Constructing accurate force fields for biological materials such as DNA and proteins is a challenging problem. In this talk, we focus on the computation of accurate and transferable net atomic charges (NACs) using Density Functional Theory followed by Density Derived Electrostatic and Chemical (DDEC) analysis. The DDEC method simultaneously optimizes net atomic charges, atomic dipoles and quadrupoles, and atomic spin moments to reproduce realistic chemical states and the electromagnetic field surrounding a material. However, this method was originally programed in a noncompiled language. This presentation describes a new Fortran program parallelized with OpenMP that achieves parallelization efficiencies of approximately 65% on 32 processors. Compared to the noncompiled serial program running on a single processor, the parallel Fortran program running on 32 processors is more than 200 times faster. The computational cost scales linearly with increasing system size. The following systems are shown as examples: DNA and proteins (containing approximately 1000 atoms). Comparisons are made for different DNA molecules and conformations to demonstrate charge transferability. This methodology will facilitate the construction of flexible force-fields for biological materials.

Improving the computational efficiency of DFT-derived net atomic charges for flexible force fields of biological materials

247th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.