Citation
Sputter Deposited Zinc-Magnesium Alloyed Transparent Conducting Oxides for Copper Indium Gallium Diselenide Solar Cells

Material Information

Title:
Sputter Deposited Zinc-Magnesium Alloyed Transparent Conducting Oxides for Copper Indium Gallium Diselenide Solar Cells
Creator:
Hicks, Albert B, III
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (134 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Chemical Engineering
Committee Chair:
ANDERSON,TIMOTHY J
Committee Co-Chair:
JIANG,PENG
Committee Members:
DAVIDSON,MARK R
BOSMAN,GIJSBERTUS
Graduation Date:
8/9/2014

Subjects

Subjects / Keywords:
Absorptivity ( jstor )
Conductivity ( jstor )
Doping ( jstor )
Electric current ( jstor )
Electrical resistivity ( jstor )
Electrons ( jstor )
Energy gaps ( jstor )
Figure of merit ( jstor )
Oxygen ( jstor )
Zinc ( jstor )
Chemical Engineering -- Dissertations, Academic -- UF
cigs -- mgzno -- photovoltaic -- sputtering -- tco -- thin-film -- zno
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Chemical Engineering thesis, Ph.D.

Notes

Abstract:
Zinc oxide alloyed with magnesium (MgZnO) has been investigated as a novel transparent conducting oxide (TCO) for Copper Indium Gallium Diselenide (CIGS) solar cells with improved transparency relative to ZnO. Quantum Efficiency (QE) of CIGS devices completed using ZnO and MgZnO TCOs revealed that the MgZnO device had a TCO band-edge cutoff of 340 nm, compared to 380 nm for the ZnO device. There is an additional 2% of the solar spectral energy available to CIGS using MgZnO. However, the realization of overall efficiency improvement of the device was limited by the CdS buffer layer, which resulted in a steady decrease in QE for wavelengths below 500 nm. Elements gallium and scandium were studied as alternative dopants to aluminum in the MgZnO system by reactive sputtering of metallic alloy targets on glass substrates. MgZnO:Ga showed comparable conductivity to that of ZnO:Al, 9 x 10-4 ohm-cm. MgZnO:Al and MgZnO:Sc were less conductive, on the order of 2 x 10-3 ohm-cm. Low power sputtering (13 to 25 watts) compared with high power (50 to 100 watts) revealed that the film quality improves as the power/deposition rate is decreased for films of the same thicknesses (600 to 800 nm). Transmission data reveals that high power films suffer transparency loss for wavelengths below 500 nm, suggesting a higher rate of near-valance-band-optically-active-defect occurrence relative to films sputtered with low power. XRD suggests improved crystal quality with lower sputtering power by an increased Scherer grain size. A novel figure of merit (FOM) technique is presented as an absolute comparison of TCO performance in terms of CIGS devices. A spectrum-energy integration is performed to quantify the TCO loss into a single optical parameter, which is then combined with the resistivity parameter for the FOM. The TCO FOM is modified to predict the loss of device output power in CIGS as a result of adjusting the TCO. A TCO of graded thickness between the metal contacts offers a reduction in TCO-associated loss in CIGS by approximately 15%. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2014.
Local:
Adviser: ANDERSON,TIMOTHY J.
Local:
Co-adviser: JIANG,PENG.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2015-02-28
Statement of Responsibility:
by Albert B Hicks.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Hicks, Albert B, III. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
2/28/2015
Classification:
LD1780 2014 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

1 SPUTTER DEPOSITED ZINC MAGNESIUM ALLOYED TRANSPARENT CONDUCTING OXIDES FOR COPPER INDIUM GALLIUM DISELENIDE SOLAR CELLS By BARRETT HICKS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 4

PAGE 2

2 © 201 4 Barrett Hicks

PAGE 3

3 I dedicate the work to my w ife, Yige Hu, who has loved me and supported me through my most difficult tim es .

PAGE 4

4 ACKNOWLEDGMENTS I acknowledge my thesis advisor, Prof. Tim Anderson, my committee members, Profs. Peng Jian g, Mark Davidson and Gis Bosman. I would like to acknowledge other faculty who have helped greatl y with advising my work: Prof. Paul Holloway. Colleagues of my research group have been very enjoyable to work with and get to know: Jo s e ph Revelli, David Wood, Ranga rajan Krishnan , Vaib hav Cha u d hari , Chris topher Muzzillo, Chris topher O onohue, Seo Y o u ng Kim, Zhi Li, Ga b ri e l Tong, Sam ua l Wilson and Hankook Kim . Colleagues from collaborating research groups have been a pleasure to work with: Yige Hu from Elec trical and Computer Engineering and Wei ran Cao from Material Science.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 LIST OF ABBREV IATIONS ................................ ................................ ................................ ........ 12 ABSTRACT ................................ ................................ ................................ ................................ ... 15 CHAPTER 1 BACKGROUND ................................ ................................ ................................ .................... 17 Motivation ................................ ................................ ................................ ............................... 17 CIGS Solar Cells ................................ ................................ ................................ ..................... 20 Performance Properties of ZnO Based TCOs ................................ ................................ ........ 21 Figure of Merit ................................ ................................ ................................ ................ 22 Conductivity ................................ ................................ ................................ .................... 24 Optical Transmis sion ................................ ................................ ................................ ....... 25 Film Crystal Quality ................................ ................................ ................................ ........ 28 Sputter Deposition ................................ ................................ ................................ .................. 29 Experimental Methods ................................ ................................ ................................ ............ 30 The Sput tering System ................................ ................................ ................................ .... 31 Materials and Handling ................................ ................................ ................................ ... 32 Characterization and Analysis ................................ ................................ ......................... 33 2 THEORETICAL TCO MODELS AND NANOSTRUCTURE ENHANCED PERFORMANCE ................................ ................................ ................................ ................... 45 TCO Performance Modeling ................................ ................................ ................................ .. 45 Medici ................................ ................................ ................................ .............................. 46 Analytical ................................ ................................ ................................ ........................ 48 Opto Electric Method ................................ ................................ ................................ ...... 50 Graded TCO ................................ ................................ ................................ .................... 54 Multiple Quantum W ells ................................ ................................ ................................ ........ 57 Density of States ................................ ................................ ................................ .............. 59 Predicted IR Enhancement ................................ ................................ .............................. 63 3 DC SPUTTERING: THERMAL AND NEGATIVE ION EFFECTS ................................ ... 79 Structure Zone Model Experiment ................................ ................................ ......................... 79 Experiment ................................ ................................ ................................ ...................... 81

PAGE 6

6 Oxygen and Discharge Characteristics ................................ ................................ ............ 81 Electrical Properties ................................ ................................ ................................ ......... 81 Crystal Characteristics ................................ ................................ ................................ ..... 82 Negative Ion Resputtering Experiment ................................ ................................ .................. 82 Electrical Properties ................................ ................................ ................................ ......... 84 XRD Analysis ................................ ................................ ................................ .................. 84 4 MGZNO:( AL, GA AND SC ) TCO BY OFFSET DC SPUTTERING ................................ .. 93 Experiment ................................ ................................ ................................ .............................. 94 Doping ................................ ................................ ................................ ............................. 94 Effects of Power ................................ ................................ ................................ .............. 95 Figure of Me rit Analysis ................................ ................................ ................................ . 96 Material Analysis Results ................................ ................................ ................................ ....... 99 Performance Merit ................................ ................................ ................................ ........... 99 Crystal Properties ................................ ................................ ................................ .......... 101 Discharg e Voltage Characteristics ................................ ................................ ................ 103 5 CIGS DEVICES WITH ZNO AND MGZNO ................................ ................................ ..... 117 Experiment ................................ ................................ ................................ ............................ 117 High Power MgZnO:Al on CIGS ................................ ................................ ......................... 118 Low Power MgZnO:Ga on CIGS ................................ ................................ ......................... 119 Discussion ................................ ................................ ................................ ............................. 120 6 CONCLUSIONS AND FUTURE WORK ................................ ................................ ........... 126 Conclusions ................................ ................................ ................................ ........................... 126 Future Work ................................ ................................ ................................ .......................... 126 LIST OF REFERENCES ................................ ................................ ................................ ............. 128 BIOGRAPH ICAL SKETCH ................................ ................................ ................................ ....... 134

PAGE 7

7 LIST OF TABLES Table page 2 1 RF Sputtered ZnO:Al electric properties ................................ ................................ ........... 65

PAGE 8

8 LIST OF FIGURES Figure page 1 1 The CIGS device structure. ................................ ................................ ............................... 37 1 2 The TCO diagram illustrating the length spanned between the electric contacts relative to the device thickness. ................................ ................................ ......................... 37 1 3 Trend of transmission and sheet resistance. ................................ ................................ ...... 38 1 4 Sample ZnO:Al transmission spectrum. ................................ ................................ ........... 38 1 5 The quarter wave length interference effect. ................................ ................................ .... 39 1 6 The target and electric field diagram. ................................ ................................ ............... 39 1 7 The sputtering system diagram. ................................ ................................ ........................ 40 1 8 The sputtering chamber configuration . ................................ ................................ ............. 40 1 9 The sputtering gun diagram. ................................ ................................ ............................. 41 1 10 The heated substrate holder diagram. ................................ ................................ ............... 41 1 11 The m ass balance of the sputtering gas. ................................ ................................ ............ 42 1 12 The TCO completion process of the CIGS device. ................................ ........................... 42 1 13 The post metallization device finishing. ................................ ................................ ........... 43 1 14 The four point probe rational. ................................ ................................ ........................... 43 1 15 ray diffraction ................................ ................................ .............. 44 1 16 The characteristics of the JV plot in relationship to the extracted performance parameters. ................................ ................................ ................................ ......................... 44 2 1 The TCO current collection structure in CIGS. ................................ ................................ 66 2 2 The material device structure for the 1D and 2D Medici simulations. ............................. 66 2 3 The Medici CIGS 1D comparison . ................................ ................................ ................... 67 2 4 The 2D CIGS efficiencies at multiple TCO thicknesses and sun concentrations. ............ 67 2 5 Curr ent flow line approximation.. ................................ ................................ ..................... 68 2 6 The constant current flux (non diode) TCO loss model. ................................ .................. 68

PAGE 9

9 2 7 The comparison of the Medici and the analytical calculated efficiencies as a function of the number of suns and sheet resistance. ................................ ................................ ....... 69 2 8 The energy loss at CIGS stages from solar input to electric output. ................................ . 69 2 9 The AM1.5 spectrum for weighted and unweighted by the CIGS thermalization. .......... 70 2 10 Quantum efficiency weighting.. ................................ ................................ ........................ 70 2 11 A sample TCO. ................................ ................................ ................................ ................. 71 2 12 Current loss trend. ................................ ................................ ................................ ............. 71 2 13 The explicit MQW CIGS concentrator. ................................ ................................ ............ 72 2 14 The inefficiency of the traditional flat TCO structure. ................................ ..................... 72 2 15 The graded TCO structure. ................................ ................................ ............................... 73 2 16 The relative performance of the flat and graded TCOs. ................................ ................... 73 2 17 The mobility of state of art heavily doped ZnO compared with ionized impurity scattering limited mobility trends. ................................ ................................ ..................... 74 2 18 The quantum well electronic band alignment of a 2DEG enhanced TCO. ...................... 74 2 19 The superlattice structure of ZnO and MgZnO. ................................ ................................ 75 2 20 The energy levels formed in the quantum well with deviation of the density of states from a continuum. ................................ ................................ ................................ .............. 75 2 21 The alpha and beta plots for obtai ning quantum well energy levels ................................ . 76 2 22 The calculated energy levels plotted as a density of state number (n) versus the energy. ................................ ................................ ................................ ................................ 76 2 23 Carrier density levels ................................ ................................ ................................ ........ 77 2 24 Doping profile design. ................................ ................................ ................................ ...... 77 2 25 Ionic scattering radius overlap with quantum wells.. ................................ ....................... 78 2 26 Modeled IR absorption.. ................................ ................................ ................................ ... 78 3 1 Structured zone model of the crystal quality as a function of temperature and pressure. ................................ ................................ ................................ ............................. 87 3 2 Illustration of the ener gy dissipation and surface movement of hot molecules in a sputtered flux. ................................ ................................ ................................ .................... 87

PAGE 10

10 3 3 The oxygen voltage discharge characteristics of the Zn:Al target during sputtering. ...... 88 3 4 The electrical resistivities of the samples as a function of substrate temperature. ........... 88 3 5 The theta 2 theta scan of the 100 C sample. ................................ ................................ ..... 89 3 6 The XRD extracted parameters. ................................ ................................ ....................... 89 3 7 The sputtering mechanism resulting in the formation of negative oxygen ions. .............. 90 3 8 The direct and offset substrate configurations relative to the sputtering target surface. ................................ ................................ ................................ ............................... 90 3 9 The electrical resistivities of the direct and the offset oriented substrates. ...................... 91 3 10 The theta 2 theta scans of the direct and offset samples at 100 C. ................................ ... 91 3 11 The XRD extracted parameters of the direct and offs et films. ................................ ......... 92 3 12 Crossectional SEM. ................................ ................................ ................................ ........... 92 4 1 The Air Mass 1.5 spectrum with the CIGS sensitive regions. ................................ ........ 105 4 2 The Sellmeier fitting of MgZnO dispersion relationship from MgZnO data and ZnO:Al bandgap shift. ................................ ................................ ................................ ..... 105 4 3 Optical transmission d ata. ................................ ................................ ............................... 106 4 4 Example of fitting optical transmission data with the reflection calculation model. ...... 106 4 5 Example of a linear extraction of the optical band gap from absorption data. ............... 107 4 7 The energy absorption coefficient as a function of the oxygen partial pressure. ........... 108 4 8 The energy absorption coefficient plotted versus the oxygen partial pressure. .............. 108 4 9 The figure of merit plotted versus the oxygen partial pressure. ................................ ..... 109 4 10 The figure of merit plotted versus the electrical resistivity. ................................ ........... 109 4 11 The figure of merit plotted versus the energy absorptio n coefficient. ............................ 110 4 12 The optical band gap potted versus the oxygen partial pressure. ................................ ... 110 4 13 The electrical resistivity plotted versus the optical band gap. ................................ ........ 111 4 14 The energy absorption coefficient plotted versus the optical band gap. ......................... 111 4 15 The figure of merit plotted versus the optical band gap. ................................ ................ 112

PAGE 11

11 4 16 The powder XRD scans of MgZnO:Al at different oxygen partial pressures. ............... 112 4 17 The grain size plotted versus the oxygen partial pressure. ................................ ............. 113 4 18 The c axis length plotted versus the oxygen partial pressure. ................................ ........ 113 4 19 The optical band gap plotted versus the c axis length. ................................ ................... 114 4 20 The electrical resistivity potted versus the c axis length. ................................ ............... 114 4 21 The figure of merit plotted versus the c axis length. ................................ ...................... 115 4 22 The DC voltage discharge characteristics versus the oxygen partial pressure. .............. 115 5 1 The device finishing diagram and label of individual cells. ................................ ........... 122 5 2 The IV data for ZnO:Al and MgZnO:Al averaged over four experiments. .................... 122 5 3 The QE of the ZnO:Al and MgZnO:Al devices averaged over four experiments.. ........ 123 5 4 The UV region of the QE data.. ................................ ................................ ...................... 123 5 5 The JV data and extracted parameters for ZnO:Al and MgZnO:Ga . .............................. 124 5 6 The QE data for ZnO:Al and MgZnO:Ga. ................................ ................................ ...... 124 5 7 The UV regions of the QE data averaged . ................................ ................................ ...... 125

PAGE 12

12 LIST OF ABBREVIATIONS 2DEG 2 Dimensional Electron Gas ALD Atomic Layer Decomposition AM1.5 Air Mass Index 1.5 C D Cadmium C D O Cadmium Oxide C D S Cadmium Sulfide CIGS Copper Indium Gallium Diselenide CO 2 Carbon Dioxide C U Copper CVD Chemical Vapor Deposition DC Direct Current EFF Efficiency E V Electron Volt FOM Figure of Merit FF Fill Factor FWHM Full Width Half Maximum G A Gallium IR Infrared IN 2 O3 Indium Oxide IISE Ion Induced Secondary Electron JSC Short Circuit Current JV Current Voltage Curve M G F 2 Magnesium Fluoride MFC Mass Flow Controller

PAGE 13

13 M G Z N O Magnesium Zinc Oxide M G Z N O:A L Aluminum Doped Magnesium Zinc Oxide M G Z N O:G A Gallium Doped Magnesium Zinc Oxide M G Z N O:S C Scandium Doped Magnesium Zinc Oxide M O Molybdenum MQW Multiple Quantum Well NIR Negative Ion Resputtering NREL National Renewable Energy Laboratory PLD Pulse Laser Deposition PV Photovoltaic QE Quantum Efficiency RF Radio Frequency RT Room Temperature SCCM Standard Cubic Centimeters per Minute S N Tin SR Spectral Response SZM Structural Zone Model TCO Transparent Conducting Oxide UHV Ultra High Vacuum VOC Open Circuit Voltage ZNO Zinc Oxide Z N O:A L Aluminum Doped Zinc Oxide Z N S Zinc Sulfide X ray Angle Light Wavelength

PAGE 14

14 Carrier Mobility Ohms

PAGE 15

15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SPUTTER DEPOSITED ZINC MAGNESIUM ALLOYED TRANSPARENT CONDUCTING OXIDES FOR COPPER INDIUM GALLIUM DISELENIDE SOLAR CELLS By Barrett Hicks A u g u s t 2014 Chair: Tim Anderson Major: Chemical Engineering Zinc oxide alloyed with magnesium (MgZnO) has been investigated as a novel transparent conducting oxide (TCO) for Copper Indium Gallium Diselenide (CIGS) solar cells with improved transparency relative to ZnO. Quantum Efficiency (QE) of CIGS devices completed using ZnO and MgZnO TCOs revealed that the MgZnO device had a TCO band edge cutoff of 340 nm, compared to 380 nm for the ZnO device . There is an additional 2% of the s olar spectral energy available to CIGS using MgZnO. However, the realization of overall efficiency improvement of the device was limited by the CdS buffer layer , which resulted in a steady decrease in QE for wavelengths below 500 nm . Elements gallium and s candium were studied as alternative dopants to aluminum in the MgZnO system by reactive sputtering of metallic alloy targets on glass substrates . MgZnO:Ga showed comparable conductivity to that of ZnO:Al, 9 x 10 4 cm. MgZnO:Al and MgZnO:Sc were less cond uctive, on the order of 2 x 10 3 cm . Low power sputtering (13 to 25 w atts) compared with high power (50 to 100 watts) revealed that the film quality improves as the power/deposition rate is decreased for films of the same thicknesses (600 to 800 nm) . Tra nsmission data reveals that high power films suffer transparency loss for wavelengths below

PAGE 16

16 500 nm, suggesting a higher rate of near valan ce band optically active defect occurrence relative to films sputtered with low power . XRD suggests improved crystal q uality with lower sputtering power by an increased Scherer grain size. A novel figure of merit (FOM) technique is presented as an absolute comparison of TCO performance in terms of CIGS devices . A spectrum energy integration is performed to quantify the TCO loss into a single optical parameter, which is then combined with the resistivity parameter for the FOM . The TCO FOM is modified to predict the loss of device output power in CIGS as a result of adjusting the TCO. A TCO of graded thickness between the metal contacts offers a reduction in TCO associated loss in CIGS by approximately 15%.

PAGE 17

17 CHAPTER 1 BACKGROUND Motivation In modern times there is a growing desire to get off the dependency on fossil fuels as the m ain energy source of the world . T he high energy density and convenient energy release during combustion of fuels as well as s energy needs . But there is a growing motivation to move to a cleaner more renewable source of energy. An important reason for scaling back fossil fuel use is the growth of man related carbon dioxide in the atmosphere [1] and the impact that it may cause on the global climate . Even if the impact of climate change does not cause negative consequences to the earth, as models suggest [ 2] [3] , there are more basic reasons to phase out their use . Many years ago Tesla believed that the future should not be too dependent on ener gy that comes out of the ground [4] : The hard fact is that unless new resources are opened up, energy derived from fuel will remain our chief relia nce. The thermodynamic process is wasteful and barbarous, especially when burning coal, the mining of which, despite of modern improvements, still involves untold hardships and dangers to the unfortunates who are condemned to toil deep in the bowels of the earth. Oil and natural gas are immensely superior in this and other respects and their use is rapidly extending. It is quite evident, though, that this squandering cannot go on indefinitely, for geological investigations prove our fuel stores to be limite d. So great has been the drain on them of late years that the specter of exhaustion is looming up threateningly in the distance, and everywhere the minds of engineers and inventors are bent upon increasing the efficiency of known methods and discovering ne w sources of power. Nicola Tesla, 1931 Up to the present there has not yet been a technology that is market ready to fully replace traditional energy sources. While nuclear has been available, there have been social impediments to its adoption. T he energy statistics published by the U.S. Energy Information Agency suggest 2040. The production of oil in the U.S. has reached the previous peak of 9.6 million ba rrels per

PAGE 18

18 day [5] . Natural gas is expected to become an even greater player in all the energy markets, growing by 56% overall, surpassing coal and nuclear electric generation and exports from the U.S. amounting to 8 trillion cubic feet per year 36 years from now . The repor t statistics anticipated a gradual reduction of household energy consumption due to efficiency improvements such as LED based interior lighting, it did not suggest that there would be a new renewable energy source . A technology anticipated as a possible fossil fuel replacement is photovoltaics. Unlike more traditional renewable and/or non polluting energy resources such as hydroelectric, biomass, geothermal, wind and oceanic, which may be limited by scale and/or region, conversio n of the geographies. The sun gives a sustained 10 17 watts to the earth on a continual basis , which is also 10,000 times the total global energy usage rate. Only 0. 1% of that sun light needs to be gathered at 10% collection efficiency to meet the current energy needs. In addition to being a very large source, it is available everywhere. Furthermore, photovoltaic modules are the most versatile alternative energy sourc e because the panels do not have moving parts and intense labor to operate and maintain. They can be used from large scale plants down to very small sca les such as roof top generation. Historically t he complexity of photovoltaic technology has prevented i t from being able to become an affordable substitute for terrestrial power . The modern semiconductor technology for the task was first implemented in 1954 [6] for the primary application motivation of space power but too expensive to consider for terrestrial power generati There have been many advances, but they have not yet provided a sufficiently efficient, reliable, and

PAGE 19

19 inexpensive solar alternative to fossil fuels. However, the past 40 years of solar research has made tremendous progress towards bringing the price per watt to a level that gives real economic incentive for businesses and residences to replace conventional electric power with solar . Rooftop ins tallations are already having impacts in areas that contain both high electricity prices and large sun supplies. In 2013, California doubled the rate of rooftop installation with the total installed capacity going from 1 to 2 billion watts [7] . At this point, there are only two barriers to full adoption. Many re searchers are working on storage advances to provide 24 hour supply from solar sourced energy. However, a dvancing solar technology to reduce the price is the most important factor at the present time . The recent explosion of rooftop solar installations in sunny places can be due to a large drop in the price of silicon panels from 2010 to 2013. The reason for the price drop was not due so much to advances in scientific knowledge and manufacturing technology, but more so a glut of inexpensive panels as the pr ice of silicon dropped, resulting in the price per watt being cut from $1 to $0.5. The p rice de c rease was attributed to a fall in the price of polycrystalline silicon [8] . The problem with silicon is the amount of high purity semiconductor material needed to make a module. Since silicon is an indirect semiconductor, a photon will pass by thousands of atoms in the lattice before impacting the rig ht phonon to result in the creation of an electron hole pair. Therefore the device thickness must be high, between 0.1 and 1 mm. The amount of material will keep the manufacturing cost of the panel limited by the market price of silicon, which could become inflated if the demand is extremely high. A technology alternative to silicon is solar based on thin film absorbers . It consists of direct bandgap semiconductors, where the light is completely adsorbed by a layer approximately 1/1000 th of the thickness of silicon solar cells . A major pla yer in the thin film industry is First

PAGE 20

20 Solar, a panel maker using the technology cadmium telluride . The panels are a few percent less efficient than polycrystalline silicon, but the price per watt recently hit a new milesto ne of $ 0.46 per watt, surpassing the cost of overseas panels [9] . Because the material costs are very low, the anticipation is tha t the panel price will continue to decrease year by year. There are concerns about the cadmium being widespread if cadmium telluride based solar cells become mass produced for terrestrial power. First Solar up to 2013 was challenged to capture the rooftop markets [10] , although it is not clear if it was due entirely to cadmium . However, one alternative thin film technology to CdTe is copper indium gallium diselenide (CIGS). It is not only a cadmium free absorber, but also holds the record thin film efficiency [11] . CIGS Solar Cells The traditional CIGS device structure, shown in Figure 1 1 , has a thin buffer layer of CdS, which is a very small portion of the overall device material in comparison to a CdTe device. Furthermore progress is being made towards a completely cadmium free CI GS cell [12] . CIGS is an extremely stable solar compound with an optical absorption coefficient high enough to allow the device thickness to be less than 2 um. The TCO work presented here is done with CIGS applications as the primary motivati on. The rest of the section will introduce the CIGS physical structure. The first step is the device construction is the substrate , which serves as a structural platform for the razor thin layers of semiconductor to rest on. The requirements are a smooth s urface and chemical stability . For minimizing material cost in a CIGS manufacturing model, low cost substrates such as glass and steel foil are often used. Next is the molybdenum layer, which serves as the back electrical contact. It also is important for promoting a firm bond for the remaining layers . The next layer deposited is the CIGS absorber . CIGS has a bandgap that can be

PAGE 21

21 tuned by altering the indium and gallium ratio , allowing the bandgap to change from 1.1 eV to about 1.4 eV. CIGS is fabricated so that there is a gradient of higher gallium towards the back side of the absorber to promote the collection of photogenerated electrons deep in the structure . The next layer is the CdS buffer layer, which is a thin p type semiconductor that sets up the pn junction with CIGS. However, a secondary junction is formed where the exchange of copper and cadmium ions creates a thin layer of electrically inverted (n type) CIGS . The role of the inversion layer is now well understood, but the obse rvation is that a thin layer is beneficial while a thick layer results in interface recombination . The thickness is determined by the heat that the device is exposed to once the CdS layer is deposited. In processing the layers after CdS the device temperat ure should not be in excess of 100 C . The next layer is the ZnO resistive electrical buffer layer which serves as an electrical barrier against processing defects that could otherwise allow the top contact to form a shunt pathway with the back contact . Th e top electric contact , which is the focus of the current work , is the ZnO:Al transparent conducting oxide (TCO) . It is a link in the electrical circuit between the top of the device and the metalized grid . But unlike the grid, which is allowed to be opaqu e because it covers a small portion of the device, the TCO must be transparent and c overing all of the device area. The next section will discuss the requirements of the TCO for thin film devices and why ZnO is suitable for the roll. Performance Properties of ZnO Based TCOs Progress and development in CIGS device efficiency requires minimizing all areas of optical loss to maximize the optical energy that goes towards photogeneration. The focus of the present work is energy that is lost is in t he transparent conducting oxide (TCO) . Theoretical and experimental methods are presented for reducing the loss . A TCO is necessary for thin film solar cells because the thin n side layers semiconductor do not provide a continuous pathway for t he

PAGE 22

22 photogenerated electron s to be collected at the metal grids, which are on the order of 2mm separation , as illustrated Figure 1 2 . Traditional silicon devices do not require the TCO because the thick layers are very conductive . It is important for the TCO to perform well in both optical transmission and electrical conduction. A ll light that reaches the active region of the device has to pass through the TCO and all the photogenerated current uses the TCO as an electrical transport pathway . In the following section the TCO design c onsiderations and performance limiting parameters are discussed starting with the figure of merit. Figure of Merit The nature of the TCO performance is for the electrical conductivity performance to be inversely related to the optical performance. When a TCO film such as ZnO:Al is deposited, there are two electrical properties that describe the performance: the bulk electrical conductivity, and the optical adsorption coefficient, ( ), which varies depending on the wavelength of light, . The absolute opti cal adsorption, A and absolute sheet resistance, Rs, are determined by multiplication of the bulk coefficients with the film thickness: (1 1) (1 2) Both A and Rs are desired to be as low as possible, but they are inversely rel ated to each other as a function of the thickness , illustrated Figure 1 3 . The optimal thickness is chosen based on the electrical performance needs. If the electrical load is light, the TCO can be made very thin so there is minimal loss due to optical ads orption . CIGS has a 1 sun CIGS load of 30 mA/cm 2 works well if the sheet resistance is in the range of 50 to 100 /sq [Chapter 5 , present work] , which is on the high side of most sheet resistance requirements for TCOs. For the heavier electrical loads Grav est suggests TCOs should be in the range of 1 to 100 W/sq for covering a

PAGE 23

23 wide range of applications in his 1993 review paper [13] . The lower sheet resistances would be required if a CIGS device were used under a concentrated sunlight intensity. Coutts et al once greatly surpassed the world record efficiency and achieved 21.5% efficiency by simply using 15x sunlight [14] . At such solar levels the efficiency of a CIGS device would lose about 3% efficiency due to series resistance with a TCO thickness optimized for 1 sun cond itions [Chapter 2, present work] . The TCO bulk optical and elect rical properties can vary over many orders of magnitude depending on the chemical composition the choice of dopant and the purity and crystal quality of the physical film . The best quality TCO s have both high conductivities and low optical absorptions. A single term that expresses the performance ratio . There are different ways to define the figure of merit depending on if it characterizes film or bulk properties . One of the early attempts to define a figure of merit was done by Haacke in 1978, as a function of the film properties [15] : (1 3) Where TCO is the ratio of merit of the TCO, T is the absolute transmission, x is a customizable parameter and R s is the sheet resistance. The adva ntage of such a definition is it includes the film thickness to rate the absolute performance of the deposited TCO. The parameter x can be changed depending on how important transmission is considered to be relative to conductivity. Haacke suggested using x = 10. The disadvantage is the complexity and ambiguity . This method also fails t o capture the spectral differences in absorption of the solar spectrum, a critical factor in the overall device performance. A simpler figure of merit considers only the bulk properties . More recently Gordon suggested the figure of merit should simply be t he ratio of the bulk properties [16] :

PAGE 24

24 (1 4) Where is the bulk electrical conductivity and is the bulk absorption coefficient. For convenience with the absolute film properties it can be expressed in terms of the sheet resistance, the absolute transmission (T) and reflectance (R). This method also fails to capture spectral differences in absorp tion of the solar spectrum. However, two modified versions of the model are presented in the second and fourth chapters, respectively, that account for the spectral effect. Conductivity TCOs are a special class of compound semiconductors. They have full sp ectrum transparency through wide bandgaps, typically 3.0 eV and higher . The oxides are typically n type due to a self compensating mechanism of interstitials and oxygen vacancies . Indium Tin Oxide (ITO) is the best overall performing TCO material. The low cost abundant element is ZnO . Research efforts are focused on bringing the reliability and performance parameters of ZnO up to that of ITO to be a suitable replacement. The conductivity is a result of mobile elect rons in the conduction band of the TCO . It is calculated by the product of the fundamental charge ( q ), the electron density ( ) and the electron mobility ( ): (1 5) A conductive TCO must have a reasonable carrier density and a reasonable mobility in order to meet the minimum requiremen ts without the CIGS film being too thick. Carrier density in ZnO is obtained by dopants such as aluminum. Heavy doping is often used to obtain high conductivity. A typical ZnO:Al property of typical conductivity of 4 x 10 4 cm [17] . However, ZnO:Al is more often on the order of 1 x 10 3 cm when deposited on CIGS substrates and the thickness is increased to 1 um to compensate [18] . However there are negative impacts of

PAGE 25

25 excessive doping on both the mobility and optical absorption loss es, which a re discussed in later sections below. Optical Transmission The first step of analyzing a TCO optical performance is to track where the incident photons go. Most should be transmitted through towards the absorber layer of the device. The once that are not transmitted are lost through either absorption and reflectio n. The optical intensity balance on the light going in and going out of the TCO is expressed as: In order to maximize transmission both the reflectance and absorption must be mini mized. The parameters are all functions of the wavelength of the incident light. There are different reasons for the different types of losses, which are explained below. For the spectrum of electromagnetic radiation important for CIGS, the reflective loss is due to Fresnel reflection is the primary loss to consider. The reflection occurring due to a difference in the index of refraction of two materials of an interface light is crossing. The amount of reflection is determined by the Fresnel equations: (1 6) (1 7) (1 8) Where Rs and Rp are the reflectances of s polarized and p polarized light, respective ly. As the difference of index of refractions increases the amount of Fresnel reflection increase approximately by the square of the difference. Minimizing loss due to reflection should avoid layers that have large differences in index of refraction. In CI GS there is consideration of the TCO air interface. The index of refraction of ZnO is approximately 1.9 at 580 nm [19] , when the

PAGE 26

26 index of refraction of air is 1.0. A MgF2 (n ~1.4) coating added to maximize the performance of high eff iciency cells [11] . Fresnel reflection loss in solar cells can be nearly eliminated through optical engineering as surface texturing methods such as zinc oxide nanocrystals provide a textured surface so that light that is reflected off the surface is bounced towards a nearby feature to enter a second time [20] . O ptical engineering solutions can theoretically suppress reflection loss to zero. But adsorption loss, however, cannot be suppressed because it is linked to the TCO that give electrical conductivity. The origin of optical adsorption in the TCO is a result of the electro magnetic nature of photon electron interactions. A sample transmission spectrum of sputtered ZnO:Al is shown in F igure 1 4 . There are two regions where loss is associated with the properties of the TCO . The first is the sharp fall in transmis sion, starting around 400 nm , that is associated with the bandgap of the material. ZnO is a direct bandgap semiconductor , indicating that there is a high probability of optical adsorption of any photon that is above the bandgap. The adsorption coefficient is around 10 5 cm 1 so tha t very little UV light survives through films thicker than 100 nm . The bandgap is a property of the semiconductor and one of the factors that goes into place when making the material choice. The cutoff is defined by the point where the photon energy is equal to the bandgap : (1 9) The expression is useful for a graphical extraction of the optical adsorption. For heavily doped TCO compounds the observed bandgap will shift upwards due to a filling of the conduction band electronic states in the well known trend called the Moss Bernstein effect, where the trend can be predicted by a Fermi gas model:

PAGE 27

27 (1 10) Where meff is the electron (n type TCO) carrier effective mass and N is the carrier concentration. The doping level is a way to adjust the actual bandgap, although a range of 0.2 eV change is accomplished by two or three orders of magn itude of doping difference . There are other consequences to having excessive carriers doped. The second feature is in the long wavelengths with a decreasin g transmission. It i s due to free electron carrier absorption . In the Figure there is a gradual decre ase of the adsorption curve going into the deeper spectrum. The effect is an adsorption of photons by free electron carriers . The adsorption coefficient of free carriers excited from one intra band state to another is well known: (1 11) Where is the wavelength of the adsorbed photon and o is the permittivity of free space. It is proportional to N, the electron carrier density. The more densely doped the TCO is, the greater the absorption loss per thickness. The loss is also inversely prop ortional to mobility. For two films with equal carrier concentrations, the film with the higher mobility will not only have a higher electrical conductivity but a lower free carrier adsorption coefficient as well. A third featur e to note is the series of optical interference fringes. Light impinging on the TCO air interface will pass through the thickness of the TCO. Some of the light will bounce off the TCO glass interface and create quarter wavelength interference , as illustrat ed in Figure 1 5 . For film thickness t, the spacing between two peaks and the wavelength of the two peaks gives the film thickness by the following equation: (1 12)

PAGE 28

28 Where 1 is the longer of the two wavelengths at which the peaks occu r. The CIGS cutting off at 1.2 eV loses the longer wave IR spectrum. However, next generation solar cells that attempt to collect the long wave radiation will have to advance in TCOs that have minimal free scattering loss. Film Crystal Quality The most inf luential parameter on the figure of merit is the electron mobility. As shown in the equations above, a higher mobility correlates with a higher electrical conductivity and a lower rate of free carrier absorption . However, achieving high mobility films is a chal lenging. The mobility is determined mainly by how often an electron carrier scatters: (1 13) is the average scattering time of an electron. Improving mobility is accomplished when scattering mechanisms are reduced. There are man y scattering mechanisms in semiconductors things that the electron can exchange momentum with, including defects and the lattice. For scattering mechanisms A, B and C there are corresponding mobilities A , B and C . The overall mobility is determined by the well (1 14) Mathematically the lowest mobility will equate to the overall mobility as the dominant scattering mechanism will make up a large percentage of the overall scattering. The most dominant tw o related to TCOs are scattering off grain boundaries and ionized impurity scatterings. In order for a high figure of merit, both scattering mechanisms must be minimized.

PAGE 29

29 Sputter Deposition ZnO can be deposited on most types of semiconductor growth syste ms including chemical vapor deposition, evaporation, spray pyrolysis pulse laser deposition and sputtering. Sputtering is a physical deposition method in which a plasma is generated and ions extracted to kinetically sputter atoms from a target material, th us forming a flux of atoms which re condense to form a film on the substrate . There are a number of advantages of sputtering ZnO. The rate of deposition can be precisely controlled by the sputtering power. The deposition rate can also adjusted over a very wide range, covering a few orders of magnitude . ZnO can sputtered either reactively, using a zinc metal target in an oxygen sputtering environment, or non reactively, using a zinc oxide target. In the present study direct current (DC) reactive sputtering i s used. The advantage is to fine tune the oxidation of the growing film. For ZnO reactive sputtering the environment is argon and oxygen and a metallic target provides the source of Zn . The energy source of the plasma is a negat ive electric bias on the target . A diagram of the sputtering target and fields is given in Figure 1 6 . The field draws in argon ions and repels electrons . As the electrons accelerate away from the electric field, they bombard neutral argon atoms to produces the following ionizatio n reaction: (1 15) The result is the creation of a new argon ion and a new electron. Both the new electron and old electron will gain new kinetic energy in the field and cause other ionization events. The rate of ionization events for electrons is greatly increased by the addition of a magnetron, or a magnetic field originating below the t arget. The electronic and the magnetic fields present result in the Lorentz force acting on the particles: (1 16)

PAGE 30

30 Where is the net force, is the electric field line, is the magnetic field line. The force on the particle is a function of the position relative to the biased target and the magnets behind the target. The doughnut shaped arrangement of the magnets causes the field lines to run from the outer ring and curve 180 and converge to t he center magnet. The electric field goes from the entire target surface outwards towards the cathode shield and the chamber walls. The result is that electrons follow a spiraling trajectory going either to or from the target. The direction depends on the charge: positive particles spiral towards the target and negative particles away . The argon ion will be immediately accelerated towards the biased target to impact the surface. While its path is also a spiral, the rotational velocity and radius are reduced due to its mass. The result of the collision is to transfer kinetic energy to surface atoms to the point that the chemical binding bonds are broken and the target particles are ejected. A result of the magnetic force on the argon ions is for the target er osion pattern to follow the magnetic field lines. Without the magnetron the plasma would be driven by electric fields alone, and would spread out over the chamber and not be efficient at sputtering. Electrons have a very long path length in the vicinity of the target. The plasma is confined close to the target surface because the electron impact events are guided by the magnetic field. Experimental Methods Experiments were performed on a multiple sourced sputtering system described below. New compounds of MgZnO as TCOs for CIGS were sputter deposited onto glass and on CIGS. T he glass TCO samples were used to probe the system dynamics without the complications of full device structure and provided useful information on the sputtering process and how it influ enced the crystal growth properties. Techniques including X ray diffraction, optical

PAGE 31

31 transmission and 4 point probe were the primary methods for gathering data. The use of new dopants revealed how the MgZnO compound c an be most effectively doped and deposi ted under direct magnetron sputtering. The MgZnO compounds were later deposited on CIGS devices that were partially completed up to the CdS level. The performance of the devices was measured under a simulated (AM1.5) solar light and also assessed for quant um efficiency. The Sputtering System The sputtering system is an in house built unit, which is commonly known by the name Rusty . It is a bell jar chamber design . For the TCO project there were extensive modifications including switching to a cryopump/loa dlock from a turbopump with a custom float ing heated substrate mount . The system schematic is shown in Figure 1 7 . The high vacuum was powered by a Cryotorr 8 model capable of reaching vacuum levels of 10 7 torr. It operates continually and is isolated fro m the chamber by a pneumatic gate valve to allow samples to be unloaded and loaded. A mechanical hoist raised and lowered the chamber to allow introduction of samples . The atmospheric pressure was mechanically pumped to a crossover pressure of about 300 mt orr by a Welch Model 1397 rotary va ne pump. The system contained four independently functioning magnetron sputtering sources with one sized for sized for . The layout of the guns is shown in F igure 1 8 . T he design is a circular gun with an outer and inner ring of magnets, internal water cooling, a dielectric insulating barrier and a metallic anode shield on the outer perimeter. The gun schematics are shown in Figure 1 9 . The guns are able to run in both DC and RF sputtering modes depending on the type of power supply that is connected. The re are four power supplies that allow the sources to be run independently . Each source can be powered by any of the DC or RF supplies. The substrate holder is illustrated in Figure 1 10 . It is a floating holder that is mounted on a rotating arm as illustrated in Figure 1 8 . The arm rotates over all four sputtering sources . There

PAGE 32

32 are two identical substrate holders, one on each end of the arm. It consi block onto which the substrates would be held by steel clamps. There was a mounted thermal couple and a halogen lamp to monitor and control the temperature. For aligning the sample in correct position, the heater lamp would cast a sharp shadow onto the sputtering gun. The sputtering gas pressure is driven by an inflow and outflow of sputtering gases a s shown in the mass flow diagram Figure 1 11 . There are two mass flow controllers (MCFs), one for argon and one for oxygen, which deliver the gas to the chamber while the cryopump removes the gas. The sputtering gas pressure is determined by the steady state conditions. In order to keep the steady state pressure at 1 m t orr, there is an adjustable shutter valve in front of the cryopum p that restricts the pumping speed approximately 50:1. Materials and Handling For the TCO sputtering study the choice of target material was chosen based on the experimental design and the n eeds of the process. The target composition , size, manufacturer a nd purity are listed: The dopant concentrations of aluminum, gallium and scandium were constant at 2.4 mol % for the respective targets. The magnesium containing targets were all 10 mol % Mg . The targets were stored under vacuum at all times. The substrate for bare TCO deposi tions was Corning soda lime glass microscope slides. Before sputtering the glass was cleaned with acetone and dry nitrogen. The CIGS substrates terminated in CdS for making complete devices with the experimental TCOs were fabricated by an industrial collab orator on substrates of thin stainless steel . The unfinished CIGS substrates were stored under dry nitrogen.

PAGE 33

33 The device growth proces s of CIGS is outlined in Figure 1 12 . The first step t o deposit the resistive layer. In the reactive sputtering process obtaining a resistive film was a matter of adjusting the oxygen to the regular TCO target, as explained in the third chapter . The next step was to deposit the conductive layer . The last deposition was the aluminum contacts, which were deposited by metal ev aporation. The grid pattern was obtained by a metal mask. The final step of the processing was to divide the CIGS into individual devices , which is outlined in Figure 1 13 . Precision was required to obtain the exact cell area for accuracy in efficiency cal culation from measurements. The metal pattern provided alignment marks. The device was cut with a razor blade . After the cells were cut, the back contact was exposed by scraping off the CIGS outside the active device area to expose the molybdenum. The fina l step was to complete the back contact with indium solder . Characterization and Analysis The following section discusses the techniques that were used to collect data. There were two types of ZnO samples prepared. One was ZnO on bare glass and the other was CIGS devices . Analysis techniques for the ZnO on the glass give the data used in the analysis for the work presented in the third chapter . One essential test for ZnO performance was the optical transmission measurement , which was performed on a Lambda 900 photospectrometer. The optical band gap can be obtained from the transmission by linearly extrapolating the optical absorption data . Films of different decreasing sheet resistance are expected to show gradually increasing bandgaps according to the Mo ss Bernstein shift . The trend plotted predicts the change in the carrier concentration and the compensation of dopants. One more key piece of information obtained from transmiss ion data is the film thickness, which is calculated by the interference fringe spacing.

PAGE 34

34 Another essential performance test was the e lectrical resistance of the film is obtained from the four point probe analysis by an in house built system . Four points of contact are so that two points can supply the electric load and another two po ints can measure the voltage difference , as illustrated in Figure 1 14 . The method eliminates distortion caused by resistance in the film and wires. The apparatus applies a known current and the sheet resistance is extracted by : (1 17) Where a proportionality constant corrects for the offset of the current and voltage probes . Microscopic analysis of the film quality was performed using theta two theta X ray diffraction . le as illustrated in Figure 1 15 . The X ray source and detector move in synchronized motion to maintain the same value of theta relative to the substrate. n = 2dsin ( 1 18 ) Where n is an integer, is the X ray wave length, d is the lattice spaci ng and is the incident and detection angle. The c axis equation for the 002 peak works out to: ( 1 19 ) Where the K is the wavelength of the copper k x rays in nm and is the x ray angle in radians. An additional aspect of the peak wa s also the spread of the signal, which was expressed as the full width half maximum (FWHM). The wider the spread, the smaller the size of the grains. In general the growth conditions that produced better performing films produced larger grains. The grain s ize can be calculated by the well known Scherer formula:

PAGE 35

35 ( 1 20 ) Where is the grain size in nm, is the spread of the instrument and 0.9 is a correction factor for the shape of the grains. Data for solar cell performance is obtained from power measurements under the terrestrial light standard: Air Mass 1.5 (AM1.5). The measurements for the solar cell performance were taken with an in house built probe station and parameter analyzer setup. Th e output is the current voltage curve, with a sample shown in Figure 1 16 . The easiest parameter to extract is the short circuit current (Jsc), measured in miliamps per square centimeter of the device. It is the vert ical intersection of the graph, V = 0 . T he second parameter to extract is the open circuit voltage (Voc), which is the intersection of th e curve at the horizontal axis, J = 0 . The third parameter to extract is the efficiency, which is calculated as: (1 21) Where Pma x is the combination of current and voltage that produces the maximum amount of power. Power is the product of current and voltage , the maximum power is found by the local maximum in plot of power, as indicated in the figure. T he final parameter, the fill factor (FF), is obtained from the max power point, the Voc and the Jsc. It is defined as follows: ( 1 22 ) The relative areas of the two squares on the figure indicate the same parameter how well the actual power fills in relative to the short circuit currents and open circuit voltages . The other important solar analysis technique is the quantum efficiency measurement, which assess the photon collection efficiency at each wavelength of light over the part of the solar spec trum that CIGS is sensitive to. The setup consisted of an in house Labview based

PAGE 36

36 setup . For ZnO purposes, the overall QE can reflect how well the TCO performs. If there is excessive transmission loss in the TCO, it will be very evident in the QE.

PAGE 37

3 7 Figure 1 1 . The CIGS device structure . Figure 1 2. The TCO diagram illustrating the length spanned between the electric contacts relative to the device thickness.

PAGE 38

38 Figure 1 3 . Trend of transmission and sheet resistance . Figure 1 4. Sample ZnO:Al transmission spectrum .

PAGE 39

39 Figure 1 5. The quarter wave length interference effect . Figure 1 6 . The target and electric field diagram.

PAGE 40

40 Figure 1 7. The sputtering system diagram . Figure 1 8. The sputtering chamber configuration .

PAGE 41

41 Figure 1 9. The sputtering gun diagram . Figure 1 10. The heated substrate holder diagram .

PAGE 42

42 Figure 1 11. The m ass balance of the sputtering gas . Figure 1 12. The TCO completion process of the CIGS device.

PAGE 43

43 Figure 1 13. The post metallization device finishing . Figure 1 14 . The four point probe.

PAGE 44

44 Figure 1 15 . ray diffraction . Figure 1 16 . The characteristics of the JV plot in relationship to the extracted performance parameters.

PAGE 45

45 CHAPTER 2 THEORETICAL TCO MODELS AND NANOSTRUCTURE ENHANCED PERFORMANCE The CIGS TCO is a critical layer that has the potential to create a bottleneck in the electrical circuit and the optical pathway. One aspect of optimization is the thickness, which balances conductivity and transparency. A full scale 2D device simulation i s performed over a range of ZnO thicknesses. A new method of predicting the optimal thickness is presented. Also, a graded structure TCO is presented as a method for improving the TCO performance. A novel nanostructure is presented as a method of minimizin g the free carrier absorption loss in the TCO . TCO Performance Modeling The test basis is a 2D simulation of a full CIGS device. 2D solar cell simulations are rare in the literature , as the prediction of device properties, other than the TCO, can be accur ately approximated in 1D structures. The TCO resistance can create a voltage gradient on the surface, which causes a deviation from the performance fit that assumes a lumped series resistance : (2 1) Where Jo is th constant, T is the temperature, Rsh is the shunt resistance and JL is the photogeneration current. The problem of the lumped series resistan ce is illustrated in Figure 2 1 wi th the illustration of the current profile in the TCO. From the centerline of the TCO, the line in the middle of the contacts, to the metal contact, there is an accumulation of current in the TCO layer. At the center line the current is zero. Moving toward s the edge of the device the current accumulates as a function of x. At each differential element, dx, the region of the device behaves as an individual diode that follows equation 2 1. The current of Wdx is given as: (2 2)

PAGE 46

46 Th e current of each differential element experiences an additional resistance towards the TCO. The additional loss at each dx depends on the position x, as there are different path length for the different currents. The total current can be represented by an integral over the entire length: (2 3) Where V is the voltage at each element from 0 to x and v is a dummy variable. The complex ity of producing the JV curve is easiest to address using a full scale 2 dimensional device simula tion. Medici The 2 dimensional device simulation is based on p revious 1 dimensional work by J . Song et al at the University of Florida [21] . The CIGS data modeled in the work was a 1999 from the National Re newable Energy Laboratory (NREL) [22 ] . The modeling software Analysis of Microelectronic and Photonic Structures 1D ( AMPS 1D ) is small scale device calculator based on solving the continuum device equations by the finite element method up to 2000 mesh nodes . For the present work the softw are Medici is employed , which is also a continuum finite element me thod but with millions of mesh nodes . Both 1D and 2D simulations were performed. The structures are illustrated in Figure 2 2. On the left side is the 1D layered structure starting with th e molybdenum, continuing with the CIGS, CdS, iZnO and the top contact. The material parameters are used from the previous work . In the horizontal direction there is only a single mesh, so there is no possibility of lateral current. In the 2D structure, sho wn on the right, the same material stack is used except for a 2D scale magnetron sputtered film [23] . The critical parameter for series resistance was the electric resistivity, which was 4.3 x 10 4 cm. In the 2D structure a mesh grid in the horizontal and

PAGE 47

47 vertical direction so that current can flow in both directions. The 1D simulation does not take into account the series resistance associated with the ZnO TCO. Therefore the 1D device is considered to be an ideal TCO of zero resistance while the 2D takes into account the loss. The J V results of the 1D and 2D sim ulation are shown in Figure 2 3 with the NREL data represented by black dots, the 1D by the red line and the 2D represented by two sample TCO thicknesses to illustrate the reduction of the JV performance by series resistance. The 1D is a good fit except for the area a round the maximum power point , where the red line bows out slightly more than the black dots. The 1D result also agreed with the work by Song. The 2D results were performed over TCO thickness ranging from 13 to 1000 nm, wher e most of the JV curves looked exactly like the 2D until the sheet resistance of the TCO was greater than 100 /sq. The 35 nm TCO run, a sheet resistance of 125 /sq, and the 13 nm TCO, a sheet resistance of 332 /sq, show a gradual broadening of the knee as a direct result of series resistance. The efficiencies of the 1D, 35 nm and 13 nm devices were 19.7%, 19.0% and 18.1%, respectively. T he material was applied to CIGS concentrator applications. A series of CIGS TCO film thicknesses ranging from 13 nm to 1000 nm were run at sun concentrations from 1 to 100 suns . The efficiencies are plotted against the sun concentration for a series of film thicknesses as well as the 1D case in Figure 2 4 . The trend of the 1D is a steady increase from below 19.7% at 1 sun to 23.1% at 100 suns . The efficiency of CIGS can be greatly boosted by as much as 3% due to saturation of defects and recombination centers. The phenomena is experimentally known to make CIGS more efficient , w here an experimental efficiency of 21.5%, was a chieved under an il lumination intensity of 16 suns [14] . But in CIGS the TCO may limit the potential. F or the 2D simulations the efficiencies fall off from the 1D efficiency at high sun concentrations .

PAGE 48

48 Analytical The dominant effect in a 2D device is the series resistance limitation by the TCO. A more simplified approach towards the device model can be to approximate some of the non TCO aspects from the traditional rigorously calculated device simulation. 2D photovolta ic simulations [24] [25] [26] and more recently [27] [28] . There are a several methods to simplify the calculation by treating parameters that should not deviate much as fixed. The f irst assumption is a uniform photon flux over the area of the device. The second assumption is that the flow lines of the current are approximately only occurring in the lateral direction. Figure 2 5 A, the current profile is mostly lateral except around t he vicinity of the contact. In part B, the current profile is approximated as being purely lateral as if the contact extended into the depth of the TCO. The lateral current only approximation is justified by the fact that the TCO is several thousand times longer than it is thick. The third assumption is ohmic contacts. The fourth is negligible resistance in the grids. An additional assumption allows for a very simplified assessment of TCO series resistance effects. Because the total loss in a solar cell as sociated with the TCO is typically only a few percent, there is not a lot of deviation in the diode current at different points between the diode and the contact. If the diode effects can be approximated in any way, the calculations are greatly simplified. In 2009, a calculation method for the TCO electrical loss based on a non diode treatment of the thin film device was presented [28] . The device is treated as a constant and uniform current source, as illustrated in Figure 2 6. Equation 2 1 is reduce d to , where J M is the current at the maximum power point. Under constant current, the current accumulation at any point is linear from 0, where I(x=0) = 0 and L, where I(x=L) = JM*W*L. The current as a function of x is: (2 4)

PAGE 49

49 Ove r each differential element there is a differential power loss associated with the TCO 2 ): (2 5) Where the resistance can be put in terms of the TCO resistivity, , the TCO thickness, t and the width, W: dx (2 6) The total power loss can be achieved by inserting 2 6 into 2 5 and integrating from x=0 to x=L: (2 7) A more convenient expression is the power loss per area of the device: (2 8) If the TCO loss is the dominant effect expected in a 2D CIGS simulation, the expectation is for equation 2 8 to predict the true device loss with r easonable accuracy. The ZnO:Al electrical properties used in medici as well as the sun concentrations and series resistances were applied to equation 2 8. The calculated loss was used to obtain the analytically estimated efficiencies and compare to the Med ici efficiencies as follows: (2 9) (2 10) The efficiency of the Medici and the a nalytical methods are p lotted in Figure 2 7 versus the sheet resistance. To compare the performance of different sun concentrations, the sheet resistance is multiplied by the number of suns. The Medici is represented by the dashed lines and the analytical by the solid lines. Col ors are coding the different intensity levels of sun. Overall the

PAGE 50

50 series resistance analytical model accurately predicts the actual device performance as a function of the series resistance. The correlation of the efficiency prediction is most accurate at low sheet resistances and low sun concentrations. As the sun Rs approaches 400 Sun /sq, the deviation of the Medici and the analytical efficiency becomes about 2%. Opto Electric Method A new meth od of analyzing the TCO electrical and optical performance in CIGS is presented. Equation 2 8 is use as the method for calculating electric loss. A method for calculating optic loss is presented and given in combination with electric loss for a new TCO rating method. The electrical loss, E L and the optical loss, O L are the components of the total loss, P L : (2 11) The optical loss, like the electric loss, should have a loss parameter that can be calculated in terms of the film thickness. Optical energy in the input solar spectrum does not relate directly to the electrical energy output. The degenerate loss mechanisms are illustrated in Figure 2 8. In addition to the optical absorption in the TCO there is the thermalization loss and the upper and lower BG cutoffs associated with CdS and CIGS, resp ectively. When the TCO film thickness is varied, the optical absorption loss varies in a proportional manner. After the TCO, the rate at which the remaining energy is converted to electrical energy is independent of the TCO thickness. Therefore the optical energy loss in the TCO can be equated to a current reduction by a calculated loss factor for the mechanisms of the rest of the device. At each wavelength of light coming into the TCO there is an absorption coefficient, . The first step is to weight t he coefficient by the AM1.5 energy intensity at that spectrum.

PAGE 51

51 The next step is to convert the energy intensity into a current intensity by the thermalization loss ratio, which is the ratio of the energy after an d before thermalization. It is calculated by the photon energy and the CIGS bandgap energy. In terms of wavelength the thermalization factor can be expressed as follows: (2 12) The AM1.5 spectrum is plotted in Figure 2 9 showing both before and after thermalized weig hting. The next step is to account for the bandgap cutoffs and the photo conversion efficiency. The information is contained in the quantum efficiency (QE) trend. A sample from the fifth chapter is shown in Figure 2 10. The shape is approximated by linear trends, which is reasonable accuracy for the TCO loss calculation purposes. The QE correction, C QE , is calculated as follows for different ranges of wavelength: < 300 300 < < 500 500 < < 100 (2 13) 1000 < < 1100 < 300 The final expression for the weighting factor of the optical absorption loss at each wavelength is as follows : (2 14) The optical transmission of a particul ar wavelength is given as: (2 15)

PAGE 52

52 Where t is the TCO film thickness. The current without absorption loss is referred to as J, loss : (2 16) by the product of the transmission loss weighted at each wavelength, integrated over the CIGS absorption range and normalized as follows: (2 17) (2 1 8 ) T he transmission can be approximated by a Taylor series: (2 19) For values of t less than 0.1the series can be truncating at the first term and have reasonable accuracy for TCO optical loss purposes. The advantage of a linear relationship with trans mission and film thickness allows equation 2 18 to be simplified by taking the t out of the inte gral: (2 20) (2 21) (2 22 ) The I absorption coefficient is effectiv ely a current absorption coefficient lumped over the entire CIGS absorption spectrum. The term is very convenient for a simplified expression of the optical power loss:

PAGE 53

53 (2 2 3 ) Where V is the voltage at the J V maximum power point . T he previously introduced J is the current at the maximum power point the assumed conditions that the solar cell would be operated under . The new expression for the optical loss and the previous expression for current loss may be combined into equation 2 11 to obtain a total power loss as a function of the film thickness: (2 2 4 ) The benefit of equation 2 2 4 is that the optimal film thickness may be conveniently found by the local minimum found by the derivative: (2 2 5 ) The optimal film thickness is obtained by solving equation 2 22 for t: (2 2 6 ) The next step is to demonstrate the model on a TCO and predict the theoretical electrical and optical loss. A sample transmission spectrum from the fourth chapter data is used to test the method. MgZnO:Sc showed good transparency for an electrical conductivity of 0.00516 cm. The optical transmission and the extracted absorption coefficient are shown in Figure 2 1 1 A and B, respectively. The absorption data is analyzed by the simplified and the rigorous method. For the simplified method the absorption data is applied to equation 2 22 to obtain the current loss coefficient. The rigorous method is the calculation of equation 2 18 at multiple film thicknesses. The fraction of current lost as a function of film thickness is plotted in Figure 2 1 2 for both the rigorous and the estimated method. The calculated electrical absorption coefficient is reduced by a few percent so that the linear approximation crosses the rigorous around 470 nm, the film

PAGE 54

54 thickness of the TCO of the transmission measur ement . The accuracy is reasonable at low film thicknesses but deviates by a larger amount at thicknesses above 1000 nm as the loss trend becomes less and less linear. The data is applied to the power loss equation. The results are plotted in Figure 2 13, w here the dashed line represents the simplified calculation and the solid line the rigorous calculation. There is a good fit of the simplified method around the optimal film thickness. At higher film thicknesses there is more disagreement between the two me thods due to the error of linearizing the optical loss calculation. The optimal film thickness of 530 nm, predicted by equation 2 23, matches the plotted calculation of the trend. The optimal film thickness from the rigorous method of 580 nm deviates. Howe ver there is only a very small difference in TCO power loss in the thickness range of 500 to 600 nm. The optimal film thic have to be perfectly accurate to be close to the maximum efficiency due to the parabolic broadening of the curve . Graded TCO One aspect of optimizing the thickness of a flat TCO is that the method does not consider how the electrical loss varies at different regions of the device. Figure 2 14 illustrates the loss. The current has a graded structure from the center po int to the contacts. At the center the TCO is too thick for the amount of current and there is unnecessary optic loss. Near the contact the TCO is too thin and there is unnecessary electric loss. A new TCO structure is proposed in Figure 2 15. The structur e is a TCO that is linearly graded in thickness from the center to the contacts. It is thinnest at the center and thickest at the contacts. The advantage of the graded structure will be analyzed by the analytical method above. The first step is to calculat e the electric loss by the method of equation 2 7. First the dR term must be modified for the graded thickness as follows:

PAGE 55

55 (2 2 7 ) Where to is the t hickness at the TCO contact side . The integral can be re written as: (2 2 8 ) In comparison to the result of equation 2 7 the result is the same except the 1/3 factor is replace by a ½ factor. When the thickness of the flat TCO is the same as the thickness of the edge of the graded TCO the graded TCO has 50% more electric loss. However, the anticipation is that the optical loss will be reduced by a greater percentage. The optical loss will be calculated. Unlike equation 2 2 3 , where the optical loss is uniform over the area o f the TCO, the optical loss is non uniform for the graded and will need to be integrated. For the differential area of length dx, width W and thickness t(x) the amount of current lost is given by the following expression: (2 2 9 ) The optical loss per unit area is integrated as follows: (2 30 ) The optical loss in the graded TCO is ½ the loss in the flat TCO. The next step is to quantify the minimal power loss of the graded TCO. The total power loss as a function of to is written as: (2 31 ) Next, the optimal to is found by the local minimum using the derivative, as done in equation 2 2 5 : (2 32 ) The optimal to is expressed similarly to equation 2 23:

PAGE 56

56 (2 33 ) The final step is to determine the relative difference of the minimal loss of the flat and the graded TCO: (2 3 4 ) The equations 2 21 and 2 28 are plugged into equation to have an expression in terms of the optimal thicknesses: (2 3 5 ) In order to put the equation in like terms, the respective optimal thickness expressions (equations 2 2 7 and 2 3 3 ) are inserted into 2 3 5 . The like terms cancel out to reduce the expression to the following fraction: (2 3 6 ) The flat TCO minimal power loss is higher by a f performance of the graded TCO was calculated by a rigorous method and the simplified analytical method. The results are shown in Figure 2 16. The flat TCO lines are blue and the graded TCO lines are red. Also, for convenience, the graded TCO was plotted versus the average thickness. The agreement of the rigorous and the simple methods for the graded TCO are similar as for the flat TCO. The minimal power loss of the graded TCO is 0.63 mW/cm2, whereas its 0.72 mW/cm2 for the flat TCO . The improvement is about 13%. Furthermore the average thickn ess of the graded TCO is 460 nm, compared with 530 for the flat TCO . The calculation suggest the overall amount of TCO material can be less compared with the flat TCO. A reduction in the amount of indium consumed in the TCO could make the use of ITO more feasible for projected large scale solar production.

PAGE 57

57 Multiple Quantum Wells High mobility electron carriers are the key to producing a TCO for CIGS that provides excellent conductivity and transp arency. Firstly, high mobility allows the optimal conductivity to be achieved with relatively fewer carriers . Secondly, a higher mobility reduces the IR absorption per carrier. Traditional TCOs such as ZnO:Al are optimized for a maximum possible electrical c onductivity. In Figure 2 17 ZnO doped with various elements, with conductivities ranging from 1 to 8 x 10 4 carrier density with the data from a rev iew by N. Minami of state of art TCO performance materials [29] . The elements aluminum and gallium result in the best overall electrical performance with the data points being in the top right part of the cluster. T he mobility in heavily doped ZnO, 10 21 e /cm 3 , is limited to a maximum of 30 cm2/V s due to ionized impurity scattering [30] . The Brooks Herring Dingle (B H T) theory suggests a maximum mobility of 90 and 30 cm 2 /V s (at 10 21 e /cm 3 ) for not considering and considering the deviation of the ZnO conduction band from parabolicity. While the a bsolute electrical conductivity of ZnO can be obtained at extreme doping levels, the drawback is the loss of IR transmission at longer wavelengths. When the free carrier absorption is modeled by the drude theory for a film of a fixed conductivity the free carrier scattering absorption coefficient is inversel y proportional to the mobility: (2 3 7 ) Pure ZnO has a mobility of 200 cm 2 /V s in bulk and in quantum confined states 250 cm2/V s . In bulk semiconductors the ionized scattering centers ar e built in with the dopants and cannot be avoided. However, an engineered electronic structure to separate carriers from the carriers can allow a high electron mobility to be realized in a doped ZnO .

PAGE 58

58 The electronic structure is a repeating well barrier pattern, as shown in Figure 2 18. First t he bandgap of the well material is lower than the bandgap of the barrier material. Second, the electron affinity of the well material is higher than the electron affinity of the barrier. Electron donors doped in the barrier layers, result in carriers will occupy the wells while the ionized donors remain in the barrier . Therefore there ionized impurity scattering effect will be separated from the carrier density of high doping . The concept was first demonstrated in th e GaAs/AlGaAs supperlattice [31] . Aluminum alloyed GaAs has a higher conduction band than GaAs so that electrons are confined in the GaAs layer and depleted from the AlGaAs. The two materials have different electronic propert ies yet very similar lattice matches. F or the strategy to work in ZnO the two sandwiching layers must also have as close as possible lattice matches. Otherwise there would be defects and strain on the interface, which reduces the electron mobility. One si mple way , which for layer by layer growth will have i ZnO and ZnO:Al layers alternately deposited. If the ZnO:Al layer is heavily doped, up to 0.3 eV band offset can be achieved due to the Moss Bernstein effect. The problem, h owever, is that . The second method, which is the more complex of the two, is using magnesium alloyed ZnO and ZnO . MgZnO was first proposed as a novel TCO material i n 1998 where t he MgZnO was deposited epitially on 0001 saphire by PLD from ceramic ta rgets in pure oxygen conditions [32] . The superlattice structu re is illustrated in Figure 2 19. There are two advantages of using MgZnO/ZnO over delta doping. First, much higher bandgap differences can be obtained, as MgZnO can reach 4.0 eV at magnesium levels above 30% . Second, the MgZnO itself can be modulation dop ed so that MgZnO:Al is sandwiched by two layers of undoped MgZnO, which has the effect of separating the scattering radius reach of the ionized donors from the electrons in

PAGE 59

59 the wells. Previous work proposed 1D continuum with Poisson/Shrodinger and molecul ar with Monte Carlo for electron scattering have shown that this system is capable of outperforming uniform doped ZnO:Al and ITO [33] [34] [35] . Experiments have verified the electron confinement of the iZnO channels by RF sputtering [36] [37] [32] , pulse laser deposition [38] , chemical vapor deposition [39] and molecular beam epitaxy [40] . Density of States The Cohen work showed that carrier confinement in the wells starts at 10 13 e /cm 2 . The exact amount of carrier confinement is challenging to predict. Experimental measurements of confinement in epitaxial grown ZnO/MgZnO superlattices suggest that the number can approach 10 14 e /cm 2 . The continuum principles employed by Cohen showed that carrier densities in the sheet are indep endent of the thickness of the well and barrier but depend primarily on the height -not the doping in the barrier either, as long as it is enough to completely fill t he well to the maximum capacity . The goal of the MQW approach is to maximize the overall N in the TCO regions where the exceeds 200 cm 2 /V s. Cohen suggested wells and barriers as thin as 2 nm. The problem, however, is the breakdown of the density of states. Cohen touch only briefly on the issue. The present work investigates the limits of car rier confinement in quantum wells considering the discretization of the occupied energy states. The model is a square quantum well, as illustrated in Figure 2 20. The input parameters are the well width (W W , nm), the barrier width (W B , nm), the barrier hei ght (H B , eV) and the electron effective mass in the well (m eff , W ) and the barrier (m eff , B ). For ZnO, the well material, the electron effective mass (ratio) is 0.38. The properties of MgZnO are for Mg levels in excess of 20% in order to maximize the carrier density level and confinement. The barrier height is 0.5 eV and the electron effective mass ratio is 1.5.

PAGE 60

60 There are three well widths considered: 5 nm, 10 nm and 15 nm. In each case, the well width and the barrier width are the same. The simple model is the square quantum well. The method and solution are commonly known. However since there are several equations involved the whole method is outlined to ensure clarity . T h e wave function in a well with finite barriers is written as For the x coordinate origin is the center of the well and the right and left boundaries of the well are W B /2 and W B /2, respectively. And the wave f unction on the barrier is zero. : (2 3 8 ) The solution to both forms of 2 35 is written as: (2 3 9 ) Where the normalization coefficients and are: (2 40 ) (2 41 ) The energy (E) is written as: (2 42 ) Where m eff is the m eff,W or m eff,B , depending on the region. The matching of the wave functions and respective derivatives at the boundaries leads to four equations: (2 43 ) (2 4 4 )

PAGE 61

61 (2 4 5 ) (2 4 6 ) Equations 2 39 through 2 42 can be combined and rearranged into conditions for allowed energy levels and transcendental equations: (2 4 7 ) (2 4 8 ) A simple approach is a graphical method with 1/2aW and 1/2bW being the equivalent of Cartesian coordinates x and y: (2 4 9 ) For the given well parameters a plot of the tangent and the circle is made in Carte sian coordinates with the above The graphs are shown in Figure 2 21. Each intersection of the circle represents an energy level. The intersection starting in the upper left is nm the max n is 3, for the 10 nm 7 and for the 15 nm 10. The resulting energy levels are plotted in terms of n levels as a function of the energy of the levels in Figure 2 22. The pattern of the steps outlines the hypothetical continuum of the density of states, which is indicated by the dashed lines. T he carrier density at each nth energy level can be calculated as follows: (2 50 ) At each n, the total carrier occupancy in the well up to n is an accumulation: (2 51 )

PAGE 62

62 In Figure 2 23 A the total sheet density is plotted as a function of the energy level. Each point represents a level. The amount of carriers versus the energy is indicated by stair step solid lines. The carriers filling the energy level should be of too high energy states such as in excess of 100 meV. There is a cluster of states for the 3 well sizes at energy from 28 to 35 meV with the sheet carrier densities ranging from 1.4 to 2.1 x 10 13 /cm 2 . For the 1 5 nm and 1 0 nm wells the next level of filling occurs at 63 and 74 meV , which gives sheet densities of 4.3 and 4.4 x 10 13 /cm 2 , respectively. The next level of the 5 nm well is at 1 20 meV, which is out of reach because it reduces the barrier separation from 0.5 to 0.38 eV. The well doping is a sort of carrier sheet density. To compare the MQW TCO performance to conventional TCOs, the bulk electric property must be calculated. The MQW TCO has conductivity as a result of the number of electrons in the wells there should be few electrons in the barrier if the barrier doping is well optimized. The thickness of 1 barrier plus 1 well is the period length. The effective bulk carrier concentration is computed by dividing the sheet density by the period thickness: (2 52 ) The effective bulk carrier concentration is plotted in Figure 2 23 B. The 5nm and 15 nm wells have a level at 1.4 x 10 1 9 /cm 3 . The 5 nm energy level is 28 meV whereas the 15 nm well is at 63 meV. The 10 nm well, however, has up to 2.2 x 10 1 9 /cm 3 doping density at 74 meV. Another possibility is the next level of the 15 nm well, which would effectively have up to 2.5 x 10 1 9 /cm 3 at 97 meV. If further raising of the electron energy level was allowable the next 5 nm level would have a doping level of 5.5 x 10 1 9 /cm 3 at 112 meV. However, once factor to consider is the ionized impurity scattering. The charged centers should be as far from the electron wells as possible. A suggested st ructure is shown in Figure 2 -

PAGE 63

63 4. The dopants are deposited in a monolayer in the middle of the barrier material. If the desired effective carrier level is between 2 and 5.5 x 10 1 9 /cm 3 , compacting the required dopants into a layer 1 nm thick out of a period of 20 nm would equate to a doping level of the 1 nm between 4 x 10 20 /cm 3 and 1.1 x 10 21 /cm 3 , which is well within the doping capabilities of ZnO. The range of an ionized impurity scattering center is determined by the field potential around the barrier. The radius where the field potential is approximately thermal energy (25 meV), the radius is 6 nm in ZnO. For the 5 nm well, the 1 nm doped would le ave 2 nm undoped MgZnO on each side as a buffer between the charged centers and the high mobility electrons. For the 10 nm well, there would be a 4.5 nm buffer and for the 15 nm well, there would be a 7 nm buffer. By a traditional understanding of ionized scatterings the 5nm and possibly the 10 nm well w ould be ruled out. However, the 2D sheet of charged donors may not have traditional scattering effects, as illustrated in Figure 2 25. In part A the donors are loosely spaced so that the effects of individua l atoms are felt by electron carriers. In part B, with donors more closely spaced, the field radii overlap into a continuous field. Predicted IR Enhancement If a channel mobility of 250 cm2/V s is realized the bulk electric conductivities of effective she et densities ranging from 1 to 5 x 10 1 9 /cm 3 , the result is resistivities that range from 2.5 x 10 3 to 4.9 x 10 4 cm. The absolute bulk electrical conductivity is unlikely to surpass the conductivity of traditional ZnO and ITO with heavy doping, which ap proach 1 x 10 4 cm. However, a tremendous advantage is the optical property of a TCO with high mobility electrons. In the following section the TCO transmission properties in the IR range are calculated and compared with traditional ZnO:Al.

PAGE 64

64 The interest in free carrier absorption for semiconductors has been silicon based photodetectors that suffer from the near infrared absorption [41] . For the application towards improving the IR properties of ZnO, however, there are very few reports in the literature. A likely reason is that IR absorption in ZnO TOCs is less important for CIGS with bandgap cutoffs around 1.2 eV. However, there is a significant portion of the solar energy between 0.7 and 1.2 eV that could be captured by future generation devices that are sensitive to the full solar spectrum. Traditional ZnO:Al with heavy doping has a large IR absorption loss. The state of art RF sputtered ZnO:Al electrical properties are shown in Table 2 1 [23] . The free carrier absorption is fitted based on the provided transmission trends. The free carrier absorption coefficient of semiconductor s is traditionally calculated using the drude theory as the basis: (2 53 ) ZnO does not fit exactly with the Drude model, but is much closer with an additional lambda term. The following equation was used to fit the data: (2 5 4 ) (The wavelength is in microns.) Next, the equation was used to predict the IR absorption of a MQW TCO 1000 nm thick and an electrical resistivity of 5 x 10 4 cm. The results are shown in Figure 2 26 B. For comparison, the AM1.5 spectrum is plotted in part A. The IR loss of heavily doped ZnO increases significantly with Al content and the increasing wavelength. Approximately 30% IR energy is lost with the 0.5% Al and 50% with 2% Al. With the MQW TCO, however, there is an excellent transparency capability. Over 95% of the IR could theoretically penetrate to the photovoltaic absorber.

PAGE 65

65 Table 2 1. RF Sputtered ZnO:Al electric properties [23] . % Al 0.5 1 2 (cm 2 /V s) 32 30 22 N (cm 3 ) 2.75 x 10 20 4.30 x 10 20 8.00 x 10 20 ( cm) 7.01 x 10 04 4.79 x 10 04 2.86 x 10 04 t (nm) 726 657 513

PAGE 66

66 Figure 2 1. The TCO current collection structure in CIGS . Figure 2 2. The material device structure for the 1D and 2D Medici simulations.

PAGE 67

67 Figure 2 3. The Medici CIGS 1D comparison [22] . Figure 2 4. The 2D CIGS efficiencies at multiple TCO thicknesses and sun concentrations.

PAGE 68

68 Figure 2 5. Current flow line approximation. A. Vertical flow in proximity of the contact. B. Vertical flow effected neglected by approximation with a contact perpendicular to the flow. Figure 2 6. The constant current flux (non diode) TCO loss model.

PAGE 69

69 Figure 2 7. The comparison of the Medi ci and the analytical calculated efficiencies as a function of the number of suns and sheet resistance. Figure 2 8. The energy loss at CIGS stages from solar input to electric output .

PAGE 70

70 Figure 2 9. The AM1.5 spectrum for weighted and unweighted by the CIGS thermalization . Figure 2 10. Quantum efficiency weighting. A. A sample QE. B. The linearly estimated QE for simplified calculations .

PAGE 71

71 Figure 2 1 1 . A sample TCO. A. The optical transmission spectrum. B. The extracted optical absorption coefficient. Figure 2 1 2 . Current loss trend.

PAGE 72

72 Figure 2 1 3 . The explicit MQW CIGS concentrator . Figure 2 14 . The inefficiency of the traditional flat TCO structure .

PAGE 73

73 Figure 2 15 . The graded TCO structure . Figure 2 16. The relative performance of the flat and graded TCOs.

PAGE 74

74 Figure 2 17. The mobility of state of art hea vily doped ZnO compared with ionized impurity scattering limited mobility trends [29] . Figure 2 18. The quantum well electronic band alignment of a 2DEG enhanced TCO.

PAGE 75

75 Figure 2 19. The superlattice structure of ZnO and M gZnO. Figure 2 20. The energy levels formed in the quantum well with deviation of the density of states from a continuum.

PAGE 76

76 Figure 2 21. The alpha and beta plots for obtaining quantum well energy levels. A. 5 nm. B. 10 nm. C. 15 nm. Figure 2 22. The calculated energy levels plotted as a density of state number (n) versus the energy.

PAGE 77

77 Figure 2 23. Carrier density levels. A. Carrier sheet density. B. Effective bulk carrier density. Figure 2 24. Doping profile design.

PAGE 78

78 Figure 2 25. Ionic scattering radius overlap with quantum wells. A. Loose spacing. B. Tight spacing. Figure 2 26. Modeled IR absorption. A. AM1.5 (for comparison). B. Fitted and predicted IR loss for ZnO:Al and ZnO MQW, respectively.

PAGE 79

79 CHAPTER 3 DC SPUTTERING : THERMAL AND NEGATIVE ION EFFECTS ZnO :Al is a wide bandgap semiconductor that is suitable for large scale solar production because of the earth abundance of element zinc and the nontoxicity of the compound. Sputtering is a fast and reliable method to depo sit films of ZnO :Al , but the drawback of sputter deposited ZnO for CIGS is that films rely on heated substrates [42] [17] or post deposition annealing [43] to obtain conductivities b elow 10 3 ohm cm . The re are two phenomena to explain the loss of conductivity at low temperatures. The first is the limit to the grain size and density that is able to form at low temperatures [44] . The second is that heating or post deposition annealing improve properties because of annealing repair to defects caused by resputtering of the growing film with high energy oxygen ions [45] . A n experimental investigation and modeling explaining the observations is presented. Structure Zone Model Experiment Grain size and quality is very important to polyc rystalline ZnO TCOs to optimize mobility. The quality of the film crystals is correlated with the growing film t emperature and the sputtering gas pressure [44] . C rystal film properties improve with changing sputtering conditions , as illustrated in Figure 3 1 . For the sputtering parameters temperature and pressure, there are four quadrants based on high and low temperature and sputtering gas pressure, respectively. For films produced at the low end of the range in this study the gra ins are small and have a lot of void space, which does not allow for optimal conductivity. For high temperature and low pressure there are very large and well organized grains with minimal void space. For the other combinations there are moderate sized gra ins.

PAGE 80

80 By crystal growth theory t he root element , determined by pressure and temperature , is the length of time that a sputtered atom has to properly crystalized from arrival at the film surface to the point where it condenses as a solid . Condensation of an ad atom is illustrated in Figure 3 2. The first step is arrival and sticking to the surface. The second step is surface motion. The third step is condensation to the permanent location. The atom arrives with high energy and loses it qu ickly. The more energy, the more mobile time a molecule will have to position on a lattice site before condensing to solid form. Temperature is the first parameter. The closer the temperature is to the melting temperature of the crystal phase of the film, the more time the atom has to move to the correct location. Pressure is the second parameter. Neutral sputtered flux consists of molecules with several eV of energy. Sputtered particles undergo collisions with other gas molecules before arriving at the su bstrate. As the molecule travels to the substrate it collides with gas molecules to loose energy. The amount of energy it loses with each collision is determined by statistics of gas kinetic theory. The kinetic gas theory predicts the mean free path length to be: (3 1) Where is the hard sphere diameter of the molecules. At 1 mtorr the mean free path length is approximately 4 cm. For a target to substrate distance of 6 cm, sputtered flux experiences an average of 1.5 collisions before arriving at the substrate surface. ZnO reports often give different values for the sputt ering gas pressures ranging from 2 to 20 mtorr lower pressures give better films. Fewer collisions mean there is more energy left over for the molecule to arrive at the substrate with. When the sputtering gas pressure is lower, the atoms arrive with more e nergy.

PAGE 81

81 Experiment The experimental approach was to deposit ZnO:Al on substrates of different temperatures to assess the effects on the electrical performance and the quality of the film crystallinity. A 3 zinc metal target was sputtered at 100 Watts under an argon oxygen mixture of a total pressure of 1 mtorr. The argon to oxygen combined flow rate is fixed at 25 standard cubic centimeters per minute (sccm ) . The oxygen and argon flow rates were co adjusted to tune the oxygen partial pressure while maintain ing a constant total pressure. The substrate to target distance was 6 cm. Oxygen and Discharge Characteristics One aspect of DC sputtering is that electrons originating from the surface are dependent on the oxidation state of the target surface. ZnO has a higher ion induced secondary electron emission c oefficient than zinc metal . Therefore an oxidized target emits more current than a metallic target during argon sputtering. When the power supply operates in constant power mode, where the voltage will decrea se as the current increases. Previous reports of DC voltage discharge trends of Zn:Al metallic targets reported the same shape of the curve [46] . The reported process window was located immediately after the steep voltage fall in the oxide region . For the above conditions with Rusty, the curve shown in Figure 3 3 , where the voltage of the cathode is plotted against the partial pressure of oxygen. The optimal oxygen flow rate of 7.5 sccm, or 0.3 mtorr resulted in films that were both conductive and transparent. Lower amounts of oxygen resulted the films quickly dar kening. Higher amounts of oxygen resulted in a quick loss of transparency. Electrical Properties The resulting electronic properties are shown in Figure 3 4 for temperatures ranging from 50 to 250 C on both glass and 001 silicon substrates. The resistivit ies are too high at values near 100 ohm cm for samples deposited at room temperature. However, as the temperature

PAGE 82

82 approaches 250 C the resistivity reaches the 0.01 ohm cm range, which is near the level expected for ZnO:Al (0.001 to 0.0001 ohm cm). Crysta l Characteristics A powder XRD measurement was taken for the glass samples over the fu ll range (30 to 80 degrees). Silicon samples could not be measured because the diffraction of the substrate would damage the detector. A theta 2 theta measurement of the 100 C sample is shown in Figure 3 5 . The dominant peak is 002, which is the c plane being parallel to the substrate surface. ZnO is known to have a 002 preference. When the texture is allowed to fully develop, it appears to be vertical hexagonal columns. T he reason for the tendency is that when randomly oriented nucleation grow, some surfaces grow faster than others. The fastest surface becomes the dominant peak in a well . The 002 peak for the glass samples was a nalyzed by the Scherer grain size formula. The results are shown in Figure 3 6, A . The grains are a reasonable size of 10 to 20 nm. The peak location was also converted to the c axis length, which is shown in Figure 3 6, B . Negative Ion Resputtering Exper iment The structural zone model predicted high temperatures and low pressures would result in high quality films. It is important for these atoms to arrive with some kinetic energy that will add to surface mobility. The assumption is that the particles eje cted from the target do not have an excessive amount of energy as to resputter the film. The observed trend with temperature shows an improvement in sheet resistance, following the SZM. However, the absolute values of t he resistances are too low. A nother aspect particular with ZnO is the tendency of the target surface chemistry to form negatively charged ions, which are ejected at high energies and do a lot of damage to the growing film by negative ion resputtering (NIR). The theory of NIR predicts reduced deposition

PAGE 83

83 rates a nd damage to the deposited film if O is present. The formation of O in the sputtering process is illustrated in Figure 3 7 . When a hot argon atom strikes the surface, the first action is the transfer of energy to the oxygen atom on the surface. Second, the oxygen atom squeezed into the z inc atom results in an electron transfer due to the high electron affinity of oxygen and electronegativity of zinc . Third, the argon ion continues to impact other metal and oxygen atoms. Most will be eje cted as neutral. But there will be some positive zinc ions and some negative oxygen ions. The positive zinc ions will be drawn back to the target by the field. The negative oxygen ions will accelerate away to extreme energies and impact the target surface. T he target voltage, which is as high as 260 volts in the present study , is expected to generate O ions of that energy . High energy ions have a very strong impact on the surface of the growing substrate, causing a resputtering effect. Furthermore, the imp acts create defects in the growing film. Negative ions are driven by the Lorentz force. Argon ions are strongly affected, as there is an erosion pattern corresponding to the field lines. Negative ions, likewise, follow a confined pathway in the vertical d irection away from the sputtering target. Reports of the ZnO film properties being positional dependent based on the erosion pattern show a strong correlation [45] . Films directly above the erosion show very high resistances and films to the side of the pattern show decreasing resistances. There are a couple of approaches taken towards dealing with the problem. Some reports use higher gas pressures with the idea to increase the number of molecular collisions a sputtered atom experiences on the path to the substrate [45] . As the high energy io ns collide with gas molecules between the target and substrate, the energy is dissipated. A problem is the neutrals also loose energy more quickly. High energy n egative ions have small collision crossections and are unlikely to impact gas molecules on thei r path from the target to the substrate, which would have the effect of dissipating energy and reducing the resputtering

PAGE 84

84 yield . Pressures above 20 mtorr allow the films deposited around the erosion pattern to come to 10 2 ohm cm, which is better than at lo w pressure but not the desired range for the TCO. The present approach is to place the substrate as far out of the path of the negative ion beam as possible without losing the n eutral flux . A diagram comparing the direct and the offset orientation is shown in Figure 3 8 . To maximize the deposition rate in the offset position, the substrate is rotated so that the normal faces directly to the center of the target. Electrical Properties For the offset position ZnO:Al was sputtered onto glass and silicon under the same oxygen condition and temperature range as the direct experiments . The resistivites of the direct and the offset samples are plotted in Figure 3 9 . There is a huge difference between direct sputtered and offset sputtered films by absolute resistiv ities. For relative resistivities, the offset sputtered films stay somewhat flat around 1 x 10 3 ohm cm from 50 C to 250 C, whereas the direct sputtered films decrease from the 1x 10+1 to 1 x 10 1 range from 50 C to 250 C. It is clear that better films are obtained by the offset orientation. The effects of temperature greatly improves the resistive films directly oriented and slightly improves the already conductive films in the offset position . XRD Analysis The powder XRD scans for the directly oriented and the offset oriented samples at 100 C are plotted in Figure 3 10. The offset oriented sample reveals a single 002 peak and a lack of the other minority peaks that appear with the direct sample indicate that the offset sam ple has much cleaner grain quality. The 103 peak appears to be a secondary orientation. The 004 peak is essentially an echo of the 002 peak. Another difference is the position of the 002 peak in the offset sample is a higher 2 theta than the directly orien ted peak. There are implications of the position for the amount of film stress.

PAGE 85

85 The Scherer grain sizes for the direct and offset samples are plotted in Figure 3 11 A. T he offset samples show consistently larger grain sizes than the direct samples . If the difference in conductivity of the observed samples were due to grain boundary scatterings, it would seem that there is not enough difference in the observed size of the grains to account for the extreme differences in resistivities between direct and offse t samples. Furthermore, there was not a clear grain size correlation with temperature Another factor to consider is film stress, which is proportional to the deviation of the c axis length from that of pure ZnO , which is 0.5200 nm. The XRD extracted c axis lengths are plotted in Figure 3 11 B. The direct samples show deviation, where the offset samples are nearly in unison with bulk ZnO. Reports often cite ZnO grain stress wit h low electrical qualities . It is, h owever, un likely that the film stress is the root cause of the high resistivity . Crystals pointing in different directions may be pushing on each other to create the film stress. Another possibility is the smaller grains, as well as the mixed orientations, indicate more empty void space between uneven grains to elongate the paths of conductive carriers . The most likely explanation is that defects thought the crystals are elongating the c axis. The defects reduce the mobility by scattering events. The defect theory is also the best evidence of negative ion resputtering, which predicts damage to the crystal quality and purity. The conclusion that can be drawn from the direct and offset comparison is that the offset orientation strongly enhances both the electr ical and the texture quality of sputtered ZnO films. The temperature does not seem to have a strong effect on the film properties except for the electrical property of the directly sputtered films, which went from resistive to mildly conductive from low to high substrate temperatures. The reason for the slight improvement in the direct films is a predicted trend of the structure zone model. However, the crystallinity pattern does not

PAGE 86

86 seem to follow it. The reason why the crystallinity is not following it is possibly that the negative ion resputtering is so intense at the low pressure and short target to substrate distance that temperature is insufficient to compensate for the negative ion damage. The improvement with temperature may be that the dopant activa tion is slightly more efficient at higher temperatures. The observed improvement substrate heating and post deposition may be contributing to repairing damage as much as promoting better grains. However, avoiding or minimizing negative ion resputtering wou ld promote good quality films for temperature sensitive CIGS.

PAGE 87

87 Figure 3 1. Structured zone model of the crystal quality as a function of temperature and pressure. Figure 3 2. Illustration of the energy dissipation and surface movement of hot mo lecules in a sputtered flux.

PAGE 88

88 Figure 3 3. The oxygen voltage discharge characteristics of the Zn:Al target during sputtering. Figure 3 4. The electrical resistivities of the samples as a function of substrate temperature.

PAGE 89

89 Figure 3 5. The the ta 2 theta scan of the 100 C sample. Figure 3 6. The XRD extracted parameters. A. Scherer grain size. B. C axis length.

PAGE 90

90 Figure 3 7. The sputtering mechanism resulting in the formation of negative oxygen ions. Figure 3 8. The direct and offset substrate configurations relative to the sputtering target surface.

PAGE 91

91 Figure 3 9. The electrical resistivities of the direct and the offset oriented substrates. Figure 3 10. The theta 2 theta scans of the direct and offset samples at 100 C.

PAGE 92

92 Fig ure 3 11. The XRD extracted parameters of the direct and offset films. A. Scherer grain size. B. c axis length. Figure 3 12. Crossectional SEM.

PAGE 93

93 CHAPTER 4 MG ZNO :( AL, GA AND SC ) TCO BY OFFSET DC SPUTTERING conditions. But the high cost and global scarcity of indium make ITO somewhat of a luxury material. ZnO is vastly superior in regards to cost and draws increasing attention as a replacement for ITO in all kinds of applications from liquid crystal displays [47] to CIGS solar cells . The resistivity of ZnO is capable of matching or surpassing that of ITO [30] . However the bandgap of ZnO limits the transmission spectrum relative to ITO, being an indirect bandgap semiconductor, which has an optical cutoff at 4.0 eV and is ideal for UV transparency [48] . The optical shortcoming o f ZnO can be improved by incorporating magnesium into ZnO to widen the UV range [49] . The chapter is focused on the optical transparency advantages, employing the wider bandgap cutoff to extend the spectrum of sunlight that CIGS is responsive to. Magnetron sputtering is the experimental technique for film deposit ion. ZnO cuts of a large portion of the UV light that reaches the CIGS absorber. In Figure 4 1 the plot of the AM1.5 spectrum is given as a function of the photon energy. The CIGS spectral sensitivity is marked by a red box, which is defined by energy that is above the conduction bandgap cutoff of 1.2 eV . The light going towards the device must first pass through the TCO . If ZnO:Al is used, the light is cutoff at 3.2 eV , which is indicated by the blue box on the figure . The energy portion that is lost is 4% of the total CIGS spectrum . Incorporating a small amount of magnesium to boost the bandgap from 3.2 to 3.6 eV expands the bandgap cutoff and allows 3% additional energy to pass throu gh the TCO towards the absorber and only 1% of the CIGS spectrum lost.

PAGE 94

94 Exp eriment Choosing the right amount of magnesium is the key to obtaining a better figure of merit . The bandgap ranges from 3.3 eV to approximately 3.9 eV as up to 44% magnesium is added and maintaining a of epitxial ly deposited MgZnO on crystalline substrates [50] . The problem of high percentages of magnesium 20%+ is a loss of electrical conductivity [51] . The reason could be due to a lower electron mobility, a decrease in the rate of dopant ionizatio n and /or a reduction of the size o f the crystal grains. Furthermore, amounts of magnesium between 5 and 15% make a much greater marginal increase in optical transparency than magnesium between 15 and 30% [50] . In the present study a small amount of magnesium, 10%, was chosen to min imize electrical property loss and maximize the electrical property improvement. Doping The choice of the dopant is important for effective donor ionization, high mobility and phase stabilit y. Dopants substitute Zn for electrical charge. +3 elements that are incorporated act as shallow donors. A number of different dopants can serve the role. Aluminum is the most efficient dopant for ZnO in terms of the total electron carrier density that can be achieved, which is at least 1.5 x 10 21 e /cm3 [52] . The influence of the dopant on crysta l and electrical properties is not well understood. For this reason, side by side comparisons of multiple dopants under the same experimental comparisons are helpful. Studies that use more than 2 dopants at a time are rare. In 1985, Minami used 4 dopants g allium, aluminum, indium and boron for ZnO [52] . The result showed that aluminum is the most effective dopant for generating high carrier concentrations whereas gallium and boron tend to result in higher electron mobilities. The ionic radius is an important factor for the solubility and chemical stability of the dopant for substituting the zinc site. In the present work there are three different dopants

PAGE 95

95 incorporated into the study for MgZnO: aluminum, gallium and scandium. The effective ionic radii of zinc, aluminum, gallium and indium are 74, 53.5, 62 and 80 pm, respectively. The low ionic radiu s of aluminum relative to zinc is one possible reason for apparent interference with the grain growth and size. Gallium has a being a closer radius. However, a particular element of interest is scandium, which has a nearly perfect effective ionic radius ma tch with zinc at 74.5 pm. In that has not been applied to MgZnO is Sc, where the ionic radius matches zinc almost perfectly. However, reports of even ZnO:Sc are very rare [53] [54] . Doping evidence suggest that scandium ionizes in the mid 10^20 e /cm3 range, though slightly less efficient than aluminum and gallium. ZnO and MgZnO doping can achieve high conductivity through extrem e doping such as 4 wt% aluminum. However, for a good ZnO TCO it is sometimes of equal interest to keep a stable structure and promote grain growth for maximum mobility. One study took high resolution SEM and TEM images of aluminum at doping levels in ZnO f rom 1 to 8% in sputtered ZnO films and found that any doping level over 1.5wt% interfere with the evolution of the texture , whereas doping from 4 to 8% resulted in severe degradation of the crystal quality [55] . In the present study a doping level of 2.4 mol% of each is used. For aluminum, tha t amount corresponds to 1wt%. The target compositions were: 87.4 Mol % Zinc 10 Mol % Magnesium 2.4 Mol % Dopant (Aluminum, Gallium or Scandium) Effects of Power One of the not well understood sputtering parameter effect on the quality of deposited ZnO film s is the sputtering power. DC magnetron sputtering is capable of extremely high rates of deposition, which is part of the motivation for using the technique towards rapid processing of

PAGE 96

96 CIGS. However, r eports studying the effect of sputtering power and depo sition rates on the grain size and electrical properties do not give consistent trends. In a case for ZnO:Ga in RF oxide target, power and deposition rate was reported to improve film electrical properties [56] [57] . The role of power can be complex and alter the balance of target oxidation, film oxidation and the ratio of film growth rate to the film resputtering rate . In the present study powers ranging from 13 to 100 watts are applied to cover a wide range. The time was increased so that the product of the power and the deposition time was a constant: 2.5 min x 100 W = 5 min x 50 watts = etc. If the deposition rate is proportional to power, which is generally true for ZnO sputtering, the film thickness is propor tional to the product of sputtering power and time. Figure of Merit Analysis A new method of assessing the energy transparency of the window layer is presented here. Traditional optical analysis of a TCO is to measure the absorption coefficient at a single wavelength, 580 nm. TCOs for solar cell applications have much more loss in the UV near the bandgap cutoff and in the IR region due to free carrier absorption. Therefor a figure of merit transmission spectrum has little relevance for the true TCO performance in a solar cell application. A new absorption term is introduced that considers the losses at all wavelengths relative to CIGS: photon s above 1.2 eV energy in the following section . A bsorption is calculated from the reflection and the transmission: (4 1) Experimentally measured reflection was not available from the Lambda 900 setup. Therefore the reflection was estimated by theoretical means. Reflected light is det ermined by the Fresnel equations. At an incidence of 90 degrees, the Fresnel reflection for each layer is given by:

PAGE 97

97 (4 2) The total reflection is modeled by Fresnel reflection at interface: the air TCO interface (R1), the TCO glass interface (R2) and the backside glass air interface (R3). The index of refraction of air and water both have very low dispersion s [31] and are treated as constants 1.00 and 1.51, respectively. The disper sion of ZnO and MgZnO, however, is a strong function of wavelength. For undoped MgZnO, the dispersion relationship for samples at magnesium levels ranging from 0% to 30% measured and analyzed with optical ellipsometry was reported by Xi Jian [58] . First order Sellmeier coefficients were given at each magnesium level. The Sellmeier technique is a n empirical fitting method for the index of refractio n, which is expressed as: (4 3) Where A (unit less) and B (cm2) are the two fitting coefficients. One problem with the given dispersion relation was ZnO and MgZnO have an increase in bandgap and a decrease in the average index o f refraction as the bandgap increases due to the Moss Burnstein effect. The result of increased carrier levels on the index of refraction of ZnO:Al was reported by Hwang et al [59] . Samples from non conductive to highly conductive had an overall average range i n the index of refraction of approximately 0.07. The effect was used to correct the dispersion relationship of the Xi Jian data, where the coefficients A and B were varied according to the observed bandgap of the films. The dispersion relationship is shown in Figure 4 2 . The next step of the analysis is to consider the quarter wavelength interference of the transmission spectrum. The TCO glass reflection R2 interferes with the incoming beam of light. The result is the intensity of the transmitted light flu ctuates in an oscillatory pattern. The reflection function for R2 is given by the following expression:

PAGE 98

98 (4 4) Where the factor of negative 1results in the average of R 2 ( ) over the spectrum being equal to the average of R 2 ( ) t imes F(t, ); hence reconstructing the interference fringes in the reflectance without changing the amount of reflection that would be expected without interference fringes . Another aspect of the model is the variation of the amplitude of the interference f ringes. An example is shown for two separate transmission samples in Figure 4 3 . Sample A has low amplitude whereas sample B has high amplitude. The variation of amplitude is corrected as follows: (4 5) Where A F accounts for the overall amplitude and B F corrects for the decreasing amplitude with wavelength. A sample of the analysis is shown in Figure 4 4 . The first item plotted is the transmission. The second item is 1 minus R, where R is the calculated reflect ance. The space between T and 1 R is the light lost due to optical absorption. The third item plotted is 1 absorption, which is referre d to as T*, the remaining light after the optical absorption loss. The final part of the analysis is the calculation of the lumped energy loss coefficient. The T* term is used in the calculation. It is averaged over the entire spectrum for CIGS, weighted by the AM1.5 intensity as follows: (4 6) The result is an energy transmission of the TCO . The numerical figure is the fraction of energy that passes through the TCO. The next step is in preparation for calculating the figure of merit for the total spectrum energy as presented in the present work . The energy transmission fraction

PAGE 99

99 is converted to a n absorption coefficient of energy in the same manner an optical coefficient is calculated: (4 7) Where E is the coefficient and t is the film thickness. The figure of merit is d efined as the ratio of conductivity to the coefficient : (4 8) The coefficient is one of the key parameters used in the analysis . An additional parameter extracted from the optical absorption data is the optical bandgap. The method of graphic ally extracting the bandgap from data was first presented by David and Mott in 1970 [60] : (4 9) Where n is a parame ter chosen for best linear fit. In the present study n=1 allows for a very clear linear plot of the absorption data. An example is shown in Figure 4 5. Material Analysis Results The MgZnO dope with the three elements were sputtered over a range of powers and oxygen conditions. Measurements were taken for conductivity, transparency and powder XRD. Data analysis was used to extract various parameters, which are compared and discussed in the present section. Performance Merit The measured electrical resistivities of the three materials at multiple powers are plotted as a function of the oxygen partial pressure, as shown in Figure 4 6 . The gallium compound overall has an exceptional conductivity, approaching the 10 4 ohm cm ran ge. All of the curves occur at different positions, but have the same shape and proportions. Oxygen stoichiometry is

PAGE 100

100 the dominant factor, as small changes can alter the conductivity by an order of magnitude. The power level that results in the lowest condu ctivity is MgZnO:Ga at 50 watts. The scandium and aluminum have very similar conductivities that minimize around 3 x 10 3 ohm cm. One trend to note is that MgZnO:Al and Sc at 100 Watts have the same range of oxygen partial pressures as MgZnO:Ga at 50 Watts . Next the energy absorption coefficient is plotted as a function of the oxygen partial pressure in Figure 4 7. The trend is a rapid rate of increase over a small change of oxygen partial pressures. Films deposited with minimal resistivity showed a tendenc y for the optical transparency to quickly fall off with little improvement in conductivity. Therefore the best performing films were not necessarily ones with the minimal conductivity, but ones with the best combination of conductivity and transparency, as defined by the figure of merit. Another view of the optical effects is shown in Figure 4 8, where the energy absorption coefficient is plotted against the log of the conductivity. The far left shows that the most conductive samples quickly increase in ene rgy absorption with little marginal improvement in conduc tivity, indicating that the performance of the dopants reached a saturation of activation. The gallium and scandium films show the best transparency. The figure s of merit are plotted versus the oxyg en partial pressure in Figure 4 9. The shape of the trend appears to be a linear relationship of increasing figure of merit with decreasing oxygen until the peak. Gallium samples show the best figures of merit. Aluminum and scandium samples had very simila r figures. The figure of merit is plotted versus the resistivity in Figure 4 10. Low resistivity was important for high figures of merit. Since gallium doping resulted in the lowest resistivities and the highest figures of merit. As the resistivity was dec reased, the figure of merit followed the black trend line until a point of saturation where the figure began to decrease

PAGE 101

101 with further lowering of resistivity. In figure 4 11, the figure of merit is plotted versus t he energy absorption coefficient. The lowe st energy absorption coefficient of all was 500 cm 1, which corresponded to approximately 7% energy lost. The best figure of merit films had energy coefficients of 1000 cm 1, or about 14% energy lost. The optical bandgaps as a function of the oxygen parti al pressure are plotted in Figure 4 12. There is an almost linear correlation with the oxygen partial pressure. The relationship between oxygen, conductivity, carrier concentration and the bandgap (Moss Bernstein shift) indicates that for each power and do pant there is almost a constant rate of change of the electrical properties with oxygen. The resistivity is plotted against the bandgap energy in Figure 4 13. The straight line trend shows a strong correlation, confirming that the bandgap data displays Mos s Bernstein behavior. The energy absorption coefficient is plotted in Figure 4 14 versus the bandgap. There is a saturation type behavior of a maximum reachable bandgap through oxygen deficiency, where further reduced oxygen simply makes the coefficient of absorption go up without improving the electrical or optical properties. The figure of merit is plotted versus the bandgap in Figure 4 15. The highest figures of merit occur at the highest bandgaps. The relationship is very similar to that of Figure 4 10. The figure of merits follow the black line trend with conductivity increasing without much loss of optical transmission. Then at a critical saturation point the optical transmission falls off. Crystal Properties Powder XRD indicated that the films were si ngle phase with a single peak for the 002 oriented textures , as shown in Figure 4 16 . The first thing to note for the MgZnO:Al is that there are very few minority peaks. MgZnO is a less stable compound as the amount of magnesium increases. The absence of o ther peaks indicates that there is not only no other phase such as MgO, but that the phase is very well textured . The powder XRD was r un for samples under the

PAGE 102

102 same power but varying oxygen so that the conductivity varied by 2 orders of magnitude . There is not an obvious change in the orientation or the appearance of any secondary phase . If conductivity loss with increasing oxygen was due to dopants being deactivated through physical displacement from the lattice sites the expectation would be to see additional peaks due to a second phase formation such as Al 2 O 3 . It is very unlikely that such a large molecule could act as an interstitial. If Al 2 O 3 were present as a second phase, it would be expected to show up on the XRD . There is no evidence on the scans. Therefore a more likely explanation is that the dopants are being compensated while remaining on site. For onsite compensation the dopants would neighbor interstitial oxygen, which adsorbs the free electron. In Figure 4 17 the grain size is showing a decreasing tendency with the amount of oxygen. This effect was also reported [ 46] , although in the present work there is not a strong correlation of the grain size with any other properties such as sputtering power, dopant identity, the figure of merit or the electrical conductivity. There was a stronger correlation with the opti cal and electrical properties with the c axis length. The c axis is plotted versus the oxygen partial pressure in Figure 4 18 . The c axis lengths are lightly lower than the pure bulk ZnO by about 0.5200 nm, which is consistent with other works [49] . Early reports on MgZnO indicated the same trend of c axis compr ession [49] . The findings were that with increasing Mg i ncorporation, the c axis became shorter and the a axis became longer by one or two percent. The cell volume is somewhat constant over the range of Mg composition. The hexagonal unit cell is getting shorter and fatter. The effective ionic radius of magnesiu m is 0.72 and of zinc is 0.74. The oxygen zinc oxygen and the oxygen magnesium oxygen bondings line up in the vertical direction of the lattice. The magnesium part is slightly

PAGE 103

103 shorter than the zinc part, and therefore pulls on the vertical direction. The s tretching of the magnesium bond and the equal squeezing of the zinc bond causes the zinc to be pushed outward. The lattice parameter shows an increases trend as the oxygen level is increased. The same effect is observed in other reports of ZnO by reactive magnetron sputtering [46] . The expected effect of oxygen interstitials is to expand the total cell volume . One soft trend is observed when the data clusters at different powers. The c axis length is increasing slightly with the decrease of the power. The lower power films do not appear to lack electrical conductivity relative to the high power films. Although M gZnO should have a c axis about 1% lower than that of bulk ZnO, a possibility is that there is an unknown relaxation mechanism compensating the slight instability of magnesium incorporation. Bandgap data is plotted versus the c axis length in Figure 4 19. There is a linear trend of increasing bandgap with decreasing oxygen. Considering the Moss bernstein relationship, there is a relationship between c axis length and conductivity, which is also demonstrated in Figure 4 20. The explanation is oxygen compensa tion is pushing the vertical bonds upwards and lengthening. In Figure 4 21, the plot of the figure of merit with c axis show a strong correlation, further supporting the relationship between c axis and oxidation state. Discharge Voltage Characteristics The sputtering discharge voltage curve reveals important information on the surface condition of the target as well as effects of negative ion resputtering. Electrons normally originate in the plasma due to the ionization of argon atoms and, to a lesser exten t, oxygen molecules. Another aspect of the DC discharge current is ion induced secondary electrons (IISEs) from Ar+ and O2+ impacts. For this reason the target voltage is very sensitive to the surface properties of the target. Metals often have different I oxides, which is reflected in the target discharge characteristics. However, experimental measurement s of

PAGE 104

104 2008 book Modeling of the Magnetro n Discharge [61] . One of the lesser studied aspects of surface discharge characteristics is the contribution of n egative oxygen ion s to the total current. In chapter 3 strong evidence of negative ion resputtering was observed. If the target surface is not very oxidized there will not be negative ions generated. However, if it i s oxidized there will be more negative ions, which will effectively increase the current and cause the controller to drive down the voltage. The present section is an analysis of the deposition rates and discharge date to determine the effects. Voltage dis charge characteristic curves are shown in Figure 4 22 for the three dopants at different power levels. One thing that is very clear is the dopants are influencing the discharge curves positions that is, the saddle point. But they appear to have the same sh ape and proportion. Another feature of the discharge curve for magnesium to explain is the upward slope after the saddle point. The trend of the increasing voltage with oxidation beyond the critical oxygen point can be explained by the lower sputtering yie ld of oxygen relative to argon. Since there is a constant total flow of gas, increasing oxygen flow corresponds to a decreasing argon flow. The gallium and scandium are both at higher voltages than the aluminum. The gallium has a higher processing window o f oxygen while the aluminum and scandium have about the same. All the processing windows have approximately the same separation from the saddle points. The atomic mass of aluminum, scandium and gallium are 26.98, 44.96 and 69.72. By the rational of more ne gative ions for lighter elements, it turns out that the curves go in that order. Gallium and scandium are very close together while aluminum is much lower.

PAGE 105

105 Figure 4 1. The Air Mass 1.5 spectrum [62] with the CIGS sensitive regions. Figure 4 2. The Sellmeier fitting of MgZnO dispersion relationship from MgZnO data [58] and ZnO:Al bandgap shif t [59] .

PAGE 106

106 Figure 4 3. Optical transmission data. A. MgZnO:Sc at 50 W atts for 5 min utes and 0.126 mT Oxygen. B. MgZnO:Al at 100 W atts for 2.5 min utes and 0.20 mT O xy Figure 4 4. Example of fitting optical transmission data with the reflection calculation model.

PAGE 107

107 Figure 4 5. Example of a linear extraction of the optical band gap from absorption data. Figure 4 6. The electrical resistivity as a function of the oxygen partial pressure.

PAGE 108

108 Figure 4 7. The energy absorption coefficient as a function of t he oxygen partial pressure. Figure 4 8. The energy absorption coefficient plotted versus the oxygen partial pressure.

PAGE 109

109 Figure 4 9. The figure of merit plotted versus the oxygen partial pressure. Figure 4 10. The figure of merit plotted versus the electrical resistivity.

PAGE 110

110 Figure 4 11. The figure of merit plotted versus the energy absorption coefficient. Figure 4 12. The optical band gap potted versus the oxygen partial pressure.

PAGE 111

111 Figure 4 13. The electrical resistivity plotted versus the optical band gap. Figure 4 14. The energy absorption coefficient plotted versus the optical band gap.

PAGE 112

112 Figure 4 1 5 . The figure of merit plotted versus the optical band gap. Figure 4 16. The powder XRD scans of MgZnO:Al at different oxygen partial pressures.

PAGE 113

113 Figure 4 17. The grain size plotted versus the oxygen partial pressure. Figure 4 18. The c axis length plotted versus the oxygen partial pressure.

PAGE 114

114 Figure 4 19. The optical band gap plotted versus the c a xis length. Figure 4 2 0 . The electrical resistivity potted versus the c axis length.

PAGE 115

115 Figure 4 2 1 . The figure of merit plotted versus the c axis length. Figure 4 22. The DC voltage discharge characteristics versus the oxygen partial pressure.

PAGE 116

116 Figure 4 2 3 . The deposition rate versus the oxygen partial pressure.

PAGE 117

117 CHAPTER 5 CIGS DEVIC E S WITH ZNO AND MGZNO The goals of the device tests are to assess the performance characteristics of the TCO films from the third and fourth chapters. I n the thir d chapter a high quality ZnO:Al film was sought without the use of heating or post deposition annealing. The f irst goal of the present chapter was to demonstrate the performance characteristics of the ZnO:Al on a complete device . In the fourth chapter MgZn O was investigated as a TCO that can be reactively sputtered with DC power to give a reliable and stable film . The second goal of the present chapter was to realized the wider bandgap effect of MgZnO:Al on the performance of CIGS . The effects of iMgZnO in comparison to iZnO have been tested on a CIGS device for the purpose of reducing interface recombination [63] , as well as interfa ce effects occurring with iZnMgO/CdS [64] . However, to the best of our knowledge doped MgZnO has not been used as a TCO material on CIGS. Experiment The CIGS devices were manufactured by an industrial collaborator that had been partially completed with the device fabri cation being terminated at the CdS level. The raw sheet of CIGS being as close to one another as possible to minimize possible nonuniformities or variations that oc cur over large area sheets. A resistive layer was first deposited on the device. It consisted of resistive ZnO for the ZnO TCOs and resi stive MgZnO for the MgZnO TCOs. The oxygen partial pressure was boosted by about 40% , r elative to the conductive films, to achieve sheet resistances on the order of 10 6 / square. The layers were approximately 70 nm thick . The conductive layer was applied to a thickness giving approximately 60 +/ 10 / square. The ZnO:Al was deposited at 60 Watts and 0.18 m torr . A time of 4 minutes gave a film thickness of 150 nm

PAGE 118

118 under 0.13 mto rr O2 to give a film thickness of 250 nm. High Power MgZnO:Al on CIGS The first set of experiments was a side by side comparison of MgZnO :Al to ZnO :Al . A total of four experimental runs are included, where a single experiment consists two CIGS samples with one having a ZnO TCO and another with a MgZnO TCO. Each sample was metalized and cut to have two devices, Cell 1 and Cell 2 corresponding to the position, as illustrated in Figure 5 1. The resulting IV data was organized by the cell and averaged: there wa s a ZnO Cell 1 and Cell 2 average as well as a MgZnO Cell 1 and Cell 2 average . The trend is plotted Figure 5 2, where the solid line represents the average and the dashed line represents the standard deviation . The standard deviation of each data point is calculated and plotted along the side of the average curve. The performance parameters are included in the figure. The device performances are overall very similar MgZnO samples slightly more efficient by about 0.4 absolute percent . Efficiency is related to the other parameters by the following equation: (5 1) The two devices are very similar in the short circuit current, suggesting there is not a difference in the photocurrent collection efficiency , which was one of the t heoretical possibilities of a wider bandgap TCO. T he open circuit voltage the MgZnO devices are lower by 4 to 5 mV , which is opposite to the trend of efficiency difference . For the fill factor, however, t here is a sharp improvement in the MgZnO over the ZnO that is outside of the indicated standard deviation . The next step was to measure the quantum efficiency. The full range of the QE was measured for the same set of CIGS devices and averaged, as shown in Figure 5 3 . The curves for

PAGE 119

119 ZnO and MgZnO have ver y similar shapes with a photon collection rate for MgZnO approximately 4% lower than to ZnO for the 800 to 1000 nm region. The visible region shows the best collection efficiency in the case of both devices, about 87% efficient. The loss is most likely opt ical reflection, as there is no antireflective layer. At 500 nm there is a sharp fall off of the efficiency due to the absorption in the CdS region. Beyond the efficiency drops below 40%. A closer look was taken at the UV range of the QE data is shown in F igure 5 4 reveals a higher efficiency of the MgZnO device from 375 to 350 nm. Low Power MgZnO:Ga on CIGS The first set of experiments show that MgZnO:Al is an effective TCO for CIGS. It is considered to be a standard material relative to the alternately d oped ones. Gallium doping showed better conductivity. Furthermore the material showed significantly improved transparency at lower sputtering powers without a loss in electric conductivity . There are two data sets which each consist of one ZnO and one MgZn O sample. They are labeled Set A and Set B. The JV results and extracted parameters are shown in Figure 5 5. The A set consists of a very healthy JV curve for both the ZnO and the MgZnO devices. There is a flat top portion, a sharp knee and a steep slope. In set A there is a devices have high efficiencies that overall range from 12.5 to 13.2%. There is not much distinction in the fill factors, the open circuit voltage and the short circuit current between the ZnO and the MgZnO samples. One of the samples of MgZnO has slightly less short circuit current than the other MgZnO samples and the two ZnO samples. Since the fill factor of the curve is healthy, the difference is possibly due to an error in the active area of the device. The B set appears to have a rel atively less healthy JV curve. There is a flat top portion and a steep right portion, but the knee is broadened for both the ZnO and the MgZnO, with the ZnO being relatively less broadened. The shape of the curve is reflected in the efficiency and fill

PAGE 120

120 fac tor parameters, which averages to about 54% for the MgZnO devices and 57% for the ZnO devices. The effect shows up in the efficiencies relative to set A. The ZnO samples are about 11% efficient whereas the MgZnO samples are about 10% efficient. The open ci rcuit voltages are healthy for the set B compared to A. The short circuit voltages are lower by about 1 mA for the set B samples relative to the set A. The reduced device quality could possibly be a series resistance effect. However, the conditions of the TCO were the same for set A and set B. The TCOs were measured at 60 ohms/square. Its possibly a shunt current effect. However, the top of the curve is flat. More than likely the result is an indication of a device with more recombination. The QE results for sets A and B are shown in Figure 5 6. In set A it appears that the MgZnO is less efficient in the IR range. I n the visible range the efficiency of the two material devices is very similar. The efficiency peaks close to 90%, corresponding with the transparency figure for the TCO on glass. In set B the efficiency of the ZnO and MgZnO materials are similar over the entire range with the oscillation pattern being that of the quarter wavelength effect. The UV region of the QE is shown in Figur e 5 7. It is evident that the more transparent and wider bandgap MgZnO:Ga at low power is giving a better performance in the region than the traditional ZnO:Al. Discussion For the MgZnO:Al test the fill factor difference could be a result of a lower series resistance in the MgZnO device. Since most series resistance is expected to be in the TCO and the TCOs of the comparison devices were matched, the difference in fill factor should not be for the reason. It is most likely that the deposition process is cre ating a change of some type with the thin CdS buffer layer, which is what the TCO is being deposited on top of. There is a much higher power density in the MgZnO:Al than with the ZnO:Al.

PAGE 121

121 For the MgZnO devices with a 4 mV lower open circuit voltage, differe nce is likely related to an electronic reason. The workfunction of MgZnO is higher than that of ZnO [65] . The shape of the IV curve at the knee is a r esult of the bias promoting cold current diffusion from the n side and competing with the forward current. The reservoir of cold electrons on the TCO side is at a higher energy in the MgZnO device . As a result the onset of diffusion is expected to happen a t a lower bias level. The efficiency of the devices shows a marginal improvement in the UV region. Integrated the improvement contributes to an additional 0.01% absolute efficiency. The limitation is in the CdS layer. Future devices that do not have the l imitation will be able to capture up to 5% additional energy. For CIGS a 2% absolute efficiency is expected.

PAGE 122

122 Figure 5 1. The device finishing diagram and label of individual cells . Figure 5 2. The IV data for ZnO:Al and MgZnO:Al averaged over four experiments . A) Devices from the first cell position. B) Devices from the second cell position.

PAGE 123

123 Figure 5 3. The QE of the ZnO:Al and MgZnO:Al devices averaged over four experiments. A) Devices from the first cell position. B) Device s from the second cell position . Figure 5 4. The UV region of the QE data. A. Devices from the first cell position. B. Devices from the second cell position.

PAGE 124

124 Figure 5 5. The JV data and extracted parame ters for ZnO:Al and MgZnO:Ga. A. Set A. B. Set B . Figure 5 6. The QE data for ZnO:Al and MgZnO:Ga. A. Set A. B. Set B.

PAGE 125

125 Figure 5 7. The UV regions of the QE data averaged . A. Device set A. B. Device set B .

PAGE 126

126 CHAPTER 6 CONCLUSIONS AND FUTURE WORK Conclusions MgZnO is a promising substitute for ZnO, with improved transparency, and TCO with earth element abundance and minimal toxicity. The dopant gallium is a likely means to solve the reduced electrical conductivity of MgZnO relative to ZnO. Future CIGS devices with CdS free buffer layers could realize the full improvement of the UV spectrum below 380 nm. The presented FOM technique is an effective way to co mpare absolute TCO performance, for CIGS applications, from a variety of compositions and sputtering condition s. The total TCO loss analysis method is a simplified alternative to the traditional and rigorous full scale device simulation. The graded bandgap TCO offers a reduction in TCO associated loss in CIGS and could save large quantities of indium for ITO compl eted CIGS. Adaption of the FOM technique combined with a f ull scale 2D Medici simulation of a CIGS device resulted in an effective analytical model to predict the optimal TCO thickness. A novel TCO structure results in approximately 15% less loss in the TC O. Future Work MgZnO film quality decrease with increased deposition rates should be studied for determining if the effect is due to an effect of the specifics of the system or due to a fundamental aspect of the rate of the film growth. Defect chemistry i s very important to the performance of the TCO. The samples should be tested by photoluminescence to quantify the defects and develop a model. The samples should be analyzed by cross sectional TEM to determine the effects of defects, dopants and power on t he microstructure and texture evolution. The graded TCO should be deposited on full CIGS devices in comparison with flat TCO CIGS devices to

PAGE 127

127 validate the theoretical improvement in cell efficiency through the reduction of overall TCO associated loss.

PAGE 128

128 LIST OF REFERENCES [1] New York Times , 10 May 2013. [2] C. Parmesan and G. Yohe Nature , vol. 421, no. 6918, pp. 37 42, Jan. 2003. [3] C. D. Thomas, A. Cameron, R. E. Green, M. Bakkenes, L. J. Beaumont, Y. C. Collingham, B. F. N. Erasmus, M. F. de Siq ueira, A. Grainger, L. Hannah, L. Hughes, B. Huntley, A. S. van Jaarsveld, G. F. Midgley, L. Miles, M. A. Ortega Huerta, A. Townsend Peterson, O. L. Nature , vol. 427, no. 6970, pp. 145 14 8, Jan. 2004. [4] Everyday Sci. Mech. , p. 26, 1931. [5] [6] n Junction Photocell for J. Appl. Phys. , vol. 25, no. 5, pp. 676 677, 1954. [7] lar Power Doubled in 2013, With Policy Battles on the Horizon for 2014. [Accessed: 30 Jan 2014]. [8] [9] http:// www.pv tech.org/news/has_first_solar_retaken_the_lowest_cost_pv_manufacturer_mantle. [Accessed: 30 Jan 2014]. [10] Available: http://www.bloomberg.com/news/2013 04 10/first solar will target japanese rooftops with tetrasun purchase.html. [Accessed: 30 Jan 2014]. [11] P. Jackson, D. Hariskos, E. Lotter, S. Paetel, R. Wuerz, R. Menner, W. Wischmann, and M. film solar cells beyond Prog. Photovolt. Res. Appl. , vol. 19, no. 7, pp. 894 897, 2011. [12] efficiency cadmium free Cu(In,Ga)Se2 thin Electron Devices IE EE Trans. On , vol. 46, no. 10, pp. 2093 2097, Oct. 1999. [13] Appl. Phys. A , vol. 57, no. 1, pp. 19 24, Jul. 1993.

PAGE 129

129 [14] J. S. Ward, K. Ramanathan, F. S. Hasoon, T. J . Coutts, J. Keane, M. A. Contreras, T. Prog. Photovolt. Res. Appl. , vol. 10, no. 1, pp. 41 46, 2002. [15] J. Appl. Phys. , vol. 47, no. 9, pp. 4086 4089, 1976. [16] MRS , pp. 52 57, 2000. [17] properties of tr ansparent conductive oxide ZnO:Al films prepared by dc magnetron reactive J. Vac. Sci. Technol. Vac. Surf. Films , vol. 19, no. 3, pp. 963 970, May 2001. [18] S. Ishizuka, K. Sakurai, A. Yamada, K. Matsubara, P. Fons, K. Iwata, S. Nakamura, Y. gap highly resistive i Int. Conf. Phys. Chem. Eng. Int. Conf. Phys. Chem. Eng. , vol. 87, no. 1 4, pp. 541 548, May 2005. [19] Appl. Phys. Lett. , vol. 72, no. 25, pp. 3261 3263, 1998. [20] Y. Nano Lett. , vol. 8, no. 5, pp. 1501 1505, Apr. 2008. [21] J. Song, S. S. Li, C. H. Huang, O Solid State Electron. , vol. 48, no. 1, pp. 73 79, Jan. 2004. [22] M. A. Contreras, B. Egaas, K. Ramanathan, J. Hiltner, A. Swartzlander, F. Hasoon, and R. film solar Prog. Photovolt. Res. Appl. , vol. 7, no. 4, pp. 311 316, 1999. [23] ts to improve J. Appl. Phys. , vol. 95, no. 4, pp. 1911 1917, Feb. 2004. [24] function of Solid State Electron. , vol. 20, no. 7, pp. 629 634, Jul. 1977. [25] Solid State Electron. , vol. 19, no. 6, pp. 471 472, Jun. 1976.

PAGE 130

130 [26] Solid State Electron. , vol. 10, no. 8, pp. 765 775, Aug. 1967. [27] Sol. Energy Mater. Sol. Cells , vol. 93 , no. 9, pp. 1499 1506, Sep. 2009. [28] D modeling of thin film Sol. Energy Mater. Sol. Cells , vol. 93, no. 3, pp. 350 354, Mar. 2009. [29] MRS Bull. , vol. 25, no. 08, pp. 38 44, 2000. [30] J. Phys. Appl. Phys. , vol. 34, no. 21, p. 3097, 2001. [ 31] modulation Appl. Phys. Lett. , vol. 33, no. 7, pp. 665 667, 1978. [32] A. Ohtomo, R. Shiroki, I. Ohkubo, H. Koinuma, and f supersaturated MgxZnxO alloy films and MgxZn1 [33] doped ZnO J. Appl . Phys. , vol. 98, no. 5, p. 053705, 2005. [34] State Electron Transport and Low Field Mobility of Wurtzite Bulk ZnO and Zn x Mg x J. Electron. Mater. , vol. 40, no. 4, pp. 466 472 472, Apr. 2011. [35] Chih I Huang, Huai An Chin, Yuh Renn Wu, I Chun Cheng, J. Z. Chen, Kuo Chuang Chiu, and Tzer Electron Devices IEEE Trans. On , vol. 57, no. 3, pp. 696 703, 2010. [36 ] [37] xMgxO transparent conducting thi n films fabricated by multi cathode RF magnetron Thin Solid Films , vol. 518, no. 11, pp. 2949 2952, Mar. 2010. [38] transparent conducting oxides with modulatable band Solid State Commun. , vol. 151, no. 3, pp. 264 267, Feb. 2011.

PAGE 131

131 [39] metal organic ch J. Electron. Mater. , vol. 35, pp. 1680 1684, 2006. [40] Appl. Phys. Lett. , v ol. 75, no. 7, pp. 980 982, 1999. [41] Solid State Circuits IEEE J. Of , vol. 13, no. 1, pp. 180 187, Feb. 1978. [42] rate Temperature Dependence of Transparent Conducting Al Jpn. J. Appl. Phys. , vol. 31, no. Part 2, No. 3A, p. L257, 1992. [43] tructural and J. Appl. Phys. , vol. 80, no. 2, pp. 1063 1073, 1996. [44] Annu. Rev. Mater. Sci. , vol. 7, no. 1, pp. 239 260, Aug. 1977. [45] K. Tominaga, T. Yua Jpn. J. Appl. Phys. , vol. 24, no. Part 1, No. 8, p. 944, 1985. [46] K. Ellmer, F. Kudella, R. Mientus, R. Schieck, and S. Fiechter Thin Solid Films , vol. 247, no. 1, pp. 15 23, Jul. 1994. [47] N. Yamamoto, H. Makino, S. Osone, A. Ujihara, T. Ito, H. Hokari, T. Maruyama, and T. Y doped ZnO transparent electrodes for liquid crystal 8th Int. Conf. Coat. Glass Plast. ICCG8 , vol. 520, no. 12, pp. 4131 4138, Apr. 2012. [48] H. Kim, C. M. Gilmore, A. Pique, J. S. Horwitz, H. Mattoussi, H. Mu rata, Z. H. Kafafi, and -tin -oxide thin films for organic light J. Appl. Phys. , vol. 86, no. 11, pp. 6451 6461, Dec. 1999. [49] A. Ohtomo, M. Kawasaki, T. Koida, K . Masubuchi, H. Koinuma, Y. Sakurai, Y. Yoshida, x]O as a II -VI widegap semiconductor Appl. Phys. Lett. , vol. 72, no. 19, pp. 2466 2468, May 1998.

PAGE 132

132 [50] n of Mg composition in Mg[sub x]Zn[sub 1 J. Vac. Sci. Technol. B Microelectron. Nanometer Struct. , vol. 23, no. 5, pp. 2030 2033, 2005. [51] D. Cohen, K. Ruthe, and S. Barne xMgxO:(Al,In ) thin [52] Jpn. J. Appl. Phys. , vol. 24, no. Part 2, No. 10, p. L781, 1985. [53] T. Minami, T. Yamamoto, Thin Solid Films , vol. 366, no. 1 2, pp. 63 68, May 2000. [54] P. Amzaga Madrid, W. Antnez Flores, J. E. Ledezma Sillas, J. G. Murillo Ramr ez, O. Sols Canto, O. E. Vega Becerra, R. Martnez Snchez, and M. Miki microstructural characterization and optical properties of undoped, V and Sc doped ZnO J. Alloys Compd. , vol. In Press, Corrected Proof. [55] I. Sieber, N. Wanderka, I. Urban, I. Dörfel, E. Schierhorn, F. Fenske, and W. Fuhs, Thin Solid Films , vol. 330, no. 2, pp. 108 113, Sep. 1998. [56] X. Yu, J. M Appl. Surf. Sci. , vol. 245, no. 1 4, pp. 310 315, May 2005. [57] X. Yu, J. Ma, F. Ji on the properties of ZnO:Ga films deposited by r.f. magnetron sputtering at low J. Cryst. Growth , vol. 274, no. 3 4, pp. 474 479, Feb. 2005. [58] Zhang Xi Jian and Ma Hong Lei and Li Yu Xiang and Wang Qing Pu and Ma Jin and Zong Fu Jian and Xiao Hong Chin. Phys. , vol. 15, no. 10, p. 2385, 2006. [59] es of post annealed Proc. IFFM2011 2011 Int. Forum Funct. Mater. AFM 2 2nd Spec. Symp. Adv. Funct. Mater. , vol. 47, no. 10, pp. 2898 2901, Oct. 2012. [60] E. A. David and N. F. Mott, Philos. Mag. , v ol. 22, p. 903, 1970. [61] Reactive Sputter Deposition , vol. 109, D. Depla and S. Mahieu, Eds. Springer Berlin Heidelberg, 2008, pp. 61 130.

PAGE 133

133 [62] Materials, ASTM G173 03, Jan. 2003. [63] Photovoltaic Specia lists Conference (PVSC), 2009 34th IEEE , 2009, pp. 000127 000130. [64] Appl. Phys. Lett. , vol. 87, no. 3, p. 032101, 2005. [65] A. Fallisch D SPICE Simulation and Analytical Calculation of Spreading Resistance Effects in Emitter Wrap Through Cells With Nonsquare Via Photovolt. IEEE J. Of , vol. 1, no. 2, pp. 153 158, 2011.

PAGE 134

134 BIOGRAPHICAL SKETCH Barrett Hicks g rew up in the City of Birmingham in Jefferson County, Alabama. He did h is B achelor of Science in c hemical e ngineering at the University of Auburn. In the Fall of 2007 he attended the c hemical e ngineering graduate p rogram at the University of Florida to pur sue his Doctor of Philosophy d egree. His research interest is photovoltaics and alternative energy. He works primarily with physical vapor deposition . In his spare ti me he likes exercise, building custom outdoor recreation gear and photography .