The goal of the proposed project is to develop timescale parallelization techniques to rapidly advance on the timescale through tracking of multiple event trajectories. This will be achieved by developing parallel strategies for the recently invented timescale methods such as parallel replica dynamics, temperature accelerated dynamics, and forward flux sampling. These techniques can then be combined with spatial parallelization techniques to provide a multiplicative boost to the number of processors that can be e ectively utilized.
Extending molecular dynamics timescales by orders of magnitude while retaining atomic level details constitutes a revolutionary advance in furthering scientific and engineering knowledge through atomistic simulations. If successful, the proposed high-risk exploratory research will significantly increase the timescale of molecular dynamics simulation capabilities.
To ensure broad impact to the molecular dynamics community, the developed methods will be made available as part of next generation LAMMPS software package, which already has a substantial user base. Other open source programs will be released to enable computational scientists to take advantage of the proposed research. Through the project, a postdoctoral research associate will be trained in interdisciplinary research that combines advanced scientific computing and computational methods aimed for petascale system