Diffusion Behaviors of Lithium Ions at the Cathode/Electrolyte Interface from Global Neural Network Potentials

The diffusion of Li ions plays a vital role and has been the central topic of the Li-ion battery (LIB) research. However, the diffusion behaviors at the cathode/electrolyte interface still remain unclear due to the complexity of interfaces. Despite some progress achieved through ab initio molecular dynamics (AIMD) and classical molecular dynamics (MD) simulations, a full understanding of Li-ion diffusion behavior requires direct simulations of the entire interface. This remains challenging due to the inherent limitations of current simulation methods. Here, we develop a global neural network potential to reveal the Li ion diffusion behaviors at the interface between LiCoO₂ cathode and liquid electrolytes (EC, DMC and LiPF₆) by performing long-term molecular dynamics simulations. We identify four kinds of interfacial diffusion behaviors by analyzing the trajectories of Li ions. While the inactive Li ions are immobile, the active Li ions can shuttle between the interface and solution regions, hop between different interfacial sites, or diffuse as they would in pure electrolytes. Among all diffusion behaviors, only those diffusion across the interface can contribute to the effective conductivity and thus the device performance. Based on the above findings, we further study the influence of electrolyte concentration and interfacial compounds on the diffusion of interfacial Li ions. We show that 1 mol/L LiPF₆ has the largest conductivity across the interface, in agreement with the experimental results that 1 mol/L LiPF₆ is the most suitable electrolyte concentration. We further propose that Li₂O could be used as interface coating to improve the Li ion conductivity across the interface. Our work provides deep atomic insights into the dynamics of Li ions at cathode/electrolyte interface and is expected to help the optimization of LIBs.