(60e) Physics-Guided Autonomous Design for Acid-Stable Water Oxidation Catalyst

A major challenge in the oxygen evolution reaction (OER) is to find a cost-effective catalyst that is active and stable. To date, OER catalysts with high activity in highly acidic environments have been noble metal-based catalysts such as IrO₂ ¹. However, since these catalysts are difficult to use due to their scarcity and high price, the formation of mixed metal oxides is required. Unfortunately, the reaction mechanism of OER, which is important for the development of OER catalysts, remains unclear ². In addition, it is a time-consuming process to test compositional combinations of numerous candidate catalysts in trial-and-error experiments because there are many things to consider, not only the optimal selection of elemental components, but also the determination of composition and crystal structure ³. Therefore, it is essential to develop a design methodology that improves catalyst performance while minimizing the number of experiments.

In our preliminary research, we tried to design the catalyst for acid oxygen evolution reaction (OER) in bimetallic system using multi-objective Bayesian optimization (MOBO) ⁴. An optimal catalyst composition that improves both current density and residual density fraction, two objective functions that are indicators of catalyst activity and stability, was proposed. As a result, the current density increased by 7.77 mA cm^-2 and the residual density fraction increased by 13% compared to the pure Ir-based catalyst. We extend the methodology of preliminary research and search for the trimetallic catalyst with ideal catalytic performance. However, as the number of alloying elements increases, the component exploration space expands, resulting in significant computational cost and an increase in the number of experiments to be performed. Therefore, we intend to develop a new methodology in which physical information such as first-principles calculations (density functional theory) is used to guide the optimization mechanism.

Here, we introduce the physics-guided autonomous discovery methodology by applying the DFT fidelity model to the BO-based optimal experimental design and present the Ir-Mo-Ni trimetallic acid stable OER catalyst. In contrast to previous studies, we add a fidelity function related to the two objective functions, perform the evaluation in two steps using an acquisition function, and approximate black-box functions with different fidelities. This allows us to efficiently search the composition space of a wide range of catalysts and reduce the number of experiments required for composition optimization. We expect to find more active and stable catalysts using the DFT-BO method than by using experimental data alone. Finally, when the physical information of materials can be roughly known in advance, as in DFT calculations, this method can be applied in many chemical fields as a generalization of physics-based BO.

References

1. Lyu, F., Wang, Q., Choi, S. M., & Yin, Y. (2019). Noble‐metal‐free electrocatalysts for oxygen evolution. Small, 15(1), 1804201.
2. Fabbri, E., & Schmidt, T. J. (2018). Oxygen evolution reaction—the enigma in water electrolysis. Acs Catalysis, 8(10), 9765-9774.
3. Reddington, E., Sapienza, A., Gurau, B., Viswanathan, R., Sarangapani, S., Smotkin, E. S., & Mallouk, T. E. (1998). Combinatorial electrochemistry: a highly parallel, optical screening method for discovery of better electrocatalysts. Science, 280(5370), 1735-1737.
4. Khan, N., Goldberg, D. E., & Pelikan, M. (2002, July). Multi-objective Bayesian optimization algorithm. In Proceedings of the 4th Annual Conference on Genetic and Evolutionary Computation(pp. 684-684).