[lammps-users] Temperature suddenly went to zero

LAMMPS LAMMPS Mailing List Mirror
1

Hi,

I’ve encountered a strange problem.
I’m tring to simulate some fluid confined in a tub, there are totally 93313 atoms in the system.

Here is the input script:

log.93313.deltat=0.005.t=0.4.eff=1.4.efw=1.4|attachment (465 KB)

2

I would run until just after this happens and print out
thermo more frequently to see what the system does
right before this happens. I would also viz the system
to see what is happening. That should give you some clues.

Steve

3

Hi Steve,

Thanks for you advice~
I’ve solved the problem today.

When I reduced the scale of my simulation, I observed that along z axis there is a shrink of the tube I constructed, which is not obvious in the former larger system I use.

Since I used the “s” boundary condition in z direction , and set a reflect wall in the zlo and zhi. It seems that the two reflect walls were continuously running toward each other according to the shrink of the boundary, and therefore squeezed the tube atoms significantly.

When I changed the boundary conditions in z to “m”, everything went OK.

And I have here in mind two questions:

  1. Provided that I write in the data file " -10 10 zlo zhi" and the max z-coordination of the atoms is 8, where would be my wall if I used the command “fix 1 all wall/reflect zhi”, z=10 or z=8?

  2. “The initial position for each atom is its location at the time the fix command was issued”, this line was in the description of command “fix spring/self”. I’m a little conused, if as addressed in the description, under the setting of “fix 1 wall spring/self 100”, is it possible for the tube atoms to be sqeezed so much and leave their original coordinations far far far …away? May be that’s why I lost more than half the atoms in the simulation before.

4

See answers below.

Steve

Hi Steve,

Thanks for you advice~
I've solved the problem today.

When I reduced the scale of my simulation, I observed that along z axis
there is a shrink of the tube I constructed, which is not obvious in the
former larger system I use.

Since I used the "s" boundary condition in z direction , and set a reflect
wall in the zlo and zhi. It seems that the two reflect walls were
continuously running toward each other according to the shrink of the
boundary, and therefore squeezed the tube atoms significantly.

When I changed the boundary conditions in z to "m", everything went OK.

And I have here in mind two questions:

1. Provided that I write in the data file " -10 10 zlo zhi" and the max
z-coordination of the atoms is 8, where would be my wall if I used the
command "fix 1 all wall/reflect zhi", z=10 or z=8?

If your boundary is style "m" then the wall will be at z=10. If
your boundary is style "s", then the box will get shrink-wrapped
to z=8, and the wall will be at z=8.

2. "The initial position for each atom is its location at the time the fix
command was issued", this line was in the description of command "fix
spring/self". I'm a little conused, if as addressed in the
description, under the setting of "fix 1 wall spring/self 100", is it
possible for the tube atoms to be sqeezed so much and leave their original
coordinations far far far ...away? May be that's why I lost more than half
the atoms in the simulation before.

Don't really know what happened in your bad version of the simulation.
You might need to put some print statements in the fix spring routines
and see what it was doing.