[2Aa01-14] データ駆動表面科学研究部会「データ駆動アプローチ研究の最近の動向」

[2Aa08] Machine learning molecular dynamics simulations of silicate minerals

*Masahiko Okumura¹ (1. Center for Computational Science and e-Systems, Japan Atomic Energy Agency)

Silicate minerals are abundant on the earth and are one of the most familiar environmental substances. However, their physical characteristics have not been entirely understood. We have applied the recently proposed "machine learning molecular dynamics method" to silicate minerals, which realizes high accuracy and low calculation cost. The new method enables us to evaluate chemical processes and physical quantities that are difficult to evaluate using the classical molecular dynamics method and first-principles molecular dynamics method. In this talk, those results will be reported.