Machine-Learned Force Fields for Lithium Manganese Oxide Cathode Materials

Lithium manganese oxides (LMO) are promising cathode materials for lithium batteries. The transport properties crucial for battery performance are heavily dependent on the electronic structure and polarons, coupling the electronic structure with lattice distortions. While advanced X-ray scattering methods can, in principle, be used to investigate such properties, detailed knowledge of LMO is necessary to guide and analyze such experiments. Computational approaches can contribute significantly here. Traditionally, this requires sophisticated quantum mechanics calculations. However, accessible system sizes and time scales are limited using such methods. Thus, they often offer only an insufficient picture of the charging and discharging process. The development and training of a machine-learned force field (MLFF) for LMO has the potential to complete this picture. However, it requires a novel class of MLFFs that treat explicit charges in a way that accounts for long-range electrostatic interactions. Based on the charge equilibration (QEq) approach implemented in our ELECTRODE software package, we envisage incorporating charges into the MLFF. It is necessary to describe the short- and long-range order of extended polaronic structures in  LMO. This study will support X-ray experiments to enable the tracking of lithium transport in  LMO. Therefore, combining atomistic simulations with experiments would provide novel insights into this material and technological advancements