Content uploaded by Mario Bähr
Author content
All content in this area was uploaded by Mario Bähr on Jun 29, 2014
Content may be subject to copyright.
LOCAL STRUCTURING OF DIELECTRIC LAYERS ON SILICON FOR IMPROVED SOLAR CELL
METALLIZATION
K. Neckermann1, S. A. G. D. Correia2, G. Andrä3, M. Bähr1, J. Lossen2, E. Ose3, L. Redlich3, I. Köhler4, H. Metzner1
1SolarZentrum Erfurt, CiS Institut für Mikrosensorik GmbH, Konrad-Zuse-Str. 14, 99099 Erfurt, Germany,
Phone: +49 - 361 - 663 12 14, Fax: +49 - 361 - 663 14 13, E-mail: kneckermann@cismst.de
2ersol Solar Energy AG, Wilhelm-Wolff-Str. 23, 99099 Erfurt, Germany
3Institut für Physikalische Hochtechnologie e. V., Albert-Einstein-Str. 9, 07745 Jena, Germany
4Merck KGaA, P.O. Box, 64271 Darmstadt, Germany
ABSTRACT: Using different types of nanosecond lasers with wavelengths in the range from 248 to 1064 nm and a
femtosecond (fs) laser with λ = 785 nm, dielectric SiNx or SiOx layers on silicon wafers were opened locally to form
diffusion and metallization masks. The wafers were all Cz-grown single-crystalline material and were either smooth, i.e.
shiny etched, or textured and had a phosphorus-doped emitter or not. For all laser types, the laser parameters which were
necessary and sufficient to open the dielectric layers were determined as optimal parameters. Additionally, screen-
printed etching pastes were employed to open the dielectric layers in parallel experiments. The opened structures were
characterized microscopically and by means of effective minority carrier lifetime measurements. The effects of the
utilized laser parameters on surface structure and carrier lifetime are discussed. Secondary ion mass spectroscopy reveals
the effects of the different laser irradiations on the phosphorus doping profile. SunsVoc measurements confirm the
capability of the investigated process steps in solar-cell processing.
Keywords: laser processing, silicon nitride, silicon oxide
1 INTRODUCTION
The efficiency of crystalline silicon solar cells out of
the standard industrial process is limited due to
restrictions imposed by the generally employed screen-
printing process for front and back-side metallization [1].
On the front side, the metallization pastes have to be
optimized simultaneously for the pervasion of the
standard dielectric SiNx layer, a good ohmic contact to
the emitter, and a low specific resistance. That is why
screen printing on the front side is, in general, only
applicable to emitters with sheet resistances beneath a
certain level. Moreover, the comparatively high series
resistances produced by the front-side metallization
reduce the fill factor and thus the efficiency of the
processed cells.
Therefore, some effort has been devoted to the
investigation of alternative metallization procedures
which are compatible to low-cost industrial mass
production [2]. Among these, laser processes offer some
advantages when employed to open defined local
structures in the anti-reflection coating on the front of the
cell. These local structures are then susceptible for
promising metallization processes including wet
chemical processes and also screen printing with low-
resistivity pastes which can be processed at
comparatively low temperatures. On the other hand, laser
processing is well-known to be capable of introducing
material modifications into silicon wafers including
lifetime limiting defects and internal stress [3]. Hence,
before these alternative metallization concepts can enter
into production their benefits have to be shown on a
laboratory scale first.
In this paper, we open locally different types of
dielectric layers, either SiNx or SiOx, by different types
of lasers in order to shed light on the interplay of laser
beams and material effects as a function of laser
parameters such as wavelength, energy fluence, pulse
length and pulse repetition time. The SiNx layers are part
of the industrial standard cell, while SiOx films play
important roles in various high-efficiency concepts. To
compare with the laser processes, a screen-printed
etching paste is used for a local opening of the dielectric
layers.
The ablation of the dielectric layers SiNx and SiOx
without damaging the underlying silicon surface using a
nanosecond laser is difficult because of the low
absorption coefficient of SiNx and SiOx in comparison to
silicon [4]. Thus, it is very important to choose adequate
wavelengths, for which absorption in the dielectric layer
takes place. As a consequence a laser wavelength of
λ < 350 nm is required for light absorption within the
SiNx layer. Still lower wavelengths are needed for
absorption within SiOx. By using very small pulse
duration in the femtosecond range non-linear effects [5]
in the absorption process result in ablation of material
nearly without heat transfer.
2 EXPERIMENTAL DETAILS
In the present experiments, different lasers or an
etching paste were employed to open dielectric SiNx or
SiOx layers on Cz-Si wafers. The latter had been
processed according to the pathways of the reaction
scheme which is depicted in figure 1. Therefore we used
an industrial standard diffusion process (ρsh = 47 Ω/sq),
an industrial standard PECVD SiNx of about 80 nm and a
APCVD SiOx of 100 nm thickness.
Figure 1: Scheme of processing of the used wafers.
The laser experiments included the following
features:
• Nd:YAG laser, λ = 1064 nm
• Nd:YAG laser, λ = 532 and 355 nm
• Excimer laser with λ = 248 nm
• Ti:sapphir laser with λ = 785 nm
The Nd:YAG lasers and the Excimer laser have pulse
durations in nanosecond range whereas the Ti:sapphire is
a femtosecond laser.
The parameters for a complete opening of the
dielectric layers were determined for each laser (optimal
parameters). To this end, the fluences of the Nd:YAG
lasers and the femtosecond laser were varied. For the
excimer laser, the number of pulses at a constant fluence
of 580 mJ/cm² was changed.
The etching paste was Solar Etch BE 01 of Merck
which was printed by means of a common screen printer.
The nominal finger width of the screen was chosen to be
80-85 µm. The etching paste was activated by a heat
treatment at 350°C for 90 s and was finally removed by
0.1 % KOH at 40°C in an ultrasonic bath for 90 s.
3 RESULTS AND DISCUSSION
3.1 Optical characterizations
As a first step, the laser fluence in combination with
the other laser parameters, which are necessary and
sufficient to remove locally the dielectric SiNx or SiOx
layer were determined. These fluences range from 0.03 to
7.75 J/cm² for the Nd:YAG lasers. The parameters, for
which the dielectric layer was completely opened while
damaging the Si surface as little as possible, will be
referred to as "optimal parameters". The fluences of the
optimal parameters are between 0.8 and 0.1 J/cm².
Figure 2 shows examples of SEM images of areas of
textured wafers, where the SiNx layer has been removed
using the optimal parameters. For the longest wavelength
of 1064 nm, the image indicates a complete surface
melting (a), while for the shorter wavelengths of 532 and
355 nm the tops of the pyramids only show signs of
melting, (b) and (c), respectively. These observations
were made after irradiation with a Nd:YAG laser,
whereas the short pulse femtosecond laser (785 nm) in
figure 2 (d) apparently did not induce any surface
modifications.
In the SEM image of Fig. 3 (a), which shows laser
irradiated and non-irradiated areas, we are confronted
with a less pronounced surface texture below a SiOx
layer, the removal of which is clearly accompanied by a
complete surface melting (Nd:YAG, 1064 nm).
Proceeding to the shorter wavelength of 532 nm, the
surface melting appears to be less complete (b), while
finally, at 355 nm (c) the topmost Si layer only was
melted and so a somewhat smoothed texture appears.
Using fluences above the optimal one at a constant
wavelength, the silicon surface is melted more and more.
Figure 2: SEM images at the textured silicon surface
after laser opening of the SiNx layer with different lasers
using the optimal parameters: a) λ = 1064 nm, b)
λ = 532 nm, c) λ = 355 nm and d) femtosecond laser with
λ = 785 nm.
Figure 3: SEM images at the textured silicon surface
after laser opening of the SiOx layer with different
Nd:YAG lasers using the optimal parameters: a)
λ = 1064 nm, b) λ = 532 nm and c) λ = 355 nm.
Surface melting is avoided when etching pastes are
used instead of lasers. Fig. 4 shows images taken with an
optical microscope which demonstrate the precise local
removal of a SiNx layer by the etching paste on a smooth
(a) and a textured surface (b). The measured line width is
about 105 µm for the shiny-etched surface and 125 µm
for the textured surface. At the edges of the lines thinner
layer thickness is observed due to sloping of the paste.
The same results were observed after local removing of
SiOx layers. Furthermore, comparable geometries can be
obtained with laser processing (not shown).
Figure 4: Optical microscopy pictures of local structured
SiNx layers on a shiny etched (a) and a textured (b)
surface.
3.2 Lifetime measurements
Minority carrier lifetime measurements using the
MWPCD (microwave-detected photoconductance decay)
were carried out in order to look for the effects of the
local opening of the dielectric films on the electronic
quality of the underlying bulk material. These
measurements were done on smooth shiny-etched wafers
without emitter, after the remaining parts of the dielectric
layers had been removed in HF and after subsequent
surface passivation using iodine ethanol solution. Fig. 5
(left) shows lifetime mappings of a wafer piece with
three columns of laser-opened stripes which are clearly
visible as stripes of low lifetime. The columns
correspond to the indicated laser wavelengths (Nd:YAG)
while the laser fluence is decreasing from top to bottom.
The stripes which relate to the optimal fluences are
encircled. Fig. 5 (right) shows the lifetime mapping of a
wafer piece which had had four quadratic windows in the
dielectric layer each opened with an excimer laser at
different pulse numbers (between 10 and 500) with
constant energy per pulse. Obviously, the number of
pulses does not very much affect the lifetime which is
lowered to a few micro-seconds in all four cases. The
encircled window corresponds to the critical number of
pulses which is 50.
Figure 5: Lifetime measurements after laser opening of
the SiNx layer using the different Nd:YAG lasers
(λ = 355-1064 nm) and the Excimer laser (λ = 248 nm)
after iodine ethanol passivation.
Fig. 6 shows analogous examples of lifetime
mappings of wafer pieces on which windows in the
covering SiNx (left) or SiO
x (right) layers had been
produced by means of fs-laser irradiation. In these
examples, the lifetime is - as in the examples of Fig. 5 -
also reduced to a few micro-seconds in the laser-treated
areas although these wafer pieces yielded a much higher
lifetime for the non-irradiated areas due to better quality
of the starting material.
Figure 6: Lifetime measurements on two wafer pieces
where the dielectric layers SiNx (left) and SiOx (right)
were locally opened with the femtosecond laser. The
structures lasered with the optimal parameters are
encircled.
In contrast to the laser treated wafers, the opening of
the dielectric layers with an etching paste does not at all
affect the lifetime. This is illustrated in Fig. 7, which
shows a sketch of the printed test structure (left) and the
corresponding lifetime mapping of the wafer piece. In
this case, a SiNx film had been opened. For SiOx layer
opening, also effects on carrier lifetime do not occur.
Figure 7: Sketch of the opened structures using the
etching paste (left) and the lifetime mapping of the
structured wafer after iodine ethanol passivation.
3.3 SIMS analyses
As mentioned above, an important possible
application of laser-opened local structures in dielectric
layers will be a modification or complete substitution of
screen printed front contacts. Therefore, the reaction of
the phosphorous doped emitter to the laser treatment will
be of major importance. Fig. 8 shows a SIMS (secondary
ion mass spectroscopy) measurement of the phosphorus
depth profile of a standard industrial emitter which is
termed "starting profile" and the profiles which result
from this starting profile due to the laser irradiation with
different lasers each with the optimal parameters. All
these measurements were done with smooth shiny-etched
wafers. Except for two cases, the depth of the P profile
increases proportionally to the laser wavelength for
locally opened structures using the optimized laser
parameters. This is probably an effect of the increasing
penetration depth of the laser light with increasing
wavelength. The excimer laser (λ = 248 nm) irradiation
was done at a fluence above the optimal one which
explains the unexpectedly deep profile in this case. In
contrast, the P profile after fs-laser irradiation is even
flatter than the starting profile and shows a lower surface
concentration. So, we assume a Si material loss at the
wafer surface due to ablation whereas the changing of the
profiles for the nanosecond lasers (Nd:YAG and
Excimer) are a result of melting processes. Note that no
signs of ablation are visible in the REM image of a fs-
laser-treated textured wafer surface [cf. Fig. 2 (d)].
Figure 8: Phosphorus depth profiles obtained after
structuring the dielectric layer using the Nd:YAG lasers,
the Excimer laser and the Ti:sapphire (fs) laser in
comparison to the starting profile (ρsh = 47 Ω/sq).
3.4 SunsVoc measurements
Since the metallization of the laser grooves is still
outstanding, we performed SunsVoc measurements on
textured wafers having a standard emitter, a standard
SiNx layer on the front side and a standard aluminium
back contact with and without laser-opened local
structures. As an example, we show, in Fig. 9, the J(V)
characteristics constructed from SunsVoc data for a
wafer covered with a standard SiNx layer and for a laser-
opened area (Nd:YAG laser, λ = 355 nm). In the latter
case, a lower JSC was assumed to account for the missing
antireflection coating. Both curves indicate rather high
VOC values of above 600 mV and also rather high pseudo
fill factors of FF = 0.8. These data suggest potential cell
efficiencies of 18 %, although this value is not affected
by series resistance. However, there appear to be good
chances to tap the full potential using an optimized
metallization, e.g. via wet-chemical methods, of the
locally laser-opened structures.
Figure 9: SunsVoc measurements confirm the capability
of the investigated process steps in solar-cell processing.
4 CONCLUSIONS
In the present paper, either laser processes or an
etching paste were employed in order to locally remove
dielectric SiNx or SiOx layers from the surface of Cz-
Silicon wafers. The effects of these opening procedures
on the surface topography, the local effective minority
carrier lifetime, and a standard P doping profile were
investigated. Additionally, the potential of the applied
process steps for solar cell processing was elucidated by
means of SunsVoc measurements.
For the different lasers, the optimal parameters were
determined for which the dielectric layer is removed at
the least damage of the underlying Silicon material. SEM
images demonstrated the complete or partial surface
melting when ns laser pulses were employed (Nd:YAG),
while fs pulses did not show visible effects on a textured
surface. These findings correspond well to the SIMS
results of P doped emitters which indicate surface
melting in case of the ns -pulse lasers, while material
ablation is occurring for the fs-laser treatment.
Comparing the laser processes to the layer removal
by the screen-printed etching paste, we find drastic local
reductions of carrier lifetimes for the former while
lifetime is not at all affected by the latter. Still, SunsVoc
data suggest the laser treatments not to have detrimental
effects on the solar cell performance since the prospected
cell parameters are the same as those without laser
irradiation.
In conclusion, locally opened structures in dielectric
layers are now ready for applications in diffusion
processes as barrier layers in terms of selective emitters
or in new metallization concepts, for example wet-
chemical plating. The benefits of etching pastes for
structuring SiNx layers with a standard finger grid pattern
in combination with the subsequent screen-printing of
dedicated metallization pastes are demonstrated in a
publication by Bähr and coworkers [6].
5 ACKNOWLEDGEMENTS
The authors would like to thank Mr. Strutzberg, CiS
Erfurt, for the SIMS analyses and Mr. Bergmann, IPHT
Jena, for his help in determining the laser parameters.
The financial support by the German ministry of BMU
for the project ProgS (No. 0327521B) is gratefully
acknowledged.
6 REFERENCES
[1] S. Kontermann, G. Emanuel, J. Benick, R. Preu, G.
Willeke, 21st EPVSEC, Dresden 2006, pp. 613-616
[2] M. McCann, I. Melnyk, E. Wefringhaus, A. Hauser,
P. Fath, S. Roberts, T. Bruton, D. Jordan, 19th
EPVSEC, Paris, 2004
[3] A.Esturo-Breton, M. Ametowobla, C. Carlson, C.
Köhler, J.H. Werner 21st EPVEC, Dresden, 2006
[4] A. Grohe, C. Harmel, A. Knorz, S.W. Glunz, R. Preu,
G.P. Willeke, 21st EPVEC, Dresden, 2006, pp. 1399-
1402
[5] P. Engelhart, N.-P. Harder, T. Neubert, H. Plagwitz,
B. Fischer, R. Meyer, R. Brendel, 21st EPVEC,
Dresden, 2006, pp. 773-776
[6] M. Bähr, S. Kim, S. Sridharan, C. Khadilkar, A.
Shaikh, I. Köhler, M. Reichardt, M. Kumar, to be
published at this conference