Sustainable Regeneration of Graphite from Spent Lithium-Ion Batteries through Machine Learning and Life Cycle Assessment

Abstract

Graphite, the major anode material in spent lithium-ion batteries, remains underutilized in current recycling systems. Regeneration is often assumed to require a trade-off between electrochemical performance and environmental impact since high reversibility is typically attained through energy-intensive processing. In this study, graphite regeneration is formulated as an integrated design problem linking process condition, graphite structure, electrochemical performance, and sustainability. A literature-curated dataset was used to train an interpretable multi-output model for predicting structural descriptors and initial coulombic efficiency. Shapley additive explanation analysis demonstrated that chemical pretreatment influences the effectiveness of subsequent thermal reconstruction, while electrochemical performance depends on coupled chemical–thermal effects. Within the explored design space, the optimized regeneration condition was predicted to yield graphite with favorable structural and electrochemical characteristics, achieving an initial coulombic efficiency of 98.18% and an interlayer spacing of 0.337 nm at 577 °C for 1 h under an Ar atmosphere. Compared with a reference case based on literature-average conditions, the global warming potential decreased from 12.26 to 5.60 kg CO2-equivalent, representing a 54.4% reduction. The results demonstrate that improved graphite quality and lower carbon footprint can be achieved simultaneously within the studied domain and provide a general, data-driven strategy for low-carbon battery recycling.