Computational Modeling of Electrolytes and the Solid–Electrolyte Interphase (SEI) in Li-Ion Batteries: A 30-Year Odyssey from DFT/MD/MC to AI–A Comprehensive Review

The solid electrolyte interphase (SEI)—the passivating nanometric film that forms on anode surfaces in lithium-ion batteries—remains one of the most critical yet least understood components in electrochemical energy storage. Unraveling its formation, structure, and evolution has proven to be a formidable odyssey, spanning decades of relentless effort and confronting researchers with challenges as daunting as those faced by the ancient voyagers. This review provides a comprehensive, 30-year perspective on this arduous journey, tracing the evolution of computational methods for modeling electrolytes and SEI from early density functional theory (DFT) calculations of isolated molecules to modern artificial intelligence (AI) and machine learning (ML) frameworks that promise to accelerate materials discovery. We organize the methodological development into five generations, each representing a milestone in this ongoing voyage: static DFT (1995–2005) establishing thermodynamic foundations; ab initio molecular dynamics (AIMD) for interfacial dynamics (2005–2010); AIMD integrated with free energy calculations to elucidate electron transfer and rare events (2011–2015); quantitative design rules, space charge effects, and mechanical properties (2016–2020); and advanced DFT techniques combined with MLIPs, generative models, large language models, and self-driving laboratories (2021–2026). We critically assess classical MD, reactive force fields (ReaxFF, eReaxFF), hybrid HAIR methods, kinetic Monte Carlo, and Monte Carlo approaches, demonstrating how each method bridges the chasm between electronic structure and mesoscale phenomena. The review further examines ML applications—from property prediction and spectroscopic signature forecasting to MLIPs achieving first-principles accuracy at classical force-field speeds, generative molecular design, high-throughput screening, active learning, and integration with autonomous experimentation. We identify persistent challenges that continue to test the field’s resolve: long-range electrostatics, transferability across chemical space, time-scale bridging, uncertainty quantification, and data standardization. Practical recommendations for method selection are provided, accompanied by a comprehensive comparative table. This review establishes that the integration of DFT, AIMD, MD, ReaxFF, kMC, MLIPs, and AI has created a powerful multiscale modeling ecosystem, transforming battery research from trial-and-error to rational, data-driven design.