DOE OSTI · 960620
Accelerating atomistic simulations through self-learning bond-boost hyperdynamics
Abstract
By altering the potential energy landscape on which molecular dynamics are carried out, the hyperdynamics method of Voter enables one to significantly accelerate the simulation state-to-state dynamics of physical systems. While very powerful, successful application of the method entails solving the subtle problem of the parametrization of the so-called bias potential. In this study, we first clarify the constraints that must be obeyed by the bias potential and demonstrate that fast sampling of the biased landscape is key to the obtention of proper kinetics. We then propose an approach by which the bond boost potential of Miron and Fichthorn can be safely parametrized based on data acquired in the course of a molecular dynamics simulation. Finally, we introduce a procedure, the Self-Learning Bond Boost method, in which the parametrization is step efficiently carried out on-the-fly for each new state that is visited during the simulation by safely ramping up the strength of the bias potential up to its optimal value. The stability and accuracy of the method are demonstrated.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Perez, Danny, Voter, Arthur F. 2008-01-01. Accelerating atomistic simulations through self-learning bond-boost hyperdynamics. https://searchnasa.com/record/osti%3A960620/accelerating-atomistic-simulations-through-self-learning-bond-boost-hyperdynamics
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Search NASA by Indiciae: BibTeX · RIS.