(509dd) Physics Based Machine Learning Modeling of O and OH Adsorption on Oxide Surfaces

Discovering materials that can efficiently perform the oxygen evolution reaction (OER) and oxygen reduction reaction (ORR) is a key to producing efficient fuel cell technologies. The primary processes that control this are the adsorption of O and OH. In this talk, we seek to examine this process on oxide materials. Understanding the adsorption properties of metal oxide surfaces is significantly more challenging than that of metals due to the complex interactions between metal centers and the neighboring non-metals as well as the intricate crystal structures of the materials. The adsorption of O and OH on oxide surfaces is studied using density functional theory (DFT) and crystal orbital Hamiltonian projections (COHP.) Based on these calculations we provide a systematic understanding of the electronic structure of O and OH adsorption on oxide materials based on ligand field theory. With this understanding, we generate relevant features that are used in conjunction with machine learning methods to construct a quantitative model of the adsorption process. Finally, regularized regression is used to build a model of the adsorption energies. This work provides a framework for understanding and quantifying the bonding of O and OH to oxide materials, allowing for the easy prediction of adsorption energies on new materials.