Atomistic-Continuum Coupling for Heterogeneous Catalysis by a Reduced Basis Approach and Multilevel On-the-fly Sparse Grid Interpolation

• Applicants: Dr. Sebastian Matera; Dr. Christian Merdon
• Subject Area: Fluid Mechanics ; Chemical and Thermal Process Engineering ; Theoretical Chemistry: Molecules, Materials, Surfaces
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 467076359

The central objective of the project is the development of an improved multiscale modeling strategy for heterogeneous catalysis which couples atomistic kinetic Monte Carlo (kMC) simulations for the surface chemistry with Computational Fluid Dynamics (CFD) for reactive flows. One particular application will be in situ surface characterization experiments, which typically require complex flow geometries and in which the interplay between atomic and macroscale aspects plays a crucial role. We will follow two complementary approaches to improve two central aspects our existing methodologies. First, the CFD of catalytic flows in complex geometries comes at huge computational costs, even for analytical kinetic models, primarily due to the highly nonlinear surface kinetics. To lift this problem, a novel mass conservative reduced basis strategy (RBS) will be developed utilizing gradient-robust finite element and conservative finite volume schemes. This RBS decouples kinetics and transport and the later only requires the solution of a set of much simpler, primarily linear, problems. The chemical kinetics will only appear in a small non-linear problem defined on the surface. Besides the reduction of the cost by the decoupling, this allows to test several reactivity models without recomputation of the RBS. Therefore, we expect it to be a tremendously beneficial tool for the research community for the interpretation of the data. Second, in a kMC+CFD coupled simulation, the very largest part of the costs results from the kMC simulations, which constitute a nonlinear boundary condition to the CFD. To reduce the number of kMC simulations, we will complement the RBS with a multilevel on-the-fly interpolation (ML-OFI) of the kMC response which will serve as a surrogate in the CFD. Based on sparse grids, fast convergence will be ensured also for many reactive species. The ML-OFI will add only those grid points, which are needed to interpolate the KMC at the query points, and adjust the sparse grid resolution as well as the kMC sampling accuracy to the progress of the nonlinear iterative solver. As the RBS, the generated surrogate can be reused in other settings, e.g. different reaction chambers, without the need to recalculate all kMC data again.The combined methodology will be applied to coupled CO and NO oxidation on the Pd(100) surface employing first principles kMC models based on quantum chemical calculations. We thereby arrive at a multiscale modeling bridging between the electronic scales of chemical bond making and breaking and the macroscopic scales of reactor response. The strategy will be employed to analyze recent experiments and those conducted by our collaborators and to address the question of the active phase of Pd(100).

• DFG Programme: Research Grants
• International Connection: Sweden
• Cooperation Partners: Professor Dr. Edvin Lundgren; Dr. Johan Zetterberg