Degradation mechanism analysis of LiNi0.5Co0.2Mn0.3O2 single crystal cathode materials through machine learning

Abstract

LiNi_0.5Co_0.2Mn_0.3O₂ (NCM523) has become one of the most popular cathode materials for current lithium-ion batteries due to its high-energy density and cost performance. However, the rapid capacity fading of NCM severely hinders its development and applications. Here, the single crystal NCM523 materials under different degradation states are characterized using scanning transmission electron microscopy (STEM). Then we developed a neural network model with a two-sequential attention block to recognize the crystal structure and locate defects in STEM images. The number of point defects in NCM523 is observed to experience a trend of increasing first and then decreasing in the degradation process. The space between the transition metal columns shrinks obviously, inducing dramatic capacity decay. This analysis sheds light on the defect evolution and chemical transformation correlated with layered material degradation. It also provides interesting hints for researchers to regenerate the electrochemical capacity and design better battery materials with longer life.