See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/257402325

Analysis Of RISE-IBC Solar Cells

CONFERENCE PAPER OCTOBER 2013

READS

73

9 AUTHORS, INCLUDING:
Emmanuel Van Kerschaver
First Solar
57 PUBLICATIONS 515 CITATIONS
SEE PROFILE

Available from: Emmanuel Van Kerschaver

Retrieved on: 23 November 2015

ANALYSIS OF RISE-IBC SOLAR CELLS

F. Dross(1), A. Merkle(2), E. Van Kerschaver(1), S. Baker-Finch, K. Cabanas-Holmen(1),
R. Peibst(2), N. Peter-Harder(2,3), R. Brendel(2), P. Basore(1)
(1)
Hanwha Solar America 2424 Walsh Avenue, Santa Clara, CA95510 - USA
(2)
ISFH - Am Ohrberg 1, D-31860 Emmerthal Germany
(3)
now at TOTAL, San Jose, California, USA
ABSTRACT: In order to test whether the lab facilities of Hanwha Solar America in Santa Clara are compatible with
high-efficiency processing, we commissioned ISFH to fabricate small-area crystalline-silicon high-efficiency solar
cells based on the RISE process. We reach efficiencies exceeding 23%, observed no statistically significant difference
between the cells processed completely at ISFH and the cells partly processed at Hanwha, and conclude that the
facilities are compatible with high-efficiency processing. Meanwhile, we analyze the performance reached by the
cells in the different splits of the experiment. We find that a passivation of the n-type surface based on SiO2 improves
the open-circuit voltage, but suffers from a lower (pseudo-) fill-factor compared to an Al2O3 passivation layer. We
also observe that increasing the emitter width increases the open-circuit voltage thanks to a lower saturation current
but reduces also the fill-factor because of increased majority-carrier resistivity.
Keywords: High-efficiency, Passivation, Back Contact, SiO, Al2O3, c-Si
1

INTRODUCTION

Most of the cost items of the levelized cost of solar

electricity scale (inversely) with the efficiency of the PV
device deployed in the system. The development of highefficiency devices is therefore of very high relevance to
remain competitive in tomorrows PV market. Hanwha
Solar America (HSA), the corporate advanced R&D
laboratory of the Hanwha Solar group, is pursuing the
development of such devices. In order to demonstrate that
HSAs lab facilities in Santa Clara (California, USA) are
compatible with high-efficiency device processing, HSA
commissioned ISFH (Germany) to make cells using
ISFHs high-efficiency RISE (Rear Interdigitated Single
Evaporation) back-junction solar cell process, and to
compare RISE cells fabricated solely at ISFH with RISE
cells fabricated in part at HSA.
HSA chose to commission ISFH for this
experimental test because the RISE-IBC [1-4] highefficiency solar cell process developed by ISFH
combines a number of interesting features. It is a selfaligned, Ag-free and lithography-free process [5] having
reached already on large area (155.1 cm2) an efficiency
of 21.4% [6]. A number of variations were tested: the
contact geometry, the type of passivation on the base
contact, and the starting material. Most importantly, the
rear-surface POCl3 diffusion at the beginning of the
process was either done at HSA (USA) or at ISFH
(Germany). That is, this study provides a comparison of
RISE-IBC solar cells that were completely fabricated at
ISFH in comparison with the same type of solar cells
made at ISFH from wafers that were pre-processed at
HSA. The resulting high-efficiency devices were
characterized and analyzed with respect to the process
variations. We describe in this paper the most significant
parameters influencing the device performance.
2

EXPERIMENTAL DETAILS

The structure under study is illustrated on Figure 1. It

is an n-type interdigitated-back-contact solar cell. It
consists of a diffused P-doped BSF, a diffused B-doped
emitter, both of them being passivated with some
dielectric layer. In one group, the passivation is the same
on both polarities and consists of ALD Al2O3 capped

with a SiNx PECVD layer; whereas in another group, a

selective passivation [7] was employed at the rear side:
the passivation over the emitter is based on the same
Al2O3/SiNx stack, and the passivation over the BSF is
based on thermal SiO2, capped with the Al2O3/SiNx
dielectric stack used for the passivation of the emitter.
The front surface is not intentionally doped and relies for
passivation entirely on a charged SiNx layer deposited by
PECVD.

Figure 1: Both types of RISE-IBC solar cells prepared in

this study. The cell structure on the left hand side has
Al2O3-passivation on the full rear side area. The structure
on the right hand side features a selective passivation
[7], i.e. the n-type BSF regions are passivated by a
thermally grown SiO2, and only the p-type emitter
regions are passivated by Al2O3.
Each 156-mm wafer contains 24 2.5x2.5-cm2 cells
(active designated area is 3.97 cm2), where the contact
geometry has been varied (4 different geometries). Two
wafer materials have been tested, both exhibiting several
milliseconds of lifetime. Two different (but similar)
POCl3 diffusion processes have been compared, one in
the HSA facility, and one in the ISFH facility. In total, 80
wafers have been processed (including monitor wafers),
196 cells have been measured, and a maximum efficiency
of 23.1% (independently confirmed) was achieved on a
3.97-cm2 designated cell area [8]. This paper intends to
reveal the effect of each of the features intentionally
varied (BSF passivation, geometry, wafer, POCl3) on
device performance.
3

RESULTS AND DISCUSSION

Because of the very high number of cells generated

we decided to limit the measurements to 90 cells. These
cells are a relatively random sample of all the splits. For
this reason, while attempting to draw conclusions on the

data, one must take special care of not introducing a

sampling bias. To this end, we first collect the
performance of a number of cells on each of the wafers.
Note that this number is not necessarily the same for each
wafer, and is certainly not the same for each geometry
within each wafer. We then calculate for each wafer the
average performance for cells with the same geometry.
This gives one data point per geometry, per wafer (4 data
points per wafer). Since we carried out measurements on
12 wafers, we end up with 48 data points which we
consider statistically independent from each other (i.e.
statistically linked only by the variables under study, such
as the wafer, the POCl3 process, the geometry, or the BSF
passivation method).
3.1 ISFH POCl3 versus HSA POCl3
We first compare the cells fabricated entirely at ISFH
to the ones partly processed at HSA. We therefore
identify 2 groups of data points (each group consists of
16 data points) which differ only by the fact that one
group was processed entirely at ISFH, and the other one
was partly processed at HSA. Table I, below, shows the
device performance (averages and 95% confidence
interval) of the two populations of data points.
Table I: Averages and 95% confidence interval for the
population fully processed at ISFH, and the population
partly processed at HSA

First, the very high performance obtained in this

experiment needs to be highlighted. The average
efficiency reached ~22% which is remarkable for a Agfree, litho-free device and confirms the performance
potential of the RISE process developed at ISFH.
Secondly, the average of one group systematically
lies within the confidence interval of the other group. We
cannot statistically differentiate the two groups and
conclude therefore that the front-end part of the Santa
Clara facilities of Hanwha Solar is suitable for highefficiency solar cell processes.
3.2 Al2O3 versus SiO2 passivation
We now compare 2 samples of data points by
isolating the cells having Al2O3 as base-region (BSF)
passivation from the cells having SiO2 as base-region
passivation. The emitter region is in any case passivated
with Al2O3 leading to the terminology selective
passivation in the case of a SiO2 base-region
passivation. Figure 2 shows the Voc and FF measured for
the whole dataset. T-test methodology indicates that there
is less than 0.5% chance that the two samples are coming
from the same population. In other words, the Voc is
statistically significantly higher (9mV on average) and
the FF is statistically significantly lower (0.8% absolute
on average) for the selective passivation than when Al2O3
is used for both polarities.

Figure 2: Voc and FF for two different BSF passivation

schemes.
The Voc trend is confirmed through monitor samples
double-side diffused and passivated where the saturation
current (J0) was extracted by QSSPC. Depending on
whether the POCl3 diffusion was done at HSA or at
ISFH, a factor ranging between 1.3 and 3 is observed
between the extracted saturation currents. The SiO2
passivation provides values around or below 100 fA.cm-2,
whereas Al2O3 provides values ranging from 100 fA.cm-2
(ISFH diffusion) to 300 fA.cm-2 (HSA diffusion),
explaining (at least qualitatively) the trend observed
experimentally. These values translate into an expected
Voc difference ranging between 7 and 30mV.
The reason why the passivation layer material affects
the FF remains to be fully understood. Shunt and series
resistance remain unaffected, but the sunsVoc FF (or
pseudo-FF) is also measured lower for the selective
passivation. Since the Voc is higher for the selective
passivation, it is clear that the additional recombination
has to exhibit an ideality factor higher than unity. One
possible reason is that the laser parameters used for
contact opening induces more damage [9,10,11] in case
of the SiO2-passivated BSF regions compared to the
Al2O3 passivation case. It is also possible that the
passivation where the pn-junction surfaces is better
without the presence of the SiO2 layer. The simultaneous
presence of SiO2 and Al2O3 in this specific location, two
dielectrics known to exhibit opposite fixed charges,
might interfere and reduce the overall passivation
properties of the dielectric stack. A more extensive
analysis is presented in another paper presented at the
same conference [8].
Furthermore, the fact that the two POCl3 diffusions
(at HSA or at ISFH) seem to give different values of
saturation currents on the monitor wafers deserves further
investigation. We isolated the data comparing
performance for Al2O3-passivated BSF (other factors
being equal) and calculate the average difference of FF
and Voc between the HSA and ISFH diffusions. This
time, the t-test allows a large (>25%) chance that the
difference observed is just due to luck. We cannot
conclude that the choice of POCl3 diffusion process (at
HSA or at ISFH) has any effect on either Voc or FF.
3.3 Effect of device geometry
In order to study the effect of the geometry, one has
to avoid, in the dataset, the contribution of the other
features (such as wafer, passivation type, POCl3 process).
This has been achieved by first calculating for each wafer
the average performance (average of the 4 data points
previously considered independent). The performance of

a given geometry within a wafer is then characterized by

its deviation from the average cell performance in that
wafer. Figure 3 displays this deviation to average wafer
performance as a function of the emitter dimension
(keeping the base dimension equal to 250!m).

Figure 3: Deviation to wafer average of Voc and FF when

varying emitter width (base dimension is kept constant)
In this case, the t-test confirms that Voc increases and FF
decreases when the emitter width increases from 900 to
1800 m. The Voc trend is explained by the lower
saturation current for the emitter than for the BSF. The
emitter saturation current is estimated from symmetrical
structures to be ~20 fA.cm-2 (compared to 100-300
fA.cm-2 for the BSF region). The FF trend is explained by
an increased series resistance of the electron current
flowing to the base contacts. The wafer material has a
resistivity between 1.5 and 2 ohm.cm, and a thickness of
~150 m after processing, resulting in an equivalent
sheet resistance of the base exceeding 100 ohm/sq,
sufficient to explain the reduction in FF for widely
separated base contacts (wide emitter regions). A PC2D
model [12] has been setup with the known characteristics
of the device. Varying the emitter width from 900 to
1800!m, we obtain a decrease of the FF by 3%points,
very similarly to what is measured. Meanwhile, using the
saturation currents measured on monitor wafers, the
model predicts an increase of Voc of 10 mV, to be
compared to ~4 mV observed experimentally.
In addition, when the emitter width is kept constant
(1800 !m), a statistically significant influence of the base
geometry on the short-circuit current has been observed.
Doubling the base width (to 500 m) reduces the current
by ~0.4 mA.cm-2. This effect is explained by electrical
shading [13,14,15], which is due to recombination of
electrons generated in the base region before they can
diffuse to the emitter for collection. This effect has also
been qualitatively witnessed by measuring external
quantum efficiency when the monochromatic beam is
aligned to the emitter, and when the beam averages the
current response in both emitter and base regions.
4

SUMMARY

In order to demonstrate the high-efficiency

compatibility of new facilities, we made use of a highefficiency reference process developed and available at
ISFH. We carried out a high-temperature step in our new
facilities and commissioned ISFH to fabricate RISE-IBC
solar cells from these pre-processed wafers in comparison
with cells processed completely at ISFH. Efficiencies
above 23% have been obtained, with no significantly
different performance for the cells partly fabricated in the

new facilities. An analysis of the set of cells reveals some

significant effects of the geometry of base and emitter
regions, as well as some influence of the passivation type
for the base region.
[1] Peter Engelhart, et al. Proc. of the 21st EU PVSEC,
Dresden, Germany, 2006, p. 773-776
[2] Robert Bock, et al. Applied Physics Letters 96,
263507 (2010)
[3] Agnes Merkle, et al. Prog. Photovolt: Res. Appl.
(2012), DOI: 10.1002/pip.2297
[4] Felix Haase, et al. Proc. of the 27th EU PVSEC,
Frankfurt, Germany, 2012, p. 580-585
[5] Nils.-P. Harder, et al. Physica Status Solidi Vol. 6 (3),
(2009), p. 736-743.
[6] Nils.-P. Harder, et al. Photovoltaics International, 7th
Edition, p.50-60, (2012)
[7] Fabian Kiefer, et al., IEEE Journal of Photovoltaics,
Vol.1, No.1, 2011
[8] R. Peibst, N.-P. Harder, A. Merkle, T. Neubert, S.
Kirstein, J. Schmidt, F. Dross, P. Basore, R. Brendel,
this conference, 2013
[9] S. Hermann, at al., Applied Physics A: Mat. Science
& Processing 99(1), p.151-158 (2010)
[10]Udo Rmer, et al. Proc. of the 22nd PV Science and
Engineering Conf., Hangzhou, China, 2012
[11] Felix Haase, et al. Proc. of the 22nd PV Science and
Engineering Conf., Hangzhou, China, 2012.
[12] Paul A. Basore and Kirsten Cabanas-Holmen, IEEE
Journal of Photovoltaics, 1, 1, July 2011, pp. 72-77,
10.1109/JPHOTOV.2011.2166376
[13]Frdric Dross, et al., Proc. of the 15th PV Science
and Engineering Conf., Shanghai, China, 2005
[14] De Ceuster DM, et al. Proc. of the 22nd EU
PVSEC, Milano., Italy, 2007, p.816-819
[15] Martin Hermle, et.al. 33th IEEE Photovoltaic
Specialist Conference, St. Diego, CA, 2008