388079 A Parallel Fortran Code to Perform Density Derived Electrostatic and Chemical (DDEC) Analysis

Tuesday, November 18, 2014: 5:10 PM
212 (Hilton Atlanta)
Nidia Gabaldon Limas and Thomas A. Manz, Chemical & Materials Engineering, New Mexico State University, Las Cruces, NM

One of the primary challenges in the study of complex materials is to extract meaningful chemical descriptors that can be correlated to experimental properties. Net atomic charges (NACs), atomic spin moments (ASMs), and effective bond orders (EBOs) are examples of chemical descriptors that are commonly used to explain electrostatic, magnetic, and chemical properties of materials. NACs, ASMs, and EBOs computed with the Density Derived Electrostatic and Chemical (DDEC) method have good correlations to a wide range of spectroscopic experiments. In this talk, we explain the development of a parallel Fortran program to compute DDEC NACs, ASMs, and EBOs. This methodology can be applied to materials with 0, 1, 2, or 3 periodic boundary conditions with no magnetism, collinear magnetism, or non-collinear magnetism.  The computational cost scales linearly with increasing number of atoms per unit cell. Shared memory parallelization achieved using OpenMP directives. When run on a single processor, the new Fortran program runs approximately 8 times faster than the previous version (which was coded in Matlab). When run on multiple processors, the new Fortran program has a parallelization efficiency of approximately 70-85% on 16 processors and approximately 60-70% on 32 processors. As applications, we studied a B-DNA decamer with 733 atoms in the unit cell, the Mn12 acetate single molecule magnet, and diisopropylammonium bromide (DIPAB) molecular crystals. We compute the electric polarization of the DIPAB ferroelectric phase and compare it to the recent experimental and computational results of Fu et al. (Science, Jan. 2013, Vol. 339, pp. 425-428).

Extended Abstract: File Not Uploaded