Defect Engineering in Cuprous Oxide (Cu 2 O) Solar Cells
by
ARCIVES
Sin Cheng Siah
MASSACH-ETT
M. S., (Mechanical Engineering)
Massachusetts Institute of Technology, 2013
B. Eng., (Engineering Science)
National University of Singapore, 2010
INSTITUTE
OFTENLLG
JUL 3 02015
LIBRARIES
Submitted to the Department of Mechanical Engineering
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Mechanical Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2015
Massachusetts Institute of Technology 2015. All rights reserved.
Signature redacted
N ......... ...
........I
..............
.
Author.............................................
Department of Mechanical Engineering
May 19, 2015
C ertified by.....................................
Signature redacted
Tonio Buonassisi
Associate Professor of Mechanical Engineering
Thesis Supervisor
Signature redacted
.........................................
David E. Hardt
Professor of Mechanical Engineering
Chairman, Department Committee on Graduate Theses
A ccep ted b y .......................................................
Defect Engineering in Cuprous Oxide (Cu 2 0) Solar Cells
by
Sin Cheng Siah
Submitted to the Department of Mechanical Engineering
on May 19, 2015 in Partial Fulfillment of the
Requirements for the Degree of
Doctor of Philosophy in Mechanical Engineering
ABSTRACT
This thesis is focused on the development of a cuprous oxide (Cu2O) thin-film (TF) solar cell
that is fabricated by manufacturing-friendly methods such as electro-deposition, sputtering and
atomic layer deposition. Due to its bandgap of close to 2 eV, it has the potential of being applied
as top cell in a tandem configuration. Firstly, I perform bottom-up cost and price analysis to
investigate the economic feasibility of TF and c-Si based tandem photovoltaic modules. Next, I
investigate the formation of good ohmic back contacts on Cu20 absorber layer and demonstrate
that low contact resistivity can be achieved with a variety of metals on heavily doped Cu20 films
by forming a tunnel junction. Then, I apply synchrotron-based X-ray absorption spectroscopy
(XAS) to characterize two front contact buffer materials: amorphous Zn-Sn-O (a-ZTO) and Sndoped Ga2O3. I elucidate a fundamental loss mechanism in the amorphous Zn-Sn-O (a-ZTO)
electron-blocking layer that has origin in local structural disorder and establish the structureprocess-property relationship of a-ZTO so that the front buffer layer can be optimized for
photovoltaics. Then, I investigate the doping mechanism of Sn dopant atoms in TFs and single
crystalline Ga2O3:Sn by revealing the doping mechanism so that Ga203:Sn can be optimized for
photovoltaics. Lastly. I apply bulk defect engineering to manipulate the intrinsic point defect
structure of Cu20 towards improved device performance. The key results will inform the
processing conditions for improving mobility and minority carrier lifetime in Cu20.
Keywords - Earth-abundant, thin-film solar cells, tandem, defect engineering, cost modeling,
synchrotron.
Thesis Supervisor: Tonio Buonassisi
Title: Associate Professor of Mechanical Engineering
3
4
ACKNOWLEDGEMENTS
I would like to express my utmost gratitude to my advisor, Professor Tonio Buonassisi, for
his guidance, and motivation over the past five years. I have not only picked up a tremendous
amount about research from Tonio, but also learnt a lot about being a good family man.
I am also thankful to Yun Seog Lee, Rafael Jaramillo, Riley E. Brandt, Jonathan P. Mailoa
and Michael Lloyd. They have each contributed in ways that have been invaluable to this
research project. Douglas M. Powell is acknowledged for his mentorship on cost modeling.
Sarah E. Sofia and Marius I. Peters are acknowledged for their contributions in the cost modeling
work. Sergio Castellanos is acknowledged for being an exercise and thesis-writing buddy. My
gratitude also goes out to all fellow colleagues of the PVLab for the constant intellectual
stimulations during group meetings and all the fun and joyful memories during the group's
recreational activities. Being so far away from home is a big challenge and would have been
much more difficult without the great support from the PVLab.
A big part of my PhD involves synchrotron-based X-ray spectroscopy techniques and I
would like to thank Grant B. Bunker for introducing me to the world of synchrotrons through the
XAS summer school in 2012. I am deeply grateful to Michael F. Toney, Carlo U. Segre and
Steve M. Heald for giving me the opportunities to conduct experiments at their beam-lines. I am
also grateful to Chengjun Sun, Tomohiro Shibata, Laura. T. Schelhas, Kipil Lim and Joshua
Wright for their help during synchrotron runs.
I am also extremely grateful to the National Research Foundation of Singapore for
supporting my graduate studies at a leading technology institute through a Clean Energy
Scholarship. Finally, I would like to thank my family, my girlfriend Serene Tan, and friends for
their unwavering love and support back home in Singapore.
5
6
CONTENTS
3
Abstract ......................................................................................................................................
Acknow ledgem ents.....................................................................................................................5
Contents......................................................................................................................................7
Figures ......................................................................................................................................
1
Citations to Published W ork......................................................................................................15
Chapter]: Introduction ..............................................................................................................
1.1
17
Photovoltaics as a Sustainable Energy Source ..................................................... 17
1.1.1
Terawatts Scalability: Earth-Abundant Thin-Film Solar Cells ........................ 18
1.1.2
State-of-the-art of Cu20 Solar Cells.............................................................
1.2
Thesis Overview ................................................................................................
21
22
Chapter 2: Bottom-Up Techno-Economic Analysis of Thin-Film and c-Si Based Tandem Solar
Cells..........................................................................................................................................25
2.1
A bstract...................................................................................................................25
2.2
Introduction.............................................................................................................25
2.3
Cost and M inim um Sustainable Price M odels......................................................
27
2.3.1
Bottom-Up Manufacturing Cost Analysis for PV Modules ..........................
27
2.3.2
M inim um Sustainable Price for PV M odules ...............................................
30
2.3.3
Calculation of Total System Installation M SP...............................................
30
2.4
Results and D iscussion.......................................................................................
31
2.4.1
Econom ic Feasibility of TF Single Junction M odules ..................................
31
2.4.2
Economic Feasibility of c-Si Based Tandem Modules ..................................
34
2.5
Conclusions.............................................................................................................38
Chapter 3: Principles of Defect Engineering ..........................................................................
39
3.1
Introduction.............................................................................................................39
3.2
Engineering Back Contact..................................................................................
40
Form ing Ohm ic Back Contact .....................................................................
40
3.2.1
7
3.2.2
Measuring Contact Resistivity and Schottky Barrier Height...........................42
3.2.3
Measuring Schottky Barrier Height..............................................................
43
3.2.4
Lowering Contact Resistivity Using a Tunneling Layer ................................
44
3.3
Engineering Front Hetero-Junction......................................................................
45
3.3.1
Importance of Recombination at Hetero-Junctions.......................................
3.3.2
Effects of Band Alignment and Interface States at Hetero-Junction............... 46
3.3.3
Measuring Band Alignments ........................................................................
3.4
Engineering Bulk Defects...................................................................................
45
47
48
3.4.1
Importance of Bulk Defects in Thin-Film Solar Cells ..................................
48
3.4.2
Extrinsic Doping in n-type buffer layers .......................................................
49
3.4.3
X-ray Absorption Spectroscopy as a Local Probe of Atomic Environment ........ 49
3.4.4
Bulk Recombination and Point Defects Structure in Cu20 ............................
51
3.4.5
Lifetime and Photoluminescence Spectroscopy.............................................
52
Chapter 4: Low Contact Resistivity of Metals on Nitrogen-Doped Cuprous Oxide (Cu2O) Thin-
F ilm s .........................................................................................................................................
55
4 .1
A b stract...................................................................................................................5
5
4 .2
Intro ductio n .............................................................................................................
56
4.3
Experimental Methods........................................................................................
57
4.4
Results and Discussions......................................................................................
58
4.4.1
Undoped
Cu20/Metal
Contacts:
Schottky
Barrier
and
Chemical
R eactiv ity ........................................................................................................................
58
4.4.2
Nitrogen
Doping
Reduces
Metal/Cu20
Contact
Height
Resistivity
and Changes
Conduction Mechanism ..............................................................................................
4.4.3
61
Nitrogen Doping Enables Low-Resistivity Contact for Earth-Abundant Metals 64
4.5
Application of Cu20:N Layer in Solar Cell..........................................................
4 .6
C onclu sio ns.............................................................................................................66
65
Chapter 5: Impact of Structural Disorder on Electron Mobility in Amorphous Zinc-Tin-Oxide
B uffer L ayers ............................................................................................................................
69
5 .1
A b stract...................................................................................................................6
9
5 .2
Intro d uctio n .............................................................................................................
70
8
5.3
Experim ental Details...........................................................................................
73
5.4
Results and Discussions ......................................................................................
74
5.5
Conclusions.............................................................................................................
8 1
Chapter 6: Dopant Activation in Sn-Doped Ga2O3................................................................
83
6.1
Abstract...................................................................................................................83
6.2
Introduction.............................................................................................................84
6.3
Experimental Details...........................................................................................
85
6.4
Results and Discussions ......................................................................................
87
6.5
Conclusions.............................................................................................................)9
1
6.6
Supplemental M aterial .........................................................................................
92
Chapter 7: Bulk Defect Engineering in CU20 ........................................................................
93
7.1
Abstract ...................................................................................................................
93
7.2
Introduction.............................................................................................................
94
7.3
Experimental M ethods .........................................................................................
95
7.4
Results and Discussions ......................................................................................
97
7.5
Conclusions...........................................................................................................100
Chapter 8: Conclusions ...........................................................................................................
103
References ..............................................................................................................................
105
9
10
FIGURES
Figure 1.1: Distribution of global primary energy demand by 2050 as predicted by the German
Advisory Council on Global Change, from Reference. 2 The scenario is predicted based on
extrapolation of current expansion rates of various renewable energies. ................................
18
Figure 1.2: Scatter plot of elemental market price against production volume. The estimated
production volume and elemental market price required to reach I TWp is indicated by the solid
lines. Adapted after Lloyd et al. ................................................
. . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
Figure 2.1: Typical manufacturing process flow of a TF solar cell.........................................27
Figure 2.2: Merging c-Si and TF cost models to yield a tandem module cost model. ............ 28
Figure 2.3: Sensitivity of estimated BoS price to module efficiency from SAM.
.......
........ . .
31
Figure 2.4: Cost breakdown and MSP for a TF module with q = 14.5% in terms of (a) process
step and (b) component. The capex for deposition of absorber is $10/M 2 and absorber-related
material cost is $5/M 2 . The MSP is estimated assuming WACC = 14%. ...............................
32
Figure 2.5: Sensitivity of (a) module MSP and (b) total system MSP of various TF technologies
to absorber deposition capex and module efficiency. The black dash line represents the module
MSP or system MSP for c-Si. The symbols represent several technologies and are estimated
based on their efficiency, estimated materials-related cost and estimated deposition capex. The
open symbols represent current record cell efficiency and the closed symbols represent the
typical m odule efficiency in the market................................................................................
33
Figure 2.6: Schematic drawings of a 2-terminal (a) mechanically stacked tandem module and (b)
monolithically integrated tandem module ..............................................................................
35
Figure 2.7: Sensitivity of (a) module MSP and (b) total system MSP of i = 30% tandems to
absorber deposition capex and module efficiency. The black dash line represents the module
MSP or system MSP for c-Si. The colored dashed lines are for single junction TF modules for
co mp ariso n ................................................................................................................................
11
36
Figure 2.8: Sensitivity of total system MSP of q = 30% tandems to absorber deposition capex as
a function of module efficiency. The black dash line represents the system MSP for q = 20.5% cSi. The colored dashed lines are for single junction TF modules for comparison. .................
37
Figure 3.1: Liebig's Law of the Minimum adapted to solar cells, from Ref.
40
28.
..........................
Figure 3.2: (a) SCAPS simulation for a metal/Cu20/Ga203/ZnO:Al solar cell with different work
function for the metal back contact and (b) device structure used for SCAPS simulations..........41
Figure 3.3: (a) Circular contact pattern for CTLM and (b) cross-section of CTLM. .............. 43
Figure 3.4: Schematic illustrating the effect of doping on holes transport across the M-S
interface, (a) thermionic emission, (b) thermionic field emission, and (c) field emission......45
Figure 3.5: (a) J-V characteristics of solar cells with and without a highly doped layer and (b)
energy band diagram at Voc condition, illustrating the narrow depletion width at the back contact
region, resulting in tunneling transport. The conduction band (Ec), electron quasi-Fermi level
(EFn),
hole quasi-Fermi level (EFp) and valence band (Ev) are indicated on the energy band
d iag ram .....................................................................................................................................
45
Figure 3.6: SCAPS-simulated results for (a) different interface states densities, and (b) different
conduction band offsets.............................................................................................................47
Figure 3.7: SCAPS-simulated J-V characteristics for Cu20-based solar cells with different buffer
d op ing d ensities. .......................................................................................................................
49
Figure 3.8: Equilibrium Fermi level and defect concentrations calculated for Cu-rich/O-poor
conditions and Cu-poor/O-rich conditions. The hole concentration and Fermi level are given both
in equilibrium at growth temperature Tg and at room temperature 298 K. This figure is from
Raeb ig er et al.
........................................................................................................................
52
Figure 4.1: Cu 2p core level photoemission for (a) Au, (b) Ag and (c) Pd samples. A peak shift
towards lower binding energies is observed for the Pd samples, indicating the lowering of the
SBH. A high binding energy shoulder due to CuO can be observed.......................................58
Figure 4.2: Illustration of method to determine the SBH from the XPS data from the valenceband spectra (right) and Cu 2p core level spectra (left) of (a) bare Cu20 film and (b) Cu20 film
12
with two nm Au overlayer. The Fermi level is calibrated using the Fermi edge of Au. All spectra
have been corrected for charging by using the adventitious C Is peak. .................................
59
Figure 4.3: Photoemission from metallic peaks of samples with 2 nm thick (a) Au, (b) Ag and
(c) Pd overlayers. A high binding energy shoulder due to PdO can be observed. ...................
60
Figure 4.4: Contact resistivity is plotted against Cu20 nitrogen doping concentration for three
contact m etals: A u, A g, and Pd.............................................................................................
62
Figure 4.5: Contact resistivity as a function of measurement temperature for Pd on CU20 with
three doping concentrations. Similar behavior (not shown) is also observed for Au and Ag
samples. The solid black line represents a fit to a thermionic emission model; the dashed red and
blue lines are guides to the eye. .................................................................................................
63
Figure 4.6: Plot of the ratio kT/Eoo as a function of hole density (p) with m*=3.36m0 and e =
7.OEo. The dotted line indicates that the ratio kT/Eoo = I and both thermionic emission and field
)
effect processes are comparable. kT/Eoo for both undoped ([N] = 0.0 at.%, p = 3.7x 1015 cm- 3
and lightly ([N]= 0.6 at.%, p = 1.8x1018 cm-3) doped samples are indicated on the plot.....64
Figure 4.7: Contact resistivity as a function of measurement temperature for Cu, Ni and Pd on
highly doped ([N] = 1.2 at. %) samples. The dashed lines are guides to the eye. ....................
65
Figure 4.8: (a) A cross-sectional SEM image of a Cu20-based TF solar cell with a Cu20:N holetransporting layer. Dashed
lines (white)
indicate interfaces between layers, and (b) J- V
characteristics of a Cu2 0:N and a control Cu20 device under I sun illuminated (AM 1.5 G, 100
mW cm-2 ) condition. Figures are from Lee et al.6 1...............................
. . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
Figure 5.1: (a) J-V characteristics of Au/Cu 2 0/a-ZTO/ZnO:AI/Al solar cells fabricated with aZTO of different compositions and (b) Relative alignments of conduction band (CB) and valence
band (VB) for a-ZTO and ZnO overlayers to Cu20 TFs as measured by XPS. Plots reproduced
fro m L ee et al.13 .
. .. .
.......................................................................................................
71
Figure 5.2: Capacitance-frequency measurements for Au/Cu20/a-ZTO/ZnO:Al/Al solar cells
fabricated with a-ZTO of different compositions. ................................................................
72
Figure 5.3: Fourier-transformed EXAFS spectra at Zn and Sn K-edges. Figures (a) and (b)
compare the peak intensity of the INN shell at the Zn and Sn K-edges respectively. The arrows
illustrate the decreasing peak intensity with increasing [Sn]. Figures (c) and (d) show the
13
measured (symbols) and fitted spectra (lines) at the Zn and Sn K-edges respectively (offsetted for
c larity ). .....................................................................................................................................
76
Figure 5.4: XANES spectra at (a) Zn and (b) Sn K-edges. The 2 isosbestic points are circled and
labelled A and B in (a)..............................................................................................................78
Figure 5.5: Field-effect/Hall-effect mobilities, device efficiency and pseudo-Debye-Waller
factors are plotted against [Sn]/([Sn]+[Zn])
..........................................................................
79
Figure 6.1: Net carrier density of P-Ga2O3:Sn single crystals as a function of varying [Sn],
compared against the deduced upper limit for a-Ga203:Sn films...........................................86
Figure 6.2: Sn edge XANES spectra for (a) SC fl-Ga203:Sn, (b) PLD a-Ga2O3:Sn and (c) ALD aGa203:Sn samples. The dashed, dashed-dotted and dotted lines represent Sn(0) metal (blue),
Sn(lI)O (green) and Sn(IV)0 2 (purple) references respectively.............................................88
Figure 6.3: Fourier-transformed EXAFS spectra plotted as the magnitude, IX(R)I, for (a) Ga and
(b) Sn K-edges. The spectra for SnO and Sn02 are included as reference...............................89
Figure 7.1: (a) A typical sample under illumination by microscope light. Sample is glowing red
due to its bandgap. (b) A microscope image of a similar as-grown foil under 50x magnification.
Scale bar represents 1000 tm ....................................................................................................
96
Figure 7.2: Time-temperature profiles for the growth and annealing process of Cu20 samples.
(N ot draw n to scale)..................................................................................................................96
Figure 7.3: Measured sheet resistance, Hall-mobility and effect hole concentration for samples
grown with varying Tanneal. The lines are guides for the eye. ..................................................
98
Figure 7.4: [t-PCD decay transients for all samples. A double exponential behavior can be
observed for samples at low Tanneai.............................................................................................98
Figure 7.5: (a) Spectrally-resolved PL for sample with no annealing, or Tanneal = 24"C and (b) PL
imaging of all samples. The defect-band that is being imaged is shown schematically in (a)......99
Figure 7.6: (a)
TPCD
plotted
TpCD
as a function of the inverse of spatially-integrated PL intensity (PLI) and (b)
as function
of NA- ND
.....
'.''.'''..''''''''''''''..''....................................................100
14
CITATIONS TO PUBLISHED WORK
Parts of this dissertation cover research reported in the following articles:
[1] S. C. Siah, Y. S. Lee, Y. Segal, and T. Buonassisi, "Low contact resistivity of metals on
nitrogen-doped cuprous oxide (Cu20) thin-films," J. Appl. Phys., vol. 12, no. 8, p. 084508,
2012.
[2] S. C. Siah, S. W. Lee, Y. S. Lee, J. Heo, T. Shibata, C. U. Segre, R. G. Gordon, and T.
Buonassisi, "X-ray absorption spectroscopy elucidates the impact of structural disorder on
electron mobility in amorphous zinc-tin-oxide thin films," Appl. Phys. Lett., vol. 104, no. 24,
p. 242113, Jun. 2014.
[3] S. C. Siah, R. E. Brandt, L. T. Schelhas, K. Lim, J. D. Perkins, R. Jaramillo, M. D.
Heinemann, D. Chua, J. Wright, C. U. Segre, R. G. Gordon, M. Toney, T. Buonassisi,
"Dopant activation in Sn-doped Ga2O3 investigated by synchrotron-based X-ray absorption
spectroscopy," (in preparation for submission).
[4] Y. S. Lee, D. Chua, R. E. Brandt, S. C. Siah, J. V Li, J. P. Mailoa, S. W. Lee, R. G. Gordon,
and T. Buonassisi, "Atomic Layer Deposited Gallium Oxide Buffer Layer Enables 1.2 V
Open-Circuit Voltage in Cuprous Oxide Solar Cells," Adv. Mater., vol. 26, no. 27, pp. 47044710, May 2014.
[5] Y. S. Lee, J. Heo, M. T. Winkler, S. C. Siah, S. B. Kim, R. G. Gordon, and T. Buonassisi,
"Nitrogen-doped cuprous oxide as a p-type hole-transporting layer in thin-film solar cells,"
J. Mater. Chem. A, vol. 1, no. 48, p. 15416, 2013.
[6] Y. S. Lee, J. Heo, S. C. Siah, J. P. Mailoa, R. E. Brandt, S. B. Kim, R. G. Gordon, and T.
Buonassisi, "Ultrathin amorphous zinc-tin-oxide buffer layer for enhancing heterojunction
interface quality in metal-oxide solar cells," Energy Environ. Sci., vol. 6, no. 7, p. 2112,
2013.
15
16
CHAPTER
1
INTRODUCTION
1.1
Photovoltaics as a Sustainable Energy Source
During the industrial revolution in the 18th century, there was a heavy demand for
mechanical machines to drive manufacturing processes. Demand for energy in the form of fossil
fuels to power these machines increased dramatically and since then, mankind never looked
back. Global energy consumption is now estimated to be more than 1.7x 1014 kW-h and is
expected to grow by 53% over the next 20 years.' While standard of living has improved as a
result of this revolution, continued unsustainable use of resources to meet our insatiable thirst for
energy has resulted in undesirable impacts on Earth's environment. The unsustainable use of
resources has led to a plethora of problems that include air and water pollution, depletion of
natural resources and global climate change. According to a report published by the International
Panel on Climate Change (IPCC), high-accuracy measurements have provided evidence for the
direct correlation between warming of the Earth's atmosphere to human activities. 1 High profile
reports by climate champions such as former U.S. Vice president Al Gore and the IPCC have
strongly advocated for the mitigation of these negative impacts on the Earth's climate by moving
towards a future that is built upon energy sustainability.
One viable option to reduce mankind's reliance on conventional energy sources is to adopt
renewable energies as alternatives. As illustrated in Figure 1.1, the German Advisory Council on
Global Change envisions several renewable energy options that are expected to be more
17
important by 2050.2 It can be seen that conventional fossil fuels are not expected to be totally
replaced but their overall contribution is expected to decrease significantly over the next 40
years. On the other hand, renewable energies are expected to increase their relative contribution
to the energy distribution. In particular, solar is an attractive alternative due to the vast amount of
energy Earth receives from the Sun every day. To put things into perspective, the Earth receives
about 5.5x 1017 kW.h of energy from the Sun annually. This is about 3200 times the world total
energy consumption in 2013. Solar energy can be harvested in various forms including heat
(solar thermal) and electricity (photovoltaics). In this thesis, the main focus will be on
photovoltaics (PV). The goal of solar PV is to convert a fraction of this enormous resource from
the sun into electricity.
700Primary energy saving
600
Savings through:
500
CHP and heat pumps
400
Direct energy generation
(wind, solar, hydro)
E Wind and solar
generated gas
M Solar power
M Wind
* Solar heat
M Geothermal power
3 Hydropower
M Biomass heat
l Biomass power
E Nuclear energy
E1 Natural gas
M Crude oil
M Coal
200
100
0
1970
1980
1990
2000
2010
Year
2020
2030
2040
2050
Figure 1.1: Distribution of global primary energy demand by 2050 as predicted by the German Advisory Council
on Global Change, from Reference. 2 The scenario is predicted based on extrapolation of current expansion rates
of various renewable energies.
Terawatts Scalability: Earth-Abundant Thin-Film Solar Cells
1.1.1
In 2014, worldwide PV production is about 38.7 GWp, 3 and solar cells based on crystalline
silicon (c-Si) make up over 90% of the industry. The remainder of the PV market consists of
18
solar cells based on thin-film (TF) technologies such as amorphous silicon (a-Si), cadmium
telluride (CdTe) and copper indium gallium diselenide (CIGS). Work by Powell et al.4 has
shown that in c-Si PV manufacturing, upfront capital cost or capital expenditure (capex)
associated with producing polysilicon as well as material cost associated with feedstock silicon
can be substantial. For example, the capex associated with building a 400 MWp c-Si solar cells
manufacturing factory can be more than $272 milliona. The cost of the silicon feedstock is about
$0.10/Wp or >15% of the total cost for c-Si PV.4 The high capex, high feedstock cost and
extensive silicon requirement associated with silicon PV might eventually be an impediment to
terawatts (TW) scalability.
The low cost per unit area of TF technologies makes it a leading candidate to achieve lower
than $0.50/Wp module cost. It also has improved potential to reach TW scalability due to a I00x
reduction in material requirement for the light-absorbing layer. However, conventional TFs solar
cells using CdTe and CIGS absorbers could be limited to sub-terawatt deployment due to
material scarcity (Te and In) and toxicity (Cd) issues. 5 To guide material selection for the TF PV
absorber layer, Lloyd et al.6 estimated the material production and price required to achieve
terawatts scalability. Figure 1.2 plots the selling price of materials against their production in
2010. To achieve I TW production annually the annual production of the chosen material should
be more than 1500 tonnes per year and the elemental selling price of the material should be less
than $250/kg (assuming a 2 ptm absorber). By assuming a binary semiconductor consisting of an
Earth-abundant cation metal (- $250/kg) and anion chalcogen or oxygen (- $1/kg), the upper
a Default
currency referred to in this thesis is U.S Dollars.
19
bound of the material cost for a 2 pm thick absorber layer is estimated to be $2.5/M 2 (yield
=50%).
Elemental Market: Modes of Production
106
Os
0
* Primary
105
U
Secondary
Primary/Secondary
Rb
104
TM
big
4
Re Ge toT
xe
S103
$250/kg
Kr
102
H
Nb
10
o
C
100
so
100B
B
101O
10-2
10-2
Na
M
Ca
10.1
100
101
102
103
104
10S
106
107
108
109
1010
10"
2010 Production (Tonnes)
Figure 1.2: Scatter plot of elemental market price against production volume. The estimated production volume
and elemental market price required to reach I TWp is indicated by the solid lines. Adapted after Lloyd et al.6
Based on a list of low-cost sustainable elements, Lee et al.7 performed first-principles
calculations to search for possible compound semiconductors using a combinatorial method.
Subsequently, extensive literature review and further first-principles calculations are used to
narrow down the list and eventually, identified cuprous oxide as a promising candidate among
others like zinc phosphide (Zn 3P 2 ), copper nitride (Cu3N), silicon diphosphide (SiP 2 ).8 While
there is certainly a long list of semiconductors that satisfy the criteria mentioned above, the
technical work performed in thesis will primarily focus on engineering and enhancing Cu20based solar cells and maximizing its potential as a future PV material. In principle, the structured
20
approach to improve performance presented herein can be generalized to other promising
compounds.
1.1.2
State-of-the-art of Cu20 Solar Cells
Prior to 2006, the efficiency of Cu 20-based solar cells remained below 2%. The main reason
for low performance, was the inability to find a good conjugate n-type partner for Cu20. In 1983,
Olsen et al.9 demonstrated a conversion efficiency of 1.8% using Cu20/Cu Schottky junction
solar cells. These Schottky cells are based on polycrystalline Cu20 wafers grown using thermal
oxidation of copper foil and demonstrated carrier collection as high as 10 pm. The relatively
high collection length resulted in short-circuit currents (Jsc) as high as 8 mA/cm 2, and it was
estimated that Jsc could reach at least 12 mA/cm 2 by incorporating a more transparent front
contact. On the other hand, the low open-circuit voltage (Voc) of below 400 mV was attributed to
the formation of a copper-rich region at the hetero-interface which limited barrier height and
consequently, the
Voc. It was hypothesized by Olsen et al.9 that the development of
homojunction or better control of interfacial properties would be critical for the success of Cu 2 0.
In 2006, Mittiga et al.' 0 improved the Voc of Cu 20-based solar cells to 600 mV by incorporating
a zinc oxide (ZnO) buffer layer to form a hetero-junction solar cell. Subsequently, the use of zinc
magnesium oxide (a-Znu.x)MgxO),1,
2
amorphous zinc tin oxide (a-ZTO),' "4 and amorphous
gallium oxide (a-Ga2O3)1,1 6 as buffer layers have contributed to Voc improvements over the
years. In particular, a record Voc of 1.2 V was achieved using a-Ga2O3 grown by atomic layer
deposition on electrodeposited Cu20."
This thesis is mainly focused on elucidating the physics that is driving these recent
improvements, and to provide outlook for future improvements. We note that tandem devices are
21
particularly sensitive to top-cell performance, because the top cell supplies roughly two thirds of
the tandem output power.
1.2
Thesis Overview
To provide the thesis with a broader perspective, Chapter 2 focuses on addressing scalability
issues pertaining to TF PV manufacturing. A cost model for a hypothetical TF PV manufacturing
factory is developed and the impact of absorber-related capex and material costs on the minimum
sustainable price of TF modules is investigated. Subsequently, the model is extended c-Si or TF
based tandem solar cells.
Chapter 3 provides a brief overview of defect engineering concepts that are relevant to the
scope of the thesis. Results from SCAPS simulations are shown to highlight the importance of
back contact, front contact and absorber bulk engineering and to provide a framework for
improving the performance of Cu20 solar cells. In addition, brief descriptions of different
characterization techniques are presented.
Chapter 4 investigates the formation of low-resistivity ohmic back contacts to Cu20. Contact
resistivity of various metals, both precious and Earth-abundant, is quantified. A scheme of
introducing a highly doped Cu20 as tunneling layer is introduced to significantly reduce contactresistivity for metals that form a Schottky barrier at the back contact.
Chapter 5 elucidates the physics behind the relationship between device efficiency and bulk
electronic and structural properties of a-ZTO buffer layers. Band alignment measurements are
performed to relate device efficiency to energy band structure at the hetero-junction. In addition,
synchrotron-based X-ray absorption spectroscopy is used to relate structural properties of a-ZTO
films to mobility and solar cell performance.
22
Chapter 6 discusses the physics underlying a high Voc obtained with a-Ga203 buffer layer
and emphasize the importance of doping the buffer layer to further improve efficiency. In
addition, synchrotron-based X-ray absorption spectroscopy is used to understand the doping
mechanism of Sn-doped Ga203 so as to help guide the development of highly conductive Ga203
thin films.
Chapter 7 examines the relationship between growth process, electronic properties and
defect structure of Cu20. Complementary characterization techniques such as spectrally-resolved
photoluminescence,
defect-band
photoluminescence
imaging,
microwave
reflection
photoconductance decay lifetime measurement and Hall measurements are used to derive useful
information about the bulk electronic properties of Cu20.
Lastly, Chapter 8 summarizes the main results of this thesis.
23
24
CHAPTER
2
BOTTOM-UP TECHNO-ECONOMIC
ANALYSIS OF THIN-FILM AND C-SI
BASED TANDEM SOLAR CELLS
2.1
Abstract
The cost advantages of thin-film (TF) and c-Si based tandem solar cells are not well studied
and are less obvious as compared to conventional single-junction solar cells. In this chapter,
module minimum sustainable selling price (MSP) and total photovoltaic (PV) system MSP
models are developed for both types of solar cells with an emphasis on understanding the impact
of TF or top cell absorber deposition on economical feasibility. Without restricting the analyses
to any particular material system but generalizing it more broadly to various manufacturing
approaches, the costs associated with TF absorber growth techniques are surveyed and
investigated. The sensitivity of MSP and total system installation MSP for TF and tandem
modules towards module efficiency and capital expenditure (capex) is investigated. It is found
that balance of system (BoS) costs drive the economic feasibility of high-efficiency approaches
and make high-efficiency c-Si based tandems more financially attractive.
2.2
Introduction
The cost-competitiveness of electricity generation is determined by the amount of money it
costs to generate one watt of electricity. For PV, it has been recognized that the following
25
simplified expression relates cost of PV to manufacturing and installation costs and module
performance:1 7 '1 8
2
2
Cost of PV [$W] = Manufacturing Cost [$/M ] + Installation Cost [$/m and $/W,]
Module Efficiency [W/m 2 ]
(2.1)
Ultimately, the cost of PV plays a huge role in determining the economically feasibility of PV
technologies and there exist many pathways to lower the cost of PV. In this chapter, two
promising ways are addressed: 1) TF solar cells with low manufacturing and upfront capital costs
and 2) c-Si based high efficiency tandem solar cells. These two approaches are not necessarily
mutually exclusive because cost-effective TF solar cells can be used to complement existing c-Si
solar modules by operating in tandem.
The chapter will begin by introducing the bottom-up cost assessment approach which is an
accounting exercise used to sum up every cost components across the PV value chain. Once the
bottom-up cost accounting is completed, the MSP can be estimated by additional financial
considerations.4,19-21 The emerging view of the topic is that capex is one of the main drivers of
MSP due to cost of capital associated with an upfront capital investment. For a fixed module
selling price, a higher capex will reduce the profit margin and leads to a slower growth rate for
the manufacturing company. In light of this, the subsequent part of this chapter will be dedicated
to understanding the influence of capex on MSP for various technologies. A strong focus will be
made on the growth of TF absorbers because other module-related costs like electrodes
deposition, laser patterning, encapsulation, framing and wiring are expected to be standard across
different TF technologies. Without restricting the analysis to any particular material system but
generalizing it more broadly to manufacturing approaches, the capex and materials cost
associated with TF absorbers fabricated using various techniques are surveyed and compared.
These techniques are broadly classified into four different categories: solution-based processing,
26
high-rate physical and chemical deposition, low-rate physical and chemical deposition and
epitaxy deposition.
The analysis is then extended to c-Si based tandem PV modules by combining the cost
model for TF solar cells with existing c-Si cost model. In both scenarios, the main issue to be
discussed pertains to the impact of absorber-related capex on module MSP and total system MSP
of the respective PV technologies. The results from these analyses can help guide selection of
materials and growth techniques, as the cost savings [$/W] attainable for tandem modules can be
expressed as a trade-off between a decrease in area-related costs due to efficiency improvements
and an increase in costs associated with top-cell manufacturing and integration. Lastly, module
efficiencies that will enable each of these manufacturing methods be economically viable and
exceed the target of US$0.50/W are estimated.
2.3
Cost and Minimum Sustainable Price Models
2.3.1
Bottom-Up Manufacturing Cost Analysis for PV Modules
Glass
Cleaning
FrntBuffer
Laser
Patterning
Laser
Patterningn
Bc
Layer
Deposition
Laser
Patterning
Absorber
Deposition
Modularization
Figure 2. 1: Typical manufacturing process flow of a TF solar cell.
Figure 2.1 shows a typical manufacturing process flow of a TF module. In this case, a
superstrate structure is assumed in which the front electrode is deposited first. Nevertheless, this
27
model is also applicable to a solar cell grown in substrate configuration since it will consist
similar number of process steps, only in a different sequence. This will also form a baseline in
which the capital and materials costs associated with absorber deposition step can be
investigated.
In general, the direct manufacturing costs associated with each process step can be divided
into five components: depreciation, direct materials, utilities, labor, and maintenance overhead.
In accrual accounting, the depreciation expense [$/Wp or $/m2 ] of each step is defined as capex
allocated over the period of time when related revenue is generated (matching principle). In this
work, a throughput-normalized capex [$/(Wp.yr) or $/(m2 .yr)] is used to calculate depreciation
expense so that different deposition techniques are compared on equal footing. Table 2.1
summarizes the various assumptions that are used for each component. In addition to the various
process steps illustrated Figure 2.1, the costs associated with building a factory and cleanrooms
are also included under 'facilities'. Repetitive steps such as laser patterning and electrode
deposition are accounted for by a multiplication factor. It is important to note that the purpose of
this model is not to predict manufacturing costs of TF PV modules with pin-point accuracy, but
to highlight dominant factors and general trends.
Bottom Cell Cost Model
c-Si
Top Cell Cost Model
thin-film
Tandem Cell Cost Model
c-Si +thin-film
Figure 2.2: Merging c-Si and TF cost models to yield a tandem module cost model.
To model the cost structure of c-Si based tandem modules, the TF module cost model is
combined with existing model for a / = 16% c-Si module4 as shown schematically in Figure 2.2.
The breakdown of costs for a c-Si module is summarized in Table 2.11.
28
Table 2.1: Summary of the assumptions used for each process step to model the cost structure of a hypothetical TF
PV manufacturing line. The numbers are not adjusted for yield and only represents a single repetition.
Step
Component
Cost [$/m 2 ]
Depreciation
1.9
Materials
0.000
Materials in operation phase is negligible.
Utilities
0.375
Total electricity consumption similar to c-Si cell plant.4
Labor
0.426
Estimated using methodology in Powell et al.4
Maintenance
5.3
28% of depreciation.
Depreciation
0.007
Estimated using data from glass cleaner supplier.
Materials
5.660
Includes a piece of glass.
Utilities
0.375
Total electricity consumption similar to c-Si cell plant.4
Labor
0.426
Estimated using methodology in Powell et al.4
Maintenance
0.002
28 % of depreciation.
Depreciation
2.000
Estimated using data from PVD equipment supplier.
Materials
3.000
Estimated using data from PVD equipment supplier.
x
10-1
x 10-6
Assumption
Similar to c-Si cell and module manufacturing.4
.5
Utilities
0.375
Total electricity consumption similar to c-Si cell plant.4
0
&
Labor
0.426
Estimated using methodology in Powell et al.4
Maintenance
0.560
28 % of depreciation.
Depreciation
x
Vary x from 0.1 to 1000.
2.500
Fixed at $2.5/M 2
Utilities
0.375
Total electricity consumption similar to c-Si cell plant.4
Labor
0.426
Estimated using methodology in Powell et al.4
Maintenance
0.28x
28 % of depreciation.
Depreciation
0.500
Estimated using data from laser supplier.
Materials
0.000
No materials needed.
Utilities
0.375
Total electricity consumption similar to c-Si cell plant. 4
Labor
0.426
Estimated using methodology in Powell et al.4
Maintenance
0.140
28 % of depreciation.
Depreciation
0.022
Estimated using data from laminator supplier.
Materials
17.82
Includes encapsulant, glass, Al frame and junction box. 4
Utilities
0.375
Total electricity consumption similar to c-Si cell plant.4
Labor
0.426
Estimated using methodology in Powell et al.4
Maintenance
0.006
28 % of depreciation.
8 Materials
e
.
o
29
Table 2.11: Cost breakdown in c-Si module manufacturing for (a) q = 16% baseline c-Si module and (b) q
20.5% advanced concept c-Si module. 4
Component
Cost [$/m 2]
Component
Cost [$/m 2]
Depreciation
Materials
Utilities
Labor
14.90
57.19
7.08
16.7
Maintenance
5.52
Depreciation
Materials
Utilities
Labor
Maintenance
Total
14.98
38.12
5.97
7.72
5.48
72.27
101.39
Total
(a)
=
(b)
Minimum Sustainable Price for PV Modules
2.3.2
Manufacturing cost does not necessarily translate directly to selling price. However,
applying certain financial considerations to a cost structure can allow the lower bound on a
sustainable price be estimated. This price is so-called the MSP and is calculated using a
discounted cash flow model for a hypothetical 400 MW factory. It is the price at which the
internal rate of return equals the WACC and more details about calculating MSP is described by
Powell et al.4 The role of the MSP is to take into account the cost of capital of initial outlay to
build a working manufacturing facility. Consequently, the difference between cost and MSP can
be mostly attributed to the weighted average cost of capital (WACC) and capex.
2.3.3
Calculation of Total System Installation MSP
To fully capture the effect of module efficiency, the total PV system installation cost is
calculated by taking into account BoS costs. The total BoS cost is sensitive to PV conversion
efficiency as many factors such as land requirement and mounting hardwares scale with total
installation area. In this work, BoS-related costs for 2015 and the SunShot target for 2020 are
obtained from the Solar Advisor Model (SAM) developed by the National Renewable Energy
Laboratory.2 3 Here, the term 'BoS MSP' is used because the profit-margin for PV installation is
included.
30
-
1nn
-BoS
in 2015
-SunShot
5
Goal in 2020
5.0
0
Ca
0.0
0.0%
10.0%
20.0%
30.0%
40.0%
Module Efficiency
Figure 2.3: Sensitivity of estimated BoS price to module efficiency from SAM. 23
Figure 2.3 shows the sensitivity of BoS price in 2015 and the projected SunShot target for
2020. At low efficiencies, the BoS price varies greatly with module efficiency due to costs that
scale with installation area. Hence, the cost-effectiveness of a PV installation depends on the
trade-off between BoS and module price.
2.4
Results and Discussion
2.4.1
Economic Feasibility of TF Single Junction Modules
Figure 2.4 (a) and (b) show the cost breakdown for a TF module with r7 = 14.5%, absorber
deposition capex = $10/m 2 and absorber materials-related cost = $2.5/m 2. Here, the capex of
$10/M2 assumes a high-rate physical vapor deposition process (based on internal market
research) and a material cost of $2.5/m2, which is determined as a reasonable upper bound for an
Earth-abundant semiconductor that has potential for TW scalability following Figure 1.2. The
model estimates a cost and MSP of $0.40 and $0.63 respectively, highlighting the prospect for
Earth-abundant TF modules to achieve < $0.50 total manufacturing cost. It can be observed that
the total cost comprises of a large fraction of materials related cost (- 56%) whereas capex
makes up about 26%. This cost structure is typical for both organic and inorganic TF modules
31
which can be made using high-throughput processes that significantly reduces capex or
processing costs.
0.70
0.70
0.63
O.63
0.50
0.50
a.
w Maintenance OH
WLabor
0.40
* Utilities
Materials
30
sP
Laser
Patterning
0.30
3CAPEX
NSP
0m0.20
(
U)*Estimated
m Top Cell
Deposition
w Contact
Deposition
Substrate Glass
Cleaning
a Facilities and OH
0.10
0.10
0.00
0.00
&
0...~
0.40
0.20
Framing and
wiring
wEncapsulation
0.60
0.60
+ Estimated MSP
(b)
(a)
Figure 2.4: Cost breakdown and MSP for a TF module with 'i = 14.5% in terms of (a) process step and (b)
2
2
component. The capex for deposition of absorber is $1 0/m and absorber-related material cost is $5/M . The MSP
is estimated assuming WACC = 14%.
Figure 2.5 (a) shows the sensitivity of MSP to absorber capex for a single-junction TF
module and compares the module MSP for various TF PV technologies. Similarly, the iso2
efficiency lines are determined by assuming an absorber materials-related cost = $2.5/m and
other parameters as summarized in Table 2.11. For the various TF technologies that are
represented in Figure 2.5, the materials-related cost and capex associated with the absorber layer
are estimated using the elemental market price chart in Figure 1.2 and internal market research
respectively. It can be observed that when capex is low, the MSP becomes limited by materials
cost and MSP reduction can be achieved by improving efficiency or reducing materials cost. The
record Cu20 device efficiency grown using a low capex method (electro-deposition) is 3.9%.15
The model suggests that to attain cost-competitiveness in module MSP with existing c-Si,
efficiency should be at least 7%. Based on the estimation carried out in this work, it can be
observed that TF modules based on CdTe and CGS are already cost-competitive with c-Si
32
modules. Increasing efficiency while keeping the same cost structure can allow both CdTe and
CIGS to approach a module MSP of below $0.60/Wp. The emerging methylammonium lead(II)
iodide perovskite (CH 3NH 3PbI3) is particularly attractive from a cost perspective due to a low
capex associated with solution-based processing methods. However, long-term stability issues
due to moisture, heat
4
and UV light" have to be resolved before CH 3NH 3PbI 3 can be
commercially viable.
10
(a)
-L
CU20
U 1
a-si/nc-Si
C ZT2
c-Si=$0.85/W,
0
S1----GS
dTe
CH 3NH 3Pbl3
0.1
10
-IL
(b) Cu20+
a-Si/nc-Si
CZT SS
-
E
CH 3NH 3Pbl3
1
c-Si=$2.12/W,
CIGS
CdTe
+ Record Cell Efficiency
4)
o Current Module Efficiency
5%- 10%-15%-20%-25%
S
0.1
0.1
10
1
100
1000
Absorber Layer Depreciation [$/m 2]
Figure 2.5: Sensitivity of (a) module MSP and (b) total system MSP of various TF technologies to absorber
deposition capex and module efficiency. The black dash line represents the module MSP or system MSP for c-Si.
The symbols represent several technologies and are estimated based on their efficiency, estimated materialsrelated cost and estimated deposition capex. The open symbols represent current record cell efficiency and the
closed symbols represent the typical module efficiency in the market.
33
As mentioned earlier, the module MSP by itself is an inadequate gauge because module
efficiency has a significant influence on costs incurred further downstream of the PV value
chain. For a more comprehensive analysis, the total system MSP is calculated and compared in
Figure 2.5 (b). The total system MSP is calculated by summing the module MSP, BoS costs, and
installer's profit margin as described in section 2.3.3. Because the total system MSP comprises a
large portion of BoS-related costs which are mostly area-related, it is more sensitive to module
efficiency. By taking this effect into account, the financial barrier for low efficiency technologies
becomes significantly higher as observed in Figure 2.5 (b). Correspondingly, r/ should
approximately be greater than 15% for TF PV to make economic sense on a utility scale if
absorber deposition capex is greater than $1 0/m2 . It is worth cautioning that apart from economic
feasibility, there are other barriers which can play an important role in commercial viability
manufacturing scalability. These include (but no limited to) module reliability (which influences
weighted
the
average
cost
of capital),
process
complexity,
yield
and
performance
reproducibility. 22
2.4.2
Economic Feasibility of c-Si Based Tandem Modules
While c-Si based tandems is a high efficiency approach which can potentially leverage on
existing c-Si manufacturing factories, the economic feasibility is not well studied and will be
address partly in this section. The main question that will be addressed pertains to the economic
viability of c-Si based tandem. Although the analyses performed in this section are for c-Si based
tandems, most learnings can be generalized to other tandem structures.
The most straightforward scenario for a tandem module to make economic sense is if the
combined MSP is cheaper than the MSPs of both the top and bottom module in a situation which
they are to operate individually:
34
MSPanemcell
<Min[M P,.ceSMSPottomcei]
(2.2)
This scenario will provide the financial incentives for both top and bottom cell manufacturers to
pursue the tandem approach. In this section, the economic feasibility of tandems will be
investigated from two different perspectives: module MSP and total system MSP. As shown in
Figure 2.6, two different tandem structures are studied in this work: (a) mechanically stacked
tandem in which a TF module is mechanically stacked on top of a c-Si module, and (b)
monolithically integrated tandem in which additionally absorber, buffer and contact layers are
grown monolithically on c-Si solar cells to yield a tandem solar cell. In the first scenario, the
entire cost structure of a top cell module is added onto c-Si's module cost whereas in the latter
approach, only the costs associated with absorber and electrodes depositions are included. The
monolithic approach is more streamlined because the total modularization cost is significantly
reduced.
++
(b)
(a)
Figure 2.6: Schematic drawings of a 2-terminal (a) mechanically stacked tandem module and (b) monolithically
integrated tandem module.
Figure 2.7 (a) shows the sensitivity of the MSP for a q = 30% tandem (mechanically stacked
and monolithically integrated) to absorber capex as compared to c-Si and single junction TF. The
mechanically stacked tandem has a higher MSP due to additional costs related to modularization.
The cost advantages of both types of tandem module are negligible especially at low top cell
absorber deposition capex because the MSP for TF top modules with q > 10% will be lower than
both q = 16% bottom c-Si and q = 30% tandem modules. As a result, the window is very narrow
35
for tandems to be successful just by comparing module MSP. Physically, it is also very
challenging to attain a q/
=
30% tandem module with a top cell q < 10%.
10
a-a
c-Si=$0.85IW~
-
0.1'
-
10
(b)
--
15%
10%
5%
20%
0.1
1
0.1
Top
2
[$/m
Depreciation
Absorber
Cell
1000
100
10
]
Figure 2.7: Sensitivity of (a) module MSP and (b) total system MSP of ' = 30% tandems to absorber deposition
module efficiency. The
capex and
colored dashed
Similarly,
that
It
The
the
total
the
be
can
total
cost
advantage
MSP
module
beyond
looking
MSPs
system
that
observed
system
line represents the module MSP
black dash
lines are for single junction TF modules for
MSP
for
and
of
the
of
both
window
a tandem
a high
efficiency
tandem
for
module
approach
the
considering
total
are
to
be
r/ =
30%
tandems
36
can
only
below
that
of
is
successfully
is either
be
instead.
MSP
system
approaches
with
or system MSP for c-Si. The
comparison.
on
par
fully
Figure
c-Si
now
and
understood
2.7
a r/ =
significantly
(mechanically
by
(b)
shows
20%
TF.
wider.
stacked)
or lower (monolithically integrated) than -both a q = 20% top TF module and a q = 16% c-Si
bottom module. The technological barrier to achieve a q = 30% tandem module with a q = 20%
top TF module and a q = 16% c-Si bottom module is also much lower, especially if the bandgap
of the top module is near optimal. These results are promising as efforts have already been made
to demonstrate significant boost in overall efficiencies by combining these perovskite cells with
c-Si or CIGS cells.26 27
10
E
c-si=$1.07/WP
Monolithic q=30% Tandem
10%
15% - - 20%
- 5%
F-0.1
0.1
1
10
100
1000
Top Cell Absorber Depreciation [$/m 2]
Figure 2.8: Sensitivity of total system MSP of I = 30% tandems to absorber deposition capex as a function of
module efficiency. The black dash line represents the system MSP for q7 = 20.5% c-Si. The colored dashed lines
are for single junction TF modules for comparison.
Projecting forward to the next 5 years, the efficiency of the bottom c-Si module is assumed
to reach q = 20.5% and achieve a MSP of $0.5 1/Wp.4 BoS price will also continue to decline as
described in Figure 2.3. Under these assumptions, the 'sweet spot' window is predicted to shrink
as shown in Figure 2.8, mainly due to reductions in the MSP of c-Si and area-related BoS costs.
Nonetheless, tandems can still remain feasible economically and have the potential to achieve
close to $1/Wp total installed MSP. In particular, a monolithic tandem will have greater success
over a mechanically integrated tandem.
37
2.5
Conclusions
In summary, module MSP and total system MSP are modeled for TF and c-Si based
tandems. It is shown that area-related BoS costs drive the economic viability of high-efficiency
PV technologies. Consequently, it becomes important to take into account total system MSP
when considering the feasibility of different technologies. For TF PV, it is found that q > 15% is
required to make economic sense on a utility scale. For both monolithically and mechanically
integrated c-Si based tandem structures with / = 30%, it is found that the total system MSPs are
below those of the bottom q = 16% c-Si module and top / = 20% TF module. This presents an
exciting opportunity for the commercialization of tandem PV technologies based on c-Si. Lastly,
projections made toward 2020 predict that there exist opportunities for tandems to be
economically feasible despite reductions in c-Si module MSP and various area-related BoS costs.
38
CHAPTER
3
PRINCIPLES OF DEFECT ENGINEERING
3.1
Introduction
Defects present in bulk, interfaces and surfaces can have detrimental effects on solar cell
performance. Consequently it is important to understand their origins and engineer solutions to
mitigate their negative impacts. A famous German chemist named Justus von Liebig came up
with the Law of the Minimum which essentially states that the water level in a barrel is limited
by the length of the shortest stave, emphasizing the need to identify and mitigate the most
limiting factors so improve performance. Glunz adapted it for silicon solar cells development as
shown in Figure 3.1 to illustrate efficiency bottlenecks.2 8 In general, the same law is applicable
to the development of TF solar cells in which identifying limiting loss mechanisms is crucial for
efficiency improvements. In this respect, some critical issues of interest typically pertain to
recombination processes in different regions of a solar cell, metal back contact formation and
optical light trapping. To this end, back contact, interface and bulk defect engineering are
discussed in relation to Cu20 solar cell performance in this chapter as background and to
motivate work in later chapters.
39
Ake&
Figure 3.1: Liebig's Law of the Minimum adapted to solar cells, from Ref.
3.2
Engineering Back Contact
3.2.1
Forming Ohmic Back Contact
28
For a solar absorber, the back metal contact is responsible for extracting the photo-generated
majority carriers (i.e., holes for p-type and electron for n-type). Consequently, forming lowresistivity ohmic contacts on a solar absorber is a critical step to reduce power loss. However, the
formation of an ohmic metal-semiconductor (M-S) contact can be complicated due to a variety of
reasons such the formation of a Schottky barrier, surface Fermi level pinning and chemical
reactions at the M-S interface which introduces secondary phases. These issues have been
discussed comprehensively by Schroder 29 and a brief summary is included in this section.
The Schottky model is the simplest model which describes the barrier height, 0B at M-S
interface is given as: 29
40
B =OM
#B
=
-X
for n-type
(3.1)
p-type
(3.2)
Eg + X -# mfor
where X is the electron affinity of the semiconductor, $m is the work function of the metal and
E. is the bandgap of the semiconductor. To the first order, a suitable metal can be chosen in
consideration of the semiconductor's X so that
#B is as
small as possible. Figure 3.2 shows the
SCAPS-simulated impact of metal work function on metal/Cu20/Ga2O3/ZnO:Al solar cell. The
parameters used for SCAPS simulations are similar to those report by Brandt. 3 0 As Cu20 has ;r
= 3.2 and E. = 2.0, a metal Om = 5.1 yields the highest efficiency due to good alignment with
the valence band of Cu20. A roll-over characteristic can be observed for Om= 4.8 eV due to the
formation of a barrier with $B= 0.3 eV.
0.0
Metal Work Function
4.8
E
E
-2.0
-
-
-
4.9
Metal
ZnO:Al
5.1
Ga 2 O 3 :Sn
5.0
-4.0c:
-6.0-8.Q
0.0
CU2
Cu O
-
2
'Metal
0.1 0.2 0.3 0.4 0.5
Voltage [V]
0.6 0.7
(b)
(a)
Figure 3.2: (a) SCAPS simulation for a metal/Cu 20/Ga 2O 3/ZnO:Al solar cell with different work function for the
metal back contact and (b) device structure used for SCAPS simulations.
In reality, the simple model does not predict trends observed for some semiconductors
including Ge and Si due to the formation of oxide, although it works well for silicides. 29 Work
by Bardeen has shown that the presence of defect states at the M-S interface can pin the Fermi
41
level and make
#B
independent of 0, .
Following Bardeen's model, Cowley and Sze has
shown that the barrier height (for n-type semiconductor) can be expressed by:
#B
=v(M - X) + (I -)(Eg
-- #0 )
(3.3)
where E. is the bandgap of the semiconductor, 0 is the position of the charge neutrality level
measured from the top of the valence band and y is a "pinning factor" which is related to the
thickness and total permittivity of the oxide layer at the interface and the density of surface
states.
While Schottky's model offers a good first-order guide for selecting an appropriate metal
back contact, it is often more complicated due the presence of surface states and oxides.
Consequently, direct measurements of contact resistivity and
#B
across the M-S interface
provide a more realistic approach to select and evaluate ohmic contacts suitable for device
applications.
3.2.2
Measuring Contact Resistivity and Schottky Barrier Height
The most common way to measure specific contact resistivity (pc ) of a M-S interface is
through the transmission line model. A linear transmission line model required mesa etching to
prevent unwanted conduction pathways while a circular transmission line model (CTLM)
eliminates the mesa etching step, hence simplifying test-structure fabrication.32
42
r
d
L
Rcontact
Rcontact
Rr
tL4v
I
(b)
(a)
Figure 3.3: (a) Circular contact pattern for CTLM and (b) cross-section of CTLM.
As shown in Figure 3.3, the test pattern consists of a central dot contact and concentric ring
contacts. The CTLM test pattern can be achieved through photoresist lift-off of metallic films.
The total resistance across the contact spacings can be expressed by:33
Rt = R In
27r
r+Lt(-+r
r-d
r
)]
r-d_
(3.4)
where RS is the sheet resistance of the Cu20 film, r is the radius of the outer circular contact pad,
d is the width of the ring, and Lt is the transfer length. Least squares curve fitting can be used to
extract the parameters R, and L, ; subsequently, pc can be obtained from the relation:33
Pc = L2 R,
3.2.3
(3.5)
Measuring Schottky Barrier Height
Temperature-dependent CTLM measurements can be performed to determine
#B.
The
standard thermionic emission (TE) model relates #B and the temperature-dependence of pc
through an Arrhenius relationship:3 3 3 4
k
qB
PC =- T exp( kT )
qA
(3.6)
where k is Boltzmann's constant, A* is the effective Richardson's constant and T is the absolute
measurement temperature.
43
In addition,
#B
is also commonly determined by analyzing the band bending at the M-S
interface through X-ray photoelectron spectroscopy (XPS). The following expressions relate the
core level bind energies as measured by XPS to
B
#B:
= (Eg - Eb) +(E, - Evbn)
for n-type,
(3.7)
#B
for p-type,
(3.8)
= Eb-(Ei - Evbm
where Eb is the binding energy of a core level in the semiconductor with a metal overlayer, Ei
is the binding energy of a core level for a bare semiconductor,
E'bm
is the position of the valence
band maximum and Eg is the bandgap of the semiconductor. All binding energies are referenced
to the Fermi level of the instrument.
3.2.4
Lowering Contact Resistivity Using a Tunneling Layer
Figure 3.2 illustrates the impact of a Schottky barrier at the back contact on solar cell
efficiency. While selecting metal with high
#m
is desirable for making ohmic contacts to p-type
materials, it is also advantageous from a manufacturing scalability point of view to be able to
engineer ways to make ohmic contacts with any metals as metals with high
#m
like Au (5.3 1-
5.47 eV),35 Pd (5.22 - 5.60 eV)3 5 and Pt (5.12 - 5.93 eV)3 5 are expensive and not manufacturing
friendly.
One potential solution is to introduce a thin highly-doped semiconductor layer to which
reduces the depletion width into the semiconductor and allows tunneling of carriers through field
emission to occur as illustrated in Figure 3.4.
44
..
(a)
.
.
(b)
(c)
Figure 3.4: Schematic illustrating the effect of doping on holes transport across the M-S interface, (a) thermionic
emission, (b) thermionic field emission, and (c) field emission.
Figure 3.5 shows the SCAPS-simulated impact of incorporating a highly-doped Cu20 layer
(1020 cm-3) at the back contact. It can be observed that the negative effect due to a low
#Mmetal
can be mitigated through a tunneling layer that is introduced at the back of the device.
.
E
2
-2.0 -
E
.
1
1
2-
-
--
1
I
-4.0-
E
-- EC
CC
:-2
-
..
C -6.0-
--3
0.
'_'_'_'_'
0.1 0.2
0.3 0.4
0.5
'
0.6
-_4
-0.5
0.7
EFp
Fn
--
E
0.0
0.5
-C
-8
1..
'........
' '
1
Metal Work Function
4.8 eV
4.8 eV with
Cu 2O doped layer
-
0.0
1.0
1.5
2.0
2.5
3.0
Thickness [ptm]
(b)
Voltage [V]
(a)
Figure 3.5: (a) J-V characteristics of solar cells with and without a highly doped layer and (b) energy band
diagram at Voc condition, illustrating the narrow depletion width at the back contact region, resulting in tunneling
transport. The conduction band (E.), electron quasi-Fermi level (EFn), hole quasi-Fermi level (EFp) and valence
band (Ev) are indicated on the energy band diagram.
3.3
Engineering Front Hetero-Junction
3.3.1
Importance of Recombination at Hetero-Junctions
Due to difficulties in achieving ambipolar doping in many semiconductors, most TF solar
cells are based on hetero-junctions. The presence of lattice mismatch will lead to strain, dangling
bonds and dislocations at the hetero-interface. These structural defects usually result in large
45
densities of deep-level defects which can increase recombination and lower device performance.
At a hetero-interface, the recombination current for holes in a p-type semiconductor, j,
can be
modeled using the following expression: 36 37
J=ShNexp(
-)
(3.9)
kBT
where Sh is the surface recombination velocity (SRV) for holes, N, is the valence band density
of states, PbO is the potential barrier for holes at the hetero-junction, kB is the Boltzmann
constant and T is the temperature.
Following Equation (3.9), this section will focus on discussing issues pertaining to interface
recombination and the effects of band alignment, presence of interface states and doping levels
in the conjugate buffer layer will be discussed. Additionally, the XPS technique used commonly
to characterize band alignments at hetero-interfaces will be discussed.
3.3.2
Effects of Band Alignment and Interface States at Hetero-Junction
Presuming that the interface recombination is the dominant recombination mechanism, the
open-circuit voltage (Voc) of a solar cell is limited to the maximum of PbO following equation
(3.9).37 Here, PbO is determined by the conduction band offset at the hetero-interface and the
donor-acceptor ratio (ND /
/NA
) between the p-type absorber and n-type buffer layer. Therefore,
Voc can be enhanced by increasing donor density in the n-type buffer layer, a smaller conduction
band and reducing interface states (decreasing SRV). 37
To illustrate the effects of conduction band alignment and interface states, SCAPS-simulated
results for Cu20-based solar cells are shown in Figure 3.6. It can be observed that increasing the
density of mid-gap defect states at the hetero-interface is detrimental to device performance.
46
Increasing conduction band offset (AEc) also decreases device performance. On the other hand,
a high Voc of 1.4 V can be achieved if the conduction bands of Cu20 and the buffer layer are
well aligned (i.e., AEc> -0.3 eV). A Voc of 1.2 V has already been achieved experimentally by
using Ga2O3 as a buffer layer, which has a conduction band well-aligned to Cu20.'
0.0
0.0
C14 E
Hetero-interface States
1210 cm3
(baseline)
10" cm-3
10 cm
cm2:1
-1015
<
-2.0
E
E
-
-
E
7t
-6~
-6.0
-
0
E
0.1
012
0.3
0.3
0.4 '
0.4
0.6
0.30
-
-0.80
-
o-1.30
0.5
--
-
-6.0
U
-8.2
*AEC
-4.0
n -4.0 C
-2.0
-
0
0.7
.0
0.5' ' -.5
0.5
-1.05 (baseline)
1.0
1.5
Voltage [V]
(b)
Voltage [V]
(a)
Figure 3.6: SCAPS-simulated results for (a) different interface states densities, and (b) different conduction band
offsets.
In summary, some of the most important considerations when managing hetero-junctions are
interface defect densities and alignment of conduction (p-type) or valence bands (n-type).
Chapter 5 will discuss issues band alignment at Cu 20/a-ZTO interface and discuss how structural
disorder present in the amorphous a-ZTO films can be correlated with device performance.
Chapter 6 will investigate the use of Ga2O3 as an n-type buffer layer to achieve high Voc and
discuss the potential of attaining higher efficiencies by doping Ga2O3 extrinsically.
Measuring Band Alignments
3.3.3
Like Schottky barriers, band alignments are commonly measured using the XPS
method. 1"
53
1 Using
Cu20/Ga2O3 hetero-interface as an example, the valence band offset (AE,)
at the hetero-junction can be calculated using the following relationship:
47
AE
where
A
/Cu2O
-
EGa
(E/Cu2o
EGa
/Cu2O
a
-(E
- ECuOa
(3.10)
- E")
is the difference in binding energy between the Cu-2p and Ga-2p
core levels measured at the Ga 2 0 3 /Cu 2 0
interface. E
-
E
3 and E
-
Ecv"O
are the
positions of the core levels referenced with respect to the respective valence band minimum
(vbm). Subsequently, AEc can be determined by:
AEC = AEv- (E
2
-C"2O)
(3.11)
where E "2o and Ela",o are the bandgaps of Cu 2 0 and Ga2O3 respectively.
3.4
Engineering Bulk Defects
3.4.1
Importance of Bulk Defects in Thin-Film Solar Cells
In the process of optimizing TF solar cells, it is crucial to understand the intricate
relationship between materials' bulk electrical properties, structural properties and process
condition. Bulk electrical properties like recombination lifetime, carrier mobility and free carrier
density are strongly influenced by defect types and densities and can have both positive or
negative impacts on solar cell performance. The common types of defects that are present in a
polycrystalline compound semiconductor can be generally classified into three categories,
namely intrinsic point defects, structural defects and extrinsic impurities. Each type of defect and
their complexes can have different influence on electrical properties.
Due to the extensive nature of this subject matter, this section will only be devoted to
discussing bulk-defects related issues in the context of improving Cu20-based solar cells: 1)
Extrinsic doping of n-type semiconductors is discussed in relation to its role in enhancing solar
cell efficiency. 2) Manipulating intrinsic point defect structure in absorber materials to control
48
bulk electrical properties is discussed in the context of reducing bulk recombination activity. 3)
Measurement techniques used characterize bulk materials properties will also be introduced.
3.4.2
Extrinsic Doping in n-type buffer layers
0.0
E
E
Buffer Layer ND
16
3
10 cm- 3 (baseline)
3
1011 cm
-2.0 - 10 cm
1
AE =-0.3
-6.08..
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Voltage [V]
Figure 3.7: SCAPS-simulated J-V characteristics for Cu 2 0-based solar cells with different buffer doping
densities.
As elucidated by both SCAPS simulations and experiments in section 3.3.2," a wellmatched conduction band has the potential of achieving high Voc but the fill-factor (FF) remains
low in a Au/Cu20/Ga2O3/ZnO:Al device stack due to high series resistance at high forward bias
resulting in the roll-over effect. This is possibly be due to the formation of a blocking diode at
the front interface between the intrinsic Ga2O3 buffer and the ZnO:Al front contact. As shown in
Figure 3.7, SCAPS simulations reveal that the FF deficit can be improved if ND of the buffer
layer is increased. Consequently, these results motivate the work performed in Chapter 6 to
develop a framework to better understand extrinsic doping mechanisms in Ga2O3.
3.4.3
X-ray Absorption Spectroscopy as a Local Probe of Atomic Environment
X-ray absorption spectroscopy (XAS) is a powerful technique for determining the local or
electronic structure of matter and is one of the tools used in this thesis to characterize local
49
environments of semiconductors in relation to their bulk properties and process conditions. The
X-ray beams used in XAS are generated in synchrotron radiation sources. The two synchrotron
beamlines used for the work performed in the scope of this thesis are the 10-ID at Advanced
8 39
Photon Source (APS) and BL 4-3 at Stanford Synchrotron Radiation Light Source (SSRL).3 ,
The energy ranges of 10-ID and BL 4-3 are 4.3 to 27 keV and 2.4 to 14.5 keV respectively.
These energy ranges allow absorption spectroscopy to be performed at the K-edges of transition
metals Zn, Ga and Sn. A typical XAS spectrum consists of two regions, X-ray absorption nearedge structure (XANES) and extended X-ray absorption fine structure (EXAFS). XANES refers
to the low energy region just above the absorption edge where photoelectrons have low kinetic
energies and multiple scattering events are dominant. Hence, it is sensitive the local geometric
coordination. The absorption edge can be used to deduce the binding energies of core shell
electrons which informs the overall charge state of the absorption atom.
The EXAFS refers to the higher-energy region of the XAS spectrum, and can be analyzed to
yield information about the distances between central absorbing and neighboring atoms, the
number of neighboring atoms, the nature of neighboring atoms, and the degree of disorder of
local bonds. The EXAFS equation is expressed as:4 0
'k - 21 N~jf, (k) 2k2U2 -2R /(k)
kSoj
kRk e _
e_
sin(2kR, +J(k)),
where j indicates shells of like atoms, S2 is the passive electron reduction factor, N
(3.12)
is the
coordination number of atoms in the jth shell, k is the photoelectron wavenumber, R, is the
distance to the neighboring atoms, o7
is the Debye-Waller factor or the mean-squared disorder
of neighbor distance and A(k) is the electron mean free path. The scattering amplitude,
, (k),
and the phase shift, 8, (k), are dependent on the atomic number of the scattering atoms. The
50
crystal structure of the material of interest is used as a starting input into the ATOMS and FEFF6
codes implemented in Artemis. 4 1 Non-linear least squares fitting is subsequently performed in
Artemis to obtain the best-fit parameters used in equation (3.12).
3.4.4
Bulk Recombination and Point Defects Structure in Cu2O
In Shockley and Queisser's detailed balance limit analysis, 4 2 band-to-band radiative
recombination in the bulk absorber material places an upper bound on the theoretical efficiency
of a p-n junction solar cells. However, the presence of bulk defects in a polycrystalline TF solar
cell can introduce other non-radiative recombination pathways that increase the reverse
saturation current jo of a diode and therefore, reduce the Voc of a solar cell. 43 Therefore, it is
important to understand the origin of these bulk recombination mechanisms so as to mitigate its
impact on device performance.
For bulk defects, states which are deeper into the bandgap are more active in ShockleyRead-Hall (SRH) recombination, 44 whereas shallow states can contribute to free carriers. The
properties of these states are related to the point defects structure. Hence, the ability to
manipulate the point defects structure allows optimization and tunability of critical properties for
solar cell performance. For example, Raebiger et al.45 performed first principles calculations to
predict the intrinsic hole concentrations (p) by calculating the concentrations of dominant point
defects in Cu20 as a function of growth temperature and oxygen partial pressure. As shown in
Figure 3.8, the preferential formation of acceptor Cu vacancies (Vcu) over donor 0 (Vo)
vacancies over a wide range of growth temperature (800 to 1500 K) gives Cu20 its intrinsic ptype behavior. Moreover, the high ionization energy of oxygen vacancies also reduces the
compensation ratio of ionized donor to acceptor in the material. Consequently, p increase with
growth temperature under both oxygen rich and poor conditions.
51
I
IL
Cu-rich/C-poor
EC T (a)
0.4
>
{--F----- - - -
0.2
102
E0,
----
FT
--- EF(T=Tg)
~--- - ----- - -
EL.
E
(b)
Ig)
-
0
I
Cu-poor/C-rich
10 (c)
EF(T=298K)
LF(-298K
(d)
CU
I
V
0ig)
0
0 10.
800
1000
1200
1400 800
Growth temperature [K]
1000
1200
1400
Growth temperature [K]
Figure 3.8: Equilibrium Fermi level and defect concentrations calculated for Cu-rich/O-poor conditions and Cupoor/0-rich conditions. The hole concentration and Fermi level are given both in equilibrium at growth
temperature Tg and at room temperature 298 K. This figure is from Raebiger et al."
In order to reduce the resistivity of Cu20, many authors rapidly quenched Cu20 films while
it is at high temperature. As the Cu vacancies do not have sufficient time to equilibrate, a high
concentration of Vcu is 'frozen-in', giving a high p at room temperature.4 6 4, 7 However, due to the
relatively high ionization energy, not all Vcu are ionized, effectively setting a lower bound to the
bulk resistivity that can be obtained using this 'freeze-in' method. Although the manipulation of
these defects allows tunability of free carriers, the impact on recombination lifetime is relatively
unknown. In Chapter 7, bulk electrical properties such as recombination lifetime, p and
mobilities of Cu20 are tuned by manipulating growth conditions. Following which, bulk
recombination activities are evaluated using lifetime and photoluminescence spectroscopy
techniques.
3.4.5
Lifetime and Photoluminescence Spectroscopy
Many techniques capable of characterizing bulk defects are commonly based on (but not
limited to) lifetime, 4 8 photoluminescence49 50 or capacitance.' Each technique has it's own pros
52
and cons but in general, they can be applied in a complementary manner to yield complete
information about bulk defects present in the system. The three defect characterization
techniques used in this thesis are microwave-photoconductance-based
transient lifetime
measurements, spectrally resolved photoluminescence, and photoluminescence imaging.
53
54
CHAPTER
4
Low CONTACT RESISTIVITY
OF
METALS ON NITROGEN-DOPED
CUPROUS OXIDE (CU20) THIN-FILMSa
4.1
Abstract
Forming low-resistivity contacts on cuprous oxide (Cu20) is an essential step toward
demonstrating its suitability as a candidate solar cell material. Contact resistivity of three noble
metals (Au, Ag and Pd) is measured on sputtered Cu20 thin-films (TFs) with a range of nitrogen
doping levels. Using the circular transmission line method, specific contact resistivity as low as
1.1xl0-4
Q-cm 2
is measured for Pd contacts on heavily nitrogen-doped
Cu20 films.
Temperature-dependent current-voltage measurements and X-ray photoemission spectroscopy
are used to determine the barrier heights formed at metal/Cu20 interfaces. Thermionic emission
is observed to dominate for undoped films whilst field emission dominates for heavily doped
films, highlighting the importance of carrier concentration on contact resistivity. It is
demonstrated that low contact resistivity can be achieved on heavily doped Cu20 films using
Published as S.C. Siah, Y.S. Lee, Y. Segal, and T. Buonassisi, Journal of Applied Physics 112, 084508 (2012).
[http://dx.doi.org/ 10.1063/l.4758305] and Y.S. Lee, J. Heo, M.T. Winkler, S.C. Siah, S.B. Kim, R.G. Gordon, and
T. Buonassisi, Journalof Materials Chemistry A 1, 15416 (2013).[ http://dx.doi.org/10. 1039/C3TA I3208K]
a
55
Earth-abundant metals such as Cu and Ni. Lastly, a heavily doped Cu20 film is applied as a
tunneling layer to improve performance of a Ag/Cu20/a-ZTO/ZnO:Al solar cell by mitigating
the series resistance due to a Schottky barrier at the back contact.
4.2
Introduction
Minimizing resistive power losses in Cu20-based devices requires the formation of ohmic
metallic contacts. For comparison, a contact resistivity (pc ) of < 2x 10- Q-cm 2 is desirable for
silicon solar cells under one sun conditions. 2 9 In previous studies, Au has been the material of
choice to form contacts on Cu20 films,5 2, 3 due to its favorably large work function and low
chemical reactivity. However, no contact resistivity measurements of metals on Cu20 films have
been reported.
In this work, we measure the pc of metal/Cu20 interface using various metals (both noble
and Earth-abundant) including Au, Ag, Pd, Cu and Ni. We study the effect of Cu20 doping level
on the contact resistivity and the carrier transport mechanism. The three noble metal candidates
are selected based on their high work function (#A
= 4.74 eV,
#Au=
5.31 eV,
Pd=
5.60 eV)35
and low reactivity with Cu20 as determined by the standard free energies of metal-oxide
formation at room temperature (A G'2 U20=
-147.2 kJ, 54 AG2Pd
= -82.1 kJ, 54 A G A2
= -10.8
kJ 54 and A G A"20 > 0). We show that it is possible to obtain pc as low as 1.1xl0-4 Q-Cm 2 on
nitrogen-doped (N-doped) Cu20 films, indicating the importance of doping to achieve low Pc.
Lastly, we demonstrate that pc < 10-3 Q-cm 2 can be achieved for Earth-abundant metals such as
Cu and Ni by forming a tunneling junction with N-doped Cu20.
56
4.3
Experimental Methods
The Cu20 TFs in this study are grown using reactive direct-current magnetron sputtering on
GE-124 fused quartz glass substrates. A metallic copper target (99.999% pure, K.J. Lesker) is
sputtered in an argon, oxygen and nitrogen ambient. The film's doping level is controlled by
varying the flow of N 2 gas between 0 and 4 sccm during the sputtering process. Film electrical
resistivities and hole concentrations are characterized by four-point probe and Hall effect
measurements respectively. X-ray diffraction is used to confirm that all films contained pure
Cu20 phase. Film thicknesses are determined by cross-sectional scanning electron microscopy.
Film nitrogen concentrations are measured using secondary ion mass spectrometry (SIMS) with
a calibration sample prepared by controlled amount of nitrogen implantation.
The parameter pc on nominally undoped and N-doped films is determined using the
circular transmission line model (CTLM) as described in section 3.2.2.32 The CTLM patterns
with inner-outer ring spacings of 4 to 12 ptm are achieved by photoresist lift-off of electronbeam-evaporated 100 nm films of Au, Ag, Pd, Cu and Ni. XPS analyses are conducted on
undoped Cu20 films with one, two, and three nm of metallic overlayers deposited using electronbeam evaporation at a rate of 0.5 A/s. The XPS measurements
are performed using a
monochromated Al-Ka source (photon energy 1486.6 eV), and the films are grounded using
silver-painted contacts. The adventitious C Is peak is monitored to account for any residual
charging. Lastly, to elucidate the electron transport mechanism across the metal/Cu20 interface,
temperature-dependent
pc
measurements
are
performed
on
a
heated
chuck
and
J-V
characteristics are recorded over the temperature range 280-390 K monitored externally through
a thermocouple.
57
4.4
Results and Discussions
4.4.1
Undoped Cu20/Metal Contacts: Schottky Barrier Height and Chemical Reactivity
To measure the Schottky barrier height (SBH) and investigate the chemical reactivity at the
contact between metal and undoped Cu 20, XPS analyses on undoped Cu20 films covered with
one, two, and three nm of metallic overlayers are performed. Figure 4.1 shows the Cu 2p core
level signals for the three different metals as a function of metal film thickness. One observes
two distinguishing features that set Pd apart from the other metals: (1) The measured peak
intensities for the samples covered in Pd attenuate more rapidly with increasing thickness
compared to Au and Ag; (2) The Cu 2p core level peak energy shifts to lower values with
increasing Pd thickness. These observations will be discussed in greater detail after a discussion
of Schottky barrier heights.
(a) Au
-Bare
-1
(c) Pd
(b) Ag
-Bare
-1
nm
--- 2 nm
nm
-3
nm
2 nm
-3nm
-Bare
-1
-2
-3
nm
nm
nm
C
936
932
936
932
936
932
BE [eV]
Figure 4.1: Cu 2p core level photoemission for (a) Au, (b) Ag and (c) Pd samples. A peak shift towards lower
binding energies is observed for the Pd samples, indicating the lowering of the SBH. A high binding energy
shoulder due to CuO can be observed.
By analyzing the band bending at the metal/Cu20 interface as illustrated in Figure 4.2, the
SBH can be determined from the XPS data using the method described in section 3.2.3. SBH
values obtained by analyzing the two nm thick overlayer signal are 0.3 0.1 eV, 0.4 0.1 eV and
58
0.2
0.1 eV for Au, Ag and Pd respectively. The two nm thick samples are compared as the Cu
2p core level signal of the Pd three nm sample is too low to be used for meaningful analysis.
(a)
E 2P-Eb
>E
Ci
(b)
-:
6
935 934 933 932 931
4
2
0
BE [eV]
Figure 4.2: Illustration of method to determine the SBH from the XPS data from the valence-band spectra (right)
and Cu 2p core level spectra (left) of (a) bare Cu20 film and (b) Cu2O film with two nm Au overlayer. The Fermi
level is calibrated using the Fermi edge of Au. All spectra have been corrected for charging by using the
adventitious C Is peak.
In addition, to obtain the SBH values at metal/Cu20 interfaces, the pc vs. temperature
curves (Figure 4.5) for the undoped samples is also fitted using the standard TE model and the
fitted values for SBH are 0.23 0.01 eV, 0.17 0.01 eV and 0.14 0.01 eV for Au, Ag and Pd
respectively.
Table 4.1: Comparison of Schottky barrier heights obtained from XPS and CTLM.
Metal
Au
Ag
Pd
XPS
0.3
0.4
0.2
CTLM (100nm)[eV]
0.23 0.01
0.17 0.01
0.14 0.01
(2nm)[eV]
0.1
0.1
0.1
Table 4.1 summarizes the values of SBH obtained using XPS and CTLM. Both XPS and
CTLM methods measure the lowest SBH for the Pd sample. We believe this is due to more
significant shifting of the Fermi level toward the valence band maximum of Cu2O at the
Pd/Cu2O interface. This effect can be explained from Figure 4.1, which shows the Cu 2p core
59
level peak shifting toward lower binding energies as the thickness of the metal overlayer
increases. 5 5 Insignificant peak shifts are observed for both Au and Ag. This results in the lowest
effective SBH for the Pd/Cu20 structure.
An additional effect may contribute to lowering the Pd-Cu20 contact resistivity relative to
Ag and Au. In XPS measurements, the core level intensity is observed to attenuate more strongly
with increasing Pd overlayer thickness than either Au or Ag. From other studies
56,
XPS peak
attenuation is known to depend on overlayer thickness, metal coverage uniformity, and
photoelectron attenuation length of the overlayer material. By comparing electron inelastic mean
free paths for Au (16.5
A,
1400 eV), Ag (15.2
A,
1 100 eV) and Pd (19.6
A,
1 100 eV) as
one would expect the Cu 2p core level signal of the Ag and Au
calculated by Tanuma et al.,
samples to attenuate most rapidly because of electron scattering. However, the Cu 2p core level
peak intensity for the Pd sample is the lowest despite having the largest electron inelastic mean
free path. This may indicate that Pd wets the surface mor effectively than Au and Ag.
(a)
Au 4,2
(b)
(C)
Ag 4d5 2
Au 4 f5/
Ag 4d312
(D) 94
Pd 3d512
Pd 3
90 88 86 84 82 376 372 368
32
344 340 336 332
BE [eV]
Figure 4.3: Photoemission from metallic peaks of samples with 2 nm thick (a) Au, (b) Ag and (c) Pd overlayers.
A high binding energy shoulder due to PdO can be observed.
The metal core level peaks for two nm overlayer thickness are shown in Figure 4.3. Sharp,
well-defined peaks can be observed for both Au and Ag, indicating that the Au and Ag are in a
60
metallic chemical state. This is in agreement with the thermodynamic data mentioned earlier, as
Au and Ag oxide formation energies are relatively unfavorable. A high binding energy shoulder
can be observed for the Pd sample, which we attribute to PdO. 5 8 The hypothesized superior
wetting characteristics of Pd on Cu20 could be related to the formation of PdO. This observation
is consistent with the findings of Pan et al.59 In this case, the formation of interfacial Pd-O bonds
might contribute to the reduction of the overall free energy and promote a Frank-van der Merwe
(layer-by-layer) type growth for Pd.5 9 The good coverage by the metal film mediated by
interfacial PdO formation may lead to higher-quality metal/Cu20 junction and lower pc for
undoped Cu20 films. If this interpretation is correct, then chemical inertness may not be the most
important parameter defining an ideal contact metal for undoped Cu20.
4.4.2
Nitrogen Doping Reduces Metal/Cu2O Contact Resistivity and Changes Conduction
Mechanism
Introducing nitrogen gas during Cu 20 film growth is observed to reduce film electrical
resistivity at room temperature."," For example, an undoped Cu20 film exhibits a resistivity of
56 Q-cm, while N-doped films have resistivities of 4.4 and 0.4 Q cm for nitrogen concentrations
([N])
0.6 and 1.2 atomic (at.) %,
respectively.
From Hall effect measurements,
hole
concentrations of undoped and N-doped ([N] = 0.6 at. %) films are measured to be 3.7x 1015 cm-3
and 1.8x1018 cm- 3 respectively. Due to a low Hall mobility, we are unable to accurately measure
the hole concentration of the heavily doped film ([N] = 1.2 at. %).
61
1n-
-u--Au
-
E
E
--Ag
-.- Pd
10
10-3
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Nitrogen Concentration [at.%]
Figure 4.4: Contact resistivity is plotted against Cu 2O nitrogen doping concentration for three contact metals: Au,
Ag, and Pd.
Figure 4.4 shows a plot of p, for the three inert contact metals (Au, Ag and Pd) and
different nitrogen doping densities. It is observed that Pd exhibits the lowest p, on undoped
Cu20 films, while pc as low as I.1x10-4 Qcm 2 is obtained on highly doped Cu20. While pc
can vary over an order of magnitude on undoped films for the three different metals, less
variation is observed for doped films.
To gain insight into the conduction mechanism, in Figure 4.5 the temperature dependence of
pc of Pd for three different Cu20 nitrogen doping concentrations is compared. For the highly Ndoped sample, pc shows a weak dependence on the measurement temperature, whereas pc of
the undoped sample exhibits an Arrhenius-type dependence on temperature. Similar behavior is
observed for Au and Ag samples. This indicates that at low doping densities, the dominating
conduction mechanism is thermionic emission (TE). In this regime, pc decreases as temperature
increases and a greater number of electrons possess sufficient thermal energy to overcome the
Schottky barrier at the metal/Cu20 interface. 29 As doping densities increase, the Schottky barrier
62
becomes sufficiently narrow and carrier transport becomes increasingly dominated by field
emission (FE), which is only weakly temperature dependent.
S[N]=O.O at.%
-[N]=0.6 at.%
-*[N1=1.2 at.%
-- TE Model Fit
-
E
10-
0
0
280
300
340
320
360
380
400
Temperature [K]
Figure 4.5: Contact resistivity as a function of measurement temperature for Pd on Cu 2O with three doping
concentrations. Similar behavior (not shown) is also observed for Au and Ag samples. The solid black line
represents a fit to a thermionic emission model; the dashed red and blue lines are guides to the eye.
To support these observations, we plot the ratio kT/Eoo as a function of doping densities, Nd
62
(Figure 4.6). The parameter Eoo is related to the tunneling probability and is defined as:
0
qh
EcO =$
2
Nd(41
-N;W_
m g
(4.1)
where h is the reduced Planck's constant, q is electronic charge, m* is taken to be
0.58mo, 47
where mo is the free electron mass and E is the dielectric constant of Cu20 taken to be 7.Eo.47 In
this case, the ratio kT/Eoo is a measure of the relative contribution of the TE to tunneling
3
process.62 It can be observed from Figure 4.6 that for lightly doped Cu20 (Nd < 1018 cm ), kT/Eoo
>> 1 and thus, TE is dominant. At intermediate doping levels (Nd
1018
cm-3 ), kT/Eoo
I and the
carrier transport is in the thermionic field emission regime where both thermionic and tunneling
3
processes are comparable. Lastly, at very high doping levels (Nd> 1019 cm- ), kT/Eoo << I and
carrier transport is dominated by FE where current is carried by holes tunneling across the
63
junction. As such, the calculated functional dependence of kT/Eoo with Nd is in good agreement
with the observations made based on the temperature dependence of pc.
40
35
30
25
N =3.7x10
cm'
LiP 20
15
10
=1.8x18 CM3
5N
5
101
-d
.
10 1
101
10
Carrier Density
101
3
[cm~ ]
102
)
Figure 4.6: Plot of the ratio kT/Eoo as a function of hole density (p) with m*=3.36m0 and c = 7.08o. The dotted
line indicates that the ratio kT/Eoo = 1 and both thermionic emission and field effect processes are comparable.
3
3
kT/Eoo for both undoped ([N] = 0.0 at.%, p = 3.7x 1015 cm- ) and lightly ([N] = 0.6 at.%, p = l.8x 1018 cmdoped samples are indicated on the plot.
4.4.3
Nitrogen Doping Enables Low-Resistivity Contact for Earth-Abundant Metals
It is demonstrated the formation of low resistivity ohmic contacts to N-doped Cu20 using
Cu and Ni, both Earth-abundant metals is possible. pc of 9.3x104 and
6 .3
x 10
Q-cm
2
for Cu
and Ni contacts, respectively, on Cu20 doped with [N] = 1.2 at. % is achieved. Temperaturedependent TLM measurements (Figure 4.7) indicate that p. for both Cu and Ni on N-doped
Cu20 films have weak temperature dependence. This suggests that FE is the dominating carrier
transport mechanism across the metal/Cu20 interface, consistent with observations involving
noble metals.
64
10,2
-+.-Cu
-.Ni
- A-Pd
*
44
T
0
o)
4
A I
280
320
300
340
360
380
400
Temperature [K]
Figure 4.7: Contact resistivity as a function of measurement temperature for Cu, Ni and Pd on highly doped ([N]
= 1.2 at. %) samples. The dashed lines are guides to the eye.
4.5
Application of Cu20:N Layer in Solar Cell
To demonstrate the potential of using Cu20:N as a tunneling or hole-transporting layer, a
20nm-thick Cu20:N with [N] = 1.2 at. %is incorporated between the Cu20 and Ag back contact.
Figure 4.8 (a) shows the cross-section SEM image of the Ag/Cu 20:N/Cu20/a-ZTO/ZnO:Al
device. A baseline Ag/Cu20/a-ZTO/ZnO:Al device without the Cu20:N is fabricated for
comparison. More details about the deposition conditions of each layer can be found in Lee et
'
al.6
Table 4.11: Comparison of photovoltaic characteristics for both control and control+Cu 20:N devices. Table are
adapted from Lee et al.6
Device
Cu20:N layer
Control
Voc [mV]
557
485
Jsc [mA/cm-2]
7.3
6.9
FF [%]
61.0
46.7
PCE [%]
2.56
1.56
From Figure 4.8 (b), it can be observed that the current density of the control device is
suppressed at bias > 0.3 V. According to SCAPS simulations shown in Section 3.2.1, the
suppressed current density can be attributed to the formation of a Schottky barrier at the back
65
contact. By inserting a highly-doped Cu20:N layer, the depletion width of the Schottky barrier is
reduced. This allows a high tunneling current through the back contact as predicted by SCAPS
simulations in section 3.2.4. Consequently, the power conversion efficiency is improved from
1.56 % to 2.56 % and the photovoltaic characteristics are compared in Table 4.11.
2
- Cu2O:N
ZnO:A
A
-control
2
Cu 2O
Cu O-N
6
Ag
-8
0.0
0.2
0.4
0.6
bias (V)
(b)
(a)
Figure 4.8: (a) A cross-sectional SEM image of a Cu 20-based TF solar cell with a Cu2 0:N hole-transporting
layer. Dashed lines (white) indicate interfaces between layers, and (b) J-V characteristics of a Cu2 O:N and a
control Cu2 O device under I sun illuminated (AM 1.5 G, 100 mW cm- 2 ) condition. Figures are from Lee et al.6
4.6
Conclusions
The doping-level dependence of contact resistance for three noble metals on Cu20 films: Pd,
Ag, and Au is examined. For all metals, an inverse dependence of contact resistivity on Cu20
nitrogen doping concentration is observed. At low carrier densities, a TE model describes the
interface charge transport mechanism. At high Cu20 carrier densities, a FE model accurately
describes conduction across the metal-Cu20 interface, indicating the formation of a tunneling
junction with kT/Eoo < 1. In such samples, p, as low as 1.1 x 10' Q cm 2 is achieved. Using this
approach, we showed that it is possible to obtain pc < 10- Qcm 2 on N-doped Cu20 films using
Earth-abundant metals such as Cu and Ni.
66
An interesting observation is that Pd is the best candidate for forming low-resistivity ohmic
contacts on undoped Cu20 due to the low effective Schottky barrier height at the meta/Cu20
interface. XPS measurements suggest that the good electrical contact of Pd on undoped Cu20
may be due to the formation of an interfacial PdO layer that aids in the wetting of the Cu20
surface, and bending of the Cu20 valence band maximum toward the metal Fermi energy at the
Pd/Cu20 interface as observed from the shift in the Cu 2p core levels. Further work may
elucidate the degree to which the best candidate contact metals on lowly doped substrates may be
predicted not solely on the basis of inertness with the underlying film, but on the basis of a
complex interrelationship between chemical inertness, wettability, stress, chemical potential, and
semiconductor free carrier density, among other parameters.
Lastly, it is demonstrated that a 20-nm thick Cu20:N can function as a tunneling layer which
mitigates the Schottky barrier at an Ag back contact, leading to an efficiency comparable with
devices formed using metals with larger work functions (e.g., Au).
67
68
CHAPTER
5
IMPACT OF STRUCTURAL DISORDER ON
ELECTRON MOBILITY IN AMORPHOUS
ZINC-TIN-OXIDE BUFFER LAYERSa
5.1
Abstract
The correlation between the atomic structures of amorphous zinc-tin-oxide TFs grown by
atomic layer deposition and their electronic transport properties is first investigated by
performing synchrotron-based X-ray absorption spectroscopy at the K-edges of Zn and Sn on aZTO samples with varying [Zn]/[Sn] compositions. In extended X-ray absorption fine structure
(EXAFS) measurements, signal attenuation from higher-order shells confirms the amorphous
structure of a-ZTO TFs. Both quantitative EXAFS modeling and X-ray absorption near edge
spectroscopy reveal that structural disorder around Zn atoms increases with increasing [Sn].
Field- and Hall-effect mobilities are observed to decrease with increasing structural disorder
around Zn atoms, suggesting that the degradation in electron mobility may be correlated with
structural changes. Moreover, increasing structural disorder around Zn atoms of a-ZTO films is
found to correlate with decreasing solar cell performance. Capacitance-frequency measurements
a
Published as S. C. Siah, S. W. Lee, Y. S. Lee, J. Heo, T. Shibata, C. U. Segre, R. G. Gordon, and T. Buonassisi,
Applied Physics Letters 104, p. 242113 (2014). [http://dx.doi.org/i0.1063/1.4884115]
69
indicate that the decrease in solar cell efficiency is correlated with an increase in defect states at
the hetero-interface. This suggests that structural disorder might also play an important role in
increasing the density of deep-level states at the Cu 20/a-ZTO interface.
5.2
Introduction
Amorphous metal-oxide semiconductors (AMOS) have attracted attention for transparent
optoelectronic applications including liquid-crystal displays and transparent conducting oxides
for organic light-emitting diodes. Zinc oxide and related oxide classes are investigated due to
their high electron mobilities compared to conventional amorphous silicon. Nomura et al.6 3
demonstrated high-performance flexible transistors using amorphous indium-gallium-zine-oxide
(a-IGZO) with electron mobilities exceeding 10 cm2 /V s. The high mobility and amorphous
structure of AMOS enable a faster operation of the devices and elimination of grain-boundary
related defects, respectively.6 4 ,65 The high electron mobility in the amorphous state originates
from the conduction band, which mainly consists of unoccupied Ns (N > 5) orbitals of metal
atoms such as In-5s in a-IGZO. 66- 68 The spherical symmetry of the s orbitals makes them less
prone to form traps and scattering centers when the metal-oxygen-metal bonds are distorted by
rotation of metal-oxygen polyhedrons.
Amorphous zinc-tin-oxide (a-ZTO) is recently highlighted as a promising candidate material
for cost-effective applications, as it is free from toxic and expensive elements (e.g., indium and
gallium).13 ,69-7 6 To synthesize high-quality a-ZTO TFs, various techniques have been used,
including
sol-gel,"
sputtering,7 4 pulsed laser deposition,7 6 and atomic layer deposition
(ALD). 69 7 5 In particular, ALD provides a precise and conformal control of the atomic
concentrations of Zn ([Zn]) and Sn ([Sn]) in the film. By controlling the [Zn]/[Sn] ratio, ALD aZTO channels in TF transistors (TFT) have demonstrated a field-effect mobility up to 13
70
cm 2 /V-s with a large on-to-off ratio of drain current (>109).69 Furthermore, controlling the
[Zn]/[Sn] ratio allows tunability of conduction-band edge positions, which is also useful for
photovoltaic
performance.
E
-
E
<
-
E
applications,
reducing
interface
recombination
and
device
improving
477
1:0.27 (Zn:Sn)
1:0.59
1:1.8
CB
znO-0.090.2
0.10
0.2 0.28 0.2
jE.,IeV
4->
oQ
) -6
0
-1.47
0.2
VB
2
-8
Cu20
0.0
0.4
0.2
(1:027) (1:0.59) (1:18
a-ZTO (Zn:Sn)
-8
ZnO
0.6
Bias/ V
(a)
(b)
Figure 5.1: (a) J-V characteristics of Au/Cu 20/a-ZTO/ZnO:A/Al solar cells fabricated with a-ZTO of different
compositions and (b) Relative alignments of conduction band (CB) and valence band (VB) for a-ZTO and ZnO
overlayers to Cu 2O TFs as measured by XPS. Plots reproduced from Lee et al."
Figure 5.1(a) shows the J-V characteristics of Au/Cu20/a-ZTO/ZnO:AI/Al
solar cells
fabricated with a-ZTO of different compositions. A strong dependence of Voc on the chemical
composition of the a-ZTO is observed. There is a clear trend in improving Voc with increasing
Sn content, up to a Zn to Sn ratio of 3.7:1. At higher Sn contents, the Voc drops. As shown in
Figure 5.1(b), XPS measurements indicate that the improvement in band alignment corresponds
to an increase in Voc. However, Sn-rich a-ZTO films are detrimental toward solar cell
performance despite a favorable band alignment. SCAPS simulations presented in section 3.3.2
suggest that this could be due to the presence of deep-level defect states in Cu2O/Sn-rich a-ZTO
hetero-interface that enhances recombination. In addition, capacitance-frequency measurements
from Figure 5.2 seem to support this hypothesis. At low frequencies, the capacitance of all solar
71
cells plateau to a depletion capacitance
(C),
which is affected by charging and discharging of
interfacial and bulk defect levels present in the depletion region. Due to identical geometry and
fabrication processes across the devices except for 5-nm-thick buffer layer materials with high
resistivity, the relative change in Cd can be attributed to the defects from the a-ZTO buffer layers.
The highest efficiency device (Zn:Sn = 1:0.27) exhibits the lowest Cd, indicative of lower defect
densities. On the contrary, the device (Zn:Sn = 1:1.8) with the lowest Voc exhibits the highest Cd
possibly due to higher densities of subgap states that contributes to additional capacitance. The
significant impact of the [Zn]/[Sn] ratio in a-ZTO TFs on both TFT and solar cells suggests a
strong correlation between the atomic structure and electronic properties of the films. Because
amorphous materials can exhibit a continuum of structures, 78 a probe of local order is needed to
investigate this relationship.
60,
E
cd
U
d
LL.
Zn:Sn
- 1:0.27
40
1:0.59
-
1:1.8
+ZnO
20
0 0 103
10 5
10 4
10 6
Frequency / H z
Figure 5.2: Capacitance-frequency measurements for Au/Cu20/a-ZTO/ZnO:A/Al solar cells fabricated with aZTO of different compositions.
In this work, the effect of a-ZTO TF atomic structure on electron transport properties is
investigated using synchrotron-based XAS. EXAFS at the Zn and Sn K-edges is used to probe
the local chemical neighborhoods of Zn and Sn atoms in a-ZTO films, respectively. The DebyeWaller factor, which gives a measure of structural disorder, 79 at the Zn K-edge increases with
increasing [Sn] in the films. The structural disorder is further investigated by XANES analysis,
which indicates the amorphization around Zn atoms with increasing [Sn] in the films. Disorder
72
as measured by EXAFS and XANES coincides with a decrease in measured electron mobility,
suggesting that the degradation in electron mobility may be correlated with structural changes for
Zn rich films ([Zn] > 0.5) whereas for Sn rich regions, larger ionic size of Sn dominates the
electron mobility. Lastly, increasing structural disorder around Zn atoms of a-ZTO films is also
found to correlate with decreasing solar cell performance. This suggests that structural disorder
might play an important role in increasing the density of deep-level states at the Cu20/a-ZTO
interface.
5.3
Experimental Details
A set of a-ZTO TFs is grown on quartz substrates by ALD at a growth temperature of
120 'C. Diethylzinc (DEZ) and a cyclic tin (II) amide ((1,3-bis(1,1-dimethylethyl)-4,5-dimethyl(4R,5R)-1,3,2-diazastannolidin-2-ylidene)Sn(II))
0
are used as the Zn and Sn precursors,
respectively. Hydrogen peroxide (H202, 50 wt.%, Sigma Aldrich) is used as the common oxygen
source. The compositions of a-ZTO films are controlled by varying the sub-cycle ratio of ZnO
and Sn02. Pure ZnO and Sn02 films are also deposited by ALD at 120 'C on quartz substrates,
respectively. The thickness of all films is -200 nm to minimize self-absorption issues in XAS
measurements. The atomic compositions of the resulting films are measured by Rutherford
backscattering spectroscopy (RBS), summarized in Table 5.1.
XAS is performed at the MRCAT Sector 10-ID beamline of the Advanced Photon Source,
Argonne National Laboratory (IL, USA). The beamline uses a cryo-cooled Si(1 11) double
crystal monochromator. In this study, we perform XAS at both the Zn and Sn K-edge transitions,
which occur at 9.66 and 29.2 keV, respectively. The TF samples are measured in fluorescence
mode with an incident beam of area 500
x
500 Rm 2 and a 13-element Canberra germanium solid-
state detector with a liquid nitrogen coolant to collect the fluorescence emission. To improve the
73
signal-to-noise ratio, multiple scans are averaged together to achieve effective counts higher than
106. The energy scale is calibrated by using the absorption edge of reference metallic Zn or Sn
TFs measured simultaneously and the XANES and EXAFS are isolated by normalizing the
absorption spectrum and subtracting the smooth atomic background absorption signal from the
measured absorption signal using the AUTOBK algorithm in Athena with Rbkg = 1.0
A.4 181 After
the background removal, the processed data are transformed from energy space to k-space using
the relationship, k 2 = 2m(E - EO) / h 2 , where k is the electron wavenumber, m is the electron
mass, Eo is the K-edge absorption energy of the respective elements, and h is Planck's constant.
The entire spectrum is weighted by k2 to compensate for amplitude decay. For further analysis,
the k2-weighted spectra data are Fourier-transformed with a Hanning window as a bandpass filter
A-1
.
to enhance the signal to noise ratio defined from k = 1.5 to 8.0
Table 5.1: Atomic composition of a-ZTO films measured by RBS and x = [Sn)/([Sn]+[Zn]).
ZnO/SnO2
Sample
5.4
sub-cycle
ratio
Atomic contents
[0]
(at.%)
[Zn]
(at.%)
[Sn]
(at.%)
X
ZITO
1/0
51.8
48.2
0
0
Z3T1
3/1
57.6
33.4
9.0
0.21
ZITI
1/1
61.7
24.2
14.0
0.37
ZIT3
1/3
65.8
12.2
22.0
0.64
ZOTI
0/1
67.9
0
32.1
1
Results and Discussions
Figure 5.3 shows the resulting Fourier-transformed spectra plotted as the magnitude, JX(R)J,
for both the Zn and Sn K-edges. The first large peak in the spectrum is indicative of the
scattering signal from the first nearest neighbor (INN) shell of atoms. 8 1 It can also be noted that
74
the amplitudes of [X(R)l from higher-order shells (R > 2 A) for all other spectra are strongly
attenuated, except for the spectrum collected at Zn K-edge for sample Z1TO (pure ZnO). This
suggests that the higher-order shells around the Zn atoms are well-ordered in the pure ZnO
sample, verifying its crystalline nature.
For all a-ZTO and SnO2 films, the EXAFS spectra exhibit limited structure beyond INN,
indicating a lack of long-range order. The lack of long-range order in the Z3T1, ZI T1, and ZI T3
samples is consistent with the amorphous structure as characterized by X-ray diffraction
measurements we have reported elsewhere. 13 Previously, we have reported that SnO2 deposited
by ALD at 120 'C is nano-crystalline (nc-SnO2). 8 2 The nc-SnO2 films are grown using a closedvalve mode, during which the nitrogen flow into the reactor is stopped until a base pressure (50
mTorr) is reached, and then a valve between the ALD reactor and the pump is closed during the
injection of the Sn precursor.2 ' In contrast to these previous results, the SnO2 films in this study
are found to be amorphous when made by an open-valve ALD method without either stopping
the nitrogen flow or pumping the reactor down to base pressure between precursor pulses.
Nevertheless, the pure ZnO films grown with open-valve ALD method are polycrystalline.
These results are found by XRD measurements (not shown). From Figure 5.3(a) and (b), it can
also be observed that the amplitude of the first shell peak decreases with increasing [Sn].
75
2.0
(a)
(b)
Zn K-edge
V
[Sn
[Sn]
]
1.21
2
2
(d)
(c)
0
0
0
0 0oZ1TO
00
0
4
-
1
Sn K-edge
-
000
Z3T1
Z3T1
Z1T1
Z1T1
Z1T3
Z1T3
a-SnO2
3
2
1
ZOT1
0
0
1
2
3
R
[A]
4
5 0
2
1
R
3
[A]
Figure 5.3: Fourier-transformed EXAFS spectra at Zn and Sn K-edges. Figures (a) and (b) compare the peak
intensity of the INN shell at the Zn and Sn K-edges respectively. The arrows illustrate the decreasing peak
intensity with increasing [Sn]. Figures (c) and (d) show the measured (symbols) and fitted spectra (lines) at the Zn
and Sn K-edges respectively (offsetted for clarity).
To gain quantitative structural information for the first shell, the INN peaks from all
samples (1.0 - 2.0 A for the Zn K-edge and 1.0 - 2.1 A for the Sn K-edge) are isolated and fitted
using the EXAFS equation described in equation (3.12). Due to the similarity in the first-shell
EXAFS spectra between a-ZTO samples and reference samples, the local structures around the
Zn and Sn atoms of the a-ZTO films are expected to resemble those of crystalline ZnO and
amorphous SnO2 respectively. Hence, the model scattering paths used in the data fitting routines
are calculated using the crystal structures of ZnO (wurtzite, space group P63mc)83 and SnO2
(rutile, space group P42/mnm) 84 as starting inputs using the ATOMS and FEFF6 codes
76
implemented in Artemis.4 1'85 In this case, the INNs of Zn and Sn atoms contain only oxygen
atoms, thus the scattering paths consist only of Zn-O and Sn-O bonds respectively. To reduce the
number of fitting parameters, S for both Zn (0.84) and Sn (1.00) are obtained by fitting the
EXAFS spectra of metallic Zn or Sn TFs collected simultaneously with each sample.
Subsequently, non-linear least squares fitting is performed in Artemis to obtain the fitting
parameters as summarized in Table 5.11.
Table 5.11: Summary of EXAFS fit parameters for Zn-O and Sn-O bonds. x = [Sn]/([Sn]+[Zn]).
EXAFS Fit Parameters
Sample
Sn K-Edge
Zn K-Edge
x
R
[A[e]
[A]]Nfac[t]r-
fa tor
[eV]
factor
R [A]
N
a2
[A 2]
r
Eo
0
1.97(3)
3.8(2)
0.002(3)
6.0(8)
0.003
-
-
-
-
Z3TI
0.21
1.97(6)
3.7(6)
0.005(4)
4.7(1)
0.002
2.04(7)
5.9(7)
0.006(0)
7.5(6)
0.004
ZITI
0.37
1.98(0)
3.7(7)
0.006(8)
4.4(3)
0.001
2.04(9)
5.9(0)
0.006(0)
7.4(3)
0.005
ZIT3
0.64
1.98(0)
3.8(9)
0.007(7)
4.0(5)
0.005
2.04(5)
5.8(9)
0.006(6)
7.4(1)
0.003
Sn02
1
-
-
-
-
-
2.04(7)
5.9(4)
0.007(1)
7.0(7)
0.003
-
ZnO
It can be observed that the bond lengths of Zn-O and Sn-O of all a-ZTO samples have very
small changes with respect to composition and are comparable with the reference ZnO and Sn02.
The fitted average bond-lengths, Rz -
1.98 A and Rs.-~ 2.05 A, are also in close agreement
with previously reported values for wurtzite ZnO 83 and rutile Sn02.8 4 In addition, the fitting
results also indicate that Zn and Sn atoms have tetrahedral and octahedral coordination
respectively, as predicted by theoretical calculations.86 87 This coordination is consistent with the
oxygen-rich atomic composition measured by RBS, which can be attributed to the use of H 202 as
the oxidizing agent. 13,82
77
2
(a) Zn K-edge
A
(D
(b) Sn K-edge
B
N1
0
- Z3T1
- Z1T1
- Z1T3
- ZOT1
Z1TO
- Z3T1
- Z1T1
- Z1T3
0
9.66
9.67
'
9.68 29.20
'
0-
29.25
29.30
Energy [keV]
Energy [keV]
Figure 5.4: XANES spectra at (a) Zn and (b) Sn K-edges. The 2 isosbestic points are circled and labelled A and B
in (a).
While both R and N have very small changes with Sn content, higher [Sn] leads to an
increase in the pseudo-Debye-Waller factor, cu2 , (the subscript is used to denote the respective
metal atom) in Zn-O path. The increase in o
can be attributed to an increase in the level of
structural disorder surrounding the Zn atoms resulting in a larger spread in the Zn-O bond-length
or distortion of the ZnO4 tetrahedra.7 9 Structural disorder further manifests itself as featural
changes in the Zn XANES spectra as shown in Figure 5.4(a). Such featural changes in the Zn
XANES spectra from 9.66 to 9.67 keV are due to a lack of the distinct multiple scattering
contributions and have been associated with the amorphization of the Zn chemical environment
as observed by Cho et al..88 In addition, isosbestic points observed in Zn-edge XANES suggest
that the local environments of Zn atoms in the three a-ZTO films comprise of a two-phase
mixture that is distinctly different from crystalline wurtzite ZnO structure. One plausible
explanation of the presence of two phases can be attributed to a segregation of ZnO and Sn02
which is likely to occur considering the distinct difference in INN coordination oxygen atoms
between Zn and Sn. One of the phases could be due to Zn atoms which lie along the interface
78
between the Zn-O tetrahedron and Sn-O octahedron and experience a different coordination
environment whereas the second phase can be due to the Zn atoms located further away from the
ZnO/SnO2 interface.
4
70
*
+
Field-Effect Mobility
Device Efficiency
0
CD
.
Hall Mobility
3
*
10
-
E
*
0
2
0
CD
0
1
8
C
X
0
6
4
2
0
Zn K-edge
- Sn K-edge
0.0
0.2
0.4
0.6
0.8
1.0
[Snl/([Sn]+[Zn])
Figure 5.5: Field-effect/HalI-effect mobilities, device efficiency and pseudo-Debye-WalIer factors are plotted
against [Sn]/([Sn]+[Zn]).
In Figure 5.5, the Hall- and field-effect mobilities together with U2 of all samples are plotted
as a function of cation composition. The field-effect mobilities of a-ZTO are measured by
fabricating TF transistors with a-ZTO layers as n-channels and more details are reported
elsewhere.6 9 The Hall mobilities are measured by using the van der Pauw configuration and a
magnetic field of 0.75 T. The ZITI and ZIT3 samples exhibit too small Hall voltages to measure
Hall mobility. 13 In Zn-rich ([Sn]/([Sn]+[Zn]) < 0.5) a-ZTO samples, increasing aL is wellcorrelated with a decrease in mobility, with respect to increasing [Sn]. The decreased mobility
suggests that structural disorder around Zn atoms reduces the film's mobility significantly. In the
79
Sn-rich ([Sn]/([Sn]+[Zn]) = 0.64) a-ZTO sample, on the other hand, the mobility remains at ~ 0.5
cm 2 /V-s with a subtle increase in
072
Zn
structural disorder of Zn is increased.
For AMOS with high ionicity, the conduction band edge typically comprises of overlapping
Ns orbitals from the metal cations. 66,68',84 The isotropic nature of these Ns orbitals causes electron
transport to be less sensitive to bonding geometries. However, prior results from other authors on
similar material systems such as amorphous ZnO 88 8 9 (a-ZnO) and amorphous ZnSn39 0 (aZnSnO3) suggest that the conduction band of a-ZTO can comprise of a mixture of Zn-4s and Sn5s orbitals that are hybridized with O-2p orbitals. In particular, Cho et al. have shown that the
hybridization strength between Zn-4s and O-2p orbitals in a-ZnO is sensitive to structural
disorder. 89 In Zn-rich a-ZTO films, the predominant conduction paths are between hybridized
Zn-4s and O-2p orbitals. Thus, the reduced mobility by increased [Sn] can be attributed to the
structural disorder between Zn and 0 atoms; reduced hybridization strength of Zn-4s and O-2p
orbitals can limit electron conduction pathways. 88' 8 9 On the other hand, Sn-rich a-ZTO films
exhibit a very slight increase in the mobilities as [Sn] increases. The dependence of electron
mobilities on the composition of the larger metal ion has similarly been observed in Al-Zn-Sn091 and In-Ga-Zn-0 68 systems where the effect has been hypothesized to be due to species with
larger ionic radii (Sn4 * and In 3+ ions (N > 5) respectively) contributing towards conduction.
These results suggest that for the a-ZTO films investigated in this work, Sn4 + ions which have
larger ionic radii thus larger Ns orbital (N > 5), may begin to dominate electron conduction via
Sn-5s - O-2p pathways, thereby increasing electron mobility closer to that of Sn02.
Additionally, a correlation between solar cell efficiency and structural disorder in the a-ZTO
buffer layer can be observed. Earlier c-f results indicate that the density of defects at the Cu20/aZTO interface increases with [Sn]. One possible explanation to reconcile both XAS and c-f
80
observations is that structural disorder in a-ZTO introduces defects near the interface. These
defects could be due to dangling bonds at the Cu20/a-ZTO interface as a result of imperfect
passivation or bulk defects in the thin layer of a-ZTO buffer layer.
5.5
Conclusions
In summary, synchrotron-based XAS is used to investigate the effect of atomic structure of
a-ZTO TFs on their electron transport properties. The a-ZTO TFs exhibit higher degree of
structural disorder, as [Sn] increases in the films. Quantitative EXAFS analysis reveals a strong
correlation between U2 and electron mobility for Zn-rich films. XANES measurements provide
further evidence of structural disorder near Zn atoms in Zn-rich a-ZTO films. The decrease in
mobility is correlated with increasing local structural disorder surrounding Zn atoms. Literature
reports suggest the decrease in mobility may be due to reduced hybridization strength of Zn-4s
and
O-2p
orbitals, which may limit electron conduction pathways. A slight increase in mobility
for the Sn-rich a-ZTO films is observed. A plausible explanation can be that the Sn atoms which
have larger Ns orbitals (N > 5) begin to dominate electron transport in the a-ZTO films, leading
to an increase in electron mobility. Lastly, XAS and c-f measurements suggest that increasing
defects near the Cu20/a-ZTO interface might be due to dangling bonds at the Cu20/a-ZTO
interface as a result of imperfect passivation or bulk defects in the thin layer of a-ZTO buffer
layer
81
82
CHAPTER
6
DOPANT ACTIVATION IN SN-DOPED
GA2O3a
6.1
Abstract
Doping activity in both beta-phase (fl-) and amorphous (a-) Sn-doped gallium oxide
(Ga 2O 3 :Sn) are investigated by X-ray absorption spectroscopy (XAS). The single crystal /3Ga2 03 is grown using edge-defined film-fed growth at 1725 'C while the a-Ga2O3:Sn thin-films
(TFs) are grown at low temperature (<300 C). XAS analyses indicate that activated Sn dopant
atoms in conductive single crystal fl-Ga203 are present as Sn4",
preferentially substituting Ga at
the octahedral (Ga2) site, as predicted by theoretical calculations. On the other hand, inactive Sn
atoms in resistive a-Ga203 are present in either +2 or +4 charge states depending on growth
conditions. These observations suggest the importance of growing Ga 2O 3 at high temperature to
obtain a crystalline phase and controlling the oxidation state of Sn during growth to achieve
dopant activation.
S. C. Siah, R. E. Brandt, L. T. Schelhas, K. Lim, J. D. Perkins, R. Jaramillo, M. D. Heinemann, D. Chua, J.
Wright, C. U. Segre, R. G. Gordon, M. Toney, and T. Buonassisi (in preparation for submission).
a
83
6.2
Introduction
Gallium oxide (Ga20 3) has many favorable optoelectronic properties that make it a
15 6 93
92
promising semiconductor for solid-state applications in field-effect transistors, solar cells, " 1'
gas sensors, 94 and lasers. 95 Furthermore, the successful epitaxial growth of GaN on beta-phase
Ga203 (p-Ga203) substrate has shown the potential for high-brightness, cost-effective GaN
LEDs.96 9 8 . In doped Ga2O 3 , series resistance losses in epitaxially grown GaN LEDs can be
reduced by enabling vertical current injection, as compared to horizontal structures which suffer
from current crowding. 96 Furthermore, good electrical conductivity couples with good optical
transparency in the visible range makes Ga203 an ideal transparent conducting oxide (TCO),
suitable for transparent electronics. More importantly, there is a need to develop TCOs which are
highly conductive and transparent, yet have very small electron affinities. Having small electron
affinities is important to pair the TCO as a window layer with new photovoltaic absorber
materials including GaN and Cu20, which have smaller electron affinities than traditional
semiconductors including Si and GaAs. A TCO with small electron affinity can potentially be
applied as an 'electron selective contact' in crystalline silicon solar cells. 93 There are very few
materials which currently satisfy these criteria, but single crystal Ga2O3:Sn have shown
promise. 96,99 The essential problem
is to replicate the simultaneous conductivity
and
transmissivity of Ga2O3:Sn bulk crystals in TF polycrystalline Ga2O3:Sn. Hence, it is important
to understand the extrinsic doping mechanism of Sn in single crystal Ga2O3 so that TFs
Ga203:Sn can be engineered during growth to attain similar high dopant activation and
transmissivity.
In this chapter, synchrotron-based XAS is used to probe differences the local atomic
structure and chemical state of both Sn dopant atoms and Ga host atoms in both 8-Ga2O3:Sn
84
single crystal and amorphous (a-) Ga2O3:Sn films grown by ALD and PLD. The local structures
of the metal cations in these different samples will provide insights into how dopant atoms are
incorporated
into the host Ga2O3 lattices and elucidate how this determines electrical
conductivity.
6.3
Experimental Details
/8-Ga2O3:Sn single crystal (SC) is purchased commercially from Tamura Corporation and is
grown from melt using the edge-defined film-fed growth (EFG) method. 92 In addition, aGa2O3:Sn TFs are grown using atomic layer deposition (ALD) and pulsed laser deposition (PLD)
methods. The ALD film is deposited at 120 'C in a custom-built cylindrical reactor with a 30 cm
long and 3 cm wide sample stage, and a chamber volume of 0.627 L. The Ga and Sn precursors
used in the ALD
process are B is(pt-dimethylamino)tetrakis(dimethylamino)digallium
and
Tetrakis(dimethylanido)tin(IV) respectively. The oxygen source is H 20. During the ALD
process, the temperatures of the Ga and Sn are maintained at 120 and 60 'C, respectively while
H 20 is kept at 25 'C. High-purity N 2 is used as a carrier gas and the exposures of the gallium
pre-cursor and H 20 are estimated to be approximately 3 and 5 Torr-s, respectively. The
deposition rate is measured to be -0.2 nm per cycle. The PLD film is prepared using Ga203 and
SnO2 targets and the energy density of the pulsed KrF excimer laser (248 nm) is set to 300 mJ
with a repetition rate of 10 Hz and a distance of 10 cm between the target and the sample
substrate. The substrate is rotated during 50 laser pulses applied to, the Ga2O3 target, and kept at
a fixed angle during 2 pulses applied to the SnO2 target. This procedure is repeated 400 times,
resulting in a homogeneous Ga2O3 film with a Sn doping gradient across the sample from
approximately I at.% to 4 at.%. The oxygen partial pressure is set to 100 pTorr for the
depositions at 400'C to ensure the formation of stoichiometric films.
85
The atomic wt. % of Sn of all 3 samples are measured by calibrated X-ray wavelength
dispersive spectroscopy (WDS). Hall and four-point probe measurements are performed for all
the samples but only the SC sample exhibits detectable signals. The Hall mobility and net carrier
density of the SC is determined to be 110 cm 2 /V.s and 5.6x 1018 cm-3 respectively. The
resistivities for the ALD and PLD samples are estimated to be > 2000 Q-cm based on the
detection limit of the four-point probe system. By assuming a mobility of > 0.1 cm 2 /V-s, the
upper limit of carrier density is estimated to be 1.5x 1016 cm-3 for the a-Ga2O3:Sn films. Figure
6.1 compares the net carrier density of all experimental samples, including data obtained from
Tamura Corporation for a suite of SC samples with varying Sn concentrations. It can be observed
that the SC samples grown via EFG have an activation ratio close to 100%, whereas the Sn
dopant atoms in the ALD and PLD grown films are either un-activated or completely
compensated, as evident through their high resistivities.
1021
-,----, .-,
1020
S
1
ft-Ga 2 0 3 Single Crystals
m PLD
* ALD
*
19
E 10
10
100% activation
1
Z011017
116
Upper Limit for a-Ga2O3 Sn
10 1517
10
18
10
19
10
N6 [cm 3 ]
20
10
21
10
Figure 6.1: Net carrier density of B-Ga 203:Sn single crystals as a function of varying [Sn], compared against the
deduced upper limit for a-Ga 2O3:Sn films.
Ga K-edge XAS is performed at BL 4-3 at the Stanford Synchrotron Radiation Lightsource
and Sn K-edge XAS at MRCAT Sector 10-ID of the Advanced Photon Source. In both
measurements, the TF samples are measured in fluorescence mode with an incident beam of
86
approximately 500 x 500 ptm2 . The K-edge fluorescence for Ga and Sn is measured by a Lytle
detector and silicon Vortex solid-state detector respectively. Reference metallic Ga or Sn thinfoils are measured to account for relative energy drifts. The X-ray absorption near-edge
structures (XANES) and extended X-ray absorption fine structures (EXAFS) are isolated by
normalizing the absorption spectrum and subtracting the smooth atomic background absorption
signal from the measured absorption signal using the AUTOBK algorithm in Athena with
1.0
A. 3 8 ,8 1, 100
Rbkg =
After background removal, the processed data are transformed from energy space
to k-space using the relationship, k 2 = 2m(E- Eo)/h2 , where k is the electron wavenumber, m
is the electron mass, Eo is the K-edge absorption energy of the respective elements, and h is
Planck's constant. The spectra are weighted by k 2 to compensate for amplitude decay. For further
analysis, the k 2-weighted spectra data are Fourier-transformed with a Hanning window as a
bandpass filter to enhance the signal to noise ratio within windows between k = 1.5 to 10.0
6.4
A-'.
Results and Discussions
Firstly, information about the chemical states of Sn for each sample can be derived by
comparing, in Figure 6.2, the respective XANES spectra with Sn metal foil, SnO powder and
SnO2 powder references. The relative chemical shifts observed in the XANES spectra are
believed to be due to changes in valency, which alters the binding energy for electrons in the first
shell.1 0 1 Figure 6.2 shows that the average charge states of Sn atoms in the SC and ALD samples
are similar to that of SnO2 (Sn 4 ). Comparing this with the resistivity data suggests that Sn4 can
function as an electron donor under the correct conditions. However, its presence does not
always result in free electrons due to other reasons including the formation of compensating
defects or formation of secondary phases. The average oxidation state in the PLD sample
corresponds to Sn 2 + (SnO) which is not believed to act as an electron donor. The presence of this
87
reduced state (compared to the SC and ALD) suggests that the growth environment could be too
reducing relative to the ALD and SC growth processes.
1.4
-
S1.2
-
-
--
Sn Foil-
Sno
SnO 2
CL
-
01.0
-
.
<0.8
N0.6 - SC
z
0 2
0.0-
ALD
-
0.4
PLD
-I-
(a
'29.1829.20
- (b)
29.18 29.20
Energy [keV]
(C)
29.18 29.20
Figure 6.2: Sn edge XANES spectra for (a) SC 8-Ga 2O3:Sn, (b) PLD a-Ga 2O3:Sn and (c) ALD a-Ga 2O3:Sn
samples. The dashed, dashed-dotted and dotted lines represent Sn(O) metal (blue), Sn(II)O (green) and Sn(IV)0 2
(purple) references respectively.
Next, EXAFS is used to investigate the structural origin of Sn doping. Figure 6.3 shows the
Fourier-transformed spectra plotted as the magnitude, [X(R)I, for both the Ga and Sn K-edges.
The first large peak in the [X(R)l spectrum is due to only single-scattering paths from the first
nearest neighbor (INN) shell of atoms, and higher-order peaks are due to single- and multiplescattering paths involving neighboring atoms in INN and higher order shells. In both sets of
spectra, the amplitudes of LX(R) from higher-order shells (R > 2 A) for the ALD and PLD grown
samples are strongly attenuated, showing limited structural order beyond INN. The lack of longrange order is consistent with the amorphous structure as characterized by X-ray diffraction
measurements. By comparing the Sn EXAFS spectra with the standard references and from the
diffraction data, any presence of crystalline SnO2 or SnO secondary phases is not observed,
although segregation of diluted nanocrystalline phases cannot be ruled out.
88
7
(a)
a K-edge
(b) n K-edge
6
6
SnO
5
23
SnO
-SC
\
-SC
2JL
PLD
U
ALD
00
ALD
1
2
3
4 0
R +Ar [A]
1
2
3
4
R +Ar [A]
Figure 6.3: Fourier-transformed EXAFS spectra plotted as the magnitude,
The spectra for SnO and SnO2 are included as reference.
IX(R), for (a) Ga and (b) Sn K-edges.
To gain quantitative local structural information, the peaks are isolated and fitted using the
EXAFS equation described in equation (3.12). The fitting parameters for each scattering path are
the changes in the half path length (AReff),
U
and energy shift (AEo). For the SC sample, the
spectra are fitted up to the second-order peak, and the scattering paths used in the data fitting
routines are calculated using the crystal structure of P-Ga2O3:Sn (space group CI 2/ml)
0210
3
as a
starting input into the ATOMS and FEFF6 codes implemented in Artemis. 4 In the unit cell of$fGa2O3,
there are two crystallographically nonequivalent Ga atoms (Gai and Ga2) and three
nonequivalent 0 atoms (Oi, 02 and 03). The Ga K-edge spectrum for the SC sample is modeled
by considering equal contributions from the Gai and Ga2 sites,1 0 4 and the Sn K-edge spectrum is
fitted by considering either substitutional Sn-on-Gai (SnGal) or Sn-on-Ga2 (SnGa2) defects.
Scattering path-lengths up to 3.5 A are considered for fitting the SC sample. For the amorphous
TF samples, only the first order peak is fitted by considering Ga-O and Sn-O bonds in the INN
89
shell. The non-linear least squares fitting routine is subsequently performed in Artemis to obtain
the best-fit parameters. The best-fit parameters for the Sn K-edge spectra are included in Table
6.1, while those for the Ga K-edge spectra are included as supplemental materials as many
scattering paths are involved.
Table 6.1: Sn K-edge EXAFS parameters for all samples. The SC is fitted by assuming either Sn substitution on
Gai and Ga2 sites. The best fit is obtained for Sn substitution at Ga 2 site. Scattering path-lengths up to 3.5 A are
considered for fitting the SC sample.
Site
Gai
Shell
First
First
SC
Ga2
Second
Path Description
N
[Sni] - 01 - [Sni]
I
[Sni]
02 - [Sni]
1
[Sni] - 03 - [Sni]
2
[Sn2]
01 - [Sn2]
2
[Sn2] - 02 - [Sn2]
3
[Sn2] - 03 - [Sn2]
I
[Sn2] - Gal - [Sn2]
7
0.11(2)
[Sn2] - Ga2 - [Sn2]
4
0.02(1)
[Sn2]
Oi - [Sn2]
2
[Sn2] - 02 - [Sn2]
2
-
03 - [Sn2]
4
-
03
[Sn2]
-
-
-
.2 [A- 2]
AEo [eV]
AReff [A]
5(2)
0.05(1)
0.000(2)
3.6(3)
0.010(3)
0.0010(5)
R-Factor
0.05
3.6(3)
-0.14(6)
0.02(1)
ALD
First
[Sn]
[Sn]
5.0(2)
4.4(3)
0.090(3)
0.0076(6)
0.005
PLD
First
[Sn] - 03 - [Sn]
5.1(2)
8.2(3)
0.177(4)
0.0128(8)
0.008
-
As shown in Figure 6.3, the good agreement between the Ga K-edge spectrum for the SC /Ga203:Sn and the model up to the second order shell corroborates the beta-phase nature of the
host lattice. By considering two different possibilities in which Sn atoms can be incorporated
into the SC 8-Ga203 host lattice, it is found that there is a preferential substitution at the Ga2
octahedral site (R-factor = 0.01) as compared to substitution at either the Gai (R-factor = 0.05).
This observation is also consistent with first-principles calculations by Varley et al.'O5 Several
other studies have also shown that transition metals like In,1 06 Cr,1 07 and Mn' 08 have a preference
for the octahedral site and can be explained by steric reasons.1
90
07 08
1
Preferential substitution at the
octahedral site can potentially impose an upper limit on doping concentration, because only 50%
of the Ga sites can be occupied.' 06 Additionally, calculations by Maccioni et al.'
06
show that the
solubility of In atoms in Ga2O3 is limited to a concentration of 12 % because the formation
energy of InGa2 defect increases with every subsequent addition of In atom. Due to the proximity
of In and Sn in the periodic table, Ga203:Sn might also suffer from similar limitation
For the ALD and PLD amorphous TFs, it is found that the average coordination number to
o
atoms in Sn's INN shell is close to 5.0, despite the difference in charge state of the central
absorbing Sn atoms. This suggests that the amorphous films might not be grown in
thermodynamic equilibrium conditions, because Sn atoms in Sn02 (Sn 4 *) and SnO (Sn2+) tend to
favor six-fold and four-fold coordination respectively. Despite similar coordination numbers, the
larger Sn2+ ions in the PLD TF increases the average Sn-O bond-length (ARef= 0.177 A) as well
as the structural disorder (U 2 = 0.0128
A-2 ) relative to the
ALD films (AReff = 0.090 A and a'
= 0.0076 A-2 ). Finally, a-Ga203 may be intrinsically limited because of its amorphous structure.
For example, the presence of electron or hole traps in amorphous structure can potentially
impose a doping limit by pinning the bulk Fermi-level
091 1 0
The intrinsic conductivity of
amorphous materials, and the formation of compensating defects, might help explain the low
conductivity achieved by the ALD sample despite the high concentrations of Sn 4" dopants.
6.5
Conclusions
In conclusion, XAS is used to investigate the differences in the local structures of
conductive single-crystal P-Ga2 O 3:Sn and resistive a-Ga2O3:Sn TFs. The results obtained in this
work can potentially be used to help engineer Ga203:Sn TFs with the favorable electrical
properties of single crystal ,-Ga2O3:Sn. XAS analyses indicate that activated Sn dopant atoms in
conductive single-crystal ,-Ga2O3:Sn are present as Sn4 *, which preferentially substitutes for Ga
91
at the octahedral Ga2 site as predicted by previous theoretical calculations. On the other hand,
inactive Sn dopants in resistive a-Ga203 are present in either +2 or +4 charge states, depending
on growth conditions. XAS results indicate a lack of structural order beyond the INN shell. It is
hypothesize that the low conductivity of a-Ga203 TFs might be due to its amorphous structure
which introduces electron traps. Hence, achieving crystalline Ga2O 3 and controlling the
oxidation state of Sn during growth are both important to attain good conductivity.
6.6
Supplemental Material
Table 6.11: Ga K-edge EXAFS parameters for all samples. Scattering path-lengths up to 3.5
A are considered
for fitting the SC sample.
Site
Shell
First
Ga,
Second
SC
First
Ga2
Second
Path Description
N
[Gai] - 01 - [Gai]
1
[Gai]
-
02 - [Gai]
1
[Gai]
-
03
[Gai]
2
0.02(2)
[Gai] - Gai - [Gai]
[Gai] - Ga2 - [Gai]
2
7
0.02(2)
-0.085(4)
[Gal]-0, -[Gal]
5
0.11(4)
[Gai] -02 - [Gai]
1
[Gai] - 03 - [Gai]
[Ga 2] - 01 - [Ga2]
4
[Ga 2]
-
02
[Ga2]
3
[Ga 2]
-
03 - [Ga2]
1
0.02(2)
[Ga2] - Gal - [Ga 2]
7
0.02(2)
[Ga2 ]- Ga2 -[Ga 2]
4
-0.085(4)
- [Ga2]
2
0.11(4)
[Ga 2] -02 - [Ga2]
2
[Ga 2]
4
[Ga 2] -0
-
-
-
03 - [Ga2]
AEo [eV]
AReff [A]
U2
[^ 2 ]
R-Factor
-0.04(2)
5.7(5)
5.7(5)
-0.04(2)
-0.01(8)
0.008(1)
0.0001(1)
0.07(2)
2
0.03
-0.04(2)
5.7(5)
5.7(5)
-0.04(2)
-0.01(8)
0.008(1)
0.0001(1)
0.07(2)
ALD
-
First
[Ga] -03- [Ga]
4.2(1)
5.0(2)
0.037(2)
0.0059(3)
0.003
PLD
-
First
[Ga] -03- [Ga]
3.9(1)
5.3(3)
0.032(3)
0.0051(5)
0.006
92
CHAPTER
7
BULK DEFECT ENGINEERING IN CU 2 Oa
7.1
Abstract
In this work, the electronic properties of wafer-based polycrystalline Cu 2 0 is tuned by
tailoring the oxygen partial pressure and temperature profile of the furnace during the postannealing step. As a result, an increase in NA
-ND
of over two orders of magnitude is
demonstrated from 1012 to 101 4 cm 3 . In addition, microwave-reflection photoconductance
lifetime of up to II ps is measured. Spectrally-resolved room-temperature photoluminescence
reveals a broad mid-gap defect band centered at 1.3 eV and samples with lower lifetime show
comparatively stronger defect-related PL emission. This suggests that radiative recombination
from mid-gap states might be detrimental to the lifetime of photo-generated carriers. It is unclear
whether the mid-gap defect band play a role in increasing NA
- ND
but it might be possible that
these defect states contribute to compensation through donating electrons or defect-pairings.
aWork
present as S. C. Siah, M. Lloyd, S. W. Johnston, R. Brandt, Y. S. Lee and T. Buonassisi at 2014 Spring MRS
Meeting & Exhibit, Symposium E, Abstract E13.02.
93
7.2
Introduction
As discussed in previous chapters, efficiency enhancement in Cu20-based solar cells has
largely been driven by interface engineering. This typically involves controlling the chemistry
and band-alignment at the hetero-interface using thin buffer amorphous oxide layers such as ZnSn-O or Ga2O3 to achieve open-circuit voltages (Voc) as high as 1.2 V. However, the shortcircuit currents (Jsc) are still below the 14 mA/cm 2 theoretical entitlement for a 2.1 eV bandgap
absorber. Quantum efficiency analysis indicates that increasing charge collection length by
reducing bulk recombination is an important next step towards higher Jsc. Furthermore, reducing
bulk recombination can also allow the electron quasi-Fermi level in the absorber to be pushed
closer to the conduction band under illumination, further increasing Voc. To this end, the ability
to engineer the intrinsic point defect structure of Cu20 during its growth process to mitigate the
effects of deleterious bulk defects and tune its conductivity can be beneficial. By utilizing
appropriate tools to characterize relevant bulk electronic properties such as mobilities, carrier
concentration, defect levels and bulk minority carrier lifetime, the growth process of the absorber
can be systematically optimized for PV devices.
High quality poly-crystalline Cu20 wafers are grown by thermal oxidation as a proxy to
study intrinsic defect engineering methods. By tailoring the oxygen partial pressure and
temperature profile of the post-annealing process, it is shown that carrier concentration can be
tuned over almost two orders-of-magnitude while achieving a high mobility of 100 cm 2 /V.s, as
measured using Hall-effect. In samples that are annealed at high temperatures (> 500 'C),
photoconductance decay lifetime as high as 10 [is is measured using microwave reflection
photoconductance. Additionally, spectrally-resolved room-temperature photoluminescence (PL)
reveals a broad mid-gap defect band centered at 1.3 eV and samples with lower lifetime show
94
comparatively stronger defect-related PL emission. This suggests that radiative recombination
from mid-gap states is detrimental and dominates recombination activity in lower lifetime
samples. Spatial PL mapping is used to gain further insights into the structural dependence of
defect-assisted recombination and PL images show that grain boundaries are relatively more
recombination active than bulk region, pointing towards the need of larger grain sizes for higher
device efficiencies. Lastly, annealed sputtered Cu20-based devices show enhancement in shortcircuit current suggesting an improvement in carrier collection length.
7.3
Experimental Methods
Large grain Cu20 samples are grown by thermal oxidation of copper foil. 100 pim thick foils
of 99.95% purity by metals basis are purchased from Goodfellow Inc. Samples of I x I cm 2 are
suspended from a quartz rack using platinum wires to minimize surface contact during hightemperature processing steps. Sample structure are created by laser cutting the copper foils.
Figure 7.1 shows a typical Cu20 and its microscope image. It can be observed that grain size up
to 1000 prm can be achieved.
Thee oxidation process is carried out in a horizontal quartz tube furnace and Figure 7.2
illustrates the time-temperature profiles used in this study. The samples are first heated up to an
oxidation temperature of 1040*C at 15*C/min in a nitrogen atmosphere supplied by 5N purity
source at a flow rate of 944 standard cubic centimeter per minute (SCCM). After stabilization of
the temperature, an oxygen-rich atmosphere is attained by introducing an argon/oxygen mixture
(4:1 ratio) at a flow rate of 47 SCCM. The oxidation of the copper foils occurs for 60 minutes
under these conditions. Following the oxidation process, the temperature of the furnace is cooled
to another temperature (Tanneal) and annealed for an additional 120-minute in a N 2 ambient.
95
Lastly, a rapid quenching step is performed by dipping the samples quickly into a beaker of cold
water.
Figure 7.1: (a) A typical sample under illumination by microscope light. Sample is glowing red due to its
bandgap. (b) A microscope image of a similar as-grown foil under 50x magnification Scale bar represents 1000
pim.
Oxidation
60 mins
Post-Annealing
120 mins
- - - --
Rapid
Quench
-
500 *C
400
E
L - - - - - -
II
-F-A
oc
300 C
24 C
Time (min)
Figure 7.2: Time-temperature profiles for the growth and annealing process of Cu20 samples. (Not drawn to
scale)
The phase and crystal structure are characterized by X-ray diffraction (XRD) using a
PANalytical X'Pert Pro diffractometer with Cu-Ka radiation. Phase purity is confirmed by XRD
as the diffraction peaks of all samples are well matched to the reference pattern of Cu20 and the
peaks of other phases e.g., Cu and CuO are not detected. XPS is used to determine phase purity
at the surface. The thickness of CuO on the surface is deduced to be less than 5 nm by comparing
the ratio of the Cu2+ to Cu+ peak intensities.
96
Hall-effect measurements are carried out in the van der Pauw (VDP) configuration to
determine resistivity, carrier density and mobility of each Cu20 foil using ohmic Au contacts (I
mm2 area, 300 nm thickness) evaporated onto each corner the square samples. Microwave
photoconductive decay (pt-PCD) in reflection mode using a 630 nm optical excitation source is
used to characterize the samples. Subsequently, the conductance decay curves are fitted at the
low-injection regime to avoid any influence from surface recombination. Further information
regarding pt-PCD is described in greater detail by Johnston et al.'"
7.4
Results and Discussions
Figure 7.3 shows the measured sheet resistance
concentration (NA
increasing in
Tanneai
-
(psheet),
Hall-mobility (p al 1) and effect hole
ND) for samples grown with varying Tanneal. A decrease in Psheet with
is observed and this is mainly due to an increase in NA
remains relatively constant. Nearly two orders of magnitude increase in NAcontrolling
-ND
ND
since
/Hal
is achieved by
Tanneal.
It is well-known that Cu20 is intrinsically p-type due to the presence of copper vacancies
(Vcu) that is a shallow acceptor. 45 4 7 First-principles calculations of Cu20 by Raebiger et al.
suggest that the equilibrium density of Vcu increases with increasing growth temperature, leading
to higher hole density. This is consistent with the experimental observations presented in here as
the purpose of a rapid quenching step is to 'freeze-in' the equilibrium intrinsic point defect
structure of higher equilibrium temperatures.
97
107
0
I
I
U
106
aI
10 5
10 4
g-
I
I
110
105
100
95 B
90
U
U
I
I
E
z
1014
in
0 1012
U
-U
U
12
I
I
I
I
0
100
200
300
400
500
T anneal 1C
Figure 7.3: Measured sheet resistance, Hall-mobility and effect hole concentration f r samples grown with
varying Tanneal. The lines are guides for the eye.
10
TTanneal
---
i/5
--
0
(D"
--
24 OC
300 0C
400 *C
500 0C
_0
N
0d 10-2
0
10
20
30
40
50
Time [pis]
Figure 7.4: p-PCD decay transients for all samples. A double exponential behavior can be observed for samples at
low Tanneal.
Figure 7.4 shows the p-PCD decay transients for all samples. A double exponential
characteristic
can be observed
for samples with
Tanneai
< 400*C,
suggesting
different
recombination at short and long time scales. In Figure 7.5, spectrally-resolved PL reveal a
defect-related a broad mid-gap defect band centered at 1.3 eV. Subsequently, PL imaging
98
focusing on the defect-band on all samples reveal that samples with faster pt-PCD transient decay
show comparatively stronger defect-related PL emission.
1.60eV
1.03eV
532 nm excitation
No anneal
RG 1000 filter
and Si camera
CSWI
Ul)
600
800
1000
1200
1400
1600
Wavelength [nm]
max
min
(b)
(a)
Figure 7.5: (a) Spectrally-resolved PL for sample with no annealing, or Tanneal
samples. The defect-band that is being imaged is shown schematically in (a).
24*C and (b) PL imaging of all
To gain quantitative information, an effective p-PCD lifetime (rPCD) is determined by
fitting the second exponential decay in Figure 7.4. A spatially-integrated PL intensity (PLI) is
also obtained from each PL images. Figure 7.6 shows the relationship between
iPCD
and both
PLI and NA-ND. A relatively high rPCD up to 11 pas is obtained for the sample that is annealed
and quenched at 500 0 C. It is possible that the high rPCD could be due to persistent conductivity
)
observed in Cu20 reported previously.4 7 Olsen et al. reported an effective diffusion length ( Leff
of 12.3 ptm based on quantum efficiency measurements on wafer-based Cu20 Schottky diodes.
2
By assuming a minority carrier of 30 cm /V.s in Cu20, 11 pts corresponds to Leff = 27 Pm. The
proximity of both results suggest that rPCD could be used to estimate the actual minority carrier
lifetime in Cu20.
99
Evidently, a higher rPCD corresponds to a lower defect-band PL emission and higher
N-ND . One possible explanation to reconcile these observation is that post-annealing results
in the suppression of defect states near mid-gap, leading to an increase to rPCD. It is unclear
whether these defect states play a role in increasing NA
-ND
but it might be possible that these
defect states contribute to compensation by donating electrons or defect-pairings. Further studies
are warranted to provide clearer insights into the process-defect structure-property relationship
of Cu 20.
C/)
U
U
-
5
5
.t,
-~10-
00
0
U)
0
0
0
U~.
C,)
U
I
10-
1
2
I
I
I
3
4
5
1/PLI [a.u.]
(a)
x10
1012
4
1013
NA- ND [C
(b)
1 014
3
rPCD as a function of the inverse of spatially-integrated PL intensity (PLI) and (b)
as function of NA -ND
7.5
10-I
6
TPCD
Pl )tted
-
Figure 7.6: (a)
'
Conclusions
In summary, the intrinsic point defect structure of Cu20 is tuned by tailoring the oxygen
partial pressure and temperature profile of the furnace during the post-annealing step. As a result,
an increase in NA- ND of over two orders of magnitude is demonstrated. In addition,
CPCD
of up
to 11 [s is measured in thermally oxidized 100 ptm Cu20 bare wafers. Spectrally-resolved roomtemperature photo luminescence reveals a broad mid-gap defect band centered at 1.3 eV and
100
samples with lower lifetime show comparatively stronger defect-related PL emission. This
suggests that radiative recombination from mid-gap states might be detrimental to solar cell
performance. It is unclear whether the mid-gap defect band play a role in increasing NA
- ND
but it might be possible that these defect states contribute to compensation through donating
electrons or defect-pairings. Further studies are warranted to provide clearer insights into the
relationship between the intrinsic point defect structure and bulk electronic properties.
101
102
CHAPTER
8
CONCLUSIONS
This thesis is focused on improving the performance of Cu20 solar cells via defect
engineering. In Chapter 2, a bottom-up cost analysis to investigate the tradeoffs between
efficiency and costs for various top-cell deposition methods. The results that are presented
inform the economic viability of manufacturing and scaling up TF solar cells.
In Chapter 3, the formation of good ohmic back contacts on Cu20 absorber layer is
investigated and it is demonstrated that low contact resistivity can be achieved with a variety of
metals on heavily doped Cu20 films. This scheme is applied to Cu20 solar cells to reduce the
contact resistance of Ag due to a Schottky barrier at the back contact.
In Chapter 4, the fundamental loss mechanism in the amorphous Zn-Sn-O (a-ZTO) electronblocking layer is investigated by XAS and is found to have origin in local structural disorder
around Zn atoms. The presence of structural disorder in the a-ZTO buffer layer is found to
correlate strongly with electron mobility and solar cell efficiency. XAS and c-f measurements
suggest that increasing defects near the Cu20/a-ZTO interface might be due to dangling bonds at
the Cu20/a-ZTO interface as a result of imperfect passivation or bulk defects in the thin layer of
a-ZTO buffer layer.
Chapter 5 is motivated by the need to develop a transparent conducting oxide (TCO) that has
low electron affinity to be a good electron selective contacts for Cu20. Ga 2O 3 is identified as a
promising candidate, but its high resistivity limits PV device fill factor. XAS analyses indicate
103
that activated Sn dopant atoms in conductive single-crystal P-Ga2O3:Sn are present as Sn,
which preferentially substitutes for Ga at the octahedral Ga2 site as predicted by previous
theoretical calculations. On the other hand, inactive Sn dopants in resistive a-Ga2O3 are present
in either +2 or +4 charge states, depending on growth conditions. XAS results also indicate a
lack of structural order beyond the INN shell. It is hypothesize that the low conductivity of aGa203 TFs might be due to its amorphous structure which introduces electron traps. Hence,
achieving crystalline Ga 2O 3 and controlling the oxidation state of Sn during growth are both
important to attain good conductivity. These information can be used to inform growth of TF
Ga2O 3 on Cu20 for optimal device performance.
In Chapter 6, bulk defect engineering is used to manipulate the intrinsic point defect
structure of Cu20 towards improved device performance. Spectrally-resolved room-temperature
photoluminescence reveals a broad mid-gap defect band centered at 1.3 eV and samples with
lower lifetime show comparatively stronger defect-related PL emission. This suggests that
radiative recombination from mid-gap states might be detrimental to solar cell performance. It is
unclear whether the mid-gap defect band plays a role in increasing NA-ND but it might be
possible that these defect states contribute to compensation through donating electrons or defectpairings. Further studies are warranted to provide clearer insights into the relationship between
the intrinsic point defect structure and bulk electronic properties.
Taken together, the technological advances presented in this thesis contributed to our recent
creation of a record-efficiency electrochemically deposited Cu20 device, and highlight the
remaining technical challenges and economic window of opportunity for Cu20-based solar cell
devices, including tandems.
104
REFERENCES
2
F. Kaspar, M. Meinshausen, and M. Schulz, IPCC Climate Change Report (2013).
German Advisory Council on Global Change (WBGU), World in Transition: A Social Contract
for Sustainability(D-10 117 Berlin, Germany, 2011).
3
International Energy Agency (IEA), 2014 Snapshot of Global PVMarkets (2015).
4
D.M. Powell, M.T. Winkler, A. Goodrich, and T. Buonassisi, IEEE J. Photovoltaics 3, 662
(2013).
5 C. Wadia, a. P. Alivisatos, and D.M. Kammen, Environ. Sci. Technol. 43, 2072 (2009).
6
M.A. Lloyd, R.E. Brandt, and T. Buonassisi, in Accel. Dev. Earth Abund Thin-Film
Photovoltaics(Pasedena, 2013).
7
Y.S. Lee, M. Bertoni, M.K. Chan, G. Ceder, and T. Buonassisi, in 34th IEEE Photovolt. Spec.
Conf (2009).
8
Y.S. Lee, Defect Engineering of Cuprous Oxide Thin-Films for Photovoltaic Applications,
Massachusetts Institute of Technology, 2013.
9 L.C.
10
Olsen, F.W. Addis, and W. Miller, Sol. Cells 7, 247 (1983).
A. Mittiga, E. Salza, F. Sarto, M. Tucci, and R. Vasanthi, Appl. Phys. Lett. 88, 2014 (2006).
" Z. Duan, A. Du Pasquier, Y. Lu, Y. Xu, and E. Garfunkel, Sol. Energy Mater. Sol. Cells 96,
292 (2012).
12
13
14
T. Minami, Y. Nishi, T. Miyata, and
Y.S. Lee, J. Heo, S.C. Siah, J.P.
Buonassisi, Energy Environ. Sci. 6,
S.W. Lee, Y.S. Lee, J. Heo, S.C.
S. Abe, M, 1 (2012).
Mailoa, R.E. Brandt, S.B. Kim, R.G. Gordon, and T.
2112 (2013).
Siah, D. Chua, R.E. Brandt, S.B. Kim, J.P. Mailoa, T.
Buonassisi, and R.G. Gordon, Adv. Energy Mater. (2014).
" Y.S. Lee, D. Chua, R.E. Brandt, S.C. Siah, J. V Li, J.P. Mailoa, S.W. Lee, R.G. Gordon, and
T. Buonassisi, Adv. Mater. 26, 4704 (2014).
16 T. Minami, Y. Nishi, and T. Miyata, Appl. Phys. Express 6, 044101 (2013).
17
T. Surek, in 3rd World Conf Photovolt. Energy Convers. (2003).
18
N.S. Lewis, Science 315, 798 (2007).
19 D.M. Powell, M.T. Winkler, H.J. Choi, C.B. Simmons, D.B. Needleman, and T. Buonassisi,
Energy Environ. Sci. 5, 5874 (2012).
20
M. Woodhouse, R. Margolis, T.L. James, D. Feldman, and A. Goodrich, (2012).
21
M. Woodhouse, A. Goodrich, R. Margolis, T. James, R. Dhere, T. Gessert, T. Barnes, R.
22
Eggert, and D. Albin, Sol. Energy Mater. Sol. Cells 115, 199 (2013).
K. Zweibel, Sol. Energy Mater. Sol. Cells 59, 1 (1999).
System Advisor Model (SAM), NREL (2015).
23
105
24
25
26
S.N. Habisreutinger, T. Leijtens, G.E. Eperon, S.D. Stranks, R.J. Nicholas, and H.J. Snaith,
Nano Lett. (2014).
T. Leijtens, G.E. Eperon, S. Pathak, A. Abate, M.M. Lee, and H.J. Snaith,
Nat. Commun. 4,
2885 (2013).
C.D. Bailie, M.G. Christoforo, J.P. Mailoa, A.R. Bowring, E.L. Unger, W.H. Nguyen, J.
Burschka, N. Pellet, J.Z. Lee, M. Gritzel, R. Noufi, T. Buonassisi, A. Salleo, and M.D.
27
28
McGehee, Energy Environ. Sci. 956 (2014).
J.P. Mailoa, C.D. Bailie, E.C. Johlin, E.T. Hoke, A.J. Akey, W.H. Nguyen, M.D.
McGehee,
and T. Buonassisi, Appl. Phys. Lett. 106, 121105 (2015).
S.W. Glunz, Adv. Optoelectron. 2007, 1 (2007).
29
D.K. Schroder, IEEE Trans. Electron Devices 31, 637 (1984).
30
R.E. Brandt, Master Thesis (2013).
31
J. Bardeen, Phys. Rev. 71, 717 (1947).
G.K. Reeves, Solid. State. Electron. 23, (1979).
32
3
H.S. Yang, Y. Li, D.P. Norton, K. Ip, S.J. Pearton, S. Jang, and F. Ren, Appl. Phys. Lett. 86,
192103 (2005).
34 S.C. Siah, Y.S. Lee, Y. Segal, and T. Buonassisi, J. Appl. Phys. 112, 084508 (2012).
3
W.M. Haynes and D.R. Lide, CRC Handbook of Chemistry and Physics, 91st Editi (2011).
36
j. V. Li, S. Grover, M. a. Contreras, K. Ramanathan, D. Kuciauskas, and R. Noufi, Sol. Energy
Mater. Sol. Cells 124, 143 (2014).
37
R.E. Brandt, M. Young, H.H. Park, A. Dameron, D. Chua, Y.S. Lee, G. Teeter, R.G. Gordon,
38
and T. Buonassisi, Appl. Phys. Lett. 263901, 1 (2014).
S.C. Siah, S.W. Lee, Y.S. Lee, J. Heo, T. Shibata, C.U. Segre, R.G. Gordon, and T.
Buonassisi, Appl. Phys. Lett. 104, 242113 (2014).
39
S.C. Siah, R.E. Brandt, L.T. Schelhas, K. Lim, J.D. Perkins, R. Jaramillo, M.D. Heinemann,
D. Chua, J. Wright, C.U. Segre, R.G. Gordon, M. Toney, and T. Buonassisi, Prep. Submiss.
(n.d.).
40
D.C. Koningsberger and R. Prins, X-Ray Absorption: Principles, Applications, Techniques of
EXAFS, SEXAFS, and XANES
Netherlands, 1988).
41
42
(Blackwell
Scientific
Publications,
Eindhoven,
B. Ravel and M. Newville, J. Synchrotron Radiat. 12, 537 (2005).
W. Shockley and H.J. Queisser, J. Appl. Phys. 32, 510 (1961).
43
S. Siebentritt, Sol. Energy Mater. Sol. Cells 95, 1471 (2011).
44
W. Shockley and W.T. Read, Physcial Rev. Lett. 87, 835 (1952).
45 H. Raebiger, S. Lany, and A. Zunger, Phys. Rev. B 76, 1 (2007).
46
J.D. Schick and D. Trivich, J. Electrochem. Soc. Solid-State Sci. Technol. 119, 376 (1967).
106
The
47
F. Biccari, Ph.D Thesis (2009).
48
S. Rein, Lifetime Spectroscopy: A Method of Defect Characterization in Silicon for
PhotovoltaicApplications (Springer, Berlin, 2006).
49
F. Yan, S. Johnston, K. Zaunbrecher, M. AI-Jassim, 0. Sidelkheir, and K. Ounadjela, Phys.
Status Solidi - Rapid Res. Lett. 6, 190 (2012).
50
T. Trupke, R. a. Bardos, M.C. Schubert, and W. Warta, Appl. Phys. Lett. 89, 044107 (2006).
J. V. Li, S.W. Johnston, X. Li, D.S. Albin, T. a. Gessert, and D.H. Levi, J. Appl. Phys. 108,
064501 (2010).
52
S. Ishizuka, K. Suzuki, Y. Okamoto, M. Yanagita, T. Sakurai, and K. Akimoto, Phys. Status
Solidi 1070, 1067 (2004).
5
T. Minami, Y. Nishi, T. Miyata, and J. Nomoto, Appl. Phys. Express 4, 062301 (2011).
14
E.. Turkdogan, Physical Chemistry of High Temperature Technology (Academic Press, 1980).
5
56
J.-S. Jang and T.-Y. Seong, J. Appl. Phys. 88, 3064 (2000).
K. a. Rickert, a. B. Ellis, J.K. Kim, J.-L. Lee, F.J. Himpsel, F. Dwikusuma, and T.F. Kuech, J.
Appl. Phys. 92, 6671 (2002).
57
S. Tanuma, C.J. Powell, and D.R. Penn, Surf. Interface Anal. 17, 927 (1991).
58
M. Brun, A. Berthet, and J.C. Bertolini, J. Electron Spectros. Relat. Phenomena 104, 55
(1999).
J. Pan, U. Diebold, L. Zhang, and T. Madey, Surf. Sci. 295, 411 (1993).
60 S.I. Shizuka, S.K. Ato, T.M. Aruyama, and K.A. Kimoto, Appl. Phys. 40,
2765 (2001).
61 Y.S. Lee, J. Heo, M.T. Winkler, S.C. Siah, S.B. Kim, R.G. Gordon, and T.
Buonassisi, J.
Mater. Chem. A 1, 15416 (2013).
62 A.Y.C. Yu, Solid. State. Electron. 13, 239
(1970).
59
63
K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, and H. Hosono, Nature 432, 488
64
T. Kamiya, K. Nomura, and H. Hosono, Sci. Technol. Adv. Mater. 11, 044305 (2010).
65
J.K. Jeong, Semicond. Sci. Technol. 26, 034008 (2011).
66
M. Orita, H. Ohta, M. Hirano, S. Narushima, and H. Hosono, Philos. Mag. Part B 81, 501
(2004).
67
68
69
70
71
(2001).
J. Robertson, Phys. Status Solidi 245, 1026 (2008).
H. Hosono, J. Non. Cryst. Solids 352, 851 (2006).
J. Heo, S. Bok Kim, and R.G. Gordon, Appl. Phys. Lett. 101, 113507 (2012).
P.T. Erslev, E.S. Sundholm, R.E. Presley, D. Hong, J.F. Wager, and J.D. Cohen, Appl. Phys.
Lett. 95, 192115 (2009).
C.-G. Lee, B. Cobb, and A. Dodabalapur, Appl. Phys. Lett. 97, 203505 (2010).
107
72
T. Moriga, Y. Hayashi, K. Kondo, Y. Nishimura, K. Murai, I. Nakabayashi, H. Fukumoto, and
K. Tominaga, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 22, 1705 (2004).
73
H.Q. Chiang, J.F. Wager, R.L. Hoffman, J. Jeong, and D.A. Keszler, Appl. Phys. Lett. 86,
013503 (2005).
74
D. Hong, H.Q. Chiang, and J.F. Wager, J. Vac. Sci. Technol. B Microelectron. Nanom. Struct.
24, L23 (2006).
75
A. Hultqvist, C. Platzer-bj6rkman, U. Zimmermann, M. Edoff, and T. T6rndahl, Prog.
Photovoltaics Res. Appl. 20, 883 (2012).
76
P. G6rrn, M. Sander, J. Meyer, M. Kr6ger, E. Becker, H.-H. Johannes, W. Kowalsky, and
T.
Riedl, Adv. Mater. 18, 738 (2006).
77
A. Hultqvist, M. Edoff, and T. Torndahl, Prog. Photovoltaics Res. Appl. 19, 478 (2011).
78
E. Johlin, N. Tabet, S. Castro-Galnares, A. Abdallah, M.I. Bertoni, T. Asafa, J.C. Grossman,
S.
Said, and T. Buonassisi, Phys. Rev. B 85, 075202 (2012).
79
C.J. Glover, G.J. Foran, and M.C. Ridgway, Nucl. Instruments Methods Phys. Res. Sect. B
Beam Interact. with Mater. Atoms 199, 195 (2003).
80
S.B. Kim, P. Sinsermsuksakul, A.S. Hock, R.D. Pike, and R.G. Gordon, Chem. Mater. in press
81
B.K. Newman, E. Ertekin, J.T. Sullivan, M.T. Winkler, M. a. Marcus,
S.C. Fakra, M.-J. Sher,
(2014).
82
83
84
85
86
87
E. Mazur, J.C. Grossman, and T. Buonassisi, J. Appl. Phys. 114, 133507 (2013).
J. Heo, A.S. Hock, and R.G. Gordon, Chem. Mater. 22, 4964 (2010).
H. Sawada, R. Wang, and A.W. Sleight, J. Solid State Chem. 122, 148 (1996).
J. Robertson, J. Phys. C Soild State Phys. 12, 4767 (1979).
M. Newville, J. Synchrotron Radiat. 8, 322 (2001).
W. K6rner, P. Gumbsch, and C. Elsdsser, Phys. Rev. B 86, 165210 (2012).
J. Lee, Y. Kang, S. Han, C. Seong Hwang, and J.-H. Choi, Appl. Phys. Lett. 103, 252102
(2013).
88
D.-Y. Cho, J.H. Kim, and C.S. Hwang, Appl. Phys. Lett. 98, 222108 (2011).
89
D.-Y. Cho, J.H. Kim, K.D. Na, J. Song, C.S. Hwang, B.-G. Park, J.-Y. Kim, C.-H. Min, and
S.-J. Oh, Appl. Phys. Lett. 95, 261903 (2009).
90
J. Lee, D.-Y. Cho, J. Jung, U. Ki Kim, S. Ho Rha, C. Seong Hwang, and J.-H. Choi, Appl.
Phys. Lett. 102, 242111 (2013).
9'
D.-H. Cho, S. Yang, C. Byun, J. Shin, M.K. Ryu, S.-H.K. Park, C.-S. Hwang, S.M. Chung,
W.-S. Cheong, S.M. Yoon, and H.-Y. Chu, Appl. Phys. Lett. 93, 142111 (2008).
92
M. Higashiwaki, K. Sasaki, A. Kuramata, T. Masui, and S. Yamakoshi, Phys. Status Solidi
Appl. Mater. Sci. 211, 21 (2014).
93
T.G. Allen and A. Cuevas, Appl. Phys. Lett. 105, 031601 (2014).
94 M. Fleischer and H. Meixner, Sensors Actuators B Chem. 4, 437 (1991).
108
E. Nogales, J. a Garcia, B. Mendez, J. Piqueras, K. Lorenz, and E. Alves, J. Phys. D. Appl.
95
Phys. 41, 065406 (2008).
96
E.G. Villora, S. Arjoca, K. Shimamura, D. Inomata, and K. Aoki, in SPIE 8987, Oxide-Based
Mater. Devices V, 89871U (March 8, 2014) (2014).
97
98
K. Shimamura, E.G. Villora, K. Domen, K. Yui, K. Aoki, and N. Ichinose, Jpn. J. Appl. Phys.
44, L7 (2004).
E.G. Villora, K. Shimamura, K. Kitamura, K. Aoki, and T. Ujiie, Appl. Phys. Lett. 90, 23
(2007).
99
100
E.G. Villora, K. Shimamura, Y. Yoshikawa, T. Ujiie, and K. Aoki, Appl. Phys. Lett. 92, 0
(2008).
S.C. Siah, R. Jaramillo, R. Chakraborty, P.T. Erslev, C. Sun, T. Weng, M.F. Toney, G.
Teeter, and T. Buonassisi, IEEE J. Photovoltaics 1 (2014).
101
102
103
104
105
106
L. Qiao, H.Y. Xiao, S.M. Heald, M.E. Bowden, T. Varga, G.J. Exarhos, M.D. Biegalski, I.N.
Ivanov, W.J. Weber, T.C. Droubay, and S.A. Chambers, J. Mater. Chem. C 1, 4527 (2013).
P. Villars, MaterialPhases Data System (MPDS) (CH-6354 Vitznau, Switzerland, 2014).
S. Grindy, B. Meredig, S. Kirklin, J.E. Saal, and C. Wolverton, Phys. Rev. B 87, 1 (2013).
T.C. Lovejoy, R. Chen, E.N. Yitamben, V. Shutthanadan, S.M. Heald, E.G. Villora, K.
Shimamura, S. Zheng, S.T. Dunham, F.S. Ohuchi, and M. a. Olmstead, J. Appl. Phys. 111,
123716 (2012).
J.B. Varley, J.R. Weber, A. Janotti, and C.G. Van de Walle, Appl. Phys. Lett. 97, 142106
(2010).
M.B. Maccioni, F. Ricci, and V. Fiorentini, Appl. Phys. Express 021102 (2015).
107
T.H. Yeom, I.G. Kim, S.H. Lee, S.H. Choh, and Y.M. Yu, J. Appl. Phys. 93, 3315 (2003).
108
I.G. Kim, T.H. Yeom, S.H. Lee, Y.M. Yu, H.W. Shin, and S.H. Choh, J. Appl. Phys. 89, 4470
(2001).
109
A.L. Shluger, K.P. McKenna, P. V. Sushko, D.M. Ramo, and A. V. Kimmel, Model. Simul.
Mater. Sci. Eng. 17, 084004 (2009).
110
K.P. McKenna and A.L. Shluger, Proc. R. Soc. London A Math. Phys. Eng. Sci. 467, 2043
(2011).
"
S. Johnston, K. Zaunbrecher, R. Ahrenkiel, D. Kuciauskas, D. Albin, and W. Metzger, IEEE
J. Photovoltaics 4, 1295 (2014).
109