Speaker
Description
Nuclear materials are subjected to extreme environments involving high-flux irradiation, strong temperatures gradients, and severe mechanical stresses. These conditions induce complex macroscopic behavior and property evolution, originating from microstructural phenomena and below, with diffusion playing a crucial role. Understanding and predicting this evolution requires multiscale modeling approaches that link atomic-level processes to continuum-scale behavior.
We focus in this work on fission gas behavior in polycrystalline uranium dioxide, integrating thermomechanical evolution to improve our understanding on fission gas release, swelling, and high-burnup restructuring. A multiphase, multifield phase-field approach simulates grain boundary and gas bubble evolution under temperature and stress gradients, with irradiation-induced defect and gas source terms. The phase-field model is parameterized using atomic-scale calculations of free energy and diffusion data.
We obtain atomic-scale diffusion coefficients through direct simulations of diffusion trajectories or via mean-field approaches, and explore diffusion in multicomponent solutions, such as mixed actinide oxides or high-entropy alloys, using a generative machine-learning approach to optimize computational time [1]. This multiscale methodology can be relevant for fusion materials, where irradiation-induced microstructural heterogeneities impact material properties and behavior. Understanding and predicting these phenomena by integrating relevant physical data and behavioral laws into higher-scale models can improve the physical reliability and predictive power of macro-scale simulations, contributing to the development of robust and safe nuclear materials.
References:
[1] M. Karcz et al., “Targeting the partition function of chemically disordered materials with a generative approach based on inverse variational autoencoders”, arXiv:2408.14928 (2024).
