AMBER Archive (2004)

Subject: RE: AMBER: Turning off non-bond interactions

From: david.evans_at_ulsop.ac.uk
Date: Wed Apr 21 2004 - 11:09:26 CDT

Thanks for taking the time to look into this.
I was using in vacuo simulations intially, sorry for not saying, so the problem isn't with PME specifically.
I'm reluctant to start hacking the core subroutines of sander - as you say I would like a way to turn off non-bonded interactions entirely with some sort of exclusion list, but I don't know if this functionality exists or could be easily added.
I'll try your suggestions of using a tiny radius (and I think epsilon needs to be non zero as well) to prevent atom overlap -- that would probably be good enough for what I'm doing.

Thanks again

Dave

---- Message from Chris Moth <chris.moth_at_vanderbilt.edu> at 2004-04-21 10:58:07 ------
>A bit more information perhaps:
>
>My vdw radii diea of course introduces undesired attractive forces... and I
>now have access to our sander source code...
>
>The non-bonded electrostatic interactions are tightly integrated into
>sander's pme implementation. I have found several 1.0/Rij^2 type terms
>in sander, but, not being a PME guru :), I don't see any quick and dirty way
>to modify the PME implementation to outright ignore atom centers.
>
>In ew_directp.h (included by ew_direct.f) there is this interesting piece of
>code - which I believe causes interactions outside the cutoff to be not
>included in the nonbonded calculations:
>
> if ( delr2 .lt. filter_cut2 )then
> icount = icount + 1
>
>You _might_ get away with modifying this to something like
>
> if ( delr2 .lt. filter_cut2 .and. delr2 .gt. .00001 )then
> icount = icount + 1
>
>but that is risky - given that I don't fully understand the pme
>implementation, and so it might have further unanticipated impacts.
>
>I also looked for a quick way to simply exclude your atoms from all non-bonded
>lists... but I did not find it and am out of time. That would be cleaner, I
>suspect.
>
>Couple of other ideas: You might also try pmemd from Robert Duke instead of
>sander. Perhaps it can better cope with the overlapping zero valued charges.
>
>If explcit solvent is not important, you might try using Generalized Born with
>sander. Then, the pme module is not called at all.
>
>If the problem is localised to just one pair of atoms, you could fudge things
>- and introduce a very small partial charge repulsion prior to the collision.
>
>Sorry to not have a better answer. Hope some of the more experienced
>folks can help you out.
>
>Chris Moth
>chris.moth_at_vanderbilt.edu
>


-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu