Uranium oxide (UO2) is the usual fuel used in nuclear fission power plants. As such, its behaviour under irradiation has been extensively studied. Irradiation creates vacancies or interstitial defects that control the evolution of the material's microstructure, which in turn impacts its physical (e.g. thermal conductivity) and mechanical properties. Interstitial clusters in particular play a major role.
On the one hand, at the smallest sizes, the diffusion of interstitials in UO2 is still relatively poorly understood. Experimentally, we observe the appearance of dislocation loops made up of interstitials as large as ten nanometres. Conversely, no cavities are observed and the vacancy defects remain sub-nanometric in size. This indicates that interstitials diffuse more rapidly than vacancies, with diffusion allowing interstitials to agglomerate and form loops. However, atomic-scale calculations show no major difference between the diffusion coefficients of vacancies and interstitials in UO2. One hypothesis to explain this apparent contradiction is that interstitial clusters diffuse rapidly (Garmon, Liu et al. 2023).
On the other hand, the three-dimensional interstitial clusters are expected to be the seeds of the dislocation loops observed by transmission electron microscopy in irradiated uranium oxide. However, the mechanisms by which the aggregates transform into loops and the nature of the loops changes remain poorly understood in uranium oxide. These mechanisms have very recently been elucidated for face-centred cubic metals (Jourdan, Goryaeva et al. 2024). It is possible that comparable mechanisms are at work in UO2 with the complication induced by the existence of two sub-lattices.
We therefore propose to study interstitial clusters in UO2 using atomic-scale simulations.
We will first study the structure of these three-dimensional subnanometric clusters. To do this, we will use artificial intelligence tools for classifying defect structures developed in the laboratory (Goryaeva, Lapointe et al. 2020). We will study the diffusion of these objects using molecular dynamics and automatic searches for migration saddle points using kinetic-ART type tools (Béland, Brommer et al. 2011). Secondly, we will study the relative stability of 3D clusters and loops of faulted and perfect dislocations and the transformations between these different objects.
This study will be based on interatomic interaction potentials. We will start by using empirical potentials available in the literature before turning to Machine Learning-type potentials (Dubois, Tranchida et al. 2024) under development at the CEA Cadarache Fuel Studies Department.

Béland, L. K., et al. (2011). ‘Kinetic activation-relaxation technique.’ Physical Review E 84(4): 046704.

Chartier, A., et al. (2016). ‘Early stages of irradiation induced dislocations in urania.’ Applied Physics Letters 109(18).

Dubois, E. T., et al. (2024). ‘Atomistic simulations of nuclear fuel UO2 with machine learning interatomic potentials.’ Physical Review Materials 8(2).

Garmon, A., et al. (2023). ‘Diffusion of small anti-Schottky clusters in UO2.’ Journal of Nuclear Materials 585: 154630.

Goryaeva, A. M., et al. (2020). ‘Reinforcing materials modelling by encoding the structures of defects in crystalline solids into distortion scores.’ Nature Communications 11(1).

Jourdan, T., et al. (2024). ‘Preferential Nucleation of Dislocation Loops under Stress Explained by A15 Frank-Kasper Nanophases in Aluminum.’ Physical Review Letters 132(22).