(8e) Uncertainty Quantification of Catalyst Structure Effects on Kinetics

Heterogeneous catalytic reactions are commonly structure sensitive and dependent on the size and shape of catalyst particles. However, the catalyst's structure effects on reactivity are often neglected in state-of-art microkinetic modeling by assuming the reaction happens only on a single crystallographic plane [1]. This material gap in identifying and modeling the real active site intrinsically hinders the understanding of complex reaction mechanisms at the atomic level. Published experimental data have long served as a valuable resource for bridging this gap. However, the reported literature data often exhibit significant differences for even identical systems. This fact complicates meaningful comparisons across sources and hinders the utilization of prior knowledge. To alleviate this problem, it is important to quantify the uncertainties caused by catalyst structure effects.

Complete methane oxidation on noble metals is chosen as a case study due to many available published literature data. The structure sensitivity of this reaction has been debated experimentally for decades [2] due to the difficulty in identifying catalyst surface structures in situ. Structure-dependent kinetic modeling provides the possibility to explore it theoretically. Recently published computational work either chooses several common facets [3] or develops a structure-dependent model for archetypical simple reactions, such as CO oxidation [4]. There is a need to create suitable microkinetic models for describing structure sensitivity and explore structure effects.

In this work, we develop a novel methodology to build a structure-descriptor-based kinetic model for complete methane oxidation based on first-principles calculations. Structure-reactivity scaling relationships, reminiscent of the generalized coordination number (GCN) [5], are developed using machine learning techniques. By incorporating these correlations into kinetic models leveraging our in-house developed software, we predict the experimental observables, such as turnover frequencies (TOF) and apparent activation energies, at a dramatically reduced computational cost compared to first-principles calculations. Additionally, uncertainty quantification is applied for exploring the effects of errors in structure-reactivity scaling relations on variable catalyst facets. This methodology enables the rapid prediction of kinetics and quantifies the uncertainties due to the catalyst structure.

References

1. Cheula, R., Soon, A., and Maestri, M. Catal. Sci. Technol. 8, 3493 (2018).

2. Beck, Irene E. et al. J. Catal. 268, 60 (2009).

3. Wang, Y. et al. J. Phys. Chem. C 124, 2501 (2020).

4. Jørgensen, M. and Grönbeck, H. ACS Catal. 7, 5054 (2017).

5. Calle-Vallejo, F., Sautet, P. and Loffreda, D., Angew. Chemie - Int. Ed. 53, 8316 (2014).