(509dc) High Throughput Surface Stability Analysis of Alloy Catalysts Using Density Functional Theory and Machine Learning

Pt-alloy catalysts, especially Pt₃Ni, have demonstrated excellent catalytic activity for a range of electrochemical reactions, including the oxygen reduction reaction (ORR)—integral to the efficient operation of a fuel cell. However, in terms of stability­—an important criterion for their practical application—several issues remain poorly understood, especially regarding the influence of mechanisms such as surface segregation, leaching, and oxidation. In the present work, we develop a generalized computational framework utilizing a combination of Density Functional Theory (DFT), ab-initio thermodynamics, and machine learning to analyze the surface stability of alloy catalysts. Applying this framework to Pt₃Ni, we discover an inverse relationship between surface coordination and “Pt-skin” stability under vacuum, thus shedding some new light on structure sensitivity to segregation. We further discuss the impact of this insight on a Pt₃Ni catalyst through a virtual nanoparticle model. The framework is next used to probe the Pt₃Ni surface in an electrochemical environment, revealing the formation of a Pt-shell at least two atomic layers thick due to the segregation and subsequent leaching of Ni. Finally, to ensure thorough navigation of the search space, a crystal graph convolutional neural network^1,2 is used to extend the scope of the framework to encompass more surface layers and complex facets, like steps. We close with a brief discussion on the application of this framework to study the stability of multimetallic alloy catalysts for the ORR.

References:

(1) Xie, T.; Grossman, J. C. Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties. Phys. Rev. Lett. 2018, 120 (14), 145301. https://doi.org/10.1103/PhysRevLett.120.145301.

(2) Palizhati, A.; Zhong, W.; Tran, K.; Back, S.; Ulissi, Z. W. Toward Predicting Intermetallics Surface Properties with High-Throughput DFT and Convolutional Neural Networks. J. Chem. Inf. Model. 2019, 59 (11), 4742–4749. https://doi.org/10.1021/acs.jcim.9b00550.