High‐Throughput Design of Active MXene Catalysts for Li─O 2 Battery Using Machine Learning

ABSTRACT The performance of lithium–oxygen (Li─O 2 ) batteries is limited by sluggish reaction kinetics, leading to issues such as poor reversibility and severe parasitic reactions. This necessitates advanced catalysts like MXenes, but their vast compositional diversity and complex structure‐activity relationships hinder traditional discovery approaches. Herein, we employ an integrated high‐throughput workflow (HTW) and machine learning (ML) framework for Li batteries for the first time to systematically investigate 2D transition metal carbides/nitrides MXenes‐based catalysts. We defined a virtual compositional space of ∼2 million MXene candidates. Guided by a combinatorial enumeration and subsequent rule‐based screening, we down‐selected this space to an HTW design set of 4896 unique MXene configurations for computation. Our developed Light Gradient Boosting Machine model achieved superior accuracy (MAE = 0.32 eV) in predicting reaction free energy change across four key steps, enabling the identification of exceptional catalysts including Mo 3 C 2 Cl 2 which exhibits an ultra‐low overpotential of 0.01 V. Our interpretability analysis reveals the intricate mechanisms by which different electronegativity terminals modulate the electronic structure and reaction mechanisms of MXenes. This work establishes an efficient computational reference for accelerating the discovery of advanced energy materials and provides fundamental insights into structure‐activity relationships in electrocatalysis.