Artificial Intelligence (AI) plays nowadays a key role in the design of innovative materials for the transition to low-carbon electricity production. In particular, AI generative methods that have led to well-known text/image generation tools can be used as well in the modeling of nuclear materials, to enhance reactor efficiency and safety. Over recent years, our laboratory has been actively working on such methods to accelerate the calculation of atomic-scale properties – a fundamental step in advancing our understanding of the physical phenomena resulting from the irradiation of these materials. Some of them are chemically disordered, which entails a random distribution of the chemical elements on the crystal lattice and inherent challenges in dealing with the astronomical number of resulting atomic configurations. The goal of the generative methods currently under investigation is to generate a set of representative configurations, for a rapid and accurate estimation of the desired property.
The objective of this thesis is to continue the development of these methods and apply them to determine the properties of crystal defects and fission gases that underlie the irradiation-induced microstructure evolution. The work will focus on actinide mixed oxides and high-entropy multicomponent alloys. The former are used to optimize the consumption of fissile material and move towards the closure of the fuel cycle, while the latter are currently seen as a highly promising alternative to conventional alloys for improving structural material properties. This project represents a cornerstone of our research efforts, as it will produce a significant amount of data for multiscale models that simulate the behavior of these materials in nuclear reactors.
The work will take place at the Nuclear Fuel Department of the IRESNE Institute at the CEA center of Cadarache in the south of France, in a team consisting of several materials modeling experts, in close collaboration with another CEA team in the Paris region specialized in artificial intelligence. The findings will be disseminated through scientific publications and participation in both national and international conferences. This PhD thesis will enable the candidate to acquire essential skills in materials science, advanced machine learning methods, data analysis, and software development, which will be valuable for a future career in academic or industrial research in the fields of AI and materials engineering.