Rationally Designing Efficient Biomass Carbon Electrocatalysts for H₂O₂ Synthesis and Near-Neutral Zn–Air Batteries with Preliminary Machine Learning Guidance

Biomass-based carbon materials are considered promising metal-free catalysts for the 2e^- oxygen reduction reaction (ORR) to synthesize H₂O₂ and act as air electrodes in Zn-air batteries. However, optimization of the catalyst structure is a complex process due to the diversity of biomass precursors and synthesis parameters. Machine learning, a new artificial intelligence technology, has recently been used in various fields owing to its ability to rapidly analyze large amounts of data and guide material synthesis. Consequently, we constructed a machine learning model based on previously reported experimental data and guided the fabrication of a boron-doped biomass carbon catalyst for the 2e^- ORR. The achieved catalytic performance exceeded most reported ORR catalysts in terms of H₂O₂ selectivity (90-95% in broad potentials of 0.30-0.68 V vs reversible hydrogen electrode), stability (maintaining over 90% selectivity for 12 h), yield (3450 mmol g_catalyst^-1 h^-1), and Faraday efficiency (over 90%). We applied the catalysts to Zn-air batteries and showed a high capacity (2856 mAh g^-1) and stability twice that of traditional commercial metal catalysts. Therefore, this study proposed an effective machine learning model to guide the fabrication of biomass-based catalysts in the field of electrocatalysis.