Development of Epitaxial Interdigitated Back Contact Solar Cell by

Development of Epitaxial Interdigitated Back Contact Solar Cell by HWCVD Process
Amirjan Nawabjan1*, Antulio Tarazona1, Darren M Bagnall2, Stuart A Boden1,
Group, Electronics and Computer Science, University of Southampton, Southampton, SO17
2School of Photovoltaic and Renewable Energy Engineering, The University of New South Wales,
Sydney, NSW 2052, Australia
*Corresponding author: [email protected]
The design of a test platform for novel
antireflection and light trapping schemes, based
on the interdigitated back contact solar cell
concept, is presented. The fabrication process
includes POCl3 diffusion to create n+ regions
and the epitaxial deposition of p+ regions using
hot wire chemical vapour deposition (HWCVD).
To provide a comparison with traditional
antireflective methods, a silicon nitride single
layer coating and an optional texturing step to
form a random pyramid array are included in the
process. The optimum thickness of the SiNx
coating deposited by HWCVD is shown to be
110 nm and the optimum KOH etching time for
complete texturing of the surface is shown to be
30 minutes.
concept has been used to manufacture largearea cells with efficiencies of more than 24 %
Most of the encouraging results in the fields of
IBC cell were realized with emitters deposited
by plasma enhanced chemical vapour
deposition (PECVD). An alternative deposition
technique is hot wire chemical vapour
deposition (HWCVD), but this has attracted
much less attention to date due to inferior
electronic properties for the films produced
compared to those prepared using PECVD.
However, HWCVD possesses some key
technological advantages in terms of deposition
rate, gas decomposition, uniformity of the final
film, and also plasma free deposition [7]. Some
groups have already reported encouraging
results using this method [8], [9].
In recent years, interest in new methods of
achieving antireflection (AR) and light trapping
(LT) for solar cells has been growing as part of
the search for ways to increase cell efficiency
and hence drive down the cost-per-watt of PV.
These include biomimetic ‘moth-eye’ structures
[1], plasmonic metal nanoparticle arrays [2] and
Mie resonator arrays [3]. This has led to the
requirement of a solar cell test platform to
facilitate the comparison of these new AR and
LT methods with more traditional thin film and
micron-scale texturing approaches. Since the
front surface of the cell is free of any contacts,
monocrystalline silicon solar cell design is well
suited to fulfil this need [4].
IBC cells offer potentially higher efficiency
compared with conventional solar cells.
Complete elimination of the front contacts leads
to a higher short circuit current due to the
absence of shading losses and lower resistive
losses. The front and back of the cell can be
fully utilized for optical and electrical
improvement, respectively. The cell design
offers easier interconnection and increased
packing density within a module [5].
Furthermore, the approach also improves
aesthetics, increasing the likelihood of large
scale adoption in building integrated PV. The
In this paper, steps towards the development of
a HWCVD based IBC cell for testing novel AR
and LT schemes against traditional thin film and
micron-scale texturing are described. The
fabrication process is described and the cell
design is presented. This is followed by the
optimizations of two traditional AR/LT schemes
for the top surface: A thin film single layer
antireflective coating (SLAR) of silicon nitride
and a random pyramidal texture.
Cell Design
The IBC cell designs in this study are illustrated
in Figure 1 and an outline of the process listing
is given in Figure 2. Two designs are shown:
Design A has a planar surface and a SiNx SLAR
whilst design B includes an additional etching
process to texture the top surface with a random
array of pyramids.
The process begins with cleaning of a silicon
wafer (n-type, <100>, 1-10 Ω.cm). For design B,
the top surface is then textured using a KOH
etch, with the back surface protected by a nitride
layer. A photolithography and dry etch stage is
then used to define the n-type contact pattern
on the rear of the cell, through which the n+
back surface field (BSF) is formed by POCl3
diffusion at 900OC. This also forms the floating
emitter at the front surface. This is followed by
nitride deposition (by PECVD or HWCVD) to
form the SLAR. A second photolithography and
dry etch stage is then used to define the p-type
contacts. The epitaxial p+ contacts are then
fabricated by HWCVD deposition and lift-off.
Finally, a third lithography process followed by
aluminium evaporation and lift-off forms the n
and p ohmic contacts. The wafer is then diced
to form 1 cm × 1 cm individual devices.
[10] to calculate the normal incidence
reflectance spectra of the PECVD and HWCVD
deposited thin films.
Figure 2: Simplified outline of fabrication
Design A
Design B
Figure 1: Schematic of the IBC cell designs for
this study. Design A has a planar surface and
Design B has a textured surface.
Figure 3: Complex refractive index (n and k)
data for silicon nitride AR layers deposited by
In this paper, optimizations for two steps of the
fabrication process are presented. Firstly, the
optimization of the thickness of the nitride SLAR
using a transfer matrix approach, with input
from experimentally-determined film properties,
is presented. This is followed by an
experimental optimization of the KOH texturing
These simulated spectra match well with
measured reflectance spectra (see figure 4),
obtained using an integrating sphere
spectrophotometer (Bentham Instruments).
This verifies that the simulation method can
accurately predict the reflectance behaviour of
the SiNx thin films on silicon and therefore can
be used to optimize the film thickness.
Optimization of SLAR
Two methods were investigated for the
deposition of a SiNx SLAR in this work, with
standard recipes being used in each case.
PECVD nitride was deposited using a
SiH4/NH3/N2 gaseous mixture with the flow rates
of 12.5/20/500 sccm. HWCVD nitride was
deposited with a 30/450 sccm SiH4/NH3
mixture. Both processes were performed at a
temperature of 300OC. Film thicknesses and
complex refractive index data (n and k) were
determined by variable angle spectroscopic
ellipsometry (VASE, J.A Wollam). The n and k
values for nitride films deposited by PECVD and
HWCVD, using the standard recipes, are shown
in Figure 3.
The complex refractive index data was used in
a model employing the transfer matrix method
Figure 4: Simulated and measured reflectance
spectra for SLAR layers on silicon.
The optimization was carried using the transfer
matrix method to calculate reflectance spectra
for a range of film thicknesses. Each reflectance
spectra was then weighted to the AM1.5 solar
spectrum (ASTM173G, global tilt) expressed as
a photon flux density (PFD), and averaged to
give a single figure of merit called the weighted
average reflectance, Rw, using equation 1.
antireflective effect. Light reflected from one
facet of the texture is incident on an adjacent
facet and so has multiple chances of being
coupled into the cell. The method also
enhances light trapping within the cell by
changing the propagation direction of light
through refraction and thereby increasing the
optical path length. In this study, a simple
maskless technique based on a recipe from
King et al. is used to form random arrays of
upright pyramids [11]. The aim is to develop a
good quality AR scheme based on the widelyused traditional approach of micron-scale
texturing that can act as a comparison to more
novel AR and LT schemes.
 max
Rw 
 R( ) PFD ( )d
 min
 max
 PFD ( )d
 min
Values of Rw versus film thickness for the
PECVD and HWCVD nitrides are plotted in
Figure 5. The optimum film thickness, read from
the plots in Figure 5, is 100 nm for the PECVD
nitride and 110 nm for the HWCVD nitride.
These give Rw values of 6.7% and 5.2% for the
PECVD and HWCVD nitride films, respectively,
demonstrating that HWCVD compares well with
PECVD as a deposition method for SiNx SLARs.
After cleaning, monocrystalline silicon wafers
(n-type, <100>, 1-10 Ωcm) were etched in a
1.5% KOH, 3.8% IPA solution at 70°C [11] for
various lengths of time. The wafers were rinsed,
dried and the inspected by scanning electron
microscopy (SEM). The spectral hemispherical
reflectance of the surfaces was then measured
using an integrating sphere spectrophotometer
(Bentham Instruments).
SEM images of silicon surfaces etched for 15
minutes, 30 minutes, and 60 minutes are
presented in Figure 6. At 15 minutes, the
pyramids are just starting to form and do not yet
cover the whole surface. At 30 minutes, the
pyramids completely cover the wafer surface
and have base sizes in the 6-8 µm range.
Further etching for 60 minutes results in smaller
pyramids with base sizes of 4-6 µm.
Reflectance spectra of the samples are
presented in Figure 7. After a 15 minute etch,
the reflectance spectrum is similar to that of
bare silicon. However, for an etch time of 30
minutes, the reflectance is reduced significantly.
The reflectance spectrum for the sample etched
for 60 minutes is similar to the 30 minute
sample, therefore a 30 minute etch is
considered to be optimum.
Figure 5: Optimum thickness calculation for
SiNx SLAR deposited by PECVD and HWCVD.
Texturing Process Development
Etching in an alkaline solution is a commonly
used texturing method to confer an
Figure 6: SEM image of random pyramids after etching for (a) 15, (b) 30, and (c) 60 minutes. The
scale bar is 4 µm
implementing novel plasmonic or biomimetic
S. A. Boden and D. M. Bagnall, Prog.
Photovoltaics Res. Appl., vol. 18, no. 3,
pp. 195–203, 2010.
T. L. Temple and D. M. Bagnall, Prog.
Photovoltaics Res. Appl., 2012.
P. Spinelli, M. A Verschuuren, and A
Polman, Nat. Commun., vol. 3, p. 692,
M. D. Lammert and R. J. Schwartz,
Electron Devices, IEEE Trans., vol. 24,
no. 4, pp. 337–342, 1977.
M. Spath, P. De Jong, I. J. Bennet, T.
P. Visser, and J. Bakker, in
Proceedings of 33rd IEEE PVSC, 2008,
pp. 1–6, 2008
P. J. Cousins, D. D. Smith, H.-C. Luan,
J. Manning, T. D. Dennis, A.
Waldhauer, K. E. Wilson, G. Harley,
and W. P. Mulligan, Proceedings of
35th IEEE PVSC., pp. 275–278, 2010.
R. E. I. Schropp, Thin Solid Films, vol.
517, no. 12, pp. 3415–3419, 2009.
S. Lien and D. Wuu, Prog.
Photovoltaics Res. Appl. pp. 489–501,
H. M. Branz, C. W. Teplin, D. L. Young,
M. R. Page, E. Iwaniczko, L. Roybal, R.
Bauer, a. H. Mahan, Y. Xu, P. Stradins,
T. Wang, and Q. Wang, Thin Solid
Films, vol. 516, no. 5, pp. 743–746,
S. A. Boden and D. Bagnall, Prog.
Photovoltaics Res. Appl., vol. 17, no. 4,
pp. 241–252, 2009.
D. L. King and M. E. Buck, in
Proceedings of the 22nd IEEE PVSC,
pp. 303–308, 1991.
J. Schmidt, T. Lauinger, A. G. Abe, and
R. Hezel, in Proceedings of the 25th
IEEE PVSC, pp. 17–20, 1996.
Figure 7: Reflectance spectra for KOH textured
silicon surfaces, etched for various times.
Besides AR, another important task of the top
surface coating is to provide passivation and
thereby reduce the front surface recombination.
PECVD nitride films have can provide excellent
surface passivation with surface recombination
velocities down to 4 cm/s being reported [12].
Further investigation is required to determine if
the HWCVD process used here can also
provide adequate surface passivation. A thin
layer of SiO2 may need to be included in the
coating design, with the thicknesses of the
layers then re-optimized to provide both
antireflection and effective surface passivation.
Conventional solar cells employ a combination
of pyramidal texturing and an SLAR to provide
AR and LT. In this report, the two schemes have
been optimized separately and so further
optimization should be possible by considering
a combination of the two.
A HWCVD-based epitaxial IBC solar cell for use
as a test platform for novel antireflection and
light trapping schemes has been designed. For
antireflective schemes have been optimized for
inclusion in the fabrication process for the full
cells. Firstly, the optimum thickness for a
HWCVD deposited SiNx film was shown to be
110 nm, giving an average reflectance,
weighted to the AM1.5 solar spectrum, of only
5.2%. Secondly, a KOH texturing process was
tested and found to produce a suitably dense
random pyramidal array after 30 minutes of
etching. The task now is to complete the
fabrication and testing of the full cell design with
these traditional AR schemes