We have harnessed the power of the Open Catalyst Project (OCP) and used machine learning (ML) models to predict oxide catalysts for the oxygen evolution reaction (OER) needed for water electrolysis to produce hydrogen (H2). Our material library spans facets up to a maximum Miller index of 1 for 4,728 oxide materials resulting in over 200,000 total energy predictions of bare slabs and over 4 million predictions for surface intermediates. Our screening criteria take into consideration the synthesizability, aqueous stability, material cost, and overpotential of each facet for every material. To increase the pool of stable candidates, nanoscale stability was also considered. Our search resulted in a small set of new candidate materials and their predicted properties could be verified with more reliable density functional theory (DFT) simulations. Despite this exhaustive search, the steady-state kinetics of the OER reaction remain sluggish stable candidate materials. Thus, we investigated promoting the OER using microkinetic modeling of forced catalyst dynamics. We found that programmable oxides could increase current density at a fixed overpotential or reduce the overpotential required to reach a fixed current density. The key parameters controlling the quality of the catalytic ratchet were the O*-to-OOH* and O*-to-OH* activation barriers. Overall, the combination of ML-assisted material screening and dynamic promotion may be a viable strategy to lower the cost of green H2 from water electrolysis.

Guest Speakers

Speaker:  Lars Grabow
Dan Luss Professor, Professor of Chemistry

Visit Their Profile...