Revealing Fast Ionic Conduction in Solid Electrolytes through Machine Learning Accelerated Raman Calculations

This dataset contains the following files for each material: XDATCAR: Dynamics (MD) trajectory in XDATCAR format produced by VASP. md_vdos.dat: Vibrational Density of states (VDOS) obtained from the MD trajectory. The first column is freuqency in inverse cm, the second column is total VDOS, the remaining columns are VDOS per atomic species. The atomic species are listed in the same order as in the XDATCAR file. md_msd.dat: Mean square displacement (MSD) obtained from the MD trajectory. The first column is time in fs, the remaining columns are MSD per atomic species. The atomic species are listed in the same order as in the XDATCAR file. alpha_trainingset.dat: Dielectric tensors used for training the ML models. These were obtained with DFPT calculations in VASP, for randomly selected snapshots. The first column contains the index of the snapshots in the trajectory, the remaining columns contain the tensor components in the order xx yy zz xy yz zx. alpha_Nt_*.dat: Timeseries of dielectric tensors predicted by the ML models. The number in the filename specifies the size of the training dataset that was used in each case. The columns contain the tensor components in the order xx yy zz xy yz zx. The sampling rate of the timeseries corresponds to 10 times the MD timestep. spectrum_Nt_*.dat: Raman spectra obtained from the ML-predicted timeseries. The number in the filename specifies the size of the training dataset that was used in each case. The first column is frequency in inverse cm, the second column is Raman intensity. spectrum_subtrajectory_*ps.dat: Raman spectra obtained from ML-predicted timeseries for subtrajectories of different lengths. All of these where generated with a training set size of Nt=300. The number in the filename specifies the length of the subtrajectory. The first column is frequency in inverse cm, the second column is Raman intensity.