distribution of metal by RTP mc-Si

solar , Berkeley . W Germany Argonne, , California published correlation precipitates °CH20850 cell cell APPLIED PHYSICS LETTERS 87, 121918 H208492005H20850 Rapid thermal processing H20849RTPH20850 of solar cells has long been considered a potentially attractive substitute for con- ventional emitter diffusion. Obvious advantages are the shorter annealing times and reduced floor space necessary for RTP equipment. However, it has been observed that some silicon materials respond poorly to RTP, resulting in lower solar cell devices efficiencies. Peters et al. 1 reported that cast multicrystalline silicon H20849mc-SiH20850 responds the poorest, edge- defined film fed growth H20849EFGH20850 responds less poorly, and single-crystalline Czochralski material responds fairly well to rapid thermal processing. It was also observed that longer and higher-temperature anneals 1–4 and faster cooling rates 2,5 lead to poorer materials performance. It is typically proposed that these effects are due to the creation of pairs of interstitial iron and substitutional boron and other Fe-related defects, 2,6 or to hydrogen out-diffusion. 4 To assess the role of metals and structural defects in decreasing cast mc-Si materials performance after RTP, we employed synchrotron-based analytical techniques 7,8 to study the distribution and average sizes of metal-silicide nanopre- cipitates along structural defects. These techniques, which include x-ray fluorescence microscopy H20849H9262-XRFH20850, x-ray ab- sorption microspectroscopy H20849H9262-XASH20850, and x-ray beam in- duced current H20849XBICH20850, are powerful techniques with submi- cron spatial resolution capable of characterizing the elemental composition, chemical state and recombination ac- tivity of an iron silicide precipitate as small as 16±3 nm in radius. 9 Three sister wafers of cast mc-Si, i.e., adjacent vertical slices of the ingot, with virtually identical initial crystal structure, were selected for this analysis. The first wafer was left unprocessed and was used as a reference. The second and the third wafers were processed into solar cells using different temperatures for emitter diffusion from a phos- phorus spin-on source. The second wafer was processed at 860 °C for 120 s, while the third wafer was processed at a higher temperature, 1000 °C for 20 s. Different annealing times were chosen to obtain comparable emitter depths. The average cooling rate at the end of RTP annealing was ap- proximately 100 °C/s. While this study pertains specifically to emitter diffusion, the obtained conclusions are equally ap- plicable to high-temperature processing during later stages of cell processing. No anti-reflection coating was deposited, as to minimize the effect of hydrogen passivation in these ex- periments. The solar cell fabricated using low-temperature H20849860 °CH20850 RTP was found to be 20% H20849rel.H20850 more efficient than the cell fabricated using high-temperature H208491000 °CH20850 RTP. Most of this change in efficiency was linked to an increase of the minority carrier diffusion length L eff , as shown in the laser-beam induced current H20849LBICH20850 maps in Fig. 1. A characteristic region of material was extracted for synchrotron-based analytical studies from the same location aH20850 Author to whom correspondence should be addressed; electronic mail: buonassisi@alumni.nd.edu bH20850 Currently at: Deutsche Cell GmbH, Freiberg, Germany. cH20850 Currently at: Institute of Microtechnology, University of Neuchâtel, Impact of metal silicide precipitate dissolution processing of multicrystalline silicon T. Buonassisi aH20850 and A. A. Istratov Department of Materials Sciences and Engineering, University Science Division, Lawrence Berkeley National Laboratory S. Peters, bH20850 C. Ballif, cH20850 J. Isenberg, dH20850 S. Riepe, W Fraunhofer Institute for Solar Energy Systems, Freiburg, Z. Cai and B. Lai Advanced Photon Source, Argonne National Laboratory, E. R. Weber Department of Materials Science and Engineering, University Divison, Lawrence Berkeley National Laboratory, Berkeley H20849Received 25 January 2005; accepted 24 July 2005; Synchrotron-based analytical x-ray microprobe techniques of iron, copper, and nickel silicide precipitates at structural response to rapid thermal processing H20849RTPH20850. A direct precipitate dissolution, increased minority carrier recombination, after high-temperature H208491000 °CH20850 RTP. In contrast, iron material remained after lower-temperature RTP H20849860 minority carrier diffusion length and better solar effectively dissolved nickel and copper silicide precipitates. structural defect reservoirs detrimentally affects the point defects and smaller precipitates. For cast multicrystalline expected by inhibiting the dissolution of these precipitates, temperature of processing steps. © 2005 American Institute Neuchâtel, Switzerland. dH20850 Currently at: Q-Cells AG, Thalheim, Germany. 0003-6951/2005/87H2084912H20850/121918/3/$22.50 87, 121918-1 Downloaded 29 Aug 2006 to 202.6.242.69. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp during rapid thermal cells of California, Berkeley and Materials , California 94720 arta, R. Schindler, and G. Willeke Illinois 60439 of California, Berkeley and Materials Science 94720 online 16 September 2005H20850 were employed to study the dissolution defects in cast multicrystalline silicon in was observed between iron silicide and decreased device performance comparable in size to as-grown ; in this case the material exhibited higher performance. RTP at both temperatures It is concluded that iron dissolved from performance, likely by forming distributed silicon, higher performance can be i.e., by reducing the time and/or of Physics. H20851DOI: 10.1063/1.2048819H20852 in all three sister wafers. Medium-resolution H9262-XRF scans were performed over the same grain boundary in all three © 2005 American Institute of Physics samples, as shown in Fig. 2. In the as-grown sample, total Fe, Cu, and Ni contents in the 10 13 –10 14 cm −3 range were determined from standard-calibrated H9262-XRF measurements and estimates of the total grain boundary surface area per FIG. 1. Laser beam induced current H20849LBICH20850 maps of minority carrier diffu- sion length in H20849aH20850 a low-temperature RTP H20849860 °C,120 sH20850 solar cell and H20849bH20850 a high-temperature H208491000 °C,20 sH20850 RTP cell. The low-temperature RTP sample is 20% H20849rel.H20850 more efficient than the high-temperature RTP sample. Note the different minority carrier diffusion length scales in H20849aH20850 and H20849bH20850. FIG. 2. H20849Color onlineH20850 Synchrotron-based analysis of the metal content and distribution at grain boundaries in three sister wafers: H20849aH20850 Unprocessed ma- terial H20849as grownH20850, H20849bH20850 high-temperature RTP H208491000 °C,20 sH20850, and H20849cH20850 low- temperature RTP H20849860 °C,120 sH20850. H9262-XRF reveals that while some large FeSi precipitates remain after low-temperature RTP, high-temperature RTP 121918-2 Buonassisi et al. 2 reduces the average metal content of FeSi 2 precipitates by 50%. The recom- bination activity of intragranular regions increases with decreasing Fe con- tent at structural defects, as seen in XBIC images and Fig. 1, suggesting the dissolved iron contaminates nearby regions. Downloaded 29 Aug 2006 to 202.6.242.69. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp unit volume, 9 in good agreement with neutron activation analysis data on similar samples. 10,11 Iron content H20849in units of H9262-XRF counts per secondH20850 for different precipitates in the three samples was analyzed via high-resolution two- dimensional H9262-XRF scans over each precipitate with 50 nm steps and optimized focus conditions. In as-grown material, six iron precipitates were analyzed which had 523, 412, 508, 496, 464, and 302 counts after background subtraction, yielding an average of 451 with a standard deviation of 83 H20849equivalent to 60±4 nm diameter via standard reference ma- terialH20850. In the low-temperature RTP sample H20849860 °C,120 sH20850, the analyzed iron precipitates had 364 and 556 counts after background subtraction, yielding an average of 460 with standard deviation of 136 H20849equivalent to 60±6 nm diameterH20850. Finally, in the high-temperature RTP H208491000 °C,20 sH20850 sample, the analyzed iron precipitates had 205, 271, 312, 114, 236, and 183 counts after background subtraction, yield- ing an average of 220 with standard deviation of 70 H20849equiva- lent to 46±6 nm diameterH20850. Data for these three samples are plotted in Fig. 3. Comparison of these samples led us to the following observations: H208491H20850 In the as-grown material, multiple iron-rich clusters can be seen decorating the grain boundary. H9262-XAS analyses H20849not shownH20850 revealed the chemical state of iron in these pre- cipitates to be most similar to iron silicide. Some copper and nickel precipitates were also observed, although in lower spatial densities. The copper was determined by H9262-XAS to be in the form of copper silicide. Some similarly large clus- ters were observed in intragranular locations, probably coin- ciding with dislocations. H208492H20850 In the “low-temperature RTP” sample, some large FeSi 2 precipitates remain, with the same count rate H20849i.e., number of metal atoms per precipitateH20850 as in the as-grown material. However, while faint traces of Ni-rich precipitates can still be detected, the number of nickel atoms per precipi- tate is much reduced in amount relative to the as-grown sample. Cu 3 Si precipitates are no longer detectable. H208493H20850 In the “high-temperature RTP” sample, FeSi 2 pre- FIG. 3. H20849Color onlineH20850 Average metal content per precipitate analyzed by high-resolution H9262-XRF in reference material H20849RefH20850, low-temperature RTP H20849860 °C,120 sH20850, and high-temperature RTP H208491000 °C,20 sH20850. Nickel and copper silicide precipitates are nearly entirely dissolved during both RTP treatments. High-temperature RTP reduces the average iron silicide precipi- tate size by nearly 50%, while low-temperature RTP dissolves iron silicide precipitates to a much lesser extent. Appl. Phys. Lett. 87, 121918 H208492005H20850 cipitates are detected, but they contain on average 50% less iron atoms than as-grown material, giving evidence for iron silicide precipitate dissolution. No Cu- nor Ni-rich precipi- tates are above the detection limits. 121918-3 Buonassisi et al. Appl. Phys. Lett. 87, 121918 H208492005H20850 These results indicate that low-temperature RTP effec- tively dissolves most Cu- and Ni-rich precipitates, while FeSi 2 precipitates are largely undisturbed. This effect is due to the orders of magnitude higher solubilities and diffusivi- ties of Cu and Ni relative to Fe. On the other hand, high- temperature RTP not only completely dissolves the Cu- and Ni-rich precipitates, but it also partially dissolves the FeSi 2 precipitates. Higher temperatures greatly enhance the solu- bility and diffusivity of iron, allowing it to diffuse away from the precipitates at structural defects and contaminate the in- tragranular regions of the material. This effect can be seen by comparing XBIC images in Fig. 2: while the as-grown sample exhibits denuded zones around the grain boundaries and other structural defects, the high-temperature RTP sample shows exactly the opposite, i.e., grain boundary “bleeding” into the grains. The observation of material deg- radation due to high-temperature anneals, especially fol- lowed by fast cools, has been previously reported in the literature. 1–6 One must conceive of iron precipitates at grain bound- aries and intragranular structural defects as effective reser- voirs of metals. When metal atoms accumulate at these res- ervoirs, the metal defect concentration elsewhere is reduced, improving the diffusion lengths in those regions. On the other hand, when metal silicide precipitates are partially dis- solved during processing, metals can diffuse from these res- ervoirs and contaminate neighboring regions, effectively in- creasing the bulk minority carrier trap density and reducing the bulk diffusion lengths, as seen in Fig. 1. This effect is most pronounced for cast mc-Si, in which slow cooling dur- ing crystal growth promotes the formation of larger metal silicide precipitates H20849and a reduction in metal point defectsH20850 in as-grown material. When processing mc-Si materials with such metal reser- voirs, two effective strategies can be pursued to minimize the amount of dissolved metals distributed within the grains as the result of solar cell processing: H208491H20850 Fully dissolve these metal silicide nanoprecipitates and remove them from the active device region. This is complicated because of the lim- ited capacity and segregation coefficient of gettering layers such as aluminum backside 12 or phosphorus doped emitter, 13 which results in a fraction of the metals remaining dissolved in the bulk. Additionally, it is known that cast mc-Si contains rather large H20849several 10’s of nmH20850 metal silicide precipitates at grain boundaries, 9 due to the slow cooling process that fa- vors the formation of large precipitates. 14 Thus, several hours are needed to fully dissolve iron silicide precipitates because of solubility and diffusivity limits, 15 and the high total metal content H20849in the 10 14 cm −3 rangeH20850. The alternative strategy is to H208492H20850 disturb Fe silicide clus- ters as little as possible by using lower-temperature treat- ments in the range of 800–900 °C. Low-temperature pro- cessing by Schultz et al. 16 has recently led to a 20% record- efficient mc-Si device, suggesting that this process may be the most desirable given the current state of technology. By the same token, short H-passivation rapid thermal anneals are also preferred over long anneals, not only to limit hydro- gen out-diffusion as suggested by Rohatgi et al., 4 but also to limit metal silicide precipitate dissolution. To summarize the findings of this study, synchrotron ra- diation based x-ray microprobe techniques were used to Downloaded 29 Aug 2006 to 202.6.242.69. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp demonstrate that short high-temperature RTP completely dis- solves copper and nickel precipitates, but only partly dis- solves iron silicide nanoprecipitates. When the rapid cooling occurs at the end of the RTP anneal, metals dissolved from the precipitates become “trapped in,” forming smaller and more distributed recombination centers H20849precipitates and point defectsH20850. This leads to a decrease in solar cell perfor- mance. To overcome this performance degradation, two strat- egies are to use shorter and preferably lower-temperature