Calculation update! New properties have been added to the website for dislocation monopole core structures, dynamic relaxes of both crystal and liquid phases, and melting temperatures! Currently, the results for these properties predominately focus on EAM-style potentials, but the results will be updated for other potentials as the associated calculations finish. Feel free to give us feedback on the new properties so we can improve their representations as needed.
Warning! Note that elemental potentials taken from alloy descriptions may not work well for the pure species. This is particularly true if the elements were fit for compounds instead of being optimized separately. As with all interatomic potentials, please check to make sure that the performance is adequate for your problem.
Citation: K. Ito (2026), "Fast and accurate Fe-H machine-learning interatomic potential for elucidating hydrogen embrittlement mechanisms", International Journal of Hydrogen Energy242, 155600. DOI: 10.1016/j.ijhydene.2026.155600.
Abstract: Highly accurate machine-learning interatomic potentials (MLIPs) for the Fe-H system are essential for elucidating hydrogen embrittlement (HE), yet the high computational cost of existing MLIPs limits their applicability in practical, large-scale simulations. In this study, we develop a new MLIP within the Performant Implementation of the Atomic Cluster Expansion framework, trained on a comprehensive HE-related dataset generated via a concurrent-learning strategy. The potential achieves density functional theory-level accuracy for lattice defects in α-Fe, including vacancies, surfaces, grain boundaries, and both screw and edge dislocations, as well as their interactions with hydrogen. Furthermore, extrapolation-grade analysis demonstrates that it reliably captures atomic configurations associated with hydrogen-induced grain boundary fracture formed during uniaxial tensile deformation of hydrogen-segregated nanopolycrystals. Despite its high accuracy, the computational cost is only several tens of times that of empirical potentials and over an order of magnitude lower than existing Fe-H MLIPs, enabling efficient HE simulations.