[gmx-users] A method to scale Coulombic 1-4 interactions seperately

[Berk Hess]

>From: chris.neale at utoronto.ca
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: gmx-users at gromacs.org
>Subject: [gmx-users] A method to scale Coulombic 1-4 interactions 
>seperately
>Date: Tue, 12 Sep 2006 23:08:45 -0400
>
>There is a recent publication about combining the Berger lipids with 
>OPLS-AA and
>getting the correct 1-4 scaling factor. New dihedrals were developed 
>because it
>is possible to set special 1-4 parameters for LJ, but not coulomb. However,
>there is another workaround to this that should be generally applicable to 
>1-4
>coulombic scaling.
>
>The lipids.itp file has already defined [ pairtypes ] so the LJ 1-4 is 
>scaled
>correctly (FudgeLJ will not be applied to the values in [ pairtypes ]). The
>problem now is that OPLS-AA requires a FudgeQQ of 0.5 and the Berger lipids
>desire a FudgeQQ of one.
>
>The solution comes in two parts:
>1. Divide the epsilon entries in the [ pairtypes ] of lipid.itp by 2.0. Now 
>the
>LJ and coulombic 1-4 interactions are both exactly half of what they should 
>be.
>2. Every entry in the [ pairs ] section of pope.itp should appear twice. 
>Now the
>LJ and coulombic 1-4 interactions are both exactly what they should be.
>
>This solution should work for an arbitrary 1-4 scaling problem, just make 
>sure
>that you get the ratios correct. A detailed method is outlined below and 
>that is
>followed by the method that I used to test this procedure. I have also run 
>a
>0.5ns simulation and the system "looks" fine (when I say looks fine I mean 
>by
>visual inspection or by g_density). 0.5ns is not nearly long enough to 
>properly
>evaluate area per lipid, but for what it's worth I get 0.53nm^2 per lipid 
>with
>regular combination rules and 0.52nm^2 per lipid with the
>half-epsilon-double-pair method outlined here.
>
>Perhaps the Gromacs crew could tell us whether a double [ pairs ] entry 
>might
>run into problems in any special situations?

No, there are no problems at all.
A slightly cleaner method would be to leave the LJ pair parameters in
the force field section unchanged and add the double pairs
with additional 0 0 parameters for the LJ.

Berk.