[gmx-users] reference on free energy with soft core electrostaticsand PME?

[Berk Hess]

>From: David Mobley <dmobley at gmail.com>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>CC: gmx3 at hotmail.com
>Subject: [gmx-users] reference on free energy with soft core 
>electrostaticsand PME?
>Date: Fri, 21 Oct 2005 12:09:58 -0700
>
>Hi,
>
>I need some information about how Gromacs handles PME for free energy
>calculations when sc-alpha is nonzero and the perturbation is 
>electrostatic.
>That is, how is PME handled when soft core potentials are used to turn off
>electrostatics for a molecule in solution, for example? For what it's 
>worth,
>I'm using the CVS version of Gromacs 3.2 referenced here (
>http://www.gromacs.org/pipermail/gmx-developers/2004-December/000982.html),
>but I assume things are handled the same in Gromacs 3.3, which I'll be
>switching to shortly.
>
>What I'd like to see is a derivation or description of a derivation for 
>what
>it means to correctly deal with PME for this case. The recent J. Comp. 
>Chem.
>paper on Gromacs (v. 26, n. 16) states that "Differences in constraints or
>in long-range interactions are
>properly handled," but doesn't explain what it means to handle these
>properly in the soft core case. The papers referenced in that section (one
>on Gromacs 3.0, and van Gunsteren's paper introducing the soft core
>potential) are similarly unhelpful.
>
>In other words, I know the claim is that Gromacs now properly handles
>electrostatics when a soft-core potential is used to turn off 
>electrostatics
>in a free energy calculation. But what does it mean to "properly handle"
>this? How exactly are they handled? Again, I'd like to see a derivation (or
>reference to one) or at the very least a description of a derivation so I
>can work through it myself.

No derivation is required.
Gromacs calculates the soft core energy according to the formula given
in the manual.
Another matter is how this is done in practice with PME.
This goes as follows:
We determine the PME reciprocal space energy twice, once with the A and
once with the B state charges. These energies are interpolated with lambda.
Then for the particle-particle part we subtract the reciprocal space energy 
and
add the normal plain Coulomb soft-core potential.
The only assumption made here is that the soft-core effect is negligible
at the cut-off, which is always the case as it decays as r^-6.

Berk.