Atomic Design of Alkyne Semihydrogenation Catalysts via Active Learning

Alkyne hydrogenation on palladium-based catalysts modified with silver is currently used in industry to eliminate trace amounts of alkynes in alkenes produced from steam cracking and alkane dehydrogenation processes. Intensive efforts have been devoted to designing an alternative catalyst for improvement, especially in terms of selectivity and catalyst cost, which is still far away from that as expected. Here, we describe an atomic design of a high-performance Ni-based intermetallic catalyst aided by active machine learning combined with density functional theory calculations. The engineered NiIn catalyst exhibits >97% selectivity to ethylene and propylene at the full conversion of acetylene and propyne at mild temperature, outperforming the reported Ni-based catalysts and even noble Pd-based ones. Detailed mechanistic studies using theoretical calculations and advanced characterizations elucidate that the atomic-level defined coordination environment of Ni sites and well-designed hybridization of Ni 3d with In 5p orbital determine the semihydrogenation pathway.