[Non-DoD Source] Re: DPD fix_nve - Verlet - electrostatic (UNCLASSIFIED)


The LAMMPS implementation uses the standard velocity-Verlet integration scheme to integrate the DPD force, where the DPD force is the sum of the conservative, dissipative and random forces. In principle, the standard velocity-Verlet integration scheme could be used, provided that a sufficiently small time step is specified to ensure that the thermostat is working to give the desired temperature. However, there are applications of DPD where this particular integration scheme requires prohibitively small time steps. There are a number of alternative integration schemes that are better suited for DPD, especially in handling the stochastic ODEs (i.e., the dissipative and random force integration). A good review of various integration schemes are provided in this reference: Nikunen et al, Computer Physics Communications, 153 (2003) 407-423. The simplest and most often used is the modified (Groot-Warren) VV integrator. The G-W VV is usually sufficient for the typical weakly repulsive fluid models used for many DPD simulations.

In principle, there is nothing in the DPD approach that doesn't allow point charges. But if the conservative force is soft (like the standard weakly repulsive model), then overlaps can occur due to strong ion pairing. Models to handle electrostatic for these soft particles have been developed (e.g., Groot, R. D., J. Chem. Phys., 118, 11265, 2003; Gao, L., and Fang, W., J. Chem. Phys., 132, 031102, 2010; Gonzalez-Melchor, M.; Mayoral, E.; Velazquez, M. E.; Alejandre, J. J. Chem. Phys., 125, 224107, 2006; Ibergay, C., Malfreyt, P., and Tildesley, D.J., J. Phys. Chem. B, 114, 7274, 2010)

As Steve mentioned, our contributions to LAMMPS will likely be released in the next 6 months or so. In our work (Comput Phys Commun, 185, 2014, 1987-1998), we apply Shardlow's splitting scheme to separate the conservative force integration from the dissipative and random force integration. In our applications of DPD, the accurate integration of the stochastic ODEs allows us to take sufficiently large time steps.

Hope that helps,