Journal of Physics D: Applied Physics

J. Phys. D: Appl. Phys. 49 (2016) 085101 (6pp) doi:10.1088/0022-3727/49/8/085101

Temperature dependent electrical

characterization of thin film Cu2ZnSnSe4
solar cells
E Kask1, J Krustok1, S Giraldo2, M Neuschitzer2, S López-Marino2
and E Saucedo2
1
Tallinn University of Technology, Ehitajate tee 5, 19086 Tallinn, Estonia
2
Catalonia Institute for Energy Research, IREC, Jardins de les Dones de Negre 1,
08930 Sant Adrià del Besòs, Barcelona, Spain

E-mail: erkki.kask@ttu.ee

Received 16 August 2015, revised 15 November 2015

Accepted for publication 18 November 2015
Published 25 January 2016

Abstract
Impedance spectroscopy (IS) and current–voltage characteristics measurements were applied
to study properties of a Cu2ZnSnSe4 (CZTSe) thin film solar cell. IS measurements were done
in the frequency range 20 Hz to 10 MHz. The measurement temperature was varied from
10 K to 325 K with a step ∆T  =  5 K. Temperature dependence of Voc revealed an activation
energy of 962 meV, which is in the vicinity of the band gap energy of CZTSe and hence
the dominating recombination mechanism in this solar cell is bulk recombination. Different
temperature ranges, where electrical properties change, were found. Interface states at grain
boundaries with different properties were revealed to play an important role in impedance
measurements. These states can be described by introducing a constant phase element in the
equivalent circuit.

Keywords: Cu2ZnSnSe4, impedance spectroscopy, current–voltage characteristics,

grain boundaries
(Some figures may appear in colour only in the online journal)

1. Introduction and different defect levels were detected. In CZTSe two

defect states were observed by admittance spectroscopy
Quaternary compound Cu2ZnSnSe4 (CZTSe) with a direct (AS). The unstable defect state showed activation energies
band gap and high absorption coefficient (>104) [1] is a prom- from 87 meV to 100 meV and was attributed to the interface
ising absorber material for photovoltaics. Moreover, CZTSe states because the possible change of interface properties
is a low-cost non-toxic semiconductor material, whose ele- with time could alter activation energy. The second state at
mental constituents are abundant in the Earth’s crust. The cur­ 74 meV was assumed to belong to Cu−Zn. These activation
rent power conversion efficiency record of a CZTSe device is energies were found by measuring impedance Z and phase
11.6% [2]. Photovoltaic performance is hindered by a large angle θ versus frequency f, and calculating capacitance C by
variety of intrinsic lattice defects, which have influence on employing an equivalent circuit, where a series resistor Rs is
optical and electrical properties of solar cells. Little informa- followed by a resistor Rp and a capacitor connected in par-
tion about the defect structure of CZTSe can be found in the allel, (Rs and Rp cover series resistance of a structure, wires,
literature. contacts, etc, and shunt resistance of the junction, accord-
Defects have been studied by photoluminescence spectr­ ingly). Inverse temperature versus the natural logarithm of
oscopy [3–6] and also by capacitive spectroscopy methods. angular frequencies ω that correspond to a capacitance step
In our previous works [7, 8] we studied so-called monograin showed linear dependence and this was used for activation
solar cells consisting of a microcrystal absorber material, energy calculation.

0022-3727/16/085101+6$33.00 1 © 2016 IOP Publishing Ltd Printed in the UK

J. Phys. D: Appl. Phys. 49 (2016) 085101 E Kask et al

Weiss et al [9] have studied modelling equivalent circuit boundaries properties. Also, IS measurements frequency
responses in kesterite solar cells and found that the evaluation range limits are shown and results of I–V measurements and
of the admittance data cannot be performed simply with the IS for a CZTSe thin film solar cell are presented and analyzed.
as-measured capacitance data, as an increasing series resist­
ance with decreasing temperature results in a capacitance step 2. Experiment
within the C–f profile. For instance, they suppose that meas-
urements data need to be discarded at higher frequencies and A CZTSe absorber for this study was prepared by reactive
at lower temperatures. In addition, they used current–voltage thermal annealing of a metallic Cu/Sn/Cu/Zn precursor stack
characteristics (I–V) to extract series and shunt resistances, deposited by dc magnetron sputtering onto Mo-coated soda
but resistances found by direct current measurements are not lime glass substrates, as described elsewhere in more detail
perfectly suitable for replacing resistance values of alternating [14]. The CZTSe absorber composition was measured by
current measurements. Moreover, they used as-measured x-ray florescence spectroscopy showing cation ratios of Cu/
capacitance (meaning no series resistance exists) and later (Zn  +  Sn)  =  0.69, Zn/Sn  =  1.45, Cu/Zn  =  1.16 and Cu/
applied series resistance to it. Sn  =  1.69. The details of the preparation of the thin film
Goodman [10] showed that accurate measurement of solar cell device are given in [15]. The individual solar cell
capacitance requires both Rs  Rp and Rs  (ωC)−1 so that (with dimensions 3  ×  3 mm2) used for this study shows power
the capacitive impedance is the dominant circuit element. conversion efficiency of η  =  6.6% with Jsc  =  27.2 mA cm−2,
Ordinarily, the quantity C is assumed to be frequency inde- Voc  =  383 mV and FF  =  64%.
pendent, but in experiments this is not always true; reasons For temperature-dependent measurements the selected
may be, for example, charge relaxation times. In general, thin film solar cells were mounted in a closed-cycle He cry-
such effects tend to disappear with increasing frequency and ostat (Janis). I–V curves were recorded using a Keithley
may place a lower limit on the trustworthy frequency range. SourceMeter 2401 with 100 mW cm−2 illumination. For a
Moreover, limitations with respect to measurement equip- light source a standard 250 W halogen lamp with calibrated
ment may place further restrictions upon the measurement intensity was used. The impedance spectra curves were
frequency. These restrictions require that the resistance and recorded using a Wayne Kerr 6500B impedance analyzer. For
geometry of the solar cell are suitably chosen. IS the impedance Z and phase angle θ were both measured
Fernandes et al [11] have studied AS of Cu2ZnSnS4 as functions of frequency f and temperature T. The temper­
(CZTS) solar cells and also applied various equivalent circuits ature was varied from 10 K to 300 K with a step ∆T  =  5 K.
to fit measured data. They calculated the deviation between The used frequency was in the range 20 Hz to 10 MHz. In
measured and fitted data and found that their fitting of meas- order to maintain the linearity of the response signal, the ac
ured data is satisfsctory only via a quite complicated model. voltage was kept as low as 10 mV. The ac measurements were
Usually, the more elements used in an equivalent circuit the carried out in the dark and dc biases of 0 V and  −0.5 V were
better fit found, but in this case the interpretation is impossible used. After each measurement, the real and imaginary parts
or inappropriate. Moreover, Friesen et al [12] studied a thin of the impedance, Z′ and  −Z″ respectively, were calculated.
film CdTe solar cell and state that as commonly applied equiv- Measured data were fitted using the ZView (Scribner, USA)
alent circuit models consist only of frequency-independent computer program, which gives values for elements used in an
circuit elements they cannot be used to describe the frequency equivalent circuit; for example, series resistance, Rs.
dispersion of a thin film solar cell. They replaced the capacitor
with a frequency-dependent non-ideal capacitor, called the 3. Results and discussion
constant phase element (CPE). It is generally assumed that
non-ideal capacitance behaviour originates from a distribution The studied solar cell exhibited an efficiency of 6.6%.
in the current density due to material inhomogeneity and/or Temperature dependence of the open circuit voltage (Voc)
the grain boundaries [12]. The use of a CPE provides a better derived from I–V measurements with varied temperature from
fitting for a depressed semicircle in the Nyquist plot. T  =  10 K to T  =  325 K was linear in the range 250 K to 325 K
According to our previous temperature-dependent meas- (see figure 1). It is known that the temperature dependence of
urements [13], three different temperature ranges, where elec- Voc near room temperature can be presented as [16, 17]
EA, Voc nkT ⎛ I00 ⎞
trical parameters change, were seen in monograin CZTS solar
cells. At very low temperatures (T  <  40 K) the effective band Voc = − ln⎜ ⎟,
⎝ IL ⎠
(1)
gap energy Eg* and series resistance Rs showed steep changes q q
and this behaviour was interpreted as hole blocking at a where IL, EA, Voc, n, k are the photocurrent, an activation energy,
CZTSe/CdS heterojunction. Therefore the interface recombi- the diode ideality factor and the Boltzmann constant, respec-
nation rate turns out to be very low at T  <  40 K. tively. The constant q is the electrical charge of the electron
Usually, impedance spectroscopy (IS) and AS measure- and I00 is obtained from the temperature dependence of the
ments are conducted at higher temperatures because models dark saturation current I0. In good solar cells  −IL  ≈  Isc, where
are then more easily applied and understood. Unfortunately, Isc is a short circuit current. I0 equals [16, 17]
⎛ EA, Voc ⎞ ⎛ E ⎞
at very low temperatures interpreting IS spectra is quite dif-
ficult. In this work, we analyze this complicated IS measure- I0 = I00 exp⎜− ⎟ ≈ AT 3 ⋅ exp⎜− A, Voc ⎟.
⎝ nkT ⎠ ⎝ nkT ⎠
(2)
ments data at very low temperatures and correlate with grain
2
J. Phys. D: Appl. Phys. 49 (2016) 085101 E Kask et al

Figure 2. Temperature dependence of jsc. Insets show found

Figure 1. Temperature dependence of Voc of a CZTSe thin film activation energies in region I and II.
solar cell.
time, we also expect an increasing role of bulk recombina-
In general, the activation energy EA, Voc and also I00 depend tion in region II, as previously seen in CZTS [13]. At low
mostly on the dominating recombination mechanism in the temper­atures (T    150 K), the carriers, here holes, can occupy
solar cell. In the case of bulk recombination EA, Voc  ≈  Eg, where shallow acceptor states in the band gap. These states can arise
Eg is the band gap energy of the absorber material. from intrinsic defects or impurities or their interaction. At a
Linear fitting of the temperature dependence of Voc showed sufficiently high concentration of these states, but below the
that EA, Voc  =  962 meV (see figure 1), which is in good agree- critical Mott concentration, overlapping of carriers wave func-
ment with the band gap energy of CZTSe that is expected to tions could form a so-called impurity band [23]. In such a case
be in the vicinity of 1 eV [4, 18], and hence the dominating charge carriers can move from one shallow impurity state to
recombination mechanism seems to be bulk recombination. another one at its spatial neighbourhood and conduction in
According to the temperature dependence of jsc (see the impurity band is expected [22]. These different processes
figure 2) three different temperature regions can be distin- could all be present and therefore we do not focus so much on
guished. Similar behaviour was also seen in CZTS [13]. At activation energy here.
temperatures T  >  150 K (region I) the series resistance Rs At low temperatures (region III, T  =  10 K–50 K) the
decreases with increasing temperature, indicating the thermal maximum of series resistance is present. By analyzing jsc
activation of carriers. In this region, Rs can be calculated by and I–V graphs (see figure 3), it can be seen that I–V curves
have similar shapes symmetrically to both sides of 50 K. Also,
RS = RS0 exp(EA /kT ),
(3) impedance curves have quite similar patterns symmetrically
where Rs0 covers all parameters that are temperature-inde- to the temper­ature of T  =  50 K in the Nyquist plot (compare
pendent and EA is an activation energy. Using equation (3), the complex impedances at 10 K and 100 K in figure 4). The
the activation energy EAI  =  23.4  ±  0.2 meV was found in sudden drop in resistance below 50 K seems to be related to
the temperature range 175 K to 285 K (region I). Similar blocking of the interface recombination. The rapid change
values in kesterites have been published before, for example in the series resistance can occur because of cooling of the
29–48 meV [19, 20]. In our sample, the activation energy recombination processes: generated holes are not able to
of 23.4 meV is related to grain boundaries or some shallow tunnel through the potential barrier into the interface region
acceptor-like defect. between CZTSe and CdS and therefore the interface recombi-
At intermediate temperatures (region II, T  =  50 K to 150 K) nation rate must be very small at these very low temperatures
another thermally activated process appears. Using equa- (T  =  10 K–50 K) [13].
tion (3), the activation energy EAII  =  33.8 meV was found IS is a powerful non-destructive tool for characterizing
in the temperature range T  =  75 K to 150 K. In our previous semiconductors. IS detects the ac current response of a system
studies [3, 4, 21] this activation energy was related to potential to a small ac voltage signal with varied frequency. The imped-
fluctuations of the valence band edge; it is the energy needed ance can be calculated by Z  =  V/I. From complex impedance
by holes to cross barriers that were created by these potential the capacitance can be calculated, but this requires a certain
fluctuations. According to the literature, the root mean square equivalent circuit to be known. The temperature dependence
depth of the potential well γ in CZTSe is usually in the range of Z′ and  −Z″ of the CZTSe thin film solar cell is shown in
γ  =  20–30 meV [3, 4, 21]. At the same time it is shown that in figure 4.
region II, the dependence of the series resistance on temper­ Our measured impedance curves in the Nyquist plot in the
ature changes and a typical Mott’s variable-range hopping whole temperature range can be fitted by applying an equiva-
conduction starts to emerge [22]. This behaviour is expected lent circuit, where a series resistor is followed by a resistor
in all heavily doped materials, where spatial potential fluc- and a capacitor connected in parallel and followed by addi-
tuations create deep potential wells for holes. But at the same tional elements (Rp2 and CPE) connected in parallel (circuit

3
J. Phys. D: Appl. Phys. 49 (2016) 085101 E Kask et al

Figure 5. Measured impedance of the CZTSe solar cell and fitting

Figure 3. Temperature dependence of jsc. Inset shows I–V results with corresponding equivalent circuits. A better fit was
curve shape similarity between both sides of 50 K at different obtained by A. Commonly used equivalent circuit B is not suitable
temperatures. at lower temperatures T    150 K. At room temperature no enormous
difference was seen.

Figure 4. Nyquist plots. Selection of impedance curves of the

CZTSe thin film solar cell in the logarithmic scale at temperatures
250 K, 100 K and 10 K. Figure 6. Fitting deviations of |Z| versus frequency at temperatures
295 K and 100 K of the CZTSe solar cell.
A in figure 5). The used complex impedance of a CPE, ZCPE
is given by
that this condition is satisfied at frequencies f  <  3.7 MHz at
1 100 K and at f  <  250 kHz at room temperature. Moreover,
Z CPE =
(4) ,
CCPE ⋅ (i ∗ ω )P measurement instrumentation capabilities have limits and
where CCPE is equal to capacitance C, if parameter P  =  1. affect measurements data. At low frequency, the fitting devia-
ω stands for angular frequency. But fitting of high temper­ tions of |Z| were not significant. Hence, IS results at higher
ature IS curves is also satisfactory with an equivalent circuit, frequencies are not trustworthy, and wide fitting deviations of
where a series resistor is followed by a resistor and a capacitor |Z| coincide with neglecting the condition Rs  (ωC)−1.
connected in parallel (circuit B in figure 5). In figure 5 both At every measured temperature values of the equivalent
theoretical fittings and experimental data curves are shown. In circuit elements can be found by exploiting the Z view pro-
figure 6 the deviation between theoretical and experimental gram for fitting. In turn, the Rs values can be used to calculate
impedance data is shown. It is clearly seen that at room the capacitance. When assuming series inductance to be insig-
temper­ature there is no difference between fittings with cir- nificant or absent, equivalent circuit B can be described by the
cuits A and B, but the deviation between fitted and measured following equation [24]:
|Z| increases at higher frequencies. At low temperatures, i.e. RP RP2Cω
T    150 K, the equivalent circuit B was not suitable at all, Z = RS +
(5) −i ,
1 + (ωRPC )2
1 + (ωRPC )2
and the equivalent circuit A fitting deviation of measured |Z|
is enormously high (more than 50%) at quite high frequen- where ω, C and RP are angular frequency, capacitance and par-
cies. Obviously, condition Rs  (ωC)−1 [10] is not satisfied at allel resistance, respectively. After rewriting (5), the capaci-
higher frequencies. It was checked by calculations and found tance can be found by

4
J. Phys. D: Appl. Phys. 49 (2016) 085101 E Kask et al

4. Conclusions

IS and current–voltage characteristics were used for charac-

terizing the CZTSe thin film solar cell. Different temperature
ranges where electrical properties change were seen. Fittings
of impedance data showed that some interface states at grain
boundaries are affected at low temperatures. These states can
be described by CPE in equivalent circuits due to grain bound-
aries with different properties. Linear fitting of temperature
dependence of Voc showed activation energy EA  =  962 meV,
which is in good agreement with the band gap of CZTSe.
Hence the dominating recombination mechanism seemed to
be bulk recombination in the studied thin film solar cell. From
temperature dependence of jsc the activation energies were
found. In the temperature range from 175 K to 285 K the acti-
Figure 7. Grain boundary located between grains in the thin vation energy of 23.4 meV is related to the energy needed
film solar cell. Some interface states are affected at quite low
temperatures and these impact fittings of measured data.
to cross the grain boundaries or some shallow acceptor-like
defect. At intermediate temperatures, from 75 K to 150 K dif-
ferent processes like carrier localization in potential wells,
−Z ″
C=
(6) , radiative recombination and Mott’s variable-range hopping
[(Z ′ − RS)2 + (−Z ″)2] ⋅ ω conduction start to come into effect, showing an activation
where Z′ and  −Z″ are the real and imaginary part of the com- energy of EAII  =  33.8 meV. In conclusion we have shown that
plex impedance, respectively. Equation (6) can be used for temperature dependent impedance and I–V studies can give
capacitance calculation, if equivalent circuit B is applied for a valuable insight into electronic processes related to grain
the fittings. In equivalent circuit A, CPE is used and in this boundaries and interfaces in solar cells.
case equations (5) and (6) would take a much more compli-
cated form and the calculation of capacitance would not be
Acknowledgment
straightforward. Thus, in the case of our studied CZTSe thin
film solar cell capacitance should be derived by using equiv- This work was supported by the Estonian Science Founda-
alent circuit B and valid results are obtained only at higher tion grant ETF 9369, by the institutional research funding IUT
temperatures (T  =  150 K to 325 K). However, IS results are 19–28 of the Estonian Ministry of Education and Research, and
also beneficial in this low temperature range, because they add by FP7 project CHEETAH, EC grant agreement no. 609788, by
complementary or confirmative details about processes. project KESTCELLS (FP7-PEOPLE-2012-ITN-316488) and
Temperature dependence of electrical parameters seen by European Regional Development Funds (ERDF, FEDER
by IS in the CZTSe thin film solar cell at T    150 K can be Programa Competitivitat de Catalunya 2007–2013). The
explained by grain boundaries, see figure 7. It seems that some authors from IREC and the University of Barcelona belong
interface states at grain boundaries are affected and change to the M-2E (Electronic Materials for Energy) Consolidated
electrical properties of the studied cell at low temperatures. Research Group and the XaRMAE Network of Excellence on
These states may not show a real capacitance, but can be Materials for Energy of the ‘Generalitat de Catalunya’. S G
explained by CPE in the equivalent circuit due to grain bound- thanks the Government of Spain for the FPI fellowship (BES-
aries with different properties. This extra capacity is caused 2014–068533) and E S for the ‘Ramon y Cajal’ fellowship
by holes being captured by interface states in grain bounda- (RYC-2011–09212).
ries, which are below the Fermi level, and released by states,
which are above the Fermi level. If the cell is contacted and
being electrically disturbed, then the band bending and hence References
crossing between Fermi level and interface states change. As
a result, charging or discharging of these states occurs and [1] Ito K and Nakazawa T 1988 Electrical and optical properties
capacitance can be detected. of stannite-type quaternary semiconductor thin films Japan.
As a result of this kind of capacitance change at low temper­ J. Appl. Phys. 27 2094
[2] Lee Y S, Gershon T, Gunawan O, Todorov T K, Gokmen T,
atures T    150 K, where grain boundary states are affected,
Virgus Y and Guha S 2015 Cu2ZnSnSe4 thin-film solar
we need to add extra elements to our equivalent circuit. At cells by thermal co-evaporation with 11.6% efficiency and
high temperatures (T  =  150 K–325 K), where grain boundary improved minority carrier diffusion length Adv. Energy
states do not contribute to capacitance, both equivalent cir- Mater. 5 1401372–5
cuits give excellent results. This low temperature (T  =  10 [3] Grossberg M, Krustok J, Raudoja J, Timmo K, Altosaar M
and Raadik T 2011 Photoluminescence and Raman study
K–150 K) behaviour, which requires the use of CPE and is
of Cu2ZnSn(SexS1−x)4 monograins for photovoltaic
present in the studied CZTSe thin film solar cells, was not applications Thin Solid Films 519 7403–6
seen in our previous research on monograin solar cells, where [4] Grossberg M, Krustok J, Timmo K and Altosaar M 2009
the role of grain boundaries was not so notable [7, 8]. Radiative recombination in Cu2ZnSnSe4 monograins