Data for "Machine Learning Modeling Reveals Unconventional Nucleation Mechanism in Phase-Change Material GeTe"

This dataset contains the trained DeePMD neural-network potential and representative 500 K molecular-dynamics trajectory files used for the main nucleation analysis of GeTe. The uploaded files include the trained model GeTe_model.pb, the full 4096-atom 500 K trajectory for 1 ns, and the corresponding Ge and Te sublattice-like structure trajectories extracted using PTM. The Ge and Te sublattice-like structure files contain the atoms identified as FCC-like MRO motifs together with part of their neighboring coordination atoms for visualization and structural interpretation.