Peculiar behavior with ewald/disp and/or dipole/sphere


I’m interested in measuring the electrostatic potential at the center of a cavity in a Stockmayer fluid. In order to do this, I want to understand the interaction of a charged ion with the homogeneous background charge and its periodic images. I have therefore constructed a very simple test of a single ion in a cubic box. The single-point energy that I get is:

E1 = -0.028647619 (lj units)

If I introduce another ghost atom into the box, by which I mean it carries no charge, dipole or lennard-jones interactions, then the energy that I get is:

E2 = -0.014366134 (lj units)

i.e. essentially half of the original value. I appreciate that dealing with charged systems under PBC can be a bit hairy, but something strikes me as odd here. Incidentally, if I perform the same test in real units (in a 50x50x50 A^3 cell), with atom_style full and the pppm solver, then I get an energy difference of ~0.025 kcal/mol, which is similar in terms of significant figures, but only ~0.2% of the absolute energy. (I therefore tried changing the relative accuracy from 1e-5 to 1e-7 in the original test, but this made little difference.)

Below are my input and data files. You can switch between the two systems simply by changing “1 atoms” to “2 atoms” in the data file. If anyone has any ideas what’s going on then it would be great to hear!


PS I have done this with both the git version, and the latest stable version (7thDec2015)

Maybe it is this: For thermo output with LJ units, the energy values are normalized

by the # of atoms. For real units they are not. See the

thermo_modify doc page on the norm keyword and default for details.


Thanks for the speedy response! This seems to have fixed things. Below are the new energies with thermo_modify norm no used instead. Thanks again for your help.

E1 = -0.028732268E2 = -0.028647619