Designing Catalyst\nDescriptors for Machine Learning\nin Oxidative Coupling of Methane

Abstract

Catalysts\ndescriptors for representing catalytic activities\nhave\nbeen challenging in regard to machine learning. Machine learning and\ncatalyst big data generated from high-throughput experiments are combined\nto explore the catalyst descriptors. Catalyst descriptors are designed\nusing the physical quantities from the periodic table in the oxidative\ncoupling of methane (OCM) reaction. Machine learning unveils the five\nkey physical quantities representing ethylene/ethane selectivity (C₂s) in the OCM reaction, where machine learning predicted three\ncatalysts to have high C₂s values. Experiments confirm\nthat the proposed three catalysts have high C₂s values\nin the OCM reaction. Hence, the physical quantities can be used as\nalternative descriptors for designing heterogeneous catalysts.