A machine-learning-assisted study of propylene adsorption behaviors on transition metals and alloys: Beyond the Dewar-Chatt-Duncanson model

Summary

The interactions between propylene and heterogeneous catalysts play a crucial role in determining the catalytic performance in various propylene-related reactions. In this work, density functional theory (DFT) calculations and machine-learning techniques have been used to examine the adsorption behaviors of propylene on elemental transition metals and alloys. To predict propylene adsorption energies without DFT calculations, a set of intrinsic features and the random forest algorithm are employed to train a surrogate model. The analysis of frontier orbitals and density of states is then used to provide a physical interpretation of the observations by machine learning. Our results suggest the transition metal-propylene interactions are not only due to the electron transfer between the d states and the π bonding and π^∗ antibonding orbitals in the C=C double bond, but they also are influenced by the filling and energy levels of the metal valence s and p orbitals, which is well beyond the Dewar-Chatt-Duncanson model.