[gmx-users] Re: Energy is not conserved

[Vishal Vaidyanathan]

Hi,
  We ran some simulations of a water box (the one provided with gromacs)
in the NVE ensemble (no temperature or pressure coupling) and found the
energy to fluctuate wildly.

  After discussions with Erik Lindahl, the problem was quite simply solved
by correcting grompp for the rlist > rvdw and rlist > rcoulomb errors. As
the manual says rlist SHOULD be 2 to 3 angstroms larger than rcoulomb or
rvdw when using coulombtype = shift or coulombtype = cut-off. This solved
our energy conservation problems.

  I think nstlist = 1 is overkill. Try playing with rlist/rcoulomb values
first. nstlist = 1 simply makes rlist superfluous.

  Hope it helps,
								Vishal


> Date: Wed, 04 Dec 2002 20:03:16 +0100
> From: Ram'on Garc'ia Fern'andez <ramon at jl1.quim.ucm.es>
> To: gmx-users at gromacs.org
> Subject: [gmx-users] Bad energy conservation.
> Reply-To: gmx-users at gromacs.org

> I am having problems with energy not being conserved with Gromacs. I
> tried
> reducing the timestep with no sucess.

> My simulation box contains 432 diatomic molecules. In reduced units,
> bond
> length is 1 and temperature is 2. My reduced time is about 10^(-4), so
> it is a very small timestep. There should be no overlaps, since the
> original
> structure is a crystal (and the initial energy is negative).

> I tried the example simulation of water included in the tutor directory
> I removed both the termostat and the barostat, and changed both the
> vdw type and coulombic type to shifted. Gromacs manual says than with
> the potential type is shifted, rlist should be higher than rvdw and
> rcoulombic,
> but grompp emits an error if rlist is higher than rvdw or rcoulombic.
> Anyway,
> I am not sure if I understand exactly what is rlist.

> The energy is not at all conserved in this simulation. Values of total
> energy are:

> step  Total energy
> 0     -7.48040e+03
> 50    -7.44887e+03
> 100   -7.41691e+03
> 150   -7.39175e+03
> 200   -7.36756e+03
> 250   -7.34220e+03
> 300   -7.30953e+03
> [...]
> 5000  -3.78888e+03
>
> Energy is not quite correct.
>
> Could anyone show any simulation with Gromacs where energy is conserved?
>
> Ramon
>
> --__--__--