How Can We Decrease Defect Density in Si Multicrystals to Realize High-Efficiency Solar Cells?

Abstract

In order to answer the question, we started with model crystal growth experiments using a seed crystal to consist of a couple of single crystals with an artificially controlled grain boundary configuration, namely tilt grain boundary. Spatially resolved X-ray rocking curve analysis and evaluation of etch-pit density clarified that “sub-grain boundaries”, which consist of a cluster of dislocations and give detrimental effect on solar cell performance, are generated from the grain boundary as crystal growth proceeds. Furthermore, the amount of dislocations was found to be strongly dependent on the grain boundary configuration. In a particular configuration, dislocations were mostly localized in one of the grains to consist of a grain boundary [1]. Since Si expands when crystallized from the melt, compressive stress is supposed to be applied owing to the contact with a rigid crucible. This led us to assume that local shear stress around a grain boundary, which would depend on the configuration of the grain boundary, might explain the experimentally observed difference in defect density. To see if this hypothesis is reasonable, we carried out threedimensional finite element analysis of shear stress for a model bi-crystal with a single grain boundary. By taking anisotropic elastic constants into account, calculations were performed for various grain boundary configurations. Interestingly, dislocation density was found to be well correlated with the calculated local shear stress around the grain boundary. The shear stress is likely to be reduced when the rotational symmetry is improved along the direction perpendicular to the plane where external stress from the crucible is introduced. This knowledge inspired us to suppress dislocations during directional solidification by controlling the coherency of the grain boundary without any seeds so that one could decrease shear stress around the grain boundary. The attempt to change the coherency of the grain boundary was made by utilizing nucleation of plural dendrite crystals. Since the upper plane of the dendrite crystal can be limited to either {110} or {112}, the coherency of the grain boundary can be controlled by the contact angle between adjacent dendrite crystals. Crystal growth was started with dendritic nucleation by establishing the amount of supercooling larger than 10 K. The contact angle was measured by using electron back scattering pattern analysis. Dislocation density measured by counting etch-pits revealed that parallel contact of adjacent dendrite crystals is useful to suppress generation of dislocations. On the other hand, increase of the contact angle resulted in increase of the dislocation density. This confirms that control of the coherency of grain boundaries at the initial stage is very important for decreasing dislocations in the whole ingot [2].