Optimizing transition interface sampling simulations

Abstract

We demonstrate that a recently proposed adaptive optimization algorithm for forward flux sampling simulations [E. E. Borrero and F. A. Escobedo, J. Chem. Phys. 129, 024115 (2008)] can be easily applied within the framework of transition interface sampling. This optimization algorithm systematically identifies the kinetic bottlenecks along the order parameter used to partition phase space via interfaces and improves the statistical accuracy of the reaction rate constant estimate. In different versions of the algorithm, the number or the placement of the interfaces (or both) are varied in order to allocate the numerical effort in a balanced way. The algorithm is demonstrated for a simple two-dimensional model and for the dipole flip transition of icelike structures inside carbon nanotubes. For these test systems, the optimization yielded an efficiency increase by a factor of 2-15.