[lammps-users] NPT and NVT integration and neighbor lists

LAMMPS LAMMPS Mailing List Mirror
1

Hi all,
I’ve got two related problems with damping parameters in Nose-Hoover integration in the canonical and isothermal-isobaric ensembles, so I’m going to lay both of them out. Sorry about the doubly long text here! I’m using the 8 Oct 2010 version of LAMMPS in serial.

First, when I run with an NVT or NPT ensemble and Tdamp < timestep (but not for Tdamp >= timestep), I get bad temperature values (as with thermo output and script below using the attached alpha-quartz data file). This seems to be independent of pair style, although I haven’t tried turning coulombic interaction off. Has anyone encountered this before? Is this a known limitation of Nose-Hoover integration in LAMMPS, or a bug, or do I just need to use unusual neighbor and neigh_modify settings to make this possible? I didn’t find any clues in a quick look at the discussion of Verlet integration with NH equations of motion in the Tuckerman paper cited in the doc/fix_nh page.
Script:

aq.data (82.9 KB)

2

tyler,

Hi all,
I've got two related problems with damping parameters in Nose-Hoover
integration in the canonical and isothermal-isobaric ensembles, so I'm going
to lay both of them out. Sorry about the doubly long text here! I'm using
the 8 Oct 2010 version of LAMMPS in serial.

First, when I run with an NVT or NPT ensemble and Tdamp < timestep (but not

why would you even _want_ to do that?

you could just as well use temp/rescale.
the whole point of nh-thermostats is to not
interfere with the dynamical properies of
the system much and ideally only provide
a very weakly coupled heat bath.
if you need to move energy in or out of
your system that drastically, you are much
better off with berendsen, langevin, or even
temp/rescale.

cheers,
     axel.

3

Hi Axel,
Thanks, that’s a good point. I was unhappy with the ~30K temperature fluctuations that I was getting in my thousand-atom system with tdamp > timestep, but I suppose up-sizing would help there, and if not, I’ll turn to other thermostats.

Any thoughts on the Pdamp problem? I get immense pressure fluctuations with reasonable values of Pdamp, but maybe again that’s because of the size of my system.

Also, thanks for making topotools and imd, they’re great!

Tyler

4

You can't use damping constants near one timestep
in value with Nose/Hoover - they will not work. You
need a damp of about 100 timesteps for temperature
and about 1000 for pressure.

Steve