Characterization, modelling and optimization of industrial silicon thin film solar cells

UNIVERSIDADE DE SANTIAGO DE COMPOSTELA Departamento de Electrónica e Computación Doctoral thesis CHARACTERIZATION, MODELLING AND OPTIMIZATION OF INDUSTRIAL SILICON THIN FILM SOLAR CELLS Submitted by: José Antonio Rodríguez González Supervised by: Dr. Antonio J. García Loureiro Dr. Ing. Michael Vetter Santiago de Compostela, September 2013 Dr. Antonio J. García Loureiro, Profesor Titular del Área de Electrónica de la Universidade de Santiago de Compostela Dr. Ing. Michael Vetter, Director del Laboratorio de I+D de la empresa T-Solar Global S.A. HACEN CONSTAR: Que la memoria titulada CHARACTERIZATION, MODELLING AND OPTIMIZATION OF INDUSTRIAL SILICON THIN FILM SOLAR CELLS ha sido realizada por D. José Antonio Rodríguez González bajo nuestra dirección en el Departamento de Electrónica e Computación de la Universidade de Santiago de Compostela, y constituye la Tesis que presenta para optar al título de Doctor. Santiago de Compostela, septiembre de 2013 Dr. Antonio J. García Loureiro Codirector de la tesis Dr. Ing. Michael Vetter Codirector de la tesis José Antonio Rodríguez González Autor de la tesis A mis padres No podemos resolver problemas pensando de la misma manera que cuando los creamos. Albert Einstein Casi todo lo que realice será insignificante, pero es muy importante que lo haga. Mahatma Gandhi Acknowledgments Fist of all, I want to kindly thank my PhD supervisors, Michael Vetter and Antonio García Loureiro, for their constant help, support and encouragement, which have allowed me to carry out this PhD. As well, for their ability to successfully manage their respective teams and their motivation and passion showed at work. I would also thank Jordi Andreu for his confidence. To my T-Solar workmates from all the departments but especially from the Laboratory: Carlos, Jacinto, Florent, José Manuel, Miguel, Óscar, Boris, Teresa and Luis. The working environment was really nice and to work with you a rewarding experience. I want to also thank you for the good moments in Ourense playing football, tennis, going to the swimming pool or making party at "Vinos". Particularly, I want to thank to Eva, Montse and Amaury, to work in group with you has been very easy and your motivation a pleasure. To my CITIUS workmates, where I have performed the simulations and wrote this memory, especially to Enrique for his help with the simulations and computer problems. I cannot forget Pablo, which has been my "brother" in Ourense during the last four years. We form a nice "tandem" and I hope it is going to continue in the future with our jump to the business world. To my parents and my brother Santi for their constant support, help and their sacrifice to let me study whatever I wanted and to let me participate in the Erasmus internship program. To my friends for being always there and for so many unforgettable moments. Finally, I acknowledge the funding of the Spanish Ministry of Economics and Competitiveness in the frame of the program "Torres Quevedo" (Contract No. PTQ-10-03524) and under project TEC2010-17320. This work was supported by the FP7 European Project HELATHIS (Grant Agreement No. 241378) and the regional government of Galicia (Projects No. IN841D-2010/14 and IN841C 2011/231). Santiago de Compostela, September 2013 xvi List of Figures Fig. 1.19 PECVD tool chamber usage for the p-i-n deposition ............. 29 Fig. 1.20 a-Si:H laser scribing step, P2 ......................... 30 Fig. 1.21 PVD sputtering tool .............................. 31 Fig. 1.22 a-Si:H and back contact laser scribing step, P3 ................ 31 Fig. 1.23 P1, P2 and P3 scribes (microscope view) ................... 32 Fig. 1.24 Final module structure which allows the current to flow among the cells . . . 32 Fig. 1.25 T-Solar production line is ready to produce modules of four different sizes . 33 Fig. 1.26 Buss tool picture and clean room to handle the PVB ............. 34 Fig. 1.27 Lamination tool and autoclave tool overall view ............... 35 Fig. 1.28 Junction box tool and in-line solar simulator tool ............... 35 Fig. 1.29 Finished full and quarter size modules ..................... 36 Fig. 1.30 Control plan scheme with all the controls made in-line and off-line ..... 38 Fig. 2.1 Generation of electron-hole pairs due to the photon incidence in a p-n junction 45 Fig. 2.2 Solar cell IV curve as result of the diode IV curve minus the Iph ....... 46 Fig. 2.3 Equivalent circuit of an a-Si:H solar cell ................... 47 Fig. 2.4 p-i-n structure ................................. 48 Fig. 2.5 Amorphous silicon structure (random network) showing a bond with a hydrogen atom ................................. 48 Fig. 2.6 DOS distribution model for intrinsic a-Si:H layer on a linear scale ...... 51 Fig. 2.7 DOS distribution model for intrinsic a-Si:H layer on a logarithmic scale . . . 51 Fig. 2.8 Different types of localized states in the band gap of a-Si:H and models that are used to calculate the recombination rate and charge occupation ..... 55 Fig. 2.9 Electronic transitions in the recombination process between a single energy level in the band gap of a semiconductor and the energy bands ........ 56 Fig. 2.10 Possible electronic transitions in the recombination process between the energy bands and an amphoteric R-G center represented by two energy levels in the band gap ................................. 59 Fig. 2.11 Band diagram for the a-Si:H p-i-n layers under 1 sun illumination and without polarization (V=0V) ........................ 63 Fig. 2.12 Structure of the simulated a-Si:H solar cell .................. 65 Fig. 2.13 Photon flux into a semiconductor ....................... 67 Fig. 2.14 Optical absorption coefficient for different PV materials ........... 68 Fig. 2.15 Complex refractive index for our a-Si:H layers ................ 69 List of Figures xvii Fig. 2.16 SEM picture of an industrial TCO sample AN10 from AGC showing fine surface roughness of the crystal grains .................... 71 Fig. 2.17 AFM picture of an industrial TCO sample AN10 from AGC ......... 71 Fig. 2.18 Solar spectrum AM0, AM1.5G and AM1.5D ................. 73 Fig. 2.19 Simplified flow chart of the method used for the numerical resolution .... 75 Fig. 2.20 Energy band diagram of a typical single junction a-Si:H solar cell under equilibrium conditions ............................. 77 Fig. 3.1 Schematic representation of the glow-discharge deposition process ..... 87 Fig. 3.2 PECVD chamber diagram ........................... 88 Fig. 3.3 Former T-Solar coupon with 1 cm2solar cells and sample holder for MCs measurements of IV curves and SR with the switch box to choose the solar cell to be measured ............................... 90 Fig. 3.4 Scratch mask, scratch pencils and Combi mask with their four pieces .... 92 Fig. 3.5 Back contact side and glass side of a coupon produced at FZJ with a magnetized mask (Combi mask) ........................ 92 Fig. 3.6 Position of the coupons with MCs along the full size panel .......... 93 Fig. 3.7 Back contact side and glass side of a coupon produced at TS through laser scribe ...................................... 94 Fig. 3.8 Position of the MMs along the full size panel ................. 95 Fig. 3.9 10 cm x 10 cm total area MM produced at TS through LSR scribe ...... 95 Fig. 3.10 Arc lamp and monochromator ......................... 97 Fig. 3.11 Incoming beam divided by the beam-splitter ................. 98 Fig. 3.12 Two lock-in amplifiers and cold light source with its illumination system . . 99 Fig. 3.13 SR equipment and links between the different devices ............ 103 Fig. 3.14 Configuration of a Czerny-Turner monochromator .............. 104 Fig. 3.15 The beam-splitter with the two arms ...................... 104 Fig. 3.16 Optical path of the light through the system .................. 105 Fig. 3.17 Relationship between hardware and software for the CSR equipment .... 106 Fig. 3.18 Comparison of the EQE provided by the manufacturer and the average EQE measured in the T-Solar equipment for the reference solar cell WPVS Cell 06-2008 .................................... 107 Fig. 3.19 Experimental IV curve and DC bias point for 4 representative LEDs ..... 111 xviii List of Figures Fig. 3.20 Peak wavelength and FWHM for a LED with narrow band width and for a LED with wide band width .......................... 112 Fig. 3.21 Experimental spectral irradiance for 23 selected LEDs in the range 370 nm - 1000 nm .................................... 112 Fig. 3.22 Decomposition of a periodic signal in its different harmonics ........ 113 Fig. 3.23 Sine wave in the time-domain and, after FFT analysis, sine wave in the frequency-domain ............................... 114 Fig. 3.24 MM with every cell connected individually with a conducting adhesive side buss....................................... 115 Fig. 3.25 Diagram of the VFSR measurement system .................. 116 Fig. 3.26 Time dependent current density curve ..................... 117 Fig. 3.27 Current density curve in the frequency-domain as resulting from FFT analysis 117 Fig. 3.28 EQE measured in a CSR equipment with monochromator and in a VFSR equipment ................................... 118 Fig. 3.29 Solar simulator and the other parts of IV curve measurement devices .... 120 Fig. 3.30 Equivalent circuit of an a-Si:H PV solar cell ................. 122 Fig. 3.31 Typical illuminated IV curve and PV curve of a photovoltaic solar cell . . . 124 Fig. 3.32 Typical dark IV curve of a photovoltaic solar cell ............... 125 Fig. 3.33 Spatial mapping result of the calibration done on 29/11/2011 ........ 127 Fig. 3.34 LabVIEW front panel for the IV curve tracer ................. 129 Fig. 4.1 Design for the former T-Solar solar cells produced with shadow mask .... 133 Fig. 4.2 JV curves for solar cells on Asahi U TCO-glass ................ 138 Fig. 4.3 Illuminated JV curve for three cells before and after performing shunt busting for a coupon sent to FZJ without back contact (first shipment) ........ 141 Fig. 4.4 Dark JV curve for the best cell (C17) before and after performing shunt busting for a coupon sent to FZJ without back contact (first shipment) .... 141 Fig. 4.5 JV curve comparison for the best cells between the samples produced on 04/02/10 only with p-i-n, with back contact layers added on 18/03/10: p-i-n without any treatment, p-i-n treated with HF etching and p-i-n + AZO . 144 Fig. 4.6 Dark JV curves for two solar cells sent with and without back contact .... 144 Fig. 4.7 EQE comparison between the samples produced on 04/02/10 with back contact layers deposited on 18/03/10: p-i-n without any treatment, p-i-n treated with HF etching, p-i-n + AZO ..................... 145 List of Figures xix Fig. 4.8 Total transmission for five TCO types with different carrier mobility, thickness and resistivity ............................ 148 Fig. 4.9 Correlation between Jsc measured in the SS and TCO-glass absorbance for five TCO-glass types with different carrier mobility, thickness and resistivity 148 Fig. 4.10 JV curve for five TCO types with different carrier mobility, thickness and resistivity .................................... 150 Fig. 4.11 EQE curve for five TCO types with different carrier mobility, thickness and resistivity .................................... 150 Fig. 4.12 Total transmission for three industrial TCO-glasses and one laboratory TCO-glass (Asahi U) .............................. 152 Fig. 4.13 JV curve for two TCO-glass types with different total transmission, carrier mobility and thickness ............................. 153 Fig. 4.14 EQE curve for two TCO-glass types with different total transmission, carrier mobility and thickness ............................. 154 Fig. 4.15 Structure of an a-Si:H solar cell including the placement of two ARC at interfaces air/glass and TCO/p-layer ...................... 155 Fig. 4.16 JV curve for four solar cells with NTO interface layer with different thickness and doping concentration deposited on Asahi U TCO-glass ......... 157 Fig. 4.17 EQE curve for four solar cells with NTO interface layer with different thickness and doping concentration deposited on Asahi U TCO-glass .... 157 Fig. 4.18 Reflection for four solar cells with NTO interface layer with different thickness and doping concentration deposited on Asahi U TCO-glass .... 158 Fig. 4.19 Total and diffuse transmission for a SnO2TCO deposited on float glass of 3.2 mm and a ZnO TCO deposited on Corning glass of 1.0mm ....... 159 Fig. 4.20 Haze for a SnO2TCO deposited on float glass of 3.2 mm and a ZnO TCO deposited on Corning glass of 1.0mm..................... 160 Fig. 4.21 JV curve for three solar cells with ZnO front TCO produced at FZJ with different doping concentration and deposition process ............ 161 Fig. 4.22 EQE curve for three solar cells with ZnO front TCO produced at FZJ with different doping concentration and deposition process ............ 162 Fig. 4.23 JV curve for two solar cells with ZnO front TCO produced at FZJ and Asahi U (SnO2) front TCO. Both samples have μc-Si p-layer ........ 163 xx List of Figures Fig. 4.24 EQE curve for two solar cells with ZnO front TCO produced at FZJ and with Asahi U (SnO2) front TCO. Both samples have μc-Si p-layer and are compared with a solar cell with ZnO front TCO with a-Si:H p-layer ..... 163 Fig. 4.25 Typical thickness mapping of a n-type a-Si:H control layer .......... 165 Fig. 4.26 Variation of the main electrical parameters for four coupons placed along the diagonal F-I of the panel ............................ 165 Fig. 4.27 EQE curve for two coupons of the same panel, one placed in the centre and the other one in a corner ............................ 166 Fig. 4.28 Degradation factors of Jsc for solar cells prepared with different i-layer thicknesses in different laboratories ...................... 168 Fig. 4.29 Degradation factors of Voc and FF for solar cells prepared with different i-layer thicknesses in different laboratories .................. 168 Fig. 4.30 Efficiency of quarter size modules in the initial and stabilized state and η degradation factors ............................... 169 Fig. 4.31 Relation between i-layer thickness reduction and throughput increase of the PECVD machine and extrapolation to the annual production capacity .... 169 Fig. 4.32 JV curve for three solar cells which represent the main three evolutions performed at T-Solar during the three years of the HELATHIS project .... 172 Fig. 4.33 EQE curve for three coupons which represent the main three evolutions performed at T-Solar during the three years of the HELATHIS project. The EQE for the world record solar cell of Oerlikon Solar-Lab is also presented . 173 Fig. 4.34 Gain of the main electrical parameters (in stabilized state) respect to evolution No. 4 for the main four evolutions performed at T-Solar during the three years of the HELATHIS project ................... 173 Fig. 4.35 EQE curve for two solar cells, one with Al back contact and another one with Ag back contact, both deposited by sputtering at UU ............. 175 Fig. 4.36 Reflectance for five different Ag and/or Al back reflectors .......... 175 Fig. 4.37 EQE of two SJ solar cells with different i-layer thickness, both measured with our CSR equipment and the new VFSR equipment ........... 176 Fig. 4.38 EQE of two SJ solar cells with different BKM evolutions (No.4 vs. No.8), both measured with the CSR and VFSR equipment .............. 178 Fig. 4.39 Jsc spatial mapping along a 10 x 10 cm2size mini module measured with the VFSR equipment ................................ 179 List of Figures xxi Fig. 4.40 Variation of short circuit current density and open circuit voltage with the i-layer thickness for three different p-layer thicknesses ............ 180 Fig. 4.41 Variation of efficiency and fill factor with the i-layer thickness for three different p-layer thicknesses .......................... 181 Fig. 4.42 Variation of short circuit current density and open circuit voltage with the i-layer thickness for three different p-layer doping concentrations ...... 181 Fig. 4.43 Variation of efficiency and fill factor with the i-layer thickness for three different p-layer doping concentrations .................... 182 Fig. 4.44 Light trapping in the active layer (i-layer) of an a-Si:H solar cell ....... 183 Fig. 4.45 Representation of the texture interface parameters .............. 184 Fig. 4.46 Simulated JV curves for flat and textured interfaces with different heights . . 185 Fig. 4.47 Simulated EQE for flat and textured interfaces with different heights .... 186 Fig. 4.48 Experimental and simulated JV curve in the initial and degraded state .... 187 Fig. 4.49 Experimental and simulated EQE in the initial and degraded state ...... 187 Fig. 4.50 Efficiency evolution in modules in the last four years ............. 188 Fig. 4.51 Efficiency histogram for similar production periods of 5.72 m2a-Si:H modules from 2009 to 2012 .......................... 189 Fig. 4.52 Evolution of the T-Solar production capacity and throughput during the period 2009 - 2012 ............................... 190 List of Tables Tabla 1.1 An overview of major PV technologies ................... 10 Tabla 1.2 Evolution of the cumulative solar electrical capacities until 2035 ...... 21 Tabla 2.1 Time rates of change in the carrier concentrations for transitions involving VB and CB tail states at an arbitrary energy level ETin the band gap .... 56 Tabla 2.2 Transition rates of change in the carrier concentrations for recombination processes involving DB states. ........................ 59 Tabla 2.3 Baseline input parameters used in the simulation .............. 66 Tabla 3.1 IV curve parameters of the calibrated solar cell (4 cm2) from FHG-ISE measured under STC ............................. 99 Tabla 3.2 Comparison of the technical data sheet values with the experimental data obtained at our lab for the 24 selected LEDs ................. 110 Tabla 3.3 Results of all the calibration controls done for the three lamps used until the moment in the solar simulator ...................... 128 Tabla 4.1 Electrical parameters comparison for the old (shadow mask) and new (laser scribe) T-Solar solar cell design ....................... 134 Tabla 4.2 Electrical parameters comparison for three different types of annealing performed in three coupons which were located side by side in the same panel. Top: Results before and after annealing and after 300 kWh/m2of light soaking. Middle: Variation for the three coupons before and after performing annealing. Bottom: LID for the three coupons ......... 135 xxiv List of Tables Tabla 4.3 Electrical parameters average for the best 5 cells of a coupon without treatment, after shunt busting and after shunt busting and annealing (top). Variation respect to values without treatment (middle). Number of good cells in four different coupons without treatment, after annealing, after shunt busting or after shunt busting and annealing (bottom) ......... 136 Tabla 4.4 Electrical parameters comparison for 1 cm2solar cells and 5.72 m2full size modules produced on 16/06/2011 ...................... 137 Tabla 4.5 Electrical parameters for solar cells on Asahi U TCO-glass produced with different production methods and treatments ................. 139 Tabla 4.6 Average electrical parameter values before and after performing shunt busting for a coupon sent to FZJ without back contact (first shipment) . . . 140 Tabla 4.7 Average electrical parameter values for the measured solar cells in each experiment (p-i-n: 8 cells, p-i-n + HF: 12 cells, p-i-n + AZO: 8 cells) . . . 143 Tabla 4.8 Electrical parameter values of the best cell for the three experiments and average values for 1 cm2cells produced (with mask) with the T-Solar standard process ............................... 143 Tabla 4.9 Summary of the main characteristics of the different TCO types ...... 146 Tabla 4.10 Transmission, absorption and sheet resistance for five TCO-glass types with different carrier mobility, thickness and resistivity .............. 147 Tabla 4.11 Electrical parameters for five TCO types with different carrier mobility, thickness and resistivity ........................... 149 Tabla 4.12 Integrated total transmission in the range 400 nm - 800 nm for three industrial TCO-glasses and one laboratory TCO-glass (Asahi U) and their gain with respect to AN10 .......................... 151 Tabla 4.13 Electrical parameters for two TCO-glass types with different total transmission, carrier mobility and thickness ................. 153 Tabla 4.14 Thickness and doping conditions for four different interface layers of NTO deposited on Asahi U ............................. 155 Tabla 4.15 Electrical parameters (initial state) for two references and four solar cells with NTO interface layer with different thickness and doping concentration deposited on Asahi U TCO-glass ....................... 156 Tabla 4.16 Material properties of a SnO2TCO deposited on float glass of 3.2 mm and a ZnO TCO deposited on Corning glass of 1.0mm ............. 158 List of Tables xxv Tabla 4.17 Electrical parameters (initial state) for three solar cells with ZnO front TCO produced at FZJ with different doping concentration and deposition process 160 Tabla 4.18 Electrical parameters (initial state) for two solar cells with ZnO front TCO produced at FZJ and Asahi U (SnO2) front TCO. Both samples have μc-Si p-layer .................................... 162 Tabla 4.19 Electrical parameters statistics for four coupons of the same panel placed along the diagonal .............................. 166 Tabla 4.20 LID for thirteen coupons produced at different laboratories (2 at UU, 3 at FZJ and 8 at TS) with different i-layer thicknesses ............. 167 Tabla 4.21 Main BKM evolution related to the PECVD process since the T-Solar production start ................................ 170 Tabla 4.22 Electrical parameters for five coupons with solar cells which represent the main five evolutions performed at T-Solar during the three years of the HELATHIS project .............................. 171 Tabla 4.23 Gain in the stabilized state respect to evolution No. 4 ............ 174 Tabla 4.24 Electrical parameters for the stabilized record T-Solar solar cell and world record solar cell of Oerlikon Solar-Lab ................... 174 Tabla 4.25 Jsc determined from SR with the VFSR and CSR equipment for solar cells with i-layer thickness (di)of260nmand200nm.............. 177 Tabla 4.26 Jsc determined from SR with the VFSR and CSR equipment for solar cells with different BKM evolution steps (No.4 vs. No.8) ............ 177 Tabla 4.27 Statistical parameters obtained from the mapping in Fig. 4.39 (VFSR equipment) and the same mapping obtained with the CSR equipment . . . 179 Tabla 4.28 Simulated electrical parameters depending on the βangle in the degraded state ...................................... 185 Tabla 4.29 Electrical parameters for an experimental and simulated solar cell in the initial and degraded state ........................... 186 xxxii List of Acronyms PV power-voltage PVB polyvinyl butyral PVD physical vapour deposition QASR quality assurance and shunt removal QE quantum efficiency rreflection R-G recombination-generation RF radiofrequency ROW rest of the World Rsseries resistance Rsh shunt resistance Rsq sheet resistance SB shunt busting SCPI standard commands for programmable instruments SEM scanning electron microscopy Si silicon SiF4silicon tetrafluoride SIGMA "SIstema de Gestión y Monitorización Avanzado" SiH4silane SiO2silicon oxide SJ single junction SnO2tin oxide List of Acronyms xxxiii SnO2:F fluorine-doped tin oxide SR spectral response SRH Shockley Read Hall SS solar simulator STC standard test conditions (1000 W/m2,25oC and AM1.5G spectral distribution) SWE Staebler-Wronski effect Ttemperature T-Solar T-Solar Global S.A. TCO transparent conductive oxide TCS thermal conditioning station TiO2titanium oxide TJ tandem junction TMM transfer matrix method TS T-Solar Global S.A. UB University de Barcelona USC University of Santiago de Compostela UU University Utrecht UV ultraviolet UVI University of Vigo VB valence band VBT valence band tail VFSR very fast spectral response xxxiv List of Acronyms VI virtual instrument Vmpp voltage at maximum power point Voc open circuit voltage WP work packages ZnO zinc oxide Introduction Nowadays the most of the energy consumption is provided from fossil fuels as oil, natural gas or coal. Nevertheless, in the last years, the world-wide energetic policies are changing and renewable energies are increasing rapidly their quota in the energy mix. The main reasons for this modification are, on one hand, that fossil fuel prices are raising due to the decreasing availability of this sources. On the other hand, the consumer countries are dependent of the producer countries. In addition, the massive use of them is producing the climate change. The European Union (EU) climate and energy targets, known as the "20-20-20" targets, set three key objectives for 2020 [1]: A 20% reduction in EU greenhouse gas emissions from 1990 levels (to come back to 450 ppm - 550 ppm range), to raise the share of EU energy consumption produced from renewable resources to 20% and a 20% improvement in the EU’s energy efficiency. During the last years the investments into renewable energy and energy efficiency sectors has been increasing. In 2011, world-wide new investments into these sectors increased to a new record of e202 billion, including e19.8 billion research and development spending. More than 85% (e173 billion) of these investments were non-governmental, non-research clean energy investments [2]. This resulted in a record of 83.5 GW of new clean energy generation capacity, bringing the total to more than 565 GW (50% more than the installed nuclear generating capacity world-wide). Specifically for solar energy, in 2011 (for the second year in a row) attracted the largest amount of new investments into renewable energies [2]. There was a 44% increase (respectively to 2010) in solar energy investments to e98.5 billion [2]. Nowadays, electricity production from photovoltaic (PV) solar systems has shown that it can be cheaper than peak prices in the electricity exchange. The electricity generation costs 2Introduction are already at the level of residential electricity prices in several countries, depending on the actual electricity price and the local solar irradiance level. Photovoltaic solar energy will continue to grow at high rates in the coming years accelerating its consolidation because it is considered to be the most powerful energy source in the future since it joins optimal conditions such as solar energy is an unlimited resource (the sun supplies 1000 times more energy than the world energy consumption) or it is a clean energy. Furthermore, the current solar module technologies are well established and provide a reliable product, with sufficient efficiency and energy output for at least 25 years lifetime. The development and recent fast growth of the PV industry has resulted in module oversupply and a fast decrease of module prices in the last years, demanding now on all actors in the PV sector an important increase in R&D activities to be competitive in the future. The investigations of this PhD thesis were performed in that frame. It was financed by the High Efficient very LArge area THIn film Silicon photovoltaic modules (HELATHIS) project and the "Torres Quevedo" program. The activity involved the Technology and R&D Department of the T-Solar Global S.A. (T-Solar) company and the Department of Electronics and Computer Science of the University of Santiago de Compostela. This PhD thesis is structured in four chapters and is based on the optical and electrical characterization of solar cells with different materials and deposition process conditions, the optimization and modelling of these cells and the development of metrology equipments. The first chapter presents a general view on the photovoltaic status analysing the different photovoltaic technologies and its applications, as well we will take a glance on the evolution of the world-wide photovoltaic production capacity and installed capacity since 2000 and the forecast until 2017. Next, we will focus in T-Solar and its hydrogenated amorphous silicon (a-Si:H) photovoltaic module factory. Finally, the main aims for the optimization of a-Si:H solar cells in the frame of the FP7 European Project HELATHIS are exposed. In the second chapter the electrical model that governs the a-Si:H solar cells is presented, as well as their equivalent circuit, both obtained from the ideal diode model and deduced from the crystalline silicon (c-Si) solar cell model. Then, the physical model is explained emphasising in the differences between the properties of a-Si:H and c-Si. The most important changes result from the density of states (DOS), the recombination-generation statistics, the graded layers and the light induced degradation produced in a-Si:H due to the Staebler-Wronski effect (SWE). Next, it is explained how the optical generation is produced in a-Si:H solar cells, how the spectral irradiance is considered and how the optical transmission, Introduction 3 reflection and absorption works in the different solar cell layers. As well, we point out the different methods adopted in the here used Sentaurus (Synopsys) simulation tool to simulate both models, with flat or textured layers. Finally, a summary is presented about how the semiconductor equations are solved. That section introduces the equations and boundary conditions that the Sentaurus software solves to get, e.g., the current-voltage (IV) curves or external quantum efficiency (EQE) curves. The simulated device is meshed forming thousands of nodes and the respective equations describing the device and the boundary conditions are discretised in every of these nodes, then, the resulting system of equations is linearised and solved with the Gummel or Newton-Raphson methods. The third chapter presents how the plasma enhanced chemical vapour deposition (PECVD) production process works, as well as how we fabricate R&D solar cells of 1 cm2 (or4cm 2) or mini modules in the industrial production line. Later, the equipment installed in the laboratory are shown. Concerning this thesis, the most important ones are a conventional spectral response (CSR) and a very fast spectral response (VFSR) equipment, which are used to determine the EQE curves, and a solar simulator (SS) equipment with its IV tracer to measure the illuminated or dark IV curves. The three equipment were developed in our laboratory. The development to produce R&D solar cells and mini modules in the T-Solar production line as well as the development of the measurement equipment have been crucial to improve the production module efficiency. In the fourth chapter we mainly present work performed in the frame of the European project HELATHIS. One of its objectives is the optimization of the solar cell structure implemented in T-Solar’s industrial production process for very large area (2.6mx2.2m) a-Si:H PV modules. In the optimization not only the achievement of the physical efficiency limit of the solar cells has to be taken into account, but also aspects like solar cell fabrication time (impacting on factory throughput), the material and energy consumption (impacting on production cost), etc. In reference to this we report here, firstly, on the evolution to fabricate highly efficient and reliable a-Si:H test solar cells in the industrial environment of T-Solar and the preparation of samples with solar cells for shipments to project partners. Next, we investigate the cell structure developments, reporting on the front transparent conductive oxide (TCO) layer and glass developments, p-i-n structure developments and back contact developments. Next, we analyse the results obtained with the VFSR equipment and simulation studies on the p-i-n structure and the front TCO layer texture. To conclude this chapter, the 4Introduction improvements in the industrial production line of T-Solar since the beginning in the year 2008 are shown. Finally, the conclusions and the main advances achieved during this PhD thesis are indicated. CHAPTER 1 PHOTOVOLTAICS STATUS AND T-SOLAR’S THIN FILM SILICON FACTORY Photovoltaic (PV) solar energy is expected to be the most powerful energy source in the future since it gathers the optimal conditions for this. For instance, solar energy is an unlimited resource (the sun supplies 1000 times more energy than the world energy consumption) and furthermore, it is a clean energy. In fact, the current solar module technologies are well established and provide a reliable product, with sufficient efficiency and energy output to have at least a 25-year-long lifetime. In addition, the increasing amount of electricity interruptions (due to grid overloads), together with the continuous price rise of electricity coming from conventional energy sources, add attractiveness to the PV systems. This chapter presents a general overview of the status of photovoltaics, analysing the different photovoltaic technologies and their applications. Likewise, we will take a glance on the evolution of the world-wide photovoltaic production capacity and the installed capacity from year 2000 to the present, and the forecast until year 2017. Next, we will focus on the company T-Solar Global S.A. and its hydrogenated amorphous silicon (a-Si:H) photovoltaic module factory. Finally, the main aims for the optimization of a-Si:H solar cells in the frame of the FP7 European Project HELATHIS are exposed. 6Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory 1.1 Photovoltaic technologies There are a wide range of PV cell technologies on the market today, using different types of materials, and even a larger number will be available in the future. PV cell technologies are usually classified into three generations, depending on the material on which they are based and on their level of commercial maturity [3]. 1.1.1 First generation PV technologies: Crystalline silicon solar cells Silicon is one of the most abundant elements in the earth’s crust. Crystalline silicon is the material most commonly used in the PV industry, therefore wafer-based crystalline silicon (c-Si) PV cells and modules dominate the current market. This is a mature technology that employs the accumulated knowledge basis developed in the electronic industry. This type of solar cell is produced in mass, individual companies produce them at a rate of several hundred MW per year, reaching in some cases the GW-scale. In Fig. 1.1 the manufacturing process of wafer-based silicon PV modules is presented. It comprises four main steps: poly-silicon production, ingot/wafer production, cell production and module assembly. Figure 1.1: Crystalline silicon module production chain [4]. 1.1. Photovoltaic technologies 7 Crystalline silicon solar cells are classified into three main types depending on how the Si wafers are made [3]. They are: – Mono-crystalline (c-Si). – Multi-crystalline (mc-Si). – Edge-defined film-fed growth (EFG) ribbon silicon. Crystalline silicon technologies accounted about 85% of the global PV sales in 2012 [5]. The efficiency of crystalline silicon modules ranges from 15% to 21% [6]. Since it is a mature technology, continued cost reductions are possible through improvements in materials and manufacturing processes. From the economies of scale, one can predict that when the market continues to grow, more high-volume manufacturers will emerge. 1.1.2 Second generation PV technologies: Thin film solar cells Thin film solar cells could potentially provide lower cost electricity than c-Si wafer-based solar cells. However, this is not certain. The lower capital costs (due to lower production and materials costs) are offset, to some extent, by the lower efficiencies of this technology. Besides, the reduction of c-Si modules’ costs will make the economics even more challenging. Thin film solar cells do not use wafers to produce the semiconductor, in contrast, they consist of successive thin layers (of about 1 μmto4μm thick) deposited on a large inexpensive substrate, such as: glass, polymer, or metal. As a consequence, they require a lot less semiconductor material to absorb the same amount of sunlight (up to 99% less material than crystalline solar cells). In addition, thin films can be packaged into flexible and lightweight structures which can be easily integrated into building components, such as roofs or facades (building integrated photovoltaics (BIPV)). The three primary commercially developed technologies employed in thin film solar cells are: amorphous silicon; cadmium telluride (CdTe); and copper, indium, gallium, selenium (CIGS)[3]. They are thoroughly explained in the following paragraphs: –Amorphous silicon solar cells (a-Si:H and a-Si:H/μc-Si:H): Amorphous silicon can be deposited on cheap and very large substrates (up to 5.7m 2of glass) based on continuous deposition techniques, thus considerably reducing manufacturing costs. Currently, amorphous silicon PV module efficiencies are in the range from 5% 14 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory –Grid-connected power plants: Can be ground-mounted, or located on large industrial/commercial buildings such as shopping malls, airport terminals or railway stations. These produce a large quantity of PV electricity at a single point. –Off-grid systems for rural electrification: Can be a small solar PV system covering the basic electricity needs of a household, or a larger solar mini-plant, providing enough power for several homes. These systems bring access to electricity to remote areas (mountain huts, villages in developing countries or small islands). –Off-grid industrial applications: Very frequent in the telecommunications and transport fields. For example for repeater stations for mobile phones, traffic signals, marine navigation aids, security phones, remote lighting, highway signs, etc. These bring cost-effective power to areas far away from the electricity grid, avoiding the high cost of installing cabled networks. –Consumer goods: Many everyday electrical appliances use PV solar cells: watches, calculators, toys, battery chargers, water sprinklers, lighting, etc. 1.3 Evolution of photovoltaic production and installed capacity The data presented in this section is mainly obtained from the reports "Global market outlook for photovoltaics 2013 - 2017" of the European Photovoltaic Industry Association (EPIA) [5] and "PV status report 2012" of the Joint Research Centre (JRC) [2]. 1.3.1 Photovoltaic market: Installations The installed PV power here presented includes only systems connected to the grid and not those that have been installed but not yet connected. In Fig. 1.5 the annual PV installations from 2000 to 2012 are presented. PV technology has grown over the past decade at a remarkable rate and is on the way to becoming a major source of power generation for the world. The annual PV installations in 2012 were about 31 GW. After the world-wide PV market more than doubled in 2010, the growth in 2011 was 80% compared to 2010, nevertheless, in 2012 the PV market stabilized. The 2012 market volume was mainly led by Germany (7.6 GW installed in 2012 / 32.4 GW cumulative installed capacity), China (5.0GW/8.3 GW), Italy (3.4GW/16.4 GW), the USA (3.3GW/7.8 GW) and Japan (2.0GW/6.9 GW). 1.3. Evolution of photovoltaic production and installed capacity 15 Ϭ ϱ ϭϬ ϭϱ ϮϬ Ϯϱ ϯϬ ϯϱ WsĂŶŶƵĂůŝŶƐƚĂůůĂƚŝŽŶƐ;'tͿ ZKt D ŚŝŶĂ ŵĞƌŝĐĂƐ W ƵƌŽƉĞ Figure 1.5: Annual PV installations from 2000 to 2012 [5]. ROW: Rest of the World, MEA: Middle East and Africa, APAC: Asia Pacific. The results in 2012 and the forecast for the coming years indicate that Europe’s leading role in the PV market is coming to an end. In 2011, Europe accounted for 74% of the world’s new PV installations (22.4 GW); in 2012 this number was around 55% (17.2 GW) [5]. In 2013 it is almost certain that the majority of new PV capacity in the world will be installed outside of Europe. Therefore, in the next years the PV growth in Europe will occur at a more stable – and sustainable – rate than it has in the last few years. In the future the driving forces will be countries like China, the USA, Japan and India. The PV market is becoming truly global. As shown in Fig. 1.6, at the end of 2012 the total cumulative PV installations world-wide were 102 GW. The 69% of the world-wide installed capacity (70 GW from the 102 GW) are installed in the European Union (EU). Moreover, Germany counts on 32 GW. In the period 2000 - 2012, Europe has increased 540 times (from 129 MW to 70 GW) its installed capacity [5], [16]. In 2012, for the second year in a row, PV was the number-one new source of electricity generation installed in Europe. PV covers 2.6% of the electricity demand in Europe [5]. The development of PV electricity in Europe is occurring at a faster rate than almost anyone had expected. As shown in Fig. 1.7, in 2012 a total of 44.9 GW of new power capacity was connected in the EU while 15.8 GW were decommissioned. This resulted in 29.1GWofnew 16 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory net capacity in the EU. From this figure, 16.7 GW (or 57%) of the new net capacity came from PV and 30.8 GW (or 106%) corresponded to renewable energies. The net capacity from non-renewable energies was negative [5]. Ϭ ϮϬ ϰϬ ϲϬ ϴϬ ϭϬϬ ϭϮϬ WsĐƵŵƵůĂƚŝǀĞŝŶƐƚĂůůĞĚĐĂƉĂĐŝƚLJ;'tͿ ZKt D ŚŝŶĂ ŵĞƌŝĐĂƐ W ƵƌŽƉĞ Figure 1.6: Cumulative PV installations from 2000 to 2012 [5]. ͲϴϬϬϬ ͲϰϬϬϬ Ϭ ϰϬϬϬ ϴϬϬϬ ϭϮϬϬϬ ϭϲϬϬϬ ϮϬϬϬϬ WsŐĞŶĞƌĂƚŝŽŶĐĂƉĂĐŝƚŝĞƐ;DtͿ ϭϲϲϳϮ ϭϭϴϵϱ ϭϬϱϯϱ ϯϬϲϱ ϭϯϯϴ ϴϯϯ ϰϮϰ ϱϬ ϳϮϮ ϲ ϱ Ͳϱϰϵϱ Ͳϰϯ ͲϭϮϬϱ ͲϯϮϬϰ Ͳϭϱϴ Ͳϱϰϰϭ ͲϮϬϳ Figure 1.7: New installed or decommissioned electricity generation capacity in Europe in 2012 [5]. 1.3. Evolution of photovoltaic production and installed capacity 17 Regarding the future PV development, EPIA considers two scenarios for PV markets. The Business-as-usual scenario assumes rather pessimistic market behaviour. The Policy-driven scenario assumes a positive market behaviour, considering PV as a major power source in the coming years [5]. For the first time in the last 12 years, the PV market in Europe decreased in 2012 in terms of new connected capacity, as shown in Fig. 1.5. The future of the European market is uncertain for the coming years and the short-term prospects are stable in the best case or declining. In the Business-as-usual scenario, the expected growth of markets outside Europe is not likely to compensate fast enough for the slowdown of the market in Europe in the two coming years. But even in this scenario, the global market could be as high as 48 GW in 2017 as indicated in Fig. 1.8 [5]. Ϭ ϭϬ ϮϬ ϯϬ ϰϬ ϱϬ ϲϬ ϳϬ ϴϬ ϵϬ ϮϬϬϴ ϮϬϬϵ ϮϬϭϬ ϮϬϭϭ ϮϬϭϮ ϮϬϭϯ ϮϬϭϰ ϮϬϭϱ ϮϬϭϲ ϮϬϭϳ ŶŶƵĂůWsŵĂƌŬĞƚƐĐĞŶĂƌŝŽƐ;'tͿ ,ŝƐƚŽƌŝĐĂůĚĂƚĂ W/ƵƐŝŶĞƐƐͲĂƐͲhƐƵĂů W/WŽůŝĐLJͲƌŝǀĞŶ Figure 1.8: Annual PV market scenarios until 2017 - Bussiness-as-usual and Policy-driven [5]. In the Policy-Driven scenario, the European market would stabilise first around 16 GW - 17 GW in 2013 before growing slowly again to around 25 GW - 28 GW five years from now. In that case, the global market could top more than 84 GW in 2017 (see Fig. 1.8), with two-thirds of this coming from new markets outside Europe [5]. EPIA expects the Asia Pacific (APAC) region (without China) to represent between 10 GW and 20 GW each year until 2017. China alone could add 10 GW of PV installations each year [5]. 18 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory As shown in Fig. 1.9, in the Business-as-Usual scenario, the 200 GW mark could be reached in between 2014 and 2016, while in the Policy-Driven scenario, more than 420 GW of PV systems could be connected to the grid over the next five years [5]. Ϭ ϱϬ ϭϬϬ ϭϱϬ ϮϬϬ ϮϱϬ ϯϬϬ ϯϱϬ ϰϬϬ ϰϱϬ ϮϬϬϴ ϮϬϬϵ ϮϬϭϬ ϮϬϭϭ ϮϬϭϮ ϮϬϭϯ ϮϬϭϰ ϮϬϭϱ ϮϬϭϲ ϮϬϭϳ WsĐƵŵƵůĂƚŝǀĞƐĐĞŶĂƌŝŽƐ;'tͿ ,ŝƐƚŽƌŝĐĂůĚĂƚĂ W/ƵƐŝŶĞƐƐͲĂƐͲhƐƵĂů W/WŽůŝĐLJͲƌŝǀĞŶ Figure 1.9: Cumulative PV market scenarios until 2017 - Bussiness-as-usual and Policy-driven [5]. 1.3.2 Photovoltaic industry In 2012, the PV industry went again through a challenging period, with political, market and industry factors affecting business along the whole value chain. Important manufacturers disappeared, were acquired or had to adapt their business plan, decrease the utilisation rate and consequently reduce significantly their production. The tough market environment in Europe has forced many important players out of the PV business. The PV global market capacity has evolved mainly in a context of production overcapacity. In the last three years, module production capacity was in between 150% - 230% higher than annual global installations [5]. This, resulted in continuous price pressure in PV systems market. However, the aggressive reduction of prices also opened new markets getting higher growth of the industry than hoped. This scenario benefits some PV companies and damages others. In Fig. 1.10 world PV cell/module production from 2009 to 2012, as well as the forecast from 2013 to 2017 is presented (bar graph). In the period from 2000 to 2011, total PV 1.3. Evolution of photovoltaic production and installed capacity 19 production increased almost by two orders of magnitude with annual growth rates between 40% and 90% [2]. According to IHS Solar, the world PV production capacity in 2012 reached 56.5 GW, increasing 24% compared to 2011. Those capacities are likely to grow continuously, with a growth rate of around 6% until 2017, reaching a production capacity up to 75 GW in 2017 [5]. Ϭй ϭϬй ϮϬй ϯϬй ϰϬй ϱϬй ϲϬй ϳϬй ϴϬй ϵϬй Ϭ ϭϬ ϮϬ ϯϬ ϰϬ ϱϬ ϲϬ ϳϬ ϴϬ ϵϬ ϮϬϬϵ ϮϬϭϬ ϮϬϭϭ ϮϬϭϮ ϮϬϭϯ ϮϬϭϰ ϮϬϭϱ ϮϬϭϲ ϮϬϭϳ WsŵŽĚƵůĞƐƉƌŽĚƵĐƚŝŽŶƐŚĂƌĞ WsŵŽĚƵůĞƐƉƌŽĚƵĐƚŝŽŶĐĂƉĂĐŝƚLJ;'tͿ KƚŚĞƌƐƚĞĐŚŶŽůŽŐŝĞƐ dŚŝŶĨŝůŵƐŚĂƌĞ;ŽƌŐĂŶŝĐнŝŶŽƌŐĂŶŝĐͿ ĐͲ^ŝƐŚĂƌĞ Figure 1.10: World PV cell/module production capacity from 2009 to 2012 and annual forecast until 2017 [5]. Since 2011, European industry represents only about 8% - 13% of the global market in terms of actual module production. China and APAC countries supply about 70% of the global PV demand [2], [5], since this region has experienced the most rapid growth in annual production over the last years. Regarding the representation per technology, the predominant c-Si technology is expected to maintain its market share at levels slightly higher than 80% (blue line in Fig. 1.10). The main advantage of c-Si technology is that complete production lines can be bought, installed and producing within a relatively short time-frame. However, the temporary shortage in silicon feedstock and the market entry of companies offering turn-key production lines for thin film solar cells, led to a massive expansion of investments into thin film capacities between 2005 and 2009 (more than 200 companies are involved) [2]. In 2005 the production of thin film solar modules reached more than 100 MW/year. In the period 2005 - 2009, compound annual growth rate (CAGR) of thin film solar module production was beyond that of the overall industry increasing the market share of thin film 20 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory products: 6% in 2005, 10% in 2007 and between 16% - 20% in 2009. Since then, the thin film share (black line in Fig. 1.10) is decreasing slowly as their ramp up of new production lines did not follow that of wafer-based silicon due to the competing market price of c-Si technology. Anyway, the forecast for thin film is to grow at a lower rate, and therefore, their market share will stabilise over the next five years. If the forecast presented in Fig. 1.10 is fulfilled, thin film production capacity could be 10 GW or 13% of the total 75 GW in 2017. The growth rate of a-Si:H technologies might be reduced by around 3% until 2017 [5]. The reason for this negative CAGR is the lower module efficiency of a-Si:H in comparison with the rapid evolution of CdTe and CIGS, limiting the market for thin film modules with efficiencies below 10% on module level. Among emerging technologies, organic photovoltaic (OPV) technologies and especially concentration photovoltaic (CPV) technologies are expected to enjoy around 1% of the market share by 2017. The development of these technologies is accelerated by the positive development of the PV market. The existing PV technology mix is a solid foundation for future growth of the sector as a whole. No single technology can satisfy all the different consumer needs. 1.3.3 Outlook In 1996, the Directorate-General for Energy of the European Commission published the study "Photovoltaics in 2010" [17]. The medium scenario forecast a cumulative installed capacity of 3 GW in the EU by 2010. The most aggressive scenario in this report predicted a cumulative installed PV capacity of 27.3 GW world-wide and 8.7 GW in EU for 2010. The reality check reveals that even the most aggressive scenario is lower than what we expect from the current developments. At the end of 2010, PV systems with a cumulative capacity of over 41 GW world-wide and over 30 GW in Europe were generating electricity (the installations increased further to about 102 GW world-wide and 70 GW in Europe at the end of 2012). Turnkey system prices as low as 1.0e/Wphave been reported for projects to be finished in 2013 [18]. After the massive cost reductions for the technical components of PV systems like modules and other components (Balance Of System, BOS) the next challenge is to lower the soft costs of PV system installations, like the permission or financing costs [2]. The different PV industry associations, as well as Greenpeace, the European Renewable Energy Council (EREC) and the International Energy Agency (IEA), have developed new scenarios for the future growth of PV. Table 1.2 shows the different scenarios of the 1.3. Evolution of photovoltaic production and installed capacity 21 Greenpeace / EREC study, as well as the different 2011 IEA World Energy Outlook scenarios and the IEA PV Technology Roadmap. It is interesting to note that the 2015 capacity values of four of the six scenarios presented in the table (in red) have already been reached or exceeded in 2012. With forecast cumulative PV installations (according to 2013 EPIA scenarios in Fig. 1.9) between 198 GW to 264 GW in 2015, even the Greenpeace revolution scenario is no longer fictional thinking [18]. Table 1.2: Evolution of the cumulative solar electrical capacities until 2035. In red, capacity values already reached or exceeded. Source: [2], [9], [19]. ƵŵƵůĂƚŝǀĞWsƉŽǁĞƌĐĂƉĂĐŝƚLJ;'tͿ ϮϬϭϮ ϮϬϭϱ ϮϬϮϬ ϮϬϯϬ ϮϬϯϱ ĐƚƵĂůŝŶƐƚĂůůĂƚŝŽŶƐ ϭϬϮ 'ƌĞĞŶƉĞĂĐĞͲZĞĨĞƌĞŶĐĞƐĐĞŶĂƌŝŽΎ ϴϴ ϭϮϰ Ϯϯϰ ϮϵϬ 'ƌĞĞŶƉĞĂĐĞͲ΀ƌ΁ĞǀŽůƵƚŝŽŶƐĐĞŶĂƌŝŽΎ Ϯϯϰ ϲϳϰ ϭϳϲϰ ϮϰϮϬ /ͲƵƌƌĞŶƚƉŽůŝĐLJƐĐĞŶĂƌŝŽΎΎ ϲϬ ϭϲϭ Ϯϲϴ ϯϭϰ /ͲEĞǁƉŽůŝĐLJƐĐĞŶĂƌŝŽΎΎ ϭϭϮ ϭϴϰ ϯϴϱ ϰϵϵ /ͲϰϱϬƉƉŵƐĐĞŶĂƌŝŽΎΎ ϳϬ ϮϮϬ ϲϮϱ ϵϬϭ /ͲWsƚĞĐŚŶŽůŽŐLJƌŽĂĚŵĂƉΎΎΎ ϳϲ ϮϭϬ ϴϳϮ ϭϯϯϬ ΎϮϬϯϱǀĂůƵĞƐĂƌĞĞdžƚƌĂƉŽůĂƚĞĚĂƐŽŶůLJϮϬϯϬĂŶĚϮϬϰϬǀĂůƵĞƐĂƌĞŐŝǀĞŶ ΎΎϮϬϭϱǀĂůƵĞƐĂƌĞĞdžƚƌĂƉŽůĂƚĞĚĂƐŽŶůLJϮϬϬϵĂŶĚϮϬϮϬǀĂůƵĞƐĂƌĞŐŝǀĞŶ ΎΎΎϮϬϭϱΘϮϬϯϱǀĂůƵĞƐĂƌĞĞdžƚƌĂƉŽůĂƚĞĚĂƐŽŶůLJϮϬϭϬϮϬϮϬϮϬϯϬΘϮϬϰϬǀĂůƵĞƐĂƌĞŐŝǀĞŶ The IEA’s Energy Technology Perspectives 2010 stated that for their current Baseline Scenario, the overall investments in energy supply and use, for the period between 2010 and 2050, totals e208 trillion [20]. The BLUE-Map scenario, which would limit the concentration of Greenhouse Gases at 450 ppm, has an additional financing need of e35.4 trillion, but at the same time the cumulative fuel savings of this scenario compared to the Baseline would be e86.2 trillion, or more than twice the investment cost. This clearly indicates the huge societal benefit of a more aggressive climate change approach. The photovoltaic industry has changed from a MW size industry into a mass-producing industry, aiming for multi GW production on the long term. The development to economy of scale that comes with large production volumes, allows new large solar cell companies to use their cost advantages to offer lower-priced products accelerating the growth rate. On the contrary, this development will influence negatively the small and medium companies. To survive they have to specialise in niche markets with high added value in their products. 22 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory Renewable energies are, contrary to conventional energy sources, the only ones to offer a reduction of prices rather than an increase in the future. We see a continuous decrease in production costs for renewable energy technologies, as a result of steep learning curves (see Fig. 1.11). Despite of this, renewable energies and photovoltaics are still perceived as being more expensive in the market than conventional energy sources. This is due to the fact that external energy costs, subsidies in conventional energies and price volatility risks are generally not yet taken into consideration. Nevertheless, electricity production from photovoltaic solar systems has already proved to be cheaper than residential consumer prices in a wide range of countries. Dat a: Nav i g an t Co n su l t in g ; EUPD m o d u l e p r i ce (si n ce 2006) Gr ap h : PSE A G 2012 All PV Technologies Mono-Si 24.5 GWp Multi-Si 31.6 GWp (incl. Ribbon-Si) Thin Film 8.5 GWp all PV Technol. (1980-2011) LR 19.52 all PV Technol. (2006-2011) LR 26.93 Mono (2006-2011) LR 26.61 Multi (2006-2011) LR 27.35 Thin Film (2006-2011) LR 20.51 Figure 1.11: Price learning curve by technology from 1980 to 2011 [21]. 1.4 Group T-Solar Global S.A. T-Solar Global S.A. (T-Solar) is a company dedicated to hydrogenated amorphous silicon (a-Si:H) PV module manufacture. It was founded in October 2006 and is part of T-Solar Group which itself is a subsidy of the multinational company Isolux. Its first PV modules were fabricated in July 2008. 1.4. Group T-Solar Global S.A. 23 T-Solar has implemented in its production line, which is fully automated and integrated, the SunFab factory technology from the American company Applied Materials Inc. (AMAT). T-Solar was the second customer of AMAT and its factory the first constructed in Europe. Its thin film module fabrication process was an innovative project on world-wide level, due to the low cost of the production process and the enormous size of the modules with dimensions of 2.2mx2.6 m and a maximum power of about 430 Wp(efficiency of 7.52%) [22]. The factory has an annual production capacity of about 72 MWp, equivalent to about 900.000 m2 of PV modules. The factory is located in the "Parque Tecnolóxico de Galicia", part of the San Cibrao das Viñas industrial area, in Ourense, Spain. The production plant and the offices take up a plot of 29.000 m2(see Fig. 1.12). The laboratories are placed in the technical annex, the orange painted part of the building in the lower right side of Fig. 1.12. Figure 1.12: T-Solar Global S.A.’s a-Si:H PV module factory. Apart from fabrication and commercialization of PV modules, T-Solar Group has also built large PV power plants. It stands currently as one of the largest producers of PV electricity in Spain. In fact, it is one of the leading independent power producers (IPPs) of solar PV energy worldwide. The group has an installed capacity of 284 MWpthroughout Spain, Italy, India, Peru, Puerto Rico and California. 30 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory Figure 1.20: a-Si:H laser scribing step, P2 (cross section view). 6. Back contact deposition in physical vapour deposition (PVD) 5.72 m2in-line sputter tool: This last thin film deposition on the glass panel creates the back contact that consists of three layers which are obtained by PVD technique (Fig. 1.21). In our case this is a plasma process with pulsed direct current (DC) sputtering from ceramic and metallic targets in Argon (Ar)-atmosphere. The sputter compartment has a cathode (target) that is bombarded by the plasma atoms (Ar) to deposit metal on the glass. The chamber is evacuated with turbo pumps to achieve low base pressure (in the range of 1 ×10−6mbar - 10 ×10−6mbar), the deposition process is carried out at around 5×10−3mbar of pressure. The following layers are implemented: – Aluminium-doped zinc oxide (ZnO:Al or AZO): Acts as transparent back contact to improve the light reflection at the back metal layer and, thus, to improve the optical confinement. Layer thickness is about 90 nm, deposited with a planar target. – Aluminium (Al): Reflective and conductive layer. Its thickness is about 200 nm, deposited simultaneously with dual rotatable targets. – Nickel vanadium (NiV): Protective and solder layer. Its thickness is about 50 nm, deposited with a single rotatable target. 7. Third laser scribing, LSR3: The function of this last laser scribing is, together with LSR1, to isolate one cell from another at the back contact. LSR3 applies a Nd:YVO4 light emission, works in Q-switch pulse mode with a maximum power of 16 W and uses a wavelength of 532 nm (green). It removes the a-Si:H layer and with it, the back contact by ablating them as presented in Fig. 1.22. The thickness of the laser line is approximately 40 μm and it is separated about 110 μm from the second laser line. The death area between outside extremes of the P1 and P3 laser scribes is about 250 μm. In Fig. 1.23 a microscope view of the three laser scribes with the separation between them and the diameter of each one is presented. 1.5. T-Solar production line: Fabrication process and process control 31 Figure 1.21: PVD sputtering tool. Figure 1.22: a-Si:H and back contact laser scribing step, P3 (cross section view). 8. Tests. Quality assurance and shunt removal: Once the module is electrically finished, some basic electrical tests are done on each module in the quality assurance and shunt removal (QASR) tool. First, a shunt removal step is applied to eliminate shunts by reverse biasing the cells with a high current for a very short time. Then, the quality assurance checks the performance of every single solar cell on the panel by measuring the shunt resistance (Rsh) and open circuit voltage (Voc) under low illumination. 32 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory Figure 1.23: P1 (left), P2 (middle) and P3 (right) scribes (microscope view). 9. Final cell structure: The final structure of each solar cell is shown in Fig. 1.24. The generated current in the p-i-n junction passes vertically to the front TCO layer and it flows laterally from one cell to another through the P2 laser scribe area (LSR2 scribe) contacting the front TCO to the back contact and allowing the series connection between cells. Laser scribes 1 and 3 separate and isolate consecutive cells. Front Glass TCO a-Si Back Contact Figure 1.24: Final module structure which allows the current to flow among the cells (cross section view). 10. Panel cutout (optional): The cutting table is used only when other sizes, rather than the full size modules, are required (quarter size or half size). Fig. 1.25 presents the four different sizes available. 11. Edge deletion by an automatic seaming: In the next fabrication step, the layers at the glass edge are removed for a width of about 12 mm from the glass border. The goal is to get a module electrically well isolated and more resistant to meteorological 1.5. T-Solar production line: Fabrication process and process control 33 1.1 m 2.2 m 1.1 m 2.2 m 1.3 m 1.3 m 2.6 m 2.6 m Figure 1.25: T-Solar production line is ready to produce modules of four different sizes. inclemency. Humidity penetration caused by atmospheric exposure is prevented, as well as the current leaks at the borders and edges that can induce electric shock (the modules operate at about 200 V and 3 A). After this seaming process the module is washed again. 12. Wiring: The buss line attachment tool (see the left side of Fig. 1.26) places buss wires on the panel to retrieve the current from the module. The side buss wires are soldered on the first and last cell of the panel collecting the current of the whole panel (due to the series connection of the cells through the laser scribe). The cross buss wires are used to connect each side buss to the junction box. 13. Polyvinyl butyral (PVB) foil and front glass positioning: A PVB sheet is placed on the front panel to ensure the merge between the front and the back glass. The PVB is a hygroscopic material, thus it must be stored at low temperature (cold room) and handled under controlled temperature and humidity conditions (clean room as the one shown in the right side of Fig. 1.26). Next, the back glass is placed over the PVB. 14. Module lamination. This process consists of two stages: – Preheating at about 140 oC with the subsequent pressure roller laminator. This step deals the air removal. – Heating at about 230 oC with the subsequent pressure roller. This step seals the edges. It is important to separate these two steps considering that, to obtain a good air removal you cannot apply high temperatures, otherwise the sealing of edges would start too 34 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory Figure 1.26: Buss tool picture (left) and clean room to handle the PVB (right). early, enclosing inside some of the air. Lamination tool is shown in the left side of Fig. 1.27. 15. Autoclave (ACL): This process occurs at high pressure and temperature to ensure the complete elimination of the remaining air in the glass/PVB/glass sandwich. The process lasts about 3 h and reaches a pressure of about 12 bar - 13 bar and a temperature of 145 oC. The operation is done in a ramp of three steps: Heating, stabilization and cooling. ACL tool is shown in the right side of Fig. 1.27. 16. Junction box attachment: The junction box is an electrical connector to extract the current generated by the module (see left side of Fig. 1.28). The box contains two connectors that are soldered to the cross buss wires and a bypass diode to prevent the array from failing under partial shadow conditions since the panels are usually connected in series. Its placement consists of three phases: – Adhesion of the box to the rear glass. – Welding the pins of the junction box to the module wires. – Sealing and insulation he connections using a polymer (pottant). 1.5. T-Solar production line: Fabrication process and process control 35 Figure 1.27: Lamination tool (left) and autoclave tool (right) overall view. 17. In-line solar simulator: When the module is finished it passes through the solar simulator in order to be electrically tested (see right side of Fig. 1.28). Modules are illuminated with a xenon lamp under standard test conditions (STC, 1000 W/m2light power, AM1.5G spectrum and 25 oC temperature) and then, their electrical parameters are measured (η,I sc,V oc, FF, Rs,R sh,V mpp,I mpp and Pmpp). The data is stored for each module and is printed on a label (together with the series number and a bar code) which is glued on the backside of the module. Figure 1.28: Junction box tool (left) and in-line solar simulator tool (right). 18. Rail bonding: Four (two) rails are placed in the full size modules (quarter size modules) on the back side of the module to fix it on the support structures when they are installed in the field. 36 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory 19. Sorting and packaging the PV modules: The modules pass the last quality control, where is checked visually if they have any imperfection. Then, they are classified and packaged depending on their power. The power class achieved in the last generation of 5.72 m2full size modules is TS410, whose maximum power is 410+20 W. Fig. 1.29 shows a full and quarter size module as example. The modules have been certified according to European standards IEC 61646 and IEC 61730. In addition, the plant where they are manufactured fulfils the standards ISO 9001, ISO 14001, OHSAS 28001 and EMAS. Figure 1.29: Finished full and quarter size modules (lower left). Since the very beginning of the production line, 100% of modules are handled in automatic mode by the material handling system (MHS). The MHS is controlled by the factory automation software (FAS) system, developed by AMAT. Another important tool to automate the production line is the "SIstema de Gestión y Monitorización Avanzado" (SIGMA) tool (advanced management and monitoring tool), which is a powerful in-home 1.5. T-Solar production line: Fabrication process and process control 37 developed interface for data collection and data mining. SIGMA software tool uploads in-line and on real time data (from tools, inspection metrology, etc) to be further studied and used for panel tracking and traceability. The information provided by SIGMA is commonly used by several departments such as management, R&D, quality, engineering, production, maintenance, logistics, etc. We divide the most important tasks of SIGMA into three fields: –Line management: SIGMA offers detailed information about the actual and previous status of all the production line, e.g. tool status, tool specifications, material availability, panel location within the production line, panel history, or current and previous recipes used in each panel. –Product quality and process control: They are done in-line and off-line. The electrical part is checked in the in-line solar simulator (ISS) (illuminated current-voltage (IV) curve), QASR (voltage measurement when the panel is electrically finished) and Brightview (PECVD layer properties). The optical properties control is mostly performed off-line in the laboratory. The visual control is done in-line by the quality operators which avoid that scrap panels with defects to go forward in the production line. In Fig. 1.30 a schematic control plan with all the controls made in-line and off-line is presented. –Tool control: The different tools are also monitored and controlled with SIGMA. Therefore, all data of sensors available at the tools is stored. 38 Chapter 1. Photovoltaics status and T-Solar’s thin film silicon factory TCO • TCO RESISTIVITY: MANUAL MEASUREMENT OF THE CONDUCTIVITY (MATERIALS CONTROL PLAN). • TCO THICKNESS : 100% OF TCOS MEASURED IN THE IN-LINE BRIGHTVIEW TOOL (SIGMA ). LASER-1 CVD LASER-2 J-BOX BUSS LASER-3 PVD • HOT PANELS INTRODUCED IN-LINE FOR A FAST FEEDBACK OF THE LASER SCBRIBE QUALITY AT THE QASR. • SAMPLING CHECK IN-LINE TO CONTROL THE QUALITY OF THE LASER SCRIBES (CONTROL PLAN). • PANEL VOLTAGE VALUES MEASURED 100% AT THE QASR TOOL (SIGMA). • ALL ELECTRICAL PARAMETERS, 100% MEASURED AT SOL (SIGMA). • PANEL VOLTAGE VALUES MEASURED 100% AT THE QASR TOOL (IN SIGMA). • ALL ELECTRICAL PARAMETERS, 100% MEASURED AT SOL (DATA IN SIGMA). • FILM THICKNESS (STACK) MEASURED AT BRIGHTVIEW TOOL (100%). • REFLECTED POWER BY CHAMBER AVAILABLE IN SIGMA (100%). • INDIVIDUAL DEPOSITED LAYERS ON QUALIFICATION PANELS ARE MEASURED INLINE (BRIGHTVIEW TOOL) & OFFLINE (LAB). • PANEL VOLTAGE VALUES MEASURED 100% AT THE QASR TOOL (IN SIGMA). • ALL ELECTRICAL PARAMETERS, 100% MEASURED AT SOL (DATA IN SIGMA). • QUALIFICATION PANELS FOR INDIVIDUAL LAYERS DEPOSITED ARE MEASURED OFFLINE (LAB). • VOLTAGE MEASUREMENT IN A „LIGHT BED“ DURING THE TRIMMING PROCESS (SAMPLING UNDER ENG. CONTROL). • ALERT IN SIGMA WHEN LOOSING VOLTAGE DUE TO A WRONG SOLDERING PROCESS. • OPTICAL INSPECTION (VISION SYSTEM) TO CHECK THE CORRECT SOLDERING PROCESS. • POTTANT RATIO AND POTTANT QUANTITY CHECK. BRIGHTVIEW - THICKNESS, UNIFORMITY & ROUGHNESS (TCO AND a-SI). QASR - VOLTAGE MEASUREMENT SOLAR SIMULATOR - ELECTRICAL PARAMETERS • HOT PANELS INTRODUCED IN-LINE FOR A FAST FEEDBACK OF THE LASER SCBRIBE QUALITY AT THE QASR (QUALITY ASURANCE TOOL). • SAMPLING CHECK IN-LINE TO CONTROL THE QUALITY OF THE LASER SCRIBES (CONTROL PLAN). • PANEL VOLTAGE VALUES MEASURED 100% AT THE QASR TOOL (IN SIGMA). • ALL ELECTRICAL PARAMETERS, 100% MEASURED AT SOL (DATA IN SIGMA). • HOT PANELS INTRODUCED IN-LINE FOR A FAST FEEDBACK OF THE LASER SCRIBE QUALITY AT THE QASR (QUALITY ASURANCE TOOL). • SAMPLING CHECK IN-LINE TO CONTROL THE QUALITY OF THE LASER SCRIBES (CONTROL PLAN). • PANEL VOLTAGE VALUES MEASURED 100% AT THE QASR TOOL (IN SIGMA). • ALL ELECTRICAL PARAMETERS, 100% MEASURED AT SOL (DATA IN SIGMA). • EDGE DELETE WIDTH MEASURED ON LINE • EDGE DELETE STEP HEIGHT MEASURED IN LINE AND ON THE LAB USING A PROFILOMETER SEAM-B Figure 1.30: Control plan scheme with all the controls made in-line and off-line. 1.6 Technology and R&D Department In order to improve the efficiency and quality of the modules and to develop the technology in the factory, T-Solar has installed a scientific laboratory. With the purpose of implementing in a fast and efficient way its research activity, this department operates directly beside the production line. In the organizational structure of T-Solar, the laboratory, where the here reported PhD activity was carried out, is located within the Technology and R&D Department. Some objectives of the Department are: – To improve the process for each machine as well as for each process step of the production line. – To optimize the optical and electrical properties of the modules, in order to increase their efficiency. – To minimize the a-Si:H degradation due to illumination (Staebler-Wronski effect (SWE)). 1.6. Technology and R&D Department 39 – To develop and to implement production control techniques, such as new metrology systems, with the objective of verifying that modules meet the design specifications. The tasks carried out by the laboratory staff are divided in two fields, control plan and R&D activities. The control plan activities are, on one hand, the control of the process parameters. This includes different measurements such as thickness, resistivity or optical properties of TCO, PECVD and PVD individual layers deposited on control panels. On the other hand, it consists on the implementation of the automation of the control plan by, e.g. developing in-line metrology equipments. Concerning the R&D activities, T-Solar is collaborating with different institutions like the University of Santiago de Compostela (USC), the University of Vigo (UVI) and Galician research centers such as "Laboratorio Oficial de Metrología de Galicia" (LOMG) and "Asociación de Investigación Metalúrgica del Noroeste" (AIMEN). Besides this, the most important research activity during the last three years, 2010-2012, was performed in the frame of the European Project HELATHIS, which is the acronym of "High Efficient very LArge area THIn film Silicon photovoltaic modules". This project was created in the program call "FP7 - ENERGY 2009.2.1.1: Efficiency and material issues for thin film photovoltaics". T-Solar was the coordinator, the other partners of the project were the Belgian subsidiary of the multinational glass producer AGC, AGC Flat Glass Europe S.A. (AGC); the photovoltaic department of the Research Centre Jülich GmbH (FZJ); the Debye Institute of the University Utrecht (UU); and the Department of Applied Physics and Optics of the University de Barcelona (UB). The participants cover the whole fabrication chain from the TCO-glass (AGC) to the final a-Si:H solar cells or small modules (UU, UB and FZJ) and of the very large area (5.72 m2) a-Si:H modules (T-Solar). In fact, T-Solar was the first company in Europe fabricating modules with size larger than 5 m2. Covering the whole fabrication process allowed the consortium to implement fast optimization cycles for the TCO-glass/p-i-n/back reflector system. In the next section, more about the main aims of this project in the points related with this thesis will be explained. T-Solar also participated in other national and regional projects such as the "Torres Quevedo" program from the Spanish Ministry of Economics and Competitiveness. The investigations of this PhD thesis were performed in the frame of the HELATHIS project and the "Torres Quevedo" program. Finally, it is noteworthy to mention that a sophisticated Outdoor test station was developed over the years by the laboratory staff. It allows monitoring and measuring a large number 46 Chapter 2. Theory and simulation models of a-Si:H solar cells           Figure 2.2: Solar cell IV curve as result of the diode IV curve minus the Iph. Fig. 2.3). Hence, in eq. 2.3, we have to substitute the voltage drop at the diode, V,byV−IRs and to add the term (V−IRs)/Rsh as a sum in the second part of the equality. To get the illuminated IV curve for a-Si:H p-i-n junction solar cells we have to add a term due to current loss by recombination within the i-layer, thus we obtain [23]: I=−Iph +Iph d2 i (μτ)eff[Vbi −(V−IRs)] +I0expq(V−IRs) nkT −1+V−IRs Rsh (2.5) where Vbi (V) is the built-in voltage, which is the difference between the electron and hole quasi-Fermi levels under illuminated conditions, its value depends on the illumination level; di(cm) is the p-i-n junction thickness for the solar cell; (μτ)ef f (cm2/V), where μ(cm2/(V·s)) is the effective carrier mobility in the intrinsic region and τ(s) the effective carrier lifetime in the intrinsic region. The equivalent circuit is shown in Fig. 2.3. With this equation we obtain the typical IV curve of an a-Si:H solar cell. The term Irec = Iph d2 i (μτ)eff[Vbi−(V−IRs)] describes the recombination in the intrinsic region. It is important to consider this term in the a-Si:H technology due to the SWE (see subsection 2.2.5). This term expresses a decrease in the carriers mobility and lifetime with illumination time. 2.2. Physical model 47      Figure 2.3: Equivalent circuit of an a-Si:H solar cell. The current sink (Irec, dashed lines) takes into account the current loses due to recombination in the i-layer of the device [23]. 2.2 Physical model Silicon (Si) belongs to IV group of the periodic element table and has four electrons in its valence band that form a tetrahedral structure through covalent bonds when it forms a crystal or an amorphous structure. When doped with a V group element, such as phosphorus (P), there is an excess of electrons and provides a n-type semiconductor material (donor). On the other hand, if III group impurities are introduced, e.g. boron (B), holes are created in the network of atoms, resulting in a p-type semiconductor material (acceptor). Crystalline semiconductors are materials with a very low concentration of contaminants and a crystalline structure with a very low density of structural defects. The effort to industrialize solar cells at lower costs has led to the development of thin film semiconductor devices such as a-Si:H solar cells. The properties of a-Si:H are different from the ones of c-Si, as a consequence, solar cells based on this material, need an special device structure. Next, we present the main differences in material properties in terms of device features: –Carrier mobility: Doped a-Si:H films present a very short carrier lifetime due to its high defect density and the resulting fast recombination processes. Therefore, one uses an intrinsic (non-doped) material with low defect density as absorber layer that allows the generated carriers (electrons or holes) to have larger lifetime. The carriers are separated due to the electric field created by very thin p and n-layers deposited on top and bottom of the intrinsic layer (see Fig. 2.4). The drift length produced by the electric field is longer than the width of the intrinsic region. Therefore, the solar cell structure of amorphous silicon (a-Si:H) modules is generally p-i-n or n-i-p type. 48 Chapter 2. Theory and simulation models of a-Si:H solar cells W/E Figure 2.4: p-i-n structure. –Structure: Amorphous silicon molecules have a local tetrahedral bonding structure similar to c-Si, but the amorphous structure has an important amount of sites with a lower electronic coordination, in these sites a non-saturated bond (or dangling bond) appears. These non-saturated bonds produce electronic states with energies close to the middle of the gap (recombination centres). Therefore, the amorphous silicon films are deposited in a plasma containing a silane (SiH4)/hydrogen (H2) atmosphere. Due to the high amount of hydrogen in the plasma, hydrogen is bonded to the silicon (Si) atoms in the sites with lower electronic coordination, saturating the dangling bonds (see Fig. 2.5). This is an important process to achieve amorphous silicon with electronic device quality with a low defect density and carrier recombination rate.       Figure 2.5: Amorphous silicon structure (random network) showing a bond with a hydrogen atom. –Electronic states: Even without long range order, a-Si:H presents band states similar to the ones of c-Si. The valence band can be understood as a product of the bonding sp3 2.2. Physical model 49 orbital and the conduction band as a product of the antibonding sp3. The unsaturated bonds produce electronic states with energies close to the middle of the gap. For non-hydrogenated silicon the density of these defects is about 1020 cm−3, hydrogen introduction decreases this density to about 1016 cm−3. The disorder also produces exponential tails of localized states near the edges of the conduction and valence bands. The energy limit between extended and localized states is called the mobility edge. –Doping: Although a-Si:H can be doped with several atoms, the doping efficiency is much lower than that in c-Si. For n-type doping the minimum separation between Fermi level and conduction band is 0.25 eV. For p-type doping the minimum distance between valence band and Fermi level is 0.40 eV. The reason of lower Fermi level shift in comparison to c-Si is because most of the electrons released by the donors in n-type a-Si:H go to a defect and do not contribute to fill the conduction band. In the same way, most of the acceptors take the electron from a deep defect state in p-type a-Si:H and do not come from the valence band. Regarding the simulations, there are different software packages available for electronic device simulation in the market. Sentaurus TCAD software (Synopsys Inc.) is one of the most advanced tools. We have chosen this tool to simulate complex 2D a-Si:H solar cell structures. Sentaurus is basically designed for simulation of electronic devices. Therefore, it can be used for the simulation of a-Si:H devices since it is based on using physical models to describe the semiconductor material properties and solving the semiconductor equations. However, to simulate a-Si:H devices a lot of basic configuration work must be done. Besides the standard models and equations used for c-Si, the main features to include in an a-Si:H solar cell simulation are [24]: – Theoretical models to describe a continuous DOS distribution in the band gap of a-Si:H. – Model for recombination-generation (R-G) rate and occupation involving the localized states in the mobility gap of a-Si:H. – Modelling of graded layers. – Introduction of textured layers (front TCO). Optical modelling including scattering at the rough interfaces. 50 Chapter 2. Theory and simulation models of a-Si:H solar cells In the next sections the description of the models of a-Si:H for the DOS distribution, R-G statistics, graded layers and textured layers is introduced. 2.2.1 Density of states models for amorphous silicon The standard model of the DOS distribution consists in a parabolic conduction band (CB) and valence band (VB) as in c-Si. In addition, in a-Si:H, the continuous distribution of the DOS in the band gap strongly affects the trapping and recombination processes and therefore, the trapped charge in the localized states cannot be ignored. The localized states are formed by: –Tail states: They are modelled by an exponential distribution in the energy gap which are called conduction band tail (CBT) and valence band tail (VBT). They behave like ordinary acceptor-like states (CB tail states) or donor-like states (VB tail states). –Dangling bond states (DB+/0and DB0/−): They are simulated by adding two equal Gaussian distributions separated by an energy U. The dangling bond (DB) are amphoteric states so they can act both as acceptor-like and donor-like states and are represented by two energy levels. The different nature of the localized states in comparison to the extended states requires that different models are applied to calculate R-G statistics through localized states. It is assumed that within the mobility gap the mobility of charge carriers is zero. Next, we show the standard model of the DOS distribution in a-Si:H on a linear scale in Fig. 2.6 and on a logarithmic scale in Fig. 2.7. It is assumed that the mobility edges of conduction and valence band (Emob Cand Emob V) are equal to the connection points (Etail Cand Etail V), thus Emob C=Etail C,Emob V=Etail V. The values of DOS at the mobility edges Emob Cand Emob Vare denoted Nmob Cand Nmob V, respectively. The energy levels ECand EV, in eV, define the band gap. In the case of a-Si:H it corresponds to the so-called optical band gap, Eopt gap (eV) (see Fig. 2.6): Eopt gap =EC−EV(2.6) When considering the transport properties of carriers in a-Si:H we have to differentiate between the extended states and the localized states in the DOS distribution. The energy levels Emob Cand Emob V, define the mobility gap, Emob gap (eV) (see Fig. 2.7): Emob gap =Emob C−Emob V(2.7) 2.2. Physical model 51 Ϭ ϭнϮϮ ϮнϮϮ ϯнϮϮ ϰнϮϮ ͲϬϮϬϬϬϮϬϰϬϲϬϴϭϬϭϮϭϰϭϲϭϴϮϬ EK^ ;ĐŵͲϯ ĞsͲϭͿ ŶĞƌŐLJͲǀ;ĞsͿ ŐĂƉŽƉƚсϭϳϲĞs  s ƚĂŝůEƚĂŝů sƚĂŝůEsƚĂŝů Figure 2.6: DOS distribution model for intrinsic a-Si:H layer on a linear scale. The density in the conduction band is lower than in the valence band resulting in a non-symmetric DOS distribution. The points (Etail C, Ntail C) and (Etail V,Ntail V) are the connection between the parabolic distribution of extended states and the exponential distribution of tails. ϭнϭϰ ϭнϭϱ ϭнϭϲ ϭнϭϳ ϭнϭϴ ϭнϭϵ ϭнϮϬ ϭнϮϭ ϭнϮϮ ϭнϮϯ ͲϬϮϬϬϬϮϬϰϬϲϬϴϭϬϭϮϭϰϭϲϭϴϮϬ EK^ ;ĐŵͲϯ ĞsͲϭͿ ŶĞƌŐLJͲǀ;ĞsͿ ŐĂƉŵŽďсϭϳϴĞs ϬͲEϬͲ нϬEнϬ ŵŽďEŵŽď sŵŽďEsŵŽď h Figure 2.7: DOS distribution model for intrinsic a-Si:H layer on a logarithmic scale. It is assumed that the mobility edges are equal to the connection points (Emob C=Etail C,Emob V=Etail V). The mathematical description of the DOS distribution in a-Si:H is given by the equations presented in the following subsections [25]. 52 Chapter 2. Theory and simulation models of a-Si:H solar cells Conduction and valence band states The DOS distribution of the conduction band is given by: NCB+CBT (E)=NCB(E)for E≥Etail C NCBT (E)for E≤Etail C (2.8) NCB(E)=N0 C(E−EC)(1/2)(2.9) NCBT (E)=Ntail Cexp−(Etail C−E Etail C0 )(2.10) The DOS distribution of the valence band is given by: NVB+VBT(E)=NVB(E)for E≤Etail V NVBT(E)for E≥Etail V (2.11) NVB(E)=N0 V(EV−E)(1/2)(2.12) NVBT(E)=Ntail Vexp−(E−Etail V Etail V0 )(2.13) where N0 Cand N0 V,incm −3, are the parameters which describes the parabolic distribution of states in the conduction and valence band respectively; Etail C0and Etail V0, in eV, are the characteristic energies describing the decay of the CB and VB tails respectively and (Etail C, Ntail C) and (Etail V,Ntail V) are the connection points of the parabolic and exponential part of the conduction and valence band, respectively. Dangling bond states To describe the continuous distribution of the DB states in the band gap, a Gaussian distribution is used. A DB can be in three charge states: positive (D+), neutral (D0) and negative (D−). A defect with three possible charge states acts in good approximation like a group of two defects consisting of a donor-like state (DB+/0) and an acceptor-like state (DB0/−) and is therefore represented by two energy levels E+/0and E0/−in the band diagram, respectively. These energy levels are called the transition energy levels, they are separated from each other by a correlation energy, U(eV), which is the energy needed to add the second electron to a singly occupied (neutral) DB. Uis assumed to be constant and positive. Under these assumptions the DB are represented by two equal Gaussian distributions in the band 2.2. Physical model 53 diagram separated from each other by a distance U. The equations to represent the density of DB states are the following: NDB+/0(E)= Ntot DB σDB√2πexp−(E−E+/0 DB )2 2σ2 DB (2.14a) NDB0/−(E)= Ntot DB σDB√2πexp−(E−E0/− DB )2 2σ2 DB (2.14b) where Ntot DB (cm−3) is the total density of defects, E+/0 DB and E0/− DB , in eV, are the energies of the peaks of Gaussian distributions for the donor-like states DB+/0and the acceptor-like states DB0/−, respectively and σDB (eV) is the standard deviation of the distribution. To correlate NDB+/0and NDB0/−and E+/0 DB and E0/− DB one uses the next relations: NDB0/−(E)=NDB+/0(E+U)(2.15) E0/− DB =E+/0 DB +U(2.16) Concentration of charge carriers The states in the conduction band above the mobility edge Emob Care extended states. They are populated with electrons that are characterized by the concentration nand the extended-state mobility μn>0. The states in the valence band below the mobility edge Emob Vare also extended states. They are populated with holes that are characterized by the concentration pand the extended-state mobility μp>0. In the operational range of a-Si:H solar cells the dominant electric transport mechanism is multiple trapping and release. The transport is then characterized by the concentration of carriers in the extended states (n,p) and the extended-state mobilities (μn,μp). Furthermore, it is assumed that the mobilities are the same in thermal equilibrium and in steady state illumination conditions. For the concentration of charge carriers we can write [25]: n=Neff Cmob expEFN −Emob C kT (2.17a) p=Neff Vmob expEmob V−EFP kT (2.17b) where Neff Cmob and Nef f Vmob ,incm −3, are the effective density of states at the CB and VB mobility edge. In case of a-Si:H the effective density of states NCand NVare equal to Nef f Cmob and Nef f Vmob , respectively. 54 Chapter 2. Theory and simulation models of a-Si:H solar cells 2.2.2 Recombination-generation statistics in amorphous silicon The recombination process in crystalline semiconductors is typically dominated by a single energy level of a recombination center in the band gap. However, in the band gap of a-Si:H there is a continuous density of allowed states which contribute to the net R-G rate. Hence, one needs to integrate the recombination rate contributions from the gap states over the whole band gap. Under the assumption that the recombination centres are non-interacting, the net R-G rate, Rnet (cm−3s−1), can be calculated from: Rnet =Emob C Emob V N(E)ηR(E)dE(2.18) where ηR(E)is the recombination rate contribution of a state at energy Eand N(E)is the DOS as function of energy in the band gap. The trapped charge in tail states is calculated using the following equations for ordinary donor-like and acceptor-like states, respectively: ρD=qEmob V Emob C ND(E)[1−f(E)]dE(2.19) ρA=−qEmob V Emob C NA(E)f(E)dE(2.20) where f(E)is the occupation function [26] [27]. In case of the amphoteric DB states the space charge is given by: ρDB =qEmob V Emob C NDB(E)[F+(E)−F−(E)]dE(2.21) In this equation F+(E)and F−(E)are the occupation functions of empty and doubly occupied DB, respectively. The assumption of non-interacting centres means that the localized states in the band gap can only interact with carriers in the extended states of the conduction and valence bands. This assumption allows to use the Shockley Read Hall (SRH) R-G statistics [28] to model the recombination process through the single-level states and the Sah and Shockley [29] multi-level R-G statistics for the amphoteric DB states. The different types of the localized states in the band gap of a-Si:H and the models that are used to calculate the recombination rate through these states and their charge occupation are schematized in Fig. 2.8. 2.2. Physical model 55 >ŽĐĂůŝnjĞĚƐƚĂƚĞƐŝŶƚŚĞďĂŶĚŐĂƉŽĨ ĂͲ^ŝ, dĂŝůƐƚĂƚĞƐ ĂŶŐůŝŶŐďŽŶĚƐƚĂƚĞƐ sƚĂŝůƐƚĂƚĞƐ ;ŽŶŽƌͲůŝŬĞƐƚĂƚĞƐͿ ƚĂŝůƐƚĂƚĞƐ ;ĐĞƉƚŽƌͲůŝŬĞƐƚĂƚĞƐͿ ŵƉŚŽƚĞƌŝĐ ƐƚĂƚĞƐ ^ŚŽĐŬůĞLJͲZĞĂĚͲ,Ăůů ;ƐŝŶŐůĞͲůĞǀĞů ƐƚĂƚŝƐƚŝĐƐͿ ^ĂŚΘ^ŚŽĐŬůĞLJ ;ŵƵůƚŝͲůĞǀĞůƐƚĂƚŝƐƚŝĐƐͿ Figure 2.8: Different types of localized states in the band gap of a-Si:H and models that are used to calculate the recombination rate and charge occupation [25]. R-G statistics of CB and VB tail states In the energy band diagram, the donor and acceptor R-G center is represented by a single energy level. The CB tail states behave like ordinary acceptor states and are neutral (T0 A)or negative (T− A). The VB tail states behave like ordinary donor states and are neutral (T0 D)or positive (T+ D). The theory that describes the recombination process involving a single energy level in the band gap of a semiconductor was developed by Shockley and Read [28] and Hall [30], [31]. The SRH theory is based on four possible transitions between an energy level ETin the band gap and the extended states of the CB and VB: 1. Electron capture at an R-G center. 2. Electron thermal emission from an R-G center. 3. Hole capture at an R-G center. 4. Hole thermal emission from an R-G center. 62 Chapter 2. Theory and simulation models of a-Si:H solar cells –Electron and hole effective masses: The electron and hole effective masses are defined with the next formula [36]: me m0=NC,300 2.540×1019 2/3 (2.43a) mh m0=NV,300 2.540×1019 2/3 (2.43b) where we set NC,300 =NV,300 =1.0×1020 (cm−3) which are the effective DOS in the CB and VB, respectively at a temperature of 300 K. To know the effective DOS at any temperature we apply the next equation: NC(T)=NC,300 ·(T/300)3/2(2.44a) NV(T)=NV,300 ·(T/300)3/2(2.44b) –Band gap and electron affinity: In crystalline semiconductors, the band gap (EG)is the difference between the lowest energy in the CB and the highest energy in the VB (Eopt Gin Fig. 2.6). In a-Si:H based material the EGrepresents the mobility gap (Emob Gin Fig. 2.7). The electron affinity (χ) is the difference between the lowest energy in the CB and the vacuum level (E0). Sentaurus models the lattice temperature–dependence of the band gap as [37]: EG(T)=EG(0)−αT2 T+β(2.45) where we have chosen β=0 and α=0.0004 eV/K [38]. The effective band gap results from the reduction caused by the temperature and doping concentration. It is presented in the next equation: EG,eff(T)=EG(T)−EBGN (2.46) Another parameter that can be obtained from the previous ones is the intrinsic density: ni(T)=NC(T)NV(T)·exp−EG(T) 2kT (2.47) 2.2. Physical model 63 The effective intrinsic density (including doping-dependent band gap narrowing) is: ni,eff =ni·exp−EBGN 2kT (2.48) where EBGN (eV) is the energy reduction in the band gap due to high doping level. Fig. 2.11 shows the band diagram for the a-Si:H p-i-n layers under illumination and without polarization. One can appreciate that the doped layers are much thinner than the intrinsic layer which has sloped energy bands as consequence of the electric field created by the doped layers. One also sees a larger energy gap for the p-layer due to its carbon content (a-SiC:H) which increases the energy band gap with the objective to absorb less photons in this region. 0 1 2 e rgy,E(eV) Ec Ev Efn Efp Ͳ2 Ͳ1 0 50 100 150 200 En e Depth,d(nm) iͲlayer nͲlayer pͲlayer Figure 2.11: Band diagram for the a-Si:H p-i-n layers under 1 sun illumination and without polarization (V=0 V). The drop in EFN at the beggining of the p-layer is because the anode is placed at the TCO/p interface. 2.2.5 Light-induced degradation. Staebler-Wronski effect The a-Si:H presents a light-induced metastability of its electronic properties (also known as Staebler-Wronski effect, SWE) [39], [40]. It causes a reduction of the conversion efficiency of a-Si:H solar cells due to light exposure. The rate of degradation during continuous illumination at 1 sun (1000 W/m2) intensity is high during the first tens of hours, but decreases over time. Finally, the cell performance stabilizes after several hundreds of hours. 64 Chapter 2. Theory and simulation models of a-Si:H solar cells The initial efficiency can be completely recovered by annealing of the device at about 150 oC for several hours. The reversibility of the device performance shows that the initial loss is not due to diffusion of ions or dopants, nor to other irreversible processes, but it is due to the increase of DB defects created in the i-layer which act as recombination centres for the photogenerated carriers. The metastable DBs have an energetic and spatial distribution, which determines their charge state. The space charge distribution of the charged states then, modifies the internal electric field profile which in turn has an effect on the carrier collection in the device [25]. The main material properties that could play a role in the SWE are the concentration of impurities, the hydrogen concentration and its complex bonding structure and the disorder of the Si network. The most commonly observed effects in the electrical parameters of p-i-n a-Si:H solar cells are: – The biggest relative changes occur in the fill factor (FF), the relative changes in short circuit current density (Jsc) and open circuit voltage (Voc) are significantly smaller. – Solar cells with a thick intrinsic layer degrade more than those with a thin intrinsic layer. – Solar cells with a high impurity concentration (above 1018 cm−3) in the i-layer degrade stronger than those with high purity i-layer. – Solar cells operated at elevated temperatures (60 oC-90oC) stabilize at a higher efficiency than those operated at room temperature or below. – Cyclic exposure results into a higher efficiency stabilization than continuous exposure. – Exposure at high intensity illumination causes stronger degradation than 1 sun. Illumination levels less than 1000 W/m2lead to reduced degradation. In order to reduce the SWE related degradation of a-Si:H solar cells several methods have been investigated, e.g.: – To make the i-layer as thin as possible in order to maintain a high electric field after degradation. At the same time, the absorbed fraction of the incident light is maximized by using optical light confinement techniques made possible by textured electrodes and enhanced multilayer back reflectors. 2.2. Physical model 65 – To assist the transport of the minority carriers (holes) in the low field region by band gap profiling [41]. – To redistribute the field by using graded low-level impurity doping [42]. – To stack layers with different carrier mobilities, making a mobility grading [43]. 2.2.6 Baseline input parameters applied in the simulation The here simulated structure is TCO/a-SiC:H/a-Si:H/a-Si:H/AZO/Al implemented in the T-Solar module fabrication. In the simulation we assume an ideal tunnel contact at the front contact and an ideal ohmic contact at the back one. To implement this, we fix the front contact at the TCO/p-layer interface and the back contact at the end of the Al layer. Hence, the electrical features of the TCO are not considered, only its optical characteristics are taken into account. The p-i-n single junction solar cell is shown in Fig. 2.12. TCO Front cont act i-layer AZO Al Back contact p-layer n-layer Glass Figure 2.12: Structure of the simulated a-Si:H solar cell. 66 Chapter 2. Theory and simulation models of a-Si:H solar cells Table 2.3 shows the baseline input parameters for the initial state used in the simulation. The values are adjusted to fit the electrical parameters and the spectral response of our typical 1cm 2laboratory a-Si:H solar cells produced in the T-Solar production line. To simulate the stabilized state (light-soaked state) we only need to increase the maximum density of dangling bonds (NDB) in the i-layer [44]. Table 2.3: Baseline input parameters used in the simulation. Some parameters were taken from [25] p-layer i-layer n-layer MATERIAL PARAMETERS d (nm) 10-30 200-300 10-20 Doping (cm−3)3×1018 1×1015 8×1018 εr7.211.911.9 χ(eV) 3.90 4.00 3.99 Emob gap (eV) 1.95 1.78 1.80 μn(cm2/(V·s)) 20 20 20 μp(cm2/(V·s)) 5 5 5 NC(cm−3)1×1020 1×1020 1×1020 NV(cm−3)1×1020 1×1020 1×1020 TAIL STATES PARAMETERS Ntail C(cm−3/eV) 2×1021 8×1021 1×1021 Ntail V(cm−3/eV) 1×1021 4×1021 2×1021 Etail C0(eV) 0.180 0.032 0.070 Etail C0(eV) 0.090 0.047 0.160 C− p,C+ n(cm3/s) 1×10−81×10−81×10−8 C0 p,C0 n(cm3/s) 1×10−10 1×10−10 1×10−10 DANGLING BOND STATES PARAMETERS σ(eV) 0.144 0.144 0.144 Ntot DB (cm−3)8×1018 5×1015 2×1019 E+/0 DB (eV) (from CB) −0.70 −0.89 −1.40 U (eV) 0.20 0.20 0.20 C− p,C+ n(cm3/s) 4×10−88×10−94×10−8 C0 p,C0 n(cm3/s) 4×10−98×10−10 4×10−9 2.3 Optical model The optical model is based on the optical generation in the solar cell which depends on the transmission and reflection coefficients of the different layers of the solar cell, the light 2.3. Optical model 67 scattering and light trapping properties of the device and the spectral irradiance of the sun or an equivalent light source. 2.3.1 Optical generation The optical generation in a solar cell is expressed by the number of electrons and holes generated by the incident photons per cm3and second in a concrete position inside the device. The amount of photons penetrating the absorber material decay exponentially when going deeper into the semiconductor as we see in Fig. 2.13.                   Figure 2.13: Photon flux into a semiconductor. The photon flux, φphotons cm2s, is given by the Lambert law as shown in the next equation: φ(x,λ)=φ(0,λ)exp−α(λ)x(2.49) where α(cm−1) is the absorption coefficient which is different for every semiconductor (see Fig. 2.14). The effective range of absorption of a-Si:H is from about 300 nm to 800 nm. 68 Chapter 2. Theory and simulation models of a-Si:H solar cells The optical generation, Gopt 1 cm3s, of a semiconductor, depending on the position and the wavelength, results in: Gopt(x,λ)=−dφ(x,λ) dx =α(λ)φ(0,λ)exp−α(λ)x(2.50) ϭнϬϭ ϭнϬϮ ϭнϬϯ ϭнϬϰ ϭнϬϱ ϭнϬϲ ϮϬϬ ϰϬϬ ϲϬϬ ϴϬϬ ϭϬϬϬ ϭϮϬϬ ϭϰϬϬ ϭϲϬϬ ϭϴϬϬ ďƐŽƌƉƚŝŽŶĐŽĞĨĨŝĐŝĞŶƚα α α α;ĐŵͲϭͿ tĂǀĞůĞŶŐƚŚλ λλ λ;ŶŵͿ ^ŝ 'Ğ Ě^ ĂͲ^ŝ, 'ĂƐ Figure 2.14: Optical absorption coefficient for different PV materials. Concerning the complex refractive index (n), it is used to describe the propagation and absorption of light in a media: n=n−ik (2.51) The refractive index (n) relates the speed of light in vacuum (c) and in the media (v): n=c v(2.52) The extinction coefficient (k) describes the absorption of the light in the media: k=αλ 4π(2.53) We have added in the material library of every layer in Sentaurus a table with the complex refractive index in the a-Si:H wavelength range of interest (300 nm - 800 nm). In Fig. 2.15 we see the refractive index and the extinction coefficient for the three p-i-n layers. We cannot measure the coefficients in the range 300 nm - 400 nm therefore, we extrapolate the trend of each one to get the data. 2.3. Optical model 69 3 4 5 6 d ex(n)andextinction e fficient(k) nfor p Ͳla y er 0 1 2 300 400 500 600 700 800 Refractivein d co e Wavelength, O (nm) p y nforiͲlayer nfornͲlayer kforpͲlayer kforiͲlayer kfornͲlayer Figure 2.15: Complex refractive index for our a-Si:H layers. 2.3.2 Improvement of light trapping and scattering through textured interfaces A crucial aspect to improve the current generation in the solar cell having thin intrinsic layers is the high dispersion at the front TCO, which is described by the haze factor. The haze is the ratio between the diffuse transmission (Tdif ) and the total transmission (T), as indicated in Eq. 2.54. Thus, a front TCO with high haze will lead to more scattered light in the device and longer absorption, increasing the probabilities of photons to be absorbed in the solar cell. This is achieved by texturing the front TCO layer which results in textured interfaces as shown in Fig. 2.12. Apart of good scattering, to increase the current generation in the 550 nm - 800 nm wavelength range, we need a highly reflective back reflector to take advantage of the non-absorbed photons in the i-layer which could be absorbed after rebounding in the back reflector. This improvement would not be possible in the long wavelength range without both the textured front TCO and the back reflector. Haze =Tdif T(2.54) To simulate the texture we have used two options: the Hegedus model (assuming a light enhancement factor, which is an approximation) and to build a regular triangular structure (which is a more precise method). 70 Chapter 2. Theory and simulation models of a-Si:H solar cells –Hegedus model: To simulate the light trapping effect we use a method to calculate an optical path enhancement factor m(λ)depending on the wavelength range using the model developed by Hegedus et al. [45]. Thus, the absorption of the i-layer (Ai) changes from Eq. 2.55a to 2.55b. This provides an adjustment of the spectral response to the respective experimental data in the long wavelength range. Ai(λ)=1−exp[−α(λ)·di](2.55a) Ai(λ)=1−exp[−α(λ)·m(λ)·di](2.55b) The use of non-textured surfaces/interfaces implies the formation of interferences in the spectral response depending mainly on the TCO and i-layer thickness. Therefore, to suppress the interference effects coming from the TCO, we implement a very thin TCO layer (d=75 nm, when usually d≈700 nm - 1000 nm). We also applied a respectively enhanced extinction coefficient (k) to approximate the real absorption properties of the TCO in the device. –Geometry with texture: We took the front TCO AGC AN10 as experimental texture reference. This TCO type is not any more the standard TCO at T-Solar, which now is ANS10ME. However, the texture of both can be considered nearly identical. Scanning electron microscopy (SEM) and atomic force microscopy (AFM) pictures of the AN10 TCO texture have been studied [46]. One example of these measurements is presented in Figs. 2.16 and 2.17. They show how the texture is irregular having, in general, pyramidal shape. To simulate the texture in 2Dwe have employed, for the interface TCO/p-layer, a regular triangular structure with height and base using the average parameters obtained from the AFM data. In the interfaces i-layer/n-layer and AZO/Al we have applied a scaled coefficient to slightly reduce the height of the texture [47], [48], [49], [50]. 2.3. Optical model 71 Figure 2.16: SEM picture of an industrial TCO sample AN10 from AGC showing fine surface roughness of the crystal grains [46].  Figure 2.17: AFM picture of an industrial TCO sample AN10 from AGC [46]. 78 Chapter 2. Theory and simulation models of a-Si:H solar cells ηn=EFn −EC kT (2.66a) ηp=EV−EFp kT (2.66b) in the equations above NCand NV,incm −3, are the effective DOS in the CB and VB. Their values are given by Eq. 2.44; F1/2is the Fermi integral of order 1/2; EFn =qφnand EFp =qφp are the quasi-Fermi energies for electron and holes, and φnand φp, in V, are the quasi-Fermi potentials for electron and holes. Alternatively, one can write the equations 2.64 as: n=γn·NC·expEFn −EC kT =γn·NC·expEFn −E0+qψ+χ kT (2.67a) p=γp·NV·expEV−EFp kT =γp·NV·expE0−qψ−χ−EG−EFp kT (2.67b) where γnand γpare function of ηnand ηp: γn=F1/2(ηn) exp(ηn)(2.68a) γp=F1/2(ηp) exp(ηp)(2.68b) Working out the value of EFn and EFp, respectively, we obtain: EFn =EC+kT lnn γnNC=E0−qψ−χ+kT lnn γnNC(2.69a) EFp =EV−kT lnp γpNV=E0−qψ−χ−EG−kT lnp γpNV(2.69b) Current transport. Drift-Diffusion model Using equations 2.69a and 2.69b in equations 2.61a and 2.61b, respectively, and after a manipulation which makes use of the Einstein relations for the diffusion coefficient for electrons and holes: Dn=kT qμn(2.70a) Dp=kT qμp(2.70b) and the fact that dE0 dx =0, we find: 2.4. Numerical methods and solvers 79 Jn=qDn dn dx   Electron diffusion current +μnn−qdψ dx −dχ dx −kT NC dNC dx    Electron drift current (2.71a) Jp=−qDp dp dx   Hole diffusion current +μpp−qdψ dx −dχ dx −dEG dx +kT NV dNV dx    Hole drift current (2.71b) In Equations 2.71 the first term represents the current due to diffusion of the carriers and the second term represents drift transport. The terms in the square brackets are considered to be the effective drift fields. In the case that the device is made of spatially uniform material (homojunction device), we obtain: dχ dx =dEG dx =dNC dx =dNV dx =0 (2.72) Therefore, the Eqs. 2.71 result in the classical current density expressions for electron and hole, which are used in c-Si homojunction non-degenerated devices analysed with Maxwell-Boltzmann statistics. 2.4.3 Boundary conditions There are two boundaries in the device, the front and the back contact. The implied conditions at the contacts fix the values of the independent model variables at these points. The boundary conditions depend on how the contacts of a device are modelled. Generally, two types of contacts are distinguished: Ohmic contacts and Schottky contacts. Our device is modelled with Ohmic contacts where the electrostatic potential and electron and hole concentrations at the boundaries of the device (x=0, anode and x=L, cathode, where Lis the device thickness) are fixed, or the minority carrier concentrations are determined by surface recombination. It is assumed that the majority carrier concentration is independent of the injection level. In case of ideal Ohmic contacts infinite surface recombination, equilibrium and charge neutrality at the contacts is assumed: ρ(0,y)=0⇒n0(0,y)−p0(0,y)=ploc −nloc +ND−NA(2.73a) ρ(L,y)=0⇒n0(L,y)−p0(L,y)=ploc −nloc +ND−NA(2.73b) n0(0,y)·p0(0,y)=n2 i,eff (2.74a) n0(L,y)·p0(L,y)=n2 i,eff (2.74b) 80 Chapter 2. Theory and simulation models of a-Si:H solar cells The space charge density ρ(x)is given by Eq. 2.60. The following conditions are introduced for the electrostatic potential ψand quasi-Fermi potentials φFn and φFp at the boundaries: ψ(0,y)=ψ0(0,y)+Vapp (2.75a) ψ(L,y)=ψ0(L,y)(2.75b) φFn(0,y)=φFp(0,y)=Vapp (2.76a) φFn(L,y)=φFp(L,y)=0 (2.76b) where ψ0(0,y)and ψ0(L,y)are the solutions of equations 2.73 andVapp is the applied external voltage. For Boltzmann statistics, these conditions can be expressed analytically. For Fermi statistics this is not possible, hence, Sentaurus Device computes the equilibrium solution numerically. Furthermore, for Ohmic contacts the surface recombination velocities for electrons, νn and holes, νp, in cm/s, determine the carrier concentrations at the boundaries. If they are specified, Sentaurus Device uses the following current density boundary conditions:  Jn·ˆn=qνn·(n−n0)(2.77a)  Jp·ˆn=qνp·(p−p0)(2.77b) where nand pare the electron and hole concentrations at the front and back contacts, n0and p0are the electron and hole concentrations in thermodynamic equilibrium at the front and back contacts and ˆnis the normal vector to the contact. Replacing the general point xby x=0 and x=L, we find:  Jn(0,y)·ˆn=q·νn0·[n(0,y)−n0(0,y)] (2.78a)  Jn(L,y)·ˆn=q·νnL ·[n(L,y)−n0(L,y)] (2.78b)  Jp(0,y)·ˆn=q·νp0·[p(0,y)−p0(0,y)] (2.79a)  Jp(L,y)·ˆn=q·νpL ·[p(L,y)−p0(L,y)] (2.79b) where νn0and νnL are the surface recombination velocities for electrons at the front and back contact, respectively, and νp0and νpL are the surface recombination velocities for holes at the front and back contact, respectively. By default, n=n0and p=p0are applied for concentrations at the Ohmic contacts since  Jn,p=0 at the boundaries. 2.4. Numerical methods and solvers 81 2.4.4 Discretization The model described previously is formed by the differential equations which model the behaviour of the corresponding semiconductor. To solve the model we need to look for a numerical solution of the system of equations together with their boundary conditions. This, results in a non-linear system of partial differential equations. The procedure used to solve this system of a simulated semiconductor device is the following [59]: – To discretise the system of non-linear equations. Hence, the continuous problem is replaced by a discrete non-linear system of equations. – To apply any linearization method to the non-linear problem. – To solve a system of linear and disperse equations in order to get the wished solution. To solve the non-linear system of 3N(where Nis the number of nodes) differential equations formed by the Poisson and the continuity carrier equations (2.57, 2.58 and 2.59) one needs to discretise the system of equations. In spite that our simulation is in 2D, we are going to simplify in the following the discretization to 1D since the resolution methods are not the aim of this thesis, they are presented just to clarify how the simulations steps are. The discretization starts dividing the device in cells. In every cell there are two kinds of points: the cell edges which are called ior x(i)and the central points i+1/2orx(i+1/2). The anode (positive electrode) or front contact is placed at the point 0 and the cathode (negative electrode) or back contact is placed at the point N−1. The equations relate three variables at the point i:ψ(i),n(i)and p(i). Therefore, there are 3Nvariables with 3Nequations in total. In addition, the three equations at each iposition depend not only on the three variable values at this location but also on the variable values at the adjacent points i−1 and i+1. To obtain the discretised equations at a point i(which is not a contact) of the cited equations 2.57, 2.58 and 2.59 we have to integrate them with limits i−1/2 and i+1/2, considering hi=x(i+1)−x(i)and hi−1=x(i)−x(i−1), we get: εi+i/2 ψi+1−ψi hi−εi−i/2 ψi−ψi−1 hi−1=hi+hi−1 2(n−p+NA−ND)|i(2.80) Jn(i+1/2)−Jn(i−1/2)=IR(i)(2.81) Jp(i+1/2)−Jp(i−1/2)=−IR(i)(2.82) 82 Chapter 2. Theory and simulation models of a-Si:H solar cells where IR(i) is the integral of the net recombination (Rnet ) which was obtained in Eq. 2.41, resulting in: IR(i)=i+1/2 i−1/2(Rtot CB +Rtot VB+Rtot DB)dx (2.83) 2.4.5 Linearization of the discretised system The coupled non-linear system of 3Nequations, discretised in the previous section, is solved using the Newton method [60]. To obtain the linear system we have to take into account that in every point ithere are 3 discretised equations (Poisson and continuity carrier equations). They can be written equalling them to zero: Gi 1=0, Gi 2=0 and Gi 3=0. Due to the discretization, the equations only depend on three variables (ψ,nand p) in three points (i−1, iand i+1). We name Withe vector whose components are the variables in this point (Wi=(ψi,ni,pi)) and dWiits differential vector. Then, we apply the Taylor series development of the functions Gi 1,Gi 2and Gi 3for a range of values of the variables together with the Newton method. It leads in each iteration to three linear equations which can be represented in matrix as follow: ⎛ ⎜ ⎜ ⎝ ∂Gi 1 ∂Wi−1 ∂Gi 1 ∂Wi ∂Gi 1 ∂Wi+1 ∂Gi 2 ∂Wi−1 ∂Gi 2 ∂Wi ∂Gi 2 ∂Wi+1 ∂Gi 3 ∂Wi−1 ∂Gi 3 ∂Wi ∂Gi 3 ∂Wi+1 ⎞ ⎟ ⎟ ⎠·⎛ ⎜ ⎝ dWi−1 dWi dWi+1⎞ ⎟ ⎠=⎛ ⎜ ⎝ −Gi 1 −Gi 2 −Gi 3 ⎞ ⎟ ⎠(2.84) 2.4.6 Numerical resolution of the non-linear equations To solve the system 2.84, coming from the drift-diffusion model, two approximation methods are usually employed [61] [59] [62], the Newton-Raphson method (coupled) and the Gummel method (uncoupled). To choose between both methods, one has to consider the operation characteristics of the device, the algorithms used to solve the linearised system of equations or the amount of memory of the computational system where the simulations are run. Newton-Raphson method The Newton-Raphson method consists in solving the complete system of 3Nequations simultaneously by applying any iterative method of type Newton [63]. This method is very robust and provides a very accurate solution. However, it may need a lot of time and memory 2.4. Numerical methods and solvers 83 resources since it solves a non-linear and non-symmetric system of equations with high number of dimensions. Gummel method The Gummel method [59] [64] uses an iteration type Gauss-Seidel/Jacobi which uncouples the equations G1,G2and G3, so that it is only necessary to solve three systems of equations of dimension N. Hence, the Poisson equation and the continuity equations of electron and holes are solved separately. This method is very useful since it is possible to reach the convergence even starting with poor initial conditions and it is a fast method. Nevertheless, for some applications such as very high injection level in the semiconductor or high recombination, the method could have convergence problems. In addition, in some cases the Gummel method converges very fast in the first iterations but later the convergence turns slower. In this case, one can combine the Gummel and Newton-Raphson methods. Thus one can approximate the solution by using the Gummel method and later change to the Newton-Raphson method in order to take advantage of its quadratic convergence properties near the solution. CHAPTER 3 INDUSTRIAL PRODUCTION PROCESS AND BASIC EQUIPMENT SET-UPS OF A-SI:H SOLAR CELLS This chapter presents how the PECVD production process works, as well as how we fabricate R&D solar cells of 1 cm2(or 4 cm2) or mini modules in the industrial production line. Later, the equipment installed in the laboratory are shown. Concerning this thesis, the most important ones are the conventional spectral response and the very fast spectral response equipment, which are used to determine the external quantum efficiency (EQE) curves; and the solar simulator equipment with its current-voltage (IV) tracer to measure the illuminated or dark IV curves. The three equipment were developed in our laboratory. The development of R&D solar cells and mini modules in the T-Solar production line as well as the measurement equipment have been crucial to improve the production modules’ efficiency. 3.1 Plasma enhanced chemical vapour deposition As explained in section 1.5, the solar cell fabrication process is based on the deposition of a-Si:H layers by PECVD in a 7 chamber cluster tool on a 2.2mx2.6 m substrate. The first layer deposition is a p-type layer doped with trimethyl borane (TMB, B(CH3)3), the second one is an intrinsic layer and, at last, a n-type layer doped with PH3. The p-layer is deposited in one chamber while the other two are deposited together in another chamber to prevent 86 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells cross contamination. The a-Si:H layers are formed through deposition of SiH4and H2. This deposition of a p-i-n structure allows the generation of the current in the cell when the module is exposed to the light. The thin film deposition process results from the decomposition of SiH4,H 2and other gases creating a plasma at moderate substrate (glass film) temperatures, around 200 oC (PECVD). In this process the silicon forms an amorphous layer mixed with hydrogen in a random bond net (random network, see Fig. 2.5). PECVD uses electrical power coupled into the plasma at a radiofrequency (RF) of 13.56 MHz to create radicals and ions of incoming gases, so that they can react to form silicon layers on the substrate glass. The whole system is kept in vacuum by a mechanical roughing system. The chambers are cleaned by dissociating nitrogen trifluoride (NF3) which removes Si remaining in the chamber. The fluorine (F) atoms formed in the plasma react with Si to a volatile molecule, silicon tetrafluoride (SiF4), which then, are pumped out of the chamber. The deposition process by an RF discharge can be described as a four step process [25], it is schematically represented in Fig. 3.1 [65]: – The primary reactions in the gas phase are electron-impact excitation, dissociation and ionization of SiH4molecules. The plasma thus, consists of neutral radicals and molecules, positive and negative ions and electrons. – Secondary reactions, between molecules and ions or radicals, are very important as they predominantly control the electronic and structural film properties. Reactive neutral species move to the substrate by diffusion, positive ions bombard the growing film and negative ions are trapped within the sheaths [66] of the counter electrode and may eventually form small particles or dust. – The third step consists of surface reactions, such as hydrogen abstraction, radical diffusion and chemical bonding. – The fourth step is the subsurface release of hydrogen molecules and relaxation of the silicon matrix. The deposition process is a very complicated matter as the physical and chemical interactions in the plasma and at the growing film surface are dependent on the RF power and frequency, the substrate temperature, the gas pressure and composition, the magnitude and the pattern of the gas flow, the electrode geometry, etc. 3.1. Plasma enhanced chemical vapour deposition 87 ZĞĐŽŵďŝŶĂƚŝŽŶ ^ŝ,ϰн,Ϯ ^ŝ,ϰ,Ϯ ĐƚŝǀĞƐ Ɖ ĞĐŝĞƐ ;LJсϬ ϭϮϯͿ ;džфϭϬLJф΀ϮdžͲϭ΁Ϳ ;njсϭϮͿ ;LJсϬ ϭϮϯͿ ŚĞŵŝĐĂůĂĚƐŽƌƉƚŝŽŶŽĨƌĂĚŝĐĂůƐĂƚŽŵƐ ,ͲĂƚŽŵƌĞĂĐƚŝŽŶƐ /ŽŶďŽŵďĂƌĚŵĞŶƚ /ŶƐĞƌƚŝŽŶ ^ƵƌĨĂĐĞƌĞĂĐƚŝŽŶƐĚŝĨĨƵƐŝŽŶ ^ƵďƐƵƌĨĂĐĞƌĞĂĐƚŝŽŶƐ,ϮƌĞůĞĂƐĞ^ŝƌĞůĂdž ^ŽůŝĚ WůĂƐŵĂ ƚĐŚŝŶŐ ĚĞƐŽƌƉƚŝŽŶ ůĞĐƚƌŽŶŝŵƉĂĐƚĚŝƐƐŝŽŶŝnj ,ͲĂƚŽŵĂďƐƚƌĂĐƚŝŽŶ /ŽŶͲŶĞƵƚƌĂůƌĞĂĐƚŝŽŶƐ + y SiH y SiH yx HSi + yx HSi H + z H Figure 3.1: Schematic representation of the glow-discharge deposition process [25]. In Fig. 3.2 we see the main parts of a PECVD chamber as well as their distribution and the place of the substrate. 94 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells Figure 3.7: Back contact side (left) and glass side (right) of a coupon produced at TS through laser scribe. high efficiency but often in bad cell yield. However, combining SB and annealing we achieve high efficiency, high yield and low dispersion in electrical parameters. Together with the very accurate cell area definition on coupons, MCs produced with LSR scribe method return very reliable data. The production process can be summarized as: Glass →TCO (SnO2)→a-Si:H →LSR scribe (P2) for grids →ZnO →Al →NiV →LSR scribe (P3) for pads →SB →Annealing The design through LSR scribe allows a flexible design, which finally has permitted the fabrication of MC sizes with 1 cm2and4cm 2, as well as MMs of several sizes (usually 10 x 10 cm2and 20 x 20 cm2of total or active surface). As an example, Fig. 3.8 presents the location of 18 MMs 20 x 20 cm2and 18 MMs 10 x 10 cm2in the full size panel. The overall design of these MMs is the same as the one of full size modules having the same measurements for the edge delete area and the contact cells at the module border. An example of a 10 x 10 cm2MM design is presented in Fig. 3.9. To measure the MMs we do not use a sample holder. We solder side buss on the contact cells to make the connection and then we perform the SB. Previously to these two steps, 3.2. Production process to obtain mini cells and mini modules: laser scribe processes, shunt busting and annealing 95   ϵϯ ϲ ϵ ϴ ϳϮ ϱ ϴ ϲ ϱϭ ϰ ϳ ϰ ϯ Ϯ ϭ   ϵ ϴ ϳ ϲ ϱ ϰ ϯ Ϯ ϭ ϭϮϯϰϱϲϳϴϵϭϮϯϰϱϲϳϴϵ ϭϲ ϭϭ ϭϰ ϭϳ ϯϲϵ ϮϲϬϬŵŵ ϮϮϬϬŵŵ Ϯϱϴ ϭϰϳ ϭϮ ϭϱ ϭϴ ϭϬ ϭϯ ϭϮ ϭϱ ϭϴ ϭϰ ϭϭ ϭϬ ϭϯ ϭϲ ϭϳ &^ LJ dž Figure 3.8: Position of the MMs along the full size panel. C1 C2 C3 C4 C5 C6 Point (0,0) Edge delete Death cell Active area Death cell Edge delete Active area Edge delete Edge del. Figure 3.9: 10 cm x 10 cm total area MM produced at TS through LSR scribe. Distances in mm. 96 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells we delete the edges by manual grinding. Most of the times, we laminate them following a process similar to the one performed in-line and we make the annealing placing the MMs in the ACL. 3.3 Conventional spectral response equipment With the SR equipment we measure the generated current in the solar cells for a determined wavelength and for the wavelength range where the solar cell is sensitive to light. This range depends on the kind of semiconductor used and on the solar cell structure. For SJ a-Si:H p-i-n structures, the interesting wavelength range goes from 300 nm to 800 nm. 3.3.1 Components for the conventional spectral response system The conventional spectral response (CSR) equipment is made up of the following devices: – Arc lamp: This lamp is a Newport 300 W Xenon Ozone free lamp supplied by a power controller to have stable illumination conditions (see left side of Fig. 3.10). This kind of lamp delivers stable radiation, from UV to near infrared (NIR) radiation with a variable intensity. However its output spectrum presents some peaks that could affect the accuracy of the final measurement. – Chopper wheel: It converts continuous light (provided by the arc lamp) into pulsed light with a constant frequency that we set. – Filter wheel: This filter wheel allows to use and to change automatically different filters when measuring over the wavelength spectrum. These filters are used to avoid illumination effects at second order wavelengths caused by the diffraction grating in the monochromator. For example, if the diffraction grating is in the position to transmit light at 800 nm wavelength, the monochromator will also transmit a small 400 nm component where our solar cell is also sensitive. This would induce a current higher than the true current generated at 800 nm. The filters eliminate second order light by just transmitting above the wavelength range of interest. As an example, if we want to scan from 400 nm to 900 nm, we will use a filter from 800 nm on, that will transmit only from wavelengths greater than 400 nm. 3.3. Conventional spectral response equipment 97 – Monochromator: It is shown in the right side of Fig. 3.10, it is the main component of the CSR equipment. It supplies the system with monochromatic light. The monochromator receives a large band spectrum light of the arc lamp which is focused on a grating which only transmits a certain wavelength to the monochromator output depending on the incidence angle of the grating. Changing the incidence angle and changing to different gratings, a wide range of wavelengths can be selected. All the mechanical manipulations in the monochromator used to select the wavelength, are operated by a computer program developed in the laboratory (to be explained in section 3.3.6). Figure 3.10: Arc lamp (left) and monochromator (right). – Beam-splitter: It is used to divide the main incoming beam into two beams of light with an angle of 90o. One is called fibre optic output (where the solar cell under test is placed) and the other one is the reference output (where the reference photodiode is placed). By using different semitransparent materials, e.g. partially covered glasses or mirrors, a beam-splitter has different transmissions properties and therefore different light intensities for the two output beams. It is important to optimize the light division to have adequate light intensities in both, the reference photodiode and the solar cell under test. – Reference photodiode and calibrated photodiode: Two photodiodes are used in the spectral response measurement principle. In the calibration procedure a calibrated photodiode, with known SR, is located at the fibre optic output (replacing the solar cell 98 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells Figure 3.11: Incoming beam divided by the beam-splitter. under test) and the signal is measured at the same time with the reference photodiode (whose SR is also known). From this measurement we obtain the calibration factor (F) for each wavelength of the optical assembly. The F is used later to calculate the SR of the solar cell under test comparing its intensity with the one of the reference photodiode which stays always at the reference output of the beam splitter. – Optical fibre: This optical fibre bundle transports the light from the beam-splitter to the solar cell under test. We use a fibre bundle to have more flexibility to illuminate easily different parts of the coupon. – Lock-in amplifiers: The small currents generated by the solar cell and the photodiode are measured with two lock-in amplifiers (see left side of Fig. 3.12). A lock-in amplifier uses a reference frequency provided by the chopper wheel to stimulate the test sample and phase locked loop technology. It is able to detect very small signals (at the same frequency as that of the reference frequency) in a noisy signal background. The two lock-in amplifiers are used to simultaneously measure the small currents generated by the momochromatic light of the reference photodiode and the solar cell under test. – Cold light: This light source, used as a bias light, illuminates the test cell with a white light during the measurement (see right side of Fig. 3.12). This illumination adjusts the cell to operating conditions similar to standard test conditions (1000 W/m2,25oC and AM1.5G spectral distribution) (STC). It is called cold light because the lamp filters the 3.3. Conventional spectral response equipment 99 infrared (IR) light so that it does not create heat during the illumination and, therefore, does not transmit heat to the illuminated solar cell. Figure 3.12: Two lock-in amplifiers (left) and cold light source with its illumination system (right). Other devices/equipments linked to the CSR equipment are: – Reference solar cell: The calibrated filtered solar cell (fabricated and calibrated by the FHG-ISE Callab Freiburg, Germany) is used to check if we are working under an irradiance of 1000 W/m2with the solar simulator (SS) and to know the error of our SR equipment. We know its SR and EQE in 10 nm steps for a spectrum from 280 nm to 930 nm, as well as its IV curve in the first quadrant under STC. The main electrical parameters are exposed in Table 3.1: Table 3.1: IV curve parameters of the calibrated solar cell (4 cm2) from FHG-ISE measured under STC [68]. Electrical parameters η(%) 7.6±0.2 Isc (mA) 61.0±1.5 Jsc (mA/cm2)15.23±0.38 Voc (mV) 624.2±3.1 FF (%) 80.1±0.8 Impp (mA) 57.0 Vmpp (mV) 532.9 Pmpp (mW) 31.0 100 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells – Spectrophotometre: The Perkin Elmer LAMBDA 950 spectrophotometre is an instrument operating from the UV to the NIR spectral range. It is made up of two monochromators, two radiation sources (a deuterium lamp and a halogen lamp) and a detector compartment. In the laboratory, this equipment allows measuring the diffuse and total transmission and reflection of thin films and thus, it can calculate the direct transmission and absorption. Related to the SR, we determine the internal quantum efficiency (IQE) by measuring the solar cell reflection (as explained in the next section). 3.3.2 Spectral response and quantum efficiency The spectral response (SR, expressed in A/W) represents the intensity of the photovoltaic effect depending on the incoming light for a given wavelength in a solar cell. It allows to determine how much current is generated by a solar cell for a specific irradiance and for each wavelength of the whole spectrum. The SR is also associated to the external quantum efficiency (EQE, dimensionless), which represents the number of electron/hole pairs generated in the cell by the incident photon flux at each wavelength [69]. Both parameters are linked by Eq. 3.1: EQE(λ)=SR(λ) q hc λ=1240SR(λ) λ(3.1) where qis the electric charge of one electron (q=1.602×10−19 C), his the Planck’s constant (h=6.626×10−34 J·s), cis the speed of light in vacuum (c=3×1017 nm/s) and λ(nm) is the wavelength of the incident light. The SR is usually used to extract information about recombination at front and rear contacts, diffusion or drift lengths of carriers, width of the depletion region and light trapping properties among the most important [70] [71]. In addition, it gives detailed information about which wavelengths have more carrier generation in the solar cell and therefore, it is essential to optimize the modules. There is another parameter related to the QE, the internal quantum efficiency (IQE) which considers the reflection loss at the solar cell’s surface. It represents the number of carriers generated in the solar cell per absorbed photon. Knowing the EQE and the reflection (r) of the device we calculate the IQE through the Eq. 3.2: IQE(λ)=EQE(λ) 1−r(λ)(3.2) 3.3. Conventional spectral response equipment 101 3.3.3 Differential spectral response method Traditionally, the differential spectral response (DSR) method is based on the detection by a lock-in amplifier of the solar cell’s response (solar cell’s current). The illumintation source is a modulated (AC) monochromatic test light. The cell is simultaneous exposed to a white bias (DC) light of irradiance in the range of 1 sun illumination. Another variation of this method is to use an incandescent lamp, filtered with a set of spectral filters, to obtain the monochromatic light source. In this version, a solar simulator is used as the bias light source [72]. The principle of the DSR method [52] is to compare the current of a reference photodiode and the current of the solar cell under test illuminated simultaneously by the same light source under STC [73]. Monitoring simultaneously these two currents for each wavelength, and knowing the calibration factor (F(λ)) for the reference cell, we can calculate the SR of the solar cell under test and, thus, the EQE and the short circuit current (Isc) (see below). 3.3.4 Determination of the calibration factor, the spectral response and the integrated short circuit current density Before measuring the solar cell in the SR equipment, we have to know a calibration factor (F(λ)) for the calculations of the solar cell SR. In the case of the CSR equipment, the F is determined by the transmission and reflection characteristics of the beam splitter and by the losses in the fibre optic. This factor takes into consideration the difference of light intensity in the two arms of the beam splitter. To determine it, we use a second photodiode (calibrated photodiode) with a known SR. In the calibration measurement the calibrated photodiode is in the position of the solar cell under test. Consequently, we will measure simultaneously the signals of the reference and calibrated photodiodes to obtain the ratio between them for each wavelength. One has to put special attention in the quantity of incident light reaching the calibrated photodiode, so it is very important to concentrate all the light in the active area of the photodiode (the same holds for the solar cells). Therefore, the beam is focused by placing a lens in the fibre optic output. Taking into account the definition of the EQE, we get the following equation: EQE(λ)= Iph(λ) q·φtot(λ)(3.3) where Iph (A) is the generated current of a solar cell and φtot (photons/s) is the incident photon flux. 102 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells Applying Eq. 3.3 for the calibrated photodiode (cal), the reference photodiode (ref) and the solar cell under test (cell), we get: EQEcal(λ)= Ical ph (λ) q·φtot(λ)·t(λ)(at f ibre optic out put)(3.4a) EQEre f (λ)= Ire f ph (λ) q·φtot(λ)·r(λ)(at direct out put)(3.4b) EQEcell(λ)= Icell ph (λ) q·φtot(λ)·t(λ)(at f ibre optic out put)(3.4c) where φtot (photons/s) is the incident photon flux on the beam-splitter, t is the transmittance of the beam splitter and r is the reflection of the beam-splitter. Combining the equations 3.4 and changing EQE for SR following the Eq. 3.1 we obtain the SR of the solar cell under test: SRcell(λ)=Icell ph (λ) Ire f ph (λ)·F(λ)·SRcal(λ)(3.5) where the F is: F(λ)=Ire f ph (λ) Ical ph (λ)(3.6) The SR equipment is configured to measure the samples in short circuit conditions. So, to calculate the Isc of the solar cell from the spectral response measurement one only has to integrate over the spectrum of interest the product of the measured SR(λ), the irradiance (G(λ)) of the AM1.5 spectrum and the surface of the solar cell under test as shown in the following equation: Isc =λf λ0 Girrad(λ)·SR(λ)·S·dλ(3.7) where Girrad (W/cm2/nm) is the irradiance provided by the solar spectrum AM1.5 by unit of surface and wavelength and S (cm2) is the surface of the solar cell. 3.3.5 Experimental set-up In Fig. 3.13 the scheme of the experimental set-up of the CSR measurement equipment is presented. The operation is based in the conversion of continuous light provided by 3.3. Conventional spectral response equipment 103 the arc lamp to pulsed light employing the chopper wheel. Then, the beam arrives to the monochromator which converts the pulsed light in monochromatic light depending on its wavelength. This beam of monochromatic light is separated into two by the beam splitter. One beam goes directly to the reference photodiode and the other one goes through a flexible optical fibre whose output illuminates just above the solar cell. A sample holder is used to establish the electrical contact with the solar cell (see Fig. 3.3, right). Lastly, the two lock-in amplifiers detect and measure the small current generated per wavelength. White lamp Beam splitter Chopper Computer Lock-in amplifier 2 Solar cell under test or calibrated photodiode Lock-in amplifier 1 Reference photodiode Monochromator Filter wheel Light path Figure 3.13: SR equipment and links between the different devices. The optical tuning during the installation of the equipment is a challenge. The optical components of the equipment must be installed step by step by optimizing the beam at each transition point, beginning at the arc lamp and finishing in the reference photodiode and the solar cell. One of the most important steps is to properly focus the collimated beam provided by the arc lamp on the input slit of the monochromator. This point is vital to use the maximum area of the diffraction grating and to have the maximum amount of light at the output slit of 110 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells Table 3.2: Comparison of the technical data sheet values with the experimental data obtained at our lab for the 24 selected LEDs. LED code Structure V(V) I(mA) V(V) I(mA) V(V) I(mA) L1 355 ͲͲ3.60 25 3.70 34.0 0.3 56 Ͳ L2 370 376 11 3.90 10 3.50 30.5 0.3 23 AlGaN L3 385 383 11 3.50 20 3.55 47.0 0.3 35 InGaN L4 400 395 12 3.70 20 3.50 18.5 0.4 26 InGaN L5 415 412 14 3.70 30 3.35 24.0 0.3 21 InGaN L6 470 470 31 3.20 30 3.40 18.0 0.4 20 InGaN L7 510 501 23 3.40 20 3.25 18.0 0.5 25 InGaN L8 525 524 38 3.20 20 3.00 37.0 0.3 33 InGaN L9 565 562 23 2.00 20 3.50 18.0 0.5 34 GaP L10 596 593 15 2.10 20 2.00 14.0 0.3 25 AlGaInP L11 625 641 16 2.20 20 2.45 35.0 0.5 41 AlGaInP L12 660 653 21 1.90 20 2.05 32.0 0.5 40 GaAlAs L13 670 672 25 1.80 20 1.80 22.0 0.2 31 GaAlAs/GaAlAs L14 680 683 21 1.90 20 2.20 47.0 0.3 33 GaAlAs L15 720 721 24 2.00 50 1.80 34.0 0.2 37 AlGaAs/AlGaAs L16 760 755 25 1.20 20 1.65 26.0 0.3 57 AlGaAs/AlGaAs L17 780 772 27 1.70 50 2.00 43.0 0.3 33 AlGaAs/AlGaAs L18 800 795 30 1.80 50 1.65 66.0 0.3 67 AlGaAs L19 820 819 30 1.60 100 1.55 42.5 0.1 43 AlGaAs L20 840 827 31 1.60 100 1.60 47.5 0.4 74 AlGaAs L21 870 859 49 1.50 100 1.45 58.5 0.1 49 AlGaAs L22 910 914 57 1.40 100 1.40 47.0 0.2 44 AlGaAs L23 950 941 44 1.40 100 1.35 80.0 0.1 100 AlGaAs/GaAs L24 985 982 43 1.25 100 1.25 54.5 0.1 63 GaAs Datasheet peak wavelength (nm) Experimental peak wavelength (nm) FWHM(exp.) (nm) DatasheetDCbias ExperimentalDCbias ExperimentalACamplitude other) is crucial [74]. To identify this region we measured the IV curve of each LED by introducing voltage to it and measuring (in series) the current that crossed. In order to automate the analysis, a program was developed in the visual platform LabVIEW. The source-multimetre used for this purpose was a Keithley 2420. Fig. 3.19 shows the IV curves of four representative LEDs. The range of operation is indicated as well as their experimental direct current (DC) operating point (chosen to be in the middle of the experimental quasi linear range). Each LED is supplied with a sinusoidal voltage. It oscillates in its linear range, therefore the operating point is the DC bias and the amplitude of the sine wave is defined by the extension of the linear range. –Peak wavelength and spectral bandwidth: To minimize the measurement error in our equipment we need to determine very well the peak wavelength (λpeak) of emission and to select LEDs with narrow band width. Fig. 3.20 shows the spectra of two LEDs measured under 25 oC presenting a large difference in the band width. The figure illustrates as well that we found some differences between the measured peak wavelengths and the ones provided by the manufacturer. We have discarded LEDs with 3.4. Very fast spectral response equipment 111 Ϭ ϮϬ ϰϬ ϲϬ ϴϬ ϭϬ ϭϱ ϮϬ Ϯϱ ƵƌƌĞŶƚ/;ŵͿ sŽůƚĂŐĞs;sͿ ϲϳϬŶŵ ϱϵϲŶŵ ϳϲϬŶŵ ϱϲϱŶŵ džƉKƉĞƌĂƚŝŶŐWŽŝŶƚ Figure 3.19: Experimental IV curve and DC bias point for 4 representative LEDs. Black lines indicate the AC amplitude. full width at half maximum (FWHM) greater than 40 nm, in addition it is important to know the exact peak wavelength, especially for LEDs illuminating in the wavelength range at the flanges of the SR curve, to prevent large errors [74]. The peak wavelengths were determined using an Ocean’s Optics USB2000+ spectrometre and supplying the LEDs with DC voltage using an Agilent N5771A. The current traversing the LED was measured connecting in series a Fluke 287 multimetre. Since the sensitivity of the mentioned spectrometre is limited from 360 nm to 1010 nm, we experienced certain restrictions while measuring the λpeak of some UV and IR LEDs. In addition, the λpeak and the FWHM were found to be independent from the current at which the LED works under our operation conditions. However, the emission of photons is proportional to the LEDs supply current [27]. The variation among the spectra of LEDs of the same model was also measured, observing a small deviation between the peak wavelengths (less than 2 nm). Since these results were not significant we continued evaluating one LED per model. Fig. 3.21 presents the individual spectra of the twenty-three selected LEDs for the VFSR measurement system. L1 at 355 nm cannot be measured with our spectrometre, in this case we use the data of the LED manufacturer. The spectral irradiance values 112 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells were obtained with LEDs operating near to the recommended working point of the provider. ϬϬϬ ϬϬϮ ϬϬϰ ϬϬϲ Ϭ Ϯ ϰ ϲ ϱϱϬ ϲϬϬ ϲϱϬ ϳϬϬ ϳϱϬ ϴϬϬ ϴϱϬ ϵϬϬ >ƐƐƉĞĐƚƌĂů/ƌƌĂĚŝĂŶĐĞ;tŵϮŶŵͿ >ƐƐƉĞĐƚƌĂůŝƌƌĂĚŝĂŶĐĞ;tŵϮŶŵͿ tĂǀĞůĞŶŐƚŚʄ ;ŶŵͿ >ϱϵϲŶŵΛϮϱŵ >ϳϬϬŶŵΛϮϱŵ ʄƉĞĂŬсϱϵϯŶŵ ʄƉĞĂŬсϳϬϲŶŵ &t,DсϲϲŶŵ &t,DсϭϱŶŵ Figure 3.20: Peak wavelength (λpeak) and FWHM for a LED with narrow band width (22 nm, blue dashed curve) and for a LED with wide band width (66 nm, red dotted curve). ϬϬϬ ϬϬϰ ϬϬϴ ϬϭϮ Ϭϭϲ ϬϮϬ ϬϮϰ ϬϮϴ Ϭ Ϯ ϰ ϲ ϴ ϭϬ ϭϮ ϭϰ ϯϱϬ ϰϬϬ ϰϱϬ ϱϬϬ ϱϱϬ ϲϬϬ ϲϱϬ ϳϬϬ ϳϱϬ ϴϬϬ ϴϱϬ ϵϬϬ ϵϱϬ ϭϬϬϬ >ϵƐƉĞĐƚƌĂůŝƌƌĂĚŝĂŶĐĞ;tŵϮŶŵͿ ^ƉĞĐƚƌĂůŝƌƌĂĚŝĂŶĐĞ;tŵϮŶŵͿ tĂǀĞůĞŐŚƚλ λλ λ;ŶŵͿ >ϯ >ϳ >ϰ >Ϯ >ϱ >ϲ >ϴ >ϭϭ >ϵ >ϭϯ >ϭϬ >ϭϰ >ϭϱ >ϭϲ>ϭϳ >ϭϴ >ϭϵ >ϮϬ >Ϯϭ >ϮϮ >Ϯϯ >Ϯϰ >ϭϮ Figure 3.21: Experimental spectral irradiance for 23 selected LEDs in the range 370 nm - 1000 nm. 3.4. Very fast spectral response equipment 113 3.4.2 Sinusoidal generators In order to reduce significantly the measurement time, in the VFSR system all LEDs work simultaneously and, to identify them, their light is modulated at different frequencies. Later, the current generated by each LED in the solar cell is determined by a FFT of the total generated current. Nowadays, we use eighteen/twenty-four sine-wave generators for SJ/TJ solar cell measurements. The sine-wave generators exit simultaneously 24 individual voltage waves of ±4.5 V of amplitude. They produce an individually optimized wave for each LED (frequency, DC bias and AC amplitude) according to their corresponding IV curve. This illumination generates sinusoidal current in the solar cell. Concerning the excitation frequencies of these signals, they were chosen to sweep between 100 Hz and 200 Hz with a step width of 4 Hz (avoiding multiples of frequencies for the LEDs and for the 50 Hz frequency of the electricity grid). The selection of the frequency applied to each LED is important to reach a proper equipment operation. The overlap of the harmonics with the main current peaks must be avoided. 3.4.3 Fast Fourier transform concept The fundamental tool used in the VFSR equipment is the FFT. This is widely utilized in digital signal processing (DSP). When signals are expressed in time-domain some information such as frequency and amplitude are coded. This data becomes evident in frequency-domain. Analytically, Fourier analysis provides the connection between time-domain and frequency-domain. Using the Fourier Transform a time dependent periodic function f(t)can be expressed as the sum of different sinusoidal functions with different frequencies wn=(n+1)·w0(see Fig. 3.22). For n=0 we say that it is the fundamental frequency. While, for n=1,2,3,4··· we say that it is the nth harmonic. Figure 3.22: Decomposition of a periodic signal in its different harmonics. 114 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells Fast Fourier transformation is a method that allows saving calculations to find the Fourier transform of a sampled signal. We can say that for a sinusoidal wave in the time-domain the FFT is a delta function placed in the frequency-domain of this signal (see Fig. 3.23). It has no harmonics, just the fundamental component. Figure 3.23: Sine wave in the time-domain (left) and, after FFT analysis, sine wave in the frequency-domain (right). The virtual platform LabVIEW has implemented a tool for FFT calculation, using different algorithms. The FFT tool allows us to determine the spectrum for the measured periodical current and display the results in terms of peak amplitude. This means that, a sine tone of amplitude Ayields a spectral value (magnitude) of Aat the sine tone frequency. 3.4.4 Measurement procedure for the very fast spectral response equipment The VFSR equipment allows the measurement of MCs, MMs or full size modules. In the case of MCs, we establish the electrical contact with the solar cells by means of the sample holder and its switch, both presented in section 3.2. The terminals of this sample holder are connected to a digital multimetre for sampling. As for MMs, we fabricated special test modules where every cell is connected individually with a side buss. This enables to perform the measurements connecting each cell directly or connecting the extremes of the mini module, as is the usual way to electrically connect a module (see Fig. 3.24). In this last case, the side buss is not soldered as usual to prevent damaging the solar cells. The used buss has a conducting adhesive that can be pasted onto the solar cells’ back contact without losing conductivity. The illumination is done placing the optical coupling element directly above the respective solar cell illuminating a small area (0.5 mm of diameter). Simultaneously, the reference photodiode is lit. We use a fast and sensitive current metre, an Agilent 34410A, to directly measure the generated current in both devices (Icell(t)and Ire f (t)). We use shielded 3.4. Very fast spectral response equipment 115 cables in the connections to minimize the noise. The next steps consist in processing this data and are explained in the following section. Figure 3.24: MM with every cell connected individually with a conducting adhesive side buss. To measure the generated current properly, one has to take in account that to analyse an analogue (time continuous) signal it is necessary to digitalise and sample it. When this is done, a phenomenon known as aliasing can appear. This effect causes different continuous signals to become indistinguishable once sampled. To avoid it, the Nyquist theorem establishes that the sampling rate (number of samples (S) taken by unit of time) must be at least the double of the frequency of the sampled signal [80]. The frequencies at which the LEDs are driven are between 100 Hz and 200 Hz. This means that according to that theorem, the sampling rate must be at least 400 S/s. In our case, the sample frequency at which the Agilent 34410A works is 1000 S/s, taking 5000 measurements (samples). The schematic diagram of the operation of the whole system is shown in Fig. 3.25. 3.4.5 Software developed for the very fast spectral response equipment As mentioned, we have implemented the VFSR program using the platform LabVIEW. For this purpose, we use standard commands for programmable instruments (SCPI) instructions in order to control the digital multimetre. We measure voltage in the front and 116 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells Figure 3.25: Diagram of the VFSR measurement system. rear terminals, taking for each one 5000 samples (S) at a rate of 1000 S/s. The transduction of voltage to current is done digitally knowing the value of a resistance placed parallel to the solar cell or photodiodes. If we are doing a calibration of the equipment, a calibrated photodiode is located in the place of the solar cell under test. The reference photodiode and calibrated photodiode currents (Ical(t)and Ire f (t)) are measured as explained in the previous section. Then, the FFT analysis will be performed over both current signals (obtaining frequency-domain signals Ical(ν)and Ire f (ν)). Subsequently, as the excitation frequencies of the LEDs are known, we can assign a current value to each wavelength (Ical (λ)and Ire f (λ)). Next, the F(λ)factor will be calculated using Eq. 3.6. Finally, the F(λ)factor will be saved in a .txt file, together with the Ire f (λ)values during this calibration (Ire fcal (λ)), the data sheet EQE(λ) of both photodiodes and the spectrum AM1.5(λ) for the given wavelengths. This file is used as input file when measuring a solar cell or module. In the case of the SR measurement of a solar cell, the reference photodiode and solar cell currents (Icell(t)and Ire f (t)) are measured in the time-domain as explained in the previous section, resulting in Fig. 3.26. Then, with the FFT analysis we obtain the frequency-domain current signals (Icell(ν)and Ire f (ν)) as presented in Fig. 3.27. In the figure it is shown that the signal to noise ratio is about 2 – 4 orders of magnitude which is sufficient to determine the SR of an individual solar cell. Apart from the peaks corresponding to the LEDs, peaks resulting from the noise produced by the electricity grid are indicated. Once the current is determined in the frequency-domain (resolution 0.2 Hz) we analyse the peak by searching the maximum in ±1.6 Hz of the reference frequency of each LED and we integrate the generated current in ±0.4 Hz from the maximum to include small deviations 3.4. Very fast spectral response equipment 117 ϴϬ ϵϬ ϭϬϬ ϭϭϬ ϭϮϬ ϭϯϬ ϬϬϬϭϬϮϬϯϬϰϬϱ ƵƌƌĞŶƚĚĞŶƐŝƚLJ:;μ μ μ μĐŵϮͿ dŝŵĞƚ;ƐͿ WĞƌŝŽĚсϮϱϬŵƐ Figure 3.26: Time dependent current density curve presents repetition period of 250 ms due to the selected frequencies with highest common divisor of 4. ϭͲϬϱ ϭͲϬϰ ϭͲϬϯ ϭͲϬϮ ϭͲϬϭ ϭнϬϬ ϭнϬϭ Ϭ ϱϬ ϭϬϬ ϭϱϬ ϮϬϬ ϮϱϬ ƵƌƌĞŶƚĚĞŶƐŝƚLJ:;μ μ μ μĐŵϮͿ &ƌĞƋƵĞŶĐLJν νν ν;,njͿ ůĞĐƚƌŝĐŐƌŝĚ &ƵŶĚΘϭƐƚ ,ĂƌŵŽŶŝĐ &ƵŶĚĨƌĞƋŽĨ ϭϬ>Ɛ &ƵŶĚĨƌĞƋŽĨ ϲ>Ɛ ůĞĐƚƌŝĐŐƌŝĚ ϮŶĚΘϯƌĚ ,ĂŵŽŶŝĐƐ EŽŝƐĞ ůĞǀĞů Figure 3.27: Current density curve in the frequency-domain as resulting from FFT analysis. in frequency of the generated signal [81]. Subsequently, as the excitation frequencies of the LEDs are known, we can assign a current value to each wavelength (Icell(λ)and Ire f (λ)). Next, the program accesses to the input file which contains the data of the reference photodiode current determined during the last calibration (Ire fcal (λ)). A comparison is done between Ire fcal (λ)and Ire f (λ), the difference must not exceed ±5%. Otherwise, the 118 Chapter 3. Industrial production process and basic equipment set-ups of a-Si:H solar cells program will be stopped, since the device needs calibration. If the actual current of the reference photodiode (Ire f (λ)) is within the tolerance range, the calculation of the EQE will be performed using the EQE of the calibrated photodiode and the calibration factor of the last calibration (through Eqs. 3.1 and 3.5). Additionally, if the EQE data measured with the CSR equipment is available, a comparison of the results can be done. Finally, the Isc is calculated following Eq. 3.7 and the results are saved in a .txt file. In Fig. 3.28 the comparison of the EQE measured with the CSR and VFSR equipment is presented. The red measurement points represent the average data of ten repetitive measurements at each wavelength, for the point at 593 nm (the one with higher standard deviation) an error bar (standard deviation) is indicated. ϬϬ ϬϮ Ϭϰ Ϭϲ Ϭϴ ϭϬ ϯϬϬ ϰϬϬ ϱϬϬ ϲϬϬ ϳϬϬ ϴϬϬ džƚĞƌŶĂůƋƵĂŶƚƵŵĞĨĨŝĐŝĞŶĐLJY tĂǀĞůĞŶŐƚŚλ λλ λ;ŶŵͿ ^ZƵƐŝŶŐŵŽŶŽĐŚƌŽŵĂƚŽƌ ^ZƵƐŝŶŐs&^Z ϯϳϱŶŵͲϴϬϬŶŵ :ƐĐ ŵŽŶŽĐсϭϮϯϳŵĐŵϮ :ƐĐ s&^ZсϭϮϯϯŵĐŵϮ ƌƌŽƌďĂƌƐ Figure 3.28: The blue curve presents the EQE measured in a CSR equipment with monochromator (5 nm wavelength step). Red dots (the point at 593 nm has an error bar indicating its standard deviation) represent the average of ten EQE measurements with the VFSR measurement system. In this case, the difference between the Jsc determined with the CSR equipment and the VFSR is 0.26%. We have found that the standard deviation in repetitive measurements is less than 2% for every wavelength. For most of the points the deviation from the traditional method is small, a larger deviation is found for the LEDs at 501 nm and 562 nm wavelengths. The reason for the larger deviation at 562 nm most probably is due to the small intensity of the LED illumination at that wavelength which is under investigation, also the larger deviation at 501 nm is under investigation but no reason has been identified so far. As shown in the insert of Fig. 3.28, the short circuit current calculated in the wavelength range from 375 nm to 800 nm is very similar 3.5. Measurements of illuminated and dark current-voltage curves 119 for both methods with a deviation of less than 1%, and comparable to values found by other researchers [75], [76]. 3.5 Measurements of illuminated and dark current-voltage curves With the measurement of illuminated and dark IV curves we are going to characterise electrically the solar cells and mini modules. Next, we present the main electrical parameters to be considered and the equipments and procedures needed to obtain them. 3.5.1 Equipment – Solar simulator: It is installed in the optical laboratory. It simulates the solar light in a 20 cm x 20 cm area specifically, it provides a wide beam of collimated and uniform light with an irradiance almost like the sun. The solar simulator (SS) equipment consists mainly in a 1300 W external electric power source, an arc lamp compounded by high pressure noble gases, an ellipsoid mirror around the lamp for collimation, a flat mirror to direct the light to the sample through a lens and an AM1.5 spectrum filter. It provides class AAA (the highest precision) illumination for a 20 cm x 20 cm area. Since we have a variable power supply by the external source, we could regulate the irradiance in our device changing the power, however, that could change the spectrum. Therefore, the illumination level is usually fine adjusted by the lamp position in respect to the ellipsoid, even if this could affect slightly the overall uniformity. To measure the IV curves under STC we set the SS output irradiance to 1000 W/m2following the steps described in section 3.5.4. In addition, the irradiance level is controlled every time we use the SS by measuring the Isc with a reference solar cell. With this equipment we measure the illuminated IV curves for small samples as our 1 cm2and4cm 2solar cells and our 10 x 10 cm2and20x20cm 2MMs. – Keithley 2400 source metre/multimetre: We use this source to inject current or to apply voltage to the samples in which we measure IV curves under illumination or dark conditions. The maximum power, voltage and current parameters for this source are: Pmax =22W,V max = 200 V, Imax = 1 A. Consequently, if we work with the maximum