The rationalization of catalytic processes relies on the fundamental understanding of competing reaction mechanisms driving reactants to products. The list of elementary steps composing the reaction networks is proposed based on chemical intuition and evaluated via density functional theory. This approach is limited by the size of the network and disregards alternative paths. Here we present the Catalytic Automated Reaction Evaluator (CARE), a flexible end-to-end framework for heterogeneous catalysis composed of (1) a rule-based reaction network generator, (2) a thermodynamic and kinetic parameter evaluator powered by state-of-the-art machine learning models and (3) a fast microkinetic solver. CARE reproduces the experimental activity trends in methanol decomposition, identifies the selectivity to C3 products in CO2 electroreduction and generates the Fischer–Tropsch synthesis mechanism including 370,000 reactions reaching C6 products. This comprehensive framework enables the exploration of thermal and electrocatalytic reactions previously not amenable to atomistic simulations. This study presents a framework for the automated generation of reaction networks in heterogeneous catalysis. Powered by state-of-the-art machine learning models, the framework enables the investigation of thermal and electrochemical processes not amenable to density functional theory. The capabilities of its kinetic module are demonstrated by simulating Fischer–Tropsch networks with 37,000 reactions.
Morandi et al. (2026) studied this question.