Back-Contact Recombination in Thin-Film Solar Cells

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 8, NO. 3, MAY 2018
871
Analysis of Back-Contact Interface Recombination in
Thin-Film Solar Cells
Sanjoy Paul , Sachit Grover, Ingrid L. Repins , Brian M. Keyes, Miguel A. Contreras, Kannan Ramanathan,
Rommel Noufi, Zhibo Zhao, Feng Liao, and Jian V. Li
Abstract—The open-circuit voltage (VO C ) in a generic TCO/
buffer/absorber/back-contact thin-film solar cell device is a key
parameter in the recombination analysis. In particular, VO C
is sensitively influenced by the interface recombination at the
buffer/absorber front interface and at the absorber/back-contact
interface. This paper reports the temperature, excitation light intensity, and wavelength-dependent open-circuit voltage analysis to
separate and quantify recombination rates in solar cells at the
front and back interfaces, in the depletion regions, and in the
quasi-neutral region. The wavelength-dependent VO C analysis is
exploited to extract the absorber/back-contact recombination coefficient. The experimentally observed results are verified using
SCAPS-1D (one dimensional—a solar cell capacitance simulator)
simulation.
Index Terms—Back-contact, interface, open-circuit voltage,
photovoltaics, recombination, SCAPS-1D.
I. INTRODUCTION
HE interface recombination is critical in thin-film solar
cell’s (TFSC’s) performance especially the open-circuit
voltage (VOC ) [1]–[5]. Extraction of interface recombination
rates enables more detailed analysis for material and device
engineering to reduce interface recombination and improve
VOC . VOC is a special operating point in the current–voltage
T
Manuscript received October 4, 2017; revised December 17, 2017 and March
6, 2018; accepted March 17, 2018. Date of current version April 19, 2018. This
work was supported in part by Taiwan Ministry of Science and Technology grant
number MOST 107-2218-E-006-022-MY3, in part by the U.S. Department of
Energy under Contract DE-EE-0007541 with Texas State University and DEAC36-08GO28308 with the National Renewable Energy Laboratory (NREL),
and in part by the collaborative research and development agreement between
First Solar, Inc. and NREL. (Corresponding authors: Sanjoy Paul; Jian V. Li.)
S. Paul is with the Department of Physics and Material Science, Engineering,
and Commercialization Program, Texas State University, San Marcos, TX 78666
USA (e-mail:,[email protected]).
S. Grover, Z. Zhao, and F. Liao are with First Solar, Inc., Perrysburg, OH
43551 USA (e-mail:,[email protected]; [email protected];
[email protected]).
I. L. Repins, B. M. Keyes, and R. Noufi are with the National Renewable Energy Lab, Golden, CO 80401 USA (e-mail:, [email protected];
[email protected]; [email protected]).
M. A. Contreras is with Ascent Solar Technologies, Inc., Thornton, CO 80241
USA (e-mail:,[email protected]).
K. Ramanathan is with the Stion Corporation, San Jose, CA 95119 USA
(e-mail:,[email protected]).
J. V. Li is with the Department of Aeronautics and Astronautics, National
Cheng Kung University, Tainan 70101, Taiwan, and also with the Department
of Physics and Material Science, Engineering, and Commercialization Program, Texas State University, San Marcos, TX 78666 USA (e-mail:, jianvli@
mail.ncku.edu.tw).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JPHOTOV.2018.2819664
characteristics in analyzing recombination in solar cells because at this point all generated charge carriers recombine. The
physics of recombination mechanism can be revealed from the
experimental VOC analysis with the variation of different experimental conditions such as temperature (T) [6], [7], excitation
light intensity (Ga , measured in suns) [8]–[11], and excitation
wavelength (λ) [12], [13]. Previous investigation with combined
T and Ga -dependent VOC has proven useful in identifying and
quantifying Shockley–Read–Hall (SRH) recombination rates
at different locations such as at the buffer/absorber (Ri,f ), in
the depletion region (Rd ), and in the quasi-neutral (Rb ) region
[14]. However, this analysis did not consider the recombination
at the absorber/back-contact interface Ri,b .
Here, in this paper, we extend the previous analysis to resolve recombination rates in the complete absorber of a TFSC
(from the buffer/absorber interface to absorber/back-contact interface) using excitation-wavelength-dependent VOC analysis.
The mathematical formulation to include recombination at both
buffer/absorber and absorber/back-contact interfaces is presented. We apply this methodology to two major types of solar
cells with nonnegligible (CdTe) and negligible (Cu(In,Ga)Se2 ;
CIGS) absorber/back-contact interfacial recombination. Additionally, our analysis provides suggestions to improve the interfaces and enhance the performance of the solar cells.
II. EXPERIMENTAL AND SIMULATIONS
Two major types of solar cells (CIGS and CdTe), which
share the same device architecture TCO/buffer/absorber/back
contact), were used in this analysis. The CIGS solar cells were
fabricated using the three-stage coevaporation method [15]. The
layer structure of the CIGS device (bottom to top) is—soda lime
glass substrate, 1 μm of bilayer Mo back contact, 2 μm CIGS absorber, CdS buffer layer, ZnO (TCO), and Ni–Al front-contact.
The efficiency of this device is 19.5%. Detail material synthesis
and device fabrication steps can be found in [15]. The CdTe solar cell with the 19.6% efficiency was fabricated by First Solar
Inc. [16].
The temperature and intensity-dependent current–voltage
(I–V) characteristics were measured using a standard AM1.5
light source (Newport 96000 solar simulator). A Keysight
B2912A source measure unit was used in this experiment. The
temperature of the samples was varied between 70 and 310 K.
Neutral-density filters were used to vary the intensity of incident light on the solar cell between 1 × 10−6 and 1 sun.
2156-3381 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
872
IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 8, NO. 3, MAY 2018
particular interest in this paper and its inclusion provides a more
accurate and complete recombination analysis in TFSCs. To
probe the absorber/back-contact recombination, we introduce
the wavelength-dependent VOC analysis. In analogy with the
prior recombination formalism, we consider Ri,f , Rd , Rb , and
Ri,b SRH recombination rates at the buffer/absorber interface,
in the depletion region, in the quasi-neutral region, and at
the absorber/back-contact interface. The temperature (T) and
the bias (V)-dependent recombination rates were expressed as
follows:
qV
Ri,f = R0i,f e k B T
qV
Rd = R0d e 2k B T
qV
Rb = R0b e k B T
qV
Ri,b = R0i,b e k B T
Fig. 1. Energy band diagram of a generic TCO/buffer/absorber (p-type)/backcontact heterojunction solar cell with the front interface, the depletion region,
quasi-neutral region, and the back-contact interface. Note, for VO C analysis using white light R i , f = R i , R i , b = 0, and S hf = S h [25]. Different
parameters are Ga the generation rate averaged over the absorber thickness
W, e/h the electron/hole, q the electronic charge, E C /E V /E F the energy of
conduction/valence-band/Fermi-level, E g the absorber bandgap, W d the depletion width, τ e the minority carrier lifetime in the quasi-neutral region, S hf /S eb
the surface recombination velocity of hole/electron at front/back interface, ϕ b 0
the potential barrier for majority carrier at the interface, ξi the Fermi-energy separation from the minority band edge at the interface, ξb the bulk Fermi-energy
separation from the valence band edge.
Capacitance–voltage (CV) measurements were performed using
an Agilent 4294A impedance analyzer at frequency of 10 kHz
with applied bias voltage ranging from –1.0 to 0.5 V. For extraction of back-contact recombination in CdTe and CIGS solar
cell structures, monochromatic LED light sources of different
wavelengths (625, 780, and 940 nm) were used. Observed experimental results for CdTe and CIGS solar cells were verified
via SCAPS-1D (one dimensional—a solar cell capacitance simulator) simulation [17]–[19]. All simulation parameters used in
this paper are in the Appendix. The SCAPS-1D program was
developed at ELIS (University of Gent, Belgium) to simulate
CdTe, CIGS, and other solar cells [19]–[24].
where, kB is the Boltzmann constant. R0i,f , R0d , R0b , and R0i,b
are bias-independent (V = 0) recombination coefficients at
the buffer/absorber interface, in the depletion region, in the
quasi-neutral region, and at the absorber/back-contact interface,
in that order.
These coefficients essentially account for the intrinsic properties of material or device and represent effectiveness of the
corresponding recombination pathways. Note that the exponential functions in (1) describe the densities of carriers, which are
characterized by respective diode ideality factors (2 for the recombination in the depletion region and 1 for the recombination
in the quasi-neutral region and at the interfaces) [14]. Utilizing the concept in [25] and from the energy band diagram (in
Fig. 1), we present the mathematical formula for recombination
coefficients (R0i,f , R0d , R0b , and R0i,b ) as follows:
R0i,f = Shf NV e
R0d =
Wd
ni
τe + τh
R0b =
W n2i
τ e NA
R0i,b = Seb NC e
III. MATHEMATICAL FORMULATION
In the recombination formalism, we present the generic
heterojunction solar cells structure (TCO/buffer/absorber/back
contact) in Fig. 1 to show different regions and interfaces of
interest. In the previous analysis [14], [25], quantitative extraction of recombination rates at different cell regions (at the
buffer/absorber interface Ri , in the depletion region Rd , and
in the quasi-neutral region Rb ) was performed using analytical
approximation and SRH recombination statistics [26]–[28]. Extraction of surface recombination velocity (Sh ) at the interface
(buffer/absorber) and minority carrier lifetime (τe ) were possible because of their dependency on recombination coefficients
Ri and Rb , respectively [25].
In the prior analysis, conspicuously, the recombination at
the absorber/back-contact interface was missing which is of
(1)
qφb0
B T
−k
−
( E g −ξ b )
kB T
.
(2)
The symbols in (2) have usual meanings (as noted in Fig. 1
caption). Interpretation of all recombination coefficients (R’s)
using (1) and (2) is as follows. The recombination coefficient
(Ri,f ) is proportional to the surface recombination velocity of
holes at the buffer/absorber interface (Shf ) and the hole density at
this interface. This expression assumes that the Fermi-level pinning occurs at the front interface which is applicable for high
recombination velocity or low injection level. Recombination
coefficients at the absorber/back contact (Ri,b ) is proportional
to the surface recombination velocity of electron (Seb ) and the
electron density at this interface. In the depletion region, the recombination coefficient (Rd ) is proportional to the space charge
width (Wd ), intrinsic carrier concentration (ni ), and inverse
sum of carrier lifetimes (i.e., τe + τe for electrons and holes). In
the quasi-neutral region, the recombination coefficient (Rb ) is
PAUL et al.: ANALYSIS OF BACK-CONTACT INTERFACE RECOMBINATION IN THIN-FILM SOLAR CELLS
proportional to the absorber thickness, square of intrinsic carrier
concentration (n2i ), inverse of minority carrier lifetime (τe ), and
density of electrons (minority carriers).
Now, focusing on the VOC (voltage at which all generated
charge carriers recombines), total generation across the absorber Ga W equated with the total recombination ( R =
Ri,f + Rd + Rb + Ri,b ) which leads to the following:
qVOC
qVOC
R0i,f + R0b + R0i,b e k B T + R0d e 2k B T − Ga W = 0. (3)
The solution of (3) which relates the VOC with the temperature
(T), the average generation rate Ga , and all recombination rates
(R0i,f , R0d , R0b , and R0i,b ) is given by the following:
⎡
VOC
⎜
×⎜
⎝
=
4Ga W
R0i,f
R0b
+
+
d 2
R0
R0i,b
Like the VOC analysis with white light illumination, there
are two steps to extract all recombination coefficients R0i,f ,
R0d , R0b , and R0i,b . They are 1) intensity-dependent VOC analysis at fixed T (say 300 K), and 2) temperature-dependent
VOC analysis at fixed intensity (1 sun). Intensity-dependent
VOC analysis with blue light (with K1Blue ) enables one to extract R0d and R0i,f + R0b . Now with known R0d , the intensitydependent VOC analysis with red light (with K1Red ) helps to extract R0i,f + R0b + R0i,b and finally R0i,b (by subtracting K1Blue
and K1Red ). Note that in this analysis, R0i,f and R0b are coupled
and need a secondary relationship (temperature-dependent VOC
analysis under 1 sun condition) to decouple them. In the limit
4W (R 0i +R 0b )
of K2 Ga =
Ga 1, (4) reduces to the following:
d 2
(R 0 )
1
R0d
2kB T ⎢
n ⎢
=
⎣
i,f
q
2 R + Rb + Ri,b
0
0
0
⎛
VOC
⎞⎤
(4)
Ea =
where
R0d
1
2 Ri,f + Rb + Ri,b
0
0
0
4W R0i,f + R0b + R0i,b
.
K2 =
d 2
R0
K1 =
(5)
It is clear from (4) that VOC is a function of T and Ga (excitation light intensity) which enables the extraction of all recombination coefficients experimentally. The extraction of the
interface recombination coefficients R0i,f and R0i,b required different experimental conditions beside temperature and intensitydependent VOC analysis based on white light illumination. The
wavelength-dependent VOC analysis extends the analysis to the
back contact with recombination coefficient R0i,b . For simplification, we denote the combined front and back interface recombination coefficient as R0i (prior analysis). For VOC analysis with
monochromatic short-wavelength light source (blue), the photons energy is well above the absorber’s bandgap and hence the
photogeneration is weighted toward the front interface, which
make R0i = R0i,f . On the other hand, photogeneration with longwavelength light (red), whose photon energy is barely above
the bandgap, makes near uniform generation profile across the
absorber depth, rendering R0i = R0i,f + R0i,b . So, VOC analysis
with monochromatic blue and red light sources, the coefficient
K1 can be written as follows:
R0d
1
= 2 Ri,f + Rb
0
0
K1Red =
R0d
1
.
2 Ri,f + Rb + Ri,b
0
0
0
Ga W
kB T
n
=
.
q
R0i,f + R0b + R0i,b
(7)
It was experimentally observed that near room temperature,
VOC is linear in T [25]. So, with the condition K2 Ga 1, the
activation energy of recombination Ea can be obtained using
the formula VOC = Eqa + T ( dVdTO C ) and on simplification
⎟⎥
⎥
+ 1 − 1⎟
⎠⎦
2kB T n K1
Ga K2 + 1 − 1
q
K1Blue
873
(6)
R0i,f Eai,f + R0b Eab + R0i,b Eai,b
R0i,f + R0b + R0i,b
(8)
where Eai,f , Eab , and Eai,b are the activation energies at the
buffer/absorber interface, in the quasi-neutral region, and at the
absorber/back-contact interface, respectively. So, recombination activation energy (Ea ) is the average activation energy of
Eai,f , Eab , and Eai,b weighted by respective recombination rates
(R0i,f , R0b , and R0i,b ) at V = 0.
Experimentally, the value of Ea can be extracted by extrapolating the T-dependent VOC data to T = 0 K (i.e., at T = 0 K,
Ea = VOC ) [25]. With the knowledge (will be discussed in the
result section) of the activation energies Eai,f , Eab , and Eai,b , one
Ri,f
can obtain R0 b which enables one to decouple R0i,f and R0b .
0
Now combining (3), (6), and (8), the four unknown recombination coefficients R0i,f , R0d , R0b , and R0i,b are readily solvable
via the matrix method—with the transformation matrix
⎞
⎛
qVOC
qVOC
qVOC
qVOC
e kB T
e 2k B T
e kB T
e kB T
⎟
⎜
⎟
⎜
1
0
1
− 2K 1B l u e
⎟
⎜
1
⎟
⎜
⎟
⎜
⎠
⎝
1
− 2K1R e d
1
1
1
i,f
b
i,b
0
Ea − Ea
Ea − E a
Ea − E a
⎛ i,f ⎞ ⎛
⎞
R0
Ga W
⎜ Rd ⎟ ⎜
0 ⎟
⎜
⎟
⎟.
× ⎜ 0b ⎟ = ⎜
(9)
⎝
0 ⎠
⎝ R0 ⎠
0
Ri,b
0
With this recombination analysis, other parameters such as
surface recombination velocity of holes at the buffer/absorber
interface (Shf ), the surface recombination velocity of electron
at the absorber/back contact (Seb ), and the electron lifetime
(τe ) can be extracted using (2). In extracting Seb , ξb (the bulk
Fermi level above valance) can be obtained from the equation
874
IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 8, NO. 3, MAY 2018
fitting experimental data with (4) yield K1 = 6.18 × 105 and
K2 = 1561.4 Sun−1 . Note that white light illumination does not
include the back-contact recombination term in the expression
of K1 and K2 . From the values of K1 and K2 , we obtain the
= 3.7 × 108 cm−2 s−1 and using this R0d
value of R0d = K2W
1K 2
value the values of R0i + R0b = 299.4 cm−2 s−1 . Note that here
both R0i and R0b are coupled and need other experiments to decouple them. The temperature-dependent VOC analysis at 1 sun
4G W (R i +R b )
condition (with verified K2 Ga = a d 02 0 1) leads to
(R 0 )
estimate activation energy of recombination Ea = 1.45 eV [see
Fig. 3(b)]. With known value of ϕb0 = 1.17 V (from capacitance measurement—data not shown) and Eg = 1.5 eV, the estimated ration of R0i and R0b is
3 × 102 cm−2 s−1 and
Fig. 2. Flow chart for the recombination analysis with the thin-film solar cell
to extract various recombination coefficients, surface recombination velocities,
and electron lifetime as mentioned in the formalism section.
R 0i
R 0b
R 0i
R 0b
≈ 0.18. Using R0i + R0b ≈
= 0.18, the estimated value of R0i =
45.8 cm−2 s−1 and R0b = 254.2 cm−2 s−1 . Considering VOC =
qVOC
888 meV, the recombination coefficients are Ri = R0i e k B T =
qVOC
4.1 × 1016 cm−2 s−1 , Rd = R0d e 2k B T = 1.1 × 1016 cm−2 s−1 ,
qVOC
Fig. 3. (a) Dependence of V OC on excitation light intensity (G) at T ࣈ 300 K.
(b) Temperature (T)-dependent V OC at one Sun illumination. In both figures,
the squares and dashed lines represent the experimental data and fitted lines,
respectively.
qϕb0 = qϕbi + ξb . The flow chart summary of extracting recombination coefficients along with various electronic parameters (surface recombination velocity and electron lifetime) using
white light and monochromatic light illumination is represented
in Fig. 2 for simplicity.
Here onward, the experimental data analysis on CdTe and
CIGS solar cells will be discussed which includes the extraction
of recombination coefficients at the buffer/absorber interface,
in the depletion region, in the quasi-neutral region, and at the
absorber/back-contact interface.
IV. RESULTS AND DISCUSSION
A. Recombination Analysis With CdTe Solar Cells
The dependence of VOC on illumination (white light) intensity at T ࣈ 300 K and on T at 1 sun illumination is
shown in Fig. 3(a) and (b), respectively. From Fig. 3(a), the
and Rb = R0b e k B T = 2.3 × 1017 cm−2 s−1 .
Using (2) (R0i,f = R0i , and Shf = Sh ) and considering ni =
9 × 105 cm−3 , NV = 1.8 × 1019 cm−3 , NA = 1 × 1015 cm−3 ,
and ϕb0 = 1.17 eV, we extract the values of Sh = 1.25 ×
102 cm/s and τe = 1.59 ns. The summary of extracted recombination coefficients, surface recombination velocity at
buffer/absorber interface, and the electron lifetime in this analysis is tabulated in Table I.
In case of CdTe device, back-contact optimization is a known
challenge [29], [30]. The previous analysis with white light
source fails to probe back-contact recombination. To extend the
recombination at the back contact, the intensity (Ga )-dependent
VOC at excitation wavelength λEx = 625 nm, and λEx = 780 nm
are shown in Fig. 4(a). Now, from the fitting using (4), the
estimated value of K1Blue = 8.64 × 105 , K1Red = 6.03 × 105 ,
K2Blue = 610.4 Sun−1 , and K2Red = 1134.4 Sun−1 , respectively.
Utilizing (6), the transformation matrix in (9), and experimental results, one can easily estimate the recombination coefficients (at V = 0) from front- to back contact (R0i,f , R0d , R0b ,
and R0i,b ). With the relation K1Blue · K2Blue = 2W
, one obtains
Rd
0
the value of R0d = 3.7 × 108 cm−2 s−1 . With simplified equaR 0i , b
R 0d
1
− 2K 1B l u e and R0d = 3.7 × 108 cm−2 s−1 ,
2K 1R e d
1
we obtain R0i,b = 93 cm−2 s−1 . Using matrix equation (9),
we estimate the R0i,f + R0b = 214 cm−2 s−1 and R0i,f + R0b +
R0i,b = 307 cm−2 s−1 and finally R0i,b = 93 cm−2 s−1 (as ob-
tion
=
served previously). Now with blue light excitation and using (8) and with known values [Ea = 1.45 eV, Eab = Eai,b =
Eg = 1.5 eV, Eai,f = ϕb0 (front) = 1.17 V] we obtain the
values of R0b = 167 cm−2 s−1 , and R0i,f = 47 cm−2 s−1 . At
V = VOC , we estimate the recombination coefficients Ri,f =
3.4 × 1016 cm−2 s−1 , Rd = 1.0 × 1016 cm−2 s−1 , Rb = 1.1 ×
1017 cm−2 s−1 , and Ri,b = 6.3 × 1016 cm−2 s−1 .
By considering NV = 1.8 × 1019 cm−3 and ϕb0 = 1.17 eV,
the observed front surface recombination velocity of hole (Shf )
PAUL et al.: ANALYSIS OF BACK-CONTACT INTERFACE RECOMBINATION IN THIN-FILM SOLAR CELLS
875
TABLE I
EXTRACTED RECOMBINATION COEFFICIENTS, SURFACE RECOMBINATION VELOCITY, AND LIFETIME OF MINORITY
CARRIER IN CDTE SOLAR CELL USING WHITE LIGHT SOURCE
Fig. 4. (a) Illumination-dependent V OC (in CdTe cell) with a different excitation light source wavelength of λEx = 625 and 780 nm. (b) Dependence of
V OC with excitation wavelength λEx = 600 and 800 nm generated using the
SCAPS-1D program.
and lifetime of electron were 1.25 × 102 cm/s and τe = 2.43 ns
using (2). The surface recombination velocity at the back contact
(Seb ) was extracted from a known value of NC = 8 × 1017 cm−3
and ϕbi = 0.92 V as 1.28 × 105 cm/s. All extracted recombination coefficients, surface recombination velocities, and electron
lifetime are in Table II.
It is important to note that the recombination rate in the
quasi-neutral region Rb ≈ 1017 cm−2 s−1 is overestimated because it implicitly includes the back-contact recombination rate
Ri,b . This results in a lower estimate of minority-carrier lifetime τe . To verify further, we simulated intensity-dependent
VOC with monochromatic light of wavelength λEx = 600 and
800 nm in SCAPS-1D (using the baseline structure: Au (back
contact)/CdTe (5 μm)/CdS (25 nm)/SnOx (0.5 μm)) as shown
in Fig. 4(b). Intensity-dependent VOC data simulated by SCAPS
-1D helps to understand our experimentally observed data. Here
we did not try to match the simulated data with the experimental
one and fit it with the discussed method. Dependency of VOC on
too many semiconducting material and device parameters made
this task difficult. The set of parameters used for the simulation
is based on experimentally extracted device parameters (such
as NA , Eg , τe , Shf , Seb ,W, etc.) while keeping other parameters
(as available in SCAPS-1D for CdTe-baseline) unchanged. The
errors associated with the experimentally observed material and
devices parameters have impact on VOC which contribute to the
difference between simulation and experiment. On the other
hand, uniform distribution of traps at mid-gap was considered
in this simulation. But in actual device nonuniform and complicated energetic distribution of traps is possible. Finally, while
simulation provides consistent comparison with the experimental results but simulation using more realistic device parameters
could be of particular interest in the future.
Extraction of ϕbi in a device with very thin absorber is challenging compared with semi-infinite thick absorber. Due to this
reason, for example, there is considerable difficulty in extracting
ϕbi with CdTe device with thickness of about 3–5 μm, which
could also be compounded by the presence of a non-ohmic back
contact. The preceding (white light) analysis on the CdTe device
was carried out using a ϕbi value of 0.92 V. Regretfully, this is an
educated guess based on VOC value instead of an actual extraction ϕbi from the Mott–Schottky plot, which would have been
0.75 V. Therefore, further methodological improvement is necessary on this topic. Moreover, because the error in ϕbi directly
translates to the same error in ϕb0 , the extraction of Sh is affected
because of its exponential dependency on ϕb0 [14], [25]. As an
example, changing ϕbi to 1.02 V—and error of 0.1 V—would
lead to a seemingly alarming change in Shf = 6.24 × 103 cm/s
as compared with Sh = 1.25 × 102 cm/s (using ϕbi = 0.92 V)
while keeping τe relatively unchanged at ∼ 1.5 ns. In this case, it
is perhaps more informative to inspect the recombination coefficients and observed values are as follows: R0i = 58 cm−2 s−1 ,
R0d = 1.1 × 108 cm−2 s−1 , and R0b = 189 cm−2 s−1 at V =
0 and Ri = 5.1 × 1016 cm−2 s−1 , Rd = 3.4 × 1015 cm−2 s−1 ,
and Rb = 1.7 × 1017 cm−2 s−2 at V = VOC which does not deviate from the previous analysis (see Table I). Note the spatial distribution of excess carrier density in steady state depends on the
absorption coefficient (generation), the diffusion length (transport and bulk recombination), and the surface recombination
(boundary conditions) [31]. If both l and 1/α are short compared
with the absorber thickness, then the back-contact recombination does not play a significant role. The opposite is true if either
or both l and 1/α is long compared with the absorber thickness.
B. Recombination Analysis With CIGS Solar Cells
The estimation of recombination coefficients at
buffer/absorber interface, in the depletion region, and in
the bulk of the absorber of CIGS devices have been investigated
876
IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 8, NO. 3, MAY 2018
TABLE II
EXTRACTED RECOMBINATION COEFFICIENTS, SURFACE RECOMBINATION VELOCITIES, AND ELECTRON LIFETIME IN CDTE
SOLAR CELLS USING BLUE AND RED LIGHT SOURCE
Intensity-dependent VOC data with a CIGS baseline structure [Mo (back contact)/p-CIGS (2 μm)/n-CdS
(50 nm)/n-ZnO (50 nm)] were generated using SCAPS-1D with
white light and different excitation wavelength (λEx = 600 and
900 nm) as shown in Fig. 5(b). In this simulation, we created the
bandgap profile via a two-segment linear compositional grad[Ga]
) in CuIn1−x Gax Se2 .
ing scheme with Ga content (x = [Ga]+[In]
The bandgap in this structure linearly varies from 1.15 eV (at
CdS/CIGS contact) to 1.05 eV at a depth of 0.5 μm in the
CIGS absorber and then to 1.3 eV (at CIGS/back contact).
The simulated data have qualitative consistency with the experimentally observed intensity-dependent VOC [see Fig. 5(a)].
Almost equal observed value of K1Blue and K1Red may be
attributed to the bandgap grading. However, the actual cell
may have complicated bandgap grading which compels further
examination.
V. CONCLUSION
Fig. 5. (a) Intensity-dependent V OC data for CIGS solar cell with two different
excitation wavelengths λEx = 625 and 940 nm as well as with white light. Here
symbols are experimental data and dotted lined are fitted data. (b) SCAPS-1D
simulated intensity-dependent V OC data with white light and three different
available wavelengths λEx = 600 and 900 nm.
and published elsewhere [25]. Here, the wavelength-dependent
VOC analysis on three-stage coevaporated CIGS devices has
been performed to confirm that the absorber/back-contact
recombination is indeed negligible. This result is presented in
Fig. 5(a) which shows intensity-dependent VOC at two excitation
laser light λEx = 625 and 940 nm. The fitting with (4) yield
almost identical K1 values K1Blue = 20918.8 ± 605.7 versus
K1Red = 21308.0 ± 2119 within experimental errors. This is
due to the bandgap gradient of the CIGS absorber deposited by
three-stage coevaporation. This bandgap gradient is manifested
as more elevated conduction band toward the back contact and
function as a back-surface field commonly observed a-Si/c-Si
heterojunction device [32]. This feature repels minority
electrons away from the back contact and effectively reduces
interface recombination rate.
We have investigated the temperature, illumination intensity,
and wavelength-dependent VOC analysis [VOC (T, I, λ) analysis] with generic TCO/buffer/absorber/back-contact solar cell
(CdTe and CIGS) in order to separate and quantify the recombination coefficients at buffer/absorber interface (Ri,f ), in the
depletion region (Rd ) and bulk of the absorber (Rb ), and at absorber/back contact (Ri,b ). The VOC (T, I) analysis with white
light illumination can be used to extract the recombination rates
Ri,f , Rd , and Rb while neglecting Ri,b . This analysis is valid
under certain conditions such as the a-Si/c-Si heterojunction
solar cells with a back-surface field/passivation and three-stage
coevaporated CIGS solar cells due to the advantageous bandgap
gradient. Ri,f depends on two main factors: 1) surface recombination velocity, and 2) the potential barrier height ϕb0 , the latter
being constrained by the alignment of the conduction band of
the TCO and the valence band of the absorber. The complete
TCO/buffer/absorber structure of the CIGS device is analyzed
to reveal the dependence of ϕb0 on bandgap, band offset, band
bending, and inversion to devise strategies for reducing the absolute rate of interface recombination as well as the relative
strength of interface recombination over bulk recombination.
Furthermore, we improved upon the aforementioned method to
extract Ri,b further via wavelength dependent VOC analysis and
tested with CdTe solar cell in which the recombination at the
back contact is nonnegligible. This characterization method will
be useful in other heterojunction devices where the interface recombination is crucial, i.e., Cu2 O [33], CuZnSnS4 [34]–[36],
and even CIGS with high Ga content where bandgap gradient is
not optimized [15].
PAUL et al.: ANALYSIS OF BACK-CONTACT INTERFACE RECOMBINATION IN THIN-FILM SOLAR CELLS
877
APPENDIX
THE MATERIALS AND DEVICE PARAMETERS OF CIGS AND CDTE SOLAR CELLS (DIFFERENT LAYERS, FRONT-CONTACT,
AND BACK CONTACT) USED IN SCAPS-1D SIMULATION
ACKNOWLEDGMENT
The authors would like to thank Dr. M. Gloecker for valuable
discussion and Dr. M. Burgelman for providing SCAPS-1D simulation software utilized in generating solar cell band structure
and to verify the experimental observation.
REFERENCES
[1] C. J. Hages et al., “Improved performance of Ge-alloyed CZTGeSSe thinfilm solar cells through control of elemental losses,” Prog. Photovolt. Res.
Appl., vol. 23, no. 3, pp. 376–384, Mar. 2015.
[2] C. S. Ponseca et al., “Organometal halide perovskite solar cell materials
rationalized: Ultrafast charge generation, high and microsecond-long balanced mobilities, and slow recombination,” J. Amer. Chem. Soc., vol. 136,
no. 14, pp. 5189–5192, Apr. 2014.
[3] M. Taguchi et al., “24.7% record efficiency HIT solar cell on thin silicon
wafer,” IEEE J. Photovolt., vol. 4, no. 1, pp. 96–99, Jan. 2014.
[4] T. K. Todorov et al., “Beyond 11% efficiency: Characteristics of state-ofthe-art Cu 2 ZnSn(S,Se) 4 solar cells,” Adv. Energy Mater., vol. 3, no. 1,
pp. 34–38, Jan. 2013.
[5] W. Wang et al., “Device characteristics of CZTSSe thin-film solar cells
with 12.6% efficiency,” Adv. Energy Mater., vol. 4, no. 7, May 2014,
Art. no. 1301465.
[6] V. Nadenau, U. Rau, A. Jasenek, and H. W. Schock, “Electronic properties
of CuGaSe2-based heterojunction solar cells. Part I. Transport analysis,”
J. Appl. Phys., vol. 87, pp. 584–593, 2000.
[7] U. Rau and H. W. Schock, “Electronic properties of Cu(In,Ga)Se 2 heterojunction solar cells-recent achievements, current understanding, and
future challenges,” Appl. Phys. A Mater. Sci. Process., vol. 69, no. 2,
pp. 131–147, Aug. 1999.
[8] D. Abou-Ras, T. Kirchartz, and U. Rau, Eds., Advanced Characterization Techniques for Thin Film Solar Cells. Weinheim, Germany: Wiley,
2011.
[9] O. Gunawan, T. Gokmen, and D. B. Mitzi, “Suns- V O C characteristics
of high performance kesterite solar cells,” J. Appl. Phys., vol. 116, no. 8,
Aug. 2014, Art. no. 084504.
[10] A. Halverson, A. Yakimov, A. Vert, O. Sulima, and B. Korevaar, “Sideby-side characterization of non-optimized and optimized CdS/CdTe solar
cells on commercial transparent conductive oxide/glass,” Thin Solid Films,
vol. 535, pp. 249–252, May 2013.
[11] R. Sinton and A. Cuevas, “A quasi-steady-state open-circuit voltage
method for solar cell characterization,” in Proc. 16th Eur. Photovolt. Sol.
Energy Conf. Proc., 2000, pp. 1152–1155.
[12] J. P. Babaro, K. G. West, and B. H. Hamadani, “Spectral response measurements of multijunction solar cells with low shunt resistance and breakdown
voltages.,” Energy Sci. Eng., vol. 4, no. 6, pp. 372–382, Nov. 2016.
[13] H. Mäckel and A. Cuevas, “Generalized models of the spectral response
of the voltage for the extraction of recombination parameters in silicon
devices,” J. Appl. Phys., vol. 98, no. 8, Oct. 2005, Art. no. 083708.
[14] S. Grover, J. V. Li, D. L. Young, P. Stradins, and H. M. Branz, “Reformulation of solar cell physics to facilitate experimental separation of
recombination pathways,” Appl. Phys. Lett., vol. 103, no. 9, Aug. 2013,
Art. no. 093502.
[15] M. A. Contreras et al., “Wide bandgap Cu(In,Ga)Se 2 solar cells with improved energy conversion efficiency,” Prog. Photovolt. Res. Appl., vol. 20,
no. 7, pp. 843–850, Nov. 2012.
[16] M. A. Green, K. Emery, Y. Hishikawa, W. Warta, and E. D. Dunlop, “Solar
cell efficiency tables (version 47),” Prog. Photovolt. Res. Appl., vol. 24,
no. 1, pp. 3–11, Jan. 2016.
[17] M. Djinkwi Wanda, S. Ouédraogo, F. Tchoffo, F. Zougmoré, and
J. M. B. Ndjaka, “Numerical investigations and analysis of Cu2 ZnSnS4
based solar cells by SCAPS-1D,” Int. J. Photoenergy, vol. 2016, 2016,
Art. no. 2152018.
[18] E. H. Ihalane, L. Atourki, H. Kirou, A. Ihlal, and K. Bouabid, “Numerical
study of thin films CIGS bilayer solar cells using SCAPS,” Mater. Today
Proc., vol. 3, no. 7, pp. 2570–2577, 2016.
[19] A. Niemegeers, M. Burgelman, K. Decock, J. Verschraegen, and
S. Degrave, “Manual for SCAPS-1D,” 2013. [Online]. Available:
http://users.elis.ugent.be/ELISgroups/solar/projects/scaps/SCAPS Manual 2 september 2013.pdf. Accessed on: Mar. 18, 2017].
878
[20] B. E. Pieters, K. Decock, M. Burgelman, R. Stangl, and T. Kirchartz,
“One-dimensional electro-optical simulations of thin-film solar cells,” in
Advanced Characterization Techniques for Thin Film Solar Cells. Weinheim, Germany: Wiley, 2011, pp. 501–527.
[21] “SCAPS-1D simulation programme,” 2017. [Online]. Available:
http://scaps.elis.ugent.be/. Accessed on: Mar. 18, 2017.
[22] S. Bansal and P. Aryal, “Evaluation of new materials for electron and hole
transport layers in perovskite-based solar cells through SCAPS-1D simulations,” in Proc. 2016 IEEE 43rd Photovolt. Spec. Conf., 2016, pp. 0747–
0750.
[23] K. Kim et al., “Simulations of chalcopyrite/c-Si tandem cells using
SCAPS-1D,” Sol. Energy, vol. 145, pp. 52–58, Jan. 2017.
[24] C.-H. Huang and W.-J. Chuang, “Dependence of performance parameters
of CdTe solar cells on semiconductor properties studied by using SCAPS1D,” Vacuum, vol. 118, pp. 32–37, Aug. 2015.
[25] J. V. Li, S. Grover, M. A. Contreras, K. Ramanathan, D. Kuciauskas, and
R. Noufi, “A recombination analysis of Cu(In,Ga)Se2 solar cells with
low and high Ga compositions,” Sol. Energy Mater. Sol. Cells, vol. 124,
pp. 143–149, May 2014.
[26] S. M. Sze and K. K. Ng, Physics of Semiconductor Devices. Hoboken, NJ,
USA: Wiley, 2007.
[27] R. N. Hall, “Electron-hole recombination in germanium,” Phys. Rev.,
vol. 87, no. 2, pp. 387–387, Jul. 1952.
[28] W. Shockley and W. T. Read, “Statistics of the Recombinations of Holes
and Electrons,” Phys. Rev., vol. 87, no. 5, pp. 835–842, Sep. 1952.
[29] N. R. Paudel and Y. Yan, “CdTe thin-film solar cells with cobaltphthalocyanine back contacts,” Appl. Phys. Lett., vol. 104, no. 14,
Apr. 2014, Art. no. 143507.
IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 8, NO. 3, MAY 2018
[30] J. Sites and J. Pan, “Strategies to increase CdTe solar-cell voltage,” Thin
Solid Films, vol. 515, no. 15, pp. 6099–6102, May 2007.
[31] M. Turek, “Interplay of bulk and surface properties for steady-state measurements of minority carrier lifetimes spectral dependence of carrier lifetimes in silicon for photovoltaic applications Interplay of bulk and surface
properties for steady-state measurements of minority carrier lifetimes,” J.
Appl. Phys., vol. 1111, no. 10, pp. 123703–233108, 2012.
[32] J. V. Li et al., “Electrical characterization of interface recombination and
its dependence on band offset, potential barrier height, and inversion in
certain heterojunction solar cells,” in Proc. 2014 IEEE 40th Photovolt.
Spec. Conf., 2014, pp. 0686–0690.
[33] Y. S. Lee et al., “Atomic Layer deposited gallium oxide buffer layer
enables 1.2 v open-circuit voltage in cuprous oxide solar cells,” Adv.
Mater., vol. 26, no. 27, pp. 4704–4710, Jul. 2014.
[34] O. Gunawan, T. K. Todorov, and D. B. Mitzi, “Loss mechanisms in
hydrazine-processed Cu2ZnSn(Se,S)4 solar cells,” Appl. Phys. Lett.,
vol. 97, no. 23, Dec. 2010, Art. no. 233506.
[35] S. Bag, O. Gunawan, T. Gokmen, Y. Zhu, and D. B. Mitzi, “Hydrazineprocessed ge-substituted CZTSe solar cells,” Chem. Mater., vol. 24, no. 23,
pp. 4588–4593, Dec. 2012.
[36] D. A. R. Barkhouse, O. Gunawan, T. Gokmen, T. K. Todorov, and
D. B. Mitzi, “Device characteristics of a 10.1% hydrazine-processed
Cu2ZnSn(Se,S)4 solar cell,” Prog. Photovolt. Res. Appl., vol. 20, no. 1,
pp. 6–11, Jan. 2012.
Authors’ photographs and biographies not available at the time of publication.