Prediction of materials properties from core-loss spectrum using neural network*

  Data driven approaches are now indispensable for modern materials characterization due to rapid increase of size and dimension of data observed in experiments and simulations. Based on this backgrounds, we are developing data-driven methods for the materials characterizations.

  The group of presenter has reported applications of machine learning for XAFS/EELS spectrum. XAFS/EELS is core-loss spectroscopy observed using X-ray/electron, and the spectral feature around near edge is called XANES/ELNES and it reflects partial density of states of conduction band.

  We have applied artificial neural networks for the core-loss spectrum [1-3]. For instance, the excited states that reflects the spectrum was predicted from their ground states [1], and radial distribution function was directly obtained from the XANES/ELNES [2]. In my presentation, I will mainly report on the predictions of materials properties from core-loss spectrum using the neural network. By using the C-K edge spectra as the descriptor for the neural network, totally 11 properties of organic molecules have been predicted [3].

[1] S. Kiyohara, et al., npj Comp. Mater., 6 (2020) 68-1-6.

[2] S. Kiyohara et al., J. Phys. Soc. Jpn (Letter), 89 (2020) 103001-1-4.

[3] K. Kikumasa et al., Adv. Intel. Sys., 3 (2021) 2100103-1-10.