uMLIP for 2D materials

Data for manuscript, entitled: "Remarkable accuracy of universal machine learning interatomic potentials for exploring complex mechanical and phononic properties of two-dimensional materials"
This dataset includes the complete training datasets generated for defective graphene, hexagonal boron nitride (hBN), and nanoporous carbon-nitride covalent organic frameworks (CN-COFs), together with the corresponding passively-fitted MTP potentials.
The datasets are provided in both CFG and XYZ formats. In the CFG and MTP files, the atomic type indices 1, 5, 6, and 7 correspond to H, B, C, and N atoms, respectively.