M., Brusic V. and Sparrow G. Znt. J. Hydrogen Energy, 9, No. 8, 707-715 (1984). Hinogami R . , Nakamura Y., Yae S . and Nakato Y. J. Phys. Chem.B, 102, No. 6,974-980 (1998). Nakato Y., Ueda K. and Tsubomura H. J. Phys. Chem., 90, No. 22,5495-5406 (1986). Bockris J. O’M. and Kainthla R. C. Znt. J. Hydrogen Energy, 13, No. 6,375-383 (1988). Nakato Y., Ueda K., Yano H. and Tsubomura H. J. Phys. Chem., 92, No. 8,2316-2324 (1988). Yae S., Kitagaki M., Hagihara T., Miyoshi Y., Matsuda H., Parkinson B. A. and Nakato Y. Electrochim. Acta, 47, 345-352 (2001). Lewis N. S . J. Electroanal. Chem., 508, 1-10 (2001). Allongue P., Souteyrand E. and Allemand L. J. Electrochem. SOC., 136, No. 4, 1027-1033 (1989). Meier A., Uhlendorf I. and Meissner D. Electrochim. Acta, 40, 1523-1535 (1995). Meier A. and Meissner D. Multiple Nan0 Contacts. In: Nanoparticles in Solids and Solutions, J. H. Fendler and I. Dekany (Eds.),pp. 421-449, Kluwer, Netherlands, (1996).
Chapter 6; A.I. Kulak
181
Arce E. M., Ibanez J. G. and Meas Y. Electrochim. Acta, 40, No. 2,263-266 (1995). Chandrasekbaran N. and Kamat P. V. J. Phys. Chem.B, 104, No. 46, 10851-10857 (2000). Sat0 S. and White J. M. Chem. Phys. Letr., 72, No. 1, 83-86 (1980). Mills A. and Porter G. J. Chem. Soc., Faraday Trans., 78,3659-3669 (1982). Kraeutler B. and Bard A. J. J. Am. Chem. Soc., 100, No. 13,4317-4318 (1978). Nosaka Y., Norimatsu K. and Miyama H. Chem. Phys. Lett., 106, No. 1-2, 128-131 (1984). Izumi I., Dunn W., Wilboum K., Fan F. and Bard A. J. J. Phys. Chem., 84,3207-321 1 (1980). Sat0 S. and White J. M. Chem. Phys. Lett. 70, No. 1, 131-134 (1980). Sat0 S. and White J. M. J. Phys. Chem. 85, No. 4,336-341 (1981). Ushida H., Katoh S. and Watanabe M. Electrochim. Acta, 43, No. 14-15, 2111-2116 (1998). Courbon H., Herrmann J.-M. and Pichat P. J. Catal., 72, No. 1, 129-138 (1981). Courbon H., Herrmann J.-M. and Pichat P. J. Catal., 95, No. 3, 539-545 (1981). Herrmann J.-M., Courbon H. and Pichat P. J. Catal., 108, No. 2,426-432 (1987). Pichat P. New J. Chem., 11, No. 1, 135-141 (1987). Dawson A. and Kamat P. V. J. Phys. Chem.B, 105, No. 5,960-966 (2001). Ohtani B., Zhang S.-W., Ogita T., Nishimoto S. and Kagiya T. J. Photochem. Photobiol. A: Chem., 71, 195-198 (1993). 39. Tada H., Teranishi K., Inubushi Y. and Ito S. Lungmuir, 16, No. 7, 3304-3309 (2000). 40. Lee W., Shen H.-S., Dwight K. and Wold A. J. Solid State Chem., 106, 288-294 (1993). 41. Sclafani A., Mozzanega M.-N. and Pichat P. J. Photochem. Photobiol. A: Chem., 59, 181-189 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38.
(1991). 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63.
Sclafani A. and Herrmann J.-M. J. Photochem. Photobiol. A: Chem., 113, 181-188 (1998). Sokmen M. and Ozkan A. J. Photochem. Photobiol. A: Chem., 147,77-81 (2002). Sokmen M., Candan F. and Sumer Z. J. Photochem. Photobiol. A: Chem., 143,241-244 (2001). Hoffmann M., Martin S., Choi W. and Bahnemann D. W. Chem. Rev., 95, No. 1,69-96 (1995). Photocatalytic Purijkation and Treatment of Water and Air, D. E. Ollis, H. Al-Ekabi (Eds.), Elsevier, Amsterdam, 1993. Blake D. M. Bibliography of Work on the Photocatalytic Removal of Hazardous Conpounds from Water and Air, National Renewal Energy Laboratory, 1994. Kamat P. V. Chem. Rev., 93, No. 2,267-300 (1993). Kulak A. I. Electrochemistry of Semiconductor Heterostructures, Univ.Press, Minsk, 1986. Kulak A. I. The investigation of electrochemical and photoelectrochemical processes on the titanium dioxidefilms, Ph. D. Thesis, Belarus State Univ., Minsk (1980) (in Russian). Kulak A. I. Photoelectrochemical and photochemical processes in systems based on the semiconductor heterostructures, D.Sc.Thesis, Belarus State Univ., Minsk (1990) (in Russian). Kulak A., Kokorin A. and Sviridov D. J. Mater. Res., 16, No. 8,2357-2361 (2001). Poznyak S. K., Pergushov V. I., Kokorin A. I., Kulak A. I. and Schlapfer C. W. J. Phys. Chem., 103, NO. 8, 1308-1315 (1999). Poznyak S. K., Kokorin A. I. and Kulak A. I. J. Electroanal. Chem.,442,99-105 (1998). Kulak A. I., Sviridov V. V., Pakhomov V. P. and Shchukin G. L. Electrokhimiya, 16, No. 1,104107 (1980) (in Russian). Streltsov E. A., Kulak A. I., Sviridov D. V. and Pakhomov V. P. Electrokhimiya, 19, No. 4, 546548 (1983) (in Russian). Streltsov E. A., Lazorenko-Manevich R. M., Pakhomov V. P. and Kulak A. I. Electrokhimiya, 19, No. 3, 365-368 (1983) (in Russian). Streltsov E., Kulak A. and Lazorenko-Manevich R. Electrokhimiya, 20, No. 2, 21 1-214 (1984). Streltsov E. A., Pakhomov V. P., Lazorenko-Manevich R. M. and Kulak A. I. Electrokhimiya, 19, No. 2,232-235 (1983) (in Russian). Frank1 D. R. Phys. Rev., 128, No. 6,2609-2613 (1962). Matsas E., Dyner L., Primachenko V. E. and Snitko 0.Surface Sci., 19, No. 1, 109-117 (1970). Boddy P. J. and Brattain W. H. J. Electrochem. SOC.,109, No. 9, 812-818 (1962). Mychko D. I., Malchenko S. N., Streltsov E. A., Branitsky G. A., Kulak A. I. and Sviridov V. V. Vesti Acad. Nauk BSSR. Ser.khim.nauk, No. 4, 42-46 (1986) (in Russian).
182
Metal Nanoparticles on Semiconductor Surfaces
64. Streltsov E. A., Sviridov V. V., Kulak A. I., Mychko D. I. and Malchenko S. N. Mater. Chem. Phys., 25, NO. 3, 315-322 (1990). 65. Bindra P., Gerischer H. and Kolb D. M. J. Electrochem. SOC., 124, No. 7, 1012-1018 (1977). 66. Allongue P. and Souteyrand E. J. Electroanal. Chem., 286,217-237 (1990). 67. Ehlers C., Konig U., Staikov N. and Schultze J. W. Electrochim. Acta, 47, No. 1-2, 379-385 (2001). 68. Streltsov E. A., Sviridov D. V and Kulak A. I. Electrokhimiya, 20, No. 12, 1671-1674 (1984). 69. Darwent J. J. Chem. Soc., Faraday Trans., 77, 1703-1709 (1981). 70. Sahyun M. R. V. and Serpone N. Lungmuir, 13, No. 19,5082-5088 (1997). 71. Nosaka Y., Ishizuka Y. and Miyama H. Ber. Bunsenges. Phys. Chem., 90, 1199-1204 (1986). 72. Ohtani B., Iwai K, Nishimoto S.-I. and Sato S. J. Phys. Chem. B., 101, 3349-3359 (1997). 73. Hemnann J.-M., Tahiri H., Ait-Ichou Y., Lassaletta G., Gonzalez-Elipe A. R., Fernandez A. Appl. Catal., B 13, NO. 3-4,219-228 (1997). 74. Aspens D. E. and Heller A. J. Phys. Chem., 87, No. 24,4919-4929 (1983). 75. Aspens D. E. and Heller A. J . Vac. Sci. Technol. B, 1, No. 3,602-607 (1983). 76. Reiche H., Dunn W. and Bard A. J. J. Phys. Chem., 83, No. 17,2248-2251 (1979). 77. Serpone N., Ah-You K., Tran R. P., Harris R., Pelizzetti E. and Hidaka H. Solar Energy, 39, No. 6,491-498 (1987). 78. Lin W. Y. and Rajeshwar K. J. Electrochem. Soc., 145, No. 8,2751-2756 (1998). 79. Tanaka K., Harada K. and Murata S . Solar Energy, 36, No. 2, 159-161 (1986). 80. Herrmann J.-M., Disdier J. and Pichat P. J. Catal., 113, No. 1,72-81 (1988). 81. Chen D. and Ray A. K. Chem. Eng. Sci., 56, 1561-1570 (2001). 82. Hilmi A., Luong J. H. and Nguyen A.-L. Chemosphere, 38, No. 4, 865-874 (1998). 83. Suarez M., Mills A., Egdell R. and Compton R. G. Electroanalysis, 12, No. 6, 413-419 (2000). 84. Kelly J. J., Rikken J. M. G., Jacobs J. W. M. and Valster A. J. Vac. Sci. Technol. B, 6, No. 1,4852 (1988). 85. Jacobs J. W. and Rikken J. M. J. Electrochem. SOC., 136, No. 12,3633-3640,3641-3647 (1989). 86. Mollers F., Tolle H. J. and Memming R. J. Electrochem. SOC., 121, No. 9, 1160-1167 (1974). 87. Kelly J. J. and Vondeling J. K. J. Electrochem. SOC.,122, No. 8, 1103-1107 (1975). 88. Mees C. E. K. and James T. H. The Theory ofthe Photographic Process, Macmillan Comp., New York (1966). 89. Poznyak S. K. and Kulak A. I. Electrochim. Acta, 35, No. 11/12, 1941-1947 (1990). 90. Poznyak S. K. and Kulak A. I. Khimiya vysokhich energij, 24, No. 1,46-51 (1990) (in Russian). 91. Sviridov V. V., in: Nonsilver Photographic Processes, A. L. Kartuzhansky (Ed.), pp. 242-307, Khimiya, Leningrad (1984) (in Russian). 92. Poznyak S . K., Sviridov V. V. and Kulak A. I. Electrokhimiya, 20, No. 7,996-1000 (1984). 93. Poznyak S. K. and Kulak A. I. J. Electroanal. Chem., 278, No. 1,227-247 (1990). 94. Kolb D. M., Przasnyski M., Gerisher H. Z. Phys. Chem. (BRD), 93, No. 1-6, 1-14 (1974). 95. Butler M. A. and Ginley D. S. Appl. Phys. Lett., 42, No. 7,582-584 (1983). 96. Cardonna M. Modulation Spectroscopy, Academic Press, New York (1969). 97. Vos K. and Krusemeyer H. J. J. Phys.C: Solid St. Phys., 10, 3893-3915 (1977). 98. Pujadas M. and Salvador P. J. Electrochem. Soc., 136, No. 3, 716-723 (1989). 99. Streltsov E. A., Lazorenko-Manevich R. M., Pakhomov V. P. and Kulak A. I. Electrokhimiya, 19, No. 8, 1148 (1983) (in Russian). 100. Gomes W. P. and Cardon F. Progr. SUI-$Sci., 12, 155-216 (1982). 101. Sviridov D. V., Kulak A. I. and Streltsov E. A. Vestnik Belarus. Univ., Ser.2, No. 1, 8-1 1 (1986). 102. Yu N. and Halley J. W. Phys. Rev. B., 51, No. 8,4768-4776 (1995). 103. Pettinger B., Schoppel H.-R. and Gerischer H. Ber. Bunsenges. Phys. Chem., 78, No. 5,450-455 (1974). 104. Sviridov D. V. and Kulak A. I. Thin Solid Films, 198, 191-198 (1991). 105. White H. S., Abruna H. D. and Bard A. J. J. Electrochem. SOC.,129, No. 2,265-271 (1982).
Chemical Physics ofNanostructured Semiconductors, pp. 183-202 A.I. Kokorin and D.W. Bahnemann (Eds.) 0VSP 2003.
CHAPTER 7
Photocatalysis: Initial Reaction Steps Detlef W. Bahnemann, Ralf Dillert* and Peter K. J. Robertson** Institut fiir Technische Chemie, UniversitAt Hannover, Germany
* EcoTRANSfair Gesellschaft fur Umwelt und Gesundheit mbH, Braunschweig, Germany
* * Centre for Environmental Engineering and Sustainable Energy, The Robert Gordon University, Aberdeen, United Kingdom
Keywords: electron transfer, heterogeneous catalysis, photochemistry, photoelectrochemistry, Titanium dioxide Summary: During the past twenty-five years research and development in the area of titanium dioxide photocatalysis have been tremendous. The present review describes the basic prin-ciples of photocatalysis, focusing in particular on important mechanistic and kinetic aspects.
7.1. Introduction Photocatalysis, i.e., using semiconductor particles under band gap irradiation as little micro reactors for the simultaneous reduction and oxidation of different redox systems, has been intensively studied during the last 25 years since the pioneering work of Carey et a1 [l]. The main focus of these studies seems to be the investigation of the principal applicability of photocatalytic system for the efficient treatment of water and air streams polluted with toxic substances. Several review articles on this topic have recently been published [ 2 ] . In some cases, pilot-scale or even commercially available reactors have already been constructed, especially when titanium dioxide is used as the photocatalyst [3]. However, due to the inherent complexity of this minute photoelectro-chemical system, details of the underlying reaction mechanism of photocatalysis are even today still far from being understood. In contrast to an ordinary photoelectrochemical cell which employs an external bias voltage to deliberately separate oxidation and reduction processes in different compartments of the reactor, in photocatalysis both processes occur on the surface of the same semiconductor particle, usually only separated by a distance of a few angstroms. Moreover, as is evident from basic principles, the reaction rate of the overall process will be limited by the
184
Photocatalysis: Initial Reaction Steps
slowest reaction step, which in most cases is not known. Consequently, most laboratory studies have in the past been restricted to the measurement of overall degradation kinetics of organic model compounds (for a review see, e.g., ref. [2]). While such unified overall kinetic constants might be sufficient for the theoretical and practical design of photocatalytic reactors and pilot plants for the application of photocatalysis to pollution abatement, a better understanding of the mechanistic details of the photocatalytic process is certainly required before several basic problems, such as the preparation of catalysts with improved photocatalytic activity or with an extended spectral response and activity in the visible wavelength region, can be addressed in any systematic way. Some information on the primary processes of photocatalysis has been obtained by laser flash photolysis studies using colloidal semiconductor solutions and suspensions. Early flash photolysis studies suffered, however, from the inherent problem that a large number of electrodhole pairs were generated in each semicon-ductor particle (usually titanium dioxide) per laser pulse [4]. Due to the considerable progress in synthetic methods to produce nanosized semiconductor clusters with extremely small and well-defined diameters, several research groups were able to employ experimental conditions which ensured that on the average less than one electronhole pair was formed per semiconductor cluster during each light pulse [5-71. Consequently Serpone et al. [ 5 ] and Bowman et al. [6] have recently published very interesting work using picosecond or even sub-picosecond time resolution to study the primary events following band gap excitation of titanium dioxide colloids. Bahnemann and co-workers [7] have discussed details of the interfacial electron transfer between semiconductor particles and solutes in the surrounding electrolyte. This paper concentrates on a detailed description of the primary events occurring immediately after the absorption of a photon within a single titanium dioxide particle in an aqueous environment. This restriction was made, because 1) titanium dioxide seems to be the most active photocatalyst, and 2) the photocatalytic treatment of polluted water seems to be a promising application for an interfacial electron transfer serving the environment. 7.2. Primary Processes upon Bandgap Irradiation of Semiconductor Particles
Absorption of a photon with an energy hv greater or equal the bandgap energy E, of the semiconductor (Le., 3.2 eV for titanium dioxide in its anatase modification) generally leads to the formation of an electronhole pair in the semiconductor particle (reaction (7.1) and Fig. 7.1).
where e-cb represents a conduction band (CB) electron e-cb and h+,,b a positive hole in the valance band (VB) of the semiconductor. Ultrafast laser flash photolysis experiments using colloidal titanium dioxide have shown that the generation of these charge carriers happened
Chapter 7; D. W. Bahnemann, et al.
185
within a few femtoseconds after the absorption of a photon [6].Immediately after the laser flash very broad featureless transient absorption spectra ranging from 400 nm to 800 nm are obtained (cfi Fig. 7.2) [5-71. These absorption spectra are generally attributed to trapped electrons and holes.
Fig. 7.1. Schematic presentation of the processes occumng in photocatalysis upon bandgap irradiation of a semiconductor particle
After their generation according to reaction (1) both conduction band electrons and valence band holes are migrating to the surface of the semiconductor particle. The transit time T needed by these charge carriers to reach the surface of the particle is given by Equation (a)
where R is the radius of the particle and D the diffusion coefficient of the excited charge carriers [8]. Taking a value of D = 5x10” cm2s-’ [9] and a particle radius of 2.5 nm (radius of particles typically used in experiments with colloidal titanium dioxide) the average transit time is only 1.3 ps. Even in bigger particles such as those used in photocatalytic systems for water treatment (e.g., 21 nm for Degussa P25 titanium dioxide [lo]) the transit time is only some ten picoseconds. Reaching the surface these charge carriers are trapped in subsurface and surface states of the particle (reactions (7.2) and (7.3)):
Photocatalysis: Initial Reaction Steps
186
e-cb
+ e-,
(7.2)
h+"b + h+,
(7.3)
where e-, and h', are representing the trapped electron and trapped hole, respectively. 0.6
*E
l'""""""l
0
0.4
0
. I
&
E 0 B 4
0.3 e
0.2 0.1
400
Absorption under N
-01
450
1
1
1
500
1
550
-a,
I
Absorption under 0, , l , I
600
650
700
.
750
Wavelength [nm] Fig. 7.2. Transient absorption spectra measured 20 ns after laser excitation (Aex = 355 nm) in 02-sat. and N2-sat.solutions, respectively, pH 2.3,1.0x104 mol L-' colloidal TiOz-particles,absorbed photon concentration per pulse: 9 ~ 1 0 mol - ~ L-', adopted from [7a].
These trapped charge carriers exhibit strong optical absorptions. The position of the absorption maximum is strongly affected by the presence of suitable electron acceptors and donors in the surrounding aqueous phase. Exploiting this effect it has been shown in early laser flash photolysis studies that the trapped electron exhibits a strong optical absorption around 650 nm (Fig. 7.3, top) while the trapped hole absorbs predominantly at shorter wave-lengths, Le., around 430 nm or even shorter (Fig. 7.3, bottom) [4c, 4d]. Thus, it might concluded that the observed absorption spectra resulted from the superposition of the spectra of the trapped charge carriers, Le., the extremely broad and featureless spectra indicate the simultaneous observation of trapped electrons and trapped holes [5,71. As has been shown by time-resolved flash photolysis measurements in colloidal titanium dioxide suspensions trapping is a very fast process. Rothenberger et al. performed picosecond and nanosecond transient absorption experiments on titanium dioxide and observed that the electron trapping time was faster than 30 ps, the time resolution of their laser system [4e]. The trapping time for holes was estimated to be < 250 ns. In a recent picosecond study by Serpone et al. on titanium dioxide colloids solutions of varying diameters it was observed that the spectra of trapped electrons as well as of trapped holes are fully developed after a laser
Chapter 7; D. W. Bahnemann, et al.
187
pulse of 30 ps [ 5 ] . Another recent picosecond study by Bowmann et al. had shown that the time needed for the full development of the spectrum of the electron is approximately 200 fs [6].
-0.3 -0.5
g 2
0.9
9
0.6
*
1 2.5
0
0.5
1.5 2 Time after Laser Pulse [ ~ s ] 1
1.2
. I
8
m
8 0.3
4
G
0
-0.3 -0.5
0 0.5 1 1.5 2 Time after Laser Pulse [)IS]
2.5
Fig. 7.3. Transient absorption vs. time signals observed upon laser excitation (Aex = 355 nm) at 680 nm and 480 nm, respectively, pH 2.3, 1 . 1 ~ 1mol 0 ~L-' colloidal Ti02-particles, absorbed photon concentration per pulse: 2 . 7 ~ 1 0mol - ~ L-', 02-sat.,adopted from [7a].
7.3. Chemical Nature of Trapped Charge Carriers Generally it is assumed, that Ti'" cations at the surface of the titanium dioxide particle are reduced by the light induced electrons forming Tiln cations [ 111 which can be considered to be intrinsic surface states localised about 0.1 eV below the conduction band edge, Le., within the bandgap [12]. An equilibrium between these trapped electrons and free electrons is assumed, but in an acidic medium nearly all electrons are trapped in surface states [lla]. On the basis of their laser flash photolysis measurements Hoffmann and co-workers have extended this mechanistic picture [13]. These authors assume that the CB electrons are trapped in two different Ti"' sites (reactions (7.4) and (7.5))
Photocatalysis: Initial Reaction Steps
188
ecb+ Ti"0H w TiInOH
(7.4)
e-cb+ Ti"
(7.5)
TiIn
where Ti"0H is a surface-trapped electron while TiIn denotes a bulk trapped electron, respectively. The dynamic equilibrium of reaction (7.4) represents a reversible trapping of a CB electron in a shallow trap. Hoffmann et al. have estimated that these trapped electrons lie in the range of 25 to 50 meV below the conduction band edge of TiOz Degussa P25 [13]. Reaction (7.5) represents the irreversible trapping in a deep trap. From the analysis of their experimental results of the investigation of the charge carrier recombination kinetics in titanium dioxide colloidal solutions and in dispersions Serpone et al. and Bowman and co-workers have also assumed the existence of two different traps [5,6]. Concerning the nature of the electron-trapping centers, several problems arise. Assuming that the absorptions of the trapped electrons around 650 nm (1.7 eV) correspond to a transition between a classical surface state and the conduction band, this surface state is located near the middle of the band gap. As has been already pointed out by Bahnemann et al. this assignment cannot be correct because it has been found experimentally that the reduction of molecular oxygen occurs via transfer of a trapped electron [7]. This process would not be possible thermo-dynamically if the electron originated from an energy state being 1.7 eV below the conduction band of Ti02. Accordingly, Bahnemann et al. assumed that the absorption is due to excitation of a trapped electron within a surface molecule [7a]. The chemical nature of the trapped holes has not been clearly clarified yet. Older reports assume that the holes are trapped at the titanium dioxide surface in adsorbed hydroxy groups yielding weakly adsorbed hydroxyl radicals (reaction (7.6)) [ 14, 151.
Howe and Gratzel deduced from esr investigations that the hole is trapped in a subsurface oxygen anion (reaction (7.7)) [l la]. hCvb+ Ti'V-02--Ti'V-OHj Ti'V-O'--Ti'V-OH
(7.7)
Other groups are assuming that the trapped hole is an oxygen radical centered at the surface of the titanium dioxide particle, having an energy state lower than the valence band edge of the semiconductor (reaction (7.8)) [ l Id]. + H+ h+vb+ Ti'V-02--Ti'V-OHj Ti1V-02--Ti1V-O'
(7.8)
On the other hand, an extended model with two different types of traps appears to be much better suited to explain the experimental observations obtained by Bahnemann and coworkers [7a]. They have therefore proposed the following modification of reaction (7.3). It is
Chapter 7; D. W. Bahnemann, et al.
189
envisaged that at least two different trap sites for holes exist on the surface of the T i 0 2 particle. While holes which are trapped in energetically deep traps, h+,,d, can be characterized by their transient absorption around 430 nm, those initially residing in shallow traps, h',,s, do not posses such spectral features. Following generation, all holes are rapidly trapped in either of these energy states (reactions (7.9) and (7.10)).
h'
@ h+,,s
(7.10)
The holes trapped in shallow traps are probably excited thermally into the valence band, so that an equilibrium with free holes is indicated in reaction (7.10). Shallowly trapped holes, h',,S, will therefore have a comparable reactivity to detrapped holes, h'. While both types of trapped holes will recombine with the trapped electrons within the first 200 ns after their generation following reaction (7.1), only holes excited thermally from the shallow traps have the chance to migrate to the energetically more favored h+,,d site (c$ reaction (7.1 1)).
It is assumed that deeply trapped holes, h+,, are chemically equivalent to surface-bound hydroxyl radicals. Weakly trapped holes, on the other hand, that are readily detrapped apparently posses an electrochemical potential close to that of free holes and can therefore be considered to be chemically similar to the latter. Their shallow traps are probably created by surface imperfections of the semiconductor nanocrystals. From these traps the charge carriers recombine or they are transferred by interfacial charge transfer to suitable electron acceptors or donors adsorbed at the surface of the semiconductor.
7.4. Fate of Trapped Charge Carriers
7.4.I. Recombination Kinetics The charge carriers formed upon absorption of light (reaction (7.1)) can recombine in a radiative or non-radiative way according to reactions (7.12) to (7.15). This is clearly seen from the rather rapid depletion of the transient absorption spectra recorded during laser flash photolysis studies (see Fig. 7.4).
+ Ti02 + energy
(7.12)
+ h+vb+ Ti02 + energy
(7.13)
e-cb t h++, e-,
Photocatalysis: Initial Reaction Steps
190
+ Ti02 + energy
(7.14)
+ h', + TiOz + energy
(7.15)
e-cb+ h',
e-,
The recombination kinetics of the charge carriers have been studied in detail by the groups of Gratzel, Serpone and Colombo [4e, 5,6]. Since recombina-tion of electrons and holes is monitored by transient absorption techniques most of the observed decay is due to reaction (7.15).
1
-40x1s
*11,5ps
--o-100 ns
--+-
-m-
3.5 ps
I
'
T
'
I
,
I
,
250 ns
0.8
8 0
3 9
e
0.6
0 . CII
8 0.4
2
0.2
n " 400
1
450
.
1
500
550
600
I
650
,
700
Wavelength [nm] Fig. 7.4. Transient absorption spectra measured at various times after laser excitation (Aex = 355 nm), pH 2.3, 1.0x104 mol L-' colloidal Ti02-particles, absorbed photon concentration per pulse: 2.2~10" mol L-', 02-saturated,adopted from [7a].
The f i s t picosecond laser spectroscopic study to examine charge carrier trapping and recombination dynamics was reported by GrPtzel, Serpone et al. [4e]. They used an argonpurged titanium dioxide sol with 12 nm particles prepared by hydrolysis of TiC14, and they documented the nature of the kinetics of recombination of the charge carriers: where the number of electrodhole pairs per particle is much less than 0.5 recombination is expected to take place by first-order processes, but when the average number of such pairs per particle is >30, recombination should occur by second order kinetics. The mean lifetime of a single electrodhole pair was 30 k 15 ns at low occupancy of electrodhole pairs in the titanium dioxide particles. At high occupancies, where recombination followed second-order kinetics,
Chapter 7; D. W. Bahnemann, et al.
191
the bulk rate coefficient for recombination was (3.2 f 1.4) x lo-" cm3s-l. Serpone et al. have examined colloidal titanium dioxide sols (prepared by hydrolysis of TiC14) with mean particle diameters of 2.1, 13.3, and 26.7 nm by picosecond transient absorption and emission spectroscopy [5]. Absorption decay for the 2.1 nm sols was found to be a simple first-order process, and electrodhole recombination was 100% complete by 10 ns. For the 13.3 and 26.7 nm sols absorption decay follows distinct second-order biphasic kinetics; the decay times of the fast components decrease with increase in particle size. 10 ns after the excitation pulse, about 90% or more of the photogenerated electrodhole pairs have recombined such that the quantum yield of photooxidations must be 10% or less. The faster components are due to the recombination of shallow-trapped charge carriers, whereas the slower components (z > 20 ns) reflect recombination of deep-trapped electrons and holes. Bowman and coworkers characterized the subpicosecond dynamics of titanium dioxide sols employing particle sizes of about 2 nm prepared by hydrolysis of titanium tetraisopropoxide [6]. From their spectral results the authors inferred that the average lifetime of an electrodhole pair is 23 k 5 ps, and substantial electrodhole recombination occurs within the first 30 ps. A second-order recombination rate constant of (1.8 k 0.7) x lo-'' cm3 s-l for trapped electrons with holes has been obtained [6a]. 7.4.2. Charge Transfer Kinetics 7.4.2.a. Interfacial Electron Transfer
In most experiments and applications with titanium dioxide photocatalysts, molecular oxygen is present to act as the primary electron acceptor. Usually the electrons trapped as Ti(II1) are transferred to dioxygen adsorbed at the semiconductor surface yielding peroxyl radical anions (reaction (7.16)) [ 161.
02'- + H'
+ HO;
(7.17)
Depending on the pH of the suspension these superoxide radical anions can also exist in the protonated form (reaction (7.17)) [17]. Beside the electron transfer from the semiconductor to adsorbed molecular oxygen also the direct transfer to an organic molecule is possible. This type of photocatalytic reaction, yielding an organic radical anion, has been found to occur with 1,4-benzoquinone [ 181, tetrachloromethane [ 191, and several nitroaromatic compounds [20]. But electrons can also be transferred very efficiently to (adsorbed) metal cations [21]. In the investigations of Bahnemann et al. different decay kinetics and evolution of the transient absorption spectra of titanium dioxide colloidal solutions upon bandgap irradiation have been observed depending upon the presence of molecular oxygen, air, or molecular ,nitrogen, respectively [7]. In every case, a biphasic decay of the transient absorption signal was
192
Photocatalysis: Initial Reaction Steps
observed. Following a fast initial decay, the remaining 20-40% of the original signal height decayed much more slowly. While in the presence of molecular nitrogen this portion of the signal appeared to be stable even over a period of 200 ms, its decay rate increased with increasing O2 concentration. Considering the limited number of data points a rate constant k = 7.6 x lo7 L mol-' s-' has been determined by Bahnemann et al. for the reaction of a trapped electron with molecular oxygen [7]. 7.4.2.b. Direct Interfacial Hole Transfer
A significant body of literature proposes that the photocatalytic oxidation of organic or inorganic solutes may occur by either indirect oxidation via a surface-bound hydroxyl radical (i.e., a trapped hole at the particle surface) or directly via the valence-band hole before it is trapped either within the particle or at the particle surface.Interfacia1 hole transfer from titanium dioxide to organic and inorganic solutes has been studied recently in [4f, 6c, 71. An example of the latter paper is shown in Fig. 7.5.
4 0
" 3 2 *
c 0
-20
0 20 40 60 80 Time after Laser Pulse [p]
100
Fig. 7.5. Transient absorption vs. time signals observed upon laser excitation (Aex = 355 nm) at 500 nrn in the presence of various DCA- concentrations, pH 2.0, 1 . 0 ~ 1 mol 0 ~ L-' colloidal - ~ L-', air-sat., TiO,Pt( 1%)-particles, absorbed photon concentration per pulse: 1 . 6 ~ 1 0 mol adopted from [7al.
Grabner et al. have shown that in titanium dioxide sols containing chloride (which is either introduced into the solution as HC1 to adjust the pH or is present on the particle surface when Tic& is used as starting compound to prepare TiO2) Cli- radical anions are formed. Their formation was postulated to occur by direct valence-band hole oxidation of surface adsorbed C1(reactions (7.18), (7.19)) [4fl.
Chapter 7; D. W. Bahnemann, et al.
193
h+"b+ C1- + C1'
(7.18)
C1' + c1- + c1;-
(7.19)
It has been observed that these Cli- radical anions oxidize phenol yielding phenoxy1 radicals (reaction (7.20)) [4fl. PhOH + Cli- + PhOH7'
+ 2 C1-
(7.20)
Interfacial hole transfer dynamics from titanium dioxide (Degussa P 25) to SCN- has been investigated by Colombo and Bowman using femtosecond time-resolved diffuse reflectance spectroscopy [6c]. A dramatic increase in the population of trapped electrons was observed within the first few picoseconds, demonstrating that interfacial charge transfer of an electron from the S C N to a hole on the photoexcited titanium dioxide effectively competes with electron-hole recombination (reactions (7.12) - (7.15)) on an ultrafast time scale [6c]. Since Bahnemann and co-workers have observed that a comparatively high amount of trapped holes are formed when partially platinized titanium dioxide particles are subjected to ultra band gap irradiation (CJ? Fig. 7.6), they have chosen this system to study the dynamics of the photocatalytic oxidation of the model compounds dichloroacetate, DCA-, and S C N [7]. To explain their experimental observations these authors have used a model assuming two energetically different types of hole traps (see our detailed discussion above). ~~
3.5
-absorption -absorption -difference
after 20 ns after 5 p
3
0.5 400
450
500 550 600 Wavelength [nm]
650
700
Fig. 7.6. Transient absorption spectra measured at 20 ns and 5 ms, respectively, after laser excitation (Aex = 355 nm) and difference spectrum, pH 2.3, ~ . O X ~ Omol - ~ L-' colloidal Ti02/F't(l%)-particles, absorbed photon concentration per pulse: 1 . 6 ~ 1 0mol - ~ L-', air-sat., adopted from [7a].
194
Photocatalysis: Initial Reaction Steps
While the initial height of the transient absorption signal attributed to energetically deep traps, h+,,d, i.e., the concentration of h+tr,d,is considerably decreased by an increasing dichloroacetate concentration, the kinetics of its decay is not effected. It was therefore concluded that h+tr,d do not react with dichloroacetate [7a]. However, since the h'tr,d concentration is reduced considerably in the presence of DCA- (cfi Fig. 7 3 , either the free holes, h', can be directly transferred to adsorbed DCA- molecules (reaction (7.1 8)) or shallowly trapped holes, h',,, are detrapped (reaction (7.10)) to react with DCA- in the nanosecond time scale via reaction (7.21). h'
+ DCA- + DCA'
(7.21)
A similar reactivity of trapped holes has previously reported by Bahnemann et al. [4c, 4d] who studied reactions in colloidal T i 0 2 P t suspensions with an average particle diameter of approximately 12 nm. While the addition of ethanol as a hole scavenger resulted in a considerable increase of the rate of disappearance of the h', absorption, the addition of citrate and acetate mainly led to a decrease of its initial absorption height. It was concluded that strongly adsorbed ionic species would primarily react with free holes while weekly adsorbed molecules will mainly react with long-lived h', in a diffusion-controlled process [4c, 4d]. The direct charge transfer to dichloroacetate proposed in reaction (7.21) requires that the scavenging molecules are adsorbed on the TiOz surface prior to the adsorption of the photon. Otherwise, this reaction could not compete with the normal hole-trapping reactions (7.9) and (7.10). So the adsorption of the model compound DCA- on the titanium dioxide surface prior to the bandgap excitation appears to be a prerequisite for an efficient hole scavenging. A detailed kinetic analysis of the time-resolved spectroscopic data revealed an extremely good correlation with independent adsorption measurements [7]. It has been calculated that 20% of all T i 0 2 particles carry on average one adsorbed DCA- anion. The direct one-electron oxidation of dichloroacetate immediately follows the hole transfer from the bulk to the TiOz surface and, in principle, a maximum photonic efficiency of 0.2 would be possible under the experimental conditions. However, much lower efficiencies have been observed during the steady-state photocatalytic oxidation of dichloroacetate in the presence of T i 0 2 colloids [ 2 2 ] , suggesting that a considerable number of holes either recombine with the electrons or are trapped at the surface hydroxyl groups yielding the transient absorption around 430 nm. These surface-bound hydroxyl-radicals are apparently unreactive toward dichloroacetate. Thus, the model incorporating the direct hole trapping by adsorbed dichloroacetate molecules, which has been proposed by Bahnemann and co-workers, appears to be probable [7]. Moreover, calculations using the Marcus electron transfer theory for adiabatic processes which result in a reorientation energy of 0.64 eV suggest that also in the case of SCN- the hole transfer occurs in the adsorbed state [7].
Chapter 7; D. W. Bahnemann, et al.
195
7.4.2.c. Hole Transfer through the Intermediate Formation of Hydroxyl Radicals
In photocatalytic degradation experiments with acetate in dioxygen-containing suspensions of TiO, evidence had been obtained that holes as well as hydroxyl radicals are acting as oxidizing species [9,231. Acetate is readily degraded when aqueous suspensions of TiO, and acetate are irradiated in the presence of molecular oxygen [9,23]. As seen in Fig. 7.7, the degradation rates of acetate depend strongly on the pH of the suspension.
x-x-
a-
0
100
200
300
400
Illumination time / min
Fig. 7.7. Photocatalytic oxidation of acetate, lOmM sodium acetate, 0.5 g/l TiOz (Degussa P25), aqueous oxygen saturated suspension, T = 298 K, adopted from [23].
In acidic suspensions (pH 3.0) formate and formaldehyde have been detected as the only products of the photocatalytic oxidation of acetate (cfi Fig. 7.8). In alkaline suspension (pH 10.6) the main products are glycolate and formate accompanied by smaller amounts of glyoxylate and formaldehyde (cfi Fig. 7.9). In less alkaline suspensions smaller amounts of glycolate and glyoxylate are formed under illumination [9, 231. Comparing this product distribution with the product distribution obtained in homogeneous solutions upon oxidation of acetate with hydroxyl radicals or by direct one-electron oxidation, e.g., on a Pt electrode, shows that both oxidizing species contribute to the photocatalytic oxidation of acetate [24]. It has been established in detailed radiation chemical investigations that hydroxyl radicals attack acetate ions mainly at the methyl group according to reaction (7.22) [24a]. CH3COO- + 'OH
+ 'CHzCOO- + HzO
(7.22)
Photocatalysis: Initial Reaction Steps
196
/
x
4
o,oo
100
300
200
400
Illumination time I min
Fig. 7.8. Formation of primary products during the photocatalytic oxidation of 10 mh4 sodium acetate in the presence of 0.5 g/l Ti02 (Degussa P25), in aqueous oxygen saturated suspension (T = 298 K) at pH 3.0, adopted from [23].
025
. I
0,20
.
-
I
-a-o-A-X-
Glyoxylate Glycolate Formate Formaldehyde
A
E
-E C
.-0
0,15
/*
L
8
t 4”
8
-X
I I I I
0,oo
0
50
X
I I
100
150
200
--
1 1 1
250
Illumination time / min
Fig. 7.9. Formation of primary products during the photocatalytic oxidation of 10 mM sodium acetate in the presence of 0.5 g/l TiOz (Degussa P25), in aqueous oxygen saturated suspension (T = 298 K) at pH 10.6, adopted from [23].
Chapter 7; D. W. Bahnemann, et al.
197
In the presence of air the radicals thus formed react quickly with molecular oxygen leading to the products given in reaction (7.23) [24b]. 'CH2COO-
+ 0 2 + 'OzCH2C00- +++
(OCH2COO-)2, CHOCOO-, CH20HCOO-, CH2O
(7.23)
Direct oxidation of acetate results in the well-known Kolbe decarboxylation with the formation of methyl radicals (reaction (7.24)) [24c]. CH3COO-
+ h' + CH3COO' + CH3' + C02
(7.24)
A considerably different product distribution results when these methyl radicals react with oxygen (reaction (7.25) [24c].
CH3' + O2 + CH300'
+++
CH300H, CH300CH3, CH20, CH30H, HCOO-
(7.25)
Figure 7.10 summarizes both described pathways as the proposed reaction mechanism.
+h', -CO,
CH,CO,+OH'
1-H,O
YO, (b)
+02
CH;+
CyO;
+++
CyOOH CyOOCH, C W CH,OH
*CH,CO,-
1+o,
HC0,-
'O,CH,CO,-
1 1 J,
(a)
(a) - ~ 4 b 1 (b) - ~ 4 ~
1
H,O,, -O,CCyOOCyCO,-, CO, CHOC0,-, CyOHCO,-, CH,O Fig. 7.10. Proposed reaction mechanism for the oxidation of acetate by h+"B or 'OHs, respectively (adopted from [23]).
The formation of glycolate and glyoxylate during its photocatalytic oxidation has been taken as evidence for the photocatalytic oxidation of acetate via hydroxyl radicals. The relative importance of this reaction path seems to be higher with increasing pH.
198
Photocatalysis: Initial Reaction Steps
In alkaline suspensions the surface of the T i 0 2 particles is negatively charged (pHZpc = 6.0 - 6.4) and the resulting electrostatic repulsion should hinder the adsorption of the negatively charged carboxyl group of the acetate anion thus favoring an attack of surface bound hydroxyl radicals onto the methyl group. On the other hand, negatively charged carboxyl groups are directed towards positively charged surface groups of the semiconductor particles at pH values below the pHzX and an attack leading to the subsequent decarboxylation of the acetate molecule is favored. It should be noted that the formation of formate does not unambiguously indicate that the oxidation of acetate occurs also via a direct electron transfer from the carboxylate group. Formate itself is the main oxidation product of glycolate and glyoxylate and thus a secondary reaction product of the photocatalytic oxidation of acetate. Furthermore, it is evident that in acidic suspensions of T i 0 2 only formaldehyde and formate are formed during the photocatalytic oxidation of acetate. Here a different mechanism appears to be operative, probably a direct oxidation of the acetate molecule via holes. It can be concluded that the formation of glycolate and glyoxylate during the photocatalytic oxidation of acetate strongly suggests that hydroxyl radicals are formed on TiOz surfaces upon band-gap illumination [9, 231. An additional support of hydroxyl radicals as reactive oxidants is the observation that the intermediates detected during the photocatalytic degradation of aromatic compounds in the presence of titanium dioxide are typically hydroxylated structures [25]. These intermediates are consistent with those found when similar aromatics are reacted with a known source of hydroxyl radicals. In addition, EPR studies have verified the existence of hydroxyl radicals in aqueous solutions of irradiated T i 0 2 [14b, 1 4 ~ 1Mao . et al. have found that the rate of the oxidation of chlorinated ethanes correlates with the C-H bond strengths of the ethanes under investigation which indicates that the abstraction of hydrogen by a hydroxyl radical is an important factor in the rate-determining step of the photocatalytic oxidation of this class of organics [26]. On the other hand, these authors have observed that trichloroacetic acid and oxalic acid (compounds which have no hydrogen atom available for abstraction by a hydroxyl radical) are oxidized primarily by valence-band holes via a photo-Kolbe reaction [26]. Kinetic isotope work by Cunningham and Srijaranai [27] and Robertson et al. [28] also provides evidence for hydroxyl radical attack. Cunningham and Srijaranai [27] observed a primary kinetic isotope effect of 3.3 for the destruction of isopropanol using Ti02. A similar effect of 3 was reported by Robertson [28] for the photocatalytic destruction of the cyanotoxin, microcystin-LR. The results of both studies suggest that the formation of the hydroxyl species may be a rate limiting process in the photocatalytic process. It was proposed that the reduced rate of photocatalytic decomposition in DzO was due to the lower quantum efficiency for the formation of 'OD radicals on the TiOz surface [27]. This would therefore result in a relatively lower surface concentration of 'OD radicals on the TiOl surface for subsequent attack on the target molecules. The lower rate of oxidation may, however, be due to the 'OD radical having a lower oxidation potential compared to the 'OH radical and therefore having a reduced oxidising
Chapter 7; D. W. Bahnemann, et al.
199
power. Whatever the reason for the influence of the kinetic isotope effect on the photocatalytic process, Cunningham proposed that such effects strengthened the supposition that the photogeneration of hydroxyl radicals was the rate determining process for the photocatalytic process. It is interesting that the magnitude of kinetic isotope effects observed by Cunningham and Robertson were so similar. Robertson [28] proposed that an additional possibility was that the destruction of the substrates may be mediated by hydroxyl radicals generated via the superoxide radical anion produced at the conduction band. This is subsequently hydrated or deuterated by the solvent. This may be rate determining since the O2 has to be generated at the conduction band prior to interaction with the solvent and subsequent formation of O H or OD' species. Therefore the kinetic isotope effect could be due to the interaction of the solvent with the superoxide species rather than the attack on the toxin. If this is the case it was suggested that a similar kinetic isotope effect would be observed no matter what substrate was being destroyed. Further kinetic isotope studies will help elucidate the potential of this proposed mechanism. Interestingly other workers have also suggested the possibility that species (02-, HOz' and H202)generated following conductance band electron transfer to oxygen were involved in photocatalytic oxidation processes [29, 301. Linsebigler and Yates used 1 8 0 2 to establish the involvement of such species in the destruction of chloromethane on Ti02 [31]. Richard found evidence that both holes and hydroxy radicals are involved in the photocatalytic oxidation of 4-hydroxybenzyl alcohol [32]. His results suggest holes and hydroxyl radicals have different regioselectivities in the photocatalytic transformation of this compound: hydroquinone is thought to result from the direct oxidation by a valence-band hole, dihydroxybenzyl alcohol from the reaction with a hydroxyl radical, while 4-hydroxybenzaldehyde is produced by both pathways. In the presence of a hydroxyl radical quencher, the formation of dihydroxybenzyl alcohol is completely inhibited while the formation of 4hydroxybenzaldehyde is inhibited. The strongest evidence for direct hole oxidation as the principal step in the photooxidation step comes from a recent study performed by Draper and Fox that failed to detect any of the expected intermediate hydroxyl radical adducts following diffuse reflectance flash photolysis of several titanium dioxidehbstrate combinations [33]. In each case where the product of hydroxyl radical-mediated oxidation was known to be different from that of direct electron transfer oxidation, the authors observed only the products of the direct electron-transfer oxidation. 7.5. Conclusions
The primary events occurring within a nanometer-sized semiconductor particle after the absorption of a photon the energy of which is exceeding the bandgap energy have been discussed in detail based upon a review of the current literature. Both, electrons and holes, are separated extremely rapidly from the initially formed exciton and trapped at or very close to
200
Photocatalysis: Initial Reaction Steps
the surface of the particle. While there is general agreement that the electrons are trapped at surficial titanium sites generating Ti(II1) species, the chemical nature of the trapped hole has not yet been fully understood. The most likely model suggests at least two energetically different trap sites: shallowly trapped holes possess a very positive one-electron redox potential and can be regarded as surface-bound hydroxyl radicals while deeply trapped holes are much weaker oxidants and exhibit a very long lifetime. In the presence of the appropriate redox couples both trapped charge carriers are subsequently transferred to the surrounding solute giving rise to the processes typically known as photocatalysis. Except for some special cases the most likely reaction of the electron appears to be its transfer to molecular oxygen initially generating superoxide radicals. Two distinctly different mechanisms explain the reactivity of the trapped holes: While many of the observed reactions can best be explained by a direct hole transfer to the solute (e.g., pollutant) molecule, there is clear evidence for the intermediacy of hydroxyl radicals in other reactions. It is important to note that hydroxyl radicals can also be formed as part of the reductive pathway following the transfer of two more electrons to the initially formed superoxide radical. As has been shown by isotopic labelling studies both pathways are apparently operative in parallel for the formation of hydroxyl radicals in photocatalytic systems. Acknowledgement This work has been funded by the European Commission under the Energy, Environment and Sustainable Development programme, contract No. EVKl -CT-2000-00077. REFERENCES 1.
2.
3.
4.
Carey J. H., Lawrence J. and Tosine H. M. Bull. Environ. Contam. Toxicol., 16, 697-701 (1976). a) Bahnemann D., Cunningham J., Fox M. A., Pelizzetti E., Pichat P. and Serpone N., in: Aquatic and Surface Photochemistry, G. R. Helz, R. G. Zepp and D. G. Crosby (Eds.), p. 261-316, Lewis Publ., Boca Raton, London, Tokyo (1994); b) Hoffmann M. R., Martin S.T., Choi W. and Bahnemann D. W. Chem. Rev., 95, No. 1,69-96 (1995); c) Bahnemann D., in: The Handbook of Environmental Chemistry, Vol. 2, Environmental Photochemistry, P. Boule (Ed.), p. 285-351, Springer, Berlin (1999). a) Dillert R., Vollmer S., Gross E., Schober M., Bahnemann D., et al., Z. Phys. Chem., 213, No. 1, 141-147 (1999); b) Dillert R., Cassano A. E., Goslich R. and Bahnemann D., Catalysis Today, 54, No. 2, 267-282 (1999); c) Alfano 0. M., Bahnemann D., Cassano A. E., Dillert R. and Goslich R. Catalysis Today, 58, NO. 1, 199-230 (2000). a) Duonghong D., Ramsden J. and Gratzel M. J. Am. Chem. Soc., 104, No. 11,2977-2985 (1982); b) Dimitijewicz N. M., Savic D., Micic 0. I. and Nozik A. J. J. Phys. Chem., 88, No. 23,42784283 (1984); c) Bahnemann D., Henglein A., Lilie J. and Spanhel L. J. Phys. Chem., 88, No. 4, 708-71 1 (1984); d) Bahnemann D., Henglein A. and Spanhel L. Faraday Discuss. Chem. Soc., 78, No. 1, 151-163 (1984); e) Rothenberger G., Moser J., Gratzel M., Serpone N. and Sharma D.K. J. Am. Chem. Soc., 107, No. 26, 8054-8059 (1985); f ) Grabner G., Li G., Quint R.M. and Getoff N. J. Chem. Soc., Faraday Trans., 87, No. 8 , 1097-1101 (1991).
Chapter 7; D. W. Bahnemann, et al.
20 1
5.
Serpone N., Lawless D., Khairutdinov R. and Pelizzetti E. J. Phys. Chem., 99, No. 45, 16655-
6.
a) Colombo D. P. Jr., Roussel K. A., Saeh J., Skinner D. E., Cavaleri J.J. and Bowman R.M. Chem. Phys. Lett., 232, No. 3,207-212 (1995); b) Colombo D. P. Jr. and Bowman R. M., J. Phys. Chem., 99, No. 30, 11752-11756 (1995); c) Colombo D. P. Jr. and Bowman R. M. J. Phys. Chem., 100, No. 47, 18445-18449 (1996). a) Bahnemann D. W., Hilgendorff M. and Memming R. J. Phys. Chem. B, 101, No. 21,4265-4275 (1997); b) Hilgendorff M. Untersuchungen zur Bedeutung der Adsorption in der Photokutalyse, Doctoral thesis, Department of Chemistry, University of Hannover, Hannover, Germany (1996). Grltzel M. and Frank A. J. J. Phys. Chem., 86, No. 16,2964-2967 (1982). Wolff K. Mechanistische Untersuchungen zum Oxidationsprozess an der belichteten TitandioxidWasser- Grenzfluche, Thesis, Department of Chemistry, University of Hannover, Hannover, Germany (1993). Degussa AG, Hochdisperse Metalloxide nach dem Aerosil-Verfahren [Schrifenreihe Pigmente, 4th ed., Frankfurt (1989). a) Howe R. F. and Grltzel M. J. Phys. Chem., 89, No. 21,4495-4499 (1985); b) Howe R. F. and Gratzel M. J. Phys. Chem., 91, No. 14, 3906-3909 (1987); c) Moser J., Punchihewa S., Infelta P. P. and Grltzel M. Langmuir, 7,No. 11, 3012-3018 (1991); d) Micic 0. I., Zhang Y., Cromack K. R., Trifunac A. D. and Thumauer M. C. J. Phys. Chem., 97, No. 26,7277-7283 (1993). a) Schindler K.-M. and Kunst M., J. Phys. Chem., 94, No. 21, 8222-8226 (1990); b) W m a n J. M., de Haas M. P., Pichat P. and Serpone N. ibid., 95, No. 22,8858-8861 (1991). a) Martin S. T., Herrmann H., Choi W. and Hoffmann M. R. Trans. Faraday SOC.,90,3315-3323 (1994); b) Martin S. T., Henmann H. and Hoffmann M. R. Trans. Faraday SOC.,90, No. 21,
16661 (1995).
7.
8. 9.
10. 11.
12. 13.
3323-3330 (1994). 14. a) Gonzalez-Elipe A. R., Munuera G. and Soria J. J. Chem. SOC.,Faraday Trans. I , 75, 748-755 (1979); b) Jaeger C. D. and Bard A. J. J. Phys. Chem., 83, No. 19, 3146-3152 (1979); c) Anpo M., Shima T. and Kubokawa Y. Chem. Lett., 1799-1802 (1985); d) Kasinski J. J., Gomez-Jahn L. A., Faran K. J., Gracewski S. M. and Miller R. J. D. J. Phys. Chem., 90, 1253-1259 (1989). 15. Lawless D., Serpone N. and Meisel D. J. Phys. Chem., 95, No. 13,5166-5170 (1991); Serpone N.,
16.
17. 18. 19.
Lawless D., Terzian R. and Meisel D., in: Electrochemistry in Colloids and Dispersions, R. A. Mackay and J. Texter (Eds.), VCH, New York, pp. 399-412 (1992). a) Bickley R. I. and Jayanty R. K. M. Faraday Discuss. Chem. SOC.,58, 194-203 (1974); b) Tafalla D. and Salvador P. Ber. Bunsenges. Phys. Chem., 91, No. 4,475-479 (1987); c) Brezovh V., Stasko A. and LapcikL., Jr., J. Photochem. Photobiol. A: Chem., 59, No. 1, 115-121 (1991); d) Peterson M. W., Tumer J. A. and Nozik A. J. J. Phys. Chem., 95, No. 1, 221-225 (1991). Buxton G., Greenstock C., Helman W. and Ross A. J. Phys. Chem. Re& Data, 17, 513 (1988). Richard C. New J. Chem., 18,443-451 (1994). a) Choi W. and Hoffmann M. R. Environ. Sci. Technol., 29, No. 6, 1646-1654 (1995); b) Hilgendorff M., Hilgendorff M. and Bahnemann D. W., in: Environmental Aspects of Electrochemistry and Photoelectrochemistr, M. Tomkiewicz, R. Haynes, H. Yoneyama, Y. Hori (Eds.), The Electrochem. SOC.,Pennington, Vol. 93-18, pp. 112-121 (1993); c) the same authors, J. Adv. Oxid. Technol., 1, No. 1, 35-43 (1996).
202
Photocatalysis: Initial Reaction Sfeps
a) Peyton G. R., Bell 0. J., Girin E. and Lefaivre M. H. Environ. Sci. Technol., 29, No. 6, 17101712 (1995); b) Nahen M., Bahnemann D., Dillert R. and Fels G. J. Photochem. Photobiol. A: Chem., 110, No. 1, 191-199 (1997). 21. a) Ward M. D. and Bard A. J. J. Phys. Chem., 86, No. 17,3599-3602 (1982); b) Sclafani A., Palmisano L. and Davi E. J. Photochem. Photobiol. A: Chem., 56, No. 1, 113-123 (1991); c) Prairie M. R., Evans L. R., Stange B. M. and Martinez S. L. Environ. Sci. Technol., 27, NO. 9,
20.
1776-1782 (1993). 22. Bahnemann D. W. Isr. J. Chem., 33, No. 1, 115-136 (1993). 23. Wolff K., Bockelmann D. and Bahnemann D. W. Proc. IS&T Symp. on Electronic and Ionic Properties of Silver Halides [44th IS&TAnnual Con&, St.Pau1, Minnesota, May 12-17, 19911, B. Levy (Ed.), pp. 259-267, IS&T, Springfiled, USA (1991). 24. a) Neta P., Simic M. and Hayon E. J. Phys. Chem., 73, No. 24,4207-4213 (1969); b) Schuchmann M. N., Zegota H. and von Sonntag C. Z. Natudorsch. B , 40, No. 1,215-221 (1985); c) Schuchmann H.-P. and von Sonntag C. ibid., ,39, No. 2,217-221 (1984). 25. a) Augugliaro V., Palmisano L., Sclafani A., Minero C. and Pelizzetti E. Toxicol. Environ. Chem., 16, No. 1, 89-109 (1988); b) Turchi C. S. and Ollis D. F. J. Catal., 122, No. 1, 178-192 (1990); c) Terzian R., Serpone N., Draper R. B., Fox M. A. and Pelizzetti E. Langmuir, 7,No. 11, 3081-3089 (1991); d) Mills G. and Hoffmann M. R. Environ. Sci. Technol., 27, No. 8, 1681-1689 (1993); e) Theurich J., Lindner M. and Bahnemann D. W. Langmuir, 12, No. 26,6368-6376 (1996); f , Theurich J., Bahnemann D. W., Vogel R., Ehamed F. E., Alhakimi G. and Rajab I. Res. Chem. Intermed., 23, No. 3,247-274 (1997); g) Theurich J. Kinetische und Mechanis-tische
26. 27. 28. 29.
Untersuchungen zum photochernischenAbbau organischer Schadstoffe in wassriger Phase, Doctoral Thesis, Dept. of Chemistry, University of Hannover, Germany, (1999). Mao Y., Schoneich C. and Asmus K. D. J. Phys. Chem., 95, No. 24, 10080-10089 (1991). Cunningham J. and Srijaranai S. J. Photochem., Photobio. A: Chem., 43, No. 2, 329-335 (1988). Robertson P. K. J., Lawton L. A., Benjamin J. P., Cornish A. and Jaspars M. J. Photochern, Phofobiol,A, Chem., 116, NO.1, 215-219 (1998). Okamoto K., Yamamoto Y., Tanaka H., Tanaka M. and Itaya A. Bull. Chem. SOC.Japan, 58, No. 6,2015-2022 (1985).
Anpo M., Chiba K., Tominari M., Coluccia S., Che M. and Fox M. A. Bull. Chem. SOC.Japan, 64, NO. 2, 543-551 (1991). 3 1. Wang C. M., Gerischer H. and Heller A. J. Am. Chem. Soc., 114, No. 13, 5230-5234 (1992). 32. Richard C. J. Photochem. Photobiol. A: Chem., 72, No. 1, 179-182 (1993). 33. Draper R. B. and Fox M. A. Langmuir, 6, No. 6, 1396-1401 (1990). 30.
Chemical Physics of Nanostructured Semiconductors, pp. 203-263 A.I. Kokorin and D.W. Bahnemann (Eds.) 0 VSP 2003.
Dedicated to the memory of m y teacher and friend
Kirill I. Zumaraev
CHAPTER 8
Electron Spin Resonance of Nanostructured Oxide Semiconductors Alexander I. Kokorin N.Semenov Institute of Chemical Physics RAS, Moscow, Russia
Keywords: Nanoparticles, structure, EPR, doping metal ions, Titanium dioxide, photoelectrochemistry
List of general symbols Bohr magneton dielectric constant frequency wavelength Planck constant hyperfine splitting (hfs) constant main anisotropic values of A (as well as A,, A,, A,) concentration local concentration amplitude of the EPR line fine coupling (dipolar) constants conduction band edge valence band edge band gap of a semiconductor g-factor of an unpaired electron (g, = 2.0023) main anisotropic values of g (as well as g,, g,, g,) magnetic field (in Gauss) EPR line width between the points of maximum slope EPR line width at half height EPR line width in the absence of dipolar or spin exchange interaction (initial) dipole-dipole broadening of the EPR line nuclear spin exchange integral average radius of a nanoparticle mean distance between paramagnetic centers electron spin specific surface area longitudinal relaxation time of an electron spin
204
T2 T NC PC, SC
ESR of Nanostructured Semiconductors
transverse relaxation time of an electron spin
temperature nanocrystals or nanosized particles polycrystals, single crystals
8.1. Introduction During the last 25 years, a lot of publications concerning catalytic, photocatalytic, photoelectrochemical, photophysical and absorptive properties of the nanostructured semiconductors of different types have been reported. Many books and reviews, for example [ 1-151, presented analytical overviews both on scientific results and practical application of nanosized semiconductor materials, first of all on the titanium dioxide (TiOz). In the previous chapters of this book there were described new interesting data on photoelectrochemical (PEC) and photocatalytic systems based on pure and doped TiOz, chalcogenide materials, CdS, hematite, oxide electrodes modified with tiny metal particles, etc. Progress in all these directions has been attained in many laboratories all over the world dealing with nanocrystalline particles, nanocolloids and nano-structured bulk electrodes. It should be pointed out that if functional properties, regularities and peculiarities, mechanisms of action of these systems, influence of various factors on them are reasonably well studied, to their structural analysis including details of the spatial distribution of active centers, defects, doping atoms, etc., in the semiconductor matrix was not given enough attention in many cases. Indeed, practically any semiconductor material has paramagnetic centers, or they are created during its action. Knowledge of their nature, properties, structure and spatial organization is very important for correct interpretation of the obtained results. Now, after more than 35 years of numerous and in many cases successful applications of the electron paramagnetickpin resonance (EPRESR) to structural, kinetic and physico-chemical studies in material science, catalysis and photocatalysis, EPR became a routine, but nevertheless a very powerful method. Indeed, this technique can be used for studying any paramagnetic centers (PCs), including transition metal ions, free radicals, trapped electrons, etc., in solid, liquid or gaseous media, any diamagnetic matrix, as well as on their surfaces and interfaces. The theory of the EPR spectroscopy is very well developed [ 16-24] and allows to make conclusions about the composition, structure and properties of bulk, dissolved and dispersed compounds. In the diamagnetic matrix with low concentration of PCs, one can characterize all the paramagnetic species, which can also be used as spin probes for getting information about their local surrounding. At rather high content of PCs, one can study their spatial organization quantitatively, measuring mean distances between them or local concentrations in the area of their location (if distribution is not random). In this chapter, we would like to present the most interesting and important (from our point of view!) results obtained by the EPR technique for colloidal and nanostructured oxide semiconductors. Kinetic, photocatalytic, PEC and spectroscopic data will be performed in addition when necessary. Below, for easier understanding of the EPR terms by the readers, we will explain shortly some basic principles of the method.
Chapter 8, A.I. Kokorin
205
The unpaired electron with its spin S = 1/2 in a sample disposed into the resonator of the EPR spectrometer interacts magnetically: a) with the external magnetic field H (Zeeman interaction); b) with the nuclear spin of the "host" atom or metal ion I (hyperfine interaction); c) with other electron spins S existing in the sample (dipole-dipole interaction). In the last case, electrons can be localized either at the same atom or ion (the so called fine interaction), for example in Ni2+, Co2+,Cr3+, high-spin Fe3+, MnZt, etc., or others. These interac-tions are characterized energetically by the appropriate spin-Hamiltonian
*=P,HgS + SAZ + S"'DS'2'
(8.1)
and by the values of the g-factor, of the hyperfine splitting constant A; by the zero-field splitting constant or the dipolar constant D [16, 18, 221. All these interactions are usually anisotropic (because of their vector nature). Thus, g, A and D values must be presented as tensors: 811 and gl, All and AL, DIIand DL in the case of axial symmetry or g,, g,, g,; A,, A,, A,, etc., in the case of three-axis anisotropy. There are simple relations for g and A parameters:
where go and a, are isotropic constants. Also, the third part of the equation (8.1) should be written as: S"'DS'2' = D[S: - %S(S + l)] + E(S:
- S;)
(8.3)
where D and E are the constants of this spin-spin (S"', S"') coupling; S,, S,, and S, are the proections of the spin S to a corresponding axis. If there is partial overlapping of orbitals of unpaired electrons, the Heisenberg spin exchange interaction can be observed [23, 251, and the fourth term JS"'S'2' should be included to the equation (8.1). Here J is the exchange integral of two electron spins S'" and S('). Till now, the best fundamental work concerning peculiarities of the dynamic and static intermolecular spin exchange in liquids and solids is [23]. We'll use their data in parts 8.5. and 8.6. for semi-quantitative estimation of local concentration of paramagnetic centers idon the semiconductor lattice. Additional information about the EPR method will be given in corresponding parts of the chapter.
8.2. EPR Signals of Oxide Semiconductors Starting from old fundamental overviews [26-281, a lot of papers have been published concerning the EPR spectra observed in the oxide semiconductor lattice, first of all for TiOz [29-381, ZrOz [39-431 and In203[44]. After photolysis at room temperature of the degassed aqueous solutions of colloidal Ti02 (anatase, 2R = 10-15 nm) in the presence of poly(viny1 alcohol) (PVA), EPR spectra recorded at 77 K showed existence of several types of paramagnetic centers in the system
206
ESR of Nanostructured Semiconductors
[29]. These signals were attributed as surface (811 = 1.88, gl = 1.93) and interstitial (811 = 1.96, g, = 1.99) Ti3+ ions (see Table 7.1). Other methods of the reduction of Ti(1V) to Ti(III), such as doping with electron donors [45, 46, 351, heat treatment [31], by hydrogen treatment [32] or electrochemically [36], gave similar results and confirmed the proposed identification (Table 8.1). Table 8.1 EPR parameters of Ti3' signals
Sample *
gl
811
Ref.
(Ti3+)sd,A, NC, at pH 2.2; PVA, I-, Ac-
1.925
1.885
29
(Ti3+)sd, NC, at pH 2.2; CH30H
1.930
1.885
29
(Ti3+)sud,NC, at pH 10.6; PVA
1.945
1.880
29
(Ti3+);m, NC
1.988
1.957
29
trapped electrons, NC, A
1.990
1.960
30
(Ti3+)sd,NC, A, Hombicat UV 100
1990
1.957
31
1.992
1.961
31
(Ti37her,NC, A, Hombicat UV 100
1.987, 1.988
33,37
(Ti3+)sd, NC, A **
1.928
33
**
1.924
(Ti3t)hler,NC, A (Ti3t)sd, NC, A
1.988
(Ti3+)lattice,NC, A
**
34 1.961; 1.958
35
1.903
36
(Ti3+)l,ttice, PC, A
1.96
32
(Ti3+)iattice,PC, A
1.990
1.959, 1.960
45,45a
(Ti3')iattice, PC, A
1.992
1.962
46
1.966, 1.965
1.946, 1.947
45a
1.973
1.946
45a
(Ti3+)sud,NC, Degussa P25
(Ti3+)iattice. PC, A, R (Ti3+)iattice, PC, R
**
1.955
47
Ti3+,NC, A
1.947
48
Ti3+,NC, R
1.967
37
Ti3', NC, A
1.97
(Ti3+)iattice, PC, R
1.90
49
* A is anatase, R is rutile, Ac is acetate; ** high concentration Indeed, irradiation of Ti02 particles with light of energy higher than the band gap (A c 390 nm) results in generation of electron-hole pairs:
Ti02 + hv
+ (e- + h')
TiOz
(8.4)
Chapter 8, A.I. Kokorin
207
These electrons and holes can be trapped both at the interior sites and on the surface of colloidal particles. Then, electrons can be located in the conduction band (e-cb) or on Ti4+ ions, at the surface (Ti3+)sdand in the bulk lattice (Ti3+)lattice. It follows from Table 8.1 that surface and lattice Ti3+centers can be distinguished by difference it their EPR parameters. Discussing the published data, authors of [35] concluded that, in the case of anatase, Ti(II1) centers with gl = 1.988, 811 = 1.958 are coordinated with lattice oxygen atoms only, with little tetragonal distortion, and their EPR spectra are not affected by the surface modification. Ti(II1) ions with gl = 1.924, 811 = 1.885 are coordinated with surface OH groups or HzO, having strong tetragonal distortion - (Ti3+)sd centers. Ti(II1) ions, coordinated with surface bound oxygen atoms from the surface modifiers (ascorbic anion), have glA = 1.955, giB= 1.934 and g11A' gllB = 1.885. Moreover, recording the X-band spectra of degassed aqueous TiOz colloids modified with ascorbic acid at 4.2 K showed [35] the splitting of the parallel component, Le. the existence of two paramag-netic species with gil =1.9885, gill = 1.9615 and glz =1.9880, gllz = 1.9581. One of the two types of (Ti3+)inte, centers both for A and R PC TiOz with gl = 1.966, 1.965 and 811 = 1.946, 1.947 (Table 8.1) has been assigned to the Ti3+ions in lattice or interstitial positions, associated with oxygen vacancies [45a]. Formation of Ti3+centers in Ti02 from trapped electrons is usually connected with generation of various radicals from trapped holes, but such reactions and species will be discussed in section 8.3. Another important and well studied paramagnetic ion in the lattice of oxide semiconductors is Zr3+ in ZrOz. Zirconia dioxide is widely used both as a catalyst of different chemical processes, and as a carrier for constructing supported metal-complex catalysts. In the last years, sulfated zirconia attracted significant interest as an active and selective catalyst in skeletal isomerization of normal alkanes at low temperatures, cracking of paraffins, alkylation and acylation of aromatics [42, 53 and Refs therein]. The appropriate experimental data are collected in the following Table 8.2. Table 8.2 EPR parameters of Zr3+signals
Sample *
g1
gti
Ref.
ZrOZ,PC
1.980
1.969
39
ZrOZ,(Zr3c)su,f, NC
1.976
1.957
40
ZrOz,NC
1.977
1.958
41
ZrOz sulfated, ( Z r 3 + ) sNC ~,
1.98
1.95
42
ZrOz sulfated, NC
1.980
1.976
43
ZrOz, (Zr3+)b,k,NC ZrOz, (Zr3+)surf, NC
1.974, 1.979
1.961, 1.962
50, 54
1.978
1.953
51
ZrOz, (Zr3+)bu~, NC ZrOz,NC
1.98 1
1.956
52
1.9755, 1.9720
1.9562
53
208
ESR of Nanostructured Semiconductors
Similar parameters and behavior have been shown for zirconia ions supported on silica Si02 in the reaction of benzene hydrogenation [42b]. These signals were contributed to the bulk Zr3' ions located at axially symmetric sites. The variation of bulk Zr3' and surface related F-center concentration as a function of S (specific surface area) was studied in [50]. The intensity of F-center signal increased and the intensity of Zr3+markedly decreased with the increase of S . At S < 16 m2/g (2R > 24.5 nm) the Zr3+signal increased sharply ([Zr3'] 2 IO" spidg), while the F-center signal practically vanished [50]. Transformation of tetragonal zirconia phase to monoclinic phase has been studied in [53], Calcination of zirconium hydroxide ZrO(OH)2 at various temperatures produced three types of paramagnetic centers assigned to trapped electrons located in oxygen vacancies of Zr02 (g = 2.0018), to adsorbed 0 2 species (see in 8.3.1.) and to Zr3+ions. g values for the latter (Table 8.2) correspond with the expected ones for a 4d' ion in an octahedral environment with strong tetragonal distortion. With the increase of calcination temperature Tcdc, the intensity of Zr3+ signal increased to 980°C [53]. A few works have also been published reporting unusual valence states of metal ions in lattices of such oxide semiconductors as In203, ZnO, Sn02 [44, 55-68]. These compounds attract researchers' attention because they are very perspective materials like thin films and ceramics for constructing new chemical sensors [ S I , as well as highly conductive thermo- and chemically stable n-type conductors (In203). In the EPR spectra of In203 there were observed signals with the following parameters [44]: g = 2.003, AH = 6-8 G (F-centers), and gl = 2.055, 811 = 2.105, Al = 7 G, All = 38 G attributed as In2' which fit to paramagnetic ions with the electron configuration 425s' t;, 4d1'5s1 with 811 > g l > g, [19,21]. Similar g- and A-values have been measured in [ S a ] . Relatively small values of All and Al constants have been connected with the localization of an unpaired electron not on the sole indium ion but on two (an In2+-In3+ couple) or several ones, as it has been experimentally observed in [57] for V4+ centers in V205. Indeed, according to [58], high electric conductivity in nonstoichiometric indium oxide is provided by intensive electron exchange between In', In2+and In3+ions, because it is known that 21n2' t;, In' + In3' with AG = 0 at 300 < T < 800 K [58]. Two EPR signals with the following parameters: gll = 2.058, All = 7 G, gill = 2.107, Alll = 38 G and gL2 = 2.059, A12 = 7 G, gl12 = 2.077, All2 = 72 G were recorded in [56], but unfortunately, they were not reasonably attributed by the authors. Probably, they characterize In2+ions in the substitutional and in the interstitial position in the In203lattice. Pure Sn02 is a dielectric material but after doping with electron donors or partial reduction it becomes a relatively good conductor. The oxidized sample of SnOz heat treated at Tcdc> 720 K provided an anisotropic EPR spectrum with g = 1.89 and AH = 25 G assigned to the Sn3' ions [60]. CO chemisorption at room temperature led to a noticeable increase of the signal intensity. The adsorption of 0 2 immediately after CO sorption was accompanied by fast decay of this spectrum and the formation of the intensive 02- signal. This fact allowed to propose that a signal with g = 1.89 should be ascribed to Sn3' cations located most probably on the surface [60]. Several Sn02 samples were synthesized in [61] by precipitation of SnC14 solutions with NaOH, KOH or NHdOH,
Chapter 8,A.I. Kokorin
209
dried at 370 K and calcinated at 770-1070 K. The authors recorded EPR spectra of four types of oxygen radicals in the system, discussed their features in detail, but unfortunately paid no attention to the signals at g e 2.0, although they have observed them. The EPR spectrum of ZnO depends on the pretreatment of the sample. Besides radical signals with g > 2, a nearly symmetric single line with g = 1.961 f 0.001 and AH = 4.3 k 0.1 G was practically always observed both for PC and NC materials [62-651. This spectrum has been assigned to unionized Zn+ donors [65], interstitial zink [66], F-centers [67] or conduction electrons [68]. We suppose that the singlet with g = 1.961 in such a nonstoichiometric n-type semiconductor as ZnO should be ascribed to rather unusual Zn' centers with electron configuration 3d"4s1. This is in agreement with [62], where UV irradiation (A c 255 nm) of the sample at 77 K produced (e--h+) pairs. Then, some of the holes were trapped by zink ion vacancies, forming 0- ions, and electrons e- reacted with Zn2+ increasing the amplitude of the signal at g = 1.961. Upon warming the sample an electron-hole recombination occurred [62]. Thus, recording and analysis of EPR spectra of lattice metal ions in their paramagnetic state, changes of the spin-Hamiltonian parameters, absolute and relative concentration of the species as a result of influence of external conditions such as heat treatment, light irradiation, chemical reactions, gas evaporation, etc., provide a valuable information about the structure and properties of oxide semiconductor materials. The results of the EPR studies of 0,- and N,O, radicals will be discussed below.
8.3. EPR of Small Molecules Adsorbed on the Semiconductor Surface After the adsorption of inorganic ( 0 2 , 0 3 , NO, NOz, SOZ,CO, C02, etc.) or organic molecules onto the semiconductor surface and especially after further illumination of a sample prepared, different stable or relatively stable radicals are easily recorded by the EPR method. Several important systems in which charge separation created organic radicals were described in detail in Chapter 1 of this book. Some additional information concerning adsorbed pentane, methane, ethylene, benzene, methylbenzenes and m-dinitrobenzene can be found in publications [41, 60, 69-74]. Further, we will shortly discuss some structural features of paramagnetic centers formed under chemical activation or irradiation of the adsorbed oxygen or N,O, molecules. 8.3.1. Oxygen Radicals
EPR studies of transition metal-oxide catalysts have shown that oxygen molecules and atoms on their surface form radicals of several types whose parameters are mainly listed in Table 8.3. Here and in further Tables, for better comparison and representation, we include the appropriate data obtained for some diamagnetic oxides and relative compounds. Usually, EPR signals of the radicals and the paramagnetic metal ions of the lattice are superimposing as it has been observed in [29-33,40-42, etc.]. In many cases for generating radicals hydrogen peroxide HzOz was used, as well as illumination, heat treatment in the presence of O2 and reduction by CO or Hz. It follows from Table 8.3 that:
210
ESR of Nanostructured Semiconductors
Table 8.3 The g-values of oxygen and related radicals Radicals *
g1
g2
g3
Ref.
0- (TiO2, A + hv)
2.020 (g,)
2.009 (g,)
2.002 (gy)
49,75
0- (Ti02,R + Ga)
2.030
2.023
2.007
31
0- (TiOz, R + Al)
2.026
2.019
2.003
31
0- (V center, Ti02 surface)
2.028
2.016
2.004
31
0- (TiOz, A, H20 colloids)
2.0273 (g,)
2.0188 (gy)
2.0073 (g,)
33
0- (TiOz, A)
2.019
2.010
2.004
80
0- (ZnO + hv)
2.022
2.021
2.003 (gz)
61,63
2.043 (gx) 2.024,2.025
2.043 (gy)
2.002 (g,)
40
2.009
2.003,2.002
29,30,49
0- (MgO + H202)
02-(Ti02,A + hv) ** 02-
(TiOz, R + 02)
2.030,2.020
2.008,2.009
2.004,2.003
81,82
02-
(TiO2, A + hv or CO)
2.0234 (g,)
2.0098
2.0035
45a
02-
(Ti02, A + CO)
2.030,2.019
2.009
2.004
85
02-(Zr02+ H202)
2.034 (g,)
2.010 (gy)
2.003 (g,)
40,42
02-(ZrOz + hv)
2.036 (g,)
2.010 (gy)
2.004 (gx)
41
02-(Sn02+ CO)
2.024
2.008
2.003
60
02-@no2+ 02)
2.033,2.029
2.005,2.010
1.986,2.003
79
02-(SnO2 + 02)
2.034,2.024
2.004,2.009
1.994,2.004
84
(SnOz + 02)
2.028
2.009
2.002
80
02-(ZnO + 02)
2.051
2.009
2.002
80,87
2.057 (g,) 2.0773
2.008 (gy) 2.0089
2.003 (g,) 2.0018
40
OH' (Ti02surface)
2.0146
2.0146
2.0032
86
OH' (MgO + H202)
2.050 (gz)
2.0137 (g,)
2.0038 (g,)
40
0-.02 (MgO + hv) T~~+o-.T~~+oH-
2.017
2.010
2.002
76-78
2.018
2.014
2.004
31
~i4+02-~i4+0-.
2.030
2.018
2.004
31
02-
02-
(MgO + HzOd
02-
(MgO)
83
* A is anatase, R is rutile; ** the same values are also in [75, 811 a) it is possible to distinguish various radical centers on the surface and incide a metal oxide, using some additional experimental approaches if necessary; b) principal parameters of different species are sometimes very close to each other, which makes a
Chapter 8, A. I. Kokorin
21 1
problem of their precise identification rather complicated; c) the definition of axis in some works is mixed-up. Recently, all kinds of paramagnetic species formed in ZrOz prepared from zirconium hydroxide by thermal dehydration were investigated by means of EPR technique [54]. Parameters for Zr3' ions are given in Table 8.2, and for 0; radicals it was measured: gl = 2.033, g2 = 2.0075, g3 = 2.003, this is in a good correlation with g, = 2.0334, g, = 2.0082, g, = 2.004 determined in [53], as well as with those in [41, 601 (Table 8.3). The intensity of this signal rapidly decreased and disappeared at calcination temperature Tcdc> 5OOOC [53]. Concerning nanosized particles, it was shown that in Ti02 (unheated anatase, possessing surface OH-groups) powders, photoproduced under UV irradiation at 77 K holes were trapped at the surface forming Ti4+0-'Ti4+OH-radicals [31], while in case of heated samples, holes were trapped at the surface as Ti4+O2Ti4+O-* radicals [31, 331. The same results have been observed if the samples were irradiated at 4.2 K [30]. In some cases, authors used 170and D2O enriched water [29, 30, 331, D20z[40, 831, 13C0 [631, C6D6 [78], Nzl'O [61] or 1 7 0 2 [81, 83, 841 for better understanding of the reactions mechanism. Several forms of the superoxide 02- radical ion formed on the surface of ZnO, MgO, CoO/MgO and Si02 have been reported in [40, 831. The species were differed by the orientation of the 0-0 residue relatively the surface and the metal ion M"'. The correlation between distances and angles in the most probable structures with the experimentally measured g, values was found, and the dynamic behaviour observed in some cases was also discussed [83]. Calculated EPR spectra of the adsorbed 0; for different charges of the metal ion M"' (2 I n I 6) showed that g, values are sensitive to the ionic charge and the increase of n+ causes the decrease of g, [83]. The z-axis of the tensor is usually in the direction of the internuclear axis and the x- direction is that of the mole-cular orbital hosting the unpaired electron. The data in Table 8.3 show that the dependen-ce of g, on n+ is, however, valid quantitatively not always because of rather many factors affecting the g, value (distances to the neighbouring atoms, orientation, local fields, etc.). Additional detailed information can be found in references cited in this section.
8.3.2. N,O, Radicals Studies of nitrogen oxide radicals in various condensed media by means of the EPR technique started about 45 years ago. Initial results were collected in [SS, 281. N,O, radicals are of interest first of all because of their toxicity and a key role in atmospheric chemistry. From this point of view, formation, stability and reactivity of these species adsorbed on the surface of nanosized metal-oxide semiconductor particles, which are photoactive and widely presented in atmosphere, are of essential importance. Principal values of g- and A-tensors for some cases are picked up in the following Table 8.4. Practically all experiments showed a case of three-axis anisotropy in EPR spectra, and the EPR parameters could be easily measured. Free radicals in the atmosphere could be detected by a method of matrix isolation and EPR suggested in [93]. Formation of the NO-: ion-radical has been proved by using 15NO: at the same g-values A, for I5NO?- was equal to 54 G instead of 38 G for l4NO;- [91]. One can also see from Fig. 2 in [91] that
212
ESR of Nanostructured Semiconductors
EPR spectra are not really axial, therefore in precise measurements g, f g,. The EPR parameters published in [loo] and concerning the NO2 radical in Argon matrix at helium temperatures (Table 8.4) are not correct because of the wrong interpretation of the spectrum presented in Fig. 12 [ 1001. Correct determination by the same spectrum gives: g, =2.004, g, =1.992, g, =2.001; A, = 58 G, A, = 46 G, A, = 62 G, which correlate well with the rest of the parameters listed in Table 8.4. Table 8.4 EPR parameters of NOz' radicals in different matrixes at 77 K gx
gY
gz
A,, G
A,, G
A,, G
Refs.
Ti02, A, NC
2.0055
1.9925
2.0023
54.4
49.6
68.4
89
Zr02, NC
2.0051
1.9925
2.0023
52.2
47.7
66.0
90
ZnO *
2.0057
2.0057
2.0026
38
91
Ice
2.0066
1.9920
2.0022
50.6
49.8
70.2
88
Ice
2.003
1.992
2.001
50
47
65
92
COZ
2.0060
1.9915
2.0030
50.8
48.3
62.9
93
NaN02
2.0057
1.9910
2.0015
46.2
43.7
63.4
94
1.995
1.995
2.004
57
57
50
95
2.0037
1.990
2.0037
57.8
47
57.8
100
Matrix
Pb(N03)z
Argon
* the NO-:
ion-radical [91]
Rather often, scientists did not measure the EPR parameters of NO2 radicals in their systems, but observed its very characteristic spectrum, measured concentrations and used these data for discussing structural properties and mechanisms of the chemical reactions occurred (for example, [96-991). Our studies of Ti02 and Zr02 nano-sized particles prepared by a sol-gel. precipitation method [ 1011 (titration with NH40H and further stabilization of the precipitate with HN03) showed interesting difference between titanium and zirconium dioxides [ 89, 901. Fig. 8.1 performs a typical EPR spectrum of the NO2 radical adsorbed on the surface of NC Ti02 thermally treated for 1 h at 200" (parameters are listed in Table 8.4). A similar one with lower intensity has been observed in the case of 2 1 - 0 2 [go]. Calcination of the Zr02 powder at temperatures 200" ITcdcI600" resulted in noticeable changes of the signals amplitude while the EPR spectrum corresponded to the same radical. [NOz'] concentration changes are shown in Fig. 8.2. A bell-shape plot has been observed with negligible amount of NO; at Tcdc< 300" and Tcdc2 600" [go]. In the case of TiOz (anatase), NO?' species were mainly presented in the sample heated at 200" ([NOZ'] l O I 7 spidg) with approximately 15-20% of the NO' centers of the total amount (Fig. 8.1). With increasing TCdc,initial NO2' radicals started to transform in-to NO' ones with -100% content at 500". All the samples were stable at room temperature for months without any changes in air. Therefore, in nanocrystalline Ti02 occurs
-
213
Chapter 8, A.I. Kokorin
thermally induced irreversible transformation NOz' reaction has not been observed.
3000
3200
3400
+ NO'
contrary to ZrOz, where such a
3600
3800
H, G
Fig. 8.1. EPR spectra at 77 K of nano-Ti02 particles calcined at 200" (l), 350" (2) and 500" (3) [89] (details see in the text).
10
L
86420 I
.
I
200
.
I
300
.
I
400
.
I
.
500
I
600
.
l
700
*
I
800
,
I
L
900
T, "C
Fig. 8.2. The effect of calcination temperature on the NO (1) [89], NO1 (2) [90], 02-(3) [49] and 213' ion (4) [54]concentration in TiOl (1,3) and Zr02 (2,4) NC powders.
214
ESR of Nanostructured Semiconductors
If during preparation of nanosized Ti02 particles, NaOH or KOH had been used to neutralize solution acidity instead of NH40H, only NO radicals were contained in the samples without any NO; ones [89]. A corresponding dependence of [NO'] vs. Tcdc is plotted in Fig. 8.2. A similar graph for 0 2 - radicals on the Ti02 surface (based on the data from [49]) is given in Fig. 8.2 for comparison. It is interesting to note that plots for [NOz'] on TiOz and for [NO'] on ZrOz are very close with the maximum at Tcdc 420-430", while it is approximately 100' higher in the case of 0; centers. Comparing spin-Hamiltonian parameters measured for NO centers in different inorganic matrixes (Table 8.5) one can conclude that for all of them A, = 30 f 2.5 G, A, c 10 G and A, < 5 G although g-tensor values are varying in a rather wide range depending on the local crystal fields in the lattice.
-
Table 8.5 EPR parameters of NO' radicals in different matrixes at 77 K gx
gY
gz
A,, G
Refs.
Ti02,A, NC
2.002
1.999
1.9275
32.4
89
Zr02sulfated, NC
1.997
1.997
1.93
31
102
ZrOZ
2.00
2.00
1.92
27.5
103
CeO2
1.993
1.993
1.90
28.4
103
Tho2
1.991
1.991
1.93
27.6
103
ZnO
1.999
1.999
1.94
30
91
ZnS
1.997
1.997
1.91
31
91
MgO
1.995
1.995
1.88
MgO Na-A Zeolite
1.996
1.996
1.89
33
105
1.980
1.987
1.905
30
106
Zn-A Zeolite
1.999
1.999
1.918
30
106
Na-Y Zeolite
1.986
1.978
1.83
29
99
Ba-Y Zeolite
1.999
1.995
1.89
34
99
Zn-Y Zeolite
2.000
1.998
1.93
30
99
Ba-Y Zeolite
1.994
1.89
30
107
Ca-Y Zeolite
1.994
1.92
-30
107
Matrix
104
It has been learnt from [ 1051 that nitric oxide is mostly adsorbed on metal oxides in a dimer form below 110 K. The heat of dissociation of the dimer adsorbate on the MgO surface: (NO)2= 2 N 0 = 3.2 kcal/mol) was measured in [ 1041. The EPR signal of the dimer form in the system { S O ~ - / Z r 0 2+ NO} has been recorded, and parameters g, = g, = 1.993, g, = 1,940, D = 195 G were calculated [102]. The triplet-state NO-NO species were also observed after NO adsorption in Na-A zeolite: g, = g, = 1.976, g, = 1.912, D =
Chapter 8, A.I. Kokorin
215
288 G [106]. These triplet species were formed within the zeolite cavities, and the distance d = 0.46 nm between the unpaired electrons has been estimated from the dipole-dipole interaction. It was observed in synthetic Linde Type Na-Y, Ba-Y and Zn-Y zeolites (a crystalline aluminosilicate with the formula Nax(A1Oz)x(SiO~),,etc.) that after the NO adsorption, only NO radicals could be detected at 77 K [99]. The use of "NO gas confirmed the results with 14N0. After UV irradiation at 77 K of the NO-treated Ba-Y, only NOz radical spectrum was observed. This photoinduced signal was stable at 77 K, decayed gradually at room temperature (7112= several hours) and disappeared completely when the sample was annealed at 50" for 30 min [99]. The role of NO, NO2 and N2O3 species, photoelectron transfer between them and their reversible transformations have anready been discussed. Ab initio B3LYP cluster model calculations have been performed to describe the adsorptive behaviour of NO on MgO solid [ 1081. The most preferable configurations of the NO, NO-: and N z O ~ surface ~complexes were determined. The calculated IR frequencies of these species accounted well for the temperature dependence of the experimental IR spectra.
8.4. Structural Aspects in the Study of Nanocrystalline Materials Principle difference between nanocrystalline and bulk solid materials is based on = SpclVpc, the great distinction of the surface-to-volume ratio. Indeed, SNCNNC>> SSCIVSC which are realized very often in essential changes of adsorptive, electrochemical, catalytic and photocatalytic properties of nanosized and massive particles. In rather many works the specific surface area S and/or the diameter of the particles 2R were determined and compared with the features [38, 50, 109-1211. S-value is usually estimated by means of the BET method at adsorption of small molecules onto the surface (N2 [ 110, 112, 115-1171, Ar [38, 1211, CH30H [49]) from the gas phase. 2R-values are calculated from X-ray diffraction (XRD) 138, 109-114, 120, 1211 or TEM [112, 113, 117-1201 data. The XRD was also used for controlling the phase state (A or R) of the Ti02 material. Existence of noticeable amounts of the brookite phase (7-18% in the range of 70-400°C) was observed in [ 1141. Thermal treatment of titanium dioxide precipitates at temperatures between 200°C and 6OOOC produced powders of agglomerated crystallites as determined by XRD. The average diameter 2R of the crystallites estimated from the half-widths of the diffraction peaks is shown as a function of the calcination temperature Tcdc in Fig.8.3, where we collected the experimental results for Ti02 from several publications. An increase in the Tcalcof the powders leads to an increase in the particle size 2R especially in the case of rutile [38]. Unfortunately, in this basic work M. Anpo et al. did not present Tcdc values at which they prepared their anatase samples for 2R and S measurements. 2R vs. Tcdc dependences reported in [l09, 111, 1131 are very similar but differ strongly from those published in [ 1121 (Fig. 8.3). Probably, this is a result of the applied alkoxide method of Ti02 synthesis (Ti[OCH(CH3)2]4was a starting compound in [ 112]), and not the conventional sol-gel method. Indeed, as it was shown in [ 1111, a crystallite
216
ESR of Nanostructured Semiconductors
size of anatase prepared from Ti(S04)2,TiC14 or Ti[OCH(CH&]4, and 2R dependence vs. Tcalcare markedly different.
150
-
1 2 3
-A-0-
-v100
-
50
-
-x-
4
m
5
0-
Fig. 8.3. Average size of TiOz anatase (1, 3-5) and rutile (2) crystallites as a function of the calcination temperature Tcdcmeasured by different authors: 1 - [109], 2 - [38], 3 [ I l l ] , 4 - [112], 5 - [113].
p 70
4-
1
-x-
2
-0-3
50
A
I
4
....p...5
...
0 I
I
200
400
600
Tu,.*
1
1
800
1000
"c
Fig. 8.4. Average size of Ti02 anatase doped with 1 wt % (1) or 10 wt % (2) of Fe [114]; 3 In203[1091,4 - {54% Ti02 + 46% In203}[109], 6 - In203[120] and 5 - Zr02 [50] crystallites as a function of the calcination temperature Tc*.
Chapter 8, A.I. Kokorin
217
Dependences of 2R on Tcdc for several pure or mixed semiconductor oxides are presented in Fig. 8.4. Iron doped titania photocatalysts with different iron contents at Tcalc below 400°C had iron ions uniformly distributed in the anatase-TiOz phase [ l 141. At Tcalc> 400"C, 1 wt % Fe samples performed the same behaviour of 2R as without iron, and at Tcdc 2 600°C in the samples with 10 wt % Fe content, the formation of hematite phases interacted with the titania phases was observed in XRD experiments. The crystalline structure of Ti02 phases was distorted at high Tcdcwhich also resulted in 5-fold decrease of 2R as compared to 1 wt % Fe case (Fig. 8.4). It was noticed that the preparation method also affects the particle size: 2R increased from 4 to 47 nm and from 7 to 40 nm in the range of 350 e Tcdc e 770 K for particles synthesized by sol-gel or gel method correspondingly [ 1171. Serious difference between 2R vs. Tcdccurves observed for In203 particles in [lo91 and [120] could be also explained by variations in synthetic procedures. The structural properties of nanosized mixtures of various oxide semiconductors: Ti02-In203 [ 1091, Ti02-ZnO [ 1161 or with silica: Ti02-SiOz [ 1211 were studied. One of the plots is given in Fig. 8.4 as an example. It can be seen that there is no any difference between the 1: 1 composite and pure In203up to 500"C, which becomes noticeable at higher temperatures. A binary nanocrystal-line mixture In203-Sn02 (65% of SnOz), prepared as thin films (Tcdc = 600°C) for CO and NO2 sensors [55b], has revealed the binary phase structure consisting of well-crystallized cubic In203 (2R -25 nm) and a highly dispersed phase of SnOz (2R = 5-10 nm). InzO3-NiO thin films (20 wt% of NiO) annealed in the air at 600°C contained the structure of In~-,Ni,03 solid solution with 2R = 20-50 nm calculated from TEM images [ S a , b]. The influence of nanoparticle size on the absorption spectra [ 1191 and diffuse reflectance UV-Vis spectra [ 1131 was studied for Ti02 and ZnO colloids during the particle growth. The specific surface area S of the same titanium dioxide powders for which 2R values were measured in [lo91 (Fig. 8.3) was determined by low-temperature nitrogen adsorption by using the BET method (Fig. 8.5) [ 1101. The surface area ST (in m2/g) can also be calculated from the experimentally measured 2R values using equation (8.5): ST = N T . S= ~ 4.103d 33'2*p*R
(8.5)
Here SI= 4nR2 is the average surface area of one nano-sized particle treated at temperature T, NT = m/p.R3.33'2is the average number of Ti02 particles in a sample with a mass of m = 1 g; (3/4)3'z is the packing factor for small globules in a large box, p = 3.9 g/cm3 is the density of anatase [119, 1221, and 2R is measured in nm. Our results of such calculation [110] are also presented in Fig. 8.5 and correlate to the respective experimental data. Analogous results were recently published in [ 1131. XRD analysis of the xerogels obtained by drying pure titanium dioxide sol at 70°C showed the presence of the nanocrystalline anatase phase [log]. Thermal treatment of this xerogel resulted in the growth of anatase crystallites up to 400°C. The anatase-to-rutile transformation began to occur at 450-500°C. This process was practically completed at 700"C, and only rutile phase existed at Tcdc 2 700OC. This feature of Ti02 xerogels is typical and well known (see, for example, [109]). Thus, it can be concluded that anataserutile transition temperature of nanosized particles is considerably lower than that of the
218
ESR of Nanostructured Semiconductors
corresponding bulk material.
-v-
1 2
-0-
3
-A-
...x...4
0
5
Fig. 8.5. Changes of the surface area [SI of TiOl anatase (1) [110], (4)[116], ( 5 ) [112], (7) [117], rutile (3) [38] and Be0 (2) [ 1151 crystallites as a function of the calcination temperature Tcdc. (6) were calculated by the equation (8.5) from the data of [lo91 (Fig. 8.3).
Serious differences in the 2R vs. Tcdcplots, which one could see in the case of Ti02 (anatase) and Inz03 (Fig. 8.3, 8.4), are also observed in S vs. Tcdcdependences (Fig. 8.5): if S values determined at the same temperatures in [110] and [116] are rather similar, those reported in [112] and [117] differ a lot. This fact can not be explained by technical variations in measurements because all the authors have used Nz as the adsorption gas. Thus, the reason of the changes observed should be found in differences of the particles morphology, i.e. in the details of their preparation. Mixed metal oxides TiOz-ZnO prepared by either homogeneous or heterogeneous co-precipitation performed a fairly high acid strength at about 7 to 57% of ZnO and very high catalytic activity and selectivity for the hydration reaction of ethylene, correlated with changes of S upon Tcdcat different ZnO content [116]. It is interesting that there was no correlation between the catalytic activity of B e 0 in the reaction of isomerization of olefins and the value of surface area S (Fig. 8.5): the maximum activity was exhibited by samples heat-treated at 900-10OO0C, while the S-values were -4 and 15 times lower than at 5OOOC [ 1151. Quantitative experiments have indicated that the F-center EPR signal intensity increased and the Zr3+signal intensity decreased with the increase in S [50]. Fig. 8.6 shows some graphs constructed from the data published in [50]. XRD analysis indicated that all nanopowders exhibited transformation from tetragonal to monoclinic ZrOz phase from 700 to 950°C. The crystalline size (DXKB) calculated from the XRD diffraction peaks, and the particle size (2RTEM)measured by TEM pattern increased with the increase in Tcdc,while the S value changed in reverse [50]. Fig. 8.6 shows also the straight correlation of the F-center signal intensity with the S-value, while Zr3+EPR signal changes in the opposite direction. Calcining of Zr02
Chapter 8, A.I. Kokorin
E I 5u C
90 --
&%/ 1 2
3
4
60--
1
$'
--
/
1
-0-
-A-
m
Q
219
Y
30--
E
vi 0
--
L 400
a , , ,
R
I
I
I
I
600
800
1000
1200
I
Tab,"C
Fig. 8.6. Variation of S (l), DmB (2), 2REM (3), Zr3+(4)and F-center (5) EPR signal intensities as a function of Tcdc(by the results of [50]).
powders at 500°C during 1 to 6 hours resulted in S decrease from 74 to 47 m2/g and correlated qualitatively with the amount of F-centers which decreased 10-fold in the same period of time. The DXmB parameter of these samples remained 9.5 nm and in the tetragonal phase only. This difference in behaviour of D X ~and B S on Tcdc may be caused by the aggregation of crystallites. The F-center EPR signal is related to the S value or particle size 2R rather than the crystallite size DmB [50]. 8.4.1. The Measurement of Local Concentration of Paramagnetic Centers (PCs)
The EPR technique allows to obtain information of three various kinds: a) characterization of the nature of different paramagnetic centers (PCs) and their content in the sample; b) relaxation and dynamic properties of PCs; c) peculiarities of the spatial organization, local concentrations or mean distances between PCs in the system. The latter is usually connected with measurements of the energy of magnetic dipole-dipole interaction between electron spins. It is known from the EPR theory that the magnitude of the dipolar broadening 6H of the EPR spectrum lines is proportional to the concentration of paramagnetic centers C in a solid specimen [ 16,201:
Here AH is the width of an individual EPR line; f i is the width in the absence of dipolar interaction; A is a coefficient, which depends on the shape of an individual line; on the character of the spatial distribution of the PCs in the sample and the longitudinal relaxation time of electron spins T I [123, 1241. Numerical values of A for several practically important cases were calculated theoretically in [20] and were subsequently
220
ESR of Nanostructured Semiconductors
confirmed in many experiments for “long” TI > lo-’ s. In our experiments for nitroxide radicals, VO(H20)52+ and Cu(en)Z(H20)? complexes (en means ethylenediamine) in values equalled 37 k 3, 35 & 1.5 vitreous at 77°K water-glycerol = 1:l solutions, the kXp [ 1231 and 36 f 1.5 G.M-’ [ 1261 correspondingly. These results correspond well to Atheor= 5.8.10-” G . c ~ =- 34.8 ~ G.M-’ calculated in [126] for the case of the EPR line width at half height AHll2 (Gaussian line shape). Equation (8.6) is valid for any case of not too high concentrations C, for which 47cr:C/3 << 1, where ro is the distance of the closest approach of PCS. The fulfilment of the linear dependence (8.6) in a fairly wide range of concentrations C with proximity of experimental values kXp to the theoretical value Atheor is an indication of the uniformity (equal probability) of PCs distribution throughout the specimen. If PCs are distributed only in a certain part of the sample, the experimentally measured value of the parameter &fi = 6WC would be appreciably greater than Atheor [ 1251. = Atheor, At the same time, if it is known independently for PCs of the certain type that kXp it becomes possible to determine the mean local concentration C1, of these PCs in systems in which they occupy only part of the sample’s volume, but are distributed randomly in that part [126, 1271:
Here A b and A should be determined in model experiments. For semi-conductor materials doped with paramagnetic metal ions with T1 2 lo-* s, it is possible to calculate such an important structural parameter as the mean distance (r) between the PCs:
Fig. 8.7 presents our experimental results plotted by the data published in [128, 1291 obtained for polycrystalline Ti02 (rutile) lattice doped with vanadium (IV) ions at different content. The linear dependence of V4+ amount (in spidg) on total vanadium content (in at.%) shows that all vanadium ions in these samples are in (+4) state while the non-linear graph of CI, allows to assume that some part of V4+centers is distributed in the lattice not randomly. Such systems will be discussed in detail in section 8.5. One can see from Fig. 8.7 that CI, and (r) values can be easily estimated by such a simple approximation as Eqs. (8.7) and (8.8). The theoretical model [ 1301 was developed as an extension of the classical theory of dipolar broadening in dilute solid solutions in the absence of exchange interactions [ 161. It was suggested in [ 1301 how to determine the dipolar part in the linewidth by subtraction of the calculated input of Heisenberg exchange interaction of pairs of exchange-coupled ions. Equations were written for the three cubic lattices as a function of ion’s concentration for various numbers of cationic sites included in a sphere of radius Rc with the assumption that clustering effects were absent. The results were compared well with experimental data on Cr3+in MgO powders. Later, this approach was used for the study of influence of vanadium concentration (from 0.01 to 1.0 mol%) on the EPR linewidth of V4+ ions in Ti02 (rutile) host lattice. Our contribution to the peakcalculation of values from the dipolar A&d
Chapter 8, A.I. Kokorin
22 1
, nm
I
0
.
I
1
.
,
2
.
,
.
3
,
4
.
,
5
Metal Content, atom %
Fig. 8.7. Total amount (1) of paramagnetic V b centers in TiOz doped with vanadium; local concentrations (2) of isolated V b ions and mean distances (r) between them (3) vs.
vanadium content. to-peak first derivative linewidth, listed in Table 1 in [131], using Eqs. (8.7) and (8.8) gave the following: 2.03, 2.34 and 2.91 nm for 1.0, 0.4 and 0.2 mol% of [V4'] respectively. This is a reasonably good correlation with the results showed in Fig. 8.7. Principle possibility of measuring hydrodynamic radii of colloidal nanoparticles rr from the reorientation correlation times zR was demonstrated in [132] for three quasispherical heteropoly anions in buffered aqueous solutions at several temperatures. The analysis of EPR line widths of V4+ ions by Kivelson equations allowed to determine TR, from which rr were estimated by the ratio:
where K is an anisotropic interaction parameter, q is the solution viscosity, k is the Boltzmann constant, and T is the absolute temperature [133]. Although the value of K is not known precisely, one can determine r, with good accuracy from the slope of plots of zR vs. q/T since 0.48 < K < 1 for vanadyl acetylacetonate in a number of solvents [ 1331, and 0.8 I ' 2 < 1. The effective hydrodynamic radii of VW501:- (3.72 f 0.07 A), a-PVW1104:(5.08 k 0.06 A), and a2-P2VW170628-(6.41 k 0.08 A) were shown to correlate well with unsolvated crystal radii for the same heteropolyacids. We believe this method could be useful for determining the approximate sizes of unknown nanosized structures. 8.4.2.Regularities and Peculiarities of Spatial Distribution of PCs
We would like to attract attention of scientists using EPR technique in the structural studies to an important problem of correct determination of local concentrations
222
ESR of Nanostructured Semiconductors
of paramagnetic centers (PCs) CI, and mean distances (r) between them in colloidal and nanostructured systems? The problem of the line shape of the magnetic resonance spectrum in condensed matter at random distribution of spins is one of the classical problems for statistical physics of the spin systems [134, 1351. This problem was discussed in detail in [16] where equations for its solution in certain cases (spin pairs, random distribution, cubic crystal lattice, etc.) were given. For a system of magnetic moments coupled by dipolar interaction in a strong Zeeman field, the line shape and the spin-lattice relaxation time are both affected by random fractional occupancy in the lattice. A systematic theory of the shape of paramagnetic resonance spectra in magnetically diluted systems was proposed in [ 1361. A quantitative analysis of the experimental data was made and showed that without taking into account correlations in spatial distribution of spins and nonhomogeneous broadening of the initial spectrum, the error in calculation of CI, could achieve up to 50%. However, in all the papers mentioned above the authors analyzed only threedimensional (3D) systems, while a two-dimensional (2D) case is also experimentally observed: surfaces of various absorbers, heterogeneous catalysts, photocatalysts, etc. In [ 1371, Fel’dman and Lacelle examined the quenched disorder average of nonequilibrium magnetization, i.e., a free induction decay G(t) and its relative fluctuations for dipolar coupled homonuclear spins in dilute substitutionally disordered lattices. The studies of NMR free induction decays and their relative fluctuations revealed that the functional form of the disorder average (G(t)), depends on the space-filling dimentionality D of the lattice. Explicit evaluations of these averages for dilute spin networks with D = 1, 2, 3 were presented in [137]:
where numerical factors are omitted for clarity (they were calculated in the Appendix to [ 1371); p = N N ,the concentration of occupied lattice sites, y is the nuclear magnetogyric ratio, B is Planck’s constant divided by 271. In D = 3, Anderson’s result for a “volume” is valid (see [134]): (G(t))3D = eXp(-471pfbd943)
(8.11)
For D = 2 and D = 1 cases correspondingly, equations should be [ 1371: (8.12) (8.13) where T(x) is the gamma function. The next step in approximation of the theory for EPR measurements was done by Atsarkin and co-workers in [ 1381. Basing on Fel’dman’s results, the researchers studied dipolar broadening and spin exchange narrowing of EPR lines from the paramagnetic centers distributed on the solid surface, and a simple formula for a signal of free precession S(t) was expressed [138]:
223
Chapter 8, A.I. Kokorin
S(t) = So.exp[-(t/T2)D"]
(8.14)
Here T2 is transverse relaxation time of an electron spin. The EPR absorption lineshape is related to free induction decay G(t) through the Fourier transformation [ 161. After averaging over all angles between the surface and the external magnetic field H the following equation was obtained for D = 2 systems [138]:
TF'= 4.64n2D3'2y%
(8.15)
Here n2D is the PCs surface density. This formula was used for estimation of the local surface density of radicals formed on the powdered chars [ 1381.
1600
1 BOO
2000
2200
2400
2600
Fig. 8.8. Fourier transformed derivative dipole frequency spectra for 1-, 2- and
3-dimentionaldistributions of paramagnetic centers *) Our analysis of the Fourier transformed dipole frequency spectra for 1-, 2- and 3dimentional distributions of PCs *) showed that at a given T, value, the dipolar 2D line, and especially 1D line, has a sharper central part and more intense distant wings than the Lorentzian 3D line (Fig. 8.8). Spectra in Fig. 8.8 were normalized by amplitude. According to [136, 1371, the dipolar line width of the EPR spectrum in a general form at a certain Ddimension should be:
where nD is the D-dimensional density of PC's, bo = 3?A2/2, and CD is a certain Ddimensional-dependent coefficient. Therefore, the line width value 1/T2 can be not a linear function of the dimensional concentration nD.
* The author is thankful to Dr. E. N. Degtyarev and Prof. A. A. Dubinsky (ICP RAS, Moscow) for their assistance in spectra calculations and useful discussions
224
ESR of Nanostructured Semiconductors
The 2-dimensional model was used to explain the results of the EPR study of the mechanism of oxygen response in carbon-based sensors [139]. It was suggested that the compounds contained two basic types of PCs. Changes in oxygen concentration lead to a reversible transformation of PCs in chars. At high oxygen concentration, the 2D dipolar interaction between PCs at the surface determined the EPR lineshape. EPR studies of Fullerenes before and after annealing at 850°C exhibited wide wings of the spectra lines [ 1401. The broadening observed was calculated in terms of the 2D interaction. Unfortunately, the author has not come so far across any publication on concerning inorganic semiconductor surfaces (2D) or linear 1D systems. The problem of correct measurement of local densities or distances between PCs in nanostructured low-dimensional systems is even more complicated. Indeed, using modem EPR technique, one can measure llT2 values up to 5-6 nm [124]. But it is the very size of colloidal and aggregated nanoparticles! Is it possible to use the “pure” 2D model in this case, or is it necessary to take into consideration an input of 3D interaction? Our group is working on this problem now, trying to understand where is a border between 3D and 2D cases in terms of quantitative analysis of dipole-dipole interaction.
8.5. Vanadium Ions “Behaviour” idon Oxide Semiconductors During the last 25 years oxide semiconductors such as Ti02, SrTi03, W03, etc. have been widely used as electrodes for photoelectrochemical conversion of solar energy, as photocatalysts for decomposition of toxic pollutants, and for preparation of industrial catalysts. There were numerous investigations of Ti02 photocatalysis since it is applied for the destruction of undesirable chemical contaminants, which appears to be a promising process for water and air pollution control. Complete mineralization of a wide variety of organic compounds to C02, HzO, and inorganic constituents has been reported. Photocatalytic efficiency of Ti02 depends upon the relative degree of branching of the reactive electrodhole pairs into interfacial charge-transfer reactions. In order to enhance them, Ti02 colloids and electrodes were modified by selective metal ion doping of the TiOz matrix [2, 9, 12, and Refs. therein]. Titanium dioxide is the most important among these oxides. However, the large bandgap of Ti02 (2 3.0 eV) makes it unsuitable for solar applications, since only a small fraction of solar energy spectrum included in the UV region is absorbed and efficiently converted into chemical energy. The problem of extending the photoresponse of Ti02 to the visible region can be solved either by decreasing its bandgap, or by increasing its absorption coefficient to the visible light. To approach this problem, the synthesis of mixed oxides of the type Ti02-M,0,, isostructural with Ti02 could be used. A lot of metals were tested as dopants, and in many cases a noticeable increase of the light absorption in visible region, i.e. at h > 400 nm was achieved. Fig. 8.9 reproduces this effect in case of vanadium [ 128, 1291. Changes in properties and their corre-lation with composition and structure of semiconductors will be discussed later, but it should be mentioned that the influence of various metal dopants on semiconductor properties is rather well studied [2, 9, 12, 141 and Refs. therein], whereas the peculiarities of their structure are studied poorly, while
Chapter 8,A.I. Kokorin
225
crystallographic structures of T i 0 2 and the rest of pure (not mixed or doped) oxides are well known [142-1441.
Fig. 8.9. Photocurrent spectra 1, of poly-crystallineTil.,V,02 electrodes at x = 0.005 ( I ) , 0.01 (2), 0.02 (3),0.03(4),0.05 (5). These spectra are normalized by the intensity of the UV-peak [129].
The EPR technique has been widely used for the study of paramagnetic ions in diamagnetic host lattices. The following paragraphs will focus on the results obtained in these studies. 8.5.1. Titanium Dioxide
ESR is known to be a very sensitive tool and can therefore be used in studying structural features of nanosized semiconductor particles doped with paramagnetic metal ions. In many studies vanadium “impurities” inside the T i 0 2 matrix or on the particle’s surface were used as dopants. Moreover, V4+ions are very convenient ESR probes since the ”V nuclei have a large magnetic moment leading to informative hyperfine structures ( S = 1/2; / = 7/2). At low vanadium concentration, the EPR spectrum has well resolved sharp lines (Fig. 8.10) allowing precise measurement of spin-Hamiltonian parameters.
1
3000
.
1
3200
.
I
.
3400
I
3600
.
I
3800
.
,
,
4000
H, G
Fig. 8.10. Experimental ESR spectra of the polycrystalline rutile doped with 0.1 at. % of V“, calcined at 120OOC in He, before (2) and after (1) further treatment at 900°C in air. T = 77 K. (3) is a spectrum of the interstitial Vb ions calculated by subtraction of (1) from (2) [129, 14.51.
226
ESR of Nanostmctured Semiconductors
Table 8.6 EPR parameters of Vh ions in Ti02 lattice (A, values are listed in lo4 cm-’)
Sample
gx
gY
gz
Ax
A,
A,
Ref.
Rutile NC
1.914
1.912
1.956
34
47.3
142.5
147
NC
1.913
1.913
1.956
27.7
40.2
142.5
148
NC
1.914
1.912
1.956
31
44
142
146
NC
1.906
1.899
1.941
27.6
45.2
137.7
118
PC
1.915
1.913
1.958
28.4
40.7
139.0
129
PC, int.
1.986
1.9935
1.9405
45
60
110.1
129
PC
1.913
1.913
1.955
31
44
142
150
PC, int.
1.986
1.993
1.940
45
60
111
150
PC
1.931
1.922
1.972
41.5
42.2
140.1
149
sc sc
1.915
1.913
1.956
31
43
142
151
1.913
1.912
1.955
30.9
44.1
141.5
152
SC, int.
1.9865
1.9930
1.9407
45.4
60.5
111.4
153
Theory
1.917
1.893
1.966
35
40
141
154
1.923
49.6
40
143
118
1.932
47.6
160
147
Anatase NC
1.967
1.940
NC
1.96
NC, int
1.96
1.96
1.932
48
48
158
146
NC
1.991
1.984
1.920
65
78
149
48
SC, int.
1.926
1.982
1.877
44.1
69.1
114.5
155
44.1
62
150.5
156
52.8
31.1
154.2
157
SC, int. Brookite
sc
1.9262
1.8970
1.9573
Int. means interstitial
In the first EPR studies of TiOz colloids doped with V4+ ions [48, 146-149, 1181, an important question about the correct attribution of the observed lines to certain coordination centers arose. First, it was necessary to distinguish between “inner” and “surface” centers, and then, to determine the local environ-ment around V4+ions in each of these main groups. This attribution could be done by comparison of the experimental spectrum parameters with those known for monocrystalline samples of the same oxides. As an example, Fig. 8.10 V4+ centers presents a significant difference between two types of
Chapter 8, A.I. Kokorin
227
in polycrystalline rutile-type Ti02 (spectrum 2) [ 1451. After 2-hour calcining at 120OOC in Helium, both substitutional (spectrum 1) and interstitial (spectrum 3) V4+ ions were formed at a ratio 2:l approximately. After further heating for two hours at 900°C in the air (under such conditions all oxygen vacancies were tempered), only substitutional V4+ centers occurred in the sample. Computer simulations of V4+signals were also very useful for the determination of the EPR parameters, and they confirmed the interpretation of the experimental spectra as it had been done in [145-1491 for vanadium doped Ti02 colloid powders and bulk samples. The EPR parameters of V4+ ions in TiOz lattices (rutile, anatase and brookite) are summarized in Table 8.6 for nano-, polycrystalline and single crystal samples (NC, PC, SC). One can see from Table 8.6 that there is noticeable difference in g- and A-values for equivalent samples reported by different authors. It is difficult to assume that there could be any other locations in the TiOz lattice except substitutional and interstitial positions, therefore, some authors seem to have calculated these values without second-order corrections [19]. It is also clear that theoretical values [154, 1581 correlate reasonably well with the experimental ones. 8.5.1.a. Substitutional and Interstitial Centers
From the crystallographic point of view, inside the lattice of mono- or polycrystals, there are two substitutional and four interstitial positions for the doping atoms in the rutile matrix [ 143, 144, 1591. The substitutional octahedral sites are equivalent to magnetic axes, whuch coincide with the directions [ 1101 and [ lIO]. All the unoccupied interstitial sites have the same arrangement of local surroundings, but the axes of the octahedra are oriented at the angle cp = k12.56' with respect to [110] and [lIO] crystal directions [159]. The corresponding EPR signals were observed for rutile and anatase structures in SC and PC samples (Table 8.6). A special technological procedure, which led to the trans-fer of the V4+ ions from substitutional to interstitial positions, was developed in [ 1531. The strong crystal field approach and the assumption of the complete ionicity of V4+ions gave a good explanation of hyperfine and g-tensors. Temperature-induced diffusion of probe V4+ ions into the matrix of TiOz was investigated by EPR technique [ 1481. A comparison with the spectrum of a reference solid solution containing 1% of vanadium (V/Ti atomic ratio) allowed a quantitative description of the V,Til.,02 solid solution formed. The evolution of x was studied versus the annealing temperature Tcdcand the annealing time at given temperatures. For Tcdc greater than 900 K, the isothermal measurements were characteristic of a bulk cationic diffusion and the activation energy of this mechanism was determined. Drastic changes in the EPR spectrum of V/TiO2 impregnated sample have been observed: an axially distorted octahedral symmetry with four 0 ligands in the equatorial plane, D4h,811 < gl, All > AJ, before, and biaxially distorted octahedron, D2h,g, > g, > g,, A, > A, > A, after annealing in the air at 1120 K. Obviously, the equivalence (randomness) of the spatial distribution of the acceptor impurities (V4+ions) in polycrystalline lattice, or its heterogeneity, should strongly affect electrophysical, photoelectrochemical and catalytic properties of nano-
228
ESR of Nanostructured Semiconductors
structured semiconductors. The knowledge of the matrix spatial organization should allow to give recommendations for improving the technologies of preparation of photoelectrodes and catalysts. Only very few articles on the subject have been published so far, though many papers reported on the complex superimposed EPR spectra (e.g., [129, 149, 160-1631, etc.). Samples of the ceramic polycrystalline TiOz (rutile) doped electrodes of the VxTil. x 0 2 composition were studied at different vanadium content (0.001 e x e 0,05) in [128, 1291. It was shown that at x S 0,003 the EPR spectra perform a well resolved hyperfine structure (hfs) typical of V4+-dopedrutile (Fig. 8.10), in which V4+ions substituted Ti4+ions in the crystal lattice. At 0.003 e x c 0.01, the dipolar broadening of the individual lines 6H occurred. At x 2 0.01, in parallel with continuing broadening of the hfs lines, a broad single line appears (Fig. 8.11). Its part in the spectrum increased with the increase of vanadium content.
,
3000
3200
3400 H,
3600
.
I
3800
,
,
.
4000
G
Fig. 8.11. The EPR spectrum of TiOz doped with 2 at.% of V4+(2) as a superimposition of the multiplet (1) and the anisotropic singlet (3). T = 77 K
This single line (go = 1.935 k 0.005, LW= 140 G) is usual for V4+centers, which are coupled by strong spin-exchange interaction [23, 1641. These V4+ions are localized in some areas with high local concentration (V4'), (the so called "nano-phases" - because of their probable size), which can be estimated from the exchange narrowing of a spectrum by the equation [23]: V,,
= K,, (V"):
= 3.0*10" (AH)-'
(8.17)
Here vs ii Ks are frequency and the rate constant of a static spin exchange in a solid state correspondingly; Ks = 1.8.10' s-'.(g-ion/l)-2 for vanadyl ions V02+in solutions frozen at 77 K, AH is the linewidth in Gauss, and (V4'), is measured in mol/l. It was calculated
229
Chapter 8, A.I. Kokorin
(V4'), = 3.5 M, Le., the mean distance (d)$ between the paramagnetic V4+ions in the areas of high local concentration are close to = 7.7 8, according to (d)s = lo8 ((V4+)s.NA)-", where NAis the Avogadro number [129]. Such (d), value agrees well with the measured range of the exchange interaction between V4+ions in rutile rc = 0.81 nm [131]. But this distance (0.77 nm) exceed more than twofold the average distance d = 3.0 Abetween the neighbouring metal ions in the lattice of TiOz and VO2 single crystals [ 142-1441. Therefore, mixed { TiO2-VOZ}nano-phases with high V4+content are formed in rutile at doping concentrations larger than 1%. Computer analysis of the EPR spectra allowed to obtain additional quantitative data about the studied system [129]. Double integration of EPR spectra showed (Fig. 8.7) that all vanadium ions after heat treatment at 900°C in the air were paramagnetic and distributed either as isolated V4+ions in the T i 0 2 matrix, or in (TiO2-VO2}nano-phases. Local concentrations C,, and mean distances (r) between V4+ions located in the main part of the matrix as isolated centers were determined by using Eqs. (8.7) and (8.8) (see Fig. 8.7, page 259). Relative concentrations of isolated (2) and aggregated (3) V4+ions at different vanadium content in the doped polycrystalline TiOz were estimated in [ 1291 and are given in Fig. 8.12.
3-
a
E C
2.
1 -
0-
0
1
2
3
4
5
Metal Content, atom %
Fig. 8.12. Photocurrent Iph (l), and relative concentrations of isolated (2), and aggregated (3) V4+ ions vs. vanadium content in polycrystalline Til.,V,02 electrodes.
The amount of isolated V4+ions in the rutile lattice remains rather constant, while the amount of aggregated V4+ions in nano-phases is gradually increasing within the range 0.005 5 x 5 0.05. One can observe tenfold drop of the photocurrent with the increase of x (Fig. 8.12), though there is a strong increase in the visible light absorption at x > 0.01 (Fig. 8.9). A reverse behaviour of curves (1) and (3) in Fig. 8.12 allowed to assume that it was the formation of vanadium aggregates (nano-phases), which caused this drastic drop of the lph.
Photo-sensitization of Ti02 electrodes by doping crystals with different transition metals was done in many works, e.g., [141, 165-1721. The results similar to ours (in
230
ESR of Nanostructured Semiconductors
decreasing the short circuit voltage, efficiency and electrical conductivity with increasing chromium content) were obtained for polycrystal-line Til-xCrx02samples [ 1711. Even niobium doping (electron donors) of n-Ti02 electrodes from 0.2 to 20 wt% of N b 2 0 5 resulted in the decrease of photo-current efficiency; the bandgap E, increases with the content of Nb [167]. Fig. 8.13 shows that in this case formation of nano-phases with high local concentration of chromium ions also occurred in parallel with areas containing isolated Cr3+centers (their EPR spectra overlap) [ 1731.
I
I
I
I
I
1000
2000
3000
4000
5000
H, G
Fig. 8.13. EPR spectra at 77 K of polycrystalline TiOz doped with 0.5% Nb5' and 0.5 (l),2.5 (2) and 4.5 at.% (3) C13'ions.
X-ray-phase analysis showed that all Ti-V and Ti-Cr mixtures had the rutile structure and were homogeneous [ 129, 1711. Our results qualitatively correspond to ESCA and optical absorption data on electronic structure of reduced TiOz and V,Ti1.,O2 single crystals [174]. Therefore, it is possible to conclude that metal doped polycrystalline Ti02 systems, being homogeneous enough at the micro-sized level, seems to be rather heterogeneous in their structure at the nano-sized level. A tendency to form metal ion aggregates in the Ti02 matrix with high local concentration C,, of doping ions is at least typical of V4+, Cr3+ and Nb5+ species. This fact has to be taken into account for the explanation of photoelectrochemical and photocatalytic results which were obtained for metal doped Ti02 systems. 8.5.1.b. Surface Doping
Surface doping of oxide colloids and nanostructured electrodes with transition metal ions and complexes is of great interest for improving efficiency and selectivity of photocatalysts and photoelectrodes. Such surface ions as electron donors or acceptors play an important role as catalytic active centers, in charge transfer and in adsorption. There were many publications on this subject and we will try to bring forward the most
Chapter 8, A.I. Kokorin
23 I
interesting ones. Experimental parameters measured by different authors are collected in Table 8.7. Table 8.7 EPR parameters of V4+and V02+complexes on the surface of different supports
€3
gl
All, cm-'
Al, cm-'
Ref.
Ti02,NC, R
1.937
1.968
158.3
49.6
148
Ti02, NC, A
1.922
1.956
163.3
50
118
Ti02,PC, R
1.950
1.983
157.5
51
149
Ti02, NC, A - -
1.93
1.96
149.6
46.7
160
1.89
1.92
151.8
43
160
TiOz,NC, A
1.922
1.991
172.3
67
48
Ti02, PC, R
1.96
2.00
163
70
161
Ti02, PC, A - -
1.907
1.97
161
-60
163
Ti02,PC, R
1.958
1.96
140
45.8
163
Ti02,PC, R
1.950
1.983
157.6
51
186
TiOz,PC, A
1.912
1.981
178.5
69.4
186
Ti02,PC, R
1.906
1.967
192
75
150
Ti02, PC, A
1.905
1.973
186
72; 64
150
Ti02, PC, A
1.943
1.980
166
72.1
187
Ti02,NC, A
1.948
1.973
150
73.7
188
1.910
1.982; 1.999
169.4
37; 37.3
188
Ti02,NC, R
1.922
1.991
172.3
67
189
Ti02
1.922
1.983
167
56
190
Sn02, PC
1.9265
1.9807
180.3
68.4
191
Sample
-''-
164.7
1.938
163
Heterogeneity of the corresponding spin-Hamiltonian parameters in Table 8.7 can be caused by several reasons: a) a large set of possible surface structures; b) ternary surface metal-complex formation with organic or/and inorganic ligands existing in a solution, as it has been observed in many cases, e.g., in [201-2071; c) changes of the pH value [208-2101; d) nonstoichiometry on T i 0 2 and metal-Ti02 interfaces [211]; e) oxygen adsorption and dynamical changes in the crystal field around the V4+ ion on the surface [149]; f , calculations of A- and g-values in the case of V02+ and V4+ions without the second-order correction [ 191, etc. Rather often, some authors involve too much fantasy explaining their own experimental results without any appropriate real data.
232
Next Page
ESR of Nanostructured Semiconductors
Table 8.7 EPR parameters of V4+and VOz+complexes on different supports (continued)
Sample
811
gl
All, cm-'
AL, cm-'
Ref.
GeOz, PC
1.929
1.976
175.5
68.2
184
Zr02, NC
1.923
1.976
166
59
38
Si02, NC
1.931
1.985
172.5
66.2
192
Si02,PC
1.907
1.993
189.3
76.6
193
Si02,NC
1.922
1.982
182
72
194
SiO2, NC
1.934
1.974
174
67.3
162
after TPR
1.930
1.979
175
65.6
162
A1203,NC
1.942
1.974
173
58
162
after TPR
1.947
1.950
163
66.5
162
A1203
1.939
1.983
161
56
190
1.952*)
1.991
158.6
63.2
195
1.946
1.987
160.8
60.3
195
y-A1203
1.940
1.998
175.7
67.2
196
--
1.949
1.998
143.8
58.8
197
A1203,Neobead
1.933
1.978
170
61
198
A1203
1.916
1.989
169
66
194
pc M003, PC
1.910
1.978
178.3
64.6
191
1.908
1.965
166
51.4
185
MgO, NC
1.954
1.965
159
69
162
after TPR
1.953
1.964
159.5
70.6
162
MgO, PC
1.928
1.978
162
62.8
199
MgO V02t(HzO)s
1.936
1.976
160
59
190
1.932
1.975
182
72
147
VO(HzO)P -''-
1.9312
1.9778
185.2
70.6
164
1.934
1.980
181.5
57.4
200
V02+(H20),
1.930
1.984
176
69
194
y-Al203, 1% V 2% v
Nb205,
*)
there was an error in Table 1 in [195].
At the adsorption onto titanium dioxide surface, vanadium ions form, at the beginning, randomly distributed isolated V4+ centers with typical Ddhsymmetry (811 e gl, All > AI for the unpaired electron). At higher vanadium concentrations, monolayers and
Previous Page
Chapter 8, A.I. Kokorin
233
double layers of vanadium pentoxide V205 are formed on different carriers [39, 48, 161163, 186-191, 196, 212, 2131. SIMS, X P S , electron microscopic [214] and ESCA [174] studies have clarified the stoichiometry of interaction between vanadium and titanium in VTi oxide catalysts. An interesting comparison of structures of vanadium oxide catalysts supported on Ti02 (anatase, rutile, mixture of AN and RU) [212] and on A1203 [195] has been done using the rectangular pulse technique, XRD, EPR, in situ IR and UV-Vis spectroscopy. When V2O5 content increased to 5 mol %, the surface of Ti02 was covered, but only partially, by 1-3 layers of V2O5 lamellae. At 10 mol % V2O5 content, about 90% of the catalyst surface was covered with 5-8 layers of V2O5 in the form of the [OlO] face of v205 [212]. The spectroscopic study of the nature of vanadium oxide {VOX} supported on a high surface area TiOz (anatase) indicated the formation of three different {VOX}structures [48]: a) isolated V4' ions, part of which was coordinatively unsaturated, strongly bonded to the surface hydroxy groups of the support; b) bidimensional clusters of {VOX}with mainly V5' after calcinations, reducible under mild conditions to V4' and also to V3' to some extent (these species weakly interact with the support surface); c) Vz05 appeared when cove-rage was about the monolayer and was presented as bulk multiplayer structures. The authors observed the existence of at least two different isolated surface V4' species, which caused splitting of the low-field hfs lines in parallel orientation. Our investigation of the surface doped with V4' ions oxide carriers (nanosized wide-gap semiconductors: Ti02 (Degussa P25, Hombicat-100, nano-Ti02 prepared by ourselves, and Zr02 nanoparticles), of their spatial distribution at different content of V4' ions, of the local concentration (V"),, mean distances ( r ) , d between them, were initiated as an attempt to clarify the structural problems formulated in [48]. The first results concerning hetero-geneity of the surface V4' paramagnetic centers were reported in [215, 2161.
Fig. 8.14. EPR spectra of the Degussa P25 particles after: 10 min (l),6 days (2), 75 days (3) of incubation in 0.15 cm3 of 0.65 M ascorbic acid in CzH50H-Hz0= 3:2 solution. [V4-'], = 1.8~1OZo~ r n - T ~ ;= 77 K [215].
234
ESR of Nanostructured Semiconductors
As an example, Fig. 8.14 presents typical EPR spectra at 77 K of the Degussa P25 particles after different time of incubation in 0.65 M ascorbic acid solution [215]. EPR analysis indicates the presence of at least two distinct V4+ species: the sharp signals, overlapping with a broad single line of the aggregated V4+ centers. Similar spectra were observed in [48, 160, 1631, etc. It follows from Fig. 8.14 that a relative part of such aggregated centers decreases in time, transforming into isolated centers. Analogous changes have also been observed for the Hombicat-100 samples [217]. In contrast, ZrOz particles doped with vanadium did not change their EPR spectra with time.
2800
2900
3000
3100
3200
H, G Fig. 8.15. Low-field lines of the EPR spectra of the Hombicat-100 particles with [V&] content: 1.6.1019~ r n (l), - ~ 4.0.1019cm-3 (2), 1.2~1OZ0 ~ r n (3) - ~ after 15 days of incubation in 0.15 cm3 of 0.75 M ascorbic acid; 4 - 0.01 M V02+in CzH50H-Hz0= 3:2 solution. T = 77 K [215].
Fig. 8.15 shows changes in the low-field part of the V4+ EPR spectrum for the samples prepared using Hombicat-100 powders at various V4+content. One can see from this figure that there exist up to three types of surface V4+ centers (a, b, c) with slightly different spin-Hamiltonian parameters: All = 180 k 2 G, g11* = 1.948 (a); All = 179 f 3 G, g11* = 1.973 (b); All = 195 f 2 G, g11* = 1.954 (c). It should be noted that gll* values were calculated without the second-order correction (A, and gl values could not be evaluated because of the over-lapping of the EPR spectra of (a), (b) and (c) species), hence, g11* parameters are relative. For V02' ions in the same experimental conditions, All = 200 f 2 G, g11* = 1.951. Comparing these data with those known from literature [19, 21, 2071, etc., the probable surface structures of V4+centers have been proposed [217] (see thje next page). The (a) centers are the most stable because they are included into the surface Ti02 matrix binding with at least three or four lattice oxygen atoms; (b) structures should also be relatively stable, but they are attached to the surface by only two lattice 0 atoms; and the (c) complexes have to be rather mobile, as they are anchored to the surface by the one oxide 0 atom. A few positions in the coordination sphere of (b) or (c) species are occupied with water molecules.
Chapter 8, A.I. Kokorin
OH
I
OH
OH
OH
(a)
I
I
I
I
(OH2)z
-TiUV-
I
(b)
0
II
I
- T i U V U T i -
235
I
I
0
I OH
0
I
I1
I
I
-Ti&V-
(c)
-Ti-OH
I
(OH&
0
H2O
0
The following results confirmed this explanation: a) double integration of the first derivative EPR spectra had been done, and showed that the total content of the paramagnetic vanadium species in the samples had no changes during 60 days (Fig. 8.16). Then, the computer analysis of the low-field parallel component of the isolated centers has been done [217]. It showed that for the time of the experiment V4+ions from “aggregates” transformed only to the (c) type complexes, while the amount of (a) and (b) ones remained constant (Fig. 8.17). The (c) complexes are still anchored to the surface, and not dissolved in a liquid phase as it is seen from Fig. 8.15: positions of the EPR low-field lines for the (c) and V02+ centers are slightly different (see page 275). This work is in progress at the moment. After solving technical difficulties in calculations, it will be possible to suggest a methodology for complete quantitative description of “what is happening” on the surface of nanostructured oxide semiconductors.
. CI)
.-P
z
8-
6-
‘0 r
X
E
4 -
o-*x)
2 2-
.
xx-x-x-x-x
3
X
0 1
1
0
*
1
10
.
I
20
.
l
.
30
I
-
40
I
.
50
‘
1
60
t, days
Fig. 8.16. Total amount of V4+ions [V6] on the surface of Ti02: Hombicat-I00 (l), Degussa P25 (2), and Zr02 (3) nanoparticles at different time of incubation in the 0.65 M ascorbic acid
solution.
ESR of Nanostructured Semiconductors
236
E I
m c L
---L
1
S-X
2
C
e
10
20
x
30
X
40
50
60
t, days
Fig. 8.17. Relative amount of V4+centers in Ti02 Hombicat-100: type “a” (l), type “b” (Z),type “c” (3). [V4+ltOtd= 1.1.1O2’spidg.
The first high field (HF) EPR investigations (at 110 GHz and 330 GHz, besides 9.5 GHz) of solid catalytic materials (V4+ supported on TiOz) have been recently published [218, 2191. A quasi-optical HF EPR spectrometer allowed to use standardf ampoules, similar to the X-band ones, and to vary frequency over a wide range without removing a sample from the resonator. For the V4+/Ti02catalyst with high (=20 wt.%) concentration of V4+ions, the X-band EPR spectrum presented a rather narrow single line (g = 1.967, AH = 98 G at 30 K) associated with the strong spin exchange interaction between paramagnetic ions [218]. Measurements at 330 GHz permitted the authors to observe the well-resolved spectra of the system. At least two types of V4+centers in different coordination could be observed in the HF EPR spectra. Unfortunately, the EPR parameters were not calculated by the authors in [218, 2191. The picosecond dynamic effects have been observed in the HF EPR spectra at 20 < T I 120 K, and at 20 K with changes of the field frequency from 330 to 110 GHz [219]. These changes were explained as a process of fast electron echange in a “cluster” containing a number of coupled vanadium ions. 8.5.2. Other Oxides In this paragraph we would like to present Table 8.8 which collects the spinHamiltonian parameters of V4+centers in the lattice of various oxide carries, differed from TiOz. This information can be useful in comparison with Table 8.6 for the relative analysis of the matrix nature influence on EPR characteristics of vanadium-doped systems. One can conclude from the data of Table 8.8 that there is a noticeable influence of the lattice nature on spin-Hamiltonian parameters, although it is not so easy to find any clear correlation. It is probably caused by serious variety of the reported values published for just the same system (compare values for VOz [175, 1761 and VzOs [178, 1791). In the latter case, both g- and A-parameters are simply inverted, while gll-values are shown in a very wide range 1.88 5 811 I 1.923 [177-180, 571. Probably, this is a manifestation of several positions for V4+ions, in which they can exist in the VZOS matrix.
237
Chapter 8, A.I. Kokorin
Pure monocrystalline VzO5 is diamagnetic, and the EPR spectra can be recorded only in the presence of paramagnetic V4+ ions. The All values were determined approximately twice smaller than in the rest of the cases [177, 1791, and the spectrum pattern included 15 equally spaced lines. This was serious evidence that there were certain defect centers in the matrix, in which V4+-V5+or V4+-V3+pairs were located. The results were interpreted in terms of a model by which an unpaired electron interacted with two equivalent nuclei separated by an oxygen vacancy. A self-consistent mechanism has been proposed for the formation of the low-temperature form of non-stoichiometry in VzO5 [179]. Table 8.8 EPR parameters of V4+ions in various oxides (A, values are given in lo4 cm-') A,
A,, All
Ref.
47
44
147
175
1.948
27
45
140
176
1.923
64.6
57.4
165.6
57
1.88
46.2
92.2
177
1977
141
46.9
178
1.983
1.911
30.6
78.5
179
1.986
1.923
62
168
180
V2O5, amor.
1.984
1.926
73
190
180
VzO5, amor.
1.98
1.913
61
157
181
1.943
21.1
41.8
140.1
182
1.942
22.6
43
140.5
183
Sample
gx, g,
gY
sc
1.895
1.930
gz, 811 1.925
v02, PC
1.950
1.950
sc v205, sc v205, sc v205, sc v205, sc
1.978
1.984
v02,
v205,
Sn02, SC
1.98 1.905
1.939
1.981
1.903
SnOz, PC
A,,
Al
48.8
Ge02,PC
1.9213
1.9213
1.9632
36.7
37.54
134.36
184
ZrOz, PC
1.977
1.942
1.889
62
13.6
140
38
M003,PC
1.976
1.974
1.921
51.7
52.5
161.4
185
It was found that on the surface of zirconia-supported vanadia catalysts vanadium was presented in the form of isolated vanadyl species or oligomeric vanadates, or as Vz05 nanocrystals, and that V5+and V4+ions coexisted in octahedral and tetrahedral coordination. Within the bulk of zirconia matrix, V4+ ions were stabilized in a VxZrl-xOzsolid solution
WI. The K-band EPR spectrum of SnOz doped with 0.5% vanadium has shown at 77 K two sets of super-hyperfine structures (shfs, I = % for both "'Sn and "'Sn) with AI? = 168 G of the two tins located along the c axis and with AI? = 28 G of the four tins lying in a diagonal plane of the unit cell containing four 0 atoms [ 1821. The ground electron level was suggested as 3 d , ~ ~ 2This . was confirmed in [ 1831, where super-hfs interaction constants of V4+with neighbouring tins AIp"(1) = 158 G and AlY(2) = 28.1 G were
238
ESR of Nanostructured Semiconductors
measured for PC Sn02 samples doped with V4+(X-band, 77 K). Such values are caused by the difference in interatomic distances: V-Sn(1) of 3.2 8, and V-Sn(2) of 3.7 A. It was concluded that a part of 3d,2+ wave function extends directly toward Sn( l), while only the indirect exchange mechanism is responsible for the interaction with the magnetic nuclei at Sn(2) site [183]. The X-band EPR spectra of vanadium-doped amorphous and PC tetra-gonal GeOz have been observed even at room temperature (Table 8.8), but there were not recorded for PC hexagonal Ge02 neither at 298 K, nor at 77 K [ 1841. 8.6. Other Paramagnetic Dopants
EPR studies of metal-doped Ti02 and other oxide colloids were used for structural and functional characterization of such materials. This information is spread in many original articles, and was partially collected in [21, 220-2221. Various paramagnetic ions such as Mo5', WSt, Cr5', Nb4+,Ta4+,Mn4+,Mn3+,Cr3+,Fe3+,Ce3+,AI3+,Pt3', Ni3+,Ni2', Nit, Co2+,Cu", etc., were used as spin dopants. As in the previous paragraph, Table 8.9 contents the spin-Hamittonian parameters of metal centers in Ti02 (rutile - R, anatase - A, brookite - B), and the same data concerning other wide bandgap semiconductor oxides are collected in Table 8.10. Doping with Fe3+, as with V4', of Ti02 colloids in aqueous dispersions with light irradiation at 77 K and room temperature resulted in the inhibition of hole-electron recombination by these ions [147]. Interstitial Mo6+in Mo-doped powders behaves as an irreversible electron trap on irradiation; substitutional Mo5+ on the other hand was a reversible hole trap. The electronic structure of the nd' ions in the crystal field with DZh symmetry: Mo5', W5+,Nb4+,V4', has been studied in [ 1761. The substitutional doping of 12-nm-sized Ti02 colloidal crystallites with Fe3' ions had a profound effect on the charge carrier recombination time [223]. Doping with 0.5% Fe3' drastically augmented the mean lifetime of the electron-hole pair from 30 f 15 nc (undoped Ti02) to minutes and hours. EPR studies showed that Fe3+ions entered the host lattice on Ti4' sites, charge compensation took place through the formation of oxygen vacancies. Valence-band holes produced under band-gap excitation reacted with these centers in the bulk, forming Fe4+.Electrons from the conduction band were trapped by Ti4+ centers at the particle surface. The spatial separation of trapped electrons and holes, presumably, inhibited their recombination [223]. The following paramagnetic centers were attributed by use of EPR: a) charge compensated Fe3+(kc= 4.295); b) Fe3+without charge compensation by an oxygen vacancy (g = 2.0023); c) Ti3+ ions located at the particle surface (811 = 1.883, gl = 1.927); d) a new Fe3' signal (g = 1.997) appeared after standing at RT . Niobium-doped Ti02 has been used both as a appropriate material for rutile masers [224, 2251, and as a photocatalyst for water cleavage processes [46]. Nb4+,Ta4+and Ce4' substituted the Ti4+ions in the lattice, and Nb4+,Ta4+at helium temperatures had short TI times suitable for maser applications [224].
Chapter 8,A.I. Kokorin
239
Table 8.9 EPR parameters of various metal ions in TiOz lattice (A, values are listed in lo4 cm-') Dopan
Sample
g,
gY
gz
Ax
A,
A,
Ref.
SC, A
1.834
1.759
1.842
32.2
36.8
74.1
226
SC, B
1.8159
1.7874
1.9148
35.3
29.0
76.5
157
SC, R
1.8117
1.7884
1.9125
24.74*
30.5*'
65.85*'
227
1.9167
1
30.5*'
65.1*'
228
24.5*'
31.1*'
66.4*'
t
Mo5+
SC, R
1.8155
1.7923
25.0*'
sc sc
1.4725
1.4431
1.5944
40.8
63.7
92.5
229
1.4731
1.4463
1,5945
40.5
63.9
92.0
230
NC, A
~1.979
-1.979
1.947
SC, R
1.973
1.981
1.948
1.66
7.93
2.32
225
SC, R
1.973
1.981
1.948
1.8
8.0
2.1
224
Ta&
SC, R
1.979
1.979
1.945
<2.5
<2.5
2.7
224
72.7
70.4
72.7
23 1
Mn4+
sc sc*3
1.995
1.9909
1.9898
72.4
70.4
72.75
232
Mn3+
SC*4
2.00
2.00
1.99
52.8
80.6
84.5
233
Ce3+
SC, R
4.394
2.069
3.866
224
Ni3+
SC
2.085
2.254
2.084
234
Ni2+
SC*'
2.10
2.10
2.20
234
Ni'
SC
2.237
2.050
2.272
234
sc sc
2.19
3.75
5.88
40
26
150
235
coz+
2.090
3.725
5.86
39.1
25
143
236
PC, R
2.200
2.188
1.998
20.0
20.5
163.2*6
237
PC, A
2.090
2.091
2.435
17.3
13.1
83.0*6
237
SC, R
2.093
2.105
2.344
29
19
88
234
BaTi03
1.950
2.459
135
0
238
w5+ Nb4+
Cu" Pt3'
1.990
45
*' average value for two isotopes; ** separated values for two isotopes; *3 D = 4.08 cm-', E = 1.307 C u ; A,(') cm-I; *4 D = -3.4 cm-I, E = 0.1 16 cm-I; * 5 D = -83.4 cm-', E = 1.375 cm-'; *6 f ~ r ~ ~for% = 0.0175 and 0.00893 cm-' respectively.
The well resolved EPR structure of V4+and Mn4+in TiOz was interpreted in terms of the shfs interaction of 47Ti and 49Ti ions occupying the nearest sites along the crystalline c
240
ESR of Nanostructured Semiconductors
Table 8.10 EPR parameters of transition metal ions in various oxides (A,, values are listed in lo4 cm-')
Ge02, am
2.056
2.318
21
Ge02,tetr
2.070
2.049
2.380
17
Ge02, hex
2.096
2.047
2.382
21.6
ZnO ( T d )
1S237
0.7383
224
2.19
MgO
164
239
43
128
239
56.4
83.2
239
198
240
19
24 1
Be0 (C3")
2.379
1.709
108
50
242
A1203 *
2.0772
2.0784
60.2
64.6
243
186
244
Nbw SnOz, PC
1.909
ZrSiOa, SC
SiOz,
1.779
1.910
58.8
81.8
1.908
1.862
123
269
245
1.92
1.89
130
274
246
* Cu3+,D = -0.1884 cm-'. EPR parameters of Mn2+ions in BaTi03 matrix for cubic (g = 2.0009, IAl = 85.0 f 0.5 G, ID1 = 15 f 5 G), tetragonal (g = 2.0023, [AI = 84.1 k 0.5 G, ID1 = 70 k 5 G) and rhombohedral (g = 2.0016, IAl = 85.4 k 0.5 G, ID1 = 60 k 5 G) phases were determined in [248]. Temperature changes of ID1 and g-values were observed for the Ti02 rutile crystal doped with Gd3+:g = 1.9930 (295 K), 1.9941 (77 K), 1.9986 (1.8 K) [249]. Amorelli et al. have studied Ti02 polycrystalline powders of rutile, anatase [250] and brookite [251] treated with transition metal ions using the EPR technique. At Tcdc below 5OO0C, both rutile and anatase forms exhibited EPR signals at g = 1.97 and g = 4.3 (ID1 values were not estimated) ascribed to surface Cr5+and Fe3+ species, respectively. At higher temperature migration of these surface bound ions into the bulk of the rutile powder was observed. In the case of both Cr and Fe substitutionally incorporated within the rutile lattice, photochromic activity was observed. Diffusion of ions into the bulk matrix upon calcination was also demonstrated by the corresponding anatase powder. The process was more difficult for the anatase case [250]. The brookite was doped with Cr3+,Fe3+and Mn2+ ions. For samples heated above 650°C, significant changes in the EPR spectra were observed [251], and they were attributed to the structural transformation of the brookite crystal host first to the anatase, and then to the rutile form of TiO2. For Mn4+(S = 312) in anatase, ID1 = 446 G [25 11 was measured. In nanocrystalline anatase colloids doped with vanadium, molybdenum or iron EPR signals of several types were observed [147]: a) Fe3+(charge compensated) g = 4.27; b) Fe3+ (charge uncompensated) g = 1.99; c) Mo5+ (sub-stitutional) 811 = 1.912, g l = 1.81; d) axis (Ti02-V4+:A," = A,T' = 2.0 G, A," = 2.4 G; Ti02-Mn4+:AxM". = AYMn c 0.5 G, AzMn= 1.0 G) [247]. The absence of such structure in the cases of Cr3+and Mn3+was discussed.
Chapter 8, A.I. Kokorin
24 1
Mo5+(interstitial) 811 = 1.941, g l = 1.89. Transformation of tetragonal (T) ZrO2 phase to monoclinic (M) phase in the presence of Fe3+ions as spin probes gave the following results: Fe3+(T):gl = 10.2, g2 = 5.4, g3 = 4.15, JDJ -0.3 cm-', JEDJ< 0.01; Fe3*(M): gl = 4.27, g2 = 9.2, g3 = 4.7, g4 = 3.9, JDJ -1.7 cm-', (ED1= 0.315; Fe3'(AhI): g = 4.27, ID1 >> 0.3 cm-', IE/DI = 0.333 (AM means the amorphous phase) [53]. Chromium was a metal widely used as a paramagnetic agent for doping Ti02 and other oxide lattices [220,22 1,250-2661. The electronic configurations of chromium centers are 3d3 (Cr3+, S = 312) and 3d' (Cr5+,S = 1/2). Spin-Hamiltonianparameters are Table 8.11 EPR parameters of chromium ions in various oxides (A - anatase, B - brookite, R -rutile, X - xerogel) Ion
cr3+
CrS'
cm-'
IEJ,cm-'
Ref.
0.14
260,261
G
Ti02,R
1.97
0.68
TiO2, A
1.972
0.0372
262
Ti02,A
1.973
0.056
21
TiOz,B
1.97
0.12
TiO2, X
1.98
Ti02,A Ti02,R
1.98, 1.973
0.0374
TiOz,R TiOz,R
1.98
0.6791
0.1433
252
MgTi03
1.98
0.553
0.272
253
Sn02
1.97
0.55
0.27
254
A1203
1.976, 1.981
0.502
A1203
1.975
0.577
Ga203
1.982
0.191
259
scZo3
1.977
0.0326
21
y203
1.96
0.544
1.97, 1.984 1.97
1.172
21
1.212
22 1
1,9782
0.0798
21
A1203
1.90
7.0
Zr02, surf.
1.97
Ti02,A
A, = 0.00097 A, = 0.00143
226
-1'-
g, = 1.949 g, = 1.968
-'I-
g, = 1.935
A, = 0.00387
- _
MgO
Cr4+
IDI, A,,
Matrix
0.03
25 1 252 252,255
263 0.278
0.133
< 0.05
258
22 1
21,220 257 -1'-
242
ESR of Nanostructured Semiconductors
listed in Table 8.11. More information concerning the EPR spectroscopic features of chromium ions in inorganic lattices one can find in the references cited in the Table. The X-band EPR and diffuse reflectance spectroscopy results provided evidence for the presence of V4+ions in the tetragonal zirconia phase [267].Three types of signals were observed at 10 K in Zrl-,V,02 at 0.02 I x 5 0.1: a broad single line at g = 1.95 of the centers with strong dipolar and spin-exchange coupling, whose intensity grew up with increasing vanadium content x, and two rather sharp signals with resolved hfs with a) 811 = 1.927,g l = 1.942;All = 151 G,AI = 76 G,and b) gl = 1.87,g2 = 1.98,g3 = 1.91;Al = 165 G,A2 = 60 G, A3 = 95 G. the latter signal corresponds well with V4+ substituting for Zr4+ in the monoclinic zirconia lattice, occupying positions with rhombic (C,) local geometry [268]. Co doped Ti02 thin films were prepared by reactive co-sputtering deposition [269]. Magnetization studies showed hysteretic behavior with the coercive field 55-65 Oe and the saturation magnetization at RT ranging from 7 emu/cm3 (2.2% Co) to 28 emu/cm3 (8.5% Co). EPR measurements at X-band revealed an anomaly in the temperature behavior of the absorption intensity at the temperature about 60 K. That effect was attributed to an unconventional spin-glass-like behavior, caused by competition of long-range dipole-dipole interaction and anisotropy fields in ferromagnetic Co nanoparticles [269]. 8.7. Identification of Adsorbed Cu” Ions at the Ti02 Surface and Electrochemical Behaviour of Copper-Modified Electrodes
The study of the interaction of transition metal ions with the titanium dioxide surface as well as the study of the properties of systems obtained as a result of this interaction is of interest from several viewpoints: a) high chemical stability and high ionexchange capacity of hydrous Ti02 allow to consider this material as a promising inorganic ion-exchanger and sorbent [270];b) Ti02 modified with ions and small particles of various metals can be used as an efficient catalyst and photocatalyst [l-6,12, 2711. There were many works devoted to the study of Ti02 systems involving ions, complexes and small particles of noble metals, but considerably less attention has been given to the processes involving Ti02 matrix doped with ions of common transition metals such as copper. At the same time, the so called “ternary surface complexes” (TSC) have been widely studied as “dark” catalysts and important objects in the area of aquatic and surface chemistry [201-
210,272-2841. Cupric ions are attached firmly to the Ti02 surface with the formation of copper complexes involving Ti02 surface groups. This process may result in the appearance of surface states at the Cu-modified Ti02 films. In order to define structural and electrochemical properties of Cu-Ti02 surface systems, the nano-structured Ti02 film electrodes modified by Cu2+ions have been studied with the use of EPR, X P S , electrolyte electroreflection (EER), and electrochemical methods [ 101, 201, 285-2881.The EER method was used to investigate the band structure, to determine flat-band potentials, and to identify surface states of semiconductor electrodes. The treatment of the powder samples with Cu2+-containingsolutions was carried out under the same conditions as in case of the electrodes [loll.Typical EPR spectra of Cu2+are shown in Fig. 8.18.
Chapter 8,A.I. Kokorin
I
2600
1
I
2800
I
I
3000
.
I
3200
,
-
I
3400
,
243
I
3600
,
I
3800
H, G
Fig. 8.18. EPR spectra of Cu (11) ions adsorbed onto Ti02 nanostructured samples from solutions containing 0.001 M Cu2+,pH 6.0 (1) and 0.1 M Cu", pH 3.1 (4). Dashed ( 2 ) and dotted (3) lines present two additive signals forming spectrum (1) as a superposition. T = 77 K. P881
Cu2' ions have the 3d9 electronic configuration with one unpaired electron (S = 1/2) and I = 3/2 for both 63Cuand 65Cuisotopes. Cu(I1) complexes are usually arranged in a slightly distorted octahedron with four ligand atoms disposed approximately in the equatorial plane of the coordination sphere, lying rather close to the copper nucleus and strongly bonded. The two other ligands are placed to the 5th and 6th axial positions, and these bonds are usually much weaker as compared to equatorial ones. The main values of gand A-tensors of such tetragonally distorted complexes in the case of the axial symmetry of the Cu(I1) complexe are gll,gl; All,Al, which are collected in Table 8.12 for oxide carriers in the absence of organic ligands. One can see from the Table that Cu2' ions, adsorbed onto any oxide support, are easily differed from copper aqua-complexes by the value of 811. It is also seen that Ti02-Cu2' surface complexes are similar to Si02-Cu2+ones by All values, but different by 811 values; with A1203-Cu2+centers there is an opposite correlation. The chemistry of adsorbed Cu(I1) and Mn(I1) complexes in aqueous T i 0 2 suspensions was studied as a function of pH in [209]. Unfortunately, copper concentration used in that work was too high, which resulted in strong dipolar broadening and did not allow to measure the precise g and A values. Additional information on the nature of copper complexes formed on the TiOz surface and their relative amount has been obtained in [285]. Fig. 8.18 represents typical EPR spectra of Cu2' ions adsorbed onto T i 0 2 nanoparticles at pH 3.1 and 6.0. At pH 3.1, the spectrum is typical of isolated mononuclear Cu(I1) complexes. Some noticeable broadening of the ESR lines corresponding to the parallel orientation of Cu(I1) complexes in the external magnetic field, which is observed even for the most magnetically diluted samples, is indicative of the existence of several paramagnetic centers with similar, but not identical structures. The increase of Cu(I1) concentration results initially in an increase of
244
ESR of Nanostructured Semiconductors
the spectrum amplitude [285]. After certain concentrations, depending on the pH value, a more complicated spectrum appeared (Fig. 8.18, curve 1). A computer analysis showedthat this spectrum is a superposition of two different ones: the initial multiplate spectrum of the isolated adsorbed Cu(I1) ions and the anisotropic singlet signal (curves 3 and 2 in Fig. 8.18, correspondingly). Table 8.12 EPR parameters of Cu(I1) surface complexes at 77 K
Sample
All.lo4, cm-’
gll
gl
Ref.
Ti02-Cu2+
142.1
2.360
2.080
287
-“-,298 K
137.1
2.350
2.076
287
Ti02-Cu2+
141.5
2.349
2.080
101
-“-,298 K
137.7
2.341
2.070
101
Ti02-Cu2+
139,8
2.358
2.075
285
TiO2-Cu2+
137
2,342
2.075
203
Ti02-CuZ+,A
134
2.351
208
-“-,B
146
2.334
208
Si02-Cu2+
138.2
2.387
2.075
287
-1‘-
137.2
2.389
2.079
205
128
2.380
2.085
282
135.7
2.390
2.088
284
s~o~-cu~+
139
2.38
2.08
203
&A1203
148
2.357
2.085
202,203
Gibbsite (A120,)
154
2.35
2.06
210
C~(Hz0)62+
132.4
2.424
2.092
101
-‘‘-
132.9
2.421
-‘I-,
300 K
-“-,298 K
298
The relative contribution of both types to a resulting spectrum depended on the pH value and Cu2+ concentration in the solution used for TiOz treatment. Such broad anisotropic singlets are typical of Cu(I1) complexes with strong magnetic dipole-dipole and spin exchange interaction. They were observed at various carriers treated with metal ions [209, 210, 2901, and were related to the formation of specific areas with very high local concentration of paramagnetic centers, which were called “associates”. The average distance between the nearest neighbouring Cu(I1) ions in “associates” on the surface of Ti02 samples was estimated according to [ 16,201 and ranged from 0.6 to 0.9 nm [285]. In such superimposed spectra as shown in Fig. 8.18, the individual spectrum shapes of a multiplate and of a single line are not changing with varying the copper concentration, i.e., the individual amounts of “associated” and isolated Cu(I1) ions can be
Chapter 8, A.I. Kokorin
245
calculated at any their ratio in the sample [285]. As an example, Fig. 8.19 shows the estimated total content of paramagnetic Cu2+ions, [Cu2+&p~, existing on the Ti02 surface, and the amount of Cu2+ions forming the “associates”, [CU~+~~,],,,, for a pH 6.0 value. This figure also shows the total amount of adsorbed copper, [Cu2+dsltot,which was estimated from the electrochemical data (using the charge of the cathodic peaks) in [285].
l0g[CuZ’],
measured Fig. 8.19. Dependence of the total amount of the adsorbed Cu(I1) species (I) electrochemically;the total amount of adsorbed paramagnetic Cu2+ions (0),and Cu2+ions forming the “associates” with high local concentration (x), measured by ESR, on logarithm of [ C U ~ +in]solutions ~ used for Cu(I1) adsorption onto the nano-Ti02particles at pH 6.0. [291]
-0,6
-0,4
-0,2
Potential (V vs. AgIAgCI)
Fig. 8.20. Dependence of the dark cathodic current on the electrode potential for copper-modified TiOz film electrodes which were equilibrated in solutions with different concentration of Cu2+ ions at pH 4.3. Electrolyte: 0.1 M Ar-saturated acetic buffer solution with pH 4.4.Potential sweep rate: 5 m v s-’.
246
ESR of Nanostructured Semiconductors
The analysis of the results published in [285, 2911 reveals that at pH 4.3, the [CU2+&]EpRcoincides very closely with [ c u ~ for + cu2+ ~ ~ concentrations ~ ~ ~ ~ in solution up to l o 2M. The treatment of the nanostructured TiOz samples with lo2M Cu2+solution at pH 6 resulted in the accumulation of about 5% copper in the TiOz film. Considering the specific area of the Ti02 oxide and taking the radius of the hydrated Cu2+ions to be 3.42.10-" m [209], adsorption of this copper amount must cover the Ti02 surface with ca. 0.9 monolayer of hydrated cupric ions. It is important to note that at such rather high levels of the surface coverage, there exists a marked amount of magnetically isolated Cu2+ions on the surface, indicating strongly inhomogeneous distribution of Cu2+ions over the TiOz surface. It was found [285, 2911 that there exist four different kinds of copper species on the Ti02 surface, which correspond to peaks at the potentiodynamic current-potential curves (Fig. 8.20): a) isolated monovalent Cu(1) ions formed as a result of partial reduction of the adsorbed Cu2+ions by electrons of the Ti02 matrix (they correspond to peak A in Fig. 8.20); b) magnetically isolated Cu(I1) ions anchored to the surface via oxygen atoms; c) "associates" of Cu(I1) ions with a strong dipolar and spin-exchange interac-tion between them; (b) and (c) species are elecprochemically reduced in the region of peak B); d) formally diamagnetic copper hydroxide particles (attributed to peak C). The following Table 8.13 presents the quantitative description of such copper species at two pH values in the case when Cu2+ions were adsorbed from [ C U ( C ~ O , )= ~]~ 0.01 M aqueous solutions. Table 8.13
Differenttypes of copper complexes existing on the Ti02 surface at 77 K [291] Type of the adsorbed complex
pH 4.3
pH 6.0
[Cu].105,M/g
% of total [CUI
[Cu].105,M/g
% of total [CUI
(TiOz-Cu')
0.9
21
8.7
10
(Ti02-CU2+)individual
2.9
67
3.4
4
{TiO*-C~2+Iassociat~ {Ti02-Cu(OH)2)
0.3
7 5
3.4
4
0.2
71.5
82
Total [CUI
4.3
100
87
100
The identification of adsorbed Cu'ions on the TiOz surface was done in [285]. The charge consumed in the region of peak A was about 10% of that corresponding to peaks B and C. The X P S studies of the nanostructured TiOz films treated with Cu2+-containing solutions, and the preliminary cathodic reduction of the electrode at a potential -1.4 V for 10 min just before Cu2+adsorption, evidently confirmed the assumption that the nature of peak A was associated with the existence of Cu+ions 12851. It was also shown that there exist formally diamagnetic Cu(I1) species on the Ti02 surface which can not be recorded by ESR method, and they are electrochemically reduced
Chapter 8, A.I. Kokorin
247
at more negative potentials than the adsorbed paramagnetic Cu(I1) ions. These species seem to have copper hydroxide nature (in copper hydroxide, all the spins are coupled and it is formally diamagnetic). In neutral and alkaline solutions, the redox potential of CU(OH)~/CUO pair is more negative than that of Cu2+/Cuopair [292]. It was shown that the fraction of diamagnetic copper species and that of “associates” increase with increasing the Cu2+ concentration in solution (Fig. 8.19) and its pH (Fig. 8.21). At pH 6.0, the diamagnetic species constitutes the most part of all adsorbed copper (about 90% at initial [CUI, > M) and, correspondingly, the cathodic peak C has become predominant on the current potential curves [285].
15 10 -
20
50-
3
I
I
I
I
4
5
6
7
pH value
Fig. 8.21. Dependence of the total amount of the adsorbed copper species (1) and the amount of the adsorbed paramagnetic Cu(I1) ions (2) on the pH value of solutions contained [CuZ+]o = lo4 M. [291]
It should be noted that there are two basically different possible structures of attaching the metal complexes to a surface: EDTA
I
Cu2+
Cu2’
I
EDTA
In the presence of organic ligands, Cu2+ ions adsorbed onto the surface of T i 0 2 nanoparticles can effectively form ternary surface complexes (TSC). The All and gll values of Cu(I1) centers are sensitive to the nature of ligands in the coordination sphere, therefore, they can be used to assign the structural rearrangements in a complex. The hyperfine splitting A1 is usually so small that the perpendicular component of the EPR spectrum shows usually no splitting at all. Spin-Hamiltonian parameters of some TSC are
248
ESR of Nanostructured Semiconductors
given in Table 8.14. So far, we have come across only a few examples of Cu(II)-TiOz-TSC studied by means of the EPR technique [201, 287, 2881. Much more data have been published for other oxide carriers (see, e.g., references in Table 8.14). We have added to the table EPR parameters for Cu2+complexes with the same ligands frozen at 77 K in vitreous solutions for easier comparing with analogous values of the adsorbed species.
Table 8.14 EPR parameters of Cu(I1) ternary surface complexes at 77 K bpy, 2,2’-bipyridine; pic, a-picolinic acid. Complex
~ , , . 1 0cm-’ ~,
gll
gL
Ref.
166.2
2.282
SiOZ-Cu(bpy)
172.4
2.280
2.067
205
-”-,298 K
171.1
2.283
2.067
205
162.6
2.262
2.074
205
-”-, 298 K
156.1
2.267
2.077
205
SiOz-Cu(pic)
161.1
2.316
2.067
205
SiO2-Cu(pi~)~
Ti02-Cu(bpy),298 K
~iOz-C~(bPY)2
287,288
185.0
2.252
2.061
205
SiOz-Cu(bpy)
165
2.290
2.067
282
SiO2-C~(bpy)~
2.073
282
150
2.259
SiOz-Cu(pic)
155
2.318
282
SiOz-Cu(pic)2
180
2.255
282
Al203-Cu(bpy)
177
2.295
203
AlzO3-Cu(pic)
170
2.300
203
WbPY)
165
2.315
2.072
293
C~(~PY)WZO)Z
166
2.308
2.068
294
WbPY )2
165
2.285
2.082
293
Cu(bPY)z
161
2.269
2.082
294
CU(~PY)~
160.5
2.271
2.073
293
W ~ P Y ) ~ Cu(pic)+
161
2.266
2.070
294
160.5
2.332
2.070
295
CU(PiC)Z
178
2.259
2.059
295
C~(pic)~-
165
2.284
2.059
295
Cu(pic)+
162
2.332
296
CU(PiC)Z
175
2.261
296
249
Chapter 8,A.I. Kokorin
In the vast majority of investigations, A-type-structures have been realized (one can replace EDTA in the scheme by any mono- or bidentate organic ligand). As a result, the EPR parameters of such TSC and of isolated mononuclear species (of analogous compositions) in frozen solutions are rather similar, differing slightly by the All values for oxides of various nature (Table 8.14). Another situation has been unambiguously observed in the case of a tetrademate ligand, N,N’-ethylenediaminetetraacetate(edta). Table 8.15 Spin-Hamiltonianparameters of Cu(I1)-edta ternary surface complexes at 77 K
Complex
~ , ~ . 1cm-’ 0~,
gll
g1
Ref.
d
TiO2{Cu-edta) a “A”
135
2.318
TiOz{Cu-edta)
“B”
142
2.343
TiOz{Cu-edta)
“B”
143.5
2.329
TiOz[Cu-edta} “C”
150
2.284
Cu(edta)
147
2.321
2.040
297
Cu(edta)2
154
2.288
2.032
297
Cu(edda)
175
2.281
298
Cu(edda)
179
2.271
298
Cu(edda)z
183
2.236
298
Cu(mal)
164
2.362
299
Cu(mal)z
172
2.329
299
CU(0X)Z
164
2.318
201
2.071
300
pH 6.9; b, pH 8.0; ‘) pH 2.86; d, g2 = 2.092; g3 = 2.027, A3 = (56 ?r 3).104 cm-’. H4edta,N,N’-ethylenediaminetetraaceticacid; Hzedda,N,N’-ethylenediaminediaceticacid; Hzmal, maleic acid; HZox,oxalic acid.
a)
Kivelson and Niemann [301] showed that both All and gllcorrelate well with the type of ligand atoms bound to Cuz+and with the polyhedron structure [301-3031.Therefore, changes in the EPR spectrum shape and parameters have to reflect rearrangements in the coordination sphere. Fig. 8.22 presents typical EPR spectra of Cu(I1) complexes adsorbed onto nanocrystalline Ti02 particles from solutions containing Cu(N03)~and edta at the ratio [Cul:[edta] = 1:l at different pH values. The line-shape analysis showed that at pH 2.9 and 8.0 the EPR spectra are a superposition of the spectra of at least two different species, while the spectrum, recorded for the sample prepared at pH 6.9 with a short (1 h) time of adsorption, indicates the formation of only one Cu2+species at the surface (type “A”). Spin-Hamiltonian parameters of complexes “B” and ”C” (Table 8.15) compared with those known from the literature (see Refs. in the Table) allowed to conclude that these axial EPR spectra are characteristic of a slightly distorted octahedral coordination sphere, as it is valid for all Cu(II) complexes in frozen solutions listed in the table [2011.
250
ESR of Nanostructured Semiconductors
The EPR spectrum of the complex “A” (Fig. 8.22 and Table 8.15) shows three different g values. This indicates that its polyhedron is significantly distorted compared to the one with axial symmetry.
I
I
I
I
2600
2800
3000
3200
.
,
3400
.
,
3600
H, G
Fig. 8.22. EPR spectra at 77 K of Cu(I1) complexes adsorbed on the Ti02 surface from solutions [CuZ’]:[H2edta2-]= 1:l ([Cu2’] = 0.001 M, 0.1 M NaC104). pH values are equal to 6.9 (l),8.0 ( 2 ) and 2.9 (3). [201]
The computer analysis of EPR spectra has been used for structural characterization of the copper complexes existing on the Ti02 surface [201]. It was concluded that the Cu-edta complexes are bound to the Ti02 surface via 0 atoms of the edta carboxylic groups, and there are no direct bonds between Cu(I1) ions and oxygen atoms of the Ti02 surface. In type “B“ complexes (low pH values), which are also not individual, only oxygen atoms of the carboxylic groups are bound to the Cu2+cations. Type “C” species (absorbed at pH 8.0) shows a considerably larger All value and smaller gll (Table 8.15). The “C” structure is formed either with participation of one N atom from the same edta residue, or with oxygen atoms of two edta molecules (Table 8.15, [297]). The coordination sphere of the “A” complex was assumed as a trigonal-bipyramid with two equatorial nitrogens and one equatorial and one apical carboxylate oxygen in strongly distorted five-coordinated structure [201]. This is in accordance with the low symmetry EPR spectrum. Two other carboxylate groups of the edta molecule are attached to the Ti02 surface. This explanation looks reasonable since in situ FT IR studies gave spectroscopic evidence for the formation of several different surface complexes of the oxalate-anion (-OOC-COO-) on TiOz in the aqueous media [304]. An additional argument for this hypothesis is known from the X-ray structure of the Cu(edta) 13051. Analogous fivecoordinated Cu(I1) species were also observed for some low-molecular and polymer-metal complexes in frozen solutions [306-3081. The probable structure of such complex “A” is drawn in Fig. 8.23 [201]:
Chapter 8, A.I. Kokorin
25 1
0 Fig. 8.23. Probable structure of the Cu2'-edta-Ti02 complex assumed in [201].
Electrochemical behaviour of the copper-modified Ti02 film electrodes confirmed this hypothesis [201]. Fig. 8.24 shows typical cyclic voltammograms obtained for untreated T i 0 2 film electrodes and for the electrodes modified by Cu2+ions or Cu(I1)-edta complexes. For untreated electrodes, a small cathodic current peak was observed at potentials slightly positive of the T i 0 2 flat band potential. This peak has no coupled anodic counter-peak and it is associated with the reduction of adsorbed 0 2 [101,285,309], traces of which can occur in the Ar-saturated electrolyte. The voltammograms obtained for the T i 0 2 electro-des modified by Cu(I1)-edta complexes were very similar to those recorded for the untreated Ti02 electrodes, while for electrodes modified by Cu2' ions only, the cathodic peak increased markedly, especially if Cu2+ions have been adsorbed from the neutral solutions (Fig. 8.24).
................... 0
-untreated . ......... CU'++EDTA I
I
-0,8
I
-0,4 -0,2 Potential / V vs. Ag/AgCI
-0,6
Fig. 8.24. Cyclic voltammograms for untreated (-)
.
5 - -400 C)
36
I
080
and copper-modifiednanostructured Ti02 film electrodes without (. .) and in the presence of edta (- - -). The Cu(I1) adsorption was carried out in solutions of [Cu(C104)2]= [H4edta]= 0.01 M against the background of 0.1 M NaC104 at 293 K. Electrolyte: 0.1 M deaerated acetic buffer solution with pH 6.0. Potential sweep rate: 5 m V d . I
252
ESR of Nanostructured Semiconductors
It follows from the results obtained that the electrochemical reduction of Cu2+ions does not occur at the nanostructured TiOz film electrodes modified by Cu(I1)-edta complexes [201], i.e., there is no electron transfer from the Ti02 matrix to the metal ion in the TSC, which exists as the B-type structure (see page 291 and Fig. 8.23). The edta in these TSC acts as an isolating bridge and inhibits the charge exchange between the Cu2+ion and the Ti02 matrix. A comparative study of the Cu(I1)-edta-TiOz ternary surface complexes by potentiometry, EPR and electrochemical methods showed that the adsorptive properties of the Cu(I1)-edta complexes are very similar to those of individual edta species [201]. The Cu(I1)-edta adsorption ratio, equal to 1:1, indicated that the complexes were adsorbed intact. The Cu(II)-edta-Ti02 surface complex with a distorted structure of the trigonal bipyramid had not been previously observed in solutions. It was revealed that Cu(I1)-edta complexes could be electroreduced at a glassy carbon electrode in the same potential region, where the nano-Ti02 electrodes were inactive [201]. The adsorption of Cu2+ions on the Ti02 forms electroactive surface states within the band gap of the oxide, whose energy position was determined by the electrolyte electroreflection method [285, 2861. These copper-induced surface states were established to be located ca. 1.1 eV below the conduction band edge. Information concerning the subbands of the surface states in the Ti02 electrodes modified with Ag, Pd, Pt and Au one can find in [286, 310, 31 11, as well as in Chapter 6 of this book. Finishing this chapter, we should note that such an interesting field as the EPR spectroscopy in studies of photochemical, photoelectrochemical and photocatalytic processes has not been reviewed. Indeed, doping oxide semicon-ductors with Fe3+,Mo”, Cu”, Cr3+,Nb”, V4+Mn2+or Rh3+,etc., one can await changes in the photoreactivity and changes of the valency of these ions involved in the redox chemical processes, which can be easily detected and interpreted, providing valuable information on the studied systems. Unfortunately, the limited space of this book, does not allow us to discuss this topic, which will be done elsewhere.
8.8. General Conclusions It has been demonstrated that ESR spectroscopy in combination with electrochemical methods allows to obtain important information on the adsorption behaviour and structure of various paramagnetic ions, inorganic and organic radicals on the titanium dioxide surface, and on the nature of species formed as a result of the adsorption. Also, a unique knowledge concerning spatial distribution and its peculiarities can be obtained from the analysis of the long-distance dipole-dipole or spin-exchange interaction between paramagnetic centers. As an example, it was described in part 8.7, how Cu(I1) ions adsorbed strongly at the surface of nanostructured Ti02 films had been reduced under cathodic polarization of the TiOz electrodes up to Cuo. The amount of different forms of the adsorbed cupric ions has been determined from the charge consumed for the electroreduction of Cu2+ions and its comparison with the EPR data recorded for the analogous nanosized powders. The EPR technique can be also used for defining the composition and structural peculiarities of metal complexes formed at the surface of different oxides.
Chapter 8, A.I. Kokorin
25 3
Some properties characteristic of metal-doped oxide semiconductors are summarized below as a result of the presented investigations: 1. A tendency to form metal ion aggregates with high local concentration C,, of the doping ions is typical of metal species both inside the oxide lattice (V4+,Cr3+,Nb5+,Cu2+), and at the surface (Cu”, V4’3 of TiOz nanoparticles. This tendency is less for ZrOz, i.e. it depends on the nature of the specimen and on the specific surface area of the semiconductor particles. 2 . Relative concentrations of different metal species of “aggregated” and “isolated” ions (V4’, Cu23 depend on a) metal content in the sample, and b) the nature of the carrier. 3. T o explanain photocatalytic and photoelectrochemical behaviour of such nanocrystalline systems one should take into consideration existence of various types of metaldoped T i 0 2 and ZrOz nanoparticles with different structural forms and spatial distribution of metal centers on their surface.
REFERENCES 1. 2. 3.
4.
5. 6.
7. 8. 9. 10. 11. 12. 13. 14. 15. 16.
Energy Resources through Photochemistry and Catalysis, M. Gratzel (Ed.), Academic Press, New York (1983). Photoelectrochemistry, Photocatalysis and Photoreactors, M. Schiavello (Ed.), Reidel Publ. Co., Dordrecht (1985). Photocatalytic Conversion of Solar Energy, K. I. Zamaraev (Ed.), V. 1, 2, Nauka, Novosibirsk (1985) (in Russian). Kulak A. I., Electrochemistry of Semiconductor Heterostructures, Nauka, Minsk (1986) (in Russian). Schiavello M., Photocatalysis and Environment: Trends and Applications, Kluwer Academic Publishers, Dordrecht (1988). Photocatalysis: Fundamentals and Applications, N. Serpone and E. Pelizzetti (Eds.), Wiley & Sons, New York (1989). Pleskov Yu. V., Photoelectrochemical Conversion of Solar Energy, Nauka, MOSCOW (1990) (in Russian). Photocatalytic Conversion of Solar Energy. Heterogeneous, Homogeneous and Molecular Structurally-Organized Systems. K. I. Zamaraev and V. N. Parmon (Eds.), Nauka, Novosibirsk (1991) (in Russian). Photochemical Conversion and Storage of Solar Energy, E. Pelizzetti and M. Schiavello (Eds.), Kluwer, Dordrecht (1991). Photocatalytic Purification and Treatment of Water and Air, D. F. Ollis, H. Al-Ekabi (Eds.j , Elsevier, Amsterdam (1993). Fujishima A., Hashimoto K. and Watanabe T., Ti02 Photocatalysis. Fundamentals and Applications. BKC Inc., Tokyo (1999). Hoffmann M. R., Martin S. T., Choi W. and Bahnemann D.W. Chem. Rev., 95, No. 1, 69-96 (1995). Hagfeldt A. and Gratzel M. Chem. Rev., 95, No. 1,49-68 (1995). Homogeneous and Heterogeneous Photocatalysis, E. Pelizzetti and N. Serpone (Eds.), Reidel Publ. Co., Dordrecht (1986). Grltzel M., Heterogeneous Photochemical Electron Transfer, CRC Press, Boca Raton, FL (1989). Abragam A., The Principles of Nuclear Magnetism, Clarendon Press, Oxford (1961).
254
ESR of Nanostructured Semiconductors
17. Blumenfeld L. A., Voevodskii V. V. and Semenov A. G., Using of Electron Paramagnetic Resonance in Chemistry, Nauka, Novosibirsk (1962) (in Russian). 18. Carrington A. and McLachlan A. D., Introduction to Magnetic Resonance with Applications to Chemistry and Chemical Physics, Harper & Row, New York (1967). 19. Kuska H. A. and Rogers M. T., ESR of First Row Transition Metal Complex Ions, Interscience Publ., New York (1968). 20. Lebedev Ya. S. and Muromtsev V. I., EPR and relaxation of the stabilized radicals, Khimiya, Moscow (1971) (in Russian). 21. Al’tshuler S. A. and Kozyrev B. M., E. P. R. of Compounds of Intermediate Group Elements, Nauka, Moscow (1972) (in Russian). 22. Wetrz J. E. and Bolton J. R., Electron Spin Resonance, McGraw-Hill, New York (1972). 23. Molin Yu. N., Salikhov K. M. and Zamaraev K. I., Spin Exchange, Springer-Verlag, Berlin, New York (1980). 24. Zhidomirov G. M., Lebedev Ya. S., Dobryakov S. N., Shteinshneider N. Ya., Chirkov A. K. and Gubanov V.A., Interpretation of Complex ESR Spectra, Nauka, Moscow, (1975) (in Russian). 25. Parmon V. N., Kokorin A. I. and Zhidomirov G. M., Stable Biradicals, Nauka, Moscow (1980) (in Russian). 26. Iyengar R. D. and Codell M. Adv. Colloid Interface Sci., 3, No. 2,365-388 (1972). 27. Lunsford J. H. Catal. Revs., 8, No. 1, 135-157 (1973). 28. Atkins P. W. and Symons M. C. R. The Structure of Inorganic Radicals, Elsevier, Amsterdam, New York (1967). 29. Howe R. F. and GrPtzel M. J. Phys. Chem., 89, No. 21,4495-4499 (1985). 30. Howe R. F. and GrPtzel M. J. Phys. Chem., 91, No. 14, 3906-3909 (1987). 31. Nakaoka Y. and Nosaka Y. J. Photochem. Photobiol. A, Chem., 110,299-305 (1997). 32. Serwicka E., Schindler R. N. and Schumacher R. Ber. Bunsenges. Phys. Chem., 85, 192-195 (1981). 33. Micic 0. I., Zhang Y., Cromack K. R., Trifunac A. D. and Thurnauer M. C. J. Phys. Chem., 97, NO. 28, 7277-7283 (1993). 34. Rajh T., Nedeljkovic J., Chen L. X., Tiede D. M. and Thurnauer M. C. J. Adv. Oxid. Technol., 3, NO. 3, 292-298 (1998). 35. Rajh T., Nedeljkovic J., Chen L. X., Poluektov 0. and Thurnauer M. C. J. Phys. Chem. B, 103, NO. 18, 3515-3519 (1999). 36. Cao F., Oskam G., Searson P. C., Stipkala J. M., Heimer T. A., Farzad F. and Meier G. J. J. Phys. Chem., 99, No. 31, 11974-11980 (1995). 37. Anpo M., Yabuta M., Kodama S. and Kubokawa Y. Bull. Chem. SOC.Japan, 59, No. 1,259264 (1986). 38. Anpo M., Shima T., Kodama S. and Kubokawa Y. J. Phys. Chem., 91, No. 16, 4305-4310 (1987). 39. Adamski A., Sojka Z., Dyrek K., Che M., Wendt G. and Albrecht S. Langmuir, 15, No. 18, 5733-5741 (1999). 40. Giamello E., Rumori P., Geobaldo F., Fubini B. and Paganini M. C. Appl. Magn. Reson., No. 2, 173-192 (1996). 41. Volodin A. M. Catal. Today, 58, 103-114 (2000). 42. (a) Bedilo A. F., Timoshok A. V. and Volodin A. M. Catal. Lett., 68, 209-214 (2000); (b) Volodin A. M., Alekseev 0. S., Bolshov V. A. and Ryndin Yu. A. Kinetika i Kataliz, 31, No. 1, 179-185 (1990) (in Russian). 43. Bobricheva I., Stavitsky I., Yermolaev V., Kotsarenko N., Shmachkova V. and Kochubey D. I. Catal. Lett., 56,23-27 (1998).
Chapter 8, A.I. Kokorin
255
44. Ivanovskaya M. I., Bogdanov P. A., Gurlo A. Ch. and Ivashkevich L. S . Russ. J. Inorg. Mater., 34, No. 3, 329-334 (1998) (in Russian). 45. Mtriaudeau P., Che M. and Jorgensen C. K. Chem. Phys. Lett., 5 , No. 3, 131-133 (1970). 45a. Gravelle P. C., Juillet F., Mtriaudeau P. and Teichner S . J. Faraday Discus. Chem. SOC.,No. 52, 140-148 (1971). 46. Kiwi J., Suss J. T. and Szapiro S . Chem. Phys. Lett., 106, No. 1, 135-138 (1984). 47. Antufiev V. V., Vasil'ev Ya. V., Votinov M. P., Kharitonova 0. K. and Kharitonov E. V. Solid State Phys., 4, No. 6, 1496-1499 (1962) (in Russian). 48. Busca G., Centi G., Marchetti L. and Trifiro F. Lungmuir, 2, No. 5, 568-577 (1986). 49. Maschenko A. I., Kazanskii V. B., Parijsky G. B. and Sharapov V. M. Kinetika I Kataliz, 8, No. 4, 853-861 (1967) (in Russian). 50. Liu H., Feng L., Zhang X. and Xue Q.J. Phys. Chem., 99, No. 1,332-334 (1995). 51. Morterra C., Giamello E., Orio L. and Volante M. J. Phys. Chem., 94, No. 7, 3111-3116 (1990). 52. Torralvo M. J, Alario M. A. and Soria J. J. Catal., 86, No. 2,473-476 (1984). 53. Matta J., Lamonier J.-F., Abi-Aad E., Zhilinskaya E. A. and Aboukai's A. Phys. Chem. Chem. Phys., 1, NO. 21,4975-4980 (1999). 54. Frolova E. V., Ivanovskaya M. I., Azarko I. I. and Sviridov V. V. Proc. Nut. Acad. Sci. Belarus, 46, No. 2, 74-77 (2002) (in Russian). 55 (a) Ivanovskaya M. and Bogdanov P. Sensors and Actyators B, 53, 44-53 (1998); (b) Ivanovskaya M., Bogdanov P., Faglia G. and Sberveglieri G. ibid., 68, 344-350 (2000); (c) Orlik D. R., Ivanovskaya M. I., Branitsky G. A. and Bogdanov P. A. ibid., 13-14, 605-607 (1993). 56. Gurlo A,, Ivanovskaya M., Pfau A., Weimar U. and Gopel. W. Thin Solid Films, 307,288-293 (1997). 57. Ioffe V. A. and Patrina I. B. Solid State Phys., 10, No. 3, 815-821 (1968) (in Russian). 58. Popov V. P. and Morozova L. V. Zh. Neorg. Khimii, 40, No. 1, 37-41 (1995). 59. Malchenko S. N., Ivanovskaya M. I., Orlik D. R., Lyakhov A. S . and Grigorenko V. I. Russ. J. Inorg. Mater., 28, No. 1, 103-107 (1992) (in Russian). 60. Volodin A. M. and Cherkashin A. E. React. Kinet. Catal. Lett., 17, No. 3-4, 323-327 (1981). 61. Ivanovskaya M. I., Orlik D. R., Vrublevsky A. I., Malchenko S . N. and Branitsky G. A. Vestnik Balarus. Univers., Ser. 2, No. 1, 19-23 (1990) (in Russian). 62. Wong N.-B., Taarit Y. B. and Lunsford J. H. J. Chem. Phys., 60, No. 5,2148-2151 (1974). 63. Volodin A. M. and Cherkashin A. E. React. Kinet. Catal. Lett., 18, No. 1-2, 243-246 (1981); ibid., 20, NO. 3-4, 335-338 (1982). 64. Volodin A. M. and Cherkashin A. E. Kinetika i Kataliz, 22, No. 4, 979- (1981) (in Russian). 65. Morrison S . R. and Bonnelle J. P. J. Catal., 25, No. 3, 416-424 (1972). 66. Muller K. A. and Schneider J. Phys. Lett., 4, No. 5, 288-291 (1963). 67. Hausmann A. Z. Physik, 237, No. 1, 86-89 (1970). 68. Kokes R. J. J. Phys. Chem., 66, No. 1, 99-103 (1962). 69. Volodin A. M. and Cherkashin A. E. React. Kinet. Catal. Lett., 20, No. 3-4, 347-350 (1982). 70. Bedilo A. F. and Volodin A. M. React. Kinet. Catal. Lett., 67, No. 1, 197-203 (1999). 71. Bedilo A. F., Volodin A. M., Zenkovets G. A. and Timoshok G. V. React. Kinet. Catal. Lett., 55, NO. 1, 183-190 (1995). 72. Timoshok G. V., Bedilo A. F. and Volodin A. M. React. Kinet. Catal. Lett., 59, NO. 1, 165171 (1996). 73. Bedilo A. F. and Volodin A. M. J. Molec. Catal. A, Chem., 158,405-408 (2000). 74. Bedilo A. F., Kim V. I. and Volodin A. M. J. Catal., 176, No. 2, 294-304 (1998).
25 6
ESR of Nanostructured Semiconductors
75. Mikheikin I. D., Maschenko A. I. and Kazanskii V. B. Kinetika I Kataliz, 8, NO. 6, 1363-1368 (1967) (in Russian). 76. Volodin A. M. React. Kinet. Catal. Lett., 28, No. 1, 189-193 (1985). 77. Volodin A. M. React. Kinet. Catal. Lett., 44, NO. 1, 171-177 (1991). 78. Volodin A. M., Bolshov V. A. and Konovalova T. A. Molec. Engineering., 4, No. 1, 201-226 (1994). 79. Van Hooff J. H. C. and van Helden J. F. J. Catal., 8, No. 2, 199-200 (1967). 80. Van Hooff J. H. C. J. Catal., 11, NO. 2,277-279 (1968). 81. Naccache C., MCriaudeau P., Che M. and Tench A. J. Trans. Faraday Soc., 67, NO. 3, 506512 (1971). 82. Svejda P., Hartmann W. and Haul R. Ber. Bunsenges. Phys. Chem., 80, No. 12, 1327-1330 (1976). 83. Giamello E. and Che M., in: Electron Magnetic Resonance of Disordered Systems, N. D. Yordanov (Ed.), pp. 112-133, World Scientific, Singapore (1991). 84. Mkriaudeau P., Naccache C. and Tench A. J. J. Catal., 21, No. 1,208-21 1 (1971). 85. Volodin A. M., Cherkashin A. E. and Zakharenko V. S. React. Kinet. Catal. Lett., 11, No. 2, 107-111 (1979). 86. Anpo M., Shima T. and Kubokawa Y . Chem. Lett., No. , 1799- (1985). 87. Lunsford J. H. and Jayne J. P. J. Chem. Phys., 44, No. , 1487- (1966). 88. Atkins P. W., Keen N. and Symons M. C. R. J. Chem. SOC.,No. ,2873-2879 (1962). 89. Kokorin A. I., Poznyak S. K. and Kulak A. I., submitted for publication. 90. Kokorin A. I., Pergushov V. I., Kulak A. I. and Vedrine J., to be published. 91. Lunsford J. H. J. Phys. Chem., 72, No. 6,2141-2144 (1968). 92. Mihelcic D., Mtisgen P. and Ehhalt D. H. J. Atmos. Chem., 3, No. 2, 341-361 (1985). 93. Mihelcic D., Ehhalt D. H., Klomfass J., Kulessa G. F., Schmidt U. and Trainer M. Ber. Bunsenges. Phys. Chem., 82, No. 1, 16-19 (1978). 94. Zeldes H. and Livingston R. J. Chem. Phys., 35, No. 2,563-567 (1961). 95. Golding R. M. and Henchman N. J. Chem. Phys., 40, No. 6, 1554-1564 (1964). 96. Mihelcic D., Volz-Thomas A., P2tz A. W. and Kley D. J. Atmos. Chem., 11, No. 2, 271-297 (1990). 97. Thomas K., Volz-Thomas A., Mihelcic D., Smit H. G. J. and Kley D. J. Atmos. Chem., 29, NO. 1, 17-43 (1998). 98. Gershenzon Yu. M., Il’in S. D., Politenkova G. G. and Remorov R. G. Russ. J. Chem. Phys., 18, NO. 1, 46-50 (1999). 94, No. 16, 5560-5566 (1972). 99. Kasai P. H. and Bishop R. J., Jr. J. Am. Chem. SOC., 100. Jen C. K., Foner S. N., Cochran E. L. and Bowers V. A. Phys. Rev., 112, No. 4, 1169-1182 (1958). 101. Poznyak S. K., Kokorin A. I. and Kulak A. I. J. Electroanalyt. Chem., 442, No. 1/2, 99-105 (1998). 102. Volodin A. M., Biglino D., Itagaki Y . ,Shiotani M. and Lund A. Chem. Phys. Lett., 327, No. , 165-170 (2000). 103. Niwa M., Minami T. and Murakami Y . Bull. Chem. SOC.Japan, 49, No. 2,565-566 (1976). 104. Furuyama S., Monmoto T. and Hirasawa R. J. Phys. Chem., 82, No. 9, 1027-1028 (1978). 105. Lunsford J. H. J. Chem. Phys., 46, No. 11,4347-4351 (1967). 106. Kasai P. H. and Gaura R. M. J. Phys. Chem., 86, No. 21,4257-4260 (1982). 107. Lunsford J. H. J. Phys. Chem., 74, No. 7, 1518-1522 (1970). 108. Lu X., Xu X., Wang N. and Zhang Q.Chem. Phys. Lett,, 300, No. 1, 109-117 (1999). 109. Poznyak S. K., Talapin D. V. and Kulak A. I. J. Phys. Chem. B., 105, No. 21, 4816-4823 (2001).
Chapter 8, A.I. Kokorin
257
110. Bahnemann D. W., Kholuiskaya S. N., Dillert R., Kulak A. I. and Kokorin A. I. Appl. Catalysis B, Environ., 36, No. 1, 161-169 (2002). 111. Tanaka K., Hisanaga T. and Rivera A. P. In: Photocatalytic Purijication and Treatment of Water andAir, D. F. Ollis, H. AI-Ekabi (Eds.), pp. 169-178, Elsevier, Amsterdam (1993). 112. Kat0 K. In: Photocatalytic Purification and Treatment of Water and Air, D. F. Ollis, H. AlEkabi (Eds.), pp. 809-814, Elsevier, Amsterdam (1993). 113. Luca V., Djajanti S. and Howe R. F. J. Phys. Chem. B, 102, No. 52, 10650-10657 (1998). 114. Wang J. A., Limas-Ballesteros R., Lopez T., et al. J. Phys. Chem., 105, No. 40, 9692-9698 (2001). 115. Halasz I. andGati Gy. React. Kinet. Catal. Lett., 11,No. 1, 57-61 (1979). 116. Tanabe K., Ishiya C., Matsuzaki I., Ichikawa I. and Hattori H. Bull. Chem. SOC. Japan, 45, NO. 1,47-51 (1972). 117. Malchenko S. N., Esmaiel E., Ivanovskaya M., Baikov M. V. and Lyakhov A. S. Russ. J. Znorg. Mater., 27, No. 11,2327-2331 (1991) (in Russian). 118. Martin S. T., Morrison C. L. and Hoffmann M. R. J. Phys. Chem., 98, No. 51, 13695-13704 (1994). 119. BahnemannD.W. IsraelJ. Chem., 33, No. 1, 115-136 (1993). 120. Ivanovskaya M., Frolova E., Orlik D. and Gurin V. Proc. of 41h Intern. Cont on Materials for Microelectronics and Nanoengineering, IOM Communications Ltd., London, UK, 259-262 (2002). 121. Anderson C. and Bard A. J. J. Phys. Chem., 99, No. 24,9882-9885 (1995). 122. Kormann C., Bahnemann D. W. and Hoffmann M. R. J. Phys. Chem., 92, No. 18,5196-5201 (1988). 123. Kokorin A. I., in: The Method of Spin Labels and Probes. Problems and Perspectives. N. M. Emanuel and R. I. Zhdanov (Eds.), pp. 61-79, Nauka, Moscow (1986) (in Russian). 124. Distance Measurements in Biological Systems by EPR, L. J. Berliner, G. R. Eaton and S. S. Eaton (Eds.), Kluwer AcadRlenum Publ., New York (2000). 125. Vasserman A. M. and Kovarsky A. L., Spin Labels and Probes in Physical Chemistry of Polymers, Nauka, Moscow (1986) (in Russian). 126. Kokorin A. I. and Shubin A. A. J. Struct. Chem., 30, No. 2, 260-267 (1989). 127. Kokorin A. I. and Shubin A. A. Sov. J. Chem. Phys., 7, No. 8, 1770-1780 (1991). 128. Kokorin A. I., Arakeljan V. M. and Arutyunyan V. M. Proc. of 13-th QUANTSOL, Kirchberg, Austria, 13-16 (2001). 129. Kokorin A. I., Arakeljan V. M. and Arutyunyan V. M. Russian Chem. Bull., 52, in press (2003). 130. De Biasi R. S. and Femandes A. A. R. J. Phys. C, Solid State Phys., 16,5481-5489 (1983). 131. De Biasi R. S., Femandes A. A. R. and Grillo M. L. N. J. Phys. Chem. Solids, 55, No. 5,453455 (1994). 132. Harmalker S. P. and Pope M. T. J. Phys. Chem., 82, No. 26,2823-2825 (1978). 133. Hwang J., Kivelson D. and Plachy W. J. Chem. Phys., 58, No. , 1753-17 (1973). 134. Anderson P. W. Phys. Rev., 82, No. , 342- (1951). 135. Kittel C. and Abrahams E. Phys. Rev., 90, No. ,238- (1953). 136. Dzheparov F. S., Lundin A. A. and Khazanovich T. N. Zh. Ehp. Teor. Fiz. (JETP),92, No. 2, 554-567 (1987) (in Russian). 137. Fel’dman E. B. and Lacelle S. J. Chem. Phys., 104, No. 5 , 2000-2009 (1996). 138. Atsarkin V. A., Vasneva G. A,, Demidov V. V., Dzheparov F. S., Odintsov B. M. and Clarkson R. B. JETP Lett., 72, No. 7,369-372 (2000). 139. Atsarkin V. A., Demidov V. V., Vasneva G. A., Dzheparov F. S., Ceroke P. J., Odintsov B. M. and Clarkson R. B. J. Magn. Reson., 149, NO. 1,85-89 (2001).
258
ESR of Nanostructured Semiconductors
140. Kulikov A. V., Kushch S. D., Fursikov P. V. and Bogatyrenko V. R. Appl. Magn. Reson., in press (2002). 141. Arutyunian V. M., in: Hydrogen Energy Progress, XII, J. C. Bolcich and T. N. Veziroglu (Eds.), EAAH, Buenos Aires, No. 1, 13-22 (1998). 142. Cromer D. T. and Herrington K. , 77, No. 9, 4708-4709 (1955). (1950) (in Russian). 143. Ormont B. F., The Structure of inorganic Compounds, GITTL, MOSCOW 144. Wyckoff R. W. G., Crystal Structures, 2-nd Ed., Vol. 1, 2, John Wiley & Sons, Inc., New York (1964). 145. Kokorin A. I., Pergushov V. I. and Arutyunyan V. M. Appl. Magnet. Reson., in press (2003). 146. Gallay R., van der Klink J. J. and Moser J. Phys. Rev. B, 34, No. 5, 3060-3068 (1986). 147. Grltzel M. and Howe R. F. J. Phys. Chem., 94, No. 6,2566-2572 (1990). 148. Davidson A. and Che M. J. Phys. Chem., 96, No. 24,9909-9915 (1992). 149. Kera Y. and Matsukaze Y. J. Phys. Chem., 90, No. 22, 5752-5755 (1986). 150. MCriaudeau P. and Vedrine J. C. Nouv. J. Chimie, 2, No. 2, 133-136 (1991). 151. Gerritsen H. J. and Lewis H. R. Phys. Rev., 119, No. 3, 1010-1012 (1960). 152. Zverev G. M. and Prokhorov A. M. JETP, 39, No. 1,222-223 (1960) (in Russian). 153. Kubec F. and Sroubek Z. J. Chem. Phys., 57, No. 4, 1660-1663 (1972). 154. Michel-Calendini F. M. and Fichelle G. Phys. Stat. Sol. B., 69, No. 2, 607-618 (1975). 155. Grunin V. S., Ioffe V. A,, Patrina I. B. and Davtyan G. D. Solid State Phys., 17, No. 11, 30343036 (1975) (in Russian). 156. Gilbrow J. R. and Winfield M. E. Molec. Phys., 25, No. , 1073- (1973). 157. Grunin V. S. and Razurneenko M. V. Solid State Phys., 22, No. 8, 2485-2488 (1980) (in Russian). 158. Rey D. K. Solid State Phys., 3, No. 9,2535-2539 (1961) (in Russian). 159. Low W. and Offenbacher E. L. Solid State Phys., 17, No. 1, 155-1 (1965). 160. Klosek S. and Raftery D. J. Phys. Chem. B, 105, No. 14,2815-2819 (2001). 161. Rusiecka M., Grzybowska B. and Gasior M. Appl. Catal., 10, No. 1, 101-110 (1984). 162. Sharma V. K., Wokaun A. and Baiker A. J. Phys. Chem., 90, No. 12,2715-2718 (1986). 163. Centi G., Giamello E., Pinelli D. and Trifiro F. J. Catal., 130, No. 2, 220-237 (1991). 164. Zamaraev K. I. and Nikitaev A. T., in: Free Radical States in Chemistry, pp. 102-112, Nauka, Novosibirsk (1971). 165. Ghosh A. K. and Maruska H. P. J. Electrochem. SOC., 124, No. 10, 1516-1522 (1977). 166. Maruska H. P. and Ghosh A. K. Solar Energy Materials, 1,237-247 (1979). 167. Salvador P. Solar Energy Materials, 2,413-421 (1980). 168. Phillips T. E., Moorjani K., Murphy J. C. and Poehler T. 0. J. Electrochem. SOC.,129, No. 6, 1210-1215 (1982). 169. Matsumoto Y., Kurirnoto J., Shimizu T. and Sato E. J. Electrochem. SOC.,128, No. 5, 10401044 (1981). 170. Sarkisyan A. G., Arakelyan V. M., Stepanian G. M., et al., Sci. Proc. Yerevan State Univ., No. 1, 79-83 (1981) (in Russian). 171. Sarkisyan A. G., Arutyunian V. M., Stepanian G. M., Pogosian A. A. and Khachaturian E. A. Electrochemistry, 21, No. 2, 261-265 (1985) (in Russian). 172. Arutyunian V. M. Optical phenomena in semiconductor photocatalysts. In: Photocatalytic Conversion of Solar Energy. K. I. Zamaraev and V. N. Parmon (Eds.), pp. 228-294, Nauka, Novosibirsk (1991) (in Russian). 173. Kokorin A. I., Arakeljan V. M. and Arutyunyan V. M. Submitted to Russian Chem. Bull. (2003). 174. Sakata K. Phys. Stat. Sol. B, 116, No. 1, 145-153 (1983). 175. Grunin V. S., Ioffe V. A. and Patrina I. B. Phys. Stat. Sol. B, 63, No. 4, 629-634 (1974).
Chapter 8, A.I. Kokorin
259
176. (a) Shimizu T. J. Phys. SOC.Japan, 23, No. 4, 848-856 (1967); (b) Shimizu T. Phys. Lett., 23, NO. 1, 20-21 (1966). 177. Ioffe V. A. and Patrina I. B. Solid State Phys., 6, No. 10, 3045-3049 (1964) (in Russian). 178. Boesman E. and Gillis E. Phys. Stat. Sol., 14, No. 2, 349-353 (1966). 179. Gillis E. and Boesman E. Phys. Stat. Sol., 14, No. 2, 337-347 (1966). 180. Kahn A., Livage J. and Collongues R. Phys. Stat. Sol. A , 26, No. 1, 175-179 (1974). 181. Nabavi M., Sanchez C. and Livage J. Phil. Mag. B, 63, No. 4,941-953 (1991). 182. Kikuchi C., Chen I., From W. H. and Dorain P. B. J. Chem. Phys., 42, No. 1, 181-185 (1965). 183. K a a i P. H. Phys. Lett., 7, NO. 1, 5-6 (1963). 184. Siege1 I. Phys. Rev., 134, No. lA, A193-A197 (1964). 185. Tai N. T., in: Modem Applications of EPWESR. From Biophysics to Materials Science. C. Z . Rudowicz, K. N. Yu and H. Hiraoka (Eds.), pp. 252-255, Springer-Verlag, Singapore (1998). 186. Kera Y., Inoue T. and Matsukaze Y. Bull. Chem. SOC.Japan, 61, No. 3, 761-768 (1988). 187. Bond G. C., S i r k h y A. J. and Parfitt G. D. J. Catal., 57, No. 3,476-493 (1979). 188. Busca G., Marchetti L., Centi G. and Trifiro F. J. Chem. Soc., Faraday Trans. I , 81, No. 6, 1003-1014 (1985). 189. Cavani F., Centi G., Foresti E., Trifirb F. and Busca G. J. Chem. SOC., Faraday Trans., 84, NO. 1, 237-254 (1988). 190. Akimoto M., Usami M. and Echigoya E. Bull. Chem. SOC. Japan, 51, No. 8, 2195-2199 (1978). 191. Kera Y. and Kawashima T. Bull. Chem. SOC.Japan, 61, No. 5, 1491-1499 (1988). 192. Che M., Canosa B. and Gonzalez-Elipe A. R. J. Phys. Chem., 90, No. 4,618-621 (1986). 193. Konnov A. A,, Pershin A. N., Shelimov B. N. and Kazanskii V. B. Kinetika I Kataliz, 24, No. 1, 161-167 (1983). 194. Van Reijen L. L. and Cossee P. Disc. Faraday SOC.,41, No. 1, 277-289 (1966). 195. Inomata M., Mori K., Miyamoto A. and Murakami Y. J. Phys. Chem., 87, No. 5, 761-768 (1983). 196. Chary K. V. R., Reddy B. M., Nag N. K., Subrahmanyam V. S. and Sunandana C. S. J. Phys. Chem., 88, No. 12, 2622-2624 (1984). 197. UedaH. Bull. Chem. SOC.Japan, 52, No. , 1905-19 (1979). 198. Takahashi H., Shiotany M., Kobayashi H. and Sohma J. J. Catal., 14, No. 1, 134-141 (1969). 199. Fenin V. A,, Shvets V. A. and Kazanskii V. B. Kinetika i Kataliz, 16, No. 4, 1046-1050 (1975) (in Russian). 200. Mamedov B. A., Kokorin A. I. and Ragimov A. V. Vysokomolek. Soed. B, 32, No. 5, 358-362 (1991). 201. Schlapfer C. W., Vlasova N. N., Poznyak S. K. and Kokorin A. I. J. Coll. Interface Sci., 239, NO. 2,200-208, (2001). 202. Motschi H. Natumissenschafren, 70, No. 3, 519-520 (1983). 203. Motschi H. Colloids & Surfaces, 9, No. 2, 333-347 (1984). 204. Bowers A. R. and Huang C. P. J. Coll. Interface Sci., 110, No. 2, 575-590 (1986). 205. Kokorin A. I., Vlasova N. N., Pridantsev A. A. and Davidenko N. K. Russian Chem. Bull., 46, NO. 10, 1670-1676 (1997). 206. Kokorin A. I., Vlasova N. N., Poznyak S. K., Pridantsev A. A., Kulak A. I. and Schlaepfer C. W., manuscript is in preparation (2003). 207. Boucher L. J., Tynan E. C. and Yen T. F. In: Electron Spin Resonance of Metal Complexes, T. F. Yen (Ed.), pp. 111-130, Adam Hilger Ltd., London 1969.
260
ESR of Nanostructured Semiconductors
208. Ottaviani M. F., Ceresa E. M. and Visca M. J. Coll. Interface Sci., 108, No. 1, 114-122 (1985). 209. Bleam W. F. and McBride M. B. J. Coll. Interface Sci., 110, No. 2, 335-346, (1986). 210. McBride M. B., Fraser A. R. and McHardy W. J. Clays & Clay Minerals, 32, No. 1, 12-18, (1984). 21 1. Wagner M., Kienzle O., Bonnell D. A. and Riihle M. J. Vac. Sci. Technol. A , 16, No. 3, 10781085 (1998). 212. Inomata M., Mori K., Miyamoto A., Ui T. and Murakami Y. J. Phys. Chem., 87, No. 5, 754761 (1983). 213. Kijenski J., Baiker A., Glinski M., Dollenmeier P. and Wokaun A. J. Catal., 101,No. 1, 1-11 (1986). 214. Pinaeva L. G., Rar A. A., Kalinkin A. V., Zaikovskii V. I., Ivanov V. P. and Balzhinimaev B. S. React. Kinet. Catal. Lett., 41, No. 2, 375-379 (1990). 215. Kokorin A. I., Galkova T. N., Pergushov V. I. and Doubinsky A. A. Proc. of 14-th QUANTSOL, Rauris , Austria, 43-46 (2002). 216. Kokorin A. I., Pridantsev A. A. and Kulak A. I. Abstracts of 5th ISTC Seminar "Nanotechnologies in the area of physics, chemistry and biotechnology", p. 108, St.Petersburg, Russia, May 27-29, (2002). 217. Kokorin A. I., Galkova T. N., Pergushov V. I. and Kulak A. I. Appl. Magnet. Reson., to be published. 218. Volodin A. M., van To1 J., Brunel L.-C., Sobolev V. I., Zenkovetz G. A. and Simakov A. V. NHMFL 2001 Annual Programs Report, 62-63, Florida (2002). 219. Volodin A. M., van To1 J., Brunel L.-C., Zenkovetz G. A. and Popova G. Ya. NHMFL 2001 Annual Programs Report, 63-64, Florida (2002). 220. Goodman B. A. and Raynor J. B. E. S. R. of Transition Metal Complexes, in: Adv. Inorg. Chem. Radiochem., H. J. EmelCus and A. G. Sharpe (Eds.), 13, 135-362 (1970). 221. Sorin L. A. and Vlasova M. V. Electron Spin Resonance of Paramagnetic Cryatals, Plenum Press, New York, London (1973). 222. Orton J. W. Electron Paramagnetic Resonance. An Introduction to Transition Group Ions in Crystals, Iliffe Books Ltd., London (1968). 223. Moser J., Grltzel M. and Gallay R. Helv. Chim. Acta, 70, No. , 1506-1604 (1987). 224. Chester P. F. J. Appl. Phys., 32, No. 5, 866-868 (1961). 225. Zimmermann P. H. Phys. Rev. B, 8, No. 8,3917-3926 (1973). 226. Grunin V. S., Patrina I. B. and Davtyan G. D. Solid State Phys., 20, No. 5, 1556-1558 (1978) (in Russian). 227. Kyi R. Phys. Rev., 128, No. 1, 151-153 (1962). 228. Chang T.-T. Phys. Rev., 136, No. 5A, A1413-AI416 (1964). 229. Chang T.-T. Phys. Rev., 147, No. 1,264-267 (1966). 230. Chang T.-T. Bull. Am. Phys. SOC.,9, No. 3,568-571 (1964). 231. Anderson H. G. Phys. Rev., 120, No. 6, 1606-1611 (1960). 232. Anderson H. G. J. Chem. Phys., 35, No. 3, 1090-1096 (1961). 233. Gerritsen H. J. and Sabisky E. S. Phys. Rev., 132, No. 4, 1507-1512 (1963). 234. Gerritsen H. J. and Sabisky E. S. Phys. Rev., 125, No. 6, 1853-1859 (1962). 235. Yamaka E. and Barnes R. G. Phys. Rev., 125, No. 5, 1568-1569 (1962). 236. Zverev G. M. and Prokhorov A. M. JETP, 43, No. 2(8), 422-425 (1962) (in Russian). Zverev G. M. and Prokhorov A. M. In: Paramagnetic Resonance, Proc. 1-st Intern. Con&, W. LOW (Ed.), Academic Press, New York, I, 13-16 (1963). 237. Ismailov E. G., Zul'fugarov Z. G., Maksimov N. G. and Anufrienko V. F. Doklady AN SSSR, 244, NO. 2, 392-395 (1979).
Chapter 8, A.I. Kokorin
261
238. Simhek E., Sroubek Z., i d h s k j l K., Kaczkr J. and Nov& L. Phys. Status Solidi, 14, No. 2, 333-335 (1966). 239. Siege1 I. and Lorenc J. A. J. Chem. Phys., 45, No. 6, 2315-2320 (1966). 240. Dietz R. E., Kamimura H., Sturge M. D. and Yariv A. Phys. Rev., 132, No. 4, 1559-1569 (1963); Kamimura H. and Yariv A. Bull. Am. Phys. Soc., 8, No. 1, 23 (1963); De Wit M. and Estle T. L. ibid., 24 (1963). 241. Coffman R. E. Phys. Lett., 21, No. 4, 381-383 (1966); Hayes W. Rep. Progr. Phys., 22, No. 2, 204-206 (1959). 242. De Wit M. and Reinberg A. R. Phys. Rev., 163, No. 2,261-265 (1967). 243. Geschwind S. and Remeika J. P. Bull. Am. Phys. SOC., 6, No. 1, 116-118 (1961). 244. From W. H., Dorain P. B. and Locher D. R. Bull. Amer. Phys. Soc., 11, No. 2, 220-223 (1966). 245. Vinokurov V. M., Zaripov M. M., Stepanov V. G., Chirkin G. K. and Shekun L. Ya. Solid Staate Phys., 5, No. 6, 2034-2035 (1963) (in Russian). 246. Yafaev N. R. and Garifyanov N. S. Solid Staate Phys., 5, No. 10, 3025-3026 (1963) (in Russian). 247. Yamaka E. and Barnes R. G. Phys. Rev., 135, No. lA, A144-A148 (1964). 248. IkushimaH. andHayakawaS. J. Phys. SOC. Japan, 19, No. 10, 1986 (1964). 249. YamakaE. J. Phys. SOC.Japan, 18, No. 7, 1557-1558 (1963). 250. Amorelli A., Evans J. C., Roulands C. C. and Egerton T. A. J. Chem. Soc., Faraday Trans. I , 83, NO. 12, 3541-3548 (1987). 251. Amorelli A,, Evans J. C. and Roulands C. C. J. Chem. Soc., Faraday Trans. 1, 85, No. 12, 4031-4038 (1989). 252. Doeuff S., Henry M., Sanchez C. and Livage J. J. Non-Crystal. Solids, 89, No. 1, 84-97 (1987). 253. Sierro J., Muller K. A. and Lacroix R. Arch. des Sci., No. 3, 122-124 (1959). 254. Gerritsen H. J., Harrison S. E., Lewis H. R. and Wittke J. P. Phys. Rev. Lett., 2, No. 4, 153155 (1959). 255. Barry T. I. Solid State Comm., 4, No. 3, 123-124 (1966). 256. Sierro J., Lacroix R. and Muller K. A. Helv. Phys. Acta, 32, No. 2,286-288 (1959). 257. Evans J. C., Owen C. R. and Rowlands C. C. J. Chem. Soc., Faraday Trans. I , 85, NO. 12, 4039-4046 (1989). 258. From W. H. Phys. Rev., 131, No. 3, 961-963 (1963). 259. De Biasi R. S. and Rodrigues D. C. S. J. Mater. Sci., 16, No. 5,968-972 (1981). 260. Gerritsen H. J., Harrison S. E. and Lewis H. R. J. Appl. Phys., 31, No. 9, 1566-1571 (1960). 261. Gemtsen H. J. In: Paramagnetic Resonance, Proc. 1-st Intern. Con$, W. Low (Ed.), Academic Press, New York, I, 3-12 (1963). 262. Grunin V. S., Davtyan G. D., Ioffe V. A., and Patrina I. B. Solid State Phys., 17, No. 7, 21742176 (1975) (in Russian). 263. Schimitschek E. J. Phys. Rev., 130, No. 6, 2199-2200 (1963). 264. O’Reilly D. E. and MacIver D. S.J. Phys. Chem., 66, No. 2, 276-282 (1962). 265. Karthein R., Motschi H., Schweiger A., Ibric S., Sulzberger B. and Stumm W. Znorg. Chem., 30, NO.7, 1606-1611 (1991). 266. Charlet L. and Karthein R. Aquatic Sci., 52, No. 1,93-102 (1990). 267. Valentin C., Folgado J. V. and Alarc6n J. Mater. Res. Bull., 36, 1615-1627 (2001). 268. Tartaj P., Serna C. J., Soria J. and Ocana M. J. Mater. Res., 13, No. 2,413-417 (1998). 269. Rameev B. Z., Yildiz F., Tagirov L. R., Aktas B., Park W. K. and Moodera J. S. J. Magn. Magn. Mater., in press (2003).
262
ESR of Nanostructured Semiconductors
270. (a) Yamashita H., Ozawa Y., Nakajima F. and Murata T. Bull. Chem. SOC.Japan., 53, No. 1, 1-5 (1980); (b) Inoue Y. and Yamazaki H. ibid., 53, No. 4, 811-815 (1980); (c) Malati M. A. and Smith A. E. Powder Technol., 22, No. 2, 279-282 (1979); (d) Bonsack J. P. J. Colloid Interface Sci., 44,430-436 (1973). 271. Choi W., Termin A. and Hoffmann M. R. J. Phys. Chem., 98, No. 51, 13669-13679 (1994). 272. Aquatic Surface Chemistry, W. Stumm (Ed.), John Willey & Sons Inc., New York (1987). 273. Ludwig Ch. and Schindler P. W. J. Coll. Interface Sci., 169, No. 2,284-288 (1995). 274. Schindler P. W. Reviews in Mineralogy, 23, No. 2, 281-297 (1990). 275. Errnakov Yu. I., Zakharov V. A. and Kuznetsov B. N. Complexes Fixed on Oxide Supports in Catalysis, Nauka, Novosibirsk (1980) (in Russian). 276. Komova 0. V., Simakov A. V., Rogov V. A., Kochubei D. I., et al. J. Mol. Catal., A: Chem., 161, NO. 1, 191-204 (2000). 277. Sekine R., Kawai M., Asakura K., et al. Surface Sci., 278, 175-181 (1992). 278. Ward M. D. and Bard A. J. J. Phys. Chem., 86, No. 18,3599-3605 (1982). 279. Reiche H., Dunn W. W. and Bard A. J. J. Phys. Chem., 83, No. 17,2248-2251 (1979). 280. Matar K., Zhao D., Goldfarb D., Azelee W., Daniel W. and Harrison P. G. J. Phys. Chem., 99, NO. 24, 9966-9973 (1995). 281. Motschi H. Adsorp. Sci. & Technol., 2, No. 1, 39-54 (1985). 282. von Zelewsky A. and Bemtgen J.-M. Inorg. Chem., 21, No. 5 , 1771-1777 (1982). 283. Dyrek K., Rokosz A., Madej A. and Bidzinska E. Appl. Magn. Reson., 10, No. 2, 319-338 (1996). 284. Xia X., Mehadi A., Taylor R. W. and Bleam W. F. J. Colloid Interface Sci., 185, 252-257 (1997). 285. Poznyak S. K., Pergushov V. I., Kokorin A. I., Kulak A. I. and Schlaepfer C. W. J. Phys. Chem. B, 103, No. 14, 1308-1315 (1999). 286. Kulak A. I., Kokorin A. I. and Sviridov D. V. J. Mater. Res., 16, No. 8,2357-2361 (2001). 287. Kokorin A. I., Pridantsev A. A. and Schlapfer C. W. Proc. of 29th AMPERE Con$ on Magn. Res., Berlin, vol. 11, 1122-1123 (1998). 288. Kokorin A. I., Poznyak S. K., Vlasova N. N., Pridantsev A. A., Kulak A. I. and Schlapfer C. W. Russian J. Chem. Phys., in press (2003). 289. Yokoi H., Otagiri M. and Isobe T. Bull. Chem. SOC.Japan, 46, No. 2,442-446 (1973). 290. Kokorin A. I., Zhilenko M. P. and Rudenko A. P., Russian J. Phys. Chem., 69, No. 10, 16321637 (1995). 291. Kokorin A. I., Kulak A. I. and Arutyunyan V. M. Proc. of 11-th QVANTSOL, Wildhaus, Switzerland, 101- 103 (1999). 292. Standard Potentials in Aqueous Solution, A. J. Bard, R. Parsons and J. Jordan (Eds.), Marcel Decker, Inc., New York (1985). 293. Noack M. and Gordon G. J. Chem. Phys., 48, No. 6,2689-2699 (1968). 294. Walker F. A. and Sigel H. Inorg. Chem., 11, No. 5 , 1162-1164 (1972). 295. Kokorin A. I. and Berentsveig V. V. Bull. Acad. Sci. USSR, Div. Chem. Sci., 37, No. 8, 15371540 (1988). 296. Radionov B. K., Molochnikov L. S . and Babkin 0. N. Russian J. Phys. Chem., 62, No. 7, 1849-1856 (1988) (in Russian). 297. Vishnevskaya G. P., Shapnik M. S . and Safh R. Sh. Zh. Neorg. Khimii, 26, No. 10, 27812785 (1981) (in Russian). 298. Molochnikov L. S. and Radionov B. K. Russian J. Phys. Chem., 67, No. 3, 466-469 (1993) (in Russian). 299. Noji S . and YamaokaK. Macromolecules, 12, No. 6, 1110-1117 (1979). 300. Walker F. A., Sigel H. and McCormick D. B. Inorg. Chem., 11, No. 11,2756-2763 (1972). 301. Kivelson D. and Neiman R. J. Chem. Phys., 35, No. 1,149-161 (1961).
Chapter 8, A.I. Kokorin
263
302. Vanngard T., in: Biological Applications of Electron Spin Resonance, H. M. Swartz, J. R. Bolton and D. C. Borg (Eds.), pp. 41 1-447, Wiley-Interscience, New York (1972). 303. Peisach J. and Blumberg W. E. Arch. Biochem. Biophys., 165, No. 2,691-708 (1974). 304. Hug S. J. and Sulzberger B. Langmuir, 10, No. 10, 3587-3597 (1994). 305. Porai-Kocic M. A., Polynova T. N. and Shkol’nikova L. M. Zh. Mendeleev Chem. SOC.,29, No. 3,283-292 (1984) (in Russian). 306. Rostovshchikova T. N., Smimov V. V. and Kokorin A. I. J. Molec. Catal. A, Chem., 129, No. 1, 141-151 (1998). 307. Blagutina V. V., Pridantsev A. A. and Kokorin A. I. Russian J. Phys. Chem., 72, No. 3, 403406 (1998). 308. Saraev V. V. and Shmidt F. K. ESR of Metal Complex Catalysrs, Izd. IGU, Irkutsk (1985) (in Russian). 309. Wahl A. And Augustynski J. J. Phys. Chem. B, 102, No. 40,7820-7828 (1998). 3 10. Kulak A. I. Photoelectrochemistv 01Semiconductor Oxides, in: Chemical Problems of the Development of New Materials and Technologies, V. V. Sviridiv (Ed.), pp. 415-430, Minsk (1998) (in Russian). 311. Kulak A. I. and Sviridov D. V. Photoelectrochemistv of Semiconductor Oxides and Molecular Aggregates, in: Selected Scientijic Works of the Belarus State University, Vol. 1, pp. 123-138, BSU Publ., Minsk (2001) (in Russian).
