The initial configuration is defined at the appropriate density and I
equilibrate the system using NVT at 300K for 100 ps and then start
collecting the results for the next 50 ps (with the NVT intact). To be sure
collecting results for only 50ps seems mightily short to me.
you may get half-way decent O-O, O-H, and H-H g(r)s,
and perhaps even Na-O, Na-H, Cl-O, and Cl-H, but for Na-Na,
Cl-Cl, and Na-Cl this would result in very, very bad statistical sampling.
that the g(r) results are not caused by initial closeness of the ions in the
system, i also computed g(r) from the very start. I find that initially g(r)
is zero for Na-Na and Cl-Cl when using 2 or 10 pairs; but with time, peaks
appear for the 10 pair case, and it seems that the peak is progressing from
a larger value of r towards r=0 (which indicated that though the like ions
were apart when the simulation was started, they started moving towards each
other with time). However, the peak seems to stop progressing with time
how many different systems with 2 and 10 randomly placed ion pairs
did you generate? from what you write it almost sounds as if you created
only one of each.
after reaching a fixed r value (close to the sigma of the ion). i tried
using VMD to plot the g(r) but apparently the XYZ dump from lammps needs
some atom names to be set appropriately during the selection process of the
vmd-rdf-plugin.
using .xyz dump output is the worst possible out of all options
supported by LAMMPS as it contains the least amount of information
(and still consumes a lot of space). especially the system cell size
information is missing which makes proper use of PBC for calculation
of correct bulk system g(r)s difficult. outside of that, VMD can only
use information that you hand it. since LAMMPS doesn't us or store
any kind of atom type labels but only numbered atom types, you either
have to use that information in VMD or find other way to reconstitute
this information. the topotools plugin in VMD has a reader module
for data files, that will under certain circumstances allow to pass
atom (and residue) label information to VMD from suitably formatted
comments in the data file. there are also ways to infer atom names
via their mass or use manual VMD scripting to reconstitute the required
information and store it in a some suitable format (e.g. .psf) for later use.
For the second point, could you suggest a suitable technique to compute the
running volume integral, or should that be done by dumping the coordinates
at different time intervals. Further, I checked the interaction energy
the volume integral is best computed from the raw data of the
g(r) calculation before the normalization step. the gofr support
in VMD has this implemented.
values between Cl-Cl, Na-Na and Na-Cl, and find that while Na-Na and Cl-Cl
have finite values, the Na-Cl is zero. This is also confusing since for
dilute systems one would expect oppositely charged ions to come close and
interact (if at all they are not completely hydrated), and not the like
charged ions.
the screening of charges by the water molecules is pretty good,
and thus it is not only a question of whether unlike charged ions
attract each other, but also how strongly like-charged repel each
other. since you have good screening, there is always a good
chance that a single ion pair will come fairly close.
that all being said, there may, of course, also be an oversight
in your input and yet again, without sufficient statistical sampling
you may just be seeing something that is less likely to happen.
the fact that it is less likely to happen, doesn't mean that it never
happens. if that were the case, people would not play the lottery.
cheers,
axel.