Towards Hydrogen Imaging and Mapping in Embrittlement-Susceptible Steels via XAS and Machine Learning

Hydrogen is an important component of future low-carbon energy systems. However, its wider use is still limited by hydrogen embrittlement, particularly in structural alloys such as stainless steels, where hydrogen can compromise strength, ductility and long-term reliability. Direct X-ray detection of hydrogen is extremely challenging because of its low electron density and weak interaction with X-rays. Alternative approaches include neutron imaging, in which neutrons are directly sensitive to hydrogen but limited by low flux and slow acquisition, whereas cryogenic atom probe tomography provides nanoscale chemical analysis but is destructive, volume-limited and susceptible to background contamination.

In this study, we use a computational model of 316L stainless steel to identify an indirect but measurable signature of hydrogen. Molecular Dynamics simulations using LAMMPS are combined with FEFF10 calculations to investigate how hydrogen-induced strain in Fe–Fe interatomic distances affects the local X-ray absorption response. The results show that these changes in the iron lattice produce subtle but quantifiable variations in the EXAFS oscillations. A supervised ML regression model is then trained to detect these spectral features, allowing both the presence and concentration of hydrogen to be inferred. These findings provide a basis for rapid, large Field-of-View indirect mapping of hydrogen using synchrotron-based X-ray Absorption Spectroscopy.