Polymer Photovoltaics A Practical Approach
Polymer Photovoltaics A Practical Approach
Frederik C. Krebs, Editor
Bellingham, Washington USA
Library of Congress Cataloging-in-Publication Data Polymer photovoltaics : a practical approach / Frederik Krebs, editor. p. cm. Includes bibliographical references and index. ISBN 978-0-8194-6781-2 1. Photovoltaic cells--Materials. 2. Conducting polymers. I. Krebs, Frederik. TK8322.P69 2008 621.31'244--dc22
2007050208
Published by SPIE P.O. Box 10 Bellingham, Washington 98227-0010 USA Phone: +1 360 676 3290 Fax: +1 360 647 1445 Email:
[email protected] Web: http://spie.org Copyright © 2008 Society of Photo-Optical Instrumentation Engineers All rights reserved. No part of this publication may be reproduced or distributed in any form or by any means without written permission of the publisher. The content of this book reflects the work and thought of the author(s). Every effort has been made to publish reliable and accurate information herein, but the publisher is not responsible for the validity of the information or for any outcomes resulting from reliance thereon. Printed in the United States of America.
Contents List of Contributors
ix
Preface
xi
List of Abbreviations 1 Introduction 1.1 Human Energy Consumption Now and in the Future 1.2 Renewable Energy Sources 1.3 Important Facts About Energy, Energy Conversion, the Earth, and the Sun 1.4 Solar Energy 1.5 The Storage and Relocation Problem 1.6 Types of Solar Cells 1.7 Current Challenges References 2 The Polymer Solar Cell 2.1 Introduction 2.2 Materials 2.2.1 Polymers 2.2.2 Molecules and oligomers 2.3 Fast and Simple Guide to a Polymer Solar Cell from Scratch 2.3.1 Equipment 2.3.2 The substrate 2.3.3 The PEDOT:PSS layer 2.3.4 The active layer 2.3.5 Evaporating the electrode 2.3.6 Applying electrodes and measuring the electrical properties of the devices 2.3.7 Device preparation and performance References
xiii 1 1 2 4 4 5 6 8 9 11 11 12 12 34 42 42 43 46 47 64 65 68 79
vi
Contents
3 Characterization of Organic Solar Cells 3.1 Taking the Sun Inside 3.1.1 Air mass 3.1.2 The ASTM E 927-05 standard and the IEC 904-9 standard 3.1.3 Types of simulators 3.1.4 Halogen lamps 3.1.5 Recording the spectrum 3.1.6 Applying filters to improve the spectrum 3.1.7 Spectral, temporal, and spatial homogeneity of the light field 3.1.8 Calibration of the sun simulator 3.2 IV-Curves and Efficiencies 3.2.1 The source meter 3.2.2 Where the electrons are and how to connect your cell to the outside world 3.2.3 Speed of IV-curve measurement, dielectric relaxation, and capacitive loading 3.2.4 Action spectra using a high-power spectrometer 3.2.5 IPCE measurements using a simple high-power spectrometer 3.2.6 Environmental effects 3.3 Outdoor Measurements 3.3.1 Why outdoor photovoltaic characterization is necessary for organic solar cells 3.3.2 Experimental procedure 3.3.3 Temperature dependence of the photovoltaic parameters of BHJ solar cells 3.3.4 Example of long-term outdoor testing of stability of organic solar cells 3.3.5 Some new experimental possibilities and suggestions for future studies 3.4 Methods for Preparation and Characterization of Thin Films 3.4.1 Controlling morphological properties 3.4.2 Techniques for monitoring morphology References
91 91 92
128 131 133 144 148
4 Lifetime and Stability Studies 4.1 Overview 4.2 Studies of Degradation Mechanisms Using TOF-SIMS 4.2.1 Principle of TOF-SIMS 4.2.2 Isotopic labeling
155 155 156 156 159
94 97 97 99 100 101 103 105 105 105 108 109 110 112 112 112 113 116 123
Contents
4.2.3 TOF-SIMS depth profiling 4.2.4 Gaining access to the various layers in the photovoltaic device 4.2.5 TOF-SIMS imaging 4.2.6 Chemical structure elucidation based on mass spectral information 4.2.7 Monitoring photooxidation in time—mapping the “history” of degradation 4.3 Studies of Degradation Mechanisms Using XPS 4.3.1 The principle of XPS 4.3.2 Chemical shifts 4.3.3 Angle-dependent studies 4.3.4 Experimental details 4.3.5 Device aging and IV measurement 4.3.6 XPS overall observations 4.3.7 Li and F distribution 4.4 Studies of Degradation Mechanisms Using RBS 4.4.1 Principles and quantitative depth profile of the composition 4.4.2 Studies of cathode degradation using RBS 4.5 Studies of Degradation Mechanisms Using Physical and/or Spectroscopic Techniques 4.5.1 Interference microscopy 4.5.2 Atomic force microscopy (AFM) 4.5.3 Scanning electron microscopy (SEM) 4.5.4 Fluorescence microscopy 4.6 Accelerated Lifetime Measurements for Extended Periods of Time 4.7 Apparatus for Lifetime Measurements and for Isotope Labeling References 5 Processing and Production of Large Modules 5.1 Printing and Coating Methods 5.1.1 R2R coating 5.1.2 Screen printing 5.1.3 Pad printing 5.1.4 Doctor blading 5.1.5 Other printing methods 5.2 Printing the Active Layer 5.2.1 Screen printing 5.3 Carrier Substrates
vii
159 167 170 178 185 187 188 189 189 189 191 192 193 197 198 201 210 211 212 213 216 217 220 223 229 232 232 234 237 237 239 239 239 266
Contents
viii
5.4 Anodes and Cathodes 5.5 Processing of the Transparent Front-Side Contact 5.5.1 PEDOT as transparent contact 5.5.2 Introduction of a conductive grid into photovoltaic devices 5.6 Processing of the Opaque Back-Side Contact 5.6.1 Ag-based pastes as back-side contacts 5.7 Encapsulation and Permeability 5.7.1 Measurement of permeability 5.7.2 Measurement of the diffusion coefficient D 5.7.3 Units 5.7.4 Apparatus 5.7.5 An example of a commercial instrument 5.7.6 The calcium test 5.7.7 Mass spectrometry 5.7.8 Tritiated water 5.7.9 Oxygen permeation in PEDOT 5.8 Practical Encapsulation Techniques 5.8.1 Rigid encasement at IMEC (Belgium) 5.8.2 Small rigid encasement at Risø National Laboratory (Denmark) 5.8.3 Large rigid encasement at Risø National Laboratory (Denmark) 5.8.4 Flexible encasement 5.9 Production and Companies 2007 5.9.1 Intellectual property rights in Europe, the United States, and Asia 2007 5.9.2 A road map for setting up a company producing OPVs in Europe 5.9.3 What production equipment is available in 2007 References 6 Outlook 6.1 Where Is the Technology Now? 6.2 Where Is It Suitable? 6.3 Where Could It Be in the Next Decades? References Index
266 267 269 272 274 275 279 279 281 282 282 283 283 285 285 285 286 287 287 289 290 291 293 294 295 295 301 301 302 303 306 307
List of Contributors Tom Aernouts IMEC Kapeldreef 75 B-3001 Leuven Belgium Rémi de Bettignies Commissariat à l’Ènergie Atomique 17, rue des Martyrs 38054 Grenoble cedex 9 France Eva Bundgaard Risø National Laboratory Technical University of Denmark Frederiksborgvej 399 DK-4000 Roskilde Denmark Stéphane Cros Commissariat à l’Ènergie Atomique 17, rue des Martyrs 38054 Grenoble cedex 9 France Muriel Firon Commissariat à l’Ènergie Atomique 17, rue des Martyrs
38054 Grenoble cedex 9 France Mikkel Jørgensen Risø National Laboratory Technical University of Denmark Frederiksborgvej 399 DK-4000 Roskilde Denmark Eugene A. Katz Ben-Gurian University of Negev P.O.B. 653 Beer-Sheva 84105 Israel Frederik C. Krebs Risø National Laboratory Technical University of Denmark Frederiksborgvej 399 DK-4000 Roskilde Denmark Kion Norrman Risø National Laboratory Technical University of Denmark Frederiksborgvej 399 DK-4000 Roskilde Denmark
Preface Polymer photovoltaics is a discovery that potentially houses the solutions to many of the problems currently encountered with traditional photovoltaic technologies. Most notably, the technology offers the possibility for ultrafast processing, low cost, light weight, flexibility, and a very low thermal budget. The technology rests on a moderately solid base of scientific literature spanning from the first prototypical literature reports. Among the most prominent contributors are the groups of C.W. Tang, R. Friend and A.J. Heeger through an impressive number of original research papers documenting a steady increase in the performance at the level of very small devices with power-conversion efficiencies of up to around 5% for single junctions, which today represent the state of the art. This base of research reports, conference proceedings, reviews, and even many books makes the topic highly accessible to the newcomer and as such there is no need for a new book on the topic from a theoretical or explanatory point of view. One of the problems when entering the field of organic photovoltaics is getting a good idea of how to actually make devices, how to study them, and how to characterize them. The ambition of this book is that it should be a practical guide in the laboratory for the experimental solar cell scientist whether he or she is involved with synthesis, device preparation, processing, or device characterization. Our feeling is that such an experimental guide will be useful to all scientists working practically in the field. This book presents the process of creating a polymer solar cell device starting with a description of materials including how they are made and characterized, followed by how the materials are processed into devices and films and how these are characterized. Following on from this, the status of two emerging fields of polymer solar cells are described, namely, degradation and stability, and large-scale processing. Frederik C. Krebs December, 2007
Acknowledgements Technical assistance by Jan Alstrup is greatly acknowledged. Rolf H. Berg is acknowledged for carrying out a detailed freedom-to-operate analysis within organic photovoltaics. Suren Gevorgyan is acknowledged for preparing small encapsulated modules for the purpose of this book.
List of Abbreviations [60]PCBM or PCBM [70]PCBM AA AFM Ag paste Al Alq3 AM ATRP BCP BE BHJ BIF Ca CAFM CdTe CEA CHA Cu(In,Ga)Se2 CuPc CV CVD DA dppp DSC DTR DUT Ea EBPVD ECN EDG EDX EQE ERDA
[6,6]-phenyl C61 -butyric acid methyl ester [6,6]-phenyl C71 -butyric acid methyl ester Atmospheric air Atomic force microscopy Silver-(epoxy) paste Aluminum Tris(8-hydroxyquinolinato)aluminum Air mass, amount of atmosphere light passes through Atom transfer radical polymerization Bathocuproin Binding energy Bulk heterojunction Barrier improvement factor Calcium Conductive atomic force microscopy Cadmium telluride Commissariat à l’Énergie Atomique Concentric hemispherical analyser Copper indium-gallium diselenide Copper phthalocyanine Cyclic voltammetry Chemical vapor deposition Dry air 1,3-bis(diphenylphosphino)propane Differential scanning calorimetry Diffusion transfer reversal Device under test Activation energy Electron beam physical vapor deposition Energy Research Centre of the Netherlands Electron donating group Energy-dispersive x-ray analysis External quantum efficiency Elastic recoil detection analysis
xiv
ESCA eV EVA EWG F8BT FF FWHM GaAs GW HC-PEDOT HOMO HPLC HTO HWE Impp IPCE IPR Isc ISE ITN ITO IV IV-curve I-V-L K kB kdeg KE KHS575 KOH LCD LED LiF LIP LUMO MALDI-TOF MDMO-PPV MEH-PPV MeOH MeV mfp
List of Abbreviations
Electron spectroscopy for chemical analysis Electron volt ethyl vinyl acetate Electron withdrawing group Poly(9,9-dioctylfluorene-co-benzothiadiazole) Fill factor full width at half maximum Gallium arsenide Gigawatt = 109 W Highly conductive PEDOT Highest occupied molecular orbital High-perfomance liquid chromatography tritium-containing water Horner-Wadsworth-Emmons Current at maximum power point Incident photon to current efficiency Intellectual property rights Short-circuit current Institut für Solare Energisysteme Isothianaphthene Indium tin oxide Current-voltage Current-voltage diode characteristics Current-voltage-luminance Acceleration factor Boltmann constant Degradation constant Kinetic energy Sun simulator from Steuernagel Lichttechnik Potassium hydroxide Liquid crystalline display Light-emitting diode Lithium fluoride Localized irradiation probe Lowest unoccupied molecular orbital Matrix-assisted laser desorption/ionization–time-of-flight Poly(2-methoxy-5-(3,7-dimethyloctyloxy)1,4-phenylenevinylene) Poly[2-methoxy-5-(2-ethylhexyloxy)1,4-phenylenevinylene] Methanol Megaelectron volt Mean free path
xv
Mg MgSO4 Mol wt MW NBS NIR NMR NRA NREL ODCB OLED OPV OTR P3CT PCE PCT PD PE PEDOT PEN PEOPT PET PFB PFO Pin PITN PLD PLED Pmax POMeOPT Pout PPV PSS PTCA PTCBI PTOPT PTV PV PVD R2R
Magnesium Magnesium sulfate Molecular weight Megawatt N-bromosuccinimide Near infrared Nuclear magnetic resonance Nuclear reaction analysis National Renewable Energy Laboratory Orto-dichlorobenzene or 1,2-dichlorobenzene Organic light-emitting device or organic light emitting diode Organic photovoltaic Oxygen transmission rate Poly(3-carboxythiophene-co-thiophene) Power conversion efficiency Patent cooperation treaty Polydispersity Poly(ethylene) Poly(ethylendioxythiophene) Poly(ethylenenaphthalate) Poly[3-(4 -(1 ,4 ),7 -trioxaoctyl)phenylthiophene] Poly(ethylene terephthalat) Poly(9,9-dioctylfluorene-co-bis-N,N-(4-butylphenyl)-bisN,N-phenyl-1,4-phenylenediamine) Poly(9,9-dioctyl-fluorene) Incoming solar power Poly(isothianaphthene) Pulsed laser deposition Polymer light-emitting device or polymer light emitting diode Maximum power Poly[3-(2 -methoxy-5 -octylphenyl)-thiophene] Output electrical power Poly(phenylenevinylene) Poly(styrene sulfonic acid) Perylene tetracerboxylic acid Perylene tetracarboxylic acid bisimide Poly[3-(4-octylphenyl)-2,2 -bithiophene] Poly(thienylenevinylene) Photovoltaic Physical vapor deposition Roll-to-roll
xvi
RBS RR Rs SD SEC SEM SIMS SMU STC Ta TCO TE TEM TEMPO Tg THF TOF TOF-SIMS TW UHV UPS UV UV-vis VD VE Vmpp Voc VTE Vth W Wp WVTR XPS XRD XRD ZJ η ηe ηn Φ Ø
List of Abbreviations
Rutherford backscattering Rectification ratio Series resistance Sputter deposition Size exclusion chromatography Scanning electron microscope Secondary-ion-mass spectrometry Source measure unit Standard test conditions Annealing temperature Transparent conductive oxide Thermal evaporation Transmission electron microscopy 2,2,6,6,-tetramethylpiperidin-1-oxyl Glass transition temperature Tetrahydrofurane Time-of-flight Time-of-flight secondary-ion-mass-spectrometry Terawatt = 1012 W Ultrahigh vacuum Ultraviolet photoelectron spectroscopy Ultraviolet Ultraviolet-visible Vapor deposition Vacuum evaporation Voltage at maximum power Open-circuit voltage Vacuum thermal evaporation Theoretical paste volume Watt Watt peak Water vapor transmission rate X-ray photoelectron spectroscopy Grazing-incidence x-ray diffraction X-ray diffraction Zettajoule = 1021 J Viscosity Energy conversion efficiency Viscosity in a spin-coating process Work function Diameter
Chapter 1
Introduction Frederik C. Krebs Photovoltaic devices convert light energy directly into electrical energy, and the primary objective of their use is the harvesting of light energy from the sun. The photovoltaic devices are silent when they operate and have no mechanical movement associated with their function. They operate under illumination and produce an electrical current that can be directly consumed or stored by chemical (batteries, hydrogen, etc.) or mechanical means (flywheels). The current conversion of power from sunlight into electrical power is not highly efficient; however, improvements are being made all the time. Aside from the very occasional solar eclipse, the sun is a reliable source of energy that the human race can depend on for the next five billion years.
1.1 Human Energy Consumption Now and in the Future The level of energy consumption by the humans on the planet in 2004 was approximately 15 terawatts (TW), with everything included. Most of this energy (87%) was derived from fossil-fuel sources.1, 2 Continuous industrialization of developing countries, growth in human population, and a general increase in human welfare is projected to increase the demands for energy in the future by a large proportion. By the year 2050, the anticipated level of energy consumption by humans is 28– 35 TW, which is a challenge we currently cannot meet with the sources of energy available. The global economy today is based on fossil fuel (coal, oil, and, gas) and, as a resource, is generally accepted to follow the Hubbert peak theory that, in brief, claims the rate of petroleum production follows a bell-shaped curve. The point at which the global demand for petroleum exceeds the rate of production is termed “peak oil.” This point in time is when the global economy is predicted to collapse with more or less disastrous consequences. Depending on who you ask, peak-oil production may already have been reached or is imminent. The Hubbert peak theory is based on the fact that the resources are finite and that the rate of production to a rough approximation follows the rate of discovery with a time lag. For any resource, the rate of discovery is small to begin with followed by an increase to a certain maximum and then finally a decrease.
Chapter 1
2
Table 1.1 Fossil fuel use (in 2004) compared with the reserves.1, 2 The estimated time left is under the assumption that the annual consumption does not increase (1 TW = 1 terawatt = 1012 W, 1 ZJ = 1 zettajoule = 1021 J).
Form
Consumption (TW)
Coal Solid 3.8 Oil Liquid 5.6 Natural gas Gaseous 3.5
Annual consumption (ZJ year−1 ) 0.12 0.18 0.11
Energy reserve (ZJ2 )
Time left (years)
290 57 30
2400 316 272
Note: There is an estimated reserve of 2500 ZJ if using the uranium from the earth’s crust with currently available extraction techniques. This assumes the use of breeder reactors that generate more fissile material than they consume. Fusion is currently not in practical use and is not currently considered as a serious alternative to energy production.
The generally accepted view is that there are still plenty of fossil fuels left (see Table 1.1) and that it may be an overreaction to change our energy consumption pattern toward renewable energy sources, nuclear fission reactors, and nuclear fusion. Nuclear energies will not be discussed in this introduction simply because they are, as it will become apparent, not necessary seen in the light of the wealth of renewable energy sources that are available. However, energy derived from fossil fuels does produce carbon dioxide and the disaster that looms may not be a shortage of fossil fuels but rather global warming as a result of our excessive abuse of the energy derived from fossil sources. In reality, there is no argument; we should urgently change horses and aim for renewable and non-CO2 -emissive sources of energy.
1 2 Renewable Energy Sources The most ironic outcome of the analysis of renewable energy sources3, 4 is that there is a multitude of sources to choose from and the energy available from them is plentiful; from some of the sources unfathomable amounts of energy are available. The renewable energy sources available are largely covered by hydropower, biomass energy, solar energy, wind energy, geothermal energy, and ocean energy. The current use of the various renewable energy sources are presented in Table 1.2. It is interesting to note that about three-quarters of the renewable energy produced is in the form of electrical energy. This is due to the extensive use of hydropower. Some of the renewable energy sources are inherently inferior, such as ocean energy that can be divided into tidal energy and wave energy. The former is due to the gravitational pull of mainly the moon but also the sun, while the latter is derived from wind that is derived from the sun and thus has a low potential due to losses with each conversion. When comparing all these technologies according to the theoretical potential (how much energy there is available), the technical potential (how much we can extract), and the current use, it quickly becomes apparent
Introduction
3
Table 1.2 The current use of renewable energy sources as electricity and heating where applicable in figures of continuous energy consumption (1 GW = 109 J s−1 ).4
Hydropower Biomass energy Solar energy Wind energy Geothermal energy Ocean energy Total
Electricity (GW) 816 44 5.4 59 9.3 0.3 934
Heating (GW) − 220 88 − 28 − 336
Total (GW) 816 264 93.4 59 37.3 0.3 1270
Table 1.3 The renewable energy resources in terawatt (1 TW = 1012 W). Global use in 2004 = 15 TW.
Hydropower Biomass energy Solar energy Wind energy Geothermal energy Ocean energy Total
Current use (TW) 0.816 0.264 0.0934 0.059 0.0373 0.0003 0.934
Technical potential (TW)a 1.6 >7.9 >51 19 158 6c >238
Theoretical potential (TW)a 4.8 92 124000 190(370b ) 4440000 235 >4560000
a
Esti–mates taken from Ref. [4]. Upper estimate. c Tidal power = 3 TW, wave power = 3 TW. b
that solar energy is used little, while there is a massive theoretical and technical potential for its use.3, 4 By comparison, hydropower and biomass are the only two well-exploited renewable energy sources. Wind energy is established and is widely used; however, the technical potential is much larger than its current use. In Table 1.3, the continuously available energy is compared in terms of the current use and the technical and theoretical potential. Table 1.3 shows that solar energy and geothermal energy outweigh all other renewable energy sources by orders of magnitude when comparing the theoretical potentials. In terms of the technical potential that comprises an estimate of our ability to make use of the theoretical potential, solar energy and geothermal energy are still the two renewable sources with the largest potential, but by a smaller margin. After this brief examination of the currently accepted renewable energy sources, it should be clear that solar energy is a renewable energy with a large potential. This is the motivation for many scientists working in this field.
4
Chapter 1
1.3 Important Facts About Energy, Energy Conversion, the Earth, and the Sun The sun is at the center of our solar system and is a type G2 star. It consists of 70% hydrogen, 28% helium, and 2% of all the other elements. While essentially being a cloud of gasses, it is a massive object held together by gravitational forces. The sun has a diameter of 1.39 × 107 m, and with a mass of 1.989 × 1030 kg it accounts for 99.8% of the total mass in our solar system. The temperature and pressure in the center of the sun reach very high levels, with temperatures of 15.6×105 K and pressures of more than 25 × 1010 atmospheres that allow for nuclear fusion processes to take place, thus producing energy. The fusion processes mainly involve hydrogen, which gives helium as a product. At the core of the sun, 7 × 1011 kg of hydrogen is converted into 695 × 109 kg of helium every second, releasing 5 × 109 kg of energy in the form of gamma rays, which is equivalent to 3.86 × 1026 J. The gamma rays make their way to the surface of the sun by absorption and reemission at lower and lower temperatures until they reach the surface mainly as visible light. The temperature at the surface that is what we can observe directly on Earth amounts to 5800 K (while the temperature of sunspots can descend to 3800 K). The distance from the sun to the earth is 149.6 × 108 m and the earth has, with its much smaller mass of 5.98 × 1024 kg, a diameter of 12.756 × 103 m. The energy from the sun is emitted in all directions meaning that the energy density at the distance from the sun where the earth is situated is lower. The energy in space just outside the earth’s atmosphere is 1366 W m−2 , and when the light energy passes the atmosphere some of the visible light energy is lost by absorption in specific regions of the spectrum. This is termed the “air mass” and is abbreviated AM followed by a number (see Fig. 1.1 and Chapter 3 for further details). The absorption loss amounts to 28%, giving about 1000 W m−2 at the surface of the earth under ideal conditions throughout the year. A year has 8760 hours and at any given location on earth half of these are nighttime, giving 4380 hours of daylight averaged over a year. In practical terms, there are many dependencies, such as latitude, earth rotation, and cloud cover, that on the average give 1800 hours of sunlight available for power production every year. In a country such as Denmark, this implies that a solar panel of 1 kWpeak will generate 850 kW-hours of electricity on an annual basis.
1.4 Solar Energy The earlier discussion of our fossil-fuel energy past gives a daunting projection of our energy future. The good news, however, is that we receive plenty of energy from the sun and the challenge is simply to make use of it. The side of the earth that is exposed to the sun receives approximately 1.2 × 105 TW from the sun continuously, which is approximately 10,000 times the energy we consumed in the year 2004. Even with the projected doubling of our energy consumption over the
Introduction
5
Figure 1.1 An illustration of the solar spectrum in space just outside the atmosphere (air mass is negligible; AM 0) and the solar spectrum at the surface of the earth at the northern latitudes of Europe (AM 1.5). The distances in the drawing of the sun, earth, and atmosphere is not to scale.
next 50 years, there is plenty of energy available from the sun alone. The plentitude is such that we can easily accommodate for energy conversion factors well below 100% and we still have to cover only a fraction of the earth’s surface with energy extraction devices such as solar cells. There are obvious logistical problems associated with solar energy because the energy we receive and convert into electricity is lost if it is not used as it is generated. Even for a large energy-consumption system such as the earth, it is useful to buffer or store the energy if there is no immediate need for it when generated. Wind, wave, and photovoltaic solar energy all suffer from the problem of not being energy technologies that inherently allow for storage. Some good examples of energy technologies that intrinsically achieve this are nuclear power, hydropower, biomass, and geothermal power. In the latter case, energy is extracted from the subsurface of the earth and is in principle wasted if not extracted; thus, it does not fully comply with the concept of storage. Hydropower is a very good example of a renewable energy source, where water is collected in a reservoir and emptied from the reservoir through a turbine when needed. Biomass is another good example, where wood can be stored until needed. By today’s standards, biomass energy also accounts for biofuels such as bioethanol and biodiesel.
1.5 The Storage and Relocation Problem Our society has evolved over centuries, and our habit of consuming energy is deeply anchored in our way of thinking, working, and behaving. Most of our methods for extracting energy are naturally centered on the most efficient energy storage materials available such as coal, oil, and natural gas. As seen from the above, this is a fantastic solution since we have access to stored energy in solid, liquid, and gaseous
6
Chapter 1
form. It is easy to relocate the energy material to the consumer by transport through a carrier or a pipeline, and we are thus used to treating energy as something that is available at any given location at any given time. The main reason for this is that the energy is efficiently stored and there are few good alternatives to coal, oil, and gas. Renewable energy sources all suffer from the weakness that they are not all reliably available or storable. With photovoltaic systems, we only have reliable operation when the sun is shining, and while this happens at regular intervals (day and night), the energy output from a photovoltaic is highly dependent on the weather (cloud cover, time of the year in some locations, etc.). Wind energy has similar problems and while it is not limited in operation to hours of sunlight, it does depend heavily on the weather. It is paramount to view renewable energy sources in a large systems perspective. They all have advantages and disadvantages and must be coupled to some means of storage. This could be conversion to a storable chemical such as hydrogen, formic acid, methanol, or methane. As an example, wind energy is widely exploited in a country such as Denmark where 20% of the electricity production is covered by wind energy. This is a large proportion of the national electricity consumption and the nature of the energy source is highly dependent on the weather, meaning that there are periods when the energy that can be produced is not used because the demand is not there. Fortunately, there is a natural gas network in Denmark and if the technology for converting excess electrical energy into methane was available, it would be possible to produce methane from excess wind energy for storage and transmission through the existing gas distribution network.
1.6 Types of Solar Cells The archetypical photovoltaic device5 is a silicon-based PN-junction. Since its first incarnation, the silicon solar cell has taken many forms and is by far the most exploited photovoltaic technology. The monocrystalline silicon solar cells are generally better than the polycrystalline silicon cells with module efficiencies of up to 20% in the case of float zone–grown monocrystalline silicon. In terms of the best reported efficiencies for small-scale laboratory devices, crystalline silicon (24.7% efficiency) is only rivaled by gallium arsenide (GaAs, 25.1% efficiency) for singlejunction devices. Generally, photovoltaics based on the monocrystalline materials are costly and various thin film-technologies exist that offer lower cost both in terms of stock material and thermal budget. Amorphous silicon is perhaps the most well-known technology, and while it suffers from degradation under illumination, the technology offers stable efficiencies of around 5–10%. Other thin-film technologies are cadmium telluride (CdTe) and copper indium-gallium diselenide [Cu(In,Ga)Se2 ], and they both offer very attractive efficiencies that rival those of silicon with best efficiencies of, respectively, 16.5% and 18.9%. Finally, solar cells falling broadly under the heading of third-generation photovoltaics can be divided into dye-sensitized, organic, and polymer solar cells. The dye-sensitized
Introduction
7
Figure 1.2 Exploded view of the organic solar cell described by Tang.7
solar cells are best exemplified by the Grätzel cell, with best reported efficiencies of ∼11%.6 One of the earliest examples of organic solar cells was described by Tang7 in 1986 and consisted of two different small molecule organic compounds sandwiched between a glass substrate with indium oxide and a silver electrode as shown in Fig. 1.2. The organic components copper phthalocyanine (CuPc) and a bis imide derivative of perylene tetracarboxylic acid (PTCA) were robust enough to be evaporated at high temperature and under high vacuum onto the substrate. This device achieved a remarkably high power-conversion efficiency of ∼1% in the conversion of light to electricity. The light comes in through the transparent substrate, continues through the indium-oxide electrode, and is then absorbed in the organic layers. The excited state (exciton) can then give rise to a pair of charge carriers, namely, an electron and a hole. This is facilitated at the interface between the two types of organic materials: a donor (CuPc) and an acceptor (PTCA). The electrodes on each side of the cell then collect the charges. The small molecule organic solar cells today have best reported efficiencies of >5%.8 The solar cells based on small organic molecules are prepared by evaporation of the active material and this is viewed by some as being impractical in a large-scale production due to the requirement for a vacuum step. The preparation of solar cells entirely via solution processing techniques such as coating and printing are highly compatible with soluble conjugated polymers, and this is the most recent type of solar cell. The polymer solar cell has been subject to many advances and today almost all efficient polymer solar cells are based on the concept of a so-called bulk heterojunction (BHJ),9 whereby an acceptor and a donor component are mixed and processed simultaneously into the active film. Today, single-layer BHJ cells give efficiencies of ∼5% (see Ref. [10]), and efficiencies of 6.5% (see Ref. [11]) have been reported for tandem cells whereby polymer solar cells, as a technology, have become competitive with amorphous silicon.
8
Chapter 1
1.7 Current Challenges The use of photovoltaics is increasing dramatically, but in the context of the global energy consumption photovoltaics only account for an insignificant part of the amount of energy that is produced (∼0.036%). Most of the established photovoltaic technologies are stable and exhibit a relatively high power-conversion efficiency in the range of 10–20%, qualifying for electrical energy production for on-grid or local use provided that the land mass is available. However, they all suffer from a very high cost, giving energy prices of a few euros per watt, in 2006 terms. This is believed to be one of the reasons that the use is mainly limited to niche products and applications at remote locations. The general view is that a photovoltaic technology has to fulfill three criteria to reach a large market and provide on-grid electricity production.12, 13 The photovoltaic technology has to be stable, efficient, and of low cost (see Fig. 1.3). Crystalline silicon photovoltaics are very stable, with estimated operational lifetimes in excess of 25 years and module power conversion efficiencies as high as 20%. The cost, however, is too high and this is seen as the main reason for the scarcity of photovoltaic technologies on the energy scene. The special focus area of this book, polymer photovoltaics, holds some promise in this respect because it is a technology that in many ways is complementary to the prototypical silicon-based solar cells. The technology offers intrinsic flexibility, low cost, a low thermal budget, solution processing, and very fast methods for fabrication. The technology succeeds where the inorganic photovoltaic technologies have failed, namely, the cost issue. Conversely, the organic photovoltaics have until recently exhibited low stability and low efficiency, while this is improving and power conversion efficiencies of up to 6.5% and estimated operational lifetimes of many years have been reported.14 The main reasons for pursuing a technology that,
Figure 1.3 The Brabec triangle6 contrasting organic PV and silicon PV. The technologies are in many ways complementary; organic PV succeeds where silicon fails and vice versa. Silicon technologies are becoming available at a lower cost and organic PV is improving in terms of stability and efficiency.
Introduction
9
Figure 1.4 The unification challenge of combining efficiency, stability, and large-scale processing for the same material. The properties have been demonstrated individually and the combination of some stability and some efficiency has been achieved with some success.15
in principle, is inferior to the silicon photovoltaic technology is that it offers low production costs and fast processing, while there are other soft advantages, such as flexibility, light weight, and environmental benefits. Seen in light of the recent advances within the field of organic photovoltaics with respect to stability and power-conversion efficiency, the current challenge is the industrial demonstration of a low-cost organic photovoltaic module with moderate stability and efficiency. The secondary challenges are a higher stability of more than 10 years and efficiencies above 10%. Most notably, the demonstrations of moderate efficiency,11 high stability,14, 15 and large-scale processing16 have not been demonstrated for the same material. However, the fact that isolated studies can reach any of the three goals does hold promise for the possibility of combining all three goals in the same material, and this is the overall current challenge known as the unification challenge as illustrated in Fig. 1.4.
References 1. http://www.eia.doe.gov/pub/international/iealf/table18.xls. 2. http://pubs.usgs.gov/dds/dds-060/ESpt4.html#Table. 3. Simms, A., “It’s time to plug into renewable power,” New Scientist, 183(2454), pp. 18–19 (2004). 4. http://www.ren21.net/globalstatusreport/download/RE_GSR_2006_ Update.pdf
10
Chapter 1
5. Markvart, T., and Castañer, L., Solar Cells: Materials, Manufacture and Operation, Elsevier, Oxford (2005). 6. O’Regan, B., and Grätzel, M., “A low-cost, high-efficiency solar cell based on dye-sensitized colloidal TiO2 films,” Nature, 353, pp. 737–740 (1991). 7. Tang, C.W., “2-Layer organic photovoltaic,” Appl. Phys. Lett., 48, pp. 183– 185 (1986). 8. Xue, J., Rand, B.P., Uchida, S., and Forrest, S.R., “A hybrid planar-mixed molecular heterojunction photovoltaic cell,” Adv. Mater., 17, pp. 66–70 (2005); Xue, J., Uchida, S., Rand, B.P., and Forrest, S.R., “Asymmetric tandem organic photovoltaic cells with hybrid planar-mixed molecular heterojunctions,” Appl. Phys. Lett., 85, pp. 5757–5759 (2004). 9. Yu, G., Gao, J., Hummelen, J.C., Wudl, F., and Heeger, A.J., “Polymer photovoltaic cells – enhanced efficiencies via a network of internal donor-acceptor heterojunctions,” Science, 270, pp. 1789–1791 (1995). 10. Li, G., Shrotriya, V., Huang, J., Yao, Y., Moriarty, T., Emery, K., and Yang, Y., “High-efficiency solution processable polymer photovoltaic cells by selforganization of polymer blends,” Nature Mater., 4, pp. 864–868 (2005); Ma, W., Yang, C., Gong, X., Lee, K., and Heeger, A.J., “Thermally stable, efficient polymer solar cells with nanoscale control of the interpenetrating network morphology,” Adv. Funct. Mater., 15, pp. 1617–1622 (2005). 11. Kim, J.Y., Lee, K., Coates, N.E., Moses, D., Nguyen, T.Q., Dante, M., and Heeger, A.J., “Efficient tandem polymer solar cells fabricated by all-solution processing,” Science, 317, pp. 222–225 (2007). 12. Brabec, C.J., “Organic photovoltaics: technology and market,” Sol. Energy Mater. Sol. Cells, 83, pp. 273–292 (2004). 13. Krebs, F.C., “Alternative PV,” Refocus, 6, pp. 38–39 (2005). 14. Krebs, F.C., and Spanggaard, H., “Significant improvement of polymer solar cell stability,” Chem. Mater., 17, pp. 5235–5237 (2005). 15. Yang, X., Loos, J., Veenstra, S.C., Verhees, W.J.H., Wienk, M.M., Kroon, J.M., Michels, M.A.J., and Janssen, R.A.J., “Nanoscale morphology of highperformance polymer solar cells,” Nano Lett., 5, pp. 579–583 (2005). 16. Krebs, F.C., Spanggaard, H., Kjær, T., Biancardo, M., and Alstrup, J., “Large area plastic solar cell modules,” Mater. Sci. Eng. B, 138, pp. 106–111 (2007).
Chapter 2
The Polymer Solar Cell Mikkel Jørgensen, Eva Bundgaard, Rémi de Bettignies and Frederik C. Krebs 2.1 Introduction The first half of this chapter briefly describes the conceptual buildup of a polymer solar cell followed by an overview of the organic materials that have been used. The second half is devoted to a practical guide to making solar cells complete with synthesis, device fabrication, and testing. Today, most polymer solar cells are based on the bulk heterojunction concept described in 1995 by Yu et al.1 In this type of solar cell, the donor material (typically a polymer) is mixed with an acceptor (a soluble fullerene) in an organic solvent and then spin coated or cast on a substrate of indium-tin oxide (ITO) on glass. During evaporation of the solvent and latter treatments, a microphase separation takes place with the formation of an interpenetrating network. A hole-blocking layer of, e.g., lithium fluoride may be added, and in a last step a metal electrode (aluminum) is evaporated on top (Fig. 2.1). The bulk heterojunction is important because a large interfacial area between the donor and acceptor materials is created where charge separation can take place.
Figure 2.1 Exploded view of a bulk heterojunction type of polymer solar cell.
12
Chapter 2
Exciton generation on excitation of the mixture is generally short-lived, with a diffusion length on the order of 10 nm. The size of the individual domains is therefore critical. It is also important that the domains are interconnected so that continuous paths exist for both electrons and holes for transportation to the external electrodes. The advantages of the BHJ-type cell have increased the efficiency to about 5–6%. Unfortunately, the optimal structure for generating the most efficient device may not be the most thermodynamically stable configuration. Much effort has therefore been devoted to creating technical recipes for optimizing the best devices. A large number of different polymers and small molecules have been tested as active materials in organic solar cells, as described in the overview below. Only very few of these new materials have been tested by more than one research group and most of them failed to produce devices with efficiencies of even 1%. As a consequence, nearly all physical studies of polymer solar cells have been conducted with only a few popular polymers. Experience shows that the efficiency can be increased by adjusting fabrication parameters. This requires, however, that sufficient quantities be available. Development of new materials is still very important and among the recent trends are low bandgap polymers that extend the harvesting of photons above 600 nm. Another is the advent of thermocleavable polymers that combine processability with high stability of the finished device.
2.2 Materials 2.2.1 Polymers Most of the common conjugated polymers have been tested as active materials for solar cells, but two classes have attained a special status. The poly(phenylenevinylene)s (PPVs) exemplified by poly[2-methoxy-5-(2-ethylhexyloxy)-1,4-phenylenevinylene] (MEH-PPV) and poly(2-methoxy-5-(3,7-dimethyloctyloxy)-1,4phenylenevinylene) (MDMO-PPV) have a conjugated PPV core, as shown in Fig. 2.2. Early examples of polymer-based solar cells include polyacetylene with an efficiency of 0.3%.2 Karg et al. were the first to investigate PPV in a photovoltaic
Figure 2.2 Chemical structures of the most commonly used conjugated organic polymers, namely, MEH-PPV, MDMO-PPV, and poly(3-hexylthiophene) (P3HT).
The Polymer Solar Cell
13
Figure 2.3 Chemical structures of PEDOT and PSS.
device in 1993.3 PPV in itself is a very intractable material due to its insolubility. Adding alkyl or alkoxy chains on the phenylene rings, as in MEH- and MDMOPPV, makes these materials processable and soluble in some organic solvents such as chloroform, chlorobenzene, or 1,2-dichlorobenzene (ODCB). Yu et al. reported the use of MEH-PPV in a dual function light-emitting diode (LED) photovoltaic device also in 1994.4 Higher efficiency devices were prepared from MDMO-PPV/ [60]PCBM blends reaching 2.5% in the so-called bulk heterojunction geometry.5 More recently, highly regioregular P3HT has come to be a material of choice. By tuning several parameters such as annealing, solvent, and film formation, efficiencies of approximately 5% have been achieved recently.6, 7 Using a tandem geometry with different polymers has extended the efficiency even to 6.5%.8 Another polythiophene material that is used extensively in organic solar cells is PEDOT:PSS or poly(3,4-ethylenedioxythiophene) poly(styrenesulfonate) (Fig. 2.3). A thin film of PEDOT:PSS is usually applied as a hole conducting material directly on top of the ITO electrode by spin coating the commercially available dispersion in water followed by removal of the water by heating. 2.2.1.1 Poly(phenylenevinylenes)
PPVs have the general structure of phenylene rings (aromatic rings) joined by ethylene bridges (Fig. 2.4). PPV and substituted versions have been prepared in numerous ways, as reviewed earlier.9 One of these methods is the sulfonium precursor route,10 where α, α -dichloro-p-xylene is treated with tetrahydrothiophene to give a bis-sulfonium salt (Scheme 2.1). Treatment of this salt with sodium hydroxide generates a quinoid intermediate that polymerizes. The remaining sulfonium groups are eliminated in a later heat treatment generating the PPV structure. This procedure has the advantage that the intermediate polymer salt is soluble and can be processed into films, while PPV itself is intractable. Karg et al. prepared photovoltaic cells from PPV with an ITO/PPV/Al structure and measured maximum monochromatic light conversion efficiencies of about 0.1% (at 514 nm).3 Marks et al. also described the use of PPVs in solar cells with different metal top layers (Al, Mg, Ca) and reported open-circuit voltages of 1.2, 1.2, and 1.7, respectively.11 Soon after, mixtures of PPV with the electron accepting C60 fullerene were tested in solar cells.
14
Chapter 2
Figure 2.4 General chemical structure of poly(phenylenevinyles) (PPVs). When R = H, it is the archetypal PPV structure. More commonly, one or more of the R groups represent flexible alkyl or alkoxy groups that aid solubility.
Scheme 2.1 Synthesis of PPV using the sulfonium precursor (Wessling) method.
Alkyl group- or alkoxy group-substituted PPVs allowed solution processing and made these materials more suitable for device preparation, such as for organic light-emitting diodes (OLEDs). The main focus has been on finding materials with emission in different regions of the optical spectrum. PPV itself emits green light, while alkoxy-substituted versions emit in the orange part of the spectrum. The major absorption band of PPVs is in the range of 400–500 nm with a bandgap of 2.1–2.7 eV (electron volts). MEH- and MDMO-PPV can be prepared using several types of reactions that all seem to be based on creating a reactive intermediate, a quinodimethane. A common starting material is hydroquinone monomethyl ether, which can be alkylated with a branched alkane such as 2-ethyl-hexyl-1-bromide or 2,7-dimethyl-octyl-1bromide. Chloromethylation can be carried out in both the 2- and 5-positions to obtain the bis-chloromethyl-dialkoxy monomer that can be utilized directly in the Gilch reaction.12 There is still some discussion about the actual mechanism of the reaction13, 14 shown in Scheme 2.2. While these polymers perform well in certain applications such as PLEDs (polymer light-emitting diodes), a number of defects (up to several mol %) are present due to incomplete reactions and side reactions. Becker et al. made a detailed investigation of the defects using nuclear magnetic resonance (NMR) to characterize them and found the presence of both fully saturated ethylene bridges and
The Polymer Solar Cell
15
Scheme 2.2 Gilch polymerization producing MEH-PPV and MDMO-PPV.
triple bonds to a level of about 2%. The conjugation, of course, breaks at many of these defects, influencing the electronic properties.15 Yu et al. were among the first to utilize MEH-PPV to prepare photovoltaic devices with the ITO/MEH-PPV/Ca geometry with an open-circuit voltage (Voc ) of 1.6 V and short-circuit current (Isc ) of 6 μA/cm2 at 20 mW/cm2 light intensity.4 A year later, in 1995, Yu et al. reported a new and more efficient cell with a blend of MEH-PPV and [6,6]-phenyl C 61-butyric acid methyl ester, or [60]PCBM, in a so-called bulk heterojunction geometry.1 Compared with previous bilayer devices, heterojunction devices have a microphase separated interpenetrating network of polymer and acceptor domains. The internal distance to a phase boundary is on the order of a few nanometers and the chance of successful charge carrier generation was improved greatly, leading to a better energy conversion efficiency (ηe ) of 2.9%. Wienk et al. have increased the efficiency even further by optimizing the devices and exchanging [60]PCBM with [70]PCBM (phenyl-C71 -butyric acid methyl ester).16 Alkyl substituted PPVs have been synthesized using a number of routes that generate the vinylene groups. These are usually condensation reactions, such as the Horner-Wadsworth-Emmons (HWE) reaction between an aromatic aldehyde and an aryl methyl phosphonate ester (Scheme 2.3). Alternatively, a Heck-type coupling reaction between an aryl halide and a styrene type monomer may be utilized. Such alkyl substituted PPVs have been shown to exhibit very high carrier mobility.17 The structure of the alkyl substituted PPVs were also investigated by x-ray powder diffraction and it was found that the relatively high carrier mobility could be explained by the good overlap between polymer chains (Fig. 2.5). Although many other variants of PPV structures have been made, few have been applied to photovoltaic devices. One instance is the use of a cyano substituted MEH-PPV (Fig. 2.6) by Granström et al.18 in a laminated bilayer structure with a polythiophene, obtaining an overall power conversion efficiency of 1.9%. CN-
Chapter 2
16
Scheme 2.3 Two examples of PPV syntheses that involve two different monomers resulting in the same polymer. The reaction on top is a classical phosphonate-ester condensation, while the bottom one is a palladium catalyzed Heck reaction.
Figure 2.5 Stereo drawing of an x-ray structure of a poly[2,5-dioctyl-(1,4-phenylene)vinylene-1,4-phenylene-vinylene] type PPV. (Reprinted with permission from Ref. [17]. Copyright 2003, American Chemical Society.)
Figure 2.6 Chemical structure of CN-MEH-PPV.
The Polymer Solar Cell
17
MEH-PPV had previously been prepared by Greenham et al. for use in LEDs.19 The electron attracting cyano groups change the material to be an acceptor compared to MEH-PPV. Organic materials with a high electron mobility are rare and the fullerenes must still be considered unrivaled in this respect. 2.2.1.2 Poly(thiophenes)
Polythiophene is, like PPV, an intractable material, although it can easily be produced by chemical or electrochemical oxidation of thiophene. Side groups of various types are therefore necessary for obtaining a usable polymer that can be cast or spin coated. The 2- and 5-positions of thiophene are the most reactive and are preferred for polymerization. 3-alkyl thiophenes can be polymerized by oxidation with, for example, ferric chloride. The initial cation radical reacts with other thiophene molecules and protons are eliminated to regain conjugation. This material is a regiorandom polymer with thiophene units joined at the 2- or 5-positions (head to head, head to tail, and tail to tail) (Fig. 2.7). In the head-to-head type dimer, the alkyl groups force the thiophenes to adopt an out-of-plane geometry, reducing the electronic overlap and limiting the intermolecular interactions. Regioregular P3HT with almost exclusive head-to-tail geometry can be prepared via two related routes. In the McCullough route,20 3-bromothiophene is alkylated with hexyl magnesium bromide in a Kumada coupling, followed by bromination in acetic acid with elemental bromine at the 2-position. The pure 2-bromo-3-hexylthiophene can then be polymerized in one pot by lithiation at the 5-position using LDA conversion to the magnesium derivative with MgBr2 · Et2 O, which is then treated with a Ni(dppp)Cl2 catalyst, where dppp is 1,3-bis(diphenylphosphino)propane (Scheme 2.4). Another procedure for the synthesis of regioregular P3HT is based on the activation of 2,5-dibromo-3-hexylthiophene with so-called Rieke-zinc. As in the McCullough route, the metal is inserted in the 5-position creating a nucleophilic species. At a low temperature (−78◦ C) and for the 3-hexyl substituted thiophene, the selectivity in this reaction was found to be 97:3. The intermediary is then polymerized with the Ni(dppp)Cl2 catalyst to create regioregular material with a molecular weight (mol wt) of 15.000.21 The UV-visible (UV-vis) spectra of the regioregular material has a maximum of 450 nm, which is red shifted some 25 nm relative to the random material. Also, the
Figure 2.7 Head-to-head, head-to-tail, and tail-to tail geometries in part of a random 3-alkyl polythiophene chain.
18
Chapter 2
Scheme 2.4 Regiorandom P3HT is produced by ferric chloride polymerization (top). The McCollough route (middle) and the Rieke route to highly regioregular P3HT (bottom).
emission spectra differ with a maximum of 570 nm for the regular versus 550 nm for the random P3HT (see later in this chapter). The electronic spectra in the solid state show a pronounced red shift of the lowest absorption from about 440 nm to more than 500 nm due to close intermolecular interactions. This is only possible for the regioregular P3HT, which forms a well-ordered structure with interdigitating side chains (Fig. 2.8). Several studies have shown that the structure of heat-treated (annealed) regioregular P3HT film is semicrystalline. Diffraction techniques, such as grazingincidence x-ray diffraction (XRD), show peaks that are interpreted as a lamellae structure with an alternating orientation of the thiophene moieties, where the alkyl groups of neighboring polymer chains interdigitate.22, 23 A regular surface structure can be observed with atomic force microscopy (AFM), showing needlelike structures in an amorphous matrix. P3HT, like all the other condensation polymers, have a broad distribution of molecular masses where the smaller masses might act like glue between crystalline domains. Other types of polythiophenes (see Fig. 2.9) that have been used in solar cells include poly[3-(4-octylphenyl)-2,2 -bithiophene] (PTOPT),24 poly[3-(2 -methoxy5 -octylphenyl)-thiophene] (POMeOPT),25 and poly[3-(4 -(1 ,4 ),7 -trioxaoctyl) phenyl)thiophene] (PEOPT).26
The Polymer Solar Cell
19
Figure 2.8 Schematic representation of the crystaline part of P3HT.
Figure 2.9 Some other polythiophenes used in organic solar cells.
2.2.1.3 Low bandgap polymers
Low bandgap polymers are defined as polymers absorbing light with wavelengths above 600 nm. The traditional polymers used in organic photovoltaics, such as MEH-PPV, have an absorption that extend to wavelengths of 550 nm. Commonly employed P3HT has an absorption that extends out to 650 nm and comparing the absorption spectra of this polymer with the solar spectrum (Fig. 2.10), a strong mismatch between the absorption spectrum of P3HT and the emission spectrum of the sun is seen.27
20
Chapter 2
Figure 2.10 Absorption spectra of P3HT and a low bandgap polymer and the solar spectrum (AM 1.5D).
This mismatch could be alleviated if a low bandgap polymer was used, that is, the absorption spectrum of the low bandgap polymer has a better overlap with the sun emission spectra. Therefore, in the past decade the focus on low bandgap polymers have increased tremendously. But why are low bandgap polymers important? And furthermore, how can we obtain the low bandgap? Before answering these questions, we take a look at the definitions of a bandgap. The bandgap of a polymer is defined as the difference between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO). The bandgap can be determined from an optical absorption spectrum (UV-vis), which gives the optical bandgap, or by cyclic voltammetry (CV), which gives the electrical bandgap. The bandgap is in principle used for the determination of the energy that can be extracted from the cell. However, neither the optical nor the electrical bandgaps give this information directly. In the optical bandgap, the binding energy of the exciton is not accounted for and in the CV experiment the energy of solvation of the electrochemical species is unknown. The optical bandgap gives a somewhat larger value. To answer the question on why we need low bandgap polymers, we take a closer look at the solar spectrum. In Fig. 2.11, the AM 1.5G spectrum of the sun is shown together with the number of photons as a function of wavelength. The spectrum clearly shows that the maximum number of photons is at wavelengths up to 900 nm. However, when we look at the integrated number of photons and integrated current as a function of the wavelength (Fig. 2.12) some important results are found. We compare two examples of polymers, that is, one absorbing light with wavelengths up to 500 nm (A) and one absorbing light with wavelengths up to 1000 nm (B). Figure 2.12 shows that polymer A will absorb 9.4% of the photons in the solar spectrum. Polymer B, however, will absorb 55.1% of the photons. By converting to the maximum theoretical current, this corresponds to Jsc = 5.1 mA cm−2 for A and Jsc = 33.9 mA cm−2 for polymer B, that is, lowering
The Polymer Solar Cell
21
Figure 2.11 Sun irradiance spectrum AM 1.5G (black) and the number of photons (red) as a function of wavelength.
Figure 2.12 Number of photons for AM 1.5G (black), integrated photons (in percent from 280 to 4000 nm), and integrated current (in mA cm−2 ) as a function of wavelength (red).
the bandgap of the polymer results in a higher theoretical current. More values are given in Table 2.1. The values given for the maximum theoretical current is calculated from the assumption that the polymer absorbs all the photons from 280 to the wavelength given. The current measured depends on the absorption of the device, which includes absorption in the material, but also reflection losses from the window and interfaces. As an example, we have plotted the absorption spectrum of a thin film of a low bandgap polymer with the photon flux and performed the same calculations (Fig. 2.13). As can be seen from Fig. 2.13, the absorption spectrum of the polymer does not correspond completely with the photon flux, and hence the maximum theoretical current is around 14 mA cm−2 and not around 20 mA cm−2 , which is indicated in
Chapter 2
22
Table 2.1 for a polymer absorbing light with wavelengths up to 750 nm. The actual current measured for an OPV depends on several factors, such as morphology, and hence, to obtain a more precise calculation the incident photon to current efficiency (IPCE) should be considered, since this is a device measurement and factors like morphology, thickness, carrier mobility, carrier lifetime, and reflection losses have thus been taken into account.
Table 2.1 The integrated photon flux and maximal theoretical current for an OPV with a polymer absorbing all wavelengths from 280 nm to the wavelength given assuming every photon is converted to an electron in the external circuit. The maximum current may increase in bulk heterojunctions due to the absorption of PCBM (the values are for direct and circumsolar AM 1.5D and AM 1.5G in brackets).
Wavelength 500 600 650 700 750 800 900 1000 1250 1500
Max. % harvested (280 nm →) 8.0 (9.4) 17.3 (19.0) 22.4 (24.3) 27.6 (29.6) 32.6 (34.7) 37.3 (39.5) 46.7 (48.8) 53.0 (55.0) 68.7 (70.4) 75.0 (76.4)
Current density in mA cm−2 (AM1.5G) 5.1 (6.47) 11.1 (13.15) 14.3 (16.77) 17.6 (20.42) 20.8 (23.9) 23.8 (27.23) 29.8 (33.67) 33.9 (38.0) 43.9 (48.57) 47.9 (52.75)
Figure 2.13 Number of photons (AM 1.5G, black), integrated photon flux (blue), current (blue) as a function of wavelength, and an absorption spectrum of a low bandgap polymer (red).
The Polymer Solar Cell
23
Above we have shown that low bandgap polymers are important to increase the short-circuit current and hence the efficiency of the OPV due to a better overlap with the solar spectrum. Another important factor that influences the efficiency of the device is the open-circuit voltage. In Fig. 2.14 it can be seen that maximum voltage obtainable decreases as a function of wavelength (as calculated from the optical bandgap). The optimum bandgap is in the region of 0.9–1.2 eV. From this crude approximation, the maximum power conversion efficiency is 44% (in reality, thermodynamical factors have to be taken into account). The maximum Voc that can be obtained from a device based on donor and acceptor components in a bulk heterojunction geometry is roughly the difference between the HOMO of the donor and the LUMO of the acceptor. The energy level alignment between the polymer, the electron acceptor, and the electrodes becomes very crucial when low bandgap polymers are used in OPV devices. In Fig. 2.15, the energy level alignment in a bulk heterojunction between a low bandgap polymer and PCBM is shown. The energies β and ΔE have to be high enough to ensure a high Voc . A decrease in the LUMO of the polymer and an increase in the LUMO of the acceptor, that is, a decrease in α, will cause an increase in the maximum Voc that can be obtained and a decrease in the bandgap assuming a better overlap with the solar spectrum; hence, a higher Isc as described above. However, α has to be high enough to ensure an efficient charge transfer from the electron donor (polymer) to the electron acceptor (most commonly PCBM). These factors are of great importance when a low bandgap polymer is applied and often it becomes difficult to align the energy levels when the LUMO of the acceptor is fixed. Therefore, when low
Figure 2.14 Maximum power obtainable in OPVs based on the AM 1.5G emission spectrum. The power is taken as the product of the integrated current assuming an IPCE of 100% and the voltage of the device as the value for the bandgap. Thus, the power is the maximum theoretical value, neglecting thermodynamic effects and losses. The pink box is the range of bandgaps where the most efficient devices can be found.
Chapter 2
24
Figure 2.15 Energy level alignment in a bulk heterojunction based on a low bandgap polymer and PCBM. The sum of ΔE and β represents the maximum voltage that can be obtained (Voc ).
Figure 2.16 Resonance forms of a fused ring system, here PITN.35
bandgap polymers are applied in OPV devices, the choice of acceptor may have to be reconsidered to ensure a good energy level alignment.27, 28 We will now focus on how to design these important low bandgap polymer materials. There are several factors that influence the bandgap of the polymer, for example, conjugation length, bond length alternation, intrachain charge transfer, intermolecular interactions, aromaticity, and substituents. The effect of an increased conjugation length, for example, is seen for poly(3alkylthiophenes), where the regioregularity of the alkyl groups are of great importance. In the randomly coupled poly(3-alkylthiophene), the torsion between the side groups causes the backbone to twist, reducing the effective conjugation length and thus increases the band.29 Planarity along the aromatic polymer backbone increases delocalization of the π-electrons, reducing the bandgap.30 This can be further enhanced by intermolecular interactions observed for P3HT and other poly(3-alkylthiophenes) when comparing the absorption spectra of solid and liquid phases of the polymer.31–34 The low bandgap polymers described in the literature are based on fused ring systems or are copolymers with alternating donor and acceptor groups. Fused ring systems, such as those shown in Fig. 2.16,35 enhance the quinoid resonance structure, which in turn reduces the bond alternation.36–41 Finally, examples in the literature demonstrate the effect of additional electron donating groups (EDGs) and electron withdrawing groups (EWGs) to the polymer backbone.42–55 For example, in a bithiophene repeating unit, one thiophene has
The Polymer Solar Cell
25
Figure 2.17 Examples of low bandgap polymers. Fused ring systems (PITN), and copolymers based on benzothiadiazole, pyrrole, and thiophene (PBPT); benzothiadiazole and thiophene (PBT); benzo-bis(thiadiazole) and thiophene (PBBT); thienopyrazine and thiophene (PTP); thienopyrazine, thiophene, and fluorene (PTTF); and benzothiadiazole, thiophene, and fluorene (PBBTF) R = alkyl, phenyl or chlorine.
an EDG and one has a EWG. This results in a donor/acceptor-based copolymer.56 In Fig. 2.17, some examples of low bandgap polymers that are described in the literature are shown. These polymers can be divided into two categories, namely, (1) a fused ring HOMO polymer, for example, poly(isothianaphthene), PITN, and (2) copolymers based on donor and acceptor majorities, for example, the polymer based on benzothiadiazole and thiophene, PBT. Below is a short summery of these low bandgap polymers and the results obtained from OPV devices prepared with some of these low bandgap polymers. First, a general summary of the polymerization methods used to prepare these polymers. The low bandgap polymers shown in Fig. 2.17 have been synthesized by different methods as described below. The most important ones are (1) Stille cross-coupling polymerization, (2) Suzuki cross-coupling polymerization, (3) Yamamoto condensation polymerization, (4) oxidative ferric chloride polymerization,
Synthesis
26
Chapter 2
Figure 2.18 Synthesis methods for low bandgap polymers: (1) Stille, (2) Suzuki, (3) Yamamoto, (4) oxidative ferric, and (5) electrochemical polymerizations.
and (5) electrochemical polymerization. These reactions are shown in a general form in Fig. 2.18. In Stille cross-coupling polymerization, a dibromo derivate and a distannyl derivate are coupled using a palladium catalyst. In Suzuki cross-coupling polymerization, a bis-boronic acid monomer and dibromo monomer are coupled, also using a palladium catalyst. In both Stille and Suzuki coupling, it is possible to synthesize polymers with two different monomers forming a donor/acceptor copolymer. Unfortunately, in these polymerization methods there are a number of side reactions such as HOMO coupling of the stannyl compounds, dehalogenation, destannylation, etc. This may result in low molecular weights of the polymer products.57 Yamamoto condensation polymerization is a reaction between dibromo monomers using a nickel catalyst. In oxidative ferric chloride polymerization, a monomer is polymerized using FeCl3 as a reagent and results in high molecular weights. Finally, in electrochemical polymerization, a potential is applied on a solution containing the monomer and a buffer. This results in formation of the polymer product at the anode. PITN Poly(isothianaphthene) (PITN) is prepared from a monomer with a benzene fused to a thiophene ring. The fused ring system causes the bandgap to decrease
The Polymer Solar Cell
27
Figure 2.19 Derivates of ITN-based polymers. (a) Poly(naphthothiophene) and (b) copolymer of thiophene and isothianaphthene p(T-ITN-T), R = alkyl or chlorine.
due to the quinoid resonance structure (Fig. 2.16).30 The bandgap of PITN was found to be 1.0 eV.58 Addition of another fused benzene ring (Fig. 2.19) caused an increase in the electron delocalization over the aromatic system, and hence a suppression of the quinoid structure and an increase in the bandgap was observed.59 The isothianaphthene (ITN) unit has also been applied in a copolymer based on donor/acceptor systems where the ITN unit is the electron acceptor and a thiophene unit is the electron donor (Fig. 2.19).60, 61 The bandgap of this copolymer was found to be 1.5 eV.60 The synthesis of PITN was first described in 1984 and was carried out by a electrochemical polymerization. Since then, the synthesis has been carried out using P4 S10 (see Refs. [62] and [63]) and by other methods.64–66 PBPT PBPT is a copolymer based on benzothiadiazole, pyrrole, and thiophene, where thiophene and pyrrole are the electron-donating units and benzothiadiazole is the electron-accepting unit. Benzothiadiazole is a fused ring system where a thiadiazole ring is fused to a benzene ring. The polymer has been synthesized by a Stille cross-coupling reaction; however, this results in low molecular weight products and the Suzuki cross-coupling polymerization was therefore used, which resulted in higher molecular weight products.37, 57, 67 The bandgap of PBPT was found to be 1.6 eV for the polymer with the highest molecular weight.67, 68 PBT PBT is also a copolymer where thiophene corresponds to the electron-donating unit and benzothiadiazole corresponds to the electron-accepting unit. The synthesis of these types of polymers have been carried out by Stille cross-coupling, by oxidative ferric chloride, or by Yamamoto coupling polymerization.69–72 Stille cross-coupling polymerization resulted in four different polymer products with a variation in the number of thiophene units between the benzothiadiazole unit (n = 1–4). The building blocks for Stille cross-coupling polymerization are shown in Fig. 2.20.69 The bandgap was shown to decrease with an increase in number of thiophenes in the repeating unit from 2.1 to 1.65 eV for the polymer with one and four thio-
28
Chapter 2
Figure 2.20 Building blocks for Stille cross-coupling polymerization of PBT. R = alkyl.
phenes in the repeating unit, respectively.69 This is ascribed to the optimization of the donor unit with respect to the acceptor unit (benzothiadiazole). It also means that there is an optimal length of the thiophene segment for one benzothiadiazole acceptor unit, and thus increasing the number of thiophenes beyond the optimum may cause an increase in the bandgap, which eventually will approach the bandgap of polythiophene.69 Studies of side groups showed that the solubility and film forming ability of the polymer was highest when 3,7,11-trimethyldodecyl was applied as a side chain.69 The position of the side chain also showed an effect on the bandgap, that is, steric hindrance of the side chain in the 3-position on thiophene, which results in an increase of the bandgap.37 PBBT PBBT is similar to PBT, having the thiophene unit as the donor and a benzo-bis(thiadiazole) unit as the acceptor. Benzo-bis(thiadiazole) has two thiadiazole rings that are fused to a benzene ring. The synthesis of PBBT was also carried out by Stille cross-coupling using the three building blocks shown in Fig. 2.21, creating two different polymer products, one with three and one with four thiophenes in the repeating unit.69 The bandgap of the two polymers was found to be 0.65 and 0.67 eV, respectiv69 ely. The low bandgap of this type of copolymer is ascribed to the quinoid form of the polymer, which is stabilized by the 1,2,5-thiazole unit (Fig. 2.22).73 It has been found that the benzo-bis(thiadiazole) has the highest electron-accepting ability among other acceptor units, for example, thienopyrazine and benzothiadiazole; hence, polymers based on benzo-bis(thiadiazole) also show the lowest bandgap.36 PTP PTP is another example of a copolymer where thiophene is the electron donor unit, and here thienopyrazine is the electron acceptor. The thienopyrazine is also a fused ring system where a pyrazine ring is fused to a thiophene ring. The synthesis of PTP was carried out by electrochemical, oxidative ferric chloride, or Yamamoto coupling polymerization.36, 74–76 The bandgap of PTP was found to be around 1.3 eV; however, a bandgap down to 1.20 eV was found for a more coplanar structure without bulky side chains on the backbone.74, 75
The Polymer Solar Cell
29
Figure 2.21 Building blocks for Stille cross-coupling polymerization of PBBT. R = alkyl.
Figure 2.22 Resonance forms in the PBBT. Table 2.2 Bandgaps of the copolymers based on fluorene shown in Fig. 2.23.
Polymer 1 2 3 4 5 6
Band gap (eV) 2.01 ∼1.9 1.27 2 1.4 1.3
λmax (nm) 530 780 780 740
Ref. [77,78] [79] [30,80,81] [82] [83] [57]
PF Several copolymers based on fluorene have been described in the literature (Fig. 2.23). In the copolymers, the fluorene and thiophene unit functions as the electron-donating unit and these are coupled with varying electron accepting groups such as benzothiadiazole (1 and 2), thienopyrazine (4), and thiadiazolequinoxaline (3, 5, and 6). The side chains on the polymer backbone was varied to ensure solubility and to vary the bandgap of the polymers. The bandgap of the polymers is summarized in Table 2.2. The polymers were synthesized by Suzuki cross-coupling reaction.36, 56, 77, 78 Photovoltaic devices Some of the low bandgap polymers described above have been applied in solar cell devices and the best reported results are summarized in Table 2.3. As can be seen from this table, the efficiency of the devices with the low bandgap polymers is small compared with the efficiency of P3HT.7 The measured Isc is several times lower than the maximum theoretical current (Table 2.1). The choice of PCBM as an electron acceptor in these devices might ex-
Chapter 2
30
Figure 2.23 Copolymers based on fluorene. R and R1 = alkyl.
plain the low Voc caused by a poor overlap between the HOMO of the polymer and the LUMO of the acceptor. For reviews on low bandgap polymers see Refs. [29] and [89]. 2.2.1.4 Other polymers
A number of less-used polymers have been studied as active materials for solar cells. They may not perform particularly efficiently in photovoltaic devices, but are included to show the scope and because some interesting and alternative principles have been used. Henckens et al. exchanged phenyl groups with thiophene in the PPV type polymer to obtain PTVs or mixed structures.84 Smith et al. prepared regioregular poly(3-dodecyl-2,5-thienylene-vinylene) by a Stille type cross-coupling between 3-dodecyl-2,5-dibromothiophene and (E)-1,2-bis(tributylstannyl)ethylene, as shown in Scheme 2.5.90
The Polymer Solar Cell
31
Table 2.3 Photovoltaic responses reported for OPV devices with low band gap shown in Fig. 2.17. The device structure was ITO/PEDOT:PSS/polymer:PCBM/Al.
Polymer PITN P(T-ITN-T) PBPT PBT PTT 1 3a
Isc (mA cm−2 ) 0.045 1.13 3.1 3.59 3.5 4.66 3.4
Voc (V) 0.55 0.88 0.72 0.61 0.56 1.04 0.58
FF (%) 30 25 37 46 58 46 35
ηe (%) 0.008 0.31 1.0 1.0 1.1 2.2 0.7
4 5a
3.0 2.4
0.78 0.61
– 40
0.9 0.59
a
IPCE (%) – 20/480 20/550 18/600 10/800 40/550 8.8/850 7/900 10/650 8/800
Ref. [84] [61] [85,86] [87] [74] [77] [88] [80,82] [80,82]
The devices was prepared with BPF70 in stead of PCBM.
Scheme 2.5 Synthesis of a poly(thienylene-vinylene).
This PTV blended with PCBM in a bulk heterojunction type solar cell had an efficiency of 0.24%, a short-circuit current of 0.8 A cm−2 , and an open-circuit voltage of 0.54 V at AM 1.5G conditions. Branched PTV type polymers have been prepared by Li et al., who found similar values for the efficiencies of the solar cell devices.49 A special type of organometallic polymer with alternating platinum alkyne bridges and bithiazole units has been prepared by Wong et al.91 (Fig. 2.24). This polymer had an absorption maximum at 460 nm and a bandgap of 2.35 eV in the solid state. A very low photo current was generated and an efficiency of ∼10−5 %. Diblock copolymers such as polystyrene-polydimethylsiloxane form a number of microphase-separated structures depending on the size of each of the blocks. This was utilized by de Boer et al. to prepare interesting examples of a selfassembling PPV C60 conjugates.92 These conjugates were prepared by first synthesizing a PPV domain using the Siegrist reaction on (4-methyl-2,5-bis-octyloxybenzylidene)-phenyl-amine. NMR indicated an approximate length of 10 units. The aldehyde group at the end of this oligomer was then used to transform the PPV into a macroinitiator by reaction with a 2,2,6,6,-tetramethylpiperidin-1-oxyl (TEMPO) derivative. Atom-transfer radical polymerization (ATRP) with a mix-
Chapter 2
32
Figure 2.24 A platinum alkyne bithiadiazole polymer.
Figure 2.25 A diblock copolymer of PPV and polystyrene/C60 (PPV-b-P(S-stat -C60 MS).
ture of styrene and 4-chloromethyl-styrene then gave a diblock copolymer with the PPV block and a polystyrene block. Some of the styrene units were functionalized with chloromethyl groups that were used to react with C60 forming the PPV-b-P(S-stat-C60 MS) diblock (Fig. 2.25). A similar diblock copolymer without C60 was investigated with SEM and showed a honeycomb structured film indicating the expected microphase separation. The photovoltaic response was measured for films with and without C60 at a single wavelength of 458 nm to give short-circuit currents of 5.8 μA cm−2 and 0.15 μA cm−2 , respectively. This is rather low, but it is encouraging that the diblock copolymer with C60 performs best. Krebs et al.93 also tried to use the diblock copolymer approach to make materials with separated channels of hole and electron transporting domains (Fig. 2.26). A dialkyl substituted PPV domain was used as the hole conducting block, while a cyanovinylene terphenylene was presumed to be electron conducting. These two domains were coupled to a porphyrin system. As in the other studies of microphase segregated systems, rather low efficiencies were obtained for devices based on this material. Furthermore, a negative photovoltaic response was seen in the part of the spectrum where the porphyrin moiety was absorbed through its Soret band. Krebs94 also tested another self-assembling system consisting of a polar head group chromophore attached to a poly(cyanovinylene-terphenylene). This mater-
The Polymer Solar Cell
33
Figure 2.26 Diblock copolymer of a central porphyrin unit joined to a dialkyl PPV and a poly(cyanovinylene terphenylene) at each end.
Figure 2.27 The self-assembling polar molecule with a poly(cyanovinylene-terphenylene) domain coupled to an aza-triangulenium dye group.
ial had previously been found to self-assemble using the Langmuir–Blodgett technique, giving films with a rectifying behavior95 (Fig. 2.27). Donor/acceptor copolymers based on a PPV, triphenylamine, and a perylene bisimide have been prepared by Hua et al.96 Lu et al. used MEH-PPV blended with (PTCA) and investigated the effect of annealing on the morphology and photovoltaic efficiency.97 Barber et al. prepared devices from a block copolymer [MEHPPV]-co-[biphenylene vinylene] blended with PCBM having a maximum power conversion efficiency of 2.4%.98 2.2.1.5 Thermocleavable materials
Alkyl and alkoxy polymers such as P3HT, MEH-PPV, and MDMO-PPV have been very successful materials in solar cells with desirable properties such as solubility in organic solvents, processability, and semicrystallinity. These properties are mainly due to the judicious choice of flexible side chains. On the other hand, they take no part in the function of the solar cell, they do not absorb light, and they do not transport carriers. The film thickness is a compromise between the wish for maximum absorbance and carrier (exciton) lifetime/mobility. Another factor is the semiliquid nature of the film that is also imparted by the side chains. This flexibility allows changes in the structure to occur either with time or with heat. The latter property is explored in the annealing procedure where the PCBM component
Chapter 2
34
Scheme 2.6 Transforming the soluble prepolymer to PPV according to the Wessling reaction.
Scheme 2.7 Thermocleavable polythiophene carboxylic acid ester. R = alkyl.
of the film is typically allowed to form microcrystallites of a certain domain. It is, however, important that this structural development is not carried too far. A possible solution would be to remove the solubilizing groups after film formation in a similar way that PPV itself is prepared from a prepolymer. This principle has been realized for a special type of polythiophene with carboxylic acid groups developed by Liu et al.99 The acid groups were esterified with a tertiary alcohol and polymerized. The alkyl ester groups make the polymer soluble in organic solvents and allow film formation through spin coating and drop casting. The ester groups can then be removed simply by heating it to 190–210◦ C for a short time. Tertiary esters are susceptible to this thermocleavage reaction, forming the free acid and an alkene as the only products (Scheme 2.7). This type of polythiophene carboxylic acid was later found by Krebs and Spanggaard to be exceptionally stable in an organic solar cell.100 2.2.2 Molecules and oligomers 2.2.2.1 Small molecules
The first organic solar cell reported by Tang in 1986 had a bilayer structure of CuPc and a bis-benzimide derivative of perylene tetracarboxylic acid bisimide (PTCBI), shown in Fig. 2.28.101 Both of these materials were vacuum deposited. The phtalocyanine acted as the electron donor part while the perylene imide was the acceptor to a given cell that had a respectable ∼1% efficiency. Vapor deposition techniques continue to be an efficient method for creating layered solar cells from smaller molecular weight organic compounds. One of the recent examples is an asymmetric tandem photovoltaic cell with a front subcell com-
The Polymer Solar Cell
35
Figure 2.28 CuPc and PTCBI used in the first organic solar cell by Tang.
Figure 2.29 Chemical structures of SubPC and BCP.
prising CuPc/CuPc:C60 /C60 /PTCBI and a back subcell similar to the front, but having a diphenyl-dimethyl-phenanthroline, bathocuproin (BCP), instead of PTCBI. The front and back subcells were sandwiched between ITO and silver metal electrodes. This type of organic solar cell was reported to give a very high efficiency of 5.7%.102 Another type of phthalocyanine with only three arms (SubPc) has been shown to enhance the open-circuit voltage with up to 0.5 V (Fig. 2.29).103 Self-organizing molecular materials have been used as an alternative method to create structure in the active layer of an organic solar cell. A substituted hexabenzocoronene donor together with N,N -bis(1-ethylpropyl)-3,4,9,10-perylenetetracarboxylic acid diimide acceptor in a 40:60 blend was used to obtain device efficiencies of about 2% (Fig. 2.30).104 The disk-shaped benzocoronene formed columnar structures with the perylene derivative in between to give a material with a very large interfacial area. Another disk-shaped molecule that has been employed in self-organized solar cells is DL-CuPC in combination C60 bathocuproine105 (Fig. 2.31). 2.2.2.2 Fullerenes
The PPVs and P3HT organic polymers are electron rich materials that can be oxidized fairly readily, having high-energy HOMO levels, and are typically hole conducting materials. Organic materials with high electron affinity are much rarer. One of the few reasonably electron conducting materials are C60 and derivatives
Chapter 2
36
Figure 2.30 Hexakis(4-dodecylphenyl) benzocoronene (left) and N,N -bis(1-ethylpropyl)3,4,9,10-perylenetetracarboxylic acid diimide (right).
Figure 2.31 Chemical structure of the mesogenic molecule DL-CuPC.
thereof. C60 is in itself a rather insoluble material, but a soluble PCBM derivative is commercially available. The active organic layer in organic solar cells is typically constructed from a mixture of polymer and PCBM that forms an interpenetrating network—the so-called bulk heterojunction.1 PCBM was first described by Hummelen et al.106 The synthesis is shown in Scheme 2.8, where the methyl ester of 4-benzoylbutyric acid tosylhydrazone is treated with sodium methoxide, creating a reactive diazo intermediate. This compound then reacts with C60 in a 1,3-dipolar addition to one of the double bonds in C60 , followed by elimination of nitrogen. Two adducts, [5, 6] and [6, 6], are possi-
The Polymer Solar Cell
37
Scheme 2.8 Synthesis of PCBM, a soluble derivative of C60 .
Figure 2.32 Variant types of soluble C60 derivatives used in solar cells instead of [60]PCBM.
ble with the [5, 6] isomer dominating the reaction product. The numbers indicate between which type of rings the former double bond was situated. Heating the mixture in 1,2-dichlorobenzene convert it to the [6, 6] isomer, leading to color change of the solution from burgundy to more brownish. It is remarkable that this particular derivative of C60 with the phenyl ring was chosen “just for synthetic convenience” and is now the material of choice with bulk heterojunction solar cells. [6, 6]PCBMs have a series of characteristic UV-vis absorption at 210, 258, 328, 430, 492, and 696 nm in hexane solution. These optical transitions have low extinction coefficients and PCBM usually does not harvest the main part of photons in a solar cell. Symmetrical C60 can be replaced by the oval egg-shaped C70 , where lowenergy transitions become allowed, increasing light absorption. Cells with blends of [70]PCBM with MDMO-PPV have obtained an efficiency of 3%.16 Variation in the substitution pattern of PCBM with electron withdrawing and electron donating groups have been shown to affect the reduction potential of these materials and, hence, also the level of the LUMO.107 PCBM forms crystallites in the active layer, and control over size and morphology has a great influence on the efficiency of the devices. It has been shown that crystallite size increases on heat treatment of a PCBM/P3HT device improves the efficiency dramatically.108 Ultimately, the PCBM crystallites may grow too large to function well in the solar cell devices and an optimum heat treatment (annealing) can be found. Diffusion
Chapter 2
38
Scheme 2.9 The Prato reaction between C60 , N-methyl-glycine, and benzaldehyde.
of PCBM is controlled both by temperature and the glass transition temperature (Tg ) of the polymer. C60 can also be coupled to aromatic aldehydes using the Prato reaction. The aldehyde is reacted with N-methyl-glycin, forming an iminium salt that can act as a 1,3-dipole toward a double bond in C60 . The product is an aryl fulleropyrrolidine, as shown below in Scheme 2.9. Nierengarten et al.109 utilized the Prato reaction to make a C60 moiety bound to a trimeric oligophenylenevinylene with two alkoxy groups, reported to give a solar cell device with a monochromatic efficiency of 1% (Isc = 10 μA cm−2 , Voc = 0.2 V) at 400 nm. Peeters et al.110 prepared a more extended version with four phenylenevinylene units, each having alkoxy groups as shown in Fig. 2.33. UV-vis spectra show that the electronic transitions of the individual oligophenylenevinylene and C60 parts are retained in the combined molecule. Other photophysical studies indicated that a photogenerated charge-separated state is formed for the larger oligophenylenevinylenes, and solar cells were prepared from the tetrameric oligophenylenevinylene -C60 adduct. A short-circuit current of 235 μA and an open-circuit voltage of 650 mV were obtained under white light illumination (65 mW cm−2 ). Other structures incorporating thiophene and naphthalene units in the oligophenylenevinylene structure reported by Guldi et al.111 gave similar results. The fluorescence of the oligophenylenevinylene fragment is effectively quenched by the C60 moiety, presumably leading to a charge-separated state. Solar cell devices made from this material gave a rather low overall white light efficiency of <0.2%, which the authors attribute to the poor fill factor of <0.3. These linear p-conjugated systems derivatized with C60 have been reviewed by Roncali112 and also by Segura et al.113 2.2.2.3 Oligomers
Condensation polymers like the PPVs invariably have rather broad molecular weight distributions and may also contain defects that interfere with the properties needed for their function in solar cells. Smaller versions, so-called oligomers, are pure single molecular compounds that can easily be characterized in a number of ways.
The Polymer Solar Cell
39
Figure 2.33 Tetrameric oligophenylenevinylene -C60 adduct.
Such oligophenylenevinylenes are thus attractive both as model compounds for the PPVs, but may also be interesting in their own right. As opposed to the polymers, it is possible to control all aspects of the chemistry of the oligomers, allowing more interesting structures and variability. One obvious possibility is to functionalize each end of the oligomer with an electron donating and an electron accepting group. The resulting molecule has a dipole, and on excitation this may be further developed to the point where an intramolecularly charge-separated state is obtained. The optical spectrum of the oligomers in solution changes with an increasing number of monomer units and ultimately approaches that of the corresponding polymers. Normally, a bathochromic shift of the lowest transition is observed converging after about seven units.114 Rational routes to such asymmetric oligophenylenevinylenes require a stepwise and unidirectional synthesis rather than relying on the separation of statistical mixtures. A simple methodology has been developed by Jørgensen and Krebs.115, 116 This oligophenylenevinylene synthesis utilizes stilbene type monomers that have a masked aldehyde functionality at one end and a methylphosphonate ester at the other end. Extension of the oligophenylenevinylene molecule then requires two simple steps, namely, an HWE (HornerWadsworth-Emmons) reaction and a deprotection of the aldehyde as shown in Scheme 2.10. Several types of these stilbene type monomers have now been prepared with various substituents and variations in the aromatic rings. This synthetic methodology allows full control over the chemistry along the oligomers and the addition of many types of special end groups. A number of trimeric oligophenylenevinylenes of this type with a donor group at one end and an acceptor at the other were pre-
Oligophenylenevinylenes
40
Chapter 2
Scheme 2.10 Two condensation/deprotection cycles of the stepwise unidirectional oligophenylenevinylene synthesis. In the first cycle, a stilbene type monomer with a phosphonate ester group and a masked aldehyde is reacted with 4-dimethylamino-benzaldehyde in a HWE reaction, and then the acetal is deprotected with acid to expose a new aldehyde function ready for the next cycle.
Figure 2.34 Oligophenylenevinylenes with donor and acceptor groups at each end prepared by the stepwise method.
pared using this methodology and investigated in simple photovoltaic cells without PCBM (Fig. 2.34). This stepwise oligomerization approach culminated in the synthesis of a very large “all-in-one” molecule with several domains. The center part of this “all-inone” molecule was comprised of zinc-porphyrin coupled to a ruthenium terpyridine complex. This is a so-called dyad, which has been shown to be effective in absorbing light and creating a long-lived charge-separated state. By attaching a hole that transports oligothiophene to the porphyrin groups and what was hoped to be a more electron-transporting oligophenylenevinylene domain at the ruthenium complex, the idea was to transport the charges farther apart and ultimately to external electrodes.115
The Polymer Solar Cell
41
Figure 2.35 An “all-in-one” molecule with multiple domains, each responsible for different functions of a solar cell. Anions are omitted for clarity.
Scheme 2.11 Selective bromination and lithiation of 3-alkylthiophenes. R = alkyl.
Scheme 2.12 Head-to-tail coupling of two thiophene units using Stille coupling. R = alkyl.
Similar to the oligophenylenevinylenes, a number of oligothiophenes have been prepared and used in solar cells. Of special importance is regioregular oligomers of 3-hexylthiophene. There are a few different routes to these oligomers, but production of reasonable amounts limits the possibilities to the route shown in Scheme 2.11. It relies on the fortuitous fact that lithiation occurs predominantly in the 5-position of 3-hexylthiophene, while bromination with NBS is directed to the less hindered 2-position. The 5-lithio-3-alkyl-thiophene intermediate is generated at low temperature and some rearrangement to the 2-isomer is inevitable. Reaction with a trialkyl tin chloride (trimethyltin chloride or tributyltin chloride) produces the stable 5-trialkylstannyl-3-alkyl-thiophene. Two thiophene units can then be joined using the Stille type coupling catalyzed by palladium, as shown in Scheme 2.12. The bis-thiophene can be lithiated and brominated, and the reaction cycle continued to obtain higher oligomers. This methodology was extended to prepare a very long oligomer containing 16 thiophene units end capped with dimethylaminophenyl groups.117
Oligothiophenes
Chapter 2
42
Figure 2.36 A 16 mer oligo-3-hexylthiophene end capped with dimethylaminophenyl groups.
Figure 2.37 Examples of trigonal terthiophene molecules for solar cells.
Cravino et al.118 prepared trigonal terthiophenes with a triphenylamine core (Fig. 2.37). Solar cells with a vapor-deposited layer of C60 as an acceptor had an efficiency of 0.32% due to a low fill factor of 0.3.
2.3 Fast and Simple Guide to a Polymer Solar Cell from Scratch The purpose of this section is to show the minimum requirements that are needed to prepare a polymer solar cell. The presentation is visual, with pictures showing what it actually looked like when these devices were prepared. As shown later in this book, it is possible to prepare the organic photovoltaic (OPV) devices under more controlled conditions in a glove box, but this is not necessary for the successful preparation of devices. The size of the devices is kept at a moderate size here to avoid delicate design schemes and precise positioning of electrodes during characterization and measurement. Each step in the practical fabrication of an organic solar cell device is exemplified. The equipment, the substrates, and the preparative organic chemistry needed for the synthesis of the most commonly used active polymers and molecules are shown in detail. Especially, the synthesis of the presently most popular material, namely, regioregular P3HT, and the acceptor material [60]PCBM is presented in detail. The device fabrication includes how to spin coat successive layers of PEDOT:PSS and the active bulk heterojunction mixture of P3HT-PCBM. The thermal evaporation of an aluminum film in high vacuum is discussed and the characterization of the solar cell device is finally described. 2.3.1 Equipment In terms of equipment, the only absolute requirement is a metal evaporator for application of the metallic cathode. The best suited is a thermal evaporator, where an electrical current heats up a high melting metal wire or boat that is typically made of tungsten or tantalum. The electrical current heats up the wire or boat and the
The Polymer Solar Cell
43
cathode metal evaporates onto the OPV device. Evaporators are available commercially or can be built by a skilled technician. In essence, the requirement for the evaporator is a vacuum chamber that can be pumped to a pressure of 1×10−6 mBar or less. Inside the chamber, the device must be fixed above the source (boat or wire) and electrical feedthroughs for heating the source must be incorporated. This is the simplest possible method, and metal evaporators are available commercially at all levels of complexity and in all price ranges. Aside from this, simple laboratory conditions are needed with a well-ventilated hood. If great caution is exercised and gloves, goggles, and a mask are worn, it is possible to do the preparation without such equipment. The reasons for undertaking such an endeavor could be educational in a school where access to expensive equipment may not be available. 2.3.2 The substrate The preparation of the OPV device starts with the substrate, which is typically a glass slide with a thin layer of ITO that is a transparent electrically conducting material. The ITO needs to be patterned to avoid making short circuits when contacting the completed device. This is achieved by etching as shown schematically in Figs. 2.38 and 2.39. The conducting side of the glass-ITO slide can be found by measuring the resistance with a simple multimeter. When handling materials, it is advisable to use gloves to avoid depositing grease from the fingers. When making connections to the back electrode (the evaporated metal cathode), there is a high risk that the device might become short-circuited. If this happens, an electrical connection will be created between the front and back electrode and the device will give low voltages and to a lesser extent a lower short-circuit current. This type of failure is eliminated by removing the ITO in the area where the back electrode connection is made. A simple way of achieving this is by masking with ordinary tape and etching the nonmasked area with an acid solution (20% w/v HCl, 5% w/v HNO3 ) at a temperature of 50◦ C. A piece of paper can be used as a tape guide so the tape can be placed to mask the desired area. The tape is carefully rubbed (Fig. 2.40) so it adheres firmly against the ITO side of the glass. Any cavities extending to the edge of the tape/glass will act as a channel for the acid, and nondesired etching is the end result. In the same way, different front electrode geometries can be made by masking the geometry with tape. The masked glass slide is submerged for 30–60 s in the heated acid mixture so the area that is going to be etched is fully submerged. The acid mixture is washed off with water and the surface resistance can be checked with a multimeter if there is doubt about whether the etching is complete. A little training allows one to visually follow the etching by catching the reflection from the surface with the eye. The thicker the ITO (the higher the sheet conductivity and the lower the sheet resistivity), the longer the etching time will be. When complete etching is observed, the tape is removed and the slide is washed with water followed by a
44
Chapter 2
Figure 2.38 A flow chart showing the cross section of a simple device-fabrication process By etching away a part of the ITO, short-circuit formation is avoided during contacting to the metal top electrode. The PEDOT:PSS layer is shown in dark gray and the active P3HTPCBM bulk heterojunction layer is shown in light gray. The aluminum top electrode is shown in black. During contacting to the top electrode, it is common to contact all the way through the electrode to the underlying layers. The contact to PEDOT:PSS is not problematic and one does not need to remove the PEDOT:PSS layer due the high sheet resistivity.
The Polymer Solar Cell
45
Figure 2.39 The conducting side of ITO coated glass can, with a little practice, be identified with a fingernail or by observing the specular reflection from a lamp or the sun. The most reliable method is by measuring the resistance, which is low for the ITO side and high for the glass side (left). The part of the ITO that is to be kept is masked by adhesive tape (right).
Figure 2.40 The tape is attached firmly to the surface (left), and the exposed side of the ITO is immersed in the acid solution (right).
water miscible organic solvent such as ethanol or tetrahydrofurane (THF). Great care has to be taken to avoid contact between flammable solvents (ethanol and THF) and nitric acid solutions. If many slides are being prepared, it can be worthwhile to have a saturated aqueous NaHCO3 (aq), where the etched slides are submerged while etching the next slide. This neutralizes the acid and avoids the continuation of etching. It is important to stress that the device must be washed immediately after etching is complete to remove acid traces. Failure to do this results in a poor definition of the etched area (Fig. 2.41). The ITO side is rubbed with chloroform on lens paper, and then placed in a beaker with chloroform in an ultrasonic bath for 5–15 min. This is to remove tape residues and other soluble organic materials. The slides can be stored in this etched form indefinitely when kept dry. There are several lithographic methods available for creating the pattern by etching; the tape method is simple and fast.
46
Chapter 2
Figure 2.41 Cleaning of the etched substrate in an ultrasonic bath (left), and the etched substrate where the etched part of the slide is clearly visible (right).
2.3.3 The PEDOT:PSS layer The etched slides have to be cleaned immediately prior to use in order to obtain the best result. The slides are submerged in 2-propanol (also known as isopropanol) in a beaker and subjected to ultrasound for a couple of minutes. If an ozone treater is available, this is desirable after the cleaning step. If no ultrasonic bath is available, careful rinsing with pure water is advised. The best method involves cleaning the substrates in several different solvents starting with cyclohexane, chloroform, and acetone, and ending with 2-propanol or water. At this stage, it is common practice to apply a hole conducting PEDOT:PSS layer that improves the surface roughness of the substrate and improves and stabilizes the electrical contact between the transparent anode and the active layer. It is possible to make devices without PEDOT:PSS if none is available; however, it is strongly advisable to not omit this step. The PEDOT:PSS is best applied by spin coating. The slide is placed in the spin coater with the ITO slide facing up (Fig 2.42). If you have a dust problem, the slide can be rinsed by spinning the slide while washing with millipore water. When the slide is dry the PEDOT:PSS can be applied while the slide is spinning, or by spinning the slide after application. The spin coating of PEDOT:PSS dispersion (typically 1.3% wt/v) has to be performed at high speed; 2000–3000 rpm gives a very thin homogenous layer. Spinning has to be continued for a few minutes because the dispersion dries slowly. The application of the PEDOT:PSS dispersion can be done with a pipette, or, as shown in Fig. 2.43, with a plastic syringe through a microfilter. After coating, the connection area to the ITO electrode is cleaned with a cotton stick dipped in water. If the spin coater is stopped before the film has set, particles may form, resulting in a poor PEDOT:PSS layer.
The Polymer Solar Cell
47
Figure 2.42 The PEDOT:PSS layer is applied by spin coating from an aqueous dispersion. The slide is fixed on the spin coater (left), and cleaned with water while spinning (right) before applying the PEDOT:PSS.
Figure 2.43 The PEDOT:PSS dispersion is applied before spinning by filtering through a microfilter (left). After the spinning, the PEDOT:PSS is removed at one end using a wetted cotton bud. This serves the purpose of making electrical contacts to the ITO easier once the device has been completed (right).
The slides should be kept in a dust free container. Before the slides are used, the PEDOT:PSS layer has to be heated in an oven or on a hot plate to a temperature of 120–200◦ C for 10–60 min to dry the film (Fig. 2.44). The drying step is performed immediately prior to use. 2.3.4 The active layer 2.3.4.1 Materials
Conjugated organic materials are semiconductors and will, if applied in a photovoltaic device geometry, yield some sort of photovoltaic response. Some materials are of course much better suited than others and the choice of material is most often made based on various criteria such as stability, purity, ease of preparation, and efficiency of the final device (Fig. 2.45).
Chapter 2
48
Figure 2.44 The freshly prepared PEDOT:PSS slides are dried in an oven prior to use (left), giving a smooth PEDOT:PSS film on the slide (right).
Figure 2.45 The materials employed here are [60]PCBM.
One of the best polymer materials available today that fulfills many of the criteria just mentioned is P3HT. It is a stable polymer material that can be prepared by different methods with a varying degree of experimental difficulty. Purification to a product suitable for photovoltaic work is easily achieved. We describe three different methods from the very simplest one that can be performed essentially in handheld equipment if no stirring apparatus is available to the more complex method where anhydrous and anaerobic conditions are required. The difference between the regiorandom P3HT product obtained by simple ferric chloride oxidation and the regioregular P3HT product obtained via nickel catalyzed aryl-aryl coupling is the position of the hexyl side chains (see Scheme 2.12). The P3HT obtained by the methods presented here is well suited for preparing homojunction devices with just the polymer materials sandwiched between two electrodes. The current state of the art, however, employs a mixture of an acceptor material such as PCBM and a donor material such as P3HT.
The Polymer Solar Cell
49
Scheme 2.13 The synthetic approaches to P3HT.
Poly(3-hexylthiophene) Regiorandom poly(3-hexylthiophene) the simplest way This is the simplest method for preparing regiorandom P3HT and very little equipment is required. The absolute minimum requirement is a balance for weighing the reactants and a flask for the reaction. The exact volume of the solvent chloroform is not critical. The mixture can be shaken by hand, but it is advantageous to have some consistent means of stirring or shaking the reaction mixture (Fig. 2.46).
Method A, direct mixing. Anhydrous FeCl3 (35 g, excess) and chloroform (500 mL) is placed in a 1-L conical flask on a magnetic stirrer with a large magnetic stirring bar. 3-hexylthiophene (16.8 g, 0.1 mol) is then added to the flask in one
50
Chapter 2
Figure 2.46 Pictures of the synthetic procedure leading to regiorandom P3HT by ferric chloride oxidation. (a) The solid ferric chloride is placed in the reaction flask. (b) The chloroform solvent is added. (c) 3-hexyl thiophene (method A) is added in one portion. (d) 3hexylthiophene in chloroform is added dropwise to the ferric chloride (method B). (e) The reaction mixtures are stirred. (f) The reaction mixtures are poured into methanol, leading to precipitation of the crude product. (g) The mixture is stirred for 1 hr. (h) The crude product is filtered.
The Polymer Solar Cell
51
portion with vigorous stirring and the mixture is stirred for 16 hr. At the end of the reaction, the mixture is poured into methanol (3 L) in a large beaker. The mixture is stirred for 1 hr before filtering the crude solid that is washed carefully with water and methanol and dried in a vacuum oven at 70◦ C. The crude yield obtained is typically in the range of 9–12 g of a black, slightly sticky material. Purification by Soxhlet extraction gives a typical yield of 4–5 g (24–30%) after reprecipitation from methanol and drying. Method B, careful addition. Anhydrous FeCl3 (35 g, excess) and chloroform (400 mL) is placed in a 1-L conical flask on a magnetic stirrer with a large magnetic stirring bar. The mixture is then stirred vigorously while 3-hexylthiophene (16.8 g, 0.1 mol) dissolved in chloroform (100 mL) is added dropwise with an addition funnel over 1 hr. After the addition, the mixture is stirred for 15 hr. Workup is identical to method A. Figure 2.46 shows a series of pictures obtained throughout the synthetic procedure by the two methods. The crude yield is typically in the range of 7–10 g of a black, slightly tarry material. Purification by Soxhlet extraction gives a typical yield of 4–6 g (24–36%) after reprecipitation from methanol and drying. Regioregular P3HT via the McCollough route119
This method gives a regioregular product that is superior with respect to light absorption and power conversion efficiency when applied in devices. More skill is, however, required when performing the synthetic reactions and great care is needed in order to exclude moisture and oxygen from the air.
2-bromo-3-hexylthiophene. 3-Hexylthiophene (60 g, 0.357 mol) is stirred with N-bromosuccinimide (NBS) (63.3 g, 0.357 mol) in a 1:1 mixture of chloroform and acetic acid (500 mL). The NBS dissolves gradually. After 15 min the solution becomes clear and everything dissolves. After 10 min of additional stirring, the mixture is poured into water (500 mL) and the phases are separated. The chloroform phase is then washed with 10% aqueous potassium hydroxide [KOH(aq)] (500 mL) and the clear colorless chloroform phase is dried (MgSO4 ) and evaporated to give a colorless oil (88.6 g) that is distilled using a water aspirator. The prefraction boiling up to 120◦ C should be discarded (6 g) and the product that distills at 134◦ C (78.4 g, 89%) should be collected. Regioregular poly(3-hexylthiophene). Diisopropylamine (5 g, excess) is mixed in dry THF (250 mL) and cooled to −50◦ C. nBuLi (30 mL, 1.5 M, 44 mmol) is added. The mixture is stirred for 40 min at −50◦ C and then cooled to −78◦ C. 3-bromohexylthiophene (10.92 g, 44.2 mmol) is added and the mixture is stirred for 1 hr with gradual heating to −40◦ C.
52
Chapter 2
The mixture is then cooled back down and MgBr2 Et2 O (11.5 g, 44.2 mmol) is added. The mixture is allowed to reach −5◦ C when Ni(dppp)Cl2 (120 mg) is added and the mixture stirred. After 1 hr, a polymer is formed as evidenced by size exclusion chromatography (SEC) and after 3 hr, a precipitate is formed. The mixture is left overnight at room temperature with stirring. 90% aqueous MeOH (2 L) is then added, the mixture is filtered, and the solid washed with methanol (Fig. 2.47). The dark colored paste is dried in a vacuum oven at 70◦ C and purified by Soxhlet extraction (Fig. 2.48), giving a typical yield of 2–3 g (27–41%) after reprecipitation from methanol and drying as a green material with a metallic luster. A good method for the purification of polymers employs Soxhlet extraction. This method allows for continuous extraction with a limited amount of solvent. The apparatus is shown in Fig. 2.48, and the polymer product is placed in an extraction thimble that is normally made from cellulose fibers, while quartz or glass fiber thimbles are also available for extraction with corrosive or strongly acidic or basic solvents. The normal procedure starts with extraction in a polar solvent such as methanol followed by acetone and then a nonpolar solvent such as hexane. These solvents will extract many salts, small organic materials, and oligomers, while the polymer is insoluble in these solvents and thus left in the thimble. The soluble polymer fraction can then be extracted using an appropriate solvent for the polymer such as chloroform, tetrahydrofurane, or toluene. The extract can then be concentrated to a smaller volume by evaporation and the product precipitated by addition of the extract solution to a large volume of methanol. Filtration and drying gives the product, which is sufficiently pure for application in devices. One further advantage of the Soxhlet extraction method is that only the soluble polymer fraction is isolated. This method applies to any crude polymer product regardless of how it was synthesized. Purification of the crude P3HT
Size-exclusion chromatography (SEC) is by far the best suited method for characterizing polymer products according to their size (Fig. 2.50). It is a versatile technique that allows for purifying materials according to size if needed. The techniques can be contrasted to high-performance liquid chromatography (HPLC), which relies on physical chemical interaction between the column material, the solvent, and the analyte. HPLC can be termed interaction chromatography. In SEC there is ideally no interaction between the column material and the analyte, and it is assumed that the experiment is performed in a good solvent for the analyte. Separation of molecules using SEC is according to its name only, based on the molecular size or, more accurately, the hydrodynamic volume of the molecule in solution. In HPLC, the column material has a specific surface chemistry that can interact with the analytes in absorption-desorption equilibria that Size-exclusion chromatography
The Polymer Solar Cell
53
Figure 2.47 Pictures of the synthetic procedure leading to regioregular P3HT by the McCollough route. (a) The thoroughly dried glassware required for the reaction equipped with a low temperature thermometer, a septum for introducing reagents, and an argon bubbler. (b) The nBuLi is added to the diisopropylamine solution in dry THF. (c) The solution of MgBr2 Et2 O is added. (d) The Ni(dppp)Cl2 has been added and the polymerization has started. (e) The product is precipitated in methanol (MeOH) and filtered. (f) The raw product is obtained as a dark pastelike material.
54
Chapter 2
Figure 2.48 (a) A schematic illustration of the Soxhlet extraction apparatus. (b) Soxhlet extraction of regioregular P3HT employing a round-bottomed flask. (c) Soxhlet extraction of regiorandom P3HT, showing all the fractions from the left—methanol, acetone, and hexane. The final extraction on the hot plate is with chloroform.
depend on the solvent composition, and typically HPLC employs a gradient of solvents throughout its run. The time from sample injection to elution of the desired analyte thus depends on the chemistry of the desired product, the impurities, the column systems, and the solvent system at hand, and is in no way predictable. In SEC, separation relies on the ability of the analyte to penetrate the pores in the passage through the column system. The column material typically consists of porous cross-linked polymer-gel particles with a diameter on the 5–25-μm scale with a well-defined pore diameter in the 100–10,000,000 Å range. Very large molecules cannot enter the pores and are transported in the volume outside the gel particles, being eluted first. Small molecules have access to all the pores and are efficiently retained in the system because they have to pass the entire free volume and are thus eluted last. The time of an SEC run is therefore a known parameter that can be chosen freely within practical limits of operating pressure and fraction switching. The time from injection to completion of the SEC experiment is given as the ratio of the free volume in the column and the solvent flow rate. A final advantage of SEC is that the amount of analyte material that can be injected is in principle only limited by the solubility of the analyte and the viscosity of the analyte solution, which should not differ significantly from the pure solvent. Generally, 10% w/w solutions can be injected. In the system shown here, injection volumes in the 1–10 mL range are used with operating pressures in the 80–140 bar range and a solvent flow of 20 mL min−1 .
The Polymer Solar Cell
55
Figure 2.49 Pictures of the P3HT products as they appear to the eye. The raw regiorandom product is black in color (left), the purified regiorandom product is red in color (middle), and the regioregular product has a greenish metallic luster (right). The ruler for scale is in centimeters.
There is one catch with respect to the determination of the molecular weight, and that is related to the fact that SEC essentially gives a measure of the size of the molecule in solution and not its actual molecular weight. The measure of size is relative to the size of the pores in the gel, which to some extent depends on the solvent used. A calibration of the time axis of the SEC chromatogram to molecular weight can thus only be achieved by separating narrow fractions followed by a determination of their individual molecular weight by other means. For the synthetic chemist this is highly impractical, and the most commonly employed method involves the use of a set of standards for a known polymer such as polystyrene. Most often, molecular weights are quoted relative to a known polystyrene standard. This is the commonly accepted and adopted method and it is quite accurate when the solvent used is good for both the calibrant (polystyrene) and the analyte. When dissolved in a good solvent, polystyrene adopts a globular shape and the calibration thus only works well if the analyte has the same properties in the solvent. Many conjugated organic molecules used for polymer solar cells are stiff, rodlike molecules and do not compare well to the polystyrene molecular weight scale, which tends to overestimate the molecular weight. A rodlike molecule will have a hydrodynamic volume
56
Chapter 2
Figure 2.50 Pictures a typical a preparative SEC apparatus. (left) Two systems differing in the mode of detection. Both systems comprise an isocratic pump (250 mL min−1 ), injection pump (50 mL min−1 ), injection loop switch, and 16-port fraction collector. One system has a UV-vis diode-array detector, and the other system has a single-wavelength UV-vis detector and a refractive-index detector. (right) Two sets of gel-column systems. One system for small molecule characterization in the range 100–10,000 g mol−1 comprises a precolumn [25 mm × 25 mm diameter(Ø)], a 100 Å pore volume (600 mm × 25 mmØ), and a 1000 Å pore volume (600 mm × 25 mmØ). Another system for large molecule characterization in the range 100–500,000 g mol−1 comprises a precolumn (25 mm × 25 mmØ), a 500 Å pore volume (600 mm × 25 mmØ), a 10,000 Å pore volume (600 mm × 25 mmØ), and a 1 × 106 Å pore volume (600 mm × 25 mmØ).
equal to a sphere that can contain it (i.e., the diameter of the sphere is equal to the length of the linear molecule). As a technique, SEC is very useful for following polymerization reactions and characterizing the polymer products. The SEC analysis of the P3HT polymers described in this chapter are shown in Fig. 2.51 for the regiorandom polymers obtained by ferric chloride polymerization and in Fig. 2.52 for the regioregular polymer obtained via the McCollough route. For regiorandom P3HT, it is evident that the raw product has a very broad distribution with polydispersities (PDs) of the sixth order. After purification and removal of the low molecular weight components, a higher molecular weight product is obtained with a lower value for PD. In contrast, the regioregular product has a much narrower molecular weight distribution in the final product that approaches the PD typically obtained for condensation polymerizations. The hexane fraction from the Soxhlet extraction gives a product with a small PD. Overall, the molecular weight for the regiorandom product is much higher than the molecular weight for the regioregular product. The regiorandom P3HT product has better solubility properties and dissolves readily in solvents such as chloroform despite its much larger molecular weight. In solution, the regiorandom and the regioregular products have very similar UV-vis spectra and give clear orangeyellow solutions. It should be noted that P3HT is known for its thermo- and sol-
Characterisation by NMR and UV-vis spectroscopy
The Polymer Solar Cell
57
Figure 2.51 SEC traces of regiorandom P3HT prepared by the ferric chloride polymerization run in chloroform. The product from the direct mixing is shown above and the product from the dropwise addition is shown below. The short dashed calibration curve is based on polystyrene standards marked by a triangle. The column system comprised a precolumn (25 mm × 25 mmØ), a 500 Å pore volume (600 mm × 25 mmØ), a 10,000 Å pore volume (600 mm × 25 mmØ), and a 1 × 106 Å (600 mm × 25 mmØ) pore volume. The solvent flow rate was 20 mL min−1 . The solid curve is the raw polymer product and the curve with the long dashes is the product that has been purified by Soxhlet extraction.
vatochromic behavior, which means that solutions may darken when cooled and when the properties of the solvent changes. These effects are mostly ascribed to the formation of aggregates. The solution spectra are very similar, while both the regiorandom and, to a larger extent, the regioregular product exhibit a significant red shift of the absorption in the solid state. Particularly, the regioregular product is subject to a dependence of the final UV-vis spectrum on the solvent employed for the film preparation as illustrated in Figs. 2.53–2.55 for the UV-vis and NMR spectra. The slower the drying of the film during spin coating, the more the vibronic fine structure is obtained and a large red shift of the absorption is observed. This is desirable in polymer solar cell applications. Synthesis and purification of PCBM The most commonly available acceptor material is a soluble derivative of Buckminster fullerene C60 known as PCBM. While
Chapter 2
58
Figure 2.52 SEC traces of regioregular P3HT prepared by the McCollough route run in chloroform. The short dashed calibration curve is based on polystyrene standards marked by a triangle. The column system comprised a precolumn (25 mm × 25 mmØ), a 500 Å pore volume (600 mm × 25 mmØ), a 10,000 Å (600 mm × 25 mmØ) pore volume, and a 1 × 106 Å (600 mm × 25 mmØ) pore volume. The solvent flow rate was 20 mL min−1 . The solid curve is product purified by Soxhlet extraction and the curve with the long dashes is the product obtained from the hexane extraction, which has a small molecular weight.
Table 2.4 Molecular weights for the products determined by SEC in chloroform based on polystyrene standards.
Compound Method A, crude product Method A, Soxhlet purified Method B, crude product Method B, Soxhlet purified McCollough route, hexane fraction McCollough route, chloroform fraction
Mn 12150 94000 6800 40000 2750 17300
Mw 75300 420750 41100 197400 3500 33700
Mp 59100 154400 33300 59100 3500 33600
PD 6.2 4.5 6.0 4.9 1.3 1.9
C60 can be employed on its own, PCBM has many advantages and it is currently used almost exclusively. The synthetic procedure is shown in Scheme 2.14, starting from commercially available materials. The synthetic procedure and workup is of medium difficulty and cannot be prepared using handheld equipment. PCBM was first prepared by Hummelen et al.120 in 1995 and has become the most widely used acceptor material in the field of polymer photovoltaics. The synthetic procedure exploits the 1,3-dipolar addition of a diazoalkane to C60 . Alternatively, the addition can be viewed as a carbine addition giving exclusively a [5, 6] adduct that on heating is transformed into the [6, 6] adduct. The reaction is by no means efficient, but unreacted C60 can be efficiently recovered. When the chemistry of the C60 and C70 fullerenes was established, one of the first properties that was discovered was their weakly oxidizing nature. The
The Polymer Solar Cell
59
Figure 2.53 UV-vis spectra of regiorandom and regioregular P3HT in solution and in spincoated films from chloroform.
Figure 2.54 UV-vis spectra of regioregular P3HT in spin-coated films obtained from chloroform and o-dichlorobenzene (left). The visual appearance of regiorandom (red) and regioregular (purple) spin-coated films (right).
fullerenes were quickly viewed as a collection of pyracyclene units that are a 4nπ system with the ability to accept two electrons to become a (4n + 2)π dianionic system or by reaction with a nucleophile to give an adduct and a cyclopentadienide anion. This reactivity was explored and many derivatives of C60 in particular were reported. Reaction with amines, phosphines, Grignard reagents, organolithium reagents, thiolate, methoxide, phenoxide, carbenes, and diazoalkanes were quickly reported and the 1,3-dipolar addition of diazoalkanes was found to initially produce a pyrazoline derivative that eliminated nitrogen to form a fulleroid. A thermal rearrangement was found to give the methanofullerene. The synthetic approach to [6, 6]PCBM (Scheme 2.14) is starting from commercially available benzoylbutyric acid. A straightforward esterification provides the methyl ester that on conversion to tosylhydrazone is reacted in situ with C60 and sodium methoxide to give the
60
Chapter 2
Figure 2.55 1 H NMR of regiorandom P3HT with an insert showing the region around the signals originating from the methylene group next to the thiophene ring (top). The signal at 2.8 ppm originates from the preferred head-to-tail isomer, while the signal at 2.6 ppm is due to the head-to-head isomer. Small extra signals in the aromatic region can also be discerned. Regioregular P3HT has a simpler spectrum (bottom) with almost no signal at 2.6 ppm and a clean singlet at 7 ppm in the aromatic region.
desired [5, 6]-PCBM fulleroid that is subsequently rearranged thermally by reflux in 1,2-dichlorobenzene to [6, 6]PCBM, as shown in Fig. 2.57. The procedure described here rests heavily on the work by Hummelen et al.120 and the description here is aimed at showing how the reaction is performed such that nonchemists can do it and obtain [6, 6]PCBM suitable for making solar cells. 5-Oxo-5-phenyl-pentanoic acid methyl ester. 4-benzoylbutyric acid (50 g, 0.26 mol) is mixed in a 500 mL round-bottomed flask with methanol (200 mL) and thionyl chloride (31 g, 0.26 mol) is added dropwise under vigorous stirring.
The Polymer Solar Cell
61
Scheme 2.14 The synthetic approaches to PCBM.
After the addition, the mixture is left for three days and evaporated to give an oil that is dissolved in ether (500 mL). The ether phase is then washed carefully with 1 M NaHCO3 (3 × 500 mL) in a separating funnel [vigorous formation of CO2 (g)]. The ether phase is then dried (MgSO4 ), filtered, and evaporated to give a clear, colorless oil that is distilled using an oil pump, with a boiling point of 145–155o C/ 7 × 10−2 mbar. This gives the product as a colorless oil in 82% yield (43.8 g). 5-toluenesulfonylhydrazono-5-phenyl-pentanoic acid methyl ester. 5-oxo5-phenyl-pentanoic acid methyl ester (20.6 g, 0.1 mol) is mixed with p-tosylhydrazide (22.3 g, 0.12 mol) in methanol (150 mL) and the mixture is refluxed for 6 hr. The mixture is then cooled in a refrigerator and the solid product is filtered, washed with a little cold methanol, and dried in a vacuum oven at 70◦ C. This produces the hydrazone in 93% yield (34.8 g). {6}-1-(3-(methoxycarbonyl)propyl)-{5}-1-phenyl[5,6]-C61 . 5-toluenesulfonylhydrazono-5-phenyl-pentanoic acid methyl ester (1.5 g, 4 mmol) is dissolved in dry pyridine (30 mL) under argon. NaOMe (225 mg, 4.14 mmol) is added and the mixture is stirred for 15 min. A solution of C60 (1.44 g, 2 mmol) in 1,2dichlorobenzene (100 mL) is then added and the mixture is stirred at 65◦ C for 24 hr in an oil bath. The mixture is evaporated to dryness on a rotary evaporator and dissolved in chlorobenzene (10 mL). The chlorobenzene solution is loaded onto a silica gel column packed in chlorobenzene. Elution with chlorobenzene results in a fast-moving purple band, which is unreacted C60 followed by the product as a purple-brown band (see Fig. 2.56). Evaporation of the fraction that contains the second band gives the product in a 30% yield (0.55 g) used directly for the conver-
62
Chapter 2
Figure 2.56 Pictures of the preparation and purification of PCBM. (a) The 1,3-dipolar addition of the in-situ generated diazoalkane to C60 . (b) Loading of the crude reaction mixture in chlorobenzene onto a silica column packed in chlorobenzene. (c) A layer of sea sand has been added to avoid stirring up the column material and elution with chlorobenzene has started. (d) The fast moving band of unreacted C60 elutes first. The PCBM band elutes next. (e) After evaporation and isolation of the [5,6]PCBM, it is converted to [6,6]PCBM by boiling in 1,2-dichlorobenzene under argon for 16 hours.
sion into the methanofullerene. It is not practical to recover unreacted C60 on the scale described here. 1-(3-(methoxycarbonyl)propyl)-1-phenyl[6,6]-C61 . {6}-1-(3-(methoxycarbonyl)propyl)-{5}-1-phenyl[5,6]-C61 (550 mg, 0.6 mmol) is refluxed in 1,2dichlorobenzene (60 mL) for 12 hr. The reaction is easily followed by UV-vis and the color of the solution changes from distinctly purple to brown (see Fig. 2.56). After conversion, the mixture is evaporated to dryness and hexane (100 mL) is added. The product is then filtered, washed with hexane and methanol, and dried to give the product in 82% yield (0.45 g).
The Polymer Solar Cell
63
Figure 2.57 The thermal rearrangement of the fulleroid adduct to the methanofullerene (top) and the changes in the UV-vis spectrum associated with the rearrangement (bottom).
2.3.4.2 Spin coating of the active layer
The active material is accurately weighed and then dissolved in a solvent such as chlorobenzene or 1,2-dichlorobenzene. To make a bulk heterojunction cell between a polymer and the soluble fullerene derivative, [60]PCBM, the materials are weighed and added at the same time. The solvent can be added by using a plastic or glass pipette. Typically, amounts are of 20-mg of polymer and 20-mg PCBM in 1 mL of solvent. The temperature of the solution can be raised on a hot plate (Fig. 2.58) to increase the speed at which the materials dissolve. When the substances have dissolved, the solution is cooled and filtered through a microfilter using a syringe. This eliminates undesirable particles from the solution. The slides can be rinsed of dust by washing with a clean organic solvent while spinning (Fig. 2.59), or by clean pressurized air. It is very important to remove dust particles if an even, homogenous film is to be obtained. The particle-free solution is spin coated (Fig. 2.60) on the dust-free glass-ITOPEDOT:PSS substrates. The connection area for the ITO contact at the end that has not been etched is cleaned with a cotton bud wetted with solvent. UV-vis spectra of the device film can be recorded at this stage.
64
Chapter 2
Figure 2.58 The materials are dissolved by gentle heating on a hot plate with the temperature set at or below the boiling point of the solvent (left). The solution of the active materials is filtered through a microfilter before use (right).
Figure 2.59 The slide can be rinsed just prior to applying the active material with clean solvents.
2.3.5 Evaporating the electrode The final step in the process of completing the device is evaporation of the back electrode that for most OPVs is aluminum; and if the devices are to be handled in air, this is the only advisable metal cathode to use. While calcium and magnesium alloys have been used, they are simply not stable under ambient conditions. The slide is placed in a suitable mask in the evaporator (Fig. 2.61). The hole defines the back electrode geometry. Printer and/or photocopy transparecies [poly(ethylene-terephthalate) (PET) foil] can be used as an easy way to make masks. The geometry can be cut with a scalpel and a liner. Once the pressure reaches the desired value, typically 1 × 10−6 mbar or lower, the metal cathode can be evaporated (Fig. 2.62). Once the evaporator has cooled down after 1–20 min, argon can
The Polymer Solar Cell
65
Figure 2.60 The solution with the active material is applied by spin coating while spinning (left) and the material is removed using a cotton bud wetted in the appropriate solvent where the ITO contact is to be made later (right).
Figure 2.61 The slides with the active material are placed in a substrate holder where the areas to be coated with the cathode metal are exposed (left) and the substrate holder is introduced into the metal evaporator (right).
be introduced into the chamber and the completed devices can be recovered and used for measurements. 2.3.6 Applying electrodes and measuring the electrical properties of the devices Once the devices have been removed from the evaporator, they are subjected to degradation due to oxygen and water from the atmosphere. Most of the degradation takes place at the interface between aluminum and the active layer (for further details, see Chapter 4). Regioregular P3HT:[60]PCBM devices are, however, quite stable both in the dark and under illumination, and there is no rush to characterize
66
Chapter 2
Figure 2.62 Evaporation of the aluminum cathode (left), and the completed devices based on regioregular P3HT:[60]PCBM (device on the left) and regiorandom P3HT:[60]PCBM (device on the right).
Figure 2.63 The electrical connections are made either by thermosetting silver epoxy (left) or by attaching adhesive aluminum tape (right).
the cell that will keep working for hours even under the poorest conditions. The first step is the application of contacts to the device. It is difficult to make point contacts to ITO, and the contact must generally be made over a large area. Two easy methods exist as shown in Fig. 2.63. Thermosetting silver epoxy glue makes a good contact. The disadvantage is that it takes some time for the glue to harden and make a reliable contact, typically 15 min at 70◦ C and much longer at room temperature. A faster method employs adhesive aluminum tape that can simply be stuck onto the evaporated electrode and the ITO electrode. Once the electrical connections have been made, the electrical characterization of the device can be performed (Fig. 2.64). Contact with or scratching of the back side of the device has to be avoided. While devices based on regiorandom P3HT do not improve the performance of annealing, devices based on regioregular P3HT can at this stage be annealed by heating to a temperature of 100–120◦ C for 5 min (Fig. 2.65). This will generally improve the performance.
The Polymer Solar Cell
67
Figure 2.64 The current-voltage diode characteristics (IV-curve) is recorded in the dark in a closed container. The device is kept in the dark box at the top (left). The IV-curve is then recorded under illumination using a sun simulator where both types of contacting schemes are shown (right).
Figure 2.65 The device is annealed on a hot plate (left) and in a preheated oven (right). Note that the oven was opened to the photographer but should be kept closed during annealing to stabilize the temperature.
The first step in the characterization of a device is recording the IV characteristics in the dark and under illumination. The device IV-curve (shown in Fig. 2.66) is measured in darkness and under AM 1.5 conditions using a solar simulator such as the Solarkonstant 575 from Steuernagel Lichttechnik GmBH, Germany. The IV-curve is recorded using a source meter such as the Keithley 2400 shown in Fig. 2.64. From the example given, the benefit of the annealing procedure is evident. The short-circuit current improves quite a lot as shown in Table 2.5. The active area of the devices is 3 cm2 . Based on this simple demonstration, anyone with experimental skills, an evaporator, and a spin-coater should be able to prepare OPV devices.
Chapter 2
68
Figure 2.66 IV-curves for the devices prepared in this section in the dark (left) and under illumination with a solar simulator (1000 W m−2 , AM 1.5).
Table 2.5 Photovoltaic properties for the devices prepared in this section and regiorandom and regioregular P3HT-[60]PCBM devices prepared in this chapter. FF = fill factor, RR = rectification ratio in the dark. The device active area was 3 cm2 . The incident light intensity was 1000 W m−2 AM 1.5.
Random Regular Regular annealed
Voc (V) 0.51 0.38 0.52
Isc (mA cm−2 ) −0.68 −5.58 −8.21
FF (%) 29 31 30
ηe (%) 0.10 0.65 1.28
RR |±1 V| 1 3 600
2.3.7 Device preparation and performance Examples of devices based on the poly(3-hexylthiophene) materials described in this chapter are given in this section. As evident from above, there is a large difference between the optical properties of P3HT in its regioregular and regiorandom form. The fast and simple approach to a polymer photovoltaic device described in this chapter was carried out in air, and in many literature reports devices are prepared entirely in air. It is also quite common to prepare devices partially in air and a glove box or entirely in a glove box. The final purpose of this section is to demonstrate the results obtained in different laboratories using exactly the same batch of polymer materials, but employing a different environment in terms of substrates, cleaning, operator, evaporator, sun simulator, and characterization method. 2.3.7.1 Device preparation and performance using regiorandom and regioregular P3HT at Risø National Laboratory (Denmark) in air
Devices were prepared using a 1:1 (w/w) mixture of P3HT and [60]PCBM in 1,2dichlorobenzene (20 mg each). The films’ active layer was spin coated (800 rpm)
The Polymer Solar Cell
69
Figure 2.67 A schematic side view of the device geometry (left). A picture showing devices based on regioregular P3HT:[60]PCBM (device to the left) and regiorandom P3HT:[60]PCBM (device to the right).
onto a PEDOT:PSS coated and etched 5–15 Ω square−1 with ITO substrates. The aluminum electrode was evaporated thermally and the active area of the device was measured to 2 cm × 1.5 cm = 3 cm2 , which can be considered as a moderately large active area (Fig. 2.67). From this point of view, there will be some ohmic losses due to the large size, and imperfections in the film will contribute to a poorer response mainly observed as a decrease in the open-circuit voltage and a decrease in the fill factor. The large-scale performance, however, does give a realistic picture of the efficiency that can be reliably obtained. After preparation, the contacts were made using silver epoxy that was allowed to thermoset at 70◦ C for 15 min. Photovoltaic characterization was carried out directly after preparation to minimize the effect of degradation caused by water and oxygen from the atmosphere. The devices were illuminated using a KHS575 solar simulator from Steuernagel Lichttechnik, Germany. The spectrum approximates AM 1.5 and the incident light intensity was set to 1000 W m−2 . The temperature of the device during measurements was 72 ± 2◦ C. After the first characterization, the devices were annealed at 115◦ C for 5 min and characterized again. In the dark there were distinct differences. Most notably the devices based on the regiorandom P3HT exhibited virtually no rectification. Furthermore, when the device was annealed the current density decreased an order of magnitude (Fig. 2.68). For the regioregular P3HT, annealing incurred a significant improvement in the rectification ratio by decreasing the current density at negative bias and increasing it at positive bias (Fig. 2.69). Under illumination, devices based on the regiorandom P3HT perform quite poorly with little rectification behavior and low fill factors. Annealing leads to poorer performance mainly due to a decrease in the short-circuit current. The devices based on regioregular P3HT perform better, while being somewhat short of the 5% benchmark.1, 6, 7 Annealing gives rise to a slight lowering of the open-circuit voltage and a significant increase in both short-circuit current and fill factor. One possible reason that the fill factor does not improve could be due to resistive losses in the nonoptimal device geometry (Table 2.6).
70
Chapter 2
Figure 2.68 Dark IV-curves for unannealed and annealed samples of regiorandom and regioregular P3HT-PCBM. The devices had an active area of 3 cm2 .
Figure 2.69 IV-curves for unannealed and annealed samples of regiorandom and regioregular P3HT-PCBM under simulated sunlight (AM 1.5 at 1000 W m−2 ). The devices had an active area of 3 cm2 .
The Polymer Solar Cell
71
Table 2.6 Photovoltaic properties of regiorandom and regioregular P3HT-PCBM devices prepared in this chapter. FF = fill factor, RR = rectification ratio in the dark. The device active area was 3 cm2 . The incident light intensity was 1000 W m−2 AM 1.5.
Random Random annealed Regular Regular annealed
Voc (V) 0.63 0.72 0.54 0.48
Isc (mA cm−2 ) −0.57 −0.37 −6.78 −8.20
FF (%) 28 26 32 36
ηe (%) 0.10 0.07 1.17 1.42
RR |±1 V| 1 1 1480 3290
Table 2.7 Photovoltaic properties for a small module based on regioregular P3HT-PCBM prepared in this chapter. FF = fill factor, RR = rectification ratio in the dark. Each of the three serially connected devices had an active area of 2 cm2 . The incident light intensity was 1000 W m−2 AM 1.5.
1st module 1st module annealed 2nd module 2nd module annealed
Voc (V) 1.00 1.50 1.27 1.52
Isc (mA cm−2 ) −5.5 −7.0 −6.3 −8.0
FF (%) 40 45 49 55
ηe (%) 0.73 1.57 1.30 2.22
Using the same material and a better device geometry, a small module was prepared by Suren Gevorgyan that had three cells in series, each with an active area of 2 cm × 1 cm = 2 cm2 , giving a total active area of 6 cm2 , but due to the serial connection a step up in the voltage was obtained, thus minimizing resistive losses. As described previously, the device was prepared and characterized in air. After preparation, it was encapsulated in a glove box.4 Two devices were prepared and they show some variation underlining the fact that device performance often improves when doing it again and that the experimenter has a choice when data is reported. Either all data can be reported; or only the best data. As faith has it, it is often only the best data that are reported, making it difficult for the newcomer to understand why his or her data are so poor. Practice and repetition is paramount. The data for the best devices are shown in Table 2.7. IV-curves and pictures of the device are shown in Fig. 2.70. The device gains performance and an increase in open-circuit voltage, short-circuit current, and fill factor after annealing. The large increase in especially the open-circuit voltage could be due to an improvement of the contacts between the cells. 2.3.7.2 Comparison between cells prepared at Risø National Laboratory (Denmark) in air and in a glove box
The previously described examples show that it is possible in a few experiments to achieve a power conversion efficiency above 1–2% using simple manipulations in air. It is also possible to prepare the entire device in a glove box. This means that
72
Chapter 2
Figure 2.70 Picture of the second module after encapsulation and under a sun simulator. The data presented were recorded before encapsulation and in ambient atmosphere. IVcurves for the unannealed and annealed module based on regioregular P3HT-PCBM under simulated sunlight (AM 1.5 at 1000 W m−2 ). The device consisted of three cells in series, each having an active area of 2 cm2 .
the ITO substrates are prepared outside the glove box and the PEDOT:PSS electrode is also applied by spin coating outside the glove box, since this is an aqueous dispersion. The ITO-PEDOT:PSS substrates are heated briefly in the glove box before application of the active layer by spin coating in the glove box. The general experience is that the device performance does not increase dramatically when performing the device preparation in a glove box as compared to preparing and handling devices in ambient atmosphere. The stability of the devices in the atmosphere is, however, not good and the devices start degrading the moment they are removed from the metal evaporator. To get comparable data for devices prepared in ambient atmosphere and in a glove box, characterization must be swift when performed in the atmosphere. In a glove box, devices are generally stable for days without significant degradation (Fig. 2.71). If the devices prepared from a particular material are subject to instability in the dark under inert conditions, they will not, needless
The Polymer Solar Cell
73
Figure 2.71 Preparation of devices in a glove box system at Risø National Laboratory (Denmark). The solutions are microfiltered (top left) and spin coated (top middle), and the coated modules are prepared for evaporation (top right). The glove-box system is shown below.
to say, benefit from being in a glove box. Devices prepared from some materials are subject to instability in the dark, such as PPV-based materials and devices prepared from materials giving a soft active layer. 2.3.7.3 Device preparation and performance using regiorandom and regioregular P3HT in CEA (France) in a glove box
The bulk heterojunction photovoltaic cells were prepared using solutions of P3HT:[60]PCBM in 1:1 weight ratios. Concentrations of the samples used were
74
Chapter 2
14 g L−1 of P3HT regioregular and regiorandom P3HT. The glass-ITO substrates (obtained from PGO, Germany) were sequentially cleaned in an ultrasonic bath with acetone and isopropanol, then rinsed with deionized water, dried in an oven at 140◦ C for 30 min, and finally treated with UV-ozone. The substrate was then spin coated with PEDOT:PSS (Baytron PH) at 1500 rpm, and oven dried for 30 min at 140◦ C to give a 40-nm film. The active P3HT:PCBM layer was deposited by spin casting from an anhydrous chlorobenzene solution heated to 40◦ C, in a dry nitrogen atmosphere (glove box) to obtain 70-nm thick films for regioregular P3HT and 150 nm for regiorandom P3HT. Following drying under reduced pressure at 2 × 10−7 mbar for 1 hr, the devices were completed by deposition of the LiF/Al (0.8 nm and 100 nm, respectively) cathode through a shadow mask with 6-mm diameter openings at approximately 2 × 10−7 mbar. All cells had an active area of 28 mm2 . The annealing process was carried out under an inert atmosphere by placing the cells directly onto a controlled hot plate. Cell performances were evaluated after free cooling to ambient temperature. Current-voltage characteristics and power conversion efficiencies of the solar cells were measured in inert atmosphere via a computer controlled Keithley source measure unit, SMU 2400, using 100 mW cm−2 (AM 1.5) simulated white light from a Steuernagel Solar Constant 575 simulator. A monocrystalline silicon solar cell, calibrated at the Fraunhofer Institut Solare Energiesysteme (Freiburg, Germany), was used as a reference cell to confirm stabilization of the 100 mW cm−2 illumination. The mismatch factor (see Chapter 3) was not taken into account. The temperature of the polymer heterojunction, measured using a thermocouple (Pt100) mounted on the ITO substrate, reached 30◦ C at the initial IV characterization. The preparation of the devices is illustrated in Figs. 2.72–2.74. The device performance was analyzed as a function of temperature. It is clear from Fig. 2.75 that the regiorandom sample does not benefit from thermal annealing, whereas the regioregular samples do, with an optimum temperature around 110◦ C. The dark rectification ratio is moderate for both regiorandom P3HT and regioregular P3HT, while it is best in the latter case with a value of ∼200 as shown in Fig. 2.76 and Table 2.8. Under simulated sunlight, the devices based on regiorandom P3HT improved with respect to fill factor but the overall efficiency decreased due to a decrease in both open-circuit voltage and short-circuit current. For the regioregular sample, large improvements in short-circuit current, fill factor, and efficiency are observed with a small decrease in open-circuit voltage. The data are summarized in Table 2.8 and Fig. 2.77. 2.3.7.4 Comparison between cells prepared under glove box conditions at Risø National Laboratory and CEA using the same materials
The device geometries employed at CEA and at Risø National Laboratory are not the same and the main difference lies in the magnitude of the active area. The
The Polymer Solar Cell
75
Figure 2.72 The glove box system for solar cell preparation at CEA. The system is commercially available from MBraun.
Figure 2.73 Spin coating the active layer inside the glove box. The PEDOT:PSS coated ITO substrate is placed on the spin coater (left). The solution containing P3HT and [60]PCBM is filtered directly onto the substrate (top right). After spinning, the sample is removed (bottom right).
76
Chapter 2
Figure 2.74 After evaporation of the aluminum electrode (not shown), the completed devices are removed from the evaporator (top left). Characterization is carried out under simulated sunlight in the glove box (top right). The completed devices have two circular cells on each substrate, each with an active area of 28 mm2 (bottom).
devices at CEA employed an active area of 0.28 cm2 , and at Risø, larger active areas of 2–3 cm2 were used for the experiment. The main influence of the active area’s size on the performance of the solar cell is that the fill factor decreases as the device area increases due to sheet resistive losses. The short-circuit current and open-circuit voltage are for small increases in active area only affected by imperfections in the device film, leading to short-circuit paths. As the active area increases,
The Polymer Solar Cell
77
Figure 2.75 Evolution of Voc , I sc , FF, and efficiency of P3HT:PCBM-based photovoltaic cells as a function of applied annealing temperature for RR and regiorandom P3HT.
Figure 2.76 Dark IV-curves for annealed samples of regiorandom and regioregular P3HTPCBM. The devices had an active area of 0.28 cm2 .
Chapter 2
78
Figure 2.77 IV-curves for unannealed and annealed samples of regiorandom and regioregular P3HT-PCBM under simulated sunlight (AM 1.5 at 1000 W m−2 ). The devices had an active area of 0.28 cm2 .
Table 2.8 Photovoltaic properties of regiorandom and regioregular P3HT-PCBM devices prepared in this chapter. FF = fill factor, RR = rectification ratio in the dark. The device active area was 28 mm2 . The incident light intensity was 1000 W m−2 AM 1.5. The devices were prepared in a glove box at CEA (France).
Random Random annealed Regular Regular annealed
Voc (V) 0.82 0.74 0.65 0.58
Isc (mA cm−2 ) −0.85 −0.60 −6.69 −8.00
FF (%) 27 31 51 59
ηe (%) 0.19 0.14 2.21 2.72
RR |±1 V| – 13 – 200
the probability for film imperfections increase and thus a poorer performance is expected. In this example, the difference in the magnitude of active area is a factor of ten: CEA employed a circular active area, while Risø National Laboratory employed a rectangular active area. As the difference in the magnitude of the active area increases beyond a factor of ten, the sheet resistive losses will start to affect the magnitude of the short-circuit current and to a lesser extent the open-circuit voltage. The type of solar simulator employed was the same in both laboratories (KHS 575 from Steuernagel Lichtechnik GmBH). As expected, the short-circuit currents were within experimental error and nearly identical, and the open-circuit voltages at Risø were only very slightly lower than those obtained at CEA. Furthermore, the main difference in the power conversion efficiency was due to the fill factor that, as expected, was lower for the larger active areas.
The Polymer Solar Cell
79
References 1. Yu, G., Gao, J., Hummelen, J.C., Wudl, F., and Heeger, A.J., “Polymer photovoltaic cells—Enhanced efficiencies via a network of internal donoracceptor heterojunctions,” Science, 270, pp. 1789–1791 (1995). 2. Weinberger, B.R., Akhtar, M., and Gau, S.C., “Polyacetylene photovoltaic devices,” Synth. Met., 4, pp. 187–197 (1982). 3. Karg, S., Riess, W., Meier, M., and Schwoerer, M., “Characterization of light-emmiting-diodes and solar-cells based on poly-phenylene-vinylene,” Synth. Met., 57, pp. 4186–4191 (1993). 4. Yu, G., Zhang, C., and Heeger, A.J., “Dual-function semiconducting polymer devices—light-emitting and photodetecting diodes,” Appl. Phys. Lett., 64, pp. 1540–1542 (1994). 5. Shaheen, S.E., Brabec, C.J., Sariciftci, N.S., Padinger, F., Fromhertz, T., and Hummelen, J.C., “2.5% efficient organic plastic solar cells,” Appl. Phys. Lett., 78, pp. 841–843 (2001). 6. Li, G., Sharotriya, V., Huang, J., Yao, Y., Moriarty, T., Emery, K., and Yang, Y., “High-efficiency solution processable polymer photovoltaic cells by selforganization of polymer blends,” Nat. Mater., 4, pp. 864–868 (2005). 7. Ma, W., Yang, C., Gong, X., Lee, K., and Heeger, A.J., “Thermally stable, efficient polymer solar cells with nanoscale control of the interpenetrating network morphology,” Adv. Funct. Mater., 15, pp. 1617–1622 (2005). 8. Kim, J.Y., Lee, K., Coates, N.E., Moses, D., Nguyen, T.-Q., Dante, M., Heeger, A.J., “Efficient tandem polymer solar cells fabricated by all-solution processing,” Science, 317, pp. 222–225 (2007). 9. Kraft, A., Grimsdale, A.C., and Holmes, A.B., “Electroluminescent conjugated polymers—Seeing polymers in a new light,” Angew. Chem., Int. Ed. 37, pp. 402–428 (1998). 10. Wessling, R.A., “The polymerization of xylene bisdialkyl sulfonium salts,” J. Polym. Sci.—Polym., 72, pp. 55–66 (1985). 11. Marks, R.N., Halls, J.J.M., Bradley, D.D.C., Friend, R.H., and Holmes, A.B., “The photovoltaic response in poly(p-phenylene vinylene) thin film devices,” J. Phys. Condens. Mater., 6, pp. 1379–1394 (1994). 12. Gilch, H.G., and Wheelwright, W.L., “Polymerization of alpha-halogenated p-xylenes with base,” J. Polym. Sci., Part A: Polym. Chem., 4, pp. 1337–1349 (1966). 13. Neff, C.F., and Ferraris, J.P., “MEH-PPV: Improved synthetic procedure and molecular weight control,” Macromolecules, 33, pp. 2311–2314 (2000).
80
Chapter 2
14. Hsieh, B.R., Yu, Y., VanLaeken, A.C., and Lee, H., “General methodology toward soluble poly(p-phenylenevinylene) derivatives,” Macromolecules, 30, 8094–8095 (1997). 15. Becker, H., Spreitzer, H., Ibrom, K., and Kreuder, W., “New insights into the microstructure of Gilch-polymerized PPVs,” Macromolecules, 32, pp. 4925– 4932 (1999). 16. Wienk, M.M., Kroon, J.M., Verhees, W.J.H., Knol, J., Hummelen, J.C., van Hal, P.A., and Janssen, R.A.J., “Efficient methano[70]fullerene/MDMOPPV bulk heterojunction photovoltaic cells,” Angew. Chem., Int. Ed., 42, pp. 3371–3375 (2003). 17. Krebs, F.C., and Jørgensen, M., “High carrier mobility in a series of new semiconducting PPV-type polymers,” Macromolecules, 36, pp. 4374–4384 (2003). 18. Granström, M., Petrisch, K., Arias, A.C., Lux, A., Andersson, M.R., and Friend, R.H., “Laminated fabrication of polymeric photovoltaic diodes,” Nature, 395, pp. 257–260 (1998). 19. Greenham, N.C., Friend, R.H., and Bradley, D.D.C., “Angular dependence of the emission from a conjugated polymer light-emitting diode: Implications for efficiency calculations,” Adv. Mater., 6, pp. 491–494 (1994). 20. McCullough, R.D., Lowe, R.D., Jayaraman, M., and Anderson, D.L., “Design, synthesis, and control of conducting polymer architectures: structurally homogeneous poly(3-alkylthiophenes),” J. Org. Chem., 58, pp. 904–912 (1993). 21. Chen, T.-A., Wu, X., and Rieke, R.D., “Regiocontrolled synthesis of poly(3alkylthiophenes) mediated by Rieke zinc: Their characterization and solidstate properties ,” J. Am. Chem. Soc., 117, pp. 233–244 (1995). 22. Sirringhaus, H., Brown, P.J., Friend, R.H., Nielsen, M.M., Bechgaard, K., Langeveld-Voss, B.M.W., Spiering, A.J.H., Janssen, R.A.J., Meijer, E.W., Herwig, P., and de Leeuw, D.M., “Two-dimensional charge transport in selforganized, high-mobility conjugated polymers,” Nature, 401, pp. 685–688 (1999). 23. Hugger, S., Thomann, R., Heinzel, T., and Thurn-Albrecht, T., “Semicrystalline morphology in thin films of poly(3-hexylthiophene),” Colloid Polym. Sci., 282, pp. 932–938 (2004). 24. Gadisa, A., and Workalemahu, B., “Photovoltaic properties of a single layer poly [3-(4-octylphenyl)-2,2 -bithiophene] (PTOPT),” Synth. Met., 129, pp. 179–185 (2002).
The Polymer Solar Cell
81
25. Roman, L.S., Arias, A.C., Theander, M., Andersson, M.R., and Inganäs, O., “Photovoltaic devices based on photo induced charge transfer in polythiophene: CN-PPV blends,” Brazil. J. Phys., 33, pp. 376–381 (2003). 26. Roman, L.S., Mammo, W., Pettersson, L.A.A., Andersson, M.R., and Inganäs, O., “High quantum efficiency polythiophene/C-60 photodiodes,” Adv. Mater., 10, pp. 774–777 (1998). 27. Scharber, M.C., Mühlbacher, D., Koppe, M., Denk, P., Waldauf, C., Heeger, A.J., and Brabec, C.J., “Design rules for donors in bulk-heterojunction solar cells—Towards 10 % energy-conversion efficiency,” Adv. Mater., 18, pp. 789– 794 (2006). 28. Koster, L.J.A., Mihailetchi, V.D., and Blom, P.W.M., “Ultimate efficiency of polymer/fullerene bulk heterojunction solar cells,” Appl. Phys. Lett., 88, 093511 1–3 (2006). 29. Winder, C., and Sariciftci, N.S., “Low bandgap polymers for photon harvesting in bulk heterojunction solar cells,” J. Mater. Chem., 14, pp. 1077–1086 (2004). 30. Chen, M., Perzon, E., Anderson, M.R., Marcinkevicius, S., Jönsson, S.K.M., Fahlman, M., Berggren, M., “1 micron wavelength photo- and electroluminescence from a conjugated polymer,” Appl. Phys. Lett., 84, pp. 3570–3572 (2004). 31. Breiby, D.W., and Samuelsen, E.J., “Quantification of preferential orientation in conjugated polymers using X-ray diffraction,” J. Polym. Sci. B: Polym. Phys., 41, pp. 2375–2393 (2003). 32. Aasmundtveit, K.E., Samuelsen, E.J., Guldstein, M., Steinsland, C., Flornes, O., Fagermo, C., Seeberg, T.M., Pettersson, L.A.A., Inganäs, O., Feidenhans’l, R., and Ferrer, S., “Structural anisotropy of poly(alkylthiophene) films,” Macromolecules, 33, pp. 3120–3127 (2000). 33. Prosa, T.J., Winokur, M.J., Moulton, J., Smith, P., and Heeger, A.J., “X-ray structural studies of poly(3-alkylthiophenes)—An example of an inverse comb,” Macromolecules, 25, pp. 4364–4372 (1992). 34. Aasmundtveit, K.E., Samuelsen, E.J., Mårdalen, J., Bakken, E., Carlsen, P.H.J., and Lienert, U., “Orientation effect in thin layers of poly(octylthiophene) on glass studied by synchrotron X-ray diffraction,” Synth. Met., 89, pp. 203–208 (1997). 35. Wudl, F., Kobayashi, M., and Heeger, A.J., “Poly(isothianaphthene),” J. Org. Chem., 49, pp. 3382–3384 (1984). 36. Kitamura, C., Tanaka, S., and Yamashita, Y., “New narrow-bandgap polymers composed of [1,2,5]thiadiazolo[3,4-g]quinoxaline and aromatic heterocycles,” Chem. Mater., 8, pp. 570–578 (1996).
82
Chapter 2
37. Jayakannan, M., van Hal, P.A., and Janssen, R.A.J., “Synthesis and structureproperty relationship of new donor-acceptor-type conjugated monomers and polymers on the basis of thiophene and benzothiadiazole,” J. Polym. Sci. A: Polym. Chem., 40, pp. 251–261 (2002). 38. Jespersen, K.G., Beenken, W.J.D., Zaushitsyn, Y., Yartsev, A., Andersson, M., Pullerits, T., and Sundström, V., “The electronic states of polyfluorene copolymers with alternating donor-acceptor units,” J. Chem. Phys., 121, pp. 12613–12617 (2004). 39. Jespersen, K.G., Yartsev, A., Pascher, T., and Sundström, V., “Excited state dynamics in alternating polyfluorene copolymers,” Synth. Met., 155, pp. 262– 265 (2005). 40. Persson, N.-K., Sun, M., Kjellberg, P., Pullerits, T., and Inganäs, O., “Optical properties of low band gap alternating copolyfluorenes for photovoltaic devices,” J. Chem. Phys., 123, 204718-1–9 (2005). 41. Yang, L., Feng, J.-K., and Ren, A.-M., “Theoretical studies on the electronic and optical properties of two thiophene-fluorene based pi-conjugated copolymers,” Polymer, 46, pp. 10970–10981 (2005). 42. Luzzati, S., Mozer, A., Denk, P., Scharber, M.C., Catellani, M., Lupsac, N.-O., Giacalone, F., Segura, J.L., Martin, N., Neugebauer, H., and Sariciftci, N.S., “Tuning of the photoinduced charge transfer process in donor-acceptor ‘double cable’ copolymers,” Proc. SPIE, 5215, pp. 41–48 (2004). 43. Catellani, M., Luzzati, S., Lupsac, N.-O., Mendichi, R., Consonni, R., Famulari, A., Meille, S.V., Giacalone, F., Segura, J.L., and Martin, N., “Donoracceptor polythiophene copolymers with tunable acceptor content for photoelectric conversion devices,” J. Mater. Chem., 14, pp. 67–74 (2004). 44. Catellani, M., Luzzati, S., Lupsac, N.-O., Mendichi, R., Consonni, R., Giacalone, F., Segura, J.L., and Martín, N., “Donor-acceptor ‘double-cable’ polythiophenes with tunable acceptor content,” Thin Solid Films, 451-452, pp. 2–6 (2004). 45. Theander, M., Yartsev, A., Zigmantas, D., Sundström, V., Mammo, W., Andersson, M.R., and Inganäs, O., “Photoluminescence quenching at a polythiophene/C-60 heterojunction,” Phys. Rev. B, 61, pp. 12957–12963 (2000). 46. Sheina, E.E., Khersonsky, S.M., Jones, E.G., and McCullough, R.D., “Highly conductive, regioregular alkoxy-functionalized polythiophenes: A new class of stable, low band gap materials,” Chem. Mater., 17, pp. 3317–3319 (2005). 47. Ogawa, K., Stafford, J.A., Rothstein, S.D., Tallman, D.E., and Rasmussen, S.C., “Nitrogen-functionalized polythiophenes: Potential routes to new low band gap materials,” Synth. Met., 152, pp. 137–140 (2005).
The Polymer Solar Cell
83
48. Andersson, M.R., Thomas, O., Mammo, W., Svensson, M., Theander, M., and Inganäs, O., “Substituted polythiophenes designed for optoelectronic devices and conductors,” J. Mater. Chem., 9, pp. 1933–1940 (1999). 49. Hou, J., Huo, L., He, C., Yang, C., and Li, Y., “Synthesis and absorption spectra of poly(3-(phenylenevinyl)thiophene)s with conjugated side chains,” Macromolecules, 39, pp. 594–603 (2006). 50. Hou, J., Yang, C., and Li, Y., “Synthesis of regioregular side-chain conjugated polythiophene and its application in photovoltaic solar cells,” Synth. Met., 153, pp. 93–96 (2005). 51. Gadisa, A., Svensson, M., Andersson, M.R., and Inganäs, O., “Correlation between oxidation potential and open-circuit voltage of composite solar cells based on blends of polythiophenes/fullerene derivative,” Appl. Phys. Lett., 84, pp. 1609–1611 (2004). 52. Zhou, E., He, C., Tan, Z., Yang, C., and Li, Y., “Effect of side-chain end groups on the optical electrochemical, and photovoltaic properties of sidechain conjugated polythiophenes,” J. Polym. Sci. A: Polym. Chem., 44, pp. 4916–4922 (2006). 53. Greenwald, Y., Cohen, G., Poplawski, J., Ehrenfreund, E., Speiser, S., and Davidor, D., “Transient photoconductivity of acceptor-substituted poly(3butylthiophene),” J. Am. Chem. Soc., 118, pp. 2980–2984 (1996). 54. Johansson, T., Mammo, W., Svensson, M., Andersson, M.R., and Inganäs, O., “Electrochemical bandgaps of substituted polythiophenes,” J. Chem. Mater., 13, pp. 1316–1323 (2003). 55. McCullough, R.D., “The chemistry of conducting polythiophenes,” Adv. Mater., 10, pp. 93–116 (1998). 56. Shi, C., Yao, Y., Yang, Y., and Pei, Q., “Regioregular copolymers of 3alkoxythiophene and their photovoltaic application,” J. Am. Chem. Soc., 128, pp. 8980–8986 (2006). 57. Dhanabalan, A., van Hal, P.A., van Duren, J.K.J., van Dogen, J.L.J., and Janssen, R.A.J., “Design and synthesis of processible functional copolymers,” Synth. Met., 119, pp. 169–170 (2001). 58. Kobayashi, M., Colaneri, N., Boysel, M., Wudl, F., and Heeger, A.J., “The electronic and electrochemical properties of poly(isothianaphthene),” J. Chem. Phys., 82, pp. 5717–5723 (1985). 59. Otto, P., and Ladik, J., “Theoretical search for low-gap polymers based on polythiophene,” Synth. Met., 36, pp. 327–335 (1990). 60. Shaheen, S.E., Vangeneugden, D.L., Kiebooms, R., Vanderzande, D., Fromherz, T., Padinger, F., Brabec, C.J., and Sariciftci, N.S., “Low band-gap polymeric photovoltaic devices,” Synth. Met., 121, pp. 1583–1584 (2001).
84
Chapter 2
61. Vangeneugden, D.L., Vanderzande, D.J.M., Salbeck, J., van Hal, P.A., Janssen, R.A.J., Hummelen, J.C., Brabec, C.J., Shaheen, S.E., and Sariciftci, N.S., “Synthesis and characterization of a poly(1,3-dithienylisothianaphthene) derivative for bulk heterojunction photovoltaic cells,” J. Phys. Chem., 105, pp. 11106–11113 (2001). 62. van Asselt, R., Hoogmartens, I., Vanderzande, D., Gelan, J., Froehling, P.E., Aussems, M., Aagaard, O., and Schellekens, R., “New synthetic routes to poly(isothianaphthene) 1. Reaction of phthalic-anhydride and phthalide with phosphorous pentasulfide,” Synth. Met., 74, pp. 65–70 (1995). 63. Hagan, A.J., Moratti, S.C., and Sage, I.C., “Synthesis of low band gap polymers: Studies in polyisothianaphthene,” Synth. Met., 119, pp. 147–148 (2001). 64. Polec, I., Henckens, A., Goris, L., Nicolas, M., Loi, M.A., Adreaensens, P.J., Lutsen, L., Manca, J.V., Vanderzande, D., and Sariciftci, N.S., “Convenient synthesis and polymerization of 5,6-disubstituted dithiophthalides toward soluble poly(isothianaphthene): An initial spectroscopic characterization of the resulting low-band-gap polymers,” J. Polym. Sci. A: Polym. Chem., 41, pp. 1034–1045 (2003). 65. Chen, S.-A., and Lee, C.-C., “Processable low band gap pi-conjugated polymer, poly(isothianaphthene),” Polymer, 37, pp. 519–522 (1996). 66. Chen, S.-A., and Lee, C.-C., “The importance of molecular-dynamics of chrystalline-structure of poly(3-dodecylthiophene),” Pure Appl. Chem., 67, pp. 1983–1990 (1995). 67. Dhanabalan, A., van Duren, J.K.J., van Hal, P.A., van Dogen, J.L.J., and Janssen, R.A.J., “Synthesis and characterization of a low bandgap conjugated polymer for bulk heterojunction photovoltaic cells,” Adv. Funct. Mater., 11, pp. 255–262 (2001). 68. van Duren, J.K.J., Dhanabalan, A., van Hal, P.A., and Janssen, R.A.J., “Lowbandgap polymer photovoltaic cells,” Synth. Met., 121, pp. 1587–1588 (2001). 69. Bundgaard, E., and Krebs, F.C., “Low-band-gap conjugated polymers based on thiophene, benzothiadiazole, and benzobis(thiadiazole),” Macromolecules, 39, pp. 2823–2831 (2006). 70. Jayakannan, M., van Hal, P.A., and Janssen, R.A.J., “Synthesis, optical, and electrochemical properties of novel copolymers on the basis of benzothiadiazole and electron-rich arene units,” J. Polym. Sci. A: Polym. Chem., 40, pp. 2360–2372 (2002). 71. Bundgaard, E., and Krebs, F.C., “A comparison of the photovoltaic response of head-to-head and head-to-tail coupled poly{(benzo-2,1,3-thiadiazol-4,7diyl)-(dihexyl[2,2 ]dithiophene-5,5 -diyl},” Polym. Bull., 55, pp. 157–164 (2005).
The Polymer Solar Cell
85
72. Wienk, M.M., Struijk, M.P., and Janssen, R.A.J., “Low band gap polymer bulk heterojunction solar cells,” Chem. Phys. Lett., 422, pp. 488–491 (2006). 73. Karikomi, M., Kitamura, C., Tanaka, S., and Yamashita, Y., “New narrowbandgap polymer composed of benzobis(1,2,5-thiadiazole) and thiophene,” J. Am. Chem. Soc., 117, pp. 6791–6792 (1995). 74. Campos, L.M., Tontcheva, A., Günes, S., Sonmez, G., Neugebauer, H., Sariciftci, N.S., and Wudl, F., “Extended photocurrent spectrum of a low band gap polymer in a bulk heterojunction solar cell,” Chem. Mater., 17, pp. 4031– 4033 (2005). 75. Wienk, M.M., Turbiez, M.G.R., Struijk, M.P., Fonrodona, M., and Janssen, R.A.J., “Low-band gap poly(di-2-thienylthienopyrazine): fullerene solar cells,” Appl. Phys. Lett., 88, 153511-1–3 (2006). 76. Kitamura, C., Tanaka, S., and Yamashita, Y., “Synthesis of new narrow bandgap polymers based on 5,7-di(2-thienyl)thieno[3,4-b] pyrazine and its derivatives,” J. Chem. Soc., Chem. Commun., pp. 1585–1586 (1994). 77. Hou, Q., Xu, Y., Yang, W., Yuan, M., Peng, J., and Cao, Y., “Novel redemitting fluorene-based copolymers,” J. Mater. Chem., 12, pp. 2887–2892 (2002). 78. Svensson, M., Zhang, F., Veenstra, S.C., Verhees, W.J.H., Hummelen, J.C., Kroon, J.M., Inganäs, O., and Andersson, M.R., “High-performance polymer solar cells of an alternating polyfluorene copolymer and a fullerene derivative,” Adv. Funct. Mater., 15, pp. 988–991 (2003). 79. Yohannes, T., Zhang, F., Svensson, M., Hummelen, J.C., Andersson, M.R., and Inganäs, O., “Polyfluorene copolymer based bulk heterojunction solar cells,” Thin Solid Films, 449, pp. 152–157 (2004). 80. Perzon, E., Wang, X., Zhang, F., Mammo, W., Delgado, J.L., de la Cruz, P., Inganäs, O., Langa, F., and Andersson, M.R., “Design, synthesis and properties of low band gap polyfluorenes for photovoltaic devices,” Synth. Met., 154, pp. 53–56 (2005). 81. Perzon, E., Wang, X., Admassie, S., Inganäs, O., and Andersson, M.R., “An alternating low band-gap polyfluorene for optoelectronic devices,” Polymer, 47, pp. 4261–4268 (2006). 82. Zhang, F., Perzon, E., Wang, X., Mammo, W., Andersson, M.R., and Inganäs, O., “Polymer solar cells based on a low-bandgap fluorene copolymer and a fullerene derivative with photocurrent extended to 850 nm,” Adv. Funct. Mater., 15, pp. 745–750 (2005). 83. Wang, X., Perzon, E., Mammo, W., Oswald, F., Admassie, S., Persson, N.-K., Langa, F., Andersson, M.R., and Inganäs, O., “Polymer solar cells
86
Chapter 2
with low-bandgap polymers blended with C-70-derivative give photocurrent at 1 μm,” Thin Solid Films, 511-512, pp. 576–580 (2006). 84. Henckens, A., Knipper, M., Polec, I., Manca, J., Lutsen, L., and Vanderzande, D., “Poly(thienylene vinylene) derivatives as low band gap polymers for photovoltaic applications,” Thin Solid Films, 451-452, pp. 572–579 (2004). 85. Brabec, C.J., Winder, C., Sariciftci, N.S., Hummelen, J.C., Dhanabalan, A., van Hal, P.A., and Janssen, R.A.J., “A low-bandgap semiconducting polymer for photovoltaic devices and infrared emitting diodes,” Adv. Funct. Mater., 12, pp. 709–712 (2002). 86. Winder, C., Mühlbacher, D., Neugebauer, H., Sariciftci, N.S., Brabec, C.J., Janssen, R.A.J., and Hummelen, J.C., “Polymer solar cells and infrared light emitting diodes: Dual function low bandgap polymer,” Mol. Cryst. Liq. Cryst., 385, pp. 93–100 (2002). 87. Bundgaard, E., Shaheen, S.E., Krebs, F.C., and Ginley, D., “Bulk heterojunctions based on a low band gap copolymer of thiophene and benzothiadiazole,” Sol. Energy Mater. Sol. Cells, 91, pp. 1631–1637 (2007). 88. Wang, X., Perzon, E., Oswald, F., Langa, F., Admassie, S., Andersson, M.R., and Inganäs, O., “Enhanced photocurrent spectral response in low-bandgap polyfluorene and C-70-derivative-based solar cells,” Adv. Funct. Mater., 15, pp. 1665–1670 (2005). 89. Bundgaard, E., and Krebs, F.C., “Low band gap polymers for organic photovoltaics,” Sol. Energy Mater. Sol. Cells, 91, pp. 954–985 (2007). 90. Smith, A.P., Smith, R.R., Taylor, B.E., and Durstock, M.F., “An investigation of poly(thienylene vinylene) in organic photovoltaic device,” Chem. Mater., 16, pp. 4687–4692 (2004). 91. Wong, W.-Y., Chan, S.-M., Choi, K.-H., Cheah, K.-W., and Chan, W.-K., “Synthesis, optical and photoconducting properties of platinum poly-yne polymers functionalized with electron-donating and electron-withdrawing bithiazole units,” Macromol. Rapid Commun., 21, pp. 453–457 (2000). 92. de Boer, B., Stalmach, U., van Hutten, P.F., Melzer, C., Krasnikov, V.V., and Hadziioannou, G., “Supramolecular self-assembly and opto-electronic properties of semiconducting block copolymers,” Polymer, 42, pp. 9097– 9109 (2001). 93. Krebs, F.C., Hagemann, O., and Jørgensen, M., “Synthesis of dye linked conducting block copolymers, dye linked conducting homopolymers and preliminary application to photovoltaics,” Sol. Energy Mater. Sol. Cells, 83, pp. 211–228 (2004). 94. Krebs, F.C., “On the photovoltaic response of the J-domain and the JAassembly,” Sol. Energy Mater. Sol. Cells, 80, pp. 257–264 (2003).
The Polymer Solar Cell
87
95. Krebs, F.C., Spanggaard, H., Rozlosnick, N., Larsen, N.B., and Jørgensen, M., “Synthesis, properties, and Langmuir-Blodgett film studies of an ionic dye terminated rigid conducting oligomer,” Langmuir, 19, pp. 7873–7880 (2003). 96. Hua, J., Meng, F., Li, J., Ding, F., Fan, X., and Tian, H., “Synthesis and characterization of new highly soluble and thermal-stable perylene-PPV copolymers containing triphenylamine moiety,” Eur. Polym. J., 42, pp. 2686–2694 (2006). 97. Lu, S., Niu, J., Li, W., Mao, J., and Jiang, J., “Photophysics and morphology investigation based on perylenetetracarboxylate/polymer photovoltaic devices,” Sol. Energy Mater. Sol. Cells, 91, pp. 261–265 (2007). 98. Barber, Jr., R.P., Gomez, R.D., Herman, W.N., and Romero, D.B., “Organic photovoltaic devices based on a block copolymer/fullerene blend,” Org. Electron., 7, pp. 508–513 (2006). 99. Liu, J.S., Kadnikova, E.N., Liu, Y.X., McGehee, M.D., and Fréchet, J.M.J., “Polythiophene containing thermally removable solubilizing groups enhances the interface and the performance of polymer-titania hybrid solar cells,” J. Am. Chem. Soc., 126, pp. 9486–9487 (2004). 100. Krebs, F.C., and Spanggaard, H., “Significant improvement of polymer solar cell stability,” Chem. Mater., 17, pp. 5235–5237 (2005). 101. Tang, C.W., “2-Layer organic photovoltaic cell,” Appl. Phys. Lett., 48, pp. 183–185 (1986). 102. Xue, J., Uchida, S., Rand, B.P., and Forrest, S.R, “Mixed donor-acceptor molecular heterojunctions for photovoltaic applications. I. Material properties,” Appl. Phys. Lett., 85, pp. 5757–5759 (2004). 103. Mutolo, K.L., Mayo, E.I., Rand, B.P., Forrest, S.R., and Thompson, M.E., “Enhanced open-circuit voltage in subphthalocyanine/C-60 organic photovoltaic cells,” J. Am. Chem. Soc., 128, pp. 8108–8109 (2006). 104. Schmidt-Mende, L., Fechtenkötter, A., Müllen, K., Moons, E., Friend, R.H., and MacKenzie, J.D., “Self-organized discotic liquid crystals for high-efficiency organic photovoltaics,” Science, 293, pp. 1119–1122 (2001). 105. Kippelen, B., Yoo, S., Haddock, J.A., Domercq, B., Barlow, S., Minch, B., Xia, W., Marder, S.R., and Armstrong, N.R., Organic Photovoltaics Mechanisms, Materials and Devices, S.-S. Sun and N. S. Sariciftci (Eds.), Chap. 11, CRC Press, Boca Raton (2005). 106. Hummelen, J.C., Knight, B.W., LePeq, F., and Wudl F., “Preparation and characterization of fulleroid and methanofullerene derivatives,” J. Org. Chem., 60, pp. 532–538 (1995).
88
Chapter 2
107. Kooistra, F.B., Knol, J., Kastenberg, F., Popescu, L.M., Verhees, W.J.H., Kroon, J.M., and Hummelen, J.C., “Increasing the open circuit voltage of bulk-heterojunction solar cells by raising the LUMO level of the acceptor,” Org. Lett., 9, pp. 551–554 (2007). 108. Padinger, F., Rittberger, R.S., and Sariciftci, N.S., “Effects of postproduction treatment on plastic solar cells,” Adv. Funct. Mater., 13, pp. 85–109 (2003). 109. Nierengarten, J.-F., Eckert, J.-F., Nicould, J.-F., Ouali, L., Krasnikov, V., and Hadziioannou, G., “Synthesis of a C-60-oligophenylenevinylene hybrid and its incorporation in a photovoltaic device,” Chem. Commun., pp. 617–618 (1999). 110. Peeters, E., van Hal, P.A., Knol, J., Brabec, C.J., Sariciftci, N.S., Hummelen, J.C., and Janssen, R.A.J., “Synthesis, photophysical properties, and photovoltaic devices of oligo(p-phenylene vinylene)-fullerene dyads,” J. Phys. Chem. B, 104, pp. 10174–10190 (2000). 111. Guldi, D.K., Luo, C., Swartz, A., Gómez, R., Segura, J.L., Martín, N., Brabec, C., and Sariciftci, N.S., “Molecular engineering of C-60-based conjugated oligomer ensembles: Modulating the competition between photoinduced energy and electron transfer processes,” J. Org. Chem., 67, pp. 1141–1152 (2002). 112. Roncali, J., “Linear pi-conjugated systems derivatized with C-60-fullerene as molecular heterojunctions for organic photovoltaics,” Chem. Soc. Rev., 34, pp. 483–495 (2005). 113. Segura, J.L., Martín, N., and Guldi, D.M., “Materials for organic solar cells: the C-60/pi-conjugated oligomer approach,” Chem. Soc. Rev., 34, pp. 31–47 (2005). 114. Meier, H., Gerold, J., Kolzhorn, H., and Müling, B., “Extension of conjugation leading to bathochromic or hypsochromic effects in OPV series,” Chem. Eur. J., 10, pp. 360–370 (2004). 115. Jørgensen, M., and Krebs, F.C., “Stepwise and directional synthesis of endfunctionalized single-oligomer OPVs and their application in organic solar cells ,” J. Org. Chem., 69, pp. 6688–6696 (2004). 116. Jørgensen, M., and Krebs, F.C., “Stepwise unidirectional synthesis of oligo phenylene vinylenes with a series of monomers. Use in plastic solar cells,” J. Org. Chem., 70, pp. 6004–6017 (2005). 117. Hagemann, O., Jørgensen, M., and Krebs F.C., “Synthesis of an all-in-one molecule (for organic solar cells),” J. Org. Chem., 71, pp. 5546–5559 (2006).
The Polymer Solar Cell
89
118. Cravino, A., Roquet, S., Alévêque, O., Leriche, P., Frère, P., and Roncali J., “Triphenylamine-oligothiophene conjugated systems as organic semiconductors for opto-electronics,” Chem. Mater., 18, pp. 2584–2590 (2006). 119. McCollough, R.D., Lowe, R.D., Jayaraman, M., and Anderson D.L., “Synthesis and physical properties of regiochemically well-defined, head-to-tal coupled poly(3-alkylthiophenes),” J. Org. Chem., 58, pp. 904–912 (1993). 120. Hummelen, J.C., Knight, B.W., LePeq, F., Wudl, F., Yao, J., and Wilkins C.L., “Preparation and characterization of fulleroid and methanofullerene derivatives,” J. Org. Chem., 60, pp. 532–538 (1995).
Chapter 3
Characterization of Organic Solar Cells Eugene A. Katz, Kion Norrman, Eva Bundgaard and Frederik C. Krebs This chapter guides you through electrical and physical characterization of photovoltaic devices. There is a special focus on the practical aspects of calibrating sun simulators and accurately determining power conversion efficiencies for devices under simulated and real sunlight conditions.
3.1 Taking the Sun Inside The sun is a reliable source of luminous energy with an essentially constant power output. From this point of view, it is an excellent light source when testing the power conversion efficiency of a device and it is of course also the real source of light energy against which one has to measure the efficiency of a given device when it comes to energy production using a photovoltaic in a real application. As practice would have it, the incident intensity of sunlight at the surface of the earth is, in many regions, highly variable and depends on the weather (cloud cover, rain, snow, etc.), the time of the day (day, night), and the time of the year (summer, winter). For this reason it is practical to have a light source that can be used inside the laboratory such that measurements can be carried out repeatedly at any point in time and also the source can be switched off. As discussed earlier, the spectral distribution of the sunlight follows blackbody radiation from a source with the surface temperature of the sun (5800 K). In space the distribution is continuous and near perfect, while at the surface of the earth various atmospheric gasses absorb part of the spectrum giving it a particular spectrum depending on where you are on the globe. The light sources that are available in the lab are exclusively electric arc lamps (while chemical light sources are possible by burning gasses with a high flame temperature). For this reason the spectrum of simulated sunlight cannot exactly reproduce the spectrum of the sun. Generally there is a tendency to have too much ultraviolet and visible light intensity and too little infrared light intensity. Furthermore, the spectrum often has discrete emission lines from the elements present in the arc lamp.
92
Chapter 3
In practice, different filters are employed and this can significantly improve on the resemblance to the solar spectrum. 3.1.1 Air mass As described above, the spectrum of the sunlight changes as it is transmitted through the earth’s atmosphere as shown below. Aside from the transmission loss, there is selective absorption in certain wavelength regions by atmospheric gasses. The amount of atmosphere that the light has to pass before reaching the surface of the earth is described as the air mass. The spectrum obtained after passage through a certain air mass is often described by the abbreviation AM followed by a numeric figure that, in essence, indicates how much air mass the sunlight has been transmitted through. In space just outside the earth’s atmosphere, the spectrum of the sun is virtually undisturbed since it has not passed through any atmosphere but mostly empty space. The solar constant there is 1366.1 W m−2 and the spectrum is termed AM 0 according to the ASTM E490-00 standard. The shortest distance light can travel through the atmosphere to reach the surface of the earth is at the equator. There, the sun’s spectrum is termed AM 1.0. At the higher latitudes most common in northern Europe and the United States, the sunlight has to travel longer distances through the atmosphere to reach the surface, and the air mass there is termed AM 1.5, which corresponds to a receiving surface tilted 37 deg toward the equator. Alternatively it can be viewed as a receiving surface where the sun has an elevation of 48.2 deg at zenith (Fig. 3.1). Aside from the transmission loss through the atmosphere due to scattering, there are some distinct bands due to absorption mainly by oxygen, ozone, water, and carbon dioxide in the atmosphere. Ozone absorbs efficiently in the ultraviolet region and water absorbs at 940 nm and at 1130 nm. Both water and carbon diox-
Figure 3.1 Illustration of the air masses AM 0, AM 1.0 and AM 1.5. The size of the earth, sun, distances and angles arenot drawn to scale for the purpose of illustration. Since the sun is much larger than earth the incident light field that bathes the earth can be considered as planar (left). Illustration of the receiving surface for the solar cell along with a calculation of the air mass. The normal of the receiving surface is tilted by an angle, t, towards the equator. The angle between the ground normal and the sun at zenith, θ, is used to calculate the air mass as the inverse of cos(θ). For AM 1.5 the tilt angle t = 37 deg and the zenith angle θ = 48.2 deg (right).
Characterization of Organic Solar Cells
93
Figure 3.2 Comparing the sun spectrum in space just outside the earth’s atmosphere (AM 0) and the sun’s spectrum as received at the surface of the earth at a latitude corresponding to northern Europe or the United States (AM 1.5G). The inset shows the spectrum up to 2000 nm on a logarithmic scale.
ide absorb at 1400 nm, 1850 nm, and 2700 nm. Oxygen gives rise only to minor absorption bands. 3.1.1.1 The AM 1.5D (or simply AM 1.5) versus the AM 1.5G spectrum
Data on the emission from the sun can be downloaded from National Renewable Energy Laboratory (NREL) in Colorado, U.S., and is contained in the standard tables for solar spectral irradiance.1 While the origin of the AM 0, AM 1, AM 1.5 (AM 1.5D), and AM 1.5G is well described, the subtle differences are not commonly appreciated by experimenters characterizing polymer photovoltaics under simulated sunlight. The AM 1.5 spectrum is also called the AM 1.5D where the D stands for direct, and circumsolar is the easiest spectrum to obtain reproducibly. It is simply the spectrum obtained when viewing the sun through a circular aperture with an opening angle of 5.8 deg (the sun being at the center of the aperture). The sunlight that is received through such an aperture includes the irradiance from the solar disk plus irradiation from the corona. The total power contained in this spectrum obtained by integration amounts to 901 W m−2 . Thus, this is the power that a photovoltaic will receive per unit area when viewing the sun. For instance, a black tube with a length defined such that the opening angle as seen from the position of the solar cell is 5.8 deg. Such a setup is well-defined but impractical from an experimenters point of view since the platform has to track the motion of the sun across the sky in order to keep the solar disk in the middle of the aperture. Furthermore, not all the available sunlight is received by the solar cell in this manner because there is a considerable amount of scattered and diffused light that can be harvested. From this point of view, a more practical setup is to use a viewing platform that faces south (for northern latitudes) and is tilted at an angle with respect to the surface of the earth at the viewing location. For AM 1.5, this angle is 37 deg.
94
Chapter 3
Figure 3.3 A comparison bewteen the AM 1.5 (AM 1.5D) and the AM 1.5G spectra. The integral of the AM 1.5G is ∼1000 W m−2 and includes contributions from difusely scattered light and ground reflection from light soil. The differences between the two spectra is only in the visible range, and at long wavelengths the two spectra are nearly identical.
Such a receiving surface receives the direct and circumsolar sunlight, but also reflection from the ground in front of the viewing platform and diffusely scattered light from the truncated hemisphere above the viewing surface. Light from the sky dome to the north of the viewing platform is not received. The total power received at such a surface is 1002 W m−2 and is called AM 1.5G, where G stands for global. The larger amount of energy in the AM 1.5G spectrum is concentrated at shorter wavelengths (Fig. 3.3) The reason for the higher concentration of photons at short wavelengths in the AM 1.5G spectrum is due to scattering of the light out of the direct beam by aerosols and gas molecules in the atmosphere. This also houses the explanation as to why the sky is blue when observed from earth. Since scattering is wavelength dependent, the effect is most dominant at short wavelengths. AM 1.5G also contains a contribution from ground reflection in front of the tilted viewing surface. From this point of view, AM 1.5G is the correct standard to use because it is what one would typically experience during an experiment on a tilted viewing surface, but it is not as rigorously defined as AM 1.5D since many components enter into the formation of it (surface albedo, aerosol depth, water vapor, ozone level carbon dioxide, etc.). The most the experimenter can do in the laboratory is to try to match the AM 1.5G spectrum and power as much as possible. 3.1.2 The ASTM E 927-05 standard and the IEC 904-9 standard The sun simulators for indoor testing of photovoltaic devices can be classified according to two standards: the ASTM E 927-05 standard and the IEC 904-9 standard.2 The performance of the sun simulator in three categories, that is, spectral match, spatial nonuniformity, and temporal instability, is classified in three classes,
Characterization of Organic Solar Cells
95
Table 3.1 Spectral distribution of irradiance performance requirements (small and large area simulators) for ASTM E 927-05.
Wavelength interval (nm) 300−400 400−500 500−600 600−700 700−800 800−900 900−1100 1100−1400
Percent of total irradiance Direct AM 1.5 Global AM 1.5 – – 16.9 18.4 19.7 19.9 18.5 18.4 15.2 14.9 12.9 12.5 16.8 15.9 – –
AM 0 8.0 16.4 16.3 13.9 11.2 9.0 13.1 12.2
Table 3.2 Reference spectral irradiance distribution for IEC 904-9.
Wavelength interval (nm) 400−500 500−600 600−700 700−800 800−900 900−1100
Percentage of total irradiance (400–1100 nm) 18.5 20.1 18.3 14.8 12.2 16.1
A, B, and C, and for a light source to be classified as a sun simulator, it has to be a least a class C. The spectral match is the ratio of the actual percentage of total irradiance to the required percentage specified in Table 3.1 and 3.2 for the ASTM E 927-05 standard and the IEC 904-9 standard, respectively. The spatial nonuniformity of irradiance is defined in Eq. (3.1), where the maximum and minimum irradiance is measured with the detector over the designated test area: nonuniformity = ±
max irradiance − min irradiance × 100. max irradiance + min irradiance
(3.1)
The temporal instability is defined in Eq. (3.2), where the maximum and minimum irradiance is measured with the detector at any particular point on the test plane during the time of data acquisition: temporal instability(%) = ±
max irradiance − min irradiance × 100. (3.2) max irradiance + min irradiance
The ASTM E 927-05 standard2 divides the simulators into two groups, that is, large and small areas, and the specifications for these simulators in the three classes are summarized in Table 3.3 and 3.4.
Chapter 3
96
Table 3.3 Classification of small area simulator performance. Small area has a test area of 30 × 30 cm or for a circular test area a diameter of 30 cm.
Classification
Class A Class B Class C
Characteristics Spectral match 0.75–1.25 0.6–1.4 0.4–2.0
Spatial nonuniformity (%) 2 5 10
Temporal instability (%) 2 5 10
Table 3.4 Classification of large area simulator performance. Large area has a test area of greater than 30 × 30 cm or for a circular test area a diameter greater than 30 cm.
Classification
Class A Class B Class C
Characteristics Spectral match 0.75–1.25 0.6–1.4 0.4–2.0
Spatial nonuniformity (%) 3 5 10
Temporal instability (%) 2 5 10
Table 3.5 Simulator classification.
Characteristic Spectral matcha Non-uniformity of irradiance Temporal instability
Class A 0.75–1.25 ≤±2% ≤±2%
Class B 0.6–1.4 ≤±5% ≤±5%
Class C 0.4–2.0 ≤±10% ≤±10%
a
The ratio of the actual percentage of total irradiance to the required percentage specified in Table 3.2 for each wavelength interval.
In the IEC 904-9 standard2 the simulators are not divided into groups; the classifications from this standard are summarized in Table 3.5. There are specific requirements, which the manufacturer has to supply with the sun simulator, and these are (1) date of issue, (2) manufacturer, (3) type of simulator, (4) date(s) of measurements used to determine the classification, (5) defined test area size, (6) distance between light source and test plane, (7) test plane depth, (8) classes for spectral match, spatial nonuniformity and temporal instability, (9) maximum and minimum irradiance, (10) spectral distribution data, (11) repeatability data, (12) map of nonuniformity of irradiance measured over the specified test area, (13) summery of temporal instability determination, (14) measurement methods used to determine classification categories, (15) percentage of total irradiance of the simulator that falls within a 30 deg field of view, and (16) recommended time interval for verification of classification.
Characterization of Organic Solar Cells
97
Figure 3.4 SolarKonstant KHS 575 solar simulators (Steuernagel Lichttechnik, Germany).
3.1.3 Types of simulators There are several solar simulators that can be used for solar cell experimenters and as such they all rely on the arc lamp that is used in terms of the quality of the spectrum. With respect to uniformity and temporal stability, the design of the solar simulator becomes important (particularly the lamp housing and electronics in the power supply). In Fig. 3.4, the solar simulator of Steuernagel Lichttechnik GmBH, Germany, is shown and the corresponding solar spectrum obtained with this simulator is shown in Fig. 3.5 in a comparison with the AM 1.5G solar spectrum. Figure 3.5 shows that the spectra obtained from the solar simulator has a higher irradiance in the ultraviolet (UV) area, and a lower irradiance in the near infrared (NIR) area of the spectra compared to the AM 1.5G sun spectra. In Fig. 3.6 some of the available solar simulators are shown; the characteristic data from these simulators are summarized in Table 3.6. 3.1.4 Halogen lamps Many laboratories do not have a certified sun simulator or a lamp that accurately simulates the real sun spectrum, while they do have the capacity to prepare solar cells. Often the characterization of the devices is performed using halogen lamps as shown in Fig. 3.7. These are good and stable sources of light commonly avail-
Chapter 3
98
Figure 3.5 A comparison of the AM 1.5G spectrum of the sun with the filtered spectrum obtained from a sun simulator.
Figure 3.6 SPI-cell tester (Courtesy of Spire Solar, United States),3 SS150 (Courtesy of Sciencetech, United States),4 and SS400A (Courtesy of Photo Emission Tech., Inc., United States).5
Table 3.6 Data for the solar simulators shown in Fig. 3.6.
Data Power output (mW cm−2 ) Spectra obtained with filters
SPI-cell tester 70–110 AM 1.5
SS150 136 AM 0, AM 1, AM 1.5, AM 2 Spectral range (nm) – 250–2500 Illumination sample size (cm2 ) 4.41 2.5 Standard/class ASTM E927/A IEC 904-9/A
SS400A 100 ± 15% AM 0, AM 1, AM 1.5 – 16 ASTM E927/A
Characterization of Organic Solar Cells
99
Figure 3.7 Pictures of various halogen light bulbs from 400 W down to 20 W. The power supply depends on the light bulb and can be either 240 V ac or 12–36 V dc. The scale of the ruler is in centimeters.
able, and as such are well suited to illuminate most organic photovoltaics. The intensity of light is a wide distribution concentrated in the 400–800-nm range. The spectral distribution is nearly independent of the wattage, shape, and manufacture of the light bulb (some bulbs have color filters, which of course changes the whitelight spectrum adversely). From this point of view, it is an excellent choice but there is a definite lack of intensity in the UV and NIR region, making the spectrum a relatively poor match with the sun spectrum, and if used as a light source for efficiency measurements the power conversion efficiency of the device is easily overestimated because all the light intensity is concentrated in the region where the organic photovoltaic exhibits the highest responsiveness. From a scientific view, the use of a halogen light can be justified if one is certain that there is no color filter in the halogen lamp and provided that the output power is measured reliably by bolometric means in conjunction with spectral analysis or by using a reference photodiode and correcting for mismatch (see Section 3.3.3 and Refs. [24,25,26]). This will make it possible for others to repeat the experiment. 3.1.5 Recording the spectrum Accurate optical spectrum analyzers have, with the advent of control by the personal computer, become low priced and readily available within a simple laboratory budget. An example is the AvaSpec-2048 from Avantes shown in Fig. 3.8, which is a complete optical spectrum analyzer that works in conjunction with a
100
Chapter 3
Figure 3.8 A calibration lamp (AvaLight DHS) containing both a deuterium and a halogen light source (left), and the optical spectrum analyzer (AvaSpec 2048) (right). The fiber optic cable with a probe head (red cable) enters the front of the analyzer and the power supply, and a USB cable enters the back.
personal computer. A light source is also available for calibration. It is, however, a very delicate matter to calibrate the spectrometer such that the spectral irradiance is measured and this is really a science in itself and requires great skill and certification. Such resources are beyond the expectation for OPV scientists who most probably concentrate efforts on making materials and devices. To the experimenter with a practical focus, the use of a simple low-cost spectrometer with a calibration lamp allows for qualified estimates of the spectral mismatch of the solar simulator, and accurate measurements of the power conversion efficiency can be made. An estimate is that the measurement of the power conversion efficiency can be performed with an accuracy of ±15% or better if care is taken with bolometric power measurement and spectral evaluation. A bolometric power measurement is carried out with a device that measures the heat generated when light impinges on a black surface thermally insulated. Such a device has a completely linear response in the wavelength range of 280–4000 nm, and is often called a pyranometer, pyrheliometer, or a radiometer. 3.1.6 Applying filters to improve the spectrum As discussed above, it is quite common for most arc lamps to produce too much light intensity in the UV-vis region and too little in the near infrared–infrared (NIRIR) region. Arc lamps have various elements in the gas that give spectral lines (es-
Characterization of Organic Solar Cells
101
pecially in the NIR). Most suppliers of solar simulators also supply filters that can efficiently remove the UV component and reduce the visual component, thus counterbalancing the effect of lower intensity in the NIR region. There is, however, no way out of the deficiency of the NIR unless a second lamp is employed. Such simulators are available with several different lamps and multiple filtering that even has stop filters in the specific regions where the atmospheric components absorb. Such solar simulators are of course highly attractive but possibly beyond the scope of the ordinary OPV experimenter, both in terms of cost and in terms of certification. 3.1.7 Spectral, temporal, and spatial homogeneity of the light field The SolarKonstant KHS 575 is a class A sun simulator up to 800 nm as measured by an Avaspec 2048 spectrometer calibrated for irradiance measurements using a halogen reference lamp (see Tables 3.1 and 3.2). Since most OPVs, aside from devices based on low bandgap materials, are responsive in the region of 400–800 nm. This means that such a simple simulator is quite efficient, while there may be a small tendency to overestimate the power conversion efficiency due to the lack of intensity in the NIR region. When bolometric intensity calibration is performed and the power is set to, say, 1000 W m−2 , this implies that the intensity is artificially high in the 400–800-nm region. Furthermore, the spectrum and intensity of a sun simulator change in time due to aging of the light bulb. Arc lamps are particularly subject to this effect since the arc length change as the electrodes burn out, depending on the geometry of the lamp. An example is shown in Fig. 3.9 where light bulbs of different ages are shown. The luminous output power can be recorded during measurements and this is highly recommended when performing long time measurements such as stability and accelerated tests (see Chapter 4). An example is shown in Fig. 3.10, where the luminous output power was measured using a bolometric pyranometer from Eppley Laboratories during the lifetime of a light bulb such as the ones shown in Fig. 3.9. The light intensity drops by about 5% during the lifetime of the light bulb and has a temporal instability during the day of 5%. It is thus a class B sun simulator with respect to temporal stability according to Tables 3.3–3.5. It is noticeable that the device current (also shown in Fig. 3.10) also drops by a certain amount. It is, from an experimental point of view, difficult to justify corrections in device degradation as a result of temporal instability of the solar simulator. In terms of the nonuniformity of the light field, the SolarKonstant KHS 575 performs moderately over a small area as shown in Fig. 3.11, and compared to the qualifications in Tables 3.3–3.5, it is a class C over a circular area with a diameter of 30 cm. Local variations over smaller areas in the center of the illuminated area are quite small, and variation in the center of a circular area with a diameter of 10 cm is less than 2%. The advantage of the SolarKonstant KHS 575 is that a large area is illuminated and it is possible to test large cells (40 cm × 40 cm), albeit at a lower intensity of 0.5–0.7 suns.
102
Chapter 3
Figure 3.9 Metal-halide light bulbs from OSRAM (HMI 575 W/GS) used in the Steuernagel SolarKonstant KHS 575. Notice how the electrodes in the arc have degraded as a function of age. A new bulb (top), slightly used (second from above), a broken bulb with one electrode broken (third from above), and a broken bulb with both electrodes broken (bottom).
Figure 3.10 Light intensity from a Steuernagel SolarKonstant KHS 575 sun simulator as a function of time at constant power setting. The measurement was performed with a precision radiometer (a bolometric device also called a pyranometer) from Eppley Laboratory Inc. The light output degrades by about 5% during the 950 hr lifetime of the light bulb. This is partly reflected in the current output from a device under test. While it is possible to correct for the changes in the incident light intensity, care has to be exercised and it is best to simply record the incident light intensity and the cell performance.
Characterization of Organic Solar Cells
103
Figure 3.11 The homogeneity of the light intensity under a Steuernagel KHS 575 solar simulator as measured with a bolometric pyranometer from Eppley Laboratory Inc. The light field is homogenous over a considerable area, while there are hot and cold spots around the edges of the 27-cm × 27-cm test area. The scale is in watts per unit area.
The largest variation in nonuniformity is close to the corners with hot and/or cold spots at the sides of the illuminated square. This is due to the rectangular shape of the reflector surface in the housing of the solar simulator. 3.1.8 Calibration of the sun simulator Once the sun simulator has been switched on and has stabilized (typically in half an hour) the calibration can be performed. The spectrum can be checked with an optical spectrometer as described in Section 3.1.5, but the most important exercise for the experimenter is to make sure that the incident light intensity is the desired one. Many laboratories employ a silicon photodiode in conjunction with a correction for spectral mismatch (see Section 3.3.3 and Refs. [24,25,26]) and this is in principle acceptable to monitor the state of the lamp and changes in intensity as a function of time, but the incident power is most reliably measured by a device that measures power. A photodiode is also a solar cell and has a spectral response. Typically, photodiodes are based on silicon and do not work above 1240 nm (or energies below the bandgap of the semiconductor). In order to measure the power accurately, a device that has a flat response over the entire wavelength region is needed: a bolometric pyranometer is such a device. It essentially measures the heating of a surface using an accurate thermometer; thus, the pyranometer measures the total
104
Chapter 3
energy delivered to the surface as power per area or J s−1 area−1 . A bandgap device only counts the photons impinging at a given area (i.e., number of photons s−1 area−1 ) and is thus not directly related to the incident power. It is highly advisable to have an integrated pyranometer in the test platform as shown in Fig. 3.12, where a small high-temperature pyranometer (CM4 from Kipp & Zonen) is shown. The advantage of this pyranometer is that it is compact and reliable. When mounted directly on the test platform, one always has a measure of the incident light intensity and it can be recorded during measurements. It also has a built-in thermocouplebased thermometer (Pt100). Once the approximate light intensity has been adjusted on the power supply of the solar simulator, the elevation table can be adjusted for the exact incident power that is desired. It is also highly advisable to have a precision pyranometer such as the one available from Eppley Laboratories shown in Fig. 3.13. It is highly accurate and can also
Figure 3.12 A test platform that can be placed under the sun simulator with a built-in pyranometer for continuously monitoring the incident power. The pyranometer is a hightemperature radiometer from Kipp & Zonen, model CM4.
Figure 3.13 Pictures of a very reliable bolometric pyranometer, the precision radiometer from Eppley Laboratory Inc. It is very accurate, certified, and provides an accurate measure of the incident light intensity. It has a double dome and is therefore very suited for outdoor measurements.
Characterization of Organic Solar Cells
105
be used for outdoor measurements. To conclude on the setting up and calibrating of the solar simulator, the most reliable procedure requires that the incident power at the sample position is measured bolometrically using a pyranometer. Furthermore, it is advisable to record the spectrum of the incident light to get a measure of the mismatch.
3.2 IV-Curves and Efficiencies When the solar simulator is switched on, stabilized, and calibrated, the experimenter is ready to conduct measurements on devices. The most common method for characterization in the laboratory involves recording the electrical response of the OPV device in the dark and under illumination. The electrical response is measured using a source-measure unit (SMU), also called a source meter. The source meter has the capacity to apply a voltage to a two-terminal device under test (DUT), such as a solar cell, and measure the current. Alternatively, a current can be passed through the device and the voltage can be measured. Most commonly and practically, the voltage is applied and the current is measured. The reason for this lies in the fundamental nature of the OPV device, which is a thin film. It is rather unpredictable what current range a given device will respond in and the passage of a particular current set by the experimenter may require unrealistic voltages, leading to breakdown and destruction of the device. Therefore, it is almost exclusively the voltage that is swept. Typical voltage ranges for single devices are from −1 V to +1 V, vice versa. 3.2.1 The source meter There are many commercially available source meters but the most extensively (and almost exclusively) used source meter is the Keithley 2400 SourceMeter shown in Fig. 3.14. It has a tangible size and is accurate and reliable. Furthermore, it has many possible means for interfacing to computers such as a serial interface, IEEE-488 interface, and an Ethernet port (IP address). It is possible to perform two-contact and four-contact measurements, but for most OPVs, two-contact measurements are performed. The electrical characteristics as measured by the source meter traces a continuous line trough 2 or 3 of the quadrants in the IV plane, as shown in Fig. 3.15. The connections to the device decide how the curve is traced through the IV plane. The connections are normally given, but they do depend on the operation of the device. It is thus instructive to discuss how an electrical current ends up in the electrical device and why it has the sign it has. 3.2.2 Where the electrons are and how to connect your cell to the outside world One of the most confusing issues when dealing with chemical systems and electrical currents is the definition of the plus and the minus! By this we do not mean in
Chapter 3
106
Figure 3.14 The Keithley 2400 SourceMeter (two units are shown).
Figure 3.15 The IV plane illustrated with a two-terminal device and the associated electrical field; a current in the different quadrants of the plane.
terms of charge, but in much simpler human terms. In this section we answer simple questions. For instance, when you connect the red wire to the circuit, how does this affect transport of an electrical current? This may seem a basic question, perhaps too basic for an account of this nature; however, it has always been problematic that experimenters get it wrong (often intuitively) when thinking about connecting the completely consistent molecular circuitry to the external world. Three important
Characterization of Organic Solar Cells
107
aspects can be used to account for this limited human aptitude and they are listed below. 1. The sign convention in chemistry has been dominated by electrochemists who traditionally define the positive electrode according to the chemistry that takes place at the electrode. A battery is a good example, where the positive electrode depends on the process, that is, charging or discharging. 2. The sign convention in electronics has been dominated by electrical engineers who base their ideas on the fact that most conductors of electricity conduct by means of transporting electrons, that is, they are electronic conductors. Furthermore, they distinguish between a circuit element that generates an electrical current and a circuit element that dissipates electrical current. 3. Finally, there is the problem that an electrical circuit may operate without us ever knowing which carrier is responsible for the transport, that is, we can observe a transport of electrical current through a device and we rarely know whether transport is by electrons in one direction, by holes in the opposite direction, or by both mechanisms. By making a short repetition throughout this text of simple physics and how it is applied, it is our hope that the reader may benefit. Two simple definitions help us and of an electrical current, i: here, namely, the definition of the electrical field, E, = − dV , E dx dq i = . dt
(3.3) (3.4)
By definition, the direction of the electrical field vector is from plus to minus and a positive current always flows in the direction of the electrical field by means of positively charged species (also called holes). Alternatively, a positive current flow in the direction of the electrical field could be by means of negatively charged species moving in a direction opposing the electrical field vector as illustrated in Fig. 3.16. This can be a problem for the experimenter because she or he may never know if the current results from holes moving in one direction or electrons moving in the other direction. To the physicist, the above situations are distinguishable and there are techniques available to probe the nature of the preponderant carrier type. To the electrical engineer, however, the result is the same; the electrical current follows the direction of the electrical field. This is a short description of the behavior of carriers of electrical current at the microscopic level in a theoretical conductor. The next level of the ladder of complexity is the consideration of a circuit element. Electronic devices with two connecting wires may either consume electrical power and are then termed passive elements, or may generate electrical power and are then termed active elements. The passive element is straightforward since the current enters the positive terminal and the behavior is just like that described in Fig. 3.16.
108
Chapter 3
Figure 3.16 Transport of an electrical current in a conductor follows the electrical field in a passive device. The carriers can be holes flowing in the direction of the electric field (left), electrons flowing in the opposite direction to the electrical field (middle), or both (right).
Figure 3.17 Current flow in an active element (left), a passive element (middle), and a complete circuit (right).
The active element, a generator of an electrical current, must, however, have the ability to send the current in the opposite direction against the direction of the electrical field. In the active element, the electrical current exits the device via the positive terminal as shown Fig. 3.17. We can connect an active and a passive element in a complete electrical circuit where current is passed around. The distinction between an active and a passive element can also be made by making a simple plot with the voltage across the terminals of the device along the x-axis and the current through the device along the y-axis as shown in Fig. 3.15. 3.2.3 Speed of IV-curve measurement, dielectric relaxation, and capacitive loading The speed of recording the IV-curve is, for small devices under ideal conditions, unimportant and the IV data can be recorded at arbitrary speeds and in arbitrary directions (i.e., −1 V → +1 V or +1 V → −1 V). When performing measurements on devices of larger areas, the capacitance of the device may influence the
Characterization of Organic Solar Cells
109
measurement. An OPV device is also a capacitor and is charged by an amount of current that can be stored at the given voltage according to the capacitance. When changing the applied voltage, the device has to charge or discharge according to the position in the IV plane and the direction of change. The impedance of the system determines how fast this process takes place and the time constant for this process is given by τ = RC ,
(3.5)
where τ is the time constant in seconds, R is the resistance through which the device has to charge/discharge in ohms, and C is the capacitance of the device in farads. Normally, the impedance of the source meter is large but many systems perform the measurement by subjecting the device under test through a low-impedance period for the purpose of charging/discharging; for such source meters this is not a problem. The impedance during measurement may be very high (109 Ω). The capacitance for OPV devices depends on the thickness of the active layer. Typical values for a 100-nm thick device have relative dielectric constants, εr , for polymer blends of around 2–4, giving device capacitances of around 20–40 nF cm−2 . If the device has a low impedance due to shunts, there will not be any problem with capacitive charging and in general this is not observed. It can always be tested by recording IV-curves very slowly and very quickly in both sweep directions. The curves should be superimposable, but if they are not, the speed should be decreased. Another related effect that can take place is dielectric relaxation in the materials of the active layer. Especially soft, lossy dielectrics exhibit this behavior where the molecular dipoles are rearranged in response to the electric field. The process is indistinguishable from the capacitive charging/discharging effect except for the time dependence. The advice is to make sure that there is no dependence of the IV-curve on speed and direction. 3.2.4 Action spectra using a high-power spectrometer Another characterization technique that is important is the action spectrum of the OPV. It shows how the solar cell operates at the different wavelengths of the solar spectrum. In general, the action spectrum follows the absorption spectrum of the active layer, but there may be processes taking place that prevent this from happening, such as internal conversion and emission phenomena. The devices prepared in Chapter 2 were characterized using this setup (Fig. 3.18). The action spectra are shown in Fig. 3.19. While both the current and voltage can be recorded, the action spectrum is best recorded as the short-circuit current as a function of wavelength. The reason for this is due to potential problems with capacitive charging when measuring voltage; and furthermore, the voltage is not directly related to the ability of a solar cell to convert sunlight into electrons.
110
Chapter 3
Figure 3.18 A setup for the measurement of action spectra and for IPCE measurements. A water cooled Xenon lamp (upper left corner) illuminates the grating focusing lenses. The grating separates the incident light according to the wavelengths, and the sample is rotated to measure the current and voltage as a function of the wavelength. The spectral resolution for the setup is 10 nm.
Figure 3.19 I sc measured as a function of wavelength for regioregular and regiorandom poly(hexyl-thiophene) devices as bulk heterojunctions with [60]PCBM.
3.2.5 IPCE measurements using a simple high-power spectrometer In Fig. 3.18, the setup for IPCE measurements is shown and the recorded action spectra are shown in Fig. 3.19. If the incident light from the lamp at the various wavelengths is given, and the incident photon to current efficiency (IPCE) can be
Characterization of Organic Solar Cells
111
determined. The result for a typical high-pressure xenon arc lamp is shown in Fig. 3.20. It is evident that there is a linear dependence from 420 nm to 800 nm. At UV wavelengths, and in the NIR region there are many lines that make accurate IPCE measurements difficult. Using the action spectra in Fig. 3.19 and the irradiance spectrum in Fig. 3.20, it becomes possible to work out the IPCE as IPCE(%) =
I × 100, e × photons
(3.6)
where I is the current density in amper per unit area, e is the elementary charge or
Figure 3.20 Irradiance (black) from a xenon arc lamp in the setup for IPCE measurements along with the number of photons (gray) as a function of the wavelength.
Figure 3.21 IPCE curves for regiorandom and regioregular poly(hexyl-thiophene). The active area of the device is 3 cm2 .
112
Chapter 3
1.602 × 10−19 C electron−1 , and photons is the number of photons from the lamp in unit area per second. The plot in Fig. 3.21 clearly shows that the IPCE is higher for the regioregular P3HT than the regiorandom P3HT. 3.2.6 Environmental effects Organic photovoltaics are inherently unstable under atmospheric conditions mainly due to the moisture and oxygen present in the atmosphere (see Chapter 4 for a detailed discussion). While some materials are more stable in the presence of both oxygen and moisture, under illumination devices they will degrade the interfaces between the electrodes and the active materials. Measurements on cells have to be performed when the cell is freshly made to minimize the effect of the testing environment if this is the ambient atmosphere. Alternatively, the measurements can be performed in a vacuum chamber or in a glove box, or the device can be encapsulated, which allows for stable transport outside the glove box (see Chapter 5).
3.3 Outdoor Measurements The above discussion deals exclusively with laboratory measurements performed under indoor conditions. The best reference source for OPV work is the sun, and there are many reasons for performing measurements under the real sun. As will be detailed in this section, there are many aspects of the art of outside testing. 3.3.1 Why outdoor photovoltaic characterization is necessary for organic solar cells Although the achieved power conversion efficiencies of solar cells, based on bulk heterojunction6 (BHJ) between donor-type conjugated polymers and acceptor-type fullerenes, of up to 5% (see Ref. [7]) clearly demonstrate the high potential of polymer photovoltaics (PVs), though considerable improvement of the cell stability under operational conditions needs to be achieved. As discussed in detail in Chapter 4, it has been demonstrated8–10 that the degradation behavior of polymer cells involves a number of photochemical mechanisms, including direct photooxidation of conjugated polymers, a photochemical reduction of the organic constituents by aluminum (from the back contact of the cells), and subsequent chemical reactions between the organoaluminum species, molecular oxygen, water, etc. However, the data on the stability of PV parameters of BHJ cells have been obtained by an accelerated indoor testing at elevated temperatures.9, 11–13 The acceleration factor is undoubtedly dependent on the PV materials and the cell architecture, and may vary with the degradation. Furthermore, there is always the risk that encapsulated cell failure might arise from particular combinations of environmental conditions that were missed by the accelerated tests. Such a situation is illustrated by the ethyl vinyl acetate (EVA) browning phenomenon in the encapsulated inorganic solar modules.14–17 The browning mechanisms were missed in the laboratory tests
Characterization of Organic Solar Cells
113
because the accelerated tests did not adequately simulate reality.15 The need of outdoor long-term characterization will increase with further improvement of stability and lifetime of organic solar cells. Thus, we believe that although accelerated tests can obviously predict some trends in the cell degradation, outdoor tests are the only reliable method for monitoring the aging of the cells and modules under natural operational conditions. At the practical level there is another reason for the importance of accurate outdoor testing of organic solar cells. Such testing can help one to quantify cell performance in a manner that may be compared from one laboratory to another. In the case of conventional inorganic solar cells, a set of standard test conditions (STC) have been defined. These correspond to an intensity of sunlight of 1000 W m−2 and a cell temperature of 25◦ C.18, 19 As for a spectral distribution of sunlight used for the STC, much effort was accordingly invested in modeling20 and measuring21 solar spectra under various atmospheric conditions, leading eventually to the international adoption of the so-called “AM 1.5G” standard global spectrum.18 In spite of the existence of such a standard, all kinds of efficiencies have been reported for organic solar cells, based on measurements performed under a wide variety of test conditions.22, 23 Only recently have a few studies24–26 been devoted to accurate indoor characterization of organic solar cells at STC using sophisticated procedures of solar simulator calibration and spectral mismatch correction. On the other hand, as it was demonstrated for inorganic solar cells and modules,27 the STC-adjusted outdoor tests can be remarkably precise with respect to quantitative findings. This section describes a procedure of accurate outdoor photovoltaic testing and adjusting of the results to STC, as well as the results of such outdoor testing24, 28, 29 at Sede Boker in the Negev Desert, Israel (lat. 30.8N, lon. 34.8E, alt. 475 m). The study was prompted by the fact that due to a fortunate coincidence of atmospheric effects at this particular site, the noontime spectrum on cloudless days is always extremely close17, 27 to the standard AM 1.5G spectrum when compared to any solar simulator. Figure 3.22 shows typical midsummer and midwinter sunlight spectra measured at Sede Boker, superimposed on the standard AM 1.5G values. The closeness of both ambient spectra to that of AM 1.5G indicates that no significant spectral corrections need to be performed on the Sede Boker outdoor measurements. 3.3.2 Experimental procedure The outdoor measurements of current-voltage (IV) characteristics of organic solar cells were performed in Sede Boker on cloudless days and at normal incidence to the incoming solar beam radiation, P in , which was measured with a calibrated thermopile pyranometer (Eppley PSP). For this purpose the studied cells and the pyranometer were mounted on a solar tracker (Fig. 3.23). As described elsewhere,30, 31 all of the pyranometers at the Sede Boker test center are calibrated once a year by comparison with a secondary standard. The sec-
114
Chapter 3
Figure 3.22 Typical clear-day, noontime, global spectra measured on a sun-tracking surface at Sede Boker; (a) summer, (b) winter. The standard AM 1.5G values18 are indicated by circles. We note that the midsummer noon declination angle at the latitude of Sede Boker (lat. 30.8N) corresponds to AM 1.0, whereas the midwinter figure is close to AM 1.7.17 (Reprinted from Ref. [17], with permission from Elsevier, copyright 1997.)
Characterization of Organic Solar Cells
115
Figure 3.23 Studied organic solar cells produced in RISØ National Laboratory and an Eppley PSP pyranometer mounted on a solar tracker. (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
ondary standard is an Eppley PSP instrument that is kept in the dark except when it is used for calibration purposes. A calibration check of the secondary standard is performed each June, when instruments owned by the Sede Boker test center, the Israeli National Physical Laboratory, and the Israeli Meteorological Service are brought together for comparison. This involves, in addition to each of the laboratory’s secondary standards, three primary standard cavity radiometers for which the calibration is directly traceable to the World Meteorological Standard. The comparisons take place outdoors on cloudless days during the noontime period. Calibration from the cavity instruments is transferred to the pyranometers under conditions of normal incidence to the incoming solar beam, using a fixed shading disk that subtends the same solid angle at one of the pyranometers as the angle of acceptance of the cavity instruments. Under such conditions, pyranometer calibration can be determined to a precision of ±0.3%. Calibration is subsequently transferred from the secondary standard to the pyranometer used for photovoltaic characterization; again, using the normal incidence method. During the IV-curve measurements, the cells had ambient temperatures varying from 25 to 50◦ C as measured by a T-type thermocouple. For study of temperature dependence of photovoltaic parameters, a thermoelectric table with the solar cells on its top surface was mounted on a solar tracker. In addition to all pairs of I and V sampled by the curve tracer, it automatically records the fitted values of the shortcircuit current, Isc , open-circuit voltage, Voc , maximum output electrical power, P out , current at the maximum power point, I mpp , voltage at the maximum power point, V mpp , efficiency, ηe , and fill factor, FF.
Chapter 3
116
The next stage of data processing should involve adjustment of the measured values of Isc , Voc , Ipp , Vpp , Pmax (maximum power), and FF to their corresponding values at the standard test conditions of 1000 W m−2 and 25◦ C. To do adjustments for organic solar cells, we suggest adopting an approach developed by D. Faiman and coauthors for inorganic cells.17, 27 In the case of Voc , Vpp , and FF, a linear adjustment can effect via the following transformations: Voc (25) = Voc (T ) − βoc (T − 25),
(3.7)
Vpp (25) = Vpp (T ) − βpp (T − 25),
(3.8)
FF(25) = FF(T ) − δ(T − 25),
(3.9)
where T is the cell temperature in degrees Celsius, and the coefficients βoc , βpp , and δ are temperature coefficients of the corresponding photovoltaic parameters. In the case of Isc , Ipp , and Pmax , the measured values should be first adjusted linearly to the standard irradiance level of 1000 W m−2 , and then given their respective linear temperature corrections to 25◦ C. The relevant transformations are as follows:
1000 Isc (1000, 25) = Isc (Pin , T ) − αsc (T − 25), P in 1000 Ipp (Pin , T ) − αpp (T − 25), Ipp (1000, 25) = P in 1000 Pmax (Pin , T ) − γ(T − 25), Pmax (1000, 25) = Pin
(3.10) (3.11) (3.12)
where αsc , αpp , and γ are are temperature coefficients of the corresponding photovoltaic parameters. Examples of measurements of temperature coefficients of organic cell parameters is given below. 3.3.3 Temperature dependence of the photovoltaic parameters of BHJ solar cells For the first time, accurate measurements of fullerene/polymer BHJ solar cells performed both indoors (under appropriately corrected simulated STC) and outdoors (under correspondingly corrected, naturally occurring STC), as well as the resulting temperature dependence for the cell parameters, was reported in Ref. [24]. In that study, the solar cells were produced at the Linz Institute for Organic Solar Cells and given their first approximate characterization using a filtered metalhalide solar simulator (Steuernagel SolarKonstant KHS 575, with calculated spectral mismatch factor M = 0.76).32 Samples were then sent to the Energy Research Centre of the Netherlands (ECN) (Petten, The Netherlands) for accurate indoor studies using a class A xenon-arc solar simulator (Spectrolab XT-10, with calculated mismatch factor M = 0.90), and to Sede Boker for accurate outdoor studies. The bulk donor-acceptor heterojunction solar cells were produced by spin casting (the production process is described elsewhere).32 MDMO-PPV was used as
Characterization of Organic Solar Cells
117
Figure 3.24 Device structure of an ITO/PEDOT/MDMO-PPV:PCBM/LiF/Al solar cell studied in Ref. [24], together with the chemical structure of compounds used for the cell active layer. (Reprinted with permission from Ref. [24]. Copyright 2001, American Institute of Physics.)
Table 3.7 Photovoltaic parameters of the BHJ polymer-fullerene cells24 measured under simulated AM 1.5 conditions in a glove box, before sealing. Measurements were performed with a solar simulator (Steuernagel SolarKonstant KHS 575) at an irradiance level of 800 W m−2 and a cell temperature of 55◦ C. Measured data were corrected to the plotted AM 1.5 values using a calculated mismatch factor of 0.76. (Reprinted with permission from Ref. [24]. Copyright 2001, American Institute of Physics.)
Cell 26
Cell 40
Voc (mV) 856 839 837 858 845 844
Isc (mA/cm2 ) 3.85 3.9 3.92 4.02 3.99 3.90
FF 0.598 0.601 0.598 0.567 0.585 0.606
Efficiency (%) 2.47 2.45 2.46 2.46 2.47 2.49
Area (mm2 ) 6.75 6.9 7.05 6.3 6.2 6.15
the electron donor while the electron acceptor was PCBM. The thickness of the spin-cast MDMO-PPV:PCBM active layer was about 100 nm. As electrodes, a transparent ITO film on one side and a LiF/Al bilayer contact on the other side was used. For improvement of the ITO contact, the ITO was coated with a thin layer of PEDOT. The device structure of ITO/PEDOT/MDMO-PPV:PCBM/LiF/Al layered solar cells is shown in Fig. 3.24. At Linz, preliminary indoor photovoltaic measurements of the as-prepared cells were carried out using a solar simulator, both before and after the cell encapsulation under inert conditions in a glove box (Table 3.7 and Fig. 3.25). At Sede Boker, the outdoor current-voltage measurements were performed during the noontime period. The solar irradiance, P in , was found to remain constant during the test runs to within approximately ±0.3% at levels that slightly
118
Chapter 3
Figure 3.25 Typical IV-curves for an as-produced BHJ MDMO-PPV:PCBM solar cell before and after sealing.24 Measurements were performed with a solar simulator (Steuernagel SolarKonstant KHS 575) at an irradiance level of 800 W m−2 and a cell temperature of 55◦ C. Measured data were corrected to the plotted AM 1.5 values using a calculated mismatch factor of 0.76. (Reprinted with permission from Ref. [24]. Copyright 2001, American Institute of Physics.)
exceeded 1000 W m−2 . Because the P in level was not, in general, precisely equal to 1000 W m−2 , it was necessary to adjust the measured values of Isc , I mpp , and P out to the STC irradiance value. To check the linear relation between Isc and P in , a solar simulator (also a Steuernagel SolarKonstant KHS 575) and wire masks were used to vary the intensity in the range of 80–550 W m−2 , without changing the spectral quality of the light. Under these circumstances, it was reasoned that if the cell current turned out to be linear with respect to changes in the intensity of the simulator lamp, then in all probability it would be safe to assume that it would remain linear under true AM 1.5 conditions—particularly for small departures from 1000 W m−2 . Figure 3.26 shows a plot of Isc versus simulator light intensity level at a variety of fixed temperatures in the range of 10–33◦ C. These curves are sufficiently linear to give us confidence that no substantial errors would be introduced into our outdoor results by adjusting measurements performed at, say, 1050 W m−2 to the standard 1000 W m−2 level. Recently,33 it was confirmed that the photocurrent behavior in BHJ cells is close to linear at T = 300 K, i.e., 1.01 , but deviates at lower temperatures (T ∼ 100 K) and irradiance levels Isc ∝ Isc (Plight ∼ 10 W m−2 ). The observed near-linear dependence is a key condition for efficient photoconversion. In simple terms, it indicates that a recombination of the photogenerated charge carriers is sufficiently slow compared to the charge transport for all of the photogenerated carriers to be collected at short circuit.34 At Petten, accurate indoor testing was performed. This employed a solar simulator, with a light spectrum that approximates the AM 1.5G spectrum, and a
Characterization of Organic Solar Cells
119
Figure 3.26 Short-circuit current (Isc ) versus irradiance level of solar simulator (Steuernagel SolarKonstant KHS 575) at various cell temperatures. (Reprinted with permission from Ref. [24]. Copyright 2001, American Institute of Physics.)
calibrated reference cell to set the intensity. These measurements can be divided into two steps, namely, (1) the determination of the simulator spectral mismatch factor M , and (2) measurements of the IV-curve of the solar cell and correction to STC. The match between a simulator spectrum, ES (λ), and the AM 1.5G reference spectrum, ER (λ), is never perfect, even for the best solar simulators. Furthermore, a spectral mismatch is introduced since the spectral responses of the device under test, ST (λ), and of the reference cell, SR (λ), are, in general, not identical. In order to correct for this, a spectral mismatch factor M can be computed via the following formula: ER (λ)SR (λ)∂λ ES (λ)ST (λ)∂λ M= ∗ , (3.13) ES (λ)SR (λ)∂λ ER (λ)ST (λ)∂λ where each integral is proportional to the short-circuit current that would be produced, at standard temperature, by the cell of stated spectral sensitivity Si (λ) under the specified spectrum Ej (λ). For inorganic solar cells, M usually lies in the range of 0.98–1.02, since stable calibrated solar cells can be constructed from more or less the same material as the test cell. However, for organic solar cells, suitable and stable reference cells cannot be fabricated yet. This implies that for the measurement of these cells, one has to use calibrated reference cells with a different spectral response from the device under test, resulting in mismatch factors significantly deviating from 1. Therefore, the first step is the measurement of the spectral sensitivity functions
120
Chapter 3
of the reference cell and the test cell, and measurement of the spectrum of the simulator. These measurements, together with the defined spectrum AM 1.5G, enable M to be calculated. It is of utmost importance to carry out the procedure as precisely as possible in order to minimize measurement errors. A monocrystalline silicon solar cell plus a KG5 filter calibrated at Institut für Solare Energi Systeme (ISE) was used as the reference cell. The spectral response of an encapsulated ITO/PEDOT/MDMO-PPV:PCBM/LiF/Al device was measured relative to the spectral response of the reference cell. Together with the spectral distribution for AM 1.5G and the simulator spectrum, a mismatch factor of 0.9 was calculated using Eq. (3.13). The value M = 0.9 was then used in the second step to correct the measured Isc values of the polymer-fullerene cell to Isc values appropriate to AM 1.5G conditions. IV-curves were measured at various cell temperatures in the wide range of the possible operating conditions (25–60◦ C). Generally, a qualitatively similar temperature behavior was observed by the indoor and outdoor IV measurements of all devices studied. Figures 3.27 and 3.28 summarize the temperature dependencies of the principal cell parameters (Voc , Jsc , ηe , and FF) derived from the outdoor and indoor IV measurements of typical devices. Outdoor and simulator measurements of Voc show a linear decrease with increasing temperature [Figs. 3.27(a) and 3.28]. Additional outdoor measurements of Voc (made while continuously varying the cell temperature, without recording the entire IV-curve) confirmed this behavior [Fig. 3.29(a)]. For all samples, the observed linear decrease had a temperature coefficient in the range βoc = dVoc /dT = −(1.40−1.65) mV/K. We are not discussing here the physical mechanisms responsible for the observed temperature dependence of Voc (for detailed analyses see, e.g, Ref. [35].) Similar to inorganic solar cells, a variety of organic photovoltaic devices demonstrate a linear decrease of Voc with increasing temperature (a deviation of the temperature dependence of Voc for organic solar cells from linearity was observed only in the very low temperature range,33 100–150 K). It allows using Eqs. (3.7) and (3.8) for adjusting the outdoor measured Voc and Vpp to the STC. Figures 3.27(b), 3.27(d), and 3.28 show a relatively large monotonic increase with temperature for Isc and FF, followed by a saturation region. A slight increase in Isc with temperature is also a common feature for inorganic solar cells. However, in the case of the studied MDMO-PPV:PCBM BHJ cells, the rate of increase is so dramatic that the increase of the short-circuit current and fill factor product with temperature overtakes the decrease of open-circuit voltage with temperature. As a result, there is an absolute increase of the power conversion efficiency ηe with temperature T reaching a maximum value at Tmax , which for different samples lies in the range of 47–60◦ C [Figs. 3.27(c) and 3.28]. In order to investigate this behavior more thoroughly, we measured Isc with continuous variation of the cell temperature without recording the entire IV-curve. The result is shown in Fig. 3.29(b) (for another sample), where a clear indication
Characterization of Organic Solar Cells
121
Figure 3.27 Temperature dependencies of the principal photovoltaic parameters for a typical BHJ MDMO-PPV:PCBM solar cell derived from outdoor measurements of its IV-curves.24 Plotted values of efficiency and I sc have been adjusted to the STC irradiance level of 1000 W m−2 . (Reprinted with permission from Ref. [24]. Copyright 2001, American Institute of Physics.)
122
Chapter 3
Figure 3.28 Temperature dependence of normalized photovoltaic parameters for a typical MDMO-PPV:PCBM BHJ solar cell derived from indoor measurements of its IV-curves.24 Ordinate axis displays all parameters normalized to their measured values at 25◦ C, namely, I sc = 3.1 mA cm−2 , V oc = 840 mV, FF = 0.55, and ηε = 1.45%. Active cell area = 7.5 mm2 . Measurements were performed with a class A solar simulator (Spectrolab X-10). Measured data were corrected to their corresponding AM 1.5 values using a mismatch factor of 0.9. (Reprinted with permission from Ref. [24]. Copyright 2001, American Institute of Physics.)
Figure 3.29 The MDMO-PPV:PCBM cell recording by continuously varying the cell temperature.24 I sc values were adjusted to the STC irradiance level of 1000 W m−2 . (Reprinted with permission from Ref. [24]. Copyright 2001, American Institute of Physics.)
of saturation sets in at around 60◦ C. This, together with the continued falloff in Voc , would result in a subsequent decrease in the efficiency with further increase in temperature. A noteworthy point, which provides further confirmation of this result, is that when the cell temperature was cycled back and forth there was no hysteresis in the Isc temperature dependence. It should be noted that positive temperature coefficients for Isc have since been observed for other organic solar cells. For example, PCBM:polyfluorene solar cells showed even higher values of the Isc temperature coefficient. For such devices, Isc was revealed to increase linearly from 0.8 to 2.2 mA cm−2 (i.e., by 175%) as temperature increased through the range 5–100◦ C.36 Possible reasons for the observed, unusually large, positive temperature coefficient for Isc of organic so-
Characterization of Organic Solar Cells
123
lar cells are still under discussion. It has been attributed to thermally activated mobility of the charge carriers in the cell’s active layer,24 difference in the temperature dependences of mobility of electrons, and holes within the fullerene and polymer transport paths,33 as well as space charge effects.34 On the other hand, Isc in PCBM:P3HT BHJ solar cells was recently shown to be almost temperature independent in the temperature range 10–60◦ C.37 These findings point to the fact that the charge collection in these particular BHJ cells is not controlled by electronic transport in the photoactive layer. The observed temperature dependence of FF [Figs. 3.27(d) and 3.28] was quite similar to that of Isc . The former, however, can be qualitatively understood in terms of the temperature-dependent series resistance of the solar cell, RS . BHJ cells have a relatively high resistivity of the organic active layer, but one that decreases with increasing temperature. The complicated character of the temperature dependence of Isc and FF of the MDMO-PPV:PCBM BHJ cells described above points to the fact that, contrary to inorganic solar cells, simple adjusting of FF, Isc , and Ipp by Eqs. (3.9) and (3.10) to the STC is not always relevant for organic devices. However, as shown for the case of the MDMO-PPV:PCBM BHJ cells, temperature dependencies of FF and Isc can be approximated by two linear adjustment and the corresponding linear adjustment [Eqs. (3.7) and (3.8)] in each region, separately. For example, for Isc of the cell shown in Fig. 3.29, these two linear regions are at 15–65◦ C and 65–70◦ C. Due to the fast degradation of the studied MDMO-PPV:PCBM BHJ cells,24 it was difficult to make quantitative comparisons among the results of photovoltaic characterization of the cells at different stages of their degradation. In particular, it is meaningless to compare the absolute results of our outdoor and indoor IV measurements. However, in addition to qualitative coincidence of the temperature dependencies of the cell parameters observed by the indoor and outdoor IV measurements, we may report that after 6 hr of irradiation during 11 successive days of the outdoor experiment, absolute values of Voc and its temperature dependence [dVoc /dT, Voc (0 K)] remained almost constant. The long-term study of the stability of various organic solar cells under outdoor operational conditions is reported in the next section. 3.3.4 Example of long-term outdoor testing of stability of organic solar cells29 Three encapsulated polymer-fullerene solar cells with photoactive area of 10 cm2 were produced at Risø National Laboratory. Details of the cell production and encapsulation are described elsewhere.38 Al and ITO/ PEDOT:PSS layers were used as back and front electrodes, respectively. The cells differed by the configuration and content of photoactive layer as follows: 1. Bulk heterojunction of MEH-PPV:PCBM 2. Bulk heterojunction of P3HT:PCBM
124
Chapter 3
Figure 3.30 Examples of current-voltage curves of the MEH-PPV/PCBM cell at different stages of degradation.29 Current values were normalized to the standard irradiance value of 1000 W m−2 . (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
3. Bilayer junction of poly(3-carboxythiophene-co-thiophene) (P3CT)/C60 The latter device does not include a PEDOT:PSS sublayer. The measurements were carried out for a period of 32 subsequent days during daylight hours, from ∼9 am to ∼5 pm (contrary to the noontime experiments described in the previous section) March 15, 2006. At night the cells were kept in a dark glove box with a nitrogen atmosphere. Because the outdoor solar irradiance, P in , varies throughout the day, in order to record stability of the PV parameters, the measured values of Isc , I mpp , and P out , were adjusted to the standard irradiance value of 1000 W m−2 , assuming linear dependence of the photocurrent on P in . The MEH-PPV:PCBM cell exhibited the fastest degradation (Figs. 3.30 and 3.31). One can see that after a few hours of exposure to sunlight the efficiency of the cell decreased to 50% of its initial level. On the other hand, even initial levels of the Isc , Voc , FF, and ηe were very low. We ascribe this in part to the large area of the cell and in part to degradation of the device during transport from Denmark to Israel, since the PPV-based devices are subject to instability in the dark. It should be noted here that data on the performance of organic solar cells are commonly based on very small devices with active areas of only a few square millimeters. Large-scale organic photovoltaics is a research field that is still relatively unexplored.39 The degradation of photovoltaic performance of the P3HT:PCBM cell was much slower (Figs. 3.32 and 3.33), while the P3CT/C60 cell was found to be the
Characterization of Organic Solar Cells
125
Figure 3.31 I sc and ηe as a function of additive time of the sunlight exposure for the MEH-PPV/PCBM cell. Vertical lines separate various days of measurements (from March 15, 2006 to March 26, 2006). I sc was normalized to the standard irradiance value of 1000 W m−2 . (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
Figure 3.32 Examples of current-voltage curves of the P3HT/PCBM cell at different stages of degradation. Current values were normalized to the standard irradiance value of 1000 W m−2 . (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
126
Chapter 3
Figure 3.33 Principle PV parameters as a function of additive time of the sunlight exposure for the P3HT/PCBM cell. Solid circles indicate the very first measurements every day (from March 15, 2006 to May 16, 2006). I sc was normalized to the standard irradiance value of 1000 W m−2 . (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
Figure 3.34 Examples of current-voltage curves of the P3CT/C60 cell at different stages of degradation. Current values were normalized to the standard irradiance value of 1000 W m−2 . (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
most stable (Figs. 3.34 and 3.35). A totally different time dependence of the FF is evident for these two devices. The fill factor of the P3HT:PCBM cell quickly degraded from 0.26 to 0.21 during the first few hours of the exposure and then started to exhibit a slow linear degradation [Fig. 3.33(b)]. The FF of the P3CT-C60 cell was almost constant along the entire test process [Fig. 3.33(b)].
Characterization of Organic Solar Cells
127
Figure 3.35 Principle PV parameters as a function of additive time of sunlight exposure for the P3CT/C60 cell. Solid circles indicate the very first measurements every day (from March 15, 2006 to May 17, 2006). I sc was normalized to the standard irradiance value of 1000 W m−2 . (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
We also observed an unusual behavior of Isc and Voc for both the P3HT:PCBM and P3CT/C60 cells [Figs. 3.33(a) and 3.33(b)]. A restoration, or recovery, effect was observed for Isc and Voc during each night (when the cells were kept in the dark). The first Isc and Voc measurements every morning (indicated by solid circles) yielded the highest values compared to those during the rest of the day. While Isc only partly recovered at nights recovered at night and exhibit significant degradation during a month, Voc values recovered completely every night and show almost no reduction on a long-term time scale. As already mentioned, at night the cells were usually kept in a dark glove box with a nitrogen atmosphere. In order to check the role of the atmosphere on the recovery effect, we kept the cells in the dark but in air, for several nights, and observed the same recovery behavior of Isc and Voc . Furthermore, we observed almost complete restoration of both Isc and Voc parameters after shadowing the P3CT/C60 cell for 30 min, during daylight (afternoon) (Fig. 3.36). Simultaneous recording of the cell temperature and the input light intensity variations during this experiment (not shown) indicated that they are not responsible for the recovery effect. Shadowing the cell for 10 min resulted only in the partial recovery of Isc and Voc (Fig. 3.36). A time threshold appears to exist for the recovery effect. For P3CT/C60 devices, this threshold is between 10 and 30 min. A similar effect of partial recovery of Isc was observed after keeping quasisolid-state dye-sensitized solar cells in the dark for 9 hr.40 Probably some restoration of Isc was observed in the accelerated lifetime measurements of nonencapsulated BHJ solar cells after changes of solar simulator light bulbs,12 in spite of the fact that the authors attributed the effect only to the variation in the emission of the lamp. One should further check if the recovery effect is reproducibly observed for other plastic cells (encapsulated and nonencapsulated). If so, it would
128
Chapter 3
Figure 3.36 I sc and V oc behavior of the P3CT/C60 cell during the last days of measurements described in Ref. [29]. Vertical lines separate various days of measurements. Solid circles indicate the very first measurements each day, while arrows and solid squares indicate the first measurements after the cell was shadowed for 10 or 30 min. I sc was normalized to the standard irradiance value of 1000 W m−2 . (Reprinted from Ref. [29], with permission from EDP Sciences, copyright 2007.)
point to the fact that in addition to the nonreversible photochemical degradation of the plastic cell, some reversible mechanisms could be in play, which may include photoinduced generation of charge traps that then slowly disappear in the dark. 3.3.5 Some new experimental possibilities and suggestions for future studies In general, variation in the outdoor photovoltaic performance throughout the day reflects simultaneous effects of changes in intensity and spectrum of sunlight and ambient temperature. For organic solar cells, the situation is even more complicated due to possible degradation of the cells during the day light hours. Production of more stable cells will allow investigating variations of the photovoltaic parameters of organic solar cells throughout the daylight hours. In order to properly interpret experimental data on such variation, further experimental programs should include measurements of the temperature dependence of photovoltaic parameters and the spectral response of the cells at various stages of degradation, together with a recording of the variation in the daytime sunlight spectrum. In general, temperature coefficients can be affected by cell degradation. Recently, an increase in the positive temperature coefficient for Isc with degradation of dye-sensitized nanocrystalline photoelectrochemical cells was observed.36 To the best of our knowledge, there
Characterization of Organic Solar Cells
129
Figure 3.37 Clear-day variation of P max measured at Sede Boker and adjusted to the standard 1000 W m−2 irradiance and 25◦ C cell temperature, for inorganic solar cells based on single crystalline (c-Si), polycrystalline (polyX-Si), and amorphous (a-Si) silicon. (Reprinted from Ref. [27], with permission from Elsevier, copyright 1999.)
is no publication on the study of organic solar cells operating under real outdoor conditions. Figure 3.37 shows the results of the variation of Pmax of inorganic solar cells on a clear day measured at Sede Boker and adjusted to the standard 1000 W m−2 irradiance and 25◦ C cell temperature.41 Since the cells based on single crystalline, polycrystalline, and amorphous silicon demonstrated no significant degradation during a single day, the results clearly reflect spectral effects. The fact that crystalline and amorphous silicon solar cells show opposite trends, the variation of Pmax was attributed by the authors to the difference in the bandgap of these semiconductors and the corresponding difference in the spectral response of the cells. In addition to the above-mentioned experiments, the results of parallel studies of organic solar cell degradation under real sun outdoor operational conditions should be compared with those of accelerated indoor studies at various elevated temperatures. The dependence of Isc on P in was measured indoors with a solar simulator. However, this experiment could be done with natural sunlight and a wide range of controlled levels of P in using a novel characterization approach and outdoor/indoor test facility based on the fiber optic/minidish concentrator (Fig. 3.38).42–45
130
Chapter 3
Figure 3.38 Schematic of the solar fiber optic/concentrator test facility.42–45 (a) Solar radiation is concentrated outdoors into the tip of a highly transmitting optical fiber, which guides the concentrated sunlight indoors onto the solar cell being tested. Radiation input is moderated via a pizza-slice iris mounted on the dish window. (b) Flux uniformity is achieved with a square cross-sectional kaleidoscope while (c) direct fiber/cell contact is used for the localized irradiation probe (LIP). (Reprinted with permission from Ref. [45]. Copyright 2006, American Institute of Physics.)
In such a facility, solar-beam radiation is collected and concentrated outdoors and is focused into a high-transmissivity optical fiber and then delivered indoors onto the solar cell being tested. Radiation on the cell is moderated with a pizzaslice iris that is mounted on the dish window and preserves the angular distribution of delivered sunlight. Near-perfect flux uniformity is achieved with a square crosssectional kaleidoscope, matching the size of the cell and placed between the distal fiber tip and the cell [Fig. 3.38(b)]. The current system can deliver sunlight with intensity from 0.1 sun up to 100 suns on a 100 mm2 cell or 4000 suns on a 4 mm2 cell (for 1 sun, Pin = 1000 W m−2 ). Removing the kaleidoscope and varying the fiber height above the cell, we can realize a wide range of flux distributions, including the extreme localized irradiation limit when the fiber touches the cell such that radiation projected beyond the fiber tip is negligible [Fig. 3.38(c)]. This localized irradiation probe allows ultrahigh local flux levels of up to 104 suns and mapping the cell performance with the fibers of 2-, 1-, 0.6-, and 0.2-mm diameters. The study of the photovoltaic performance of organic cells at elevated concentrations of sunlight using this approach will provide important information on the possibility of using this kind of cell under irradiation levels higher than that of one sun. Some hope for a positive answer on this question appeared after the demonstration by the Princeton group that a specific series resistant C60 /copper phthalocyanine (CuPc) heterojunction solar cells can be reduced down to 0.1 Ω cm−2 .41, 46
Characterization of Organic Solar Cells
131
A high FF of 0.61 was achieved, which was only slightly reduced at the intense illumination. As a result, ηe was found to increase with the incident light intensity, reaching a maximum of ∼4.2% under 4–12 suns simulated AM 1.5G illumination. The system shown in Fig. 3.38 will also enable investigating the long-term stability of organic solar cells as a function of sunlight intensity in a wide range of P in .
3.4 Methods for Preparation and Characterization of Thin Films Morphology is an important contributor to the factors governing the photovoltaic properties of organic solar cells. Optimization of efficiency has primarily been the center of activity when performing morphological studies. To improve efficiency, it is imperative to gain a better understanding of how to predict and thus control the morphology. This is achieved from detailed understanding of the underlying photophysics that are in play during operation of the photovoltaic device. When an organic solar cell is illuminated the active material absorbs light. The relatively high absorption coefficient of organic materials (compared to silicon) enables the use of very thin films of the active material, that is, it only requires ∼100 nm to absorb a sufficient amount of light, which is straightforward from an architectural point of view. The incident photons create so-called excitons that have to dissociate into free charges in order to be transported to the electrodes. Exciton dissociation is a charge transfer process and occurs efficiently at the appropriate interfaces, depending on the architecture, and controlled by the relative electron affinities and ionization potentials of the materials involved. The photoexcited excitons have to reach an appropriate interface before they decay (i.e., recombine), otherwise the photons are wasted. Exciton diffusion is controlled by dipole coupling between molecules. The diffusion length should be at least equal to the layer thickness, which constitutes a challenge. The exciton diffusion range in a typical conjugated polymer is only ∼10 nm, which is not straightforward to compensate for. Single-layer cells are the simplest form of organic solar cells, consisting of only one semiconductor material sandwiched between two electrodes (the Schottky type). One of the drawbacks of this architecture is a limited coverage of the absorption range. Furthermore, due to the short exciton diffusion range, the photoactive region is typically very thin. Finally, due to the fact that the electrons and holes may travel through the same material, there is a significant loss of efficiency due to recombination. In double-layered cells, an electron acceptor layer is placed between the active material (i.e., electron donor material) and the cathode. Exciton dissociation takes place at the donor-acceptor interface. There are advantages associated with this type of architecture, namely, the electrons and holes are spatially separated and travel in different layers so recombination is significantly reduced; furthermore, the active region is larger than for a single-layer device since the exciton diffusion lengths of both materials now define the active region. Consequently, a thicker active layer can be used since excitons are created from both sides of the donor-
132
Chapter 3
acceptor interface. Finally, if the donor and acceptor materials complement each other, a broader spectral absorption range can be obtained, which will contribute to the overall device performance. However, in spite of a significant improvement compared to single-layer cells, the limited interface thickness is still a matter of concern for this type of architecture. The introduction of blended cells constitutes a major breakthrough in the architectural development of organic solar cells. Mixing an electron donor with an electron acceptor material produces a large interfacial area (i.e., increases the effective surface of the junction) on a scale similar to the exciton diffusion range. After exciton dissociation, percolated pathways are required so that the charges can reach the respective electrodes. This is achieved by careful development of a suitable nanoscale morphology resulting in an interpenetrating network. This type of architecture greatly improves the quantum yield and thus the efficiency. However, due to low-charge carrier mobilities and imperfect network structures, charge transport is still the limiting factor, especially electron transport. As is evident from the previous discussion, a control of the morphology must be effective on the scale of the exciton diffusion length to avoid loss of carriers. Another important parameter that can be morphologically controlled is the charge mobility. Charge mobility depends on the physicochemical properties of the materials, but also on the morphology. There are two types of charge transport. The first type is specific to conjugated polymers—intramolecular charge transport along the polymer chain. The second type is less specific and less efficient—intermolecular charge transport between polymers (or molecules). Increasing the charge mobility will improve the photovoltaic performance. Mobility is affected by (i.e., controllable by) the molecular packing. Higher molecular ordering is consistent with a higher mobility, that is, it is desirable to increase the mesoscopic order and crystallinity. For a blended cell, it is thus desirable to have a nanoscale interpenetrating network with crystalline order of both components for the active layer in the organic solar cell. A detailed understanding of the underlying photophysics that are in play during operation of an organic solar cell enables us to predict the required morphological control necessary for improving the photovoltaic performance. Morphological control is specifically needed to create well-defined thin films of organic material and well-defined thin electrodes, and for blended cells, control of the miscibility between semiconductor materials and the ability to control phase separation is required. The morphology is controlled by various methods depending on the application and architecture. Typical methods will be presented in the following sections such as spin coating, thermal evaporation, and annealing. An important aspect of controlling morphology is the ability to monitor the morphology. This is carried out by employing characterization techniques such as scanning electron microscopy (SEM) and transmission electron microscopy (TEM), atomic force microscopy (AFM) and conductive atomic force microscopy (CAFM), optical mi-
Characterization of Organic Solar Cells
133
croscopy (phase contrast, interference, and fluorescence), and scattering techniques [x-ray and Rutherford backscattering (RBS)]. The applicability of these techniques will be demonstrated in the following based on examples from the literature. 3.4.1 Controlling morphological properties Controlling the morphology in organic solar cells is manifested in the practical ability to create well-defined organic and metal thin films. Different deposition techniques typically produce films with different properties. This is due to the fact that the optical and electrical properties are strongly affected by structure, morphology, and the nature of possible impurities. The resulting film properties of one particular deposition technique can differ due to the involvement of various deposition parameters. It is thus important to gain detailed knowledge on the relationship between film properties and the method of deposition. The number of existing deposition technologies is overwhelming. Consequently, only the most common deposition techniques relevant for organic solar cells are described in this section, including a description of the annealing phenomenon and its relevance to organic solar cells. Spin coating and thermal evaporation are both deposition methods that, from a production point of view, are well established, fairly cheap, fast, and produce sufficiently high-quality thin films, that is, production-friendly methods. 3.4.1.1 Spin coating
Spin coating,47 also known as spin casting, is the most successful and widely employed method for the highly reproducible fabrication of organic thin films with high structural uniformity. For a review of spin-coated polymer films, see Ref. [47]. The principle of spin coating is fairly simple (Fig. 3.39). It is the detailed understanding of the spin-coating processes that allows us, via judicious selection of experimental parameters, to control the morphology of the resulting thin film. When a polymer solution is deposited on a horizontal rotating disk, a uniform liquid film is produced. The disk should either be static or rotating at a low angular velocity followed by a rapid acceleration to a high spin speed. The adhesive forces at the solution-substrate interface and the centrifugal forces acting on the ro-
Figure 3.39 Schematic of the major spin-coating processes.
134
Chapter 3
Figure 3.40 Schematic of the relation between film thickness, angular velocity, and concentration (C n ) or viscosity (ηn ) in a spin-coating process.
tating solution result in strong sheering of the solution, which causes a radial flow in which most of the polymer solution is rapidly ejected from the disk, as shown in Fig. 3.39. The solvent then starts to evaporate, causing the polymer concentration to increase (and thus the viscosity) at the liquid-vapor interface. After evaporation of most of the remaining solvent, a uniform practically solid polymer film is consequently formed. The two most important morphological parameters to control are the film thickness and the uniformity of the film. The film thickness is affected by (i.e., controlled by) the following spin-coating parameters: angular velocity, solution viscosity, and solution concentration. Spin-coating parameters such as the amount of solution initially deposited on the disk, the rate at which it is deposited, the history of rotational acceleration prior to the final acceleration, and the total spin time have limited or no effects on the film thickness. Figure 3.40 shows the relationship between the parameters controlling the film thickness. Increasing the angular velocity (i.e., spin speed) for a given concentration and viscosity decreases the film thickness. Decreasing the concentration or viscosity for a given angular velocity decreases the film thickness. For polymer solutions, the concentration is related to the viscosity such that an increase in the viscosity corresponds to an increase in concentration (typically not linearly). The above examples are valid for a given polymer and a given solvent, that is, different polymer-solvent combinations might produce different film thicknesses in absolute terms. For example, highly volatile solvents produce thicker films at a given polymer concentration and initial viscosity compared to low-volatility solvents. The dynamic range of controllability with respect to the film thickness is typically in the range of a submonolayer (where the degree of surface coverage is a more appropriate term) up to several hundreds of micrometers. The uniformity of the resulting thin film is affected and thus controlled mainly by the choice of solvent. Solvent evaporation changes the physical and thus the rheological properties of the solution during the spin-coating process, that is, highly volatile solvents could decrease the temperature to an extent that could induce nonNewtonian behavior such as elasticity and shear thinning. Spangler et al.48 studied
Characterization of Organic Solar Cells
135
the effect of solvent on the surface topography of spin-coated polymers. Based on their findings, it can be concluded that surface topography of spin-coated polymer films is affected and thus controlled by mainly two parameters, namely, (1) solvent volatility, and (2) polymer-solvent interactions. Spangler et al.48 correlated polymer-solvent compatibility to the degree of coiling of the polymer chain in solution. Good compatibility corresponds to energetically favorable interactions between polymer and solvent, that is, the interactions are more favorable than the polymer-polymer and the solvent-solvent interactions, causing the polymer chain to expand. For solutions of comparable volatilities, only solvents with good polymer compatibility produce uniform polymer films. If only solvents that have good polymer compatibilities are considered, then solvents with lower volatilities produce more uniform polymer films. If the solvent is of sufficiently high volatility, a performance, known as “skinning” can become significant. A so-called solid “skin” can form at the free surface. Consequently, defects can form in the film if the convective flow is not completed when the skin starts to form. This phenomenon can be prevented by partially saturating the atmosphere above the solution-coated rotating disk with solvent vapor, or by using a binary solvent system, that is, containing both a high- and a lowvolatility solvent. The principle of spin coating may sound simple, however, there are factors not discussed in this material that can affect the outcome of the spin-coating process and thus limit the morphological controllability. Besides the angular velocity and the physical properties of the solution, the outcome is very sensitive to parameters such as temperature, airflow velocity (see Fig. 3.39), relative humidity, and thermal surroundings for the evaporating solvent. This is the reason why it is very difficult to reproduce experimental conditions for spin coating. Spin coating is excellent when it comes to creating a well-defined film with a homogeneous lateral and vertical polymer distribution. Alternative techniques for depositing coatings from liquid media include dip coating, flow coating, spray coating, roller coating, pressure-curtain coating, brushing, offset printing, and doctor blading. These methods have varying applicability when it comes to organic solar cells. One method has proven to be a useful alternative to spin coating. The dip-coating (also known as drop-casting) method is a crude version of spin coating, where the substrate is immersed in a polymer solution and then withdrawn. The film thickness can be controlled by the withdrawal speed, and by the concentration or viscosity. It has been shown that dip coating produces semiconducting polymer films that exhibit charge carrier mobilities that are significantly higher than the corresponding spin-coated version. The explanation lies in the solvent evaporation speed—for spin coating, the solvent evaporation is fast (seconds) compared to dip coating (minutes), which leaves little time for the polymer chains to align. For dip coating, the polymer molecules can self-organize over a long time to form the thermodynamically favored structure.
136
Chapter 3
The relationship between molecular ordering (and thus mobility), choice of solvent, and choice of coating method (i.e., spin coating or dip coating) has been investigated by several researchers.49–57 A lot of this work has focused on the wellstudied P-type semiconducting polymer material P3HT that has self-organizing properties to form microcrystalline structures. Sirringhaus et al.49–51 investigated the correlation between microstructures and charge-carrier mobilities in P3HT. The P3HT films prepared by dip coating exhibited charge carrier mobilities that were 100-fold larger than the spin coated P3HT films. It was found that the two preparation methods resulted in two lamella microstructures, that is, two different orientations of ordered P3HT domains relative to the substrate. In the spin-coated films, conjugated lamellae and the direction of π-π stacking are oriented normal to the substrate, and in the dip-coated films, the orientation is parallel to the substrate. The dip-coating method is clearly advantageous compared to spin coating when it comes to alignment of the molecules (i.e., mobilities). However, dip coating is inferior when it comes to creating a well-defined film with a homogeneous lateral and vertical polymer distribution. The success of the dip-coating method lies in the relatively slow speed of solvent evaporation; thus, if the speed of solvent evaporation is slowed down in the spin-coating process, it would constitute a significant improvement in mobility when spin coating semiconductor materials for the purpose of fabricating organic solar cells. Sirringhaus et al.49–51 investigated the relationship between solvent volatility and mobility for the well-studied P3HT when employing spin coating. It was found that the lowest mobilities were observed for chloroform, which had the lowest boiling point (60.5–61.5◦ C), and the highest mobility was observed for 1,2,4trichlorobenzene, which had the highest boiling point (218–219◦ C). The mobility for 1,2,4-trichlorobenzene was 0.12 cm2 V−1 s−1 , which is tenfold higher than the corresponding value for chloroform. Based on the previous discussion, it is tempting to conclude that one should always use the higher boiling solvents when spin coating the active layer of an organic solar cell; however, there are exceptions where it is advantageous to use the lower boiling solvent. Nanoscale phase-separation mechanisms in blended cells can be controlled by varying the spin-coating parameters in which solvent volatility plays an important role. Arias et al.58, 59 studied the aspects of morphology (controlled by spin-coating parameters) on the performance of organic solar cells. The authors explored vertically segregated polymer-blend photovoltaic thin-film structures through surface-mediated solution processing. A polymer blend was created composed of equal amounts of the hole acceptor poly(9,9-dioctylfluorene-co-bisN,N-(4-butylphenyl)-bis-N,N-phenyl-1,4-phenylenediamine) (PFB) and the electron acceptor poly(9,9-dioctylfluorene-co-benzothiadiazole) (F8BT), which were dissolved in xylene or chloroform. The homopolymer solutions and polymer blend solution were spin coated on ITO substrates followed by Al electrode deposition. The photovoltaic response was systematically investigated as a function of solvent
Characterization of Organic Solar Cells
137
and substrate temperature. Furthermore, by comparison, the solutions were also dip coated. Arias et al.58, 59 found that the blend solutions resulted in higher efficiencies than the homopolymers. The cells prepared from chloroform solutions demonstrated higher efficiencies compared to the more slowly evaporating xylene solution. This was taken to suggest that the rapid evaporation impedes the rearrangement of the polymer chains and quenches the phase separation on a scale similar to the exciton diffusion length. The phase separation was observed to be much slower when employing dip coating, which resulted in phase separation on a much larger scale, that is, a lower efficiency compared to spin coating. Heating the substrate (to 40◦ C) was found to increase the evaporation rate, resulting in a significantly increased efficiency. 3.4.1.2 Thermal evaporation
Thermal evaporation (TE), also known as vapor deposition (VD), vacuum evaporation (VE), or vacuum thermal evaporation (VTE), is one of the oldest techniques for depositing thin films. The technique has mainly been used for depositing metals or metal alloys. However, in correlation with the growing semiconductor industry, the demand for depositing organic thin films is growing. For a more thorough description of the technique, see Ref. [60]. The principle of TE is (like spin coating) fairly simple (Fig. 3.41). A judicious selection of experimental parameters that are based on a detailed understanding of the physical processes allows us to control the morphology of the resulting thin film. The process of TE involves several sequential steps, namely, (1) a material that can be sublimed is placed in a vessel (boat) in vacuo and the vessel is heated; (2) when sufficient heating is obtained, the material starts to boil or sublime; (3) the evaporated material then travels from the heated vessel in all accessible directions in straight lines (i.e., ballistic propagation) until an obstacle is reached, for example, the substrate or any other surface in the vacuum chamber; (4) if a substrate is placed appropriately behind a mask, then the material condenses on the substrate surface on an area defined by the mask. The adsorbed material then typically
Figure 3.41 Schematic of the major thermal evaporation processes.
138
Chapter 3
diffuses some distance before being incorporated into the surface. The material’s molecules/clusters then react with other material’s molecules/clusters, or with the substrate consequently forming aggregates, which is known as a nucleation process. Further growth results in the formation of structure or morphology, and topography and crystallography now become part of the film’s properties. The evaporation of the material is carried out by heating the boat containing the material, that is, the source of the material. This is typically done using a heating coil where the applied current controls the temperature. However, there are alternative methods to evaporate the material. The material can be exposed to a beam of energetic electrons, photons, or positive ions. These alternative methods are known as electron-beam-physical-vapor deposition (EBPVD), pulsed-laser deposition (PLD), and sputter deposition (SD). The evaporation methods mentioned thus far are categorized as physical vapor deposition (PVD), contrary to chemical vapor deposition (CVD), where the substrate is typically pretreated to become chemically reactive toward the incoming source material, which can be a gas, an evaporating liquid, or a chemically gasified solid. For simplicity and clarity, only TE is discussed in this section, that is, PVD using a heating coil to evaporate the material. The uniformity and film thickness of the resulting film is influenced by several process parameters of which some are interrelated. The uniformity of the resulting film is influenced by the uniformity of the arrival rate (the equivalent to the evaporation rate) of the material at the substrate. In high vacuum, the material travels in straight lines from the source to the substrate (i.e., ballistic propagation). The arrival rate uniformity is thus affected by the relative geometry between the source and the substrate. A longer distance between the boat source and the substrate corresponds to a better uniformity of the resulting film. Furthermore, the arrival rate uniformity is also affected by the ability to maintain a constant heating of the boat source, which is not straightforward. The process of the actual depositing of the material on the substrate is influenced by three factors, namely, (1) substrate surface conditions such as roughness, substrate reactivity toward the incoming material, degree of contamination, and possibly by crystallographic parameters; (2) reactivity of the incoming material; and (3) substrate temperature, which affects the mobility of the material on the surface, and thus, affects the film forming properties. There are actually not that many controllable parameters in a TE process. Evaporation time is controllable from the ability to switch on and off the current to the heating coils that produce the heat required for the evaporation process to occur. The evaporation time controls the film thickness at a given evaporation rate, that is, longer evaporation times produces thicker films. The source temperature is controllable from the amount of current applied to the heating coils. The source temperature controls the evaporation rate (i.e., arrival rate at the substrate), which controls the film thickness for a given evaporation time (i.e., higher evaporation rates produce thicker films). The precision of the layer thickness can typically be
Characterization of Organic Solar Cells
139
controlled within ±0.5 nm. The ability to control the source temperature accurately governs the uniformity of the evaporation rate, which governs the uniformity of the film. The source temperature controls the evaporation rate, which affects the nucleation behavior, that is, a higher evaporation rate will produce less coarse nucleation and thus produce a smoother film. However, in some cases it is advantageous to use a slow evaporation rate in order to produce a polycrystalline film with large grain sizes and thus a corresponding low concentration of grain boundaries. Grain boundaries can act as electrical carrier traps in semiconductors and as channels for diffusion of undesirable species, for example, oxygen and water diffusion through the cathode or barrier layer of an organic solar cell. The substrate temperature is (for some instruments) controllable and affects the deposition behavior. Lowering the substrate temperature will inhibit molecular diffusion and thus inhibit the nucleation and coalescence processes, that is, the growth is quenched, resulting in more smooth films. This phenomenon is known as ballistic deposition. Rotating the substrate (if possible) produces a more homogeneous film. However, this requires the ability to control the angular movement accurately. In spite of TE being a well-established and well-developed technique, there are still some problematic aspects. If the pumping is sufficiently low, the pressure could become so high due to foreign species (e.g., air and water) that the transport of the material from the source to the substrate could be affected due to collisions with other species, resulting in an altered path to the substrate, which will consequently affect the uniformity of the film. The background pressure should be kept below 10−4 mbar in order to maintain a ballistic propagation of the vapor. If the material is an organic, it will typically be thermally insulating and will thus only be heated in regions where it is in direct contact with the heated boat. This could create small voids in the material that occasionally, due to gravity, will collapse, resulting in rapid changes in the evaporation rate, which will affect the uniformity of the resulting film. Impurities could originate from another previously evaporated material. It is unavoidable that previously used material to some degree desorbs from the inner walls of the vacuum chamber and readsorbs at the substrate. Previously evaporated material could also desorb from the inner walls as flakes, which could cause problems with dust ending up on the substrate. Impurities could also originate from a number of other places. The effect of impurities on the deposition process depends on the reactivity of the impurity and the reactivity of the deposited material. For example, gold has a low reactivity in general and produces highly pure films (even in a poor vacuum) in contrast to the highly reactive aluminum, which reacts with and incorporates almost all contaminants arriving at the surface. The source material could also be contaminated. For some metals such as, for example, aluminum, the outer surface will consist of the corresponding oxide. Aluminum oxide melts at a higher temperature than aluminum metal; thus when the metal melts a crust of aluminum oxide forms that can coevaporate with the metal and form an intermediate layer unless the substrate can be shielded from the evaporation source during
140
Chapter 3
initial heating. For organic solar cells, this constitutes a problem. Aluminum oxide is an insulator; thus, it acts like an electrical barrier layer, which is undesirable in an organic solar cell where charge transport is vital for the cell performance. 3.4.1.3 Annealing
Annealing is a heat treatment of a material inflicting possible changes in the physicochemical properties such as strength, hardness, internal stress, and thermal and electric conductivity. In an annealing process, the temperature is maintained at a constant suitable level, which produces equilibrium conditions, followed by very slow cooling. The basic annealing process can be described in three stages: (1) removal of crystal defects (that cause internal stress), resulting in softening of the material; (2) recrystallization by nucleation and initial growth of new grains that replace those deformed by the internal stresses; (3) further grain growth, resulting in a coarsening of the microstructure that continues until a state of equilibrium is reached. Annealing can also be used to incorporate dopants into substitutional positions in the crystal lattice, which can be used to control the electrical properties of the material. There are only few controllable annealing parameters, which include (1) annealing temperature, (2) annealing time, and (3) cooling rate. The annealing temperature (Ta ) is chosen according to the glass transition temperature (Tg ), that is, recrystallization occurs for Ta ≥ Tg but not for Ta < Tg . The annealing time controls the extent of grain growth; shorter annealing times produce smaller grains. The cooling rate controls the uniformity of the film, that is, slower cooling rates produce more uniform films. Organic solar cells can be annealed as entire devices or at various fabrication steps. Either approach is problematic since it is not possible to anneal the individual layers separately, for example, the cathode is applied last and if a certain set of annealing conditions is used to control the morphology of the electrode, then the entire device will experience these annealing conditions, which are not necessarily suited for the other device components. Annealing is a simple process, but when applied on multilayer films such as organic solar cells, the applicability could become problematic depending on the application. The most important purpose of annealing organic solar cells is to control the crystallographic properties, that is, molecular packing and morphology, that is, nanoscale phase separation. Charge mobility is affected or controllable by the molecular packing, wherein a higher molecular ordering corresponds to a higher charge mobility, which corresponds to an improved photovoltaic performance. Furthermore, it is desirable to control the grain growth; for example, it is advantageous to produce polycrystalline films with large grain sizes with a corresponding low concentration of grain boundaries due to the fact that (discussed previously) grain boundaries can act as electrical carrier traps in semiconductors and as channels for diffusion of undesirable species.
Characterization of Organic Solar Cells
141
It is common to employ annealing after the fabrication of the organic solar cell. The problems with controlling the morphology of all the layers at the same time is typically surpassed by the overall improvements in photovoltaic performance. It makes sense to anneal after fabrication but before operation, because during operation of the organic solar cell the device will unavoidably be heated (i.e., annealed) from the illumination; thus, it is preferable to perform the annealing under controlled circumstances in order to obtain the optimal result. The annealing process is a separate treatment, but can also be used in conjunction with deposition methods to assist the morphological control. In that context, it is typically not regarded as an annealing process even though the same principles apply, although on a different time scale. For thermal evaporation, the substrate temperature can be used to control the deposition behavior, that is, lowering the substrate temperature inhibits molecular diffusion and thus inhibits the nucleation and coalescence processes. Heating the substrate for a spin-coating process is a bit more complicated. Besides from the effect of the annealing process on the morphology, the substrate heating will accelerate the solvent evaporation, which, according to previous discussion, is also a parameter affecting the morphology. Much of the work presented in the literature, especially the last couple of years, regarding the effect of annealing on organic solar cells, has focused on devices where the active layer is a blend of P3HT and PCBM.61–68 This is because this system is one of the most promising polymer solar cells developed to date in terms of stability and efficiency. The interesting thing about this system is the fact that the efficiency appears to be strongly dependent on the processing conditions, particularly the thermal annealing step. Kim et al.61 studied the device’s annealing effect on the photovoltaic performance of devices with a Al/P3HT:PCBM/PEDOT:PSS/ITO composition using blends of regioregular P3HT and PCBM. The outcome is shown in Fig. 3.42. They found, irrespective of the solvent (chlorobenzene or 1,2-dichlorobenzene), that when the annealing temperature is systematically increased, the optimal photovoltaic performance is observed at 140◦ C (when annealed for 15 min). The improved effect is more pronounced for the lower-boiling solvent (chlorobenzene), which the authors explain from the solvent evaporation effect on the morphology. The higher-boiling solvent (1,2-dichlorobenzene) evaporates slower and thus allow the P3HT to segregate toward the PEDOT:PSS layer. Vertical phase segregation could then form a structure that could help direct the charges toward the electrodes and reduce recombination, which is expected to improve generation of the current. On annealing, the device fabricated with the lower-boiling solvent (chlorobenzene) will be able to undergo phase segregation and develop a similar or equivalent morphology to that of the higher-boiling solvent (1,2-dichlorobenzene), that is, the annealing process compensates for the poor properties of the lower-boiling solvent. Ma et al.67 performed a study similar to that of Kim et al.,61 and obtained an opti-
142
Chapter 3
Figure 3.42 I sc (squares in top panel), FF (circles in top panel), V oc (squares in bottom panel), and power conversion efficiency (PCE) (circles in bottom panel) as a function of annealing temperature for chlorobenzene (filled symbols) and 1,2-dichlorobenzene (open symbols) devices annealed for 4 min in air. (Reprinted with permission from Ref. [61]. Copyright 2005, American Institute of Physics.)
mal efficiency at 150◦ C [see Fig. 3.43(a)] in good agreement with the 140◦ C found by Kim et al.61 In addition, the authors measured the efficiency as a function of annealing temperature at the optimal annealing temperature (150◦ C), and found that the efficiency tends toward saturation after ∼30 min [Fig. 3.43(b)]. They observed a stable performance even after annealing for hours at 150◦ C. The sudden increase in efficiency at ∼50◦ C was explained from the fact that it is close to the glass transition temperature of P3HT, a fact that was confirmed by differential scanning calorimetry (DSC) measurements. Earlier it was common to observe a sudden drop in efficiency after a short time when annealing at high temperatures.53, 65 This was attributed to the formation of large PCBM aggregates.69 Today experimentalists have become better at controlling the morphology based on a better understanding of the effect of process conditions on the photovoltaic performance. Ma et al.67 performed another interesting experiment wherein it was shown that the Al/P3HT:PCBM interface plays an important role in the annealing process in regard to the efficiency. Two sets of devices were prepared: one where the annealing was carried out before deposition of the Al electrode, and one where the annealing was carried out after deposition of the Al electrode. The efficiency was measured for both sets of devices as a function of annealing temperature. The results are shown in Fig. 3.44.
Characterization of Organic Solar Cells
143
Figure 3.43 (a) Device efficiency versus annealing temperature. For these data, the annealing time was 15 min. (b) Evolution of device efficiency with thermal annealing time at 150◦ C. The device is composed of Al/P3HT:PCBM/PEDOT:PSS/ITO. (Reprinted from Ref. [67], with permission from Wiley-VCH Verlag GmbH & Co., copyright 2005.)
It is clear from Fig. 3.44 that annealing after the Al electrode deposition improves the efficiency at higher temperatures. The authors suggest that at elevated temperatures, processes such as Al diffusion and formation of chemical bonds between Al and P3HT:PCBM (e.g., C-Al or C-O-Al)70, 71 could result in stronger contacts and increased interfacial contact area. These conclusions were supported by
144
Chapter 3
Figure 3.44 Evolution of device efficiency with thermal annealing time at 150◦ C for devices composed of Al/P3HT:PCBM/PEDOT:PSS/ITO. Open squares correspond to annealing before deposition of the Al electrode. Closed squares correspond to annealing after deposition of the Al electrode. (Reprinted from Ref. [67], with permission from Wiley-VCH Verlag GmbH & Co., copyright 2005.)
AFM measurements. A significant increase in interfacial contact area is expected to result in a significant reduction in series resistance. Furthermore, the authors suggest that the presence of an Al electrode could possibly prevent P3HT:PCBM crystal overgrowth during annealing and thus prevent degradation of the morphology of the donor-acceptor interpenetrating networks. 3.4.2 Techniques for monitoring morphology A vital part of controlling morphology in an organic solar cell is being able to monitor it. The tools we use for monitoring morphology are various characterization techniques of which the application determines the choice, because different techniques produce different kinds of information. The techniques are typically physical and/or spectroscopic in nature. Chemical characterization techniques are not that common in morphology studies of organic solar cells. In the next chapter, various chemical characterization techniques are described in relation to degradation studies. The most common characterization techniques used for monitoring morphology in organic solar cells are SEM, TEM, AFM, and x-ray methodologies. For a more thorough description on morphology studies of organic solar cells using the most common characterization techniques, see Ref. [26].
Characterization of Organic Solar Cells
145
3.4.2.1 Electron microscopy: SEM and TEM
The basic principle of SEM analysis consist of bombarding a highly focused (1–5 nm) primary electron beam onto a surface. Secondary electrons are consequently emitted from the sample surface and detected. When rastering the primary electron beam over the sample surface, a secondary electron image is obtained, an SEM image. SEM has become a standard tool for visualizing morphology. The strength of the technique is the superior lateral resolution and the capability of analyzing a broad range of scales from the nanometer range to the millimeter range. The weakness is the lack of a depth scale and the fact that it is a vacuum technique. In some cases, depending on the detector, a transparency effect can complicate the interpretation of the image. This is demonstrated in the next chapter in relation to degradation studies of organic solar cells, where information stored below the Al electrode is visible in the SEM image. The superior lateral resolution of SEM can be utilized to visualize the bulk morphology by imaging a cross section of the device. Figure 3.45 shows SEM images of cross sections of the blended material MDMO-PPV:PCBM spin coated onto an ITO substrate using two different solvents (toluene and chlorobenzene). The figure illustrates the importance of the choice of solvent used for the spincoating process. The left image shows (chlorobenzene) spheres with diameters of ∼20 nm embedded in a matrix, that is, optimal morphology of the donor-acceptor interpenetrating networks. The right image (toluene) shows that large clusters have formed during the spin-coating process, that is, a poor morphology of the donoracceptor interpenetrating networks. In TEM, an electron beam is (like SEM) focused onto the sample surface, however, in this method the electrons are transmitted through the sample and detected on a fluorescent screen. Areas where the electrons are scattered appear dark on the
Figure 3.45 Cross-sectional SEM images of MDMO-PPV:PCBM (1:4 weight ratio) blend films spin coated from toluene (right) or chlorobenzene (left) solutions. (Reprinted from Ref. [73], with permission from Elsevier, copyright 2006.)
146
Chapter 3
Figure 3.46 High-resolution TEM images of regioregular P3HT:PCBM (1:0.8) blended films spin coated from chlorobenzene. (a) As spin coated. (b) Annealed at 150◦ C for 30 min. (c) Annealed at 150◦ C for 2 hr. (Reprinted from Ref. [67], with permission from Wiley-VCH Verlag GmbH & Co., copyright 2005.)
screen. TEM has been used extensively to analyze the nanostructure of the bulk heterojunction with a combination of a cross-sectional and planar view. Figure 3.46 demonstrates the applicability of TEM to monitor the effect of annealing (150◦ C) on the nanostructure of regioregular P3HT:PCBM blended films spin coated using chlorobenzene. Figure 3.46(a) represents the morphology before annealing, the donor-acceptor interpenetrating networks are not well developed (i.e., it is difficult to distinguish the donor-acceptor domains). After annealing for 30 min [Fig. 3.46(b)] the donor-acceptor domains starts to become visible, and after 2 hr [Fig. 3.46(c)] the domains are clear and easily visible as ∼10 nm spherical P3HT nanostructures, which constitutes an optimal morphology of the donor-acceptor interpenetrating networks. 3.4.2.2 Atomic force microscopy (AFM)
AFM produces a topographic map by scanning a needle across the surface. The weakness of the technique is the fact that only small areas can be analyzed (∼100 ×100 μm2 ) and the acquisition time is long; thus, it is not suited for screening purposes. The strength of AFM is the excellent height resolution and lateral resolution both in the angstrom range. Figure 3.47 is a good example of the usefulness of AFM to monitor the effect of a solvent on the nanostructure of MDMO-PPV:PCBM blended films spin coated using chlorobenzene [Fig. 3.47(a)] or toluene [Fig. 3.47(b)]. Figure 3.47(a) shows a nanostructure with features visible around 50 nm, and Fig. 3.47(b) shows features of several hundreds of nanometers, which is a large difference in the scale of phase separation. Furthermore, the toluene spin-coated film exhibits height variations
Characterization of Organic Solar Cells
147
Figure 3.47 AFM images of MDMO-PPV:PCBM (1:4 by weight) blended films spin coated using (a) chlorobenzene and (b) toluene. (Reprinted from Ref. [69], with permission from Wiley-VCH Verlag GmbH & Co., copyright 2004.)
that are one order of magnitude larger than those for the chlorobenzene spin-coated film. 3.4.2.3 X-ray diffraction (XRD)
In x-ray diffraction, a diffraction pattern of the sample is obtained by illuminating the sample with x rays. When the x rays hit the surface of the sample (or the single crystal), the electrons interact with the x rays and these are diffracted by their interactions with the electrons. The diffraction pattern is recorded by detectors. There are two primary methods in x-ray diffraction, which are powder diffraction and single-crystal diffraction. The data collected from the diffraction pattern is a reciprocal space representation of the crystal lattice. X-ray diffraction is particularly useful to monitor the structural molecular ordering of materials, which is highly relevant for applications regarding optimization of the charge transport properties in materials used for organic solar cells. Figure 3.48 demonstrates how XRD can be used to monitor the effect of annealing (150◦ C) on the crystallinity within the phase-separated networks of regioregular P3HT:PCBM blended films drop casted onto a PEDOT:PSS/ITO substrate. Figure 3.48 shows that annealing increases the crystallinity. The inset shows the P3HT crystal structure. The increased intensity (as a result of the annealing process) observed at ∼5 deg corresponds to the interchain spacing in P3HT associated with the interdigitated alkyl chains. The increased intensity at ∼23 deg corresponds to the interchain spacing associated with face-to-face packing of the thiophene rings.
148
Chapter 3
Figure 3.48 XRD spectra of a P3HT:PCBM blended film drop casted onto a PEDOT:PSS/ITO substrate with and without annealing at 150◦ C for 30 min. The inset shows the P3HT crystal structure. (Reprinted from Ref. [67], with permission from Wiley-VCH Verlag GmbH & Co., copyright 2005.)
References 1. http://rredc.nrel.gov/solar/spectra/am1.5/; ASTM international, Standard Tables for Reference Solar Spectral Irradiances: Direct Normal and Hemispherical on 37◦ Tilted Surface, G 173-03 (2003). 2. ASTM international, Standard Specification for solar simulation for photovoltaic testing, E 927-05, 2005; IEC 904-9, Photovoltaic devices—Part 9: Solar simulator performance requirements, IEC 904-9:1995(E) (1995). 3. http://www.spirecorp.com/spire-solar/solar-manufacturing-equipment/ testning-andsort-cells/spi-cell-tester.php (2007). 4. http://www.sciencetech-inc.com/Midicart/shop/Brochures%5CSS150% 20Brochure.pdf (2007). 5. http://www.photoemission.com/ (2007). 6. Yu, G., Gao, J., Hummelen, J.C., Wudl, F., and Heeger, A.J., “Polymer photovoltaic cells—enhanced efficiencies via a network of internal donor-acceptor heterojunctions,” Science, 270(5243), pp. 1789–1791 (1995). 7. Reyes-Reyes, M., Kim, K., and Carroll, D.L., “High-efficiency photovoltaic devices based on annealed poly(3-hexylthiophene) and 1-(3-methoxycar-
Characterization of Organic Solar Cells
149
bonyl)-propyl-1-phenyl-(6,6)C-61 blends,” Appl. Phys. Lett., 87, 083506 (2005). 8. Neugebauer, H., Brabec, C.J., Hummelen, J.C., and Sariciftci, N.S., “Stability and photodegradation mechanisms of conjugated polymer/fullerene plastic solar cells,” Sol. Energy Mater. Sol. Cells, 61, pp. 35–42 (2000). 9. Krebs, F.C., Carlé, J.E., Cruys-Bagger, N., Andersen, M., Lilliedal, M.R., Hammond, M.A., and Hvidt, S., “Lifetimes of organic photovoltaics: photochemistry, atmosphere effects and barrier layers in ITO-MEHPPV: PCBMaluminium devices,” Sol. Energy Mater. Sol. Cells, 86, pp. 499–516 (2005). 10. Norrman, K., and Krebs, F.C., “Lifetimes of organic photovoltaics: Using TOF-SIMS and O-18(2) isotopic labelling to characterise chemical degradation mechanisms,” Sol. Energy Mater. Sol. Cells, 90, pp. 213–227 (2006). 11. Schuller, S., Schilinsky, P., Hauch, J., and Brabec, C.J., “Determination of the degradation constant of bulk heterojunction solar cells by accelerated lifetime measurements,” Appl. Phys. A, 79, pp. 37–40 (2004). 12. Krebs, F.C., and Spanggaard, H., “Significant improvement of polymer solar cell stability,” Chem. Mater., 17, pp. 5235–5237 (2005). 13. De Bettignies, R., Leroy, J., Firon, M., and Sentein, C., “Accelerated lifetime measurements of P3HT : PCBM solar cells,” Synth. Met., 156, pp. 510–513 (2006). 14. Czanderna, A.W., and Pern, F.J., “Encapsulation of PV modules using ethylene vinyl acetate copolymer as a pottant: A critical review,” Sol. Energy Mater. Sol. Cells, 43, pp. 101–181 (1996). 15. Rosenthal, A.L., and Lane, C.G., “Field-test results for the 6 MW Carrizo solar photovoltaic power plant,” Sol. Cells, 30, pp. 563–571 (1991). 16. Berman, D., Biryukov, S., and Faiman, D., “EVA laminate browing after 5 years in a grid-connected, mirror-assisted photovoltaic system in the Negev desert—Effect on module efficiency,” Sol. Energy Mater. Sol. Cells, 36, pp. 421–432 (1995). 17. Berman, D., and Faiman, D., “EVA browning and the time-dependence of I-V curve parameters on PV modules with and without mirror-enhancement in a desert environment,” Sol. Energy Mater. Sol. Cells, 45, pp. 401–412 (1997). 18. International Standard CEI/IEC 60904-3: Photovoltaic devices, Part 3, Measurement principles for terrestrial photovoltaic (PV) solar devices with reference spectral irradiance data, International Electrotechnical Commission, Geneva, Switzerland (1989). 19. International Standard CEI/IEC 60904-1: Photovoltaic devices, Part 1, Measurement of current–voltage characteristics, International Electrotechnical Commission, Geneva, Switzerland (1989).
150
Chapter 3
20. Bird, R.E., and Hulstrom, R.L., “Review, evaluation, and improvement of direct irradiance models,” ASME J. Sol. Energy Eng., 103, pp. 182–192 (1981). 21. C. Riordan, D. Myers, M. Rymes, R. Hulstrom, W. Marion, C. Jennings, and C. Whitaker, “Spectral solar radiation data base at SERI,” Sol. Energy, 42, pp. 67–79 (1989). 22. Katz, E.A., “Fullerenes for photovoltaics,” Encyclopedia of Nanoscience and Nanotechnology, H.S. Nalwa (Ed.), American Scientific Publishers, Los Angeles, Vol. 3, pp. 661–683 (2004). 23. Katz, E.A., “Fullerene thin films as photovoltaic material,” Nanostructured Materials for Solar Energy Conversion, T. Soga (Ed.), Elsevier, New York, pp. 359–446 (2006). 24. Katz, E.A., Faiman, D., Tuladhar, S.M., Kroon, J.M., Wienk, M.M., Fromherz, T., Padinger, F., Brabec, C.J., and Sariciftci, N.S., “Temperature dependence for the photovoltaic device parameters of polymer-fullerene solar cells under operating conditions,” J. Appl. Phys., 90, pp. 5343–5350 (2001). 25. Kroon, J.M., Wienk, M.M., Verhees, W.J.H., and Hummelen, J.C., “Accurate efficiency determination and stability studies of conjugated polymer/fullerene solar cells,” Thin Solid Films, 403, pp. 223–228 (2002). 26. Shrotriya, V., Li, G., Yao, Y., Moriarty, T., Emery, K., and Yang, Y., “Accurate measurement and characterization of organic solar cells,” Adv. Funct. Mater., 16, pp. 2016–2023 (2006). 27. Berman, D., Faiman, D., and Farhi, B., “Sinusoidal spectral correction for high precision outdoor module characterization,” Sol. Energy Mater. Sol. Cells, 58, pp. 253–264 (1999). 28. Katz, E.A., Faiman, D., Cohen, Y., Padinger, F., Brabec, C., and Sariciftci, N.S., “Temperature and irradiance effect on the photovoltaic parameters of a fullerene/conjugated-polymer solar cell,” Proc. SPIE 4108, pp. 117–124 (2000). 29. Katz, E.A., Gevorgyan, S., Orynbayev, M.S., and Krebs, F.C., “Out-door testing and long-term stability of plastic solar cells,” Eur. Phys.–Appl. Phys., 36, pp. 307–311 (2006). 30. Faiman, D., Feuermann, D., and Zemel, A., “Accurate field calibration of pyranometers,” Sol. Energy, 49, pp. 489–492 (1992). 31. Faiman, D., Feuermann, D., Ibbetson, P., Medwed, B., Zemel, A., Ianetz, A., Liubansky, V., Setter, I., and Suraqui, S., “The Negev radiation survey,” ASME J. Sol. Energy Eng., 126, pp. 906–914 (2004). 32. Shaheen, S.E., Brabec, C.J., Sariciftci, N.S., Padinger, F., Fromherz, T., and Hummelen, J.C., “2.5% efficient organic plastic solar cells,” Appl. Phys. Lett., 78, pp. 841–843 (2001).
Characterization of Organic Solar Cells
151
33. Dyakonov, V., “The polymer-fullerene interpenetrating network: one route to a solar cell approach,” Physica E, 14, pp. 53–60 (2002). 34. Nelson, J., “Diffusion-limited recombination in polymer-fullerene blends and its influence on photocurrent collection,” Phys. Rev. B, 67, 155209 (2003). 35. Koster, L.J.A., Mihailetchi, V.D., Ramaker, R., and Blom, P.W.M., “Light intensity dependence of open-circuit voltage of polymer:fullerene solar cells,” Appl. Phys. Lett., 86, 123509 (2005). 36. Pacios, R., Bradley, D.D.C., Nelson, J., and Brabec, C.J., “Efficient polyfluorene based solar cells,” Synth. Met., 137, pp. 1469–1470 (2003). 37. Riedel, I., and Dyakonov, V., “Influence of electronic transport properties of polymer-fullerene blends on the performance of bulk heterojunction photovoltaic devices,” Physica Status Solidi A, 201, pp. 1332–1341 (2004). 38. Krebs, F.C., “Encapsulation of polymer photovoltaic prototypes,” Sol. Energy Mater. Sol. Cells, 90, pp. 3633–3643 (2006). 39. Krebs, F.C., Alstrup, J., Biancardo, M., and Spanggaard, H., “Large area polymer solar cells,” Proc. SPIE, 5938, 593804 (2005). 40. Biancardo, M., West, K., and Krebs, F.C., “Optimizations of large area quasisolid-state dye-sensitized solar cells,” Sol. Energy Mater. Sol. Cells, 16, pp. 2575–2588 (2006). 41. Katz, E.A., Zinevich, S., Faiman, D., Melnichak, V., and Kroon, J., “Outdoor testing of dye-sensitised nanocrystalline solar cells,” Proc. of 11th Sede Boqer Symposium on Solar Electricity Production, D. Faiman (Ed.), pp. 97–100 (2002). 42. Gordon, J.M., Katz, E.A., Feuermann, D., and Huleihil, M., “Toward ultrahigh-flux photovoltaic concentration,” Appl. Phys. Lett., 84, pp. 3642–3644 (2004). 43. Gordon, J.M., Katz, E.A., Tassew, W., and Feuermann, D., “Photovoltaic hysteresis and its ramifications for concentrator solar cell design and diagnostics,” Appl. Phys. Lett., 86, 073508 (2005). 44. Katz, E.A., Gordon, J.M., and Feuermann, D., “Effects of ultra-high flux and intensity distribution in multi-junction solar cells,” Progr. Photovoltaics, 14, pp. 297–303 (2006). 45. Katz, E.A., Gordon, J.M., Tassew, W., and Feuermann, D., “Photovoltaic characterization of concentrator solar cells by localized irradiation,” J. Appl. Phys., 100, 044514 (2006). 46. Xue, J., Uchida, S., Rand, B.P., and Forrest, S.R., “4.2% efficient organic photovoltaic cells with low series resistances,” Appl. Phys. Lett., 84, pp. 3013– 3015 (2004).
152
Chapter 3
47. Norrman, K., Larsen, N.B., Ghanbari-Siahkali, A., “Studies of spin-coated polymer films,” Ann. Rep. Progr. Chem., Sect. C: Phys. Chem., 101, pp. 174– 201 (2005). 48. Spangler, L.L., Torkelson, J.M., and Royal, J.S., “Influence of solvent and molecular-weight on thickness and surface-topography of spin-coated polymer-films,” Polym. Eng. Sci., 30, pp. 644–653 (1990). 49. Sirringhaus, H., Tessler, N., and Friend, R.H., “Integrated optoelectronic devices based on conjugated polymers,” Science, 280, pp. 1741–1744 (1998). 50. Sirringhaus, H., Brown, P.J., Friend, R.H., Nielsen, M.M., Bechgaard, K., Langeveld-Voss, B.M.W., Spiering, A.J.H., Janssen, R.A.J., Meijer, E.W., Herwig, P., and de Leeuw, D.M., “Two-dimensional charge transport in selforganized, high-mobility conjugated polymers,” Nature, 401, pp. 685–688 (1999). 51. Sirringhaus, H., Brown, P.J., Friend, R.H., Nielsen, M.M., Bechgaard, K., Langeveld-Voss, B.M.W., Spiering, A.J.H., Janssen, R.A.J., and Meijer, E.W., “Microstructure-mobility correlation in self-organised, conjugated polymer field-effect transistors,” Synth. Met., 111-112, pp. 129–132 (2000). 52. Apperloo, J.J., Janssen, R.A.J., Nielsen, M.M., and Bechgaard, K., “Doping in solution as an order-inducing tool prior to film formation of regio-irregular polyalkylthiophenes,” Adv. Mater., 12, pp. 1594–1597 (2000). 53. Chang, J.-F., Sun, B., Breiby, D.W., Nielsen, M.M., Sølling, T.I., Giles, M., McCulloch, I., and Sirringhaus, H., “Enhanced mobility of poly(3-hexylthiophene) transistors by spin-coating from high-boiling-point solvents,” Chem. Mater., 16, pp. 4772–4776 (2004). 54. Krebs, F.C., Hoffmann, S.V., and Jørgensen, M., “Orientation effects in selforganized, highly conducting regioregular poly(3-hexylthiophene) determined by vacuum ultraviolet spectroscopy,” Synth. Mater., 138, pp. 471–474 (2003). 55. Kline, R.J., McGehee, M.D., Kadnikova, E.N., Liu, J., and Fréchet, J.M.J., “Controlling the field-effect mobility of regioregular polythiophene by changing the molecular weight,” Adv. Mater., 15, pp. 1519–1522 (2003). 56. Bao, Z., Dodabalapur, A., and Lovinger, A.J., “Soluble and processable regioregular poly(3-hexylthiophene) for thin film field-effect transistor applications with high mobility,” Appl. Phys. Lett., 69, pp. 4108–4110 (1996). 57. Wang, G., Swensen, J., Moses, D., and Heeger, A.J., “Increased mobility from regioregular poly(3-hexylthiophene) field-effect transistors,” J. Appl. Phys., 93, pp. 6137–6141 (2003). 58. Arias, A.C., Corcoran, M., Banach, M., Friend, R.H., and MacKenzie, J.D., “Vertically segregated polymer-blend photovoltaic thin-film structures through
Characterization of Organic Solar Cells
153
surface-mediated solution processing,” Appl. Phys. Lett., 80, pp. 1695–1697 (2002). 59. Arias, A.C., MacKenzie, J.D., Stevenson, R., Halls, J.J.M., Inbasekaran, M., Woo, E.P., Richards, D., and Friend, R.H., “Photovoltaic performance and morphology of polyfluorene blends: A combined microscopic and photovoltaic investigation,” Macromolecules, 34, pp. 6005–6013 (2001). 60. Smith, D.L., Thin Film Deposition Principles and Practice, McGraw-Hill, New York (1995). 61. Kim, Y., Choulis, S.A., Nelson, J., Bradley, D.D.C., Cook, S., and Durrant, J.R., “Device annealing effect in organic solar cells with blends of regioregular poly(3-hexylthiophene) and soluble fullerene,” Appl. Phys. Lett., 86, 063502 (2005). 62. Kim, Y., Choulis, S.A., Nelson, J., Bradley, D.D.C., Cook, S., and Durrant, J.R., “Composition and annealing effects in polythiophene/fullerene solar cells,” J. Mater. Sci., 40, pp. 1371–1376 (2005). 63. Padinger, F., Rittberger, R.S., and Sariciftci, N.S., “Effects of postproduction treatment on plastic solar cells,” Adv. Funct. Mater., 13, pp. 85–88 (2003). 64. Camaioni, N., Ridolfi, G., Casalbore-Miceli, G., Possamai, G., and Maggini, M., “The effect of a mild thermal treatment on the performance of poly(3allkylthiophene)/fullerene solar cells,” Adv. Mater., 14, pp. 1735–1738 (2002). 65. Chirvase, D., Parisi, J., Hummelen, J.C., and Dyakonov, V., “Influence of nanomorphology on the photovoltaic action of polymer-fullerene composites,” Nanotechnology, 15, pp. 1317–1323 (2004). 66. Yang, X., Loos, J., Veenstra, S.C., Verhees, W.J.H., Wienk, M.M., Kroon, J.M., Michels, M.A.J., and Janssen, R.A.J., “Nanoscale morphology of highperformance polymer solar cells,” Nano Lett., 5, pp. 579–583 (2005). 67. Ma, W., Yang, C., Gong, X., Lee, K., and Heeger, A.J., “Thermally stable, efficient polymer solar cells with nanoscale control of the interpenetrating network morphology,” Adv. Funct. Mater., 15, pp. 1617–1622 (2005). 68. Chirvase, D., Chiguvare, Z., Knipper, M., Parisi, J., Dyakonov, V., and Hummelen, J.C., “Temperature dependent characteristics of poly(3 hexylthiophene)-fullerene based heterojunction organic solar cells,” J. Appl. Phys., 93, pp. 3376–3383 (2003). 69. Hoppe, H., Niggemann, M., Winder, C., Kraut, J., Hiesgen, R., Hinsch, A., Meissner, D., and Sariciftei, N.S., “Nanoscale morphology of conjugated polymer/fullerene-based bulk-heterojunction solar cells,” Adv. Funct. Mater., 14, pp. 1005–1011 (2004).
154
Chapter 3
70. Birgerson, J., Fahlman, M., Broms, P., and Salaneck, W.R., “Conjugated polymer surfaces and interfaces: A mini-review and some new results,” Synth. Met., 80, pp. 125–130 (1996). 71. Logdlund, M., and Bredas, J.-L., “Theoretical-studies of the interaction between aluminium and poly(p-phenylenevinylene) and derivatives,” J. Chem. Phys., 101, pp. 4357–4364 (1994). 72. Hoppe, H., and Sariciftci, N.S., “Morphology of polymer/fullerene bulk heterojunction solar cells,” J. Mater. Chem., 16, pp. 45–61 (2006). 73. Mozer, A.J., and Sariciftci, N.S., “Conjugated polymer photovoltaic devices and materials,” Comptes Rend. Chim., 9, pp. 568–577 (2006).
Chapter 4
Lifetime and Stability Studies Kion Norrman, Stéphane Cros, Rémi de Bettignies, Muriel Firon and Frederik C. Krebs 4.1 Overview The development of organic solar cells has almost exclusively focused on optimizing the device performance by studying the influence of device geometry and phase morphology on the photovoltaic response.1–4 Efficiencies of ∼5% have been achieved and are sufficient for various low-energy applications.5–7 The major limitation for widespread applications is the relatively short lifetime of the device manifested in insufficient device stability. One challenge that remains to be resolved is the problem of stability. All organic solar cells are to some extent unstable and degrade over time. Many degradation mechanisms have been identified, but not necessarily corrected for in Refs. [8], [9], [10], [11] and [12]. Furthermore, it is not always straightforward to determine to what extent a particular degradation mechanism contributes to the overall degradation of the photovoltaic response. The most common degradation mechanisms are as follows: • Diffusion of oxygen and water into the device and subsequent reaction with the active components. • Direct photodegradation of the active components. • Thermal degradation of the active components. • Corrosion at electrode interfaces. • Interlayer diffusion, for example, diffusion of electrode material into the device and subsequent possible reaction with the active components. • Particle formation within the device. The first three mechanisms described above are externally applied conditions, and the last three are internal processes. The mechanisms are interrelated; for example, illumination increases the temperature, and both of these properties facilitate oxygen- and water-induced oxidation. The temperature affects the diffusion phenomena. A lot of effort has been put into the development of barrier layers in order to remove or minimize diffusion of oxygen and water into the device.
156
Chapter 4
The archetypical polymer solar-cell devices based on PPV materials are notoriously unstable. From the moment the metal electrode is applied, the device is subject to degradation in vacuum, both in the dark and under illumination. The polythiophenes are more stable and the current state of the art employs morphologically stable bulk heterojunctions of regioregular P3HT as the donor material and the soluble fullerene material PCBM as the acceptor material. Stable device operation for a thousand hours has been documented for this system;13 however, wellfounded literature reports that stable polymer photovoltaics exist are scarce. Stable operation for more than ten thousand hours have been reported for bilayer heterojunctions between regiorandom poly(3-carboxydithiophene) (P3CT) and C60 .14 The degradation behavior of organic solar cells is complex and cannot be described by a single process, but rather as multiple processes that could be interrelated, thus adding to the complexity. To prevent or minimize degradation, it is a prerequisite to identify and understand the details of the degradation mechanisms. This is typically attempted by systematic experiments involving the use of physical and chemical characterization techniques. Typical and useful examples of characterization techniques used as tools to study degradation mechanisms in organic solar cells are presented in the following sections. This involves a fairly thorough description of time-of-flight secondary-ion-mass-spectrometry (TOF-SIMS) methodologies that are used for chemical characterization, including examples that illustrate the applicability of the technique for degradation studies. The following sections deal with x-ray photoelectron spectroscopy (XPS) and Rutherford backscattering (RBS) studies of degradation mechanisms, which include descriptions of the techniques. Then a description is given of the applicability of various physical and/or spectroscopic characterization techniques in regard to degradation studies, including examples of the usefulness of correlating with results obtained from chemical characterization. A brief description of accelerated lifetime measurements for extended periods of time is then presented. Finally, in the last section an apparatus for lifetime measurements and for isotope labeling is described.
4.2 Studies of Degradation Mechanisms Using TOF-SIMS TOF-SIMS has proven to be a valuable tool to study degradation mechanisms in organic solar cells.8–12 However, the technique is not well established within the field of organic photovoltaics. Furthermore, TOF-SIMS is a versatile technique that can be utilized in different ways depending on the application.15 The basic principles are briefly described in order to clarify the different modes of TOF-SIMS and what type of information that can be extracted from such an analysis. 4.2.1 Principle of TOF-SIMS A sample surface is bombarded in vacuo with ions of high kinetic energy (primary ions). The resulting cascade effect in the material causes species from the top ∼1 nm of the surface to be ejected. A small fraction of these species have either a positive or negative charge (secondary ions). The ions are typically a mixture
Lifetime and Stability Studies
157
Figure 4.1 Schematic showing (A) the SIMS ionization process, (B) the SEM [energy-dispersive x-ray analysis (EDX)] process, (C) the matrix-assisted laser desorption/ionization–time-of-flight (MALDI-TOF) process, and (D) the XPS process.
of atomic ions, molecular fragment ions, cluster ions, and molecular ions (if the molecular ion is sufficiently stable and not too large). The described process is schematically shown in Fig. 4.1(A). The process is known as the SIMS ionization process. There are similarities between the SIMS process and other well-known vacuum surface techniques such as SEM [Fig. 4.1(B)], MALDI-TOF [Fig. 4.1(C)], and XPS [Fig. 4.1(D)]. All the processes involve bombarding the surface with primary ions or photons and detecting the secondary ions or photons. The similarities for which the various analysis processes are performed are not necessarily reflected in the type of information obtained, that is, complementary information is often the outcome. If the ejected secondary ions for the SIMS process [Fig. 4.1(A)] are subsequently analyzed in a time-of-flight (TOF) analyzer that performs a mass analysis, then the technique is referred to as TOF-SIMS.15 The basic outcome of a TOF-SIMS analysis is a mass spectrum of the sample surface. The outcome of a MALDI-TOF analysis is also a mass spectrum, but the typical use involves mixing the sample with a matrix and applying it to a surface (i.e., not a sample surface). In a TOF-SIMS analysis it is the surface of the sample that is directly analyzed. The primary ion beam in a TOF-SIMS analysis is rastered over the surface area in order to minimize surface damage, and the average surface-mass spectrum is used. The analyzed surface area is divided into pixels where each pixel represents a mass spectrum. The intensity distribution of a mass spectral marker for a given species on the surface then provides an ion intensity image of that species over the analyzed area. A color intensity bar is used to visualize the lateral (in-plane) distribution of the species in question. The process is schematically shown in Fig. 4.2 for an ion image consisting of 9 × 9 pixels. The number of pixels commonly used is in the range 64 × 64 pixels to 2048 × 2048 pixels, depending on the application.
158
Chapter 4
Figure 4.2 Schematic of creating an ion image from TOF-SIMS data. A particular peak intensity from the mass spectrum (A) is converted to a color via a color intensity bar (B) and the pixel from which the mass spectrum is based is assigned this color (C). If the procedure is performed for all the pixels, an ion image is obtained (D).
Figure 4.3 Schematic of a TOF-SIMS depth profile through a multilayer film (A). The surface is bombarded with sputter ions (B) that consequently remove material from the surface, thus creating a hole. By switching between sputter ions (B) and analysis ions (primary ions), secondary ions (C) can be collected/detected as a function of the sputter time. A depth profile is constructed by plotting the intensity of the secondary ions (specific ions from each layer) against the sputter time (D).
Whereas TOF-SIMS imaging is used to visualize lateral (in-plane) distribution of chemistry, TOF-SIMS depth profiling is used to visualize vertical (in-depth) distribution of chemistry. This is done by complementing the analysis process with a sputter process. Sputter ions have different properties from analysis ions (primary ions). When sputter ions are rastered over the sample surface, material is consequently removed from the surface thus creating a hole. Switching between sputter ions and analysis ions (primary ions) enables the collection/detection of secondary ions as a function of sputter time. The depth profile is constructed by plotting the intensity of the secondary ions as a function of sputter depth. The sputter rate varies for different materials; thus, it is not straightforward to convert the sputter time to depth. Complementary techniques are typically used to determine the film thickness. A schematic view of the depth profiling process is shown in Fig. 4.3.
Lifetime and Stability Studies
159
4.2.2 Isotopic labeling TOF-SIMS being a mass spectrometric technique, it is capable of distinguishing between different isotopes. This can be utilized in various ways. One obvious way is to study oxygen or water diffusion into an organic solar cell, for example, during operation. If an organic solar cell is exposed to 18 O2 or H2 18 O, it is possible to track 18 O in a subsequent TOF-SIMS analysis, which makes it possible to identify the oxidative degradation mechanisms that take place during an experiment, and thus distinguish these from oxidative degradation prior to the experiment (e.g., during construction and handling of the device) and oxidative degradation after the experiment but before the analysis. The experimental setup is schematically shown in Fig. 4.4. Unfortunately, it is not just a question of detecting the intensity of 18 O (or 18 O derivatives) in various parts of the device after the experiment. However, there is one minor problem: ambient oxygen contains 0.2% 18 O. To avoid this problem the 18 O/16 O ratio is monitored instead. Oxygen incorporation (via 18 O or H 18 O) is 2 2 thus evident if the measured ratio is higher than the natural ratio of 0.2%. Furthermore, it only makes sense to quantitatively compare measured ratios within the same layer in the device since the various layers must be expected to contain different concentrations of natural oxygen due to different concentrations of possible native oxygen, contamination, and/or reaction products from 16 O2 or H2 16 O. In addition, monitoring the 18 O/16 O ratio removes possible undesired matrix effects; that is, the matrix effect should be the same for 16 O and 18 O. 4.2.3 TOF-SIMS depth profiling Organic solar cells have a multilayered thin-film structure with a total thickness in the nanometer range (disregarding the substrate), and a width and length in the
Figure 4.4 Schematic setup of an experiment where an isotopically labeled atmosphere is utilized to map the oxidative degradation mechanisms.
160
Chapter 4
Figure 4.5 TOF-SIMS depth profiles of an organic solar cell with the configuration Al/C60 /C12 -PPV/PEDOT:PSS/ITO. The study is described in greater detail in Refs. [11] and [12]. The analysis ions consisted of 1 pA of 25-keV Ga+ and the sputter ions con− sisted of 6.5 nA of 1.5-keV Xe+ . AlO− 2 is a marker for aluminum oxide, C4 is a marker for C60 , and C2 Al− is a marker for the Al/C60 interface. C12 corresponds to dodecyl.
centimeter range (small laboratory test cells). Such devices are obvious candidates for TOF-SIMS depth profiling analysis. However, even though the dimensions of the device seem ideal for the analysis, there are many aspects of the analysis that are problematic. These problems are discussed in this section and examples are given on degradation studies where TOF-SIMS depth profiling has proven to be a useful tool in spite of the complex nature of the technique. An organic solar cell with the configuration Al/C60 /C12 -PPV/PEDOT:PSS/ITO was subjected to illumination (1000 W m−2 , AM 1.5) in an 18 O2 /N2 (20:80) atmosphere (1 atm) for 45 hr, which corresponds to a 99% decrease in efficiency (ηe ). The device was subsequently subjected to a TOF-SIMS depth profiling analysis. Figure 4.5 shows selected profiles for the first 200 min of the sputter process corresponding to the Al layer and part of the C60 layer. The sudden increase in intensity at 65 min for the mass spectral marker C− 4 reveals the temporal location of the Al/C60 interface, that is, C− 4 is a marker for C60 . One of the drawbacks of the sputter process is degradation of the molecular structure of the surface. This is the reason why the molecular ion C− 60 is not available as a marker for C60 , which would have been a unique marker. It is typically necessary to use atomic or small fragment ions as markers for the individual layers. Sometimes there are no unique markers available. The mass spectral marker C2 Al− (Fig. 4.5) is interesting in that it could be the result of Al having formed an adduct with C60 . It is known that certain metals can react and form adducts with double bonds in organic species. This is, however, not necessarily the case in this experiment; C2 Al− could have formed in the ionization
Lifetime and Stability Studies
161
plasma from Al and C60 (or C60 fragments). This is a good example of how important it is to take care when attempting to interpret observations based on TOF-SIMS data. Regardless of the origin, C2 Al− can be regarded as a very specific marker for the Al/C60 interface. C60 produces only Cn − and Cm Hn − clusters during the ionization process (if it is not in physical contact with other matter). The same fragment ions are formed from all organic species; thus, all the Cn − and Cm Hn − fragment ions are generic and not optimal as a marker, for example, organic contamination could interfere.15 16 − The AlO− 2 (i.e., Al O2 ) profile provides interesting information. Aluminum oxide must form before the experiment begins that is, during fabrication and handling of the device. The outer aluminum oxide layer is expected to be formed instantly when exposed to ambient air, but according to the AlO− 2 profile (Fig. 4.5), the oxide is present through the entire Al electrode, and accumulated at the Al/C60 interface. There are two possible explanations for this phenomenon. First, oxygen diffusion through the Al electrode would result in the formation of aluminum oxide through the entire electrode. At the Al/C60 interface, nonreacted oxygen could then react with possible adducts between aluminum and C60 forming, once again, aluminum oxide. The thickness of the Al electrode for the device in question (Fig. 4.5) is ∼22 nm. The AlO− 2 profile for a 100-nm thick Al electrode is shown in Fig. 4.6. It is evident that aluminum oxide is only present on the outer Al electrode surface and in the Al/C60 interface. The fabrication and handling time for the two devices (Figs. 4.5 and 4.6) were approximately the same. Since no aluminum oxide is observed for the thicker Al electrode, it implies that oxygen (16 O2 ) does not have time to diffuse through the Al electrode to any significant degree during fabrication and handling. The second, and more likely, explanation deals with the fact that when the Al electrode is deposited on the device via evaporation, there will unavoidably be aluminum oxide present on the surface of the aluminum being evaporated. However, the boiling point for aluminum is ∼460◦ C lower than for aluminum oxide, which suggests that the temperature might have been too high during evaporation, causing aluminum oxide to be mixed in initially during the deposition until the relatively small amounts of aluminum oxide are used up. It is not possible in a TOF-SIMS depth-profiling analysis to distinguish between aluminum oxide and a mixture of aluminum oxide and aluminum. The presence of oxygen compounds in the material has a tendency to facilitate the ionization process and thus increase the total ion intensity. This is known as the matrix effect. This phenomenon is particularly pronounced for metal oxides that are good electron conductors, for example, ITO. The phenomenon is evident for the C− 4 profile (Figs. 4.5 and 4.6), which exhibits an increased intensity at the Al/C60 interface caused by the presence of aluminum oxide. Aluminum oxide is undesirable because of the significantly lower electric conductivity compared to aluminum. Specifically, the aluminum oxide is unwanted in the bulk of the aluminum electrode and especially at the Al/C60 interface. A thin layer of aluminum oxide on the outer Al electrode is
162
Chapter 4
Figure 4.6 TOF-SIMS depth profiles of an organic solar cell with the configuration Al/C60 /P3CT/ITO. P3CT corresponds to poly(3-carboxydithiophene). The analysis ions con− + sisted of 200 fA of 25-keV Bi+ 3 and the sputter ions consisted of 19 nA of 3-keV Xe . AlO2 − − is a marker for aluminum oxide, C4 is a marker for C60 , S is a marker for P3CT, and InO− 2 is a marker for ITO.
of less importance. The depth profiling results suggest that the Al electrode should be deposited without the presence of aluminum oxide. If this is not possible, the next best thing is to attempt to control the evaporation temperature as much as possible, and to avoid using Al electrodes that are too thin (e.g., ∼22 nm). The device (Fig. 4.5) was illuminated in an 18 O2 /N2 atmosphere in order to attempt to track the oxidative degradation mechanism, that is, possible reactions between 18 O2 and device components. The 18 O/16 O ratio profile in Fig. 4.5 provides valuable information. The 18 O/16 O ratio exceeds the natural ratio of 0.2% all the way through the Al electrode and through the Al/C60 interface. 18 O2 has reacted either with aluminum and/or exchanged oxygen with aluminum oxide. At the Al/C60 interface, the same is possible together with the possibility that 18 O2 has reacted with possible adducts between aluminum and C60 . In the C60 material, the 18 O− signal becomes so weak that the profile becomes too noisy for an adequate measurement. The sputter process significantly decreases the sensitivity; thus, it is possible that 18 O2 has reacted even further into the device. Based on the 18 O/16 O ratio profile, it can be concluded that oxygen diffuses through the Al electrode when illuminated in an oxygen-containing atmosphere. Figure 4.6 depicts depth profiles through an entire organic solar cell with the configuration Al/C60 /P3CT/ITO. The device was subjected to illumination (1000 W m−2 , AM 1.5) in vacuum for 10,760 hr until the short circuit current Isc had dropped to 10% of the initial value. The device was then cooled (72→ 25◦ C) and an atmosphere, composed of 18 O2 /N2 (20:80) to a pressure of 1 atm, was introduced. The device was then left in the dark for another 2240 hr until a total
Lifetime and Stability Studies
163
experimental time of 13,000 hr was reached. The device was subsequently subjected to a TOF-SIMS depth profiling analysis. The device described in Fig. 4.5 was subjected to illumination for a relatively short time (45 hr) in an 18 O2 /N2 (20:80) atmosphere. This corresponds to activated conditions (light and heat from the light) without any encapsulation, that is, a worst-case scenario. The device described in Fig. 4.6 was subjected to illumination for a very long time (10,760 hr) and then stored in darkness for a very long time (2240 hr) in an 18 O2 /N2 (20:80) atmosphere. This corresponds to activated conditions with practically perfect encapsulation followed by nonactivated conditions with no encapsulation. The mass spectral marker C− 4 is produced by the C60 layer as well as by the P3CT layer. However, S− is produced solely by the P3CT layer; thus, it is still pos− sible to distinguish the two layers. The intensities of the two markers (C− 4 and S ) are observed to increase drastically from ∼105 min (Fig. 4.6). This is caused by the phenomenon described earlier, the so-called matrix effect, in this particular case caused by the ITO layer. ITO produces a very strong matrix effect. It is sufficient to be in the vicinity of the ITO layer. When going from the P3CT layer to the P3CT/ITO interface, the concentration of P3CT decreases, which should cause less intense marker signals to be produced. However, the matrix effect is so powerful that it completely dominates and thus initially increases the signal instead. At some point when the P3CT concentration becomes sufficiently low, the concentration effect becomes dominant and the signal starts to decrease and finally disappears. − This explains the rather odd shapes of the P3CT marker profiles (C− 4 and S ) in Fig. 4.6, and it illustrates how TOF-SIMS data is not always straightforward to interpret. The 18 O/16 O ratio in the bulk Al electrode (at 5–35 min) for the device described in Fig. 4.6 reaches ∼10%; thus, 18 O incorporation is evident. The noisiness of the profile in that time window (5–35 min) is a consequence of the extremely low concentration of 16 O in that time window, that is, the denominator value becomes small and oscillates in time because the 16 O concentration is close to the noise level. At the Al/C60 interface, the 18 O/16 O ratio is ∼1% with a less noisy profile due to the higher concentration of 16 O. It is not possible to compare the 18 O/16 O ratio (∼0.4%) in the Al/C60 interface in Fig. 4.5 with the corresponding ratio (∼1%) in Fig. 4.6 since there is no way of knowing whether the 16 O concentration is the same for the two Al/C60 interfaces. However, it is justified to compare the 18 O/16 O ratios in the bulk C60 for the two devices (Figs. 4.5 and 4.6). The 18 O/16 O ratio in the bulk C60 (∼80 min.) in Fig. 4.6 is ∼17%, but it is not possible to measure any 18 O/16 O ratio in the bulk C for the device in Fig. 4.5, at least not by TOF-SIMS 60 depth profiling. Activated conditions (light and heat) for a relatively short period of time (45 hr) thus results in no oxygen incorporation in the C60 layer, or, at best, trace amounts for a device without encapsulation (Fig. 4.5). Nonactivated conditions (darkness at room temperature) for a relatively long period of time (2240 hr)
164
Chapter 4
results in significant oxygen incorporation in the C60 layer for a device without encapsulation (Fig. 4.6). The comparison of the two experiments would have been more informative if the experimental time had been the same. A device analogous to that described in Fig. 4.5 was subjected to the same conditions except for being wrapped in aluminum foil, which would prevent the device from being activated by light, but only by the heat produced by the light. The device showed oxygen incorporation to a lesser degree (not shown). There have been no systematic studies comparing, in mechanistic detail, the degradation caused by heat with degradation caused by light. Figure 4.7 shows TOF-SIMS depth profiles for a device that was illuminated (1000 W m−2 , AM 1.5) until the current (Isc ) had dropped to 1% of the initiated value. The atmosphere consisted of saturated H18 2 O, which should reveal information on water-induced degradation mechanisms. This example illustrates that besides oxygen, water also diffuses into the device. 18 O incorporation is clearly observed throughout the entire Al electrode and partly into the active layer of MEHPPV:PCBM. The device in Fig. 4.7 was manufactured with a relatively thick layer of MEHPPV:PCBM. Thick devices introduce further complications with respect to TOFSIMS depth profiling. First of all, the thicker the device the longer it takes to analyze. The total acquisition time for the depth profiling analysis of the device in Fig. 4.7 exceeded 13 hr, which can be explained from the fact that the sputter process is just one of several steps in a depth profiling analysis. It is very impractical to have acquisition times of this order; it makes it very difficult to per-
Figure 4.7 TOF-SIMS depth profiles of an organic solar cell with the configuration Al/MEHPPV:PCBM/PEDOT:PSS/ITO. The analysis ions consisted of 25-keV Bi+ 3 and the sputter − ions consisted of 3-keV Xe+ . AlO− 2 is a marker for aluminum oxide, C4 is a marker for MEH-PPV:PCBM and PEDOT:PSS, S− is a marker for PEDOT:PSS, and InO− 2 is a marker for ITO.
Lifetime and Stability Studies
165
form systematic studies involving a large number of devices. Another problem with thick devices and thus long sputter times is the phenomenon known as interlayer mixing caused by the sputter process (not to be mistaken for interlayer diffusion). The interlayer mixing is at its minimum at the beginning of the sputter process and gets more severe for longer sputter times. This explains why the PEDOT:PSS and ITO profiles are only just separated. Gradually increasing the interlayer mixing must be taken into account when interpreting TOF-SIMS depth profiling data. The sputter conditions should be chosen such that the acquisition time is practical, but not so harsh that interlayer mixing becomes dominant from the beginning. When performing a TOF-SIMS analysis, it is only possible to analyze for positive or negative secondary ions, and the choice depends on what kind of information is desired. Sometimes it is useful to perform the analysis twice in order to collect both positive and negative secondary ions (it is not technically possible to collect both at the same time). The device in Fig. 4.6 (Al/C60 /P3CT/ITO) was analyzed for both positive and negative secondary ions. The depth profiles for the negative secondary ions were discussed in a previous paragraph. The depth profiles for the positive secondary ions provide additional information not obtainable for negative secondary ions. Figure 4.8 shows depth profiles for Al+ and In+ that are mass spectral markers for the two aluminum and ITO electrodes. The two markers are detected through the entire device (disregarding the substrate). The explanation is found in the phenomenon called interlayer diffusion (not to be mistaken for interlayer mixing). In this case the electrode material is slowly dissolved into the other layers, that is, the electrode material is diffusing through the entire device, contributing to its overall degradation. This phenomenon is not observed using
Figure 4.8 TOF-SIMS depth profiles of an organic solar cell with the configuration Al/C60 /P3CT/ITO. P3CT corresponds to poly(3-carboxydithiophene). The analysis ions con+ + sisted of 200 fA of 25-keV Bi+ 3 and the sputter ions consisted of 19 nA of 3-keV Xe . Al is a marker for aluminum oxide, and In+ is a marker for ITO.
166
Chapter 4
− negative mass spectral markers AlO− 2 and InO2 . This suggests that it is the nonoxide forms of the electrode material that are diffusing through the remaining part of the device, possibly the cationic forms catalyzed by the electrochemical processes during operation of the organic solar cell. Once again the shapes of the profiles are influenced by the concentration and the chemical surroundings. The latter changes throughout the device, which is pronounced in the aluminum oxide layers where both signals (Al+ and In+ ) are enhanced. Interlayer mixing will, as discussed earlier, have an increasing negative effect on the depth profiles for longer sputter times, resulting in an undesirable lowering of the depth resolution at longer sputter times. There is an additional complication related to interlayer mixing. Some material could (but not necessarily) be continuously pushed further into the sublayers. This phenomenon complicates mapping of interlayer diffusion in the sputter direction. For example, the Al+ depth profile in Fig. 4.8 could be influenced by this phenomenon, and it is not straightforward to determine the magnitude of the contribution. If it was technically possible to perform the depth profile from the other side (sputter→ITO/P3CT/C60 /Al instead of sputter→Al/C60 /P3CT/ITO), it would be possible to get an estimate of the contribution by comparing the two depth profiles. However, it is not possible, based on Fig. 4.8, to conclude for certain that Al+ has diffused through the entire device, but for In+ this conclusion can be made for certain since the diffusion is opposite the sputter direction. The shape, however, of the In+ depth profile could to some degree be affected by interlayer mixing. With respect to interlayer diffusion in the sputter direction, it is necessary to employ complementary analysis to ensure that the observations are consistent with reality. In summary, organic solar cells have perfect dimensions for TOF-SIMS depth profiling analysis. However, some aspects of the analysis are problematic and require consideration when interpreting the data. If an organic solar cell is operated or stored in an atmosphere containing 18 O2 or H18 2 O, it then becomes possible to map the in-depth diffusion of oxygen or water into the device and the subsequent reaction with the constituents of the device by monitoring the 18 O/16 O ratio as a function of sputter time. Interface chemistry can be mapped by monitoring appropriate mass spectral markers as a function of sputter time (i.e., depth); for example, the vertical distribution of AlO− 2 revealed that aluminum oxide is present in an Al/C60 interface. Interlayer diffusion can be mapped by monitoring appropriate mass spectral markers as a function of sputter time; for example, it was shown that the nonoxide form of indium can diffuse through an entire device to the outer side of the counter electrode. Some of the problematic aspects of performing TOF-SIMS depth profiling on organic solar cells are as follows:
• There is a significant depreciation of sensitivity due to the sputter process. • TOF-SIMS is not (directly) a quantitative technique; thus, it is, for example, not possible to determine the degree of oxidation in absolute terms.
Lifetime and Stability Studies
167
• The shapes of the depth profiles are influenced by matrix effects, that is, the effects of the chemical surroundings on the mass spectral response. • Interlayer mixing caused by the sputter process could influence the shapes of the depth profiles and could complicate the mapping of interlayer diffusion in the sputter direction. • Interlayer mixing gradually causes the depth resolution to worsen for longer sputter times. If the conditions are tuned to diminish interlayer mixing, the acquisition time could then become impractical. • The sputter rate is material dependent; thus, if it is not possible to use a depth scale, the sputter time is used instead. • There is not necessarily the same information stored in positive and negative secondary ions; therefore, a choice has to be made depending on what type of information is required. Alternatively, two depth profiles should be obtained. In conclusion, TOF-SIMS depth profiling provides information on the vertical distribution (in-depth) of chemistry and is well suited for studying diffusion phenomena such as water or oxygen diffusion into the device, or interlayer diffusion. Furthermore, interface chemistry can be identified and monitored. This valuable information comes with a price in the form of the associated problematic aspects listed above. Before the rather resourceful TOF-SIMS depth profiling analysis is performed, one should evaluate whether it is at all possible to obtain the desired information: are the problematic aspects obstacles, or can they be bypassed? Finally, some of the problematic aspects cause some conclusions to be questionable. In these cases, it is advisable to employ complementary TOF-SIMS methods (e.g., TOF-SIMS imaging). In doing so, it is necessary to gain access to the various layers and interfaces of the device. 4.2.4 Gaining access to the various layers in the photovoltaic device To gain access to the various layers/interfaces, ideally each layer should be selectively removed. This is not always possible. Another downside is the risk of modifying the exposed surfaces. If, for example, the layers are removed in ambient air, which is often the case, then oxygen instantly reacts with the exposed surface, which has to be taken into account when interpreting the analysis data. Another problematic phenomenon is surface segregation. Once a layer has been removed, it could induce surface segregation of species from at least the bulk layer material. This possible phenomenon should be considered when interpreting the analysis data. Figure 4.9 exemplifies how an organic solar cell device with the composition Al/C60 /C12 -PPV/PEDOT:PSS/ITO is systematically taken partially apart as follows. Removal of Al electrode: The Al electrode is removed from the intact device [Fig. 4.9(A)] by carefully applying a piece of adhesive tape on only part of the active area. The tape is then pulled off very quickly, exposing the Al/C60 interface
168
Chapter 4
Figure 4.9 Schematic showing the various steps involved in getting access to the various layers/interfaces of a device composed of Al/C60 /C12 -PPV/PEDOT:PSS/ITO. (A) The intact organic solar cell device. (B) The outcome of peeling off the Al electrode using adhesive tape. (C) The outcome of having washed off C60 and C12 -PPV using chloroform. (D) The outcome of having washed off PEDOT:PSS using water. C12 corresponds to dodecyl.
[Fig. 4.9(B)]. Oxygen from the ambient air quickly reacts with the exposed C60 and possibly with aluminum of the nonoxide form. Both the exposed device surface as well as the tape surface can now be analyzed. The device, in this case, detaches at the Al/C60 interface.12 Removal of the C60 and C12 -PPV layers: Unfortunately, it is not possible to selectively remove each layer of C60 and C12 -PPV. The layers are simultaneously removed by dipping a cotton swab in chloroform, and gently sweeping it across the exposed surface. After one sweep, the cotton swab is rotated 180 deg (in order to avoid reattachment of material) and another sweep is performed. The cotton swab is then replaced with a clean one and the procedure is repeated ∼30 times. The sweeping should preferably be in the same direction. If the cleaning procedure is inadequate, it shows up in a TOF-SIMS imaging analysis as stripes in the sweeping direction. The exposed PEDOT:PSS surface [Fig. 4.9(C)] can now be analyzed. Removal of the PEDOT:PSS layer: The PEDOT:PSS layer is not soluble in chloroform but it is in water; thus, the chloroform procedure is simply repeated using (ultrapure) water instead, which exposes the ITO surface [Fig. 4.9(D)] for further surface analysis.
Lifetime and Stability Studies
169
Figure 4.10 Photographs of an organic solar cell with the configuration Al/C60 /P3CT/ITO at various steps of layer removal. P3CT corresponds to poly(3-carboxydithiophene). (A) The intact organic solar cell device. (B) The outcome of peeling off the Al electrode. The inset is an enlargement of a sputter hole (∼0.4 × 0.4 mm2 ). (C) The outcome of having peeled off another piece of the Al electrode and having washed off the C60 layer. (D) The outcome of having peeled off yet another piece of the Al electrode and having washed off the P3CT layer.
Figure 4.10 consists of photographs that show how the process of stepwise layer removal looks like in real life for an organic solar cell with the configuration Al/C60 /P3CT/ITO. Figure 4.10(A) shows the intact device before analysis. When performing a TOF-SIMS depth profile through the entire device a sputter hole is created that penetrates the entire device and a minor part of the glass substrate. One of the sputter holes are shown in Fig. 4.10(B) as an inset (the arrow points to the actual area that was enlarged). Once the Al electrode is removed the exposed C60 surface degrades faster. This explains why only part of Al electrode is removed [Fig. 4.10(B)]. The device should be introduced into the vacuum system of the TOF-SIMS instrument as quickly as possible after removing the Al electrode in order to minimize degradation of the exposed C60 by ambient oxygen and light. If available, a transfer vessel with a controlled atmosphere (e.g., argon) can be used to transfer the device from a glove box to the TOF-SIMS instrument in order to avoid oxygen exposure. However, light exposure is impossible to avoid since one has to see the device while handling and analyzing. A complete and thorough TOF-SIMS analysis using all the available modes of TOF-SIMS on a device could take up to week; therefore, in order to minimize degradation during analysis, a new piece of Al electrode is removed after a couple of days (preferably in an oxygen-free atmosphere). Figures 4.10(B)–(D) reveal that in this particular case a fresh piece of Al electrode was removed three times. If the Al electrode is removed using two-sided tape, it becomes possible to mount the removed tape/Al sample on the device outside the active area in such
170
Chapter 4
a way that the exposed Al surface (i.e., the Al/C60 interface) is pointing upward, which enables surface analysis. This is shown in Fig. 4.10(C), where an arrow points to the adhered tape/Al sample that is positioned in such away that it corresponds to the Al electrode having been flipped over. This makes it possible to locate and analyze the same spot/area on the exposed surfaces. On part of the exposed surface in Fig. 4.10(D), the C60 layer was removed, but this is not visible. In this case the P3CT material provides the red/brown color. The C60 was removed on the area below the arrow in Fig. 4.10(D). The arrow in Fig. 4.10(D) indicates the area where the P3CT material was removed, exposing the (transparent) ITO electrode.
4.2.5 TOF-SIMS imaging TOF-SIMS imaging can be performed over very small areas (∼10 × 10 μm2 ) as well as over large areas (∼10 × 10 cm2 ). Small areas are analyzed by scanning the primary ion beam controlled by manipulated fields, across a surface area. This is extremely fast; typical acquisition times are from seconds to minutes for a sufficient amount of accumulated scans. This surface scanning mode is used on area sizes up to ∼0.5 × 0.5 mm2 . For larger areas the sample stage is used to scan across an area. This is, however, a very slow process with acquisition times ranging from many minutes to many hours, and in extreme cases up to a day and a night. Therefore TOF-SIMS imaging is not suited for screening many samples over large areas. Two primary ion-beam modes are typically used, namely, one that produces high image resolution (i.e., narrow primary ion beam diameter) and low mass resolution (i.e., low peak resolution), and one that produces low image resolution and high mass resolution. The type of desired information determines the analysis conditions: primary ion beam diameter, time between primary ion beam pulses (determines the mass range), analysis area, number of image pixels, number of accumulated scans, primary ion identity, primary ion dose, and so on. Examples presented in this section demonstrate the applicability of TOF-SIMS imaging for degradation studies of organic solar cells. A lot of different types of valuable information can be extracted from TOF-SIMS images in spite of the complex nature of the technique. However, just as for TOF-SIMS depth profiling, care should be taken when interpreting the TOF-SIMS imaging data. Figure 4.11 shows TOF-SIMS images of an organic solar cell with the configuration Al/C60 /P3CT/ITO. The analyzed area corresponds to the white dashed square in Fig. 4.10(B), that is, the exposed C60 . It is classified as a large area (10 × 2.8 mm2 ) so sample stage scanning was employed. The total acquisition time was 15 hr, which is clearly impractical for screening purposes and statistical studies. The long acquisition time is a consequence of the combination of large area (10 × 2.8 mm2 ), a large number of pixels in the image (1024 × 287 pixels), and
Lifetime and Stability Studies
171
Figure 4.11 TOF-SIMS images (10 × 2.8 mm2 ) of an organic solar cell with the configuration Al/C60 /P3CT/ITO after the Al electrode was peeled off. P3CT corresponds to poly(3carboxydithiophene). The imaged part of the surface corresponds to the white dashed square in Fig. 4.10(B). The images visualize the normalized lateral intensity distribution of selected ejected negative secondary ions from the surfaces. C− 60 was used as a marker for C60 (A), and S− was used as a marker for P3CT (B). The lower image (C) visualizes the normalized lateral intensity distribution of the 18 O/16 O ratio.
sample stage scanning. These parameters were necessary in order to, among other things, visualize possible macroscopic effects. Figure 4.11(A) shows the distribution of C60 on the exposed surface. Various macroscopic effects are observed. There is a defect in the lower left corner [Fig. 4.11(A)] together with various diagonal line defects. These defects/fractures are believed to be a result of the peeling process; therefore, not interesting from a degradation point of view. There are various small perfectly horizontal lines in the images. These lines are instrument effects caused by momentary loss of primary ion beam current, that is, loss of signal. Another problem with the primary ion beam current is stability over extended periods of time. It is impossible to maintain the same primary ion gun conditions for 15 hr; thus, the outcome is typically a slow, subtle drift in the total secondary ion signal and thus in the individual signal intensities, which introduces an instrument effect in the ion images. This particular instrument effect was removed by normalizing the secondary ion images with the total ion image. The ion images were then further normalized individually by assigning the pixel with the highest intensity to the color white according to the intensity color scale in Fig. 4.11. This is done for practical reasons since there is no information stored in the absolute intensities. The information is stored in the relative intensities.
172
Chapter 4
Figure 4.11(B) visualizes the distribution of P3CT on the sample surface. The P3CT occurrence matches perfectly with the observed defects in Fig. 4.11(A); thus, the peel process exposes the sublayer of P3CT. The observation is not interesting from a degradation point of view, but it is very useful. A mass spectrum can be obtained from these exposed areas and compared with the mass spectrum of the P3CT surface that was exposed by the solvent washing procedure. The outcome is fairly consistent mass spectra (not shown) supporting the assumption that the washing procedure does not significantly modify the exposed P3CT surface. The device in Fig. 4.11 is the device that had undergone a TOF-SIMS depth profiling analysis after having been stored in darkness in an 18 O2 /N2 (20:80) atmosphere for 2240 hr. The vertical (in-depth) oxygen incorporation was described by monitoring the 18 O/16 O ratio against the sputter time (i.e., depth). It is also obvious to describe the lateral (in-plane) distribution of the 18 O/16 O ratio. This is shown in Fig. 4.11(C). The 18 O/16 O ratio appears to be fairly homogeneous in its distribution on a macroscopic scale. However, there seems to be some fine structure on a smaller scale that deserves more attention. The 10-mm horizontal analysis area consists of 1024 pixels; thus, each pixel represents 10 μm. The smallest primary-ion-beam diameter (that normally defines the image resolution) is much smaller than 10 μm. When the image resolution defined by the area size and the number of pixels exceeds the image resolution defined by the largest possible primary-ion-beam diameter (i.e., high mass resolution), then the area size and number of pixels define the image resolution. It is thus possible to conclude that the scale of the observed inhomogeneities must be significantly larger than 10 μm. These inhomogeneities are discussed in greater detail in a subsequent section. The analyzed area shown in Fig. 4.11 that corresponds to the white dashed square in Fig. 4.10(B) was not randomly chosen. One of the purposes was to measure the possible introduction of 18 O2 via the edges. This explains why the analyzed area extends over the edge of the active solar cell area [Fig. 4.10(B)]. It is possible to extract an average of all horizontal line profiles from the image in Fig. 4.11(C), which produces a plot showing the average 18 O/16 O ratio as a function of horizontal distance. Figure 4.12(B) shows this plot and on the scale in question (10-μm image resolution) and it is evident that there is clearly no increased 18 O/16 O ratio close to the edge of the active area. It is thus not possible to detect introduction of 18 O via the edges of the active area of the device, suggesting that this is not an 2 important degradation mechanism, or at least not within the time frame of the 18 O2 exposure (2240 hr). One interesting observation was, however, made in Fig. 4.12. At the edge (at ∼8.5 mm), the Al electrode detached at the C60 /P3CT interface. The 18 O/16 O ratio corresponds approximately to the natural ratio (0.2%) on the surface of P3CT indicated by the red arrow in Fig. 4.12(A). It can be concluded that no (or practically no) 18 O incorporation is observed at the C60 /P3CT interface.
Lifetime and Stability Studies
173
Figure 4.12 The ion image from above [Fig. 4.11(B)] that shows the distribution of P3CT (A). An average of all horizontal line profiles are plotted visualizing the 18 O/16 O ratio as a function of distance from the edge of the active area (B).
An investigation of the peeled off Al electrode on the corresponding edge area did not reveal an increased 18 O/16 O ratio either. The apparent small-scale inhomogeneities in Fig. 4.11(C) can be further investigated by performing the TOF-SIMS imaging analysis on a smaller scale. An area size of 0.5 × 0.5 mm2 was chosen to take advantage of the fast primary-ion-beam rastering capability. Figure 4.13(D) shows the small-scale 18 O/16 O ratio distribution in the Al/C60 interface of the device with the configuration Al/C60 /P3CT/ITO. It is now clear that the 18 O incorporation has happened laterally on circular areas or clusters of circular areas (i.e., overlapping circular areas). The circular areas have diameters up to ∼100 μm and the clusters are up to ∼200 μm in size. The circular oriented incorporation in the lateral plane is a consequence of microscopic holes in the corresponding Al electrode. 18 O2 diffuses vertically through the holes and expands in all lateral directions reacting with at least C60 , after which the 18 O fixates in degradation products, which enables mass spectrometric detection by TOF-SIMS methodologies. An equivalent measurement was performed on the device with the configuration Al/C60 /C12 -PPV/PEDOT:PSS/ITO. The devices are not equivalent, but both contain an Al/C60 interface. The 18 O/16 O ratio distribution pattern for the Al/C60 /C12 PPV/PEDOT:PSS/ITO device [Fig. 4.13(A)] is clearly different from that in Fig. 4.13(D). The difference is manifested in the different experimental conditions during exposure to 18 O2 . The device corresponding to Fig. 4.13(A) was illuminated in an 18 O2 /N2 atmosphere (1 atm) for a relatively short time (45 hr), and the device corresponding to Fig. 4.13(D) was stored in an equivalent atmosphere, but
174
Chapter 4
Figure 4.13 TOF-SIMS images (0.5 × 0.5 mm2 ) of two organic solar cells with the configuration Al/C60 /C12 -PPV/PEDOT:PSS/ITO (A–C) and Al/C60 /P3CT/ITO (D–F) after the Al electrode was peeled off. P3CT corresponds to poly(3-carboxydithiophene) and C12 corresponds to dodecyl. The images visualize the normalized lateral intensity distribution of the 18 O/16 O ratio. The 18 O/16 O ratio values in the lower right corners are averages of 10 measurements each covering 0.1 × 0.1 mm2 from various surface locations.
in darkness for a relatively long time (2240 hr). The distribution patterns suggest that oxygen reacts faster when illuminated,that is, the oxygen reacts in close vicinity of the holes in the lateral plane. The device stored in darkness reacts slower and has time to diffuse further in the lateral plane, resulting in bigger but more blurred circular areas [Fig. 4.13(D)] compared to the smaller, more concentrated circular areas in Fig. 4.13(A). This example illustrates that it is possible to use the 18 O/16 O ratio distribution to get information on possible holes in the Al electrode, and get information on the rate of 18 O2 diffusion. Even though it makes no sense to compare Figs. 4.13(B) and (C) with Figs. 4.13(E) and (F), there is still interesting information to be extracted from each set of images. The circular areas of the 18 O/16 O ratio is also observed in Fig. 4.13(B), revealing that PEDOT:PSS is also susceptible to oxygen-induced degradation. The C60 /P3CT interface shows a homogeneous 18 O/16 O ratio distribution (on the scale in question) and the average 18 O/16 O ratio value (lower right image corner), based on ten different surface locations, reveals a ratio very close to the natural ratio (0.2%). The oxygen diffusion rate is presumably the same in the lateral and vertical directions (for the same material), and considering the fact that the C60 layer is only ∼100-nm thick, it should be safe to assume that 18 O2 has reached the P3CT layer. Based on the measured 18 O/16 O ratio of 0.3 ± 0.03% in the C60 /P3CT interface [Fig. 4.13(B)], it can be concluded that P3CT is not susceptible to oxi-
Lifetime and Stability Studies
175
dation by oxygen when it is incorporated in a photovoltaic device and stored in darkness. This conclusion is in perfect agreement with the conclusion based on Fig. 4.12. Figures 4.13(C) and (F) show homogeneous 18 O/16 O ratio distributions in both ITO interfaces with average 18 O/16 O ratios larger than the natural ratio. It is now possible to conclude that oxygen diffuses through the entire organic solar cell, all the way to the counter electrode, where part of the oxygen (16 O) in ITO is exchanged with 18 O. Considering the high concentration of natural oxygen (i.e., 16 O) in ITO, the measured 18 O/16 O ratios of 0.43 ± 0.02% and 0.45 ± 0.04% must correspond to significant 18 O incorporation. The lack of apparent circular incorporation in the ITO interfaces suggests homogenous incorporation of 18 O, which again suggests an alternative entrance channel for oxygen. A possible alternative and less dominating entrance channel for oxygen could be diffusion through the Al grains in the Al electrode. However, there is nothing in the literature to confirm this. The macroscopic images in Fig. 4.11 revealed that during the peel process some of the P3CT was accidentally exposed. Figure 4.14 is an optical image (0.77 × 0.77 mm2 ) of an area where this phenomenon is especially pronounced. The red-dashed line represents the area (0.5×0.5 mm2 ) that was subjected to TOFSIMS imaging. Normalization procedures can sometimes complicate the comparison of different images. By performing the TOF-SIMS imaging analysis over an
Figure 4.14 Optical image (0.77 × 0.77 mm2 ) of an organic solar cell with the configuration Al/C60 /P3CT/ITO after the Al electrode was peeled off. P3CT corresponds to poly(3carboxydithiophene). The area indicated by the red dashed square (0.5 × 0.5 mm2 ) was subjected to TOF-SIMS imaging. This particular area was chosen because it shows part of the remaining Al electrode, the C60 surface, and the P3CT surface.
176
Chapter 4
area where several surfaces/interfaces are exposed, these problems are minimized or absent. Furthermore, the possible effects of the solvent washing procedures described in a previous section are absent. The only surface/interface not exposed in this particular example is the ITO surface/interface. Selected ion images from a TOF-SIMS imaging analysis of the area corresponding to the red-dashed square in Fig. 4.14 are presented in Fig. 4.15. The ion images in Fig. 4.15 are very informative and support earlier conclusions made for this system based on a TOF-SIMS depth profiling analysis. It is quite easy to distinguish the various surfaces/interfaces from the intensity distribution patterns. The area indicated by 1 in Fig. 4.15(B) corresponds to the Al surface, 2 corresponds to the exposed C60 surface, and 3 corresponds to the exposed P3CT surface. Figure 4.15(E) shows the distribution of natural oxygen species to be present everywhere on the analyzed surface. This is expected since oxygen is a natural part of aluminum oxide and P3CT. Exposed C60 reacts with ambient air after peeling off the Al electrode; thus, “natural” oxygen species in the form of degraded products are expected on the exposed C60 surface. The atmosphere is in direct contact with the outer aluminum oxide surface, so it is not surprising that a significant 18 O incorporation/exchange is observed at the outer part of the
Figure 4.15 TOF-SIMS images (0.5 × 0.5 mm2 ) of an organic solar cell with the configuration Al/C60 /P3CT/ITO after the Al electrode was peeled off. P3CT corresponds to poly(3carboxydithiophene). The imaged part of the surface corresponds to the red-dashed square in Fig. 4.14. The images visualize the normalized lateral intensity distribution of selected ejected negative or positive secondary ions from the surfaces. Al+ was used as a marker − was used as a marker for for aluminum (A), C− 60 was used as a marker for C60 (B), S + 16 − P3CT (C), In was used as a marker for indium (D), O was used as a marker for natural oxygen (E), and 16 O− and 18 O− were used as a markers for the 18 O/16 O ratio (F).
Lifetime and Stability Studies
177
Al electrode [Fig. 4.15(F)]. Figures 4.15(B) and (C) have a high degree of contrast, that is, C60 is almost only observed where C60 is exposed and P3CT is only observed where P3CT is exposed. The opposite behavior is observed for the aluminum [Fig. 4.15(A)] and indium [Fig. 4.15(D)] ion images, where only a weak contrast is evident, that is, aluminum and indium are observed everywhere on the analyzed surface. This is related to the phenomenon called interlayer diffusion. Aluminum and indium (presumably the cations) are small and mobile; therefore, these species (and possibly other mobile species) can diffuse through the various layers of the device during operation and/or storing, which explains why the electrode materials are observed everywhere on the analyzed surface. The TOF-SIMS images support the conclusion based on the TOF-SIMS depth profiling analysis: that both electrode materials diffuse through the remaining parts of the device to the counter electrode. Indium is even found on the outer Al electrode surface. The device in Fig. 4.15 (Al/C60 /P3CT/ITO) was subjected to TOF-SIMS imaging on all surfaces/interfaces (1–4 in Fig. 4.16). The average relative signal intensities of the electrode materials (Al+ /In+ and/or In+ /Al+ ) can, for each surface/interface, be extracted from the corresponding ion images. The outcome is shown in Fig. 4.17. If it is assumed that the two electrode materials (aluminum and indium) experience the same matrix effects when analyzed, then the expected trend would be a systematic decrease in the Al+ /In+ ratio when going from the Al electrode to the ITO electrode (1→ 4 in Fig. 4.17). According to Fig. 4.17, this is true except for one inconsistency: the Al+ /In+ ratio is too low (or the Al+ /In+ ratio is too high) for 2 in Fig. 4.17 (Al/C60 interface). Judging from Figs. 4.15(A) and (D) it seems to be the In+ intensity that is especially enhanced on the C60 surface. A possible explanation for this phenomenon could be surface segregation of indium as a consequence of the Al electrode peeling. If it is more energetically favored for the indium to be interfaced with air or vacuum, then the outcome would be an accumulation of indium on the C60 surface after the peeling procedure. This could possibly also explain why indium is observed on the outer Al electrode.
Figure 4.16 Schematic cross section of an organic solar cell composed of Al/C60 /P3CT/ITO with part of each layer successively removed, exposing the various surfaces/interfaces (1–4).
178
Chapter 4
Figure 4.17 The diagram shows average Al+ /In+ and In+ /Al+ intensity ratios for surfaces 1–4 schematically presented in Fig. 4.16. The measured intensity ratios are extracted from an average of 10 TOF-SIMS imaging analyses on different surface locations (100 × 100 μm2 areas).
In conclusion, in order to perform a TOF-SIMS imaging analysis, it is a necessity to take the device apart, that is, expose the interfaces. TOF-SIMS imaging provides information on the lateral distribution (in-plane) of chemistry on surfaces and exposed interfaces of organic solar cells. Various scales can be analyzed, from the low micrometer range to several centimeters with an image resolution ranging from several micrometers to ∼100 nm. TOF-SIMS imaging provides supportive information to TOF-SIMS depth profiling analysis; for example, conclusions on diffusion phenomena can be supported. Furthermore, TOF-SIMS imaging can also be used to identify and monitor interface chemistry. However, caution is advised when interpreting TOF-SIMS images. When analyzing organic solar cells, it is important to distinguish between instrument effects, effects of the peel process, and actual degradation phenomena. 4.2.6 Chemical structure elucidation based on mass spectral information In the previous sections, it was demonstrated how applicable TOF-SIMS depth profiling and imaging are for degradation studies of organic solar cells. In this section, it will be demonstrated how basic mass spectral information can be used to identify unknown substances that are presumably related to degradation phenomena in organic solar cells. Practically in this area is described in the literature on particle formation in organic solar cells, so uncertain what the effect of particle formation is on solar cell performance. Norrman et al.11 have studied particle formation in an organic solar cell with the configuration Al/C60 /C12 -PPV/PEDOT:PSS/ITO. The authors needed a partial organic solar cell for reference purposes; thus, device fabrication was stopped after applying the C12 -PPV layer, resulting in a
Lifetime and Stability Studies
179
Figure 4.18 Schematic cross section of a partial organic solar cell composed of C12 PPV/PEDOT:PSS/ITO, demonstrating particle formation on the surface of C12 -PPV. C12 corresponds to dodecyl.
Figure 4.19 TOF-SIMS ion images (0.5×0.5 and 0.05×0.05 mm2 ) of a partial organic solar cell composed of C12 -PPV/PEDOT:PSS/ITO. The right image (B) is a small-scale image of the area indicated by a white dashed square in the left image (A). The images visualize the normalized lateral intensity distribution of ejected negative secondary ions from the surfaces at m/z 137. C12 corresponds to dodecyl. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
partial organic solar cell with the configuration C12 -PPV/PEDOT:PSS/ITO. The partial device was then stored in darkness in ambient air after having served its purpose. After a few months, the surface topography of the device was mapped (not shown) in order to determine the density of microscopic holes. The topography analysis revealed that the surface was covered by microscopic particles ∼25 μm in size in the lateral plane with a height of ∼12 nm. The first step in attempting to identify the mechanism of the observed particle formation is identification of the chemical composition of the particles. The most practical approach is to obtain a mass spectrum of the particles and attempt to extract possible information on the molecular structure. First of all a TOF-SIMS imaging analysis was performed. An image for each mass spectral peak was obtained (not shown) in order to search for mass spectral markers that would provide a contrast between the particles and the surrounding material in the image. Figure 4.19 shows the intensity distribution for one such mass spectral marker. The peak at m/z 137 originates only from the particle area and can be used to define
180
Chapter 4
the area from which the mass spectra should be extracted. Figure 4.19(A) is the outcome of a TOF-SIMS imaging analysis using low image resolution and a high mass (i.e., peak) resolution. Figure 4.19(B) is an analysis of the one particle indicated by the dashed-white square in Fig. 4.19(A) using high image resolution and low mass resolution. The TOF-SIMS imaging data corresponding to Fig. 4.19(A) (i.e., high mass resolution) were used to extract appropriate mass spectral information. The mass spectra in Fig. 4.20 were extracted from the particle area [Fig. 4.20(B)] and from the surrounding area [Fig. 4.20(A)]. The area surrounding the particles consists of peaks originating from degraded C12 -PPV (oxygen and/or water reaction products). It is unavoidable for some degraded C12 -PPV to end up on the particles. Furthermore, the particles produce at least some mass spectral fragments that are also produced from the surrounding area. In order to identify peaks that are unique to the particles, a careful comparison was made between the two mass spectra in Fig. 4.20. Those peaks that are unique to the particles were labeled with a number (1–7) and the corresponding element composition was identified based on a combination of exact mass determination and the isotope distribution.
Figure 4.20 TOF-SIMS mass spectra of the nonparticle area (A) and of the particle area (B) extracted from the TOF-SIMS imaging analysis described in Fig. 4.19(A). The peaks indicated by numbers are peaks that are unique for the particle area, that is, not found in the surrounding material. The element compositions are based on a combination of exact mass determination and isotope distribution. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
Lifetime and Stability Studies
181
As is evident from Fig. 4.20(B), seven peaks were found unique to the particles, and six out of seven peaks correspond to an element composition containing sulfur. The sublayer of PEDOT:PSS (Fig. 4.18) is the only material in the device that contains sulfur. Furthermore, the distribution of the seven peaks in Fig. 4.20(B) resembles (i.e., they are not identical) the mass spectrum of PSS (not shown). The mass spectral observations suggest that the particles could be a PSS derivative. The next step is to put the pieces of the puzzle together in the correct fashion. If it indeed is a PSS derivative, then one would intuitively expect the sulfonic acid group (being the only functional group in PSS) to be involved. It then becomes fairly straightforward to add a part to the sulfonic group that will consequently explain the seven unique mass spectral peaks shown in Fig. 4.20(B). The molecular structure shown in Fig. 4.21 explains all of the seven unique mass spectral peaks. Now that the molecular structure of the particles is proposed, the next step is to attempt to propose a mechanism for the particle formation. At first it seems tempting to suggest that hydroxy benzoic acid has reacted with PSS. However, this theory suffers from the fact that hydroxy benzoic acid is not a degraded product of C12 -PPV, which is manifested in the lack of mass-to-change ratio (m/z) 137 [5 in Fig. 4.20(B)] in the area surrounding the particles (Fig. 4.19). Norrman et al.11 instead proposed the mechanism shown in Fig. 4.22 where PSS undergoes a so-called oxido-de-sulfonato substitution, forming the phenolate. This step is usually performed under extremely basic conditions, which is unlikely in this case. There could easily be alternative experimental conditions producing the same outcome, possibly by alternative mechanisms. The phenolate can then react with PSS forming two PSS chains linked together via a sulfonic ester group. The carbon in-
Figure 4.21 Proposed molecular structure of the particles shown in Fig. 4.19, and proposed origins of the observed mass spectral fragments produced during the ionization process. The numbers in brackets correspond to the numbers in Fig. 4.20(B). (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
182
Chapter 4
dicated by an asterisk in Fig. 4.22 is the carbon most susceptible for oxidation and will consequently form the proposed molecular structure of the particles observed in Fig. 4.19. The complex nature of an organic solar cell device is manifested in the considerable amount of layers and interfaces, and the many more or less interrelated degradation mechanisms that take place more or less at the same time. Another approach to study degradation mechanisms in organic solar cells is to simplify the situation. One such way is to focus on, for example, the active material by performing experiments directly on the material, that is, while it is not incorporated in a device. Norrman et al.11 performed such an experiment where the active material was specially designed to facilitate a TOF-SIMS mass spectrometry analysis. The compound (active material) is shown in Fig. 4.23. The compound in Fig. 4.23 is a trimer of the PPV type with two different end groups that introduce asymmetry in the molecule, which is an advantage when interpreting the corresponding mass spectral data. The molecule is sufficiently small
Figure 4.22 Proposed mechanism for the formation of the particles shown in Fig. 4.19. The asterisk indicates the carbon most susceptible to oxidation. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
Figure 4.23 Trimer of the PPV type with two different end groups that were chosen to facilitate a TOF-SIMS mass spectrometry analysis. The systematic name is {4-[2-(4-{2-[4-[2-(4-{2-[4-[2-(4-{2-[4-[2-(4-chloro-phenyl)-vinyl]-2,5-bis-(2-ethyl-hexyloxy)phenyl]-vinyl}-phenyl)-vinyl]-2,5-bis-(2-ethyl-hexyloxy)-phenyl]-vinyl}-phenyl)-vinyl]-2,5-bis(2-ethyl-hexyloxy)-phenyl]-vinyl}-phenyl)-vinyl]-phenyl}-dimethylamine. EH corresponds to 2-ethylhexyl. For a more detailed description, see Refs. [9] and [10].
Lifetime and Stability Studies
183
(mol wt = 1639.87 g mol−1 ) to desorb from the surface during ionization, and the molecular ion is sufficiently stable for it to be detectable in a TOF-SIMS mass spectrometry analysis. The trimer was spin coated on a glass substrate and subsequently exposed to simulated sunlight (1000 W m−2 , AM 1.5) for 20 hr in ambient air. These conditions are more than enough to degrade all the trimer molecules corresponding to at least the probe depth of TOF-SIMS, which is ∼1 nm when sputtering is not considered. The goal is now to attempt to identify the degradation products in order to gain information on the actual chemical degradation mechanisms. This is possible by identifying the element compositions of the observed fragment ions. The element compositions are, as mentioned previously, determined from a combination of exact mass determination and isotope distribution of the mass spectral peaks. Figure 4.24 shows a TOF-SIMS mass spectrum in the mass range between m/z 100 and m/z 180 of the sample surface after illumination. The trick is now to disregard generic peaks and focus on the remaining dominant peaks. The peaks indicated by number in Fig. 4.24 were identified and listed. The inset corresponds to a narrow window around m/z 60 and is included because it constitutes an interesting observation (discussed later). The next step is not as straightforward. Based on knowledge of gas-phase ion chemistry, analytical mass spectrometry, and chemical intuition, one has to propose from where in the intact trimer molecule, the fragment ion, could have originated. Figure 4.25 shows, among other things, from where in the trimer molecule the peaks 5–9 (Fig. 4.24) presumably originate from. Few of the assignments are straightforward, for example, C7 H4 ClO+ (peak labeled 7 in Fig. 4.24) must (because of the chlorine atom) originate from the (left) end group shown in Fig. 4.25. The four hydrogen atoms suggest that it is an intact
Figure 4.24 Part of a TOF-SIMS mass spectrum of the trimer from Fig. 4.23 after it was spin coated on a glass substrate and exposed to simulated sunlight (1000 W m−2 , AM 1.5) for 20 hr. The element compositions are based on a combination of exact mass determination and isotope distribution. For a more detailed description, see Refs. [9] and [10].
184
Chapter 4
benzene ring. Intuitively, the oxygen atom is probably situated on the remaining carbon. The interpretation suggests that the vinylene group has reacted with oxygen. Carbon double bonds are known to be susceptible for oxidation by oxygen. Other assignments are less specific, for example, the peaks labeled 5, 6, and 8 could, because it is a trimer, originate from several places in the molecule. The fragment ion labeled 9 is a specific fragment ion that can only (because of the nitrogen atom) originate from the (right) end group shown in Fig. 4.25. The fragment ion labeled 9 in Fig. 4.25 is one of the three peaks at m/z 60 in the mass spectrum (Fig. 4.24). Before illumination, there were no peaks at all at m/z 60; thus, the three peaks must originate from degradation products. It is also certain that they are the result of three different degradation mechanisms. Peak 9 (C2 H6 NO+ ) originates from oxygen reacting with the nitrogen; peak 10 (C3 H10 N+ ) must be the outcome of benzene ring degradation, and peak 11 (C2 H4 O+ 2 ) must originate from a part of the trimer molecule where oxygen has reacted either with the benzene ring or with the side chains. The fact that practically all the dominant peaks shown in Fig. 4.24 correspond to fragment ions that are the outcome of oxygen having reacted with the vinylene groups suggests that this is an important degradation mechanism. In conclusion, basic mass spectral data can be used to extract information on element compositions of fragment ions by determining the exact mass correlated with the isotope distribution. Knowledge of the element compositions of the fragment ions enables conclusions to be made regarding the molecular structure of the fragment ions. The molecular structure of the fragment ions provides information on the identity of unknown species (e.g., particles) and thus of the origin of these species. Furthermore, the molecular structure of the fragment ions provides information on degrading mechanisms, for example, preferred reaction sites with respect to oxygen.
Figure 4.25 Drawing showing the trimer molecule from Fig. 4.23, proposed fragment ions, and an indication of the possible origin of the proposed fragment ions. It should be emphasized that the fragment ions do not necessarily have the indicated molecular structures in the mass spectrometer. For a more detailed description, see Refs. [9] and [10].
Lifetime and Stability Studies
185
4.2.7 Monitoring photo-oxidation in time—mapping the “history” of degradation The complexity of studying degradation of organic solar cells is manifested in the many different mechanisms that are in play, some of which some are interrelated and some are not; some are affected by externally applied conditions and some are not. A fact that has not been discussed in the previous sections is that not all degradation mechanisms occur at the same time (or at the same place). The introduction of time as a parameter in study of degradation mechanisms provides additional, more detailed information, making it possible to map the “history” of degradation. The experiment involving the compound shown in Fig. 4.23 was extended to include time as a parameter. The fragment ions shown in Fig. 4.25 were monitored as a function of time. This was done by spin coating the compound (Fig. 4.23) on a glass slide and subsequently obtaining a TOF-SIMS mass spectrum. The glass slide was then exposed to simulated sunlight (1000 W m−2 , AM 1.5) for a given amount of time in ambient air. The glass slide was then, once again, transferred to the TOF-SIMS instrument where another mass spectrum was acquired. This procedure was continued with the same glass slide until only negligible differences were observed in the mass spectra. Figure 4.26 shows the resulting temporal plots of selected fragment ion intensities. The peak intensity at m/z 1638 represents the intact molecule, that is, the peak intensity is a semiquantitative measure of the population of intact molecules (disregarding possible matrix effects). Once the intact molecule is involved in any form of degradation such as dissociation and/or addition of oxygen, it will no longer
Figure 4.26 TOF-SIMS normalized secondary ion intensities versus illumination time for the molecular ion (1) and selected fragment ions from Fig. 4.25 (2, 3, and 7). The sum of all isotopes was used for the molecular ion. The points in parentheses are artifacts caused by mass spectral peak overlap. The small increase at m/z 1638 from 0 s to 100 s is real and believed to be a subtle matrix effect caused by the increasing oxygen incorporation in the surrounding material. For a more detailed description, see Refs. [9] and [10].
186
Chapter 4
produce a peak at m/z 1638 in the mass spectrum. According to Fig. 4.26, the population of intact molecules starts to decrease drastically from ∼100 s and at ∼500 s there are practically no intact molecules left on the surface. One can only speculate on the reason for this sudden decrease; perhaps oxygen incorporation facilitates further degradation when a certain level of oxygen incorporation is reached. The peak at m/z 1521 (2 in Fig. 4.25) represents the result of oxygen having reacted with a specific double bond in an intact molecule. The population of this specific degradation product starts naturally to increase as the population of intact molecules starts to decrease. This degradation product is still a large molecule, and the larger the molecule the higher the probability is for further degradation; thus, the population is only present in a relatively narrow time window. The degradation product represented by the peak intensity at m/z 1521 originates from the intact molecule and thus must have a maximum population within the time window for the intact molecule. The peak at m/z 959 (3 in Fig. 4.25) represents the result of oxygen having reacted with, once again, a specific double bond. However, in this case the degradation product does not necessarily originate from the intact molecule; it can also originate from other degradation products that are the result of the right part (Fig. 4.25) of the molecule having degraded; for example, the degradation product represented by the peak at m/z 1521. The time window for the peak intensity at m/z 959 should be wider than the time window for the peak intensity at m/z 1521. However, this is difficult to evaluate due to the abnormalities in the m/z 1521 intensity plot (the enclosed points in Fig. 4.26). The maximum population for the degradation product represented by the peak at m/z 959 should be at least within the time window for the peak intensity at m/z 1521. The specific reaction product that is the result of oxygen that reacted with the first vinylene group from the left (Fig. 4.25) is represented by the peak intensity at m/z 139. In line with the previous discussion, this peak intensity should have a very broad time window and a maximum population at a relatively late time, which is also the case (Fig. 4.26). The peak intensity plots in Fig. 4.27 represent different situations compared to the plots in Fig. 4.26. The peak intensity at m/z 60 (9 in Fig. 4.25) represents the results of oxygen that reacted with the nitrogen (i.e., not a double bond) in the intact molecule or in a degraded form of the molecule. Thus far, there has been no mention of the alkyl chains (“EH” in Fig. 4.25). This is because the alkyl chains produce hydrocarbon ions during the ionization process of the analysis. Hydrocarbon ions are generic (i.e., formed from anything containing carbon); thus, the degradation behavior of the “EH” chains is not directly available in this experiment. However, it is possible to obtain some indirect information. The peak intensity at m/z 491 (4 in Fig. 4.25) contains information on the degradation history of the alkyl chains. As expected, the peak intensity plot has a broader time window and a maximum population later than the peak intensity
Lifetime and Stability Studies
187
Figure 4.27 TOF-SIMS normalized secondary ion intensities versus illumination time for selected fragment ions from Fig. 4.25 (4, 5, and 9). For a more detailed description, see Refs. [9] and [10].
plot for m/z 959 (Fig. 4.26). The population of the degradation product represented by m/z 491 disappears somewhere between 104 s and 105 s. The fragment ion in question (4 in Fig. 4.25) contains (presumably) intact alkyl chains; thus, this is indirect evidence of the survival of at least some of the alkyl chains for ∼104 s (167 minutes). Fragment ion 5 in Fig. 4.25 represents the result of a high degree of degradation, that is, both alkyl chains are gone and both vinylene groups have reacted with oxygen. Considering the size of the fragment ion, the peak intensity plot is expected to have a time window width and a maximum population time in between that of the peak intensities of m/z 491 and m/z 60, which is, according to Fig. 4.27, also the case. In conclusion, monitoring variations in mass spectra over time provides detailed information on the degradation behavior on a molecular level. In some cases, a specific degradation mechanism can be monitored; for example, photooxidation of a specific double bond in a molecule. Comparing the temporal development for different fragmented ion-peak intensities provides information on the relative time window for the occurrence of different degradation products, that is, the degradation history. However, this procedure suffers from the fact that it relates to the active material from an organic solar cell before it is incorporated into a device. It is not possible in practice to perform the procedure on an intact device, which would otherwise, among other things, provide information on the effect of the various barrier layers and interfaces.
4.3 Studies of Degradation Mechanisms Using XPS X-ray photoelectron spectroscopy16 (XPS) dates back to the mid-1960s, when K. Siegbahn and his research group worked on the technique. The 1981 Nobel Prize in physics was awarded to K. Siegbahn for his development of XPS as a tech-
Chapter 4
188
nique. The technique is based on the photoelectric effect, which has roots back to Einstein who, in 1905, first described and used it to explain the ejection of electrons from a surface that is illuminated. For XPS, Al K alpha (1486.6 eV) or Mg K alpha (1253.6 eV) are the photon energies that are employed most often. Other x-ray lines can also be chosen such as Ti K alpha (2040 eV) or a synchrotron source. The XPS technique is highly surface specific. This is due to the short range the excited photoelectrons can travel within the solid. The energies of the photoelectrons ejected from the sample are measured using a concentric hemispherical analyzer (CHA), and this gives a spectrum with a series of photoelectron peaks. The binding energies of the peaks are characteristic of each element. The area of the peaks can be used (with appropriate sensitivity factors) to determine the composition of the surface of any material. The technique gives a quantitative analysis of the chemical composition of the surface and is sometimes also called electron spectroscopy for chemical analysis (ESCA). The shape of each peak and the binding energy, depending on the chemical state of the emitting atom and XPS, also provide information on the chemical bonding. XPS is not sensitive to the light elements hydrogen or helium, but can detect all other elements. XPS must be carried out in ultrahigh vacuum (UHV) conditions. When used in conjunction with argon-ion gun sputtering of the surface that is analyzed, it is possible to obtain a depth profile of the material. 4.3.1 The principle of XPS∗ In the XPS experiment, the surface of the sample is illuminated by a monochromatic x-ray beam. The x-ray photons that are absorbed by the material near the surface (with an energy hν) is partly used to excite core levels electrons such that hν = BE+KE, where BE is the binding energy and KE is the kinetic energy of the excited electron. BE is characteristic of a given electron level for a given element. With hν fixed and KE measured, it is possible to determine BE and thus the type of excited atom (Fig. 4.28).
Figure 4.28 XPS principle. ∗
[http://www.chem.qmul.ac.uk/surfaces/scc/scat5_3.htm].
Lifetime and Stability Studies
189
4.3.2 Chemical shifts The exact binding energy of an electron depends not only on the level from which photoemission is occurring, but also on the following: 1. The formal oxidation state of the atom. 2. The local chemical and physical environment. Such shifts are readily observable and interpretable in XPS spectra (unlike in Auger spectra) because the technique (1) is of high intrinsic resolution (as core levels are discrete and generally of a well-defined energy) and (2) is a one-electron process (thus simplifying the interpretation). Atoms of a higher positive oxidation state exhibit a higher binding energy due to the extra coulombic interaction between the photoemitted electron and the ion core. This ability to discriminate between different oxidation states and chemical environments is one of the major strengths of the XPS technique. 4.3.3 Angle-dependent studies The degree of surface sensitivity of an electron-based technique such as XPS may be varied by collecting photoelectrons emitted at different emission angles to the surface plane. This approach may be used to perform nondestructive analysis of the variation of surface composition with depth (with chemical state specificity). 4.3.4 Experimental details The basic requirements for a photoemission experiment (XPS or UPS) are as follows: 1. A source of fixed-energy radiation [an x-ray source for XPS or, typically, a Hedischarge lamp for ultraviolet photoelectron spectroscopy (UPS)]. 2. An electron energy analyzer (which can disperse the emitted electrons according to their kinetic energy and thereby measure the flux of emitted electrons of a particular energy). 3. A high-vacuum environment (to enable the emitted photoelectrons to be analyzed without interference from gas-phase collisions). If studies of photo- and thermodegradation of the organic active layer, which can play an important role in the loss of the optoelectronic properties of the devices, are crucial in order to improve the stability of organic solar cells.17–19 Another essential point to understand is the chemical and morphological evolution during aging of the electrodes and principally at the electrode-organic active layer interfaces. It has previously been shown that the initial degradation of OPVs strongly depends on the material used as the cathode.20 In most of the studied organic solar cells, a thin lithium-fluoride (LiF) layer adjunction between metal cathode (generally aluminum) and organic active layer has demonstrated an overall improvement of electrical characteristics.21 It is well known that insertion of a thin barrier layer prevents interaction between the deposited cathode and active layer. It may be
190
Chapter 4
an explanation of the improvement of the devices compared to an Al-only cathode, for example, but is not an exhaustive explanation. During the past 10 years of development for OLEDs, the insertion of a LiF layer between a stable metal such as Al and the organic emitting layer has provided better efficiencies, particularly with a decrease in the operating voltage.22–24 Using UV photoelectron spectroscopy, this improvement was ascribed to an enhanced electron injection at the cathode by a reduction of the barrier height between the LUMO of the organic layer and the work function (Φelectrode ) of the aluminum cathode and band bending phenomena in the LiF layer.25,26 This reduction improves as the built-in potential (VBI = Φanode − Φcathode ) of the device and can be directly related to the decrease of Φcathode . The variation of the work function for aluminum is found to be dependent on the thickness of LiF.27 Two explanations can be found in the literature that describe how Φcathode decreases, as follows: 1. Dissociation of alkali fluoride. This is still controversial but observed in the case of small molecules such as tris(8-hydroxyquinolinato)aluminum (Alq3 ) and for polymers such as MEH-PPV, giving rise of a low work function contact and/or doping phenomena in the organic layer with free Li.28,29 2. Generation of a dipole in the LiF layer that induces a shift between the vacuum levels of the cathode and organic compound that lowers the Φcathode .26 Because of the hygroscopic nature of LiF, deposition conditions seem to be important in interfacial mechanisms that occur. On one hand, H2 O molecules promote reaction between aluminum and LiF,30 and on the other hand, chemisorption of a few monolayers of H2 O result in a strong interface dipole between aluminum and LiF.26 Experimentally, the LiF layer does not behave as a good insulator31 and has a nonuniform coverage at low thickness32 that can diminish the voltage drop at the interface. Studies at Linköping University regarding interfaces between Al, Li, or Al/Li cathodes and PFO [poly(9,9-dioctyl-fluorene)],33 and studies by Gennip and coworkers with poly(p-phenylene vinylene) and PCBM,34 are particularly interesting. In the case of PFO, neither appearance of new gap states nor dissociation of LiF takes place for Al/LiF/PFO devices. No indication of interaction at the interface Al/LiF/MDMO (or PCBM) is observed. Formation of AlF3 as well as Li doping was eliminated as a possibility by SIMS studies. Evidence of the formation of a surface dipole at the LiF/Al interface and the observation that the LiF layer protects PFO and PCBM from interaction with Al (no interaction between MDMO-PPV and aluminum is observed) has been found. A thin layer of LiF does not seem to interact with the cathode or the organic layer. In this section, we describe results obtained by XPS studies of the cathodeactive layer interface for a typical Al/LiF/MDMO-PPV:PCBM/PEDOT:PSS/ITO/ glass solar cell in its freshly prepared form and after aging (for tens of hours) under illumination in an oxidative atmosphere. Several possible physicochemical changes
Lifetime and Stability Studies
191
were revealed concerning the (not well-defined) action of the LiF layer that could be responsible for the rapid degradation of the unencapsulated devices. To study the chemical and morphological changes that take place at the cathode– active layer interface, we have decided to look into three types of samples. To demonstrate alterations that occurred at the interface under aging, we have compared a complete solar cell kept in the dark in ambient atmosphere (labeled “unaged cell”). One cell was aged during ten hours under illumination in an oxidative atmosphere (labeled “aged cell”). Keeping in mind that we use ionic sputtering to probe the interface in this device, we have also directly characterized reference devices without cathode, with a 1-nm LiF cathode, and 2–10-nm Al on 1 nm of LiF. By comparison of the reference devices with the unaged solar cell, we were able to distinguish “real” aging effects from the effects created during etching using argon-ion sputtering. Argon-ion implantation and collision with the constituents of the the exposed volume of the cathode could lead to movement and rearrangement of the target components.35 4.3.5 Device aging and IV measurement The aging chamber was under a controlled oxidative atmosphere. During aging, the current-voltage characteristics were recorded with a Keithley SMU 2400 unit. The devices were illuminated with a halogen photo-optic lamp (Xenophot) with a light intensity at the sample position of 63 mW cm−2 illumination. The temperature in the aging chamber was measured using a thermocouple (Pt100) and was found to be constant around 30◦ C during aging. Before introduction of the oxidative atmosphere in the test chamber, the IV characteristics of the cell were recorded (Fig. 4.29). After the first hours of aging, we observed a quick decrease of the opencircuit voltage from 0.75 to 0.70 V and a decrease in the fill factor (FF) from 0.52 to 0.45. The decrease of the FF is mainly linked to the increase of the series resistance
Figure 4.29 Evolution of IV characteristic during aging of an Al(50 nm)/LiF(1 nm)/MDMOPPV:PCBM(70 nm)/PEDOT:PSS/ITO/glass solar cell.
192
Chapter 4
Rs of the cell. The shunt resistance (Rsh ) stays almost constant even after 50 hr of operation.36–38 The short-circuit current density starts to decrease significantly after 10 hr of illumination. After 50 hr of aging under the experimental conditions, the power conversion efficiency drops to about 30%. The evolution of the IV characteristics of the solar cells during the first hours of operation is similar to the IV characteristics of ITO/PEDOT/MDMO-PPV:PCBM/ Al with or without an LiF layer.32, 40 These macroscopic observations strengthen the idea that chemical and/or morphological changes occur around the Al/LiF activelayer interface and that the proposed beneficial properties of the LiF layer can be questioned. 4.3.6 XPS overall observations Characterization of the interface between the active layer and the cathode was performed in an ultrahigh vacuum setup, where manipulation of the sample between two chambers was possible at a pressure of around 7 × 10−9 mbar as follows: 1. The first chamber was an etching station comprising an AG21 argon-ion gun (VG Microtech). One of the manipulation difficulties was to probe the interface buried under the cathode. Sputtering with Ar ions allows for etching through the aluminum cathode. To avoid too many target degradations during ion etching, a low energy was applied. However, SRIM03 simulation, using Monte Carlo calculations to make detailed calculations of the energy transferred to every target atom by collision, revealed that correct energies for sputtering/etching, that is, sputtering yield = (number of sputtered atoms)/(number of incident ions) larger than 4, were above 2 keV. In the experiments reported here, an energy of 3 keV was used for Ar-ion sputtering with currents of a few microamperes. Experiments with 2 keV Ar ions showed identical results in depth profiles. 2. The second chamber has the XPS source and analyzer. Photoemission studies were performed with a vacuum generator Escalab 210 spectrometer, using the monochromatized Al-Kα line at 1486.6 eV. A fixed analyzer that passes energy of 20 eV was used for element core level scans and 50 eV for survey scans. The photoelectron takeoff angle was 90 deg with respect to the sample plane, which provides an integrated sampling depth of approximately 10 nm. The energy scale of the instrument was calibrated by setting Au 4f7/2 = 84.00 eV [with full width at half maximum (FWHM) at 0.95 eV], Ag3d5/2 = 368.70 eV.39 The core level signals from the elements were decomposed using mixed symmetrical Gaussian-Lorentzian curves. For all the analysis, the Gaussian character varied between 50% and 90% and the FWHM between 0.8 and 1.8 eV, except for the mentioned elements. No energy compensation was applied. The fitting procedure used the standard vacuum generator instrument software, which solves the matrix of derivatives of the peak parameters with a Gaussian-Newtonian method to locate a minimum in the residual root-mean square between the experimental
Lifetime and Stability Studies
193
data and the fitting curve. The best fit obtained by varying the number of peaks and their location on the energy scale was retained. Going back and forth between sputtering station and XPS, we were able to trace a depth profile of the different elements found in the device. Depth profiles of the aged and nonaged diodes permitted us to follow elements detected from the surface of the aluminum cathode to the active layer (Fig. 4.30). First, in Fig. 4.30(a), we observe mainly Al and O in the Al volume. Then, when the metal-organic interface is reached (Fig. 4.30(b)], Al peaks start to decrease and F, Li, and C increase until the organic layer. When Al, F, and Li disappear, only C, O, and S elements can be seen (Fig. 4.30). The presence of sulfur is attributed to the MDMO-PPV precursor. The presence of argon is due to its implantation during etching; its content is relatively low (< 4 atom%). The first aging effect is detected when indium and tin appears in an oxidized form at the interface of the aged device [Fig. 4.30(b)]. These two elements are not observed in the case of unaged solar cells. The ratio of In:Sn found is around 10:1. These peaks are still present but at lower amounts in the active layer [Fig. 4.30(c)]. A part of the ITO anode compounds [In2 O3 or In(OH)3 , and SnO2 or Sn(OH)4 species with binding energy 445eV for In(3d5/2 ) and 486.7eV for Sn(3d5/2 )]42–45 diffuse across both organic layers (PEDOT:PPS and MDMO-PPV:PCBM) as far as the cathode-organic interface when the cell is operating under illumination. This result gives an extension of the formation of an interfacial layer with indium and tin from the ITO that has diffused through the active layer. This effect has been observed in aged diodes.46 In Fig. 4.31, the quantitative evolution of O compared with Al and C gives information about oxygen distribution and its diffusion during aging of the cells. First, in the case of the unaged cell, there is some native oxidation of aluminum at the surface of the cathode and a notable amount of oxygen at the metal-organic interface is also present. After aging, a presence of oxygen in the bulk of the Al cathode is observed and this is ascribed to the porosity of the Al cathode. Furthermore, the oxygen content increases at the metal-organic interface. Oxygen accumulation at the metal-organic interface has already been reported in a TOF-SIMS study of the Al/C60 interface of a Al/C60 /C12 -PPV/PEDOT:PSS/ITO/glass device under illumination in the presence of an isotopically labeled oxidative atmosphere.12 4.3.7 Li and F distribution To properly compare Li and F in XPS studies, it is important to take the mean free path (mfp) of emitted photoelectrons into account because of the great difference of the binding energy of these elements [around 56 eV for mean peak Li(1s), and 686 eV for F(1s)]. According to Penn’s calculations,47 mfp(Li) is around 50% greater than mfp(F). In the case of a homogenous distribution of Li and F, elements in the LiF layer Li should be observed first when probing the cathode–active layer interface. Another important point is that the XPS detection of lithium is more
194
Chapter 4
Figure 4.30 Evolution of XPS survey scans after cathode sputtering into aged cells: (a) Al cathode, (b) at the Al/LiF/active layer interface, (c) organic active layer (only the greatest peaks are noted).
difficult than fluorine because of the smaller cross section of Li atoms than F atoms. To compensate for this point, a great number of scans have to be accumulated for Li atoms (typically a few hundred scans) compared to F atoms (tens of scans). Quantitative depth profiles of Li and F elements are shown in Fig. 4.32. Before and especially after illumination of the solar cell, different depth profiles for F
Lifetime and Stability Studies
195
Figure 4.31 Percent atomic depth profile of main elements Al, C, and O read. Unaged cell in solid line, and aged cell in dashed line.
Figure 4.32 Percent atomic depth profile of main elements Al, C and O read. Unaged cell in full line and aged cell in dashed line.
196
Chapter 4
atoms are observed when compared to the depth profiles of Li atoms. Fluorine seems to be concentrated near the Al/LiF interface, whereas lithium is concentrated near the LiF active-layer interface. Thus, the distributions of Li and F elements are not homogeneous, as discussed above. For aged devices, the fluorine distribution does not seem to vary with depth, while the lithium distribution seems to diffuse toward the polymer layer. To determine if this diffusion phenomenon is only due to the sputtering process or if it is a redistribution or reorganization of the interface, we have considered two cases. SRIM03 simulation of the Al/LiF/polymer multilayer structure under ionic sputtering does not show any prevalence of transfer of Li or F atoms (Fig. 4.33). The simulation informs us about the overall distribution of Li, and F as obtained experimentally (especially around the interfaces). SRIM03 does not take chemical effects of Ar sputtering process into account and gives only a clue about the origin of the different depth distributions of Li and F atoms. We have observed lithium and fluorine depth profiles in nontreated Al(4 nm)/ LiF(1 nm)/MDMO:PPV systems by variation of the angle between XPS detector and the normal of the sample (Fig. 4.34). The greater angle that is used, the nearer surface of the device is probed. In order to try to understand the observed quantitative profile, we have to take into account both photoelectron mfp and heterogeneous distribution of Li and F at the interfaces.
Figure 4.33 SRIM03 simulation of recoiled Li and F atoms distributions in [Al (5 nm)/LiF (1 nm)/polymer (70 nm)] system after Ar+ (3 keV) sputtering (after 5000 counts). Densities applied were 2.702 for Al, 0.8225 for LiF, and 2.3 for polymer.
Lifetime and Stability Studies
197
Figure 4.34 The relative percent of the atomic angle resolved by XPS of Li and F on an Al(4 nm)/LiF(1 nm)/MDMO:PPV sample.
At a high angle, a slight majority of lithium is detected because of its longer photoelectron mfp compared to fluorine. In the case of a homogenous distribution of Li and F atoms at the interfaces, we should observe a slight difference in the lithium amount at a smaller angle until a stoichiometry of 1:1 for Li and F is reached, when all the Li and F photoelectrons can be detected. But this is not the case; as the α angle is decreased, the XPS observation depth gradually increases and the relative fluorine amount becomes dominant as the photoelectrons from fluorine become increasingly more detectable. As the α angle increases further and the detector reaches normal incidence, the Li:F ratio becomes stoichiometric. The LiF layer can thus be considered as a heterogeneous fluorine and lithium volume with a maximum amount of fluorine near the Al/LiF interface and a maximum amount of lithium nearer the LiF–active layer interface. This observation does not totally exclude the possibility that the Li and F distribution is affected by the ionic treatment. We can, however, conclude that the slight diffusion phenomenon is not only due to forward implantation during the sputtering operation. It is noticeable that this separation process of Li and F also has been observed in the case of the Al/LiF/Alq3 system.28 This observation could suggest that a chemical reaction that decomposes the LiF takes place.
4.4 Studies of Degradation Mechanisms Using RBS Rutherford backscattering spectroscopy (RBS), nuclear reaction analysis (NRA), and elastic recoil detection analysis (ERDA) with incident MeV ions are powerful methods for the quantitative analysis of thin films and depth profiling near the surface of solids. These are useful tools to investigate, especially, the electrode degradation mechanisms in solar cells.
Chapter 4
198
Figure 4.35 Sample chamber in a RBS experiment. (Source: Evans Analytical Group, 2007.)
4.4.1 Principles and quantitative depth profile of the composition RBS is an analytical technique in materials science. A high-energy beam [2–4 MeV (mega–electron volt)] of low mass ions (e.g., 4 He+ , 4 He++ ) is directed at a sample (see Fig. 4.35). A detector is placed so that scattered particles from the sample are collected at an angle as close to 180 deg as possible. The energy of these ions depends on their incident energy and on the mass of the sample atom with which they collided. The amount of energy transferred to the sample atom in the collision depends on the ratio of masses between the ion and the sample atom. Thus, measuring the energy of scattered ions indicates the chemical composition of the sample. Additionally, in the case that the incident ion does not hit any of the atoms near the surface of the sample but instead hits an atom deeper than the surface, the incident ion gradually loses the energy as it passes through the solid, and again as it leaves the solid. Hence, RBS is an efficient method to perform a quantitative depth profile of the composition of a sample. This is especially useful in analysis of thin-film materials like metallic electrodes; for example, films about half a micrometer in thickness can be profiled using a 2 MeV 4 He++ beam. 4.4.1.1 Sensitivity
The relative number of particles backscattered from a target atom into a given solid angle for a given number of incident particles is related to the differential scattering cross section. The scattering cross section is basically proportional to the square of the atomic number of the target atom, as seen in Eq. (4.1). Consequently, RBS is over 100 times more sensitive for heavy elements than for light elements due to the larger scattering cross sections of the heavier elements (see Fig. 4.36):
∂σ Z1 Z2 e2 = ∂Ω 4E
2
4 {cos θ + 1 − [(M1 sin θ)/M2 ]2 }2 , sin4 θ 1 − [(M1 sin θ)/M2 ]2
(4.1)
Lifetime and Stability Studies
199
Figure 4.36 Relative yields for He backscattering from selected elements at an incident He energy of 2 MeV. (Source: Evans Analytical Group, 2007.)
Figure 4.37 Scheme of a Van der Graaff accelerator.
where Z1 is the atomic number of the incident ion, Z2 is the atomic number of the target atom, E is the energy of the incident ion, M1 is the mass of the incident ion, M2 is the mass of the target atom, and θ is the angle of incidence. 4.4.1.2 Technical facts and sample preparation
In order to correctly analyze an RBS spectrum, it is necessary to calibrate the spectra with a well-known energy. We analyzed a reference sample after every new series of samples. This reference sample is based on a 3-nm thick layer of gold and palladium (30 atom% of Pd) deposited on an aluminum substrate. Figure 4.38 presents a theoretical adjustment by SIMNRA software of experimental data from a reference sample. We can then deduce experimental parameters extracted from this sample, which are used for the whole series:
Chapter 4
200
Figure 4.38 RBS spectrum of an Au:Pd reference sample (— theoretical fit, –•– raw data).
Incidence angle Exit angle Detection angle Energy shift Energy per canal Detector resolution Detector solid angle
0 deg 10 deg 175 deg 12.945 keV 4.8538 keV 22 keV 0.1152 sr
Samples can be thin layers deposited on thick substrates such as vitreous carbon. Sensitive samples are transferred from the glove box to the analysis enclosure of the accelerator using a portable container under a controlled atmosphere. Modeling and fitting of the RBS spectra is achieved with the simulation program SIMNRA, which allows for the analysis of the backscattering spectra for ion beams with MeV ions. When complete cells were analyzed, a “bubbling phenomenon” was observed due to a too-large surface current density. Organic layers can be highly degraded because of the current density. 4.4.1.3 Elastic recoil detection analysis (ERDA)
Light elements (e.g., H, D) can be detected by elastic recoil detection analysis (ERDA). Incident ions penetrate the sample with a flat incident angle (3–10 deg). Some of them are scattered on sample atoms. Several of the recoiled sample atoms leave the probe as ions and can be detected. From the measured energy spectrum of the recoil, a concentration depth profile can be calculated in hydrogen or deuterium.
Lifetime and Stability Studies
201
4.4.1.4 Nuclear reaction analysis (NRA)
Under proton ion irradiation, target elements can undergo a nuclear reaction under resonance conditions for sharply defined resonance energy. The reaction product is usually a nucleus in an excited state, which immediately decays, emitting ionizing radiation. For example, the reaction 18 O(p)15 N to profile 18 O is 18
O + 1 H → 15 N + α + p
(4.2)
with a resonance at 628 keV. The energetic emitted α ray is characteristic of the reaction, and the number that is detected at any incident energy is proportional to the concentration at the respective depth of 18 O in the sample. The reactions used are highly specific and relatively low bombarding energies can be used. The method is particularly useful to determine isotopic oxygen at or near the surface of solids, such as oxide films, silicates, etc., and for the isotopic analysis of very small quantities of organic materials. Amounts in the order of 10−12 g of 18 O can be detected. 4.4.2 Studies of cathode degradation using RBS 4.4.2.1 Ca cathode degradation
Calcium is largely used as a cathode material and as a probe for highly sensitive permeation measurements.48 Indeed, organic materials used in organic solar cells are easily degraded under atmospheric conditions. These devices require encapsulation with a barrier material that exhibits extremely low permeation rates for water vapor and oxygen. In order to determine these low permeation rates, the degradation of a metallic calcium layer encapsulated with the barrier material is usually monitored. However, the individual contributions of oxygen and water vapor to this degradation have not been clearly distinguished. In this study, the corrosion of metallic calcium was investigated using isotopic markers in connection with ion beam analysis. We have shown that calcium reacts only with water at room temperature. During the degradation of calcium in the presence of water vapor and/or oxygen, calcium passes quickly from a reflective metallic state to a transparent oxidized state. It is supposed that the following reactions are taking place:49 Ca + H2 O → CaO + H2 ,
(4.3)
2Ca + O2 → 2CaO,
(4.4)
CaO + H2 O → Ca(OH)2 .
(4.5)
It is known, however, that calcium is not readily oxidized in dry air at room temperature, although the reaction is fast under moist atmospheric conditions.50, 51 Indeed, the contribution of oxygen in a humid atmosphere has not been clearly established.52 This work was thus performed in order to clearly distinguish between the two values of the oxygen transmission rate (OTR) and the water-vapor transmission rate (WVTR).
202
Chapter 4
4.4.2.2 Methods
Isotopic markers and ion beam analysis were used to distinguish between the degradation products from the reactions of water vapor and oxygen with calcium. Thermal evaporation was used to grow thin-layer (100 nm) calcium samples on 2-mm thick vitreous-carbon substrates. The calcium layer was then placed sequentially in different atmospheres containing various oxidizing species. To prevent degradation during the deposition process and subsequent sample transfer, the evaporation chamber was integrated into a glove box containing an inert atmosphere. The sample was transferred from the glove box to a hermetic enclosure using a portable container under anhydrous nitrogen. Oxygen depth profiles in the calcium layer were determined by RBS using a 0.5-mm diameter collimated beam of 2.0 MeV 4 He+ . Hydrogen depth profiles were obtained using elastic recoil detection analysis53 (ERDA) with a 2 MeV 4 He+ . Finally, nuclear reaction analysis (NRA) experiments were performed with a perpendicular incident beam of protons. 4.4.2.3 Results and discussion
In order to verify the nonreactivity of calcium to a dry atmosphere, samples were placed in contact with dry air (DA) and then with atmospheric air (AA) at room temperature. Figures 4.39 and 4.40 show, respectively, the evolution of O and H within the calcium layer, sequentially exposed to dry and wet atmospheres for relatively short times. The spectrum of the raw calcium shows a low degree of contamination with oxygen that probably occurred during the transfer of the sample from the glove box to the analysis tool. Under a dry atmosphere, constant levels of O and H indicated that the calcium did not react at room temperature without water. However, in a humid atmosphere, the strong reactivity of calcium resulted in
Figure 4.39 Oxygen analysis (RBS spectrum) of the calcium layer exposed sequentially in dry (DA) and atmospheric (AA) atmospheres for periods of 300 s and compared against the raw Ca layer.
Lifetime and Stability Studies
203
Figure 4.40 Hydrogen analysis (ERDA spectrum) of a calcium layer sequentially exposed to dry (DA) and atmospheric (AA) atmospheres for 300 s against that of the raw Ca layer.
an increase in O and H levels until a complete passivation of the layer was attained. The stoichiometry of the final compound indicated that Ca(OH)2 was formed. To determine the origin of oxygen incorporation in calcium in a wet atmosphere (i.e., water and oxygen gas or only water reaction), oxygen isotopic markers were used (18 O2 versus D2 16 O). Deuterated water was used for the ERDA analysis of the deuterium and hydrogen incorporation within the calcium layer. Characterization by RBS allowed for a distinction to be made between 18 O and 16 O. However, we also studied 18 O depth profiles by NRA using the 18 O(p,α)15 N reaction with 628 keV resonance.54 The NRA analysis of the degraded calcium layer was compared with a standard sample of a thin layer of Si18 O2 . The hydrogen and deuterium evolution were measured using ERDA. Both RBS and ERDA experiments were performed simultaneously using the same parameters as the first experiment. An 18 O2 (20%)/N2 (80%) atmosphere with a constant humidity of D2 16 O and a pressure of 80 Torr was introduced in a hermetic preparation chamber at room temperature. A relative hygrometry of 80% was fixed using a saturated sodiumchloride solution after an equilibrium time of 24 hr.55 As in the first experiment, a 80-nm calcium layer was deposited on a 2-mm thick vitreous carbon substrate and transferred into a hermetic enclosure. Then, calcium was placed in the 18 O2 /N2 / D2 16 O atmosphere for increasing periods of time. Between each exposure, the calcium was stored under dry nitrogen or analyzed in vacuum. As observed in the first experiment, the RBS spectrum of the raw calcium showed some oxygen contamination (Fig. 4.41), although at a lower level than observed in the first experiment. The RBS spectra of the degraded Ca showed an increase in the O peak at the surface as compared to the bulk of the material. After 4265 s in the wet atmosphere, the oxygen peak reached a maximum, correspond-
204
Chapter 4
Figure 4.41 RBS spectra of the calcium layer exposed to an 18 O2 (20%)/N2 (80%)/D2 O(80% relative hygrometry) atmosphere for increasing periods of exposure.
Figure 4.42 Experimental RBS spectra of the raw (2) and the totally passivated calcium layer (×) compared to the SIMNRA simulations of the totally passivated calcium layer including only 16 O (—) or 16 O (80%) + 18 O (20%) (- - -).
ing to complete Ca passivation. Moreover, the oxygen profile was similar to the previous profile (see Fig. 4.39), indicating that only 16 O was detected. Figure 4.42 compares an experimental and a SIMNRA simulation56 of RBS spectra of raw and totally degraded Ca layers. If the simulation parameters include only 16 O (neglecting the natural isotopic distribution of 0.2% 18 O), a good agree-
Lifetime and Stability Studies
205
Figure 4.43 Oxygen 18 profile of the totally passivated calcium layer in 18 O2 (20%)/N2 (80%)/D2 O(80% RH) atmosphere (1) and Si18 O2 standard ("), deduced from the nuclear reaction analysis of 18 O with the 18 O(p,α)15 N resonance at 628 keV.
ment between the experimental and simulated data is observed. If 20% of 18 O was incorporated in the Ca layer, the corresponding spectrum should show a second peak in the 800–900 keV range. When compared to the standard Si18 O2 profile (Fig. 4.43), it was not possible to detect 18 O on the NRA profile of the totally passivated Ca. This good agreement between RBS and NRA analysis confirms the absence of a detectable amount of 18 O in the Ca layer. It is therefore possible to state that there was no reaction between metallic Ca and oxygen gas, even in a humid atmosphere. The calcium test previously described is only sensitive to H2 O. The calcium layer can be used as a probe of H2 O to measure the WVTR, but not the OTR. 4.4.2.4 Ca/Ag or Ca/Al cathode degradation
Based on the study of Ca cathode degradation, a methodology to protect the Ca layer from water can be considered. One idea is to cover the Ca layer with Ag or Al to convey protection from the beginning. First, we discuss results from studies of the Ca/Ag electrode. In order to study the degradation of the Ca/Ag cathode, we used RBS to analyze this metal layer. The first results obtained on a Ca/Ag double layer cathode evaporated on a vitreous carbon substrate indicated the formation of a Ca-Ag alloy at the interface between the Ca and Ag layers regardless of the thickness of Ag. The Ca/Ag bilayer is partially or totally transformed into a Ca-Ag alloy at the interface layer depending on the thickness of the Ag layer. Figure 4.44 shows the comparison of simulated (SIMNRA) and experimental spectra of 2 MeV alpha particles backscat-
206
Chapter 4
Figure 4.44 RBS spectra of the stack C/Ca(20 nm)/Ag(60 nm), incident energy = 2 MeV; Dashed line is the double layer simulation. Solid line is the alloy simulation. Experimental data are represented in open squares.
tered on the sample. The hypothesis of a double layer is presented as a dashed line. The simulation clearly shows that the interface between the two metallic layers is not abrupt. This is confirmed by the second simulation (shown as a solid line), which models the four-layer structure C/Ca/Cax :Ag1−x /Cay :Ag1−y /Ag. The concentration gradient of Ca and Ag of the electrode depends on the thickness of the different layers. For a Ca/Ag (20/20 nm) electrode, there is very slight sign of pure Ca in the first layer, but not pure Ag. In the case of a thicker electrode, a pure Ca layer is observed followed by the alloy layer of Ca and Ag, and then the pure Ag layer on top. Table 4.1 gives the complete composition of the layers of three electrodes, of 20/20 nm, 20/60 nm, and 30/80 nm. Notice that oxygen traces are only observed for the thinnest electrode (20/20 nm). We reproduced the same experiments for Ca/Al electrodes and we noticed here a pure double layer of Ca and Al (Table 4.2). We also varied the thickness of the different layers (10/80 nm, 20/80 nm, and 50/60 nm) without noticing the formation of an alloy. The RBS spectra in Fig. 4.45 clearly show the presence of oxygen at the Ca/Al interface and also the existence of aluminum oxide (Al2 O3 ) at the surface. 4.4.2.5 Influence of the nature of the electrode on aging
After studying the electrode by RBS, we prepared a series of solar cells to correlate the previous results with the performance in time of cells made with different electrodes. The accelerated lifetime study20 was carried out under an inert atmosphere
Lifetime and Stability Studies
207
Table 4.1 Element concentration (in ×1015 at/cm2 ) of the simulated layers for three Ca/Ag electrodes: 20/20 nm, 20/60 nm and 30/80 nm.
Ca/Ag electrode thickness
Element Concentration (× 1015 at/cm2 ) Ag H O Ca 6 15 9 35 10 5 43 7 20
20/20 nm
Layer 1 Layer 2 Layer 3
20/60 nm
Layer 1 Layer 2 Layer 3
221 54
Layer 1 Layer 2 Layer 3
186 177
30/80 nm
15 5
28 9
14 2 4
60 6
Figure 4.45 RBS spectra of the stack C/Ca(X nm)/Al(X nm), incident energy = 2 MeV. Solid line is the double layer simulation. Experimental data are represented in open circles.
at a temperature of 70±5◦ C imposed by the illumination (AM 1.5, 100 mW cm−2 ). All the parameters that can be extracted from the IV curve, such as Voc , Isc , FF, ηe , Rs , and Rsh , were recorded during illumination every hour. We have mainly focused on the first four parameters to follow the degradation process. The cells
Chapter 4
208
Table 4.2 Element concentration (in ×1015 at/cm2 ) of the simulated layers for three Ca/Al electrodes: 10/80 nm, 20/80 nm and 50/60 nm.
Ca/Al electrode thickness 10/80 nm
Layer 1 Layer 2 Layer 3
20/80 nm
Layer 1 Layer 2 Layer 3 Layer 1 Layer 2 Layer 3
50/60 nm
Element Concentration (×1015 at/cm2 ) Al H O Ca 21 9 400 7 17 27 400
16 1
16 9
40
15
19
100
21 313
Figure 4.46 Evolution of power conversion efficiency (up) and shunt resistance (down) of P3HT:PCBM cells based on a Ca/Ag cathode.
chosen for this study present the average of eight cells and the results between cells were easily reproducible in performance. Overall, three phases appear in the degradation process that can be rationalized by physicochemical processes. The shape of the degradation of the power conversion efficiency obtained over 600 hr is shown on Fig. 4.46. The first phase can be ascribed to a slow annealing of the device and can be viewed as a stabilization time (this phase is not observed for all device geometries where different cathodes are involved). In the second phase, the first signs of degradation are observed as a change of slope. This degradation is ascribed to changes at the cathode or the
Lifetime and Stability Studies
209
interface at the cathode. The last phase shows an increase in the slope and may possibly involve the bulk of the active layer. Solar cell devices based on Ca/Al and Ca/Ag electrodes were prepared. The test conditions remained the same (70 ± 5◦ C, AM 1.5G illumination). For Ca/Ag or Ca/Al electrodes, we chose to evaporate a thick layer of silver (200 nm) or aluminum (120 nm) over 20 nm of Ca. (This important point is discussed and developed in the paragraph below.) The evolution of those two electrodes over time is shown in Fig. 4.47. The progression of the cells barely differs in the first 200 hr. The Voc and Isc of the two resulting series of cells are stable over the first 200 hr. A higher and more stable FF for Ca/Al electrodes is the reason for the better performance of this series of cells in the first 200 hr. The tendency is inverted after this time, and the evolution of Ca/Al cells show a 40% loss for Isc and 11% loss for Voc , and FF decreases dramatically below 0.25. The power conversion efficiency passes below 1% after 410 hr. At the end, Ca/Ag electrodes show the best stability. After 600 hr, we measure less than 9% loss for Isc and around 18% for Voc . FF is still higher than 0.3 after this time. For Ca/Ag electrodes, the power conversion efficiency of the cells decreases below 1% after 600 hr, or 200 hr later than the Ca/Al electrode cells. In the case of Ca/Ag, Fig. 4.47 shows clearly two steps in the degradation kinetics of the Ca/Ag cathode solar cell. This first part of the degradation process
Figure 4.47 Evolution of the main PV parameters over 500 hr of standard P3HT:PCBM cells based on Ca/Ag (open square) and Ca/Al electrodes (open star).
210
Chapter 4
Figure 4.48 Evolution of short-circuit current density normalized from 0 to 400 hr for different thicknesses of Ag above the 20-nm Ca layer—20 nm (square), 60 nm (triangle), and 200 nm (star).
needs a more careful study to correlate the lifetime of devices to the nature of their cathode. Hence, the thickness of the silver layer of the Ca/Ag cathode was varied as follows. Three series of cells with silver layers of 20 nm, 60 nm, and 200 nm were prepared with the same calcium thickness (20 nm). The duration of the first phase of the degradation increases with the thickness of the Ag layer, as shown in Fig. 4.48. For the thinner layer (20/20 nm), the change of the slope already appears after 4 hr, whereas for the thicker layer (20/200 nm), signs of degradation appear after 120 hr. RBS can be applied to understand the photo-oxidation of the Ca/Ag and Ca/Al cathodes and their composition. We have shown that the duration of the first aging step (cathode aging) can be correlated to the Ag-Ca alloy thickness. The thickness of this layer is probably stable after the complete reaction of the Ca atoms with Ag to form an alloy. Thus, increasing the thickness of silver seems to increase the protective barrier effect of silver against residual water or oxygen.
4.5 Studies of Degradation Mechanisms Using Physical and/or Spectroscopic Techniques The previous sections have described the use of TOF-SIMS, XPS, and RBS methodologies to study degradation mechanisms in organic solar cells. TOF-SIMS and XPS are chemical characterization techniques, and RBS is a physicochemical characterization technique. The prospect of complementary information makes it advantageous to combine chemical characterization techniques with physical and/or spectroscopic characterization techniques. Physical and/or spectroscopic charac-
Lifetime and Stability Studies
211
terization techniques have typically been used to study the morphology in organic solar cells with the purpose of optimizing the photovoltaic properties. In the following sections, examples are given that illustrate the applicability of physical and/or spectroscopic methodologies to study degradation mechanisms. The extracted information is correlated with the corresponding information obtained from chemical characterization. 4.5.1 Interference microscopy Interference microscopy is a spectroscopic method based on light. The outcome is a topographic map of the analyzed surface. The strength of interference microscopy is the excellent height resolution in the Angstrom range, combined with the ability to analyze large areas in the millimeter range fairly quickly (it is not a vacuum technique). The phenomenon of oxygen diffusion through the electrode (e.g., Al) as opposed to the electrode (e.g., ITO) that is situated on a glass substrate was discussed in previous sections. One of the examples described the phenomenon for an organic solar cell composed of Al/C60 /C12 -PPV/PEDOT:PSS/ITO. TOF-SIMS imaging on the exposed C60 surface revealed circularly oriented incorporation of oxygen in the lateral plane, suggesting the presence of microscopic holes in the corresponding Al electrode. In order to verify the microscopic holes in the Al electrode, the exposed C60 surface as well as the surface of the Al electrode were subjected to an interference microscopy analysis. The resulting topographical map is shown in Fig. 4.49(A) for the exposed C60 surface. The surface is covered by circular protrusions of varying size. A line profile is shown in Fig. 4.49(B) from the part of the image indicated by the white horizontal line in Fig. 4.49(A). The topographical map for the Al electrode surface (not shown) has equivalent properties, suggesting that the Al electrode has adopted the shape of the C60 surface. The larger protrusions appear to have a hole in the center. The topographical map confirmed the presence of at least some holes on the surface of the Al electrode, and the fact that the holes are also observed on the surface of the exposed C60 surface suggests that the holes penetrate at least the Al electrode. The analysis also revealed that each hole is centered in a protrusion. The theory is that oxygen diffuses through the holes in the Al electrode and once it reaches the sublayer of C60 , it continuously diffuses in both the vertical and horizontal planes. When oxygen reacts with the C60 , the oxygen uptake causes the material to expand in the horizontal and vertical planes. The expansion in the vertical plane causes protrusions to be formed on the Al electrode surface. This conclusion was possible only because results from physical characterization (i.e., interference microscopy) was correlated with the result from chemical characterization (i.e., TOF-SIMS imaging), which clearly demonstrates the importance of employing complementary techniques.
212
Chapter 4
Figure 4.49 Interference microscopy image (A) covering 0.134 × 0.179 mm2 of an exposed C60 surface from an organic solar cell with the composition Al/C60 /C12 PPV/PEDOT:PSS/ITO. The lower plot is a line profile through the image (B) indicated by the white line. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
4.5.2 Atomic force microscopy (AFM) AFM produces a topographical map by scanning a needle across the surface. The weakness of the technique is the fact that only small areas can be analyzed (∼100× 100 μm2 ) and the acquisition time is long; thus, it is not suited for screening purposes. The strength of AFM is the excellent height resolution and lateral resolution, both in the angstrom range. The interference microscopy image in Fig. 4.49(A) confirmed the presence of microscopic holes in the center of the largest of the protrusions on the Al electrode surface. The problem is that the lateral resolution of the interference microscope is not sufficient enough to detect all microscopic holes on the Al electrode surface. AFM has, however, a superior lateral resolution and is thus applicable for topographical mapping on smaller scales. The AFM image in Fig. 4.50 was acquired on the surface of the Al electrode from an organic solar cell with the composition Al/C60 /C12 -PPV/PEDOT:PSS/ITO. AFM is well suited for analyzing single protrusions. Figure 4.50(A) is an AFM image of a single protrusion. The protrusion has a diameter of ∼12 μm and the
Lifetime and Stability Studies
213
Figure 4.50 AFM image (A) covering 40 × 40 μm2 of the surface of an Al electrode from an organic solar cell with the composition Al/C60 /C12 -PPV/PEDOT:PSS/ITO. The lower plot is a line profile through the image (B) indicated by the black dashed line. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
hole in the center is 1–2 μm. The line profile in Fig. 4.50(B), corresponding to the black dashed line in Fig. 4.50(A), reveals the depth of the hole to be ∼60 nm. The Al electrode is only ∼22-nm thick and the C60 layer is ∼100-nm thick; thus, the hole penetrates the Al electrode and continues further into the C60 layer. The hole could actually be even deeper; that is, the needle dimension could prevent the entire depth of the hole to be measured. By analyzing all protrusion within a given area using AFM, it becomes possible to conclude that all protrusions have a hole in the center. The AFM results made it possible to confirm the conclusions made from the TOF-SIMS imaging and interference microscopy analyses. 4.5.3 Scanning electron microscopy (SEM) SEM has become a standard tool for visualizing morphology. The strength of the technique is the superior lateral resolution and the capability of analyzing a broad range of scales from the nanometer range to the millimeter range. The weakness is the lack of a depth scale and the fact that it is a vacuum technique. There numerous examples in the literature on SEM analysis of organic solar cells; however, the work has, in one way or another, been focused on optimization of the morphology for the purpose of optimizing the photovoltaic properties. The organic solar cell with the composition Al/C60 /C12 -PPV/PEDOT:PSS/ITO has been extensively described in the previous sections. SEM analysis of this device produced additional information that is interesting from a degradation point
214
Chapter 4
Figure 4.51 SEM images from different surface locations (A–C) of an Al electrode from an organic solar cell with the composition Al/C60 /C12 -PPV/PEDOT:PSS/ITO. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
of view. Figure 4.51 shows SEM images of the Al electrode surface at different surface locations. The first image [Fig. 4.51(A)] is a small-scale (400 × 400 nm2 ) image clearly revealing that the Al electrode is composed of aluminum grains. This raises the questions: Is there an alternative entrance channel for oxygen? Can oxygen diffuse in between the aluminum grains? It has for a long time been a known fact that an organic solar cell degrades very quickly (1–2 days) in ambient air if not encapsulated, which is indirect evidence of oxygen having entered the device. There has been no work in the literature that describes specific entrance channels for oxygen into an organic solar cell. Figure 4.51(C) shows a hole so small that it would have been impossible to detect with interference microscopy. The high lateral resolution enables fine details to become visible, which could possibly give an indication to the origin of the holes. The hole in question [Fig. 4.51(C)] appears to have been created from something like an air bubble or a particle that forced its way out. There are no reports in the literature describing how holes are formed in electrodes used in organic solar cells. Figure 4.51(B) is a good example of the difficulties that can be encountered when interpreting SEM images. The two white areas could be protrusions, but they could also be an entirely different phenomenon. The white shading indicates an elevated SEM response that could be caused either by an elevated position on the surface (i.e., protrusion) or by another species in one of the sublayers. There is a certain degree of transparency in an SEM analysis that typically becomes significant when analyzing multilayered thin films such as organic solar cells. Figure 4.52 is a large-scale (0.75 × 1.12 mm2 ) SEM image of the surface of the organic solar cell (Al/C60 /C12 -PPV/PEDOT:PSS/ITO) that has been extensively discussed in the previous sections. Part of the Al electrode was peeled off, exposing the C60 surface. The SEM image is acquired over an area covering both the Al elec-
Lifetime and Stability Studies
215
Figure 4.52 SEM image of a surface area partly covering an Al electrode and partly covering an exposed C60 surface as a result of the Al electrode having been peeled off. The composition of the organic solar cell is Al/C60 /C12 -PPV/PEDOT:PSS/ITO. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
trode surface and the exposed C60 surface. The image contains a lot of information, of which not everything is straightforward to interpret. The almost diagonal lines are believed to be macroscopic fractures caused by the peeling process. The elongated twisted objects in the right part of the image are only observed on the Al electrode surface in the vicinity of the exposed C60 ; thus, the phenomenon is most likely related to the peeling process, and thus not interesting from a degradation point of view. The most interesting observation in Fig. 4.52 is the circular dark areas. This phenomenon is the same as that observed using TOF-SIMS imaging (Fig. 4.13), interference microscopy (Fig. 4.49), and AFM (Fig. 4.50). Oxygen has diffused through microscopic holes in the Al electrode and continued in the lateral and vertical directions, causing circular areas (in the lateral plane) to be oxidized (i.e., degraded). The dark circular areas are also observed on the part of the surface where the Al electrode is still attached. This is due to the aforementioned transparency effect, which confirms that the phenomenon is not a result of the peeling process. The center of the circular areas appears to provide an intense SEM response. Two explanations to this phenomenon come to mind, namely, (1) there could be a particle (e.g., as observed in Fig. 4.19) situated in one of the sublayers, or (2) the confirmed hole in the center could be so deep that the ITO layer either directly causes or induces an elevated SEM response. The authors used a combination of light microscopy and fluorescence microscopy to conclude that
216
Chapter 4
there is no correlation between the position of the holes and the position of the particles, which excludes particles from having anything to do with the phenomenon. 4.5.4 Fluorescence microscopy Fluorescence microscopy is not a vacuum technique; it is easy to use with fast acquisition times, is a practical tool. The resulting image visualizes possible fluorescence from the analyzed material, or in the case of more than one species being present, visualizes contrast in emitted fluorescence. Typical cathodes (e.g., Al, Ca, or Ag) are nontransparent; thus, the organic solar cell is analyzed from the transparent substrate side (e.g., glass). A state-of-the-art fluorescence microscope is typically a confocal-laser-scanning-fluorescence microscope. The confocal utility enables fluorescence to be detected from well-defined depths; thus, when scanning the depth, three-dimensional information is obtained. However, a typical organic solar cell has a thickness of 300–500 nm (disregarding the substrate), which is too thin compared to the depth resolution of confocal-laser-scanning-fluorescence microscopy analysis; thus, for multilayered thin films such as organic solar cells, possible fluorescence is detected from all transparent layers. Fluorescence microscopy has been used by Norrman et al.11 on organic solar cells with the composition Al/C60 /C12 -PPV/PEDOT:PSS/ITO. Fluorescence images were obtained by two devices; one had been illuminated in air [Figs. 4.53(C) and (D)] and one had been stored in darkness and air [Figs. 4.53(A) and (B)]. The right-hand images correspond to the areas indicated by white dashed squares in the left-hand images. There are distinct differences between the image of the illuminated device and the image of the device that was stored in darkness. The image of the illuminated device shows the dark circular areas that were also observed using various other techniques (discussed in previous sections) as a chemical and/or physical contrast (Figs. 4.13, 4.49, 4.50, and 4.52). The device that was stored in darkness [Figs. 4.53(A) and (B)] does not show dark circular areas, which is a useful observation. The active material C12 -PPV is highly fluorescence, but C60 is not (it actually quenches fluorescence). Furthermore, oxidized/degraded C60 does not exhibit fluorescence; thus, no information should come from the C60 layer. Oxidized/degraded C12 -PPV does not exhibit fluorescence, or at least not to any significant degree. It can thus be concluded that storing the device in darkness and air causes only the C60 to be oxidized/degraded. Furthermore, illumination in air causes both C60 and at least the sublayer of C12 -PPV to be oxidized/degraded; that is, the illumination appears to accelerate the oxidation/degradation process. The images in Fig. 4.53 contain additional information that is interesting from a degradation point of view. It is especially clear from the enlarged images to the right [Figs. 4.53(B) and (D)] that the devices contain particles that exhibit fluorescence. The sizes of the particles shown in Fig. 4.53(B) are a few micrometers, and
Lifetime and Stability Studies
217
Figure 4.53 Fluorescence microscopy images (excitation at 488 nm, emission at wavelengths >505 nm) obtained from the glass slides of organic solar cells with the composition Al/C60 /C12 -PPV/PEDOT:PSS/ITO after illumination in air (C and D) or after having been stored in darkness and air (A and B). The right-hand images correspond to the areas indicated by white dashed squares in the left-hand images. (Reprinted from Ref. [11], with permission from Elsevier, copyright 2006.)
the particles shown in Fig. 4.53(D) have more varying sizes. It was not investigated whether this was statistically significant. Unfortunately, it was not possible to determine in which layer or layers the particles are situated. It would otherwise have been useful to determine the mechanism of particle formation. It would be pure speculation to suggest that the particles possibly could be the same particles shown in Fig. 4.19.
4.6 Accelerated Lifetime Measurements for Extended Periods of Time Once the newly prepared OPV device has been prepared and characterized with respect to efficiency and IPCE, the following and perhaps most important questions arise: 1. 2. 3. 4.
How stable is the device under illumination? How stable is the device in the dark? How long of an operational lifetime can be anticipated? What degradation or failure mechanisms can be distinguished?
In the previous part of this chapter, a handful of techniques are presented that have been successfully employed for the elucidation of degradation paths for dif-
218
Chapter 4
ferent materials and devices. Methods that are helpful in the process of answering the last question have thus been developed and described. The first three questions, however, are not as simple to answer. Furthermore, the answer sought depends on the viewpoint of the experimenter. From a scientific point of view, the answer requires complete analysis of the degradation and failure in the devices, whereas a producer or consumer may only require that the device is stable enough for a given purpose. For a particular application, a short shelf life of, for instance, three months and a longer operational lifetime of 24 months may be required, whereas a medical application for a disposable device may require a long shelf life of 12 months and an operational lifetime of a few days. A few examples of very stable materials do currently exist13, 14, 20, 57 and from the experimenters point of view some means of speeding up the degradation in the laboratory is needed. When the lifetime exceeds a few months, it becomes impractical to do complete lifetime analysis and accelerated methods is the only available approach.14 The nature of an accelerated measurement is a simple engineering approach, where the device in question is subjected to conditions much harsher than the conditions the device will meet under normal operation. The conditions that are typically altered by engineers in an accelerated lifetime measurement are the temperature, light intensity, voltage, humidity, radiation, and mechanical stress. These parameters may be extreme (i.e., higher than normal) or they may be cycled up and down in value. A most common parameter that is increased beyond normal is temperature. The reasons for this are many, but the assumption that many degradation paths are chemical in nature and that they are thermally activated is a reasonable one. This type of behavior is accurately described by the Arrhenius equation, which is shown for a first-order process below as follows: Ea kdeg = A exp − , (4.6) kB T where kdeg is the degradation constant, A is the preexponential constant, Ea is the activation energy, kB is the Boltzmann constant, and finally, T is the temperature. In this simple approximation there is one parameter that the experimenter can adjust, and that is temperature. A higher temperature implies that the degradation is faster. The model is most often expanded to also include other parameters. In the simple case above, the degradation of the device is monitored by recording some parameters related to solar cell operation as a function of time. The best parameter to choose is the short-circuit current from the device, because this is directly related to the machinery of an OPV. Photons (light particles) are absorbed by the OPV and converted into electrons that are measured as a current. There is thus proportionality between the current and the state of operation of the device. The open-circuit voltage, for instance, is a poor parameter to monitor because there is no simple proportionality between the value recorded and the state of operation of the OPV. The efficiency and fill factor can also be chosen, but the most reliable parameter is the short-circuit current. In the experiment, the device is simply mounted under
Lifetime and Stability Studies
219
a set of conditions and the short-circuit current is measured as a function of time. The shape of the degradation curve is not necessarily a simple exponential decay. It may be linear, exponential, biexponential, or a combination—or something entirely different. It should be emphasized that the OPV is a complex device and once it has been prepared it is subject to degradation that is inevitable. The processes by which it degrades may be many, and the degradation may take place at different times during the history of an OPV device life. Different conditions may also stimulate different degradation paths to take place. The view of the degradation should thus be that it is a complex process. By keeping constancy in the conditions, a comparison can be made if just one parameter is altered. To the OPV experimenter, this parameter is either the temperature, the light intensity, or the atmosphere (oxygen and humidity). In the simple experiment where constancy in all parameters except temperature is kept, simple observations of the degradation constant, kdeg , becomes possible as a linear or simple exponential relationship, Linear decay:
Isc (t) = It=0 (1 − kdeg t),
Exponential decay:
Isc (t) = It=0 exp(−kdeg t),
Biexponential decay: a b 1 2 t) + t) . Isc (t) = It=0 exp(−kdeg exp(−kdeg a+b a+b
(4.7) (4.8)
(4.9)
The acceleration factor, K, for testing at elevated temperature can be extracted by determining kdeg at different temperatures and taking the ratio of the degradation constant at the elevated temperature and at the degradation constant at the assumed reference temperature (i.e., normal operating conditions) as Acceleration constant:
kdeg (Taccelerated ) kdeg (Treference ) Ea 1 1 = exp − . kb Treference Taccelerated (4.10)
K =
Ideally, kdeg should be determined at many different temperatures whereby both K and the activation energy, Ea , can be determined. Most often a complex bior multiexponential decay is observed. In such a case the acceleration factor becomes a function of time. In the biexponential case, the degradation is typically explained by the presence of two degradation mechanisms, namely, a fast one and a slow one. Several degradation mechanisms may be in play in each time domain (fast and slow) and cannot be distinguished because only their sum is observed. The acceleration factors for each time domain can then be determined. The chosen parameter (i.e., temperature) may influence the different mechanisms differently, and it is not uncommon to have different acceleration constants at different times during the device lifetime. When estimates of shelf life are required, the devices
220
Chapter 4
are most reliably kept in the dark under a given set of conditions and then subjected to a brief characterization (IV-curve) at given time intervals. The magnitude of acceleration factors are expected to be somewhere between 5 and 50 when comparing temperatures of 25◦ C and 80◦ C. In one literature example, an acceleration factor of 10 was found for a MDMOPPV-[60]PCBM blend when operated at 0.3 sun in an inert atmosphere.58 In practical terms, the temperature elevation implies that lifetime analysis in the laboratory can be shortened from years to months or weeks. It should be emphasized that accelerated testing is a very practical tool, but no substitute for real lifetime analysis under real-time conditions.
4.7 Apparatus for Lifetime Measurements and for Isotope Labeling The equipment that is required for accelerated lifetime measurement and isotopic labeling requires rigorous control of the conditions in terms of temperature, atmosphere, and light intensity. This is most easily achieved with an atmosphere chamber as shown in Fig. 4.54 where the device can be fixed inside and electrical connections can be made. The control of the temperature of the device under operation can be difficult since intense light from the solar simulator heats the device and a considerable amount of heat has to be removed to maintain a temperature of around 25◦ C. The most efficient method of achieving this is by physical con-
Figure 4.54 The apparatus for accelerated lifetime tests under the sun simulator (left) and a photograph of the chamber without the radiation shield. In such a chamber, two separate measurements can be carried out simultaneously (i.e., dark and light), while controlling the light intensity, temperature, and atmosphere. Under the sun simulator (left) there is a shield with two holes for the viewport. There is fan cooling of the chamber to avoid the illumination heating up the system. The chamber (right) has two viewports for cells, thermocouples for measuring the temperature of the atmosphere in the chamber, and for measuring the surface temperature of the devices. In the back of the annular chamber there is a fan that circulates the atmosphere in the chamber and a heat exchanger that controls the temperature of the atmosphere. There are inlets and outlets for gasses, and a port for vacuum pumping. Since the temperature control is by means of a circulating atmosphere, temperature control is not possible when measurements are carried out in vacuum.
Lifetime and Stability Studies
221
tact and heat transfer through a metallic block. For an OPV this is, however, not very practical since light needs access from the front side, which is normally the mechanically durable side. The other side is typically the back side and it is not unproblematic to make thermal contact to this side without destroying the electrical contact. Another possible method employs a circulating atmosphere with a controlled temperature. This is gentle on the device and only has the disadvantage that it does not work in vacuum. The apparatus shown in Fig. 4.54 has been employed for studies in a pure nitrogen, dry oxygen, isotopically labeled oxygen, and isotopically labeled humid oxygen-free nitrogen atmosphere. The system can also be operated in vacuum without temperature control. Isotopic labeling of oxygen or water in the atmosphere is a very useful tool when it comes to studying degradation and failure mechanisms by TOF-SIMS methods described earlier in this chapter. Since oxygen is a gas, it is readily introduced by mixing with pure N2 using simple vacuum techniques. The test chamber is simply pumped down to a low pressure of 1 × 10−4 mbar or lower using a turbomolecular pump, and 18 O2 is introduced through a valve to the desired pressure (i.e., 200 mbar), followed by the addition of pure nitrogen to a total pressure of 1 bar. This gives approximate atmospheric conditions with zero humidity. The only requirements are a good pumping system and a pressure gauge that spans the range from 1 bar down to 1 × 10−6 mbar or better. These are readily available and their use is highly advised.
Figure 4.55 Isotopic labeling is achieved by illuminating the device in a chamber, as shown in Fig. 4.54, while introducing isotopically labeled materials such as molecular 18 O2 from a pressurized cylinder (right) or H2 18 O labeled water as shown in its liquid from in small vials (left).
222
Chapter 4
Figure 4.56 A compact system for isotopic labeling comprising a turbomolecular pump, wide-range pressure gauge, gas mixing valves, and testing chamber (left). Under solar simulation (middle and right).
In the case of adding isotopically labeled water to the atmosphere, the volume of the chamber must be accurately known. The amount of H2 18 O for the desired humidity level is then added to the evacuated system through a syringe, followed by pure nitrogen, 18 O2 , oxygen, or any other desired gas mixture. A compact system for isotopic labeling is shown if Fig. 4.56, where a turbomolecular pump is attached to the chamber. There is gas mixing valves for the 18 O2 gas cylinder and for pure nitrogen, or any other gas mixture.
Lifetime and Stability Studies
223
References 1. Brabec, C.J., Sariciftci, N.S., and Hummelen, J.C., “Plastic solar cells,” Adv. Funct. Mater., 11, pp. 15–26 (2001). 2. Spanggaard, H., and Krebs, F.C., “A brief history of the development of organic and polymeric photovoltaics,” Sol. Energy Mater. Sol. Cells, 83, pp. 125–146 (2004). 3. Coakley, K.M., and McGehee, M.D., “Conjugated polymer photovoltaic cells,” Chem. Mater., 16, 4533–4542 (2004). 4. Hoppe, H., and Sariciftci, N.S., “Organic solar cells: an overview,” J. Mater. Res., 19, pp. 1924–1945 (2004). 5. Li, G., Shrotriya, V., Huang, J., Yao, Y., Moriarty, T., Emery, K., and Yang, Y., “High-efficiency solution processable polymer photovoltaic cells by selforganization of polymer blends,” Nat. Mater., 4, pp. 864–868 (2005). 6. Ma, W., Yang, C., Gong, X., Lee, K., and Heeger, A.J., “Thermally stable, efficient polymer solar cells with nanoscale control of the interpenetrating network morphology,” Adv. Funct. Mater., 15, pp. 1617–1622 (2005). 7. Reyes-Reyes, M., Kim, K., and Carroll, D.L., “High-efficiency photovoltaic devices based on annealed poly(3-hexylthiophene) and 1-(3-methoxycarbonyl)-propyl-1-phenyl-(6,6)C-61 blends,” Appl. Phys. Lett., 87, 083506 (2005). 8. Lira-Cantu, M., Norrman, K., Andreasen, J.W., and Krebs, F.C., “Semiconductor oxides and oxygen exchange in hybrid solar cells,” Chem. Mater., 18, pp. 5684–5690 (2006). 9. Norrman, K., Alstrup, J., Jørgensen, M., and Krebs, F.C., “Lifetimes of organic photovoltaics: photooxidative degradation of a model compound,” Surf. Interface Anal., 38, pp. 1302–1310 (2006). 10. Alstrup, J., Norrman, K., Jørgensen, M., and Krebs, F.C., “Lifetimes of organic photovoltaics: Design and synthesis of single oligomer molecules in order to study chemical degradation mechanisms,” Sol. Energy Mater. Sol. Cells, 90, pp. 2777–2792 (2006). 11. Norrman, K., Larsen, N.B., and Krebs, F.C., “Lifetimes of organic photovoltaics: Combining chemical and physical characterisation techniques to study degradation mechanisms,” Sol. Energy Mater. Sol. Cells, 90, pp. 2793– 2814 (2006). 12. Norrman, K., and Krebs, F.C., “Lifetimes of organic photovoltaics: using TOF-SIMS and 18 O2 isotopic labelling to characterise chemical degradation mechanisms,” Sol. Energy Mater. Sol. Cells, 90, pp. 213–227 (2006).
224
Chapter 4
13. Yang, X., Loos, J., Veenstra, S.C., Verhees, W.J.H., Wienk, M.M., Kroon, J.M., Michels, M.A.J., and Janssen, R.A.J., “Nanoscale morphology of highperformance polymer solar cells,” Nano Lett., 5, pp. 579–583 (2005). 14. Krebs, F.C., and Spanggaard, H., “Significant improvement of polymer solar cell stability,” Chem. Mater., 17, pp. 5235–5237 (2005). 15. Vickerman, J.C., and Briggs, D., TOF-SIMS Surface Analysis by Mass Spectrometry, IM Publications and SurfaceSpectra Ltd., West Sussex, UK (2001). 16. Briggs, D., and Grant, J.T. (Eds.), Surface Analysis by Auger and X-ray Photoelectron Spectroscopy, IM Publications and SurfaceSpectra Ltd., Manchester, UK (2003). 17. Brabec, C.J., Shaheen, S.E., Winder, C., Sariciftci, N.S., and Denk, P., “Effect of LiF/metal electrodes on the performance of plastic solar cells,” Appl. Phys. Lett., 80, pp. 1288–1290 (2002). 18. Scurlock, R.D., Wang, B., Ogilby, P.R., Sheats, J.R., and Clough, R.L., “Singlet oxygen as a reactive intermediate in the photodegradation of an electroluminescent polymer,” J. Am. Chem. Soc., 117, pp. 10194–10202 (1995). 19. Chambon, S., Rivaton, A., Gardette, J.-L., and Firon, M., “Photo- and thermal degradation of MDMO-PPV:PCBM blends,” Sol. Energy Mater. Sol. Cells, 91, pp. 394–398 (2007). 20. De Bettignies, R., Leroy, J., Firon, M., and Sentein, C., “Accelerated lifetime measurements of P3HT:PCBM solar cells,” Synth. Met., 156, pp. 510–513 (2006). 21. Brabec, C.J., Shaheen, S.E., Fromherz, T., Padinger, F., Hummelen, J.C., Dhanabalan, A., Janssen R.A.J., and Sariciftci, N.S., “Organic photovoltaic devices produced from conjugated polymer/methanofullerene bulk heterojunctions,” Synth. Met., 121, pp. 1517–1520 (2001). 22. Hung, L.S., Tang, C.W., and Mason, M.G., “Enhanced electron injection in organic electroluminescence devices using Al/LiF electrode,” Appl. Phys. Lett., 70, pp. 152–154 (1997). 23. Li, F., Tang, H., Anderegg, J., and Shinar, J., “Fabrication and electroluminescence of double-layered organic light-emitting diodes with the Al2O3/Al cathode,” Appl. Phys. Lett., 70, pp. 1233–1235 (1997). 24. Jabbour, G.E., Kawabe, Y., Shaheen, S.E., Wang, J.F., Morrell, M.M., and Kippelen, B., “Highly efficient and bright organic electroluminescent devices with an aluminum cathode,” Appl. Phys. Lett., 71, pp. 1762–1764 (1997). 25. Mori, T., Fujikawa, H., Tokito, S., and Taga, Y., “Electronic structure of 8hydroxyquinoline aluminum/LiF/Al interface for organic electroluminescent
Lifetime and Stability Studies
225
device studied by ultraviolet photoelectron spectroscopy,” Appl. Phys. Lett., 73, pp. 2763–2765 (1998). 26. Schlaf, Parkinson, R., B.A., Lee, P.A., Nebesny, K.W., Jabbour, G., Kippelen, B., Peyghambarian, N., and Armstrong, N.R., “Photoemission spectroscopy of LiF coated Al and Pt electrodes,” J. Appl. Phys., 84, pp. 6729–6736 (1998). 27. Shaheen, S.E., Jabbour, G.E., Morrell, M.M., Kawabe, Y., Kippelen, B., Peyghambarian, N., Nabor, M.-F., Schlaf, R., Mash, E.A., and Armstrong, N.R., “Bright blue organic light-emitting diode with improved color purity using a LiF/Al cathode,” J. Appl. Phys., 84, pp. 2324–2327 (1998). 28. Le, Q.T., Yan, L., Gao, Y., Mason, M.G., Giesen, D.J., and Tang, C.W., “Photoemission study of aluminum/tris-(8-hydroxyquinoline) aluminum and aluminum/LiF/tris-(8-hydroxyquinoline) aluminum interfaces,” J. Appl. Phys., 87, pp. 375–379 (2000). 29. Piromreun, P., Oh, H., Shen, Y., Malliaras, G.G., Scott, J.C., and Brock, P.J., “Role of CsF on electron injection into a conjugated polymer,” Appl. Phys. Lett., 77, pp. 2403–2405 (2000). 30. Heil, H., Steiger, J., Karg, S., Gastel, M., Ortner, H., Von Seggern, H., and Stöβel, M., “Mechanisms of injection enhancement in organic light-emitting diodes through an Al/LiF electrode,” J. Appl. Phys., 89, pp. 420–424 (2001). 31. Brown, T.M., Friend, R.H., Millard, I.S., Lacey, D.J., Butler, T., Burroughes, J.H., and Cacialli, F., “Electronic line-up in light-emitting diodes with alkalihalide/metal cathodes,” J. Appl. Phys., 93, pp. 6159–6172 (2003). 32. Brown, T.M., Friend, R.H., Millard, I., Lacey, D., Burroughes, J.H., and Cacialli, F., “LiF/Al cathodes and the effect of LiF thickness on the device characteristics and built-in potential of polymer light-emitting diodes,” Appl. Phys. Lett., 77, pp. 3096–3098 (2000). 33. Greczynski, G., Fahlman, M., and Salaneck, W.R., “An experimental study of poly(9,9-dioctyl-fluorene) and its interface with Li and LiF,” Appl. Surf. Sci., 166, pp. 380–386 (2000). 34. van Gennip, W.J.H., van Duren, J.K.J., Thüne, P.C., Janssen, R.A., and Niemantsverdriet, J.W., “The interfaces of poly(p-phenylene vinylene) and fullerene derivatives with Al, LiF, and Al/LiF studied by secondary ion mass spectroscopy and x-ray photoelectron spectroscopy: Formation of AlF3 disproved,” J. Chem. Phys., 117, pp. 5031–5035 (2002). 35. Ziegler, J.F., Handbook of Ion Implantation Technology, Elsevier, New York (1992). 36. Kroon, J.M., Wienk, M.M., Verhees, W.J.H., and Hummelen, J.C., “Accurate efficiency determination and stability studies of conjugated polymer/fullerene solar cells,” Thin Solid Films, 403, pp. 223–228 (2002).
226
Chapter 4
37. Neugebauer, H., Brabec, C.J., Hummelen, J.C., Janssen, R.A.J., and Sariftci, N.S., “Stability studies and degradation analysis of plastic solar cell materials by FTIR spectroscopy,” Synth. Met., 102, pp. 1002–1003 (1999). 38. Padinger, F., Fromherz, T., Denk, P., Brabec, C.J., Zettner, J., Hierl, T., and Sariciftci, N.S., “Stability studies and degradation analysis of plastic solar cell materials by FTIR spectroscopy,” Synth. Met., 121, pp. 1605–1606 (2001). 39. Briggs, D., and Seah, M.P., Practical Surface Analysis, 2nd ed., Vol. 1, Wiley, New York (1990). 40. Brown, T.M., Friend, R.H., Millard, I., Lacey, D., Burroughes, J.H., and Cacialli, F., “Efficient electron injection in blue-emitting polymer light-emitting diodes with LiF/Ca/Al cathodes,” Appl. Phys. Lett., 79, pp. 174–176 (2000). 41. de Jong, M.P., “Interface stability in polymer light emitting diodes,” Ph.D. thesis, Eindhoven (2000). 42. McGuire, G.E., Schweitzer, G.K., and Carlson, T.A., “Study of core electron binding energies in some group IIIA, VB, and VIB compounds,” Inorg. Chem., 12 pp. 2450–2453 (1973). 43. Barr, T.L., “Recent advances in X-ray photoelectron spectroscopy studies,” J. Vacuum Sci. Technol., 9, pp. 1793–1805 (1991). 44. Zeller, M.V., Grutsch, P.A., and Fehlner, T.P., “Photoelectron spectroscopy of tin compounds,” Inorg. Chem., 12, pp. 1431–1433 (1973). 45. Barr, T.L., “ESCA study of termination of passivation of elemental metals,” J. Phys. Chem., 82, pp. 1801–1810 (1978). 46. Gautier, E., Lorin, A., Nunzi, J.-M., Shalchi, A., Benattar, J.-J., and Vital, D., “Electrode interface effects on indium-tin-oxide polymer/metal light emitting diodes,” Appl. Phys. Lett., 69, pp. 1071–1073 (1996). 47. Penn, D.R., “Quantitative chemical analysis by ESCA,” Journal of Electron Spectrosc. Relat. Phenom., 9, pp. 29–40 (1976). 48. Cros, S., Firon, M., Lenfant, S., Trouslard, P., and Beck, L., “Study of thin calcium electrode degradation by ion beam analysis,” Nucl. Instrum. Methods Phys. Res. B, 251, pp. 257–260 (2006). 49. Nisato, G., Bouten, P.C.P., Slikkerveer, P.J., Bennet, W.D., Graaf, G.L., Rutherford, N., and Wiese, L., “Evaluating high performance diffusion barriers: the calcium test,” Proc. of Asia Display IDW’01, pp. 1435–1438 (2001). 50. Burrows, P.E., Graff, G.L., Gross, M.E., Martin, P.M., Shi, M.K., Hall, M., Mast, E., Bonham, C., Bennet, W., and Sullivan, M.B., “Ultra barrier flexible substrates for flat panel displays,” Displays, 22, pp. 65–69 (2001).
Lifetime and Stability Studies
227
51. Pascal, P., Nouveau Traité de Chimie Minérale, Vol. 4, Masson, Paris, p. 412 (1963). 52. Wu, Z., Wang, L., Chang, C., and Qiu, Y., “A hybrid encapsulation of organic light-emitting devices,” J. Phys. D, 38, pp. 981–984 (2005). 53. Tirira, J., Serruys, Y., and Trocellier, P., Forward Recoil Spectrometry, Plenum, New York (1996). 54. Amsel, G., and Samuel, D., “Microanalysis of stable isotopes of oxygen by means of nuclear reactions,” Anal. Chem., 39, pp. 1689–1698 (1967). 55. Greenspan, L., “Humidity fixed points of binary saturated aqueous solutions,” J. Res. Nat. Bur. Stand. Sect. A: Phys. Chem., 81, pp. 89–96 (1977). 56. Mayer, M., “SIMNRA User’s Guide,” Technical Report No. IPP9/113, Max Planck Institut für Plasmaphysik, Garching, Germany (1997). 57. Katz, E.A., Gevorgyan, S., Orynbayev, M.S., and Krebs, F.C., “Out-door testing and long-term stability of plastic solar cells,” Eur. J. Appl. Phys., 36, pp. 307–311 (2007). 58. Schuller, S., Schilinsky, P., Hauch, J., and Brabec, C.J.,“Determination of the degradation constant of bulk heterojunction solar cells by accelerated lifetime measurements,” Appl. Phys. A, 79, pp. 37–40 (2004).
Chapter 5
Processing and Production of Large Modules Tom Aernouts, Stéphane Cros and Frederik C. Krebs In the previous chapters, we have focused on the study of single organic solar cells. To power an actual electrical tool with such a device, one might think of simply enlarging the active device area, thereby nominally generating a higher output current and a higher amount of electrical energy. Unfortunately, from a technological and practical point of view, this is not always the best solution. Therefore, integration of solar cells into electrical applications creates the need for interconnecting a number of single cells with each other into a larger structure. We have seen that a standard single solar cell is characterized by an opencircuit voltage below 1 V. The voltage at the maximal power point is still somewhat lower than this, even at high illumination levels. The actual output current varies, of course, strongly with the light intensity, but the size of the active area of the device is also crucial. However, due to many technological issues, the dimensions of single solar cells will be limited. Small inhomogeneities in the active layer may deteriorate the performance,1 especially for the very thin films that are used in organic photovoltaic cells. These effects scale with the size of the device. Furthermore, the collection of the photogenerated current for large solar cells can be substantially influenced by the limited conductivity of the electrode material.2 Thus, the output of a single solar cell is too small to power most practical applications. Therefore, interconnecting separate devices into a module to enlarge the overall output has to be considered.3 This can be achieved by externally connecting solar cells that have been fabricated separately. This is what is commonly done in the wafer-based Si solar cell industry. Besides this, an integrated approach in which monolithic production of photovoltaic modules is achieved can also be considered. There, the different layers are processed on one substrate in such a way that the different cells are directly interconnected. This is, of course, a very attractive production process that is pursued for many thin-film solar cell technologies and also in the field of organic solar cells. Accurate structuring and patterning of the subsequent layers is absolutely necessary here. Irrespective of the approach,
230
Chapter 5
some general aspects of interconnecting solar cells have to be considered.4, 5 These are discussed here. A series connection of two cells is schematically depicted in Fig. 5.1(a). The metallic back-side contact of the first cell is therefore connected with the transparent front-side electrode of the second one. The external contacts are then situated at the transparent electrode of the first and the metallic contact of the second cell. The corresponding series connection of the equivalent circuits is represented in Fig. 5.1(b). From this, it can be clearly seen that under shortcircuit conditions for two identical cells, the externally extracted current equals the photogenerated current of a single cell. On the other hand, for an open circuit, the voltage is the sum of the open-circuit voltages of the two separate devices. Therefore, a series connection of two identical solar cells results in the IV characteristic as given in Fig. 5.1(c). The result is that the maximum power generated by the two cells equals the sum of the individually developed powers, with a doubling of the output voltage for a series connection. In commonly available Si-based modules, a sufficient amount of solar cells, for example, 34 or 36, is put in series to have an optimum output voltage that is higher than 12 V. In this way, it is possible to efficiently charge standard batteries that are still the most widely used storage medium in photovoltaic applications. For nonidentical solar cells with different short-circuit currents, the situation is more complex. Different short-circuit currents can be caused by internal differ-
Figure 5.1 (a) Schematic representation of a series connection of two solar cells; (b) corresponding equivalent circuit representation; (c) IV characteristic for a series connection of two identical solar cells; (d) IV characteristic for a series connection of two nonidentical solar cells.
Processing and Production of Large Modules
231
ences in the performance of the device or by different external influences such as the shadowing of one of the cells. From the equivalent circuit, the weakest cell that generates the smallest short-circuit current, strongly limits the total performance. Whereas the total open-circuit voltage is not strongly influenced by the mismatching, the total current in this case is almost completely determined by the weakest cell. If the IV characteristics of the individual cells are known, then the curve for the series interconnected cells can be predicted. For each current, the different voltages of the individual cells have to be added, as illustrated in Fig. 5.1(d). Obviously, the total power generated by the two mismatched cells is substantially less than the addition of the power produced by the individual cells. Therefore, series connections of solar cells should only be considered for cells with equal short-circuit currents. Similar considerations can be made for the parallel connection of solar cells. For two solar cells, such a configuration is schematically depicted in Fig. 5.2(a), and the corresponding equivalent circuit is represented in Fig. 5.2(b). For an open
Figure 5.2 (a) Schematic representation of a parallel connection of two solar cells; (b) corresponding equivalent circuit representation; (c) IV characteristic for a parallel connection of two identical solar cells; (d) IV characteristic for a parallel connection of two nonidentical solar cells.
232
Chapter 5
circuit of two identical cells, the total voltage equals the open-circuit voltage of a single cell. Under short-circuit conditions, the externally extracted current is the sum of the photogenerated currents of the two separate devices. Therefore, a parallel connection of two identical solar cells results in the IV characteristic as given in Fig. 5.2(c). This results with the maximum power generated by the two cells equaling the sum of the individually developed powers, with a doubling of the output current. Again, the situation is more complex for nonidentical solar cells. From the equivalent circuit, mismatching mainly effects the output voltage. The IV characteristic of the parallel connection can be found by adding the currents of the individual cells at each voltage. The result for nonidentical solar cells is given in Fig. 5.2(d). It shows that the parallel connection of nonidentical cells is indeed limited by the cell that generates the lowest output voltage. Although this effect seems to cause minor problems, the total power can still be almost equal to the sum of the individual powers. Nevertheless, it is obvious that mismatching of the cells has a detrimental influence on the overall performance of a system of interconnected solar cells, and it therefore must be avoided as much as possible during module production. Conclusively, a well-defined structuring of the different layers of the solar cell is necessary to achieve good interconnecting and matching of the different devices in a module. To obtain this for organic solar cells, new deposition techniques have to be introduced.
5.1 Printing and Coating Methods In this section, new technologies such as printing are introduced. As opposed to spin coating, these new technologies offer accurate and direct patterning of the deposited layers. From the outset, this can result in reduced material consumption since the material is only deposited in the areas were it is actually needed in the final device structure. Also, the formation of photovoltaic module structures becomes possible. Furthermore, the use of printing technology opens opportunities for easy integration of organic devices into other applications. In this respect, improved aesthetics of the final product can even be considered.6, 7 5.1.1 R2R coating The roll-to-roll (R2R) coating method as shown in Fig. 5.3 is well established as a technique and is very well suited for high-speed processing of thin-film devices. OPVs require several coating steps, namely, (1) transparent anode, (2) active layer, and (3) cathode. Each of these steps may require several substeps; especially the cathode, which with the current state of the art is going to involve a vacuum coating step. This means that there are several process steps that differ fundamentally in makeup. Setting up an R2R process line can be done in discrete steps or in an integrated in-line process. The in-line process has the advantage that there is less
Processing and Production of Large Modules
233
Figure 5.3 An illustration of discrete and integrated R2R processing of a multistep product such as the OPV.
damage to the films during spooling. The disadvantage is that one of the process steps may be slower; thus setting the pace of the entire production. In the discrete setup, this can be solved by dimensioning the process equipment of each step for a similar pace. At the time of writing this book, there are no scientific reports on large-scale R2R processing of OPVs based on conjugated polymers. This is not to say that it has not taken place, since industry often keeps details and processes to themselves. There is, however, another good reason that R2R coating is not the scientists’ first choice. First, it is inherently a high-volume process and in order to do a smallscale test, large amounts of substrates and active material are needed. R2R is thus a scale-up process and is not relevant until the right choice of active material and device disposition has been made. Second, the cost and effort put into making one device is large; thus, unless you are close to the final version of your device it is not efficient. Small batch type methods as shown in Fig. 5.4 are much more suited for the incremental small-scale studies (spin coating, screen printing, doctor blading, and flexo/pad printing). Commercially available coating can be tailored to suit the needs of any future OPV production plant. The printing process can thus be divided into the feed method and the printing method. R2R will undoubtedly be the choice when it comes to the feed method in a future process for OPV production. The choice of printing method depends on many factors such as the desired film thickness, the viscosity of the ink, and the postprinting processes required. Regardless of the choice of printing method, it will most likely be compatible with R2R processing. Some examples of printing methods follow in this section that are compatible with small-scale printing in batch mode (for scientific testing and development) and with the possibility to do R2R upscaling.
234
Chapter 5
Figure 5.4 Industrially available equipment for solution coating (left) and vacuum evaporation (right).
5.1.2 Screen printing As a low-cost production technique, screen printing8–12 can be preferred over other printing techniques for several reasons. Compared to techniques such as offset, gravure, or flexography,13 screen printing requires less complex equipment, which makes it highly suitable for research activities. Furthermore, upscaling the technique is easily accessible by its fully industrial variant in the form of rotary screen printing. This could result in an adequate production technology with a throughput that is only a factor of three to five lower than for the other techniques mentioned. With respect to ink-jet printing,14 the main benefit is believed to lie in obtaining a more appropriate resolution level. The main attributes needed in the screen printing process are the screen itself and a squeegee.15–17 The printing screen18–23 is made of a porous, finely woven fabric. This used to be silk, therefore the process is also known as silk-screening. Since the 1940s, typical fabrics used are polyester and nylon, but also stainless steel is commonly used today. These materials are woven into a fabric such that open areas are left in between the different wires or fibers, as depicted in Fig. 5.5(a). It is through such openings that the printed solution or ink will pass during the final process. Different types of screens are characterized clearly by the material from which they are made, but also by the filaments and wire diameter, the orientation and the number of wires, the weave texture etc. The two most important properties of the fabric are the mesh opening w and the wire diameter d. They are indicated in Fig. 5.5(b). The number of wires n per unit of length is called the mesh number and it defines the fineness of the screen. The
Processing and Production of Large Modules
235
diameter of the wire is responsible for the open area a0 in the screen, as follows: a0 =
w2 . (w + d)2
(5.1)
So, the open area denotes the percentage of all openings in relation to the total fabric area. The larger the open screen area, the larger the passage of the solution or ink to be printed. The theoretical maximal amount of ink that can go into the mesh openings during printing is called the theoretical paste volume Vth . It indicates the amount of printing solution or ink that can be present in a certain area of the screen fabric and therefore relates clearly to its open area and also to its thickness D, which is governed by the wire diameter and the weaving technique. The volume Vth (typically expressed in units of cm3 m−2 ) is a first approximation to the wetlayer thickness to be printed by the chosen fabric:24 Vth =
w2 D = a0 D. (w + d)2
(5.2)
After tight stretching, the fabric is mounted onto a printing frame. Materials such as, for example, wood, aluminum, or steel are mostly preferred since the frame should be as resistant as possible to deformations. The most commonly used technique to adhere the fabric to the printing frame is by gluing. The wires of the woven structure can thereby be oriented at specified angles. The choice of the frame size depends mainly on the desired printable area; there should always be an adequate zone reserved between this area and the frame. The printable area is defined onto the screen fabric by imaging of a photosensitive emulsion layer. A mask covers certain parts of the emulsion layer, whereas the other parts are exposed to UV light. The exposed parts cure and harden, resulting in impermeable material. The other parts can be washed afterwards. As such, a well-defined pattern in the emulsion layer is obtained. In Fig. 5.5(c), a screen partly coated with an emulsion layer is depicted. There are several ways to adhere
Figure 5.5 (a) Schematic representation of a screen fabric; (b) the different dimensions of a screen fabric; (c) screen partially coated by an emulsion layer.
236
Chapter 5
the emulsion layer onto the screen. It can be coated from a liquid form directly on the fabric and the imaging procedure is then carried out afterward. Alternately, it can also be simply pressed onto the screen from the film, which could be already patterned beforehand. Methods in between these also exist in which the emulsion coating is applied onto the woven structure by capillary forces due to wetting agents and whereby the imaging can be implemented either before or after. The method used can have some influence on the final solution or ink deposition, which can be basically related to the buildup of the emulsion, that is, the part of this layer that extends out or on top of the woven structure. This effect may raise the thickness of the printed wet layer because the volume in the screen that can be taken by the solution or ink has potentially enlarged. The second attribute for the screen printing process is the squeegee.25, 26 It is schematically depicted in Fig. 5.6(a), showing a blade that extends for a large part out of a holder. Commonly used materials are rubber or polyurethane, chosen according to the process, since each of them can have different chemical and mechanical properties. Properties such as the hardness and stiffness of the squeegee blade can play an important role since they can influence the angle of attack. The squeegee can be positioned at a certain angle with respect to the screen. However, applying a downward pressure onto the squeegee can result in the bending of the squeegee blade, thereby altering the actual angle at the surface, as depicted in Fig. 5.6(b). This effect can further be influenced by the choice of the squeegee edge profile. Some examples of different profiles are depicted in Fig. 5.6(c). Besides these intentionally chosen influences, the condition of the squeegee blade is also very important. A well-formed squeegee blade will help to control more precisely the quality of the final print, while a damaged blade can result in, for example, streaky prints. Figure 5.7 depicts the basics of the screen printing process. The screen, attached onto a frame, is placed above the substrate at a certain offset, the snapoff distance. A printing solution or ink is deposited onto this screen in front of
Figure 5.6 (a) Profile of a typical squeegee; (b) effect of a downward pressure on the angle between the squeegee blade and the screen; (c) examples of some different squeegee edge profiles.
Processing and Production of Large Modules
237
Figure 5.7 Basic scheme of the screen printing process.
the squeegee. By moving the squeegee, the ink is spread over the screen. Applying sufficient pressure onto the squeegee, it deflects the screen downward to make contact with the substrate. The ink is then forced onto the substrate through the open areas of the screen not filled by the emulsion coating. As the squeegee passes a given point, screen fabric tension snaps the screen back, leaving the ink behind. 5.1.3 Pad printing The pad printing method shown in Fig. 5.8 is, like screen printing, highly compatible with making a single print (batch printing). The main advantage is that printing on corrugated or curved surfaces is possible. Also, the ink requirement is a lower viscosity than that required for screen printing. It is also possible to use volatile solvents since the ink is never exposed for long time, thus drying is not a problem. This can be an advantage compared to screen printing where drying in the mesh can be problematic. The technique involves a cliché with a gravure (etching) that corresponds to the pattern that is to be printed. The gravure on the cliché is filled by a doctor blade or a sweeping process with the ink and the ink is subsequently transferred to a soft stamp, which is typically a silicone rubber. The stamp is then pressed onto the substrate and the motif is transferred. The gravure can also be a roll and the stamp can be a roll giving full R2R compatibility, where very fast feed speeds are possible (1–10 m s−1 ). 5.1.4 Doctor blading The doctor blade technique is excellently suited for preparation of small prototypes on the laboratory scale. The strength of the technique is that a relatively large area can be coated evenly with a precise thickness of ink. The method relies on fixing the substrate on a planar surface (a glass plate). A steel carrier with a comblike structure is drawn at constant speed over the substrate (1–100 mm s−1 ). Ink is placed in front of the comb and as it is drawn across the surface of the substrate, an even layer is applied between the teeth in the comb. The depth of
238
Chapter 5
the comb defines the thickness of the film and is typically on the order of micrometers. Unlike the techniques mentioned above, doctor blading can only make one-dimensional patterns (i.e., stripes). A typical doctor blading machine is shown in Fig. 5.9.
Figure 5.8 A picture of a pad printing machine (LPE 70 from Pad Print Machinery in Vermont) and picture of the pad, the gravure plate, and the cup holding the coating solution (right).
Figure 5.9 A common setup for doctor blading. Shown is the Erichsen Coatmaster 509 MC-I.
Processing and Production of Large Modules
239
5.1.5 Other printing methods The above presentation was just a brief overview of the available printing techniques or coating techniques that are particularly suited to solution-based processes for making OPVs. Many of the techniques are very closely related and differ only in the nature of the ink that is used or the way that the motif is transferred. Gravure, flexographic, lithographic, letterpress, plateless, and ink-jet printing are just a few of the other available techniques. The general choice of printing method in reality depends on the nature of the substrate.
5.2 Printing the Active Layer All the techniques mentioned above have been explored extensively for many years and are all well-proven, established industrial techniques. When it comes to printing the active layer of OPVs, which is most often based on a polymer-fullerene mixture, few examples have been reported and most literature reports have employed spin coating of the active layer, while a few reports have employed doctor blading and screen printing. 5.2.1 Screen printing The variables that affect the screen printing process can be divided into several main areas such as machine setup, substrate and environment, screen manufacture, squeegee, ink paste, and screen.27 However, for each of these areas many particular parameters can be defined such that the final result would be a very extensive list of possible print conditions, as depicted in Fig. 5.10. This fish-bone diagram also indicates that the printing process is determined by completely different parameters with respect to those discussed before for the spin-coating technique. In the experiments described further on, we restrict the discussion to certain variables such as machine setup, ink paste, and screen. Concerning the machine setup, we mainly focus on the influence of the printing speed, the snap-off distance, and the squeegee pressure on the printing results. Screens with different mesh sizes are studied as well as ink pastes with different compositions.
Figure 5.10 Diagram representing the many parameters in the screen printing process.
240
Chapter 5
In the experiments conducted, the printing was always carried out using a manually operated SP002-R screen printer (ESSEMTEC, Switzerland) with a polyurethane squeegee, since this is the most resistant against the solvents used in the process examined here. With a rectangular shape, a shore hardness of 85, and an angle set at 45 deg, it can be assumed that the angle of attack is kept constant during the process and similar for different applied processing conditions. This should result in an even balancing of the pressure over the full contacting surface between the squeegee and the screen. Polyester screens were commercially purchased (KOENEN Gmbh, Germany), with an emulsion coating and adhered to an aluminum frame processed at their manufacture facilities. The printing itself is done in an ambient atmosphere. 5.2.1.1 Process optimization
Single layers from solutions containing only the donor material are processed first. The conjugated polymer MEH-PPV is used an results in the fabrication and characterization of a light-emitting device. Prior to this, the influence of different printing parameters on the quality and thickness of the deposited layer was investigated. 5.2.1.2 Printing speed
The speed of the squeegee’s movement was studied as a first printing parameter since it turned out that it could have a serious influence on the quality of the deposited layer. Therefore, MEH-PPV solutions were printed onto glass/ITO substrates under different printing speeds. Since the printer was operated manually, it was not obvious how to quantify the printing speed. (A more elaborate study is presented at a later stage.) For now, the effect is discussed using digital scans of the deposited layers. Such photographs of MEH-PPV layers printed at different squeegee speeds are given in Fig. 5.11, where the scans have an actual width of 0.8 cm. With a resolution of 200 pixels per line, this results in pixel sizes of 40 μm × 40 μm.
Figure 5.11 Digital scans of MEH-PPV films printed at (a) low speed, (b) medium speed, and (c) high speed of the squeegee’s movement.
Processing and Production of Large Modules
241
The film after printing with low speed of the squeegee movement is nonhomogeneous with very uneven spreading of the material over the substrate. Areas with almost no material present are contrasting with places of thick dropletlike zones. It is important to note that the dimensions of these areas do not at all correspond with any significant dimension of the applied screens. The wire diameter as well as the mesh opening of the fabric is on the order of some tens of microns. Therefore, the observed film morphology is at first instance clearly not determined by the choice of screen parameters. Furthermore, visual inspections as well as profilometric measurements clarify that the different color intensities in the layer can be attributed to variations in the thickness. So, in the lighter areas almost no material is deposited, while it is strongly concentrated in the darker dropletlike zones. When the squeegee movement is increased to a medium speed level, the morphology of the deposited layer is altered, as clarified by Fig. 5.11(b). Still, an inhomogeneous spreading of the material can be observed. However, the dropletlike features are present but a two-phase deposition occurs. This phenomenon is often called mottle. Areas with lighter color and therefore less material alternate with darker zones. The contrast is, however, not as large as in the previously processed layer. Also, the darker areas are more largely spread over the full surface. On the other hand, when a sufficiently high printing speed is applied, a much more homogeneous film is formed [see Fig. 5.11(c)]. A uniform color occurs for the whole printed surface. This indicates an even coverage of the substrate and good spreading of the material. A more quantitative study of the digital scans can be obtained by constructing color histograms. This is done by attributing a color code to each pixel in the scan. The color histogram then displays on the y-axis a distribution that gives the number of pixels that have a certain color code attributed and displayed along the x-axis. Such a distribution can then be quantified by values such as its mean or median color code and its standard deviation. These can then be used as criteria to define the quality of the film. Since a low standard deviation means the color codes of all pixels of the scan are approximately the same, this would indeed indicate that a homogeneous film was printed. This procedure is now performed onto the scans of Fig. 5.11, using the existing software in Corel PHOTO-PAINT. Here, the intensity of the color in a red-greenblue color scheme is chosen. The attributed numbers range between 0 and 255, with the lowest value for completely dark (black) pixels and oppositely the highest number for white pixels. The resulting color histograms are given in Fig. 5.12; the mean value and the standard deviation of the distributions in each histogram are discussed below. The histogram in Fig. 5.12(a) for the film printed at low squeegee speed displays a wide distribution with a standard deviation as large as 24, indicating a broad range of pixel colors. More specifically, it shows in this case that there is a certain
242
Chapter 5
Figure 5.12 Color histograms of the films shown in Fig. 5.11 printed at (a) low speed, (b) medium speed, and (c) high speed of the squeegee’s movement.
distribution around a rather high color-code value with a tail extending into the lower values. This confirms the observation that in most areas a thin layer of material is deposited, resulting in a bright, almost white color, whereas most material is concentrated in smaller, darker spots. The resulting mean value is therefore as high as 227. In the case of the film deposited with a medium printing speed, the spread of the pixel colors is already much narrower, as shown in Fig. 5.12(b). This indicates that a more uniform material deposition is obtained here, which is no longer dominated by large areas with only a thin film formed. This is confirmed by a lower mean value of 197. The obtained standard deviation of 9 is still rather high. This can be attributed to the specific feature of the distribution, clearly displaying two large, separated peaks. They are almost equal in size with a separation that is not too large, indicating the appearance of two basic color distributions that are rather similar. Therefore, an almost perfectly homogeneous layer formation is represented by this histogram. The histogram in Fig. 5.12(c) for the film printed at high speed shows a very narrow distribution with a standard deviation of only 1. Thus, a high-quality, uniform film is deposited under these circumstances. It can be noted that the mean value here is 189. This corresponds very well to the position of the lower peak in the distribution for the previous film, printed at medium speed. This indicates that in that situation a large portion of the film has a quality that is almost equal to the one obtained at high printing speeds. The fact that the two peaks in the previous histogram are very similar in size reveals that already up to half of the layer printed at medium speed has acquired this high quality. Thus, the introduced method of applying color histograms has shown that it offers a way to have a more quantitative comparison of the quality of the printed films. The obtained results clearly correspond with the general observations made on the material distribution. In the cases studied here, it confirms and quantifies that an increased printing speed results in much more homogeneously deposited layers. A more elaborate study also involving color histograms will come up again at a later stage. For now it has to be specified that the results described here are for 1% concentrated MEH-PPV solutions. Similar data have been produced for more concentrated
Processing and Production of Large Modules
243
solutions, indicating the general importance of the applied printing speed. Therefore, all subsequent results are obtained by printing at high speed of the squeegee movement. The influence of some other printing parameters on the formation of the organic layer is studied; the results are described in subsequent sections and some trends clarified. 5.2.1.3 Mesh size
Screens with different mesh sizes were tested, varying from 120 over 180 to 190 threads per cm. The characteristics of the screens were chosen so that the theoretical paste volume Vth decreases from 18 cm3 m−2 over 9 to 5 cm3 m−2 , respectively. Figure 5.13 depicts the final film thickness obtained for layers printed with these different screens. The thickness of the deposited layers was measured with a Dektak profilometer. Results are given for 1% and 2% concentrated MEH-PPV solutions. Quite generally, a trend of increasing film thickness with higher theoretical paste volume of the screens is observed.28 In other words, the amount of material that can be pushed through the screen openings correlates with the final layer thickness. However, the increase in film thickness is not as large as would be expected for the given theoretical paste volumes. Between the lowest and highest Vth values, a difference of more than a factor of three is seen, which is not observed in the measured layer thicknesses. Please note that the data present average values from a large variety of printer settings for each of the screens, which could partly explain the discrepancy. More likely, the results show that the theoretical paste volume is only, to a very first approximation, an indication for the obtained layer thickness. Other effects, for example, the method of applying the emulsion layer as mentioned before, seem to also play an important role. On the other hand, it is shown that the layer thickness relates quite well to the polymer concentration in
Figure 5.13 The average film thickness as a function of the theoretical paste volume of the different screens, for the different MEH-PPV solution concentrations.
244
Chapter 5
the solutions used.28 An increase of a factor of two for the solution concentration seems to result in a similar increase of the deposited layer thickness. For the 2% polymer solution, an average thickness on the order of 100 nm can already be obtained, which is within the range commonly applied to organic devices. The layer’s thickness can still be altered by other parameters. 5.2.1.4 Squeegee pressure and snap-off distance
Here, the effect of the downward squeegee pressure and of the snap-off distance is studied. Figure 5.14 shows the variation of the final film thickness with these parameters. It can be seen that by increasing pressure on the squeegee, this final film thickness also increases. In the used printer setup, the pressure on the squeegee can only be changed by a spring mechanism. Therefore, the displayed values on the x-axis are arbitrary. The observed relationship is most likely related to pressing more of the solution through the openings of the screen when higher pressure is applied, resulting in the formation of thicker films Furthermore, the graph also shows that for increased snap-off distances the final layer thickness also increases. The variation of the film thickness with snapoff distance is, however, rather limited. Furthermore, the graph also shows that it holds only to a certain extent, because when the distance becomes too large the screen would no longer be touching the substrate surface and no material would be transferred at all. A possible explanation for the observed relation is that the screen is simply stretched more for higher snap-off distances. The actual mesh openings are thereby enlarged, allowing more material to flow through the screen. 5.2.1.5 Evaluation of the printed film as an active layer in a light-emitting diode
It is shown that print parameters such as the printing speed, the snap-off distance, and the squeegee pressure have a clear influence on the formation, homogeneity, and thickness of the deposited film. Furthermore, the mesh size of the screen as
Figure 5.14 The influence of the squeegee’s pressure on the final thickness of the deposited film for several snap-off distances.
Processing and Production of Large Modules
245
well as the ink formulation also affect the print result. In accordance with the previous experiments to study the different materials, inks, and printer settings, we will fabricate an electronic device with a screen printed active layer based on a solution of pure MEH-PPV. The resulting device, therefore, is a light-emitting diode (LED)29, 30 and is also characterized as such.31 The ITO film on the glass substrate was patterned beforehand by UV photolithography. Prior to printing the polymer solution, a thin PEDOT:PSS layer was spin coated on this patterned substrate. To finalize the device structure, a metallic back-side contact of aluminium was evaporated in high vacuum on top of the active polymer film through a shadow mask. The active device area amounts to 0.09 cm2 . Polymer solutions with either a 1% concentration or a 2% concentration of MEHPPV in chlorobenzene were printed as the active layer. The printing conditions were such that in both cases a homogeneous layer was deposited with an approximate thickness of 100 nm for the 2% and only half of that for the 1% concentrated solution. IV characteristics are determined with the standard measurement setup described earlier. The light output is measured with a Si photodiode. A typical current-voltage-luminance (I-V-L) curve for the device based on the 2% solution is presented in Fig. 5.15(a). Clear diode behavior is observed with a rectification ratio of almost 103 at 2 V. Light output begining at 2 V with the forward-applied bias. The external quantum efficiency (EQE), as the ratio between the number of photons emitted by the device over the amount of charges injected into it, is calculated for devices based on 1% and 2% polymer solutions, and the results are given in Fig. 5.15(b). There is clearly a difference when 2% concentration ink is used instead of 1%. The maximum EQE for the latter is only 1.2 × 10−2 %, whereas for the 2% concentration paste it reaches over 4 × 10−2 %. This change in performance might be related to the different film thicknesses28 obtained for these solutions, as mentioned earlier.
Figure 5.15 (a) I-V-L curve for an LED with screen printed MEH-PPV emissive layer from a 2% concentrated MEH-PPV solution; (b) external quantum efficiency of organic LEDs with screen printed active layer of different concentration MEH-PPV solutions.
246
Chapter 5
Figure 5.16 Picture of an operational LED with a screen printed active layer of MEH-PPV.
On the basis of these results, another LED with a screen printed active layer of MEH-PPV was processed. In this case, the ITO layer was patterned to have a buss bar connected to the letters “IMEC.” On top of it a thin layer of PEDOT was spin coated and the active layer was printed. Finally, an aluminum back-side contact was deposited by vacuum evaporation. A picture of the resulting device is given in Fig. 5.16. There, the light output is shown at an applied forward bias of approximately 10 V. The total width of the displayed word is around 4 cm. This offers a glance at the possibilities of the printing technique in the fabrication of fully operational optoelectronic devices based on organic materials. At this time, no optimization of the performance of the LED has been addressed. Improved performance can be expected, for example, from a more appropriate choice of backside contacts and charge transport layers.28 Device optimization will focus later on photovoltaic cells and modules. 5.2.1.6 Rheological characterization
It is clear that in printing processes the general flow behavior of the ink has an influence on the quality of the deposited layers.32–39 This urges us to quantify some of the rheological properties of the polymer-based solutions that we are applying in the screen printing process. In rheology, the deformation and the flow of materials are studied.40–42 More specifically, one tries to relate the deformation or strain with the stress that is present in a material when an external force is applied onto it. A common way to induce a deformation in a liquidlike solution is by shearing it. Placing the material in between two parallel plates, as illustrated in Fig. 5.17, can do this. Force K now displaces the upper plate so that it moves parallel to the fixed lower plate with
Processing and Production of Large Modules
247
Figure 5.17 Shearing of a material between two parallel plates.
velocity v. A shearing stress τ is induced to resist this force as τ=
K , A
(5.3)
where A is the area of the material that is sheared. The resulting deformation throughout the material can then be envisaged as a movement of different layers parallel to each other, but with different rates. The displacement with respect to its position above the fixed lower plate is a measure for the induced strain. The shear γ is therefore, in the case of Fig. 5.17, expressed as follows: X , h
(5.4)
dγ v = . dt h
(5.5)
γ = tg α = so that the shear rate γ yields γ =
Since viscosity η is now the measure of the resistance of the liquid to flow, it can be defined as the ratio of the shearing stress τ to the shear rate γ : η=
τ . γ
(5.6)
For an ideal viscous material, the viscosity is constant regardless of the rate of shearing or shear stress. This is called a Newtonian fluid. A single measurement of the viscosity is therefore enough to quantify the rheological behavior of such materials. However, many liquids such as suspensions, emulsions, and polymer solutions display non-Newtonian behavior. If we subject such materials to a flow experiment, it will turn out that their viscosity is a function of the shear rate. The result is no longer a single-material constant but a material function, as depicted in Fig. 5.18. A graph that displays the relationship between shear stress and the shear rate is called a flow curve [see Fig. 5.18(a)]. For a Newtonian fluid, this results in a straight line with a constant slope that indicates the viscosity of the fluid. Therefore, in a viscosity curve, a constant value is displayed regardless of the shear rate [see Fig. 5.18(b)]. Many materials, however, have a nonlinear relation between the shear stress and shear rate. Mostly pseudo-plasticity or shear thinning is observed with
Chapter 5
248
Figure 5.18 Some general cases of rheological behavior represented by (a) flow curves and (b) viscosity curves.
Figure 5.19 General viscosity curve for polymer-based solutions.
a decreasing viscosity when the shear rate increases. Also, flow behavior with an increased viscosity for increasing shear rate can be observed in some cases. Such a shear-thickening relation is also called dilatancy. It also has to be mentioned that many materials only start to flow if a certain critical amount of stress is applied onto them. This is denoted with an apparent yield stress and results in a flow curve that starts at a stress value differing from zero. It can occur for all of the previous flow behaviors, and the material is then named a Bingham fluid. When measuring in a sufficiently large shear rate range, three different areas can generally be distinguished for a polymer-based solution, as shown in Fig. 5.19. At first a so-called Newtonian plateau can be observed. The polymer chains in the solution are randomly oriented and therefore the resistance against flow is maximal at these low shear rates. A maximal zero-shear viscosity is measured. On increasing the shear rate, polymer chains can start to orient themselves along the flow direction. At a certain point, a critical shear rate is reached where this behav-
Processing and Production of Large Modules
249
ior occurs. The viscosity decreases and a pseudo-plastic flow regime appears. The critical shear rate is dependent on many solution parameters, such as the polymer material itself, the solvent, the concentration, etc. Again, at a further increase of the shear rate, the polymer reaches its maximal orientation along the flow direction. The viscosity can therefore no longer decrease. A second Newtonian plateau appears with a constant viscosity. The shear-thinning behavior of a fluid is often described by an empirical expression. The most commonly used relation between shear stress τ and shear rate γ is the power law equation: τ = kγn
with n < 1,
(5.7)
where k and n are parameters characteristic to the material system. Combined with Eq. (5.6), this shows that the viscosity η varies with shear rate as: η=
τ = kγn−1 = kγm , γ
with m = n − 1 < 0.
(5.8)
This equation indeed indicates a decrease of the viscosity with increased shear rate. It is therefore suitable to characterize the pseudo-plastic flow regime given in Fig. 5.19 in between the two Newtonian plateaus. For n = 1, the relation would reduce to Newtonian fluid behavior, whereas the power law description can also be used for shear-thickening flows when n > 1. Another feature of polymer solutions that can also be related to the internal structure and chain orientation is the time-dependent behavior. If a large stress is applied, the solution flows at a low viscosity. When the stressing stops, the fluid regains its initial internal structure. This behavior is called thixotropy and is characterized by the time scale at which the orientation of chains is induced or relieved. Some types of materials display the opposite, that is, the internal resistance against flow and the viscosity increase when applying a high stress. This is then called rheopexy. 5.2.1.7 Rheology measurement setup
Rheological measurements are carried out at University Hasselt (B) using a CarriMed rheometer (type CSL2 500). A schematic representation of the basic elements is given in Fig. 5.20. The fluid that has to be examined is placed on a flat surface. This part of the rheometer remains steady during measurements. The moving part consists of a conelike geometer that is positioned on top of the fluid solution. The amount of liquid has to be such that with the chosen gap, the space between the rotating cone and the steady plate is nicely filled. An important advantage of this cone/plate geometry is the realization of a shear rate that is uniformly distributed over the solution. Cones with different angles and diameters are available. The one used for the measurements described later has an angle of 2 deg and a diameter of 4 cm.
250
Chapter 5
Figure 5.20 Schematic representation of the basic elements of the rheological measurement setup.
It is made of stainless steel to be resistant to the chlorobenzene solvent. Furthermore, it is a relatively heavy material, resulting in higher stability of the system during measurements, and it has minimal expansion in the examined temperature range. Heating and cooling of the liquid is controlled by a Peltier element in the lower plate. A temperature range from −10◦ C to 99◦ C is thereby easily accessible. Since solvent evaporation from the liquid solution could occur during the time of measurement, a solvent trap is placed over the cone. A small cavity on top of the cone is filled by a certain amount of pure solvent and a cap is then placed over it. This has an edge that enters the solvent and the space under the cap gets saturated with solvent vapor. As such, evaporation of solvent from the solution to be measured is prevented. The liquid sample is now studied by applying a known stress onto it by rotating the cone. Subsequently, the resulting shear rate is measured. 5.2.1.8 Temperature-dependent viscosity measurements
We have examined solutions with different concentrations of the MEH-PPV polymer as well as a blend of the donor and acceptor material in a 1:4 weight ratio. The solvent was chosen to be, in all cases, chlorobenzene. Figure 5.21(a) depicts the results of rheological measurements for such different solutions at a constant shear rate of 100 s−1 , but whereby the temperature was gradually increased over a certain range. The measurements were each time repeated with a decreasing temperature profile. High reproducibility of the data is seen for all examined solutions. Generally, it can be observed that the viscosity increases when the temperature is lowered. A gradual change with the temperature occurs and at no point is a strong increase seen. This indicates that no gelation phenomena are appearing in this temperature range for any of these solutions. Taking the three uppermost curves of Fig. 5.21(a) into account, the variation of the viscosity for different polymer concentrations is shown. The 1% concentrated
Processing and Production of Large Modules
251
Figure 5.21 Viscosity measurements at a constant shear rate of 100 s−1 for MEH-PPV solutions with different concentrations and for an MEH-PPV:PCBM blend in a 1:4 ratio as function of (a) temperature and (b) the inverse of the absolute temperature.
MEH-PPV solution has a viscosity ranging from 7 × 10−2 Pa s at 10◦ C to 4 × 10−2 Pa s at 55◦ C. These values triple when the concentration is doubled and even increase almost 12 times for 3% concentrated solutions. It was not possible to study the behavior of higher concentrated MEH-PPV solutions due to limited solubility of the polymer in chlorobenzene. The lower curve in Fig. 5.21(a) shows the behavior of a 1% polymer solution to which PCBM is added in a 1:4 ratio by weight. What is clearly remarkable here is that the overall viscosity of the mixed solution is lower than the one obtained for the pure 1% MEH-PPV solution. The values vary from 3 × 10−2 Pa s at 10◦ C to 1.6 × 10−2 Pa s at 55◦ C for the solution containing PCBM. So, the viscosity decreases despite the addition of solid material. In Fig. 5.21(b), the viscosity data are represented as a function of the inverse absolute temperature T , thereby displaying a typical Arrhenius-like behavior as
E0 η = A exp RT
T0 =A , T
(5.9)
where R is the universal gas constant, A is a constant characteristic for the examined material system, E0 is the activation energy, and T0 is the corresponding activation temperature. The activation temperatures obtained are all very similar with values around 1000 K. Therefore, a substantial addition of the acceptor material PCBM does not seem to alter the temperature-dependent viscosity behavior of the given solution. The viscosity values obtained at a temperature of 20◦ C for the three different polymer concentrations are displayed again in Fig. 5.22. The full line is a fit of these data with a power-law relation, indicating that the viscosity increases with a
Chapter 5
252
Figure 5.22 Viscosity data for different concentrations of MEH-PPV, determined at a temperature of 20◦ C with a shear rate of 100 s−1 .
polymer concentration as follows: η ∝ concα ,
where α = 3.
(5.10)
We can now use this relation the other way around to position the viscosity value of the MEH-PPV:PCBM solution obtained at the same measurement conditions (20◦ C, shear rate of 100 s−1 ). This would then correspond to a solution concentration of 0.6%. Thus, the addition of PCBM to a 1% concentrated MEH-PPV solution corresponds to the situation of diluting the polymer solution to a concentration of only 0.6%. The temperature-dependent flow behavior is thereby not substantially altered. The general decrease of the viscosity despite the addition of solid material is examined further in the next section. 5.2.1.9 Shear rate–dependent viscosity measurements
Figure 5.23(a) shows the result of a measurement on the 1% MEH-PPV polymer solution whereby the shear rate was varied. This was done for different temperatures ranging from 10◦ C to 55◦ C. It can again be observed that the viscosity increases when the temperature is lowered. Furthermore, there is a clear dependence of the viscosity on the shear rate in such a way that it decreases for higher shear rates. So, for this polymer solution, we are in the shear-thinning regime. Neither for the low shear rates nor for the higher ones is any occurrence of a Newtonian plateau indicated. A similar measurement is performed on the 1% concentrated solution after the addition of PCBM in a 1:4 weight ratio. The result is given in Fig. 5.23(b), showing also in this case the same temperature dependence as measured before. If we compare the values of the blend with those obtained for the pure polymer solution, we see that the viscosity is lower for the blend, under similar measurement conditions.
Processing and Production of Large Modules
253
Figure 5.23 Viscosity as function of shear rate for different temperatures for (a) 1% MEHPPV solution and (b) 1% MEH-PPV solution with additional PCBM in a 1:4 weight ratio.
This confirms the observation of the previous measurement. Furthermore, the data indicate that for this blended solution we are also in the shear-thinning regime with no appearance of Newtonian plateaus. Because PCBM is a rather small molecule with respect to the long polymer chains of MEH-PPV, it appears to have a lubricating effect on the pure solution. In this way the resistance against flow is reduced, leading to a decrease of viscosity. Nevertheless, the observed dependence of the viscosity on the shear rate indicates that there is still substantial interaction between the polymer chains, influencing the flow behavior. So, the addition of PCBM decreases the viscosity of the solution, but the overall pseudo-plastic flow behavior remains. This is further quantified by fitting the measurement data according to the empirical relation. The full lines in Fig. 5.23 represent the results of such calculations. For the temperature range examined here, values for the power coefficient n are varying around 0.7 for the pure polymer solution, whereas they are on the order of 0.8 for the MEH-PPV:PCBM blend. So, in both cases clear shear-thinning behavior is identified on a rather similar order. The addition of a substantial amount of PCBM to the polymer solution alters the behavior only slightly to a somewhat more Newtonian flow. To sum up, temperature-dependent as well as shear rate–dependent viscosity measurements have been performed. The results indicate that the addition of PCBM in a 1:4 ratio to a 1% concentrated MEH-PPV solution alters the rheology of the polymer solution only to a limited extent. Contrary to expectations, since more solid material is present in the solution, the overall viscosity of the blend is lower than the corresponding pure polymer solution. However, the general flow behavior is still shear thinning on a similar order for the polymer solution as for the MEH-PPV:PCBM blend. Therefore, it can be concluded that screen printing con-
254
Chapter 5
ditions for the mixed donor/acceptor solution would be very similar to what was examined earlier for the pure polymer solution. So, applying such screen printed composites as an active layer in photovoltaic devices is studied in the following sections. 5.2.1.10 Photovoltaic devices with screen printed active layer
First, we will elaborate on the quality of the obtained films by screen printing a donor/acceptor blend. Color histograms are used to quantify these results, as described earlier. Especially the effect of the printing speed and the relation to the viscosity measurements are examined. Following these findings, photovoltaic devices with screen printed active layers are fabricated and characterized. Glass as well as flexible substrates are used and a first module concept is studied. 5.2.1.11 Screen printed layers of donor/acceptor blends
Similar to what was done in the rheological study, chlorobenzene-based solutions of pure MEH-PPV in different concentrations (here 1% and 2%) and a blend of MEH-PPV:PCBM (1:4) are tested in the screen printing procedure. To study the effect of the printing speed, a semiautomatic screen printer (AT701, Alraun Technik, Germany) was used. Films of the different polymer and blended solutions mentioned earlier were printed with speeds varying from 150 mm s−1 to 550 mm s−1 . They were then characterized with color histograms as described above. The resulting mean and standard deviation values are depicted in Fig. 5.24(a) and (b), respectively. Since the mean values generally indicate the dominant intensity, it is difficult to compare these results with each other for the different compositions that are tested. However, interesting observations can still be made. For all compositions, the mean value decreased if a higher printing speed is applied. This corresponds with earlier trends reported for the manually printed films. This indicates that the overall picture becomes darker when printed at higher speed, since the attributed numbers are lower for darker pixels and higher for lighter ones. At this time it is, however, not really clear what causes this trend. Furthermore, the trend is more pronounced for some compositions than for others, which is visualized by the slope of the curves in Fig. 5.24(a). This shows that the variation with speed is limited for the 2% concentrated MEH-PPV solution, while it is already much stronger for the 1% concentrated polymer solution. But it is clearly most present for the blend of MEH-PPV:PCBM. Interestingly, the variation of the slope therefore becomes stronger if the viscosity of the solution decreases. This can be seen by comparing the described observation with the rheological results obtained earlier and depicted in Figs. 5.21 and 5.23. Further insights on such relations are deduced from the standard deviation values.
Processing and Production of Large Modules
255
Figure 5.24 (a) Mean and (b) standard deviation values extracted from color histograms of screen printed films of 1% and 2% concentrated MEH-PPV solutions and of an MEHPPV:PCBM (1:4) blend.
In Fig. 5.24(b), the standard deviation of the color histograms for different polymers and the blended solutions is depicted. It shows that in all cases the standard deviation decreases clearly with an increased printing speed. This indicates that indeed an improved quality of the films can be obtained with faster printing, as described earlier, because a low standard deviation means that a homogeneous spreading of the material occurs. Moreover, the graph shows that the lowest values are obtained for the 2% concentrated MEH-PPV solution, whereas they become larger for the 1% polymer solution. They are highest for the blended MEH-PPV:PCBM solution. Therefore, comparing this with the results depicted in Figs. 5.21 and 5.23, again a trend occurs that for less viscous solutions higher standard deviation values are obtained. This indicates that printing speeds become less important for more viscous solutions. Nevertheless, for the low-viscous MEH-PPV:PCBM blend low standard deviation values are obtained if a sufficiently high printing speed is applied. If the viscosity of the solution expresses its cohesion, then the observed behavior shows that the film formation might be strongly influenced by the interplay between these cohesion forces and the adhesion of the solution to the screen. The adhesion effect results in inhomogeneous deposition of the solution by a phenomenon called film splitting. This is depicted in Fig. 5.25, and it resembles what generally happens when trying to separate two sheets that are held together by a fluid in between. Also, the fluid separates during this process, thereby spreading unevenly over the two different sheets. If adhesion of the fluid onto one of the sheets is less than onto the other, film splitting can be prevented when the process happens fast enough. In the screen printing procedure, the interaction of the solution with the screen can be substantially smaller than with the substrate due to
256
Chapter 5
Figure 5.25 Graphical representation of layer formation for a rather low-viscous solution applied in the screen printing process.
the holes in the screen. When applying a higher printing speed, the screen snaps off faster so that adhesion of the solution onto the screen can be prevented. Therefore, a higher printing speed overcomes this effect and results in better film quality. However, for less viscous solutions it is necessary to apply a larger printing speed. Although we have seen before that the flow behavior of the polymer solution is not substantially altered due to the addition of PCBM, the overall lowering of viscosity forces us to adjust the printing procedure. An increased printing speed is necessary to obtain blended MEH-PPV:PCBM films with a sufficiently high quality so that they can be used as an active layer in photovoltaic devices. Subsequent results show that this indeed can be achieved. 5.2.1.12 A screen printed photovoltaic module based on MEH-PPV-[60]PCBM
In accordance with the above, we have fabricated photovoltaic devices with a screen printed active layer. We used a donor/acceptor blend based on a 1% MEH-PPV solution with additional PCBM in a 1:4 weight ratio. The printing of this solution was done with a manually operated machine. At first, devices were assembled onto glass substrates coated with ITO that was patterned according to the standard UV-photolithography procedure. Prior to the printing of the polymer solution, a thin PEDOT:PSS layer was spin coated on these patterned substrates. To finalize the device structure, a metallic back-side contact of LiF/Al was evaporated in high vacuum through a shadow mask. The IV characteristic of such a device is presented in Fig. 5.26(a), under standardized simulated solar illumination (AM 1.5, 100 mW cm−2 ). A short-circuit current density Isc of 3.4 mA cm−2 , an open-circuit voltage Voc equal to 845 mV, and a fill factor FF of 44% are obtained. This results in an overall energy conversion
Processing and Production of Large Modules
257
Figure 5.26 IV characteristics of photovoltaic devices with a screen printed active layer of MEH-PPV:PCBM (1:4) on (a) a glass substrate and (b) a flexible PET substrate (AM 1.5 simulated solar illumination, 100 mW cm−2 ).
efficiency of 1.25%. These results are very comparable to what was achieved beforehand for devices with a spin-coated active layer. Especially the high Voc value points out that no losses due to shunting occur in the device here. This indicates that a high-quality active layer is produced by the screen printing technique without pinholes that could cause direct contact of the evaporated top layer with the bottom electrode. A similar processing is tested on flexible substrates. PET flexible foils coated with ITO have been commercially purchased from Vitex Systems (U.S.). High optical transparency occurs for these sheets but the resistivity of the ITO coating is substantially higher than what is achieved for the glass substrates. Sheet resistance on the order of 60 Ω square−1 is claimed for these foils. The patterning of the ITO is done by a UV-photolithography procedure very similar to that used for the glass substrates. Care had to be taken when high temperature steps were involved. Also, the oxygen plasma treatment had to be reduced in time to only ∼1 min. A thin PEDOT:PSS layer was spin coated onto the substrates prior to the printing of the active layer. The device structure was finalized by the evaporation of LiF/Al through a shadow mask. The IV characteristic obtained under standardized simulated solar illumination (AM 1.5, 100 mW cm−2 ) is depicted in Fig. 5.26(b). Again, a high open-circuit value is measured, pointing to the good quality of the printed active layer. This indicates that the studied processing technique for the active layer can easily be applied to different substrate materials. An Isc of 2.4 mA cm−2 is obtained here, which is somewhat lower than what was measured for the structure on glass [Fig. 5.26(a)]. Also, a diminished value of 0.31 for the FF is seen here. From the IV curves, it is
258
Chapter 5
quite obvious that these changes can be attributed to a substantially increased series resistance for the device on PET. As mentioned before, this can be due to a higher sheet resistance of the ITO on this foil. Although, another effect most likely has a more pronounced influence on the series resistance. It is observed that the metallic back-side contact displayed a white shade instead of being completely shiny. This indicated an imperfect surface with high roughness. It can possibly be attributed to some thermal stress from which the PET substrate suffered during the vacuum evaporation of the metallic back-side contact. Unfortunately, it led to an increased resistivity of this back-side contact, which was already encountered when trying to contact this electrode for electrical characterization. Therefore, all this resulted in a limited conversion efficiency of 0.65% for this device with a screen printed active layer on a PET-foil substrate. Nevertheless, a first attempt was made to process a photovoltaic module on a flexible PET-foil substrate. Similar processing of the substrate and for the deposition of the subsequent layers was done as described earlier; however, the masking was altered to obtain a device structure as depicted in Fig. 5.27(a). It consists of long ITO line positioned alternately either more to the left or to the right on the substrate. PEDOT:PSS is spin coated on top of that and a square pattern of the active material is subsequently printed. Long metal electrodes are evaporated to finalize the device, similar to but with an alternating positioning opposite of the ITO contacts. Single cells with an active area of ∼2 cm2 are constructed this way. External contacting makes it possible to easily connect neighboring devices in series with each other. The resulting IV characteristics under standardized AM 1.5 simulated solar illumination (100 mW cm−2 ) are depicted in Fig. 5.27(b). First, it
Figure 5.27 (a) Schematic representation of the photovoltaic module structure with four single cells to be connected in series; (b) IV characteristic of a photovoltaic module on a flexible PET-foil substrate with a screen printed active layer of MEH-PPV:PCBM (1:4) (AM 1.5 simulated solar illumination, 100 mW cm−2 ).
Processing and Production of Large Modules
259
shows the result for a single solar cell to which afterward an additional device is repeatedly connected in series. Finally, the output of four interconnected cells on one substrate is displayed. The graphs show that this was successfully achieved since a steady increase of Voc can clearly be observed. Furthermore, Isc also remains rather stable for the interconnected devices, pointing to a good reproducibility of the single device output characteristics. It was explained earlier that a series connection of several photovoltaic devices could only be successfully applied if the current output of the different cells is well matched. The result here indicates that this is achieved for the screen printed active layer deposition over the full substrate. The steady increase of the output voltage when connecting more cells in series results in Voc well above 3 V. This corresponds very well indeed with a fourfold multiplication of the single cell value for the series interconnection of four cells in one module. The Isc of the module structure amounts to ∼2 mA. With a fill factor of ∼0.3, this gives η = 0.24% under the standardized AM 1.5 simulated solar illumination. Despite the limited power-conversion efficiency obtained, the picture in Fig. 5.28 shows that this output is sufficient for common practical applications. A solar cell module fabricated on a flexible PET-foil substrate as described above is connected to a small electronic calculator. Even under bending conditions, enough power is supplied by the flexible module to run the calculator under ambient illumination. Direct patterning of the active layer of an organic bulk heterojunction solar cell is demonstrated. The donor/acceptor blend consisted of the standard materials MEH-PPV and PCBM and was deposited from a solution by the screen printing technique. Thorough rheological characterization was performed for this mater-
Figure 5.28 Picture of an organic solar cell module under bending conditions powering a calculator when illuminated by ambient light.
260
Chapter 5
ial system, thereby pointing to a lubricating effect of the addition of the fullerene derivative to the polymer-based solution. In combination with an optimization of the printing parameters, we were able to fabricate organic solar cells with such a printed active layer in the standard device structure, yielding an energy conversion efficiency of 1.25% under standardized solar illumination. Processing onto flexible substrates resulted in a somewhat lower performance of the photovoltaic device, mainly due to increased resistive effects at the contacts. Nevertheless, it was shown that the introduction of printing technology for the deposition of the organic active layer could result in the direct fabrication of photovoltaic modules. The operation of a small application tool such as a calculator was shown with such a fully flexible organic solar cell module. 5.2.1.13 A larger screen printed photovoltaic module based on MEH-PPV
The advantage of screen printing is that it is easily scaled to motif sizes of a few square meters. One of the first documented examples of the industrial production of a small serially connected module is shown in Figs. 5.29, 5.30, and 5.31.12 The module had three cells in series and the main purpose of the study was to print the cells under industrial conditions. MEH-PPV was printed from a chlorobenzene solution in the ambient atmosphere onto PET substrates with an etched ITO electrode pattern. The ITO was industrial grade and of relatively low conductivity. Screen printed silver connections were printed and cured before printing the MEH-PPV layer. The cells were kept in dark air and completed by evaporation of a thick layer of C60 before evaporation of the aluminum electrodes. The outline of the module is shown in Fig. 5.30. The individual cells measured 3.5 cm × 2 cm and thus had an active area of 7 cm2 . The active area of the module was 21 cm2 , while the total area of the module was 10 cm × 10 cm = 100 cm2 . The geometric fill factor in
Figure 5.29 One of the small screen printers at MEKOPRINT A/S, Denmark, (left) and a view of the screen. The three light yellow rectangles define the printed (active) area of the individual solar cells (right).
Processing and Production of Large Modules
261
Figure 5.30 A small screen printed module with three cells connected in series based on MEH-PPV-C60 . The completed cell from the ITO side (left), from the aluminum side (middle), and the PET-ITO substrate with screen printed silver connections and screen printed MEHPPV (right).
Figure 5.31 A large module where nine of the modules shown in Fig. 5.30 were connected in parallel.
this module is rather low (21%) because most of the total aperture area is spent for interconnecting the individual cells and for spacing between cells. The completed module could be laminated between PET sheets using an office laminator, and while this gave no real barrier to water and oxygen, it did provide a mechanical barrier that allowed for handling the module. The operational lifetime was short (63 hr) and the practical limit to applicability of the module was around 300 hr. This relatively long operational lifetime for a PPV-based device is ascribed to the fact that both the MEH-PPV layer and the C60 layer were very thick. While
262
Chapter 5
the thickness conveyed a good stability, it also gave a much lower efficiency as compared to the cells described in Section 5.2.1.12. Many of these modules (∼100) were produced, and attempts to make an even larger module were made. Nine of the small modules were connected in parallel, giving a module with 27 individual cells. This module produced around 5 mA and 1 V during the first few days, and could do useful work such as turning the propeller on a small electrical motor. The performance of these cells was low; typically, efficiencies of 0.005% were obtained and the best efficiencies were 0.0125%. 5.2.1.14 An even larger screen printed photovoltaic module based on MEH-PPV
The quality of the transparent conductor always puts a limit on the current that can be extracted. For the most efficient polymer photovoltaics, current densities of >10 mA cm−2 are possible. If this current is going to be extracted without significant losses on the scale of square meters, an efficient contacting scheme is required. The obvious method is to connect the individual cells in series, but there may also be requirements for extraction schemes at the individual cell level such as buried contacts or high-conductivity grids (vide supra). The process line at Risø National Laboratory employs a small screen printer from Alraun Technik, as shown in Fig. 5.32. A large module was prepared having a total area of 0.1 m2 .43–45 The substrate was 200 μm PET foils with an etched ITO electrode pattern. The nominal sheet resistivity of the ITO foil was 50 Ω square−1 but in practical terms, it was found to vary considerably over the total area of the module and the local variations in sheet resistivity were large, tending toward higher values. The design of the module aimed at having as high a geometric fill factor as possible while keeping a good margin between cells and at interconnections to ensure that practical work in the laboratory was not hampered by sharp tolerance requirements. It turned out that
Figure 5.32 The AT701 screen printer from Alraun Technik (left) and the MEH-PPV solution being poured onto the screen before printing (right).
Processing and Production of Large Modules
263
geometric fill factors above 70% are hard to attain without great accuracy in aligning the substrate during screen printing and during cathode evaporation. A geometric fill factor of 65% was chosen as a practical solution. The individual cells of the module are shown in Fig. 5.33, also a simple device with the active layer sandwiched between the ITO anode and the aluminum cathode. During later experiments, extra layers were included and both PEDOT:PSS between ITO and the active layer were screen printed, and a sublimed layer of C60 was employed between the active layer and the cathode. The active area of each cell was ∼7 cm2 and the module consisted of 91 such cells connected as serial lines of 13 cells and a parallel connection of 7 of the parallel lines. The current extraction was achieved at opposite ends of the module with screen printed silver buss lines. The total active area of the device was ∼650 cm2 .
Figure 5.33 Side and top views of the individual cells in the module (top) and a side view of the serial connection and a top view of the entire module (below).
264
Chapter 5
The performance of the large modules was only slightly poorer than the example in Section 5.2.1.12. It was found the e-beam evaporation of aluminum gave a poorer device, and thermal evaporation of aluminum was preferred. The device geometries prepared were ITO-MEH-PPV-Al, ITO-MEH-PPV-C60 -Al, ITO-MEHPPV:[60]PCBM-Al, and ITO-MEH-PPV:[60]PCBM-C60 -Al, and the best devices were generally the ones with ITO-MEH-PPV-C60 -Al. Typical module open-circuit voltages for the devices were 5V, and module short-circuit currents of around 1 mA could be obtained.44
Figure 5.34 The module after evaporation still mounted behind the shadow mask (top left). A complete laminated module (top right). A large 1 m2 module obtained by connecting nine of the modules (scientists shown for scale).
Processing and Production of Large Modules
265
In terms of stability, the devices lasted outside for about one week. The modules had a simple lamination of PET as a mechanical protection, but were otherwise unprotected. They were subject to delamination and fast bleaching when subjected to mechanical movement, which underlines the need for efficient encapsulation of OPV modules as shown in Figs. 5.34 and 5.35.
Figure 5.35 An outside test of a module (top left) and the module current and incident sunlight as measured with a pyranometer (top right). A comparison of a delaminated module that has been tested outside with high humidity (lower left) and a module that has been tested under laboratory conditions (lower right).
266
Chapter 5
5.3 Carrier Substrates The carrier substrate is a very important part of the photovoltaic device and while very few substrates have been explored, its importance will become apparent once full-scale industrial production is at hand. It may seem ironic that the cost of the substrate may set the lower cost for OPVs in the very high volume approximation. There are several requirements to the substrate that must be met. The most fundamental is linked to the stability of the OPV and relates to permeation of water and oxygen from the atmosphere. If a rigid substrate can be chosen, one of the better choices is a simple glass substrate because it is transparent, durable, and chemically stable and has a moderately planar surface. It is also impervious to water and oxygen. Since glass is rigid, the choice of back plate would most probably also be rigid. It could be a material such as glass, steel, or aluminum. With such a choice of carrier substrate and back plate, there are no problems in achieving a package that is impervious to any material, and a completely sealed device can be made. The disadvantage is that the material’s cost will always be high and it is difficult to envisage packaging costs below 10 € m−2 . If the choice of substrate is a flexible one, there are many good polymer materials with excellent thermal stability. Examples are poly(ethylene-terephthalate) (PET) and poly(ethylenenaphthalate) (PEN). The disadvantage of these materials is that they, on their own, do not provide any significant barrier toward permeants such as water and oxygen. Both PET and PEN are available with a layer of ITO that can be patterned for use in a module; and while ITO is a significant barrier of both oxygen and water, it does not meet the requirements in terms of OTR and WVTR and mechanical stability (vide supra). The metallic cathode cannot be considered as a significant barrier for either water or oxygen, and the choice of PET or PEN as a carrier substrate implies the need for encapsulation. There are low-permeability barriers and ultrabarriers available that rely on multilayer structures of a polymer and an inorganic oxide such as SiO2 . The claim is that these are truly flexible barriers for electroactive devices that meet the requirements for OTR and WVTR. Examples of companies and organizations that produce ultrabarrier materials are General Atomics,46 VITEX,47 and NOVAPLASMA.48 The latter has been used in a study of the shelf life of polymer photovoltaics. It was shown that when the devices were encapsulated using the ultrabarrier and kept in the dark, very long shelf lives of 6000 hours or more were possible.49
5.4 Anodes and Cathodes The electrical contact between the active layer and the outside world goes through the anode and the cathode. The anode collects the holes and the cathode collects the electrons; traditionally, the anode has been the transparent contact that allows light to enter the device. There are two main reasons for this. First, nearly all the high-conductivity transparent conductors available are oxides that are deposited by processes involving oxygen plasma or high temperatures. Second, the requirement
Processing and Production of Large Modules
267
for a low work function cathode is most easily obtained by evaporating a reactive metal such as aluminum, calcium, or a magnesium alloy that is not transparent. In principle, the oxide could be deposited by sputtering on top of the active layer, but this would bring the active layer into contact with the reactive oxygen plasma, destroying the electronic properties at the surface. Also, it should be possible to coat or print the active layer on top of the cathode metal. Clean metal surfaces of aluminum, calcium, and magnesium are, however, extremely reactive and form an electrically insulating oxide at the surface very quickly, even under glove box conditions. It is simply not practical to apply the active layer on top of the cathode, except in a few instances such as titanium, where oxide is an excellent electronic conductor and suits the purpose of a cathode well.50 These facts of nature imply that the steps taken to build the OPV device start with the transparent oxide followed by the active layer and completed by the cathode. Transparent oxides have been used extensively in liquid crystal displays, and the easiest choice in the laboratory has been to choose commercially available glass substrates with a transparent conductive layer. Indium tin oxide (ITO) is without doubt the most commonly employed transparent conductor, and sheet resistivities as low as 4–8 Ω square−1 are commercially available. ITO is normally not chosen to be in contact with the active layer since its work function is highly variable and depends on the treatments it has been subjected to before spin coating the active layer. In practice, it is simply used as a high-conductivity transparent contact. A layer such as PEDOT:PSS is normally applied between the ITO and the active layer to serve as a hole contact and to provide a much smoother surface than ITO. Another commonly applied barrier layer is LiF51 or a sublimed layer of C60 ,52 which is applied between the active layer and the cathode. The benefit of LiF is to improve electron transport and the fill factor, while C60 has been shown to significantly improve the device’s lifetime when aluminum is used as the cathode. When considering the transparent contact in a high-volume industrial context, ITO is not applicable due to the cost of indium and its lack of abundance on earth. As it stands now, indium makes up more than 50% of the cost of a plastic solar cell module; and from this point of view, ITO is only relevant for laboratory work.
5.5 Processing of the Transparent Front-Side Contact Inorganic transparent conductive oxide (TCO) layers (e.g., ITO) can be commonly processed via different techniques.53 To obtain low resistivity for such layers in combination with a high optical transparency, a good stoichiometric composition is required. For most of these inorganic oxides, this means that processing temperatures generally above 150◦ C are needed. High-quality films of ITO coated on a glass or a ceramic substrate can therefore easily be obtained. Processing on plastic foils such as PET or PEN is, however, limited to temperatures generally below 100◦ C such that the quality of the oxide layer is not optimal. This results in increased surface roughness and reduced conductivity of the coating that are not
268
Chapter 5
desirable for the photovoltaic applications studied here. The detrimental effect of this lower ITO quality has already been observed in the devices described in the previous section. Another drawback concerning TCO layers on plastic substrates is their generally brittle character. The mechanical flexibility of the active layer of organic optoelectronic devices is seen as an interesting advantage over commonly used materials.54 As fully flexible devices, this offers a high potential for their commercialization in many new areas. The need for inorganic coatings might, however, limit this broad application area. Bending the device processed on a plastic substrate could damage these brittle inorganic oxide layers, thereby reducing their electrical properties. So, the performance of the complete device deteriorates considerably under such conditions. This is easily demonstrated by the following experiment. Sheets of PET coated with a 100-nm thin layer of ITO are repeatedly bent. The films are subjected to a curvature with a radius of ∼0.5 cm. At different stages, they are characterized with a measurement of the sheet resistance. Pictures of the films with an optical microscope are shown in Fig. 5.36. Before any curving is applied to the film, no specific features could be observed with the microscope
Figure 5.36 Optical microscope pictures of ITO layers on a flexible PET substrate, (a) before bending; (b) after bending 10 times in two perpendicular directions; (c) after bending 100 times in two perpendicular directions.
Processing and Production of Large Modules
269
[Fig. 5.36(a)]. A sheet resistance of 57 Ω square−1 is measured for the ITO layer at this initial stage. Second, the flexible sheet is bent 10 times in one direction and again 10 times in a direction perpendicular to the first one, with a bending radius of ∼0.5 cm. The picture in Fig. 5.36(c) reveals that this procedure creates many cracks in the oxide layer, in both directions. The sheet resistance increased to a value of 560 Ω square−1 . The procedure is repeated by bending a sheet now 100 times in both perpendicular directions. An even more cracks can be observed in this case, as shown in Fig. 5.36(c). The sheet resistance of the ITO layer after this procedure has a value of 3700 Ω square−1 . These results show that the performance of flexible electronic devices incorporating a conductive oxide layer such as ITO can dramatically deteriorate if bending occurs. It is also considered that bad adhesion of the oxide layer onto the plastic substrate under bending conditions substantially reduces the lifetime of organic optoelectronic devices. The problem of a brittle oxide layer can be tackled to a large extent by optimal construction of the full device. It is therefore required that the packaging is such that the oxide layer is positioned in the so-called neutral zone of the different layers. The stress in this zone is then minimal under bending conditions so that cracks or other damage to the oxide layer are prevented. This might, however, complicate the production process substantially and require high optimization of the different layers with respect to their mechanical properties. Furthermore, it does not yet deal with the problem of lower quality of the oxide layer itself processed at lower temperatures. So, it can be useful to look into new materials that might have the possibility to fully replace such oxide layers and that can easily be deposited on plastic substrates. If these new materials would be of an organic nature, detrimental effects due to a brittle character would no longer have to be considered. Furthermore, the limited availability of a material such as indium results in a high cost for ITO coatings. Indium-tin-oxide coatings are also commonly used as conductive anode material for liquid crystalline display (LCD) panels. Due to the growing demand for flat-panel displays, the price of indium has increased from $60 per kilogram in 2002 to nearly $1000 today. Thus, development of new materials is expected to be beneficial from an economical point of view. In the next section, PEDOT:PSS films are investigated as just such a replacement material for ITO as the transparent electrode in the organic solar cell. This work was conducted in strong collaboration with Agfa-Gevaert part (B). They were able to provide a PEDOT-based material with considerably higher conductivity than the standard dispersion used in previously described experiments. Therefore, their material is denoted as highly conductive PEDOT (HC-PEDOT) to distinguish both compounds. 5.5.1 PEDOT as transparent contact A comparison is made between solar cells with either ITO or HC-PEDOT as a transparent contact. In all cases, this electrode material was finally covered with
270
Chapter 5
a thin standard low-conductive PEDOT film to have a fair comparison with previously described results. After this preparation of the respective anodes, the active layer of the devices was spin coated from a chlorobenzene solution. It consists of a donor/acceptor blend of MEH-PPV and PCBM in a 1:4 (w/w) ratio. The active area of the solar cells amounts are ∼1.2 cm2 after deposition of the aluminum top electrode under high vacuum. At first, a highly conductive PEDOT dispersion that could be deposited by spin coating was examined. Processing was done on a glass substrate. A sheet resistance in the order of 400 Ω square−1 was obtained for such coatings. Patterning of this layer to obtain a conductive electrode only in certain well-defined areas was done by similar UV lithography as described earlier for ITO contacts. After deposition, illumination, and development of a suitable resistance, the unprotected HC-PEDOT is etched by a sodiumhypochlorite (NaClO) solution. Immersion in such etching material for more than 90 s is needed to remove the layer. However, if the HCPEDOT is exposed only for a limited time (e.g., 10 s) to the NaClO solution, full deactivation occurs. This means that the conductivity of the unprotected material is completely reduced by chemical interactions. IV characteristics measured under AM 1.5 simulated solar illumination of photovoltaic devices constructed on glass substrates with either ITO or spin-coated HC-PEDOT as the anode are depicted in Fig. 5.37(a). For the ITO device, Isc of 09.3 × 10−4 A and Voc of 770 mV were obtained. A somewhat limited value of the fill factor (∼0.35) can readily be accounted for by the larger surface area of the
Figure 5.37 IV characteristics measured under simulated solar illumination (AM 1.5, 100 mW cm−2 ) of photovoltaic devices constructed on glass substrates, (a) with either ITO or spin-coated HC-PEDOT as anode; (b) with screen printed HC-PEDOT as the anode, also showing the result of a series connection of two similar devices.
Processing and Production of Large Modules
271
device than commonly studied in previous experiments. For the device structures here, an influence of the sheet resistance on the transport of charges through the anode layer can be expected. This can be observed by the presence of a rather large series resistance effect in the IV-curves at a higher forward bias. The solar cell with the HC-PEDOT anode exhibited the same Voc value, but a lower short-circuit current of 2.4×10−4 A was measured in this case. Because both devices differ only in the anode, this reduced value has to be attributed to properties of the electrode material used. A distinct electrical difference of the anode materials is the higher sheet resistance of the HC-PEDOT by at least a factor 10 with respect to that of the ITO layer. These results therefore indicate the importance of a sufficiently high conductivity of the applied transparent electrode. The HC-PEDOT dispersion used up to now has been deposited by the spincoating technique. In this way, full coverage of the substrate can be easily obtained. However, only certain well-defined areas have to be covered with a transparent conductive layer for organic solar cell applications. Therefore, photoresist-based patterning steps have to be included in the processing of these spin-coated layers. To circumvent this, a screen printable HC-PEDOT dispersion (Orgacon EL-P 3040) has been developed. This offers the advantage of depositing patterned features in a single step. Organic solar cells with such screen printed anodes instead of spincoated ones were processed on glass substrates. IV-curves measured under AM 1.5 simulated solar illumination are displayed in Fig. 5.37(b). A single device with such an anode exhibits a short-circuit current of 10−4 A and its Voc value amounts to 820 mV. This is quite comparable to what is obtained for a device with a spincoated HC-PEDOT transparent electrode. The current is also somewhat lower than for the case of an ITO electrode, pointing again to the need for sufficiently high conductivity of the anode. On the other hand, a series connection of two cells on the same substrate with a screen printed HC-PEDOT electrode is examined. The resulting IV-curve is also depicted in Fig. 5.37(b). The open-circuit voltage of this series connection has a value of 1620 mV, which almost doubles the value of the single device. Also, the short-circuit current of such serially connected devices remains approximately unchanged with respect to the single device. This demonstrates the good reproducibility of devices constructed by this method. Therefore, a series connection of photovoltaic devices exhibiting a linear increase of the output voltage with no substantial loss in collected current can only be achieved for devices that perform equally well, as explained earlier. Consequently, HC-PEDOT can, to a certain extent, be a replacement material for ITO as a transparent anode material in organic photovoltaic devices. This has a high potential due to its ease of processing. Moreover, direct patterning of such electrodes can be achieved by low-cost methods such as screen printing. Unfortunately, its lower conductivity still limits the performance of the solar cells. In the next section, a possible solution for this problem is investigated.
272
Chapter 5
Furthermore, the applicability of HC-PEDOT on flexible substrates are be demonstrated. 5.5.2 Introduction of a conductive grid into photovoltaic devices A further decrease of the resistivity of HC-PEDOT-based anodes in organic photovoltaics is tested by applying a metallic grid in between the substrate and the transparent electrode material. The deposition of this metallic grid onto the substrate is achieved by the diffusion transfer reversal (DTR) technique that has been optimized to result in a conductive Ag pattern that exhibits a sheet resistance of ∼2 Ω square−1 . In the design used, conductive lines are chosen to be 40-μm thick at a distance of 400 μm from one another. Light transmission of the final electrode is thereby blocked in less than 10% of the active area of the photovoltaic device. It is possible to further optimize this in a later stage since DTR is not limited to these grid sizes. Simple optical mask illumination allows that this technique can handle many kinds of features. The incorporation of such an Ag grid in the transparent electrode of an organic solar cell is demonstrated in Fig. 5.38. Because DTR is a process best suited to be used on flexible substrates, a comparison has first been made between devices with spin-coated HC-PEDOT anodes either on glass or on a PET substrate. The active layer consists of a donor/acceptor blend of MEH-PPV and PCBM in a 1:4 (w/w) ratio. The devices were finalized by the evaporation of an aluminum top electrode that again sets the active area of the solar cells to ∼1.2 cm2 . IV-curves for these solar cells are given in Fig. 5.39(a), measured in an inert N2 atmosphere under halogen lamp illumination of ∼100 mW cm−2 . The device on glass exhibited an open-circuit voltage of 835 mV; whereas for the photovoltaic cell on PET, this was somewhat lower (Voc = 785 mV). Moreover,
Figure 5.38 Picture of an Ag grid incorporated in the transparent electrode of an organic solar cell.
Processing and Production of Large Modules
273
Figure 5.39 IV characteristics of photovoltaic devices, measured under halogen lamp illumination (∼100 mW cm−2 ) with (a) spin-coated HC-PEDOT as anode material, constructed on either glass, PET, or PET with Ag-grid substrates, and (b) either spin-coated or screen printed HC-PEDOT as the anode material, constructed on PET substrates with Ag grid.
concerning the short-circuit current, very similar values were obtained, namely, 0.65 mA and 0.56 mA for the device on glass and on PET, respectively. This indicates that a transfer of the full process to a PET substrate is possible without a substantial loss in performance of the solar cell. Besides, it can be assumed that similar processing will also be applicable to other flexible substrates. Additionally, the result for a device with an Ag grid between the PET substrate and the spin-coated PEDOT:PSS anode is displayed in Fig. 5.39(a). Compared to the solar cell without such a metallic layer, a slight decrease of 85 mV in Voc is seen. However, a more than threefold improvement for the short-circuit current has been achieved by applying the extra layer (Isc = 1.8 mA). In this way, it is clearly shown that lower resistivity of the organic anode can be obtained by the application of a highly conductive grid design, improving the performance of the photovoltaic device. It was shown earlier that the spin-coated HC-PEDOT layer can be replaced by the deposition of a directly patterned HC-PEDOT layer by applying screen printing. Also, this has been examined here in combination with the conductive Ag grid. A comparison of the result with such a screen printed HC-PEDOT anode to that of the previously described spin coated anode is displayed in Fig. 5.39(b). Both types of devices were assembled on a PET substrate and have an extra Ag grid in between the substrate and the organic anode, deposited using the above-described DTR technique. A clear difference in the short-circuit current can be seen; for the screen printed device, Isc = 5.2 mA, which is an almost threefold increase with respect to the cell that contains the spin-coated HC-PEDOT anode. Also, Voc
274
Chapter 5
improved to a value of 735 mV for the device with the screen printed anode. An important observation was that during the patterning process for the spin-coated HC-PEDOT film, the underlying Ag grid was partly destroyed. Screen printing the conductive organic dispersion prohibits such detrimental effects, leading to a higher conductivity of the resulting composite anode. It can thereby be observed that the effect of series resistance at higher forward bias is substantially lower for the screen printed anode layer than for the spin-coated one. Therefore, the results for the device with the Ag grid and the screen printed HC-PEDOT anode show that an organic solar cell can be successfully processed onto this composite electrode on a flexible substrate. The use of highly conductive PEDOT dispersions as transparent contacting materials in the bulk heterojunction solar cell structure is investigated as a replacement for ITO. Both spin coating and screen printing are evaluated as a deposition technique for this layer. The latter technique offered the possibility of easily interconnecting different photovoltaic cells with each other in a modular structure. Unfortunately, the obtained results showed that the limited conductivity of the PEDOT dispersions lowered the solar cell performance compared to ITO-based devices. Therefore, the introduction of an additional metallic grid was examined in combination with the PEDOT contacting layer to construct a transparent electrical contact with sufficiently high conductivity. The processing of this metallic grid involved Ag halide-based photographic techniques. The integration of this additional layer was successful and resulted in photovoltaic devices with a performance highly comparable to that based on an ITO transparent electrode. Furthermore, successful processing of organic bulk heterojunction solar cells with this newly developed transparent electrode structure was obtained on flexible substrates.
5.6 Processing of the Opaque Back-Side Contact After considering the processing of the active layer and the transparent front-side contact, we now look into the back-side contact. Traditionally, we have deposited this by high-vacuum thermal evaporation. Such technology was highly effective for our research purposes, but involves an expensive tool. Also, large-scale processing is not to be expected with this technique, certainly not with a throughput that would be desired for industrial processing. Furthermore, existing industrial techniques such as sputtering are not easily compatible with large-scale patterning. Therefore, the possibilities of low-cost processing techniques for the deposition of a metallic back-side contact are explored here. One option would be to consider the photographic techniques described earlier. However, the use of different solution processing steps in which either water- or solvent-containing solutions are needed clearly have large detrimental effects on the active layer already present at this stage. Instead, we consider, for the purpose here, the deposition of metallic-
Processing and Production of Large Modules
275
based, conductive pastes. This would be compatible with low-cost processing such as printing techniques. Most commonly available conductive inks and pastes are based on metallic silver particles. Metallic Ag has a work function comparable to that of aluminum, but its stability against oxidation is far better. Therefore, Ag can be considered a suitable material for initial experimental investigations on alternative deposition techniques for the back-side contact of organic solar cells. The experiments described further in this section are based on Ag pastes provided by Emerson and Cumming part (B). 5.6.1 Ag-based pastes as back-side contacts Two main systems exist for the preparation of Ag-based pastes. At first, solventbased dispersions of silver particles in a carrier material can be processed. Resistivity below 50 μΩ cm−1 can be easily reached with these solutions. Also, processing can be done in many different ways since they can be made available in a large viscosity range. Very fluid inks suitable for techniques such as ink jetting are easily obtained, but also more viscous versions for direct painting or dispensing can be produced. Unfortunately, the active layer of the organic devices that we would like to study is only ∼100-nm thick. Moreover, it contains materials that dissolve in most common solvents. So, it is rather likely that deposition of such solvent-based Ag pastes will result in shunting paths between the upper and lower electrodes in the studied organic solar cell devices. These solvent-based Ag dispersions have, however, already been shown to be highly suitable for external contacting of the solar cells when sealing these devices is considered, as explained in the fabrication of the standard device in Chapter 2. The silver particles can be dispersed in epoxy-based carrier materials. This excludes the usage of solvents, but introduces some additional steps in the deposition of these pastes. In a first stage, the epoxy is fluidlike and has to be converted into a solid state after deposition. Applying a two-component system commonly does this. Thereby, the basic epoxy parts, also containing silver particles, are mixed with a binder shortly before processing. The mixing initiates the cross-linking of the epoxy parts into a solid structure. This process can take up to more than 24 hr at room temperature, but heating can facilitate it such that it takes even less than a minute for certain epoxy material compositions. To obtain a highly conductive film, percolation of the Ag particles in the epoxy matrix has to occur. This is achieved by shrinking the matrix, for which an annealing step is again required. In some cases, such annealing is combined with the application of pressure, especially when, besides shrinkage of the epoxy, an adhesive effect also has to be obtained. The resulting resistivity is minimally on the order of 100 μΩ cm−1 . The viscosity of these pastes is mostly higher than that of the solvent-based conductive inks. Therefore, deposition techniques such as dispensing or screen printing are more suitable for the epoxy-based Ag pastes.
276
Chapter 5
Feasibility tests were done with several epoxy-based pastes as back-side contacts in organic solar cells. At first Amicon pastes were tested. They resulted in a resistivity of ∼100 μΩ cm−1 after a curing procedure of about 10 min at a temperature of 150◦ C. Since these conditions could result in degradation effects of the underlying organic active layer when a polymer such as MEH-PPV is incorporated in it, another system was investigated. A standard glass/ITO/PEDOT substrate was used. The active layer was deposited onto it by spin coating a donor/acceptor blend from a chlorobenzene solution. The acceptor material was PCBM, whereas the donor was a poly(thiophene vinylene) derivative. So, a mixture of such a PTV precursor with PCBM was deposited. Afterward, a conversion of the polymer precursor into its conjugated form was conducted by heating the sample in an inert atmosphere up to a temperature of 185◦ C. Finally, drops of the Ag paste were dispensed onto this active layer. Curing of the paste at a temperature of 150◦ C for 10 min in an inert nitrogen atmosphere was also done. We showed earlier that not only the PTV polymer but also PCBM obviously remain stable at these annealing conditions. IV characteristics of such a glass/ITO/PEDOT/PTV:PCBM/Ag paste device are given in Fig. 5.40. For the result under dark measurement conditions, it can be observed that a rectification ratio of only ∼10 at 1 V is obtained. Comparing this with the previously obtained results on devices with a similar active layer, it can mainly be attributed to a large current density in the reverse bias region. This may indicate that there are many shunting paths between the electrodes present in the device here. These might be induced due to the deposition procedure of the Ag paste. The dispensing itself could create these shunts, or the subsequent annealing step could induce some of the paste particles to diffuse into the active layer. On the other hand, in forward bias a current density comparable to what was obtained
Figure 5.40 Dark and illuminated IV characteristic of an organic solar cell with a PTV:PCBM active layer and a Ag-paste back-side contact (illumination was done with a halogen lamp at an intensity of ∼100 mW cm−2 ).
Processing and Production of Large Modules
277
previously can also be observed here. This indicates that a low resistivity of the Ag-paste contact has been achieved such that the device characteristic is not much altered in this regime. The IV characteristic under illumination was obtained with a halogen lamp at an intensity of ∼100 mW cm−2 . A short-circuit current of 0.12 mA cm−2 and an open-circuit voltage of 75 mV were measured. It is clear that the strong shunting effects, discussed earlier, limited both these values. It also reflects on the fill factor, which is only 0.26. Conclusively, an energy conversion efficiency of only 0.002% could be achieved here. This is more than a factor of 10 lower than what was measured for a similar device with an evaporated Al back-side contact. This decreased performance is clearly related to minor values for Isc and Voc . Thus, also from the illuminated characteristic, it can be concluded that mainly the induced shunting due to the dispensed electrode has a detrimental effect on the electrical behavior of the photovoltaic cell. In a second stage, Amicon pastes were tested on device structures with an active layer of spin-coated MEH-PPV:PCBM (1:4). Unfortunately, in all cases a current density below 10−6 A cm−2 was observed at an applied bias of 2 V. This is most likely due to degradation of the organic layer during the curing of the dispensed Ag paste. In many cases, a slight color change of the active layer could be observed despite processing in an inert atmosphere. Therefore, an Eccobond paste was subsequently investigated as a back-side contact. It is a two-component epoxy-based Ag paste that cures already at a temperature of 65◦ C within 2–3 min. The resulting resistivity is, however, somewhat higher than for the Amicon paste, with a value of over 500 μΩ cm−1 . It was dispensed onto a device structure with an active layer of spin-coated MEHPPV:PCBM (1:4). Measured IV characteristics of such a glass/ITO/PEDOT/MEHPPV:PCBM/Ag paste device are shown in Fig. 5.41. For the curve under dark measurement conditions, a rectification ratio is obtained as high as ∼4×104 at 1 V. A low current density in reverse bias is measured, indicating very limited shunting effects in this device. On the other hand, a sufficiently high conductivity of the back-side contact is achieved in order to not limit the current density in forward bias. This also shows that the previously mentioned effects due to high curing temperatures are absent here. Under illumination with a halogen lamp at an intensity of ∼100 mW cm−2 , a short-circuit current of 0.5 mA cm−2 and an open-circuit voltage of 715 mV were measured. These values are of course both lower than what is obtained earlier for a device with an optimized evaporated contact. It shows, however, that a reasonable result can be achieved with a low-cost processing technique for the back-side contact if material properties are made more compatible with each other. Limitations remain, of course, due to nonoptimal intrinsic or extrinsic properties. Ag as a back-side electrode results in an energy level offset that is not as appropriate as that of LiF/Al or Ca in a bulk heterojunction solar cell. This can affect the Voc as well
278
Chapter 5
Figure 5.41 Dark and illuminated IV characteristic of a glass/ITO/PEDOT/MEHPPV:PCBM/Ag paste solar cell (illumination was done with a halogen lamp at an intensity of ∼100 mW cm−2 ).
as create a limitation to the extraction of the photogenerated current. Moreover, the pastelike consistency of the electrode can result in a disrupted coverage of the organic layer by metallic silver, inducing additional resistance effects. Also, the reduced conductivity of the paste with respect to an evaporated, fully metallic layer limits the performance of the device. Therefore, further optimization of the pastes could lead to a suitable low-cost deposition of back-side electrodes for organic solar cells. In this respect, some other types of conductive Ag films have been examined. At first, a film of Ag particles dispersed in an epoxy matrix coated onto a glass fiber-woven structure was considered. This can be applied onto nonflat, flexible surfaces and has adhesive properties. Therefore, a second material has to be put on top of it. In the final device structure, this additional layer could be considered to have a sealing function so that an encapsulation is provided at once. However, the conversion of the Ag film into its conductive state involves temperature curing combined with a pressure step. Organic active layers can be developed that are stable at an annealing of 1–2 hr at 95◦ C as previously shown. However, applying a pressure of even a few kilograms per square centimeter is rather incompatible with an active layer thickness on the order of ∼100 nm. Even the smallest irregularities of the adhesive film are on the order of micrometers and will therefore easily perforate the active layer during a pressure step. Test results showed that shunting could not be avoided for such Ag films. The advantages of combining the adhesive effect with an encapsulating material and the high flexibility of the layer make it an attractive material composition if pressure steps could be minimized or ultimately avoided. Also, an Ag paste was investigated with a consistency suitable for screen printing and a curing temperature as low as 80◦ C. Unfortunately, the fabricated devices
Processing and Production of Large Modules
279
were again short-circuited, most likely due to perforating the underlying active layer during the printing process. Therefore, further optimization of this process is required. Nevertheless, its feasibility is already shown for light-emitting devices with an active layer of conjugated materials. Working LEDs with such screen printed back-side contacts have been demonstrated. For the Ag paste studied here, we could observe that a good mechanical flexibility is obtained for the printed electrode without any substantial decrease of the conductivity during or after bending it. Resistivity below 500 μΩ cm−1 could be achieved for such printed Ag layers. Experimental investigations on alternative deposition techniques for the backside contact of organic solar cells are described. For this purpose, we focused on Ag-based pastes. Solvent-based dispersions of silver particles in a carrier material were considered not to be useful. Therefore, dispersions of silver particles in an epoxy-based carrier material were examined. Feasibility tests showed that dispensing such pastes on the photovoltaic active layer could result in successful formation of a back-side contact. Nevertheless, processing conditions such as curing temperature and time had to be highly compatible with the active layer materials to prevent degradation effects. A pressure-sensitive adhesive layer of Ag pastes or the introduction of a technology such as screen printing to fabricate the back-side contact was also considered. Whereas advantages such as large-scale, high-throughput processing and flexible devices could be expected from these approaches, we have not yet been able to fabricate working solar cells with them. Mainly the creation of shunting paths due to the pressure involved in both processes onto the very thin organic active layer was detrimental for successful device fabrication.
5.7 Encapsulation and Permeability Encapsulation of OPVs is a relatively unexplored area, and the available techniques are most often developed in a laboratory to suit the needs of scientific experimentation. For this reason, the encapsulation procedures that have been reported to the public are mostly rigid and by no means effective in terms of cost and quality. Encapsulation will be one of the most important aspects of a successful introduction of an OPV product to the market, and it is expected that once the developmental stage reaches the industry and the doorstep of the consumer, innovative solutions will appear. Generally, the encapsulation or encasement techniques can be either rigid or flexible. From an experimenters point of view, rigid encapsulation is the most appealing since the device will be in a tangible form. From the consumers point of view, flexible encapsulation will be the most versatile. 5.7.1 Measurement of permeability Polymer solar cell devices with reactive metal electrodes require encapsulation with a barrier material that exhibits extremely low permeation rates for water vapor and oxygen. These transmission rates in oxygen and water, deducted from the require-
Chapter 5
280
ment of OLEDs, are on the order of 10−3 cm3 m−2 day−1 for oxygen (OTR) and 10−6 g m−2 day−1 for water vapor (WVTR), see Fig. 5.42.64 Barrier materials have to follow the same driving aspects as organic devices, such as transparency, flexibility, and processing, with low-cost techniques, for example, coating or lamination.65 The permeation of low molecular weight chemical species through a polymeric matrix is generally envisaged as a combination of two processes, namely, solubility and diffusion (Fig. 5.43). A permeant gas is dissolved into the upstream face of the material and then undergoes molecular diffusion to the downstream face, where it evaporates into the external phase again. A solubility/diffusion mechanism is thus applied, which can be formally expressed in terms of permeability P , solubility S, and diffusion D coefficients by67 P = D · S.
(5.11)
The solubility coefficient S is thermodynamic in nature, and is defined as the ratio of the equilibrium concentration of the dissolved permeant in the polymer to its partial pressure p in the gas phase (Henry’s law) as follows: C1 = S · p1 .
(5.12)
The diffusion coefficient D characterizes the average ability of the sorbed permeant gas to move through material, and is determined from Fick’s first law of
Figure 5.42 WVTR requirement for common flexible electronic devices and barrier performance provided by available materials.66
Figure 5.43 An illustration of the solution process.
Processing and Production of Large Modules
281
Figure 5.44 Typical flux evolution during the measurement.
diffusion, namely, the flux of the permeant J is proportional to the local gradient of concentration c through the thickness of the polymer: J = −D ·
∂C . ∂x
(5.13)
During the measurement of J,57 surfaces of the sheet in x = 0 and x = l are maintained at constant concentration C1 and C2 . In the most common experimental arrangement, C2 is equal to 0. After a time, a steady state is reached, 0, in which the concentration remains constant at all points of the sheet. In the steady state, the flux J∞ can be expressed as follows (Fig. 5.44): J∞ = C te = D
p 1 − p2 Δp C1 − C2 = DS =P . l l l
(5.14)
5.7.2 Measurement of the diffusion coefficient D 67 By integrating J with respect to time, we obtain the total amount of diffusing substance Qt , which has passed through the membrane in time t. If C1 , at t = 0, is equal to 0 and C2 is maintained at 0, Qt =
∞ (−1)n 2 Dt 1 2 2 t − − exp Dn π , l2 6 π2 1 n2 l2
(5.15)
which, at t → ∞, approaches the line
DC1 l2 t− . Qt = l 6D
(5.16)
This has an intercept T1 on the t-axis given by (Fig. 5.45 and Table 5.1) D=
l2 . 6TL
(5.17)
Chapter 5
282
Figure 5.45 Cumulative flux versus time.
Table 5.1 Usual units for permability measurements.
[Dimension]
Usual unit
Flux J
[Quantitypermeant ] [Time]
g(cm3 )∗ day
Flux J (by surface unit)
[Quantitypermeant ] [Surfacematerial ]×[Time]
g(cm3 )∗ m2 ×day
Permeability constant P
[Quantitypermeant ]×[Thicknessmaterial ] [Surfacematerial ]×[Time]×[pressureupstream ]
g(cm3 )∗ ×cm m2 ×day×atm
∗
cm3 are used if a gas such as oxygen is the permeant.
5.7.3 Units The improvement of the substrate barrier properties by a coating is usually described by the barrier improvement factor (BIF), which is the ratio between the barrier performances of the coated material and the substrate alone. 5.7.4 Apparatus In each of the permeability measurements, the upstream partial pressure p1 of the permeant is maintained constant and the downstream pressure p2 is maintained at 0. Different methods are possible to provide these conditions (e.g., vector gas or vacuum) and to measure the flux of the permeant gas in the downstream side. The most common instruments used for permeation measurement are not sensitive enough to measure very low transmission rates that correspond to the organic solar cells requirement. Up until now, commercially available instruments were configured for higher transmission rates, but because of the new optoelectronic applications such as OLEDs, new commercial or experimental instruments are beginning to become available.
Processing and Production of Large Modules
283
5.7.5 An example of a commercial instrument The sensor of the Aquatran (Fig. 5.46) is a coulometric phosphorous pentoxide that is absolute. The downstream pressure p2 is maintained at 0 by a dry nitrogen gas flow. 5.7.6 The calcium test64 Based on the corrosion of reactive metal films, the calcium test is an optical method used to quantify the water transmission rate of substrates provided with high-performance diffusion barriers. Effective permeation constants in the range of 10−1 to 10−6 g H2 O m−2 day−1 have been determined experimentally (Table 5.2). The calcium test can also be advantageously used to identify micromechanical damage of thin, brittle diffusion barrier layers on polymer-based substrates. A thin layer of metallic calcium is deposited on a substrate of interest, such as a polymerbased substrate, and encapsulated with a transparent powder-blasted lid as shown Table 5.2 High-sensitivity permeation instruments and their specifications.
Principle
O2 detection limit (cm3 /m2 /day)
H2 O vapor detection limit (g/m2 /day)
Commercial OxTran 2/21 PermaTran-W 3/33 Mocon
Coulometric (oxygen) Infrared (water)
∼5 × 10−3
∼5 × 10−3
Commercial Aquatran 1 Mocon
Coulometric sensor Film 50 cm2 2–50 cm2 test cells per module Expandable up to 10 modules
/
Experimental Mass spectromery With isotopic elements68
Mass spectrometry
Nd
∼5 × 10−5
Experimental Calcium-test69
Measurement of the degradation by water of an encapsulated calcium layer70
/
∼1 × 10−6
Experimental Tritiated water71
Counter
/
∼2 × 10−7
5 × 10−4
Chapter 5
284
Figure 5.46 Scheme of the Aquatran. (Source: Mocon, Inc.)
Figure 5.47 Schematic representation of the cell used in the calcium test.
in Fig. 5.47. At room temperature, calcium reacts only with water70 entering the test cell and becomes progressively transparent: Ca + H2 O → CaO + H2 , CaO + H2 O → Ca(OH)2 . This leads to a change in the optical transmission of the test cell, which can be monitored over time. The calcium layer can be deposited following patterns that simplify image analysis. The total area covered by metallic calcium as well as the optical transmission of this area changes in time. The change in transmission of the cells is then used to quantify effective permeation rates.
Processing and Production of Large Modules
285
Figure 5.48 Measure of permeability with isotopic elements (INES/CEA/LCS).
5.7.7 Mass spectrometry68 Mass spectrometer detection systems are nonspecific (elemental discrimination), and several test gases/environments can therefore be tested in principle, thus providing more flexibility than single-detector machines. Indirect permeation from a vacuum seal is a limiting factor (O2 and H2 O vapor are present around the machines), and the sensitivity of the detection system does not mean much if the baseline is not controlled properly. Using isotopic labelling the partial pressure of the baseline can be much lower. A sensitivity around 5×10−5 g m−2 day−1 is achieved as follows (Fig. 5.48): 1. To form a mixture of isotopic gases, each isotopic gas corresponding to a target gas is mixed in a mixing enclosure. 2. The mixture of isotopic gases fills a first chamber of the permeation enclosure comprising the first and second chambers (high-vacuum chamber), the first chamber being separated from the second chamber by the material being tested. 3. The detection of the isotopic gases having permeated through the material and being present in the second chamber is performed by the mass spectrometer. 5.7.8 Tritiated water71 This permeation measurement method uses tritium-containing water (HTO) as a tracer material. The theoretical detection limit is very low (2 × 10−7 g m−2 day−1 ). The detection is performed using an ionization chamber that has a counter. 5.7.9 Oxygen permeation in PEDOT Few examples exist in the literature on permeation through electroactive materials, and the OTR and WVTR are often given for substrate materials and not for the
286
Chapter 5
Figure 5.49 Measure of permeabilty with HTO. (Reprinted from Ref. [72], with permission from IEEE, copyright 2005.)
active materials. The problem is that large areas of film are often required to get the desired response, and the active layers are often very thin films prepared by coating methods and are not easily obtainable as thick freestanding films. It is possible to obtain permeation data on a carrier substrate using laminate theory, and in this way obtain permeation data. This method has been used to estimate the OTR in PEDOT.73 Thin films of PEDOT were prepared on poly(ethylene) (PE) carrier substrates and the OTR could be extracted. A requirement of the technique is that OTR or WVTR of the carrier substrate is comparable to or larger than the material that is analyzed.
5.8 Practical Encapsulation Techniques Much development in this area is needed and probably underway. Many solutions are currently being sought for the encapsulation of OLED and PLED technologies; these techniques should be more or less directly transferable to OPVs. It is also very likely that the encapsulation procedure will involve the use of a getter material such as Dryflex from SAES.74 Getter materials can be included inside the package and have the capacity to actively remove a certain amount of water and oxygen, thus prolonging the life of the packaged device by reducing the degradation mechanisms pertaining to water and oxygen. There are other water- and oxygen-independent degradation mechanisms that a getter material will not reduce, and as such the view of the encapsulation should be to achieve the required level of stability, because OPV technology is inherently unstable and its performance degrades. The game is
Processing and Production of Large Modules
287
Figure 5.50 Top view of several solar cells on a typical sample and a representation of a sealed sample.
thus to encapsulate the OPV device in such a manner that the required packaged device lifetime is achieved. 5.8.1 Rigid encasement at IMEC (Belgium) Moisture, humidity, and oxygen are known to lead to degradation of organic materials; therefore, encapsulation of these devices is a requirement. A fully encapsulated device on a glass substrate is depicted in Fig. 5.50. First, metal strips are attached at the edges with silver paste (Eccocoat CC2, Emerson and Cuming) to make extra electrical contacting pads with the individual ITO and evaporated metal lines. Second, a rubber spacer (Viton, Eriks) is put on the sample, leaving some unprotected area at the edges of the substrate. It is at these edges that an electrically insulating two-component epoxy glue (EPO-TEK 302-3M, Gentec) is applied. Finally, an extra glass cover plate is put over the substrate. The rubber spacer prevents the epoxy glue from flowing from the edges to the center of the substrate and makes sure that the cover plate is not damaging the actual solar cells. The epoxy glue cures overnight at room temperature. Recently, faster-curing glues have become available from the above-mentioned company. As a result, the cover plate and the extra contacting pads are firmly attached to the substrate. Electrical measurements on the actual devices are therefore still possible and can even be carried out in an ambient atmosphere since the epoxy and the glass cover form a strong barrier for any kind of moisture, humidity, or oxygen. Typically, the penetration through such encapsulation at room temperature and a relative humidity of ∼50% is reduced for water vapor to a level of (or even below) 10−5 g m−2 day−1 ; and also for oxygen, values below 10−3 cm3 m−2 day−1 can be obtained.75 5.8.2 Small rigid encasement at Risø National Laboratory (Denmark) An encapsulation technique has been developed so that they can be exchanged with other laboratories, used for demonstration purposes, or tested in different locations.76 The choice of substrate was glass, where typically an ITO coating is
288
Chapter 5
Figure 5.51 An example of a cell that is sealed in glass. The opaque plastic tubes have been glued on for mechanical protection of the fragile glass seal.
employed. The dimensions of the glass substrate is 75 mm × 50 mm and a milled aluminum back plate is employed that allows for an active area of 40 mm × 30 mm. The challenge for all encapsulation techniques is the seal. A glass seal is, for instance, the best you can get because it is an absolute seal, where no diffusion takes place (Fig. 5.51). There are other diffusion-free sealing techniques such as different soldering alloys and ceramics, but common to all of them is that they employ a high temperature. The challenge is to create a seal by a low-temperature process <80–100◦ C due to the low morphological stability of many OPV materials at high temperatures. It is preferable to have a sealing temperature of 70◦ C or at the maximum operational temperature of the device. Most sealing methods for this temperature interval employ some sort of glue that is polymer based. Regardless of what polymer material it is, it will never qualify as a tight seal. The method for curing the sealing materials or glues can be thermal or by UV light. The advantage of using a UV-curable material is that lower temperatures can be employed. Since permeation is directly proportional to the cross-sectional area of the seal and inversely proportional to the length of the seal, much can be gained by having a very thin seal over a large area. An even better seal is obtained if the sealant is filled with an impervious material such as glass fibers. By applying a mechanical pressure during sealing, the effective cross-sectional area is diminished and a much more efficient seal is obtained as shown schematically in Figs. 5.52 and 5.53.
Processing and Production of Large Modules
289
Figure 5.52 An illustration of how glass filling makes a better seal. The green areas are the permeable sealant, and the gray boxes and black circles are impermeable. The gray boxes are part of the package and the filled circles are cross sections of glass fibers.
By using a glass-filled thermosetting epoxy that becomes liquid when heated, it is possible to make a very good seal by applying pressure during sealing, thus forcing the package against the fibers and minimizing the area available for permeation. The sealing method is quite efficient, and devices have been kept for more than two years with degradation to only about half the initial performance. It is possible to include a getter material in the milled area of the back plate.
5.8.3 Large rigid encasement at Risø National Laboratory (Denmark) The encapsulation method shown above can be scaled to larger sizes, and in this section a module encasement with outside dimensions of 30 cm × 30 cm is described. Four 10 cm × 10 cm areas are available for realization of small modules. The back plate was 5-mm thick milled aluminum and the front plate was 1-mm ITO-coated glass. The device’s active layers could be printed directly onto an etched ITO pattern. The sealing procedure is exactly the same as in Section 5.8.2. As shown in the collection of images, Dryflex sheets from SAES getters were attached to the back plate before sealing the device. The device consisted of screen printed PEDOT:PSS Orgacon EL-3040 from AGFA, with P3HT:[60]PCBM as the active layer. The cathode was evaporated aluminum. The procedure for including the Dryflex getter material is shown in Fig. 5.54. The product is easy to use and comes with a self-adhesive backing that makes it easy to attach to that back plate. The device was then completed by evaporating the aluminum electrode through a shadow mask. The device was placed onto the back plate and sealed at 70◦ C for 12 h (Fig. 5.55).
290
Chapter 5
Figure 5.53 The back plate (top left) and the back plate with glass fiber-filled thermosetting epoxy (top right; the sealed device from the front and back side (below).
5.8.4 Flexible encasement The flexible encasement procedures that have been reported for OPVs are scarce and mostly limited to a simple encasement in a polymer that serves as a mechanical barrier only and does not convey any barrier properties toward oxygen and water. While the permeation through the flexible barrier substrate may be very low and perhaps approaching that of glass, there is still the problem of sealing the device at the edges, where epoxy glue has been used in one example (Fig. 5.56).77
Processing and Production of Large Modules
291
Figure 5.54 The back plate is prepared with sealant and fixed in the jig on the hot plate (top left) with epoxy glue. The Dryflex sheets are removed from their enclosure (top right). The foil protecting the adhesive is removed (middle) and the Dryflex sheets are fixed onto the back plate (bottom). All operations were performed in a glove box.
As OPVs make their way to the market, many more solutions for flexible encapsulation will be reported and become available for further development and exploitation.
5.9 Production and Companies 2007 At the time of writing this book, OPVs have moved very close to commercialization. There have been several initiatives aimed at commercializing some form
292
Chapter 5
Figure 5.55 The OPV module is mounted on the back side of the shadow mask (left) and after the aluminum has been evaporated (middle), the glass plate with the printed module is carefully placed on top of the back plate for sealing (right). All operations were performed in a glove box.
Figure 5.56 An example of a small module based on MDMOPPV:[60]PCBM prepared on a flexible ultrabarrier material. The schematic cross section of the device is shown (top) along with the bendable module (below).
Processing and Production of Large Modules
293
of OPV. The most consistent performer started with Siemens AG, who sold their effort to Konarka Inc. Konarka is based in the U.S. and Europe and has leading efforts within dye-sensitized technologies and OPVs in the form of polymer photovoltaics. Konarka is well founded with several Nobel Prize laureates on the board of directors and highly distinguished scientists as scientific advisors. If there is any commercial future for OPVs, it can, from an investor’s point of view, simply not go wrong for Konarka. The most notable achievement is the collection of intellectual property rights (IPRs) over a wide range of solar cell technologies that Konarka have at their disposal. Konarka licensed part of their technology to a company named G24i (Ref. [78]) that has purchased a complete process line for the production of semiflexible solar cells on titanium foil. While very limited information has been released to the public, the technology allegedly relies on a quasi–solid state dye-sensitized solar cell with a limited operational lifetime for military applications (possibly for the replacement of batteries on the battle field). The production line has been produced by Solarcoating Machinery GmBH (Ref. [79]) and delivered in Cardiff (Wales) in early 2007. The production line has a 30 MW annual production capacity, and G24i has announced that there are plans for expanding the production capacity to 200 MW by 2008.
5.9.1 Intellectual property rights in Europe, the United States, and Asia 2007 When it comes to setting up a company within OPVs, whether it is a production company or a company that employs OPV technology in their product, it is crucial to have an IPR portfolio that gives enough freedom to operate without too many dominating patents, where potentially costly licenses have to be secured. Within OPVs there are thousands of patents, and most of them are unimportant or easily avoided because no commercial technology is available and most patents that claim process rights are likely to be of little value since they were written in anticipation of a particular mode of production spawned by the vision of solution processing and printing of electronics. In such a situation, the first commercial exploitation of OPVs will undoubtedly be subject to a wealth of patent holders claiming that their IPRs have been violated with respect to makeup, process, or design. While this will be a time-consuming exercise for a newly started company, it is hardly going to be critical. There are, however, a core of patents that are highly generic and give little freedom to operate unless a license is available. Two heterojunction U.S. patents by Alan J. Heeger and N. S. Sariciftci as inventors appear to be the first IPR, listed as follows: • US Patent No. 5,331,183 (Filed August 17, 1992; issued July 19, 1994; expires October 2011) • US Patent No. 5,454,880 (continuation of US Patent No. 5,331,183) (Filed January 12, 1994; issued October 3, 1995; expires October 2012)
294
Chapter 5
These patents are highly generic and claim exclusive rights over the existing state of the art, namely, polymer-fullerene heterojunctions, polymer-fullerene bulk heterojunctions, polymer-polymer heterojunctions and polymer-polymer bulk heterojunctions, and donor-acceptor concepts, all of which are described in detail and claimed in the patents. Furthermore, the use of conjugated polymers of diverse constitution is claimed, and it is difficult to envisage any significant IPR on generic OPVs aside from these two patents. The University of California seems to be the sole owner of the patents. No assignments to other parties have been recorded at the USPTO, whereas Konarka Technologies Inc. has obtained a limited sole license in a specific field of use and related to certain types of hybrid photovoltaic cells.80 A pending patent application in Japan (JP-20006-080530A) was published on March 23, 2006, claiming priority from U.S. application 07/930,161 (and WO/1994/005045). If the Japanese patent is issued, it will expire in October 2013. The most interesting aspect is that there is no protection in Europe! A patent cooperation treaty (PCT) application (WO/1994/005045) claiming priority from U.S. application 07/930,161 (now issued as U.S. Patent No. 5,331,183) was filed on August 17, 1993. Based on this PCT application, an EP patent application 93920199.2 was filed, but later deemed withdrawn (due to lack of paying filing and search fees), effective as of March 18, 1995. This means that there is an enormous freedom to operate in Europe that is nonexisting in the United States and Japan for the next five to six years. This could mean that it will be in Europe that OPVs will be commercialized first by companies other than Konarka Inc.
5.9.2 A road map for setting up a company producing OPVs in Europe Given this fortunate situation of freedom to operate in Europe, the outline for setting up a company producing OPVs in Europe becomes simple and a road map comprises the following: • • • • •
Composition of matter patent(s) Low bandgap material(s) Stable under illumination Fabrication friendly properties IPR on preparation and processing
With these simple things at hand it would be possible to start producing OPVs based on the low bandgap polymer material for which a composition of matter patent is available. The reason for specifically choosing a low bandgap material is that the amount of IPR on these is rather scarce and the path to higher-efficiency devices pass through low bandgap materials, where more of the solar spectrum is covered (as described in Chapter 3).
Processing and Production of Large Modules
295
5.9.3 What production equipment is available in 2007 Because there is no widespread commercial production of OPVs, it is needless to say that the industry that produces equipment for fabrication of OPVs is very limited. Currently the only industrial partner that furnishes production equipment dedicated to fabrication of OPVs is Solarcoating Machinery (SCM), as mentioned above. When production of OPVs by printing and coating methods becomes widespread, the potential players will be everyone with a business in coating and printing. It is thus expected that there will be a wide range of suppliers, and professional help will be available to start-up companies. Production equipment will be available in the form of standard printing and coating solutions at competitive prices.
References 1. Rau, U., Grabitz, P.O., and Werner, J.H., “Resistive limitations to spatially inhomogeneous electronic losses in solar cells,” Appl. Phys. Lett., 85, pp. 6010– 6012 (2004). 2. Al-Ibrahim, M., Roth, H.K., and Sensfuss, S., “Efficient large-area polymer solar cells on flexible substrates,” Appl. Phys. Lett., 85, pp. 1481–1483 (2004). 3. Goetzberger, A., Hebling, C., and Schock, H.-W., “Photovoltaic materials, history, status and outlook,” Materials Science and Engineering, R 40, pp. 1–46 (2003). 4. Green, M.A., Thin-Film Solar Cells: Review of Materials, Technologies and Commercial Status Solar Cells: Operating Principles, Technology and System Applications, Prentice Hall, Englewood Cliffs, NJ (1982). 5. Van Overstraten, R.J., and Mertens, R.P., Physics, Technology and Use of Photovoltaics, Adam Hilger Ltd., Bristol (1986). 6. Krebs, F.C., Biancardo, M., Winther-Jensen, B., Spanggard, H., and Alstrup, J., “Strategies for incorporation of polymer photovoltaics into garments and textiles,” Sol. Energy Mater. Sol. Cells, 90, pp. 1058–1067 (2006). 7. Flexible Organic Solar Cells for Power-Generating Textiles, Final report of Project No. PA/09, Belgian Science Policy, Brussels (2006); http://www. belspo.be 8. Leonida, G., Handbook of Printed Circuit Design, Manufacture, Components and Assembly, Electrochemical Publications, Port Erin, UK (1981). 9. Scarlett, J.A., An Introduction to Printed Circuit Board Technology, Electrochemical Publications, Port Erin, UK (1984). 10. Coombs, C.F., Printed Circuits Handbook, McGraw-Hill, New York (1988).
296
Chapter 5
11. Shaheen, S.E., Radspinner, R., Peyghambarian, N., and Jabbour, G.E., “Fabrication of bulk heterojunction plastic solar cells by screen printing,” Appl. Phys. Lett., 79, pp. 2996–2998 (2001). 12. Krebs, F.C., Alstrup, J., Spanggaard, H., Larsen, K., and Kold, E., “Production of large-area polymer solar cells by industrial silk screen printing, lifetime considerations and lamination with polyethyleneterephthalate,” Sol. Energy Mater. Sol. Cells, 83, pp. 293–300 (2004). 13. Brollier, B.W., Printing Methods and Their Possibilities, Proc. of IMAPS, Advanced Technology Workshop on Printing an Intelligent Future, March 8–10, Incline Village, NV (2002). 14. de Gans, B.-J., Duineveld, P.C., and Schubert, U.S., “Inkjet printing of polymers: State of the art and future developments,” Adv. Mater., 16, pp. 203–213 (2004). 15. Handbook for Screen Printers, Sefar Printing Division, Bangkok, Thailand (1999). 16. SST Handbuch für den Sieb- und Textildrucker, Reprinted by Schweiz, Seidengazefabrik AG, Thal (1996). 17. Cheek, G.C., “The development of low-cost fabrication technologies for semicrystalline silicon solar cells,” Ph.D. thesis, Catholic University Leuven, Belgium (1983). 18. Franconville, F., Kurzweil, K., and Stalnecker, S.G., “Screen-essential tool for thick-film printing,” Solid State Technol., 17, pp. 61–68 (1974). 19. Dubey, G.C., “Screens for screen printing of electronic circuits,” Microelectron. Reliab., 13, pp. 203–207 (1974). 20. Durgin, S.H., “The emulsion screen versus the metal mask stencil,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 341–344 (1987). 21. Riemer, D.E., “The function and performance of the stainless steel screen during the screen-print ink transfer process,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 826–831 (1986). 22. Huner, B., Rangchi, H., and Ajmera, P.K., “On the release of the printing screen from the substrate in the breakaway region,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 328–334 (1987). 23. Brown, D.O., “Screen printing-an integrated system,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 582–590 (1986).
Processing and Production of Large Modules
297
24. Frecska, T., “Theoretical model for multilevel screen printing of thick-film compositions,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 314–319, 1987. 25. Dubey, G.C., “Sqeegee in printing of electronic circuits,” Microelectron. Reliab., 14, pp. 427–429 (1975). 26. Atkinson, R.W., “Sqeegee preasure and thick-film resistor fabrication,” Solid State Technol., 14, pp. 51–56 (1971). 27. Berry, C.W., and Jacobson, R.Y., “Screen printing design of experiments for low temperature co-fired ceramic substrates,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 538–543 (1999). 28. Jabbour, G.E., Radspinner, R., and Peyghambarian, N., “Screen printing for the fabrication of organic light-emitting devices,” IEEE J. Select. Top. Quant. Electron., 7, pp. 769–773 (2001). 29. Mori, K., Ning, T.L., Ichikawa, M., Koyama, T., and Taniguchi, Y., “Organic light-emitting devices patterned by screen-printing,” Jpn. J. Appl. Phys., 39, pp. L942–L944 (2000). 30. Pardo, D.A., Jabour, G.E., and Peyghambarian, N., “Application of screen printing in the fabrication of organic light-emitting devices,” Adv. Mater., 12, pp. 1249–1252 (2000). 31. Forrest, S.R., Bradley, D.D.C., and Thompson, M.E., “Measuring the efficiency of organic light-emitting devices,” Adv. Radspinner Mater., 15, pp. 1043–1048 (2003). 32. Miler, L.F., “Screenability and rheology,” Solid State Technology, pp. 54–58 (1974). 33. Riemer, D.E., “Ink hydrodynamics of screen printing,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 52–58 (1986). 34. Moeller-Johnson, G., “The importance of rheology in screen-printing ink. Newtonian vs. shear-thinning,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 320–327 (1987). 35. Riemer, D.E., “The shear and flow experience of ink during screen printing,” Proc. of the International Symposium on Microelectronics, International Society of Hybrid and Microelectronics, Reston, VA, pp. 335–340 (1987). 36. Huner, B., “A simplified analysis of blade coating with applications to the theory of screen printing,” Int. J. Hybrid Microelectron. 12, pp. 88–94 (1989).
298
Chapter 5
37. Owczarek, J.A., and Howland, F.L., “A study of the off-contact screen printing process 1 model of the printing process and some results derived from experiments,” IEEE Trans. Compon. Hybrids Manuf. Technol., 13, pp. 358– 367 (1990). 38. Owczarek, J.A., and Howland, F.L., “A study of the off-contact screen printing process 2 Analysis of the model of the printing process,” IEEE Trans. Compon. Hybrids Manuf. Technol., 13, pp. 368–375 (1990). 39. Leach, R., and Pierce, R., The Printing Ink Manual, Springer, New York (1999). 40. Morrison, F.A., Understanding Rheology, Oxford University Press, London (2001). 41. Macosko, C.W., and Larson, R.G., Rheology: Principles, Measurements, and Applications, VCH, New York (1994). 42. Blom, C., Jongschaap, R.J.J., and Mellema, J., Inleiding in de Reologie: Reometrie, Dispersiereologie, Polymeerreologie, Kluwer Technische Boeken, Deventer (1988). 43. Krebs, F.C., Spanggaard, H., Kjær, T., Biancardo, M., and Alstrup, J., “Large area plastic solar cell modules,” Proc. Soc. Vacuum Coaters, 48th Annual Conference, p. 669 (2005). 44. Krebs, F.C., Alstrup, J., Biancardo, M., and Spanggaard, H., “Large area polymer solar cells,” Proc. SPIE, 5938, 593804-1–11 (2005). 45. Krebs, F.C., Spanggaard, H., Kjær, T., Biancardo, M., and Alstrup, J., “Large area plastic solar cell modules,” Mater. Sci. Eng. B, 138, pp. 106–111 (2007). 46. http://displayproducts.ga.com/thinfilm.html (2007). 47. http://www.vitexsys.com/new/index.htm (2007). 48. http://www.nova-plasma.com/ (2007). 49. Lungenschmied, C., Dennler, G., Neugebauer, H., Sariciftci, N.S., Glatthaar, M., Meyer, T., and Meyer, A., “Flexible, long-lived, large-area, organic solar cells,” Sol. Energy Mater. Sol. Cells, 91, pp. 379–384 (2007). 50. Zimmermann, B., Glatthaar, M., Niggemann, M., Riede, M.K., Hinsch, A., and Gombert, A., “ITO-free wrap through organic solar cells—A module concept for cost-efficient reel-to-reel production,” Sol. Energy Mater. Sol. Cells, 91, pp. 374–378 (2007). 51. Hung, L.S., Tang, C.W., and Mason, M.G., “Enhanced electron injection in organic electroluminescence devices using an Al/LiF electrode,” Appl. Phys. Lett., 70, pp. 152-154, 1997; Brabec, C.J., Shaheen, S.E., Winder, C., Sariciftci, N.S., Denk, P., “Effect of LiF/metal electrodes on the performance of plastic solar cells,” Appl. Phys. Lett., 80, pp. 1288–1290 (2002).
Processing and Production of Large Modules
299
52. Krebs, F.C., Carlé, J.E., Cruys-Bagger, N., Andersen, M., Lilliedal, M.R., Hammond, M.A., and Hvidt, S., “Lifetimes of organic photovoltaics: photochemistry, atmosphere effects and barrier layers in ITO-MEHPPV: PCBMaluminium devices,” Sol. Energy Mater. Sol. Cells, 86, pp. 499–516 (2005). 53. H.L. Hartnagel, A.L. Dawar, A.K. Jain, and C. Jagadish, Semiconducting Transparent Thin Films, Institute of Physics Publishing, Philadelphia, USA (1995). 54. Paetzold, R., Heuser, K., Henseler, D., Roeger, S., Wittmann, G., and Winnacker, A., “Performance of flexible polymeric light-emitting diodes under bending conditions,” Appl. Phys. Lett., 82, pp. 3342–3344 (2003). 55. Groenendaal, L., Jonas, F., Freitag, D., Pielartzik, H., and Reynolds, J.R., “Poly(3,4-ethylenedioxythiophene) and its derivatives: Past, present, and future,” Adv. Mater., 12, pp. 481–494 (2000). 56. Ghosh, S., and Inganäs, O., “Self-assembly of a conducting polymer nanostructure by physical crosslinking: applications to conducting blends and modified electrodes,” Synth. Met., 101, pp. 413–416 (1999). 57. Pettersson, L.A.A., Carlsson, F., Inganäs, O., and Arwin, H., “Spectroscopic ellipsometry studies of the optical properties of doped poly(3,4-ethylenedioxythiophene): An anisotropic metal,” Thin Solid Films, 313, pp. 356–361 (1998). 58. De Roeck, J., Basiscursus Fotografie, Agfa-Gevaert, Mortsel, Belgium (1986). 59. Ray, S.F., Imaging Systems for Photography, Film and Video, Focal Press, St. Louis, MO (1988). 60. Neblette, C.B., Fundamentals of Photography, Van Nostrand Reinhold, New York (1970). 61. Mason, L.F.A., Photographic Processing Chemistry, Focal Press, St. Louis, MO (1984). 62. Van Hunsel, J., Coppens, P., Deprez, L., Odeurs, R., and VandenBergh, D., “Silver diffusion transfer technology for CTP imaging,” Proceedings of the International Congress on Image Science, ICPS’98, 7, 9, pp. 1435–1438 (1998). 63. Rott, A., and Weyde, E., Photographic Silver Halide Diffusion Processes, Focal Press, St. Louis, MO (1973). 64. Nisato, G., Bouten, P.C., Slickkerveer, P.J., Bennet, W.D., Graaf, G.L., Rutherford, N., and Wiese, L., “Evaluating high performance diffusion barriers: the calcium test,” Proc. of Asia Display, IDW01, Society for Information Display, Nagoya, Japan, pp. 1435–1438 (2001).
300
Chapter 5
65. Brabec, C.J., “Organic photovoltaics: technology and market,” Sol. Energy Mater. Sol. Cells, 83, pp. 273–292 (2004). 66. Lewis, J., “Material challenge for flexible organic devices,” Mater. Today, 9, pp. 38–45 (2006). 67. Crank, J., and Park, G.S., Diffusion in Polymers, Academic Press, New York (1968). 68. Firon, M., Trouslard, P., and Cros, S., CEA/DRT/LITEN/LCS, French Patent No. 06/01320. 69. Burrows, P.E., Graff, G.L., Gross, M.E., Martin, P.M., Shi, M.K., Hall, M., Mast, E., Bonham, C., Bennett, W., and Sullivan, M.B., “Ultra barrier flexible substrates for flat panel displays,” Displays, 22, pp. 65–69 (2001). 70. Cros, S., Firon, M., Lenfant, S., Trouslard, P., and Beck, L., “Study of thin calcium electrode degradation by ion beam analysis,” Nucl. Instrum. Methods Phys. Res. B, 251, pp. 257–260 (2006). 71. Bujas, R., and Roko, S., General Atomic, U.S. Patent No. 6,804,989. 72. Dunkel, R., Bujas, R., Klein, A., and Horndt, V., “Method of measuring ultralow water vapor permeation for OLED displays,” Proc. IEEE, 93, pp. 1478– 1482 (2005). 73. Andersen, M., Carlé, J.E., Cruys-Bagger, N., Lilliedal, M.R., Hammond, M.A., Winther-Jensen, B., and Krebs, F.C., “Transparent anodes for polymer photovoltaics: Oxygen permeability of PEDOT,” Sol. Energy Mater. Sol. Cells, 91, pp. 539–543 (2007). 74. http://www.saesgetters.com/ 75. Pichler, K., “Conjugated polymer electroluminescence: Technical aspects from basic devices to commercial products,” Philos. Trans. R. Soc. London Ser. A, 355, pp. 829–842 (1997). 76. Krebs, F.C., “Encapsulation of polymer photovoltaic prototypes,” Sol. Energy Mater. Sol. Cells, 90, pp. 3633–3643 (2007). 77. Dennler, G., Lungenschmied, C., Neugebauer, H., Sariciftci, N.S., Latrèche, M., Czeremuszkin, G., and Wertheimer, M.R., “A new encapsulation solution for flexible organic solar cells,” Thin Solid Films, 511, pp. 349–353 (2006). 78. http://www.g24i.com/ 79. http://www.solarcoating.de/ 80. www.konarka.com/news_and_events/press_releases/2004/2_february/0224_ dupont.php
Chapter 6
Outlook Frederik C. Krebs The field of organic photovoltaics (OPVs) is moving rapidly and is, in many ways, advancing quickly in the wake of the OLED and PLED technologies, which have experienced a strong commercial push. OPVs share many of the technological challenges and design advantages with the aforementioned technologies. Since the device geometry in many aspects is identical and all three technologies employ a thin active layer sandwiched between two electrodes, the available modes of preparation for all three technologies are similar. Many solutions have been developed for OLEDs and PLEDs that are directly transferable to OPVs, and most notably the device structure is in many ways very similar. The desire to have a flexible device that is of low cost and that can be tailored to meet the needs of demanding designers and consumers have also driven the vision of ubiquitous plastic solar cells. Globally, the market for consumer products has seen a change and as the cost of consumer products have decreased during the past two decades, there has also been a shift in the attitude of the consumer, which is projected directly to the producers and developers of technology. The low-cost and ready availability has induced a state in the consumers where the technology is expected to be cheap, reliable, and easily replaceable. During the microelectronics era, from the 1960s through the 1980s, electronics were expensive and the rigid conditions of production led the consumers and designers to accept the size, shape, feel, color, etc. of the technology. In a sense, the technology dictated the available products. From the 1990s and until today, this has changed and consumers and designers now dictate what the technology has to look like and how it has to perform. This leaves researchers under a lot more pressure, and the high pedestal that technologists used to enjoy has now been reduced to a somewhat more balanced and service-oriented one.
6.1 Where Is the Technology Now? OPV technology is currently at a crossroad between large-scale production for ongrid electrical energy production and production for niche applications. While there is some development missing in terms of setting up a production line for OPVs, the technology as such is ready and the decision rests in many hands, including the hands of investors, and the faith is in the hands of the consumers. The general
302
Chapter 6
consensus among scientists and technologists is that the technology is not ready for large-scale on-grid electrical energy production, and the major reason for that is the relatively low power-conversion efficiency. As discussed in previous chapters, the major showstopper has been a very short operational lifetime, but solutions to this have been demonstrated; and from that point of view, niche applications are possible. It is important that the right path is chosen, and the current opinion is that this path runs through niche applications. The development of production techniques for niche applications and the establishment of a niche market are believed to serve as leverage for further exploitation as the technology ripens with respect to stability and power-conversion efficiency. Large-scale on-grid electricity production at low cost [<1 euro/Wp (Wp = watt peak)] is then a likely future scenario. There is no doubt that the OPV technology is applicable, and the worst thing that can happen is that the initial choice of products and applications fail in the eyes of the consumer because this could make future developments problematic in terms of reluctance from investors. It must be remembered that OPV technology currently is a novel technology around 20-years old, and as such is immature. It has to enter a market and compete with an existing and established silicon technology that has been around for 50 years. Furthermore, it is in many aspects inferior to technologies such as silicon-based PVs. Thus, a careful scheme for market entry will have to be chosen that encompasses all the strengths of the technology, namely, low cost, flexibility, fast production, low thermal budget, and environmentally friendly modes of disposal. At the same time, the weaknesses have to be avoided, including low power-conversion efficiency and short lifetime.
6.2 Where Is It Suitable? This is a very important question and, like any example of technology-driven innovation, it is very hard to predict who your consumer is going to be and what type of intellectual property rights the producer will need to establish and maintain market shares. During the next decade or so, OPV technology will have to face the music in the sense that commercial exploitation is inevitable. Manufacturers of equipment for OPV production and manufacturers of devices that employ OPVs will, in this time span, uncover the suitability of the technology. The most likely applications will be disposable products where a source of electricity is needed, and the market segment that the technology competes with in this aspect is small batteries and power supplies. The edge for the technology in such a case is mainly environmental friendliness; and since the product is assumed to have a short service life, the major requirement is an acceptable shelf life. Currently, shelf lives of many months are possible. A second type of application could be in medical or sensory equipment where the OPV is a part of a sensor that is disposable. In terms of more stable products, the technology should replace electrical supplies in the form of wires and, in a sense, give the consumer a freedom to apply the product wherever needed. There are many types of products where an OPV technology can enter directly, includ-
Outlook
303
ing pocket calculators, timers, outside thermometers, USB chargers, MP3 players, small radios, etc. It will be in competition with existing low-cost PV technologies such as amorphous silicon or cadmium telluride. From this point of view, the battle for market segments from existing and established suppliers could be a losing one, and caution in the choice of product should be advised.
6.3 Where Could It Be in the Next Decades? The vision of a plastic future is there, and in many ways the answer to this question lies in the hands of designers and developers who have a feel for the needs of the people. It is a mind game and, given that designers are provided with a fabric that is malleable enough for them to shape and form into products that change the world around us, it is very likely that a symbiosis between science and design will become pervasive. It could be very productive and the largest challenge is for the scientists to teach designers about the limitations and requirements of the technology. As a scientist or technologist, the work with designers can be both stimulating and frustrating. The work is stimulating in the sense that the scientist will attempt ideas that they would not have, had they not met the designer in the first place (or at least the likelihood of such attempts would be very small). To round off this very short discussion with an example, attempts have been made to incorporate OPVs into clothing. From the designers point of view, most examples of PVs in clothing consist of stitching a PV onto the piece of clothing in question. From a design point of view, this does not solve any problem and does not justify the existence of the designer because no design value is added and only the functionality of the two products (the clothing and the PV) are combined. If the PV could be truly incorporated into the clothing in such a way that the clothing had the function of an electrical energy supply without suffering from it in an esthetic sense, then the designer would have fulfilled his or her purpose. A first attempt1 involved sewing flexible PVs into both male and female clothing.2 As shown in Fig. 6.1, the integration of the flexible PET foil with an MEH-PPV-C60 heterojunction was successful. The designer succeeded in choosing a particular shape of the device (triangle and diamond shaped) and further chose a pattern for the aluminum counter electrode. The shape of the aluminum electrode was chosen from an esthetic point of view. The technological approach would imply an aluminum electrode that would cover as much of the area as possible and that would minimize resistive losses. Such an electrode would most likely consist of stripes. The reaction of the designer to this was the pattern shown in figure, which has obvious problems with current extraction. The message and challenge is that designer and scientist need to work together and perhaps create a product that will both sacrifice and gain in terms of science, technology, and design. The scientist will sacrifice efficiency but will gain value, and the designer will sacrifice esthetics due to limits in choice of shape but will gain in usefulness. The designers also experimented with the idea of a spatial design of the clothing (pyramids) that stick out and, in
304
Chapter 6
Figure 6.1 Pictures of a design project where a PET-ITO-MEH-PPV-C60 -Al heterojunction was employed. The shape of the aluminum counter electrode is seen as a dark color in the red devices (triangles and diamond shapes). The power supply for the light bulbs is not the OPV, but external. Design by Tine Hertz.
principle, optimize the angle with the incidence of the light at least some of the time. In the above example, the designer successfully incorporated the OPV into a dress and a futuristic belt, but failed somewhat in terms of making the OPV blend in with the clothing. A second attempt was thus made employing a screen printed OPV that was prepared directly on cloth. In this way, flexibility was gained and the OPV was naturally incorporated into the clothing as a pattern that looks no different from any other pattern or print on contempory clothing.1 This is shown in Fig. 6.2 where it is evident that the geometric fill factor is low with the active solar cell area being less then 20% of the total area. There
Outlook
305
Figure 6.2 A picture of a screen printed PE-PEDOT-MEHPPV-C60 -Al device. The device was printed on PE covered cellulose based woven cloth. There is good diffuse transparency. Design by Tine Hertz and Maria Langberg.
are also obvious problems with current extraction, which is at opposite corners. The design idea, however, was that the produced pieces of cloth could be sewn together, thereby connecting the solar cells; and while the device performance suffered dramatically, the design gain was large and the objective was reached. The combined scientific, technological, and design work won the INDEX 2005 prize (for amateurs).3 It should be evident from the above examples that there is no product that will suit the average consumer, but a clear vision of a product has been
306
Chapter 6
presented and materialized in the form of a functional device prototypes. Even though the devices presented here were not stable and certainly would not wash, it does show that there are possibilities, and the answer to the question of where OPV technology could be in a decade lies in the combined efforts of scientists, technologists, designers, producers, investors, and consumers. The technology is flexible and adaptable. The possibilities are all there and it will be exciting to follow how OPV technology makes its way into our everyday life.
References 1. Krebs, F.C., Biancardo, M., Winther-Jensen, B., Spanggaard, H., and Alstrup, J., “Strategies for incorporation of polymer photovoltaics into garments and textiles,” Sol. Energy Mater. Sol. Cells, 90, pp. 1058–1067 (2006). 2. www.risoe.dk/solarcells (2007). 3. http://www.indexaward.dk (2007).
Index 1,2,4-trichlorobenzene, 136 1,2-dichlorobenzene (ODCB), 13, 141 1,3-dipolar addition, 36 1 m2 module, 264 1-(3-(methoxycarbonyl)propyl)-1phenyl[6,6]-C61 , 62 3-alkyl thiophenes, 17 5-oxo-5-phenyl-pentanoic acid methyl ester, 61 5-toluenesulfonylhydrazono-5-phenylpentanoic acid methyl ester, 61 {6}-1-(3-(methoxycarbonyl)propyl){5}-1-phenyl[5,6]-C61 , 61 [60]PCBM, 15, 42, 48 [70]PCBM, 15, 37 π–π stacking, 136 A absorption coefficient, 131 accelerated lifetime measurements, 156, 217 apparatus, 220 study, 206 acceleration factor, 219 activation temperatures, 251 active device area, 245 addition of oxygen, 185 adhesion effect, 255 adhesive aluminum tape, 66 adhesive forces, 133 Ag pastes, 275 aggregates, 138 air, 139 mass, 4, 92 airflow, 135 Al diffusion, 143 Al electrode, 142, 143 Al+ , 165 Al+ /In+ , 177 Al/C60 /C12 -PPV/PEDOT:PSS/ITO, 173 Al/C60 /P3CT/ITO, 173
Al/P3HT:PCBM/PEDOT:PSS/ITO, 141, 143, 144 alignment, 136 alkyl substituted PPVs, 15 all-in-one molecule, 40 AlO− 2 , 161, 166 Alq3 , 190 aluminum, 139, 177, 267 oxide, 139 AM 0, 93 AM 1.5G, 93 Amicon pastes, 277 angle of attack, 236 angular velocity, 134, 135 annealing, 132, 133, 140, 142, 143, 146–148 conditions, 140 temperature, 140–143 time, 140 anode, 266 apparent yield stress, 248 applying electrodes, 65 applying filters to improve the spectrum, 100 arc lamp, 91 architecture, 131, 132 Arrhenius equation, 218 like behavior, 251 arrival rate, 138 uniformity, 138 ASTM E 927-05, 94 E490-00, 92 atmosphere, 135 atomic force microscopy (AFM), 18, 144, 146, 212 Auger spectra, 189 B background pressure, 139 ballistic deposition, 139 propagation, 137–139 bandgap, 20
Index
308
barrier improvement factor, 282 layer, 139, 155 materials, 280 batch printing, 237 type methods, 233 bathocuproin (BCP), 35 batteries, 1 bendable module, 292 benzo-bis(thiadiazole), 28 benzothiadiazole, 25, 27 biexponential decay, 219 binary solvent, 135 Bingham fluid, 248 biodiesel, 5 bioethanol, 5 biofuels, 5 biomass energy, 3 bithiazole, 31 blend cell, 132, 136 boat, 137, 139 source, 138 boiling point, 136 bolometric power measurement, 100 bolometric pyranometer, 101 Brabec triangle, 8 browning phenomenon, 112 brushing, 135 bulk heterojunction, 15, 36, 146 C C12 -PPV, 160 C2 Al− , 160 C− 4 , 160 C60 , 13 Ca passivation, 204 cadmium telluride (CdTe), 6 calcium, 201, 267 test, 283 calibration of the sun simulator, 103 capacitance, 109 capacitive loading, 108 capillary forces, 236 carrier substrates, 266 cathode, 131, 140, 266 aging, 210 degradation, 201 cell performance, 140 centrifugal forces, 133 characterisation by NMR and UV-vis spectroscopy, 57 of organic solar cells, 91 techniques, 144 charge carrier mobilities, 132, 135 mobility, 132, 140
transport, 140, 147 chemical characterization, 144 chemical degradation mechanisms, 183 chemical shifts, 189 chemical structure elucidation, 178 chemical vapor deposition, 138 chemically gasified solid, 138 chemisorption, 190 chlorobenzene, 141, 145–147 chloroform, 136, 137 cliché, 237 clothing, 304 clusters, 138 CN-MEH-PPV, 17 coal, 2 coalescence processes, 139, 141 collisions, 139 color code, 241 color histograms, 241 columnar structure, 35 compatibility, 135 concentration, 134 conductive inks, 275 conductivity, 140 confocal laser scanning fluorescence microscope, 216 contamination, 138 controllability, 134 convective flow, 135 cooling, 140 rate, 140 copper indium-gallium diselenide [Cu(In,Ga)Se2 ], 6 copper phthalocyanine (CuPc), 7 corrosion of metallic calcium, 201 critical shear rate, 248 cross section, 145 crystal defects, 140 lattice, 140, 147 overgrowth, 144 crystalline order, 132 crystallinity, 132, 147 crystallographic parameters, 138 crystallographic properties, 140 crystallography, 138 curing glues, 287 current-voltage-luminance (I-V-L), 245 cyano substituted MEH-PPV, 15 cyanovinylene terphenylene, 32 cyclic voltammetry, 20 D defects, 15, 135 degradation constant, 218
309
mechanisms, 155, 210 using XPS, 187 products, 183 degree of coiling, 135 Dektak profilometer, 243 delaminated module, 265 delamination, 265 deposition, 133 behavior, 139 techniques, 133 depth profiling, 158 scale, 145 determination of the molecular weight, 55 device aging, 191 preparation and performance, 68 dielectric relaxation, 108 differential scanning calorimetry (DSC) measurements, 142 diffraction pattern, 147 diffusion, 138, 139 coefficient, 280 free sealing techniques, 288 phenomena, 167 through the Al grains, 175 dilatancy, 248 dip coating, 135–137 diphenyl-dimethyl-phenanthroline, 35 disk, 133 distribution of P3CT, 172 DL-CuPC, 35 doctor blade, 237 doctor blading, 233, 237 donor-acceptor, 131 domains, 146 interface, 132 interpenetrating networks, 144–146 double-layered cells, 131 drop-casting, 135 Dryflex getter, 289 dust, 139 dyad, 40 dye-sensitized nanocrystalline photoelectrochemical cells, 129 E e-beam evaporation of aluminum, 264 Eccobond paste, 277 ECN solar energy, 116 efficiency, 132, 142–144 efficient encapsulation of OPV modules, 265 elastic recoil detection analysis (ERDA), 197 elasticity, 135
electric arc lamps, 91 electrical barrier layer, 140 electrical carrier traps, 139 electrical field, 107 electron acceptor, 131, 132, 136 beam physical vapor deposition, 138 donor, 131, 132 microscopy, 132, 145 emulsion coating, 237 encapsulation, 279 and permeability, 279 techniques, 286 environmental effects, 112 EP patent application 93920199.2, 294 epoxy-based Ag paste, 277 epoxy-based carrier materials, 275 equilibrium conditions, 140 equivalent circuit, 231 evaporate, 134, 138 evaporating the electrode, 64 evaporation, 138 rate, 138, 139 time, 138 exciton, 131 exciton diffusion length, 131, 137 range, 131, 132 dissociation, 131, 132 experimental parameters, 137 exponential decay, 219 external quantum efficiency (EQE), 245 F failure mechanism, 217, 221 female clothing, 303 ferric chloride polymerization, 18 fiber-optic/minidish concentrator, 129 Fick’s first law, 280 film splitting, 255 thickness, 134, 138 fish-bone diagram, 239 flakes, 139 flat-panel displays, 269 flexibility, 280 flexible encasement, 290 foils, 257 PET foil, 303 flexible ultrabarrier material, 292 flexo/pad printing, 233 flexography, 234, 239 flow coating, 135
Index
310
curve, 247 fluorene, 29 fluorescence, 216 images, 216 microscopy, 216 screen, 145 flywheels, 1 fossil fuel, 1 fragment ion, 184 G G24i, 293 gallium arsenide (GaAs), 6 gas-phase ion chemistry, 183 gelation phenomena, 250 geometric fill factors, 263 geothermal energy, 3 getter materials, 286 Gilch polymerization, 15 reaction, 14 glass transition temperature, 140, 142 glove box, 68 gold, 139 grain, 140 boundaries, 139, 140 growth, 140 sizes, 139 gravure, 234, 239 growth, 138 H H18 2 O, 159, 222 halogen lamps, 97 hardness, 140 head-to-tail, 17 heating boat, 138 heating coil, 138 Heck-type coupling, 15 height resolution, 146 hexabenzocoronene, 35 high carrier mobility, 15 high-performance liquid chromatography (HPLC), 52 high-power spectrometer, 109 high-sensitivity permeation instruments, 283 highly conductive PEDOT (HC-PEDOT), 269 history of degradation, 185 hole acceptor poly(9,9-dioctylfluoreneco-bis-N,N-(4-butylphenyl)-bisN,N-phenyl-1,4-phenylenediamine, 136 homopolymers, 137 Horner-Wadsworth-Emmons reaction, 15
human energy consumption, 1 hydrogen, 1 hydropower, 3 I IEC 904-9 standard, 94 impedance, 109 impurities, 133, 139 In+ , 165 In+ /Al+ , 177 incident photon to current efficiency, 110 incorporate dopants, 140 incorporated, 138 INDEX 2005 prize, 305 indium, 177 tin oxide, 267 induced strain, 247 ink, 234 formulation, 245 jet printing, 234, 239 paste, 239 InO− 2 , 166 inorganic transparent conductive oxide, 267 insulator, 140 integration of solar cells, 229 intellectual property rights, 293 intensity distribution patterns, 176 interaction chromatography, 52 interchain spacing, 147 interdigitated alkyl chains, 147 interface, 131, 133 Al/C60 , 166 Al/LiF, 196 Al/P3HT:PCBM, 142 chemistry, 166 ITO, 175 interfacial area, 132 contact area, 143 interference microscopy, 211 interlayer diffusion, 155, 167 mixing, 165 internal stress, 140 interpenetrating network, 132 intramolecular charge transport, 132 IPCE measurements, 110 IPR portfolio, 293 isothianaphthene (ITN), 27 isotope labeling, 156, 159, 220 isotopic markers, 201 isotopically labeled water, 222 Israeli Meteorological Service, 115 Israeli National Physical Laboratory, 115
311
J JP-20006-080530A, 294 junction, 132 K Keithley 2400, 67 Konarka Inc., 293 Kumada coupling, 17 L lamella microstructures, 136 laminated, 261 lamination, 280 Langmuir–Blodgett technique, 33 large module, 262 large rigid encasement, 289 larger screen printed photovoltaic module based on MEH-PPV, 260 lateral resolution, 145, 146 LED, 245 letterpress, 239 Li and F distribution, 193 LiF layer, 190 lifetime, 155 light source, 91 linear decay, 219 Linz Institute for Organic Solar Cells, 116 liquid crystalline display (LCD) panels, 269 liquid-vapor interface, 134 lithographic, 239 long-term characterization, 113 long-term outdoor testing of stability of organic solar cells, 123 low-conductive PEDOT, 270 M magnesium, 267 MALDI-TOF, 157 mapping the history of degradation, 185 mask, 137, 235 mass spectral information, 178 mass spectral marker, 160 mass spectral markers, 165 mass spectrometry, 285 materials, 47 matrix, 145 McCullough route, 17 MDMO-PPV, 12 MDMO-PPV:PCBM, 145, 147 measurement of permeability, 279 of the diffusion coefficient, 281 mechanical barrier, 261 mechanical flexibility, 268
mechanism for the particle formation, 181 median color code, 241 MEH-PPV, 12, 190 MEH-PPV:PCBM, 164 mesh size, 243 mesoscopic order, 132 metal, 137 alloys, 137 evaporator, 42 metallic grid, 272 microcrystalline structures, 136 microscopic holes, 212 microstructure, 136, 140 millimeter range, 145 miscibility, 132 mismatched cells, 231 mobility, 136, 138 molecular diffusion, 139, 141 molecular ordering, 132, 136, 147 molecular packing, 132, 140 molecules, 34 monitor, 144 monitoring photooxidation, 185 monocrystalline silicon solar cell, 6, 120 monolithic production, 229 morphological control, 141 morphological controllability, 135 morphological parameters, 134 morphology, 131–133, 136, 137, 141, 142, 144, 145 mottle, 241 N nanometer range, 145 nanoscale interpenetrating network, 132 morphology, 132 phase-separation, 136 nanostructure, 146 natural gas, 2 Negev Desert, 113 Newtonian fluid, 247 Newtonian plateau, 248 non-Newtonian behavior, 134, 247 nuclear fusion, 2 reactors, 2 nuclear reaction analysis (NRA), 197, 202 nucleation, 138–141 nylon, 234 O
18
O incorporation, 163 O incorporation/excange, 176 18 O2 , 159, 222 18
312
18
O2 gas cylinder, 222 O/16 O ratio, 159 ocean energy, 3 offset, 234 printing, 135 oil, 2 OLED, 14, 301 oligomers, 34, 38 oligophenylenevinylene, 38, 39 oligothiophenes, 41 one-dimensional patterns, 238 operational lifetime, 218 optical microscopy, 133 optical spectrum analyzers, 99 optoelectronic devices, 246, 268 Orgacon EL-P 3040, 271 organic photovoltaics (OPV), 301 into clothing, 303 other printing methods, 239 outdoor IV measurements, 120 outdoor measurements, 112 outdoor solar irradiance, 124 outside test, 265 oxidative ferric chloride polymerization, 26 oxide, 139 oxido-de-sulfonato substitution, 181 oxygen, 139, 155 accumulation, 193 and/or water reaction products, 180 diffusion, 161, 167, 211 incorporation, 203 permeation in PEDOT, 285 transmission rate (OTR), 286 ozone treater, 46 18
P P3HT, 136, 141, 142 crystal structure, 147, 148 purification of the crude, 52 P3HT:PCBM, 147, 148 pad printing, 237 parallel connection of solar cells, 231 particle formation, 155 in organic solar cells, 178 paste volume, 235 PBBT, 28 PBPT, 27 PBT, 25 PCBM, 13, 36, 141 aggregates, 142 PCBM:polyfluorene solar cells, 123 PEDOT:PSS, 13, 46, 141 Peltier element, 250 PEOPT, 18 percolated pathways, 132 permeability coefficient, 280
Index
perylene tetracarboxylic acid (PTCA), 7 PET, 257 PF, 29 PFO, 190 phase separation, 137 -separated networks, 147 photodegradation, 155, 189 photo-oxidation, 210 photophysics, 131, 132 photoresist, 271 photosensitive emulsion layer, 235 photovoltaic performance, 132, 141, 142 phthalocyanine, 35 physical processes, 137 physical vapor deposition, 138 physicochemical properties, 132, 140 plateless, 239 PLEDs (polymer light-emitting diodes), 14, 301 poly(3,4-ethylenedioxythiophene), 13 poly(3-carboxydithiophene) (P3CT), 156, 162 marker profiles, 163 poly(3-hexylthiophene), 49 poly(9,9-dioctyl-fluorene), 190 poly(9,9-dioctylfluorene-cobenzothiadiazole, 136 polycrystalline film, 139 polycrystalline silicon cells, 6 polyester, 234 poly(ethylene-terephthalate), 266 poly(ethylenenaphthalate), 266 poly(isothianaphthene) (PITN), 25 poly(isothianaphthene) (PITN) with a monomer of a benzene ring fused to a thiophene, 26 polymer chain, 137 compatibility, 135 solar cell from scratch, 42 -solvent compatibility, 135 interactions, 135 poly(phenylenevinylenes) (PPV), 12, 13 polystyrene-polydimethylsiloxane, 31 poly(styrenesulfonate), 13 poly(thiophenes), 17 polyurethane, 236 POMeOPT, 18 porphyrin, 32 powder diffraction, 147 practical fabrication, 42 Prato reaction, 38
313
pressure-curtain coating, 135 pressure-sensitive adhesive layer, 279 primary electron beam, 145 principle of spin coating, 133 print conditions, 239 printing and coating methods, 232 frame, 235 speed, 240, 255 the active layer, 239 process parameters, 138 processing and production of large modules, 229 processing conditions, 141 processing of the opaque back-side contact, 274 processing of the transparent front-side contact, 267 production and companies 2007, 291 profilometric measurements, 241 protective barrier, 210 protrusions, 212 pseudo-plasticity, 247 pseudo-plastic flow, 253 PSS derivative, 181 PTOPT, 18 PTP, 28 PTV, 30 pulsed laser deposition, 138 pyranometer, 115, 265 pyrrole, 27 Q quantum yield, 132 quasi-solid-state dye-sensitized solar cells, 127 quinoid resonance structure, 27 R R2R coating, 232 radial flow, 134 reactive oxygen plasma, 267 reciprocal space, 147 recombination, 131, 141 recording the spectrum, 99 recovery effect, 127 recrystallization, 140 rectification ratio, 74 red-green-blue color scheme, 241 regiorandom P3HT, 18, 49 regioregular, 17, 42 P3HT, 13, 141 P3HT via the McCollough route, 50 P3HT:PCBM, 146 relative humidity, 135 removal of Al electrode, 167 removal of the C60 and C12 -PPV layers, 168
removal of the PEDOT:PSS layer, 168 renewable energy sources, 2 rheological characterization, 246 rheological properties, 134 rheology, 246 rheometer, 249 rheopexy, 249 Rieke-zinc, 17 rigid encapsulation, 279 rigid encasement, 287 roll-to-roll coating, 232 roller coating, 135 rotary screen printing, 234 rotational acceleration, 134 roughness, 138 rubber, 236 ruthenium terpyridine complex, 40 Rutherford backscattering spectroscopy (RBS), 156, 197 S S− , 163 SAES getters, 289 scale of phase separation, 146 scanning electron microscopy (SEM), 144, 213 image, 145, 214 scanning probe microscopy, 132 scattering techniques, 133 Schottky, 131 screen printed active layer, 254 screen printed layers of donor/acceptor blends, 254 screen printed MEH-PPV, 261 screen printed OPV, 304 screen printed silver connections, 261 screen printing, 233, 234, 239 SEC chromatogram, 55 secondary electron image, 145 secondary electrons, 145 Sede Boker, 113 self-assembling, 32 self-organize, 135 self-organizing molecular materials, 35 self-organizing properties, 136 semiautomatic screen printer, 254 semiconducting polymer, 136 series connection, 230 series resistance, 144 sewing flexible PV, 303 shadow mask, 245 shear rate–dependent viscosity measurements, 252 shear thinning, 134, 247 shear-thickening, 248 shear-thinning regime, 252
314
shearing stress, 247 sheering, 134 sheet resistance, 257 sheet resistivities, 267 shelf life, 218 Si photodiode, 245 Siegrist reaction, 31 Siemens AG, 293 silicon solar cell, 6 SIMNRA, 200 SIMS ionization process, 157 single-crystal diffraction, 147 single-layer cells, 131 single-layer device, 131 size-exclusion chromatography, 52 skin, 135 skinning, 135 small rigid encasement, 287 small screen printed module, 261 snap-off distance, 236, 244 solar energy, 3, 4 irradiance, 117 simulator, 97 spectrum, 19 Solarcoating Machinery GmbH (SCM), 293, 295 Solarkonstant 575, 67 solubility coefficient, 280 solution concentration, 134 viscosity, 134 solvent, 134, 145, 146 evaporation, 134–136, 141 -solvent interactions, 135 vapor, 135 volatility, 135, 136 Soret band, 32 source meter, 67, 106 temperature, 138, 139 spheres, 145 spherical P3HT nanostructures, 146 spin casting, 133 spin coating, 132, 133, 135, 136, 233 parameters, 136 processes, 133 spin speed, 133, 134 spray coating, 135 sputter deposition, 138 squeegee, 236 blade, 236 edge profile, 236 pressure, 244 stability, 155 and lifetime of organic solar cells, 113
Index
standard test conditions, 113 stepwise and unidirectional synthesis, 39 Steuernagel Lichttechnik GmbH, 67 Stille cross-coupling reaction, 27 strength, 140 structure, 133 SubPc, 35 substrate, 43 reactivity, 138 temperature, 137–139, 141 sulfonium precursor route, 13 sun simulator, 67 surface segregation, 167 surface topography, 135 Suzuki cross-coupling polymerization, 27 synthesis and purification of PCBM, 58 T taking the sun inside, 91 Tang, 7 TE, 137 TEM, 144, 145 temperature dependence of the photovoltaic parameters, 116 -dependent viscosity measurements, 250 temporal stability, 97 thermal annealing, 141 time, 144 thermal degradation, 155 thermal evaporation, 132, 133, 137, 141 thermal evaporation of aluminum, 264 thermal evaporator, 42 thermocleavable materials, 33 thermocleavage reaction, 34 thermodegradation, 189 thermosetting silver epoxy, 66 thiadiazolequinoxaline, 29 thin film, 133, 134 thin film-technologies, 6 thixotropy, 249 titanium, 267 TOF-SIMS, 156 depth profiling, 159, 166 imaging, 167, 170, 174 toluene, 145, 147 topographic map, 146, 211 topography, 138 transparency, 280 effect, 145 transparent oxides, 267 tritiated water, 285
315
two-component system, 275 types of simulators, 97 U ultrabarrier materials, 266 ultrasonic bath, 46 unification challenge, 9 uniformity, 134, 138–140 University of California, 294 U.S. application 07/930,161, 294 U.S. patents, 293 No. 5,331,183, 293, 294 No. 5,454,880, 293 UV photolithography, 245 V vacuum chamber, 137 evaporation, 137 thermal evaporation, 137 vapor, 139 deposition, 137 vertical distribution, 167 vertical phase segregation, 141 viscosity, 134, 247 curve, 247 voids, 139 volatile solvents, 134
W water, 139, 155 vapor, 201 wetting agents, 236 wind energy, 3 withdrawal speed, 135 WO/1994/005045, 294 World Meteorological Standard, 115 woven fabric, 234 WVTR, 280 X x-ray, 144, 147 diffraction (XRD), 147 grazing-incidence, 18 spectra, 148 photoelectron spectroscopy (XPS), 156, 157, 187 principle, 188 spectra, 189 xenon arc lamp, 111 xylene, 136, 137 Y Yamamoto coupling, 27 Z zero-shear viscosity, 248 zinc-porphyrin, 40
Frederik C. Krebs began his university studies in Aberdeen (Scotland) where he obtained bachelor-ofscience degrees in chemistry (1993) and biochemistry/immunology (1994). He then went to Université de Nantes (France) where he obtained a DEA (1995) in the areas of solid state chemistry. He returned to Denmark and studied dielectric materials with pyroelectric properties leading to a Cand. Scient. degree (1996). Further work centered on the synthesis of organic dielectrics with a polar axis and studies of their crystals by both neutron and synchrotron x-ray methods led to a Ph.D. (2000) at the Technical University of Denmark. Post doctoral studies at Risø National laboratory (2001–2002) were directed toward plastic solar cells starting mainly with synthetic efforts and materials characterization using synchrotron based ultraviolet photoelectron spectroscopy and mobility measurements employing transient microwave measurements by radiation doping through irradiation with high energy electrons. He was then employed as Senior Scientist at Risø National Laboratory (2002–present) and today the efforts are concentrated on large-scale preparation of polymer photovoltaic devices and their characterization. His group currently has three areas of focus; synthesis of materials with low band gap and properties that give stable solar cells and that can be processed, stability and degradation studies using sensitive techniques with the aim of improving polymer solar cells stability, processes and techniques for producing large area polymer solar cells. Currently he acts as associate editor for the international journal Solar Energy Materials and Solar Cells and has published more than 160 peer reviewed papers, conference proceedings, editorials, book reviews, patents and reports. Tom Aernouts received his master-of-science degree in physics in 1999 from the Katholieke Universiteit Leuven on the characterization and simulation of organic oligomer-based diode structures. Continuing the research on organic semiconductors, he joined the organic photovoltaics group at the Interuniversity Micro-Electronics Center (IMEC) in Leuven, Belgium, where his work focused on the processing and characterization of polymer-based organic solar cells and monolithic modules, which resulted in several journal publications, conference contributions and invited talks. In September 2006, he received his Ph.D. degree on this topic from the Katholieke Universiteit Leuven. Dr. Aernouts is currently a senior scientist at IMEC, supervising the polymer-based work at the
Organic Photovoltaics group. His main research interest is in the introduction of printing technology in this field. Rémi de Bettignies graduated from the engineering schools Ecole Superieure d’Electricité (Supélec, Paris) in 2000, majoring in electronics and solidstate physics. In parallel, he obtained in 2000 a DEA (Diplômes D’Etudes Approfondies) from University Paris VI in optoelectronics. He received his Ph.D. in 2003 from the University of Angers at the Organic Photovoltaic Solar Cells group, directed by J.M. Nunzy and J. Roncali. From 2003 to 2005, he worked as a postdoctoral fellow in the Laboratory of Organic Components at the CEA/DRT (Saclay), where he studied the modelling of organic solar cells and the ageing processes. He has authored two patents and several publications. Since 2005, he has been working for the CEA-INES laboratory, the French National Institute for Solar Energy, in Chambéry, on organic photovoltaic solar cells. Eva Bundgaard graduated from the Technical University of Denmark (DTU) in 2003 as a master of science in chemistry, specializing in organic chemistry, where she conducted several projects concerning, for example, synthesis of pectin (carbohydrate chemistry) and natural compounds from carbohydrates (metalorgano-chemistry). In 2004, she began her Ph.D. project at Risø National Laboratory and Roskilde University Centre (RUC), focusing on the synthesis of low band-gap polymers for organic photovoltaics. She received her Ph.D. in 2007 and is currently working at Risø National Laboratory as a post.doc. While studying at both DTU and at Risø National Laboratory, she has conducted research projects abroad. At the Macquarie University in Sydney in 2002, the research project focused on the synthesis of a novel human UV filter compound. She also spent six months at the National Renewable Energy Laboratory in Colorado, U.S., in 2006 where she conducted research in organic photovoltaics with low bandgap polymers she prepared at Risø National Laboratory. Ms. Bundgaard has published six peer-reviewed journal papers and three peer-reviewed proceedings papers within the field of organic synthesis, organic photovoltaics, and low band-gap polymers.
Stéphane Cros obtained a DEA (Diplômes D’Etudes Approfondies) in physics of solids in 1998, and entered in 2002 a Ph.D. program in ESPCI (Ecole Supérieure de Physique et de Chimie Industrielle) in Paris emphasizing synthesis and mechanical properties of multilayered PMMA-nanosilica materials by interface modification. In parallel of the first year of his Ph.D. thesis, he obtained a DEA from Université Pierre & Marie Curie (Paris VI) in macromolecular physics and chemistry. He received his Ph.D. in 2002 and joined an ATER position (research and teaching) in the Laboratory of Macromolecular Materials at the CNAM (Conservatoire National des Arts et Métiers) in Paris where I focused on industrial polymer processing and characterisation. He then spent 6 months working on polymer blends for the CRITT Picardie (center of technology transfer) to write a technical rapport for PME (little firms). In 2004, Dr. Cros joined a postdoctoral position in the Laboratory of Organic Components at the CEA-Saclay, where he studied the barrier properties of sealing materials for organic solar cells. His research interest was focused on organic-inorganic multilayered materials, permeability measurement with a new experimental permeameter and ion beam analysis of the degradation of inorganic layers. He currently has a permanent position in the CEA-INES laboratory, the French National Institute for Solar Energy, in Chambéry with the same research subjects. Muriel Firon received her Ph.D. in physical chemistry at the University of Paris VI in 1994, while working for IBM. She worked on plasma deposition processes (RFPECVD) and contributed to optimizing steps of semiconductor production. In June of 1995 she began working on the deposition and electrical characterization of thin silicon oxide films deposited by PECVD (DECR plasma) at Pr. Bernard AGIUS laboratory (University of Paris XI Orsay). She then joined the CEA/DAM in 1995 and was in charge of the development of thin solid films deposited by PVD on large plastic areas with roll-to-roll processing for optical applications. In 2000, she joined the Physics of Accelerator Group to work on the ageing of nuclear and nonnuclear materials with ion and electron beams (0.5 MeV–6 MeV). Since 2002, she has worked at the CEA/DRT (Saclay) on the development and the characterization of plastic solar cells. Her research interests were focused on organic-inorganic interfaces in multilayers or nanocomposite materials, ageing mechanisms of organic layers and photovoltaic organic devices and
ion beam analysis. She has authored four patents and several publications. In 2006, she joined the Nuclear Energy Direction to manage R&D projects on conditioning, storage, and disposal of nuclear waste. Mikkel Jørgensen obtained his Ph.D. in organic chemistry in 1990 from the University of Copenhagen emphasizing the synthesis and characterization of new organic compounds/polymers with applications in materials science. During positions as an industrial chemist he investigated stable, nontoxic organic radicals for use as contrast agents in new types of magnetic resonance imaging (MRI) (NycoMed 1987– 1990); synthesis of peptide nucleic acids (PNA analogues of DNA) (PNA Diagnostics, 1990) to be used as diagnostic probes; new redox active polymers with application in electrochromic windows. After obtaining a position as senior scientist at Risø National Laboratory in 1994, Dr. Jorgensen, has investigated molecular recognition and self-assembly and utilized it for construction of sensor molecules for biologically interesting analytes such as glucose, creatinine, and ephedrine. In 2003, he became involved in the organic solar cell programme led by Frederik C. Krebs. This has involved a lot of fascinating organic chemistry on conjugated polymers and oligomers, which has resulted in more than 70 articles and patents. Eugene A. Katz received his master-of-science degree in semiconductor materials science in 1982, and his Ph.D. in physics in 1990 from the Moscow Institute of Steel and Alloys. In 1995, as a visiting scientist at the Israel National Solar Energy Center of the Ben-Gurion University, he started to investigate the growth, structure and photoelectrical properties of fullerene thin films. In 1997, Dr. Katz joined the BenGurion University’s Institute for Desert Research and has since been working in the Department for Solar Energy and Environmental Physics. In 2006, he became a member of the Ilse-Katz Center for Meso- and Nanoscale Science and Technology at the Ben-Gurion University. He is reviewer for 15 physical and materials science journals. Dr. Katz’s research interests include areas of applied solar energy, photovoltaics based on nontraditional semiconductors (fullerenes, carbon nanotubes, conjugated polymers, etc), photovoltaic characterization of AIIIBV concentrator solar cells at ultra-high concentration of natural sunlight (1,000 suns and more), and synthesis of carbon and inorganic fullerenes and
nanotubes by concentrated sunlight. He has published more than 120 scientific papers and book chapters on the above-mentioned topics (including 48 peer-reviewed papers in international journals), and 13 popular articles on fullerene-like structures in carbon nanomaterials, living organisms, and their architectures.
Kion Norrman obtained his Ph.D. in physical-organic chemistry from the University of Copenhagen in 1996. His research interests have since shifted from gasphase ion chemistry using theoretical and mass spectrometry based methods to surface science with focus on modification and characterization of polymer surfaces using a suite of physical and chemical characterization techniques. After having worked one year at the Danish Technological Institute, Dr. Norrman obtained a position at Risø National Laboratory in 1999. His work focused around time-of-flight secondary ion mass spectrometry (TOF-SIMS) and eventually photooxidation of polymer surfaces, which consequently led to collaboration with colleague Frederik Krebs and a lot of fascinating results within the field of degradation of organic solar cells.