How to Discover New Electrolyte Additives for Lithium Batteries: A Computational Roadmap

Electrolyte additives, present at a few weight percent or less, decide whether a lithium battery forms a stable solidelectrolyte interphase, survives high-voltage operation, tolerates trace water and HF, or resists fire. For three decades most additives were found by trial and error, and the few design rules that emerged, most famously “a lower LUMO than the solvent”, were known to fail almost as soon as they were formulated. Here we organize the computational literature on additive discovery, from the first quantum-chemical studies of vinylene carbonate (2002) to self-driving laboratories, chemical foundation models and LLM agents (2026), into a step-by-step roadmap: define the function the additive must perform; choose a descriptor that is actually predictive for that function; define and enumerate the chemical space; screen it with tiered quantum-chemical funnels; establish the decomposition mechanism; accelerate with machine learning; generate candidates beyond existing libraries; simulate the interphase the additive builds; and close the loop with automated experiments. For each step we summarize what has been established, what has only been correlated, and what has failed. We show that the hierarchy of redox descriptors (frontier orbitals → vertical and adiabatic electron affinities and ionization potentials → thermodynamic cycles → Li+-and anion-containing clusters → explicit-interface dynamics) is now quantitatively documented, that solvation-and product-level descriptors are emerging as the missing axes, and that the decisive bottleneck has shifted from candidate generation to reliable, fast, function-specific experimental labels. We close with a table of what each class of method can and cannot predict and with a realistic outline of a closed-loop additive-discovery framework.