A general-purpose machine-learning interatomic potential for FeCr steel: Atomistic insights into high-temperature mechanical behavior

FeCr alloys are promising for cladding due to their thermal stability and radiation resistance, but their atomic-scale mechanical behaviors under varying temperatures is not yet well understood. Traditional empirical potentials are unreliable at high temperatures due to oversimplified assumptions. The deep potential (DP) model offers a more accurate and efficient alternative for predicting high-temperature alloy behavior. Here, we develop a deep potential model for FeCr alloys using a dataset obtained from density-functional theory (DFT) and the DP-GEN active learning framework. Molecular dynamics(MD) simulations based on the DP model show that a typical Fe3Cr alloy has a tensile strength of 15 GPa at 1200 K with a 25 % reduction in stress. This difference is attributed to the pinning effect of Cr atoms on dislocation slip and the strengthening induced by short-range ordering in Fe3Cr bonds. Compared to the MEAM potential, the DP model predicts a fracture strain of 32 % for FeCr alloys, which is in agreement with ductile characteristics observed in experiments. These results elucidate the microscopic mechanical behavior and failure mechanisms of FeCr alloys, paving the way for the development of high-performance FeCr alloys for high-temperature applications.