(558ac) Discover Perovskites for Electrocatalytic Water Oxidation Via Adaptive Machine Learning

Perovskites with the ABO₃-type structure have shown great promise for catalyzing oxygen evolution in the (photo)-electrochemical systems. However, it is very time-consuming and costly to search for highly optimized catalysts by either empirical testing or quantum-chemical calculations. In this regard, a low cost and accurate predictive model is attractive. In recent years, we have seen a dramatic rise of the implementation of machine-learning techniques for accelerating discovery of metallic catalysts[1–3]; Nevertheless, developing a machine-learning approach for the search of complex oxide catalysts is much more challenging due to the insufficient knowledge about the materials’ structure-activity relationships and heterogeneity/scarcity of available data.

In this talk, we demonstrated a probabilistic adaptive machine learning approach to navigate through a rich chemical space ~4000 double perovskites quickly for identifying the stabilized cubic structures with promising OER catalytic activity within the established descriptor-based kinetic framework. Notably, our model exhibits high predictive reliability for predicting the adsorption properties (e.g., *O and *OH), which is attributed to a heterogeneous set of descriptors with multi-fidelity informatics, i.e., compositional and electronic structure descriptor, and an adaptive learning process. By iteratively selecting the appropriate candidates for DFT validations where the Gaussian process regression model is most uncertain about in terms of the OER over potential, we are able to improve the model progressively by expending the training datasets toward as much chemical space as possible with the smallest computational effort. Our model successfully identifies most of the recognized perovskites along with ~50 potential candidates that have not been reported in the previous studies. Moreover, by interpreting the perovskites’ probability distributions with Kullback-Leibler (KL) divergence analysis, we draw important molecular orbital insight that the metal B site e_g orbitals of the perovskite bulk electronic structure is a significant underlying descriptor that governs the OER activity.

References

[1] Z. Li, S. Wang, W.S. Chin, L.E. Achenie, H. Xin, High-throughput screening of bimetallic catalysts enabled by machine learning, J. Mater. Chem. A Mater. Energy Sustain. (2017). doi:10.1039/C7TA01812F.

[2] Z. Li, X. Ma, H. Xin, Feature engineering of machine-learning chemisorption models for catalyst design, Catal. Today. 280 (2017) 232–238.

[3] X. Ma, Z. Li, L.E.K. Achenie, H. Xin, Machine-Learning-Augmented Chemisorption Model for CO2 Electroreduction Catalyst Screening, J. Phys. Chem. Lett. 6 (2015) 3528–3533.