Solar Energy 146 (2017) 443–452

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

Optimisation of rear reflectance in ultra-thin CIGS solar cells towards

>20% efficiency
Olivier Poncelet a, Ratan Kotipalli a,⇑, Bart Vermang b,c, Angus Macleod d, Laurent A. Francis a, Denis Flandre a
a
ICTEAM, Université catholique de Louvain, Louvain-la-Neuve 1348, Belgium
b
ESAT-KU Leuven, University of Leuven, Leuven 3001, Belgium
c
IMEC, Kapeldreef 75, Leuven 3001, Belgium
d
Thin Film Center Inc, 2745 E Via Rotunda, Tucson, AZ 85716-5227, USA

a r t i c l e i n f o a b s t r a c t

Article history: In order to decrease their cost and the use of rare metal elements, thin film solar cell thicknesses are con-
Received 23 December 2016 tinuously reduced at the expense of their efficiency, due to a lack of absorption for long wavelengths.
Received in revised form 21 February 2017 Optimisation of cells rear reflectance (Rb ) thus becomes meaningful to provide non-absorbed light a sec-
Accepted 1 March 2017
ond chance to be harvested by the active cell layer. In this sense, we present a way to keep the rear reflec-
tance in advanced Cu(In, Ga) Se2 (CIGS) cell as high as possible while keeping in mind the progress
already done regarding the rear passivation techniques. We show that introducing a stack of thin
Keywords:
Al2O3 and aluminium between the CIGS layer and the rear molybdenum electrode increases Rb up to
Rear reflection
just CIGS solar cells
92% in the long wavelength 800–1100 nm range. Several other stacks, using MgF2, SiO2 or TiO2, are opti-
Dielectric passivation mised in order to investigate the best trade-off between passivation, material consumption and perfor-
SCAPS-1D mances, resulting in Rb ranging from 42% (moderate case) to 99% in the best case. Those CIGS rear
Transfer matrix interface reflectance optimisations were performed by using a standard transfer matrix method (TMM)
in the long wavelength range. Seven interesting stacks are then analysed for solar cell performances using
SCAPS simulation software to understand the impact of rear reflectance on short circuit current density
(Jsc) and eventually on the cell efficiency (g), for ultra-thin CIGS absorber thicknesses (<1 lm). Based on
these results, we propose Rb optimisation to achieve Jsc > 40 mA/cm2 and g > 20% with a 500 nm-thick
CIGS absorber film using CIGS-Al2O3-Mo stack, where the Al2O3 thickness can be chosen in between
104 and 139 nm. This way, we can ensure good rear reflectance (Rb = 65%) and reduced interface recom-
bination while being industrially feasible with present technologies.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction mising the rear reflectance (Rb ) to give light a second chance to
be absorbed while taking into consideration the highly recombina-
CIGS solar cells are presently considered to be the best thin film tive CIGS/Molybdenum (Mo) rear contact interface is then manda-
absorbing material in terms of their excellent light-to-power con- tory in order to retain high efficiency devices (Krc et al., 2016).
version efficiencies exceeding 20% (Ramanathan et al., 2003; Although rear interface passivation is already addressed by
Jackson et al., 2011). Unfortunately, their use of rare metals is a using aluminium oxide (Al2O3) (Vermang et al., 2014b,c; Kotipalli
challenge when aiming for a long-term marketability. Since other et al., 2015b, in press), efforts are still needed to improve Rb . Up
industrial applications (e.g., LED, high-frequency transistors, infra- to now, the most interesting approach is to use nano-structures
red detectors, . . .) also use gallium and indium, which are in short (Ji et al., 2013; Bednar et al., 2015; Yin et al., 2016) or nano-
supply, it is of the utmost importance to reduce their use if CIGS particles on the rear side to build a highly reflective structure
thin-film solar cells are to be produced in large volumes. While which scatters the light back into the CIGS layer (Vermang et al.,
moving towards ultra-thin (
500 nm) CIGS solar cells will reduce 2015a; Lare et al., 2015). Although these methods are very effec-
the costs and thus bring economic advantages, it will also induce tive, their optimisation and their homogeneity on large areas are
a loss of absorption within the cells (Andreani et al., 2012). Opti- still a challenge. Some studies have already shown that replacing
molybdenum with gold or silver improve the rear reflectance, thus
increasing the cell performances (Dahan et al., 2012; Li-Kao et al.,
⇑ Corresponding author. 2012). Other metals like W, Cr or Ta (Orgassa et al., 2003a),
E-mail address: raja.kotipalli@uclouvain.be (R. Kotipalli).

http://dx.doi.org/10.1016/j.solener.2017.03.001
0038-092X/Ó 2017 Elsevier Ltd. All rights reserved.
444 O. Poncelet et al. / Solar Energy 146 (2017) 443–452

zirconium nitride (ZrN) (Krc et al., 2016; Schleussner et al., 2009) 1 lm-thick CIGS is already enough to absorb 97% of the input irra-
or titanium nitride (TiN) (Mahieu et al., 2011) have been tested diance for wavelengths comprised between 300 and 800 nm, while
and have shown good results. The backwall supersaturate struc- silicon exhibits the same behaviour only between 300 nm and
ture, where the glass substrate is used as the front-part of the 434 nm (Fig. 1b). This statement does not hold true for long wave-
device, gives an other effective alternative to improve ultra-thin length. In the case of 1 lm-thick CIGS, wavelengths above 800 nm
film solar cell performances by using silver (Ag) reflector on the become less and less absorbed until the corresponding energy is
rear side (Larsen et al., 2014; Simchi et al., 2014). equal to the bandgap energy. This absorption loss becomes really
We propose with this work an alternative industrially viable critical for thicknesses below 1 lm (Fig. 1) because a substantial
and scientifically interesting method using optimised rear material part of the input irradiance is not fully absorbed, thus decreasing
stacking instead of complex photonics nano-structures. This the external quantum efficiency (EQE) of the device. Managing rear
means, by the use of few layers of dielectric and metal using indus- reflection (Rb ) while decreasing the active thickness in ultra thin
trial or other deposition techniques, we can achieve similar Rb CIGS cells thus becomes mandatory to keep the efficiency as high
while simultaneously providing the rear surface passivation and as possible (Vermang et al., 2014b; Kotipalli et al., in press).
an excellent diffusion barrier against elements moving from the
base substrate (especially stainless steel) during the high temper- 2.2. Theoretical background
ature process step.
In this work, rear reflectance optimisation is addressed by using In thin-film optics, the ratio between the magnetic field ampli-
a transfer-matrix method. These methods are well-known in thin- tude H and the electric field amplitude E of a single linearly-
film optics to investigate optical properties of material stacks (Hass polarised plane wave in a non-magnetic (lr ¼ 0) medium is given by
and Gerstenberg, 1964; Pedrotti and Pedrotti, 1993; Angus
H
Macleod, 2001). The method allows the study of Rb , that is the ~Y 0
¼y¼n
E
amount of light reflected back into the CIGS, independently of
the rest of the cell. The impact of Rb on the current density (Jsc) ~
where y is the characteristic optical admittance of the medium, n
qffiffiffiffi
and cell efficiency (g) is then addressed by incorporating the trans-
0
the complex refractive index and Y 0 ¼ l the admittance of the
0
fer matrix results in a 1-D Solar Cell CaPacitance Simulator (SCAPS,
free space (Angus Macleod, 2001). Reflection on a diopter at normal
University of Ghent, Belgium), to give guidelines for cells
incidence is simply given by
fabrication.
This article is divided in two parts. The first part exposes the y0  ys
r¼ and R ¼ r  r
theoretical optical issues when CIGS thickness is reduced and pre- y0 þ ys
sents succinctly the transfer matrix method used in this work, nec-
where r is the reflection coefficient, R the reflectance, y0 the charac-
essary to support the results and the discussion. The second part of
teristic admittance of the incident medium and ys the characteristic
this work presents the effect of some optimised thin-film stacks
admittance of the substrate. A high contrast between y0 and ys will
from the first part on the global CIGS cell parameters (Jsc and effi-
give a high reflectance.
ciency) by using SCAPS.
When the admittances are replaced by their definitions, we
come back to the well-known Fresnel coefficient. In a thin-film sys-
2. Optical simulations tem (Fig. 2), the reflection coefficient keeps the same form (Angus
Macleod, 2001),
In the following Sections 2.1–2.5, we go through CIGS rear
y0  Y
reflectance optimisation and highlight the best stacks studied here. r¼
All the material optical parameters to study a CIGS cell can be
y0 þ Y
found in Orgassa et al. (2003b), Han et al. (2004), Chen et al. but Y is now the input optical admittance of the system and is
(2015), Rubio-Bollinger et al. (2015), Bernal-Correa et al. (2016), defined by H1 =E1 where H1 (E1 ) is the amplitude of the magnetic
and Onwudinanti et al. (2016) and are summarised in the supple- field (electric field) at the input interface (denoted by 1 in Fig. 2).
mentary data. The input fields are linked with the output fields through a charac-
teristic matrix M which only depends on the thin-film parameters
2.1. CIGS absorption (Hass and Gerstenberg, 1964; Pedrotti and Pedrotti, 1993):
    " # 
i
E1 E2 cos d sin d E2
Optical properties of materials are mainly explained by their ¼M ¼ y1

complex refractive index n ~ ðkÞ ¼ nðkÞ  ikðkÞ, where n is the real H1 H2 iy1 sin d cos d H2
refractive index and k is the extinction coefficient, both being a
where y1 is the thin-film characteristic admittance and d ¼ 2py1 d=k
function of k, the wavelength at which materials are exposed.
is the phase factor due to the physical thickness d of the film. By
The extinction coefficient governs how the spectral irradiance is
normalising the fields by E2 , we get:
decreasing while the light propagates through the material. This " #
  
is expressed through the Lambert-Beer’s law as E1 =E2 cos d i
sin d 1
¼ y1

IðkÞ 4pk H1 =E2 iy1 sin d cos d ys

¼ eaz with a ¼
I0 ðkÞ k We can then express the input admittance Y as a function of the
where IðkÞ is the irradiance after a propagation length z (in metre) in thin-film parameters and the substrate admittance:
the medium, I0 ðkÞ is the initial spectral irradiance and a is the H1 ys cos d þ iy1 sin d
absorption coefficient. In a medium free from absorption, k(k) = 0 Y¼ ¼
E1 cos d þ i yys sin d
1
while it can reach high values in absorbing media such as metals
or semiconductors. If the thin-film is replaced by a multilayer, the characteristic
Computing this equation for CIGS material using the AM1.5 D matrix of the system is the product of the characteristic matrices
solar spectrum as I0 ðkÞ shows that CIGS is an excellent absorber of all layers M ¼ M 1 M 2 . . . M n . TMM consists to compute M the
compared with silicon (Fig. 1a), due to its direct bandgap. A characteristic matrix of the studied system to obtain r then R.
 O. Poncelet et al. / Solar Energy 146 (2017) 443–452 445

(a) (b)

Fig. 1. (a) AM1.5 D spectral irradiance before and after a propagation through different CIGS thicknesses. The curves are computed using the Lambert-Beer law. The labels
correspond to the propagation length (z). (b) Relative absorption of the solar spectrum by CIGS. This graph shows the amount of light already absorbed by the CIGS. Silicon
(red line) is added as a reference. Front reflection and diffusion through CIGS losses are not taken into account. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

material refractive index. That will be used in this work to validate

the results.
Finally, TMM used here does not explicitly incorporate the
effect of the roughness of the layers. It is still possible to take
roughness into account in TMM by using statistical representation
of the roughness but it depends on the correlation length and
needs thorough experimental measurements. More insight could
be found in the following literature (Mitsas and Siapkas, 1995;
Katsidis and Siapkas, 2002; Tikhonravov et al., 2003; Dahan et al.,
2013).
Fig. 2. Thin film model used in transfer matrix. An example of corresponding
materials are indicated between brackets. 2.3. CIGS rear interfaces

An advanced schematic cross-section of a CIGS cell is shown in

The foregoing is always valid as long as the incident material is Fig. 3a (Vermang et al., 2014a,b,c, 2015a,b; Krc et al., 2006). There
free from losses, i.e. y0 is real (for non-absorbing medium). When are two possible kinds of optical rear interfaces, identified by (1)
y0 becomes complex, as it is the case when analysing the rear side and (2):
of a solar cell, the reflectance cannot be theoretically defined by
R ¼ r  r anymore because the incident irradiance and the reflected (1) CIGS-Mo
irradiance are now coupled together (Angus Macleod, 2001). The or CIGS-MoSe2-Mo
coupling term is called the mixed Poynting vector (Macleod, (2) CIGS-Passivation-Mo
2014) and is essentially an interference term. Ignoring this term
by using R as it is defined above could lead to Rb > 1 in some cases, The first interface type is CIGS-rear molybdenum electrode con-
which is obviously unphysical (see Fig. 3 in supplementary data). tact. More specifically, a MoSe2 layer is grown during the seleniza-
On the other hands, the reflection coefficient r is still valid. tion step to avoid Schottky barrier (Schlenker et al., 2007) and to
It is interesting here to raise this theoretical issue because soft- ensure an ohmic contact between CIGS and Mo (Cao et al., 2011).
wares like SCAPS (Khoshsirat and Yunus, 2013; Mostefaoui et al., The second interface is formed by a passivation layer between CIGS
2015) still require to provide a rear reflectance (Rb ), which cannot and Mo.
be defined in absorbing media. Fortunately, the mixed Poynting The passivation layer will mostly be alumina (Al2O3) due to its
vector will only be important with high extinction coefficient, i.e. field-effect passivation (Vermang et al., 2014c; Kotipalli et al.,
when the incident irradiance will be absorbed anyway, before 2015b, in press) and is then mainly studied in this work. It should
reaching the rear side. In fact, the mixed Poynting vector is, in be noticed here that we only focus on the relative rear reflectance
pffiffiffiffiffiffiffiffiffiffi
the best interference condition, of about 2 nk r  r (Macleod, (Rb ) of optical interfaces. Absolute reflectance could be obtained by
2014). On the other hand, it will not be important when k is small multiplying the residual irradiance at the rear interface (see
and n is high enough, as it is often the case for semiconductors in Fig. 1a) by the relative reflectance.
long wavelength, especially for CIGS in the 800–1100 nm range. In The wavelength range selected here corresponds to long wave-
this work, we thus set the extinction coefficient of CIGS kCIGS to zero length between 800 and 1100 nm. We first point out from Fig. 3b
so that R ¼ r  r is valid but tainted with a small error proportional that reflection on CIGS-Mo interface is low (around 30%, dashed
to the mixed Poynting vector. curve). In general, bare Mo is not an effective reflector, only reflect-
That being said, the validity of the results obtained by setting ing about 65% of the light while aluminium reflects more than 90%
kCIGS = 0 can be checked by looking at the transmittance instead (Fig. 3b). It becomes even worse when growing the necessary
of the reflectance. We know from the perfect absorber concept MoSe2 (Fig. 3b). This growth drastically increases matching
(Macleod, 2013) that optimising Rb means to cancel light transmis- between exit Y sMoSe2 Mo and input yCIGS
0 admittance, which acts as
sion (T) into the substrate. It implies that the exit admittance must an anti reflective coating (ARC, Fig. 4). The worst case is 22.4 nm
be purely imaginary, i.e. Re(Y) = 0, that is the locus of Y in the of MoSe2, where its locus crosses the real admittance axis.
admittance diagram has to cross the imaginary axis, or at least On the other hand, adding a dielectric layer, such as amorphous
be as close as possible. The computation of the exit admittance Al2O3 between CIGS and Mo, increases the amount of light
does not require to cancel the imaginary part of the incident reflected back into CIGS (Fig. 3b). In the admittance diagram,
446 O. Poncelet et al. / Solar Energy 146 (2017) 443–452

1
(a) Air-Al (b)
0.8

Reflectance
0.6 Air-Mo

0.4 CIGS-Al 2 O 3 -Mo

CIGS-Mo
0.2
CIGS-MoSe -Mo
2
0
800 900 1000 1100
Wavelength [nm]

Fig. 3. (a) Schematic cross-section of a CIGS cell. Aluminium-doped zinc-oxide (AZO) is ZnO:Al. (b) Relative reflectance of few optical interfaces. Thin-film thicknesses are set
to 50 nm for Al2O3 and 20 nm for MoSe2. (1) and (2) show two different interfaces in (a) and their respective reflectance in (b).

0 Reflectance spectra corresponding to worst and best cases are

presented in Fig. 5b. Rb is still limited to 50% in the best condition.
The problem here is that the MoSe2 layer thickness is experimen-
-1 tally complex to manage accurately. Moreover, thick MoSe2 layer
also induces a series resistance Rs , thus reducing the cell fill factor
(FF) and efficiency (Polizzotti et al., 2013; Lin et al., 2016).
Im(Y)

-2 Decreasing CIGS-MoSe2-Mo interface area by using point-

contact instead of full metal contact on the rear side is considered
to be one of the best options (Vermang et al., 2014a,c). This method
-3
is already used in silicon PERC-family in order to decrease contact
resistance and to reduce rear surface recombination velocity (Sb,
-4 rear SRV).
0 1 2 3 4
Re(Y)
2.4.2. CIGS-passivation-Mo interface
Fig. 4. Admittance diagram at 900 nm, in free space units (Y 0 = 1). Only part of the Passivation layers are used in silicon cells in order to decrease
loci are plotted. Arrows show direction of admittance evolution when growing thin the SRV of the minority carriers and to minimise the optical losses
films. on front and rear sides of a cell (Kotipalli et al., 2013a). In the case
of a p-type semiconductor, Al2O3 is widely used for its chemical
increasing Al2O3 thickness does not change a lot the contrast and field effect passivations due to its high negative charge density
between exit and input admittances but it drastically decreases (Agostinelli et al., 2006; Hoex et al., 2008; Terlinden et al., 2010;
the real part of Y s (Fig. 4), thus decreasing T into Mo and improving Werner et al., 2011; Kotipalli et al., 2013b) as well as a barrier
Rb . The last statement remains true as long as the layer between against diffusion (Van Delft et al., 2012).
CIGS and Mo is transparent, where R þ T ¼ 1 (R þ T þ A ¼ 1 other- Adding Al2O3 between CIGS and Mo greatly increases R b
wise, where A is the absorption within the intermediary layer). The (Fig. 6a). The best case corresponds to an Al2O3 thickness of
optimum Al2O3 thickness that maximises Rb at 900 nm is then 121 nm while the worst case happens at 268 nm, resulting in a
117.6 nm. Only part of the admittance loci are plotted in Fig. 4 to mean rear reflectance R  b of 65% and 31% respectively. As it is the
make it easier to follow. Full locus for a dielectric is a circle traced case for MoSe2, optimisation at 900 nm gives a thickness of
in clockwise direction and centred on the real axis. 117.6 nm, the solution found with admittance diagram (Fig. 4).
Although admittance diagram is a convenient way to optimise This highlights that optimising dielectric thickness with either
dielectric thickness as it yields the exact solution at one wave- the admittance diagram or by computing R  b with the extinction
length, optimising Rb on a range rather than at one wavelength coefficient of the incident medium set to zero brings the same
becomes more complex. The following optimisations are thus per- solutions. The only thing that could be altered is the absolute value
formed by computing Rb in the selected range. In order to get the of Rb .
best Rb for long wavelength range, we used arithmetic mean Reflectance spectrum corresponding to worst and best cases for
 b ðkÞ over the 800–1100 nm range as the performance criteria.
R Al2O3 are presented in Fig. 6b. The worst case gives approximately
the same R  b than the bare interface, that is CIGS-Mo interface. Add-
2.4. Rb optimisation ing Al2O3 is then always profitable, at least for the sake of the
passivation.
2.4.1. CIGS-MoSe2-Mo interface Thickness variation around the best case is not a problem and
Growing MoSe2 is detrimental for R  b below 50 nm (Fig. 5a). The
allows inaccuracy during the deposition.
worst case happens with a 25.2 nm-thick MoSe2 where R  b = 10%. In fact, thickness in a 104–139 nm range still reflects more than
Above 50 nm, R  b becomes higher than bare CIGS-Mo interface 64%. This 14%-error on the thickness is by far greater than a stan-
and reaches a maximum at 73.6 nm. The optimisation at one dard coating-system accuracy (around 10%).
unique wavelength (900 nm) gives roughly the same results. The Although Al2O3 is one of the best materials for passivation, a
damping behaviour above 85 nm is due to MoSe2 small absorption few others are of great interest from a research point of view: mag-
around 800 nm. nesium fluoride (MgF2) and silicon dioxide (SiO2) for their low
At 900 nm, the worst case is 22.4 nm of MoSe2, which is exactly characteristic admittance (or low refractive index) and titanium
the thickness found with the admittance diagram. dioxide (TiO2) for its diffusion barrier capability (Paz and Heller,
 O. Poncelet et al. / Solar Energy 146 (2017) 443–452 447

1 1
CIGS- 25.2 CIGS-Mo
nm MoSe -Mo
CIGS-MoSe 2 -Mo at 900 nm (a) 2 (b)
Rb(800-1100 nm) CIGS-73.6 nm MoSe25.2
CIGS- 2
-Mo nm MoSe 2-Mo
0.8 0.8
CIGS-73.6 nm MoSe 2-Mo

0.6 0.6
CIGS-Mo

b
b

R
R 0.4 0.4
CIGS-Mo

0.2 0.2

0 0
0 50 100 150 800 900 1000 1100
MoSe thickness [nm] Wavelength [nm]
2

Fig. 5. (a) Reflectance of CIGS-MoSe2-Mo interfaces depending on the MoSe2 thickness, at 900 nm and its average on the 800–1100 nm range. (b) Reflectance spectrum of
CIGS-MoSe2-Mo interfaces for the best/worst MoSe2 thickness. Bare CIGS-Mo spectrum is given as a reference.

1 1
at 900 nm CIGS-Mo
CIGS-268 nm Al O -Mo
CIGS-Al O -Mo
2 3
(a) CIGS-268
CIGS-121 nm nm
2 3
Al OAl-Mo
O -Mo
(b)
Rb(800-1100 nm) 2 2 3
3
0.8 data3 0.8 CIGS-121 nm Al 2O 3-Mo
data4
0.6 0.6
Rb
b
R

0.4 0.4

0.2 0.2
CIGS-Mo
CIGS-Mo
0 0
0 100 200 300 800 900 1000 1100
Al2O 3 thickness [nm] Wavelength [nm]

Fig. 6. (a) Reflectance of CIGS-Al2O3-Mo interfaces depending on the Al2O3 thickness, at 900 nm and its average on the 800–1100 nm range. (b) Reflectance spectrum of CIGS-
Al2O3-Mo interfaces for the best/worst Al2O3 thickness. Bare CIGS-Mo spectrum is given as a reference.

1997; Finley and Gillery, 1998; Ueranantasun et al., 2003). Evolu- experimentally. This layer is small enough to avoid bad optical
tion of R  b with thickness for these 3 dielectrics are shown in impact and thick enough for the field effect to be effective
Fig. 7a and their respective best spectra in Fig. 7b. (Kotipalli et al., 2013a). Adding this layer between CIGS and
Out of the four dielectrics, MgF2 gives the best R  b , about 72% for a MgF2 doesn’t change R  b (still 72%) but decreases the optimum
thickness of 144 nm. On the other hand, TiO2 is the worst material, MgF2 thickness down to 131 nm (Fig. 8a & b). This is the best
with a maximum R  b of 42%. Best case for each material are sum- trade-off regarding the optical losses, passivation and materials
marised in Table 1. The thickness analysed here corresponds to thickness.
thickness around the quarter-wave condition at 900 nm. Even with On the other hand, a thin aluminium layer between dielectrics
 b is still relatively low due to the non-appropriated optical
MgF2 ; R and molybdenum greatly enhances rear reflection. The stack
CIGS-Al2O3-Al-Mo (stack 8 in Table 1) improves R  b up to 92%.
admittance of Mo. Finishing the stack with an effective metal reflec-
tor such as aluminium is thus mandatory to reach higher R b . Adding the same aluminium layer on the previous best option
(stack 6) improves R  b up to 94%, where the stack is CIGS-Al2O3-
In the following, a 15 nm-thick layer of Al2O3 is kept on the CIGS
rear interface for passivation purpose, as it is often done MgF2-Al-Mo (stack 10) in this case (Fig. 8a & b).

1 1
(a) (a)
0.8 0.8
R (800-1100 nm)

0.6 0.6
b
R

0.4 0.4
b

MgF
MgF 2
2
0.2 SiO 2 0.2 SiO 2
CIGS-Mo CIGS-Passiv.-Mo CIGS-Passiv.-Mo Al O
Al 2O 3 2 3
CIGS-Mo TiO2
TiO2
0 0
0 50 100 150 200 250 300 800 900 1000 1100
Dielectric thickness [nm] Wavelength [nm]

Fig. 7. (a) Average rear reflectance versus dielectric (passivation) thickness, for 4 materials. The vertical lines are guidelines to indicate the optimised thicknesses. (b) Rear
reflectance of the optimised dielectric-thickness from (a), for the 4 materials. Bare CIGS-Mo spectrum is given as a reference.
448 O. Poncelet et al. / Solar Energy 146 (2017) 443–452

Table 1
Material and thickness optimisation of R b , defined as being the average Rb on 800–1100 nm wavelength range. C = CIGS, M = MgF2, T = TiO2. #Layers is the number of layers


between CIGS and Mo. Thickness(es) is/are the thickness(es) of the dielectric layers between CIGS and Mo. Re(Y s ) is the real part of the exit admittance (at 900 nm). drange is the
dielectric thickness range for which 99% of R  b (last column) can be obtained and
the maximum related error allowed (compared with the maximum one). dAl is the minimum
aluminium thickness to get 99% of the max R  b (last column), obtained with 90 nm-thick Al in the simulations.

Case N. & stack #Layers Thickness(es) [nm] Re(Ys)
dAl [nm]  b [%]
drange [nm] R

Simple stacks
1. CIGS-Mo 0 Bare interface 3.063 / / 33
2. C-TiO2-Mo 1 75 0.672 62–8817.3% / 42
3. C-Al2O3-Mo 1 121 0.340 104—13914% / 65
4. C-SiO2-Mo 1 135 0.286 115—15514:8% / 69
5. C-MgF2-Mo 1 144 0.258 122—16514:6% / 72
6. C-Al2O3-MgF2-Mo 2 15–131 0.262 109—15216% / 72

7. C-Al2O3-TiO2-Al-Mo 3 15–52 0.139 25—7952% 33 86

8. C-Al2O3-Al-Mo 2 126 0.069 88—16530:2% 28 92
80 . C-Al2O3-Cu-Mo 2 122 0.021 58—18324:6% 45Cu 97
800 . C-Al2O3-Ag-Mo 2 124 0.011 43—20363:7% 43Ag 98

9. C-Al2O3-SiO2-Al-Mo 3 15–126 0.058 82—17034:9% 26 93.5

10. C-Al2O3-MgF2-Al-Mo 3 15–135 0.053 88—18234:8% 25 94

Bragg multilayer stacks

11. C-Al2O3-MTM-Mo 4 15-M149-T99 0.096 / / 88
12. C-Al2O3-MTMTM-Mo 6 15-M154-T101 0.036 / / 94
13. C-Al2O3-MTM-Al-Mo 5 15-M150-T98 0.019 / 17 98
14. C-Al2O3-MTMTM-Al-Mo 7 15-M155-T99 0.007 / 10 99

(a) (b)

Fig. 8. (a) Average rear reflectance versus dielectric (passivation) thickness, for 4 different stacks. (b) Rear reflectance of the optimised dielectric-thickness for the 4 stacks.

Aluminium layer could be ultra thin as 99% of the two previous 2.4.3. CIGS-multilayer-Mo interface
 b are already obtained with 27 nm and 25 nm-thick Al respec-
R Although periodic stacks (Bragg mirror) are known to improve
tively, while reaching R  b of 92% and 94% for 70 nm and 63 nm. reflectance, it is not worth to add many layers to gain few % more.
We kept in simulation Al-thickness to 90 nm (thick enough to We limited our analysis at 5 layers maximum. The stacks have the
optically act as bulk metal) in each case to get the highest achiev- following structure: CIGS-Al2O3-L-(Al)-Mo, CIGS-Al2O3-LHL-(Al)-
able R b . The thickness to reach 99% of this value is then indicated in Mo and CIGS-Al2O3-LHLHL-(Al)-Mo, with or without Al, where L
Table 1. stands for Low (admittance or refractive index) and H stands for
The optimum thickness of each dielectric layer are slightly High. A higher contrast between L and H will result in higher per-
modified by adding the 15 nm-thick Al2O3 and the ultra-thin Al formances, so that the only interesting case here is L = MgF2=M and
layer. All these results are summarised in Table 1. H = TiO2 = T.
TiO2 becomes optically interesting when adding an Al layer. Thickness optimisation leads to MgF2 = 149 nm/TiO2 = 99 nm
 b = 86% at the opti-  b = 88% for LHL stack and MgF2 = 150 nm/TiO2 = 97 -
(Fig. 9a) with R
Indeed, the stack CIGS-Al2O3-TiO2-Al-Mo gives R
nm (Fig. 9b) with R  b = 98% for LHL-Al stack. We clearly see that an
mum thickness of 52 nm (Fig. 8a & b). The stack CIGS-Al2O3-SiO2-
 b = 93.5%, thus being the second best solution (Not error on the thickness is not detrimental for R  b . Optimisation for
Al-Mo gives R
traced on Fig. 8 for visibility purpose). The ultra-thin Al layer could the LHLHL-(Al) stacks gives roughly the same thickness (see
easily be added in the already existing bottom-up process Table 1) and is not shown here. All the optimised Rb spectra are
(Vermang et al., 2014a). presented in Fig. 10.
The stack 8 has also been simulated with copper (Cu) and silver Increasing the number of layers improves R  b but is much less
(Ag) instead of aluminium. The results are summarised in Table 1 effective than using an ultra-thin Al layer. In fact, the stack CIGS-
as stack 80 and 800 . These two metals are also promising due to their Al2O3-M-Al-Mo is as good as the very thick CIGS-Al2O3-MTMTM-
high reflectance in infrared (IR) region. On another hand, gold (Au) Mo (Fig. 10). Obviously, the CIGS-Al2O3-MTMTM-Al-Mo stack gives
has also demonstrated excellent reflective properties in the IR- the best performance, with R  b = 99% but is also industrially unreal-
region, but is expensive for large scale industrial production. istic (costly and low throughput).
 O. Poncelet et al. / Solar Energy 146 (2017) 443–452 449

(a) (b)
 b optimisation versus TiO2 and MgF2 thickness for (a) MTM stack and (b) MTM-Al stack. R
Fig. 9. R  b iso-contour is set at 2 nm-step in (a) and 1 nm-step in (b).

The first solution is to add an ultra-thin IR metal-reflector layer

between dielectric and Mo (cases 7–10). The best trade-off here is
the stack Al2O3-Al so that only 2 materials are used, while keeping
passivation for a relatively good R  b . Obviously, adding a metal layer
in a material stack could lead to thermal problem like delamination.
Other kinds of metals can be used for the same purpose as long as
their bare reflectivity in long wavelength is high (Ag, Cu, Au, . . .).
In this work, we focused on aluminium because it is widely used
in solar industry and available in a lot of different coating systems.
The second solution is to use a Bragg mirror, i.e. a periodical stack
of high and low refractive index materials. Although this system is
well-known to provide high reflectance, a large number of layers
is often needed. In the case of MgF2-TiO2 stack, we need at least 5
layers to become comparable to a dielectric-aluminium stack.
Fig. 10. Optimised spectra for 6 different stacks, where M = MgF2 and T = TiO2. Another disadvantage with Bragg mirror is the precise thickness it
requires to be fully useful. Even if Bragg mirror could be deposited
by atomic layer deposition (ALD) (Poncelet et al., 2015, 2016), it is
2.5. Discussion time consuming and expensive. Fortunately, R  b is in this work opti-
mised on a wavelength range rather than on a unique value.
Table 1 summarises all the relevant informations from the fore- The optimum-thickness map (Fig. 9) shows that error on both
going simulations. It provides the best R  b achievable for all the  b are quite
thicknesses are acceptable as contours of constant R
stacks studied in this work, their respective optimised thickness large, but it is difficult to set a safety-range because of the
and the number of layers they are made of. It is clear that decreasing ellipse-shape of those contours.
transmission into the substrate by decreasing Re(Ys) greatly Although it is a common material used in photovoltaic, Si3N4 is
increase Rb as there is no absorption in the stacks used. The thick- not considered here because of its high positive charge density
ness range mentioned in Table 1 is a target-range that ensures to (+Qf) which contributes to increase the surface recombination
get at least 99% of the optimised R  b . The most critical case is the
velocity (Sb) (Kotipalli et al., in press). Nevertheless, its moderate
stack 3, where the thickness error on Al2O3 is limited to 14% maxi- refractive index is still interesting from the optical point of view.
mum. This is still higher than common coating-system accuracy It should be noticed that the results presented here may also
(
10%), which ensures a robust process step. On the other hand, vary a little bit with the material properties, resulting from differ-
adding a metal thin-layer (Al, Cu or Ag) increases the target-range, ent coating techniques used. Nevertheless, the impact will be lim-
allowing the dielectric coating to vary while still keeping a high R b .
ited, as it is the case with thickness variation.
The bare CIGS-Mo interface (case 1) is bad in the optical point of
view. Adding a dielectric between them can already double R  b and,
the lower its refractive index is, the better it will be (cases 2–5). 3. Solar cell simulations
However, using MgF2 alone gives only 7% more reflectance at the
price of the passivation loss (case 5). In this section, we discuss the impact of rear reflectance on the
This passivation loss will be even higher if SiO2 is used alone solar cell parameters (Jsc and g). This has been achieved by using
(case 4), due to its slightly positive charge density. It is thus always SCAPS simulation software, where the model accommodates rear
recommended to keep at least 15 nm of Al2O3 and then use a sec- dielectric surface passivation effects (i.e. the field effect passivation
ond low-refractive index material such as SiO2 or MgF2 (cases 6– due to negative fixed charges and interface state densities) using
14). Thickness around 15 nm will still be acceptable as it is far experimentally extracted CIGS/Al2O3 interface properties as
away from any quarter-wave conditions. It will slightly modify described in Kotipalli et al. (in press).
the optimised thickness presented in Table 1.
Unfortunately, optical constraint due to molybdenum admit- 3.1. Opto-electronic simulations
tance doesn’t allow R  b to be higher than 72% so that two other
alternatives can be exposed (apart from complex texturization Jsc and g are generated for complete CIGS solar cell structure for
and photonic crystals). seven different rear reflectance stacks from Table 1 and four
450 O. Poncelet et al. / Solar Energy 146 (2017) 443–452

(a) (b)

Fig. 11. Simulated cell parameters for four different CIGS absorber thicknesses for seven different rear reflectance stacks. (a) Short circuit current density. (b) Cells efficiency.
Number 1, 3, 6, 7, 8, 10 and 14 represent the respective stack number in Table 1.

different absorber thicknesses (i.e. 250, 500, 750 and 1000 nm). In this note, for a 250 nm CIGS absorber film with improving rear
our simulation model, the rear reflectance files are modified with reflectance (from UP-stack to P-stack 3), a considerable gain in both
Rb vs wavelength profiles of the seven proposed stacks. All the sim- Jsc and cell efficiency of 2.2 mA/cm2 and 6.6% (absolute) respectively
ulations were performed under AM1.5 solar spectrum using base- were obtained. Such gain in cell efficiency also includes the contri-
line properties of respective materials provided in Ref. (Kotipalli bution of improved Voc due to reduced interface and bulk recombi-
et al., in press). However, a high quality CIGS film is chosen in nations. This even makes a very thin CIGS film (250 nm) with
our simulation model with improved deep defect concentration improved rear reflectance and passivation more effective than a
of 1014 cm3 compared to the present CIGS baseline properties conventional thick one (1 lm) without dielectric (Fig. 11b).
given in Kotipalli et al. (in press) in order to clearly distinguish Next, a further improvement in rear reflectance (from P-stacks
the optical effect trends. The front surface reflectance is main- 3–14) yields gains in Jsc and efficiency of 1.6 mA/cm2 and 0.86%
tained at 5% for all the CIGS thicknesses. The gallium to gallium respectively. Similar gain trends were observed for 500 nm absor-
plus indium ratio GGI = Ga/(Ga + In) of the CIGS absorber ber films between UP-stack 1 and P-stack 3 with Jsc and cell effi-
(Kotipalli et al., 2015a) is kept at 0.3. The CIGS absorber layers have ciencies of 2.0 mA/cm2 and 5.79% respectively. Further increase
uniform gallium profiles, in order to avoid complementary optical in R  b from P-stack 3–14 yields gains in Jsc and efficiency of
gains (i.e. enhanced band gap). More detailed device mechanisms 0.6 mA/cm2 and 0.33% respectively. Thus overall, we observe a
related to the effects of negative fixed charges, interface trap den- decreasing trend in the gains while moving towards thicker CIGS
sities, rear reflectance and their combinations can be found in absorber films even with increasing R  b (from UP-stack 1 to P-
Kotipalli et al. (in press). stack 3 and from P-stack 3 to P-stack 14).
Next, for the sake of clarity only gains related to Jsc and g are Based on these simulation results and analysis, we confirm that
shown and discussed in order to quantify the impact of Rb only, a CIGS thickness of 250 nm is not sufficient to achieve >20% effi-
and not the impact due to the rear surface recombination (i.e. ciencies even when implementing excellent Rb (stack 7, 98%). The
improvement in the open-circuit voltage Voc and FF). In the follow- minimum thickness of CIGS target thickness is found to be around
ing, we will refer to the reference case, i.e. CIGS-Mo, by unpassi- 500 nm with moderate to excellent rear reflectance optimisation
vated stack (UP-stack, case 1) and the six others by passivated (P-stacks 3–14: 65–98%).
stack (P-stack, cases 3, 6, 7, 8, 10, and 14 in Table 1). As a trade-off between the number of stacking materials (i.e.
process feasibility), gains in Jsc and efficiency, we suggest targeting
3.2. Discussion rear reflectance between 65% and 92% to be sufficient for achieving
efficiencies >20%. On the other hand, keeping rear surface passiva-
In Fig. 11, UP-stack is considered to be the reference case with tion and industrial feasibility in considerations, targeting Al2O3
no Rb optimisation (i.e. R b = 33%), compared to other optimised P- thickness in a range between 104 nm and 139 nm would ensure
 b ranging between 65% and 98%. More generally, all Rb ¼ 65
0:65% and g > 20%.
stack with R
the stacks show an increasing trend in Jsc and cell efficiency with
increasing absorber layer thickness. Such gain in Jsc is mainly due 4. Conclusion
to the improved optical absorption with increasing absorber thick-
ness. The overall trends in both Jsc and efficiencies for thicker In this work, we investigated the optical properties of a rear
absorbers (750 nm and 1000 nm) with rear reflectance beyond CIGS cell interface using rigorously transfer matrix method
65% can be considered to reach a saturation plateau. (TMM) in order to optimise Rb , the rear reflectance. We highlighted
However, for thinner absorber films (250–500 nm), significant the weak reflectance of the standard CIGS-Mo interface and we
gains in the Jsc (2.9–4.1 mA/cm2) are observed with increasing R b proposed to use standard thin film stacks to optimise the reflec-
tance, where our performance criteria was R  b , the average Rb , in
(i.e. from 65% to 98%). Lesser gains in Jsc (0.6–1.2 mA/cm2) and
(0.2–0.6 mA/cm2) were observed for further increase in the CIGS the range 800–1100 nm.
thickness, i.e. from 500 nm to 750 nm, and 750 nm to 1000 nm In order to keep the rear CIGS surface recombination velocity
respectively. It means that a CIGS thickness of 500 nm appears suf- (SRV) as low as possible, a fixed 15 nm-thick Al2O3 was placed
ficient while considering the rear reflectance optimisation concepts. between the CIGS and the dielectric-Mo stack (MgF2, SiO2, Al2O3
For a constant CIGS absorber thickness, the gain in Jsc between  b was 72% by using the stack CIGS-
or TiO2). The best achievable R
UP-stack and P-stack 3 is due to both reduced rear surface recombi- Al2O3 (15 nm)-MgF2 (131 nm)-Mo.
nation (Sb) and improved rear reflectance (Rb ), whereas the gain We then proposed to use either an ultra-thin aluminium layer
obtained between P-stacks 3–14 is solely due to improved Rb . On or a multilayer film, bringing R  b from 86% with CIGS-Al2O3
 O. Poncelet et al. / Solar Energy 146 (2017) 443–452 451

(15 nm)-TiO2 (52 nm)-Al (33 nm)-Mo up to 99% with a 5-layers solar cell: optical modeling and characterization. In: SPIE OPTO. International
TiO2/MgF2 Bragg stack. The best trade-off between high R  b , passi- Society for Optics and Photonics, pp. 862019–862019.
Finley, James J., Gillery, F. Howard, 1998. Alkali Metal Diffusion Barrier Layer,
vation and the number of layers was the stack CIGS-Al2O3 November 3 1998. US Patent 5,830,252.
(126 nm)-Al (28 nm)-Mo, giving a R  b of 92%. We have also shown Han, Sung-Ho, Hermann, Allen M., Hasoon, F.S., Al-Thani, H.A., Levi, D.H., 2004.
Effect of cu deficiency on the optical properties and electronic structure of
that other good IR metal-reflector such as copper (Cu) or silver CuInSe2 and CuIn0. 8Ga0. 2Se2 determined by spectroscopic ellipsometry. Appl.
(Ag) provide high rear reflectance. Phys. Lett. 85, 576.
The impact of R  b on cell performance parameters (Jsc and effi- Hass, Georg, Gerstenberg, D., 1964. Physics of Thin Films. The Electrochemical
Society, pp. 87–90 (Chapter 2 - Theory and Calculations of Optical Thin Films).
ciency) in function of the CIGS thickness were then performed with Hoex, B., Gielis, J.J.H., Van de Sanden, M.C.M., Kessels, W.M.M., 2008. On the c-si
SCAPS. We have shown that R  b plays an important role in very thin surface passivation mechanism by the negative-charge-dielectric Al2O3. J. Appl.
Phys. 104 (11), 113703.
CIGS-thicknesses (thinner than 750 nm) and becomes less detri-
Jackson, Philip, Hariskos, Dimitrios, Lotter, Erwin, Paetel, Stefan, Wuerz, Roland,
mental for thicker CIGS. We have also shown that a R  b of 65%, Menner, Richard, Wischmann, Wiltraud, Powalla, Michael, 2011. New world
standing for a single Al2O3 layer of around 121 nm, is already record efficiency for Cu (In, Ga) Se2 thin-film solar cells beyond 20%. Prog.
Photovolt.: Res. Appl. 19 (7), 894–897.
enough to achieve high cell performances, i.e. g > 20% and Jsc > 40 - Ji, Liming, Nam, Wook Jun, Fonash, Stephen, 2013. Highly ordered nano-cone back
mA/cm2. Finally, we desired through this work to provide to any reflector arrays for ultra-thin high performance cigs cells. In: Photovoltaic
researcher the right dielectric thickness ranges to use on a rear Specialists Conference (PVSC), 2013 IEEE 39th. IEEE, pp. 1977–1979.
Katsidis, Charalambos C., Siapkas, Dimitrios I., 2002. General transfer-matrix
CIGS interface in order to get the best performances. Our next step
method for optical multilayer systems with coherent, partially coherent, and
will be to experimentally implement those stacks within our own incoherent interference. Appl. Opt. 41 (19), 3978–3987.
CIGS fabrication process steps. Khoshsirat, N., Yunus, N.A.Md., 2013. Numerical simulation of CIGS thin film solar
cells using SCAPS-1D. In: 2013 IEEE Conference on Sustainable Utilization and
Development in Engineering and Technology (CSUDET). IEEE, pp. 63–67.
Acknowledgments Kotipalli, Ratan, Delamare, Romain, Poncelet, Olivier, Tang, Xiaohui, Francis, Laurent
A., Flandre, Denis, 2013a. Passivation effects of atomic-layer-deposited
aluminum oxide. EPJ Photovolt. 4, 45107.
This work is supported by the Actions de Recherches Concertées Kotipalli, Raja Venkata Ratan, Delamare, Romain, Henry, Frédéric, Proost, Joris,
of the Académie Universitaire Louvain. The project has received Flandre, Denis, et al., 2013b. Thermal stability analysis of DC-sputtered Al2O3
funding from the European Union’s Horizon 2020 research and films for surface passivation of c-Si solar cells. In: 28th European Photovoltaic
Solar Energy Conference and Exhibition (EU PVSEC 2013).
innovation program under grant agreement No. 720887, and from Kotipalli, Ratan, Vermang, Bart, Fjällström, Viktor, Edoff, Marika, Delamare, Romain,
the European Research Council (ERC) under the European Union’s Flandre, Denis, 2015a. Influence of Ga/(Ga+ In) grading on deep-defect states of
Horizon 2020 research and innovation programme (grant agree- Cu (In, Ga) Se2 solar cells. Phys. Status Solidi (RRL)-Rapid Res. Lett. 9 (3), 157–160.
Kotipalli, R., Vermang, Bart, Joel, J., Rajkumar, R., Edoff, Marika, Flandre, Denis,
ment No. 715027). Additionally, R. Kotipalli acknowledges the 2015b. Investigating the electronic properties of Al2O3/Cu (In, Ga) Se2 interface.
financial support of the Belgian National Fund for Scientific AIP Adv. 5 (10), 107101.
Research (F.R.S.-FNRS). Finally, B. Vermang acknowledges the Kotipalli, R., Poncelet, O., Guoli, L., Francis, L.A., Vermang, B., Flandre, D., 2016.
Addressing the impact of rear surface passivation mechanisms on ultra-thin Cu
financial support of the Flemish Research Foundation FWO (man-
(In,Ga)Se2 solar cell performances using SCAPS 1-D model (in press).
date 12O4215N). Krc, J., Cernivec, G., Campa, A., Malmström, Jonas, Edoff, Marika, Smole, F., Topic, M.,
2006. Optical and electrical modeling of Cu (in, Ga) Se2 solar cells. Opt. Quant.
Electron. 38 (12), 1115–1123.
Appendix A. Supplementary material Krc, J., Sever, M., Campa, A., Lokar, Z., Lipovsek, B., Topic, M., 2016. Optical
confinement in chalcopyrite based solar cells. Thin Solid Films.
Lare, Claire van, Yin, Guanchao, Polman, Albert, Schmid, Martina, 2015. Light
Supplementary data associated with this article can be found, in coupling and trapping in ultrathin Cu (In, Ga) Se2 solar cells using dielectric
the online version, at http://dx.doi.org/10.1016/j.solener.2017.03. scattering patterns. ACS Nano 9 (10), 9603–9613.
001. Larsen, Jes K., Simchi, H., Xin, P., Kim, K., Shafarman, W.N., 2014. Backwall
superstrate configuration for ultrathin Cu (In, Ga) Se2 solar cells. Appl. Phys.
Lett. 104 (3), 033901.
References Li-Kao, Zacharie Jehl, Naghavi, Negar, Erfurth, Felix, Guillemoles, Jean François,
Gérard, Isabelle, Etcheberry, Arnaud, Pelouard, Jean Luc, Collin, Stephane,
Voorwinden, Georg, Lincot, Daniel, 2012. Towards ultrathin copper indium
Agostinelli, G., Delabie, A., Vitanov, P., Alexieva, Z., Dekkers, H.F.W., De Wolf, S.,
gallium diselenide solar cells: proof of concept study by chemical etching and
Beaucarne, G., 2006. Very low surface recombination velocities on p-type silicon
gold back contact engineering. Prog. Photovolt.: Res. Appl. 20 (5), 582–587.
wafers passivated with a dielectric with fixed negative charge. Sol. Energy
Lin, Yi-Cheng, Hsieh, Yin-Ting, Lai, Chien-Mu, Hsu, Hung-Ru, 2016. Impact of mo
Mater. Sol. Cells 90 (18), 3438–3443.
barrier layer on the formation of MoSe2 in Cu (in, Ga) Se2 solar cells. J. Alloys
Andreani, L., Kowalczewski, P., Mura, C., Patrini, Maddalena, Acciarri, Maurizio,
Compd. 661, 168–175.
Binetti, Simona, Sassella, Adele, Marchionna, Stefano, 2012. Towards CIGS solar
Macleod, Angus, 2013. Optical absorption. SVC Bull. Fall, 28–33.
cells with reduced film thickness: a study of optical properties and of photonic
Macleod, Angus, 2014. The mixed poynting vector. SVC Bull. Summer, 30–34.
structures for light trapping. In: Proceedings of the 27th European Photovoltaic
Mahieu, Stijn, Leroy, W.P., Aeken, Koen Van, Wolter, M., Colaux, J., Lucas, S., Abadias,
Solar Energy Conference and Exhibition, Frankfurt, Germany, pp. 24–28.
G., Matthys, Paul, Depla, Diederik, 2011. Sputter deposited transition metal
Angus Macleod, H., 2001. Thin-film Optical Filters. CRC Press, pp. 12–50 (Chapter 2 -
nitrides as back electrode for cigs solar cells. Solar Energy 85 (3), 538–544.
Basic Theory).
Mitsas, C.L., Siapkas, D.I., 1995. Generalized matrix method for analysis of coherent
Bednar, Nikola, Severino, Noemi, Adamovic, Nadja, 2015. Optical simulation of light
and incoherent reflectance and transmittance of multilayer structures with
management in CIGS thin-film solar cells using finite element method. Appl. Sci.
rough surfaces, interfaces, and finite substrates. Appl. Opt. 34 (10), 1678–1683.
5 (4), 1735–1744.
Mostefaoui, M., Mazari, H., Khelifi, S., Bouraiou, A., Dabou, R., 2015. Simulation of
Bernal-Correa, R., Morales-Acevedo, A., Montes-Monsalve, J., Pulzara-Mora, A., 2016.
high efficiency CIGS solar cells with SCAPS-1D software. Energy Procedia 74,
Design of the TCO (ZnO: Al) thickness for glass/TCO/CdS/CIGS/Mo solar cells. J.
736–744.
Phys. D: Appl. Phys. 49 (12), 125601.
Onwudinanti, Chidozie, Vismara, Robin, Isabella, Olindo, Grenet, Louis, Emieux,
Cao, Qing, Gunawan, Oki, Copel, Matthew, Reuter, Kathleen B., Jay Chey, S., Deline,
Fabrice, Zeman, Miro, 2016. Advanced light management based on periodic
Vaughn R., Mitzi, David B., 2011. Defects in Cu (In, Ga)Se2 chalcopyrite
textures for Cu (In, Ga) Se2 thin-film solar cells. Opt. Express 24 (6), A693–A707.
semiconductors: a comparative study of material properties, defect states, and
Orgassa, Kay, Schock, Hans Werner, Werner, J.H., 2003a. Alternative back contact
photovoltaic performance. Adv. Energy Mater. 1 (5), 845–853.
materials for thin film Cu (In, Ga) Se2 solar cells. Thin Solid Films 431, 387–391.
Chen, Chang-Wen, Hsiao, Sheng-Yi, Chen, Chien-Yu, Kang, Hao-Wei, Huang, Zheng-
Orgassa, Kay, Rau, U., Schock, H.-W., Werner, J.H., 2003b. Optical constants of Cu (ln,
Yu, Lin, Hao-Wu, 2015. Optical properties of organometal halide perovskite thin
Ga) Se/sub 2/thin films from normal incidence transmittance and reflectance.
films and general device structure design rules for perovskite single and tandem
Proceedings of 3rd World Conference on Photovoltaic Energy Conversion, vol. 1.
solar cells. J. Mater. Chem. A 3 (17), 9152–9159.
IEEE, pp. 372–375.
Dahan, Nir, Jehl, Zacharie, Hildebrandt, Thomas, Greffet, J.-J., Guillemoles, J.-F.,
Paz, Yrn, Heller, A., 1997. Photo-oxidatively self-cleaning transparent titanium
Lincot, Daniel, Naghavi, N., 2012. Optical approaches to improve the
dioxide films on soda lime glass: the deleterious effect of sodium contamination
photocurrent generation in Cu (In, Ga) Se2 solar cells with absorber
and its prevention. J. Mater. Res. 12 (10), 2759–2766.
thicknesses down to 0.5 l m. J. Appl. Phys. 112 (9), 094902.
Pedrotti, F.L., Pedrotti, L.S., 1993. Introduction to Optics. Prentice-Hall International,
Dahan, Nir, Jehl, Zacharie, Guillemoles, Jean-François, Lincot, Daniel, Naghavi,
Inc., pp. 391–396 (Chapter 19 - Theory of Multilayer Films).
Negar, Greffet, Jean-Jacques, 2013. Lambertian back reflector in Cu (inGa) Se2
452 O. Poncelet et al. / Solar Energy 146 (2017) 443–452

Polizzotti, Alex, Repins, Ingrid L., Noufi, Rommel, Wei, Su-Huai, Mitzi, David B., Ueranantasun, Attachai, Richards, Bryce S, Honsberg, Christiana B, Cotter, Jeffrey E.,
2013. The state and future prospects of kesterite photovoltaics. Energy Environ. 2003. Titanium dioxide film as a phosphorus diffusion barrier in silicon solar
Sci. 6 (11), 3171–3182. cells. Proceedings of 3rd World Conference on Photovoltaic Energy Conversion,
Poncelet, Olivier, Tallier, Guillaume, Simonis, Priscilla, Cornet, Alain, Francis, vol. 2, pp. 1411–1414.
Laurent A., 2015. Synthesis of bio-inspired multilayer polarizers and their Van Delft, J.A., Garcia-Alonso, D., Kessels, W.M.M., 2012. Atomic layer deposition for
application to anti-counterfeiting. Bioinspirat. Biomimet. 10 (2), 026004. photovoltaics: applications and prospects for solar cell manufacturing.
Poncelet, Olivier, Tallier, Guillaume, Mouchet, Sébastien R., Crahay, André, Rasson, Semicond. Sci. Technol. 27 (7), 074002.
Jonathan, Kotipalli, Ratan, Deparis, Olivier, Francis, Laurent A., 2016. Vapour Vermang, B., Watjen, J.T., Frisk, C., Fjallstrom, V., Rostvall, F., Edoff, M., Salome, P.,
sensitivity of an ALD hierarchical photonic structure inspired by Morpho. Borme, J., Nicoara, N., Sadewasser, S., 2014a. Introduction of Si PERC rear
Bioinspirat. Biomimet. 11 (3), 036011. contacting design to boost efficiency of Cu (In, Ga) Se solar cells. IEEE J.
Ramanathan, Kannan, Contreras, Miguel A., Perkins, Craig L., Asher, Sally, Hasoon, Photovolt. 4 (6), 1644–1649.
Falah S., Keane, James, Young, David, Romero, Manuel, Metzger, Wyatt, Noufi, Vermang, Bart, Wätjen, Jörn Timo, Fjällström, Viktor, Rostvall, Fredrik, Edoff,
Rommel, 2003. Properties of 19.2% efficiency ZnO/CdS/CuInGaSe2 thin-film Marika, Kotipalli, Ratan, Henry, Frederic, Flandre, Denis, 2014b. Employing Si
solar cells. Prog. Photovolt.: Res. Appl. 11 (4), 225–230. solar cell technology to increase efficiency of ultra-thin Cu (In, Ga) Se2 solar
Rubio-Bollinger, Gabino, Guerrero, Ruben, Lara, David Perez de, Quereda, Jorge, cells. Prog. Photovolt.: Res. Appl. 22 (10), 1023–1029.
Vaquero-Garzon, Luis, Agraït, Nicolas, Bratschitsch, Rudolf, Castellanos-Gomez, Vermang, B., Fjallstrom, V., Gao, X., Edoff, M., 2014c. Improved rear surface
Andres, 2015. Enhanced visibility of MoS2, MoSe2, WSe2 and black- passivation of Cu (In, Ga) se solar cells: a combination of an AlO rear surface
phosphorus: making optical identification of 2D semiconductors easier. passivation layer and nanosized local rear point contacts. IEEE J. Photovolt. 4
Electronics 4 (4), 847–856. (1), 486–492.
Schlenker, E., Mertens, V., Parisi, J., Reineke-Koch, R., Köntges, M., 2007. Schottky Vermang, B., Watjen, J.T., Fjallstrom, V., Rostvall, F., Edoff, M., Gunnarsson, R., Pilch,
contact analysis of photovoltaic chalcopyrite thin film absorbers. Phys. Lett. A I., Helmersson, U., Kotipalli, R., Henry, F., 2015a. Highly reflective rear surface
362 (2), 229–233. passivation design for ultra-thin Cu (In, Ga) Se2 solar cells. Thin Solid Films 582,
Schleussner, Sebastian, Kubart, Tomas, Törndahl, Tobias, Edoff, Marika, 2009. 300–303.
Reactively sputtered ZrN for application as reflecting back contact in Cu (In, Ga) Vismara, Robin, Onwudinanti, Chidozie, Grenet, Louis, Emieux, Fabrice, Isabella,
Se2 solar cells. Thin Solid Films 517 (18), 5548–5552. Olindo, Zeman, Miro, 2015b. Optical analysis of cigs solar cells on periodic
Simchi, Hamed, Larsen, Jes K., Kim, Kihwan, Shafarman, William N., 2014. Improved gratings. In: Optical Nanostructures and Advanced Materials for Photovoltaics.
performance of ultrathin Cu (InGa) Se2 solar cells with a backwall superstrate Optical Society of America, pp. PTu4B-5.
configuration. IEEE J. Photovolt. 4 (6), 1630–1635. Werner, Florian, Veith, Boris, Zielke, Dimitri, Kühnemund, Lisa, Tegenkamp,
Terlinden, N.M., Dingemans, G., Van de Sanden, M.C.M., Kessels, W.M.M., 2010. Role Christoph, Seibt, Michael, Brendel, Rolf, Schmidt, Jan, 2011. Electronic and
of field-effect on c-Si surface passivation by ultrathin (2–20 nm) atomic layer chemical properties of the c-Si/Al2O3 interface. J. Appl. Phys. 109 (11), 113701.
depositedAl2O3. Appl. Phys. Lett. 96, 112101. Yin, Guanchao, Song, Min, Duan, Shengkai, Manley, Phillip, Greiner, Dieter, Kaufmann,
Tikhonravov, Alexander V., Trubetskov, Michael K., Tikhonravov, Andrei A., Duparré, Christian A., Schmid, Martina, 2016. Well-controlled dielectric nanomeshes by
Angela, 2003. Effects of interface roughness on the spectral properties of thin colloidal nanosphere lithography for optoelectronic enhancement of ultrathin
films and multilayers. Appl. Opt. 42 (25), 5140–5148. Cu (In, Ga) Se2 solar cells. ACS Appl. Mater. Interf. 8 (46), 31646–31652.