[gmx-users] Re: Anomaly in Temperature Behavior

S. Alireza Bagherzadeh s.a.bagherzadeh.h at gmail.com
Tue Feb 5 01:46:39 CET 2013

Dear Andrey,

Today's Topics:
>    4. Re: Anomaly in Temperature Behavior (Andrey Frolov)
> ------------------------------
> Message: 4
> Date: Mon, 4 Feb 2013 00:07:24 -0800 (PST)
> From: Andrey Frolov <andrey.i.frolov at mail.ru>
> Subject: [gmx-users] Re: Anomaly in Temperature Behavior
> To: gmx-users at gromacs.org
> Message-ID: <1359965244629-5005142.post at n6.nabble.com>
> Content-Type: text/plain; charset=us-ascii
> S. Alireza B wrote
> > It appears that my solid phase completely melts after 700 ps. Please see
> > the video below.
> >  http://www.youtube.com/watch?v=6iidpiivVRI&feature=youtu.be
> > Can I conclude that my solid phase is not really far from the equilibrium
> > as it does not suddenly melt/collapse?
> Yes, i would conclude the same looking at this video.
> 1) Do you still have the problem that the temperature keeps increasing (but
> not decreasing) over the 700 ps  simulation ? If this is true, then it
> looks
> strange to me. Then, i would think that the problem could with with the
> poor
> energy conservation in the system.

Yes, the T keeps increasing... :-(

> Such, gmx developers suggest to use PME-switch instead of PME.
> "
> PME-Switch
>     A combination of PME and a switch function for the direct-space part
> (see above). rcoulomb is allowed to be smaller than rlist. This is mainly
> useful constant energy simulations.
> "
> Here, one should note that PME in gromacs is not exactly the same as in
> other MD codes. In gromacs PME there is no cut-off for the real space Ewald
> sum. But one could force it to do the cut-off with the switching function
> on
> a tiny distance interval using "PME-Switch". Whether it improves the energy
> conservation, i do not know.
> In this discussion "http://redmine.gromacs.org/issues/196"  it is stated
> that Gromacs PME "... can conserve energy better in single precision than
> either Desmond, NAMD, Amber or Charmm do in double." Thus, to me it is not
> clear why actually in the online manual it is suggested to use PME-Switch
> for energy conservation.

Not clear to me either!

> Anyway, i never did NVE simulations, so can not be much of help here.

Thanks a lot. You have already been much of help to me.
I will test the PME-Switch and will let you know about the outcome.

> 2) Comment 1: are you aware of "ewald_geometry = 3dc" ?  If you want to
> simulate 2d periodic system (slab) in vacuum then this would be the correct
> setup. If you want a 3d periodic system of infinite number of slabs, which
> interact with each other, then the current setup is correct.

My system does appear to be slab-like. Dimensions of the simulation box are
as the followings:
   3.60900   6.60800  28.64862
However, I would think that regular ewald geometry (3d) is the more general
form of the 3dc and therefore using the 3d option is not incorrect in
It could be more computationally expensive though...

> 3) Comment 2: you start your NVE simulation with zero velocities on the
> molecules of solid phase (and, thus, zero temperature of the solid phase).
> In this setup to melt the solid, one has to 1) heat the solid phase up to
> its melting temperature and 2) melt it. Is it what you want to do ?

Yes indeed. I am trying to achieve this by unfreezing the solid phase and
letting the heat to be transferred to my solid phase which causes its
temperature (atom velocities) to go up and ultimately leads to melting.
Is there any better way to do his?

> Andrey
> --

More information about the gromacs.org_gmx-users mailing list