[9p-S301-1~12] 23.1 Joint Session N "Informatics"

1:00 PM - 1:15 PM JST(4:00 AM - 4:15 AM UTC)

[9p-S301-1] Evaluation of phase transition and dielectric properties using GNN model

〇(M1)Koki Yoshimochi¹, Alex Kutana¹, Ryouji Asahi¹ (1.Nagoya Univ.)

Keywords:

machine learning,graph neural network,phase transition

We investigated phase transitions and dielectric properties of CaTiO₃ using newly developed rotational-equivariant graph neural network model (EquivarGNN). By performing local structural phase classification, we observed high-temperature phase transitions, which are not clearly determined through macroscopic analysis such as lattice constants. The predicted dielectric constants at 0 K reproduced DFPT results, and their temperature dependencies were consistent with experimental observations.