University of Southern California Dissertations and Theses

Multilayer grown ultrathin nanostructured GaAs solar cells towards high-efficiency, cost-competitive III-V photovoltaics

(USC Thesis Other)

Multilayer grown ultrathin nanostructured GaAs solar cells towards high-efficiency, cost-competitive III-V photovoltaics

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2019 Boju Gai
Multilayer Grown Ultrathin Nanostructured GaAs Solar Cells
Towards High-efficiency, Cost-competitive III-V Photovoltaics

by

Boju Gai

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(MATERIALS SCIENCE)
May 2019

i

Table of Contents
ACKNOWLEDGEMENT ............................................................................................................ v
ABSTRACT ................................................................................................................................. vii
Chapter 1. Introduction ................................................................................................................ 1
1.1 GaAs for solar cell applications ........................................................................................ 1
1.2 Cost consideration of GaAs solar cells ............................................................................. 3
1.3 Epitaxial lift off (ELO) for GaAs solar cells .................................................................... 4
1.4 Transfer printing................................................................................................................ 8
1.5 Multi-layer grown GaAs solar cells .................................................................................11
1.6 Nanoscale photon management for ultrathin GaAs solar cells ....................................... 15
1.7 Inverted metamorphic multijunction (IMM) solar cells ................................................. 17
1.8 References ....................................................................................................................... 22
Chapter 2. Methods .................................................................................................................... 28
2.1. Materials synthesis ......................................................................................................... 28
2.1.1. Epitaxial growth of triple-stack GaAs solar cells by molecular beam epitaxy
(MBE) ........................................................................................................................... 28
2.1.2. Epitaxial growth of ten-fold-stack GaAs solar cells by molecular beam epitaxy
(MBE) ........................................................................................................................... 29

ii

2.1.3. Epitaxial growth of triple-junction inverted metamorphic (3J IMM)
GaInP/GaAs/InGaAs solar cells by organometallic vapor phase epitaxy (OMVPE) ... 31
2.2. Device fabrication .......................................................................................................... 33
2.2.1. Fabrication of GaAs solar cells with a vertical contact configuration ................ 33
2.2.2. Fabrication of GaAs solar cells with TiO2 nanoposts (NPs) ............................... 38
2.2.3. Fabrication of 3J IMM microcells ...................................................................... 39
2.3. Materials and device characterizations .......................................................................... 43
2.4. Numerical Optical Modeling ......................................................................................... 45
2.5 References ....................................................................................................................... 46
Chapter 3. Multilayer-grown ultrathin nanostructured GaAs solar cells ............................. 48
3.1 Introduction ..................................................................................................................... 48
3.2 Results and discussion .................................................................................................... 51
3.3 Conclusion ...................................................................................................................... 68
3.4 References ....................................................................................................................... 70
Chapter 4. Ten-fold stack multilayer-grown nanomembrane GaAs solar cells .................... 75
4.1 Introduction ..................................................................................................................... 75
4.2 Results and discussion .................................................................................................... 77
4.3 Conclusion ...................................................................................................................... 89
4.4 References ....................................................................................................................... 91

iii

Chapter 5. Transfer printed 3J inverted metamorphic multijunction (IMM) solar cells .... 95
5.1 Introduction ..................................................................................................................... 95
5.2 Results and discussion .................................................................................................... 97
5.3 Conclusion .................................................................................................................... 107
5.4 References ..................................................................................................................... 108
Chapter 6. Future work ............................................................................................................. 115
6.1 Fully-depleted ultrathin GaAs solar cells ......................................................................115
6.2 References ..................................................................................................................... 120
Appendices ................................................................................................................................. 121
A.1 Source code for Chapter 4.2 ......................................................................................... 121
A.1.1 Lumerical script to generate multilayer structure ............................................. 121
A.1.2 Lumerical script to calculate for each stacks .................................................... 123
A.1.3 MATLAB script to summarize data .................................................................. 124
A.2 Source code for Chapter 6.1 ......................................................................................... 126
A.2.1 Lumerical script for simulation I-V for different doping concentration ........... 126
A.2.2 MATLAB script for summarize I-V data .......................................................... 128
A.2.3 Lumerical script for export generation for each wavelength from FDTD for QE
simulation .................................................................................................................... 129
A.2.4 MATLAB script for summarize the generation for each wavelength ............... 132

iv

A.2.5 Lumerical script for import generation for each wavelength to DEVICE ........ 133
A.2.6 Lumerical script for simulation QE for different doping concentration ........... 134
A.2.7 MATLAB script for summarize QE data .......................................................... 136

v

ACKNOWLEDGEMENT
First, I would like to show my sincere gratefulness to my advisor, Professor Jongseung
Yoon. He accepted me to do research work when I had little experience and knowledge for this
field, and cultivated and educated me the thought, skill and knowledge as an academic research.
Without his high and strict standard, it is not possible to finish my PhD.
Secondly, I would like to thank the help and support from my college group members: Dr
Dongseok Kang, Dr Sung-min Lee, Weigu Li, Lang Shen, Lesley Chan, Hajira Hunter, Abinasha
Kalita, Anthony Kwong, Hyungwoo Choi, Yuzhou Xie, Haneol Lim, Huandong Chen, Lauren
Liaw, Qian Yi, Phillip Weiner, Emily Palmer, Fan Sang, Chiye Huang, Moonchul Jung, Tingyun
Bi, Wanting He, Kyunghwan Min, Yuntao Lu, Theodore Somekh, Sangbom Kim, Qiang Fu,
Mike Shao, Goeon Lee. There company, support and unconditional support are invaluable for my
PhD career.
Thirdly I would like to thank our collaborating group: Prof Minjoo Lee, Yukun Sun, Dr
Joseph Faucher from Yale University and University of Illinois at Urbana-Champaign, Dr John
Geisz, Dr Daniel Friedman, Dr James Young, Dr Walter Klein and Dr Todd Deutsch from
National Renewable Energy Laboratory, Dr Bryant Thompson, Prof Noah Malmstadt from
University of Southern California, and Dr Purnim Dhar, Angelo Montenegro, Dr Chayan Dutta
and Prof Alexander Benderskii from University of Southern California, for their effort, insight
and providing materials.

vi

I also want to express my gratitude to my dissertation committee: Prof Michael Kassner
and Prof Wei Wu, for the suggestions and positive comments. I also thank Donghai Zhu and
Alfonso Jimenez for their effort to maintain the equipment and environment for Keck Photonics
Laboratory, Prof Willie Ng for the SEM, and John Curulli for helping with the equipment in
CEMMA.
Last but not least, I need to thank my parent’s supporting and understanding during these
years when I was pursuing my PhD.

vii

ABSTRACT
Large-scale deployment of GaAs solar cells in terrestrial photovoltaics demands
significant cost reduction for preparing device-quality epitaxial materials. Although multilayer
epitaxial growth has been proposed as a promising route to achieve this goal, their practical
implementation remains challenging owing to the degradation of materials properties and
resulting nonuniform device performance between solar cells grown in different sequences. In
this thesis, we studied alternative approach to circumvent these limitations and enable
multilayer-grown GaAs solar cells with uniform photovoltaic performance. Ultrathin
single-junction GaAs solar cells having a 300-nm-thick absorber (i.e., emitter and base) are
epitaxially grown in triple- (Chapter 3) and tenfold-stack (Chapter 4) releasable multilayer
assemblies by molecular beam epitaxy using beryllium as a p-type impurity. Microscale GaAs
solar cells fabricated from respective device layers exhibit excellent uniformity of photovoltaic
performance and contact properties owing to the suppressed diffusion of p-type dopant as well as
substantially reduced time of epitaxial growth associated with ultrathin device configuration.
Bifacial photon management employing hexagonally periodic TiO 2 nanoposts and a vertical
p-type metal contact serving as a metallic back-surface reflector together with specialized
epitaxial design to minimize parasitic optical losses for efficient light trapping synergistically
enable significantly enhanced photovoltaic performance of such ultrathin absorbers, where ∼17.2%
solar-to-electric power conversion efficiency under simulated AM1.5G illumination is

viii

demonstrated from 420-nm-thick single-junction GaAs solar cells.
Inverted metamorphic (IMM) multijunction solar cells represent a promising materials
platform for ultrahigh efficiency photovoltaic systems (UHPVs) with a clear pathway to beyond
50% efficiency. The conventional device processing of IMM solar cells, however, typically
involves wafer bonding of a centimeter-scale die and destructive substrate removal, thereby
imposing severe restrictions in achievable cell size, type of module substrate, spatial layout, as
well as cost-effectiveness. In this thesis (Chapter 5), we studied materials design and fabrication
strategies for microscale triple-junction inverted metamorphic (3J IMM) Ga 0.51In0.49P
/GaAs/In0.26Ga0.74As solar cells that can overcome these difficulties. Specialized schemes of
delineation and undercut etching enable the defect-free release of microscale IMM solar cells
and printed assemblies on a glass substrate in a manner that preserves the growth substrate,
where efficiencies of 27.3% and 33.9% are demonstrated at simulated AM1.5D one- and 351 sun
illumination, respectively. A composite carrier substrate where released IMM microcells are
formed in fully-functional, print-ready configurations allows high-throughput transfer printing of
individual IMM microcells in a programmable spatial layout on versatile choices of module
substrate, all desired for CPV applications.

1

Chapter 1. Introduction
1.1 GaAs for solar cell applications
Solar energy represents a promising source of renewable energy in the future. The widely
used fossil fuels are considered eventually depleting, while the global energy consumption is
expected to be doubled within the next half century
1
. To this end, human beings will face the
energy crisis unless a renewable source of energy is widely used. On the other hand, utilizing
fossil-fuel-based energy also has many environmental concerns, such as air pollution, oil spills in
the ocean, and greenhouse gases that affect the global climate. The energy earth receives from
the sun is as large as 3×10
24
Joule per year
1
, meaning that if we have 0.1% coverage of solar cell
on earth with 10% efficiency, we could meet the requirement of current energy demand need for
the entire world
1
. Furthermore, the usage of solar cell is environmentally clean, without
combustion or any other reactions which may cause environmental problems
1
.
III-V materials are the compound semiconductors composed of group III and group V
elements in the periodic table. Many of III-V materials are excellent semiconductors for solar
energy harvesting due to many favorable attributes such as high electron mobility, a wide range
of direct bandgap energies by changing the material composition, and possibility to
monolithically form multi-junction tandem devices
2, 3
. Among various III-V materials, gallium
arsenide (GaAs) is one of the most promising materials for photovoltaic energy harvesting. GaAs
has nearly ideal bandgap (~1.42 eV) for the absorption of solar spectrum
4
, a direct bandgap
5
,

2

high electron mobility (~8500 cm
2
V
-1
s
-1
)
6
, and. Based on such compelling materials properties,
GaAs currently has the record-high solar-to-electric conversion efficiency of 28.8%
7
, which is
more efficient than crystalline silicon solar cells (25.6%)
7
.
GaAs has the zinc blende crystal structure (i.e. cubic ZnS crystal) as depicted in Fig 1.1,
which is a face-centered-cubic (FCC) Bravais lattice with a basis consisting of gallium (or
arsenic) occupying 0 0 0 site and arsenic (or gallium) occupying 1/4 1/4 1/4 site. The lattice
constant (i.e. “a” in Figure 1.1) of GaAs is 0.565 nm
8
. Many other III-V materials have the zinc
blende crystal structures with various lattice constants (Table 1.1, also include some group IV
semiconductors that have diamond cubic crystal structure, which is similar to zinc blende crystal
structure except that a basis is composed of the same type of element.

Figure 1.1: Structure of GaAs crystal, note that if Ga and As sites are exchanged, the structure will still be
equivalent
8
.

3

Material Lattice constant (nm) Bandgap (eV)
e/h mobility
(cm
2
V
-1
s
-1
)
GaAs 0.565 1.42 8500/400
AlAs 0.566 2.17 185/160
Al 0.52In 0.48P 0.565 2.35 100/10
Ga 0.51In 0.49P 0.565 1.85 500/30
Ge 0.566 0.66 3900/1900
InP 0.587 1.34 5400/200
Si 0.543 1.12 1400/450
GaSb 0.610 0.73 3000/1000
Table 1.1: Lattice constant of several III-V and IV semiconductors with zinc blende or diamond sturcture
6, 9-12
.

1.2 Cost consideration of GaAs solar cells
Despite many compelling advantage of GaAs for solar energy harvesting, one of the
major challenges to hinder its widespread deployment is the cost of preparing device-quality
single-crystalline materials. Considering the elemental abundance in the earth crust, the mass
fraction of gallium is 19 ppm (35
th
-most-abundant among all the elements) and arsenic is only
2.1 ppm (48
th
-most-abundant). Compared to silicon that is most widely used in PV, whose
abundance is 270,000 ppm (2
nd
-most-abundant)
13
, GaAs is inherently expensive even without
considering the cost of processes for epitaxially growing single-crystalline materials.
The cost of epi-ready GaAs wafer is expensive. From a recent cost analysis, the cost of an

4

epi-ready “solar grade” GaAs wafer of 6-inch diameter is $150, the overall wafer cost only is
already $11,300/m
2
, which will be not practical for terrestrial applications
14
. Assuming a GaAs
solar cell with 28% the efficiency
7
, the cost of the substrate itself is already ~$40/W
(=$11,300/m
2
/ 280W/m
2
). In contrast, the current cost of crystalline silicon solar cell is
~$0.75/W
15
, which is the number that GaAs solar cells cannot compete with. This is the reason
that GaAs solar cells and other III-V solar cells have been mainly used for space applications,
where the cost is not a primary consideration, while other factors such as efficiency, light weight,
and radiation hardness are more important to consider.
The high cost of GaAs solar cells are also associated with the epitaxial growth process
such as molecular beam epitaxy (MBE) and metal organic vapor phase epitaxy (MOVPE).
Source materials have to be in a very high purity, and the equipment for the epitaxial growth is
also expensive. Nevertheless, the wafer cost still accounts for the largest portion
14
. In this regards,
if we want to make GaAs solar cells affordable for terrestrial purposes, approaches that can
reduce the cost of growth wafer would be the first priority and most effective.

1.3 Epitaxial lift off (ELO) for GaAs solar cells
Epitaxial growth means the deposition of a crystalline overlayer on a crystalline
substrate
16, 17
. As introduced in section 1.1, GaAs and many other III-V materials have the
zincblende crystal structure. If we tune the composition carefully while maintaining the same

5

lattice constant, then we can grow a single crystal film on a single crystal GaAs substrate with
different chemical compositions. On the other hand, for the materials composed with different
compositions, the etching behavior could be quite different, i.e. etching selectivity, which means
that both materials could be exposed to a certain etchant at the same time but only the designated
materials can be selectively etched, or “sacrificed” while GaAs remains almost intact (or vice
versa in some other etchants). Lattice constant and bandgap of several common III-V materials
are in Figure 1.2
9
. Some potential sacrificial layers which is lattice-matched with GaAs (include
GaAs itself) are listed in Table 1.2. The etching rate dependence of Al xGa1-xAs in diluted HF
solution is in Figure 1.3
18
.
Materials Lattice constant (nm)
Etching selectivity to GaAs in
terms of etch rate
GaAs 0.565 -
AlAs 0.566 ~10
6
(in HF)
19, 20

Al 0.52In 0.48P 0.565 ∞ (ΔG>0 in HCl)
21, 22

Ga 0.51In 0.49P 0.565 ∞ (ΔG>0 in HCl)
21, 22

Note: “ΔG>0 in HCl” means the Gibbs free energy (ΔG) of reaction of GaAs + 3HCl -> GaCl 3 + AsH 3 is positive,
so it is thermodynamically unfavorable.
Table 1.2: Sacrificial layers for GaAs ELO.

6

Figure 1.2: Lattice constant and bandgap of several common III-V materials
9
.

Figure 1.3: Etching rate dependence for Al xGa 1-xAs in diluted HF etchant (Reproduced)
18
.

Epitaxial lift off (ELO) has been proposed as a method to reuse the wafer multiple times
and thus reduce the wafer cost. The ELO process is described in the following steps: First, we

7

should epitaxially grow the sacrificial layer onto the cleaned single crystalline GaAs wafer.
Second, the desired device structure should be further grown onto the as grown sacrificial layer.
Subsequently, the as grown wafer is soaked into the etchant that selectively etches the sacrificial
layer while not affecting the GaAs, and wait until the entire sacrificial layer was removed and
device film released. Finally, retrieve the device layer onto a non-native handle substrate
(typically much cheaper than GaAs substrate), and freed GaAs substrate can be further used for
the next epitaxial growth
16, 17
. In Figure 1.4, there is a brief schematic showing the process of
ELO.

Figure 1.4: Schematic for ELO
22
.

By applying the ELO method, we can reuse the GaAs growth wafer multiple times to
reduce the manufacturing cost of solar cells. On the other hand, making GaAs solar cells thin is
also advantageous for their performance
7
. Since GaAs is a direct bandgap semiconductor, the
light is absorbed efficiently, so the active light absorption part is only first several micrometer
thick and remaining of the wafer is “dead” part. Removing the “dead” part enables photon

8

recycling to further enhance the efficiency
23
. Besides, thin film solar cells are also good for
making the device thin, light weight and flexible, which can also broaden their applications.

1.4 Transfer printing
Transfer printing is a technique that allows deterministic materials assemblies for
heterogeneous integration of various materials on different types of substrates by an elastomeric
stamp
24
. While the wafer-scale thin film is peeled off during the ELO procedure, the technique of
transfer printing can be utilized to release individual devices in micro- and nanometer length
scales.
The epitaxially-grown wafers require several fabrication steps to be ready for transfer
printing. First, all required fabrication steps for functional devices including etching and
metallization are completed on the source wafer. The device stack is typically isolated into
smaller blocks (cells) (from nanometer to millimeter scale). Subsequently, the sacrificial layer
(e.g. AlAs) needs to be selectively etched by a wet chemical etchant (e.g. HF) to isolate a device
stack from the growth wafer while specially designed polymeric “anchors” make the cells, or the
“ink”, tethered to the source wafer. Therefore, the cells could maintain lithographically defined
layouts and are ready for transfer printing.
To conduct the transfer printing, an elastomeric stamp with a flat and clean surface made
of thermally cured poly(dimethylsiloxane) (PDMS), is employed. The low modulus and low

9

surface energy nature of the elastomeric stamp enables a conformal and soft contact to the
surface of the “ink” when it is laminated. The stamp is then peeled off, while the anchor part is
fractured in a controlled manner. The process of transfer printing was completed by placing the
ink element onto the target substrate and peeling off the stamp. Figure 1.5 shows a schematic
illustration of these processes
24, 25
.

Figure 1.5: Schematic of transfer printing (reproduced)
24, 25
.

The pick-up and printing processes can be facilitated by exploiting kinetic dependence of
adhesion strength of viscoelastic materials, as described subsequently. We define the energy
release rate (G = F / w, where F is peeling force and w is the width of the contact region), as
shown in the schematic in Figure 1.6. The critical energy release rate 𝐺 is defined as the

10

smallest G for steady-state crack propagation, i.e. if G > 𝐺 , the new interface of crack
propagation is formed between two layers (either between the stamp and the “ink” or between
the “ink” and the substrate), or “crack” will spontaneously propagate and finally result in the
detach of these two layers. This indicates that the greater 𝐺 is, the more difficult for forming
the crack, and stronger the adhesion. The high yield of the transfer printing is therefore
achievable by carefully controlling 𝐺 of the interfaces. For viscoelastic materials, this can be
is achieved by taking advantage of the dependence of 𝐺 upon the velocity (v) of peel-off
between the viscoelastic stamp and the elastic ink due to viscoelasticity nature of the stamp,
which could be described as the equation: 𝐺 (𝑣 ) = 𝐺 [1 + ]. For successful picking-up
(i.e. retrieving “ink” from source substrate), the critical energy release rate between the stamp
and the “ink” 𝐺 / , should be greater than the one between the substrate and the ink
𝐺 / , so that the crack is more preferentially formed between the “ink” and the
substrate. According to the equation, a greater v could increase 𝐺 / to get a high
picking-up yield. On the other hand, for printing (i.e. delivering “ink” to target substrate), the
velocity (v) can be reduced to get smaller 𝐺 / to make it smaller than 𝐺 / ,
similarly. With such control of v, the transfer printing process can be optimized for high yield
pick-up and printing. Besides, the red line and green line in Figure 1.6 indicate when the 𝐺
between ink and substrate is too strong (e.g. adhesive is applied and cured) or too weak, no
matter how large v is, it will be always “printing only” or “picking-up only”, respectively.

11

Figure 1.6: Schematic of critical energy release rate (G crit) and peeling off rate (v) relationship (reproduced)
26
.

1.5 Multi-layer grown GaAs solar cells
In section of 1.2 and 1.3 we discussed about the high cost of GaAs wafer and ELO
method to reduce the cost, and ideally if we could reuse the GaAs substrate for many times, the
cost of the wafer will drop to correspondingly the inverse of the number of times of reuse, or, in
the extreme case, zero when it is reused infinitively. However, the practical application of ELO
to actually reduce the cost is challenging.
First, based on the fact that the cost of wafer is $40/W when it is used for only once, to
reduce the wafer cost which could be competing silicon solar cells requires hundreds times of
reusing
14
. The thermal cycle because of repeating epitaxial growth (elevated temperature) and
ELO procedure (around room temperature) could eventually degrade the wafer quality, by
introducing defects such as dislocations and slips, which would be inherited by later epitaxial
grown device layer and thus degrade the performance, considering the large number
14
of reuse

12

times required even the degradation of device performance could be optimized to minimal within
several or even several tens of times. Second, to restore the wafer to epitaxial ready status, the
cleaning and polishing steps are not trivial in terms of both engineering and economic.
Chemical-mechanical polishing (CMP) is an effective method to restore the surface condition,
however, every time after the CMP, not only the surface is cleaned, but also some GaAs material,
typically several micrometers in terms of thickness, is lost during the polishing
27
. Considering
the typical thickness of commercially available GaAs wafer (several hundred micrometers), the
times of reuse of a single wafer will be limited to just several tens of times
28
. Besides, even
everything going ideally, the cost of CMP itself is expensive. Because of the polishing, the
“effective” wafer cost will drop dramatically initially, but will eventually be a constant number
because the polishing cost is dominant at a large number of reuse, rather than just reduced to
zero
14
. This analysis is detailed described in Figure 1.7
14
.

13

Figure 1.7: Wafer cost versus times of reuse(reproduced)
14
.

However, there could be a way to mitigate all these issues in regard of wafer reusing,
which is multi-layer growth. Rather than reusing the wafer directly, we grow the device layer and
sacrificial layer alternatively many times in single epitaxial growth. Then, perform the ELO
procedure layer by layer, without the need of putting the wafer into the epitaxial chamber again.
Therefore, the wafer is used only once but considering the production of device layers, the wafer
is “reused” for many times. A schematic showing how this method is done is in Figure 1.8.

14

Figure 1.8: Conventional ELO versus multi-layer growth technique.

There are already several pioneering works done using this strategy in this field before
this dissertation. In 2010, Yoon and co-workers reported a triple-stack, n-on-p GaAs multi-layer
grown solar cells system using zinc (Zn) as the p-type dopant
29
. This work has shown the
possibility of multilayer grown GaAs solar cell enabled by transfer-printing method. However,
its application is limited by degradation of earlier grown layers due to severe diffusion of the
p-type dopant, zinc (Zn). Kang and co-workers reported a triple-stack, p-on-n system
30
later in
2013, with p-type dopant as carbon (C). Due to lower tendency of diffusion nature of C in GaAs,
the uniformity improved compare to previously reported Zn as p-type dopant case
29
, while some
systematically degradation still observed due to carbon related point defects. Choi et al in 2014
reported a multi-layer system using p-on-n configuration, Zn as the p-type dopant, and GaInP as
window layer and back surface field (BSF) layer material, in a 3-stack system
31
. This work
showed quite excellent uniformity, by applying p-on-n design and let GaInP act as both

15

passivation layer (window and BSF) as well as diffusion barrier. This work, however, the solar to
electricity efficiency less than 11%, for samples without ARC under 1 sun measurement, while
having relatively optically thick (emitter + base ~ 4 μm) absorber design, which is not satisfying.
In this dissertation, we are pursuing an alternative way to get uniform performance within
different stacks, which is to reduce the thickness ~ 10 times to several hundred nanometers, to
minimize undesired heat treatment and relieve any deterioration of performance for early grown
layers.

1.6 Nanoscale photon management for ultrathin GaAs solar cells
Although the uniformity of performance in multi-layer grown GaAs solar cells could be
improved by employing ultra-thin epitaxial design, the efficiency will be still limited by
insufficient absorption of light, especially in the range of longer wavelengths, due to the optically
thin active layers (i.e. emitter + base). In this regard, the poor absorption could be compensated
by strategies of photon management using periodic nanostructure and metallic reflector (BSR) on
the front- and rear-surface of the solar cell, respectively. Figure 1.9 schematically illustrates
one-dimensionally periodic grating formed on the top surface of solar cells. When light is
incident on the surface of nanostructured grating, diffraction will occur because of the
comparable dimension of the grating to the wavelength of light. The 1D grating equation is given
by
32, 33

16

𝑛 𝑠𝑖𝑛𝜃 = 𝑛 𝑠𝑖𝑛𝜃 =
(1-1)
, where n1 and n2 stand for the refractive index for air and the absorber, θ1 and θ2 are the
diffraction angles in the air and absorber, m is the diffraction order, d is the grating period, and λ
is the wavelength of light. According to the equation (1-1), there will be always more
propagation modes available (i.e. m  0) in the absorber that has a higher refractive index than
air, indicating most of incident light will be coupled to the absorber and the reflection to the air
will be effectively suppressed. In addition to the effect of antireflection, diffraction can
effectively increase the optical path-length of light within the optically thin absorber. Lastly,
when the diffracted light satisfies the condition of total internal reflection (TIR), the diffracted
light can be trapped within the absorber, further increasing the chance of being absorbed. All of
these optical effects are beneficial for enhancing the optical absorption and photovoltaic
performance of optically thin absorber.

17

Figure 1.9: Schematic illustration of a 1D periodic diffraction grating.

In our group, Lee et al already reported forming nanostructure on dielectric layers
(titanium dioxide, TiO2) on top of ultrathin (absorber emitter + base ~ 200 nm) GaAs solar cells
by nano-imprinting lithography (NIL) and reactive ion etching (RIE) can enhance the
performance to the efficiency of ~ 16.2%
34
. Detailed introduction about the fabrication of
nanostructure and discussion about applying this technique to multilayer grown system will be in
the following chapters.

1.7 Inverted metamorphic multijunction (IMM) solar cells
In Chapter 5, the transfer printed inverted metamorphic multijunction (IMM) solar cells
are also being discussed. Although GaAs can achieve a relative higher efficiency
7
among single
junction solar cells, there is a limitation to prevent perusing even potential higher efficiency

18

called Shockley-Queisser limit
4
, which is the limitation of the efficiency for a single junction
solar cells while assuming the ideal quantum efficiency, i.e. all the sub-bandgap photons are all
absorbed and converted into the electricity. For a certain material, a photon with wavelength
longer than the bandgap threshold will not be absorbed, and on the other hand, a photon with
very short wavelength, its energy will be wasted due to thermalization, and in result still provide
the same voltage limited by the material bandgap. A schematics showing this is in Figure 1.10.
One way to circumvent this limit is to make multijunction tandem solar cells, which is using
several materials with different bandgap working together, i.e. stacking larger bandgap material
on the top to absorb shorter wavelength, or higher energy photons, while placing smaller
bandgap material on the bottom to utilize the photons that cannot be absorbed by the upper
structure. Currently, the world record for solar cell efficiency is made by such type of solar cells
7
:
46.0% (2014), using such strategy. As discussed previously, III-V is a versatile platform for
integrate materials with different bandgap, and the conventional III-V GaInP/GaAs/Ge system
has been developed for many years and it is already widely used in aerospace applications due to
its favored efficiency and light weight advantages
35
. However, there is a problem with this
system, which is that the bandgap of germanium (Ge) is too small and in result the current
provided by this junction is too large and not well matching with top two junctions, such that the
efficiency is limited. A better choice for the bottom junction according to the calculation to work
with GaInP/GaAs top structures is a material with around 1.0 eV
36
, such as a type of material

19

called “diluted nitride”
37, 38
, since these types of materials can tune the bandgap within a range by
changing its composition and while still keep the lattice constant matching with GaAs or Ge
epitaxial substrate. However, such diluted nitride is not totally satisfying because the nitrogen
related intrinsic defects and such defects will act as the non-radiative recombination center. In
this regard, the best scenario for this purpose will be that we could have better freedom in terms
of lattice constant while still keeping the quality of the materials.

Figure 1.10: Schematics of single junction and multijunction tandem solar cells.

Metamorphic solar cells is invented following this idea, which is introducing a graded
buffer with gradually changing its lattice constant to be matched on both sides of this layer

20

connected, while with larger bandgap for photons that meant to be absorbed by lower layers can
passing through, i.e. optically transparent, and heavily doped for electricity conducting. Because
the lattice constant is only gradually changed throughout the epitaxial growth, the undesired
dislocations will be more tend to confined within the layer and the later grown absorber layer can
still kept in acceptable quality. This metamorphic scheme can be designed as upright
metamorphic, which is growing smaller bandgap material (bottom junction) first, and then this
graded buffer layer, and then larger bandgap material structure
39, 40
. This design can achieve
some higher efficiency, but with the disadvantage of the upper junctions are grown after the
grading buffer, and the defects generated during the buffer layer growth, due to its lattice
constant mismatching nature, such as dislocations, will be succeeded by upper layers, even with
careful design and controlling, and these structures are more important due to they will
contribute more on efficiency
36, 40
. Because of this, we can implement this strategy in an inverted
way, i.e. growing the upper junctions first, then the graded buffer, and finally lower junctions last,
called inverted metamorphic (IMM), to mitigate the issues of the upright design. The inverted
design requires some way to get rid of the substrate wafer, flip the whole structure over to let the
light incident the top structure first. Currently, the method to achieve this is to bond the whole
wafer on to a handling substrate (i.e. polymer, glass or any cheap, stable substrate can act as an
operating platform), selectively remove the whole wafer by wet chemical etching, and start the
fabrication. A schematics showing the process is in Figure 1.11
41
.

21

Figure 1.11: Schematics of conventional way of IMM fabrication
41
.

This way of making IMM has some limitations. First, the substrate, either GaAs or Ge, is
not cheap as we discussed in 1.2, while etching the whole wafer will eliminate the possibility of
wafer recycle completely. Second, the method of wafer bonding onto a handling substrate is
limiting the possible size that can be handled while keeping the areal efficiency, while the
module consisted of smaller cells is advantageous for concentrated photovoltaic (CPV)
applications
42
. Herein, we propose a way of using transfer-printing for IMM solar cells to show
an alternative technical route, to get more freedom of designing and fabrication. Further
discussion will be in the following chapters.

22

1.8 References
1. Gratzel, M., Photoelectrochemical cells. nature 2001, 414, 338.
2. Dimroth, F., High-efficiency solar cells from III-V compound semiconductors. physica
status solidi (c) 2006, 3, 373-379.
3. Yamaguchi, M.; Takamoto, T.; Araki, K.; Ekins-Daukes, N., Multi-junction III–V solar
cells: current status and future potential. Solar Energy 2005, 79, 78-85.
4. Shockley, W.; Queisser, H. J., Detailed balance limit of efficiency of p‐n junction solar
cells. Journal of applied physics 1961, 32, 510-519.
5. Blakemore, J., Semiconducting and other major properties of gallium arsenide. Journal of
Applied Physics 1982, 53, R123-R181.
6. http://www.ioffe.ru/SV A/NSM/Semicond/.
7. Green, M. A.; Emery, K.; Hishikawa, Y .; Warta, W.; Dunlop, E. D., Solar cell efficiency
tables (Version 45). Progress in photovoltaics: research and applications 2015, 23, 1-9.
8. Wikipedia. https://en.wikipedia.org/wiki/Gallium_arsenide/.
9. Schubert, E. F., Light-emitting diodes. E. Fred Schubert: 2018.
10. Vaisman, M.; Mukherjee, K.; Masuda, T.; Yaung, K. N.; Fitzgerald, E. A.; Lee, M. L.,
Direct-Gap 2.1–2.2 eV AlInP solar cells on GaInAs/GaAs metamorphic buffers. IEEE
Journal of Photovoltaics 2016, 6, 571-577.
11. Shitara, T.; Eberl, K., Electronic properties of InGaP grown by solid ‐ source

23

molecular‐beam epitaxy with a GaP decomposition source. Applied physics letters 1994,
65, 356-358.
12. Gudovskikh, A.; Kaluzhniy, N.; Lantratov, V .; Mintairov, S.; Shvarts, M.; Andreev, V.,
Numerical modelling of GaInP solar cells with AlInP and AlGaAs windows. Thin Solid
Films 2008, 516, 6739-6743.
13. Webelements. https://www.webelements.com/periodicity/abundance_crust/.
14. Woodhouse, M.; Goodrich, A. A manufacturing cost analysis relevant to single-and
dual-junction photovoltaic cells fabricated with III-Vs and III-Vs grown on Czochralski
silicon; National Renewable Energy Laboratory: 2013.
15. Energysage.com.
http://news.energysage.com/how-much-does-the-average-solar-panel-installation-cost-in-
the-u-s/.
16. Konagai, M.; Sugimoto, M.; Takahashi, K., High efficiency GaAs thin film solar cells by
peeled film technology. Journal of crystal growth 1978, 45, 277-280.
17. Yablonovitch, E.; Gmitter, T.; Harbison, J.; Bhat, R., Extreme selectivity in the lift‐off
of epitaxial GaAs films. Applied Physics Letters 1987, 51, 2222-2224.
18. Kumar, P.; Kanakaraju, S.; Devoe, D., Sacrificial etching of Al x Ga 1-x As for III–V
MEMS surface micromachining. Applied Physics A 2007, 88, 711-714.
19. V oncken, M.; Schermer, J.; Bauhuis, G.; Mulder, P.; Larsen, P., Multiple release layer

24

study of the intrinsic lateral etch rate of the epitaxial lift-off process. Applied Physics A:
Materials Science & Processing 2004, 79, 1801-1807.
20. Wu, X.; Coldren, L.; Merz, J., Selective etching characteristics of HF for
AlxGa1-xAs/GaAs. Electronics Letters 1985, 21, 558-559.
21. Pearton, S., Critical issues of III–V compound semiconductor processing. Materials
Science and Engineering: B 1997, 44, 1-7.
22. Cheng, C.-W.; Shiu, K.-T.; Li, N.; Han, S.-J.; Shi, L.; Sadana, D. K., Epitaxial lift-off
process for gallium arsenide substrate reuse and flexible electronics. Nature
communications 2013, 4, 1577.
23. Wang, X.; Khan, M. R.; Gray, J. L.; Alam, M. A.; Lundstrom, M. S., Design of GaAs
solar cells operating close to the Shockley–Queisser limit. IEEE Journal of Photovoltaics
2013, 3, 737-744.
24. Yoon, J.; Lee, S. M.; Kang, D.; Meitl, M. A.; Bower, C. A.; Rogers, J., Heterogeneously
integrated optoelectronic devices enabled by micro‐transfer printing. Advanced Optical
Materials 2015, 3, 1313-1335.
25. Meitl, M. A.; Zhu, Z.-T.; Kumar, V .; Lee, K. J.; Feng, X.; Huang, Y . Y.; Adesida, I.;
Nuzzo, R. G.; Rogers, J. A., Transfer printing by kinetic control of adhesion to an
elastomeric stamp. Nature materials 2006, 5, 33.
26. Feng, X.; Meitl, M. A.; Bowen, A. M.; Huang, Y.; Nuzzo, R. G.; Rogers, J. A.,

25

Competing fracture in kinetically controlled transfer printing. Langmuir 2007, 23,
12555-12560.
27. Smeenk, N.; Engel, J.; Mulder, P.; Bauhuis, G.; Bissels, G.; Schermer, J.; Vlieg, E.; Kelly,
J., Arsenic formation on GaAs during etching in HF solutions: relevance for the epitaxial
lift-off process. 2013.
28. Bauhuis, G.; Mulder, P.; Haverkamp, E.; Schermer, J.; Bongers, E.; Oomen, G.; Köstler,
W.; Strobl, G., Wafer reuse for repeated growth of III–V solar cells. Progress in
Photovoltaics: Research and Applications 2010, 18, 155-159.
29. Yoon, J.; Jo, S.; Chun, I. S.; Jung, I.; Kim, H.-S.; Meitl, M.; Menard, E.; Li, X.; Coleman,
J. J.; Paik, U., GaAs photovoltaics and optoelectronics using releasable multilayer
epitaxial assemblies. Nature 2010, 465, 329.
30. Kang, D.; Arab, S.; Cronin, S. B.; Li, X.; Rogers, J. A.; Yoon, J., Carbon-doped GaAs
single junction solar microcells grown in multilayer epitaxial assemblies. Applied Physics
Letters 2013, 102, 253902.
31. Choi, W.; Kim, C. Z.; Kim, C. S.; Heo, W.; Joo, T.; Ryu, S. Y .; Kim, H.; Kim, H.; Kang,
H. K.; Jo, S., A Repeatable Epitaxial Lift‐Off Process from a Single GaAs Substrate for
Low‐Cost and High‐Efficiency III‐V Solar Cells. Advanced Energy Materials 2014,
4, 1400589.
32. Chong, T. K.; Wilson, J.; Mokkapati, S.; Catchpole, K. R., Optimal wavelength scale

26

diffraction gratings for light trapping in solar cells. Journal of Optics 2012, 14, 024012.
33. Mokkapati, S.; Catchpole, K., Nanophotonic light trapping in solar cells. Journal of
applied physics 2012, 112, 101101.
34. Lee, S.-M.; Kwong, A.; Jung, D.; Faucher, J.; Biswas, R.; Shen, L.; Kang, D.; Lee, M. L.;
Yoon, J., High performance ultrathin GaAs solar cells enabled with heterogeneously
integrated dielectric periodic nanostructures. ACS nano 2015, 9, 10356-10365.
35. Karam, N. H.; King, R. R.; Haddad, M.; Ermer, J. H.; Yoon, H.; Cotal, H. L.;
Sudharsanan, R.; Eldredge, J. W.; Edmondson, K.; Joslin, D. E., Recent developments in
high-efficiency Ga0. 5In0. 5P/GaAs/Ge dual-and triple-junction solar cells: steps to
next-generation PV cells. Solar Energy Materials and Solar Cells 2001, 66, 453-466.
36. Geisz, J.; Kurtz, S.; Wanlass, M.; Ward, J.; Duda, A.; Friedman, D.; Olson, J.; McMahon,
W.; Moriarty, T.; Kiehl, J., High-efficiency Ga In P∕ Ga As∕ In Ga As triple-junction
solar cells grown inverted with a metamorphic bottom junction. Applied Physics Letters
2007, 91, 023502.
37. Kurtz, S.; Johnston, S.; Branz, H. M., Capacitance-spectroscopy identification of a key
defect in N-degraded GalnNAs solar cells. Applied Physics Letters 2005, 86, 113506.
38. Derkacs, D.; Jones-Albertus, R.; Suarez, F.; Fidaner, O., Lattice-matched multijunction
solar cells employing a 1 eV GaInNAsSb bottom cell. Journal of Photonics for Energy
2012, 2, 021805.

27

39. King, R.; Law, D.; Edmondson, K.; Fetzer, C.; Kinsey, G.; Yoon, H.; Sherif, R.; Karam,
N., 40% efficient metamorphic GaInP∕ GaInAs∕ Ge multijunction solar cells.
Applied physics letters 2007, 90, 183516.
40. Guter, W.; Schöne, J.; Philipps, S. P.; Steiner, M.; Siefer, G.; Wekkeli, A.; Welser, E.;
Oliva, E.; Bett, A. W.; Dimroth, F., Current-matched triple-junction solar cell reaching
41.1% conversion efficiency under concentrated sunlight. Applied Physics Letters 2009,
94, 223504.
41. Sasaki, K.; Agui, T.; Nakaido, K.; Takahashi, N.; Onitsuka, R.; Takamoto, T. In
Development of InGaP/GaAs/InGaAs inverted triple junction concentrator solar cells,
AIP Conference Proceedings, AIP: 2013; pp 22-25.
42. Price, J. S.; Sheng, X.; Meulblok, B. M.; Rogers, J. A.; Giebink, N. C., Wide-angle planar
microtracking for quasi-static microcell concentrating photovoltaics. Nature
communications 2015, 6, 6223.

28

Chapter 2. Methods
This chapter will describe experimental and numerical methods employed in this
dissertation, including epitaxial growth and microfabrication processes of triple- and
ten-fold-stack ultrathin GaAs solar cells and inverted metamorphic triple-junction
GaInP/GaAs/InGaAs solar cells, their optical, electrical, and morphological characterizations, as
well as numerical optical modeling of ultrathin GaAs solar cells with TiO 2 photonic
nanostructures.

2.1. Materials synthesis
2.1.1. Epitaxial growth of triple-stack GaAs solar cells by molecular beam epitaxy (MBE)
All III-V materials for GaAs solar cells used in this thesis work were grown by Mr.
Yukun Sun in Dr. Minjoo Lee’s group at the University of Illinois Urbana Champaign (UIUC) by
molecular beam epitaxy (MBE). For the MBE growth of triple-stack ultrathin GaAs solar cells
described in Chapter 3, a 200-nm-thick GaAs buffer layer was grown after de-oxidation,
followed by the growth of Al0.90Ga0.10As (0.5 µm/hr, V/III = 50), p
+
-GaAs (1 µm/hr, V/III = 15),
p
+
-Al0.4Ga0.6As (1.67 µm/hr, V/III = 20), p-GaAs (1 µm/hr, V/III = 15), n-GaAs (1 µm/hr, V/III =
15). The substrate was then cooled to 480 °C for the growth of n-Al 0.52In0.48P (1 µm/hr, V/III =
10) and n
+
-GaAs (1 µm/hr, V/III = 15) to complete the growth of a single device stack. Two
more device stacks were grown by repeating the above-described steps after increasing the

29

substrate temperature to 610 °C. Detailed growth conditions are summarized below (Table 2.1,
Credit: Dr. Minjoo Lee at UIUC).
Items Material
Thickness
(nm)
Growth
Temperature
(°C)
Growth Rate
(μm/h)
V/III Ratio
Buffer uid-GaAs 200 610 1.0 15
Sacrificial uid-Al 0.90Ga 0.10As 400 610 0.5 50
Bottom Contact p
+
-GaAs 50 610 1.0 15
BSF p
+
-Al 0.40Ga 0.60As 50 610 1.67 20
Base p-GaAs 250 610 1.0 15
Emitter n-GaAs 50 610 1.0 15
Window n-Al 0.52In 0.48P 20 480 1.0 10
Top Contact n
+
-GaAs 200 480 1.0 15
Table 2.1: Experimental conditions for the MBE growth of triple-stack ultrathin GaAs solar cells. All layers except
for “buffer” were repeated in the same way for 3 times.

2.1.2. Epitaxial growth of ten-fold-stack GaAs solar cells by molecular beam epitaxy (MBE)
For the demonstration of 10-fold-stack ultrathin GaAs solar cells described in Chapter 4,
a GaAs substrate was heated up to 650 °C for deoxidation and reduced to 610 °C before the
growth began. A buffer layer of 200 nm GaAs (1 μm/h, V/III = 15, unintentionally doped (UID))
was then grown, followed by the growth of 400 nm Al0.9Ga0.1As (0.5 μm/h, V/III = 50,
unintentionally doped), 50 nm p
+
-GaAs (0.5 μm/h, V/III = 25, Be-doped, 1 × 10
19
cm
−3
), 50 nm

30

p
+
-Al0.3Ga0.7As (0.71 μm/h, V/III = 20, Be-doped, 5 × 10
18
cm
−3
), 250 nm p-GaAs (1 μm/h, V/III
= 15, Be-doped, 3 × 10
17
cm
−3
), 50 nm n- GaAs (1 μm/h, V/III = 15, Si-doped, 2 × 10
18
cm
−3
), 20
nm n-Al0.4Ga0.6As (0.83 μm/h, V/III = 20, Si-doped, 2 × 10
18
cm
−3
), and 200 nm n
+
-GaAs (0.5
μm/h, V/III = 30, Si-doped, 5 × 10
18
cm
−3
). Nine more device stacks were grown by repeating the
above-described steps and substrate temperature was kept at 610 °C throughout the growth of all
device layers. Detailed growth conditions are summarized below (Table 2.2, Credit: Dr. Minjoo
Lee at UIUC).
Items Material
Thickness
(nm)
Growth
Temperature
(°C)
Growth Rate
(μm/h)
V/III Ratio
Buffer uid-GaAs 200 610 1 15
Sacrificial uid-Al 0.90Ga 0.10As 400 610 0.5 50
Bottom Contact p
+
-GaAs 50 610 0.5 25
BSF p
+
-Al 0.30Ga 0.70As 50 610 0.71 20
Base p-GaAs 250 610 1 15
Emitter n-GaAs 50 610 1 15
Window n- Al 0.4Ga 0.6As 20 610 0.83 20
Top Contact n
+
-GaAs 200 610 0.5 30
Table 2.2: Experimental conditions for the MBE growth of ten-fold-stack ultrathin GaAs solar cells. All layers
except for “buffer” were repeated in the same way for 10 times.

31

2.1.3. Epitaxial growth of triple-junction inverted metamorphic (3J IMM)
GaInP/GaAs/InGaAs solar cells by organometallic vapor phase epitaxy (OMVPE)
Epitaxial materials for triple-junction inverted metamorphic (3J IMM) solar cells used in
Chapter 5 were grown by Drs. John F. Geisz and Daniel Friedman at National Renewable Energy
Laboratory (NREL). The epitaxial growth was performed on (001) GaAs substrate miscut 2 
toward (111)B by atmospheric-pressure organometallic vapor phase epitaxy (OMVPE) at
temperatures ranging from 570 - 720°C using trimethylgallium, triethylgallium, trimethylindium,
trimethylaluminum, arsine, phosphine and dopant sources. The 3J IMM structure
1
is composed
of lattice-matched top Ga0.51In0.49P (Eg ~1.8 eV), middle GaAs (Eg ~1.4 eV), and metamorphic
bottom In0.26Ga0.74As (Eg ~1.0 eV) junctions, where two tunnel junctions of p
++
-Al0.6Ga0.4As
(C-doped)/n
++
-GaAs (Se-doped) lattice-matched to GaAs were incorporated to monolithically
connect the subcells in series. Notably, the GaInP and GaAs subcells utilized the
rear-heterojunction structure that is effective for reducing non-radiative Sah-Noyce-Shockley
(SNS) junction recombination and enhancing the external radiative efficiency
2, 3
. To suppress
strain-induced defects in the lattice-mismatched bottom junction, the composition of Ga xIn1-xP
buffer layers was step-graded from Ga0.51In0.49P (i.e. lattice matched to GaAs growth substrate, a0
= 5.66 Å) to Ga0.26In0.74P (i.e. lattice matched to In0.26Ga0.74As bottom junction, a0 = 5.76 Å). A
sacrificial layer (AlAs, ~0.1 m) was added before the growth of the typical 3J IMM structure to
enable the release of IMM solar cells without requiring the destructive removal of the growth

32

substrate. Detailed epitaxial design of 3J IMM solar cells appears in Table 2.3.
Items Material
Thickness
(μm)
Dopant
Bottom Contact Ga 0.744In 0.256As 0.2 Zn
BSF Ga 0.262In 0.738P 0.3 Zn
Base Ga 0.739In 0.261As 2.5 Zn
Emitter Ga 0.739In 0.261As 0.05 -
Window Ga 0.262In 0.738P 0.1 Si
Window Ga 0.262In 0.738P 0.9 Si
Window Ga 0.260In 0.740P 0.05 Si
Grading Buffer Ga 0.232In 0.768P 1 Si
Grading Buffer Ga 0.262In 0.738P 0.25 Si
Grading Buffer Ga 0.296In 0.704P 0.25 Si
Grading Buffer Ga 0.328In 0.672P 0.25 Si
Grading Buffer Ga 0.362In 0.638P 0.25 Si
Grading Buffer Ga 0.396In 0.604P 0.25 Si
Grading Buffer Ga 0.430In 0.570P 0.25 Si
Grading Buffer Ga 0.464In 0.536P 0.25 Si
Grading Buffer Ga 0.506In 0.494P 0.2 Si
Grading Buffer GaAs 0.03 Si
Tunnel Junction Ga 0.404Al 0.596As 0.0494 Se/Si
Tunnel Junction GaAs 0.0089 Se
Tunnel Junction Ga 0.404Al 0.596As 0.0197 C
BSF Ga 0.506In 0.494P 0.01 Zn

33

BSF Ga 0.506In 0.494P 0.3 Zn
Base GaAs 0.02 -
Emitter GaAs 3.5 Si
Window Ga 0.506In 0.494P 0.02 Si
Tunnel Junction Ga 0.404Al 0.596As 0.0494 Se/Si
Tunnel Junction GaAs 0.0089 Se
Tunnel Junction Ga 0.404Al 0.596As 0.0197 C
BSF Ga 0.506In 0.494P 0.01 Zn
BSF Ga 0.260In 0.470Al 0.270P 0.2 Zn
Base Ga 0.506In 0.494P 0.01 -
Emitter Ga 0.506In 0.494P 0.8 Si
Window In 0.480Al 0.520P 0.02 Se
Top Contact GaAs 0.1 Se
Top Contact GaAs 0.1 Se
Top Contact Ga 0.97In 0.03As 0.1 Se
Etch Stop Ga 0.51In 0.49P 0.05 -
Etch Stop Ga 0.506In 0.494P 0.5 Se
Sacrificial AlAs 0.1 -
Epi-buffer GaAs 0.1 Si
Table 2.3: Detailed epitaxial design of 3J IMM GaInP/GaAs/InGaAs solar cells.

2.2. Device fabrication
2.2.1. Fabrication of GaAs solar cells with a vertical contact configuration
The fabrication of triple-stack-grown GaAs microcells with a vertical contact

34

configuration started with the formation of n-type metal contact (490 × 30 m
2
, Pd/Ge/Au: 5
nm/35 nm/80 nm) by electron beam evaporation (Temescal) and lift-off. Active cell areas (~500
× 500 m
2
) were delineated by photolithography (AZ1518, AZ 400K 4:1 developer, Merk KGaA)
and successive wet chemical etching of n
+
-GaAs in the mixture (4:1, by volume) of citric acid
(100 g of citric acid monohydrate (≥99.0%, Sigma Aldrich) in 83 ml deionized (DI) water) and
hydrogen peroxide (H2O2, 30%-32%, Macron), n-Al0.52In0.48P in a dilute (1:1, by volume with DI
water) hydrochloric acid (36.5%-38.0%, EMD), and n-GaAs (50 nm)/p-GaAs (100 nm) in citric
acid/H2O2 (20:1, by volume). Subsequently, an additional mesa (560 × 560 m
2
) structure was
formed to isolate individual microcells by photolithography and wet chemical etching by citric
acid/H2O2 (20:1, by volume). The exposed Al0.90Ga0.10As was partially etched by hydrofluoric
acid (48.5%-50.5%, EMD), followed by the spin-coating of a photoresist (as a polymeric anchor)
and the formation of etch holes on the second mesa by photolithography and wet chemical
etching (citric acid/H2O2, 20:1, by volume). After the undercut etching of Al 0.90Ga0.10As in
HCl/DI water (1:4, by volume), arrays of microcells were retrieved by an elastomeric stamp
made of poly(dimethylsiloxane) (PDMS, Sylgard
TM
184, Dow Corning), and metals (Pt/Ti/Pt/Au:
10 nm/40 nm/10 nm/80 nm or Cr/Ag/Au: 3 nm/100 nm/40 nm) were deposited on the exposed
p
+
-GaAs as a p-type metal contact. Retrieved microcells were printed on a glass substrate using a
thin (~1 m) photocurable adhesive followed by thermal annealing (175 C, 60 min in N2) to
form ohmic contacts. Subsequently, the n
+
-GaAs top contact layer was removed except beneath

35

the metal-deposited region in a mixture of ammonium hydroxide (28.0%-30.0%, EMD),
hydrogen peroxide, and DI water (2:1:50, by volume). The bottom metal contact (490 × 30 m
2
)
was exposed by photolithography and wet chemical etching in a mixture of phosphoric acid
(85%, Fischer Scientific), H2O2, and H2O (1:13:12, by volume). Chemicals used and associated
processing conditions are summarized in Table 2.4 and 2.5, respectively.

Chemical Name
Chemical
Formula
Supplier Concentration Grade
Hydrochloric acid HCl EMD 36.5-38.0 wt.% GR ACS
Citric acid C 6H 8O 7 Sigma-Aldrich
100 g of citric acid
monohydrate in 83
mL of deionized
(DI) water
≥99.0%
Hydrogen peroxide H 2O 2 Macron 30-32 wt.%
Microelectronic
Grade
Hydrofluoric acid HF EMD 48.5-50.5 wt.% GR ACS
Phosphorus acid H 3PO 4 Fischer Scientific 85 wt.% GR ACS
Ammonium
hydroxide
NH 4OH EMD 28.0-30.0 wt.% GR ACS
Table 2.4: Chemicals for wet etching processes employed in this thesis work. The information in this table applies
to all the processes related to wet etching in this dissertation, unless specifically mentioned.

36

Process Name Process Conditions
Native oxide removal
(before spin-coating PR and before metal
deposition)
NH 4OH:DI water = 1:10 for 1 min
Native oxide removal
(before GaAs or AlGaAs etching)
HCl:DI water = 1:1 for 30 s
Mesa etching
Citric acid:H 2O 2 = 4:1 for 1 min (top contact layer
GaAs 200 nm)
HCl:DI water = 1:1 for 5 s (window layer
Al 0.52In 0.48P 20 nm)
Citric acid:H 2O 2 = 4:1 for 1.5 min (emitter and base
layer GaAs 300 nm)
Isolation etching
Citric acid:H 2O 2 = 20:1 for 2 min (BSF layer
Al 0.40Ga 0.60As 50 nm and bottom contact layer GaAs
50 nm)
Partial etching of sacrificial layer HF:DI water = 1:10 for 25 s
Undercut etching
Citric acid:H 2O 2 = 20:1 for 3 min (BSF layer
Al 0.40Ga 0.60As 50 nm and bottom contact layer GaAs
50 nm)
HCl:DI water=1:1 for 5 min (part of Al 0.9Ga 0.1As
sacrificial layer)
HCl:DI water=1:4 for ~6.5 hours (all of
Al 0.90Ga 0.10As sacrificial layer) (~3 hours for TLM
cells)

37

Bottom contact etching
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 8 s (top contact
layer GaAs 200nm)
HCl:DI water = 1:1 for 5 s (window layer
Al 0.52In 0.48P 20 nm)
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 15 s (all the
remaining semiconductor layers)
Top contact etching
NH 4OH:H 2O 2:DI water = 2:1:50 for 40 s (top
contact layer GaAs 200nm)
Device stack overlayer etching
It is the same recipe as mesa etching, isolation
etching and sacrificial partial etching together for
transfer printing case, with repeating for once to
expose the middle stack, or repeating twice to
expose the bottom stack, respectively.
Top contact etching
(for on wafer sample)
Citric acid:H 2O 2 = 4:1 for 1 min (top contact layer
GaAs 200 nm)
Bottom contact etching
(for on wafer sample)
HCl:DI water = 1:1 for 5 s (window layer
Al 0.52In 0.48P 20 nm)
Citric acid:H 2O 2 = 4:1 for 1.5 min (emitter and base
layer GaAs 300 nm)
Citric acid:H 2O 2 = 20:1 for 40 s (BSF layer
Al 0.4Ga 0.6As 50 nm)
Isolation etching (for on wafer sample)
It is the same recipe as mesa etching and isolation
etching together for transfer printing case.
Table 2.5: Processing conditions of wet chemical etching for the fabrication of GaAs solar cells.

38

2.2.2. Fabrication of GaAs solar cells with TiO2 nanoposts (NPs)
GaAs microcells with TiO2 NPs were fabricated by procedures adopted from our previous
work
4
, as schematically illustrated in Figure 2.1. After wet chemical etching of the top contact
layer except the region beneath the metal contact, TiO 2 (~180 nm) was deposited by RF
magnetron sputtering (Orion 5, AJA International, Ar: 15 sccm, 3 mTorr, 100 W). Hexagonally
periodic Cr islands as a mask for subsequent dry etching were fabricated on the TiO 2-deposited
surface by softimprint lithography, oxygen reactive ion etching (Plasmalab, 10 W, 20 mTorr, 80
s), and electron beam evaporation of Cr (8 nm). For nanoimprinting, poly(methyl methacrylate)
(PMMA, M.W. = 996,000 g/mol, Sigma-Aldrich) solution (1.8 wt% in anisole) and dilute SU-8
2000.5 (Microchem, 40wt% in SU-8 thinner) were successively spin-coated. Subsequently,
inductively coupled plasma reactive ion etching (ICP RIE, Plasmalab system 100, Oxford) was
performed to produce TiO2 NPs via Pseudo-Bosch process (SF6:C4F8 = 10 sccm : 20 sccm, 10
mTorr, RF/ICP = 50W/500W, 120 s), followed by etching a residual C 4F8 layer by oxygen RIE
(10 W, 100 mTorr, 1 min) and removing Cr islands in an Cr etchant (CR-7, KMG). After the
formation of TiO2 NPs, above-described processing steps for bare GaAs solar cells were
performed, starting from the mesa (500 × 500 m
2
) definition.

39

Figure 2.1: Fabrication procedures of implementing TiO 2 NPs on GaAs solar cells.

2.2.3. Fabrication of 3J IMM microcells
The fabrication process of dense-array IMM microcells began with the deposition of
p-type ohmic contact (Cr/Au = 5/70 nm, ~467  471 μm
2
) by electron beam evaporation
(Temescal) and lift-off (AZ 5214 E, Microchemicals GmbH). Subsequently, the active junction
area (~500  500 m
2
) and isolated cell boundary (580  580 m
2
) were defined by
photolithography (AZ 1518 for active defining active junction area, AZ 4620 for isolation)
and wet chemical etching. The etchant for phosphide layers (Ga xInyAl1-x-yP) was hydrochloric
acid (HCl, 36.5-38.0 wt.%, EMD) without dilution, while arsenide layers (Ga xInyAl1-x-yAs) were
etched by the mixture of phosphoric acid (H3PO4, ~85%, EMD), hydrogen peroxide (H2O2,
30%-32%, Macron), and deionized (DI) water (H2O) with the volume ratio of 1:13:12. Undercut

40

etching of a sacrificial layer (i.e. AlAs) was performed using diluted hydrofluoric acid (HF
(48.5%-50.5%, EMD): DI = 1:100 by volume). The entire array of undercut-etched microcells
was picked up by an elastomeric stamp made of poly(dimethylsiloxane) (PDMS, Sylgard
TM
184,
Dow Corning) and printed on a glass substrate using photocurable polyurethane (NOA 61,
Norland Products Inc.) as an adhesive. After the printing, the GaInP buffer layer was removed
and the bottom p-type contact was exposed by wet chemical etching. The fabrication of IMM
microcells was completed by the deposition of n-type ohmic contact (Cr/Cu/Au = 15/2000/100
nm) and removal of GaAs contact layer. More processing details of wet chemical etching for
IMM microcells appear in Table 2.6.
Process Name Process Conditions
Native oxide removal
(before spin-coating PR and before metal
deposition)
NH 4OH:DI water = 1:10 for 1 min
Native oxide removal
(before GaAs or AlGaAs etching)
HCl:DI water = 1:1 for 30 s

41

Mesa etching
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 8 s (Ga 0.74In 0.26As
bottom contact layer 200 nm)
HCl concentrated for 15 s (Ga 0.26In 0.74P BSF layer
300 nm)
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 140 s (Ga 0.74In 0.26As
bottom cell layer 2550 nm)
HCl concentrated for 60 s (Ga 0.26In 0.74P graded
buffer layer 4000 nm)
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 5 s
(Al 0.60Ga 0.40As/GaAs tunnel junction layer 108 nm)
HCl concentrated for 20 s (Ga 0.51In 0.49P BSF layer
310 nm)
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 125 s (GaAs middle
cell layer 3520 nm)
HCl concentrated for 20 s (Ga 0.51In 0.49P window
layer 20 nm)
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 5 s
(Al 0.60Ga 0.40As/GaAs tunnel junction layer 78 nm)
HCl concentrated for 20 s (Ga 0.51In 0.49P top cell
layer 1040 nm)
Isolation etching
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 12 s (GaAs top
contact layer 300 nm)
HCl concentrated for 60 s (Ga 0.51In 0.49P buffer layer
550 nm)

42

Undercut etching
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 18 s (GaAs top
contact layer 300 nm)
HCl concentrated for 5 min (Ga 0.51In 0.49P buffer
layer 550 nm)
HF:H 2O = 1:100 for ~12 h (AlAs sacrificial layer
100 nm)
Bottom contact etching (Reversed sequence of “Mesa etching”)
Top contact etching
H 3PO 4:H 2O:H 2O 2 = 1:12:13 for 12 s (GaAs top
contact layer 300 nm)
Table 2.6: Processing conditions of wet chemical etching for the fabrication of IMM solar cells.

To fabricate sparse-array module of IMM microcells, a (100) silicon wafer was employed
as a temporary carrier substrate for print-ready IMM microcells after the surface was roughened
by metal-assisted wet chemical etching. Subsequently, a triple-stack of metals (Cr/Ag/Cr) was
deposited onto the etched surface as a sacrificial layer by electron-beam evaporation, followed
by the spin-coating and curing (180°C, 1 h in N2) of polyimide (PI 2555, HD MicroSystems
TM
)
as a supporting ‘base’ layer (~ 2 m). The released IMM microcells from the growth wafer
were then transfer-printed onto the PI layer coated with a thin (~1 m) layer of adhesive. After
the printing, same procedures as on a module substrate of dense cell arrays were performed for
etching the GaInP buffer layer, depositing n-type contact, and exposing the bottom p-type contact
to yield fully-function 3J IMM microcells in an upright configuration. To make the printed
microcells into a print-ready form, the isolation of an individual microcells was performed by

43

photolithography (AZ 4620) and reactive ion etching (CHF3:O2 = 5 sccm: 45 sccm, 100 mTorr,
100W, ~10 min, PlasmaPro 80, Oxford Instruments). After forming a polymeric anchor structure,
undercut etching was performed using a chromium etchant (CR-7, KMG Electronic Chemicals®)
for 1 h. Released microcells were individually pick up and printed onto a glass substrate using an
elastomeric stamp with relief features.

2.3. Materials and device characterizations
Reflectance spectra were measured using a home-made optical setup including a white
light source (HL-2000, Ocean optics) and a fiber-optic spectrometer (Flame-T-VIS-NIR, Ocean
Optics).
4
The source light was collimated by an achromatic doublet lens (f = 19 mm) and then
focused on the cell region (beam diameter = ~50 m) through an objective lens (20X, N.A. =
0.4). The reflected light was collected with the same objective lens and guided to the
spectrometer (Flame-T-VIS-NIR, Ocean Optics) through a multimode optical fiber using an
achromatic fiber-port lens. A silver mirror deposited on fused silica (Thorlabs) was used as a 100%
calibration standard for reflectance measurements. In characterization of materials properties and
device performance at specific device stacks, overlying epitaxial layers were removed by
repetitively etching device stack (in a mixture of phosphoric acid (85%, Fischer Scientific),
hydrogen peroxide (H2O2, 30%-32%, Macron), and deionized (DI) water (H 2O) with the volume
ratio of 1:13:12) and sacrificial layers (in diluted hydrofluoric acid (HF (48.5%-50.5%, EMD):

44

DI = 1:10 by volume). Microscale (~500 x 500 m
2
) GaAs solar cells on each device stack with
a recessed bottom contact were fabricated by previously reported procedures
5, 6
. The photovoltaic
performance was characterized using a semiconductor parameter analyzer (4156C, Agilent
Technologies) and a full-spectrum solar simulator (94042A, Oriel). External and internal
quantum efficiencies were obtained from QE measurement system (QEX7, PV Measurements).
SIMS measurements were performed by a commercial vendor (EAG Laboratories). Scanning
electron microscopy images were taken by a Hitachi S-4800 field emission SEM.
Photovoltaic performance of IMM microcells in dense arrays was measured under
simulated AM1.5D illumination (1000 W/m
2
) at NREL after adjusting the spectrum with Xenon
lamp and LED illumination, where the correction of spectral mismatch was made using isotype
reference cells. The photovoltaic performance of IMM microcells in sparse arrays was measured
under simulated AM1.5G illumination (1000 W/m
2
) at USC using a semiconductor parameter
analyzer (4156C, Agilent Technologies) and a full-spectrum solar simulator (94042A, Oriel).
The performance of printed 3J IMM microcells under high concentration (up to ~2500 suns) was
characterized using a high intensity pulsed solar simulator (HIPSS) without precise spectral
control, where the short-circuit current at one sun was used as a reference to estimate
concentration ratio assuming the linearity of the short-circuit current with the light intensity.

45

2.4. Numerical Optical Modeling
Optical FDTD calculations were performed to obtain reflectance and absorption spectra
of ultrathin GaAs solar cells by a commercial software package (FDTD Solutions 8.15,
Lumerical)
6, 7
. A 3D simulation volume comprising epitaxial layers identical to the experimental
GaAs solar cell was defined, where a periodic boundary condition in the x- and y directions and
a perfectly matched layer (PML) boundary condition in the z-direction were applied. A
continuous plane-wave source with a broad Gaussian frequency spectrum (330−750 THz) was
normally incident in the −z-direction to the simulation volume.

46

2.5 References
1. Geisz, J. F.; Kurtz, S. R.; Wanlass, M. W.; Ward, J. S.; Duda, A.; Friedman, D. J.; Olson, J.
M.; McMahon, W. E.; Moriarty, T.; Kiehl, J., High-efficiency GaInP/GaAs/InGaAs
triple-junction solar cells grown inverted with a metamorphic bottom junction. Appl.
Phys. Lett. 2007, 91, 023502.
2. Geisz, J. F.; Steiner, M. A.; García, I.; Kurtz, S. R.; Friedman, D. J., Enhanced external
radiative efficiency for 20.8% efficient single-junction GaInP solar cells. Applied Physics
Letters 2013, 103, 041118.
3. Essig, S.; Allebé, C.; Remo, T.; Geisz, J. F.; Steiner, M. A.; Horowitz, K.; Barraud, L.;
Ward, J. S.; Schnabel, M.; Descoeudres, A.; Young, David L.; Woodhouse, M.; Despeisse,
M.; Ballif, C.; Tamboli, A., Raising the one-sun conversion efficiency of III–V/Si solar
cells to 32.8% for two junctions and 35.9% for three junctions. Nature Energy 2017, 2,
17144.
4. Lee, S. M.; Kwong, A.; Jung, D.; Faucher, J.; Biswas, R.; Shen, L.; Kang, D.; Lee, M. L.;
Yoon, J., High Performance Ultrathin GaAs Solar Cells Enabled with Heterogeneously
Integrated Dielectric Periodic Nanostructures. Acs Nano 2015, 9, 10356-10365.
5. Gai, B.; Sun, Y .; Lim, H.; Chen, H.; Faucher, J.; Lee, M. L.; Yoon, J., Multilayer-Grown
Ultrathin Nanostructured GaAs Solar Cells as a Cost-Competitive Materials Platform for
III–V Photovoltaics. ACS nano 2017, 11, 992-999.

47

6. Lee, S.-M.; Kwong, A.; Jung, D.; Faucher, J.; Biswas, R.; Shen, L.; Kang, D.; Lee, M. L.;
Yoon, J., High performance ultrathin GaAs solar cells enabled with heterogeneously
integrated dielectric periodic nanostructures. ACS nano 2015, 9, 10356-10365.
7. Chan, L.; Kang, D.; Lee, S.-M.; Li, W.; Hunter, H.; Yoon, J., Broadband antireflection
and absorption enhancement of ultrathin silicon solar microcells enabled with
density-graded surface nanostructures. Applied Physics Letters 2014, 104, 223905.

48

Chapter 3. Multilayer-grown ultrathin nanostructured GaAs solar
cells
3.1 Introduction
Despite well-suited materials properties and the record-high efficiency approaching their
theoretical limit, practical application of single-junction gallium arsenide (GaAs) solar cells in
terrestrial photovoltaics has been largely precluded due to excessively high materials costs for
growing device-quality epitaxial materials on an expensive native substrate.
1-3
While epitaxial
liftoff (ELO) has been proposed as a promising concept to potentially realize significant
reduction of the cell cost by reusing the growth substrate,
4-7
restoring an epi-ready condition of
the used wafer and repeatedly performing epitaxial growth naturally involve many processing
steps and thermal cycles, leading to the accumulation of crystalline defects and deterioration of
manufacturing yields.
8
In this regard, Yoon and co-workers have reported an alternative approach
to circumvent these drawbacks in conventional ELO by growing a large number of individually
releasable device stacks on a single growth wafer such that an excessive number (e.g. >100) of
substrate reclaims may not be strictly necessary for driving the cost reduction.
9, 10
However, the
practical implementation of this idea in GaAs photovoltaics has been hampered by difficulties in
preserving uniform materials properties and performance characteristics between device layers
grown in different sequences. For example, in previously reported ‘n-on-p’ type triple-stack (i.e.
top, middle, bottom) GaAs solar cells grown by metal organic chemical vapor deposition

49

(MOCVD), the efficiencies of solar cells exhibited systematic degradation from top, middle, to
bottom device stacks owing to the thermally-driven diffusion of zinc incorporated as a p-type
dopant and the resulting modification of optimized junction properties as well as carrier
compensation in n-type regions.
9, 11
More recently, carbon was used as an alternative p-type
impurity in the ‘p-on-n’ type cell configuration with the rationale of its low tendency of thermal
diffusion and reduced thickness of carbon-doped regions.
10, 12
However, carbon-related point
defects associated with prolonged heat treatments in the multilayer growth
13-15
as well as dopant
passivation by hydrogen
16, 17
resulted in the noticeable deterioration of device performance,
where the solar-to-electric conversion efficiencies at middle and bottom layer devices were ~95%
and ~91% of the efficiency from top devices.
10
From these studies, the diffusion of p-type dopant
and extended heat treatments at a high growth temperature were identified as major
technological barriers that must be addressed to permit this approach to be practical for
photovoltaic applications. In this context, GaAs solar cells having ultrathin (e.g. 200-300 nm)
active layers were conceived as an attractive device configuration for multilayer epitaxy as they
can improve the overall economics of materials preparation but also substantially decrease the
growth time to minimize thermally-induced deterioration of materials properties. Furthermore, it
was recently demonstrated that the performance of such ultrathin GaAs solar cells, when
combined with optimized schemes of photon management, can be significantly enhanced to a
level close to optically thick absorbers, thereby further improving the cost-effectiveness obtained

50

from the reduced materials utilization.
18-20
Motivated by these advances, herein we present a
cost-competitive materials platform for multilayer-grown GaAs solar cells that can circumvent
prior challenges and achieve uniform (< 3% relative) interlayer cell efficiencies, by employing
an order of magnitude thinner active layers (i.e. emitter and base: 300 nm) and specialized
epitaxial design to minimize parasitic optical losses for reflected photons from a back-side
reflector (BSR) in bifacial light management. Beryllium, which is known as a less mobile in
GaAs than zinc,
21, 22
was used as a p-type dopant in the multilayer epitaxial growth by molecular
beam epitaxy (MBE) to suppress the effect of dopant diffusion. Enabled with printing-based
materials assemblies,
23-26
hexagonally periodic TiO2 nanoposts and metallic BSR were
incorporated to the front and rear-surface of multilayer-grown ultrathin GaAs solar cells in a
vertical contact configuration to compensate for the insufficient light absorption near the
bandgap energy, leading to a 17.2% solar-to-electric energy conversion efficiency under
simulated AM1.5G solar illumination with 420-nm-thick single-junction GaAs solar cells. In this
work, systematic studies of electrical and optical properties, secondary ion mass spectrometry
(SIMS), and photovoltaic device characteristics of multilayer-grown ultrathin GaAs solar cells in
various materials configurations, together with optical modeling based on a finite-difference
time-domain (FDTD) method, provide quantitative descriptions of the materials design and
fabrication concept, as well as key design rules for multilayer-grown single-junction GaAs cells.

51

3.2 Results and discussion
Figure 3.1a illustrates schematically the processing steps for multilayer-grown ultrathin
GaAs solar cells in a ‘vertical’ contact configuration, where a p-type ohmic contact is formed
directly underneath the bottom contact layer. The triple-stack ultrathin GaAs solar cells were
grown in a solid-source molecular beam epitaxy (MBE) system equipped with effusion cells (for
Ga, In, Al, Si and Be) and valved cracker sources (for As and P). Phosphide layers and n-type
GaAs contact layers were grown at 480 °C while all arsenide layers were grown at 610 °C. The
fabrication step started with the deposition of metals (Pd/Ge/Au: 5 nm/35 nm/80 nm) to form
n-type ohmic contact by electron beam evaporation and lift-off. Active junction areas (~500 ×
500 m
2
) of microscale solar cells (i.e. microcells) were delineated by photolithography and wet
chemical etching (Figure 3.1b; upper left).
19
Similarly, an additional mesa (580 × 580 m
2
)
structure was formed to isolate individual microcells such that etch holes for the delivery of
etchant to a sacrificial layer (Al0.90Ga0.10As) were made outside the active cell area (Figure 3.1b;
upper right) to minimize undesired performance degradation by surface recombination. Partial
etching of the sacrificial layer and spin-coating of photoresist formed a heterogeneous ‘anchor’
structure,
10, 19
followed by the formation of etch holes on the second mesa and the selective
removal of Al0.90Ga0.10As using dilute hydrochloric acid (HCl). After the undercut etching, arrays
of microcells were retrieved by an elastomeric stamp made of poly(dimethylsiloxane) (PDMS),
and the exposed bottom surface (i.e. p
+
-GaAs) of microcell was deposited with metals

52

(Pt/Ti/Pt/Au: 10 nm/40 nm/10 nm/80 nm or Cr/Ag/Au: 3 nm/100 nm/40 nm) to form p-type
ohmic contact that can also serve as a specular BSR (Figure 3.1b; lower left). Microcells
implemented with vertical metal contacts were then printed onto a glass substrate using a thin
(~1 m) layer of photocurable adhesive (Figure 3.1b; lower right).
10, 27
Wet chemical etching to
remove the top contact layer and expose the bottom metal contact completed the entire
fabrication processes. Figure 3.1c depicts photographic images of completed ultrathin GaAs
solar cell arrays printed on a glass substrate. The same fabrication procedure for the top device
layer can be applied to underlying device stacks in a repetitive manner.

53

Figure 3.1: (a) Schematic illustration of the fabrication procedures (b) Optical micrographs of arrays of GaAs
microcells at various processing steps (c) Photographic image of 10 × 10 arrays GaAs solar microcells with inset
shows the magnified view.

The individual ultrathin single-junction GaAs solar cell is composed of n
+
-GaAs top
contact (200 nm, Si-doped, 5 x 10
18
cm
-3
), n-Al0.52In0.48P window (20 nm, Si-doped, 2 × 10
18

cm
-3
), n-GaAs emitter (50 nm, Si-doped, 2 × 10
18
cm
-3
), p-GaAs base (250 nm, Be-doped, 3 ×

54

10
17
cm
-3
), p
+
-Al0.40Ga0.60As back surface field (BSF) (50 nm, Be-doped, 5 × 10
18
cm
-3
), and
p
+
-GaAs bottom contact (50 nm, Be-doped, 1 × 10
19
cm
-3
) (Figure 3.2a). Al0.90Ga0.10As (400 nm)
was incorporated as a sacrificial layer underneath the bottom contact (i.e. p
+
-GaAs) of each
device stack to enable the sequential release of individual solar cells from the multilayer stacks.
Notably, the thickness (i.e. 50 nm) of the bottom contact layer was intentionally thinned down to
minimize parasitic optical losses and maximize the effect of bifacial light trapping. Figure 3.2b
shows cross-sectional scanning electron microscope (SEM) image of the triple-stack GaAs solar
cells, where MBE-grown defect-free epitaxial layers including dark regions of sacrificial
Al0.90Ga0.10As between device stacks are clearly identified. As the first step to evaluate the
performance of multilayer-grown solar cells, p- and n-type contact properties including contact
resistance (Rc) and specific contact resistivity ( c) were characterized by a standard transmission
line model (TLM) method
10, 28
. For p-type contact properties, microcells released from each
device stack were flipped over by a two-step transfer process using a PDMS stamp such that the
p
+
-GaAs is exposed to the front surface for the subsequent deposition of TLM metal pads (see
the inset of Figure 3.2d). As summarized in Figures 3.2c and 3.2d, and Table 3.1, both p- and
n-type contacts after a single-step thermal annealing (175 C, 60 min under N2 atmosphere)
exhibited a perfect ohmic behavior with a comparable range of specific contact resistivity
between device stacks (see Figures 3.3 and 3.4), supporting uniform concentration of activated
p- (beryllium) and n-type (silicon) dopants in contact layers of top, middle, and bottom device

55

layers.

Figure 3.2: (a) Detailed epitaxial design (b) cross-sectional view SEM image. (c) n-type for total resistance (R) as a
function of metal pad spacing (x) from standard transmission line model (TLM) measurements and (d) for p-type.

Items Top Middle Bottom
n-type
Rc (Ω) 3.5 4.5 4.0
ρc (Ω cm2) 9.3 × 10
-5
1.4 × 10
-4
1.1 × 10
-4

p-type
Rc (Ω) 36.9 35.5 33.1
ρc (Ω cm2) 2.1 × 10
-4
2.0 × 10
-4
1.6 × 10
-4

Table 3.1: Calculated contact resistance (R c) and contact resistivity (ρ c) from TLM Results.

56

Figure 3.3: I-V curve for n-type TLM measurements.

57

Figure 3.4: I-V curve for p-type TLM measurements.

One of key challenges in the growth of multilayer GaAs solar cells is to preserve the
concentration profile of activated p-type dopants throughout the entire device stacks to ensure
uniform electronic configuration of solar cells grown in different sequences. Given that both zinc
and carbon are not ideal for the multilayer epitaxial growth owing to the fast diffusion and
generation of dopant-related point defects, respectively,
9, 10, 29, 30
we chose beryllium as an
alternative p-type dopant with advantages such as the ease of ionization and comparatively lower
tendency of diffusion in GaAs than zinc.
21, 22
To quantitatively assess the extent of beryllium

58

diffusion in the triple stack GaAs solar cells, we carried out secondary ion mass spectrometry
(SIMS) measurements with as-grown triple-stack samples. Figures 3.5a, b, and c show SIMS
depth profiles for top, middle, and bottom device stacks, where red, blue, and green lines are
atomic concentration (in atoms per cm
3
) of beryllium and silicon, and the secondary ion intensity
(in counts per cm
3
) from aluminum, respectively. Both p- and n-type dopant concentrations
matched well with the target values in the top device layer, while the beryllium profile became
gradually less sharp in the middle and bottom device stacks due to slight thermal diffusion
during the growth of upper device layers. The average concentration of beryllium in the base
layer has been slightly changed from ~2.9 × 10
17
cm
-3
(top) to ~8.2 × 10
17
cm
-3
(middle) and ~7.5
× 10
17
cm
-3
(bottom). Nevertheless, the dopant concentrations in both top and bottom contact
layers were still comparable in all device stacks, consistent with above-described TLM studies.
Overall, the p-type dopant (i.e. beryllium) is much less diffused than in the previous zinc-doped
system grown by MOCVD. Due to the extended heat treatment during the multilayer epitaxy, the
concentration of zinc in the n-type contact region was almost comparable to the concentration of
silicon, while it increased to the level over 10
17
cm
-3
in the emitter, causing severe carrier
compensation and modification of junction characteristics.
9

59

Figure 3.5: SIMS results for top (a), middle (b) and bottom (c).

Photovoltaic performance of triple-stack ultrathin GaAs solar cells was examined on the
growth wafer as well as after transfer printing on a glass substrate. Figure 3.6a shows
representative current density (J)-voltage (V) curves of GaAs microcells on wafer, measured
under simulated AM1.5G solar illumination (1000 W/m
2
) without any additional optimization
such as an antireflection coating (ARC). For cells tested on the growth wafer, the p-type ohmic
contact was evaporated in a coplanar, ‘recessed’ layout (~480 × 30 m
2
, see the inset of Figure
3.6a),
9, 19
while it was formed on the entire bottom surface of the released cell to yield a ‘vertical’

60

contact design in printed devices. As summarized in Figure 3.6b, top, middle, and bottom layer
devices exhibited comparable photovoltaic device characteristics, where the average short-circuit
current densities (Jsc) were 14.6, 14.5, and 14.3 mA/cm
2
, open-circuit voltages (Voc) were 0.950,
0.942, 0.941 V , fill-factors (FF) were 0.78, 0.77, 0.78, efficiencies ( ) were 10.8, 10.6, 10.5%,
respectively. The Jsc and  were evaluated based on the active junction area excluding metal
contacts.
9
The slight variation (< ~3% relative) of efficiencies is within the range of experimental
errors (e.g. fluctuation of the simulated light intensity), which is significantly smaller than
previously reported systems. In terms of efficiency, the degradation between top and bottom
devices was ~16.0% and ~10.4% for zinc and carbon-doped GaAs solar cells, respectively, while
it is 2.8% for the beryllium doped ultrathin devices. Notably, a little increase of beryllium
concentration in the base and emitter for the middle and bottom device stacks as shown in SIMS
measurements, which might cause a slight modification in the electronic configuration such as
the junction position and depletion region width, did not translate to the noticeable deterioration
of solar cell performance. Cells measured after printing on a glass substrate also performed as
comparably as those on wafer (Figures 3.6c and d), where the average Jsc were 14.2, 14.1, and
14.5 mA/cm
2
, Voc were 0.928, 0.939, 0.949 V , FF were 0.79, 0.79, 0.79,  were 10.4, 10.5,
10.8%, respectively. As expected, the vertical contact design in the printed cells promoted more
efficient current spreading at the bottom contact and thus appreciable reduction of series
resistance as evidenced in the dark JV curves (Figures 3.6e and 3.7, Table 3.2). It should be

61

mentioned that the almost comparable performance of printed cells to those on the wafer is
attributed to the slight (~10%) variation of absorber layer thickness across the diameter of the
growth wafer as well as moderate reflectance of the ohmic metal contact (Pt/Ti/Pt/Au) used here
(Figure 3.8). The internal (IQE) and external (EQE) quantum efficiencies (Figure 3.6f)
measured on wafer also supported the uniform device performance of multilayer grown ultrathin
GaAs solar cells, where the wavelength (~450 nm) of maximum IQE is shorter than optically
thick GaAs solar cells owing to the insufficient absorption of long wavelength photons in a
single optical pass.

62

Figure 3.6: Photovoltaic performance for on wafer sample (a), (b), printed sample (c), (d), dark I-V (e), and
quantum efficiency (f) for top, middle and bottom devices.

63

Bottom
Contact
Configuration
Stack I0 (A) n Rs (Ω) Rsh (Ω)
Vertical
Top 1.0 × 10
-12
2.1 4.2 8.9 × 10
9

Middle 5.5 × 10
-12
2.3 4.7 1.6 × 10
9

Bottom 4.3 × 10
-12
2.3 2.2 1.3 × 10
9

Resessed
Top 5.4 × 10
-13
2.1 233.0 2.0 × 10
10

Middle 5.4 × 10
-13
2.1 326.1 1.0 × 10
10

Bottom 7.3 × 10
-13
2.1 320.1 1.5 × 10
10

Table 3.2: I 0, n, R s, R sh information from fitted dark I-V results.

64

Figure 3.7: Fitting curve and model for Dark IV .

65

Figure 3.8: Reflectance for bottom contact metal.

Given the limited absorption of near-bandgap photons in optically thin GaAs solar cells,
incorporating optimized schemes of photon management is essential to enhance their
photovoltaic performance and thus cost-effectiveness. Accordingly, we incorporated hexagonally
periodic TiO2 nanoposts (NPs) on top of the Al0.52In0.48P window layer of multilayer-grown GaAs
solar cells through procedures adopted from our previous work including radio-frequency
magnetron sputtering, softimprint lithography, and reactive ion etching.
19
To determine optimal
designs of TiO2 NPs, the integrated solar flux absorption (S_abs) weighted over the simulated
AM1.5G solar illumination was calculated by,
𝑆 (%) =
∫
( ) . ( ) ∫
. ( ) × 100 (3-1)
, where h, c, A( ), and I1.5G ( ) are Planck’s constant, the speed of light, absorption calculated by

66

a finite-difference time-domain method (FDTD), and the standard solar irradiance (AM 1.5G;
ASTM G-173), respectively.
19
Figure 3.9a shows a contour plot of S_abs for nanostructured
420-nm-thick GaAs solar cells at the NP period (p) of 500 nm as a function of diameter (D) and
height (h) of NPs, where TiO2 NPs and a silver BSR were implemented on the front and rear
surfaces of the cells, respectively (Figure 3.10). The two local absorption maxima in Fig. 3.9(a)
are 77.6% and 78.4% at the NP diameter (D)/height (h) of 300/150 nm and 300/350 nm,
respectively. These calculations were further extended over a range of NP periods (200-800 nm)
to identify the optimal design of TiO2 NPs as shown in Figure 3.9b. The maximum integrated
absorption is ~78.6% at the NP period of 400 nm (D = 280 nm, h = 120 nm), which is even
higher than the case (S_abs = 73.7%) with a double layer ARC (DLARC, TiO2/SiO2 = 60
nm/100 nm) due to the combined optical effects of antireflection, diffraction, and light
trapping.
19, 24
Figure 3.9c depicts measured (solid line) and calculated (dotted) reflectance
spectra of ultrathin GaAs solar cells on wafer as well as after the printing, with and without the
near-optimal design of TiO2 NPs (D = 330 nm, h = 140 nm, p = 500 nm, base layer thickness =
20 nm), in which the broadband ARC effect of TiO2 NPs is evidently shown. As summarized in
Figure 3.9d, the photovoltaic performance of nanostructured GaAs solar cells significantly
improved owing to the synergistic effects of bifacial photon management and the ultrathin (~50
nm) bottom contact layer to minimize the parasitic optical losses. By exploiting TiO 2 NPs and
p-type metal contact (Cr/Ag/Au = 3 nm/100 nm/40 nm, Figure 3.10) of high reflectance as a

67

specular BSR, the short-circuit current density increased from ~14.6 (bare GaAs on wafer) to
~22.3 mA/cm
2
(TiO2 NPs after printing) with a corresponding solar-to-electric power conversion
efficiency of ~17.2% from 420-nm-thick single-junction GaAs solar cells.

Figure 3.9: Performance enhanced by bi-facial photon management.

68

Figure 3.10: Schematics of the bi-facial photon management.

3.3 Conclusion
In conclusion, we demonstrated the growth and fabrication of multilayer-grown ultrathin
GaAs solar cells having uniform photovoltaic efficiencies between device stacks grown in
different sequences. Ultrathin GaAs solar cells with 300-nm-thick absorber layer were grown by
MBE using beryllium as a p-type impurity to address prior difficulties in multilayer-grown GaAs
solar cells, including dopant diffusion, defect generation from prolonged heat treatments, and
associated performance degradation in early-grown devices. The triple-stack ultrathin GaAs
microcells on the growth wafer exhibited comparable contact properties and uniform
photovoltaic device characteristics as a result of the suppressed dopant diffusion and preservation
of optimized electronic configuration during the multilayer epitaxy, consistent with SIMS and
IQE/EQE measurements. The bifacial photon management using ex-situ deposited TiO 2 NPs and

69

a specular BSR in a vertical contact design, as well as the specialized epitaxial design for
multi-pass light trapping collectively contributed to the significant increase of the absorption of
optically thin GaAs solar cells, where 17.2% solar-to-electric conversion efficiency was achieved
in 420-nm-thick GaAs solar cells.
Materials and fabrication strategies presented here will constitute an important advance in
the continuing efforts of making III-V solar cells more practical for terrestrial photovoltaics. In
future work, the p-type doping in the base can be minimized or eliminated to further reduce
unintentional diffusion while switching to carbon doping in the BSF and bottom contact layers.
Such a scheme would minimize the defect issues associated with carbon doping in active regions,
while taking advantage of the low diffusivity of carbon in the contact regions. Continued
evolution of the device design and doping may enable deposition of dozens of high-efficiency,
ultra-thin device stacks to be grown in a single run by either MBE or MOCVD, and will
effectively reduce the substrate cost to zero.

70

3.4 References
1. Woodhouse, M.; Goodrich, A., A Manufacturing Cost Analysis Relevant to Single- and
Dual-Junction Photovoltaic Cells Fabricated with III-Vs and III-Vs Grown on
Czochralski Silicon. NREL report 2014, PR-6A20-60126, 92.
2. Horowitz, K. A. W.; Woodhouse, M.; Lee, H.; Smestad, G. P., A Bottom-Up Cost
Analysis of a High Concentration PV Module. Aip Conf Proc 2015, 1679.
3. Dimroth, F., High-efficiency solar cells from III-V compound semiconductors. Phys
Status Solidi C 2006, 3, 373-379.
4. Konagai, M.; Sugimoto, M.; Takahashi, K., High efficiency GaAs thin film solar cells by
peeled film technology. Journal of crystal growth 1978, 45, 277-280.
5. Yablonovitch, E.; Gmitter, T.; Harbison, J. P.; Bhat, R., Extreme selectivity in the lift-off
of epitaxial GaAs films. Appl. Phys. Lett. 1987, 51, 2222-2224.
6. Yablonovitch, E.; Hwang, D. M.; Gmitter, T. J.; Florez, L. T.; Harbison, J. P., Van Der
Waals bonding of GaAs epitaxial liftoff flims onto arbitrary substrates. Appl. Phys. Lett.
1990, 56, 2419-2421.
7. Cheng, C. W.; Shiu, K. T.; Li, N.; Han, S. J.; Shi, L.; Sadana, D. K., Epitaxial lift-off
process for gallium arsenide substrate reuse and flexible electronics. Nat. Commun. 2013,
4, 1577.
8. Bauhuis, G. J.; Mulder, P.; Haverkamp, E. J.; Schermer, J. J.; Bongers, E.; Oomen, G.;

71

Köstler, W.; Strobl, G., Wafer reuse for repeated growth of III–V solar cells. Progress in
Photovoltaics: Research and Applications 2010, 18, 155-159.
9. Yoon, J.; Jo, S.; Chun, I. S.; Jung, I.; Kim, H. S.; Meitl, M.; Menard, E.; Li, X. L.;
Coleman, J. J.; Paik, U.; Rogers, J. A., GaAs photovoltaics and optoelectronics using
releasable multilayer epitaxial assemblies. Nature 2010, 465, 329-U80.
10. Kang, D. S.; Arab, S.; Cronin, S. B.; Li, X. L.; Rogers, J. A.; Yoon, J., Carbon-doped
GaAs single junction solar microcells grown in multilayer epitaxial assemblies. Appl.
Phys. Lett. 2013, 102, 253902.
11. Yu, S.; Tan, T. Y .; Gosele, U., Diffusion mechanism of zinc and beryllium in GaAs.
Journal of Applied Physics 1991, 69, 3547-3565.
12. Cunningham, B. T.; Guido, L. J.; Baker, J. E.; Major, J. S.; Holonyak, N.; Stillman, G. E.,
Carbon diffusion in undoped, n-type, and p-type GaAs. Appl. Phys. Lett. 1989, 55,
687-689.
13. Watanabe, K.; Yamazaki, H., Annealing effect on the electrical properties of heavily
C-doped p+GaAs. Appl. Phys. Lett. 1991, 59, 434-436.
14. Fushimi, H.; Wada, K., Carbon-related defects in carbon-doped GaAs by
high-temperature annealing. Journal of Applied Physics 1997, 82, 1208-1213.
15. Delyon, T. J.; Woodall, J. M.; Goorsky, M. S.; Kirchner, P. D., Lattice contraction due to
carbon doping of GaAs grown by metalorganic molecular beam epitaxy. Appl. Phys. Lett.

72

1990, 56, 1040-1042.
16. Pan, N.; Bose, S. S.; Kim, M. H.; Stillman, G. E.; Chambers, F.; Devane, G.; Ito, C. R.;
Feng, M., Hydrogen passivation of C acceptors in high-purity GaAs. Appl. Phys. Lett.
1987, 51, 596-598.
17. Rahbi, R.; Pajot, B.; Chevallier, J.; Marbeuf, A.; Logan, R. C.; Gavand, M., Hydrogen
diffusion and acceptor passivation in p-type GaAs. Journal of Applied Physics 1993, 73,
1723-1731.
18. Vandamme, N.; Chen, H.-L.; Gaucher, A.; Behaghel, B.; Lemaitre, A.; Cattoni, A.;
Dupuis, C.; Bardou, N.; Guillemoles, J. F.; Collin, S., Ultrathin GaAs Solar Cells With a
Silver Back Mirror. IEEE J. Photovolt. 2015, 5, 565-570.
19. Lee, S. M.; Kwong, A.; Jung, D.; Faucher, J.; Biswas, R.; Shen, L.; Kang, D.; Lee, M. L.;
Yoon, J., High Performance Ultrathin GaAs Solar Cells Enabled with Heterogeneously
Integrated Dielectric Periodic Nanostructures. Acs Nano 2015, 9, 10356-10365.
20. Yang, W.; Becker, J.; Liu, S.; Kuo, Y .-S.; Li, J.-J.; Landini, B.; Campman, K.; Zhang,
Y .-H., Ultra-thin GaAs single-junction solar cells integrated with a reflective back
scattering layer. J. Appl. Phys. 2014, 115, 203105.
21. Ilegems, M., Beryllium doping and diffusion in molecular‐beam epitaxy of GaAs and
AlxGa1−xAs. Journal of Applied Physics 1977, 48, 1278-1287.
22. Kohnke, G. E.; Koch, M. W.; Wood, C. E. C.; Wicks, G. W., Beryllium diffusion in

73

GaAs/AlGaAs single-quantum-well separate-confinement heterostructure laser active
regions. Appl. Phys. Lett. 1995, 66, 2786-2788.
23. Kang, D.; Lee, S. M.; Li, Z. W.; Seyedi, A.; O'Brien, J.; Xiao, J. L.; Yoon, J., Compliant,
Heterogeneously Integrated GaAs Micro-VCSELs towards Wearable and Implantable
Integrated Optoelectronics Platforms. Adv. Opt. Mater. 2014, 2, 373-381.
24. Chan, L.; Kang, D.; Lee, S.-M.; Li, W.; Hunter, H.; Yoon, J., Broadband antireflection
and absorption enhancement of ultrathin silicon solar microcells enabled with
density-graded surface nanostructures. Applied Physics Letters 2014, 104, 223905.
25. Lee, S.-M.; Biswas, R.; Li, W.; Kang, D.; Chan, L.; Yoon, J., Printable nanostructured
silicon solar cells for high-performance, large-area flexible photovoltaics. ACS nano 2014,
8, 10507-10516.
26. Kang, D.; Lee, S.-M.; Kwong, A.; Yoon, J., Dramatically Enhanced Performance of
Flexible Micro-VCSELs via Thermally Engineered Heterogeneous Composite
Assemblies. Advanced Optical Materials 2015, 3, 1072.
27. Kang, D.; Gai, B.; Thompson, B.; Lee, S.-M.; Malmstadt, N.; Yoon, J., Flexible
Opto-Fluidic Fluorescence Sensors Based on Heterogeneously Integrated Micro-VCSELs
and Silicon Photodiodes. ACS Photonics 2016, 3, 912-918.
28. Reeves, G. K.; Harrison, H. B., Obtaining the Specific Contact Resistance from
Transmission-Line Model Measurements. Electron Devic Lett 1982, 3, 111-113.

74

29. Tuck, B.; Kadhim, M. A. H., Anomalous diffusion profiles of Zinc in GaAs. Journal of
Materials Science 1972, 7, 581-591.
30. van Ommen, A. H., Examination of models for Zn diffusion in GaAs. Journal of Applied
Physics 1983, 54, 5055-5058.

75

Chapter 4. Ten-fold stack multilayer-grown nanomembrane GaAs
solar cells
4.1 Introduction
Gallium arsenide (GaAs) is one of the most promising materials candidates for high
efficiency photovoltaic systems owing to a number of unique advantages including direct
bandgap, ideal absorption band against solar spectrum, superior photo-physical properties, as
well as established growth technology for single-crystalline materials
1-5
. While GaAs-based
single-junction solar cells currently have the record-high efficiency and promise extremely low
power-to-weight ratio, the prohibitive cost of preparing device-quality epitaxial materials has
prevented them from being widely deployed in residential photovoltaic applications where
silicon-based solar cells are currently predominant
6, 7
. Given such compelling advantages,
tremendous research efforts have been devoted over the past decades for identifying alternative
routes to economically prepare high quality GaAs solar cells
8-10
. Among various approaches
pursued including epitaxial liftoff (ELO)
8, 11
and hydride vapor phase epitaxy (HVPE)
12-14
,
growing multiple layers of solar cells on a single growth substrate, in conjunction with sequential
liftoff via transfer printing, has been of special interest due to its ability to circumvent critical
limitations of conventional ELO, where an excessively large number (e.g. >500) of substrate
reuses is mandatory to attain cost-competitiveness
7
. Critically, multilayer epitaxy has potential
to significantly lower the cost of materials growth by aggressively reducing the contribution

76

from equipment depreciation; the time-consuming load-unload procedure is performed just once
for many device growths, which is not achievable in conventional ELO or HVPE. In the
pioneering work by Yoon and co-workers, triple-stack GaAs solar cells grown by metal organic
vapor phase epitaxy (MOVPE) were investigated to examine the feasibility of multilayer epitaxy,
where zinc and carbon were used as the p-type dopant for n-on-p and p-on-n type single-junction
GaAs solar cells, respectively
15, 16
. In both studies, moderate performance degradation in the
middle and bottom layer devices was observed primarily because of the diffusion of p-type
dopant (for Zn-doped) and increased levels of point defects (for C-doped), respectively. More
recently, Gai et al introduced ultrathin (emitter + base: ~300 nm) device configuration into
multilayer epitaxy to minimize the adverse effects of the extended thermal soaking in
early-grown materials and therefore allow excellent uniformity (< 3% relative) of photovoltaic
performance
17
. Despite such optically-thin absorber layers, bifacial nanophotonic light
management employing hexagonally periodic TiO2 nanoposts and vertical p-type metal contact
serving as a back-surface reflector enabled 17.2% one-sun efficiency from 420-nm-thick
single-junction GaAs solar cells
17, 18
. Motivated by the successful outcome, herein we
investigated tenfold-stack (i.e. composed of 10 device layers) ultrathin GaAs solar cells for the
first time to assess the possibility to accommodate massive numbers (e.g. ~100) of device stacks
in multilayer epitaxy and to elucidate the evolution of materials properties and resulting device
performance under the prolonged growth conditions. In the following, we present systematic

77

studies of electrical, optical, and morphological properties of multilayer-grown ultrathin GaAs
solar cells in tenfold-stack epitaxial assemblies, together with thermally-driven changes in
dopant profiles as well as photovoltaic performance characteristics, revealing the underlying
materials science and design rules for the reported system.

4.2 Results and discussion
Tenfold-stack, n-on-p type ultrathin GaAs solar cells were grown on a (100) GaAs
substrate by molecular beam epitaxy (MBE)
17
. A single-junction GaAs solar cell at each device
stack consists of n
+
-GaAs top contact (200 nm, Si-doped, 5 × 10
18
cm
-3
), n-Al0.4Ga0.6As window
(20 nm, Si-doped, 2 × 10
18
cm
-3
), n-GaAs emitter (50 nm, Si-doped, 2 × 10
18
cm
-3
), p-GaAs base
(250 nm, Be-doped, 3 × 10
17
cm
-3
), p
+
-Al0.4Ga0.6As back surface field (BSF) (50 nm, Be-doped,
5 × 10
18
cm
-3
), and p
+
-GaAs bottom contact (50 nm, Be-doped, 1 × 10
19
cm
-3
) (Figure 4.1a).
Cross-sectional scanning electron microscope (SEM) image (Figure 4.1b) shows defect-free
epitaxial layers of tenfold-stack GaAs solar cells, where 400-nm-thick sacrificial layers
(Al0.90Ga0.10As) were inserted between respective device stacks to facilitate their sequential
lift-off from the growth substrate and printing-enabled integration on foreign substrates
17
. In
total, the materials growth took ~20 hours, meaning that the first-grown device stack experienced
thermal soaking of ~18 hours at 610 C.

78

Figure 4.1: (a) epitaxy design (b) cross-sectional SEM for 10-fold stack GaAs ultra-thin solar cells, (c) TLM results
for n-type and (d) p-type contact, respectively.

Implementing ohmic or rectifying metal contact is a key requirement for high efficiency
photovoltaic devices
19
. Accordingly, contact properties such as contact resistance (Rc) and
specific contact resistivity ( c) were first examined by a standard transmission length method
(TLM) in n-type (Pd/Ge/Au) and p-type (Pt/Ti/Pt/Au) metal contacts made on 1
st
, 5
th
, 7
th
and 10
th

device stacks
17, 20
. For characterization of early-grown (i.e. 2
nd
– 10
th
) device stacks, overlying
epitaxial layers were intentionally removed by wet chemical etching as in our previous works
17
.

79

As summarized in Figure 4.1c, d, and Table 4.1, perfect ohmic characteristics were obtained for
both n- and p-type contacts in all device stacks after the thermal annealing (175 C, 1 h under N2).
N-type metal contacts exhibited a comparable range of contact resistance and specific contact
resistivity among different device stacks, while p-type contact properties progressively degraded
from 1
st
to 10
th
layers due to the out-diffusion of beryllium during the growth of overlying stacks
and resultant decrease of effective dopant concentration in the p
+
-GaAs contact layer.
1
st
5
th
7
th
10
th

n-type
Rc (Ω) 3.8 5.6 3.1 5.8
ρc
(Ω cm
2
)
1.30 × 10
-4
1.80 × 10
-4
4.90 × 10
-5
1.80 × 10
-4

p-type
Rc (Ω) 108.3 303.7 389.2 347.1
ρc
(Ω cm
2
)
8.60 × 10
-4
4.00 × 10
-3
5.30 × 10
-3
4.50 × 10
-3

Table 4.1: TLM Results.

To further elucidate the effect of multilayer epitaxy on the evolution of materials
properties, dopant profiles of tenfold-stack GaAs solar cells were studied by secondary ion mass
spectrometry (SIMS) with as-grown epitaxial materials before further processing. In multilayer
epitaxy, early-grown device stacks are subject to the extended heat treatment during the growth
of overlying layers, resulting in thermally activated diffusion of dopants and noticeable deviation
from the optimized electronic configuration
15, 17
. One important question is whether the

80

diffusion-related degradation will continually proceed with the increasing number of device
stacks grown. Figure 4.2a and b show SIMS concentration profiles of silicon and beryllium
obtained from 1
st
(red), 5
th
(blue), 7
th
(green), and 10
th
(orange) device stacks, respectively, where
the data from 5
th
, 7
th
, and 10
th
layers were rearranged on the same depth (x-axis) coordinate as
the 1
st
layer (i.e. all profiles start from the top contact layer) such that the spatial variation of
dopant concentration at constituting layers of solar cells can be easily compared among different
device stacks. The thickness-averaged concentrations of silicon and beryllium extracted from the
SIMS data are shown in Figure 4.2c and d, respectively. The average concentration of silicon
in the top contact layer (i.e. n
+
-GaAs) was maintained nearly constant throughout all device
stacks, consistent with the uniform n-type contact properties (Figure 4.1c). The slight increase
of silicon concentration in the base layer of early-grown stacks attests some tendency for silicon
diffusion in GaAs under the current growth condition of MBE. As expected, the extent of
dopant migration was more pronounced with beryllium due to its high diffusion coefficient
associated with substitutional-interstitial diffusion mechanism
21-23
, while the profile change is
comparatively less severe than zinc
22
. Consequently, the concentration of beryllium in the
bottom contact layer substantially decreased in long annealed stacks, while it increased in the
base and emitter. The “pile-up” of beryllium near the junction seen in Figure 4.2c is caused by
lower diffusivity of beryllium in n-GaAs due to the electric field effect with the positively
charged beryllium interstitials, similar to what was reported by Enquist et al
24
. The reduced

81

beryllium concentration in the bottom contact layer directly translated to the degradation of
p-type contact properties in early-grown device stacks. It is also noteworthy that the rate of
concentration variation was largest between 1
st
and 2
nd
device stacks (Figure 4.3), while the
layer-to-layer discrepancy gradually diminished in long-annealed stacks, which is understandable
given that the concentration gradient becomes progressively smaller as the diffusion proceeds.
The more uniform beryllium profile in long-annealed device stacks may also be partially
explained by the phenomenon of decreasing diffusivity of beryllium with annealing time,
associated with the change of non-equilibrium point defect concentrations as observed
previously in MBE-grown GaAs
25
.

Figure 4.2: SIMS results for 1
st
, 5
th
, 7
th
and 10
th
.

82

Figure 4.3: SIMS results for 1
st
~ 5
th
.

Figure 4.4a shows representative current density (J)-voltage (V) curves of solar cells
fabricated from 1
st
, 5
th
, 7
th
, and 10
th
device stacks, measured on the wafer under simulated
AM1.5G solar illumination (1000 W/m
2
). To evaluate photovoltaic performance, microscale
(~500 × 500 m
2
) solar cells (i.e. microcells) with a ‘recessed’ bottom contact were fabricated on
the wafer using previously reported procedures after the removal of overlying device stacks
15-17
.
The average device characteristics including short-circuit current density (Jsc), open-circuit
voltage (Voc), fill factor (FF), and solar-to-electric energy conversion efficiency ( ) are also

83

summarized in Figure 4.4b and Table 4.2. As expected, the efficiency was highest at the top (1
st
)
device stack that did not undergo any post-growth annealing, but slightly degraded in all
early-grown layers, where the difference between 1
st
and 10
th
stacks is ~11% (relative). Based on
our previous works
17, 18
, these cells would be expected to yield ~14-16% efficiency with bifacial
nanoscale photon management. Notably, such performance degradation was not continuously
aggravated but rather saturated with the increase of thermal soaking at early-grown stacks,
matching with the trend in dopant profiles. The decrease in Voc suggests that the degree of carrier
recombination increased in early-grown materials possibly due to the unintentional degradation
of ultrathin (~20 nm) Al0.40Ga0.60As window layer (e.g. during the etching of top contact layer).
Slightly reduced fill factor might be attributed to the increased series resistance associated with
the elevation of emitter resistance associated with the carrier compensation due to the beryllium
diffusion as well as the degraded p-type contact properties. Such characteristics of Voc and FF are
also captured in dark IV measurements as summarized in Figure 4.4c, d, Table 4.3, and Figure
4.5. Reverse bias saturation current (I0) of solar cells significantly increased at 5
th
, 7
th
, 10
th
layers
comparing to the 1
st
layer device, consistent with the degradation of Voc. It is also shown that
series resistance increased while shunt resistance decreased at early-grown devices, supporting
the degrading trend of FF. While the lower performance of early-grown devices must be
addressed in future work, the similar characteristics of early-grown device stacks is promising
for further scale-up to enable ultralow costs.

84

Figure 4.4: I-V data for PV performance and dark.

1
st
5
th
7
th
10
th

Avg.
Err.+/
Avg.
Err.+/
Avg.
Err.+/
Avg.
Err.+/
Err.- Err.- Err.- Err.-
Voc (V) 0.952
0.006/
0.914
0.004/
0.908
0.010/
0.919
0.005/
0.011 0.006 0.011 0.005
FF 0.79
0.006/
0.755
0.004/
0.751
0.010/
0.767
0.008/
0.005 0.004 0.006 0.009

85

Jsc
(mA/cm
2
)
11.2
(11.2)
0.2/0.2
11.8
(11.9)
0.1/0.1
11.1
(11.2)
0.1/0.1
10.7
(10.8)
0.1/0.1
 (%)  8.5 0.2/0.1 8.2 0.1/0.1 7.6 0.2/0.2 7.5 0.1/0.1
Table 4.2: I-V data for photovoltaic performance.

1
st
5
th
7
th
10
th

Avg.
Err.+/
Avg.
Err.+/
Avg.
Err.+/
Avg.
Err.+/
Err.- Err.- Err.- Err.-
- log
I0(A)
13.1 0.2/0.2 11.4 0.2/0.4 11.5 0.3/0.4 11.2 0.5/0.2
n 1.919
0.038/
2.243
0.094/
2.212
0.101/
2.337
0.070/
0.05 0.057 0.071 0.133
Rs (Ω) 297.8 3.7/4.6 645.4
104.9/
616.6 20.4/20.0 408 24.5/11.2
65.3
log Rsh
(Ω∙cm
2
)
8 0.2/0.3 6.9 0.4/0.6 7 0.3/0.4 6.6 0.3/0.3
Table 4.3: Dark I-V data.

86

Figure 4.5: Curve fitting of dark IV results.

To provide additional insight into the origin of performance degradation, internal (IQE)
and external quantum efficiencies (EQE) were measured on the wafer with 5
th
, 7
th
, and 10
th

device stacks (Figure 4.6a), where the short-circuit current densities derived from QE data
matched well with the average values obtained from PV measurements (Figure 4.6b). Similar
to the observation in JV characteristics, the QE spectra also exhibited systematic degradation at

87

early-grown stacks, while the discrepancy was most pronounced in the shorter wavelength range
below ~450 nm, supporting the relevance of performance degradation to the deteriorated
materials properties in window and emitter layers.

Figure 4.6: Quantum efficiency results.

Besides, since all the samples are prepared on their epi-growth substrate but not
transfer-printed, the substrate condition is not exact the same. Specifically, the sample prepared
for 1
st
layer stack has 9 repeating layers beneath it, and the one for 2
nd
layer has 8, etc. Because
the cell is ultrathin, the condition of the substrate may affect the PV performance. In order to
study the effect of the substrate, an optical simulation was conducted using Lumerical
TM
FDTD
software. Related scripts are in appendices A.1, and the parameters used are referred from this
website
26
. The results for the absorption and reflection are in Figure 4.7. As we can see, even the
GaAs is ultrathin, still most of the light within the bandgap is absorbed, and thus make the light
reach to the substrate quite small, so there is only small difference near the bandgap. On the other

88

hand, the reflectance out of the bandgap fluctuates drastically due to the interference, but it will
not affect our performance.

Figure 4.7: Simulated absorption and reflection for each layers.

Figure 4.8 is the comparison for 1
st
, 5
th
, 7
th
and 10
th
layers between simulated data and
the experimental data. The trend of the peaks fitted well, where the peak position difference
could be because of the actual thickness variation. Besides, the interference caused sharp peak
from the simulation data also agrees with the experimental data, and it is not observed in the 10
th

layer which beneath layer causing the interference is absent. To sum up, although the substrate
condition can affect the results a little, and the simulation fits with the experiments quite well, it
cannot explain the difference for the performance for each stack.

89

Figure 4.8: Comparasion between experiment and simulation data of reflection for 1
st
, 5
th
, 7
th
and 10
th
.

4.3 Conclusion
In conclusion, we systematically studied the evolution of materials properties and device
performance in tenfold-stack ultrathin GaAs solar cells grown by MBE, which is the largest
number of epitaxially grown solar cell stacks that have been explored. The one-sun
photovoltaic efficiency of solar cells exhibited promising uniformity even in ten stacks, only
with slight degradation by ~11% (relative) between 1
st
and 10
th
device stacks, while the

90

discrepancy between stacks gradually diminished at early-grown epitaxial stacks, consistent with
the saturation behavior in beryllium diffusion associated with the decrease of concentration
gradient and beryllium diffusivity at long annealing times. Since virtually all of the device
degradation can be linked to the motion of beryllium from the heavily doped contact layer into
the base, emitter, and window, a switch to carbon in the p-type contact and back-surface field
should greatly improve the uniformity of performance; the carbon will not be present in the
active regions of the device, and thus challenges associated with carbon-related point defects will
be minimized
16
. Future work should also focus on the use of higher growth rates to significantly
reduce the time spent at high-temperature for early-grown device layers. These two measures
will enable more uniform device performance, and thus very large numbers of device stacks
beyond the ten shown here. We believe this study will serve as a firm foundation for further
development of this technology towards cost-effective GaAs solar cells.

91

4.4 References
1. Bett, A.; Dimroth, F.; Stollwerck, G.; Sulima, O., III-V compounds for solar cell
applications. Applied Physics A 1999, 69, 119-129.
2. Bosi, M.; Pelosi, C., The potential of III‐V semiconductors as terrestrial photovoltaic
devices. Progress in Photovoltaics: Research and Applications 2007, 15, 51-68.
3. Geisz, J.; Friedman, D., III–N–V semiconductors for solar photovoltaic applications.
Semiconductor Science and Technology 2002, 17, 769.
4. Goetzberger, A.; Hebling, C.; Schock, H.-W., Photovoltaic materials, history, status and
outlook. Materials Science and Engineering: R: Reports 2003, 40, 1-46.
5. Blakemore, J., Semiconducting and other major properties of gallium arsenide. Journal of
Applied Physics 1982, 53, R123-R181.
6. Green, M. A.; Hishikawa, Y .; Warta, W.; Dunlop, E. D.; Levi, D. H.; Hohl‐Ebinger, J.;
Ho‐Baillie, A. W., Solar cell efficiency tables (version 50). Progress in Photovoltaics:
Research and Applications 2017, 25, 668-676.
7. Woodhouse, M.; Goodrich, A. A manufacturing cost analysis relevant to single-and
dual-junction photovoltaic cells fabricated with III-Vs and III-Vs grown on Czochralski
silicon; National Renewable Energy Laboratory: 2013.
8. Konagai, M.; Sugimoto, M.; Takahashi, K., High efficiency GaAs thin film solar cells by
peeled film technology. Journal of crystal growth 1978, 45, 277-280.

92

9. Chopra, K. L.; Das, S. R., Why thin film solar cells? In Thin film solar cells, Springer:
1983; pp 1-18.
10. Lee, K.; Zimmerman, J. D.; Hughes, T. W.; Forrest, S. R., Non‐Destructive Wafer
Recycling for Low‐Cost Thin‐Film Flexible Optoelectronics. Advanced Functional
Materials 2014, 24, 4284-4291.
11. Yablonovitch, E.; Gmitter, T.; Harbison, J.; Bhat, R., Extreme selectivity in the lift‐off
of epitaxial GaAs films. Applied Physics Letters 1987, 51, 2222-2224.
12. Deschler, M.; Grüter, K.; Schlegel, A.; Beccard, R.; Jürgensen, H.; Balk, P., Very rapid
growth of high quality GaAs, InP and related III-V compounds. Le Journal de Physique
Colloques 1988, 49, C4-689-C4-692.
13. Simon, J.; Schulte, K. L.; Jain, N.; Johnston, S.; Young, M.; Young, M. R.; Young, D. L.;
Ptak, A. J., Upright and inverted single-junction GaAs solar cells grown by hydride vapor
phase epitaxy. IEEE Journal of Photovoltaics 2017, 7, 157-161.
14. Bozler, C. O.; Fan, J. C., High‐efficiency GaAs shallow‐homojunction solar cells.
Applied Physics Letters 1977, 31, 629-631.
15. Yoon, J.; Jo, S.; Chun, I. S.; Jung, I.; Kim, H.-S.; Meitl, M.; Menard, E.; Li, X.; Coleman,
J. J.; Paik, U., GaAs photovoltaics and optoelectronics using releasable multilayer
epitaxial assemblies. Nature 2010, 465, 329.
16. Kang, D.; Arab, S.; Cronin, S. B.; Li, X.; Rogers, J. A.; Yoon, J., Carbon-doped GaAs

93

single junction solar microcells grown in multilayer epitaxial assemblies. Applied Physics
Letters 2013, 102, 253902.
17. Gai, B.; Sun, Y .; Lim, H.; Chen, H.; Faucher, J.; Lee, M. L.; Yoon, J., Multilayer-Grown
Ultrathin Nanostructured GaAs Solar Cells as a Cost-Competitive Materials Platform for
III–V Photovoltaics. ACS nano 2017, 11, 992-999.
18. Lee, S.-M.; Kwong, A.; Jung, D.; Faucher, J.; Biswas, R.; Shen, L.; Kang, D.; Lee, M. L.;
Yoon, J., High performance ultrathin GaAs solar cells enabled with heterogeneously
integrated dielectric periodic nanostructures. ACS nano 2015, 9, 10356-10365.
19. Green, M. A., Accuracy of analytical expressions for solar cell fill factors. Solar Cells
1982, 7, 337-340.
20. Reeves, G.; Harrison, H., Obtaining the specific contact resistance from transmission line
model measurements. IEEE Electron device letters 1982, 3, 111-113.
21. Weisberg, L. R.; Blanc, J., Diffusion with Interstitial-Substitutional Equilibrium. Zinc in
GaAs. Physical Review 1963, 131, 1548-1552.
22. Ilegems, M., Beryllium doping and diffusion in molecular‐beam epitaxy of GaAs and
Al x Ga1− x As. Journal of Applied Physics 1977, 48, 1278-1287.
23. Gösele, U.; Morehead, F., Diffusion of zinc in gallium arsenide: A new model. Journal of
Applied Physics 1981, 52, 4617-4619.
24. Enquist, P.; Wicks, G. W.; Eastman, L. F.; Hitzman, C., Anomalous redistribution of

94

beryllium in GaAs grown by molecular beam epitaxy. Journal of Applied Physics 1985,
58, 4130-4134.
25. Hu, J. C.; Deal, M. D.; Plummer, J. D., Modeling the diffusion of grown‐in Be in
molecular beam epitaxy GaAs. Journal of Applied Physics 1995, 78, 1595-1605.
26. Refractive index database. https://refractiveindex.info/.

95

Chapter 5. Transfer printed 3J inverted metamorphic multijunction
(IMM) solar cells
5.1 Introduction
Inverted metamorphic (IMM) multijunction solar cells have been of intense research
interest over the past decade as a promising materials platform for ultrahigh (>50%) efficiency
photovoltaic systems (UHPVs) owing to their unique capabilities to incorporate materials with
dissimilar lattice constants into monolithic two-terminal devices
1-7
without involving the direct
bonding of two separately-prepared solar cells
8, 9
. Such greatly expanded choices of materials
allow the accommodation of more ideal bandgap (Eg) combinations in terms of absorbing solar
spectrum than those possible with lattice-matched systems, thereby promising significantly
higher theoretical efficiencies of solar cells and thus lowered levelized cost of energy
10, 11
. In
the epitaxial growth of IMM multijunction solar cells by organometallic vapor phase epitaxy
(OMVPE), optically transparent buffer layers such as GaxIn1-xP or AlxGayIn1-x-yAs with gradually
varying atomic compositions are incorporated between lattice-matched (e.g. GaAs and
Ga0.51In0.49P) and lattice-mismatched (e.g. InxGa1-xAs) subcells such that defects associated with
a lattice mismatch including misfit and threading dislocations are mostly confined within the
buffer layers while enabling the growth of strain-free lattice-mismatched subcells
4, 12, 13
. To
date, four-junction IMM devices composed of Ga0.51In0.49P (1.8 eV), GaAs (1.4 eV),
In0.28Ga0.72As (1.0 eV), and In0.57Ga0.43As (0.7 eV) subcells have demonstrated 45.7%

96

solar-to-electricity power conversion efficiency at 690 suns
2
and research towards six-junction
IMM systems consisting of AlGaInP (2.1 eV), AlGaAs (1.7 eV), GaAs (1.4 eV), InGaAsP (1.13
eV), InGaAsP (0.91 eV), and InGaAs (0.7 eV) targeting over 50% efficiency is currently
underway
14
.
In the conventional IMM cell processing technology, however, the fabrication of
functional solar cells typically involves wafer bonding of a centimeter-scale die followed by
destructive removal of an expensive growth wafer by wet chemical etching or liftoff of epitaxial
layers, thereby imposing unavoidable restrictions in minimum cell size, type of module substrate,
or spatial layout of cells but also deteriorating the cost effectiveness
1, 15-17
. In this regard,
transfer printing, pioneered by Rogers and coworkers, has been successfully demonstrated to be
a manufacturable pathway for silicon and III-V solar cells with dimensions smaller than a
millimeter (e.g. microscale) into one-sun as well as concentrator photovoltaic (CPV) systems
18-25
.
The resulting thin, microscale cell design offered many unique advantages in the context of CPV
including the accommodation of lightweight, low-profile concentrating optics, efficient materials
utilization, high-throughput manufacturing, as well as, facile heat dissipation at high
concentration without an active cooling system
20, 23, 26
. Despite such compelling merits,
however, microscale IMM multijunction solar cells compatible with transfer printing, which are
expected to greatly expand their application possibilities (e.g. micro-optic CPV module) but also
substantially reduce the cost of materials (e.g. by avoiding wafer dicing and reusing the growth

97

substrate), have not been demonstrated yet owing to the added complexity of processing
associated with inverted cell configuration
4, 15
. Here we report materials design and fabrication
strategies for microscale IMM triple-junction (3J) Ga0.51In0.49P (referred to as
GaInP)/GaAs/In0.26Ga0.74As (referred to as InGaAs) solar cells that allow their defect-free release
from a growth wafer and printed assemblies on non-native substrates in ways that can preserve
their growth substrate for reuse, and also enable programmable spatial layout of printed
microcells and versatile choices of module substrate, as described in detail subsequently.

5.2 Results and discussion
The fabrication process started with the deposition of metals (Cr/Au: 5 nm/100 nm) to
form a p-type ohmic contact by electron beam evaporation and lift-off (Figure 5.1).
Photolithography and successive steps of wet chemical etching delineated the mesa-shaped
junction area (~500 × 500 m
2
) until the n
+
-GaAs contact layer was exposed, where chemical
reagents based on hydrochloric acid (HCl : deionized (DI) water (H 2O) = 1:1 by volume) and
phosphoric acid (H3PO4:H2O2:H2O = 1:12:13 by volume) were used for etching phosphide (i.e.
GaxIn1-xP) and arsenide (i.e. InxGa1-xAs) layers, respectively. Subsequently, isolated arrays of
individual IMM microcells (~560 × 560 m
2
) were defined by photolithography and wet
etching, followed by the partial removal of the sacrificial layer (i.e. AlAs) and formation of
photoresist (PR) ‘anchors’ to maintain the lithographically defined cell layout and protect active

98

layers during the undercut etching of a sacrificial layer
27
. Notably, channels to deliver the
etchant to the sacrificial layer were made outside the active junction area to minimize the
exposed sidewall area that can serve as sites for surface recombination of photogenerated
carriers
28
.

Figure 5.1: Schematics of procedures for transfer printed IMM solar cells.

One of critical challenges for the undercut etching and release of 3J IMM microcells from
the growth wafer is the formation of fracture defects at the center of the microcell caused by the
residual stress and resultant bowing effect of the released cell membrane. In particular, when
etch holes were made at four-sides of cell boundary as in conventional processes of upright
devices, the etch-front propagates from the edge to the center of microcell (Figure 5.2a). Due

99

to the compressive residual stresses at the metamorphic layers with larger lattice constants, top
and bottom portions of the released microcell experience tensile and compressive strains,
respectively, to cause the concave bowing towards the front surface after the cell is released from
the growth substrate and printed (Figure 5.3). When the remaining area of sacrificial layer
becomes small near the completion of undercut etching (i.e. forming umbrella-shaped structure),
stresses are concentrated where the cell is still attached near the center of the device and result in
the fracture. To address this challenge, we purposefully designed the etch holes placed at only
two sides of the cell boundary (Figure 5.2b). In this case, the etching front proceeded
diagonally from one corner to the other such that the undercut etching completed outside the
active device area without producing any defects. The required time for undercut etching for
samples with two-side etch-holes increased by about two times compared to those with four-side
etch-holes under the present experimental condition due to the increased distance between
etch-holes.

100

Figure 5.2: Schematic illustration of etching-front propagation during the undercut etching for (a), four- and (b),
two-side etch holes.

Figure 5.3: (a) Photographic image of printed arrays of IMM microcells on glass, where the arrow indicates the
scan direction of surface profilometer. (b) Surface profile of printed IMM microcells. (c) Zoomed-in image of the
shaded region in (b).

After the completion of undercut etching in dilute hydrofluoric acid (HF:H 2O = 1:100 by
volume), IMM microcells were picked up from the growth wafer using an elastomeric stamp
made of poly(dimethylsiloxane) (PDMS) and printed onto a glass substrate by two-step printing

101

processes
28
such that the top (i.e. high bandgap) GaInP junction is placed at the front surface of
the module substrate, while the bottom (i.e. low-bandgap) InGaAs junction that was originally at
the top surface of the growth wafer was faced down and embedded in a photocurable polymer
matrix. Glass was chosen as a module substrate in this study for the proof of concept to allow
easy visualization of the rear surface of the printed cell. In most CPV systems, substrates with
higher thermal conductivity such as metals are usually employed to facilitate improved heat
dissipation under concentrated solar illumination
29, 30
Subsequently, the GaInP buffer layer was
removed by wet chemical etching, followed by the deposition of n-type ohmic metal contact,
exposure of p-type bottom contact, and deposition of ZnS/MgF 2/ZnS/MgF2 (50 nm/20 nm/15
nm/100 nm) as a broadband antireflection coating (ARC) by thermal evaporation
1, 31
. Figure
5.4a show optical micrographs of triple-junction IMM solar cells at various fabrication steps
after the deposition of p-type ohmic contact (top left), delineation of microcell layout (top right),
1
st
transfer to a temporary PDMS substrate (bottom left), and 2
nd
transfer to a module substrate
(bottom right). Figure 5.4b shows a photographic image of dense arrays of 3J IMM microcells
printed on a glass substrate before the deposition of ARC, where the spatial layout of microcells
defined on the growth wafer was maintained by transferring the entire cell array.

102

Figure 5.4: (a) optical microscopic images for each fabrication step of the transfer printed IMM solar cells and (b)
optical photographic and microscopic images of the transfer printed IMM solar cells after fabrication.

Photovoltaic performance of an individual IMM microcell printed on a glass substrate
was measured under simulated AM1.5D illumination (1000 W/m
2
) by adjusting the spectrum
with Xe-lamp and LED illumination, where the spectral mismatch correction was made using
isotype reference cells
32
. Figure 5.5 shows a representative current density (J)-voltage (V)
curve of an individual 3J IMM microcell deposited with a quadruple-stack ARC
(ZnS/MgF2/ZnS/MgF2). The average values of open-circuit voltage (Voc), short-circuit current
density (Jsc), fill factor (FF), and solar-to-electric power conversion efficiency (Eff.) were 10.7
mA/cm
2
, 2.922 V , 0.876, and 27.3%, respectively, where the active junction area except the
region for the metal busbar was considered for the calculation of current density and efficiency.
While this performance is respectable, the Voc of these devices is slightly lower than similar 3J
IMM devices processed using conventional procedures
4, 15, 33
.

103

Figure 5.5: PV data from transfer printed IMM solar cells.

The most likely degradation mechanism for these transfer printed samples is the use of
phosphoric acid (H3PO4:H2O2:H2O) to etch the n-type contact layer (i.e. GaAs) and unintentional
damage to AlInP window layer
15
, which we expect can be avoided by using an alternative
etchant (e.g. NH4OH:H2O2:H2O) for the removal of the top contact layer.
With the same set of samples for one-sun measurements, we also characterized the
performance of printed 3J IMM microcells under concentrated illumination up to ~2500 suns
using a high intensity pulsed solar simulator (HIPSS) without precise spectral control, where the
short-circuit current at one sun was employed as a reference to estimate the concentration ratio
assuming the linearity of the short-circuit current with the intensity of illumination (Figure 5.6a).
Although the performance characteristics obtained from these measurements are not as accurate
as those measured under the concentrator standard testing condition (CSTC)
31
, they provide a

104

semi-quantitative assessment of the effect of series resistance upon the cell performance under
the illumination of ultrahigh concentration. The cell performance as a function of concentration
is summarized in Figure 5.6b, c, d. As expected for a well-behaved concentrator cell, the Voc
logarithmically increased with the light intensity up to ~1330 sun concentration. The fill factor
decreased over 100 suns due to series resistance, leading to the highest efficiency of ~33.9% at
~351 sun concentration.

Figure 5.6: PV performance of transfer printed IMM solar cells under concentrated light.

To form a concentrator solar module, the released IMM microcells need to be printed

105

individually into a distributed array whose areal coverage is much larger than that on the growth
wafer. To this end, although working well for the transfer of microcells without modifying the
original footprint defined on the growth wafer, above-described fabrication schemes are not
suitable for transfer printing of individual microcells into sparse arrays with arbitrary cell layout
on the module substrate. To circumvent these difficulties, we devised an alternative processing
route that can facilitate the high-throughput transfer printing of individual IMM microcells while
preserving the growth substrate. The key aspect of the approach is to employ a temporary carrier
substrate on which fully-functional IMM microcells are formed in an upright and print-ready
configuration as a solid-state ink element for transfer printing. The surface of (001) silicon wafer
as a carrier substrate was intentionally roughened by metal-assisted wet chemical etching to
facilitate the release of microcells
34
. Subsequently, a triple-stack of metals (Cr/Ag/Cr = 5 nm/100
nm/100 nm) was deposited onto the roughened surface of a carrier substrate as a sacrificial layer
by electron-beam evaporation, followed by the spin-coating and curing of polyimide (PI, ~2.0
m) as a supporting ‘base’ layer. The released IMM microcells (Figure 5.4a) were then
transfer-printed onto the composite carrier substrate (i.e. PI/metal/Si) coated with a thin layer (~1
m) of photocurable adhesive, followed by above-described post-printing processes, including
the etching of GaInP, deposition of n-type contact, and exposure of p-type contact, to yield
fully-functional ‘upright’ IMM microcells (Figure 5.7a). Subsequently, photolithography and
dry etching of PI delineated isolated arrays of print-ready IMM microcells with a PI base layer

106

and polymeric anchor structure. After the undercut etching of metal sacrificial layer using a wet
chemical etchant (e.g. CR-7), the fully-functional IMM microcells were individually picked up
using an elastomeric stamp with relief features and printed onto a glass substrate in a sparse and
user-definable layout. Figure 5.7b shows photographic image of such 3  3 sparse arrays of 3J
IMM microcells printed on a glass substrate. The photovoltaic performance between devices on
temporary and module substrates, measured under simulated AM1.5G illumination (1000 W/m
2
)
were nearly identical (Figure 5.7c), where slight discrepancy is attributed to the wave-guided
photon flux incident outside the cell area on a metal-coated temporary substrate. Although not
shown here, these printed cell arrays can be further interconnected by processes of thin film
metallization and integrated with optics components to form a concentrated solar module of
IMM solar microcells.

Figure 5.7: (a) the schematics of the procedures of individually printed IMM as a sparse array (b) photographic
image and optical microscope image of finished sparse array (c) respective photovoltaic performance.

107

5.3 Conclusion
We demonstrated printed assemblies of microscale inverted metamorphic 3J
GaInP/GaAs/InGaAs solar cells. Specialized schemes of delineation and undercut etching
enabled the release of IMM solar microcells without destructive removal of a growth substrate
and printed assemblies on a glass substrate. The printed 3J IMM devices exhibited efficiencies
of 27.3% and 33.9% at one and 351 suns, respectively. A composite carrier substrate where
fully-functional IMM microcells were formed in print-ready forms as solid-state ink elements
allowed high-throughput transfer printing of individual IMM microcells in user-definable spatial
layout on versatile choices of module substrate. The approach presented here represents an
important step towards the realization of ultrahigh efficiency, cost-competitive CPV module with
ultrahigh efficiency IMM multijunction solar cells. Although demonstrated here with 3J IMM
systems, we expect the reported approach will be readily applicable to other IMM systems with a
larger number of junctions.

108

5.4 References
1. France, R. M.; Geisz, J. F.; Garcia, I.; Steiner, M. A.; McMahon, W. E.; Friedman, D. J.;
Moriarty, T. E.; Osterwald, C.; Ward, J. S.; Duda, A.; Young, M.; Olavarria, W. J.,
Quadruple-Junction Inverted Metamorphic Concentrator Devices. IEEE Journal of
Photovoltaics 2015, 5, 432-437.
2. France, R. M.; Geisz, J. F.; Garcia, I.; Steiner, M. A.; McMahon, W. E.; Friedman, D. J.;
Moriarty, T. E.; Osterwald, C. R.; Ward, J. S.; Duda, A.; Young, M.; Olavarria, W. J.,
Design Flexibility of Ultra-High Efficiency Four-Junction Inverted Metamorphic Solar
Cells. IEEE Journal of Photovoltaics 2016, 6, 578-583.
3. Geisz, J. F.; Friedman, D. J.; Ward, J. S.; Duda, A.; Olavarria, W. J.; Moriarty, T. E.; Kiehl,
J. T.; Romero, M. J.; Norman, A. G.; Jones, K. M., 40.8% efficient inverted
triple-junction solar cell with two independently metamorphic junctions. Applied Physics
Letters 2008, 93.
4. Geisz, J. F.; Kurtz, S. R.; Wanlass, M. W.; Ward, J. S.; Duda, A.; Friedman, D. J.; Olson, J.
M.; McMahon, W. E.; Moriarty, T.; Kiehl, J., High-efficiency GaInP/GaAs/InGaAs
triple-junction solar cells grown inverted with a metamorphic bottom junction. Appl.
Phys. Lett. 2007, 91, 023502.
5. Steiner, M. A.; Geisz, J. F.; Garcia, I.; Friedman, D. J.; Duda, A.; Kurtz, S. R., Optical
enhancement of the open-circuit voltage in high quality GaAs solar cells. Journal of

109

Applied Physics 2013, 113, 123109.
6. Law, D. C.; King, R. R.; Yoon, H.; Archer, M. J.; Boca, A.; Fetzer, C. M.; Mesropian, S.;
Isshiki, T.; Haddad, M.; Edmondson, K. M.; Bhusari, D.; Yen, J.; Sherif, R. A.; Atwater,
H. A.; Karam, N. H., Future technology pathways of terrestrial III-V multijunction solar
cells for concentrator photovoltaic systems. Solar Energy Materials and Solar Cells 2010,
94, 1314-1318.
7. Dimroth, F.; Grave, M.; Beutel, P.; Fiedeler, U.; Karcher, C.; Tibbits, T. N. D.; Oliva, E.;
Siefer, G.; Schachtner, M.; Wekkeli, A.; Bett, A. W.; Krause, R.; Piccin, M.; Blanc, N.;
Drazek, C.; Guiot, E.; Ghyselen, B.; Salvetat, T.; Tauzin, A.; Signamarcheix, T.; Dobrich,
A.; Hannappel, T.; Schwarzburg, K., Wafer bonded four-junction
GaInP/GaAs//GaInAsP/GaInAs concentrator solar cells with 44.7% efficiency. Progress
in Photovoltaics: Research and Applications 2014, 22, 277-282.
8. Chiu, P. T.; Law, D. C.; Woo, R. L.; Singer, S. B.; Bhusari, D.; Hong, W. D.; Zakaria, A.;
Boisvert, J.; Mesropian, S.; King, R. R.; Karam, N. H. In 35.8% space and 38.8%
terrestrial 5J direct bonded cells, 2014 IEEE 40th Photovoltaic Specialist Conference
(PVSC), 8-13 June 2014; 2014; pp 0011-0013.
9. Essig, S.; Allebé, C.; Remo, T.; Geisz, J. F.; Steiner, M. A.; Horowitz, K.; Barraud, L.;
Ward, J. S.; Schnabel, M.; Descoeudres, A.; Young, David L.; Woodhouse, M.; Despeisse,
M.; Ballif, C.; Tamboli, A., Raising the one-sun conversion efficiency of III–V/Si solar

110

cells to 32.8% for two junctions and 35.9% for three junctions. Nature Energy 2017, 2,
17144.
10. Cotal, H.; Fetzer, C.; Boisvert, J.; Kinsey, G.; King, R.; Hebert, P.; Yoon, H.; Karam, N.,
III–V multijunction solar cells for concentrating photovoltaics. Energy & Environmental
Science 2009, 2, 174-192.
11. Stan, M.; Aiken, D.; Cho, B.; Cornfeld, A.; Ley, V .; Patel, P.; Sharps, P.; Varghese, T.,
High-efficiency quadruple junction solar cells using OMVPE with inverted metamorphic
device structures. Journal of Crystal Growth 2010, 312, 1370-1374.
12. King, R. R.; Law, D. C.; Edmondson, K. M.; Fetzer, C. M.; Kinsey, G. S.; Yoon, H.;
Sherif, R. A.; Karam, N. H., 40% efficient metamorphic GaInP∕ GaInAs∕ Ge
multijunction solar cells. Applied Physics Letters 2007, 90, 183516.
13. Schulte, K. L.; France, R. M.; Guthrey, H. L.; Geisz, J. F., Strategies for Thinning Graded
Buffer Regions in Metamorphic Solar Cells and Performance Tradeoffs. IEEE Journal of
Photovoltaics 2018, 8, 1349-1354.
14. Geisz, J. F.; Steiner, M. A.; Jain, N.; Schulte, K. L.; France, R. M.; McMahon, W. E.; Perl,
E. E.; Friedman, D. J., Building a Six-Junction Inverted Metamorphic Concentrator Solar
Cell. IEEE Journal of Photovoltaics 2018, 8, 626-632.
15. Duda, A.; Ward, S.; Young, M. Inverted Metamorphic Multijunction (IMM) Cell
Processing Instructions; NREL/TP-5200-54049; TRN: US201206%%57 United States

111

10.2172/1036035 TRN: US201206%%57 NREL English; ; National Renewable Energy
Lab. (NREL), Golden, CO (United States): 2012; p Medium: ED; Size: 9 pp.
16. Adams, J.; Elarde, V .; Hains, A.; Stender, C.; Tuminello, F.; Youtsey, C.; Wibowo, A.;
Osowski, M., Demonstration of multiple substrate reuses for inverted metamorphic solar
cells. IEEE Journal of Photovoltaics 2013, 3, 899-903.
17. Tatavarti, R.; Wibowo, A.; Martin, G.; Tuminello, F.; Youtsey, C.; Hillier, G.; Pan, N.;
Wanlass, M. W.; Romero, M. In InGaP/GaAs/InGaAs inverted metamorphic (IMM) solar
cells on 4″ epitaxial lifted off (ELO) wafers, 2010 35th IEEE Photovoltaic Specialists
Conference, 20-25 June 2010; 2010; pp 002125-002128.
18. Yoon, J.; Baca, A. J.; Park, S. I.; Elvikis, P.; Geddes, J. B.; Li, L. F.; Kim, R. H.; Xiao, J.
L.; Wang, S. D.; Kim, T. H.; Motala, M. J.; Ahn, B. Y .; Duoss, E. B.; Lewis, J. A.; Nuzzo,
R. G.; Ferreira, P. M.; Huang, Y . G.; Rockett, A.; Rogers, J. A., Ultrathin silicon solar
microcells for semitransparent, mechanically flexible and microconcentrator module
designs. Nature Materials 2008, 7, 907-915.
19. Yoon, J.; Jo, S.; Chun, I. S.; Jung, I.; Kim, H. S.; Meitl, M.; Menard, E.; Li, X. L.;
Coleman, J. J.; Paik, U.; Rogers, J. A., GaAs photovoltaics and optoelectronics using
releasable multilayer epitaxial assemblies. Nature 2010, 465, 329-U80.
20. Furman, B.; Menard, E.; Gray, A.; Meitl, M.; Bonafede, S.; Kneeburg, D.; Ghosal, K.;
Bukovnik, R.; Wagner, W.; Gabriel, J.; Seel, S.; Burroughs, S.; Ieee, A HIGH

112

CONCENTRATION PHOTOVOLTAIC MODULE UTILIZING MICRO-TRANSFER
PRINTING AND SURFACE MOUNT TECHNOLOGY . 35th IEEE Photovoltaic
Specialists Conference 2010, 475-480.
21. Gai, B.; Sun, Y .; Lim, H.; Chen, H.; Faucher, J.; Lee, M. L.; Yoon, J., Multilayer-Grown
Ultrathin Nanostructured GaAs Solar Cells as a Cost-Competitive Materials Platform for
III–V Photovoltaics. ACS Nano 2017, 11, 992-999.
22. Lee, S.-M.; Kwong, A.; Jung, D.; Faucher, J.; Biswas, R.; Shen, L.; Kang, D.; Lee, M. L.;
Yoon, J., High performance ultrathin GaAs solar cells enabled with heterogeneously
integrated dielectric periodic nanostructures. ACS nano 2015, 9, 10356-10365.
23. Sheng, X.; Bower, C. A.; Bonafede, S.; Wilson, J. W.; Fisher, B.; Meitl, M.; Yuen, H.;
Wang, S.; Shen, L.; Banks, A. R.; Corcoran, C. J.; Nuzzo, R. G.; Burroughs, S.; Rogers, J.
A., Printing-based assembly of quadruple-junction four-terminal microscale solar cells
and their use in high-efficiency modules. Nature Materials 2014, 13, 593-598.
24. Lumb, M. P.; Meitl, M.; Schmieder, K. J.; Gonzalez, M.; Mack, S.; Yakes, M. K.; Bennett,
M. F.; Frantz, J.; Steiner, M. A.; Geisz, J. F.; Friedman, D. J.; Slocum, M. A.; Hubbard, S.
M.; Fisher, B.; Burroughs, S.; Walters, R. J. In Towards the ultimate multi-junction solar
cell using transfer printing, 2017 IEEE 44th Photovoltaic Specialist Conference, PVSC
2017, 2018; pp 1-6.
25. Gai, B.; Sun, Y.; Chen, H.; Lee, M. L.; Yoon, J., 10-Fold-Stack Multilayer-Grown

113

Nanomembrane GaAs Solar Cells. ACS Photonics 2018, 5, 2786-2790.
26. Ghosal, K.; Burroughs, S.; Heuser, K.; Setz, D.; Garralaga-Rojas, E., Performance results
from micro-cell based high concentration photovoltaic research development and
demonstration systems. Prog. Photovoltaics 2013, 21, 1370-1376.
27. Kang, D.; Lee, S. M.; Li, Z. W.; Seyedi, A.; O'Brien, J.; Xiao, J. L.; Yoon, J., Compliant,
Heterogeneously Integrated GaAs Micro-VCSELs towards Wearable and Implantable
Integrated Optoelectronics Platforms. Adv Opt Mater 2014, 2, 373-381.
28. Kang, D.; Young, J. L.; Lim, H.; Klein, W. E.; Chen, H.; Xi, Y .; Gai, B.; Deutsch, T. G.;
Yoon, J., Printed assemblies of GaAs photoelectrodes with decoupled optical and reactive
interfaces for unassisted solar water splitting. Nature Energy 2017, 2, 17043.
29. Furman, B.; Menard, E.; Gray, A.; Meitl, M.; Bonafede, S.; Kneeburg, D.; Ghosal, K.;
Bukovnik, R.; Wagner, W.; Gabriel, J.; Seel, S.; Burroughs, S. In A high concentration
photovoltaic module utilizing micro-transfer printing and surface mount technology,
2010 35th IEEE Photovoltaic Specialists Conference, 20-25 June 2010; 2010; pp
000475-000480.
30. Sheng, X.; Bower, C. A.; Bonafede, S.; Wilson, J. W.; Fisher, B.; Meitl, M.; Yuen, H.;
Wang, S.; Shen, L.; Banks, A. R.; Corcoran, C. J.; Nuzzo, R. G.; Burroughs, S.; Rogers, J.
A., Printing-based assembly of quadruple-junction four-terminal microscale solar cells
and their use in high-efficiency modules. Nature Materials 2014, 13, 593.

114

31. Aiken, D. J., High performance anti-reflection coatings for broadband multi-junction
solar cells. Solar Energy Materials and Solar Cells 2000, 64, 393-404.
32. Osterwald, C. R.; Emery, K. A.; Myers, D. R.; Hart, R. E. In Primary reference cell
calibrations, at SERI: history and methods, IEEE Conference on Photovoltaic Specialists,
21-25 May 1990; 1990; pp 1062-1067 vol.2.
33. Geisz, J. F.; Steiner, M. A.; García, I.; France, R. M.; Friedman, D. J.; Kurtz, S. R.,
Implications of redesigned, high-radiative-efficiency GaInP junctions on III-V
multijunction concentrator solar cells. IEEE Journal of Photovoltaics 2015, 5, 418-424.
34. Chan, L.; Kang, D.; Lee, S.-M.; Li, W.; Hunter, H.; Yoon, J., Broadband antireflection
and absorption enhancement of ultrathin silicon solar microcells enabled with
density-graded surface nanostructures. Applied Physics Letters 2014, 104, 223905.

115

Chapter 6. Future work
6.1 Fully-depleted ultrathin GaAs solar cells
Even that the diffusion for beryllium is better than zinc, the diffusion still happened and
observed, and this issue will be amplified when the number of stacks getting more and more, as
discussed
1, 2
in chapter 1, 3 and 4. Therefore, one way to thoroughly eliminate any diffusion
related issue is desired. Carbon as the p-type dopant in metal organic vapor phase epitaxy
(MOVPE) is a possible choice since carbon is much less tending to diffuse in GaAs. However,
on the other hand, carbon is not an ideal p-type dopant for the absorber due to it will form
undesired carbon related defects caused non-radiative recombination sites and thus affect the
performance, as discussed
3
in chapter 1.
In order to solve the dilemma for carbon introduced above, here, the proposed method is
only using carbon in BSF and bottom contact layer, which provide little contribution to the
photocurrent and thus the carbon related defects will not affect the performance too much. At the
meant time, the carbon is doped heavy enough for these two layers for reflecting minority
carriers and enable tunneling for ohmic contact, For the absorber layer however, the doping
concentration is suppressed, preferably neither using Zn or Be which can diffuse, nor C which
can cause the defects, and the acceptor level is dominated by the un-intentional doping. By doing
in this way, the uniformity of performance could be guaranteed by un-mobile carbon and silicon,
from the dopant diffusion point of view.

116

Electrically the effect of decrease the doping concentration in the base layer could be
predicted in the following several points: First, as the doping concentration in base decreased, the
depletion width in the base layer will getting larger, and up to some point, the whole base layer
will be depleted, and it will be similar to p-i-n design commonly seen in silicon solar cells, from
which we call it as the “fully-depleted ultrathin GaAs solar cells”. In a more specific analysis, if
we fix the doping concentration of emitter as 2×10
18
cm
-3
, which is the same as the one used in
chapter 3 and 4, assuming all the dopant atoms are ideally ionized (N D = 2×10
18
cm
-3
), the device
was operated in room temperature (300K), and the intrinsic carrier concentration (n i) in GaAs at
this temperature is 2×10
6
cm
-3
, dielectric constant (εr) is 12.9, and the doping concentration in
base is always orders of magnitude lower than the one in emitter so N A + ND ≈ ND, we have
4
:
x
= ( )
V
≈ (6.1)
, where
V
=
ln (
) (6.2)
, then
x
= ln (
) = ( . )( . × )
( . × ) (0.0259) ln ( × ) × ( )
(𝑐𝑚 ) =
. × ( )
ln[5 × 10
(𝑐𝑚 )N
]
(6.3)
x
=
x
(6.4)
W = x
+ x
(6.5)

117

If NA ranges from 1×10
15
cm
-3
(about the lowest level detected from SIMS for undoped
part) to 3×10
17
cm
-3
(the design used in chapter 3 and 4), we can plot a graph as in Figure 6.1,
based on equation 6.3. As we can see in this graph, if the doping concentration of base layer is
lower than approximately 3×10
16
cm
-3
, the depletion width of p-type region is larger than the
actual base layer thickness, and thus even that all the holes participate in neutralize the electrons,
it is still not enough, or in other words, it is fully depleted base layer.

Figure 6.1: Relationship of x p and N A.

Second effect out of this low doping concentration design will be the actual performance
effect to the device. In order to know the actual effect of the doping concentration, the device
simulation is conducted in Lumerical
TM
DEVICE software pack. The optical model is established
similar as the chapter 4, while the device model and parameters used is in Figure 6.2 and Table

118

6.1. The scripts used for the simulation are in the appendices A.2.

Figure 6.2: Device model used in the simulation.

Cell size 2 × 10
-9
m
2

Series resistance 35000 Ω
Shunt resistance ∞
Surface recom. vel. between GaAs and AlGaAs 1 × 10
3
cm/s
Auger recom. coefficients (e=h) GaAs 7 × 10
-30
cm
6
/s
Auger recom. coefficients (e=h) AlGaAs 7 × 10
-30
cm
6
/s
Radiative recom. rate GaAs 5 × 10
-10
cm
3
/s
Radiative recom. rate AlGaAs 1.4 × 10
-20
cm
3
/s
Other Material Parameters (as same as the reference)
5

Table 6.1: Parameters used in the simulation.

119

In the simulation, both the I-V data under AM 1.5G 1 sun illumination, and the quantum
efficiency was calculated, as shown in Figure 6.3. Note that there are actually 4 curves in each
graph, but overlapped, which indicates that at least under the condition of the simulation
conducted here, which is quite ideal compare to the actual sample made, lowering the doping
concentration to 1×10
14
cm
-3
will not affect the performance too much. The results in reality will
need to be tested by experiments.

Figure 6.3: Device simulated data for light I-V and quantum efficiency.

120

6.2 References
1. Ilegems, M., Beryllium doping and diffusion in molecular‐beam epitaxy of GaAs and
Al x Ga1− x As. Journal of Applied Physics 1977, 48, 1278-1287.
2. Weisberg, L. R.; Blanc, J., Diffusion with interstitial-substitutional equilibrium. Zinc in
GaAs. Physical Review 1963, 131, 1548.
3. Kang, D.; Arab, S.; Cronin, S. B.; Li, X.; Rogers, J. A.; Yoon, J., Carbon-doped GaAs
single junction solar microcells grown in multilayer epitaxial assemblies. Applied Physics
Letters 2013, 102, 253902.
4. Pierret, R. F., Semiconductor device fundamentals. Pearson Education India: 1996.
5. http://www.ioffe.ru/SV A/NSM/Semicond/.

121

Appendices
The following scripts run at certain version of the software. Version for Lumerical™
FDTD is 8.15.758, the version of Lumerical™ DEVICE is 5.0.786, and version of MATLAB
script is R2016B. A different version may not be compatible for the codes showing here.

A.1 Source code for Chapter 4.2
A.1.1 Lumerical script to generate multilayer structure
switchtolayout;
clear;
##################
Num_of_Layers = 10;
##################
for (i=1:(Num_of_Layers-1))
{
groupscope("::model::Structure::Semiconductor");
select("Window");
copy(0,0,i*(-1020e-9));
set("name","Window"+num2str(i+1));
select("Emitter_Base");

122

copy(0,0,i*(-1020e-9));
set("name","Emitter_Base"+num2str(i+1));
select("BSF");
copy(0,0,i*(-1020e-9));
set("name","BSF"+num2str(i+1));
select("BottomContact");
copy(0,0,i*(-1020e-9));
set("name","BottomContact"+num2str(i+1));
groupscope("::model::Structure::Substrate");
select("Sacrificial");
copy(0,0,i*(-1020e-9));
set("name","Sacrificial"+num2str(i+1));
select("TopContact");
copy(0,0,i*(-1020e-9));
set("name","TopContact"+num2str(i+1));
}
groupscope("::model");

123

A.1.2 Lumerical script to calculate for each stacks
switchtolayout;
clear;

Num_layers = 10;
for (i=1:Num_layers)
{
?"Running for number "+ num2str(11-i)+" layer, please wait...";
select("::model::Simulation::FDTD");
set("z min",-920e-9+(i-1)*(-1020e-9));
run;
f=getdata("Monitor::m0","f");
w=c/f*1e9;
R=(transmission("Monitor::m0"))*100;
T1=-(transmission("Monitor::m1"))*100;
T2=-(transmission("Monitor::m2"))*100;
T3=-(transmission("Monitor::m3"))*100;
T4=-(transmission("Monitor::m4"))*100;
T5=-(transmission("Monitor::m5"))*100;

124

A=T1-T4;
A0=100-T5-R;
A1=T1-T2;
A2=T2-T3;
A3=T3-T4;
A4=T4-T5;
plot(w,R,A,"wavelength (nm)","R or A","Number "+num2str(11-i)+" Layer Result");
legend("R","A");
matlabsave("Number "+num2str(11-i)+" Layer Result",R,A,A0,A1,A2,w,A3,A4);
switchtolayout;
}

A.1.3 MATLAB script to summarize data
close all;
clear all;
%%%%%%%%%%%%%%%%%%%%%
Number_layers = 10;
%%%%%%%%%%%%%%%%%%%%%
load('Number 1 Layer Result.mat');

125

lam=w;
Number_points=size(A,1);
Absorption=zeros(Number_points,Number_layers);
Reflection=zeros(Number_points,Number_layers);
for i=1:Number_layers
load(char(string('Number ')+ string(i) + string(' Layer Result.mat')))
for j=1:Number_points
Absorption(j,i)=A(j);
Reflection(j,i)=R(j);
end
end
save('10_layers_summary.mat','lam','Absorption','Reflection');
for i=1:Number_layers
plot(lam,Absorption(:,i),'DisplayName',char(string('#')+string(i)));
hold on;
end
legend('show','Location','Southwest');
title('Absorption');
saveas(gcf,'Absorption_summary.jpg','jpeg');

126

hold off;
figure;
for i=1:Number_layers
plot(lam,Reflection(:,i),'DisplayName',char(string('#')+string(i)));
hold on;
end
legend('show','Location','Northwest');
title('Reflection');
saveas(gcf,'Reflection_summary.jpg','jpeg');
hold off;

A.2 Source code for Chapter 6.1
A.2.1 Lumerical script for simulation I-V for different doping concentration
clear;
switchtolayout;
matlabload("top_bias.mat");
Data_points = length(V_set);
select("CHARGE::doping_3e17");
set("enabled",0);

127

select("CHARGE::doping_3e16");
set("enabled",0);
select("CHARGE::doping_3e15");
set("enabled",0);
select("CHARGE::doping_3e14");
set("enabled",0);
####################################
select("CHARGE::doping_3e17");
set("enabled",1);
for (i=1:Data_points)
{
setbc("top","steady state","voltage",V_set(i));
run;
Results = getresult('CHARGE','bottom');
V = Results.V_top;
I = Results.I;
In = Results.In;
Ip = Results.Ip;
matlabsave("I_V_Results_1e17_"+num2str(i)+".mat",V ,I,In,Ip);

128

switchtolayout;
}
select("CHARGE::doping_3e17");
set("enabled",0);
###################################
## Change 3e17 for other doping

A.2.2 MATLAB script for summarize I-V data
close all;
clear;
%%%%%%%%%%%%%%%%%%%%%%
File_name = 'top_bias.mat';
Cell_area = 2.0e-9; %unit: m^2
%%%%%%%%%%%%%%%%%%%%%%
load (File_name);
Num_points = length(V_set);
Results_mat = NaN(Num_points,2);
% Change 3e17 for corresponding doping to get the data
for i=1:Num_points

129

if exist(char(string('I_V_Results_3e17') + string(i) + string('.mat')),'file') == 2
load (char(string('I_V_Results_3e17') + string(i) + string('.mat')));
Results_mat(i,1) = -V;
Results_mat(i,2) = 0.1*I/Cell_area; %unit: mA/cm^2
end
end
plot(Results_mat(:,1),Results_mat(:,2));

A.2.3 Lumerical script for export generation for each wavelength from FDTD for QE
simulation
clear;
f=getdata("refl_gen::field","f");
x=getdata("refl_gen::field","x",1);
y=getdata("refl_gen::field","y",1);
z=getdata("refl_gen::field","z",1);
delta_x=getdata("refl_gen::field","delta_x",1);
delta_y=getdata("refl_gen::field","delta_y",1);
if (havedata("refl_gen::field","dimension")==3) {
delta_z=getdata("refl_gen::field","delta_z",1);

130

} else {
delta_z = 0;
}
Nx = length(x);
Ny = length(y);
Nz = length(z);
Nf = length(f);
#########################################
lam_nm = solar(0)*1e9; # solar spectrum wavelength vector, in units of nm.
Psolar = solar(1)*1e-9; # solar power spectrum, in units of Watts/m^2/nm
f_solar = 1e9*c/lam_nm; # solar spectrum frequency vector, in units of Hz
Nf_solar = length(f_solar);
# select the region of solar spectrum covered by the monitor data
fl = max([min(f_solar),min(f)]);
fh = min([max(f_solar),max(f)]);
fi = find((f_solar>=fl)&(f_solar<=fh));
f_solar = f_solar(fi);
lam_nm = lam_nm(fi);
Psolar = Psolar(fi);

131

Nf_solar = length(f_solar);
nm = Psolar/sourceintensity(f_solar); # units of 1/nm
nm = meshgrid4d(4,x,y,z,nm);
# Calculate number of absorbed photon per unit volume
# assume this is equal to the number of generated electron/hole pairs
if (havedata("refl_gen::index","index_x")) {
gx = abs(getdata("refl_gen::field","Ex",1))^2 *
imag(eps0*getdata("refl_gen::index","index_x",1)^2);
gy = abs(getdata("refl_gen::field","Ey",1))^2 *
imag(eps0*getdata("refl_gen::index","index_y",1)^2);
} else {
gx = matrix(Nx,Ny,Nz,Nf);
gy = matrix(Nx,Ny,Nz,Nf);
}
if (havedata("refl_gen::index","index_z")) {
gz = abs(getdata("refl_gen::field","Ez",1))^2 *
imag(eps0*getdata("refl_gen::index","index_z",1)^2);
} else {
gz = matrix(Nx,Ny,Nz,Nf);

132

}
# sum contribution from each component, multiply by required constants, and
# interpolate absorption to standard mesh cell locations and solar frequency vector
g = 0.5 * ( interp(gx,x+delta_x,y,z,f,x,y,z,f_solar) +
interp(gy,x,y+delta_y,z,f,x,y,z,f_solar) +
interp(gz,x,y,z+delta_z,f,x,y,z,f_solar) ) / hbar;
gx=0; gy=0; gz=0; # clear variables, to free memory

# Calculate the generation rate by integrating 'g' over wavelength
# The generate rate is the number of electron hole pairs per unit volume per second.
(units: 1/m^3/s)
G_matrix = integrate2(g*nm,4,lam_nm);
matlabsave("Output_test.mat",g,x,y,z,lam_nm,nm);

A.2.4 MATLAB script for summarize the generation for each wavelength
clear;
%%%%%%%%%%%%%%%%%%%%%
Input_filename = 'Output_test.mat';
Output_filename = 'Generation_';

133

%%%%%%%%%%%%%%%%%%%%%
load(Input_filename);
z = [-1e-6 1e-6]';
Wavelength_points = length(lam_nm);
for i=1:Wavelength_points
G(:,:,1,1) = g(:,:,:,i);
G(:,:,2,1) = g(:,:,:,i);
save(char(string(Output_filename) + num2str(lam_nm(i)) +
string('_nm.mat')),'G','x','y','z');
end

A.2.5 Lumerical script for import generation for each wavelength to DEVICE
clear;
######################
Wavelength_low_limit = 400; #unit in nm
Wavelength_high_limit = 999;
######################
for (i=Wavelength_low_limit:Wavelength_high_limit)
{

134

addimportgen;
set("name","gen_" + num2str(i));
set("x",0);
set("y",0);
set("z",0);
matlabload("Generation_" + num2str(i) + "_nm.mat");
gen = rectilineardataset("gen",x,y,z);
gen.addattribute("G",G);
importdataset(gen);
set("enabled",0);
}

A.2.6 Lumerical script for simulation QE for different doping concentration
clear;
switchtolayout;
###################################
Wavelength_start = 400;
Wavelength_end = 999; ## unit in nm
###################################

135

select("CHARGE::doping_3e17");
set("enabled",0);
select("CHARGE::doping_3e16");
set("enabled",0);
select("CHARGE::doping_3e15");
set("enabled",0);
select("CHARGE::doping_3e14");
set("enabled",0);
####################################
setbc("top","steady state","voltage",0);
## set bias condition, QE uses 0.
####################################
select("CHARGE::doping_3e17");
set("enabled",1);
for (i=Wavelength_start:Wavelength_end)
{
select("CHARGE::gen_" + num2str(i));
set("enabled",1);
run;

136

Results = getresult('CHARGE','bottom');
V = Results.V_top;
I = Results.I;
In = Results.In;
Ip = Results.Ip;
lambda = i;
matlabsave("doping_3e17_"+num2str(i)+"_nm.mat",V ,I,In,Ip,lambda);
switchtolayout;
select("CHARGE::gen_" + num2str(i));
set("enabled",0);
}
####################################
select("CHARGE::doping_3e17");
set("enabled",0);
## Change 3e17 for other doping

A.2.7 MATLAB script for summarize QE data
clear;
close all;

137

%%%%%%%%%%%%%%%%
%This is for directly using g as G, G=g*(Psolar/sourceintensity)*lam_nm,
%where (Psolar/sourceintensity) has the unit of nm^-1, and lam_nm has the
%unit of nm, so g and G has the same unit as #*m^-3*s^-1.
%%%%%%%%%%%%%%%%
Wavelength_start = 400; %unit: nm
Wavelength_end = 999; %unit: nm
File_name = ' doping_3e17_'; % Change it for other doping
Cell_area = 2.0e-9; %unit: m^2
Plank_const = 6.626e-34; %unit: J*s
Speed_of_light = 3e8; %unit: m*s^-1
Unit_charge = 1.602e-19; %unit: columb*#^-1
Source_intensity = 0.0013; %unit: W*m^-2
%%%%%%%%%%%%%%%%
Wavelength_plot = (Wavelength_start:1:Wavelength_end)';
Current = NaN(1,Wavelength_end-Wavelength_start+1)';
Photon_energy = Plank_const * Speed_of_light ./ (1e-9*Wavelength_plot); %unit: J
Num_of_photons_density = Source_intensity ./ Photon_energy; %unit: #*m^-2*s-1
for i = Wavelength_start:Wavelength_end

138

load (char(string(File_name) + string(i) + string('_nm.mat')));
Current (i-Wavelength_start+1) = -I; %unit: A, because g and G has the same unit.
end
Current_density = Current/Cell_area; %unit: A*m^-2
Num_of_carriers_density = Current_density / Unit_charge; %unit: #*m^-2*s-1
EQE = Num_of_carriers_density ./ Num_of_photons_density; %unit: 1
plot (Wavelength_plot,EQE);
save ('Analyzed_data.mat','Wavelength_plot','EQE');

Abstract (if available)

Abstract Large-scale deployment of GaAs solar cells in terrestrial photovoltaics demands significant cost reduction for preparing device-quality epitaxial materials. Although multilayer epitaxial growth has been proposed as a promising route to achieve this goal, their practical implementation remains challenging owing to the degradation of materials properties and resulting nonuniform device performance between solar cells grown in different sequences. In this thesis, we studied alternative approach to circumvent these limitations and enable multilayer-grown GaAs solar cells with uniform photovoltaic performance. Ultrathin single-junction GaAs solar cells having a 300-nm-thick absorber (i.e., emitter and base) are epitaxially grown in triple- (Chapter 3) and tenfold-stack (Chapter 4) releasable multilayer assemblies by molecular beam epitaxy using beryllium as a p-type impurity. Microscale GaAs solar cells fabricated from respective device layers exhibit excellent uniformity of photovoltaic performance and contact properties owing to the suppressed diffusion of p-type dopant as well as substantially reduced time of epitaxial growth associated with ultrathin device configuration. Bifacial photon management employing hexagonally periodic TiO₂ nanoposts and a vertical p-type metal contact serving as a metallic back-surface reflector together with specialized epitaxial design to minimize parasitic optical losses for efficient light trapping synergistically enable significantly enhanced photovoltaic performance of such ultrathin absorbers, where ∼17.2% solar-to-electric power conversion efficiency under simulated AM1.5G illumination is demonstrated from 420-nm-thick single-junction GaAs solar cells. ❧ Inverted metamorphic (IMM) multijunction solar cells represent a promising materials platform for ultrahigh efficiency photovoltaic systems (UHPVs) with a clear pathway to beyond 50% efficiency. The conventional device processing of IMM solar cells, however, typically involves wafer bonding of a centimeter-scale die and destructive substrate removal, thereby imposing severe restrictions in achievable cell size, type of module substrate, spatial layout, as well as cost-effectiveness. In this thesis (Chapter 5), we studied materials design and fabrication strategies for microscale triple-junction inverted metamorphic (3J IMM) Ga₀.₅₁In₀.₄₉P/GaAs/In₀.₂₆Ga₀.₇₄As solar cells that can overcome these difficulties. Specialized schemes of delineation and undercut etching enable the defect-free release of microscale IMM solar cells and printed assemblies on a glass substrate in a manner that preserves the growth substrate, where efficiencies of 27.3% and 33.9% are demonstrated at simulated AM1.5D one- and 351 sun illumination, respectively. A composite carrier substrate where released IMM microcells are formed in fully-functional, print-ready configurations allows high-throughput transfer printing of individual IMM microcells in a programmable spatial layout on versatile choices of module substrate, all desired for CPV applications.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

GaAs nanowire optoelectronic and carbon nanotube electronic device applications

PDF

Growth, characterization of gallium arsenide based nanowires and application in photovoltaic cells

PDF

Light management in nanostructures: nanowire solar cells and optical epitaxial growth

PDF

Enhanced photocatalysis on titanium oxide passivated III-V semiconductors

PDF

Complementary hydrogen bonding enables efficient and mechanically robust polymer solar cells

PDF

Non-radiative energy transfer for photovoltaic solar energy conversion: lead sulfide quantum dots on silicon nanopillars

Asset Metadata

Creator Gai, Boju (author)

Core Title Multilayer grown ultrathin nanostructured GaAs solar cells towards high-efficiency, cost-competitive III-V photovoltaics

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Materials Science

Publication Date 04/26/2019

Defense Date 08/10/2018

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag bifacial photon management,III−V solar cells,inverted metamorphic,microcells,multijunction solar cells,multilayer epitaxial assemblies,nanomembrane,OAI-PMH Harvest,transfer printing,ultrathin gallium arsenide

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Yoon, Jongseung (committee chair), Kassner, Michael Ernest (committee member), Wu, Wei (committee member)

Creator Email bgai@usc.edu,gaiboju@foxmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-148707

Unique identifier UC11660204

Identifier etd-GaiBoju-7277.pdf (filename),usctheses-c89-148707 (legacy record id)

Legacy Identifier etd-GaiBoju-7277.pdf

Dmrecord 148707

Document Type Dissertation

Format application/pdf (imt)

Rights Gai, Boju

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA