[gmx-users] Questions about Thermostats

[Dommert Florian]

Hi Tsjerk,

sorry for the strong statement. I should have said:

should be applied ...

instead of

only applicable.

You are right, the question is how big is the difference and actually
one would also expect, that the differences vanish with 1/N. However, so
far it is unknown, what kind of distributions Berendsen does produce and
how this is related to the true canonical ensemble. For this reason, I
would be very catious. In my studies I realized big differences in
density and dynamical properties, if Berendsen instead of PR is used. 

Cheers,

Flo


On Wed, 2012-05-16 at 12:11 +0200, Tsjerk Wassenaar wrote: 
> Hi Florian,
> 
> > Hi, there has been recently a discussion about this topic on this
> > mailing list. Check the archives for the information you look for.
> > However, Berendsen is not producing any kind of known ensemble, and
> > therefore only applicable for equilibration.
> 
> That's a very strong statement, and scientifically unsound. From a
> theoretical statistical-mechanical perspective, the ensemble with the
> Berendsen thermostat does not fall into the known classes. This
> implies that using the Berendsen thermostat yields a physically
> different ensemble. Yet the relevant question is whether the ensemble
> is statistically significantly different, or even practically
> significantly different. A next question is whether the difference
> between two simulations with the Berendsen thermostat and two
> simulations with one of the other ones is consistent or not. And when
> comparing with experiments, are the predictions with the Berendsen
> thermostat different from one of the others?
> 
> This would be a very nice matter of debate if there was a difference
> in performance using different thermostats. However (unfortunately?
> ;)) there is not. So for the same cost there are thermostats that do
> produce the desired ensemble, which makes these preferable over the
> one from Berendsen (also for equilibration). That is not the same as
> stating that it is only applicable for equilibration though.
> 
> For the barostat it's a bit more complicated. Parrinello-Rahman can
> not be used for equilibration, because a large deviation from the
> target pressure may well cause large fluctuations that are unphysical
> and may cause instability. So at the moment the Berendsen barostat
> appears the only _one_ applicable for equilibration, but that doesn't
> (yet) disqualify it for further use.
> 
> Cheers,
> 
> Tsjerk
> 
> 
> -- 
> Tsjerk A. Wassenaar, Ph.D.
> 
> post-doctoral researcher
> Molecular Dynamics Group
> * Groningen Institute for Biomolecular Research and Biotechnology
> * Zernike Institute for Advanced Materials
> University of Groningen
> The Netherlands

-- 
Florian Dommert
Dipl. - Phys.

Institute for Computational Physics
University Stuttgart

Pfaffenwaldring 27
70569 Stuttgart

EMail: dommert at icp.uni-stuttgart.de
Homepage: http://www.icp.uni-stuttgart.de/~icp/Florian_Dommert

Tel.: +49 - (0)711 - 68563613
Fax.: +49 - (0)711 - 68563658
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20120516/ded409b3/attachment.sig>