Data Speaks: LLM-Enabled Evidence Auditing Resets the “Migration Barrier” Playbook for Solid-State Electrolytes

Reported activation barriers for ion transport in solid-state electrolytes often disagree, not because the physics is capricious, but because our computational workflows privilege convenience over coverage. By joining a domain-curated database with a language model that can normalize entities and assemble comparable cohorts, a recent pioneering study by Wang et al. shows that long-standing assumptions about “standard” activation barrier calculations fail a simple test: accuracy across families. Single-path CI-NEB and high-temperature AIMD extrapolated to device conditions deviate systematically for specific classes of materials, whereas ab initio metadynamics (MetaD) aligns more closely with experiment and repeatedly reveals a two-step migration mechanism in neutral-molecule-coordinated hydrides. The point is not to anoint a universal winner; it is to treat method’s reliability as an object of study, measured against data at scale.