Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Optical Modeling and Optimizations of Cu Znsnse Solar Cells Using The Modified Transfer Matrix Method

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Vol. 24, No.

18 | 5 Sep 2016 | OPTICS EXPRESS A1201

Optical modeling and optimizations of


Cu2ZnSnSe4 solar cells using the modified
transfer matrix method
DARIO COZZA,1 CARMEN M. RUIZ,1,* DAVID DUCHÉ,1 SERGIO GIRALDO,2 EDGARDO
SAUCEDO,2 JEAN JACQUES SIMON,1 AND LUDOVIC ESCOUBAS1
1
Aix-Marseille University - IM2NP, CNRS (UMR 7334), France
2
Catalonia Institute for Energy Research (IREC), Jardins de les Dones de Negre 1 2pl, 08930 Sant
Adrià del Besòs, Barcelona, Spain
*
carmen.ruiz-herrero@im2np.fr

Abstract: The fast and computationally inexpensive Modified Transfer Matrix Method
(MTM) is employed to simulate the optical response of kesterite Cu2ZnSnSe4 solar cells. This
method can partially take into account the scattering effects due to roughness at the interfaces
between the layers of the stack. We analyzed the optical behavior of the whole cell structure
by varying the thickness of the TCO layer (iZnO + ITO) between 50 and 1200 nm and the
buffer CdS layer between 0 and 100 nm. We propose optimal combinations of the TCO/CdS
thicknesses that can locally maximize the device photocurrent. We provide experimental data
that qualitatively confirm our theoretical predictions.
©2016 Optical Society of America
OCIS codes: (350.6050) Solar energy; (310.6845) Thin film devices and applications; (310.4165) Multilayer design;
(240.5770) Roughness.

References and links


1. K. Maraun and A. Yamiya, “Solar Frontier achieves world record thin-film solar cell efficiency: 22.3%,” (Solar
Frontier, 2015), http://www.solar-frontier.com/eng/news/2015/C051171.html.
2. X. Liu, Y. Feng, H. Cui, F. Liu, X. Hao, G. Conibeer, D. B. Mitzi, and M. Green, “The current status and future
prospects of kesterite solar cells: a brief review,” Prog. Photovolt. Res. Appl. 24(6), 879–898 (2016).
3. Y. S. Lee, T. Gershon, O. Gunawan, T. K. Todorov, T. Gokmen, Y. Virgus, and S. Guha, “Cu2ZnSnSe4 thin-film
solar cells by thermal co-evaporation with 11.6% efficiency and improved minority carrier diffusion length,”
Adv. Energy Mater. 5(7), n/a–n/a (2015).
4. M. A. Green, K. Emery, Y. Hishikawa, W. Warta, and E. D. Dunlop, “Solar cell efficiency tables (version 46),”
Prog. Photovolt. Res. Appl. 23(7), 805–812 (2015).
5. M. Winkler, W. Wang, O. Gunawan, H. J. Hovel, T. K. Todorov, and D. B. Mitzi, “Optical designs that improve
the efficiency of Cu2ZnSn(S,Se)4 solar cells,” Energy Environ. Sci. 7(3), 1029–1036 (2014).
6. L. A. A. Pettersson, L. S. Roman, and O. Inganas, “Modeling photocurrent action spectra of photovoltaic devices
based on organic thin films,” J. Appl. Phys. 86(1), 487–496 (1999).
7. J. Mann, J. Li, I. Repins, K. Ramanathan, S. Glynn, C. DeHart, and R. Noufi, “Reflection optimization for
alternative thin-film photovoltaics,” IEEE J. Photovoltaics 3(1), 472–475 (2013).
8. I. Luck, “Competitiveness of CIGS technology in the light of recent PV developments – part 1: The state of the
art in CIGS production,” Photovoltaics Int. 24, 69–72 (2014).
9. M. D. Benoy, E. M. Mohammed, M. Suresh Babu, P. J. Binu, and B. Pradeep, “Thickness dependence of the
properties of Indium Tin Oxide (ITO) films prepared by activated reactive evaporation,” Braz. J. Phys. 39(4),
629–632 (2009).
10. F. Lisco, A. Abbas, B. Maniscalco, P. M. Kaminski, M. Losurdo, K. Bass, G. Claudio, and J. M. Walls, “Pinhole
free thin film CdS deposited by chemical bath using a substrate reactive plasma treatment,” J. Renewable
Sustainable Energy 6(1), 011202 (2014).
11. D. Cozza, “Modeling and physical studies of kesterite solar cells,” PhD dissertation, Aix Marseille University,
(2016).
12. S. Niki, M. Contreras, I. Repins, M. Powalla, K. Kushiya, S. Ishizuka, and K. Matsubara, “CIGS absorbers and
processes,” Prog. Photovolt. Res. Appl. 18(6), 453–466 (2010).
13. M. Burgelman, P. Nollet, and S. Degrave, “Modelling polycrystalline semiconductor solar cells,” Thin Solid
Films 361, 527–532 (2000).
14. Z. Holman, M. Filipic, A. Descoeudres, S. De Wolf, F. Smole, M. Topic, and C. Ballif, “Infrared light
management in high-efficiency Silicon heterojunction and rear-passivated solar cells,” J. Appl. Phys. 113(1),
013107 (2013).

#262194 http://dx.doi.org/10.1364/OE.24.0A1201
Journal © 2016 Received 31 Mar 2016; revised 27 May 2016; accepted 10 Jun 2016; published 26 Jul 2016
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1202

15. M. Richter, C. Schubbert, P. Eraerds, I. Riedel, J. Keller, J. Parisi, T. Dalibor, and A. Avellán-Hampe, “Optical
characterization and modeling of Cu(In,Ga)(Se,S)2 solar cells with spectroscopic ellipsometry and coherent
numerical simulation,” Thin Solid Films 535, 331–335 (2013).
16. C. Persson, “Electronic and optical properties of Cu2ZnSnS4 and Cu2ZnSnSe4,” J. Appl. Phys. 107(5), 053710
(2010).
17. Ö. Demircioğlu, M. Mousel, A. Redinger, G. Rey, T. Weiss, S. Siebentritt, I. Riedel, and L. Gütay, “Detection of
a MoSe2 secondary phase layer in CZTSe by spectroscopic ellipsometry,” J. Appl. Phys. 118(18), 185302 (2015).
18. Optical Data from Sopra SA, “Molybdenum optical constants,” http://www.sspectra.com/sopra.html.
19. M. Neuschitzer, Y. Sanchez, S. Lopez-Marino, H. Xie, A. Fairbrother, M. Placidi, S. Haass, V. Izquierdo-Roca,
A. Perez-Rodriguez, and E. Saucedo, “Optimization of CdS buffer layer for high-performance Cu2ZnSnSe4 solar
cells and the effects of light soaking: elimination of crossover and red kink,” Prog. Photovolt. Res. Appl. 23(11),
1660–1667 (2015).
20. S. Siebentritt, G. Rey, A. Finger, D. Regesch, J. Sendler, T. P. Weiss, and T. Bertram, “What is the bandgap of
kesterite?” Sol. Energy Mater. Sol. Cells in press (2015).
21. M. S. Huh, B. S. Yang, J. Song, J. Heo, S.-J. Won, J. K. Jeong, C. S. Hwang, and H. J. Kim, “Improving the
morphological and optical properties of sputtered Indium Tin Oxide thin films by adopting ultralow-pressure
sputtering,” J. Electrochem. Soc. 156(1), 6–11 (2009).
22. G. G. Pethuraja, R. E. Welser, A. K. Sood, C. Lee, N. J. Alexander, H. Efstathiadis, P. Haldar, and J. L. Harvey,
“Current-voltage characteristics of ITO/p-Si and ITO/n-Si contact interfaces,” Adv. Mater. Phys. Chem. 2(2),
59–62 (2012).
23. J. M. Bernhard, “Work function study of Iridium Oxide and Molybdenum using UPS and simultaneous Fowler-
Norheim I-V plots with field emission energy distributions,” PhD dissertation. University of North Texas, p. 140,
(1999).
24. G. Yin, C. Merschjann, and M. Schmid, “The effect of surface roughness on the determination of optical
constants of CuInSe2 and CuGaSe2 thin films,” J. Appl. Phys. 113(21), 213510 (2013).
25. S. Byrnes, “Multilayer optical calculations,” (Cornell University Library, 2016), arXiv:1603.02720.
26. K. Han and C.-H. Chang, “Numerical modeling of sub-wavelength anti-reflective structures for solar module
applications,” Nanomaterials (Basel) 4(1), 87–128 (2014).
27. S. Giraldo, M. Neuschitzer, T. Thersleff, S. López-Marino, Y. Sánchez, H. Xie, M. Colina, M. Placidi, P. Pistor,
V. Izquierdo-Roca, K. Leifer, A. Pérez-Rodríguez, and E. Saucedo, “Large efficiency improvement in
Cu2ZnSnSe4 solar cells by introducing a superficial Ge nanolayer,” Adv. Energy Mater. 5(21), 1501070 (2015).
28. D. Raoufi and F. Hosseinpanahi, “Surface morphology dynamics in ITO thin films,” J. Mod. Phys. 03(08), 645–
651 (2012).
29. K. Nanda, S. Sarangi, and S. Sahu, “Measurement of surface roughness by atomic force microscopy and
Rutherford backscattering spectrometry of CdS nanocrystalline films,” Appl. Surf. Sci. 133(4), 293–297 (1998).
30. Z. Ma, Z. Tang, E. Wang, M. R. Andersson, O. Inganas, and F. Zhang, “Influences of surface roughness of ZnO
electron transport layer on the photovoltaic performance of organic inverted solar cells,” J. Phys. Chem. C
116(46), 24462–24468 (2012).

1. Introduction
Kesterite-based solar cells attracted significant attention during the last years as a possible
alternative to Cu(In,Ga)Se2 (CIGS) solar cells. High efficiency CIGS solar cells with η =
22.3% have been demonstrated [1] but their large-scale industrial production could be limited
because of indium scarcity and growth in its price (high demand from the display industry)
[2]. Copper–zinc–tin–sulfide/selenide absorbers (kesterite) could overcome this issue thanks
to their cheap and earth-abundant composition. The record device using a pure selenide
Cu2ZnSnSe4 absorber (CZTSe, band gap Eg ≈1 eV) showed an efficiency η = 11.6%, open
circuit voltage VOC = 0.423V and short circuit current JSC = 40.6mA/cm2. The best kesterite
solar cell, obtained with a Cu2ZnSn(S,Se)4 absorber (Eg = 1.13eV), has η = 12.6%, VOC =
0.513V and JSC = 35.2mA/cm2 [2,3].
The performances of kesterite solar cells today are still low to start rivaling the more
mature CIGS technology. The low VOC remains the main issue of this technology, with values
that can be around 150-200 mV lower than the ones of CIGS cells with equivalent Eg [2,4].
But also sub-optimal values of JSC can furtherly lower the device efficiency: Winkler et al.
stated that 75% of the possible JSC-deficit in kesterite cells is due to optical losses related to
the parasitic absorption of the buffer/TCO layers and the reflection of photons from the top
surface [5]. The optimization of the thickness of the top semi-transparent layers is a possible
way to minimize the optical losses due to photons reflection which lead to improvements of
the device JSC.
In this work we consider a Molybdenum/MoSe2/CZTSe/CdS/iZnO/ITO structure and we
simulate its optical response (reflectance and photocurrent) as a function of the thickness of
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1203

the TCO (iZnO + ITO) and the buffer (CdS) layers. From the analysis of the simulated device
photocurrent we can identify local maxima (optimal points) and minima in the space of the
TCO/CdS thicknesses: this phenomenon is due to the constructive/destructive interference
effects generated by the multiple reflections of photons inside the top transparent layers [6].
Interference fringes can be experimentally observed in the Reflectance and External Quantum
Efficiency (EQE) spectra of CIGS and Kesterite solar cells.
Previous similar works [5,7] proposed ultra-thin thicknesses for the TCO/CdS layers (≤
60nm) such that a large improvement of the photocurrent generation was mainly due to the
minimization of the parasitic absorption of these layers. This strategy is an absolute optimum
from the optical transparency point of view but it is not ideal in terms of higher lateral series
resistance [8,9] and other possible technological issues (e.g. pinholes [10]) that can
compromise the device efficiency. This partially explains why a wide range of TCO/CdS
thicknesses has been proposed for Kesterite solar cells in literature: TCO from 60 to ~500 nm
and CdS from 25 to ~100 nm at the research scale [11]. At the module scale even thicker
thicknesses for the TCO layer could be required: even though kesterite industrial modules
have yet to be reported, the standard TCO thickness for CIGS industrial modules is about 1
μm [12].
We investigated all the range of cases by performing simulations of the whole kesterite
solar cells varying the thickness of the TCO from 50 to 1200 nm and the CdS from 0 to 100
nm. We employed 1D numerical optical models to calculate the percentage of photons lost by
reflection and the percentage of photons absorbed in the CZTSe, as a function of the
TCO/CdS layers thicknesses. The software SCAPS [13] was also employed to compute the
EQE of the simulated devices. The theoretical results are compared to experimental
measurements of devices having the same structure of our model and variable thicknesses for
the TCO layer between 400 and 110 nm and fixed thickness for the CdS layer equal to ~50
nm.
2. Methodology
The optical constants of the simulated stack (see Fig. 1) were selected from literature [14–18]
with the exception of the CdS layer, whose optical constants (n-k coefficients) were measured
by Spectroscopic Ellipsometry (SE). For the CZTSe layer we chose optical constants
computed by ab-initio simulation (by Persson [16]) because the available experimental works
from literature reporting n-k coefficients of kesterites are often inaccurate in the wavelengths
range near the fundamental absorption edge (λ~1250nm for CZTSe) because of issues related
to the high surface roughness of the polycrystalline kesterite absorbers [5]. Anyhow, with the
optical constants from Persson [16], the simulated EQE underestimates the absorption in the
wavelengths between 1100 and 1300 nm compared to the experimental one (see Fig. 2): this
is expected to have some limited quantitative impacts on our presented results. The longer
absorption edge visible in the experimental EQE measurements has been explained to be
related to several possible non-idealities of the kesterite absorbers, like tail states and
inhomogeneities [20].
The electrical parameters for our model were discussed in detail in previous works [11]. In
Table 1 we have summarized the most relevant among them. The Molybdenum (Mo) back
contact was modeled with a work function Wf = 4.95eV [23]. Specific defect states and
tunneling at the interfaces between the layers were not implemented.
The software SCAPS, which was employed for the computation of EQE, natively
performs its optical simulations considering a user-defined constant reflection (between 0 and
100%) for the first and the last interfaces of the stack and it applies the Beer-Lambert law for
the estimation of photons absorption inside the layers of the stack. Alternatively the user can
input files representing the spectral reflectance for the first/last interfaces, obtained by other
optical simulation methods or also by experimental measurements.
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1204

Fig. 1. (a) Refractive indices n and (b) extinction coefficients k (reported in logarithmic scale)
of the materials considered in the optical simulations: ITO [14], iZnO [15], CdS (SE), CZTSe
[16], MoSe2 [17], Mo [18].

Table 1. Electrical properties of simulated materials


Material ITO (n) iZnO (n) CdS (n) CZTSe (p) MoSe2 (p)
Energy Bandgap (eV) 3.4a 3.3 2.4 1.05 1.1
Electron affinity (eV) 4.7b 4.45 4.2 4.54 4.14
Net Carrier Concentration (cm−3) 1e20a 1e18 1e17 1e16 1e16
Electron/Hole mobility (cm2/Vs) 100/25 100/25 100/25 40/10 100/25
Minority carrier lifetime (ns) 3e-3 3e-3 5e-3 7 2.4
a
From [21]. b From [22]. All the other materials parameters are from [11].

We carried out 1D optical simulations using two different methods: the standard Transfer
Matrix Method (TMM) and the “Modified Transfer Matrix” (MTM) method that was
previously employed to model CuInSe2 and CuGaSe2 solar cells by Yin et al. [24]. The TMM
is an exact analytical method to compute the optical response of multilayer stacks with
perfectly flat interfaces [6]: it only requires the optical constants n and k and the thickness of
each layer. The incoming electromagnetic radiation is modeled as plane waves in normal
incidence. As we anticipated above, kesterite absorbers present significant surface roughness
that can be up to a hundred of nanometers [5]. Also the CdS and TCO layers introduce
roughness features at smaller length scales. These interface roughness components can lead to
scattering within the films and results in substantial intensity reduction of reflectance (R)
especially at short wavelengths [24]. For TMM calculations we employed a code written in
Python by S. Byrnes [25]. The MTM method is identical to the TMM method except for the
substitution of the Fresnel coefficients ra,b and ta,b (complex reflection and transmission
coefficients) at the interface from the layer a to the layer b, with the following quantities [24]:

r 'a ,b = ra ,b exp  −2 ( 2πσ λ ) na 2 


2
(1)
 

t 'a ,b = ta ,b exp  −2 ( 2πσ λ ) ( na − nb ) 2 


2 2
(2)
 
where na and nb are the refractive indexes (real part) of the layers a and b respectively, and σ
is the surface roughness (unit nm) in terms of root mean square (RMS) between the layers a
and b. The MTM can partially consider the effects of interfaces roughness, when σ/λ<<1 [24].
Both the TMM and MTM are relatively fast and computationally inexpensive compared to
other more complex and time consuming optical methods (e.g. the FDTD, Finite-difference
time-domain, methods) [26].
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1205

Our baseline model was based on a device with a CZTSe absorber (Eg ≈1.02 eV)
fabricated by reactive thermal annealing of metallic precursor stacks deposited by DC
sputtering (Alliance Concepts AC450), and described more in detail in Neuschitzer et al. [19]
and Giraldo et al. [27]. CdS was deposited using chemical bath deposition, while the TCO
was deposited using DC-pulsed sputtering (Alliance Concepts CT100). In order to provide an
experimental validation of our theoretical results, four more solar cells samples with variable
ITO thicknesses (denoted as ITO1-4) were fabricated. Several individual solar cells with an
area of 3 × 3mm2 were obtained on each of the ITO1-4 samples, using a manual
microdiamond scriber with a scribed line width of 20μm. When dealing with these four
samples in the following paragraphs, we will consider only the best-efficiency cell present on
each of them.
Both the baseline and the ITO1-4 samples have a 1800 nm thick CZTSe layer and about
380 nm of MoSe2, estimated from Scanning Electron Microscopy (SEM) cross section
images. The nominal thicknesses of the ITO, iZnO and CdS layers of our cell are equal to
350, 50 and 50 nm respectively [19], whereas the ITO1-4 samples differ from the baseline for
the ITO thickness (as a result of shorter deposition time). These “nominal” thicknesses are
values obtained from the deposition systems calibrations over flat reference samples, like
glass or silicon [19]: when deposited to different substrates and rough surfaces, as it is the
case of kesterite films, these thicknesses are expected to slightly vary. But after growth over
the kesterite, it is difficult to experimentally determine their actual thickness by analysis of
SEM cross sections because of the rugosity of the underlying CZTSe layer and a non-perfect
thickness homogeneity over the whole sample.
For these reasons we refined the estimation of the thickness of the ITO, iZnO and CdS
layers using a DOE (design of experiments) fit method based on TMM and MTM simulations,
looking for the best matching of the simulated/measured spectral reflectance in terms of
position of the local maxima and minima points. The considered parameter space for the
fitting procedure was the thickness of the three top layers (ITO/iZnO/CdS), because the
position of the interference fringes in the spectral reflectance is closely correlated to the
thickness of these layers as well as their optical constants. The thickness of the considered
layers was varied over a range of ± 50 nm with respect to the known nominal values. The best
fitting combinations obtained with this procedure are summarized in Table 2 and no other
similar results could be obtained with considerably different combinations inside the
considered ranges.
Table 2. Thickness (in nm) of the top layers estimated by TMM and MTM
TMM MTM
Sample CdS iZnO ITO TCO CdS iZnO ITO TCO
Baseline 45 (50) 50 (50) 380 (350) 430 45 50 380 430
ITO1 61 (50) 62 (50) 340 (350) 402 60 50 350 400
ITO2 45 (50) 45 (50) 245 (270) 290 45 45 245 290
ITO3 55 (50) 45 (50) 147 (180) 192 55 45 145 190
ITO4 53 (50) 46 (50) 75 (90) 121 53 40 70 110
The values inside the parentheses correspond to the nominal thickness values, estimated considering the
deposition systems calibrations on flat substrates. The TCO correspond to the sum of the thicknesses of
iZnO + ITO.
In Table 3 we summarize the values of interfaces roughnesses (σ) employed for the MTM
method: these values were estimated with a similar procedure as the one explained above. The
estimated σ for the Air/ITO/iZnO/CdS interfaces are well inside the range of standard values
that are typically reported in literature for these layers (σ from 5 to 25 nm) [28–30]. Surface
roughness σ for kesterite films can typically be in the order of magnitude from tens to a
hundred of nm [3,5], which are values close to the limit of validity of the MTM method
(σ/λ<<1) for the wavelength range considered in this work (λ ∈ [280,1300] nm). We
estimated a value of σ = 35 nm for the CdS/CZTSe interface that, for this technology,
represents a relatively moderate amount of roughness.
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1206

Table 3. Surface roughness (σ) values, in terms of RMS for the MTM method
Air ITO iZnO CdS CZTSe MoSe2 Mo
15 nm 20 nm 15 nm 35 nm 0 nm 0 nm

EQE measurements were made using a Bentham PVE300 system. Total reflectance
measurements R were made with a Perkin Elmer R LAMBDA 950 UV/Vis
Spectrophotometer (integrating sphere). The JSC of all the solar cells present on the ITO1-4
samples was measured using a Sun 3000 class AAA solar simulator (Abet Technologies Inc.).
3. Results
In Fig. 2, spectra of simulated reflectance (R) and EQE obtained with four different
approaches are compared to the measured EQE/R of the baseline experimental cell [19]. In
Fig. 2(a) we can observe the results obtained with the TMM and MTM methods. In Fig. 2(b)
we show the EQE simulations obtained with a constant reflection value (R = 10%) and a so
called Hybrid model, such that the experimental R of the baseline cell (red solid line) is input
in our model in order to compute the Hybrid EQE (blue dots).

Fig. 2. Experimental (baseline [19]) and simulated (Model) EQE/R spectra.

The simulation with a constant reflectance (Fig. 2(b)) represents the basic result that can
be obtained with SCAPS only, which obviously cannot reproduce the interference fringes.
The TMM model, visible in Fig. 2(a), is able to simulate the interference fringes in the
Reflectance spectrum but it strongly overestimates their amplitudes with respect to the
experimental one. Finally the MTM method produces better results than the TMM, in terms of
reduction of the amplitude of the interference oscillations of the simulated reflectance: as
anticipated in the previous section, this is due to the implementation of the scattering effects
due to the interfaces roughness.
The Hybrid model (in Fig. 2(b)) ideally should produce an EQE identical to the
experimental one. The differences between the experimental EQE (black line) and Hybrid
model (blue circles), visible in Fig. 2(b) for some wavelengths ranges, can be explained as
follow:
• the differences in the [1100-1300] nm range are due to the optical constants of the
CZTSe layer, as we have already explained in the previous section;
• the differences in the [750-1000] nm range can be due to the optical constants of the
ITO layer (underestimation of the long-wavelengths absorption) as well as to defect
interfacial states at the back contact (not implemented);
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1207

• the differences in the [400-500] nm range can be due to the optical constants of the CdS
layer (underestimation of absorption) as well as to defect interfacial states between
the CdS-CZTSe layers (not implemented).
Even though the Hybrid model can reproduce the experimental EQE better than all the
other considered models, it cannot be employed to perform predictions or investigations of
device optimizations because it requires experimental reflectance spectra and it does not
provide information of the layers internal absorption required to compute the device
photocurrent. From the comparison of the TMM and the MTM models we show that the
MTM produces a closer agreement between the simulated and experimental EQE/R spectra
(Fig. 2(a)), thus it has been selected as the preferential method to perform the optimization
study that will be described in the followings.
In Fig. 3 we report the results of multiple optical simulations performed using the MTM
method.

Fig. 3. Results of MTM optical simulations varying the thicknesses of the TCO (iZnO + ITO)
and CdS layers: (a) percentage of reflected photons R% considering the wavelengths range
[400-1100] nm and (b) photocurrent Jph generated in the CZTSe absorber. The iZnO thickness
is constant and equal to 50nm.

Figure 3(a) represents the reflected photons percentage (R%) and Fig. 3(b) represents the
photocurrent generated in the CZTSe layer (Jph). Both graphs share the same X and Y-axis:
on the X-axis we varied the thickness of the TCO layer from 50 to 1200 nm and on the Y-axis
we varied the thickness of the CdS layer from 0 to 100 nm. Actually, the TCO layer is
composed by a thin intrinsic Zinc Oxide (iZnO) layer of constant thickness equal to 50 nm
and an ITO layer with variable thicknesses between 0 and 1150 nm. The photocurrent Jph has
been computed as follow:
1300 nm
J ph = q  N ph − AM 1.5 ( λ ) ACZTSe ( λ )d λ  mA / cm 2  (3)
280 nm

where q is the electron charge, Nph-AM1.5(λ) is the photon flux in the AM1.5 solar spectrum and
ACZTSe(λ) is the spectral absorption of photons inside the CZTSe layer computed by a MTM
simulation. Jph represents an upper bound of the maximum achievable JSC of the device [5]
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1208

thus the qualitative trends observed in Fig. 3(b) related to Jph are expected to be valid also for
JSC. The wavelength integration range between 280 and 1300 nm takes into account the onset
of the AM1.5 spectrum (280nm) and the limit of absorption of the CZTSe absorber in our
structure (below 1300nm, as seen in Fig. 1(b)). The Reflected Photons percentage R% has
been computed as follow:

[ %]
1100 nm 1100 nm
R% =  N ph − AM 1.5 ( λ ) R ( λ )d λ  N ph − AM 1.5 ( λ )d λ (4)
400 nm 400 nm

where R(λ) is the spectral reflectance at the top interface of the device, computed by MTM.
The integration interval for R% is between [400-1100] nm of wavelength because, for the
considered materials, the minimization of reflectance plays an important role mainly in this
range: for λ<400nm the absorption of the CdS/TCO layers is dominant and for λ>1100nm the
absorption of the CZTSe layer is weak (see Figs. 1(b) and 2(b)).
The dots plotted in Figs. 3(a) and 3(b) represent the simulated values of R%/Jph for the
thicknesses of CdS/TCO corresponding to the experimental samples (see also Table 2, MTM
columns). In Fig. 3(b) we can see that the samples ITO1 and ITO3 are inside two different
non-optimal regions. The sample ITO2 is inside an optimal region and the sample ITO4 seems
to be at the border between a non-optimal and an optimal region.
Comparing the Fig. 3(b) with the Fig. 3(a) we can clearly see an inverse dependency
between the local maxima of Jph with the local minima of R%. This can be also observed in
Fig. 4(a) where we plotted the simulated Jph and R% for a constant thickness of the CdS layer
equal to 50 nm (black and red solid lines respectively). Considering this specific case, five
local maxima are present in the Jph curve for TCO thicknesses equal to 1100, 845, 570, 280
nm and for a thickness lower than 100 nm. The grey line in Fig. 4(a), representing the linear
regression of the simulated Jph curve, highlights the lowering of Jph itself that is directly due
to the increasing parasitic absorption of the ITO layer with its increasing thickness.

Fig. 4. (a) Simulated Jph and R% (solid line) and experimental JSC and R% (dashed lines). The
simulations consider a varying ITO thickness and constant thickness for the CdS and iZnO
layers, both equal to 50nm. Figure (b) is a zoom-in of figure (a) in the [50-450] nm range,
showing only Jph and JSC: the empty circles represent the MTM simulations considering the
exact values of CdS-iZnO-ITO thickness corresponding to the experimental samples (Table 2)
while the black solid line is the same as the one plotted in figure (a).

In Fig. 4(b) our theoretical results in terms of Jph (solid line) are compared to the
experimental JSC measurements related to the solar cells of the samples ITO1-4 (blue crosses).
Qualitatively the average values of the experimental JSC (dashed line) seem to confirm the
expected trends of simulated Jph obtained by MTM. In Fig. 4(a) we can see that also the
simulated trends in terms of R% (red solid line) are confirmed by the experimental R% (red
dashed line) computed applying the Eq. (4) on the measured spectral reflectances of the
samples ITO1-4 visible in Fig. 5 (red lines). The simulations show some bigger discrepancies
for the case of sample ITO4, having the thinner value of ITO thickness equal to 70 nm (iZnO
Vol. 24, No. 18 | 5 Sep 2016 | OPTICS EXPRESS A1209

+ ITO = 110 nm). The experimental ITO4 exhibits the best average JSC whereas in our model
the best Jph is obtained in the case of ITO2 (290 nm thick TCO). From the analysis of Figs.
4(a) and 5 we can see that the MTM simulations exceed the estimation of the reflectance for
the thinner thicknesses of the TCO, motivating the underestimation of Jph for the simulated
ITO4. Besides this specific case, a lower mismatch between the simulated/measured EQE/R
spectra are observed for all the other samples (see Fig. 5).

Fig. 5. Measured and simulated EQE/R spectra of the four samples ITO1-4 (in this order from
(a) to (d)). The black/red solid lines represent the experimental EQE/R. The green/brown
dashed lines represent the simulated EQE/R obtained by MTM.

4. Conclusions
In this work we employed the MTM numerical method to investigate the optical behavior of a
CZTSe solar cell over the variation of the thickness of its TCO and CdS layers. Considering a
TCO thickness from 50 to 1200 nm and a CdS thickness from 0 to 100 nm we individuated
the combinations of the TCO/CdS thicknesses that can provide improvements on the device
photocurrent due to local minimizations of the Reflected Photons percentage (see Figs. 3(a)-
3(b)). Comparisons with experimental data of devices with variable TCO thicknesses
(between 400 and 110 nm) and constant CdS thickness (~50 nm) provided evidences of the
consistency of our calculations in terms of expected trends, even though we could not obtain
an exact quantitative estimation of the improvements because of some limitations of the
model itself. In general the choice of a thinner TCO thickness is preferable to obtain record
cells at the research scale (parasitic absorption minimization) whereas a thicker TCO is
required for scaling up the devices to the sub-module stage (sheet resistance minimization and
better reliability). From a methodology point of view we discussed the reasons for the
selection of the MTM over the pure TMM: the MTM improves the matching between the
simulated and experimental EQE/R spectra thanks to its capability of partially consider the
effects of scattering due to the interfaces roughness in the CZTSe solar cell. Although other
more complex methods (e.g. FDTD) could be employed to furtherly improve the accuracy of
the optical simulations, they would not guarantee the same performances of the MTM
method, in terms of high speed and low computational cost, that were required to perform the
proposed optimizations.
Acknowledgments
This research was supported by the Framework 7 program under the project KESTCELLS
(FP7-PEOPLE-2012-ITN316488). S.G. thanks the Government of Spain for the FPI
fellowship (BES-2014-068533), and E.S. for the “Ramón y Cajal” fellowship (RYC-2011-
09212).

You might also like