USF Libraries
USF Digital Collections

Copper gallium diselenide solar cells

MISSING IMAGE

Material Information

Title:
Copper gallium diselenide solar cells processing, characterization and simulation studies
Physical Description:
Book
Language:
English
Creator:
Panse, Pushkaraj
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Gallium compounds   ( lcsh )
Copper indium selenide   ( lcsh )
Solar cells -- Design and construction   ( lcsh )
photovoltaics
solar
cis
cgs
cigs
indium
Dissertations, Academic -- Electrical Engineering -- Doctoral -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: The goal of this research project was to contribute to the understanding of CuGaSe2/CdS photovoltaic devices, and to improve the performance of these devices. The initial part of the research dealt with the optimization of a Sequential Deposition process for CuIn(Ga)Se2 absorber formation. As an extension of this, a recipe (Type I Process) for CuGaSe2 absorber layer fabrication was developed, and the deposition parameters were optimized. Electrical characterization of the thin films and completed devices was carried out using techniques such as Two-Probe and Three-Probe Current-Voltage, Capacitance-Frequency, Capacitance-Voltage, and Spectral Response measurements. Structural/chemical characterization was done using XRD and EDS analysis. Current densities of up to 15.2 mA/cm2, and Fill Factors of up to 58% were obtained using the Type I CuGaSe2 Process. VOC's, however, were limited to less than 700 mV. Several process variations, such as changes in the rate/order/temperature of depositions and changes in the thickness of layers, resulted in little improvement. With the aim of breaking through this VOC performance ceiling, a new absorber recipe (Type II Process) was developed. VOC's of up to 735 mV without annealing, and those of up to 775 mV after annealing, were observed. Fill Factors were comparable to those obtained with Type I Process, whereas the Current Densities were found to be reduced (typically, 10-12 mA/cm2, with the best value of 12.6 mA/cm2). This performance of Type II devices was correlated to a better intermixing of the elements during the absorber formation. To gain an understanding of the performance limitations, two simulation techniques, viz. SCAPS and AMPS, were used to model our devices. Several processing experiments and SCAPS modeling indicate that a defective interface between CuGaSe2 and CdS, and perhaps a defective absorber layer, are the cause of the VOC limitation. AMPS simulation studies, on the other hand, suggest that the back contact is limiting the performance. Attempts to change the physical back contact, by changes in the absorber processing, were unsuccessful. Processing experiments and simulations also suggest that the CuGaSe2/CdS solar cell involves a true heterojunction between these two layers.
Thesis:
Thesis (Ph.D.)--University of South Florida, 2003.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Pushkaraj R Panse.
General Note:
Includes vita.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 204 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001430562
oclc - 52032611 500/3::INCLUDES VITA.
notis - AJL4023
usfldc doi - E14-SFE0000080
usfldc handle - e14.80
System ID:
SFS0024776:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001430562
003 fts
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 031007s2003 flu sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000080
035
(OCoLC)52032611 500/3::Includes vita.
9
AJL4023
b SE
SFE0000080
040
FHM
c FHM
049
FHME
090
TK145
1 100
Panse, Pushkaraj.
0 245
Copper gallium diselenide solar cells
h [electronic resource] :
processing, characterization and simulation studies /
by Pushkaraj R Panse.
260
[Tampa, Fla.] :
University of South Florida,
2003.
502
Thesis (Ph.D.)--University of South Florida, 2003.
504
Includes bibliographical references.
500
Includes vita.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 204 pages.
520
ABSTRACT: The goal of this research project was to contribute to the understanding of CuGaSe2/CdS photovoltaic devices, and to improve the performance of these devices. The initial part of the research dealt with the optimization of a Sequential Deposition process for CuIn(Ga)Se2 absorber formation. As an extension of this, a recipe (Type I Process) for CuGaSe2 absorber layer fabrication was developed, and the deposition parameters were optimized. Electrical characterization of the thin films and completed devices was carried out using techniques such as Two-Probe and Three-Probe Current-Voltage, Capacitance-Frequency, Capacitance-Voltage, and Spectral Response measurements. Structural/chemical characterization was done using XRD and EDS analysis. Current densities of up to 15.2 mA/cm2, and Fill Factors of up to 58% were obtained using the Type I CuGaSe2 Process. VOC's, however, were limited to less than 700 mV. Several process variations, such as changes in the rate/order/temperature of depositions and changes in the thickness of layers, resulted in little improvement. With the aim of breaking through this VOC performance ceiling, a new absorber recipe (Type II Process) was developed. VOC's of up to 735 mV without annealing, and those of up to 775 mV after annealing, were observed. Fill Factors were comparable to those obtained with Type I Process, whereas the Current Densities were found to be reduced (typically, 10-12 mA/cm2, with the best value of 12.6 mA/cm2). This performance of Type II devices was correlated to a better intermixing of the elements during the absorber formation. To gain an understanding of the performance limitations, two simulation techniques, viz. SCAPS and AMPS, were used to model our devices. Several processing experiments and SCAPS modeling indicate that a defective interface between CuGaSe2 and CdS, and perhaps a defective absorber layer, are the cause of the VOC limitation. AMPS simulation studies, on the other hand, suggest that the back contact is limiting the performance. Attempts to change the physical back contact, by changes in the absorber processing, were unsuccessful. Processing experiments and simulations also suggest that the CuGaSe2/CdS solar cell involves a true heterojunction between these two layers.
590
Adviser: Morel, Don
650
Gallium compounds.
653
photovoltaics.
Copper indium selenide.
Solar cells
x Design and construction.
solar.
cis.
cgs.
cigs.
indium.
690
Dissertations, Academic
z USF
Electrical Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.80



PAGE 1

COPPER GALLIUM DISELENIDE SOLAR CELLS: PROCESSING, CHARACTERIZATION AND SIMULATION STUDIES by PUSHKARAJ PANSE A dissertation submitted in partial fulfillment Of the requirements for the degree of Doctor of Philosophy Department of Electrical Engineering College of Engineering University of South Florida Major Professor: Don Morel, Ph.D. Committee Member: Christos Ferekides, Ph.D. Committee Member: Elias K. Stefanakos, Ph.D. Committee Member: Luis H. Garcia Rubio, Ph.D. Committee Member: Jo seph Stanko, Ph.D. Date of Approval: March 28, 2003 Keywords: photovoltaics, cgs, cis, cigs, indium Copyright 2003, Pushkaraj Panse

PAGE 2

DEDICATION I dedicate this research and this Dissertation to my Mom, Mrs. Arundhati Panse, a nd my Dad, Prof. Ramesh Panse. They both have played a major role in what I have accomplished in my life.

PAGE 3

ACKNOWLEDGEMENTS I am deeply indebted to Dr. Don Morel, my Major Professor. He guided and motivated me throughout this proje ct, but his teachings went much beyond that. He always came across as a person of the highest integrity. He will always be a father figure to me. Many thanks to Dr. Christos Ferekides, who taught me a lot about solar cells and vacuum systems, and was alw ays there when I needed help. Thanks to Dr. E. Stafanakos, Dr. L. H. Garcia Rubio, and Dr. J. Stanko, for agreeing to be on my committee. I am also grateful to the many co workers and graduate students I have had the pleasure of working with. Special th anks to Harish Sankaranarayanan, for helping me every step of the way. Also, many thanks to Vijay, John, Raj, Ramesh, Bhaskar, Dima, Sveta, Zafar, Hari, Balaji, and all other solar cell lab mates, for making my research project a truly enriching experienc e. I am what I am, because of my family in India, my friends throughout the years, and, because of my wife. To talk about repaying their debt is an insult to their unconditional love, affection, and friendship. So Ill just say I consider myself the luck iest person alive.

PAGE 4

i TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES ABSTRACT PREFACE CHAPTER 1. INTRODUCTION 1.1. Everything under the Sun 1.2. The Game of Power 1.3. Traditional Electricity Generation Technologies 1.4. Renewable Electricity Generation Technologies 1.4.1. Survey 1.4.2. Solar Energy Conversion 1.4.3. The Past and Future of Renewables CHAPTER 2. BACKGROUND I: BASIC PHYSICS 2.1. Whats a Solar Cell? 2.2. A P N Junction in the Dark 2.3. Interaction with Light 2.3.1. Photocurrent 2.3.2. I V Characteristics 2.4. Heterojunctions CHAPTER 3. BACKGROUND II: THIN FILM PHOTOVOLTAICS 3.1. Histo rical Background of Photovoltaics 3.2. Thin Film Photovoltaics 3.3. CuInSe 2 Family Based Thin Film Photovoltaics 3.3.1. CuInSe 2 Family and Device Issues 3.3.2. Bandgap Engineering 3.3.3. CuIn(Ga)Se 2 3.4. CuGaSe 2 Solar Cells iv v ix 1 3 3 4 5 8 8 10 12 19 19 20 23 23 25 30 36 36 39 41 41 47 49 53

PAGE 5

ii 3.4.1. Advantages of a High Bandgap 3.4.2. CuGaSe 2 Device Issues 3.4.2.1. High Bandgap and V OC Limitation 3.4.2.2. Band Discontinuities and Surface Inversion 3.4.2.3. CuGaSe 2 Transport Mechanisms 3.4.2.4. Effect of Buffer Deposition 3.4.2.5. Effect of Post Deposition Treatments 3.4.2.6. Othe r CuGaSe 2 Issues CHAPTER 4. OUR SOLAR CELLS: FABRICATION AND CHARACTERIZATION 4.1. The Device Structure 4.2. Fabrication of CuGaSe 2 Solar Cells 4.2.1. A Manufacturing Friendly Process 4.2.2. The Substrate 4.2.3. The Back Contact 4.2.4. The Heterojunction Partner/ Buffer 4.2.5. The Front Contact 4.3. The Absorber 4.3.1. Type I Versus Type II 4.4. Character ization of CuGaSe 2 Solar Cells 4.4.1. Current Voltage (I V) Measurements 4.4.2. Spectral Response 4.4.3. Capacitance Measurements 4.4.4. I SC V OC Measurements CHAPTER 5. RESULTS AND DISCUSSION 5.1. PART I: CuIn(Ga)Se 2 and CuGaSe 2 Processing Results 5.1.1. Relative Positions of the substrate and the Sources 5.1.2. Ga Evaporation and the Sample Numbering System 5.1.3. CuIn(Ga)Se 2 Processing 5.1.3.1. CuIn(Ga)Se 2 Sample # P020 5.1.3.2. CuIn(Ga)Se 2 Samples # P030 and # P031 5.1.4. Type I CuGaSe 2 5.1.4.1. CuGaSe 2 Sample # P041 (Type I) 5.1.4.2. CuGaSe 2 Sample # P042: Reduction in Cu 5.1.4.3. CuGaSe 2 Sample # P043: Continued Reduction in Cu 5.1.4.4. Samples # P060, P061: Continued Reduction in C u 5.1.4.5. Sample # P063: Variation in the Initial Ga 5.1.4.6. Samples # P062 and P082: I V Curve shapes and Absorber Thickness 5.1.5. Type II CuGaS e 2 5.1.5.1. EDS and XRD Characterization for Type II 53 54 54 57 59 65 69 70 72 72 73 73 74 75 76 77 79 80 83 84 85 86 89 92 92 92 93 94 95 99 101 102 103 106 111 112 114 118

PAGE 6

iii CuGaSe 2 5.1.5.2. Type II Samples # P111, # P115: Effect of Se 5.1.5.3. Type II Sam ple # P119: Initial Ga 5.1.5.4. Current Density Performance of Type II 5.1.5.5. Type II Sample # P132: Cu rich Cu poor Transition 5.1.6. Type I B a nd Type II B CuGaSe 2 : Cu Se Co evaporation 5.1.7. CdS and Other Buffer Layers 5.1.8. Light Soaking and Annealing Experiments for CuGaSe2 5.1.9. Capacitance Studies of CuGaSe 2 5.2. PART II: CuIn(Ga)Se 2 an d CuGaSe 2 Simulation/Modeling Results 5.2.1. SCAPS Modeling 5.2.2. AMPS Modeling CHAPTER 6. CONCLUSION REFERENCES APPENDICES APPENDIX A. EDS and XRD Results ABOUT THE AUTHOR 119 120 122 122 124 126 129 133 143 148 148 160 181 185 188 189 End Page

PAGE 7

iv LIST OF TABLES Table 1.1. U.S. Renewable Energy Consumption, in Quadrillion BTUs 13 Table 3.1. Various Ternary Absorber Materials with their Bandgaps 48 Table 5.1. SCAPS Simulations: Layers and Typical Parameter Values 149 Table 5.2. SCAPS Simulation Parameters and Results: Interface States Only 150 Table 5.3. SCAPS Simulation Parameters and Results: Bulk and Interface States 153

PAGE 8

v LIST OF FIGURES Figure 1.1. The Alarming Increase in the Conce ntration of Atmospheric CO 2 7 Figure 1.2. 1998 U.S. Energy Consumption by Energy Source 13 Figure 1.3. Todays Energy Mix 14 Figure 2.1. Formation of a Homojunction p n Diode 21 Figure 2.2. Photocurrent Generation in a p n Homojunction Solar Cell 25 Figure 2.3. I V Characteristic of an Ideal p n Junction Solar Cell 25 Figure 2.4. An Equivalent Circuit of a Solar Cell 26 Figure 2.5. Effect of a Non Zero Series Resistance on the I V Characteristic 29 Figure 2.6. Effect of a Finite Shunt Resistance on the I V Characteristic 30 Figure 2.7. Two Materials, Before Heterojunction Formation 31 Figure 2.8. Heterojunction Formation 32 Figure 2.9. Band Diagram of the CuIn(Ga)Se 2 /CdS/ZnO Solar Cell 34 Figure 3.1. The CuInSe 2 Structure 42 Figure 3.2. The Structure of a Special Defect Pair 46 Figure 3.3. The CuInSe 2 Structure 49 Figure 3.4. Band Bending with (a) No Grading, and (b) Grading 51 Figure 3.5. Band Offsets of Three Chalcopyrite Compounds 58 Figure 3.6. Valence and Conduction Band Offsets 58

PAGE 9

vi Figure 3.7. Band Diagram (a) at Equilibrium, (b) at an Applied Bias 60 Figure 3.8. Equilibrium Band Diagram for the ZnO/CdS/CuGaSe 2 Structure 64 Figure 3.9. Variation of the Inverse Diode Factor with Temperature 68 Figure 4.1. Structure of Our Typical Solar Cell Device 72 Figure 4.2. Arrangement of the Substrate and the Sources 79 Figure 4.3. Time Temperature Profile for Type I CuGaSe 2 81 Figure 4.4. Time Temperature Profile for Type II CuGaSe 2 83 Figure 4.5. A Representative 3 Probe I V Curve Plot for a USF CuGaSe 2 Cell 85 Figure 4.6. A Representative Spectral Response Curve for a USF CuGaSe 2 Cell 86 Figure 4.7. A 1/C 2 V Plot for a USF CuInGaSe 2 Cell 89 Figure 4.8. An I SC V OC Plot for a USF CuInGaSe 2 Cell 91 Figure 5.1. Arrangement of the Substr ate and the Sources 93 Figure 5.2. Two Probe V OC Numbers for CuIn(Ga)Se 2 Sample # P020 95 Figure 5.3. Two Probe V OC Numbers for CuIn(Ga)Se 2 Sample # P020 95 Figure 5.4. Fill Factor Vs. Position of Device, for 8 Devices on Sample # P020 97 Figure 5.5. T wo Probe V OC Numbers for CuIn(Ga)Se 2 Sample # P030 100 Figure 5.6. Two Probe V OC Numbers for CuIn(Ga)Se 2 Sample # P031 100 Figure 5.7. Two Probe V OC Numbers for CuGaSe 2 Sample # P041 103 Figure 5.8. Two Probe V OC Numbers for CuGaSe 2 Sample # P042 104 F igure 5.9. Short Circuit Current (I SC ) Vs. Position of Device, for Sample # P042 104 Figure 5.10. Two Probe V OC Numbers for CuGaSe 2 Sample # P043 107 Figure 5.11. Two Probe I SC Numbers for CuGaSe 2 Sample # P043 107 Figure 5.12. Spectral Response Curve f or CuGaSe 2 Sample # P043, Device # 7 108

PAGE 10

vii Figure 5.13. Spectral Response Curve for CuGaSe 2 Sample #P043, Device # 8 109 Figure 5.14. Two Probe V OC Numbers for CuGaSe 2 Sample # P060 111 Figure 5.15. Two Probe V OC Numbers for CuGaSe 2 Sample # P061 112 Fig ure 5.16. Two Probe V OC Numbers for CuGaSe 2 Sample # P063 113 Figure 5.17. Three probe I V Curve for Device # 15 from CuGaSe 2 Sample # P062 115 Figure 5.18. Three Probe V OC Numbers for CuGaSe 2 Sample #P082 116 Figure 5.19. Three Probe I V Curves for Dev ices # 11, 12 for Sample # P082 117 Figure 5.20. Three Probe V OC Numbers for CuGaSe 2 Sample # P111 120 Figure 5.21. Three Probe V OC Numbers for CuGaSe 2 Sample # P115 121 Figure 5.22. Three Probe V OC Numbers for CuGaSe 2 Sample # P119 122 Figure 5.23. Sp ectral Response for CuGaSe 2 Type II Sample # P115, Device # 12 123 Figure 5.24. Three Probe V OC Numbers for CuGaSe 2 Sample # P132 125 Figure 5.25. Three Probe J SC Numbers (mA/cm 2 ) for CuGaSe 2 Sample # P132 125 Figure 5.26. Two Probe V OC Numbers for CuGa Se 2 Type I B Sample #P093 127 Figure 5.27. Two Probe V OC Numbers for CuGaSe 2 Type II B Sample # P151 127 Figure 5.28a. Three probe I V for P098 06: As dep/Light Soak 134 Figure 5.28b. Three probe I V for P098 06: Anneal 1/Anneal 2 135 Figure 5.29a. Thr ee probe Plots for P098 24: As dep/Light Soak 138 Figure 5.29b. Three probe Plots for P098 24: Anneal 1/Anneal 2 139 Figure 5.30a. Three probe for Sample # P082, Device # 12: As dep 140 Figure 5.30b. Three probe for Sample # P082, Device # 12: After Ann eal 141 Figure 5.31. Three Probe V OC Numbers for CuGaSe 2 Sample # P115 143 Figure 5.32. A 2 /C 2 Vs. V Curve for Device # 1 on CuGaSe 2 Sample # P082 144

PAGE 11

viii Figure 5.33. Truncated A 2 /C 2 Vs. V Curve for Device # 1 on Sample # P082 145 Figure 5.34. Dark Depleti on Widths for Several Devices on 3 CuGaSe 2 Samples 147 Figure 5.35. I V Plots for SCAPS Simulations (Interface Defects Only) 151 Figure 5.36. I V Plots for SCAPS Simulation (Bulk and Interface Defects) 154 Figure 5.37. I V Plots for SCAPS Simulation: Ba ck Contact Work Function 155 Figure 5.38. I V Plots for SCAPS Simulation 157 Figure 5.39. I V Kink in Device # 1 from CuGaSe 2 Sample # P137 158 Figure 5.40. AMPS 1 167 Figure 5.41. AMPS 2 168 Figure 5.42. AMPS 3 169 Figure 5.43. AMPS 4 170 Figure 5. 44. AMPS 5 171 Figure 5.45. AMPS 6 172 Figure 5.46. AMPS 7 173 Figure 5.47. AMPS 8 174 Figure 5.48. AMPS 9 175 Figure 5.49. AMPS 10 176 Figure 5.50. AMPS 11 177 Figure 5.51. AMPS 12 178 Figure 5.52. AMPS 13 179 Figure 5.53. AMPS 14 180

PAGE 12

ix COPP ER GALLIUM DISELENID E SOLAR CELLS: PROCESSING, CHARACTE RIZATION AND SIMULAT ION STUDIES PUSHKARAJ PANSE ABSTRACT The goal of this research project was to contribute to the understanding of CuGaSe 2 /CdS photovoltaic devices, and to improve the performance of these devices. The initial part of the research dealt with the optimization of a Sequential Deposition process for CuIn(Ga)Se 2 absorber formation. As an extension of this, a recipe (Type I Process) for CuGaSe 2 absorber layer fabrication was developed, an d the deposition parameters were optimized. Electrical characterization of the thin films and completed devices was carried out using techniques such as Two Probe and Three Probe Current Voltage, Capacitance Frequency, Capacitance Voltage, and Spectral Re sponse measurements. Structural/chemical characterization was done using XRD and EDS analysis. Current densities of up to 15.2 mA/cm 2 and Fill Factors of up to 58% were obtained using the Type I CuGaSe 2 Process. V OC s, however, were limited to less th an 700 mV. Several process variations, such as changes in the rate/order/temperature of depositions and changes in the thickness of layers, resulted in little improvement. With the aim of breaking through this V OC performance ceiling, a new absorber reci pe (Type II Process) was developed. V OC s of up to 735 mV without annealing, and those of up to

PAGE 13

x 775 mV after annealing, were observed. Fill Factors were comparable to those obtained with Type I Process, whereas the Current Densities were found to be redu ced (typically, 10 12 mA/cm 2 with the best value of 12.6 mA/cm 2 ). This performance of Type II devices was correlated to a better intermixing of the elements during the absorber formation. To gain an understanding of the performance limitations, two sim ulation techniques, viz. SCAPS and AMPS, were used to model our devices. Several processing experiments and SCAPS modeling indicate that a defective interface between CuGaSe 2 and CdS, and perhaps a defective absorber layer, are the cause of the V OC limita tion. AMPS simulation studies, on the other hand, suggest that the back contact is limiting the performance. Attempts to change the physical back contact, by changes in the absorber processing, were unsuccessful. Processing experiments and simulations also suggest that the CuGaSe 2 /CdS solar cell involves a true heterojunction between these two layers.

PAGE 14

1 PREFACE Photovoltaics is the method of converting sunlight into electricity. It is a simple and environmentally friendly method of producing electricity. One promising way to make such solar electricity affordable for the layman is to use various thin film technologies such as Copper Indium Gallium Diselenide (CuIn(Ga)Se 2 ) and Cadmium Telluride (CdTe). Copper Gallium Diselenide (Cu GaSe 2 ), which is a variation of CuIn(Ga)Se 2 has the potential to be used as a high voltage producing solar cell material, and, also, to be used as a material for multi structured tandem solar cell systems. This document discusses the research project tha t focused on the development of a manufacturing friendly process for making CuGaSe 2 solar cells, and the characterization, as well as the computer simulation studies of these solar cells. The first chapter, Introduction presents a detailed review of trad itional (nonrenewable), as well as renewable electricity generation technologies. Although much of this chapter does not relate directly to the specific project undertaken during this research, it serves a dual purpose. First, it attempts to make a stron g case (and, hopefully, succeeds in doing so) for the support of solar energy research. Secondly, it was deemed necessary, by the author of this document, to introduce the reader to the general topic of renewables. It is a need of the present time that e very truly concerned citizen not only take notice of renewables and their positive effect on the environment, but also recognize the responsibility to help bring about the transition to renewables from the traditional

PAGE 15

2 non renewable technologies. If the re ader is already familiar with such topics, he or she should feel free to skip to the next chapter of this dissertation. The second topic, Background has two parts. Background I deals with the basic physics of solar cells. Background II reviews thin film photovoltaics in general, and then specializes into CuIn(Ga)Se 2 thin film photovoltaics. In addition to the treatment of fundamental workings of the relevant thin film devices, this chapter also provides historical perspectives, along with numerous refer ences to the past research carried out in this area. The third chapter introduces the reader to the fabrication, characterization and simulation techniques that have been used in this research endeavor. The next topic, Results and Discussion has been di vided into two parts. Part I deals with the processing and characterization results obtained with CuIn(Ga)Se 2 and CuGaSe 2 while Part II presents the computer simulation/modeling results. Lastly, Chapter 6 presents our conclusions.

PAGE 16

3 CHAPTER 1. INTRODUCTION "We're altering the environment far faster than we can possibly predict the consequences. This is bound to lead to some surprises." -Dr. Stephen Schneider, National Center for Atmospheric Research. 1.1. Everything Under The Sun! In our busy everyday routine, we hardly have the time and the willingness to stop and think about something as basic as Solar Energy. When we do (if we do), we often limit ourselves to thinking about the electricity generated using solar energy. And th ats only natural, because, in this hi tech world that we are living, we are primarily concerned with only those things that can make our everyday life easier (and make ourselves lazier). The fact is that life on earth has always depended on solar energy. This was true before the invention of electricity, and is equally true now. Yes, of course, all of us have learnt, back in kindergarten, that the Sun is the star that gave birth to our mother Earth; that it was the solar energy that kept our planet warm enough so life could sustain here; and that without it, there would be no photosynthesis of the plants, and no light and warmth for the organisms to live and evolve. But who cares? We are here now, and the Suns here to stay (so we were told, at least). And we need our heater turned on because its too

PAGE 17

4 darn cold, and the lights turned on because its kind of dark. And would you be kind enough to turn the TV on and pass me the remote please? Well, maybe its a fact that we, in this modern commercialized world, have little time to think about anything outside the little circle that only contains our families, workplaces, and a few friends that our busy lifestyles can afford. Topics such as protecting the global environment of the very planet that we live on rarely attract our attentions anymore. 1.2. The Game of Power Electric power has now become the fundamental platform that supports most of our physical needs. Indeed, it was only a couple of centuries ago that there was no electricity. But, today, it is so difficult to imagine ourselves without it. No wonder the California Power Crisis has scared many a folk, and is making the headlines at CNN everyday. Even scarier the fact is that theres a possibility that the whole power situation is going t o get a lot worse than it is now. There are several factors that contribute towards the immense increase of electricity usage in the world seen in the recent times. Firstly, theres the population growth. More people to use the power, so more power is n eeded. The birth of the official 6 billionth baby was recently celebrated. Official because there are 250 babies born around the world in a single minute, and 15,020 in a single hour [http://www.census.gov]. The second most important factor is the fas t industrialization of the developing and underdeveloped countries. There are more factors, such as the

PAGE 18

5 gradual spread of the (so called modern) style of living, where larger joint families are often broken into smaller ones, so the houses and workplaces arent shared anymore, thereby increasing the total power usage. As an example of how fast the demand for electricity is rising, in the United States alone, electricity sales increased by about 80% from 1975 to 1997 [Mcveigh, 2000]. Obviously, to keep up with this ever increasing demand, the generation of electric power has to increase. Several technologies have been, and are being, used to generate this power. These can vaguely be divided into two types, traditional and non traditional (renewable) t echnologies, and are reviewed below. 1.3. Traditional Electricity Generation Technologies Traditionally, most of the electricity has been generated using the following three technologies: (i) Power from Fossil Fuels: These fuels include oil, coal and natural gas, (ii) Nuclear Power, and (iii) Hydroelectric Power. The first one, viz. fossil fuels technology, can be categorized as a non renewable technology, meaning that the sources used for this technology cannot be recycled. The other two can be classified as renewable s, and will be described in a later section. As of today, a majority (70%) of todays power is, in fact, generated using fossil fuels [Sweet, 2001]. This is not at all surprising, for fossil fuels are abundant at present, and so the

PAGE 19

6 resulting power is ve ry cheap for the consumer. However, there are a few important aspects of this technology that show up as distinct disadvantages when we consider the future of electricity generation. Firstly, although there seem to be enough coal and oil available right n ow, we know that the supply of these fuels cannot be infinite. While the demands shoot up, the fuels are gradually being depleted. There is going to be a time when this will start affecting the consumers wallet, and then there is going to be another tim e when there just are no more fossil fuels left that can be easily accessed by the humankind. There is disagreement among scientists and forecasters about when this might happen. According to some, it could be as early as the next decade, while others fe el confident that newer fossil locations would be discovered that could delay this situation by decades, or even, centuries. Whichever direction one might choose to believe in, there can hardly be any disagreement about one thing: it is safer to find alte rnatives that can take up the burden of the fossil fuel technology, rather than to wait until the last minute. Secondly, there are strong political implications of the fact that fossil fuel source locations are distributed unevenly around the world. One of the biggest proofs of this came about when the fossil fuel prices shot up in the wake of the Gulf War. Although such major events are rare, even minor uncertainties associated with the changes in international political relations that are results of in ternal policy changes of various countries, can have serious impact on the electricity bill that a consumer pays. The third aspect is perhaps the most important one. There are serious environmental concerns associated with the use of fossil fuels as the s ource of electric power. These fuels -coal, oil and natural gas -were created chiefly by the decay of

PAGE 20

7 plants that flourished millions of years ago. Burning these fuels unlocks the carbon stored by these plants and releases it to the air as carbon dio xide. For instance, burning one gallon of gasoline generates 22 pounds of carbon dioxide. In other words, it takes a pound of coal to generate the electricity to light a 100 watt bulb for 10 hours. For every pound of coal we burn, nearly three pounds of carbon dioxide go into the atmosphere. Since 1750, carbon dioxide in the air has risen by more than 30%. It could double by the year 2065! This atmospheric CO 2 rise over the years is depicted in the following figure. Figure 1.1. The Alarming Increase in the Concentration of Atmospheric CO 2 (Redrawn from www.enviroweb.org ) Billions of tons of carbon dioxide are released into the atmosphere every year. This pollutes our atmosphere, but that is not where it stops. It is the cause of another permanent damage. The atmosphere has always contained carbon dioxide, methane and nitrous oxide. These gases, together with water vapor, trap some of the Sun's energy an d keep the Earth warm enough to sustain life. This process is called the Greenhouse Effect 350 CO 2 conc. (ppm) 330 Year 310 290 270 250 1750 1 800 1850 1900 1950 2000

PAGE 21

8 which is a natural phenomenon. However, human activities increase some of the gases, and add new ones, thereby intensifying the natural greenhouse effect. This, according to many researchers, can eventually cause excessive warming of the atmosphere. Such a permanent change can potentially result in devastating effects such as the melting of polar ice, gigantic floods and rise of ocean levels. Of course, there a re skeptics that dont yet believe in the greenhouse theory. However, in the words of James Hansen, Greenhouse Researcher, Goddard Institute for Space Studies, It is time to stop waffling so much and say that the evidence is pretty strong that the greenh ouse effect is here." Carbon dioxide accounts for three fourths of the predicted increase in the greenhouse effect. In addition to carbon dioxide, burning coal and other fossil fuels also releases sulfur dioxide, nitrogen oxides, and particulates, adding to the air pollution. One partial (and temporary) solution to this problem is to start using more natural gas, as natural gas releases lesser amounts of carbon dioxide. However, even these amounts are substantial, so alternatives to fossil fuels are deem ed necessary. The so called renewable technologies are the ideal alternatives. Lets see why. 1.4. Renewable Electricity Generation Technologies 1.4.1. Survey Lets now take a look at the class of renewable technologies, two traditional ones of which are the nuclear power and the hydroelectric power.

PAGE 22

9 The nuclear power alternative, although being used quite extensively, has drawn serious concerns regarding safety. Accidents such as at Chernobyl are still fresh in our minds. Moreover, the radioactiv e waste that is created in a nuclear plant has to go through an expensive and elaborate disposal. Such disposal practices have been proven to be quite controversial, and the public is becoming more and more aware of the uncertainties involved. Hydroelectr ic power has also been developed extensively, taking up a major share of the electricity generation in developing countries. However, this, too, comes at the expense of the environment, essentially destroying the river ecosystems that the hydroelectric pl ants are built on. Therefore, significant expansion of this resource faces severe opposition from environmentalists. (An example of how this technology has had devastating effects on human life is the Narmada river/ Sardar Lake project in central India: w ww.narmadabachao.com ) Because of the these reasons, although nuclear and hydroelectric are renewables, they cannot be considered exactly environment friendly. This, then, leaves us with the five renewable technologies that can be considered environmenta lly clean, which are: (i) Wind: Wind spins blades, which turn a generator to produce power. (ii) Solar Photovoltaics (the topic of this dissertation): Sunlight is converted directly to electricity using appropriate semiconductor materials. (iii) Solar thermal: Sunlight, reflected with the help of mirrors, is then used to boil water that runs a turbine to produce electricity. (iv) Geothermal: Makes use of the natural heat present inside the earth. Steam coming up through wells is used to produce electricity.

PAGE 23

10 (v) Biomass : Burning /extracting fuels from fast growing plants produces power. Because solar electricity is the topic of this research project, only this technology will be described in detail, and that is the subject of the next section. 1.4.2. Solar Energy Conversion T his energy is created by the fusion reactions that take place in the Sun, and is practically unlimited. The Earth receives about 7.45 x 10 17 kWh of this energy annually from the Sun in the form of sunlight. The annual power consumption of the world is ap prox. 400 Quadrillion BTUs (1.2 x 10 14 kWh). This means that if we have an economical and easy way of making use of the sunlight to produce electric power, our needs can be satisfied. Ideally, a source of energy must be inexpensive, widely and easily a vailable, easy to use, environmentally friendly and renewable. Solar energy satisfies most of these criteria. Its free, environmentally clean and renewable. The only shortcoming is regarding the availability. There are some places around the world, wh ich do not receive enough sunlight during a major part of the year. Obviously, solar energy would be a poor choice for such locations. However, even at the sunniest places, the sun is not available round the clock. (This drawback is also shared by the w ind technology.) The result is an intermittent generation potential. Hence, it is vitally important that an appropriate storage technology such as hydroelectric pump storage or batteries to store electricity and/or potential energy is available. Otherwi se, it may not be economical to employ this technology. It may, however, be viable to employ it as a secondary source to

PAGE 24

11 complement an existing conventional electricity grid, provided the availability of the sun coincides with the periods of peak energy d emand. Photovoltaics is the process of conversion of solar energy into usable electric power. A typical photovoltaic cell is an integrated device consisting of layers of semiconductor materials and electric contacts. Several such cells are usually interc onnected to form an integrated assembly that is called a solar cell module. Such modules are then placed at the appropriate places where they get exposure to bright sun. The absorption of this sunlight produces electricity, which can either be used direc tly, or, more often, is stored in some sort of energy storage system for later use. Todays photovoltaic market is 151 Megawatts per year, corresponding to a value of about 0.7 1 billion US Dollars [Goetz]. Presently, there are about 34 photovoltaics modu le manufacturers in the U.S. [national center for photovoltaics, www.nrel.gov]. According to the Energy Information Administration [doe: www.eia.doe.gov ], photovoltaic (PV) cells and modules shipments had reached ab out 50 peak megawatts in 1998. (Module shipments accounted for 32 peak megawatts, while cell shipments accounted for 18 peak megawatts.) Cells and modules that used crystalline silicon dominated the PV industry in 1998, accounting for 93 percent of total shipments, the remaining 7 percent of the share going to Thin film technologies (to be explained later). In 1998, the average price for modules (dollars per peak watt) was about $3.94. Although the solar power market growth in the last decade was between 15% and 20%, the consumption statistics indicate very clearly that solar technologies are not yet getting a major share of the energy production market. The reason is economical. The current cost of solar energy generation is about 30 cents per a kilowa tt hour (The current

PAGE 25

12 cost of capacity, which is the capital cost measured by the dollar expenditure for the rated capacity, is about $6000/kilowatt) [Mcveigh, 2000]. This cost of generation includes the various costs at the point of production, such as th ose for the capital, the fuel, and the operation and maintenance, and is then leveled with respect to the total costs of production over the lifetime of the production facility. The cost, when compared with the traditional electricity cost of about 6 to 8 cents/kilowatt hour, is still very high. However, it is expected that, in the coming years, the cost of solar photovoltaics will continue to decline. This feat can be accomplished by improving the conversion efficiencies of the solar cells, while also im proving the methods of capturing solar radiation. 1.4.3. The Past and Future of Renewables In 1998, the renewable energy consumption in the United States was 7 quadrillion Btu, accounting for almost 8 percent of the total U.S. energy consumption. The di vision of total energy consumption into individual generation technologies for 1998 is depicted in Figure 1.2 on the next page [Energy Information Administration: www.eia.doe.gov]. As can be seen from this figure, hydroelectric power and biomass dominated the renewable energy market, with 50 percent and 43 percent shares, respectively. The remaining 7% was split among solar, wind, and geothermal technologies. Table 1.1, shown on the next page, contains information about U.S. renewable energy consumption by energy source, for five years: 1994 1998. Note that this includes hydroelectric power, and as can be seen from the numbers, increase in this power is mainly responsible for the total renewable power increase from 1994 to 1998.

PAGE 26

13 Figure 1.2. 1998 U.S. Energy Consumption by Energy Source Table 1.1. U.S. Renewable Energy Consumption, in Quadrillion BTUs Energy Source 1994 1995 1996 1997 1998 Conventional Hydroelectric 2.971 3.474 3.913 3.922 3.540 Geothermal 0.39 5 0.339 0.352 0.328 0.334 Biomass 2.917 3.048 3.108 2.981 3.052 Solar 0.072 0.073 0.075 0.074 0.074 Wind 0.036 0.033 0.035 0.034 0.031 Total Renewable 6.390 6.968 7.483 7.339 7.032 According to the Electric Power Research Institute, todays U.S. ene rgy consumption by energy source is as shown below [Sweet, 2001]. Petroleum 39% Natural Gas 23% Coal 23% Nuclear 8% Renewables 8% Wind < 0.5% Geo Thermal 5% Biomass 43% Solar 1% Hydro 50%

PAGE 27

14 Figure 1.3. Todays Energy Mix It will be illuminating to compare todays energy mix (Figure 1.3) to that of 1998. Several conclusions can be drawn from such a c omparison. Firstly, the proportion of fossil fuel power has decreased from 85% to 70%, which is a positive step towards reducing the devastating effects on the environment. However, inside of fossil fuels, the share taken up by natural gas has decreased and this goes in the opposite direction, because natural gas is the least harmful than coal or oil. Similarly, although non nuclear renewables (includes hydroelectric) percentage has increased from 8 to 12, the share of non nuclear, non hydroelectric r enewables has actually gone down from about 4% to 2%. Some of the renewable energy technologies have been in development for a few decades. In spite of this, renewables have failed to emerge as a prominent component of the energy generation, as can be see n from the above analysis. As mentioned before, the Coal 52% Nuclear 18% Hydro 10% Natural gas 15% Oil 3% Renewables 2%

PAGE 28

15 main reason for this is the fact that the power obtained from these technologies is still expensive, when compared to that obtained from traditional generation technologies. It has hence been argued tha t renewables have not met the goals and claims that were set by their supporters. Mcveigh, et al. have addressed this issue in a recent publication [Mcveigh, 2000]. They have provided an evaluation of the performance of five renewable energy technologies those being biomass, geothermal, solar photovoltaics, solar thermal and wind. Their findings refute the above argument. The authors conclude that renewable technologies have failed to meet expectations only with respect to market penetration. However, in terms of meeting the goals with respect to their cost, these technologies have succeeded, sometimes even exceeding those expectations. An important thing to remember is that the main motivation for developing renewable energy technologies is the desi re to get away from fossil fuels with their adverse effect on the environment. Use of such technologies will help both, to slow global warming, and to reduce air pollution. Traditional technologies may produce cheap power in terms of cost to the consumer However, environmental, political, and health costs are not reflected in this cost. Were a cost assessed for the degradation of the environment and health, and were the other costs shifted from the taxpayer to the consumer, the increased cost would be significant [Gabor]. Tsur, et al. have used dynamic optimization methods to analyze the development of solar technologies in light of the increasing scarcity and environmental pollution associated with fossil fuel combustion [Tsur, 2000]. They have includ ed shadow prices to account for this scarcity and pollution, to allow a valid evaluation of social costs and benefits of alternative energy options. Based on the analysis, the authors predict that

PAGE 29

16 alternative energy sources will eventually capture an incr easing share of energy supply. Moreover, their model advocates substantial early engagement in solar R&D programs that should precede, rather than follow, future increases in the price of fossil fuels. Indeed, the only argument against the renewables i s that these technologies are not cost effective for the consumer. Here's an excerpt from Home Power Magazine, written by Randy Udall, explaining why we have to move past this "cost effective" argument: "Building 110 nuclear power plants before figuring o ut what to do with the waste is cost effective. Drowning the Columbia river and its priceless salmon runs is cost effective. Spending $50 billion a year to defend the Persian Gulf oil fields is prudent. Strip mining pays nice dividends: Wyoming coal is literally cheaper than dirt. Chernobyl was a superb investment. . Conventional energy economics is a value system masquerading as mathematics. At its heart is one key assumption: the future is worthless and the environment doesn't matter. . Switchi ng to renewables from the fossil fuels seems to be the only solution, when one considers the environmental factors. The advantages gained from such a switch are evident in the following example. A one kilowatt PV system: (i) Prevents 150 lbs. of coal fr om being mined, (ii) Prevents 300 lbs. of CO 2 from entering the atmosphere, (iii) Keeps 105 gallons of water from being consumed, (iv) Keeps NO and SO 2 from being released into the environment, each month! [www.solarenergy.org]

PAGE 30

17 Unfortunately, because of the high retail cost, it is unfair to expect the consumer to opt for a solar panel rather than a connection to the conventional grid electricity. Therefore, there has to be an integrated effort from the layman, the Government and the Private Sector to ma ke this switch from non renewables to renewables possible. More public awareness will help build the public support, which hopefully will bring in the required change in the policies on the part of the Government. Recently, there has been a reduction in the amounts of Government funds for the development of such technologies as solar photovoltaics. Such policies have been, and are being, criticized both by technical and political supporters of photovoltaics. In Resources for the Future John F. Ahearne writes about such issues [Ahearne, www.ulib.org]: Although nuclear support (of the Government) has been productive, the large dollar amounts spent on such projects as the Clinch River Breeder Reactor could have been spent much better elsewhere. As far bac k as 1955, Greenewalt wrote, "I wonder what our position would have been today had the amounts of money and effort equivalent to those expended on atomic energy been devoted to the utilization of solar energy. That same statement could have been made in 1965 and in 1975, and it can be made today. However, there is hope. The awareness about the environmental concerns is growing fast, thanks to various environmentalists groups and other non profit organizations, and to the information technology. Even p rivate sector companies seem to be taking notice. An example is the automotive industry. Most experts agree that within the next handful of years, consumers will see fuel cell vehicles 100% clean engines that run on hydrogen and produce only water as a b yproduct hit the roads. [ Why is BMW driving itself crazy? Sue Zesiger, Fortune, 2000, www.fortune.com ] In fact, a recent publication of possible energy scenarios up to the year 2060 predicts a multi Gigawatt energ y production by renewable technologies [Shell, 1997]. What then remains to be

PAGE 31

18 seen is whether renewables will be used to their fullest potential, and whether they are able to replace the harmful fossil fuels in the near future. Because, in the words of M ichael Oppenheimer, Senior Scientist of the Environmental Defense Fund: "We have an obligation to weigh the risks of inaction against the cost of action. In that regard, global warming is no different than any other problem. But global warming is novel in one respect. It brings with it the possibility of a global disaster, and we have only one Earth to experiment on."

PAGE 32

19 CHAPTER 2. BACKGROUND I: BASIC PHYSICS 2.1. Whats a Solar Cell? Photovoltaics (PV) is the conversion of light into electricity. When the source of light is the sun, the process is called Solar Photovoltaics (Although the solar part will be assumed hereafter). Because of the simplicity of this process, and the abundance of the source that it uses, it appe ars to be one of the most promising ways of meeting the increasing energy demands of our planet. A solar cell, of the type that is used in this research, is essentially formed by sandwiching together a p type semiconductor and an n type semiconductor. Meta llic contacts are made to both these semiconductors. The semiconductors are chosen in such a way that, when light is shone on the device, one of them will absorb a significant portion of the light. Absorption of the light creates mobile carriers, both ne gative (electrons) as well as positive (holes) in the material. Ordinarily, such generated carriers recombine in a semiconductor. However, a good solar cell is designed in such a way that most of these generated carriers, after they are swept across the junction, are collected by the metallic contacts. Such carriers are then made to flow in an external circuit, and their energy can be utilized. The phenomenon of solar energy conversion thus involves the processes of absorption of radiation, generation o f carriers, transport of these carriers to the junction, separation of the carriers at the junction, collection of the separated carriers, and finally the utilization of the power generated.

PAGE 33

20 A brief survey of the basic properties of p n junctions, and the electronic processes involved when light interacts with a solar cell, is in order. 2.2. A P N Junction in the Dark An n type semiconductor material has a large concentration of electrons, and a few holes, whereas a p type semiconductor has a lot of holes and a few electrons. When two such materials are appropriately joined together, diffusion of carriers takes place because of the large concentration gradients at the junction. Each electron leaving the n side leaves behind an uncompensated positively c harged donor ion, and every hole going across the junction leaves a negatively charged acceptor ion. These ionized donors and acceptors, present in the region depleted of carriers (called the depletion region W), build up an electric field. This field is set up such that it creates a drift component of current that opposes the diffusion of carriers. At thermodynamic equilibrium, when there is no net flow of charge across the junction, an equilibrium contact potential, also called a Built in potential, V 0 is thus set up across the depletion region. This potential difference produces a bending of the energy bands of the semiconductors. Such a band bending is shown in the following figure, for the case of a homojunction p n diode (made by using p and n d oped parts of a single semiconductor).

PAGE 34

21 Figure 2.1. Formation of a Homojunction p n Diode (a) Metallurgical Junction, (b) Electrostatic Potential, and (c) Energy Bands. When an external voltage is applied to the p n junction diode shown above, one of t wo things can happen. If the bias is forward, i.e. a positive voltage V f is applied to the p side, the height of the potential barrier is reduced from V 0 to V 0 V f thereby reducing the band bending. This increases the diffusion current of majority carr ier electrons from the n side surmounting the barrier to diffuse to the p side, and holes surmounting their barrier from p to n. A large current, directed from the p to the n side, hence, flows in the

PAGE 35

22 forward bias. On the other hand, if a reverse bias i s applied to the junction (p side negative with respect to the n side), the bend bending increases, and this decreases the diffusion current from p to n to negligible values. The other current component that flows in a p n junction is the so called genera tion current, which is directed from n to p (opposite to the diffusion current). This is the drift component, composed of minority carriers from both the sides of the junction, and is relatively insensitive to the height of the potential barrier. These m inority carriers are generated by thermal excitation of electron hole pairs (EHPs), at or near the junction, and are swept to the other side of the junction because of the electric field. In reverse bias, this is the only current present (because the dif fusion current is negligible), and hence this current component is sometimes referred to as the reverse saturation current, I 0 with the corresponding current density denoted by J 0 Note that the letters I and J, hereafter, will refer to the currents and the corresponding current densities. The total current density in a p n junction in the dark can be written as: Where the J and J 0 designate the total and the reverse current densities, respectively, V is the applied voltage, k is the Boltzmann consta nt, q is the electronic charge, and T is the absolute temperature. As can be easily seen, at equilibrium (V = 0), the net current is zero. The above equation defines the I V characteristic of the junction diode, which is shown graphically in the next sec tion. = 1 exp 0 kT qV J J (Eq. 2.1)

PAGE 36

23 2.3. Interaction with Light 2.3.1. Photocurrent When light is shone on the junction, the photons that have energies greater than the bandgap of the semiconductor have a high probability of being absorbed. The absorption of light can be described by relating the radiation intensity I 0 falling on a semiconductor surface to the intensity I that remains after the light has penetrated a distance x: ( ) ( ) [ ] x I I l a l = exp 0 The parameter a which is a function of the wavelength of the light, is a charac teristic of the material, and is called as the absorption coefficient. The value of the absorption coefficient must be high for the absorber material used in a solar cell device, so that most of the light is absorbed in a useful way. Each photon that is absorbed in the absorber material generates an EHP. Such minority carriers, if generated within a certain distance of the junction (called a diffusion length), can diffuse to the junction, be swept to the other side, and be collected by appropriate contac ts. When a monochromatic light of wavelength l is incident on the surface of a solar cell, the photocurrent and spectral response, that is, the number of carriers collected per incident photon at each wavelength, can be derived as follows [Sze, 1981]. Th e generation rate of electron hole pairs at a distance x from the semiconductor surface is given by: (Eq. 2.2)

PAGE 37

24 ( ) ( ) ( ) ( ) [ ] ( ) [ ] x R F x G l a l l l a l = exp 1 Where F( l ) is the number of incident photons per cm2 per unit bandwidth, and R( l ) the fraction of these photons reflected from the surface. Using the appropriate boundary conditions, and assuming low injection conditions, the internal spectral response (SR) is given by ( ) ( ) ( ) [ ] ( ) ( ) ( ) [ ] l l l l l l dr n p J J J R qF SR + + = 1 1 Where J p ( l ), J n ( l ), and J dr ( l ) are the photocurrent contributions from the p region, the n region, and the depletion region, respectively. Once the SR is known, the total photocurrent density obtained from the solar spectrum distribution F( l ) is given by ( ) ( ) [ ] ( ) l l l l l d SR R F q J m L = 1 0 The generation of EHPs because of light gives rise to an added generation rate g op given in EHP/cm 3 s, which produces a current from the n to the p side (opposite to the dark forward diffusion current). If L P and L N are the diffusion lengths for the minority carrier holes and electrons, respectively, then the resul ting optically generated current for a junction of area A cm 2 and depletion region width w can be written as: Figure 2.2 depicts the current generation in the p n junction under illumination. Since this current is from n to p, it subtracts from the tota l current from p to n. ( ) w L L qAg J N P OP L + + = (Eq. 2.3) (Eq. 2.4) (Eq. 2.5) (Eq. 2.6)

PAGE 38

25 Figure 2.2. Photocurrent Generation in a p n Homojunction Solar Cell 2.3.2. I V Characteristics The resulting I V characteristic of the diode, in dark as well as in light, is shown in the following figure. Figure 2.3. I V Chara cteristic of an Ideal p n Junction Solar Cell

PAGE 39

26 Depending upon the intended application, the diode can be operated either in the third or the fourth quadrants of the I V characteristic. Power is delivered to the device from the external circuit when the cur rent and junction voltage are both positive or both negative. If operated in the fourth quadrant, however, power is delivered from the junction to the external circuit, and this is the principle of operation of a solar cell device. The next figure shows an equivalent circuit of a solar cell. The generation of the photocurrent I L is represented by a current generator, in parallel with a diode that represents the p n junction. Figure 2.4. An Equivalent Circuit of a Sola r Cell There are two resistances shown in the above figure. R S is the series resistance, which should ideally be zero, but always exists, in a practical solar cell. It involves the bulk resistance of the absorber semiconductor, as well as any other resi stances in the device such as those coming from the contact materials used. The parallel (or shunt) resistance R P represents any parallel paths for the junction current to flow (an example is metal particulates shunting the junction). Ideally, such para llel paths shouldnt exist, making R P infinite. I L I D R P R S

PAGE 40

27 For simplicity, let us assume that the current generated by light can be added to the current flowing in the dark (superposition), and also that R S = 0, and R P = (ideal case). Then, the current density J flowing in the device in the presence of light can be expressed as: Here, the first term on the right is the forward current driven by the voltage V, and the second term is the (reverse) light generated counterpart. J 0 is often referred to as the rever se saturation current. A few important terms that are commonly used as measures of solar cell performance need to be defined. The short circuit current density J SC is simply the light generated current J L The open circuit voltage can be obtained by se tting J = 0. It can be easily seen that, while the J SC depends only on the light assisted generation, the V OC depends on the current generation recombination processes as well as on the nature of the junction transport (A and J 0 ). Both I SC and V OC are shown in the I V characteristic above. No power can be generated under short or open circuit. The maximum power P MAX produced by a device is reached at a point on the characteristic where the product L J AkT qV J J = 1 exp 0 + = 1 ln 0 J J q kt V L OC L SC J J = (Eq. 2.7) (Eq. 2.9) (Eq. 2.8)

PAGE 41

28 IV is maximum, that is, when the area covered by t he power rectangle shown in the figure is maximum. The Fill Factor (ff) is defined as: The FF, therefore, is a measure of the squareness of the characteristic. The efficiency h of a solar cell is defined as: Which, in terms of V OC J SC and FF, beco mes: Where P RAD is the power of the radiation incident upon the cell. The standard conditions used to calculate the solar cell efficiency are: an irradiance of 100mW/cm 2 standard reference AM1.5 spectrum, and a temperature of 25 0 C. When we consider a practical solar cell, the above equation for the current transport has to be modified. Real cells usually have a non zero series resistance R S and a finite shunt resistance R P The equation for the current I (= J Area) then becomes: I = I 0 {exp[q(V IRs)/AkT] 1} + (V IRs)/Rp ( ) OC OSC MAX MAX V I V I ff FillFactor = RAD MAX P P = h RAD OC SC P ff V J = h (Eq. 2.11) (Eq. 2.10) (Eq. 2.12) (Eq. 2.13)

PAGE 42

29 The factor A, in the denominator of the exponential, is the so called ideality factor which relates to the mechanism of the junction transport in a practical device. The value of A usually varies between 1 and 2. A value of 1 usually means that the junction transport is by diffusion, whereas a value of 2 signifies that the transport is controlled by recombination in the depletion region. If the values of R S and 1/R P are significant, then the I V characteristic of the device gets affected, as shown in the next two figures. To the first order, V OC is unchanged by a reasonably low R S whereas I SC decreases slightly. On the other hand, a finite R P usually decreases V OC while I SC is unaffected. Figure 2.5. Effect of a Non Ze ro Series Resistance on the I V Characteristic

PAGE 43

30 Figure 2.6. Effect of a Finite Shunt Resistance on the I V Characteristic The most common types of junction that are used to form solar cells are: (i) Homojunction: p n junction within the same semiconductor m aterial. (ii) Heteroface structure: similar to a homojunction, but with an added window layer made of a larger band gap semiconductor. (iii) Heterojunction: p n junction between two different semiconductor materials. (iv) Schottky barrier: metal semiconductor junction. 2 .4. Heterojunctions When semiconductors of different bandgaps and electron affinities are brought together to form a junction, as in a heterojunction, discontinuities are produced in the

PAGE 44

31 energy bands, as the Fermi level of the different materials line up at equilibrium. The discontinuities in the valence ( D E V ) and the conduction bands ( D E C ) accommodate the difference in the bandgaps. Figure 2.7 (on the next page) shows an example of such a heterojunction system, and the important parameters, before the t wo semiconductors are joined together. The band bending that occurs after the two are joined together is depicted in Figure 2.8. It can be seen, from the resulting band bending, that a spike has appeared in the conduction band, where the two materials mee t. Such a spike is the result of properties specific to the materials used, such as the electron affinities c s. A discontinuity such as this limits the electron current that flows from the p side to the n side when the solar cell is placed in the light, and hence should be avoided by proper selection of the semiconductor materials, and by appropriate processing. Figure 2.7. Two Materials, Before Heterojunction Formation

PAGE 45

32 Figure 2.8. Heterojunction Formation It should be noted that, in heterojunction s, the light can either be incident on the larger band gap material (backwall type) or on a thin layer of the smaller band gap material (frontwall type). Similarly, in Schottky barriers, it is possible to have the light incident on either the semitranspar ent metal forming the barrier (frontwall), or through the semiconductor (backwall) [Bube].

PAGE 46

33 The solar cells that are the topic of this research can be considered to be backwall type heterojunction solar cells. A heterojunction can be either isotype, where both the semiconductors have the same type of conductivity, or anisotype, where the conductivities are different. Again, the CuGaSe 2 /CdS junction used in this research belongs to the latter type, viz. anisotype. The principle advantage of using a direct bandgap heterojunction system for a solar cell, such as the one used in this research can be seen as follows. Consider the case of an indirect bandgap homojunction solar cell, an example of which is a Silicon solar cell. Here, because of the low absorpt ion coefficient associated with the indirect bandgap, a large thickness of the material is needed to absorb enough light. If we consider replacing this system by a direct bandgap homojunction system, another problem arises. Because the light needs to be absorbed as close to the junction as possible (so that the generated carriers are easily collected by the junction), the top layer (say, n layer) needs to be fairly thin, with a thicker p layer underneath it. In such a structure, the carriers generated in the n layer have a high probability of diffusing away from the junction, towards the front contact, and eventually getting lost because of the high surface recombination velocity at the contact surface. Now, if the system is a direct bandgap heterojuncti on, then it can be designed in such a way that the top n layer is made of a wider bandgap material which will absorb little light in the spectrum of interest. Most of the light hence will reach the junction and the underlying p type absorber, thereby sign ificantly reducing the likeliness of surface recombination at the front contact. The p CuIn(Ga)Se 2 /n CdS solar cell structure has been optimized in this way, and the resulting band diagram is shown below, along with the n type ZnO which acts as the

PAGE 47

34 front contact. The CuGaSe 2 /CdS cell has a very similar structure, except that the conduction band of the absorber is raised further above, a direct effect higher bandgap of CuGaSe 2 Figure 2.9. Band Diagram of the CuIn(Ga)Se 2 /CdS/ZnO Solar Cell For heteroj uction solar cell structures, an added complication is the increased defect states at the interface. These mainly arise because of the lattice mismatch between the two semiconductors. However, processing conditions may also have a strong effect. Therefo re, unlike in homojunctions, the carrier transport properties in heterojunctions are usually dominated by phenomena in the interface region. The current transport in the

PAGE 48

35 depletion layer has been attributed to either the recombination, or the tunneling, or a combination of both. This transport is aided by the defect energy levels at or near the interface. A large density of electrically active interface states provides two mechanisms: (i) The charge stored in these states distorts the band profile, and (i i) The states give rise to a high density of recombination centers, thereby producing high forward current (J 0 ) values. In some cases, the extremely high density of charged states at specific energy levels at the interface is sufficient to pin the surface (or interface) Fermi level at that energy.

PAGE 49

36 CHAPTER 3. BACKGROUND II: THIN FILM PHOTOVOLTAICS 3.1. Historical Background of Photovoltaics Becquerel reported the photovoltaic effect in 1839, when he found that a light depen dent voltage developed between electrodes immersed in an electrolyte. In 1876, this effect was observed in an all solid state Selenium system. Subsequent work on the PV effects in selenium and cuprous oxide led to the development of the selenium PV cell that was widely used in photographic exposure meters. The modern era of PV began in 1954, when Chapin et al., at the Bell laboratories, successfully developed a silicon single crystal solar cell. This device represented a major development because it was the first photovoltaic structure that converted light to electricity with a reasonable efficiency (6%). Until the 1960s, the main interest in the development of solar cells was their application as power sources in spacecraft. The early 1970s saw a g rowing interest in the development of PV technologies for terrestrial use. More recently, the focus has shifted from single crystal technology to the low cost alternative of thin film technology. Before going into the specifics of PV technologies, it will be worthwhile to outline the general requirements for such a technology. The most important of these are listed below.

PAGE 50

37 (i) Conversion efficiency should be high for laboratory cells as well as for modules. (As the cost decreases, this requirement becomes les s important. However, realistically, to keep the area related costs down, module level efficiencies of at least 10% are necessary.) (ii) Constituent materials (semiconductors, metals) should be readily available, and should be inexpensive. (iii) A simple but reprodu cible deposition method that is suitable for large area production should be available. (iv) The cells/modules must be stable over long periods of time. (v) Total (capital + maintenance) cost should be low. (vi) Constituent materials should be non toxic/environmentally friendly. Currently, the most widely used PV technologies are the single crystal silicon technology and the polycrystalline silicon technology. Together, these two forms of silicon constitute about 86% of the solar cell market today [Goetzberger]. Howeve r, there are distinct disadvantages of using silicon as the absorber material for solar cells, as described below. Silicon is an indirect bandgap semiconductor, with a relatively low absorption coefficient for absorbing sunlight. Consequently, a considera ble thickness (about a 100 microns) of silicon is needed to absorb the light, thereby increasing the material cost. This, in turn, means that the photogenerated carriers have to traverse long distances to reach the junction, which is near the front surfac e. The diffusion length of the minority carriers has to be very high, which can happen only when the material is of very high purity and of high crystalline perfection. This, then, increases the processing costs.

PAGE 51

38 Moreover, the single crystal or polycrys talline silicon wafers are cut from ingots grown by the Czochralski method, or by controlled solidification in a crucible or mold. Sawing of these wafers results in material loss, adding to the total material cost. Although the laboratory efficiencies of these cells have exceeded 24%, the commercially available module efficiencies are usually limited to less than 16%. It is surprising that, in spite of these shortcomings, silicon is the dominant solar cell technology. At least a major part of the reason l ies with the fact that this technology has benefited tremendously from the high standard of silicon technology that was originally developed for transistors, and later for integrated circuits. The resulting silicon based solar cells have exhibited high ef ficiency and good stability. The best laboratory efficiency for a single crystal silicon solar cell is 24.5% [Green, 1999], while the best production cells have efficiencies of 15 16%. However, the resulting electricity costs are still relatively high, w hen compared to the cost of conventional electricity, making it necessary to look for new materials and technologies to replace silicon. One alternative to the silicon technology that has been extensively investigated is the gallium arsenide (GaAs) technol ogy. GaAs is a direct bandgap semiconductor with a high absorption coefficient, with the bandgap of 1.43 eV that is well suited to the solar spectrum. The effect of the direct bandgap is easily appreciated when it is recognized that, for a 90% light abso rption, it takes only 1 m m of GaAs, versus 100 m m of silicon. This technology, however, is quite expensive, and, as a result, more and more scientists and researchers are getting interested in the development of the low cost alternative, viz. thin film ph otovoltaics.

PAGE 52

39 3.2. Thin Film Photovoltaics The three thin film technologies that hold the greatest promise are: Amorphous Silicon (a Si), Cadmium Telluride (CdTe) and Copper Indium Diselenide (CuInSe 2 ). The a Si technology, which uses a silicon hydrogen a lloy (containing 20 30% hydrogen) as the absorber material, has been around for a couple of decades. The first amorphous solar cells were prepared in 1976 [Carlson, 1976]. The a Si technology currently dominates the thin film photovoltaics market. (The d ominance of silicon in its crystalline and amorphous forms is an overwhelming 99% of the total photovoltaics market. Most of the remaining 1% is taken up by CdTe, with CuInSe 2 only recently beginning to show up on the commercial scene.) The cuprous sulfi de/cadmium sulfide heterojunction was the first all thin film photovoltaic system developed. Currently, two of the most promising thin film polycrystalline technologies are Cadmium Telluride (CdTe) and Copper Indium Diselenide (CuInSe 2 ), both of which use Cadmium Sulfide (CdS) as the (n type) heterojunction partner. Both CuInSe 2 as well as CdTe are direct bandgap materials. Such polycrystalline thin film PV technologies offer several advantages, which can be weighed against the shortcomings of single crys tal and poly silicon cells that are listed above. These are: (i) Thin film technologies often involve semiconductor materials that have direct bandgaps, and hence have very high absorption coefficients for the wavelengths of interest. Therefore, only a small thickness, usually a few

PAGE 53

40 micrometers, is enough to absorb all of the sunlight incident on the absorber layer. This provides for significant savings in the material costs. (ii) Because of the low consumption of the active solar cell material, rare and expensiv e materials can be considered. (iii) A variety of relatively inexpensive vacuum deposition techniques can be employed for the processing of thin film solar cells, thereby reducing the processing costs. These techniques include RF and DC magnetron sputtering, va cuum evaporation, close space sublimation, etc. (iv) There are no small wafers to wire together, while making solar cell modules. Separate cells can be monolithically integrated on the module by scribing steps between depositions [Gabor, 1995]. This makes pac kaging and wiring easier, and also allows high voltage to be produced with smaller areas [Goetzberger]. (v) Thin films can be deposited on flexible, lightweight substrates, thereby making the cells viable for a larger variety of applications. Based on this lis t of desirable properties, one might begin to think that thin film technologies are clearly the one solution that will get rid of all the hurdles that PV faces. However, in spite of all these advantages, these technologies havent been able to get the ele ctricity cost down enough, thanks to the following shortcomings. (i) Most of the thin film technologies involve heterojunctions, and hence face the problem of faulty interfaces, arising because of lattice mismatches between the materials. (ii) Difficulty of getting different films to adhere to each other well.

PAGE 54

41 (iii) Difficulty in achieving uniformity of thickness, composition, and quality across a large substrate. (iv) Difficulty in achieving stability of the films over many years. (v) Toxicity of some constituents involved (for i nstance, Cd in the case of CdTe/CdS, and, to a lesser extent, CuInSe 2 /CdS). Although thin film PV technology is still in its infancy, both CdTe as well as CuInSe 2 technologies have shown tremendous promise. Laboratory efficiency numbers have exceeded 18% for CuIn(Ga)Se 2 and 15% for CdTe, whereas commercially available thin film modules have shown conversion efficiencies in the neighborhood of 10 12% One of the main problems facing the CdTe technology is the toxicity of cadmium. This necessitates end of life recycling programs for CdTe modules, thus adding to the total cost. CuInSe 2 on the other hand, has consistently passed the toxicity tests, and hence can be thought of as the leader among all current thin film technologies. 3.3. CuInSe 2 Family Base d Thin Film Photovoltaics 3.3.1. CuInSe 2 Family and Device Issues The CuInSe 2 family of thin films belongs to the I III VI class of thin film semiconductors, and has shown great promise for photovoltaic applications. With a direct bandgap of 1.0 eV, CuI nSe 2 has the highest reported absorption coefficient of

PAGE 55

42 about 3.6 x 10 5 cm 1 The CuInSe 2 family includes several I III VI compounds, which will be described in the next section. The typical structure of a completed solar cell device based on the CuInSe 2 absorber material would be: Glass substrate/ Molybdenum / CuInSe 2 / CdS/ Zinc Oxide. Here, the junction is essentially formed between the p type CuInSe 2 and the n type CdS. The molybdenum and zinc oxide thin films are used as the back contact and the fro nt contact, respectively. The light is incident from the front (zinc oxide) side. CuInSe 2 is a I III VI ternary (i.e., three elements) compound. The following figure shows the so called chalcopyrite structure of this compound, which, essentially, is a di amond like lattice made up of face centered tetragonal unit cells. Figure 3.1. The CuInSe 2 Structure [Zhang, 1998]

PAGE 56

43 CuInSe 2 is a self doped (intrinsically doped) material, which means that, when the compound is formed, it automatically becomes either p or n type, depending upon the composition. The primary intrinsic defects, which are also called native defects, include copper vacancies (V Cu ), copper on indium (Cu In ) antisites, indium on copper antisites (In Cu ), and selenium vacancies (V Se ). The forme r two produce acceptor type defects, whereas the latter two give rise to donor type defects. Depending upon the ratio of the Group I to Group III (commonly referred to as the metal ratio ), the CuInSe 2 material can be made either Cu rich or In rich. Cu ri ch material is highly conductive, mainly because of the presence of unreacted, and highly conductive, copper selenide species. This material is generally p type, due to a large concentration of Cu In defects. The performance of the solar cells that have C u rich absorber layers is usually diminished. This has been attributed to the above mentioned copper selenide forming between the grain boundaries, thereby shorting the p n junction. The In rich CuInSe 2 material, on the other hand, does not contain coppe r selenide species. This type of material can be either n or p type. Usually, the In Cu donor defects and the V Cu acceptor defects are present in this material at the same time, reducing the conductivity of the layer (the so called compensation effect). The efficient self doping ability of CuInSe 2 has been attributed to the exceptionally low formation energy of Cu vacancies and to the existence of a shallow Cu vacancy acceptor level [Zhang, 1998]. In general, there are two methods that have primarily bee n used to carry out the vacuum physical vapor deposition of CuInSe 2 absorber films. One predominant method is the simultaneous co evaporation of all the elements onto the substrate material. High quality thin films can be obtained by this technique. In fact, the best CuInSe 2 device

PAGE 57

44 reported in the literature has been fabricated at NREL using a three stage co evaporation approach. After an anti reflective coating, the device parameters reported were the following: Area = 0.395cm 2 h = 13.2%, V OC = 484mV, J SC = 36.29 mA/cm 2 and FF = 75.10% [Contreras, 1994]. A small area of co evaporated CuInSe 2 thin films was reported to have an efficiency of more than 15% [Tuttle, 1996]. There are, however, some disadvantages of the co evaporation technique. Firstly, the technique requires a very high control of deposition parameters, especially because three, or sometimes even four, elements are being deposited at the same time. Secondly, evaporation, by its very nature, is an expensive process, in terms of material usage. Thirdly, large area uniform depositions are difficult to achieve with evaporation (compared to, say, sputtering). This, therefore, makes it difficult to scale this method for a high volume commercial production. The other approach that is being e xplored for the formation of the CuInSe 2 type absorbers usually includes two steps. The first step involves the deposition of the so called precursors which essentially are alloys of copper and indium (and sometimes gallium), by a physical vapor depositi on technique such as evaporation or sputtering. In the second high temperature step, commonly referred to as selenization the precursor films are exposed to a high flux of selenium containing vapors, by using either elemental selenium, or a selenium comp ound such as hydrogen selenide (H 2 Se). With this method, it is sometimes difficult to achieve a highly homogeneous absorber film. Although high efficiency solar cells have been fabricated using this method, the performance is usually inferior to that obt ained by the co evaporation technique.

PAGE 58

45 In a recent paper, Kim, et al. reported a study of the CuInSe 2 selenization parameters [Kim, 2000]. The CuInSe 2 absorber was prepared using a two step method. To accomplish a homogeneous precursor layer, they used DC magnetron co sputtering from a Cu In alloy. Only two phases, Cu 11 In 9 and CuIn 2 were formed over a wide range of compositions, suggesting that a high degree of elemental mixing occurred. (Others have reported existence of other phases, such as Cu and Cu In in sputtered or evaporated bi layer or multi layer Cu In precursor films.) The selenization of the precursors involved two stages. In the first stage, selenium was incorporated into the alloy precursor film at a lower temperature of 250 0 C. The sec ond stage involved a re crystallization process, which was performed at an elevated temperature of 400 550 0 C. The authors then compared selenization at two different pressures: 10mTorr vacuum, and 1atm. At atmospheric pressure, the scattering because of the Argon present in the system reduced the energy of the Se atoms. These low energy atoms induced localized reactions, resulting in several intermediate compounds. In vacuum, on the other hand, the Se atoms had higher energy, and could migrate easily on the surface to promote a reaction with the metals. This indicated that the formation of CuInSe 2 single phase needed higher temperature treatment to obtain enough Se energy. Several recent studies have identified the presence of a thin (a few hundred ang stroms) n type layer at the surface of the CuInSe 2 absorber films. Such layers have been referred to as ordered vacancy compounds (OVCs), ordered defect compounds (ODCs) or chalcopyrite defect compounds (CDCs). The improved CuInSe 2 device performance has been attributed to these defect layers. Abulfotuh, et al. characterized CuInSe 2 layers using photoluminescence, and detected the presence of a 250A 0 thick In

PAGE 59

46 rich defect layer. They also found that this layer, which was previously identified as CuIn 3 Se 5 had a gradual change of composition (Cu content) with depth, which resulted in a gradual change in the optical properties of the films [Abulfotuh, 1996]. Later, Zhang, et al. showed that the ODCs in CuInSe 2 resulted from the unusual stability of a s pecial defect pair: (In Cu 2+ + 2V Cu ), i.e., two Cu vacancies next to an In on Cu antisite. Evidently, a periodic spatial repetition of this pair gives the ODCs [Zhang, 1998]. The electrically benign character of the large defect population in CuInSe2 ha s also been explained in terms of an electronic passivation of the In Cu 2+ by the 2V Cu Such a special defect pair can be seen in the following figure. Figure 3.2. The Structure of a Special Defect Pair [Zhang, 1998]

PAGE 60

47 3.3.2. Bandgap Engineering The open circuit voltage produced by a solar cell is proportional to the band bending that is produced at the junction where the p and n layers meet. This band bending, in turn, increases as the bandgap of the absorber layer increases. However, such an inc rease in V OC with an increase in the absorber bandgap, comes at the cost of a reduced short circuit current. This happens because, as the bandgap is raised, the minimum energy of photons that can generate carriers in the semiconductor increases. Hence, fewer photons can now be useful for current generation. As it turns out, however, solar cell modules can actually benefit from this effect, because when the current (I SC ) decreases, the resistive losses in the modules also decrease. In addition to these effects on V OC and I SC there are a few more advantages of having a high bandgap absorber layer, and these will be discussed in a later section dealing with CuGaSe 2 One of the advantages of compound semiconductor thin film technologies is that different c ompounds can be alloyed together to form newer compounds, to achieve the desired material properties. The CuInSe 2 family includes a variety of such ternary semiconductor compounds that can be considered for alloying, some examples of which are tabulated o n the next page (Table 3.1), along with their respective bandgaps values [Gabor, 1995].

PAGE 61

48 Table 3.1. Various Ternary Absorber Materials with their Bandgaps [Gabor, 1995] Absorber Compound Bandgap (eV) CuInSe 2 1.04 CuGaSe 2 1.70 CuAlSe 2 2.70 CuInS 2 1.55 AgInSe 2 1.24 CuInTe 2 0.95 Recent calculations show that, given the shape of the AM1.5 global solar power spectrum, the ideal bandgap for efficient solar energy conversion is around 1.14eV [Ward, 1993]. It will, therefore, be advantageous to all oy CuInSe 2 with higher bandgap materials to achieve a better match to the solar spectrum. One of the most promising ways to increase the bandgap of CuInSe 2 is to incorporate gallium (Ga). It can be seen, from the first two entries of the above list, that if In is replaced by Ga, the bandgap of the absorber increases from 1.04 to about 1.70eV, a jump of 0.66eV! The specific advantages of the CuGaSe 2 material will be discussed later. An extremely important case, however, is a quaternary compound CuIn(Ga) Se 2 where only a part of the In in the basic CuInSe 2 is replaced by Ga, so that the resulting material becomes an alloy of CuInSe 2 and CuGaSe 2 The bandgap of this alloy material can be varied between the two extreme values mentioned above, and it can al so be varied as a function of the depth within the absorber layer itself. This CuIn(Ga)Se 2 material has been extensively explored in the recent past, and is discussed next.

PAGE 62

49 3.3.3. CuIn(Ga)Se 2 CuIn(Ga)Se 2 solar cells, prepared by incorporation of a cont rolled amount of gallium, have recently reached 18.8% efficiency, which is the highest efficiency ever reported for a thin film solar cell [Contreras, 1999]. As mentioned in the last section, when In in CuInSe 2 is replaced by Ga, the bandgap tends to incr ease. Evidently, this is an effect of the smaller size of the Ga atom (when compared with In), and the various formation energies involved. Albin carried out optical absorption measurements, and determined the bandgaps of CuInSe 2 CuGaSe 2 alloys over the full range of compositions [Albin, 1990]. According to him, for CuIn 1 X Ga X Se 2 the bandgap varies according to ( ) x x x E g + = 1 249 0 664 0 011 1 (Eq. 3.1) and, for films with slight Cu deficiencies, the relation becomes linear, with ( ) x E g + = 1 71369 0 0032 1 (Eq. 3.2) Both these functions are shown in the following figure. Figure 3.3. The CuInSe 2 Structure [Albin, 1990]

PAGE 63

50 With the addition of Ga, and the corresponding change in the bandgap, some of the material properties also change. T hese include structural properties like lattice constants, film morphology and adhesion, and chemical changes such as defect levels, affinities and carrier concentrations. Therefore, the In to Ga ratio must be optimized to achieve the appropriate set of f ilm properties, and thereby obtain highest possible performance of the solar cell devices. At present, the best CuIn 1 X (Ga X )Se 2 solar cells are made with x <= 0.3. Wei, et al. have listed the effects of Ga addition to CuInSe 2 which are the following [We i, 1998]. First, Ga incorporation increases the bandgap, according to [Albin, 1991]: ( ) ( ) ( ) ( ) ( ) x bx xE E x x E CuGaSe g CuInSe g g + = 1 1 2 2 (Eq. 3.3) where b is the (measured) bowing coefficient that depends on growth. The theoretical value of b has been calculated to be 0 .21, in good agreement with the most reproducible experimental values of 0.15 to 0.24 eV. Second, the hole concentration in the stoichiometric 1:1:2 compound increases significantly. In addition, the stability domain of the 1:1:2 compound in the phase di agram increases, i.e., the chalcopyrite phase becomes more stable, while the 1:3:5 ordered defect compounds (ODC) now have a narrower domain of existence in the phase diagram. As x Ga increases, the cell efficiency initially increases. However, when x > 0 .3, the efficiency drops off. The 1:1:2 phase can no longer be made n type. It has been previously suggested that the reason for this performance deterioration is related to strain, which comes from the lattice mismatch between the 1:1:2 and the 1:3:5 ph ases at the interface, as x Ga goes over 0.3, causing structural defects. However, the calculation of Wei, et al. shows that the change of the

PAGE 64

51 lattice mismatch due to Ga addition is very small, and hence is unlikely to be the main reason for device deterio ration. As mentioned once before, the bandgap in the p type CuIn(Ga)Se 2 absorber layer can actually be engineered so that it changes rather gradually, from the metallurgical junction towards the inside of the absorber. This is made possible by changing t he Ga concentration (that is, the Ga/(Ga + In) ratio) with the depth of the film. An illuminating account of how this can be achieved has been given by Gabor [Gabor, 1995]. The effect can be briefly explained as follows. Figure 3.4. Ba nd Bending with (a) No Grading, and (b) Grading The above figure depicts two different structures, one with a single bandgap throughout the absorber layer, and the other with a graded bandgap (the bandgap increasing towards the back). As will be discusse d later, in CuIn(Ga)Se 2 the bandgap increase, arising because of Ga incorporation, seems to be accommodated by the conduction band edge moving upwards. If a single bandgap exists throughout the thickness of the absorber layer, then the band bending is co nfined to the front portion of the layer, in the region where the depletion region penetrates. (This, of course, depends (a) (b)

PAGE 65

52 on the doping level in the film.) Hence, there is no electric field outside this region, towards the back of the device. The minorit y carriers (electrons in the conduction band, in the case of CuIn(Ga)Se 2 ), that are generated outside the depletion region, therefore, must rely on the diffusion mechanism to reach the junction. If, however, the minority carrier diffusion length is small, compared to the depth of the absorber layer beyond the depletion region, then carriers generated far away from the depletion region (i.e., towards the back of the layer) have only a small probability of being collected, and of contributing to the photocur rent. On the other hand is the other structure, where a bandgap grading, and the resulting conduction band edge bending, exists. In this case, the quasielectric field helps the electrons move towards the front of the device, thereby increasing the probab ility of their collection. The above example demonstrates that it is important to gain a precise control over the composition throughout the depth of the film. By having such a control, the intended grading profile can be carefully accomplished. On the other hand, if no grading is intended, any grading can be avoided, using carefully controlled compositions. Any unintentional grading in the opposite direction (bandgap decreasing towards the back) may seriously hurt the collection efficiency of the devic e. Fortunately, the latter effect, of the existence of an opposite grading, is easily avoided in CuIn(Ga)Se 2 processing. It turns out that Ga in the films has a strong tendency to move towards the back of the device, with In staying at the front. Such an effect means that it is easy to get the bandgap to increase towards the back. However, this creates another challenge for the processing engineers. Because the Ga will always try to go deeper, it becomes more difficult to create a thin higher bandgap la yer at the

PAGE 66

53 front, when one intends to utilize the increased bandgap offered by CuGaSe 2 to increase the V OC of the solar cell. 3.4. CuGaSe 2 Solar Cells 3.4.1. Advantages of a High Bandgap Before we go on to the specifics of the CuGaSe 2 solar cell devices it will be worthwhile to list the various advantages of having a higher bandgap for the absorber material (some of which have been mentioned before). (i) The open circuit voltage, obtainable from a solar cell, is proportional to the bandgap of the absorber. Hence, as the bandgap increases, so does the V OC The current density, on the other hand, decreases with an increasing bandgap. However, this loss in the current density means lower ohmic losses in the solar cell modules, which is an important advantage (ii) With increased V OC s, fewer cells are needed to obtain the given voltage. Consequently, the number of interconnects within the module is reduced, thereby lowering the optical losses. (iii) The relative loss of the open circuit voltage with increased temperatu re is significantly lower for wider bandgap materials [Nadenau, 1999]. (iv) High bandgap materials have the potential to be used in tandem solar cells. (v) (Specific to devices based on CuInSe 2 type absorbers.) In the solar cells based on CuInSe 2 the bulk of the s eries resistance comes from the ZnO, which is the

PAGE 67

54 common front contact/window material. To reduce this resistance, the doping in ZnO has to be made very high. However, this gives rise to another problem. At around 1.1 eV (approximately 1100nm), the high ly doped ZnO starts to show high free carrier absorption, thereby reducing the photocurrent in that wavelength range. If the bandgap of the absorber is increased significantly such that the photocurrent no longer depends on these high wavelengths, then th e loss because of the free carrier absorption does not hurt the device any more. This is easily accomplished with CuGaSe 2 as the relevant absorption wavelengths are well below the above mentioned range. 3.4.2. CuGaSe 2 Device Issues 3.4.2.1. High Bandga p and V OC Limitation With a bandgap of 1.68 eV at room temperature, CuGaSe 2 is a good candidate for high voltage single cell devices, as well as for the top cell in tandem systems. Due to its high optical absorption coefficient, it is suitable for thin f ilm applications. The typical device structure (one which has shown most promise for high performance) for a solar cell device using a CuGaSe 2 absorber layer is: Glass/ Mo Back Contact/ p CuGaSe 2 / n CdS/ ZnO Front Contact, which is essentially the same as that for a CuInSe 2 device, with the CuInSe 2 absorber layer replaced by CuGaSe 2 As mentioned previously, the higher bandgap of CuGaSe 2 means less current density. In the case of CuGaSe 2 only the photons below about 750 nm are absorbed

PAGE 68

55 strongly, leading t o fewer photo generated carriers. In CuInSe 2 the corresponding wavelength is about 1200 nm. The unused range (difference between the two) of almost 450 nm usually translates into a loss of more than 20mA/cm 2 for a typical laboratory device. Equations (3 .1) and (3.2), presented before, related the bandgap increase to the amount of Ga incorporated into CuInSe 2 In fact, a phenomenological relation can be given for the V OC s in Ga containing devices, which is [Nadenau, 1999]: mV q E V g OC 500 = (Eq. 3.4) Here E g is the bandgap, and q is the elementary charge. According to this relation, CuGaSe 2 with its bandgap of 1.68 eV, should exhibit voltages as high as 1.2 V. Until 1997, even after a couple of decades of CuGaSe 2 research, the b est V OC s were still limited to about 750 mV. Later that year, Nadenau et al. used a new processing approach, and succeeded in preparing CuGaSe 2 devices with a V OC of 870 mV and a conversion efficiency of 9.3 % [Nadenau, 1997]. This performance has been the best so far, for thin film polycrystalline CuGaSe 2 The cells in this case were processed using a newly optimized deposition temperature for the CdS buffer layer. For comparison, the best efficiency for a single crystal CuGaSe 2 solar cell device is 9 .7% [Saad, 1996]. As already mentioned, V OC s of 870 mV have been achieved for CuGaSe 2 solar cells. However, the theoretical and phenomenological models seem to indicate that there is another about 300 mV that should be obtainable for this material. A fe w studies have been carried out to investigate the limiting mechanisms that seem to have held the V OC s hostage, and to understand the transport mechanisms that are involved. An extensive review of these studies will now be presented. But before that, le ts revisit a couple of

PAGE 69

56 important aspects of the CuInSe 2 /CuGaSe 2 absorber materials, so that the survey of the results can be easily understood. Firstly, there is the change in the band diagram, and hence in the band bending, because of Ga incorporation. When a band gap increases, the change could be because of the valence band edge E V moving downwards, or the conduction band edge E C moving upwards, or a little of both. The evidence seems to suggest that the bandgap change in the case of CuGaSe 2 comes fro m the change in the electron affinity, and hence from the conduction band moving upwards. The second important aspect relates to the various point defects, the important ones being the vacancies, antisite defects and interstitials. It should be remembere d that these are native defects, and hence are not very easily controlled by changes in the processing conditions. Nevertheless, factors such as the formation energies of the defects have tremendous impact on the doping levels and other parameters that go vern the performance of the devices. Another aspect that is closely related with the defects is the doping inversion (the ODCs) that is present in the CuInSe 2 absorber. Such a layer could significantly alter the device performance. For example, if the inverted layer is thick enough, it essentially makes the junction a buried homojunction, rather than a true heterojunction. Because the electrical junction is now well below the metallurgical CuInSe 2 /CdS junction, it is relatively protected from any possi ble structural defects or strains that can form at the metallurgical interface between the two semiconductors. Whether or not such an inverted layer can be formed (and if it is formed, how much its thickness and degree of inversion is), therefore, will pl ay a significant role in determining the junction characteristics.

PAGE 70

57 3.4.2.2. Band Discontinuities and Surface Inversion Wei, et al. used a first principles band structure method to theoretically study the effects of Ga addition on the electronic and struc tural properties of CuInSe 2 [Wei, 1998]. Some of their findings are summarized below. The band offset D E V between the valence band maxima of CuGaSe 2 and CuInSe 2 was calculated to be only 0.04 eV, when CuGaSe 2 and CuInSe 2 each have their own equilibrium la ttice constants. Therefore, it could be concluded that the conduction band minimum (E C ) of CuGaSe 2 was about 0.6 eV higher than that of CuInSe 2 (The relation D E C = D E G D E V has been used.) This is shown in the Figure 3.5, where the band diagrams of th ree ternary chalcopyrite compounds, CuInSe 2 CuGaSe 2 and CuAlSe 2 are compared. The above calculation also suggested that p type doping in CuInSe 2 and CuGaSe 2 should be similar, while n type doping should be more difficult in CuGaSe 2 than in CuInSe 2 Fig ure 3.6 presents the calculated band offsets, in eV, between CdS, CuInSe 2 and CuIn 3 Se 5 [Zhang, 1998]. The offset between CdS and CuGaSe 2 can be easily visualized, with the CB of the absorber moving upwards by about 0.6 eV.

PAGE 71

58 Figure 3.5. Band Offsets of Three Chalcopyrite Compounds [Redrawn, Wei, 1995] Figure 3.6. Valence and Conduction Band Offsets [Redrawn, Zhang, 1998] The calculated defect formation energies D E of single acc eptor defects (V Cu V Ga and Cu Ga ) in CuGaSe 2 were found to be similar to their counterparts in CuInSe 2 meaning that the acceptor densities in the two compounds are similar. However, the formation 1.04 0.80 1.37 2.67 0.26 1.88 0.04 0.77 0.22 2.67 CuAlSe 2 CuGaSe 2 CuInSe 2 CuAlSe 2 2.42 1.07 0.31 1.04 0.34 0.17 1.21 CdS Cu InSe2 CuIn3Se5

PAGE 72

59 energies of single donor defects (Ga Cu 0 Cu i 0 ) in CuGaSe 2 were larger in CuGaSe 2 when compared to those in CuInSe 2 so that the donor density in CuGaSe 2 could be lower than that in CuInSe 2 under similar growth conditions. A comparison of the defect transition energy levels showed that the acceptor levels in Cu GaSe 2 were similar to, or slightly shallower than, those in CuInSe 2 suggesting the presence of slightly more holes in CuGaSe 2 On the other hand, the Ga Cu antisite donor levels in CuGaSe 2 were much deeper than the In Cu donor levels in CuInSe 2 This mean t that, as far as the contribution of III on I antisite defects to n typeness was concerned, CuGaSe 2 would be less n type than CuInSe 2 3.4.2.3. CuGaSe 2 Transport Mechanisms In a recent paper, Nadenau, et al. presented a systematic study of the electroni c transport mechanisms of CuGaSe 2 based solar cells [Nadenau, 2000]. They tried to relate these mechanisms to the stoichiometry deviations, the substrates, and the buffer layers. Their findings are discussed below in detail. The evaluation models used b y the authors are briefly mentioned, followed by their experimental results. First, the authors recognize that in the case of low Ga content (Ga/(Ga + In) below 0.3), the Fermi level at the surface of the absorber is closer to the conduction band, and hen ce this surface is inverted. They hence argue that this type inversion at the interface decreases the number of available holes at the interface, thereby diminishing interface recombination, so that the recombination in the space charge region (SCR) becom es the dominant loss mechanism. With the help of admittance spectroscopy, the authors show

PAGE 73

60 that this type inversion is not exhibited in CuGaSe 2 / CdS/ ZnO devices. The following heterostructure band diagram (reproduced from the publication), under an appl ied bias voltage, depicts the possible recombination paths in this solar cell structure. Figure 3.7. Band Diagram (a) at Equilibrium, (b) at an Applied Bias Second, the authors argue that the space charge in these solar cells extends into the CdS buf fer as well as the ZnO window layer, and a large density of electrons is

PAGE 74

61 available in the buffer layer due to the negative band offset between the absorber and the buffer material. The authors then go on to treat the Tunneling Enhanced Interface Recombinat ion as well as the Tunneling Enhanced Bulk Recombination These equations are then combined to formally describe the forward current density J as = = AkT qV AkT E J AkT qV J J a exp exp exp 00 0 (Eq. 3.5) with J 0 as the saturation current density. The open circuit vo ltage can then be given as SC a SC OC J J q AkT q E J J q AkT V 00 0 ln ln = (Eq. 3.6) If A, J SC and J 00 are independent of temperature, a plot of V OC vs T should yield a straight line and the extrapolation of this line to T = 0 0 K should give the activation energy Ea. This activation energy corresponds to the flatband barrier F b f ( F b p if the barrier height is not field dependent) in the case of interface recombination and to the bandgap energy Eg in the case of bulk recombination Also, when tunneling is important, the ideality factor becomes temperature dependent, and hence the authors use the J 0 equation to get ( ) ( ) 00 0 ln ln J A kT E J A a + = (Eq. 3.7) Here, the plot of A ln (J 0 ) Vs. inverse temperature 1/T should yield a straight line with a slope corre sponding to the activation energy Ea. This activation energy is the flatband barrier F b f and bandgap energy Eg, in the case of interface recombination and bulk recombination, respectively.

PAGE 75

62 The authors examined two sets of CuGaSe 2 samples: Cu rich and Cu p oor (i.e., Ga rich). Each of these categories had one Na containing and one Na free sample. The temperature dependence of the V OC was measured, and the activation energy Ea for the dominant recombination process was found from the extrapolation of the pl ot on the voltage axis. For the Cu rich devices, an Ea ~1.25eV was identified as the flat band barrier, assuming interface recombination (alternatively, this Ea could also be identified with the presence of Cu 7 Se 4 precipitates). In contrast, the Na conta ining Ga rich devices had an Ea ~ 1.6eV, thereby indicating that the recombination mechanism which limited V OC occurred in the volume (bulk) of the absorber material. The temperature dependence (125 350 0 K) of the inverse (diode) ideality factors (1/A) was used for the quantitative analysis of the I V data of the different CuGaSe 2 devices. The Cu rich, Na containing device showed the largest ideality factors over the whole temperature range, with the product AT being independent of temperature, suggesting that tunneling was the dominant recombination mechanism. These devices also exhibited higher values of saturation current densities. On the other hand, the Ga rich samples showed reduced values of saturation current densities near room temperature, and a lso lower values of diode ideality factor. These were fitted to obtain lower values of tunneling energy and of charge density, than those of the Cu rich samples. The authors then monitored the change in the capacitance and V OC during the light soaking of the samples using red light of wavelength > 630nm. For both, Cu rich/ Na free as well as Ga rich/Na containing, samples, the capacitance rose during illumination, indicating increased space charge density (and a corresponding decrease in

PAGE 76

63 the SCR width). The V OC however, behaved differently for the two devices. The Cu rich device showed a decrease, whereas the Ga rich device exhibited an increase in the V OC It was then concluded that tunneling enhanced interface recombination process dominated the Cu rich devices, whereas the dominant process in the Ga rich devices was recombination in the SCR without significant contribution from tunneling It was also proposed by the authors that the reduction of space charge density, in the case of the Ga rich devic es optimized with the improved CdS buffer deposition process, was the result of Cd diffusion. The formation of Cd Cu + according to the authors, might have compensated for the high concentration of negatively charged Cu vacancies (V cu ) within the defectiv e surface layer. This issue of the effect of the buffer layer deposition will be discussed in more detail in the next section. In the part II of the above mentioned study, Jasenek, et al. studied the electronic properties of the Cu rich and Ga rich CuGaSe 2 devices using admittance spectroscopy, DLTS, and C V measurements [Jasenek, 2000]. Using a recently determined band offset value D E V of 0.9 eV at the CuGaSe 2 /CdS interface, the energetic difference E F E V was calculated to be 0.8 eV, as shown in the foll owing figure.

PAGE 77

64 Figure 3.8. Equilibrium Band Diagram for the ZnO/CdS/CuGaSe 2 Structure [Jasenek, 2000] In the above model, the CdS buffer layer is assumed to be completely depleted. It can also be easily seen that there is no type inversion at the surf ace of the CuGaSe 2 absorber layer. For the Cu rich samples, the defect spectra resulted from two different emissions, A1 and A2, which were correlated to two different acceptor like bulk traps with activation energies of 240 meV and 375 meV, respective ly. Trap A1, with a concentration of 4 x 10 17 cm 3 eV 1 yielded the dominant emission, whereas trap A2 concentration was lower by a factor of 5. Air annealing was found to reduce the density of A1 to some extent, whereas the density of A2 was drastically reduced, thereby implying that annealing affected deeper traps more It was also suggested that the defect A2 might reflect a Ga vacancy V Ga and A1 might be correlated to either another transition of V Ga or a Cu Ga antisite defect.

PAGE 78

65 The Ga rich sample, o n the other hand, exhibited a tail like energy distribution of acceptor defects with the maximum lying at about 250 meV. This defect, it was concluded, provided the dominant recombination path in high efficiency CuGaSe 2 solar cells based on Ga rich absor bers. Another important result was that the performance limitation of these CuGaSe 2 devices mainly originated from the low electronic quality of the absorber and not that of the film surface. 3.4.2.4. Effect of Buffer Deposition It must be remember ed that, in the substrate type CuGaSe 2 solar cell preparation, the CdS buffer/heterojunction partner is deposited after the absorber layer. Hence, the substrate sees the atmosphere while it is being transferred from the absorber deposition vacuum chamber to the CdS bath. It may take several minutes before the sample is actually dipped in the CdS bath. A strong sensitivity of CuGaSe 2 based solar cells to air exposure time of the absorber surface after growth before deposition of the buffer layer has been reported leading to a drastic degradation of the device performance [Nadenau, 1997]. Such a strong dependence of the performance on the processing variables relating to the surface of the absorber suggests that the junction formation and/or the junction t ransport mechanisms in the case of CuGaSe 2 may correspond to those of a true heterojunction, rather than to a buried homojunction. Even in the case of a CuIn(Ga)Se 2 /CdS interface, with a buried electronic junction, intermixing has been observed at the in terface. For example, Heske, et al. carried out a combination of x ray emission spectroscopy and x ray photoelectron

PAGE 79

66 spectroscopy using high brightness synchrotron radiation of such a buried interface [Heske, 1999]. Samples were prepared by rapid thermal processing of Cu, In, Ga and Se layers, followed by chemical deposition of CdS. Intermixing processes involving S, Se and In were identified in the analysis. As mentioned earlier, the best performance for a CuGaSe 2 solar cell has been obtained by using a newly optimized buffer layer [Nadenau, 1997]. The older buffer deposition process had used a temperature of 60 0 C. The chemistry of the CdS bath was changed (the concentration of NH 3 in the bath was increased) so that the deposition temperature in the ne w process rose to 80 0 C. A CuGaSe 2 device that used a Ga rich absorber layer had the best performance. There are several possible reasons for this performance improvement resulting from the new CdS process, and these are outlined below. Nadenau, et al. co mpared two Cu rich samples treated with KCN after the deposition [Nadenau, 1999]. The KCN treatment was carried out to remove the unwanted Cu Se species present at/near the absorber surface. One of the samples went through a 60 0 C CdS process, while the o ther went through an 80 0 C process. The performance of the 80 0 C sample was actually diminished, compared to the 60 0 C sample (a trend opposite to that seen for CuGaSe 2 made using Ga rich absorber film). EDX linescans across the interfaces of these samples showed that the interaction between Cu and the buffer layer was much stronger for the 80 0 C CdS sample. From the analysis of the microstructures of these Cu rich and Ga rich samples, it could be concluded that:

PAGE 80

67 (i) For all devices based on Cu rich (KCN treated ) absorber layers (both, 60 0 C, as well as 80 0 C CdS), and also for the device based on Ga rich layer with 60 0 C CdS, the recombination at the interface was dominant. (ii) For Ga rich absorber devices with 80 0 C CdS, recombination in the space charge layer was domi nant. The interfacial region for this sample was found to be spatially enlarged, leading to a lower density of interfacial states. Moreover, the lattice mismatch between CuGaSe 2 and CdS was reduced due to the dominant cubic phase of CdS for the 80 0 C reci pe. Using XPS measurement results, it was also proposed that sulfur was incorporated in the place of selenium at the top of the absorber layer, thereby decreasing the valence band energy there. This created a front surface field close to the interface, p ushing the holes back into the absorber layer, thereby reducing the number of carriers contributing to the interfacial recombination. In another study (that has been previously cited) Nadenau, et al. have compared the diode ideality factors for two Ga rich Na containing samples with different CdS deposition temperatures [Nadenau, 2000]. The following figure, reproduced from the publication, shows the variation of the inverse ideality factors with the absolute temperature.

PAGE 81

68 Figure 3.9. Variation of the In verse Diode Factor with Temperature The 80 0 C CdS device exhibited the lowest values of A, which were below 2 at room temperature. The calculated tunneling energy, deduced using the tunneling theory (briefly outlined before), for the 80 0 C device (23 meV) was much lower than that for the 60 0 C device (42 meV). It was hence concluded that the reduction of tunneling losses by the higher temperature CdS process was crucial for the better performance of these devices. The model that was proposed to explain the beneficial effect of the increased CdS bath temperature was the following. It had been suggested, from earlier studies, that the Cu poor surface layer of high efficiency CuIn(Ga)Se 2 films was a result of Cu removal from the surface via the creation of Cu vacancies V Cu and the migration of Cu interstitials into the bulk of the absorber material [Klein, 1999]. This migration led to a high concentration of negatively charged V Cu within a defective surface layer [Niemegeers, 1998]. Such a high charge dens ity would enhance tunneling. However, if Cd ions could diffuse into the grains of the absorber, the formation of Cd Cu + could

PAGE 82

69 compensate a part of this charge, thereby reducing the tunneling. The observed decrease of the tunneling energy for the higher te mperature CdS deposition process was hence consistent with a stronger Cd diffusion into the absorber. The above model agrees well with another study, performed by Nakada, et al., for the case of high efficiency CuIn(Ga)Se 2 solar cells [Nakada, 1999]. Th ey investigated the diffusion behavior at the CuIn(Ga)Se 2 /CdS interface, using EDS. The analysis revealed that Cd was present in the CuIn(Ga)Se 2 layer approximately 100 A 0 from the interface boundary, thereby giving a direct evidence of Cd diffusion. Als o, Cu concentration was found to be decreased near the surface of the absorber film, suggesting substitution of Cd for Cu atoms. 3.4.2.5. Effect of Post Deposition Treatments Several groups have reported an enhancement of the performance of CuInSe 2 or CuIn(Ga)Se 2 devices after a post deposition air annealing treatment. It has been proposed that oxygen passivates surface dangling bonds related to Se deficiencies. This, in turn, reduces grain boundary recombination, enhances the net p type doping of the absorber, and facilitates inter grain transport [Cahen, 1991]. However, at times, contradictory results are reported for different fabrication processes used. Rau, et al. studied air annealing effects on CuIn(Ga)Se 2 films with the help of photoelectron spectroscopy and admittance spectroscopy [Rau, 1999]. UV photoelectron spectroscopy revealed type inversion at the surface of as made films, which disappeared after exposure of several minutes to air, due to the passivation of surface Se deficiencies. XP S revealed

PAGE 83

70 that air annealing at 200 0 C led to a decreased Cu concentration at the film surface. Admittance spectroscopy of completed CuIn(Ga)Se 2 /CdS/ZnO devices showed that the absorber surface type inversion was restored by the chemical bath used for CdS deposition. Air annealing of these finished devices at 200 0 C reduced the type inversion again, due to defect passivation. Moreover, the study also showed that oxygenation led to a charge redistribution and to a significant compensation of the effective acceptor density in the bulk of the absorber, suggesting the release of Cu from the surface and its redistribution in the bulk. For Ga rich CuGaSe 2 devices, Jasenek, et al. found improved performance after an air anneal. A study of the change in the diod e ideality factors and a numerical fit to the tunneling model indicated that the tunneling energy reduced significantly after the anneal. However, an Arrhenius plot of [A ln(J 0 )] Vs. 1/T showed that the activation energy remained nearly constant, indicati ng that recombination in the SCR was still the dominant recombination mechanism. 3.4.2.6. Other CuGaSe 2 Issues Kampschulte, et al. carried out measurements of some important parameters of the CuGaSe 2 material, such as mobility and resistivity [Kampschul te]. They studied CuGaSe 2 epitaxial layers that were grown on GaAs(001) substrates by low pressure metalorganic vapor phase epitaxy (MOVPE), exclusively with metalorganic precursors. XRD measurements revealed a predominantly c[001] oriented growth. All these CuGaSe 2 layers showed p type conductivity with net carrier concentrations of the order

PAGE 84

71 of 10 17 /cm 3 and Hall mobilities of approximately 30 cm 2 /V s. Also, the resistivity of 260 nm thick layers was found to be in the range of 0.5 1 W cm. MOVPE was ut ilized with the future purpose of using the same reactor sequentially to grow the heterojunction partner n ZnSe, so that the sample would not be exposed in between.

PAGE 85

72 CHAPTER 4. OUR SOLAR CELLS: FABRICATION AND CHARACT ERIZATION 4.1. The Device Structure The fabrication of the CuIn(Ga)Se 2 or CuGaSe 2 solar cells involves a sequential deposition of the various thin films that serve specific purposes, such as the absorber material, the buffer, and the front and the back c ontacts. The overall structure can be written as: Glass substrate/ Molybdenum (Back Contact)/ CuIn(Ga)Se 2 or CuGaSe 2 (Absorber material)/ CdS (n type Buffer)/ ZnO (Front contact), deposited in this sequence. The following figure depicts this structure. Figure 4.1. Structure of Our Typical Solar Cell Device Soda lime Glass Moly Back Contact ZnO Front Contact Absorber CdS External Contacts Light

PAGE 86

73 4.2. Fabrication of CuGaSe 2 Solar Cells 4.2.1. A Manufacturing Friendly Process Our laboratory has been developing a manufacturing friendly process for CuIn(Ga)Se 2 and Cu GaSe 2 solar cell preparation for the past few years. The manufacturing friendliness of the process has two aspects, as discussed below. Firstly, our fabrication involves a two step process for the preparation of the absorber layer. The first step is the deposition of the metal precursors (deposited sequentially from individual metal sources), and the second step is the high temperature selenization. As mentioned previously, such a process has distinct advantages compared with the more common co evaporat ion process used by some other laboratories. The co evaporation technique (obviously) requires a very high degree of control, because a number of elements are being evaporated from different sources simultaneously. This is especially important in the cas e of ternary or quaternary compounds such as those dealt with in this research. Therefore, the size of the substrate can be severely limited when co evaporation is being used, for larger substrate sizes usually mean more non uniformity of the film composi tion. The two step process, on the other hand, is relatively easily scaleable to commercial production, and hence can be termed as manufacturing friendly. The second aspect of manufacturing friendliness comes from a rather undesirable, however, unavoidab le, situation. Our laboratory, by the virtue of being a part of a University, faces several restrictions. First of all, theres always the financial funding problem. This often means that there are limitations as regards the quality of vacuum

PAGE 87

74 systems, t hat of the maintenance, and such. There are several graduate students walking around, sometimes leaving the doors to the laboratory wide open, and bringing the outside dust and other impurities in. Its not exactly a clean room situation. Theres limite d space to move around, and the samples have to be carried by hand from one room to another between depositions. A process that is developed in our laboratory, therefore, would automatically be a robust process, and any further improvements in terms of de position systems or the processing environment, such as those that would ensue in a highly maintained commercial method, would likely lead to an improvement of the device performance. The processing of the various thin films will now be described, along with the important characteristics of each of the material layers. (The absorber material, which is the most important material of interest, will be discussed last.) It should be kept in mind that one of the main aims of solar cell development is to arriv e at a process that offers a low cost electricity generation method. Hence, simple and cost effective deposition techniques, as well as inexpensive materials, have been preferred over their costly alternatives. 4.2.2. The Substrate The choice for the substrate material is Soda Lime Glass. It is an inexpensive substrate material, and offers good resistance to corrosion. It is also easily available at local hardware stores. There are other advantages associated with the use of glass as the substrate, such as, the substrate can be used as a packaging material. This becomes even

PAGE 88

75 more important when the solar cells are superstrate type, where the light is shone through the glass to reach the absorber. Another benefit of using soda lime glass is the diff usion of sodium (Na) from the glass to the deposited layers. In CuInSe 2 type cells, the V OC s have been shown to increase because of this Na reaching the absorber films. There are a few disadvantages that are associated with the use of soda lime glass. F irstly, the operating temperatures have to be limited to about 600 0 C, otherwise the glass is prone to warping, or even breakage, because of the stress. Secondly, the glass pieces purchased from local stores sometimes have scratches and/or spots on them, w hich can degrade the structure, and hence the performance, of the films deposited on them. It, therefore, becomes crucially important that the glasses are thoroughly cleaned before they are used. The cleaning procedure involves a soap/DI water soak s tep, followed by a soap scrub and DI water rinse. The glass pieces then go through an ultrasonic clean in a chemical (trichlorotrifuoroethane) that removes the organics from the glass. This is again followed by a DI water rinse. Lastly, the glass is blo w dried with high purity nitrogen. 4.2.3. The Back Contact Molybdenum (Mo) is a refractory metal that has been widely used as a contact material for CuInSe 2 type solar cells. It forms a good ohmic contact, and has a high resistance to selenium corrosio n. A 1 m m thick molybdenum layer is deposited using DC magnetron sputtering. Before this deposition, the glass is often heated in vacuum, to get rid of the moisture

PAGE 89

76 present. It has been shown that the sputtering pressure is crucially important to the qu ality of the films. If deposited at higher pressures, the films exhibit a rough surface morphology and poor resistivity, but they adhere well to the underlying glass substrate. On the other hand, films deposited at lower sputtering pressures have improve d resistivity and smoother surfaces, but they suffer from adhesion problems due to compressive stress. To circumvent this problem, Scofield, et al. used a bi layer of Mo, where the two layers were deposited at two different pressures (Scofield, 1995). Su ch a bi layer has been used for our molybdenum deposition. A thin first layer (of about a 1000 A 0 ), is deposited at a higher pressure of about 5 mTorr, to get the improved adhesion, followed by a low pressure (~1.5 mTorr) layer that gives excellent conduc tivity (typical numbers for the resistivity are in the low 10 5 W cm range). 4.2.4. The Heterojunction Partner/ Buffer Cadmium Sulfide (CdS) has been used extensively as the n type semiconductor material to form the p n junction with the p type CuGaSe 2 a bsorber material. CdS has a direct bandgap of 2.4 eV, and has an absorption edge at around 510 nm. This means that some of the light in the blue region of the visible solar spectrum (that below 510 nm) is absorbed in the CdS layer. These absorbed photon s can generate carriers, and such carriers can also contribute to the total photogenerated current. All the photons that have energy lower than 2.4 eV (i.e. wavelength higher than 510 nm) are transmitted through the CdS layer into the absorber layer.

PAGE 90

77 The CdS film is deposited by the Chemical Bath Deposition (CBD) technique. CBD is a non vacuum process, which turns out to be the best method for the CdS deposition so far. Other techniques, such as Close Space Sublimation and Sputtering, have been experimen ted with. However, these methods have produced CdS layers that severely limit the performance of the solar cells. This has been attributed to the possible cleaning and/or passivation of absorber surface by the chemicals involved in the CdS bath. The Cd S layer, evidently, plays an additional role. The CdS deposition comes after the absorber layer, and before the ZnO front contact. ZnO is deposited by RF sputtering, which, if done directly after the absorber, could severely damage the surface of the abs orber layer. The CdS essentially protects the absorber surface from this damage. Because of this, the CdS has often been referred to as the buffer layer in the literature. The chemistry of the deposition involves Cadmium Acetate (a source for the Cadmiu m), Thiourea (a source for Sulfur), and Ammonium Hydroxide (which acts as a complexing agent, controlling the rate of the reaction). 4.2.5. The Front Contact A good front contact needs to satisfy two requirements. It has to be highly conductive, so tha t the current generated by the photons can easily be conducted into the external circuit. The sheet resistance of this layer needs to be as low as possible, because often the external metal grid is a set of thin metal fingers, separated by a significant

PAGE 91

78 d istance between them. The current, hence, has to also flow in a direction perpendicular to the direction of incident light. Secondly, the front contact material has to be as transparent to the incident photons as possible, so that they go through the lay er unabsorbed, to reach the absorber layer. Zinc Oxide (ZnO) is used as the front contact for our solar cells. ZnO transmits about 90% of the incident light between 400 and 1000 nm. Transmission drops off at higher wavelengths, due to the free carrier a bsorption, which increases with increased doping. Therefore, a compromise has to be made in terms of achieving a low resistivity value and low free carrier absorption. Resistivity numbers of high 10 4 W cm are routinely achieved in our process. After th e CdS layer deposition, the sample is transferred to an RF sputtering system. An undoped ZnO layer of about 500 A 0 is first deposited. This is then followed by a thicker (about 4500 A 0 ) ZnO layer, which uses a ZnO target, along with several Aluminum piec es, to provide aluminum doping. During the ZnO deposition, a mask is used to divide the 2X 2 substrate into 25 individual circular device dots, each approximately of area 0.1 cm 2 This makes it easier to obtain good, working devices, without significan t shunts. It also allows the performance variation between the individual devices to be correlated to the locations of the dots with respect to the deposition sources (this will be more clearly seen in a figure in the next section).

PAGE 92

79 4.3. The Absorber The absorber deposition system contains a substrate holder, to hold the 2 X 2 substrate in place, and a heater system to heat the substrate. The evaporation sources for Cu, In, Ga and Se are located on four different sides, with respect to the substr ate. This is depicted in the following figure. Figure 4.2. Arrangement of the Substrate and the Sources The above figure also shows four of the 25 circular device dots on one substrate (brought about with the help of a mask during the ZnO deposition), along with the number system that is used to differentiate the individual devices. Ga In Cu Se 5 21 25 1 15 11 A single device

PAGE 93

80 As mentioned earlier, a sequential deposition process is more suitable than a co deposition process, from the point of view of large area comme rcial production of solar cell modules. Our process for the deposition of the absorber layer is a two stage sequential process, these stages being: (i) Precursor deposition: A low temperature sequential deposition of Cu and Ga metals, either with or without a Se flux. (Cu, In and Ga in the case of CuIn(Ga)Se 2 .) (ii) Selenization: A relatively high temperature step where the precursor film is annealed in a high flux of Se vapor. It should be noted that, although the major part of this research project dealt with the development of a process for CuGaSe 2 absorber deposition, it did start out with optimization of our (regular) CuIn(Ga)Se 2 absorber process. In the course of this research, various parameters for the absorber deposition were changed from time to time. Mos t of these films, however, could be divided into two major categories, depending on the sequence of the metal (Cu, Ga) depositions and the selenization temperature profiles. These two recipes are referred to as Type I and Type II CuGaSe 2 respectively, an d are described next. 4.3.1. Type I Versus Type II The first recipe used for CuGaSe 2 deposition, called Type I, was a natural extension of one of our CuIn(Ga)Se 2 deposition recipes, and involved the following sequence of depositions.

PAGE 94

81 Type I CuGaSe 2 : (i) Pr ecursor Deposition (275 0 C): a. Initial Ga, 100 A 0 b. Ga and Se co evaporation, c. Cu. (ii) Selenization (Se evaporation, a flux of ~ 28 A 0 /s about 28 minutes total): a. Ramp up from 275 to 450 0 C, b. 7 minutes at 450 0 C, c. Ramp up from 450 to 550 0 C, d. 7 minutes at 550 0 C (30 40 A 0 Top Cu optional), e. Ramp down from 550 to 425 0 C. f. Cool down to room temperature, in vacuum. This sequence is summarized in the following figure. Figure 4.3. Time Temperature Profile for Type I CuGaSe 2 550 450 275 0 4 11 15 22 28 Time (min) Temp ( 0 C) Se Ga/Ga Se/Cu

PAGE 95

82 The Type II CuGaSe 2 also started with the same precursor deposition sequence (Ga/Ga Se/Cu). However, about 4/5 th of the Ga needed for the absorber formation was deposited before the Cu. After the Cu deposition was over, the temperature was gradually increased al l the way to 550 0 C (eliminating the 450 0 C step), along with a low Se flux (of about 12 A 0 /s). The sample stayed at 550 0 C for about 10 minutes, after which the remainder (about 4/5 th of the total amount) of the Ga was deposited, along with Se. This was then followed by 28 minutes of selenization in a high Se flux. The entire sequence can be divided into three parts as follows: Type 2 CuGaSe 2 : (i) Precursor Deposition I (275 0 C): a. Initial Ga, 100 A 0 b. Approx. 4/5 th Ga and Se co evaporation, c. Cu. (ii) Precursor De position II a. Ramp up from 275 to 550 0 C (low Se flux), b. 10 minutes at 550 0 C (low Se flux), c. Remaining Ga (low Se flux). (iii) Selenization (Se evaporation, about 28 minutes total): a. 22 minutes at 550 0 C, b. Ramp down from 550 to 425 0 C. c. Cool down in vacuum. This seque nce of events is depicted in the following figure.

PAGE 96

83 Figure 4.4. Time Temperature Profile for Type II CuGaSe 2 4.4. Characterization of CuGaSe 2 Solar Cells We relied heavily on our routine characterization techniques such as Two Probe and Three Probe Current Voltage (I V) measurements, and Spectral Response. Variations were attempted in the fabrication procedure, and feedback was gained from the above techniques to correlate these variations to the performance of the devices. Wavelength Dependent I V, Capacitance Frequency (C F), Capacitance Voltage (C V), and I SC V OC measurements were done on selected samples from time to time, to gain knowledge about workings of specific regions of, or specific phenomena in, the sol ar cells. 550 275 Se (high) 28 min Time (min) Temp ( 0 C) Ga/Ga Se/Cu Ga Se (low)

PAGE 97

84 4.4.1. Current Voltage (I V) Measurements An HP 4145B Semiconductor Measurement Analyzer is utilized in either a 2 probe or a 3 probe configuration. The 2 probe I V measurement is the first measurement done on a completed solar cell devic e. The theory of I V has been covered in the Background chapter. To measure one device out of the 25 device dots (defined by the ZnO mask), one probe is placed on the dot being measured, and the other probe is placed on the Molybdenum exposed (by scrapin g off the top layers) between the dots. The dark and the light (1 Sun intensity using a Solar Simulator) I V curves for all 25 devices on a sample are obtained in this manner. Normally, this measurement provides the initial assessment of a sample, includ ing the effect of any composition gradient effects. If the V OC s, I SC s, and the curve shapes are within the acceptable range, the next step is the 3 probe I V measurement. In the 3 probe set up, two, instead of just one, probes are placed on the device dot (to touch the ZnO front contact layer). This is done so as to eliminate any contact (series) resistance effects, so that a more reliable measurement can be obtained. Because this measurement is cumbersome and time consuming, all 25 devices from a sa mple are rarely measured. More often than not, one row and one column, or two rows and two columns, are measured. The FFs can be calculated manually for each measured device by finding out the maximum power point. The V OC and the FF values from the 3 p robe data are used, in conjunction with the J SC values from the Spectral Response (Quantum Efficiency) measurement, to calculate the conversion efficiency values for the solar cells.

PAGE 98

85 Figure 4.5. A Representative 3 Probe I V Curve Plot for a USF CuGaSe 2 Cell 4.4.2. Spectral Response A Spex 4700 spectrometer is used to determine the external quantum efficiency (Q.E.) as a function of wavelength. The light source is calibrated using a Silicon Standard Cell Reference obtained from the National Renewable Energy Laboratory (NREL). The output, J SC of the cell at each wavelength is normalized against the Si reference to get the Q.E. versus wavelength curve. This curve is integrated against a reference AM 1.5 global spectrum to get the J SC of the device. T he next figure depicts a representative Spectral Response curve.

PAGE 99

86 Figure 4.6. A Representative Spectral Response Curve Plot for a USF CuGaSe 2 Cell 4.4.3. Capacitance Measurements Capacitance measurements can provide valuable information about the p n j unction, such as the depletion width, the doping densities and the doping profiles, and the built in voltage. The Capacitance Voltage (C V) measurement relies on the fact that the width of the depletion region varies with the applied voltage bias. The ju nction depletion region can be considered as a capacitor, with the capacitance per unit area given by W C e = (Eq. 4.1)

PAGE 100

87 Where e is the dielectric constant of the semiconductor, and W is the width of the depletion (space cha rge) region. If the doping densities of the p and the n side of the junction differ by more than two orders of magnitude, then the junction can be treated as a one sided abrupt junction, with its entire depletion region lying on the lower doping side. If this doping density is N A then the depletion width W can be given by ( ) 2 1 2 = A bi s qN V V W e (Eq. 4.2) Where q is the electronic charge, V is the applied voltage, and V bi is the built in junction voltage of the diode. The capacitance c an then be written as ( ) 2 1 2 = V V N q C bi A s e (Eq. 4.3) Ideally, a plot of 1/C 2 vs. V should be a straight line, with a slope of A 2 2 N A q 2 dV C 1 d e = (Eq. 4.4) The depletion width and the doping concentration can be ca lculated as follows. C W s e = (Eq. 4.5) ( ) slope q N s A e 2 = (Eq. 4.6) Moreover, the x axis intercept of this plot gives a value of the built in voltage V bi

PAGE 101

88 An HP 4194 Gain Phase Analyzer is used for our capacitance measurements. The I V measurement that is done previously is used as a screening measurement to eliminate any devices that show obvious shunting in the 3 probe I V plots. This is then followed by a Capacitance Frequency measurement. Here, t he frequency is varied from 100 Hz to 1 MHz. Typically, in the good devices, the capacitance signal is large at lower frequencies, but drops quickly to a lower value, and then saturates at this low value. Larger variations mean that the device is shunted and such devices are not used for the following C V measurement. The C V is carried out at a frequency of 500 Hz. This value is chosen because, at this frequency, the interface states or the stray capacitance from the leads and the set up do not contri bute to the measurement. The device is biased by applying a dc voltage V, which is varied from 3.0 to about 0.5 Volts. A sinusoidal ac voltage of a small amplitude (about 10 mV) is superimposed on the dc voltage. The plots for C vs. V and 1/C 2 vs. V ar e then obtained. An example of a such a 1/C 2 Vs. V plot, obtained for a CIS solar cell, is shown on the next page [Karthikeyan, 1997]. As can be seen, the variation 1/C 2 Vs. V for this particular device is highly linear.

PAGE 102

89 Figure 4.7. A 1/C 2 V Plot for a USF CuInGaSe 2 Cell [Karthikeyan, 1997] 4.4.4. I SC V OC Measurements When the mechanism of the junction transport differs from the ideal symmetrically doped p n homojunction, a diode (ideality) factor A is introduced in the basic equation for the juncti on current, which is reproduced below for convenience. ( ) sh s L s R IR V I AkT IR V q I I + = 1 exp 0 (Eq. 4.7) Classically, the value of A is between 1 and 2. However, an A value greater than 2 has been observed in several studies, and this has been attributed to an asymmetry in the doping arising because of non uniform spatial distribution of recombination centers at or near the junction interface [Bube, Photovoltaic Materials]. The diode factor, therefore,

PAGE 103

90 can provide important information about the junction me chanisms in a solar cell. The I SC V OC technique is one popular method to calculate this factor. Assuming that Rsh is large enough so as not to affect the characteristic, the V OC can be given as (setting I = 0) + = 0 0 ln I I I q AkT V L OC (Eq. 4.8) Recognizing that I L + I 0 ~ I L = I SC ( ) ( ) [ ] 0 ln ln I I q AkT V SC OC = (Eq. 4.9) And hence ( ) ( ) 0 ln ln I V AkT q I OC SC + = (Eq. 4.10) If a set of different I SC values and the corresponding V OC values can be obtained, they can be plotted as a straight line, and the value of A can be derived from the slope of this line. It is possible to generate the set of values using neutral density filters. The A values can then be correlated with the data from other measurement techniques, to gain valua ble insights into the junction mechanisms. An example of an I SC Vs. V OC plot, obtained for a CIS solar cell, is shown on the next page [Karthikeyan, 1997].

PAGE 104

91 Figure 4.8. An I SC V OC Plot for a USF CuInGaSe 2 Cell [Karthikeyan, 1997]

PAGE 105

92 CHAPTER 5. RESULTS AND DISCUSSION The results obtained for our CuIn(Ga)Se 2 and CuGaSe 2 solar cell devices will be presented and discussed in the following manner. Part I of this chapter will deal extensively with the processing results for CuIn(Ga)Se 2 and CuGaSe 2 In Part II, the modeling of our CuIn(Ga)Se 2 and CuGaSe 2 devices, carried out using two simulation techniques, will be discussed. 5.1. PART I: CuIn(Ga)Se 2 and CuGaSe 2 Processing Results 5.1.1. Relative Positions of the Substrate and the Sou rces Before beginning the presentation of the results, heres one important thing to remember: The discussion is, to a large extent, about the absorber deposition, and the effects of variations in the absorber recipe. Therefore, the relative positions o f the substrate and the various sources (Cu, In, Ga and Se for CuIn(Ga)Se 2 ; Cu, Ga and Se for CuGaSe 2 ) play a crucially important role, in terms of the gradients of these elements in the final absorber film. For the readers convenience, the figure that d epicts these relative positions is reproduced below.

PAGE 106

93 Figure 5.1. Arrangement of the Substrate and the Sources 5.1.2. Ga Evaporation and the Sample Numbering System During a single vacuum run, the four constituent element s, Cu, In (for CuIn(Ga)Se 2 ), Ga and Se, are deposited sequentially on a substrate, to accomplish the absorber layer deposition. Initially, a sputtering method was being used for the deposition of Ga, while the other constituent elements (Cu, In, Se) were deposited using an evaporation method. When this research project began, the above mentioned Ga sputtering was still in use. With this process, for CuIn(Ga)Se 2 open circuit voltages of the order of 425 450 mV had been achieved in our laboratory, with th e short circuit current densities exceeding 40 mA/cm 2 Later, however, it started becoming more and Ga In Cu Se 5 21 25 1 15 11 A single device

PAGE 107

94 difficult to control the sputtering and the evaporation systems in the same chamber, during a single run. The deposition uniformity on the substrate, as w ell as the run to run reproducibility, started getting affected. It was, therefore, decided that the sputtering gun for Ga be removed, and a Ga evaporation source be installed. For this, a few more changes had to be done to the internal geometry of the c hamber. Consequently, our base process control was lost. The next task, therefore, was to carry out the necessary calibrations and retrieve the base process. Around this time, with the new Ga evaporation, the sample # P001 was processed. All the sampl es that followed are numbered sequentially. The initial samples were CuIn(Ga)Se 2 devices. As a rule of thumb, unless otherwise specified, the samples until and including P033 were CuIn(Ga)Se 2 devices, whereas those beginning with P041 were CuGaSe 2 devi ces. (Samples P034 to P040 were an attempt to process CuGaSe 2 CuIn(Ga)Se 2 bi layer devices, and these wont be discussed here.) 5.1.3. CuIn(Ga)Se 2 Processing As mentioned previously, a number of conclusions can been drawn on the basis of the I V charac teristics of various samples. In our study, some correlations have been observed, between the I V parameters and the amounts/ratios of the elements deposited in the absorber layer. These will be discussed next, with the help of the results obtained for s pecific samples.

PAGE 108

95 5.1.3.1. CuIn(Ga)Se 2 Sample # P020 Sample # P020 was one of the first CuIn(Ga)Se 2 samples, with the new Ga evaporation system, to exhibit a decent I V performance. This sample had many devices that had open circuit voltages of 400 and above, with the highest one at 425 mV. The following figure contains the data for the V OC of the 25 devices on this sample. 425 415 385 345 325 395 415 415 425 425 395 415 425 415 415 385 405 405 395 405 395 385 395 405 405 Figure 5.2. Two Probe V OC Numbers for CuIn(Ga)Se 2 Sample # P020 Figure 5.3. Two Probe V OC Numbers for CuIn(Ga)Se 2 Sample # P020 The following observations can be made from the above figures. First, with the exception of the two devices at the top right corner (the Cu Se corner), the V OC s of all 0 100 200 300 400 500 P020 Cu In Se Ga Device #1 Device #25

PAGE 109

96 the other devices are in the range of 385 mV to 425 mV. Typically, such a variation of performance is seen from one side of the substrate to the other side, and originates from the slight, but definite, variation in the rat ios of the constituent elements. The two bad devices, with V OC s of 345 and 325, respectively, are the result of one or more of several possible effects, which are discussed next. First, the occurrence of the low V OC s may be an outcome of a high (grea ter than 1.0) metal ratio (i.e. the ratio of Group I/Group III, or, in this case Cu/(In+Ga)). This is because these devices are closer to the Cu source, while at the same time being farthest from Ga, as well as In. At the first glance, it might seem rathe r strange that there is such a sudden drop in the V OC s for the two devices, while the two neighboring devices in the next row are 425 mV each. However, such an effect has been observed in a few other samples, where the devices with a metal ratio very clo se to 1 were the best among the lot, and as soon as the ratio went above 1, the performance dropped rather abruptly. This phenomenon, of Cu rich devices being poorer in performance than their Ga or In rich counterparts, has often been discussed in the li terature. The second possible cause for the above mentioned low V OC s may be edge effects. It is possible that this corner of the sample was damaged, either during the handling or because of unwanted deposition effects such as masking because of the samp le holder. The third reason may be a poor quality glass substrate piece, or even a non uniformity produced during the Molybdenum deposition. However, these reasons seem less likely than the first one, primarily because the rest of the sample shows very c onsistent gradients, and even the lesser performing devices show their own internal gradient -the 325 mV device does, in fact, have a

PAGE 110

97 slightly higher I/III ratio than the 345 mV device, making the I/III ratio the most likely cause of this behavior. The f igure on the next page shows the variation of the fill factors (calculated from 3 Probe I V curves) with the change in position, for two rows of devices on this sample [Shankaradas, Thesis]. The four backside columns in this figure represent four devices from the third row: # 3, 8, 13 and 18, whereas the four front columns represent four devices from the second row: # 7, 12, 17, 22. The following observations can be made from this figure. First, the FF numbers range from about 58% to about 50%, which is q uite typical of our devices. Second, the FF performance is better towards the Se side of the sample, compared to the Ga side. A more or less similar gradient can be observed, for the two rows of devices, in the V OC numbers. This is not surprising, as th e V OC s and FFs are often found to go hand in hand. 1 2 3 4 44 46 48 50 52 54 56 58 60 Fill Factor (%) Second row Third row In Se Cu Ga Figure 5.4. Fill Factor Vs. Position of Device, for 8 Devices on Sample # P020

PAGE 111

98 A similar performance variation has been observed in other CuIn(Ga)Se 2 samples processed in t his research. Some of these results will be presented later. It is reasonable to say that this may be the result of one of the following two reasons. First, it is possible that the amount of Se deposited was insufficient to the point that the Ga side of the substrate (the side opposite to Se) did not receive the amount of Se that it needed for the formation of the proper absorber phase. This can readily explain the above gradient in performance. However, this does not seem likely, as great care is usua lly taken to provide more than enough Se during the deposition process. The second, more likely reason could be that the Ga was responsible for the degradation of the devices. This could happen, for example, if the Ga deposited during the precursor layer deposition did not bond very well with the other elements. Such un bonded Ga could produce defects in the material, thereby leading to the deterioration of the devices. The above sample (# P020) had the following amounts of constituent elements deposit ed during the absorber precursor layer formation (as measured by the thickness monitors ): 1360 A 0 Cu (this is called the Bulk Cu), 2900 A 0 In, 875 A 0 Ga, and about 45000 A 0 Se (this is the precursor Se, as against that deposited during the final selenizati on step). Also, towards the end of the selenization step, 30 A 0 of additional Cu (hereafter referred to as Top Cu) was deposited. At this point, it should be noted that the above thickness numbers are not the actual numbers that get deposited on the subst rate, and that there is a correction factor associated with it, depending on the distance of the source from the substrate. However, because these distances are constant from run to run, the correction factors are constant, too. Therefore, comparing the thickness numbers from the thickness monitors is quite

PAGE 112

99 reasonable, and provides an easy way to track the run to run variations in the amounts/ratios of elements. (Typically, the variation in the thickness of a particular element across the sample was abou t 7 10 %.) In an attempt to further understand the performance of our samples, the following experiments were undertaken, where the amounts and the ratios of different elements were systematically changed. 5.1.3.2. CuIn(Ga)Se 2 Samples # P030 and # P031 A s mentioned previously, for our CuIn(Ga)Se 2 samples, Cu was deposited in two stages. Most of it was deposited early in the deposition sequence, in the precursor layer. This was called bulk Cu as it was thought to participate in the formation of the bulk of the absorber layer. A very small amount, typically, about 30 to 50 A 0 was deposited towards the end of the selenization stage. This top Cu was thought to play a key role in the formation of the surface layer of the absorber [Zafar, Dissertation]. Sa mple # P030 had 1350 A 0 of bulk Cu (as opposed to 1360 A 0 in # P020), whereas the top Cu was increased to 40 A 0 (from 30 A 0 of # P020). The total Cu amount essentially remained the same. The amounts of the other constituent elements were held constant. (The total Se amount during the precursor formation varied by about 10 %. However, as mentioned before, ample Se was usually available during the process. Moreover, much more extra Se was typically available during the selenization stage, to more than co mpensate for any possible deficiencies.) The V OC numbers for this sample are shown below.

PAGE 113

100 395 415 425 415 385 385 405 415 405 295 285 385 375 355 365 335 335 365 355 365 275 305 365 345 365 Figure 5.5. Two Probe V OC Numbers for CuIn(Ga)Se 2 Sample # P030 Once again, a trend, of increasing V OC numbers towards the Cu side, can be noticed. This effect is even more pronounced than that in sample # P020. The higher V OC s are once again in the range of 415 425 mV, and are present in the first two rows, w hich are the rows closest to the Cu source. This indicates that a small change in the amounts of the top and the bulk Cu, while maintaining the same total amount, did not have much effect on the performance of the devices. For the next sample, # P031, th e bulk Cu was kept constant, while the top Cu was increased from 40 A 0 to about 55 A 0 The following V OC numbers were obtained. 235 365 205 365 365 365 375 375 345 365 355 335 295 ----395 375 335 305 335 375 375 355 355 355 Figure 5.6. Two Prob e V OC Numbers for CuIn(Ga)Se 2 Sample # P031 It can easily be seen that the increase in the top Cu has had an overall detrimental effect on the performance of this sample. Although it looks like there might have been some other problems, as witnessed by t he two dead devices depicted by -- and the poor performance in the area surrounding these bad devices, the highest voltage numbers have definitely shifted away from the Cu source (shifted down in the figure). This meant

PAGE 114

101 that 55 A 0 of Cu was too much, a nd, therefore, it would have to be cut back down. This conclusion has been corroborated by several other experiments with CuIn(Ga)Se 2 samples. Keeping the above observations and conclusions in mind, we now turn our attention to CuGaSe 2 processing results. An important thing to be noted is that CuGaSe 2 cells had been processed previously in our laboratory [DAmico, Thesis]. The highest V OC and J SC values obtained in this research (on separate devices) were 675 mV and 14.3 mA/cm 2 respectively. The results from the above CuGaSe 2 experience helped establish the starting point for the CuGaSe 2 project discussed in this document. However, as discussed before, because of an important modification to the existing deposition system (in the Ga deposition method), t his project virtually began from scratch, by re establishing the CuIn(Ga)Se 2 and CuGaSe 2 base processes. Therefore, the benefits of previous CuGaSe 2 experience were rather limited. 5.1.4. Type I CuGaSe 2 Calibration experiments were carried out to find out the equivalent amount of Ga to replace the In, for the formation of CuGaSe 2 absorber layers. Initially, the overall amounts of different elements were somewhat lower than those for CuIn(Ga)Se 2 Hence, the overall final thickness of the CuGaSe 2 absorb er layer was less than that of a typical CuIn(Ga)Se 2 absorber (approximately 1.5 m m, instead of ~ 2 m m). The following description relates to the CuGaSe 2 processing experiments.

PAGE 115

102 The first few results that are presented below relate to the Type 1 CuGaSe 2 process, which has been described in detail elsewhere. As a reminder, the recipe is reproduced below. (i) Precursor Deposition (275 0 C): a. Initial Ga evaporation, 100 A 0 (without any Se), b. Ga and Se co evaporation, c. Cu evaporation. (ii) Selenization (Se evaporatio n, a flux of ~ 28 A 0 /s for about 28 minutes total): a. Ramp up from 275 to 450 0 C, b. 7 minutes at 450 0 C, c. Ramp up from 450 to 550 0 C, d. 7 minutes at 550 0 C (30 40 A 0 Top Cu optional), e. Ramp down from 550 to 425 0 C, f. Cool down to room temperature, in vacuum (no Se during this phase). In the following sections, we present the results for Type I CuGaSe 2 samples. 5.1.4.1. CuGaSe 2 Sample # P041 (Type I) Sample # P041 was the first CuGaSe 2 sample to exhibit a V OC number higher than that obtained with CuIn(Ga)Se 2 O nly one device, # 15, on this sample showed 505 mV, while all other V OC numbers were much lower. The precursors for this sample contained the following amounts of constituent elements: (i) Cu: 1200 A 0 Bulk, 25 A 0 Top.

PAGE 116

103 (ii) Ga: 100 A 0 Initial, 4000 A 0 Bulk. (iii) Se ( deposited with Ga precursor): 53 kA 0 (iv) Se (in the Selenization part): ~ 45 kA 0 The following table shows the V OC numbers for sample # P041. 205 255 165 275 --265 235 295 185 215 255 235 265 225 305 125 265 305 165 205 185 215 505 155 185 Figure 5.7 Two Probe V OC Numbers for CuGaSe 2 Sample # P041 The appearance of a 500 mV device meant that the metal ratio (Cu to Ga ratio, in the case of CuGaSe 2 ) was somewhere in the ballpark of what was needed. This ratio would now have to be adjusted so as to ob tain more devices with better V OC numbers. 5.1.4.2. CuGaSe 2 Sample # P042: Reduction in Cu In sample # P041, shown above, the only device that had 505 mV was located on the side opposite to that of the Cu source. This, then, indicated that the sample n eeded less Cu to be able to perform better. For the next sample, # P042, the amount of the bulk Cu was reduced from 1200 A 0 to 1075 A 0 while the top Cu, deposited during the selenization stage, was maintained at 25 A 0

PAGE 117

104 445 345 345 ----605 545 515 345 --545 555 535 425 195 525 545 525 365 445 305 245 325 455 335 Figure 5.8. Two Probe V OC Numbers for CuGaSe 2 Sample # P042 As can be easily seen from the following table, many more devices showed V OC s in the range of 500 to 600 mV, with th e highest one at 605 mV. It is, therefore, easily observed that the decrease in the amount of bulk Cu had helped the process. The higher V OC numbers were now centered on the middle of the sample, as regards the North South direction (North is the Cu sour ce side. Note that there is no In at the South, as we are dealing with CuGaSe 2 here, and not Cu In (Ga)Se 2 .) However, in the side to side (West East), i.e., the Ga Se source direction, the improved devices seemed to be located towards the left side, which was nearer to the Ga source, and away from Se. A similar trend was observed for sample # P043, which is presented later. Lets now consider the variation in the I SC (current) numbers, as a function of the position of the individual device on the substrate for sample # P042. .759 .631 .727 ----.868 1.036 .847 .899 --1.119 1.132 .869 .570 .534 1.048 1.123 .990 .884 .716 .958 1.015 1.006 1.009 .684 Figure 5.9. Short Circuit Current (I SC ) Vs. Position of Device, for Sample # P042

PAGE 118

105 From the I SC Vs position plot, it can be seen that the devices closer to the Se Cu corner have poorer performance compared to those near the Ga side. An important thing to be noted here is that these I SC numbers depend on the areas of the individual devices. Because o f the variations in the device areas (originating from the non uniform ZnO contact deposition through the mask, to be described later), only limited conclusions can be drawn from this comparison. Nevertheless, this data can be taken as a general guideline The main objective, throughout this project, had been to gain improvements in the V OC s. Therefore, much more emphasis was given to the V OC performance, than that of the I SC s. However, whenever warranted, the current density (J SC ) numbers, which esse ntially eliminated the area dependence, were used to compare the current performance of the devices. To be able to compare the trends in the I SC numbers, and eventually relate these trends to the absorber deposition parameters, another thing should be ke pt in mind. The current through the entire device structure also depends upon the quality of the top contact, which, in the case of our solar cells, was ZnO. As mentioned above, the active area of the device was defined by a mask during the ZnO depositio n process. In this process, the ZnO sputtering target was located off axis, towards one side of the substrate. Also, because of the substrate holder arrangement during the CdS chemical bath deposition process (which precedes the ZnO deposition), the sub strate had to be cut into two separate pieces. The geometry of the ZnO process could affect the devices in various ways. First, because of the angle involved in the masked ZnO deposition, shadowing effects originated. The areas (and, to a much lesser ex tent, ZnO thickness) of different devices were, therefore, different. Second, the two pieces (of the same initial substrate)

PAGE 119

106 could receive slightly different amounts of ZnO, because of the fact that they were placed at two different corners of the ZnO sub strate holder. The former effect was usually quite noticeable, while the latter had been proved to be negligible. A third effect, that was a consequence of the off set target and the standard orientation used, was observed during the early stages of CuGa Se 2 development. Because there could be significant atomic/molecular bombardment during the ZnO sputtering process, the devices closer to the target may have faced significantly more damage than those farther away. This may have led to a gradient in perf ormance. Such a gradient, arising because of ZnO variations, was in the East West direction, the same direction for the gradient arising because of Ga Se variations. This often complicated the analysis of the trends originating from the variation in the metal ratios. As a solution to this problem, a small change was introduced in the procedure, for all of the samples that were processed this point forward. The two pieces of a single sample (cut prior to CdS, and hence ZnO) were now arranged for the ZnO deposition in a way such that the bottom half was inverted with respect to the top half of the sample. This way, if there was a certain East West gradient evident in the performance of the finished sample, the Ga Se effect could easily be distinguished fr om the ZnO effect. 5.1.4.3. CuGaSe 2 Sample # P043: Continued Reduction in Cu Because of the performance improvement that resulted from the reduction of the amount of Cu, it was decided to further reduce the bulk Cu, for the next sample (# P043). This s ample had 1025 A 0 of bulk Cu (down from 1075 A 0 ), while the top Cu amount was

PAGE 120

107 kept approximately the same (30 A 0 while the previous sample had 25 A 0 with an estimated inaccuracy of +/ 2 A 0 ). The following figure shows the V OC distribution for this samp le. 525 555 345 255 205 505 475 375 325 375 475 575 355 365 385 365 585 405 365 275 275 375 285 305 305 Figure 5.10. Two Probe V OC Numbers for CuGaSe 2 Sample # P043 Although the number of devices with V OC s above 500 mV was smaller than that in th e previous sample, most of these higher voltage devices were still located in a somewhat similar region of the substrate. The difference was that now there were two good devices (>500 mV) in the first row, which was closest to the Cu source. For the prev ious sample, the better devices were present in rows 2, 3, and 4, and there were none in the first row. This was to be expected, because the amount of Cu was lowered here, whereas there was no change in the amount of Ga or Se (East West direction), compar ed to the previous sample. Next, we present the data for the current (I SC ) values for sample # P043. .883 1.212 1.092 1.111 .961 1.188 1.224 1.206 1.244 .915 1.240 1.527 1.575 1.480 1.166 1.064 1.293 1.446 1.415 1.045 .858 .987 1.022 .960 .842 Figu re 5.11. Two Probe I SC Numbers for CuGaSe 2 Sample # P043

PAGE 121

108 It can be seen that, except for a few corner devices, the I SC numbers for most devices turned out to be above 1 mA. Although there was some variation in these numbers, they were much more uniform than the past samples. A few devices from near the Ga Cu corner were selected for the J SC measurement, to be carried out using the Spectral Response technique. The following two figures depict the spectral response curves for devices # 7 and 8 from sampl e P043 (these are the two devices with bold faced current values above). Figure 5.12. Spectral Response Curve for CuGaSe 2 Sample # P043, Device # 7

PAGE 122

109 Figure 5.13. Spectral Response Curve for CuGaSe 2 Sample #P043, Device # 8 The J SC numbers, calculated from the above spectral response curves, were 14.8 mA/cm 2 and 15.2 mA/cm 2 for devices # 7 and 8, respectively. This was very encouraging, because these current density values were close to the highest that could be expected of the CuGaSe 2 devices, with the theoretical maximum J SC predicted to be near 20 mA/cm 2 (However, as will be seen in the remainder of this report, it turned out to be impossible to further improve these J SC values.) A couple of other important observations can be made from the above spectral response curves. First, the extrapolation of the drop in the curves near long wavelengths shows that the bandgap of this CuGaSe 2 material was around 1.63 to 1.64 eV. This was close to, but slightly less than, the theoretical bandgap value of 1. 68 eV. This indicated

PAGE 123

110 towards the formation of a material that was very close to the ideal CuGaSe 2 absorber material. On the short wavelength side, the response starts to drop around 420 to 430 nm. This is rather interesting, because this drop, which wa s present because of the absorption in the CdS window/buffer layer on the top of the CuGaSe 2 was expected to be around 500 nm. This indicated towards the presence of a very thin CdS layer, perhaps less than a 100 A 0 It is possible that this layer was n ot a true CdS layer, but was rather present as an intermediate phase, or mixture, between CdS and the adjoining ZnO top contact layers. Such a material, with a bandgap value between the bandgaps of CdS and ZnO (2.4 eV and 3.2 eV, respectively), would mani fest itself as a shift of the start of the drop to shorter wavelengths. The presence of such a layer could not be proved in this work, because of the lack of availability of certain advanced characterization techniques. (For example, such a layer could p erhaps be identified with the help of Transmission Electron Microscopy.) However, simulation studies of our CuIn(Ga)Se 2 solar cells have indicated towards the possibility of such a CdS ZnO intermixing [Shankaradas, Thesis]. Alternatively, the CdS layer c ould be a part of the space charge (depletion) layer for the solar cell. Another salient feature, which typically exhibited itself in all CuGaSe 2 spectral response curves, was the gradual decrease of the response in the long wavelengths from about 600 nm, up to the bandgap edge of 750 nm. This region was expected to represent the bulk of the absorber (as against the absorber surface). Therefore, such a drop may have been indicative of poor quality absorber material in the bulk. One possibility was th at a slightly different phase of the CuGaSe 2 material was present in the bulk, which decreased the absorption of the incident light. A second possibility was the presence of

PAGE 124

111 structural defects in the bulk, either originating in the absorber itself, or pro pagating from the underlying Mo back contact layer, or even from the glass substrate itself. 5.1.4.4. Samples # P060, P061: Continued Reduction in Cu We now turned our attention back to improving the V OC s of our CuGaSe 2 cells. The above samples had in dicated that a reduction in Cu, and hence, a reduction in the Cu/Ga metal ratio, had helped improve the performance. It was, then, decided to continue to reduce Cu, until we saw a drastic reduction in the performance. This point, then, would define one e xtreme of the metal ratio. This happened at around 900 A 0 Cu, while the amounts of the other elements remained constant, with the metal ratio of about 0.85. It was found, along the way, that a Cu amount of about 950 A 0 seemed optimal, for the total thick ness that was being used. The following table shows the V OC performance for sample # P060, which had 950 A 0 of bulk Cu, and 20 A 0 of top Cu. 655 555 715 695 675 575 625 695 555 665 575 575 605 515 385 605 665 635 635 565 295 475 445 375 235 Figure 5.14. Two Probe V OC Numbers for CuGaSe 2 Sample # P060 It is clearly evident that the metal ratio adjustment helped improve the performance significantly. Several devices had V OC s over 600 mV, with one above 700 mV. The last row of devices, however, had suffered, either from the edge effects, or from a bad region of the starting substrate.

PAGE 125

112 However, there are other places on this sample where the V OC was low, even if all the surrounding devices had higher numbers. Such non uniformity could again be a r esult of low quality substrate or a poor uniformity of the Mo back contact deposition. These non uniformities proved to be very difficult, even impossible at times, to track. Therefore, a decision was made to focus on increasing the highest V OC numbers, rather than worrying too much about such local fluctuations. Henceforth, we decided to track only those devices that had V OC s higher than 600 mV, as a measure of the V OC performance of the sample. Sample # P061 had amounts of elements similar to w hat # P060 had. The following figure shows the V OC performance. Although there were (slightly) less number of devices with V OC s > 600 mV, the high V OC devices were located in the same region of the substrate. These numbers were still in the top row, wi th the highest one, once again, approaching 700 mV. 620 695 655 675 665 535 535 595 225 245 545 675 645 635 545 395 555 605 515 575 345 485 545 545 585 Figure 5.15. Two Probe V OC Numbers for CuGaSe 2 Sample # P061 5.1.4.5. Sample # P063: Variation i n the Initial Ga All the above samples used an initial Ga layer of about 100 A 0 As a reminder, this initial Ga was the very first precursor layer deposited, before the Ga+Se co evaporation. We decided to investigate the effect that this layer had on the performance.

PAGE 126

113 The initial Ga thickness was reduced to 50 A 0 for one sample, reduced to zero for another, and was increased to 150 A 0 for yet another sample. The 150 A 0 sample, as well as the 0 A 0 sample exhibited a diminished performance. The sample with 50 A 0 initial Ga, sample # P063, had the following V OC s. 685 665 645 645 645 655 665 655 675 665 635 675 665 685 665 645 675 675 685 595 605 645 655 605 665 Figure 5.16. Two Probe V OC Numbers for CuGaSe 2 Sample # P063 Two things can easily b e noticed. The performance here was much more uniform, and almost all the V OC s were fairly high, with most of them above 650 mV. However, although more number of devices had voltages of > 600 mV, the highest number (of 685 mV) was, in fact, lower than t hat from the previous two samples, # P060 and P061 (705 mV and 695 mV, respectively). It was quite possible that, for this particular absorber thickness, a 50 A 0 initial Ga was more suitable. However, in the subsequent experiments, the total thickness of the absorber layer was increased, in an attempt to get away from the possible shunting effects in the absorber film, as well as, to reduce the occurrence of peeling of the absorber film. (Such a peeling was seen earlier to be resulting from deposition of Se, without a buffer layer such as the initial Ga layer.) For this increased thickness, a 100 A 0 initial Ga did, indeed, produce uniformity similar to what was seen above, along with improved V OC numbers. The 100 A 0 initial Ga layer was, therefore, once again established as the first step in the absorber deposition process,

PAGE 127

114 for our Type I samples. (Changing the thickness of the initial Ga layer did not significantly affect the Type II samples.) 5.1.4.6. Samples # P062 and P082: I V Curve shapes and Ab sorber Thickness All CuGaSe 2 samples processed until this time had fill factors in the range of about 40% 50%. The following figure shows an example of a rather poor I V curve, drawn for device # 15 of sample # P062. The following observations can be made from the figure. First of all, the curve in the third quadrant, after the turn on of the device, is far from being vertical. This usually means that there is an unwanted series resistance in this sample, which bends the curve away from the vertical. Although all practical devices will always have some finite series resistance, the effect in this particular device is rather large. The series resistance in a solar cell device generally comprises of the bulk resistance of the absorber material, and an y contact resistances that may be present. As the contacts used in this particular structure are highly conductive, the resistance, more likely, is coming from the bulk of the absorber material itself (although some contribution from the top contact ZnO i s a possibility). (The external measurement contact resistance is eliminated by using a third probe on the top contact, during measurement.)

PAGE 128

115 Figure 5.17. Three probe I V Curve for Device # 15 from CuGaSe 2 Sample # P062 In the first quadrant (low reverse bias), an ideal I V curve should be a horizontal line. The I V curve presented above has a slight slope in this region. This may be the result of one or more of the following two reasons. Firstly, if there is shunting in the device, perhaps beca use of defects introduced during the deposition, this will bend the curve away from the horizontal. Alternatively, such a bend could occur because of poor collection of the photo generated carriers. When the reverse bias increases, the collection improve s, and hence the photocurrent slowly increases accordingly. This may happen because, with an increased reverse bias, the depletion region extends more into the bulk of the device, thereby increasing the number of carriers that can reach the depletion regi on and be collected on the other side. To investigate the effect of shunting, and possibly reduce any shunting through the absorber layer, it was decided that the overall thickness of the absorber be increased.

PAGE 129

116 Sample # P082 was processed in such a way t hat the absorber film was about 20% thicker than that in the past, while maintaining the same metal (Cu/Ga) ratio as before. The following table shows the V OC performance of this sample. (Unlike the previous V OC numbers, which were 2 probe, these numbers are 3 Probe numbers.) 659 653 693 641 693 681 635 699 640 685 550 599 468 643 653 638 633 645 638 629 457 --632 670 647 Figure 5.18. Three Probe V OC Numbers for CuGaSe 2 Sample #P082 As can be easily noticed, most of the V OC s are above 600 mV, with the highest one, Device #12, at 699 mV. The following figure, presented on the next page, depicts the 3 probe I V curve plots for devices #11 and #12, with V OC s of 693 and 699, respectively. A comparison of these next I V curves to the one present ed before (from # P062) indicates that, for this sample (# P082), the slope of the curve in the 1 st quadrant is closer to the horizontal. It is reasonable to say that this is because of reduced shunting through the absorber layer, as only the thickness of the absorber was changed (increased) for this sample. Any change in the shunting or series resistance behavior of a device should show up in the squared ness of the I V curve in the 2 nd quadrant of the 3 probe curve. Therefore, this reduction in shuntin g can be quantitatively measured in terms of the fill factor. Indeed, the two devices shown, from sample #P082, had substantially increased fill factors 56% and 55%, for devices #11 and #12, respectively. The highest fill factor, for device #21, was 58 %, which resulted in a conversion efficiency of about 4.8%.

PAGE 130

117 Figure 5.19. Three Probe I V Curve for Devices # 11, 12 for Sample # P082

PAGE 131

118 5.1.5. Type II CuGaSe 2 As described in the literature survey, NREL researchers have used a three stage recipe for t he CuIn(Ga)Se 2 absorber preparation. In this recipe, the film starts out being Cu poor, then goes through a Cu rich phase (at a high temperature), and then goes back to being Cu poor towards the end of the deposition. According to the literature, the int ermediate Cu rich phase helps form larger grains in the film, thereby improving the performance of the absorber. We decided to explore such a Cu rich to Cu poor conversion for our CuGaSe 2 absorbers. The process that resulted was called the Type II recipe and it went through the following deposition sequence. (i) Precursor Deposition I (275 0 C): a. Initial Ga, 100 A 0 b. Approx. 4/5 th part of (Ga and Se) co evaporation, c. Cu. (ii) Precursor Deposition II: a. Ramp up from 275 to 550 0 C (low Se flux), b. 10 minutes at 550 0 C (l ow Se flux), c. Remaining 1/5 th part (Ga and Se) co evaporation. (iii) Selenization (Se evaporation, about 28 minutes total): a. 22 minutes at 550 0 C, b. Ramp down from 550 to 425 0 C, c. Cool down to room temperature (no Se during this phase).

PAGE 132

119 In short, the (Ga + Se) layer, which was deposited all at once in the case of Type I, was now split into two layers. One of these (a major portion) was deposited before Cu, while the other after Cu, at high (550 0 C) temperature. One thing to be remembered about these Type II devices is that the Se amount used (i.e., evaporated) during the deposition was much higher than that normally used during a Type I process. This is because, for Type II, the substrate sat at a higher temperature for a much longer time. During this interval, if a constant low Se flux was not on, the Ga already present in the sample (from previous steps) might have left the sample, in the form of volatile Ga Se species (this phenomenon is described before). Of course, it was also assumed that, although more Se wa s used, only that amount, which could combine with either Ga or Cu, or both, would be incorporated in the absorber film. (Such an assumption would not hold if the Se amount were excessively high.) Lets now look at some of the samples processed with this recipe. Initially, to be able to see whether the composition and structure of the Type II absorber is close to what we needed it to be, EDS and XRD characterizations were carried out on a sample. This characterization is discussed next. 5.1.5.1. EDS and XRD Characterization Results for Type II CuGaSe 2 Energy Dispersive Spectroscopy and X Ray Diffraction were carried out on a CuGaSe 2 absorber film deposited on a glass substrate that was coated with Mo (Sample # P131, Type II CuGaSe 2 ). The EDS and XRD pl ots are included in Appendix 1 of this report.

PAGE 133

120 EDS results showed that the ratio of the three (Cu, Ga, Se) elements present in the absorber layer was close to 1:1:2, which meant that the material was close to being CuGaSe 2 Because this was a standard les s EDS, another EDS scan was done on a CuGaSe 2 standard provided by NREL. The comparison between these two was used to derive the above conclusion. It should be remembered that EDS probes the top few thousand Angstroms of the material in question, and hen ce, the results are representative of the top portion of the film only. The deeper bulk material may have had a different ratio of elements, and hence a different phase, which would possibly give rise to a drop in the spectral response for longer waveleng ths, as was discussed in the previous section. XRD analysis showed that the structure of the analyzed CuGaSe 2 film was polycrystalline, with a preferred orientation along the [112] direction. This data is consistent with the recent data in the literature. 5.1.5.2. Type II Samples # P111, # P115: Effect of Se For sample # P111, which was a Type II sample, the amounts of the individual elements were: 1080 A 0 Cu, 4400 A 0 Ga (4100 before Cu, and the remaining 300 after Cu). The following was the V OC perfor mance. 585 615 635 525 715 315 675 685 705 675 655 695 645 665 625 645 465 695 645 555 605 365 695 685 665 Figure 5.20. Three Probe V OC Numbers for CuGaSe 2 Sample # P111

PAGE 134

121 Most of the V OC s are above 600 mV, with the highest one at 715 mV. One thing to be noted here is that the total Se amount could have been excessively large, when compared to the next sample that is presented below. Sample # P115, which had similar numbers for Cu and Ga, had less Se (a deposition rate of 13 A 0 /s, as opposed to 20 A 0 /s, for the same period of time), during the first Ga Se deposition step. Another change introduced was that the bottom piece of this sample (the bottom three rows) was annealed in air for 10 minutes, at 200 0 C, immediately after the CdS buffer was depo sited. The top piece (top two rows) was the control piece, meaning that it was processed with a regular recipe, without any annealing. This was an attempt to see if an intermediate annealing, which had been claimed in some research papers to be helpful, would improve our devices as well. The following is the V OC performance of this sample. 705 685 655 685 665 665 735 725 635 665 735 775 775 775 725 535 695 755 765 735 495 715 595 505 555 Figure 5.21. Three Probe V OC Numbers for CuGaSe 2 Sample # P115 It can be seen, from the above figure, that the V OC performance of this sample is, in fact, better than that of the previous one. Two conclusions can be drawn from this result. Firstly, the reduction in the Se flux has helped the process (the top t wo rows). This has produced the highest V OC (non anneal) seen so far in this research: 735 mV.

PAGE 135

122 Secondly, in addition to the decreased Se, an air anneal after CdS also has helped improve the V OC s. (More about annealing in a later section.) 5.1.5.3. Typ e II Sample # P119: Initial Ga For sample # P119, no initial Ga was deposited. The 100 A 0 reduction in the amount of Ga was compensated by increasing the Ga amount in the Ga+Se layer. The V OC performance was as follows. 425 705 665 685 685 675 695 665 305 665 675 695 635 645 685 655 705 695 695 675 625 605 725 705 505 Figure 5.22. Three Probe V OC Numbers for CuGaSe 2 Sample # P119 The above table clearly shows that the elimination of the initial Ga layer did not change the V OC performance signifi cantly. This, in fact, points towards the possibility that the structure of a Type II absorber is quite different from a Type I absorber This issue will be discussed in more detail in a later section. 5.1.5.4. Current Density Performance of Type II A lthough improved V OC s, in general, were obtained with the Type II recipe, the J SC (current density) performance of these Type II samples was diminished. The next figure depicts the spectral response curve for device # 12 from sample # P115.

PAGE 136

123 Figure 5.23 Spectral Response for CuGaSe 2 Type II Sample # P115, Device # 12

PAGE 137

124 The J SC value calculated from this curve was 10.2 mA/cm 2 (As a reminder, the above device had a V OC of 725 mV.) Other Type II samples have consistently shown current density values in t he range 10 11 mA/cm 2 Type I J SC values were generally higher by 2 3 mA/cm 2 ranging from about 12.5 mA/cm 2 to more than 15 mA/cm 2 A number of experiments were carried out to improve the currents for Type II, but little success was achieved. This rea lly points towards the basic structural difference between the two types of recipes The very mechanism that lead to the improved V OC s seems to be the reason why the J SC s are lower in Type II devices. 5.1.5.5. Type II Sample # P132: Cu rich Cu poor T ransition The main purpose behind designing the Type II recipe was to explore the possibility of a Cu rich to Cu poor transition in the absorber film. Although careful calibrations showed that the amounts of Ga deposited in the first and the last step (i.e., before and after Cu) would have carried the absorber film through this transition, it was deemed necessary to confirm this with some more experimentation. For sample # P132, the thickness of the absorber layer was increased by about 7.5 %. However, to make absolutely sure that the sample goes through a Cu rich to Cu poor transition, the increase in the Ga came only in the second Ga deposition (i.e., in the second Ga+Se layer, which comes after the Cu, at high temperature). This would, then, carry t he absorber film through a transition from a metal ratio of about 1.05 to that of about 0.9. Following is the V OC performance of this sample.

PAGE 138

125 675 675 645 685 705 615 635 695 695 705 595 675 635 665 605 635 675 665 635 565 665 655 655 645 635 Figure 5.24. Three Probe V OC Numbers for CuGaSe 2 Sample # P132 It can easily be observed that the sample performance is more uniform, in terms of the V OC numbers. (One easy way is to look at the minimum V OC which, in this case, is 595 mV, fairly close to bei ng a 600 mV device.) In addition to the above, the current density numbers were higher as well. The following table shows these J SC numbers for 9 devices on this sample. (Because of the time consuming nature of this measurement, only a limited number of devices were measured.) 11.7 11.4 12.1 10.9 12.4 12.5 12.6 11.8 11.1 Figure 5.25. Three Probe J SC Numbers (mA/cm 2 ) for CuGaSe 2 Sample # P132 It can be concluded, from the results for # P132, that increasing the thickness of the ab sorber, while making sure that there is a Cu rich to Cu poor transition of the phases during the deposition, has helped form a better quality absorber. It has been mentioned before that, according to the literature, such a transition helps form larger gra ins. If these larger grains are preserved until the end, it is understandable that the overall current density would be increased. This is because the current density changes as the grain size

PAGE 139

126 of the final film changes, especially if this size increase i s in the direction of the current flow. Moreover, the total increased thickness seems to have helped even out the local fluctuations/differences, to produce a more uniform absorber film. This could have been a result of one or more of the following. Fir st, because of the increased thickness, the surface of the absorber, which is supposed to play a crucial role in terms of the V OC s, is placed farther from the back contact as well as the glass substrate. This would, then, protect the surface from any non uniformity originating in either the back contact or the glass substrate. Secondly, if there were any vertical shunting paths present (in the direction of current flow though the absorber film), perhaps because of pin holes or metal particulates, an incr eased thickness may have helped keep these shunts away from reaching all the way through to the surface of the absorber. A third possibility is that, while increasing the total absorber thickness, the sample saw a longer period of high temperature deposit ion, along with more amount of Se. These could have helped form a more uniform absorber, thereby evening out any possible fluctuations. Until this point, it was clearly evident that Type II produced slightly better voltages, and more uniformity, but reduc ed current densities, when compared to Type I. 5.1.6. Type I B and Type II B CuGaSe 2 : Cu Se Co evaporation In an attempt to improve the V OC s and the overall performance, a small variation in the Type I and Type II recipes was explored. These two proce sses were, therefore, named as Type I B CuGaSe 2 and Type II B CuGaSe 2 respectively. In the regular Type I and Type II processes, Cu was evaporated alone, in the absence of any Se flux. In the I B

PAGE 140

127 and II B processes, Cu was deposited along with a constan t flux of Se. The primary motivation here was two fold. First, we wanted to see if the Se flux helped in the making of a more uniform material, thereby improving the properties. Secondly, if there was any significant loss of Ga Se species during the Cu deposition, the extra Se flux would help compensate for this loss. Sample # P093 was processed using the Type I B recipe. The following was the V OC performance of this sample. 645 715 705 665 725 665 715 645 715 725 665 675 635 625 615 665 655 645 505 665 405 545 605 625 445 Figure 5.26. Two Probe V OC Numbers for CuGaSe 2 Type I B Sample #P093 The highest V OC was improved to 725 mV. The rest of the sample also had a decent performance; with most devices exhibiting V OC s greater than 600 mV. Samp le # P151 was a Type II B sample (Cu Se co evaporation), with the following performance. 455 445 565 715 715 695 675 725 735 725 715 715 655 725 695 665 715 695 715 705 705 695 695 615 655 Figure 5.27. Two Probe V OC Numbers for CuGaSe 2 Type II B Sam ple # P151

PAGE 141

128 It can be easily seen that the highest V OC (735 mV) is as high as the best numbers found with any other Type I or Type II samples. Interestingly, this sample also has the maximum number of 700s so far -12 devices. These 12 devices are fair ly scattered across the entire substrate, indicating that the metal ratios across the entire substrate are not very far from each other (or, at least, not far enough so as to diminish the V OC performance significantly). A couple of conclusions can be drawn from the above two pieces of data. First, for both, Type I as well as Type II, the Cu Se co evaporated samples (The B Type samples) are better than the regular recipe samples. This means that the Cu Se co evaporation works (at least) slightly better tha n the Cu evaporation, in terms of improving the V OC performance. Second, Type II B seems to be better than Type I B. This is mainly evident in terms of the uniformity on the substrate (many more 700s for Type II B, and they are scattered over a larger r egion of the sample). The problem with this Cu Se co evaporation recipe, however, was that it was extremely difficult to control the amount of Cu in the samples. The reason was as follows. Because of the geometry of the sources in the CuGaSe 2 deposition chamber, and because of the high flux of Se that was always used during the Ga+Se co evaporation step, the inside of the chamber usually got coated with Se. Some of this Se inevitably found its way on to the Cu source (a tungsten/tantalum boat that held t he Cu pieces), in spite of the presence of a small separator shield between the Se and Cu sources. In the regular Type I and II recipes, it was fairly easy to get rid of this Se sitting in and around the Cu boat, by heating up the Cu source before the sub strate was exposed to the Cu evaporation (by opening the substrate shutter only after this Se had evaporated).

PAGE 142

129 However, during the Type B processes, because Cu and Se had to be evaporated simultaneously, it was extremely difficult to decide how much of th e material coming from the Cu source was indeed Cu, and how much of it was extra Se that might have been bouncing off the hot surfaces around the Cu boat. Because of this limitation, although slightly higher voltages were obtained, the Type B recipes coul d not be continued. 5.1.7. CdS and Other Buffer Layers At this point, it was evident that, no matter what we did to the absorber, the V OC which was the main parameter we were trying to improve, was limited to the low 700s. For the device characteristi cs, the next important layer, in the CuGaSe 2 solar cell structure, was the CdS buffer layer, along with the i ZnO layer. We turned our attention to these two layers. As mentioned in the literature survey, CdS has worked out best, as the buffer layer for C uIn(Ga)Se 2 as well as for CuGaSe 2 In our laboratory, a chemical bath deposition (CBD) of CdS had always been used as the standard buffer layer. (As a reminder: this CdS layer is usually followed by a thin layer of undoped ZnO, and then a much thicker la yer of the top contact, viz. doped ZnO.) It is, perhaps, important to mention here that this CBD CdS is one of the least understood steps in our processing sequence. Because our CuGaSe 2 recipe had evolved over time, it was warranted to carry out experime nts that would tell us if CdS was, indeed, the best buffer layer for our

PAGE 143

130 devices, and, if it was, then it needed to be seen if we could, somehow, optimize the CdS process for our CuGaSe 2 recipe. At this point, it is important for the reader to be aware of some details of our CBD process. The process typically began with a mixture of 150 ml water and 27.5 ml Ammonium Hydroxide (1 Molar), which was then heated slowly, while being stirred. When the mixture reached 30 0 C, the sample was placed in this solution and 22 ml each of Cadmium Acetate (0.015 Molar) and Thiourea (0.15 Molar) were added. When the solution/sample temperature reached about 74 75 0 C, the solution typically started turning yellow, because of the sulfur containing precipitation. About a hal f minute or so later, at about 77 0 C, the sample was taken out of the solution, rinsed, and blow dried. Meanwhile, it is also worthwhile to recall an important piece of information from the recent literature, where the best CuGaSe 2 performance so far was a ccomplished by tweaking the CBD CDS process, by raising the CdS precipitation temperature to about 80 0 C [Rau, et al.]. This information was used in designing the last two of the experiments described below. To see how the CdS process affected our devices, we set out by processing a sample where we skipped the CdS layer completely. Not surprisingly, this sample turned out to be ridiculously low in performance. We then carried out a series of experiments where we used the following processing variations, o ne by one, with the Type I CuGaSe 2 absorber recipe: (i) A double CdS layer was deposited, where the entire sequence of a CBD deposition was repeated. The performance was very poor.

PAGE 144

131 (ii) A set of experiments involved samples that were heated outside, independently of the CdS bath, and immersed in the solution directly at various stages of the precipitation (in the range of 74 78 0 C). The performance was still quite poor. (iii) (#P161) The sample was taken out approximately at the point where the solution would start turni ng yellow (74 0 C). The sample, hence, was thought to have a thinner CdS layer in this case. The performance was as good as a standard CdS process, in terms of the voltages and currents. (iv) The amount of Thiourea was increased by about 25% for a sample. This resulted in a poor performance. (v) In order to raise the precipitation temperature, the sample was heated outside to about 88 0 C, and then immersed in the CdS bath, at about 75 0 C. The exact temperature of the sample at the time of precipitation could not be known, although, it had to be somewhere between the two temperatures stated above. The performance was low, and there were a number of small particulates on the sample. (vi) In order to raise the precipitation temperature, the amount of Ammonium Hydroxide was significantly increased (approximately doubled). The precipitation temperature was raised to 80 0 C, and the sample showed good V OC s. However, the currents were extremely low, indicating that the absorber was attacked severely by the chemicals. It can be seen, from the above list, that many of the experiments did not produce good performance. In the very last experiment, although the CuGaSe 2 precipitation could be increased to 80 0 C, there were other problems, and, hence, this direction of investigation w as abandoned (in favor of other things).

PAGE 145

132 It is interesting to note, however, that the third experiment in the above list, where the sample was taken out of the CdS bath rather prematurely (before the precipitation began), yielded excellent results. Most o f the devices on this sample exhibited a V OC of >600 mV. Moreover, the V OC numbers were more evenly spread out, indicating that this sample was better in terms of the uniformity. This was an important result in two ways. It was thought (at least in our laboratory) that the precipitation stage was crucially important in terms of forming the much needed buffer layer on the absorber film. Secondly, it was known, from past literature, that the CdS bath also passivated the absorber surface, thereby reducing the possibility of interface defects. It was not known when this passivation exactly occurred, or when the passivation was completed, during the CdS process. From the above experiment, it could be concluded that the passivation was completed around the t ime the precipitation began. It is possible that, in the experiment, the precipitation had actually begun, but wasnt quite visible yet, when the sample was taken out. In any case, it is reasonable to say that either the necessary CdS layer was formed be fore significant precipitation occurred, or a thinner CdS layer was actually sufficient for the sample. Because of practical limitations, we were unable to find out which one of these was the case. Also, taking the sample out of the bath before a signifi cant amount of precipitation occurred resulted in better uniformity, suggesting that leaving the sample in the solution for too long may hurt the sample. At this point, we decided to try out other buffer layers as possible replacements for the CBD CdS lay er. A number of other layers, such as evaporated Gallium Selenide (Ga 2 Se 3 ) with and without a vacuum break, evaporated Indium Selenide (In 2 Se 3 ), a CdS layer followed by another layer of evaporated In 2 Se 3 evaporated Zinc Indium Selenide

PAGE 146

133 (ZISe), and a CdS layer followed by evaporated ZISe, were attempted. Along with some of these experiments, the i ZnO layer thickness was varied, to see the effect on the properties of the devices. None of these new buffer layers worked nearly as good as the CBD CdS. The refore, the results are not presented here. CBD CdS, as the prevailed winner, was continued as the buffer layer of choice. 5.1.8. Light Soaking and Annealing Experiments for CuGaSe 2 To better understand the characteristics of our CuGaSe 2 samples, we dec ided to explore the effects of Light Soaking and high temperature annealing. Some of the CuGaSe 2 samples were light soaked under one sun illumination, for about 10 minutes. A three probe I V measurement was then carried out on selected devices, after let ting the soaked sample cool down for about 30 minutes. The measurement was then compared with that before the light soaking (i.e., the as deposited measurement). Some of the samples were annealed in air, at two different temperatures: Anneal #1 at 125 0 C, and Anneal #2 at 200 0 C. Some other samples were annealed only at the higher temperature (200 0 C). The results obtained are discussed next. Sample # P098 was processed with a Type II CuGaSe 2 recipe. The following figures (shown on the next 2 pages) depic t the three probe curves for device # 6 on this sample. The four figures show the three probe I V curves for the device at 4 different times: As deposited, after light soak, after the 1 st anneal, and after the 2 nd anneal.

PAGE 147

134 Figure 5.28a. Three probe I V for P098 06: As dep/Light Soak

PAGE 148

135 Figure 5.28b. Three probe I V for P098 06: Anneal 1/Anneal 2

PAGE 149

136 The following observations can be made from these plots. First of all, the V OC progressively changed from 571 to 600, to 616, to 660. This means that the l ight soak has had some effect on the device, increasing the V OC by 29 mV. The first, low temperature (125 0 C) anneal improved the V OC by another 16 mV, whereas the second, high temperature (200 0 C) anneal increased it by 44 mV. We will compare this V OC beh avior to another set of 4 figures, which are for another device from the same sample. However, before that, there is another thing that can be noticed in the above plots. The very first (as deposited) curve shows some amount of crossover of the dark and t he light curves, in the 3 rd quadrant of the graph. This (unwanted) behavior has been seen in a number of CuGaSe 2 (and CuIn(Ga)Se 2 ) samples, to various degrees. The worst (maximum) crossover we have seen seems to be present in samples where the ZnO deposi tion process had problems. An example of such problems is: the sample going through an extra heating cycle before the ZnO deposition, because the run had to be shut down due to problems with gas pressures or the ZnO target. It is quite possible that, bec ause of such oddities in the processes after the absorber deposition, either the junction interface and/or the top layer of the absorber got disturbed, perhaps leading to additional defect formations, which showed up as crossovers in the I V characteristic s. The amount of this crossover is almost equal for the first two plots, then it increases some, in the 3 rd plot (this is rather hard to see because of the changed scale), and increases substantially in the 4 th plot. This seems to indicate that, for a sa mple that had some crossover to begin with, the light soaking has a small effect, whereas the annealing, especially the one at high temperature, seems to have a substantial effect.

PAGE 150

137 We now turn our attention to another device, # 24, from Sample # P098 (the same sample as above). The following four figures (on the next two pages) show the behavior of this device before and after light soaking and annealing, just like the previous set of figures. In this case, the V OC started out at 520 mV, then decreased by 12 mV after the light soaking, and increased by 3 mV after the first anneal. Up until this point, the change was rather small. However, when the sample was annealed at a high temperature, the V OC increased by 31 mV. Comparing the above observations to the results of the previous device, the main difference seems to be the behavior after the light soaking. In the previous case, the V OC increased after the light soaking, whereas, here, it decreased a little. In fact, after carrying out similar study on a number of other samples, it became evident that the V OC fluctuated in both directions after light soaking, and there was no unique, common trend. However, it became very clear, that the annealing, especially the high temperature (200 0 C) one, always res ulted in substantial increases in the V OC numbers. It is also interesting to note that, just like the previous device, this device shows changes in the crossover phenomenon. Although the device started out with a minimal crossover, which remained fairl y constant after the light soak, it became worse after the first anneal, and the second anneal deteriorated it drastically. Well analyze the annealing behavior more, with the help of a few more examples.

PAGE 151

138 Figure 5.29a. Three probe plots for P098 24: As dep/Light Soak

PAGE 152

139 Figure 5.29b. Three probe plots for P098 24: Anneal 1/Anneal 2

PAGE 153

140 Let us, now, look at the annealing results for another sample, # P082. Unlike the previous one, this was a Type I CuGaSe 2 sample. The V OC results for this sample were p resented in the CuGaSe 2 Type I section. At that time, it was also mentioned that device #12 from this sample had a decent fill factor, of 55%. The sample was later annealed at 200 0 C, and the following figures show the before anneal and after anneal resul ts for device #12. Figure 5.30a. Three probe for Sample # P082, Device # 12: As dep

PAGE 154

141 Figure 5.30b. Three probe for Sample # P082, Device # 12: After Anneal From the above figures, it can again be seen that the annealing has improved the V OC from 699 mV to 714 mV. Although this increase was rather small, compared to the increases seen in the V OC s for devices from the previous sample, it has to be remembered that the starting V OC (699 mV) was much higher in this case. The increase in this case was pr obably limited by another mechanism controlling the interface properties. It should also be recalled that the high V OC limit experienced with Type I samples is lower than that experienced with Type II samples. Because the sample in question is a Type I s ample, with a V OC as high as 699 mV, it probably already was closing in on the limit, and therefore, the V OC improvement was limited.

PAGE 155

142 Another interesting observation that can be made from the above figures is that, as the V OC increased because of the high temperature annealing, the I SC has actually gone down, from about 1.16 mA to about 0.73 mA. This behavior has been quite representative of the annealed samples, meaning that the I SC always seemed to go in the opposite direction of the V OC This is rather discouraging, because this meant that the V OC and the I SC behaviors could not be separated, and hence, could not be improved independently of each other. However, this behavior seems to be consistent with the fact that, while the best V OC s obtained with the Type II recipe are higher than those obtained with Type I, the I SC s (and the J SC s), are, in fact, lower. One primary difference between the two recipes is that the Type II absorber is exposed to a higher temperature for a much longer period of time This indicates to an effect of longer periods of high temperature on the absorber properties. Although the recipe change relates directly to the formation of the absorber, whereas the annealing could be thought of only re arranging the absorber or the interface, both do involve the application of temperature over long periods of time. Another experiment was attempted, where a sample was annealed in between the depositions, unlike the above samples, which were annealed after all the depositions were com pleted. The sample was # P115, and was presented in the Type II section before (although, the annealing wasnt discussed in detail). The bottom piece of this sample (the bottom three rows) was annealed in air for 10 minutes, at 200 0 C, right after the Cd S buffer was deposited. The top piece (top two rows) was the control piece, meaning that it was processed without any annealing. This was an attempt to see if an intermediate annealing, which had been claimed in some research papers to be helpful, would improve our devices as well. The following was the V OC performance of this sample.

PAGE 156

143 705 685 655 685 665 665 735 725 635 665 735 775 775 775 725 535 695 755 765 735 495 715 595 505 555 Figure 5.31. Three Probe V OC Numbers for CuGaSe 2 Sample # P115 The top two control rows contain devices that have V OC s of 735 and 725, the two highest numbers seen for our regular (non treated) CuGaSe 2 processing. What is even more interesting is that the bottom three rows, which were annealed after CdS, produced V O C numbers as high as 775 mV! This number was the highest of all samples produced (treated or non treated) in this research project. It is possible that the annealing helped the CuGaSe 2 CdS interface properties by passivating it better, by activating some kind of diffusion mechanisms, whereby, ions traveled into the absorber to reduce the defect density present at or near the interface. Such a defect reduction would then result in the reduction of the dark current, thereby producing a higher V OC 5.1.9. Capacitance Studies of CuGaSe 2 To gain more understanding about characteristics such as the nature of the junction formed in the CuGaSe 2 the depletion width, and the doping concentration, we decided to study the capacitance behavior of our samples. Typ ically, a Capacitance Frequency curve was obtained for several devices on a sample, as a screening mechanism. On a few good devices, a dark C V and a light C V measurement was carried out. This procedure has been described in detail in a previous chapter

PAGE 157

144 Sample # P082 was a Type I sample that yielded decent performance, and the I V characteristics of this sample have been described in detail before. A few devices on this sample were then selected for the C V measurements. The following figure shows t he A 2 /C 2 vs. V curves (dark and light) for device # 1 on this sample. (A is the area of the device, C, the capacitance, and V, the voltage applied.) P082-01 0.00E+00 1.00E+15 2.00E+15 3.00E+15 4.00E+15 5.00E+15 6.00E+15 7.00E+15 8.00E+15 9.00E+15 -3 -2 -1 0 1 Voltage (V) A2/C2 (cm2/F2) Dark Light Figure 5.32. A 2 /C 2 Vs. V Curve for Device # 1 on CuGaSe 2 Sample # P082 As c an be seen in the figure, the dark curve and the light curve do not overlap. We have seen this behavior with all of our CuGaSe 2 devices, as well as the CuIn(Ga)Se 2 devices processed in our laboratory [Jayapalan, et al.]. The underlying mechanism for the increased capacitance in light has been known to be the trapping of light generated carriers, which results in the change in the width of the space charge (depletion) region [Shankaradas]. In the case of our CuIn(Ga)Se 2 devices, these traps have been corr elated

PAGE 158

145 to recombination centers that seem to influence the V OC of the devices [Jayapalan]. It is evident that such traps are present in our CuGaSe 2 devices as well. However, attempts to find out a correlation between them and the V OC were unsuccessful. T he calculation of the depletion width in dark and light, in the above case, yielded the values of 617 and 515 A 0 respectively. To find out the doping concentration, a slope value is needed, from the A 2 /C 2 Vs. V curve. For this purpose, the near flat reg ion of the curve, from 1.5 Volts to 1.0 Volts, was selected. This is primarily because, at lower voltages (less reverse bias), the capacitance value may be affected by the forward capacitance. The following figure reproduces these selected regions of th e above curves, along with the linear equations derived by curve fitting. P082-01 y = -1E+15x + 3E+15 y = -1E+15x + 6E+15 0.00E+00 1.00E+15 2.00E+15 3.00E+15 4.00E+15 5.00E+15 6.00E+15 7.00E+15 8.00E+15 -2 -1.5 -1 -0.5 0 Voltage (V) A2/C2 (cm2/F2) Dark Light Linear (Light) Linear (Dark) Figure 5.33. Truncated A 2 /C 2 Vs. V Curve for Device # 1 on Sample # P082 For the above device, the doping concentration was calculated to be 5E15/cm 3 This value agreed with another device from the same sample. Most of the devices

PAGE 159

146 yielded the values in the range of 1E15 to 5E15/cm 3 with the exception of one device, which had a value as high as 6E16. (This particular device seemed to have an unusually small depletion width, as discussed later.) According to the theory of the C V measurement, the extrapolation of the straight line C V curve to the voltage axis is supposed to yield the built in voltage for the device. In the above case, it can be seen that such an extrapolation would give a value of about 2 Volts. A few other devices on this sample had a similar behavior. This, obviously, is too high to be the real built in voltage. This may be a result of the inaccuracy originating from any shunting that may be present in these devices, thereby limiting the usefulness of the measurement itself. Similar measurements were done on a few other selected samples. Sample # P082, presented above, was processed with a Type I CuGaSe 2 recipe. Another sample, # P115, was a Type II sample. A third sample selected, P093, was processed with a Type I B recipe. As a reminder, the B type recipes included a low Se flux while the Cu was being deposited. The primary motivation behind this variation was two fold. Fi rst, we wanted to see if the Se flux helped in the making of a more uniform material, thereby improving the properties. Secondly, if there was any significant loss of Ga Se species during the Cu deposition, the extra Se flux would compensate for this loss C V measurements were carried out on several devices from the above 3 samples, and the following figure presents the dark depletion width values (in Angstroms) for these devices. (The devices came from various locations from a sample, and the x axis me rely represents the number of device, not to be confused with the number representing the location with respect to the sources.)

PAGE 160

147 Dark Depletion Widths 0 500 1000 1500 2000 2500 0 2 4 6 8 10 Device # Depletion width (A) P082-TYPE I P093-TYPE I with Cu-Se P115-TYPE II Figure 5.34. Dark Depletion Widths for Several Devices on 3 CuGaSe 2 Samples Keeping the limitedne ss of the measurements in mind, a couple of important observations can be made from the above figure. First, there is a large variation among the depletion width values of devices from sample # P082. The values range from about a 100 to over 2000 A 0 Th e smallest value of depletion width corresponded with the device that showed a high doping concentration of 6E16 before. However, four of the seven values are around 500 to 700 A 0 and it seems likely that this is a good indication of where the true value s may be. This is possible, especially because the variation in the values for devices from the other two samples is much smaller. For sample # P093, the values are between 100 and 200 A 0 while those for sample P115 are mostly between 30 to 50 A 0 Alt hough all of these values (especially for # P093 and # P115) are too low compared to some of the values we have previously seen with our CuIn(Ga)Se 2 samples (several hundred nanometers), there is a trend that can be observed. As a reminder, Type I devices always had better current performance than Type II devices, and it is reasonable

PAGE 161

148 to see Type I depletion widths to be larger than those of Type II devices. Therefore, in the present case, it is no surprise that both, Type I (P082), as well as Type IB (P0 93), samples have better depletion widths than the Type II sample (P115). 5.2. PART II: CuGaSe 2 Simulation/Modeling Results Even after a number of processing experiments, the V OC s of our CuGaSe 2 devices were still limited to about 700 800 mV. To be abl e to better understand this performance limitation, we decided to employ two simulation techniques SCAPS and AMPS to model the device behavior. The primary motivation was to see whether the problem existed at the heterojunction interface, or in the deep absorber bulk, or at the back contact. The following sections describe selected results from this study. 5.2.1. SCAPS Modeling The first technique, called Solar Cell CAPacitance Simulator (SCAPS), was developed by Prof. Marc Burgelman and his colleague s at University of Gent, in Belgium. The technique had previously been used to model CdTe and CuIn(Ga)Se 2 solar cells. For our simulations, we used the SCAPS 1D, version 2.1, which was the latest version available at the time. First, a little bit about the SCAPS technique. The program simulates the electrical characteristics for thin film heterojunction solar cell structures. An arbitrary number of semiconductor layers, with arbitrary doping profiles (as a function of position) and

PAGE 162

149 arbitrary energetic distributions of deep donor and/or acceptor levels can be introduced in the semiconductor bulk or at the heterojunction interface, and the effects of these can be observed in the electrical characteristics such as I V, C V, etc. For our simulation purpos es, we focused on the I V behavior of CuGaSe 2 solar cell structures. We used the following typical values for the various parameters, unless otherwise specified later in the discussion. Table 5.1. SCAPS Simulations: Layers and Typical Parameter Values La yer # Layer Function in structure Parameter name Parameter value 1 Back contact (Moly) Work function 4.80 2 Absorber (p type) (CuGaSe 2 ) Thickness 1.5 m m Bandgap 1.68 eV Affinity 3.32 Acceptor density 1E+17 Absorption constant 1E+5 3 Heteroju nction partner (CdS) Thickness 550 A 0 Bandgap 2.42 Affinity 4.0 Donor density 1E+14 4 Buffer layer (i ZnO) Thickness 550 A 0 Bandgap 3.45 Affinity 3.7 Donor density 1E+16 5 Front contact (ZnO) Work function 3.7 A number of simulation runs were carried out, where various defect levels and defect densities were introduced in the absorber bulk and/or at the interface between the

PAGE 163

150 absorber and the heterojunction partner. The following pages contain the summary of these simulations. The t able below has information about simulations where there were only interface defects present (no bulk defects). The first column denotes the file name, i.e. the name of the simulation run. The next column has the information about the type, density, and location (from the valence band) of the interface defects. The last four columns show the results obtained for the respective I V simulation, in terms of the V OC the J SC the fill factor, and the conversion Efficiency. (Note: The very first run had no d efect states at all, and is, therefore, named as J_base, meaning the base run ) Table 5.2. SCAPS Simulation Parameters and Results: Interface States Only (No defects for J_base, the first row) File name Interface defects Type/ density/ location V OC Volts J SC mA FF, % Eff, % J_base No defects 1.28 14.36 69.0 12.65 J_i1_1 Neutral/ E12/ 0.60 0.486 14.32 70.2 4.88 J_i1_2 Donor/ E12/ 0.60 0.380 14.49 61.1 3.36 J_i1_3 Acceptor/ E12/ 0.60 0.487 14.32 70.1 4.89 J_i1_6 Neutral/ E12/ 0.40 0.485 14.33 70.2 4. 88 J_i1_7 Donor/ E12/ 0.40 0.383 14.40 63.5 3.50 J_i1_8 Acceptor/ E12/ 0.40 0.488 14.29 70.5 4.91 J_i1_9 Neutral/ E12/ 0.20 0.522 14.36 77.0 5.77 J_i1_10 Donor/ E12/ 0.20 0.513 14.36 78.3 5.77 J_i1_11 Acceptor/ E12/ 0.20 0.554 14.11 77.5 6.05 The f ollowing figure shows the light I V graphs drawn for some of the above simulation runs, viz. J_base, J_i1_1, 2, 9, 10, and 11.

PAGE 164

151 Figure 5.35. I V Plots for SCAPS Simulations (Interface Defects only) A couple of important observations can be made from the results of this set of simulations. Firstly, the J_base run, where there were neither bulk states nor interface states introduced, produced a V OC of 1.26 Volts, with current and FF values of 14.36 mA/cm 2 and 69%, respectively. Because the a im was to simulate a device as close to our processed devices as possible, we had to introduce defect states so as to bring this simulated V OC down. A series of runs, where there were different types of interface defect states introduced at different loca tions from the valence band edge, all with the J_base J_i1_11 (dotted) J_i1_11 (solid ) J_i1_2 J_i1_9 J_i1_1

PAGE 165

152 density of 1E12, brought this V OC down in the 400 500 mV range. Specifically, the last three runs, where the V OC s are in the 500s, and the J SC values are around 14 mA, closely resemble the best results prod uced with our early Type I CuGaSe 2 devices. However, our best fill factors were less than 60%, which does not agree with the fill factors simulated here, of about 77%. This fact is also reflected in the conversion efficiency numbers, where the simulated numbers of 5 6% are on the higher side of the best efficiencies achieved with our processed sample (slightly less than 5%). This may primarily be because of the extra series resistance that was present in our processed samples, perhaps coming from the pre sence of multiple (not so conductive) phases of the ternary absorber, and defects present in the absorber bulk. One of the conclusions from our Type I and Type II CuGaSe 2 processing experience was that as we successfully raised the V OC (from 500s to 600 s and rarely 700s), the J SC values went down from 13 15 mA to about 10 12 mA. For example, for our Type II samples, the typical average V OC s were in high 600s, whereas the J SC s were in the range of 10 12 mA. To be able to simulate this behavior, we n eeded to introduce other defect states into the structure. The following table includes the parameters used, and the results obtained, for the simulations, where both, interface defects as well as bulk defects were introduced in the solar cell structure. Here, the bulk defect states were two acceptor type densities of 2E17 and 1E17, at 0.25 and 0.13 eV from the valence band edge respectively. These numbers were borrowed from a recent publication where these specific levels were suggested to be present in CuGaSe 2 [Zunger].

PAGE 166

153 Table 5.3. SCAPS Simulation Parameters and Results: Bulk and Interface States (Bulk defect states: acceptor/ 2E17/ 0.25, and acceptor/ 1E17/ 0.13) File name Interface defects Type/ density/ location V OC Volts J SC mA FF, % Eff, % J_p2_1 No defects 1.26 10.81 60.5 8.23 J_p2i1_2 Neutral/ E12/ 0.80 0.521 10.79 73.4 4.12 J_p2i1_3 Donor/ E12/ 0.80 0.433 11.34 74.4 3.65 J_p2i1_4 Acceptor/ E12/ 0.80 0.521 10.79 73.4 4.12 J_p2i1_5 Neutral/ E12/ 0.60 0.520 10.78 73.4 4.11 J_p2i1_6 Don or/ E12/ 0.60 0.363 11.26 60.3 2.46 J_p2i1_7 Acceptor/ E12/ 0.60 0.520 10.78 73.4 4.11 J_p2i1_8 Neutral/ E12/ 0.40 0.520 10.78 73.4 4.11 J_p2i1_9 Donor/ E12/ 0.40 0.369 11.10 65.6 2.68 J_p2i1_10 Acceptor/ E12/ 0.40 0.520 10.78 73.4 4.11 J_p2i1_11 Neut ral/ E12/ 0.20 0.532 10.81 75.2 4.32 J_p2i1_12 Donor/ E12/ 0.20 0.509 10.84 77.6 4.28 J_p2i1_13 Acceptor/ E12/ 0.20 0.552 10.45 76.8 4.43 J_p2i1_14 Neutral/ E10 / 0.80 0.670 10.81 77.2 5.60 J_p2i1_15 Donor/ E10/ 0.80 0.669 10.82 77.2 5.58 J_p2i1_16 Acc eptor/ E10/ 0.80 0.670 10.81 77.2 5.60 J_p2i1_17 Neutral/ E10 / 0.20 0.675 10.81 77.7 5.66 J_p2i1_18 Donor/ E10/ 0.20 0.673 10.81 77.8 5.66 J_p2i1_19 Acceptor/ E10/ 0.20 0.675 10.80 77.7 5.67 As can be seen from the above table, the first run, where there were two bulk defects present (but no interface defects) produced a very high V OC of 1.26 V. However, the J SC number, of 10.81 mA, was already in the range of that for the processed devices. When, in addition to the two bulk defects, some interface defects were introduced at the density of 1E12, the V OC dropped down to the 400s and 500s. These

PAGE 167

154 numbers, however, were too low, when compared with the V OC s obtained with our processed samples (most of which are in 600s and 700s). Reducing the defe ct state density down to 1E10 brought the V OC back up in high 600s, while still maintaining the J SC value at around 11 mA. However, once again, the simulated fill factors are somewhat high, and this fact is reflected in the high efficiency of about 5.6%. The following figure depicts the I V characteristics for some of the simulation runs mentioned above. The high fill factors are quite evident from the significant squared ness of the curves. Figure 5.36. I V Plots for SCAPS Simulation (Bulk and Interface Defects) J_p2_1 J_p2_i1_6 J_p2_i1_14 J_p2_i1_13 J_p2_i1_5

PAGE 168

155 The above results closely resemble our processed devices, except for the fill factor values. The low fill factors, obtained for the actual devices, may be a result of a combination of a high series resistance and a low shunt resi stance in our samples, which could not be accurately simulated. Another factor that may affect the V OC and overall performance, of samples is the back contact. We decided to see how the change in the work function of the back contact affected the I V cha racteristics. The following figure shows the I V curves for three values of the back contact work function: 4.53, 4.80 (base value), and 4.65. The absorber layer had the same bulk defects as before: 2E17 at 0.25 eV, and 1E17 at 0.13 eV. Figure 5.37. I V Plots for SCAPS Simulation: Back Contact Work Function

PAGE 169

156 As is easily evident from the above figure, although the currents seem to be okay, the V OC s are still very high. It proved impossible to bring the voltage down substantially, while keeping other p arameters in a reasonable range, by varying the back contact work function. The question about whether CuGaSe 2 forms a true heterojunction has been heavily debated in the recent past. The alternative to a true heterojunction, of course, is a buried homoju nction inside the CuGaSe 2 absorber layer. For this to happen, a thin layer near the surface of the absorber would have to be n type with respect to the deeper, p type bulk of the absorber. Some recent publications, as mentioned in the literature survey, suggest the presence of such an n type layer in CuIn(Ga)Se 2 However, CuGaSe 2 has been suggested to be favoring a true homojunction. We decided to employ the SCAPS technique to see what such an n type top layer, if present, would do to our devices. To accomplish this, a thin (0.1 m m) CuGaSe 2 layer, called CGS2, with 1E14 shallow donors, was introduced on the top of the regular CuGaSe 2 absorber layer. Moreover, we also wanted to see the effect of presence or absence of other layers, such as CdS and ZnO. The following figure shows 7 I V curves. Each of these curves is associated with a specific set of conditions (as depicted with roman numerals in parentheses beside the curves), and the conditions are described in the text below the figure.

PAGE 170

157 Figure 5.38. I V Plots for SCAPS Simulation (various conditions, described in text) The regular CuGaSe 2 layer had two acceptor type defect states, with densities 2E17 and 1E17, at 0.25 eV and 0.13 eV from the valence band edge, respectively (same as t he simulations presented before). Other conditions, for the CGS2 layer, as well as other layers in the structure, were: (i) No CdS or i ZnO present; CGS2 had 2 defect states, same as the CuGaSe 2 layer. (ii) No CdS or i ZnO present; No defect states for CGS2. (iii) CdS, i ZnO present; CGS2 had 2 defect states. # (i) # (ii) # (iii) # (iv) and (vii) # (v) # (vi)

PAGE 171

158 (iv) CdS, i ZnO present; No defect states for CGS2. (v) CdS present; No i ZnO; No defect states for CGS2. (vi) i ZnO present; No CdS; No defect states for CGS2. (vii) CdS, I ZnO present; 1E12 donors in CGS2. The most salient feature of the curves is that, whenever there is either CdS or i ZnO (or both) present, there is a distinct kink that can be seen near the voltage (x) axis. This indicates that, because of the extra n type CGS2 layer present in the structure, there is an occurrence of some sort of a double junction in the structure. The only curves that look normal are the ones where there is neither CdS nor i ZnO present. Interestingly enough, a few devices from one of our samples had shown the presence of a strong kink near the voltage axis, somewhat like the one seen in the simulated figure. The I V plot of one of these devices is reproduced below. (It should be noted that the 3 probe I V plot is oriented differently, compared to the Simulation I V.) Figure 5.39. I V kink in Device # 1 from CuGaSe 2 Sample # P137

PAGE 172

159 However, there was a problem during the ZnO deposition for the above sample. The sample was heated before the actual deposition, but it was realized that the run had to be abandoned because of an issue with the Z nO target. The ZnO deposition was done later, after the issue was resolved. However, the sample had gone through an additional anneal (at about 120 0 C) for a few hours, in inert atmosphere. The occurrence of a kink such as the one shown above suggests th at there may have been a double junction present in this particular device. This could happen, for example, because of processing problems, such as incorporation of unwanted elements which may have been left behind in the absorber deposition chamber, afte r, say, CuIn(Ga)Se 2 processing. However, the ZnO processing problem suggests that the extended annealing may have hurt the junction, and perhaps, given rise to unwanted diffusion of elements, resulting in the above behavior. However, as mentioned above, such behavior is very rare, and, therefore, seems to suggest that the simulated results dont really match with our processed devices. It also insinuates that the CuGaSe 2 absorber film does not have an n type surface layer, so the CuGaSe 2 devices are the true heterojunction type devices. In summary, we had limited success in simulating the behavior of the actual devices, by using SCAPS. The three important conclusions were: (i) Simultaneous introduction of Interface and Bulk defects in the solar cell structu re generated results that were close to the actual device results, (ii) Changes in the back contact work function could not bring the voltage down to a reasonable value, and (iii) N layer simulation experiments indicated the presence of a heterojunction.

PAGE 173

160 5.2.2. AMP S Modeling AMPS 1D (Analysis of Microelectronic and Photonic Structures: One Dimensional Approach) was developed by the Pennsylvania State University, in collaboration with the Electric Power Research Institute. The main difference between the SCAPS and AMPS techniques, as it relates to the modeling of our devices, was that SCAPS allowed the incorporation of interface defect states using a rolled up parameter called interface recombination velocity (somewhat like the surface recombination velocity used a t the semiconductor and metal surfaces). In a way, this made it easier to experiment with the interface defects, just by changing this velocity number. In AMPS, on the other hand, no such parameter was available. The same effect had to be produced by in corporating defects in the layers that formed the interface, and by manipulating other characteristics such as the affinity and the Bandgap, etc. Another relevant difference between the two techniques was that AMPS used the so called back contact energy to specify where the back contact energy bands were, with respect to the bands of the semiconductor (absorber) layer. This value is calculated from the conduction band edge (Ec) of the CuGaSe 2 absorber. In SCAPS, the relevant parameter was the work funct ion of the contact. In terms of the graphical representation of I V characteristics, the AMPS plots are inverted, as will be seen shortly. Also, because the AMPS plots show only the one side of the voltage axis, the forward curve can be seen on the same side of the voltage axis, and this results in a negative sign being attached to the fill factor and V OC numbers, with the J SC positive. Similarly, the AMPS energy band diagrams typically show the

PAGE 174

161 equilibrium band diagram at the bottom, and the non equili brium diagram at the top of the figure. Once these minor details are understood, its easy to study these AMPS plots. The simulation runs that we carried out are described next. The figures are all together, after the end of this discussion. Most of th e parameters used for the solar cell structure were the same as the ones used for the SCAPS simulations, unless otherwise specified. However, our approach to AMPS was a little different. The CuGaSe 2 absorber was created using the main, base layer, along with a thinner (0.1 m m) top layer, so that properties of the bulk and the surface (or near surface) of the absorber could be controlled rather independently. The thin top layer will sometimes be referred to as the n layer or n CuGaSe 2 layer, although, al l this means is that this layer may be similar to, or slightly n type with respect to, the base layer, depending upon the doping levels. It has been suggested by Zunger, et al. that the increase in the band gap (Eg) of CuIn(Ga)Se 2 because of increased Ga percentage is due to the conduction band edge (E C ) moving up, while the E V stays the same. We, hence, started out our CuGaSe 2 simulations by using the known CuIn(Ga)Se 2 parameters, and then lowering the affinity to 3.35, and increasing the band gap to 1. 6 eV. The front electrode contact was assumed to be ZnO, the same as that for CuIn(Ga)Se 2 The affinity value used for ZnO was 3.7 eV. This first run is labeled as basecgs and Figure 5.40 depicts the two band diagrams for this run. The bottom diagram is at equilibrium, while the top one is drawn at 1 Volt forward bias. In forward bias, the bands in the CuGaSe 2 are beyond the flat band condition. However, because of the low value of the affinity, the barrier presented by the high E C is quite high, and prevents electrons from entering. Figure 5.41 shows the light I V characteristics for

PAGE 175

162 the basecgs run (The solid lines. The dashed curve is discussed next). It can be seen, from the I V curve, that the efficiency is nearly 18%, with J SC at 18 mA/cm 2 and V OC at 1.17 Volts, with an FF of about 80%. The next task was to lower the V OC It has been claimed that the easily formed defect Cu Ga is at the same location (0.29 eV from E V ) as the defect Cu In [Zunger]. Hence, a defect was added to the n and base lay ers, at 1.31 eV from the E C at the densities of 1E17 and 1E18, respectively. The dashed I V curve in Figure 5.41 is the result of this simulation. As can be seen, there is only a small drop in the V OC whereas the J SC and ff are significantly reduced. N ext, acceptor type defects were placed at midgap, but this did not seem to have a strong effect on the characteristics. The large band gap of the absorber material seemed to control the behavior. Therefore, the n layer was doped to the 1E17 level. This would create the junction inside of the absorber material, between the n and base layers, thereby making it a buried homojunction. Figures 5.42 and 5.43 show the resulting band diagram and the I V behavior, respectively. As can be seen, while the J SC ff and efficiency were all strongly affected, the V OC still remained at over 1 Volt. Moreover, this seemed to distort the I V curve, with a distinct kink like behavior, which is not normally observed in the processed samples. Until this point, it was clea r that nothing could bring the V OC down to the level seen in our processed devices, although the other parameters were very much in the range of the processing results. As the primary cause of this was thought to be the height of the Ec above the contact, we decided to increase the affinity value of CuGaSe 2 to see if anything changed. The affinity was increased to 3.65 eV. The results are displayed in

PAGE 176

163 Figure 5.44 (solid lines). As can be easily seen, the I V curve shape was recovered, but the V OC still stayed high. Next, the acceptor defect density in the n layer was raised to the 1E21 level. As seen in the dashed curve in figure 5.44, this brought the V OC down to 761 mV. However, the J SC dropped to a meager 2.7 mA/cm 2 an unreasonably low value. It seemed that, to be able to lower the V OC s, while maintaining the J SC in the range of 10 13 mA/cm 2 the affinity would have to be increased further. However, this did not agree with the recent theoretical work in the literature, and, therefore, we decided to try a different approach, as described next. Figure 5.45 presents the spectral response curves for the above simulation experiments. Another parameter that has a strong effect on V OC is the back contact. We, therefore, decided to vary the back contac t energy in the next few simulation runs. As a reminder, this energy is calculated from the conduction band edge of the bulk CuGaSe 2 layer. For the basecgs run, we had used a value of 1.6 eV. This value was now lowered, first to 1.0 eV, and then to 0.8 eV. The results are shown in Figure 5.46. It was found that, once the contact was above the E f a proportional drop occurred in the V OC value. For the 1.0 eV back contact energy case, the V OC dropped down to 833 mV. Other values were: J SC = 15.717 mA/c m2, efficiency = 8.303% and ff = 60%. Although the J SC and ff were around the highest we had seen in our Type I devices (15.2 mA/cm2 and 58%, respectively), the other numbers were still too high. The V OC needed to be around 600 700 mV, and the efficiency around 5%. The back contact energy value of 0.8 eV yielded such results, those being: V OC = 633 mV, J SC = 15.437 mA/cm 2 efficiency = 5.977%, ff = 58.1%. This seemed to be in reasonably good agreement with our (Type I) devices. The

PAGE 177

16 4 slight discrepancy i n the efficiency is a result of the simulated J SC being slightly on the higher side of the J SC value of the processed samples. The same figure (Figure 5.46) includes a few more curves, obtained by adding some defects in the absorber. It can be easily see n that these defects have essentially distorted the I V curve shapes, and, hence, cannot be accepted as describing our processed samples. The two cases mentioned above (1.0 eV and 0.8 eV) are the ones that did not have these additional defects, and these yielded normal curve shapes. There is yet another similarity between the above simulation run and the results of our processed devices. For the processed samples, the I V curve typically shows an upward slope as the reverse bias increases. This charact eristic slope is also observed in the above mentioned simulated I V curves. Such a behavior could either be indicative of shunting in the samples, or, more likely, could result from widening of the depletion region, leading to enhanced carrier collection, as the reverse bias increases. This, then, means that the back contact has a stronger effect on lowering the V OC for the CuGaSe 2 devices. (Changes in other parameters, such as defects, can lower the V OC to some extent, but typically result in poor curve shapes.) Indeed, the textbook value (Sze) of 4.8 eV, for the work function of the Molybdenum back contact, would place the contact near the 0.8 eV case mentioned above. At this point, it was realized that a better fit to our actual data could be obtaine d if the band gap of CuGaSe 2 was increased to 1.65 eV (from the value of 1.6 eV used earlier). This new base simulation run was then labeled as basecgs1.65 and Figures 5.47 and 5.48 depict the band diagram and the I V characteristics, respectively. The spectral response curves, with and without a forward bias of 1 Volt, are shown in Figure 5.49. As

PAGE 178

165 can be easily seen from comparison of the two curves, there is a loss of current because of poor collection, in the forward bias. Poor transport properties that result into this behavior may be the reason why it is difficult to maintain the J SC while trying to lower the V OC value. Next, with the new Eg value of 1.65 eV, the back contact energy was progressively lowered, from 1.4 to 1.2, to 1.0, and lastly, to 0.8 eV. Figure 5.50 presents the I V results for these experiments. Once again, the 0.8 eV case offers the best fit to the results obtained with processed (Type I) samples. The next figure, Figure 5.51, shows the band bending that results from the lo wering of the back contact energy. Because the large band gap of ZnO is the cause of the high V OC s in the preceding simulations, in the next few runs, we decided to leave the ZnO, as well CdS out. The base run with this structure is now labeled as basec gs1.65 no cds or zno The I V characteristics for this run is shown in Figure 5.52, along with those for runs where defect levels were introduced in this structure. It is noteworthy that the efficiency for the base run is as high as 18.769%. It is also inte resting to note that high defect levels quickly destroy the J SC just like the case when both CdS and ZnO were present. Ideally, with the band gap of 1.65, and appropriate doping of the absorber, it is possible to raise the efficiency of a CuGaSe 2 structu re to above 20%. Figure 5.53 shows the simulated I V curves for such high efficiency devices. In summary, the AMPS simulations that involved variations in the back contact properties yielded a close match with the experimental results. In addition, these simulations also showed that, ideally, an efficiency of above 20% is possible with these solar cells.

PAGE 179

166 It should be mentioned here that, in some our processing experiments, we did try to change the back contact of the absorber material. The way we tried t o do this was not by replacing Molybdenum, but by depositing In, instead of Ga, as the first layer of the absorber material. The hope was that this would modify the actual back contact that the solar cell had. However, the performance of these runs was d iminished, and it was attributed to the mixing of In with other elements of the absorber (Cu, Ga, and Se), thereby forming a CuInSe 2 CuGaSe 2 mix compound. The efforts to change the back contact, hence, were unsuccessful. The next few pages contain fig ures 5.40 5.51, which relate to the above discussion, after which, the Conclusions of our research are presented.

PAGE 180

167 Figure 5.40. AMPS 1 Thermodyna,ic Equilibrium 1 Volt Forward Bias Basecgs Run: Ec Ec Ev Ef Ev

PAGE 181

168 Figure 5.41. AMPS 2

PAGE 182

169 Figure 5.42. AMPS 3

PAGE 183

170 Figure 5.43. AMPS 4

PAGE 184

171 Figure 5.44. AMPS 5

PAGE 185

172 Figure 5.45. AMPS 6

PAGE 186

173 Figure 5.46. AMPS 7 1 2 3 4 5 6 # 1 #2 # 3 # 4 # 5 # 6

PAGE 187

174 Figure 5.47. AMPS 8 Basecgs1.65 Run: Thermodynamic Equilibrium 1 Volt Forward Bias

PAGE 188

175 Figure 5.48. AMPS 9

PAGE 189

176 Figure 5.49. AMPS 10

PAGE 190

177 Figure 5.50. AMPS 11

PAGE 191

178 Figure 5.51. AMPS 12 Basecgs1.65 Basecgs1.65, with 0.8eV back contact energy

PAGE 192

179 Figure 5.52. AMPS 13

PAGE 193

180 Figure 5.53. AMPS 14

PAGE 194

181 CHAPTER 6. CONCLUSION The accomplishment of this research project was two fold: (i) The (modified) manufacturing friendly, sequential deposition process was used to establish a base CuGaSe 2 process, and was then optimized to gain substantial improvements in the performance (especially, the V OC ) of our CuGaSe 2 solar cells, (ii) The V OC improvement gained in this project is belittled by the high V OC value theoretically predicted for a CuGaSe 2 structure. This pointed to a performance ceiling for the material and t he process. Therefore, numerous physical, as well as simulation experiments were carried out, which helped improve our understanding of this limitation. Our EDS results (on a Type II sample) showed that the ratios of the three (Cu, Ga, Se) elements presen t in the absorber layer were close to 1:1:2, which meant that the material was close to being CuGaSe 2 at least in the top region of the film. XRD analysis (also on a Type II sample) showed that the structure of the analyzed CuGaSe 2 film was polycrystalli ne, with a preferred orientation along the [112] direction. The spectral response curves showed that the bandgap of our CuGaSe 2 material was around 1.63 to 1.64 eV. This indicated towards the formation of a material that was very close to the ideal CuGaSe 2 absorber material. Also, the response curves showed the presence of a very thin CdS layer, perhaps less than a 100 A 0 It is possible that this layer

PAGE 195

182 was not a true CdS layer, but was rather present as an intermediate phase, or mixture, between CdS and the adjoining ZnO top contact layers. This is one area where more systematic experimentation is needed. The major part of this study entailed the processing of CuGaSe 2 solar cells using the Type I and Type II recipes for the absorber formation. There ar e several differences between the two recipes, in terms of the processing and the results they accomplished. These are listed below, along with the best I V characteristics obtained for each of the recipes: (i) In Type I, all of the Ga was deposited up front, before Cu, whereas, for Type II, the Ga was split into two layers, one before, and another after, Cu. (ii) Typically, a Type II sample saw a higher temperature for a longer period of time. Because this made the substrate more vulnerable to the loss of III VI species, the process was adjusted so that much more Se was available during a Type II run. (iii) Variations in the thickness of the initial Ga layer changed the outcome of Type I samples, whereas they had little effect on Type II samples. (iv) Type II samples had imp roved V OC values, compared to Type I. The better V OC values for Type II samples were around 725 mV, with the best one at 735 mV. Annealing treatment increased this number to 775 mV. The highest V OC for a Type I sample was 699 mV, with most other high values between 650 and 699 mV. (v) The J SC (current density) performance of Type II samples was diminished, when compared to that of Type I samples. Typical values for Type I were between 13 and 15 mA/cm 2 with the best one at 15.2 mA/cm 2 (not in the same sam ple that showed the best V OC ). Typical values for Type II samples were between 10 and 12 mA/cm 2

PAGE 196

183 The above results indicate towards a basic structural difference between the Type I and Type II processing recipes. Such a difference was intended, in that T ype II was an attempt towards raising the V OC s by utilizing the Cu poor to Cu rich film conversion. Although this V OC improvement was accomplished, it was at the cost of the J SC keeping the overall efficiency nearly the same. Repeated efforts to decoup le the voltage and current behavior were unsuccessful. Such a decoupling is needed, in order to improve the overall performance. The shunting effect (as seen in the I V characteristics) was successfully reduced by optimizing the thickness of the absorber layer. This resulted in a better curve shape, and improved fill factor values, with the best one at 58%. This also resulted in the best efficiency value of 4.8%. As mentioned above, annealing, at a high temperature (200 0 C), improved the V OC s, while some what reducing the current performance. Neither annealing at a lower temperature, nor light soaking, showed any specific trend. Experiments with other layers in the solar cell structure shed some light on the complexity of the material. Efforts to replace CdS, as the heterojunction partner, had very limited success, once again re establishing CdS as the right choice. These experiments, however, improved our understanding about the CdS process itself. The best CuGaSe 2 results have been achieved, elsewhere by making variations in the CdS process. Although all the details of this work couldnt be known, a similar effort with our CdS was unsuccessful. Further research is needed, to focus on this aspect. Efforts were also made to intentionally form an n typ e layer at the top surface of the absorber, resulting in little success. This experience, along with the SCAPS

PAGE 197

184 simulation experiments, indicates towards the strong possibility of the junction being a true heterojunction. The two simulation techniques were helpful in separating the effects of various parameters on the device performance. SCAPS simulation produced a close match to our Type I (processed) devices. This needed the introduction of interface, as well as bulk defects in the solar cell structure. AMPS simulation, on the other hand, placed the blame on the back contact. (SCAPS produced no such results when the back contact properties were changed.) Because of the unavailability of an alternate back contact material (and process) in our laboratory actual (physical) experimentation with the back contact could not be carried out. Such experimentation is needed, to make use of the simulated results, and to better understand the source of the limitation on the V OC of the CuGaSe 2 /CdS solar cell struct ure.

PAGE 198

185 REFERENCES Abulfotuh, F., et. al., 25 th IEEE Photovoltaics Specialists Conference, p. 993 (1996) Albin, D.S., et al., IEEE Photovoltaics Specialists Conference, p. 562 (1990) Albin, D.S., et al., Materials Research Society Symposium Proceedings 228, p. 267 (1991) Bube, R., and Fahrenbruch, A., Fundamentals of Solar Cells, Academic Press, Inc. (1983) Bube, R., Book: Photovoltaic Materials, Imperial College Press (1998) Cahen, D., Noufi, R., Solar Cells 30, 53 (1991) Carl son, D., Wronski, C., Appl. Phys. Lett. 28, p. 671 (1976) Contreras, M.A., Egaas, B., Ramanathan, K., Hiltner, J., Swartzlander, A., Haason, F., Noufi, R., Prog. Photov. Res. Appl. 7, p. 311 (1999) Contreras, M.A., et al., Proceedings of the first world conference on photovoltaic energy conversion. Rec. IEEE Photovoltaics Specialists Conference, Vol. 24, p. 68 (1994) DAmico, J., Masters thesis, University of South Florida (1997) Gabor, A., Ph. D. dissertation, University of Colorado (1995) Goetzberge r, A., Hebling, C., Solar Energy Materials and Solar Cells 62, p.1 19 (2000) Green, M., et al. Prog. Photovolt. Res. Appl. 7, 31 (1999) Heske, C., et al., Appl. Phys. Lett. 74, 10, p.1451 (1999) Jayapalan, A., Sankaranarayanan, H., Lin, H., Rarayanaswa my, R., Ferekides, C.S., Morel, D, Proceedings of the 2 nd World PV Conference, Vienna (July 1998) A. Jayapalan, P. Panse, D. Morel, et al., Interface mechanisms in CuIn(Ga)Se 2 solar cells, 2 nd World Conference on Photovoltaic Solar Energy Conversion (19 98)

PAGE 199

186 Jasenek, A., Rau, U., Nadenau, V. and Schock, H., Journal of Applied Physics, Vol. 87, No. 1, p. 594 (2000) T. Kampschulte, J. Albert, U. Blieske, A. Bauknecht, M. Saad, S. Chichibu and M. Lux Steiner, 26 th IEEE Photovoltaic Specialists Conference ( 1997) Kim, S., et al., Solar Energy Materials and Solar Cells 62, p. 357 (2000) Klein, A., Jaegermann, W., Appl. Phys. Lett. 74, p. 2283 (1999) Mcveigh, J., Bertraw, D., Darmstadter, J., Palmer, K., Solar Energy Vol. 68, No. 3, p. 237 255 (2000) Nadena u, V., Hariskos, D. and Schock, H., Proceedings of the 14 th European Photovoltaic Solar Energy Conference, Barcelona, Spain, p. 1250 (1997) Nadenau, V., Hariskos, D. and Schock, H., J. of Appl. Phys., 85, 1, p. 534 (1999) Nadenau, V., Rau, U., Jasenek, A and Schock, H., Journal of Applied Physics, Vol. 87, No. 1, p. 584 (2000) Nakada, T., Appl. Phys. Lett. 74, 17, p. 2444 (1999) Natarajan, H., Masters Thesis, University of South Florida, (1997) Niemegeers, A., et al., Prog. Photovolt. Res. Appl. 6, p 407 (1998) Panse, P., Sankaranarayanan, H., Narayanaswamy, R., Shankaradas, M., Ying, Y., Ferekides, C., and Morel, D., IEEE Photovoltaics Specialists Conference (2000) Rau, U., et al., J. of Appl. Phys. 86, 1, p. 497 (1999) Scofield, J. H., et al., T hin Solid Films, 260, p. 26 (1995) Shankaradas, M., Ying, Y., Sankaranarayanan, H., Panse, P., Ferekides, C.S., Morel, D., IEEE Photovoltaics Specialists Conference (2000) Shell Renewable Energy Information Brochure, (1997) Sweet, W., article, the Elect ric Power Research Institute, (Jan. 2001) Sze, S. M., Book: Physics of Semiconductor Devices, Wiley Eastern Limited, (1981) Tsur, Y., Zemel, A., Solar Energy Vol. 68, No. 5, pp.379 392 (2000)

PAGE 200

187 Tuttle, J., et al., Proceedings of the 1996 Spring MRS Meetin g, San Francisco, CA, USA, p. 143 (1996) Ward, J.S., et al., 23 rd IEEE Photovoltaics Spec. Conf., Louisville, Kentucky, p. 650, (1993) Wei, S., Zunger, A., J. Appl. Phys., 78, 6, p. 3846 (1995) Wei, S., Zhang, S. and Zunger, A., Appl. Phys. Lett., Vol. 72, No.24, (1998) Wei, S., Zhang, S. and Zunger, J. Appl. Phys. 85, 7214 (1999) Zafar, S., Panse, Morel, D., et al., NREL/SNL Photovoltaics Program Review, (1996) Zhang, S., Wei, S., and Zunger, A., Phys. Rev. Lett. 78, 4059 (1997) Zhang, S., Wei, S., Zunger, A., and Katayama Yoshida, H., Phys. Rev. B, 57, p. 9642 (1998) Zunger, A., Zhang, S., and Wei, S., National Renewable Energy Laboratory Report No. CP 450 23581 (2002)

PAGE 201

188 APPENDICES

PAGE 202

189 APPENDIX A. EDS and XRD Results The EDS are XRD results for a CuGaSe 2 Type II sample are presented below. EDS Results for Sample # P131

PAGE 203

190 APPENDIX A (Continued) XRD Results for Sample # P131

PAGE 204

ABOUT THE AUTHOR Although born and brought up in India, Mr. Panse has a strong disbelief in Arranged Marriages He is married to his childhood friend Pradnya. Moreover, Mr. Panse is a confirmed (Russelian) agnostic and Reason (along with Murphys Laws, and the wife) is his guiding principle in l ife. He also holds highest the virtues of peace, freedom, democracy, and non violence. Mr. Panse grew up in the city of Thane, in Western India. He obtained his Bachelors in Physics from the University of Bombay (now Mumbai). Later, he came to the Unit ed States, to attend the University of South Florida, for his graduate study in Physics. He, then, went on to obtain his Ph.D. in Electrical Engineering. In addition to using it for his familys well being, Mr. Panse plans to use his knowledge and experie nce for the betterment of the community. Right now, he is not sure how. But he will figure it out, eventually!