20 likes | 147 Views
A GPU Implementation of Coulomb Interaction in Molecular Dynamics Monica Olvera, Northwestern University, DMR 0907781.
E N D
A GPU Implementation of Coulomb Interaction in Molecular DynamicsMonica Olvera, Northwestern University, DMR 0907781 Ever increasing market demand for fast and realistic graphics has driven a rapid development of inexpensive GPU devices, with a doubling of computational power every 12 months. In recent years, the GPU hardware has become available to non-graphical applications through the advent of general-purpose programmability of the device. A notable example is molecular dynamics (MD) with reports of GPU implementations achieving speed-ups in excess of 100 times compared to standard MD codes. We report a GPU implementation of long-range electrostatic interactions. Our implementation is significantly faster than the CPU implementation of the standard Ewald method for small to a sizable number of charged particles (~105) in electrolyte solutions. Prateek K. Jha, Rastko Sknepnek, Guillermo Ivan Guerrero-Garcia, Monica Olvera de la Cruz Journal of Chemical Theory and Computation20106 (10), 3058-3065
A GPU Implementation of Coulomb Interaction in Molecular Dynamics Monica Olvera, Northwestern University, DMR 0907781 Long-range electrostatic interaction plays a central role in a large number of scientific and technological problems. In addition, electrostatics is the driving force behind most biological processes in cells. Due to its long-range nature, exact treatment of electrostatic interaction, like the widely used Ewald summation, is very costly. On the other hand, the GPU implementation of an angle-averaged scheme is very fast and proven to be reliable when applied to electrolytes. The implementation in the HOOMD package, which will become public in the near future, will be useful to scientists and engineers modeling a broad range of problems involving charges. The physical properties of electrolytes are unknown in many important industrial and biological settings due to the difficulty in introducing explicitly all the charged components in simulations. The new GPU implementation in molecular dynamics will aid to obtain efficiently and accurately thermodynamics and structural properties of complex electrolytes.