(538g) Automated DFT Calculations for the High-Throughput Screening of Oxygen Evolution Reaction Catalysts

For the past decades, quantum chemistry calculations have become reasonably accurate, and computer simulation-based materials discovery has become more feasible due to the exponential increase in computing power. More excitingly, recent advances in machine-learning (ML) have opened up the possibility of a high-throughput screening with the minimal number of expensive density functional theory (DFT) calculations.¹

DFT calculations on metal catalysts have elucidated reaction mechanisms and active sites, guiding experimentalists to further improve the catalytic properties such as activity and selectivity. For example, DFT studies have suggested that the Cu (211)² stepped surface and the Pt (111)³ surface are responsible for high catalytic activity for CO₂ electrochemical reduction and O₂ reduction reaction, respectively. On the other hand, ceramics such as metal carbides, oxides, sulfides, and nitrides have rarely been thoroughly studied since they are extremely demanding to model various facets/coverages/terminations due to their structural complexity. Therefore, the current strategy in the field has been to select the most stable facet and consider only a certain coverage limit in modelling, which could falsely predict the catalytic properties in some cases. In this work, we present a workflow to investigate complicated catalyst systems that enable us to consider multiple structures, facets, and coverages in a high-throughput fashion. We also propose a reasonable ML model to predict the binding energies and catalytic activities, which will further help to design high activity catalysts. As a prototypical example, we modelled several facets of many (existing and hypothetical) crystal structures of IrO_x (IrO₂ and IrO₃), and investigated O₂ evolution reaction (OER) activity of all possible unique sites.

Toward the high-throughput screening for oxide OER catalysts, we demonstrate our progress on the computational framework that automates DFT calculations for various oxides and utilizes ML techniques to reduce the number of DFT calculations.

1. Tran, K.; Ulissi, Z. W., Nature Catalysis 2018, 1 (9), 696.
2. Peterson, A. A.; Abild-Pedersen, F.; Studt, F.; Rossmeisl, J.; Nørskov, J. K., Energy & Environmental Science 2010, 3 (9), 1311-1315.
3. Markovic, N. M.; Gasteiger, H. A.; Ross, P. N., The Journal of Physical Chemistry 1996, 100 (16), 6715-6721.