[21a-B01-1~10] 23.1 Joint Session N "Informatics"

[21a-B01-1] [Young Scientist Presentation Award Speech] Development of data-driven analytical methods for ELNES/XANES

〇Shin Kiyohara¹, Masashi Tsubaki², Teruyasu Mizoguchi¹ (1.IIS, Univ. Tokyo, 2.AIST)

Keywords:ELNES/XANES, machine learning, neural network

ELNES/XANES reflects local conduction states, and thus we can analyze local atomic structure and chemical bonds by combining with theoretical calculations. However, it is not straightforward to extract such information from ELNES/XANES spectra. Here, we tried to quantify material properties and physical quantities directly by artificial neural network. As the results, we achieved to predict geometrical and chemical bonding properties directly and apply the constructed prediction model to an experimental spectrum.