Double-absorber thin-film solar cell with 34%

efficiency
Cite as: Appl. Phys. Lett. 117, 033901 (2020); [Link]
Submitted: 10 June 2020 . Accepted: 08 July 2020 . Published Online: 20 July 2020

Faiz Ahmad , Akhlesh Lakhtakia , and Peter B. Monk

Appl. Phys. Lett. 117, 033901 (2020); [Link] 117, 033901

© 2020 Author(s).
 Applied Physics Letters ARTICLE [Link]/journal/apl

Double-absorber thin-film solar cell with 34%

efficiency
Cite as: Appl. Phys. Lett. 117, 033901 (2020); doi: 10.1063/5.0017916
Submitted: 10 June 2020 . Accepted: 8 July 2020 .
Published Online: 20 July 2020

Faiz Ahmad,1 Akhlesh Lakhtakia,1,a) and Peter B. Monk2

AFFILIATIONS
1
Department of Engineering Science and Mechanics, Pennsylvania State University, University Park, Pennsylvania 16802, USA
2
Department of Mathematical Sciences, University of Delaware, Newark, Delaware 19716, USA

a)
Author to whom correspondence should be addressed: akhlesh@[Link]

ABSTRACT
Power-conversion efficiency is a critical factor for the wider adoption of solar-cell modules. Thin-film solar cells are cheap and easy to manu-
facture, but their efficiencies are low compared to crystalline-silicon solar cells and need to be improved. A thin-film solar cell with two
absorber layers (instead of only one), with bandgap energy graded in both, can capture solar photons in a wider spectral range. With a 300-
nm-thick CuIn1n1 Gan1 Se2 absorber layer and an 870-nm-thick Cu2 ZnSnðSn2 Se1n2 Þ4 absorber layer, an efficiency of 34.45% is predicted by
a detailed optoelectronic model, provided that the grading of bandgap energy is optimal in both absorber layers.
Published under license by AIP Publishing. [Link]

A photovoltaic solar cell consists of a metallic layer that serves generation of electricity for human progress to become truly uncon-
as an optical reflector as well as the electrical back-contact, a back- strained by energy economics.
passivation layer, at least two semiconductor layers in which electri- Thin-film solar cells can fill the need for in-device microwatt-
cal charges are generated by the absorption of solar photons, an scale ubiquitous generation of electricity. However, the efficiencies of
electrical front-contact layer, and one or two antireflection coatings thin-film solar cells are lower than those of c-Si solar cells. The highest
that are illuminated by the sun. Some solar cells have a buffer or reported efficiencies of thin-film CdTe, CuIn1n1 Gan1 Se2 (i.e., CIGS),
front-passivation layer as well, the role of any passivation layer being and Cu2 ZnSnðSn2 Se1n2 Þ4 (i.e., CZTSSe) solar cells are 21.0%, 22.6%,
to prevent recombination of two charges of opposite polarity. When and 12.6%, respectively,4 while the efficiency of the c-Si solar cell is
the solar cell is exposed to sunlight, photons with energy larger than 26.7%.4 Thin-film solar cells require improvements.
the bandgap energy are absorbed in the semiconductor layers. The In a typical solar cell, there is one thick semiconductor layer that
absorbed energy excites electrons in the valence band. The excited is the dominant site for photons to be absorbed and is, therefore, the
electrons move to the conduction band and leave holes behind in the major contributor to the electric current generated. The absorber layer
valence band. Thus, electron–hole pairs (EHPs) are created. When can be either an n-type or a p-type semiconductor. c-Si, CIGS, and
an electron and a hole recombine, energy is lost by conversion to CZTSSe solar cells have a single p-type thick absorber layer. The
heat and/or light. If a voltage is applied across the semiconductor bandgap energy of the absorber layer plays a crucial role in the effi-
layers, the electrons and the holes move in separate directions, creat- ciency of the solar cell to convert solar (photonic) energy into electrical
ing an electric current that depends on the density of impurity atoms energy. The short-circuit current density Jsc is high/low, but the open-
in each semiconductor layer. This is the basic principle of a photovol- circuit voltage Voc is low/high when the absorber layer has small/large
taic solar cell.1 bandgap energy.5 Hence, the efficiency g of a solar cell can be
Laudable technological and economic developments made on the improved by the optimal design of the absorber layer.
commercially dominant crystalline-silicon (c-Si) solar cells have dra- The bandgap energy of an absorber layer made of a compound
matically decreased investment expenditure,2 in line with what is semiconductor such as CIGS and CZTSSe can be optimally fixed
needed to cope with climate emergency.3 However, solar parks take within reasonable upper and lower bounds by correctly choosing the
land that could be used for other purposes such as farming, and the composition of the semiconductor.6–9 Thus, the bandgap energy of
transportation of electrical energy adds transmission losses. In addi- CIGS depends on n1 2 ½0; 1 and that of CZTSSe on n2 2 ½0; 1.
tion to a grid of solar parks, there is a need for in-device energy Compositional grading of the absorber layer (i.e., grading n1 for CIGS

Appl. Phys. Lett. 117, 033901 (2020); doi: 10.1063/5.0017916 117, 033901-1
Published under license by AIP Publishing
 Applied Physics Letters ARTICLE [Link]/journal/apl

and n2 for CZTSSe) during fabrication can be exploited to grade the single device with compatible fabrication techniques. CIGS and
bandgap energy of that layer in the thickness direction. Linear grading CZTSSe are almost lattice matched12,13 and can be fabricated using
of the bandgap energy of the absorber layer has been experimentally vapor deposition techniques.14
shown to improve the open-circuit voltage of CIGS solar cells.6,9 Hence, the double-absorber CIGS-CZTSSe solar cell shown in
Similarly, it has been experimentally demonstrated that Jsc can be Fig. 1(c) is proposed and theoretically studied in this communication,
improved without reducing Voc by grading the bandgap energy of the using a detailed optoelectronic model15 coupled with the differential
absorber layer in CZTSSe solar cells.7 Detailed optoelectronic model- evolution algorithm (DEA)16 for optimization. In conformance with
ing indicates that the proper grading of the bandgap energy of the existing thin-film solar cells, the thicknesses of various layers were
absorber layer can enhance both Voc and Jsc in thin-film solar cells;10,11 fixed as follows: LMgF2 ¼ 110 nm, LAZO ¼ 100 nm, LZnO ¼ 80 nm,
efficiencies as high as 27.70% and 21.74% have been predicted for LCdS ¼ 70 nm, LAl2 O3 ¼ 10 nm, and LMo ¼ 500 nm. Also in confor-
CIGS solar cells10 and CZTSSe solar cells,11 respectively. mance with existing solar cells, the thicknesses LCIGS  2200 nm and
Even with bandgap-energy grading, the CIGS and CZTSSe LCZTSSe  2200 nm of the two absorber layers were kept variable.
absorber layers absorb only a part of the optical energy available in the With the exposed surface of the MgF2 layer identified as the plane
solar spectrum. The bandgap energy of CIGS can be varied between z ¼ 0 and the z axis pointing into the solar cell (as shown in Fig. 1), the
0.947 eV and 1.626 eV, and that of CZTSSe between 0.91 eV and z-dependent bandgap energy (in eV) was modeled in the CIGS layer
1.49 eV. One way to absorb solar photons in a wider spectral range is as10
to combine a CIGS solar cell and a CZTSSe solar cell in a tandem
Eg ðzÞ ¼ Ea1 þ A1 ðEb1  Ea1 Þ
structure.5 However, the current densities created in the two constitu-       a1
ent solar cells will be different, and a two-terminal device with both 1 z  L1
 sin 2p K1  w1 þ1 ;
solar cells in series will not be efficient.5 Combining a CIGS solar cell 2 LCIGS
and a CZTSSe solar cells in a four-terminal device will require addi- z 2 ½L1 ; L2 ; (1)
tional circuitry to be fabricated and managed, resulting in parasitic
losses, effectively reducing the overall efficiency of the tandem solar and in the CZTSSe layer as11
cell. We propose here another option to harvest photons over a wider
Eg ðzÞ ¼ Ea2 þ A2 ðEb2  Ea2 Þ
spectral range and improve the performance of thin-film solar cells.       a2
The structures of the CIGS and CZTSSe solar cells—shown in 1 z  L2
 sin 2p K2  w2 þ1 ;
Figs. 1(a) and 1(b), respectively—are identical, except for having a p- 2 LCZTSSe
type absorber layer made of a specific material. Solar cells of both types z 2 ½L2 ; L3 ; (2)
have a molybdenum (Mo) back-contact layer, an aluminum-oxide
(Al2 O3 ) back-passivation layer, the absorber layer of a p-type semi- where Eb1 ¼ 1:626 eV, Eb2 ¼ 1:49 eV, L1 ¼ LMgF2 þ LAZO þ LZnO
conductor, a semiconductor layer of n-type cadmium sulfide (CdS), an þLCdS , L2 ¼ L1 þ LCIGS , and L3 ¼ L2 þ LCZTSSe . Whereas Eg ðzÞ in
oxygen-deficient zinc-oxide (od-ZnO) front-passivation layer, a front- the CIGS layer can be engineered through n1 ðzÞ,6 Eg ðzÞ in the
contact layer made of aluminum-doped zinc oxide (AZO), and an CZTSSe layer can be engineered through n2 ðzÞ.12,17
antireflection coating of magnesium fluoride (MgF2 ). Equations (1) and (2) can represent a wide variety of bandgap-
If absorber layers of both types were present in a single two- energy profiles. Based on experience, optimization was carried out in
terminal solar cell, then parasitic impedances and additional circuitry the parameter space defined as follows: Ea1 2 ½0:947; 1:626 eV,
will be avoided. However, the absorbers of both types must have mini- A1 2 ½0; 1, K1 2 ½0; 8; w1 2 ½0; 1; a1 2 ½0; 8, Ea2 2 ½0:91; 1:49
mum lattice difference and should be capable of being deposited in a eV, A2 2 ½0;1, K2 2 ½0;8; w2 2 ½0;1, a2 2 ½0;8, LCIGS 2 ½0;2200nm,
LCZTSSe 2 ½0;2200 nm, and 0 < LCIGS þ LCZTSSe  2200nm.
The optoelectronic model15 has a photonic step and an elec-
tronic step. In the photonic step, the transfer-matrix method18,19 was
used to determine the electric and magnetic fields everywhere inside
the solar cell due to normally incident monochromatic radiation.
The transfer-matrix method is an efficient technique to solve the
frequency-domain Maxwell equations. With the assumption that the
absorption of every photon in the ZnO, CdS, CIGS, and CZTSSe
layers of the double-absorber solar cell excites an EHP, the EHP gen-
eration rate G(z) was determined in those layers,10 assuming normal
illumination by unpolarized polychromatic light endowed with the
AM1.5G solar spectrum.20 The frequency-dependent relative permit-
tivity of every material in the double-absorber solar cell is available
elsewhere.10,11,21
In the electronic step, G(z) was used as an input to the 1D drift-
FIG. 1. Schematic of the thin-film solar cell based on (a) CIGS absorber layer, (b)
diffusion equations1 applied to the ZnO, CdS, CIGS, and CZTSSe
CZTSSe absorber layer, and (c) CIGS-CZTSSe double-absorber layer. The thick-
ness of the absorber layer in both (a) and (b) equals 2200 nm for the highest effi- layers. These equations for the gradients of the electron current density
ciency reported in the literature.4 The optimal thicknesses of the two absorber Jn ðzÞ and the hole current density Jp ðzÞ also contain the electron–hole
layers in (c) are predicted in this paper. recombination rate R(z). The nonlinear Shockley–Read–Hall (SRH)

Appl. Phys. Lett. 117, 033901 (2020); doi: 10.1063/5.0017916 117, 033901-2
Published under license by AIP Publishing
 Applied Physics Letters ARTICLE [Link]/journal/apl

and radiative contributions to R(z) were incorporated for each of the If the CZTSSe absorber layer is absent but LCIGS ¼ 300 nm,
four semiconductor layers. In addition, the 1D Poisson equation for the highest efficiency predicted is 19.01%, Jsc ¼ 25:98 mA cm2,
the dc electric potential was simultaneously considered with the elec- Voc ¼ 1023 mV, and FF ¼ 0:73; these values were obtained with
tron density n(z), the hole density p(z), and the fixed charge density Ea1 ¼ 0:95 eV, A1 ¼ 0:98, K1 ¼ 1:5; w1 ¼ 0:74, and a1 ¼ 6. If the
(including the doping density and trap/defect density) contributing to CIGS absorber layer is absent but LCZTSSe ¼ 870 nm, the highest effi-
the source term.15 The Boltzmann approximation was adopted. Both ciency is 21.74%, Jsc ¼ 37:39 mA cm2, Voc ¼ 772 mV, and
the front- and the back-contact layers were modeled as ideally Ohmic, FF ¼ 0:75; these values were obtained with Ea2 ¼ 0:92 eV,
and local quasi-thermal equilibrium was assumed to determine A2 ¼ 0:98, K2 ¼ 2; w2 ¼ 0:75, and a2 ¼ 6. Notice that the optimal
boundary conditions. All photo-generated minority carriers in both double-absorber solar cell outperforms both single-absorber cells in all
CIGS and CZTSSe absorber layers were taken to be collected across four performance parameters: g, Jsc , Voc , and FF. The double-absorber
the CdS/CIGS junction. The effect of band-tail states in CZTSSe was solar cell appears to derive the high value of Jsc from the CZTSSe
incorporated by reducing the bandgap energy of CZTSSe for electrical absorber layer and the high value of Voc from the CIGS absorber layer.
calculations.11 Surface recombination was neglected because the incor- Equations (1) and (2) encompass both absorber layers being
poration of surface traps/defects has minimal consequences.10,11 homogeneous in the space of the parameters chosen for optimization.
Electrical data for ZnO, CdS, CIGS, and CZTSSe are available Therefore, the highest efficiency with the graded-bandgap-energy
elsewhere.10,11,21 absorber layers will necessarily exceed (or equal) the highest efficiency
A set of six nonlinear differential equations15 was solved using a with homogeneous-bandgap-energy absorber layers. Indeed, if we fix
hybridizable discontinuous Galerkin (HDG) scheme22,23 to determine LCIGS ¼ 300 nm, LCZTSSe ¼ 870 nm, and A1 ¼ A2 ¼ 0, the highest
the z-independent current density Jdev ¼ Jn ðzÞ þ Jp ðzÞ and the electri- efficiency predicted is 11.08% with Ea1 ¼ 0:95 eV and Ea2 ¼ 0:91 eV;
cal power density P ¼ Jdev Vext as functions of the bias voltage Vext correspondingly, Jsc ¼ 35:90 mA cm2, Voc ¼ 455 mV, and
under steady-state conditions. In turn, the Jdev –Vext and the P–Vext FF ¼ 0:67. If LCIGS ¼ 300 nm and LCZTSSe ¼ 0, the highest efficiency
curves yielded Jsc ; Voc , g, and a figure of merit called the fill factor predicted is 11.59% with Ea1 ¼ 1:25 eV, Jsc ¼ 22:56 mA cm2, Voc
FF 2 ½0; 1 which should be as high as possible.1 The model has been ¼ 681 mV, and FF ¼ 0:76. Conversely, if LCIGS ¼ 0 and LCZTSSe
validated against experimental results.10,11 ¼ 870 nm, the highest efficiency predicted is 11.84% with Ea2 ¼ 1:20
Finally, the widely used DEA16 was adopted to maximize g in an eV, Jsc ¼ 30:13 mA cm2, Voc ¼ 558 mV, and FF ¼ 0:70. Thus, even
11-dimensional parameter space formed by Ea1 , A1, K1 , w1, a1, Ea2 , though the double-absorber solar cell exceeds both single-absorber
A2 , K2, w2, a2, LCIGS , and LCZTSSe . The DEA was implemented using solar cells in Jsc , it underperforms both in Voc so much so that its effi-
R
MATLABV version R2019a and run over a search time of eight weeks. ciency is somewhat lower than either’s. Grading of the bandgap energy
Given an initial guess in the search space, the underlying strategy in of both absorber layers is the key to significantly higher efficiency.
the metaheuristic DEA is to improve the candidate solution at every The profiles of Eg ðzÞ, G(z), R(z), n(z), and p(z) in the ZnO/CdS/
iteration step and does not require explicit gradients of the cost func- CIGS/CZTSSe region of the optimal double-absorber solar cell are
tion (i.e., g). depicted in Fig. 2. Whereas G(z) is higher in regions with lower
The highest value of g predicted for the double-absorber bandgap energy and vice versa, R(z) is higher in regions with higher
solar cell is 34.45%; correspondingly, Jsc ¼ 38:11 mA cm2, Voc Eg ðzÞ due to higher fixed charge density caused by higher gallium or
¼ 1085 mV, and FF ¼ 0:83. The optimal thicknesses of the absorber sulfur content in those regions.
layers in the double-absorber solar cell are LCIGS ¼ 300 nm and Although Eg is independent of z in both the ZnO and CdS layers
LCZTSSe ¼ 870 nm. The corresponding bandgap-energy parameters (by design), it varies with z in both absorber layers. This variation
are as follows: Ea1 ¼ 0:95 eV, A1 ¼ 0:91, K1 ¼ 1:88; w1 ¼ w2 comprises constant-Eg regions separated by regions with large Eg gra-
¼ 0:75; a1 ¼ a2 ¼ 6, Ea2 ¼ 0:91 eV, A2 ¼ 0:99, and K2 ¼ 2. The dients. The bandgap energy is low in the constant-Eg regions, these
efficiency drops to no less than 34.43% if any of the foregoing optimal regions being responsible for elevating the EHP generation rate
bandgap-energy parameters is altered by 1%. because less energy is required to excite an EHP across a narrower
Whereas the standard absorber thickness is about 2200 nm in the bandgap.29 Figure 2(b) confirms that G(z) exceeds R(z) in the con-
single-absorber CIGS and CZTSSe solar cells,24,25 both absorbers in stant-Eg regions.
the optimal double-absorber solar cells are together only 1170-nm The large Eg gradient close to the back surface in the CZTSSe
thick. The absorber thickness should be less than the diffusion length layer enhances the drift field to reduce the back-surface recombination
of the minority carriers so that they have a reasonable probability of rate, thereby supplementing the role of the Al2 O3 passivation
being collected. The diffusion length of compositionally graded CIGS layer.30,31 Since the bandgap energy is high close to both faces of each
varies from 500 nm to 2000 nm,26 so that LCIGS ¼ 300 nm in the opti- absorber layer, Voc is high in the optimal design.30,32,33 The triangular
mal double-absorber solar cell is definitely lower. However, the minor- regions in the middle of each absorber layer in Fig. 2(a) also create an
ity carriers generated in the CZTSSe layer also need to traverse the additional drift field that favors charge-carrier collection deep inside
CIGS region to be collected across the p–n junction (i.e., the CdS/ the absorber layer.8
CIGS interface). As the minority-carrier diffusion length is around The profiles of n(z) and p(z) in Fig. 2(c) are discontinuous at the
1200 nm in CZTSSe,27 even the minority carriers generated deep CIGS/CZTSSe interface. The discontinuity in the minority charge-
inside the CZTSSe layer have diffusion lengths longer than or compa- carrier density n(z) is considerably smaller than in the majority
rable to LCIGS þ LCZTSSe ¼ 1170 nm. Furthermore, the drift-field cre- charge-carrier density p(z). These are consistent with the small discon-
ated by the Eg -gradient helps to accelerate the minority carriers tinuity of the conduction band energy Ec ðzÞ and the large discontinu-
toward the p–n junction.8 ity in valence band energy Ev ðzÞ (both not shown), which make

Appl. Phys. Lett. 117, 033901 (2020); doi: 10.1063/5.0017916 117, 033901-3
Published under license by AIP Publishing
 Applied Physics Letters ARTICLE [Link]/journal/apl

FIG. 3. Plots of Jdev and P vs Vext of the optimal double-absorber solar cell. The
values of Jdev and Vext for maximum P are identified.

based solar cell is promising as a cheap option for the ubiquitous har-
nessing of solar energy.
To conclude, CIGS solar cells deliver higher efficiency than
CZTSSe solar cells whether the bandgap energy in the absorber layer
is homogeneous4 or graded.10,11 But CIGS contains indium which is
not plentiful on our planet, whereas no constituent element of
CZTSSe suffers from that constraint. If absorber layers of both CIGS
and CZTSSe are used in a single solar cell, then the efficiency can be
boosted highly to 34.45% with a fill factor of 0.83, provided the
bandgap energy is optimally graded, with the CIGS layer being only
300-nm thick and the CZTSSe layer being 870-nm thick. Fabrication
of this double-absorber thin-film solar cell, or an approximative var-
iant thereof, will require the attention of experimentalists for careful
compositional grading and may not perform as well as predicted by
a detailed optoelectronic model. Nevertheless, this solar cell is prom-
ising for ubiquitous in-device microwatt-scale generation of
electricity.

This research was supported by (i) the U.S. National Science

Foundation under Grant Nos. DMS-1619901 and DMS-1619904
and (ii) the Charles Godfrey Binder Endowment at Penn State.
FIG. 2. Profiles of (a) Eg ðzÞ, (b) G(z) and R(z), and (c) n(z) and p(z) in the ZnO/
CdS/CIGS/CZTSSe region of the optimal double-absorber solar cell. DATA AVAILABILITY
The data that support the findings of this study are available
Eg ðzÞ ¼ Ec ðzÞ  Ev ðzÞ also discontinuous at the CIGS/CZTSSe within the article.
interface.
The Jdev –Vext characteristics of the optimal double-absorber solar REFERENCES
1
cell are shown in Fig. 3. The optoelectronic model predicts that J. Nelson, The Physics of Solar Cells (Imperial College Press, London, UK,
the solar cell should be operated with Vext ¼ 965 mV to deliver Jdev 2003).
2
¼ 35:71 mA cm2; then 34.45 mW cm2 is predicted as the maxi- 3
M. A. Green, Joule 3, 631 (2019).
Drawdown, edited by P. Hawken (Penguin, New York, NY, 2017).
mum extractable power density when the incident solar flux is 4
M. A. Green, E. D. Dunlop, J. Hohl-Ebinger, M. Yoshita, N. Kopidakis, and A.
100 mW cm2. Since FF is high, g ¼ 33% even when Vext is reduced W. Y. Ho-Baillie, Prog. Photovoltaics: Res. Appl. 28, 3 (2020).
to 900 mV. 5
M. A. Green, Prog. Energy 1, 013001 (2019).
The commercially dominant c-Si solar cell delivers at most an 6
C. Frisk, C. Platzer-Bj€ orkman, J. Olsson, P. Szaniawski, J. T. W€atjen, V.
efficiency of 26.7%4 with 100-lm-thick highly pure silicon wafers. Fj€allstr€
om, P. Salome, and M. Edoff, J. Phys. D 47, 485104 (2014).
7
III–V multijunction solar cells have been shown experimentally to K. Woo, Y. Kim, W. Yang, K. Kim, I. Kim, Y. Oh, J. K. Kim, and J. Moon, Sci.
deliver the record efficiency of 46% with concentrated sunlight (508 Rep. 3, 03069 (2013).
8
J. A. Hutchby, Appl. Phys. Lett. 26, 457 (1975).
suns),28 but they are prohibitively expensive for terrestrial applications 9
S. Schleussner, U. Zimmermann, T. W€atjen, K. Leifer, and M. Edoff, Sol.
because they use scarce materials and require a costly and time- Energy Mater. Sol. Cells 95, 721 (2011).
consuming manufacturing process (metal–organic vapor–phase epi- 10
F. Ahmad, T. H. Anderson, P. B. Monk, and A. Lakhtakia, Appl. Opt. 58, 6067
taxy) as well as expensive substrates.5 A 34.45%-efficient thin-film (2019); 59, 2615 (2020).

Appl. Phys. Lett. 117, 033901 (2020); doi: 10.1063/5.0017916 117, 033901-4
Published under license by AIP Publishing
 Applied Physics Letters ARTICLE [Link]/journal/apl

11 23
F. Ahmad, A. Lakhtakia, T. H. Anderson, and P. B. Monk, J. Phys.: Energy 2, F. Brezzi, L. D. Marini, S. Micheletti, P. Pietra, R. Sacco, and S. Wang, Handb.
025004 (2020); 2, 039501 (2020). Numer. Anal. 13, 317 (2005).
12 24
S. Adachi, Earth-Abundant Materials for Solar Cells (Wiley, Chichester, West P. Jackson, R. Wuerz, D. Hariskos, E. Lotter, W. Witte, and M. Powalla, Phys.
Sussex, UK, 2015). Status Solidi RRL 10, 583–586 (2016).
13 25
T. Klinkert, M. Jubault, F. Donsanti, D. Lincot, and J.-F. Guillemoles, Thin W. Wang, M. T. Winkler, O. Gunawan, T. Gokmen, T. K. Todorov, Y. Zhu,
Solid Films 558, 47 (2014). and D. B. Mitzi, Adv. Energy Mater. 4, 1301465 (2014).
14 26
K. L. Chopra, P. D. Paulson, and V. Dutta, Prog. Photovoltaics: Res. Appl. 12, G. Brown, V. Faifer, A. Pudov, S. Anikeev, E. Bykov, M. Contreras, and J. Wu,
69 (2004). Appl. Phys. Lett. 96, 022104 (2010).
15 27
T. H. Anderson, B. J. Civiletti, P. B. Monk, and A. Lakhtakia, J. Comput. Phys. T. Gokmen, O. Gunawan, and D. B. Mitzi, J. Appl. Phys. 114, 114511 (2013).
28
407, 109242 (2020). F. Dimroth, T. N. D. Tibbits, M. Niemeyer, F. Predan, P. Beutel, C. Karcher, E.
16
R. Storn and K. Price, J. Global Optim. 11, 341 (1997). Oliva, G. Siefer, D. Lackner, P. Fuß-Kailuweit, A. W. Bett, R. Krause, C.
17
A. Kanevce, I. Repins, and S.-H. Wei, Sol. Energy Mater. Sol. Cells 133, 119 Drazek, E. Guiot, J. Wasselin, A. Tauzin, and T. Signamarcheix, IEEE J.
(2015). Photovoltaics 6, 343–349 (2016).
18 29
D. W. Berreman, J. Opt. Soc. Am. 62, 502 (1972). S. J. Fonash, Solar Cell Device Physics, 2nd ed. (Academic, Burlington, MA,
19
T. G. Mackay and A. Lakhtakia, The Transfer-Matrix Method in 2010).
30
Electromagnetics and Optics (Morgan & Claypool, San Rafael, CA, USA, 2020). T. Dullweber, O. Lundberg, J. Malmstr€ om, M. Bodegård, L. Stolt, U. Rau, H.
20
National Renewable Energy Laboratory, “Reference Solar Spectral Irradiance: Air W. Schock, and J. H. Werner, Thin Solid Films 387, 11 (2001).
31
Mass 1.5,” [Link] (accessed 9 June, 2020). P. Casper, R. H€ unig, G. Gomard, O. Kiowski, C. Reitz, U. Lemmer, M. Powalla,
21
F. Ahmad, “Optoelectronic modeling and optimization of graded-bandgap and M. Hetterich, Phys. Status Solidi RRL 10, 376 (2016).
32
thin-film solar cells,” Ph.D. dissertation (Pennsylvania State University, K.-J. Yang, D.-H. Son, S.-J. Sung, J.-H. Sim, Y.-I. Kim, S.-N. Park, D.-H. Jeon, J.
University Park, PA, 2020). Kim, D.-K. Hwang, C. W. Jeon, D. Nam, H. Cheong, J.-K. Kang, and D.-H.
22
D. Brinkman, K. Fellner, P. Markowich, and M.-T. Wolfram, Math. Models Kim, J. Mater. Chem. A 4, 10151 (2016).
33
Methods Appl. Sci. 23, 839 (2013). M. Gloeckler and J. R. Sites, J. Phys. Chem. Solids 66, 1891 (2005).

Appl. Phys. Lett. 117, 033901 (2020); doi: 10.1063/5.0017916 117, 033901-5
Published under license by AIP Publishing