| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Innovations in Energy Materials: Unveiling Future Possibilities of Computational Modelling and Atomically Controlled Experiments
|
| Presentation Title |
Specialized Machine Learning Interatomic Potential to assess Self-Healing at a W Grain Boundary |
| Author(s) |
Jorge Suárez-Recio, Pablo Piaggi, Javier Domínguez-Gutiérrez, Raquel González-Arrabal, Roberto Iglesias |
| On-Site Speaker (Planned) |
Roberto Iglesias |
| Abstract Scope |
The development of plasma-facing materials capable of withstanding extreme conditions of irradiation and thermal loads is a cornerstone of nuclear fusion research. Nanostructured tungsten with its density of grain boundaries and eventual self-healing properties is a prominent candidate. DFT simulations may describe physical phenomena such as electronic effects at interfaces or open boundary free-surfaces at an expensive computational cost. Machine learning algorithms may help in addressing the accuracy versus efficiency dilemma in molecular dynamics simulations. Here, the development of a specialized MLIP for a tungsten GB is described and applied to study the recombination of intrinsic defects as a function of defect density and temperature in the presence of light impurity atoms. We employ he open source toolkit FaVAD for automated defect analysis of crystalline materials, providing a distance metric suitable for classification. The advantages and limitations of our methodologies and a comparison to published research will be presented. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, Nuclear Materials |