Predicting Short-Range Order in Disordered Rocksalt Li-Oxide Cathode Materials with Density Functional Theory and Crystal Graph Neural Networks

Li-excess disordered rocksalts have recently emerged as a promising class of materials to replace conventional layered nickel manganese cobalt oxide cathodes as a means to enhance the reversible capacity of Li-ion batteries. However, Li ions on the cation sublattice have been shown to form localized clusters of chemical short-range order, which can affect capacity and Li mobility, high values of which are crucial to the macroscale performance of the cathode. To address this challenge, we employ both ab initio calculations and crystal graph machine learning methods to reveal structural factors contributing to short-range order formation and explore a large compositional space to identify materials that exhibit a thermodynamically stable disordered phase. Ultimately, this high-throughput screening process results in a subset of compositions that are strong candidates for future experimental characterization.