Unraveling the bilayer-cooperative transformation mechanism at the α/β-Si3N4 interface via machine-learning simulations
Silicon nitride (Si3N4) is a strong, thermally stable covalent ceramic typically regarded as brittle with limited deformability. Recent experimental and density functional theory (DFT) studies indicate that the α/β interface undergoes a β→α transformation via sliding followed by bond-switching, suggesting a pathway to achieve plasticity, but DFT's spatiotemporal reach prevents a full mechanistic picture. Here, we develop a physics-informed high-accuracy neural network interatomic potential (NNAP) model with DFT-level accuracy for phase transformations and use it to perform large-scale atomistic simulations. NNAP-guided simulations show that structural relaxation during relative sliding between α- and β-phases at the interface triggers pronounced atomic-layer rearrangements and lowers the energy barrier by nearly 60%. We further find that the ensuing phase transformation does not proceed by isolated layer-by-layer switching but instead follows in-plane nucleation and growth mediated by a bilayer cooperative mechanism, which further reduces kinetic barriers and facilitates the transformation. CI-NEB calculations reveal that the bilayer cooperative pathway has an energy barrier of 0.018 eV/Ų, lower than the independent layer-by-layer manner (0.020 eV/Ų), indicating enhanced kinetic accessibility. These results provide new atomistic insights into interface-driven phase transformations in dual-phase Si3N4 and offer guidance for designing more deformable covalent ceramics.