Resumen
Complex metal oxides (CMOs) are used broadly in applications including electroreactive forms found in lithium-ion battery technology. Computational chemistry can provide unique information about how the properties of CMO cathode materials change in response to changes in stoichiometry, for example, changes of the lithium (Li) content during the charge?discharge cycle of the battery. However, this is difficult to measure experimentally due to the small cross-sectional area of the cations. Outside of operational conditions, the Li content can influence the transformations of the CMO when exposed to the environment. For example, metal release from CMOs in aqueous settings has been identified as a cross-cutting mechanism important to CMO degradation. Computational studies investigating metal release from CMOs show that the thermodynamics depend on the oxidation states of lattice cations, which is expected to vary with the lithium content. In this work, computational studies track changes in metal release trends as a function of Li content in Lix(Ni1/3Mn1/3Co1/3)O2 (NMC). The resulting dataset is used to construct a random forest tree (RFT) machine learning (ML) model. A modeling challenge in delithiation studies is the large configurational space to sample. Through investigating multiple configurations at each lithium fraction, we find structural features associated with favorable energies to chemically guide the identification of relevant structures and adequately predict voltage values.