For catalyst screening, estimation of binding energy of adsorbates on transition metal surfaces constitute a key step, wherein adsorbates are expected to bind optimally with the surface, following the Sabatier principle. In this regard, previous attempts have focused on estimating the d-band center of the transition metal catalysts, which serves as a descriptor to dictate the strength of adsorbate-metal interaction via the hybridization of the adsorbates’ atomic orbitals and surface d-band of the metal<sup>1</sup>. However, this descriptor tend to fail on the surfaces of Group 11 elements (Cu, Ag, and Au), since d-band is mostly filled and repulsive interaction due to Pauli’s exclusion principle dominates. Since most of the single atom alloys (SAAs) of transition metal catalysts developed today for various reactions employ a lesser active host from the Group 11 elements, modeling the catalytic activity of such surface poses a greater challenge. Herein, an attempt is made to develop a comprehensive descriptor to predict the binding energy of carbon and oxygen atoms on single atom alloys (SAAs) of Group 11 elements using a machine learning (ML) approach. We have selected a gradient boosting regression (GBR) algorithm to predict the binding energies on SAAs of Cu, Ag and Au by screening an array of available methods (viz. Linear Regression, Support Vector Regressor, Random Forest Regressor and Extra Tree Regressor). Readily available periodic table properties of the transition metals are utilized as input features in the model and their relative contribution in adsorbate-metal interaction is accessed. The ML model is trained with adsorption energies calculated from density functional theory (DFT) simulations. In test runs, ML based model is observed to predict carbon and oxygen binding energies comparable to DFT simulations, with significantly reduced and negligible (0.0007 s) computational time. Overall, the ML model provides a rational for feature selection from periodic properties of the elements to synthesize catalytically active SAAs with the Group 11 elements for high throughput in-silico screening.

Understanding the trends in binding energies of single atom alloys of Cu, Ag, and Au using a machine learning approach

ACS Spring Meeting 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.