Adsorption Energies on Electrocatalyst Surfaces Using High-Throughput Experimentation and Machine Learning

One of the most important ways to close the anthropogenic carbon dioxide (CO₂) cycle is by electro-reduction of CO₂ into value-added chemicals and sustainable fuels. One of the main aspects in electrochemical CO₂ reduction reaction (CO₂RR) is the catalyst, which decreases the dissociation energy of CO₂. Various materials have been investigated as catalysts for CO₂RR, yet, generally, high selectivity for desired products of interest, namely C₂₊ carbonates, is low. One approach for increasing product selectivity is to discover new materials, based on metals and metal oxides with more components (<i>e.g.</i>, quaternary systems), which may improve activity, as a result of synergistic effects between the composing elements. Yet, the parameter space for these systems is huge and must be focused. One feature that describes the reaction propagation pathway is intermediate adsorption energy, <i>e.g.</i>, *H, *CO, which could assist in determination of product selectivity. However, adsorption energies of known and new materials are not always experimentally determined and have not been studied systematically.<br/>Here, we tackle the problem of determining adsorption energies by experimental measurement and machine learning predictions. We built a machine learning model that aims to predict the adsorption energies of known materials, which we plan to implement on new materials. The model is based describing the adsorption environment using simple and direct features, and more elaborate descriptors related to material structure and geometry using orbital field matrix.^[1] We also use high-throughput synthesis and characterization methods to gain more relevant data points of adsorption energies on known and new multinary materials. The coupling of machine learning and high-throughput materials science assists in narrowing the large multinary materials parameter space, which will lead to new selective CO₂RR catalysts.<br/><br/><br/><br/><br/><br/><br/><br/><br/>[1] T. Lam Pham, H. Kino, K. Terakura, T. Miyake, K. Tsuda, I. Takigawa, H. Chi Dam, <i>Sci. Technol. Adv. Mater.</i> <b>2017</b>, <i>18</i>, 756.