[18a-A401-1~9] 23.1 Joint Session N "Informatics"

[18a-A401-2] Shape Prediction of Carbon K-edge ELNES/XANES for Organic Polymer Analysis via Machine Learning

〇Kiyou Shibata¹, Teruyasu Mizoguchi¹ (1.IIS, the Univ. of Tokyo)

Keywords:core loss spectrum, machine learning, database

Carbon-containing polymers and composites have been the subject of research in various fields, requiring the development of methods for microscopic analysis of their structures and compositions. In this study, we applied machine learning to a carbon K-edge spectral data set obtained by first-principles calculations to predict the shape of carbon K-edge spectra from the structure of organic molecules at low cost. As a result, the shape of the carbon K-edge spectrum was successfully predicted from the molecular structure using a graph convolutional neural network.