[gmx-users] Dispersion correction and cutoffs

[Michael Shirts]

DispCorr is described more in Appendix C, p 181-183

The dispersion correction doesn't affect the structure directly.  At each
step, it corrects the energy and the pressure, thus affecting the density,
thus affecting the structure indirectly.  The point to the dispersion
correction is that with the dispersion correction, the VdW contribution to the
virial will be almost independent of the VdW cutoff, since g(r)=1 for
water/most solvents outside of 1 nm or so.  -> much lower cutoff dependent
results.

Now, it occurs to me that you may also be referring to the solvent-solute
energy changing with VdW cutoff -- in which case this isn't that helpful :(
you could apply the same math described in Appendix C and calculate it
post-facto - It works well for small molecules -- not so much for large or
macro molecules.  I describe this process (roughly) in a JCP paper (Shirts et
al, v. 119, no 9., 1 September 2003).  OR, you could do a mdrun -rerun
afterwards with a much longer rVdW cutoff -- the extra solute-solvent energy
probably wouldn't change the solvent density that much, so the error from that
would be low.

I would NOT advise using abrupt cutoffs.  The problem is, the dispersion
correction is only implemented for abrupt cutoffs.  The code I gave to Erik
Lindahl (last summer, I think) corrects this.  As is, it probably will still
give more correct densities.

Cheers,
Michael

From: "Ilya Chorny" <ichorny at maxwell.compbio.ucsf.edu>
To: <gmx-users at gromacs.org>
Subject: RE: [gmx-users] Re: cut offs
Date: Wed, 7 Jan 2004 15:24:10 -0800
Reply-To: gmx-users at gromacs.org

I looked at the code and Dispcorr seems to be a after the fact
correction to the energy. It doesn't change the dynamics or in my case
the resulting Coulomb structure.

Ilya



Cheers,
Michael