Improved Prediction\nfor the Methane Activation Mechanism\non Rutile Metal Oxides by a Machine Learning Model with Geometrical\nDescriptors

Abstract

Methane activation could occur via either the radical-like\nor the\nsurface-stabilized mechanism\non metal oxides. The linear Brønsted–Evans–Polanyi\n(BEP) relationship between activation energies and the adsorption\nenergies of products has made it possible to swiftly predict some\nreaction mechanisms. However, it is not accurate enough to predict\nthe preferential methane activation mechanism on metal oxides. Herein,\nto improve the prediction for the methane activation mechanism, the\nmachine learning method percentile-LASSO was developed to extract\nenergetic and geometrical descriptors on the basis of a series of\nsurface-stabilized and radical-like transition states of methane activation\non rutile-type metal oxides from density functional theory calculations.\nRevised relations are capable of classifying those two mechanisms\non the same surface with a higher accuracy, which will facilitate\nhigh-throughput catalyst screening for methane activation on metal\noxides.