Predicting incoherent interface structures in Cu–Ni–Si-Mn alloys using machine-learning interatomic potentials

Interfaces are everywhere in condensed matters and materials systems. While coherent or semicoherent interfaces can be described using conventional periodic models, however, incoherent interfaces at grain boundaries remains elusive due to their large-scale or aperiodic atomic nature. In this work, we employ a machine-learning interatomic potential (MLP) to simulate large-scale interface atomic structure within a commensurate approximation. We first examine the Cu//Ni2Si coherent interface and confirm that the MLP accurately reproduce the coherent continuous precipitation nature in Cu-Ni-Si alloys. We further extend the MLP approach to the incoherent interfaces observed in Cu-Ni-Si-Mn alloys, revealing low-energy, film-like grain boundary phase morphology. Our results demonstrate that MLPs enable realistic atomic modeling of incoherent interfaces beyond first-pinrciples limit. This work was published in ref.[1].