- Email: [email protected]
Features of Bridgman-grown CuInSe2
Thin Solid Films 431 – 432 (2003) 68–72
Features of Bridgman-grown CuInSe2 C.H. Champness*, I. Shih, H. Du Electrical and Computer Engineering Department, McGill University, 3480 University Street, Montreal, Que., Canada H3A 2A7
Abstract Using a one-ampoule method, ingots of CuInSe2 were grown by the vertical Bridgman technique with specific nonstoichiometric proportions of the starting elements copper, indium and selenium. With stoichiometry or an excess of selenium, the ingots were p-type. With a deficiency of selenium, n-type conductivity was obtained but with binary phases present, such as InSe, in the last zone to freeze of the ingot. Ingots were also prepared with excess copper for melt compositions corresponding to CuxInSe2 , where x was 1.1, 1.2 and 1.3. In the last zone to freeze of all these p-type ingots, precipitation occurred of copper containing dissolved selenium or indium. In the precipitate-free main part of these copper-excess ingots, no change of electrical resistivity was observed with increased copper content. However, in regions of the chalcopyrite adjacent to precipitated copper areas, at the end of each ingot, the resistivity was apparently increased. 䊚 2003 Elsevier Science B.V. All rights reserved. Keywords: Copper–indium–diselenide; Bridgman-growth; Nonstoichiometry; Ingot; Monocrystal
1. Introduction While quaternary Cu(InGa)Se2 compounds are of special interest for thin film photovoltaic solar cells, the simpler ternary chalcopyrite is a more convenient material for studying the effects of deviation from stoichiometry, particularly by avoiding the influence of variable gallium spacial distributions in the quaternary w1x. Furthermore, using bulk monocrystals of the ternary, for such an investigation, reduces the influence of grain boundaries on the results. Given that some studies of nonstoichiometry in CuInSe2 have already been made previously w2–6x, it was considered helpful to elucidate further the effects of deviation from stoichiometry, specifically of selenium and copper in the compound. Accordingly, in the present study, ingots of the ternary were prepared by the Bridgman method with significant deviations from stoichiometry in the starting proportions of these two elements. 2. Ingot growth The preparation of the ternary compound ingots, in *Corresponding author. Tel.: q1-514-398-7121; fax: q1-514-3984470. E-mail address: [email protected] (C.H. Champness).
this work, was carried out using a vertical Bridgman method with techniques previously developed in this laboratory w7x. The processes are outlined as follows. 2.1. Ampoule preparation Pellets of high purity 99.999% elemental copper, indium and selenium were first surface etched to remove any oxide layers and then weighed out in the appropriate proportion for the composition desired. Thus, for a simple stoichiometric preparation, the atomic proportion was 1:1:2 of the elements Cu, In and Se, respectively. The pellets were introduced into a specially cleaned and prepared thick-walled quartz tube, closed at one end. This was pumped down to a vacuum of 10y6 Torr and sealed off to form an ampoule, closed at both ends. One aspect of the preparation of the ampoule, prior to charging it with the elements, was especially important, whereby the inner wall of the ampoule was coated with a layer of baked-in boron nitride powder, using a process described previously w8x. The BN coating has been found to prevent sticking of the ingot to the quartz inner wall after growth, by the action of gettering oxygen from the copper, which, unless specially treated, usually contains some of this gas, despite its nominal high purity. The BN also reduces voids in the grown ingots.
0040-6090/03/$ - see front matter 䊚 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0040-6090(03)00197-4
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
2.2. Bridgman-growth The sealed ampoule, with its charge of elements, was then slowly heated in a furnace to a temperature of 300 8C to allow the selenium and indium to react together exothermically. When the heat from this reaction had dissipated, the ampoule was inserted into the vertical heating tube of a Bridgman furnace and the temperature raised to 1100 8C. This temperature was maintained for at least 24 h with mechanical rocking to enable the compound synthesis to be completed and to ensure compositional homogeneity of the melt. On rocking the ampoule back and forth, a clear flow of liquid was a good indication that ingot sticking would not be a problem after solidification. Next, the ampoule was lowered at a rate of approximately 5 mmyh through a temperature gradient of 70 8Cycm into a lower section of the Bridgman furnace, set at a temperature of 700 8C. At the end of its vertical travel, the ampoule was cooled by decreasing the lower furnace temperature at a controlled rate of approximately 30 8Cyh and removed from the furnace at room temperature. The ampoule was then broken open to retrieve the grown ingot, which was free of adherence to the quartz wall, with few or no voids, as a result of the BN pre-treatment. 3. Ingot evaluation Each ingot after growth, was first inspected visually for external features and any small pieces of it, broken away from the fracturing of the ampoule, were examined
69
under the optical microscope. Slices were abrasively cut across each ingot, with thicknesses of 1–3 mm, and polished abrasively. Then four point probe and hot probe tests were made on each of the two faces of every wafer, keeping the voltage to the hot point heater the same (30 V) from one test to the next. Powder X-ray diffraction patterns were obtained on some samples. 4. General growth results The results on some 10 ternary ingots are reported here, prepared with the starting proportions given in Tables 1 and 2. Inspection of fragments broken from the inside of the ingots indicated the presence of the monocrystals from the many observed {1 1 2} and {1 0 1} cleavage facets and {1 1 2} twin lines w9x. The surface of each ingot was molded by the curved inner wall of the ampoule and thus had few special features. However, when an external cavity was present, it presented itself as an opportunity to observe within it crystal growth completely free of wall restriction. Fig. 1 shows, within such a cavity, features in the form of equilateral triangles, which represent triangular growth facets on a {1 1 2} plane. Such facets were approximately parallel to the ingot axis, so that the normal to the {1 1 2} plane (in a N2 2 1M direction) would be perpendicular to this axis. This is not necessarily contrary to an earlier observation w10x that the growth direction along the ingot axis is N2 2 1M, since the angle between adjacent {1 1 2} planes is 70.538, which approaches a right angle.
70
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
5. Ingots with selenium nonstoichiometry Table 1 gives results for 5 ingots, including one prepared with an excess of selenium, corresponding to a melt composition of CuInSe2.2 and another with a deficiency of this element corresponding to a composition of CuInSe1.8. The former was confirmed by hot probe to be p-type and the latter n-type. The n-type material was especially interesting, since the end of the ingot, that is the last zone to freeze, was distinctly different from the rest of the material. Here, extra phases in globular form were visible, as indicated in Fig. 2a. By contrast, no such phases were observed in the rest of the ingot (Fig. 2b). X-ray diffraction of powdered material from the end region (not shown) indicated the presence of InSe. For a similar preparation of composition CuInSe1.7, Shukri et al. w6x also found Cu7In4 and Cu9In4 in this last zone but, so far, these binaries have not been confirmed by X-ray diffraction in the present work. Hot probing in this region showed lower thermoelectric voltages than in the rest of the n-type ingot, suggesting a higher electron concentration in the last zone to freeze. The p-type ingot with excess selenium, corresponding to the melt composition CuInSe2.2 (column 1) was expected, from previous work, to have a higher hole concentration than that in the stoichiometric ingots (column 2) but the four point probe and hot probe values for the middle of the ingot indicated little change. However, the resistivity at the end of the excess selenium ingot was reduced. The stoichiometrically prepared ingot of column 4
was found to be darker in appearance than usual. This arose from surface oxidation, due to a crack in the ampoule, which occurred during the Bridgman-growth and also caused the middle of the ingot to be n-type by partial loss of selenium via the crack. Thus, any oxygen present, acting as an acceptor w11x, did not manage to keep the material p-type in this region, although the ingot end was still strongly p-type. 6. Ingots with copper excess Ingots were prepared with increasing proportions of copper, corresponding to melt compositions of CuInSe2, Cu1.1InSe2, Cu1.2InSe2 and Cu1.3InSe2, as indicated in Table 2. Except for the stoichiometrically prepared ingots of column 1, precipitated copper was freely visible on the end surface of each copper excess ingot; that is on the last surface to freeze in the growth process. This precipitated copper was in the form of a fibrous mass, as shown in Fig. 3, consisting of filaments with a diameter of approximately 30 mm. X-ray diffraction of this fibrous material indicated that it was essentially copper with displaced diffraction lines as shown in Fig. 4, suggesting lattice distortion by dissolved impurities, presumed to be selenium or possibly indium. The displacements of the peaks correspond to a lattice parameter increase of approximately 2%. Sections cut through the last zone to freeze of the copper excess ingots also showed copper phase regions here. Examination of the four point resistivity and hot probe numbers in Table 2 indicates no significant change in apparent hole concentration with increase in copper
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
71
Fig. 1. Triangular growth facets, in a {1 1 2} plane, as seen at the bottom of a side cavity in a stoichiometric CuInSe2 Bridgman-grown ingot.
content from one ingot to the next. However, surprisingly, at the end of each copper excess ingot, the resistivity adjacent to the precipitated copper areas was significantly higher than that at the middle of the ingot, where copper areas were absent. When the distinct copperprecipitated areas at the ingot ends were encountered in the probe tests, the hot probe voltage was small at approximately q0.45 mV and the resistivity was too low to measure, consistent with the high electron concentration in the metal.
Fig. 2. Surface of two slices cut across ingot 7, prepared with a selenium deficiency corresponding to a melt composition of CuInSe1.8; (a) from last end to freeze after etching, (b) from middle of ingot.
7. Discussion The Bridgman-growth experiments with CuInSe2, confirm the important role of the selenium stoichiometry in the ternary compound, particularly for a deficiency of this element. In this case, it would seem that the progressive freezing in the Bridgman-growth process takes place at first with near stoichiometric proportions but still with a sufficient selenium deficiency to make the material n-type; eventually with continued solidifi-
Fig. 3. Copper fibres with a diameter of about 30 mm precipitated at the last surface to freeze of one of the excess copper ingots.
Fig. 4. X-ray diffractogram of the copper fibres shown in Fig. 3, together with four diffraction lines for pure copper.
72
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
cation, the selenium supply effectively runs out or is greatly diminished in the last zone to freeze, giving rise to the formation of the binary compounds. In the ingots prepared with excess copper, this element was precipitated at the last surface to freeze, even for the ingot with the lowest excess having the melt composition Cu1.1InSe2. The precipitation was in the form of 30-mm diameter filaments of copper with evidence of dissolved selenium or indium. The probe tests on the copper excess ingots showed no evidence of increased hole concentration, in the middle of the ingot, with increased copper content. Thus, these results are consistent with the ternary phase diagrams, indicating that the solubility of copper in CuInSe2 is very small. More surprising is the observation of apparent increased resistivity in the chalcopyrite in areas adjacent to precipitated copper regions within the last zone to freeze, compared to material in the rest of the ingot, which was free of precipitated copper. Such a result would seem to imply that the CuInSe2 may be more intrinsic in regions near the precipitated copper than elsewhere. This needs to be checked in future work, along with further analysis of other nonstoichiometric preparations.
References w1x T. Dullweber, O. Lundberg, J. Malmstrom, ¨ M. Bodegard, L. ˚ Stolt, R. Rau, H.W. Schock, J. Werner, Thin Solid Films 387 (2001) 11. w2x L.S. Yip, W.S. Weng, I. Shih, C.H. Champness, Tech Digest Fifth Int. Photovoltaic Sci. Eng. Conf., (Pub. Int. PVSEC-5 Prof. Hiroyuki Matsunami, printed Showada Insatsu K.K. Japan) (1990) 401. w3x L.S. Yip, W.S. Weng, I. Shih, C.H. Champness, Proc. 21st. Photovoltaic Spec. Conf. (1990) 768. w4x L.S. Yip, W.S. Weng, Z.A. Shukri, I. Shih, C.H. Champness, in: K.J. Bachmann, H.L. Hwang, C. Schab (Eds.), Proc. Symp. A3 on Nonstoichiometry in Semiconductors, North Holland, Amsterdam, 1991, 1992, p. 119. w5x C.A. Mullan, C.J. Keily, S.M. Casey, M. Imanieh, M.V. Yakushev, R.D. Tomlinson, J. Cryst. Growth 171 (1997) 415. w6x Z.A. Shukri, C.H. Champness, J. Cryst. Growth 191 (1998) 97. w7x C.H. Champness, J. Mater. Sci: Mater. Electron. 10 (1999) 605. w8x Z.A. Shukri, C.H. Champness, I. Shih, J. Cryst. Growth 129 (1993) 107. w9x Z.A. Shukri, C.H. Champness, Acta Cryst. B 53 (1997) 620. w10x J. Parkes, R.D. Tomlinson, M.K. Hampshire, J. Cryst. Growth 20 (1973) 315. w11x L.L. Kazmerski, Proc. 23rd IEEE Photovoltaic Spec. Conf. (1993) 1.
Features of Bridgman-grown CuInSe2 C.H. Champness*, I. Shih, H. Du Electrical and Computer Engineering Department, McGill University, 3480 University Street, Montreal, Que., Canada H3A 2A7
Abstract Using a one-ampoule method, ingots of CuInSe2 were grown by the vertical Bridgman technique with specific nonstoichiometric proportions of the starting elements copper, indium and selenium. With stoichiometry or an excess of selenium, the ingots were p-type. With a deficiency of selenium, n-type conductivity was obtained but with binary phases present, such as InSe, in the last zone to freeze of the ingot. Ingots were also prepared with excess copper for melt compositions corresponding to CuxInSe2 , where x was 1.1, 1.2 and 1.3. In the last zone to freeze of all these p-type ingots, precipitation occurred of copper containing dissolved selenium or indium. In the precipitate-free main part of these copper-excess ingots, no change of electrical resistivity was observed with increased copper content. However, in regions of the chalcopyrite adjacent to precipitated copper areas, at the end of each ingot, the resistivity was apparently increased. 䊚 2003 Elsevier Science B.V. All rights reserved. Keywords: Copper–indium–diselenide; Bridgman-growth; Nonstoichiometry; Ingot; Monocrystal
1. Introduction While quaternary Cu(InGa)Se2 compounds are of special interest for thin film photovoltaic solar cells, the simpler ternary chalcopyrite is a more convenient material for studying the effects of deviation from stoichiometry, particularly by avoiding the influence of variable gallium spacial distributions in the quaternary w1x. Furthermore, using bulk monocrystals of the ternary, for such an investigation, reduces the influence of grain boundaries on the results. Given that some studies of nonstoichiometry in CuInSe2 have already been made previously w2–6x, it was considered helpful to elucidate further the effects of deviation from stoichiometry, specifically of selenium and copper in the compound. Accordingly, in the present study, ingots of the ternary were prepared by the Bridgman method with significant deviations from stoichiometry in the starting proportions of these two elements. 2. Ingot growth The preparation of the ternary compound ingots, in *Corresponding author. Tel.: q1-514-398-7121; fax: q1-514-3984470. E-mail address: [email protected] (C.H. Champness).
this work, was carried out using a vertical Bridgman method with techniques previously developed in this laboratory w7x. The processes are outlined as follows. 2.1. Ampoule preparation Pellets of high purity 99.999% elemental copper, indium and selenium were first surface etched to remove any oxide layers and then weighed out in the appropriate proportion for the composition desired. Thus, for a simple stoichiometric preparation, the atomic proportion was 1:1:2 of the elements Cu, In and Se, respectively. The pellets were introduced into a specially cleaned and prepared thick-walled quartz tube, closed at one end. This was pumped down to a vacuum of 10y6 Torr and sealed off to form an ampoule, closed at both ends. One aspect of the preparation of the ampoule, prior to charging it with the elements, was especially important, whereby the inner wall of the ampoule was coated with a layer of baked-in boron nitride powder, using a process described previously w8x. The BN coating has been found to prevent sticking of the ingot to the quartz inner wall after growth, by the action of gettering oxygen from the copper, which, unless specially treated, usually contains some of this gas, despite its nominal high purity. The BN also reduces voids in the grown ingots.
0040-6090/03/$ - see front matter 䊚 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0040-6090(03)00197-4
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
2.2. Bridgman-growth The sealed ampoule, with its charge of elements, was then slowly heated in a furnace to a temperature of 300 8C to allow the selenium and indium to react together exothermically. When the heat from this reaction had dissipated, the ampoule was inserted into the vertical heating tube of a Bridgman furnace and the temperature raised to 1100 8C. This temperature was maintained for at least 24 h with mechanical rocking to enable the compound synthesis to be completed and to ensure compositional homogeneity of the melt. On rocking the ampoule back and forth, a clear flow of liquid was a good indication that ingot sticking would not be a problem after solidification. Next, the ampoule was lowered at a rate of approximately 5 mmyh through a temperature gradient of 70 8Cycm into a lower section of the Bridgman furnace, set at a temperature of 700 8C. At the end of its vertical travel, the ampoule was cooled by decreasing the lower furnace temperature at a controlled rate of approximately 30 8Cyh and removed from the furnace at room temperature. The ampoule was then broken open to retrieve the grown ingot, which was free of adherence to the quartz wall, with few or no voids, as a result of the BN pre-treatment. 3. Ingot evaluation Each ingot after growth, was first inspected visually for external features and any small pieces of it, broken away from the fracturing of the ampoule, were examined
69
under the optical microscope. Slices were abrasively cut across each ingot, with thicknesses of 1–3 mm, and polished abrasively. Then four point probe and hot probe tests were made on each of the two faces of every wafer, keeping the voltage to the hot point heater the same (30 V) from one test to the next. Powder X-ray diffraction patterns were obtained on some samples. 4. General growth results The results on some 10 ternary ingots are reported here, prepared with the starting proportions given in Tables 1 and 2. Inspection of fragments broken from the inside of the ingots indicated the presence of the monocrystals from the many observed {1 1 2} and {1 0 1} cleavage facets and {1 1 2} twin lines w9x. The surface of each ingot was molded by the curved inner wall of the ampoule and thus had few special features. However, when an external cavity was present, it presented itself as an opportunity to observe within it crystal growth completely free of wall restriction. Fig. 1 shows, within such a cavity, features in the form of equilateral triangles, which represent triangular growth facets on a {1 1 2} plane. Such facets were approximately parallel to the ingot axis, so that the normal to the {1 1 2} plane (in a N2 2 1M direction) would be perpendicular to this axis. This is not necessarily contrary to an earlier observation w10x that the growth direction along the ingot axis is N2 2 1M, since the angle between adjacent {1 1 2} planes is 70.538, which approaches a right angle.
70
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
5. Ingots with selenium nonstoichiometry Table 1 gives results for 5 ingots, including one prepared with an excess of selenium, corresponding to a melt composition of CuInSe2.2 and another with a deficiency of this element corresponding to a composition of CuInSe1.8. The former was confirmed by hot probe to be p-type and the latter n-type. The n-type material was especially interesting, since the end of the ingot, that is the last zone to freeze, was distinctly different from the rest of the material. Here, extra phases in globular form were visible, as indicated in Fig. 2a. By contrast, no such phases were observed in the rest of the ingot (Fig. 2b). X-ray diffraction of powdered material from the end region (not shown) indicated the presence of InSe. For a similar preparation of composition CuInSe1.7, Shukri et al. w6x also found Cu7In4 and Cu9In4 in this last zone but, so far, these binaries have not been confirmed by X-ray diffraction in the present work. Hot probing in this region showed lower thermoelectric voltages than in the rest of the n-type ingot, suggesting a higher electron concentration in the last zone to freeze. The p-type ingot with excess selenium, corresponding to the melt composition CuInSe2.2 (column 1) was expected, from previous work, to have a higher hole concentration than that in the stoichiometric ingots (column 2) but the four point probe and hot probe values for the middle of the ingot indicated little change. However, the resistivity at the end of the excess selenium ingot was reduced. The stoichiometrically prepared ingot of column 4
was found to be darker in appearance than usual. This arose from surface oxidation, due to a crack in the ampoule, which occurred during the Bridgman-growth and also caused the middle of the ingot to be n-type by partial loss of selenium via the crack. Thus, any oxygen present, acting as an acceptor w11x, did not manage to keep the material p-type in this region, although the ingot end was still strongly p-type. 6. Ingots with copper excess Ingots were prepared with increasing proportions of copper, corresponding to melt compositions of CuInSe2, Cu1.1InSe2, Cu1.2InSe2 and Cu1.3InSe2, as indicated in Table 2. Except for the stoichiometrically prepared ingots of column 1, precipitated copper was freely visible on the end surface of each copper excess ingot; that is on the last surface to freeze in the growth process. This precipitated copper was in the form of a fibrous mass, as shown in Fig. 3, consisting of filaments with a diameter of approximately 30 mm. X-ray diffraction of this fibrous material indicated that it was essentially copper with displaced diffraction lines as shown in Fig. 4, suggesting lattice distortion by dissolved impurities, presumed to be selenium or possibly indium. The displacements of the peaks correspond to a lattice parameter increase of approximately 2%. Sections cut through the last zone to freeze of the copper excess ingots also showed copper phase regions here. Examination of the four point resistivity and hot probe numbers in Table 2 indicates no significant change in apparent hole concentration with increase in copper
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
71
Fig. 1. Triangular growth facets, in a {1 1 2} plane, as seen at the bottom of a side cavity in a stoichiometric CuInSe2 Bridgman-grown ingot.
content from one ingot to the next. However, surprisingly, at the end of each copper excess ingot, the resistivity adjacent to the precipitated copper areas was significantly higher than that at the middle of the ingot, where copper areas were absent. When the distinct copperprecipitated areas at the ingot ends were encountered in the probe tests, the hot probe voltage was small at approximately q0.45 mV and the resistivity was too low to measure, consistent with the high electron concentration in the metal.
Fig. 2. Surface of two slices cut across ingot 7, prepared with a selenium deficiency corresponding to a melt composition of CuInSe1.8; (a) from last end to freeze after etching, (b) from middle of ingot.
7. Discussion The Bridgman-growth experiments with CuInSe2, confirm the important role of the selenium stoichiometry in the ternary compound, particularly for a deficiency of this element. In this case, it would seem that the progressive freezing in the Bridgman-growth process takes place at first with near stoichiometric proportions but still with a sufficient selenium deficiency to make the material n-type; eventually with continued solidifi-
Fig. 3. Copper fibres with a diameter of about 30 mm precipitated at the last surface to freeze of one of the excess copper ingots.
Fig. 4. X-ray diffractogram of the copper fibres shown in Fig. 3, together with four diffraction lines for pure copper.
72
C.H. Champness et al. / Thin Solid Films 431 – 432 (2003) 68–72
cation, the selenium supply effectively runs out or is greatly diminished in the last zone to freeze, giving rise to the formation of the binary compounds. In the ingots prepared with excess copper, this element was precipitated at the last surface to freeze, even for the ingot with the lowest excess having the melt composition Cu1.1InSe2. The precipitation was in the form of 30-mm diameter filaments of copper with evidence of dissolved selenium or indium. The probe tests on the copper excess ingots showed no evidence of increased hole concentration, in the middle of the ingot, with increased copper content. Thus, these results are consistent with the ternary phase diagrams, indicating that the solubility of copper in CuInSe2 is very small. More surprising is the observation of apparent increased resistivity in the chalcopyrite in areas adjacent to precipitated copper regions within the last zone to freeze, compared to material in the rest of the ingot, which was free of precipitated copper. Such a result would seem to imply that the CuInSe2 may be more intrinsic in regions near the precipitated copper than elsewhere. This needs to be checked in future work, along with further analysis of other nonstoichiometric preparations.
References w1x T. Dullweber, O. Lundberg, J. Malmstrom, ¨ M. Bodegard, L. ˚ Stolt, R. Rau, H.W. Schock, J. Werner, Thin Solid Films 387 (2001) 11. w2x L.S. Yip, W.S. Weng, I. Shih, C.H. Champness, Tech Digest Fifth Int. Photovoltaic Sci. Eng. Conf., (Pub. Int. PVSEC-5 Prof. Hiroyuki Matsunami, printed Showada Insatsu K.K. Japan) (1990) 401. w3x L.S. Yip, W.S. Weng, I. Shih, C.H. Champness, Proc. 21st. Photovoltaic Spec. Conf. (1990) 768. w4x L.S. Yip, W.S. Weng, Z.A. Shukri, I. Shih, C.H. Champness, in: K.J. Bachmann, H.L. Hwang, C. Schab (Eds.), Proc. Symp. A3 on Nonstoichiometry in Semiconductors, North Holland, Amsterdam, 1991, 1992, p. 119. w5x C.A. Mullan, C.J. Keily, S.M. Casey, M. Imanieh, M.V. Yakushev, R.D. Tomlinson, J. Cryst. Growth 171 (1997) 415. w6x Z.A. Shukri, C.H. Champness, J. Cryst. Growth 191 (1998) 97. w7x C.H. Champness, J. Mater. Sci: Mater. Electron. 10 (1999) 605. w8x Z.A. Shukri, C.H. Champness, I. Shih, J. Cryst. Growth 129 (1993) 107. w9x Z.A. Shukri, C.H. Champness, Acta Cryst. B 53 (1997) 620. w10x J. Parkes, R.D. Tomlinson, M.K. Hampshire, J. Cryst. Growth 20 (1973) 315. w11x L.L. Kazmerski, Proc. 23rd IEEE Photovoltaic Spec. Conf. (1993) 1.