Modelling hydrogen embrittlement of nuclear fuel cladding

Hydrogen embrittlement is the limiting factor that affects how long we can keep fuel in the reactor, and how much energy we can extract from fuel before discharging it. This project uses atomic-scale modelling techniques, accelerated by machine-learning methods to simulate the formation and growth of hydrides in the zirconium fuel cladding. The researchers are also using state-of-the-art experimental techniques – including in situ electron backscatter diffraction (EBSD) and cryogenic atom probe tomography (cryo-APT) – to investigate the hydrogen-trapping mechanism that may be used to develop a mitigation strategy for hydrogen embrittlement.

In 2024, the team developed new machine-learning inter-atomic potentials that accelerate the simulation by several orders of magnitude, while maintaining most of the accuracy of quantum mechanical methods.