Presentation Information
[2G04]Study of electric field-induced ion diffusion in defect-laden tetragonal ZrO2 with charge-aware machine learning potentials
*Anh Khoa Augustin Lu1,2, Naoki Maekawa1, Akane Ikeda1, Koji Shimizu3, Hiroshi Masuda1, Hidehiro Yoshida1, Satoshi Watanabe1 (1. The University of Tokyo, 2. National Institute for Materials Science, 3. National Institute of Advanced Industrial Science and Technology)
High-strength structural ceramics are typically limited by brittleness. However, under a strong electric field, yttria-stabilized zirconia (YSZ) undergoes ultrafast sintering with enhanced mechanical properties. As Joule heating alone cannot explain this, electric field-induced nonthermal phenomena in tetragonal zirconia must be elucidated. To study these effects, we developed a neural network potential that includes prediction of Born effective charges (BEC). Training data was generated by first-principles calculations, including both pristine and defect-laden zirconia. Our model achieves high accuracy with a root mean squared error of 5.1 meV/atom for energy, 0.099 eV/Å for forces and 0.0936 e/atom for BEC. Molecular dynamics simulations for zirconia using this model showed the formation of defects and enhanced ion mobility under electric field. Our findings represent a significant milestone for investigations of defect-laden materials under an electric field.