Unraveling the Complexity of Catalytic Reactions via Kinetic Monte Carlo Simulation: Current Status and Frontiers

Abstract

Over the past two decades, the necessity for predictive models of chemical kinetics on catalytic surfaces has motivated the development of ab initio kinetic Monte Carlo (KMC) simulation frameworks. These frameworks have been successfully used to investigate chemistries of academic interest and industrial importance, such as CO oxidation, NO oxidation and reduction, ethylene hydrogenation, CO hydrogenation to ethanol, and water-gas shift. These studies have shed light on the effect of catalyst composition, surface structure, lateral interactions, and operating conditions on the apparent turnover frequency of the chemistries of interest. Yet, extending the existing KMC approaches to study large chemistries on complex catalytic structures poses several challenges. In this review, we discuss the recent milestones in the area of KMC simulation of chemical kinetics on catalytic surfaces and review a number of studies that have furthered our fundamental understanding of specific chemistries. In addition, we provide directions for future research aiming toward incorporating detailed physics and chemistry, as well as assessing and improving the accuracy of KMC methods, toward developing quantitative models of surface kinetics.

Information-theoretic tools for parametrized coarse-graining of non-equilibrium extended systems

“…As is typical in the KMC framework, we assume that the atomistic model can be described by a spatial, continuous-time Markov jump process, 36 . The path-space measure of this KMC process, see for the Markov Chain analogue of the path measure (6), is parametrized as P = P θ .…”

“…In the case of continuous time processes such as Kinetic Monte Carlo, the coarse-grained stochastic process is defined in terms of coarse transition ratesc(η, η ′ ) which captures macroscopic information from the fine scale rates c(σ, σ ′ ). For example, for stochastic lattice systems, approximate coarse rate functions are explicitly known from coarse graining (CG) techniques of 10,21,25 , see (36). Similarly, when we consider temporally discretized stochastic processes such as Langevin Dynamics, the coarsegrained process is given in terms of transition probabilitiesp(η, η ′ ) which capture macroscopic information from the fine scale transition probabilities p(σ, σ ′ ).…”

“…Nonequilibrium systems at transient or steady state regimes are typical in applied science and engineering, and are the result of coupling between different physicochemical mechanisms, driven by external couplings or boundary conditions. For example reaction-diffusion systems in heteroepitaxial catalytic materials, 36 are typically nonreversible, i.e., their steady state is not a Gibbs distribution. This is due to the fact that different mechanisms between species (reaction, diffusion, adsorption, etc.)…”

“…Typical examples include reaction-diffusion systems in heteroepitaxial catalytic materials, polymeric flows and separation processes in microporous materials, 32,35,36 . In this paper we develop reliable model-reduction methods, i.e., having controlled fidelity of approximation, and capable to handle extended, non-equilibrium statistical mechanics models.…”

A machine learning approach to graph-theoretical cluster expansions of the energy of adsorbate layers

“…[12][13][14] The advantage of model Hamiltonians is that they yield results much faster than the underlying DFT computations, while they are able to properly assess the configurational average and evolution of a surface under reactive conditions. In principle, the model Hamiltonian needs to account for three energy contributions: i) The energy of the surface, which includes deformation energies, ii) the interaction of adsorbates with the surface and iii) the interaction between adsorbates (also called lateral interactions).…”