[gmx-users] NVE simulations

[David L. Bostick]

Hi Sid,

I believe Dr. van der Spoel was speaking of using ewald methods rather than
a cutoff for the electrostatics.  So instead of just employing a cut-off
for your non-bonded potentials (vdw and coulomb) you can use the following
sort of options:

;Neighbor Searching
nstlist                 = 10
ns_type                 = grid
pbc                     = xyz
rlist                   = 0.9
;
;
;Electrostatics and VDW
coulombtype             = PME
rcoulomb                = 0.9
rvdw                    = 0.9
fourierspacing          = 0.20
pme_order               = 6
ewald_rtol              =1e-5
optimize_fft            =yes

This is what I use for my simulations.  If you are employing pbcs for a
bulk liquid like CCl4 then you are right .. you need pbcs, but PME offers a
drastic improvement for electrostatic treatment.  I would also suggest
starting the simulation at a smaller time step.  If you are using lincs or
shake to constrain bonds and so forth you may be able to work up to a large
timestep, but try to equilibrate the system at a smaller timestep first..
say .002 ps.  Right now, I see from your mdp file that you are using a 10
fs timestep.  This is huge!! .. generally speaking.  Also, the energy will
be conserved better using a smaller time step (an effect of simple
integration like the verlet algorithm) so you can get an idea of
what the total energy should be.  I would say that if you are sure your
initial configuration has the correct total density, you should try
simulating for at least a few hundred picoseconds before moving on to a
larger timestep.  The convergence of the total energy should happen before
you make any other moves.

David

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
David Bostick					Office: 262 Venable Hall
Dept. of Physics and Astronomy			Phone:  (919)962-0165
Program in Molecular and Cellular Biophysics
UNC-Chapel Hill
CB #3255 Phillips Hall				dbostick at physics.unc.edu
Chapel Hill, NC 27599	           		http://www.unc.edu/~dbostick
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

On Mon, 8 Jul 2002, Sidney Elmer wrote:

> I tried starting the NVE simulation from the last snapshot from the
> equilibrated NVT ensemble and I still got the same results (ie, the total
> energy relaxes to it's minimum value).  I just wanted to emphasize that I'm not
> concerned about what the temperature is doing; whether it falls to 0 or goes
> through the roof isn't as much of a concern to me as the Total Energy falling
> to it's minimum value.  It should be a constant energy simulation (and of
> course it will have fluctuations due to the discrete time steps used by the
> integrator), but the total energy is not constant.
>
> I should give some context as to why I want a constant Energy simulation.  I am
> using a model for Carbon Tetrachloride solvent and I am using a rigid model
> using Shake to constrain all distances between carbon-chlorine and
> chlorine-chlorine atoms.  Since everything is rigid, I would like to increase
> the time step.  Since the root-mean-square-deviation of the total energy in a
> microcanonical ensemble is proportional to the time step, I would like to run
> several simulations, using different time steps to see just how large I can
> make the time step before I run into problems.  It's a simple little exercise.
> But since the total energy is not constant when I run without temperature or
> pressure control, I'm concerned if there may be a bug or if I'm missing
> something in my parameter file that will cause the total energy to remain
> constant.  Here is my .mdp file as an attachment.  Thanks for your help on this
> matter.
>
> Sid
>
> PS.  I also received an email for Dr. van der Spoel, that the total energy
> should remain constant without temperature or pressure control and no
> cut-offs.  I tried running with no cutoffs, but grompp gave me errors that I
> could only run without cutoffs if I disables periodic boundary conditions.  I
> need pbc, so is there any other way I can run without cutoffs using pdc?  Maybe
> cutoffs are the problem that the energy is not constant????  I'd like to check
> it out.  Thanks again.
>
> "David L. Bostick" wrote:
>
> > It all depends on that initial configuration and velocities for an NVE
> > simulation.  The total energy can remain the same while the temperature
> > drops.  This is why in the old days people used to scale their velocities
> > each step for some time before actually simulating the system.  You can try
> > equilibrating the system in NVT and then change to NVE with initial
> > coordinates and velocities taken from the last snapshot of the NVT
> > simulation.
> >
> > Generally, if the initial configuration is energetically high, the
> > temperature will go through the roof.  If it is very low, the potential at
> > the starting point will be turned into kinetic energy plus whatever
> > velocities you add in the beginning using the velocity generation option
> > and the temperature can still fall.
> >
> > David
> >
> > -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> > David Bostick                                   Office: 262 Venable Hall
> > Dept. of Physics and Astronomy                  Phone:  (919)962-0165
> > Program in Molecular and Cellular Biophysics
> > UNC-Chapel Hill
> > CB #3255 Phillips Hall                          dbostick at physics.unc.edu
> > Chapel Hill, NC 27599                           http://www.unc.edu/~dbostick
> > =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
> >
> > On Mon, 8 Jul 2002, Sidney Elmer wrote:
> >
> > > Hi,
> > >
> > > I have a question about the pure MD (constant NVE) application.
> > > Whenever I try to run my system at constant Energy, the system always
> > > cools down to zero temperature.  Instead of being a constant energy
> > > simulation, it appears rather that it is a variant of an energy
> > > minimization.  Has anyone else had this happen to them? or is there
> > > something that I am missing in my parameter file?  Thanks.
> > >
> > > Sid
> > >
> > > _______________________________________________
> > > 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.
> > >
>