Dresden 2017 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 107: Heterogeneous Catalysis: Theory II
O 107.8: Vortrag
Freitag, 24. März 2017, 12:15–12:30, TRE Phy
Solving the master equation without kinetic Monte Carlo: Tensor train approximations for a CO oxidation model — •Patrick Gelß1, Sebastian Matera1, and Christof Schütte1,2 — 1Freie Universität Berlin, Germany — 2Zuse Institut Berlin, Germany
Kinetic Monte Carlo (kMC) simulations have become an important tool for modeling chemical kinetics on catalytic surfaces. Their appealing feature is the unbiased solution of the Markovian master equation, which results from a given microkinetic mechanism, while not being affected by the curse of dimensionality. However, the need to perform one reaction step after the other makes kMC ineffective for stiff problems, which are characterized by a large disparity in the values of the employed rate constants. We have developed an alternative approach with tunable accuracy, which directly solves the master equation by exploiting the Tensor Train Format[1]. Using a reduced model for the CO oxidation on RuO2(110), we benchmark the approach against highly accurate kMC simulations. We demonstrate that numerical accuracy and linear scaling in the system size can be achieved for a large range of input parameters. The advantage over the kMC approach is illustrated for a problem with increasing stiffness, where our approach is hardly affected but the computational costs for kMC explode.
[1] P. Gelß, S. Matera, C. Schütte, J. Comput. Phys., 314, pp. 489–502 (2016)