Machine-learning-driven global exploration reveals atomic-scale degradation of LiNiO2 during delithiation

Abstract

Increasing the nickel content enhances the energy density of lithium-ion battery (LIB) cathodes, yet Ni-rich layered materials suffer from structural degradation and safety concerns. To address these, we choose LiNiO2 as an ideal model. Employing the stochastic surface walking (SSW) method and a validated global neural network potential, we explore the bulk structure of LixNiO2 under different lithium contents. A synergistic combination of the SSW, molecular dynamics simulations, and the double-ended surface walking method is employed to investigate surface degradation on the (001), (012), and (104) facets. Additionally, the effects of Mg2+, Al3+, and Ti4+ doping are studied. The research finds that lithium stoichiometry governs three structural regimes in the bulk of LixNiO2, with high lithium content maintaining layer stability, intermediate content leading to Li/Ni mixing and O dimer formation, and low content causing structural collapse and O2 release. On the surface, the fully delithiated (001) surface stabilizes into a spinel-like configuration thermodynamically but retains the layered structure kinetically in the short term; the (012) surface degrades into a rock-salt phase through two competitive pathways; and the (104) surface undergoes an irreversible transformation into a rock-salt phase via a two-stage process with multiple concurrent pathways. Different dopants show facet-dependent stabilization effects, with Al3+ generally enhancing oxygen stability, Mg2+ being less effective in stabilizing the surface structure. This study provides a comprehensive atomic-scale understanding, deepens the comprehension of degradation mechanisms, and proposes an innovative doping strategy, offering a new direction for developing high-performance cathodes in next generation LIBs.