Calculation update! The crystal structure tables have been updated as they now use the current Materials Project (mp-) reference structures, and calculations that previously threw errors were re-ran after a minor bug fix.
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: F-S Meng, S Shinzato, K Matsubara, J-P Du, P Yu, and S Ogata (2025), "A Neural Network Interatomic Potential for the Ternary α-Fe-C-H System: Toward Million-Atom Simulations of Hydrogen Embrittlement in Steel".
Abstract: A neural network interatomic potential (NNIP) has been developed for the ternary system of α-iron, carbon, and hydrogen to clarify the degradation behavior of Fe-C steels in hydrogen-rich environments. The NNIP was trained on an extensive reference database generated from spin-polarized density functional theory (DFT) calculations. It demonstrates remarkable performance in various scenarios relevant to Fe and Fe-C systems under hydrogen, including the diffusion kinetics of H and C in Fe and their thermodynamic interactions with iron vacancies, grain boundaries, screw dislocations, cementite, and cementite-ferrite interfaces. Using this NNIP, we conducted large-scale (one-million-atom) molecular dynamics (MD) simulations of uniaxial tensile tests on C-containing α-Fe both with and without H, showing that hydrogen enhances defect accumulation during plastic deformation, which may eventually lead to material failure.
Notes: Fan-Shun Meng notes that "This neural network interatomic potential was trained using the n2p2 package and validated in the associated publication. This potential was designed for the ternary system of 𝛼-Fe-C-H. Additionally, the potential can also be used for pure 𝛼-Fe, 𝛼-Fe-C, 𝛼-Fe-H."
See Computed Properties Notes: This file was provided by Fan-Shun Meng on July 13, 2025. To use the potential in LAMMPS, the pair_style of hdnnp should be adopted, and the package of ML-HDNNP should be compiled (see ML-HDNNP documentation) File(s):