Solid State Phenomena Vols. 95-96 (2004) pp 187-196
Online available since 2003/Sep/30 at www.scientific.net
© (2004) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/SSP.95-96.187
Light-induced Degradation in Crystalline Silicon Solar Cells
Jan Schmidt
Institut für Solarenergieforschung Hameln/Emmerthal (ISFH),
Am Ohrberg 1, D-31860 Emmerthal, Germany, Email: j.schmidt@isfh.de
Keywords: Silicon Solar Cell, Degradation, Carrier Lifetime, Metastability, Czochralski Silicon,
Multicrystalline Silicon
Abstract. Solar cells manufactured on single-crystalline boron-doped Czochralski-grown silicon
(Cz-Si) degrade in efficiency by up to 10% (relative) when exposed to light or minority carriers are
injected in the dark until a stable level of performance is reached. This effect, which is now known
for 30 years, is due to the activation of a specific metastable defect in the silicon bulk. Although a
conclusive explanation of the effect is still to be found, recent investigations have clearly revealed
that the metastable defect is correlated with the boron and the oxygen concentration in the material.
In block-cast multicrystalline silicon (mc-Si) the same degradation behavior can be observed, but
due to the lower oxygen content of mc-Si compared to Cz-Si, the degradation occurs to a lesser
extent. The first part of this paper reviews the current status of the physical understanding of the
degradation effect and gives an overview of the defect models proposed in the literature. In the
second part, an overview of different strategies for avoiding or reducing the degradation is
presented.
Introduction
93% of the present world solar cell production is based on boron-doped crystalline silicon, with
Czochralski-grown monocrystalline silicon (Cz-Si) having a market share of about 36%, block-cast
multicrystalline silicon (mc-Si) having a share of 52%, and ribbon-grown silicon sheets amount to
5% [1]. The main problem of solar cells manufactured on Cz-Si is that their initial efficiency
degrades under illumination until a stable performance level, well below the initial efficiency, is
reached. In the case of high-efficiency laboratory solar cells, such as the PERL cell, the efficiency
was found to degrade by up to 10% relative [2], while in commercially manufactured solar cells an
efficiency degradation by typically 3-7% (relative) has been reported [3]. Despite the fact that the
efficiencies obtained on monocrystalline Cz-Si are initially much higher compared to those attained
on the cheaper cast mc-Si materials, efficiencies closely approach after a few hours of illumination.
Despite the fact that mc-Si solar cells show, in principle, the same degradation effect [4], it occurs
to a much lesser extent than in Cz-Si cells. Since, in general, Cz-Si is more expensive than mc-Si,
the future of solar-grade Cz-Si crucially depends on whether it is usable for the mass-production of
high-efficiency solar cells or not. Hence, in recent years, a lot of research has been devoted to the
light-induced degradation problem, which is presently the main obstacle for making solar-grade CzSi a perfect high-efficiency solar cell material. This paper discusses the present physical
understanding of the degradation effect and gives an overview of the different approaches for
reducing or even completely avoiding it.
Fundamental Understanding
Discovery of the Effect and Early Models. To our knowledge, the first observation of lightinduced degradation in non-particle-irradiated solar cells fabricated on boron-doped Cz-Si wafers
was made by Fischer and Pschunder in 1973 [5]. They recorded a strong degradation of the shortcircuit current density and the open-circuit voltage during the first few hours of illumination until a
stable level was reached. Interestingly, the initial cell performance could be completely recovered by
All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of TTP,
www.ttp.net. (ID: 130.203.136.75, Pennsylvania State University, University Park, United States of America-04/06/14,04:47:03)
188
Gettering and Defect Engineering in Semiconductor Technology X
a low-temperature anneal at only 200°C. Additional photoconductance decay measurements
indicated that the observed effect was due to a bulk carrier lifetime varying between two levels,
corresponding to two different states, A and B, of the material. State A is associated with a high
lifetime and requires low-temperature annealing, while state B is associated with a low carrier
lifetime and is caused by illumination. Both levels were found to have the tendency to saturate and
can reversibly be changed by applying the appropriate treatment [5].
While the study performed by Fischer and Pschunder was more of phenomenological nature, in the
following years, several attempts were made to develop a defect model which explains the observed
lifetime instabilities in boron-doped Cz-Si solar cells. Weizer et al. [6] proposed a model attributing
the effect to a complex of a lattice defect and a silver atom or a cluster of atoms. Graff and Pieper [7]
proposed a vacancy-gold complex as the lifetime-limiting recombination center in the material.
However, none of these models was capable of explaining the complete degradation/recovery cycle
observed in boron-doped Cz silicon. Corbett et al. [8] noted that the lifetime variations in Cz silicon
are probably not due to a direct creation of the metastable defect by photons and suggested that the
lifetime degradation might be due to the dissociation of donor-acceptor defect pairs caused by excess
carriers via a recombination-enhanced mechanism. A recombination-enhanced defect reaction was
believed to be the responsible physical mechanism for the defect formation, as it explains the
important experimental observation that the degradation of the carrier lifetime occurs not only under
light exposure, but also in the dark if a forward-bias voltage is applied [2,6,9].
One of the probably best studied donor-acceptor complexes in crystalline silicon is the iron-boron
pair. This defect pair dissociates under illumination and interstitial iron is formed [10]. Interstitial
iron is, under low-injection conditions, a more effective recombination center than the iron-boron
pair, and hence leads to a strong degradation in the carrier lifetime. Reiss et al. [11] investigated solar
cells fabricated on boron-doped Cz silicon wafers with comparatively high iron contamination levels
and hence very low bulk lifetimes (~ 4 ms). They observed a pronounced degradation in cell
performance during illumination, during the application of a forward-bias voltage in the dark, and
during thermal treatment above ~250°C. In order to recover the cell efficiency, the cell had to be
stored in the dark at temperatures below 100°C. The authors clearly showed that this behavior can
undoubtedly be explained by iron-boron pairs.
The annealing behavior of the material investigated by Reiss et al. [11] is very different to the one
reported by Fischer and Pschunder [5] who measured a complete lifetime recovery during annealing
at temperatures above about 200°C and no lifetime recovery when storing the cells at room
temperature in the dark. The latter researchers investigated Cz silicon with negligible iron content.
Hence, it is very likely that the fundamental degradation effect reported in Ref. 5 was screened in the
solar cells of Reiss et al. by the superimposed iron-related effect. At this point it should be noted that
several inconsistencies in the literature are due to the fact that often no clear distinction between both
effects is made. The present paper only deals with the fundamental degradation observed even in
highest-purity electronic-grade boron-doped Cz-Si with virtually no metal contamination.
Structure of the Metastable Defect and Transformation Mechanism. While all former
attempts to clarify the light-induced degradation of the carrier lifetime in Cz-Si were based on the
formation of metal-containing defect complexes, it was not until 1997 that a complete defect
reaction model was proposed by Schmidt et al. [12] that did not involve any metallic impurities.
This was also the first model capable of explaining the lifetime degradation under illumination (or
minority-carrier injection in the dark) as well as the lifetime recovery during annealing at
temperatures above ~200°C. The most important feature of this model is the formation of a defect
pair composed of one interstitial boron and one interstitial oxygen atom (BiOi) during minoritycarrier injection. The qualitative correlation of the magnitude of degradation with the boron and the
oxygen concentration in the material was demonstrated by means of carrier lifetime measurements
Solid State Phenomena Vols. 95-96
189
0°
C
50
°C
25
0°
20 C
0°
C
15
0°
C
10
0°
C
performed before and after light-soaking on Cz wafers with different boron and oxygen
concentrations [12]. Furthermore, photoconductance decay measurements on gallium-doped p-type
Cz silicon and on phosphorus-doped n-type Cz material did not show any lifetime degradation [12]
Based on these results, several methods for reducing the lifetime degradation in Cz silicon solar
cells were proposed, which will be discussed in detail below. According to recent theoretical
considerations of Ohshita et al. [13], the BiOi pair could only exist in a stable configuration if a
substitutional silicon atom is sited between the boron and the oxygen atom.
In 1998, Glunz et al. [14] confirmed the strong correlation between the lifetime degradation in
Cz-Si and the boron as well as the oxygen concentration by means of lifetime measurements on a
large number of boron-doped Cz-samples. Whereas they found an approximately proportional
increase of the lifetime degradation with boron doping concentration, a strongly superlinear increase
with interstitial oxygen concentration, approximately to the power of five, was observed. It is
important to note that, in the study of Glunz et al. [14], surface recombination of the Cz-Si wafers
was suppressed by a high-temperature oxidation step at 1050°C and a subsequent deposition of
corona charges on top of the thermally grown silicon dioxide layer, while in the upper study of
Schmidt et al. [12] silicon nitride films deposited at low temperature (375°C) were used for surface
passivation. The results of Glunz et al. [14] gave rise to the suspicion that the light-induced
recombination center is probably associated with a defect complex different from the BiOi pair.
Besides, it is questionable if any interstitial boron exists in non-particle-irradiated silicon at all.
These considerations were supported by measurements of Schmidt and Cuevas [15] using injectiondependent lifetime spectroscopy (IDLS). They showed that the energy level Et of the light-induced
recombination center is relatively close to the intrinsic Fermi level Ei of silicon ( Et-Ei < 0.2 eV)
and hence it is very different from that of the BiOi pair, lying at Et = EC - 0.26 eV as known from
deep-level transient spectroscopy measurements carried out on electron-irradiated Cz-Si [16]. On
the basis of these results, a new structure consisting of one substitutional boron Bs and several
oxygen atoms was proposed [15]. Recently, Rein et al. [17,18] determined the energy level of the
metastable defect (see Fig. 1) using temperature-dependent lifetime spectroscopy (TDLS). Because
in the relevant temperature range the increase in lifetime with temperature is superimposed by the
defect annihilation, the latter process was suppressed by illuminating the sample between each
measurement point. The accuracy of their measurements was shown to be very high and an energy
B-doped Cz-Si (3.5 Wcm)
0
LLI
t eff / T (µs/K)
10
10
Lifetime Measurement
Linear Fit
Fit to Entire TDLS Curve
Et = EC - 0.4 eV
-1
2.0
2.5
3.0
3.5
4.0
1000/T (1/K)
Figure 1: Determination of the energy level Et
of the metastable defect by means of temperature
dependent lifetime spectroscopy (TDLS) (data
taken from Ref. 18).
Figure 2: Stable (S) and metastable (M)
configuration of the boron-oxygen
complex proposed by Bourgoin et al. (from
Ref. 19).
190
Gettering and Defect Engineering in Semiconductor Technology X
10-1
10-1
17
-3
3
cm /µs)
[Oi] = (7.5 ± 0.5) ´ 10 cm
-16
10-2
Nt*/ Ndop (10
Normalized Defect Concentration Nt* (1/µs)
level of Et = EC - 0.4 eV was determined for the metastable defect (see Fig. 1).
Bourgoin et al. [19] proposed a possible atomic configuration of the boron-oxygen complex, in
which the Bs atom is surrounded by three Oi atoms (see Fig. 2). They also suggested a new
degradation mechanism where electron trapping induces a Jahn-Teller distortion, shifting one of the
energy levels of the defect to the middle of the silicon band gap. However, an experimental
verification of the proposed mechanism was not given.
In a recent lifetime study, it was shown that the defect generation rate is virtually independent of
the light intensity above 1/100 sun [17]. Similar results were obtained by Hashigami et al. [20] on
solar cells. Interestingly, at very low intensities (< 1/100 sun) a proportional increase of the defect
generation rate with intensity was observed [21]. The latter result implies a proportionality between
defect generation rate and photo-generation rate, suggesting a recombination-enhanced defect
formation mechanism because the photo-generation rate is directly proportional to the
recombination rate. This interpretation is also consistent with the results of temperature-dependent
bias-voltage induced solar cell degradation experiments [22].
Temperature-dependent measurements of the defect generation rate have revealed that the defect
generation is in fact a thermally activated process with a relatively low barrier energy of Egen = 0.4
eV [23]. This finding points towards a diffusion-limited defect formation process, where the
minority-carrier injection plays merely an indirect role.
The defect annihilation process has been shown to be thermally activated as well [17,23].
Isothermal annealing experiments were performed by different groups using different experimental
approaches and revealed barrier energies in the vicinity of Eann = 1.3 eV [17,23,24], well above the
barrier energy of the defect generation process of Egen = 0.4 eV.
~[Bs]
10-3
1015
1016
Substitutional Boron Concentration [Bs] (cm-3)
Figure 3: Normalized defect concentration Nt*
as a function of the substitutional boron
concentration [Bs] (data taken from Ref. 21).
~[Oi]1.9
10-2
4
5
6
7
8 9 10
Interstitial Oxygen Concentration [Oi] (1017 cm-3)
Figure 4: Normalized defect concentration Nt*
on the interstitial oxygen concentration [Oi]
(data taken from Ref. 21).
Most recently, the impact of boron and oxygen on the metastable defect concentration in Cz-Si
was re-examined on a very large number of Cz-Si samples [21,23,24] using the quasi-steady-state
photoconductance method [25], enabling the accurate measurement of carrier lifetimes at defined
injection levels. Figure 3 shows the measured normalized defect concentration Nt*, determined by
subtracting the inverse lifetimes after and before light soaking Nt* º 1/td -1/t0, as a function of the
substitutional boron concentration [Bs] for Cz-Si wafers with similar levels of oxygen
contamination [Oi] = (7-8)´1017 cm-3. In perfect agreement with previous studies [14,15], a
proportional increase of Nt* with [Bs] was found. As exactly the same dependence was measured in
several independent studies, the proportionality between Nt* and [Bs] can now be regarded as well
Solid State Phenomena Vols. 95-96
191
established. Figure 4 shows the measured Nt* values divided by the corresponding boron doping
concentration Ndop as a function of the interstitial oxygen concentration [Oi]. Plotting Nt*/Ndop
instead of Nt* makes it possible to include Cz-Si wafers of different doping concentrations. The
most striking result of Fig. 4 is the fact that Nt* shows an approximately quadratic increase with
[Oi], which is a much weaker dependence compared to that found by Glunz et al. [14]. This
discrepancy might be due to the different measurement techniques applied and/or the fact that in
Ref. 14 a high-temperature oxidation, which may change the silicon bulk properties, was performed
to passivate the surfaces, whereas in Ref. 23 low-temperature silicon nitride films were used for
surface passivation. Furthermore, the scatter in the data of Glunz et al. has been much larger,
making these results less reliable. In fact, very recently Rein et al. confirmed the quadratic
dependence of Nt* on the interstitial oxygen concentration [26]. The measured dependencies on the
boron and oxygen concentrations clearly suggest that the metastable defect is composed of one
substitutional boron and two interstitial oxygen atoms. Interestingly, for Oi concentrations below
~4´1017 cm-3 a weaker, approximately linear dependence was measured [24]. The wafers used for
these measurements have been grown by the magnetic-field assisted Cz method (MCz). It cannot be
excluded that the different growth technique might be responsible for this unusual behavior.
Emig
Si
Oi
Ediss
Bs
Ebind
O2i
Figure 5: Energy diagram of the Bs-O2i
interaction (from Ref. 23).
Bs
Oi
Figure 6: Possible configuration of the
metastable Bs-O2i complex in the silicon host
lattice (from Ref. 21).
Based on the measured linear Nt*([Bs]) and quadratic Nt*([Oi]) dependencies, a new defect
reaction model was developed by Schmidt et al. [23]. In this model, fast-diffusing oxygen dimers
O2i are captured by substitutional boron Bs to form a Bs-O2i complex, acting as highly effective
recombination center (see Fig. 5). Due to the fact that the tetrahedral covalent radius of the Bs atom
is 25% smaller than that of the Si host atom [27], the O2i is preferentially accommodated in the
vicinity of a Bs atom. In the case of Ga- and In-doped Cz-Si, no degradation was observed [21],
which according to the oxygen-dimer model is due to the larger tetrahedral covalent radii of Ga and
In compared to Si, making it less likely to accommodate the oxygen dimer. The defect formation
process in B-doped Cz-Si is governed by the diffusion of the oxygen dimer and, hence, is a
thermally activated process, in good agreement with the experimental results in Ref. 23. Figure 6
shows an estimated configuration of the metastable Bs-O2i complex, which is based on theoretical
calculations of the stable O2i configuration [28].
Two mechanisms were proposed concerning the role of the illumination or, generally speaking,
the minority-carrier injection, as the degradation also occurs in the dark if a forward-bias voltage is
applied to a pn junction. The minority carriers are in both cases not directly involved in the defect
reaction, but their presence triggers the reaction by (i) changing the charge state of the O2i, which
might lead to an increase in the O2i diffusivity, or (ii) enhancing the O2i diffusivity via a
recombination-enhanced change of the configuration [21,22,29]. Recent voltage- and temperaturedependent measurements of the dark degradation behavior of Cz-Si solar cells have unambiguously
192
Gettering and Defect Engineering in Semiconductor Technology X
annealed
100
light-soaked
Carrier Lifetime (µs)
1000
10
B-dop.Cz-Si
1.0 Wcm
[Oi] = 14 ppm
Ga-dop.Cz-Si
1.2 Wcm
[Oi] = 19 ppm
P-dop.Cz-Si
0.6 Wcm
[Oi] = 16 ppm
B-dop. MCz-Si
1.2 Wcm
[Oi] = 1.5 ppm
Figure 7: Measured carrier lifetimes of of B-, Ga-, and P-doped Cz-Si and B-doped MCz-Si
before and after light soaking (data taken from Refs. 12, 36).
excluded mechanism (i) [22]. On the other hand, mechanism (ii) would be in good agreement with
the observed proportionality between defect generation rate and photo-generation rate observed at
low light intensities (< 0.01 suns). The observed saturation in the defect generation rate for light
intensities above 0.01 suns may be attributed to the relatively low concentration of O2i in the silicon
lattice. The experimental observations would support the hypothesis that the energy set free during a
recombination event is used to bring the O2i in a configuration in which it has an increased
diffusivity.
The proposed oxygen-dimer model is similar to the thermal donor (TD) formation mechanism
first proposed by Gösele and Tan [30], in which the O2i molecules are captured by other oxygen
atoms or clusters to form the different TD levels. In fact, a detailed theoretical and experimental
analysis [31] shows that during the first few hours of TD formation, the concentration of the O2i
dimer shows a pronounced decrease. As less O2i molecules are available, the model proposed in
Ref. 23 predicts a decrease in the Bs-O2i concentration and hence a reduced light degradation of the
carrier lifetime after TD formation. In order to verify this hypothesis, Bothe et al. [32] have
annealed different boron-doped Cz-Si wafers at 450°C for up to 32 h. These conditions are ideal for
the formation of TDs. After TD formation, they measured a pronounced reduction in the metastable
defect concentration by up to a factor of 3, giving an indirect confirmation of the proposed dimer
model.
Strategies for Suppressing the Degradation
Alternative Cz-Si Materials. The direct correlation of the magnitude of degradation with the boron
and the oxygen concentration in a crystalline silicon wafer has been proven by means of carrier
lifetime measurements on a very large number of Cz-Si wafers from different manufacturers
[12,14,21,23,24,26]. Furthermore, measurements on Ga-doped p-type Cz silicon as well as on Pdoped n-type Cz material have shown no degradation of the carrier lifetime [12]. Based on these
experimental results, several methods for reducing the lifetime degradation in Cz silicon solar cells
were proposed [12]. The two most promising approaches proposed in Ref. 12 were: (i) replacement
of B with another dopant element, such as Ga or P, and (ii) reduction of the oxygen concentration in
the Cz material. In a recent international joint research project, organized by Saitoh [33], a large
number of Cz-Si wafers with different dopant elements (B, Ga) and different oxygen concentrations
(including conventional and magnetic-field assisted Cz-Si) were manufactured by Shin-Etsu Handotai
(SEH) and supplied to various international institutions (Sharp Co., Hitachi Ltd., FhG-ISE, Georgia
Tech, ISFH, UNSW, TUAT) for characterization and solar cell fabrication. Figure 7 shows the typical
Solid State Phenomena Vols. 95-96
193
FZ(B)
Cz(P)
Cz(Ga)
FZ(B)
MCz(B)
FZ(B)
Cz(Ga)
21
MCz(B)
22
Cz(B)
20
19
18
17
Cz(B)
Efficiency (%)
23
Cz(Ga)
24
MCz(B)
25
RP-PERC solar cells OECO solar cells
FhG-ISE
ISFH
PERT solar cells
UNSW
Figure 8: Stable efficiencies of solar cells manufactured on alternative Cz-Si and FZ-Si (gray
boxes). For comparison, cell efficiencies obtained on conventional B-doped Cz-Si before (white
boxes) and after (black boxes) light degradation are shown (data taken from Refs. 34-39).
behavior of the different materials: while the carrier lifetime of conventional B-doped Cz-Si (solargrade as well as electronic-grade material) degrades under illumination, Ga-doped Cz-Si of similar
doping concentration has a stable lifetime on a much higher level, comparable to that of B-doped
float-zone (FZ) silicon, even if the interstitial oxygen concentration is approximately as high as in the
case of the B-doped Cz material. The high oxygen concentration in silicon ingots grown with a
conventional Cz single crystal puller is due to the partial dissolution of the silica crucible during the
growth process. The oxygen content can be strongly reduced by damping the melt flows with
magnetic fields. This so-called magnetic-field assisted Cz (MCz) silicon with strongly reduced Oi
concentration shows an almost vanishing light degradation (see Fig. 7) and has bulk lifetimes which
are comparable to that of Ga-doped Cz-Si. Also included in Fig. 7 is a P-doped Cz-Si material with
high Oi content that is also completely stable under illumination. However, since the latter material is
an n-type semiconductor and most industrial solar cell processes have been developed for p-type base
material, at present the most promising alternative Cz silicon materials seem to be Ga-doped Cz-Si
and B-doped MCz-Si.
High-efficiency solar cell processes were applied to the alternative Cz materials at different
institutes and stable efficiencies well above 20% were obtained on Ga-doped Cz-Si, B-doped MCz
material, and P-doped n-type Cz-Si (see Fig. 8) [34-39]. For comparison, cell results obtained on Bdoped FZ-Si and B-doped Cz-Si are also included in Fig. 8. Solar cell efficiencies above 20% can
only be achieved with the new Cz materials, while the stable efficiency of cells fabricated on
conventional 1-Wcm B-doped Cz-Si is always well below 20%. Note that the degree of complexity of
the three cell processes compared in Fig. 8 is very different, with the photolithography-based PERT
(passivated emitter rear totally diffused cell) process being by far the most complex manufacturing
process, whereas the OECO process, on the other hand, uses obliquely evaporated contacts and is
completely avoiding any photolithography and alignment steps. The latter process has already
successfully been transferred to industrial pilot production.
The segregation coefficient of Ga in Si is two orders of magnitude lower compared to that of B in
Si. Hence, Ga-doped Cz-Si crystals exhibit a considerably higher variation in resistivity along their
growth axis compared to B-doped crystals. Metz et al. have investigated the usability of a complete
6” Ga-doped Cz-Si crystal in the OECO solar cell process [37]. The resistivity of the crystal varied
from 1.3 Wcm at the top to 0.4 Wcm in the tail region. Figure 9 shows the measured OECO solar cell
efficiencies as a function of the base resistivity. Peak efficiencies of up to 21% were obtained on 0.4Wcm material and, more important, in the broad resistivity range between 0.25 and 1.34 Wcm, cell
194
Gettering and Defect Engineering in Semiconductor Technology X
22
Efficiency [%]
21
2
STC-conditions, aperture: 4 cm ,
after 24h illumination
20
Resistivity range of
Ga-doped crystal
19
18
Ga-doped Cz Si
B-doped Cz Si
B-doped FZ Si
17
16
0.1
1
Resistivity [W cm]
Figure 9: Measured efficiencies of OECO solar cells manufactured on Ga-doped Cz-Si as a
function of base resistivity (from Ref. 37).
efficiencies were found to reach more than 97% of the peak value, demonstrating that the inherent
resistivity variations in Ga-doped Cz-Si crystals are well within the range tolerable for the
manufacturing of high efficiency solar cells.
Process Optimization. Several approaches aiming at reducing the concentration of the metastable
defect in B-doped Cz-Si during the solar cell manufacturing process have been investigated. In
particular, certain high-temperature steps, optimized for Cz-Si, were found to be capable of
significantly reducing the magnitude of degradation. In a recent comprehensive study, the hightemperature (~1050°C) thermal oxidation process required for the growth of masking oxides and
passivation layers was studied in detail [14,40]. It was found that the concentration of the metastable
defect can be reduced by up to a factor of 4 if the ramping conditions are chosen properly. However,
the absolute values of the stable lifetimes of conventional B-doped Cz-Si materials were still found to
be on a relatively low level between about 20 and 45 µs after the optimized oxidation step, which is
well below the lifetimes measured on Ga-doped Cz-Si and MCz materials. In a more recent study, a
phosphorus emitter diffusion step at ~850°C with optimized ramping conditions was also found to
have a beneficial effect on the magnitude of degradation, leading to a maximum reduction in the
light-induced defect concentration by a factor of 3.5 [32]. This effect was attributed to the
temperature profile only and was shown to be not due to a gettering of impurities [32]. A permanent
improvement of the carrier lifetime in Cz-Si is not only possible using conventional quartz-tube
furnaces. A short annealing step of a few seconds at temperatures around 800°C in a belt furnace was
found to halve the light-induced defect concentration [41]. Similar results were obtained using rapid
thermal processing [42]. It has also been conjectured that hydrogenation might reduce the degradation
[43], but definite experimental evidence is still lacking.
Other Approaches. An alternative method to reduce the harmful effects of lifetime degradation
on cell efficiency is to modify the cell design. By simply reducing the thickness of their solar cells to
100 µm, Münzer et al. [44] obtained a reduced degradation and an improved stable efficiency.
However, this approach requires a very efficient rear surface passivation of the solar cell [44,45].
Other solar cell structures, such as the BACK-OECO cell [46] or the Emitter-Wrap-Through (EWT)
cell [47], have also the potential to reduce the light degradation considerably.
On the basis of the oxygen-dimer model proposed in Ref. 23, it was suggested that carbon-rich Bdoped Cz-Si may exhibit a reduced lifetime degradation, because the formation of Cs-O2i complexes
could be a competitive process to the formation of the lifetime-limiting Bs-O2i complex. However, an
experimental verification of this promising new way of reducing the metastable defect concentration
in B-doped Cz-Si has not been produced up to now.
Solid State Phenomena Vols. 95-96
195
Conclusions
The concentration of the metastable boron-oxygen complex in crystalline silicon depends
proportionally on the substitutional boron concentration and quadratically on the interstitial oxygen
content. This important finding together with results from measurements of the defect generation and
annihilation kinetics support a recently proposed defect model based on fast-diffusing oxygen dimers,
which are captured by substitutional boron. Several approaches to reduce or even completely avoid
the degradation have been studied. Alternative dopants, such as gallium or phosporus, have been
demonstrated to completely avoid the degradation. Reducing the oxygen content by applying a
magnetic field during crystal growth is another way of improving the stable cell efficiency. In Bdoped Cz-Si material, thermal treatment can significantly reduce the magnitude of degradation. In
combination with alternative solar cell structures, being currently developed in several labs, strongly
improved stable efficiencies may also be expected in the near future on standard B-doped Cz-Si.
Acknowledgements
The author thanks R. Hezel for his continuous support and encouragement and K. Bothe for his
valuable contributions. Funding was provided by the State of Lower Saxony and the German
Ministry of Education and Research (BMBF) under contract no. 01SF0009. The ISFH is a member
of the German Forschungsverbund Sonnenenergie.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
M. Schmela, Photon International March 2003, p. 42.
J. Knobloch, S. Glunz, V. Henninger, W. Warta, W. Wettling, F. Schomann, W. Schmidt, A. Endrös,
K. Münzer, Proc. 13th European Photovolt. Solar Energy Conf. (Stephens, Bedford, 1995), p. 9.
S. Sterk, K. Münzer, and S. Glunz, Proc. 14th European Photovolt. Solar Energy Conf. (Stephens,
Bedford, 1997), p. 85.
H. Nagel, J. Schmidt, A.G. Aberle, and R. Hezel, Proc. 14th European Photovolt. Solar Energy Conf.
(Stephens, Bedford, 1997), p. 762.
H. Fischer and W. Pschunder, Proc. 10th IEEE Photovolt. Spec. Conf. (IEEE, New York, 1973), p.
404.
V.G. Weizer, H.W. Brandhorst, J.D. Broder, R.E. Hart, and J.H. Lamneck, J. Appl. Phys. 50, 1979, p.
4443.
K. Graff and H. Pieper, phys. stat. sol. (a) 49, 1978, p. 137.
J. W. Corbett, A. Jaworowski, R.L. Kleinhenz, C.B. Pierce, and N.D. Wilsey, Solar Cells 2, 1980, p.
11.
J. Knobloch, S.W. Glunz, D. Biro, W. Warta, E. Schäffer, and W. Wettling, Proc. 25th IEEE
Photovolt. Spec. Conf. (IEEE, New York, 1996), p. 405.
K. Graff and H. Pieper, J. Electrochem. Soc. 128, 1981, p. 669.
J.H. Reiss, R.R. King, and K.W. Mitchell, Appl. Phys. Lett. 68, 1996, p. 3302.
J. Schmidt, A.G. Aberle, and R. Hezel, Proc. 26th IEEE Photovolt. Spec. Conf. (IEEE, New York,
1997), p. 13.
Y. Ohshita, T. Khanh Vu, and M. Yamaguchi, J. Appl. Phys. 91, 2002, p. 3741.
S. W. Glunz, S. Rein, W. Warta, J. Knobloch, and W. Wettling, Proc. 2nd World Conf. Photovolt.
Solar Energy Conv. (EC, Ispra, 1998), p. 1343.
J. Schmidt and A. Cuevas, J. Appl. Phys. 86, 1999, p. 3175.
L.C. Kimmerling, M.T. Asom, J.L. Benton, P.J. Drevinsky, and C.E. Caefer, Mat. Sci. For. 38-41,
1989, p. 141
S. Rein, T. Rehrl, W. Warta, S. Glunz, and G. Willeke, Proc. 17th European Photovolt. Solar Energy
Conf. (WIP-ETA, Munich, 2001), p. 1555.
S. Rein and S. Glunz, Appl. Phys. Lett. 82, 2003, p. 1054.
J. C. Bourgoin, N. de Angelis, and G. Strobl, Proc. 16th European Photovolt. Solar Energy Conf.
(James & James, London, 2000), p. 1356.
196
Gettering and Defect Engineering in Semiconductor Technology X
[20] H. Hashigami, Y. Itakura, A. Takaki, S. Rein, S. Glunz, and T. Saitoh, Proc. 17th European
Photovolt. Solar Energy Conf. (WIP-ETA, Munich, 2001), p. 1483.
[21] J. Schmidt, K. Bothe, and R. Hezel, Proc. 3rd World Conf. Photovolt. Energy Conv., 2003, in press.
[22] S. Glunz, E. Schäffer, S. Rein, K. Bothe, and J. Schmidt, Proc. 3rd World Conf. Photovolt. Energy
Conv., 2003, in press.
[23] J. Schmidt, K. Bothe, and R. Hezel, Proc. 29th IEEE Photovolt. Spec. Conf. (IEEE, New York, 2002),
p. 178.
[24] K. Bothe, J. Schmidt, and R. Hezel, Proc. 3rd World Conf. Photovolt. Energy Conv., 2003, in press.
[25] R. A. Sinton and A. Cuevas, Appl. Phys. Lett. 69, 1996, p. 2510.
[26] S. Rein, S. Diez, R. Falster, and S. Glunz, Proc. 3rd World Conf. Photovolt. Energy Conv., 2003, in
press.
[27] F. Shimura, Semiconductor Silicon Crystal Technology (Academic Press, San Diego, CA, 1989).
[28] C. P. Ewels, Density Functional Modelling of Point Defects in Semiconductors, PhD. Thesis,
University of Exeter, UK, 1997.
[29] J. D. Weeks, J. C. Tully, and L. C. Kimerling, Phys. Rev. B 12, 1975, p. 3286.
[30] U. Gösele and T. Y. Tan, Appl. Phys. A 28, 1982, p. 79.
[31] D. Åberg, B. G. Svensson, T. Hallberg, and J. L. Lindström, Phys. Rev. B 58, 1998, p. 12944.
[32] K. Bothe, J. Schmidt, and R. Hezel, Proc. 29th IEEE Photovolt. Spec. Conf. (IEEE, New York, 2002),
p. 194.
[33] T. Saitoh, H. Hashigami, X. Wang, T. Abe, T. Igarashi, S. Glunz, S. Rein, W. Wettling, A. Ebong,
B.M. Damiani, A. Rohatgi, I. Yamasaki, T. Nunoi, H. Sawai, H. Ohtuka, T. Warabisako, J. Zhao, M.
Green, J. Schmidt, A. Cuevas, A. Metz, and R. Hezel, Proc. 16th European Photovolt. Solar Energy
Conf., 2000, p. 1206.
[34] S. Glunz, S. Rein, J. Knobloch, W. Wettling, and T. Abe, Prog. Photovolt. 7, 1999, p. 463.
[35] S. Glunz, S. Rein, J. Lee, and W. Warta, J. Appl. Phys. 90, 2001, p. 2397.
[36] A. Metz and R. Hezel, presented at the 1st Workshop on Light Degradation of Carrier Lifetimes in
Cz-Si Solar Cells, Sapporo, Japan, 1999.
[37] A. Metz, T. Abe, and R. Hezel, Proc. 16th European Photovolt. Solar Energy Conf. (James & James,
London, 2000), p. 1189.
[38] J. Zhao, A. Wang, and M. Green, Proc. 16th European Photovolt. Solar Energy Conf. (James &
James, London, 2000), p. 1100.
[39] J. Zhao, A. Wang, and M. Green, Prog. Photovolt. 8, 2000, p. 549.
[40] S. Glunz, S. Rein, W. Warta, J. Knobloch, and W. Wettling, Sol. Energ. Mat. Sol. Cells 65, 2001, p.
219.
[41] H. Nagel, A. Merkle, A. Metz, and R. Hezel, Proc. 16th European Photovolt. Solar Energy Conf.
(James & James, London, 2000), p. 1197.
[42] J. Lee, S. Peters, S. Rein, and, S. Glunz, Prog. Photovolt. 9, 2001, p. 417.
[43] J. Schmidt and A. Cuevas, Proc. 16th European Photovolt. Solar Energy Conf. (James & James,
London, 2000), p. 1193.
[44] K. Münzer, K. Holdermann, R. Schlosser, S. Sterk, IEEE Trans. Electron Dev. 46, 1999, p. 2055.
[45] B. Damiani, A. Ristow, A. Ebong, and A. Rohatgi, Prog. Photov. 10, 2002, p. 185.
[46] R. Hezel, Proc. 29th IEEE Photovolt. Spec. Conf. (IEEE, New York, 2002), p. 114.
[47] S. Glunz, J. Dicker, J. Lee, R. Preu, S. Rein, E. Schneiderlöchner, J. Sölter, W. Warta, and G.
Willeke, Proc. 17th European Photovolt. Solar Energy Conf. (WIP-ETA, Munich, 2001), p. 1287.
Gettering and Defect Engineering in Semiconductor Technology X
10.4028/www.scientific.net/SSP.95-96
Light-Induced Degradation in Crystalline Silicon Solar Cells
10.4028/www.scientific.net/SSP.95-96.187
DOI References
[6] V.G. Weizer, H.W. Brandhorst, J.D. Broder, R.E. Hart, and J.H. Lamneck, J. Appl. Phys. 50, 1979, p.
443.
doi:10.1063/1.326437
[8] J. W. Corbett, A. Jaworowski, R.L. Kleinhenz, C.B. Pierce, and N.D. Wilsey, Solar Cells 2, 1980, p.
doi:10.1557/PROC-2-1
[15] J. Schmidt and A. Cuevas, J. Appl. Phys. 86, 1999, p. 3175.
doi:10.1002/9783527613885
[17] S. Rein, T. Rehrl, W. Warta, S. Glunz, and G. Willeke, Proc. 17th European Photovolt. Solar Energy onf.
(WIP-ETA, Munich, 2001), p. 1555.
doi:10.1016/S0927-0248(00)00098-2
[29] J. D. Weeks, J. C. Tully, and L. C. Kimerling, Phys. Rev. B 12, 1975, p. 3286.
doi:10.1103/PhysRevB.12.3286
[34] S. Glunz, S. Rein, J. Knobloch, W. Wettling, and T. Abe, Prog. Photovolt. 7, 1999, p. 463.
doi:10.1002/(SICI)1099-159X(199911/12)7:6<463::AID-PIP293>3.3.CO;2-8
[39] J. Zhao, A. Wang, and M. Green, Prog. Photovolt. 8, 2000, p. 549.
doi:10.1002/1099-159X(200009/10)8:5<549::AID-PIP346>3.0.CO;2-Y
[20] H. Hashigami, Y. Itakura, A. Takaki, S. Rein, S. Glunz, and T. Saitoh, Proc. 17th European Photovolt.
Solar Energy Conf. (WIP-ETA, Munich, 2001), p. 1483.
doi:10.1016/S0927-0248(00)00103-3