[18p-D103-1~23] 15.7 Crystal characterization, impurities and crystal defects

[18p-D103-16] Pattern Recognition of Etchpits on Unpolished Surface of Multicrystalline Silicon Using Machine Learning

〇Takuto Kojima¹, Atsushi Ogura¹, Kenji Fukui², Manabu Komoda², Junichi Atobe² (1.Meiji Univ., 2.Kyocera Corporation)

Keywords:etchpit, machine learning, multicrystalline silicon

The carrier lifetime of the substrate is an important factor determining the conversion efficiency. On the other hand, the carrier lifetime differs depending on the distribution of light elements and point defects in the sensitivity to the thermal process in the fabrication process. Therefore, the relationship between the carrier lifetime before the process and the final conversion efficiency is complicated, and evaluation in combination with other indicators becomes important. One of the indices of the substrate quality is an etchpits density in which defects are manifested by a selective etching solution. In this study, we attempted automated counting of etchpits coexisting slice marks using machine learning on non-polished substrate.