Sustainable batteries necessitate high-performance hard carbon negative electrodes derived from abundant biomass. However, realizing their full potential is significantly limited by the inherent diversity of biomass feedstocks, the intricate control over carbonization and resulting microstructures, and the complex interplay between processing, structure, and electrochemical performance. Here, we introduce “intelligent carbonization”, a strategy integrating programmable Joule heating (1000-2000 °C, 10-60 s) with machine learning to substantially accelerate the discovery and optimization of biomass-derived hard carbons. By mapping over 1000 synthetic pathways and decoding the multidimensional feature space, we reveal a performance-correlated factor that serves as a crucial predictor of capacity, complementing conventional graphitic descriptors (in-plane crystallite size/ interlayer spacing). By a minimal energy input (0.1 kWh g−1), our strategy converts biochar into advanced hard carbon delivering 369 mAh g−1 reversible capacity, high rate capability, and improved cycling stability (>5000 cycles at a specific current of 3 A g−1). This data-centric approach allows low-cost and intelligent manufacturing of diverse biomass resources into performance-unified hard carbon negative electrodes, thereby paving the way for practical and large-scale biomass valorization towards sustainable energy storage solutions. Biomass-derived hard carbon negative electrodes can cut cost, but properties vary by feedstock. Here, authors use data-driven intelligent carbonization to unify performance across diverse biomass, yielding high-capacity, long-life negative electrodes with lower footprints.
Cui et al. (Fri,) studied this question.