Scalable molecular dynamics

Molecular dynamics simulations enable the study of the time evolution of molecular systems by taking many small successive time steps under atomic forces that are calculated from a parameterized set of interaction functions. These are simple functions describing bonded and non-bonded atomic interactions, so that large molecular systems can be simulated for many time steps. The simulations provide energetic and kinetic properties in the form of statistical ensemble averages. The resulting trajectories can be analyzed for a variety of geometric and kinetic properties and correlations between them. These simulations have been carried out for many different systems, especially in computational biochemistry. NWChem(1) is the computational chemistry software suite developed and maintained by the Molecular Sciences Software Group (MSSG) of the Environmental Molecular Sciences Laboratory at the Pacific Northwest National Laboratory. This software has been designed to run efficiently on a variety of massively parallel computer architectures, including distributed memory (e.g. IBM-SP, Linux clusters) and shared memory (e.g. CRAY-T3E) architectures. In NWChem the molecular dynamics simulations 2 module is based on domain decomposition. This allows reduction of memory requirements through the distribution of data across many processors as well as reduction of communication requirements because to the locality of interactions in cut-off based atomic interaction lists. This type of decomposition does require sophisticated load balancing techniques to be implemented, however, because of the heterogeneity of biological systems. In NWChem we have implemented a combination of global and local load balancing techniques, resulting in a significant increase of the parallel scaling that can be achieved. In NWChem we use the AMBER force field 3 with the GLYCAM extensions(4) for simulations of saccharides.