GPU-Accelerated Molecular Dynamics Simulation of Rigid Water

All simulations of physical systems employ the use of numerical algorithms and approximations to accurately mimic a system\u27s behavior. In classical molecular dynamics (MD), there exists an important tradeoff between the desired accuracy and time scale of a simulation. To decrease computational load and preserve long time scales, MD simulations have often ignored long-range physical interactions among particles after some arbitrary cutoff distance. In many cases, as this cutoff distance is decreased, the accuracy of a simulation significantly degrades. In biochemical systems with relatively large amounts of polar water molecules, the error introduced through the use of cutoff distances or other approximation schemes could be significant. The purpose of this research is to determine if long-range coulombic contributions from polar water molecules have a nontrivial impact on the accuracy of recent and historical MD simulations. As a first step, we have constructed and partially validated a classical MD simulation of rigid (i.e., fixed-geometry) water molecules. As future simulations will be highly computationally-intensive, the underlying code has been built to utilize the recently-realized power of the graphics processing unit (GPU)