(683g) Microkinetic Modeling Strategy Considering Topologic Aspect of a Cu Catalytic Surface Based on Kinetic Monte-Carlo

It is obvious that surface reactions occur between “neighboring” reactants adsorbed on the surface, but conventional microkinetic modeling has not been conducted majorly in this respect. Traditional microkinetic modeling methods, also in our previous study too, consist only of surface coverage, which is the ratio how much a reactant is adsorbed on the surface, -oriented rate equations. Moreover, the reaction of the gaseous reactants is usually calculated by equations based on chemical equilibrium, and the surface reactions are usually calculated by equations based on reaction kinetics. We thought that this inconsistency should be improved so that we used kinetic Monte-Carlo method on microkinetic modeling.

Incorporating kMC simulation into a microkinetic model provides three advantages:

1. The model can take it account into that the topologic effects of surface reactions.
2. All reactions, including gas phase reactions, can be modeled based on reaction kinetics not on equilibriums.
3. The diffusion reactions on the catalytic surface which is not possible with conventional microkinetic plug-flow reactor (PFR) 1D modeling, can be considered.

On the other hand, there are two major disadvantages of the method of simulating the catalytic surface with kMC:

1. The computational load is greatly affected by the size of the catalytic surface.
2. The calculation time is much long compared to the general microkinetic model, which is disadvantageous for parameter estimation.

In order to solve the first problem, this study used periodic boundary unit cell structure to overcome the finite size effect problem.[2,3] Also, to solve the second problem, we derived the correlation between kinetic constants in two different scales of which are molecular level based on kMC, -number oriented form and continuum level based on general microkinetic model, -concentration oriented form. To drive that correlation, we used Gaussian Process (GP) based on Bayesian Optimization using Deep Neural Network (DNN) and performed parameter estimation using microkinetic model using the rate constant oriented from kMC simulations. [4]

Materials and Methods

Our system is Cu catalytic system consists of syngas which is a composition of CH4, CO and CO2 gases. The Cu catalytic surface of our system is based on Cu (111), and the lattice structures of unit cell are varied with 10-by-10, 20-by-20 and 50-by-50. Experimental data of adsorption / desorption reactions on Cu surface at equilibrium will be used to estimate kMC parameter, especially on desorption kinetics.

The reactor type is PFR and it is modeled by a series of Continuous Stirred Tank Reactors (CSTRs). With no assumption of rate determinate step (RDS), the model will be a microkinetic model with total 26 reactions which are taken from the work of C. V. Ovesen. et al. consist with five adsorption/desorption reactions and eight reversible, forward and backward surface reactions. [5] Kinetic constants, pre-exponential factors, activation energies and binding energies are taken from our previous study and Ovesen, C.V., et al.’s work which are tuned with experimental data. [1,5]

The tools we used are total two. One for DNN is TensorFlow® library of python and another for kMC simulation and optimization of microkinetic model with kinetics is MATLAB®.

Significance

In point of science, it has its novelty at taking topological reaction into account microkinetic modeling. In addition, we can present the adequate strategy for modeling a desorption kinetics from continuum scale experimental data which has been unclear yet.