New Concepts for Front Side Metallization of Industrial Silicon Solar Cells Dissertation zur Erlangung des Doktorgrades der Fakultät für Angewandte Wissenschaften der Albert-Ludwigs-Universität Freiburg im Breisgau vorgelegt von Ansgar Mette Fraunhofer-Institut für Solare Energiesysteme Freiburg im Breisgau 2007 Dekan: Hauptreferent: Koreferent: Datum der Prüfung: Prof. Dr. Bernhard Nebel PD Dr. Volker Wittwer Prof. Dr. Jürgen Wilde 26. November 2007 To my parents, with love and gratitude. Table of contents INTRODUCTION..........................................................................................1 Thesis motivation ...................................................................................................... 1 Thesis outline ............................................................................................................ 3 1 CRYSTALLINE SILICON SOLAR CELLS ............................................5 1.1 Structure and operation principle.................................................................... 5 1.1.1 Current-Voltage curve.............................................................................. 6 1.1.2 Physical and technological efficiency limitations ................................... 9 1.2 Loss mechanisms due to front side metallization pattern............................. 13 1.2.1 Optical losses ......................................................................................... 14 1.2.2 Electrical losses...................................................................................... 15 1.2.3 Grid optimization for maximum power point........................................ 19 1.3 2 Metal-semiconductor contact........................................................................ 20 CELL PROCESSING AND METALLIZATION TECHNOLOGIES ......26 2.1 Processing of silicon solar cells .................................................................... 26 2.1.1 Screen-printed solar cell......................................................................... 26 2.1.2 PERL and LFC solar cell ....................................................................... 28 2.1.3 Comparison of an evaporated with a screen-printed front..................... 29 2.2 Metallization technologies ............................................................................ 30 2.2.1 Thick- and thin-film metallization technologies.................................... 30 2.2.2 Screen-printing....................................................................................... 32 2.2.3 Stencil-printing....................................................................................... 36 2.2.4 Pad-printing............................................................................................ 37 2.2.5 Ink-jet printing ....................................................................................... 39 2.2.6 Dispensing.............................................................................................. 41 2.2.7 Photolithographical definition and evaporation..................................... 42 2.2.8 Laser micro-sintering ............................................................................. 43 ii Table of contents _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2.2.9 Nickel plating ......................................................................................... 43 2.2.10 Thickening of metal contacts by means of plating ................................ 45 2.3 3 Two-layer concept......................................................................................... 45 GRID DESIGN OF THE TWO-LAYER CONTACT STRUCTURE.......47 3.1 Calculation parameters and assumptions ...................................................... 47 3.2 Solar cell with two busbars ........................................................................... 50 3.2.1 Assuming a cell with a sheet resistance of 55 Ω/sq............................... 50 3.2.2 3.2.3 3.2.4 Variation of the emitter sheet resistance ................................................ 51 Influence of the contact resistivity ......................................................... 52 Considering irradiation reflection from the contact............................... 54 3.3 Solar cell with alternative busbar geometries ............................................... 55 3.3.1 Three busbars ......................................................................................... 56 3.3.2 Mesh of wires ......................................................................................... 57 3.4 Losses caused by busbar and tab resistance.................................................. 59 3.4.1 Busbar losses .......................................................................................... 59 3.4.2 Tab losses ............................................................................................... 59 3.5 4 Chapter summary .......................................................................................... 60 SERIES RESISTANCE DETERMINATION FROM IV-CURVES.........62 4.1 Series resistance determination methods ...................................................... 62 4.1.1 Fitting of the two-diode equation to a dark IV-curve............................. 62 4.1.2 Comparison of the dark with the one-sun IV-curve ............................... 63 4.1.3 Comparison of the Suns-Voc with the one-sun IV-curve ........................ 64 4.1.4 Comparison of at least two IV-curves measured at different irradiation intensities ............................................................................................... 64 4.1.5 Comparison of jsc and Voc values with the one-sun IV-curve................. 66 4.1.6 Integration of the area under an IV-curve .............................................. 66 4.2 Correlation between fill factor and series resistance .................................... 67 Table of contents iii _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 4.3 Experimental comparison of determination methods ................................... 68 4.4 Influence of non-ideal IV-curves on rs determination .................................. 71 4.5 Detailed discussion of presented methods .................................................... 72 5 5.1 EVAPORATION OF DIFFERENT CONTACT METALS .....................74 Solar cell processing ..................................................................................... 74 5.2 IV-Parameters of processed solar cells ......................................................... 75 5.2.1 Different metals evaporated on inverted pyramid front......................... 75 5.2.2 Different metals evaporated on a random pyramid front....................... 76 5.3 Discussion of the different metals ................................................................ 77 5.3.1 Aluminum .............................................................................................. 78 5.3.2 Palladium................................................................................................ 79 5.3.3 Silver ...................................................................................................... 80 5.3.4 Titanium-Palladium ............................................................................... 80 5.3.5 Titanium ................................................................................................. 82 5.3.6 Nickel ..................................................................................................... 82 5.3.7 Chromium .............................................................................................. 83 5.4 6 Chapter summary .......................................................................................... 84 SCREEN-PRINTING OF HOTMELT SILVER PASTE.........................85 6.1 Comparison of hotmelt and conventional paste............................................ 85 6.1.1 Composition ........................................................................................... 85 6.1.2 Rheology ................................................................................................ 87 6.1.3 Thermogravimetrical analysis................................................................ 89 6.2 Characterization of hotmelt printed contacts ................................................ 90 6.2.1 Hardware setup....................................................................................... 90 6.2.2 Optical characterization ......................................................................... 91 6.2.3 Electrical characterization...................................................................... 94 6.3 Solar cell processing and results ................................................................... 95 iv Table of contents _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6.3.1 6.3.2 6.3.3 Monocrystalline silicon solar cells......................................................... 95 Multicrystalline silicon solar cells ......................................................... 96 Variation of the emitter sheet resistance ................................................ 98 6.4 Fine-line printing and high sheet resistance emitter ..................................... 99 6.4.1 Motivation .............................................................................................. 99 6.4.2 Contacting high sheet resistance emitter.............................................. 100 6.4.3 Fine-line screen-printing ...................................................................... 102 6.5 7 7.1 Chapter summary ........................................................................................ 103 LIGHT-INDUCED PLATING..............................................................105 Introduction ................................................................................................. 105 7.2 Principle of light-induced plating ............................................................... 107 7.2.1 Characterization of the silver cyanide electroplating bath................... 107 7.2.2 Plating of solar cell contacts ................................................................ 109 7.2.3 Plating without external power supply................................................. 115 7.2.4 Plating of n-type silicon solar cells ...................................................... 117 7.2.5 Plating of non-silver contacts............................................................... 118 7.3 Light-induced plating of screen-printed contacts ....................................... 118 7.3.1 Setup..................................................................................................... 118 7.3.2 Small-area monocrystalline silicon solar cells..................................... 119 7.3.3 Large-area multicrystalline silicon solar cells ..................................... 121 7.3.4 Cyanide-free plating for industrial application .................................... 124 7.4 Direct light-induced plating of silver onto silicon surface ......................... 125 7.5 Chapter summary ........................................................................................ 127 8 HIGH-EFFICIENCY SCREEN-PRINTED & PLATED SOLAR CELL 129 8.1 Production sequence for LFC and Al-BSF solar cells................................ 129 8.2 Aluminum back surface field cell structure ................................................ 130 Table of contents v _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 8.3 Laser-fired contact cell structure ................................................................ 133 8.3.1 Variation of the firing temperature ...................................................... 133 8.3.2 Variation of the emitter doping concentration ..................................... 135 8.4 Laser-fired contact versus aluminum back surface field ............................ 137 8.5 Chapter summary ........................................................................................ 140 9 PAD-PRINTING: TECHNOLOGY AND ANALYSIS..........................142 9.1 Printing of hotmelt paste............................................................................. 142 9.2 Pad-printed and plated contacts .................................................................. 145 9.2.1 Solar cell processing ............................................................................ 145 9.2.2 Variation of the grid design and emitter doping concentration ........... 146 9.2.3 Influence of a sintering and second plating step.................................. 147 9.3 Chapter summary ........................................................................................ 150 10 METAL AEROSOL JET PRINTING ...............................................151 10.1 Working principle of the metal aerosol jet printer .................................. 151 10.1.1 Deposition head.................................................................................... 152 10.1.2 Pneumatic and ultrasonic atomizer ...................................................... 153 10.2 Nano-particle inks.................................................................................... 155 10.2.1 Metallic nano-particles......................................................................... 156 10.2.2 Ink components .................................................................................... 158 10.2.3 Stabilization of nano-particle inks ....................................................... 158 10.3 Set-up of the printer at Fraunhofer ISE ................................................... 159 10.4 Characterization and optimization of printing process ........................... 161 10.5 Solar cell processing and results ............................................................. 165 10.5.1 First aerosol jet printed solar cells using nano particle inks ................ 165 10.5.2 Multicrystalline silicon solar cells ....................................................... 165 10.5.3 Monocrystalline silicon solar cells....................................................... 167 vi Table of contents _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 10.6 11 11.1 Chapter summary..................................................................................... 171 MICROSTRUCTURE ANALYSIS OF CONTACTS........................172 Review of current models for screen-printed contact formation............. 172 11.2 Solar cell results of pad-printed and plated contacts............................... 175 11.2.1 Design of experiment ........................................................................... 175 11.2.2 IV-parameters ....................................................................................... 176 11.3 Dependence of contact resistance on contact width and height .............. 178 11.3.1 Contact resistance measurement .......................................................... 178 11.3.2 Optical analysis of the interface layer.................................................. 180 11.4 Influence of a plating step on the contact resistance ............................... 184 11.4.1 Electrical measurements....................................................................... 184 11.4.2 Micro-structural investigations ............................................................ 186 11.5 12 Conclusion: New current paths for plated contacts................................. 192 SUMMARY .....................................................................................196 DEUTSCHE ZUSAMMENFASSUNG ......................................................199 APPENDIX...............................................................................................202 LIST OF SYMBOLS, ACRONYMS, INDICES AND CONSTANTS .........211 LIST OF PUBLICATIONS .......................................................................216 BIBLIOGRAPHY .....................................................................................218 DANKSAGUNG.......................................................................................230 Introduction Thesis motivation As the title: „ New concepts for front side metallization of industrial silicon solar cells” indicates, this thesis addresses the field of renewable energy resources which is one of today’s major challenges of humankind to maintain a world, worth living in for next generations. Fossil energy is the today’s most used form of energy. Apart from the limited resources of fossil energy, the climatic aspect forces us to look for alternative energy conversion systems. Already in the first IPCC report [1] the correlation between the increase of the global temperature and the increase of anthropogenic emitted greenhouse gases was proven on a scientific base (see Fig. 0.1). With a share of 70% of the so called greenhouse effect, carbon dioxide takes up an exceptional position under the greenhouse gases. According to the fourth IPCC-Report, “Carbon dioxide is the most important anthropogenic greenhouse gas produced by combustion of fossil raw materials. The atmospheric concentration of carbon dioxide in 2005 exceeded by far the natural range over the last 650,000 years as determined from ice cores (Fig. 0.1 left-hand). The primary source of the increased atmospheric concentration of carbon dioxide since the pre-industrial period results from fossil fuel use, with land use change providing another significant but smaller contribution” [2]. Fig. 0.1: Left-hand: Atmospheric concentration of carbon dioxide in the past. Measurements are shown from ice cores (different colors for different models) and atmospheric samples (red lines). Right-hand: Comparison of observed changes in surface temperature (black line) with results simulated by climate models using natural (blue band) and anthropogenic forcing. The red band shows the sum of natural and anthropogenic forcing. [2] 2 Introduction _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Therefore a carbon dioxide neutral and sustainable energy production is of major importance, which can solely be achieved by renewable energies. Especially the direct conversion of incident sun irradiation into electrical power is a clear technological advantage of photovoltaic (PV) compared to other energy conversion systems. PV systems have already nowadays a life cycle green house gas emission in the range of 25 – 35 g/kWh at Southern European locations, which is considerably lower than all fossil options [3] (e.g. hard coal ≈ 980 g/kWh, gas ≈ 430 g/kWh). A critical aspect for the dissemination of PV is the actual price for a kWh of photovoltaic of 25 to 50 cent. [4]. To render solar cell technology economically competitive, the manufacturing costs of PV must be reduced to reach comparable values as for other energy sources, like gas, biogas, wind, water, nuclear power as well as brown and hard coal. Nevertheless in the last years the photovoltaic market has grown by 40% per year. Considering the learning curve and by estimating an installed capacity of 2000 GW in the year 2050, costs of 5 to 9 cent/kWh will be reached until 2030 at southern European locations [4]. Solar cell module costs can be divided into 40% for the module manufacturing process, 40% for the wafer and 20% for the solar cell technology [5]. The highest impact on cost can be achieved by increasing the efficiency of the solar cell and reducing the wafer thickness at the same time to achieve high material utilization. Focusing onto the cell technology, a lot of successful effort has been recently reported to improve the rear side of silicon solar cells surface passivation process schemes [6-8]. The front side metallization gives room for an even higher efficiency benefit according to an analysis of the loss mechanism associated to the well established screen-printing process [9]. At the moment, due to the multifunctionality of screen-printing metallization process, this technology has reached and maintained a market share of more than 85% [10]. For optimization two routes can be followed, either by an improvement of the current technology or by an introduction of new technologies. This work follows both approaches. The screen-printing process is optimized and new technologies and processes for the front-side metallization are proposed. The main goal is to close the efficiency gap between solar cells processed using a laboratory high-efficiency and an industrial screen-printing front side structure. Introduction 3 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Thesis outline In Chapter 1 the basic structure and operation principle of a silicon solar cell is described including technological and physical efficiency limitations. Equations and assumptions concerning the front side metallization pattern which are used throughout this work are introduced in detail; the physics of the metalsemiconductor interface is reviewed. The production sequence of a screen-printed industrial silicon solar cell is compared to a high-efficiency laboratory solar cell with a photolithographically defined front in Chapter 2. Different front side metallization technologies are analyzed and discussed. Special attention is drawn to apply a two-layer contact structure, with a low contact resistivity to the underlying emitter surface being required by the first layer and a high conductivity being required by the second layer. In Chapter 3 a theoretical analysis of optical and electrical losses concerning a two-layer contact structure is performed and compared to a conventional screenprinted contact. The influences of the contact width, of the emitter sheet resistance and of the contact resistivity on the efficiency potential are investigated. Furthermore different busbar geometries are considered. Electrical losses caused by the busbar and tab are presented in detail. Based on this investigation, appropriate calculations for the grid optimization in this work are derived. Since the series resistance is a crucial parameter for solar cell characterization, an accurate series resistance determination is of major importance for this work. A minimum between electrical (resistance) and optical (shading) losses due to the front pattern needs to be determined. Different determination methods for measuring the series resistance which is not directly accessible have been proposed in the past. Based on a thorough experimental investigation these approaches are reviewed and discussed in Chapter 4. As a low metal semiconductor contact resistance of the first layer of the twolayer system is most important, different metals are analyzed in their properties as contact material in Chapter 5. The evaporation of the metals is carried out in a high vacuum process on a photolithographically structured front side of highefficiency solar cells. Ni-Ag, Al-Ag, Ag, Pd-Ag, Ti-Ag and Cr-Ag are tested and compared to the standard stack system Ti-Pd-Ag, which is commonly used for metallization of high-efficiency cells. 4 Introduction _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Screen-printing, the industrial state-of-the-art metallization technology, is characterized by achieving high throughput rates using a relatively simple process sequence. Its limitation lies in achieving fingers of fine width while keeping the finger conductivity high. The use of hotmelt silver paste is analyzed with the goal to improve contact geometry resulting in higher energy conversion efficiencies of processed solar cells as presented in Chapter 6. Screen-printing of fine lines is a way to create the first layer of the two layer contact structure as discussed in the second part of Chapter 7. In the first part of the Chapter the light-induced plating process is introduced being the key technology of forming the conductive layer of the proposed two-layer contact structure. A detailed analysis of the electrochemical process is presented. In Chapter 8 the proposed two-layer front side structure has been combined with a high-efficiency rear side featuring dielectric passivation. Fine-line screenprinted contacts on emitters of different phosphorus doping concentrations thickened by the light-induced plating process are used to create the front and the laser-fired contact process to define the rear structure. The processed cells are compared to ones with a conventional aluminum back surface field. An alternative technology to achieve contacts of small width is pad-printing being presented in Chapter 9. In combination with the light-induced plating process, solar cells of high efficiencies are fabricated. In Chapter 10 a new metallization technology is introduced. This metal aerosol jet printer is a non-contact direct-write system with the ability to create contacts of small width. The technology was applied and optimized for solar cell metallization. In Chapter 11 the metal semiconductor contact of fine-line printed and plated contacts is macroscopically and microscopically analyzed. The effect of the contact geometry and the plating step on the contact resistance is discussed. New additional current paths for a plated contact are proposed. The calculation of the different contributions to the series resistance is explained in the Appendix as well as different methods to determine the front side metal semiconductor resistance. Crystalline silicon solar cells 5 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 1 Crystalline silicon solar cells This thesis is about multicrystalline and monocrystalline silicon solar cells with a p-doped base and an n-doped emitter. In this chapter the working principle of this type of solar cells is explained and physical and technological efficiency limitations are described. Furthermore a detailed analysis of losses caused by the front side metallization pattern of the solar cell is presented as well as the physics of the metal-semiconductor contact. 1.1 Structure and operation principle A typical industrial solar cell structure fabricated on p-type crystalline silicon is illustrated in Fig. 1.1. This solar cell is in principle a large-area diode with a relatively thin n-doped emitter layer of 0.2 µm to 2 µm thickness and a 50 µm to 300 µm thick p-doped base substrate. Between the n-doped emitter and the pdoped base a space-charge region1 is formed. The emitter is located on the irradifront contact quasi neutral region sunlight space charge region _ _ + + + + + _ _ _ _ load + + + _ + _ electron, majority carrier (mobile) ionized atom (imobile) + _ E + _ + _ + " hole" , majority carrier antireflection coating n-type emitter p-type base rear contact Fig. 1.1: Schematic drawing of a solar cell. Created electron-hole pairs are separated by the electric field of the space charge region and extracted at the metal contacts. The enlarged section illustrates the p-n-junction. While the space charge region can be regarded as carrier-free, electrons are majority carriers in the n-region and “holes” in the p-region. 1 Space charge region: When bringing a phosphorus doped (n-doped) and a boron doped (p-doped) region into intimate contact, at the interface layer the free electrons of the nregion recombine by diffusion with a defect-electron or “hole” of the p-region, due to concentration differences. In the n-region a positively charged (immobile) phosphorus atom and in the p-region a negatively charged boron atom remains. A static electric field is created resulting in a drift current flow in a direction opposite to the diffusion transport. The equilibrium is reached when drift and diffusion currents have the same magnitude. The region is also called depletion region, as it is depleted of free charge carriers. 6 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ ated side. Electron-hole pairs are generated under irradiation. The minority carriers diffuse to the space charge region and reach the other region. The minority carriers become majority carriers, which reduces the recombination probability dramatically. The photon-generated current flow can be extracted at the front and rear side contacts via an external load. The rear contact usually covers the whole rear side, whereas the front contact is made up of a fine metal mesh to reduce shading losses. To further reduce optical losses on the front side, the emitter surface is textured and an antireflection coating is applied. 1.1.1 Current-Voltage curve FF= 20 jscVoc 40 dark j-V jmppVmpp 2 40 mpp 20 jV 0 0 Vmpp -20 -40 0 jmpp Voc -20 illuminated j-V -40 jsc Power density [mW/cm ] 2 Current density j [mA/cm ] For solar cell characterization the most important parameter is the energy conversion efficiency η, which describes the fraction of the incident power density of the photons being converted into electrical power. In Fig. 1.2 the current density (j) and the power density vs. voltage (V) curve of a solar cell are illustrated. 100 200 300 400 500 600 700 Voltage V [mV] Fig. 1.2: Illuminated and dark IV-curve and the power density vs. voltage curve of a solar cell, illustrating the IV-parameters. The fill factor FF can be regarded as the ratio of the largest rectangle fitting under the IV-curve to the rectangle of side length Voc and jsc. The open-circuit voltage Voc and the short-circuit current density jsc are the intersection points of the current voltage (IV) curve with the voltage and current axis, respectively. The point (jmpp, Vmpp) at which the product of current and voltage has its maximum is called the maximum power point (mpp). The energy conversion efficiency is usually measured under standard testing conditions (25°C, Crystalline silicon solar cells 7 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 1000 W/m², spectral distribution AM 1.5g2) and is described by the following equation: η= Vmpp j mpp Φ = j scVoc FF Φ (1.1) with FF being the fill factor. FF is the ratio of the largest rectangular fitting under the IV-curve to the rectangular of side length Voc and jsc. The parameters defining the performance of a solar cell can be extracted from the current-voltage (IV) curve, which can be measured under dark and under illuminated conditions. Based on the following assumptions [11], it is possible to explain the theoretical IV-curve from the continuity and transport equations of semiconductor physics: - For the occupation density of the bands, the Boltzmann distribution can be used instead of the exact Fermi-Dirac distribution, because the difference between the energy of the conduction/ valence band and the energy of the Fermi level is sufficiently large [12]. - For describing the electrical field distribution, the density of charged doping atoms in the space charge region is relevant. Outside the space charge region the solar cell is field-free; within the space charge region the number of free carriers is negligible. - In the space charge region the difference between drift and diffusion current is small compared to the currents itself and the majority carrier density is far smaller than the minority carrier density. - The current in the quasi neutral regions flows by diffusion and not by drift mechanisms. The ideal IV-curve can then be described by the one-diode-model taking into account the basic physical principles of charge carrier transport. q ⋅V j (V ) = j 0 exp − 1 − j ph k ⋅T (1.2) with q being the elementary charge, k the Boltzmann’s constant, j0 the dark saturation current density and jph the photo-generated current density. As a real solar cell is also influenced by series resistance rs and parallel (or shunt) resistance rp the one-diode-model can be extended taking these characteristics into account. An equivalent circuit diagram is presented in Fig. 1.3. 2 AM 1.5g (Air mass 1.5 global) is the reference solar spectral distribution for terrestrial application defined in the IEC 600904-3. It corresponds to an irradiance of 1000 W/m² on a sun-facing plane surface tilted at 37° to the horizontal. 8 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ The second diode connected in parallel, models the recombination in the space charge region under the assumption of a single recombination centre in the middle of the “forbidden” band gap and a constant recombination rate across the spacecharge region. The two-diode equation is expressed as follows: q (V − j ⋅ rs j (V ) = j01 exp n1 ⋅ k ⋅ T ) q (V − j ⋅ rs + j02 exp − 1 n ⋅ k ⋅T 2 ) V − j ⋅ rs − 1 − j ph + rp (1.3) j01 represents the recombination current density in emitter and base, j02 the recombination current density in the space-charge region and n1 and n2 the ideality factors. rS j01 jph j02 jrp I Vrs sunlight Va D2 D1 V rP Fig. 1.3: Equivalent circuit of a solar cell based on the two-diode equation (eq. (1.3)). The first and second diode (D1 and D2) represent recombination currents in the emitter/base (j01) and space charge region (j02), respectively, the series (rs) and parallel resistance (rp) model electrical losses. The effect of the series and parallel resistance on the IV-curve is illustrated in Fig. 1.4. The maximum power point is directly affected by the series and parallel resistance, whereas the open-circuit voltage (V-intercept) and short-circuit current density (j-intercept) are only reduced for very high rs or low rp values. 5 10 -20 20 -10 a) Voltage V [mV] b) 500 100 50 -20 -10 0 0 100 200 300 400 500 600 700 2 rp=10000 Ω cm -30 2 -30 0 0 -40 2 rs=0.0 Ω cm 0.5 1 2 j [mA/cm ] 2 j [mA/cm ] -40 100 200 300 400 500 600 700 Voltage V [mV] Fig. 1.4: Effect of series resistance rs (a) and parallel resistance rp (b) on the IV-curve. Crystalline silicon solar cells 9 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 1.1.2 Physical and technological efficiency limitations Spectral irradiance [W/m²/µm] The theoretical maximum energy conversion efficiency for a silicon solar cell at one-sun irradiance is about 29% [13]. The four main physical limitations are the following: - Photons with energy smaller than the band gap (1.12 eV) can not generate an electron-hole pair (see Fig. 1.5). - Each photon of sufficient energy can create just one electron-hole pair. Photon energy exceeding 1.12 eV is transferred to heat (see Fig. 1.5). - The maximum achievable open-circuit voltage is far below 1.12 V, since not the separation of the band gap but only the separation of the quasi-Fermi levels3 defines the maximum achievable voltage [14]. High-efficiency silicon solar cells achieve open-circuit voltages of 680 mV to 720 mV. - The maximum power that can be collected from a solar cell is not equal to the product of open-circuit voltage and short-circuit current density, as the current depends exponentially on the voltage. The IV-curve has not a rectangular characteristic (compare Fig. 1.2). This limits the fill factor to about 85% due to non-avoidable recombination currents. In addition to the above mentioned loss mechanisms, optical losses due to reflection of incident light and electrical losses due to series and parallel resistance. The highest energy conversion efficiency reported for a silicon solar cell measured under standard testing conditions is 24.7% [15]. 1500 AM 1.5g spectrum not used energy of short wavelength photons 1000 used energy optimum energy photons of long wavelength having not enough energy to generate carriers 500 0 500 1000 1500 2000 Wavelength [nm] 2500 Fig. 1.5: Schematic drawing of spectral irradiance of the AM1.5g spectrum plotted versus the wavelength. The red shaded area illustrates the spectrum utilization of a silicon solar cell. 3 The quasi-Fermi levels are defining the occupation probability for electrons and holes at irradiation. In contrast to the dark case, for an illuminated semiconductor two quasiFermi levels are necessary, one describing the electrons the other describing the holes. 10 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Typically industrially processed solar cells achieve efficiencies of about 15% to 16% on multicrystalline and 16% to 17% on monocrystalline silicon solar cells. Optical losses To achieve high energy conversion efficiencies, it is important that the incident photons are converted to electrical energy. Reflections of irradiation from the front surface or absorption by the rear need to be prevented. Some of the incident light will be reflected by the front side electrodes, which typically results in 5% to 10% shading loss (see section 1.2.1). The reflectivity of a flat silicon surface is about 36%. Applying a front side texture increases the probability of absorption due to multiple reflections, which reduces the reflectivity to less than 10% (see Fig. 1.6b). Applying an appropriate dielectric layer (single layer or double layer) on top of the textured surface further reduces the reflectivity to about 5%. A further optical loss mechanism is the transmission of photons by the rear side or the absorption in the rear electrode, respectively. The rear side needs to be optically refined, such that photons will be internally reflected. For long wavelength photons the absorption length can be several times the thickness of the wafer (see Fig. 1.6-a). To contribute to electron-hole pair generation, the photons need to be internally reflected by the rear and front side several times. light beam Lichtstrahl 100% 9% 30% texturierte Vorderseite textured front side 70% 21% flat ebener Rückseitenspiegel rear side mirror a) b) Fig. 1.6: Absorption coefficient and absorption length in silicon (left-hand) [16]. The absorption coefficient α is the fraction of incident radiant energy absorbed per unit thickness; the absorption length xa is the distance light needs to travel in a medium till the intensity is reduced to 1/e (37%) of the incident radiant energy. While short wavelength photons are absorbed close to the surface, long wavelength photons need several internal reflections before being absorbed as illustrated in the right-hand schematic [6]. Crystalline silicon solar cells 11 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Recombination losses The light-generated minority carriers need to reach the p-n-junction by diffusion. Unlike the generation, electron-hole pairs are exposed to several recombination mechanisms, occurring independently from each other. For each recombination process i a recombination rate U and a minority carrier lifetime τ is assigned. The carrier lifetime τ is defined as average time an electron or hole is traveling in the crystal before recombining. τi = ∆n Ui or τi = ∆p Ui (1.4) EC EC EV EV surface states with ∆n and ∆p being the excess carrier concentration. The recombination mechanisms are illustrated in Fig. 1.7. light a) b) c) d) Fig. 1.7: Recombination mechanisms occurring in a silicon solar cell a) radiative recombination; b) Auger recombination; c) recombination via recombination centers in the bulk and (d) via surface states. Radiative recombination Radiative recombination is the reverse process of optical absorption. An electron and a hole recombine under the emission of a photon (Fig. 1.7-a). This process is unlikely for the indirect semiconductor silicon as in addition the participation of a phonon is required. For the usual doping concentration radiative recombination is negligible. Auger recombination The energy produced when electron and hole recombine can be absorbed by a second electron or hole in the same energy level. This activated second electron emits its energy in form of phonons to the lattice structure (Fig. 1.7-b). The Auger recombination τAuger is especially active in highly doped semiconductors, as the amount of free electrons n in the conduction band or holes p in the valence band is 12 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ increased. This recombination process can be described for a p- and n-doped region by: τ Auger , n = 1 Cnp + De n ² τ Auger , p = 1 Cnp + D p p ² (1.5) with C Auger recombination rate constant and De, Dp diffusion constant of electrons and holes, respectively. For low injection the minority carrier lifetime is inversely proportional to the square of carrier concentration. Therefore lowly doped emitter and base material lead to less Auger recombination. Recombination through impurity centers and crystal defects in the bulk Impurities and dislocation in the crystal lattice create energy states in the “forbidden” bandgap in the semiconductor. These states create very effective recombination centers. Electrons fall from the conduction band onto the energy state of the trap and from there into the valence band (Fig. 1.7-c). This process was first described analytically by Shockley, Read [17] and Hall [18]: τ SRH = σ p−1 (n0 + n1 + ∆n) + σ n−1 ( p 0 + p1 + ∆n) vth N T ( n0 + p 0 + ∆n) (1.6) with n1 and p1 defined as: E − EV n1 = N C exp T kT E − ET p1 = N V exp V kT (1.7) ET is the energy level of the trap, σn and σp the capture cross section for electrons and holes, vth the thermal velocity of the charge carrier (vth,300K ~ 107 cm/s) and NT the trap density. Hence, the lifetime τSRH depends on the amount of trap levels, the energy state of the trap in the forbidden bandgap and its capture cross section. The closer the energy state to the middle of the bandgap and the higher the capture cross section, the lower is the carrier lifetime. While for FZ-silicon the recombination over traps in the bulk material can be neglected, for typical multicrystalline and Cz-silicon material, this is the main recombination process. Especially for “cheap” silicon, as metallurgical silicon, the amount of traps in the bulk material is high and increases this recombination process dramatically. If contamination of the material can not be avoided, the lifetime can be increased by removing impurities by gettering, by forming clusters of impurities and by passivating trap levels Crystalline silicon solar cells 13 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Recombination at the surface The surface of a semiconductor has a high density of defects due to the abrupt ending of the crystal lattice. The energy levels of these defects can be found over the complete range of the bandgap (Fig. 1.7-c). For analytical description of the recombination rate Us at the surface an extended Shockley-Read-Hall model is used [19]. The recombination activity of all energy levels between valence and conduction band is integrated, assuming a continuous distribution throughout the surface. Similarly as for the lifetime of the bulk, a surface recombination velocity S can be defined, which is typically used for quantifying surface recombination processes: S= ( ) Us n p − ni2 vth = s s ∆n ∆n EC ∫ (n EV s Dit (ET ) d ( ET ) + n1 ) / σ p (ET ) + ( p s + p1 ) / σ n (ET ) (1.8) ns and ps being the electron and hole concentration at the surface and Dit the surface interface density of the defects, ni the electron/ hole concentration for the intrinsic state. The reduction of the surface recombination rate can technologically be achieved by reducing the minority carrier density at the surface and by reducing the interface state density. The reduction of minority carriers can be achieved by either doping the surface layer (e.g. back surface field) or by building-in fixed charges in an overlying dielectric layer (as e.g. SiNx on p-doped surface). This is also called field effect passivation. A decrease of interface states can be achieved by covering the surface by an appropriate dielectric layer like SiO2 or SiN. In this way dangling bonds are passivated by oxygen or hydrogen atoms. 1.2 Loss mechanisms due to front side metallization pattern As this thesis deals primarily with the improvement of the front side metallization of silicon solar cells, a detailed presentation of optical and electrical losses due to the front side contacts is presented. Focusing on the front side metallization process, a reduction of all losses is needed to optimize the solar cell performance. Optical losses due to reflection of incident light on the metal grid and electrical losses due to the series resistance have to be balanced. The series resistance mainly affects the fill factor of a solar cell. For large-area industrial silicon solar cells the fill factor drops by about 4.5%abs to 5.5%abs per 1 Ω cm² increase in series resistance (compare section 4.2). Fig. 1.8 illustrates the series 14 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ resistance contribution of an industrial silicon solar cell, as used for most investigations within this work. The current on the rear side is collected by a full metallized aluminum layer, on the front side by an H-grid pattern with fine metal fingers. The total series resistance rs is equal to the sum of all resistance contributors (series connection). rtab rbus rsj rf rc re rb rm_rs rrc rm_rs: resistance of rear side metal layer rrc: contact resistance of rear side to base rb: resistance of base re: resistance of emitter rc: contact resistance of front grid to emitter rf: line resistance of finger rbus: line resistance of busbar rsj: contact resistance of soldering joint rtab: line resistance of tab rext: line resistance tab extension Fig. 1.8: Solar cell cross-sectional diagram showing the different series resistance contributors. 1.2.1 Optical losses On the front side of the silicon solar cell, some of the incident light impinging the metal contact will be reflected and lost for current generation. Usually the contact covered area is determined and divided by the total surface area in order to calculate the loss fraction due to grid shading [20]. The total coverage fraction pc is the sum of the coverage fraction of finger pc,f and busbar pc,bus: pc = pc,bus + pc, f = wbuc w f l f s ⋅ wbuc + l f ⋅ w f + = a s⋅a s⋅a (1.9) with wbuc being half the busbar width; wf the finger width; lf the finger length; s the finger separation distance and a the sum of wbuc and lf (see also Fig. 1.10). The coverage fraction of typical industrial silicon solar cells is about pc = 7% - 10%. Fig. 1.9: Direct reflection from the metallization pattern and indirect reflection from the glass-air interface onto the silicon surface. From left to right: Geometry of a typically plated contact, an advanced plated contact [21] and a screen-printed contact. Crystalline silicon solar cells 15 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ However, this calculation neglects that some of the incident light that hits the metal grid, still will be absorbed in the solar cell due to (multiple-) reflection as illustrated in Fig. 1.9. Some of the incident light that hits the contact will be directly reflected onto the silicon surface. For direct reflection especially triangular shaped contacts or roundish contacts are preferable. Multiple reflections occur when the solar cell is encapsulated in a module [22]. Some of the light will be reflected from the glass-air interface back onto the active cell area. For a thin plated contact up to 60% of the irradiation hitting the finger will be reused after encapsulating the cell in a module, as published by Blakers [22]. Burgers stated that also for e.g. a screen-printed contact up to 55% of the light hitting a finger is reflected into the cell. This is the case when the incident light will be completely diffused on the rough contact surface [23]. Taking reflection effects into account an effective contact transparency factor was introduced for the busbar/tab tb and the finger tf [23]. The shading fraction of busbar ps,bus and finger ps,f is equal: ps ,bus = (1 − tb ) wbuc a ps , f = ( (1 − t f ) w f l f + wbuc tb ) (1.10) s⋅a and the total shading fraction ps by: p s = p s ,bus + p s , f = ( )( s ⋅ wbuc (1 − t b ) + w f l f + wbuc t b 1 − t f ) (1.11) s⋅a 1.2.2 Electrical losses Electrical losses are caused by the series and parallel resistance. The individual contribution of the series resistance is illustrated in Fig. 3.2 of Chapter 3.2.1. For ohmic loss calculation the H-grid pattern can be broken down into different unit cells. Each unit cell represents the smallest symmetry element of an individual conductor. The individual resistance contribution to the total cell resistance is achieved by connecting the conductor x times in parallel (x equals number of correspondent unit cells). This value can be calculated by dividing the resistance of a unit cell by the number of unit cells. Typically instead of calculating the total series resistance contribution the resistance is weighted to the area. In this way the series resistance of solar cells of different sizes can be easily compared. As illustrated in Fig. 1.10, unit cell I is equal to all current entering the finger from one side, unit cell II to half the current collected by one soldering joint and unit cell III corresponds to the current collected by one tab. In Table 1.2 the different resistance contributions are summarized. For detailed resistance calculation see Appendix A. 16 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 1.1: Definition of symbols used for loss calculation and parameters describing the H-grid pattern of Table 1.2. Description unit contact resistivity (front) Ω m² Nsj number soldering joints contact resistivity (rear – base Ω m² line resistivity busbar Ωm p Af power loss fraction finger cross section area m² base resistivity Ωm Auc area unit cell m² line resistivity finger Ωm a length unit cell I m line resistivity metal rear side Ωm b width unit cell II m line resistivity tab Ωm d thickness of wafer m Rsh sheet resistance emitter Ω/ Rsk sheet resistance under contact Ω/ Rsj resistance soldering joint Ω Reff effective series resistance Ω ρc ρrc ρbus ρb ρf ρm_rs ρtab r ja Va tb tf area weighted series resistance Ω m² current density active cell area mA/cm² voltage active cell area V effective transparency factor busbar effective transparency factor finger - Description unit - hbus height busbar m height finger m hf hm_rs height metal rear side s lf lte, wb(uc) wf wtab m finger separation distance m length finger m length tab extension width busbar (unit cell) width finger width tab wbuc=1/2 wb m m m m wf a lf unit cell: I unit cell: III wb unit cell: II s b 2a lte Fig. 1.10: Η-grid metallization pattern illustrating unit cell I, II and III used for resistance calculations. Geometrical parameters are described in Table 1.1 Crystalline silicon solar cells 17 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 1.2: Resistance contribution of a solar cell with an H-grid pattern and a fully metallized rear side. For each conductor the equation for the effective resistance Reff of a unit cell and the area-weighted resistance r is presented. The area weighted resistance is achieved by multiplying the effective resistance with the area of the unit cell. For definition of symbols see Table 1.1 and Fig. 1.10. Resistance unit cell [m²] Emitter I: a ⋅ s 2 Emitter contact I: a ⋅ s 2 Finger I: a ⋅ s 2 Reff: Resistance unit cell [Ω] r : area-weighted resistance [Ω m²] s − wf 1 Rsh 6 lf s − wf 1 Rsh as 12 lf ( R sk ρ c lf wf coth 2 ) Rsk ρ c Rsk s a ρ c 2 b2 1 ρbus a hbus wbuc 3 4 ⋅ RS ⋅ a ⋅ b II: 2 ⋅ a ⋅ b Solder joint II: 2 ⋅ a ⋅ b 2 RS Tab III: 4 ⋅ a ⋅ b ⋅ N s 2 b 2N S + 1 ρt ht wt 3N s Tab extension III: 4 ⋅ a ⋅ b ⋅ N s Metal layer rear side lf wf coth 2 lf 1 ρf as Af 3 Busbar Base contact Rsk ρ c ) lf 2 ρf 3 Af 2 b ρ bus 3 hbus wbuc Base ( ( ρt ) lte ht wt total cell area ρb ⋅ d 8 ⋅ a ⋅ b ⋅ Ns 8⋅ Ns ⋅ a ⋅b total cell area ρ rc 8 ⋅ a ⋅ b ⋅ Ns 8⋅ Ns ⋅ a ⋅b 2 ⋅ a ⋅b ⋅ Ns 1 ρ m _ rs ⋅ l f 3 2 ⋅ b ⋅ N s ⋅ hm _ rs a ⋅ lbus 2 2 ρT 3 ht ⋅ wt 4 ⋅ ρt 1 + 1 2N s 2 lte Ns ⋅ a ⋅b ht wt ρb ⋅ d ρ rc 1 ρ m _ rs ⋅ l f a 3 hf Power output of a unit cell, not limited by electrical and optical losses: The loss fraction contributed by each individual resistance is the ratio of the power loss caused by the resistance of a unit cell and the maximum power that could theoretically be achieved for the unit cell. The maximum theoretical power output Pmax of a unit cell of size Auc, assuming no shading or ohmic losses, is defined by: Pmax = j max ⋅ Va ⋅ Auc (1.12) Va and jmax are voltage and current density without shading and electrical losses due to the front grid. The current density j is limited by shading losses. Assuming no shading losses the maximum current density jmax is equal to: 18 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ j max = j 1 − ps (1.13) with ps being the shading fraction caused by the front grid (eq. (1.11)). Thus, the current density ja between the contacts (active cell area) would also be equal to jmax, if no light is reflected from the contact fingers into the cell. However, due to reflection of light from the contacts onto the active cell area, the current density ja is even higher than jmax. ja = j 1 − pc (1.14) with pc being the coverage fraction of the contact grid (eq. (1.9)). Contrary to the current j, the voltage output V is limited by electrical, not by shading losses. The voltage across the active cell assuming no electrical losses can be calculated by: Va = V + rs j (1.15) Applying Eq (1.15) and (1.13) in (1.12) results in: Pmax = j (V + rs ⋅ j )Auc 1 − ps (1.16) Power loss due to an individual resistance: Furthermore the loss contribution of an individual resistance is determined. As the current density is not the same at each point in the conductor, an effective series resistance Reff and an area-weighted resistance r is introduced (see Table 1.2). Typically the current increases linearly from 0 to I along the conductor, as it is e.g. the case for the emitter, finger and busbar. For a detailed determination of the effective series resistance see Appendix A. The power loss Ploss in the conductor can be calculated by the following equation: 2 2 Ploss = Reff I uc (i ) = Reff ( jAuc ) Ploss = r I uc (i ) 2 Auc (i ) = r ⋅ j 2 Auc (1.17) Ploss is the ohmic power loss in the conductor and Iuc(i) is the corresponding unit cell. Crystalline silicon solar cells 19 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Power loss fraction of an individual resistance Finally the power loss fraction caused by each conductor is determined by dividing the power loss caused by the conductor Ploss by the power gain Pmax of the correspondent unit cell. Using equations (1.16) and (1.17) the power loss can be calculated: p e, x = Ploss j (1 − p s ) ⋅ r = Pgain Va (1.18) The total electrical loss fraction pe is the sum of all individual loss fraction contribution (pe = pe,m_rs + pe,rc + pe,b + pe,e + pe,c + pe,f + pe,bus + pe,sj + pe,t + pe,ext)4. 1.2.3 Grid optimization for maximum power point A widely used approach is optimizing the grid for maximum power point at standard testing conditions [23], which will be briefly presented. Other approaches are optimising for yearly yield or using the two diode equation [24]. The current density at mpp using equation (1.13) is equal to: J mpp = J mpp ,a (1 − p s ) (1.19) and the voltage at mpp: Vmpp = Vmpp ,a − RS ⋅ J mpp = Vmpp − Rs (J mpp ,a (1 − p s )) (1.20) Hence the power at maximum power point can be calculated by: Pmppp = J mppVmpp = (J mpp ,a (1 − p s ))(Vmpp − Rs (J mpp ,a (1 − p s ))) 2 = J mpp ,aVmpp ,a − J mpp ,aVmpp ,a p s − (1 − p s ) Rs J 2 mpp ,a (1.21) Assuming no power loss the maximum power output Pmpp,max would be equal to: Pmpp ,max = J mpp ,aVmpp ,a (1.22) Using eq. (1.21) and (1.22) the total loss fraction p can than be calculated by: p= Pmpp Pmax,mpp = 1 − p s − (1 − p s )2 Rs J mpp ,a Vmpp ,a = 1 − p s − p e,mpp (1.23) In order to retrieve the minimum loss, the sum of ps and pe,mpp need to be minimized. 4 Indices: m_rs: metal rear side, rc: rear side contact; b: base; e: emitter; c: contact front side; f: finger; bus: busbar; sj: soldering joint; t: tab; ext: tab extension 20 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 1.3 Metal-semiconductor contact For the front side metallization of industrial solar cells the contact resistivity becomes important. Whereas the resistivity is low for evaporated contacts which are used for high-efficiency cells, the contact resistivity of industrial screen-printed silicon solar cells is not negligible. Thus current flow mechanism and contact resistivity calculation of a metal-semiconductor contact will be briefly presented. The ideal contact for solar cells is an ohmic contact having a negligible contact resistance compared to the bulk resistance or series resistance of the semiconductor. In addition the contact should not degrade the device performance. However, the first metal-semiconductor contacts had a rectifying behavior, which was explained by a potential barrier formed in-between the two materials in 1938 by Schottky [25,26]. This model is illustrated in Fig. 1.11 and is known as the Schottky barrier. The potential barrier φΒ that is formed can be described for an ntype semiconductor by: qφ Bn = qφ M − qχ S (1.24) with qφm being the metal work function, representing the potential necessary to excite an electron from the Fermi level to the vacuum level and qχ the electron affinity defined as potential difference between the bottom of the conduction band and the vacuum level. Vbi is the built-in potential in the semiconductor. qVbi = φ Bn − qVn (1.25) with Vn being the voltage between the bottom edge of the conduction band and the Fermi level. Vacuum level φM EF Vacuum level χS φ S Ec EF φB Vn EV Metal Vbi Ec EF EV Semiconductor Metal Semiconductor Fig. 1.11: Energy band diagram of a metal-semiconductor contact before and after bringing into intimate contatact. The Schottky barrier height φB is illustrated. Crystalline silicon solar cells 21 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ If equation (1.24) would describe real contacts correctly, a metal need to be chosen with a metal work function φM smaller than the electron affinity χs of the semiconductor. This type of contact is called accumulation contact. However, for a real contact the direct proportionality between potential barrier and metal work function could never be measured. Fig. 1.12 shows measured barrier heights of different metals for n-type silicon. Although there is a relationship between barrier height and metal work function, it is not as strong as predicted by equation (1.24). Surface states at the interface layer, acting as donors or acceptors, play an important role for contact formation, as first pointed out by Bardeen [27]. A review paper of different contact formation models was published by Meier and Schroder, which is suggested for more detailed information [28]. Fig. 1.12: Measured barrier heights φBn for different metals plotted versus the metal work function φM for n-type silicon. [28] Image force effect If an electron of charge q in the vacuum is located at a distance x away from the metal interface, the free carriers in the metal are arranged in a way as if effectively at the position -x the charge q would be present. This scenario is called image force. The image force F(x) of the coulomb potential of both carriers attracts the electron in the vacuum. To excite an electron from the metal to the vacuum level, an external electrical field is applied. This potential is superimposed by the potential of the image force. The metal work function of the electron is reduced. The point xm at which the potential energy is maximal is illustrated in Fig. 1.13. The same principle can be applied for a metal-semiconductor interface, for which also a reduction of the potential barrier ∆φBn is achieved by the image force [29]: 1 ∆φBn q 3 N D (Vbi − V − k BT / q ) 4 = 2 3 8 π ε S (1.26) 22 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ with εs being the dielectric constant of the semiconductor. The potential barrier lowering ∆φBn depends on the maximum field intensity, the width of the space charge region in the semiconductor and is hence indirectly dependent on the doping concentration and the externally applied voltage. Fig. 1.13: Graph showing the location dependence of the potential energy (solid line) externally from the metal and the contribution of the image force potential and the externally applied electrical field [29]. Current flow mechanism and contact resistance The current flow over the metal-semiconductor contact is primarily defined by majority carriers and can be described by three processes (see Fig. 1.14): - Thermionic Emission (TE): Transport of electrons over the potential barrier. - Field Emission (FE): Tunneling of electrons through the potential barrier. - Thermionic field emission (TFE): Combination of field and thermionic emission. EC EC EF EF EV EV Metal Semiconductor TE Semiconductor TFE Semiconductor FE Fig. 1.14: Current flow mechanism between metal and semiconductor. From left to right: thermionic emission (TE), thermionic field emission (TFE) and field emission (FE). Crystalline silicon solar cells 23 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ In the following the current flow processes and the calculation of the contact resistance are presented more in detail. The resistivity is generally calculated by: ∂V ρ c = ∂j V =0 (1.27) Thermionic emission: The thermionic field emission theory from Bethe assumes that the current crossing the metal semiconductor interface depends on the height of the potential barrier (Schottky barrier) [29]. Only those electrons with energy greater than the barrier height contribute to the current transport. The total current flow is equal to: q ⋅ (φ Bn − ∆φ Bn ) q ⋅ V jTE = A * ⋅T 2 exp − ⋅ exp − 1 k ⋅T k ⋅T (1.28) with A* ≈ (m*/m0)·120 A m-2 K-2 being the effective Richardson constant, m* the effective electron mass and m0 the free-electron mass. The contact resistance ρc for thermionic emission is given by: ρ c (TE ) = k q (φ Bn − ∆φ Bn ) exp − q ⋅ A * ⋅T k ⋅T (1.29) Field emission: For very low temperatures and/ or high doping concentration the current flow can not be described by the thermionic emission model. The transport mechanism has changed. As the low temperature region is not important for solar cell application, only the high doping case will be discussed. With increasing doping concentration ND the width w of the space charge region shrinks. At a doping concentration of ND ≥ 1019 cm-3 the width of the space-charge region of the metalsemiconductor interface is strongly reduced, that the probability for quantum mechanical tunneling is significant. The proportionality can be expressed by: kT 4π N D Vbi − V − q qV j FE ∝ exp − exp k ⋅ T − 1. ∗ ( ) h m ε φ − ∆ φ tunnel S Bn Bn (1.30) The contact resistance for the field emission is equal to: ρ c (FE ) = C1 with q (φ Bn − ∆φ Bn ) k exp q ⋅ A * ⋅T E 00 (1.31) 24 Crystalline silicon solar cells _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ C1 = π π ⋅ k ⋅ T 4 ⋅ φ Bn sin ln 2 E ⋅ 00 EF 4φ E F ln Bn E 2 E 00 F exp − − 2 E 00 φ k ⋅ T ln 4 B EF (1.32) and E 00 = q⋅h 4 ⋅π ND ∗ mtunnel ε S (1.33) The characteristic energy E00 is in relationship to the quantum mechanical tunneling probability of the electrons and rises to the square with the doping concentration ND. Thermionic field emission The thermionic field emission is a combination of the two previously described transport processes. The electrons are thermally activated, until the width w of the potential barrier is thin enough to allow quantum mechanical tunneling [30,31]. The current flow jTFE can be expressed by: jTFE ∗ 2 ε S mtunnel ∗ 2 1 exp =AT C2 qh (φ Bn − ∆φ Bn ) qV − 1 exp E 00 k ⋅ T N D cosh k ⋅T (1.34) with E E k BT cosh 00 coth 00 EF EF k BT k BT − C2 = exp k BT π (φB + EF )E00 E00 coth E00 k T B (1.35) The contact resistivity can be written as [30]: kB q (φ Bn − ∆φ Bn ) ρ C (TFE ) = C 2 exp E q ⋅ A∗ ⋅ T E 00 coth 00 k ⋅T (1.36) The ratio k·T/E00 is a measure of the ratio of thermionic emission current to field emission current - k·T/E00 >>1: For lowly doped semiconductors thermionic emission (TE) dominates (ND<1017 cm-3) - k·T/E00 <<1: For high doping concentration field emission (FE) dominates (ND > 1019 cm-3). Crystalline silicon solar cells 25 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ - k·T/E00 ≈ 1: The thermionic field emission (TFE) dominates for moderately doped semiconductors (1017 cm-3 < ND < 1019 cm-3). Fig. 1.15 and Fig. 1.16 illustrate the dependence of the contact resistivity on the doping concentration. The strong reduction of the contact resistivity with TE ln ρc kT(qA)-1exp(φB/k/T) TFE FE φB 2(m*tunnel ε)1/2 (q h)-1 ND-1/2 Fig. 1.15: The logarithm of the specific contact resistance is plotted against the inverse root of the doping concentration. The regions in which each of the three current transport mechanisms (thermionc emission TE, thermionic field emission TFE and field emission FE) dominates can be clearly identified ([30] redrawn). increasing doping concentration is clearly visible in the region of field emission. For thermionic emission the contact resistance is independent of the doping concentration and solely attributed to the barrier height and temperature. Inbetween is the region of thermionic field emission. The curve progression for TE and FE can be directly extracted from equation (1.29) and (1.31). m*/m=1 Ks=12 ni=1.4 x 1010 cm-3 106 104 1 eV φ B= eV 0.8 ρc (Ω cm²) 102 0. 6 1 10-2 0.4 e eV V 10-4 0.2 eV 10-6 10-8 1016 1017 1018 1019 -3 ND (cm ) 1020 1021 Fig. 1.16: Contact resistivity ρc as a function of doping concentration ND and barrier height φB ([28] redrawn). 26 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2 Cell processing and metallization technologies The common industrial and laboratory process sequence of a silicon solar cell will be briefly presented. Different technologies being used for the metallization process as well as new technologies that were considered for solar cell front side metallization are reviewed in detail. 2.1 Processing of silicon solar cells 2.1.1 Screen-printed solar cell In industrial mass production the screen-printed (SP) solar cell is the most common type of solar cell fabricated (see Fig. 2.3-a). This cell is characterized by a relatively simple and well-known production sequence with high throughput rates. For its production typically boron-doped multicrystalline or monocrystalline wafers with an area of 156 cm² to 243 cm² and a thickness of 180 µm to 250 µm are used. To increase irradiation absorption, the front side is textured using an acid solution in case of multicrystalline wafers or an alkaline solution in case of monocrystalline wafers. In the next step the p-n junction is created by phosphorus diffusion resulting in an emitter sheet resistance of about Rsh = 55 Ω/sq – 60 Ω/sq and a surface doping concentration of about 2·1020 cm-3. An improved irradiation absorption and surface passivation is achieved by applying a SiN antireflection coating by PECVD (plasma enhanced chemical vapor deposition) or sputtering technology [32]. Front and rear side contacts are deposited using the screenprinting technology. The rear side is completely covered by an aluminum paste except for small Al/Ag solder points, the front side by a grid like structure of silver paste. In the high temperature contact formation process, the paste on the front etches through the SiN antireflection coating. An electrical contact of good adhesion with sufficient low line conductivity is created to the relatively highly doped emitter layer. During the high temperature process the aluminum on the rear side creates a p+ region, the so-called aluminum back surface field (Al-BSF). After the described process steps, the n-region is usually electrically connected with the p-region over the edges of the wafer, resulting in a shunt. This connection is separated either by a plasma etch, by a wet chemical single side silicon etch or by a laser edge isolation step. Cell processing and metallization technologies 27 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ monocrystalline wafer Texture KOH bath multicrystalline wafer Saw damage etch NaOH Deposition of contact material Screen-printing of Al, Ag paste Emitter diffusion Tube furnace Contact formation Inline fast firing belt furnace Edge isolation Laser edge isolation process Phosphor silicate glass removal HF bath IV-Characterization Steady state sun simulator Deposition of antireflection coating SiNx:H sputtering machine solar cell Fig. 2.1: Schematic drawing of the base line processes and technologies applied for the production of a screen-printed solar cell as used within this work. 1E21 Em40: Rsh = 40 Ω/sq Em60: Rsh = 60 Ω/sq 1E20 Em90: Rsh = 90 Ω/sq Em120: Rsh = 120 Ω/sq -3 Dopant concentration [cm ] Fig. 2.1 shows the production sequence and applied technologies for solar cell fabrication as used for most of the solar cells within this work. One important parameter for contact formation and hence for this work is the sheet resistance of the emitter, including emitter profile and surface doping concentration (NDS). Profiles of some emitters that were frequently used for this work were measured by secondary ion mass spectrometry (SIMS, see Fig. 2.2). The emitter with a sheet resistance of 40 Ω/sq (em40), 60 Ω/sq (em60) and 90 Ω/sq (em90) correspond to a typically industrial created one, whereas the emitter (em120) with a sheet resistance of 120 Ω/sq was used for laboratory highefficiency cells. The depth of em120 is significantly increased while the surface doping concentration is reduced. This is advantageous to increase the blue response (to reduce the emitter dark saturation current). A high surface doping concentration is beneficial to reduce contact resistivity between contact finger and silicon. Especially for screen-printed solar cells, the surface donor concentration needs to be high (NDS > 10-20 cm-3), in order to achieve a sufficiently low contact resistivity. 1E19 1E18 1E17 1E16 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Emitter depth [µm] Fig. 2.2: Phosphorus concentration plotted versus emitter depth for different sheet resistance emitters. Profiles were measured by SIMS. 28 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2.1.2 PERL and LFC solar cell The type of solar cell with probably the highest efficiency potential is the PERL (passivated emitter, rear locally-diffused) solar cell [33]. Minimizing the recombination losses at the front and rear surface is of major importance. This is achieved by applying dielectric layers to both sides. In addition the area under the point contacts is locally p+ doped and the area under the front contacts n++ doped in order to achieve low surface recombination velocities and low contact resistance. In-between the point contacts on the rear the dielectric layer (covering more than 99% of the cell area) passivates the rear surface and serves in combination with the aluminum on top as a nearly perfect optical mirror. Effective surface recombination velocities5 below 100 cm/s can be achieved at the rear surface [34]. In combination with a textured front side, optimal light trapping is achieved. Compared to a screen-printed solar cell, the emitter sheet resistance is increased (Rsh > 100 Ω/sq). Hence, Auger recombination in the emitter layer is also reduced and an improved front surface passivation can be achieved. Instead of creating random pyramids, an inverted pyramids texture is applied, by photolithographical definition. Another photolithographical step is used to define the emitter contacts. Applying this cell structure the world record efficiency of 24.7% on a monocrystalline silicon solar cell has been achieved. However, this process sequence is very complex and therefore not used in industrial application. The laser-fired contact (LFC) process is a simpler way to fabricate solar cells with a high quality rear surface [35] (see Fig. 2.3-b). A dielectric layer is deposited a) SP-solar cell b) LFC solar cell Fig. 2.3: Sectional drawing of a screen-printed (SP) and laser-fired contact (LFC) solar cell. 5 The effective surface recombination velocity is an averaged value of the relatively high recombination velocity of the point contacts and the low velocity of the passivated areas. Cell processing and metallization technologies 29 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ onto the rear side. On top a layer of 2 µm aluminum is evaporated and point contacts are created by a laser process, which results in a p+ region directly under the point contacts. The rear side damage caused by the e-gun evaporation and laser process is annealed at 400°C for 10 minutes under forming gas ambient. Using this process the world record efficiency on multicrystalline silicon (η = 20.3%) was achieved [36]. However, for the front side structure an easy and efficient contact formation process with an efficiency potential comparable to photolithographically defined contacts has not been achieved up to date. 2.1.3 Comparison of an evaporated with a screen-printed front The front structure of an industrial screen-printed solar cell differs strongly from the one of a laboratory high-efficiency cell. Model calculations of quantitative breakdown losses comparing both types of cells clearly illustrate the strong limitation of the screen-printed front [37]. The loss can be subdivided into three categories: contact quality, shortwavelength response and reflectance. The screen-printed (SP) contact leads to 0.5% absolute efficiency reduction compared to a photolithorgraphically (PL) defined contact due to the relative low line conductivity, high emitter sheet resistance and high contact resistance between emitter surface and contact finger. The specific line resistance of a SP contact is about a factor of two higher than for a PL defined contact of pure silver. Additionally, the finger separation distance needs to be increased due to broader fingers. This enlarges the current collecting area of each finger, resulting in a higher current density in the conductor. The enlarged finger separation distance is the reason for elevated resistance losses in the emitter layer, even though the emitter doping concentration is higher. The enlarged grid shading is responsible for another 0.5% efficiency loss caused by the five to ten times broader fingers. The third category, the poor short wavelength response of an SP cell, is responsible for 0.7% loss in efficiency compared to a PL cell. The poor response is due to the higher doping concentration of the emitter layer, necessary to achieve acceptable contact resistance. The higher emitter doping concentration elevates the Auger recombination and does not allow front surface passivation. Altogether, losses due to the different front structure of a SP solar cell lead to an efficiency reduction of about 1.7% absolute compared to a PL cell [37]. In Table 30 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2.1 typical values of geometrical, electrical and process correlated parameters of a screen-printed and a high-efficiency front are presented. Table 2.1: Typical values for geometrical and electrical parameters of a front side structure with screen-printed (SP) and photolithographically (PL) defined contacts. parameter symbol contact width wc finger width wf average finger height hf aspect ratio (height to width) AR finger separation distance s metal coverage (finger + busbar) pc finger resistivity ρf emitter sheet resistance unit µm µm µm mm % Ωm SP cell 100 – 150 100 – 150 10 – 15 1:8 - 1:10 2 – 2.5 7 –10 3.0 – 3.5·10-8 PL cell 3 – 10 15 – 30 10 – 15 1:2 - 1:3 0.8 – 1.5 2–4 1.6 – 1.9 10-8 Rsh Ω/sq 40 - 60 100 – 120 contact resistivity ρc temperature of contact formation throughput rate process complexity T - mΩ cm² °C wafer/hour - >1 700 - 850 1000 low < 0.3 300 - 400 n.a. high 2.2 Metallization technologies In the following subsection the two main field of metallization technologies are briefly discussed; thick-film and thin-film technology. However, it is difficult to distinguish the discussed processes into thick-film or thin-film technologies. All technologies commonly regarded as thick-film technology as screen-printing and pad-printing can also be applied to deposit thin-films. 2.2.1 Thick- and thin-film metallization technologies Thick-film technologies Applying thick-film technology the entire contact layer is deposited in one process step. Thick-film is defined by a layer with a thickness of 5 µm or greater formed by ink, paste, coating or spray-printing and subsequent sintering or firing. One distinguishes between direct e.g. screen-printing and indirect e.g. pad-printing Cell processing and metallization technologies 31 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ thick-film technologies. An overview of different thick-film technologies is given by e.g. Hahne [38]. Thick-film paste: For thick-film front side metallization of solar cells typically a silver containing paste is used. A large variety of commercial pastes are available for screen-printing application. These pastes consist of silver particles, glass frit, solvents and binding agents. The final contact should feature a high aspect ratio, elevated line conductivity, a low contact resistance to the underlying emitter layer and a good mechanical adhesion to the silicon surface. The silver particles with a size of 1 µm to about 7 µm are either fine spherical or large flake-like in order to obtain a high packing density and large contact area, but without causing severe agglomeration. However, agglomeration of silver particles is a common fact, so that the screen openings should be at least 5 times larger than the largest particle size, in order to avoid screen clogging [39]. Solvents prevent silver particles from interaction. In combination with binding agents the rheological behavior is defined to achieve optimum print settings. The viscosity of the paste needs to be low, when the paste is forced through the screen and high as soon as the paste is in contact with the substrate. The relaxation time6 of the paste should be on the one hand as long as necessary to level out meshmarks7, on the other hand short enough to keep line broadening as small as possible (see also Chapter 6.1.2). The glass frit, typically consisting of lead-oxide-containing silicate glass, influences the adhesion to the silicon, the contact formation process and even the conductivity of the finger. It penetrates the SiNX layer, reduces the melting point of silver and is responsible for silver crystallite formation (see Chapter 11.1). Contact formation process: After the printing process the contact is formed in a high temperature process. This process is commonly carried out in an inline fast firing belt furnace with typically three to four temperature zones (see Fig. 2.4). In the first zone the solvents in the paste are evaporated at a temperature up to 300°C. In the next zone between 300°C and 500°C the organic binders are burned. The contact is formed in the third zone in which the wafer is rapidly elevated to a temperature of about 800°C and held at this temperature for a period of one to five seconds. In the last zone the wafer is either directly cooled down to room 6 Relaxation time is defined as the time the paste or another system needs to go return from its low viscosity due to the shearing force back to its high viscosity. 7 Meshmarks are fine crosshatch imprints left by the mesh of the screen during the screen-printing process and are a result of the paste rheologie. 32 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ temperature or held at an elevated temperature between 300°C and 500°C for a short period which has been reported to improve the finger conductivity [40]. contact formation Temperature [°C] 800 600 resolidification of Si-Al at 573°C burn out of organics cooling down 400 evaporation of solvents 200 0 20 40 60 Time [s] 80 100 Fig. 2.4: Schematic drawing of the wafer’s temperature during the contact formation process. The firing process can be subdivided into four functional zones: Evaporation of solvents, burn out of organics, contact formation and cooling down. Thin-film technologies The film thickness created with thin-film technologies can range from a layer of single atoms or molecules to a thickness of a few micrometers, but usually refers to those of less than 1 µm. A thin-film is able to change properties such as color, reflectivity, and friction coefficient of the substrate, while the shape of the substrate is left practically unchanged. Thin film technologies applied for solar cell manufacturing are e.g. the sputtering or chemical vapor deposition technology used to apply a thin layer of SiNX to the substrate surface. For solar cell metallization the vacuum evaporation process of metal contacts can be regarded as a thin film technology as well as the plating process of a thin nickel layer as used for the laser buried contact solar cells (see below). 2.2.2 Screen-printing Screen-printing constitutes a fast and reliable metallization technology with the advantage that the front side structure is printed in one process step. The contact formation takes place in the following high temperature process. For solar cell metallization three printing steps are performed. In the first step two busbars are printed using Ag/Al paste onto the rear, followed by printing the remaining area Cell processing and metallization technologies 33 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ with Al paste8. The front structure is either printed before or after the rear side printing process. In-between the printing steps, the drying of paste take place in a conveyor belt furnace. A single industrial screen-printing line has a throughput of about 1000 wafers/hour. This implies that one printing process, including transfer of the wafer, alignment and the printing step takes less than 3 seconds. Printing Step: The printing step in which the paste is forced through the openings of the emulsion layer onto the surface of the wafer, can be subdivided into three consecutive phases (see Fig. 2.5). In the filling phase, the open areas of the screen are flooded by moving a squeegee (floodbar) over the surface of the screen without applying a vertical force. Depending on the application, the filling phase can also be carried out directly before the contact phase by the paste bead lying in front of the printing-squeegee (Fig. 2.5 (1)). A floodbar would not be necessary in this case. However, direct flooding after the print through process, as typically performed for solar cell metallization, has the advantage that a drying of not released paste in the screen openings is significantly reduced. In the contact phase a vertical force is applied to the printing-squeegee, pressing the screen onto the wafer and forcing paste through the screen openings. The paste sticks to the substrate due to adhesion forces. In the final phase, the paste is released out of the screen [41]. The quality of the print image depends mainly on the screen, the paste and on the printing parameters. Fig. 2.5: Screen-printing process: (1) The openings in the screen are filled with paste; (2) the squeegee brings the screen in intimate contact with the substrate and presses paste through the openings. (3) While the screen is lifted up, paste is released from the screen and sticks to the substrate. 8 For interconnection of the solar cells into a module, a tab is soldered onto the Ag/Al busbar on the rear. Soldering on a pure aluminum rear is not yet possible. 34 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ As illustrated in Fig. 2.6-a, the screen consists of an aluminum frame, a mesh of wires being clamped to the frame and an emulsion layer, which is photolithographically structured with the desired printing image. The size of the frame needs to be large enough that the mesh releases from the substrate and paste during the snap-off (see below). In order not to damage the mesh, the screen tension must be smaller than the elasticity limit of the used wire material. frame wire mesh emulsion layer frame a) b) Fig. 2.6: Structure of a screen illustrating the mesh, the emulsion and the frame as well as characteristic parameters: ws is the thickness of the screen mesh, wst the thickness of the emulsion layer, wsg total thickness of the screen, c the cross section of the wire and d the wire separation distance. Fig. 2.6-b shows a side view (left-hand) and top view (right-hand) of a wire mesh illustrating parameters for characterization. The screen opening fraction a0, describing the ratio of the screen opening area to the total screen area, is defined by: a0 = d² (2.1) (c + d )2 Especially for fine-line printing being the case for the front side metallization process of solar cells, a mesh with fine wires and a high mesh separation distance is desired, as this leads to a high screen opening fraction. However, the danger of screen breakage rises strongly with reduced finger cross section area and increased finger separation distance. That is the reason why mainly steel wires are used possessing a high tensile strength. In addition electrostatic charging of steal wires does not occur. The minimum line width wf_min that can be achieved was defined by Scheer [42] in dependence of the wire cross section c and wire separation distance d by an empirical equation: w f _ min = c (2c + d ) d 2 (2.2) Cell processing and metallization technologies 35 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ For a screen with a wire thickness of c = 25 µm and a 280 mesh9, the wire separation distance is equal to d = 65.7 µm. Applying the latter equation results in a minimum finger width of wf_min = 62 µm. To achieve very fine contacts of e.g. wf_min = 30 µm width, a wire thickness of e.g. c = 14 µm and a 370 mesh would be necessary. The theoretical paste volume vpaste being transferred can be calculated by [42]: V paste = wst ⋅ d ² (2.3) This theoretical value neglects that the area above and below the wire is also flooded with paste, which would increase the paste transfer slightly. On the other hand some of the paste is not released during the printing process and adheres to the emulsion layer and wire, reducing the paste transfer. As a high volume of paste is desired, a thick emulsion layer is preferred. However, with increasing emulsion layer thickness, the release is reduced. From the theoretical paste volume, the transferred silver amount and thus the expected finger conductivity can be estimated. The amount of silver in a paste is typically given in weight percent and not in volume percent. The volume percent of silver vag is a screen-printing paste can be calculated by: v paste,ag mag mag m gl m = o ⋅ o + o + sol ρ ag ρ *ag ρ gl ρ o s −1 (2.4) with mag, mgl, msol being the mass percent of silver, glass frit and solvent in a paste, respectively, ρ°ag, ρ°gl, ρ°sol are the appropriate densities10. Assuming a density ρ°gl = 5 g/cm3 [43], of ρ°sol = 1 g/cm3 and of ρ°ag = 10.5 g/cm3 and a mass percent of mag = 84%, mgl = 3% and msol = 13% results in a silver volume of vag = 37% in the paste. The minimum achievable line resistance Rline (resistance per finger length) can then be estimated by: Rline ≈ ρ ag wsc ⋅ wst ⋅ v ag ⋅ a 0 [Ω/m] (2.5) with ρag being the specific resistivity of silver (1.6·10-8 Ω m) and wf the finger width in the screen. Assuming a finger width of wf = 100 µm, a wire thickness c of 25 µm, a wire separation distance d of 65.7 µm (280 mesh), an emulsion layer thickness wst of 50 µm and a silver content in the paste vag of 33% leads to a finger resistance of Rline = 18 Ω/m. This calculated value is slightly lower than 9 The acuteness of a screen with a steal wires is usually specified in mesh (mesh = inch-1) 10 ρ° is used as symbol for density as ρ is used throughout this work for conductivity 36 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ experimentally determined values being in the range of 30 Ω/m for the appropriate screen design. Possible explanations are the incomplete release of paste out of the screen and the difference of the specific conductivity of the sintered silver structure to the one of bulk silver. Snap-off distance: The snap-off distance is defined as the distance between the bottom of the printing form and the solar cell surface. The snap-off determines the upward movement of the screen, after the squeegee has passed. For too small snapoff distances the paste will not detach from the screen. If the distance is too large, the tension in the screen is decreased, which reduces its lifetime. Printing speed: The printing speed is defined as the speed of the squeegee moving over the screen during the print through process. The choice of printing speed depends on the past rheology, the screen thickness, the screen itself, the positioning of the squeegee and the printing image. Due to the increased applied force with elevated printing speed, the viscosity of the paste is reduced, simplifying the print through process. For industrial application a fast printing speed is required, leading to high throughput rates. The printing speed in industrial environment is in the range of 100 cm/s to 200 cm/s. Print squeegee and print squeegee pressure: The squeegee, as central paste transfer tool, fulfils a number of functions [42]. Most importantly the squeegee brings the screen and the solar cell surface into intimate contact, which allows paste transfer. Dependent on the hardness of the squeegee material and its geometry an adaptation of the substrate surface occurs. Additionally the squeegee removes excessive paste from the print-image area. The intimate contact between screen and substrate surface is achieved by applying an appropriate vertical force to the squeegee. The pressure should be slightly higher than necessary for the kissprint-point11, allowing tolerances in the process. An overestimated pressure results in a reduced printing quality of fine-line printed contacts and due to the increased abrasion, the lifetime of the screenprinting form and the squeegee is shortened. 2.2.3 Stencil-printing Stencil printing is an established metallization technology for printed circuit boards. The same base equipment and paste as for screen-printing applications can be used. The main advantage of stencil printing over screen-printing is the ability Cell processing and metallization technologies 37 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ to print finer lines with a higher aspect (height to width) ratio and the non-wear character of the stencil. This leads to a higher efficiency output and smaller standard deviation of processed cells on a large scale, as reported by Hoornstra et al. [44,45]. The printing image is typically etched into a metal foil or nickelelectroformed12. Another way of stencil fabrication is by pulse laser processing and chemical etching. Line definitions down to 30 µm were achieved [46], which makes this technology of high interest for solar cell processing. 2.2.4 Pad-printing Transfer pad-printing for the front side metallization of silicon solar cells has been extensively investigated at Fraunhofer ISE since 1998 [38,47,48]. The economical and technological potential of this technology was demonstrated [49]. The pad-printing process is illustrated in Fig. 2.7. A photopolymer plate with the print image etched into the surface is mounted on the machine by means of a magnetic plate holder. A blade spreads ink from the ink reservoir over the plate filling the etched areas. In a next step the doctor blade moves down and retracts the ink from the surface, leaving ink just in the etched areas. The solvents in the paste start to evaporate and the surface becomes tacky. Then the pad moves downwards onto the plate and rolls across its surface. The roundish surface of the pad is important, as this prevents air to be trapped between pad and plate surface. When the pad moves upwards, some of the ink stacks to its surface. At this point some of the solvents evaporate from the surface of the paste which adheres to the pad making this surface tacky. The plate is moved away and the pad moves downwards onto the substrate. The ink sticks to the surface of the wafer and is released from the pad as the pad lifts upwards. The paste composition for the pad-printing process is relatively similar to the screen-printing one. However, the viscosity of the paste needs to be slightly lower and a resin is added to the paste, which makes the surface of the ink tacky. The plates used within this work are photopolymer plates from BASF (Nyloprint 11 Kiss-print point: The minimum squeegee pressure necessary to obtain complete transfer of the printing image. 12 Electroformed stencils are created by nickel plating onto the surface of a conductive mandrel. The mandrel is covered by a structured photoresist carrying the negative of the printing image. When the required thickness of the nickel layer is reached, the resulting foil is separated from the mandrel after photoresist removal. These stencils are available mesh mounted onto industry-standard frames or as foils. 38 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ squeegee doctor blade solvents ink/paste plate paste reservoir pad substrate Fig. 2.7: Transfer pad-printing process sequence. ST52K). These plates can be easily structured in-house with the desired printing image and the resolution is high, but in contrast to steel plates the lifetime is shorter and the blade may easily damage the surface. The structure of a photopolymer plate is shown in Fig. 2.8. The base layer is steel, to which the photopolymer plate is stuck. The top surface is covered by a film to protect it against damage. Structuring of the plate with the desired print image can be subdivided into five processing steps (see Fig. 2.9): Pre exposure, main exposure, washing out, post exposure and drying. In the pre exposure the photosensitive material is hardened up to a certain level. The photopolymer starts hardening from the bottom side Fig. 2.8: Cross section of photopolymer plate [50]. Cell processing and metallization technologies 39 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ upwards. By this, the maximal possible etching depth can be defined. In the main exposure the actual print image is defined. A photopositive is laid on the resist and illuminated with UV-light. The resist which is open to the UV-light is hardened while the areas covered by the black emulsion of the photopositive remain soft. An intimate contact of the photopositive and the resist is important as well as an accurate defined exposure time in order to prevent underexposure. The soft areas can be washed out with an alcohol water based fluid. In the following post exposure the photopolymer is fully hardened. Finally the plate is dried in an oven to evaporate the washout fluid absorbed by the polymer. UV-light soluble resist hardened resist pre-exposure mask main exposure brush washing out post exposure Fig. 2.9: Process sequence for structuring a photopolymer plate. For solar cell mass production a rotary pad-printing machine has advantages over a transfer pad-printing machine. On the one hand the pressure during the paste transfer on the surface of the wafer is reduced; on the other hand the throughput rate of a rotary pad-printer is higher due to the different process sequences applied [38]. Roller printing as presented by Huster et al. [51] is similar to rotary padprinting, except that the complete pad is covered by paste. This technique can be applied to metallize the full rear side or a specially structured front. 2.2.5 Ink-jet printing Ink-jet printing is a direct write, non-contact deposition technology with high line resolution. Compared to photolithographical definition and evaporation, this technology has the advantage of a high material efficiency and a reduced process complexity. The main benefit compared to screen-printing is the higher line 40 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ resolution (line width < 20 µm) of at least a factor of two compared to the minimum line width achievable with screen-printing technology [52] as well as the non-contact deposition of the material, which is especially useful for uneven or thin wafers. In 2003 the development of an “ultra-fine ink-jet system” has been reported by Murata achieving dots of less than one micron [53]. Ink-jet printers can be divided into two main families; drop-on-demand (DOD) and continuous-jet system (Fig. 2.10). In the continuous-jet mode a stream of ink is ejected under high pressure through the nozzle. This stream disintegrates into a train of droplets, which are directed by electrical signals. Droplets not used, are diverted by the electrical field, collected by a gutter and reused. In the drop-ondemand mode for example a piezoelectric transducer generates droplets on demand. These droplets are ejected through the nozzle tip and directed onto the surface [54]. Fig. 2.10: Schematic drawing of a drop on demand system (push-mode piezoelectric, left-hand) and a continuous stream ink-jet system (binary deflection, right-hand) [55]. First experimental results applying the ink-jet technology for solar cell metallization was published in 1988 by Teng and Vest [56]. However, the viscosity of the ink needs to be much lower to be printable with an ink-jet printer compared to a screen-printer. This means that the silver content in the ink is low, whereas the amount of solvents is high. Teng and Vest used a metal organic ink with a silver load of 21% by mass achieving a 2 µm thick layer as deposited, which was thermally decomposed during the firing process to 0.3-0.5 µm. In order to achieve sufficient line conductivity of the final contact, multiple prints are necessary. Applying the ink-jet technology for solar cell metallization is still under current investigation by the National Renewable Energy Laboratory in the USA with promising results [57-59]. Nevertheless, achieving sufficient line conductivity for solar cell application seems to be a strong technological barrier. The ink-jet printing technology in combination with a plating step could be an alternative to Cell processing and metallization technologies 41 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ achieve a high-efficiency front side structure. Having this in mind another technology, the aerosolization of ink or paste, has been used for metallization within this work. The functionality of the metal aerosol jet system is comparable to an ink-jet system. However, the working principle differs strongly. The ink is atomized and directed as an aerosol jet onto the substrate surface using a specially designed printing head [60,61]. The width of the deposited line can be far smaller than the width of the nozzle outlet implying that the danger of screen clogging is minimized. This new technology has been adapted and optimized within this work for the front side metallization of silicon solar cells and is described in detail in Chapter 8. 2.2.6 Dispensing Using dispensing technologies the material is deposited in a non-contact mode making it suitable for rough surfaces. The printing material is extruded through a tip, whereas the tip should always have the same distance to the substrate surface. For uneven surfaces the use of a good z-axis control is recommended. The maximum size of particles in the ink or paste should be less than seven times the size of the outlet diameter of the tip. Additionally the paste should not separate under higher pressure, under which the system operates. For automatic dispensing of continuous lines the time pressure dispenser and the screw dispenser are most common. The time-pressure dispenser is the oldest and simplest dispensing system. A cartridge, filled with paste, is set continuously or for a certain time under high pneumatic pressure, extruding the paste through the nozzle. For depositions of good repeatability an auger valve is needed, which consists of an Archimedes screw, powered by an electrical motor. Paste is fed to the screw from a syringe under a constant low pressure. Fig. 2.11: Schematic drawing of the dispensing process. 42 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2.2.7 Photolithographical definition and evaporation With the photolithographical and evaporation process, well defined contacts of very fine dimensions can be produced. This process is commonly used for the front side metallization of high-efficiency solar cells, consisting of the following process sequence. First a photoresist is deposited on the front using the spin-on technology and dried to drive out excess solvents. A mask with the desired front side structure is placed on top of the resist and illuminated. The illuminated areas of the photopositive resist become soluble and can be washed out using a developer solution. Typically the wafer is then “hard-baked” to solidify the resist to make it more stable for the following wet chemical etch process of the SiNX and/or SiO layer. In the subsequent high vacuum evaporation process the metal or a stack layer of metals is deposited on the front. A common stack system is titaniumpalladium-silver. Titanium forms a relatively low contact resistance of less than 10-5 Ω cm² to an n-doped emitter with a surface doping concentration above 1·1019 cm-3. In addition titanium reduces the native silicon oxide and forms a contact of good adhesion. Palladium is used as diffusion barrier and adhesion promoter between titanium and silver. The top silver layer is used as conductor to transport the current as loss-free as possible to the busbars. In a last step the rest of the photoresist covered by the excessive metal is removed in the lift off process. Solvents like acetone penetrate into the very thin metal layer at the edge of the resist openings. As a consequence a reaction with the resist occurs and the resist and the excessive metal can be lifted off. The thin metal stack system is typically thickened by a silver plating step and sintered. UV-light photoresist antireflection coating n-silicon 1. spinning-on photoresist cured photoresist mask 2. drying off photoresist 3. masking and illuminating solvent evaporated metal 4. illuminated area washed out 6.evaporating of metal 5. antireflection coating etched contact 7. removal of resist and excessive metal Fig. 2.12: Simplified sketch of the photolithography and evaporation process sequence applied for the front side metallization process of high-efficiency cells. Cell processing and metallization technologies 43 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2.2.8 Laser micro-sintering Laser micro-sintering is a rapid prototype technology. It is used to create three dimensional objects by a layer-by-layer process. A laser is applied to fuse a thin layer of powder material into a solid object. After each layer is completed, a thin layer of the powdered material is spread across the top for the fusing of the next layer. With this technology very fine three dimensional objects can be created. At Fraunhofer ISE laser micro-sintering has been used the first time for solar cell front side metallization in cooperation with the Laserinstitut Mittweida [62]. A thin layer of metal powder is raked over the cell surface [63]. A laser beam of appropriate energy scans over the thin powder bed, being sintered and bonded to the emitter surface. The powder in the focus of the laser is melted/ sintered, but by virtue of the short laser pulses the heat is not transferred to adjacent powder. The excessive powder can be removed from the wafer. To achieve sufficient line conductivity the micro-sintered layer is thickened by a plating process. The main challenge for solar cell metallization is the investigation of laser processing parameters applying enough energy for the powder fusing and bonding to the cell surface without damaging the thin, preferably lowly doped emitter. First solar cell results have been reported by Alemán et al. achieving efficiencies up to 14.5% using tungsten as contact material [62]. Plated Finger laser pulse fused powder metal powder SiN Laser-sintered Seed Layer silicon Fig. 2.13: Schematic drawing of the laser micro-sintering process (left-hand) and SEM picture of a laser sintered and plated contact after partly removing the plated silver [62]. 2.2.9 Nickel plating Plating is a metallization technique describing the deposition of metal ions out of a chemical solution onto a surface. One distinguishes between electroplating and electroless plating. For electroplating an anode is immersed into the solution, an external potential is applied and a current flows between anode and cathode. At the cathode metal is reduced, at the anode metal is oxidized. For solar cell application a thin film of metal needs to be deposited on the surface prior to plating. In an 44 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ electroless plating solution first developed by Brenner and Riddel [64-66] no external potential needs to be applied, but the surface has to be autocatalytic for the plating reaction. An extensive investigation about creating ohmic contacts to nand p-doped silicon from electroless nickel plating solutions has been published in 1957 by Sullivan and Eigler [67]. About 20 years later plating became popular for silicon solar cells. Up to date numerous papers have been published about plating for solar cell contacts. One of the earliest was probably published 1980 by Saha [68] and by Anderson [69], investigating nickel plating on n-type silicon. Since then, many research groups around the world have used plating for silicon solar cells. High efficiencies up to 21.4% for a Ni/Cu plated solar cell were reported by Kim et al. in 2005 [70]. The probably best-known example for an industrial processed solar cell with nickel plated contacts is the laser grooved buried contact cell (LGBC, see Fig. 2.14) patented by Green and Wenham in 1985 [71]. Electroless nickel and copper plating are used to fill the heavily doped grooves to form a contact with a low contact resistivity and high line conductivity. Also electrochemical deposition of copper for LGBC cells was under research [72]. Fig. 2.14: Schematic cross section of a laser grooved buried contact solar cell [72]. A further approach for nickel plating, patented by van der Putten in 1991 [73], was presented by Horzel et al. [21]. By adding additives (stabilizers) to the plating solution, they achieved triangular shaped contacts, while conventional solutions tend to plate isotropically and thus dielectric layers neighboring the contact opening are overplated. Direct reflection of irradiation from the sidewalls of these triangular contacts into the silicon bulk is high. At Fraunhofer ISE nickel plating is also under investigation [74]. Contrary to the LGBC process the simplification of the process sequence is investigated by removing a line of the front surface SiNx layer with a laser process (similar to the Cell processing and metallization technologies 45 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ approach presented by Dubé et al. [75]) and to work on non-grooved medium doped emitters. In first experiments small-area FZ-silicon solar cells were processed with photolithographically defined contact openings. First solar cells processed achieved efficiencies up to 18.9% [74]. 2.2.10 Thickening of metal contacts by means of plating Instead of plating the contact layer directly, an alternative constitutes in plating the conducting layer on top of a pre-deposited metal layer. This metal layer, which should have a low contact resistivity to the semiconductor surface, could be for example an evaporated contact. This deposition combination (evaporation + plating) is widely used in the high-efficiency process (compare section 2.2.7). The earliest publication found is a patent from 1980 by Corwin et al. [76]. A completely plated contact has been presented by Münzer et. al. [77]. Low cost fabrication processes allowing high volume production were used for the so called high-efficiency plated contact (HEPC) solar cell [78]. That the approach of thickening contacts by plating can also be performed on thick-film contacts was published by Lyman in 1980 [79]. He reported that Photowatt Inc., USA, developed a special nickel-based ink together with a thickfilm paste manufacturer. This ink was screen-printed onto the silicon surface, fired in room ambient and finally thickened by copper electroplating. A similar approach was patented 1992 by Holdermann et. al. [80]. The conductivity of a silver thick-film contact is increased by electroplating silver. The challenge is that the fine electrodes on the front side need to be electrically contacted, which is difficult in industrial mass production. Another solution of electro-plating the complete contact is presented in Chapter 7.4. 2.3 Two-layer concept In the author’s opinion there is no thick-film metallization technology which is able to improve the front side structure significantly compared to a screen-printed one. Even though the efficiency potential of e.g. stencil-printing might be higher compared to screen-printing, replacing the screen-printing technology will be difficult. Screen-printing is a relatively inexpensive technology with high throughput rates. In addition the screen-printing technology has an experience advance in the field of solar cell mass production of more than 15 years. However, with shrinking contact width also the line conductivity is reduced. Even if the same aspect ratio as for wide lines can be achieved, halving the line 46 Cell processing and metallization technologies _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ width results in a conductivity decrease by about a factor of four. Another disadvantage for today’s thick-film contacts is the fact that the underlying emitter needs to be highly doped to achieve an acceptable contact resistance. To overcome these limitations a two-layer contact structure is thought to be the “magic bullet” for front side metallization. This has the great advantage that both layers can be optimized individually. The first layer should form a good electrical and mechanical contact to the preferable lowly doped emitter, be stable in oxidizing ambient and have a good device characteristic and lifetime. In addition it should act as a seed layer for the plating process, used to form the second layer. This second layer should have a good line conductivity to minimize resistivity losses in the finger, which is the reason, why silver (conductivity silver: 63·106 S/m) or copper (conductivity copper: 59.6·106 S/m) are preferred materials. If the cheaper material copper is used, the first layer in addition has to act as a diffusion barrier to prevent copper from diffusion into the silicon bulk and the surface of the copper has to be covered by a thin layer of e.g. tin or silver to protect it against detrimental reaction with the module material. Applying this two-layer contact structure changes the demands for the contact layer and hence also for the deposition technology. The probably most important demand for the applied deposition technology is the ability to create contacts of fine widths. The contact height (and finger conductivity), the limiting factor for many deposition technologies formerly, is of secondary importance. Furthermore the technology should be direct-write, non-contact, stress-free and should have a high throughput rate. However, all metallization technologies presented in this section could be suitable to create the first layer. Transfer pad-printing, high vacuum evaporation, fine-line screen-printing and metal aerosol jet printing will be investigated more in detail in this work. In the following section ohmic and shading loss calculation for the two layer concept will be presented, emphasizing the increased efficiency potential of this two-layer process compared to the standard screen-printing process. Grid design of the two-layer contact structure 47 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 3 Grid design of the two-layer contact structure In this chapter an analysis of the optical and electrical losses of a two-layer contact structure is presented and compared to a conventional screen-printed contact. For different cell designs the optimum finger height and finger separation distance were determined. 3.1 Calculation parameters and assumptions Plated Contact In a calculation based on eq. 1.19-1.23 the dependence of ohmic and shading losses on the contact width was investigated. The two layer contact structure consists of a thin deposited contact layer (thickness <1 µm), which is isotropically thickened by light-induced silver plating. The final contact can be divided into two zones differing in its geometry. The “middle zone” of the finger has a flat surface, whereas the “edge zone” is roundish (Fig. 3.1 left-hand). edge zone middle zone edge zone plated silver hf hf contact layer wc antireflection coating 120 µm hf screen printed contact hf 15 µm wf silicon silicon Fig. 3.1: Assumed contact geometry for a plated (left-hand) and a screen-printed (right-hand) contact finger. The finger width wf of the plated contact is equal to the sum of twice the plating height hf and the contact width wc. As cross-section area, the area encapsulated by the dashed line is assumed. The assumed contact finger would have a width wf of two times the plating height hf plus the width of the contact layer wc: ( ) w f h f = 2h f + wc (3.1) The cross section area Af of the plated finger can be calculated by: ( ) A f h f = wc ⋅ h f + 1 ⋅π ⋅ hf 2 ( )2 (3.2) Applying equation (1.11) and assuming transparency factors of tb = 0 and tf = 0 (no reflection from contacts into the active cell area), the shading loss fraction ps can be expressed in dependence of the finger height hf and separation distance s by: 48 Grid design of the two-layer contact structure _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ ( ) p s s, h f = ( ) = s ⋅ wbuc + l f ⋅ (2h f s ⋅ wbuc + l f ⋅ w f h f s⋅a + wc s⋅a ) (3.3) Shading losses will decrease with increased finger separation distance as well as with reduced finger height. However, a larger finger separation distance will increase losses due to the emitter resistance, and a reduced finger height, losses due to a decreased finger conductivity. The area-weighted resistance r is the sum of finger resistance rf, contact resistance rc and emitter resistance re. The resistance of the busbar is not taken into account as it is independent of finger width and finger separation distance optimization (compare section 3.4). ( ) ( ) ( ) ( r s, h f = re s, h f + rc s, h f + r f s, h f ) (3.4) The resistance will be low, if the contact is high and the finger separation distance is small. However, this will lead to high shading losses. The electrical power loss fraction can be expressed using eq. (1.18) and eq. (1.15) by: ( ) p e s, h f = j mpp ( ) Vmpp + r s, h f j mpp (1 − p s (s, h f ))⋅ r (s, h f ) (3.5) The total loss fraction p is the sum of the electrical pe and optical ps loss fractions. ( ) ( p s, h f = p s (s, hf ) + p e s, h f ) (3.6) An optimum between optical and electrical losses needs to be found in order to minimize the total loss fraction. In the following for a given contact layer wc the finger separation distance s as well as the finger height hf have been determined to achieve minimum total loss. For the metal-semiconductor interface a contact resistivity ρc of 0.5 mΩ cm² has been assumed. This is a quite low value for a conventional screen-printed contact. However, in the author’s opinion, this value will be achieved for a contact solely optimized to give a low contact resistivity (regardless of the finger conductivity) to an emitter sheet resistance of Rsh = 55 Ω/sq, having a surface doping concentration of about 2·1020 cm-3. Within this work contact resistivity down to ρc = 0.7 mΩ cm² were measured using conventional screen-printing paste. Values of ρc = 0.6 mΩ cm² have been reported by Hilali et al. [37] contacting an emitter of Rsh = 100 Ω/sq, also using commercially available screen-printing paste. The influence of the contact resistivity on the grid design for the two layer contact structure is discussed more in detail in Chapter 3.2.3. Grid design of the two-layer contact structure 49 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Screen-printed contact: For the screen-printed contact a width wf_sp of 120 µm and a contact height (height averaged across the finger width) of 15 µm has been considered (see Fig. 3.1 right-hand). A resistivity ρf of 3.2·10-8 Ω m for the finger bulk has been considered and the same low contact resistivity (ρc = 0.5 mΩ cm²) as for the plated contact. The parameters used for calculation are presented in Table 3.1. Table 3.1: Parameters used for loss calculation. Assumed parameters: symbol definition used value a Acell lf hf_sp wbuc wf_sp wc length unit cell for 2 busbars; 3 busbars cell area finger length 2 busbars, 3 busbars average height of screen-printed contact width busbar unit cell: 2, 3 busbars width screen-printed contact width contact layer 39 mm; 26 mm 15.6 x 15.6 cm² 38 mm; 25.3 mm 15 µm 1 mm, 0.75 mm 120 µm 1 µm – 80 µm ρc ρAg_pl ρAg_sp contact resistivity 0.5 mΩ cm² conductivity plated silver 1.9·10-8 Ωm conductivity screen-printed silver tf ; tb transparency factor of finger and busbar 3.2·10-8 Ωm 0% Rsh sheet resistance of emitter pe,bus; pe,tab elec. power loss fraction busbar, tab 55, 90, 120 Ω/sq 0% jmpp current density at mpp 34 mA/cm² Vmpp voltage at mpp 520 mV Calculated values: symbol definition s hf wf re , rc , rf pe,e pe,c, pe,f ps, p finger separation distance plating height ≈ finger height total contact width area weighted resistance emitter, contact, finger electrical power loss fraction emitter, contact, finger optical power loss fraction, total power loss fraction unit m m m Ω m² 50 Grid design of the two-layer contact structure _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 3.2 Solar cell with two busbars 3.2.1 Assuming a cell with a sheet resistance of 55 Ω/sq 20 4 15 3 10 2 finger height [µm] finger separation distance [mm] 5 0 1 s [mm] hf [µm] Loss calculations have been performed for a solar cell with a sheet resistance of Rsh = 55 Ω/sq and an area of Acell = 15.6 x 15.6 cm². The current from the fingers is collected by two busbars. As the busbar and tab resistance do not influence the optimization of the finger height and finger separation distance, this loss fraction has not been taken into account. An electrical loss calculation concerning busbar and tab resistance can be found in Chapter 3.4. However, the optical losses of ps,bus = 2.56% due to the 2 mm broad busbars were considered, as this is necessary for later calculations in which different busbar geometries are assumed. Other parameters used for the calculation are listed in Table 3.1. In Fig. 3.2 (bottom figure) the minimal total loss fraction and the different loss contributors are plotted versus the contact width. In addition the correspondent finger height and finger separation distance are plotted versus the contact width (Fig. 3.2 top figure). 0 10 Loss [%rel] total loss 8 6 ps finger 4 ps busbar pe emitter pe finger pe contact 2 0 0 20 40 60 Contact width wc [µm] 80 SP cell Fig. 3.2: Calculation of electrical (pe) and shading (ps) loss fractions of a plated solar cell in dependence of the contact width. The plating height hf and finger separation distance s versus contact width wc for minimum total loss is illustrated in the top figure. For the SP cell a width wf of 120 µm and a height hf of 15 µm were considered. The calculation is based on a solar cell of size 15.6 x 15.6 cm² with two busbars, a sheet resistance of Rsh = 55 Ω/sq and a contact resistivity of 0.5 mΩ cm². Further parameters are listed in Table 3.1. For the assumed parameters the seed layer width resulting in minimum total loss of pmin = 7.8% is achieved at wc = 12 µm. With increasing contact width, the total loss rises by about ∆p = 0.3% per 10 µm. The finger separation distance s needs to Grid design of the two-layer contact structure 51 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ be increased to compensate the higher shaded area which in addition increases losses due to the emitter sheet resistance. If the contact width is reduced below the minimum total loss value, the contact resistance rises strongly, limiting cell performance. In this case for optimum cell performance the amount of fingers needs to be increased to obtain a larger contact area. This on the other hand increases the shaded area again. Compared to a screen-printed contact of width wf_sp = 120 µm (SP cell) and averaged height of hf_sp = 15 µm the total loss is reduced by ∆p = 2.4%. 3.2.2 Variation of the emitter sheet resistance Total loss p [%rel] 20 15 10 5 0 11 10 4 2 Rsh = 120 Ω/sq s [mm] hf [µm] In a further calculation the emitter sheet resistance was varied (Fig. 3.3). For the same contact width the total grid loss for cells with lower doping concentration is higher. This is mainly due to the increased sheet resistance as well as the enlarged shaded area as the finger separation distance needs to be reduced to compensate emitter losses. Of course on the other hand a reduced emitter doping will result in lower recombination losses. 0 Rsh = 90 Ω/sq Rsh = 55 Ω/sq 9 8 7 0 20 40 60 80 Contact width wc [µm] SP cell Fig. 3.3: Total loss p, finger height hf and finger separation distance s for different sheet resistance emitters plotted versus the contact width wc. In addition the total loss p for a screenprinted contact is illustrated. As in the previous calculation, the advantage of fine-line printing becomes obvious. The finger separation distance is reduced, which leads to lower emitter losses (see Table 3.2). For example comparing the loss p of a solar cell with 52 Grid design of the two-layer contact structure _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Rsh = 55Ω/sq to a cell with Rsh = 120 Ω/sq the difference is ∆p = 0.5% for the optimum seed layer width. This rises to ∆p = 1.1% for a seed layer width of wc = 50 µm. For a screen-printed contact the loss difference is even more significant. The loss difference ∆p is 1.5% comparing the loss of a screen-printed cell featuring a sheet resistance of Rsh = 50 Ω/sq to Rsh = 120 Ω/sq. Thus, in order to achieve higher efficiencies due to the better internal quantum efficiency in the short wavelength region of less doped emitters, the higher electrical and optical losses first need to be overcompensated. Table 3.2: Calculation results for a varying emitter sheet resistance and contact widths. The optimum finger width, finger height, finger separation distance as well as different loss contributions are listed. sheet contact finger finger finger resistance width height distance width [µm] [µm] [mm] [µm] [Ω/sq] emitter electrical optical loss loss loss [%] [%] [%] plated-contact for minimum total loss 55 12 17 1.5 46 90 10 15 1.2 40 120 9 14 1.1 37 plated-contact with a seed layer width wc=50 µm 55 15 2.1 81 90 50 13 1.8 76 120 12 1.6 74 screen-printed contact 55 2.5 90 120 15 2.2 120 120 2 total loss [%] 0.6 0.7 0.7 2.3 2.4 2.5 5.5 5.7 5.8 7.8 8.1 8.3 1.2 1.4 1.6 2.4 2.6 2.8 6.2 6.7 6.9 8.6 9.3 9.7 1.6 2.0 2.2 3 3.2 3.4 7.3 8 8.4 10.3 11.2 11.8 3.2.3 Influence of the contact resistivity The influence of the contact resistivity on the total power loss fraction in dependence on the seed layer width is illustrated in Fig. 3.4. The left-hand figure shows the loss calculation for a solar cell with a sheet resistance of Rsh = 55 Ω/sq and the right-hand figure for a solar cell with Rsh = 120 Ω/sq. The other parameters used were the same as for the previous calculation (see Table 3.1). Grid design of the two-layer contact structure 53 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 12 12 2 ρc> 10 mΩ cm 2 Total loss p [%rel] 11 Total loss p [%rel] ρc> 10 mΩ cm 10 10 minimum 5 9 3 8 0.1 7 6 1 1 0.5 2 ρc= 0.01 mΩ cm 10 20 Rsh = 55 Ω/sq 30 40 Contact width wc [µm] 50 10 11 minimum 10 5 3 9 8 0.5 1 2 2 7 6 1 ρc= 0.01 mΩ cm ρc= 0.1 mΩ cm Rsh = 120 Ω/sq 10 20 30 40 50 Contact width wc [µm] Fig. 3.4: Total loss p of a solar cell with an emitter sheet resistance of Rsh = 55 Ω/sq (left-hand) and Rsh = 120 Ω/sq (right-hand) in dependence of the contact width plotted for different contact resistivity. From the left-hand figure it can be concluded that a contact with a contact resistivity below ρc = 0.01 mΩ cm² has no significant influence on the total loss and a contact resistivity smaller than ρc = 0.1 mΩ cm² has just an affect for contacts with a width smaller wc = 8 µm. The influence of a contact resistivity ρc > 0.5 mΩ cm² is at least ∆p = 0.1%, independent of the contact width. In addition with increasing contact resistance the minimum of the total loss is found at broader contact widths. E.g. a solar cell with a contact featuring a contact resistivity of ρc = 0.5 mΩ cm² and of ρc = 3 mΩ cm² has its minimum total loss at a contact width of wc = 12 µm and wc = 27 µm, respectively. Comparing the calculations for a solar cell with a sheet resistance of Rsh = 55 Ω/sq to the one with Rsh = 120 Ω/sq the curve progression is relatively similar. The loss p for the less doped emitter is on a higher level, mainly due to the increased emitter sheet resistance as explained in the previous chapter. Optimum solar cell performance (minimum loss) for the less doped emitter is achieved at a lower contact width wc. For these cells the number of fingers for optimum cell performance (minimum total loss p) is higher, increasing the total contact area. The following conclusions can be drawn from the previous calculations. The contact resistivity ρc should be as low as possible. If the contact resistivity ρc is not sufficiently low, fine-line printed contacts can even increase the total loss fraction p. Nevertheless, even for standard thick-film pastes assuming a contact resistance of ρc = 3 mΩ cm² the contact width can be reduced down to wc = 27 µm before total loss rises again. 54 Grid design of the two-layer contact structure _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 3.3: Contact width for minimum total loss presented under variation of the contact resistivity. contact contact finger finger finger resistivity width height distance width ρc [mΩ cm²] [µm] [µm] [mm] [µm] losses [%] 0.01 0.1 0.5 1 3 10 2 6 12 17 27 43 20 19 17 15 14 11 1.5 1.5 1.5 1.5 1.5 1.5 42 44 46 49 55 66 0.05 0.2 0.4 0.6 1.0 2.1 electrical optical losses losses total loss [%] [%] [%] 2.0 2.1 2.3 2.5 2.9 3.7 5.2 5.3 5.5 5.7 6.1 6.9 7.2 7.4 7.8 8.1 8.9 10.6 3.2.4 Considering irradiation reflection from the contact In the previous calculations it has been assumed that the sunlight hitting the contact is not used for current generation (tf = 0, tb = 0). However, as discussed in Chapter 1.2.1 some of this irradiation will be reflected onto the active area of the cell. It is most probably that reflection from the roundish part of the contact is higher, compared to the flat part (compare Fig. 3.1). Thus two different transparency factors are introduced for the contact finger. The transparency factor tf,mz is the amount of reflected irradiation from the middle zone of the finger into the cell and tf,ez the one of the edge zone. Comparable to eq. (1.11) the shading fraction of a plated finger ps,pf can then be written as: ( ) [(1 − t f , mz )⋅ wc + s2⋅⋅a(1 − t f ,ez )⋅ hf ]⋅ lF p s , pf s, h f = (3.7) The total shading fraction of a plated grid ps,p can be described by: ( ) [(1 − t f , mz )⋅ wc + 2 ⋅ (1 −st ⋅f a,ez )⋅ hf ]⋅ (lF + t B ⋅ wbuc ) p s , p s, h f = (3.8) The total loss fraction p of the front grid in dependence on the contact width wc under variation of the “edge zone” transparency factor is illustrated in Fig. 3.5. Obviously the total loss declines with increasing transparency factor. The effective shaded area is reduced, which leads to a decreased finger distance for optimum cell performance, which again reduces losses due to the emitter sheet resistance, contact resistance and finger resistance. Grid design of the two-layer contact structure 55 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Furthermore, for broad fingers the absolute loss reduction due to the transparency of the finger edge is less than for small contact widths. This is because with a reduction of the contact width also the fraction of the “middle zone” declines, implying that the effect of the “edge zone” increases. Total loss p [%rel] 10 9 minimum 0% 8 20% 40% 7 Transparency factor finger edge zone 60% 6 1 10 20 30 40 50 60 70 Contact width wc [µm] 80 Fig. 3.5: Total loss p of the front grid plotted versus the contact (seed layer) width wc assuming different transparency factors for the edge zone tf,ez of the contact finger. The transparency factor of the finger middle zone tf,mz was assumed 0%. For each contact width, minimum total loss was calculated. 3.3 Solar cell with alternative busbar geometries For an H-grid pattern with two busbars the finger length of unit cell I is 38 mm (compare Fig. 1.10). As the current increases linear and the power loss to the square with finger length, a high line conductance of the finger is necessary to minimize resistive losses. This is achieved by using a highly conductive material (silver) and a large cross section area of the finger. Due to the mentioned dependence of plating height and contact width, an optimum between the electrical loss of the finger and the loss due to finger shading need to be determined. For the 2 busbar grid the plated thickness of silver for minimum total loss is 18 µm for a 10 µm wide contact. The height drops slightly to 15 µm for a contact width exceeding 50 µm. This actually means that the final contact will be 30 µm to 36 µm wider than the width of the printed contact layer. Reducing the finger length of a unit cell by introducing e.g. a third busbar, the conductivity of the finger can be reduced, which will yield in less shading and electrical losses. 56 Grid design of the two-layer contact structure _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 3.3.1 Three busbars 20 4 15 3 10 2 finger height hf [µm] 5 1 finger separation distance s [mm] 0 s [mm] hf [µm] Front side grid with three busbars The length of the finger for a unit cell was reduced by adding an additional busbar to the front side grid. Instead of using two 2 mm wide busbars, three 1.5 mm busbars were used for this calculation, implying that the finger length of unit cell I decreases from 38 mm to 26 mm. A significant loss reduction is expected. The loss due to grid shading and series resistance is plotted versus the width of the contact layer in Fig. 3.3. 0 10 Loss [%rel] total loss 8 6 ps finger 4 ps busbar pe emitter pe finger pe contact 2 0 0 20 40 60 80 Contact width wc [µm] SP cell Fig. 3.6: Three busbars design: Calculation of electrical (pe) and optical (ps) loss fractions of a plated solar cell in dependence of the contact width. The loss fractions for a screen-printed cell (SP) are presented in the right part of the figure. In the top figure the finger height and finger separation distance for minimum total loss is illustrated. Despite the 0.3% increase in shaded area due to the three busbars compared to the two busbars, the total loss is reduced by about 0.6% absolute (compare Table 3.4). In addition the electrical loss contribution of the busbar, which has not been considered in the above calculation, is slightly reduced compared to a 2 busbar grid structure (compare Chapter 3.4). Furthermore the amount of plated silver required is reduced. The mass mAg_pl of the deposited silver can be calculated by: m Ag _ pl ≈ A f ⋅ l sc ⋅ wsc o o ⋅ ρ ag + N bus ⋅ Abus ⋅ wsc ⋅ ρ ag s (3.9) with lsc being the length of the solar cell (in finger direction), wsc the width of the solar cell (in busbar direction), Nbus the number of busbars, Abus the cross section area of the busbar (Abus = wbus·hf) and ρ°ag the density of silver. For minimum total Grid design of the two-layer contact structure 57 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ loss the deposited mass of silver is equal to mAg_pl = 220 mg for the grid with two busbars and mAg_pl = 152 mg for the grid with three busbars. This is a reduction of 31% silver by mass, which reduces the material cost, but also the cost correlated with the plating time. 3.3.2 Mesh of wires From the previous calculation can be reasoned that a reduction of the finger length of unit cell I is preferable, if at the same time shading and/or electrical losses due to the busbar is not increased. In the following calculation the benefit of fine-line printing in combination with a recently introduced module concept, the Day4TM Electrode concept [81,82] is presented. Instead of using two 2 mm wide busbars, a mesh of about 10 to 20 copper wires are used to collect the current from the fingers (see Fig. 3.7). This reduces the finger length of a unit cell to just 4 mm to 8 mm. The conductance of the finger in order to transport the current with low loss to the copper wires needs to be a fraction compared to the prior calculations. For this calculation 16 copper wires with a cross section of 200 µm were assumed, reducing the finger length of unit cell I to 4.8 mm. Calculation results are compared to the previous grid patterns assuming two and three busbars (see Fig. 3.8 and Table 3.4). Total loss of less than 5% can be achieved for optimum contact width. This is a significant reduction compared to 7.8% for the two busbars and 7.2% for the three busbar structure. As for the previous calculations the electrical losses in the busbars (copper lines) were not taken into account, but the losses due to shading, which are equal to ps,bus = 2.0%. The amount of plated silver necessary for this Fig. 3.7: Solar cell with Day4TM Electrode contact. Coated copper wires are embedded into a polymeric film/adhesive compound and contact cell fingers [81]. 58 Grid design of the two-layer contact structure Total loss p [%rel] 20 4 15 3 10 2 5 1 0 10 0 s [mm] hf [µm] _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 8 6 4 2 busbars 3 busbars mesh of 16 wires 2 0 0 20 40 60 Contact width wc [µm] 80 SP cell Fig. 3.8: Comparison of the total loss p caused by the front grid for a 2 busbars, three busbars and a mesh of wires design. Additionally, the plating height hf and finger separation distance s is illustrated. concept is about 10% to 20% compared to the calculation with 2 and 3 busbars. Also the plating time is drastically reduced, as just a plating thickness of 2 µm to 4 µm is necessary. Table 3.4: Calculation results of characteristic parameters listed for different front busbar geometries and contact widths. busbar design contact finger finger finger optical electrical optical width height distance width losses bus losses losses [µm] [µm] [mm] [µm] [%] [%] [%] plated-contact for minimum total loss 2 busbars 12 17 1.5 46 2.6 2.3 5.5 3 busbars 11 12 1.4 34 2.9 1.9 5.3 mesh 6 2 0.8 10 2.0 1.0 3.3 plated-contact with a seed layer width of 50 µm 2 busbars 15 2.1 81 2.6 2.4 6.2 3 busbars 50 10 2.1 70 2.9 2.1 6.2 mesh 1.9 2.0 54 2.0 1.5 4.6 screen-printed contact 2 busbars 2.5 2.6 3 7.3 3 busbars 120 15 2.6 120 2.9 2.5 7.4 mesh 2.8 2.0 2.2 6.2 total loss [%] 7.8 7.2 4.3 8.6 8.2 6.1 10.3 9.9 8.4 Grid design of the two-layer contact structure 59 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 3.4 Losses caused by busbar and tab resistance 3.4.1 Busbar losses Loss due to busbar resistance [%rel] In the previous calculations electrical losses due to the busbar and tab have not been taken into account. As the electrical loss of a busbar is independent on the contact/finger width and finger separation distance, it does not influence the previous optimizations. Another reason for not taken the busbar losses into account is the strong dependence on the busbar geometry and the interconnection to the tab. The ohmic power loss fraction versus the number of soldering joints for two 2 mm, two 1.5 mm and three 1.5 mm wide busbars with an average height of 15 µm is illustrated in Fig. 3.9. A high number of soldering joints between busbar and tab is necessary to keep busbar resistance losses low [83]. 1.0 2 busbars; 1.5 mm wide 2 busbars; 2 mm wide 3 busbars; 1.5 mm wide 0.8 0.6 0.4 0.2 0.0 2 4 6 8 10 12 Number of soldering joints Ns Fig. 3.9: Losses caused by the busbar resistance. For calculation the following parameters were assumed: Cell size 15.6 x 15.6 cm², jmpp = 34 mA/cm², Vmpp = 530 mV, ρbus = 3.2·10-8 Ω m, ρtab = 1.6·10-8 Ω m and hbus =15 µm. In addition to the electrical losses, shading losses need to be added, to achieve the total loss contribution of the busbar. This is about 2.56% for the 2 mm wide, 2.88% for the three 1.5 mm wide and 1.92% for the two 1.5 mm wide busbars. 3.4.2 Tab losses On top of the busbar a tab is soldered to collect the current out of the busbar and transport it to the rear side of the next solar cell. Especially for large-area silicon solar cells the photo-generated current is high. For a typically 15.6 cm x 15.6 cm solar cell with a current density at mpp of jmpp = 34 mA/cm² and 2 busbars the 60 Grid design of the two-layer contact structure _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ current at the end of one tab is about four ampere. Obviously the tab conductance needs to be high. Copper is used as a highly conductive and low-cost (compared to silver) material. The typically used tab in industrial environment featuring a thickness of 100 µm to 150 µm and a width of 2 mm leads to an additional power loss of 2% to 3% (Fig. 3.10). Using a tab as thick as possible is recommended. Otherwise the use of three tabs of width 1.5 mm is suggested, which in addition would be advantageous for fine-line printed and plated contacts. Fig. 3.10 also illustrates the electrical losses pe of 2.45% calculated for a mesh of 16 wires featuring a cross section of 200 µm and a resistivity of 1.7·10-8 Ω m. Losses due to tab resistance [%rel] 4 2 tabs 2 mm wide 2 tabs 1.5 mm wide 3 tabs 1.5 mm wide 3 16 round wires 200 µm cross-section width 2 1 0 typically used in industry 100 150 200 250 300 Tab thickness [µm] Fig. 3.10: Losses due to tab resistance plotted versus the tab height. For calculation the same parameters as for Fig. 3.9 were assumed. The green star illustrates the electrical loss calculated for a mesh of 16 copper wires with a cross section diameter of 200 µm and a resistivity of 1.7·10-8 Ω m. Losses were calculated up to the edge of the solar cell, additionally losses due to the tab extension need to be added. 3.5 Chapter summary Loss calculations for different front side grid designs have been presented based on solar cells with a realistic and easy-to-achieve contact geometry. The two-layer contact structure consists of a contact layer which is isotropically thickened by silver plating. The finger height and finger separation distance yielding minimum total loss has been determined. Compared to a conventional screen-printed solar cell a significant reduction of losses was demonstrated. Furthermore, while reducing the contact width, an increasing efficiency potential is achieved, up to the width at which the contact resistivity becomes the limiting factor. Grid design of the two-layer contact structure 61 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Instead of using a two busbar front grid design, a three busbar or the Day4TM Electrode contact structure would be beneficial, as the efficiency potential rises and the silver consumption in the plating step is reduced. A contact resistivity value below ρc = 1 mΩ cm² needs to be achieved, in order not to effect the solar cell performance significantly. Especially today’s thick-film pastes need to be further optimized to yield a low contact resistivity. The line conductivity is of minor importance as the plated silver is used as conducting layer. If a more advanced contact geometry is used than the presented one, the efficiency potential of fine-line printed contact layers will be even further increased. More advanced contact geometry is e.g. a contact with a higher aspect ratio. This can be achieved by using a thick photoresist for structuring the metallization pattern [84] or by adding additives into the plating bath that suppresses plating in horizontal direction [21]. 62 Series resistance determination from IV-curves _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 4 Series resistance determination from IV-curves Using the one- and two-diode solar cell model all resistive losses are aggregated in one lumped series resistance. This series resistance can be determined from IVcurve measurements. In the following, six different series resistance determination methods are presented. These methods are furthermore experimentally compared among each other and discussed. The purpose of this investigation was to find a measuring system that is simple, fast and reliable and can be applied in an industrial environment. 4.1 Series resistance determination methods 4.1.1 Fitting of the two-diode equation to a dark IV-curve This method is based on an analytical description of a solar cell by the twodiode model. The current density j in dark is equal to: q (V − j ⋅ rs ,dark ) q (V − j ⋅ rs ,dark ) V − j ⋅ rs ,dark − 1 + j02 exp − 1 + j = j01 exp n ⋅ k ⋅ T n ⋅ k ⋅ T rP 1 2 (4.1) Current density j [mA/cm²] with j01 and j02 being the recombination current densities of the emitter/ base and the space-charge region, respectively. rP is the parallel resistance, k the Boltzmann constant, T the temperature, and n1 and n2 the ideality factors. The two-diode equation is fitted to the IV-points of the dark IV-curve. In the upper voltage part the series resistance rs,dark can be extracted, since in this region the power loss caused by rs,dark is the most significant loss mechanism (Fig. 4.1). 10 2 10 1 10 0 10 -1 10 -2 10 -3 "fit region" to extract rs,dark 0 200 400 600 Voltage V [mV] Fig. 4.1: Dark IV-curve fitted by the two diode equation function. Series resistance determination from IV-curves 63 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ As stated by Aberle et al. [85] the series resistance measured in the dark is lower than the one measured under illuminated conditions due to different current flow pattern in both states. In dark IV-measurements an external current is applied to the electrodes. In this case the operating voltage is highest near the electrodes and decreases towards the central emitter region due to the emitter sheet resistance. Under irradiation the current flow is reversed, hence the operating voltage is highest in the central emitter region and decreases towards the electrodes. 4.1.2 Comparison of the dark with the one-sun IV-curve Using this method the series resistance is determined by comparing the one-sun IV-curve at the maximum power point with the dark IV-curve shifted by the shortcircuit current density jsc: rs ,light _ dark = (Vdark , jmpp − Vrs,dark ) − Vmpp (4.2) j mpp with ( ) Vrs ,dark = j sc − j mpp ⋅ rs ,dark (4.3) and rs ,dark = Vdark , jsc − Voc (4.4) j sc -40 -30 -20 jsc jmpp one-sun IV-curve shifted dark IV-curve -10 Vmpp 0 200 400 V dark,jmpp Voltage V [mV] Vdark,jsc Voc Fig. 4.2: One-sun and by jsc shifted dark IVmeasurement. Current density j [mA/cm²] Current density j [mA/cm²] For definition of Vdark,jmpp and Vdark,jsc see Fig. 4.2. Equation (4.2) is taken from Aberle et al. [85] including the correction that at Vdark,mpp of the dark IV-curve, the losses due to the series resistance rs,dark can not be neglected. At Vdark,jmpp the current-flow of the dark IV-curve is low, but has a value equal to jsc – jmpp, which is -40 -35 -30 -25 -20 -15 jsc jmpp one-sun IV curve Suns-Voc curve -10 Vmpp -5 0 200 400 Voltage V [mV] Voc 600 Vsuns,jmpp Fig. 4.3: One-sun IV- and Suns-Voc curve. 64 Series resistance determination from IV-curves _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ at least 5% of the short-circuit current density. This means that also the generated voltage drop due to the series resistance of the dark IV-curve is not neglectable. Thus, Dicker [86] has added the term Vrs,dark (eq. (4.3)) taking this voltage drop into account. The series resistance rs,dark of the dark IV-curve can be calculated by comparing the voltage difference between Voc and Vdark,jsc [87]. While the series resistance at Voc is not active, at Vdark,jsc the current flow is equal to jsc, causing an additional voltage drop due to the series resistance rs,dark. Thus rs,dark can be calculated applying equation (4.4). 4.1.3 Comparison of the Suns-Voc with the one-sun IV-curve The basic assumption for this series resistance determination method is similar to the method described before. A series-resistance-free measurement, the shifted jsc-Voc curve, is compared with the series-resistance-affected one-sun IV-curve at its maximum power point, as illustrated in Fig. 4.3. This method was proposed by Wolf in 1963 [88] and is described by the following equation: rs ,Suns VOC = ∆V j mpp = VSuns , jmpp − Vmpp j mpp (4.5) Vsuns,jmpp is equal to the voltage drop of the shifted jsc-Voc curve at the maximum power point current density of the one-sun IV-curve (Fig. 4.3). jsc is unaffected from rs as long as rs is less than 10 Ω cm², Voc is unaffected as there is no current flow [89]. Hence the jsc-Voc curve is not influenced by rs, but it is still affected by j02 and rp. A simple way of measuring the jsc-Voc curve has been described by Sinton [90,91] based on the principle of superposition. The open-circuit voltage values Voc of a solar cell with known short-circuit current density at one-sun irradiance are determined under variation of the irradiation intensity by a flash. The corresponding value of the current density is calculated via a calibrated reference cell. This quick method to determine the jsc-Voc curve of a cell was used in this series resistance determination method. 4.1.4 Comparison of at least two IV-curves measured at different irradiation intensities The comparison of two IV-curves at different irradiances was first proposed by Swanson in 1960 [92]. The jsc difference of these curves is due to the linear proportionality of the irradiance to the generated photo-current, while the Voc shift Series resistance determination from IV-curves 65 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ is due to the smaller voltage drop of the series resistance at a lower current density ∆V=rs ∆jsc [88,93], which is utilized in this method. As proposed in the European Standard EN 60891 [94] IV-curves at three irradiation intensities should be measured. On each IV-curve a point ji is chosen in the way that Vi(ji) is slightly above Vi,mpp and that ji =jsc,i-∆j, as illustrated in Fig. 4.4. The series resistance rs,,int_var can then be determined by calculating the mean value of rs1, rS2 and rs3 with: rs1 = V2 − V1 j1 − j 2 rs 2 = V3 − V1 j1 − j3 V3 − V2 j 2 − j3 rs 3 = (4.6) -35 -30 2 -40 V1,j1 ∆j ∆j V2,j2 linear fit -25 -20 -15 -10 -5 0 Current density j [mA/cm ] 2 Current density j [mA/cm ] Instead of calculating the mean value, it would also be possible to linearly fit the points (Vi,ji), the inverted value of the slope is equal to rs,int_var. As stated by Altermatt et al. [89] the assumption for the irradiance variation method is, that series resistance, recombination currents and ideality factors are similar at all points (Vi,ji). This assumption is only justified as long as the difference in short-circuit current density ∆jsc is small. rs is similar at different irradiances as long as the sheet resistance of the base is higher than the sheet resistance of the emitter and/or j·rs ≤ n·k·T/q, otherwise the series resistance declines from a maximum value at the maximum power point to its lowest value at open-circuit voltage [95]. To avoid the series resistance dependence at different irradiances, Altermatt et al. proposed to use the IV-curves at 0.9 suns and one-sun irradiance, instead of at 0.5 suns and one-sun irradiance as stated in an earlier article [23]. V3,j3 ∆j one-sun IV-curve 0.9 suns IV-curve 0.5 suns IV-curve 200 400 600 Voltage V [mV] Fig. 4.4: Three IV-curves of the same solar cell measured at different irradiation intensities. -40 -35 ∆j=jsc,shaded -30 VA -25 one-sun IV-curve shaded IV-curve -20 -15 -10 -5 Voc,shaded jsc,shaded ∆j 0 200 400 600 Voltage V [mV] Fig. 4.5: One-sun and 0.1 suns IV-curve. 66 Series resistance determination from IV-curves _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 4.1.5 Comparison of jsc and Voc values with the one-sun IV-curve This method proposed by Bowden and Rohatgi [96] is based on the irradiance variation method and the comparison of the shifted jsc-Voc curve with the one-sun IV-curve. The series resistance is determined by comparing the one-sun IV-curve with an IV-curve shaded to about 0.1 suns irradiance; actually just the short-circuit current density jsc,shaded and the open-circuit voltage Voc,shaded of the shaded curve need to be determined (see Fig. 4.5). Furthermore the voltage point VA is determined of the one-sun IV-curve at the current density jsc-jsc,shaded. The series resistance can then be calculated, similar as in the previous described method, by: rs ,shaded = Voc ,shaded − V A j sc − j sc ,shaded (4.7) As Voc,shaded is not affected by series resistance, this method is not influenced by the dependence of the series resistance from the irradiance. To determine rs at the maximum power point the shading level should be chosen in that way, that jscjsc,shaded is equal to jmpp. 4.1.6 Integration of the area under an IV-curve A very elegant way to determine the series resistance would be to extract the resistance directly from the one-sun IV-curve, since just one measurement is necessary. In 1982 a method was suggested by Araujo and Sánchez [97] based on the rs-modified one diode model equation: V + jrS k ⋅T ⋅ n1 j (V ) = j0 e − 1 − j Ph . (4.8) This implicit equation in current density is solved for the voltage V and then integrated from zero to short-circuit current density to gain the area A bordered by the one-sun IV-curve (see Fig. 4.6). j = jSC A= ∫ V ( j )dj. (4.9) j =0 Equation (4.9) can now be solved for the series resistance and simplified to: V A k ⋅ T 1 rs ,integral = 2 ⋅ oc − − n ⋅ 1 j sc j 2 q j sc sc (4.10) As this method is based on the one-diode exponential equation, the series resistance determination will also be affected by a low parallel resistance or a high Series resistance determination from IV-curves 67 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2 Current density j [mA/cm ] recombination current. In addition the series resistance is averaged over the whole IV-curve, which can underestimate the series resistance at the maximum power point as explained before (see Chapter 4.1.4). -40 -30 -20 -10 one-sun IV-curve Area A bordered by the IV-curve 0 200 400 600 Voltage V [mV] Fig. 4.6: A one-sun IV-curve is presented of which the voltage integral from zero to short-circuit current density is calculated to determine the series resistance. 4.2 Correlation between fill factor and series resistance A simulation using the two-diode equation (eq. (4.1)) was performed to analyze the effect of the series resistance on the fill factor. For the processed cells the following parameters were assumed: j01 = 1.3·10-12 Ω cm², j02 = 1.1 10-8 Ω cm², n1 = 1, n2 = 2 and rp = 5 kΩ cm², whereas the series resistance was varied between rs = 0 Ω cm² and rs = 2 Ω cm². The fill factor has been extracted and plotted versus the series resistance for photo-generated current densities of jph = 32 mA/cm², jph = 36 mA/cm², and jph = 40 mA/cm² as illustrated in Fig. 4.7. The fill factor declines more rapidly for solar cells with higher current densities due to the higher voltage drop at the series resistance, whereas FF0, the fill factor at rs =0 Ω cm², almost remains the same. For typical industrial silicon solar cells with a maximum series resistance of rs = 2.0 Ω cm² the data points for each curve can be fitted linearly. For the linear fit with a photo-generated current density of jph = 36 mA/cm² the slope is equal to m = 5.1 ± 0.01 %/Ω cm² and FF0 = 82.2 ± 0.01%. The slope of the linear fit can also be calculated directly from the IV-parameters with the assumption that jsc, Voc and jmpp are constant for all solar cells. This assumption is also theoretically justified since it is known that a change of rS mainly creates a voltage drop and not a change in current [98]. An additional series 68 Series resistance determination from IV-curves _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fill factor FF [%] 82 Photo current density jph = 40 mA/cm² 80 jph = 36 mA/cm² jph = 32 mA/cm² 78 76 74 Linear Fit 82.3% - 5.7 %/Ωcm² x rs 72 82.2% - 5.1 %/Ωcm² x rs 82.0% - 4.6 %/Ωcm² x rs 70 0.0 0.5 1.0 1.5 Series resistance rs [Ω cm²] 2.0 Fig. 4.7: Calculated fill factor plotted versus the series resistance for different photo-current densities. Data extracted from the two-diode equation and linearly fitted between rs = 0 Ω cm² and rs = 2.0 Ω cm². Following parameters were assumed: j01 = 1.3·10-12 Ω cm², j02 = 1.1·10-8 Ω cm², n1 = 1, n2 = 2 and rp =5 kΩ cm². resistance will just cause an additional voltage drop ∆V at jmpp. ∆ rs = (V mpp + ∆V@ jmpp j mpp )− V mpp j mpp = ∆V@ jmpp (4.11) j mpp Therefore it should be possible to calculate the slope mcalc of the linear fit, by applying the above made assumptions in the definition for the FF and calculating the FF-difference. mcalc ( ) 2 j mpp ∆V@ j j mpp V1 + ∆V@ jmpp − j mppV1 j mpp ∆FF FF2 − FF1 mpp = = = = = ∆rs ∆rs ∆rs ⋅ Voc ⋅ j sc Voc j sc ∆rS Voc j sc (4.12) Hence the expected series resistance rs,exp for a solar cell with a distinctive fill factor can be approximated by: rs ,exp ≈ FF0 − FFmeasured . mcalc (4.13) 4.3 Experimental comparison of determination methods For the experimental comparison 20 solar cells from a single batch with an emitter sheet resistance of Rsh=55 Ω/sq were used (IV-parameters are presented in Chapter 6.3.3). As an extensive firing variation has been performed, the fill factor ranged from 73% to 80% due to the differences in the series resistance. All 20 solar cells used for further investigation had a parallel resistance rp above Series resistance determination from IV-curves 69 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 3 kΩ cm². In order to determine the series resistance for each solar cell and each method, IV-curves of the processed cells at four different irradiation intensities (1 sun, 0.9 suns, 0.5 suns and 0.1 suns) were measured using appropriate filters to reduce the irradiance. Additionally, the dark IV- and the Suns-Voc curve have been measured. The pseudo fill factor PFF of the series resistance free Suns-Voc IVcurve has been extracted for each cell. The mean pseudo fill factor for all cells is equal to PFF = 82.2% ± 0.2% showing that the cells do not suffer from extensive shunts or space charge recombination currents. The determined series resistance values for each method were plotted versus the fill factor and linearly fitted as shown in Fig. 4.8. From these fits the slope mfit and the FF-axis intersection point FF0,fit for each rs determination method were extracted (see Table 4.1). 84 rs,dark rs,light_dark Fill factor FF [%] 82 rs,suns rs,int_var 80 rs,integral rs,shaded 78 Calculation from 2-diode eq. 76 74 0.0 0.5 1.0 1.5 2.0 Series resistance rs [Ω cm²] Fig. 4.8: Fill factor of the processed solar cells plotted versus the six series resistance determination methods. Data points of each method were fitted linearly. Furthermore the fill factor versus series resistance slope was calculated for each solar cell as described in the previous section. This can be done as the values for jmpp (jmpp = 34.0 ± 0.3 mA/cm²), Voc (Voc = 619 ± 2 mV) and jsc (jsc = 36.3 ± 0.2 mA/cm²) can be assumed constant for all solar cells with an rp > 3 kΩ cm² and a fill factor above FF = 72%. This procedure results in a mean value and standard deviation for the calculated value of mcalc = 5.1%/Ω cm² ± 0.1 %/Ω cm², being close to the simulated value in Chapter 4.2. At this point the mean values of mcalc and PFF can be compared with each mfit and FF0,fit from the linear fit equations (see Table 4.1). 70 Series resistance determination from IV-curves _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 4.1: Experimentally determined values for each rs method of the fit of the FF versus rs plot and the correspondent intersection point with the FF-axis compared to theoretical obtained ones.. rs method slope m (%/ Ω cm2) FF0 (%) PFF - 82.2 ± 0.2 calc. value -5.1 ± 0.1 - Experimentally determined values -5.2 ± 0.1 rs,light_dark -5.1 ± 0.1 rs,Suns -5.3 ± 0.1 rs,shaded 82.4 ± 0.1 82.3 ± 0.1 82.4 ± 0.1 rs,int_var -5.3 ± 0.2 82.3 ± 0.2 rs,integral -6.9 ± 0.1 83.4 ± 0.1 rs,dark -11.5 ± 0.5 84.8 ± 0.4 From this analysis the following methods seems to be equally good: - the comparison of jsc and Voc values of a shaded IV-curve with the illuminated IV-curve (rs,shaded) - the comparison of the dark and the illuminated IV-curve (rs,light_dark) - the comparison of the Suns-Voc and the illuminated IV-curve (rs,suns) - and the light intensity variation method (rs,int_var). All four methods are within the uncertainty range of the theoretical calculated value (see Fig. 4.9 and Table 4.1). Results will be further discussed in 4.5. 0.1 min ∆rs [Ω cm²] 0.0 -0.1 -0.2 max standard deviation median mean -0.3 -0.4 -0.5 -0.6 'dark' 'light_dark' 'suns' 'int_var' 'integral' 'shaded' rs determination method Fig. 4.9: The absolute difference to the expected series resistance is plotted for each determination methods, including mean, median, minimum, maximum value and standard deviation. The expected series resistance value was calculated using eq. (4.13) including the uncertainty value (dashed line). Series resistance determination from IV-curves 71 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 4.4 Influence of non-ideal IV-curves on rs determination The fill factor of a solar cell is strongly influenced by the series resistance, but is also affected by high recombination and shunt currents in the space-charge region. Thus in a further investigation the influence of the different methods on high recombination currents was tested. The series resistance of solar cells was determined before and after the surface of the emitter was scratched. Parallel to the fingers (not intercepting any), scratches were distributed over the cell surface. This resulted in an intended increase of the space-charge region recombination current density j02. Fig. 4.10 shows dark IV-measurements; illustrating the increase in recombination currents due to the scratching process, whereas the series resistance seems to be not affected. Note that by the scratching process the lumped series resistance can also be slightly affected due to a different current flow in the emitter layer. However, one would expect a reduction of the lumped series resistance due to less current collected by a finger located close to the scratch. Thus the same or a slightly reduced series resistance value would be expected before and after adding scratches. The series resistance values were determined using the above discussed methods, except the one in which the jsc and Voc values at 0.1 suns irradiance is compared with the one-sun IV-curve as data were not available. After obtaining the series resistance values, the difference between the two states (∆rs = rs,after-rs,before) as well as the mean value and standard deviation were calculated (see Fig. 4.11). The Suns-Voc and light-dark comparison method do not seem to be influenced 1 10 0 constant increased recombination 10 -1 10 -2 10 before surface scratch after surface scratch -3 10 0 200 400 600 800 Voltage V [mV] Fig. 4.10: Dark IV-curve of the same solar cell before and after adding scratches to the front surface. The recombination currents are increased. Counts of ∆rS [counts] Current density j [mA/cm²] rS remains 4 2 0 4 2 0 4 2 0 4 2 0 ∆r = 0.04 ± 0.05 Ω cm 2 ∆r = -0.05 ± 0.05 Ω cm ∆r = -0.06 ± 0.12 Ω cm ∆r = 0.17 ± 0.09 Ω cm -0.4 -0.2 int_var 2 2 2 0.0 0.2 ∆rs [Ω cm²] suns light_dark integral 0.4 Fig. 4.11: The four histograms show the distribution of the ∆rs = rs,after-rs,before values in 0.05 Ω cm² intervals for the different series resistance determination methods. 72 Series resistance determination from IV-curves _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ by the artificial damage. Also the intensity variation method seems to be (if at all) minor affected by the damage. The average series resistance value rises slightly. However, the integration method is strongly influenced by shunt-currents. The apparent series resistance values after the scratch process are all higher than the values determined before. The area under the IV-curve, which is used for the series resistance calculation (equation (4.10)) is reduced by low shunt resistance. 4.5 Detailed discussion of presented methods As expected from theory [85] the value of the series resistance determined from the dark IV-curve fit compared to the expected value is too low, due to the different current flow pattern under irradiated and dark condition. For large-area screenprinted silicon solar cells this effect is even increased. On the one hand the voltage drop across the typically 30 mm long fingers of a 12.5 x 12.5 cm² sized solar cell is not negligible, on the other hand the contact area of about 4.8% under illuminated condition (the current is collected mainly by the fingers) is significantly smaller than the contact area of about 8% (including two 2 mm wide busbars) under dark condition. The integration method seems to result in fairly exact rs values between 0.5 Ω cm² and 0.8 Ω cm², but underestimates the series resistance for fill factors below 78%. Assuming a constant series resistance over the whole IV-curve might be the reason. When applying the former boundary equation j·rs ≤ n·k·T/q with j = 34 mA/cm², the series resistance is assumed to be constant up to a value of about 0.76 Ω cm². If all data points between 0.5 Ω cm² and 0.76 Ω cm² are linearly fitted, fit values of mfit = -5.5%/Ω cm2 and FF0,fit = 82.5%, are achieved, a significant improvement compared to the fit parameters over the whole measuring range (compare Table 4.1). Nevertheless, this method can not be recommended for series resistance determination. It is based on the modified one-diode model equation, and therefore strongly influenced by low parallel resistance and high junction and edge recombination currents. Although the other four methods seem to be accurate, they need to be carefully applied. A good agreement with the theoretical values is achieved by comparing the series resistance free jsc-Voc curve with the illuminated IV-curve. The fitted slope is close to the calculated one and also the standard deviation is quite small. Nevertheless, if measuring the jsc-Voc curve as proposed by Sinton et al. [90], it is important that the base resistivity, the thickness of the cell, the temperature during the measurement as well as the short-circuit current density at one-sun irradiance Series resistance determination from IV-curves 73 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ are exactly known. In addition one should be aware that two different measurement setups are used, producing spectral mismatch between the spectral irradiance of the flash and the steady state sun simulator. Besides the matching series resistance results the great advantage in using this measurement method is that from the SunsVoc measurement additional cell data as e.g. the pseudo fill factor can be obtained. The irradiance variation values as presented in Table 3 and Figure 7 were achieved by linearly fitting the Vi,ji data points of the one-sun, the 0.9 suns and the 0.5 suns IV-curves. Hereby a small error is introduced as the series resistance for different irradiances might be different for high rs values. Due to the good agreement with the expected values the introduced error seems to be small. Whereas, when only using the IV-data at one-sun and 0.9 suns irradiance, the voltage difference at point V1,j1 and V2,j2 for a solar cell with a series resistance of about 0.6 Ω cm² is only 2 mV. It is obvious that an accurate IV-measurement is crucial, depending on the irradiance fluctuation of the sun tester, the number of data points measured around Vi,ji, the quality of the interpolation function to determine the correct intersection point and the accuracy of the used measuring systems. Since the “perfect” measuring condition as proposed in [89] are difficult to achieve in an industrial environment, using the Vi,ji points of the one-sun, 0.9 suns and 0.5 suns are more reliable under practical aspects. Nevertheless rs values for high series resistance might be slightly underestimated. The advantage of the irradiance variation measurement is that the solar cell is measured with the same contacting set-up. Comparing the jsc and Voc values of a 0.1 suns irradiated with the one-sun IVcurve, the determined series resistance values are in good agreement with the theoretical obtained ones. This method combines the advantages of the previous two measuring methods discussed; the same measuring set-up and irradiation source can be used and the method is independent of rs variation at different irradiances as long as jmpp of the one-sun IV-curve is about jsc-jsc,shaded. In addition, this method is very simple to integrate as the jsc and Voc values of the shaded IVcurve need to be determined. The comparison of the dark and illuminated IV-curve as modified by Dicker [86] results in very reliable series resistance values compared to the calculated ones. A further advantage of this method is that the same contacting set-up can be used. By measuring the dark IV-curve, additional information as e.g. the recombination currents and the parallel resistance can be directly extracted. This method was used throughout this work for series resistance determination. 74 Evaporation of different contact metals _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 5 Evaporation of different contact metals Contacts of fine geometry can be fabricated by photolithographical definition of the front side structure and evaporation of metals. Due to a relatively complex production sequence and high material waste, this process is costly and hence typically used for space solar cells or concentrating systems, but not for industrially manufactured terrestrial silicon solar cells. Titanium is the preferred contact material because of its relatively low barrier height to n-doped silicon and its good adhesion. In addition, the diffusion coefficient of titanium in silicon is low. In this chapter, different metals were investigated in their properties as contact layer and compared to the standard stack system of titanium-palladium-silver. These metals were chromium, titanium, palladium, silver, nickel and aluminum covered by silver as conducting layer. 5.1 Solar cell processing Solar cells with different contact materials were processed using a highefficiency production sequence as illustrated in Fig. 5.1. Four inch boron-doped FZ-silicon wafers with a thickness of 250 µm and a base resistivity of 0.5 Ω cm were used for solar cell processing. On each wafer seven solar cells with a size of 2 x 2 cm² were processed. After front side texture, a phosphorus emitter was diffused, resulting in a sheet resistance of Rsh = 120 Ω/sq. A single layer antireflection coating (SiO2) was thermally grown with a thickness of 100 nm. By this thermal step, the emitter was driven in, which results in a reduced surface dopant concentration of phosphorus of 4 inch FZ-Wafer (250 µm, 0.5 Ω cm, boron doped) Evaporation of aluminum on rear Wet chemical cleaning Laser-fired contacts Masking oxidation Annealing/ sintering Selective oxide removal from front (areas of 2x2 cm²) Photolithographic definition of front side structure Front surface texture Evaporation of metal on the front & lift-off POCl3 planar diffusion (120 Ω/sq) Light-induced silver plating PSG and oxide etching, cleaning Annealing/ sintering Thermal growing of oxide on front and rear (100 nm) Characterization Fig. 5.1: Process flow diagram of the process sequence of high-efficiency solar cells. Evaporation of different contact metals 75 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ about 7·1018 cm-3. The emitter profile is presented in Fig. 2.2 of Chapter 2.1.1. On the rear side, on top of the oxide layer, 2 µm of aluminum was evaporated, followed by the laser-fired contact process. To complete the rear side process, and hence to be independent from further front side processing, the solar cells were annealed at 425°C for 25 minutes in forming gas ambient to remove the laserinduced damage and alneal the SiO2 layer [99]. At this stage a variety of contact metals were evaporated by electron gun onto the photolithographically defined front side structure (see Chapter 2.2.7). The width of the evaporated contacts wc was about 7 µm and the height was 100 nm. To prevent the metal surface from oxidization and to create at the same time a start layer for the plating process, a 50 nm thick layer of silver was deposited on top of the first metal layer in one go. Afterwards the metal lines were further thickened by light-induced plating and the solar cells were separated by laser scribing and cleaving. At the end of the process sequence, a sintering step of 10 minutes followed in which the sintering temperature was varied between 100°C and 600°C. 5.2 IV-Parameters of processed solar cells 5.2.1 Different metals evaporated on inverted pyramid front In a first batch an inverted pyramids texture was applied to the front side. This texture is achieved by photolithographical definition, similar to structuring the front side metal grid: A photoresist is deposited on the front side, masked, illuminated and the illuminated areas are washed out. The uncovered silicon is wet chemically etched, resulting in a well defined inverted pyramids structure (see Fig. 5.2a). In addition, flat areas for the metallization pattern are defined. 12 wafers were processed using titanium-palladium, palladium, silver, nickel or aluminum as contact layer on the front side. The IV-parameters are presented in Table 5.1. High efficiencies of 21.4% for cells with Ti-Pd-Ag contacts and 21.5% for the one with Ni-Ag contact were achieved, whereas the efficiency for the PdAg contact cells was slightly lower due to the reduced fill factor. For these three metal stacks the open-circuit voltage and short-circuit current density are comparable and on a relatively high level. The best of the aluminum solar cells achieved an efficiency of 16.6% at a significantly reduced fill factor of 66%. The fill factor of most of the other solar cells with aluminum as first layer was on an 76 Evaporation of different contact metals _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ evaporated and plated finger screen-printed and plated finger 90 µm random pyramids a) b) inverted pyramids Fig. 5.2: SEM image of an inverted (a) and a random (b) pyramids front side structure. The left-hand picture shows an evaporated contact, the right-hand one a screen-printed contact, both thickened by plated silver. even lower level. Dark IV-measurements revealed a low shunt resistance. For the cells with silver as contact material the adhesion to the silicon surface was poor. The complete silver layer was removed in the lift-off process. From this experiment it can be concluded that evaporated silver and aluminum are not suitable as contact layer to a lowly n-doped silicon surface, whereas the Ni, Ti and Pd contacts gave good results. Table 5.1: 1st batch: IV-parameters of solar cells with different contact materials achieving highest efficiency. The solar cells in this first batch received an inverted pyramids front side texture. metal Voc [mV] jsc [mA/cm2] Ti-Pd-Ag 678.4 39.3 Ni-Ag 676.4 39.5 Pd-Ag 676.7 39.2 Al-Ag 640.8 38.1 Ag no adhesion after lift off η FF [%] [%] 80.3 80.5 77.2 66.2 21.4 21.5 20.5 16.1 5.2.2 Different metals evaporated on a random pyramid front A second batch consisting of 12 wafers was processed. To simplify the manufacturing of the cells, a random pyramids texture process without photolithographical definition was used instead of an inverted one. Silver and palladium was replaced by titanium and chromium as contact layer. The sintering temperature was varied between 100°C and 600°C, a few cells were not sintered at Evaporation of different contact metals 77 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ all. Ti-Pd-Ag, Ti-Ag, Al-Ag and Cr-Ag were deposited on two wafers, respectively, and Ni was deposited on the remaining four wafers. Hence for each type of metal 14 solar cells (28 cells for Ni) were processed. The solar cell results for the best cells of each type of metal under investigation are presented in Table 5.2. The efficiency η of 20.8% to 21.0% for all types of contact layers is very similar, as well as the open-circuit voltage, the short-circuit current density and the fill factor. Compared to the first batch, the open-circuit voltage and short-circuit current density are on a slightly lower level, possibly due to a higher effective front surface recombination velocity caused by the increased contact area13. The enlarged contact area or different crystal orientation might be responsible for the higher fill factor achieved. Table 5.2: 2nd batch: IV-parameters of solar cells with different contact materials achieving highest efficiencies. The solar cells in this second batch received a random pyramids front side texture and were sintered at different temperatures Tsint. metal stack Tsint [°C] Voc [mV] jsc [mA/cm2] FF [%] [%] Ti-Pd-Ag Ni-Ag Al-Ag Ti-Ag Cr-Ag 450 400 500 450 100 663.1 663.0 662.5 662.9 662.5 38.8 38.8 38.7 38.9 39.0 81.3 81.6 81.2 81.5 81.2 20.9 21.0 20.8 21.0 21.0 η 5.3 Discussion of the different metals In the following, the use of different metals for solar cell front side metallization will be briefly discussed. This analysis is based on the second batch, except from palladium having been used only as contact material in the first batch. A wide sintering temperature range has been applied from 100°C to 600°C. The IVparameters of the different types of contact metals differ strongly over the temperature range, as presented in Fig. 5.3 and Fig. 5.4. Each data point represents the mean value of the IV-parameters of two to three solar cells. In the following, the characteristics for each contact metal over the temperature range are described 13 The contact area of the random pyramids texture is about 1.7 times larger than the flat contact area which was photolithographically defined. 78 Evaporation of different contact metals _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ in more detail. A common feature of all processed solar cells is the significant efficiency drop above 550°. The oxide passivation layer on the rear side is damaged by the aluminum layer at temperatures above 550°C [100]. 85 20 75 Efficiency η [%] Fill factor FF [%] 80 70 65 60 55 35 0 18 16 14 3.5 100 200 300 400 500 0 600 100 200 300 400 500 600 680 39 670 660 Voc [mV] jsc [mA/cm²] 38 37 36 35 10 0 200 300 400 Temperature Tsint [°C] 500 640 630 Ti-Pd-Ag Al-Ag st Pd-Ag (1 batch) 100 650 620 600 610 0 100 200 300 400 500 600 Temperature Tsint [°C] Fig. 5.3: IV-parameters of solar cells with Ti-Pd, Al, Pd as contact materials. Whereas the data from Ti-Pd and Al are taken from the second batch, data for Pd are from the first batch. Each data point presents the mean value of two to three cells. 5.3.1 Aluminum The solar cells with aluminum-silver contacts have the largest efficiency distribution over the performed temperature range. Whereas relatively high efficiencies of over 20% were achieved at temperatures between 450°C and 500°C, the lowest efficiency (14%) was obtained 100°C lower, at 350°C. The fill factor is responsible for this variation in efficiency. The fill factor, on a low level at room temperature (75%), dropped further to its minimum value of 59% at 350°C. Then the FF sharply rose to values above 80% between 455°C and 550°C. As measurements have shown, the series resistance is responsible for the fill factor variation and not a shunt as one could expect due to Al spiking. Evaporation of different contact metals 79 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ The short-circuit current density rises from a relatively low value (compared to the other metals) of 37 mA/cm² to above 38 mA/cm² at temperatures between 350°C and 500°C, having its peak mean value of 38.7 mA/cm² at 400°C. The open-circuit voltage curve over the temperature range can be explained by the jsc and FF curve. It drops to values of 620 mV between 200°C and 300°C, but recovers again at temperatures between 400°C and 500°C to values of about 660 mV. Trying to recover the efficiency of cells sintered at 350°C by applying a second sinter step at 450°C failed, whereas the efficiency of cells which were sintered at 100°C could be completely shifted to the same efficiency level compared to the ones that were directly sintered at 450°C. One possible explanation would be that below 100°C a tunneling contact is formed over the native oxide layer. This native oxide is penetrated at elevated temperatures. At 400°C aluminum diffusion in silicon starts, probably resulting in a good electrical contact. As recent measurements have shown, the solar cells with aluminum as seed layer completely degraded after 24 months. The fill factor of nearly all cells dropped to values under 60%. Dark IV-measurements revealed that neither recombination currents (j02 < 2·10-9 A/cm²) were increased nor the parallel resistance was reduced (rP > 2·106 Ω cm²). However, the measured series resistance of the cells was elevated. Values in the range of 4 Ω cm² to 50 Ω cm² were obtained, which would result in a fill factor drop of at least 20%. Probably the interface layer between aluminum and silver is damaged. 5.3.2 Palladium Palladium is of high importance in the semiconductor industry, as the Pd2Si phase has a low enthalpy of formation, starting at 200°C. For example 30 nm of Pd2Si are formed at a sintering temperature of 275°C for 20 minutes. The formation is independent from the silicon crystal orientation. Pd2Si in contact to psilicon is especially suitable as the barrier height is just φBp = 0.2 eV, the barrier height of φBn = 0.8 eV to n-silicon is significantly higher [28,101]. The results presented in Fig. 5.3 are based on the first batch, because palladium was not used as contact layer in the second one. Hence the data for palladium are not directly comparable with the other ones. Due to the slightly different process applied, the maximum achieved short-circuit current density and open-circuit voltage in the previous experiment for all types of metals is on a higher level 80 Evaporation of different contact metals _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ (compare Table 5.1 and Table 5.2). Nevertheless, the behavior over the temperature range for palladium should be representative. Palladium does not seem to form an electrical contact at room temperature. Between 200°C and 600°C sintering temperature, the trend in Voc and jsc is comparable to the trend for Ti-Pd: jsc stays on a high level between 200°C and 500°C with a maximum mean value of 39.4 mA/cm² at 350°C, Voc increases steeply from 645 mV at room temperature to its maximum value of 677 mV at 425°C. However, the fill factor is on a low level over the whole temperature range. The maximum value of just 77.2% is reached at 425°C. This low performance is theoretically expected due to the barrier height to n-silicon. Thus the maximum achieved efficiency of 20.5% is about 1% absolute less compared to cell results with nickel or titanium-palladium used as contact layer (compare Table 5.1). The low cell performance was the reason why palladium has not been used for solar cell processing in the second batch. 5.3.3 Silver Silver was tested as contact layer in the first batch only. The adhesion was poor so that it was completely removed during the lift-off process. Evaporated silver seems to be unsuitable as contact layer to moderately n-doped silicon. This shows that silver can only be used in combination with glass frits or similar materials as e.g. in screen-printing pastes. 5.3.4 Titanium-Palladium The solar cells with Ti-Pd, Ti, Cr and Ni as contact layer resulted in IVparameters on a high level (see Fig. 5.4). The trend of the IV-parameters over the temperature range for the system titanium-palladium-silver is described in more detail. The other contact materials are compared to this one. Titanium-palladium-silver as stack system for solar cell contacts was developed for space solar cells [102,103]. As mentioned above, titanium forms a good electrical and mechanical contact to silicon, palladium acts as diffusion barrier between titanium and silver and the silver layer is used as seed layer for the silver plating process. Titanium, which is very reactive, reduces the native silicon oxide during the evaporation process; hence already at room temperature a contact with a low contact resistance is achieved. The drawbacks of this stack system are the high Evaporation of different contact metals 81 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ material costs, especially for palladium, and the three layer stack system instead of a two layer one. The open-circuit voltage is between 650 mV and 655 mV for sintering temperatures up to 200°C and then rises by about 10 mV for temperatures between 350°C and 450°C (see Fig. 5.4). The electron gun evaporation of titanium and palladium (as well as for Ni and Al) might damage the front surface passivation layer. This induced damage is annealed at temperatures above 200°C, resulting in a voltage increase. The short-circuit current density has a similar curve progression as the open-circuit voltage, rising slightly from 38.3 mA/cm² to a maximum value of 38.8 mA/cm² at 450°C. The fill factor of 80.9% at room temperature rises to a value of 81.6% at 100°C, but then drops below 81% between 350°C and 425°C. However, if a sintering temperature of 450°C or 500°C is applied, the fill factor obtained is again well above 81%. Due to the small amount of samples processed, a further investigation would be advisable to validate if the observed fill factor dependence in this temperature region can be reproduced. 21.0 82 20.8 η [%] FF [%] 20.6 81 80 18 16 14 0 100 200 300 Temperature T 400 500 600 [°C] 39 660 38 650 37 36 35 0 Ni-Ag Cr-Ag Ti-Ag Ti-Pd-Ag 100 200 100 200 300 Temperature T 670 Voc [mV] jsc [mA/cm²] 40 20.2 20.0 70 60 0 20.4 400 500 600 500 600 [°C] 640 630 300 400 500 Temperature Tsint [°C] 600 620 0 100 200 300 400 Temperature Tsint [°C] Fig. 5.4: IV-parameters plotted versus the sintering temperature for cells with different front side contact materials. Each data point is the mean value of 2 – 3 samples. 82 Evaporation of different contact metals _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ The efficiency of all solar cells sintered between room temperature and 500°C is between 20% and 21%. The highest mean efficiency of 21.0% is achieved at 500°C. The efficiency of the cells measured again after 24 months was at the same level. 5.3.5 Titanium Instead of using palladium as diffusion barrier and adhesion layer between titanium and silver, silver was directly evaporated on titanium. The adhesion of silver to titanium at high humidity is low [104]. However, as the intermediate palladium layer is expensive, this simplified stack system was tested. The opencircuit voltage of cells with titanium as contact material has a similar curve progression for the analyzed temperature range compared to the previously discussed Ti-Pd stack system: Voc rises by about 10 mV at temperatures above 200°C. However, the short-circuit current density and the fill factor remain quite constant over the whole temperature range. The fill factor FF reaches its maximum value at 500°C, jsc at 400°C. The efficiency is on a constant high level between 350°C and 500°C. The maximum mean efficiency of 20.8% is obtained at 400°C. Similar efficiencies were measured after 24 months. However, due to the reported reactivity between titanium and silver, a damp heat test of encapsulated cells would be of interest. 5.3.6 Nickel Nickel has six intermetallic phases below 950°C with silicon, three of importance for thin films. The effect of different sintering temperatures and times on the solar cell performance was already investigated in 1980 by Anderson and Peterson [69]. The phase Ni2Si is formed at temperatures between 200°C and 400°C, the NiSi phase between 400°C and 750°C and the third phase above 800°C. The growth of the Ni2Si layer is diffusion controlled having an activation energy of 1.3 eV to 1.5 eV. Barrier heights of approximately 0.62 eV for Ni2Si/Si and 0.67 eV for NiSi diodes were reported, respectively [105]. The open-circuit voltage and short-circuit current density for cells with nickel contacts also have a similar trend as the one of Ti-Pd. The maximum mean Voc is achieved at 425°C, the maximum mean jsc at 400°C. Evaporation of different contact metals 83 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Compared to cells with Ti, Ti-Pd and Cr contacts, the fill factor and efficiency of cells with Ni contacts at room temperature are the lowest. But already at a temperature of 100°C the fill factor ascends significantly. The maximum mean FF value of 81.6% is achieved at 350°C. The efficiency of these cells is on a constantly high level between 350°C and 425°C. The maximum mean efficiency value of this batch is achieved at 425°C having a value of 20.9%. The cell performance was unchanged after 24 months. 5.3.7 Chromium Chromium disilicide (CrSi2) penetrates deeply into the silicon [106,107]. Nevertheless, chromium was tested as contact material to analyze if it is suited for low temperature contact formation to the n-doped emitter surface of solar cells. Compared to the other tested contact materials, chromium shows some significant differences in the IV-parameter – temperature curve progression. The open-circuit voltage is already at room temperature on the high level which is reached for other type of material only after applying a sintering step above 200°C. The higher Voc might be explained by the lower acceleration voltage used during the electron-gun evaporation process for chromium. The acceleration voltage was 8.5 kV for chromium evaporation and about 11 kV to 12 kV for the other processes. This is an indication, that the bremsstrahlung during the chromium process is reduced, probably damaging the silicon surface less compared to the processes with higher bremsstrahlung. The maximum Voc of about 664 mV is reached at a temperature between 200°C and 350°C. Above this temperature Voc is slightly reduced. The short-circuit current density has its maximum with 39.0 mA/cm² already at room temperature. Then jsc drops slightly by 0.5 mA/cm² to 38.5 mA/cm² for temperatures between 200°C and 500°C. The fill factor rises up to a temperature of 400°C, but declines significantly above 450°C. Contrary to all other tested contact materials, the efficiency level is highest between room temperature and 200°C. The maximum efficiency of 20.8% is reached at 100°C. For temperatures above or equal to 350°C, the efficiency declines. Measuring the efficiencies of these cells after 24 months, no significant change in IV-parameters could be recorded. 84 Evaporation of different contact metals _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 5.4 Chapter summary High-efficiency solar cells have been processed with evaporated and plated metal contacts. Metals under investigation were Ti-Pd-Ag, Ti-Ag, Al-Ag, Ag, PdAg, Ni-Ag and Cr-Ag. Applying a sintering temperature of 500°C or above resulted in an efficiency drop for all solar cells, probably due to a damage of the oxide passivated rear side. As expected the efficiency of about 21% for cells with Ti-Pd-Ag contacts was on a high level. The adhesion of pure silver contacts to the silicon surface was poor. The measured contact resistivity of palladium was high. Using Ti-Ag, Cr-Ag and Ni-Ag as contact system resulted in cells of high-efficiency, comparable to the standard stack Ti-Pd-Ag. No degradation of these contact materials was observed after 24 months of storage under ambient air. The highest efficiencies for solar cells with chromium contacts were achieved applying a sintering temperature not exceeding 200°C. At higher temperatures the efficiency slightly dropped. The cells with Ni-Ag contacts performed very similar to cells with Ti-Pd-Ag contacts, achieving efficiencies on a high level between 350°C and 500°C. These investigations showed that also Ni-Ag and Cr-Ag contacts could be well suited for solar cell metallization. However, further adhesion tests, long term stability tests and damp heat tests are recommended. Screen-printing of hotmelt silver paste 85 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6 Screen-printing of hotmelt silver paste Screen-printing is the state-of-the-art technology for metallization of industrial silicon solar cells. The two main challenges for screen-printing of solar cell front contacts are firstly achieving a high line resolution of the final contact, and secondly achieving a low contact resistance to a lowly doped emitter. Up to now, screen-printed contacts are limited to a rather coarse line resolution, especially since the finger conductivity has to be kept on a high level at the same time. Even if the same aspect ratio (height : width) can be achieved, halving the finger width would result in a four times smaller cross section area, which means that the conductance of the fingers would also be reduced by a factor of four. In this chapter the use of hotmelt paste for the front side metallization of silicon solar cells is discussed. Hotmelt paste is solid at room temperature and can be processed like conventional paste at elevated temperatures [108]. Due to the increased silver fraction and the different processes applied, fingers with a high aspect ratio can be printed. Energy conversion efficiencies of η = 18.0% were achieved on 12.5 x 12.5 cm² sized Cz-silicon solar cells. Furthermore the width of the printed contacts was reduced and solar cells having an emitter sheet resistance of Rsh = 75 Ω/sq were processed. 6.1 Comparison of hotmelt and conventional paste 6.1.1 Composition The basic composition of hotmelt and conventional silver paste is similar. They mainly differ in the silver content, in the binding agent and the solvent. Table 6.1 lists the ingredients of the conventional paste and the hotmelt paste by its mass percent as stated on the Material Safety Data Sheet (MSDS) [109]. The silver content of the hotmelt paste is increased, while the solvent content is reduced, resulting in different flow properties compared to conventional paste. According to the manufacturer this effect is compensated by rheological additives. A long chain solvent is used to increase the melting point of the paste. Opposed to the short-chain solvent ethanol, which is fluid at room temperature, the long chain solvent 1-docosanol has a melting point Tmelt between 68°C and 71°C. In Table 6.2 the characteristics of both solvents are presented. 86 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 6.1: Composition of conventional paste (Ferro CN33-462) and hotmelt paste (Ferro CN33-473) by its mass percent. Data taken from the MSDS [109]. paste composition conventional paste [m%] hotmelt paste [m%] 70 – 85 10 – 25 - 80 – 90 5 – 10 1–5 1–5 1–5 1–5 1–5 1–5 silver particles solvent: - ethanol -1-docosanol binding agent: - modified TBP Polyethylene - cellulose ether boron silicate glass (glass frit) additives Depending on the length of the carbon compound of the solvent, the melting point can be adjusted. This can be utilized when two or even all three printing steps are performed using the hotmelt technology [110]. If the melting point of the paste is reduced between consecutive printing steps, a drying step in-between is not necessary. The temperature difference between the prints needs to be large enough to prevent a re-melting of the already printed contacts. In Fig. 6.1 a suitable process flow of the metallization process is illustrated. In the first step the aluminum-silver pads are printed on the rear using a hotmelt paste with a melting point of e.g. Tmelt = 85°C. Afterwards the remaining area of the rear is printed with an aluminum hotmelt paste having a lower melting point of e.g. Tmelt = 70°C. In the last printing step the front side structure is printed using a low melting point silver Table 6.2: Comparison of the solvents ethanol and 1-docosanol, which are used for the investigated conventional and hotmelt paste, respectively. type of solvent: structure formula: ethanol 1-docosanol short-chain solvent long-chain solvent H H H H H H–C–C–O–H H H color: melting point Tmelt: boiling point Tboil: colorless -114°C 78°C H–C–C–…C–O–H H H H white 68°C – 71°C 180°C Screen-printing of hotmelt silver paste 87 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ hotmelt paste (e.g. Tmelt = 55°C). The wafer is dried at a temperature Tdry between 300°C and 450°C until all solvents and organic components are removed. Then the contact formation process is performed in an inline fast firing furnace. printing rear side taps Ag/ Al-hotmelt paste Tmelt = 85°C printing rear side Al-hotmelt paste Tmelt = 70°C printing front side grid drying and contact firing Ag-hotmelt paste Tmelt = 55°C Fig. 6.1: Suitable printing process flow, if all three printing steps are performed using hotmelt paste. A drying step is not necessary in-between the prints. 6.1.2 Rheology The comparability of hotmelt and conventional paste during the printing process was analyzed by rheological investigations. The science of rheology distinguishes between fluids and solids. While ideal fluids are deformed irreversibly by force effects, non-destructive deformations of solids are reversible. Printing pastes like many other dispersions are neither ideal fluids nor ideal solids. From their characteristics they are in-between these two states, called pseudoplastic or thixotropic14. Printing pastes show time-dependent elastic and viscous behavior under shear stress. The printability of a paste depends on its pseudoplastic behavior, its rheology. Silver screen-printing paste, which is used for solar cell metallization, shows a sharp flow limit in its shear thinning behavior. This provides a high viscosity at a low load. If the applied force is increased, the viscosity will be reduced rapidly, which is the case in the print-through process. The paste is fluid enough, to be printable. On the substrate the paste is transferred into a recovery process. If the paste stays in the low viscosity, it tends to flow on the substrate. This tendency is reduced by the thixotropic flow behavior, in which the deformation is recovered. The recovery process needs to take long enough to level off the metallization15, but fast enough to keep line broadening low, as the paste flows in all directions. 14 Thixotropic: A full-bodied material which undergoes a reduction in body when shaken, stirred or otherwise mechanically disturbed and which readily recovers the original fullbodies condition on standing [42]. 15 Due to the mesh the paste supply is high at parts not covered by the wire and low at the node points. 88 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ The dynamic viscosity16 µ of both pastes was measured using a Bohlin coneplate viscometer. This system uses a cone of very shallow angle in bare contact to a flat plate. The dynamic viscosity is measured by the ratio of shear stress τshear to the rate of shear g under steady flow conditions. In Fig. 6.2-a the measured dynamic viscosity of the hotmelt paste is plotted versus the temperature at a g of 5 s-1 and 10 s-1. While the viscosity is high at low temperatures it decreases with rising temperature. Fig. 6.2-b shows the viscosity as a function of the shearing rate for the conventional paste at room temperature and for the hotmelt paste at a temperature of 80°C. A shearing rate γ between 1 s-1 and 100 s-1, which was chosen for the measurement, is equal to the applied force in the printing process [38]. A rate g between 1 s-1 and 10 s-1 simulates the filling phase, whereas the range from 10 s-1and 100 s-1 is equal to the applied load in the print-through process. The viscosity of hotmelt and conventional paste versus shearing rate behave similarly over the whole range measured. Hence, also the printing process should be comparable. 300 200 150 100 50 0 70 a) 75 80 85 90 Temperature Tpaste [°C] conventional paste (Tpaste= 25°C) 300 Viscosity µ [Pa s] 250 Viscosity µ [Pa s] 350 -1 shearing rate γ = 5 s -1 shearing rate γ = 10 s 200 150 100 50 0 0 95 b) hotmelt paste (Tpaste= 80°C) 250 10 20 30 -1 40 50 Shearing rate γ [s ] Fig. 6.2: a) Viscosity µ of the hotmelt paste plotted versus the paste temperature Tpaste. b) Viscosity µ of the hotmelt and conventional paste plotted versus the shearing rate γ. 16 Dynamic viscosity µ: If a fluid with a viscosity µ of one pascal second is placed between two plates, and one plate is pushed sideways with a shear stress γ of one pascal, it moves a distance equal to the thickness of the layer between the plates in one second. Screen-printing of hotmelt silver paste 89 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6.1.3 Thermogravimetrical analysis Mass percent m [m%] 100 0.0 -0.5 96 -1.0 92 88 84 0 -1.5 hotmelt paste TGA DTGA conv. paste TGA DTGA temperature gradiente: 10 K / min 100 200 300 400 500 -2.0 -2.5 Differential TGA [% / min] Two different thermal analyses were performed to investigate the paste properties under heat treatment: Thermogravimetry (TG) is a method to analyze the mass loss of the paste sample as a function of the temperature, whereas differential thermogravimetry (DTG) illustrates the volatile rate of solvents or organic compounds. This is useful information for the drying and firing process: Both analyses were performed at the Crystallographic Institute of the University of Freiburg using the measuring system Netzsch STA 449C with oxygen and argon as process gases and a temperature gradient of dT/dt = 10 K/min. The TGA and DTGA measurements of a conventional and a hotmelt paste from Ferro are illustrated in Fig. 6.3. At a temperature of 190°C the loss in mass percent starts for the investigated hotmelt paste, which is a temperature just above the boiling point of the solvent 1docosanol (Tboil = 180°C). The differential thermogravimetric analysis (DTGA) has its first peak at a temperature of about 310°C (see Table 6.3). At this temperature the volatized product consists of evaporated organic components and of the binding agent which starts to depolymerize and combust. The maximum vaporization rate of the cellulose ether is at 400°C (second minimum). At a temperature above 455°C, all organic components are removed from the paste; the mass of the paste stays constant. The solid content of the paste is 88% by mass. The thermogravimetrical analysis of the conventional paste shows a similar curve progression as the hotmelt paste, having distinctive minima. The evaporation of solvents in the conventional paste starts at a temperature below 100°C (boiling 600 Temperature Tpaste [°C] Fig. 6.3: Thermogravimetric analysis (TGA) and differential thermogravimetric analysis (DTGA) of a hotmelt and a conventional paste. 90 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ point ethanol: Tboil = 78°C) and reaches its maximum volatilization rate at a temperature of about 240°C. The second minimum is caused by the depolymerization of the binding agent TBP-polyethylene at a temperature of 350°C, and hence on a lower temperature level compared to the binding agent of the hotmelt paste. At a temperature above 390°C, all organic components are removed. Compared to hotmelt paste the conventional paste consists of four mass percent less solid fraction (84% by mass). These thermal analyzes revealed that the temperature profile of the drying/ burn out process needed to be adjusted for hotmelt printed contacts due to the higher boiling point of the paste components. All organic components should be completely removed from the paste before the contact is formed in the hightemperature process. Hence, if a sufficiently long burn-out time at a temperature above 420°C is chosen before the firing ramp the printed structure will consist of inorganic components only; all organic components are removed. Table 6.3: Results of the thermogravimetric (TG) and differential thermogravimetric (DTG) analysis. Paste [unit] [°C] 1st minima (DTGA) [°C] 2nd minima (DTGA) solid fraction [m%] volume fraction17 applying eq. (2.4) [v%] recommended drying temperature [°C] recommended burn-out temperature [°C] conventional hotmelt 240 350 84 37 250 – 300 > 350 310 - 330 420 88 45 350 – 400 > 420 6.2 Characterization of hotmelt printed contacts 6.2.1 Hardware setup The hotmelt paste is printable like the conventional one, but it requires elevated paste temperatures Tpaste of around 50°C to 90°C (depending on the solvent). All components in contact with the paste have to be heated, which includes the 17 For both pastes a glass frit fraction of 3% by mass was assumed (ρglass = 5 g/cm³) as well as a solvent fraction of 9% and 13% by mass for the hotmelt and conventional paste, respectively (ρsol = 1 g/cm³). Screen-printing of hotmelt silver paste 91 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ squeegees (Tsqueegee), the print nest (Tnest) and the screen (Tscreen). For the investigations, a standard half automated screen-printer from EKRA (meanwhile Asys) was upgraded with a heatable print nest as well as heatable squeegees. Electrical connections were placed inside the system to plug in the resistively heated screen. Hot Screens® manufactured by Koenen GmbH, Germany, were tested and used. Besides insulating the metal mesh from the aluminum frame, the trampoline clothing of the screen has a consistent blade pressure on the metal mesh over the total screen length [111]. A picture showing the screen placed inside the screen-printer with still solid hotmelt paste is presented in Fig. 6.4. Fig. 6.4: View inside the screen-printer, showing the electrically connected screen with not molten hotmelt paste. Doctor blade (squeegee), metal mesh, print nest and table are heatable. 6.2.2 Optical characterization Optical characterization of hotmelt printed fingers was performed. The aim was the development of a process of printing contacts with a width close to the printing pattern in the screen and a high aspect ratio. Prior to hotmelt, conventional paste with similar rheological properties was printed. The same type of screen as for the hotmelt investigations was used. Optimum settings as for e.g. the snap distance, print pressure, print speed, and emulsion thickness were evaluated and used as a starting point for the hotmelt technology investigation (compare Chapter 2.2.2). For printing hotmelt paste, the temperature settings of the heated screen and print nest primarily determine the line width and height after printing. The screen’s temperature Tscreen needs to be adjusted to about 5°C to 10°C above the melting point Tmelt of the paste to allow printing. The temperatures of the squeegees Tsqueegee were adjusted to the one of the heated screen to ensure lateral temperature 92 Screen-printing of hotmelt silver paste 150 60 140 55 130 120 screen clogging over time paste not printable 110 45 40 100 70 50 35 finger opening in silkscreen 75 80 85 90 95 Finger height hf [µm] Finger width wf [µm] _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 100 Screen temperature Tscreen [°C] Fig. 6.5: Line width and height as a function of the screen temperature for the hotmelt paste CN33-473. The melting point Tmelt of the solvent in the paste is 72°C. homogeneity of the paste. Fig. 6.5 shows the impact of the screen temperature on the finger height and width. Using a paste with a melting point of Tmelt = 72°C, best printing results were obtained setting the screen temperature Tscreen between 80°C and 90°C. The finger width wf in this case was about 15 µm to 20 µm wider than the finger width in the screen. Rising the temperature above Tscreen = 90°C, resulted in increased spreading of the paste on the wafer due to its lower viscosity. When setting the screen temperature too high, the paste solidifies on the screen, leading to a complete screen clogging over time. This clogging is not reversible. The screen needs to be cleaned or replaced. The print nest temperature Tnest also has a strong influence on the width and height of the printed fingers. When choosing the optimum temperature setting, finger heights of hf = 40 µm at a width of wf = 120 µm were achieved after the firing process. A print nest temperature Tnest below the optimum reduces the release of paste out of the screen. Thus, the paste is transferred only partly or not at all. With increasing temperature Tscreen the printed fingers are broader and less high. These results can be determined from Fig. 6.6-b showing the finger aspect ratio before and after the firing process. The best aspect ratio AR after the firing process is about 1 to 3. The reduction of the aspect ratio during the firing process is due to shrinking of the finger height. At too high print nest temperatures Tnest, the same volume of paste is transferred, but the aspect ratio declines as a result of broader and less high fingers due to flow effects on the wafer. The high aspect ratio after printing at low print nest temperature is due to sticking of paste to the screen during the lift off process, leaving a structure of low density with several high but Screen-printing of hotmelt silver paste 93 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 90 Aspect ratio (height:width) Print nest temperature Tnest [°C] b) 40 180 35 160 140 30 120 100 80 a) finger width finger height 50 60 70 80 25 Finger height hf [µm] Finger width wf [µm] 200 0.6 0.5 - 1:2 0.4 - 1:3 0.3 - 1:4 - 1:5 0.2 after printing process after firing process 0.1 50 60 70 80 90 Print nest temperature Tnest [°C] Fig. 6.6: a) Finger width and finger height as a function of the print nest temperature after the firing process. b) Aspect ratio before and after the firing process. thin peaks on the substrate. This profile heavily shrinks in height during the subsequent firing step. Beside the better conductivity, another advantage of the high aspect ratio of the hotmelt-printed fingers is illustrated in Fig. 6.7. Two 3D images of an untextured multicrystalline silicon solar cell surface are shown at a location where a grain boundary is crossed by a grid finger. The effect of the grain boundary after alkaline etch on the line conductivity is negligible for the high hotmelt printed finger (see Fig. 6.7-a), but not for the conventional one (Fig. 6.7-b), where the cross section area is significantly smaller or even finger interruption could occur at the crossing point. From these optical investigations it can be concluded that the hotmelt paste has a significant advantage over conventional paste due to improved finger geometry. Electrical characteristics of hotmelt printed contacts are presented in the following. hotmelt printed contact finger a) conventional printed contact finger grain boundary b) Fig. 6.7: 3D-images of multicrystalline silicon solar cell surface, at a location where a grain boundary is crossed by a finger printed with a) hotmelt and b) conventional paste. 94 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6.2.3 Electrical characterization To characterize different contacts printed with hotmelt paste and the quality of the firing process electrically, test structures for line conductivity and specific contact resistance investigations were printed. Firing variations were carried out for all evaluated pastes and the finger resistivity ρf and the contact resistivity between finger and semiconductor ρc were determined. For the firing process a rapid thermal single wafer firing furnace (RTF) [112] was used. Parameters of the firing profile, the peak temperature, the process time as well as the gas flow and ratio have been varied for different firing processes. To prevent paste re-melting with the risk of line spreading, the firing process was carefully prepared. Before applying the fast firing process, the wafer was kept at a temperature of about 450°C until all solvents and organics were evaporated. Fig. 6.8-a shows a cross section of a printed and fired contact finger attained using the hotmelt and the conventional technology. While the finger width of wf = 120 µm is the same, the height of hf = 30 µm for the hotmelt printed contact is significantly improved compared to the conventional printed one with a maximum height of hf = 13 µm. The line resistivity ρf of 2.9 to 3.5·10-8 Ω m for hotmelt printed fingers is similar or even slightly lower than the one for conventional paste (ρf ≈ 3.2·10-8 Ω m). But due to the higher aspect ratio and hence finger cross section area, the resistance per line length Rline decreases significantly. Rline of about 14 Ω/m was measured using a four point probe setup, a low value for a single screen-printed contact of width wf = 120 µm. For the conventional printed contact a line resistance of Rline = 34 Ω/m was obtained. Fig. 6.8-b presents a Finger height [µm] 50 40 Conventional: Area = 1000 µm² Hotmelt: Area = 2200 µm² 30 20 hotmelt 10 conventional 0 a) 0 50 100 Finger width [µm] b) Fig. 6.8: a) Cross sections of contact fingers formed by the hotmelt and the conventional technology measured with an optical profilometer. b) Microscope picture of a cross section of a contact after the firing process with a width of wf = 150 µm and a height of hf = 45 µm. Screen-printing of hotmelt silver paste 95 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ microscope picture of a cross section of a hotmelt printed contact after firing with a maximum height of hf = 45 µm at a width of wf = 150 µm. The specific contact resistivity ρc was measured using the test structure proposed by Meier and Schroder [101] (see Fig. B.2 in Appendix B). For the first hotmelt pastes provided, the voltage drop across the metal semiconductor interface was high. Despite extensive variations in the firing process, a specific contact resistivity ρc below 40 mΩ cm² could not be achieved. These high values have a strong influence on the series resistance and thus on the efficiency of the solar cell (see Chapter 4.2). The paste was continuously improved by Ferro in close cooperation with Fraunhofer ISE. The latest hotmelt paste CN33-473 delivered, reached a significantly lower value of specific contact resistance. This improvement, in combination with an optimized firing process, resulted in a specific contact resistivity as low as 1 mΩ cm² to an emitter with a sheet resistance of Rsh = 40 Ω/sq. 6.3 Solar cell processing and results In this section, results of solar cells are presented that were produced using an industrial process sequence. Hotmelt screen-printed cells are compared to conventional printed ones. Cell results with different sheet resistance emitters and a reduced contact width are discussed. 6.3.1 Monocrystalline silicon solar cells Solar cells were processed on 270 µm thick boron-doped Cz-silicon wafers having a base resistivity of 3 Ω cm to 6 Ω cm and a size of 12.5 cm x 12.5 cm. The process sequence is illustrated in the flow diagram of Fig. 6.9. These cells exhibit a textured surface with an emitter sheet resistance of Rsh = 40 Ω/sq covered by a 125 x 125 mm² Cz-Si wafer Drying Wet chemical texturing & cleaning Screen printing of front grid (Ag hotmelt paste) POCl3 diffusion, PSG etch, cleaning Contact firing (fast firing single wafer furnace) Sputtered SiNx:H antireflection coating Edge isolation (laser scribing + cleaving) Screen printing of rear contact (conv. Al paste) Characterization Fig. 6.9: Process flow diagram of screen-printed silicon solar cells using hotmelt silver paste. 96 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ sputtered SiNX:H antireflection coating. After conventional aluminum screenprinting and drying of the rear side, the front side of one part of the batch was printed with hotmelt, the rest of the batch with conventional paste. A drying step for the hotmelt printed contacts was not necessary. The wafers were co-fired in a fast firing single wafer furnace under oxygen-rich atmosphere and then edgeisolated by laser scribing and cleaving. With this industrial-like process, hotmelt screen-printed cells with energy conversion efficiencies up to 18.0% and fill factors of 80.2% were fabricated (see Table 6.4). The efficiency of 17.9% averaged over eight cells from the same batch, printed and fired with the same process parameters, points out the reproducibility. For the conventional printed cells, efficiencies up to 17.7% could be achieved. Comparing the mean values, the higher fill factor of about 0.8% absolute for the hotmelt printed contacts compared to the reference cells is mainly attributed to the improved finger conductivity, leading to a lower series resistance of the solar cell. Other parameters as e.g. the emitter sheet resistance and the base material affecting the fill factor are very similar for both types of solar cells due to the same cell structure and the same screen used for printing. The efficiency gain for hotmelt screen-printed solar cells is 0.3% absolute compared to conventional printed cells. Table 6.4: IV-parameters of Cz-Si solar cells with front contacts printed with hotmelt and conventional paste. *Calibrated measurement at Fraunhofer ISE Calibration Laboratory. Voc jsc FF η 2 [%] [mV] [mA/cm ] [%] Cells printed with hotmelt paste Best cell* 624.8 36.0 80.2 18.0 Avg. of 8 cells 621 ± 1 36.0 ± 0.1 80.0 ± 0.3 17.9 ± 0.1 Cells printed with conventional paste Best cell 620.7 36.0 79.2 17.7 Avg. of 4 cells 621 ± 1 35.8 ± 0.2 79.2 ± 0.1 17.6 ± 0.1 Cz-Silicon 6.3.2 Multicrystalline silicon solar cells Multicrystalline silicon solar cells with a size of 12.5 cm x 12.5 cm, a thickness of 250 µm and a base resistivity of 0.5 Ω cm to 2 Ω cm, p-doped, were processed like the mono-crystalline cells except that wet chemical texturing was replaced by alkaline saw damage etch. Efficiencies up to η = 15.6% and fill factors of Screen-printing of hotmelt silver paste 97 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 6.5: IV-parameters of untextured multicrystalline silicon solar cells with front contacts printed with hotmelt and conventional paste. The wafers were fired under oxygen-rich atmosphere. mc-silicon Voc [mV] jsc [mA/cm²] FF [%] η [%] Cells printed with hotmelt ink Best cell 616.8 31.3 80.6 15.6 Avg. of 6 cells 616 ± 1 31.5 ± 0.1 79.9 ± 0.5 15.5 ± 0.1 Cells printed with conventional ink Best cell 618.0 31.8 78.8 15.5 Avg. of 4 cells 618 ± 1 31.8 ± 0.1 78.7 ± 0.2 15.5 ± 0.1 FF = 80.6% were achieved for solar cells which were fired under an oxygen-rich atmosphere. In a further batch, the gas atmosphere (nitrogen to oxygen) ratio in the firing process was varied. The IV-parameters of these cells are presented in Table 6.6. Fill factors up to 80.4% were measured when firing under oxygen-rich atmosphere, resulting in efficiencies up to 15.8%. However, the fill factor dropped significantly, when the contact formation atmosphere had a nitrogen to oxygen ratio similar to room ambient [113]; fill factors just over 77% were obtained. This is due to a high contact resistance ρc in the range of 6 mΩ cm² to 14 mΩ cm². For the type of hotmelt paste (Id: CN33-473) printed, the contact formation process is different for textured mono- and non-textured multicrystalline silicon solar cells. The formation process depends mainly on the emitter doping, surface treatment, the SiN deposition technology and also on the paste’s etching behavior regarding different crystal orientations. Table 6.6: IV-parameters of 12.5 cm x 12.5 cm non-textured multicrystalline silicon solar cells, which were fired under different nitrogen to oxygen ratios. η N2:O2 ratio Voc [mV] jsc [mA/cm2] FF [%] [%] 3:1 1:1 1:3 616.9 616.8 615.9 31.7 31.8 31.9 77.3 79.7 80.2 15.2 15.6 15.8 98 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6.3.3 Variation of the emitter sheet resistance In a subsequent batch textured monocrystalline Cz-silicon solar cells of size 12.5 cm x 12.5 cm, a thickness of 250 µm and a sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq were processed in an industrial production line including PECVD SiN antireflection coating. The rear side was screen-printed at Fraunhofer ISE with conventional aluminum paste and dried; the front pattern was screenprinted with silver hotmelt paste and co-fired in a standard inline fast firing belt furnace. Finally, the wafers were edge isolated by laser scribing and cleaving. High energy conversion efficiencies up to 18.0% for solar cells with an emitter sheet resistance of Rsh = 55 Ω/sq and up to 17.8% for Rsh = 40 Ω/sq have been achieved (Table 6.7). The average efficiency of 17.9% for 12 solar cells processed under the same conditions states the reproducibility of the process on a high level. Comparing the IV-parameters of the solar cells with the two different emitters, the open-circuit voltage nearly remains the same, whereas the high fill factor FF of 79.4% drops by about ∆FF = 0.5% absolute to FF=78.9% for the less doped emitter. This is due to the increased contact and emitter sheet resistance, increasing the series resistance rs of the solar cell by about 0.1 Ω cm². However, the shortcircuit current density jsc increases by ∆jsc = 1.7% (relative) as a result of the improved internal quantum efficiency in the short wavelength region. This significant increase in jsc surpasses the loss in fill factor and results in an efficiency increase of ∆η = 1.1% (relative). Table 6.7: IV-parameters of 12.5 cm x 12.5 cm Cz-Si solar cells with front contacts printed with hotmelt paste featuring an emitter sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq. Cz-Silicon Voc [mV] jsc [mA/cm2] FF [%] η [%] 40 Ω/sq emitter sheet resistance Best cell 621.0 35.7 80.4 17.8 Avg. of 12 cells 619 ± 1 35.9 ± 0.1 79.4 ± 0.4 17.7 ± 0.1 55 Ω/sq emitter sheet resistance Best cell 621.0 36.5 79.5 18.0 Avg. of 12 cells 620 ± 1 36.5 ± 0.2 78.9 ± 0.3 17.9 ± 0.1 Screen-printing of hotmelt silver paste 99 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6.4 Fine-line printing and high sheet resistance emitter 6.4.1 Motivation To further increase the efficiency of screen-printed silicon solar cells, the opencircuit voltage Voc and the short-circuit current density jsc need to be further improved. Focusing on the front side, this can be achieved by using less doped emitters. However, due to less doping the series resistance rs of the solar cell rises resulting in a reduced fill factor FF. One effect constitutes the increasing contribution of the emitter sheet resistance to the total series resistance. Compensation can be achieved by reducing the finger separation distance. But at the same time the line width needs to be reduced in order to keep shading losses low. The more significant effect is the difficulty to form an adequate electrical contact to the emitter. The contact resistance between finger and emitter rises and is the limiting factor at a certain doping level [114]. To utilize the advantages of a lowly doped emitter, the formation of selective emitters could be an alternative to avoid contact resistance problems [115,116]. Nevertheless, this leads to additional process steps. Another approach is the optimization of the overall metallization process to form a contact of good quality to lowly doped emitters. Within the last years, the standard sheet resistance emitter in industry increased from about Rsh = 40 Ω/sq to currently Rsh = 55 Ω/sq, which is also due to an improvement in the front side metallization paste. Solar cells with a size of 12.5 cm x 12.5 cm and a base resistivity ρb of 0.5 Ω cm - 2 Ω cm were processed on 270 µm thick boron-doped Cz-silicon wafers according to the process flow shown in Fig. 6.9. The solar cells were wet chemically textured, diffused in a POCl3 diffusion furnace and the front surface was covered by a sputtered SiNx:H antireflection coating. After the rear side had been conventionally Al screen-printed and dried, the front side was printed with hotmelt paste. Finally the wafers were co-fired in an inline fast firing furnace and edge isolated. For high sheet resistance investigations, an emitter variation of Rsh = 40 Ω/sq, Rsh = 55 Ω/sq and Rsh = 75 Ω/sq was performed. The front side metallization pattern consisted of three 5 cm x 5 cm grids with a finger width wf of 70 µm, 80 µm and 100 µm as well as test structures for optical and electrical 100 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ investigations. This included structures for “fine-line printability” as well as contact and line resistance measurements. For each finger width an optimum cell design was calculated, expecting a fairly good contact resistance and line conductivity. Due to processing problems, the final wafer thickness of the solar cells with a sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq was 200 µm, compared to 250 µm for cells with Rsh = 75 Ω/sq. 6.4.2 Contacting high sheet resistance emitter Representative IV-parameters for each type of emitter doping level are presented in Table 6.8. The fill factor declines from a value of about 80% for cells with an emitter sheet resistance of Rsh = 40 Ω/sq to about 78% for Rsh = 75 Ω/sq. Corescan [117,118] and contact resistivity measurements were performed to compare the voltage drop due to the emitter sheet resistance as well as the contact resistance between the finger and the emitter. Table 6.8: IV-parameters of processed 5 cm x 5 cm Cz-silicon solar cells printed with different sheet resistance emitters. The contact fingers of about 100 µm width were printed with hotmelt paste. The sheet resistance was varied. Rsh thickness Voc jsc FF [µm] [mV] [mA/cm2] [%] [Ω/sq] 40 55 75 200 200 250 620.6 621.7 620.3 34.7 35.0 36.3 79.9 79.6 78.0 η [%] rs [Ω cm²] 17.2 17.3 17.6 0.5 0.5 0.7 Contact resistance measurements showed that for both, the cells with an emitter sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq, the contact resistivity between emitter and finger is sufficiently low and hence not to influence the series resistance significantly. Values in the range of ρc=2 - 6 mΩ cm² were measured. The contact resistivity for the cells with an emitter sheet resistance of Rsh =75 Ω/sq was in the range of ρc = 9 - 15 mΩ cm². This relatively high value has certainly an effect on the series resistance of the solar cell. The low voltage Voc also indicates a j02 problem. Nevertheless, fill factors up to 78% could be achieved. The Corescan measurements showed that for all cells the contact and sheet resistance were homogeneous over the cell surface (see Fig. 6.10-a). As illustrated in Fig. 6.10-b, the voltage potential between the fingers increases with higher emitter sheet resistance. On the one hand this is due to the increased voltage Screen-printing of hotmelt silver paste 101 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Rsh = 40 Ω/sq Voltage Rsh = 55 Ω/sq parabola due to sheet resistance jump due to contact resistivity Rsh = 75 Ω/sq a) Length finger Fig. 6.10: a) Potential map of the front surface over the length of four fingers, measured by Corescan. b) Corescan measurement of half the cell size for an emitter sheet resistance of Rsh = 40 Ω/sq, Rsh = 55 Ω/sq and Rsh = 75 Ω/sq. Reflection, IQE parabola caused by the emitter sheet resistance; on the other hand it is caused by the increased voltage jump at the finger edge due to the contact resistivity. The higher current jsc for the cells with Rsh = 55 Ω/sq compared to the ones with Rsh = 40 Ω/sq can be explained by the improved internal quantum efficiency in the short wavelength region as illustrated in Fig. 6.11. The same is valid for the cells with Rsh = 75 Ω/sq. The higher IQE in the long wavelength region of these cells is due to the thickness of the solar cell. 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 IQE Rsh = 75 Ω/sq Rsh = 55 Ω/sq Rsh = 40 Ω/sq Reflection 400 600 800 1000 1200 Wavelength [nm] Fig. 6.11: IQE and reflectance data of solar cells with a sheet resistance of Rsh = 40 Ω/sq, Rsh = 55 Ω/sq and Rsh = 75 Ω/sq. 102 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6.4.3 Fine-line screen-printing An advantage of printing thinner lines is the ability to reduce the shaded area and/or to increase the finger density. This increases the current density and reduces power losses caused by the emitter sheet resistance, respectively (see Chapter 3.2.1). The main challenges are printing continuous lines with a high aspect ratio as well as a low contact resistance to the emitter surface, as the total contact area is typically reduced. Printed lines with a width wf between 50 µm and 120 µm were optically and electrically characterized. With decreasing width, also the height hf of the finger declines as illustrated by the measured cross sections in Fig. 6.12-a. As illustrated in Fig. 6.12-b, the line resistance increases slowly down to a printed width wf of about 80 µm. Below this value the risk of line interruptions rises significantly, causing the line resistance for those fingers to grow strongly. However, down to a width of 80 µm, the line resistance is less than for a typical conventional printed finger with a width of 120 µm. 50 conventional paste hotmelt paste 20 a) conventional printed contact 40 Rline [Ω/m] Finger height hf [µm] 40 30 20 10 0 -50 -25 0 25 Finger width wf [µm] 0 50 b) 70 80 90 100 110 120 Finger width wf [µm] Fig. 6.12: a) Cross sections of contact fingers with different finger width measured using an optical profilometer. b) Line resistance Rline plotted versus the finger width. As mentioned above, 5 cm x 5 cm Cz-silicon solar cells were processed with different finger widths wf and emitter sheet resistances Rsh. The finger separation distance s was equal to 1.8 mm, 2.0 mm and 2.2 mm for the solar cells with a finger width wf of 80 µm, 100 µm and 110 µm, respectively. From the IV-parameters in Table 6.9, it can be concluded that the fill factor declines with decreasing finger width. Comparing the fill factor for cells with an emitter sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq the difference between cells with 100 µm and 110 µm broad fingers is negligible, whereas the difference to the ones with wf = 80 µm is about ∆FF = 0.5% - 0.6% (relative). The relative difference comparing the line widths for cells with Rsh = 75 Ω/sq are even Screen-printing of hotmelt silver paste 103 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 6.9: Best IV-parameters of processed 5 cm x 5 cm solar cells with different finger widths and emitter sheet resistances. Rsh [Ω/sq] thickness [µm] 40 200 55 200 75 250 width [µm] Voc [mV] jsc [mA/cm2] FF [%] [%] 80 100 110 80 100 110 80 100 110 619 620 620 620 622 619 619 620 620 34.6 34.4 34.2 34.7 35.1 34.5 36.3 36.3 36.1 79.3 79.6 79.8 79.1 79.4 79.5 76.3 78.0 78.3 17.0 17.0 16.9 17.0 17.3 17.0 17.2 17.6 17.5 η larger. The FF difference is 0.4% (relative) between cells with wf = 100 µm and wf = 110 µm broad fingers and ∆FF = 2.6% (relative) for wf = 80 µm and wf = 110 µm. Because the relative difference ∆FF increases with decreasing emitter doping, the limiting factor is the contact resistivity. To fully utilize the advantages of fine-line printing on high sheet resistance emitters, the hotmelt paste needs to be further improved with respect to the contact resistance. 6.5 Chapter summary A detailed investigation of the hotmelt technology for the front side metallization of silicon solar cells has been presented. Despite the different composition of hotmelt paste compared to conventional paste, the rheology of the paste behaves similar at elevated paste temperatures. Optimum temperature settings for the print nest and the screen have been defined. The main advantage of screen-printing hotmelt paste in comparison to conventional paste, is the achievement of high finger aspect ratios, increasing the line conductivity and hence the efficiency of the solar cell. A drying process between prints is not necessary. Efficiencies up to 18.0% on 12.5 cm x 12.5 cm Cz-silicon wafers with an emitter sheet resistance of Rsh = 40 Ω/sq have been achieved, an increase of ∆η = 0.3% absolute compared to conventional printed cells of the same batch. Czsilicon solar cells with a size of 12.5 cm x 12.5 cm partly processed in an industrial 104 Screen-printing of hotmelt silver paste _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ production line achieved efficiencies of 17.8% and 18.0%, with an emitter sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq, respectively. Emitter with a sheet resistance of Rsh = 75 Ω/sq could be successfully contacted with the hotmelt technology. These cells showed a significant increase in the shortcircuit current density in comparison to higher doped emitters. To fully exhibit the advantage of contacting high sheet resistance emitters, fine-line printing of hotmelt paste has been investigated. Printing continuous lines with a width of 70 µm and a height of 20 µm was demonstrated. Further paste optimization need to be performed to achieve lower contact resistance between finger and emitter surface, especially as the doping level of the emitter is reduced. Light-induced plating 105 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 7 Light-induced plating Light-induced silver plating is the key technology to form the highly conductive layer of the proposed two layer contact structure. This plating technique has been optimized for high-efficiency cells since 1992 at Fraunhofer ISE and constitutes a fast method to homogeneously thicken metal contacts on n-doped material. A detailed analysis of the electrochemical process is presented in the first part of this chapter. In the second part, an efficiency improvement of large-area screenprinted silicon solar cells of about 0.3% to 0.4% absolute by light-induced plating will be discussed. The series resistance of screen-printed silicon solar cells is reduced, as the line conductivity of the front side metal conductors is improved. Efficiencies of 18.4% on 2 x 2 cm² screen-printed FZ silicon solar cells and 16.6% on 15.6 x 15.6 cm² industrially processed mc-Si solar cells have been obtained. 7.1 Introduction The efficiency potential of plated fine-line contacts is higher than for screenprinted ones. Plated contacts feature a higher aspect ratio and increased line conductivity (compare Chapter 2.2.10). One distinguishes between two different plating technologies: electroless plating and electro plating. Electroless plating techniques allow the growth of fine electrodes without the requirement of establishing an electrical contact. Since the deposition rate for an electroless plating solution is low, the chemical consumption is high and an appropriate pH and temperature control is needed, an electrolytic plating bath is beneficial [119]. By using the photovoltaic effect of a solar cell, it becomes possible to work with the advantageous electrolytic plating solutions without needing to contact the front side metal grid. The solar cell itself generates the required current. To the best knowledge of the author, the first publication in which the photovoltaic effect of a semiconductor was utilized is presented in a German patent from 1973 by W. Späth [120]: A device with a p-n junction is placed only with the front surface into the plating bath. The p-doped rear side, which is not in contact with the electrolyte of the plating bath, is directly connected to the anode. A photogenerated current is generated when the device is illuminated. Metal cations of the solution are attracted by the n-doped areas and at the same time deposited on it (see Fig. 7.1). In another embodiment, it is proposed to apply an external negative 106 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ potential to the p-doped surface, which in this case, is also immersed into the bath. Metal cations will be deposited on the p-doped regions. If the device is illuminated at the same time, deposition from this p-doped region will be partly oxidized, dissolved into the electrolyte again and deposited on the n-doped regions. In this way the simultaneous deposition of silver on p-doped and n-doped regions is possible by conventional electroplating and photocurrent, respectively. 1: cross section of p-doped semiconductor 2-4: n-doped area 5: light permeable container 6-8: metal layer 9-14: SiO2 area 18: anode 19: metallic suction system 20: metal cations 21: vacuum pump 22: irradiation of light 23: optical device 24: optical filter 25: light source 26: suck direction 27: not covered surface 28: plating bath 29: electrical connector 30-32: p-n junction Fig. 7.1: Schematic drawing of the plating process setup as proposed in the German patent DE 2348182 from 1973 by Späth [120]. First publications in which the use of light-induced plating for solar cell contacts is described, can be found in several patents from the end of the 1970’s to the beginning 1980’s [76,121,122]. Probably the first patent by Durkee was filed on Nov 30th, 1977: “… Plating electrical contacts onto one or more surfaces of a solar cell having an electrical junction therein are accomplished by immersing the device so that platable ions in the electrolyte will be attracted to an oppositely charged surface of the cell…” [121]. In this patent the inventor proposes to first plate the rear side of the p-doped base by conventional electroless or electroplating techniques. In a further step the wafer, positioned in an electroless plating solution is exposed to light, which means that from the rear side some of the prior deposited metal was oxidized in the solution again and re-deposited on the negatively charged contact bars of the n-doped front side. Light-induced plating 107 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ This idea was further refined by Grenon in 1980 [122-124]. He suggests connecting the rear side to an anode immersed in the bath and applying a voltage in-between. This allows simultaneous plating on both sides. The plating rate at the front and rear surfaces are independently controlled by adjusting the light intensity and the external current, respectively. Furthermore it is proposed to deposit a metal on the cell surface (by e.g. evaporation technologies), to form a silicide by sintering and to apply a barrier layer (by e.g. evaporation) overlying the metal silicide prior the plating process. Light-induced silver plating at Fraunhofer ISE is based on this latter process, but has been extensively optimized for plating the front contacts of high-efficiency silicon solar cells since 1992 [84]. 7.2 Principle of light-induced plating In the following a more detailed analysis of the light-induced plating process as used for this work is presented. The chemical reaction in the bath itself is explained very briefly. Further information can be found in technical literature [125-128]. 7.2.1 Characterization of the silver cyanide electroplating bath The investigations are based on a potassium silver cyanide (K[Ag(CN)2]) bath, which was patented in 1940. A schematic of the electroplating process is presented in Fig. 7.2. Silver is oxidized to Ag+ cations from the silver anode and reduced on the negatively charged cathode. A V H2O K[Ag(CN)2] Ag1+ CN1- Anode: Ag+ + CN- AgCN AgCN + KCN K[Ag(CN)2] Cathode: K[Ag(CN)2] AgCN + KCN AgCN Ag++ CN- KCN AgCN Ag-anode Ag(s) e- + Ag+ Side reaction: 2KCN + CO2 + H2O K2CO3 + 2 HCN Ag-cathode + e + Ag Ag(s) K[Ag(CN)2] Fig. 7.2: Schematic of electrolytic plating process for the potassium silver cyanide bath. Silver is oxidized from the anode and reduced at the cathode when an external power is applied. In addition a small amount of carbonate and hydrocyanic acid will be produced. 108 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Due to the reduction of silver cations, the ion concentration at the cathode is reduced resulting in a depletion zone. The compensation of the concentration gradient is achieved by diffusion. At some level, the limiting current density is achieved. A further increase in voltage does not result in a current gain, until the potential of hydrogen reduction is reached. If this threshold current density, which is characteristic for each electrolyte composition, is exceeded, the current contribution beyond the threshold value is used for hydrogen formation (water is decomposed). The silver deposition will then be of poor quality (porous, burnt). Up to the limiting current density the deposition rate rises with increasing current flow. The externally applied current ILIP is controlled, so that the current density jLIP at the cathode of size Ametal_surface is: I LIP = j LIP ⋅ Ametal _ surface 7.1 As mentioned above, if the current density is too high, the deposited metal will be of lower quality and at some point H2 formation will take place at the cathode and/or anode. If the current density is too low the energy is not high enough to break up the silver complex in the bath and no deposition will occur. The deposited mass mAg of silver for optimum current density can be calculated by: m Ag = η bath M Ag I LIP ⋅ t = I LIP ⋅ t ⋅ 1.118 ⋅ mg C −1 z Ag F 7.2 MAg is the molar mass of silver, zAg the oxidation number of silver (number of electrons used for the reduction of one silver ion), I the current, t the plating time, F the Faraday constant and ηbath the efficiency of the bath. The efficiency is defined as the amount of current used for the actual reduction process compared to the current that goes into side reactions as e.g. heat or carbonate production. The efficiency of the used bath at optimum current density is between 98% and 100%. For further calculations an efficiency of 100% is assumed. Applying eq. 7.2, the deposited mass is equal to about 4 grams per ampere-hour. This was experimentally confirmed. Furthermore the height hLIP of the deposition can be calculated by: hLIP = m Ag ρ o Ag ⋅ Ametal _ surface = η bath M Ag ρ o Ag ⋅z Ag j LIP ⋅ t F 7.3 with ρ°Ag the density of silver (10.5 g/cm3). Note that for e.g. a roundish contact the total surface area of the contact need to be considered (and not just the coverage area). Applying eq. 7.3 the height increases at a current density of 1 A/dm² by about 0.67 µm per minute. Light-induced plating 109 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 4 6 3 5 2 4 1 3 Resistance [Ω] 2 Current density [A/dm ] The optimum current density for the silver cyanide plating bath was determined by measuring the current voltage curve between the silver anode and a metal plate. The lowest resistance (Rbath = 2.8 Ω) at a voltage of 310 mV was obtained with a current density of about 1.1 A/dm², which is the optimum density for plating (see Fig. 7.3). Between 0.6 A/dm² and 3.0 A/dm² the current density is also in a good range for plating, as the resistance is still on a low level. minimum 0 0 200 400 600 800 1000 1200 2 Voltage V1 [mV] Fig. 7.3: Current density and bath resistance plotted versus the voltage between anode and cathode for the K[Ag(CN)2] electroplating bath. Measurements were taken 30 seconds after changing the applied voltage. 7.2.2 Plating of solar cell contacts The solar cell’s ability to generate a photocurrent under irradiation is the key factor allowing a light-induced plating process. As the cell is illuminated the ndoped front side is negatively charged (cathode). The positively charged metal cations in the solution are then deposited on the negatively charged surface. To establish a closed electrical system an electrical contact has to be made to the cell. Due to the photovoltaic effect it is sufficient to establish this contact at the rear side. There is no need to contact any of the fine electrodes at the front side, which simplifies the handling process enormously. Another advantage is that the front side under irradiation has the same voltage potential over the entire area, leading to homogeneous plating on the front surface. Standard electroplating processes have homogeneity problems caused by the loss of voltage potential with increasing distance from the contacting point. If e.g. the front grid would be contacted at the end of one busbar, then the thickness of the deposit would decay with increasing distance from the contact. This effect is even 110 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ more pronounced if the contact layer becomes thin, which is desired for the production of high-efficiency cells. The electrochemical reaction of the light-induced plating process was extensively investigated by performing different experiments; some of them will be presented in the following. As illustrated in Fig. 7.4 the rear side was covered by a resist. This was performed to avoid any influence on the measurements (e.g. oxidation or reduction at the rear side) induced by the wide variation of applied potentials and irradiances. For standard processing, covering of the rear side is not necessary, when using the right settings for the applied voltage and irradiance. ϕA ϕR V3 ILIP V2 V1 ϕF Ag+ Ag+ Ag-anode Ag-cathode ϕA: potential anode ϕR : potential rear side ϕF: potential front side Al-BSF p-base finger n-emitter V1: ϕA - ϕF V2: ϕR - ϕF V3: ϕA - ϕR V1 = V2 + V3 ARC resist Fig. 7.4: Schematic of the light-induced plating process. The illuminated front side acts as cathode. For this investigation the rear side of the solar cell is covered by a resist to isolate it from the process. At first, the irradiance at the front surface of a solar cell (with known shortcircuit current density at one-sun irradiance) placed in the electrolytic plating bath, was determined. The short-circuit current density was measured for different light intensities. The irradiance was calculated by comparing the short-circuit current density with the measured values. The current ILIP and the voltages V1, V2, V3 were measured. V2 is the cell voltage, V1 the voltage drop between anode and front electrodes and V3 the voltage drop between anode and rear side electrode (power supply + connector lines). Hence V1 is equal to the sum of V3 and V2. The measurements were taken 15 seconds after the applied voltage was changed. Forward and backward curves were measured, because the thickness of the contacts increased over time, which could affect the measurement. Solar cells with relatively wide and flat contacts were used to keep the relative increase of the contact area low. The solar cells of area Acell = 50 cm² had a metal coverage pc of about 10%. Due to the rough surface of the screen-printed contacts the surface area of the metal Light-induced plating 111 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Ametal_surface could not be exactly determined, but was assumed to be in the range of 5 cm² to 6 cm². Hence, the maximum current for plating assuming a current density of 3 A/dm² is between 150 mA and 180 mA. Fig. 7.5 shows the current ILIP plotted against the cell voltage V2 measured at different irradiation levels. The cell was placed in the plating bath as illustrated in Fig. 7.4. Up to an irradiance of 0.1 suns the maximum current flow is limited by the solar cell itself and thus follows in principle the IV-characteristics of the cell with an increased series resistance. The current flow can not exceed the photo-generated current of the solar cell, unless the applied voltage in reverse bias was higher than the breakdown voltage. However, for the 0.3 suns irradiance, the maximum current flow is limited by the plating bath, since for low voltages V2 the threshold current density is exceeded (compare Fig. 7.3). Current ILIP [mA] 200 current limited by bath (~ 3 A/dm²) 150 100 Suns 0.3 0.1 0.06 0.03 0.01 50 0 -100 0 100 200 300 400 500 600 Voltage V2 [mV] Fig. 7.5: IV-curves of a solar cell with a size of about 50 cm² measured at different irradiation intensities. The solar cell was placed into the plating bath as illustrated in Fig. 7.4. For irradiances up to 0.1 suns the current is limited by the solar cell itself, whereas for higher irradiances the maximum achievable current flow is limited by the plating bath (current density at the contacts too high). The high external load (plating bath) limits the cell performance, which becomes even more crucial with growing cell size. The question now is how to adjust the irradiation intensities and the external power to achieve optimum plating conditions. This will be explained in the following. Fig. 7.6 illustrates the solar cell potential of the rear (V3) and front side (V1) plotted against the current for different irradiances. The potential of the anode is defined as the reference potential. As discussed above, the measurements were performed with the rear side covered by a resist. In this case just the potential between anode and front side defines the plating condition. With increasing voltage difference between anode and front side, the current and hence the plating rate increases, up to the point at 112 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Potential [mV] 600 A 400 200 B 0 -200 D -400 ϕR potential rear side -600 ϕF potential front side -800 0 10 a) 0.03 suns illumination 600 potential anode 20 30 E 40 50 Current ILIP [mA] A current density limit (~ 3 A/dm²) Potential [mV] 400 B 200 C 0 D -200 -400 ϕR potential rear side -600 ϕF potential front side -800 0 b) 0.1 suns illumination Potential [mV] 600 50 E 100 150 Current ILIP [mA] A current density limit (~ 3 A/dm²) 400 200 B 0 -200 -400 -600 -800 potential anode potential anode ϕR potential rear side ϕF potential front side 0 c) 0.3 suns illumination 50 100 150 Current ILIP [mA] Fig. 7.6: Potential of rear- and front side of the solar cell plotted against the current, for 0.03 (a), 0.1 (b) and 0.3 (c) suns irradiance. The potential of anode is defined as reference potential (0 V). For definition of points A to E see Fig. 7.7, which also explains why optimum plating conditions for a solar cell with silver pads on the rear side is achieved for case “C” when the voltage difference between anode and rear side is equal to the one between rear- and front side. which either the solar cell (Fig. 7.6-a, -b) or the plating bath (Fig. 7.6-c) limit the current flow. However, if the rear side (especially the silver pads) is not covered by a resist or some other dielectric layer, plating conditions are different. Using the same measurements as illustrated in Fig. 7.6 different conditions are discussed, assuming the rear side is in direct contact with the electrolyte of the plating bath. The Light-induced plating 113 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ different cases are illustrated in the schematic drawing of Fig. 7.7. This discussion is based on a simplified model, as it does not take into account e.g. the different field intensities on the front and rear side and the different bath resistance at different current flows. Rear ϕΑ V3 V1 V2 Potential ϕR 0V Anode Front Reference potential anode Anode Rear Anode ϕF Anode ϕA = potential anode ϕR = potential rear side ϕF= potential front side V 1 = ϕA – ϕF V 2 = ϕR - ϕF V 3 = ϕA - ϕR Anode Anode Rear Front Front Front Front Rear Front Oxidation process Reduction process Rear Rear Case A Case B Case C Case D Case E Case F Fig. 7.7: Schematic drawing of the potential distribution for anode, rear and front for the different cases discussed in the text. Anode is used as reference potential. Case A (ϕA = ϕF < ϕR): The potential of anode and front side is equal, but on a lower level than that of the rear side. Silver from the rear will be oxidized and reduced at the front side as well as on the anode. Case B (ϕA = ϕR > ϕF): Potential of rear side is equal to the one of the anode. Silver from both, the rear side and the anode will be oxidized and reduced at the front side. Case C (ϕA > ϕR = 1/2 ϕF): The potential of the anode is highest. The voltage difference between anode and rear side is equal to the difference between rear and front side. In this case silver will be oxidized from the anode and partly reduced on the rear side, partly on the front side, whereas the deposition rate on the front side will be greater. However, as the rear side potential is higher than the potential of the front, silver will be also oxidized from the rear and deposited on the front. Hence oxidation and reduction processes of silver on the rear side will have in the ideal case the same value, resulting in exceptional front side silver deposition. 114 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Case D (ϕA > ϕF = ϕR): Rear- and front side are on the same negative potential referred to the potential of the anode. Silver will be reduced on the front as well as on the rear side. Case E (ϕA > ϕF > ϕR): This case is analogue to case “C”, but in this case with the potential of the front side is half the potential difference between anode and rear side. Silver deposition will occur on the rear side, not on the front. Case F (V1>V2>V3): If the potential difference between anode and rear side is two times higher than the potential difference between anode and front side, oxidation processes on the front would exceed reduction processes. The front side electrodes would completely oxidize over time. Thus, optimum plating condition, defined for the case that silver deposition just occurs on the front side whereas the rear side remains unaffected, is achieved for case “C”. The voltage difference between anode and rear side is equal to the one between rear side and front side, as long as the maximum “allowed” current flow is not exceeded. Silver from the anode will be oxidized and deposited on the front side, whereas the rear side is not affected. For a high deposition rate the current flow should be in the upper “allowed” range close to the current limit of the solution. Therefore the irradiance needs to be adjusted carefully. If the irradiance is too low, case “C” will be achieved at fairly low current flows (Fig. 7.6-a); in addition points “C”, “B” and “D” are lying at very similar current flows. Setting the irradiance too high, the necessary current flow to achieve point “C” would be higher than the maximum “allowed” current flow of the bath (Fig. 7.6-c). Hence, a good irradiance is reached when point “C” is at a high current flow, as presented in Fig. 7.6-b. Under real working conditions it is difficult to achieve case “C”, as e.g. it is quite complicated to measure the different voltages, while the solar cell is placed in the bath. To ensure that no silver oxidizes from the rear the settings should be chosen that the potentials are in-between case “C” and “D”. This can be described by the relationship (ϕA > ϕR with ϕR < 1/2 ϕF). The rear side is protected against oxidation, but silver deposition may occur on that side. On the other hand, it might be desired to ensure that silver is deposited only on the front electrodes, while some oxidation of silver from the rear side pads is accepted. In this case settings should be chosen in that way, that the potentials are between case “B” and “C”. This relationship can be expressed by ϕA >> ϕR > ϕF with ϕR < ϕF. In addition it might be beneficial to ensure that the current flow can not exceed the limiting current density, as a too high applied voltage results in no useful Light-induced plating 115 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ deposition of silver whereas a too low voltage “just” reduces the deposition rate. This is possible by choosing the irradiation level XSuns in that way, that the shortcircuit current of the solar cell is equal or slightly below the maximum “allowed” current for deposition ILIP_max (Fig. 7.6-b). This irradiance can be calculated by: X Suns = j LIP _ max ⋅ Ametal _ surface j sc ⋅ Asolar _ cell = I LIP _ max 7.4 I sc with Ametal_surface being the area of the contacts surface, Acell the cell area, jLIP_max the maximum current density of the plating bath and jsc the short-circuit current density of the solar cell. In summary, good plating condition can be achieved by following this proposed six step sequence: 1. Defining the desired deposition height. 2. Determining the metal surface area of the front side grid. 3. Calculating the optimum total current flow ILIP (eq. 7.1). 4. Calculating the mass, that is necessary to obtain the desired deposition height (eq. 7.3). 5. Calculating the plating time (eq. 7.2). 6. Adjusting the irradiance, so that the voltage drop between rear side and anode is slightly above half the voltage drop between anode and front side for the optimal current flow ILIP. 7.2.3 Plating without external power supply It would be beneficial, if the required current for plating is solely produced by the solar cell itself without requiring an external power supply. Therefore, in a further investigation the rear side (covered by a resist) was directly connected to the anode by short-circuiting the power supply in the schematic of Fig. 7.4. The IVmeasurements for different irradiation intensities are illustrated in Fig. 7.8. The current rises with increasing irradiance and silver is deposited on the contacts. The current flow can be adjusted via the irradiation intensity. However, as the rear side is always on a more positive potential than the anode, a direct contact with the electrolyte of the plating bath should be avoided. Otherwise silver from the rear would be oxidized. In spite of the expectations, the current flow does not rise linearly with the irradiance. This can be explained by the external load of about 3.5 Ω, which is the 116 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Irradiance [suns] Potential ϕ [mV] 0 0.015 0 0.04 0.2 0.4 0.09 potential anode -100 -200 ϕR (potential rear side) -300 ϕF (potential front side) 0 20 40 60 80 Current [mA] 100 120 140 Fig. 7.8: Current plotted versus voltage V3 (anode - rear) and V1 (anode - front). The rear side contact of the solar cell is directly connected to the anode. With increasing irradiance the current flow increases. But also the voltage drop across the plating bath and the connector lines rises, which will limit the current flow of the solar cell at elevated irradiance due to the high external load. sum of bath resistance and resistance of the connector lines. This load limits the cell performance as the simulation in Fig. 7.9 illustrates. A simulation of solar cell IV-curves for different irradiances is presented assuming an external load of 3.5 Ω and a cell size of 43 cm². On each IV-curve the point is marked at which the solar cell actually operates, illustrating that for high irradiances the external load limits the maximum achievable current flow. 250 working point 0.4 suns 0.28 suns Current [mA] 200 0.17 suns 150 0.09 suns 100 0.04 suns 50 0.015 suns 0 0 100 200 300 400 500 600 Voltage [mV] Fig. 7.9: Simulation of IV-curves based on the two diode equation with a high external load of 3.5 Ω, which is equal to an area weighted series resistance of about 150 Ω cm² for a cell with an area of about 43 cm². The crossed dots mark the actual measured values. Due to the high external load, the cell performance is poor, limiting the current for increasing irradiances. Light-induced plating 117 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Especially in this “power supply free” mode, the resistance of the connector lines should be low and the conductivity of the plating bath high to keep the external load as small as possible. Nevertheless, the rear side needs to be protected for this operating mode and the deposition rate will be limited by the “high” external load, especially for large-area silicon solar cells. The use of an external power supply is recommended. 7.2.4 Plating of n-type silicon solar cells The light-induced plating process as described above would not work for n-type silicon solar cells. Under irradiation the front side is always on a more positive potential compared to the rear side and/or anode. Selective plating on the front side is not possible. For n-type silicon solar cells either the front side electrodes need to be contacted or an electroless plating solution has to be used. However, under certain conditions, applying a contact to the rear side and plating on the front side is possible. When the solar cell is not illuminated and a negative potential is applied to the rear side, rear- and front side would be on the same negative potential (cell is in forward bias) compared to the potential of the anode. Silver would be oxidized from the anode and homogeneously reduced on the front- as well as on the rear side. If the deposition of silver on the rear is not desired, either the rear needs to be protected by a resist or it has to be ensured from the machine side, that the rear side is not in direct contact with the electrolyte of the plating solution. Rear Ni Rear Ni potential anode 500 mV Front Ag Potential 1050 mV 550 mV 0V Anode Ag Front Ag Anode Ag Anode Ag 400 mV Rear Ni Oxidation process 500 mV Reduction process dark Front Ag illumination 1.45 V externally applied + illumination Fig. 7.10: Potential of the silver anode and of a nickel rear and silver front side of an n-type silicon solar cell. If the cell is illuminated and a voltage externally applied, silver deposition mainly on the front side electrodes would be possible. 118 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Another theoretical solution for selective (light-induced) plating of the front side electrodes based on case C/D in Fig. 7.7 is illustrated in the schematic drawing of Fig. 7.10. The rear side surface could be completely covered by another material as e.g. nickel. The difference in the electrochemical potential between nickel and silver is around 1.05 V. If the n-type solar cell is illuminated achieving a cell voltage of e.g. V2 = 550 mV, the front side would be on a 500 mV higher potential than the rear side. If in addition a voltage of V3 = 1.45 V between anode and rear side is externally applied, the potential of the rear side would be slightly lower than half the potential of the front side. The nickel rear side would be covered by a thin layer of reduced silver protecting it against oxidation, while main silver deposition would occur on the front side electrodes. 7.2.5 Plating of non-silver contacts For light-induced plating the seed layer does not need to be made up of silver. To achieve a low contact resistance to the n-doped silicon surface, other contact materials as e.g. nickel could be better suited (compare Chapter 5.2.2). However, as the electrochemical potential of nickel is XNi=-0.25 V and the one for silver is XAg=+0.80 V, the voltage between silver anode and nickel front side must be at least 1.05 V already when the solar cell is placed into the electrolyte. Otherwise nickel would be oxidized from the front and reduced at the silver anode. As soon as a layer of silver is deposited on the front, the voltage between front and anode needs to be reduced. To ensure that the correct voltage is always applied, the external power supply should be current-controlled. 7.3 Light-induced plating of screen-printed contacts 7.3.1 Setup A simplified schematic of the light-induced plating process as used for this work is illustrated in Fig. 7.11. A solar cell with screen-printed silver front contacts and an aluminum rear side metallization including silver pads is placed in the electroplating bath. Electron-hole pairs are created under irradiation. The emitter is on a negative potential compared to the anode and attracts positively charged metal cations of the solution; the existing metal layer is thickened by the plating process. As the potential over the whole front side area is homogeneous, the silver is Light-induced plating 119 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ reduced quite uniformly on top of the contacts, even if the initial metal layer is extraordinary thin. To protect the silver pads against oxidation, the rear side is connected by an external power supply to the metal anode (see Chapter 7.2.2). The specific finger conductivity was extracted by measuring the finger resistance with the four point probe technique and its geometry with an optical profilometer. Thus the conductivity of the deposited silver was determined, by comparing the specific finger conductivity before and after the plating process. The resistivity of 1.9 x10-8 Ω m of the plated silver matches nearly the one of the bulk silver (1.6 x10-8 Ω m) and is significantly lower than the resistivity of about 3.2 x10-8 Ω m for a screen-printed contact. Fig. 7.11: Scheme of the light-induced plating process. The positively charged metal cations are deposited onto the negative n-doped front surface (cathode), induced by the photovoltaic effect under irradiation. 7.3.2 Small-area monocrystalline silicon solar cells Due to the long-term experience of light-induced plating in the field of thickening evaporated Ti-Pd-Ag contacts of high-efficiency silicon solar cells [84], a similar wafer layout was used to transfer the process to screen-printed contacts. Seven solar cells with a size of 2 x 2 cm² were processed on four inch 0.5 Ω cm pdoped FZ wafer. The wafers were wet chemically textured, had a POCl3 emitter wit a sheet resistance of Rsh = 55 Ω/sq and Rsh = 40 Ω/sq, a SiNX antireflection coating and a screen-printed aluminum rear side. Fine metal lines were screenprinted on the front side using a screen with 60 µm mesh openings for the fingers 120 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ and hotmelt paste. Finally the wafers were co-fired in a fast firing belt furnace under air ambience. The metal fingers of about 70 µm width and a 6 µm height were thin and flat resulting in a high line resistance and hence series resistance, limiting the fill factor to values of about 60% to 70%. The contact resistivity was determined to be in the range of 2 mΩ cm², not influencing the series resistance significantly. After the light-induced plating step, the series resistance was considerably reduced. The best solar cells with a sheet resistance of 40 Ω/sq and 55 Ω/sq had high efficiencies of 18.2% and 18.4%, respectively (Table 7.1). Table 7.1: IV-parameters of best 2 cm x 2 cm FZ-silicon solar cells with fine-line screen-printed front contacts thickened by light-induced plating. The best cell with an emitter sheet resistance of 40 Ω/sq and 55 Ω/sq had a finger separation distance s of 2 mm and 1.67 mm at a firing temperature of 790°C and 810°C, respectively. The series resistance was measured comparing one-sun and dark IV-curve (see Chapter 4.1.2). Rsh [Ω/sq] firing temp. finger distance 40 790°C 2 mm 55 810°C 1.67mm process step Voc jsc [mV] [mA/cm2] FF [%] [%] rs [Ω cm²] prior LIP 633.2 36.0 63.6 14.5 >2 after LIP 634.2 35.7 80.4 18.2 0.5 prior LIP 635.6 36.4 65.6 14.5 >2 after LIP 634.5 35.8 80.9 18.4 0.5 η In this batch the front side grid of the solar cells were printed partly with a finger separation distance of 1.67 mm, partly with 2 mm. Whereas the efficiency for the cells with a sheet resistance of 40 Ω/sq is highest for a separation distance of 2 mm, for the cells with an emitter sheet resistance of 55 Ω/sq the smaller finger separation distance showed the better performance. The gain due to less ohmic losses in the emitter for these cells overcompensates the losses in current due to the increased shaded area of about 0.7% absolute. The short-circuit current density for the less doped emitter cell is still on the same level compared to the cell with Rsh = 40 Ω/sq as a result of the improved internal quantum efficiency in the shortwavelength region, despite the increased shaded area. The high fill factor for cells with Rsh = 55 Ω/sq is due to the reduced separation distance. However, the open circuit voltage could not be improved. Fig. 7.12 illustrates the IV-parameters of the best cells plotted against the peak firing temperature. While the efficiency for the cells with Rsh = 40 Ω/sq has its Light-induced plating 121 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 35.5 Rsh=40 Ω/sq 35.0 Rsh=55 Ω/sq 780 790 800 810 820 830 34.5 18.4 18.2 18.0 17.8 17.6 Firing temperature [°C] 82 81 80 780 790 800 810 820 830 79 Fill factor [%] 36.0 Efficiency [%] 640 635 630 625 620 jsc [mA/cm²] Voc [mV] maximum at a temperature of 790°C, the optimum firing temperature of 810°C for the emitter cells with Rsh = 55 Ω/sq seems to be slightly higher. Firing temperature [°C] Fig. 7.12: IV-parameters after light-induced plating plotted versus the firing temperature for the best solar cells. Highest efficiency is achieved at 790°C for cells with an emitter sheet resistance of 40 Ω/sq and at 810°C for the one with Rsh = 55Ω/sq. 7.3.3 Large-area multicrystalline silicon solar cells Standard production solar cells: Further optimization of the plating process for screen-printed contacts was carried out using standard production solar cells of a modest material quality from a solar cell manufacturer. The processed 15.6 x 15.6 cm² sized multi-crystalline cells had a low open-circuit voltage, whereas the fill factor was in the usual production range. For plating, especially the rear side potential needed to be adjusted. If the applied potential is too low, the silver-pads on the rear side will be oxidized, whereas if it is too high deposition of silver might also occur on the rear, which is not desired either (see Chapter 7.2.1). Table 7.2: IV-parameter of five 15.6 x 15.6 cm² standard production multi-crystalline silicon solar cells before and after the light-induced plating process (LIP). jsc [mA/cm2] FF [%] η Process step Voc [mV] Before LIP 600 ± 1 31.9 ± 0.1 76.5 ± 0.3 14.6 ± 0.1 After LIP 600 ± 1 31.8 ± 0.1 78.1 ± 0.3 15.0 ± 0.1 [%] The efficiency of the solar cells could be increased by 0.4% absolute due to the increase of the fill factor from 76.5% before to 78.1% after a short plating time (see Table 7.2). Fig. 7.13 shows a cross section of a screen-printed contact plated with silver and Fig. 7.14 a typical deposition rate used for the silver bath. 122 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Height hf [µm] 20 15 10 5 0 0 5 10 15 Plating time [min] Fig. 7.13: Cross section of a screen-printed Fig. 7.14: Deposited height plotted against contact plated with silver at Fraunhofer ISE. the plating time. Deposition rate used was (Image – courtesy of Rohm and Haas Electronic Materials) about 1.1 µm per minute. Fine-line screen-printed solar cells: In a subsequent experiment solar cells were processed in the same way as standard production cells, except that the screen-printed fingers were thinner, having a width of about 70 µm and 90 µm. The amount of fingers printed was not changed. The small cross section of the fingers reduced their conductivity, leading to a higher series resistance (rs > 1 Ω cm²), which resulted in fill factors of about 76% for the 90 µm broad fingers and 71% for the 70 µm broad ones (Table 7.3). The contact resistivity ρc for both types of metal grid was between 2 mΩ cm² and 3 mΩ cm². As presented in Fig. 7.15 the maximum efficiency increase for cells with 90 µm and 70 µm fine fingers was reached after silver plating of 6 to 10 minutes and 8 to 12 minutes, respectively. While the open-circuit voltage stays constant over the plating time, the fill factor rises sharply first and then saturates, because the Table 7.3: IV-parameters before and after the light-induced plating process of 15.6 x 15:6 cm2 mc-silicon solar cells. Presented are the best value and the mean value of ten cells for each group. The plating time was 6 to 10 minutes for the cells with 90 µm and 8 to 12 minutes for ten with 70 µm finger width. Process step Voc [mV] jsc [mA/cm2] FF [%] η [%] Cells with 90 µm finger width Prior LIP 612 ± 1 33.6 ± 0.1 75.9 ±0.4 15.6 ± 0.1 After LIP 612 ± 2 33.2 ± 0.1 78.5 ±0.5 16.0 ± 0.1 Cells with 70 µm finger width Prior LIP 611 ± 1 33.9 ± 0.2 71.0 ±0.6 14.7 ± 0.1 After LIP 613 ± 1 33.4 ± 0.2 77.5 ±0.8 15.9 ± 0.1 rs [Ω cm²] 1.2 0.5 2.3 0.6 Light-induced plating 123 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ influence of the electrical loss of the front grid on the total series resistance becomes negligible. The maximum efficiency gain is found where the sum of relative current loss and fill factor gain is maximal. FF Relative increase [%] Relative increase [%] 10 FF 4 η 2 Voc 0 -2 0 jsc 2 4 6 8 10 12 14 16 8 η 6 4 2 Voc 0 -2 0 Plating time [min] jsc 2 4 6 8 10 12 14 16 Plating time [min] Fig. 7.15: Graphs showing the relative gain of the IV parameters over the plating time for cells having 90 µm (right-hand) and 70 µm wide fingers (left-hand). Note the different scale. Each data point represents the mean value of three measurements. The plating rate was about 1 to 1.2 µm per minute. Solar cells with an optimized front side structure for LIP: Screen-printed solar cells were processed with a front side grid optimized in finger width and separation distance for the plating process, which means optimized finger distance and width. These cells had a high series resistance and would not be suitable for the standard cell process. Table 4 shows the IVparameters before and after plating. Due to the rise of the fill factor from 74% to 78%, the efficiency could be increased by 0.7% absolute from 15.6% to 16.3%. The best solar cell processed, reached an efficiency of 16.6% with a fill factor of 79.2%. Table 7.4: IV-parameters of 15.6 x 15.6 cm² screen-printed mc-Si solar cells before and after the light-induced plating (LIP) process with an optimized grid design. η Cz-Silicon Voc [mV] jsc [mA/cm2] FF [%] [%] Before LIP Best cell Avg. of 9 cells 612.1 34.6 74.9 15.9 610±2 34.5±0.1 74.3±0.6 15.6±0.2 613.4 34.1 79.2 16.6 611±2 34.1±0.1 77.9±0.9 16.3±0.2 After LIP Best cell Avg. of 9 124 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 7.3.4 Cyanide-free plating for industrial application Due to its toxicity, the acceptance of electroplating solutions containing cyanide in the photovoltaic industry is low, which was regarded as entrance barrier for the light-induced plating process. Thus a recently developed cyanide-free bath from Rohm and Haas Electronic Materials was tested and compared to the cyanide containing solution. As presented in Table 7.5 similar efficiencies were achieved for both plating solutions. However, for these first experiments a sinter step (300°C under air ambience) at the end of the process was necessary to achieve the same efficiency level compared to the cells plated from the cyanide-containing bath. Table 7.5: IV-parameters of best 15.6 x 15.6 cm² multi-crystalline silicon solar cells using a cyanide containing and a cyanide free plating solution. The finger width before plating was 70 µm. Process step Voc [mV] jsc [mA/cm2] FF [%] η [%] cyanide containing bath before LIP 611.0 after LIP 613.7 cyanide free bath before LIP 612.3 after LIP 613.1 after sintering 614.2 33.7 33.3 71.1 78.2 14.6 16.0 33.8 33.3 33.6 70.9 77.1 78.2 14.7 15.8 16.1 Further process optimization could make the sinter step also for the cyanide-free bath unnecessary. These promising results led to the development of a horizontal inline lightinduced plating machine in cooperation with a solar cell manufacturer, a chemical supplier (Rohm and Hass Electronic Materials) and an inline equipment fabricator (Schmid GmbH). This machine is currently used for a large scale experiment at the solar cell manufacturer site. Under industrial environment 0.3% to 0.5% absolute efficiency gain compared to standard production cells were achieved as will be reported by Allardyce [129]. Light-induced plating 125 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 7.16: Inline light-induced plating machine developed in cooperation with a solar cell manufacturer (Q-Cells), an inline equipment fabricator (Schmid GmbH) and a chemical supplier (Rohm and Haas). 7.4 Direct light-induced plating of silver onto silicon surface Direct plating of silver onto the silicon surface would have the benefit that no metal seed layer would be necessary and the overall process complexity would be reduced. Such direct electroplating of silver on a moderately doped emitter (Rsh = 40 Ω/sq - 120 Ω/sq) has not been achieved so far. However, when structuring the surface in a special way direct plating is possible. Prior to emitter diffusion trenches with different depths were cut into the silicon surface by a dicing saw. In each case two trenches were made adjacent to each other, such that a sharp edge in-between was formed. The wafer was placed into the electrolytic plating bath and the rear side connected over an external power supply to the anode (compare Fig. 7.4). Different settings for the light-induced plating process were tested. The plating process seems to be enhanced by the field density at sharp edges which is higher than on a plane surface [130]. In addition to the applied irradiance, the geometry of the trenches is of major importance. When the depth of the trenches is too low, no plating occurs (Fig. 7.17-a). When the trenches are too deep, not only the edge in-between the trenches is plated, but also the opposite edge as illustrated in Fig. 7.17-b. Optimum plating could be achieved at a depth between 50 µm and 70 µm and an angle of 60°. The sharper the edge, the lower the depth needs to be, to enhance plating, but the higher is the danger of edge breakage. Adhesion of plated silver after a sintering step was of high quality. 126 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 7.17: SEM images of a silicon surface with trenches. Due to higher field intensities at the sharp edge between two adjacent grooves plating is enhanced. a) No plating occurred at the edges as the grooves were not deep enough. b) All edges were plated, even edges with a small plateau. The trenches were too deep or the field intensity too high. c) Thick plated silver line with good adhesion due to a sintering step after a thin layer has been deposited. d) After sintering the adhesion to the silicon is good. When performing adhesion tests, if at all, the edge of the silicon is ripped off; the silver still adheres to the silicon. When performing adhesion tests, the whole edge broke, but the silver still adhered to the silicon as illustrated in Fig. 7.17-c. However, experiments have shown that it might be necessary to sinter the wafer after a thin layer of silver has been deposited to achieve good adhesion already at this point. Otherwise with increasing plating thickness the adhesion of the silver to the silicon surface might not be sufficient and the silver will come off during the plating process. If a sinter step was carried out, a first metal other than silver could be used in order to achieve a lower contact resistance to the emitter surface. The thickness of the contacts can be varied as a function of the plating time. Fig. 7.17-d shows a contact with a cross section of about 40 µm achieved after 15 minutes plating time. Light-induced plating 127 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ A further advantage of this contact structure is that the contact geometry is roundish. As illustrated in Fig. 1.9 of Chapter 1.2.1 a roundish contact has the advantage that a considerable amount of the incident light hitting the metal surface will be reflected into the solar cell and can be used for electron–hole pair generation. The “W” shaped trenches can be made using a laser system with e.g. a double beam or as performed within this work mechanically using a dicing saw. Also the use of a laser chemical etch process could be suitable with the advantage that the laser induced damage of the trench formation process is directly removed [131]. Difficulties for direct plating arise due to the fact that as the thickness of industrial processed wafers is reduced continuously, the depth of the trenches should be certainly as low as possible to guarantee an adequate stability. However, as discussed above, a certain minimum depth is necessary to initialize the plating process. Fig. 7.18 shows a possible process flow for direct plating in industrial production which – as already mentioned – has not been done up to date. formation of trenches (laser) wet chemical texturing (also laser induced damage is removed) antireflection coating light-induced plating rear-side passivation (e.g. SiC, SiO2) sinter and annealing step emitter diffusion & PSG etch Al sputtering rear-side light-induced plating laser fired contacts Fig. 7.18: Possible process flow to manufacture solar cells with directly plated contacts. 7.5 Chapter summary The light-induced plating process is used to form the second layer of the proposed two layer contact structure for the front side metallization of silicon solar cells. The basic working principle of the plating process is based on the photovoltaic effect. The front side of an illuminated solar cell, which is placed in a silver electrolytic plating bath, is on a more negative potential compared to the silver anode, which is connected to the rear side of the wafer. Positive metal cations of the solution are reduced at the front side contacts serving as cathode. Due to theoretical as well as practical investigations the chemical reaction in the bath is well understood and optimum plating conditions were found. The lightinduced plating process was successfully transferred from small-area high- 128 Light-induced plating _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ efficiency cells with evaporated contacts to large-area industrial processed cells with screen-printed contacts. For first trials a similar wafer layout as for the highefficiency cells was used. Efficiencies up to 18.4% on 2 x 2 cm² FZ solar cells were achieved. These cells feature fine-line screen-printed and plated contacts, an emitter sheet resistance of 55 Ω/sq and a full metallized Al-BSF. The light-induced plating process was further adapted for plating the contacts of large-area screenprinted solar cells. On industrial processed multicrystalline silicon solar cells of size 15.6 x 15.6 cm² efficiencies up to 16.6% were achieved after plating. Experiments have shown that an efficiency increase of 0.3% to 0.5% for screenprinted solar cells in industrial environment is realistic. In addition, cost calculations have revealed that the significant increase of efficiency combined with the reduction of the amount of screen-printing paste required, overcompensates the cost for this additional step at the end of the process sequence. These results led to the development of a horizontal inline light-induced plating machine together with industrial partners (Q-Cells, Rohm and Haas Electronic Materials and Schmid). This machine, using a cyanide free plating solution, is currently undergoing large scale experiments in industrial environment. In the last part of this chapter, it was demonstrated that direct silver plating on a structured silicon surface is possible. Silver was deposited on the sharp edge of a “W” shaped trench due to its high field density. The adhesion of this silver layer was of high quality after a sintering step. High-efficiency screen-printed & plated solar cell 129 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 8 High-efficiency screen-printed & plated solar cell In this chapter two different investigations will be presented, both aiming for higher efficiencies for solar cells with fine-line screen-printed and plated front side contacts. The first investigation was based on the question whether a lowly doped emitter can be contacted by a screen-printing and plating process. In the second investigation a high-efficiency rear side structure, the laser-fired contact (LFC), was used to substitute the standard Al-BSF in order to increase the efficiency. Both approaches were successful, leading to efficiencies up to 18.7% for a conventional Al-BSF and up to 19.3% for a LFC rear side structure on 2 x 2 cm² sized FZ silicon solar cells. 8.1 Production sequence for LFC and Al-BSF solar cells The production sequence for both, the laser-fired contact (LFC [35]) structure and the Al-BSF is presented in Fig. 8.1. In both batches solar cells of size 2 cm x 2 cm were processed on 4 inch boron-doped FZ-wafers with a base resistivity of 0.5 Ω cm and a thickness of 250 µm. A sectional drawing of both cell types is presented in Fig. 2.3 of Chapter 2.1.2. Al-BSF process sequence: The process sequence for cells with an Al-BSF is similar to the process presented in the previous chapter: The wafers were wet chemically textured, had a POCl3 emitter with a sheet resistance of 40 Ω/sq, 60 Ω/sq and 90 Ω/sq (planar diffusion), a SiNX antireflection coating and a screenprinted aluminum rear side. Fine metal lines were screen-printed on the front side using hotmelt paste. A screen with 80 µm mesh openings was used. Finally the wafers were co-fired in a fast firing belt furnace under ambient air. LFC process sequence: After oxidation and selective removal of oxide from the front [36], the front side was wet chemically textured. An emitter variation resulting in a sheet resistance of 40 Ω/sq, 60 Ω/sq and 90 Ω/sq was performed (planar diffusion). The front contacts were screen-printed using hotmelt paste and had a final width of about 80 µm to 100 µm. In contrast to the conventional industrial process sequence, the front contact was formed before evaporating an aluminum layer on top of the rear side oxide and a subsequent laser firing of point contacts. The solar cells were IV-characterized at different steps in the fabrication process: after the laser-fired contact process, after an annealing step and finally 130 High-efficiency screen-printed & plated solar cell _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 4’’ FZ wafer (250µm thick, 0.5 Ω cm, p-doped, bright etched) Wet chemical cleaning Dry oxidation Wet oxidation Selective removal of oxide from front side (2 x 2 cm² sized areas) Random pyramids texture of front POCl3 planar diffusion (40, 60, 90 Ω/sq) PSG etching, cleaning Removal of oxide from rear Sputtered SiNx:H antireflection coating Screen-printing of rear contacts (conv. Al paste) Screen-printing of front contacts (Ag hotmelt paste) Contact formation (fast firing belt furnace) Evaporation of aluminum on rear Laser fired contacts (LFC) Annealing of rear surface Light-induced plating of front contacts Al-BSF process sequence LFC process sequence Fig. 8.1: Process flow diagram of manufactured solar cells featuring screen-printed front contacts thickened by light-induced plating and an Al-BSF (left-hand) and a LFC (right-hand) rear side structure. after the light-induced plating process. In addition to the emitter sheet resistance, the firing temperature of the inline fast firing belt furnace and the temperature of the annealing step were varied. 8.2 Aluminum back surface field cell structure In Chapter 7.3 it was demonstrated that a significant efficiency gain can be achieved for light-induced plated cells. Thus, for the Al-BSF cells mainly IVparameters after the plating process will be presented. In Table 8.1 IV-parameters, pseudo fill factor (PFF) and series resistance of the wafer (arithmetic average of all 7 cells) and the cell with the highest efficiency is presented. These results correlate with findings of previous experiments (see Chapter 6.3.3). The efficiency rises with reduced doping concentration as long as the series resistance and thus the fill factor is not the limiting factor. Indeed, the series resistance rises with reduced emitter doping from a quite low value of 0.46 Ω cm² to 0.74 Ω cm², but due to the gain in short-circuit current density from 34.8 mA/cm² to 36.3 mA/cm², the efficiency increases by 0.6% absolute. To High-efficiency screen-printed & plated solar cell 131 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 8.1: IV-parameters of the best cell and best wafer (arithmetic average of 7 cells) for each emitter doping level. For the best cells also the pseudo fill factor PFF and series resistance (lightdark comparison method) were determined. The cell size is 2 x 2 cm². Wafer/Cell-ID Rsh [Ω/sq] Voc [mV] jsc [mA/cm2] Cell 6.4 Avg. Wafer 6 Cell 15.4 Avg. Wafer 15 Cell 24.4 Avg. Wafer 20 40 632 632±1 632 632±1 640 642±1 35.2 34.8±0.3 35.6 35.4±0.1 36.5 36.3±0.3 60 90 FF [%] η [%] rs PFF [%] [Ω cm²] 81.2 18.1 83.5 80.9±1.2 17.8 ± 0.2 81.2 18.3 83.7 80.2±1.2 17.9±0.3 79.8 18.7 83.6 78.5±1.2 18.3±0.4 - 0.46 0.49 0.74 - Reflectance, IQE illustrate this effect, the internal quantum efficiency of three cells with different emitter sheet resistance is presented in Fig. 8.2. In Fig. 8.3, a picture of the wafer constituting the best processed cell is presented. The middle cell reached an efficiency of 18.7%. Also the mean efficiency of 18.5% of the six cells (excluding the cell with 12.6%) on the same wafer is on a high level. The cell with the 12.6% efficiency had a low parallel resistance due to a shunt in the space charge region caused by a problem during the printing process. 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 IQE Al-BSF cells 24.4: Rsh = 90 Ω/sq 15.4: Rsh = 60 Ω/sq 5.4: Rsh = 40 Ω/sq reflectance 400 600 800 1000 1200 Wavelength [nm] Fig. 8.2: Internal quantum efficiency and reflectance measurement of cells with different emitter sheet resistances. A significant improvement in the short wavelength region can be observed. 132 High-efficiency screen-printed & plated solar cell _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 8.3: Picture of the wafer with the 18.7% efficient solar cell featuring a 90 Ω/sq emitter and an Al-BSF. The mean efficiency of the six solar cells on this wafer (excluding the 12.6% efficient cell) is equal to 18.5%. Before applying the plating process, 18 solar cells (at least one of each wafer) with an emitter sheet resistance of 90 Ω/sq were characterized. The IV-parameters before and after the plating step of these cells are presented in Table 8.2. Whereas the open-circuit voltage remains equal, the short-circuit current density drops by 0.8 mA/cm² caused by the enlarged shaded area. Nevertheless, due to the plating step the fill factor of 58% (series resistance of about 10 Ω cm²) could strongly be improved to 79.0%. This improvement can not be solely explained by a better line conductivity. Measurements have shown that the contact resistivity could be reduced from about 19 mΩ cm² to about 1.6 mΩ cm² - 5 mΩ cm². This indicates that the plating process influences the contact resistivity (compare Chapter 11.4). Table 8.2: IV-parameters of 18 FZ cells (2 x 2 cm²) having a sheet resistance of 90 Ω/sq measured before and after the light-induced plating process (LIP). Mean value and standard deviation are presented. Status before LIP after LIP Rsh [Ω/sq] Voc [mV] 90 639 ± 5 641 ± 3 jsc [mA/cm2] FF [%] η [%] 36.8 ± 0.2 57.6 ± 4.7 13.6 ± 1.1 36.0 ± 0.4 79.0 ± 0.9 18.2 ± 0.2 Fig. 8.4 shows histograms of the fill factor and efficiency distribution of all solar cells processed and having an emitter sheet resistance of Rsh = 90 Ω/sq. Just two of the 42 solar cells showed poor efficiencies, both can be explained by shunts due to a failure in the production process. The mean fill factor of the remaining 40 cells is High-efficiency screen-printed & plated solar cell 133 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ equal to 78.8%, the mean efficiency to 18.3%. This demonstrates a stable and reproducible process on laboratory scale. 12 Rsh = 90 Ω/sq and Al-BSF 10 FF = 78.8 % Rsh = 90 Ω/sq and Al-BSF 10 η = 18.3% (40 cells with FF > 60%) 6 4 2 0 40 50 8 Counts Counts 8 12 (40 cells with η > 13%) 6 4 2 75 76 77 78 79 80 81 0 8 Fill factor FF [%] 12 17.5 18.0 18.5 Efficiency η [%] 19.0 Fig. 8.4: The fill factor and efficiency distribution of all cells featuring an emitter sheet resistance of 90 Ω/sq is illustrated. The mean fill factor and the mean efficiency over all 40 cells (excluding the 12.6% and 9.2%) are equal to 78.8% and 18.3%, respectively. 8.3 Laser-fired contact cell structure The laser-fired contact cells were processed as described in Chapter 8.1. The IVparameters of all cells were measured after the LFC process, after an annealing step and after the light-induced plating process. The fill factor should be on a high level already after the LFC process, as the contact formation process on rear and front was completed. For conventional laser-fired contact cells with an evaporated front side structure a temperature step (350°C to 450°C) boosts the short-circuit current density and open-circuit voltage due to an annealing of the rear side. However, in this step the screen-printed front side could be damaged by the annealing process, causing a significant drop in the fill factor due to increased recombination currents. By the light-induced plating step, the conductivity of the relatively fine electrodes is improved and probably the contact resistance reduced (see Chapter 11.4). 8.3.1 Variation of the firing temperature As the optimum firing temperature for the applied process was unknown, wafers were fired at three different temperatures T1, T2 and T3, with T1 < T2 < T3. For the cells with a sheet resistance of Rsh = 40 Ω/sq it turned out that cells fired at 134 High-efficiency screen-printed & plated solar cell _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ temperature T1 achieved highest efficiency after the firing as well as after the annealing process. The mean fill factor of 78% (maximum value 79.1%) measured after firing dropped slightly during the annealing process by about 1% absolute. This could be due to the boost in the short-circuit current density of about 2 mA/cm² increasing the current flow in the fingers. Thus, the conductivity of the fingers was improved by light-induced silver plating. By this jsc dropped slightly due to the higher shaded area, but the fill factor could be increased even above the level after firing. Highest efficiencies with a mean value of 18.2% at a fill factor of 78.9% were achieved after the plating process for cells with Rsh = 40 Ω/sq. The efficiency of the “slightly overfired” solar cells was a slightly lower (0.3% absolute) after the contact formation process compared to the optimum fired ones. This is a result of increased contact resistance for “overfired” cells [132]. A forming gas anneal, as proposed by e.g. Schubert [133], did not result in a fill factor gain. However, after the plating process the fill factor was considerably raised to nearly the same level as for optimum fired cells. This can be solely 20 Efficiency η [%] Fill factor FF [%] 80 70 60 Firing Annealing Plating 18 16 14 12 40 firing temperature Annealing Plating 650 T1: optimum T2: slighlty overfired T3: overfired 640 Voc [mV] jsc [mA/cm²] Firing 35 630 620 610 30 Firing Annealing Plating 600 Firing Annealing Plating Fig. 8.5: IV-parameters of LFC-solar cells (Rsh = 40 Ω/sq), fired at different temperatures and measured after different fabrication steps. Each data point represents the mean value and standard deviation of nine to twelve samples. High-efficiency screen-printed & plated solar cell 135 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ explained by a better contact resistance to the emitter layer or a different current flow in the contact itself, as further discussed in Chapter 11.4. Similar results were achieved with cells fired at temperature T3. These “overfired” cells had a low fill factor of about 62% after firing and annealing. After plating a strong boost was achieved to a mean fill factor of 78.8%. As a result, the efficiency increases by 3.7% absolute to a mean value of 18.3%, the same high level as for optimum fired contacts. Due to the slightly higher open-circuit voltage for the “overfired” cells, the highest efficiency of 18.8% having a fill factor of 80.0% was obtained at this temperature. This effect could only be observed for the cells with a sheet resistance of 40 Ω/sq, as the fill factor over the whole temperature range for the less doped emitter were on a too low level before plating. 8.3.2 Variation of the emitter doping concentration In the following, IV-parameters measured at various steps in the processing sequence are presented for cells having a different emitter sheet resistance. The mean values and standard deviation of 21 cells each with a sheet resistance of 40 Ω/sq and 90 Ω/sq and 40 cells with Rsh = 60 Ω/sq are illustrated in Fig. 1.8. The short-circuit current density and open-circuit voltage rose sharply after the annealing step owing to the improved rear side passivation. The fill factor was very high for the cells with an emitter sheet resistance of 40 Ω/sq, achieving values up to 79.6% even before the annealing step. After the annealing step the fill factor dropped slightly, which is explained by the considerably increased short-circuit current density and insufficient conductivity of the fingers. Therefore light-induced silver plating was performed. Due to plating, jsc dropped slightly attributable to the enlarged shaded area, but the fill factor was completely recovered. The highest efficiencies were achieved after the plating step. However, the fill factor for cells with a high sheet resistance (Rsh = 60 Ω/sq and Rsh = 90 Ω/sq) is low after the firing and the annealing step. Since investigations [40] have shown, that the contact resistance of screen-printed cells rises remarkably with decreasing emitter doping concentration. High ρc values of 13 mΩ cm² and even over 100 mΩ cm² were measured for cells with an emitter sheet resistance of 60 Ω/sq and 90 Ω/sq, respectively. This is one to two orders of magnitude higher than the contact resistivity of 2 mΩ cm² measured for the emitter 136 High-efficiency screen-printed & plated solar cell _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 20 18 Efficiency η [%] Fill factor FF [%] 80 70 60 50 40 Firing Annealing 16 14 12 10 Plating 655 38 650 37 Voc [mV] jsc [mA/cm²] Annealing Plating 660 39 36 40 Ω/sq 60 Ω/sq 90 Ω/sq 645 640 635 630 35 34 Firing 625 Firing Annealing 620 Plating Firing Annealing Plating Fig. 8.6: IV-parameters for different sheet resistance emitters and after different processing steps. with a sheet resistance of 40 Ω/sq. After thickening these contacts by light-induced silver plating, a remarkable improvement was observed. The fill factor increased from a very low level to values between 75% and 80% even for the 90 Ω/sq emitter. To ensure that the contact resistivity is significantly reduced, measurements were performed on these cells again. As for the Al-BSF cells the contact resistance could be enhanced by the plating step to values of 3 mΩ cm² (compare Chapter 11.4). Due to the boost in fill factor, efficiencies for the cells with the lowest doping concentration were maximal due to the improved shortcircuit current density in the short wavelength region. Measurements of the internal quantum efficiency are presented in Fig. 8.7. The increasing reflectivity and the slightly higher internal quantum efficiency in the long wavelength region are related to the reduced free-carrier absorption in the emitter layer [134]. The best solar cell results after plating are presented in Table 8.3. Fraunhofer ISE Calibration Laboratory (CalLab) confirmed the efficiency of 19.3% for the cell (ID-19.2) having a sheet resistance of 90 Ω/sq, which is to the knowledge of the author, the highest value for a cell with screen-printed front contacts up to date. High-efficiency screen-printed & plated solar cell 137 Reflectance, IQE _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 IQE 90 Ω/sq LFC cell 19.2: Rsh = 90 Ω/sq 13.1: Rsh = 60 Ω/sq 4.1: Rsh = 40 Ω/sq 60 Ω/sq 40 Ω/sq reflectance 400 600 800 1000 1200 Wavelength [nm] Fig. 8.7: Internal quantum efficiency and reflectance measurement of LFC cells with different emitter sheet resistances. The investigation has shown that the annealing temperature has not a detrimental influence on the efficiency as long as the temperature is in the range of 300°C to 400°C. As further investigations have revealed, a sintering step after the light-induced plating process can even further improve the contact resistance (see Chapter 11.4.1) Table 8.3: IV-parameters, pseudo fill factor and series resistance of best LFC solar cells for different emitter sheet resistances are presented. *Calibrated measurement at Fraunhofer ISE Calibration Laboratory. Cell-ID 33_7.1 33_13.1 33_19.2* Rsh [Ω/sq] Voc [mV] jsc [mA/cm2] FF [%] 40 60 90 638.5 647.4 655.4 36.7 37.1 38.2 80.0 79.6 76.9 η [%] PFF [%] rs [Ω cm²] 18.8 19.1 19.3 83.1 83.0 0.60 1.19 8.4 Laser-fired contact versus aluminum back surface field Comparing the IV-parameters of the best solar cells with the laser-fired rear side structure to those with an aluminum back surface field, a clear efficiency gain of about 0.6% to 0.7% absolute was achieved (Table 8.4). There is a strong gain in jsc (∆jsc≈+4.2%) and Voc (∆Voc≈+1.1% for the 40 Ω/sq and +2.3% for the 60 Ω/sq and 90 Ω/sq emitter), which can be explained by the improved internal reflectance and 138 High-efficiency screen-printed & plated solar cell _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 8.4: IV-parameters, series and parallel resistance of the best solar cells featuring an Al-BSF and a LFC rear side structure. rs was determined by comparing illuminated and dark IV-curve. Wafer. Cell-ID Rsh [Ω/sq] 32_6.4 33_7.1 32_15.4 33_13.1 32_24.4 33_19.2 40 60 90 η rear side Voc [mV] jsc [mA/cm2] FF [%] [%] rs [Ω cm²] Al-BSF LFC Al-BSF LFC Al-BSF LFC 632 639 632 647 640 655 35.2 36.7 35.6 37.1 36.5 38.2 81.2 80.0 81.2 79.6 79.8 76.9 18.1 18.8 18.3 19.1 18.7 19.3 0.46 0.6 0.49 n.a. 0.74 1.19 Reflectance, IQE the high-quality passivated rear side, achieved by the thermally grown oxide and the laser-fired point contacts. The improvement of the rear side reflectance and of the internal quantum efficiency is illustrated in the spectral response measurements in Fig. 8.8. The lower fill factor is due to a higher series resistance. Comparing the solar cells with an emitter sheet resistance of Rsh = 90 Ω/sq, the series resistance of 1.2 Ω cm² for the LFC cell is significantly higher than for the Al-BSF cell with a resistance of 0.74 Ω cm². Possible reasons are insufficient finger conductivity, elevated contact 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 PC1D simulation LFC cells 19.2 - 90 Ω/sq 13.1 - 60 Ω/sq 4.1 - 40 Ω/sq Al-BSF cells 24.4 - 90 Ω/sq 15.4 - 60 Ω/sq 5.1 - 40 Ω/sq 400 600 800 LFC 90 Ω/sq 60 Ω/sq 40 Ω/sq Al-BSF 1000 1200 Wavelength [nm] Fig. 8.8: Internal quantum efficiency and reflectance measurements of solar cells with different rear side structures and emitter sheet resistances. Clear improvement for the LFC rear side structure for long wavelengths and for higher sheet resistance in the short wavelength region. The curves with a sheet resistance of 90 Ω/sq were simulated using the simulation software PC1D. High-efficiency screen-printed & plated solar cell 139 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ resistance or higher rear side resistance. As this was the first batch combining a LFC rear side with a screen-printed and plated front side structure, an improvement of the series resistance for future batches is likely. The Al-BSF and LFC solar cells with a sheet resistance of Rsh = 90 Ω/sq were modeled using the simulation software PC1D. For the Al-BSF cell an internal rear surface reflection of 72% (diffuse reflection) and a back surface field recombination rate of 1385 cm/s (not the low base doping ρb = 0.5 Ω cm) were assumed, for the LFC cell a recombination rate of 152 cm/s and a reflectance of 94 % (specular reflection, Table 8.5), which resulted in a good modeling of the rear side. From the IQE data (see Fig. 8.8) it is apparent that the emitter of the Al-BSF has a higher internal quantum efficiency compared to the one of the LFC cell. To be independent from technological differences caused by the front side, for both cells the simulated curve was fitted to the Al-BSF IQE data points in the shortwavelength region. This is the reason why the short-circuit current density of the modeled LFC cell is higher compared to the fabricated one. In addition the processed LFC solar cell was limited by relative high series resistance, resulting in 2000 18.2% Effective Rear Surface Recombination Velocity [cm/s] 1800 1600 1400 18.8% 1200 18.4% 1000 800 18.6% 600 18.8% 400 19.0% 19.2% 19.4% 200 19.6% 19.8% 20.0% 0 0 20 40 60 80 20.2% 100 Internal Rear Reflection [%] Fig. 8.9: Simulation results illustrating the absolute efficiency of solar cells under the variation of Sback and Rback. The front reflectivity and series resistance as measured at the Al-BSF cell was used. Modeling refers to a 250 µm thick solar cell wit a base doping of 0.5 Ω cm and an emitter sheet resistance of 90 Ω/sq. The 18.8% efficiency point relates to the Al-BSF parameters, the 20.2% efficiency point to the LFC parameters. 140 High-efficiency screen-printed & plated solar cell _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 8.5: Internal reflectivities (Rback) and rear surface recombination velocities (Sback) as extracted from solar cell modeling ((d) = diffuse and (s) specular reflection). In addition the IVparameter simulation results are presented. Rsh Wafer Cell-ID [Ω/sq] simu 1 simu 2 simu 3 90 90 90 rear side Sback [cm/s] Al-BSF 1385 LFC 152 LFC 152 η rs Rback Voc jsc FF [%] [Ω cm²] [mV] [mA/cm²] [%] [%] 72 (d) 94 (s) 94 (s) 18.8 19.5 20.2 0.74 1.19 0.74 645 654 654 36.5 38.6 38.7 79.9 77.2 79.8 a modeled efficiency of 19.5%. Assuming the same series resistance for the LFC cell as for the Al-BSF cell, efficiencies up to 20.2% can be achieved, which is an increase of 1.4% absolute compared to the Al-BSF cell (see Fig. 8.9). 8.5 Chapter summary Fine-line screen-printed contacts were thickened by light-induced plating. The results of the emitter variation clearly demonstrate the higher efficiency potential for lowly doped emitters due to the higher internal quantum efficiency in the short wavelength region and the reduced emitter recombination current density. However, the major challenge is to achieve a low contact resistance on a lowly doped emitter. In this investigation the 90 Ω/sq emitter could successfully be contacted only if the plating process was applied. The plated layer reduced the contact resistance between emitter surface and contact significantly. The best solar cells (2 x 2 cm², FZ-Si) with an Al-BSF achieved an efficiency of 18.1%, 18.3% and 18.7% having a sheet resistance of 40 Ω/sq, 60 Ω/sq and 90 Ω/sq, respectively. The mean efficiency of 18.3% for the 40 cells processed with an emitter sheet resistance of Rsh = 90 Ω/sq demonstrates the high level reproducibility on laboratory scale. Furthermore, the latter investigation was repeated with a modified rear side structure. A laser-fired contact rear side was combined with the screen-printed and plated front grid. High efficiencies up to 18.7% were achieved for the cells with an emitter sheet resistance of 40 Ω/sq even before the plating process. In contrast, cells with a sheet resistance of 60 Ω/sq and 90 Ω/sq showed poor performance still after the annealing process. The efficiency could be significantly improved after the plating process due to a reduction in the contact and consequently series High-efficiency screen-printed & plated solar cell 141 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ resistance. The highest efficiencies achieved for cells with an emitter sheet resistance of 60 Ω/sq and 90 Ω/sq were 19.1% and 19.3%, respectively. Both values exceed the highest efficiency for a cell with a screen-printed front side reported up to date. Nevertheless, the short-circuit current density as well as the fill factor of the LFC cells can be further improved as demonstrated by model calculations, which would result in efficiencies exceeding 20%. Furthermore the covered area by the busbar (covered area pc,bus ≈ 2%) and the relatively broad contacts of approximately 100 µm (covered area pc,f > 5%) is about 7%, which is similar as for standard production cells. Replacing the screen-printed front contact by e.g. metal aerosol jet printing (see Chapter 10) would yield in an even higher efficiency potential than just reported. 142 Pad-printing: Technology and analysis _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 9 Pad-printing: Technology and analysis The pad-printing technology for front side grid processing has been extensively investigated at Fraunhofer ISE [38,47-49,135]. Economical evaluation performed by Huljic et al. [49] have shown that pad-printing is able to compete with the state-of-the-art screen-printing technology. The main advantages of the padprinting technology are the ability to produce fine lines, to print on uneven surfaces and at the same time to achieve high throughput rates. However, padprinted solar cells have been limited by poor line conductivity, caused by a not sufficient paste transfer from the cliché (printing plate) to the wafer. After the firing process the finger was 1.5 µm to 2.5 µm high [49]. In order to achieve a sufficiently high line conductivity, multiple prints are necessary, increasing the overall process complexity, especially due to an obligatory paste drying step inbetween the prints. This limitation can be overcome by using pad-printed contacts as seed layer for the light-induced plating process, which will be presented in this chapter. Furthermore pad-printing of hotmelt paste is discussed. 9.1 Printing of hotmelt paste Instead of pad-printing conventional paste, hotmelt paste was used assuming that multiple prints are possible without additional drying steps in-between. This assumption was confirmed in first experiments by Huljic [136]. The process was further applied and optimized in this work. The process sequence for printing hotmelt paste is equal to the standard padprinting one (see Chapter 2.2.4). As for screen-printing hotmelt paste, different temperatures need to be applied to the system including the paste reservoir, the cliché and the print nest. Hotmelt paste is placed in the reservoir and heated up to a temperature at which the viscosity of the paste is low enough to be printable. In a further step the cliché being held at the same temperature as the reservoir, is inked and raked. The reason causing the paste to stick to the pad is not completely understood. It is supposed, that owing to temperature difference of the pad and the cliché, the top surface of the paste cools slightly down when the pad contacts the paste. This causes an augmentation of its viscosity and makes it therefore tackier. The paste adheres to the pad and is transferred to the wafer. The same process is repeated by contacting the wafer with the pad. The substrate, which temperature is Pad-printing: Technology and analysis 143 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 9.1: Pad-printer set-up used for printing hotmelt paste. The ink reservoir, cliché and printing table are heatable. lower than the surface temperature of the paste, causes the paste to tack to the wafer. As the pad is not heated, it is important that the paste transfer from the cliché to the substrate is carried out fast. Otherwise the paste cools down on the pad and re-solidifies. As a consequence, a paste transfer would be impossible. A picture of the pad-printer set-up at Fraunhofer ISE is illustrated in Fig. 9.1. The temperature settings of the ink reservoir, cliché and printing table as well as the time of the paste transfer are of crucial importance. The ink reservoir and cliché temperature need to be well above the melting point of the paste, to achieve printability. Compared to the screen-printing hotmelt paste the melting point of the pad printing paste was reduced to Tmelt = 56°C. In addition solvents were added to the paste to make it more fluid at lower temperatures. A careful temperature adjustment of the printing table needs to be done. Therefore a new vacuum table was designed with a lateral temperature homogeneity of ∆Tnest = ±1°C. At optimum temperature setting the complete print image from the pad was transferred onto the wafer surface. For temperatures slightly below or above this optimum temperature setting (±2°C < ∆Tnest < ±4°C) the print image was just partly transferred. At a temperature deviation of ∆Tnest > ± 4°C from the optimum temperature no transfer at all occurred. For conventional paste the evaporation rate of the solvents in the ink is one of the key factors. If the solvents evaporate too quickly, the ink might not be picked 144 Pad-printing: Technology and analysis _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 14 160 12 140 120 10 100 8 80 6 60 4 contact width wc 2 0 finger height hf 1 2 3 40 20 Contact width wc [µm] finger height hf [µm] up from the cliché or transferred to the wafer, because it has dried in the etched areas of the cliché or on the pad. If the evaporation rate of the solvents happens too slowly the ink surface may not be tacky enough to stick to the surface of the pad or substrate. In both cases little ink or no ink transfer at all is the result. Even if the right evaporation rate of the solvent has been determined, adding solvents during the print process is required. The necessary amount of solvents to be added to the paste depends on the process environment, as e.g. the temperature, humidity and throughput rate. For pad-printing hotmelt paste, the importance of the solvents plays a minor role. The paste transfer appears to rely more on the temperature gradient during the printing process. Long term printability of the paste seems to be quite good as the same paste was used for printing on different days without the need of adding solvents. For fine-line printing a relatively hard pad is recommended [50,137]. When using hotmelt paste a complete transfer of paste from the pad to the wafer was solely achieved by applying a hard pad of Shore18 12. Using a softer pad resulted in little or no paste transfer at all. The disadvantage of a hard pad is the increased mechanical stress for the wafer. Fig. 9.2 illustrates the finger widths and heights for a single, double and triple printed contact after the firing process. Each data point represents the mean value and standard deviation of six samples. The single printed contact was about 10 µm broader than the width of the finger opening in the cliché and 4 µm high. Each 0 Number of prints Fig. 9.2: Width and height of pad-printed contacts plotted versus the number of prints. 18 The hardness of a pad is measured in Shore. An iron bowl (aglet) is pressed on the surface of the material being measured. The harder the material is, the more the aglet sinks into the measurement device at a certain pressure applied (penetration depth: 2.5 mm = 0 Shore, 0 mm = 100 Shore). Pad-printing: Technology and analysis 145 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ additional printing step broadens the contact width by 20 µm to 30 µm and increases the contact height by about 2 µm. Contact broadening due to multiple prints was caused by the limited print repetition accuracy of the machine and by flattening the paste by each printing step. Solar cells were processed with 90 µm wide fingers on untextured multicrystalline silicon solar cells featuring a sheet resistance of Rsh = 40 Ω/sq. For solar cells with single printed contacts efficiencies up to 11.7% were achieved. However, the solar cells were strongly limited by the series resistance yielding fill factors of just 61.4%. For a triple printed contact the fill factor could be improved to 73% and the efficiency to 13.1%. Due to the enlarged shaded area, jsc dropped by 1.3 mA/cm² to 29.8 mA/cm². The efficiency of solar cells with triple padprinted contacts was still limited by the poor line conductivity. However, in combination with a subsequent light-induced plating process, single pad-printed contacts could be ideal to form the seed layer. 9.2 Pad-printed and plated contacts 9.2.1 Solar cell processing A batch of 22 Cz-solar cells with a size of 10 x 10 cm² and a thickness of 250 µm was processed. The wafers were divided into four groups: The emitter sheet resistance was varied (Rsh = 40 Ω/sq and Rsh = 60 Ω/sq) as well as the front side grid (finger width wc and separation distance s). One part of the wafers was pad-printed with 68 fingers of 50 µm widths and the other part with 43 fingers of 90 µm widths. The process sequence is illustrated in Fig. 9.3. Cz-Si wafer; 10x10 cm², 1-3 Ω cm, 250 µm Drying Wet chemical texturing and cleaning Pad printing of front grid (wf: 50 µm; 90 µm) POCl3 diffusion (Rsh=60 Ω/sq, Rsh=40 Ω/sq) Contact firing (fast firing furnace) PSG etch and cleaning Edge isolation (laser scribing + cleaving) Sputtered SiNx:H antireflection coating Light-induced plating Screen printing of rear contact Sintering under room ambient Fig. 9.3: Process flow diagram of fabricated pad-printed solar cells. 146 Pad-printing: Technology and analysis _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 9.2.2 Variation of the grid design and emitter doping concentration 18.0 Efficiency η [%] 80 79 78 77 76 75 74 73 72 71 790 17.5 17.0 16.5 16.0 810 37.0 640 36.5 635 36.0 630 Voc [mV] jsc [mA/cm²] Fill factor FF [%] IV-parameters measured after the sintering step are presented in Fig. 9.4. The optimum firing temperature of about 790°C to 810°C for the used paste and process was determined in a previous batch. Highest efficiencies for all groups were achieved at 790°C. Fig. 9.5 illustrates a solar cell with 50 µm broad fingers. Comparing jsc and Voc of the solar cells with a sheet resistance of Rsh = 40 Ω/sq to cells with Rsh = 60 Ω/sq, the same conclusions can be drawn as for the hotmelt screen-printed solar cells (compare Chapter 6.3.3): The average efficiency of the less doped emitter cells was higher, due to an increase in the short-circuit current density as well as in the open-circuit voltage. The fill factor of less doped emitter cells was on a slightly lower level, caused by an increased contact and emitter sheet resistance. The efficiency of solar cells with 50 µm broad fingers was on a lower level than the 90 µm broad ones. Short-circuit current density and open-circuit voltage were on a similar level. The shaded area of both grid types was comparable (pc ≈ 6 – 35.5 35.0 34.5 620 615 Rsh = 40 Ω/sq, wc = 50µm 610 Rsh = 40 Ω/sq, wc = 90µm 605 33.5 600 810 Firing temperature [°C] 810 625 34.0 790 790 Rsh = 60 Ω/sq, wc = 50µm Rsh = 60 Ω/sq, wc = 90µm 790 810 Firing temperature [°C] Fig. 9.4: IV-parameters after the sintering process of solar cells with pad-printed contacts thickened by light-induced plating. Pad-printing: Technology and analysis 147 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 9.5: Pad-printed solar cell of size 10 x 10 cm² with 50 µm fine-line printed and plated fingers. 7%) due to the increased density of fingers that had a width of 50 µm. However, the fill factor for the solar cells with the 50 µm wide fingers was significantly lower. 9.2.3 Influence of a sintering and second plating step The line conductivity of four solar cells was increased by a second plating step increasing the finger height and width by about 5 µm and 10 µm, respectively. Furthermore the wafers were sintered for 10 minutes at 300°C under air ambience. The IV-parameters measured at different steps in the process sequence are illustrated in Fig. 9.6. Short-circuit current density: After the first plating step, the short-circuit current density for the two cells with an emitter sheet resistance of Rsh = 60 Ω/sq were on a higher level than the ones of Rsh = 40 Ω/sq. The same grid was used for both types of solar cells. After performing a second plating step jsc of all cells dropped due to the increased shaded area. However, after performing a further sinter step jsc of the cell with Rsh = 60 Ω/sq increased again whereas jsc of the less doped ones remained on the same level. Open-circuit voltage: The open-circuit voltage after the first plating step was on a slightly higher level for the less doped emitter cells probably due to a lower 148 Pad-printing: Technology and analysis _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 81 18.0 Efficiency η [%] Fill factor FF [%] 80 79 78 77 76 plating sintering 16.5 plating sintering 36.5 plating sintering 630 Voc [mV] 35.5 35.0 34.5 34.0 plating sintering 640 36.0 2 17.0 75 74 jsc [mA/cm ] 17.5 620 Rsh = 60 Ω/sq, wc = 50 µm 610 Rsh = 60 Ω/sq, wc = 90 µm Rsh = 40 Ω/sq, wc = 50 µm plating sintering plating sintering Process step 600 Rsh = 40 Ω/sq, wc = 90 µm plating sintering plating sintering Process step Fig. 9.6: IV-parameters of four solar cells measured after different steps in the processing sequence. emitter dark saturation current (less Auger recombination) and a reduced surface recombination velocity. Voc seemed to be unaffected by the sintering step, but was slightly reduced after the second light-induced plating step. The final sintering step increased Voc again. Fill factor: The fill factor was the parameter affected most strongly by the different process steps applied. Considering the solar cell with Rsh = 40 Ω/sq and wc = 90 µm, the fill factor was on a high level directly after plating. Neither a sintering step nor a second plating step improved the fill factor. However, a sintering step after the second light-induced plating step boosted the fill factor to a value of 80.1%. Considering solar cells with the same sheet resistance (Rsh = 40 Ω/sq) and finer lines the fill factor was significantly elevated after each process step; the sintering, the second plating and sintering, raised the fill factor from 76% to 79.6%. The same was valid for the solar cell with a sheet resistance of Rsh = 60 Ω/sq and wc = 90 µm broad lines, but the improvement was not as strong compared to the cells with Rsh = 40 Ω/sq. The highest boost in the fill factor was achieved for the solar cell having a sheet resistance of Rsh = 60 Ω/sq and fingers of Pad-printing: Technology and analysis 149 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ wc = 50 µm. After a second plating step, the fill factor increased by more than 4% absolute from 75.3% to 79.6%. This high value could not be improved by a further sintering step. While a second light-induced plating step hardly affected the fill factor of solar cells with 90 µm broad lines, the FF of the ones with 50 µm broad lines was significantly improved. The same level as for the wider lines was achieved. The finger conductivity after the first light-induced plating step was still limiting cell performance, being changed by further thickening the contacts with a layer of silver. The increase of the fill factor by sintering plated contacts could be caused by an improvement of the interface layer of the screen-printed silver contact and the overlying plated silver and/or by a reduction of the metal-semiconductor contact resistance (compare Chapter 11.3.1). Efficiency: The efficiency of fine-line printed contacts was on a relatively low value after the first plating step and was increased throughout the different process steps by about 0.8%abs to 17.3% and 17.8% for cells with a sheet resistance of 40 Ω/sq and 60 Ω/sq, respectively. This was different for solar cells with 90 µm broad contacts. The efficiency of cells for both types of emitter was increased after the first sintering step. However, after the second light-induced plating step the efficiency dropped due to the increased shaded area decreasing jsc, without improving FF or Voc. In the final sintering step the efficiency could be nearly completely recovered. Nevertheless, the highest efficiency of pad-printed cells with 90 µm broad lines was achieved after the first sintering process. The IV-parameters of the highest efficiencies achieved for pad-printed and plated cells are presented in Table 9.1. Table 9.1: IV-parameters of the best 10 cm x 10 cm Cz-monocrystalline silicon solar cells with pad-printed and plated contacts. Efficiencies were obtained after a sintering step at the end of the process sequence (s: finger separation distance). Cell-ID HM14_6 HM14_2 HM14_19 HM14_12 Rsh [Ω] 40 60 wc [µm] 50 90 50 90 s [mm] 1.5 2.3 1.5 2.3 Voc [mV] 622 621 630 624 jsc [mA/cm2] 35.5 35.9 35.6 36.1 FF [%] 79.5 78.7 79.6 79.7 η [%] 17.5 17.5 17.8 17.9 150 Pad-printing: Technology and analysis _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 9.3 Chapter summary Pad-printing the front side grid of the solar cell is an alternative technology to fine-line screen-printing. Using hotmelt paste, a complete transfer of the print image from the pad to the wafer was obtained. Therefore a relatively hard pad and a precise temperature setting of the printing table were necessary. The finger height and hence the finger conductivity of a single pad-printed contact is insufficient. Multiple pad-printing resulted in an increase of the contact width per printing step of at least 20 µm. For that reason the proposed two layer contact structure has been adapted. The pad-printing technology was used to print the contact layer, followed by the lightinduced plating step. Especially for fine-line printed front side contacts a sintering step at the end of the process sequence improved the efficiency of processed solar cells significantly. Cz-silicon solar cells of size 10 x 10 cm² have been processed. Efficiencies up to η = 17.8% and η = 17.9% were achieved using a front side grid with 50 µm and 90 µm broad contacts, respectively. To exhibit the advantage of fine-line printing further ink optimization is necessary. For industrial mass production a rotary pad-printing machine is recommended in order to avoid excessive pressure to the wafer surface and to achieve high throughput rates. Metal aerosol jet printing 151 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 10 Metal aerosol jet printing After the detailed review of printing technologies that have been applied for solar cell metallization, it becomes apparent that all technologies have some drawbacks for industrial mass production of high-efficiency solar cells. However, the metal aerosol jet system discussed in this chapter seems to combine many advantages of different technologies. The metal aerosol jet printer is a non-contact system, allowing printing on thin and fragile wafers and wafers having an uneven surface. Additionally, structuring prior to printing is not necessary due to the direct-write technology and the ability to print fine-lines of less than 15 µm widths. Furthermore, the material utilization is efficient and the range of physical and chemical properties of the metal inks and pastes which can be used is wide. These advantages can be achieved despite the relatively simple working principle of the printer. In this chapter the metal aerosol jet system is introduced in detail. The optimization for solar cell metallization and printing characteristics are discussed. In the last part solar cell results achieved by metal aerosol jet printing and lightinduced plating of the front grid are presented. 10.1 Working principle of the metal aerosol jet printer The metal aerosol jet printer is a maskless non-contact printing system developed by Optomec Inc., USA. In contrast to an ink-jet system (see Chapter 2.2.5) the metal-containing ink is not printed directly. The ink is atomized pneumatically or ultrasonically (see below). The aerosol is transported to the deposition head via a heatable tube, where the viscosity of the aerosol can be controlled. The gas flow diagram is presented in Fig. 10.1. If the pneumatic atomizer is used, some of the gas needs to be extracted in the so-called virtual impactor to separate the excessive transport gas from the aerosol. In the deposition head the aerosol stream is focused by a second gas stream (sheath gas) and deposited through a tip onto the surface of the substrate. The continuous aerosol stream can be interrupted by a mechanical shutter. The metal aerosol jet printer is operated via a graphical user interface, over which the gas flows, temperature settings etc. are set and the pressures are monitored. AutoCAD drawings can be translated into machine code and directly printed which makes the system very flexible for e.g. testing different grid designs. 152 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ process gas process control module carrier gas exhaust gas process control module pneumatic atomizer ultrasonic atomizer carrier gas virtual impactor aerosol heating tube focusing gas aerosol heating tube deposition head tip process gas shutter deposition head tip focusing gas shutter Substrate substrate Substrate substrate Fig. 10.1: Gas flow diagram of the metal aerosol jet system for the ultrasonic (left-hand) and the pneumatic (right-hand) atomization mode. 10.1.1 Deposition head The deposition head is the key part of the printer (see Fig. 10.2). In the head, the aerosol is surrounded by a second gas flow, which focuses the aerosol to a small diameter, prevents contact between aerosol and nozzle tip and accelerates the droplets in the aerosol. As the aerosol stream is focused by the laminar gas stream, the width of the deposited line can be much smaller than the width of the nozzle tip. A simulation of the different gas flows in the deposition head shows that the aerosol is completely surrounded by the focusing gas. This simulation was focusing gas deposition head aerosol tip substrate Fig. 10.2: Schematic drawing of the working Fig. 10.3: Simulation of the gas flow in the principle of the deposition head. The aerosol is deposition head. A droplet size of ddrop =1 µm surrounded by a second gas stream. and N2 as process gas was assumed. Metal aerosol jet printing 153 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ line width [µm] performed by S. Hörteis using the CFD flow modeling software Fluent. In the simulation a droplet diameter of ddrop = 1 µm and N2 as focusing gas was assumed. After leaving the tip, droplets have enough momentum to remain collimated, which means that the distance between tip outlet and surface may vary a few millimeters without a significant change of the deposited line width (see Fig. 10.4). This makes the technology also suitable for printing on uneven surfaces. However, if the distance between surface and tip outlet is too short, the line width increases due to turbulences (Fig. 10.4 (#1)). If the distance is too large, the aerosol stream will be defocused also resulting in an increased line width (Fig. 10.4 (#3)). Noninterrupted lines printed over a 500 µm deep trench without changing the working distance are illustrated in the SEM image of Fig. 10.5. #1 #2 #3 working distance [mm] Fig. 10.4: Line width plotted versus the distance between tip Fig. 10.5: SEM picture of outlet and substrate for too short (1), optimum (2) and too large 60 µm lines printed over a (3) working distances (drawings taken from [138]) 500 µm deep trench [138]. 10.1.2 Pneumatic and ultrasonic atomizer A metal containing ink is atomized pneumatically or ultrasonically into small droplets and transported as an aerosol stream to the deposition head. The working principle of the pneumatic atomizer is presented in Fig. 10.6. Compressed gas is expanded through the atomizer nozzle, producing a high-velocity gas stream. Due to the Venturi principle the ink is drawn off the reservoir through a riser tube into the atomizer nozzle. Subsequently, the high-velocity gas stream breaks the liquid stream into droplets and suspends them in the flow. The gas jet containing the droplets impinges against the sidewalls of the ink reservoir so that the larger droplets are removed. The smaller particles remain suspended in the gas and are transported to the deposition head. As this requires a high gas flow, some of the carrier gas needs to be extracted in the so-called virtual impactor. The concentrated 154 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 10.6: Schematic drawing of the Fig. 10.7: Working principle of the pneumatic pneumatic atomizer (taken from [138]). atomizer. aerosol is then transported to the deposition head. In the ultrasonic atomizer a small bottle partly filled with ink is positioned over a piezoelectric transducer (see Fig. 10.8). This transducer generates high frequency pressure waves, which are transferred via a liquid (water) to the bottle containing the ink. The tips of the thereby generated capillary waves are pinched off resulting in the atomization of small droplets. These atomized droplets are then entrained in a gas stream and transported to the deposition head. The characteristics of the inks used for the two atomization modes differ strongly. The main characteristics/requirements for each mode are summarized in Table 10.1. One important difference is the maximum particle size of the ink. For Fig. 10.8: Schematic drawing (taken from [138]) and working principle of the ultrasonic atomizer. Metal aerosol jet printing 155 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ the ultrasonic atomizer metal organic and nano particle inks can be used. Because nano-particle inks are expensive and often have a short shelf life, they are probably not suitable for solar cell mass production. Thus, the use of the ultrasonic atomizer in combination with nano-particle inks is rather more suited for laboratory use to test small ink volumes. Nevertheless, the ultrasonic atomizer in combination with metal organic inks might be an interesting field for future research. The pneumatic atomizer can be used for larger particle sizes and a wide viscosity range up to 1000 cP. This makes it more suitable for industrial application as the metal content of the ink can be much higher (40 - 60% by mass) compared to the ultrasonic atomization. To prevent the ink from drying during the atomization process, the solvents need to have a low to medium vapor pressure. Since a high gas flow is necessary for atomization, some of the carrier gas needs to be extracted, which increases the overall process complexity. However, the pneumatic atomizer seems to be best suited for industrial application due to its robustness and the wide range of inks that can be used. For laboratory application the advantage of each of the atomization modes can be utilized. Table 10.1: Requirements for the ink for each atomization mode [138]. particle size viscosity range minimum amount of ink vapor pressure of solvents unit ultrasonic pneumatic [nm] [mPa s] [ml] [hPa] < 50 0.7 - 30 0.1 low to high < 500 1 - 1000 10 low to medium 10.2 Nano-particle inks Production of nano-particle inks for ink-jet application is a highly specialized field, since the inks need to have very well defined physical and chemical properties in order to guarantee good printing results and optimum functionality of the printing system [139]. For optimum printability with the metal aerosol jet system it is important that the dynamic viscosity µ , the size of the nano-particles as well as the vapor pressure pD is within the specified range of the used atomization mode (see Table 10.1). Furthermore the ink has to remain stable during a long period of time, at least during the printing process. For a successful transfer of the 156 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ metal aerosol jet system into industrial application, the production of ink needs to be also economic. The composition of inks can be of organic or inorganic nature as well as soluble or insoluble in water. Conventional metal particle based ink-jet inks usually contain particles not exceeding 200 nm in size [140]. 10.2.1 Metallic nano-particles One of the most important components of the ink for solar cell metallization is the metal loading in form of small particles. The metal itself should form a low contact resistance to the silicon surface. A high conductivity is of minor importance, as this will be achieved by the second highly conductive layer. This is different for screen-printing, where both, the conducting and the contact layer have to be formed in one step, which is the reason why silver as a highly conductive material is used. As discussed in Chapter 5.2.2, Ni, Cr and Ti could be the materials of first choice for the contact layer. At present there is little experience in producing stable nano-particle inks from most metals, which probably will change over the next few years due to the world wide boom in nano technology. Metallic nano-particles have, compared to the respective solid material, an increased surface reactivity and a lowered melting point [140]. Silver nano particles, like many other materials, easily oxidize and are explosive [141]. The production of aqueous dispersions containing silver nano-particles is most often performed by chemical reduction of a suitable silver-salt with reducing agents such as sodium and potassium-boron-hydride, trisodium citrate, hydrazine, ascorbic acid, hydrocarbons or gaseous hydrogen [140]. Nano-particles of small size can be obtained in this way. Fig. 10.9-a shows an SEM image of an ink from the company Harima with particle sizes of around 3 nm to 7 nm. Another way to produce silver nano-particles is by physical evaporation of a silver wire, as performed by the Fraunhofer Institute for Manufacturing Technology and Applied Materials and Research (IFAM) [142]. The primary particles have a cross-section diameter of 50 nm to 100 nm. These particles agglomerate during the process to structures having a cross-section of 5 µm to 10 µm (see Fig. 10.9-b). These agglomerates are ground using a mill and dispersed in solvent. The final particle size in the solvent can be reduced to less than 100 nm. Conventional silver screen-printing paste for solar cell metallization is produced in the so called “down mill” process, in which in different steps bulk silver is Metal aerosol jet printing 157 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ ground down to the required size. Pastes with a particle size of typically 1 µm to 7 µm are produced for screen-printing applications. The smaller the particle size, the more costly is the production process. Fig. 10.9-c shows an SEM image of nickel powder with a maximum particle size of 1 µm used for a paste designed for this work by E. Kasper from Ferro/Germany. a) b) c) Fig. 10.9: SEM image of a) dispersed silver nano-particles with a cross section of about 3 to 7 nm [143] b) silver nano-particle agglomerates as used for the silver ink produced by Fraunhofer IFAM [142] c) nickel powder used for nickel screen-printing paste. As presented in Chapter 10.5 the best solar cell results so far were achieved with a modified commercially available silver screen-printing paste. The screen-printing pastes are highly optimized for solar cell metallization, achieving a relatively low contact resistance to a moderately doped emitter surface in combination with the screen-printing technology. However, the use of screen-printing paste for aerosol jet printing showed amazingly good results despite a particle size of up to 7 µm which exceeds the aerosol jet printing specifications by far. A SEM image of the printed screen-printing paste is presented in Fig. 10.10. Fig. 10.10: SEM image of a screen-printing paste which was slightly modified prior to printing with the metal aerosol jet system. 158 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 10.2.2 Ink components Besides the metal particles, the ink consists of solvents, binding agents and adhesion promoters. By volume, the main component of the inks is the solvent. In the solvent the nano-particles and stabilizers are dispersed. Depending on the vapor pressure and viscosity the suitability of the inks for one of the two atomization modes is defined (see Table 10.1). Viscosity and vapor pressure of commonly used solvents are presented in Table 10.2. Another important component of the ink is the binding agent. This organic material which is non-volatile at room temperature forms a three-dimensional lattice between the ink components and prevents the ink from spreading on the substrate after printing. Furthermore adhesion promoters such as bismuth oxide or silicate glass can be added to the ink to form a contact of high mechanical strength and to promote penetration through the antireflection coating if applied. Table 10.2: Vapor pressure pD and dynamic viscosity µ at 20°C of commonly used solvents for inks containing metal particles. solvent structure pD [hPa] µ [mPa s] water ethanol terpineol ethylene glycol n-Hexane H 2O C2H6O C10H18O C2H6O2 C7H16 23.4 58.7 0.24 0.05 160 1 1.2 n.a. 21 0.3 10.2.3 Stabilization of nano-particle inks Metal nano-particles in dispersion are also called colloid particles. They develop strong attractive forces between each other (van der Waals forces), which lead to flocculation and/or coagulation of particles. Repulsive forces between them can be produced by electrostatic or steric stabilization, which prevent the particles from approaching close enough to cause agglomeration. Electrostatic stabilization of nano-particles in polar solvents such as water is based on repulsive forces due to a polarization of the particles (see Fig. 10.11 lefthand). Polarized groups of the particles either dissociate by themselves or through the deposition of ionic dispersants or polyelectrolytes onto the particle. A Metal aerosol jet printing 159 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ polyelectrolyte is a polymer that dissociates in polar solvents, so that the polymer molecule is negatively charged and the corresponding positively charged counterion is located in the surrounding solution. Thus the particles carry a high elementary charge of the same polarity while the same amount of oppositely charged ions are located in the solvent [144]. Classic colloidal science explains electrostatic stabilization by an electrical double-layer. The surface of the particle is electrically charged and a diffuse cloud of oppositely charged ions is located around it. As two particles approach each other the charge effectively separates them and prevents closer particle interactions. Stabilization increases with the thickness of this layer [145]. Steric stabilization relies on the adsorption of a layer of resin or polymer chains on the surface of the particles (see Fig. 10.11 right-hand). When particles approach each other the adsorbed polymeric chains come into contact. This interaction can be regarded as a reduction in entropy, which is unfavorable and provides the necessary barrier to prevent further attraction [141]. Steric stabilization is typically applied for the production of stable suspensions with silver nano-particles and is influenced by e.g. temperature, pressure, molecular structure and composition and quality of the solvent [141]. Fig. 10.11: Model of electrostatic (left-hand) and of steric stabilization (right-hand) [145]. 10.3 Set-up of the printer at Fraunhofer ISE The metal aerosol jet printer was set-up at Fraunhofer ISE in 2004. It was one of the first systems ever built and the first one delivered to Europe. The printer was placed on a granite table under a flow box. In order to have on the one hand a relatively clean printing atmosphere and on the other hand to protect the operators from hazardous metal aerosols, the printer was shielded and connected to the 160 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 10.12: Metal aerosol jet system set up at Fraunhofer ISE. exhaust system of the building. Pictures of the set up are presented in Fig. 10.12. In the start-up phase the system including software and hardware was adapted for solar cell processing. Assembling and disassembling of parts of the system is a frequent task, which has to be performed when for example the ink is replaced, the atomization mode is changed or when components of the system need to be cleaned. As a lot of pipes and components are connected to each other, leakage can be a problem. This means, when using the same settings for the gas flow, different pressures will build up and printing results will differ completely. Therefore, in order to have a reproducible working condition, a “leakage check” routine has been introduced prior to printing: Depending on the size of the nozzle outlet a certain pressure is built up in the system at a particular gas flow. This pressure is measured by sensors and can be compared to reference values. If the values measured are too low there is a leak in the system, which can be located by further routine tests. In this way also clogging of the tip can be determined. If the measured pressure values are too high, clogging of the nozzle is probably the reason. Fig. 10.13 illustrates an SEM image of the tip as well as a top view of the tip outlet before and after clogging. Despite what one might expect from the working principle of the system, clogging of the nozzle tip caused by metal particles was a major problem using the pneumatic atomization mode. The responsible part was probably the virtual impactor that was located directly above the deposition head. Droplets built up in the impactor, which fell at some point into the deposition head, initiating clogging. Recently the virtual impactor and the deposition head were re-engineered by Optomec. In addition the impactor was relocated and connected directly to the Metal aerosol jet printing 161 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ atomizer aerosol outlet. This improved the uptime of the machine significantly. However, all results presented in this work are based on the first set-up of the system. Fig. 10.13: Left: SEM image of a nozzle tip having a cross section H of 150 µm [146]. Middle and right: Microscope top view of the tip outlet before and after clogging by metal particles. 10.4 Characterization and optimization of printing process A wide variety of inks were tested and optimized printing settings for each type were determined. The printed lines were optically characterized. This included edge definition, line width and height determination as well as SEM and EDX analysis. The viscosity and the maximum particle size of the inks especially influence the print settings. The main system settings that can be varied include the gas flow of the focusing gas, exhaust gas and transport gas as well as temperature settings of the tube and the printing table. The influence of the printing speed is illustrated in Fig. 10.14. If a lot of material is atomized and the viscosity of the aerosol droplets is low, increasing the printing Fig. 10.14: Influence of printing speed on line Fig. 10.15: Influence of the tube heater on line width (Top: 5mm/s. Bottom: 45mm/s). width (Top: 23°C. Middle: 70°C. Bottom: 90°C). 162 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Contact width wc [µm] speed will reduce the line width significantly. On the other hand the line width can also be reduced by heating the aerosol stream with the tube heater, as illustrated in Fig. 10.15. Some of the solvents will be evaporated during transport from the atomizer to the deposition head. The ink deposited on the substrate will have a higher viscosity and flow effects will be reduced or not present at all. However, at some temperature level all solvents will be evaporated before the droplets reach the substrate. In this case a powder is deposited, with neither good adhesion to the surface nor a good line definition. Another important setting is the focusing gas flow. The focusing gas prevents the aerosol droplets from hitting the inner walls of the nozzle (see Fig. 10.3) and focuses the aerosol droplets to a small diameter. With rising amount of material atomized, the focusing gas needs to be increased to achieve fine-line definition, as presented in Fig. 10.16. However, if the focusing gas flow is too high, turbulences may occur, resulting in poor line definition. In both cases, if the gas flow of the focusing gas is too low or too high, the danger of tip clogging rises. For the inks under investigation EDX measurements were carried out after the sintering process to analyze if all organic components of the ink were evaporated 200 150 100 50 0 30 40 50 60 70 80 90 Focusing gas [cm³/min] Fig. 10.16: Influence of the focusing gas on the line width; too low gas flow (30 cm³/min, lefthand) and too high gas flow (90 cm³/min, right-hand) for a fixed amount of aerosol stream. Metal aerosol jet printing 163 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ during the temperature treatment. This analysis was important as the composition of the inks differs from conventional screen-printing paste and any organic residues may influence the solar cell performance. As an example, the measurement for the silver containing ink from the company Harima is presented. The EDX analysis of the contact illustrates that the remaining part of the contact is pure silver. Gold which was also detected is not part of the paste, but was sputtered on the surface prior to the analyses to have a high conductance of the silicon surface needed for the EDX analysis. Remaining organic compounds were not detected. This was different for other printed and sintered contacts, in which a significant amount of carbon was detected. The carbon content of one silver ink sintered at 350°C even exceeded 60%, whereas the silver content was just 3%. These inks were not used for further investigation. Ag: 79.8 atom%; 68.4 mass% Au: 20.2 atom%, 31.6 mass% Fig. 10.17: Result of the EDX analysis (marked area of the SEM image) of a sintered contact printed with the silver nano-particle ink from Harima. Analyses were carried out at the University of Freiburg. Depending on the printing settings, the finger width varies. Fig. 10.18 shows printed lines on glass substrate using metal organic silver ink. The line width could be reduced to 14 µm, which is far less than the 100 µm nozzle outlet diameter. The adhesion of the deposited line after annealing was tested (scotch tape test). Although the adhesion of most nano-particle inks and metal organic inks on glass substrates after sintering was good, it was not sufficient on silicon surfaces. This is why modified commercially available screen-printing pastes were tested. Despite the relatively large particle size up to about 7 µm (see Fig. 10.10), the paste was successfully printed with the metal aerosol jet system. After conventional contact firing in a rapid thermal firing furnace, the adhesion to the silicon surface was of high quality. Unexpectedly the thin deposited layer even 164 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 10.18: Microscope pictures of metal aerosol jet printed lines on glass substrate using a metal organic silver ink and a nozzle with a 100 µm outlet diameter. etched through the SiNX antireflection coating during the firing process, yielding a good electrical contact. A cross section SEM image of a plated contact on a non-textured silicon solar cell is presented in Fig. 10.19-a. The seed layer width is about 35 µm and the plated silver height about 10 µm. A topographic picture of a further contact finger printed with the metal aerosol jet technique and plated by the light-induced silver plating process is shown in Fig. 10.19-b. The seed layer had a width of about 30 µm and a height of less than 1 µm. After the plating process the contact was about 66 µm wide and 20 µm high. a) b) Fig. 10.19: a) SEM cross section image of an aerosol jet printed contact plated with silver. b) Topographic image of another printed and plated contact with a seed layer width of about 30 µm. Metal aerosol jet printing 165 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 10.5 Solar cell processing and results 10.5.1 First aerosol jet printed solar cells using nano particle inks The first solar cells ever processed with the metal aerosol jet system were printed with a nickel-containing ink and a silver ink designed for ink-jet applications (Fig. 10.20). The efficiencies for these untextured cells of 9% (silverink) and 12% (nickel ink) were quite promising (Table 10.3). The ink with nickel particles even etched through the antireflection coating. However, due to the possibility of using modified silver screen-printing paste which is already optimized for solar cell application, the emphasis for further solar cell processing was based on those inks, as presented in the following section. Table 10.3: IV-parameters of first solar cells processed with the metal aerosol jet system. The low short-circuit current density is due to the non-textured surface and the conservative grid design shading about 15% of the solar cell surface. (ARC: antireflection coating) ink ARC no yes no modified Ni yes ink-jet Ag Voc jsc FF 2 [mV] [mA/cm ] [%] η [%] 600 594 598 21.5 2.6 20.7 71.5 26.4 71.5 9.2 0.4 8.9 612 27.8 69.6 11.8 Fig. 10.20: Picture of first processed solar cell. 10.5.2 Multicrystalline silicon solar cells 15.6 cm x 15.6 cm multicrystalline silicon wafers with a thickness of 220 µm were processed by an industrial partner up to the SiNX-layer with an emitter sheet resistance of about Rsh = 55 Ω/sq and an iso-textured surface. At Fraunhofer ISE the rear side of the wafers was conventionally screen-printed and dried. Then on the front side the contact layer was deposited using the metal aerosol jet system. For printing, a commercially available silver screen-printing paste, modified in its viscosity, was used. On each 15.6 cm x 15.6 cm wafer several grids with a size of 5 cm x 5 cm were printed. The busbar was created by printing several lines adjacent to each other with a slight overlap. Afterwards the cells were fired in an inline fast firing furnace at temperatures between 750°C and 900°C. The formation 166 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ of the second layer, the highly conductive layer, was performed by light-induced silver plating. Optionally, an annealing step at 400°C for 10 min in forming gas ambient followed. The solar cells were finally separated by rear side laser scribing and breaking. A picture of a solar cell is shown in Fig. 10.21 (right-hand). Fig. 10.21 Left-hand: Microscope picture of contacts using different printing settings resulting in line widths of 70 µm and 160 µm after silver plating.: Right-hand: 5 cm x 5 cm sized multicrystalline silicon solar cell with metal aerosol jet printed contacts thickened by the lightinduced plating process. Using the same tip of an outlet of 200 µm diameter, metal lines with a width between 50 µm and 140 µm were deposited on different wafers. After the fast firing process step, the contacts were thickened for 10 minutes by the light-induced plating process resulting in a finger height between 10 µm and 15 µm. The total contact width increased by approximately 20 µm to 30 µm. Microscope pictures of plated contact fingers are presented in Fig. 10.21 left-hand. The IV-parameter of solar cells with 70 µm and 160 µm wide contact lines (after plating) before and after the annealing step are presented in Table 10.4. The lower jsc for the solar cell with the wider finger can be explained with the larger shaded area. The 25 fingers with a width of 160 µm shade about 8% of the 5 x 5 cm² sized solar cell surface compared to 3.5% for the solar cell with 70 µm wide fingers. This is the main reason for the about 4.5% lower jsc value. The low fill factor of the cell with 70 µm broad contacts is probably caused by the low parallel resistance of rp = 700 Ω cm². Nevertheless, the series resistance for the cells with the thinner fingers is slightly higher (∆rs = 0.3 Ω cm²). The pastes need Metal aerosol jet printing 167 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Table 10.4: IV-parameters of the best 5 cm x 5 cm sized multicrystalline silicon solar cells measured before and after the annealing step. The best screen-printed “reference” solar cell with a size of 15.6 x 15.6 cm² achieved a short-circuit current density of 33.5 mA/cm², a fill factor of 76.0% and an efficiency of 15.8%. Cell-ID Process step Voc jsc [mV] [mA/cm2] η FF [%] [%] rs rp [Ω cm²] [kΩ cm²] 160 µm finger width after plating 83.2 after annealing 619 617 32.7 32.7 78.4 79.4 15.8 16.0 0.7 0.5 12.3 12.2 70 µm finger width after plating 93.6 after annealing 619 618 34.0 34.2 76.2 77.4 16.1 16.4 1.0 0.8 0.7 0.7 to be optimized to achieve a low contact resistance to the emitter surface, which is especially necessary for fine-line printed contacts. The solar cells were characterized before and after the annealing step. For both cells the efficiency increased by 0.2% to 0.3% absolute after the annealing step, caused by a gain in the fill factor and a reduction of the series resistance. It is assumed that the contact between the deposited and the plated metal is improved or the plated silver has a direct contact to the silicon surface. The effect of the FF improvement will be further investigated in Chapter 11. The appropriate annealing temperature and time depends on the one hand on the emitter profile, on the other hand on the deposited ink. For a contact of good quality a high annealing temperature is preferable, but this increases the danger of damaging the space charge region at the same time. For the processed cells the optimum annealing temperature was between 300°C and 400°C for a time of about 5 minutes to 10 minutes. 10.5.3 Monocrystalline silicon solar cells In a further experiment four large-area Cz-silicon solar cells have been processed with metal aerosol jet printed front contacts. These wafers had remained from another batch that was used for optimizing the hotmelt screen-printing process. For the hotmelt experiment 50 wafers were processed on 12.5 x 12.5 cm², 1 Ω cm, boron-doped Cz-silicon wafers with a thickness of 250 µm. The cells exhibit a textured surface covered by a SiNX antireflection coating and an emitter 168 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ sheet resistance of Rsh = 45 Ω/sq. The rear side was conventionally screen-printed and dried. Most of the wafers were printed with hotmelt paste, a few were screenprinted with conventional paste and more than a month later four remaining wafers were printed with the metal aerosol jet system. For comparison, the hotmelt screenprinted and conventional screen-printed cell results will be briefly presented. These cells were fired in an inline fast firing belt furnace. While for the hotmelt cells a wide temperature variation was performed, the conventional screen-printed cells were fired at the standard firing temperature for that paste. The best of the 35 solar cells with hotmelt printed contacts achieved an efficiency of η = 17.9%, while the average efficiency of eight cells fired at optimum temperature was ηavg = 17.5% (see Table 10.6). The relatively low efficiency of the best conventional screenprinted cell of η = 17.2% is probably due to a slightly too low firing temperature resulting in a FF of just 77% and an open-circuit voltage of 619 mV. For the front side grid of the metal aerosol jet printed cells, a final contact width of 70 µm and a height of 12 µm was assumed, so that in combination with other assumed cell data (e.g. jmpp, Vmpp) an optimum finger separation distance of s = 1.9 mm was calculated by grid simulation (see Chapter 3), compared to the hotmelt printed cells having a separation distance of s = 2.3 mm. For each paste the contact layer of one wafer was single printed, the other one triple printed, with the purpose to reveal the effect on the contact height. The printing speed was set to vprint = 20 mm/s. This led to a total printing time of about 15 minutes and 45 minutes for a single and triple printed contact grid, respectively. Half the time was necessary to print the two 1.5 mm wide busbars. The finger height hf after the firing process was between 1 µm and 2 µm for the single printed contact layer and between 3 µm and 6 µm for the triple printed one. Two different pastes were used for printing. Paste “B” was the same paste as used for the multicrystalline cells (see above) with the addition of 1% by mass phosphorus powder. The idea was to form a higher doping concentration directly under the contact, in order to achieve good adhesion and low contact resistivity to the emitter surface [147,148]. Paste “A” was a different silver screen-printing paste (from another manufacturer), which was modified in its viscosity. To keep full control of the optimum firing condition, the wafers were fired in a lab-type fast firing furnace, the front contacts were thickened by light-induced plating and the wafer’s edge isolated by laser scribing and breaking. After IVmeasurement some solar cells were plated a second time to increase the finger Metal aerosol jet printing 169 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ conductivity and annealed under room ambience for 10 minutes at 300°C. Fig. 10.22 illustrates a processed solar cell. Fig. 10.22: Metal aerosol jet printed and plated solar cell of size 12.5 x 12.5 cm² achieving an efficiency of 18.3%. IV-parameters of the best solar cell measured at different steps in the process sequence are presented in Table 10.5. This cell constitutes the highest efficiency (η = 18.3%) and highest fill factor (FF = 81.0%) achieved within this work for a large-area silicon solar cell. This cell featured a single printed contact using paste B. Whether the addition of phosphorus to the paste had an effect, could not be clarified within this work. After the first light-induced plating process the cell was limited by the high series resistance of rs = 2.1 Ω cm². To increase the finger conductivity further, the fingers were thickened by the light-induced plating process a second time. The enlarged shaded area reduced jsc significantly, but due to the boost in fill factor by 9%abs, the efficiency reached a value of η = 17.8%. In Table 10.5: IV-parameters of the best 12.5 cm x 12.5 cm sized monocrystalline silicon solar cell measured at different steps in the process sequence. Cell-ID 36_3 Process Step Voc [mV] jsc [mA/cm2] FF [%] [%] LIP 625 36.5 71.3 16.3 2.1 10 2nd LIP Sintering 623 624 35.8 36.1 80.2 81.0 17.8 18.3 n.a. 0.4 10 8 η rs rp [Ω cm²] [kΩ cm²] 170 Metal aerosol jet printing _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ a following sintering step the fill factor was further increased, resulting in the highefficiency value of 18.3%. IV-parameters measured after the plating process of all four wafers are presented in Table 10.6. The second wafer metallized with paste B, featuring a triple printed contact layer, also achieved an increased FF of 80.7%. Compared to the single printed contact jsc is lower, because the triple printed contact is broader. The efficiency of η = 17.5% of the solar cell with a single printed contact layer using paste A was also on a high level. However, the parallel resistance rp and pseudo fill factor PFF were reduced, indicating increased recombination currents. Both, PFF and rp are further reduced for the triple printed contact layer. In combination with the high series resistance, the measured efficiency for this cell was low. It is assumed that paste A has a more aggressive etching behavior and damages the space charge region, which might be avoided by reducing the firing temperature for this paste. Nevertheless, a single printed contact layer featuring a thickness of 1 µm to 2 µm was sufficient to achieve a contact of good electrical quality. Table 10.6: IV-parameters of the processed 12.5 x 12.5 cm² sized monocrystalline silicon solar cells. Paste “B” is the same paste as used for the multicrystalline silicon solar cells with the addition of 1% red phosphorus by mass. Paste “A” is a conventional silver screen-printing paste modified in its viscosity from a different manufacturer. Hotmelt and conventional screen-printed (SP) cells were processed on wafers from the same batch. Cell-ID Paste Prints Voc jsc [mV] [mA/cm2] η FF [%] [%] rs rp PFF [%] [Ω cm²] [Ω cm²] 36_4 A single 623 35.3 79.5 17.5 81.9 0.6 4 000 36_1 36_3 36_2 A B B triple single triple 623 624 622 34.5 36.1 35.2 68.1 81.0 80.7 14.7 18.3 17.7 80.9 82.5 82.4 2.6 0.4 0.5 700 10 000 8 000 624 36.2 79.1 17.9 36.0±0.1 78.1±1.2 36.1 76.7 Best hotmelt cell Average of 8 hotmelt cells 624±1 Best conventional SP cell 619 17.5±0.3 17.2 Metal aerosol jet printing 171 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 10.6 Chapter summary Metal aerosol jet printing in combination with a light-induced plating process has been developed for creating the front side contact layer of the proposed two layer contact structure within this work. Opposed to most other metallization technologies the metal aerosol jet system is a non-contact direct-write technique, making it suitable for printing on uneven surfaces and thin and fragile wafers. However, the biggest advantage of this technology lies in the working principle of the deposition head: The metal aerosol stream is surrounded by a second gas stream and then focused through the nozzle onto the substrate. This prevents clogging of the tip. The width of the deposited line is much smaller than the size of the nozzle outlet. Lines of 14 µm width were printed using metal organic inks and a nozzle with an outlet diameter of 100 µm. Up to date a lot of improvements of the system have been applied by Optomec as well as by Fraunhofer ISE. Especially the uptime and the process stability of the machine could be significantly increased. For industrial mass production a multi nozzle system is required. In order to achieve a throughput rate comparable to screen-printing systems, a printing speed of 50 mm/s to 100 mm/s is required, assuming that each finger of the wafer is printed by one individual nozzle of the multi nozzle system. Within this work 12.5 cm x 12.5 cm textured monocrystalline silicon solar cells with a standard Al-BSF have been successfully processed with a modified commercially available screen-printing paste. The best solar cell processed achieved an efficiency of η = 18.3% at a fill factor of FF = 81.0%. Nevertheless, the full potential of the metal aerosol jet system will be achieved when using optimized inks giving a low contact resistivity to a lowly doped silicon surface. 172 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 11 Microstructure analysis of contacts The main focus of this chapter is the analysis of fine-line printed and plated contacts on a macroscopic as well as on a microscopic level. In the first part of this chapter, the current understanding of the contact formation process for screen-printed contacts is reviewed. In the second and third part, the dependencies of the contact resistance on the contact geometry and on the plating process will be discussed, which is based on electrical and optical investigations. The results are summarized and possible new current paths between semiconductor and contact demonstrated. 11.1 Review of current models for screen-printed contact formation As presented in Chapter 2.2.2 a thick-film contact consists of a porous silver structure, a glass layer and silver crystallites at the interface between silicon surface and contact material. Within the last years, the understanding of the contact formation process of screen-printed contacts has been extensively improved [37,40,132,149-155]. Among them, Schubert [40] presents a detailed analysis incorporating the models and investigations of the above mentioned references. In his experiments he separated competing processes for contact formation, analyzing the interaction between glass, silver, SiNX and lead in all relevant combinations for the formation process. The actual contact formation process occurs during the firing step at temperatures of 500°C to 850°C (see Fig. 2.4 of Chapter 2.2.2). The basic process of the model is divided into three temperature ranges and the cooling down phase (Fig. 11.1). T<550°C: Organic components and solvents are completely decomposed. Initial sintering of silver particles may start and the viscosity of the glass frit is reduced (Fig. 11.1-a). 550°C<T<700°C: The viscosity of the glass frit is further reduced and the glass completely wets the interface layer between SiNX and silver particles. Rapid silver particle sintering occurs resulting in a porous structure of the bulk silver. Additionally silver is dissolved in the glass and penetrates significantly into the SiNX-layer at temperatures above 625°C. A metal oxide (mainly lead oxide) contained in the glass frit reduces the SiNX (Fig. 11.1-b). Microstructure analysis of contacts 173 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 700°C<T<800°C: At locations where the SiNX layer is completely penetrated, a redox reaction between the lead oxide and the silicon occurs (2PbOglass + Si2 Pb + SiO2 [132]). Epitaxial growth of silver crystals in silicon is assumed to start. Two processes are likely, silver growth from silver saturated glass and/or via a liquid lead phase (Fig. 11.1-c-e). Fig. 11.1: Simplified model of contact formation as published by Schubert [40]. Cooling down: Silver and lead separate. Silver recrystallizes epitaxially along the <111>-planes resulting in inverted pyramids, lead is found at the bottom side of the silver fingers (Fig. 11.1-f). The amount of crystals depends on the surface doping concentration and the firing temperature [132]. With increasing doping concentration and/or ascending firing temperature, the amount of crystallites is enlarged. The theory of a highquality contact formation due to the epitaxially growth of the crystallites is generally accepted [156]. Huljic and Ballif [151] determined the specific contact resistance of a silver crystallite in contact to an n-emitter having a sheet resistance of Rsh = 30 Ω/sq to be between 4·10-8 Ω cm² and 2·10-7 Ω cm² by conductive AFM measurement [150]. Values in the same order of magnitude were determined by Schubert [40], measuring the contact resistivity to a highly doped emitter and by determining optically the crystallite density and crystallite geometry under the 174 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ contact. The limiting factor to achieve a low contact resistance between the silicon surface and the bulk silver is the non- or very low conductive glass layer [157], which is located in-between. For low contact resistance, either the glass layer needs to be very thin (wglass < 2 nm) allowing a quantum mechanical tunneling of electrons [158], precipitates in the glass layer allow multiple quantum mechanical tunneling of electrons [159], or the glass layer is locally completely removed, so that the silicon surface or silver crystallites are in direct contact with the bulk silver [159,160]. The latter assumption was made by many authors, but it is very difficult to experimentally observe this direct contact. Fig. 11.2 shows an SEM image of a cleaved edge of a contact finger prepared in this work. The crystallite (in the front) shows the same geometry as described in most microstructure analyzing publications: structure of an inverted pyramid facing with the top to the n-silicon side and the roundish bottom towards the finger. The glass frit is interrupted above the crystallite. Either the missing part of the glass frit is located at the fractured counter-part or the bulk silver was in direct contact with the crystallite. The piece pointed out in the background is most probably also a silver crystallite. This crystallite is in direct contact to the bulk silver and was not separated by the cleaving process. A direct connection between crystallite and bulk silver would result in a very low contact resistance. If the contact area covered by crystallites is equal to Acrystallite = 1% and if the area of the crystallites in direct contact to the bulk bulk silver glass f rit cryst allit e silicon Fig. 11.2: SEM image of a cleaved edge across a contact finger, showing crystallites, glass frit, bulk silver and the silicon surface. Microstructure analysis of contacts 175 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ silver Adirct_contact is 1%, assuming a microscopic contact resistivity of ρc,micro = 2·10-7 Ω cm² [150], would theoretically yield in a macroscopic contact resistivity of ρc,macro = 2 mΩ cm². ρ c,macro = ρ c,micro ⋅ Acrystallite ⋅ Adirect _ contact (11.1) In practice the number of crystallites in direct contact needs to be higher to avoid current crowding effects of the point contacts [28,161]. 11.2 Solar cell results of pad-printed and plated contacts 11.2.1 Design of experiment The results presented in this chapter are mainly based on a batch specially processed for analyzing the metal-semiconductor contact. This investigation was carried out on 25 Czochralski <100> wafers of size 10 cm x 10 cm, having a boron base doping ρb of 1 Ω cm - 2 Ω cm. After performing a saw damage etch, the batch was divided into two groups, receiving an n-emitter diffusion of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq, respectively. Both groups obtained a sputtered SiNx antireflection coating and an aluminum screen-printed rear side. On the front side of the wafers a 5 cm x 5 cm solar cell front side grid and test structures for line and contact resistivity measurements were printed using the pad-printing technology. The number of prints to form one contact was varied between 1 and 4 prints. All wafers were fired in a rapid thermal single wafer reactor at a peak firing temperature of 770°C. This is the optimum firing temperature for the flat wafers and the hotmelt paste as determined in a previous experiment. The different structures on the front side were separated using a dicing saw. The solar cells were characterized after firing and plating, while the contact resistivity was measured after the firing, the plating and the sintering process. The print design was made up of the following structures (see Fig. 11.3): 1. Solar cell front side grid of size 5 x 5 cm² with 80 µm broad fingers: The purpose of the solar cell grid design was the ability to directly compare IVparameters to contact resistivity measurements. Relatively broad fingers of 80 µm were printed to ensure a constant printing image and to minimize processing problems. To ensure a stable and reproducible process was of major importance. 176 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 2. Contact resistivity determination structures (see Fig. B.2 in Appendix B) with varying finger width (wc = 20 µm - 100 µm): The aim was to measure the contact resistivity in dependence of the finger width. A previous experiment confirmed that the optimum firing temperature is independent of the finger width. 3. Line conductivity determination structures of different finger width (wc = 20 µm - 110 µm): As the line conductivity is not of major importance for the contact layer, it was just ensured that the printed line was not interrupted. For lines with a finger width of wc = 30 µm or broader, the fingers were continuous. structure for contact resistivity ρc measurement (TLM) structure for line conductivity measurement 5 x 5 cm² solar cell front grid 5 mm separation cut for ρc measurement Fig. 11.3: Picture of the test structure used for the experiment featuring a solar cell front side grid of 5 cm x 5 cm in the centre. In addition, line conductivity and contact resistivity determination structures of different finger width were placed around the cell. 11.2.2 IV-parameters The IV-parameters of the 5 x 5 cm² solar cell were determined after the firing and after the light-induced plating process. In Fig. 11.4 the mean IV-parameters are illustrated for all print repetitions. Cells with an emitter sheet resistance of Rsh = 40 Ω/sq: The fill factor for a single printed contact showed relative low values prior the plating process due to insufficient line conductivity. After the plating process, the fill factor for a single printed contact was increased up to FF = 76%, however, not reaching the same level as for a multiple printed contacts. After a second light-induced plating step the fill factor was further increased to FF = 77%. The remaining loss the in fill Microstructure analysis of contacts 177 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 80 15 75 14 70 65 60 after LIP before LIP 1 2 33 4 2 1 Number of prints 3 13 12 11 30 28 0 4 2 3 1 Number of prints Rsh = 55 Ω/sq Rsh = 40 Ω/sq 4 2 1 Number of prints 3 after LIP before LIP 2 1 2 2 3 29 1 1 4 rs [Ω cm²] 31 after LIP before LIP 5 after LIP before LIP 32 jsc [mA/cm²] Efficiency η [%] Fill factor FF [%] factor might be due to the non-optimal print image, but also might be caused by an elevated contact resistance (see Chapter 11.3). As expected from the fill factor values, the series resistance was significantly reduced after the plating step. The short-circuit current density jsc reached maximum values for the single printed contacts. In this experiment the positioning accuracy of the pad-printer was very low, so that multiple prints resulted in contact broadening instead of an increased height. Hence the shaded area was significantly enlarged, reducing the short-circuit current density, the reason why the maximum efficiency is reached for single printed contacts. Cells with an emitter sheet resistance of Rsh = 55 Ω/sq: The trend of the IVparameters was similar as for the cells with the higher doping concentration. Main differences were a higher short-circuit current density and a lower fill factor prior 2 3 4 1 Number of prints Rsh = 55 Ω/sq Rsh = 40 Ω/sq 1 2 Fig. 11.4: IV-parameters and series resistance rs of pad-printed cells plotted versus number of consecutive prints before and after the light-induced plating process (LIP). A sheet resistance variation of Rsh = 40 and Rsh = 55 Ω/sq was performed. The open-circuit voltage of all solar cells was in the range of Voc = 605 mV - 615 mV, not showing any dependence on the investigated parameters. 178 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ the plating step for the single printed cells. Improved internal quantum efficiency in the short wavelength region caused by lower emitter recombination explained the higher current density. The low fill factor was caused by a high series resistance of rs > 4.5 Ω cm². If solely the poor line conductivity was responsible for the elevated series resistance, a value comparable to the cells of Rsh = 40 Ω/sq emitter would be expected. Hence an additional significant contributor to the series resistance rs needed to exist, which was the contact resistiance. However, after the plating step the fill factor FF and series resistance rs improved, reaching the same level as for the cells with single printed contacts and Rsh = 40 Ω/sq. This means, that the contact resistance was improved by the plating step, which was proven by contact resistance measurements (see Chapter 11.4) In summary, the processed solar cells were in the normal efficiency range for the type of cell structure used. Notable is the lower fill factor for the single printed contacts even after plating compared to multiple printed ones. The width and/or height for a single printed contact are less than for a multiple printed one. Hence in a further investigation the dependence on the contact resistance on the contact width and height was analyzed. 11.3 Dependence of contact resistance on contact width and height 11.3.1 Contact resistance measurement The cells from the batch presented in the previous section were further analyzed in respect to the contact and sheet resistance. Contact and sheet resistance measurements were performed after the firing process using the test structure proposed by Meier and Schroder [101] (see Fig B.2 in Appendix B). In combination with the end resistance measurement [162] the transfer length LT and the contact resistivity ρc were determined (see Appendix B). Although the contact resistivity is from a physical point of view the more preferred characterization value, the directly extracted contact resistance Rc is used in this work. As discussed in Appendix B the advantage of a directly measured value is to be independent of a second measurement (the end resistance measurement) and a theoretical calculation, necessary to finally determine the contact resistivity. Furthermore the model of the transfer length is based on the assumption that the contact layer is Microstructure analysis of contacts 179 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Contact resistance Rc [Ω] homogeneous over the contact width. This is true for e.g. evaporated contacts; however for thick-film contacts the metal–semiconductor contact differs strongly over its width wc. The SiNX-layer is not completely etched, the shape and density of silver crystallites differs locally and the thickness of the glass layer varies. After the printing process the contact width wc of a single print was about 10 µm to 40 µm broader than the width in the printing plate. As the positioning accuracy of the pad-printer was low, multiple prints resulted rather in an increased contact width than in an increased height as desired. This was especially the case for printed lines below 80 µm, and was the reason why the emphasis was put upon comparing single printed contacts of different widths among each other. Fig. 11.5 shows the contact resistance measurements to the emitter featuring a sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq. Two trends can be observed correlating with expectation from theory. The contact resistance between metal and emitter surface is higher for the less doped emitter and the resistance first stays constant for relatively broad lines (transfer length LT < contact width wc) and rises when the transfer length LT exceeds the contact width. 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 Rsh = 55 Ω/sq Rsh = 40 Ω/sq 40 60 80 100 120 140 160 Contact width wc [µm] Fig. 11.5: The contact resistance Rc between a single pad-printed contact and two different emitter sheet resistances for solar cell processing is plotted versus the contact width. When comparing the measurements with theoretical curves, it seems that the contact resistance rises disproportionally with decreasing width. Fig. 11.6 illustrates the measured points and calculated curves for an emitter sheet resistance of Rsh = 40 Ω/sq. It can be observed that the contact resistivity rises with decreasing finger width. The resistance was calculated applying equation (11.2) 180 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ assuming a constant emitter sheet resistance Rsk of 40 Ω/sq under the contact (see Appendix B). Rc = ρ c Rsk Rsh l c Rsh wc coth ρ L c T (11.2) Contact resistance Rc [Ω] As mentioned above, the investigation is based on a hotmelt screen-printing paste, not optimized for flat and thin fingers as obtained by the pad-printing technology. To achieve a better understanding why the contact resistivity rises with decreasing finger width, microscope analyses were performed. 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 ρc = 2.8 mΩ cm² 2.4 mΩ cm² 1.8 mΩ cm² 1.2 mΩ cm² ρc: contact resistivity 40 60 80 100 120 140 160 Contact width wc [µm] Fig. 11.6: Measured (marked points) and calculated contact resistance (solid lines) in dependence on the contact width for Rsh = 40 Ω/sq. The calculated contact resistance curves are based on eq. (11.2). 11.3.2 Optical analysis of the interface layer The contact height is reduced with shrinking contact width. The hypothetical explanation of the increased contact resistivity is, that the height and width of the printed structure influences the contact formation process. The height profiles of different contacts were measured. In a following step the silver contact was removed by a wet chemical process, resulting in an imprint of the silver contact. The silicon surface as well as the SiNX layer was not affected by the process, whereas the silver, glass frit and silver crystallites were removed. The imprint of the finger, consisting of partly removed SiNX and of crystallite imprints, was optically investigated using a scanning electron microscope (SEM). The Microstructure analysis of contacts 181 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ b) a) c) Fig. 11.7: A finger imprint of an etched contact is illustrated in the left-hand SEM image. Enlarged images were taken over the contact width wc (b) and for each image the amount of crystallites and the etched SiNX area were determined using appropriate filters (c). fraction of the etched SiNX-layer and the amount of the silver crystallites were determined in order to analyze any dependence to the contact height. As illustrated in Fig. 11.7-a the finger imprint was divided into small areas across its width. From each area, an SEM image was taken and analyzed using computer software. Fig. 11.7-c shows the resulting computer image to determine the etched area. This black/white picture is achieved by applying filters to the original one (Fig. 11.7-b) and setting a threshold value in the grayscale distribution. Pixels above the threshold value are transformed into white pixels and below the value into black ones. A similar approach was applied for retrieving the amount of crystallites. By this means, the amount of crystallites and the area of SiNX that was etched, were determined as well as the correspondent contact height. Fig. 11.8 shows the result of the evaluation for a single printed contact. A correlation between both, the crystallite density and the etched silicon surface is clearly visible. At the edges of the finger where its height was smallest also the etched SiNx area and crystallite density was lowest, rising with increasing contact height. 182 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 6 finger height hf 0.8 etched surface area rSi/SiNX 5 number of crystallites per unit area ncr 4 0.6 -2 ncr [µm ] 3 0.4 2 0.2 0.0 1 0 20 40 60 Contact width wc [µm] Finger height hf [µm] rSi/SiNx 1.0 0 Fig. 11.8: Number of crystallite per unit area ncr (number of crystallites per µm²), ratio of etched SiNX area (rSi/SiN) and finger height hf plotted versus the contact width wc. The etched area and the crystallite density rose with increasing finger height. To improve the accuracy this investigation was repeated for 13 samples. The etched surface area, the density of crystallites and the corresponding finger profile were determined. Fig. 11.9 illustrates the measurements achieved in the middle of a finger imprint. Both, the crystallite density ncr and the etched SiNX area rSi/SiN rose with increasing finger height. 80 -2 ncr [µm ] 0.4 60 0.3 40 0.2 ncr 0.1 0.0 1 rSi/SiN 2 3 4 Finger height hf [µm] 20 0 5 Ratio Si/SiNX rSi/SiNx [%] 0.5 Fig. 11.9: Crystallite density ncr and fraction of etched area rSi/SiNx plotted versus the finger height hf demonstrating that etched area and crystallite density rose with increasing finger height hf. In addition the ratio of Si/SiNX (rSi/SiN) was plotted versus the crystallite density (see Fig. 11.10). The crystallite density has a more or less linear dependence on the etched SiNX area for wafers fired at the same temperature and having the same Microstructure analysis of contacts 183 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ emitter sheet resistance. This can be explained as crystallites can be formed just at places where the glass frit is in direct contact to the silicon, which means at places at which the SiNX layer is completely penetrated. Due to the similar dependence of silver crystallite density and etched SiNX area on finger height, it is difficult to conclude whether the crystallite or the direct silver-silicon contact is responsible for the main current flow. Ratio Si/SiN rSi/SiN 0.8 0.6 0.4 0.2 0.0 0.0 0.1 0.2 0.3 -2 Crystallite density ncr [µm ] Fig. 11.10: Ratio of etched SiNX area plotted versus the number of crystallites per unit area. The density rises with increased etched surface. The etch rate and number of crystallites per unit area depend on the applied firing temperature, the surface doping concentration and on the paste composition. A deeper correlation analysis between finger height, crystallite density and etched area has not been performed. The target of this investigation was to show why the contact resistivity ρc rose with decreasing finger width wc. The reason was the insufficient etch ratio of the SiN-layer. The pastes used for these investigations were designed for screen-printing purposes, expecting a significantly higher contact finger. As it is believed that the glass frit completely diffuses to the metalsemiconductor interface [40], the amount of glass frit in the paste for a reduced contact height is not sufficient. It is strongly recommended to increase the glass frit amount for fine-line pad-printed contacts. A direct dependence between glass frit amount and etching rate could not be proven within this work since the paste manufacturer have not been willing to reveal sufficient information on the composition of the pastes. 184 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 11.4 Influence of a plating step on the contact resistance Among other investigations, also the previous one led to the hypothesis that the contact resistance was improved by the plating process. For example in the previously discussed section the series resistance of processed cells featuring an emitter sheet resistance of Rsh = 55 Ω/sq was significantly higher than the cells of Rsh = 40 Ω/sq (∆rs > 1.5 Ω cm²). This was mainly attributed to a high contact resistance. The increase in the emitter resistance due to the emitter layer would only account for ∆rs ≈ 0.1 Ω cm². After the plating step the series resistance of the lowly doped emitter cells was reduced to nearly the same low value as for the cells with an emitter sheet resistance of Rsh = 40 Ω/sq, which was reasoned to be due to an improved contact resistance. As this effect was unexpected an investigation was performed to prove this assumption. Contact resistance measurements were performed at different steps in the process sequence. The results together with optical investigation are presented in the following. 11.4.1 Electrical measurements To confirm that the contact resistance was improved, measurements were performed before and after the plating process. As illustrated in Fig. 11.3 two parallel cuts with a separation distance of 5 mm were sawn into the front surface, separating the test structure from the remaining emitter layer. In Fig. 11.11 the difference of the contact resistance ∆Rc before (Rc,before) and after (Rc,after) the plating step is plotted versus the initial resistance value (Rc,before). Rc,before-Rc,after [Ω] 0.6 0.4 0.2 improvement 0.0 -0.2 0.0 deterioration 0.2 0.4 0.6 0.8 1.0 1.2 Correspondent Rc,before [Ω] Fig. 11.11: The absolute change in contact resistance (Rc,before - Rc,after) due to the plating process Microstructure analysis of contacts 185 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ is plotted versus the initial resistance. Contact resistance Rc [Ω] The measurements indicate that the plating step had a positive influence on the contact resistance. But while contacts that had a low resistance before plating were not further improved (or just marginally), the improvement of contacts having a relatively high resistance was significant. To further analyze the effect of the plating process the same measurements were plotted versus the contact width wc distinguishing between the emitter sheet resistance of Rsh = 40 Ω/sq and Rsh = 55 Ω/sq (see Fig. 11.12). Especially for contacts of relatively small width wc a significant improvement of the resistance was achieved. Broad contacts, which were already on a low level before the plating process, could be just slightly improved. Additionally, as expected from theory, the contact resistance of less doped emitter is higher (compare Fig. 11.5). 1.2 40 Ω/sq before LIP after LIP 55 Ω/sq before LIP after LIP 1.0 0.8 0.6 0.4 0.2 0.0 50 100 150 200 250 Contact width wc [µm] Fig. 11.12 The contact resistance before and after a plating step is plotted versus the contact width. In addition the effect of a sintering process under ambient air on the contact resistance was investigated. The contact resistance measurement was performed before and after a sinter step of pad-printed contacts and pad-printed contacts already thickened by a layer of plated silver. Fig. 11.13 shows the absolute change in resistance. It seems that a sinter step of plated contacts further improved the contact resistance, although the resistance was already on a very low value before sintering. Contacts that were not thickened by silver plating were not significantly influenced by the thermal step. And again it seems that contacts with a higher contact resistance were improved while those with an already lower resistance remained unaffected or even slightly increased. To reveal a significant trend further research with more samples are necessary 186 Microstructure analysis of contacts ∆Rc = Rc,before - Rc,afer [Ω] _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 0.6 sintering of fired contact sintering of fired and plated contact 0.4 0.2 improvement 0.0 deterioration -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Correspondent Rc,before [Ω] Fig. 11.13: Effect of a sintering step on the contact resistance for a plated and non plated contact. The absolute change in contact resistance ∆Rc is plotted versus the initial resistance. 11.4.2 Micro-structural investigations The improvement of the contact resistance by the light-induced plating process was revealed by electrical measurements. However, these measurements did not explain why the contact resistance was reduced. An improvement of the contact resistance would mean that the plated silver creates a new current path between the silicon surface and the bulk silver with a sufficiently low resistance. This is at first glance unexpected. Different current flow paths were considered: a) The porous thick-film contact was completely infiltrated by the silver solution and a direct contact or a tunneling contact between plated silver and silicon surface and/or silver crystallite was created (Fig. 11.14-a). b) At the edge where the thick-film contact was thinnest, plated silver fills the porous structure establishing a contact between silicon/crystallite and plated silver (Fig. 11.14-b). c) The SiNX layer in the vicinity of the printed contact had pinholes. Plated silver deposited on the antireflection coating had a direct or tunneling contact to the silicon (Fig. 11.14-c). During these investigations a new effect was observed. Next to the printed contact structure the SiNX antireflection coating was covered by a layer of glass, as confirmed by EDX measurements. This region was about wedge = 5 µm broad and consisted of holes and precipitates (presumably silver or lead, see Fig. 11.15-a). To achieve a topographic image the layer was further analyzed using an Atomic Microstructure analysis of contacts 187 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 11.14: Schematic drawing how a thick-film contact may look like after silver plating. a) The porous thick-film contact is completely infiltrated by the silver solution. b) The thick-film contact is partly infiltrated by silver solution. c) The SiNX layer has pinholes. Plated silver has a direct contact to the silicon. Force Microscope (AFM19). The height profile of the layer across one of the “black dots” was measured (see Fig. 11.15-b). This measurement confirmed that the “black dots” were holes. One of the larger holes had a cross section diameter of whole = 200 nm. The hole was at least 200 nm deep, but it is also likely that it reached down to the silicon surface as the AFM might be limited in depth resolution for such small structures. The following two explanations or a combination of both are proposed for the formation of the edge layer: 1. Glass-frit leaked out of the contact during the firing process and etched the SiNX layer. 2. The contact shrunk in width during the firing process. Glass-frit was left behind and etched the SiNX layer. Topographic measurements of screen-printed contacts showed that the contact did not just shrink in height, but also in width during the firing process. Thus, it is 19 AFM (Atomic Force Microscope): The attractive or repulsive force between a sample surface and a sharp tip, which is attached to a cantilever, is measured. Using a laser the displacement of the cantilever is measured as the tip is scanned across the sample, leading to surface visualisation. 188 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 11.15: a) Top view SEM image of the interface layer between thick-film contact and silicon nitride. Next to the contact an about 5 µm wide layer is covered by a layer of glass in which holes and precipitates were observed. b) The glass layer was further analyzed using an atomic force microscope. The topographic scan in the bottom right image confirmed that the “black dots” are holes. The width of one of the larger holes is about 200 nm and it is probable that it reaches down to the silicon surface. apparent that some of the fluid glass frit remained behind during the shrinking process. Nevertheless, glass-frit leaking out of the contact is also likely but could not be confirmed within this work. Fig. 11.16: SEM image of a finger cross section Fig. 11.17: Top view of an etched area close showing silver crystallites near the edge. to the contact, showing crystallite imprints. Microstructure analysis of contacts 189 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ A cross section image of a contact edge is presented in Fig. 11.16. The glass-frit wetted the layer next to the silver contact and silver crystallites from dissolved silver in the glass frit were observed. Most often crystallites were found at the edge zone of the glass layer as shown in the SEM image in Fig. 11.17. The exact composition of glass frit, silver particles and lead seems to be flushed out to the fringe of the glass layer to allow crystallite formation. Conventional screen-printed contacts were analyzed to ensure that this “edge effect” is not due to the hotmelt paste or the deposition technology applied. As illustrated in Fig. 11.18 the edge effect is also present for a conventional screenprinted contact. In addition the glass layer was found in the valleys of the random pyramids, which was a hint that the fluid glass frit also leaked out of the contact during the firing process. Fig. 11.18: SEM image of a textured silicon surface with a conventional screen-printed contact, confirming that the edge effect is not attributed to the different paste and process applied. Furthermore a screen-printed and plated solar cell featuring a sheet resistance of Rsh = 90 Ω/sq was optically analyzed. The fill factor of this solar cell was increased by the plating process from FF = 40% to FF = 78% (compare Chapter 8.3.2), which was traced back to a significant reduction in the contact resistance. The contact imprints to an emitter sheet resistance of Rsh = 90 Ω/sq were compared to those with an emitter sheet resistance of Rsh = 40 Ω/sq (see Fig. 11.19). Measurements were performed in the middle of the imprint of six samples for each type of emitter. While the penetrated area is similar, the amount of crystallite imprints to the higher sheet resistance emitter is far lower (about a factor of four). For these contacts, crystallite imprints were mainly detected on the fringe of the pyramids, while for the higher doped emitters crystallite imprints were detected on the fringe as well as on the side planes of the random pyramids. The reduced 190 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 11.19: Top view SEM images of finger imprint on textured monocrystalline silicon solar cells with a sheet resistance of Rsh = 40 Ω/sq (left-hand) and of Rsh = 90 Ω/sq (right-hand). Crystallite density and shape differ strongly, while the penetrated SiN area (dark regions) is relatively similar. crystallite density and the missing crystallites on the side planes of the crystallite might explain the high contact resistance and low fill factor for emitters with Rsh = 90 Ω/sq. In order to achieve a better understanding about possible current paths that might have been created by the plating process, a cross section of the contact finger was prepared. With the purpose of achieving a cross section of highest quality, focused ion beam (FIB20) machining has been performed at the Fraunhofer IWM in Halle. SEM pictures of the cross section are illustrated in Fig. 11.20. In Fig. 11.20a an overview of the cross section is presented. It shows the random pyramids textured silicon surface and the about 90 µm broad hotmelt screen-printed contact, which is covered by a 5 µm to 10 µm thick plated silver layer. Fig. 11.20-c and -d show the interface layer of the thick-film contact and the silicon surface. A layer of glass (confirmed by EDX measurements), which is at some locations very thin (wglass < 10 nm), completely wets the silicon surface. A direct (screen-printed) silver-silicon contact was not observed over the whole cross section, but at some locations the glass layer was very thin that maybe a tunneling of electrons was possible. Fig. 11.20-b illustrates the interface between plated silver and silicon surface. It seems as if a direct contact between both materials was created. 20 Focused ion beam (FIB) machining is a technique that removes microscopic portion of material from the surface by sputtering with accelerated and focused ions. The small diameter down to 7 nm of the focused ion beam makes it possible to achieve cross sections of high accuracy and quality. The sputtering process can be monitored using SEM. Microstructure analysis of contacts 191 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fig. 11.20-g shows a section of the right side of the contact. Clearly visible is the interface layer between the plated and the screen-printed silver. Fig. 11.20-e and -f illustrate the interface layer silicon-silver. Several precipitates are located in the glass layer, close to the emitter surface. However, silver crystallites could not be detected across the whole cross section (see also Fig. 11.19). As the fill factor strongly rises after the plating process, it is likely that the main current transport is achieved by a direct silicon-silver contact or a tunneling contact via a thin layer of glass or precipitates in the glass. Fig. 11.20: Cross section SEM images of a screen-printed and plated contact. The cross section was prepared by focused ion beam (FIB) machining. 192 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 11.5 Conclusion: New current paths for plated contacts In this section possible current paths for plated contacts are summarized, which are based on the previous investigations. As illustrated in Fig. 11.21, the porous screen-printed contact is covered by a thin layer of plated silver. It is likely that during the plating process some of the silver also fills the porous structure of at least the top layer of the contact structure. In the following seven different current paths will be discussed. The first four presented cases (see Fig. 11.21: 1-4) are due to the plated silver layer, whereas the last three cases correspond to a conventional screen-printed contact (see Fig. 11.21: 5-7): Plated silver in direct contact with silver crystallites (Fig. 11.21 case 1): During the firing process either glass leaks out of the printed contact structure and/or a layer of glass is left behind as the contact shrinks in width. This glass etches the SiNX layer. Mainly at the edges of this layer silver crystallites were observed (see Fig. 11.16 and Fig. 11.17). It seems as if the right constitution of silver, lead and glass is flushed out to the edge and crystallites are formed. It is also likely that Fig. 11.21: Schematic drawing of a plated thick-film contact, illustrating different connections to silver crystallites. 1: crystallite in direct contact with plated silver at the “edge zone”; 2: silicon in direct contact with plated silver at the edge of the contact; 3: high amount of precipitates in the glass layer creating a multiple tunneling contact 4: plated silver fills the porous structure at the edge of the screen-printed contact and has a direct contact to the silicon or silver crystallite; 5: crystallite covered by a thick layer of glass; 6: silicon/ crystallite in direct contact with screenprinted contact; 7: thin layer of glass in-between crystallite and screen-printed contact. Microstructure analysis of contacts 193 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ some of these crystallites are in direct contact to the plated silver. Due to the reported low contact resistance between crystallite and silicon surface [40,151], just a few crystallites in direct contact with the plated silver would be sufficient to reduce the contact resistance significantly. The contact resistivity of a crystallite to an emitter with a sheet resistance of Rsh = 30 Ω/sq was reported to be in the range of 7·10-8 Ω cm² to 2·10-7 Ω cm² [151]. If a crystallite is covered by a thin layer of glass also a tunneling contact might be established. Plated silver in direct contact with the silicon surface (Fig. 11.21 case 2): At locations at which the SiNX layer is completely penetrated and the plated silver has a direct contact to the silicon surface a current path is created (see Fig. 11.20-b). Assuming a barrier height of φB = 0.84 eV between silver and silicon and a surface doping concentration of NDS = 2·1020 cm-3 for an emitter featuring a sheet resistance of Rsh = 55 Ω/sq, field emission is the current transport mechanism, leading to a theoretical contact resistivity in the range of ρc = 1·10-6 Ω cm² - 1·107 Ω cm² (compare Fig. 1.16 of Chapter 1.3). As reported by e.g. Grupp et al. [132] also the silicon surface is slightly etched by the glass frit, which reduces the surface doping concentration of the underlying emitter. This would cause a rise of the contact resistivity. Even if the silicon surface is completely covered by a layer of glass, plated silver might be in direct contact to the silicon through holes in the glass layer, which have been observed by optical investigations (AFM and SEM, see Fig. 11.15). (Multiple) tunneling contact between silicon surface and plated contact over precipitates in the glass frit (Fig. 11.21 case 3): Optical analyzes also proved a large amount of precipitates in the glass layer next to the thick-film contact. Direct tunneling or multiple tunneling (via metal precipitates in the glass layer) of electrons from the silicon surface and/or silver crystallites to the plated silver might be possible. Plated silver infiltrates the contact at its edge and is in direct contact with a crystallite (Fig. 11.21 case 4): It is probable that silver ions out of the plating bath infiltrate at least the top layer of the porous screen-printed contact structure. At the edge where the contact is thinnest plated silver might be in direct contact to a crystallite and/or the silicon surface. The same current paths as described for case 1 might be formed. Current flow over the glass layer (Fig. 11.21 case 5): Crystallites are formed at places at which the SiNX layer is penetrated by the glass frit. Thus it is likely that a 194 Microstructure analysis of contacts _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ crystallite is covered by a layer of glass. The question that arises is, if a current flow over the glass is possible. Investigations have shown that the glass has a high resistance or is practically not conductive at all and consequently current transport is unlikely [163]. This is one of the reasons why the contact resistivity of about 1·10-3 Ω cm² for a screen-printed contact is far higher than the contact resistivity of a single crystallite, with reported values of about 2·10-7 Ω cm² [156]. Crystallite/ silicon surface in direct contact with the bulk silver of the thick-film contact (Fig. 11.21 case 6): A silver crystallite is in direct contact with the bulk silver of the thick-film contact. It is assumed that this kind of contact is responsible for the main current transport of a thick-film contact [132]. In addition it might be possible that the silicon is in direct contact to the screen-printed silver. Tunneling contact between crystallite and bulk silver of the thick-film contact (Fig. 11.21 case 7): After the firing process some crystallites might be covered by a very thin layer of glass. Quantum mechanical tunneling through the glass layer might be possible, if the layer is sufficiently thin (wglass < 2 nm). Multiple tunneling via metal precipitates in the glass layer is also possible [159]. All presented current paths might be created for a plated silver contact. However, as presented in the previous section also a sintering step of plated contacts seems to improve the contact resistance. Two effects are likely. Either the contact between the plated and the screen-printed silver is reduced and/or the actual contact interface between silicon/crystallite and plated silver is improved. Long term stability tests need to be performed to prove the contact resistance improvement over a long period of time. The main current flow is expected to be across the silver crystallites which are in direct or tunneling contact to the bulk silver. Hence a high number of crystallites is important. As has been demonstrated the number of crystallites depends on the etched surface area, whereas the etched surface area depends on the finger height. In addition the surface doping concentration of the emitter layer and the firing temperature influences the crystallite formation process. Both, an increased doping concentration and elevated firing temperatures enhance the crystallite formation process [132,164]. However, for the contact formation process, the glass frit plays an important role. The amount and the etching properties of glass frit in a paste need to be carefully adjusted. If the amount of glass frit is too high a thick layer of glass will be located between the silicon surface and the bulk silver of the contact after the Microstructure analysis of contacts 195 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ firing process. The probability for a direct or tunneling contact between silicon/crystallite and bulk silver is significantly reduced. On the other hand, if the amount of glass frit is insufficient, the etched area and crystallite density will be strongly reduced, again reducing the probability for a direct contact formation between semiconductor and metal layer. The amount of glass frit in the paste used for the presented analyzes is optimized for screen-printed contacts. This contact is significantly higher and broader compared to the width and height of pad-printed contacts. For fine-line printed and flat contacts an increased amount of glass frit in the paste probably will reduce contact resistance. 196 Summary _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Summary This work focuses on the optimization of the front side metallization process of industrial silicon solar cells. Since the state-of-the-art screen-printed contact has some well-known limitations, as a high contact resistance, a low lateral conductance and a low aspect ratio (height : width), alternative metallization technologies suitable for a one-layer grid structure of solar cells were reviewed. These include stencil-printing, pad-printing, ink-jet printing and dispensing. However, none of these technologies fulfills all the demands of a high-efficiency contact. Thus in this work the development of new concepts for the front side metallization was based on the two-layer contact structure. This structure has the advantage that every layer can be optimized individually. The requirement for the first layer is a small contact width, a low contact resistance, a sufficient adhesion to the underlying silicon surface as well as a good adhesion to the metal of the second layer. The task of the second layer is to collect the current from the first one, so its critical parameter is high line conductivity. The increased efficiency potential of the two-layer contact structure compared to a screen-printed contact was demonstrated by loss calculations. With respect to the analysis of new contact schemes, especially the total series resistance of a solar cell is an important parameter for quantifying electrical losses. The challenge consists in finding a minimum between electrical (resistance) and optical (shading) losses due to the front pattern. Six different determination methods for measuring the series resistance were reviewed and compared experimentally. Methods based on the comparison of two IV-curves measured under different illumination conditions resulted in a reliable determination of the series resistance under operating conditions. Due to its major importance in the state-of-the-art industrial production, screenprinting was investigated as a reference process. The front side screen-printing process could be optimized by using hotmelt silver paste. Despite the different composition of the hotmelt compared to conventional paste, the rheology and hence the printability of the paste behaves similarly at elevated temperatures. The advantage of screen-printing hotmelt paste is the achievement of relatively high finger aspect ratios (1 : 3 - 1 : 4). Efficiencies of 18.0%21 on 12.5 x 12.5 cm² Czsilicon with an Al-BSF were obtained, an efficiency increase of 0.3% absolute compared to conventionally printed cells of the same batch. Summary 197 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Evaporation of metals onto the photolithographically structured front surface can be regarded as the laboratory reference process for a multi layer stack system used to process high-efficiency solar cells. Within this work, different metals (Ni, Cr, Al, Pd, Ti and Ag) were evaluated with respect to their properties as contact layer. Efficiencies of 21.5% for solar cells with Ni-Ag and 21.3% for solar cells with Cr-Ag contacts and a high-efficiency rear structure were achieved, which is comparable to the reference stack system Ti-Pd-Ag. Light-induced plating is used for the high-efficiency process in order to form the second layer of the proposed two-layer contact structure. Theoretical as well as experimental investigations allowed a profound understanding of the chemical reaction in the bath and optimum plating conditions were found. The light-induced plating process was transferred successfully from small-area high-efficiency with evaporated contacts to large-area processed solar cells with fine-line screen-printed contacts. An efficiency increase of 0.3% to 0.4% absolute was demonstrated. This promising result led to the development of a light-induced plating machine with horizontal inline transport. Efficiencies up to 18.7% were achieved on 2 x 2 cm² FZ silicon solar cells with fine-line screen-printed and plated contacts on the front and an aluminum back surface field (Al-BSF) on the rear. In combination with a laser-fired rear-contact structure, efficiencies up to 19.3%21 (independently confirmed) were achieved, representing the highest efficiency for a screen-printed front metallization scheme up to date. Apart from fine-line screen-printing, other technologies are also suitable to deposit the contact layer. A promising alternative for achieving contacts of small width is pad-printing. In this work hotmelt paste was pad-printed resulting in a complete transfer of the print image from the pad to the wafer. In combination with the light-induced plating process, efficiencies up to 17.9% were achieved on 10 x 10 cm² Cz-silicon solar cells with an Al-BSF. A metallization technology not applying any load to the wafer is based on aerosolization an ink. The patented metal aerosol jet technology was further developed in this work. Line widths down to 14 µm were achieved printing metal organic inks. 12.5 x 12.5 cm² Cz-silicon solar cells with aerosol jet printed and plated front contacts were processed using modified screen-printing paste achieving efficiencies up to 18.3% and a fill factor of 81.0%. 21 Independently confirmed by Fraunhofer ISE CalLab. 198 Summary _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ In order to reach a better understanding of the contact formation process of plated and fine-line printed fingers, a micro structure analysis was performed. An increase of the contact resistivity with decreasing finger width was traced back to a reduced etched fraction of the SiNx layer and a lower silver crystallite density under the contacts. Furthermore, an improvement of the contact resistance by the plating step was revealed, which is probably caused by new current paths between silicon and plated silver in the edge zone of the contact. In summary, it could be demonstrated that the favored two-layer contact structure shows not just theoretically, but also experimentally a significant higher efficiency potential. Especially, when using the metal aerosol jet technology for creating the seed layer in combination with adapted metal inks, a significant efficiency increase in industrial production can be expected. Deutsche Zusammenfassung 199 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Deutsche Zusammenfassung Schwerpunkt der vorliegenden Arbeit ist die Optimierung der Vorderseitenstruktur industriell gefertigter Siliziumsolarzellen. Da der standardmäßig mit dem Siebdruckverfahren hergestellte Kontakt stark limitiert ist, das heißt einen hohen Kontaktwiderstand zum Silizium, eine geringe Fingerleitfähigkeit und ein geringeres Aspektverhältnis (Höhe : Breite) aufweist, wurden alternative Metallisierungstechnologien für die Solarzellenvorderseite evaluiert. Hierzu zählen Schablonen-, Tampon-, Tintenstrahldruck und Dispensen. Allerdings erfüllt keine dieser Technologien alle Anforderungen eines hocheffizienten Vorderseitenkontaktes. Aus diesem Grund basieren die in dieser Arbeit entwickelten Konzepte überwiegend auf einem Zweischichtenkontaktsystem. Die erste Schicht soll dabei eine schmale Kontaktbreite, einen niedrigen Kontaktwiderstand, eine ausreichende Haftfestigkeit sowie eine gute Anbindung zu dem Metall der darüber liegenden zweiten Schicht aufweisen. Die Aufgabe der zweiten Schicht ist es, den Strom der ersten Schicht einzusammeln und abzuführen. Daher spielt hier eine hohe spezifische Leitfähigkeit die zentrale Rolle. Das gesteigerte Wirkungsgradpotential eines solchen Zweischichtensystems im Vergleich zu einem siebgedruckten Standardkontakt wurde anhand von Verlustrechnungen demonstriert. Für die Beurteilung der neuen Kontaktierungsprozesse ist der Gesamtserienwiderstand der Solarzelle ein zentraler Charakterisierungsparameter, um die elektrischen Verluste zu quantifizieren. Die Herausforderung besteht darin, ein Optimum zwischen elektrischen (Widerstands-) und optischen (Abschattungs-) Verlusten des Vorderseitengrids zu ermitteln. Sechs unterschiedliche Bestimmungsmethoden wurden untersucht und experimentell verglichen. Methoden basierend auf dem Vergleich zweier Kennlinien bei verschiedenen Bestrahlungsbedingungen ergaben zuverlässige Serienwiderstandswerte für Solarzellen unter Arbeitsbedingungen. Aufgrund der großen Bedeutung der Siebdrucktechnologie in der industriellen Solarzellenherstellung wurde dieser Prozess ebenfalls untersucht. Der Vorderseitensiebdruckprozess konnte unter Einsatz von hochschmelzenden Pasten optimiert werden. Trotz der unterschiedlichen Zusammensetzung der hochschmelzenden Paste im Vergleich zu herkömmlichen Pasten ist das 200 Deutsche Zusammenfassung _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Fließverhalten und somit die Druckfähigkeit ähnlich, wenn eine erhöhte Temperatur gewählt wird. Der Vorteil der siebgedruckten hochschmelzenden Paste ist das Erreichen von relativ hohen Aspektverhältnissen (1 : 3 – 1 : 4). Wirkungsgrade von bis zu 18.0%22 konnten auf 12.5 x 12.5 cm² großen Cz-SiliziumSolarzellen erreicht werden, eine Steigerung um ∆η = 0.3% absolut verglichen mit konventionell gedruckten Solarzellen aus der gleichen Charge. Hochvakuumaufdampfen von Metallen auf photolithographisch strukturierten Vorderseiten ist der Laborreferenzprozess für ein Mehrschichtsystem. Dieser wird zur Herstellung hocheffizienter Solarzellen benutzt. Im Rahmen dieser Arbeit wurden die Kontakteigenschaften mehrere Metalle (Ni, Cr, Al, Pd, Ti und Ag) untersucht. Solarzellenwirkungsgrade von 21.5% wurden für Ni-Ag und 21.3% für Cr-Ag Kontakte erreicht. Diese sind vergleichbar zu dem Referenzschichtsystem Ti-Pd-Ag. Für den Hocheffizienzprozess wird die so genannte Licht-induzierte Galvanik benutzt, um die zweite Schicht des vorgeschlagenen Zweischichtenkontaktsystems herzustellen. Durch theoretische sowie experimentelle Untersuchungen wurde ein tiefer gehendes Verständnis der chemischen Abläufe erreicht. So konnten optimale Betriebsbedingungen für den Galvanikprozess ermittelt werden. Dies ermöglichte den erfolgreichen Übertrag von kleinflächigen, hocheffizienten Solarzellen mit aufgedampften Kontakten auf großflächig hergestellten Solarzellen mit feinen Siebdruckkontakten und einem ganzflächigen Aluminium Rückseitenkontaktfeld (Al-BSF) als Rückseitenstruktur. Ein Wirkungsgradgewinn von 0.3% bis 0.4% absolut wurde realisiert. Dieses Ergebnis führte zu der Entwicklung einer industriellen Licht-induzierten Galvanikanlage im Durchlaufverfahren. Auf 2 x 2 cm² großen FZ-Silizium-Solarzellen mit feinen siebgedruckten und galvanisierten Kontakten sowie einem Al-BSF wurden Wirkungsgrade bis zu 18.7% erreicht. Unter Verwendung des LFC Rückseitenprozesses wurden Wirkungsgrade bis zu 19.3%22 erreicht, der nach Wissen des Autors höchste berichtete Wirkungsgrad für Solarzellen mit einer Siebdruckvorderseite. Anstelle des Feinliniensiebdrucks gibt es eine Vielzahl von Technologien, die sich zur Herstellung der Kontaktschicht eignen. Eine vielversprechende Alternative bietet der Tampondruck. In dieser Arbeit wurde der Tampondruckprozess mit hochschmelzenden Pasten durchgeführt, bei dem das komplette Druckbild vom Tampon zum Wafer übertragen werden konnte. In Verbindung mit der Licht22 Unabhängig bestätigt vom Fraunhofer ISE CalLab. Deutsche Zusammenfassung 201 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ induzierten Galvanik wurden Wirkungsgrade bis zu 17.9% auf 10 x 10 cm² CzSilizium-Solarzellen mit einem Al-BSF erreicht. Eine Metallisierungstechnologie, die den Wafer mechanisch nicht belastet, basiert auf der Vernebelung von Metall-Tinten. Mit dem patentierten und in dieser Arbeit weiterentwickelten Aerosolstrahldruckverfahren konnten mit metallorganischen Tinten unterbrechungsfreie Linien von nur 14 µm Breite erreicht werden. Unter Verwendung modifizierter Siebdruckpasten wurden auf CzSilizium-Solarzellen mit einer Größe von 12.5 x 12.5 cm² Wirkungsgrade bis zu 18.3% und einem Füllfaktor von 81.0% mit Aerosolstrahldruck gedruckten und galvanisierten Frontkontakten erzielt. Um ein besseres Verständnis des Kontaktbildungsprozesses der galvanisierten und dünn gedruckten Finger zu erreichen, wurden Mikrostrukturuntersuchungen durchgeführt. Die Zunahme des Kontaktwiderstandes mit abnehmender Kontaktbreite konnte auf eine geringere Ätzung der Siliziumnitridfläche und einer geringeren Silber-Kristallitdichte unter dem Kontakt zurückgeführt werden. Zur Erklärung der beobachteten Verbesserung des Kontaktwiderstandes durch den Galvanikprozess, werden neue Strompfade zwischen Galvaniksilber und Silizium am Randbereich des Dickfilmkontaktes vorgeschlagen. Zusammenfassend konnte also gezeigt werden, dass das in dieser Arbeit favorisierte Zweischichtenkontaktsystem nicht nur theoretisch sondern auch experimentell ein signifikant erhöhtes Wirkunsgradpotential aufweist. Insbesondere bei der Verwendung der Aerosolstrahltechnik zur Erzeugung der Saatschicht und angepasster Metallisierungpaste können beachtliche Wirkungsgraderhöhungen in der industriellen Fertigung erwartet werden. 202 Appendix _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Appendix Appendix A – Analytical description of electrical losses An analytical description of the individual loss contribution due to the front side metallization grid is presented. This includes losses caused by the emitter, finger, busbar, tab, tab extension and soldering joint. Compare also Chapter 1.2.2. A.1 Loss contribution of the emitter The ohmic loss contribution of the emitter depends mainly on the emitter sheet resistance and the finger separation distance. The current in the emitter layer in the middle between two fingers is zero (I(x=0) = 0) and increases linear towards the finger edge, defined in the following as x-direction (see Fig. A.1) The calculation for the loss contribution of finger and busbar is analogue. a lf unit cell: I wb unit cell: III s wf unit cell: II 2a b X=0 wb dX lte Fig. A.1: Η-grid metallization pattern illustrating unit cell I, II and III used for resistance calculations. Geometrical parameters are described in Table 1.1 of Chapter 1.2.2. The bottom righthand schematic illustrates the current flow in the emitter and finger. The effective (or lumped) series resistance Reff is defined by: Reff ,i = Pe _ loss 2 I uc (A.1) Appendix 203 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ with Pe_loss being the electrical power loss due to the correspondent resistance and Iuc the total current generated by the correspondent unit cell. The current at position I(x) and the total current IucI of unit cell I entering the finger from one side can be described by: I (x ) = j a ⋅ l f ⋅ x I ucI = j a ⋅ l f s − wf (A.2) 2 The resistance of the emitter layer in x-direction to the finger by: dRe = Rsh dx lf (A.3) This leads to the determination of the power loss in the emitter Pe,e_loss. dPe,e _ loss = I ( x )2 dR s−wf Pe,e _ loss = j a2 ⋅l f 2 2 R ⋅ sh lf ∫ 0 s − wf 1 x dx = j a2 ⋅ l f ⋅ Rsh ⋅ 3 2 2 3 (A.4) Applying equation (A.1) the effective series resistance of the emitter Reff,e is achieved: Reff ,e = Pe,e _ loss 2 I ucI = ( ) 1 s − wf Rsh 6 lf (A.5) This is equal to an emitter resistance weighted to the cell area re of: re = Reff ,e AucI = Reff ,e a ⋅ ( ) s − wf s 1 = Rsh a⋅s 2 12 lf (A.6) The maximum theoretical power output of unit cell I Pmax,ucI is equal to: Pmax,ucI = j max ⋅ Va ⋅ a ⋅ s 2 (A.7) The power loss fraction pe,e of the emitter is finally determined by dividing the power loss of the emitter by the theoretical maximum power output of the same area. p e ,e = Pe,e Pmax,ucI ( s − wf 1 j (1 − p s ) = l R f sh 12 Va (1 − p c )2 s⋅a )3 = j (1 − p s )re,e Va (A.8) A.2 Loss contribution of the contact finger The loss contribution of the contact finger to the total loss contribution depends mainly on its length, its conductivity and on the finger separation distance (current 204 Appendix _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ collecting area). The loss calculation is similar as for the emitter. The current at the end of the finger (or just in-between two busbars) is zero and increases linear with its length, reaching its maximum value at the busbar. Thus, the effective series resistance of the finger Reff,f and the area weighted resistance rf are calculated by: Reff , f = lf 1 ρf hf wf 3 rf = lf 1 ρf ⋅a⋅s hf wf 3 (A.9) The power loss fraction of the contact finger by: p e, f = J (1 − p s )r f Va (A.10) A.3 Loss contribution of the busbar To keep the power loss due to the busbar resistance small, a sufficient amount of soldering joints to the overlying tab is required. The effective series resistance and loss fraction determination of the busbar is analogue to the one for the emitter. The assumption is that the current in the busbar at the soldering joint is maximal and decreases linear to zero at the point in-between two joints. The effective series resistance Reff,bus for unit cell II and the normalized resistance rbus are determined by, Reff ,bus = b 1 ρ bus 3 hbus wbuc rbus = b2 1 ρ bus a 3 hbus wbuc (A.11) the power loss fraction of the busbar by: p e,bus = j (1 − p s )rbus Va (A.12) A.4 Loss contribution of the tab Power losses due to the tab and tab extension can be high for large area silicon solar cells, if the width and the height of the tab itself are not sufficient. An analytical description of the tab and tab extension loss contribution is presented. The current collected by each soldering joint Ijs is equal to: 1− w f I js = 4 ⋅ j a ⋅ l f ⋅ b ⋅ s (A.13) Thus, the maximum current IT,max at the end of the tab with a length of lbus = 2 Nsj b (Nsj equals total amount of soldering joints) is determined by: Appendix 205 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 1− w f I T ,max = N sj ⋅ I f = N sj ⋅ 4 ⋅ j a ⋅ l f ⋅ b ⋅ s = 2 ⋅ lbus ⋅ j a ⋅ l f 1− w f ⋅ s (A.14) The current increases step wise in the tab. At each soldering joint the current Isj is collected by the tab; the current at the first soldering joint is I1, at the 2nd 2·I1 at the third 3·I1 and so on. The power loss for each tab separation distance can be calculated by: with Ptab (i ) = N sj (i ) ⋅ I sj2 ⋅ Rtab Rt = 2 ⋅ b ⋅ ρt ht ⋅ wt (A.15) The total power loss Pt of a tab can then be calculated by adding all power losses together: ) )2 ⋅ Rt + (N sj ⋅ I1 )2 ⋅ R2t (( Pt = I12 ⋅ Rt + (2 I1 )2 ⋅ Rt + (3I1 )2 ⋅ Rt + ... + N sj − 1 I1 N Sj −1 N Sj N sj 2 N sj 2 2 2 2 = I1 ⋅ Rt ⋅ ∑ n + = I1 ⋅ Rt ⋅ ∑ n − 2 2 n = 1 n=1 N sj ⋅ N sj + 1 2 N sj + 1 N sj 2 N sj ⋅ 2 N sj 2 + 1 2 2 = I1 ⋅ Rt ⋅ = I1 ⋅ Rt ⋅ − 6 2 6 2 ( )( ( ) (A.16) ) Using equation (A.1) the effective series resistance of the tab Reff,t can be calculated by: Reff ,t = Pt I t ,max 2 (2 N =R t 2 sj )= 1ρ ( +1 6 N sj 3 t ) 2 b 2 N sj + 1 1 b ⋅ l B = ρt ht wt N sj 3 ht wt 1 + 1 2N 2 sj (A.17) Multiplication of the area of unit cell III leads to the area weighted resistance rt: rt = AucIII ⋅ Reff ,t = 4 ⋅ N sj ⋅ b ⋅ Reff ,t a ⋅ lb 2 2 = ρt 3 ht ⋅ wt 1 + 1 2N 2 sj (A.18) A.5 Loss contribution of the tab extension The series resistance Reff,te and the area weighted resistance of the tab extension rte can be determined by: Reff ,te = ρ t lte hte ⋅ wte rte = 2 ⋅ ρ t lte ⋅ lb Ns ⋅ a hte ⋅ wte (A.19) 206 Appendix _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ A.6 Loss contribution of the soldering joint The area weighted resistance of the soldering joint rs depends on the size of the current collecting area and on the quality of the contact resistance between busbar and tab: rs = Rs ⋅ (2 ⋅ AucII ) = 4 ⋅ b ⋅ a ⋅ Rs (A.20) The current entering each soldering joint Is is equal to: wf I s = (2b ) ⋅ 2l f j a 1 − s ( ) (A.21) The power loss due the soldering joint: Ps = I s2 ⋅ R s = 16 ⋅ l f 2 b 2 j a2 1 − 2 wf Rs s (A.22) Appendix B – Contact resistance and resistivity The contact resistance and resistivity in this work was determined using the test structure proposed by Meier and Schroder [101]. More information also for other determination methods can be found in the following literature [161,162,165-168]. The contact resistance and contact resistivity are the most suitable parameters for describing the electrical quality of the metal-semiconductor interface. While the contact resistance is a directly measurable value the contact resistivity (also called specific contact resistance) is weighted to the contact area. As illustrated in Fig. B.1 the current flow from the emitter layer into the contact can be regarded as a lateral flow. The current density entering the contact is not homogeneous distributed over the contact width. It is highest at the edges, and drops towards the middle of the contact. The distance at which the current density has dropped to 1/e from the initial value is called transfer length (LT). The emitter and contact potential is illustrated in the schematic drawing of Fig. B.1-b, assuming the metal contact is everywhere on the same potential. For a contact with the same contact quality over its width, the contact resistivity decreases continuously with shrinking contact width, whereas the contact Appendix 207 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ contact Interface semiconductor lc a) b) contact resistance Rc Rc, ρc l(w) w contact resistivity ρc V(w) emitter potential 0 metal potential c) w=0 width w w = wc contact width wc d) Fig. B.1: a) Schematic drawing of the metal semiconductor interface with lc the contact length. [28] b) Equivalent circuit of a metal semiconductor contact with Rsk the sheet resistance under the contact. [28] c) Potential in the finger and in the emitter under the finger [23]. d) Schematic drawing of the contact resistance Rc and contact resistivity ρc versus the contact width wc. resistance for broad contacts first stays constant (wc >> LT) and then rises significantly with declining contact width (see Fig. B.1-d). Each contact (see Fig. B.1-a) can be approximated by a resistance network as illustrated in Fig. B.1-b. A differential equation for I(z) and V(z) can be set up. l − dI (w) = V (w ) c dw ρc and R dV (w) = I (w) sk dw lc (B.1) By combining the latter two equations, the following differential equation of second order is achieved: d 2V (w) 2 − d w Rsk ρc V (w) = 0 (B.2) This homogeneous differential equation can be solved by the following general approach (e.g. Bartsch 1998 p. 497): 1 1 V (w) = A ⋅ exp ⋅ w + B ⋅ exp − ⋅ w LT LT 1 1 wc I (w) = − A exp ⋅ w + B ⋅ exp − ⋅ w LT Rsh LT LT (B.3) with LT equals the transfer length: LT = ρc Rsk (B.4) 208 Appendix _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Two boundary conditions are introduced: I(w=0) = -I(w=wc) = I0 and V(w=0) = V0. Current I(w) and voltage V(w) can then be expressed by: I (w) = − w lc ⋅ 2 ⋅ A exp c LT Rsk 2 ⋅ LT 1 w ⋅ sinh c − w LT 2 (B.5) and w V (w) = 2 ⋅ A ⋅ exp c 2 ⋅ LT 1 w ⋅ cosh c − w LT 2 (B.6) Finally the contact resistance Rc,sz of a solar cell in which the current enters the contact from both sides can be calculated: Rc,sz = w V (0 ) LT Rsk = coth c I (0 ) lc 2 ⋅ LT = ρ c Rsk Rsk l c Rsk wc coth ρ 2⋅ L c T (B.7) This equation can be used to determine the contact resistivity ρc. The area weighted resistance rc as presented in Table 1.2 is equal to: rc = w LT Rsk coth c lc 2 ⋅ LT s ⋅ ⋅ a 2 (B.8) However, typically the contact resistivity ρc, the sheet resistance under the contact Rsk and the transfer length LT are not known. In this thesis the contact resistance was determined using test structures based on the transfer length model. A schematic drawing of the test structure is presented in Fig. B.2-a. RE measurement Rc measurement 50 measurment data of a TLM structure RTLM [Ω] 40 lc s contact stripe Rc Rc Rem a) Rc Rem 20 2 x RC 10 semiconductor Rc 30 Rc 0 0.0 0.5 slope = Rem 1.0 1.5 2.0 2.5 Finger separation distance s [mm] Rem b) Fig. B.2: a) Schematic drawing of the test structure proposed by Meier and Schroder [101] for contact resistance Rc and end resistance RE measurement (drawing taken from [169]). In the bottom figure the resistance equivalent circuit is presented. b) The measured resistance RTLM is plotted versus the finger separation distance s. By liner fitting the data points the sheet resistance Rsh and contact resistance Rc can be extracted. Appendix 209 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Using a four point probe setup the contact resistance Rc can be extracted by applying a constant dc current I between two adjacent contacts and by measuring the voltage drop V (Fig. B.2-a, Rc measurement). The measured resistance RTLM is equal to: RTLM = Vi ,i +1 I i ,i +1 = 2 ⋅ Rc + Rem = 2 ⋅ Rc + Rsh ⋅ s lc (B.9) with Rem the resistance of the emitter between the contacts, Rsh the emitter sheet resistance, lc the length of the contact and s the contact separation distance. If the sheet resistance Rsh is exactly known, the contact resistance can be directly calculated. Otherwise the resistance RTLM is extracted for different contact separation distances s. These RTLM measurements are plotted versus the finger separation distance s and the data points are linearly fitted. The y-axis intersection point is equal to two times Rc (RTLM(s=0) = 2 Rc) and the slope of the fit function is equal to Rem. The contact resistance Rc and the emitter sheet resistance Rsh can be calculated. Applying this measurement the contact resistance Rc is different from the one of a solar cell, because the current enters the finger just from one side. However, for qualifying the electrical performance of a contact, this simplified measuring setup was used. But the boundary conditions from equation (B.5) have changed to I(w=0) = I0, I(wc) = 0 and V(w = 0)=V0. V(w) emitter potential metal potential w=0 Length w w = wc Fig. B.3: Schematic drawing of the potential under the finger for current entering the finger from both sides (solid line) or for current entering the finger from one side (dashed line). This leads to a contact resistance Rc of: Rc = w LT Rsk coth c = lc LT ρ c Rsk lc Rsk coth w ρ c c (B.10) 210 Appendix _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Assuming the same sheet resistance under the contact as between the contacts (Rsh = Rsk), the transfer length LT and the contact resistivity ρc can be determined. However, it is expected that the doping concentration under contacts is changed, especially when using pastes that have an etching behavior. Applying an additional measurement the sheet resistance under the contact Rsk and furthermore the contact resistivity ρc can be calculated. As illustrated in Fig. B.2-a (RE measurement) the voltage drop is measured between a finger of the current-circuit and an adjacent finger as illustrated in Fig. B.2-a (RE measurement). The end resistance is then defined as: RE = V (wc ) LT ⋅ Rsk = I (0 ) lc 1 L sinh wc (B.11) If Rc and RE have been measured an experimental value for the transfer length LT can be achieved: w Rc = cosh c RE LT (B.12) Applying equation (B.11) a value for the sheet resistance under the contact Rsk can be determined and applying equation (B.4) the contact resistivity ρc. In this work mainly the contact resistance Rc (equation (B.10)) weighted to a contact of length lc = 1 cm was used to compare the electrical quality of different contacts. The contact resistivity ρc assumes the same contact quality over the whole contact width, which is not the case for a contact formed by etching pastes as crystallite density, etched SiN area and the thickness of the glass layer varies over the contact width. Furthermore for contacts of good quality the end resistance is relatively low, which increases the measuring uncertainty strongly. In contrast the contact resistance Rc is a relatively easy to measure value and a precise parameter to compare the electrical performance of different contacts. List of symbols, acronyms, indices and constants 211 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ List of symbols, acronyms, indices and constants symbol description µ a a0 Af Auc b c d E00 dynamic viscosity length of unit cell I screen opening fraction cross section area of finger area of unit cell width of unit cell II cross section diameter of screen wire wire separation distance characteristic energy for metal semiconductor interface energy level of conduction band edge Fermi- energy level energy level of trap energy level of valence band edge height electric current current density recombination current density in emitter and base recombination current density in space charge region current density of the active cell area current density of bath current density at maximum power point current density at mpp of the active cell area photo-generated current density short-circuit current density EC EF ET EV h I j j01, j02 ja jLip jmpp jmpp,a jph jsc lc lf lte, m n n1,n2 ncr unit Pa s m % m² m² m m m eV eV eV eV eV m A A m-2 A m-2 A m-2 A m-2 A m-² A m-2 A m-2 A m-2 A m-2 contact length for ρc test structure (in this work lc = 1 cm) m finger length m length tab extension m mass kg density of free electrons m-3 diode ideality factors crystallite density m-2 212 List of symbols, acronyms, indices and constants _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ symbol description unit ND NDS Ns NT p pc pe Ploss Pmax Pmpp ps q r concentration of donor atoms concentration of donor atoms at surface number of soldering joints between busbar and tab density of recombination defects/traps power loss fraction fraction of front side covered by metallization pattern electrical loss fraction electric power loss power gain of corresponding unit cell electric power density at maximum power point shading fraction caused by front metallization pattern elementary charge area weighted resistance m-3 m-3 R resistance Ω Rc Ω Reff contact resistance (in this work Rc always corresponds to the resistance for a contact of side length lc = 1 cm) effective series resistance Rline resistance of the contact finger per unit length Ω m-1 rp shunt (or parallel) resistance Ω m² rs series resistance Ω m² Rsh emitter sheet resistance Ω/sq rSi/SiN rSi/SiNx coverage fraction of silicon to SiN ratio of etched SiN area Rsk sheet resistance under contact s S Sback Seff Sfront t T tb, tf Tboil finger separation distance surface recombination velocity surface recombination velocity on the rear effective surface recombination velocity surface recombination velocity on the front time temperature transparency factor of busbar; finger boiling temperature of hotmelt paste m-3 % % % W m-2 W m-2 W m-2 % C Ω m² Ω Ω/sq m m s-1 m s-1 m s-1 m s-1 s °C or K °C List of symbols, acronyms, indices and constants 213 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ symbol description tf,ez, tf,mz Tpaste Tplate; Tnest Tscreen V Va Vbi Vmpp,a Vn Voc Vpaste Vrs vprint vth w x Xsuns transparency factor finger edge zone and middle zone temperature of hotmelt paste temperature of printing plate; nest screen temperature voltage voltage active cell area built-in voltage voltage at maximum power point of active cell area voltage between conduction band and Fermi level open-circuit voltage theoretical paste volume voltage drop at the series resistance velocity of print through process thermal velocity width distance, penetration depth number representing the irradiance in suns incident power density of photons Φ χS εs φBn φΜ η ηbath ϕ λ ρbus, ρb, ρf, ρm_rs , ρtab ρc ρrc ρ° τ τshear unit °C °C °C V V V V V V m-3 V m/s C·m s-1 m m W m-2 electron affinity of the semiconductor eV dielectric constant of the semiconductor C² J-1·m-1 potential barrier to n-silicon potential barrier of metal efficiency of solar cell efficiency of plating bath potential wavelength m line resistivity busbar, base, finger, metal of rear side, tab Ω m contact resistivity (emitter – front contact) Ω m² contact resistivity (base – rear contact) density of material Ω m² kg cm-3 minority carrier lifetime s shear stress s-1 214 List of symbols, acronyms, indices and constants _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ indices description A b bus c e ext f F m_rs metal_surface rc sj st tab uc(i) edge anode of plating bath base busbar front side contact emitter tab extension finger front side of solar cell metal rear side surface area of the contact finger rear side contact soldering joint emulsion layer tab unit cell i with i the number of the correspondent cell edge layer of the contact constant description value A* c F (m*/m0)·120 A m-2 K-2 299792458 m s-1 96485 C mol-1 η effective Richardson constant velocity of light Faraday constant Planck’s constant k Boltzmann’s constant m* MAg NC 1.3806⋅10-23 J K-1 1.08·m 107.87 g mol-1 ni effective electron mass molar mass of silver effective density of states of conduction band 2.84⋅1019 cm-3 intrinsic carrier density 1.00⋅1010 cm-3 NV effective density of states of valence band 2.68⋅1019 cm-3 q, m elementary charge of electrons zAg oxidation number of silver 1.602⋅10-19 C 1 6.62608⋅10-34 J s List of symbols, acronyms, indices and constants 215 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ acronym description AFM atomic force microscopy Al-BSF AM1.5g AR ARC aluminium back surface field air mass 1.5 global spectrum aspect ratio (height to width) antireflection coating Cz D DTGA monocrystalline silicon produced with the Czochralsky method diode differential thermogravimetric analysis EDX energy dispersive X-ray FE FIB field emission focused ion beam FZ i IV LFC PECVD PERL PL PSG SEM SIMS SP TE TFE monocrystalline silicon produced with the floating zone method recombination process current-voltage laser-fired contacts plasma enhanced chemical vapour deposition passivated emitter rear locally-diffused (solar cell structure) photolithography phosphorus silicate glass scanning electron microscope secondary ion mass spectroscopy screen printed thermionic emission thermionic field emission TGA Thermogravimetric analysis 216 List of Publications _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ List of Publications 1. A. Mette, D. Erath, R. Ruiz, G. Emanuel, E. Kasper, and R. Preu, "Hot melt ink for the front side metallisation of silicon solar cells", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 873876 (2005). 2. A. Mette, C. Schetter, D. Wissen, S. Lust, S.W. Glunz, and G. Willeke, "Increasing the efficiency of screen-printed silicon solar cells by light-induced silver plating", in Proceedings of the 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, USA, pp. 1056-1059 (2006) 3. R. Preu, D. Biro, F. Clement, G. Emanuel, A. Grohe, M. Hofmann, A. Mette, C. Voyer, W. Wolke, S. W. Glunz, and G. Willeke, "The status of silicon solar cell production technology development at Fraunhofer ISE," in Proceedings of the 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, USA, pp. 1040-1043 (2006). 4. A. Mette, P.L. Richter, S.W. Glunz "Novel metal jet printing technique for the front side metallization of highly efficient industrial silicon solar cells", in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 1174-1177 (2006) 5. A Mette, G. Emanuel, D. Erath, R. Preu, G. Willeke," High efficiencies on large area screen printed silicon solar cells and contacting high sheet resistance emitters using hotmelt ink," Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 709-712 (2006). 6. S.W. Glunz, A. Mette, M. Alemán, P.L. Richter, A. Filipovic, and G. Willeke, "New concepts for the front side metallization of silicon solar cells," in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 746-749 (2006). 7. M. Alemán, A. Streek, P. Regenfuß, A. Mette, R. Ebert, H. Exner, S.W. Glunz, and G. Willeke, "Laser micro-sintering as a new metallization technique for silicon solar cells," in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany (2006) pp. 705-708. 8. A. Mette, D. Pysch, G. Emanuel, D. Erath, R. Preu, and S. Glunz, "Series resistance characterization of industrial silicon solar cells with screen-printed contacts using hotmelt paste", Progress in Photovoltaics: Research and Applications, vol. 15, pp. 493-505 (2007). 9. D. Erath, A. Mette, and G. Hübner, "Neue Siebdrucktechnologie erhöht den Wirkungsgrad von Solarzellen," Horizonte, vol. 29 (2006) pp. 24-26. List of Publications 217 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ 10. A. Mette, P.L. Richter, M. Hörteis, and S.W. Glunz, "Metal aerosol jet printing for solar cell metallization", Progress in Photovoltaics: Research and Applications, vol. 15, pp. 621-627 (2007). 11. G. Allardyce, J. Cahalen, T. Ridler, J. Rasch, O. Weigel, H. Fröhlich, H. Kappler, C. Rattey, A. Mette, C. Schetter, and S. W. Glunz, "The Commercial Application of Light Induced Electroplating for Improving the Efficiency of Crystalline Silicon Solar Cells," in Proceedings of the 22nd European Photovoltaic Solar Energy Conference, Milan, Italy, pp. 1578-1580 (2007) 12. M. Hörteis, A. Mette, P. L. Richter, and S. W. Glunz, "Further progress in metal aerosol jet printing as front side metallization technique", in Proceedings of the 22nd European Photovoltaic Solar Energy Conference, Milan, Italy, pp. 1039-1041 (2007) 13. D. Pysch, A. Mette, A. Filipovic, S.W. Glunz, "Detailed analysis for fine-line printed solar cell contacts", in Proceedings of the 22nd European Photovoltaic Solar Energy Conference, Milan, Italy, pp. 1238-1243 (2007) 14. O. Schultz, A. Mette, R. Preu, S.W. Glunz, "Silicon solar cells with screenprinted front side metallization exceeding 19% efficiency," in Proceedings of the 22nd European Photovoltaic Solar Energy Conference, Milan, Italy, pp. 980-983 (2007) 15. D. Pysch, A. Mette, S.W. Glunz, “A review and comparison of different methods to determine the series resistance of solar cells”, Solar Energy Materials and Solar Cells, vol. 91, pp. 1698-1706 (2007). Patents 1. A. Mette, S. Glunz, R. Preu, and C. Schetter, "Halbleiterelement mit einem auf mindestens einer Oberfläche angeordneten elektronischen Kontakt" Internationales Patent No.WO 2006/008080 (2006). 2. A. Mette, C. Schetter, S.W. Glunz, P.L. Richter, M. Hörteis, "Verfahren zur Herstellung einer metallischen Kontaktstruktur einer Solarzelle" patent pending: priority date June 30th 2006 3. K. Mayer, D. Kray, S. Baumann, D. Biro, A. Mette, "Verfahren zur Präzisonsbearbeitung von Festkörpern mittels flüssigkeitsunterstütztem Laser" patent pending: priority date January 22nd 2006 4. M. Aleman, A. Mette, S.W. Glunz, R.Preu, "Verfahren zum Aufbringen von elektrischen Kontakten auf halbleitende Substrate, halbleitendes Substrat und Verwendung des Verfahrens” German patent pending No. 10 2006 040 352.523, priority date: August 29th 2006 218 Bibliography _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Bibliography [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] Bundesministerium für Umwelt, Naturschautz und Reaktorsicherheit: Agenda 21-Konferenz der Vereinten Nationen für Umwelt und Entwicklung, Juli 1992 in Rio de Janeiro (2002). "Climate Change 2007: The Physical Science Basis - Summary for Policymakers", Working Group I of the Intergovernmental Panel on Climate Change, Intergovermental panel on climate change (2007). E. A. Alsema, M. J. de Wild-Scholten, and V. M. Fthenakis, "Enviromental impacts of PV electricity generation - A critical comparison of energy supply options", in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 3201-3207 (2006). S. Teske, A. Zervor, and O. Schäfer, "Global energy [r]evolution", Greenpeace International, European Renewable Energy Council (2007). L. Frantzis, E. Jones, C. Lee, M. Wood, and P. Wormser, "Opportunities for cost reductions in photovoltaic modules", in Proceedings of the 16th European Photovoltaic Solar Energy Conference, Glasgow, UK, pp. 21002103 (2000). E. Schneiderlöchner, Dissertation Thesis, Albert-Ludwigs-Universität, Freiburg im Breisgau, Germany (2004). F. Huster, "Aluminium-back surface field: bow investigation and elimination", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 635-638 (2005). F. Huster, "Investigation of the alloying process of screen printed aluminium pastes for the BSF formation of silicon solar cells", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 1466-1469 (2005). W. P. Mulligan and S. R.M., "High efficiency, one-sun solar cell processing", in 12th Workshop on Crystalline Silicon Solar Cell Materials and Processes, Golden, CO, USA, pp. 30-37 (2002). W. P. Hirshman and M. Schmela, "Silicon shortage - so what! Market survey on cell and module production 2005" in Photon International; Vol. 3 pp 100-125 (2006). M. A. Green, "Solar cells: operating principles, technology and system applications", UNSW, Kensington (1986). J. O. Schumacher and W. Wettling, "Device physics of silicon solar cells", Vol. 3, Imperial College Press, London, Great Britain (2000). R. M. Swanson, "Approaching the 29% limit efficiency of silicon solar cells", in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, USA (2005). P. Würfel, "Physik der Solarzellen", Spektrum Akademischer Verlag, Heidelberg, Berlin, Oxford (1995). Bibliography 219 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [15] J. Zhao, A. Wang, and M. A. Green, "24·5% Efficiency silicon PERT cells on MCZ substrates and 24·7% efficiency PERL cells on FZ substrates", Progress in Photovoltaics: Research and Applications, vol. 7, pp. 471-474 (1999). [16] R. Preu, Dissertation Thesis, Fern Universität Hagen, Germany (2000). [17] W. Shockley and W. T. J. Read, "Statistics of the recombinations of holes and electrons", Physical Review, vol. 87, pp. 835-842 (1952). [18] R. N. Hall, "Electron-hole recombination in germanium", Physical Review, vol. 87, pp. 387 (1952). [19] S. W. Glunz, Dissertation Thesis, University Freiburg (1995). [20] M. A. Green, "Solar cells: operating principles, technology and system applications", UNSW, Kensingotn (1998). [21] J. Horzel, K. de Clercq, O. Evrard, J. Szlufcik, L. Frisson, F. Duerinckx, J. Nijs, and R. Mertens, "Advantages of a new metallisation structure for the front side of solar cells", in Proceedings of the 13th European Photovoltaic Solar Energy Conference, Nice, France, pp. 1368-1373 (1995). [22] A. W. Blakers, "Shading losses of solar-cell metal grids", Journal of Applied Physics, vol. 71, pp. 5237-5241 (1992). [23] A. R. Burgers, Dissertation Thesis, University Utrecht, Netherlands (2005). [24] A. R. Burgers and J. A. Eikelboom, "Optimizing metalization patterns for yearly yield [solar cell fabrication]", in Proceedings of the 26th IEEE Photovoltaic Specialists Conference, Anaheim, California, USA, pp. 219222 (1997). [25] Schottky, "Halbleitertheorie der Sperrschicht", Naturwissenschaften, vol. 26 (1938). [26] W. Schottky, "Zur Halbleitertheorie der Sperrschichtund Spitzengleichrichter", Z. Physik, vol. 113, pp. 367-414 (1939). [27] J. Bardeen, "Surface states and rectification at a metal semi-conductor contact", Physical Review, vol. 71, pp. 717-727 (1947). [28] D. K. Schroder and D. L. Meier, "Solar cell contact resistance - a review", IEEE Transactions on Electron Devices, vol. ED-31, pp. 637-647 (1984). [29] S. M. Sze, "Physics of Semiconductor Devices", 2nd ed., John Wiley & Sons, New York (1981). [30] A. Y. C. Yu, "Electron tunneling and contact resistance of metal-silicon contact barriers", Solid-State Electronics, vol. 13, pp. 239-247 (1970). [31] F. A. Padovani and R. Stratton, "Field and thermionic-field emission in Schottky barriers", Solid-State Electronics, vol. 9, pp. 695-707 (1966). [32] W. Wolke, J. Catoir, G. Emanuel, J. Liu, M. Ruske, and R. Preu, "Surface passivation for solar cells by large scale inline sputtering of silicon nitride", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 733-736 (2005). 220 Bibliography _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [33] J. Zhao, A. Wang, P. P. Altermatt, S. R. Wenham, and M. A. Green, "24% efficient silicon solar cells", in Proceedings of the 1st World Conference on Photovoltaic Energy Conversion, Hawaii, USA, pp. 1477-1480 (1994). [34] M. Hermle, E. Schneiderlöchner, G. Grupp, and S. W. Glunz, "Comprehensive comparison of different rear side contacting methods for high-efficiency solar cells", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 810-813 (2005). [35] E. Schneiderlöchner, R. Preu, R. Lüdemann, and S. W. Glunz, "Laser-fired rear contacts for crystalline silicon solar cells", Progress in Photovoltaics: Research and Applications, vol. 10, pp. 29-34 (2002). [36] O. Schultz, S. W. Glunz, and G. Willeke, "Multicrystalline Silicon Solar Cells Exceeding 20 % Efficiency", Progress in Photovoltaics: Research and Applications, vol. 12, pp. 553-558 (2004). [37] M. M. Hilali, B. To and A. Rohatgi, " A review and understanding of screen-printed contacts and selective-emitter formation", in Proceedings of the 14th Workshop on Crystalline Silicon Solar Cells and Modules NREL, Winter Park, Colorado, USA, pp. 109-116 (2004). [38] P. Hahne, "Innovative Drucktechnologien. Siebdruck - Tampondruck", Der Siebdruck, Lübeck (2001). [39] H. H. C. de Moor, A. W. Weeber, J. Hoornstra, and W. C. Sinke, "Fine-line screen printing for silicon solar cells", in Snowmass Conference, Colorado, USA, pp. 154-170 (1996). [40] G. Schubert, Dissertation Thesis, University Konstanz, Germany (2006). [41] A. S. Scherr, "Siebdruck-Handbuch", Verlag Der Siebdruck (1999). [42] H.-J. Homann, "Lehrbuch Siebdruck. Druckformherstellung", Verlag Ursula Homann, Emmendingen (1995). [43] O. V. Mazurin, M. V. Streltsina, and T. P. Shvaiko-Shvaikovskaya, Amsterdamm: Elsevier Science Publisher B.V. (1987). [44] J. Hoornstra, H. de Moor, A. Weeber, and P. Wyers, "Improved front side metallization on silicon solar cells with stencil printing", in Proceedings of the 16th European Photovoltaic Solar Energy Conference, Glasgow, UK, pp. 1416-1419 (2000). [45] H. H. C. de Moor, J. Hoornstra, A. W. Weeber, A. R. Burgers, and W. C. Sinke, "Printing high and fine metal lines using stencils", in Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 404-407 (1997). [46] G. Yao, J. Zhao, and J. E. Cotter, "Metal stencil-printed selective emitter silicon solar cell fabrication on CZ wafer", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 749-752 (2005). [47] P. Hahne, I. E. Reis, E. Hirth, D. Huljic, and R. Preu, "Pad printing - a novel thick-film technique of fine-line printing for solar cells", in Proceedings of Bibliography 221 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] the 2nd World Conference on Photovoltaic Energy Conversion, Vienna, Austria, pp. 1646-1649 (1998). D. M. Huljic, E. Hirth, R. Lüdemann, and G. Willeke, "Large area crystalline silicon solar cells with pad printed front side metallisation", in Proceedings of the 17th European Photovoltaic Solar Energy Conference, Munich, Germany, pp. 1582-1585 (2001). D. M. Huljic, S. Thormann, R. Preu, R. Lüdemann, and G. Willeke, "Pad printed front contacts for c-Si solar cells - a technological and economical evaluation", in Proc. 29th IEEE PVSC, New Orleans, Louisiana, USA, pp. 126-129 (2002). P. Consulting, "Pad printing technical manual", Vol. 2, Barnsley (1993). F. Huster, P. Fath, and E. Bucher, "Roller printed mc-Si solar cells with optimized fill factors of 78%", in Proceedings of the 17th European Photovoltaic Solar Energy Conference, Munich, Germany, pp. 1743-1746 (2001). M. Y. Ghannam, J. Poortmans, and J. F. Nijs, "Quantitative comparison between the performance of buried contact and of screenprinted contact industrial silicon solar cells", Kuwait Journal of Science and Engineering, vol. 25, pp. 363-377 (1998). K. Murata, "Super-fine ink-jet printing for nanotechnology", in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, pp. 346-349 (2003). J. Heinzl and C. H. Hertz, "Ink-jet printing", Advances in Electronics and Electon Physics, vol. 65, pp. 91-171 (1985). H. P. Le, "Progress and trends in ink-jet printing technology - part 2", Journal of Imaging Science and Technology, vol. 42, pp. 1-6 (1998). K. F. Teng and R. W. Vest, "Application of ink jet technology on photovoltaic metallization", IEEE Electron Device Letters, vol. 9, pp. 591593 (1988). C. Curtis, T. Rivkin, A. Miedaner, J. Alleman, J. Perkins, L. Smith, and D. Ginley, "Metallizations by direct-write inkjet printing", in NCPV Program Review Meeting, Lakewood, Colorado, USA (2001). T. Rivkin, C. Curtis, A. Miedaner, J. Perkins, J. Alleman, and D. Ginley, "Direct write processing for photovoltaic cells", in Proceedings of the 29th IEEE Photovoltaics Specialists Conference, New Orleans, Louisiana, USA, pp. 1326-1329 (2002). T. Kaydanova, M. F. A. M. van Hest, A. Miedaner, C. J. Curtis, J. L. Alleman, M. S. Dabney, E. Garnett, S. Shaheen, L. Smith, R. Collins, J. I. Hanoka, A. M. Gabor, and D. S. Ginley, "Direct write contacts for solar cells", in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, USA, pp. 1305-1308 (2005). 222 Bibliography _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [60] M. Carter, J. Colvin, and J. Sears, "Characterization of conductive inks deposited with maskless mesoscale material deposition (M3D)", in Proceedings of the TMS Annual Meeting (2006). [61] B. King, M. Renn and M.Essien, “Maskless Mesoscale Material Deposition,” EP&P, Electronic Packaging &Production, 43:2, pp 18-20, (2003). [62] M. Alemán, A. Streek, P. Regenfuß, A. Mette, R. Ebert, H. Exner, S. W. Glunz, and G. Willeke, "Laser micro-sintering as a new metallization technique for silicon solar cells", in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 705-708 (2006). [63] P. Regenfuß, R. Ebert, S. Klötzer, L. Hartwig, H. Exner, T. Brabant, and T. Petsch, "Industrial laser micro sintering", in Proceedings of the 4. LANE 2004, Erlangen, pp. 413-424 (2004). [64] A. Brenner, "Electroless plating comes of age", Metal Finishing, pp. 61-68 (1954). [65] A. Brenner, "Electroless plating comes of age", Metal Finishing, pp. 68-76 (1954). [66] A. Brenner and E. Riddell, "Nickel plating on steal by chemical reduction", Journal of Research of the National Bureau of Standards, vol. 37, pp. 31-34 (1946). [67] M. V. Sullivan and J. H. Eigler, "Electroless nickel plating for making ohmic contacts", Journal of the Electrochemical Society, vol. 104, pp. 226230 (1957). [68] N. R. Saha, D. Biswas, and S. Dhar, "Room temperature electroless nickel plating of n-type silicon, its possible application in solar cells", Indian Journal of Physics, vol. 54A, pp. 158-163 (1980). [69] J. R. Anderson and R. C. Petersen, "Nickel contacts for low cost solar cells", in Fourteenth IEEE Photovoltaic Specialists Conference 1980, San Diego, CA, USA, pp. 948-951 (1980). [70] D. S. Kim, E. J. Lee, J. Kim, and S. H. Lee, "Low-cost contact formation of high-efficiency crystalline silicon solar cells by plating", Jorunal of the Korean Physical Society, vol. 46, pp. 1208-1212 (2005). [71] M. A. Green and S. R. Wenham, in Australian Patent No 5703309 Australia, (1984). [72] J. D. Jensen, P. Moller, T. Bruton, N. Mason, R. Russell, J. Hadley, P. Verhoeven, and A. Matthewson, "Electrochemical deposition of buried contacts in high-efficiency crystalline silicon photovoltaic cells", Journal of the Electrochemical Society, vol. 150, pp. 49-57 (2003). [73] A. M. T. van der Putten and J.-W. G. de Bakker, "Anisotropic deposition of electroless nickel", Journal of the Electrochemical Society, vol. 140, pp. 2229-2235 (1993). Bibliography 223 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [74] S. W. Glunz, A. Mette, M. Alemán, P. L. Richter, A. Filipovic, and G. Willeke, "New concepts for the front side metallization of silicon solar cells", in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 746-749 (2006). [75] C. E. Dubé and R. C. Gonsiorawski, "Improved contact metallization for high efficiency EFG polycrystalline silicon solar cells", in Proceedings of the 21st IEEE Photovoltaic Specialists Conference, Kissimmee, Florida, USA, pp. 624-628 (1990). [76] R. E. Corwin, D. E. Riemer, and B. J. Stanbery, Electrodes for concentrator solar cells and methods for manufacture thereof in US Patent No. 4 320 250 The Boeing Company, USA (1982). [77] K. A. Münzer, R. R. King, R. E. Schlosser, H. J. Schmidt, J. Schmalzbauer, S. Sterk, and H. L. Mayr, "Manufacturing of back surface field for industrial application", in Proceedings of the 13th European Photovoltaic Solar Energy Conference, Nice, France, pp. 1398-1401 (1995). [78] K. A. Münzer, K. T. Holdermann, H. L. Mayr, R. E. Schlosser, and H. J. Schmidt, "High efficiency plated contact silicon solar cells and modules", in Proceedings of the 12th European Photovoltaic Solar Energy Conference, Amsterdam, The Netherlands, pp. 753-756 (1994). [79] J. Lyman, "Thick film, plating cut cost of laying down solar cell contacts and conductors", Electronics, vol. 53, pp. 40-41 (1980). [80] K. Holdermann, Verfahren zum Erzeugen feiner Elektrodenstrukturen in EP 0 542 148 B1 Siemens Solar GmbH, Europe (1993). [81] A. Schneider, L. Rubin, and G. Rubin, "Solar cell efficiency improvement by new metallization techniques -the day4TM electrode concepts", in Proceedings of the 4th World Conference on Photovoltaic Energy Conversion, Hawaii, USA, pp. 1095-1098 (2006). [82] R. Kopecek, T. Buck, A. Kränzl, J. Libal, K. Peter, A. Schneider, P. Sánchez-Friera, I. Röver, K. Wambach, E. Wefringhaus, and P. Fath, "Module Interconnection with alternative p- and n-type Si colar cells", in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 111-115 (2006). [83] S. Roberts, K. C. Heasman, and T. M. Bruton, "The reduction of module power losses by optimisation of the tabbing ribbon", in Proceedings of the 16th European Photovoltaic Solar Energy Conference, Glasgow, UK, pp. 2378-2381 (2000). [84] S. W. Glunz, J. Knobloch, D. Biro, and W. Wettling, "Optimized highefficiency silicon solar cells with Jsc=42 mA/cm2 and η=23.3 %", in Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 392-395 (1997). [85] A. G. Aberle, S. R. Wenham, and M. A. Green, "A new method for accurate measurements of the lumped series resistance of solar cells", in Proceedings 224 Bibliography _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] of the 23rd IEEE Photovoltaic Specialists Conference, Louisville, Kentucky, USA, pp. 133-139 (1993). J. Dicker, Dissertation Thesis, University Konstanz (2003). A. Rohatgi, J. R. Davis, R. H. Hopkins, P. Rai-Choudhury, and P. G. McMullin, "Effect of titanium, copper and iron on silicon solar cells", SolidState Electronics, vol. 23, pp. 415-422 (1980). M. Wolf and H. Rauschenbach, "Series resistance effects on solar cell measurements", Advanced Energy Conversion, vol. 3, pp. 455-479 (1963). P. P. Altermatt, G. Heiser, A. G. Aberle, A. Wang, J. Zhao, S. J. Robinson, S. Bowden, and M. A. Green, "Spatially resolved analysis and minimization of resistive losses in high-efficiency Si solar cells", Progress in Photovoltaics: Research and Applications, vol. 4, pp. 399-414 (1996). R. A. Sinton and A. Cuevas, "A quasi-steady-state open-circuit voltage method for solar cell characterization", in Proceedings of the 16th European Photovoltaic Solar Energy Conference, Glasgow, UK, pp. 1152-1155 (2000). R. A. Sinton, "Possibilities for process-control monitoring of electronic material properties during solar-cell manufacture", in Proc. 9th Workshop on Crystalline Silicon Solar Cell Materials and Processes, Golden, Colorado, USA, pp. 67-73 (1999). L. D. Swanson, private communication with M. Wolf and H. Rauschenbach (1960). R. J. Handy, "Theoretical analysis of the series resistance of a solar cell", Solid-State Electronics, vol. 10, pp. 765-775 (1967). European Committee for Electronical Standardization, "Procedure for temperature and irradiance corrections to measured I-V characteristics of silicon photovoltaic devices" European Standard EN60891, Brüssel (1994). A. Cuevas, G. L. Araújo, and J. M. Ruiz, "Variation of the internal series resistance with the operating conditions of a solar cell: dark and illuminated cases", in Proceedings of the 5th European Photovoltaic Solar Energy Conference, Athens, Greece, pp. 114-118 (1983). S. Bowden and A. Rohatgi, "Rapid and accurate determination of series resistance and fill factor losses in industrial silicon solar cells", in Proceedings of the 17th European Photovoltaic Solar Energy Conference, Munich, Germany, pp. 1802-1806 (2001). G. L. Araújo and E. Sánchez, "A new method for experimental determination of the series resistance of a solar cell", IEEE Transactions on Electron Devices, vol. ED-29, pp. 1511-1513 (1982). R. M. Swanson and R. A. Sinton, in Advances in solar energy; Vol. 6, edited by K. W. Böer, Plenum Press, New York, pp. 427-484 (1990). E. H. Nicollian and J. R. Brews, "MOS (metal oxide semiconductor) physics and technology", Wiley, New York Chichester Brisbane Toronto Singapore (1982). Bibliography 225 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [100] A. Grohe, B. Fleischhauer, R. Preu, S. W. Glunz, and G. P. Willeke, "Boundary conditions of the industrial production of LFC cells", in Proceedings of the 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, USA, pp. 1001-1004 (2006). [101] D. L. Meier and D. K. Schroder, "Contact resistance: its measurement and relative importance to power loss in a solar cell", IEEE Transactions on Electron Devices, vol. ED-31, pp. 647-653 (1984). [102] R. B. Campbell and A. Rohatgi, "Investigation of contact metallization systems for solar cells", Journal of the Electrochemical Society: Solid-State Science and Technology, pp. 2702-2704 (1980). [103] V. E. Lowe and A. C. Day, "High-temperature-stable contact metallization for silicon solar cells", IEEE Transactions on Electron Devices, vol. ED-31, pp. 626-629 (1984). [104] H. von Seefeld, N. W. Cheung, M. Mäenpää, and M.-A. Nicolet, "Investigation of titanium-nitride layers for solar -cell contacts", IEEE Transactions on Electron Devices, vol. 27, pp. 873-876 (1980). [105] S. Zhu, R. L. Van Meirhaeghe, S. Forment, G. Ru, and B. Li, "Effects of the annealing temperature on Ni silicide/n-Si(100) Schottky contacts", SolidState Electronics, vol. 48, pp. 29-35 (2004). [106] C. Canali, S. U. Campisano, S. S. Lau, Z. L. Liau, and J. W. Mayer, "Solidphase epitaxial growth of Si through palladium silicide layers", Journal of Applied Physics, vol. 46, pp. 2831-2836 (1975). [107] R. W. Bower and J. W. Mayer, "Growth Kinetics Observed in the Formation of Metal Silicides on Silicon", Journal Applied Physics, vol. 20, pp. 359-361 (1972). [108] B. R. Olaisen, A. Holt, E. S. Marstein, E. Sauar, A. Shaikh, K. McVicker, J. Salami, H. Miranda, and S. S. Kim, "Hot-melt screen-printing of front contacts on crystalline silicon solar cells", in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, USA, pp. 1084-1087 (2005). [109] Ferro (2007) online available: www.ferro.com [110] T. Williams, K. McVicker, A. Shaikh, T. Koval, S. Shea, B. Kinsey, and D. Hetzer, "Hot melt ink technology for crystalline silicon solar cells", in Proceedings of the 29th IEEE Photovoltaics Specialists Conference, New Orleans, Louisiana, USA, pp. 352-355 (2002). [111] Koenen (2007) online available: http://www.koenen.de [112] D. M. Huljic, D. Biro, R. Preu, C. C. Castillo, and R. Lüdemann, "Rapid thermal firing of screen printed contacts for large area crystalline silicon solar cells", in Proceedings of the 28th IEEE Photovoltaics Specialists Conference, Anchorage, USA, pp. 379-382 (2000). [113] D. M. Huljic, G. Grupp, R. Preu, and J. Horzel, "Comprehensive study of rapid thermal firing for industrial production of crystalline silicon thick-film solar cells", in Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, France, pp. 580-583 (2004). 226 Bibliography _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [114] G. Schubert, J. Horzel, R. Kopecek, F. Huster, and P. Fath, "Silver thick film contact formation on lowly doped phosphorous emitters", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 934-937 (2005). [115] D. Biro, O. Schultz, R. Preu, C. Ballif, D. Zickermann, and R. Lüdemann, "Screen printed selective emitters prepared by fast and rapid thermal diffusion", in Proceedings of the 17th European Photovoltaic Solar Energy Conference, Munich, Germany, pp. 1593-1596 (2001). [116] J. Horzel, J. Szlufcik, and J. Nijs, "High efficiency industrial screen printed selective emitter solar cells", in Proceedings of the 16th European Photovoltaic Solar Energy Conference, Glasgow, UK, pp. 1112-1115 (2000). [117] A. S. H. van der Heide, A. Schönecker J. H. Bultman W. C. Sinke "Explanation of high solar cell diode factors by nonuniform contact resistance", Progress in Photovoltaics: Research and Applications, vol. 13, pp. 3-16 (2005). [118] A. S. H. van der Heide, J. H. Bultman, J. Hoornstra, and A. Schönecker, "Error diagnosis and optimisation of c-Si solar cell processing using contact resistances determined with the CoreScanner", Solar Energy Materials and Solar Cells, vol. 74, pp. 43-50 (2002). [119] J. Horzel, Dissertation Thesis, Catholic University Leuven, Netherlands (1999). [120] W. Späth, Verfahren zur galvanischen Abscheidung einer Metallschicht auf der Oberfläche eines Halbleiterkörpers in Patentschrift 23 48 182, Siemens AG, Germany (1979). [121] L. F. Durkee, "Method of plating by means of light" in US Patent No. 4 144 139 Solarex Corporation, Rockville, Md., USA (1979). [122] L. A. Grenon, "Electroplating method" in US Patent No. 4 251 327 Motorola, Inc., Schaumburg, Ill., USA (1981). [123] L. A. Grenon, "Photoelectrolytic plating on silicon solar cells", in Proceedings of the Symposium on Materials and New Processing Technologies for Photovoltaics, Pennington, NJ, Electrochemical Society, Inc., 1981, Hollywood, Florida, USA, pp. 326-339 (1980). [124] L. A. Grenon, "Verfahren zum Elektroplattieren" in Offenlegungsschrift 30 06 716 Motorola, Inc., Schaumburg, Ill. (USA), Germany (1980). [125] V. M. Schmidt, "Elektrochemische Verfahrenstechnik", WILEY-VCH, Weinheim (2003). [126] R. Memming, "Semiconductor Electrochemistry", WILEY-VCH, Weinheim (1999). [127] B. Gaida, "Galvanotechnik in Frage und Antwort", 4th ed., Eugen G. Leuze, Saulgau (1983). [128] H. Gerischer, "The impact of semiconductors on the concepts of electrochemistry", Electrochimica Acta, vol. 35, pp. 1677-1699 (1990). Bibliography 227 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [129] G. Allardyce, J. Cahalen, T. Ridler, J. Rasch, O. Weigel, H. Fröhlich, H. Kappler, C. Rattey, A. Mette, C. Schetter, and S. Glunz, "The Commercial Application of Light Induced Electroplating for Improving the Efficiency of Crystalline Silicon Solar Cells", accepted for publication 22nd European Photovoltaic Solar Energy Conference, Milan, Italy (2007). [130] J. D. Jackson, "Classical Electrodynamics", 2nd ed., John Wiley & Sons, New York, London, Sydney, Toronto, USA (1975). [131] D. Kray, S. Baumann, K. Mayer, A. Eyer, and G. P. Willeke, "Novel techniques for low-damage microstructuring of silicon", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 156-159 (2005). [132] G. Grupp, D. Biro, G. Emanuel, R. Preu, F. Schitthelm, and G. Willeke, "Analysis of silver thick-film contact formation on industrial silicon solar cells", in Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, Florida, pp. 1289-1292 (2005). [133] G. Schubert, F. Huster, and P. Fath, "Physical understanding of printed thick-film front contacts of crystalline Si solar cells—Review of existing models and recent developments", Solar Energy Materials & Solar Cells, vol. 90, pp. 3399-3406 (2006). [134] D. Kray, M. Hermle, and S. W. Glunz, "Theory and experiments on the back side reflectance of silicon wafer solar cells", Progr. Photovolt., published online March 2007. [135] P. Hahne, E. Hirth, I. E. Reis, K. Schwichtenberg, W. Richtering, F. M. Horn, and U. Eggenweiler, "Progress in thick-film pad printing technique for solar cells ", Solar Energy Materials & Solar Cells, vol. 65, pp. 399-407 (2001). [136] D. Huljic, Verfahren zum strukturierten Aufbringen einer thermoplastischen Paste auf ein Substrat und dessen Verwendung in Deutsches Patent (DE 102004056492B3) Fraunhofer Gesellschaft, Germany (2006). [137] PDS-Consulting, "Pad printing technical manual", Vol. 1, Barnsley (1993). [138] Optomec™, "M3D™ Aerosol Jet Deposition System Manual", Albuqerque, New Mexico, USA (2005). [139] B. Wiley, Y. Sun, B. Mayers, and Y. Xia, "Shape-Controlled Synthesis of Metal Nanostructures: The Case of Silver", ChemInform, vol. 37 (2006). [140] A. Kamyshny, M. Ben-Moshe, S. Aviezer, and S. Magdassi, "Ink-Jet Printing of Metallic Nanoparticles and Microemulsions", Macromolecular Rapid Communications, vol. 26, pp. 281-288 (2005). [141] B. Y. Tay and M. J. Edirsinghe, "Dispersion and stability of silver inks", Journal of Materials Science, vol. 37, pp. 4653-4661 (2004). [142] D. Woiki, B. H. Günther, F. Petzhold, and M. Wagener, "Processing of nano-sized silver for new antimicrobial composite materials", in Proceedings of the International Conference on Nanotechnology & PM2, Providence, Rhode Island, pp. 119-125 (2003). 228 Bibliography _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [143] [144] [145] [146] [147] [148] [149] [150] [151] [152] [153] [154] [155] [156] Harima Chemicals (2007) online available: http://www.harima.co.jp/ Welt der Physik (2007) online available: http://www.weltderphysik.de SpecialChem (2007) online available:http://www.specialchem4coatings.com Gaiser Tool Company (2007) online available: www.gaiser.com L. M. Porter, A. Teicher, and D. L. Meier, "Phosphorus-doped, silver-based pastes for self-doping ohmic contacts for crystalline silicon solar cells", Solar Energy Materials and Solar Cells, vol. 73, pp. 209-219 (2002). D. L. Meier, H. P. Davis, R. A. Garcia, J. A. Jessup, and A. F. Carroll, "Selfdoping contacts to silicon using silver coated with a dopant source", in Proceedings of the 28th IEEE Photovoltaics Specialists Conference, pp. 6974 (2000). C. Khadilkar, S. Sridharan, T. Pham, A. Shaikh, and S. Kim, "Characterization of front contact in a silicon solar cell", in Proceedings of the 14th International Photovoltaic Science and Engineering Conference, Bangkok, Thailand, pp. 443-444 (2004). D. M. Huljic, C. Ballif, A. Hessler-Wyser, and G. Willeke, "Microstructural analyses of Ag thick-film contacts on n-type silicon emitters", in Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, Japan, pp. 83-86 (2003). C. Ballif, D. M. Huljic, G. Willeke, and A. Hessler-Wyser, "Silver thickfilm contacts on highly doped n-type silicon emitters: structural and electronic properties of the interface", Applied Physics Letters, vol. 82, pp. 1878-1880 (2003). M. M. Hilali, M. M. Al-Jassim, B. To, H. Moutinho, A. Rohatgi, and S. Asher, "Understanding the formation and temperature dependence of thick film Ag contacts on high-sheet-resistance Si emitters for solar cells ", Journal of the Electrochemical Society, vol. 152, pp. 742-749 (2005). C. Khadilkar, S. Sridharan, D. Gnizak, T. Pham, S. Kim, and A. Shaikh, "Effect of glass chemistry and silicon orientation on the front contact microstructure formation in a silicon solar cell", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 1291-1296 (2005). A. F. Carroll, D. E. Poser, R. J. S. Young, J. Raby, and M. Rose, "Advances in PV metallisation technology", in Proceedings of the 20th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 906-908 (2005). S. Sridharan, C. Khadilkar, T. Pham, and A. Shaikh, "Characterization of silver/glass/silicon front contact interface in a silicon solar cell", in 13th NREL Workshop on Crystalline Silicon Solar Cell Materials and Processes, Vail, Colorado, USA, pp. 162-165 (2003). C. Ballif, D. M. Huijic, A. Hessler-Wyser, and G. Willeke, "Nature of the Ag-Si interface in screen-printed contacts: a detailed transmission electron microscopy study of cross-sectional structures", in Proceedings of the 29th Bibliography 229 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ [157] [158] [159] [160] [161] [162] [163] [164] [165] [166] [167] [168] [169] IEEE Photovoltaics Specialists Conference, New Orleans, Louisiana, USA, pp. 360-363 (2002). O. Gzowski, L. Murawski, and K. Trzebiatowski, "The surface conductivity of lead glasses", Journal of physics, vol. 15, pp. 1097-1101 (1982). C. Ballif, S. Peters, and D. Borchert, "Local series resistance mapping of silicon solar cells by microwave photoconductivity decay measurements", Progress in Photovoltaics: Research and Applications, vol. 11, pp. 309-317 (2003). K. Firor, S. J. Hogan, J. M. Barrett, and R. T. Coyle, "Series resistance associated with thick-film contacts to solar cells", in Proceedings of the 16th IEEE Photovoltaic Specialists Conference, San Diego, California, USA, pp. 824-827 (1982). G. C. Cheek, R. P. Mertens, R. van Overstraeten, and L. Frisson, "Thickfilm metallization for solar cell applications", IEEE Transactions on Electron Devices, vol. ED-31, pp. 602-609 (1984). H. Murrmann and D. Widmann, "Current crowding on metal contacts to planar devices", IEEE Transactions on Electron Devices, vol. 16, pp. 10221024 (1969). G. K. Reeves and H. B. Harrison, "Obtaining the specific contact resistance from transmission line model measurements", IEEE Electron Device Letters, vol. 3, pp. 111-113 (1982). O. Gzowski, L. Murawski, and K. Trzebiatowski, "The surface conductivity of lead glasses", Journal of Physics D (Applied Physics), vol. 15, pp. 10971101 (1982). S. Kontermann, G. Emanuel, J. Benick, R. Preu, and G. Willeke, "Characterisation of silver thick-film contact formation on textured monocrystalline silicon solar cells", in Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, Germany, pp. 613-616 (2006). S. J. Proctor and L. W. Linholm, "A direct measurement of interfacial contact resistance", IEEE Electron Device Letters, vol. 3, pp. 294-296 (1982). D. K. Schroder, "Semiconductor Material and Device Characterization", John Wiley & Sons, New York (1990). W. M. Loh, S. E. Swirhun, T. A. Schreyer, R. M. Swanson, and K. C. Saraswat, "Modeling and measurement of contact resistances", in Proceedings of the 19th IEEE Photovoltaic Specialists Conference, New Orleans, Louisiana, USA, pp. 512-524 (1987). H. H. Berger, "Contact resistance and contact resistivity", Journal of the Electrochemical Society, vol. 119, pp. 507-514 (1972). D. Huljic, Diplomarbeit Thesis, University Konstanz, Germany (1998). 230 Danksagung _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Danksagung Auch wenn ich die vorliegende Dissertation, in Inhalt und Form selbst zu verantworten habe, ist sie ein Produkt, das ohne die Unterstützung vieler so hätte nicht zustande kommen können und auch nicht zustande gekommen wäre. Einige standen mir innerhalb der Dissertationszeit besonders nahe. Sie möchte ich hier ausdrücklich nennen und Ihnen mein herzliches Dankeschön sagen: - PD Dr. Volker Wittwer und Prof. Dr. Jürgen Wilde danke ich für die Begutachtung der Arbeit, sowie für wertvolle Tipps und Hinweise während der Anfertigung. - Stefan Glunz möchte ich ganz besonders danken für seine unermüdliche Unterstützung der letzten Jahre, für die große Freiheit meine Arbeit selbstständig durchführen zu können, für seine mitreißende Begeisterung nicht nur auf dem Gebiet der kristallinen Siliziumsolarzellen. - Ralf Preu möchte ich ebenfalls ganz besonders danken, für die vielen fruchtbaren Diskussionen, der großen Unterstützung besonders auch in der Anfangsphase meiner Arbeit, für die industrielle Sicht der Solarzellenherstellung. - Ein weiterer besonderer Dank gebührt Oliver Schultz, der mich während meiner gesamten Dissertationszeit wissenschaftlich begleitet hat, besonders auch dafür, dass er in der holprigen Anfangsphase ein wichtiger Ansprechpartner war. Auch Daniel Kray danke ich sehr für die vielen wertvollen Diskussionen und für seinen mitreißenden Humor. - Ein weiterer wichtiger Wegbegleiter war Gernot Emanuel, der mir durch sein unersetzliches Wissen auf dem Gebiet der industriellen Solarzellenherstellung sehr zum Erfolg dieser Arbeit verholfen hat. - Besonders bedanken möchte ich mich auch bei meinen Diplomanden Denis Erath, Aleksander Filipovic, Damian Pysch, Philipp Richter und Rodrigo Ruiz mit denen ich sehr gern zusammengearbeitet habe und die durch Ihre hervorragenden Arbeiten wesentlich zum Gelingen dieser Arbeit beigetragen haben. -Christian Schetter danke ich für die hervorragende Vorarbeit auf dem Gebiet der lichtinduzierten Galvanik und die vielen Diskussionen auf dem Gebiet der Chemie. -Elisabeth Schäffer danke ich für das Charakterisieren von Solarzellen mit den verschiedensten Methoden und der steten Unterstützung, auch wenn es mal nicht funktioniert hat. Danksagung 231 _______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ - Daniel Spinner danke ich für die unkomplizierte und umgehende Hilfe bei kleinen und großen computerrelevanten Problemen. - Ein weitere Dank gebührt Antonio Leimenstoll, Harald Lautenschlager, Sonja Seitz und Franz Josef Kammerewerd, für die Prozessierung hocheffizienter Solarzellen, Katinka Kordelos für unermüdliches IV-messen von Solarzellen, sowie Anke Herbolzheimer und Katja Krüger für die hervorragende Unterstützung im Bereich der lichtinduizierten Galvanik. - Auch Dank sagen möchte ich den Praktikanten und wissenschaftlichen Hilfskräften Georg Simon, Enrique Barrigon und Daniel Hertkorn, die durch zahlreiche Messungen und Analysen mich sehr unterstützt haben. - Meinen Doktoranden-Mitstreitern Monica Alemán, Sybille Hopman, Jan Benick, Florian Clement, Stephan Diez, Filip Granek, Andreas Grohe, Daniela Grote, Matthias Hörteis, Martin Hermle, Marc Hofmann, Stefan Kontermann, Nicola Mingurelli, Thomas Roth und Martin Schubert danke ich für die gute Zusammenarbeit und vielen Diskussionen. - Eric Schneiderlöchner und Wibke Wittmann danke ich für engagiertes Korrekturlesen. - Sandra Lust, Dirk Wissen, Andreas Hubert von Q-Cells, Joachim Rasch, George Allardyce, Oliver Weigel von RHEM, Werner Andreas Maurer, Heinz Kappler und Holger Fröhlich von Schmid für die gute Zusammenarbeit auf dem Gebiet der lichtinduzierten Galvanik. - Marcelino Essien und King Bruce von Optomec für die Unterstützung auf dem Gebiet des Aerosolstrahldruckens, sowie Frank Fidorra und Willi Brendle von QCells für die Bereitstellung vorprozessierter Solarzellen. - Bianca Böttge und Sandy Bennemann vom Fraunhofer IWM in Halle danke ich für die kurzfristige Anfertigung von FIB Schnitten am Ende meiner Arbeit. - Auch möchte ich dem gesamten SolProV Konsortium für die interessante Projektarbeit auf dem Gebiet der Hotmelt Technologie danken. Explizit erwähnen möchte ich Elisabeth Kasper und André Noppe von Ferro. - Meinen Eltern danke ich für die stete Unterstützung und dass Sie mich immer meinen Weg gehen lassen haben; meinen beiden Geschwistern dafür, dass sie immer im richtigen Moment für mich da sind. - Mein größter Dank geht an meine Britta, für Ihre unendlich große Geduld, für Ihre wertvollen Ratschläge, besonders aber für Ihre Liebe und Kraft die Sie mir jeden Tag aufs Neue schenkt.