258250 Building a Better Pppm: Incorporating Long-Range Dispersion Interactions

Monday, October 29, 2012: 3:15 PM
415 (Convention Center )
Rolf E. Isele-Holder1, Paul S. Crozier2 and Ahmed E. Ismail1, (1)Mechanical Engineering, RWTH Aachen University, Aachen, Germany, (2)Multiscale Dynamic Material Modeling, Sandia National Laboraotry, DOE, Albuquerque, NM

The necessity and cost of long-range electrostatics in fully atomistic molecular simulations are by now universally accepted. While the necessity of including long-range dispersion forces in inhomogeneous systems, especially those containing surfaces, is  also well known, until recently, none of the major commercial or open-source molecular simulation packages have included long-range dispersion solvers. Thus, simulations have been forced to rely on approximations or post-processing corrections, both of which having their own shortcomings. We have extended the particle-particle particle-mesh (PPPM) method to treat dispersion (r-6) interactions, and implemented this method in the LAMMPS molecular dynamics package. To obtain maximum efficiency and flexibility, we included a derivation of the potential in reciprocal space (ik-scheme) as well as a derivation of the potential in real space (analytic differentiation scheme) for calculating the forces on the particles. We provide an error measure that enables an intelligent choice of grid parameters to improve both accuracy and efficiency. To further reduce the cost of the algorithm, we have introduced a data-transfer shortcut that reduces the number of Fast Fourier transforms required, which is of high benefit when using the ik scheme or a Lennard-Jones potential that uses the Lorentz-Berthelot mixing rule. In addition, this technique can be applied when performing simulations with particles that interact via both Coulomb and dispersion interactions. We will demonstrate under which conditions this feature is advantageous in these simulations. Furthermore, we present results on the accuracy, efficiency, and scaling of the implementation and compare it to other solvers to outline its utility.

Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000.


Extended Abstract: File Not Uploaded