Data used in or calculated from the Li-Al paper.
crystal_structures/ includes POSCAR files of all stable sturctures predicted by this paper;
machine_learning/nep_train are files (dataset and input) used to train the machine learning potential of Li6Al ranging from 50 to 150 GPa, covering solid, superionic and liquid states. Using cat train.xyz* > train.xyz to restore training dataset;
machine_learning/mlmd/lammps and machine_learning/mlmd/gpumd are inputs of machine learning molecular dynamics including the configuration, the potential file nep.txt and the MD software inputs. We tested nep.txt using different MD software including LAMMPS and GPUMD. GPUMD is the package we developed for NEP machine learning training and coresponding MD simulations. See details in https://doi.org/10.1063/5.0106617 or https://github.com/brucefan1983/GPUMD.