22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy
EFFECT OF IMPURITIES ON THE MINORITY CARRIER LIFETIME OF SILICON MADE BY THE
METALLURGICAL ROUTE
1
Arve Holt1, Erik Enebakk2 and Anne-Karin Soiland2
Institute for Energy Technology, P.O. Box 24, NO-2027 Kjeller, Norway, mailto:arve.holt@ife.no
2
Elkem Solar AS, P.O. Box 8040, NO-4675 Vågsbygd, Norway
ABSTRACT: The objective of this study has been to investigate the relationship between the impurity level and the minority
carrier lifetime of solar grade silicon made by the metallurgical route by Elkem Solar (ES). The wafers have been
characterised by microwave photo conductance decay (µ-PCD), FeB-pair splitting and neutron activation analyses (NAA).
The results of the minority carrier lifetime measurements show higher minority carrier lifetime of ES-wafers compared with
results from wafers based on poly silicon material. A P-gettering processes has successfully been introduced, showing an
average improvement by a factor larger than two. At [Fe] of about 1013 atoms/cm3 clustering starts to form in the red zone
near the border of edge wafers. It is assumed that these clusters are electrically less active. This is supported by the increase
of lifetime near the edge in the red zone. Most likely iron silicide clusters are formed.
Keywords: defect engineering, P-gettering, annealing
1
INTRODUCTION
One of the bottlenecks in order to ramp up the
production of silicon solar cells and to reduce the price
of solar panels further are the shortage of silicon
feedstock of high enough quality. Traditionally feed
stock for the silicon based solar cell industry has been
produced through the silane or TCS route. However
today several new methods for making silicon feedstock
are under development. Elkem Solar (ES) has been
focusing making solar grade silicon made by the
metallurgical route [1].
The impurity concentration has been measured by
neutron activation analyses (NAA).
All the samples in this study were chemically
polished at room temperature in a fresh CP4 solution
consisting of 10 parts concentrated HNO3, 5 parts of
concentrated CH3COOH and 2 parts 48% HF. Typically
15-20 microns were etched away on each side. The
samples were then etched in a fresh piranha solution
consisting of 3 parts of H2O2 and 10 parts of H2SO4.
Before the P-gettering processing the samples were
etched in a 5% HF solution prior to the process of
spraying on the liquid diffusion source on both side of the
samples. The diffusion source solvent was baked away by
a heat treatment in an own for 10 minutes at 200 °C. In
this study different temperature profiles in our belt
furnace were used during P-gettering processing. After
the gettering process the phosphorous rich surface layer
was etched away by CP4 followed by a piranha etch.
The objective of this study has been to investigate the
relationship between the impurity level and the minority
carrier lifetime of ES solar grade silicon and to compare
the electrical properties of the wafers based on ES
material with wafers based on material produced through
the silane route.
2
EXPERIMENTAL SETUP
For life time measurements the both front and back
surfaces were passivated by going through a CP4 etch,
piranha etch and a HF etch, followed by adding
amorphous silicon on both sides by using PECVD
technique [2,3].
Detailed mapping of lifetime of minority carriers
have been studied at nine different positions in an ES
edge block. In parallel wafers from an edge block
produced from poly silicon by another manufacturer have
also been investigated as a reference. The lifetime has
been measured on as cut wafers and after different
processing steps: i) temperature annealing, ii) emitter
formation, and iii) high temperature gettering.
3
RESULTS AND DISCUSSIONS
3.1 Initial lifetime of samples before P-gettering
The results for the lifetime of minority carriers in
nine different positions in an ES edge block are shown in
Table 1 and on a reference edge block in Table 2.
Position one corresponds to samples at the bottom of the
block and samples from position nine correspond to
samples from the top of the block.
The lifetime of the minority carriers has been
characterised by microwave photo conductance decay (µPCD) method. The bulk [Fe] have been studied by FeBpair splitting by comparing the lifetime measurements
before and after light soaking. The calculations where
based on the equation
Table 1: Initial lifetime of an ES edge block
N Fe
!
$ 1
1 '
))
= C µ "PCD &&
"
% # before # after (
Pos
1
2
3
4
5
6
7
,
where Cµ -PCD = 3.4 1013 µs/cm3 were used. Although the
correctness of the C µ-PCD value is questionable, this
method is useful to see the [Fe] variation across the
wafer.
1155
! (µs)
1,2
3
7,8
20,4
40,9
49,5
51,2
! in bulk (µs)
1,5
4,4
11,3
29,7
62,1
72,8
77,0
! edge zone (µs)
0,8
0,5
0,7
0,9
1,0
1,1
1,1
22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy
8
9
60,2
54,6
90,5
80,0
It is therefore expected a negative gradient of the [Fe] all
the way from the edge.
1,0
1,2
Table 2: Initial lifetime of a reference edge block
Pos
1
2
3
4
5
6
7
8
9
! (µs)
34
! in bulk (µs)
59,2
! edge zone (µs)
3,5
46
38
41
50
60
48
44
60,1
63,0
43
63
70
50
54
5,7
5,9
10,0
5,6
7,1
12
14
Both tables show three columns of results. The first
column shows the average lifetime (!) of the sample. The
second column shows the average lifetime of wafere
without taking into account the contributions from the
"red" edge zones shown in Figure 1 and in Figure 2. The
third column gives an average value of the red zone
before lifetime starts to increase significantly.
Figure 2: A lifetime map of a sample from position 7 in
the reference edge block.
Figure 1: A lifetime map of a sample from position 7 of
an ES edge block.
Figure 3: Initial lifetime versus distance from the edge of
an ES edge block
As seen in Table 1 the lifetime increases from about 1
µs to about 90 µs when going from bottom to the top of
the ES the edge block. This increase is not present in the
reference block. In the latter case the lifetime shows a
constant value of about 60 µs from bottom to top.
Interestingly the lifetime in the ES block goes through a
minimum in the red zone when going from the edge and
inwards, shown in Figure 3. This effect is not observed in
the reference block.
The effect lower Fe concentration near the edge is
suggested to be due to formation stable Fe defect clusters
in matrix above a certain concentration. These clusters
are shown to be stable during the light soaking process.
The clustering near the edge may also explain the
enhancement in lifetime near the edge in the red zone.
The [Fe] has been measured by Fe-B pair splitting
and the results show that the [Fe] goes through a
maximum in the red zone shown in Figure 4.
The clustering starts to occur at [Fe] just below 1013
atoms/cm3 where the peak in iron concentration is
observed. In the reference block the iron concentration
never reaches values higher than about 1012 atoms/cm3 in
the red zone. Within the red zone of the reference block
there is no maximum in the [Fe] observed.
The red zone is formed due to in-diffusion of
impurities like Fe from the crucible and crucible coating.
The high concentration of iron in the red zone of the
ES wafers may be caused by long annealing times during
1156
22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy
the crystallisation process and/or due to high impurity
concentration in the crucible and/or crucible coating
used. In other words the electrical properties of the red
zone is process related and not related to the feedstock
material.
compared with the ES-wafers in this study.
Figure 5: A lifetime map of a sample from position 7 of
an ES edge block after P-gettering.
Figure 4: [Fe] versus distance from edge of an ES edge
block.
3.2 Lifetime of samples after P-gettering
In this case the samples were diffused at 890°C for
20 minutes, after which the temperature was abruptly
lowered to 800°C in order to slow down the P diffusion
and then gradually ramped down to 700°C during one
hour[4]. The resulted sheet resistance after diffusion
showed about 30"/square.
The results of the lifetime after gettering at different
positions in the ES edge block is shown in Table 3 in
columns named ! (µs), ! in bulk (µs), and ! edge zone
(µs). The results from the reference block is also given in
the column named !ref(µs).
Table 3: Lifetime after P-gettering for ES Edge block
Pos
1
2
3
4
5
6
7
8
9
! (µs)
86,7
117,8
130,6
134,7
136,9
163,7
178,5
170,4
!ref(µs)
88
76
79
83
77
81
85
70
51
! in bulk (µs)
105,0
145,0
153,0
164
170
201
221
! red zone (µs)
24,5
24,7
24,4
27,1
23,5
25,8
24,7
202
26,5
Figure 6: A lifetime map of a sample from position 7 in
the reference edge block after P-gettering.
Interestingly for the reference samples, the red zone area
is completely removed after the P gettering process,
while for ES material the red zone has been reduced
dramatically.
Figure 5 shows a lifetime map of an ES sample and
Figure 6 shows an example from a reference block. As
seen from the results above, the P-gettering process
improves the lifetime dramatically for the ES material. In
the bulk the average lifetime increases from about 70 to
170 µs. However in the regions with low lifetime initially
there is not observed any improvements after the
gettering process, seen by comparing Figure 1 and Figure
5. Improved lifetime after gettering has been shown by
many and most recently by Tan et al. [5], Bentzen et
al.[6] and by P. Manshanden and L. Geerligs[7]. However
these studies shows overall lower lifetime values
Figure 9 shows line profiles before and after P-gettering
from position 2 of an ES edge block. As seen from the
figure the red zone after gettering of the ES material
reflects the area where clustering of defects have
occurred. However note the remaining red zone area has
been proved with more than an order of magnitude. Line
1157
22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy
profiles of the lifetime from the edge are shown in Figure
7.
3.4 Annealing
The lifetime has been studied after annealing at
different temperatures. The results of the annealing
studies are shown in Figure 9 for one position in the ES
edge block. Also included are the results from the initial
and P-gettered wafer at the same position. As shown, the
effect after annealing is highest in the temperature range
500-600 °C. This improvement has also previously been
reported by others [8]. However in this study the Pgettering process shows much higher improvement with
respect on lifetime. This is the case both in the red zone
region and in the bulk region not contaminated during
crystallisation.
Interestingly the lifetime profiles shows two peaks in
the red zone part after annealing. This may be due to
different cluster regimes. At 600 °C this effect is largest.
Figure 7: Lifetime versus distance from the edge of an ES
edge block after gettering
3.3 P-gettering at different temperatures
P-gettering has been preformed at different
temperatures both for the ES and the reference material
for 150 minutes. The results are shown in Table 4 and in
Figure 8. At 800-900 °C the effect of gettering is highest
with more than doubling of the lifetime performance. Pgettering also in this case has the largest impact on the
wafers based on ES material.
Table 4: Results of lifetime measurements given in !s
from P-gettering processes at different temperatures
700 °C
800 °C
900 °C
890°C *
ESm
50
118
129
159
ESr
3,8
6,4
16,3
25,8
Refm
59
106
110
81
*
Temperature ramps and hold times equal to the set-up
described in section 3.2.
Figure 9: Lifetime profiles after annealing at different
temperatures from position 2 of an ES edge block
The results of lifetime measurements on the reference
block after annealing at different temperatures are shown
in Figure 10. In this case annealing shows best results at
the lowest temperature. Note also the presence the red
zone also after annealing.
Figure 8: Lifetime profiles after P-gettering at different
temperatures from position 5 of an ES edge block.
Figure 10: Lifetime profiles after annealing at different
temperatures of a reference edge block.
1158
22nd European Photovoltaic Solar Energy Conference, 3-7 September 2007, Milan, Italy
At [Fe] of about 1013 atoms/cm3 clustering starts to form
in the red zone near the edge of the sample. It is assumed
that these clusters are electrically less active. This is
supported by the increase of lifetime near the edge in red
zone. Most likely iron silicide clusters are formed.
3.5 NAA analyses
Neutron activation analyse has been used to
determined the bulk impurity level. Samples from two
blocks have investigated one centre block from ES and
one reference edge block made from poly silicon. The
results
from
the
study
are
shown
in
Table 5.
Low temperature annealing improves the material in
certain regions. However P-gettering overall gives better
results.
Except for the [Na] and the [Co], the overall impurity
concentration is higher for the ES material. Please also
note the reference block is also contaminated from the
crucible and coating since it is an edge block. Even
though, the measured lifetime is higher for the ES material.
This may due to presence of other elements causing
internal gettering. These inclusions are currently being
investigated.
By comparing NAA and FeB-pair splitting measurements
only 0.1 % of the present Fe atoms are solved as free
interstitial atoms. Almost all the Fe impurities present are
passivated, bonded around defects or in defect clusters or
are present in form of silicide particles.
5
Table 5 Impurity concentration given in atoms/cm
measured by using NAA.
ESB
ESC
EST
RefB
RefC
RefT
Fe
4.8e14
5.3e14
5.2e14
4.2e14
1.9e14
1.8e14
Cu
4.3e15
4.1e15
4.1e15
4.4e15
2.6e15
3.0e15
Cr
3.8e13
4.9e13
3.7e13
6.6e13
2.5e13
2.9e13
Co
1.2e13
7.3e12
1.8e13
5.5e12
4.1e12
4.6e12
Sb
5.2e12
2.8e12
1.5e12
1.4e12
2.2e12
2.2e12
Zn
5.1e13
4.5e13
6.1e13
4.3e13
2.6e13
3.3e13
As
4e13
5.2e13
7.2e13
1.7e13
2.3e13
4.3e13
Na
3.8e13
4.3e13
6.5e13
4.5e13
5.9e13
3.5e13
4
REFERENCES
3
[1] C. Zahedi. et. al., "Solar grade silicon from
metallurgical route " PVSEC-14, Bangkok, (2004)
26-30
[2] A. Bentzen, A. Ulyashin. A. Suphellen, E. Sauar, D.
Grambole, D.N. Wright, E.S. Marstein, B.G.
Svensson, and A. Holt, "Surface passivation of
silicon solar cells by amorphous silicon/silicon nitride
dual layers", Technical Digest of the 15th
International Photovoltaic Science and Engineering
Conference, (2005) 31
[3] A.G. Ulyashin, A. Bentzen, S. Diplas, A.E.
Gunnaes, A. Olsen, B.G. Svensson, A. Suphellen,
E.S. Marstein, A. Holt, D. Grambole, E. Sauar, "
Hydrogen release and defect formation during heat
treatments of SiNx:H/a-Si:H double passivation layer
on c-Si substrate" Proc. WCPEC-4 , IEEE,
(2006)1354-1357
CONCLUSIONS
The results of the minority carrier lifetime
measurements show higher minority carrier lifetime of
ES-wafers compared with results from wafers based on
poly silicon material.
Furthermore a P-gettering
processes has successfully been introduced, showing on
average a further improvement by a factor larger than two
times, shown in Figure 11.
[4] A. Bentzen, E. S. Marstein, R. Kopecek, and A. Holt,
"Phosphorus diffusion and gettering in multicrystalline silicon solar cell processing " Proceedings
of the 19th European Photovoltaic Solar Energy
Conference, Paris, France, (2004) 935
[5] J. Tan, D. Macdonald, N. Bennett, D. Kong, and A.
Cuevas, " Dissolution of metal precipitates in
multicrystalline silicon during annealing and the
protective effect of phosphorus emitters" Applied
Physics Letter P91 (2007) 043505
[6] A. Bentzen, A. Holt, R. Kopecek, G. Stokkan, J. S.
Christensen, and B. G. Svensson, "Gettering of
transition metal impurities during phosphorus emitter
diffusion in multicrystalline silicon solar cell
processing" , Journal of Applied Physics, 99:093509,
2006
[7] P. Manshanden and L. Geerligs, "Improved
phosphorous gettering of multicrystalline silicon",
Sol. Energy Mater. Sol. Cells 90 (2006) 998
[8] A.A. Istratov, T. Buonassisi , M.D. Pickett , M.
Heuer, E.R. Weber, "Control of metal impurities in
“dirty” multicrystalline silicon for solar cells",
Materials Science and Engineering B 134 (2006)
282–286
Figure 11: Bulk lifetime of minority carriers before and
after gettering at different positions of both the ES and
reference edge block.
1159