Statistically derived proxy potentials accelerate geometry optimisation of crystal structures

Geometry optimisation is a ubiquitous task in computational materials science, but the number of systems for which fast and reliable force field models have been developed is limited. This leaves local energy minimisation using computationally expensive density functional theory (DFT) as the only reliable route to structure optimisation. At the same time, the available crystal structures of known materials contain the information about the interatomic interactions that produced these stable compounds, expressed as the interatomic distances between all the constituent atoms. We use this relationship to statistically learn the effective interatomic interactions in crystalline inorganic solids from their structures. By analysing pairwise interatomic distances in the reported crystallographic data for inorganic materials, we have constructed statistically derived proxy potentials (SPPs) that can be used for structure optimisation. We apply such optimisation step to markedly improve the quality of the input crystal structures for DFT calculations and demonstrate that the SPPs accelerate geometry optimisation for three systems relevant to battery materials. As this approach is chemistry-agnostic and can be used at scale, we produced a database of all possible pair potentials in a tabulated form ready to use.