Ultra-thin GaAs single-junction solar cells integrated with a reflective back scattering
layer
Weiquan Yang, Jacob Becker, Shi Liu, Ying-Shen Kuo, Jing-Jing Li, Barbara Landini, Ken Campman, and YongHang Zhang
Citation: Journal of Applied Physics 115, 203105 (2014); doi: 10.1063/1.4878156
View online: http://dx.doi.org/10.1063/1.4878156
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/115/20?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
13.2% efficiency double-hetero structure single-junction InGaAsN solar cells grown by MOVPE
J. Vac. Sci. Technol. A 33, 021205 (2015); 10.1116/1.4906511
Experimental determination of band offsets of NiO-based thin film heterojunctions
J. Appl. Phys. 116, 163108 (2014); 10.1063/1.4900737
The impact of oxygen incorporation during intrinsic ZnO sputtering on the performance of Cu(In,Ga)Se2 thin film
solar cells
Appl. Phys. Lett. 105, 083906 (2014); 10.1063/1.4894214
Enhancing the efficiency of SnS solar cells via band-offset engineering with a zinc oxysulfide buffer layer
Appl. Phys. Lett. 102, 053901 (2013); 10.1063/1.4789855
Analyzing nanotextured transparent conductive oxides for efficient light trapping in silicon thin film solar cells
Appl. Phys. Lett. 101, 103903 (2012); 10.1063/1.4750242

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58

JOURNAL OF APPLIED PHYSICS 115, 203105 (2014)

Ultra-thin GaAs single-junction solar cells integrated with a reflective back
scattering layer
Weiquan Yang,1 Jacob Becker,1 Shi Liu,1 Ying-Shen Kuo,1 Jing-Jing Li,1 Barbara Landini,2
Ken Campman,2 and Yong-Hang Zhang1

1
Center for Photonics Innovation and School of Electrical, Computer and Energy Engineering, Arizona State
University, Tempe, Arizona 85287, USA
2
Sumika Electronic Materials, Inc., Phoenix, Arizona 85034, USA

(Received 31 October 2013; accepted 5 May 2014; published online 28 May 2014)
This paper reports the proposal, design, and demonstration of ultra-thin GaAs single-junction solar
cells integrated with a reflective back scattering layer to optimize light management and minimize
non-radiative recombination. According to our recently developed semi-analytical model, this
design offers one of the highest potential achievable efficiencies for GaAs solar cells possessing
typical non-radiative recombination rates found among commercially available III-V arsenide
and phosphide materials. The structure of the demonstrated solar cells consists of an
In0.49Ga0.51P/GaAs/In0.49Ga0.51P double-heterostructure PN junction with an ultra-thin 300 nm
thick GaAs absorber, combined with a 5 lm thick Al0.52In0.48P layer with a textured as-grown
surface coated with Au used as a reflective back scattering layer. The final devices were fabricated
using a substrate-removal and flip-chip bonding process. Solar cells with a top metal contact
coverage of 9.7%, and a MgF2/ZnS anti-reflective coating demonstrated open-circuit voltages (Voc)
up to 1.00 V, short-circuit current densities (Jsc) up to 24.5 mA/cm2, and power conversion
efficiencies up to 19.1%; demonstrating the feasibility of this design approach. If a commonly used
2% metal grid coverage is assumed, the anticipated Jsc and conversion efficiency of these devices
C 2014 AIP Publishing LLC.
are expected to reach 26.6 mA/cm2 and 20.7%, respectively. V
[http://dx.doi.org/10.1063/1.4878156]

INTRODUCTION

GaAs has the highest demonstrated power conversion
efficiency among all types of single-junction solar cells, and
is the most promising material for achieving efficiencies that
approach the Shockley-Queisser (SQ) limit. Efficiency
records of GaAs solar cells have been steadily broken over
the past several years.1–3 Most of the GaAs solar cells with
high efficiencies reported up to now use exceptionally highquality materials for relatively thick absorbers, and planar
structures with smooth top surfaces and flat backside
mirrors.2–4 Different designs to enhance the efficiency have
also been proposed and reported, such as the use of planer
thin absorbers.5–11 In order to increase the efficiency and
reduce manufacturing cost by using commonly available
materials with routinely achievable quality, it becomes necessary to implement light management structures (e.g., textured surface)12–14 to achieve maximum absorption and
minimum non-radiative recombination in a thin absorber.
Our recently reported theoretical analysis of different optical
designs5,6,15,16 shows that the use of an ultra-thin (submicron) absorber and a reflective back scattering layer can
potentially result in the maximal achievable conversion efficiency for single-junction GaAs solar cells.
In this paper, a solar cell structure consisting of an
In0.49Ga0.51P/GaAs/In0.49Ga0.51P double-heterostructure PN
junction with an ultra-thin GaAs absorber and a textured
Al0.52In0.48P/Au back reflective layer is modeled, designed,
and fabricated. Ultra-thin absorbers and textured back reflective layers are chosen to achieve the optimal efficiency using
0021-8979/2014/115(20)/203105/7/$30.00

routinely available materials with non-radiative recombination rates similar to those reported in the literature.5,6,17,18
The textured back scattering layer, offering light trapping,
makes the ultra-thin absorber optically thick but physically
thin, and thus optimizes the tradeoff between maximizing
absorption of incoming photons and minimizing
non-radiative recombination in the absorber. In addition, the
textured interface on the backside of the Al0.52In0.48P scattering layer is separated from the GaAs absorber so that the
surface-related non-radiative recombination is minimized,
while the light scattering is maximized independently. The
devices were fabricated using a substrate-removal flip-chip
bonding process, and an anti-reflective (AR) coating comprised of a MgF2/ZnS double-layer. This work demonstrates
the feasibility of the use of ultra-thin absorbers and textured
back reflective layers in single-junction solar cells to optimize light management and to minimize non-radiative
recombination, and thus achieve high power conversion
efficiencies.
DEVICE DESIGN AND MODELING

The overall device structure is shown in Fig. 1(a). The
double-heterostructural GaAs PN junction consists of a
300 nm GaAs absorber (p-type emitter and n-type base)
sandwiched between two 30 nm doped In0.49Ga0.51P window
and back-surface-field (BSF) layers. A textured surface is
placed at the backside of the thin-film solar cell to provide
optimal scattering of the transmitted light, which is mainly at
red and infrared wavelengths. The reflective back scattering

115, 203105-1

C 2014 AIP Publishing LLC
V

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58

203105-2

Yang et al.

J. Appl. Phys. 115, 203105 (2014)

FIG. 1. Schematic layer structure of
(a) the finished In0.49Ga0.51P/GaAs/In0.49
Ga0.51P double-heterostructural solar
cell with an Al0.52In0.48P layer for reflective back scattering, and (b)
as-grown solar cell wafers.

layer underneath the In0.49Ga0.51P BSF layer is a 5 lm thick,
n-type Al0.52In0.48P layer with a textured surface coated with
Au. The entire device is bonded on to a Si carrier substrate
using indium. Fig. 1(b) shows the detailed layer structure,
including the layer thicknesses and the doping concentrations of the epitaxial wafer prior to device processing.
The optimal thickness of the GaAs absorber is chosen
according to simulation using the semi-analytical model5,6
that takes into account non-radiative recombination, emittance, and absorptance calculated from the reported absorption coefficients19 including the tail below the bandgap
wavelength (Urbach tail). Fig. 2(a) shows plots of simulated
power conversion efficiencies versus absorber thickness of
GaAs single-junction solar cells with ideal textured reflective
back surfaces providing Lambertian scattering for given
Shockley-Read-Hall (SRH) recombination saturation current
densities per unit length (JSRH/d). The Auger recombination is
assumed to be zero, which is reasonable for typical GaAs
materials. JSRH/d is proportional to the non-radiative recombination rate per unit volume in the GaAs absorber, and is therefore an effective factor for characterizing the quality of GaAs
material. The green dashed curve with JSRH/d equaling
10 A/cm2/lm represents the lowest SRH recombination rate
reported in the literature for GaAs.5,6 A maximum efficiency
of 30.9% is thus potentially achievable with a 230 nm thick
absorber using the proposed structure. The optimal absorber
thickness for each structure with a given JSRH/d is the result of
a compromise between maximizing the absorption and minimizing the non-radiative recombination loss. The curve of
JSRH/d ¼ 0 represents the case of no non-radiative recombination present; thus, the efficiency approaches the detailed balance limit asymptotically with the absorber thickness
approaching infinity. However, as the SRH recombination saturation current density increases with degrading material quality, the maximum achievable efficiency decreases and there
exists a decreasing optimal thickness. This trend reveals the
advantage of using ultra-thin absorbers when the material has
substantial non-radiative recombination loss, such as in the
case of solar cells made of thin-film polycrystalline materials.
Utilizing textured surfaces for light management in
GaAs-based single- or multi-junction solar cells is challenging. This is mainly due to the fact that the absorbers in those
cells usually have large absorption coefficients and are,
therefore, very thin with a typical thickness of only a few

microns. These thin layers make it difficult to texture the
absorber surfaces. Texturing becomes even more difficult for
ultra-thin GaAs absorbers on the order of a few hundred
nanometers. Furthermore, textured absorbers have large surface or interface areas and thus result in significantly
increased non-radiative surface/interface recombination. The
solution proposed here is to integrate the solar cell with a
wide-bandgap back scattering layer coated with a reflective
mirror. The key advantage of this design is that the textured
surface and the absorber/BSF interface are separated. The
textures are fabricated on the thick wide-bandgap light scattering layer, and the surfaces of the ultra-thin GaAs absorber
remain planar. Therefore, the non-radiative interface recombination of the GaAs absorber will not be increased. It is
essential that the material used for the back scattering layer
has a large bandgap to minimize the parasitic optical absorption of the transmitted light, as well as be lattice-matched to
GaAs to minimize the number of misfit dislocations formed
during epitaxial growth. The simulated power conversion
efficiencies of GaAs single-junction solar cells integrated
with 5 lm thick reflective back scattering layers comprised
of several different III-V materials (Al0.50In0.50P,
Al0.80Ga0.20As, Al0.59Ga0.41As, and Ga0.50In0.50P) are plotted
versus absorber thickness in Fig. 2(b). The reported absorption coefficients20–22 are used for the various back scattering
materials. The best reported value of JSRH/d ¼ 10 A/cm2/lm
and a typical Auger recombination saturation current density
per unit length JAuger/d ¼ 120 A/cm2/lm for GaAs are used.6
Ideal AR coatings with no reflection loss and no top contact
grid coverage are also assumed here. Note that the choice of
5 lm as the back scattering layer thickness in the modeling is
due to growth considerations, which will be discussed below.
The figure clearly indicates the benefits of using textured
Al0.50In0.50P layer for back scattering, as it offers the largest
efficiency among the four candidates due to its widest
bandgap and correspondingly low parasitic absorption of the
transmitted light. From the simulation, the optimal thickness
of the GaAs absorber with the Al0.50In0.50P back scattering
layer is 280 nm. It is noted here that the model assumes ideal
Lambertian back scattering. A slightly thicker GaAs
absorber (300 nm) is chosen in this work to compensate for
the non-ideal scattering typically found in practical devices.
A MgF2/ZnS double-layer AR coating is designed and
optimized by minimizing the front surface reflectance

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58

203105-3

Yang et al.

J. Appl. Phys. 115, 203105 (2014)

In0.49Ga0.51P window, and BSF layers in the transfer matrix
calculation. Once each reflection spectrum is calculated
using published complex refractive indices,19,20 it is
weighted against the solar spectrum and then integrated
below the GaAs bandgap wavelength (280 nm–870 nm) to
produce a total reflectance for every thickness combination
of the MgF2/ZnS double-layer, as shown by the equation
below
ð 870nm
RðkÞ  Nsun ðkÞdk
;
(1)
Rtotal ¼ 280nm
ð 870nm
Nsun ðkÞdk
280nm

where Rtotal is the total reflectance of the AR coating, R(k) is
the wavelength-dependent reflectance of the AR coating, and
Nsun(k) is the photon flux per wavelength from solar radiation. Fig. 2(c) depicts the simulated total reflectance (Rtotal)
contours versus variable MgF2 and ZnS thicknesses. The
optimal design (96 nm MgF2/47 nm ZnS pair) is denoted in
the figure by the grey dot in the approximate center, where a
minimum Rtotal of 1.60% is achieved.

MATERIAL GROWTH AND DEVICE FABRICATION

FIG. 2. Simulated power conversion efficiency of GaAs single-junction solar cells (a) versus GaAs absorber thickness for various normalized SRH
non-radiative recombination saturation current densities per unit length
JSRH/d (A/cm2/lm); (b) versus GaAs absorber thickness for four different
III-V materials lattice-matched to GaAs integrated as the back scattering
layer; (c) simulated reflectance contour R total for MgF2/ZnS double-layer
AR coating on a GaAs solar cell.

calculated by the transfer matrix method. Note that since the
thickness of the GaAs absorber is only 300 nm, it is necessary to take into account the GaAs ultra-thin absorber,

The solar cell structures shown in Fig. 1(b) were grown
using MOCVD at Sumika Inc., a commercial epi foundry.
The double-heterostructural n-In0.49Ga0.51P/n-GaAs/p-GaAs/
p-In0.49Ga0.51P PN junction was deposited on an Al0.9Ga0.1As
sacrificial layer grown on a GaAs substrate. The p-GaAs and
n-GaAs layers have thicknesses of 30 nm and 270 nm, and
corresponding doping concentrations of 1.25  1017 cm3 and
7  1016 cm3, respectively. The window layer and BSF layer
are In0.49Ga0.51P offering a high quality interface with GaAs.
A 5 lm thick n-type Al0.52In0.48P layer with a textured
surface is grown on top of the second In0.49Ga0.51P layer.
The cone-shaped textures on the surface of the Al0.52In0.48P
layer were developed during the growth; hence, no additional
processing was required to produce the textured surface for
light scattering. Several calibration growths of Al0.52In0.48P
layers with different thicknesses were carried out to investigate the correlation between texture distribution and the
Al0.52In0.48P thickness. Figs. 3(a)–3(c) show the
cross-sectional SEM images of the Al0.52In0.48P surface with
thicknesses of 1 lm, 3.8 lm, and 5 lm. When the
Al0.52In0.48P thickness was only 1 lm, most of the surface
was smooth and sparse discrete cone textures were formed.
The 3.8 lm Al0.52In0.48P layer showed denser surface textures. However, most of the textures were still discrete, with
only a few features beginning to connect with each other.
For the 5 lm Al0.52In0.48P layer, all of the texture features
are connected with each other to cover the entire surface.
Therefore, 5 lm was chosen as the thickness of the
Al0.52In0.48P back scattering layer. A zoom-in cross-sectional
view of a texture cone with a height of 1 lm and a bottom diameter of around 5 lm is shown in Fig. 3(c) inset. The scale
in the vertical axis is the same as that in the horizontal axis.
Fig. 3(d) shows a top-view microscope image of the

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58

203105-4

Yang et al.

FIG. 3. Cross-section SEM images of as-grown Al0.52In0.48P surfaces when
the Al0.52In0.48P thickness is (a) 1 lm, (b) 3.8 lm, and (c) 5 lm; (d) top-view
microscopic images of as-grown Al0.52In0.48P surface when the Al0.52In0.48P
thickness is 5 lm. Inset in (c) gives a zoomed-in SEM image of a typical
cone texture (height  1 lm, diameter  5 lm).

as-grown Al0.52In0.48P surface with high-density cone
features.
The device processing begins with the as-grown solar
cell structures shown in Fig. 4(a). The initial stage was to
prepare the back contact and mirror prior to the flip-chip
mounting of the device. The commonly used alloyed metal
contact (e.g., Ni/Ge/Au/Ti/Au) for n-GaAs has excellent
contact conductance but poor reflectance. Conversely, a pure
Au layer has excellent reflectance but poor contact conductance. Therefore, in order to maximize the reflectivity of the
back mirror and minimize the contact resistance of the back
contact, point contacts (Ni/Ge/Au/Ti/Au) were used as
shown in Fig. 4(b). Each point contact has a size of
10  10 lm2 with a distance of 200 lm between adjacent
points. Thus, the total area of the point contacts accounts for
only 0.25% of the back surface. The point contacts were
made using a standard photolithography, e-beam deposition,
and lift-off process. The sample was then annealed using a
rapid thermal annealing process at 425  C to form ohmic

J. Appl. Phys. 115, 203105 (2014)

contacts. The remaining GaAs back contact layer was then
removed using a citric acid solution revealing the
Al0.52In0.48P back scattering layer. A Au film was then deposited on the textured Al0.52In0.48P layer to function as the
mirror. No annealing process was used after the mirror deposition in order to maintain a high reflectivity at that surface.
In addition to the initial Au layer, a Cr layer was also deposited as a diffusion barrier to prevent degradation of the mirror quality from the diffusion of indium used in subsequent
steps. A second Au layer was then deposited to function as a
wetting layer to better adhere to the indium. Fig. 4(c) shows,
schematically, the mirror deposited on the textured
Al0.52In0.48P layer along with the point contacts. Indium was
used as the bonding medium to secure the GaAs wafer to the
Si carrier wafer. A Ti/Au film deposited onto the Si carrier
wafer using E-beam evaporation prior to the bonding process
acts as a wetting layer for better adhesion to the indium, and
also as the back probing pad of the solar cells during the final
measurement. Indium shot was melted across the surface of
the Si wafer using a hot plate at 210  C, and the GaAs wafer
was mounted on the indium, mirror side down. Fig. 4(d)
shows the structure after the sample was bonded to the Si
carrier wafer.
The GaAs substrate was then removed through a
chemical-etching process over a period of approximately 3 h
using an NH4OH/H2O2 solution. The Al0.9Ga0.1As etch-stop
layer prevented the absorber from being removed. This
Al0.9Ga0.1As etch-stop layer was then removed using an
HF/H2O solution as shown in Fig. 4(e). The top Ti/Pt/Au
contact grids were deposited using a photolithography,
E-beam evaporation, and lift-off process. The exposed GaAs
contact layer was then removed using a citric acid etchant to
expose the In0.49Ga0.51P window layer underneath, as shown
in Fig. 4(f). The individual devices were isolated using
chemical etching between the mesas, as shown in Fig. 4(g),
with HNO4/HCl/H2O used as the etching solution. The final
cell geometries are 0.3 mm  0.3 mm, 0.6 mm  0.6 mm, and
1 mm  1 mm. The thickness-optimized MgF2/ZnS layers for
the AR coating were then deposited using a standard

FIG. 4. Schematic flow of the device growth and fabrication processes: (a) MOCVD growth of the double-heterostructure PN junction device with an
A0.52In0.48P light scattering layer on a GaAs substrate; (b) metals deposition for point contacts; (c) deposition of Au mirror; (d) flip-chip bonding to a Si carrier
substrate; (e) removal of GaAs substrate; (f) top contact deposition; (g) mesa etching; and (h) MgF2/ZnS AR coating deposition.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58

203105-5

Yang et al.

J. Appl. Phys. 115, 203105 (2014)

FIG. 5. Measured reflectance of a 97 nm MgF2/68 nm ZnS AR coating on a
blank GaAs calibration wafer.

FIG. 6. Current–voltage curves of TLM contacts with pad spacing distances
of 20, 40, 60, 80, and 100 lm.

photolithography, thermal evaporation, and lift-off process.
Fig. 4(h) shows the final device schematic including the AR
coating ready for test.
A custom built thermal evaporation tool was used for
the AR coating deposition. During deposition, a blank GaAs
calibration wafer was placed alongside the actual devices to
permit straightforward measurement of the AR coating
thickness and reflectance. The MgF2 and ZnS thicknesses on
the calibration wafer were measured to be 97 nm and 68 nm
by Filmetrics F-40 reflectance measurement, respectively.
The reflectance curve of the calibration wafer is shown in
Fig. 5. This reflectance curve is then weighted against the solar spectrum and integrated below the GaAs bandgap wavelength as described above. A total reflectance (Rtotal) of
3.77% was obtained for this MgF2/ZnS bilayer stack. Due to
the limitation of the custom tool used for AR coating deposition, there is a large spatial variation of the deposited film
thickness. As the calibration wafer and the solar cell devices
are at different locations, the final AR coating thicknesses on
the solar cell samples may be slightly different from that of
the calibration wafer.

100 lm) are shown in Fig. 6. The linear I-V curves reveal
that the top metal p-GaAs contacts are ohmic with a specific
contact resistivity of 6.51  103 Xcm2.
The finished devices have various areas ranging from
0.09 mm2 to 1 mm2 as shown in the optical micrograph (Fig.
7). Devices with an area of 0.09 mm2 and a 9.7% top contact
grid coverage were characterized at room temperature under
1 sun AM 1.5G solar spectrum. A typical I-V plot of the 0.09
mm2 solar cell is shown in Fig. 8. Short-circuit current densities (Jsc) as high as 24.5 mA/cm2, open-circuit voltages
(Voc) up to 1.00 V, fill factors (FF) of 77.8%, and power conversion efficiencies (g) up to 19.1% were attained. The measured performance is compared with the best achievable
performance, both of which are summarized in Table I. The
best achievable performance is modeled based on the
assumptions of Lambertian scattering on the back side of the
cell, a back mirror with 100% reflectivity, an AR coating
with 2% reflectance, a 2% top contact grid coverage, the best
material quality with the longest minority carrier lifetime
published, and no series or shunt resistance.
The demonstrated GaAs single-junction solar cells with
a 300 nm absorber have a maximum achievable Jsc of

DEVICE CHARACTERIZATION AND ANALYSIS

The current-voltage (I-V) characteristics of the processed solar cell devices were measured using an Oriel ClassA solar simulator (AM 1.5G, 0.1 W/cm2) and a Keithley
2400 source meter. The spectral reflectance measurements of
the MgF2/ZnS AR coating were carried out at normal incidence using a Perkin Elmer Lambda 18 UV-Vis spectrometer. The external quantum efficiencies (EQE) of the devices
were measured at room temperature using a Newport
QE/IPCE measurement setup.
Because the devices were not thermally annealed after
the front-contact deposition to avoid degradation of the Au
mirror and the melting of the indium, the device structure
utilizes a 6 nm highly doped (1  1019 cm3) pþ-GaAs layer
on the 200 nm p-GaAs as a non-alloyed ohmic contact layer.
The contact resistance of the top p-contact grids was characterized using the transfer length method (TLM). The I–V
characteristics measured between two adjacent electrodes
with different spacing (20 lm, 40 lm, 60 lm, 80 lm, and

FIG. 7. Optical micrograph of fabricated solar cells possessing different
areas.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58

203105-6

Yang et al.

J. Appl. Phys. 115, 203105 (2014)

FIG. 8. Measured room-temperature I-V characteristics of the GaAs solar
cells under 1 sun (AM 1.5G). Inset: calculated short-circuit current density
(Jsc) as a function of Phong exponent (m).

30.7 mA/cm2. The best device with a 300 nm absorber tested
has a measured Jsc of 24.5 mA/cm2. The causes of this discrepancy from the maximum achievable Jsc are: (1) the optical scattering in the actual devices is not as efficient as
Lambertian scattering; (2) the Au mirror has less than 100%
reflectivity; (3) the AR coating has a larger reflectance
(3.8%) than that used in the model (2%); and (4) the metal
grid coverage (9.7%) is much higher than that used in the
model (2%). If a 2% metal grid shadowing is assumed as in
conventional devices and all other factors remain the same,
the Jsc and power conversion efficiency of these actual devices could reach 26.6 mA/cm2 and 20.7%, respectively. The
resulting 4.1 mA/cm2 difference between the projected and
maximum achievable Jsc is due in part to non-ideal scattering. The impacts of non-ideal scattering are addressed as
below.
Phong’s distribution can be used to describe non-ideal
scattering deviating from Lambertian scattering.23 The angular light intensity is proportional to cosm(h), where h is the
angle between the scattered light and the surface normal, and
the Phong exponent (m) can be used to describe the angular
intensity distribution. Note here that m ¼ 1 corresponds to
the Lambertian distribution, and higher values of m result in
narrower angular intensity distributions. The calculated Jsc
as a function of m value is shown in Fig. 8 inset. The calculation assumes a reflectivity of 95% for the back
Al0.52In0.48P/Au mirror, a metal grid shadow area of 9.7%,
and a front surface reflection loss of 3.8%. The value of 95%
is chosen as an approximation of the reflectivity of the rough
Al0.52In0.48P/Au mirror based on the Au reflectivity spectrum
and its angular distribution. The calculation results indicate
that the Al0.52In0.48P/Au reflectivity averaged against the
Lambertian distribution (m ¼ 1) at the GaAs band edge

wavelength (870 nm) is 94.8%, while the value for the narrowest Phong’s distribution (m ¼ 1) is 94.4%. Fig. 8 inset
shows that Jsc decreases from 27 mA/cm2 to 23 mA/cm2
as the m value increases from 1 to 40, which represents that
the scattering gradually deviates from a Lambertian distribution to a Phong’s distribution. The m value of the Phong’s
distribution in our device is determined to be 15 by fitting
the modeled Jsc to the experimental results.
The over 0.1 V difference between the measured and the
theoretically predicted, best achievable Voc could be partly
due to SRH and interface/surface recombination losses. The
impact of interface/surface recombination becomes relatively more important for such a thin absorber. The current
low FF value of 77.8% is mainly due to the series resistance.
A specific series resistivity of 0.5 Xcm2 is determined by fitting the deviation of the dark I-V curve from the ideal diode
equation at a positive bias voltage of 1.5 V. Poor current
spreading in the devices with ultra-thin absorber and window
layers, sparse back point contacts, as well as non-annealed
top contact grids all contribute to this relatively high series
resistance. The FF can be potentially improved by using our
recently proposed conductive AR coating.24
Fig. 9 shows the EQE of the measured solar cell. The
EQE increases from 80% at 400 nm to 90% at 470 nm. Note
that the light at a wavelength near 400 nm is mainly absorbed
by the In0.49Ga0.51P window layer. The high EQE near
400 nm indicates that the highly efficient extraction of those
photo-generated carriers in the window layer is due to the
shallow PN junction of the ultra-thin absorber and the extension of the depletion region into the window layer. The maximum value of the EQE is only 90% mainly due to the fact
that the top grid has a shadow area of 9.7%. The EQE is
close to 65% at 800 nm and above 40% at the bandgap wavelength of 870 nm. The light absorption at 800 nm is calculated to be 54% for devices with specular reflective back
surfaces without light scattering. These calculations assume
shadow area of 9.7%, a reflectivity of 95% for the back
Al0.52In0.48P/Au mirror, and a reflectance of 2% at 800 nm
for AR coating. The 11% increase in the measured EQE
compared with the calculated specular light absorption at

TABLE I. Comparison between the measured and modeled best achievable
device performance of 300 nm GaAs single-junction solar cell.

Modeled best achievable
Measured

Voc (V)

Jsc (mA/cm2)

FF (%)

g (%)

1.13
1.00

30.7
24.5

85.4
77.8

29.5
19.1

FIG. 9. External quantum efficiency as a function of wavelength for the
GaAs solar cell.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58

203105-7

Yang et al.

800 nm indicates that the textured back scattering has
resulted in an increase in the EQE at these wavelengths. No
interference fringes in the EQE spectrum between 550 nm
and 870 nm are observed, indicating that the textured
Al0.52In0.48P surface offers reasonably effective light scattering.17 However, the presence of a lower EQE at the longer
wavelengths compared to that at the shorter wavelengths
indicates that the reflective back scattering is not as efficient
as Lambertian scattering/reflection.
CONCLUSION

Semi-analytical modeling shows that one of the highest
potential efficiencies for GaAs single-junction solar cells
with conventional material quality can be obtained through
the combination of an ultra-thin absorber and a reflective
back scattering layer. Ultra-thin, double-heterostructural,
single-junction In0.49Ga0.51P/GaAs/In0.49Ga0.51P solar cells,
monolithically integrated with a lattice-matched textured
Al0.52In0.48P layer coated with a Au mirror have been successfully demonstrated. Open-circuit voltages up to 1.00 V,
short-circuit current densities up to 24.5 mA/cm2, and a maximum power conversion efficiency of 19.1% are measured
under 1 sun AM 1.5G solar radiation. These results were
obtained for substrate-removed and flip-chip bonded devices
with an ultra-thin 300 nm GaAs absorber, a textured
Al0.52In0.48P surface, a MgF2/ZnS AR coating, and a contact
grid layout covering 9.7% of the front surface area. If a more
conventional 2% metal grid shadow is assumed, a Jsc of
26.6 mA/cm2 and a conversion efficiency of 20.7% can be
expected for these devices. Deviations in device performance
from the ideal model are discussed. These initial results demonstrate the feasibility and potential of using an ultra-thin
GaAs single-junction solar cell design integrated with a reflective back scattering layer to optimize light management
and minimize non-radiative recombination.
ACKNOWLEDGMENTS

We thank C. Allen, A. P. Kirk, D. Ding, and Y. Zhao for
their helpful comments and stimulating discussions. We
gratefully acknowledge the use of facilities within the Center
for Solid State Science as well as the Center for Solid State
Electronics Research at Arizona State University. This work
was supported in part by the Science Foundation Arizona,

J. Appl. Phys. 115, 203105 (2014)

Contract No. SRG 0339-08, and NSF grant Contract No.
1002114.
1

M. A. Green, K. Emery, Y. Hishikawa, W. Warta, and E. D. Dunlop, Prog.
Photovoltaics 21, 827 (2013).
2
G. J. Bauhuis, P. Mulder, E. J. Haverkamp, J. C. C. M. Uijben, and J. J.
Schermer, Sol. Energy Mater. Sol. Cell 93, 1488 (2009).
3
B. M. Kayes, H. Nie, R. Twist, S. G. Spruytte, F. Reinhardt, I. C. Kizilalli,
and G. S. Higashi, in Proceedings of 37th Photovoltaic Specialists
Conference (IEEE, Seattle, WA, 2011), p. 4.
4
M. A. Steiner, J. F. Geisz, I. Garcia, D. J. Friedman, A. Duda, and S. R.
Kurtz, J. Appl. Phys. 113, 123109 (2013).
5
D. Ding, S. R. Johnson, and Y.-H. Zhang, in Proceedings of 35th
Photovoltaic Specialists Conference (IEEE, Honolulu, HI, 2010), p. 2908.
6
D. Ding, S. R. Johnson, S.-Q. Yu, S.-N. Wu, and Y.-H. Zhang, J. Appl.
Phys. 110, 123104 (2011).
7
O. D. Miller, E. Yablonovitch, and S. R. Kurtz, IEEE J. Photovoltaics 2,
303 (2012).
8
E. D. Kosten, J. H. Atwater, J. Parsons, A. Polman, and H. A. Atwater,
Light: Sci. Appl. 2, e45 (2013).
9
X. Wang, M. R. Khan, J. L. Gray, A. Alam, and M. S. Lundstrom, IEEE J.
Photovoltaics 3, 737 (2013).
10
G. Lush and M. Lundstrom, Sol. Cells 30, 337 (1991).
11
J. J. Schermer, G. J. Bauhuis, P. Mulder, E. J. Haverkamp, J. van Deelen,
A. T. J. van Niftrik, and P. K. Larsen, Thin Solid Films 511–512, 645
(2006).
12
E. Yablonovitch and G. D. Cody, IEEE Trans. Electron Devices ED-29,
300 (1982).
13
M. Boroditsky, R. Ragan, and E. Yablonovitch, Sol. Energy Mater. Sol.
Cells 57, 1 (1999).
14
D. Liang, Y. Kang, Y. Huo, Y. Chen, Y. Cui, and J. Harries, Nano Lett.
13, 4850 (2013).
15
S. Liu, D. Ding, S. R. Johnson, and Y.-H. Zhang, Proc. SPIE 8256,
82560M (2012).
16
S. Liu, D. Ding, S. R. Johnson, and Y.-H. Zhang, in Proceedings of 38th
Photovoltaic Specialists Conference (IEEE, Austin, TX, 2012), p. 002082.
17
W. Yang, C. Allen, J.-J. Li, H. Cotal, C. Fetzer, S. Liu, D. Ding, S. Farrell,
Z. He, H. Li, H. Dettlaff, N. Karam, and Y.-H. Zhang, in Proceedings of
38th Photovoltaic Specialists Conference (IEEE, Austin, TX, 2012), p.
000978.
18
W. Yang, J. Becker, Y.-S. Kuo, J.-J. Li, S. Liu, B. Landini, K. Campman,
and Y.-H. Zhang, in Proceedings of 39th Photovoltaic Specialists
Conference (IEEE, Tampa, FL, 2013), p. 003329.
19
E. D. Palik, Handbook of Optical Constants of Solids (Academic, New
York, 1985).
20
S. Adachi, Optical Constants of Crystalline and Amorphous
Semiconductors: Numerical Data and Graphical Information (Kluwer
Academic Publishers, Massachusetts, 1999).
21
D. E. Aspnes, S. M. Kelso, R. A. Logan, and R. Bhat, J. Appl. Phys. 60,
754 (1986).
22
S. Adachi, J. Appl. Phys. 58, R1 (1985).
23
B. T. Phong, Commun. ACM 18, 311–318 (1975).
24
S. Liu, J. Becker, S. Farrell, W. Yang, and Y.-H. Zhang, in Proceedings of
39th Photovoltaic Specialists Conference (IEEE, Tampa, FL, 2013), p.
002105.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
209.147.144.15 On: Wed, 28 Jan 2015 18:27:58