[18p-S2_204-1~10] 23.1 Joint Session N "Informatics"

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

[18p-S2_204-4] Elucidating the temperature dependence of dielectric constant in BaTiO₃ using machine learning interatomic potentials

〇Daiki Nakajima^1,2, Shin Kiyohara², Atsushi Takigawa^1,2, Yu Kumagai^2,3 (1.GSE,Tohoku Univ., 2.INR,Tohoku Univ., 3.OAS,Tohoku Univ.)

Keywords:

Materials Informatics,Machine Learning,Ferroelectrics

BaTiO₃exhibits a sharp change in dielectric constant in the vicinity of the Curie temperature, and flattening this temperature dependence is an important issue for applications in electronic components. In this study, we constructed a machine learning interatomic potential, Allegro-pol, which can directly treat dielectric responses, using datasets obtained from Ab Initio Molecular Dynamics and Density Functional Perturbation Theory. As a result, we confirmed that physical quantities such as energy, forces, polarization, and polarizability can be reproduced with an accuracy comparable to that of First-principles calculations. These results will be reported in the presentation.