Relationship between dislocation density and nucleation of multicrystalline silicon

Relationship between dislocation density and nucleation of multicrystalline silicon

Available online at www.sciencedirect.com Acta Materialia 58 (2010) 3223–3229 www.elsevier.com/locate/actamat Relationship between dislocation densi...

654KB Sizes 0 Downloads 48 Views

Available online at www.sciencedirect.com

Acta Materialia 58 (2010) 3223–3229 www.elsevier.com/locate/actamat

Relationship between dislocation density and nucleation of multicrystalline silicon G. Stokkan * Norwegian University of Science and Technology, Department of Materials Science and Engineering, Norway Received 6 October 2009; received in revised form 20 January 2010; accepted 26 January 2010 Available online 4 March 2010

Abstract Dislocation density and crystal orientation were investigated for a set of multicrystalline silicon ingots grown in a pilot scale furnace. Both low and high dislocation density ingots were observed. The low dislocation density ingots showed a dominating orientation close to (2 1 1) in contrast to the high dislocation density ingots. The orientations are consistent with growth on dendritic crystals formed along the crucible bottom and dendritic crystals with an angle towards the melt, respectively. During crystal growth, the power that was dissipated to the crystallization furnace showed a marked drop shortly after the onset of crystallization for low dislocation density ingots, an indication of fast release of crystallization heat from dendritic growth. Ingots that were not dominated by a high dislocation density instead had a high occurrence of twinned areas. Favourable orientation of the [1  1 0] vector in the growth plane is suggested to be the cause of growth dominated by multiple twin faceting. This favourable orientation existed for crystals grown from dendrites grown along the crucible bottom, and this is suggested as an explanation for why these crystals are dominated by multiple twins rather than dislocations. Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Dislocations; Directional solidification; Dendritic growth; Grain boundary twin; Silicon

1. Introduction Silicon is the material most widely used in the production of solar cells, with a market share of 87.5% in 2008 [1]. The obtainable cell efficiency varies greatly, from 15– 16% for standard industrial cells to 25% for high-efficiency cells which employ advanced cell architecture to high-quality material [2,3]. Both the impurity content and the density of crystal defects affect the efficiency obtainable. Monocrystalline silicon, either Float Zone (FZ) or Czochralski (CZ), has both low defect density and impurity content and is used for high-efficiency cells. The main low-cost alternative to this material is directionally solidified multicrystalline (MC) silicon. This material has higher degree of contamination due to contact with impure materials during high-temperature processing, and higher crystal defect con*

Tel.: +47 73597089; fax: +47 73550203. E-mail address: [email protected]

tent (grain boundaries, sub-grain boundaries and dislocations) due to the lack of a seeding process and less ideal thermal conditions. Impurities [4] and crystal defects [5,6] reduce the efficiency of the solar cells. The occurrence of crystal defects and impurities also contribute to reducing the efficiency; it is well known that dislocations are made electrically active by decoration of impurities [7] and that sub-grain boundaries and grain boundaries of higher planar mismatch [8–10] are more detrimental than higher symmetry grain boundaries such as coincidence site lattice (CSL) boundaries and twins, also due to decoration by impurities. Cell areas of high dislocation density have also been shown to be difficult to improve by gettering [11]. These factors make MC silicon less suitable for use in high-efficiency cells. In recent years, much attention has been given to a method proposed by Fujiwara et al. [12] to create a seed in situ by growing dendrites in a particular crystal direction in a thin region near the crucible bottom, from which

1359-6454/$36.00 Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2010.01.042

3224

G. Stokkan / Acta Materialia 58 (2010) 3223–3229

columnar growth of crystals starts. By controlling supercooling to a certain level within a thin region, it has been reported that dendrites grow in the [0  1 1] or [1  1 2] directions parallel to the crucible bottom, forming upper planes of (2 1 1) and (1 1 0), respectively, as shown in Fig. 1. Such controlled nucleation will create large grains with a favourable crystal orientation for the following cell processing. Also, the formation of fewer grain boundaries of higher CSL symmetry and which run more parallel to the growth direction may result in less generation of dislocation clusters [13]. These factors make MC silicon grown by this process a promising candidate for low-cost material suitable for high-efficiency solar cells. In commercial directional solidification systems, a layer of higher supercooling is formed near the bottom of the crucible, and traces of dendritic growth are often observed on the bottom of the ingot. Also, the dislocation density and density of twins vary between materials grown in different crystallization furnaces and even between ingots grown in the same furnace. In this paper, this effect is investigated in a set of ingots grown in a pilot scale directional solidification furnace. Correlation between dislocation density and crystal orientation is observed, and the crystal orientation is determined at the nucleation stage, probably by the growth of dendritic crystals along or close to the crucible bottom. 2. Experimental work

to produce CZ crystals. Some of the growth parameters, such as time in the different phases (heat up of furnace, melting of charge, crystallization, holding at high temperature, and cool down to room temperature) were varied. The furnace programmes of ingots 4 and 5 were kept as identical as possible. From the centre of the ingots, blocks 100  100 mm2 were cut. A bottom layer of the ingot 5–10 mm was removed prior to wafer sawing. The block was cut into 250–350 lm-thick wafers. A selection of wafers from bottom to top of all the ingots were mechanically and chemically mechanically polished before they were cleaned and etched in a Sopori etch for 30 s [14]. This reveals dislocations penetrating the surface as etch pits, and the etch pit density is taken as a measure of the dislocation density. The dislocation density was measured using PVScan 6000 [15], and an example of two different dislocation density scans (wafer 26 from ingot 4 and wafer 23 from ingot 5) is shown in Fig. 2. It has been shown that PVScan gives a measurement error in areas of high twin density, shifting the results towards an apparently higher density [16]. The twinned areas will have a distinctive pattern of parallel lines. Since the twinned areas may or may not also contain dislocations, it is not possible to filter the results accurately using this feature, but it can be used to indicate manually a magnitude of the error, which is done in Fig. 3. Fig. 3 displays the fraction of the wafer area that has a dislocation density >105 cm2. Minority carrier lifetime was measured on selected samples using the carrier density imaging (CDI) technique [17].

A set of five ingots were produced in a pilot scale directional solidification furnace (Crystalox DS-250) producing cylindrical ingots of 10–12 kg with diameter 250 mm and height 100–120 mm. Note that two of the ingots (1 and 3) were produced in standard flat-bottomed slip cast crucibles, whereas three (2, 4 and 5) were made in electronic-grade concave-bottomed crucibles, normally used

Fig. 1. Plane orientations and directions resulting from dendritic growth in the [0  1 1] and [1  1 2] directions. The colours are taken from the colourcoded reference triangle, as seen in Fig. 4f, and the two colours on each dendrite represent directions slightly misoriented from and symmetrically arranged around the main upper plane orientation, i.e., (2 1 1) and (1 1 0). An odd number of twins at the centre of the dendrite is chosen in order to illustrate absolute orientations on each side of the twins.

Fig. 2. Dislocation density plots of wafer 26 from ingot 4 and wafer 23 from ingot 5 measured by PVScan 6000 as well as minority carrier lifetime images of wafer 22 from ingot 4 and wafer 22 from ingot 5 as measured by CDI. The grey scale is logarithmic, showing the sensitivity range of 104– 106 cm2 of the instrument. The numbers on the axes of the dislocation density plots show the dimensions measured in centimetres. The green squares indicate where the lifetime images were made on the adjoining wafers, and the dimensions of the lifetime image are 44  44 mm.

G. Stokkan / Acta Materialia 58 (2010) 3223–3229

3225

Fig. 3. Fraction of wafer area with dislocation density >105 cm2 as a function of height. The error bars indicate manual correction of misinterpreted twins. When the correction is taken into account, the ingots may be grouped into a high and low dislocation density class.

Lifetime images are shown in Fig. 2, along with the dislocation density maps. It has been shown that lifetimes measured on rough wafer surfaces require a correction for increased emission to reveal the absolute lifetime [18]. This correction was not performed on these samples, and the measured lifetime values are not to be considered absolute, neither are the values in the two images comparable. The relative differences within one image are, however, correct. Crystal orientation was measured using electron backscattering diffraction (EBSD) on a large area (>50  50 mm2) for bottom wafers from all the ingots. The system used was a Zeiss Supra low-vacuum field emission gun scanning electron microscope equipped with TSL software for EBSD data collection and analysis. A combination of scanning the electron beam and moving the sample stage was used to construct the measurement of a large area. The EBSD system may give a systematic error in the absolute measured orientation. The measurements were therefore compared against an FZ sample with known (1 0 0) orientation. In the case of high deviation, the measurements were corrected using an inverse Euler matrix based on measurement of the FZ sample. The results have a deviation of less than 5°. Orientation imaging micrographs (OIM) showing the directions of the plane normals, as well as the inverse pole figures (IPF) for the five ingots are shown in Fig. 4.

class. The area of high dislocation density (>105 cm3) may be twice as large or more for the first group than for the second. The minority carrier lifetime images in Fig. 2 show that reduced lifetime is generally correlated to areas of high dislocation density or to grain boundaries. The areas of multiple twins do not show reduced lifetime. This indicates that the twins are not decorated by dislocations and that the manual correction of dislocation density in Fig. 3 is justified. The OIM and IPF for the five ingots in Fig. 4 show the following: Ingots 3 and 4, which are low dislocation density ingots, are dominated by normal orientations close to, but not exactly (2 1 1). Ingot 2, which is also a low dislocation density ingot, is dominated by orientations (2 3 0) and (4 4 1). These orientations are fairly close to (1 1 0). However, ingot 1 is dominated by orientations (2 1 2) and (0 0 1), and ingot 5 is dominated by (2 1 3) and (1 0 2). These two ingots are high dislocation density ingots. Neither of the observed orientations for these two ingots is close to the (2 1 1) or (1 1 0) orientations expected if the grains were grown from [1 1 0] or [2 1 1] dendrites growing close to the crucible bottom plane. All the ingots are, however, dominated by large grains, including the high dislocation density ingots. This is seen for ingot 5, in particular. The dominating orientations are marked with capital letters in Fig. 4, and the results are summarized in Table 1.

3. Results 4. Discussion The dislocation density measurements in Fig. 3 show that, for all the ingots, the dislocation density increases towards the top. Furthermore, the ingots divide into two groups of high and low dislocation density, respectively. Ingot 4 is apparently placed between the two levels of high and low density, but this is a result of the high twin density of this ingot. The manually estimated errors show that it is proper to group this ingot in the low dislocation density

The large grain size and the occurrence of a few preferred orientations for all the ingots show that the crystals are probably a result of growth from large planar crystals formed on the crucible bottom, not from random nucleation on the crucible bottom. This is confirmed by visual inspection of the ingot bottom: Large crystals are evident on the bottom plane, in contrast to the many small crystals

3226

G. Stokkan / Acta Materialia 58 (2010) 3223–3229

Fig. 4. EBSD data for bottom wafers from all the ingots. Both the OIM and the IPF are shown: (a) ingot 1; (b) ingot 2 (this image was obtained by scanning the stage, a method that gives more accurate orientation measurements, which explains the smaller amount of scattering in the IPF); (c) ingot 3 (the {2 1 1} orientation is marked by a red point in the IPF); (d) ingot 4 (note that the IPF was made on a different wafer from a higher position although showing the same crystal morphology; the measurement on this wafer was of higher quality, which allows for a discrimination of the different clusters in the IPF); (e) ingot 5 and (f) coloured IPF to interpret the OIM.

Table 1 Dominating plane normal directions for all the marked areas in Fig. 4. Area

Ingot 1

Ingot 2

Ingot 3

Ingot 4

Ingot 5

A B C D E F G H

849 1 1 30

16 1 25 15 5 18

522 19 12 12 6 5 19 11 10 13

12 12 19 9 8 22 6 5 15 15 13 21 15 13 18 9 7 27 13 13 16 5 5 18

231 10 2 19 9 1 22 653

expected as a result of random nucleation. A similar image is very often also seen on industrially produced ingots. The question is then whether the observed orientations can be linked to the [0  1 1] or [1  1 2] dendrites described by [12], i.e., if the physical principle is the same for this material, but produced in a much less controlled manner. In a patent application [19], the structure controlled growth is described to be achieved by growing dendrites along the crucible bottom such that the upper plane of the dendrite has an angle <10° to the (2 1 1) or (1 1 0) planes. It is interesting to investigate which angle the (2 1 1) or (1 1 0) planes would have to the normal direction if the observed crystal orientations were to be a result of growth from such dendrites. To investigate this, the IPF for the ingots shown in Fig. 4 were studied. The clusters of measurement data are marked with capital letters correspond-

ing to the areas marked in the OIM in the same figure. Consider ingot 3 as an example. The area consisting of the twinned grains marked A and B can be found in the IPF to be symmetrically arranged around the point marking the (2 1 1) orientation. Similarly, the area marked C and D is also symmetrically arranged around (2 1 1). The angles between A and B and (2 1 1) are 5.8° and 6.5°, whereas the C and D areas make angles 13.1° and 13.6° to (2 1 1), respectively. Thus, areas AB and CD can be explained to be nucleated from a [0 1 1] dendrite with a (2 1 1) plane rotated an angle 6° and 13°, respectively, to the growth direction. The twinned areas in ingot 4 show a similar symmetric arrangement around (2 11). The twinned areas of ingot 5 are not symmetrically arranged around (2 1 1) but around (1 1 0). The deviation angles are much larger, however. Ingot 2 is similar to ingot 5, but the angles to the (1 1 0) direction are much smaller. Given the symmetry observed in the IPF and the relation to the (2 1 1) and (1 1 0) plane directions, it is therefore reasonable to believe that ingots 3, 4, 2 and 5 have been nucleated by faceted dendrites with an angle to the crucible bottom. The IPF of ingot 1 does not show symmetry to either (2 1 1) or (1 1 0). The angles are summarized in Table 2. The misorientation between the crystal orientation and the (2 1 1) plane in the case of ingot 3 can be achieved in one of two ways: Either the [0 1 1] growth direction of the dendrite is parallel to the crucible bottom and the

G. Stokkan / Acta Materialia 58 (2010) 3223–3229

3227

Table 2 Angles between dominating orientations and [2 1 1] and [1 1 0] directions. For Ingot 2, 4 and 6 the angles to the [2 1 1] direction are reported and for ingot 3 and 9 the angles to [1 1 0] are reported.

A/B C/D E/F G/H

Ingot 1

Ingot 2

Ingot 3

Ingot 4

Ingot 5

15.2°/32.6°

12.5°/13.1°

5.8°/6.5° 13.1°/13.6°

6.5°/6.8° 8.2°/8.5° 12.8°/12.8° 13.7°/13.8°

19.1°/18.0° 22.9°/21.6°

(2 1 1) plane is tilted around this axis, or the [0  1 1] direction has an angle to the bottom and the plane may be tilted or not, i.e., the dendrite grows away from the crucible bottom at a shallow angle. It is interesting to see that, for ingot 3, the angle between the plane orientation and the [0 1 1] vector in area B is 90°, whereas the angle to the [1 1 1] vector lying in the dendrite plane is 95.8°. This means that the misorientation of the plane is caused exclusively by tilting of the dendrite plane, not by the dendrite growing at an angle to the crucible bottom. Similar numbers are found for the other areas for ingot 3. However, area A in ingot 5 shows an angle of crystal orientation of 83.4° to the growth direction [1  1 2] of the dendrite, whereas the angle to the [ 1 1 1] direction in the plane is 72.0°. This indicates that the misorientation is caused by the dendrite growing away from the crucible bottom. Since this will cause the dendrite to move into a region of insufficient supercooling, it is expected that the growing crystal will change its macroscopic direction of growth to counteract this, creating a crystal that grows along the crucible bottom in the process, as sketched in Fig. 5. Also in this respect, ingot 2 appears to be similar to ingot 5, but the angles are narrower (84.5° to the assumed growth direction [1  1 2] and 78.8° to [ 1 1 1] for area A).

Fig. 5. Schematic illustration of dendritic growth along the crucible bottom (left-hand side) and dendritic growth at an angle to the crucible bottom with direction changes (right-hand side) starting at a nucleus (small cube): (a) view from the top; (b) view from the side. The dendrite changes direction and grows downwards when it reaches a region of insufficient supercooling. Note that the dendrite growing at an angle to the crucible here shows the bottom face. The colours are taken from the reference triangle in Fig. 4f to represent a dendrite growing in a [2 1 1] direction.

Fig. 6. Furnace heating power as function of time after opening the heat leak. The furnace is temperature controlled, which means that the system will react to a sudden release of crystallization heat by a corresponding sudden power drop, as seen for ingots 2, 3 and 4.

If nucleation takes place as fast dendrites growing in a narrow supercooled region near the crucible bottom, the available volume will be filled quickly, resulting in a substantial amount of crystallization heat being released during a short time. If the dendrites grow into the melt, away from the crucible bottom, this may not necessarily be so, since the process of covering the bottom plane with crystals by the changing growth direction is slower. This difference may possibly be observed in the temperature and power profiles during growth. Indeed, the power profiles shown in Fig. 6 reveal that ingots 3 and 4 show a marked drop in the power profile immediately after nucleation has started. Since the furnace uses temperature as the control parameter, it will react on quick dissipation of heat into the system by reducing the power. Also ingot 2 shows this decrease in power, whereas ingots 1 and 5 show a gradual decrease in power. This is interpreted as resulting from dendrites growing along the crucible bottom (ingots 3 and 4) in a shallow region close to the bottom (ingot 2), and from dendrites growing away from the bottom at wider angles (ingots 1 and 5). Note that the concave crucible bottom of ingot 2 may influence the orientation of the dendrites so that this ingot may also be a result of nucleation along the bottom. Ingots 2, 3 and 4, which are believed to be a result of dendritic nucleation in a plane close to the crucible bottom, are characterized by low dislocation density, but also high occurrence of areas containing multiple twins. The opposite is true for ingots 1 and 5, i.e., high dislocation density and low twin density. This phenomenon of exclusive twin and dislocation occurrence is well known [20], [21] (for EFG material). Does the present study give any clues to the origin of this? The twin planes observed in MC silicon lie close to orthogonal to the growth interface, often extending along the entire height of the crystal. The twins are therefore most likely a growth phenomenon rather than a deformation phenomenon, a conclusion also supported by energetic calculations [22]. Fujiwara et al. [23] suggested

3228

G. Stokkan / Acta Materialia 58 (2010) 3223–3229

a mechanism for the formation of double twins from a faceted growth surface. In this model, a twin boundary is formed on one of the {1 1 1} facets. As growth proceeds, a twinned region is formed, with a twinned facet towards the melt. When this facet is overtaken by the advancing original facet, a new twin boundary is formed, parallel to the first twin boundary. This mechanism was suggested for the formation of parallel double twins in dendritic growth, but may also be valid for nucleation of twins in the growing MC ingot. However, when a two-dimensional growth interface is considered, the facets may form either long ridges or tilted pyramids, depending on the orientation of the {1 1 1} planes against the macroscopic growth interface. If twins are to be nucleated across the width of the grain by the suggested mechanism, it necessarily has to happen in a configuration allowing long ridges; otherwise the extent of the twin boundaries will be very limited. Fig. 7 shows a schematic of the faceted interface created on a (2 3 1) and a (5 2 2) macroscopic plane. Such long faceted ridges may only occur if the macroscopic growth plane contains a h1 1 0i vector, which is the trace of one {1 1 1} plane crossing another. The (2 1 1) plane that results from dendritic growth along the crucible bottom in the [0  1 1] direction necessarily contains

this vector. The (0 1 1) plane resulting from growth in the [2 1 1] direction contains the [0 1 1] vector. Thus it can be seen that both the planes resulting from the described dendrites growing along the crucible bottom have a preferred configuration for nucleation of twins by the suggested mechanism. For the material in this study, it can be seen that the macroscopic orientations of all the considered areas are very close to orthogonal to a h1 1 0i vector for ingots 3 and 4; most are actually orthogonal, and the worst deviation is an area of 87° rotation. Furthermore, the twins between A and B as well as C and D are parallel to h1 1 0i directions (interestingly the long twin separating AB from CD is parallel to a h2 1 1i direction). Ingot 5, in contrast, shows no macroscopic planes containing h1 1 0i vectors; the angles to the plane normal are 79.1° and 74.8° for the large area marked AB in Fig. 4e. The areas A and B in ingot 2 do not contain the h1 1 0i vector, but area B has an angle of 86.9° between the normal vector and the h1 1 0i vector. It also contains smaller areas with a favourable orientation close to (2 1 1). The planes of ingot 1 actually do contain a h1 1 0i vector without twin formation being promoted. It must be stressed that it is not necessary for the crystals to be nucleated by dendrites growing along the crucible bottom to have a favourable configuration for this nucleation mechanism of parallel twins, only that the planes contain a h1 1 0i vector. However, the ingots nucleated in this manner do have this favourable configuration, which may explain the high occurrence of twins in this material. Material nucleated by dendrites growing along the crucible bottom have large grains and may, according to the mechanism suggested by Ryningen et al. [13], therefore develop fewer dislocations. This, combined with the suggested mechanism for preferential twin nucleation in such material, may explain why a high defect density and a high occurrence of multiply twinned areas seem to be exclusive phenomena. Another possible explanation may be the microscopic aspects of atomic adhesion during growth. It has been described [24] that the growth interface requires steps where adhesion of atoms may readily occur. A continuous supply of such growth-assisting steps can be multiple twins as well as screw dislocations [24]. If certain orientations have a preference for nucleating such multiple twins and the nucleation phase promotes the existence of such planes, it is possible for twin-dominated growth to develop for such systems, depressing the evolution of screw dislocations which could subsequently act as sources for dislocation multiplication through the process of cross slipping and formation of Frank Read sources [13]. 5. Conclusion

Fig. 7. Schematic view of a faceted two-dimensional solid–liquid interface: (a) {1 1 1} facets on a macroscopic (2 3 1) plane; (b) {1 1 1} facets on a (5 2 2) plane forming long ridges, from which multiple twins may be nucleated along the width of the crystal.

It was shown that silicon ingots produced in a pilot scale directional solidification furnace may have a low or high density of dislocation clusters. The low dislocation density ingots show a high twin density and vice versa. Lifetime

G. Stokkan / Acta Materialia 58 (2010) 3223–3229

measurements indicate that the twinned areas have an increased lifetime compared with the areas of high dislocation density. From studies of OIM and IPF, it was shown that the low dislocation density ingots are most likely caused by nucleation on dendrites grown on the crucible bottom or at a shallow angle to it. The high dislocation density ingots are probably also grown from such dendrites, but here the dendrites have probably grown away from the crucible bottom at a wider angle, with subsequent change in growth direction. This is similar to the process described in Ref. [12], but occurs here unintentionally and with less control. The study thus provides empirical evidence that such nucleation on dendrites provides an environment that generates fewer crystal defects and has potential for improved electrical properties. The low dislocation density ingots have a high density of twins. It has been shown that the dominating crystal orientations of these ingots have growth planes that contain h1 1 0i vectors or have a shallow angle to this vector. Double twin nucleation from long ridges between h1 1 1i facets on the growth interface may therefore readily occur to fill the entire width of a crystal. Both the described types of dendrites growing along the crucible bottom will result in the formation of planes containing this vector. This has been suggested as the reason for the high occurrence of twins in this material. Furthermore, the formation of favourable adhesion points at these twin facets rather than on screw dislocations has been suggested as a possible reason for the reduced dislocation density. Acknowledgments This work was performed in the project “Crystalline Silicon Solar Cells—Cost Reduction” and “Defect Engineering in Crystalline Silicon Solar Cells” sponsored by the Norwegian Research Council, Elkem Solar and REC. Otto Lohne and Lars Arnberg are acknowledged for helpful discussions, and Espen Olsen, Eivind Øvrelid and Arve Holt for cooperation in growing the crystals studied and preparing the samples.

3229

References [1] Hirshman WP, Faidas M, Finis A-L, Jin Y, Knoll B, Li R. Photon International, vol. 3. Aachen: Photon Europe GmbH; 2009. p. 170. [2] Swanson RM. Prog Photovoltaics Res Appl 2006;14:443. [3] Green MA, Emery K, Hishikawa Y, Warta W. Prog Photovoltaics Res Appl 2009;17:85. [4] Davis Jr JR, Rohatgi A, Hopkins RH, Blais PD, Choudhury PR, McCormick JR, et al. IEEE Trans Electron Devices 1980;ED-27:677. [5] Sopori B, Li C, Narayanan S, Carlson D. Efficiency limitations of multicrystalline silicon solar cells due to defect clusters. Warrendale (PA) USA: Materials Research Society; 2005. p. 233. [6] Stokkan G, Riepe S, Lohne O, Warta W. J Appl Phys 2007;101:53515. [7] Kveder V, Kittler M, Schroter W. Phys Rev B (Condens Matter Mater Phys) 2001;63:115208. [8] Buonassisi T, Pickett MD, Istratov AA, Sauar E, Lommasson TC, Marstein E, et al. In: Conference record of the 2006 IEEE 4th world conference on photovoltaic energy conversion, vol. 1. 2006. p. 944. [9] Chen J, Sekiguchi T, Xie R, Ahmet P, Chikyo T, Yang D, et al. Scripta Mater 2005;52:1211. [10] Chen J, Sekiguchi T, Yang D, Yin F, Kido K, Tsurekawa S. J Appl Phys 2004;96:5490. [11] Bentzen A, Holt A, Kopecek R, Stokkan G, Christensen JS, Svensson BG. J Appl Phys 2006;99:93509. [12] Fujiwara K, Pan W, Usami N, Sawada K, Tokairin M, Nose Y, et al. Acta Mater 2006;54:3191. [13] Ryningen B, Stokkan G, Lohne O. Growth of Dislocation Clusters in Directionally Solidified Multicrystalline Silicon. 23rd European Photovoltaic Solar Energy Conference. Valencia, SPAIN: WIP Munich; 2008. [14] Sopori BL. J Electrochem Soc 1984;131:667. [15] Sopori B, Wei C, Yi Z, Madjdpour J. J Cryst Growth 2000;210:346. [16] Stokkan G. In: 21st European photovoltaic solar energy conference. Milan, Italy: WIP-Munich; 2007. p. 4. [17] Schubert M, Isenberg J, Warta W. J Appl Phys 2003;94:4139. [18] Schubert MC, Pingel S, The M, Warta W. J Appl Phys 2007;101. [19] Fujiwara K, Nakajima K, editors. In: Office USPaT. Japan: Tohoku University; 2009. p. 18. No. 20090000536. [20] Khattak CP, Schmid F, Cudzinovic M, Symko M, Sopori BL. Analysis and control of the performance-limiting defects in HEM-grown silicon for solar cells. Pittsburgh (PA, USA): Mater Res Soc; 1995. p. 767. [21] Mo¨ller HJ, Funke C, Rinio M, Scholz S. Thin Solid Films 2005;487:179. [22] Pirouz P. Scripta Metall 1987;21:1463. [23] Fujiwara K, Maeda K, Usami N, Sazaki G, Nose Y, Nakajima K. Scripta Mater 2007;57:81. [24] Porter DA, Easterling KE. Phase transformations in metals and alloys. Boca Raton (FL): CRC Press; 2004.