| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Meeting Materials Challenges for the Future of Fusion Energy
|
| Presentation Title |
Predicting High-Dose Irradiation Damage Using Statistical Surrogate Models of Cascade Defect Production |
| Author(s) |
Chris Race, Bartosz Barzdajn |
| On-Site Speaker (Planned) |
Chris Race |
| Abstract Scope |
Molecular dynamics simulations of collision cascades are now routine and accessible with modest computing infrastructure. However, damage accumulation at longer timescales involves the cumulative effects of many, many individual sequential damage events, occurring in parallel with diffusive processes of microstructural evolution. Direct simulation to high dose with molecular dynamics therefore remains prohibitively expensive, other than for a small number of benchmarking studies.
Here we discuss the use of hierarchical statistical models to capture the statistics of damage production as a function of PKA energy, irradiation temperature and preexisting microstructural features. Viewed as a probabilistic generative surrogate model for cascade damage, this approach provides a means of producing samples of damage production with atomic resolution at near-zero computational cost. Damage distributions drawn as samples from these models can be used as inputs for longer length and timescale models and as part of a modelling chain to predict high-dose damage accumulation. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Nuclear Materials, Modeling and Simulation, |