<p>Lithium thiophosphate (LPS) has demonstrated promising properties for use as a solid electrolyte for the next-generation of lithium ion batteries. However, the high reactivity of LPS with common contaminants such as atmospheric water hinders commercialization of the cells. We employ a machine learning interatomic potential to gain fundamental, atomistic understanding on the mechanical, chemical and electronic properties of an LPS surface. Our focus lies first on the identification of relevant surface complexions formed by surface reconstructions, which differ greatly in properties from the bulk and define the form of the reactive sites. Then, we reveal the complex reactivity of the LPS surface with water and identify the surface moieties that play a relevant role in the material’s degradation. Thus, we uncover the impact of dynamical changes of the surfaces and their electronic structure on the material’s reactivity.<br>This dataset includes all files necessary to reproduce the data reported in <a href="https://doi.org/10.1103/5hf9-hlj6">Reconstructions and dynamics of 𝛽 -lithium thiophosphate surfaces</a>.<br>This includes the MLIP model and dataset, generated surface structures with certain Miller indices, Wulff constructions, as well the input and result files of LAMMPS and DFT calculations.</p>