Machine‐Learning‐Guided Dopant‐Class Design Enables Near‐Zero‐Strain P2‐Type Cathodes for High‐Voltage Sodium‐Ion Batteries

This is the authors' abstract. We don't add key points for this paper.

Manganese‐based layered oxides are promising cathodes for sodium‐ion batteries, yet their practical deployment is restricted by Mn 3+ ‐induced Jahn‐Teller distortion, irreversible high‐voltage P2‐O2 phase transitions, and oxygen‐related side reactions that collectively cause structural deterioration and rapid capacity fading. Here, based on literature‐correlation analysis of reported Ni─Mn P2‐type cathodes, we establish a machine‐learning‐guided dopant‐class design strategy. The analysis reveals distinct roles of metal, non‐metal, and metalloid dopants in governing cycling stability and redox reversibility, identifying metalloid regulation as the most favorable direction for enhancing high‐voltage structural and redox stability. Guided by this insight, an Al─B─F co‐doped P2‐type cathode is rationally designed, in which boron reinforces the oxygen framework through strong B─O covalency, while Al and F provide complementary cationic and anionic regulation. The optimized cathode delivers 142.8 mAh g −1 at a high average voltage of 3.53 V and exhibits superior low‐temperature capacity retention of 98.19% after 300 cycles at −5 C. In situ XRD, soft x‐ray absorption spectroscopy, ex situ XPS, operando DEMS, and DFT calculations reveal a highly reversible reaction pathway featuring stabilized oxygen redox and near‐zero volumetric strain (ΔV/V ≈ 0.025%). This work establishes machine‐learning‐guided dopant‐class design as a general strategy for developing durable high‐voltage layered sodium‐ion cathodes.