Battery Cathodes and Energy Storage

The battery industry faces a fundamental generation bottleneck: the chemical space of possible cathode materials is astronomically large, yet experimental synthesis is slow and expensive. Traditional cathode development relies on incremental modifications to known chemistries — NMC, LFP, NCA — because there is no systematic way to navigate the vast combinatorial landscape of multi-element oxide, sulfide, and phosphate compositions. Promising regions of composition space remain unexplored not because they lack potential, but because human-guided exploration cannot cover the search space at the pace the energy transition demands. The result is an industry locked into marginal improvements on decades-old chemical families while transformative compositions go undiscovered.

Existing computational approaches to cathode design rely on density functional theory screening of pre-enumerated candidate lists, high-throughput combinatorial synthesis of limited composition ranges, or machine learning surrogate models trained on existing databases. DFT screening is computationally expensive and limited to evaluating candidates that someone has already proposed. Combinatorial synthesis covers narrow composition windows dictated by available precursors. ML surrogates interpolate within known chemical families but struggle to extrapolate to genuinely novel chemistries — they predict properties of familiar materials well but cannot reliably generate candidates in unexplored regions of composition-structure space where the most impactful breakthroughs reside.