Publication date: 10th April 2024
The development of safe, sustainable, and mass-deployable solid-state batteries requires fast-ion conducting solid electrolytes derived from earth-abundant elements. In the search for new solid electrolytes, a common approach is to alter the composition of already known phases by doping on anion or cation sites to modify the structure and conductivity. However, searching for new high-conductivity phases from structural modification only is less common. Motivated by this idea, we have implemented a computational materials discovery and screening procedure that combines first-principles structure prediction with machine-learning-derived force-field molecular dynamics. By applying this procedure to the Li–Si–S phase space, we find 21 phases within 30 meV/atom of the convex hull. Four of these phases are predicted to show high Li-ion conductivities, of which three structures are previously unknown and are targets for future synthesis. This combined computational structure prediction and screening methodology holds promise for use in the discovery of novel fast-ion conductors beyond the Li–Si–S chemical space.