Influence of carbon and hydrogen segregation on the electrical properties of grain boundaries in polycrystalline silicon sheets

Influence of carbon and hydrogen segregation on the electrical properties of grain boundaries in polycrystalline silicon sheets


948KB Sizes 5 Downloads 51 Views

Solar Cells, 9 (1983) 149 - 157



Laboratoire de MStaUurgie Physique, Universitd Paris-Sud (France) CATHERINE TEXIER-HERVO, MICHEL MAUTREF and CHRISTIAN BELOUET

Laboratoires de Marcoussis, Centre de Recherches de la Compagnie Gdndrale d'Electricitd, route de Nozay, 91460 Marcoussis (France) (Received June 6, 1982; accepted October 12, 1982)

Summary In this article the properties of grain and twin boundaries in polycrystalline silicon layers grown by the ribbon-against-drop process are reported; the thermal diffusion of phosphorus, the electrical activity of these defects and their passivation by hydrogen are discussed. Autoradiography experiments did not evidence any significant intergranular diffusion of phosphorus; crossed electron-beam-induced current (or light-beam-induced current) and autoradiography experiments strongly suggest that the recombination at grain and twin boundaries is dominated by impurities. Finally, it is shown that the hydrogen passivation is associated with retarded diffusion and large segregation effects.

1. Introduction

The growth of polycrystalline silicon ribbons and supported sheets has received considerable attention with a view to achieving low cost solar cells. However, the growth defects in these materials impair their conversion efficiency; in particular, grain boundaries (GBs) and twin boundaries (TBs) have been found to reduce the open-circuit voltage and the fill factor of devices whereas they hardly affect the density of photocurrent generation owing to the large-sized grains at present obtained (see, for example, ref. 1 ). This article is concerned with polycrystalline layers grown on a carbon support by the ribbon-against-drop (RAD) process [2]. The electrical conductivity and the crystallographic structure of the GBs and TBs in this material were studied by the transient photoconductivity and transmission electron microscopy (TEM) techniques respectively [3]. Below we report the results *Present address: Department of Physics, University of Fluminense, Niteroi, Brazil. 0379-6787/83/$3.00

© Elsevier Sequoia/Printed in The Netherlands

150 of further comprehensive studies of phosphorus intergranular diffusion, h. purity segregation and hydrogen passivation at GBs and TBs, the aim of this work being to understand and eventually to suppress the adverse effects of these defects on the solar cell performance.

2. Experimental details The electrical activity of extended defects was mapped in a scanning electron microscope using electron-beam-induced current (EBIC) contrast or by local photoresponse measurement using light-beam-induced current (LBIC) contrast in combination with the electrolyte-semiconductor junction [ 4]. The phosphorus profiles were measured on n÷/p homojunctions 1 cm 2 in size. The layers (separated from their carbon support in this study) were mechanically polished to obtain flat surfaces and submitted to a slight etch before diffusion. Diffusions were carried out from a phosphine-doped silica source in the range 900 - 1150 °C with diffusion times of 2 - 5 h; (111) singlecrystal wafers were processed simultaneously in order to evaluate the junction edge position. After elimination of the silica source, the samples were irradiated in a flux of thermal neutrons. A fraction of 31p was converted to 32p, a/3 emitter at 1.71 MeV with a half-life of 14.3 days. The 32p profile was determined by the residual radioactivity m e t h o d [5]. In the vicinity of the junction edge, autoradiographs were taken after each chemical lapping in order to evidence possible accelerated intergranular diffusion of phosphorus. The affinity of GBs and TBs for carbon and hydrogen segregation and the distribution of the impurities across the layers were studied by means of conventional and high resolution autoradiography techniques, as follows. ~4C was incorporated in the layers by diffusion at 1200 °C for 21 days from a silicon 14C source in an evacuated and sealed ampoule. 3H was incorporated at 150 °C by cathodic charging. The electrolyte was the (molten NaHSO4)-57%H20-43%KHSO4 eutectic [6]. The sample potential was set at --4 V relative to the Ag[Ag+ reference electrode and the duration of the charging was 1 - 4 h. The high resolution autoradiographs were taken for thick and thin samples. The spatial resolutions for 14C and 3H autoradiographs are 300 nm and 120 nm respectively [7]. The autoradiographs were obtained by scanning electron microscopy (SEM), and by transmission electron microscopy (TEM) for samples less than 1 ttm thick. In the latter case the nature of the crystallographic defects and the location of the segregated radio-isotope, labelled by silver grains, could be determined simultaneously. The silver grains appear to be superimposed on the microstructure of the sample as white dots in SEM micrographs (secondary electron emission mode) and black aggregates in TEM micrographs. Finally, all the results below refer to p-type 1 ~2 cm polycrystalline layers.

151 3. Results

3.1. Phosphorus diffusion Accelerated intergranular diffusion of phosphorus has been observed in bicrystal studies [8]. However, the accumulated information a b o u t phosphorus diffusion in polycrystalline silicon suggests that intergranular diffusion is very sensitive to the boundary structure; this observation is corroborated by other studies of the diffusion of antimony in chemically vapour-deposited and SILSO polycrystalline materials which evidenced an accelerated intergranular diffusion in the chemically vapour-deposited materials only [9]. The 32p profiles measured in RAD layers are typical of a bulk diffusion process and the successive autoradiographs taken at increasing depths never evidenced any preferential diffusion at GBs or TBs. The activation energy of the bulk diffusion was found to be 248 kJ mo1-1 {compared with 221 kJ mo1-1 for the SILSO material) [5]; this value is consistent with the vacancyassisted diffusion process already reported for single crystals [10]. 3.2. Electrical activity o f extended defects TEM work showed that GBs were high order TBs whereas planar faults inside the grains were first-order TBs, paired twins or stacking faults. In both GBs and to a lesser extent TBs, extrinsic dislocation lines and sometimes precipitates have been observed [3]. The a m o u n t of recombination of the minority carriers was found to vary considerably for different GBs and also along a single boundary; this effect is illustrated dramatically in Fig. 1 where a GB (denoted by 1 ), otherwise electrically active, locally exhibits an EBIC signal that is enhanced with respect to that of the adjacent grains. Similar observations were recently reported by Turner et al. for polycrystalline silicon ingots [ 11 ]. Figure 2 shows SEM and EBIC images of an active boundary at a large magnification. It can be seen in the SEM image that the GB is macroscopically built of regions aligned with the directions of twins (denoted by 2) in one of the adjacent grains; although the EBIC contrast is poor at this magnification, significant variations in the EBIC signal can be correlated with changes in orientation of this boundary. Finally, Fig. 3 shows a strong and broad recombination at particular GBs and TBs. This phenomenon was generally observed in large grains typically a few millimetres wide. In such grains, subgrain boundaries associated with a morphological modification of the surface (the SEM image in Fig. 3) delineate regions presenting minute misorientations as inferred by X-ray Laiie pattern analysis [12]. Decorations made using a Sirtl etch reveal a high density of etch pits which spread to large distances on either side of the subgrain boundary. These extended defects are systematically found to exhibit a strong recombination effect as illustrated in the example of the EBIC image in Fig. 3. In addition, unlike most TBs, the TBs next to the subgrain boundary are generally electrically active.


Fig. 1. SEM micrograph (left) and EBIC image (right) o f the same area. The recombination (black contrast) is not constant along the GB denoted by 1. The indices used are common to all the figures, i.e. 1 denotes a GB and 2 denotes a TB.

Fig. 2. The same arrangement as in Fig. 1 ; the GB is built o f facets oriented in the directions of TBs in the contiguous grain.

T h e s e qualitative b u t r e p e a t e d m a c r o s c o p i c o b s e r v a t i o n s s t r o n g l y suggest t h a t t h e r e c o m b i n a t i o n rate o f m i n o r i t y carriers at GBs is sensitive t o the GB s t r u c t u r a l characteristics. T E M studies have b e e n initiated in o r d e r t o c h a r a c t e r i z e distinctly t h e s t r u c t u r e o f t h e GBs distinguished b y t h e i r electrical a c t i v i t y [ 13 ].

3.3. Carbon segregation and recombination of minority carriers T h e o b s e r v a t i o n s o f 14C d i s t r i b u t i o n s as described above, in s a m p l e s p r e v i o u s l y m a p p e d f o r r e c o m b i n i n g d e f e c t s b y E B I C e x p e r i m e n t s , established t h a t 14C did segregate w h e r e r e c o m b i n a t i o n o c c u r r e d a t GBs a n d TBs [5, 1 4 ] . Figure 4 illustrates t h e results o f t w o sets o f crossed E B I C and a u t o r a d i -


Fig. 3. The same arrangement as in Fig. 1; 5 denotes a surface defect; the EBIC signal (origin denoted by o ebic) along the scan path (white line) shows a strong recombination at the GB and nearby TBs; other TBs are not active.




(bl) (b2) (b3) Fig. 4. The circled areas (black and white) in the EBIC images were selected for the autoradiography experiments; the arrow next to each circled area points to the adjacent and corresponding autoradiograph (e.g. the circled area at the top of the EBIC image (a2) is presented in autoradiograph (a3) as indicated by the white arrow). The arrows in the SEM micrographs point to GBs where a large a m o u n t of 14C segregation takes place.

154 ography experiments; the optical micrographs of autoradiographs (Figs. 4(al) and 4(aa) ) evidence the 14C segregation at recombining GBs (the EBIC picture in Fig. 4(a2)) and the SEM micrographs of autoradiographs (Figs. 4(bl) and 4(ba) ) show a large 14C segregation at recombining TBs only (the EBIC picture in Fig. 4{b2)). Additional studies not developed here revealed large variations in 14C segregation along GBs and high resolution autoradiographs of thin films observed by TEM indicated an enhanced segregation at extrinsic dislocation lines pinned in the boundary. Finally, asymmetric '4C distributions on either side of GBs were also observed. On the assumption that 14C segregation occurs where carbon (or any metallic impurity reactive with carbon) is already present in the native material, these results suggest that (i) impurities are preferentially but not evenly segregated along GBs, their distribution being sensitive to the crystallographic structure of the boundary, (ii) impurity profiles may stretch far away from the GBs and (iii) TBs decorated by ~4C may have a composition different from that of the predominant undecorated TBs. Lastly, the extrinsic dislocation lines evidenced in active TBs and present in most GBs may contribute to the segregation of impurities along both types of extended defect. Within the scope of the above hypothesis, the similarity between the spatial distributions along GBs and TBs of segregated impurities and electrically active areas do suggest a role for carbon and possibly other impurities in contributing to the recombination rate of minority carriers along GBs and TBs.

3.4. Hydrogen passivation of grain boundaries The passivation of GBs in polycrystalline silicon by hydrogen plasma has recently been given much attention [15, 16]. Experiments conducted in our laboratory on RAD polycrystalline layers also indicated that high temperature annealings (600 - 800 °C) in a hydrogen ambient improved the minority carrier diffusion length in the bulk material [17]. These annealing experiments were conducted over periods of 1 - 4 h in various ambients (nitrogen, argon and hydrogen) on layers coated with a phosphorus-doped silica source prior to (P samples) or after (A samples) the actual diffusion. Although quantitative comparisons of the characteristics of the solar cells thus obtained could n o t be made because the difference in the quality of the layers used was too great, systematic increases in the photocurrent density Jph and the open-circuit voltage Voc (for P samples in particular) were observed for hydrogen-annealed samples; typical values of the increments AJph and AVoc were 1 - 3 mA cm -2 and 10 - 40 mV respectively. The increase in Jph could n o t be explained only in terms of GB passivation due to the large width of the grains b u t rather was believed to be a result of an increase in the bulk minority carrier diffusion length. Finally, this hydrogen effect was found to be most effective at 800 °C and to be considerably reduced in the absence o f a phosphorus-doped solid source during the annealing. Then, further experiments were conducted in order to investigate on a point-by-point basis the location of hydrogen segregation and the






Fig. 5. 3H autoradiographs: (a), (b) SEM; (c), (d) TEM. 3H segregation can be observed along TBs in (a) and (d), at dislocations in GBs in (c) and also inside homogeneous areas in (b) and (d).

Distance (gm) Fig. 6. LBIC signals before ( -' ) and after ( ....... ) h y d r o g e n passivation. The scan paths are indicated in the SEM micrograph. R e c o m b i n a t i o n at the grain tip has been eliminated at the GB a n d TB and inside the grain.

modification of the local recombination using the techniques described in the experimental section of this paper. The early results available indicate that the diffusion of hydrogen at 150 °C was considerably slower than that predicted on the basis of data obtained with single crystals; substantial

156 segregation effects were noted at GBs and TBs and segregation at extrinsic dislocation lines was observed, as illustrated in Fig. 5. In other experiments, LBIC scans of n÷/p homojunction cells before and after hydrogen cathodic charging revealed strong passivation effects, at least for some GBs and inside the grains as well (Fig. 6).

4. Concluding remarks The rapid survey of the properties of GBs and TBs in this article shows that these defects do not lead to appreciable intergranular diffusion of phosphorus, at least in the range of conventional diffusion temperatures. A tentative interpretation of the 14C segregation at these defects is that they are presumably decorated by segregated impurities (including carbon) with profiles which may stretch to large distances from the GBs. The incorporation of impurities in these defects is presumably controlled by both the local impurity redistribution ahead of the growth front and the crystallographic structure of the boundary; extrinsic dislocation lines eventually contribute to the segregation. Crossed EBIC and autoradiography experiments support the hypothesis that the electrical activity of GBs and TBs is related to impurity segregation in these defects. Finally, h:~drogen incorporated by cathodic charging at 150 °C was shown (i) to segregate preferentially at GBs and TBs, (ii) to passivate some of the electrically active boundaries and (iii) to reduce the recombination in the bulk material. This work is being continued with a quantitative evaluation of the effect of hydrogen passivation on the solar cell electrical characteristics.

Acknowledgments The authors wish to t h a n k J. Hervo and P. Bacle for their valuable assistance in the experimental part of the work. This work was partly sponsored by the Commissariat ~ l'Energie Solaire (COMES) and the Commission of the European Communities (CEC).

References 1 G. W. Cullen and T. Surek (eds.), Shaped Crystal Growth, North-Holland, Amsterdam, 1980. 2 C. Belouet, in K. V. Ravi and B. O'Hara (eds.), NovelSilicon Growth Methods 1980, Electrochemical Society Meet., Saint Louis, MO, 1980, Vol. 80-5, Electrochemical Society, Princeton, NJ, 1980, pp. 195 - 212.

157 3 A. Johan, J. Ayerza, D. Bielle Daspet, A. Rocher and C. Fontaine, in R. Van Overstraeten and W. Palz (eds.), Proc. 2nd Commission o f the European Communities Conf. on Photovoltaic Solar Energy, West Berlin, 1979, Reidel, Dordrecht, 1979, p. 750. 4 E. Fabre and M. Mautref, Acta Electron., 18 (4) (1975) 331. 5 A. Chari, Th~se de Troisidme Cycle, Universit~ Paris-Sud, July 1980. 6 J. Chene, Mdt.:Corros. Ind., 262 (1977) 623. 7 J. P. Laurent and G. Lapasset, Int. J. Appl. Radiat. Isot., 24 (1973) 213. 8 R. Biberian and J. Cabane, Universitd Saint-J6r0me, Marseille, personal communication, 1981. 9 H. J. Queisser, K. Hubner and W. Shockley, Phys. Rev., 123 (1961) 1245. 10 -M. Yoshiba, Jpn. J. Appl. Phys., 19 (12) (1980) 2427. 11 G. B. Turner, D. Tarrant, D. Aldrich, R. Pressley and R. Press, Proc. 4th Commission o f the European Communities Conf. on Photovoltaic Solar Energy, Stresa, May 10 14, 1982, Reidel, Dordrecht, 1982, p. 1039. 12 O. Rallon, M. Aucouturier, C. Texier-Hervo, M. Mautref and C. Belouet, unpublished results, 1982. 13 R. Sharko, A. Gervais and C. Texier-Hervo, Proc. Conf. on Polycrystalline Semiconductors, Perpignan, September I - 4 , 1982, in J. Phys. (Paris), Colloq. C1, 16 (43) (1982) 129. 14 C. Texier, [email protected] de Troisi~me Cycle, Universit6 Paris VII, June 1980. 15 C. H. Seager and D. S. Ginley, Appl. Phys. Lett., 34 (1979) 337. 16 O. Rallon, M. Aucouturier, M. Mautref and C. Belouet, Proc. Conf. on PolycrystaUine Semiconductors, Perpignan, September 1 - 4, 1982, in J. Phys. (Paris), Colloq. C1, 10 (43) (1982) 117. 17 E. Fabre, Final Rep., 1981 (Commissariat ~ l'Energie Solaire Contract COMES 78.007).