[gmx-users] Parrinello-Rahman coupling and constant slight evolution of box sides

[Jean-Paul Becker]

Hello gmx users,

I am working with gmx version 3.3.1 and
i am using the ffG43A2x force field.
I am currently running 4 simulations :
- In 2 simulations the system is a POPE bilayer
in water (454 pope with 15800 spc water molecules)
- In the 2 other simulations a protein is embedded in
the bilayer (something like 1 protein, 360 POPE,
42000 spc water molecules....)

The method of simulation for the 4 systems is as follow  :

1)careful minimization of  the system energy
2)5 ns NPT with Berendsen weak coupling (1.0 bar, 310 K)
The pressure coupling is semi-isotropic.
3)NPT with Nose-Hoover and Parrinello-Rahman couplings
(1.0 bar, 310 K)
The pressure coupling is still semi-isotropic.
I have attached the ".mdp" file of the bilayer simulations.
I use quite the same one in the protein simulation (only
temperature coupling groups are different).

The bilayer simulations are now running for ~35 ns and the protein
simulations for 58 ns with N-H and P-R couplings.
The mean values of pressure and temperature correspond to the ones
expected.
But i have problem with the lengths of the box sides :
- After 15 ns of simulations (that is to say between 15 and
35 ns for the POPE simulations and 15 to 58 ns for the protein
simulations with N-H and P-R couplings) if you consider a small
part of the trajectories (5 to 10 ns in length) the values
of the box sides (Box-X, Box-Y and Box-Z) seem to have converged
to more or less constant values.
But, if you look to the whole periode, (15 ns to 35 ns
or 15 ns to 58 ns) you realize that Box-X and Box-Y are slightly
decreasing in a continuous way while Box-Z is increasing.
(The values of Box-X and Box-Y are around 10 nm and the decrease
is of 0.1 nm to 0.2 nm depending on the system)

This results in a slow but constant decrease in the area per lipid
just like the one which can be observed in the paper :
Biophysical Journal march 2004 vol. 86 p 1601-1609 on figure 2.

I have to say that the 454 pope bilayer has been simulated
for several nanoseconds with isotropic Berendsen weak coupling
before its use in this project. Of course, in this case, the area per
lipid was almost constant.

This decrease looks like a real tendancy and i have been simulating
my systems for quite a little time :
So,
- do i have to wait a little more for the area per lipid to tend to converge ? 
I am afraid it won't happen before a very long time.
- do i have to change something in the method ? : is these
slight decreases of Box-X and Box-Y and increase of Box-Z
something frequent with semi-isotropic P-R weak pressure-coupling ?
- Is his problem coming from the lipid parameters ? I have red several
papers looking at this point.

I want to thank you for the time you take dealing with my problem
and i would really appreciate your advises.

J-Paul.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pope_mdout.mdp
Type: application/octet-stream
Size: 9034 bytes
Desc: not available
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20080117/5448dd0e/attachment.obj>