[2P01-29] Poster session

[2P05] Toward electric-field molecular dynamics near ZrO₂ grain boundaries with an electric-field-aware machine learning interatomic potential

*Shungo Arai¹, Philip Loche², Anh Khoa Augustin Lu^1,3, Marcel F. Langer⁴, Michele Ceriotti⁴, Satoshi Watanabe¹ (1. The University of Tokyo (Japan), 2. Technical University of Munich (Germany), 3. National Institute for Materials Science (Japan), 4. École Polytechnique Fédérale de Lausanne (Switzerland))

Electric-field-assisted plastic deformation in ZrO₂ ceramics is thought to involve enhanced ion transport near grain boundaries (GBs), but its atomistic mechanism remains unclear. To address this issue, we developed an electric-field-aware machine-learning interatomic potential using the LOREM architecture. The model predicts energies, forces, and Born effective charge tensors with long-range electrostatic effects, allowing field-induced forces to be evaluated efficiently in molecular dynamics. It was trained on 11,419 pristine and defective ZrO₂ structures, including oxygen-vacancy, zirconium-vacancy, and Frenkel-defect configurations, and showed accurate validation performance. The model also transferred to unseen Σ5(310)/[001] GB structures. Preliminary MD simulations show a tendency toward enhanced ion drift under external electric fields near the ZrO₂ GB structures compared with the bulk.

Password required to view