APPLIED PHYSICS LETTERS 102, 151111 (2013)

Enhanced light absorption in thin film solar cells with embedded dielectric
nanoparticles: Induced texture dominates Mie scattering
James R. Nagel1,a) and Michael A. Scarpulla1,2,b)
1
Department of Electrical and Computer Engineering, University of Utah, Salt Lake City, Utah 84112, USA
2
Department of Materials Science and Engineering, University of Utah, Salt Lake City, Utah 84112, USA
(Received 10 February 2013; accepted 29 March 2013; published online 19 April 2013)
We elucidate the dominant physical mechanisms for enhanced light absorption in thin film solar
cells with embedded dielectric nanoparticles and conformal surface texturing. A prior work
reported enhancement in short-circuit current for such a structure and attributed this gain to
increased photon absorption due to Mie scattering by the embedded nanoparticles [S. Nunomura,
A. Minowa, H. Sai, and M. Kondo, Appl. Phys. Lett. 97(6), 063507 (2010)]. On the contrary,
simulations indicate that absorption gain is categorically dominated by the roughened surface
texture that resulted from conformal overgrowth of amorphous silicon above the deposited
C 2013 AIP Publishing LLC [http://dx.doi.org/10.1063/1.4802718]
dielectric nanoparticles. V

With the advent of optically thin silicon films for photo-
voltaic power generation, efficient light-trapping has become
an important area of current research. In principle, the goal
is to scatter sunlight within the solar cell after it has entered
through the top surface, thereby increasing the overall
path length of propagation within the film and absorbing
more energy. Some examples of experimental embodiments
include random surface roughening,1 photonic crystals,2 and
plasmonic nanoparticles.3 In this work, we specifically focus
on the use of light-scattering nanoparticles deposited directly
within the film region of the device. Similar to surface
roughening at the back contact, this scheme has the advant-
age of working cooperatively with antireflective coatings
(ARCs), thereby allowing for independent maximization of
light injection in concert with light trapping.4,5
Nunomura et al., in a combined experimental and theo-
retical paper, investigated the effects of embedded dielectric
nanoparticles in a tandem solar cell structure consisting of
both crystalline silicon (c-Si) and hydrogenated amorphous
silicon (a-Si:H) layers.6 It was reported that embedded nano-
spheres of silicon dioxide (SiO2) placed between the tandem
layers resulted in a 16% improvement in short-circuit current
density Jsc, which was then attributed entirely to Mie scatter-
ing by the particles. However, it was also reported that
growth of the a-Si:H top cell above the dispersed SiO2 nano-
particles was accomplished by conformal chemical vapor
deposition. Inclusion of the embedded nanoparticles was
therefore accompanied by a randomly roughened surface tex-
ture with wavelength-scale feature sizes. Since these features
were not present in the reference device, it is unclear what
portions of the observed efficiency gain were due to scattering
by the embedded nanoparticles as opposed to scattering and
antireflective effects by the roughened surface. We therefore
seek to answer this question for the reported implementation
and then generalize the results as far as possible.
In our earlier work, we extensively studied the concept of
dielectric nanoparticles embedded in thin Si films with planar
a)
Electronic mail: nageljr@ieee.org
b)
Electronic mail: scarpulla@eng.utah.edu
interfaces.4,5 We have also solved for exact field profiles
found in guided mode propagation through lossy slab wave-
guides and showed how they differ from typical low-loss
approximations.7 In this work, we use finite-difference time-
domain (FDTD) simulations to probe the broadband response
of light-trapping structures incorporating wavelength-scale
scatterers and front surface texturing. Data were collected by
modeling a thin-film photovoltaic cell similar to the device
reported in Ref. 6 and calculating the absorbed photon flux
due to excitation by the AM 1.5 solar spectrum. Comparisons
were then made against a few simple implementations of sur-
face texture, both with and without the embedded dielectric
nanospheres. Numerical experiments were then repeated on
two variations of the same principle design as a test for gener-
ality. In all cases, the gain in absorbed photon flux from the
surface textures greatly outweighed those from the embedded
nanoparticles.
The baseline simulation model is depicted in Figure 1(a)
and consists of a planar, layered structure with no surface
features or light scatterers. This model represents an ideal-
ized (no roughness) tandem solar cell design similar to that
employed by Nunomura. The top layer was a 75 nm film of
indium tin oxide (ITO) which serves as both a front contact
and ARC. Directly underneath the ARC was a 350 nm layer
of a-Si:H followed by a 1.0 lm c-Si layer. At the bottom was
a thin, 35 nm zinc-oxide (ZnO) layer followed by a silver
(Ag) back contact. The top boundary of the simulation vol-
ume was defined by a perfectly matched layer (PML) placed
400 nm above the device surface while the bottom boundary
was a perfect electrical conductor (PEC) placed 200 nm
below the ZnO/Ag interface (well beyond the skin depth
over the band of interest). Periodic boundary conditions
along the x and y boundaries were used to mimic the effect
of an infinite square lattice every 1.0 lm, which is the ap-
proximate average spacing between nanoparticles as given in
Ref. 6. Optical properties for each material can then be found
in appropriate references throughout the literature.8–10 The
differences in refractive index between doped and undoped
semiconductors are generally negligible over the visible
spectrum.

0003-6951/2013/102(15)/151111/4/$30.00 102, 151111-1 C 2013 AIP Publishing LLC

V

Downloaded 19 Apr 2013 to 155.98.26.154. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions
151111-2 J. R. Nagel and M. A. Scarpulla
FIG. 1. Simulated solar cell geometry. (a) Baseline geometry for referencing
absorption gain. (b) Model including embedded dielectric nanosphere and
conformal surface texture. The inner radius R1 ¼ 480 nm while the outer ra-
dius R2 ¼ 555 nm. The sphere is made of SiO2 and has a radius of 130 nm.
The periodic unit cell repeats every 1.0 lm.
The embedded spheres of SiO2 and accompanying sur-
face texture are illustrated by the cross-sectional model in
Figure 1(b). The mean radius of the embedded spheres was
r ¼ 130 nm with a mean separation distance of 1.0 lm. The
spheres were allegedly deposited through spin-coating along
the surface of the c-Si layer, after which the a-Si:H layer was
deposited by chemical vapor deposition. Analysis with an
atomic force microscope also revealed a peak-to-peak eleva-
tion of the surface texture that was approximately 200 nm,
which we modeled as a hemispherical surface texture centered
on the SiO2 nanoparticles. The radius R1 ¼ 480 nm denotes
the bulge in the a-Si:H layer due to conformal deposition
around the nanoparticle while R2 ¼ 555 nm includes the outer
surface of the ITO layer. The system was excited in our simula-
tion by a plane wave pulse at normal incidence and covered the
entire visible spectrum from 300 to 1200 nm in wavelength.
The output of interest from our simulations is the
absorption factor AðkÞ, defined as the fraction of incident
power absorbed within the c-Si and a-Si:H layers at a speci-
fied wavelength k. For an ideal photovoltaic cell, AðkÞ would
be unity across the entire solar spectrum, but the low absorp-
tion coefficient of silicon prevents this from occurring in thin
films. This is especially true at longer wavelengths where the
absorption coefficient of c-Si is very small.
FIG. 2. Absorption spectra for the a-Si:H/c-Ci tandem structure, including
embedded spheres, the surface texture, and both at once.
Downloaded 19 Apr 2013 to 155.98.26.154. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions
Appl. Phys. Lett. 102, 151111 (2013)
TABLE I. Absorption gain (%) relative to baseline geometries.
a-Si:H/c-Si tandem 3-Layer spherical 3-Layer pyramidal
Sphere 2.0 5.2 5.2
Texture 14.9 47.4 38.4
Both 15.1 52.3 46.5
Once the absorption factor is determined for a given
model, we may then calculate the solar flux absorbance SðkÞ,
defined as the absorption factor weighted by the AM 1.5
photon flux UðkÞ as obtained from the NREL database.11
Writing this out explicitly gives
SðkÞ ¼ AðkÞUðkÞ : (1)
Note that the units of SðkÞ are given by photons
s1 m2 nm1 . Finally, the short-circuit current density Jsc of
the device is calculated by integrating SðkÞ with the internal
quantum efficiency IQEðkÞ, written as
ð
1
Jsc ¼ q SðkÞ IQEðkÞ dk ; (2)
0
where q is the electron charge. In principle, if the IQE were a
known function, any simulation output could be used to com-
pute Jsc under ideal AM 1.5 illumination. Because our inves-
tigation is concerned only with optical effects of embedded
particles and surface texturing, we simply postulate that IQE
is unity across the majority of the solar spectrum and then
abruptly decays to zero near the bandgap wavelength of c-Si
(kg ¼ 1130 nm). In other words, we assume any gain in light
absorption within the semiconductor layers of the solar cell
manifests as a proportional increase in Jsc.
Figure 2 shows the resultant absorption spectra for each
simulation model. The “Baseline” curve indicates the planar
structure as depicted in Figure 1(a) and has no embedded
scatterers or surface texturing. The “Spheres” curve denotes
the addition of embedded nanospheres, but no surface tex-
ture, while the “Texture” curve represents a surface texture
without any embedded spheres. Finally, the “Both” curve
denotes both the embedded nanoparticles and conformal
FIG. 3. Simulation geometries for the simplified 3-layer models. (a)
Baseline structure consisting of a 3-layer stack with a 1:0 lm active layer
of c-Si, a Si3N4 ARC, and PEC back contact. (b) Embedded nanosphere
and conformal surface texture with R1 ¼ 400 nm and R2 ¼ 475 nm. (c)
Embedded nanosphere and pyramidal surface texture with h ¼ 200 nm.
Depth and diameter for the SiO2 sphere are the same in both cases with
z0 ¼ D ¼ 200 nm.
151111-3 J. R. Nagel and M. A. Scarpulla
surface texture acting together as shown in Figure 1(b).
Clearly, the presence of the texture and particles is having an
effect on the spectra, which can be quantified through nu-
merical integration of Eq. (2). We may therefore define the
absorption gain as the percent change in Jsc relative to the
baseline geometry.
Table I summarizes the calculated absorption gains for
this numerical experiment, labeled by the “a-Si:H/c-Si
Tandem” column to indicate the tandem structure in Fig. 1.
We found that the addition of embedded dielectric nano-
spheres into a flat surface only increased Jsc by 2.0%. In con-
trast, the inclusion of a conformal surface texture pushed this
value up toward 15%. This was true whether or not the
spheres themselves were actually included. Such results are
consistent with the gain of 16% reported by Nunomura.
These predictions are also consistent with numerical experi-
ments in our own prior simulations,5 where SiO2 nanospheres
were embedded into a flat substrate of c-Si. Using a periodic-
ity of 1.0 lm between the embedded spheres, we found that
absorption gain was only 5% under a perfectly flat surface
and that greater gain could only be obtained by bringing the
spheres much closer together. These results strongly indicate
that the conformal surface texture was mostly responsible for
absorption gain observed in Ref. 6, and not the direct Mie
scattering of the embedded SiO2 nanospheres.
One possible explanation for these results is that the
additional volume of absorbing material within the surface
texture itself provides an added path length for light propaga-
tion through a lossy medium. However, simple calculations
clearly show that this cannot be the case. Even if the a-Si:H
layer is increased to a thickness of 480 nm (the peak height of
the conformal bump), the absorption gain relative to the base-
line is only 7.4% for a flattened structure. It therefore appears
the most likely explanation is that the scattering properties of
the roughened texture are dominating the enhanced light
absorption.
If this explanation is indeed accurate, then we should
expect to produce similar results under variations of the same
theme. We therefore considered the geometries depicted in
Figure 3 for numerical simulation. Figure 3(a) illustrates the
baseline model, which is now changed to a simplified, 3-layer
structure intended to capture only the critical elements of a
thin c-Si solar cell. The device consisted of 1.0 lm of c-Si
with a perfect electrical conductor at the back contact and
Downloaded 19 Apr 2013 to 155.98.26.154. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions
Appl. Phys. Lett. 102, 151111 (2013)
FIG. 4. Absorption spectra for the (a)
conformal surface texture and (b) py-
ramidal surface texture.
75 nm layer of lossless Si3N4 serving as the ARC. The em-
bedded spheres were given a diameter of D ¼ 200 nm and a
depth of z0 ¼ 200 nm below the planar surface, which is sim-
ilar to other generic models used in various numerical stud-
ies.5,12,13 The lateral periodicity was finally kept to the same
1.0 lm value as before. Two variations on surface texture
were then tested, with the first being a conformal spherical
coating as in the previous models. Depicted in Figure 3(b) we
have R1 ¼ 400 nm and R2 ¼ 475 nm. The second texture is
a pyramidal geometry depicted in Figure 3(c) with a height of
h ¼ 200 nm.
Spectral results for the two simulations are shown in
Figures 4(a) and 4(b), while the percent absorption gains rel-
ative to the baseline geometry are summarized by the corre-
sponding columns in Table I. Once again, we find that
surface texturing dominates the absorption gain, and even
more so than the tandem structure of the previous section.
The reason for this is likely due to the removal of the a-Si:H
layer, which has a much stronger absorption than c-Si and
therefore benefits less from light trapping. We also see that
the absorption gain is relatively insensitive to the exact sur-
face geometry, since both the conformal geometry and py-
ramidal geometry produced absorption gain well beyond the
embedded spheres alone. These results are consistent with
the work of Han and Chen,14 which showed that how a well-
engineered surface texture can potentially meet (and even
exceed) the Lambertian limit for light trapping. Likewise, it
has been shown that embedded nanospheres of SiO2 without
any surface roughening at all can barely induce 25% absorp-
tion gain at most.5
The numerical results in this paper strongly indicate that
surface texturing is the dominant source of light trapping
when compared to Mie scattering by embedded spheres of
SiO2. The gain in Jsc reported by Nunomura was therefore
most likely the result of the conformal texture along the sur-
face rather than Mie scattering from the embedded spheres
themselves. Embedded spheres of SiO2 therefore appear bet-
ter suited as a method for generating roughened surface tex-
ture rather than any direct source of absorption gain via light
scattering.
1
S. Koynov, M. S. Brandt, and M. Stutzmann, Appl. Phys. Lett. 88, 203107
(2006).
2
S. B. Mallick, M. Agrawal, and P. Peumans, Opt. Express 18, 5691
(2010).
151111-4 J. R. Nagel and M. A. Scarpulla Appl. Phys. Lett. 102, 151111 (2013)

3 9
C. H€agglund, M. Z€ach, G. Petersson, and B. Kasemo, Appl. Phys. Lett. 92, S. Laux, N. Kaiser, A. Z€oller, R. G€
otzelmann, H. Lauth, and H. Bernitzki,
053110 (2008). Thin Solid Films 335, 1 (1998).
4 10
J. R. Nagel and M. A. Scarpulla, Opt. Express 18, A139 (2010). N. A. Bakr, A. M. Funde, V. S. Waman, M. M. Kamble, R. R. Hawaldar,
5
J. R. Nagel and M. A. Scarpulla, “Design principles for light trapping in D. P. Amalnerkar, S. W. Gosavi, and S. R. Jadkar, Pramana J. Phys. 76,
thin silicon films with embedded dielectric nanoparticles,” Prog. 519–531 (2011).
11
Photovoltaics (published online). C. J. Chen, Physics of Solar Energy (Wiley, Hoboken, NJ, 2011).
6 12
S. Nunomura, A. Minowa, H. Sai, and M. Kondo, Appl. Phys. Lett. 97, H. Sai, Y. Kanamori, K. Arafune, Y. Ohshita, and M. Yamaguchi, Prog.
063507 (2010). Photovoltaics 15, 415 (2007).
7 13
J. R. Nagel, S. Blair, and M. A. Scarpulla, Proc. SPIE 8256, 825606 (2012). D. Wan, H.-L. Chen, T.-C. Tseng, C.-Y. Fang, Y.-S. Lai, and F.-Y. Yeh,
8
Handbook of Optical Constants of Solids, edited by E. D. Palik (Academic Adv. Funct. Mater. 20, 3064 (2010).
14
Press, 1998). S. E. Han and G. Chen, Nano Lett. 10, 4692 (2010).

Downloaded 19 Apr 2013 to 155.98.26.154. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions