Accurate Thermophysical Properties of Water using Machine-Learned Potentials

Files generated for the publication "Accurate Thermophysical Properties of Water using Machine-Learned Potentials" (https://doi.org/10.1063/5.0325444), as well as scripts used to generate them.The following data sets are provided: water_trainingset.xyz and ML_AB.water_training contain the water structures each containing 63 or 63 water molecules as used for training MLPs, in the format used for MACE and KbP training, respectively. water_testset.xyz contains hold-out water structures used as test set. water_testset_128m.xyz contains hold-out water structures each containing 128 water molecules. sample_ _ .xyz are the samples drawn from MD simulations of the potential , at temperature . For further detail on the databases please refer to the publication. Input scripts, as well as example output files are provided in the zip files: DFT.zip contains input files and example outputs for the 2 stage DFT calculations using VASP. Examples are provided for both 64 molecule as well as 128 molecule calculations. KbP_training.zip contains input files and example logs for the "select" and "refit" runs used in VASP to generate KbPs. MACE_training.zip contains input files and example logs for training MACE models with L=2 and L=0, as described in the publication.