SKM 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 19: Development of Computational Methods: Simulation Methods – Theory

MM 19.2: Vortrag

Dienstag, 28. März 2023, 14:30–14:45, SCH A 251

Increasing the Diversity of One-Sided Transition State Searches with Biasing Potentials — •Nils Gönnheimer, King Chun Lai, Karsten Reuter, and Johannes T. Margraf — Fritz-Haber-Institut der Max-Planck-Gesellschaft, Berlin, Germany

Chemical processes on catalytic surfaces are commonly described in terms of complex networks of chemical reactions. One-sided transition state (TS) searches are a powerful tool for exploring these networks, when not all reactants, products and intermediates are known. To this end, the dimer method is commonly used, since it avoids the computationally demanding calculation of second derivatives. However, a prominent disadvantage of the dimer method is the convergence to the same few TSs even with differently initialised simulations, which raises the possibility of overlooking reaction pathways. Here, we propose a method to overcome this limitation. Specifically, we use a biasing potential to modify the potential energy surface (PES) based on information gathered from previous calculations. This biasing potential is adaptively generated in a series of dimer searches to drive the searching algorithm away from already known TSs. Intuitively, such potentials can be built within the Cartesian space by placing an energy bias at the atomic positions of a found TS. Alternatively, a similarity kernel between the structure and the known TS based on a local environment descriptor can be employed. This ensures that the biasing potential is invariant to permutations between identical atoms, as well as rotations and translations. The performance of both approaches is discussed for adatom diffusion and island formation on the Pd(100) surface.