About this Abstract |
Meeting |
2020 TMS Annual Meeting & Exhibition
|
Symposium
|
Computational Materials Science and Engineering of Materials in Nuclear Reactors
|
Presentation Title |
Development and Testing of Machine Learning Interatomic Potentials for Radiation Damage Calculations |
Author(s) |
Kai Nordlund, Ali Hamedani, Jesper Byggmästar, Flyura Djurabekova |
On-Site Speaker (Planned) |
Kai Nordlund |
Abstract Scope |
Machine learning interatomic potentials have gained major attention in recent years due to their increased flexibility compared to analytical potentials. However, most potentials are trained against equilibrium properties only, and it is not clear whether they also work in radiation damage calculations. We have now taken into use the Gaussian approximation potentials (GAP) framework for radiation damage calculations in Si and W. Tests of the as-published potentials showed that they do not describe the high-energy repulsive part of the interaction in any realistic manner, and hence it is crucial to modify the GAP framework for dealing with radiation effects. In the talk, we describe our approach for doing this and show initial results on damage calculations in low-energy cascades. |
Proceedings Inclusion? |
Planned: Supplemental Proceedings volume |