About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Advances in Ceramic Materials and Processing
|
| Presentation Title |
Neural network interatomic potentials for simulating dislocation plasticity in ceramics at the atomic scale |
| Author(s) |
Shihao Zhang, Shigenobu Ogata |
| On-Site Speaker (Planned) |
Shihao Zhang |
| Abstract Scope |
Dislocations in ceramics are gaining increasing attention for their potential to enhance the toughness of brittle materials and to engineer functional properties. However, atomistic simulation of dislocation behavior remains difficult due to the complex interplay of ionic and covalent bonding and the presence of highly distorted, non-stoichiometric dislocation cores in intricate crystal structures—features that are poorly described by empirical interatomic potentials. In this study, we develop neural network potentials (NNPs) to model dislocation plasticity in representative ceramics, including ZnO, GaN, and SrTiO₃. These NNPs offer DFT-level accuracy and transferability with far greater computational efficiency. They accurately reproduce charged dislocation cores and slip barriers, as validated against DFT and experiments. Using the developed NNPs, we perform large-scale simulations of nanoindentation and nanopillar compression that reveal detailed dislocation behavior in excellent agreement with experimental observations. This work demonstrates the effectiveness of NNPs for reliable and efficient atomistic simulations of ceramic plasticity. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Ceramics, Machine Learning, Modeling and Simulation |