Boron Carbide - HDNNP dataset

This dataset contains the essential components used in the development of a high-dimensional neural network potential (HDNNP) for boron carbide, including both the training data and the final model. The following components are included: 1) Monte Carlo-Generated Structures The full set of approximately 3 million boron carbide structures generated using a Monte Carlo algorithm. These structures span a wide range of stoichiometries and atomic configurations, supporting robust sampling of the potential energy surface. 2) DFT Calculations The entire set of density functional theory (DFT) results corresponds to the sampled subset of structures from the MC data. This includes total energies and atomic forces calculated using VASP. 3) Atom-Centered Symmetry Function (ACSF) Parameters The symmetry function data used to describe local atomic environments for training the potential. 4) Fine-Tuning Structures A small set of additional structures used for fine-tuning and improving the accuracy of the HDNNP. 5) Trained Model and Weight Files The complete HDNNP model, including the weights, which can be directly used in molecular dynamics simulations.