Novel Uses of Titanium Dioxide for Silicon Solar Cells
A thesis submitted as partial fulfillment of the requirement for the Degree of
Doctor of Philosophy
by
Bryce Sydney Richards
at the
Centre for Photovoltaic Engineering and the School of Electrical Engineering University of New South Wales Sydney 2052 New South Wales
Australia
April 2002
CENTRE FOR PHOTOVOLTAIC ENGINEERING UNSW Certificate of Originality
I hereby declare that this submission is my own work and that, to the best of my knowledge and belief, it contains no material previously published or written by any other person nor material which to a substantial extent has been accepted for the award of any other degree or diploma of a university or other institute of higher learning, except where due acknowledgment is made in the text.
I also declare that the intellectual content of this thesis is the product of my own work, even though I may have received assistance from others on style, presentation and language expression.
Bryce Richards
Richards, Bryce Sydney Novel Uses of Titanium Dioxide for Silicon Solar Cells PhD Thesis Centre for Photovoltaic Engineering The University of New South Wales Sydney, NSW 2052, Australia Copyright c 2002 all rights reserved
ISBN 0 7334 1971 2 Abstract
Titanium dioxide (TiO2) thin films have a long history in silicon photovoltaics (PV) as antireflection (AR) coatings due to their excellent optical properties and low deposition cost. This work explores several novel areas where TiO2 thin films could be use to enhance silicon (Si) solar cell performance while reducing device fabrication costs.
Amorphous, anatase and rutile TiO2 thin films are deposited using ultrasonic spray- deposition (USD) and chemical vapour deposition (CVD) systems, both designed and con- structed by the author. Initial experiments confirmed that no degradation in the bulk minority carrier lifetime (τbulk) occurred during high-temperature processing, although the stability of the USD-deposited TiO2 films was dependent on the furnace ambient.
A major disadvantage of TiO2 AR coatings is that they afford little surface passivation. In this work, a novel method of achieving excellent surface passivation on TiO2-coated silicon wafers is presented. This involved growing a 6 nm-thick SiO2 layer at the TiO2:Si interface by oxidising the wafer after TiO2 film deposition. The increase in surface passivation afforded by the interfacial SiO2 layer results in a decrease in the emitter dark saturation current density −14 2 (J0e) by nearly two orders of magnitude to 4.7 − 7.7 × 10 A/cm . This demonstrates the compatibility of the TiO2/SiO2 stack with high-efficiency solar cells designs.
By varying the film deposition and annealing conditions, TiO2 refractive indices in the rangeof1.726 − 2.633 (at λ = 600 nm) could be achieved. Subsequently, a double-layer antireflection (DLAR) coating was designed comprised of low and high TiO2 refractive index material. The best experimental weighted average reflectance (Rw) achieved was 6.5% on a planar silicon wafer in air. TiO2 DLAR coatings are ideally suited to multicrystalline silicon (mc-Si) wafers, which do not respond well to chemical texturing.
Modelling performed for a glass and ethyl vinyl acetate (EVA) encapsulated buried-contact solar cell indicated that a TiO2 DLAR coating afforded a 7% increase in the short circuit current density, when compared to a standard, commercially-deposited TiO2 single-layer AR coating.
Finally, it is demonstrated that chemical reactions with phosphorus prevent TiO2 from acting as a successful phosphorus diffusion barrier or dopant source. The applicability of TiO2 thin films to various silicon solar cell structures is discussed. Acknowledgements
Many people contributed to the success of this work and my survival throughout.
First, and foremost, I need to thank (yes, thank!) my partner, Andrea Sch¨afer, for leading me along the path to the PhD. Somehow witnessing all the good and the bad moments during her PhD, ended up creating a positive image for me! Andrea also provided invaluable guidance and tips along the way, and created many shortcuts through the bureaucracy for me. I will never forget your assistance Andrea, and am deeply indebted to you. Vielen Dank, and may our love only grow stronger.
Another big Danke, goes out to our daughter, Moana Sch¨afer,who witnessed just over half of my efforts. Thanks for keeping my feet firmly planted on the ground and not letting me drift too far off into “PhD land”! Thank you for all our fun times, and my apologies for the times when my patience wasn’t sufficient to see your needs.
Naturally, I would like to thank the input from my supervisors of the years: to Stuart Wenham (UNSW) for his enthusiasm and encouragement; to Christiana Honsberg (UNSW and Georgia Institute of Technology, U.S.A.) for her moral and monetary support; to Francesca Ferrazza (Eurosolare S.p.A.) for the opportunities to see the “real” side of photovoltaics and for being a true friend; and to Jeff Cotter for his valuable advice during the latter stage of the thesis.
Several members brought their own special personalities to the Centre and made it a fun and challenging work place. These people include Keith McIntosh, Hamid Mehrvarz, Holger Neuhaus, Alex Slade, Bernhard Vogl, Rob Bardos, Matt Boreland, Martin Bruahart and Tom Puzzer. Thanks for all the ethical, moral and technical conversations. Thanks too for the great computer support, Laurie!
I would like to thank other people who assisted with TiO2 thin film characterisation: Dr. Tom Puzzer (UNSW) for SEM/AFM training; Prof. Robert Lamb (UNSW) and Dr. Matt Boreland (Toyota Technical Institute, Japan) for XPS analysis; Prof. David Jamieson (Univ. of Melbourne) for RBS analysis; Sally Rowlands and Prof. Trevor Redgrave (both Univ. of Western Australia) for training and access to the variable-angle spectroscopic ellipsometer (VASE); Dr. Alistair Sproul (UNSW), author of the forthcoming book “Ellipsometry for Dummies”; and to my father, Dr. Ray Richards (Lower Hutt, New Zealand), for his assistance in bringing me up to speed on thermochemistry analysis. 2
I am grateful for the guidance in my career provided by Dr. Andrea Sch¨afer,Prof. Mark Wain- wright, Prof. Stuart Wenham and Prof. Martin Green. The financial support provided by the Faculty of Engineering, the School of Electrical Engineering and the Centre for Photovoltaic Engineering was greatly appreciated. Publications Resulting from this Thesis (to date)
B.S. Richards (2004) Comparison of Dielectric Coatings for Buried-Contact Solar Cells: A Review, Progress in Photovoltaics 12 (in press).
B.S. Richards, S.R. Richards, M.B. Boreland, D.N. Jamieson (2004) High Temperature
Processing of TiO2 Thin Films for Application in Silicon Solar Cells, Journal of Vacuum Science and Technology A, 22(2): 339-348.
B.S. Richards, S.F. Rowlands, A. Ueranatasun, J.E. Cotter, C.B. Honsberg (2004) Reducing the Production Costs of Buried-Contact Solar Cells using Titanium Dioxide Thin Films, Solar Energy, 76(1-3): 269-276.
B.S. Richards (2003) Single-Material TiO2 Double-Layer Antireflection Coatings, Solar Energy Materials and Solar Cells, 79(3), 369-390.
B.S. Richards, S.F. Rowlands, C.B. Honsberg, J.E. Cotter (2003) TiO2 DLAR Coatings for Planar Silicon Solar Cells, Progress in Photovoltaics, 11(1), 27-32.
B.S. Richards, J.E. Cotter and C.B. Honsberg (2002) Enhancing the surface passivation of
TiO2 coated silicon wafers, Appl. Phys. Letters, 80(7), 1123-1125.
B.S. Richards, S.F. Rowlands, A. Ueranatasun, J.E. Cotter, and C.B. Honsberg (2001) Reducing the production costs of buried-contact solar cells using titanium dioxide thin films, Intl. Solar Energy Society Solar World Congress, 26-30 November, Adelaide.
B.S. Richards, J.E. Cotter, C.B. Honsberg and S.R. Wenham (2000) Novel Uses of TiO2 Films in Crystalline Silicon Solar Cells, 28th IEEE Photovoltaic Specialists Conference, Alaska, 375-378.
C.B. Honsberg, J.E. Cotter, K.R. McIntosh, S. Pritchard, B.S. Richards and S.R. Wenham, (1999), Design strategies for commercial solar cells using the buried contact technology, IEEE Trans. Electron Devices, 46(10), 1984-92.
B.S. Richards, J.E. Cotter, F. Ferrazza, C.B. Honsberg and S.R. Wenham (1998) Lowering the cost of commercial silicon solar cells, Proc. of the Environmental Engineering Research Event 1998, Avoca Beach, New South Wales, 303-308.
J.E. Cotter, B.S. Richards, F. Ferrazza, C.B. Honsberg, T.W. Leong, H.R. Mehrvarz, G.A. Naik and S.R. Wenham (1998) Design of a simplified emitter structure for buried contact solar cells, 2nd World Conference Photovoltaic Energy Conversion, Vienna, 1511-1514. Contents
1 Introduction 9
1.1MotivationforthisWork...... 9
1.2 Australia’s Solar Energy Resource ...... 11
1.3 Brief Theory of Solar Cell Operation ...... 12
1.4 Commercially Produced Silicon Solar Cells ...... 14
1.4.1 Screen-Printed Solar Cells ...... 14
1.4.2 Buried-Contact Solar Cells ...... 15
1.4.3 Buried-Contact Solar Cell Fabrication Sequence ...... 16
1.4.4 Simplified Buried-Contact Solar Cell ...... 16
1.5 Multicrystalline Silicon ...... 18
1.5.1 Issues with Multicrystalline Silicon ...... 19
1.6 Why use Titanium Dioxide? ...... 20
1.6.1 TiO2 Thin Films in Photovoltaics ...... 21
1.7 Thesis Overview and Goal ...... 25
2 Common Properties of TiO2 Thin Films 29
2.1Introduction...... 29
2.2PhysicalProperties...... 30
2.2.1 The Amorphous − Anatase − Rutile Phase Transformations...... 30
2.2.2 The Effect of Impurities on the Anatase − Rutile Phase Transformation 32
2.2.3 Substrate Type ...... 33
3 4 CONTENTS
2.2.4 Film Defects ...... 34
2.2.5 Film Density ...... 34
2.2.6 Non-Stoichiometric TiO2−x Thin Films ...... 36
2.3OpticalProperties...... 37
2.3.1 Refractive Index, Extinction Coefficient and Scattering...... 37
2.3.2 TiO2 Thin Film vs. Single Crystal ...... 39
2.3.3 Variation with Deposition and Annealing Temperature ...... 40
2.3.4 Variation with Deposition and Annealing Ambient ...... 45
2.3.5 Variation with Other Deposition Conditions ...... 48
2.3.6 Optical Properties of Highly Porous TiO2 Films ...... 48
2.4ElectricalProperties...... 50
2.4.1 TiO2: Insulator or Conductor? ...... 50
2.4.2 Non-Stoichiometric TiO2−x Thin Films ...... 50
2.4.3 Variation with Deposition or Annealing Ambient ...... 51
2.4.4 Doped TiO2 ...... 51
2.5 Chemical Properties ...... 52
2.5.1 Chemicals used in Making Solar Cells ...... 52
2.5.2 HydrofluoricAcid...... 54
2.5.3 Other Acids and Bases ...... 56
2.6 Conclusions ...... 57
3TiO2 Thin Film Deposition Equipment 59
3.1Introduction...... 59
3.2 Overview of TiO2 Thin Film Deposition Methods ...... 62
3.3UltrasonicSprayDeposition...... 65
3.4 Theory of Ultrasonic Spray Deposition ...... 67
3.5 TPT: The TiO2 Precursor ...... 70
3.5.1 WhyTPT?...... 70 CONTENTS 5
3.5.2 The TPT→TiO2 Reaction...... 72
3.6DesignofUltrasonicSprayDepositionSystem...... 74
3.6.1 Selection of Ultrasonic Nozzle ...... 74
3.6.2 UltrasonicNozzlePerformance...... 76
3.6.3 Liquid Delivery ...... 77
3.6.4 Substrate Heater ...... 78
3.6.5 MotorizedStage...... 81
3.6.6 Spray Shaping ...... 82
3.6.7 Miscellaneous Equipment ...... 82
3.6.8 Operation of the TiO2 SpraySystem...... 85
3.7DesignofCVDSystem...... 87
3.7.1 Motivation...... 87
3.7.2 TPT Bubbler and Temperature Control ...... 87
3.7.3 Water Vapour Bubbler ...... 88
3.7.4 OperationoftheCVDSystem...... 88
3.8 Conclusions ...... 89
4 Characterisation of TiO2 Thin Films 91
4.1Introduction...... 91
4.2FTIRSpectroscopy...... 92
4.3RamanSpectroscopy...... 95
4.4X-rayPhotoelectronSpectroscopy...... 98
4.5RutherfordBack-ScatteringSpectroscopy...... 99
4.6 Ellipsometry ...... 101
4.6.1 Overview...... 101
4.6.2 Ellipsometers ...... 104
4.6.3 Lorentz Oscillator Model ...... 106
4.6.4 SurfaceRoughnessModel...... 107
4.6.5 Ellipsometric measurements of Spray Deposited TiO2 Thin Films . . 107 6 CONTENTS
4.6.6 SE measurements of CVD TiO2 Thin Films ...... 109
4.7 Reflectance Spectrophotometry ...... 109
4.8ElectronMicroscopy...... 109
4.8.1 USD-Deposited TiO2 Thin Films ...... 110
4.8.2 APCVD-Deposited TiO2 Thin Films ...... 110
4.8.3 CVD-Deposited TiO2 Thin Films ...... 110
4.9AtomicForceMicroscopy...... 115
4.10 Chemical Resistance ...... 117
4.11 Conclusions ...... 118
5 Enhancing the Passivation of TiO2-coated Wafers 121
5.1Introduction...... 121
5.2 Stability of TiO2 atHigh-Temperatures...... 123
5.2.1 Titanium Contamination of Silicon ...... 123
5.2.2 Reduction of TiO2 ...... 125
5.3 Methods of Achieving Surface Passivation with TiO2 Thin Films ...... 130
5.3.1 Growth of SiO2 at the TiO2:SiInterface...... 133
5.3.2 TiO2 onPSG...... 137
5.4 Conclusions ...... 139
6 Performance of TiO2 Thin Films as a Phosphorus Diffusion Barrier and Dopant Source 141
6.1Introduction...... 141
6.2 TiO2 as a Phosphorus Diffusion Barrier ...... 142
6.2.1 Experiment...... 144
6.2.2 Results...... 144
6.2.3 Discussion...... 146
6.3 TiO2 asaPhosphorusDopantSource...... 149
6.3.1 Experiment...... 151
6.3.2 ResultsandDiscussion...... 151 CONTENTS 7
6.4 Conclusions ...... 152
7TiO2 Antireflection Coatings 153
7.1Introduction...... 153
7.2 Previous Developments in AR Coatings ...... 155
7.2.1 TheoryandDesignofARCoatings...... 155
7.2.2 TiO2 ARCoatings...... 160
7.2.3 Silicon Nitride AR Coatings ...... 165
7.3 Varying the Optical Properties of TiO2 ...... 167
7.3.1 DepositionTemperature...... 168
7.3.2 Annealing Temperature ...... 174
7.3.3 Deposition Ambient ...... 178
7.3.4 Annealing Ambient ...... 181
7.4 Development of Novel TiO2 ARCoatings...... 183
7.4.1 Single-layer TiO2 ARCoatings...... 183
7.4.2 Double-layer TiO2 ARCoatings...... 186
7.5 Performance of TiO2 DLAR-Coated Solar Cells ...... 193
7.6 Conclusions ...... 196
8 Conclusions 199
8.1 Summary ...... 199
8.2 Applicability to Various Solar Cell Processes ...... 203
8.3SuggestionsforFurtherWork...... 205
ATiO2 AR Coating Modelling Parameters 207
A.1 Variation of n and k withDepositionTemperature...... 208
A.2 Variation of n and k with Annealing Temperature ...... 212
A.3 Variation of n and k with Deposition Ambient ...... 218
A.4 Variation of n and k with Annealing Ambient ...... 220
A.5 TiO2 DLARCoatings...... 225 8 CONTENTS Chapter 1
Introduction
Photovoltaics (PV) will play an important role in the world’s future energy trends, however the major hurdles faced in widespread implementation of renewable energy are of a social, not technical, nature. This chapter briefly discusses Australia’s potential to become a major player in the future solar energy industry. The differences between the two dominant com- mercially produced silicon solar cells, screen-print (SP) and buried-contact (BC), are briefly examined. The BC technology was introduced in the mid-1980’s and designed for single crystal silicon (c-Si) wafers. A trend over recent years has seen lower-cost, multicrystalline silicon (mc-Si) wafers now dominate the marketplace. The BC fabrication sequence has several lengthy high-temperature processing steps fabrication. The processing costs currently outweigh the performance enhancement offered by the buried-contact technology when using mc-Si wafers and, today, only SP solar cells are produced on mc-Si substrates.
The most crucial high-temperature processing step in the buried-contact fabrication sequence is the growth of a thick, thermal oxide layer. If this layer could be replaced by a thin di- electric film, deposited at low temperature and such a film was able to withstand processing in phosphorus-containing ambients, a buried-contact solar cell could be fabricated with about one hour of high-temperature processing steps. This proposition seemed attractive for ap- plication of the BC technology to multicrystalline silicon wafers in a cost-effective manner.
One logical choice for this film is titanium dioxide (TiO2), due to its prevalence in the PV industry, high optical performance and low cost. A review of roles that TiO2 films have played in the photovoltaics industry is presented, before discussing the overview of the remainder of the thesis.
1.1 Motivation for this Work
The amount of solar energy that strikes earth in a period of a few days is greater than the amount of fuel burnt overt the course of the whole human history,1 which encourages one to
9 10 1. Introduction think about ways in which this energy could be effectively harnessed in order to satisfy our ever-increasing demand for energy. While this may sound like a gross oversimplification of the impending energy crisis facing humankind, it brings into perspective the amount of power that is constantly being radiated by the sun. The primary focus of most developed nations needs at this stage should be looking at ways of radically reducing the current consumption of energy and raw materials. In fact, we are all so used to consuming that our language now defines our role in society as “consumers”. A 1995 study stated that for each American, 20 tonnes of new materials have to be provided every year, including energy equivalent to 7.6 tonnes of oil (or 12 tonnes of coal).2 If the world’s projected population in the year 2070 consumed energy at this rate, the world energy production would have to be fourteen times greater than its current capacity and all potentially recoverable energy resources would be depleted in about eighteen years2!
Therefore, if close to 11 billion people are to live on this planet in 70-years time we need to become “conservers” rather than consumers. The earth does not contain enough resources for each person in a developed nation to sustain our resource-intensive lifestyle. In the same manner, the economies of Western nations simply cannot continue to grow at targeted rate of 4% p.a., no matter what the politicians say. The Australian environmental thinker and activist, Ted Trainer, claims that the only type of economic growth that can be said to be truly ecologically sustainable is one that has a 0% growth rate per annum.2 Trainer also argues that “technological fixes” will not get us there either, as the main problems faced are social.2 So while the pursuit of renewable energy is a worthy cause, a truly sustainable solution can only be reached if we were to consume a small fraction of what we use in our affluent lifestyle today.
If photovoltaics (PV), the direct conversion of sunlight into electricity via solar cells, is to have a major and timely impact upon our current global predicament, the costs of production need to be significantly reduced before PV can compete head to head with fossil fuels. The cost of generating electricity directly from solar cells is slowly, but steadily, reducing. Figure 1.1 shows that the cost of purchasing PV modules is expected to reduce from today’s 1 US$4 /watt-peak (Wp)to 100.00 1978 1980 1985 1990 10.00 1995 1998 2000 1.00 1 10 100 1000 10000 Average selling price (1998 price US$/Watt) Average selling Accumulated shipments (megawatts) Figure 1.1: Historical cost of purchasing PV modules from 1978 to 2000, and an extrapolation of the linear trend to increased shipments in the future (from Green1). 1.2 Australia’s Solar Energy Resource Australia has recently implemented a PV rebate scheme has at both State and Federal level in an attempt to help Australia meet its target of generating an additional 9500 GWh (about 1%) of electricity from renewable energy sources by the year 2010. There is definite potential for a much greater expansion in PV in Australia, considering that many areas of Australia receive more than twice the amount of solar radiation than countries such as Germany (as shown in Figure 1.2), which already has a large installed PV capacity. It is also apparent from Figure 1.2(a) that the majority of Australians do not live in the sunny central and northern regions of the country. However, the cities of Perth, Adelaide, and all of the densely populated East Coast (Sydney and further north) still receive four or more sunshine hours a day of full sunshine (1 kW/m2). This is an excellent (and free) resource, which remains relatively untapped. Additionally, PV panels can easily be mounted on existing north-facing roofs and included into fa¸cades of office buildings. This means that power can be generated from within densely populated areas, obviating the requirement of setting aside large amounts of land for such purposes. Australia’s excellent research record in the field of solar energy along with its abundant resources make it uniquely positioned to become a major player in the future solar energy industry. One only needs to look to Denmark in the case of wind energy to see how a strong domestic market can lead to a small country dominating the world market. 12 1. Introduction Figure 1.2: Maps indicating the level of solar radiation received on a horizontal plane in (a) Australia and (b) Europe (in units of kWh/m2).4 1.3 Brief Theory of Solar Cell Operation A typical silicon solar cell is shown in Figure 1.3(a). The standard n+-p silicon solar cell has a shallow junction formed near the front surface, a front ohmic contact in the form of fingers and a busbar, and a full metal rear ohmic contact. Light that is absorbed by the solar cell generates an electron-hole (e-h) pair that is able to contribute to current flow from the device. For small photon energies, the majority of the current is generated in the base (about 300 µm thick), while photons of energy greater than 2.5 eV generate current from within the first 1 µm of silicon. This is the reason why the recombination velocity at the front surface can have such a profound effect on this high-energy photocurrent. For high-efficiency solar cells the effect of rear surface recombination also becomes an important design parameter. The area-normalised current density J (mA/cm2), of the solar cell is given by V +JRs V + JR ( nkT /q ) s J = JL − J0 exp −1 + ,(1.1) Rsh 2 where J0 is the dark saturation current density (mA/cm ), JL is the light generated current density (mA/cm2), q is the electronic charge (1.602 × 10−19 C), V is the operating voltage, −23 k is Boltzmann’s constant (1.380 × 10 J/K), Rs is the series resistance, Rsh is the shunt resistance, n is the diode ideality factor, and T is the operating temperature (K). The special case, where the voltage is zero and J = JL is defined to be the short-circuit current density, Jsc. Figure 1.3(a) shows the equivalent circuit of a solar cell, including series and shunt resistances. The Jsc is limited by optical losses such as reflectance from the front surface, shading of the front surface due to the metal contacts, and transmission of lower energy light out the back of the cell. An additional source of current loss is due to minority carriers recombining at the surfaces before they are collected by the junction. 1.3 Brief Theory of Solar Cell Operation 13 J – Pmpp, Vmpp Jsc Rs J V V mpp, mpp q(V-JRs )/nkT JL J0(exp -1) Rsh Current Density, Power Current Density, + Voltage Voc (a) (b) Figure 1.3: (a) Equivalent circuit of the basic p-n junction solar cell, and (b) a typical I − V (solid) and P − V (dashed) curve of an illuminated solar cell. The operating point where J = 0 mA/cm2 in Equation 1.1 defines the open-circuit voltage Voc of the solar cell, as shown in Figure 1.3(b). The minimization of both bulk and surface recombination are important in order to maximise the voltage at maximum power point. Excellent surface passivation can be achieved by using a thermally grown silicon dioxide (SiO2) layer, as discussed in Section 5.3.1. Bulk recombination will always be higher in mc-Si wafers compared to high quality c-Si FZ wafers, due to the grain boundaries between the crystals, although lifetime enhancement measures such as gettering and hydrogenation may improve the situation somewhat. In practice, the highest Voc’s of mc-Si and c-Si solar cells have been limited to 657 mV5 and 710 mV.6 The most efficient operating point of the solar cell is at the maximum-power-point voltage Vmpp and maximum-power-point current Impp, as shown in Figure 1.3(b). At this point, solar energy is converted into electrical energy with an efficiency of I V η = mpp mpp , (1.2) Pin where Pin is the incident solar power. Another parameter commonly used is the fill factor (FF), which is a measure of the “squareness” of the I − V curve and is defined as I V FF = mpp mpp .(1.3) Isc Voc Recombination in the depletion region will lower a solar cells fill factor, as will a small shunt resistance or a large series resistance. The interested reader is referred to the books by Green7, 8 and van Overstraeten and Mertens9 for a more detailed discussion on device physics and operation. 14 1. Introduction 1.4 Commercially Produced Silicon Solar Cells 1.4.1 Screen-Printed Solar Cells The first reference to a commercial screen-printed (SP) solar cell production line can be found in 1975.10 Since this time the process has remained relatively unchanged. Figure 1.4(a) shows the structure of a screen-printed (SP) solar cell. As the name implies, the metal contacts in Figure 1.4(a) are formed by screen-printing a metallic paste through a mask. However, the contacts are the main limitation of the device, which has an efficiency of typically 12 − 13%. This is because of the paste’s poor conductivity, its poor contact resistance to silicon, the poor aspect ratios achieved, and the inability to reliably produce thin lines in production. The nett result is that the metal lines are much wider than desirable (150−200 µm). To still allow a significant fraction of the light (92%) to strike the front surface of the silicon, the fingers are spaced widely (about 3 mm) apart. The electrons are then required to travel a large lateral distance through the thin emitter region before reaching the metal contact. For this reason the emitter is heavily doped with n-type (negative) carriers, affording better lateral conductivity. This, however, gives the cell a very poor response to high-energy (ultraviolet-blue) light which is absorbed very close to the front surface. An aluminium (Al) or silver/aluminium (Ag/Al) paste is screen-printed onto the rear of the cell and fired in a furnace to form a p-type ohmic contact. The aluminium also creates a back- surface-field (BSF), which repels electrons that travel towards the rear of the cell instead of towards the junction. A titanium dioxide (TiO2) antireflection (AR) coating is deposited near the end of the process, increasing the amount of light absorbed by the silicon. The total amount of high-temperature (> 750◦C) processing involved in fabricating a SP solar cell is typically about 30 − 45 min. Modules fabricated today with SP cells typically cost about US$3.50 /Wp. Figure 1.4: Schematic diagrams of a (a) screen-printed solar cell and a (b) buried-contact solar cell (adapted from Green8). 1.4 Commercially Produced Silicon Solar Cells 15 1.4.2 Buried-Contact Solar Cells The performance of commercial silicon solar cells was enhanced greatly through the devel- opment of the buried-contact (BC) solar cell at the University of New South Wales in the mid-1980’s.11, 12 The BC solar cell, shown in Figure 1.4(b), is currently commercially pro- duced in large volumes under license by BP Solar. Conversion efficiencies of greater than 20% have been achieved on laboratory scale cells at UNSW,11 and an independently mea- sured production BC solar cell had an efficiency of 16.7%.13 The fingers are only 20 µm wide but are 50 µm deep, the grooves being made either with a laser or a mechanical dicing saw. The cell is less shaded due to the reduced metal area (about 3%) on the front surface, and therefore the fingers can be placed closer together, permitting the use of a lightly-doped emitter and giving the cell an excellent response to blue light. Figure 1.5 provides a visual comparison of the metal contact areas of a SP and BC solar cell. It can be easily seen that the fingers of the BC solar cell are much finer and occupy a much smaller fraction of the solar cell front surface. Figure 1.5: Scanned images of (a) a Solarex screen-printed solar cell on mc-Si, and (b) a BP Solar buried-contact solar cell on textured c-Si. The AR coatings used are titanium dioxide (TiO2) and silicon nitride for the Solarex and BP Solar cell, respectively (not to scale). The drawback of the BC process is that although there are substantial materials costs savings relative to SP cells, the processing costs are higher, over 30% of which can be attributed to the high-temperature processing steps14 (see Section 1.4.3). In making the BC technology commercially-viable, BP Solar have removed the lengthy high-temperature oxidation step and have replaced it with an alternative dielectric, namely silicon nitride.15, 16 The silicon nitride is deposited using a low-pressure chemical vapour deposition (LPCVD) system.16 LPCVD silicon nitride is typically deposited at a temperature of about 700◦C,17 and is 16 1. Introduction capable of acting as a phosphorus diffusion barrier and metallisation mask as well as an AR coating.16 While the LPCVD silicon nitride film acts in many ways as a drop-in replacement for the thermally-grown silicon dioxide (SiO2) layer, the high deposition temperature means that the surface passivation benefits normally associated with silicon nitride are not realised in that process.16, 18 1.4.3 Buried-Contact Solar Cell Fabrication Sequence Figure 1.6 describes the fabrication sequence for a single sided BC solar cell on a high quality float zone (FZ) c-Si wafer. Four high-temperature (> 750◦C) steps of varying length are involved in fabricating a BC solar cell: i) a deep, high-quality emitter is formed by performing a light (low dose) n-type (phos- phorus, P) diffusion on the wafers. This creates the collecting junction in the p-type wafers. While this process is relatively short (15 min), the length of the following processes are in the order of hours each. ii) a thick SiO2 layer is grown a) as a diffusion barrier to protect the lightly doped emitter from the heavy groove diffusion, b) to bond with atoms of the disrupted silicon lattice at the surface, thereby improving the surface passivation, c) to facilitate electroless metal plating of the front contacts, and d) to act as an AR coating, reducing reflection losses from the cell’s front surface. iii) a heavy phosphorus diffusion (close to the solid solubility limit) in the grooves permits good electrical contact between the silicon and the metal. iv) an evaporated Al film is sintered for a few hours to form an ohmic contact and create a BSF at the rear. 1.4.4 Simplified Buried-Contact Solar Cell The simplified buried-contact (SBC) solar cell was developed in an attempt to reduce the number of high-temperature processing steps in the standard BC fabrication sequence.19, 20 In the SBC solar cell a single emitter and groove diffusion is performed, as opposed to the two separate diffusions in the standard BC process. This reduces the number of high- temperature steps by one. More importantly, by performing the homogeneous diffusion before the deposition of the AR coating, the limitations on the choice of front-surface dielec- tric layer are considerably relaxed. The dielectric film now only has to act as a metallisation mask and a good AR coating, and does not need to act as a phosphorus diffusion barrier. This opens the door for low-temperature deposited films such as plasma-enhanced chemical 1.4 Commercially Produced Silicon Solar Cells 17 6DZGDPDJHHWFK 1D2+IRUPLQDW±G& )XOOFOHDQ 5&$5&$5&$+)GLS /LJKWSKRVSKRUXVHPLWWHUGLIIXVLRQ &IRUPLQ 7&$DQGZHWR[LGDWLRQKUWRWDODW ±G&JURZaQP6L2 )URQWJURRYHVFULELQJ[P 1D2+JURRYHHWFK 1D2+IRUPLQDW±G& )XOOFOHDQ5&$5&$5&$ +HDY\SKRVSKRUXVJURRYHGLIIXVLRQ &IRUPLQ 5HDUDOXPLQLXPHYDSRUDWLRQ ±PWKLFNQHVV $OXPLQLXPDOOR\&IRUKUV PLQ%+)WRUHPRYH6L2 IURPJURRYHV 1LFNHODQGFRSSHUHOHFWUROHVV PHWDOSODWLQJ 6L2WKLQQLQJDQGHGJHLVRODWLRQ Figure 1.6: Fabrication sequence for a BC solar cell on a FZ silicon wafer. 18 1. Introduction vapour deposited (PECVD) hydrogenated silicon nitride (a-SiN:H) and APCVD-deposited TiO2. Cotter et al. performed PC1D21 modelling and determined that efficiencies greater than 16.5% could be achieved for a single emitter and groove diffusion with a sheet resistance greater than 40 Ω/2 as long as the front-surface recombination velocity (SRVf ) was kept 22 below 20000 cm/s. This condition is relatively easy to achieve with SiO2 passivation which 23 typically results in SRVf ≈ 1000 cm/s. A necessary modification to the BC process has been the optimisation of the nickel sintering step. The electrolessly-plated nickel (Ni) film is normally sintered at 350◦C in order to form a nickel silicide (Ni2Si) ohmic contact. Additionally, the (Ni2Si) layer may act as a diffusion barrier to the subsequently deposited copper layer at normal module operating temperatures. It was found that performing a 350◦C Ni sintering step with a 45 Ω/2 groove diffusion resulted in low fill factors (< 75%) and a open-circuit voltage (Voc) that is degraded by up to 40 mV.24 It has been postulated that the degradation is caused by small-area Schottky contacts between the metal and the p-type base.24 This is attributable to either the nickel punching-through the n-doped grooves during sintering or the lack of a diffused region in some areas of the grooves. By reducing the Ni sintering temperature to 250◦C this problem has been circumvented.24 This has resulted in the fabrication of 9 cm2 solar cells with a conversion efficiency of 16.9 − 17.1%, with a sheet resistances of 39 − 50 Ω/2 on 1 Ω cm boron-doped untextured CZ wafers.24 The use of a PECVD a-SiN:H film as an AR coating 2 enabled a respectable short-circuit current density (Jsc)ofupto34.8 mA/cm to be achieved. 25 These cells did not require any SiO2 passivation layer. The disadvantage of the current SBC process is that it is not nearly as robust as the standard BC fabrication sequence, and requires further refining before it is able to withstand the rigours of a production environment. 1.5 Multicrystalline Silicon The cost of the silicon wafers alone represent the largest fraction in the cost breakdown of a solar cell. The cost of the growing the silicon ingot and cutting it into wafers contributes 46% towards the final PV module cost.26 This is not because silicon is a rare material, it is in fact the most abundant element on earth, however the purification and ingot growth processes are extremely energy intensive. Additionally, the PV industry is somewhat reliant on silicon scrap and off-cuts from the semiconductor industry for their feedstock. The refining processes to obtain the necessary purity for good electrical performance and cutting the ingots into 350 µm thick wafers costs about US$1.50 /Wp. The price is not strongly linked to economies of scale, as the PV industry in the past has been able to rely on the much larger semiconductor industry for technological advances in crystal growth. Additionally. as the semiconductor industry has moved towards larger and larger wafer sizes - 300 mm is the current standard - 1.5 Multicrystalline Silicon 19 the PV industry has been able to purchase the outdated equipment relatively cheaply. Only recently, with the strong growth of the PV industry, has dedicated equipment for PV technologies been developed in order to reduce the cost of silicon wafers. Firstly, there is an industry trend towards fabricating solar cells on thinner wafers. There are several handling issues in production to be overcome before high yields can be obtained on 150 µm-thick wafers. However, the French PV manufacturer, Photowatt, has demonstrated a high yield with 200 µm thick wafers on an automated production line.27 Secondly, different methods of casting silicon into multicrystalline silicon ingots have been developed. 1.5.1 Issues with Multicrystalline Silicon Multicrystalline silicon (mc-Si) ingots are now used by many of the world’s leading solar cell manufacturers, including BP Solar (incorporating Solarex), Kyocera, Eurosolare, and Photowatt, to name a few. Mc-Si ingots of up to 65 × 65 cm and weighing 230 kg are now being grown.27 There are several advantages to such an ingot compared to the traditional 100−150 mm round Czochralski (CZ) wafers. Firstly, the cost of mc-Si wafers is on the order of 15% less than c-Si wafers grown by the CZ process. Secondly, the overall geometrical yield from a 150 mm diameter c-Si ingot is only about 66%.27 This is because the ingot must be trimmed into pseudo-square wafers to permit a higher packing density of the solar cells once they are laminated into modules. If this is not performed then module costs will increase due to the extra amounts of glass required to fabricate a module with the same power rating. In comparison, the wafer yield from a 65 × 65 cm mc-Si ingot is about 84%. There are several disadvantages to using mc-Si wafers. Firstly, electrical properties of the material quality are slightly poorer. This is due to increased numbers of minority carriers recombining at the grain boundaries between the crystallites, before they can be collected by the junction. Secondly, due to the random orientation of the crystallites in mc-Si, the benefits achieved from the standard alkaline chemical texturing are minimal and a good AR coating is necessary to prevent large reflection losses. Thirdly, all mc-Si material is different, varying from manufacturer to manufacturer and from batch to batch. The material behaves differently under high processing temperatures (> 950◦C) and while the particular parameters used for lifetime enhancement processes, like hydrogenation and gettering, may work well for one material they may actually degrade the quality of another material. This indicates that a robust technology is required that can tolerate these kinds of variations in a production environment. For screen-printed solar cells these disadvantages represent only a small loss in performance. The mc-Si wafers are still chemically textured, even though the benefits are minimal. Screen- printed solar cells fabricated on CZ c-Si substrates maintain a slight performance advantage over mc-Si wafers, by about 0.5 − 1.0 % absolute efficiency. The device efficiency is limited, 20 1. Introduction not by the substrate, but by the screen-print process itself, and therefore can be more cost- effective on mc-Si wafers. This fact, along with the reduced cost of mc-Si wafers has seen mc-Si screen-printed solar cells capture a major share of the PV market. Additionally, large amounts of money are being invested in new facilities that will fabricate tens of megawatts more of this product. Although investment in renewable energy is applauded, an opportunity is being passed up by many companies to invest in more efficient technologies that make better use of the silicon substrate. The upper efficiency limit for a mc-Si solar cell has been demonstrated at UNSW. A laboratory scale (1×1 cm) passivated-emitter rear-locally-diffused (PERL) structure fabricated on Eurosil P48 material (Eurosolare S.p.A., Italy) resulted in a conversion efficiency of 19.8%.28 This result indicates that a significant improvement margin exists for the 12− 13% efficient commercially-produced solar cells if the right technology can be found. This work investigated the applicability of the BC process to mc-Si wafers, using low-temperature deposited titanium dioxide thin films as the replacement dielectric coating. 1.6 Why use Titanium Dioxide? There are several motivations for investigating titanium dioxide (TiO2) thin films in this work. TiO2 thin films are used currently as the PV industry standard AR coating on the vast majority of screen-printed solar cells. The important implications of this are, firstly, that the industry is familiar with the technology and will not be reluctant to adapt to fabrication processes based around TiO2 and, secondly, that the necessary deposition equipment is operating today on the factory floor. Thus, the development of a new silicon solar cell technology that included TiO2 processing steps could be readily adopted by the PV industry without the typical long lead-in time for a new technology. TiO2 exists in nature as the minerals rutile, anatase, and brookite. Titanium dioxide of the rutile form is a relatively abundant mineral,29 however anatase and brookite are extremely 30 rare in nature. TiO2 thin films are generally amorphous for deposition temperatures ≤ 350◦C, above which anatase is formed. The most stable crystalline phase, rutile, is formed at temperatures greater than about 800◦C. The brookite phase is rarely observed in deposited thin films. The functional properties of TiO2 films, powders and ceramics are strongly dependent on the phase of the material. For many applications, the size of crystals that are present also alter the behaviour of the film. Typical properties of TiO2 include: • Electrical: high electrical resistance - resistivities of 1014 Ωcm29 and dielectric constants of up to 180 are possible for rutile crystals.31 • Mechanical: high durability32 and high hardness.33 • Optical: very high refractive index - up to 2.70 − 2.71 (at a wavelength of 600nm) for rutile thin films33, 34 - and excellent transmittance in the visible region. 1.6 Why use Titanium Dioxide? 21 • Chemical: good chemical resistance and high chemical stability.35, 36 TiO2 powders and thin films are used in an extremely wide range of commercial applications and research areas, including: • Powders: as a white pigment in paint, plastic, inks, paper, and cosmetics; in washing powder, toothpaste, sunscreen, foodstuffs, pharmaceuticals, photographic plates, for creating synthetic gemstones; and as a catalyst. • Thin films: for ultra-thin capacitors and MOSFETs due to its extremely high dielec- tric constant; as humidity and oxygen sensor due to the dependence of its electrical conductance on the gases present; as an optical coating and a material for waveguides due to its high refractive index; as a protective coating and corrosion resistant barrier; and as a photoanode in solar cells due its photoelectric activity. 1.6.1 TiO2 Thin Films in Photovoltaics The use of TiO2 films has already been explored to a certain degree in the field of PV. Antireflection Coating Lord Rayleigh first observed the antireflection (AR) effect in 1887, and Bauer presented the first theoretical treatment based on interference effects in 1934.37 Since this time the theory and application of optical coatings have been well developed. The use of an AR coating for solar cells is a more recent application, beginning in the 1960’s with the advent of PV as a remote power source in space. For a good theoretical treatment of AR coatings, mainly on glass substrates, the reader is referred to Heavens,38 and some theory is also presented in Chapter 7. Many references to TiO2 thin films being experimented with as a solar cell AR coating appeared in the early 1970’s.39–44 These early experiments were aimed at increasing the efficiency of space cells, which commonly employed silicon monoxide (SiO) AR coatings. At a wavelength of 600 nm, the refractive index of silicon is 3.94, glass about 1.52, SiO about 1.9, and TiOx in the range 1.9−2.4 (it is difficult to know the exact stoichiometry of titanium dioxide thin films that are deposited via evaporation or sputtering, and therefore these films are denoted as TiOx). As discussed in Section 7.2.1, an AR coating with a refractive index of about 2.45 is optimal for achieving minimal reflection losses for a glass-encapsulated silicon solar cell. Thus, the use of TiOx films became widespread in the PV industry for providing better optical coupling of light into the silicon. Although TiOx films are more absorbing to short wavelength light, this is of little importance for SP solar cells, which already exhibit a poor blue response due to the phosphorus ”dead-layer” at the front surface.8 It should be 22 1. Introduction noted that the ethyl-vinyl-acetate (EVA) films used to encapsulate that silicon solar cells and the glass cover plates also exhibit significant absorption of short-wavelength light.45 Since the early 1970s, TiOx has been the main AR coating employed by the PV industry. Nearly all SP solar cell production lines use an APCVD- or spray-deposited TiO2 AR coat- ing.16, 19 Kern and Tracy provide an extensive review of early AR coatings for silicon solar 46 cells, focussing on TiO2. Several improvements have been made to the process, includ- ing optimisation for the firing-through of screen-printed contacts47, 48 and for deposition on 49 textured surfaces, and since 1994 TiO2 has been investigated for application in BC solar 16, 19, 20, 22, 24, 35, 50 cells. TiO2 thin films are also employed as AR coatings on glass, transmitting visible light while reflecting heat-producing IR radiation.51 Surface Passivation Surface passivation is an extremely important design consideration for high-efficiency silicon solar cells, especially at the front surface where the majority of the light is collected. The predominant recombination losses in c-Si are via defect levels within the bandgap and the large number of non-saturated Si bonds at the surfaces dominate those defect levels. In order to reduce these recombination losses and achieve high conversion efficiency the surfaces must be electronically passivated, and, in the case of solar cells, the passivation scheme should be stable under ultraviolet (UV) illumination for at least 20 years.52 Two common and well- characterized methods for silicon surface passivation are thermal oxidation at temperatures ◦ of about 1000 CtogrowSiO2 and plasma-enhanced chemical vapour deposition (PECVD) of silicon nitride. Methods of achieving surface passivation with TiO2 thin films will be discussed in detail in Chapter 5. Metallisation Mask TiO2 (anatase) films have been used as a dielectric mask for preventing electroless nickel and copper plating from occurring on the front surface of solar cells. TiO2 thin films of about 70 nm in thickness have successfully replaced the 350 nm thick SiO2 employed in the BC solar cell fabrication sequence for use as a metallisation mask.20 The necessary film properties for this application include chemical resistance and a dense, continuous film with no pinholes. ◦ TiO2 thin films deposited by spray-deposition at temperatures above 400 C have satisfied these criteria.20 Research performed at the University of Konstanz (Germany) has shown that the use of PECVD deposited a-SiN:H thin films as a metallisation mask is problematic.53 As mentioned previously, BP Solar employ an LPCVD silicon nitride coating on their BC solar cell production line. Films deposited by LPCVD are typically very dense and are well suited to acting as a metallisation mask. The primary difference between silicon nitride films deposited by PECVD and LPCVD is that the former can contain as much as 30 at. % 1.6 Why use Titanium Dioxide? 23 hydrogen. This, along with the significantly lower deposition temperature of PECVD films, results in the PECVD film having a significantly lower density.17 Ohmic Contacts 54 Thin films of TiO2 have been used to create ohmic contacts to p-type mc-Si solar cells. The TiO2 layers were deposited in between layers of aluminium screen-print paste and p- type silicon. The screen-print paste was then fired (850◦Cfor5− 30 min), which allowed the aluminium to interdiffuse with the silicon, creating a titanium silicide (TixSiy) film in the process. With the above firing conditions, the TixSiy films yielded a low contact resistivity of 1 − 13 × 10−5 Ωcm2. MIS Solar Cells Metal-insulator-semiconductor (MIS) solar cells rely on quantum mechanical tunnelling through a very thin oxide layer, less than 2 nm thick, for carrier transport.7 This is made possible by the extreme work functions of the metal. The top contact can be a thin metallic layer (< 10 nm), which is essentially transparent to light. As this layer will have a high resistivity a thicker contact grid is required to transport the current. Alternatively, very fine (5 − 10 µm) and closely spaced (50 − 100 µm) metallic fingers can be used. In this case, carriers that are generated between the fingers can be collected by a nearby grating line before recombining. The latter MIS grid structure was investigated using a 2 nm thick SiO2 layer, followed by a 55 100 nm thick TiOx layer. The TiOx was intended to act as an inversion layer, inducing a layer of minority carrier near the front surface. Although this apparently improved efficien- 56 cies, it was later found that the TiOx was acting as an accumulation layer on p-type silicon. This resulted in further research being performed with tantalum pentoxide, although TiOx would be suitable for n-type solar cells. Transparent Conducting Oxide Layers In the PV industry, transparent conducting oxide (TCO) layers are most commonly used in thin-film solar cells where low current densities are present. The most common TCO is indium-tin-oxide (ITO). However, more recently TCO layers have been employed for silicon wafer based heterojunction solar cells from Sanyo.57 These cells have a p-andn-type amor- phous silicon films deposited onto the front and rear of the n-type silicon substrate. A TCO layer is employed to reduce the front contact area required. There is always an electrical vs. optical trade-off with TCO layers, and even though a relatively high short circuit cur- 2 rent density is achieved (Jsc =36.7 mA/cm ) the spectral response curve exhibits significant 24 1. Introduction absorption at wavelengths less than 700 nm.57 The TCO layers typically have a sheet resistance of about 10 − 50 Ω/2. Niobium-doped TiO2 films with this sheet resistance and a refractive index of 2.2 − 2.5 have been deposited as TCO layers.58 However, a film thickness of greater than 1.5 µm was required to achieve the 50 Ω/2 sheet resistance. With thinner films a logarithmic dependence of the resistivity on the film thickness was observed. Therefore, these films are too thick to act as an effective AR coating. As discussed in Sections 2.4.2 and 2.4.4, oxygen deficient TiO2−x thin films can have resistivities down to 102 − 10−3 Ω/2, indicating their potential as a TCO. TiO2/c-Si Heterojunction Solar Cells Investigations have been performed on the addition of indium (In) to TiO2 to form p-TiO2/n- 59 60 Si heterojunction solar cells. The TiO2 films were deposited by a ”spray-CVD” process. The In-doped films lead to an efficiency increase of about 40% over pure TiO2 films, resulting in an efficiency of 14.1% under 100 mW/cm2 (AM1) illumination. The open circuit voltage approached 650 mV and a high fill factor of 0.82 was achieved. Gordon deposited doped TiO2 film as an electrode underneath a fluorine-doped tin oxide 61 (F:SnO2) layer to form a F:SnO2/TiO2/p-Si heterojunction solar cell. The patent deals mainly with niobium-doped TiO2, however other possible dopants that are also discussed include tantalum, tungsten, phosphorus, arsenic, antimony and vanadium. The function of the TiO2 layer is to, firstly, overcome the “interfacial resistance” observed in SnO2/p-Si heterojunction cells, and, secondly, to act as an intermediate AR coating between the SnO2 61 62 and Si. The SnO2 has a refractive index of about 1.85 at 600 nm. It is noted that the TiO2 does not exhibit sufficiently low resistance to act as a TCO. In both solar cells described here the TiO2 layer was about 100 nm thick. The dye-sensitized solar cells described below are also heterojunction solar cells, however as these involve a liquid p-type electrolyte they will discussed in a separate section. Dye-sensitized TiO2 Solar Cells In 1991 a very different solar cell concept was presented based on dye-sensitized nanocrys- 63 talline TiO2 thin films and an iodine/iodide electrolyte. The sample structure is shown in Figure 1.7. The TiO2 is n-type while the dye is p-type. The cell works by conversion of photons to electrons by the dye and the subsequent transfer of electrons to the glass electrode by the TiO2 layer. The device had an efficiency of 7.1% under full sunlight, which increased to 12% under diffuse lighting. This solar cell has a large potential market due to drasti- cally reduced fabrication costs and conversion efficiencies that are comparable to amorphous silicon solar cells.64 The dye, in this case ruthenium based, is used to photosensitize the 1.7 Thesis Overview and Goal 25 TiO2 film. A highly porous TiO2 layer with a large surface area to volume ratio is used to increase the amount of adsorbed dye. This increases the absorption properties of the device in the visible spectral region. Spray-deposition techniques have been used for depositing 65 the nanocrystalline TiO2 films and the highly porous, CVD-deposited films presented in Chapter 7 could potentially be used for this type of solar cell. Figure 1.7: Sample structure of a photoelectrochemical cell using nanocrystalline TiO2. The arrows indicate the direction of light (from Li et al.64). 1.7 Thesis Overview and Goal The aim of many silicon-wafer based PV research groups worldwide is to develop a new, commercially-viable, fabrication technology suitable for mc-Si wafers, in order to bridge the gap between the “five-year plan” for thin film dominance of the PV marketplace and the stock-standard product of the last 20 years, screen-printed solar cells. The objective of this thesis project is to develop, understand and evaluate novel applica- tionsofTiO2 thin films to silicon solar cells. TiO2 is identified as an unique material with significant potential owing to its excellent optical and electrical properties. TiO2 thin films also appeared to be an attractive option due to the possibility of depositing them at a low- cost. While amorphous TiO2 thin films have a long history of being used as an AR coating on screen-printed solar cells, very few examples of the application of polycrystalline TiO2 (especially anatase) thin films to photovoltaics can be found in the literature. Anatase TiO2 thin films exhibit a high refractive index and low absorption coefficient. This, along with their insulating properties and excellent chemical resistance, suggested 26 1. Introduction that anatase thin films could be used as a direct replacement for the thermally-grown SiO2 layer in the standard BC technology. The replacement of the SiO2 layer with TiO2 promised much lower thermal budgets, simplified fabrication sequences and reduced processing costs. The lack of literature regarding the behaviour and stability of TiO2 thin films under high- temperatures and different gas ambients meant that a significant amount of time was spent increasing this knowledge base. In order for TiO2 to successfully replace SiO2 in the standard BC process, several key pa- rameters have to be explored. Firstly, it needs to be demonstrated that TiO2 thin films do not reduce the minority carrier lifetime of silicon wafers when processed at temperatures up ◦ to 1000 C. Secondly, TiO2 is known to be a poor option for passivating the surfaces of a silicon wafer, so therefore a successful method for enhancing the surface passivation of TiO2 coated silicon wafers needs to be developed in order to achieve high efficiencies. Fourthly, one crucial role of the SiO2 layer in the BC process is to act as a phosphorus diffusion barrier. The performance of a TiO2 thin film in this role needs to be evaluated. Fifthly, an additional reduction of thermal budget is envisaged by combining the emitter diffusion and AR coating steps. In this manner, a TiO2 film doped with phosphorus atoms would be deposited onto the p-type wafer and, during a subsequent firing process, the phosphorus atoms diffuse out of the TiO2 and form an n-type emitter. Finally, as well as acting as a chemically resistant layer and an electroless metal plating mask, the optical performance of TiO2 AR coatings need to be optimised. The primary goal is to develop a 16−17% efficient BC solar cell on planar mc-Si wafers. The application of the BC technology to textured c-Si wafers has been demonstrated by industry, however the current BC technology cannot be economically applied to mc-Si wafers due to the high-processing costs. Therefore, a simplified process, centred around using TiO2 as the thin dielectric film is sought. If TiO2 films prove to be successful then one expected outcome would be the evolution of a new solar cell technology that is readily applicable to today’s PV industry. Following this introduction, Chapter 2 presents an extensive literature review, necessary to understand the physical, optical, electrical and chemical properties of TiO2. A review of thin film deposition techniques and a description of two deposition systems designed and constructed by the author are described in detail in Chapter 3. Due to the novel ways that TiO2 thin films were being implemented into solar cell processing sequences, extensive film characterisation was necessary to determine the variation of film properties with process conditions (Chapter 4). A novel method of overcoming the limited surface passivation achievable with TiO2 coated silicon surfaces is discussed in Chapter 5. Issues such as film stability at high temperatures and contamination are also addressed in this chapter. Novel PV applications of TiO2, such as its ability to act as a phosphorus diffusion barrier and phosphorus dopant source, are investigated in Chapter 6. A high-performance, commercially viable, double-layer antireflection (DLAR) coating, based on two or more TiO2 1.7 Thesis Overview and Goal 27 films with differing refractive indices, is demonstrated in Chapter 7. Modelling results of the performance of this DLAR coating on a planar BC solar cell will also be presented. Finally, Chapter 8 summarises the work, and presents opportunities for further research in the area. 28 1. Introduction Chapter 2 Common Properties of TiO2 Thin Films The physical, optical, electrical and chemical properties of titanium dioxide (TiO2) depend greatly on the amorphous or crystalline phase of the material. TiO2 is a complex material with three crystalline phases, two of which are commonly observed in thin films - anatase and rutile. Anatase is commonly observed at film deposition temperatures of 350 − 700◦C, while higher temperatures promote the growth of rutile. Deposition temperatures lower than ◦ 300 C generally result in the formation of amorphous TiO2. Amorphous TiO2 is a highest bandgap material (about 3.5 eV), and exhibits a low refractive index (about 1.9 − 2.0 for 600 nm wavelength light) and extinction coefficient. The chemical resistance of amorphous TiO2 films is poor in many acidic and basic solutions. Polycrys- talline anatase thin films, with an optical bandgap of about 3.2 eV, exhibit a much higher re- fractive index (as high as 2.532 at 600 nm for single crystal material) and a slightly increased absorption coefficient. With the crystalline structure comes increased chemical resistance, and dense anatase films are insoluble in many acids and bases. Rutile thin films (3.05 eV bandgap) have extremely high refractive indices (up to 2.70 at 600 nm for single crystal rutile) and below the bandgap absorption is still low. The chemical resistance of rutile is excellent, and after annealing at temperatures above 1000◦C it is insoluble in nearly all acids and bases. 2.1 Introduction The aim of this work was to evaluate the performance of titanium dioxide (TiO2) as a drop-in replacement for the thick, thermally grown silicon dioxide (SiO2) layer in the buried-contact solar cell. The use of TiO2 immediately obviates one of the high-temperature processing steps, required to grow the SiO2 layer. TiO2 was chosen due to the ability of depositing films at low temperatures and at atmospheric pressure; the non-toxicity of the liquid precursor; 29 30 2. Common Properties of TiO2 Thin Films the familiarity of solar cell manufacturers with this film; and, their existing ownership of the necessary deposition equipment. It was anticipated that these factors would facilitate an easy transfer of a successful laboratory device into a commercial environment. However, TiO2 is a relatively complex material, and three crystalline phases as well as the amorphous form of TiO2 exist. Since each of the these materials has different optical, electrical, and chemical properties it was necessary to perform an extensive literature review in order to predict how the TiO2 films would behave in different processing conditions. This chapter will describe the properties of the crystalline phases most commonly observed in thin films, that of rutile and anatase, and amorphous TiO2. The third crystalline phase, brookite, is a less stable and common form of TiO2 is rarely observed in deposited thin films and will not be discussed here. There are many different parameters that affect the phase of a deposited TiO2 thin film. Some of these parameters are deposition method, deposition temperature, annealing temperature, deposition rate, deposition pressure, precursor type, reaction atmosphere, impurities present, and substrate type. The resulting phase or mixture of phases, plays a large role in determining the physical, optical, chemical, and electrical properties of the film. This work relies greatly on the excellent optical properties of TiO2 thin films, as well as its chemical resistance and insulating properties. A summary of the physical, optical, electrical and chemical properties reported in the literature, with an emphasis on those relevant to solar cell fabrication, is presented in the following sections of this chapter. 2.2 Physical Properties 2.2.1 The Amorphous − Anatase − Rutile Phase Transformations ◦ 66, 67 Amorphous TiO2 thin films can be deposited at temperatures as low as 100 − 150 C. Amorphous TiO2 does not have a strict crystallographic structure, often incorporates voids within the material, and has a relatively low density. For TiO2 thin films formed by chemical reaction, the lowest temperature crystalline phase of TiO2 that can be obtained is anatase. To obtain polycrystalline anatase, the film can be either deposited as amorphous TiO2 and then crystallised by annealing at a higher temperature, or deposited as polycrystalline ma- terial directly. Nearly all published results indicate that the transition from an amorphous to anatase film occurs at about 300 − 365◦C, regardless of whether this is the deposition or annealing temperature. Rutile films are initially observed on silicon substrates at deposition temperatures above 700◦C, and more typically from 900 − 1100◦C. It should be noted that anatase is a metastable phase of TiO2, and the conversion to rutile involves a collapse of the anatase structure, which is irreversible.68, 69 Figure 2.1 indicates the structure of an anatase 2.2 Physical Properties 31 and rutile crystal. Although rutile and anatase are both of tetragonal crystallographic struc- ture, rutile is more densely packed and thus possesses a greater density. Figure 2.1: Models showing the tetragonal structure of both anatase and rutile, and the denser structure of the rutile phase of the latter (adapted from Du Pont, Inc.70). The TiO2 thin films deposited in this work are formed by chemical reaction, using chemical vapour deposition (CVD), and spray pyrolysis and hydrolysis systems. In this scenario, the substrate temperature is the primary means of controlling the deposited phase of the mate- rial. In contrast, physical vapour deposition (PVD) systems, such as evaporation, sputtering, and ion-beam deposition, the resulting phase and film structure is determined primarily by the kinetic energy of the impinging atoms. Therefore, the progression through the amor- phous, anatase, and rutile phases may not necessarily be expected. This is confirmed by the occurrence of rutile films at low deposition temperatures (< 450◦C) by carefully optimised deposition methods,34, 71 ion-assisted deposition,72 and reactive evaporation.73 However, the bulk of the discussion here pertains to TiO2 films formed by a chemical reaction, and where the substrate temperature dominates film growth characteristics. Several researchers73–75 observed that the processing temperatures required to convert an anatase film into a rutile one are much higher than temperature required to deposit a rutile film directly. Agreement with this observation can be found in the literature, where the rutile phase is only present at low temperatures when it is deposited directly at that tem- perature.34, 71, 73, 74, 76–78 The formation of rutile at lower temperatures is facilitated by the kinetic energy possessed by the TiO2 molecules during the deposition process, enabling the lowest energy state to be reached on the substrate. In contrast to the above observation, Fitzgibbons had previously claimed that the variation in physical and chemical properties of the films is determined solely by the maximum processing temperature, whether this be the deposition temperature or a subsequent annealing temperature.67 Amores et al. have published an excellent diagram indicating how the high-temperature sintering process and transformation of anatase to rutile crystals proceeds, shown here in Figure 2.2.79 The pro- 32 2. Common Properties of TiO2 Thin Films posed mechanism for the sintering and transformation of anatase into rutile involves several steps. Initially, the smallest particles (a) coalesce, forming bigger particles (b). The frac- tion of particles that are already large have been shown not to undergo sintering. The heat evolved from the exothermic sintering process causes the local nucleation of the rutile phase (c). Finally, as the conversion to rutile is also an exothermic process, this results in the transformation of the whole particle to rutile (d). Figure 2.2: Proposed mechanism for the sintering and transformation of anatase into rutile. The smallest particles (a) coalesce, forming bigger particles (b). The fraction of particles that are already large have been shown not to undergo sintering. Heat evolved from the exothermic sin- tering process causes the local nucleation of rutile (c). The conversion to rutile is also an exothermic process, leading to the transformation of the whole particle to rutile (d) (adapted from Amores et al.79). 2.2.2 The Effect of Impurities on the Anatase − Rutile Phase Transformation Many researchers have observed that the inclusion of a certain amount of impurities into TiO2 can drastically alter the physical properties of the film. It has been shown that silicon and phosphorus inhibit the transformation from anatase to rutile, with 100% anatase phase being retained at temperatures as high as 870◦C for up to 3 hr for thin films80 and 1500 K for bulk samples.81 The retardation of the anatase-rutile transformation can be achieved with 3− 2− 3+ 80 68 68, 81 79 the following impurities: PO4 ,SO4 and Al ; AlPO4; SiO2; Co3O4 and MoO3; 82 + 83 84, 85 85 81, 86 Ce and Nb; K ; WO3; Na2O; and P2O5. Conversely, it is well known that other impurities enhance the formation of rutile at lower temperatures. These impurities 79, 85 79, 87 85 include CuO2; V2O5; and NiO, CoO, MnO2,Fe2O3. Most researchers agree that oxygen vacancies are responsible for the overall transformation mechanism.68, 69, 88 Thus, the oxides and fluorides (such as Li+1,Co+2 and Mn+4)thatassist 2.2 Physical Properties 33 the transformation can substitute for Ti+4 in the anatase lattice, resulting in the creation of −3 −2 oxygen vacancies. On the other hand, the inhibiting effect of other impurities (PO4 ,SO4 , Nb2O5) has been explained by the reduction of oxygen vacancies due to the substitution of Nb+5 and S+6 into the anatase lattice. Oxygen vacancies are also known to be created in hydrogen ambients, thereby favouring the transformation to rutile.69, 85, 88 It should be noted that during the growth of TiO2 thin films, contamination from the various chemical precursors can result. Titanium alkoxides are common TiO2 precursors, with the most frequently used being titanium isopropoxide (also called tetraisopropyl titanate, TPT). The residue of the organic binders results in carbon contamination of typically a few at.%, but as high as 13 at.%, being observed.64, 89–99 It is likely that carbon incorporation could be higher at low growth temperatures, as when higher temperatures were used the carbonate species decomposed, resulting in the removal of hydrocarbon fragments.64, 94, 99, 100 Titanium tetrachloride (TiCl4) is another common TiO2 precursor, and this results in chlorine con- tamination of the deposited film.91, 93, 101 2.2.3 Substrate Type The effect of substrate type upon the deposited TiO2 film will be discussed only briefly, as nearly all experiments performed in this work employ silicon substrates. Several researchers have studied the properties of TiO2 films deposited onto various substrate types, including silicon, silica, quartz, alumina, titanium, copper, gallium arsenide, stainless steel, as well as several types of glass. The different substrate types influence the physical properties of the deposited film, including the phase, texture or surface roughness. Optical and chemical properties also change, however these are primarily dependent on the phase and density of the polycrystalline TiO2 film. Possible reasons for the dependence of the TiO2 phase and properties on the substrate include the substrate’s surface conditions affecting the orientation and packing density of the molecules, and, with glass, the diffusion of metal ions into the film.102 Battiston et al. observed that anatase was the only phase present with films deposited onto titanium and stainless steel substrates and annealed at 750◦C.103 In the same work, an anatase-rutile mixed-phase was observed on barium fluosilicate glass substrates at 1100◦C and alumina substrates at 900◦C. At 1100◦C a single rutile phase was detected on the alu- mina. This is in agreement with other results, where either single-phase rutile or epitaxially- grown rutile films have been achieved upon alumina substrates.94, 104–107 Film depositions on glass substrates are typically limited to anatase, or anatase-rutile mixed phases due to the low melting point (typically ≤ 650◦C) of most glasses. However, results of depositions on quartz indicate a preference for the formation of anatase, even at temper- atures greater than 1000◦C.67, 108, 109 This is also true for glasses containing sodium, where 34 2. Common Properties of TiO2 Thin Films it is possible for the Na+ ion to out-diffuse from the substrate and retard the formation of rutile.85, 102, 107, 110 In contrast, aluminosilicate glasses such as Corning 7059 are known 107, 110 to favour the formation of the rutile phase due to Al2O3 impurity. Out-diffusion of substrate elements has also been observed by Yuan and Tsujikawa, where up to 30% copper ◦ 111 concentration was found in a TiO2 film deposited onto copper sheet and fired at 800 C. 2.2.4 Film Defects It important that the deposited films are relatively uniform in thickness, and do not exhibit pinholes. Several works have reported pinholes or similar defects in TiO2 films. Using scanning electron microscopy (SEM), Fitzgibbons observed an occasional pinhole at 20000× magnification in films deposited by CVD at 150◦C, however no pinholes were observed after annealing (300 − 1000◦C).67 Additionally, it was found that tensile stress caused the films to crack when the film thickness reached 400 − 500 nm. Nishide and Mizukami experimented with spin-coating of TiO2 films and found that some films exhibited 1 µm diameter pores when fired at temperatures below 500◦C, however these defects disappeared when the firing temperature was increased to 550◦C.112 This is in accordance with other results indicating 113 that TiO2 films deposited at lower temperatures are generally more heavily defected. The existence of craters was noticed by Szlufcik et al. when screen-printing a titanium organo-metallic based ink, however the problem was alleviated with the addition on butanol 114 to the ink. Kern and Tracy commented that the existence of micro-pinholes and TiO2 particulates observed in pneumatically sprayed TiO2 antireflection (AR) coating did not im- pair solar cell performance.46 More recently, Golego et al. observed that some spray droplets can react on their way to the substrate, form a particulate and then become incorporated into the film.115 Spray depositions performed at very low temperatures (90◦C) resulted in the liquid layer cracking upon drying. Again, as the deposition temperature was increased slightly (to 120◦C) these defects disappeared. Kurtz and Gordon noted that by maintaining a large temperature differential between the substrate and the deposition equipment, avoids 116 TiO2 particulates from forming on the substrate. 2.2.5 Film Density While both rutile and anatase possess a tetragonal crystallographic structure, rutile is more densely packed and thus possesses a greater density (4.26 g/cm3) than anatase (ρ = 3 70 3.84 g/cm ). For TiO2 thin films the highest observed density published to date is the 3 33, 117 range 4.09 − 4.10 g/cm for a rutile film. Amorphous TiO2 films exhibit a wide range of densities, from 2.4g/cm3 for porous films118 to more typical values of 3.2 − 3.65 g/cm3,119 while films deposited with a high kinetic energy have achieved densities in the range 3 33 3.6 − 3.8g/cm . It has been noted that the TiO2 films with a lower density can favour 2.2 Physical Properties 35 impurity diffusion.120 It is widely accepted that there is a linear relationship between density and refractive index 33, 67, 121–125 of a TiO2 thin film. The linear variation of the refractive index (measured at 550 nm) with density is shown in Figure 2.3 for films deposited by five different techniques.33 The equation of the line in Figure 2.3 is nf =0.42751 ρ +0.91933 . (2.1) This can be more usefully expressed for this work as n − 0.91933 ρ = f ,(2.2) 0.42751 3 where ρ and nf are the TiO2 film density (in g/cm ) and refractive index, respectively. Figure 2.3: Experimental data from several researchers indicating that a linear correlation between TiO2 film density and refractive index is observed over for a wide range of values. Previously published data from Ottermann and Bange,121 Fitzgibbons et al.,67 Bendavid et al.,33 Hass34 and Ribarsky 126 was used. (adapted from Bendavid et al.33). Additionally, the porosity of the film can be determined using the following relation127 n2 − 1 Porosity = 1 − f , (2.3) n2 − b 1 where nb is the refractive index of the bulk single crystal material. It should be emphasised that this value is an approximation due to the fact that, firstly, both anatase and rutile crystals exhibit strong birefringence and, secondly, that mixed anatase/rutile phases can exist. The values of mean refractive indices used in this thesis are 2.70 for rutile126 and 36 2. Common Properties of TiO2 Thin Films 34 2.532 for anatase, both measured at λ = 600 nm. Several researchers have noted that TiO2 films (anatase) derived from sol-gels tend to be highly porous in nature, sometimes up to 49%.128, 129 When TiO2 films are annealed at a temperature higher than their deposition temperature, particle sintering and crystallisation contribute to an increase in film density and refractive index, and, accordingly, a reduction in film thickness (see Wong et al. for example130). Generally, the type of film structure can be estimated before the deposition takes place, based on the Movchan-Demchishin structure zone model (SZM).72 Guenther expanded the SZM (see Figure 2.4) to include high density vitreous phases that can be achieved using PVD techniques.72 However, for films deposited via CVD or spray deposition, the structure of the film is predicted by dividing the substrate temperature (Tsub) during TiO2 deposition ◦ 131 by its melting point, which for TiO2 is Tmelt = 1832 C. columnar dense vitreous porous dense polycryst. amorphous 0.3 0.4 1.0 Normalised substrate temperature (Tsub /Tmelt) Figure 2.4: Structure zone model, expanded to include vitreous phases 72 observed with TiO2 (adapted from Guenther ). 2.2.6 Non-Stoichiometric TiO2−x Thin Films Titanium dioxide thin films exhibit a bluish, purplish or greyish hue once they are reduced 29, 73, 88, 132–136 (become poor in oxygen) to TiO2−x or TiyOx. Oxygen deficiency can occur as a result of the deposition conditions. This is a common problem with evaporated thin films, where the choice of source material (e.g., Ti, TiO, TiO2 or Ti3O5) and oxygen partial pressure of the system are critical. The formation of TiOx can also result from annealing TiO2 in a vacuum or hydrogen (reducing) ambient. In this case, oxygen is lost from the lattice to the furnace ambient. The optical and electrical properties of TiO2−x thin films will be presented in Sections 2.3.4 and 2.4.2, respectively. Non-stoichiometric thin films observed in this thesis will be discussed in Section 5.2.2. 2.3 Optical Properties 37 2.3 Optical Properties 2.3.1 Refractive Index, Extinction Coefficient and Scattering The refractive index n of a material is primarily determined by the polarizability of the valence electrons.132, 137 Increased shielding of the positive nucleus results from elements with higher atomic weight, and this increases the polarizability of the electrons and consequently the refractive index. Silicon and germanium are good examples with refractive indices in the infrared spectrum of 3.4 and 4.0. respectively. In compounds, the type of bonding also affects the refractive index, with highly covalently bonded compounds yielding higher refractive indices than predominantly ionically bonded compounds.137 The extinction coefficient k plays an attenuating role in the material. When the attenuation is solely due to absorption, it is termed the absorption coefficient 4πk α = . (2.4) λ Figure 2.5 displays a theoretical transmission spectrum for an optical thin film. It can be seen that a region of high transmittance (region II) is located between the region of short- wavelength fundamental absorption (region I) and the long wavelength limit (region III). The region of fundamental absorption is dependent on the electronic structure of the material, while absorption in the long-wavelength region is due to lattice vibrations and/or free carrier absorption. The transmission of region II is strongly linked to the stoichiometry and purity of the thin film. The extinction coefficient of a film can also be increased by the scattering of light by surface and volume imperfections, such as surface roughness, porous microstructure, and density fluctuations due to crystallinity, and is thus dependent on the deposition method.132, 137 The term optical loss L is defined as being L = A + S (2.5) with 1=R + T + L, (2.6) where A is absorptance, S is the scattered component, R is reflectance and T is transmit- tance. Wang and Chao noted that an increase in the extinction coefficient of amorphous 138 TiO2 thin films deposited by sputtering upon annealing. The drastic increase observed in k was attributed, firstly, to the formation of the anatase phase, and, secondly, to an increase in the surface roughness due to the polycrystalline nature of the annealed film. The scatter- ing loss of a rough surface is related to the root-mean-squared (RMS) surface roughness σ 38 2. Common Properties of TiO2 Thin Films band-band transparent lattice vibration absorption region absorption drop caused by free carrier impurity absorption absorption I II III Figure 2.5: Schematic of a theoretical transmission curve for an optical thin film. Three regions of absorption are shown, and each region has a different mechanism for optical absorption, as explained in the text (from Pulker 137). by139 4πσ2 TIS = , (2.7) λ2 where TIS is the total integrated scattering. Wang and Chao proposed that the extinction coefficient k could be divided into two components138 k = αa + αs , (2.8) where αa is the absorption coefficient (as defined in Equation 2.4) and αs is termed a scatter- ing coefficient. Figure 2.6 shows how the two contributions to the extinction coefficient, αa 138 and αs, vary as a function of annealing temperature in an oxygen ambient. It is observed that αa decreases with temperature, due to the reduction of oxygen vacancies in the film, while αs increases as surface roughness of the film becomes greater with the formation poly- crystalline anatase. If this evaluation was performed, say, at λ = 400 nm (near the bandgap) ◦ αa would increase at about 350 C due to lower bandgap of anatase (about 3.4 eV). It should be emphasised that Figure 2.6 was derived empirically for sputtered TiO2 films, and that wrinkling, cracking, or peeling of the film at temperatures above 300◦C may not occur for alternative deposition methods. For stoichiometric films deposited by spray deposition or CVD, αa would not decrease at low annealing temperatures as there are no oxygen vacancies. ◦ At a temperature of about 800 C, αa would increase again due to the increase rutile fraction in the film (the bandgap of rutile is about 0.2 eV lower than that of anatase). The scattering coefficient αs would most likely level off with the increased rutile fraction, as the grain size 2.3 Optical Properties 39 for anatase thin films is typically an order of magnitude greater than that of rutile films (e.g., 20 nm versus 200 nm, respectively).119 Figure 2.6: Qualitative illustration indicating the behaviour of the ex- tinction coefficient k, and its absorption component αa and its scattering component αs with increased annealing temperature in an oxygen ambient (from Wang and Chao138). 2.3.2 TiO2 Thin Film vs. Single Crystal The refractive index of a TiO2 thin film is typically much less than that of an anatase or rutile crystal, while the extinction coefficient of the deposited material will generally be greater than that of the bulk material. Figure 2.7 displays the refractive index for anatase single crystals. Due to the limited studies performed on anatase single crystals the data used for the anatase curves is shown in Table 2.1. Data for rutile is shown in Figure 2.7 for comparison.140 As both anatase and rutile are birefringent crystals it is necessary to calculate the mean refractive index nmean for a randomly oriented polycrystalline thin film. This is done using Equation 2.9 below 2n⊥ + n n = ,(2.9) mean 3 where n⊥ and n are for oscillations perpendicular and parallel to the optical axis, respec- tively. The mean refractive indices of 2.70 for rutile126 and 2.532 for anatase34 (both at 40 2. Common Properties of TiO2 Thin Films λ = 600 nm) were selected as a reference bulk value to represent dense polycrystalline ma- terial. Figure 2.7: Published values for the refractive index of single crystal anatase, taken from Meyer and Pietsch,141 Hass,34 Fitzgibbons et al.,67 Kingery et al.,142 Washburn,143 and Kim.144 The dispersive curve for single crystal rutile from Kim is also given.144 Very little anatase and rutile single crystal absorption data has been published in the visible as these materials are essentially transparent. The optical bandgap of anatase and rutile 145, 146 145 is about 3.2eV and 3.05 eV, respectively. The optical bandgap of amorphous TiO2 is commonly reported as being around 3.5 eV.147 Hence, optical absorption will increase with the successive transformations from amorphous to anatase to rutile material. It can be seen in Figure 2.8 that anatase has an absorption edge with a lower steepness, which is attributed to the presence of excitons and more imperfections and disorder in anatase crystals.120 Figure 2.9 shows the exponential dependence of the absorption coefficient at 10 K for different polarisations when illuminated with UV light.120 2.3.3 Variation with Deposition and Annealing Temperature As previously discussed, many researchers have observed a direct linear relationship between 33, 67, 121–125 the refractive index and the density of the film. With TiO2 it is possible to deposit increasingly dense films, approaching the density of the bulk material, due to the mixture of amorphous, anatase, and rutile phases that can exist. Furthermore, while high refractive indices can be achieved for an amorphous film, a different set of deposition conditions can yield an anatase or rutile film with a higher refractive index.33 Thus, the optical properties 2.3 Optical Properties 41 Table 2.1: Refractive index data for anatase single crystal. The data for the anatase and rutile curves in Figure 2.7 are not presented in the table and can be found in Kim.144 The asterisked n values (∗) are for rutile. Wavelength n n⊥ nmean Reference λ (nm) 435.83 2.7688 2.6576 2.732 Meyer and Pietsch141 546.07 2.5948 2.5161 2.569 589.30 2.5612 2.488 2.537 690.70 2.5097 2.4447 2.488 405 2.8760 2.7395 2.785 Washburn143 436 2.7688 2.6576 2.69467 492 2.6586 2.5691 2.59893 546 2.5955 2.5169 2.5431 578 2.5694 2.4950 2.5198 589 2.534 2.488 2.50333 623 2.5407 2.4709 2.49417 691 2.5106 2.4456 2.46727 706 2.5052 2.4409 2.46233 450 − − 2.703 Hass34 500 − − 2.615 550 − − 2.565 600 − − 2.532 700 − − 2.485 550 − − 2.57 Fitzgibbons et al.67 600 − − 2.52 Kingery et al.142 600 2.60∗ 2.90∗ 2.70∗ Ribarsky126 of a TiO2 film can be effectively tuned by adjusting the deposition temperature, bearing in mind that, in general, the extinction coefficient of a polycrystalline TiO2 thin film will be higher than that of an amorphous thin film. Several works that have included extensive optical characterisation of TiO2 films will be re- viewed here. Hovel has published two excellent graphs demonstrating the trend of increasing refractive index with increasing deposition temperature, as shown below in Figures 2.10(a) 148 and (b). The TiO2 films were deposited by thermal spraying and the refractive index in Figure 2.10(a) was measured using ellipsometry at 633 nm. Absorption values are reported in a variety of ways, including calculations of α or the TiO2 bandgap energy, and plots of optical transmittance or absorptance versus wavelength. As previously discussed, contributions towards optical losses in film arise from the fundamental 42 2. Common Properties of TiO2 Thin Films Figure 2.8: Fundamental absorption edge of anatase and rutile single crystals, measured at a temperature of 10 K (adapted from Tang et al.120). Figure 2.9: The exponential dependence of the absorption coefficient of single crystal anatase, measured at 10 K with light polarized in Ec and E⊥c directions (adapted from Tang et al.120). absorption edge of the TiO2 film as well as surface and volume imperfections. Figure 2.11 indicates the general trend of increasing absorption coefficient with annealing temperature. This is in agreement with the formation of the anatase phase, which possesses a lower bandgap than amorphous TiO2. DeLoach et al. found that the absorption edge for sputtered TiO2 films decreased from 3.41 eV for films with a small rutile fraction (< 0.17) to 3.22 eV for films with a large rutile fraction (> 0.7).149 2.3 Optical Properties 43 Figure 2.10: (a)Variation of TiO2 refractive index with annealing temper- ature, and (b) corresponding dispersive refractive index relations (adapted from Hovel 148). Figure 2.11: Absorption coefficients of TiO2 films deposited at various temperatures (from Hovel 148). Kamataki et al. performed optical characterisation of TiO2 thin films deposited by atmo- spheric pressure chemical vapour deposition (APCVD) at 300◦C in the wavelength range 250 − 850 nm.150 Measurements were also performed on samples that were annealed for 1 hr ◦ ◦ ◦ in both oxygen (O2) and nitrogen (N2) ambients at temperatures of 500 C, 700 C and 900 C. Figure 2.12(a) shows a maximum in n at about 310 nm, while Figure 2.12(b) indicates that there is negligible absorption in the films at wavelengths greater than 350 nm. Both n and k exhibit a trend of increasing with higher annealing temperatures. The optical constants of electron-beam (e-beam) evaporated TiO2 thin films were measured 44 2. Common Properties of TiO2 Thin Films Figure 2.12: (a) Refractive index and (b) extinction coefficient of APCVD-deposited and annealed TiO2 thin films (adapted from Kamataki et al.150). by Kim using spectroscopic ellipsometry (SE) in the spectral region 1.5 − 5.5 eV.140, 144 A double oscillator model for amorphous materials151 is used to fit n and k values. Kim also successfully modelled the film as being polycrystalline anatase with a 16% void content.140 Figure 2.13 displays n and k for the TiO2 evaporated film along with the “void-free” equiva- lent film. A three-layer model with varying void fraction was used to model the 96.4 nm-thick (total) film. The surface roughness was successfully modelled by 13.1 nm-thick top layer with a 34% void incorporation. ◦ ◦ Mardare and Hones deposited TiO2 thin films at 100 C (sample B) and 250 C (sample A) onto glass substrates using RF sputtering and determined n and k using SE.82 Figure 2.14(a) indicates that high refractive indices were achieved even at these low deposition temperatures, while Figure 2.14(b) demonstrates that the extinction coefficient remained below 0.02 for the sample deposited at 250◦C for wavelengths > 400 nm. The dispersive relations for Samples C and D in Figures 2.14(a) and (b) are for doped TiO2 thin films and will not be discussed here. Szlufcik et al. investigated screen-printed TiO2 AR coatings and found that the refractive index increased linearly with annealing temperature up to a maximum of 2.30 at 800◦C.114 2.3 Optical Properties 45 Figure 2.13: Dispersive n and k values calculated from SE data for an e-beam evaporated TiO2 film (solid lines)and the equivalent “void-free” film (open circles). The refractive indices of polycrystalline anatase and rutile are also shown for comparison (from Kim144). Figure 2.14: (a) Refractive index and (b) extinction coefficient of RF 82 sputtered TiO2 thin films. (from Mardare and Hones ). 2.3.4 Variation with Deposition and Annealing Ambient Photovoltaic researchers have noted that the optical absorption of non-stoichiometric TiO2−x thin films increased in the short wavelength region, precisely where higher efficiency solar cells 46 2. Common Properties of TiO2 Thin Films 43 exhibited their improved response. As mentioned previously, non-stoichiometric TiO2−x films can be achieved under a variety of deposition and annealing conditions. The effect of 152 depositing TiO2 thin films in an oxygen poor environment is demonstrated in Figure 2.15. Films were evaporated at two different base pressures, 5 × 10−5 Torr and 1 × 10−4 Torr, and at one oxygen partial pressure. As shown in Figure 2.15, films evaporated without the presence of oxygen exhibited a much lower transmittance over the visible spectrum than the film deposited with an oxygen partial pressure of 5 × 10−5 Torr. The exact stoichiometry of the films is unknown, however a direct relationship between base pressure and transmittance can be observed. Figure 2.15: Variation of transmittance of evaporated TiO2 films with base an oxygen partial pressure (from Jiao and Anderson152). Zakrzewska et al. correlated the optical and physical properties of sputtered TiO2−x thin films deposited with different stoichiometries.153 The measure of stoichiometry was deter- mined as the fraction I/I0, which is the ratio of the titanium plasma line intensity during deposition to the 100% metallic titanium plasma line intensity. Higher I/I0 values indicate a greater departure from stoichiometry. The main graph in Figure 2.16 shows the increase in the film absorption coefficient with increasing I/I0. The inset graph in Figure 2.16 shows that the refractive index, measured at λ = 800 nm, increases linearly with decreasing oxidation state of the TiO2−x film. For TiO2 thin films deposited by ion-beam sputtering, it was found that the refractive index exhibited a maximum (2.52 at λ = 633 nm) with an oxygen concentration during 154 the deposition of [O2]=30%. The level of absorption in the film was found to decrease dramatically for [O2]< 30%, especially at the shorter wavelengths of 500 nm and 633 nm. A difference in the optical properties of APCVD-deposited TiO2 films annealed in O2 and N2 ambients was noted by Kamataki et al., as shown in Figure 2.17. This was attributed to the growth of SiO2 at the TiO2:Si interface of O2 annealed samples. The analysis of Kamataki ◦ et al. determined that after 1 hr at 500 Can11.6 nm-thick SiO2 layerhadgrownatthe 2.3 Optical Properties 47 Figure 2.16: Variation of refractive index (inset) and absorption coeffi- cient (main graph) with oxidation state of the TiO2−x thin films (from Zakrzewska et al.153). interface. This SiO2 growth rate is much greater than other researchers have observed at the same temperature.17 Figure 2.17: Effect of annealing ambient on the (a) refractive index and (b) extinction coefficient of APCVD-deposited TiO2 thin films annealed at 500◦C (adapted from Kamataki et al.150). Fuyuki and Matsunami noted that the presence of a small amount of water vapour may 147 cause scattered values of the dielectric constant in CVD-deposited TiO2 films. Subse- 48 2. Common Properties of TiO2 Thin Films quent research demonstrated that the presence of water vapour altered the refractive index 155 from about 2.0 (no H2O) to about 2.15 (300 ppm H2O). Ardakani observed that as the hydrogen pressure in the laser ablation chamber increased, the measured reflectance of the 132 films decreased due to the higher concentration of oxygen-deficient phases of TiO2−x. ◦ Golego et al. determined that annealing spray-deposited TiO2 films in hydrogen at 500 C did not affect the absorption spectra, but an absorption peak at 400 − 600 nm appeared after the annealing temperature was increased to 900◦C.115 This indicates that absorption in the visible spectrum occurs after the film is reduced to TiO(2−x). Ottermann et al. deter- mined that there was a clear relationship between the hydrogen content of TiO2 thin films, deposited using different production processes, and their refractive index.156 An increase in refractive index was observed for decreasing hydrogen content, regardless of whether the films were amorphous or anatase. Films formed by chemical reaction methods, such as the sol-gel technique and dip-coating, exhibited high hydrogen contents. Thus, it is assumed that the organic molecules from the liquid precursor have not been fully decomposed by high-temperature processing. Yoldas and O’Keeffe noted that spin-on TiO2 samples annealed in air had a refractive index of 2.1 (at λ = 546.1 nm), whereas the same coating annealed in vacuum had a refractive index of 2.4.157 This increase in refractive index was also observed for films annealed in argon or nitrogen, and was attributed to the increased densification resulting from the thermochemical reactions occurring with the residual terminal bonds in the absence of oxygen.102 2.3.5 Variation with Other Deposition Conditions Depending on the type of deposition system used, there are many different parameters that may be tuned to adjust the optical properties of the TiO2 thin film. The results of Bendavid et al. are included here due to the extremely high refractive indices achieved - 2.56, 2.62 and 2.72 (at λ = 550 nm) for amorphous, anatase and rutile films, respectively. The films were deposited by filtered arc deposition (FAD) and a range of bias voltages were applied to the substrate. Figure 2.18 shows the refractive index and extinction coefficient of two TiO2 films deposited with different bias voltages, 0 V and −400 V, which were determined to be anatase and rutile, respectively. 2.3.6 Optical Properties of Highly Porous TiO2 Films ATiO2 film with a high porosity, or low packing density, will have a reduced refractive index. The optical properties of porous films can be determined using a Bruggeman effective medium approximation (EMA), which can be expressed as158, 159 ε m − ε 1 − ε (1 − fv) + fv =0, (2.10) ε m +2ε 1+2ε 2.3 Optical Properties 49 Figure 2.18: Refractive indices and extinction coefficients of FAD- deposited TiO2 thin films. The film deposited with 0 V substrate bias is anatase and the −400 V film is rutile (from Bendavid et al.33). where fv is the volume fraction of void in the film, and ε and εm are the dielectric functions of the unknown film and the main constituent material, respectively. The dielectric function of the void is taken to be unity, while the complex dielectric constant of a material is related to its refractive index and extinction coefficient by ε =(n + ik)2 . (2.11) TiO2 thin films deposited using the sol-gel method have exhibited void fractions in the 128 ranging from 25% at the exposed surface to 38% at the TiO2:substrate interface. This variation in void content translates directly into a variation in refractive index of the film, which, in this case, was 2.15 (at λ = 500 nm) at the outer surface and 2.05 at the inner surface.128, 160 The higher refractive index at the outer surface was attributed to densification and crystallisation processes that begin at the outer surface and gradually progress through the depth of the film during annealing.128 The void fraction observed in evaporated films 161 was slightly lower at 19.6 − 24.5%. Kim determined that TiO2 thin films deposited by e-beam evaporation had a void content of 16% when compared to dense, polycrystalline 140, 144 anatase. TiO2 thin films deposited by RF sputtering exhibited a much lower void content (< 5%),82 which can be attributed to much greater kinetic energy possessed by the atoms as the impinge on the surface. Another issue related to porous films is the sorption (either adsorption or absorption) of water vapour. When the optical properties of a thin porous film are measured in a vacuum the results may be very different to those achieved in air. This is because on exposure to the atmosphere, the sorption of water vapour, which has a high refractive index than air n . 162 H2O =133, results in a marked increase in the refractive index of the film. Borgogno et al. measured the refractive index of a TiO2 thin film in vacuum and in air, and attributed the 0.1 higher n in air to the sorption of water vapour,163 while Leprince-Wang et al. and Nguyen Van et al. noted a 3% increase in n after exposure to air.161, 164 Ben Amor et al. 50 2. Common Properties of TiO2 Thin Films observed that it was possible to detect the adsorbed water vapour using Fourier transform infrared (FTIR) spectroscopy.165 2.4 Electrical Properties 2.4.1 TiO2: Insulator or Conductor? TiO2 is an n-type semiconductor with a relatively wide bandgap of 3.05 − 3.5 eV. Single 13 crystal titanium dioxide TiO2 has a resistivity of about 10 Ω cm at room temperature, and about 107 Ωcm at 250◦C.29, 116, 155 These values are similar to conductivities reported for single crystal rutile:141 at 30◦C the conductivity was 5 × 10−14 Ω−1 cm−1 while at 260◦C −9 −1 −1 it had decreased to 3.3 × 10 Ω cm . Therefore, TiO2 is generally considered to be an insulator at temperatures less than 200◦C.141, 166 There are a wide number of applications for highly insulating TiO2 films, including its use as a super-thin gate dielectric in MOSFET devices.167 However, the electrical properties of the TiO2 film can be altered to become highly conductive for various applications, such as humidity and gas sensors,83, 168 or as an corrosion-resistant electrode.132, 169 The properties of TiO2 thin films have been likened to those of zinc oxide (ZnO), with typically a large bandgap, small grain size, low carrier density, n-type conductivity, long photoconductivity relaxation, and low mobility.115 This section will briefly discuss the relationship between the electrical properties of TiO2 and the deposition parameters, annealing ambients, and dopant atoms. 2.4.2 Non-Stoichiometric TiO2−x Thin Films The semiconducting properties of non-stoichiometric titanium dioxide are very dependent on the extent of the oxygen deficiency in the film.132, 170 The conductivity mechanism in undoped TiO2 relies on a deficiency of oxygen atoms in the material. These oxygen deficiencies behave like n-type defects, with a typical density of 1018−1019 cm−3.171 As the stoichiometry departs from the ideal Ti:O ratio of 1:2, several changes in the film result. Firstly, as previously mentioned in Section 2.2.6, the TiO2−x films exhibit a blue, purple or black colour depending on the film stoichiometry. Secondly, due to the increased number of oxygen vacancies, the optical absorption in the visible is dramatically increased for these films (see Section 2.3.4). Additionally, these sub-oxides, especially TiO and Ti2O3, are known to exhibit metallic conduction or semiconducting properties.58, 172, 173 The conductivity is observed to increase −10 −1 −1 dramatically with slight departures from stoichiometric TiO2, with < 10 Ω cm for −1 −1 −1 −1 −1 2 −1 −1 TiO2.00;10 Ω cm for TiO1.9995;0.8Ω cm for TiO1.995;and10 Ω cm for 141 −3 −1 −1 TiO1.75. Rao et al. determined that the resistivity of Ti2O3 is about 94 × 10 Ω cm 2.4 Electrical Properties 51 174 at room temperature. Feuersanger noted that the deposition of TiO2 thin films by direct evaporation of a TiO2 source results in the loss of O2, and the films were semiconducting 66 rather than insulating. Tsuda et al. showed that Ti2O3 exhibits a sharp transition in conductivity at 227◦C, exhibiting metallic conductivity above that temperature and semi- conducting below it.173 Thus, TiO2−x thin films would seem to be incompatible with high-efficiency solar cells, as the increased optical absorption is undesirable in an AR coating the buried-contact fabrication sequence requires an insulating film to act as a metallisation mask. All films used in this work were intentionally stoichiometric. 2.4.3 Variation with Deposition or Annealing Ambient Much work has been performed determining the electrical properties of as-deposited TiO2 films, and the effect of the deposition or annealing ambient. It is commonly reported that thin films deposited in a low oxygen partial pressure result in the formation of sub-oxides such 113, 135 as TiO and Ti2O3. This results in increased electrical conductivity in the deposited films.175 When stoichiometric TiO2 films are annealed in a vacuum, hydrogen, or low oxygen pressure ambient, loss of oxygen from the film results.116,132,169,176 However, at room temperature it has been demonstrated that the film conductance does not increase when placed in a vacuum chamber and dry hydrogen gas is introduced.177 Yeung and Lam found that very low conductivity (10−13 Ω−1 cm−1) films could be achieved by performing an extended oxidation 178 of up to 30 hours. Stoichiometric, as-deposited TiO2 thin films possess a negative fixed charge density, in the order of −5 × 10−8 Ccm−2.117,178,179 Erkov et al. observed that after annealing in N2O this fixed charge density became zero, while annealing in vacuum created a 117 film with a positive fixed charge density. In comparison, SiO2 has a positive fixed charge density.179, 180 Takahashi et al. demonstrated that the resistivity of TiO2 thin films could vary by more than an order of magnitude depending on the measurement ambient.170 The lowest resistivity was observed in a hydrogen (reducing) ambient, consistent with the properties of an n-type semiconductor. Additionally, the film conductivity was observed to increase by 104 when illuminated with a fluorescent lamp. 2.4.4 Doped TiO2 58 Doped TiO2 films have been used as transparent conducting oxides (TCO), for creating heterojunction solar cells,59, 61 for suppressing leakage currents in capacitors,166, 181 increasing the sensitivity of gas detectors115, 168 and as electrodes for photoelectrolysis devices.63–65 52 2. Common Properties of TiO2 Thin Films Dopant atoms have also been introduced into TiO2 thin films to create a diffusion source for an underlying silicon layer. These atoms then diffuse into the silicon during a subsequent high-temperature anneal. This will be discussed in more detail in Chapter 6. The majority of dopants enhance the n-type semiconducting properties of TiO2. These dopants include niobium (Nb),116, 168, 182 tantalum (Ta),116 vanadium (V),116 fluorine (F),116 and hydrogen (H).116, 172 Dopants that change the film to being p-type include aluminium (Al),116 iron (Fe),116, 182, 183 and indium (In).59 There are instability issues with both hydrogen-doped (n-type) and all p-type TiO2 thin films. In the former case, the hydro- gen is very mobile and is likely to result in electrically unstable devices.116 In the latter case, the electrical behaviour of the Al- or Fe-doped film depends on the oxygen concentration, and it is possible for the film to revert to n-type.116, 182 Additionally, dopants such as alu- minium, iron,170 chromium and cadmium,59 are known to extend the photoactive response from the TiO2 film under visible light. Table 2.2 includes some references to published works on both doped and undoped TiO2 crystals and films, and their corresponding resistivity. 2.5 Chemical Properties 2.5.1 Chemicals used in Making Solar Cells In general, the resistance of thin films to chemical attack is a highly desirable property for most applications.137 Schr¨oder stated the requirements for a film to be considered chemically resistant. A chemically resistant film must36 i) be insoluble to the attacking medium ii) be impervious to the attacking medium, and iii) form a solid bond to the substrate, preventing chemical attack at edges or through defects in the coating. A large variety of chemicals are used to fabricate solar cells, but this is dependent on the specific type of solar cell and whether it is produced in a laboratory or industrial environment. A summary of the typical chemicals used at different stages in fabricating buried-contact solar cells include: • Sodium hydroxide (NaOH), potassium hydroxide (KOH), or a mixture of hydrofluoric (HF) and nitric (HNO3) acids: for texturing (basic solutions only) or etching silicon • Ammonium hydroxide (NH4OH), hydrochloric acid (HCl) or sulphuric acid (H2SO4): for removing organic and metallic contaminants from the surfaces of the wafers − each 2.5 Chemical Properties 53 Table 2.2: Measured resistivity of anatase and rutile crystals and thin films after annealing in various ambients and/or being doped with various impurities (adapted from Kurtz and Gordon116). Resist. Sample Preparation Reference (Ω cm) Single Crystals - Undoped 1014 Stoichiometric rutile Clark29 ◦ 184 1 Rutile reduced at 1000 CwithanO2 partial Gautron et al. pressure of 10−11 Pa. ◦ 185 3 Rutile reduced at 700 Cfor2hrinH2, result- Butler ing in a carrier conc. of 4 × 1018 cm−3 2 −3 ◦ 10 − 10 Rutile reduced in H2 or vacuum at 750 Cfor Ginley and upto2hr Knotek186 −1 1 187 10 / 10 Anatase: as-grown/after repeated O2 anneal- Forro et al. ing Single Crystals - Doped 6 188 3 × 10 TiO2 doped with 1 mol % boron Johnson 4 188 2 × 10 TiO2 doped with 1 mol % phosphorus Johnson 3 188 3 × 10 TiO2 doped with 1 mol % niobium (Nb) Johnson 189 1.5 TiO2−xFx (x=0.002) Subbarao et al. 20 −3 190 0.1 TiO2 doped with 10 Nb atoms cm Bogomolov et al. 184 0.3 TiO2 doped with 1% Nb and reduced at Gautron et al. ◦ 1000 CwithanO2 partial pressure of 10−11 Pa Thin Films - Undoped 10 191 4 × 10 Evaporated films annealed in O2 for up to Yokota et al. 14 hr 1010 − 109 MOCVD at 200◦C as-deposited Fuyuki and Mat- sunami147 107 / Reactively sputtered anatase as-deposited / Tang 192 5 × 10−2 reduced at 450◦C in vacuum 6 4 −3 −6 135 10 − 10 Evaporation of Ti under 10 − 10 Torr O2 Chen et al. 107 − 102 1.6 − 17.2 µm thick films (higher deposition Takahashi et al.71 temperature leads to higher resistivity film) 103 Films deposited via spray-CVD method at Badawy and El- 400◦C Taher193 Thin Films - Doped 7 3 170 10 − 10 CVD of Al-doped TiO2 films, deposited at Takahashi et al. 435 − 480◦C 4 3 170 10 − 10 CVD of Fe-doped TiO2 films, deposited at Takahashi et al. 415 − 470◦C 2 170 4 × 10 CVD of Cr-doped TiO2 films, deposited at Takahashi et al. 435◦C 54 2. Common Properties of TiO2 Thin Films of these chemicals is mixed with hydrogen peroxide (H2O2) and de-ionised (DI) water at a ratio of 1:1:5. With NH4OH:H2O2:H2O this is referred to as an ”RCA1” clean, 194 while HCl:H2O2:H2O is called an ”RCA2” clean. • HF: for removing silicon dioxide (SiO2) and doped SiO2 layers − diluted to about 5% in DI water • Nickelex − commercial product from Transene Inc. (MA, U.S.A.) that contains nickel chloride. It is used for forming a thin electrolessly plated nickel layer in the grooved contacts in buried-contact solar cells. • Enplate Cu-704 − another commercial product (Enthone, Melbourne) that has three components (A, B, M), containing copper sulphate (CuSO4), formaldehyde (HCHO), NaOH, potassium cyanate (KCN), as well as other organic wetting agents. It forms the main bulk of the conductor in the grooves. • Silver plating − contains (AgCN) silver cyanate. It is used as a flash coating to improve solderability During the silicon etching procedures, the TiO2 film would not normally be present on the wafer, however the TiO2 would ideally be resistant to all of the chemicals used in the subsequent cleaning and metal plating processes. As the majority of results for the chemical resistance of TiO2 films are against HF this acid will be dealt with in a separate section. 2.5.2 Hydrofluoric Acid A critical step in the buried-contact solar cell fabrication sequence is the removal of the P2O5:SiO2 from the grooves before performing electroless metal plating, using either dilute hydrofluoric acid (HF) or buffered HF (1:15, HF:NH4F). This allows nickel to plate to the heavily doped silicon, however if even a thin oxide is present the metal will not plate. In general, the crystalline phases of TiO2, anatase and rutile, are much more chemically resistant than amorphous TiO2. Kurtz and Gordon noted that TiO2 films deposited via atmospheric pressure chemical vapour deposition at 400 − 600◦C were very chemically inert and that surpassed the chemical resistance of glass to attack by common solvents and acids.116 These films could not be removed mechanically or chemically, and etching away the glass substrate in HF left an intact TiO2 film. Feuersanger reported that films deposited at 150◦C were easily etched in 10% HF.66 Fitzgib- bons observed that films deposited by CVD at 150◦C etched at 50 − 75 A/s˚ in 0.5% HF and very rapidly in 48% HF.67 Upon subsequent annealing at 350◦C, dilute HF undercut the TiO2 layer by dissolving a thin interfacial SiO2 layer, while concentrated HF etched the film slowly and unevenly. Annealing at 700 − 1000◦C made the films resistant to dilute HF, and 2.5 Chemical Properties 55 concentrated HF resulted in undercutting. Spray deposited films annealed for 30 s at 450◦C were removed in 2 − 5 min when subjected to 1 − 5% aqueous solutions of HF,46 however the aim of this work was to ensure the TiO2 films were amorphous. Yokozawa found that TiO2 ◦ (anatase) films deposited by CVD at temperatures of 530 − 700 CinN2 hardly dissolved (< 0.03 A/s),eveninconcentratedHF.˚ 118 On the other hand, films deposited at the same temperatures, but in an N2 and O2 ambient, were less crystalline and etched at 4− 60 A/s˚ in diluted HF. Fuyuki et al. deposited TiO2 film deposited by metal-organic CVD (MOCVD) with water vapour present. The etch rates of these films in 10% HF were about 5 nm/min at 400◦C and 200 − 1000 nm/min at 200◦C.155 Brown and Grannemann used buffered HF to ◦ remove TiO2 films annealed in O2 at 1000 C, by undercutting grown SiO2 and leaving the 31 ◦ TiO2 film intact. Rausch and Burte observed that films deposited at 450 C by low-pressure CVD (LPCVD) etched in 50% buffered HF.195 Frenck et al. determined that parameters other than in situ and ex situ temperature have only a minor influence on etch rate in buffered HF.124 If no post-deposition annealing step is performed, the PECVD films are only chemically resistant to buffered HF with deposition temperatures greater than 270◦C. Lee noted that the film etch rates in buffered HF decreased from 2.9 nm/s to 0.14 nm/s as the anatase fraction of the film increased from 0.35 up to 0.7.113 Thus, there is a trend of increased chemical resistance of TiO2 thin films deposited or an- nealed at higher deposition temperatures. The majority of films treated at temperatures greater than 300◦C either could not be etched, or only etched slowly, in HF. This tempera- ture coincides with the amorphous−anatase transformation. There are a number of published works that contrast the above trend of increased chemical resistance with increased annealing or deposition temperatures. Rausch and Burte claim ◦ that the etch rate of TiO2 films annealed at 900 C was four times greater than the as- deposited layer, which etched at a rate of 10 nm/min in 50% buffered HF.195 However later in the paper, it is stated that the annealed TiO2 film lifted off, exhibiting therefore a greater chemical resistance. Secondly, Harbison and Taylor reported an etch rate of 700 A/min˚ in 48% HF for material grown by hydrolysis at 800 − 1000◦C.179 This anomalous result does not agree with the above trend, especially considering the high growth temperature. Other ◦ results of TiO2 films deposited at relatively high temperatures (550 − 650 C) that etched, include the films prepared by Balog et al. using MOCVD, which etched in a solution of 5% HF.196 Burns also found that films formed by rapid thermal annealing (30 s at 550 − 650◦C) of titanium metal at these temperatures could be etched in HF.197 It is claimed that the films are polycrystalline rutile, which is in agreement with the high dielectric constants reported, but not the poor chemical resistance. It has been noted that the chemical resistance to HF 141 is highly dependent on the film deposition technique and water-content of the TiO2 films. The inclusion of a small fraction of SiO2 (6.25 at. %) into the TiO2 film results in rapid etching (70 nm/min) in buffered HF, even after annealing at 1260◦C for 10 min.198 56 2. Common Properties of TiO2 Thin Films 2.5.3 Other Acids and Bases Kurtz and Gordon found that TiO2 films were chemically resistant to attack by common solvents and acids, and surpassed the chemical resistance of glass.116 The chemical resistance 141 of TiO2 to sulphuric acid (H2SO4) is very dependent on the film preparation technique, but water-free TiO2 is insoluble in all other acids and bases. This is in agreement with Barksdale, who observed that TiO2 is known to be slightly soluble in H2SO4, HF, and a few strong alkalis, however after annealing at 1000◦C it is almost completely chemically inert.199 ◦ TiO2 films prepared by Balog using MOCVD at 550 − 650 C were able to be etched in a 196 solution of 70% H2SO4. Fitzgibbons found that the chemical resistance increased with temperature, with films deposited at 150◦C etching at 25 − 40 A/s.˚ 67 However, even films ◦ annealed at 1000 C still etched slowly (1000 A/hr˚ in boiling pure H2SO4. At the same an- nealing temperature, TiO2 films were observed to etch very slowly in 85% H3PO4. Yoldas and O’Keeffe found that for 1 wt. % concentrations of H2SO4,H3PO4, and HNO3 no observ- able deterioration was observed after 75 days.157 That research also showed an increase in ◦ chemical resistance to acids for TiO2 films annealed at higher temperatures (up to 400 C). ◦ TiO2 films deposited by the sol-gel technique and baked at 120 C were observed to etch in 200 boiling 0.3 mol HNO3 (pH=0.5). ◦ Schr¨oderfound that an observable colour change with TiO2 films baked at 200 C occurred after about 4 hours immersion in a 10% HCl solution, whereas after baking at 550◦Cthe same colour change occurred after more than 200 hours.36 Szlufcik et al. determined that a ◦ 1 wt. % HCl solution completely deteriorated TiO2 films fired at 300 C after 26 days, however no observable deterioration was achieved after firing at temperatures greater than 600◦C.114 For spray-deposited films (450◦C) no observable deterioration was found after placing the films in a 1% solution of boiling HCl for 1 hour.46 The resistance of TiO2 thin films to various bases has also been reported in the literature. Schr¨oder performed experiments with TiO2 thin films in a 10% NaOH solution, and found that for films annealed at 200◦C an observable colour change occurred after 5 hr.36 Increas- ing the annealing temperature to 500◦C delayed the same colour change observation to more ◦ than 200 hours. TiO2 films baked at 120 C were badly corroded after placing them in boil- ing 0.5 mol NaOH solution (pH=13.5) for 30 min.200 The addition of 1 at. % boron oxide (B2O3) significantly increased the chemical resistance of the film. Kern and Tracy found ◦ no observable deterioration for spray-deposited TiO2 films annealed at 450 Cina1%so- 46 lution of NH4OH. For screen-printed TiO2 films the chemical resistance to 1 wt. % NaOH ◦ 114 and NH4OH solutions was poor for annealing temperatures of 600 C. No observable de- terioration was detected after increasing the annealing temperature to 800◦C. Honsberg et al. found that cleaning 60 nm-thick, spray-deposited TiO2 films on silicon wafers in RCA solutions resulted in thickness reduction in the film of 6 nm.20 This is most likely due to the 194 NH4OH in the RCA1 clean. Yoldas and O’Keeffe found that the chemical resistance of 2.6 Conclusions 57 ◦ TiO2 to 1 wt. % NH4OH films fired in a vacuum at 500 C) was greater than that of films fired in air.157 For the air-baked films, the deterioration was obvious after 7 days, and complete after 10 − 20 days. NH4OH based etches can also be used for removing unreacted titanium in films.113 Changes in the reflectance spectra of TiO2 films containing small fractions of SiO2 that were fired at 400◦C were noted for 1% HCl solutions after 24 hours.201 No change was noted for a 1% NH4OH solution after 192 hours, but the 1% NaOH solution resulted in the swelling of the film and a reduction in refractive index. The hierarchy of chemical attack for these films 36 was given as NaOH > HCl > NH4OH. This is in agreement with the results of Schr¨oder, but 157 in contrast with the NH4OH results from other research. However, Yoldas and O’Keeffe postulated that the increased chemical resistance to sol-gel films baked at only 80◦ could be due to the retention of organic groups.157 The results of Schr¨oder indicate that the chemical 36 resistance of TiO2 containing SiO2 is significantly poorer than pure SiO2 films. ◦ TiO2 films deposited at above 400 C have also been shown to be resistant to the nickel and copper electroless metal plating solutions used in the buried-contact solar cell fabrication sequence.20 The chemical resistance to the copper plating solution is significant as it is strongly basic (pH=11) and contains NaOH. Thus, increased deposition and annealing temperature result in greater chemical resistance for the majority of other acids and bases - a similar trend to that observed with hydrofluoric acid. 2.6 Conclusions The properties of TiO2 depend greatly on whether the sample is bulk material or a thin film and the phase of the material. With CVD techniques the phase is usually directly related to ◦ the substrate temperature, with amorphous TiO2 forming at temperatures less than 300 C, the metastable crystalline phase of anatase in the temperature range 300 − 700◦C, and the stable crystalline phase of rutile at temperatures greater than 700◦C. Various impurities and substrate types are known to partially or totally impair or enhance the transformation of aTiO2 thin film to rutile. A useful linear relationship can be found between the refractive index and TiO2 film density. While optical constant data for rutile is fairly easy to find, it was necessary to collect small sets of data for anatase from over many decades in order to construct a dispersive refractive index model. Anatase has a fundamental absorption edge with a lower steepness than rutile, due to the increased disorder observed in anatase crystals. A trend of increasing refractive and extinction coefficient with increasing deposition temperature is commonly observed. Thus, accurate control of the temperature results in accurate tuning of the TiO2 thin film’s 58 2. Common Properties of TiO2 Thin Films refractive index and, to a certain extent, extinction coefficient. The electrical properties are briefly discussed. Although TiO2 is a wide bandgap n-type semiconductor, at room temperature it behaves like an insulator. The film resistivity is extremely sensitive to the deposition method and on the availability of oxygen to the system. Films deposited in oxygen poor ambients exhibit greatly increased electrical conductivity and optical absorption, however subsequent furnace oxidation processes can reduce these oxygen vacancies. Doping of TiO2 thin films in order to increase the electrical conductivity or photoconductivity of the film has been experimented with in many different applications. The chemical resistance of TiO2 thin films change markedly during their amorphous- polycrystalline transition. Amorphous TiO2 film are highly soluble in hydrofluoric acid, while dense, polycrystalline films can be insoluble. TiO2 films seem to be most susceptible to etching in strong basic solutions such as sodium hydroxide and ammonium hydroxide. The chemical resistance to sulphuric acid is dependent on the film preparation technique. There is a definite trend of increased chemical resistance to all chemicals with increased film deposition or annealing temperature. Chapter 3 TiO2 Thin Film Deposition Equipment After evaluating the desired film properties and performing a literature survey on the possi- ble deposition methods, the author designed and constructed two TiO2 thin film deposition systems. The first system used an ultrasonic atomisation spray nozzle in order to create an aerosol of the TiO2 precursor. The reasons for choosing ultrasonic spray deposition (USD) and the TiO2 precursor, tetraisopropyl titanate (TPT) are discussed. A diagram of the sys- tem is presented and the necessary components described. With this system, very dense ◦ TiO2 films could be deposited at a temperatures of 450 C and the very shallow deposition angles successfully prevented TiO2 film deposition in grooves scribed in the front surface of the wafer. Thickness uniformity and chemical resistance problems with the electroless metal plating solutions, used for contact formation in the buried-contact solar cell fabrication se- quence, arose due to the frequent inclusion of TiO2 particulates (1 − 30 µm in diameter) in the 70 nm thick films. For this reason the TPT was placed in a stainless steel bubbler, resulting in the development of a simple atmospheric pressure chemical vapour deposition (CVD) system. TiO2 films deposited using the CVD system exhibited much greater thickness uniformity and a lack of particulates. Additionally, it was also possible to deposit films anywhere in the temperature range 150 − 450◦C, enabling the refractive index to be tuned. The tradeoff with the new system was that the film density decreased significantly. 3.1 Introduction The previous chapter described the physical, optical, electrical and chemical properties of TiO2 thin films, primarily as a function of deposition or annealing temperature. Based on previous experience with a TiO2 deposition system at UNSW, TiO2 thin films with the 59 60 3. TiO2 Thin Film Deposition Equipment following properties were desirable for this work: • Physical: Dense, defect-free films with a thickness of about 70 nm. The thickness uniformity should be within ±10%. Films that are dense and lack defects (such as large, amorphous TiO2 particulates) will be impervious to chemical attack. Denser films will typically perform better as a diffusion barrier as well.120 • Optical: Dense TiO2 films will also enable a refractive index of about 2.4 for light of 600 nm wavelength to be achieved at low deposition temperatures (< 450◦C). This value is the optimum refractive index for an antireflection (AR) coating for a silicon solar cell encapsulated under glass. An excellent AR coating is necessary to reduce the amount of reflected light from the planar multicrystalline silicon (mc-Si) wafers. • Electrical: The TiO2 film must be insulating in order to act as a metallisation mask to the electroless metal plating solutions. This requires that the films are stoichiomet- ric, and do not exhibit the oxygen vacancies that form the conduction mechanism in reduced TiO2−x thin films. • Chemical: The TiO2 films should be polycrystalline anatase or rutile in order to with- stand the necessary wet chemical processing during solar cell fabrication. The films need to withstand RCA cleaning, dilute hydrofluoric acid, and the electroless metal plating solutions. A simple TiO2 pressurized spray system had already been in operation for some years at the UNSW, using tetraisopropyl titanate (TPT) as the precursor.202 This system used a spray-painting gun pressurized with nitrogen, and the TPT flow was controlled via a needle valve. A TPT aerosol was formed and transported to the wafer by the nitrogen flow. The wafer was held by vacuum onto a stainless steel block that was placed on top of a 2 kW stove element. The whole deposition process was carried out inside a fumecupboard. This system was used for several years, and it demonstrated that geometry could be used to direct the emerging aerosol, so that when spraying at shallow angles the TiO2 could be kept out of the grooves.202, 203 There were a number of problems with the TiO2 films and the deposition system. Firstly, it took over an hour to deposit a 70 nm thick TiO2 film onto a 2”-diameter wafer. This was not due to the low TPT flow rates, although the needle valve was being operated at its limits to avoid a too high a flow of TPT from emerging from the nozzle, but because of the cooling effect of the aerosol on the stainless steel block. After several seconds spraying, the temperature of the block dropped by over 30◦C, and it was necessary to wait a few minutes before continuing spraying. This highlighted the second problem, which was that the heater was very inefficient at transferring heat to the stainless steel block, and this excess heat made it uncomfortable for the user. Thirdly, there was no way of controlling the relative humidity, 3.1 Introduction 61 and the relative humidity on rainy or overcast days (up to 65%) made spraying impossible due to the formation of white TiO2 particulates in the film. A replacement TiO2 system was to be designed and constructed by the author. In order for the TiO2 thin films to exhibit the desired properties, listed above, the requirements for the new system were determined as follows: • Temperature: Significant fluctuations in the temperature at about 400◦C could result in the TiO2 film having a mixed amorphous-anatase phase. Naturally, this outcome is undesirable, so a target of achieving a maximum variation of substrate temperature of ±10◦Cwasset. • Relative Humidity: Excess humidity will result in large TiO2 particulates sticking to deposited film.202, 204 Therefore, a necessary feature of the system was to have adjustable and repeatable humidity control. It was anticipated that mixtures of dry nitrogen (N2) and wet nitrogen (N2+H2O) could be fed into the system in order to control the relative humidity. A meter should display the current relative humidity in the system. • Deposition Time: To achieve accurate control of the film thickness, a deposition time of about 5 − 10 min per wafer was decided upon. The film thickness would be judged visibly. • Deposition Area: As this project had commercial relevance, it was desirable that a 4”- square wafer could be TiO2-coated, although the research would typically be performed on 2”-round and 5 × 5 cm-square wafers. • Geometry Control: To prevent TiO2 from being deposited in the laser-scribed grooves, film deposition was to occur on the very top surface only. Any TiO2 that entered the grooves would inhibit electroless metal plating. The pressurized spray-system at UNSW had successfully fulfilled this goal by spraying a TPT aerosol onto the wafers at an angle of 10 − 20◦. • Substrate Heater: A dedicated substrate heater is required to efficiently transfer heat into the substrate. The temperature should be accurately controlled by dedicated electronics. • Automation: The previous pressurized spray system required the user to manually operate a spray-gun. In the new system it was deemed necessary that the user should only need to load and unload the wafer from the system, the user’s hands remaining free during film deposition. • Safety: The system should be enclosed so that it can safely operate on a standard laboratory bench. It also needs to have exhaust facilities. It was desirable to continue 62 3. TiO2 Thin Film Deposition Equipment using TPT as the precursor as it is a safe, non-toxic liquid, and obviates the need for expensive gas handling systems. • Cost: There was a budget of A$10,000 for the new TiO2 deposition system. Initially, a new system was designed using an ultrasonic atomisation nozzle. Figure 3.1 shows the USD system in its final form, including the ultrasonic nozzle and generator, syringe pump, cartridge heaters embedded in the stainless steel block, motorized sample stage, and temperature controller. Some components have been omitted for clarity, including a nitrogen heater, air-knife heater, motor speed controller, both the oxygen and humidity sensors, as well as regulators and flow meters to accurately control gas flows. The following section will describe the literature review performed to evaluate simple, flexible and cost-effective techniques for depositing TiO2 films. Subsequently, the theory of ultrasonic spraying will be discussed, and a thorough description of the necessary system components will be presented. 3.2 Overview of TiO2 Thin Film Deposition Methods Titanium dioxide has been deposited by many different techniques, including • hydrolysis and pyrolysis,49, 60, 65–67, 71, 74, 78, 115, 148, 168, 177, 179, 193, 205–207 • pneumatic spraying,46, 208 • ultrasonic spraying,65, 106, 206, 209–212 • dip coating100, 109, 111, 156, 157, 200, 207, 213 • plasma enhanced chemical vapour deposition (PECVD),93, 97, 113, 121, 124, 134, 214–216 • atmospheric pressure chemical vapour deposition (APCVD),116, 130, 150, 204, 217–221 • metal organic chemical vapour deposition (MOCVD),58, 71, 77, 92, 94, 96, 103, 105, 106, 108, 147, 155, 167, 170, 195, 196, 211, 220–227 • ultra-high vacuum chemical vapour deposition (UHV-CVD),228 • low pressure chemical vapour deposition (LPCVD),90, 91 • evaporation,31, 34, 73, 76, 119, 121, 137, 156, 161, 213, 229–231 • spin-on methods,89, 100, 112, 121, 157, 198, 200, 232, 233 • sputtering,119, 120, 145, 149, 154, 165, 234–236 3.2OverviewofTiO2 Thin Film Deposition Methods 63 r PID contolle temperature C C q q r 445 450 DC 1.8 W 12 V ultrasonic generato line switch (x2) switch thermo -couple vacuum st. steel block relay housing TPT storage TPT bottle rings vacuum motor 3-way valve TPT heaters cartridge nozzle 2 ultrasonic hot N syringe air-knife syringe pump syringe 0.02 ml/min Figure 3.1: TiO2 ultrasonic spray deposition system. 64 3. TiO2 Thin Film Deposition Equipment • ion assisted deposition,72, 107, 140, 237, 238 • plasma anodisation,113 • reactive ion plating,121,156,213,231 • laser ablation,132, 239 • filtered arc deposition,33 • atomic layer epitaxy,240 and • screen-printing.114, 241 A primary consideration is that the growth morphology, crystalline structure and stoichiom- 73, 110, 122, 237, 242 etry of TiO2 thin films are very sensitive to the deposition conditions. This a disadvantage for many physical vapour deposition methods, such as evaporation, where a large variation in the observed optical properties arise from only a small changes in the de- 107, 161, 164 position conditions. Therefore, the need for stoichiometric TiO2 films with minimal absorption suggested that a deposition method where the film stoichiometry is controlled by a chemical reaction would enable more consistent results. The chemical reaction of a TiO2-precursor to form TiO2 can either proceed by hydrolysis or pyrolysis. With hydrolysis systems, separate gas lines of nitrogen or argon are bubbled through heated baths of a liquid TiO2 precursor and water. The two delivery lines are then brought together close to the substrate where the reaction takes place. Pyrolysis systems are similar, except that a water bath is not required as the TiO2 precursor decomposes upon reaching the heated substrate. Both of these systems have the advantage of simplicity, although they may be relatively inflexible, as tubing diameters have to be designed around predicted flow rates. To keep the deposition system as simple as possible, maximise throughput, and keep costs at a minimum, systems with a vacuum chamber were not considered to be a viable option. This excluded evaporation, sputtering, and the majority of CVD systems. However some exper- iments were performed with an APCVD system,204 owned by Eurosolare S.p.A. (Nettuno, Italy). The APCVD system is designed around a belt furnace, and, as the name implies, the depositions are performed at atmospheric pressure, so no vacuum is required. The system 217 is capable of depositing TiO2 onto 580 solar cells per hour, which corresponds to about a 5 MW yearly throughput for screen-printed solar cells. Spin-on methods were not seriously considered, as the throughput of any such system would be limited in a production environment. Screen-printing is commonly used in the PV in- dustry for depositing metallic contacts, about 30 − 50 µm thick, to solar cells. Szlufcik et 114 al. demonstrated that screen-printing could also be used for depositing TiO2 thin films. As the thickness of the metallic contacts are about 500 times thicker than the TiO2 thin 3.3 Ultrasonic Spray Deposition 65 films it is not known how the thinner films behaved with regard to reproducibility, squeegee wear, and thickness uniformity. Following the screen-printing of the organometallic ink, the samples were fired in a three-step process of 30 min duration. The firing is a relatively slow process as first the thick film needs to settle for 15 min to obtain a uniform film, with subse- quent drying performed at 125◦C for 5 min. The final crystallisation was performed in a belt furnace at temperatures between 500◦C and 900◦C. Lengthy drying procedures are required to remove the substantial amount of organic solvents added to the TiO2 precursor. This was also necessary in the pneumatic spraying technique used by Kern and Tracy.46 Kern and Tracy developed a system for production was developed that was capable of coating 4500 cells per hour (around 30 MW per year), based on batch processing. Each batch took 30 s to receive a coating, followed by three separate heating steps to remove organic groups in the film. It has been noted by other researchers that the temperature required to crys- tallise an amorphous film is significantly greater than the temperature needed to grow such a crystalline film.74 Therefore, to lower thermal the budget and processing costs it would be desirable to deposit a polycrystalline TiO2 thin film in one step without subsequent heat treatment steps. 3.3 Ultrasonic Spray Deposition 35, 65, 106, 206, 210, 212 Several researchers have used USD for depositing TiO2 and other thin films.209, 211, 243 Blandenet et al. describes a method for depositing many different metal- lic oxides based on the pyrolysis of an aerosol.210 If an ultrasonic beam is focussed on the surface of a liquid the vibrations result in the formation of on aerosol. The chemical to be sprayed is contained in a glass container attached to a high frequency generator that vibrates at 800 − 1000 kHz. The chemical precursor for TiO2 films was butyl orthotitanate diluted in acetyl-acetone and butanol. Air or nitrogen is passed through the glass container, transporting the aerosol close to the heated substrate, which is subsequently decomposed by pyrolysis. Other researchers have also used commercially available ultrasonic nebulizers,65, 211 while more recently ultrasonic atomizing nozzles have been used.35, 58, 106, 212, 243 Versteeg et al. used an ultrasonic nozzle to inject small quantities of a TiO2 precursor into a vacuum chamber, maintained at a pressure of 0.1−1 Torr.106 It is stated that the low pressure cham- ber facilitates film uniformity over large areas. Liang implemented an ultrasonic nozzle as a modified injector in an APCVD system.58 DeSisto and Henry deposited magnesium oxide thin films by USD.243 The primary advantages of USD are that, firstly, there is a very narrow size distribution of the droplets in the aerosol. Secondly, by altering this droplet size, the droplet→solid reaction mechanism can also be changed. This is also dependent on gas flows and the nozzle to substrate distance. For example, smaller droplets will have already decomposed by pyrolysis and will strike the wafer as a solid, while larger droplets will not have had time to vaporise 66 3. TiO2 Thin Film Deposition Equipment completely. Alternately, the spraying conditions can be adjusted so that the majority of the aerosol is a gas when it contacts the wafer, resulting in a process similar to CVD.209, 210, 244 Additional advantages of USD that were potentially relevant for this work, included: i) The potential flexibility of the system and excellent parameter control. This included low flow rates, variation of spray angle, variation of power to atomize liquids of different viscosities and at different flow rates, and control over the shape of the spray plume using air-jets.245, 246 ii) The possibility of spraying at a shallow angle to keep the TiO2 out of the grooves, scribed on the front side of a buried-contact solar cell. In this manner, the simple geometry of the system prevents the spray from entering the grooves.20, 202, 203 With other deposition systems a lot more care would need be taken to keep the grooves free of TiO2. This is especially true for the majority CVD systems which provide a conformal coating of the surface. iii) Since the distribution of the droplet sizes can be varied by choosing the excitation frequency (see Section 3.4), it was postulated that larger droplet sizes (greater than the groove width) would not enter the grooves, but would still coat the front surface of the wafer.247 iv) It is possible to pulse-feed the liquid into the nozzle, enabling very low flow rates to be achieved.106, 245 v) It is possible to shape the spray plume by directing jets of gas across it. This can be used for focussing the spray plume or to spread it out.245 vi) As decomposition by hydrolysis or pyrolysis is a chemical process, this would en- sure that stoichiometric TiO2 was deposited. The deposition of TiO2 by physical vapour deposition methods, evaporation and sputtering for example, can result non- stoichiometric TiOx films. TiOx films exhibit very different properties including in- creased absorption and a metallic or semiconducting behaviour. vii) No vacuum chamber would be required. viii) The nozzles do not clog or wear out.245 ix) The system would be easy to clean by flushing with a solvent,248 such as isopropyl alcohol. x) Since overspraying is avoided, material consumption can be reduced by up to 80%.245 As well as avoiding waste, it was anticipated that this would reduce the build-up of TiO2 powder in the system. 3.4 Theory of Ultrasonic Spray Deposition 67 xi) Spraying may be suitable for production if accompanied by a suitable pump, with a continuous flow, such as a gear pump or pressurized feed, and either multiple nozzles or a single traversing nozzle.46 xii) An USD system can be used to deposit a variety of films, including AR coatings and SiO2 passivation layers. Additionally, it is possible to add other dopant liquids to the main precursor.65 ◦ xiii) TiO2 films deposited by spray pyrolysis at 450 C are known to produce nearly dense, optically transparent, anatase films.177, 206 The use of an ultrasonic spray nozzle appeared to offer the best flexibility and potential for the new system. Out of the five surveyed models available on the US and Australian markets, only one (Sono-Tek Corp., U.S.A.) had been previously used in published scientific literature. This brand also had the advantage of being able to operate nozzles with a different atomisation frequency from the one ultrasonic generator. 3.4 Theory of Ultrasonic Spray Deposition Blandenet et al. provide a good description of USD-deposited films produced by an aerosol.210 In the case of most metal oxide depositions, the aerosol is a colloidal disper- sion of a organometallic liquid in a carrier gas. The liquid is atomized in a glass vessel with a transducer operating in the kHz to MHz frequency range. The carrier gas transports the aerosol close to the heated substrate, where it is decomposed by pyrolysis. Alternately, the aerosol can react with water vapour to complete the reaction by hydrolysis. In such systems, unlike pneumatic spraying, the gas flow rate is independent of the aerosol flow rate. In an ultrasonic nozzle, two piezoelectric transducers are contained within the nozzle, as shown in the cut-away view in Figure 3.2. This sets up a standing wave and both ends of the nozzle become anti-nodes. The junction between the two transducers is a node, a point of zero amplitude, because the transducers have their polarities opposed. This causes the transducers to either expand or contract against each other. The standing waves and anti-nodes are illustrated in Figure 3.3.248 As the TPT passes through the nozzle, the ultrasonic vibrations (48 − 120 kHz) form an aerosol. The mean diameter of the droplets produced depend upon the excitation frequency f, the surface tension σ, and the density of the liquid ρ being atomized210, 245 8πσ d = 3 k ρf 2 8πσ ∼ 3 0.34 ,(3.1) = ρf 2 68 3. TiO2 Thin Film Deposition Equipment Figure 3.2: Cut-away view of Sono-Tek ultrasonic nozzle.248 Figure 3.3: Standing waves inside ultrasonic nozzle.248 where k is a constant, and its value of 0.34 determined by Lang.249 This relationship is plotted for a range of frequencies in Figure 3.4 for water (ρ =1g/cm3 and σ =0.073 N/m) and isopropyl alcohol (ρ =0.80 g/cm3 and σ =0.0217 N/m) at temperatures of 20−25◦C.245 The intercepts on the graph indicate the median droplet diameters for water at the excitation frequencies used in this work, 3.1µm at 48kHz and 1.7 µm at 120 kHz. Although the density of TPT is similar to that of water at 0.955 g/cm3,250 no data on the surface tension could be found. Therefore, the size of the TPT aerosol droplets could not be accurately estimated. Although small droplet sizes are achievable with pneumatic spraying, the main advantage of USD is the narrow size distribution of the droplets.210 Figure 3.5 shows the distribution of water droplet sizes for the range of excitation frequencies 25 − 120 kHz.245 The number median diameter defines the 50% value of drop size, meaning that one half of the drops 3.4 Theory of Ultrasonic Spray Deposition 69 103 m) Water µ ( d Isopropyl Alcohol (at T=20-25oC) 102 101 Median Droplet Diameter, 100 1 10 100 1000 Excitation Frequency, f (kHz) Figure 3.4: Median droplet diameter as a function of excitation frequency for water and isopropyl alcohol. have diameters larger than this, while the remaining half are smaller. The number mean is obtained by summing the diameters of each drop together and then dividing by the number of drops in the sample. The weight mean diameter is calculated by taking the adding together the volume of each drop in a spray sample, taking the cube root of this sum, and dividing by the number of drops. The Sauter mean diameter measures the effective ratio of drop volume to surface area and is primarily used for combustion applications. Other factors affecting the operation of ultrasonic spraying are liquid viscosity, solids con- tent, and the miscibility of components.245 Although there are no hard-and-fast rules for determining the suitability of a liquid for ultrasonic spraying, there are several guidelines. Generally, the higher the viscosity or solids-content of a liquid, the lower the maximum flow rate. Liquids with a viscosity of up to 500 mPa s (or 50 centipoise) can be readily atom- ized,248 where water and TPT have a viscosity of 1 mPa s and 3.5 mPa s,251 respectively, at 20◦C. Liquids containing long polymer chains can interfere with the atomization process and may inhibit the formation of discrete droplets. It has been found that liquids with a solids content of up to 40% can be successfully atomized, however the particle size should 245 be less than one-tenth of the droplet diameter. The maximum flow rate Fmax has been empirically determined to be A F = k (3.2) max f 2/3 where the constant k has the value k ≈ 28500 l Hz2/3/(s m2), and A is the atomizing surface area. Figure 3.6 plots Equation 3.2 for the range of ultrasonic nozzles produced by Sono-Tek. The specific flow rate r is the maximum flow rate Fmax divided by the atomizing area A. The specific flow rate for the 120 kHz nozzle, not shown in Figure 3.6, is 11 l s−1m2.More importantly for research applications though, is the minimum flow rate. The minimum flow 70 3. TiO2 Thin Film Deposition Equipment Figure 3.5: Drop size distribution for ultrasonic nozzles operating at var- ious frequencies (from Berger 245). rate is typically about 20% of the maximum flow rate, as below this rate the liquid emerges from the nozzle in a non-uniform manner and the spray plume becomes distorted.245 3.5 TPT: The TiO2 Precursor 3.5.1 Why TPT? Titanium isopropoxide, also known as tetraisopropyl titanate (TPT), was chosen as the TiO2 precursor. Apart from being the most commonly used precursor in the literature, this chemical is also used on solar cell production lines. The use of any precursor can result in contamination of the TiO2 film with by-products of the precursor. Metal-organic precursors, such as TPT, often result in carbon contamination due to the residue of the organic binders.89–93, 95, 96, 98 This is typically in the order of a 1 − 2 at. % for films deposited at low temperatures. However, several researchers have observed that at higher deposition or annealing temperatures (400 − 600◦C) the carbonate species can decompose, resulting in the removal of hydrocarbon fragments.64, 99, 252 Chen et al. noted that the decomposition of TPT to TiO2 is a very clean process in which carbon is not significantly trapped either 3.5 TPT: The TiO2 Precursor 71 Figure 3.6: Specific flow rates as a function of excitation frequency (from Berger.245) 135 within the crystalline film or at the grain boundaries. Titanium tetrachloride (TiCl4)is 91, 93, 101 another common TiO2 precursor, which results in chlorine contamination. In addition corrosive by-products (HCl) are produced in the reaction.66, 91 In one instance the chlorine contamination was so high that it prevented crystallisation of the film and poor film adhesion onto the substrate resulted.101 In any case, the level of contamination observed with TPT 97 is much smaller than with TiCl4. Further advantages of TPT are that: i) It is non-corrosive124 and non-toxic, listed as being a mild skin and eye irritant.250 ii) It can be highly purified and has an almost indefinite shelf-life.124 iii) As a liquid, it is relatively easy to handle,97 although it should not be exposed to a naked flame.250 The fact that it is not dangerous makes the addition of TPT to a CVD system a relatively easy and safe task, as no special gas handling equipment is required.97 iv) It is very volatile at low temperatures (50◦C), which means that it will be readily decomposed.97 v) It can be ultrasonically sprayed directly without dilution.49 vi) It has been observed that there is enough oxygen in the TPT molecule that the reaction 94, 99, 101, 124, 221 to form TiO2 can proceed without additional oxygen in the ambient. 72 3. TiO2 Thin Film Deposition Equipment 3.5.2 The TPT→TiO2 Reaction The mechanism of the reaction of TPT aerosol to form TiO2 depends on the droplet size. If the majority of the aerosol is a gas when it contacts the substrate then the deposition conditions are similar to a CVD process.210 Aerosols with larger droplet sizes will not have had time to vaporise completely, while smaller droplets will be decomposed by pyrolysis be- fore striking the substrate. Since the droplet size distribution is small in ultrasonic spraying, the same decomposition conditions will apply to nearly all of the aerosol. The deposition conditions vary also with temperature, as the diagram for pyrolysis in Figure 3.7 shows, however other factors such as flow rates and geometry also play a role. In process A, the decomposition rate at very low temperatures (< 100◦C) will be slower than the deposition rate and a liquid film will form on the surface. This layer will slowly dry, however it will still contain many organics and probably cracks.115 In process B, the droplets evaporate before reaching the surface and a precipitate strikes the substrate where decomposition occurs. In process C, the solid precipitate melts and vaporises (or sublimes) and the vapour diffuses to the substrate and undergoes a reaction there. This corresponds to true CVD. At higher temperatures (process D), the vapour undergoes a chemical reaction before impinging upon the substrate. The droplets in the aerosol have formed solid particles that stick to the surface of the substrate. The product information sheet on DuPont’s “TYZOR” TPT notes that TPT pyrolyses at temperatures greater than 350◦C, and that films deposited in this method at 500 − 600◦C are considerably harder than films produced by hydrolysis and contain no organic residue.252 VXEVWUDWH ILQHO\GLYLGHG VROLGSURGXFW YDSRXU SUHFLSLWDWH { { { { GURSOHWV $ % & ' /RZWHPSHUDWXUH +LJKWHPSHUDWXUH Figure 3.7: Pyrolysis decomposition as a function of temperature (adapted from Vigui`eand Spitz 244). 253 The decomposition of TPT (by pyrolysis) to form TiO2 proceeds as follows: Ti(OC3H7)4 −→ TiO2 +2C3H7(OH) + olefins , (3.3) 3.5 TPT: The TiO2 Precursor 73 For the reaction of TPT to form TiO2 by hydrolysis, the reaction product will be strongly dependent on the amount of water vapour present in the system, as well as the substrate temperature. Wong et al. found that for APCVD-deposited TiO2, there is not enough oxygen to form TiO2 and the TPT remains unreacted on the wafer when less than 30% relative humidity exists.204 The reason why the TPT did not react by pyrolysis in this ◦ case (Tdep =250C) remains unclear. When spraying in an environment with greater than 45% relative humidity, the TPT reacts before long before reaching the wafer and a powdery white deposit results.46, 202, 203 Wong et al. mentioned that laboratory scale CVD experiments showed it was possible to deposit films with 15% thickness uniformity at humidities above 45%.204 In general, the recommended range in relative humidity is from 30% to 45%, which results in a transparent homogeneous film being deposited.202, 203 In this case, the reaction rate is limited by the amount of TPT reaching the wafer,205 which decomposes by hydrolysis onto the heated wafer, according to the two-step hydrolysis/degradation reaction in Equation 3.4.210, 253 The “Tyzor” TPT product information sheet (Du Pont, Inc.) contains a complete explanation of all the steps involved in the hydrolysis reaction.252 In the information it is also noted that whether hydrous titanium dioxide (TiO2·H2O) or TiO2 itself is formed as the final product is dependent on the temperature and the rate at which water is added to the system. Ti(OC3H7)4 +2H2O −→ Ti(OH)4 +4C3H7(OH) Ti(OH)4 +4C3H7(OH) −→ TiO2 +4CH3CH(OH)CH3 (3.4) In Equation 3.4, one mole of TPT reacts with two moles of water vapour to form one mole of TiO2 and 4 moles of by-product (mostly 2-propanol). Practically however, the amount of water vapour will depend on the geometry of the reaction zone, and gas flow and exhaust 217 rates. TiO2 films deposited in an APCVD system required more than four times this amount of water vapour for the above reaction to occur.204 As the substrate temperature is increased, the deposition chamber, or environment, will heat up and the relative humidity will decrease. However this is not strictly of great concern, as the number of moles of water vapour going into the system will remain constant, and this is what is required. The amount of TPT and water vapour required for the reaction to occur, as given in Equation 3.4, can be calculated. The molar mass mm of TiO2 is 79.9 g/mol. The number of moles M 2 −6 3 of TiO2 to cover a 25 cm surface area with a 70 nm thick film (volume, ν =175× 10 cm ) is given by the following equation, where the density ρ of rutile is 4.26 g/cm3 ρν M = mm −6 =9.3 × 10 moles TiO2 (3.5) −6 −6 This means we require 9.3 × 10 mol of TPT and 18.6 × 10 mol of H2O for the reac- tion in Equation 3.4 to occur. The molar mass of TPT is 284.26 g/mol and its density is 74 3. TiO2 Thin Film Deposition Equipment 0.955 g/cm3.250 Rearranging Equation 3.5 we obtain Mm ν = m ρ =2.8 µl of TPT to cover the 25 cm2 area with a 70 nm thick film. (3.6) The amount of water vapour required has been estimated at being four times greater again due to unreacted water vapour being extracted out the exhaust.204 However, in the system constructed for this work, many times this amount (approximately 0.5 − 1 ml) of TPT is needed to be sprayed as the film density is lower than that of single crystal rutile and not all the TPT sprayed is deposited within this area. This is especially true for shallow deposition angles, where the majority of the TPT passes across the surface of the wafer. Based on the requirement of 1 ml of TPT being required for each 5 cm × 5 cm wafer we can expect that at least 4 ml of water vapour would be required. 3.6 Design of Ultrasonic Spray Deposition System 3.6.1 Selection of Ultrasonic Nozzle Two ultrasonic atomizing nozzles (UAN) were purchased from Sono-Tek Corporation (Mil- ton, N.Y., U.S.A.). There were several reasons for selecting an ultrasonic nozzle, rather than an ultrasonic nebulizer. Firstly, with a nozzle the spray could be directed at different angles. Secondly, it was believed that a spray system where the reaction mechanism could be altered with the drop size could be tuned to behave similarly to a belt-furnace APCVD system204 as used in the PV industry. Thirdly, the potential of a solely spray-based processing system for fabricating solar cells was attractive. As well AR coating deposition, spray systems have been used for depositing passivation layers for solar cells.208 The two nozzles operated at frequencies of 48 kHz and 120 kHz, respectively. One motivating factor for nozzles manufactured by Sono-Tek was that only one broadband ultrasonic gener- ator (BUG) was required to operate any of their nozzles. In contrast, other manufacturers require that a separate BUG be purchased for each nozzle. The 48 kHz nozzle was selected in order to evaluate the system performance at high flow rates, and also to determine whether large droplets would be prevented from entering the 20 − 30 µm wide grooves on the front surface of the wafer. The 120 kHz nozzle was selected because it was able to tolerate the lowest flow rates. Later on in the project the 48 kHz nozzle was exchanged for a newly released nozzle called the MicroSpray. This nozzle used a 120 kHz atomizing frequency as well, however it was fitted with a micro-bore tube which significantly reduced the inside diameter of the nozzle. This enabled extremely low flow rates to be tolerated, while still achieving acceptable spray plumes. Figure 3.8 illustrates the 48 kHz standard nozzle and the 120 kHz MicroSpray nozzle purchased from Sono-Tek. The 120 kHz standard nozzle is 3.6 Design of Ultrasonic Spray Deposition System 75 similar in appearance to the MicroSpray nozzle, except that it possessed a conical tip (see Figure 3.8). Figure 3.8: The 48 kHz (left) and 120 kHz (right) ultrasonic nozzles pur- chased from Sono-Tek Corp (adapted from Sono-Tek Corp.246). The front and rear horns of the nozzles are manufactured from titanium alloy, Ti-6Al-4V, while the nozzle housing and liquid inlet are made from 316 stainless steel. The titanium alloy is chosen for is high mechanical strength, good acoustical properties and excellent chemical resistance.245 Most chemicals, except hydrofluoric and sulphuric acid and strong oxidizing agents, are compatible with these spray nozzles. Table 3.1 provides technical details on the Sono-Tek MicroSpray ultrasonic atomizing nozzle as well as the standard 120 kHz and 48 kHz nozzles. The dimensions in Table 3.1 correspond to the schematic of the nozzle shown in Figure 3.9. Figure 3.9: Dimensions of the Sono-Tek ultrasonic atomizing nozzles that correspond to the values given in Table 3.1 (adapted from Sono-Tek Corp.246). 76 3. TiO2 Thin Film Deposition Equipment Table 3.1: Sono-Tek ultrasonic atomization nozzle data (from Sono-Tek Corp.246). Note that dimensions are not exact as they are converted from inches. So few experiments were performed with the 48 kHz nozzle that some parameters are unknown and are labelled in the table as not applicable (n/a). Nozzle Type Microspray UAN UAN Parameter UAN Atomisation frequency (kHz) 120 120 48 Orifice diameter (mm) 0.38 1.32 2.18 Maximum flow rate (ml/min) 2.4 21 72 Recommended minimum flow rate (ml/min) 0.48 4.2 14.4 Actual minimum flow rate (ml/min) 0.02 0.35 n/a Atomization power at minimum flow rate (W) 1.1 − 2.0 1.1 − 2.0 n/a Median drop diameter (µm) 18 18 38 Weight (g) 196 309 Dimensions: A1 (mm) − 5.8 11.7 B1 (mm) − 11.2 26.9 A2 (mm) 2.5 − − B3 (mm) 11.5 − − C (mm) 29.3 25.4 37.3 D (mm) 36.6 36.6 38.1 E (mm) 12.7 12.7 42.9 F (mm) 10 8.6 − 3.6.2 Ultrasonic Nozzle Performance It was quickly determined that the 48 kHz nozzle would be unsuitable for our application. This was due to the extremely high flow rates, which caused a much thicker film than desired to be deposited in less than one second and was difficult to control. The 120 kHz proved to be better with spraying times in the order of 1 min for a 70 nm thick film at the lowest flow rates. Both of these nozzles possessed a conical tip, as shown in the left hand side of Figure 3.9, which is designed to spread the spray plume out over several inches. This tip was initially favoured as it was believed that this coverage would be sufficient to coat a 2” wafer. However, to achieve longer deposition times, the lowest flow rates had to be used. Although a fine mist often emerged from the nozzle the low volumes being pumped were not sufficient to create a plume of several inches in diameter. Occasionally, the nozzle would also “stall” and spit larger droplets of liquid onto the substrate. This resulted in 1 − 30 µm diameter TiO2 particulates being incorporated into the film. These defects reduced the chemical resistance of the films. To improve the film uniformity, reduce particulate incorporation, and to achieve lower flow 3.6 Design of Ultrasonic Spray Deposition System 77 and film deposition rates the 48 kHz nozzle was exchanged for a 120 kHz MicroSpray nozzle. This nozzle had a much narrower bore enabling very low flow rates could be used (down to 0.02 ml/min). Additionally, the tapered tip (see tip on right-hand side of Figure 3.9) resulted in a much finer line of spray. Achieving a consistent spray was aided by the use of the small-bore 1 ml gas-tight syringes (see Section 3.6.3). This setup resulted in a deposition time of 1 − 2 min and the best performance of all configurations tried for the spray system. The power required from the broadband ultrasonic generator (BUG) to atomize the TPT with the 120 kHz nozzles was in the order of 1.5W. 3.6.3 Liquid Delivery Pump Selecting the correct pump is a crucial step in designing the spray deposition system. This is because the ultrasonic nozzle will atomize any liquid that reaches the atomizing surface. A syringe pump (Yale YA-12, Kent Scientific, U.S.A.) was chosen primarily because it can pump continuously in the µl/min range and pulsed volumes of nanolitres can be achieved.245 The upper limit on the pumping rate and the dose is limited by the maximum syringe size, in this case 60 ml. The pumping action is very smooth, which is important for achieving good spraying. The pump is programmable and can be controlled via TTL signals or an RS-232 port. In this work it was controlled via the front panel only. This enabled a flow rate, typically 0.02 − 1.00 ml, to be set and the pumped volume to be monitored on the display. It is also possible to inject or withdraw with the pump. The ability to withdraw liquid was used in conjunction with a three-way valve in order to refill the smallest (1 ml) syringes (refer Figure 3.11). Figure 3.10: The Yale YA-12 syringe pump. 78 3. TiO2 Thin Film Deposition Equipment Syringes, Tubing, Valves and Bottles Initially experiments were performed with disposable 5 ml and 10 ml plastic syringes with rubber-tipped plungers. By observing the spray plume it was noticed that the plunger was periodically sticking as it travelled down the syringe barrel. It was discovered that the ability of a syringe pump to dose at consistently low flow rates is influenced by the diameter of the syringe. Therefore, several high-quality 1 ml and 2.5 ml gas-tight syringes equipped with Teflon plungers were purchased from a chromatography supplier (SGE, Melbourne, Australia). These performed excellently and could be flushed with dilute HF to remove any build-up of TiO2. All the syringes used were of the Luer-lock variety to ensure an air-tight and leak-proof connection. A variety of chemically resistant plastic (HDPE) fittings were purchased (Chro- malytic Technology, Victoria) to adapt from Luer-lock to barbed fittings for 1/8” outer diameter (1/16” inner diameter) tubing. The 1/8” plastic tubing (LDPE) then connected directly to the 1/8” Swagelok liquid inlet fitting on the Sono-Tek nozzle. As mentioned previously, a Teflon-lined gas-tight three-way valve (SGE, Australia) was later installed to permit withdrawal of new TPT precursor from a 100 ml screw-top chemical bottle. This was convenient for the 1 ml syringes since coating a 2”diameter wafer with a 70 nm thick TiO2 typically required 0.5 ml of TPT. The gas-tight syringes were able to screw directly into the valve, however it was necessary to purchase Kalrez fittings (SGE, Australia) to connect the 1/8” tubing. Figure 3.11 indicates how the Schott chemical bottles were modified to contain TPT. Initially a hole was drilled in plastic lid and a thread tapped into the hole. A female Luer adapter was screwed into the lid, sealed with a Viton o-ring, and the 1/8” tubing extended down into the bottle. A Teflon solvent filter (SGE, Australia) was pressed onto the end of the tubing, and this rested about 5 mm above the bottom of the bottle. This was to remove as many particulates as possible from the TPT. Particulates can slowly form due to the bottle being opened and closed, and also if it is not totally air-tight. 3.6.4 Substrate Heater The previous substrate heater for the pneumatic spray system at UNSW consisted of a stain- less steel block that was placed on a 2 kW stove element.203 This was less than ideal as the majority of heat generated did not pass into the block. Therefore, a new block was designed. Stainless steel was still used as it was anticipated that the block would need to withstand temperatures of 500◦C or more. Other materials considered included copper, graphite, nickel, molybdenum and titanium. Copper is an undesirable material for semiconductor devices due to the speed that which it can diffuse through silicon. The remainder of the materials were 3.6 Design of Ultrasonic Spray Deposition System 79 2' (QG /'3(WXELQJ FDS 0DOH/XHUWR EDUEDGDSWHU )HPDOH/XHUILWWLQJ VFUHZHGLQWROLGZLWK 9LWRQ IOXRULQDWHG UXEEHU RULQJ 2' /'3(WXELQJ 6ROYHQW )LOWHU Figure 3.11: The creation of a filtered TPT reservoir from a chemical bottle. ruled out because of expense. Therefore, a 150 mm×150 mm×20 mm stainless steel block was milled out for use as a substrate heater. Figure 3.12 shows the workshop drawings of the block, including precisely milled holes for the six heater cartridges, thermocouple and vacuum line. Heater cartridges were selected as they would result in the best power to heat conversion. The heater cartridges chosen were a split design (Dalton Electric Co., MA, U.S.A.) that expand upon heating to ensure good thermal contact to the block (see Figure 3.13). This avoided the use of a copper thermally-conducting paste. The power rating of each cartridge was 250 W and could be run off mains power (240 VAC . The cartridges were custom designed so that a 20 mm length near the leads was not embedded into the block. This cool-zone was important for keeping the shielding around the leads from burning or melting and possibly shorting out. Double layer fibre-glass/ceramic sheets (6 mm thick each) were placed on five sides of the block, and were contained by a stainless steel box with 1 mm thick walls. The insulation was designed to prevent too much heat from being transmitted downwards, as this could warp the rails of the driving mechanism below. A 3 mm diameter K-type thermocouple was inserted into the block. The tip of the thermocouple was located underneath the centre of the wafer. The output from the thermocouple was fed into a BTC-9090 PID temperature controller. The output of the PID controller to the six heater cartridges was via a 240 VAC 10 A relay. As the heater cartridges could draw a maximum of 1.5 kW there was no risk of 80 3. TiO2 Thin Film Deposition Equipment 3 BACK Hole for thermocouple 3mm 20 VIEW diameter, 70mm deep 150 5 75 3 45 35 25 15 Vacuum 3mm diam. hole channels: for thermocouple 1mm wide, 75 1mm deep 150 TOP VIEW 5 3mm diam. vacuum line 1/8” NPT 10 thread 50 25 6x holes for heater 25 cartridge = 0.377” 50 50 3 10 FRONT 0.377” VIEW 20 Figure 3.12: Schematic diagram of the stainless steel heater block. 3.6 Design of Ultrasonic Spray Deposition System 81 Figure 3.13: Six of these 6” long 250 W Dalton Watt-Flex heater car- tridges were embedded in the stainless steel block to ensure maximum heat transfer.254 blowing the fuse on the output side of PID controller. The block took about 15 min to reach 450◦C from room temperature, and typically temperature fluctuations while spraying were not more than 5◦C. 3.6.5 Motorized Stage A motorized translation stage was designed so that the substrate could pass back-and-forth beneath the spraying zone. The 12 VDC motor (RS Components, Sydney) had a continuous torque rating of 300 mNm and a maximum torque rating of 600 mNm. At 12 VDC the motor speed was 220 rpm. The motor was directly connected to a double-start threaded spindle with a 2 mm thread pitch. This spindle passed through a Teflon nut that was mounted on a plate connected to the underside of the heater block. The significant mass of the heater block (about 6 kg) was supported by two 12 mm diameter stainless steel rails. These 50 cm long rails were machined precisely to accommodate four linear bearings. Once correctly aligned, the block slid effortlessly along the rails. Thus, by applying either a positive or negative voltage to the motor the spindle turned and moved the block forwards or backwards. At 220 rpm the block travelled at a speed of 7 mm/s. As it was necessary to keep the door to the spray chamber shut during depositions (see Section 3.6.7), a simple circuit using a double-pole change-over (DPCO) 12 VDC relay was implemented in order to reverse the direction of the stage. The relay has two pairs of inputs. The first pair consisted of +12 VDC and -12 VDC, while the second had the polarity switched. Two sealed and chemically resistant momentary push-button micro-switches (RS Components, Sydney) were used to trigger the state of the relay. Pressing one switch caused the relay to “change-over” from its first input to its second input pair, while pressing the second switch caused the relay to go back to its initial state. Thus, the stage started moving in one direction (the direction it was last travelling) until reaching the end of the rails upon which the microswitch was depressed, and the stage changed direction. A fuse was also included in this simple circuit to prevent the motor from burning out. A motor speed controller was also added to the system to enable slower translation of the stage, however this was rarely used. A 2 A 12 VDC power supply was sufficient to power all the electrics. Finally, in order to keep the rails free from TiO2 dust, concertina-style rubber boots were fitted over the rails. These boots could be compressed to about 25% of their standard length. 82 3. TiO2 Thin Film Deposition Equipment 3.6.6 Spray Shaping Spray shaping and directional control was necessary for several reasons. In early experiments it was found that when the nozzle was placed directly above the substrate heater the fine, low-velocity (about 10 cm/s) mist did not fall on the wafer. This was because of hot air rising from the 450◦C block and carrying the spray droplets away with it. Experiments were performed with the two “air shrouds” that were supplied with the nozzles (Sono-Tek Corp.). This enabled the spray to reach the heated substrate, however there were many particulates in the film and the thickness uniformity was very poor. After discussing our requirements with Sono-Tek Corp., a loan of a “vertical spray assembly” was arranged. The vertical spray assembly, depicted in Figure 3.14, is designed to produce wide spray patterns. The two streams of gas are slightly off-centre, resulting in the shearing of spray plume.245 Wide spray patterns were able to be generated with the vertical spray assembly, however these were far from uniform. Additionally, whatever nozzle-substrate distance was used resulted in the incorporation of many particulates into the film. It is believed that the TiO2 particulates arise from droplets reacting with remnant humidity in the air, entrained into the nitrogen gas stream. Since the nozzle is directly above the wafer any particulates formed will fall onto the wafer and be incorporated into the film. As a solution, an “air-knife”, sometimes also called an air-guide, was purchased (Exair Corp., U.S.A.). The air-knife was 6” long and consists of two halves of an aluminium casing bolted together. There is an 1/4” NPT fitting at one end for the nitrogen inlet. At the front there is a small slit running almost the whole length of the air-knife, which creates a high-velocity sheet or curtain of nitrogen. The slit height is set by a plastic shim inside the air-knife. It was found that by intersecting the emerging spray plume from the nozzle with a sheet of nitrogen from the air-knife enabled the deposition of a visually acceptable film on the wafer. As the 6” wide curtain cooled the substrate significantly a new shim was inserted, which had created a slit with 4 mm width and 0.02” height. This reduced the amount of cooling of the block significantly. 3.6.7 Miscellaneous Equipment Relative Humidity Sensor As discussed previously in Section 3.5.2, the reaction of TPT to form TiO2 is extremely sensitive to the relative humidity. Relative humidity RH is defined as the partial pressure p of water vapour in air divided by the vapour pressure of water ps at the same temperature. To be able to monitor the relative humidity in the spray system, a sensor and digital meter were purchased (Elan Technical Corp., CT, U.S.A.). 3.6 Design of Ultrasonic Spray Deposition System 83 Figure 3.14: Sono-Tek’s vertical spray assembly, designed for generating wide spray patterns (from Sono-Tek Corp.246). Oxygen Concentration Sensor An oxygen sensor was added to the system because it was anticipated that spray depositions with TPT diluted in a solvent, such as isopropanol, would be performed. Although the volumes of solvents sprayed would be less than 1 ml, ensuring that the oxygen concentration was low would inhibit the combustion of the solvent. Therefore an oxygen sensor (Electrovac GmbH, Austria) was added to the system. As shown in Figure 3.15(a), when a voltage is applied across the zirconia electrolyte cell, oxygen is pumped through the cell because oxygen ions carry the current through the cell. By attaching a cap with a pinhole on the cathode side of the cell and increasing the voltage, the current becomes saturated due to limited transfer of oxygen ions to the cathode. This current is proportional to the ambient oxygen concentration.255 The advantages of this sensor included having a linear output signal, no cross-sensitivities to other gases, long life, and a very low temperature dependence of the 84 3. TiO2 Thin Film Deposition Equipment signal. Due to a heater inside the package the sensor required the application of 2 V while warming up (30 s) and then 4 V for constant operation. Figure 3.15(b) shows an image of the oxygen sensor. Designing this power supply and installation of the sensor and a suitable analogue meter was completed by a German practicum student, Manfred Fahr. Figure 3.15: (a) Schematic indicating oxygen sensor operation, and (b) Image of the oxygen sensor (adapted from Electrovac GmbH 255). Spray Chamber The spray chamber was fabricated from steel “speed-frame” and clear perspex sheeting, with internal dimensions of 70 ×70 ×70 cm. The entire front side of the chamber was hinged along one edge and could be opened to provide full access to the chamber. A series of 25 mm diameter holes were drilled in the door and covered with a sliding plate. This was in order to allow the entry of ambient air after spraying was completed. The top of the chamber was connected to a 6” diameter exhaust line. The exhaust flow could be controlled via a large butterfly valve. Nitrogen Heater Due to the relatively high gas flow rates used (see Section 3.6.8) significant cooling of the block occurred. With the block temperature set at 450◦C the top surface of the wafer was measured to be about 330◦C while spraying. In order to reduce the amount of cooling occurring it was decided to control the temperature of the nitrogen gas stream. The nitrogen heater used a heater cartridge that was inserted inside a length of 3/8” stainless steel tubing. It was designed to operate upright so that if the nitrogen gas flow stopped the hot air for around the heater would rise up to the top of the tube where a thermocouple was placed. The system was designed to operate at temperatures up to 400◦C, being controlled by a BTC-2020 PID controller (ECE Fast, Melbourne). 3.6 Design of Ultrasonic Spray Deposition System 85 Air-Knife Heater Although the nitrogen heater worked well, a with heated lines the temperature of the emerg- ing gas had dropped only 30◦C after passing through a 1 m length of tubing (with the heater set to 200◦C). However, when connected to the aluminium air knife, which acted as a very efficient heat sink, the emerging gas temperature was only a few degrees above ambient temperature. Therefore a 240 VAC 100 W heater pad was obtained (RS Components, Aus- tralia) that was both flexible and had a high-temperature adhesive applied to one side. This was carefully folded around the air-knife and firmly held in place. This permitted heating (not-controllable) of the air-knife up to temperatures of 190◦C, which enabled the heated nitrogen gas to reach the substrate. 3.6.8 Operation of the TiO2 Spray System In its final form, the spray system incorporated all of the above componentry, as well as regulators and flow meters to accurately control gas flows (as shown in Figure 3.1). All the films deposited using this deposition system were formed by spray pyrolysis, due to the very low relative humidity. The diagram in Figure 3.16 indicates the necessary steps to set-up and perform depositions with the TiO2 spray system. Table 3.2 in Section 3.7 lists the various deposition parameters of the system. The gas flow rates listed were optimised in order to extend the film coverage in the forward direction, while trying to reduce the amount of cooling of the substrate heater. At lower flow rates, the film was not deposited onto the wafer due to heat arising from the 450◦C block, while little benefit was achieved by using higher flow rates. A typical atomisation power of 1.5 − 2.0W was required from the BUG for flow rates between 0.02 − 0.10 ml/min. Spray depositions were only performed at the maximum system operating temperature of 450◦Casatlower temperatures the number of particulates in the film was unacceptable. It should be noted that the deposition temperatures quoted throughout this work may be slightly higher than the actual deposition temperature due to the cooling of the wafer by the gas flow. The deposition time and the efficiency at which the TPT is converted to TiO2 is greatly influenced by the alignment of the system. The height of the nozzle and whether the tip sits directly in the nitrogen flow from the air-knife or just above it is important. The best results were achieved by adjusting the height of the nozzle with the nitrogen flowing, and when it could be heard that the tip just entered the gas stream the nozzle was fixed in that position. The optimum angle for spraying was determined to be 5◦ below the horizontal. At this angle the nitrogen emerging from the air-knife travels up towards the block and then adheres to the top surface of the block as it passes across it. The adherence of a gas to a solid surface is called the Coanda effect. In general, spraying at angles closer to the horizontal reduced the number of particulates observed in the film, and also enabled only the top surface of 86 3. TiO2 Thin Film Deposition Equipment Check that tip of spray nozzle is free TiO2 powder Install a new 0.5 m length of LDPE 1/8" OD tubing between three-way valve and nozzle Withdraw 1 ml of TPT into gas-tight syringe Syringe pump: set syringe diameter and desired flow rate Purge chamber with N2 until RH ≈10% Wait for substrate heater to reach 450°C Inject TPT to fill "dead-space" in the tubing, valves and nozzle Turn on motor to move stage away from centre and place wafer on substrate heater Using three-way valve, withdraw TPT from bottle into syringe To begin spraying: adjust power on BUG to 1.2 W, start TPT flow, and turn on air-knife N2 Leave motor off momentarily as there are often a few "spits" more wafers at the start of spraying y ra p Turn on motor and monitor colour of TiO2 film on wafer - stop s To when it appears dark blue Turn off syringe pump, motor and BUG, and remove wafer. Open exhaust valve and purge system for 1 min. To finish spraying, turn off heater block, remove nozzle and syringe and rinse with IPA, discard tubing in waste container Figure 3.16: Diagram indicating the necessary steps to set-up and per- form depositions with the TiO2 spray system. 3.7 Design of CVD System 87 groovedwaferstobecoated. The order in which the various components are turned on and off are important. The tubing, valve and nozzle should be filled with TPT before starting. Any excess TPT can be wiped off the tip of the nozzle using a tissue and some isopropyl alcohol (IPA). The stage should be positioned so that the initial spray plume will not land on the wafer. This is because of a tendency for there to be more particulates in the initial spray plume. Secondly, the air-knife flow should be started and the BUG should be turned on. After a few seconds of spraying the motor can be switched on and coating of the wafer will begin. To finish, it is best to turn off the BUG first, to stop he atomisation process as quickly as possible, then the syringe pump, and finally the motor and air-knife flow. For the deposition of doped TiO2 films (as in Section 6.3 of this work) the dopant liquid was mixed together with the TPT in a small chemical bottle in a fumecupboard before spraying. If many doped films, especially of different doping concentrations, were required it would be recommended to replace the three-way valve with a four-way valve and to have separate pure bottles of TPT and the dopant liquid. The syringe should be shaken to ensure good mixing of the two liquids. 3.7 Design of CVD System 3.7.1 Motivation Although the spray pyrolysis system produced quite dense anatase films, the thickness uni- formity on a macroscopic level was poor. On a scale of nanometres to micrometers the films were quite uniform, however, on a scale of millimetres to centimetres, thickness variations of a similar magnitude to the average film thickness were observed. This meant that, for a 70 nm thick film, there were many points across the wafer that remained virtually uncoated. This, along with the incorporation of particulates, typically 1 − 30 µm in diameter, had se- rious implications for chemical resistance and the electroless metal plating process. Other limitations included only being able to operate the spray deposition system at the maximum temperature of 450◦C, in order to limit the number of particulates in the film. 3.7.2 TPT Bubbler and Temperature Control Therefore a simple CVD system was designed to replace the spray system. Attachai Uer- anatusan (MEngSc, UNSW) had performed initial investigations into a CVD nozzle that was fed with a vapour from a glass bubbler. This idea was adapted by the author into the existing structure of the spray system. A one-litre stainless steel bubbler was purchased (Meriter, U.S.A.) for safety as it could handle pressures up to 42 psi, much greater than the 88 3. TiO2 Thin Film Deposition Equipment quartz bubbler. The use of electropolished stainless steel bubblers have been successfully demonstrated in the semiconductor industry.256 It was found that there was essentially no assay degradation when comparing the performance of quartz and stainless steel bubblers (J.C. Schumacher, CA, U.S.A.) were the same and as long as moisture was prevented from entering the bubbler. This is achieved by using metal gaskets instead of Teflon. A pressure relief valve with a threshold of 35 psi was placed in parallel with the bubbler to prevent too high a pressure being applied to the bubbler. The outlet of the pressure relief valve went directly into the exhaust system. A Schumacher temperature controller and temperature control unit were used to maintain the bubbler temperature at 50◦C. At this temperature, the TPT has a vapour pressure of 1 mbar. The Teflon tubing leading from the bubbler to the nozzle was heated using a 1 m length of 240 VAC heater tape with a power of 90 W. This was insulated with glass- fibre insulation tape, held in place with high-temperature heat-shrink tubing. This was to prevent condensation of the TPT onto any cool surface so as to avoid the build up of TiO2 particulates in the tubing and nozzle. It was not necessary to hear the nozzle itself as its close proximity to the heater block ensured that it maintained a temperature well above 50◦C. 3.7.3 Water Vapour Bubbler After initial success with TiO2 films deposited using the CVD system, a second bubbler containing de-ionised (DI) water was added to the system. This was designed to permit reactions by hydrolysis. This was a standard quartz bubbler typically used for performing wet oxidations in a tube furnace. A separate nitrogen regulator was used to limit the pressure that could be applied to the bubbler to 5 psi. The bubbler had a matching base with an integrated heater. The temperature was monitored with a thermometer and manually maintained at approximately 100◦C. A 3-way valve was attached to the outlet of the water bubbler so that the water vapour could be ”switched-off” by redirecting the flow to an empty flask. Figure 3.17 shows a diagram of the final CVD system. 3.7.4 Operation of the CVD System Operation of the TiO2 CVD system was quite similar to the spray system. About three hours before depositions were performed it was necessary to switch on the temperature controllers for the TPT and water bubblers. A timer was implemented for doing this so that CVD depositions could begin early in the morning. Depositions could now be performed at substrate temperatures of 150 − 450◦C. Below 150◦C the film coverage became very non- uniform, probably due to the TPT striking the substrate as a liquid. As with the spray system the air-knife angle and nozzle position were critical. It was found that the optimum 3.8 Conclusions 89 angle for the air-knife was about 2◦ above the horizontal, and the height on the nozzle was adjusted in the same manner as previously described in Section 3.6.8. Again, the gas flows listed in Table 3.2 are the minimum flows for successful TiO2 deposition to occur, while trying to reduce the amount of cooling of the block occurring. Table 3.2: Deposition conditions for the USD and CVD TiO2 films Process parameters USD CVD Liquid TPT flow rate (cm3) 0.02 − 0.10 n/a TPT bubbler flow rate (scm3) n/a 1700 Air-knife N2 flow rate (slpm) 5 13 Atomisation power (W) 1.5 − 2.0 n/a Chamber pressure (kPa) ≈ 101 ≈ 101 TPT temperature (◦C) 25 50 Substrate temperature (◦C) 450 150 − 450 Nozzle-substrate distance (cm) 5 − 7 3 − 5 Relative humidity (%) < 10 < 10 Deposition angle (◦) −5 ≈ 2 Deposition time (min) 1 − 2 10 − 20 3.8 Conclusions A good understanding of basic TiO2 material properties along with knowledge of a previous TiO2 spray deposition system at UNSW, formed the requirements for the TiO2 thin films in this application. Films were required to be dense and defect-free, exhibit good thickness uniformity, possess a high refractive index and low optical absorption, and be insulating. Examination of deposition techniques described in the literature lead to the author design- ing and constructing an ultrasonic spray deposition (USD) system. USD offered several many potential advantages, including shallow-angle depositions, low deposition rates, and the ability to deposit for a wide range of thin films via this method using different liquid precursors. Many of the desired TiO2 film properties were obtained from films deposited using the USD system. However, the limitations of the USD system were, firstly, that the thickness uniformity of these films was poor, secondly, that large particulates were commonly embedded into the thin films (these films will be characterised in detail in Chapter 4). Spray depositions were performed at 450◦C as a dramatic increase in the number of particulates was observed at lower temperatures. Thus, although the films exhibited a high refractive index, the ability to tune the refractive index by varying the substrate temperature could not be realised. This lead to the design and construction of a simple chemical vapour deposition (CVD) system that could operate at atmospheric pressure. TiO2 films deposited with this simple CVD system are used throughout the majority of this work. 90 3. TiO2 Thin Film Deposition Equipment PID contoller temperature contoller temperature TPT bubbler C C C C q q q q 445 450 49.9 50.0 DC 12 V 2 N line switch (x2) switch thermo -couple vacuum st. steel block relay housing st. steel bubbler TPT heater rings vacuum motor + 1/8" st. 1/8" steel nozzles 2 TPT N heaters lines cartridge heated O + 2 2 H N 2 N O 2 r heate quartz H quartz air-knife bubbler and bubbler 2 N Figure 3.17: TiO2 CVD deposition system. The humidity sensor, exhaust valve on TPT bubbler, and the 3-way valve on the water bubbler have been omitted for clarity. Chapter 4 Characterisation of TiO2 Thin Films Extensive characterisation of TiO2 films deposited using ultrasonic spray deposition (USD) and chemical vapour deposition (CVD) was performed in order to determine the physical, optical, electrical and chemical properties of the films. All films deposited at 450◦C were of the anatase phase. A surprising result was that USD anatase films did not convert to rutile ◦ 3 after lengthy annealing at 950 C. USD TiO2 films were found to be dense (3.64 g/cm ) and exhibited a high refractive index (2.45 at 600 nm), ideal for acting as an antireflection (AR) coating on a glass encapsulated silicon wafer. Although the USD films appeared continuous over a microscopic level, macroscopically the films exhibited large variations in thickness and the occasional pinhole. TiO2 films deposited using CVD were found to exhibit much lower surface roughness and better thickness uniformity, although the density was significantly lower than that of the USD-deposited films. Shallow-angle depositions were successful in maintaining the grooves free of TiO2. However, shallow-angle depositions were not successful on textured crystalline silicon wafers, with large tree-like structures growing at the tips of the pyramids. The chemical resistance of all films was excellent against acids, but relatively poor against alkaline solutions, although the etch resistance did improve upon annealing of the TiO2 film. 4.1 Introduction The literature review in Chapter 2 described the many different material phases of TiO2, and the variation of the material properties with those phases. The front-surface dielectric film in the buried-contact (BC) solar cell fabrication sequence is subjected to a number of high-temperature processing steps in oxygen, nitrogen, and phosphorus-containing furnace ambients. Therefore, before replacing the silicon dioxide (SiO2) layer in the original BC solar cell with a TiO2 film, it was necessary to perform extensive characterisation of the TiO2 thin films to, firstly, optimise the deposition parameters to obtain a TiO2 film exhibiting the 91 92 4. Characterisation of TiO2 Thin Films desired qualities and, secondly, to determine the behaviour of the films under typical BC fabrication sequence conditions. Different characterisation techniques were utilised in order to determine the physical, optical, chemical and electrical properties of the as-deposited and annealed TiO2 thin films, deposited using ultrasonic spray deposition and chemical vapour deposition (CVD). Characterisation techniques can be roughly divided into the following four categories: • Physical Properties: The crystalline phase of the film and the existence of oxygen vacancies were determined using Fourier-transform infrared (FTIR) and Raman spec- troscopy. The surface roughness of the films was measured using atomic force mi- croscopy (AFM). The presence of particulates and the density of the films was ob- served using scanning electron microscopy (SEM). SEM work also provided a rough indication of the grain size of the material. Elemental analysis of film was performed using both X-ray photoelectron spectroscopy (XPS) and Rutherford back-scattering (RBS) spectroscopy. • Optical Properties: The refractive index and extinction coefficient of the films were measured using spectroscopic ellipsometry (350 − 1150 nm), ellipsometry (633 nm), and reflectance measurements (300 − 1200 nm). • Electrical Properties: Conductivity observed in some TiO2 films that had undergone a reaction was determined using a four-point probe (FPP). The degree of surface passiva- tion afforded by TiO2 films and TiO2/SiO2 stacks was determined using the transient photoconductance decay (transient-PCD) technique (discussed further in Chapter 5). • Chemical Properties: The chemical resistance of TiO2 thin films were determined by placing films in various solutions commonly used in solar cell processing. A brief introduction to each of these characterisation techniques will be provided here and the results presented throughout this section. 4.2 FTIR Spectroscopy FTIR spectra provide information regarding the bonding between atoms in a sample, and are primarily used in the semiconductor industry to determine the existence of dopant or impurity atoms. The nature of the technique is very quantitative in identifying the impurity type, but very qualitative in determining the concentration of the impurity. FTIR spectra are also demonstrated in this work to be useful for understanding changes occurring in TiO2 films during high-temperature processing. For a discussion of the theory of FTIR refer to works by Nakamoto257 and Griffiths and de Haseth.258 4.2 FTIR Spectroscopy 93 FTIR measurements in this work were performed with a Nicolet 520 spectrometer in the range 250 − 5000 cm−1.A4cm−1 resolution was used and 256 scans were collected per spectrum. Any oxygen and carbon in the float-zone (FZ) wafers was accounted for by first performing a reference spectrum with only the FZ wafer. This reference spectrum was then subtracted from subsequent sample spectra to yield information about the TiO2 film. Erkov et al. published some excellent FTIR spectra of 110 nm thick, LPCVD deposited TiO2 thin films, as shown in Figure 4.1(a) and (b).117 The rutile films are deposited on Si wafers ◦ at 630 C using a TiCl4 precursor with N2O and H2 ambient gases. The spectra shown in Figure 4.1(a) are the as-deposited TiO2 film (curve 1), samples annealed in a vacuum (curve 2), structures annealed in N2O ambient (curve 3), and samples that had 10 nm of SiO2 grown prior to TiO2 deposition (curve 4). In Figure 4.1(b) the IR spectra of a bare Si wafer with 0.4 − 0.6 nm natural oxide (curve 5) and of sample 3 after etching off the TiO2 in a hot −1 H2SO4 etch (curve 6). The strong absorption peak at 608 cm is inherent to the rutile phase, however this is somewhat obscured by a weaker silicon absorption peak at 608−610 cm−1,as seen in curve 5. The overlapping peaks at 423 cm−1 and 460 cm−1 are also characteristic of −1 the rutile modification of TiO2. An additional absorption peak is observed at 470−480 cm for samples annealed in N2O (curve 3) and for structures with a 10 nm interfacial SiO2 layer −1 (curve 4). In the former case the peak at 480 cm is attributed to the formation of Ti2O3. This will be discussed in more detail in Section 5.2.2. The small peak at 515 cm−1 has been observed for TiO2 samples annealed in vacuum only (curve 2). The existence of a peak −1 at 1108 cm is indicative of the formation of SiO2 at the Si:TiO2 interface (curves 2, 3, 4 and 6). This peak is even observed in the vacuum (10−2 Torr) annealed samples where the availability of oxygen would be extremely limited. Figure 4.2 shows the FTIR spectra of two USD-deposited TiO2 films from this work (TO-11 and TO-12), over the range where the spectra of coated wafers differ markedly from the bare silicon reference wafer. It can be seen that at wavenumbers greater than 570 cm−1 the features observed in all spectra are similar and originate from the silicon substrate. A broad −1 absorption peak at 1083 cm is the only difference between the TiO2 coated wafers and the bare substrate. This peak is likely to be due to the formation of SiO2 at the TiO2:Si interface, and is normally observed at 1108 cm−1.117 When comparing these thin film FTIR spectra to reference spectra for bulk anatase and 259, 260 rutile TiO2 obtained from the literature it is immediately apparent that there is little agreement between the absorption peaks. Only a local transmittance peak at about 380 cm−1 can be observed in all samples. The spectra of the bulk TiO2 samples may differ greatly due to other incorporated impurities or the measurements may not have been obtained at room temperature. TiO2 samples TO-11 and TO-12 were loaded into the furnace under different conditions, and the effect of the ambient furnace gas will be discussed in detail in Section 5.2.2. The spectrum of TiO2 film TO-11 has been successfully modelled by the author using the polarisation-dependent model recently applied to single crystal anatase.261, 262 The model 94 4. Characterisation of TiO2 Thin Films Figure 4.1: (a) and (b) FTIR spectra of TiO2 thin films on silicon wafers. The spectra are of 1) an as-deposited TiO2 film, 2) TiO2 films annealed in a vacuum, 3) TiO2 films annealed in N2O, 4) TiO2 films deposited on 10 nm of SiO2, 5) a bare Si wafer with 0.4 − 0.6 nm natural oxide, and 117 6) the spectra of sample 3 after etching off the TiO2. presented in Equation 4.1 is based on the factorised form of the complex dielectric function ε (ν).261 ε2 − ε2 ıγ ε LOn + LOn ε (ν)=ε1(ν) − ıε2(ν)=ε∞ (4.1) ε2 − ε2 + ıγ ε n TOn TOn The longitudinal optical (LO) and transverse optical (TO) phonon oscillator frequency ν (cm−1) and damping γ (cm−1) values are given in Table 4.1. The complex dielectric constant is related to the complex refractive index by ε = n 2 (refer to Section 2.3.6 for more detail). A similar model has been published in the past for rutile.263 The sample structure used in the anatase model was a 300 µm thick polished silicon wafer with a 74 nm thick TiO2 film deposited onto both surfaces. The dielectric constant of the TiO2 film was determined using a Bruggemann effective medium approximation158 of 55.7% Ec-axis and 44.3% E⊥c- axis. The modelling was performed using the WVASE32 software package.62 The best fit 4.3 Raman Spectroscopy 95 obtained with the model is plotted in Figure 4.2 using the anatase values from Table 4.1. It can be seen that the location of the absorption peaks are in excellent agreement with the experimental results. This result is somewhat surprising as the as-deposited films had been subjected to 90 min of annealing at 950◦C, which should be more than sufficient to observe a transformation from anatase to rutile. The only similar result that can be found in the ◦ literature (for undoped TiO2) is for anatase films, deposited at 330 C, remained anatase after annealing at 850◦C.75 In order to confirm that only the anatase phase was present Raman spectroscopy measurements were performed on the films. Figure 4.2: Comparison of the FTIR spectra of spray deposited TiO2 thin films and the modelled anatase result obtained using Equation 4.1. 4.3 Raman Spectroscopy Raman scattering occurs when light interacts with the optical phonons in a material. If the photon gives up part of its energy to the lattice, in the form of a lattice vibration or phonon, the photon emerges with a lower energy. It is also possible for a photon to absorb a phonon and emerge at a higher energy, although these events much weaker. The theory of Raman spectroscopy is discussed in detail in the works of Long264 and Nakamoto.257 Since the intensity of Raman scattered light is very weak, the intense monochromatic beam from a laser is required. The collected signal is usually passed through a double monochro- mator and detected by a photodetector. The Renishaw Model 2000 machine located in the School of Materials Science at UNSW is equipped with three lasers, with wavelengths of 514.5 nm, 632.8 nm and 780 nm. The system uses a prism to disperse the signal onto a CCD 96 4. Characterisation of TiO2 Thin Films Table 4.1: LO and TO phonon frequencies for anatase and rutile to fit di- electric function in Equation 4.1 (adapted from Gonzalez,261 and Gervais and Piriou263). Anatase Rutile Mode Frequency Damping Frequency Damping ν (cm−1) γ (cm−1) ν (cm−1) γ (cm−1) E c-axis TO 367 68 172 76 (A2u) LO 755 79 796 38 E⊥ c-axis TO 262 36 189 27 (Eu) LO 366 4.1 367 10 TO 435 32 381.5 16.5 LO 876 33 443.5 21.5 ε∞(Ec-axis) = 5.41 ε∞(Ec-axis) = 7.8 ε∞(E⊥c-axis) = 5.8 ε∞(E⊥c-axis) = 6.0 array. While this makes measurements very fast it also limits the accuracy of the instru- ment to about 1.6 cm−1. An additional complication with the Renishaw machine is that the absorption filter designed to remove the laser line at 0 cm−1 begins absorbing at about 200 cm−1. This interferes with the dominant peak of anatase which lies at 143 − 144 cm−1, and rutile also has a smaller peak that lies at the same frequency. Figure 4.3 illustrates the 265 Raman spectra for the three crystalline phases of TiO2 as well as the amorphous phase. Figure 4.3: Raman spectra of the various phases of TiO2, obtained from powder samples.265 4.3 Raman Spectroscopy 97 It can be seen that the peaks from each TiO2 phase are clearly separated in frequency and therefore easily distinguishable. Table 4.2 presents the Raman peak assignments for single crystal anatase and rutile, and also silicon. Silicon is included as the absorption depth of even the most absorbing TiO2 films deposited in this work is just over 1 µmatλ = 514.5 nm. Since the films are typically 70 nm thick, photons interact with the silicon, which has a similar absorption depth to TiO2 at this wavelength, creating Raman scattering events. This is confirmed in the literature where the Raman peaks of silicon were still observed at 300 cm−1 −1 and 520 cm when performing measurements on 700 nm thick TiO2 films deposited onto 266 silicon wafers using a laser of 532 nm. Thus, it is not possible to observe the A1g and B1g modes of anatase for a thin film deposited onto silicon. Table 4.2: Raman active phonon frequencies for anatase and rutile (adapted from Ohsaka et al.267 and Gonzalez 261). Anatase Rutile Mode Frequency Mode Frequency ν (cm−1) ν (cm−1) Eg 144 B1g 143 Eg 197 Eg 447 B1g 399 A1g 612 A1g 514 B2g 826 B1g 514 Eg 639 Figure 4.4 shows the behaviour of the Raman spectra during annealing to temperatures for a fixed time of 40 min.265 At about 250◦C, the 141 cm−1 anatase peak appears and by 425◦C the conversion to anatase is complete.265 The anatase to rutile transformation is clearly observable at 800◦C, and at 1000◦C the conversion to rutile is complete. Raman measurements from this work are given in Figure 4.5. Note that the data for the 520 cm−1 silicon peak has been omitted for clarity. Samples 1 and 2 were deposited by APCVD at Eurosolare S.p.A., Italy. Sample 1, deposited at 320◦C, exhibits no anatase peaks. Silicon peaks at 300 cm−1 and were observed, and the peaks at 622 cm−1 and 834 cm−1 could possibly be attributed to the A1g and B2g modes of rutile. The A1g peak is typically very strong, while the B2g mode is very weak. Therefore it is concluded that the sample is predominantly amorphous with a very small fraction of rutile. With increased temperature (450◦C) anatase peaks emerge at 143 cm−1,396cm−1 and 637 cm−1. The location of these peaks is in excellent agreement with the bulk values given in Table 4.2. Samples 3 and 4 were both deposited at 450◦C using ultrasonic spray pyrolysis, and sample 4 received a subsequent 90 min anneal at 950◦C. These results confirm the earlier result that only the anatase phase is present after a lengthy high-temperature anneal. The broad peak at about 98 4. Characterisation of TiO2 Thin Films Figure 4.4: Raman spectra measured after annealing TiO2 powder sam- ples at various temperatures.265 950 cm−1 is attributed to silicon-oxygen-titanium (Si-O-Ti) bond formation.268 Note that a peak at 960 cm−1 is observed in silicon, however this is a multiple phonon event and it is 8 very weak. The Si-O-Ti peak indicates an interaction between the TiO2 layer and either 266 a thin or native SiO2 layer. This result is consistent with the FTIR finding that a SiO2 layer is formed at the TiO2:Si interface. Fitzgibbons et al. noted that TiO2 depositions on quartz substrates remained anatase after annealing at 1000◦C for 20 hr.67 Therefore, it is postulated that the presence of an interfacial SiO2 layer has resulted in the TiO2 maintaining the anatase preferentially. 4.4 X-ray Photoelectron Spectroscopy X-ray photoelectron spectroscopy (XPS) is primarily used to identify chemical species at the surface of a sample. When high-energy photons (X-rays) interact with atoms of the sample via the photoelectric effect, electrons are ejected from the core levels. However only photoelectrons ejected from atoms in the top 5 − 50 A˚ can escape and therefore be detected. The primary strength of XPS is that it allows chemical and not only elemental identification. Depth profiling is also possible by performing ion-beam sputtering of the sample. Further information on XPS theory and measurement techniques can be found in Briggs and Seah.269 The spectrometers used in this work were, firstly, a VG Scientific ESCALAB 220i-XL, using monochromated Al Kα (1486.6 eV) radiation, with an accuracy of 0.2 at. %. Figure 4.6 shows the XPS spectrum resulting from a wide energy scan of a spray deposited TiO2 film on silicon. This was a surface scan performed with no etching of the film. Apart from the expected titanium and oxygen peaks, a significant carbon peak is observed. This peak is primarily due to adsorbed carbon from the atmosphere, as the sample was stored in 4.5 Rutherford Back-Scattering Spectroscopy 99 Figure 4.5: Raman spectra of TiO2 thin films deposited by APCVD (sam- ples 1 and 2) and ultrasonic spray pyrolysis (samples 3 and 4). Anatase (A) and silicon (Si) Raman peak assignments are shown. Samples 1 and 2 were deposited at 320◦C and 450◦C, respectively, while samples 3 and 4 were both deposited at 450◦C. Sample 4 received a subsequent 90 min anneal at 950◦C. air. This adsorbed carbon had an atomic concentration of 13% at the top surface. As the film was etched this reduced quickly to below 1%, indicating that 1% was the level of contamination resulting from the organic precursor (see Section 5.3.1). The location of the carbon peak at 285 eV serves the useful purpose of providing a calibration point for the spectra, accounting for any charging effects.94 Windows around the titanium, oxygen, carbon and silicon peaks were drawn and examined in greater resolution. The O1s peak at 531 eV exhibits a slight shoulder at exhibited a shoulder at about 533 eV, and this has been attributed to the adsorption of water vapour onto the top surface.94 The results of further XPS experiments will be discussed in Section 5.3.1. 4.5 Rutherford Back-Scattering Spectroscopy Rutherford back-scattering spectroscopy (RBS) is based on bombarding a sample with high- energy helium ions and measuring the energy of these back-scattered ions. With this method, the masses of elements and their depth distribution in the sample can be determined. The depth resolution of RBS is typically 100 A,˚ and atomic concentrations down to 10 − 100 ppm can be detected. Figures 4.7(a) and (b) indicates how the measured RBS spectrum corre- sponds to the elements in the sample structure.270 Since Au, Ag and N are only on the surface in Figure 4.7(a), the RBS signals have a narrow spectral distribution. It also demon- 100 4. Characterisation of TiO2 Thin Films Figure 4.6: XPS surface scan over a wide range of binding energies of a TiO2 film spray deposited onto a silicon wafer. strates that, firstly, the RBS yield increases with atomic number. Secondly, the RBS signal of elements lighter than the substrate (nitrogen in our example) will appear superimposed on the substrate signal, while heavier elements will be displayed as separate peaks.270Figure 4.7(b) demonstrates how the RBS signal for gold broadens as ions back-scattered from deeper within the gold film have lower energies due to additional energy loss mechanisms within the film. More detailed theory on RBS can be obtained from Schroder270 or Breese et al.271 The RBS results in this work were obtained with a collimated beam of 4He2+ ions at an energy of 2 MeV. The beam was scanned over an area of 1 × 1mm2. Backscattered ions were detected with a surface barrier detector with a solid angle of 28 msr at a scattering angle of 145◦. The beam was provided by the University of Melbourne 5U pelletron accelerator. Figure 4.8 shows the RBS spectrum obtained from a TiO2 film deposited by ultrasonic spray 272 pyrolysis onto a silicon wafer and a RUMP simulation of a 100 nm thick TiO2 film. As discussed in Section 2.2.5, the density of the film can also be determined from the RBS spectrum. The areal densities of the titanium and oxygen in the film in Figure 4.8 were 17 2 17 2 ρareal =1.51 ± 0.01 × 10 atoms/cm ρareal =3.1 ± 0.2 × 10 atoms/cm , respectively. The resulting stoichiometry is determined to be 2.02 ± 0.13, being limited by the noise in the oxygen statistics.273 Further results from RBS will discussed in Chapter 6. 4.6 Ellipsometry 101 Figure 4.7: (a) Calculated RBS spectrum for a silicon sample with the elements gold (Au), silver (Au) and nitrogen (N) on the top surface. Note the narrow peaks. (b) Au depth profile information corresponds to an increased Au peak width.270 Figure 4.8: RBS spectrum of a TiO2 film on a silicon wafer (solid line) 272 and a RUMP simulation of a 100 nm thick TiO2 film. 4.6 Ellipsometry 4.6.1 Overview Ellipsometry is primarily used in the semiconductor industry for accurately measuring the thickness of thin dielectric films. However, ellipsometry is also a very powerful tool for determining the optical constants of a film. The following brief introduction to ellipsometry has been adapted from J.A. Woollam.62 102 4. Characterisation of TiO2 Thin Films Ellipsometry measures the change in the state of polarisation of light that is reflected from the front surface of a sample. The measured ellipsometric parameters Ψ and ∆ are related to the ratio of the Fresnel coefficients Rp and Rs for p- and s-polarised light, respectively. Figure 4.9 shows an incoming linearly polarised beam, with the p-direction lying in the plane of incidence (the plane that contains the incident and reflected beams) and the s-direction (from Senkrecht, German for perpendicular) lies perpendicular to the p-direction. Figure 4.9: An ellipsometry experiment, showing the p- and s-directions and the electric field E (from J.A. Woollam62). Figure 4.10 compares linearly, circularly and elliptically polarised light. By looking at the electric field vector E in a plane perpendicular to the direction of propagation the polarisation of the light can be determined. For linearly polarised light (a), E lies in one line at all times and there is no phase difference. In the case of circularly polarised light (b) the Ex and Ey components of E are of equal magnitude but 90◦ out of phase. Circularly polarised light may precess either clockwise or counter-clockwise around the circle. In general, the Ex and Ey-fields do not have to be of equal magnitude and could possess any phase relationship. In this case (c) the tip of the electric field vector E traces out an ellipse as a function of time. The ratio of the reflection coefficients Rp and Rs is defined as the complex reflection ratio ρ, which is related to the measured ellipsometric parameters Ψ and ∆ by R ρ p ı∆ = = tan(Ψ) exp (4.2) Rs The schematic diagram in Figure 4.11 shows the typical componentry of a simple ellipsome- ter. A collimated beam of unpolarised light becomes linearly polarised after passing through the polariser (P). The compensator (C), or retarder, changes this linearly polarised light the elliptically polarised light. The compensator contains a fast and slow optical axis per- pendicular to the direction of transmission. Therefore one component will become retarded in phase relative to the other component. After being reflected off the front surface of the sample the linearly polarised light enters the analyser (A) (similar to the polariser). With 4.6 Ellipsometry 103 Figure 4.10: Diagram looking into the propagating beam, showing (a) linearly, (b) circularly and (c) elliptically polarised light (adapted from J.A. Woollam62). null ellipsometry the signal is extinguished by the analyser and a zero output is observed at the detector. Then by fixing certain angles the number of solutions is reduced down to one pair of Ψ and ∆. There are other ellipsometric configurations (e.g., rotating analyser like the VASE instrument) and the reader is referred to J.A. Woollam62 for further information. In the example shown in Figure 4.11 light is reflected at an air-substrate interface. In this Figure 4.11: Schematic diagram of a typical ellipsometry setup (adapted from270). simple case Fresnel’s equations can be used to show that the complex refractive index n can be determined from the measured Ψ and ∆ values:270 n n − ık = 1 1 4ρ 2 = n0 tan(φ) 1 − sin (φ) . (4.3) (1 + ρ)2 104 4. Characterisation of TiO2 Thin Films If the ellipsometric ratio ρ from Equation 4.2 is measured at the incident angle φ and n0 is known (unity for air) then the refractive index n and extinction coefficient k can be calculated. In Figure 4.12(a) and (b) the behaviour of the p- and s-components of the reflectance and the ellipsometric parameters Ψ and ∆ are plotted for a bare crystalline silicon substrate for light of λ = 633 nm. It can be seen that as the angle of incidence increases Ψ reaches a minimum at 75.5◦. At this angle ∆ changes from very close to 180◦ to 0◦. This angle is known as the Brewster angle and is determined by n −1 1 φB =tan ,(4.4) n 0