VMD-L Mailing List

From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Thu Sep 18 2008 - 22:10:46 CDT

Hi Chris,

from the setup that you've described (particularly the fact that there's
water on all sides of the slab), there's no real reason why your system
*should* keep the slab parallel to the x-y plane. It will be knocked
about by langevin forces and solvent molecules and should begin to
randomly translate and rotate just like any other large molecule. If you
want to maintain its initial location you should probably apply harmonic
restraints to some subset of the atoms in the slab to keep it in one
place. Alternatively, depending on what your slab is made of, if the
initial box size is such that the slab is in direct contact with its
periodic images and thus forms one large infinitely tiled slab, it's
more likely to stay put; this is what is done when people do membrane or
nanopore simulations, for example (if you do this you'll probably want
to have either useConstantArea or useConstantRatio on). You'll need to
think carefully about what system you're trying to physically model, and
apply proper periodic boundary conditions and (if necessary) constraints
to yield either a small slab fixed in space, or a large slab tiled
infinitely through space.

> 6) Also, can anyone explain what the margin value represents and what
> a margin of 5 vs. lets say 2 does to the outcome?
>
Margin will have no effect on the outcome, as noted in the manual. All
it does is affect how the system is spatially decomposed; in general the
system will be divided into patches with side lengths greater than or
equal to pairlistdist plus small padding that is longer than any bonds
to hydrogen atoms. If you set a nonzero margin, the margin is added to
the minimum patch side length.

Because during constant pressure simulations the patches will change
size, if any of the patch edge lengths drop below the minimum set by the
pairlist distance the simulation will need to be restarted. Having a
larger margin allows the cell to shrink further before it crashes;
however, this is not a good practice except during equilibration because
it adversely affects your parallel scaling and because your periodic
cell size shouldn't be changing much except during equilibration (and
even in that case, having your unit cell shrink significantly probably
means your system is undersolvated, and you should consider using
programs such as Grubmuller's solvate to get better initial conditions).

Best,
Peter

> Thank you very much in advance.
>
> Best,
> Chris M Hartshorn
> Washington State University
> cmhartshorn_at_wsu.edu <mailto:cmhartshorn_at_wsu.edu>
>
>