Rate-Determining Step or Rate-Determining Configuration? The Deacon Reaction over RuO₂(110) Studied by DFT-Based KMC Simulations

Abstract

Ab initio kinetic Monte Carlo (KMC) is successfully applied to simulate the experimentally observed promoting effect of O 2 on the HCl oxidation reaction (Deacon process) catalyzed by RuO 2 (110). Density functional theory (DFT) calculations provide, in addition to the adsorption energies of reaction intermediates and activation energies, also interaction energies between the adsorbates within the cluster expansion approach. KMC simulations with this extended set of energy parameters were analyzed employing the concept of "degree of rate control". In contrast to previous propositions, our simulations indicate that neither the dissociative O 2 adsorption (the sterically hindered first reaction step) nor the associative desorption of chlorine (the step with the highest activation energy) are rate determining under typical Deacon conditions. Instead, the hydrogen transfer in the water formation determines the rate of the overall reaction. These hydrogen transfer processes are not highly activated but turn out to be strongly configuration controlled.

Efficient Implementation of Cluster Expansion Models in Surface Kinetic Monte Carlo Simulations with Lateral Interactions: Subtraction Schemes, Supersites, and the Supercluster Contraction

“…Equation is used only for the computation of the energy of the basins, that is, the energies of initial ( E I ( σ )) and final states ( E F ( σ )) of reactions on the surface under the influence of lateral interactions. For the computation of the energies of transition states, we follow the approach of Nielsen et al, where the activation energy under the influence of lateral interactions, E A (σ), is computed from the activation energy extrapolated to zero coverage, E A ,0 . Defining a reaction energy and similarly, a reaction energy at zero coverage: the activation energy under the influence of lateral interactions is computed as …”

“…On the other hand, the term from the search step will be even less noticeable. To demonstrate that our findings for the SSS‐VSSM algorithm represent the performance of real KMC simulations, we provide a benchmark for full KMC simulation of the HCl oxidation over RuO 2 (110) in section S3 in the Supporting Information. We find that SSS‐VSSM performs equally well in the simulation of a relevant surface reaction as it does for the toy system.…”

“…With the help of Grand canonical Monte Carlo (GCMC) and KMC simulations, it has become possible for the first time to explicitly study the effect of lateral interactions and adsorbate layer ordering on the kinetics of surface reactions . It has been thus shown that lateral interactions can have a dramatic influence on the reaction rate (several orders of magnitude), microkinetic reaction parameters (apparent activation energies and reaction orders), and even the relative catalytic activity of different crystal facets .…”

“…However, in systems with lateral interactions, M scales exponentially with the number of interaction parameters ( M ≫ N ), which can be an issue even in simulations with “relatively simple” interaction models . In more practical models, such as our models for the HCl and CO oxidation over RuO 2 (110), for instance, the number of distinct adsorption energies is around 10 9 . With approximately 10 11 different rate constants, rate class/type/catalog‐based algorithms become inadequate.…”

“…In the KMC method, the active sites of the catalyst are mapped onto a lattice, where the chemical master equation (CME) is solved directly through stochastic simulation. This allows for the simultaneous, exact treatment of spatial correlations on the surface that may be caused by adsorbate confinement or slow diffusion, [26,27] island formation, [28][29][30] lateral interactions, [1,14,18,19,[31][32][33][34][35] or substrate heterogeneity. [30,36,37] The numerical solution of the CME by KMC can be quite costly and, for some problems, there may be alternative approaches, [2,[38][39][40] such as solving the CME in Fourier space [41] or the tensor-train approximation.…”