Modern ab initio methods have rapidly increased our understanding of materials chemistry. However, due to computational expense, direct ab initio calculations are often unfeasible for large or complex systems such as those of interest for heterogeneous catalysis. Recent approaches in developing atomistic neural network (ANN) potentials based on ab initio data have opened up new opportunities for using classical molecular dynamics simulations to explore such systems with close to first-principles accuracy. In this talk, I will discuss our work developing and using ANN potentials to model several catalyst-water interfaces of interest for electrochemical water splitting and aqueous CO2 reduction. In addition, I will briefly discuss our efforts to develop neural network charge density functionals for various materials properties, including band gap and oxygen evolution reaction activity on perovskite oxides.

Combined ab initio and machine learning approaches for modeling catalyst-water interfaces

255th National Meeting (2018)

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