[18p-A24-1~13] 23.1 Joint Session N "Informatics"

2:45 PM - 3:00 PM JST(5:45 AM - 6:00 AM UTC)

[18p-A24-5] [The 56th Young Scientist Presentation Award Speech] Direct inverse analysis of defect distribution and electron transport in field-effect transistors by tandem neural network

〇Masatoshi Kimura¹, Keisuke Ide¹, Hideo Hosono¹, Toshio Kamiya¹ (1.Tokyo Tech)

Keywords:

Machine learning,Inverse problem analysis,Thin film transistor

We have developed a machine-learning method for multi-valued inverse problem analysis and applied it to prediction of material properties in thin film transistors directly from their operation characteristics. To solve the issue of the multi-valued analysis, a forward analytical model is connected after a general inverse analytical model in which descriptors and objective variables are swapped. It is found that material properties reflecting the multivaluedness are predicted, and TFT characteristics can be reproduced from these material properties with high accuracy coefficient of determination of approximately 0.99.