[2P01-53] Poster Presentation

[2P13] Analysis of the ionic behavior in zirconia under an applied electric field using machine learning potentials

*Naoki Maekawa¹, Koji Shimizu¹, Satoshi Watanabe¹ (1. Graduate School of Engineering, the University of Tokyo)

As a first step to clarify the microscopic mechanism of ductility enhancement of yttria-stabilized zirconia ceramics by applying an electric field, a high-dimensional neural network potential (HDNNP) that can predict ion behaviors under electric fields has been constructed for zirconia. The constructed HDNNP demonstrated accurate predictions for energy and forces. To consider ion behaviors under electric fields, a neural network (NN) was constructed for prediction of the Born effective charge. In generating the training dataset for the Born effective charge, anomalously high values were observed in some cases. They were attributed to metallic behavior due to Fermi level positioning, and refining computational conditions led to improved predictions. In the presentation, results of MD calculation using the constructed HDNNP and NN will also be shown.