Universal Neural Network Potential-Driven Molecular Dynamics Study of CO₂/O₂ Evolution at the Ethylene...

Long-term durability and safety are required to develop Li-ion batteries that can operate at high voltages. However, side reactions, including the release of O₂ from the electrode and CO₂ from the organic electrolyte, occur at the positive-electrode/electrolyte interface during charging at high voltages. In this study, universal neural network potential (UNNP)-driven molecular dynamics (MD) calculations are used to investigate the mechanism of the reaction between Li_xCoO₂ (0 ≤ x ≤ 1) or Li_xNiO₂ (0 ≤ x ≤ 1), as the positive-electrode material, and an ethylene-carbonate-based electrolyte, with a solid-liquid interface composed of ∼1700 atoms. Molecular CO₂ and O₂ evolve from the partially or fully Li-deintercalated Li_xNiO₂, while no gas-evolution reactions are observed for Li_xCoO₂. Hence, compared Li_xNiO₂, the LiCoO₂ electrode is more stable toward the decomposition of ethylene carbonate in the charged state. The decomposition reactions at the solid-liquid interface during charging are also analyzed using a NN force field. This study provides a robust approach involving MD simulations using UNNP to better understand the side reactions in electrochemical devices, which can guide manufacturers in selecting appropriate materials.