Catalysis and Chemical Processing

Catalyst design is one of the most consequential materials challenges in the chemical industry — catalysts underpin over 80% of industrial chemical processes, yet discovering new catalytic materials remains dominated by trial-and-error synthesis campaigns and incremental modification of known formulations. The space of possible catalyst compositions, support interactions, promoter combinations, and nanostructure geometries is combinatorially vast. A single heterogeneous catalyst system involves choices across active metal, oxidation state, support material, particle morphology, promoter elements, and preparation method, creating a design space that empirical exploration can only sample sparsely. High-performing catalysts for emerging applications — CO₂ reduction, green hydrogen, biomass conversion — remain undiscovered because the search space dwarfs available experimental throughput.

Current catalyst generation methods rely on computational screening using density functional theory to evaluate adsorption energies on pre-defined surface facets, descriptor-based models that correlate activity with a small number of electronic properties, or combinatorial library synthesis covering narrow composition ranges. DFT screening evaluates only the candidates researchers think to propose and is limited to idealized surface models that may not represent real operating conditions. Descriptor models like d-band center correlations work within known catalyst families but fail when applied to novel multi-metallic or single-atom systems where the underlying scaling relations break down. None of these approaches truly generate — they filter from a human-curated search space.