Machine Learning Guided Processing and Electrochemical Analysis of Recycled Graphite Anodes for Li-Ion Batteries

Le résumé fourni par la source

Bolstering graphite supply chains for lithium-ion batteries requires multiple complementary pathways, including biomass-derived graphite, reclamation of graphite from black mass, and direct recovery of spent anodes. Catalytic graphitization of biomass, such as Fe-catalyzed conversion of pyrolysis bio-oil to graphitic carbon, has emerged as a promising route to induce structural ordering in low-cost carbon precursors. Related catalytic approaches have also been explored for upgrading recycled black mass into graphitic materials. 1 However, even when graphite is recovered directly from spent anodes, post-recovery purification and morphological optimization remain essential to achieve competitive electrochemical performance. In this study, graphite was recovered directly from anode electrodes collected from lithium-ion batteries with three distinct states-of-health: calendar-aged (CA), lightly cycled (LC), and heavily cycled (HC). Unlike black-mass-derived graphite, the electrode-derived feedstock contained minimal aluminum contamination, eliminating the need for NaOH roasting or aggressive HF treatment typically required to remove Al 2 O 3 impurities. Building on prior work, a machine-learning (ML) model was used to downselect optimized milling parameters that minimize an “F-score,” a quantitative metric benchmarking roundness, median particle size, density, and surface area relative to commercial spherical graphite. Milling conditions previously optimized for commercial spherical graphite were applied to all recycled graphite sources using both ZrO 2 and polypropylene (PP) milling media. All samples, after controlled milling, were characterized using XRD, Raman spectroscopy, SEM, BET surface area analysis, and particle size analysis. Electrochemical performance was evaluated in Gr/lithium half-cells and Gr/NMC622 full cells. The optimized recycled graphite exhibits improved electrochemical performance relative to baseline materials, demonstrating the effectiveness of machine-learning-guided processing in enabling high-performance reuse of recycled graphite anode materials.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.