BULK PASSIVATION IN SILICON RIBBONS –
A LIFETIME STUDY FOR AN ENHANCED HIGH EFFICIENCY PROCESS
Martin Kaes1, Giso Hahn1, Axel Metz2
1University of Konstanz, Department of Physics, 78457 Konstanz, Germany
2RWE SCHOTT Solar GmbH, Carl-Zeiss-Str. 4, 63755 Alzenau, Germany
ABSTRACT
The effectiveness of hydrogenation either by deposition plus firing of a PECVD SiN layer in a conventional belt furnace or by remote H-plasma was compared quantitatively using spatially resolved lifetime measurements for EFG and String Ribbon. Additionally the effect of a preceding phosphorous gettering on the hydrogenation and the presence of a screen printed rear side aluminum during firing was analyzed. Wafer areas with the presence of a rear side aluminum show additional lifetime improvements for both hydrogenation methods probably due to a gettering effect. With preceding P- gettering hydrogenation by SiN deposition plus firing is superior to remote H-plasma. A synergetic effect of a rear side aluminum as described elsewhere is not obtained.
First high efficiency EFG solar cells using a PECVD SiN fired in a conventional belt furnace were processed with efficiencies in the 17-18 % range.
INTRODUCTION
Evergreen Solar’s String Ribbon (SR) and RWE Schott Solar`s Edge-defined Film-fed Growth (EFG) process have a high potential to significantly reduce Wp
production costs because no material consuming sawing steps are needed after the crystallization process.
Both ribbons have in common that the wafer area can be distinguished in two classes. The one are wafer areas with high minority carrier lifetimes well above 50 μs after cell processing. Large diffusion lengths above cell thickness in these areas do hardly limit the cell efficiencies of solar cells. The other are wafer areas with minority carrier lifetimes in the range of only a few μs.
With diffusion lengths below cell thickness these areas mainly limit cell efficiency. Such areas of higher and lower lifetimes can be traced along the direction of crystallization over several centimeters. For enhanced record efficiencies on these ribbons one could look out for a wafer containing a high lifetim e area that is large enough to place a high efficiency cell (typically 2x2 cm²) into. The more meaningful way is to push the diffusion lengths of the worse wafer quality areas towards the range of the cell thickness and higher. An effective bulk passivation for EFG and SR was demonstrated by Rohatgi et al. [1] with new record efficiencies for EFG
(18.2 %) and String Ribbon (17.9 %) by using plasma enhanced chemical vapor deposition (PECVD) SiN hydrogenation with rapid thermal processing (RTP) firing at 740-750 °C for very short times. But the process applied in [1] led to efficiencies unstable under illumination [2]. On the other hand we could publish stable cell efficiencies of 17.7 % (SR) and 16.7 % (EFG) using a microwave-induced remote hydrogen plasma (MIRHP) step for hydrogenation [3]. In this presentation we try to get more insight in the hydrogenation process of good and lower wafer quality areas in EFG and SR. Therefore the effectiveness of several hydrogenation methods was tested on lifetime level.
LIFETIME TESTS Experimental Set-Up
Four 5x5 cm² wafers were laser cut out of a 12.5x12.5 cm² EFG wafer (3 Ωcm) and four 4x5 cm² out of a 8x15 cm² SR wafer (3 Ωcm). Thus two 5x5 cm² pairs of wafer for each EFG and SR wafer are adjacent in direction of crystallization with comparable crystal properties (figure 1) and they can be compared concerning different process steps. Within one wafer the effect of an Al
Fig. 1: Geometry for the lifetime analysis. Four wafers are cut from EFG and SR wafers.
923
Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-2-t6n538yx0q3b4
Erschienen in: Conference record of the Thirty-First IEEE Photovoltaic Specialists Conference - 2005 : Coronado Springs Resort, Lake Buena Vista, FL, January 3 - 7, 2005. - Piscataway, NJ : IEEE Operations Center, 2005. - S. 923-926. - ISBN 0-7803-8707-4
back surface field (BSF) can be revealed by partly covered rear side screen printing.
The experiment has been carried out with and without a previous P-gettering step (POCl3 diffused, 80-100 Ω/sq).
Wafers have been hydrogenated either by SiN deposition using low frequency PECVD plus conventional belt furnace firing, or by MIRHP (1h at 350°C). The processing sequence for the lifetime analysis is illustrated in figure 2, left. All spatially resolved minority carrier lifetime measurements were carried out using an iodine-ethanol surface passivation with a standard microwave photo conductance decay (μ-PCD) equipment under low level injection conditions. The lifetime measurements were carried out on as-grown wafers and after hydrogenation.
Prior to the second μ-PCD measurement Al-BSF, SiN and emitter have been etched off. The additional Al etching and evaporation in the MIRHP sequence (figure 2, left) is carried out in analogy to the old standard high efficiency process (figure 2, right) [3].
Fig. 2: Process sequence for the lifetime analysis (left) and high efficiency solar cell processes (right).
Lifetime Results
Bulk lifetimes in EFG and SR vary over several orders of magnitude. In a single measurement only a small lifetime range can be reliably detected. Therefore several measurements have to be performed for different lifetime ranges and merged together [4]. Figure 3 shows the result of the as-grown EFG wafers and the result after hydrogenation without preceding phosphorous gettering (compare with figure 1). The as-grown measurement shows typical lifetime behavior of EFG, with elongated grains showing very high initial lifetimes above 30 μs and widespread areas of low lifetimes. It can be stated that the lifetime results for hydrogenation via SiN plus firing and MIRHP are comparable. The positive effect of an
Fig. 3: Bulk lifetimes of the four 5x5 cm² as-grown EFG wafers originating from one 12.5x12.5 cm² wafer (left).
Bulk lifetimes of the hydrogenated EFG wafers (no POCl3, right).
aluminum layer present during the hydrogenation on the lifetimes is demonstrated within a single wafer.
Surprisingly, in absence of aluminum the SiN plus firing passivation in contrast to MIRHP leads only to poor lifetime improvements and even some areas with decreased lifetimes are observed. This could be due to indiffusion of impurities from the metal belt during firing as no Al capping layer is present.
Significantly higher lifetimes can be reached if a P- gettering step precedes hydrogenation as shown in figure 4 for EFG and figure 5 for SR (compare with [4]). Although our applied standard belt furnace firing is “slower” and
“hotter” (~10 s >700°C with typical wafer peak temperatures of about 800°C) compared to the “faster”
and “colder“ RTP firing (1 s at 750°C [1]) a tremendous boost in lifetim es above 40 μs of nearly all lower quality wafer areas can be observed. MIRHP, however, shows only moderate lifetime im provements below 10 μs in the lower quality wafer areas. In addition, we cannot observe a synergetic effect of an Al-layer present at the backside during firing as it was described elsewhere [6]. The SR wafers shows the same behavior in lifetime improvement as the EFG wafers.
Fig. 4: Bulk lifetimes of the four 5x5 cm² as-grown EFG wafers originating from one 12.5x12.5 cm² wafer (left).
Bulk lifetimes of the hydrogenated EFG wafers with preceding phosphorous (POCl3) gettering step (right).
924
Fig. 5: Bulk lifetimes of the four 4x5 cm² as-grown SR wafers originating from one 8x15 cm² wafer (left). Bulk lifetimes of the hydrogenated SR wafers with preceding phosphorous (POCl3) gettering step (right).
Trapping of hydrogen at the low temperatures used for the MIRHP passivation (1h at 350°C) may constrain the hydrogen kinetics resulting in a less effective bulk passivation compared to the SiN plus firing with temperatures above 700°C.
SOLAR CELL PROCESSING MIRHP Sequence
The new insights of the lifetime analysis were transferred to the high efficiency process currently applied to EFG and SR material at University of Konstanz. The altered sequences implemented in the cell process are shown in figure 2, right (compare with [1]).
Using the MIRHP processing sequence cell efficiencies of 11.5 % without antireflective coating could be reached recently on EFG material. This should lead to a cell efficiency of 17.0 % with an optimized ZnS-MgF2
double antireflection coating (DARC). With a better bulk passivation in the SiN sequence enhanced cell efficiencies are expected according to the results of [1]
even though a SiN-MgF2 DARC might lead to higher reflection than a ZnS-MgF2 DARC.
SiN Sequence - First Results
We processed first solar cells using the SiN sequence according to figure 2 with non optim ized firing parameters and standard PECVD SiN (refractive index of 2.0). 1 Ωcm multicrystalline SOLSIX material was used as a reference material.
The SiN layer is now used as hydrogen source for the passivation during firing replacing MIRHP, it provides surface passivation replacing thermal oxidation and it is used as an antireflection coating layer replacing ZnS of the former ZnS/MgF2-DARC. After contact sintering and cutting four 2x2 cm² solar cells out of each processed 5x5 cm² wafer using a dicing saw (0) we performed a MIRHP treatment of the cells for 30 min at 350°C. The observed increase especially in fill factor might be due to a successful hydrogen passivation of the cell’s edge and
Table 1: IV-data of the best EFG (3 Ωcm) and SOLSIX (1 Ωcm) mc cell after different steps. The steps are: (0) cell process, (1) MIRHP, (2) MgF2 deposition (DARC), (3) anneal, (4) light soaking.
Step Voc
[mV]
Jsc
[mA/cm2] FF [%] η
[%]
EFG 0 604 33.9 77.8 16.0
SOLSIX 0 626 33.7 79.4 16.7
EFG 1 606 34.1 78.7 16.2
SOLSIX 1 627 33.7 80.3 17.0
EFG 2 608 36.0 78.8 17.2
SOLSIX 2 626 34.7 80.2 17.4
EFG 3 609 36.0 78.7 17.2
SOLSIX 3 630 35.4 80.2 17.8
EFG 4 609 36.0 78.7 17.2
SOLSIX 4 630 34.8 79.6 17.5
has to be investigated further. The efficiency benefit after evaporation of MgF2 as the second layer of the DARC (2) is restricted to about 1 % absolute because of the SiN refractive index of 2.0 being too low. It is noteworthy that the improvement for the SOLSIX cell is smaller compared to EFG. This seems to be related to a slight degradation effect observed in this SOLSIX cell during the initial IV measurement on a sub-second time scale. This degradation can be temporarily reversed by a thermal anneal at 200°C for 10 min (3). This might be the same effect already reported for Cz material [5]. We could not observe any degradation on EFG after multiple IV measurements (4) and after 24 h under 1 sun illumination.
The excellent EFG material quality after hydrogenation can be seen in figure 6 by the IQE in the long wavelength range (0.83 at 1000 nm). The reflectance loss especially in the wavelength range of 400 to 600 nm is related to the low SiN refractive index of 2.0.
Figure 7 shows the spatially resolved IQE at 980 nm for the best EFG and SOLSIX cells. The dark areas
Fig. 6: IQE measured with bias illumination and reflection of former record cells (EFG1 and SR2, DARC: ZnS/MgF2) at UKN [3] and of the best EFG3 and SOLSIX4 cells (DARC: SiN/MgF2) of our first SiN based high efficiency solar cell process.
925
Fig. 7: Spatially resolved IQE at 980 nm of the best EFG (left) and SOLSIX (right) solar cell (2x2 cm²) without bias illumination.
visible in the EFG cell may lim it the open circuit voltage and might be further improved with optimized firing parameters.
SiN – Influence of refractive index n
Fig. 8: Reflection properties and IQE of EFG solar cells with a DARC using SiN with different refractive indices between 2.0 and 2.2.
Using PECVD SiN with n = 2.0 and MgF2 (n = 1.38) as a double layer antireflective coating the optical properties are not yet satisfying (see figure 6). Therefore we tried to use SiN with a higher n to reach lower reflectivity.
Solar cells of EFG material with 2.2 Ωcm were processed with different refractive indices for the SiN layer. As illustrated in figure 8 the reflection using n = 2.2 can be considerable decreased compared to n = 2.0.
Further on, a higher n could have an influence on surface passivation [7], bulk passivation (because of a lower hydrogen content) and optical absorption. All these effects have to be investigated and evaluated. E.g., the lower IQE for n = 2.2 at wavelengths < 500 nm might be the result of a higher optical absorption in the SiN layer. In a first estimation Jsc can be increased by more than 1 mA/cm² using n = 2.2 instead of n = 2.0 despite of the higher absorption. The decreasing content of hydrogen with increasing refractive index seems not to have an effect on the bulk passivation as the IQE in the wavelength range
above 900 nm is comparably good for all refractive indices used.
SUMMARY
Bulk passivation of ribbon silicon using a hydrogen rich PECVD SiN layer followed by firing is a promising alternative for enhancing minority carrier diffusion lengths as compared to hydrogenation using MIRHP which was applied in our photolithography-based cell process. The presented lifetime experiment set-up is well suited for analyzing of different gettering and hydrogenation steps.
High lifetimes as demonstrated in [1] with PECVD SiN plus RTP firing could be proven spatially resolved for a standard belt furnace firing step. First processed EFG solar cells using the new PECVD SiN based process sequence lead to stable efficiencies of 17.2 % on EFG with enough room for further optimizations. The observed degradation of record cells as demonstrated in [4] could only be observed to a certain extent for standard block cast multicrystalline material, but not for EFG cells. The reflectance of the solar cells with the SiN/MgF2-DARC using a refractive index of n = 2 for the SiN can be significantly decreased if a refractive index of n = 2.2 is used. Thus Jsc could be increased by about 1 mA/cm² and more using higher n. All presented new IV data in this contribution are not yet independently confirmed.
ACKNOWLEDGEMENTS
Part of this work was funded by the German BMU in the frame of the ASiS project (0329846J) and by the EC within the Crystal Clear project (SES6-CT-2003-502583).
LITERATURE
[1] A. Rohatgi, D.S. Kim, V. Yelundur, K. Nakayashiki, A.
Upadhyaya, M. Hilali, V. Meemongkolkiat, Technical Digest of the 14th PVSEC, Bangkok 2004, p. 635
[2] B. Damiani, K. Nakayashiki, D.S. Kim, V. Yelundur, S.
Ostapenko, I. Tarasov, A. Rohatgi, Proc. 3rd WCPEC Osaka 2003, p. 927
[3] G. Hahn, P. Geiger, Progr. Photovolt.: Res. Appl. 11, 2003, p. 347
[4] P. Geiger, G. Kragler, G. Hahn, P. Fath, E. Bucher:
Solar Energy Materials & Solar Cells 85, 2005, p. 559
[5] S.W. Glunz, E. Schaeffer, S. Rein: Proc. 3rd WCPEC, Osaka 2003, p. 919
[6] A. Rohatgi, V. Yelundur, J. Jeong, A. Ebong, M.D.
Rosenblum, J.I. Hanoka: Solar Energy Materials & Solar Cells 74, 2002, p. 117
[7] T. Lauinger, J. Schmidt, A. G. Aberle, and R. Hezel:
Appl. Phys. Lett. 68, 1996, p. 1232
926
