Neural network potential for pyrophosphate hydrolysis

Neural Network Potential (NNP) trained with DeePMD-kit v2 for P2O7(4-), MgP2O7(2-), HP2O7(3-) and H2P2O7(4-) hydrolysis. The dataset.zip folder contains the whole training dataset used to train the NNP. It has 51467 structures, and every one contains 518 atoms. The units of energy are in eV, the units of forces in eV/angstrom and the units of xyz coordinates in angstrom. The training set format follows the conventions used by DeePMD-kit (see https://docs.deepmodeling.com/projects/deepmd/en/v3.0.3/data/data-conv.html). The validation_dataset.zip folder contains an independet set of structures unseen by the NNP and selected along the reaction path. It has 800 structures for each different chemical system of pyrophosphate. The README file that contains basic instructions to run a molecular dynamics simulation using the NNP. The NNP_compressed.pb file is the NNP file that can be used to run molecular dynamics. The lammps.in file is the input file of a LAMMPS calculation. The ppmg_r_1.lmp file contains a reactant structure of solvated MgP2O7(2-) in LAMMPS format. The ppmg_r_1.xyz file contains a reactant structure of solvated MgP2O7(2-) in xyz format. The r_ppmg.lammpstrj file contains the resulting trajectory of the MD simulation. The lammps.log file contains the output data of the MD simulation.