Predicting Oxygen Release and Surface Reconstructions in Li-Ion Battery Cathodes Via Automated Construction of Computational Surface Phase Diagrams and Machine Learning

Although LiNiO 2 (LNO) is chemically very similar to LiCoO 2 (LCO), LNO and related Co-free Ni-rich cathodes suffer from continuing surface degradation via oxygen gas release during electrochemical cycling that leads to the formation of surface phases with high impedance. While the surface degradation of LNO and related cathode compositions have been characterized experimentally on a phenomenological level, an understanding of the surface reconstructions that form on the atomic scale and the intrinsic surface instability of LNO compared with other related cathode compositions is still lacking. To shed light on surface reactivity, we developed a thermodynamic methodology for the prediction of voltage and temperature-dependent surface electrode reconstructions [1] and a computational framework to automate the time-consuming enumeration of surface reconstructions, the construction of symmetric surface-slab models, and the analysis of surface phase diagrams [2]. By applying first-principles atomistic modeling, we determined the self-reduction mechanism of LNO and compared the stable surface reconstructions with those of LCO. To further assess the surface stability of more complicated NMC/NCA cathodes with the help of our own generated LNO and LCO databases, we developed a supervised machine learning (ML) model to train on geometrical and electronic fingerprints from surface and bulk models, respectively. Our results provide insight into the initial stages of surface degradation in Ni-rich cathodes and lay the foundation for the computational design of stable cathode materials against oxygen release. Li, X.; Wang, Q.; Guo, H.; Artrith, N.; Urban, A. ACS Appl. Energy Mater. 2022, 5 (5), 5730–5741. Li, X.; Qu, J.; Yeu, I.; Li, Z.; Rodríguez-López, J.; Urban, A., in preparation , 2023