AdsorbML: How AI Is Transforming Catalyst Discovery for Clean Energy and Chemicals

Researchers developed AdsorbML, a hybrid AI + physics method that finds the best way molecules bind to catalyst surfaces—critical for designing fuels and chemicals. Using machine-learning potentials with selective Density Functional Theory calculations, the system identifies optimal adsorption energies ~2,000× faster than traditional approaches. The team also released the Open Catalyst 2020-Dense Dataset, a benchmark with ~100,000 configurations to standardize catalyst discovery research.