Photovoltaic modules of dye solar cells

Photovoltaic modules of dye solar cells Dissertation zur Erlangung des Doktorgrades der Fakultät für Mathematik und Physik der Albert-Ludwigs-Universität Freiburg im Breisgau Ronald Sastrawan Freiburg im Breisgau Juni 26

Dekan: Leiter der Arbeit: Referent: Korreferent: Tag der Verkündung des Prüfungsergebnisses: Prof. Dr. Josef Honerkamp Prof. Dr. Joachim Luther Prof. Dr. Joachim Luther Prof. Dr. Matthias Weidemüller 13. September 26

Contents 1. Introduction... 1 1.1 The gap between research on the dye solar cell and large area applications... 2 1.2 Module-related physics of the dye solar cell... 3 1.2.1 Partial shading or electrical mismatching... 4 1.2.2 Leak in the internal sealing or photoinduced electrophoresis... 5 1.2.3 Optimization of module design... 5 1.3 Production technology for large area dye solar modules... 5 1.4 Outline of this thesis... 6 2. Physics of the dye solar cell... 9 2.1 The architecture of the dye solar cell... 1 2.1.1 Energy diagram and working principle of a dye solar cell... 11 2.1.2 Recombination losses in a dye solar cell... 13 2.2 Electrochemical potential... 14 2.3 Mass transport in an electrochemical system... 15 2.3.1 Gradient in the electrochemical potential... 16 2.3.2 Fick s diffusion laws... 17 2.4 Electrode-electrolyte interface... 17 2.4.1 Helmholtz double layer... 18 2.5 Charge transfer over an electrode-electrolyte interface... 18 2.5.1 Potential controlled reaction: Butler Vollmer equation... 19 2.5.2 Diffusion controlled reaction... 2 2.5.3 General case... 22 2.5.4 Nernst diffusion layer... 22 2.5.5 Diffusion-limited current... 23 2.6 I-V characteristics of a solar cell... 24 2.6.1 One-diode model... 25

3. Module-relevant physics of solid-state solar cells...27 3.1 From cell to module: methods of interconnecting solar cells...28 3.1.1 Methods of interconnecting dye solar cells...3 3.2 Electrical mismatching of solar cells or partial shading of a module...3 3.2.1 Physics of reverse biased solar cells...31 3.2.2 I-V curve of a solar module under partial shading...33 3.2.3 Electrical mismatching of a dye solar module...34 3.3 Electrical shunt between solar cells in a module...34 3.3.1 Shunts in a dye solar module...35 3.4 Optimization of module design...36 3.4.1 Distributed series resistances...37 3.4.2 Approximation of the distributed series resistances by a lumped series resistance (one-third rule)...38 3.4.3 Distributed series resistances in dye solar modules...39 3.5 Encapsulation of the module...39 3.5.1 Encapsulation of dye solar modules...39 3.6 Conclusions of Chapter 3...41 4. Integrated series connection in dye solar modules...43 4.1 Z-connection...44 4.2 W-connection...45 4.3 Monolithic connection...45 4.4 Conclusions of Chapter 4...47 5. Charge transfer under reverse bias potential...49 5.1 Measured I-V characteristic of a dye solar cell under forward and reverse bias potential...5 5.1.1 Electrical mismatching of dye solar cells in series connection...51 5.2 Investigation of the charge transfer in reverse-bias...52 5.2.1 I-V characteristics of electrode set-ups to determine the dominating charge transfer route in reverse bias...54 5.2.2 Interfaces studied by electrical impedance spectroscopy...55 5.3 Model for the I-V characteristics of a dye solar cell under forward and reverse bias potential...57 5.3.1 One-diode model for the dye solar cell under forward bias potential...58 5.3.2 Butler-Vollmer model for the dye solar cell under reverse bias potential...58 5.3.3 Complete model for the I-V characteristic of a dye solar cell over the full voltage range...6 5.3.4 Model for the electrical mismatching of dye solar cell in series connection 61 5.4 Long term stability of reverse biased DSC...63 5.4.1 Charge transfer under reverse bias suppressed by a dense TiO 2 blocking layer...65 5.5 Pattern in the scattering layer of dye solar modules...66 5.5.1 Applications of decorative solar modules...69

5.6 Spatially resolved photocurrent imaging technique for large area, series interconnected dye solar modules... 71 5.7 Conclusions of chapter 5... 74 6. Photoinduced electrophoresis in the electrolyte of dye solar modules... 77 6.1 Model of the photoinduced electrophoresis in dye solar modules... 78 6.1.1 Energy levels in a series connection of DSC... 78 6.1.2 Model system for the theory of photoinduced electrophoresis... 8 6.1.3 Potential gradients in a Pt/electrolyte/Pt cell with applied voltage and electrolyte barrier... 81 6.1.4 Modelling the ion concentration profiles under photoinduced electrophoresis in open-circuit... 83 6.1.5 Modelling the regeneration process through diffusion... 84 6.1.6 Modelling the short-circuit current densities under photoinduced electrophoresis... 85 6.1.7 Set of parameters... 86 6.2 Experimental... 87 6.2.1 Apparatus... 87 6.3 Measurement and simulation of the photoinduced electrophoresis... 88 6.3.1 Photoinduced electrophoresis in a low viscous electrolyte... 88 6.3.2 Photoinduced electrophoresis in an undiluted, high viscous ionic liquid... 89 6.3.3 Charge transport in highly concentrated ionic liquids... 9 6.3.4 Regeneration in the dark... 91 6.3.5 Regeneration with an externally applied voltage... 92 6.3.6 Diffusion current... 93 6.3.7 Influence of parameters on the theoretical model of the photoinduced electrophoresis... 94 6.3.8 Requirements on the barrier properties of the internal sealing material... 96 6.4 Conclusions of Chapter 6... 98 7. Optimization of module design... 99 7.1 Modelling the I-V curve of a dye solar module... 1 7.1.1 Modelling distributed series resistances... 1 7.1.2 Modelling discrete series resistances... 12 7.1.3 Modelling the efficiency of the module... 12 7.2 Influence of parameter variations on the I-V curve of a dye solar module... 13 7.2.1 Standard set of parameters... 13 7.2.2 Variation of the width of photoactive area... 16 7.2.3 Variation of the width of photoinactive area... 17 7.2.4 Variation of TCO sheet resistance... 18 7.2.5 Variation of contact resistance... 19 7.2.6 Variation of illumination intensity... 11 7.3 Module design I: Strip module... 114 7.3.1 Electrolyte canal... 114 7.3.2 Experimental... 116

7.4 Module design II: Interdigital meander module...117 7.4.1 Experimental...12 7.5 Conclusions of chapter 7...122 8. Long term stability...123 8.1 Glass frit sealing...124 8.1.2 Thermal stability of glass frit...126 8.1.3 Electrode distance in glass frit dye solar modules...127 8.1.4 Resistance of the series interconnection in glass frit dye solar modules...128 8.1.5 Up-scalability of glass frit sealing technology...129 8.1.6 Colourful glass frit...13 8.2 Accelerated ageing of small test cells...133 8.2.1 Long term stability under visible light soaking...134 8.2.2 Estimating the number of stable turnovers for the Ruthenium dye...136 8.2.3 Long term stability under 85 C in the dark...137 8.2.4 Combined ageing under 85 C in the dark and subsequent ageing under visible light...139 8.2.5 Thermal cycling test...14 8.3 Model for the degradation under thermal ageing...141 8.3.1 TCO layer...141 8.3.2 TiO 2 layer...141 8.3.3 Platinum layer...142 8.3.4 Dye...142 8.3.5 Electrolyte...142 8.3.6 Modelling the degradation of the short-circuit current density under thermal ageing...144 8.4 Conclusions of Chapter 8...145 9. State of the dye solar module technology...147 9.1 Research and development on dye solar module technology...148 9.1.1 Sharp Corporation, Ecological Technology Development Centre...148 9.1.2 Toyota Central R&D Laboratory and Aisin, Seiki...148 9.1.3 Sony, Materials Science Laboratories...149 9.1.4 Fujikura Ltd., Material Technology Laboratory...149 9.1.5 Dyesol...149 9.1.6 Solaronix...15 9.1.7 Hitachi...15 9.1.8 Electronic and Telecommunications Research Institute...151 9.1.9 Energy Centre Netherlands...151 9.1.1 Peccell...151 9.1.11 Institute of Plasma Physics...152 9.1.12 IVF, Industrial Research and Development Corporation...152 9.1.13 Industrial Technology Research Institute...153 9.1.14 Fraunhofer Institute for Solar Energy Systems...153 9.2 Conclusions of chapter 9...154

1. Summary and Conclusions... 155 1.1 Module-related physics of the dye solar cell... 155 1.1.1 Partial shading or electrical mismatching (Chapter 5)... 156 1.1.2 Spatially resolved photocurrent imaging technique (Chapter 5)... 157 1.1.3 Photoinduced electrophoresis (Chapter 6)... 157 1.1.4 Optimization of module design (Chapter 7)... 158 1.2 Production technology of dye solar modules... 158 1.2.1 Hermetic sealing material (Chapter 8)... 159 1.2.2 Long term stability (Chapter 8)... 159 1.2.3 Interdigital meander design (Chapter 7)... 16 1.3 Applications of results of this thesis... 16 Appendix A1. Experimental... 165 A1.1. Module manufacturing... 165 A1.2. Test cells: masterplates... 168 A1.3. Accelerating ageing procedures... 168 A1.4. Electrolyte cells for the characterisation of the platinum electrode... 169 A2. Approximation of the transient current in photoinduced electrophoresis171 A3. List of symbols, physical constants and abbreviations... 175 A4. References... 179 A5. Publications... 185 A5.1. Publications in reviewed journals... 185 A5.2. Conference proceedings, oral presentations, poster presentations... 186 A5.3. Patents... 189 A6. Acknowledgements... 191

1. Introduction 1. Introduction Photovoltaic (PV) research, development and industry have been expanding rapidly for a number of years now. Especially in Germany, the Renewable Energy Sources Act (EEG) has triggered an average annual economic growth of 9 % in the PV industry sector for the last 5 years. In spite of this solar boom, to date, merely.2 % of electricity in Germany is produced by PV. On the other hand, this technology has an enormous available potential, since PV is expected to play a major role in future energy supply. [Luther '5] The pillar of today s PV technology is the crystalline silicon solar cell. Crystalline silicon solar cells will continue to dominate the PV sector in the foreseeable future. However, new solar technologies are expected to gain market shares in the coming years. And in a long term perspective, the solar age [Scheer '99] will be founded on one or several of these new technologies. New PV technologies have the potential of becoming a true mass product with significant cost-reduction potential and new applications. New PV technologies competing for their market share today, are highly efficient concentrator solar cells (based on III/V compounds, such as GaAs), and thin film solar cells such as amorphous silicon (a-si), copper indium diselenide (CIS), cadmium telluride (CdTe) and crystalline silicon thin film (CSiTF) solar cells. New concepts, such as the organic solar cell (OSC) and the dye solar cell (DSC) are still under fundamental investigation. However, the DSC technology has meanwhile progressed to a stage very close to a possible commercialization. This work focuses on the module-relevant physics of the DSC technology, with special attention to the producibility of large area applications. 1

1. Introduction 1.1 The gap between research on the dye solar cell and large area applications Figure 1-1: Architecture of a dye solar cell. The nanoporous titanium dioxide and the liquid electrolyte are located between two glass plates, coated with transparent conducting oxide. The dye solar cell (DSC) technology has attracted wide attention since its invention in 1991 by M. Grätzel [O'Regan '91]. Soon laboratory scale efficiencies of over 1 % were reported [Grätzel ']. Due to the potential for low production costs and attractive colour and design, considerable efforts are being increasingly undertaken to enable a commercial up-scaling of this new type of solar cell. The structure and working principle of a DSC differs fundamentally to solid-state solar cells. The DSC is an electrochemical solar cell. Between two transparent conducting electrodes a dye-covered, nanocrystalline titanium dioxide (TiO 2 ) layer and a liquid electrolyte is encapsulated (Figure 1-1). In order to understand a DSC, many fields of science must come together. Nanoparticles, electrochemistry of redox electrolytes, catalysts and dye synthesis all find their application in a DSC. Thus, the DSC has fascinated many scientists around the world and extensive research has been invested in understanding and optimizing the DSC. Long term stability a crucial requirement for solar cells has shown good improvement [Hinsch '1]. Recently, Wang et al. reported a stable 8 % efficient DSC based on a low volatile electrolyte under thermal stress of 8 C for 1 hours [Wang '5]. It may be stated, that the fundamentals of the single cell are understood reasonably well and that a single DSC is a controllable system. However, the initial enthusiasm about this new type of solar cell weakened when attempts were made to bring this technology from laboratory scale to industrial, large area applications. The electrochemical nature of the DSC, in particular the liquid electrolyte imposed problems on the scale-up process. A module technology could not be adopted from solid-state solar technologies, but new solutions had to be found. Apart 2

1. Introduction from technological problems, also new, module-related problems appeared. For example, first attempts of interconnecting DSC in series failed, because the photoinduced electrophoresis of the electrolyte had not been anticipated (see section 1.2.2 below). Even today, 15 years after the discovery of the DSC, a gap exists between the laboratory research on single DSCs and the technological development of large area DSC modules. This work is intended to close this gap. Mainly, this work focuses on two points: The investigation of the module-related physics of the DSC The development of a producible technology for large area DSC modules A central issue of large area DSC modules is that although stable single cells have been demonstrated new module-related degradation mechanisms occur in a series connection of DSCs. 1.2 Module-related physics of the dye solar cell This research is not primarily concerned with the investigation and optimization of single DSCs. Rather, single DSCs are interconnected in series, to form large area modules. In this work, the term module-related physics refers to all process that are significant in a series connections of large area DSCs. A schematic cross section of a series connection of DSC is shown in Figure 1-2. single DSC single DSC single DSC series interconnect series interconnect electron flow conductor (silver) sealing material glas transparent conducting oxide titanium dioxide electrolyte platinum transparent conducting oxide glas break in transparent conducting oxide Figure 1-2: Schematic cross section of a series connection of DSC. The architecture of the series interconnect is shown in the enlargement. 3

1. Introduction The architecture of the cells and the series interconnect is shown in the enlargement in Figure 1-2. The exact architecture is not important in this chapter and will be discussed in detail in the course of this work. Here, the series connection of DSCs may be viewed as a series connection of direct current (DC) voltage sources in parallel with a diode. Each cell produces a voltage and current. In series connection, the voltages are added, while an equal current passes through each cell. The 3 major module-related aspects, which are investigated in this work are partial shading leak in the internal sealing optimization of module design 1.2.1 Partial shading or electrical mismatching In Figure 1-3 it can be seen, that if one cell in the series connection is shaded (or electrically mismatched), the current still has to pass this cell. The simplest equivalent circuit of a conventional pn-junction solar cell is a DC source with a diode in parallel (see Figure 1-3 bottom). Then the current passes the shaded cell in reverse bias direction through the diode. This might pose a problem. In order to clarify the behaviour of a DSC module under partial shading, the physics of reverse biased DSCs is investigated. These results are presented in detail in Chapter 5: Charge transfer under reverse bias potential. shaded cell Figure 1-3: When one cell in a series connection is shaded, the current has to pass this cell in reverse bias. The simplest equivalent circuit of a solar cell is a DC source in parallel with a diode. 4

1. Introduction Figure 1-4: A leak in the internal sealing will cause electrophoresis of the electrolyte under illumination. electrolyte leak in internal sealing Figure 1-5: The width of the individual cells must be optimized, because of series resistances. distributed series resistance optimization of cell width discrete series resistance 1.2.2 Leak in the internal sealing or photoinduced electrophoresis Another potential module-related degradation mechanism is shown in Figure 1-4: If there were a possibility of mass transport between the electrolyte of neighbouring cells, e.g. by a leak in the sealing material, an electrolytical shunt would occur. An electrolytical shunt in a DSC module causes electrophoresis of the electrolyte under illumination. During this photoinduced electrophoresis the redox couple (here: triiodide and iodide) is separated from each other. This research investigates the physics of the photoinduced electrophoresis in Chapter 6: Photoinduced electrophoresis in the electrolyte of dye solar modules. 1.2.3 Optimization of module design For large area solar modules, the design of the module is crucial for optimal photovoltaic performance. Essentially, the DSC does not differ here from conventional solid-state solar technologies. Upon increasing the individual cells, both a higher current is produced and the series resistance rises (Figure 1-5). In Chapter 7: Optimization of module design, the aim is to numerically investigate the influence of module parameters on overall efficiency. 1.3 Production technology for large area dye solar modules Based on the investigation of the module-related physics of DSCs, a producible technology for large area DSC modules can be developed. In particular, this includes the development of a hermetic sealing material and a (industrially) producible module design based on screen printing technology. 5

1. Introduction 1.4 Outline of this thesis In Chapter 2, the physics of the single DSC are discussed. The focus is placed on the charge and mass transport in a liquid redox electrolyte and over an electrolyte-electrode interface. In Chapter 3, the module-relevant physics of solid-state solar cells are studied in terms of their relevance to DSC. Module-relevant aspects such as partial shading and optimization of module design have already been investigated thoroughly for conventional solar technologies. New solar technologies like the DSC technology can benefit from these experiences. In this chapter, the solutions from solid-state solar technologies are discussed and evaluated in terms of their applicability to DSC. In Chapter 4, the loss factors of three types of integrated series connection in DSC modules are discussed. For the further development of the module concept in this work, the best type of integrated series connection will be chosen. Chapter 2, 3 and 4 describe knowledge from literature and up-to-date publications. My own research is presented in the following chapters. The partial shading (or electrical mismatching) of a DSC module is studied in Chapter 5. As the underlying process in partial shading, the charge transfer under reverse bias potential is investigated experimentally and theoretically. Based on these results, the partial shading as a module-relevant degradation mechanism is examined. Furthermore, DSC modules are presented, which are deliberately electrically mismatched. The electrical mismatching is achieved by a pattern in the photoactive area. And additionally, a method is presented, which allows the measurement of the spatially resolved photocurrent image of a large, series interconnected DSC module. This method is closely related to the operation of a DSC under reverse bias potential. The photoinduced electrophoresis of the electrolyte as a module-related degradation mechanism is investigated in Chapter 6. Here, the photoinduced electrophoresis in the electrolyte of DSC modules through an internal sealing is studied in a theoretical diffusion model and in experiments. The model allows the estimation of the required barrier properties of the internal sealing material, with respect to the diffusion constant of triiodide. In Chapter 7, the theoretical optimization of module design is carried out. In a numerical model of distributed series resistances, the influence of module and material parameters on the overall efficiency of the module is studied. In the model, the ratio of photoactive and photo-inactive area is varied and the series resistance of the interconnects and the contact material. Using these results, a module design is developed, which is very similar to solid-state thin film solar modules: the individual cells are strip-shaped. Additionally, in order to improve the industrial producibility, a DSC-specific design is 6

1. Introduction developed. This new design features meander shaped cells and a significantly smaller number of filling holes. In Chapter 8, the long term stability of single DSCs is tested with a newly developed glass frit sealing and a low-volatile electrolyte based on ionic liquids. The properties of the glass frit are investigated, both in terms of stability and up-scalability. In particular, electrode distance, resistance of interconnect, thermal stability, lead oxide (PbO) content and colour of the glass frit is examined. The glass frit sealing is applied to 3 x 3 cm² modules. Chapter 9 gives a brief overview on the state of the DSC module technology. Here the research and development of DSC modules in institutes and companies around the world is presented. 7

2. Physics of the dye solar cell 2. Physics of the dye solar cell In this chapter, the physics are summarized, which are required for an understanding of the dye solar cell (DSC). In section 2.1 the structure and working principle of the dye solar cell is introduced. The DSC is an electrochemical solar cell, in particular it utilizes a liquid electrolyte. Therefore, the interface between a metal (or semiconductor) and an electrolyte must be very well understood in order familiarize oneself with the processes in a DSC. The central quantity here is the electrochemical potential, which is a general concept, that can be applied to liquid electrolytes, as well as solid state bodies. It is introduced in section 2.2. Mass transport in an electrochemical system occurs down the gradient of the electrochemical potential. In a redox electrolyte, however, the dominant form of mass transport is diffusion, i.e. the field driven current (migration) can be neglected. This is discussed in section 2.3. The interface between a solid-state electrode and a liquid electrolyte is further addressed in 2.4 and 2.5. In particular, the charge transfer over such an interface is of importance. And finally, in section 2.6 the I-V curve of a DSC (and a solar cell in general) is discussed. Important electrical parameters of solar cells are introduced and a model for the I-V curve is presented. 9

2. Physics of the dye solar cell 2.1 The architecture of the dye solar cell The architecture of a dye solar cell (DSC) is shown in Figure 2-1. The DSC consists of a dye-covered, nanoporous TiO 2 (titanium dioxide) layer and an electrolyte encapsulated between two glass plates. Figure 2-1: Architecture of a dye solar cell. The nanoporous semiconductor (TiO 2 ) and the redox-electrolyte are located between two glass plates, coated with transparent conducting oxide (TCO). The TiO 2 is covered with a monolayer of dye and the counter electrode is coated with a thin platinum layer. Front and counter substrates are coated with a transparent conducting oxide (TCO). Fluorine doped tin oxide (SnO 2 :F), FTO is most commonly used. The FTO at the counter electrode is coated with few atomic layers of platinum (Pt), in order to catalyze the redox reaction with the electrolyte. The front electrode is coated with a nanocrystalline TiO 2 layer with average particle sizes of 5-2 nm. Assuming a layer thickness of 1 μm, the resulting effective surface is about 1 times larger as compared to a dense, compact TiO 2 layer. Three modification of TiO 2 exist: rutile, anatase and brookit. In the DSC preferably only the anatase modification is used. On the surface of the TiO 2, a monolayer of dye molecules is adsorbed. The huge nanoporous surface allows for an adsorption of a sufficiently large number of dye molecules for efficient light harvesting. The employed dye molecule is usually a ruthenium (Ru) metal-organic complex. The spectral absorption of the dye lies between 3 nm and 8 nm. The chemical structure of the most widespread dye molecule in DSC, the socalled N719 [Nazeeruddin '93] is shown in Figure 2-2. Good adsorption of the dye to the TiO 2 is important and is achieved over the two carboxylic groups of the ligand (L=2,2'- bipyridyl-4,4'-dicarboxylic acid) of the RuL 2 (NCS) 2 Between the two glass substrates, a liquid redox electrolyte is encapsulated. In particular, the liquid electrolyte is able to penetrate the nanopores of the TiO 2. The redox couple iodide/triiodide (I - /I 3 - ) is commonly used. Iodide is usually applied as a room temperature molten salt (ionic liquid), e.g. an imidazolium iodide (Figure 2-3). The ionic iodide liquid acts as a solvent for iodine (I 2 ), which reacts with iodide to form triiodide (I 3- ): 1

2. Physics of the dye solar cell I 2-1 + 2 I I3 During cell operation the following redox reaction takes place: I 3 - + 2e 3I 2-2 In general, best efficiencies are obtained upon illumination from the TiO 2 side. However, the DSC is usually semi-transparent and may be illuminated from the platinum side as well. Figure 2-2: Chemical structure of the N719 dye molecule: RuL 2 (NCS) 2 with L = 2,2'- bipyridyl-4,4'-dicarboxylic acid Figure 2-3: Chemical structure of propyl-methyl-immidazolium iodide (PMII): a room temperature molten salt (ionic liquid) 2.1.1 Energy diagram and working principle of a dye solar cell The energy diagram and electron transfer paths of a DSC are shown in Figure 2-4. The working principle of a DSC is based on the kinetics of the shown electron transfer reactions. Electrons are injected from the dye into the TiO 2 and the hole is injected into the electrolyte. Therefore, charge separation and charge transport occurs in different media and is spatially separated. By absorption of a photon, the dye molecule is excited. An electron is excited from the highest occupied molecular orbital (HOMO) to the lowest unoccupied molecular orbital (LUMO). The excited electron is rapidly injected into the conduction band of the TiO 2. 11

2. Physics of the dye solar cell TCO TiO 2 Dye Electrolyte Pt TCO E CB e - e - D * /D + E F n light e V e - I - Energy D/D + E REDOX I 3 - E VB front electrode counter electrode e - e - external load Figure 2-4: Energy scheme and electron transfer paths of a dye solar cell. D is the ground state of the dye and D * the exited state. E CB and E VB are the energy of the conduction band and valence band of the TiO 2, respectively. E n F is the quasi-fermi level of the electrons in the conduction band of TiO 2. E REDOX is the redox energy of the redox electrolyte, with redox couple I - /I - 3. It should be noted, that TiO 2 is a semi conductor with a large band gap of 3.2 ev, corresponding to a wavelength of λ=39 nm. Accordingly, visible light is not absorbed by the TiO 2. Direct absorption by UV-light is unwanted, since the created holes in the valence band of the TiO 2 are highly reactive and produce side reactions in the electrolyte, which are destructive for the cell in long term operation [Kern '1]. Charge transport occurs in the conduction band of the TiO 2 by pure diffusion of electrons to the FTO electrode. Electric fields in the TiO 2 are screened by the cations in the electrolyte, which penetrates the pores of the TiO 2 on a nano-scale [Würfel '6]. Upon reaching the TCO electrode, the electrons are conducted to the counter electrode via the external circuit. Catalyzed by the platinum on the counter electrode, the electrons are accepted by the electrolyte. This means, that the holes in the electrolyte (the - I 3- ) recombine with electrons to form the negative charge carriers, I + 2e 3I 3. By diffusion, the negative charge (I - ) is transported back and reduces the oxidized dye molecule (D + ). Triiodide (I 3 - ) is formed and the electrical circuit is completed: + 2D + 3I I + 3 2D 2-3 From Figure 2-4 it is clear, that the negative pole of the DSC is at the TiO 2 coated TCO substrate and the positive pole is at the platinum-coated TCO substrate. 12

2. Physics of the dye solar cell 2.1.2 Recombination losses in a dye solar cell Apart from these current-producing processes, loss processes occur in the DSC. An excited dye molecule may directly relax into its ground state, without injection of an electron into the TiO 2. This process is negligible, as injection is about 1 times faster [Nazeeruddin '93]. Also, electrons from the conduction band of the TiO 2 may recombine with the oxidized dye molecule, before the dye is reduced by the electrolyte. However, reduction by the electrolyte is about 1 times faster [Hagfeldt '95]. The most significant loss mechanism in the DSC is the recombination of TiO 2 conduction band electrons with the holes in the electrolyte, i.e. I 3 -. The electron transport by diffusion in the TiO 2, and their recombination with the electrolyte are the two competing processes in the DSC [Peter ']. It is important to realize, that for this reason the triiodide concentration in a DSC must be small. On the other hand, the triiodide concentration must be high enough as to provide enough recombination partners for the electrons at the platinum counter electrode. If this is not the case, the maximum current of the DSC will be diffusion-limited, i.e. limited by the diffusion of triiodide. Additionally, it should be noted, that the electrolyte penetrates the TiO 2 and is also in contact with the front FTO electrode. If the charge transfer resistance at the front FTO electrode would be the same as at the counter electrode, the DSC would indeed not operate properly. However, the charge transfer resistance at the counter electrode is reduced by many orders of magnitude by the platinum catalyst. Whereas the charge transfer resistance between pure, uncoated FTO and the iodide/triiodide redox couple is sufficiently high. This cannot be taken for granted, especially with redox couples other than iodide/triiodide. As optional layer in the architecture of a DSC, a very thin (<1 nm), dense TiO 2 layer may applied beneath the TiO 2. This layer is aimed to further suppress charge transfer from the front FTO to the electrolyte [Cameron '5]. Additionally, an optional light scattering layer may be applied on top of the TiO 2 in order to increase light harvesting [Hore '6]. Thus, transmitted light will be reflected back into the TiO 2 layer. This layer must be an insulator and porous, i.e. penetrable by the electrolyte. Such a scattering layer (e.g. ZrO 2 ) leads to an opaque (not transparent) DSC. 13

2.2 Electrochemical potential 2. Physics of the dye solar cell An essential concept in understanding the DSC is the electrochemical potential. In this section, it will be shown, that the electrochemical potential is a general concept, which is equivalent to the fermi energy of a metal or semiconductor and the redox energy of a redox electrolyte. Such a general concept is needed in the DSC, since charge transfer occurs between different media (semiconductor, metal, redox electrolyte). The electrochemical potential η is the sum of the chemical potential μ and the electrical potential ϕ [Würfel '95]: η = μ + ze ϕ 2-4 Here, e o is the elementary charge and z the number of charges of the considered species. It should be noted, that the electrical potential has the dimension of a voltage. It is given by the macroscopic distribution of charges. The chemical potential μ of a species is determined by its concentration c: c μ = μ + kbt ln 2-5 c st Here, μ is determined by the surroundings, k B is the Boltzmann constant, T the temperature and c st the standard concentration. Obviously, μ is the value of the chemical potential, if the considered species is present under standard concentration c=c st. The electrochemical potential corresponds to the energy, which is required to bring a particle of the considered species from the inside of the interacting system to the interaction-free point of infinity (vacuum level). This definition is equivalent to the definition of the fermi energy E F or the work function of a metal or semiconductor. Therefore, the electrochemical potential and the fermi energy are equivalent [Würfel '95]: μ = 2-6 E F From semiconductor physics it is known, that the quasi-fermi energy of electrons the conduction band of a semiconductor is given by [Ibach '95]: n n E F ECB + k BT ln N C n E F in = 2-7 Here E CB is the energy level of the conduction band, n is the density of electrons in the conduction band and N C is density of states in the conduction band. The analogy with Equation 2-4 and 2-5 is immediately obvious. 14

2. Physics of the dye solar cell The same analogy applies for a redox electrolyte. According to the general redox reaction ν Ox + m e ν Red 2-8 Ox Red a redox electrolyte is able to accept or release electrons. Here, Ox is the oxidized species and Red the reduced species. ν are the stoichiometric coefficient and m the number of transferred electrons. The redox energy is the energy which is required to remove an electron from the redox electrolyte, similar to the fermi energy of a metal. Therefore, the redox energy of a redox electrolyte E Redox is equivalent to the electrochemical potential of an electron in the electrolyte [Lewerenz '95]: η = E Redox 2-9 The Nernst equation allows to calculate the redox potential from the concentrations c of the redox species [Atkins '9]: With st νred ( c ) ( ) st νox c νox k BT = = + c Ox η ERedox ERedox ln 2-1 m νred cred E Redox the standard redox potential and c st the standard redox concentration. 2.3 Mass transport in an electrochemical system In general, in electrochemical systems, it is necessary to consider two modes of mass transport. Mass transport occurs due to [Southampton '85] a gradient in the electrochemical potential and convection Convection is the movement of a species due to mechanical forces. However, convection is not an important form of mass transport in this work, since the electrolyte layers in a DSC are very thin (<5 μm). In such thin layers, diffusion dominates and no convection regions are formed in the DSC [Hauch '98]. 15

2. Physics of the dye solar cell 2.3.1 Gradient in the electrochemical potential In a system, which consists of many subsystems, between which mass and charge transport is possible, a constant electrochemical potential throughout the system defines the equilibrium state. In equilibrium no net charge or mass transport occurs. On the other hand, if the electrochemical potential is not equal throughout the system, mass transport will occur in order to compensate all electrochemical potential differences. The electron current density j is determined by the gradient in the electrochemical potential [Würfel '95]: r σ r = η j 2-11 e Here, σ depicts the conductivity of the species. It is given by [Würfel '95]: σ 2-12 = bc e With b, the mobility of the species, which already includes the number of charges z. According to equation 2-4 and 2-11, the current is determined by the sum of the field driven current (migration) and diffusion. Migration, i.e. the field driven current, is the movement of charged species due to an electrical potential gradient, and it is usually the origin of mass and charge transport through the electrolyte. The current of electrons through the external circuit must be balanced by the passage of ions through the solution between the electrodes. It is, however, not necessarily an important mechanism of mass transport for the redox active species even if it is charged. The forces leading to the field driven current (migration) are purely electrostatic, and hence the charge can be carried by any ionic species in the solution [Southampton '85]. As a result, if a large excess of inert (redox inactive) electrolyte is in the solution, this balances the charge, and little redox active species is transported by migration. The inert electrolyte in a DSC is the cation of the iodide salt, usually Li + or an imidazolium + ion. These inert charges screen the inside of the electrolyte from electrostatic forces by the formation of a Helmholtz double layer at the electrodes (this will be explained in section 2.4.1). It should be noted, that in the electrolyte of a DSC the reduced and the oxidized redox species are negatively charged. In case of the hole (I - 3 ) migration would therefore work against the intended direction of mass transport. Diffusion is the movement of a species down a concentration gradient, and it must occur whenever there is a chemical change at a surface. In a redox electrolyte, especially in the DSC, diffusion is the dominant form of mass transport. 16

2.3.2 Fick s diffusion laws 2. Physics of the dye solar cell For the case, that the gradient in the electrical potential can be neglected, mass transport will only occur by diffusion. Convection is not an important form of mass transport in this work, since the electrolyte layers are very thin (<5 μm). According to equation 2-4, 2-5 and 2-11, the current is therefore: r j = σ r μ e + k B T ln c c st 2-13 r σ 1 r j = kbt c e c 2-14 r r j = bk T c 2-15 B With the Einstein relation and the diffusion coefficient D, D k BT b e = 2-16 this leads to the well-known 1 st Fick s Law for electrons: r r j = e D c 2-17 With the continuity equation dimension): n div j = t e, the 2 nd Fick s Law is obtained (here: in one c t 2 c = D 2 x 2-18 2.4 Electrode-electrolyte interface If a metal (or in general: any solid state body) is immersed into an electrolyte, it is referred to as an electrode. In general, the electrochemical potentials of the metal and the electrolyte are not equal. Upon contact, equilibrium will be established by an exchange of species. In case of a redox electrolyte and a noble metal electrode, only electrons are exchanged. In particular, no metal ions from the electrode will go into solution. Such an electrode is then called redox electrode. 17

2. Physics of the dye solar cell The exchange of electrons leads to an electrical potential difference between electrolyte and electrode, the so-called built-in potential. Since the charges in the electrolyte are freely mobile, opposite charges will arrange themselves at the metalelectrolyte interface. This double layer is called the Helmholtz double layer [Lewerenz '95]. 2.4.1 Helmholtz double layer The Helmholtz double layer acts like a capacitor with a small (a few angström) plate distance. The capacitance can be measured by impedance spectroscopy. In the simplest model, the potential gradient is linear (Figure 2-5). - - - - + + redox electrode + + redox electrolyte Figure 2-5: Arrangement of opposite charges in a Helmholtz double layer. x ϕ In particular, the complete electrical potential drops over the Helmholtz layer. Thus, in the inside of the electrolyte no electrical field exists. In a redox electrolyte, the electrical field can be very efficiently screened. It is important to note, that a redox electrolyte contains charged species, which do not participate in the redox reaction (redox inactive or inert species). Those species can very effectively arrange themselves in a Helmholtz layer, in order to screen any electrical potential differences. In case of the DSC, the cation of the iodide salt is the redox-inactive species, e.g. an imidazolium + or Li +. 2.5 Charge transfer over an electrode-electrolyte interface In equilibrium, no net charge transport occurs over the electrode-electrolyte interface. The Helmholtz double layer is formed with the built-in potential (see previous section). However, if an external potential is applied to the electrode, a current will flow over the electrode-electrolyte interface. In general, the charge transfer reaction is controlled by the applied potential. A higher potential will lead to a higher current flow over the interface. Additionally, the charge transfer will be determined by the concentration of the redox species in the electrolyte at the interface. Here, two cases can be differentiated. If the charge transfer reaction does not depend on the concentration of the redox ions at the interface, the reaction is called potential controlled. This is the case, if the charge transfer is slow, and/or the concentration of redox partners is high. 18

2. Physics of the dye solar cell If the charge transfer reaction does depend on the concentration of the redox ions at the interface, the reaction is called diffusion controlled. This is the case, if the charge transfer over the interface occurs very rapidly, and/or the concentration of redox partners is too low. Then the reaction can only proceed, if new redox partners reach the electrode by diffusion. 2.5.1 Potential controlled reaction: Butler Vollmer equation In a potential controlled charge transfer, the current over the interface is independent of the concentrations of the redox ions at the interface. It is only dependant on the applied potential. In this case, the current-voltage characteristics can be described by the Butler-Vollmer equation [Vetter '61]: e e j( V ) = j V V exp β exp (1 β ) 2-19 k BT k BT The current density j over the interface is determined by the applied potential V, the exchange current density j and the symmetry parameter β = 1. A symmetry parameter of.5 describes a charge transfer reaction, which is symmetric under positive (forward) and negative (reverse) potential V. An asymmetric reaction can be accounted for by β.5, as seen in Figure 2-6. current density j / j 1 8 6 4 2-2 -4-6 -8-1 -1-5 5 1 applied voltage V / mv α=.75 α=.5 α=.25 Figure 2-6: Current-voltage characteristics of the Butler-Vollmer equation for varying symmetry parameter β. 19

2. Physics of the dye solar cell For small potential V 1mV the Butler Vollmer equation can be approximated by first term Taylor expansion to: e j( V ) = j V kbt 123 1 = R CT 2-2 Thus, the charge transfer resistance R CT is defined. R CT is of dimension Ω cm². This is the same dimension as for a contact resistance, the total resistance decreases with electrode (or contact) area. In the DSC, the charge transfer resistance at the TCO-electrolyte interface can be reduce drastically by the catalyst platinum. Then, if the electrolyte is not diffusion-limited, the approximation 2-2 is valid. If the TCO electrode is not coated with platinum, however, the charge transfer is inhibited and the Butler-Vollmer equation 2-19 must be used. 2.5.2 Diffusion controlled reaction In a diffusion controlled charge transfer reaction, the concentration of redox ions at the interface is reduced drastically, because the redox ions are consumed faster, than they can diffuse to the electrode. The change in electrochemical potential caused by the change in concentration in the redox electrolyte is described by the Nernst equation 2-1. In the DSC, electrons from the platinum counter electrode react with the triiodide (I - 3 ) in the electrolyte. The change in electrochemical potential Δη due to a change in triodide concentration is then: 2 k BT c Δ = I3 η ln 2-21 E m c I3 E Here, c and c are the triiodide concentrations at the electrode, and far away from I 3 I 3 the electrode (the constant, initial concentration), respectively. Assuming a linear concentration profile from the electrode to a point inside the electrolyte, where the initial (constant) concentration of triodide is present, the current density is: E c c I3 I3 j = ed 2-22 δ Here, δ is the distance to the inside of the electrolyte, where the initial concentration of triiodide is existing, and D is the diffusion constant of triiodide.

2. Physics of the dye solar cell By using the diffusion limited current density j lim of an electrolyte (this will be described in 2.5.5), the current-voltage characteristics can be obtained. The diffusion limited current density j lim is the current, which flows when the concentration at the electrode has E diminished completely c : = I 3 c I3 jlim = ed 2-23 δ It follows: j j lim = c I 3 c c I 3 E I 3 c = 1 c E I 3 I 3 c c I3 E I3 = j j lim lim j 2-24 Substituting this expression into 2-21 yields: V = Δη = e k BT jlim ln em jlim j 2-25 The current-voltage characteristics for a diffusion controlled charge transfer reaction is thus: e = m j( V ) j lim exp V 1 2-26 k BT This curve is shown qualitatively in Figure 2-7. For small voltages the current increases roughly linearly and then approaches the diffusion-limited current j lim. 1..8 j/j lim.6.4.2. voltage / a.u. Figure 2-7: Current-voltage characteristics for a diffusion controlled charge transfer reaction of an electrode-electrolyte interface. The current j is related to the diffusion limited current j lim 21

2. Physics of the dye solar cell In the DSC, the catalytic platinum layer at the counter electrode reduces the charge transfer resistance drastically. The qualitative curve progression at the platinum counter electrode is therefore given by Figure 2-7 (diffusion controlled) rather than Figure 2-6 (Butler Vollmer). 2.5.3 General case In general, a combination of the two discussed cases occurs. For small applied potentials, a Butler Vollmerian behaviour is observed and for large applied potentials and large currents, the current approaches the diffusion limited current. In between, a mixed region exists, where both processes are combined (Figure 2-8). j/j lim 1 potential controlled combined region diffusion controlled voltage / a.u. Figure 2-8: Current-voltage characteristics of the charge transfer reaction at an electrode-electrolyte interface. In general, three regions can be distinguished: a potential controlled region at low voltages and a diffusion controlled region, where the current approaches the diffusion limited current j lim of the electrolyte. In between, a region exists where both processes are combined. 2.5.4 Nernst diffusion layer From equation 2-18 it is clear, that the steady state concentration profile will be a linear profile in most cases. Especially, for a one or two electrode set-up, where species are generated and consumed only at one or two points (in a one-dimensional problem), this is the case. From equation 2-17 it is clear, that the steady state current is then easily determined by the concentration of the considered species. Figure 2-9 and Figure 2-1 show the qualitative steady state concentration profiles in a one and two electrode setup, respectively. Upon increasing the applied potential, the redox species is consumed at one electrode and a linear steady state profile is formed. 22

2. Physics of the dye solar cell concentration / a.u. c initial electrode increasing potential δ distance from electrode / a.u. concentration / a.u. 2c initial c initial electrode electrode increasing potential δ 2δ distance from electrode / a.u. Figure 2-9: Qualitative steady state concentration profiles in a one electrode setup. Figure 2-1: Qualitative steady state concentration profiles in a two electrode setup. For the case of the one electrode setup the Nernst diffusion layer is defined to thickness δ, where far away from the electrode the initial redox concentrations exist (due to convection). The Nernst diffusion layer is typically about 1 μm in a DSC, but exact values are usually not known and depend on the diffusion constant and convection [Southampton '85]. For the case of the two electrode setup, with electrode distances small enough to prevent convection, the Nernst diffusion layer extends through the whole system. In the two electrode setup, redox species are produced at the 2 nd electrode (of opposite polarity), as they are consumed at the 1 st electrode. Here, the Nernst diffusion layer is defined to thickness δ, which is half the electrode distance. 2.5.5 Diffusion-limited current In Figure 2-9 and Figure 2-1 the steady state current is then given by: j cinitial c() = ed 2-27 δ If the applied potential is increased, the concentration of redox species at the electrode c() diminishes. Obviously, if the concentration reaches zero, the current cannot increase further, even if the potential is increased further. The maximum steady state current is therefore the so-called diffusion limited current j lim : cinitial j lim = e D 2-28 δ With the definitions used, it is important to note, that the Nernst diffusion layer δ is half the electrode distance in a two electrode setup. 23

2. Physics of the dye solar cell Furthermore, it should be mentioned, that the diffusion limited current is the maximum possible steady state current. Transient currents can indeed be much larger. Especially, when the potential is stepped from equilibrium to that for a diffusion controlled reaction, the initial currents are very large, since the initial concentration at the electrode is consumed first. The current then drops until it reaches the diffusion limited current. 2.6 I-V characteristics of a solar cell A typical current-voltage characteristics of a solar cell is shown Figure 2-11. The important electrical parameters are the short-circuit current density j SC, the open-circuit voltage V OC and the maximum power point P MAX (or MPP) with its corresponding currentdensity and voltage (j MP, V MP ). j SC current density / ma cm -2 j MP voltage / V V MP P MAX V OC Figure 2-11: Typical I-V curve of a solar cell. The important electrical parameters are the short-circuit current density j SC, the open-circuit voltage V OC and the maximum power point P MAX with its corresponding current-density and voltage (j MP, V OC ) In this work, the current I of a solar cell is in the majority of cases related to the area of the cell, in order to obtain a current density j, for obvious reasons. The current densityvoltage characteristics is nevertheless referred to as I-V curve. If the current is plotted instead of the current density, the short-circuit current density is always additionally stated in the Figure caption. The fill factor FF of a solar cell is defined by: MAX FF = 2-29 j P SC V OC Here, P MAX is also related to area with the dimension W/cm². 24

2. Physics of the dye solar cell And the efficiency eta: MAX eta 2-3 = P Φ incident Here, Φ incident is the power of the incident irradiation with dimension W/cm². The efficiency of a solar cell is highly dependant on the incident irradiation. The standard reporting conditions (SRC) are under the so-called AM 1.5 global spectrum. It is equivalent to the spectrum of the sun with an incident angle, such that the light path through the atmosphere is 1.5 times longer than for perpendicular incident irradiation. The intensity of AM 1.5 global is 1 mw/cm² or 1 W/m² [Goetzberger '94]. This light intensity is also referred to as 1 sun. 2.6.1 One-diode model The I-V curve of a solar cell can be modelled by the one-diode model [Würfel '95]. It is valid if the current is not limited by transport resistances. In a DSC, these would be the diffusion in the electrolyte or the diffusion in the TiO 2. Thus, if transport limitations can be neglected, the one-diode model from solid-state semiconductor physics can be adopted for the electrochemical DSC. Many detailed models have been developed, which explain the I-V curve of a DSC from microscopic parameters [Stangl '98,Ferber '1]. In this work, however, it is sufficient to describe the macroscopic I-V curve of a DSC by the one-diode model [Würfel '95]: ev I ( V ) = Isat exp 1 + ISC mdk BT 2-31 Here, I sat is the diode saturation current and m D is the diode ideality factor. The diode ideality factor is 1 from semiconductor theory. However, it has been empirically found, that real diodes deviate from the ideal diode equation (m D =1). Therefore, the ideality factor has been introduced. In real silicon diodes the ideality factor is 2 for low currents and 1 for high currents. The one-diode model is sometimes also called the standard solar cell equation. In a real solar cell, the series resistance and the shunt resistance have to be taken into account (Figure 2-12). The I-V curve then becomes: I( V ) = I e V e R I exp mdk BT S sat 1 + I SC V RSI + RSh 14243 I Sh 2-32 25

2. Physics of the dye solar cell Figure 2-12: Equivalent circuit of the one-diode model with series resistance R s and shunt resistance R Sh. The influence of the series and shunt resistance is shown in Figure 2-13. In Figure 2-13 it can be seen, that the one-diode model has been introduced in a form, which yields negative photocurrents. Considering, that a solar cell is essentially a diode this make sense from a physicist s point of view: especially in the dark, a positive voltage leads to a positive current. In solar cell engineering, however, the I-V curve of a solar cell is often displayed as in Figure 2-11: the photocurrent, and thus the produced power is of positive sign. In this work, both conventions will be used. In particular, if the I-V curves of setups are investigated, which exhibit diode behaviour, but are not a complete solar cell, the physicist s convention will be used. When complete solar cells are investigated, the engineer s convention is mostly used, as to follow the style of most publications in this field. current density / ma cm -2-5 -1-15 -2..2.4.6.8 voltage / V R SH = Ω, R S = Ω R SH = Ω, R S =1 Ω R SH =1 Ω, R S = Ω Figure 2-13: I-V curves of a solar cell with the one-diode model. The series resistance R s and shunt resistance R Sh have been varied. 26

3. Module-relevant physics of solid-state solar cells 3. Module-relevant physics of solid-state solar cells To date, a large variety of different photovoltaic (PV) technologies exist, ranging from multi- or monocrystalline silicon solar cells over high-efficient GaAs solar cells to thin film photovoltaics such as amorphous silicon solar cells (a-si), copper indium diselenide (CIS) or CdTe solar cells. Although each of these technologies uses different materials, working principle or production technology, they all share one aspect in common: at one point in the value creation chain, single cells are interconnected to form modules. On its way to commercialization, the DSC technology now faces the same challenge. Answers must be sought to questions that arise when interconnecting cells in modules. In this work, a PV module or PV panel, refers to an array of identical solar cells, which are all interconnected in series. In series connection, the voltage of the single cells add up, while the current of the module is identical to the current of one cell. A module is desirable, because a high, utilizable voltage is reached. Furthermore, a low current leads to lower ohmic losses. When interconnecting single cells in series to form a module, a number of new, related questions arise. These module relevant aspects have already been investigated thoroughly for the above mentioned solar cell technologies. New solar technologies like the DSC technology can benefit from these experiences. However, in contrast to conventional solar technologies, which are all based on a pn-junction in a semiconductor, the DSC is an 27

3. Module-relevant physics of solid-state solar cells electrochemical solar cell. In particular, the DSC utilizes a liquid electrolyte, a completely new situation compared to solid state solar technologies. In this chapter, the module relevant physics of conventional, established solar cell technologies are presented and related to their counterpart (if existing) of DSCs. In general, photovoltaic module technology must deal with: Methods of interconnecting solar cells Electrical mismatching of solar cells or partial shading of a module Electrical shunt between solar cells in a module Optimization of module design Encapsulation of the module The system integration of DSC modules is not considered in this work. Here, ACconverters (inverters) should not pose a problem. Whereas, the maximum power point tracking of a DSC module is more complex due to the high capacitance of the DSC [Wheatley '6]. 3.1 From cell to module: methods of interconnecting solar cells In general, two different types of series connection can be found in commercial solar modules. The individual solar cells are connected externally and are then assembled into a solar module. This technique is commonly used for crystalline silicon modules. The second option, as in thin film solar modules, is that all cells of the module are manufactured simultaneously already with an integrated series connection. The processing of cells to make a module accounts for up to 4 % of the total production costs when interconnecting the cells externally. Therefore, an integrated series connection is generally favoured, if it is technically achievable. For crystalline silicon cells an integrated series connection is not possible for a large area module, because of the limited wafer size. Thin film photovoltaic modules like a-si, CIS or CdTe all incorporate an integrated series connection. An external series connection results in a solar module which consists of many cells with the dimensions of the wafer (about 15 x 15 cm²). While the integrated series connection of most commercial thin film photovoltaics results in a strip design: the individual cells are strip-shaped. 28

3. Module-relevant physics of solid-state solar cells Figure 3-1: Typical monocrystalline silicon solar module with external series connection (Shell Solar). Figure 3-2: Typical CIS solar module with integrated series connection (thin vertical lines) (Shell Solar). Therefore, thin film solar modules have a very homogenous optical appearance. Figure 3-1 and Figure 3-2 show two typical examples of an externally interconnected crystalline silicon solar module and a thin film solar module (CIS) with integrated series connection, respectively. Figure 3-3: Layer sequence and production steps of a monolithic, integrated series connection for a CdTe solar module. The TCO layer is structured by laser scribing in step 1 In step 2 the absorber materials are deposited. Then a line is cut into the absorber layer in step 3. The backcontact (e.g. Mo) is deposited in step 4 and a line is again cut in step 5. Figure 3-4: Many individual cells are aligned on a module interconnected by a monolithic series connection. 29

3. Module-relevant physics of solid-state solar cells The production steps of a monolithic, integrated series connection are shown in Figure 3-3 for a CdTe solar module [Brecl '5]. As substrate a float glass is used, which already acts as the module cover glass for further cost reduction. The glass is coated by a thin layer of a transparent conducting oxide (TCO), in this case sputtered indium tin oxide (ITO). The TCO layer is structured by laser scribing in step 1 in order to electrically insulate the individual cells from each other. In step 2 the absorber materials (the solar cell, CdS and CdTe) are deposited by e.g. close spaced sublimation (CSS), a physical vapour deposition method. Then a line is cut into the absorber layer either mechanically or by laser scribing in step 3. The backcontact (e.g. Mo or Ni) is sputtered in step 4 and a line is again cut in step 5. In a module many cells are then just aligned next to each other as seen in Figure 3-4. 3.1.1 Methods of interconnecting dye solar cells Due to the different set up of the DSC and its electrochemical nature, the above type of integrated series connection cannot just be adopted. But new solutions must be (and have been) found to fabricate an integrated series connection in a DSC module. The DSC technology allows the fabrication of an integrated series connection in 3 different ways: the monolithic type, the W-type and the Z-type. These will be discussed in detail in Chapter 4: Integrated series connection in dye solar modules. Of course it is also possible to manufacture DSCs in tiles of for example 1x1 cm² and interconnect those externally in a module. But the DSC modules developed in this research all incorporate an integrated series connection. As mentioned above, the processing of a module with integrated series connection is much more cost-effective and the aesthetic appearance more favourable, than of a module with external series connection. 3.2 Electrical mismatching of solar cells or partial shading of a module Building integrated photovoltaics (BIPV) is a new branch in modern solar architecture. It is a synthesis between architecture and engineering to integrate a photovoltaic panel into a facade or a roof of a building so that it is both aesthetically pleasing and energetically useful. In an urban environment, however, its is not always possible to install a complete photovoltaic array on a flat plane, without shading, and to insure a homogenous incident irradiation on the array. Inhomogeneities in the incident irradiation can lead to electrical mismatching between the individual cells of a module. When electrical mismatching occurs, the weakest cell in the panel becomes reverse biased and is an energy load rather than an energy producer (see Figure 3-5). Thus optimal module performance is not achieved. Since a PV module is an electrical series connection of many cells, the ratio of energy loss and shaded area is disproportionately high. 3

3. Module-relevant physics of solid-state solar cells Figure 3-5: If one cell in a PV array is shaded, it still has to transport the current of the module. A certain voltage is needed to drive the current through this shaded cell in reverse direction. The cell consumes energy, a hot spot occurs. Additionally, localized points of dangerously high temperatures (more than 15 C) occur in the shaded, reverse biased cell, which are called hot spots. Hot spots may cause cell cracking and damage to module encapsulation. In crystalline silicon solar modules with external series connection, bypass diodes are therefore incorporated across individual cells or groups of cells. In thin film modules with integrated series connection the individual cells are not directly accessible, so bypass diodes can only be installed across module terminals. Hence, hot spot protection is not possible. On the other hand, thin film modules have a strip pattern, making the extensive shading of a single cell very unlikely under outdoor operation. The product qualification tests for solar modules include the verification of hot spot resistance. [Kovach '95] 3.2.1 Physics of reverse biased solar cells To determine the I-V curve of a solar module under partial shading, the I-V curves of the individual cells must be known under forward and reverse bias. Figure 3-6 shows the I-V curve of a crystalline silicon solar cell under forward and reverse bias in the dark. Under forward bias the I-V curve resembles a typical semiconductor diode. For most purposes the one-diode model is sufficient to describe the I-V curve of a solar cell under forward bias [Würfel '95] (compare Chapter 2.6.1). I( V ) = I e V e R I S sat exp 1 m D k B T + I SC V RSI + RSh 14243 I Sh 3-1 When the applied voltage under reverse bias is large enough, the pn-junction will break down, and a theoretically infinite amount of current will be conducted. The largest sustainable reverse biased voltage is called the breakdown voltage, V br. Under reverse bias the breakdown voltage is an important parameter. The exact value of V br can vary even for cells with identical characteristics under forward bias. For crystalline silicon solar cells V br can be higher than 2 V [Kovach '95]. 31

3. Module-relevant physics of solid-state solar cells current / A 4 2-2 -4 V br V < reverse bias V > forward bias -25-2 -15-1 -5 voltage / V Figure 3-6: I-V curve of a crystalline silicon solar cell under forward and reverse bias in the dark. The two mechanisms leading to breakdown in silicon solar cells are avalanche multiplication and tunnelling. Essentially, junction breakdown is not inherently destructive for the cell. However, damage does occur when excessively large reverse currents lead to points of localized overheating (hot spots). Tunnelling or Zener breakdown occurs at high electric fields, when the lattice structure of silicon becomes distorted. When the electric field increases even further, the electron has enough energy to tunnel through the forbidden energy gap. Avalanche breakdown is due to the destruction of a covalent bond by the collision of a free carrier. When an electron collides with the lattice structure to break a covalent bond, an avalanching effect is set off, whereby two carriers are created upon the impact of one. Avalanche breakdown is the dominant breakdown mechanism in silicon solar cells.[bishop '89,Kovach '95] The complete I-V curve (forward and reverse bias) of a solar cell can be described by modifying the one-diode model. In the Bishop model the shunt current I Sh in Equation 3-1 is multiplied by the avalanche factor to account for the reverse biased current [Bishop '88]. m a V + IRS V + IRS I = + Sh 1 a 1 3-2 R Sh Vbr 1442444 3 avalanche factor Here, a is the fraction of ohmic current involved in the breakdown and m a is the avalanche breakdown coefficient. The avalanche factor approaches infinity as ( V + IRS ) approaches V br. 32

3. Module-relevant physics of solid-state solar cells In literature the value of the avalanche factor (Equation 3-2) has been determined empirically. Due to the complexity an uniqueness of each cell, the value of the avalanche factor can vary even between different regions of one cell and from cell to cell. The complete I-V curve is then described as: m a qv qr SI V + IRS V + IRS I ( V ) = I exp 1 + + 1 + 1 sat I SC a 3-3 mdk BT RSh Vbr 3.2.2 I-V curve of a solar module under partial shading When solar cells are interconnected in series, voltages are summed at equal currents. V total n ( I) = V ( I) 3-4 i= 1 i Equation 3-3 has to be solved for V in order to obtain V i (I). Obviously, Equation 3-3 is a transcendental equation and can only be solved numerically. In order to model the partial shading of an array, the photocurrent of the individual cells (I SC ) are chosen according to the amount and geometry of the shading. I SC is proportional to the incident irradiation. The partial shading of an individual cell is equivalent to reducing the incident irradiation on this cell homogenously [Quaschning '96a,Quaschning '96b]. In a small array the total short-circuit current of the array is dictated by the weakest cell. A small array is an array with a total open-circuit voltage, that is smaller than the breakdown voltage of one cell. In a large array the situation is different. Here, the unshaded cells provide enough voltage to drive a current through the shaded cell in reverse bias. 33

3. Module-relevant physics of solid-state solar cells one cell partially shaded resulting module of 72 cells. 71 cells in series unshaded + = -.5 current / A -1. -1.5-2. V br of one cell Isc/2 voltages add up -2.5 Isc -3. -2-1 1 2 3 4 voltage / V Figure 3-7: Qualitative I-V curve of a solar module with 72 cells in series. One cell is shaded by 5 %. Voltages add up at equal currents. A bump or step in the I-V curve occurs. Figure 3-7 shows the qualitative I-V curve of a module with 72 cells in series. One cell is shaded by 5 %, the remaining 71 cells are not shaded. A bump or step in the I-V curve occurs due to the voltage loss, which is required to operate the shaded cell in reverse bias at currents higher than the short-circuit current of the shaded cell (I SC /2). The breakdown voltage of the total module is over 1 V. It is not shown in Figure 3-7. 3.2.3 Electrical mismatching of a dye solar module Essentially, a DSC module under electrical mismatching behaves like a conventional solid state solar module. However, a special, very important feature of the DSC is its extremely low breakdown voltage (about 5 mv). For that reason a DSC may be seen as a solar cell with its own incorporated bypass diode. Obviously, this is advantageous in terms of power output of a partially shaded DSC module. Additionally, no hot spots occur since much less power is consumed when reverse biasing a DSC. The reason for this low breakdown voltage lies in the electrochemical nature of the DSC. One of the main results of this research is the investigation of the physics of reverse biased DSCs. These results and their significance to DSC modules are presented in detail in Chapter 5: Charge transfer under reverse bias potential. 3.3 Electrical shunt between solar cells in a module During the fabrication steps of an integrated series connection of the type shown in Figure 3-3, its is unavoidable to incorporate electrical shunts between the individual cells 34

3. Module-relevant physics of solid-state solar cells Figure 3-8: Electrical shunts (R SH ) in an integrated series connection of a CdTe solar module. of the module [Brecl '5]. A close up of the integrated series connection is shown in Figure 3-8. One shunt (R SH ) inherently occurs over the line in the TCO across the absorber material. A second shunt occurs when backcontact material (e.g. Mo) is transferred onto the TCO during the last cutting process. These shunt resistances should be known and taken into account in optimization of module design [Burgelman '98]. Obviously, these shunt resistances reduce the performance of the module. It should be mentioned, that the loss in performance due to shunt conductances is more severe at lower light intensities (when the total current is lower), and less severe at higher light intensities. 3.3.1 Shunts in a dye solar module In a DSC module the situation is completely different. The 3 types of integrated series connection (monolithic, W and Z, see Chapter 4: Integrated series connection in dye solar modules ) can be manufactured without inherently integrating an electrical shunt resistance between individual cells in a module. Here, a precise screen printing technology in principle allows the accurate positioning of all layers. The absence of shunt conductances is especially advantageous for operation at low light intensities. However, due to its electrochemical nature, a DSC module might feature a different type of shunt. In a DSC module, the individual cells must not only be insulated electrically from each other, but also electrolytically. In other words, each cell in a series connection of DSC must be a sealed, closed compartment in order to prevent penetration of the liquid (or quasi solid-state) electrolyte. If there was a possibility of mass transport between the electrolyte of neighbouring cells, e.g. by a leak in the sealing material, an electrolytical shunt would occur. At first glance, one would expect similar consequences as of an electrical shunt. The module performance would be reduced and the shunt would be taken into account during optimization of module design. However, an electrolytical shunt in a DSC module is much more severe than an electrical shunt in conventional solar modules. An electrolytical shunt in a DSC module causes the separation of the redox couple between the two shunted cells. In a time period of several hours to days, this would lead 35

3. Module-relevant physics of solid-state solar cells to the complete degradation of module performance and eventually irreversible destruction of the module. In this work, the process of the separation of the redox couple is called photoinduced electrophoresis. One of the main results of this research is the investigation of the physics of the photoinduced electrophoresis in Chapter 6: Photoinduced electrophoresis in the electrolyte of dye solar modules. In a theoretical model the long term stability of a DSC module is estimated depending on the diffusion coefficient of triiodide in the sealing material. Long term stability is one major obstacle for the DSC technology on its way to industrialization. For single DSCs many critical degradation mechanisms have been identified. Single cells could be demonstrated, which are stable under high temperatures (85 C), visible light and UV light (with additives to the electrolyte). The photoinduced electrophoresis is an additional, module-related degradation mechanism and must be thoroughly studied. The hermetic sealing of DSCs is a critical issue for DSC researchers around the world. Apart from (obviously) avoiding leaks (holes) in the sealing material, one must realize that even a slight permeability of the sealing material with respect to the electrolyte might lead to a slow, but continuous degradation of module performance. 3.4 Optimization of module design The up-scaling of solar cells from small, single cells to large area modules is inevitably accompanied with a loss in efficiency. This efficiency gap is usually several (absolute) percent. The main reason for this loss in efficiency is a loss in active area, the series resistance and the shunt conductance [Burgelman '98]. The shunt conductances are shown in Figure 3-8. Figure 3-9 shows the series resistances and the ratio of active area and dead area of a CdTe module. Figure 3-9: Distributed sheet resistances (R sheet ) and discrete contact resistances (R contact ) in a CdTe module. The active area is reduced due to the dead area of the integrated series resistance. 36

3. Module-relevant physics of solid-state solar cells In general, two types of series resistances are distinguished. The discrete series resistance mainly consists of the contact resistance of the integrated contact (R contact ) and all other ohmic resistances, which have to conduct to total current of the module. The second type, the distributed series resistances, only conduct part of the current (R sheet ). In Figure 3-9 the current increases from right to left in the backcontact material. A sheet resistance is defined as the resistance of a layer, measured in a configuration as seen in Figure 3-1. The current passes along the layer, not through it. The sheet resistance is measured in Ω/ for a=b in Figure 3-1. a Ω b Figure 3-1: A sheet resistance is measured over two sides of a layer. The current passes along the layer, not through it. Sheet resistance is measured in Ω/. For a very large width of the active area in Figure 3-9, the area loss is small, but the loss due to distributed series resistances is high. On the other hand, for a very small width of active area, the area loss is high, but the ohmic losses are small. Obviously, an optimal width of active area exists. 3.4.1 Distributed series resistances In order to calculate the distributed series resistances we shall consider the unit cell shown in Figure 3-11. The ideal unit cell is assumed to have no TCO series resistance and a I-V characteristic j(v). The TCO has a specific sheet resistance ρ TCO. Figure 3-11: Unit cell with an I-V characteristic j(v(x)) and sheet resistance ρ TCO. In particular, each unit cell is operated at a different working point j(v) due to the sheet resistance of the TCO. If the total cell is (for example) operated in short-circuit (V cell =), the unit cell at x= is not operated in short-circuit. The voltage drop dv in the infinitesimal element dx is given by the current in the TCO I(x) (Ohm s law): dx dv = I( x) ρ TCO 3-5 L 37

3. Module-relevant physics of solid-state solar cells dv = I( x) ρ 3-6 TCO dx L is the length of one cell, such that the sheet resistance x ρ TCO is of dimension Ω. L The current I(x) is given by all current contributions of all elements of the unit cell from to x: x I( x) = L dx' j( v( x')) 3-7 This leads to the following differential equation [Nielson '82,de Vos '84]: d 2 dx v 2 = ρ j( v( x)) 3-8 TCO Equation 3-8 cannot be solved analytically, except for very trivial functions j(v(x)). For the one-diode model (Equation 3-1) the exact solution must be found numerically. 3.4.2 Approximation of the distributed series resistances by a lumped series resistance (one-third rule) It is common in literature to approximate the distributed series resistances with an equivalent lumped series resistance. Assuming, that the current is generated homogenously throughout the cell ( j ( v( x)) = j = const ), one may deduce from Equation 3-7: I( x) = j x L 3-9 This approximation is not valid for large sheet resistances or large width of active area. As mentioned above, all unit cells are operated at different working points at different x- coordinates. For small sheet resistances, however, this approximation may be used and it 2 follows for the power loss ( P = R I ): loss dp loss 2 2 2 dx = ρ TCO j x L 3-1 L p loss 1 ( x) x 3 2 3 = ρ TCO j L 3-11 38

3. Module-relevant physics of solid-state solar cells p loss 1 w a ( w a ) = ρ TCO I 14243 3 L lumped series resistance 2 Mod 3-12 with I = Mod j w al 3-13 Here, w a is the width of the active area and L is the length of one cell. The total module current is then given by Equation 3-13. Equation 3-12 shows that the lumped series resistance may be approximated by the total sheet resistance with the factor 3 1. This approximation is also known as the one-third rule [de Vos '84]. 3.4.3 Distributed series resistances in dye solar modules The optimization of module design for DSC modules, especially the treatment of distributed series resistance may exactly be adopted from thin film photovoltaics like CdTe. Furthermore, with today s computers it is convenient to find an exact numerical solution to Equation 3-8, rather than using the approximation of a lumped series resistance. In Chapter 7: Optimization of module performance Equation 3-8 is solved numerically to find the optimal cell width for the case of a DSC module. 3.5 Encapsulation of the module Long term stability of solar modules is often related to the encapsulation of the module. The most common encapsulation method is to laminate a cover glass on the module using an EVA foil (ethylene vinyl acetate). Degradation of module performance occurs, if humidity corrodes the metal contacts. Furthermore, inadequate UV stability leads to browning of the EVA, thus lowering the current and the efficiency of the module (relative about 5 % in about 1 years [Dunlop '6]). Much research is conducted to improve the encapsulation of conventional solar modules. However, this work is not related and cannot be transferred to the development of an encapsulation of DSC modules. 3.5.1 Encapsulation of dye solar modules The encapsulation of DSC modules is much more critical (and difficult) than the encapsulation of conventional solar modules. Again, the liquid electrolyte distinguishes the DSC from solid-state solar technologies and complicates encapsulation. The development of a hermetic seal is one major challenge in attaining long term stable DSCs. 39

3. Module-relevant physics of solid-state solar cells The liquid electrolyte in a DSC contains the I - 3 /I - redox couple, a very reactive and corrosive substance. Therefore, the sealing material of DSC modules must guarantee (for 2 years): Protection of the metal contacts (corrosion) Hermetic internal separation of the individual cells (photoinduced electrophoresis) Hermetic external encapsulation (evaporation of the electrolyte) And the sealing material must be (for 2 years): Inert against the I 3 - /I - redox couple UV-stable Stable at high temperature (85 C) Many sealing materials have been tested by various groups. The most common material used for encapsulation is Surlyn (DuPont), a polymer hotmelt foil. Epoxy glues have been extensively tested, silicon, UV adhesives and many others. Up to now, the results obtained with these sealants are not satisfactory and do not guarantee long term stability (in terms of the product qualification test for solar modules). Especially, the damp heat test (85 C) remains a challenge. In this work, glass frit is proposed as a sealing material, similar to that used for the sealing of plasma display panels (PDP). Glass frit has the advantage that it has very stable thermal, chemical and mechanical properties. Furthermore, it can be processed costeffectively by screen printing. It has the disadvantage of a high processing temperature. Therefore, the dye has to be introduced into the module after the glass fusing process. The external sealing of a PDP with glass frit may be adopted for the DSC technology. In PDP sealing, front and bottom substrates are fused together with glass frit. A hole remains in the bottom substrate to exhaust and fill the panel. A tubulation or exhaust tube is fused to the hole with glass frit. The tube is usually glass and can be melted off, forming an air tight seal [Boeuf '3]. This sealing procedure sealing of glass plates and sealing of filling holes is comparable to the sealing procedure required for DSC modules. Since the encapsulation issue is of such crucial importance to the DSC technology, the properties of the glass frit encapsulation are presented in Chapter 8: Long term stability along with the results of long term stability tests on single, very well sealed, glass frit DSCs. 4

3. Module-relevant physics of solid-state solar cells 3.6 Conclusions of Chapter 3 The most important module-relevant physics of solar modules were presented. On the basis of conventional solar cell technologies, especially crystalline silicon and CdTe solar modules, it was shown how these module-relevant aspects have been dealt with in the past. The DSC technology can benefit from these experiences. Some of the results obtained in other solar cell technologies may be directly transferred to the development of DSC modules, such as the treatment of distributed series resistances to find the optimal cell width. For others, such as the treatment of electrical mismatching (partial shading) or the method of interconnecting cells in series, the approach may be adopted, but new solutions must be found, taking into account the electrochemical nature of the DSC. And there are also new aspects, which are not known from established solar cell technologies. In particular, the process of photoinduced electrophoresis the shunt across the electrolyte has no counterpart in conventional solar cell technologies and must be investigated. And last, but not least, new materials (glass frit) must be researched to develop a stable encapsulation method, which is capable to hermetically seal the liquid, very corrosive electrolyte (I 3 - /I - ). In terms of module-relevant physics, it may be concluded that the DSC technology has advantages as well as disadvantages in comparison to established solar cell technologies. From a module-based point of view the low breakdown voltage of the DSC (incorporated bypass diode) is advantageous when interconnecting DSC in series. But there are also major drawbacks, such as the sealing issue and the (related) photoinduced electrophoresis. 41

4. Integrated series connection in dye solar modules 4. Integrated series connection in dye solar modules As in other thin film photovoltaic technologies, DSC modules can be manufactured with an integrated series connection. Due to various reasons (e.g. simplified processing, cost-efficiency, aesthetic appearance) an integrated series resistance is generally favoured over an external series connection of individual cells, if it is technically achievable. The design of an integrated series connection for DSC modules differs from the design for the integrated series connections known from other thin film solar technologies (e.g. CdTe, see Chapter 3.1). The unique set up and the electrochemical nature of the DSC must be accounted for in the design of an integrated series connection for DSCs. The individual cells must not only be insulated from each other electrically, but also electrolytically. Otherwise photoinduced electrophoresis would occur (see Chapter 3.3.1). Furthermore, the electrical interconnects must be protected against the corrosive electrolyte. The use of metals, which are resistant against the I 3 - /I - couple, such as Ti, W or Ni was rejected due to low conductivity. Furthermore, these materials oxidize strongly at temperatures needed for the glass fusing process, used to seal the devices in this work. Therefore, silver (Ag) is used as the electrical interconnect, protected by a glass frit barrier. Silver can be applied by screen printing. This is a great advantage, since all other layers are applied by screen printing as well. An integrated series connection for DSC modules can be designed in 3 ways [Tulloch '4]: 43

Z-connection W-connection Monolithic connection 4. Integrated series connection in dye solar modules In this chapter, all three types of integrated series connection are discussed and evaluated. A favoured concept is chosen for the DSC modules developed in this work. It should be mentioned, that all 3 types of integrated series connection are fabricated by screen printing. This technology allows a precise positioning of all layers. In contrast to the monolithic series connection of e.g. a CdTe module (see Chapter 3.3), no shunt conductances are therefore inherently incorporated into an integrated series connection for DSC modules. 4.1 Z-connection Figure 4-1 shows the cross section of the fabrication of an integrated series connection of Z-type. First the transparent conducting oxide (TCO) layer on the glass substrate is structured by laser scribing (1). Then the TiO 2 layer, the platinum layer, the silver lines and the glass frit are screen printed on the respective substrates (2) and dried (3). Optionally, ZrO 2 is screen printed as a scattering layer on top of the TiO 2 layer. The glass frit is screen printed as a protective barrier on both sides of the silver lines. After sintering (4) the counter electrode is aligned on top of the working electrode and fused at high temperatures (5). Thus, the glass frit forms a hermetic seal around the silver. Provided that the layer thickness of the glass frit and the silver match, the electrical Z-contact is formed during the glass fusing process. The dead area is given by the width of the glass frit and the width of the silver line. Figure 4-1: Integrated series connection of Z-type. First the TCO layer is structured by laser scribing (1), Then the TiO 2 layer, the platinum layer, the silver lines and the glass frit are screen printed on the respective substrates (2) and dried (3). After sintering (4) the module is fused at high temperatures (5). A hermetic seal is formed around the silver. The electrical Z-contact is formed during the glass fusing process. 44

4.2 W-connection 4. Integrated series connection in dye solar modules Figure 4-2 shows the cross section of the fabrication of an integrated series connection of W-type. First the transparent conducting oxide (TCO) layer on the glass substrate is structured by laser scribing (1). Then the TiO 2 layer, the platinum layer and the glass frit are screen printed on the respective substrates (2) and dried (3). After sintering (4) the counter electrode is aligned on top of the working electrode and fused at high temperatures (5). Thus, the glass frit forms a hermetic seal between the cells. The electrical W-contact is achieved by aligning neighbouring cells alternatingly in an inverted pattern. The dead area is given by the width of the glass frit. This design requires no silver lines. This is an advantage in terms of cost and ratio of active area to dead area. A disadvantage is, that no scattering layer (e.g. ZrO 2 ) can be applied, since half of the cells are illuminated from the backside. Even without scattering layer, the cells, which are illuminated from the backside are not electrically matched to the cells, which are illuminated from the front side. A solution would be to adjust (increase) the cell widths of the backside-illuminated cells [Koide '6]. 4.3 Monolithic connection Figure 4-3 shows the cross section of the fabrication of an integrated series connection of monolithic-type [Kay '96]. First the transparent conducting oxide (TCO) layer on the glass substrate is structured by laser scribing (1). Then the TiO 2 layer, the ZrO 2 layer, the graphite layer and the glass frit are screen printed on the respective substrates (2) and dried (3). After sintering (4) the cover glass is placed on top of the working electrode and fused at high temperatures (5). Thus, the glass frit forms a hermetic seal between the cells. Figure 4-2: Integrated series connection of W-type. First the TCO layer is structured by laser scribing (1), Then the TiO 2 layer, the platinum layer and the glass frit are screen printed on the respective substrates (2) and dried (3). After sintering (4) the module is fused at high temperatures (5). A hermetic seal is formed between the cells. The electrical W-contact is achieved by aligning neighbouring cells alternatingly in an inverted pattern. No silver is required. 45

4. Integrated series connection in dye solar modules The monolithical contact is achieved by just aligning neighbouring cells next to each other. The dead area is given by overlap of the ZrO 2 and the graphite and the width of the glass frit. All layers including the ZrO 2 and the graphite layer are porous and are penetrable for the electrolyte. This design requires no silver lines. Moreover, it requires only one TCO glass. This is an advantage in terms of costs, since especially the TCO glass is one of the greatest cost factors of the DSC. Additionally, this design allows to adjust the electrode distance by varying the thickness of the ZrO 2 layer. This is an advantage when making use of pure, undiluted ionic liquids as electrolytes. In general, those have a low diffusion limited current, requiring to minimize the electrode distance. A disadvantage is, that a scattering layer (e.g. ZrO 2 ) must be applied as a spacer. Therefore no semi-transparent cells are practicable with a monolithic layer sequence. A further disadvantage is the required graphite layer, which must feature a low sheet resistance and good catalytic activity (like platinum). The catalytic activity can be increased by incorporating an additional thin layer of platinum between the ZrO 2 and the graphite. A low sheet resistance is more difficult to achieve, especially during the high temperature of the fusing process. Figure 4-3: Integrated series connection of monolithic-type. First the TCO layer is structured by laser scribing (1), Then the TiO 2 layer, the ZrO 2 layer, the graphite layer and the glass frit are screen printed on the respective substrates (2) and dried (3). After sintering (4) the module is fused at high temperatures (5). A hermetic seal is formed between the cells. The electrical monolithical contact is formed, requiring only one TCO glass. No silver is needed.. 46

4. Integrated series connection in dye solar modules 4.4 Conclusions of Chapter 4 Although each of the 3 types of integrated series connection has it advantages, one has to be chosen that will be incorporated into the module concept of this research. Without doubt, the highest efficiencies with DSCs around the world are achieved with the layer sequence TiO 2 /scattering layer/electrolyte/pt. This layer sequence is only attainable with the integrated series connection of Z type. The W-connection is discarded in this work, because the slight simplification in processing and cost reduction (omission of the Ag line), does not outweigh the loss in efficiency (no scattering layer, electrical mismatching) and the inhomogeneous optical appearance (back- and front side illumination). The monolithic series connection is discarded in this work, because the efficiencies are usually even lower, than for transparent DSCs. The reason for this is the high sheet resistance of the graphite layer. In this work, glass frit is used a sealing material. Under the temperatures of the glass fusing the sheet conductance of the graphite is reduced further (under ambient atmosphere). Additionally, the (thick) graphite layer complicates the colouring process. It should be noted, however, that the omission of one TCO glass in the monolithic series connection is a strong argument of further cost reduction for the DSC. The good applicability of pure, undiluted ionic liquids in the monolithic connection, because of the potentially low electrode distance is a further advantage. In conclusion, the Z-connection is chosen as the favoured type of integrated series connection. Beyond question, the highest efficiencies are attained with this design. 47

5. Charge transfer under reverse bias potential 5. Charge transfer under reverse bias potential The up-scaling of the dye solar cell (DSC) technology involves a series connection of single cells to form modules. In series connection, the voltages of the single cells add up, while the current of the module is identical to the current of one cell. A module is desirable, because a high, utilizable voltage is reached. In an outdoor situation, one or more cells of the module might be shaded. As a result, electrical mismatching occurs and the shaded cells have to transport the current of the module in reverse bias, as explained in Chapter 3.2. In order to understand and predict the behaviour of a DSC module under electrical mismatching it is therefore essential to investigate the I-V characteristic of a single DSC under forward and reverse bias potential. The I-V characteristic of a DSC under reverse bias differs significantly from a typical diode I-V characteristic. Compared to other solar cell technologies, in particular crystalline silicon solar cells, the largest sustainable reversed biased voltage (breakdown voltage) of a DSC is extremely low (approx. 5 mv) (section 5.1). In this chapter, the processes and interfaces involved in the electron transport of a DSC under reverse bias are investigated (section 5.2). Based on these results, a model is developed, which describes the I-V curve of a DSC both under forward and reverse biased voltages (section 5.3). Models have been developed to explain and analyse the behaviour of DSC under forward bias, but usually not under reverse bias [Ferber '98,Stangl '98]. With the model for the I-V curve over the complete voltage range, the behaviour of a 49

5. Charge transfer under reverse bias potential series connection of DSCs under electrical mismatching (partial shading) can be predicted (section 5.3.4). Furthermore, the long term stability of a DSC operated in reverse bias is investigated (section 5.4). Studies on DSC have been reported previously, which showed that a reverse biased potential of ~15 mv can result in a degradation of the cell s performance, due to damaging of the dye [Wheatley '3]. However, in a realistic module design, the maximum current conducted through the shaded solar cell would be equivalent to its own short-circuit current. When this is the case, the voltage drop is much lower, as will be shown in this chapter. Then, the effect of a dense TiO 2 blocking layer on the long term stability of reverse biased DSCs is investigated (section 5.4.1). A dense TiO 2 blocking layer can be used to increase the fill factor of a DSC under forward bias [Cameron '3]. However, the effect on operation under reverse bias is undesirable, as will be shown in this chapter. In section 5.5 a scattering layer with a pattern is applied to the DSC modules. The pattern allows to incorporate an image onto the module area, in order to create a deco art or logo module. And finally, in section 5.6, a method is presented, which allows the measurement of the spatially resolved photocurrent image (SRPI) of series interconnected DSC modules. 5.1 Measured I-V characteristic of a dye solar cell under forward and reverse bias potential In crystalline silicon solar cells, first at large reverse biased voltages, the so-called breakdown voltage, a sharp rise in the current occurs due to the avalanche breakdown. The value of this largest sustainable reverse biased voltage varies and can be 2 V or greater (compare Chapter 3.2.1). In a DSC, current flows in reverse bias, but no actual breakdown occurs. Also, the current will not rise as steeply. Nevertheless, the term breakdown voltage will be used in this paper to describe the reverse biased voltage, which is required to obtain a reverse current in the magnitude of the short-circuit current of the cell, i.e. about 2 ma/cm². 5

5. Charge transfer under reverse bias potential current density / ma cm -2 16 8-8 -16 V br 16 8-8 -16 Figure 5-1: Measured I-V curve of a typical DSC (area=2.5 cm²) in the dark. The breakdown voltage (V br ) is at -5 mv. -6-4 -2 2 4 6 voltage [mv] Figure 5-1 shows a typical I-V curve of a DSC (area=2.5 cm²) in the dark 1. The breakdown voltage (V br ) of about -5 mv is extremely low compared to crystalline silicon solar cells. This is an advantage when connecting solar cells in series. The power output of a solar module is much less sensitive to partial shading when each solar cell has a low breakdown voltage, as described in Chapter 3.2.2. 5.1.1 Electrical mismatching of dye solar cells in series connection In a series connection the total voltage is determined by the sum of the voltages of each cell at equal currents. Figure 5-2 shows the I-V curves of 5 identical DSC connected in series without bypass diodes. In Figure 5-2, when one cell is completely shaded, the open circuit voltage is reduced by the open-circuit voltage of this cell (about 7 mv) plus the voltage required to run the shaded cell in reverse bias (about 3 mv). The short-circuit current is only reduced very slightly, because of the low breakdown voltage of the shaded cell. Figure 5-3 shows the same measurement with only partially shaded cells. The characteristic steps or bumps in the I-V curve are clearly visible (compare Chapter 3.2.2). A similar behaviour is obtained as with crystalline silicon cells with installed bypass diodes. However, in this case no bypass diodes have been applied. Furthermore, the module in Figure 5-3 only consists of 5 individual cells. The voltage of these 5 cells is sufficient to operate the shaded cells in reverse bias. Clearly, this is only possible because of the low breakdown voltage of DSCs. A DSC can therefore be treated as a solar cell with incorporated bypass diode. 1 The manufacturing procedure is equivalent to the test cells (masterplates), which is described in the Appendix A1.2. Electrolyte composition:.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile. 51

5. Charge transfer under reverse bias potential 5 cells illuminated 1 cell shaded 2 cells shaded 3 cells shaded 4 cells shaded current density / ma cm -2 14 12 1 8 6 4 2-2..5 1. 1.5 2. 2.5 3. 3.5 voltage V [V] Figure 5-2: Measured I-V curves under illumination (1 W/m²) of an array of 5 DSC connected in series. The cells are successively shaded completely. 16 5 cells illuminated 1 cell partially shaded 2 cells partially shaded current density / ma cm -2 12 8 4..5 1. 1.5 2. 2.5 3. 3.5 voltage V [V] Figure 5-3: Measured I-V curves under illumination (1 W/m²) of an array of 5 DSC connected in series. The cells are successively partially shaded. 5.2 Investigation of the charge transfer in reverse-bias In order to identify the processes and the interfaces involved in the charge carrier transport under reverse bias, 6 electrode set-ups that consisted of single components of a standard DSC were built 2 (Figure 5-4). These electrode set-ups were chosen in such a way, that all interfaces in the DSC can be investigated independently. In this way, the dominant route of charge transfer under reverse bias will be determined. 2 The manufacturing procedure is equivalent to the test cells (masterplates), which is described in the Appendix A1.2. Electrolyte composition:.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile. 52

5. Charge transfer under reverse bias potential A Glass TCO Electrolyte B Glass TCO Dye Electrolyte TCO Glass Dye TCO Glass C Glass TCO Pt Electrolyte D Glass TCO Pt Electrolyte TCO Glass Dye TCO Glass E Glass TCO Pt Electrolyte F Glass TCO Pt Electrolyte TiO 2 TCO Glass Dyed TiO 2 TCO Glass Figure 5-4: Special electrode set-ups used for the experimental analysis of the electron transport in DSC under reverse bias. For explanation see text. Set-up A consists of two TCO-electrodes and the electrolyte. Set-up B consists of TCOelectrodes that have been flushed with dye before injecting the electrolyte. Set-up C consists of one platinum-coated electrode and one TCO-electrode and electrolyte. Set-up D is the same as C, only flushed with dye before injecting the electrolyte. Set-up E is a standard DSC set-up without dye (TCO/Pt/electrolyte/TiO 2 /TCO). Set-up F is a standard DSC flushed with dye. In the following, reverse bias refers to an electrically negative platinum electrode (electron transport from the platinum electrode to the electrolyte). Whereas, forward bias refers to an electrically positive platinum electrode (electron transport from the electrolyte to the platinum electrode). This is illustrated in Figure 5-5. Obviously, set-ups A and B are symmetric, so no direction of current can be denoted. e - forward direction reverse direction e - Glass TCO Pt Electrolyte TCO Glass Figure 5-5: Definition of forward and reverse direction of electrons in the electrode set-ups. U>: forward bias U<: reverse bias 53

5. Charge transfer under reverse bias potential 5.2.1 I-V characteristics of electrode set-ups to determine the dominating charge transfer route in reverse bias Figure 5-6 shows the I-V curves of set-ups A-E measured in the dark. As can be seen, set-ups A (TCO/electrolyte/TCO) and B (TCO/electrolyte/TCO flushed with dye) do not conduct any current at applied biases from 1.5 V to +1.5 V. At higher voltages a current would flow (about 3. V). However, such a high voltage is destructive for the electrolyte and has thus not been applied. If one electrode is coated with platinum (set-up C), the charge transfer resistance at the platinum electrode is reduced drastically. The resistance of the platinum/electrolyte interface may be neglected when compared to the uncoated TCO/electrolyte interface. Thus, the I-V characteristic is only determined by the charge transfer reaction at the uncoated TCO/electrolyte interface. The magnitude of the current densities are equal at equal forward and reverse bias potentials, as can be seen in Figure 5-6. This means, that the reaction at the TCO/electrolyte interface of set-up C is symmetric. When the TCO/electrolyte interface is covered with dye molecules (set-up D), the direction from the TCO to the electrolyte (forward bias) seems unaffected. In reverse bias, however, a much larger current density is observed, compared to the uncoated TCO/electrolyte interface. In other words, the dye catalyses the electron transfer from the electrolyte to the TCO (reverse bias), the charge transfer reaction is asymmetric. Presumably, the dye participates in the electron transfer reaction under reverse bias. It is likely that the dye is being oxidised at voltages of about.5 V. This oxidation is reversible as the dye is being reduced by the iodide again. When the TCO is coated with a nanocrystalline layer of TiO 2 (set-up E), the I-V curve shows that the electron transfer from the electrolyte to the TCO (reverse bias) is not significantly affected by the nanoporous TiO 2 layer. However, the electron transfer from the TCO to the electrolyte (forward bias) is facilitated by the TiO 2. Under forward bias potential, electrons are injected from the TCO to the conduction band of the TiO 2. From the TiO 2 the electrons can recombine with the triiodide in the electrolyte. This route is the dominant recombination mechanism in DSCs. The involved processes of electron transport in the nanocrystalline TiO 2 and the electron recombination from the TiO 2 to the electrolyte are well known from detailed models and studies describing the DSC under forward bias in the dark [Dloczik '97,Nelson '99]. 54

5. Charge transfer under reverse bias potential 4 current densitiy / ma/cm 2 2-2 -4-1.5-1. -.5..5 1. 1.5 voltage / V A: TCO/El/TCO without dye B: TCO/El/TCO with dye C: TCO/El/Pt/TCO without dye D: TCO/El/Pt/TCO with dye E: TCO/TiO 2 /El/Pt/TCO without dye Figure 5-6: Measured I-V curves of set-ups A-E measured in the dark. 5.2.2 Interfaces studied by electrical impedance spectroscopy Electrode set-ups C, D and F were studied using electrical impedance spectroscopy (EIS). All EIS measurements were carried out in the dark with different bias voltages. The amplitude of the modulated voltage was 5 mv. The interpretation of EIS data of DSC is carried out according to Kern et al. [Kern '2] Figure 5-7 shows the impedance of set-up C (TCO/Pt/electrolyte/TCO) compared to setup D (TCO/Pt/electrolyte/TCO flushed with dye) at a reverse biased voltage of.8 V. Without the dye (C), the impedance is determined by the high resistance of the TCO/electrolyte interface. When dye molecules are attached to the TCO electrode (D), the resistance drops by one magnitude under reverse bias. In addition to the peak in the phase at 1 khz associated with the Pt/electrolyte interface, a peak is visible at about 5 Hz, which can be associated with the electron transfer at the TCO/dye/electrolyte interface. The resistance of the TCO/dye/electrolyte interface now has a magnitude which is approximately equivalent to that of the Pt/electrolyte interface. Figure 5-8 again shows the impedance of set-up D (TCO/Pt/electrolyte/TCO flushed with dye), at reverse and forward biased voltages of -.8 V and +.8 V. Under a forward bias of +.8 V, the TCO/dye/electrolyte interface acts like set-up C: the dye is not catalysing the electron transfer in this direction. Under reverse bias the dye catalyses the electron transfer. The resistance drops and the two peaks associated with the platinum/electrolyte interface and TCO/dye/electrolyte interface are visible, like in Figure 5-7. Figure 5-9 shows the impedance of set-up F (a complete DSC) at reverse and forward biased voltages of +.8 V and.8 V. A phase peak at about 1 Hz is associated with electrons recombining from the TiO 2 to the electrolyte, and a phase peak at about 1 khz is 55

5. Charge transfer under reverse bias potential associated with the electron transfer at the Pt/electrolyte interface. Under reverse bias, the TiO 2 is not significantly involved in the electron transfer, as the peak in the phase at 1 Hz is missing. The high resistance of the TCO/electrolyte interface determines the impedance signal. The attached dye at the TCO/electrolyte interface does not reduce the resistance as drastically as for set-up D, since in this case the available surface of the TCO is much smaller due to the TiO 2. Electrode set-up C: TCO/Pt/Electrolyte/TCO Electrode set-up D: TCO/Pt/Electrolyte/TCO flushed with dye Impedance / Ω 1-6 -4 1-2 Reverse-bias: -.8 V 1 1 1 1 1k 1k 1k Frequency / Hz Phase / deg Figure 5-7: Measured EIS of set-ups C and D under reverse bias of.8 V. open symbols: Z ; solid symbols: phase (Z) Electrode set-up D: TCO/Pt/Electrolyte/TCO flushed with dye Rerverse-bias -.8 V Forward-bias +.8 V Impedance / Ω 1 1 1 1 1 1 1k 1k 1k Frequency / Hz -6-4 -2 Phase / deg Figure 5-8: Measured EIS of set-up D biased with +.8 V and.8 V. open symbols: Z ; solid symbols: phase (Z) 56

5. Charge transfer under reverse bias potential Electrode set-up F: complete DSC Rerverse-bias -.8 V Forward-bias +.8 V 1-6 Impedance / Ω 1-4 -2 Phase / deg Figure 5-9: Measured EIS of set-up F biased with +.8 V and.8 V. open symbols: Z ; solid symbols: phase (Z) 1 1 1 1 1k 1k 1k Frequency / Hz 5.3 Model for the I-V characteristics of a dye solar cell under forward and reverse bias potential In the previous section, the charge transfer routes under reverse and forward bias have been investigated. In principle, the four possible routes of charge transfer at the TiO 2 front electrode shown in Figure 5-1 are under forward bias: Route 1: recombination of electrons from the conduction band or the surface states of the TiO 2 with the electrolyte. Route 3: reduction of I 3 - at the TCO substrate under forward bias. Figure 5-1: The four possible routes of electron transfer under forward and reverse bias at the TiO 2 electrode. 57

5. Charge transfer under reverse bias potential under reverse bias: Route 2: injection of electrons directly from the electrolyte (I - ) into the TiO 2. Route 4: oxidation of I - at the TCO substrate under reverse bias. As the most important result from the previous section, it has been found that the dominant route of charge transfer under reverse bias is route 4 in Figure 5-1. Route 2 may be neglected. Furthermore, route 4 is catalysed by the dye, but route 3 is not catalysed by the dye. For the functioning of the DSC it is crucial, that route 3 has a high charge transfer resistance and is not catalysed by the dye [Cameron '5]. 5.3.1 One-diode model for the dye solar cell under forward bias potential Under forward bias, electron recombination mainly occurs from the TiO 2 conduction band or surface states with the electrolyte (Figure 5-1, route 1). Electron recombination - by reduction of the I 3 ions at the TCO substrate can be neglected under short-circuit conditions, but becomes important at open-circuit conditions (Figure 5-1, route 3). Especially under low light intensities, the back reaction of electrons via the TCO substrate is the dominant route [Cameron '5]. The I-V characteristic under forward bias is well known from detailed models [Ferber '98,Stangl '98]. For the purpose of this work, however, the one-diode model is sufficient to describe the I-V curve in forward bias (Chapter 2.6.1): j( V ) = j sat ev ers j( V ) exp 1 + mdk BT j SC V RS j( V ) + RSh 14243 j Sh 5-1 j sat : saturation current density of the diode in ma/cm² m D : diode ideality factor, dimensionless j SC : photocurrent density (short-circuit current density) in ma/cm² R s : internal series resistance related to area in Ωcm² R sh ; internal shunt resistance related to area in Ωcm² j Sh : shunt current density in ma/cm² Here, the diode ideality factor (m D =1 2) empirically accounts for the fact, the real diodes deviate from the ideal diode equation, where m D =1. 5.3.2 Butler-Vollmer model for the dye solar cell under reverse bias potential Under reverse bias, electron transfer mainly occurs via the TCO/electrolyte interface by oxidation of the I - ions (Figure 5-1, route 4). A compact TiO 2 blocking layer can suppress this route [Hore '5]. However, in this work for DSC modules an easy electron transfer 58

5. Charge transfer under reverse bias potential under reverse bias over the substrate is desired. Therefore, blocking layers have not been applied. The TCO layer, which is not coated with platinum behaves essentially as a noncatalytic metal. In the absence of diffusion limitations, the total current density over the TCO/electrolyte interface is described by the Butler-Vollmer equation (compare Chapter 2.5.1). In general, the Butler-Vollmer equation describes the charge transfer over a metal/electrolyte interface, which is not determined by diffusion limitation in the electrolyte. Then the charge transfer only depends on the applied potential: e e j( V ) = j V V exp β exp (1 β ) 5-2 kbt k BT The relation between the current density j and the overvoltage V is determined by two parameters: the exchange current density j and the symmetry parameter β. A symmetry parameter of.5 describes a charge transfer reaction, which is symmetric under forward and reverse bias. The exchange current density varies on the type of TCO glass used, as well as on thermal treatment of the glass [Cameron '5]. Adsorbed impurities on the TCO may act as catalytic sites. In section 5.2 it has been shown, that in particular, the dye molecules are adsorbed to the TCO substrate and catalyse the electron transfer reaction. In this case the symmetry parameter greatly differs from.5, favouring one direction of the electron exchange. In Figure 5-11 the Butler-Vollmer equation is fitted to the I-V curves of set-ups C-E. The fits do not agree very well to the experimental results, because the Butler-Vollmer equation only describes the electron transfer at one TCO/electrolyte interface. The charge transport in the electrolyte and the charge transfer at the platinum electrode are neglected in this approach. However, the fits account for the asymmetric behaviour of the measured I-V characteristics. The reaction at the TCO/electrolyte interface of set-up C is symmetric, yielding a symmetry parameter of β =.5. When the TCO/electrolyte interface is covered with dye molecules (set-up D), the exchange current density over this interface increases over 3 magnitudes and the symmetry parameter is about.35. In other words, the dye catalyses the electron transfer from the electrolyte to the TCO (reverse bias). The direction from the TCO to the electrolyte (forward bias) seems unaffected. When the TCO is coated with a nanocrystalline layer of TiO 2 (set-up E), a symmetry parameter of β =.54 is obtained. In this case, the Butler-Vollmer equation is not at all suited to describe the I-V characteristic under forward bias. It is only used to determine a symmetry parameter for comparison. Under forward bias, the one-diode model should be used, as will be shown in the next section. 59

5. Charge transfer under reverse bias potential 3 current density / ma/cm 2 2 1-1 -2-3 -1.5-1. -.5..5 1. 1.5 voltage / V C: TCO/El/Pt/TCO without dye D: TCO/El/Pt/TCO with dye E: TCO/TiO 2 /El/Pt/TCO without dye Figure 5-11: The I-V curves of set-ups C,D and E were fitted to the Butler-Vollmer equation. The fit parameters were: C: j = 1.4 1-1 ma/cm² β=.5 D: j = 1.4 1-7 ma/cm² β=.35 E: j = 9.5 1-9 ma/cm² β=.54 5.3.3 Complete model for the I-V characteristic of a dye solar cell over the full voltage range In order to obtain a complete model for the I-V characteristic of a DSC over the full voltage range, the one-diode model and the Butler-Vollmer model are combined. The one-diode model (Equation 5-1) is extended in reverse bias by the Butler-Vollmer model (Equation 5-2). The Butler-Vollmer model is only valid in reverse bias. In reverse bias (V<), the first exponential term may be neglected and under forward bias (V>) the second exponential term does not contribute to the current density: j( V ) = e e j exp V exp (1 β ) V kbt kbt 144 243 4 14442444 3 for V < for V > β 5-3 The complete model for the I-V characteristic of a dye solar cell over the full voltage range may therefore be written as: 6

5. Charge transfer under reverse bias potential ev ers j( V ) V RS j( V ) e j( V ) = jsat exp 1 SC exp (1 ) Dk + j + j β V B Sh k 5-4 B 14 m T 4444444 2444444444 R 3 144 T 44 24444 3 one diode model Butler Vollmer model j Sat : saturation current density of the diode in ma/cm² m D : diode factor, dimensionless j SC : photocurrent density (short-circuit current density) in ma/cm² R s : internal series resistance related to area in Ωcm² R sh ; internal shunt resistance related to area in Ωcm² β: symmetry parameter, dimensionless j : exchange current density in ma/cm² Figure 5-12 shows a measured I-V curve of a DSC with a fit of Equation 5-4. As can be seen, the model is well suited to describe the I-V curve of a DSC under reverse and forward bias potential. 5.3.4 Model for the electrical mismatching of dye solar cell in series connection To calculate the I-V curve of a series connection of DSC, the voltages of the individual cells are added at equal currents: V total n ( j) = V ( j) 5-5 i= 1 i Here, V i (j) is determined by numerically inverting Equation 5-4. Figure 5-13 shows the calculated I-V curve of 2 DSCs interconnected in series. The two individual DSCs (red and green) are not electrically matched. A bump occurs in the resulting I-V curve (black). Obviously, the short-circuit current densities of the individual cells can be determined by the steps in the resulting I-V curve. 61

5. Charge transfer under reverse bias potential current densitiy / ma cm -2 2 15 1 5-5 -1-15 Model Measurement reverse bias: Butler-Volmer model forward bias: one-diode model -2-8 -6-4 -2 2 4 6 8 voltage / mv Figure 5-12: Measured I-V curve of a DSC with a fit of the model for forward and reverse bias potential. current density / ma cm -2-5 -1-15 -2 + = -25-1. -.5..5 1. 1.5 voltage / V Figure 5-13: Calculated I-V curve of 2 DSCs interconnected in series. The voltages of the individual DSCs add up at equal currents. The two individual DSCs (red and green) are not electrically matched. A bump occurs in the resulting I-V curve (black). 62

5. Charge transfer under reverse bias potential 5.4 Long term stability of reverse biased DSC In terms of power output, the low breakdown voltage of a DSC is advantageous for the performance of DSC modules under partial shading. However, the long term stability of a reverse biased DSC has to be addressed. A shaded cell still has to transport the total current of the module. The voltage which is required to operate the shaded cell under this reverse current is determined by its I-V characteristic in reverse bias. A sufficiently large reverse biased voltage might therefore damage the cell, either by decomposition of the electrolyte solvent, or by damage to the dye. In practical terms, a potential greater than 15 mv may damage a cell irreversibly [Wheatley '3]. The highest reverse current and thus the highest reverse biased voltage will occur when one cell is completely shaded and the module is operated under short-circuit conditions. Since all cells in a module should be electrically matched, the shaded cell then would conduct the equivalent of its own short-circuit current. In order to simulate this worst-case situation, in Figure 5-14 a DSC is shown, which is operated at a constant reverse current of 14 ma/cm² in the dark, corresponding to the cell s short-circuit current density under 1 sun. The voltage drop over the cell was monitored for more than 1 hours. This seems to be a relevant time scale for the occasional partial shading of a module under outdoor conditions. For the first 6 hours the voltage drop slowly rises at about 1.5 mv/h. It reaches saturation at a voltage below.7 V. The voltage seems to stay well below critical values of 1.5 V which would damage the cell. In Figure 5-15 the efficiency of the cell was measured during the time of exposure to the reverse current. The efficiency is rising slightly and reaches saturation after about 2 hours. The slight rise in efficiency was due to an increase of the fill factor. In Figure 5-16 the electrical impedance spectra of the cell are shown at different points in time during the exposure to the reverse current. The measurements were carried out in the dark at 72 mv forward bias, corresponding to the open-circuit voltage of the cell under 1 sun. The amplitude of the modulated voltage was 5 mv. The shift of the high frequency peak corresponds to a decrease in the charge transfer resistance at the Pt/electrolyte interface [Kern '2]. This conforms to the observed rise in fill factor. 63

5. Charge transfer under reverse bias potential U / V -.8 -.6 -.4 -.2. 15 3 45 6 75 9 15 12 135 time / hours Figure 5-14: A constant reverse current of 14 ma/cm² is applied to the cell. The voltage drop over the cell is monitored over time. A time-scale of 13 hours is relevant only for the occasional partial shading of a module. eta / % 7 6 5 4 3 2 1 2 4 6 8 1 12 time / hours Figure 5-15: A constant reverse current of 14 ma/cm² is applied to the cell. The efficiency of the cell is monitored over time. 1 before applying 14 ma/cm 2 after 18 h after 37 h after 15 h -3 Impedance / Ω 1-25 -2-15 -1-5 Phase / deg Figure 5-16: A constant reverse current of 14 ma/cm² is applied to the cell. Electrical impedance spectra were measured at subsequent points in time. open symbols: Z ; solid symbols: phase (Z) 1 1m 1 1 1 1k 1k Frequenz / Hz It should be mentioned, that a blocking layer (apart from decreasing the power output of a partially shaded DSC module) might also reduce the stability of a partially shaded DSC module. The voltage required to run a DSC in reverse bias is significantly higher with an incorporated blocking layer. This will be discussed in the next section. Furthermore, it should be noted, that the stability of a DSC under reverse bias has only been demonstrated for a time scale, which seems relevant for the occasional partial shading of a module under outdoor conditions. In this work, all test cells showed an overall increase in conversion efficiency during the first 1 hours of exposure to the reverse current. However, long term testing (>1 h) of reverse biased DSCs might still reveal degradation effects due to operation under reverse bias. The slight decrease in conversion efficiency between 9 h and 11 h in Figure 5-15 might indicate the beginning of such a long term degradation. However, on a longer time scale more data must be 64

5. Charge transfer under reverse bias potential collected to safely attribute this degradation to reverse biasing. In Chapter 8 long term stability tests are performed for more than 1 hours. Since long term stability is of general concern for DSCs, a (slow) degradation of reversed biased DSCs might have other reasons, than the reverse biased voltage. For a relevant time scale for partial shading it can be concluded, however, that reverse biasing with the appropriate reverse biased voltage is not an additional degradation mechanism. 5.4.1 Charge transfer under reverse bias suppressed by a dense TiO 2 blocking layer To verify the hypothesis that a dense TiO 2 blocking layer suppresses the charge transfer route under reverse bias, DSC with and without blocking layer were manufactured. Figure 5-17 shows the I-V characteristics of DSCs with and without blocking layer. The blocking layer was about 2-5 nm and was deposited by spray pyrolysis of 3.5 ml acetyl acetone with 1 ml titanium isopropoxide and 12 ml ethanol. The charge transfer from the electrolyte to the TCO is clearly suppressed by a TiO 2 blocking layer up to a reverse bias of at least -1.5 V. without blocking layer with blocking layer current density / ma cm -2 4 2-2 -4-1.5-1. -.5..5 1. voltage / V Figure 5-17: I-V characteristics of DSC with and without blocking layer. Charge transfer from the electrolyte to the TCO is suppressed by a TiO 2 blocking layer. If the reverse biased voltage is increased, charge transfer occurs through the dense blocking layer, at about -1 V. This has been measured in Figure 5-18. However, the voltages needed to operated the cell at a reverse current, which is about equal to its own short-circuit current density, i.e. about -1 V, are almost immediately destructive for the dye and the electrolyte. The cell which has been measured in Figure 5-18 was completely destroyed in less than 5 minutes, upon application of a constant reverse current of 14 ma/cm². 65

5. Charge transfer under reverse bias potential 8 current density / ma cm -2 6 4 2-2 -4-6 -14-12 -1-8 -6-4 -2 2 voltage / V Figure 5-18: I-V characteristics of DSC with blocking layer in the dark. At higher reverse biased voltage a current starts to flow through the blocking layer. However, the high voltages are destructive for the cell. 5.5 Pattern in the scattering layer of dye solar modules The results of this chapter confirm, that DSC modules are comparatively insensitive to partial shading both in terms of power output and long term stability (without blocking layer). In the course of this research, the idea evolved to demonstrate this property of DSCs in a potentially new application for solar cells [Brandt '6]. Since DSCs can be manufactured semi-transparent without scattering layer and opaque with scattering layer, it is possible to incorporate an image or pattern on the solar cell. The image may be extended over all cells in a module to produce an art deco or logo module. The architecture of a DSC with a structured scattering layer is shown in Figure 5-19. Art deco modules and logo modules with such inhomogeneous design result in a variation of the short-circuit current of individual cells. This problem can be treated the same way as the partial shading of a module. 66

5. Charge transfer under reverse bias potential Figure 5-19: Architecture of a DSC with structured scattering layer. Using the module architecture, which is described in detail in Chapter 7.4 (interdigital meander design), art deco and logo modules of 3 different designs have been realized. A module for the presentation at the Hanover Trade Fair 26 features the logo of the Hanover Fair. Additionally, an imaginative DSC logo has been designed. And to demonstrate an art deco module, the sun module was created. In Figure 5-2, Figure 5-21 and Figure 5-22 these DSC modules are shown, respectively. The I-V characteristics are shown as measured immediately following fabrication 3. The electrolyte of the module consisted of.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile for the Hannover Fair module and the DSC logo module. The sun module contained an electrolyte of.8 M propylmethylimidazolium iodide,.15 M I 2,.5 M n-methylbenzimidazole and.1 M EMI-SCN in acetonitrile. The measurement was performed under a sulphur lamp at AM 1.5 global (1 W/m²) taking into account spectral mismatch at a temperature of 38 C. The measured I-V curves were fitted with the model described in section 5.3 applied on a series connection of 6 DSCs. The electrical mismatching can be seen clearly in the bumps or steps in the I-V curve. In the fits, the short-circuit current of each cell was determined, taking into account the known layout of transparent and opaque area. Transparent areas (without scattering layer) produce less current than opaque areas (with scattering layer). The different shortcircuit current densities (j transparent and j opaque ) can be calculated from the steps in the measured I-V curves [Brandt '6]. 3 The manufacturing procedure is described in the Appendix A1.1. 67

5. Charge transfer under reverse bias potential Figure 5-2: Hanover Fair logo module and I-V curve. current / A.2. -.2 -.4 -.6 -.8 Fit Measurement -1. -4-3 -2-1 1 2 3 4 voltage / V Isc = -758.14 ma jsc of one cell = -7.74 ma/cm 2 Voc = 4. V voc of one cell = 667 mv FF=58 % eta = 2.98 % A active = 588 cm 2 A one cell = 98 cm 2 The slight discrepancy of the calculated and measured I-V curves has mainly two reasons. Firstly, j transparent and j opaque are assumed to be constant over the whole module area. This is not the case, as will be shown in the next section. And secondly, the breakdown voltage has been assumed to be equal for all 6 cells. This is also not the case, since the breakdown voltage can vary even for cells with the same characteristics under forward bias. From the I-V curve in e.g. Figure 5-2 it is clear, that valuable information can be gained about the short-circuit currents of individual cells in a DSC module by the steps in the I-V curve, also at reverse bias potentials. Therefore, even non-patterned DSC modules should always be measured over the full voltage range. 68

5. Charge transfer under reverse bias potential Figure 5-21: DSC logo module and I-V curve. current / A.2. -.2 -.4 -.6 -.8 Fit Measurement -1. -4-3 -2-1 1 2 3 4 voltage / V Isc = -761.39 ma jsc of one cell = -7.77 ma/cm 2 Voc = 4.1 V voc of one cell = 683 mv FF=59 % eta = 3.14 % A active = 588 cm 2 A one cell = 98 cm 2 5.5.1 Applications of decorative solar modules Potential first commercial applications of DSCs are expected in products, which do not primarily require high efficiencies. In solar architecture a photovoltaic facade or roof light might be desired with a different optical appearance, than achievable with conventional solar technologies. Here, DSC modules can offer a different colour, particularly amber, and semi-transparency. Additionally, an image, logo or pattern can be designed into the module, as demonstrated here. 69

5. Charge transfer under reverse bias potential Figure 5-22: Sun module and I-V curve. current / A.2. -.2 -.4 -.6 -.8 Fit Measurement Isc= -582 ma jsc of one cell= -6.5 ma/cm² Voc=4.5 V voc of one cell= 75 mv FF=69.6 % eta=3.1 % A active= 588 cm² A one cell=98 cm² -1. -4-3 -2-1 1 2 3 4 voltage / V Additionally, solar advertising signs offer an enormous potential. A large area solar advertising sign would collect electrical energy during daytime. During the night the electrical energy would be used to illuminate the advertising sign. The advantage of DSCs is the integration of the solar cell directly into the advertising sign, thus saving area and electrical power at night. The skyline in many mega cities with luminous advertising at night is a huge energy load. In Shanghai, for example, the luminous advertising is switched off at about 11 pm due to the enormous energy costs. 7

5. Charge transfer under reverse bias potential 5.6 Spatially resolved photocurrent imaging technique for large area, series interconnected dye solar modules In this section, a spatially resolved photocurrent imaging (SRPI) technique for large area, series interconnected DSC modules is developed. SRPI is a very valuable measuring technique to investigate the up-scalability of DSCs. With SRPI it is possible to study inhomogeneities in the screen printed layers, as well as inhomogeneities, which might occur due to the colouring process or electrolyte distribution. In SRPI a monochromatic, focused light beam is scanned over the area of the module and the spatial resolved photocurrent is recorded. The module is kept in the dark, and is only illuminated with the focused, monochromatic light beam. SRPI is an easy measuring technique and has been applied to single DSCs before [Macht '2]. However, in a module of 6 series interconnected cells, the light induced photocurrent cannot be measured in short-circuit conditions (V=), because only one cell is slightly illuminated and the other 5 cells are kept in the dark, thus blocking the current. With the results of this chapter, it is possible to measure the light induced photocurrent of one cell under reverse bias potential. Figure 5-23 shows a simulation of a series connection of 6 cells. In the simulation, 5 cells are kept in the dark (I sc =) and 1 cell is slightly illuminated (I sc =.5 ma or 1 ma). It can be seen, that the light beam induced current can be measured at a reverse bias potential of about -1. V. The optimal reverse biased voltage to measure the light beam induced current varies slightly with the magnitude of the induced current. Assuming, that the inhomogeneities over the area are small, this variation may be neglected. Thus, the spatially resolved, light beam induced current can be measured. From Figure 5-23, however, it is clear, that even for areas, which actually produce no current at all, a small offset will be measured..5 current / ma. -.5-1. 5 cells with I sc = ma, 1 cell with I sc = ma I sc =.5 ma I sc = 1 ma -1.5-2. -1.5-1. -.5..5 1. 1.5 2. voltage / V Figure 5-23: Calculated I-V curve of a series connection of 6 DSCs. The module is kept in the dark and only 1 cells illuminated with a light beam. 71

5. Charge transfer under reverse bias potential Figure 5-24: Spatially resolved photocurrent image of a DSC module. Figure 5-24 shows the measured SRPI for a DSC module with a homogeneous scattering layer. The light beam had a wavelength of 532 nm and the illuminated spot had a size of 2 mm x 2 mm. The reverse bias potential was -1.4 V. The measurement was carried out at Matthias Klockhaus in Gelsenkirchen. All currents, which are lower than 64 % of the maximum current are displayed in black colour. As mentioned above, a small current is measured even for areas which actually produce no current due to the applied reverse bias potential. Additionally, the illumination spot of 2 mm x 2 mm is too large to obtain a good resolution. The glass frit grid (dead area) has a width of 2.5 mm. Moreover, reflectance of the glass reduces the resolution further. In the upper right corner, a cable from the measuring apparatus can be seen. Overall, the current is reasonably homogenous (the green parts). However, there are some parts which generate considerably less current (the blue and black parts). Most of the black parts can be identified as air bubbles in the electrolyte. The filling holes of this module have been (inadequately) sealed with a polymer hotmelt foil. Surprisingly, there are some parts which generate significantly more current than average (the red parts). The reason for this might be a slight variation of electrode distance, which could also influence the colouring process. By visual inspection, the module s layer and colour appears homogenous. It can be calculated, that if all areas would produce the maximum (red) current, the total current of the module could be increased by 2.3 %. 72

5. Charge transfer under reverse bias potential Figure 5-25: Spatially resolved photocurrent image of a DSC module with a structured scattering layer (DSC logo). Figure 5-25 shows the measured SRPI for a DSC module with a structured scattering layer (DSC logo). The light beam had a wavelength of 532 nm and a the illuminated spot had a size of 2 mm x 2 mm. The reverse bias potential was -1.4 V. The pattern in the scattering layer can clearly be seen. The current at the positions with scattering layer is reasonably homogenous (the green parts). And the current at the positions of the transparent cell is reasonably homogenous as well (blue parts). The black parts can be identified as air bubbles in the electrolyte. However, again there are some parts which generate significantly more current than average (the red parts). It can be noticed, that the highest currents always occur in the curves of a meander. This leads to the assumptions, that the inhomogeneities in the current are produced by the colouring process. It can be calculated, that if all areas would produce the maximum (red) current, the total current of the module could be increased by 5.2 %. Of course, this would assume a homogenous scattering layer over the whole module area. 73

5. Charge transfer under reverse bias potential Figure 5-26: Spatially resolved photocurrent image of a DSC module with a structured scattering layer (DSC logo). A part of the module area is scanned with a higher resolution. In Figure 5-26 a section of the module in Figure 5-25 has been scanned with a higher resolution. The illuminated spot has a size of.2 mm x.2 mm. The chosen section is part of the bottom cell in Figure 5-25, where localized spot of high currents were measured. It can bee seen, that the current is higher at the bottom of the magnified section. Furthermore, the area of the high currents seems to resemble the flow direction of the colouring solution. In particular, the points of maximum current do not reach into the corners in the shown meander. The area of low current (blue) in the upper part of Figure 5-26 is produced by the structure in the scattering layer. Here, the lower part of the letters S and C can be seen. The small stripes in the lower part are artefacts from screen printing. These stripes can be seen clearly by visual inspection, too. 5.7 Conclusions of chapter 5 The processes and interfaces involved in the electron transport of a DSC under reverse bias were studied. 74

5. Charge transfer under reverse bias potential It was shown that charge transport under reverse bias directly occurs via the electrolyte/tco interface, catalysed by the dye molecules. The charge transfer is highly asymmetrical, with a symmetry parameter of approx..35, using a simple Butler-Vollmer model. Thus, the dye catalyses the electron transfer at the TCO/electrolyte interface only in reverse bias, but not under forward bias. Presumably, the dye undergoes a reversible oxidation, participating in the electron transfer reaction under reverse bias. Electrical impedance spectroscopy revealed that in contrast to forward bias conditions, the TiO 2 -layer is not significantly involved in the electron transport under reverse bias. Thus, a model for the I-V curve of a DSC was developed over the whole voltage range. In forward bias, the one-diode model has been used, and in reverse bias, the Butler- Vollmer model. In this way, a series connection of DSCs can be modelled under electrical mismatching. Due to the low breakdown voltage (approx. 5 mv under reverse bias), no bypass diodes are required when interconnecting DSC in modules, with respect to power output. With respect to long term stability, no degradation in overall efficiency was observed after operating a DSC at a constant reverse current for 1 hours. This seems a relevant time scale for the operation of a DSC module under outdoor conditions subjected to occasional partial shading. The magnitude of the reverse current applied to the DSC corresponded to the short-circuit current of an illuminated cell in the module, in order to simulate the condition in a partially shaded DSC module. It has to be noted though, that bypass diodes must be used over module terminals in case a cell or a group of cells undergoes degradation, for example, by leakage. A dense TiO 2 blocking layer, which can be applied in order to increase the fill factor, suppresses the charge transfer under reverse (and forward) bias over the TCO/electrolyte interface. In a DSC module this is not desired. Under partial shading, the voltage drop over the shaded cell would reach critical values due to the blocking layer, thus destroying the module. Art deco and logo modules have been produced, which feature an image over the whole module area. The image is attained by structuring the scattering layer of the module. Thus, each cell in the module produces a different current. The resulting I-V curves can be calculated by the developed model. Furthermore, a method was developed to measure the spatially resolved photocurrent image (SRPI) of a series interconnected DSC module. The small light beam induced photocurrent of one cell can be measured in reverse bias, although all other cells in the series connection are kept in the dark. Additionally it should be mentioned, that valuable information can be gained about the short-circuit currents of individual cells in a DSC module by the steps in the I-V curve, also at reverse bias potentials. Therefore, even non-patterned DSC modules should always be measured over the full voltage range. 75

6. Photoinduced electrophoresis in the electrolyte of dye solar modules 6. Photoinduced electrophoresis in the electrolyte of dye solar modules As the DSC technology progresses from laboratory scale to large area applications, long term stability is one major issue. Especially for large area DSC modules, stability is related to hermetic sealing. The sealing material must be mechanically and thermally stable and be chemically inert against the I - / I 3 - redox couple. The sealing material must protect the conductor grid and ensure hermetic external sealing. In a series connection of DSCs the sealing material must not only guarantee hermetic external sealing but also hermetic internal sealing. Each cell in a series connection of DSCs must be a closed compartment, i.e. mass transport of the electrolyte between neighbouring cells must be prevented. When cells in series connection are illuminated under open-circuit conditions, the individual cells have different electrochemical potentials. If there were a possibility for ion exchange between the redox electrolytes of neighbouring cells, the redox couple (triiodide and iodide) would be separated from each other. In this work, this process is called photoinduced electrophoresis. During the photoinduced electrophoresis the electrical parameters of the module would slowly drop, due to the change of ion concentrations in the respective cells. Therefore, photoinduced electrophoresis is a potential degradation mechanism as a result of inadequate internal sealing of DSC modules with integrated series connection [Ziegler '5]. 77

6. Photoinduced electrophoresis in the electrolyte of dye solar modules In this chapter a thorough theoretical and experimental study of the photoinduced electrophoresis is presented. In a diffusion model, the decrease of the short-circuit current density of a DSC module with inadequate internal sealing is calculated under illumination (section 6.1). The quality of the internal sealing regarding its barrier properties against diffusion of the I - / I 3 - redox couple may be estimated using this model. The theoretical model is verified by photoinduced electrophoresis experiments on DSC modules, which were manufactured with a defined electrolyte canal between two cells in series connection (section 6.3). In the dark, the separation of the redox couple should be reversed by diffusion, as predicted by the model. However, in the experiments the degradation of electrical parameters never fully reversed. This implies, that the separation of the redox couple leads to further, irreversible degradation mechanisms, which are unaccounted for in the model. 6.1 Model of the photoinduced electrophoresis in dye solar modules In an isothermal system consisting of an undissociated solvent and dissolved substances or ion species whose concentrations are by orders of magnitude smaller than that of the solvent, mass transport is generally the sum of diffusion, migration, and convection. In the electrochemical system of the DSC, we will neglect convection, since the typical thickness of a DSC is smaller than 5 μm and typical diffusion layers are in the order of 1 μm in a DSC [Hauch '98]. The field driven current (migration) will also be neglected, since the electrolyte contains an excess of cations, which are not participating in the redox reaction and are counterbalancing the electrical field in the electrolyte. Thus, only diffusion will be considered as the dominant form of mass transport [Papageorgiou '96b] (compare 2.3.1). 6.1.1 Energy levels in a series connection of DSC Figure 6-1 shows the energy levels of two DSCs in series connection under short-circuit conditions. HOMO and LUMO are the highest occupied and the lowest unoccupied molecular orbital of the dye, respectively. In series connection, the quasi fermi level of the electrons in the conduction band of the TiO 2 (E n f TiO 2 ) of cell 1 and the redox level of the electrolyte (E REDOX ) of cell 2, are on the same level at the point of series connection. The metallic (Ag) contact interconnects the TCO of the front electrode of cell 1 with the platinum electrode of cell 2. The fermi energies of electrons in the platinum electrode and in the electrolyte are assumed to be equal at the interface. Under short-circuit conditions, the energies are on the same level at the point of external contact. Figure 6-1 is a schematic diagram, in particular the gradient of the redox energy of the electrolyte is exaggerated for illustration purpose. 78

6. Photoinduced electrophoresis in the electrolyte of dye solar modules Figure 6-1: Qualitative energy scheme of two DSCs in series connection under shortcircuit conditions. HOMO and LUMO are the highest occupied and the lowest unoccupied molecular orbital of the dye, respectively. In a series connection, the quasi fermi level of the electrons in the conduction band of the TiO 2 (E n f TiO 2 ) of cell 1 and the redox level of the electrolyte (E REDOX ) of cell 2, are on the same level at the point of series connection. Under short-circuit conditions, the energies are on the same level at the point of external contact. The arrows indicate the direction of transport for the negative charge carriers (electrons or iodide) The Figure is a schematic diagram, in particular the gradient of the redox energy of the electrolyte is exaggerated for illustration purpose. If there were a possibility for ion exchange between the electrolytes of both cells, no separation of the redox couple would occur under short-circuit conditions, since no direction is favoured in Figure 6-1. Figure 6-2 shows the same situation under open-circuit conditions. In open-circuit, no external current flows. Therefore, no significant gradient in the fermi energies is formed. In one cell an electron is photo-excited from the HOMO to the LUMO and injected into the conduction band of the TiO 2. From the conduction band of the TiO 2 the electrons recombine with the electrolyte, which reduces the oxidized dye. In series connection under open-circuit, the redox energies of the electrolyte of both cells are clearly on different levels. If ion exchange was possible, the separation of the redox couple would now occur (the red path in Figure 6-2). In the following, we will therefore only consider the situation under open-circuit conditions, since this is the operating point of maximal occurring photoinduced electrophoresis. 79

6. Photoinduced electrophoresis in the electrolyte of dye solar modules Figure 6-2: Qualitative energy scheme of two DSCs in series connection under opencircuit conditions. Electrons are photo-excited from the HOMO to the LUMO and injected into the conduction band of the TiO 2, from there the electrons recombine with the electrolyte. In series connection, the redox energy (E REDOX ) of the electrolyte of cell 2 is on the same level as the quasi-fermi energy of the electrons in the TiO 2 of cell 1. If a path for the electrolyte existed between both cells, separation of the redox couple would occur (red path). 6.1.2 Model system for the theory of photoinduced electrophoresis Figure 6-3 shows the cross section of two DSCs in series connection in open-circuit with a possibility for ion exchange in the electrolyte. The greatest difference in fermi energy is between the platinum electrode of cell 1 (+) and the TiO 2 of cell 2 (2-). The redox energies of the electrolytes of both cells are on a different level, as described in the previous section. The difference is exactly the open-circuit voltage of one cell, because the fermi energy of electrons in the platinum electrode is equal to the fermi energy of the electrons in the electrolyte at the interface: E 6-1 F, Redox EF,Pt _ TCO TiO 2 TCO TiO 2 2 _ I 3 - platinum + TCO I - platinum TCO Figure 6-3: Cross section of two DSCs in series connection with a possibility for ion exchange in the electrolyte. Under open-circuit conditions, the situation regarding mass transport in the electrolyte is then equivalent to one Pt/electrolyte/Pt cell with large electrode spacing and applied bias voltage of about 7 mv (open-circuit voltage of the DSC). 8

6. Photoinduced electrophoresis in the electrolyte of dye solar modules The electrolytes of both cells are now allowed to exchange ions. It is assumed, that the ion concentrations are constant throughout each cell, and only change in an electrolyte canal, which connects both cells. Under open-circuit conditions, the situation regarding mass transport in the electrolyte is then equivalent to one Pt/electrolyte/Pt cell with large electrode spacing and applied bias voltage of about 7 mv (open-circuit voltage of the DSC). Thus, we obtain the model system shown in Figure 6-4: two platinum electrodes in a distance d (length of electrolyte canal). At the surface of the platinum electrodes (x= and x=d) the ion concentration is that of the respective cells. Furthermore, we will only consider the triiodide ions, since these are the current limiting charge carriers in the process of photoinduced electrophoresis. This is the case in electrolytes, which contain an excess of iodide (.45 M) compared to the concentration of triodide (.5 M). In the model system, the diffusion constant of triiodide inside the cells D cell may be different from the diffusion constant inside the electrolyte canal D. In this way, the electrolyte canal may be considered a membrane with a known diffusive permeability for triiodide. In other words, the electrolyte canal may be filled with a sealing material of varying quality. Figure 6-4: The model system consist of two DSCs (cell 1 and cell 2), which are connected by an electrolyte canal. The platinum electrodes of both cells are combined into one Pt/electrolyte/Pt cell. 6.1.3 Potential gradients in a Pt/electrolyte/Pt cell with applied voltage and electrolyte barrier The following figures illustrate the electrochemical, electrical and chemical energy levels inside a platinum/electrolyte/platinum cell in a qualitative way. In particular, the electrical energy inside the electrolyte is constant. The electrical field inside the electrolyte is screened by an excess of cations and only drops over the Helmholtz double layers. 81

6. Photoinduced electrophoresis in the electrolyte of dye solar modules Equilibrium: In equilibrium, the electrochemical potential is constant. A Helmholtz double layer is formed at the metal/electrolyte interfaces. The gradient in the electrical and chemical potential is not drawn to scale. The width of the Helmholtz double layer is only a few angströms. η: electrochemical potential of electrons ϕ: electrical potential of electrons μ: chemical potential of electrons Applied voltage, initial situation: An applied voltage results in a difference in electrochemical potential. This difference is entirely a difference in electrical potential at the contacts. The drop in electrical potential over the two Helmholtz double layers is increased to about half the applied voltage each by rearrangement of cations (Li + ). Thus, the inside of the electrolyte is electrostatically screened. The chemical potential is not changed significantly immediately after the voltage is applied. Applied voltage, steady state: The high drop of electrical potential over the Helmholtz double layers, induces a diffusion controlled reaction (2e - + I - 3 3 I - ). This results in a high current flow over the Pt/electrolyte interfaces, which changes the chemical potential in the electrolyte. The current drops, until it reaches the diffusion limited current, when the triiodide concentration is zero at one interface and a linear gradient of triiodide concentration is formed inside the electrolyte (c(i - 3 )). A linear concentration gradient results in a logarithmic gradient of the chemical potential. 82

6. Photoinduced electrophoresis in the electrolyte of dye solar modules 6.1.4 Modelling the ion concentration profiles under photoinduced electrophoresis in open-circuit Since mass transport inside the electrolyte mainly occurs via diffusion, the timedependant change of triiodide concentration is determined by Fick s second law (Chapter 2.3.2): n t n 2 I3 I3 = D 2 x 6-2 As initial condition we assume a constant concentration profile with initial concentration n. I 3 n 6-3 ( x, t = ) = n I3 I 3 Now the module is illuminated. This corresponds to stepping the potential of the working electrode from equilibrium to about 7 mv in the model system. This potential results in a diffusion controlled reduction of triiodide at the cathode I 3 + 2 e 3I. Since an excess of iodide is present in the electrolyte only the reverse reaction occurs at the counter electrode. Thus, a concentration gradient forms at x= and x=d, resulting in a current flow over the electrode/electrolyte interfaces. This current flow is very large immediately after switching on the potential and then converges against the diffusion limited current[southampton '85]. The transient concentration gradient at x= and x=d can be approximated as ([Southampton '85], Appendix A2): n I t 2n 4n I I 3 D t 3 3 3 = 1 exp[ ] + exp 2 x τ S d d d x= and x= d 6-4 t Here, the term 1 exp[ ] numerically accounts for stepping the potential of the τ S working electrode from equilibrium to that for diffusion controlled reduction of triiodide. The stepping of the potential occurs very fast (e.g. τ S =.1s ). Equation 6-4 is derived an approximation by a Taylor series (Appendix A2). Initially, this leads to a factor of 24 instead of 3 in Equation 6-4. However, in the numerical calculations it turned out, that the current in Equation 6-4 not drops fast enough with a factor of 24. Negative values for the triiodide concentration occurred, because triiodide was consumed too fast for small times t. Since Equation 6-4 is in itself an approximation, it seems justified and physically reasonable to empirically increase the exponential decay slightly. Thus, the current drops a little bit faster and in the numerical calculation of the concentration profiles all values stay positive. The physical characteristics of the curve 83

6. Photoinduced electrophoresis in the electrolyte of dye solar modules remain the same: an exponentially decaying current, which converges against the diffusion limited current. It is clear from Equation 6-4, that the steady state concentration profile ( t ) is simply 2n / d. This is the concentration gradient of a diffusion limited current. The I 3 current is found by multiplying Equation 6-4, with ν D. e e t = s, 1 s, 1 s, 5 s, 1 s, 2 s Concentration of I - 3 /cm-3 6x1 19 increasing time 5x1 19 4x1 19 3x1 19 2x1 19 1x1 19 increasing time..2.4.6.8 1. 1.2 1.4 x / cm Figure 6-5: Calculated development of the concentration profile during the photoinduced electrophoresis under open-circuit conditions. Equation 6-2 can be solved numerically using the initial condition 6-3 and the two boundary conditions of Equation 6-4. Figure 6-5 shows the calculated evolution of the triiodide concentration profile for the parameter set of the standard electrolyte composition of Table 6-1. 6.1.5 Modelling the regeneration process through diffusion In order to calculate the regeneration of the concentration profile in the dark, Equation 6-2 is solved with different boundary and initial conditions. In the dark, i.e. at equilibrium potential at the working electrode, no current will flow over the electrode/electrolyte interfaces: n x I 3 = x= and x= d 6-5 The initial concentration profile before the regeneration in the dark is a linear profile with the slope of 2n / d. However, we will use an initial concentration profile, which I 3 satisfies the boundary conditions 6-5 and approximates the linear profile for <x<d: 84

6. Photoinduced electrophoresis in the electrolyte of dye solar modules n ( x, t = ) = n I3 i3 π x 1 cos d 6-6 This profile is shown in Figure 6-6 in comparison to the linear profile of 2n I / d. 3 Then, Equation 6-2 can be solved numerically with the initial condition 6-8 and the two boundary conditions of Equation 6-5. 2n I - 3 concentration of I - 3 linear profile cosine profile x d Figure 6-6: Calculated development of the concentration profile during the photoinduced electrophoresis under open-circuit conditions. 6.1.6 Modelling the short-circuit current densities under photoinduced electrophoresis From the thus calculated triiodide concentration profiles ( x, t), we can now determine the triiodide concentration in cell 1 and 2. Within the scope of the model system, the triiodide concentrations in cell 1 and 2 are just the concentrations at the boundaries x= and x=d, respectively: Cell 1 Cell 2 n (, t) = n ( t) n ( d, t) n ( t) I3 I 3 I3 I 3 ni 3 = 6-7 This yields the diffusion limited current density in cell i: j i lim Cell i 2 n ( t) I3 ( t) = ν e e D 6-8 Cell d Cell Here, D Cell is the diffusion coefficient of triiodide in the cells, which may differ from the diffusion coefficient D of triiodide in the electrolyte canal, i.e. the sealing material. And d cell is the thickness of the DSC. Before the photoinduced electrophoresis process, the short-circuit current density j SC of the cells is assumed not to be limited by diffusion of triiodide in the electrolyte. When the triiodide concentration decreases in cell 1 the diffusion limited current density of cell 1 85

6. Photoinduced electrophoresis in the electrolyte of dye solar modules (Equation 6-8) decreases. As long as j ( j the current remains unchanged. But if t) lim j ( ) drops below j, the short-circuit current density will be determined by j ( ). lim t SC SC i i ( jsc für jsc jlim ( t) j SC t) = i ( ) > i 6-9 jlim t für jsc jlim ( t) lim t 6.1.7 Set of parameters The model requires 5 parameters. Table 6-1 summarises the set of parameter, which are used in the simulations for the standard electrolyte composition and the undiluted ionic liquid electrolyte. The diffusion constant of triiodide in acetonitrile is 8.5x1-6 cm²/s. The diffusion constant in hexylmethylimidazolium iodide (HMII) is about 9.x1-8 cm²/s [Papageorgiou '96a]. This value will also be used for the diffusion constant in propylmethylimidazolium iodide (PMII). The diffusion constants of triiodide inside the electrolyte canal is the same as inside the cells, since the electrolyte canal is only filled with electrolyte (no sealing material). Parameter Symbol Standard electrolyte Undiluted ionic liquid electrolyte length of electrolyte canal d 1.45 cm 1.45 cm thickness of cell d Cell 5 μm 5 μm initial concentration of I 3 - n I3 3. 1 19 cm -3 3. 1 2 cm -3 diffusion constant of I 3 - inside electrolyte canal D 8.5 1-6 cm²/s 9. 1-8 cm²/s diffusion constant of I 3 - inside cell D Cell 8.5 1-6 cm²/s 9. 1-8 cm²/s Table 6-1: Set of parameters for the standard electrolyte and the undiluted ionic liquid electrolyte 86

6. Photoinduced electrophoresis in the electrolyte of dye solar modules Figure 6-7: Design of the test cells with integrated electrolyte canal to study the photoinduced electrophoresis under open-circuit conditions. 6.2 Experimental The design of a masterplate with electrolyte canal is shown in Figure 6-7 (silver grid is not shown). The cells with the electrolyte canal are externally interconnected in series. The manufacturing procedure is equivalent to the test cells (masterplates), which is described in the Appendix A1.2. 6.2.1 Apparatus For the photoinduced electrophoresis experiments the masterplates are placed under continuous light soaking immediately after fabrication. Each series connected module is kept at open circuit conditions. During the photoinduced electrophoresis process the short-circuit current densities of the individual cells are measured in regular intervals. For the regeneration process, the masterplates are kept in the dark. Continuous strong light soaking is done with a sulphur plasma lamp. The sulphur lamp spectrum is continuous from 4 to 8 nm with its maximum at 5 nm, approximating quite well the solar spectrum in the spectral sensitivity range of the DSC. The lamp spectrum contains practically no UV light below 4 nm. Silicon reference cells control the lamp intensity. The light intensity was adjusted to approximately 1 W/m² at 35 C. 87

6. Photoinduced electrophoresis in the electrolyte of dye solar modules 6.3 Measurement and simulation of the photoinduced electrophoresis 6.3.1 Photoinduced electrophoresis in a low viscous electrolyte Figure 6-8 shows the measurement and the simulation of the normalised short-circuit current of two cells in series connections under photoinduced electrophoresis in opencircuit for 5 hours. The standard electrolyte composition based on acetonitrile has been used. The initial value of the short-circuit current densities j is indicated with respect to the area of one cell of the module. After about 1 hours cell 1 is completely depleted of triiodide and the short-circuit current has dropped to zero. The simulation of the short-circuit current of cell 1 with the parameters of Table 6-1 matches the measurement sufficiently well. The short-circuit current of cell 2 drops to about 8 % during the first 1 hours and then remains constant. The drop during the first 2 hours is likely to be due to an accumulation of the orange coloured triiodide in cell 1, which absorbs part of the incident light. This effect is not accounted for in the model. Therefore, the simulation of the short-circuit current of cell 2 remains constant over the whole 5 hours. Measurement cell 1 (j =5.3 ma/cm 2 ) Simulation cell 1 Measurement cell 2 (j =4.2 ma/cm 2 ) Simulation cell 2 relative short ciruit current density j/j 1..8.6.4.2. 1 2 3 4 5 time / hours Figure 6-8: Calculated (open symbols) and measured (closed symbols) development of the normalised short-circuit currents in cell 1 and cell 2 for the photoinduced electrophoresis process under open-circuit conditions with the standard, low viscous electrolyte. 88

6. Photoinduced electrophoresis in the electrolyte of dye solar modules 6.3.2 Photoinduced electrophoresis in an undiluted, high viscous ionic liquid Figure 6-9 shows the measurement and the simulation of the normalised short-circuit current of two cells in series connections under photoinduced electrophoresis in opencircuit for more than 2 hours. In this case, the electrolyte composition based on an undiluted, high viscous ionic liquid has been used. The initial value of the short-circuit current densities j is indicated with respect to the area of one cell of the module. The short-circuit currents of both cells drop equally in this case. But the photoinduced electrophoresis is much slower than for the low viscous electrolyte composition. For an illumination time of more than 2 hours the current only drops to about 5 %. The simulation of the current of cell 1 with the parameters for the ionic liquid electrolyte composition from Table 6-1, also drops much slower due to the high initial concentration of triiodide and the lower diffusion coefficient. However, the simulated current of cell 1 still drops about 2 times faster than the measured current. The reason for this might be, that the charge transport mechanism in undiluted, high viscous ionic liquids cannot be described by pure diffusion, as assumed in the model. The measured short-circuit currents of cell 1 and cell 2 decrease equally, in contrast to the simulated currents. It is likely, that for the ionic liquid electrolyte, which has a very high triiodide concentration, the iodide also depletes in cell 2. In the model, the iodide is assumed to be present in excess, and only the triiodide depletion has been considered. Measurement cell 1 (j =1,9 ma/cm 2 ) Simulation cell 1 Measurement cell 2 (j =2, ma/cm 2 ) Simulation cell 2 relative short ciruit current density j/j 1..8.6.4.2. 5 1 15 2 time / hours Figure 6-9: Calculated (open symbols) and measured (closed symbols) development of the normalised short-circuit currents in cell 1 and cell 2 for the photoinduced electrophoresis under open-circuit conditions with the undiluted, high viscous ionic liquid electrolyte. 89

6. Photoinduced electrophoresis in the electrolyte of dye solar modules 6.3.3 Charge transport in highly concentrated ionic liquids In the model, charge transport is described by pure diffusion. This is valid under the assumption, that an undissociated solvent is present in excess in the solution. In particular, the diffusion constant D and the viscosity η are related by the Einstein-Stokes equation: k T D = B 6-1 6π η r Here, r is the radius of the ion. The product of viscosity and diffusion coefficient (the Walden product) is approximately Dη constant and the Einstein-Stokes ratio should be constant for a constant T temperature. The first deviation reported in literature is the well known high mobility of protons in water, where according to Grotthus another transport mechanism has been postulated [Robinson '59]. For undiluted, highly concentrated ionic liquids a similar mechanism has been suggested, as generally higher diffusion coefficients as expected have been observed in binary mixtures of ionic liquids [Papageorgiou '96a,Kawano '4]. In a binary electrolyte of PMII and I 2 non-stokesian behaviour has been repeatedly measured. This means, that the diffusion coefficient is increasing with increasing iodide concentration, despite increasing viscosity of the blend. A possible explanation for this performance is a change in the mechanism of mass- or charge transfer. Apparently, the triiodide transport is not only created by an ordinary diffusion process, but also by a Grotthus-type exchange mechanism increasing with increasing iodide concentration. This mechanism was suggested by Yanagida et al. [Yanagida '5] and by Papageorgiou et al. [Papageorgiou '96a] and can be described as follows: Figure 6-1: Grotthus-type charge transport based on exchange mechanism According to this mechanism, the transport of triiodide is facilitated for high iodide concentration, a fact which corresponds with the experiments [Zistler '6]. 9

6. Photoinduced electrophoresis in the electrolyte of dye solar modules However, as a consequence the presented model of photoinduced electrophoresis is valid for electrolytes with an excess of solvent, but has only limited validity for non- Stokesian ionic liquids. 6.3.4 Regeneration in the dark Photophoresis Regeneration in the dark Concentration of I - 3 / cm-3 3.x1 19 2.5x1 19 2.x1 19 1.5x1 19 1.x1 19 5.x1 18. 5 1 15 2 25 3 35 time / hours Figure 6-11: Calculated development of the triiodide concentration in cell 1 during the photoinduced electrophoresis under open-circuit conditions (closed symbols) and the regeneration in the dark (open symbols) with the standard, low viscous electrolyte. Figure 6-11 shows the simulated concentration of triiodide in cell 1 during the photoinduced electrophoresis and subsequent regeneration in the dark. Regeneration in the dark is much slower than the photoinduced electrophoresis. This implies that in an outdoor situation, the regeneration over night is not sufficient to completely reverse the process of photoinduced electrophoresis. In Figure 6-11 the module is illuminated for about 8 hours and after regeneration in the dark for the following 16 hours (total 24 hours), the separation of the redox couple is only reversed to about 9 %. Figure 6-12 shows the simulation for the regeneration of the triiodide concentration and the short-circuit current. The short-circuit current reaches its initial value after only about 2.5 hours, because the initial triiodide concentration of the electrolyte is much higher than required to conduct a short-circuit current density of 1 ma/cm². Within the scope of the model, regeneration over night should therefore be sufficient to reverse the degradation of the short-circuit current. 91

6. Photoinduced electrophoresis in the electrolyte of dye solar modules short circuit current density j SC / ma cm -2 Simulation: j sc of cell 1 Simulation: concentration of I - in cell 1 3 1 8 3.x1 19 2.5x1 19 6 2.x1 19 1.5x1 19 4 1.x1 19 2 5.x1 18 5 1 15 2 25. 3 time / hours Concentration of I - 3 / cm-3 Figure 6-12: Calculated development of the short-circuit current (closed symbols) and the triiodide concentration (open symbols) in cell 1 during the regeneration in the dark with the standard, low viscous electrolyte. However, the measurements never showed a complete regeneration of the short-circuit current in the dark. Even after a regeneration time of more than 8 hours, only 3 % of the initial value of the short-circuit current was recovered for the standard electrolyte composition. For the electrolyte composition based on undiluted, high viscous ionic liquids, the short-circuit current recovered to about 8 % of the initial value for a regeneration time of more than 8 hours in the dark. In this case, the value for the shortcircuit current was still about 32 % before the regeneration process, because the photoinduced electrophoresis process is very slow in undiluted, high viscous ionic liquids. These measurements imply, that the separation of the redox couple during the photoinduced electrophoresis leads to additional, irreversible degradation mechanisms. These could include a degradation of the dye molecules, due to depletion of iodide. 6.3.5 Regeneration with an externally applied voltage It seems worthwhile to attempt the regeneration of the photoinduced electrophoresis by an externally applied voltage. By applying a external reverse voltage to the module in the dark, the regeneration process might be accelerated. This has been experimentally studied. However, although the regeneration process could be considerably accelerated, no complete regeneration could ever be achieved. The highest regeneration of the current, which was observed, was 4 % of the initial value. Therefore, it can be concluded, that a regeneration with an applied voltage is no solution to the process of photoinduced electrophoresis. On the other hand, it was also experimentally shown, that the photoinduced electrophoresis is not equivalent to a purely electrical electrophoresis, initiated by an 92

6. Photoinduced electrophoresis in the electrolyte of dye solar modules external forward bias. On the first glance, this seems surprising. However, the external voltage is applied between the platinum layer of cell 1 and the TiO 2 layer of cell 2. This does not result in the same energy potentials as in shown Figure 6-2. 6.3.6 Diffusion current In the experimental set-up it is possible to measure the diffusion current. In Figure 6-13 the cross section of two DSCs in series connection with a possibility for ion exchange in the electrolyte is shown. The current which is carried through the electrolyte by diffusion, must also flow in the electrical series connection (I Diff ). This current was calculated in Equation 6-4. _ TCO TiO 2 TCO TiO 2 2 _ I Diff I 3 - platinum + TCO I - platinum TCO Figure 6-13: Cross section of two DSCs in series connection with a possibility for ion exchange in the electrolyte. The current which is carried through the electrolyte by diffusion, must flow in the electrical series connection (I Diff ) The calculated diffusion current is shown in Figure 6-14 for the set of parameters of the standard electrolyte and the undiluted ionic liquid electrolyte (Table 6-1). The steady state diffusion current (for large t) is identical to the diffusion limited current: I lim Cell i 2 n ( t) I3 ( t) = Aν e e D 6-11 Cell d Cell Here, A is the area of the cross section of the electrolyte canal. The measured diffusion current is shown in Figure 6-15 for the two electrolyte compositions. Qualitatively, the measurement and the calculation show the same trends. However, the measured diffusion current is by more than one magnitude larger than the calculated diffusion current. The currents are related to the cross section area of the electrolyte canal A, which is of the dimension.5 cm x 5 μm. The reason of this huge discrepancy is attributed to a difficulty in defining the correct area. In the model system, the charge is transferred over the Pt/electrolyte interface, which is of the area 2.5 cm² and then diffuses through the electrolyte canal, which has the cross section area of.25 cm². 93

6. Photoinduced electrophoresis in the electrolyte of dye solar modules diffusion current density / μa cm -2 35 3 25 2 15 1 5 standard electrolyte undiluted ionic liquid electrolyte 5 1 15 2 25 3 time / hours Figure 6-14: Calculated diffusion current for the set of parameters in Table 6-1. diffusion current density / μa cm -2 12 1 8 6 4 2 standard electrolyte undiluted ionic liquid electrolyte 5 1 15 2 25 time / hours Figure 6-15: Measured diffusion current for the two electrolyte compositions. 6.3.7 Influence of parameters on the theoretical model of the photoinduced electrophoresis In order to illustrate variations of the model parameters on the photoinduced electrophoresis, we use the half-life of the photoinduced electrophoresis. The half-life of the photoinduced electrophoresis is the time in which the short-circuit current of cell 1 drops to 5 % of its initial value. 94

6. Photoinduced electrophoresis in the electrolyte of dye solar modules Figure 6-16A shows the half-life in dependence of the initial concentration of triiodide in a logarithmic scale. As expected, the photoinduced electrophoresis can be slowed down, by adding more iodine (triiodide) to the electrolyte. Figure 6-16B shows the half-life in dependence of the length of the electrolyte canal. The half-life increases with increasing distance between the cells. Figure 6-16C shows the half-life in dependence of the diffusion constant of triiodide inside the electrolyte canal in a double logarithmic scale. As expected, the half-life increases with decreasing diffusion constant. On the basis of Figure 6-16C, the quality of the internal sealing regarding its barrier properties against diffusion of the I - / I 3 - redox couple may be estimated. half life / hours 1 8 6 4 2 A half life / hours 3 25 2 15 1 5 B 5x1 19 1x1 2 2x1 2 3x1 2 initial concentration of I - 3 / cm-3 1 2 3 4 5 length of electrolyte canal d / cm Figure 6-16: half life / hours 1 1 1 C A: The half-life of the photoinduced electrophoresis in dependence of the initial triiodide concentration. B: The half-life of the photoinduced electrophoresis in dependence of the length of the electrolyte canal. 1 1-9 1-8 1-7 1-6 diffusion constant of I - of sealing material / 3 cm2 s -1 C: The half-life of the photoinduced electrophoresis in dependence of the diffusion constant of triiodide of the sealing material. All other parameters are for the low viscous, standard electrolyte. 95

6. Photoinduced electrophoresis in the electrolyte of dye solar modules 6.3.8 Requirements on the barrier properties of the internal sealing material In Figure 6-16C the standard set of parameters has been used for the calculation. In the particular, the length of the electrolyte canal has been 1.45 cm, as in the model system. In a realistic module design, however, the width of the sealing material is much smaller. The DSC modules developed in this work use a sealing of about 2 mm width (see Chapter 7: "Optimization of module ). 1 6 1 5 half life / hours 1 4 1 3 1 2 1 1 1 1-12 1-11 1-1 1-9 1-8 1-7 1-6 diffusion constant of I - in sealing material / 3 cm2 s -1 Figure 6-17: Half life of the short-circuit current under the photoinduced electrophoresis effect versus the diffusion constant of triiodide inside the sealing material. Width of sealing material: 2 mm. In Figure 6-17 the half life of the short-circuit current under photoinduced electrophoresis has been calculated in dependence of the diffusion constant of triiodide in the sealing material, taking into account the width of the sealing material of 2 mm. As can be seen in Figure 6-17, the diffusion constant for triiodide in the sealing material of the interconnect has to be less than D[I 3- ] < 1-11 cm²/s in order to achieve lifetimes higher than 1, hours of illumination (1 years). In this estimation, one year in Europe was assumed to have 1 hours of 1 sun illumination. With organic sealants such a requirement is very hard to realize. In literature the diffusion constant of triiodide in a specific polymer is hard to find. However, the diffusion constant of O 2 and H 2 O can be found for a variety of polymers. 96

6. Photoinduced electrophoresis in the electrolyte of dye solar modules Diffusion constant of Polymer H 2 O [cm²/s] O 2 [cm²/s] PET (Poly Ethylene Terephthalate) 5.6E-9 PP (Poly Propylene) 2.5E-7 HDPE (High Density Poly Ethylene) 1.E-7 1.7E-7 LDPE (Low Density Poly Ethylene) 4.6E-7 PC (Poly Carbonate) 3.E-9 4.2E-8 PS (Poly Styrene) 1.1E-7 PVC (Poly Vinyl Chloride) 1.2E-8 Butyl Rubber (Poly Isobutylene.) 8.1E-8 Table 6-2: Diffusion constants of O 2 and H 2 O for some typical polymers (www.diffusion-polymers.com) Table 6-2 summarizes the diffusion constants of O 2 and (for some polymers) H 2 O in the most common polymers. As can be seen that the calculated requirement of a diffusion constant smaller than 1-11 cm²/s is not achievable with the stated polymers with respect to oxygen or water. A widespread polymer used for the sealing of DSC is Surlyn (poly(ethylene-comethacrylic acid) by DuPont). Although no values of diffusion constants in Surlyn could be found in literature, it can be expected, that the material is comparable to the polymers listed in Table 6-2. In general, barriers are classified as seen in Table 6-3. Only glass is classified as very high barrier. Therefore, glass frit will be used as sealing material in this work. Barrier classification Very high barrier Material Glass High barrier multilayer ultra barrier foils, SiOx coatings Medium barrier PVC, PET Low barrier HDPE, LDPE, PP Table 6-3: Barrier classifications for some materials. 97

6. Photoinduced electrophoresis in the electrolyte of dye solar modules 6.4 Conclusions of Chapter 6 In this chapter the photoinduced electrophoresis has been identified as a potential degradation mechanism as a result of inadequate internal sealing of DSC modules with integrated series connection. A thorough theoretical and experimental study of the photoinduced electrophoresis was presented. In a diffusion model, the decrease of the short-circuit current density of a DSC module with inadequate internal sealing was calculated under illumination (section 6.1). The theoretical model was verified by photoinduced electrophoresis experiments on DSC modules, which were manufactured with a defined electrolyte canal between two cells in series connection (section 6.3). In the dark, the separation of the redox couple should be reversed by diffusion, as predicted by the model. However, in the experiments the degradation of electrical parameters never fully reversed. This implies, that the separation of the redox couple leads to further, irreversible degradation mechanisms, which are unaccounted for in the model. The quality of the internal sealing regarding its barrier properties against diffusion of the I - - / I 3 redox couple was estimated using this model. It was calculated, that the diffusion constant of triiodide in the sealing material must be smaller than D[I 3- ] < 1-11 cm²/s in order to achieve life times higher than 1, hours of illumination (1 years). With organic sealants such a requirement is very hard to realize. Therefore, glass frit will be used as sealing material in this work, which exhibits excellent barrier properties. 98

7. Optimization of module design 7. Optimization of module design The up-scaling of solar cells from small, single cells to large area modules is inevitably accompanied with a loss in efficiency. This efficiency gap is usually several (absolute) percent. The main reason for this loss in efficiency is a loss in active area and the series resistances. In this chapter, a model is developed, which allows the calculation of module efficiency depending on module parameters (section 7.1). This allows an optimization of module performance by calculating the correct layout for the module design. As described in Chapter 3.4, the main task in calculating the I-V curve of a module from the I-V curve of an ideal unit cell is taking into account the distributed series resistance of the contact material (i.e. TCO sheet resistance). In particular, this means that due to the sheet resistance of the TCO, one cell is operated at different points on its I-V characteristic along the direction of current flow. The most significant module parameters are the ratio of active to inactive area, the discrete series resistances, the distributed series resistances and the I-V characteristics of a unit cell. The influence of those parameters on the model is shown in section 7.2. Based on these results an optimal width of active area is determined. The resulting module design, which is made of strip-shaped cells with the determined optimal cell width is discussed in section 7.3. This module design is referred to as strip module. 99

7. Optimization of module design In the course of this research it became evident, that a good DSC module would require a new DSC-specific module design. Such a new design should take into account the electrochemical nature of the DSC. Based on the design of the strip module, an interdigital meander design was developed. This DSC-specific module design is presented in section 7.4. Its main advantage is that the number of filling holes for the electrolyte is reduced drastically. 7.1 Modelling the I-V curve of a dye solar module Starting from an ideal I-V curve of a unit cell, the I-V curve of the module is calculated in this section. Obviously, the efficiency of the module will be lower than the efficiency of the ideal unit cell. The reasons are the following: The distributed series resistance The area loss The discrete series resistance The shunt conductance The loss of incident photon flux, due to absorption of light in the TCO The shunt conductance may be neglected in a DSC module due to the specific layer sequence of the integrated Z-connection and precise screen printing technology (see Chapter 4). Given a flawless manufacturing process, no electrical shunt conductance is incorporated between the cells of a module, Furthermore, the loss due to light absorption of the TCO layer is not taken into account separately. Rather, it is accounted for in the I-V curve of the ideal unit cell. A detailed model was developed by Burgelman et al., which relates the absorption of the TCO (thickness) to its specific sheet resistance and uses this data to calculate module efficiencies [Burgelman '98,Burgelman '99]. 7.1.1 Modelling distributed series resistances As described in Chapter 3.4, a unit cell shall be considered to calculate the influence of the distributed series resistance (Figure 7-1) on the I-V curve of the module. For the architecture of a DSC, the TCO sheet resistance ρ TCO is doubled, because the resistance of both TCO sheets (top and bottom) add up. Then the following 2 equations must be solved (compare Equation 3-5 and 3-7): 1

7. Optimization of module design Figure 7-1: Unit cell with an I-V characteristic j(v(x)) and sheet resistance ρ TCO. x dx v( x) = I( x')2ρ ' TCO 7-1 L x I( x) = L dx' j( v( x')) 7-2 L is the length of one cell, such that I(x) is a current with the dimension ma and x ρ TCO is of dimension Ω. L The I-V characteristic of the unit cell is described by the one-diode model. j( V ) = j Sat ev ers j( V ) exp 1 mdk BT j SC V RS j( V ) + RSh 14243 j Sh 7-3 j sat : saturation current density of the diode in ma/cm² m D : diode factor, dimensionless j SC : photocurrent density (short-circuit current density) in ma/cm² R s : internal series resistance related to area in Ωcm² R sh ; internal shunt resistance related to area in Ωcm² j Sh : shunt current density in ma/cm² Note that Equation 7-3 is a transcendental equation and must be solved numerically. The following algorithm is used to numerically integrated Equations 7-1 and 7-2: At the starting point x=, an arbitrary voltage v()=v is set. V should be in the interval V [ 1mV, 8mV ]. The current in the TCO at x= is obviously I()=. This leads to the starting values: [ 1mV, mv ] v ) = V V 8 7-4 ( ( ) = I 7-5 11

7. Optimization of module design The length w a is divided into e.g. 1 discrete intervals. And Equations 7-1 and 7-2 are integrated numerically from to w a. This yields the first current-voltage pair (I cell, V cell ), where w a is the width of active area: v w ) = V ( V ) 7-6 ( a cell I w ) = I ( V ) 7-7 ( a cell To obtain the whole I-V curve, this procedure is repeated for all V [ 1mV, 8mV ]. 7.1.2 Modelling discrete series resistances The discrete series resistances consist of the TCO sheet resistance, from the end of the active area until the series contact and the contact resistance of the series contact. Thus, the voltage of the module V mod is found: I V ) = I ( V ) 7-8 mod ( cell V rcontact w d ( V ) = Vcell( V ) Icell( V ) + ρ TCO 7-9 L w contact L mod 2 Here, r contact is the contact resistance [Ω cm²] of the series contact and w contact the width of the contact area. And w d is the distance from the end of the active area until the series contact. The factor 2 again results from the need to account for both TCO glasses (top and bottom) 7.1.3 Modelling the efficiency of the module To obtain the efficiency, the maximum power P max is found from the I-V curve of the w + 2 w + w, as seen in Figure module. The width of the total area is given by ( ) 7-2. a d contact Figure 7-2: Dimensions of a unit cell. 12

The efficiency is then determined by: Here, 7. Optimization of module design Pmax η mod = 7-1 L ( w a + 2w d + w contact ) Φ incident Φ incident is the incident irradiation (usually 1 sun=1 mw/cm²). 7.2 Influence of parameter variations on the I-V curve of a dye solar module 7.2.1 Standard set of parameters To study the influence of parameter variations on the I-V curve of a DSC module, a set of standard parameters is needed. The model requires the internal parameters of the ideal unit cell and the external parameters of the module. Internal parameters The internal parameters determine j(v) of the ideal unit cell, i.e. a cell without TCO sheet resistance. The ideal unit cell is not accessible experimentally. However, the model of distributed series resistance (section 7.1.1) allows the calculation of real unit cells with TCO sheet resistance from ideal unit cells. The I-V curves of real unit cells are known from publications and own experiments. We will define the internal electrical parameters of the ideal unit cell (V oc, I sc, FF, η) in such a way, that the calculated I-V curve of the real unit cell (I cell, V cell ) resembles an experimentally measured solar cell. In the following, real unit cells are assumed to have a width of.5 cm, as is the case at Fraunhofer ISE. Furthermore, real unit cells have a low series resistance of about 1 Ωcm² (in the one-diode model). A high series resistance would distort the I-V curve in an unrealistic way upon variation of light intensity. Since the electrical parameters of a DSC can vary depending on materials and production process, 3 classes of DSCs are used in the following. The world-class DSC resembles the current world record cell by Sharp (A=1 cm²) [Chiba '5]. The standard DSC has electrical parameters, which can be obtained by using excellent materials on an area of about 2.5 cm² [Kroon '1]. The mediocre DSC is obtained by using medium-quality materials on an area of about 2.5 cm². These classifications are arbitrary. But they will account for the fact, that different electrical parameters of a cell, might require different optimal width of active area. 13

7. Optimization of module design Internal electrical parameters of the ideal unit cell j sc [ma/cm²] V oc [mv] FF [%] η [%] world-class DSC 21.9 723 7 11.1 standard DSC 15.8 73 74.8 8.6 mediocre DSC 8.1 694 7.1 3.9 Resulting electrical parameters of a (measured) real unit cell of width.5 cm world-class DSC 21.9 723 65.6 1.4 standard DSC 15.8 73 71.5 8.2 mediocre DSC 8.1 694 68.6 3.8 Table 7-1: Set of standard internal electrical parameter of the ideal unit cell. The electrical parameters of the real unit cell result from taking into account the distributed series resistance of the TCO. The real unit cells correspond to experimental cells of.5 cm width. 3 classes of DSCs are distinguished. The set of electrical parameters for the 3 classes of DSCs are shown in Table 7-1. It is noticeable, that for the world-class DSC, the loss in efficiency due to.5 cm of TCO is already.7 % (absolute). External parameters The external parameters of the module are shown in Table 7-2. The dimensions of the integrated series resistance (w d and w contact ) is imposed by technological constraints. Here, it is determined by the glass frit sealing technology. The total width of photoinactive area is 2.5 mm for all modules in this work. Clearly, w d and w contact is a very important parameter in the evaluation of the optimal width of active area. In principle, ( 2w d + w contact ) could be decreased further especially by using more precise glass frit screen printing technology. The contact resistance of the series resistance r contact has been measured in specifically designed devices, in which r contact was directly accessible (compare Chapter 8.1.4). The value for ρ TCO is taken from the technical data sheet of the TCO glass (8 Ω/, Pilkington TEC glass). This glass has an optical transmission of 77 %, as of manufacture s specification. However, this value is measured with an interface of both glass surfaces with air. In the DSC only one glass/air interface occurs, and one glass/electrolyte interface. The glass/electrolyte interface has a much lower refraction index. Thus, the optical transmisson of the TEC 8 glass in a DSC is about 81 %. Alternatively, a TEC 15 glass with sheet resistance of 15 Ω/ may be considered. The higher sheet resistance might be tolerated in trade-off for higher optical transmisson. TEC 15 glass has an optical transmission of 83 %, as of manufacture s specification. With the same argument as above, the optical transmission in a DSC is then about 87 %. 14

7. Optimization of module design External parameters of the module Parameter Description Value w a width of active area 7 mm w d distance from end of active area to series contact 1 mm w contact with of contact area.5 mm (2 w d + w contact ) width of inactive area 2.5 mm ρ TCO specific sheet resistance of TCO 8 Ω/ r contact contact resistance of series connection.2 Ω cm² Table 7-2: Set of standard external parameters of the module. The standard value for the width of active area w a is set to 7 mm. Obviously, w a is the parameter of greatest interest and will be determined by variation. For this standard set of external module parameters and the internal parameters for the standard DSC, the resulting I-V curve of the module is shown in Figure 7-3-A. The efficiency of the ideal unit cell (8.6 %) is reduced to 5.6 %. The values of V oc and j sc are almost unchanged (related to active area). The loss in fill factor (71.3 % to 66.6 %) is attributed to the distributed and discrete series resistances. The main loss, however, is due to a loss in photoactive area. The contribution to the total current of each ideal unit cell (j(x)) along the x-axis is shown in Figure 7-3-B under short-circuit condition, maximum power-point and opencircuit condition. At short- and open-circuit condition, all unit cells deliver or full current, respectively. At maximum power-point, the unit cells further away from the contact deliver less current, since those cells are not operated at maximum power-point, but rather at a point closer to open-circuit. 15

7. Optimization of module design current density / ma cm -2 related to total area 12 1 8 6 4 2 V oc = 729 mv j sc (total area) = 11.6 ma/cm 2 j sc (active area) = 15.8 ma/cm 2 FF = 66.3 % eta total area = 5.6 % eta active area =7.6 % 1 2 3 4 5 6 7 8 voltage / mv A currrent density of unit cell / ma cm -2 16 14 12 1 8 6 4 2 short-circuit maximum power-point open-circuit..1.2.3.4.5.6.7 x / cm B current density in TCO / ma cm -2 16 14 12 1 8 6 4 2 short-circuit maximum power-point open-circuit..1.2.3.4.5.6.7 C current density related to active area voltage in the TCO / mv 8 7 6 5 4 3 2 1 short-circuit maximum power-point open-circuit..1.2.3.4.5.6.7 D x / cm x / cm Figure 7-3: For the set of standard external parameter and the internal parameters of the standard DSC: A. Calculated I-V curve of the module B. Calculated current density of the unit cells along the x-axis C. Calculated current density in the TCO D. Calculated voltage distribution in the TCO The current density in the TCO (I(x)) is shown in Figure 7-3-C. Obviously, the current density remains under open-circuit condition. At maximum power-point and opencircuit, the current rises steadily, as all unit cells contribute to the current. The voltage distribution in the TCO (v(x)) is shown in Figure 7-3-D. Under short-circuit condition the voltage at the contact is, but at x=, the voltage is still about 7 mv. A similar situation occur at maximum power-point, here the additional voltage produced by the cells further away from the contact is lost for the module, reducing the fill factor. 7.2.2 Variation of the width of photoactive area The width of the active area w a in a solar module is the most important parameter in optimizing module performance. In contrast to all other parameters, the width of active area is easy to modify. 16

7. Optimization of module design For a very large width of the active area, the area loss is small, but the loss due to distributed series resistances is large. On the other hand, for a very small width of active area, the area loss is high, but the ohmic losses are small. Obviously, an optimal width of active area exists. Efficiency / % world-class DSC standard DSC mediocre DSC 7. 6.5 6. 6-7 mm 5.5 5. 7-8 mm 4.5 4. 3.5 3. 2.5 2. 1-11 mm 1.5 1..5...5 1. 1.5 2. Width of active area w a / cm Figure 7-4: Calculated module efficiency versus width of active area for the set of standard parameters. Sheet resistance of the TCO is assumed to be ρ TCO =8 Ω/. Optimal cell width is shown for the 3 classes of DSCs. Figure 7-4 shows the calculated module efficiency versus width of active area w a for the set of standard parameters. The sheet resistance of the TCO is assumed to be ρ TCO =8 Ω/. The optimal cell width is shown for the 3 classes of DSCs. The efficiency of the module with optimal cell width is 6.9 %, 5.6 % and 2.8 % for the 3 classes of DSC respectively. From Figure 7-4 it is clear, that the efficiency of the module is more sensitive to a variation of w a, when the current density of the unit cell is higher. 7.2.3 Variation of the width of photoinactive area The width of photoinactive area is imposed by the glass frit sealing technology. In principle, it is possible to reduce the width of inactive area. However, a certain width is required for a hermetical seal. In this work, the width of photoinactive area 2 w + is fixed to 2.5 mm. ( ) d w contact Although it is not an easy task to reduce the width of inactive area, it is well worth the effort. For the module designer, the width of inactive area is the main reason for efficiency loss. In Figure 7-5 the distance from active area to series contact w d is varied. The total 2 w +, w contact remains constant. In Figure 7-5 the width of inactive area is ( ) d w contact active area w a is varied for each w d. The efficiency of the module is shown at optimal width of active area. 17

7. Optimization of module design Efficiency / % 9. 8.5 8. 7.5 7. World-class DSC 6.5..2.4.6.8 1..7.6.5 optimal width of active area / cm Efficiency / % 7.5 7. 6.5 6. Standard DSC 5.5..2.4.6.8 1..8.7.6 optimal width of active area / cm distance from end of active area to series interconnect: w d / mm distance from end of active area to series interconnect: w d / mm Efficiency / % 3.4 3.2 3. Mediocre DSC 2.8..2.4.6.8 1. 1..9.8.7 optimal width of active area / cm Figure 7-5: Calculated module efficiency versus width of inactive area w d for the set of standard parameters. For each w d the optimal cell width is calculated and module efficiency is shown for the optimal cell width. distance from end of active area to series interconnect: w d / mm The steps in optimal cell width in Figure 7-5 result from the increment in the numerical calculation (1 mm). The harsh effect of w d on module efficiency can clearly be seen in Figure 7-5. E.g a reduction of w d from 1 mm to.5 mm results in a gain in efficiency of about 1 % (relative). This corresponds to a reduction of the width of inactive area ( 2 w d + w contact ) from 2.5 mm to 1.5 mm, which seems realistic with screen printed glass frit. 7.2.4 Variation of TCO sheet resistance The TCO sheet resistance is a parameter which is accessible to the module designer within certain restrictions. Two common, commercially available TCO (SnO 2 :F) glasses have sheet resistances of ρ TCO =8 Ω/ and ρ TCO =15 Ω/, respectively. As described in section 7.2.1, a higher sheet resistance might be tolerated in trade off for higher optical transmission. Resulting electrical parameters of a real unit cell of width.5 cm with ρ TCO = 15 Ω/ world-class DSC 23.5 727 61.4 1.5 standard DSC 17 732 68.2 8.5 mediocre DSC 8.7 698 67.5 4.1 Table 7-3: Calculated set of electrical parameters of the real unit cells with TCO sheet resistance of ρ TCO =15 Ω/. 18

7. Optimization of module design Efficiency / % world-class DSC standard DSC mediocre DSC 7. 6.5 6. 5-6 mm 5.5 5. 4.5 5-6 mm 4. 3.5 3. 2.5 2. 7-8 mm 1.5 1..5...5 1. 1.5 2. Width of active area w a / cm Figure 7-6: Calculated module efficiency versus width of active area for the set of standard parameters. Sheet resistance of the TCO is ρ TCO =8 Ω/ (closed symbols) and ρ TCO =15 Ω/ with higher optical transmission (open symbols). Optimal cell width is shown for ρ TCO =15 Ω/. Table 7-3 shows the calculated set of electrical parameters of the real unit cells with TCO sheet resistance of ρ TCO =15 Ω/. Compared to the electrical parameters of the real unit cells with TCO sheet resistance of ρ TCO =8 Ω/ (Table 7-1), the efficiencies are slightly higher, due to a higher current density resulting from the higher optical transmission. Figure 7-6 shows the calculated module efficiency versus the width of active area for the case of ρ TCO =8 Ω/ and ρ TCO =15 Ω/. It can be seen, that in a module the use of a higher transparent and higher resistive TCO is unfavourable. Although the unit cell is more efficient, the high resistivity leads to a lower optimal cell width. This results in a higher ratio of photoinactive to photoactive area. It should be mentioned that TCO layers are under investigation, which possess significantly lower sheet resistance (<2 Ω/) [Kumara '6]. However, even a sheet resistance of only 1 Ω/ would still result in an optimal width of active layer of only 15 mm for the world-class DSC and 17 mm for the standard DSC (not taking into account additional absorption losses). 7.2.5 Variation of contact resistance The discrete series resistance of the interconnect, i.e. the contact resistance is also a very important parameter affecting module performance. Obviously, the contact resistance should be as low as possible. 19

7. Optimization of module design Efficiency / % 7. 6.5 6. 5.5 5. 4.5 4. 3.5 3. 2.5 2. 1.5 1..5. 1E-3.1.1 1 1 Contact resistance / Ω cm 2 world-class DSC w a =6 mm standard DSC w a =7 mm mediocre DSC w a =1 mm Figure 7-7: Calculated module efficiency versus contact resistance of the series interconnect. For the different classes of DSC, different (optimal) width of active area is used in the calculation. Figure 7-7 shows the calculated module efficiency versus contact resistance of the series interconnect for the set of standard parameters. For the different classes of DSC, different (optimal) width of active area is used in the calculation. The efficiency is very sensitive to a change of contact resistance in the region of.1 Ω cm²-5 Ω cm². The contact resistance should definitely be lower than.1 Ω cm², and preferably lower than.5 Ω cm². The contact resistances measured in this work, are in the magnitude of.2 Ω cm². In the region of Ω cm²-.5 Ω cm², the efficiency is nearly constant. A poor contact resistance can be compensated by increasing the width of the interconnect, but at the expense of active area. 7.2.6 Variation of illumination intensity The standard reporting conditions (SRC) for the electrical parameters of a solar cell are under AM 1.5, corresponding to an illumination intensity of 1 W/m². The rated power output and efficiency of a solar module are also related to this illumination intensity of 1 sun. In section 7.2.2, the optimal cell width was determined, assuming an illumination intensity of 1 W/m². This is reasonable, since the module efficiency will be reported under SRC. Under real outdoor conditions, however, the incident irradiation is mostly below 1 sun. A lower illumination intensity leads to lower current densities, which again lead to a higher optimal width of active area. Thus, the lower ratio of photo-inactive to photoactive area may result in a higher module efficiency at lower light intensities [Randall '3]. In contrast, the unit cell will always increase in efficiency upon an increase of illumination intensity, due to a proportional rise in current and logarithmic rise in voltage. 11

7. Optimization of module design Efficiency / % 7.2 7. 6.8 6.6 6.4 6.2 World-class DSC 6...2.4.6.8 1. Illumination intensity / 1 W/m 2 2.4 2.2 2. 1.8 1.6 1.4 1.2 1..8.6.4 optimal width of active area / cm Figure 7-8: Calculated module efficiency versus illumination intensity for the parameters of a world-class DSC. At each illumination intensity, the optimal width of active area determined. ρ TCO =8 Ω/ (closed symbols) and ρ TCO =15 Ω/ (open symbols Efficiency / % 6. 5.8 5.6 5.4 5.2 Standard DSC 5...2.4.6.8 1. Illumination intensity / 1 W/m 2 2.4 2.2 2. 1.8 1.6 1.4 1.2 1..8.6.4 optimal width of active area / cm Figure 7-9: Calculated module efficiency versus illumination intensity for the parameters of a standard DSC. At each illumination intensity, the optimal width of active area determined. ρ TCO =8 Ω/ (closed symbols) and ρ TCO =15 Ω/ (open symbols Figure 7-8 illustrates the behaviour of the module efficiency upon variation of illumination intensity for the parameters of a world-class DSC. In this calculation, the optimal width of active area is re-evaluated at each illumination intensity. The unevenness of the curve results from the increments in the numerical calculation. The calculation is carried out for ρ TCO =8 Ω/ (closed symbols) and ρ TCO =15 Ω/ (open symbols), taking into account the higher optical transmission for ρ TCO =15 Ω/. The highest achievable efficiency for the parameters of the world-class DSC is 7.2 % at an illumination intensity of 4 W/m² in a module with a cell width of 1. cm² and TCO sheet resistance of ρ TCO =8 Ω/. 111

7. Optimization of module design Efficiency / % Mediocre DSC 3. 3. 2.8 2.8 2.6 2.6 2.4 2.4 2.2 2.2 2. 2. 1.8 1.6 1.8 1.4 1.6 1.2 1.4 1. 1.2.8.6 1...5 1. 1.5 2. Illumination intensity / 1 W/m 2 optimal width of active area / cm Figure 7-1: Calculated module efficiency versus illumination intensity for the parameters of a mediocre DSC. At each illumination intensity, the optimal width of active area determined. ρ TCO =8 Ω/ (closed symbols) and ρ TCO =15 Ω/ (open symbols Figure 7-9 shows the equivalent calculation for the parameters of the standard DSC. Here the highest achievable efficiency is 6. % at 2 W/m², w a =1.1 cm² and TCO sheet resistance of ρ TCO =15 Ω/. In this case, it is even advantageous to use the TEC 15 TCO glass, although only at very low illumination intensities. For the TEC 8 glass the highest achievable efficiency is 5.9 % at 3 W/m², w a =1.2 cm. Figure 7-1 shows the equivalent calculations for the parameters of the mediocre DSC. Here the highest achievable efficiencies is 2.8 % at 12 W/m² and w a =.9 cm² and TCO sheet resistance of ρ TCO =8 Ω/. Since the photocurrent of the mediocre DSC is low, the optimal illumination intensity is higher than for the other two classes of DSC. However, the efficiency of the mediocre DSC module is again much less sensitive to parameter variation. From these calculations it follows, that for real outdoor operation, when the illumination intensity is mostly below 1 W/m², a higher power output can be achieved, when the module is optimized for correct illumination intensity [Maine '5]. Furthermore, it is important to know the parameters of the unit cell (the class of DSC). For high efficient cells correct optimization is much more critical than for the mediocre DSC. Since the illumination intensity varies under outdoor operation, it is valuable to know the dependence of the module efficiency upon illumination intensity for a fixed width of active area w a. Figure 7-11 shows the calculated module efficiency for the parameters of the worldclass DSC in dependence of the illumination intensity. Here, two different fixed widths of active area are used. For illumination intensities below.7 sun, w a =1. cm yields higher efficiencies than w a =.6 cm as optimized for 1 sun operation. 112

7. Optimization of module design Efficiency / % w a =.6 cm optimized at 1 sun w a =1. cm optimized at.4 sun 7.2 7. 6.8 6.6 6.4 6.2 6. 5.8 5.6 5.4 5.2 World-class DSC 5...2.4.6.8 1. 1.2 1.4 1.6 1.8 2. Illumination intensity / 1 W/m 2 Figure 7-11: Calculated module efficiency versus illumination intensity for the parameters of a world-class DSC. The optimal width of active area is fixed to the stated values. ρ TCO =8 Ω/. Efficiency / % w a =.7 cm optimized at 1 sun w a =1.2 cm optimized at.3 sun 6. 5.8 5.6 5.4 5.2 5. 4.8 4.6 4.4 4.2 Standard DSC 4...2.4.6.8 1. 1.2 1.4 1.6 1.8 2. Illumination intensity / 1 W/m 2 Figure 7-12: Calculated module efficiency versus illumination intensity for the parameters of a standard DSC. The optimal width of active area is fixed to the stated values. ρ TCO =8 Ω/. Figure 7-12 shows the calculated module efficiency for the parameters of the standard DSC in dependence of the illumination intensity. Again, two different fixed widths of active area are used. For illumination intensities below.6 sun, w a =1.2 cm yields higher efficiencies than w a =.7 cm as optimized for 1 sun operation. Figure 7-13 shows the calculated module efficiency for the parameters of the mediocre DSC in dependence of the illumination intensity with three different fixed widths of active area. Because of the low current densities of the mediocre DSC, optimization is far less critical. 113

7. Optimization of module design Efficiency / % w a =1. cm optimized at 1 sun w a =.9 cm optimized at 1.2 sun w 3. a =1.6 cm optimized at.3 sun 2.8 2.6 2.4 2.2 2. 1.8 1.6 1.4 1.2 Mediocre DSC 1...5 1. 1.5 2. Illumination intensity / 1 W/m 2 Figure 7-13: Calculated module efficiency versus illumination intensity for the parameters of a mediocre DSC. The optimal width of active area is fixed to the stated values. ρ TCO =8 Ω/. 7.3 Module design I: Strip module In this chapter the resulting design for a DSC module is presented. In this module design, the cells are strip-shaped and aligned next to each other transverse to current flow. The resulting design (strip module) is well-known from other thin film solar technologies with integrated series connection (see Chapter 3.1.1). The integrated series interconnect is of Z-type, as illustrated in Chapter 4. As protecting material and hermetic sealant an encapsulant based on glass frit is used (compare Chapter 3.5.1). Optimisation of the width of active area has been demonstrated for different sets of parameters in section 7.2. The module in this work, is optimized for the internal parameters of the standard DSC (see section 7.2) and illumination conditions of AM 1.5, i.e. 1 W/m². The external module parameters are consequently those listed in Table 7-2. In particular, this results in a width of active area of 7 mm. 7.3.1 Electrolyte canal Each cell in a series connection of DSCs must be a closed compartment, i.e. mass transport of the electrolyte between neighbouring cells must be prevented. Otherwise photoinduced electrophoresis would occur. Consequently, each cell compartment needs 2 openings for colouration and electrolyte filling. (Compare Chapter 3.3.1 and Chapter 6) 114

7. Optimization of module design To avoid the drilling and sealing of a very large number of holes in the glass substrate, an electrolyte canal is integrated in the module to distribute the dye and electrolyte to all cell compartments. The module is coloured and filled with electrolyte via 2 holes which are located in the upper left corner and in the lower right corner of the electrolyte canals (Figure 7-14). The electrolyte canals are broken off after the filling process and the small openings on the module edges are sealed. The sealing of the edges is also done with glass frit, which is locally applied and melted onto the cell openings. This procedure, however, is not yet technically mature and fully developed as a hermetic seal is not achieved reproducibly. Other approaches, in particular inorganic glues (cements) have also shown promise according to first helium leak tests. Alternatively, the edges with electrolyte canal may remain on the module and the 2 holes are sealed with Surlyn polymer hot-melt foil (Dupont) and a small glass plate. Such modules with an electrolyte canal can be electrically characterised under illumination since the photoinduced electrophoresis over the electrolyte canal is a very slow process. But long term testing cannot be conducted on modules with the electrolyte canal intact. The resulting module design is shown schematically in Figure 7-14. The module consists of 29 cells connected in series. Between two neighbouring cells an electrical interconnect of Z-type is screen printed (see Chapter 4) as shown in the enlargement in Figure 7-14. The electrolyte canals are located on the left and right side of the modules in Figure 7-14. The glass substrate is 3 x 3 cm². The active area is 55.5 cm², which is about 74 % of the total area (683 cm² without electrolyte canal and external electrical contacts of 217 cm²). filling hole Electrolyte canal filling hole Figure 7-14: Schematic of the module design. 29 cells connected in series. The glass substrate is 3 x 3 cm² The active area is 55.5 cm², about 74 % of the total area (without electrolyte canal). Labels in mm. 115

7. Optimization of module design 7.3.2 Experimental The manufacturing process of the module is illustrated in detail in the Appendix A1.1. The I-V characteristics of a module is shown in Figure 7-15 as measured immediately following fabrication. The electrolyte of the module consisted of.6 M hexylmethylimidazolium iodide;.1 M LiI;.5 M I 2 ;.5 M tert-butylpyridine in acetonitrile. The measurement was performed under a sulphur lamp at AM1.5 global (1 W/m²) taking into account spectral mismatch at a temperature of 38 C. Due to the high capacitance of the DSC module, the scanning rate of the I-V curve measurement must be very slow, at least.5 seconds per measuring point. The values of the open-circuit voltage V OC and the short-circuit current density j SC correspond to the values of a 2.5 cm² DSC. Typical values for a 2.5 cm² DSC manufactured with the same materials are an open-circuit voltage V OC of 7 mv, a shortcircuit current density j SC of 11 ma/cm² and a fill factor of 65 %. The fill factor is much lower since low melting, lead containing glass frit has still been used in this case resulting in a large series resistance for the Z-contact of about.2 Ω cm². Moreover, the current drop in the lower voltage (-12 V) range indicates a photo shunt, which might be related to inadequate laser scribing of the TCO. The resulting efficiency is 3.5 % on active area, which corresponds to about 2.6 % on total area (without electrolyte canal and external contacts). Current / ma 18 16 Voc=2. V 14 Voc (one cell) = 689.7 mv 12 Isc=168.6 ma 1 8 jsc (one cell) = 9.7 ma/cm 2 6 FF =.53 4 eta 2 act. area = 3.5 % 5 1 15 2 Voltage / V Figure 7-15: I-V characteristics of a 3 x 3 cm² module. The measurement was performed under a sulphur lamp at AM1.5 global (1 W/m²) taking into account spectral mismatch at a temperature of 38 C. 116

7. Optimization of module design Figure 7-16: 3 x 3 cm² DSC module sealed with glass frit. Figure 7-17: Transparent 3 x 3 cm² DSC module. Figure 7-16 shows a picture of an opaque module. The electrolyte canal is still present under the black tape at the left and right side of the module. Transparent modules were built by just omitting the ZrO 2 layer as seen in Figure 7-17. 7.4 Module design II: Interdigital meander module In this Chapter, a new interdigital design for large area dye solar modules is developed for an area of 3 x 3 cm². This DSC-specific design requires fewer holes in the glass substrate for electrolyte filling, than the conventional strip design. In a strip pattern the optimal cell width for DSC is 7-8 mm, as discussed in section 7.2.2. This design results in a large number of cell compartments and consequently a large number of filling holes. One solution to reduce the number of filling holes is to incorporate an electrolyte canal, which connects all cells and which is broken off after electrolyte filling (section 7.3.1). However, the sealing of the resulting openings on the edges of the module is a technological problem. In the course of this research it became clear, that a sealing technology for the edges is harder to develop than a sealing technology for holes on the glass surface. Furthermore, an electrolyte canal such as in section 7.3.1 does not actually reduce the number of openings, but only transfers the openings from the glass surface to the glass edges. The new interdigital meander design presents a way to actually reduce the number of filling holes on the glass surface, without transferring the openings to the modules edges. To reduce the number of filling holes in the glass substrate, the width of one cell in the series connection has to be increased. By screen printing a silver grid on front and counter electrode, the effective sheet resistance of the TCO substrate can be reduced and the cell width may be increased. 117

7. Optimization of module design Figure 7-18 shows the front and counter electrode of 2 cell compartments. The silver grid of the front electrode is shifted with respect to the silver grid of the counter electrode. This interdigital silver grid results in a meander shaped cell area, allowing an easy coloration and filling process via 2 holes per cell. When front and counter electrode are aligned on top of each other (Figure 7-19), the silver grids of front and counter electrode only overlap on the boundary of two neighbouring cells, thus forming an electrical series connection. The TCO layer of the substrate is structured accordingly. The silver grid is protected by screen printed glass frit on each side (not shown in Figure 7-18 and Figure 7-19). Figure 7-18: Front and counter electrode of 2 cell compartments of the interdigital design. Figure 7-19: When front and counter electrode are aligned on top of each other the silver grids form an integrated series connection 118

7. Optimization of module design Figure 7-2: Silver grid on front and counter electrode of a 3 x 3 cm² DSC module of meander design. An interdigital silver grid results in a meander shaped cell area. Figure 7-21:The silver grids form an electrical series connection, when positioned on top of each other. The module consists of 6 cells, each with a width of about 5 cm. Such an unit cell can be scaled up in x-direction in order to increase the current and in y-direction to increase the voltage of the module as desired. Figure 7-2 shows front and counter electrode of a 3 x 3 cm² module with meander shaped cells. As seen in Figure 7-21, the module consists of 6 cells. Each cell has a width of about 5 cm. Figure 7-22 shows a section of the interdigital silver grid in top view. The width of the dead area D is determined by the glass frit screen-printing technique (here: D=2.5 mm). The width of the gap G should be minimal for optimal electrical performance. However, a certain gap width G is required for a reasonable flow rate of the dye solution. Here, G is about 8 mm. Optimisation of the meander width W is done according to section 7.2. In the model, the cell width of a strip module pattern is optimised. Here, the optimal meander width W is approximated by assuming a gap G of zero. Thus, an optimal meander width W of 7-8 mm is determined. 119

7. Optimization of module design The specific resistance r of a screen-printed silver finger of width.5 mm and height of 1 μm was assumed to be.8 Ω/cm. One silver finger has to collect the current from both sides (2 W), thus the voltage drop V loss over a length L will be: V loss L = r j 2 W x dx 7-11 Assuming a current density j of 15 ma/cm² and tolerating a voltage drop of about 3 mv, the length of the silver finger L may not be greater than 5.6 cm. Therefore, (L+G) was chosen to be about 5 cm, as to fit 6 cells on a 3 x 3 cm² module. Using these values, the effective sheet resistance of the TCO is reduced from 8 Ω/square to about.4 Ω/square as measured on TCO glass with such a screen printed silver grid. Figure 7-22: Section of the interdigital silver grid in top view. Optimal values for W and L are calculated. The resulting module design has an active area of 588 cm² (7 %), consisting of 6 cells of 98 cm² each. The total area is 846 cm² excluding external electrical contacts. 7.4.1 Experimental The manufacturing process of the module is illustrated in detail in the Appendix A1.1. The I-V characteristics of a transparent 3 x 3 cm² module with interdigital meander design is shown in Figure 7-23 as measured immediately following fabrication. The electrolyte of the module consisted of.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile. This electrolyte is highly volatile and thus has not been used in the stability tests (Chapter 8.2). However, this electrolyte is used as standard electrolyte to optimise the performance of the 3 x 3 cm² module. The measurement was performed under a sulphur lamp at AM 1.5 global (1 W/m²) taking into account spectral mismatch at a temperature of 38 C. Due to the high capacitance of the DSC module, the scanning rate of the I-V curve measurement must be very slow, at least.5 seconds per measuring point. 12

7. Optimization of module design The values of the open-circuit voltage V OC and the fill factor correspond approximately to the values of a 2.5 cm² DSC. Typical values for a 2.5 cm² DSC manufactured with the same materials are an open-circuit voltage V OC of 7 mv, a short-circuit current density j SC of 11 ma/cm² and a fill factor of 65 %. The short-circuit current density j SC is rather low since a very thin and transparent TiO 2 has been used. The somewhat low fill factor might be related to a slight difference in the short-circuit current densities of the individual cells. A mismatch in the j SC s of the individual cells does not result in a lowering of the module s j SC compared to the cell with the highest j SC, but rather in a lowering of the fill factor (compare Chapter 5). The resulting efficiency is 3.1 % on active area, which corresponds to 2.1 % on total area. Figure 7-24 shows a picture of a transparent meander module. The high level of transparency is illustrated in front of the cherry tree in Figure 7-25. Current / ma 6 Voc=4.5 V Voc (one cell)= 75 mv 4 Isc=655 ma jsc (one cell)=6.7 ma/cm 2 2 FF=.61 eta act area =3.1 % 1 2 3 4 Voltage / V Figure 7-23: I-V characteristics of a transparent 3 x 3 cm² meander shaped module sealed with glass frit. Figure 7-24: 3 x 3 cm² meander shaped module sealed with glass frit. Figure 7-25: Transparent 3 x 3 cm² DSC meander module sealed with glass frit. 121

7. Optimization of module design The I-V characteristics of an opaque (with scattering layer) 3 x 3 cm² module with interdigital meander design is shown in Figure 7-26 as measured immediately following fabrication. The electrolyte of the module consisted of.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile. The resulting efficiency is 3.8 % on active area, which corresponds to 2.9 % on total area. -.8 current / A -.6 -.4 -.2. Voc = 4,4 V voc (one cell)= 733 mv Isc = 758,52 ma jsc (one cell) = 7,74 ma/cm 2 FF =,68 eta act. area = 3,8 %.2-2 -1 1 2 3 4 voltage / V Figure 7-26: I-V characteristics of an opaque (with scattering layer) 3 x 3 cm² meander shaped module sealed with glass frit. 7.5 Conclusions of chapter 7 In this chapter, a model was developed, which allows the calculation of module efficiency depending on module parameters (section 7.1). This allows an optimization of module performance by calculating the correct layout for the module design.. Based on these results an optimal width of active area of 7-8 mm was determined for the modules developed in this work. The resulting module design, which is made of strip-shaped cells with the determined optimal cell width was discussed and realized in section 7.3. Efficiencies of about 3.5 % on active area were achieved, resulting in an efficiency on total area of about 2.6 %. In the course of this research it became evident, that a good DSC module would require a new DSC-specific module design. Thus, one of the main achievements of this work was to develop and realize a new interdigital meander design. This DSC-specific module design was presented in section 7.4. Its main advantage is that the number of filling holes for the electrolyte is reduced drastically. With this design efficiencies of 3.8 % on active area were achieved, resulting in an efficiency on total area of about 2.9 %. 122

8. Long term stability 8. Long term stability For all solar cell technologies long term stability is a basic requirement. Especially for building integrated applications, long term stability must be warranted for at least 2 years. The product certification for solar modules is based on international standards from the series IEC 68 Environmental Test Procedures. A different standard exists for crystalline silicon solar modules 4 and thin film solar modules 5. The IEC qualification testing includes a number of tests, from hot-spot endurance over mechanical load test to hail endurance. Central tests are the heat test under 85 C in the dark, continuous strong light soaking and UV stability. After each test, the degradation of maximum power at standard testing conditions should not exceed 5 %. As the DSC technology progresses from laboratory scale to large area applications, long term stability is one major obstacle. For small test cells (A 1.-2.5 cm²) good progress was made in the field of long term stability of DSC particularly by the use of electrolytes based on ionic liquids [Hinsch '1,Kroon '1]. Recently, Wang et al. reported a stable 8 % efficient DSC based on a low volatile electrolyte under thermal stress of 8 C for 1 hours [Wang '4,Wang '5]. Especially for large area DSC modules, stability is related to hermetic sealing. The sealing material must be mechanically and thermally stable and be chemically inert against 4 IEC 61215 5 IEC 61646 123

8. Long term stability the I - / I 3- redox couple. The sealing material must protect the conductor grid and prevent mass transport between the electrolytes of neighbouring cells. In section 8.1 an encapsulant based on glass frit is presented, similar to that used for the sealing of plasma display panels. Properties of the glass frit, such as thermal stability and effect on the internal components of the DSC are studied. Additionally, module parameters, which are affected by the glass frit, such as electrode distance and series resistance of the interconnect are investigated. Furthermore, the up-scalability of this new sealing technology is evaluated. Intrinsic stability of the DSC can only be studied if the extrinsic stability (the encapsulation) is assured. This is the case for glass frit sealing. Therefore, intrinsic long term stability tests are performed on small test cells (A = 2.5 cm²) completely sealed with glass frit in section 8.2. As electrolyte, undiluted ionic liquids based on propylmethylimmidazolium iodide are used. The accelerated ageing tests include continuous strong light soaking, thermal stress of 85 C in the dark and a thermal cycling between -4 C and +85 C. Good stability, but a very complex degradation behaviour is observed. 8.1 Glass frit sealing The sealing of plasma display panels is conducted with glass frit in a very comparable way as required for DSC modules (Chapter 3.5.1). Glass frit has previously been reported as a sealant for DSCs by Hinsch et al. [Hinsch '1] and Hanke et al. [Hanke '99] and is being increasingly noticed as a sealing material for DSC [Lee '6]. Glass frit has the advantage that it is cost-effectively applied via screen printing and has very stable thermal, chemical and mechanical properties. It has the disadvantage of a high processing temperature. Therefore, the dye has to be introduced into the module after the glass fusing process, similar as described in [Späth '3] and [Dai '4]. Details on the colouring unit, which was used and developed in this work are presented in the Appendix A1.1. Glass frit usually contains a high amount of lead oxide (PbO) which facilitates the wetting of the glass surface. In section 8.1.1 results are presented, which strongly imply that the lead content of some glass frit powders contaminate the platinum catalyst at the counter electrode of DSCs. A lead-free glass frit composition was found, which does not affect the catalytic activity of the platinum layer. The thermal stability of the glass frit sealing is tested in a thermal cycling test. These results are presented in section 8.1.2. The glass frit layer influences the electrode distance, as well as the series resistance of the interconnect. These parameters are determined in section 8.1.3 and 8.1.4. 124

8. Long term stability In section 8.1.5 it is shown that glass frit sealing is applicable for the up-scaling of the DSC technology to large areas. Colourful glass frit is examined and applied to DSC modules in section 8.1.6. 8.1.1 Effect of lead oxide on the catalytic activity of the platinum electrode Electrolyte cells were built to examine the influence of the lead oxide (PbO) content in the glass frit on the platinum counter electrode. These electrochemical cells consist of two identical, platinum coated TCO glass substrates in sandwich configuration filled with electrolyte [Hauch '1]. In electrolyte cells, the catalytic activity of the platinum electrode is especially easy to determine. The manufacturing procedure of electrolyte cells is described in detail in the Appendix A1.4. As primary sealing material lead-containing and lead-free glass frit was used and processed at a temperature of 63 C. Reference cells have been sealed with a polymerbased hotmelt foil (Surlyn, Dupont), which is applied at 13 C. To be able to compare the platinum electrodes of the polymer sealed and glass frit sealed cells, the platinum electrodes of the polymer sealed cells were exposed to the temperatures of the glass fusing process as well (63 C). The electrolyte consisted of.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile. A measure of the catalytic activity of the platinum electrode is the charge transfer resistance (compare Chapter 2.5.1). The charge transfer resistance of the platinum counter electrode is measured by electrical impedance spectroscopy (EIS) as described in [Kern '2]. All impedance measurements were performed with no bias potential with an amplitude of 5 mv. The EIS measurements were carried out with the Impedance Measuring Unit (IM 6) from Zahner. Table 8-1 shows the charge transfer resistances of platinum electrodes as measured in electrolyte cells by EIS. As sealing material PbO-containing and PbO-free glass frit has been used. As a reference, cells have also been sealed with a polymer hot-melt foil. Using a polymer sealing allows an investigation of the influence of the processing temperature on the charge transfer resistance. A processing temperature of 63 C results in a higher charge transfer resistance (4.3 Ωcm²) compared to the a low processing temperature of 45 C (.7 Ωcm²). Charge transfer resistance at the platinum electrodes Sealing material Temperature Polymer Lead-containing glass frit Lead-free glass frit 45 C.7 W. cm² - - 63 C 4.3 W. cm² 41 W. cm².2 W. cm² Table 8-1: Charge transfer resistances of platinum electrodes as measured in electrolyte cells by EIS. The sealing procedure differs in material and temperature 125

8. Long term stability When glass frit is used as a sealing material, a high processing temperature is required (63 C). From Table 8-1 it is clear, that the sealing with lead-containing glass frit results in a very high charge transfer resistance at the platinum electrode (41 Ωcm²), compared to the remarkable low charge transfer resistance, which is obtained when PbO-free glass frit is used (.2 Ωcm²). It is even lower than the charge transfer resistance achieved with polymer sealing. However, the reason for this might be that a different (newer) batch of platinum paste was used in the cells with lead-free glass frit. 1 Lead-containing glass frit FF = 37% Lead-free glass frit FF = 71% Current density / ma cm -2 8 6 4 2 1 2 3 4 5 6 7 8 Voltage (mv) Figure 8-1: I-V curves of DSCs sealed with lead-containing and lead-free glass frit. All other materials are identical. Figure 8-1 shows the I-V curves of 2 DSCs manufactured on test cells (2.5 cm²) which have been sealed with lead-containing and lead-free glass frit, respectively. All other materials were identical. The electrolyte of the cells consisted of.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile. The measurement was performed under a halogen lamp at AM 1.5 global (1 W/m²) taking into account spectral mismatch at a temperature of about 25 C. The influence of the sealing material on the fill factor can be seen clearly. The leadcontaining glass frit contaminates the platinum counter electrode. This results in a high charge transfer resistance, thus lowering the fill factor. The catalytic activity of the platinum counter electrode is not affected by the lead-free glass frit. 8.1.2 Thermal stability of glass frit In order to examine the thermal and mechanical stability of the glass frit seal and the Z-contact, special test devices were built. These devices resemble a strip-module (Chapter 7.3) with just the glass frit and silver layer and without electrolyte. The Z-interconnect (TCO / Silver / TCO) of these devices is directly accessible. 126

8. Long term stability Resistance / Ohm 1 8 6 4 Series connection in module 2 PT1 5 5 1 15 2 25 time / hours Figure 8-2: Thermal cycling of the series connections of a 3 x 3 cm² module protected by glass frit. One thermal cycle lasts 5 hours with a temperature variation from 4 C to 8 C. The temperature is recorded by a PT1. The resistance of the series connections follows the trend in temperature. For 5 cycles (25 hours) no degradation of the Z-contacts is observed. 25 2 15 1 Resistance / Ohm These test devices were exposed to thermal cycling tests. During one thermal cycle of 5 hours the temperature was varied from 4 C to 8 C. The resistance of the series connection was recorded (Figure 8-2). The resistance followed the trend of the temperature (PT1). For 5 cycles (25 hours) no degradation was observed as seen in Figure 8-2. These results imply, that the glass frit withstands the thermal cycling. Otherwise the electrical contact resistance of the series interconnect, which is located inside the glass frit barrier, would increase. It has to be mentioned though, that this test is not sufficient for mechanical stability testing as it does not incorporate the rates of cooling and differential conditions that can occur in use. 8.1.3 Electrode distance in glass frit dye solar modules Figure 8-3 shows a SEM picture of a cross section of the 3 x 3 cm² module. As can be seen, the electrode spacing is adjustable via the glass frit layer thickness. At the applied fusing temperatures the glass plates must soften to level out any unevenness of the glass, without reducing the electrode distance. At the applied temperatures, the glass frit must keep a high viscosity to maintain the electrode distance and prevent being drawn into the pores of the TiO 2 by capillary force. The cooling process must be slow enough (here: cooling over night), so that no strain is incorporated into the glass. The residual stress in the glass is checked with an optical glass strain detector (polarimeter by Arnold). 127

8. Long term stability 34.7 μm TiO 2 Glas Frit 33.8 μm Figure 8-3: SEM picture of a cross section of the 3 x 3 cm² module. The TiO 2 layer thickness is about 1 μm. The electrode spacing is adjusted via the glass frit layer thickness of about 35 μm (left). The electrode spacing only varies slightly over the whole cell width, dropping to about 34 μm in the middle of the cell (right). Using this technique the electrode spacing can be adjusted even over large areas. Thus, electrode distances of about 3 μm (Figure 8-3) to more than 6 μm have been realized in this work. The right choice of electrode distance depends on the electrolyte which is used. In general, an electrolyte with a low diffusion limited current density should be used in a DSC with a low electrode distance. On the other hand, a low electrode distance will increase the time of colouring. The flow rate of the dye solution reacts very sensitive to a change in electrode distance. It is worthwhile to mention, that the flow resistance of a liquid with viscosity η through two plates with distance d cell is given by the law of Hagen- Poiseuille [Gerthsen '3]: R flow 8η l = 8-1 d 2 cell Here, l is the length of the path through the plates. Accordingly, the flow resistance is 1 proportional to. 2 d cell 8.1.4 Resistance of the series interconnection in glass frit dye solar modules The resistance of the series connection is a crucial parameter affecting the fill factor of the module (see Chapter 7.2.5). To measure the resistance of the Z-contact (TCO / Silver / TCO) test devices (without electrolyte) of 3 x 3 cm² were manufactured, in which the Z-contacts were directly accessible (compare section 8.1.2). The value of the resistance of the Z-contact depends very much on the type of glass frit used. Leadcontaining glass frit with a low melting point flows under the silver and damages the silver / TCO contact resulting in a total resistance for the Z-contact of about.2 Ω cm². This value is improved by using lead-free glass frit with a high melting point, which preserves its layer thickness at the fusing temperatures and does not flow under the silver. Thus values of.2 Ω cm² are reached. Furthermore, lead-free glass frit is essential to 128

8. Long term stability ensure a low charge transfer resistance at the Pt counter electrode. The lead of the glass frit is contaminating the catalytic activity of the Pt layer during sintering (compare section 8.1.1). In addition, it is compulsory that the thickness of the glass frit layer is the same as the thickness of the silver layer. The silver layers of front and counter electrode must touch (at least pointwise) in order to form an electrical contact. A situation in which this is not the case can be seen in Figure 8-4. Figure 8-4: SEM picture of a silver interconnect with incorrect layer thickness. The silver layers form no electrical contact. 8.1.5 Up-scalability of glass frit sealing technology To demonstrate the possibility of up-scaling the glass frit DSC technology to even larger areas, 6 x 1 cm² blanks were built. The glass frit layer and the TiO 2 layer have been applied on these prototypes. The laser scribing has also been demonstrated on these large areas. It could be shown, that the glass fusing process works well on 6 x 1 cm² areas. Such a blank module sealed with glass frit is shown in Figure 8-5. It could be shown, that liquid can be injected into such prototypes over pneumatic cartridges with a pressure up to 7 bar without breakage of the glass frit sealing. This is a first indication of the mechanical stability of the glass frit sealing on these large areas. 129

8. Long term stability Figure 8-5: A 6 x 1 cm² blank module. The up-scaling of the glass frit sealing technique is demonstrated. In comparison to the sealing of DSCs with polymer hot melt foils, a screen printable glass frit sealing has a number of advantages apart from the thermal and chemical stability. Screen printing is a cost effective technique. Furthermore, it easily allows the precise positioning of the sealant between the cells of the module. When sealing DSCs with polymer hot melt foil, very flat glass plates are required, especially for large area modules. Whereas the high temperature during the glass fusing levels out any uneveness of the glass plates. In addition, the modules are clean and dry after the glass fusing and already sealed apart from the filling holes, allowing easy handling of the modules. A disadvantage of the glass frit sealing is the necessity of a colouring system, which pumps the dye solution through the module after the glass fusing. In an industrial production a number of colouring systems would be needed dependent on the desired throughput. However, with this colouring method the dye solution remains in a closed system and a smaller quantity of dye solution is required, when compared to immersing large area modules in a bath of dye solution. Details of the colouring system developed in this work are presented in the Appendix A1.1. 8.1.6 Colourful glass frit Another advantage of glass frit, is the possibility to vary its colour if desired. It is expected, that first products of DSC modules are not able to compete with conventional solar cell technologies in terms of power output and price alone. Rather, DSC modules are aimed at providing an additional architectural element for building integrated photovoltaics (BIPV). Unique selling points of a DSC facade in BIPV are colour, transparency and adaptable design. Another possible application are large area advertisement signs, which generate electricity at daytime and can be illuminated at night. 13

8. Long term stability With these applications in mind, glass frits of various colours have been tested, in order to examine their influence on the photovoltaic performance of DSC. Negative effects on the fill factor, similar to the contamination of the Pt counter electrode with PbO were expected. However, no significant effects on the photovoltaic performance were observed. Figure 8-6 shows the I-V curves of two interdigital, meander module with DSC logo (compare Chapter 5.5.1). Blue and grey (untreated) glass frit has been applied. Obviously, the photovoltaic performance of the module is not reduced by the application of blue glass frit as compared to the application of grey (untreated) glass frit. current / A -1. -.8 -.6 -.4 -.2. blue glass frit grey (untreated) glass frit Voc=4. V voc (one cell)= 667 mv Isc=832 ma jsc (one cell) = 8.49 ma/cm 2 FF=.51 eta act. area =3.2 %.2-4 -3-2 -1 1 2 3 4 voltage / V Figure 8-6: I-V curves of two interdigital, meander DSC module with DSC logo. Blue and grey (untreated) has been applied. Figure 8-7: Photo of a DSC logo module with blue glass frit. 131

132 8. Long term stability

8. Long term stability 8.2 Accelerated ageing of small test cells Accelerated ageing tests have been performed on DSCs. As electrolyte, ionic liquids based on propylmethylimidazole iodide (PMII) have been used. The accelerated ageing tests included continuous illumination with visible light (section 8.2.1) and temperature treatment at 85 C in the dark (section 8.2.3). Additionally, an ageing procedure, which consists of a temperature treatment at 85 C in the dark and subsequent visible light soaking is presented in section 8.2.4. Following this ageing procedure a thermal cycling test between -4 C and +85 C has been performed in section 8.2.5. Details on the accelerating ageing procedure is given in the Appendix A1.3. Tests on UV-stability have not been performed in this work. It is known, that UV-light is a critical degradation factor for DSCs [Pettersson '1]. The mechanism of UV-degradation in DSCs is described in detail in [Kern '1]. Progress has been made in improving UVstability by introducing additives to the electrolyte. Such UV-stabilizing additives, e.g. CaI 2 or MgI 2, dramatically enhance UV-stability [Hinsch '1]. However, for outdoor applications an easy and effective option to achieve UV-stability is the application of an UV-filtering polymer foil [Pettersson '1]. For the accelerated ageing tests, DSCs were built on so-called masterplates. One masterplate consists of 5 identical cells with areas of 5 x.5 cm². The masterplates were sealed completely with glass frit. The filling holes have also been sealed by glass frit. To seal a hole with glass frit, a small ceramic plug is fused into the holes during the glass fusing process. The ceramic plug features a hole for the colouring and filling. The hole in the ceramic plug is then sealed by glass frit, which is locally applied and melted onto the ceramic. The ceramic acts as a thermally insulating spacer. Details on the manufacturing process of masterplates are given in the Appendix A1.2. Thus, a reliable solution for sealing the holes on the glass surface has been found and tested on 2.5 cm² cells. The method is similar to the sealing of the exhaust hole in plasma display panels (see Chapter 3.5.1). The up-scalability of this hole sealing method has yet to be demonstrated. 133

8. Long term stability jsc / jsc Voc / Voc FF eta / eta 1.2 1..8.6.4.2. 1..8.6.4.2. 1..8.6.4.2. 1..8.6.4.2. 5 1 15 2 25 3 time / hours 5 1 15 2 25 3 time / hours 5 1 15 2 25 3 time / hours 1.2 5 1 15 2 25 3 time / hours Figure 8-8: Relative change of electrical parameters in a under continuous illumination with visible light. The test cells contained: Dye: N719 Electrolyte composition:.2 Mol I 2.5 Mol NMBI.1 Mol GSCN solvent: PMII : EMISCN (65:35) Initial electrical parameters: eta = 3.4 % jsc = 7.1 ma/cm² Voc = 7 mv FF =.69 8.2.1 Long term stability under visible light soaking Continuous strong light soaking has been done with a sulphur plasma lamp. During light soaking the test cells were operated at maximum power point. 134

8. Long term stability jsc / jsc Voc / Voc FF / FF eta / eta 1..8.6.4.2. 1..8.6.4.2. 1.2 1..8.6.4.2. 1..8.6.4.2. 5 1 15 time / hours 2 5 1 15 2 time / hours 5 1 15 2 time / hours.97 5 1 15 2 time / hours Figure 8-9: Relative change of electrical parameters in a under continuous illumination with visible light. The test cells contained: Dye: N719 Electrolyte composition:.5 M I 2.45 M NMBI solvent: PMII Initial electrical parameters: eta = 1.4 % jsc = 5. ma/cm² Voc = 62 mv FF =.43 A large number of cells (>5) with various electrolyte compositions has been tested. Overall, it showed that continuous visible light soaking is not a significant cause of degradation for DSCs. This confirms results published previously [Hinsch '1]. Figure 8-8 and Figure 8-9 show two examples of the most stable electrolyte candidates observed in this work. In the figures, the electrical parameters of the solar cells are recorded during visible light soaking. The electrical parameters are normalized with respect to their initial values. The initial values are given in the figure captions, as well as the electrolyte compositions and the type of dye. In case of the efficiency, the final normalized value is noted in the graph. 135

8. Long term stability 8.2.2 Estimating the number of stable turnovers for the Ruthenium dye The preceding ageing test shows clearly, that DSCs can be very stable under visible light soaking. In particular this means, that the dye is reduced fast enough from its oxidized state by the iodide. The number of turnovers of the dye (excitation of dye, oxidation and immediate reduction) can be estimated as follows: Assuming a short-circuit current density j sc =1 ma/cm² j e SC = 6.25 1 15 Oxidations s cm² 8-2 occur in the DSC. Assuming the density of dye on area of TiO 2 n dye = 1 g/m² and a molar weight of N719 of M dye =1187.7 g/mol, the number of dye molecules in 1 cm² is: dye molecules n 1 5.7 1 16 dye N A = 8-3 M cm² dye Here, N A is the Avogadro number. It follows that one dye molecule undergoes excitation, oxidation and reduction at a rate R turnover of: 1 R turnover = 1.23 8-4 s In other words, upon illumination a dye molecule absorbs about 1 photon each second. This is a remarkably slow rate, as has also been estimated by [Kay '94], using a different approach. The rate of reduction of an oxidized dye molecule by an iodide molecule has been measured to 1 8 s -1 [Hagfeldt '95]. It follows, that the dye is in its (stable) ground state most of the time. It was found that the N719 dye can survive 1 8 oxidations [Kohle '97]. Assuming 1 h of full sunlight in one year in Europe, the dye could therefore be stable for about 2 years. Assuming a higher short-circuit current density of j sc =2 ma/cm², the dye would still survive about 1 years. However, the number of oxidations which an average dye molecule can survive depends on many factors. The concentration of the reducing agent (iodide) and the reaction rate of reduction is an important parameter. Also, surface states on the TiO 2 may lead to irreversible reactions of an oxidized dye molecule. Altogether, the stable number of turnovers could well be only 1 7 as reported by Tributsch et al [Tributsch '4]. The measurements in Figure 8-8 and Figure 8-9, on the other hand, are direct evidence of at least 1 7 stable turnovers (24 hours). 136

8. Long term stability jsc / jsc Voc / Voc FF / FF eta / eta 1.2 1..8.6.4.2. 1..8.6.4.2. 1..8.6.4.2. 1..8.6.4.2. 2 4 6 8 1 12 time / hours 2 4 6 8 1 12 time / hours 2 4 6 8 1 12 time / hours.75 (.87) 2 4 6 8 1 12 time / hours Figure 8-1: Relative change of electrical parameters under heat test of 85 C in the dark. The test cells contained: Dye: Z97 Electrolyte composition:.5 M I 2 solvent: PMII Initial electrical parameters: eta = 3.1 % jsc = 1.3 ma/cm² Voc = 54 mv FF =.58 8.2.3 Long term stability under 85 C in the dark Accelerated ageing at 85 C in the dark was performed in hot-air convection ovens. A large number of cells (>5) with various electrolyte compositions has been tested. Overall, it showed that the heat test at 85 C is the part of the IEC test, which is the most difficult to pass. Although the glass frit sealing is reliable, degradation occurs depending on electrolyte composition and purity. The purity of the electrolyte turned out to be a central requirement in achieving long term stability at elevated temperatures. In particular, the water content of the electrolytes impairs long term stability. For that reason, the hydrophobic dye Z97 [Wang '4] [cis-rull'(scn)(2) (L=4,4'-dicarboxylic acid-2,2'- bipyridine, L'=4,4'-dinonyl-2,2'-bipyridine)] yields better stability results than the N719. 137

8. Long term stability jsc / jsc Voc / Voc FF / FF eta / eta 1.2 1..8.6.4.2. 1..8.6.4.2. 1..8.6.4.2. 1.2 1..8.6.4.2. 2 4 6 8 1 12 time / hours 2 4 6 8 1 12 time / hours 2 4 6 8 1 12 time / hours.97 (.82) 2 4 6 8 1 12 time / hours Figure 8-11: Relative change of electrical parameters under heat test of 85 C in the dark. The test cells contained: Dye: Z97 Electrolyte composition:.5 M I 2.45 M NMBI solvent: PMII Initial electrical parameters: eta = 2.7 % jsc = 7.4 ma/cm² Voc = 62 mv FF =.59 Figure 8-1 and Figure 8-11 show two examples of the most stable electrolyte candidates observed in this work. In the figures, the electrical parameters of the solar cells are recorded during the heat test at 85 C in the dark. The electrical parameters are normalized with respect to their initial values. The initial values are given in the figure captions, as well as the electrolyte compositions and the type of dye. In case of the efficiency, the final normalized value is noted in the graph. Figure 8-1 and Figure 8-11 both show a rapid initial change (rise or fall) in cell performance during the first 2 hours of thermal ageing at 85 C in the dark (preageing), which is followed by stabilization for more than 1 hours at the same conditions. The final efficiency normalized with respect to the efficiency after 2 h is noted in the graph in parentheses. 138

8. Long term stability eta / eta eta = 3.1 %; jsc = 1.3 ma/cm 2 ; Voc = 54 mv; FF =.58 1. thermal ageing constant illumination.8.6 light soaking.4 pre-ageing 85 C in visible light.2 the dark soaking. 5 1 15 2 25 3 time / hours.78 Figure 8-12: Relative change of conversion efficiency in a accelerated ageing test of a DSC. Dye: Z97; Electrolyte composition 1:.5 M I 2, solvent: PMII eta / eta eta = 2.7 %; jsc = 7.4 ma/cm 2 ; Voc = 62 mv; FF =.59 1.4 1.2 1..8.6.4.2. pre-ageing thermal ageing 85 C in the dark constant illumination.97 light soaking visible light soaking 5 1 15 2 25 3 time / hours Figure 8-13: Relative change of conversion efficiency in a accelerated ageing test of a DSC. Dye: Z97; Electrolyte composition 2:.5 M I 2,.45 M NMBI, solvent: PMII 8.2.4 Combined ageing under 85 C in the dark and subsequent ageing under visible light A combined ageing procedure of heat test and visible light soaking has been applied to a large number of cells (>5) with various electrolyte compositions. This procedure consisted of an accelerated ageing tests under 85 C in the dark (about 14 hours) followed by subsequent illumination with visible light (1 sun, about 17 hours). In Figure 8-12 and Figure 8-13 the relative change of conversion efficiency is shown for two of the most stable electrolyte compositions. A very complex ageing behaviour is observed. The results often show a rapid initial change (rise or fall) in cell performance during the first 2 hours of thermal ageing at 85 C in the dark (pre-ageing in Figure 8-12), which is followed by only a slight degradation for more than 1 hours at the same conditions (thermal ageing in Figure 8-12). Under subsequent illumination with 1 sun equivalent light intensity a recovery effect has often been observed over 2 hours (light soaking in Figure 8-12), which is followed by only a slight degradation for more than 1 hours at the same conditions (constant illumination in Figure 8-12). 139

8. Long term stability These 4 regions could be identified in the ageing curves of all tested electrolyte compositions. The recovery effect has already been observed by Sommeling et al. [Sommeling '4]. Table 8-2 shows the gradual change of electrical parameters in all 4 regions for the cells of Figure 8-12 and Figure 8-13. The electrolytes contained.5 M I 2 in PMII (electrolyte 1) and.5 M I 2,.45 M n-methylbenzimidazole (NMBI) in PMII (electrolyte 2). The best stability was obtained with electrolyte 2, showing a total change in efficiency of only 3 % for the complete ageing procedure. Figure 8-14 shows the measured I-V curves of the DSC containing electrolyte 1 at the different ageing regions of Figure 8-12. A limited current can be seen after thermal ageing in forward bias (the deviating slope). The recovery effect to this current limitation due to light soaking can be seen as well. 8.2.5 Thermal cycling test Temperature cycling tests were performed at Schott Solar in Alzenau, Germany. After the ageing procedure of Figure 8-12 all test cells were exposed to a thermal cycling test. The temperature was cycled between -4 C and +85 C at a rate of 2.8 cycles per day for 43 hours (5 cycles). The most stable result was also obtained with electrolyte 2, showing a degradation in efficiency of again only 3 % for the thermal cycling test. The most important result of this thermal cycling is that the cells are stable under -4 C. Potential stability at 85 C has already been shown section 8.2.3 and potential stability of the sealing has already been shown in 8.1.2. Gradual change in performance Preageing Thermal ageing Light soaking Constant illumination Total change excluding preageing Total change PMII + Iod eta -13% -12% 12% -8% -1% -22% I sc 11% -7% 12% 3% 6% 18% V oc -16% -8% 2% -7% -13% -27% FF -7% 3% -1% -4% -3% -9% PMII+Iod+NMBI eta 17% -23% 9% -1% -17% -3% I sc 2% -15% 6% -2% -12% 5% V oc -8% -1% 3% % -7% -15% FF 6% % 1% % 1% 7% Table 8-2: Gradual change of electrical parameters for 2 different electrolyte compositions for the combined ageing procedure. Negative values indicate a degradation, positive values indicate a recovery. 14

8. Long term stability Current / ma cm -2 12 8 4-4 -8-12 -16..2.4.6 Voltage / V after pre-ageing (2 h) after thermal ageing (14 h) after continuous illumination (3 h) Figure 8-14: I-V curves of a DSC (A = 2.5 cm²) containing electrolyte 1 after preageing, thermal ageing and continuous illumination. 8.3 Model for the degradation under thermal ageing In order to find the cause of the degradation under thermal ageing, all components of the DSC must be considered. The thermal ageing at 85 C occurs in the dark. Therefore possible ageing mechanisms, which might occur under elevated temperature and illumination are excluded here, as investigated in [Sommeling '4]. Furthermore, the degradation mechanism must be reversible, at least to some extent, as the recovery effect suggests. And additionally, the degradation must lead to a limitation in current as the measurement in Figure 8-14 suggests. 8.3.1 TCO layer A degradation of the TCO layer under a temperature of 85 C can be excluded. The TCO layer (SnO 2 :F) is stable at temperature over 6 C, as are applied in the glass fusing process. Furthermore, SnO 2 :F is chemically very stable. A high charge transfer resistance at the TCO/electrolyte interface at the front (TiO 2 ) electrode is required in a DSC. Electrolyte substances or impurities might adsorb on the TCO at high temperatures and reduce the charge transfer resistance. This would decrease the open-circuit voltage. But this would not limit the current in the fourth quadrant of Figure 8-14. 8.3.2 TiO 2 layer The same argument applies for the TiO 2 layer. The TiO 2 layer is stable at temperature over 6 C, as are applied in the glass fusing process. 141

8. Long term stability However, electrolyte substances or impurities might adsorb on the surface of the TiO 2 at high temperatures and form recombination centres. This would decrease the shortcircuit current. But again, this would not limit the current in the fourth quadrant of Figure 8-14. 8.3.3 Platinum layer The platinum layer itself is stable at 85 C, as it has been treated with over 6 C in the glass fusing process. The platinum layer might degrade in the presence of the electrolyte at high temperature, either by desorption or contamination. However, a degradation of the platinum layer would result in an increase of series resistance and thus lead to a decrease of the fill factor. The fill factor, though, is not reduced under thermal ageing (Table 8-2). 8.3.4 Dye The dye could be decomposed under high temperatures. Furthermore, dye desorption from the TiO 2 could be a degradation mechanism. However, this would not explain the recovery effect upon illumination, nor the current limitation in the fourth quadrant of Figure 8-14. 8.3.5 Electrolyte This leaves only the electrolyte as the component which degrades under thermal ageing in the dark. It is known, that certain solvents react under high temperatures in the presence of water. For example, methoxyacetonitrile (MACN) reacts to form methoxyacetamid [Kern '1]. However, the electrolytes in this work, especially electrolyte 1 and 2 do not contain any solvents. And a recovery mechanism for the decomposition of the solvent is extremely unlikely as well. It is more likely, that the thermal ageing is due to a consumption of triiodide, which is then regenerated under light soaking. A bleaching of the electrolyte, which means depletion of triiodide is often observed under degradation, especially under illumination with UV-light [Kern '1]. Here, the recovery effect is easily explained: the regeneration of triiodide occurs, when the dye is reduced by iodide upon photo-excitation forming triiodide. The initial rise of photocurrent is then attributed to an optical gain, since triiodide strongly absorbs in the blue spectral range. The current limitation in the fourth quadrant of Figure 8-14 further supports the hypothesis of triiodide depletion under thermal ageing. A lower triiodide concentration leads to a lower diffusion limited current of the electrolyte. 142

8. Long term stability A hypothetical mechanism for the depletion of triiodide is given by Tributsch et al. [Tributsch '4]. Upon illumination and in a water-free and oxygen-free environment, the oxidized state of the dye D + is reduced by iodide leading to a formation of triiodide: + 2D + 3I I + 3 2D 8-5 In the presence of water and oxygen, however, and upon illumination, iodate may be formed within a certain probability instead of triiodide: + + 2D + 3I + 4 O2 + H2O 3IO3 + 2H + 2D 8-6 This mechanism leads to a depletion of triodide, since triodide is consumed at the platinum counter electrode but not regenerated completed upon reduction of the oxidized dye. However, this reaction mechanism is only valid under illumination. But under illumination at low temperatures (35 C) no degradation was observed (section 8.2.1). During thermal ageing in the dark, there are various possible chemical reactions of iodine (triiodide) with water: + I2 + H2O HIO + I + H HIO + OI + H + (hydrolysis, formation of HIO (hypoiodous acid)) (hypoiodous acid dissociation) 8-7 8-8 3HIO IO - 3 + 2I + 3H + (disproportionation of HIO) 8-9 + 3I2 + 3H2O IO3 + 5I + 6H (iodate formation) 8-1 I (triiodide formation) 8-11 + 2 I I3 All the above reactions, except for those that form iodate (IO 3- ), are very rapid and reach equilibrium quickly. A possible reaction mechanism for the depletion of triiodide at high temperatures in the dark would therefore be the disproportionation reaction (reactions 8-1 and 8-11): + 3I3 + 3H2O IO3 + 8 I + 6 H 8-12 The formation of H + in the electrolyte increases the binding energy of the electrons in the TiO 2, thus lowering the conduction band and decreasing the open-circuit voltage. This is also observed in the measurements (Table 8-2). In the following, we will assume, that triodide is consumed by water by this reaction. However, the following calculation will remain valid for any other reaction mechanism in 143

8. Long term stability which triiodide is consumed. The impurity in the electrolyte which reacts with triiodide may also be some other substance than water. 8.3.6 Modelling the degradation of the short-circuit current density under thermal ageing Especially, in high-viscous electrolytes when the short-circuit current density is already limited by the diffusion limited current, a reduction of triiodide concentration leads (immediately) to a lowering of the short-circuit current density. The diffusion limited current is: j lim 2 n ( t) I3 ( t) = ν e D 8-13 e I3 d Cell Here, ν e- is the number of electrons transferred in the redox reaction (ν e- =2). The initial 2 1 concentration of triiodide is.5 M, n () = 3.11 1. The diffusion coefficient of I3 3 cm 7 cm² triiodide in PMII is about D = 1.8 1 [Rau '5] and the cell thickness is about I3 s d = 3 to 35 μm in the architecture of the masterplates. Cell With these parameters ( d Cell = 33 μm ) we obtain: ma jlim () = 1.5 8-14 cm² This means, that the short-circuit current density of the cell in Figure 8-12 with electrolyte 1 is determined by the diffusion limited current of triiodide even before ageing. After 14 hours of ageing under 85 C, the short-circuit current density has degraded by 7 % (Table 8-2). Assuming that this degradation is due to triiodide consumption, the concentration of triiodide has also decreased by 7 %, i.e. the consumption of triiodide has 19 1 been Δn = 2.1 1. A change of only 7 % in triiodide concentration, is not I 3 3 cm visible by inspection. Thus, no actual bleaching can be observed. According to Equation 8-12, the same number of molecules of water are required to form iodate from triiodide. It follows, that at least 6.3 1-4 g/ml of water must be present in the electrolyte, i.e. about.6 weight %. It is likely that water impurities of this amount are present in the (hydrophilic) electrolytes used. In an empirical attempt to find a model for the degradation of the short-circuit current density based on these measurements, an exponential decay of triiodide concentration is assumed: 144

8. Long term stability n t) = n () exp( k t) 8-15 ( I deg 3 I3 Here, k deg is the rate of degradation of triiodide. Most chemical reactions are faster at higher temperatures. For the degradation rate k deg one may therefore assume an Arrhenius behaviour: k deg ( T ) E A'exp( k T B ) A = 8-16 Here A is a pre-exponential factor and E A the activation energy of the reaction. However, these values are not known. For k deg =1.46 1-8 s -1 the calculated degradation of the short-circuit current density is shown in Figure 8-15. The degradation rate has been chosen such, that a degradation of 7 % has occurred after 14 hours as in the ageing test of Figure 8-12. The extrapolation in Figure 8-15 yields a short-circuit current density of 6.2 ma/cm² after 1 hours, i.e a degradation of about 4 %. 8.4 Conclusions of Chapter 8 The most important result of this chapter, is the successful identification of a reliable and realizable encapsulation method for DSC modules. It was shown, that screen printed glass frit is applicable for sealing large area DSC modules. But it is essential, that the glass frit contains no lead oxide (PbO), which has been shown to reduce the catalytic activity of the platinum counter electrode. jsc / ma cm -2 12 1 8 6 4 2 2 4 6 8 1 time / hours Figure 8-15: Calculated degradation of the short-circuit current density. The model attributes the degradation to a consumption of triiodide. The parameters are fitted to the measurements for the first 14 hours. Extrapolation to 1 hours yields a short-circuit current density of 6.2 ma/cm². 145

8. Long term stability Using glass frit sealed test cells (A = 2.5 cm²), long term stability tests have been conducted. It can be concluded, that visible light is not a degradation factor for DSCs. Especially, the dye shows no degradation for more than 1 7 turnovers. The heat test under 85 C in the dark, however, remains a challenge. Here, stable cells with degradation of less than 5 % for more than 1 hours of heat testing have been observed. The most stable electrolytes in the heat test contained.5 M I 2 in PMII (electrolyte 1) and.5 M I 2,.45 M n-methylbenzimidazole (NMBI) in PMII (electrolyte 2). These electrolytes show an initial overall efficiency of about 3 %. In a subsequently performed thermal cycling test (-4 C to +85 C, 5 cycles) a 2.5 cm² DSC with electrolyte 2 also showed only a slight degradation of less than 5 % in conversion efficiency. Nevertheless, reproducibility has not been satisfactory. This leads to the conclusion, that the purity of the electrolyte is a crucial factor. Especially, the water content in the electrolyte is a dominant degradation cause. Measurements imply that the degradation of the short-circuit current density may be attributed to the decrease of the diffusion limited current. This leads to the assumption, that triodide is consumed during thermal ageing. In a model, a reaction mechanism is proposed in which the formation of iodate occurs from triiodide in the presence of water. The degradation of the short-circuit current density can thus roughly be estimated. 146

9. State of the dye solar module technology 9. State of the dye solar module technology The dye solar cell (DSC) technology has developed very quickly since laboratory scale efficiencies of over 1 % have been reported. Because of the potential for low production costs and attractive colour and design, considerable efforts are being increasingly undertaken to enable a commercial up-scaling of this new type of solar cell. The up-scaling of the DSC technology to large areas is an exciting new branch in photovoltaic research and development (R&D). Many industrial companies and research groups are dedicated to develop a large area DSC technology. Because of the electrochemical working principle and the novelty of this technology, numerous approaches exist for large area DSC modules. Therefore, it is not surprising that each of the R&D institutions has developed its own original and innovative approach. In this chapter, a brief review will be presented on the existing module concepts and the state of the R&D on DSC module technology. It should be mentioned, however, that to date no commercial products exist. The state of the research on single DSCs can be found in good review papers, which have been presented by e.g. Grätzel [Grätzel '3,Grätzel '5] and Tributsch [Tributsch '4]. Assessment of costs are presented by McConnel [McConnell '2] and Halme [Halme '2]. Environmental aspects have been studied by Greijer [Greijer '1] and Bowerman [Bowermann '1]. 147

9. State of the dye solar module technology 9.1 Research and development on dye solar module technology A number of R&D groups are working on the development of large area DSC modules. The following discussion of research results will be presented by group rather than by technological issues. This list has not the intention to be complete, but is based on recent publications and press releases. Furthermore, details on the exact module architecture (an efficiency) is not always publicly available. Some new start-up companies are not mentioned here, as no publications or exhibition of prototypes have been presented yet. New start-up companies in the field of DSC modules are, e.g. Orionsolar, Solaris Nanosciences, PEC Solar Cell, Konarka, Solar Technologies S.A., Helios Technologies and Solexant. 9.1.1 Sharp Corporation, Ecological Technology Development Centre The Sharp group (Japan) holds the current, certified world record efficiency on DSCs on an area of 1 cm² with 1.4 % [Chiba '5]. On smaller areas (.219 cm²) an efficiency of 11.1 % was demonstrated (certified). Furthermore, the Sharp group is working on DSC modules of 25 x 25 cm² with a W-interconnect (Figure 9-1). 9.1.2 Toyota Central R&D Laboratory and Aisin, Seiki A widespread DSC technology at present is the parallel module design with a currentcollecting silver grid, which is protected by a polymer-based hot melt foil. Large scale DSC panels have thus been developed and tested under outdoor conditions for half a year by the Toyota group (Japan). Toyoda at el. reported an increase of 1-2 % power under outdoor conditions as compared to their rated power output under standard reporting conditions [Toyoda '4] (Figure 9-3). Furthermore, monolithic DSC module have been presented (Figure 9-2). The Toyota group exhibited an artistic DSC module façade on the EXPO 25 (PAPI Dreamhouse). Here, semi-transparent DSC modules have been furnished with a reflective background. The appearance of the facade therefore changes significantly with direct and diffusive illumination (Figure 9-4). 148

9. State of the dye solar module technology Figure 9-1: DSC module of size 25 x 25 cm² developed by the Sharp group. Here, a black dye is used and a W-interconnect. Figure 9-2: 3 x 3 cm² monolithic DSC module by Toyota. 9.1.3 Sony, Materials Science Laboratories The Sony group in Stuttgart, Germany has developed a tandem DSC with efficiencies of 1.5 % using a polymer gel electrolyte [Dürr '4]. 9.1.4 Fujikura Ltd., Material Technology Laboratory An electroplated nickel grid protected by a fluorine doped SnO 2 layer has been successfully tested on 1 x 1 cm² cells [Okada '4]. Furthermore, the Fujikura group (Japan) developed a 119 x 84 cm² module consisting of 16 cells (Figure 9-5). 9.1.5 Dyesol Dyesol (the former Sustainable Technologies Australia) has demonstrated a DSC facade of 2 m² on the CISRO Energy Centre in Newcastle, NSW Australia (Figure 9-7). However, no long term stability results have been published so far [Tulloch '4]. sunny day cloudy day Figure 9-3: 1 x 1 m² DSC module used in outdoor tests by Toyota. Figure 9-4: Expo PAPI Dreamhouse DSC facade by Toyota. 149

9. State of the dye solar module technology Figure 9-5: 119 x 84 cm² module consisting of 16 cell by the company Fujikura. Figure 9-6: Small, monolithic DSC modules for indoor application developed by the IVF in Sweden. 9.1.6 Solaronix The Swiss company Solaronix has developed a 5 x 5 cm² DSC module with W- connection (Figure 9-8). 9.1.7 Hitachi Hitachi (Japan) has presented a 3 x 3 cm² DSC module (Figure 9-9) in a press release in 24. Figure 9-7: DSC facade of 2 m² on the CISRO Energy Centre in Newcastle, NSW Australia by the company Dyesol. 15

9. State of the dye solar module technology Figure 9-8: 5 x 5 cm² DSC module with W-connection by the company Solaronix. Figure 9-9: 3 x 3 cm² DSC module by the company Hitachi. 9.1.8 Electronic and Telecommunications Research Institute The ETRI (Korea) has published a transparent DSC window of 3 x 3 cm² [Kang '3]. 9.1.9 Energy Centre Netherlands Parallel connected 1 x 1 cm² DSC modules have been demonstrated by Späth et al. from the ECN group (Netherlands) by protecting a current-collecting silver grid with a polymer-based hot melt foil [Späth '3]. This technology has been scaled up to an area of 3 x 3 cm² by a series connection of 4 cells (Figure 9-11). 9.1.1 Peccell Peccell, a company dedicated to commercialization of DSCs in Japan, has demonstrated 3 x 3 cm² DSC modules on a flexible polymer substrate (Figure 9-13). Figure 9-1: Transparent DSC window of 3 x 3 cm² by ETRI. Figure 9-11: 3 x 3 cm² DSC modules have been demonstrated by the ECN group 151

9. State of the dye solar module technology Figure 9-12: 5 W DSC power station demonstrated by the IPP of the Chinese Academy of Sciences. 9.1.11 Institute of Plasma Physics The IPP of the Chinese Academy of Sciences demonstrated a 5 W DSC power station of several 4 x 6 cm² panels. Furthermore, DSC modules of size 15 x 2 cm² with 6 % efficiency have been published [Dai '5]. Stability results have not been reported so far. 9.1.12 IVF, Industrial Research and Development Corporation The IVF, Industrial Research and Development Corporation in Sweden has developed small, monolilthic DSC modules for indoor application (Figure 9-6) [Pettersson '3]. Excellent long term stability could be shown. The stability was best under short-circuit conditions and a degradation could be observed under open-circuit conditions under illumination. With the results of Chapter 6, the hypothesis may be formulated, that the degradation has been due to photoinduced electrophoresis in this monolithically interconnected modules. Figure 9-13: 3 x 3 cm² DSC modules on a flexible polymer substrate by the company Peccell. Figure 9-14: 3 x 3 cm² DSC module of parallel interconnection developed by ITRI. 152

9. State of the dye solar module technology 9.1.13 Industrial Technology Research Institute The ITRI in Taiwan has developed a 3 x 3 cm² DSC module of parallel interconnection (Figure 9-14). 9.1.14 Fraunhofer Institute for Solar Energy Systems And last, but not least, the ISE group in Freiburg, Germany, where this research has been carried out, has developed 3 x3 cm² DSC modules. A semi-transparent, decorative DSC facade has been exhibited at the Hanover Trade Fair Industry 26 (Figure 9-15). Logo or art-deco DSC modules with different design and colourful glass frit have also been exhibited (Figure 9-16) [Photon '6]. Compared to other DSC module concepts, these modules are unique in terms of Glass frit sealing Interdigital meander architecture A design in the scattering layer Figure 9-15: A semi-transparent, decorative DSC facade has been exhibited at the Hanover Trade Fair Industry 26 by the Fraunhofer ISE. 153

9. State of the dye solar module technology Figure 9-16: Logo or art-deco DSC modules as exhibited at the Hanover Trade Fair Industry 26 by the Fraunhofer ISE. 9.2 Conclusions of chapter 9 In this chapter, a brief overview of the recent R&D status on the DSC module technology has been presented. The DSC module concepts of the major R&D groups have been portrayed. These clearly show, that a technological development of DSC modules is intensively pursued. However, only few publications exist on the investigation of the module-related physics of DSC. Such an investigation has been presented here. The results of this work are relevant for all of the shown DSC module concepts. The study of reverse bias on DSC is relevant for the partial shading of DSC modules in general (Chapter 5). The investigation of the photoinduced electrophoresis applies to all module concepts with integrated series connection (Chapter 6). Optimization of module design is obviously required for all solar technologies (Chapter 7). The long term stability testing, especially the developed glass frit sealing (Chapter 8) may be incorporated in all presented module concepts, except those on flexible polymer foil. The work conducted on decorative module design, using a patterned scattering layer (Chapter 5.5) and colourful glass frit (Chapter 8.1.6), may be applied to all module concepts, except modules with the monolithic or the W-connection. 154

1. Summary and Conclusions 1. Summary and Conclusions The dye solar cell (DSC) technology is a new solar technology very close to a possible commercialization. With the DSC technology, solar applications are realizable, which drastically differ in colour, design and transparency from conventional solid-state solar technologies. In this dissertation, photovoltaic modules of dye solar cells were developed and investigated. The conclusions of this work are summarized in the following. 1.1 Module-related physics of the dye solar cell Module-related physics of the DSC is concerned with the processes that are significant when single cells are interconnected in series to form a large area module. The modulerelated aspects investigated in this work were: partial shading leak in the internal sealing (photoinduced electrophoresis) optimization of module design 155

1. Summary and Conclusions In particular, the module-related degradation mechanisms were identified. Additionally, a characterization method for DSC modules (SRPI) was developed. 1.1.1 Partial shading or electrical mismatching (Chapter 5) When one DSC in a series connection is shaded, the current of the module passes this cell in reverse bias. Then, the charge transport directly occurs from the electrolyte to the front TCO electrode (see Figure 1-1). The TiO 2 is not significantly involved in reverse bias operation. The charge transfer at the front TCO electrode-electrolyte interface can be modelled with a Butler-Vollmer equation. Furthermore, the charge transfer is catalyzed by dye molecules, which are adsorbed on the TCO surface. This results in an extremely low breakdown voltage of a DSC, i.e. a large reverse current is already conducted at very low reverse bias voltages (ca. 5 mv). Hence, the I-V characteristics of a DSC in reverse bias differs significantly from a typical diode behaviour. In that respect, the DSC behaves like a solar cell with integrated bypass diode (see Figure 1-2). TCO substrate TiO 2 electrolyte charge transfer under reverse bias Figure 1-1: The electron transfer under reverse bias occurs directly from the electrolyte into the transparent conduction oxide (TCO). current density / ma cm -2 16 8-8 -16 V br 16 8-8 -16 Figure 1-2: Measured I-V curve of a typical DSC (area=2.5 cm²) in the dark. The breakdown voltage (V br ) is at -5 mv. -6-4 -2 2 4 6 voltage [mv] As a consequence, it may be concluded, that the partial shading of a DSC module is much less critical than of a conventional solar module. Both in terms of power loss due to partial shading, and degradation due to hot-spots (which occur in conventional solar cells), no bypass diodes are required. This has been experimentally validated for a time scale, which is relevant for the occasional partial shading under outdoor conditions. I.e. long term tests on DSC under reverse bias operation have shown no degradation in overall efficiency for more than 1 hours. 156

1. Summary and Conclusions It should be mentioned, however, that a very thin (<1 nm) dense blocking layer of TiO 2 on the front TCO substrate can suppress the charge transfer route under reverse bias. This results in an increase of the breakdown voltage (>1 V). Then, the high voltage drop under reverse bias operation almost immediately destroys the cell. Therefore, such a blocking layer though slightly increasing the efficiency by suppressing the recombination under forward bias is not recommended for the use in series interconnected DSC modules. Using these results, a model for the I-V curve of a DSC has been presented, which is valid over the complete voltage range. Thus, the calculation of I-V curves of partially shaded DSC modules is possible. 1.1.2 Spatially resolved photocurrent imaging technique (Chapter 5) Based on this model, a method was developed to measure the spatially resolved photocurrent image (SRPI) of a series interconnected DSC module. The small light beam induced photocurrent of one cell can be measured in reverse bias, although all other cells in the series connection are kept in the dark. Without applied bias voltage a measurement is not possible, since the non-illuminated cells would block the photoinduced current. This is a very valuable technique to investigate inhomogeneities in the scale-up process. Additionally it should be mentioned, that valuable information can be gained about the short-circuit currents of individual cells in a DSC module by the possible steps in the I-V curve especially at reverse bias potentials. Therefore, DSC modules should always be measured over the full voltage range. 1.1.3 Photoinduced electrophoresis (Chapter 6) Photoinduced electrophoresis, the separation of the redox couple (triiodide and iodide) in the electrolyte of series interconnected DSCs, is a relevant module-related degradation mechanism. The photoinduced electrophoresis has been studied experimentally and has been modelled in a transient diffusion model. The model allows the calculation of the degrading short-circuit current under illumination. The key parameter is the diffusion constant of triiodide in the internal sealing material. The simulations showed, that very high requirements on the barrier properties of the internal sealing material must be fulfilled. It was calculated that the diffusion constant of triiodide in the internal sealing material has to be less than D[I 3- ] < 1-11 cm²/s in order to achieve module lifetimes higher than 1 years. Such low permeability is not achievable with standard polymers. Therefore glass frit, which has excellent barrier properties is recommended for the use in DSC modules with integrated series connection. 157

1. Summary and Conclusions 1.1.4 Optimization of module design (Chapter 7) The electrical optimization of module design has been performed with a numerical model of distributed series resistances. The module parameters include the resistances of the TCO substrate and the electrical interconnects, as well as the width of photoactive and photo-inactive area. Essentially, the free parameter in designing a module is the width of photoactive area, which was calculated to be optimal at 7-8 mm for typical electrical parameters of a DSC. In this work, the width of the photo-inactive area was (technologically) fixed at 2.5 mm. In this case, a DSC with intrinsic efficiency of 8.6 % leads to a module efficiency of 5.6 %. A DSC with mediocre materials of only 3.9 % yields a module efficiency of 2.8 %. However, the calculations showed that a reduced width of photo-inactive area of 1.5 mm already results in an increase of overall efficiency of about 1 % (relative). A width of 1.5 mm seems very realistic with screen printing technology. Additionally, the calculations revealed, that a higher efficiency is obtained, when the module is optimized for light intensities lower than 1 W/m². Although the efficiency of the intrinsic DSC decreases with lower light intensity, the optimal width of photoactive area increases due to lower current densities. This results in a better ratio of photoactive to photo-inactive area, thus yielding higher module efficiencies. In the calculations, a module with 5.6 % under 1 W/m², reached an efficiency of 6. % under 2 W/m² with an optimal width of photoactive area of 11 mm. Since the illumination intensity in Europe is mostly below 1 W/m², this point should be carefully considered by the module designer. 1.2 Production technology of dye solar modules Based on these investigations, DSC modules of size 3 x 3 cm² have been developed. The manufacturing process is based on screen printing. Three main results are summarized in the following: The investigation of a hermetic sealing material The results of long term tests on single DSCs The development of a new DSC-specific module design 158

1. Summary and Conclusions 1.2.1 Hermetic sealing material (Chapter 8) The photoinduced electrophoresis, which was studied in Chapter 6, turned out to be a relevant degradation mechanism in DSC modules with integrated series connection. The calculated diffusion constant of triiodide, which is required inside the internal sealing material is extremely low. Such low permeability is not achievable with standard polymers. Therefore, glass frit which has excellent barrier properties has been investigated and successfully identified as a reliable and realizable encapsulation method for DSC modules. Evidently, screen printed glass frit is applicable for sealing large area DSC modules. The possibility of up-scaling the glass frit DSC technology was shown to areas up to 6 x 1 cm². But it is essential, that the glass frit contains no lead oxide (PbO), which reduces the catalytic activity of the platinum counter electrode. 1.2.2 Long term stability (Chapter 8) Using glass frit sealed test cells (A = 2.5 cm²), long term stability tests have been performed. It can be concluded, that visible light is not a degradation factor for DSCs. Especially, the dye shows no degradation for more than 1 7 turnovers, as estimated from stable illumination test of about 3 hours. The heat test under 85 C in the dark, however, remains a challenge. Here, stable cells with degradation of less than 5 % for more than 1 hours of heat testing have been observed. The most stable electrolytes in the heat test contained.5 M I 2 in PMII (electrolyte 1) and.5 M I 2,.45 M n-methylbenzimidazole (NMBI) in PMII (electrolyte 2). These electrolytes showed an initial overall efficiency of about 3 %. In a subsequently performed thermal cycling test (-4 C to +85 C, 5 cycles) a 2.5 cm² DSC with electrolyte 2 also showed only a slight degradation of less than 5 % in conversion efficiency. Nevertheless, degradation has been observed in the heat tests. While the fill factor remains stable in most cases, the current and voltage decrease. Accompanying the degradation, a decreasing diffusion-limited current can be observed in the I-V curves. Upon illumination with visible light a recovery effect regarding the electrical parameters as well as the diffusion-limited current occurs. In this thesis, a reaction mechanism has been - proposed in which the formation of IO 3 (iodate) and H + occurs from triiodide in the presence of water. A depletion of triiodide leads to a decrease in the diffusion limited current, while an increase of H + leads to a decrease in the voltage. Upon illumination, triiodide is produced again in a DSC, thus explaining the recovery effect. In a model, the degradation of the short-circuit current density can thus be roughly estimated. This leads to the conclusion, that the purity of the electrolyte is a crucial factor. Especially, the water content in the electrolyte is a dominant degradation cause. 159

1. Summary and Conclusions 1.2.3 Interdigital meander design (Chapter 7) In Chapter 7, the optimal cell width in a DSC module was calculated to be 7-8 mm. This result applies to a design of strip-shaped cells. Based on these calculations, a new DSC-specific module design was developed with meander-shaped cells. The main advantage of this interdigital meander design is that the number of filling holes for the electrolyte is reduced drastically. In terms of industrial producibility, this a significant advantage. With this design efficiencies of 3.8 % on active area were achieved, resulting in an efficiency on total area of about 2.9 %. An interdigital meander module and its I-V curve can be seen in Figure 1-3. Figure 1-3: Left: 3 x 3 cm² meander shaped module sealed with glass frit. 6 cells are interconnected in series. Bottom: I-V characteristics of an opaque (with scattering layer) 3 x 3 cm² meander shaped module sealed with glass frit. -.8 current / A -.6 -.4 -.2. Voc = 4,4 V voc (one cell)= 733 mv Isc = 758,52 ma jsc (one cell) = 7,74 ma/cm 2 FF =,68 eta act. area = 3,8 %.2-2 -1 1 2 3 4 voltage / V 1.3 Applications of results of this thesis Art deco and logo modules have been produced, which feature an image over the whole module area. The image is attained by structuring the scattering layer of the module. Thus, each cell in the module produces a different current. The resulting I-V curves can be calculated with the developed model for partial shading, since in this case, the individual cells are deliberately electrically mismatched. The resulting module 16

1. Summary and Conclusions efficiencies are only slightly lower than for an non-patterned DSC module (3.1 % compared to 3.8 %). With the aim to promote first commercial applications of DSC modules, this design effect has been demonstrated in various prototypes. It is expected, that first products of DSC modules take advantage of the adaptable screen printed design. Thus, first products of DSC modules do not require to compete with conventional solar cell technologies in terms of power output and price alone. Rather, DSC modules are aimed at providing an additional architectural element for building integrated photovoltaics (BIPV). Unique selling points of a DSC facade in BIPV are colour, transparency and adaptable design. Another possible application are large area advertisement signs, which generate electricity during the daytime that can be used to illuminate the sign at night. Solar advertising signs offer an enormous potential. The skyline in many mega cities with luminous advertising is a huge energy load at night. At the Hanover Trade Fair 26, a semi-transparent, decorative DSC facade has been exhibited (Figure 1-4). Figure 1-4: A semi-transparent, decorative DSC facade has been exhibited at the Hanover Trade Fair Industry 26 by the Fraunhofer ISE. 161

Appendix Appendix 163

A1. Experimental A1. Experimental A1.1. Module manufacturing The flow chart in Figure A illustrates the production process for a module. The pretreatment of the glass substrate includes the laser scribing of the TCO layer, the drilling of the holes and the washing of the glass. The silver paste is commercially available from Ferro (SP 1289). The TiO 2 paste is prepared by dispersing Degussa P9 or Kemira ANX Type N pigments in a terpineol/ethylcellulose mixture using a perl mill. These TiO 2 pigments are not produced by the commonly used sol-gel method. Using these commercial TiO 2 particles for the paste, makes the manufacturing process much simpler particularly for large quantities, but the resulting current densities are slightly lower (about 1-15 %). The ZrO 2 paste is prepared from a precursor consisting of Noramium M2C (Ceca), terpineol and titanium tetraisopropoxid. The paste is then obtained by mixing the precursor, ethylcellulose and terpineol with zirconium dioxide TZ- 3YS from Tosoh-Krahn-Chemie in a Dispermat AE by VMA Getzmann. 165

A1. Experimental Figure A: Production process for a DSC module with glass frit sealing. Platinum paste is obtained by dispersing hexachloroplatinic acid from Aldrich in a terpineol/ethylcellulose mixture using a dissolver Dispermat AE by VMA Getzmann. The glass frit paste is prepared by mixing glass frit powder in a terpineol/ethylcellulose mixture using a dissolver Dispermat AE by VMA Getzmann. All layers are then subsequently screen printed on the glass substrates. The screenprinting of 6 x 1 cm² prototypes has been carried out by the company Bischoff- Glastechnik, Germany. After each screen printing step the layer is dried in a convection furnace at 15 C for about 1 minutes. Then the layers are sintered at 58 C for 1 minutes in a glass fusing oven HRF 6 (KSO Gobi). After sintering the counter electrode is aligned on top of the working electrode and fused at temperatures higher than 6 C in standard atmosphere. A colouring unit has been constructed (Figure B) with an automatic positioning system and temperature control. The module is coloured by injecting the dye solution through one hole into the electrolyte canal and the cell compartments and collecting the dye solution from the second hole for reprocessing. The dye solution is applied at high pressures up to 7 bar by a pneumatic cartridge at room temperature (Figure C). The end of the coloration process is determined by visual inspection. It takes about 2 hours using a dye solution of 1 mm N719 purchased from Solaronix in ethanol or acetonitrile/tertbutanol (1:1). This time corresponds to our experience with the colouring time for 2.5 cm² cells and might be decreased by using higher concentrations of dye solution. After the colouring is complete, the electrolyte is pumped through the module for about 1 minute until the cells are clean from residues of the dye solution. Then the process is stopped and the electrolyte remains in the module. 166

A1. Experimental When using undiluted ionic liquids as electrolyte, the module is heated up to 7 C to reduce the viscosity. Using this technique, filling of 3 x 3 cm² modules can be carried out with high viscosity ionic liquids. Then the cell openings are sealed with a polymer-based hotmelt foil (Surlyn, Dupont). Ideally, the sealing of the holes is also done with glass frit. To seal a hole with glass frit, a small ceramic plug is fused into the holes during the glass fusing process. The ceramic plug features a hole for the colouring and filling. The hole in the ceramic plug is then sealed by glass frit, which is locally applied and melted onto the ceramic. The ceramic acts as a thermally insulating spacer. Figure B: Colouring machine for a 3 x 3 cm² DSC module. Figure C: Colouring process of a meander DSC module. The dye solution is injected with pneumatic cartridges. In case of the strip design, with integrated electrolyte canal, the electrolyte canal is broken off using standard glass cutting techniques. The sealing of the edges is also done with glass frit, which is locally applied and melted onto the cell openings. This procedure, however, is not yet technically mature and fully developed as a hermetic seal is not achieved reproducibly. Other approaches, in particular inorganic glues (cements) have also shown promise according to first helium leak tests. Alternatively, the edges with electrolyte canal may remain on the module and the 2 holes are sealed with Surlyn polymer hot-melt foil (Dupont) and a small glass plate. Such modules with an electrolyte canal can be electrically characterised under illumination since 167

A1. Experimental the photoinduced electrophoresis over the electrolyte canal is a very slow process. But long term testing cannot be conducted on modules with the electrolyte canal intact. A1.2. Test cells: masterplates All test cells in this work were built on so-called masterplates. One masterplate consists of 5 identical cells with areas of 5 x.5 cm². The masterplates were either sealed completely with glass frit, using a ceramic plug to seal the holes. Alternatively, the holes have been sealed with a polymer hotmelt foil (Surlyn) and a thin glass plate. The manufacturing process of a masterplate is entirely equivalent to the manufacturing of a 3 x 3 module. A colouring unit for masterplates can be seen in Figure D [Späth '3]. Figure D: Colouring machine for a masterplate A1.3. Accelerating ageing procedures Continuous strong light soaking has been done with a sulphur plasma lamp. The sulphur lamp spectrum is continuous from 4 to 8 nm with its maximum at 5 nm, approximating quite well the solar spectrum in the spectral sensitivity range of the DSC. The lamp spectrum contains practically no UV light below 4 nm. Silicon reference cells control the lamp intensity. The light intensity was adjusted to approximately 1 W/m² at 35 C. Accelerated ageing at 85 C in the dark was performed in hot-air convection ovens. 168

A1. Experimental Temperature cycling tests were performed at RWE Schott Solar in Alzenau, Germany. A1.4. Electrolyte cells for the characterisation of the platinum electrode Electrolyte cells were built to examine the influence of the lead content in the glass frit on the platinum counter electrode. These electrochemical cells consist of two identical, platinum coated TCO glass substrates in sandwich configuration filled with electrolyte [Hauch '1]. The platinum is applied by screen printing and sintered at 45 C. As primary sealing material lead-containing and lead-free glass frit was used and processed at a temperature of 63 C. Reference cells have been sealed with a polymer-based hotmelt foil (Surlyn, Dupont), which is applied at 13 C. To be able to compare the platinum electrodes of the polymer sealed and glass frit sealed cells, the platinum electrodes of the polymer sealed cells were exposed to the temperatures of the glass fusing process as well. The electrolyte consisted of.6 M hexylmethylimidazolium iodide,.1 M LiI,.5 M I 2,.5 M tert-butylpyridine in acetonitrile The charge transfer resistance of the platinum counter electrode was measured by electrical impedance spectroscopy (EIS) as described in [Hauch '1] and [Kern '2]. All impedance measurements were performed with no bias potential with an amplitude of 5 mv. The EIS measurements were carried out with the Impedance Measuring Unit (IM 6) from Zahner. 169

A2. Approximation of the transient current in photoinduced electrophoresis A2. Approximation of the transient current in photoinduced electrophoresis The calculation of the transient current I(t) response in a thin layer cell bounded by two electrodes acting as working and counter electrode will be shown in this section as demonstrated by [Southampton '85]. The potential of the working electrode is stepped - from equilibrium to that for diffusion controlled reduction of triiodide I3 + 2e 3I. The solution contains an access of iodide, so only the equation for triiodide needs to be solved: n t n 2 I3 I3 = D 2 x A1 As initial condition we assume a constant concentration profile with initial concentration n. I 3 ( x, t = ) = n I3 I 3 n A2 171

A2. Approximation of the transient current in photoinduced electrophoresis 172 By stepping the potential of the working electrode from equilibrium to about 7 mv a diffusion controlled reduction of triiodide starts. This means, at x= the concentration of triiodide is approximately zero: ) (, 3 I = t n A3 And the currents at the electrodes are equal but opposite in sign: d x I x I 3 3 = = = x n x n A4 Laplace transforming equations A1 to A4 we obtain: 2 I 2 I 3 I 3 3 ~ ~ x n D n sn = A5 Here s is the Laplace parameter. The boundary conditions are: ~ 3 I = n at = x A6 and d x I x I 3 3 ~ ~ = = = x n x n A7 Equation A5 has the solution: s n x D s B x D s A n I I3 2 1 2 1 3 exp exp ~ + + = A8 And A and B are determined using equations and. It follows that ( ) s n d D s s x D s n d D s s x d D s n n I I 3 2 1 2 1 I3 2 1 2 1 I3 3 exp 1 exp exp 1 exp ~ + + = A9

A2. Approximation of the transient current in photoinduced electrophoresis 173 + = = d D s d D s s D n x n 2 1 2 1 2 1 2 1 I 3 3 exp 1 exp 1 ~ x I A1 The inverse of this expression does not appear in the table of transforms. For long times, i.e. small s, we can rewrite Equation A1 in the form (i.e. multiplying top and bottom by ( ( ) d D s 2 1 / exp 2 1 ): + = = d D s d D s d D s d D s s D n x n 2 1 2 1 2 1 2 1 2 1 2 1 I 3 3 2 1 exp 2 1 exp 2 1 exp 2 1 exp ~ x I A11 = = d D s d D s s D n x n 2 1 2 1 2 1 2 1 I 3 3 2 1 sinh 2 1 cosh ~ x I A12... 48 1 2 1... 8 1 1 ~ 3 2 x I 2 3 2 1 2 1 2 1 I 3 3 + + + + = = d D s d D s d D s s D n x n A13 + + + = = 2 2 x I 24 1 1 4 24 1 1 2 ~ I3 I3 3 d D s D d n d D s sd n x n A14 Resolving the first term into partial fractions, Equation A14 becomes: + + + = = 2 2 x I 24 1 1 4 24 1 1 12 2 ~ I3 I3 I3 3 d D s D d n d D s D d n sd n x n A15

A2. Approximation of the transient current in photoinduced electrophoresis n ~ I x 3 x= = 2n I3 sd n d I3 + 1 s 6D 1 + 24 D d 2 A16 n ~ I x 3 2n I = 3 sd 4n I + 3 24D d + 2 d x= s A17 And the final expression is readily inverted using the table of transforms. n 2n 4n I 3 = I3 + I3 24D exp t 2 x d d d x= A18 The current is found by multiplying Equation A18, with ν D. It is clear from Equation A18, that the steady state current is the diffusion limited current2ν e D n / d. At t= a high current flows, which then converges to the e I3 diffusion limited current. In the numerical calculations of Chapter 6.1.4 it turned out, that the current in Equation A18 actually not drops fast enough. Negative values for the triiodide concentration occurred in the numerics, because triiodide was consumed too fast for small times t. Since Equation A18 is in itself an approximation, it seems justified and physically reasonable to increase the exponential decay slightly. Thus, the current drops a little bit faster and in the numerical calculation of the concentration profiles all values stay positive. The physical characteristics of the curve remain the same: an exponentially decaying current, which converges against the diffusion limited current. Eventually the following gradient was used as a boundary condition: e e n 2n 4n I 3 = I3 + I3 3 exp t 2 x d d d x= A19 174

A3. List of symbols, physical constants and abbreviations A3. List of symbols, physical constants and abbreviations List of Symbols Symbol Dimension Description Φ incident W/m² incident irradiation β 1 symmetry parameter in Butler Vollmer δ m Nernst diffusion layer η J electrochemical potential η N s m -2 viscosity η % sometimes: efficiency ϕ V electrical potential μ J chemical potential ν ox, ν ox 1 stoichiometric coefficient ρ TCO Ω m² sheet resistance of TCO σ Ω -1 m -1 conductivity a 1 fraction of ohmic current involved in the breakdown A m² area b m² V -1 s -1 mobility c mol/l initial concentration c st mol/l standard concentration c x mol/l concentration of species x 175

A3. List of symbols, physical constants and abbreviations Symbol Dimension Description D m² s -1 diffusion coefficient d m length of electrolyte canal d cell m thickness of the DSC D c ell m² s -1 diffusion constant of triiodide inside the cell E F (n) J (quasi) fermi energy E Redox J redox energy eta % efficiency FF 1 fill factor I A current I cell A current of cell I Diff A diffusion current in photophoresis I lim A diffusion limited I mod A current of module I sat A diode saturation current I sc A short-circuit current j A/m² current density j A/m² exchange current density j lim A/m² diffusion limited current density j SC A/m² short-circuit current density k deg 1/s rate of degradation of triiodide L m length of one cell m 1 number of transferred electrons m a 1 avalanche breakdown coefficient. m D 1 diode factor M dye g/mol molar weight of dye MPP W maximum power point n m -3 density of a species N C m -3 density of states in the conduction band n dye g/m² density of dye on area of TiO 2 n x m -3 density of a species x P W power P max W maximum power point R Ω resistance r m radius r Ω/m specific resistance R contact Ω m² contact resistance r contact Ω m² contact resistance R CT Ω m² charge transfer resistance R s Ω series resistance R Sh Ω shunt resistance R sheet Ω sheet resistance R turnover 1/s turnover rate of dye T K temperature t s time V V voltage v (x ) V voltage at point x in TCO V br V breakdown voltage 176

A3. List of symbols, physical constants and abbreviations Symbol Dimension Description V cell V voltage of cell V mod V voltage of module V OC V open-circuit voltage w a m width of the active area w contact m width of the contact area w d m distance from end of active area until series contact x m coordinate z 1 number of charges (-1 for electrons) Physical constants Symbol Dimension Descripton e 1.6 1-19 As elementary charge k B 1.38 1-23 J/K Boltzmann constant N A 6.22 1 23 1/mol Avogadro constant π 3.14159 Pi Abbreviations Abbreviation ACN Ag AM a-si BIPV CdS CdTe CIS D DSC EIS El EVA FTO GaAs HMII HOMO I - I 2 Description acetonitrile silver atmosphere mass amorphous silicon building integrated photovoltaic cadmium sulfide cadmium telluride copper indium diselenide dye molecule dye solar cell electrical impedance spectroscopy electrolyte ethylene vinyl acetate flourine doped tin oxide gallium arsenide hexyl methyl immidazolium iodide highest occupied molecular orbital iodide iodine 177

A3. List of symbols, physical constants and abbreviations Abbreviation - I 3 - IO 3 ITO Li LUMO MACN Mo N719 NMBI OSC Ox PbO PDP PMII Pt PV R&D Red Ru SEM SnO 2 :F SRC SRPI TBP TCO TiO 2 UV Z97 ZrO 2 Description triiodide iodate indium doped tin oxide lithium lowest unoccupied molecular orbital methoxyacetonitrile molybdenum most common dye molecule used in DSC n-methylbenzimidazole organic solar cell oxidized species lead oxide plasma display panel propyl methyl immidazolium iodide platinum photovoltaic research and development reduced species ruthenium scanning electron microscopy flourine doped tin oxide standard reporting conditions spatially resolved photocurrent image tert-butylpyridine transparent conductive oxide titanium dioxide ultra violet hydrophobic dye zirconium dioxide 178

A4.References A4. References [Atkins '9] [Bishop '88] [Bishop '89] P. W. Atkins (199). Physikalische Chemie, VCH Verlag. J. W. Bishop. Computer Simulation of the Effects of Electrical Mismatches in Photovoltaic Cell Interconnection Circuits. Solar Cells 1988 25: 73-89. J. W. Bishop. Microplasma Breakdown and Hot-Spots in Silicon Solar-Cells. Solar Cells 1989 26(4): 335-349. [Boeuf '3] J. Boeuf. Plasma display panels: physics, recent developments and key issues. J. Phys. D: Appl. Phys. 23 36: R53 R79. [Bowermann '1] B. Bowermann,V. Fthenakis. EH&S analysis of dye-sensitized phtovoltaic solar cell production. Brookhaven National Laboratory (21), Report DE-AC2-98CH1886 [Brandt '6] [Brecl '5] H. Brandt. Einfluss variabler Designeffekte auf den Wirkungsgrad von seriell verschalteten Farbstoffsolarmodulen. Umwelttechnik / Regenerative Energien, Fachhochschule für Technik und Wirtschaft Berlin, Diploma Thesis 26 K. Brecl, M. Topic,F. Smole. A Detailed Study of Monolithic Contacts and Electrical Losses in a Large-area Thin-film Module. Progress in Photovoltaics: Res. Appl. 25 13: 297 31. [Burgelman '98] M. Burgelman,A. Niemegeers. Calculation of CIS and CdTe module efficiencies. Solar Energy Materials and Solar Cells 1998 51(2): 129-143. [Burgelman '99] M. Burgelman,A. Niemegeers. Influence of illumination conditions on the design of thin-film modules. Solar Energy Materials and Solar Cells 1999 57(1): 85-95. [Cameron '5] P. Cameron, L. Peter,S. Hore. How Important is the Back Reaction of Electrons via the Substrate in Dye-Sensitized Nanocrystalline Solar Cells? J. Phys. Chem. B 25 19: 93-936. 179

A4.References [Cameron '3] P. J. Cameron,L. M. Peter. Characterization of titanium dioxide blocking layers in dye-sensitized nanocrystalline solar cells. J. Phys. Chem. B 23 17(51): 14394-144. [Chiba '5] Y. Chiba, A. Islam, K. Kakutani, R. Komiya, N. Koide,L. Han. High Efficiency of Dye- Sensitized Solar Cells. 15th International Photovoltaic Science & Engineering Conference, Shanghai, China 25. [Dai '5] S. Dai, K. Wang, J. Weng, Y. Sui, Y. Huang, S. Xiao, S. Chen, L. Hu, F. Kong, X. Pan, C. Shi,L. Guo. Design of DSC panel with efficiency more than 6%. Solar Energy Materials and Solar Cells 25 85(3): 447-455. [Dai '4] S. Dai, J. Weng, Y. F. Sui, C. W. Shi, Y. Huang, S. H. Chen, X. Pan, X. Q. Fang, L. H. Hu, F. T. Kong,K. J. Wang. Dye-sensitized solar cells, from cell to module. Solar Energy Materials and Solar Cells 24 84(1-4): 125-133. [de Vos '84] A. de Vos. The distributed series resistance problem in solar cells. Solar Cells 1984 12: 311-327. [Dloczik '97] L. Dloczik, O. Ileperuma, I. Lauermann, L. M. Peter, E. A. Ponomarev, G. Redmond, N. J. Shaw,I. Uhlendorf. Dynamic response of dye-sensitized nanocrystalline solar cells: Characterization by intensity-modulated photocurrent spectroscopy. J. Phys. Chem. B 1997 11(49): 1281-1289. [Dunlop '6] [Dürr '4] [Ferber '1] [Ferber '98] E. Dunlop,D. Halton. The Performance of Crystalline Silicon Photovoltaic Solar Modules after 22 Years of Continuous Outdoor Exposure. Progress in Photovoltaics: Res. Appl. 26 14: 53 64. M. Dürr, A. Bamedi, A. Yasuda,G. Nelles. Tandem dye-sensitized solar cell for improved power conversion efficiencies. Applied Physics Letters 24 84(17): 3397-3399. J. Ferber,J. Luther. Modeling of photovoltage and photocurrent in dye-sensitized titanium dioxide solar cells. J. Phys. Chem. B 21 15(21): 4895-493. J. Ferber, R. Stangl,J. Luther. An electrical model of the dye-sensitized solar cell. Solar Energy Materials and Solar Cells 1998 53(1-2): 29-54. [Gerthsen '3] C. Gerthsen,D. Meschede (23). Physik. Berlin, Springer. [Goetzberger '94] A. Goetzberger, B. Voß,J. Knoobloch (1994). Sonnenenergie: Photovoltaik, Teubner Studienbücher. [Grätzel '] [Grätzel '3] [Grätzel '5] [Greijer '1] [Hagfeldt '95] [Halme '2] [Hanke '99] M. Grätzel. Perspectives for dye-sensitized nanocrystalline solar cells. Progress in Photovoltaics 2 8(1): 171-185. M. Grätzel. Dye-sensitized solar cells. Journal of Photochemistry and Photobiology C-Photochemistry Reviews 23 4(2): 145-153. M. Grätzel. Solar energy conversion by dye-sensitized photovoltaic cells. Inorganic Chemistry 25 44(2): 6841-6851. H. Greijer, L. Karlson, S. E. Lindquist,A. Hagfeldt. Environmental aspects of electricity generation from a nanocrystalline dye sensitized solar cell system. Renewable Energy 21 23(1): 27-39. A. Hagfeldt,M. Grätzel. Light-Induced Redox Reactions in Nanocrystalline Systems. Chemical Reviews 1995 95(1): 49-68. J. Halme. Dye-sensitized nanostructured and organic photovoltaic cells: technical review and preliminary tests. Department of engineering, physics and mathematics, Helsinki University of Technology, Master Thesis 22 K. P. Hanke. Upscaling of the dye sensitized solar cell. Zeitschrift Fur Physikalische Chemie-International Journal of Research in Physical Chemistry & Chemical Physics 1999 212: 1-9. 18

A4.References [Hauch '98] A. Hauch. Diffusion und Grenzflächen an der Platinelektrode in farbstoffsensibilisierten Solarzellen. Fakultät für Physik, Albert-Ludwigs-Universität, Diploma Thesis 1998 [Hauch '1] [Hinsch '1] [Hore '5] [Hore '6] [Ibach '95] [Kang '3] [Kawano '4] [Kay '94] [Kay '96] [Kern '1] [Kern '2] [Kohle '97] [Koide '6] [Kovach '95] A. Hauch,A. Georg. Diffusion in the electrolyte and charge-transfer reaction at the platinum electrode in dye-sensitized solar cells. Electrochimica Acta 21 46(22): 3457-3466. A. Hinsch, J. M. Kroon, R. Kern, I. Uhlendorf, J. Holzbock, A. Meyer,J. Ferber. Longterm stability of dye-sensitised solar cells. Progress in Photovoltaics 21 9(6): 425-438. S. Hore,R. Kern. Implication of device functioning due to back reaction of electrons via the conducting glass substrate in dye sensitized solar cells. Applied Physics Letters 25 87: 26354. S. Hore, C. Vetter, R. Kern, H. Smit,A. Hinsch. Influence of scattering layers on efficiency of dye-sensitized solar cells. Solar Energy Materials and Solar Cells 26 9(9): 1176-1188. H. Ibach,H. Lüth (1995). Festkörperphysik, Einführung in Grundlagen, Springer Verlag. M. G. Kang, N. G. Park, Y. J. Park, K. S. Ryu,S. H. Chang. Manufacturing method for transparent electric windows using dye-sensitized TiO2 solar cells. Solar Energy Materials and Solar Cells 23 75(3-4): 475-479. R. Kawano, H. Matsui, C. Matsuyama, A. Sato, M. Susan, N. Tanabe,M. Watanabe. High performance dye-sensitized solar cells using ionic liquids as their electrolytes. Journal of Photochemistry and Photobiology a-chemistry 24 164(1-3): 87-92. A. Kay. Solar cells based on dye-sensitized nanocrystalline TiO 2 electrodes. Departement de Chimie, Ecole Polytechnique federale de Lausanne, Doctoral Dissertation 1994 A. Kay,M. Gratzel. Low cost photovoltaic modules based on dye sensitized nanocrystalline titanium dioxide and carbon powder. Solar Energy Materials and Solar Cells 1996 44(1): 99-117. R. Kern. Untersuchungen zur Langzeitstabilität von Farbstoffsolarzellen mittels optischer und elektrischer Impedanzspektroskopie. Fakultät für Physik, Albert- Ludwigs Universität, Doctoral Dissertation 21 R. Kern, R. Sastrawan, J. Ferber, R. Stangl,J. Luther. Modeling and interpretation of electrical impedance spectra of dye solar cells operated under open-circuit conditions. Electrochimica Acta 22 47(26): 4213-4225. O. Kohle, M. Gratzel, A. F. Meyer,T. B. Meyer. The photovoltaic stability of bis(isothiocyanato)ruthenium(ii)-bis-2,2'-bipyridine-4,4'-dicarboxylic acid and related sensitizers. Advanced Materials 1997 9(11): 94-&. N. Koide, Liyuan Han, Atsushi Fukui, Nobuhiro Fuke,R. Yamanaka. High Efficiency of Dye-Sensitized Solar Cell and Module. IEEE 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii 26. A. Kovach (1995). Effect of Partial Shading on the Ènergy Performance of Photovoltaic Arrays Integrated onto Buildings, VDI-Verlag GmBH. [Kroon '1] J. M. Kroon, A. Hinsch, J.A.M. van Roosmalen, N.P.G. van der Burg, N.J. Bakker, R. Kinderman, P.M. Sommeling, M. Späth, R. Kern, R. Sastrawan, J. Ferber, M. Schubert, G. Hasenhindl, C. Schill, M. Lorenz, R. Stangl, S. Baumgärtner, C. Peter, A. Meyer, T. Meyer, I. Uhlendorf, J. Holzbock,R. Niepmann. Long Term Stability of Dye Sensitised Solar Cells for large area power applications. THE EUROPEAN COMMISSION PUBLISHABLE REPORT 21: JOR3-CT98-261. 181

[Kumara '6] [Lee '6] 182 A4.References G. R. A. Kumara, S. Kaneko, A. Konno, M. Okuya, K. Murakami, B. Onwona- Agyeman,K. Tennakone. Large Area Dye-sensitized Solar Cells: Material Aspects of Fabrication. Progress in Photovoltaics: Res. Appl. 26: DOI: 1.12/pip.695. W. J. Lee, E. Ramasamy, D. Y. Lee,J. S. Song. Glass frit overcoated silver grid lines for nano-crystalline dye-sensitized solar cells. Journal of Photochemistry and Photobiology A: Chemistry 26: doi:1.116/j.jphotochem.26.3.6. [Lewerenz '95] H. J. Lewerenz,H. Jungblut (1995). Photovoltaik, Springer Verlag. [Luther '5] J. Luther. Photovoltaic - Energielieferant der Zukunft. IHKS - Fachjournal 25: 14-2. [Macht '2] [Maine '5] B. Macht, M. Turrion, A. Barkschat, P. Salvador, K. Ellmer,H. Tributsch. Patterns of efficiency and degradation in dye sensitization solar cells measured with imaging techniques. Solar Energy Materials and Solar Cells 22 73: 163-173. J. A. Maine, G. Phani, J. M. Bell,I. L. Skryabin. Minimisation of the cost of generated electricity from dye-sensitised solar cells using numerical analysis. Solar Energy Materials and Solar Cells 25 87(1-4): 133-148. [McConnell '2] R. D. McConnell. Assessment of the dye-sensitized solar cell. Renewable & Sustainable Energy Reviews 22 6(3): 273-295. [Nazeeruddin '93] M. K. Nazeeruddin, A. Kay, I. Rodicio, R. Humphrybaker, E. Muller, P. Liska, N. Vlachopoulos,M. Gratzel. Conversion of Light to Electricity by Cis- X2bis(2,2'-Bipyridyl-4,4'-Dicarboxylate)Ruthenium(Ii) Charge-Transfer Sensitizers (X = Cl-, Br-, I-, Cn-, and Scn-) on Nanocrystalline Tio2 Electrodes. Journal of the American Chemical Society 1993 115(14): 6382-639. [Nelson '99] [Nielson '82] J. Nelson. Continuous-time random-walk model of electron transport in nanocrystalline TiO2 electrodes. Physical Review B 1999 59(23): 15374-1538. L. Nielson. Distributed series resistance effects in solar cells. IEEE Transactions on electron devices 1982 29(5): 821-826. [O'Regan '91] B. O'Regan,M. Gratzel. A Low-Cost, High-Efficiency Solar-Cell Based On Dye- Sensitized Colloidal TiO 2 Films. Nature 1991 353(6346): 737-74. [Okada '4] K. Okada, H. Matsui, T. Kawashima, T. Ezure,N. Tanabe. 1 mm x 1 mm largesized dye sensitized solar cells. Journal of Photochemistry and Photobiology A: Chemistry 24 164(1-3): 193-198. [Papageorgiou '96a] N. Papageorgiou, Y. Athanassov, M. Armand, P. Bonhote, H. Pettersson, A. Azam,M. Gratzel. The performance and stability of ambient temperature molten salts for solar cell applications. Journal of the Electrochemical Society 1996a 143(1): 399-318. [Papageorgiou '96b] N. Papageorgiou, M. Gratzel,P. P. Infelta. On the relevance of mass transport in thin layer nanocrystalline photoelectrochemical solar cells. Solar Energy Materials and Solar Cells 1996b 44(4): 45-438. [Peter '] L. M. Peter,K. G. U. Wijayantha. Electron transport and back reaction in dye sensitised nanocrystalline photovoltaic cells. Electrochimica Acta 2 45(28): 4543-4551. [Pettersson '1] H. Pettersson,T. Gruszecki. Long-term stability of low-power dye-sensitised solar cells prepared by industrial methods. Solar Energy Materials and Solar Cells 21 7(2): 23-212. [Pettersson '3] H. Pettersson, T. Gruszecki, L. H. Johansson,P. Johander. Manufacturing method for monolithic dye-sensitised solar cells permitting long-term stable low-power modules. Solar Energy Materials and Solar Cells 23 77(4): 45-413. [Photon '6] Fraunhofer Institute searches for industrial partner for DSC. Photon International, the photovoltaic magazine, May 26, 17

A4.References [Quaschning '96a] V. Quaschning (1996a). Simulation der Abschattungsverluste bei solarelektrischen Systemen. Berlin, Verlag Dr. Köster. [Quaschning '96b] V. Quaschning,R. Hanitsch. Numerical simulation of current-voltage characteristics of photovoltaic systems with shaded solar cells. Solar Energy 1996b 56(6): 513-52. [Randall '3] [Rau '5] J. F. Randall,J. Jacot. Is AM1.5 applicable in practice? Modelling eight photovoltaic materials with respect to light intensity and two spectra. Renewable Energy 23 28(12): 1851-1864. U. Rau. (25) Transport analysis of dye sensitized solar cells. Bad Wildbad: Presentation of measurements at project meeting [Robinson '59] R. Robinson,R. Stokes (1959). Electrolyte Solutions. London, Butterworths. [Scheer '99] H. Scheer (1999). Solare Weltwirtschaft. Strategie für die ökologische Moderne München, Verlag Antje Kunstmann. [Sommeling '4] P. M. Sommeling, M. Spath, H. J. P. Smit, N. J. Bakker,J. M. Kroon. Long-term stability testing of dye-sensitized solar cells. Journal of Photochemistry and Photobiology a-chemistry 24 164(1-3): 137-144. [Southampton '85] Southampton, Electrochemistry,Group (1985). Instrumental Methods in Electrochemistry, John Wiley & Sons. [Späth '3] M. Späth, P. M. Sommeling, J. A. M. van Roosmalen, H. J. P. Smit, N. P. G. van der Burg, D. R. Mahieu, N. J. Bakker,J. M. Kroon. Reproducible manufacturing of dyesensitized solar cells on a semi-automated baseline. Progress in Photovoltaics 23 11(3): 27-22. [Stangl '98] R. Stangl, J. Ferber,J. Luther. On the modeling of the dye-sensitized solar cell. Solar Energy Materials and Solar Cells 1998 54(1-4): 255-264. [Toyoda '4] T. Toyoda, T. Sano, J. Nakajima, S. Doi, S. Fukumoto, A. Ito, T. Tohyama, M. Yoshida, T. Kanagawa, T. Motohiro, T. Shiga, K. Higuchi, K. Tanaka, Y. Takeda, T. Fukano, N. Katoh, A. Takeichi, K. Takechi,M. Shiozawa. Outdoor performance of large scale DSC modules. Journal of Photochemistry and Photobiology A: Chemistry 24 164(1-3): 23-27. [Tributsch '4] H. Tributsch. Dye sensitization solar cells: a critical assessment of the learning curve. Coordination Chemistry Reviews 24 248(13-14): 1511-153. [Tulloch '4] [Vetter '61] G. E. Tulloch. Light and energy - dye solar cells for the 21st century. Journal of Photochemistry and Photobiology A:Chemistry 24 164(1-3): 29-219. K. J. Vetter (1961). Elektrochemische Kinetik, Springer Verlag. [Wang '5] P. Wang, C. Klein, R. Humphry-Baker, M. Zakeeruddin,M. Grätzel. Stable 8% efficient nanocrystalline dye-sensitized solar cell based on an electrolyte of low volatility. Applied Physics Letters 25 86: 12358. [Wang '4] P. Wang, S. M. Zakeeruddin, J. E. Moser, R. Humphry-Baker, P. Comte, V. Aranyos, A. Hagfeldt, M. K. Nazeeruddin,M. Gratzel. Stable new sensitizer with improved light harvesting for nanocrystalline dye-sensitized solar cells. Advanced Materials 24 16(2): 186-+. [Wheatley '6] M. Wheatley, D. Snowdon,M. Brungs. On the effects of highly capacitive cells on maximum point power trackers. Industrialisation of dye solar cells from research to product, Canberra, Australia 26. [Wheatley '3] M. G. Wheatley, A. M. McDonagh, M. P. Brungs, R. P. Chaplin,E. Sizgek. A study of reverse bias in a dye sensitised photoelectrochemical device. Sol. Energy Mat. Sol. Cells 23 76(2): 175-181. [Würfel '95] P. Würfel (1995). Physik der Solarzellen. Heidelberg, Spektrum, Akad. Verl. 183

[Würfel '6] A4.References U. Würfel. Untersuchung zum Elektronentransport im nanoporösen TiO 2 von Farbstoffsolarzellen. Fakultät für Physik, Albert-Ludwigs Universität, Doctoral Dissertation 26 [Yanagida '5] S. Yanagida, M. Watanabe, H. Matsui, K. Okada, H. Usui, T. Ezure,N. Tanabe. Dyesensitzed solar cells using nanocomposite ion-gel. Fujikura Technical Review 25: www.fujikura.co.jp. [Ziegler '5] T. Ziegler. Photophorese in seriell verschalteten Farbstoffsolarmodulen. Umwelttechnik / Regenerative Energien, Fachhochschule für Technik und Wirtschaft Berlin, Diploma Thesis 25 [Zistler '6] M. Zistler, P. Wachter, P. Wasserscheid, D. Gerhard, A. Hinsch, R. Sastrawan,H. J. Gores. Comparison of Electrochemical Methods for Triiodide Diffusion Coefficient Measurements and Observation of Non-Stokesian Diffusion Behaviour in Binary Mixtures of two Ionic Liquids. Electrochimica Acta 26: in press. 184

A5. Publications A5. Publications A5.1. Publications in reviewed journals Comparison of electrochemical methods for triiodide diffusion coefficient measurements and observation of non-stokesian diffusion behaviour in binary mixtures of two ionic liquids M. Zistler, P. Wachter, P. Wasserscheid, D. Gerhard, A. Hinsch, R. Sastrawan and H. J. Gores Electrochimica Acta, in press 26 Production and characterization of ITO-Pt semiconductor powder containing nanoscale noble metal particles catalytically active in dye-sensitized solar cells S. Katusic, P. Albers, R. Kern, F. M. Petrat, R. Sastrawan, S. Hore, A. Hinsch and A. Gutsch, Solar Energy Materials and Solar Cells, Volume 9, Issue 13, 15 August 26, Pages 1983-1999 New Interdigital Design for Large Area Dye Solar Modules Using a Lead-Free Glass Frit Sealing R. Sastrawan, J. Beier, U. Belledin, S. Hemming, A. Hinsch, R. Kern, C. Vetter, F. M. Petrat, A. Prodi- Schwab, P. Lechner and W. Hoffmann Progress in Photovoltaics: Research and Applications, Editorial ref. code: PIP/S 1862, in press 26 185

A5. Publications Nanocrystalline dye-sensitized solar cells having maximum performance J. M. Kroon, N. J. Bakker, H. J. P Smit, P. Liska, K. R. Thampi, M. Grätzel, A. Hinsch, U. Würfel, R. Sastrawan, S. Hore, J. R. Durrant, E. Palomares, H. Pettersson, T. Gruszecki, J. Walter, K. Skupien, G. Tulloch Progress in Photovoltaics: Research and Applications, in press 26 A glass frit-sealed dye solar cell module with integrated series connections R. Sastrawan, J. Beier, U. Belledin, S. Hemming, A. Hinsch, R. Kern, C. Vetter, F.M. Petrat, A. Prodi- Schwab, P. Lechner and W. Hoffmann Solar Energy Materials and Solar Cells, Volume 9, Issue 11, 6 July 26, Pages 168-1691 Interconnecting dye solar cells in modules I V characteristics under reverse bias R. Sastrawan, J. Renz, C. Prahl, J. Beier, A. Hinsch and R. Kern Journal of Photochemistry and Photobiology A: Chemistry, Volume 178, Issue 1, 2 February 26, Pages 33-4 Modelling and interpretation of electrical impedance spectra of dye solar cells operated under open-circuit conditions R. Kern, R. Sastrawan, J. Ferber, R. Stangl and J. Luther Electrochimica Acta, Volume 47, Issue 26, 9 October 22, Pages 4213-4225 A5.2. Conference proceedings, oral presentations, poster presentations Developments for Dye Solar Modules: Results from an Integrated Approach A. Hinsch, H. Bönnemann, A. Drewitz, F. Einsele, D. Fassler, D. Gerhard, H. Gores, S. Himmler, G. Kheshliavilli, D. Koch, G. Nasmudinova, U. Opara-Kras, M. Peters, P. Putira, U. Rau, R. Sastrawan, T. Schauer, S. Sensfuß, C. Siegers, K. Skupien, J. Walter, P. Wasserscheid, U. Würfel, M. Zistler Proceedings, 21st European Photovoltaic Solar Energy Conference and Exhibition, Dresden, Germany, September 26, oral presentation (to be published 26) Ionic liquid based electrolyte solidified with SiO2 nanoparticles for Dye Sensitized Solar Cells M. Berginc, U. Opara Krašovec, A. Hinsch, R. Sastrawan, M. Topič Proceedings, 6th International Conference on Coatings on Glass and Plastics ICCG, Berlin, Germany, June 26 Glass frit sealed dye solar modules with adaptable screen printed design A. Hinsch, U. Belledin, H. Brandt, F. Einsele, S. Hemming, D. Koch, U. Rau, R. Sastrawan, T. Schauer Proceedings, 4th IEEE World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, May 26, oral presentation 186

A5. Publications Quasi-solid state polymer electrolytes for dye-sensitized solar cells: effect of the electrolyte components variation on the triiodide ion diffusion properties and charge-transport resistance at platinum electrode G. Nazmutdinova, S. Sensfuss, M. Schrödner, A. Hinsch, R. Sastrawan, D. Gerhard, S. Himmler, P. Wasserscheid Proceedings, E-MRS IUMPS ISEM 26, Nice, France, May 26, poster presentation Physical Characterization and Electrochemical Study of the Pt:SnO 2 Powder Electrocatalyst for Printable Dye Solar Cells G. Khelashvili, S. Behrens, A. Hinsch, W. Habicht, D. Schild, A. Eichhoefer, R. Sastrawan, H. Bönnemann Proceedings, E-MRS IUMPS ISEM 26, Nice, France, May 26, oral presentation Studies on PVDF - HFP based gel polymer electrolytes for dye-sensitized solar cells application G. Nazmutdinova*, S. Sensfuss, M. Schroedner, A. Hinsch, R. Sastrawan, D. Gerhard, S. Himmler, P. Wasserscheid Technologies for Polymer Electronics - TPE 6, Rudolstadt, Germany, May 26, poster presentation Efficiency limitations in dye-sensitized solar cells with ionic liquids F. Einsele, M. Hlusiak, U. Rau, R. Sastrawan, R. Kern, A. Hinsch DPG Tagung, Dresden, Germany, March 26, oral presentation Upscaling dye solar modules with durable glass frit sealing R. Sastrawan, A. Hinsch, J. Beier, U. Belledin, S. Hemming, R. Kern, C. Vetter Proceedings, 15th International Photovoltaic Science and Engineering Conf Conference and Solar Energy Exhibition, Shanghai, China, October 25, oral presentation Towards Manufacturing Dye Solar Cells R. Sastrawan, A. Hinsch, J. Beier, U. Belledin, S. Hemming, S. Hore, R. Kern, C. Prahl, C. Vetter, U. Würfel Proceedings, 2th European Photovoltaic Solar Energy Conference and Exhibition, Barcelona, Spain, June 25, oral presentation Interconnecting Dye Solar Cells in Modules I-V Characteristics under Reverse Bias R. Sastrawan, J. Renz, C. Prahl, J. Beier, R. Kern, A. Hinsch Proceedings, 2th European Photovoltaic Solar Energy Conference and Exhibition, Barcelona, Spain, June 25, poster presentation Module and design related aspects of dye solar cells A. Hinsch, J. Beier, R. Kern, C. Vetter, U. Belledin, S. Hemming, R. Sastrawan Photovoltaics of the Future: Empa, Switzerland, April 25, poster presentation Dependence of spectral response of Dye Solar Cells on bias light illumination J. Hohl-Ebinger, A. Hinsch, R. Sastrawan, W. Warta, U. Würfel Proceedings, 19th European Photovoltaic Solar Energy Conference, Paris, France, June 24, poster presentation 187

A5. Publications The Wrap Through Electrode Concept for Organic Solar Cells M. Glatthaar, A. Hinsch, J. Luther, M. Niggemann, M. Riede, R. Sastrawan, J. Wagner, B. Zimmermann EuroSun 24, Freiburg, Germany, June 24, poster presentation Electrochemical And Optical Impedance Spectroscopy On Dye Solar Cells R. Sastrawan, R. Kern, A. Hinsch, J. Ferber Extended abstracts, International Conference Spectroelectrochemistry of conducting polymers, Moscow, Russia, October 22, poster presentation Long-term stability and efficiency of dye-sensitized solar cells A. Hinsch, J. M. Kroon, R. Kern, R. Sastrawan, A. Meyer, I. Uhlendorf Proceedings: 17 th European Photovoltaic Solar Conference and Exhibition, Munich, Germany, October 21, oral presentation Influence of TiO 2 properties and tert-butylpyridine addition on electron lifetime and diffusion coefficients in dye sensitized solar cells S. Baumgärtner, R. Sastrawan, M. Schubert, J. Ferber, J. Luther 17 th European Photovoltaic Solar Conference and Exhibition, Munich, Germany, October 21, poster presentation Long-term stability of dye-sensitized solar cells R. Kern, J. Ferber, A. Hinsch, J. M. Kroon, J. Luther, A. Meyer, R. Sastrawan Proceedings, E-MRS 21 Spring Meeting, Strasbourg, France, June 21, oral presentation Electrical and optical impedance spectroscopy on dye-sensitized solar cells R. Sastrawan, R. Kern, J. Ferber, J. Luther E-MRS 21 Spring Meeting, Strasbourg, France, June 21, poster presentation Investigation of the long-term stability of dye-sensitized solar cells by optical and electrochemical impedance spectroscopy R. Kern, J. Ferber, A. Hinsch, J. M. Kroon, J. Luther, A. Meyer, R. Sastrawan, I. Uhlendorf Proceedings, 13 th Workshop on quantum solar energy conversion, Kirchberg, Tyrol, Austria, March 21, oral presentation Untersuchungen zur Langzeitstabilität von farbstoffsensiblisierten TiO 2 -Solarzellen J. Ferber, R. Kern, R. Sastrawan, C. Schill, M. Schubert, A. Hinsch, J. Kroon, M. Späth, J. Roosmalen, N. J. Bakker, P. Sommeling, N. Burg, R. Kinderman, A. Meyer, T. Meyer, I. Uhlendorf, J. Holzbock, R. Niepmann 12. Internationales Sonnenforum, Freiburg, Germany, July 2, poster presentation 188

A5. Publications A5.3. Patents Photo-electrochemical solar cell module WO 25/96391 A2, October 25 Filing persons: Andreas Hinsch, Udo Belledin, Ronald Sastrawan, Andreas Georg Dye-sensitized photovoltaic cells, method for the production of said photovoltaic cells and use thereof WO 24/38745 A3, May 24 Filing persons: Frank-Martin Petrat, Bernhard Stützel, Friedrich Georg Schmidt, Günther Michael, Andreas Gutsch, Stipan Katusic, Andreas Hinsch, Rainer Kern, Ronald Sastrawan 189

A6. Acknowledgements A6. Acknowledgements The successful completion of this work would not have been possible without the support of mentors, colleagues, friends and family to whom I am very thankful. In particular, I would like to thank: Prof. Joachim Luther, my doctoral supervisor, for his guidance and support. Dr. Andreas Hinsch, for the excellent supervision of this work, his expertise and enthusiasm in the field of the DSC and the great collaboration. PD Dr. habil. Andreas Gombert, for his support. Dr. Rainer Kern, for the excellent collaboration and his support. Tina Ziegler and Henning Brandt, the team from FHTW Berlin, for their excellent collaboration and dedication during their diploma thesis at Fraunhofer ISE and for their friendship. Simon Hemming, for his wonderful colouring system, his support and his friendship. Dr. Jutta Beier, for the excellent collaboration in the development of DSC modules. Carmen Vetter, for the excellent collaboration in the long term stability tests. Udo Belledin, for the excellent collaboration and his support. 191

A6. Acknowledgements Uli Würfel, for the lively discussions, his blocking layers and his friendship. Marius Peters, for his never-ending enthusiasm in building masterplates in his diploma thesis. Markus Glatthaar, for the lively discussions, his guidance in numerical methods and his friendship. Conrad Siegers, for his expertise in Chemistry. Melanie Schumann, for the SEM measurements. Vera Walliser for her support. All members of the Organic and Dye Solar Cells team at Freiburg Materials Research Centre, especially Birger Zimmermann, Audrey Dobbins, Christof Prahl, Moritz Riede, Dr. Jochen Wagner, Dr. Michael Niggemann, Dr. Anneke Georg and Dr. Simone Baumgärtner. Krzysztof Skupien, for his first-rate platinum pastes. Dennis Koch, from FPL Stuttgart for his beautiful coloured glass frit. PD Dr. habil. Uwe Rau, IPE Stuttgart for the lively discussions about electrolyte diffusion. Dr. Bernhard Stützel, Dr. Frank-Martin Petrat and Dr. Anna Prodi-Schwab, from Creavis for the collaboration and their support. Dr. Winfried Hoffmann, Dr. Peter Lechner, from Schott Solar for the collaboration and their support. Mathias Klockhaus, for the SRPI measurements. All partners from the Network DSC, especially Prof. Peter Wasserscheid and PD. Dr. habil. Heiner Gores and Dr. Matthias Künzel. My father Dr. Budhi Sastrawan, my mother Marianne Klute-Sastrawan and my sister Rita Sastrawan, for their support. Corinna Strauss, for the great time together, the upcoming world-trip with her and just everything. 192