Collectivity effect in cluster catalysis under operational conditions

Subnanometer metal clusters are promising catalysts but are limited by structural heterogeneity and dynamics under operational conditions. Herein, we employ artificial intelligence-enhanced multiscale modeling, integrated with statistical analysis, to exhaustively explore the catalytic sites of cluster catalysts under reaction conditions. We discover that numerous sites across varying sizes, compositions, isomers, and locations collectively contribute to overall activity due to their high intrinsic activity and abundance. The collectivity of active sites, despite their distinct local environments, configurations and reaction mechanisms, arises from their high intrinsic activity and considerable population. Data-driven machine learning reveals that this collectivity is governed by the balance between local atomic coordination and the adsorption energy. Using CO oxidation on Cu/CeO2 as an example, we validate the collective effect via agreement between computed mechanisms/kinetics and experimental data. This work provides insights into active sites in heterogeneous catalysis and highlights the potential of leveraging collective effects in cluster catalysis. Subnanometer metal clusters are limited by structural heterogeneity and dynamics under reaction conditions. Here, AI-driven multiscale modeling reveals collective activity of diverse Cu/CeO₂ sites via coordination-adsorption balance.