[gmx-users] hints for core/shell optimization?

[Justin Lemkul]

On 7/25/14, 11:22 AM, Tamas Karpati wrote:
> Dear Justin,
>
>> Does your topology specify the proper intramolecular exclusions?  What
>> is(are) the molecule(s)?
>
> No bonds, no exclusions. The whole crystal is modelled by ions
> interacting via forces of the Coulomb and Buckingham types.
> In fact, there is an X-Y-X angle force type which does have an
> effect on forces and energy. Shell particles, on the other hand,
> are defined via the [polarization] section as quasi-bonds. That's all.
>

Coulombic interactions fail at short distances; you probably need to apply Thole 
screening to avoid polarization catastrophe.  Ions are particularly problematic 
in this regard.

>> Try nstlist = 1.  The shell positions are solved via SCF (EM), so you need
>> to update the neighbor list very frequently.
>
> Thanks for the trick. Tried but to no avail (very same results).
> Although shells should have their weight in the model, I expect
> them not to frequently change partner -each is ultimately connected
> to its core atomic site.
> In addition, there are no more bonds in the system i, thus, don't
> see the point in regenerating pairs. Is there a process -I should
> get to know about- which autogenerates bonds?
>

If your model doesn't use them, then there's nothing to be done here.  I was 
asking about bonded structure and exclusions and such because of the Thole issue 
I noted above.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================