[gmx-users] diverging temperature with pressure coupling

Mon Apr 18 13:41:42 CEST 2011

On Mon, 2011-04-18 at 11:42 +0200, David van der Spoel wrote: 
> On 2011-04-18 11.22, Dommert Florian wrote:
> > Hello,
> >
> > oh this nasty ILs ;) I am currently investigating a similar problem,
> > that deals with PR-coupling and my impression is that Leap-Frog is the
> > problem and not PR. Currently I am running simulations with md-vv and
> > MTTK to verify my ideas. On the other hand I am using two tc_groups for
> > my ILs due their difference in size. Unfortunately you have not written
> > which IL you deal with, but thinking about long chain ILs like
> > [BMIM][Cl], or [BMIM][PF6], the degrees of freedom for the different
> > molecules differ strongly. Here the big question is why does the
> > barostat influence the thermostat. It seems one has to be very very
> > careful, when choosing the coupling parameter for PR. I did a bunch of
> > simulations just varying the coupling parameter and obtained large
> > differences in the dynamics of my system, while static properties like
> > RDF, density of mass, and pressure look quite fine. For example if I
> > calculate the conductivity including correlation effects the values
> > differ from 2 S/m up to 12 S/m. But so far I also have no idea why ???
> > I am really happy when I finally got my results with vv and if you are
> > interested I can report back to the list or off-list, just let me know.
> >
> > On the other hand I do not know if the force field could be a problem.
> > PR changes the box size according to a Lagrangian equation of motion in
> > contrast to Berendsen, which just does size rescaling. I do not know how
> > sensible this equations of motions are in respect to the motion of the
> > atoms. Perhaps one can try to perform simulation with a higher accuracy
> > for the electrostatic interactions. I realized that the standard PME
> > settings are quite poor for ILs. Perhaps the Berendsen pressure coupling
> > can deal with this, because it forces the box to scale. PR follows its
> > equations of motions and PERHAPS introduces artefacts if the integration
> > of atomic trajectories is too inaccurate.
> >
> > So my PME settings are like this:
> >
> > rcut=1.3 ( here the force field tells you the number )
> > pme_order=6
> > fourierspacing=0.05
> > rtol=1e-8
> >
> > This are just approximate values, exact numbers can be calculated by
> > g_pme_error for your system.
> >
> Florian, these PME values are scary! Are you running in double precision 
> too, otherwise I guess it will be difficult to get the errors as low as 
> you want them.

Hello David,

these settings yield an accuracy of 10^-4 kJ/(mol nm), which should be
captured by a single precision calculation, if I take into account that
I have forces around 100kJ/(mol nm). Am I wrong ?

> As regards the Temperature diversion (no pun intended), this is a known 
> issue also with Berendsen T,P coupling. The reason for this is lack of 
> interaction between the different subsystems. It is interesting to hear 
> the PR makes it worse though, because of the better ensemble properties.
> 
> If this also has an effect on dynamic properties all of us are in even 
> more trouble. I recently had problems reproducing dielectric constants. 
> I will now  redo my simulation with Berendsen P-coupling instead of PR, 
> even though grompp tells me not to!
> 

So I assume the rerun with Berendsen will give you faster dynamics,
which I had when using Berendsen. However as you mentioned you get into
problems for creating a correct canonical ensemble.

Cheers,

Flo 

> > Cheers,
> >
> > Flo
> >
> >
> > On Sun, 2011-04-17 at 03:10 -0400, Roland Schulz wrote:
> >> Forwarding this email from my group colleague:
> >>
> >>
> >> Dear Gromacs users,
> >>
> >>
> >>
> >> I am trying to simulate a cellulose fiber in an ionic liquid solution
> >> in the NPT ensemble.  During the simulation, the entire system is
> >> coupled to a thermostat.  Yet, I observe an inhomogeneous temperature
> >> distribution throughout my system (hot-solvent/cold-solute) when I use
> >> Parrinello-Rahman pressure coupling but NOT when I employ Berendsen
> >> pressure coupling.  I have tested velocity-rescaling and the
> >> Nose-Hoover scheme to keep the temperature constant and in both cases
> >> Parrinello-Rahman pressure coupling seems to cause the solute’s
> >> temperature to become significantly lower than the solvent’s (to
> >> decompose temperatures, I am using “mdrun -rerun” with a run input
> >> that defines tc_grps separately).
> >>
> >>
> >>
> >> I was wondering whether there were any known algorithmic reasons for
> >> this unphysical temperature gradient when using Parrinello-Rahman
> >> pressure coupling.
> >>
> >> Thank you.
> >>
> >> Barmak
> >>
> >>
> >> Comment from me: The effect is large. The ionic liquid is 5 degrees
> >> higher and the cellulose is 50 degrees lower (after 50ps, after that
> >> it stays constant). With Berendsen pressure both parts fluctuate
> >> around the same target temperature (as one would expect). Any reason
> >> why one doesn't get the correct temperature with rerun? Or is their a
> >> better way to get the temperature for different groups(for a
> >> simulation with just one tc-group)? Any reason why Parrinello-Rahman
> >> pressure coupling would have this effect on the temperature?
> >>
> >>
> >> Roland
> >>
> >>
> >> --
> >> ORNL/UT Center for Molecular Biophysics cmb.ornl.gov
> >> 865-241-1537, ORNL PO BOX 2008 MS6309
> >> --
> >> gmx-users mailing list    gmx-users at gromacs.org
> >> http://lists.gromacs.org/mailman/listinfo/gmx-users
> >> Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> >> Please don't post (un)subscribe requests to the list. Use the
> >> www interface or send it to gmx-users-request at gromacs.org.
> >> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> >
> 
> 
> -- 
> David van der Spoel, Ph.D., Professor of Biology
> Dept. of Cell & Molec. Biol., Uppsala University.
> Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
> spoel at xray.bmc.uu.se    http://folding.bmc.uu.se

-- 
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/20110418/8c5b089d/attachment.sig>