Fix npt problem - uniaxial deformation

Dear all,

I have a problem using one of the latest versions of Lammps (December 2012)
in order to deform uniaxially (at the x-axis) a PE rubber system at the
NPT ensemble (orthogonal simulation box). In particular, I use the following
fix npt command:

fix 3 all npt temp 450.0 450.0 10.0 x -20.0 -20.0 4000.0 y 1.0 1.0 4000.0 z 1.0 1.0 4000.0 couple yz

(all in real units)

The problem is that Lx variable continuously increases without reaching a plateau value.

I ran the same simulation with a previous Lammps version (July 2009) using the
older syntax for the fix npt command:

fix 3 all npt 450.0 450.0 10.0 yz -20.0 -20.0 1.0 1.0 1.0 1.0 4000.0

without any problem (Lx initially increases until it reaches a plateau value).

I think that there is a problem with my parametrs of the fix npt command.
Due to the change of the fix npt command (in 7 April 2010) the default
values of some parameters have changed while new parameters were added. Trying
to make the new fix npt command “backwards compatible” I also used the following
syntax:

fix 3 all npt temp 450.0 450.0 10.0 x -20.0 -20.0 4000.0 y 1.0 1.0 4000.0 z 1.0 1.0 4000.0 couple yz mtk no tchain 1 pchain 0

where I turned off Martyna, Tuckerman, and Klein correction terms, I set tchain=1
in order to use the original Nose-Hoover thermostat and finally I set pchain=0
in order not to use thermostatting of the barostat variables. Unfortunately the
problem remains the same.

Could you please help? Thank you very much.

Stefanos.

I'll let Aidan comment on this. I don't remember the
old syntax precisely, but it appears you are not controlling
the x dimension at all in the old command. So I don't
see why it would change, nor that it would be driven
to adjust to any new value. In the new command,
you are requesting the x-dim pressure be a negative
value, which will drive the box to expand. If it never
reaches that negative pressure, it could expand forever.

Steve

The "yz" in your old simulation is replaced by "couple yz" in the new
version of the fix, which I see you have used correctly. A big issue with
NPT simulations is what happens during the initial equilibration. If your
initial state is far from the target pressure, you may see big swings in
volume, and you may end up in the basin of attraction of the zero-density
gas. Using Nose-Hoover chains helps with this problem. But, you might also
want to increase the time constant. It's definition might have changed
since 2009. You can also try using a drag value.