[lammps-users] thermo: pressure units question

LAMMPS LAMMPS Mailing List Mirror
1

I am using LAMMPS version lammps-30Nov05.

My objective is to 'thermalize' a system composed soley of TIP3P water
molecules at 300 K and 1 atmosphere. My LAMMPS input file
contains the following relevant lines:

dimension 3
boundary p p p
units real
timestep 2.0
thermo 5000
fix 3 all npt 300.0 300.0 100.0 xyz 1.0 1.0 200.0

The log file for the corresponding portion of the simulation contains
the following lines:

Step Temp E_pair E_mol TotEng Press Volume
   50000 300.26691 -16453.569 2606.2144 -11163.155 765.16096
23163.727
   55000 302.22421 -16345.575 2576.9987 -11066.88 3027.7438
23265.413
   60000 298.2265 -16717.18 2669.2943 -11381.927 -2068.6822
23385.803
   65000 293.63985 -16647.839 2801.2188 -11221.663 354.33068
23387.336
   70000 304.61083 -16512.807 2727.571 -11062.204 2051.8615
23445.992
   75000 296.91324 -16832.354 2689.7354 -11488.399 -2164.4052
23561.782

Because the units are real, my 'fix npt' pressure units are in
atmospheres. I then assumed that the 'thermo' output pressure units
are also atmospheres, If so, the system pressure is fluctating by
thousands of atmospheres, both plus and minus. This seems excessive
to me. (I reduced timestep to 1.0, and obtained (qualitatively) the
same results.)

Am I making an obvious mistake somewhere?

Ken.

2

Hi Ken. This is normal behavior for an MD simulation
of this size and type. I would be worried if you
weren't seeing the huge fluctuations in pressure. This
is a FAQ that should probably be posted in a place
where people searching for it can easily find it. If
you are convinced that you need to reduce those
fluctuations, simulate a bigger system --- the bigger
the system, the smaller the fluctuations. Also, you
can reduce the volume fluctuations and have some
(probably small) reduction of the pressure
fluctuations by increasing the pressure damping
constant, i.e.:

fix 3 all npt 300.0 300.0 100.0 xyz 1.0 1.0 2000.0

instead of:

fix 3 all npt 300.0 300.0 100.0 xyz 1.0 1.0 200.0

Another idea is to add a drag term (see:
http://www.cs.sandia.gov/~sjplimp/lammps/doc/fix_npt.html
)

Paul