Work

On-the-fly Neural Network based Kinetic Monte Carlo Simulations of NH3-SCR over Cu-CHA

Year: 2026
Type: preprint
Abstract: Copper exchanged chabazite (Cu-CHA) is the preferred choice for abatement of NOx from lean burn engines. The low-temperature activity depends critically on the presence of paired [Cu(NH3)2]2 + complexes, which are required for O2 adsorption. The pairing of [Cu(NH3)2]2 + complexes has so far been studied by ab initio molecular dynamics or more recently molecular dynamics using machine learning force-fields. An alternative to molecular dynamics is kinetic Monte Carlo (kMC) simulations, which enable simulations over longer time scales. In this study, an on-the-fly kMC model is developed where the rates for [Cu(NH3)2]2 + diffusion are evaluated on the fly using a machine learning force field (ML-FF). We find that the fraction of paired complexes depends sensitively on the zeolite composition and the aluminum distribution. The results reveal a large distribution of Cu pair stabilities and that only few Cu ions are paired, demonstrating the heterogeneous nature of Cu-CHA. A mean field model based on the fraction of paired complexes is used to explore how changes in zeolite composition affect the catalytic activity. The methodology of combining kMC and ML-FF to account for dynamic changes in catalyst structure during reaction conditions is general and provide a possibility to go beyond the often used static approximation when evaluating reaction rates.
Source: ChemRxiv
Language: English
Open Access status: green
OpenAlex ID: W7225901629
Cites:
Cited by:
Related to: 10