Dresden 2014 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
CPP: Fachverband Chemische Physik und Polymerphysik
CPP 37: Poster Session 2
CPP 37.52: Poster
Mittwoch, 2. April 2014, 15:00–19:00, P3
Automatic, optimized interface placement in Forward Flux Sampling simulations — •Kai Kratzer1, Axel Arnold1, and Rosalind J. Allen2 — 1Institute for Computational Physics, University of Stuttgart, Allmandring 3, 70569 Stuttgart, DE — 2SUPA, School of Physics, The University of Edinburgh, Mayfield Road, Edinburgh EH9 3JZ, UK
With Forward flux sampling (FFS) rare events like crystallization phenomena in equilibrium or non-equilibrium systems can be simulated by driving the system from its initial state to the final state via a set of interfaces in phase space. The efficiency as a combination of computational cost and statistical error of FFS depends on the locations of these interfaces. We present two methods for placing the interfaces automatically at optimal locations and on-the-fly during the simulation, without knowledge of the underlying energy landscape of the physical system. Using these methods only the borders of the initial state and the final state in terms of an order parameter must be defined. Then, the simulation finds its own way through phase space by means of inexpensive exploring simulations. This facilitates the setup of an FFS simulations and increases the efficiency of FFS tremendously by placing more interfaces at bottlenecks of the system where the probability of advancement is lower. The method aims at establishing a constant flux through all interfaces, which not only maximizes the efficiency, but also avoids biasing due to bad sampling at interfaces with very low fluxes.