Artificial intelligence-boosted quantitative deformation mapping for spatially clarifying electrochemically irreversible anion intercalation in graphite interlayers

Rechargeable aluminum-based batteries, particularly aluminum-graphite batteries (AGBs), represent promising candidates for stable energy storage owing to their low cost and intrinsic safety features. However, the intercalation/deintercalation of bulky anions induces severe interlayer deformation in graphite positive electrodes which is linked to the unexpectedly low initial Coulombic efficiency (CE) in the early cycles. Herein, an artificial intelligence (AI)-boosted lattice deformation identification method, i.e., AI-DeformSnap, is developed to enable high-efficiency and high-accuracy recognition of deformation states across more than 400,000 lattices extracted from 200 transmission electron microscopy (TEM) images. The density-based spatial clustering of applications with noise algorithm is incorporated to quantitatively analyze the spatial distribution and connectivity of deformed regions. Nine-interval classification criteria are proposed to clarify the competitive coexistence and heterogeneous distribution of electromechanically compressed and expanded regions, suggesting that irreversible deformation regions are strongly correlated with structural disorder in graphite interlayers and may contribute to the low initial CE. This work establishes an AI-boosted quantitative characterization and analysis framework for understanding the mechanism of microstructural degradation and evolution behaviors of energy storage based on anion intercalation, which provides insights for optimizing the graphite interlayer structures for high-performance AGBs.