Machine Learning-Accelerated Development of Li-Rich Layered Oxides with High Reversible Capacity and Coulombic Efficiency

Li-rich layered oxides are promising high-energy-density cathodes for next-generation Li-ion batteries, yet their practical application is hindered by structural instability arising from oxygen redox activity, which typically results in a trade-off between achieving a high discharge specific capacity and maintaining a high Coulombic efficiency. Herein, we employ machine learning to identify key synthesis factors governing the initial Coulombic efficiency in Li_1.2Ni_0.13Co_0.13Mn_0.54O₂. Four machine learning models were trained with a data set of 203 samples. Among them, the gradient boosting decision tree exhibited superior predictive accuracy (<i>R</i>² = 0.802 on the test set) and identified the lithium-to-transition metal ratio and presintering atmosphere as critical parameters. Machine learning-guided synthesis reveals that presintering in air at low temperatures reduces the Li₂MnO₃ phase proportion, promotes the exposure of the {010} planes favorable for Li⁺ transport, and mitigates the formation of surface rock-salt. This approach yielded a high discharge capacity of 301.02 mAh g^-1 and an initial Coulombic efficiency of 81.05%, highlighting the effective integration of data-driven design with experimental synthesis for advanced Li-rich layered oxide optimization.