Developing Ion Parameters Using Shared GPU Accelerator Hardware
Faculty Mentor(s)
Dr. Brent Krueger
Document Type
Poster
Event Date
4-10-2015
Abstract
Molecular dynamics (MD) simulations are used to model the structure and movement of macromolecules. The motion of finite particles is modeled by twice numerically integrating the forces on the atoms such as charge-charge interactions, van der Waals interactions using Lennard-Jones (LJ) potentials, Hookian bond length and angle interactions, and sinusoidal bond torsion interactions. Periodic boundary conditions (PBC) are used to approximate an infinite system of particles even though only a finite number are described. In many MD simulations water is simulated using a model with single van der Waals potential and mass. Additional detail is added through mass-less point charges that simulate the electrostatic properties (ESP) of the water molecule. TIP3P is the most popular simple water model. It includes a charge on the oxygen and the two hydrogens to give it a dipole moment. TIP4P-Ew is a popular water model that is similar to TIP3P but instead of having a charge on the oxygen atom it places a charge where the lone pairs would be. Another style includes TIP5P which has a charge for each lone pair, and a charge for each hydrogen atom. Recently, a new water model, OPC, has been developed that uses the same style of charge distribution as TIP4P-Ew and has results that compare better to experiment than the TIP4P-Ew model. For this new water model to be useful, LJ parameters must be developed for at least a few monovalent ions. This study looks into developing these parameters using MD simulations running on a time sharing computer cluster and GPU accelerator hardware. MD simulations running on entry level GPU hardware run around 1.5 times faster than on high end traditional computing hardware. This has allowed us to produce preliminary results that show first peaks of RDFs for a variety of Na+ and Cl- LJ parameters within the OPC water model.
Recommended Citation
A recommended citation will become available once a downloadable file has been added to this entry.
Comments
This work is supported by the NSF-MRI under grant No. CHE- 1039925 and the NSF-RUI award No. CHE-1058981