[gmx-users] Re: Re: Energy conservation of rigid water models

[milton sonoda]

I strongly sugest you to try double precision compilation
for energy conservation in NVE.

Single precision is enough for NPT/NVT due to
convenient algorithms (velocity and volume reescaling)
that keeps the system at the desired termodynamic state
(constante pressure and constante temperature) masking
the truncation errors of the single precision simulation.

In NVE no algorithm is used, so the double precision
is necessary to prevent/postpone energy drift due to
this error acumulation.


2006/2/16, Janne Hirvi <janne.hirvi at joensuu.fi>:
>
> > Have you compiled gromacs in single precision?
> > I think double precision compilation is critical
> > for energy conservation in NVE.
>
> Yes, I have used only single precision compilation because it has been
> enough
> accurate until now (NPT/NVT). Actually I just want to know if employed
> time
> step of 2fs is applicable and so I think that I should use same
> compilation
> also for energy conservation simulation in NVE ensemble. Or may I prove
> suitability of this long time step with double precision compilation even
> I
> have done other simulations with single precision compilation?
>
> Janne
>
>
> > 2006/2/16, Janne Hirvi <janne.hirvi at joensuu.fi>:
> > >
> > > Hello!
> > >
> > > I have read several messages about energy conservation problems and
> > > solutions
> > > but however I am not sure if I have done something stupid. I have used
> > > time
> > > step of 2.0fs for rigid SPC and SPC/E water models in simulations of
> bulk
> > > water
> > > (NPT) and water droplets on frozen polymer surfaces (NVT). Everything
> > > seemed to
> > > be just fine but when I tested energy conservation of bulk model in
> NVE
> > > ensemble I was terrified.
> > >
> > > NVE bulk simulation of 1372 rigid SPC water molecules with 2fs time
> step
> > > (total
> > > time 500ps) resulted in huge drift in total energy. I used same
> parameters
> > > as
> > > for other simulations and even those aren't optimal I expected atleast
> > > something better. Energy conservation seemed to be adequate only when
> I
> > > used as
> > > short time step as 0.5fs.
> > >
> > >
> > > Time step (fs)      Total Energy (kJ/mol)      Drift
> > >
> > > 2.0                 -45596.6                   1235.5
> > > 1.0                 -46147.5                   330.974
> > > 0.5                 -46331.1                   <10 (not yet finished)
> > >
> > >
> > > Parameters:
> > >
> > > comm-mode                = None or Linear
> > > nstlist                  = 5 or 10
> > > rlist                    = 1.2
> > > coulombtype              = PME
> > > rcoulomb                 = 1.2
> > > vdw-type                 = Cut-off
> > > rvdw                     = 1.2
> > > fourierspacing           = 0.12
> > > pme_order                = 4
> > > ewald_rtol               = 1e-05
> > > ewald_geometry           = 3d
> > > tcoupl                   = No
> > > Pcoupl                   = No
> > >
> > >
> > > Bad energy conservation indicates that employed time step of 2.0fs is
> too
> > > long
> > > but however other results from NPT bulk simulations are reasonable.
> > > Moreover I
> > > think that its quite normal to use this long time step for rigid water
> > > molecules. So I am willing to know if this large energy conservation
> > > problems
> > > are acceptable and are they atleast "partially compensated" in NPT or
> NVT
> > > ensemble when temperature coupling prevents heating of the system?
> > >
> > >
> > >
> > > Thanks for any help or comments!
> > >
> > > Janne
> > >
> > >
> > >
> >
>
> ------------------------------------------------------------------------------
> > > Janne Hirvi, MSc(Physical Chemistry), Researcher
> > > University of Joensuu, Department of Chemistry, P.O.Box 111 80101
> Joensuu,
> > > FI
> > > Tel: +358 13 2514544 & +358 50 3474223
> > > E-mail: Janne.Hirvi at joensuu.fi & hirvi at cc.joensuu.fi
> > >
> > >
> >
>
> ------------------------------------------------------------------------------
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> 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/mailing_lists/users.php
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20060216/a342cd0b/attachment.html>