[gmx-users] RE: Coul-LR in PME and in Reaction Field

[Berk Hess]

>From: pascal.baillod at epfl.ch
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: gmx-users at gromacs.org
>Subject: [gmx-users] RE: Coul-LR in PME and in Reaction Field
>Date: Thu, 27 Apr 2006 16:31:48 +0200 (MEST)
>
>Thank you very much, Berk!
>
>1) That only leaves one unanswered point, that seems logical but of which I
>would like to be certain of.. Is there any accessible term in the reaction 
>field
>implementation accounting for long-range electrostatic interaction due to 
>the
>environnement dielectric? Or is this contribution simply added to Coul-SR?
>
>2) Last, I have a situation that I cannot understand, comparing RF and PME 
>for a
>100 residue protein in a water cubic box of roughly 70*70*70A. Comparing
>contributions of energy groups (g_energy), I obtain:
>
>-Reaction field: |coul-SR prot-prot| > |coul-SR prot-other|
>-PME:            |coul-SR prot-prot| < |coul-SR prot-other| (the opposite)
>(with "||" for absolute values)
>
>Would that be because of the long-range electrostatic contribution (with 
>the
>environnement dielectric), added to coul-SR as I suggest in 1)?

You can not simply compare this terms.
Reaction-field is an approximation. It assumes a homogeneous dielectric
medium beyond the cut-off which is clearly not true for your system.
Your protein probably has charged residues which cause significant
error with RF.

Secondly the Coul-SR term does not contain 1/r.
With RF it is 1/r plus the effect of the reaction field force.
This might be what you call "long-range electrostatic contribution",
but it is not really long-range.
With PME is it erfc(beta*r)/r.

Berk.