APPLIED PHYSICS LETTERS 98, 153508 共2011兲

Silicon diffusion in aluminum for rear passivated solar cells

Elias Urrejola,1,a兲 Kristian Peter,2 Heiko Plagwitz,2 and Gunnar Schubert2
1
International Solar Energy Research Center (ISC)–Konstanz, Rudolf-Diesel-Str. 15,
D-78467 Konstanz, Germany
2
Sunways AG, Macairestrasse 3-5, D-78467 Konstanz, Germany
共Received 7 March 2011; accepted 28 March 2011; published online 12 April 2011兲
We show that the lateral spread of silicon in a screen-printed aluminum layer increases by
共1.50⫾ 0.06兲 ␮m / ° C, when increasing the peak firing temperature within an industrially applicable
range. In this way, the maximum spread limit of diffused silicon in aluminum is predictable and does
not depend on the contact area size but on the firing temperature. Therefore, the geometry of the rear
side pattern can influence not only series resistance losses within the solar cell but the process of
contact formation itself. In addition, too fast cooling lead to Kirkendall void formations instead of
an eutectic layer. © 2011 American Institute of Physics. 关doi:10.1063/1.3579541兴

The improvement of solar cell efficiency in industrial
production has motivated important contributions to rear sur-
face passivation techniques.1–3 However, the optimization of
the local contacts between screen-printed aluminum pastes
and silicon is not trivial and requires deeper understanding of
the metal–semiconductor interaction. Indeed, a compromise
between the contact area and finger spacing is an essential
issue when reducing series resistance.4–6 On the other hand,
the local formation of a high-quality p+-doped layer 关back-
surface-field 共BSF兲兴 to improve the cell performance,7–9 is
still a challenge. Recently, it has been shown that the high
overlapping of aluminum on each side of the local contact
opening is essential for a well-formed local BSF and the
minimization of the contact resistivity.10 Thereby, the design
of the rear side pattern can influence the series resistance
losses and the process of contact formation. In this letter, we
search for the minimum contact spacing allowed for an op-
timal rear side pattern based on the interdiffusion between
liquid aluminum and silicon. These results may have appli-
cations on screen-printed back-contacted and rear passivated
silicon solar cells.
Figure 1共a兲 shows the cross-section of a silicon solar
cell with a passivated and locally contacted rear side, as a
model for this letter. Three variables describe the rear side
structure—the width of the local contact openings 共LCO兲, d1;
the maximum spread limit of diffused silicon in aluminum
layer, d2; and the contact spacing, L p. Thus, 共d2 − d1兲 / 2 rep-
resents the spread of silicon in aluminum on each side of the
LCO 共away from the contact area兲. As shown in the cross-
sectional model of Fig. 1共a兲, the contact area between silicon
and aluminum is restricted to the LCO, d1. An aluminum
layer fully covers the rear surface. The local BSF forms in
the LCO at the rear of the device structure due to local
aluminum-silicon interaction. The microscope image of Fig.
1共b兲 shows a section of the rear side of a processed solar cell
with the same rear structure as shown by Fig. 1共a兲. The dark-
gray regions 共within d2兲, which are visible after firing in the
aluminum layer, do not represent the local BSF formation
because they are wider than the LCO 共d2 ⬎ d1兲. The under-
standing of this phenomenon motivated the development of
the present study.
a兲
Electronic mail: elias.urrejola@isc-konstanz.de.
Polished Czochralski p-type silicon wafers with
1.5⫾ 0.5 ⍀ cm resistivity were fully covered with a dielec-
tric insulation layer deposited by the plasma-enhanced
chemical vapor deposition approach. The LCO 共d1兲 were
achieved by screen printing an etching paste, which contains
phosphoric acid, a useful etchant of dielectric films.11 The
etching of the dielectric was performed by heating the wafers
in an infrared belt furnace for 4 min at 330 ° C. The dried
etching paste was removed within a few seconds in an
ultrasonic bath filled with 0.2% potassium hydroxide diluted
with deionized water. A broad range of d1 was chosen be-
tween 100– 500 ␮m in steps of 50 ␮m 共the real values are
⬇20 ␮m broader due to the spreading of the etching paste兲.
A state of the art 20 ␮m thick aluminum contact was screen-
printed, fully covering the rear passivation layer and the
opening lines. The alloy was formed after sintering the
samples following a standard firing furnace profile. Three
peak firing temperatures were applied: 750, 850, and
950 ° C. The sharp limits of the visible dark-gray regions
共d2兲 were measured by optical microscopy.
The aluminum layer was characterized by scanning elec-
tron microscopy 共SEM兲 and energy dispersive spectrometry/
energy dispersive x-ray 共EDS/EDX兲. The three layers 共A, B,
C兲, as shown by Fig. 1共a兲, are as follows:10,12 the local BSF
共A, composition Si–1%Al兲; the eutectic layer or aluminum-
silicon alloy formation 共B, composition Al–12.6%Si兲; and
the aluminum layer in porous state, C. For rear passivated
solar cells, we redefine the aluminum layer in two regions;
共1兲 the visible dark-gray region 共within d2兲 in hypereutectic
alloy 共composition ⬎Al– 12.6%Si兲 and 共2兲 the rest of the
aluminum layer formed by solid particles of the aluminum
paste. Figure 2 shows on the y-axis the optical measurement
of the dark-gray regions 共spread limit of silicon in aluminum,
FIG. 1. 共Color online兲 共a兲 Rear passivated solar cell in cross section 共front
side simplified兲. d1: width of LCO. d2: spread limit of silicon in aluminum.
L p: contact spacing. A: BSF. B: Eutectic layer. C: Aluminum layer. 共b兲
Section of the rear side of a solar cell showing dark-gray lines 共not to scale兲.

0003-6951/2011/98共15兲/153508/3/$30.00 98, 153508-1 © 2011 American Institute of Physics

Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
153508-2 Urrejola et al.
FIG. 2. Temperature linear dependency of the spread limit of silicon in the
screen-printed aluminum layer. Error bars shown for d1 = 500 ␮m.
d2兲, versus the peak firing temperature, for different LCO
widths, d1. For a LCO of 500 ␮m, the error bars plotted
show small standard deviation in the optical microscopy
analysis. For three d1 values, a linear fit is shown as a
guide for the eye. The same linear dependency on the
peak firing temperature is remarkable for all data. The
slope of the line is 共3 ⫾ 0.12兲 ␮m / ° C, which means, on
each side of the LCO the spread of silicon increases by
共1.50⫾ 0.06兲 ␮m / ° C. Additionally, the lateral spread limit
of silicon from the edge of the LCO, 关d2 − d1兴 / 2, is deter-
mined as follows: 共75⫾ 9兲 ␮m at 750 ° C, 共225⫾ 30兲 ␮m
at 850 ° C, and 共375⫾ 90兲 ␮m at 950 ° C. These results
show that for a certain firing temperature the spread limit of
silicon in the aluminum layer is constant and does not de-
pend on the width of the LCO. Thus, the maximum spread
limit of diffused silicon in aluminum is predictable.
A line scanning was performed by EDX in the
aluminum-silicon alloying structure for two samples fired at
the high peak firing temperature of 950 ° C. The SEM mi-
crographs and the EDX results are shown in Fig. 3. After
laser cutting, the cross-sectional samples were analyzed un-
der the following conditions: acceleration voltage 10 kV,
specimen current 780 pA, scanning speed 5.4 ␮s / pxl, and
electron beam focused to 660 nm in diameter. Counts at the
FIG. 3. SEM/EDX analysis. 共a兲 80 ␮m LCO, eutectic and BSF formed. 共b兲
EDX analysis of 共a兲 and Gaussian fit, following the silicon concentration to
the spread limit at 375 ␮m. 共c兲 125 ␮m LCO, Kirkendall void and BSF
formed. 共d兲 EDX analysis of 共c兲. The dotted lines in 共a兲, 共c兲 delimit the BSF;
in 共b兲, 共d兲 the region showed by the SEM figures.
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
Appl. Phys. Lett. 98, 153508 共2011兲
Si K␣ line were taken for 300 s with a line scan width of
10 ␮m. The silicon content was measured from the center
of the LCO, 200 ␮m to the left and right. Although the
results of this method are influenced by structured inhomo-
geneities of the aluminum layer,14 we see that the silicon
concentration in the aluminum decreases with increasing dis-
tance from the center of interface 共LCO兲, as already pub-
lished elsewhere15,16 关see Fig. 3共b兲兴.
Figure 3共a兲 shows an eutectic layer formed below the
aluminum layer for a d1 ⬇ 80 ␮m. Its silicon concentration,
as shown by Fig. 3共b兲, follows a Gaussian fit centered at the
LCO, decreasing to the left and right, to the predicted spread
limit of 375 ␮m. From the Gaussian fit, the maximum con-
centration is represented by 27 counts, while the concentra-
tion at the spread limit 共d2 − d1兲 / 2 is represented approxi-
mately by three counts. For a high peak firing temperature
of 950 ° C a silicon concentration of C关共d2 − d1兲 / 2兴 / C0
= 11.11% is still present in the aluminum layer at the spread
limit of silicon, 375 ␮m far away from the LCO. The BSF
forms homogeneously below the contacts, up to 8 ␮m deep.
Figure 3共c兲 shows no eutectic layer formed below the
contacts but shows instead a void. For this sample, the
aluminum was deposited on a broader contact opening 共d1
⬇ 125 ␮m兲 and fired at the high peak firing temperature
共fast cooling兲. The void present is normally found in samples
with broader LCO and which are cooled too fast. Contrary to
other authors,17 a homogeneous BSF forms below the void.
The silicon composition in the aluminum layer, for the
sample with a void, is shown by Fig. 3共d兲. It is described as
an irregular form with a higher concentration at the center of
the opening and two maxima located more than 150 ␮m
away from the center of the opening, to the left and right. An
explanation for this phenomenon is given by the following
model of the local contact formation between aluminum and
silicon.
The formation of the aluminum-silicon alloy begins with
the melting of aluminum. The liquid aluminum wets the sili-
con surface in the local openings made in the dielectric, and
then silicon dissolves in the aluminum layer. The depth of
penetration of aluminum in silicon is a function of the tem-
perature and of the aluminum spherical diameter.21 For a
peak firing temperature of 950 ° C, the depth of penetration
is 20 ␮m, as shown by Figs. 3共a兲 and 3共c兲. Due to the higher
solubility of silicon in aluminum than that of aluminum in
silicon,16,20 a higher volume of silicon atoms diffuse in the
aluminum than aluminum atoms in the silicon. If the peak
firing temperature is too high, the diffusion is enhanced and
silicon spreads deeper in the aluminum layer. At the inter-
face, the aluminum in direct contact with the silicon bulk
saturates first, then the silicon diffusion proceeds from the
edges of the interface in both lateral directions into the wide
amount of liquid aluminum mass, which is not yet saturated
by silicon. After few seconds, a large amount of silicon at-
oms is found in the aluminum layer, more than 370 ␮m
away from the contact area. The strong presence of alumi-
num atoms converts the immediately adjacent surface at the
penetration, to p-type silicon 共local BSF兲. Once the firing
temperature has reached the maxima, the liquid solution
saturates.
When the contact size is too broad, more silicon surface
interacts with the aluminum particles, and the melt located in
the middle part of the interface 共LCO兲 will saturate too fast.
153508-3 Urrejola et al.
Furthermore, below too high firing temperatures 共fast cool-
ing兲, a high elastic stress field may occur in the melt during
the alloying.18 Due to the Kirkendall effect,19 which explains
that the rates at which two types of atoms diffuse are not the
same, the high generation of vacancies during the silicon-
aluminum interdiffusion may coalesce in the melt, causing
the nucleation and formation of Kirkendall voids18 instead of
an eutectic layer.
By cooling down the sample, the composition of silicon
in the liquid-phase has to decrease following the liquid curve
of the binary system,22 however, only the edges of the LCO
are still in direct contact with the aluminum layer due to the
presence of the void. By reaching the eutectic temperature,
no more silicon segregates and the whole remaining liquid
solidifies, leaving a high concentration of silicon in alumi-
num, away from the surface of contact 关shown by the two
maxima of Fig. 3共d兲兴. If the interaction between aluminum
and silicon takes place only within d2, where a visible dark-
gray region is sharply delimited, the spread of silicon in the
aluminum layer is diffusion-limited. Thereby, an exponential
decrease with the distance from the center of the interface is
evident due to the Fick’s law of diffusion.13
Because the diffusion of silicon in the liquid aluminum
proceeds laterally within the aluminum layer, a large amount
of aluminum should overlap each side of the dielectric open-
ing to achieve an optimal metal–semiconductor interaction.
This overlapping of aluminum on narrow dielectric openings
is evident due to the reduction in the contact resistivity and
the homogeneous formation of the local BSF.10 We suggest
that the maximum spread limit of diffused silicon, d2, may
determine the minimum width of the aluminum metallization
needed for an optimal contact and full local BSF formation.
Thus, the aluminum overlapping is determined by d2. In this
way, the finger spacing may also be determined by the spread
limit of silicon in the screen-printed aluminum layer 共L p
ⱖ d2兲. Another application of the aluminum overlapping may
be found on interdigitated back-contact n-type silicon solar
cells to obtain high-quality screen-printed aluminum-alloyed
emitters.
In conclusion, we give further understanding of the local
contact formation between an screen-printed aluminum
paste and narrow silicon contact areas. The spread of
silicon in a screen printed aluminum layer increases by
共1.5⫾ 0.06兲 ␮m / ° C when increasing the peak firing tem-
perature in a range of 750– 950 ° C. This lateral spread limit
of silicon, on each side of the dielectric opening, does not
depend on the contact area size but on the firing temperature,
Author complimentary copy. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
Appl. Phys. Lett. 98, 153508 共2011兲
and is predicted as 75 ␮m, 225 ␮m, and 375 ␮m for
750 ° C, 850 ° C, and 950 ° C, respectively. Thus, the width
of the aluminum contact may be determined by the maxi-
mum spread limit of diffused silicon in an aluminum layer.
We suggest that the contact spacing should be equal or larger
than the maximum spread limit of diffused silicon in alumi-
num. Additionally, the void formation below the aluminum-
silicon contacts may be explained by the Kirkendall effect.
Our result may be applied to develop high-efficiency screen-
printed back-contacted and rear passivated solar cells suit-
able for industry.
The authors gratefully acknowledge the financial support
by the German Federal Ministry of Education and Research
under Contract No. 03SSF0335I, and Merck KGaA for the
kind supply of the etching paste.
1
D. Niinobe, H. Morikawa, S. Hiza, T. Sato, S. Matsuno, H. Fujioka, T.
Katsura, T. Okamoto, S. Hamamoto, T. Ishihara, and S. Arimoto, Sol.
Energy Mater. Sol. Cells 95, 49 共2011兲.
2
D. Y. Lee, H. H. Lee, J. Y. Ahn, H. J. Park, J. H. Kim, H. J. Kwon, and J.
W. Jeong, Sol. Energy Mater. Sol. Cells 95, 26 共2011兲.
3
M. Hofmann, S. Janz, C. Schmidt, S. Kambor, D. Suwito, N. Kohn, J.
Rentsch, R. Preu, and S. W. Glunz, Sol. Energy Mater. Sol. Cells 93, 1074
共2009兲.
4
S. M. Yang and J. Pl, Solid-State Electron. 53, 925 共2009兲.
5
H. Plagwitz and R. Brendel, Prog. Photovoltaics 14, 1 共2006兲.
6
J. Müller, K. Bothe, S. Gatz, F. Haase, C. Mader, and R. Brendel, J. Appl.
Phys. 108, 124513 共2010兲.
7
A. Urueña, J. L. Hernández, J. John, J. Poortmans, and R. Mertens, Pro-
ceedings of 25th EUPVSEC, Hamburg, Germany, 2010, pp. 2562–2564.
8
F. S. Grasso, L. Gautero, J. Rentsch, R. Preu, and R. Lanzafame, Proceed-
ings of 25th EUPVSEC, Hamburg, Germany, 2010, pp. 371–374.
9
J. Krause, R. Woehl, and D. Biro, Proceedings of 25th EUPVSEC, Ham-
10
burg, Germany, 2010, pp. 1899–1904.
E. Urrejola, K. Peter, H. Plagwitz, and G. Schubert, J. Appl. Phys. 107,
124516 共2010兲.
11
W. van Gelder and V. E. Hauser, J. Electrochem. Soc. 114, 869 共1967兲.
12
V. A. Popovich, M. Janssen, I. M. Richardson, T. van Amstel, and I. J.
Bennett, Sol. Energy Mater. Sol. Cells 95, 93 共2011兲.
13
P. G. Shewmon, Diffusion in Solids 共McGraw-Hill, New York, 1963兲, pp.
9–14.
14
V. D. Kurochkin, L. P. Kravchenko, L. N. Kuz’menko, and L. A. Tsurpal,
Powder Metall. Met. Ceram. 44, 191 共2005兲.
15
A. Paccagnella, G. Ottaviani, P. Fabri, G. Ferla, and G. Queirolo, Thin
Solid Films 128, 217 共1985兲.
16
J. O. McCaldin, Appl. Phys. Lett. 19, 524 共1971兲.
17
T. Lauermann, T. Lüder, S. Scholz, B. Raabe, G. Hahn, and B. Terheiden,
Proceedings of 35th IEEE PVSC, Hawaii, USA, 2010, p. 000028.
18
A. Paul, Ph.D. thesis, Technische Universiteit Eindhoven, 2004.
19
E. O. Kirkendall, Trans. AIME 147, 104 共1942兲.
20
S. I. Fujikawa, K. I. Hirano, and Y. Fukushima, Metall. Trans. A 9, 1811
共1978兲.
21
T. Chung, J. Electrochem. Soc. 109, 229 共1962兲.
22
J. L. Murray and A. J. McAlister, J. Phase Equilib. 5, 74 共1984兲.