Improved Prediction for the Methane Activation Mechanism on Rutile Metal Oxides by a Machine Learning Model with Geometrical Descriptors

Abstract

Methane activation could occur via either the radical-like or the surface-stabilized mechanism on metal oxides. The linear Brønsted–Evans–Polanyi (BEP) relationship between activation energies and the adsorption energies of products has made it possible to swiftly predict some reaction mechanisms. However, it is not accurate enough to predict the preferential methane activation mechanism on metal oxides. Herein, to improve the prediction for the methane activation mechanism, the machine learning method percentile-LASSO was developed to extract energetic and geometrical descriptors on the basis of a series of surface-stabilized and radical-like transition states of methane activation on rutile-type metal oxides from density functional theory calculations. Revised relations are capable of classifying those two mechanisms on the same surface with a higher accuracy, which will facilitate high-throughput catalyst screening for methane activation on metal oxides.