A high-throughput molecular dynamics study with a machine-learning method on predicting diffusion coefficient of hydrogen in α-iron grain boundaries

C Sun and Y Yuan and M Xu and X Li and MF Li and CE Zhou, MATERIALS TODAY COMMUNICATIONS, 44, 111997 (2025).

DOI: 10.1016/j.mtcomm.2025.111997

Engineering on the grain boundaries (GBs) of metal is aiming for reducing susceptibility to hydrogen embrittlement (HE) by trapping hydrogen within the GBs, which slows down hydrogen diffusion and thus facilitates hydrogen gas transport through existing natural gas pipelines. An optimized GB engineering requires a comprehensive understanding of the diffusion behaviors of hydrogen in the intergranular regions and in the bulk. This study develops a high- throughput computational framework involving a Monte-Carlo sampling from the SO (3) group, a parameterized molecular dynamics modeling and simulation code, and a machine-learning algorithm for data analysis. A total of 512 GB structures of alpha-iron are generated and investigated. A database of diffusion coefficients of hydrogen and a predictive model are obtained. A multi-scale scheme for estimating hydrogen diffusion coefficient is proposed. Configurations of the modeled GBs correspond to the lowest, the most common and the highest diffusion coefficient of hydrogen are analyzed. The diffusion of hydrogen within GBs is markedly reduced compared to that in bulk BCC iron, with a 99 % decrease, which highlights the potential of GB engineering as an effective strategy to mitigate HE.

Return to Publications page