Speeding up the prediction of C-O cleavage through bond valence and charge on iron carbides

The activation of CO on iron-based materials is a key elementary reaction for many chemical processes. In this work, we investigate CO adsorption and dissociation on a series of Fe, Fe₃C, Fe₅C₂, and Fe₂C catalysts through density functional theory calculations. We detect dramatically different performances for CO adsorption and activation on diverse surfaces and sites. The activation of CO is dependent not only on the local coordination of the molecule to the surface but also the bulk phase of the underlying catalyst. The different local bonding environment affects the bulk properties leading to varying interactions between the adsorbed CO and the surface, thus yielding different activation levels of the C-O bond. We also examine the prediction of CO adsorption upon different types of Fe-based catalysts by performing machine learning through linear regression models. We combine the features originating from both surfaces and bulk phases to enhance the prediction of the activation energies. Eight different linear regressions utilizing feature engineering of polynomial representations were performed. Among them, a ridge linear regression with 2nd-degree polynomial feature generation predicted the best CO activation energy with a mean absolute error of 0.269 eV.