Understanding the Source of Error in First-Principles-Based Micro-Kinetic Modeling: Density Functional Theory Calculations Versus the Mean-Field Approximation

Abstract

First-principles-based micro-kinetic modeling has become an essential tool in the rational design of heterogeneous catalysts. Currently, most theoretical understandings and predictions are provided by this model, which combines results from density functional theory (DFT) calculations at the generalized gradient approximation level with the simulations using the mean-field micro-kinetic modeling (MF-MKM). In spite of its popularity and success in the catalyst design and screening, sometimes this combination yields undesired predictions that significantly deviate from experimental observations. It is, therefore, important to understand its success and failure and to locate the bottleneck of applying/improving the conventional MF-MKM. To this end, we have compared the errors that come from different DFT functionals with those introduced by the MF approximations against the kinetic Monte Carlo (KMC) results for the Pt-catalyzed water gas shift reaction system. It was found that the shape of the volcano map is less sensitive to the differences in the DFT functionals, even though their predictions of the key energetic parameters are distinctly different. On the other hand, the top of the volcano map, the detailed mechanism, as well as some key kinetic properties are markedly affected by the change of the energetic data. In comparison, the traditional MF-MKM that neglects the adsorbate–adsorbate interactions totally fails to describe the reaction kinetics. The inclusion of the adsorbate–adsorbate interaction effect can significantly improve the performance of the MF-MKM, which results in largely similar predictions to the reference data provided by KMC simulations. Nevertheless, the missing spatial correlation in the MF treatment can lead to an improper description in the high coverage region. Our analyses demonstrate that the accuracy of the MKM might not always be improved by only modifying the DFT-calculated energetic parameters, and close attention should be paid to the kinetic simulation method employed, particularly when quantitative kinetic properties are to be expected.