Speaker
Description
The resilience of plasma-facing materials (PFMs) under intense radiation is a significant challenge for feasible fusion reactors. Computational predictions of cumulative radiation damage are needed to explain experimental observations but are obscured due to the complex ultrafast dynamics that occur after energetic collisions with the lattice. As a result, scientific understanding of PFM degradation is inadequate, creating the need for model advancement. This project seeks to overcome aforementioned challenges through advanced computational techniques to study radiation damage accumulation in tungsten.
We introduce a novel class of highly precise machine-learned interatomic potentials (ML-IAP) for Molecular Dynamics (MD) simulations that directly account for electronic structure changes when shifting from ground state to warm-dense matter. ML-IAP are trained on a diverse set of density functional theory (DFT) defect calculations to directly capture the altered electronic states that occur during initial stages of neutron collisions. Employment of these excited state ML-IAP can significantly improve the accuracy of mesoscale damage accumulation models (NRT-, CRC-DPA) as MD is the determining factor in their parameterization. In particular, the consequences of accounting for excited state dynamics in MD are amplified at higher energies (>10keV) of the recoil spectra. This research demonstrates advancements in machine learning for materials development in extreme conditions and illustrates the importance of semiclassical dynamics in understanding radiation damage accumulation in PFM. Extensions of the work to other excited state processes in material is also discussed.