[gmx-users] Re: v-rescale - harmonic oscillator

[servaas]

Hi, 

Thanks for your confirmation of this problem. One thing is not clear to
me from your answer. Has v-rescale fundamental problems with a Harmonic
oscillator or is it a problem with the GROMACS implementation?


Kind regards,

Servaas


> Message: 3
> Date: Fri, 8 May 2009 10:12:45 +0200
> From: Berk Hess <gmx3 at hotmail.com>
> Subject: RE: [gmx-users] Re: v-rescale - harmonic oscillator
> To: Discussion list for GROMACS users <gmx-users at gromacs.org>
> Message-ID: <COL113-W10F68330EA113929CBC3518E640 at phx.gbl>
> Content-Type: text/plain; charset="iso-8859-1"
> 
> 
> Hi,
> 
> I have come to the conclusion that the v-rescale thermostat is only correct
> in the limit of large numbers of degrees of freedom, which is equivalent to a small scaling limit.
> The conservation of the conserved energy quantity improves as the system gets bigger.
> 
> But I noticed that in 4.0 the energy conservation is not good.
> It can be improved significantly by changing line 164 or src/mdlib/update.c from
> vv             = lg*(vn + f[n][d]*w_dt);
> to
> vv             = lg*vn + f[n][d]*w_dt;
> 
> There are some other similar lines in case you are PR p-coupling or cosine acceleration.
> 
> I will probably not fix this for 4.0.5, but only for 4.1, so we do not change
> the reproducibility in minor revisions.
> 
> Berk
> 
> 
> > From: servaas.michielssens at student.kuleuven.be
> > To: gmx-users at gromacs.org
> > Date: Thu, 7 May 2009 15:27:37 +0200
> > Subject: [gmx-users] Re: v-rescale - harmonic oscillator
> > 
> > Hi Berk,
> > 
> > Thank you for your concern on this issue. It did not expect it to be a
> > problem from the paper of Bussi et al., but it is of course possible. 
> > I am very curious to hear what they have to say about this. Are you
> > contacting him or will I do it?
> > 
> > For the rotational degrees of freedom would removing them with
> > comm_mode=angular be a problem then because you are using restraints?
> > (you use orire than?)
> > 
> > Thanks again,
> > 
> > Servaas
> > 
> > > Hi,
> > > 
> > > With a diatomic molecule there is the issue that the rotation of the molecule has two degrees of freedom
> > > which are completely uncoupled from the rest of the system.
> > > 
> > > Anyhow, I made a 1D harmonic oscillator using a position restraint and even there the energy
> > > does not seem to be conserved.
> > > I am not sure if a harmonic oscillator is a pathetic case, like with Nose-Hoover,
> > > or if this is an issue with the integrator implementation in Gromacs.
> > > I might need to ask Bussi about this.
> > > 
> > > Berk
> > > 
> > > > From: servaas.michielssens at student.kuleuven.be
> > > > To: gmx-users at gromacs.org
> > > > Date: Thu, 7 May 2009 14:03:30 +0200
> > > > Subject: [gmx-users] v-rescale - harmonic oscillator
> > > > 
> > > > Hi,
> > > > 
> > > > It is just a diatomic molecule without other interactions (so only
> > > > bonded interaction, LJ=0 and charges=0 on the atoms). What other details
> > > > would you like to know? Should I send you the tpr file?
> > > > 
> > > > Servaas
> > > > 
> > > > 
> > > > 
> > > > > 
> > > > > 
> > > > > Hi,
> > > > > 
> > > > > No, only for small values is should be off, maybe tau_t < 10 or 100 * delta_t.
> > > > > 
> > > > > How did you set up your harmonic oscillator?
> > > > > 
> > > > > Berk
> > > > > 
> > > > > > From: servaas.michielssens at student.kuleuven.be
> > > > > > To: gmx-users at gromacs.org
> > > > > > Date: Thu, 7 May 2009 12:54:54 +0200
> > > > > > Subject: [gmx-users] Re:  v-rescale - harmonic oscillator
> > > > > > 
> > > > > > Hi,
> > > > > > 
> > > > > > Thank you for your reply.
> > > > > > 
> > > > > > Yes by effective energy I mean the Gromacs conserved energy term.
> > > > > > 
> > > > > > So you would say the cause of my problem here is that I need a small
> > > > > > tau_t to thermostat this system but for small tau_t values the algorithm
> > > > > > is not 100% ok yet? And what are very small tau_t values? I tested with
> > > > > > quite a big range here.
> > > > > > 
> > > > > > The problem occurred to me when I was experimenting with a hybrid monte
> > > > > > carlo algorithm. I first tested it on a noble gas and compared to MD
> > > > > > simulation with nose-hoover and v-rescale. Those results were 100%
> > > > > > identical. Then I tried it on a dipeptide in vacuum, results were a
> > > > > > little off here (I would expect this for nose-hoover thermostat, but not
> > > > > > for v-rescale). So I decided to go the the extreme simple case of a
> > > > > > harmonic oscillator. Those results were seriously off, and then I
> > > > > > checked the conserved energy term, which seems impossible to keep it
> > > > > > conserved for this system. (of course there can also be an error in my
> > > > > > hybrid monte carlo code, but the fact that the conserved energy is not
> > > > > > conserved is disturbing me here)
> > > > > > 
> > > > > > 
> > > > > > Servaas
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > Hi,
> > > > > > 
> > > > > > What do you mean with effective energy?
> > > > > > The Gromacs conserved energy term?
> > > > > > 
> > > > > > For very small tau_t the current implementation does not work well.
> > > > > > Bussi mailed me a proper implementation that I will put in when I have time.
> > > > > > 
> > > > > > Berk
> > > > > > 
> > > > > > > From: servaas.michielssens at student.kuleuven.be
> > > > > > > To: gmx-users at gromacs.org
> > > > > > > Date: Thu, 7 May 2009 09:02:52 +0200
> > > > > > > Subject: [gmx-users] v-rescale - harmonic oscillator
> > > > > > > 
> > > > > > > Hi,
> > > > > > > 
> > > > > > > I did some experiments with a harmonic oscillator (diatomic molecule
> > > > > > > without charge en LJ parameters) using the v-rescale thermostat. First I
> > > > > > > ran a simulation in the NVE ensemble with a time-step of 0.0001ps, total
> > > > > > > energy was constant here. 
> > > > > > > 
> > > > > > > Then I tried simulating the system with the v-rescale thermostat, same
> > > > > > > time-step of 0.0001ps and tau_t=100,10,1,0.1,0.01,0.001. For none of
> > > > > > > those simulations the effective energy was conserved. 
> > > > > > > (effective energy should be a conserved quantity for NVT simulations
> > > > > > > with v-rescale thermostat, JOURNAL OF CHEMICAL PHYSICS    Volume: 126
> > > > > > > Issue: 1  Article Number: 014101    Published: JAN 7 2007 )
> > > > > > > 
> > > > > > > So what is going wrong here? Would one expect this thermostat to fail
> > > > > > > for such system (e.g. a simple Nose-Hoover is known to fail for the
> > > > > > > harmonic oscillator)? 
> > > > > > > 
> > > > > > > 
> > > > > > > Thanks in advance,
> > > > > > > 
> > > > > > > Servaas
> > > > > > > 
> > > > > > 
> > > > > > _______________________________________________
> > > > > > gmx-users mailing list    gmx-users at gromacs.org
> > > > > > http://www.gromacs.org/mailman/listinfo/gmx-users
> > > > > > Please search the archive at http://www.gromacs.org/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/mailing_lists/users.php
> > > > >