[gmx-developers] ewald shift

hess at sbc.su.se hess at sbc.su.se
Sat Aug 15 21:18:27 CEST 2009


But that is what I meant.
A plain cut-off lead to serious artifacts,
at least for liquid systems. In water you get very strong dipole-dipole
anti-correlation/correlation at the cut-off radius. This effects
increases (!) with increasing cut-off distance.
My thesis has an example of this.
In liquid systems plain cut-off are almost not used anymore.
Referees will probably reject work done with a plain cut-off.
Reaction-field is better, but even that is mostly replaced by PME.

For something as rigid as a crystal it might work.
Shifted Coulomb is available in Gromacs with the mdp option
coulombtype = shift


> Hi Berk:
> My apologies for posting on the same list.  Regarding the utility of
> the Wolf paper - it does not artificially truncate the Coulomb
> interactions at a cut-off.  It does use a spherical cut-off radius,
> but then it corrects the Coulomb energy by ensuring charge
> neutralization.  Doing so leads to a very simple expression for the
> Coulombic energies (equation 5.10) and forces (equation 5.21).  In
> that way, the Coulombic energy (and thus forces) can be calculated
> exactly using spherical truncation, as is often used in MD
> simulations, with a small cut-off radius.
> I am aware that there are other methods that are used for non-periodic
> systems in slab geometries, but the Wolf method is general, and thus
> not utilize arguments to make the slab non-periodic in certain
> directions, while using Ewald in the other periodic directions.  In
> contrast, the Wolf method can be used both at surfaces and for bulk
> systems.
> Regards,
> Harold
>> --
>> Message: 1
>> Date: Fri, 14 Aug 2009 16:36:05 +0200
>> From: Berk Hess <hess at cbr.su.se>
>> Subject: Re: [gmx-developers] ewald shift question
>> To: Discussion list for GROMACS development
>> 	<gmx-developers at gromacs.org>
>> Message-ID: <4A857655.9090009 at cbr.su.se>
>> Content-Type: text/plain; charset=ISO-8859-1
>> Hi,
>> Please do not post the same question on multiple lists.
>> People might answer the same questions multiple times and the
>> discussion
>> gets confusing.
>> You can simply call shift_LRcorrection if you want, right?
>> Also ewald_LRcorrection does nearly the same thing, but with a
>> different
>> functional shape.
>> Both function are called at nearly the same place in do_force_lowlevel
>> in src/mdlib/force.c.
>> But having looked quickly trough the paper, it would think that you
>> don't need these functions
>> at all.  You want to shift the pair potential for all pairs, or am I
>> wrong?
>> Both functions in Gromacs only operate on excluded pairs.
>> Shifting a pair potential can be done simply with user tables.
>> BTW I don't see the advantage of this electrostatics method.
>> Artificially truncating all Coulomb pair interactions at a cut-off is
>> exactly the same
>> as using a pure cut-off (although this method is probably
>> computationally more efficient).
>> I would consider the artifacts due to this worse than the artificial
>> periodicity introduced by
>> Ewald and PME.
>> For interface systems Gromacs has a correction term for Ewald and PME
>> (ewald_geometry=3dc)
>> which works very well and there are also exact methods in the
>> literature
>> for this geometry.
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-developers-request at gromacs.org.

More information about the gromacs.org_gmx-developers mailing list