Data-Driven Discovery of Mg–Sr–X Ternary Anodes for High-Performance Mg–Air Batteries

Abstract

Developing new magnesium (Mg) alloy anodes is a crucial pathway to improve the performance of Mg–air batteries, however, identifying suitable elements and alloy compositions is impractical by “trial-and-error” method. In this study, a data-driven framework integrating eXtreme Gradient Boosting (XGB), SHapley Additive exPlanations (SHAP), and operations-research-guided sensitivity analysis is developed to accelerate the discovery of high-performance Mg–air battery anodes. This approach overcomes limitations of binary alloy design by optimizing multicomponent Mg–Sr–X systems. Leveraging a curated data set of 819 experimental records, the XGB model achieves exceptional predictive accuracy for specific capacity and specific energy. Sparse-aware SHAP analysis identifies dopants (Ge, Y, In) as critical enhancers, defining optimal compositional windows: 0–1 wt % Ge, 0–4 wt % Y, and 0–1.8 wt % In. Sensitivity screening (Lmax metric) reveals Mg–Sr–Ge as the most promising unexplored system. Experimentally validated Mg–0.1Sr–0.1Ge delivers breakthrough performance: 1391.0 mAh g–1 specific capacity and 1949.5 Wh kg–1 specific energy at 10 mA cm–2, closely aligning with predictions. This work introduces a data-driven and predictive paradigm for designing new Mg alloys.