Structural Changes and Oxygen Dimerization in Li-Rich Layered Oxide Cathodes: An Atomic-Scale Study

Li-rich transition metal oxides as cathode materials are drawing attention for their potential to dramatically increase energy storage for heavy-duty vehicles and aerospace applications. The increased capacity and energy density are achieved through additional charge compensation from oxygen at high voltages. However, the extra capacity gained from coupled cationic and anionic redox is accompanied by significant voltage fading and hysteresis with cycling. Here, we employ atomistic machine-learning methods to unravel the underlying structural changes at play in two representative Li-rich systems. We develop a machine-learning interatomic potential (MLIP) using an equivariant neural network trained on a data set generated using Density Functional Theory (DFT) for Li2Ni0.75Mn0.25O2 and Li1.2Ni0.6Mn0.2O2, both with a 3:1 Ni:Mn ratio. The training data set includes pristine as well as oxygen-bonded configurations, such as peroxide, superoxide, and O2 species, which were generated via an Ewald preconditioning scheme. Using our MLIP, we extract structural transformations driven by lithium migration from octahedral to tetrahedral sites and reveal the critical role of Mn in promoting Li occupation in locations different from the usual octahedral environment. Oxygen dimers are found to be energetically favorable only in LixNi0.6Mn0.2O2, where their formation is enabled by the presence of Li–O–Li configurations and the cation vacancies resulting from the migration of Li. Our study gives insight into the atomic-scale irreversible processes responsible for performance degradation in Li-rich layered oxide cathodes, providing a foundation for developing mitigation strategies.