Los Alamos National Laboratory
Long-timescale Simulations with Accelerated Molecular Dynamics and the EXAALT package
One of the common motivations coarse graining atomistic simulations is the inability of standard molecular dynamics (MD) to reach sufficiently long timescales. An alternative approach that can be particularly useful for systems that evolve through rare, thermally activated, events can be thought of as a coarse graining in time. Accelerated MD is a family of methods that implement different variants of this idea. I will review the current state of Accelerated MD, with an emphasis on the recently introduced Parallel Trajectory Splicing (ParSplice). I will then briefly describe the implementation of these methods in the EXAALT software package that combines ParSplice, the LAMMPS MD code, and the LATTE density functional tight binding code. This combination of software tools deployed on large-scale computers will enable simulations over a very large portion of the time/size/accuracy simulation space and also allow for acceleration of multiscale systems.
Danny is a staff member in the Theoretical Division at Los Alamos National Laboratory. His main research interest lies in the development, implementation, and application of methods for long-timescale atomistic simulations, with an emphasis on the Accelerated Molecular Dynamics techniques developed at LANL. He is particularly interested in methods that can exploit massive parallelism to extend simulation timescales. He applies these methods to a range of problems, including materials for fusion and fission energy production, and nano-materials. He is also the lead developer for the Accelerated MD module in EXAALT, a DOE exascale computing project that aims at developing the next generation of scalable MD simulation tools.