|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
COPPER GALLIUM DISELENIDE THINT FILM AB SORBER GROWTH
FOR SOLAR CELL DEVICE FABRICATION
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
O 2007 Ryan Kaczynski
To my family I love you
First, I would like to thank Prof. A. Brad Anton at Cornell University for encouraging me
to pursue my doctorate degree when I had no idea what I wanted to do in the future. I express
my sincerest gratitude to Dr. Oscar Crisalle for taking me under his guidance. It has been a
pleasure working for him. His relaxed attitude has been very beneficial to our working
I would like to acknowledge the many members of the CIS solar cell team at the
University of Florida: Dr. Timothy Anderson and Dr. Sheng Li for your expertise in the solar cell
Hield and bringing these many excellent graduate students together, Billy Stanbery for starting
this proj ect, Serkan Kincal for training me on the PMEE reactor, Suku Kim for training me on
fi1m deposition, Ryan Acher for helping with reactor maintenance (by far the toughest j ob) and
fi1m characterization, Woo Kyoung Kim and Seokhyun Yoon for growth support, Jiyon Song
and Xuege Wang for device characterization, and Andre Baran and Wei Liu for device
fabrication. Each one of you was very integral to my success on this project. I would also like to
recognize the staff at MICROFABRITECH, especially Scott Gapinski and Diane Badylak.
TABLE OF CONTENTS
ACKNOWLEDGMENT S .............. ...............4.....
LI ST OF T ABLE S ........._... ......___ ...............8....
LIST OF FIGURES .............. ...............12....
AB S TRAC T ............._. .......... ..............._ 15...
1 INTRODUCTION ................. ...............17.......... ......
System Description............... ..............1
P MEE Reactor ................. ................. 18..............
Chamber .............. ...............20....
Chalcogen Zone ................. ...............21.................
Heater Zone .............. ...............22....
Metal s Z one ................. ...............22.......... ......
Control ................ ...............23.......... ......
Problem Statement ................. ...............24.................
2 SOLAR CELLS .............. ...............26....
Energy ................. ...............26.......... ......
Sunlight ................. ...............26.................
C osts .............. .. ...............28...
Photovoltaic Systems............... ...............29
Future ................. ...............30.................
Solar Cell History ................. ...............31.......... ....
Solar Cell Device Physics............... ...............32
Band Gap .............. ...............3 2....
Electric Field .............. ...............33....
Defects ................. ...............34.................
Thin Films............... .... .. .... ..........3
Direct vs Indirect Band Gap ................ ...............36...............
Ab sorption Length ................. ...............36........... ....
Diffusion Length .............. ...............36....
CulnSe2-Based Solar Cells .............. ...............37....
Material Properties .............. ...............37....
Defects ................. ...............39.................
Gallium Addition............... ...............39
Defects ................. ...............41........._.....
Type inversion ......_. ................ .......__. ..........4
Sulfide-based Chalcopyrites ................ ...............43.................
Deposition Processes .............. ...............44....
M ultij uncti ons ................. ........... ...............45......
Theoretical Multijunctions .............. ...............46....
Tandem Structure .............. ...............47....
Monolithic vs Mechanical .............. ...............48....
3 ABSORBER GROWTH AND DEVICE FABRICATION .................. ................5
Growth Calibration .............. ...............52....
Standard Growth Procedure............... ...............5
Growth Schemes ................. ...............56.................
Absorber Characterization ................. ...............59.................
ICP ................ ...............59.......... ......
SE M ................ ...............60.......... ......
X RD ................ ...............60.......... ......
Device Fabrication..................... ...........6
Substrate and Back Contact............... ...............61
Post Absorber Deposition ................. ...............63........... ....
Buffer Layer .............. ...............63....
Alternative Buffers .............. ...............65....
W indow Layers .............. ...............66....
M etal liz ati on ................. ...............67........... ...
Anti-Reflective Coating............... ...............67
Device Characterization............... ............6
Current-Voltage ................. ...............67.................
I-V Measurement Technique ................. ...............69........... ....
Quantum Efficiency............... ...............6
QE Measurement Technique .............. ...............70....
4 COPPER GALLIUM DISELENIDE AB SORBER GROWTH.............__ .........___.......73
Growth M atrix .............. ...............73....
Absorber Characterization ............. ...... .__ ...............82...
5 COPPER GALLIUM DISELENIDE DEVICE FABRICATION............. ............._ ..120
Best Devices in the Literature................ ..............12
Device Fabrication............... ..............12
Device Characterization............... ...........12
6 CIGS ABSORBER GROWTH AND DEVICE FABRICATION ................... ...............13
Best Devices in the Literature................ ..............13
Growth M atrix .............. ...............134....
Absorber Characterization ................. ...............138................
Ori entati on ........._.__....... .__ ...............138...
M orphology ................. ...............140......... ......
Device Fabrication............... ..............14
Device Characterization............... ...........14
7 DYNAMIC REACTOR MODEL .............. ...............150....
Flux M odeling .............. ...............150....
PMEE Reactor Modeling ................. ...............152................
8 CONCLUSIONS AND FUTURE WORK ................. ...............159........... ...
Future Work............... ...............160.
A GROWTH RUN DATA ................ ...............161...............
B REACTOR MODEL .............. ...............236....
InputPM EE.m .............. ...............236....
PMEE model.m ................. ...............238................
LIST OF REFERENCES ................. ...............243................
BIOGRAPHICAL SKETCH .............. ...............250....
LIST OF TABLES
2-1. Efficiencies of copper chalcopyrites. .............. ...............50....
4-1. First CGS growth series. .............. ...............91....
4-2. Second CGS growth series............... ...............91.
4-3. Third CGS growth series. ............. ...............91.....
4-4. Fourth CGS growth series............... ...............91.
4-5. Fifth CGS growth series............... ...............92.
4-6. Sixth CGS growth series ................. ...............92........... ...
4-7. Seventh CGS Growth Series ................. ...............92........... ...
4-8. Eighth CGS growth series............... ...............93.
5-1. Device parameters of record CGS cells produced at NREL............._ .........._ ..... 130
5-2. Device parameters for the second CGS absorber growth series. .................. ...............130
5-3. Device parameters for the fourth CGS absorber growth series. .................. ...............130
5-4. Device parameters for the fifth CGS absorber growth series. ............. .....................13
5-5. Device parameters for the sixth CGS absorber growth series. ............. .....................13
5-6. Device parameters for the seventh CGS absorber growth series. .........._... ........._.....13 1
5-7. Device parameters for the eighth CGS absorber growth series. .................. ...............131
6-1. CIGS growth series. .............. ...............144....
6-2. Device parameters for the CIGS growth series............... ...............148
A-1. Reactor conditions for Growth Run #443 ....._ .....___ ........__ ...........6
A-2. Reactor conditions for Growth Run #444 ................. ...............163........... ..
A-3. Reactor conditions for Growth Run #445 ................ ...............164......___..
A-4. Reactor conditions for Growth Run #446 ................. ...............165........... ..
A-5. Reactor conditions for Growth Run #447 ................. ...............166........... ..
A-6. Reactor conditions for Growth Run #452 ................. ............_ .....167.........
A-7. Reactor conditions for Growth Run #453 .....___.....__.___ .......____ ...........6
A-8. Reactor conditions for Growth Run #454 ................. ............_ .....169.........
A-9. Reactor conditions for Growth Run #455 ...._. ......_._._ .......__. ...........7
A-10. Reactor conditions for Growth Run #456 ................. ............_ .....171.........
A1.Reactor conditions for Growth Run #457 ................. ............_ .....172.........
A-12. Reactor conditions for Growth Run #458 ................. ............_ .....173.........
A-13. Reactor conditions for Growth Run #459 ................. ............_ .....174.........
A-14. Reactor conditions for Growth Run #472 ................. ............_ .....175.........
A-15. Reactor conditions for Growth Run #474 ................. ............_ .....176.........
A-16. Reactor conditions for Growth Run #475 ....__. ...._.._.._ ......._.... ...........7
A-17. Reactor conditions for Growth Run #476 ................. ............_ .....178.........
A-18. Reactor conditions for Growth Run #477 ................. ............_ .....179.........
A-19. Reactor conditions for Growth Run #478 ................. ............_ .....180.........
A-20. Reactor conditions for Growth Run #479 ................. ............_ .....181.........
A-21. Reactor conditions for Growth Run #480. ................ ...._.._ .....__.. .......18
A-22. Reactor conditions for Growth Run #510. ................ ...._.._ ...............183.
A-23. Reactor conditions for Growth Run #511 .........._._ ... ........ ....._.... ......18
A-24. Reactor conditions for Growth Run #512. ................. ....._.. ............ ......18
A-25. Reactor conditions for Growth Run #513 .........._._ ... ........ ....._.._..........8
A-26. Reactor conditions for Growth Run #514 ................. ..........._._.....187...... ..
A-27. Reactor conditions for Growth Run #515 ....._.. ................ .. ........ ...... ... 18
A-28. Reactor conditions for Growth Run #516. ................. ....._.. ............ ......18
A-29. Reactor conditions for Growth Run #521 .........._._ ........... ......_.. ........19
A-30. Reactor conditions for Growth Run #522 ................. ............_ .....191.........
A-31. Reactor conditions for Growth Run #523 ................ ...............192....._._. .
A-32. Reactor conditions for Growth Run #524 ................. ............_ .....193.........
A-33. Reactor conditions for Growth Run #525 ................ ...............194........... ..
A-34. Reactor conditions for Growth Run #535 .........._._ ........... ......_.. ........19
A-3 5. Reactor conditions for Growth Run #536. ................. ....._.. ............ ......19
A-36. Reactor conditions for Growth Run #537 ................. ............_ .....197.........
A-37. Reactor conditions for Growth Run #53 8. ................ ....___ ....___ .........19
A-38. Reactor conditions for Growth Run #540 ................. ..........._._.....199...... ..
A-39. Reactor conditions for Growth Run #541 ...._.._.._ ................._._. .........20
A-40. Reactor conditions for Growth Run #542. .............. ...............201....
A-41. Reactor conditions for Growth Run #569. .............. ...............202....
A-42. Reactor conditions for Growth Run #575. ............. ...............203....
A-43. Reactor conditions for Growth Run #578. ............. ...............204....
A-44. Reactor conditions for Growth Run #579. .............. ...............205....
A-45. Reactor conditions for Growth Run #582. .............. ...............206....
A-46. Reactor conditions for Growth Run #586. .............. ...............207....
A-47. Reactor conditions for Growth Run #587. .............. ...............208....
A-48. Reactor conditions for Growth Run #588. ............. ...............209....
A-49. Reactor conditions for Growth Run #628. ............. ...............210....
A-50. Reactor conditions for Growth Run #629 ................. ............_ .....211.........
A-51. Reactor conditions for Growth Run #630. .............. ...............212....
A-52. Reactor conditions for Growth Run #634. .............. ...............213....
A-53. Reactor conditions for Growth Run #63 5 ................ ...............214.............
A-54. Reactor conditions for Growth Run #636. .............. ...............215....
A-55. Reactor conditions for Growth Run #637. .............. ...............216....
A-56. Reactor conditions for Growth Run #63 8 ...._.__._ ......_._. ....._._. .........21
A-57. Reactor conditions for Growth Run #639. .............. ...............218....
A-58. Reactor conditions for Growth Run #640. .............. ...............219....
A-59. Reactor conditions for Growth Run #641 ...._.._.._ ................._._. .........20
A-60. Reactor conditions for Growth Run #647. .............. ...............221....
A-61. Reactor conditions for Growth Run #648. ............. ...............222....
A-62. Reactor conditions for Growth Run #649. .............. ...............223....
A-63. Reactor conditions for Growth Run #652. .............. ...............224....
A-64. Reactor conditions for Growth Run #653. ............. ...............225....
A-65. Reactor conditions for Growth Run #654. .............. ...............226....
A-66. Reactor conditions for Growth Run #655. ............. ...............227....
A-67. Reactor conditions for Growth Run #656. .............. ...............228....
A-68. Reactor conditions for Growth Run #657. .............. ...............229....
A-69. Reactor conditions for Growth Run #658. ............. ...............230....
A-70. Reactor conditions for Growth Run #659 ................. ............_ .....231.........
A-71. Reactor conditions for Growth Run #660. .............. ...............232....
A-72. Reactor conditions for Growth Run #661 ...._.._.._ ................._._. .........23
A-73. Reactor conditions for Growth Run #662. .............. ...............234....
A-74. Reactor conditions for Growth Run #666. .............. ...............235....
LIST OF FIGURES
1-1. Top view of the PMEE reactor ........... ..... ._ ...............25.
2-1. Spectral irradiance versus wavelength under AMO and AM1.5 conditions. .....................49
2-2. Photovoltaic system. ............. ...............49.....
2-3. Chalcopyrite structure of CulnSe2. ............. ...............50.....
2-4. CIGS/CGS monolithic tandem device structure ................. ...............51...............
3-1. UF growth recipes ................. ...............71................
3-2. Typical CIGS device structure ................. ...............72........... ...
4-1. Morphologies of films grown at lower growth temperatures by similar growth
recipes. ............. ...............94.....
4-2. Morphology of a film grown at a higher growth temperature. ............. .....................9
4-3. Morphologies of the Cu-rich domain region of CGS films grown by the same recipe
at different growth temperatures. .............. ...............95....
4-4. Morphologies of the Ga-rich matrix region of CGS films grown by the same recipe
at different growth temperatures.. ............. ...............95.....
4-5. Morphologies of the Cu-rich domain region of CGS films grown with different
growth recipes at 4910C ................. ...............95...............
4-6. Morphologies of the Ga-rich matrix region of CGS films grown with different
growth recipes at 4910C ................. ...............96...............
4-7. Morphologies of CGS films grown by the emulated 3-stage process at 4910C
(X3 0,000) ................. ...............96........... ....
4-8. Morphology of a Cu-rich film (#542) with large grains and a uniform surface. ...............96
4-9. Diffraction patterns of films grown at different temperatures with the same modified
three-stage process.. ............. ...............97.....
4-10. Diffraction patterns of films grown at different temperatures with the same modified
three-stage process featuring an initial GaSe layer. ....._._._ .... ... .__ ........_........98
4-11. Diffraction patterns of films grown at different rotational speeds ................. ................99
4-12. Diffracti on patterns of film s grown at different levels of overall Cu-ri chnes s................1 00
4-13. Effect of KCN-etch on the diffraction pattern of a Cu-rich fi1m.. ................ .................101
4-14. Diffraction patterns of films grown by the Constant Cu Rate Process. .................. .........102
4-15. Diffraction patterns of films grown with varying levels of peak Cu-richness. ................ 103
4-16. Diffraction patterns of films grown by the Emulated 3-Stage Process. ................... ........104
4-17. Diffraction pattern of a film grown by the Emulated 3-Stage Process that was never
Cu-rich. ............. ...............105....
4-18. Surface morphology of films grown by the Constant Cu Rate process (X100).. ........._...106
4-19. Surface morphology of a Cu-rich film grown by the Constant Cu Rate Process
(X 5000). .............. ...............107....
4-20. Surface morphology of a Cu-rich film grown by the Constant Cu Rate Process
(X 10,000)............... ...............10
4-21. Surface morphology of a Ga-rich film grown by the Constant Cu Rate Process. ...........109
4-22. Effect of KCN-etch on the surface morphology of the island region of a Cu-rich film..1 10
4-23. Effect of KCN-etch on the surface morphology of the field region of a Cu-rich film. ...11 1
4-24. Surface morphology of a Ga-rich film with rings around the islands..............._._...........112
4-25. Distinct grain structure of a Ga-rich film with rings around its islands. ................... .......1 13
4-26. Surface morphology of a film grown by the Emulated 3-Stage Process. ................... .....1 14
4-27. Surface morphology of a Cu-rich film grown by the Emulated 3-Stage Process. ...........1 15
4-28. Surface morphology of a Ga-rich film grown by the Emulated 3-Stage Process. ...........1 16
4-29. Diffraction pattern of a Ga-rich film grown by the Constant Cu Rate process. ..............1 17
4-30. Diffraction patterns of films grown by the Constant Cu Rate process. ................... ........1 18
4-31i. Diffraction patterns of films grown by 3-stage process ................. ................. .... 1 19
5-1. Dark and illuminated I-V curves for Device #523............... ...............132.
5-2. Spectral response curves comparing Device #523 and #452. ..............._ .............. ....132
5-3. Photo I-V curve for Device # 640............... ...............133..
5-4. Photo I-V curve for Device # 655............... ...............133..
6-1. Diffraction pattern of CIS film #582. ............. .....................144
6-2. Diffraction patterns of CIGS films grown to different thicknesses.. ............ ..................145
6-3. Diffraction patterns of CIGS films grown with different Cu/III ratios............................ 146
6-4. Diffraction patterns of CIGS films grown with different Ga/III ratios.............._.._.. ........147
6-5. Illuminated I-V curve for Device #582............... ...............148.
6-6. Comparison of illuminated I-V curves of Device #575 and #588. ............. ... ..........._..149
6-7. Comparison of the illuminated I-V curves of Device #588 and the calibration cell. ......149
7-1. Metal source crucible. ................. ...............157_.._ .....
7-2. Deposition flux Fs(r) (atoms/cm2-S) on the substrate at nine different melt levels. ........157
7-3. Positioning of the sources in the reactor. ...._ ......_____ .......___ ..........5
7-4. Cryoshroud ................. ...............158....._.. .....
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
COPPER GALLIUM DISELENIDE THIN FILM AB SORBER GROWTH
FOR SOLAR CELL DEVICE FABRICATION
Chair: Oscar Crisalle
Major: Chemical Engineering
A custom-built migration-enhanced epitaxy reactor originally optimized for CulnSe2 (CIS)
deposition was modified to grow gallium-containing compound semiconductor thin films, such
as CuGaSe2 (CGS) and Cunlnt-GaxSe2 (CIGS). The addition of gallium allows for the
manufacturing of solar cell absorber layers with wider band gaps.
Three distinct growth recipes under several growth temperatures and a wide range of
metal-composition ratios are used to deposit polycrystalline CGS thin films. The surface
morphology of gallium-rich fi1ms is typically very uniform, with long needle-like grains when
grown by the first recipe, a constant copper-rate process. In contrast, copper-rich fi1ms grown by
this same recipe or by a modified three-stage process have island structures with very large
grains embedded in a matrix region that possesses small grains. The surface morphology
becomes more uniform and the grains in the matrix region become larger when a higher growth
temperature is used. The third recipe, an emulated three-stage process, does not produce films
with an island-matrix structure, and the grains are uniformly large.
The highest conversion efficiency achieved for solar cells based on CGS is 5.3%, delivered
by a copper-rich absorber deposited at the highest sustainable growth temperature of 4910C.
This device has a large fill factor of 66 %, but the open-circuit voltage of 0.48 V is lower than
what is expected from a wide band-gap absorber. A set of CIGS solar cells was completely
fabricated and characterized in-house. This led to the most efficient device produced from an
absorber grown in our reactor, in the form of a 9 % CIS solar cell featuring a one-micron film
deposited at 4910C.
Finally, a dynamic reactor model was created to describe the deposition environment in our
epitaxial reactor. All relevant physical features are incorporated, including the cyclic motion of a
rotating platen and the spatial distribution of the flux produced by three metal effusion sources.
Reaction occurs under an excess of selenium, and operational variables such as rotational speed
and melt height can be simulated. The outputs are predicted film thickness and composition.
Further work is proposed to identify the values of adjustable sticking coefficients using
This first chapter is intended to give the reader an overall description of the system under
study. This information will be followed by the statement of the obj ectives of the proj ect and
Einally by the proposed solution strategy to the problems. Not many details will be covered in
this chapter; the purpose is to give the reader an overall idea of the rationale behind the
remaining sections of the report.
The second chapter is an introduction to photovoltaics and thin film solar cells, especially
those based on copper chalcopyrites. Its purpose is to familiarize the reader with energy
production, solar cells and the important parameters that affect their efficiency. This chapter
may be skipped by readers who are already familiar with the field without a loss of continuity.
The third chapter is an in-depth description of the absorber growth and device fabrication
procedures. It is complimented by an explanation of the various techniques used to characterize
the respective films and devices. Our techniques are compared to those in the literature.
Chapter 4 is an account of Copper Gallium Diselenide absorber growth in our modified
molecular beam epitaxy reactor. The main motivation is the characterization of the films grown
under various processing conditions. This chapter is accompanied by Appendix A, which
includes the growth conditions for each absorber film described within the course of this work.
The fifth chapter is very similar in structure to the previous one, with the focal point shifting
to the device fabrication of CuGaSe2 Solar cells. The thin films grown in the PlVEE reactor that
are discussed in Chapter 4 are used as the absorber layers in solar cell devices. Most cells were
finished at the National Renewable Energy Laboratory (NREL), except for the final set which
was completely fabricated in-house at the University of Florida.
In Chapter 6, low gallium content Cu(In,Ga)Se2 absorbers are grown and completely
fabricated within our facilities. The as-grown absorbers and subsequent devices were
characterized to determine the effect of Ga composition on films grown by a simple single-stage
process at a substrate temperature below 5000C. Appendix A also contains the reactor
conditions pertaining to each of the growth runs.
In the seventh chapter, the focus is shifted to modeling of the reactor. A flux model had
already been developed and needed to be incorporated into overall dynamic reactor model. This
chapter is complemented by appendix B, which includes the details of model development.
The final two sections of the manuscript are conclusions and the list of references. The
conclusions chapter also contains a list of possible future directions this research can take.
This proj ect was initiated after the Boeing Company decided to terminate its photovoltaics
research program and donated some research equipment to the University of Florida. Billy
Stanbery, who was part of the Boeing Team, decided to enroll at the University of Florida to
pursue a PhD degree. This jump-started a comprehensive, multi-faceted and multidisciplinary
CIS solar cell research effort at the University of Florida.
Physical Vapor Deposition (PVD) describes semiconductor thin film growth in a reactor
whose high vacuum conditions cause material to flow in the molecular regime. Molecular Beam
Epitaxy (MBE) describes this deposition process when epitaxial growth results. The main
attributes of MBE compared to other techniques are a low growth temperature that limits
diffusion, a slow growth rate that ensures two-dimensional growth, a simple growth mechanism,
and compatibility with in situ analysis. Because of its unprecedented control down to the atomic
scale, MBE has been employed for the growth of many novel devices that require "band gap
Migration Enhanced Epitaxy (MEE) is a variant of MBE based on sequential rather than
simultaneous exposure of the substrate to source fluxes. Rather than using shutters to control the
material deposition on the substrates, the substrates rotate on a donut-shaped platen that takes
them through the different deposition zones as well as fluxless relaxation steps in between. Each
substrate is sequentially exposed during a complete cycle to Cu+In+Ga, background vacuum
ambient, Se, and the background vacuum ambient again . Chalcopyrite films have been
grown by MBE for nearly 30 years , but the rotating platen, which is the main concept of
MEE, makes our work unique compared to other research groups.
Our own reactor has been named with the acronym PMEE (plasma-assisted migration
enhanced epitaxy) because of the incorporation of a plasma cracker for selenium or sulfur
deposition. This reactor was originally designed to deposit CulnSe2 (CIS) absorber layers and
then was modified to support the growth of CuGaSe2 (CGS) and Cunlnt-GaxSe2 (CIGS) films as
well. The PMEE reactor can deposit Cu, In, Ga, Se, S, and Na, allowing for the deposition of a
wide-variety of Cu-chalcopyrite thin films. It can support device manufacturing based on
polycrystalline co-deposited CIS-based films, as well as an assortment of studies such as single
crystal growth and bilayer precursor design for RTP studies. This process is low throughput and
thus not economically feasible, but this research is concerned with investigating the film
properties, not large-scale production. The ultra-high vacuum (UHV) creates an extremely clean
condition and makes it possible to generate the molecular beam of each source so that the growth
system can be used to grow epitaxial CIS thin films of high crystalline quality. It can overcome,
to a certain extent, a disadvantage of MBE, which is low productivity by processing nine
samples in one batch.
There are some disadvantages of the system. Due to the rotational movement of the platen
and thus the substrates, direct in-situ measurement of the substrate temperature is virtually
impossible. The thermocouple is currently located in the gap between the platen and the heater
and reading sort of an average value of those two temperatures. The localized heater position
creates non-uniformity of the temperature distribution on the substrates. The growth rate is
significantly limited by the selenium flux delivery. Even with a high Se flux rate
([Se]/([Cu]+[In])>5), it is hard to obtain sufficient Se incorporation into a growing fi1m under a
high temperature condition since the Se deposition zone is confined, and the high vapor-pressure
material is easily re-evaporated from the surface. As a result, maximum flux rates of Cu and In
are limited, which makes it difficult to achieve high growth rate. A time-consuming and costly
problem is that the PMEE reactor has been custom designed and therefore requires extensive
customizations to the regular sources available from manufacturers.
The reactor can be divided into four zones as is shown in Figure 1-1: load-lock, chalcogen,
heater, and metals. Materials are sequentially deposited as the substrate passes through their
respective zones rather than co-deposited. The total flux is highly enriched in the specie
evaporating from the nearer of the two sources as the substrate initially enters the metals
deposition zone from either side. Hence, substrate rotation direction can result in substantially
different compositions within the first metal layer. Counterclockwise rotation results in an initial
metal flux during each MEE cycle that is substantially Cu-enriched.
The high-vacuum chamber is divided vertically into two zones by the cryoshroud and is
maintained at a base pressure of around 10s torr by a series of diffusion and mechanical pumps.
Above the cryoshroud is the growth zone, where all the sources and substrates are located.
Liquid nitrogen, circulated in the cryoshroud, further reduces the pressure in the growth zone to a
base pressure around 10-9 torr. Pressure during deposition is in the range of 10-7 to 10-s torr
depending on the operating conditions. If the chamber is brought to atmospheric pressure, it
takes a few days to get the appropriate low pressure back.
A load-lock, attached to a port at the substrate platen level of the chamber, allows the
system to remain under vacuum for months during operation. The load-lock is independently
pumped with a small turbomolecular pump (TMP) and isolated by a gate valve from the loading
chamber. Chamber venting uses argon gas and the load-lock is equipped with a Venturi pump
and a liquid nitrogen sorption pump for rough-pumping down to the TMP's crossover pressure of
10-3 torr. Up to nine 2" diameter wafers or 2" x 2" square substrates can be loaded onto the
donut-shaped molybdenum platen via a two-prong fork, which is then rotated so that each
substrate travels through all four zones.
The chalcogen zone is where a thermal cracker for Selenium (Se) and a low-capacity
plasma cracker that can be charged with Selenium or Sulfur (S) are installed. Since Se does not
evaporate as a very reactive species, it must be further cracked to be incorporated into the film.
The thermal cracker breaks the large molecules into smaller, more reactive molecules by heating
them to very high temperatures in a double-oven reactor before deposition. Hence the
temperature in the evaporation zone of this double oven controls the flux of the material and the
temperature in the cracking zone controls the species distribution . The plasma cracker
accelerates particles to make them more effectively reactive, which is an alternative to the high
temperature of the thermal cracker. No sensors are used to measure the flux since the Se is
deposited in excess; only the temperatures of the sources are measured. Excess Se deposits
everywhere so a chalcogen zone shield was incorporated to isolate this zone from the rest of the
After passing through the chalcogen zone, substrates are brought to the desired operation
temperature in the heater zone with radiative heating by a boron nitride-coated radiation heater.
Most of the substrate platen heating is provided here. In other zones, the substrates are slowly
cooled down since there is no direct heating there. Some extent of non-uniformity of the
temperature distribution on the platen is expected due to the complex design .
Finally, the substrates enter the metal deposition zone where they are sequentially exposed
to Copper (Cu), Indium (In), and Gallium (Ga) fluxes. The Cu, In, and Ga sources are thermal
evaporation sources with conical shaped crucibles and free-evaporating surfaces. Deposition
uniformity is improved significantly with a conical instead of a cylindrical crucible . The
effusion sources are identical in structure: 7.50 tapered angle, 30 cc capacity, and constructed
with Pyrolithic Boron Nitride (PBN). Cu and Ga have dual filament heating structures due to
their properties, whereas In has only one. The tip filament keeps the tip of the crucible hot so no
impurities can condense on the surface. Shielding prevents deposition on each individual
sub state except during that portion of each rotation cycle of the substrate platen when it is inside
The platen then rotates back into the load-lock zone where the entire cycle restarts. The
dopant source is located in this area to introduce small quantities of impurities. It is charged with
a very small amount of NaF. The metals zone was expanded into the previously allocated load-
lock area so that the Gallium source could be added.
Every material source is equipped with thermocouples for monitoring the temperature.
The source temperature is actually measured indirectly by putting the thermocouple in thermal
contact with the crucible. These are mainly c-type thermocouples, with the exception of the
evaporation zone in the two crackers which are fitted by k-type thermocouples due to the lower
temperatures involved. The substrate temperature is measured indirectly by means of a c-type
thermocouple suspended in between the rotating platen and the substrate heater. The rotating
nature of the platen prevents getting direct temperature readings on the substrate .
Rate control is provided by a Leybold-Heraeus Inficon Sentinel III with both EIES
(Electron Impact Emission Spectroscopy) sensors for monitoring and controlling the metals
deposition process and quartz crystal monitors (QCM) for calibration. Closed-loop feedback
control is conducted along with the in-situ rate measurement employed by the EIES sensors for
Cu and In sources. EIES sensors are calibrated by a QCM that is located right over the source
cells whenever source material is reloaded. A single QCM is used to monitor the Ga flux
because the reactor modifications were limited by space and the current setup. The dopant flux
is monitored by QCM because it doesn't need to be controlled tightly. No instrument is used to
measure the Se flux rate in-situ so closed-loop feedback control based on temperature has been
An instrumentation and control interface for the PMEE reactor was designed to enable the
implementation of advanced control strategies envisioned for the local sources as well as the
supervisory control structure. A human-machine interface is programmed on a LABVIEW
platform so that real time control of the PMEE reactor can be administered through a central
computer . Using microprocessor control, one can ensure run-to-run repeatability by
constantly monitoring and adjusting the various growth parameters.
The purpose of this proj ect is to explore the key processing issues associated with growing
copper chalcopyrite fi1ms containing gallium in a migration enhanced epitaxy reactor. Copper
Gallium Diselenide is theoretically a good candidate as the top cell in a tandem solar device,
owing to its nearly ideal band gap of 1.7 eV and maximum theoretical efficiency of 26%.
Practical use in a tandem cell will require efficiencies greater than the current best cell efficiency
of approximately 10%. A low temperature process for the growth of CuGaSe2 absorber layers
should be developed to avoid the degradation of the junctions located underneath the top cell in a
monolithic tandem structure.
To grow high-quality CGS absorbers, several steps were required. First, a gallium source
needed to be retro-fit into a custom-built IVEE reactor used for the deposition of CulnSe2
absorber films. Then it needed to be shown that this reactor could grow polycrystalline CuGaSe2
thin films that produced working solar cell devices. Various growth recipes were investigated,
along with varying material compositions, growth temperatures, and post-deposition processes.
The structural and morphological properties of the films were characterized along with electrical
properties of the subsequently fabricated devices.
Another goal was to achieve complete in-house fabrication and characterization of CGS
devices. This was intended to decrease the feedback time needed in investigating the effect of
processing changes on electrical properties of the solar cell. This required a team effort of
several graduate students.
Finally, the flux models of the effusion sources needed to be incorporated into a dynamic
reactor model. This model will be the basis for a control feedback scheme that will correlate
film properties and hence, device properties with the input conditions of the reactor.
Top view of the PMEE reactor.
Rapid industrialization combined with an expanding population is driving the world's
demand for energy, which is proj ected to triple by the end of the century (from 13 TW to 46 TW)
. The fossil fuel reserves that currently furnish power to the globe will fall short of this
demand over the long term, and their continued use produces harmful side effects such as
pollution. Finding sufficient supplies of clean energy for the future may be civilization's most
difficult challenge. Alternative renewable fuels are currently not competitive with fossil fuels in
cost and production capacity, but solar cells have the potential to become a key part of the
solution to this problem.
Photovoltaic (PV) systems exploit an inexhaustible resource that is free to use and
available anywhere in the world. More energy from sunlight reaches the earth's surface in one
hour (4.3 x 1020 J) than is consumed by civilization in an entire year (4. 1 x 1020 J) . If 0. 16%
of the land on Earth was covered with 10% efficient solar cells, 20 TW of power would be
provided, which is more than the world' s current consumption rate of fossil energy . This
illustrates the impressive magnitude of the solar resource and the potential harbored by solar cell
The sun emits energy as a blackbody radiator at a temperature of approximately 6000 K
with a spectrum ranging from the ultraviolet (3.5 eV), through the visible, into the infrared (0.5
eV). The energy of the visible region ranges from 3.1 eV (violet) to 1.8 eV (red) with the peak
power of the sun occurring at approximately 2.25 eV. This distribution of photons in the
spectrum is one of the greatest limiting factors on solar cell performance. Under monochromatic
light, a typical PV cell might be able to convert 60% of the light to electricity, but under the
multicolored solar spectrum, the same cell would only be able to convert 10% of the light' s
energy to electricity . Diffuse light is the portion of sunlight that has been refracted or
scattered in the atmosphere before it reaches the Earth's surface. When the sky is completely
clear, about 10% of the sunlight is diffuse. Most losses in the spectrum at lower energies are
caused by light being absorbed by molecules of water vapor while at energies higher than 3 eV,
almost all sunlight is absorbed by ozone .
The AM1 solar spectrum represents the sunlight on the Earth' s surface when the sun is at
its peak. At a solar zenith angle of 48.20, the equivalent of 1.5 of these noontime air masses
(AM1.5) is diminishing the intensity of the sunlight . The AM1.5 condition has an incident
power of 84.4 mW/cm2 and is the most appropriate for calculating the conversion efficiency of a
solar cell in the terrestrial environment . The irradiant power of the sun under AMO
conditions is 135.3 mW/cm2, which is the spectrum measured outside the earth's atmosphere.
AM1.5 is compared to AMO in Figure 2-1 . The "peak watt" (WP) rating is the power (in
watts) produced by a solar cell module illuminated under the following standard conditions: 100
mW/cm2 intensity, 250C ambient temperature, and AM1.5. Because of day/night and time-of-
day variations and cloud cover, the average electrical power produced by a solar cell over a year
is about 20% of its WP rating .
The two most important variables controlling the amount of annual sunlight are latitude
and local cloud cover. Latitude is important to the amount of annual sunlight for two reasons:
the angle of the sun and the length of the days. The Northeast and Northwest United States are
the cloudiest regions, while the Southwest usually has the clearest skies. In Buffalo, NY, there is
about 60% and in Sacramento, CA, about 85% of the solar energy available in Albuquerque, NM
. Almost 90% of the country gets between 6 and 8 kWh/m2 daily, which is enough for the
effective use of PV . There is also a seasonal effect on the energy produced by PV modules
because in the summer the sun spends more time at high elevation angles (low air mass) than in
the winter (high air mass)
Clearly, solar energy can be exploited on the needed scale to meet the global energy
demand. Sunlight is readily available and its use does not harm the environment through
pollution or the climate through greenhouse gases. Yet, U.S. electricity production by solar cells
currently represents a tiny fraction (<0.02%) of the total electricity supply . The wide-spread
use of PV has been hampered by the relatively high price of the solar cell module. The huge gap
between our present consumption of solar energy and its enormous undeveloped potential
defines a grand challenge in energy research.
Solar cells typically have a lifetime of at least 30 years and incur no fuel expenses, but they
do involve a capital cost. The cost for the electricity produced by the cell is estimated by
spreading the total capital cost over the entire lifetime of the cell while considering the total
electrical energy that will be produced during that time. Higher conversion efficiency thus
directly impacts the overall electricity cost because higher efficiency cells will produce more
electrical energy per unit cell area over the cell lifetime. The most useful cost calculation for PV
cell modules ($/WP) is determined by the ratio of module cost per unit area ($/m2) divided by the
maximum amount of electric power delivered per unit of area (module efficiency multiplied by
1000 W/m2, the peak isolation power) . In addition to module costs, a PV system also has
costs related to the non-photoactive parts of the system, called balance of system (BOS) costs.
To compete with electricity produced from fossil fuels, solar cell costs must eventually approach
Efficiency = pwrot(2-1)
For a solar cell, this is the ratio of the electric power produced by the cell at any time
versus the power of the sunlight arriving at the cell. Efficiencies do not fluctuate much over the
life of a cell unless it is degrading. By definition, the higher a photovoltaic device's efficiency,
the more electricity it produces for a given exposed area. Besides device efficiency, the
definition of performance must also include uniformity, reproducibility, throughput, materials
utilization, and yield .
A single cell is only useful when powering wristwatches and calculators so many cells are
connected within a module. The module serves two purposes: it protects the solar cells from the
outside environment and it delivers a higher voltage than a single cell. Individual small-area
cells must be connected in series so some part of the module area is lost for the interconnection
of cells. Another important effect which reduces the performance of modules is the non-
uniformity of voltage and current density over a large area, which is mainly related to the
compositional variation of the absorber layer . A critical issue that must be resolved is the
large gap in efficiency between small-area laboratory devices and large-area modules.
The basic components of a photovoltaic system are the modules, support structures, land,
and possibly sun trackers. There are two different kinds of modules: flat plates and
concentrators. Flat plates are large panels that can be assembled into even larger arrays. To
maximize energy production, a module can be mounted onto a two-axis solar tracker so that it
always points directly at the sun . Concentrators use large, concentrating lenses to focus
sunlight onto small cells. These lenses replace large areas of expensive semiconductor material,
reducing the total module cost. A concentrator does not produce any energy when the weather is
cloudy because it cannot focus diffuse light. Hence, they are geographically limited to sunny
locations as opposed to flat plates, which can be useful in cloudy areas.
There are two major market sectors: grid-connected and stand-alone systems. The former
delivers power directly to the grid by converting the direct current of solar modules into
alternating current by an inverter. The latter supplies power to isolated sites and to small-scale
consumer products. Since solar energy is not always available, effective storage and distribution
are critical to matching supply with demand. Storage drives up the cost of solar cells. A
simplified photovoltaic system is shown Figure 2-2. Besides the terrestrial market, there is also
the space market, which has different material and cost requirements.
The primary obj ective of worldwide solar cell research and development is to reduce the
cost of photovoltaics to a level that will be competitive with conventional ways of generating
power. The world market grew from less than 10 MWP/yr in 1980, to 80 MW in 1996 sold at
prices close to $10/WP , to finally exceeding 1000 MW in 2004 at less than $7/WP .
Costs need to be reduced even more for solar cells to become competitive with the likes of oil,
coal, and natural gas. Otherwise, future large-scale use of PV might depend more on
environmental concerns rather than economic competitiveness.
Behind the progress of photovoltaics as a technology is a loyalty to PV as an idea. There
are technical obstacles that must be overcome and their solutions depend on the funding of
fundamental research. As these issues are resolved, the costs will continue to fall. "Solar
electricity is part of America' s present and its future as is the R&D that enables it" Larry
Kazmerski, one of the pioneers of thin film solar cells .
Solar Cell History
Photovoltaics had its beginnings in the nineteenth century. French Scientist Alexandre-
Edmond Becquerel discovered the PV effect in 1839. He observed that a voltage and a current
were produced when two electrodes in a beaker full of fluid were exposed to sunlight. And in
1873, Willoughby Smith found that the element selenium conducted far more electricity when it
was illuminated than it did when it was dark .
Many consider the work accomplished during the 1950s at Bell Laboratories to be the true
origin of photovoltaics. Cal Fuller, Darryl Chapin, and Gordon Pearson made a silicon cell that
was able to convert 6% of sunlight into electricity , which was a large improvement over
selenium cells. Fuller and Chapin eventually reached 10% conversion efficiency, but compared
to traditional electricity in the 1950s, the cost of PV-produced power was a thousand times as
From the mid 1950s to the early 1970s, PV research and development was directed
primarily toward space applications where it is the conventional power source. In space, payload
weight is critical and solar cells weigh very little compared to the power that they produce.
Nearly every communications satellite, military satellite, and scientific probe is powered by PV.
Satellite manufacturers can endure a high cost because it is a small fraction of their total cost, but
they cannot tolerate an unreliable power source that may jeopardize their entire investment.
Then in 1973, a greatly increased level of research and development on solar cells was initiated
following the oil embargo in that year, which led to the creation of the U. S. Department of
Energy, along with its PV program, a few years later.
Researchers at Bell Labs showed three decades ago that the I-III-VI2 Semiconductor
CulnSe2 (CIS) allows for efficient solar-to-electrical energy conversion. A single crystal CIS
cell of 12% efficiency was made at Bell Laboratories in 1975  and the University of Maine
produced a polycrystalline CulnSe2 Cell Of nearly 6% efficiency in the following year .
Although Bell Labs and Maine initiated the study of CIS, the most important early research was
done by a small group at Boeing Aerospace Corporation in Seattle. Boeing's team, led by Reid
Mickelsen and Wen Chen, achieved over 10% efficiency with CIS cells in 1982 using elemental
co-evaporation of the source materials in vacuum .
Solar Cell Device Physics
Photovoltaics (PV) is the direct conversion of sunlight into electricity; solar cells absorb
sunlight and change it continuously into electricity. According to quantum theory, light can
behave either as waves or as particles. Discrete particle-like packets of light are called photons,
and each photon has a well-defined energy and wavelength.
The valence and conduction bands of an inorganic semiconductor are separated by a
forbidden energy range that the electrons cannot occupy . This minimum threshold energy is
called the energy gap or band gap (E,), and it varies for different materials because each of them
has a different bond strength. Semiconductors with weak bonds have small energy band gaps.
Conduction is only possible if we can impart kinetic energy to an electron. When photons with
energies hv > E, impinge upon a pn junction solar cell, they are absorbed and the rate of
generation of electron-hole pairs as a function of distance x from the surface of the solar cell is
given by the following equation :
gE(X) = a00(1-R)e-ax (2-2)
where a is the optical absorption coefficient, Go is the incident photon flux density per unit
bandwidth per second, and R is the reflection coefficient. Photons with energies below the band
gap pass through the absorber without being absorbed.
Most semiconductor devices incorporate both positive and negative regions, and it is the
space-charge region formed between them that leads to their useful electrical characteristics .
After a certain number of electrons and holes have flowed from one region to the other, an
electric field will be built up, preventing further net flow of the carriers. The greater the density
of free carriers initially on each side of the interface, the greater will be the electric field that
forms when they mix. This electric field is not tenuous; it does not come and go. The area of the
field is also called the depletion region because in that region there are no free carriers. They all
are either in bonds or swept away by the field. Under equilibrium conditions, electron-hole pairs
are continuously generated everywhere within the semiconductor and in the absence of an
applied voltage, the electron-hole pairs recombine and therefore no current flow results.
However, when a positive voltage is applied to the n-region of a diode with respect to the p-
region, the electron-hole pairs, once generated, will be separated and their probability of
recombination is diminished. The most important attribute of these p-n junctions is that they
rectify, i.e. they permit the passage of electric current in only one direction .
The electric field that drives the current is proportional to the material's band gap. A
semiconductor with a small band gap produces an insignificant voltage. As the band gap is
increased, more solar photons within the spectrum will lack the energy to produce electrons and
holes so a very large band gap would produce a high voltage with a tiny current. For a
semiconductor to be a sufficient absorber layer in a solar cell, it must have a band gap that allows
for both reasonable current and voltage.
Radiative recombination, in which a hole reacts with an electron and produces a photon, is
exactly the reverse of absorption; it is the spontaneous transition of an electron from the
conduction band to an unoccupied state in the valence band . In real solar cells,
recombination via impurities is the predominant recombination process. Recombination centers
located within the electric field can severely reduce the field's strength and thus the voltage.
These are called shunts . Defects located at the interface can also be efficient recombination
centers because they introduce deep trap levels into the band gap .
In compound semiconductors, intrinsic point defects are introduced to compensate for
deviations from the stoichiometry . The simplest point defect is the vacancy in which a
single atom is missing from the lattice. An interstitial is an extra atom occupying space between
the normal lattice sites. A component atom may also occur on a site intended for another. This
is called an antisite defect. Point defects influence the bulk properties of a semiconductor as
opposed to grain boundaries or interfaces that only affect the film locally .
A grain boundary is the complete fracturing of bonds along an entire surface. They occur
when a lattice takes shape in such a way that a defect spreads and makes it impossible for nearby
atoms to bond together. According to the grain boundary carrier-trapping theory, grain
boundaries work as trapping centers and therefore hinder the transport of charge carriers towards
the pn interface . Cell performance would suffer considerably because these lost electrons
would not contribute to the cell current. It is very common for crystal defects to cause a
polycrystalline structure where the grain boundaries separate regions of different crystallographic
orientation. Many of the electrical and optical properties of these materials are determined by
the corresponding properties of the grain boundaries and these may differ considerably from a
single crystal .
When grain boundaries cannot be avoided, they need to be passivated to minimize their
effect. This consists of adding some extra material, usually oxygen, that can make the grain
boundary defects less harmful. Most added materials diffuse preferentially down grain
boundaries so it goes directly to the region where it can have the most effect. The added oxygen
might passivate defects in the grain boundary by grabbing loosely bound electrons, removing
many of them as undesirable recombination centers. Some materials have relatively harmless
grain boundaries like CulnSe2; they may be self-passivating or some growth step may passivate
them. They can be made inexpensively and yet behave almost as if they were single crystals.
Others have very harmful grain boundaries, like gallium arsenide (GaAs) and silicon (Si ) .
Crystalline silicon currently dominates the photovoltaic market despite a complicated
manufacturing process and a high production cost. Its advantages are a readily available raw
material, mature processing technology, and non-toxicity. Si-based products are also very
reliable and are capable of achieving high efficiencies . The preeminence of the element Si,
in its amorphous or crystalline form, within the market is an overwhelming 99% .
A strategy for reducing costs is to use thin film materials that have a very high absorptivity
for solar photons. The leading thin film technologies, a-Si, CdTe, and CIS, offer the potential for
significant manufacturing advantages over crystalline Si. They have a lower consumption of
materials, independence from Si shortages, fewer processing steps, and a monolithic circuit
design so no assembly of individual solar cells into a Einal product is needed . Less material
usage leads to lower material costs, thinner layers leads to faster processes and lower capital
costs, and the processing of large-area devices leads to reduced handling costs. They can also be
made into flexible and light-weight modules on alternative substrates, which provide multiple
advantages in processing . The most serious threat to large-scale deployment of existing thin
fi1ms is materials availability of key elements such an In, Te, and Ge. Thus, Si-based technology
may always be relevant.
Direct vs Indirect Band Gap
Si has an indirect band gap so light absorption is much weaker than for thin films, which
are direct band gap semiconductors. The difference in absorption strength between direct and
indirect band gap semiconductors comes from the different processes by which they absorb
individual photons. Direct transitions are transitions in which the momentum of the electron-
hole pair does not change. Light of sufficient energy to free an electron from its fixed state is
absorbed by the electron, which is then freed. Light-absorption is more complicated in an
indirect band gap material. Promotion of an electron to the conduction band requires the
simultaneous interaction of a photon and a thermal vibration of the crystal lattice, called a
phonon. When a photon and a phonon of the proper energies are both absorbed, at the same
time, by a bound electron, a free electron-hole pair results. Almost all of the energy needed to
generate the electron-hole pair is carried by the photon; the phonon just acts to catalyze the
The absorption length of light in a semiconductor is determined by the likelihood that a
photon will be absorbed. The absorption length for crystalline Si is approximately 30 microns,
while it is about 0.3 microns for CIS. Higher energy photons have a greater probability of
interacting with a bound electron and being absorbed than lower energy photons, even though
both have more than the band gap energy. For instance, in CIS (Eg = 1.0 eV), the absorption
length of high-energy photons (>2.5 eV) is less than 0. 1 microns, but a beam of low-energy
photons (1.1 eV) might require 1 micron to be equally satisfied .
Free electrons on the p-side move around randomly before recombining. During this short
period of time, called the lifetime, they have a finite chance of encountering the electric field and
being sent to the n-type side of the device. This separation is a result of diffusion and the
average distance that minority carriers can move toward the built-in field before they return to
their fixed states is called the diffusion length. Materials with longer diffusion lengths are able
to produce more current. In solar cells, they can vary from less than a micron to over 100
microns in some single-crystal semiconductors. Large diffusion lengths are a necessity for
indirect band gap materials like crystalline silicon, but direct band gap materials absorb enough
light within their depletion region that their performance is not critical to diffusion length. If the
ratio of diffusion length to absorption length is greater than one, most carriers will be separated.
Crystal quality is the most important factor in determining a material's diffusion length because
diffusion is constrained by the propensity of free carriers to recombine.
CulnSe2-Based Solar Cells
CulnSe2-based solar cells are the most promising of the thin film solar cells and are the
basis for the investigations within this work. Cu-chalcopyrites, of the general composition
Cu(Inl-xGax)(SSexlv)2, Offer a wide range of band gaps from 1.0 eV for CulnSe2 to 2.4 eV for
CuGaS2. Recently, new world record total-area efficiencies of 15.0, 19.5, and 10.2% for
CdS/CIGS solar cells have been achieved for x=0 (CIS), x~0.28 (CIGS) and x=1 (CGS),
respectively . Table 2-1 shows these record efficiencies alongside their theoretical output.
The direct optical band gap of single crystal CulnSe2 has a value near 1 eV at room
temperature . The absorption coefficient of 105 cm-l for light greater than the band gap
means the thickness of the absorber can be reduced theoretically to less than 1 micron . CIS-
based materials belong to the group of I-III-VI2 Semiconducting compounds. Their lattice
elements are tetrahedrally coordinated similar to diamond-like semiconductors . The
chalcopyrite cell consists of two zinc blende cells with Cu and In occupying the same lattice sites
in the upper and lower cell, alternately as seen in Figure 2-3 . In the chalcopyrite structure,
each I (Cu) or III (In, Ga) atom has four bonds to the VI atom (Se). In turn, each Se atom has
two bonds to the Cu and two to the In (or Ga). Because the strengths of the I-VI and III-VI
bonds are different, the ratio of the lattice constants c/a is not exactly 2; it varies from 2.01 for
CIS  to 1.96 in CGS .
Doping in chalcopyrites is mainly controlled by compositional variation and thereby
induced defects, which lead to strong compensation . This makes the systematic study of
their electronic properties more difficult than in extrinsically dopable materials. The band gap
can be engineered, which offers a greater possibility of Einding the optimum photovoltaic
material with respect to cost, efficiency, and stability.
Four different phases have been found to be relevant: a-phase (CulnSe2), P-phase
(Culn3Ses), 6-phase (high temperature sphalerite phase), and Cu2-ySe . The existence range
of single-phase CulnSe2 is Very small and does not include the stoichiometric composition of
25% Cu. The Cu content of absorbers for efficient thin-fi1m solar cells typically varies between
22 and 24% (at.) Cu. At the growth temperature of 500-5500C, this region lies within the single-
phase region of the a-phase.
Some interesting features of CIS-based solar cells include low open-circuit voltage to band
gap ratio compared to Si and III-V devices, insensitivity of the conversion efficiency to the
[Cu]/[III] ratio over a wide range, exceptional tolerance to grain boundaries, and loss of
performance for CIGS alloys with x greater than 0.3 . Single crystal CIGS efficiencies also
lag behind those of polycrystalline CIGS. Optimal sodium incorporation is beneficial to device
performance, and excess sodium is detrimental. Na depth profiles typically exhibit some
qualitative features: enrichment at the CIS surface and a relatively lower concentration in the
bulk of the CIS with concentration increasing toward a maximum at the CIS/Mo interface .
For Chalcopyrite compounds, intrinsic defects are introduced to maintain the crystal
structure for non-stoichiometric composition. There are twelve intrinsic point defects that can be
formed in the ABX2 chalcopyrite lattice: three vacancies (VA, VB, Vx), three interstitials (Ai, Bi,
Xi), and six antisite defects. In addition to the antisite defects that can be formed by an exchange
of anions (X) and cations (A,B) as in binary compounds (Ax, Bx, XA, XB), One can also form
antisite defects on the same sublattice by an exchange of the cations (AB, BA) . The Cu
vacancy, Vco, is considered to be the dominant acceptor in Cu-poor p-type material, while the Se
vacancy, Vse, is considered to be the dominant donor in n-type material . Most of the off-
stoichiometry defects must be electronically inactive to allow for large deviations from
stoichiometry without the deterioration of the electronic quality of the film . A critical issue
in regard to CIS-based solar cells is the control of these point defects, which are responsible for
recombination in the space charge region of the devices .
When indium is replaced by gallium, the band gap increases. This is an effect of the
smaller size of the Ga atom, when compared to In, and the subsequent formation energies
involved. With the addition of gallium, some of the material properties also change. These
include structural properties like lattice constants, film morphology, and adhesion, and chemical
changes such as defect levels, affinities, and carrier concentrations. The band gap of Cunlnt
xSe2GaxSe2 can be tuned from 1.0 to 1.7 eV by adjusting the Ga content (x) from 0 to 1.
Although the Ga/(Ga+In) ratio of roughly 0.65 would provide the optimal band gap by
theory, the actual Ga content in the current record device is 28%, which corresponds to a band
gap of approximately 1.2 eV . This disparity implies that there are factors effecting
conversion efficiency other than band gap. The increase in efficiency in the range of 0 to 28%
Ga is due mainly to the increase of the band gap, and the potential on the grain boundary in this
gallium content range stays strong. However, at higher Ga content, the absence of the potential
on the grain boundary seems to be significant in comparison to the effect of the band-gap
Copper Gallium Diselenide, as a member of the I-III-VI2 COmpound semiconductors, has a
direct energy band gap of 1.7 eV , a very high absorption coefficient a = 3 x 104 Cm-1 at 1.7
eV , and an easily controllable electrical resistivity in a wide range of 10-1 to 105 ohm-cm
. The band gap of CGS decreases with increasing Cu/Ga ratio . The temperature
coefficient is lower for wide gap materials, which means that the efficiency loss at operating
temperature is less than for smaller band gap semiconductors . To date, device efficiency of
greater than 9.5% is reported using less than 2 microns of CGS absorber film by NREL .
CuGaSe2 has been investigated for more than 25 years, but in comparison to the rapid
progress made for CulnSe2 and Cu(In,Ga)Se2, the efficiency of CGS-based solar cells is still
relatively low. CGS did not overcome a limitation of 6.2% until fairly recently when efficiencies
of up to 9.3% for thin films and 9.7% for single crystal devices were achieved . These
reported values lie well below the theoretical limit for CGS of 9 = 26% . The most critical
drawback of CGS solar cells is the low open-circuit voltage (Voc) compared to band gap. In the
case of CIS or CIGS, the Voc follows a relationship with the band gap according to Voc ~ E,/q-
500 mV, where q is the elementary charge. A Voc of 1.2 V should therefore be possible for
CGS-based devices, but Cu-rich absorbers have been limited to a Voc of 750 mV , while
high efficiency Ga-rich absorbers have peaked at 900 mV . The improved device
performance of Ga-rich films results from decreased defect densities in the bulk and a decrease
of tunneling-enhanced recombination . Cell performance may be fundamentally constrained
since it seems that the Fermi level might be limited to values less than 800 meV above the
valence band edge, which would always make CGS p-type .
Wide gap chalcopyrites, such as CGS seem to have extensive material and growth-related
problems that limit the device performance. The tetragonal structure is capable of sustaining a
large concentration of vacancy and antisite defects in CGS . Devices based on slightly Cu-
poor CuGaSe2 absorbers have shown better performances than stoichiometric ones. In Cu-poor
material, the deviation from stoichiometry is not facilitated by the formation of a second phase,
as in the Cu-rich case, but the material develops a high density of defects. This leads to a high
degree of compensation, which in turn causes lateral potential fluctuations in the concentrations
of the charged defects . This density of defects is higher at the increased Ga contents
because the lattice mismatch in that composition range is anticipated to be larger. Better lattice
matching between the surface and the bulk leads to better performance . Optimal growth
conditions of CGS thin films are realized by a trade-off between growing Ga-rich films, reducing
the density of states in the band gap, and growing stoichiometric CGS to achieve optimal grain
size and crystallinity .
The actual path of recombination, which limits open circuit voltage, is important; it can
take place at the interface or within the space charge region, or it can be tunneling supported
. Increasing the Ga content in CIGS intensifies the contribution of tunneling, which is
observed as a larger characteristic tunneling energy, Eoo. This facilitates recombination and in
turn increases the recombination loss. For CGS, Rau et al. obtained significant values offoo and
found that the devices with the highest Voc are those with the lowest charge density and the
lowest tunneling currents . At room temperature, the tunneling contribution to
recombination is insignificant for low gallium content CIGS . Hence, there is no
fundamental difference between CIS and CGS with regard to recombination path, but
recombination losses in CGS are enhanced due to a higher contribution of tunneling .
Cu-rich CGS devices are controlled by high Eoo values due to tunneling enhanced interface
recombination. The efficiency gain achieved by the use of Ga-rich absorbers is mainly explained
by a reduced doping level and the decreased tunneling rate. All the beneficial device
modifications like air-annealing or the increase of CdS deposition temperature lead to a further
decrease of Eoo. The increased Cd diffusion into the absorber material during CBD explains the
reduction of tunneling [51i].
The efficiency of a CulnSe2 device is aided by the fact that its surfaces can be inverted to
become n-type even though the bulk of the sample is p-type. This is done via deposition of CdS,
which leads to band bending. The amount of band bending equals the shift in Fermi level with
respect to the valence band maximum from the p-type to the n-type region. A large amount of
band bending results in a Fermi level close to the conduction band minimum, which is needed to
get n-type conditions. The beneficial effects of this weakly n-type surface layer include the
reduction of the recombination rate and the enhancement of the carrier collection efficiency by
shifting the electrical junction away from the interface between CdS and the absorber .
The photovoltaic performance of CIGS devices deteriorates for x>0.3. Due to the lack of
type inversion for CIGS with x>0.3, the pn junction moves to the 'real' CdS/CIGS interface
causing higher recombination losses . CGS has been reported p-type for all compositions,
for thin films as well as single crystals; this is true for all deviations from stoichiometry and
molecularity. The doping-pinning rule predicts both n- and p-type behavior for CIS and only p-
type behavior for CGS. The maximum Fermi level positions are basically the same for both
these materials . Thus, it depends on the relative position of the band edges with respect to
those maximum Fermi level positions whether the material can be p-type or n-type or both. As
one applies to CGS the same process that converts CIS to n-type, the Fermi level does not rise
towards the conduction band minimum . The larger band gap of CGS leads to a higher
difference between the Fermi level and the upper edge of the valence band at the interface .
In CGS, type inversion cannot be achieved under normal conditions, but it can be reached
due to doping via non-equilibrium effects . Schon et al. have demonstrated that n-CGS can
be obtained using ion-implantation and Zn, Ge co-doping . As-grown p-type CGS single
crystals are first doped by Ge-implantation and then heated in vacuum. Finally, annealing of the
implanted samples in Zn atmosphere results in n-type conduction of CGS .
CulnS2 has an optical band gap (E,) of approximately 1.5 eV , which is an excellent
match for the solar spectrum. Its high absorption coefficient, a = 104 Cm-1 (at h = 500 nm) is also
very good for solar cell devices . Masse et al. calculated a maximum solar energy
conversion efficiency of 28% for a CulnS2 homojunction , but to date, the highest reported
efficiency for a solar cell is about 12% . Although CulnS2 Solar cells are theoretically
expected to possess higher efficiencies than CIGS, sulfides have reached only about 60% of the
performance of selenides so far. Selenide-based chalcopyrite solar cells that are slightly copper
poor have shown the best efficiencies, but sulfide absorbers must be prepared Cu-rich to produce
a working device . CuGaS2 has a band gap E, = 2.5 eV , but it has not been used in the
PV industry to date .
The progress of the solar industry depends not only on conversion efficiency, but also on
the development of techniques, which must be conducive to producing large-area devices at a
low cost. The absorber deposition method generally has a significant impact on the resulting
fi1m properties as well as on production cost. Common thin film deposition methods for CIS-
based solar cells are co-evaporation from elemental sources, selenization of metallic precursor
layers, evaporation from compound sources, chemical vapor deposition, closed-space vapor
transport, and low-temperature liquid phase methods like electrodeposition, spray pyrolysis, and
particle deposition techniques.
Siemens Solar Industries (SSI) was the first company in the world to produce CIS modules
using the selenization process. Low temperature (2000C) CIS precursor deposition is followed
by high temperature selenization . At 5000C, a complete recrystallization of the precursor
film occurs . The material quality of the absorber is determined by the structural features of
the precursor and the experimental conditions during selenization. Some disadvantages of this
process are that it involves complicated intermediate phases, interdiffusion, and reaction, which
can affect the controllability of the fi1m quality.
Usually high-quality I-III-VI2 absorber thin films are prepared on a laboratory scale by
physical vapor deposition (PVD) or molecular beam epitaxy (1VBE). These techniques require
high temperatures for the source metal evaporation . Molecular beam epitaxy is used to
grow CGS epilayers  while two and three-stage co-evaporation are the benchmark methods
for depositing polycrystalline CIGS. Although these growth methods are relatively easy to
implement on a small R&D scale, scale-up to a commercial level proves to be challenging.
High-quality CIGS deposition requires a high substrate temperature (>5000C) which limits the
selection of substrate materials and decreases throughput due to heat-up and cool down periods.
Co-deposition of the elements requires precise control of the flux of each element and an
overpressure of chalcogens (Se or S) during deposition, which results in low material utilization
and high equipment maintenance costs. An alternative low temperature route to the formation
of CIS is the rapid thermal processing of stacked metal/Se layers .
Most PVD deposition techniques are quite wasteful of materials, but printing or
electrodeposition techniques are quite efficient in depositing materials. The co-electrodeposition
technique , where Cu-In-Ga-Se species are present in the same chemical bath, is a simple
process to prepare low-cost thin films. It is crucial to control the deposition parameters like pH,
chemical bath composition, deposition time, deposition temperature, and the applied potential,
owing to their influence on the film properties and quality.
Printing, spraying, or coating of inks involves the deposition of particulate precursor
materials onto substrates at low temperatures and the subsequent sintering under chalcogen
overpressure. The reaction kinetics of nanoparticle-derived CIGS precursor films, typically 0.5-
2.0 microns, is much different than those prepared by evaporation . ISET's non-vacuum
process utilizes nanoparticles of mixed oxides of Cu, In, and Ga with a Eixed Cu/(In+Ga) ratio
that are synthesized into precursor inks .
The maj or reason for losses in simple photovoltaic devices is the inefficient use of the solar
spectrum by cells that have only one built-in electric Hield. Some proportion of the sunlight is
not used because certain photons do not have enough energy to be absorbed and to free electrons.
For those photons that do have enough energy, there is no distinction between them; they are all
treated as if they have just enough energy to free an electron. The band gap at which these
spectrum-driven losses are smallest in single-junction cells is about 1.4 eV . A range of band
gap values at which losses are still manageable extends from about 1.0 eV to 1.8 eV.
The single junction thermodynamic limit for solar cell conversion efficiency was
determined to be 32% by Shockley and Queisser , but the practical lab limit of
polycrystalline thin film solar cells is about 20% under 1-sun illumination. The major
assumption in the calculation of the theoretical limit is that electrons and holes created by the
absorption of photons with energies above the band gap lose their excess energy by phonon
One way to achieve efficiencies above the Shockley-Queisser limit is to use a series of
semiconductor pn junctions arranged in tandem configuration. The efficiency of solar cells can
be significantly increased by stacking several cells with different band gaps such that the gap
energy decreases from top to bottom. This multijunction cell uses more than one electric field to
separate electrons and holes. Light is incident on the top cell which has a high band gap.
Photons with energies greater than the band gap of the top cell are absorbed while those with
lower energies pass through to the next semiconductor where they are absorbed if their energy is
greater than the band gap of that cell. Thus, the solar spectrum is split so that photons are used
more efficiently; losses due to the mismatch between the energies of the photons and the cell's
band gap are reduced. Two cells in series connection have a maximum theoretical efficiency of
41.9% and with a larger number of cells, 50% efficiency can be exceeded . The
thermodynamic limit for solar energy conversion is significantly higher still, 66% at 1-sun and
86% at full solar concentration (46,200 suns), for an infinite tandem . The grand challenge is
to push solar cell efficiency towards its theoretical limit while maintaining low cost.
The efficiency benefit of a tandem solar cell to that of a single junction has been known for
quite some time, but it has only been practically observed in expensive crystalline III-V
materials. Multijunction cells under concentrated light have just recently exceeded 40%
efficiency (Spectrolab). Traditionally, they have been used to power satellites and other
spacecraft. The use of multijunction cells to generate clean energy for terrestrial applications has
been sought because, when combined with high concentration, multijunction cell modules have
the potential of producing the lowest $/watt amongst solar cell technologies .
Coutts et al. identified optimum band gaps for two-junction tandem thin film solar cells. A
current-matched, 28% efficient tandem is possible with a top cell absorber of 1.72 eV and a
bottom cell absorber of 1.14 eV . These band gaps are ideally matched to the CIS-CGS
material system. Low Ga content CIGS has the band gap and performance to be the low gap
cell. The wide band gap top cell material of the tandem is critical; it is estimated that
approximately two-thirds of the tandem cell efficiency originates there . A high band gap,
transparent top cell with efficiency greater than 17% is needed to form a tandem with an
efficiency of at least 25% .
In typical CIGS thin film solar cells, metallic Mo back contacts are used, which makes it
impossible for light to pass through this layer. However, a semitransparent solar cell is required
for the top cell of tandem devices. Nakada et al. report that the cell performances of CIGS
devices incorporating tin oxide (SnO2) and indium tin oxide (ITO) back contacts are similar to
those using molybdenum . The superstrate configuration, where the glass substrate is not
only used as a supporting structure, but also as a window for illumination, has an advantage of
easy and reliable encapsulation. Since the diffusion of sodium from the soda-lime glass is
strongly inhibited by the front contact in this design, Na-doping is necessary. The addition of Na
from co-evaporated Na2Se has been reported to more than double the efficiency in superstrate
Monolithic vs Mechanical
There are pros and cons to monolithic or mechanical tandems. In a two terminal device, the
stacked cells are connected at a common boundary where the bottom contact of one is the top
contact of another. Current flows continuously between the cells under illumination. Using this
monolithic approach, only one thick transparent conducting oxide (TCO), one grid, and one anti-
reflective coating (ARC) would be needed. However, current-matching and thermal stability
issues arise. The lowest current will limit the entire device so band gaps must be chosen that
split the spectrum equally: half of the sunlight absorbed on the top and half transmitted to the
bottom cell and absorbed there.
Several difficult technical issues need to be addressed in order for high efficiency
monolithic tandem cells to be developed. The bottom, first deposited, cell must not be destroyed
by the processing conditions of the top cell. High-efficiency CIGS devices are vulnerable to
temperatures greater than 2000C where diffusion destroys the pn junction. Therefore, a
successful tandem device fabrication procedure will require a bottom cell that is not affected by
the processing conditions of the top cell or a top cell that can be grown at a much lower
processing temperature . Some clever tandem structures are being investigated because of
the need to grow thin films at temperatures greater than 5000C to obtain high-quality absorbers
. Another critical issue for a monolithically interconnected tandem cell is providing a
transparent interconnect between the top and bottom cells .
The mechanical stack may appear much simpler, but there are other issues involved. In a
four-terminal device, each cell has a top and bottom contact connected to an external circuit so
their output is taken off separately. Performance of each cell is independent so the spectrum
doesn't need to be split between them. More materials (ARCs, TCOs, and glass) are needed for
the overall structure, which increases the cost .
1 1.5 2 2.5
Spectral irradiance versus wavelength under AMO and AM1.5 conditions.
Table 2-1. Efficiencies of copper chalcopyrites.
Material Band gap (eV) Theor. rl (%)
CulnSe2 1.0  25 
Culn0.72~Ga0.2Se2 1.1  27.5 
CulnS2 1.5  28.5 
CuGaSe2 1.7  26 
Achieved rl (%)
Chalcopyrite structure of CulnSe2.
0.5-1.5 Cpm ITO 3.7 eV
0.03-0.05 pm CdS 2.4 eV
1.0-2.0 CLm CGS 1.7 eV
01.01 11m tunnel ing junctio~n 3.3 eVc~
0.01 mITO 3.7 e~V
0.03-0.05 um CdS 2.4 eV
]1.5-2.0 Cpm CIGS 1.1 eV
0.5-1.0 I-m Mo
CIGS/CGS monolithic tandem device structure.
ABSORBER GROWTH AND DEVICE FABRICATION
Fabricating and testing a working solar cell requires multiple steps including equipment
calibration, the deposition of multiple layers, and film and device characterization. When the
source materials are depleted, the reactor must be shut down and brought to atmospheric
conditions. Copper (Cu) is annealed in a hydrogen furnace to remove the oxide film present on
the Cu pellets received from the supplier. The oxides have a higher melting point than copper so
their presence in the source may cause sputtering and thus non-uniformity in the flux
distribution. The indium (In) and gallium (Ga) source materials are introduced into the vacuum
system in the same condition that they are received from the manufacturer. To add Cu or In to
the reactor, the respective source shutter must be disconnected while an optical port must be
removed to add Ga.
After the source material is replenished, the Sentinel III rate controller, using Electron
Impact Emission Spectroscopy (EIES) sensors, is calibrated by Quartz Crystal Monitors (QCM).
Both the Cu and In sources are equipped with an EIES optical sensor located adj acent to the
rotating platen at substrate level and a QCM sensor located directly above the source center at a
fixed distance above the substrates. EIES is a system of evaporant excitation by electrons that
uses the optical intensity of the subsequent de-excitation as a means of process control. These
sensors are used for online measurement while the QCMs provide an absolute value of the flux
for the calibration of the optical sensors and cannot be used online because they are located
directly above the substrates. The Ga source is equipped with a single sensor, which is a QCM
that is in an identical position to the EIES sensors used for the Cu and In sources. QCM rate
control of gallium was inadequate so a temperature control scheme was implemented. Once the
Sentinel's parameters are adjusted, fi1ms can be deposited based upon the deposition rate sensed
by the EIES sensors.
Before a growth run series is initiated, the Cu and/or In deposition rates must be calibrated
with a specific Ga source temperature. Assumed Cu-rich, Ga-rich, and near-stoichiometric thin
fi1ms, typically 0.25-0.5 microns, are grown and the composition is measured through ICP
analysis. It is assumed that the Ga deposition rate is Eixed throughout a run since the temperature
is manually controlled to a specific value. Cu and/or In rates can be adjusted based on the ICP
results of the previous run. For example, an average Cu rate is determined from the previous run
and is divided by the Cu/Ga ratio verified from ICP to give the Cu rate needed to produce
stoichiometric CGS. The Cu rate for the current run can be adjusted appropriately to give the
desired overall composition. ICP feedback results must be maintained throughout a growth
series because reactor conditions may change. This procedure is only as good as the
repeatability of reactor conditions between successive runs.
Standard Growth Procedure
System startup is a lengthy process during which stringent guidelines must be followed to
ensure proper operation of the reactor. A cryotrap is filled with liquid nitrogen so that the reactor
chamber reaches a certain crossover pressure to safely switch the pumping to the diffusion pump,
which is necessary to get to high vacuum. As the trap fills, samples can be loaded into and
unloaded out of the reactor through the load-lock. After switching the system into high vacuum,
the ionization gauges are degassed for a minute to remove any deposits from them. The platen is
then started to its desired rotating speed, which is typically 12 rotations per minute. Platen
rotation must be started before the substrate heater is engaged so as to not warp it. The PMEE
reactor Supervisory Control Panel is then opened on the attached system PC. Film pre-
deposition parameters are set for the metals such as soak power, rise time, and soak time.
Heating layers are also set to determine the order in which the heaters are turned on and how
much power should be supplied.
When the appropriate parameters are entered into the LABVIEW program, heating can
begin. The pyrolytic boron nitride (PBN) substrate heater is started first and brought up to the
growth temperature. The inputted temperature is actually the temperature in the gap between the
heater and the platen as a thermocouple cannot be directly placed on the rotating platen. As the
PBN heater' s power is increased, the Cu tip is turned on and the power is gradually increased
manually. Before turning on the selenium source heaters, the cryoshroud that surrounds the
metal sources is filled with liquid nitrogen. The cryoshroud helps keep the excess Se in its
designated reactor zone. First the cracker is heated and then the crucible. As the Se crucible
approaches its final temperature set point, the metals primary heaters are switched on. The
practice for source preparation in this IVBE system before initiating deposition is to hold the
sources at a particular soak power for a set period of time. The metal sources first go through a
period of rising temperature and then a soaking period so that the solid metal sources become
melts. Since the Ga source is manually temperature controlled, control is taken over manually
after an initial rise time and the power is adjusted to reach the Ga temperature set point before
deposition. The Ga shutter needs to be manually opened when the other metals shutters open
automatically after their soak period. If a certain metal source is not being deposited in the film,
its heaters are not turned on and its shutter remains closed throughout the run. Selenium is
depositing on the substrates as they pass through the chalcogen zone as the Se crucible
temperature approaches its final value prior to the start of metals deposition. Startup time
leading up to deposition is approximately two hours.
The deposition time begins once the metal shutters are opened and metal beam fluxes are
impinging upon the rotating substrates. Cu and In rates are controlled by adjusting the local set
point and corresponding offset. Desired element deposition rates are set in advance and layer
thickness is controlled by adjusting open shutter time. The average deposition rate over a certain
period is calculated and these parameters can be adjusted appropriately to achieve the desired
composition. The Se source is kept at a constant temperature during evaporation. The power of
the Ga source is manually adjusted to maintain temperature control since QCM rate control was
ineffective. The reactor conditions are closely monitored with the cryoshroud periodically being
filled throughout the run. Different growth strategies can be administered by closing certain
metal shutters during the growth run; selenium is supplied in excess while the Cu and In rates
can be adjusted. Dopant NaF can be added at a constant rate for a set period of time. When the
appropriate thickness is reached, the metals' shutters are closed and the heaters are shut down.
Se can be deposited in an annealing procedure under the designated growth temperature for a set
period of time as the metal heaters cool down. Otherwise, power to the Se crucible and the
substrate heater are decreased at the end of metals deposition.
When all heaters are cooled to below 2000C, which typically takes at least an hour and a
half, the system is taken out ofHi-Vac. Once the system cools completely, the grown films can
be removed from the reactor by way of the load-lock. Films to be used for device fabrication are
immediately vacuum sealed to isolate them from the atmospheric conditions prior to buffer
deposition. The films are also re-sealed after the buffer layer is added and prior to ZnO
sputtering. One sample can be cut up and used for absorber characterization. Sometimes, the
samples were exposed to a normal room temperature air ambient for over a month between
deposition and analysis. This was also the case for absorbers used for device fabrication prior to
the purchase of a vacuum sealing system.
High quality absorber layers that are well-controlled are essential to the fabrication of high
efficiency solar cells. Thermal evaporation processes have been mainly designed on a basis of
experience and intuition to grow polycrystalline thin film layers . Therefore a fundamental
understanding of CIGS film deposition is necessary to design the best absorbers.
In the simplest single-step process, all rates as well as the substrate temperature are kept
constant during the whole process. A one-stage process typically produces low-quality material
when compared to the bilayer or three-stage processes. In CIGS growth, three-stage co-
evaporation leads to an absorber with a graded band gap, while single step co-deposition results
in a uniform band gap .
The first growth strategy used to synthesize highly-efficient CulnSe2 filmS was developed
at Boeing by Mickelesen and Chen. In the Boeing bilayer process, a two-phase film containing
CIS and Cu2-xSe is first deposited at low temperature and then reacted with a Cu-deficient flux of
co-evaporated Cu, In, and Se vapors at a higher temperature. The precursor deposition and re-
growth chemistry are shown in the following equations :
Cu(v) + In(v )+ Se(v) -CulnSe2(S) + CU2-xSe(s) (3-1)
CulnSe2(S) + CU2-xSe(s) + Cu(v) + In(v )+ Se(v) -CulnSe2(S) (3 -2)
Copper-rich material tends to form larger grains. This is typically true above
approximately 5250C because of a liquid phase assisted re-growth process due to the melting of
Cu2-xSe in the presence of excess Se. Thus, by first depositing a layer containing excess copper,
larger CIS grains are formed. In-rich layers generally have smaller grains, but when grown on
top of Cu-rich layers they are inclined to conform to the same growth pattern .
In the mid 1990s, NREL developed the three-stage process to grow high-quality CIGS
fi1ms . Indium, gallium, and selenium are evaporated at 2600C to form a (In,Ga)2Se3
precursor. The temperature is then ramped up to 5500C within a Se flux. At this point, sufficient
Cu is co-deposited with Se to make the fi1m Cu-rich. Additional In, Ga, and Se are added to
bring the overall composition back to Cu-poor. The amount of In and Ga deposited in the third
stage is usually 10% of the total in the first and third stages combined. The film is finally cooled
down within a flux of Se at about 3500C. The three-stage process is based on the following
reaction chemistry (precursor deposition, re-growth, and titration) :
In(v)>+ Se(v) In2Se3(S) (3-3)
In2Se3(S) + Cu(v )+ Se(v) -CulnSe2(S) + CU1-xSe(1) (3-4)
CulnSe2(S) + CU1-xSe(1) + In(v )+ Se(v) -CulnSe2(S) (3-5)
The intermediate Cu-rich growth stage has been shown to be beneficial to the morphology
and electronic quality of CIGS layers. The Cu2-xSe secondary phase has a higher emissivity in
the IR range than Cu-deficient CIGS, and the increased emission of heat radiation leads to a
lowering of the substrate temperature. Cu2-xSe segregations begin to appear once the film
reaches stoichiometric composition, i.e. [Cu]/[III] =1.00. This is reflected in a drop of the
substrate temperature, which is recorded by the thermocouple on the sample rear side. When the
substrate temperature is ramped up to 5500C after the completion of the first stage, the PID
temperature controller is cut off and a constant heating power is supplied . The third stage is
terminated when the temperature reading reaches the value recorded before the film became
stoichiometric during the second stage.
Growth Strategies in the PMEE reactor
Three different growth recipes were investigated for the deposition of CuGaSe2 absorber
layers. Another growth strategy was employed in the reactor during the earliest investigations of
CGS growth, an initial Cu-rich layer followed by a Ga-rich layer similar to the Boeing bilayer
process, but it never resulted in quality devices. We refer to each strategy as follows:
Constant Copper Flux Process (Figure 3-1A)
Modified Three-Stage Process (Figure 3-1B)
Emulated Three-Stage Process (Figure 3-1C)
The Constant Cu Flux Process is illustrated in Figure 3-la. The selenium crucible is
maintained at a constant temperature, typically 2650C, so that selenium is provided in excess for
all film growth in our PMEE reactor. Since temperature control is used for gallium deposition,
we maintained a specific gallium temperature for each growth run. Thus, we can change the
overall composition of the absorber by manipulating the copper flux. Calibration runs were
performed to determine the relationship between the average copper deposition rate and the
Cu/Ga ratio. This strategy simply keeps the same Cu flux over the entire growth run so that the
absorber maintains either Cu-richness or Ga-richness throughout the deposition of the film.
Figure 3-1b shows the Modified Three-Stage Process. This growth recipe starts by
depositing GaSe for a set period of time, followed by a Cu-rich layer, and ending with a Ga-rich
layer. Since the Ga temperature is maintained for the complete growth run, the Cu flux must be
adjusted to achieve either a Cu-rich layer or a Ga-rich layer. The overall composition and peak
Cu-richness can be adjusted by varying the duration of growth and the Cu flux employed for
The Emulated Three-Stage Process is illustrated in Figure 3-10. In contrast to NREL's
approach, our Emulated Three-Stage Process does not use end-point detection, but is based on
the composition results of previous runs. The thermocouple for substrate temperature
measurement is placed in the gap between the heater and the platen for temperature measurement
since it could not be placed in contact with the substrate because of the rotating nature of the
platen. An additional possibility to control the composition is to monitor the emissivity of the
films, but problems with pyrometry were presented by selenium condensation on the optical
ports. In this recipe, we deposit GaSe for a certain amount of time and then deposit CuSe until
we reach the desired thickness. The greatest copper-richness is reached at this point, and then
GaSe is deposited until the overall composition becomes Ga rich. The gallium temperature
remains constant throughout the first and third stages, while the same Cu rate is maintained
during the second stage.
Characterization of the absorber film is integral to the production of high quality devices.
Many research groups have implemented in-situ techniques to observe the growing film, but this
is not possible inside our reactor. Typically, one absorber is set aside strictly for characterization
purposes. The techniques described below are used extensively within this research.
Inductively Coupled Plasma-Optical Emission Spectroscopy (ICP-OES) is most commonly
used for bulk analysis of liquid samples or solids dissolved in liquids . The ICP operates on
the principle of atomic emission by atoms ionized in an argon plasma. Photons are emitted as
electrons return to the ground state of the ionized elements, which allows for the quantitative
identification of the species that are present. The strengths of ICP-OES include its speed, low
detection limits, and relatively small interference effects, but it is a destructive technique that
provides only elemental composition. Calibration curves must be made using a series of
standards to relate emission intensities to the concentration of each element of interest. The
Perkin-Elmer Plasma 3200 ICP system used in this study is located in the Particle Engineering
Research Center, University of Florida. This system is capable of analyzing materials with a
detection limit range of less than 1 part per million.
Known concentrations of Cu-In-Ga-Se dissolved in solution (0, 1, 5, and 10 ppm) are used
to create a calibration curve for ICP characterization. A small piece of the characterization
absorber, typically 2 cm x 1 cm, is dissolved in a 10 mL nitric acid solution. After the film
reacts for a few hours, the solution is diluted with 50 mL of deionized water. The overall
composition determined from this small sample may not be representative of the entire film if it
does not have uniform morphology.
Scanning Electron Microscopy (SEM) can be used to determine the grain size and shape of
absorber films . The SEM is commonly used for image analysis by focusing a source
electron beam into a fine probe and rastering over the surface of the sample. Secondary electron
and backscattered images are obtained to provide the surface topographical information. SEM,
using the SEM JEOL JSM 6400, characterization measurements were done at the Major
Analytical Instrumentation Center, Department of Materials Science and Engineering, University
X-ray diffraction (XRD) is a powerful technique used to uniquely identify the crystalline
phases present in materials and to measure the structural properties of these phases .
Polycrystalline thin films can have a distribution of orientations, which influences the thin-film
properties. When sizes of crystal grains are less than about 100 nm, x-ray diffraction lines will
become broadened. Hence, grain size can be estimated by measuring the broadening of a
particular peak. XRD is noncontact and nondestructive, which makes it ideal for in-situ studies.
Characterization measurements, using the XRD Philips APD 3720, were done at the Maj or
Analytical Instrumentation Center, Department of Materials Science and Engineering, University
The most important use of thin-film XRD is phase identification. XRD provides positive
phase identification by comparing the measured d-spacings in the diffraction pattern with known
standards in the JCPDS Powder Diffraction File. Some thin films have a preferred orientation,
but the JCPDS file contains measurements for films with random orientations so there can be
some disagreement between the measured values and the standard. For films possessing several
phases, the proportion of each phase can be determined from the integrated intensities in the
The most commonly used structure for CIS-based solar cells is the substrate configuration;
the absorber layer is evaporated on Mo-coated glass, and on top of this is a thin CdS buffer layer
and a transparent ZnO front contact. The universally accepted device design for fabricating high
efficiency, thin-film CIGS solar cells: MgF2/ZnO/CdS/CIGS/Mo/SLG is shown in Figure 3-2
Substrate and Back Contact
Soda-lime glass (SLG) is commonly used as the substrate for high-efficiency solar cells,
but it deforms at the temperatures used for highest device efficiencies, 550-6000C. Soda-lime
glass contains significant amounts (15.6 wt %) of sodium in the form ofNa20 , and it is
typically coated with molybdenum, which serves as the back contact. When the substrate
temperature approaches the softening point of the glass, Na ions diffuse from the glass through
the Mo back contact into the growing CIGS film. The extent of Na diffusion is related to the Mo
sputtering pressure. At low pressure, the amount of Na out-diffused from the SLG is low, while
at the highest pressure, the amount of Na out-diffused exceeds the optimal value required for
high-quality devices due the formation of microvoids and microcracks .
For good electronic device properties, the formation of an ohmic contact for the maj ority
carriers (holes) from the p-type CIGS and a low recombination rate for the minority carriers
(electrons) at the CIGS/back contact interface is essential. The back contact should be inert to
the highly corrosive environment during deposition and it must impede the diffusion of
impurities from the substrate into the absorber. Finally, a high optical reflectance is necessary to
minimize optical losses. Molybdenum, the historical back contact material for CIGS solar cells,
complies well with most requirements; it is inert during deposition, allows for the growth of
large grains, and forms an ohmic contact via an intermediate MoSe2 layer .
CIGS can also be deposited onto various substrates other than glass, even flexible ones.
Stainless Steel (SS) can be heated up over 5000C which is necessary to achieve high-quality
absorbers, but Na-doping is needed since SS doesn't contain sodium like soda-lime glass.
Unlike metal foils, polymer substrates are electrically insulating, which simplifies
monolithically-integrated module fabrication. However, polymer substrates have a limited
maximum operating temperature and can cause adhesion problems between the CIGS film and
the Mo back contact . Stainless steel substrates have generated CIGS films with 17%
efficiency  while applying a low temperature, 4500C, CIGS deposition process and a
reliable method for controlled Na incorporation on polyimide substrates has yielded cells with
14% efficiency .
Post Absorber Deposition
Exposure of CIGS absorbers to the atmosphere for significant amounts of time prior to the
buffer layer deposition step leads to surface oxidation. Yamada et al. observed large oxidation
rates for polycrystalline films; the surface of the fi1m is likely to be oxidized to a depth of a
couple nm in a brief time after removal from the growth chamber . Oxidation of the surface
can also lead to large changes in resistivity of up to three orders of magnitude in just a few days
. This shows clearly that it would indeed make a difference if the application of the buffer
layer and the completion of the solar cells is done immediately after absorber deposition, a few
days, or even a few hours later. Nadenau reported a dramatic decrease in cell performance if the
CGS layers were exposed to air for one day prior to the CdS deposition .
CIGS grown under a large Cu-excess condition contains copper selenide phases at the
surface and along the grain boundaries. A cyanide-based chemical treatment is used to remove
any secondary phases while being inert to the CIGS phase . The remaining material after
potassium cyanide (KCN) etching is expected to be the stoichiometric phase, but a rough surface
may remain . This type of morphology may affect the device performance adversely
because it can lead to poor metallurgical contact at the interface .
The main purpose of the buffer layer in a solar cell device is to act as barrier to diffusion of
impurities from the transparent conducting oxide layer into the absorber. Other benefits may
include interface passivation and the establishment of an inverted region in the absorber .
The properties of a buffer layer often depend on the deposition technique used in its fabrication
and the ability to control the growth parameters for each technique.
Most high-efficiency CIGS device structures employ a high-resistivity CdS buffer layer
deposited by chemical bath deposition (CBD). Because the band gap of CdS is too low (2.4 eV)
to permit transmission of all useful light, a balancing act is employed to optimize the structure.
If the CdS layer is too thick, unacceptable absorption occurs, leading to the reduction in short
circuit current (Jsc). If the CdS layer is too thin, shunt paths are generated, which leads to a
decrease in the open-circuit voltage (Voc) . The high efficiencies that have been achieved
by the CBD process are the result of a set of critical interactions that can produce n-type doping
or inversion, compositional grading, and interface passivation . The direct diffusion of
cadmium into the CIGS layer has been observed and this may lead to the formation of a buried
A CBD bath temperature at 800C instead of 600C, which is the common standard for
CIGS, is employed for CGS devices. The growth speed increase due to the elevated temperature
so the concentrations in the solution are modified. The quality of the buffer layer and the
interface with the absorber are improved by the 800C procedure . Chemical bath deposition
also affects the defects in the bulk of the absorber material so tunneling recombination is reduced
in samples grown with the higher CdS deposition temperature.
There are a few problems associated with cadmium sulfide technology. The band gap of
the CdS layer is still low enough to limit the short wavelength part of the solar spectrum that can
reach the absorber, and this leads to a reduction in the current that can be collected. This current
reduction becomes proportionally more severe for higher band gap cells . The substitution
of the heavy metal compound, CdS, is also desirable from an environmental and economical
point of view. A large-scale CBD-CdS buffer deposition process creates extra costs for the
necessary safety precautions needed for the handling and disposal of toxic material, especially
for such an inefficient and exceedingly wasteful process .
An alternative buffer layer such as ZnS is attractive due to its wide optical band gap and
reduced ecological issues. By widening the E, beyond that of CdS, a higher short-wavelength
quantum efficiency is expected in CIGS solar cells, thereby increasing the short-circuit current
. Using a ZnS CBD buffer, a champion cell of 18.6% was achieved by Hariskos et al.
Based on Anderson' s model of heterojunctions, the electron affinities of both layers should
match in order to obtain the maximum built-in potential and hence a high open-circuit voltage.
The electron affinity of CdS (4.5 eV) is larger than that of CGS (3.9 eV) so current loss can be
minimized by mixing ZnS with CdS to form a ternary compound ZnxCdl-xS in which the electron
affinity can be varied from 3.7 eV to 4.5 eV. Electron affinities are equal at about x = 0.78 and
any variation in the buffer layer composition from this optimum value results in a deterioration
of the cell performance . Voc increases with increasing Zn concentration whereas the Jsc
decreases. Ramakrishnu et al. produced a CGS cell with moderate efficiency (rl ~ 5 %) with a
Zn composition fixed at x = 0.5 .
Due to good lattice matching and to ideal electronic band offsets, ZnSe is expected to
provide a perfect buffer layer for CGS . For the CdS/CGS interface, the CdS conduction
band minimum is below that of CGS resulting in a "cliff" structure. Devices with this type of
band alignment show interface recombination dominated behavior, and hence suffer from a loss
in Voc. In order to avoid this effect, a material with a smaller valence band offset and a larger
gap is required. ZnSe has a band gap of 2.7 eV and the lattice constants of ZnSe and CGS are
closely matched, which should result in an almost strain-free interface . Rusu et al.
produced ZnSe/CGS heterojunctions with very high voltages, but poor conversion efficiency due
to very low current (Voc ~0.96 V and Jsc ~2 mA/cm2) .
From the aspect of band alignment, the II-VI buffer layer could be omitted, but solar cells
prepared with a direct CIGS/ZnO heterocontact show only poor efficiencies . Ramanathan
et al. observed that the direct sputtering of ZnO on CIGS typically yields only 2-5% devices
. These cells are characterized by enhanced current losses probably due to tunneling or
recombination processes via trap levels associated with impurities that diffuse into the absorber
during transparent conducting oxide (TCO) deposition .
The basic properties for making high quality transparent conductors are high conductivity,
high optical transmission, minimal surface roughness, thermal stability to withstand the
processing temperature, chemical stability, and crystallinity. Typical TCOs used in solar cell
fabrication have band gaps in the range of 3.3-3.8 eV, carrier concentrations in the range of 1020_
1021 CA-12, a conductivity of 104 (Ohm-cm)- a sheet resistance of about 5-10 ohms/square, and an
optical transmission greater than 85% over the visible part of the spectrum . Almost all of
the well-known TCOs that are used in solar cell devices, such as ZnO, In203, and SnO2 have n-
type conductivity .
In the fabrication of CIGS solar cells, it is customary to use a high/low resistivity grading
of the ZnO layer. An undoped layer of ZnO (high resistivity) is first deposited on CdS, followed
by the deposition of a doped layer. Ramanathan reports that solar cells made without the
undoped ZnO layer are identical to those made with the bilayer so the undoped ZnO layer may
be unnecessary even when CdS is very thin .
To collect the current, contacts are placed across the entire surface of a PV cell. This is
normally done with a "grid" of metal strips. However, placing a large, opaque grid on top of the
cell shades the active parts of the cell from the sun so they are designed with many thin,
conductive "fingers" spreading to every part of the cell's surface. The Eingers of the grid must be
thick enough to conduct well with low resistance, but thin enough not to block much of the
incoming light. Ni/Al grids are deposited by e-beam evaporation using a mask. Cell areas are
then delineated by mechanical scribing to give individual cell areas of 0.429 cm2 and a Einal In
contact is soldered on after the fi1m is scratched away to reveal the Mo back contact far away
from the grid.
Bare solar cells can reflect about 30% of the sunlight. Since the power output is
proportional to the amount of sunlight that is absorbed, these losses are detrimental to the device
performance. Surface reflection loss can be reduced by adding anti-reflection coatings (ARC) to
the solar cell. An ARC is typically deposited onto CIGS devices by the E-beam evaporation of
MgF2 to a thickness of 800-1200 Angstroms. The gain in short-circuit current is typically 4-8%
with a corresponding enhancement in conversion efficiency .
Current-voltage (I-V) analysis is a critical tool used to study solar cell performance. The
electrical parameters, including the conversion efficiency (rl), open-circuit voltage (Voc), short
circuit current density (Jsc), fill factor (FF), series resistance (Rs), shunt resistance (RSH), diode
ideality factor (n), and saturation current density (Jo), of a device can be determined from the
measured illuminated and dark I-V curves. The conversion efficiency is defined by
FF V, I,
4 = ocs (3-6)
PIN is the total power of incident light. Considering the general expressions for Voc and Isc, the
key material parameters that determine the efficiency of the solar cell are the lifetime and
mobility of the minority charge carriers and the surface recombination velocities .
The power that a cell provides is a product of its operating current and voltage. Under
short-circuit conditions, current is maximum, but voltage is nearly zero so almost no power is
provided to the circuit. Under open-circuit conditions, voltage is highest, but no current flows so
power is again zero. Fill factor is the percentage of maximum power as compared to the product
of open-circuit voltage and short-circuit current. Some cells can have good Voc and good Jsc,
but a poor FF, which results in not much power and a low efficiency. The series resistance can
affect the shape of the photo I-V curve, mainly the FF .
To make a good measurement, two other parameters are controlled: total power in the light
and the temperature of the cell. Standard power is 100 mW/cm2, which is approximately the
power density of sunlight at the Earth's surface at noon on a cloudless day. The cell is held at a
Eixed temperature of 250C because cell voltage and thus power output varies with temperature.
To get an accurate efficiency, the precise cell area must be known because the amount of input
sunlight depends on the cell area. Total area is the total area of the top surface of the cell, while
active area is the surface area of the cell without counting metal contacts even if those are on top
of the active portion of the cell. The solar cell to be measured is exposed to simulated sunlight,
and as a resistive load is varied from open-circuit voltage through short-circuit conditions, the
cell's I-V characteristics are measured .
I-V Measurement Technique
The reference cell method, which basically uses a reference cell to adjust the illumination
level of the solar simulator, is employed in the performance measurement of CIGS (and CGS)
solar cells in this study. The solar simulator intensity is adjusted by changing the distance
between the tungsten-halogen lamp and the test plane so that the measured Jsc of the reference
cell is equal to its calibrated value at the standard measurement intensity of 100 mW/cm2. We
use a CIGS solar cell calibrated against a primary reference cell and the global reference
spectrum by NREL to set the illumination level of the solar simulator.
The open-circuit voltage of CIGS solar cells decreases with increasing temperature at 100
mW/cm2. The temperature of the test cell is maintained at 250C & loC by a temperature
controller that circulates cooling water through the assembly during the illuminated I-V
measurement. The temperature controller of the cooling system is set at 200C to keep the
reading of the thermocouple and hence the temperature of the test cell at 25 A loC. The semi-
automated I-V measurement system is controlled by a personal computer with the data
acquisition and data analysis software LabVIEW .
Quantum efficiency (QE) is defined as the number of electron-hole pairs generated per
absorbed photon and is a measure of the effectiveness of a cell in converting light of various
energies into electricity . The cell is illuminated with monochromatic light while its electrical
output is being recorded. We know the number of photons in the monochromatic light and we
can measure the resulting electric current, which tells us how many electrons are being produced
by the cell. By slowly changing the monochromatic light to various energies, we can measure
the cell's response to the spectrum of solar photons. If the photons making up the
monochromatic light have less energy than the cell's band gap, they will pass through it without
producing any current (QE = 0). Just above the band gap of the cell, light will be very weakly
absorbed and unless the material's diffusion length is very large, the quantum efficiency will be
small. QE will begin to rise sharply as the energy of the incident photons is increased. In very
good cells, quantum efficiency of over 90% can be reached across most of the solar spectrum .
QE Measurement Technique
A spectral response measurement system employing a grating monochromator is used to
analyze the quantum efficiency of our solar cells. The monochromator, which is controlled via a
computer program written in LabVIEW, scans the spectral range from 400 to 1400 nm using 10
nm incremental steps. Two order sorting fi1ters are use to block the undesired harmonic terms
from the monochromator; one is applied for the range from 630 to 1000 nm and the other for
1000 to 1400 nm. The incident power density on the test plane is first measured by calibrated
silicon and germanium photodetectors, and it is saved on the hard disk of the computer. The
measured spectral response is calculated from the data stored in the computer previously and the
measured photocurrent of the test cell (Itest cel1(h)). Finally, the external quantum efficiency as a
function of the wavelength can be converted from the spectral response using the following
h-c-I ,, l(Al)
QE (l) = ts x 100% (3-7)
q Ai power densitydetector Ae test cell
where h, c, q, and h are Planck' s constant, the speed of light, the electronic charge, and the
photon wavelength, respectively.
0 50 100 150 200 250 300
0 50 100 150
0100 200 300 400
Figure 3-1. UF growth recipes. A) Constant Cu-Flux. B) Modified 3-Stage. C) Emulated 3-
Figure 3-2. Typical CIGS device structure.
COPPER GALLIUM DISELENIDE AB SORBER GROWTH
Like other I-III-VI2 COmpounds, CuGaSe2 has a wide phase stability region. The existence
range of CGS extends down to a Cu/Ga ratio of approximately 0.7 . Optical and electrical
properties are greatly affected by film composition because intrinsic point defects exist in the
material as its composition deviates from stoichiometry. It is not surprising that different
research groups have presented slightly different results depending on their specific preparation
method since the defects present are strongly dependent on the growth method and thermal
During the last few years, CGS films were grown in order to optimize the performance of
devices. Eight sets of CGS films were grown on Mo coated soda-lime glass substrates in the
PMEE system under different growth conditions. The specific PMEE reactor conditions for each
growth run described within this document are available in Appendix A. Growth temperature
and recipe were adjusted for the films and the growth rate fluctuated from about 0.4 to 0.9 8Js.
The thickness estimated for each run was varied from 1.0 to 1.5 Cpm based upon the total Cu
thickness sensed by the Sentinel III. This estimation method has been a better predictor of
thickness for films containing indium rather than gallium. ICP analysis of film composition
gives the parts per million of each species in the solution. Knowing the area of the
characterization sample and the density of each species allows you to determine a thickness
estimate of the film. There is of course error involved in the exact measurement of the
characterization sample's area, but this procedure estimates the actual thickness range to be
approximately 0.6 to 1.3 pm. Each table related to the specific growth run series summarizes the
growth process, as-grown composition and if available, the composition of the film after a KCN
etch for each of the films. The column labeled "growth process" shows the intended
composition of each layer of the film.
The first set of films, samples 443 through 447 shown in Table 4-1, was deposited at
386oC. They were grown mostly gallium rich by following a procedure similar to Boeing's
bilayer process: Cu-rich deposition followed by a Ga-rich layer. Three different compositional
stages were incorporated: an initial Cu-rich stage, an intermediate stage that varied from less Cu-
rich, to near stoichiometric, to Ga-rich depending on the desired overall Cu/Ga ratio, and a final
Ga-rich layer. The layers were approximately the same thickness with each constituting one-
third of the total thickness. The growth rate for these films was approximately 0.5 a/s and the
film thickness was estimated to be about 1.2 pm.
Films with Cu-rich initial stages seem to give poor quality absorber films when grown in
the PMEE reactor. We believed this may have to do with poor adhesion of these films to the
molybdenum, like in Stacked Elemental Layer processing where only gallium as the first layer
led to good adhesion of the absorber layer to the Mo back contact . Klenk et al. suggest
that Cu is not useful as the first material deposited onto the Mo as it caused severe adhesion
problems . Films that start and finish with Ga sequences should adhere better to the
substrate and result in a more uniform morphology . We intended to test this by changing
our growth procedure to either include a Ga-rich CGS initial layer or a thin GaSe layer.
The second set of films, samples 452-459 shown in Table 4-2, were also grown at 386oC.
The first part of this growth series consisted of films grown overall copper rich followed by
gallium rich samples. The intended growth procedure was a Modified Three-Stage Process. A
rather thin initial Ga-rich layer was deposited followed by a Cu-rich layer. Most of the films had
a thicker final Ga-rich layer deposited on them except for 452, which had only 2 stages, and 453,
whose Einal layer was composed of GaSe. This Ga-rich layer was typically half the total film
thickness. These fi1ms were grown at a similar rate to the previous set, but they were much
thinner, approximately 0.8 pm. Selenium annealing was performed for 30 minutes after metal
deposition was complete at the growth temperature for all the samples. The Se flux remained the
same as it was during metals deposition.
Copper selenide is a degenerate p-type solar cell found in CGS cells with overall Cu/Ga
ratios greater than one. If present, the Cu2-xSe phase will tend to short-circuit the device, not
allowing for a measure of device performance that is representative of the underlying absorber
material quality. It must be removed before deposition of the buffer layer . A cyanide etch
should remove any Cu2-xSe phase at the surface and between the network of grains in the fi1m so
we dipped the CuGaSe2 film in a 10% KCN solution for fiye minutes. After the removal of the
Cu-Se surface phase, the composition of Cu-rich CGS films has been found to be near
stoichiometry . It has also been shown to increase the photoluminescence intensity of Cu-
rich grown CIGS film up to five times by Keyes . However, the removal of the secondary
phase did not result in the Cu-rich sample becoming more like the (In, Ga)-rich samples since the
dominant defects and recombination processes are inherent to the CIGS phase. This result
should be analogous to CuGaSe2 Samples.
The third set of films, samples 472 to 480 shown in Table 4-3, consisted of films that were
mostly gallium rich or near-stoichiometric. For many of the films, an initial layer of gallium
selenide was grown. This was done to promote adhesion of the film to the substrate. The initial
GaSe layer was followed by a Cu-rich CGS layer, and a thin GaSe or Ga-rich layer was
deposited on top. For two of these films, the Emulated Three-Stage growth recipe was used.
The final GaSe layer was very thin so it is assumed that since the final composition is slightly
Ga-rich for these two samples that the film may have never been Cu-rich during the growth
procedure. The gallium primary temperature was increased to 10070C, compared to 9750C that
was used for the previous growth series, to increase the growth rate. The growth rate for these
samples ranged from 0.8 to 0.9 sis, except for the emulated three-stage process, which had a
growth rate of about half that value. The film thickness was approximately 0.9 pm. The same
final Se vapor treatment was performed after deposition.
In the PlVEE system, the substrates are radiatively heated by a resistive heater located
above the rotating substrate platen. This platen carries the substrates through each of the four
zones in the system, the metals zone, the fluxless load-lock zone, the chalcogen zone, and the
heater zone. A thermocouple is placed in the gap between the heater and the rotating platen for
temperature measurement. This temperature is then controlled during growth. Previous work
had been done to determine the relationship between the gap thermocouple temperature, T,, and
the actual temperature, Ts, of the substrates, yielding the relationship :
Ts = 0.5247 T, + 18.856 (4-1)
It had been previously believed that for the PlVEE system, using a T, greater than 7000C
may result in damage to the heater. However, the efficiencies of devices grown at this
temperature were relatively low. It was decided that a higher growth temperature was needed to
improve device efficiencies. After much investigation, it was concluded that the system could
likely be used safely at a higher temperature.
CGS solar cells with the highest performance were for a long time based on Cu-rich
composition, but the current record cells incorporate absorber films with an overall as-grown
composition that is gallium rich. The disadvantages of an overall as-grown Cu-rich composition
are a high doping density and a high concentration of electrically active traps [1 11]. Air
annealing mainly diminishes the density of deep traps in Cu-rich CGS . It is well known that
the grain size of Cu-rich chalcopyrite fi1ms is larger as compared to Cu-poor films . The
Cu-content of the fi1m determines the activation energy for grain boundary motion. Higher Cu-
contents lead to lower activation energies for grain boundary motion and therefore to the
formation of larger grains .
A Cu-rich growth period has been deemed to have beneficial effects on device
performance by some, but others have defined these benefits as limited to depositions at reduced
temperatures or times. Comparing different flux profiles, it was shown that at 4000C, Cu-rich
growth is necessary to achieve good performance in CIGS by Shafarman et al. At higher
substrate temperatures, device performance is insensitive to growth sequence allowing greater
process flexibility . Two-stage, rather than three-stage growth was utilized there. The
simultaneous deposition of group I and III atoms during the two-stage process may provide more
time for the necessary reactions and lessen the benefit of the Cu-rich growth period since the
benefit of the Cu-rich growth period has been surmised to come from a fluxing of the CIGS
grains by excess liquid Cu2Se .
The lack of quality absorbers obtained by deposition in the PlVEE from overall Ga-rich
composition is a perplexing phenomenon. It is possible that the fi1ms were not grown Ga-rich
enough, but many were in the compositional existence range of high quality absorbers. It is
possible that the processing conditions arrived at after years of optimizing the performance of the
low-Ga absorber material are not optimal for use with higher Ga-containing films, especially
when the Cu/III ratio falls much below 1 .
Since previous CGS absorber layers grown by PlVEE with overall Cu-rich compositions
had shown higher efficiencies, subsequent film growth series were grown copper rich. Table 4-4
displays the fourth set, which consisted of seven CGS films. Samples 510-516 were grown at T,
= 7500C, which corresponds to a substrate temperature of 4120C. These samples used growth
recipes that were similar to the ones that resulted in our previous best cells at a slightly elevated
growth temperature. The growth rate was similar to the previous runs, about 0.4 to 0.5 8Js, and
the films were slightly thinner at 0.6-0.8 pm. Cu-rich samples became nearly stoichiometric
after etching them in a 10% KCN solution for 5 minutes to remove the unwanted copper selenide
secondary phase. Se annealing was also performed for these films.
High substrate temperatures may be even more important for CGS deposition than for CIS
or low gallium content CIGS growth. Purwins et al. found that the formation of CIS is finished
before CGS starts to form at an elevated temperature of approximately 3900C . Most of our
film growth occurred at a substrate temperature near this lower limit of CuGaSe2 COnstitution.
Growth temperature seemed to be a limiting factor in producing high-quality films so a
much higher gap temperature of Tg = 900oC, corresponding to a growth temperature of 491oC,
was used for the fifth set of films, 521-525, shown in Table 4-5. These films had a similar
growth rate and thickness to the previous set. Film #523 was grown at a very low growth rate of
0.4 8Js to a thickness of only about 0.6 Gum. The final selenium vapor treatment for 30 minutes
occurred at the new elevated growth temperature of 4910C.
The growth temperature used while depositing CGS films in our PMEE reactor is lower
than ideal resulting in lower quality films. This may be a result of inadequate sodium
incorporation into the film. Na presence is due to diffusion from the soda-lime glass. During
absorber growth and potential annealing steps, Na diffuses through the Mo film into the
absorber, improving the doping concentration of the absorber. Ideally, a higher Na amount is
found closer to the Mo contact and the concentration gradually decreases in the bulk. Rusu et al
found a sodium concentration of I atomic percent at the surface . A moderate level of Na
improves the efficiency of the cells by enhancing the p-type conductivity. It incorporates into
the lattice of the CIGS by reacting with Se and forming Na-Se compounds . These
compounds slow the growth of CIGS and facilitate the Se incorporation into the film. Excessive
Na diffusion may limit the efficiency of the cells because of the introduction of additional deep
Table 4-6 shows a sixth set of films, samples 535-542, which were mostly grown copper
rich at 491oC. Growth runs 535-540 used an elevated gallium primary crucible temperature of
10050C to increase the Ga flux. Previous attempts at growing quality absorbers by the Emulated
Three-Stage Process were unsuccessful so the procedure was attempted again for some of these
films while maintaining an overall Cu-rich composition. The initial GaSe layer was grown at
450oC and the final two stages were deposited at 491oC. Films #541 and #542 used the same
processing sequence as our best absorber to date, #523. Se annealing, as described above, was
only performed on absorbers #541 and #542. The thickness of the films ranged from 0.6 to 0.8
ym while employing a growth rate of approximately 0.4 a/s.
Three-stage co-evaporation imposes stringent limits on the parameter space if highly
efficient devices are to result. The growth kinetics, substrate temperature profile, and reaction
time will make the outcome of local equilibria unique to the growth process. The Ga and Se
delivered in the third stage reacts with the CuxSe to form additional CuGaSe2 until the CuxSe is
consumed. Cu must diffuse out of the CGS grains to react with new Ga and Se while some Ga
will diffuse into the bulk grains to bring them to more Cu-poor compositions. By varying the
temperature during this stage, the counterdiffusion process can be enhanced or inhibited such
that the thickness and/or composition of surface Cu-poor phases can be controlled . The
evolution of the intrinsic defects depends on the dynamics of the reaction pathway, i.e. the
composition changes that occur when the film transitions from Cu-rich to Ga-rich . The
degree of overall Cu-richness or lack of Cu-richness that the film has after the second stage may
have a large effect on the absorber quality.
The three growth strategies described in the preceding chapter were incorporated into a
seventh set of runs, 628-641, that produced CGS absorber layers. The second column of Table
4-7 identifies the growth process utilized. For example, Sample #628 was grown using the
Constant Cu Flux Process described previously. When the Modified Three-Stage Process was
utilized, the second column of the table indicates the target composition of each sublayer of the
resulting CGS film. For example, Sample #634 has a growth process denoted as
"O/1.2/1.62/1.2". This indicates that the absorber film in this sample is composed of four layers.
In the first layer, the ratio of the copper flux to the gallium flux is zero, indicating that the
effective metal flux reaching the substrate was composed only of gallium. In the second layer
the ratio of the copper flux to the gallium flux is 1.2, indicating that there was a 20% excess of
copper relative to gallium reaching the substrate and thus this sublayer was grown under copper-
rich conditions. In an analogous fashion, the ratio values of 1.62 and 1.2 describe the
relationship between the fluxes imposed when growing the third and fourth sublayers. Finally,
Sample #640 incorporates an Emulated Three-Stage Process designated as "GaSe/CuSe/GaSe".
This indicates that the growth was done with the intent of defining three sublayers, where the
bottom most and the top sublayers are grown under gallium and selenium fluxes, while the
middle sublayer is grown under copper and selenium fluxes.
The third column of Table 4-7 indicates the overall ratio of copper to gallium content in the
final CGS film, as measured via ICP, and whenever applicable, the fourth column gives the
copper to gallium ratio in the film after a 5 min etch procedure in a 10% KCN solution carried
out to eliminate surface CuSe material. For example, Sample #629 had a copper-to-gallium ratio
of 1.17 as determined by ICP, and hence was a copper-rich film. The last column shows that
after the KCN etch procedure the ratio of copper to gallium in Sample #629 was reduced to 1.00,
putting the film in a stoichiometric Cu:Ga composition.
In every growth run the gallium source primary temperature was maintained at 9700C for
each run, and the substrate temperature was estimated to be approximately 4400C. Although an
elevated substrate temperature of approximately 4900C gave higher quality CGS absorbers, the
heat being generated was also warping the rotating platen, which led to scraping and erratic
rotational movement of the substrates. A gap temperature of 8000C, corresponding to a substrate
temperature of 4400C, was deemed safe so this is the maximum growth temperature that is used
in the PMEE reactor. Each absorber layer was estimated to be grown to a thickness of 1.5
microns except for Sample #641, which was grown to a thickness of 2 microns. The actual
thickness is more likely to be in the range of 1.1 to 1.3 Cpm for those grown to an estimated
thickness of 1.5 Cpm by analysis of the component masses over a defined film surface area. A
thickness of 1 micron is sufficient for the absorption of photons up to 750 nm, however thicker
layers result in a better performance of the solar cell . The film growth rate is estimated to
be about 0.8 AJs, except for those using the emulated three-stage process where the growth rate
is about half of this value.
Three samples were grown under the constant copper flux strategy: one was grown under
near a 1:1 (i.e., stoichiometric) Ga:Cu fluxes (Sample #628), one under Cu-rich conditions
(Sample #629), and one under Ga-rich conditions (Sample #630). Thicker versions of the
process that resulted in our best absorber, namely #523, were grown. Overall Ga-rich and near-
stoichiometric fi1ms were grown incorporating the modified three-stage process with varying
levels of peak copper richness. For this process, a thin initial GaSe layer was deposited followed
by Cu-rich layer and then a Ga-rich layer. The Einal two layers had similar thicknesses
incorporating half of the total film thickness.
Gallium rich samples in this growth series were slightly more selenium rich than KCN-
etched copper rich samples. Etched samples ranged from 0.490 to 0.494 selenium composition
while the overall gallium rich samples varied from 0.494 to 0.500. Epitaxial CGS films grown at
a substrate temperature of 5000C by Gu et al. showed a similar trend. The Se content value was
slightly higher than 50 at.% in the Ga-rich region and slightly lower than 50 at.% for the Cu-rich
A final set of films, 647-662, were grown at 4400C by the Constant Copper Flux Process
over a range of Cu/Ga ratios from approximately 0.9 to 1.25. Films #648, grown by the
procedure that produced our best absorber to date, and #649, grown by the three-stage process,
were included for comparison. All of the films were grown at a rate of 0.7 to 0.9 a/s, except for
#649, which was grown at a rate around 0.45 is. Approximate film thickness varied from about
1.0 to 1.3 pm.
A growth series of bilayer precursors was started, represented by growth run #666 in
Appendix A, but the Ga source was damaged during or following GaSe deposition. It is likely
that the gallium crucible cracked and the metal leaked out and shorted the source. The intent was
to grow a low temperature CuSe/GaSe stack that was to be rapidly thermally processed (RTP).
The effect of different growth conditions on the film morphology such as growth
temperature, overall Cu to Ga ratio and growth recipe were shtdied by using Scanning Electron
Microscopy (SEM). All CGS films having overall Cu to Ga ratio greater than one at some point
during the growth showed the morphology that had matrix and domain structure. In this
structure, the domain region showed highly Cu-rich composition and large grains while the
matrix region had small grains and stoichiometric or Ga rich composition. It means at the point
we had the overall Cu-rich composition for the film, there was a formation of liquid-like CuSe
secondary phase in the film and it made the grain size in the domain region larger than that in the
matrix region. As this kind ofinhomogeniety was observed in Cu-rich films with large grains
that lead to better efficiencies, the growth temperature was increased to get more uniform films.
Figure 4-2 shows that film #523 has better uniformity than the films grown by a similar
process at a lower growth temperature as seen in Figure 4-1. More uniform films are more likely
to produce high-quality CGS absorbers. Figure 4-3 and 4-4 shows the morphologies of films
grown with the similar Cu to Ga ratio profile but at different growth temperatures during the
growth. As seen in those figures, the grain size in the domain region appears to be largest at the
lowest growth temperature, while the grain size in the matrix region is the smallest. The grain
size in domain region was smaller at an intermediate growth temperature and got bigger at the
highest growth temperature. It appears that the grains in the matrix region got bigger as the
growth temperature increased. This means there might have been possible phase separation
between the Cu-rich domain region and Ga-rich matrix region at the lowest growth temperature,
and the film got more homogeneous as growth temperature increased. This improved bulk
crystal quality may be due to a more ideal incorporation of sodium into the film from the soda-
lime glass substrate at an elevated temperature .
Shafarman et al. showed that with Cu-rich growth of CIGS, the mean lateral grain area
decreases from 1.8 to 0.3 squared microns as substrate temperature is reduced from 550 to
4000C, but only at the highest substrate temperature does the grain size depend on growth recipe
. Films deposited at 4000C have a greater average sodium concentration than those
deposited at higher substrate temperatures. Thus, improved device performance with increased
substrate temperature cannot be explained by greater availability of Na. Films with smaller grain
size or a greater density of grain boundaries may have greater average sodium concentration
since nearly all Na probably resides along those boundaries.
In Eigure 4-5 and 4-6, the morphologies of films grown at the highest growth temperature
were compared for three different growth processes. Film #521 was grown by using a reverse
Boeing process, which used Ga-rich and then Cu-rich conditions. Film #523 was grown by the
Modified Three-Stage Process to utilize the liquid-like CuSe secondary phase to get larger
grains. Finally, film #525 was made by a process similar to Boeing's. As shown in those
figures, absorbers #521 and #523 have similar morphologies, both for the matrix and the domain
regions. And the grain size in domain region is only slightly larger than that in the matrix region
for those two films. For film #525, we could see that the grain sizes in two regions appear to be
much different from each other and there might have been possible phase separation again.
In figure 4-7, the morphologies of Cu-rich and Ga-rich films grown by the Emulated 3-
Stage Process were compared. Even though both films showed good homogeneity, the grains of
Cu-rich film are much larger than that of Ga-rich film. The Ga-rich film (#536) has very small
grains (~100 nm) while the Cu-rich film has larger grain size (300 ~ 900 nm). Films #541 and
#542 have a more uniform morphology, as seen in Figure 4-8, than those grown at lower
substrate temperature. They also didn't show the domain (large grain size region) and matrix
structure so it is assumed that devices made from both absorbers #541 and #542 will show good
Preference for a certain orientation seems to be dependent upon the growth recipe that was
followed in the PMEE reactor. Films grown by similar growth recipes, but at different substrate
temperatures, exhibit nearly identical XRD patterns. Shafarman et al. claim that XRD
measurements did not show any significant difference in the film orientation for different
processes or substrate temperatures. All their films had nearly random orientation . Figure
4-9 shows that films #511 and #522 both have a preference for the (1 12) orientation of CGS
while Figure 4-10 shows that #515 and #523 have comparatively less of a preference for this
orientation. Absorbers #515 and #523 were grown with an initial GaSe layer while #511 and
#522 had a Ga-rich initial layer and a final layer of GaSe. The films that had no initial Cu flux
during deposition had much more intense (220) peaks compared to the (204) peaks, whereas
films without this good adhesion layer seem to have slightly more intense (204) peaks than (220)
peaks. The full width half maximum of the (1 12) diffraction peak of film #523 is sufficiently
small indicating that the crystalline quality is fairly good. Some groups claim that there is a clear
correlation between higher (1 12) orientation and smoothness of the films , but that does not
seem to be the case for CGS films grown by PMEE at lower than ideal substrate temperatures.
Figure 4-11 shows two films with the same growth conditions, but grown at different
rotational speeds. The XRD patterns are very similar, but the film grown at the higher rotational
speed, #479, has slightly sharper peaks. A higher rotational speed may lead to larger grains. The
film grown at the lower speed, #476, also seems to have a stronger preference for (112)
XRD patterns also show which secondary phases are present. Films #452 and #455 were
grown at the same low growth temperature, but film #455 has an as-grown Cu/Ga ratio of about
1.4 while #452 has a ratio of 1.1. Figure 4-12 shows that the Cu2-xSe peaks are much more
intense for the very Cu-rich film. After the 10 % KCN etch for Hyve minutes, these Cu2-xSe peaks
disappear as shown in Figure 4-13.
As shown earlier, the growth recipe can have an effect on the preferred film orientation.
Films grown by the Constant Cu Rate Process have similar XRD patterns no matter the
composition. Figure 4-14 shows that the peaks for the KCN-etched Cu-rich film, #629, are
nearly identical to the Ga-rich film, #630. The extent to which a film goes copper rich in the
Modified Three-Stage Process can also have an affect on the pattern. Films #635 and #636 have
nearly identical compositions and similar growth processes, but #636 becomes more Cu-rich
during deposition. Figure 4-15 shows a greater (220) peak intensity for the film that has a lower
copper peak composition.
The Emulated Three-Stage Process produced absorbers that favored the (204) orientation
of CGS rather than (1 12), which is the more prevalent configuration for the other two growth
recipes. Figure 4-16 shows that Cu-rich and Ga-rich films employing GaSe deposition followed
by CuSe deposition have (220) CGS peaks that are more intense than even the (112) peaks. The
preference for these orientations may be due to the fact that the Emulated Three-Stage Process
deposits copper and gallium in separate layers, while they are deposited concurrently in the other
growth strategies. Yet, the XRD pattern for film #478 shown in Figure 4-17, which was also
grown by a three stage process, but at a lower growth temperature does not show these same
characteristic peaks. Due to the very thin GaSe final layer and overall Ga-rich composition, film
#478 may not have ever been Cu-rich and hence may have had different growth kinetics.
The surface morphology of films #628 through #641, grown at 4400C by various growth
processes, was investigated by SEM. Figure 4-18 shows the Constant Cu Rate Process of a Cu-
rich (#629) and a Ga-rich film (#630) at 100X magnification. Both Samples were grown under a
constant Cu flux throughout the entire deposition run with no Cu:Ga profile grading in any
sublayers. The Ga-rich film is very uniform while the Cu-rich film has many island structures.
Figures 4-19 and 4-20 show the Cu-rich film's grain structure in the field region and the island
region at 5000X and 10,000X, respectively. The field region shows very small grains and the
island structures have large grains up to a micron in size. The Ga-rich film morphology that is
shown in Figure 4-21 exhibits long needle-like grains. The Cu-rich films have Cu2-xSe
secondary phase on their surface. Figure 4-22 shows the island region and Figure 4-23 shows
the field region of film #634 before and after the 10 % KCN etch. The gaps left by the etching of
the copper nodules are very apparent in the island region while the field region appears to be
Co-evaporation of Ga-rich samples in an in-line deposition process revealed that a bilayer
process yields large, columnar grains, whereas a single layer process leads to absorbers with very
small grains. The bilayer-like process had a Cu-rich growth regime at the beginning and the
single-layer like process had constant rates throughout . Shafarman et al. showed that at
lower temperatures, the uniform flux process appears to give more columnar grains and a
smoother surface than CIGS films with a Cu-rich growth period. There is no apparent difference
between films grown with Cu-rich flux at either the beginning or middle of deposition .
The morphology of CGS films is strongly dependent on composition. Haug et al. observe
long and needle-like grains that are of small size for Cu/Ga~0.3. Grains of somewhat Ga-rich
CGS layers with Cu/Ga ratios between 0.9 and 0.7 are triangular, and layers deposited at higher
temperatures have an increased grain size. The Cu-rich layer consists of grains of irregular
shapes with a typical grain size of 1-3 microns (Cu/Ga~-1.1) .
Orsal et al. observed slightly different morphology . For slightly Ga-rich, there are
grains in the background with thin and long shaped grains starting to grow on the surface. With
an even higher Ga-content, the morphology is again homogenous and constituted of platelet-
shaped grains that are tilted on the surface. Cu-rich films exhibit small and homogenous grains
on the bottom with large polyhedral and packed grains on the surface. While it is not the case for
Cu-rich or Ga-rich films, the morphology is very sensitive near stoichiometric composition and
seems to depend on growth temperature and kind of substrate. Triangular crystallites are more
evident at 4500C with a grain size of approximately 0.6 microns. At 4000C, grains are
polyhedral whereas the layer is composed of a melt of triangular and polyhedral grains at 5000C.
Columnar growth is observed at each growth temperature.
Films 635-637 and #639 show some very peculiar surface morphology. The growth recipe
involved an initial GaSe layer followed by a Cu-rich layer and a Ga-rich layer. The overall
composition was either Ga-rich or near stoichiometric. The surface has islands surrounded by
rings that show a grain transition of large grains to smaller grains to long thin grains as can be
seen in Figure 4-24. Figure 4-25 shows the long needle-like grains present in the field region
that are the same as those in films grown Ga-rich by the Constant Cu Rate Process and the large
grains in the island region that are of similar size and shape to the ones found in the island region
of Cu-rich films grown by the Constant Cu Rate Process. It is likely that these films may have
copper selenide phases in the island region although the overall composition may be Ga-rich.
The sample taken for ICP compositional characterization may not have been representative of
the entire film since it is very non-uniform. This could be the case for film #639 since the Cu/Ga
ratio changed to 0.95 after the KCN etch from 1.04. Typically the film becomes nearly
stoichiometric, usually very slightly Ga-rich, after the cyanide etch. The pieces of the
characterization absorber used for the as-grown and post-etch composition analysis may have
started with drastically different overall compositions because of the non-uniformity of the film.
Smooth surface morphology is characteristic of films deposited by the three-stage process
. The Emulated Three-Stage Process produced very uniform films as can be seen in Figure
4-26. Figure 4-27 shows that the Cu-rich process that did not incorporate a final GaSe layer has
long tubular grains that have a larger axial diameter than the needle-like grains of the Ga-rich
films. A SEM picture on a 450 tilt in Figure 4-26 gives a clear view of these tubular grains. The
Ga-rich absorber film, #640 shown in Figure 4-28, seems to have somewhat triangular-shaped
grains that have not been produced by any other growth recipe used in the PMEE reactor.
Films 647-662 were mostly grown with a Constant Copper Rate Process to investigate the
difference in the absorber film properties based on composition. Figure 4-29 shows a slightly
different XRD pattern for the films grown the most Ga-rich like #647. The peak intensity for
(220) CGS is slightly greater than that of (204) CGS. Nearer stoichiometric Ga-rich films and
Cu-rich films, as can be seen in Figure 4-30, demonstrate the characteristics peaks of a constant
copper rate process that were seen in the previous growth series. Figure 4-31 also shows that the
Modified and Emulated Three-Stage Processes exhibit the same orientation as those in previous
In all the diffraction patterns of the CGS films grown in the PMEE reactor, the (220) and
(204) peaks are clearly separated, showing that the films have the chalcopyrite type
crystallographic structure. This was true for the near-stoichiometric and Cu-rich CGS films
grown by MBE by Yamada et al., but they also observed sphalerite crystallite structure for films
with Ga-rich composition . But this was for very Ga-rich (Cu/Ga~0.66) films, which we
never grew in these growth run series.
Growth temperature, growth recipe, and overall Cu/Ga ratio each had varying yet
substantial effects on the film morphology and orientation. Growth temperature seems to be the
most critical variable in achieving high-quality absorber films. The fact that Cu-rich absorbers
grown by PlVEE have been more successful than Ga-rich ones is likely due to the lower
deposition temperature. Ga-rich absorbers produce the highest efficiency CGS cells in the
literature, but they are also grown at an elevated temperature of at least 5500C. The intrinsic
defects produced at different processing conditions results in absorbers with distinct properties.
For the final growth series, we were able to maintain very consistent growth conditions
between runs in our reactor, which gave us great confidence in the compositional results for each
run, even though we lack in-situ measurement techniques. A packaging system allows us to
vacuum seal the absorbers after growth to lessen any degradation that may occur before the cell
has been processed. The biggest drawback in using the PlVEE reactor to grow polycrystalline
films is the limit that we must observe on the maximum substrate growth temperature. A more
effective technique may be to grow CGS bilayer precursors at a low substrate temperature and
then utilize an RTP system to rapidly raise the temperature for a brief period of time. Klenk et
al. performed a post-growth rapid thermal treatment at 5500C for 6 minutes on stacked elemental
layers that were deposited by evaporation at a low deposition temperature . Our research
group has been successful in the past employing an analogous strategy to grow CIS films via the
RTP processing of a bilayer ; we anticipate that the RTP processing route is likely to
produce similar satisfactory results for the CGS material.
Table 4-1. First CGS growth series.
Film # Process Cu/Ga ratio
443 1.1/0.9/0.7 0.89
444 1.15/0.95/0.75 1.00
445 1.15/0.95/0.75 0.96
446 1.3/1.1/0.9 1.13
447 1.15/0.95/0.75 0.98
Note: Process refers to the intended Cu/Ga ratio of each graded layer.
Table 4-2. Second CGS growth series.
Table 4-3. Third CGS growth series.
Film # Process
Table 4-4. Fourth CGS growth series.
Film # Process
Cu/Ga after KCN-etch
Cu/Ga after KCN-etch
Table 4-5. Fifth CGS growth series.
Film # Process
Cu/Ga ratio Cu/Ga after KCN-etch
Table 4-6. Sixth CGS growth series.
Film # Process Cu/Ga ratio
535 GaSe/CuSe/GaSe 1.04
536 GaSe/CuSe/GaSe 0.98
537 GaSe/CuSe/GaSe 1.12
538 0/1.3 1.58
540 GaSe/CuSe/GaSe 1.54
Note: These samples were etched by KCN, but the composition of the etched
samples was not measured.
Table 4-7. Seventh CGS Growth Series.
Film # Process
Note: Ga-rich samples were not KCN-ete
Cu/Ga after KCN-etch
Table 4-8. Eighth CGS growth series.
Film # Process
Figure 4-1. Morphologies of films grown at lower growth temperatures by similar growth
recipes (X1000). A) Tsub = 3860C. B) Tsub = 4120C.
Figure 4-2. Morphology of a film grown at a higher growth temperature (X3000). Tsub
Morphologies of the Cu-rich domain region of CGS films grown by the same
recipe at different growth temperatures (X30,000). A) Tsub = 3860C. B) Tsub'
4120C. C) Tsub= 4910C.
Figure 4-4. Morphologies of the Ga-rich matrix region of CGS films grown by the same
recipe at different growth temperatures (X30,000). A) Tsub = 3860C. B) Tsub
4120C. C) Tsub= 4910C.
A B C
Morphologies of the Cu-rich domain region of CGS films grown with different
growth recipes at 4910C (X30,000). A) (.8/1.25). B) (0/.9/1.45/.9). C)
Morphologies of the Ga-rich matrix region of CGS films grown with different
growth recipes at 4910C (X30,000). A) (.8/1.25). B) (0/.9/1.45/.9). C)
Morphologies of CGS films grown by the emulated 3-stage process at 4910C
(X30,000). A) Ga-rich. B) Cu-rich.
Figure 4-8. Morphology of a Cu-rich film (#542) with large grains and a uniform surface. A)
X3000. B) X30,000.
10 20 30 40 50 60 70
2 theta (degrees)
2000 -1 IICGS(220/204)
10 20 30 40 50 60 70
2 theta (degrees)
Figure 4-9. Diffraction patterns of films grown at different temperatures with the same
modified three-stage process. A) Tsub = 4120C. B) Tsub = 4910C.
Diffraction patterns of films grown at different temperatures with the same
modified three-stage process featuring an initial GaSe layer. A) Tsub = 4120C. B)
Tsub = 4910C.
10 20 30 40 50 60 7(
2 theta (degrees)
2 theta (degrees)
10 20 30 40 50 60 70
2 theta (degrees)
CGS 112) 47
10 20 30 40 50 60 70
2 theta (degrees)
Figure 4-11i. Diffraction patterns of films grown at different rotational speeds. A) 12 RPM. B)
2000 G (0)CGS(312)
Cu2-xSe I cU2-xSell CU2-xSe lCGS(116)
10 20 30 40 50 60 70
2 theta (degrees)
6000 -1 CGS(112)
2000 -1 IICGS(220/204)
1000 C2-xSe~l C~~ CU2-xSe
10 20 30 40 50 60 70
2 theta (degrees)
Figure 4-12. Diffraction patterns of films grown at different levels of overall Cu-richness. A)
Cu/Ga = 1.11. B) Cu/Ga = 1.40.