Atomic Scale Modelling of the Properties and Behaviour of Oxide Nuclear Fuels

Ensuring the safety and performance of nuclear reactors requires a comprehensive understanding of the behaviour of nuclear fuels under various conditions. Fuel performance codes, such as the PLEIADES multiphysics platform developed by CEA in collaboration with EDF and Framatome, are essential tools for this purpose. Central to this effort is the precise knowledge of physico-chemical properties across various fuel types, coupled with the formulation of behaviour laws integrating experimental data and modelling results. Leveraging a multiscale modelling approach that starts at the atomic scale, the activities in our laboratory aim at providing input data and laws to the PLEIADES platform. In this context, machine-learning methods are important complementary tools to accelerate traditional methods and facilitate the bridging between the modelling scales.<br/>We will present our activities that make use of a combination of electronic-structure and empirical potential simulations to gain deeper insights into fuel properties and behaviour. We will discuss the development and application of various interatomic potentials for actinide oxides (UO₂, (U,Pu)O₂, (U,Am)O₂), including machine-learning potentials trained on extensive electronic-structure databases. We will then discuss the investigation of the temperature-dependent and diffusion properties that are crucial for understanding the thermal and irradiation-induced microstructure evolution, respectively. We use the results obtained to provide useful insights on the fuel physical behaviour and experimental measurements. Finally, we will demonstrate how generative machine-learning tools can help accelerating the study of compounds exhibiting chemical disorder, such as mixed actinide oxides.<br/><br/>These activities are included in the PUMMA and PATRICIA projects, which have received funding from the Euratom research and training programme 2019-2020 under grant agreements No 850945022 and 850945077.