Simulation of CZTSSe thin film solar cells in COMSOL: 3D coupled optical,

electrical, and thermal model

Soma Zandia , Prateek Saxenab , Mohammad Razaghia , Nima E. Gorjic,∗

a Department of Electrical Engineering, Faculty of Engineering, University of Kurdistan, Sanandaj, Iran
b Sustainable Manufacturing Systems Centre, School of Aerospace, Transport and Manufacturing, Cranfield University, Bedfordshire, UK
c Faculty of Electrical and Electronics Engineering, Ton Duc Thang University, Ho Chi Minh City, VietNam

Abstract
The Cu2 ZnSnSx Se4−x (CZTSSe) thin film solar cells have attracted the attention of researchers due to it’s earth-
abundant composition containing Copper, Zinc, Tin and Sulfur and Selenide with 12.6% record efficiency (2013-IBM). A
3D simulation analysis is presented here on the optical, electrical and thermal characteristics of CZTSSe solar cell using
COMSOL Multiphysics 3D simulation package. COMSOL is capable of calculating the coupled optical-electrical-thermal
models through Electromagnetic Wave, Semiconductor, and Heat Transfer modules for a finely meshed structure. Using
this capability, we have calculated the optical photo-generation rate of the a Mo/Mo(S,Se)2 /CZTSSe/CdS/ZnO/ITO/air
structure by inserting the refractive index and extinction coefficient of every layer in Wave optic module in COMSOL.
We also calculated the total optical generation rate for two structures with and without Mo(S,Se)2 layer at the junction
of Mo and CZTSSe layers. The current-voltage curve, electric field profile and the recombination rate of the cell has
also been calculated by Semiconductor module coupled to wave optic module. The current-voltage characteristics shows
an improvement in Voc for the cell with Mo(S,Se)2 layer (0.46 V to 0.513 V) which was also suggested by IBM for a
record cell efficiency. Finally, the thermal maps of the cell has been calculated by Heat Transfer Module coupled to
Semiconductor module considering the Shockley-Read-Hall (SRH) recombination heat, Joule Heat and conductive heat
flux. The total heat flux magnitude of the cell was also mapped as a result out of these heat generation and cooling
sources. The SRH heat is maximum within the depletion width at the CZTSSe/CdS interface whereas the Joule heating
is intensive at the Mo/Mo(S,Se)2 /CZTSSe side. Interesting is to see that the heat is mainly conducted to environment
from Mo side presented by the conductive heat map. The total heat flux is intensive at both top and bottom interfaces
which means the heat is generated at both top and bottom sides of the cells and not only from the illuminated part.
Keywords: CZTSSe, Thin films, Solar cell, COMSOL,3D Simulation.

1. Introduction and simulation analysis are also reported in the literature

to gain insight into the photo-absorption, carrier genera-
Kesterite Thin film solar cells (based on CZTSSe mate- tion, carrier transport and thermal stability of such de-
rials) attracted the attention of many research groups as an vices [3, 6, 7]. Most of these simulation analysis have been
alternative material for Cu(In,Ga)Se2 chalcogenides since performed by 1D simulation platforms such as SCAPS or
In is not an abundant element in earth. CZTSSe thin films AMPS which cannot consider the 3D features of a solar cell
have slowly raised their efficiency from 8% to 12% within [8, 9]. However, 1D simulators do not allow thermal anal-
few years [1]. The record efficiency of 12.6% patented by ysis and mapping the heat distribution or heat emission
IBM company in 2013 has not been beaten yet [2]. The to environment and conduction to other layers. COMSOL
current challenge with CZTSSe devices is the low open- multiphysics simulation package has been rarely used for
circuit voltage (Voc ) of 0.513 V (IBM-2013). This shows a simulation of CZTSSe devices [10, 11]. COMSOL con-
large difference of (Eg /q) − Voc in compare to CIGSSe so- tains semiconductor module for solving the drift-diffusion
lar cells reported in [2, 3]. Currently, the low open-circuit model of the carrier transport and also the Electromag-
voltage of CZTSSe solar cells is low (513 mV) compared netic Waves module to map the optical photo-generation
to CdTe and CIGS thin film cells which is around 900- of the cell, as well as the heat-transfer module for ther-
1000 mV. Fabrication process, phase formation, grain size mal distribution mapping across the device structure [12].
and morphology of these cells are still under excessive re- COMSOL also allows to couple these modules to simul-
search by many research groups [4, 5]. However, modeling taneously map the photo-generation, recombination and
the heat profile of the cell under the operation condition
∗
Email address: Corresponding author:
and in 3D mode. We have recently developed a coupled
[email protected] (Nima E. Gorji) optical-electrical-thermal module for the perovskite solar
 Figure 2: a. The mesh quality for skewness (overall geometry), and
b. Mesh quality for Mo(S,Se)2 layer

Figure 1: a. The schematic and geometry of the CZTSSe thin film

solar cell as described in Ref. [2], and, b. the meshed structure. the structure presented in a reputed literature report in
Ref. [2] given by Mo/Mo(S,Se)2 /CZTSSe/CdS/ZnO/ITO
structure which has given the world record efficiency of
cell including a cell with reduced Graphene oxide electrode 12.6%. We have calculated the coupled optical and elec-
[13, 14]. trical simulation to include in the photo-absorption, car-
In this paper, we present our numerical simulation anal- rier photo-generation, carrier collection, and efficiency of
ysis of the carrier transport, optical generation and ther- this device. For this purpose three multi-physics modules
mal/heat distribution map of the CZTSSe thin film solar were coupled namely, Electromagnetic Waves (Frequency
cells using COMSOL package. The optical generation rate domain), Semiconductor module, and the Heat Transfer
across the cell structure, electric field profile, recombina- in Solid modules. The Electromagnetic Waves module re-
tion rate and it’s impact on device performance, as well quired the optical constants, illumination intensity and
as the heat distribution at a static steady-state has been the perfect conductors properties to calculated the total
simulated. We present a 3D analysis by solving a coupled photo-generation (Gtot ). The semiconductor module re-
optical-electrical-thermal profiles which has been rarely quired the materials and electrical parameters and includes
presented in literature. The two structures were compared the trap assisted recombination and Shockley-Read-Hall:
in this paper: the cell with and without Mo(S,Se)2 layer at SRH recombination. This module uses the Gtot to solve
the bottom junction of Mo and CZTSSe layers. This layer the current-continuity equation as a result of coupling with
was suggested by IBM in 2013 to improve the open-circuit Semiconductor and Electromagnetic wave modules. Heat
voltage of the cell without losing the short-circuit current transfer in solid module required the heat coefficients of
density.To our best of knowledge, it’s the first time that the layers as listed in Table 1 and calculates the Joule
the thermal mapping of a CZTSSe thin film solar cell (at heat, SRH heat and temperature distribution across the
least in 3D) has been presented in literature. cell structure. The thermal/heat coefficients were taken
from appropriate references. The heat coefficient of Mo
was taken from Ref. [15, 16].
2. Physics, Mesh & Simulation
The Cu2 ZnSnSx Se4−x (CZTSSe) thin film solar cell is 2.1. Model Geometry
modelled using COMSOL Multiphysics 5.4a. Fig. 1 shows The solar cell model is made up of seven layers in to-
the device structure which is made of up to seven lay- tal. The top-most layer is modelled as air, extending up
ers on a glass substrate (not shown here). We selected to 800 nm in height. The input power of the plane wave
2
Table 1: The values of selected electrical and thermal parameters
used in COMSOL as partially reported in Refs. [2, 15, 16, 17, 18,
19, 23]
.
Parameter Mo Mo(S,Se)2 CZTSSe CdS ZnO ITO
Thickness (nm) 380 170 2000 50 50 300
εr - 10 13.6 9 9 9
Nc (cm−3 ) - 0.7 × 1018 2.2 × 1018 2.2 × 1018 2.2 × 1018 -
Nv (cm−3 ) - 0.3 × 1019 1.8 × 1019 1.8 × 1019 1.8 × 1019 -
µn /µp (cm2 / (V.s)) - 100/25 100/25 100/25 100/25 -
χ(eV) - 4.14 4.2 4.2 4.4 -
Bandgap (eV) - 1.1 1.13 2.4 3.3 -
NA (cm−3 ) - 1 × 1015 1 × 1015 1 × 1013 1 × 1013 -
ND (cm−3 ) - 5 × 1013 5 × 1013 5 × 1017 1 × 1018 -
τn /τp (ns) - 2.4/ 2.4 5.4/5.4 5 × 10−3 /5 × 10−3 3 × 10−3 / 3×10−3 -
ρ(kg/m3 ) 4690 6900 4560 4820 5610 7120
K(W/m.K) 138 103 2.95 6.2 23.4 10
Cp (J/kg.K) 277 100-130 410 210 494 340-400
h(W/(m2 K)) 363 × 106 605 × 106 14.75 × 104 124 × 106 468 × 106 33 × 106
Figure 3: The total Photo-generation rate (Gtot ) across the cell
structure in a) 2D and b) 3D format out of our optical calculations
in COMSOL.
(300–1000 nm) was AM1.5 G, with normal incident angle
only. This layer is followed by ITO layer extending upto
a depth of 300 nm. The interface between ITO layer and
CZTSSe layer is made by modelling ZnO and CdS layer Skewness close to or equal to 1 is indicative of an ideal
for a thickness of 50 nm each. The thickness of CZTSSe mesh element while skewness close to 0 represents a de-
layer is kept to be 2000 nm. This layer is further followed generated element. Thus, based on the skewness quality
by MO(S,Se)2 layer upto 170 nm in height followed by metric, the defined mesh is found suitable for carrying out
the last layer (Mo) which is 380 nm. The overall dimen- further analysis.
sion of the cell is 400 nm × 400 nm × 3750 nm. The
cell is assumed to be operating at the room temperature. 3. Results & Discussion
Material properties are imported from COMSOL Multi-
physics’ material database and from the cited literature 3.1. Optical Simulation
[2, 15, 16, 17, 18, 19, 23]. The optical photo-generation map of the cell has been
calculated by inserting the optical constants of every layer
2.2. Meshing Configuration in COMSOL environment. The refractive index and ex-
The mesh configuration is similar to the previous we tinction coefficient of the layers were extracted from Refs.
presented in Ref. [14]. As seen in Fig. 1 user-controlled [17, 18, 19] following a Beer-Lambert optical transmission
mesh is specified for the geometry. The maximum element law. To enhance the photo-absorption in CZTS layer, the
size is set to be 0.118 µm and minimum element size equal reflection must be minimized. However, the top layers
to 50.7 nm. The maximum element growth rate is defined (CdS, ZnO and ITO) have an absorption and this cre-
to be 1.35 with curvature factor as 0.3. The resolution of ates a trade-off between minimizing surface reflection and
narrow regions is specified to be 0.85. A swept type mesh absorption in the thickness of these top layers. Fig. 3
is defined for all the domains with face-meshing method presented the photo-generation map of the cell for a range
set to quadrilateral. The tessellation method is set as ’au- of wavelengths between 300 nm and 1000 nm. The Gtot
tomatic’. The distribution type of the mesh elements in across the cell structure with and without Mo(S,Se)2 . We
ZnO layer, is defined to be fixed with number of elements have also compared the generation rate across the cell with
as 5. For CdS layer, in addition to fixed number of ele- and without the Mo(S,Se)2 . The photogeneration has not
ments, an element ratio of 1 is specified. These two layers changed significantly since most photons were already ab-
are the thinnest of all the other layers in geometry and sorbed in CZTSSe layer and do not reach the Mo(S,Se)2 .
thus need to be very finely meshed. The number of el- Also, the latter is thin enough to have a negligible impact
ements and element ratio for CZTSSe and Mo(S,Se)2 is on Gtot . The same peak has been observed at the inter-
set to 70 and 0.1 respectively. Quality metrics are chosen face of CZTSSe for other CZTSSe/CdS based solar cells
for ensuring the mesh consistency. A measure of quality in Fig. 3d in Ref. [20]. Saha et al. have also reported
can be described based on skewness, maximum angle, vol- that Gtot has a high peak at the CZTSe side of the inter-
ume versus circumradius, volume versus length, growth face since CZTSSe has a higher absorption compared to
rate and condition number. In the scope of this work, CdS [20]. In fact, CdS layer has a wider bandgap and ab-
skewness quality metric is chosen. The mesh quality is sorbs high-energy photons not absorbed in CZTSSe layer
shown in Fig. 2. The basis of choosing this metric over and the photons with higher energy than the bandgap of
others is because this metric is considered to be highly re- CdS. Also, Abdelraouf et al. have simulated the optical
liable and is most popular in modelling techniques. The photogeneration of CZTS based thin film solar cells based
metric defines the mesh elements according to their angu- on coupled optical and electrical modeling (in COMSOL)
lar skewness. Elements with deviations from the optimal and reported a same trend of 3D mapping for the Gtot at
mesh (with regard to large or small angles) are penalized. λ=300 nm and 800 nm [11].
3
 Figure 5: a) The J-V curves were calculated for both structures
Figure 4: The Electric Field profile as extends across the cell for two with and without Mo(S,Se)2 layer at the bottom junction of the Mo
structures with and without Mo(S,Se)2 , b) the electric field in 3D and CZTSSe. The Mo work-function was set to 4.8 eV in electrical
calculated from COMSOL electrical simulations coupled with optical simulation. b) the 3D map of the SRH non-radiative recombination
module. rate.

3.2. Electrical Simulation characteristics of the two CZTSSe cell structures with and
The Doping profile and electrical characteristics of the without Mo(S,Se)2 layer (Fig. 5a). The results are in
modelled solar cell are presented in Table 1. At Mo/CZTSSe agreement with the record device metrics of the CZTSSe
junction a Schottky barrier is built due to imbalance be- cell reported by IBM [2]. The IBM researchers reported
tween the work functions of the Mo and CZTSSe layers. Voc = 0.46 V without Mo(S,Se)2 layer which jumped to
This barrier prevents the carrier transport to the Mo elec- 0.513 V after adding this buffer layer at the junction of
trode and reduces the Voc . To reduce the Schottky barrier, Mo/CZTSSe. Also, the 3D map of the SRH non-radiative
a thin layer of Mo(S,Se)2 is introduced between the Mo and recombination rate has been shown in Fig. 5b where the
CZTSSe [21]. This thin layer will efficiently increase the maximum recombination occurs within the depletion width
Voc by boosting the carrier collection at this junction. For of the CZTSSe layer near the top interface. The SRH map
band diagram analysis, please refer to Ref. [17] especially at the CZTSSe/CdS junction explains the level
The profile of the electric-field across the cell thickness of obtained Voc .
has been also calculated in 2D as shown in Fig. 4a. By
comparing the slope of electric field of the two structures 3.3. Thermal Simulation
in back contact regions, we observed that the Mo(S,Se)2 Generally, in a solar cell, the majority of incident sun-
layer decreases the electric field slope in that region. The light is not converted to electricity but us lost to heat
electric field slope is directly related to bending of the band generation in device. This heat can be generated via re-
diagram at the junction through Poisson equation. From combination and absorption mechanisms and will rise the
the electric field profile and the J-V curves presented later, temperature [22]. Wang et al. have also shown that non-
it seems that Mo(S,Se)2 balances up the band bending at radiative SRH recombination and thermalization are the
Mo/CZTSSe interface and reduces the harmful effects of leading energy loss in a cell [22]. Finally, we have also
the Schottky contact and reduces the recombination rate coupled the thermal module to the previously established
therein, which in turn, boosts the Voc . Moreover, the elec- optical and electrical modules in COMSOL package. The
tric field shows a Lorentzian peak at the bottom interface thermal analysis allows the optimization of device stabil-
with Mo contact where the carrier collection is improved ity by determining which heating source is the cause of
due to the amplified electric-field to enhance the Voc as performance drop over time. We have taken several heat
also reported experimentally for such a structure in Ref. sources into consideration including SRH non-radiative re-
[2]. The same electric field profile has also been reported combination, Joule heating and the conductive heat flux
in Ref. [20] where the peaks at the interfaces is noticeable magnitude (thermalization). The 3D maps of these heat
and supports an improved Voc . The electric field distri- generation sources have been presented in Fig. 6. Also,
bution has also been shown in 3D (Fig. 4b) where the the 3D map of the total heat flux magnitude has been pre-
interface at the top is the most intensive and the bottom sented in order to show where the heat is emitted from
junction is less intensive not even zero electric field yet as the cell after all. Ahmed et. al have also shown that the
expected from any pn junction. interface of Mo/CZTS layers are critical for cooling the
Using the optical module coupled to electrical mod- cell and keeping the temperature gradient isothermal [21].
ule, we have calculated the current density - voltage (J-V) The thermal conductivity of the Mo(S,Se)2 layer is also
4
 been performed to calculate the Shockley-Read-Hall non-
radiative recombination heating, Joule heating as well as
the conductive heat flux. The 3D maps of these heat gen-
eration and cooling sources helps to realize the 3D map of
the total heat flux magnitude. Thermal maps show that
SRH recombination heating is intensive at the depletion
width on CZTSSe side, while Joule heating is intensive at
the bottom electrode. A better heat conduction occurs
from the MO electrode and accelerate from the Mo(S,Se)2
layer. The total heat flux magnitude is more intensive
at the ITO electrode and the CZTSSe/CdS interface and
Figure 6: Thermal maps of the cell generated from different mecha-
nisms in the cell: a) SRH recombination heat, b) Joule heating, c)
within the ZnO layer.
heat conductive flux, and d) total heat flux magnitude.
5. Acknowledgment

This research is funded by the Foundation for Science

high enough to accelerate the heat emission out of the de-
and Technology Development of Ton Duc Thang Univer-
vice as stated by Peng et al. [16]. In Fig. 6c the thermal
sity (FOSTECT), website: http://fostect.tdtu.edu.vn, un-
gradient at the bottom of the cell shows a high tempera-
der Grant FOSTECT.2019.18.
ture within the Mo(S,Se)2 layer in order of 1.2×104 W/m2
which decreases to 0.6×104 W/m2 in the Mo layer. In op-
posite, the thermal gradient at the top electrode is not References
intensive but is well confined in the ZnO and CdS layers
[1] https://www.nrel.gov/pv/assets/pdfs/best-research-cell-
and cools down rapidly when it reaches ITO and air lay- efficiencies.20191106.pdf
ers on top. The total heat flux magnitude shows a more [2] W. Wang, M. T. Winkler, O. Gunawan, T. Gokmen, T.
interesting trend where the thermal gradient at top elec- K. Todorov, Y. Zhu, D. B. Mitzi, “Device characteristics of
trode is non-uniform (hemisphere) shape since the heat CZTSSe thin-film solar cells with 12.6% efficiency”, Adv. En-
ergy Mater., 1-5 (2013).
starts to generate from the top layers by light absorption [3] S.A. Khalate, R.S. Kate, R.J. Deokate, “A review on energy
therein. The SRH heating is dominated in the bulk region economics and the recent research and development in energy
of CZTSSe layer as expected due to non-radiative recom- and the Cu2 ZnSnS4 solar cells: A focus towards efficiency”,
bination whereas the Joule heating is 1 order of magnitude Solar Energy, 169 (2018) 616-633.
[4] N. Saini, J.K. Larsen, K.V. Sopiha, (...), N. Ross, C. Platzer-
smaller here (0.6×1010 W/m2 vs. 0.5×109 W/m2 ). Björkman, “Germanium Incorporation in Cu2 ZnSnS4 and For-
mation of a Sn–Ge Gradient”, Physica Status Solidi (A) Appli-
cations and Materials Science, 216:22 (2019) 1900492.
4. Conclusion [5] A. Davydova, K. Rudisch, J.J.S. Scragg, “The Single Phase Re-
gion in Cu2 ZnSnS4 Thin Films from Theory and Combinatorial
A simulation analysis has been performed on the opti- Experiments”, Chemistry of Materials, 30:14 (2018) 4624-4638
cal, electrical, and thermal characteristics of CZTSSe thin [6] A. Haddout, A. Raidou, M. Fahoume, A review on the numer-
ical modeling of CdS/CZTS-based solar cells”, Applied Physics
film solar cells using COMSOL multiphysics 3D simulation A, 125:124 (2019) 1-16.
package. The wave optics, semiconductor module and heat [7] A. Cherouan, R. Labbani, “Numerical simulation of CZTS solar
transfer in solids modules were coupled and run under a cell with silicon back surface field”, Materials Today: Proceed-
reasonably fine mesh of the cell structure. The simulation ings, 5 (2018) 13795-13799.
[8] A. Cherouana, R. Labbani, “Study of CZTS and CZTSSe So-
results of the optical and electrical coupled modules are lar Cells for Buffer Layers Selection”, Applied Surface Science,
in agreement with the record efficiency device parameters 424:2 (2017) 251-255.
of the Mo/Mo(S,Se)2 /CZTSSe/CdS/ZnO/ITO cell with [9] W. Zhao, W. Zhou, X. Miao, “Numerical Simulation of CZTS
12.6% presented by IBM in recent years. The current- Thin Film Solar Cell”, NEMS proceedings of IEEE, Kyoto,
Japan, March (2012) 502-506.
voltage characteristics of the two cells with and without [10] O. A.M. Abdelraouf, Nageh K. Allam, ” Nanostructuring for
Mo(S,Se)2 has been compared and the open-circuit voltage enhanced absorption and carrier collection in CZTS based solar
of the cell with Mo(S,Se)2 inserted between the Mo bottom cells: Coupled optical and electrical modeling”, Optical Mate-
contact and CZTSSe absorber layer shows an improvement rials, 54 (2016) 84–88.
[11] O. A. M. Abdelraouf, M. I. Abdelrahaman, N. K. Allam, “Plas-
from 0.46 V to 0.513 V in agreement with IBM record de- monic scattering nanostructures for efficient light trapping in
vice metrics. The electric field profile displays a maximum flat CZTS solar cells”, Proceedings of SPIE, (2017) 10227-
peak at the bottom contact around the Mo(S,Se)2 layer 1022712.
[12] https://www.comsol.com/story/simulations-for-solar-42671
which can explain the improvement in Voc . The optical
[13] P. Saxena, N. E. Gorji, “COMSOL Simulation of Heat Distri-
generation of the two cell has not been changing much by bution in Perovskite Solar Cells: Coupled Optical-Electrical-
adding the Mo(S,Se)2 layer as it is a very thin buffer layer Thermal 3D Analysis”, IEEE J. of Photovoltaics, 9:6 (2019)
with a wide bandgap. The thermal simulation has also 1693-1698.

5
[14] S. Zandi, P. Saxena, N.E. Gorji, “Numerical simulation of
heat distribution in RGO-contacted perovskite solar cells using
COMSOL”, Solar Energy, 197 (2020) 105-110.
[15] S.-D. Guo, “Phonon transport in Janus monolayer MoSSe: a
first-principles study”, Phys. Chem. Chem. Phys, (2018) pccp.
[16] B. Peng, H. Zhang, H. Shao, Y. Xu, X. Zhang, H. Zhu. “Thermal
conductivity of monolayer MoS2 , MoSe2 , and WS2 : Interplay
of mass effect, interatomic bonding and anharmonicity”, RSC
Adv., 6 (2016) 5767.
[17] D. Cozza, C. M. Ruiz, D. Duche, S. Giraldo, E. Saucedo, J.
J. Simon, L. Escoubas, “Optical modeling and optimizations
of Cu2 ZnSnSe4 solar cells using the modified transfer matrix
method”, Optics Express, 24:18 (2016) A1201.
[18] N. Song, M. A. Green, J. Huang, Y. Hu, X. Hao, “Study of
sputtered Cu2 ZnSnS4 thin films on Si”, Applied Surface Sci-
ence, 459 (2018) 700-706.
[19] G. Reya, G. Larramonab, S. Bourdaisb, C. Chonéb, B. Dela-
toucheb, A. Jacobb, G. Dennlerb, S. Siebentritta, “On the origin
of band-tails in kesterite”, Solar Energy Materials & Solar Cells,
179 (2018) 142–151.
[20] U. Saha, Md. K. Alam, “Boosting the efficiency of single junc-
tion kesterite solar cell using Ag mixed Cu2 ZnSnS4 active
layer”, RSC Adv., 8 (2018) 4905-4913.
[21] M. Ahmed, X. Hao, J. Park, E. R. Hawkes, M. A. Green, “Diode
Laser Annealing of CZTS Thin Film Solar Cells”, IEEE 42th
proceedings, 2015, 15664443, LA, USA.
[22] A. Wang, Y. Xuan, “A detailed study on loss processes in solar
cells ”, Energy, 144 (2018) 490-500.
[23] D. Cozza, C. Ruiz, M. Duché, J. J. Simon, L. Escoubas, “Mod-
eling the back contact of Cu2 ZnSnS4 solar cells” IEEE Journal
of Photovoltaics, 6 (2016) 1292-1297.