[18p-A401-1~13] 23.1 Joint Session N "Informatics"

[18p-A401-10] Feature Selection in Machine Learning to Discover Sublimable Substances for Pattern Collapse Mitigation

〇Shogo Kunieda¹, Yuta Sasaki¹, Yuichiro Hikida¹, Koki Uonami¹, Kazuyuki Hashimoto², Yosuke Hanawa¹ (1.SCREEN Holdings, 2.SCREEN Advanced System Solutions)

Keywords:semiconductor process, Materials Informatics, sublimation drying

As the miniaturization and three-dimensionalization of semiconductor devices progresses, the problem of pattern collapse during the semiconductor cleaning process is becoming more serious. In the sublimation drying method, it is important to find a sublimation material that can effectively mitigate the pattern collapse phenomenon. In this study, by creating a machine learning model that predicts the experimental pattern collapse rate from chemical structures of organic molecules, we have selected a sublimation material that is highly effective in suppressing pattern collapse phenomenon, and determined the key factors from the feature importance.