|About this Abstract
||2016 TMS Annual Meeting & Exhibition
||Computational Methods for Spatio-temporal Scale-bridging: from Atomistics to Mesoscale
||Further Development of the Local Hyperdynamics Method for Accelerated Molecular Dynamics
||Dipanjan Ray, Danny Perez, Arthur Voter
|On-Site Speaker (Planned)
Recently, a new formulation of hyperdynamics for rare-event atomic systems was put forward, where an applied bias potential is determined separately for each bond, based only on its local environment. This modification allows for maintenance of a constant boost factor with increasing system size at a computational cost which scales linearly. While arguments for the correctness of local hyperdynamics have only been advanced for homogeneous systems, it has proven to give accurate event rates for many realistic large-scale systems with greatly enhanced acceleration compared to other methods, while retaining the equal treatment of the various reaction pathways characteristic of accelerated molecular dynamics. Here, we describe recent developments with the theory, implementation, and accuracy of local hyperdynamics, including the investigation of various local bias potentials and of inhomogeneous systems, and progress towards a massively parallel implementation which can track the continuous dynamical evolution of a system while running in linear time.
||Planned: A print-only volume