Can I restart simulations from dump files?

LAMMPS LAMMPS Mailing List Mirror
1

Dear all lammps users

Simulation running on my desktop suddenly crashed (computer stopped), and I have no restart files.

Is it possible to restart simulation with information in dump files?

I’ve tried to restart simulation with position and velocity information in dumps, but i got warning with “Inconsistent image flags”, and pressure seems to fluctuate so much than previous simulation I runned.

I think this problem is related with round-off, but i desperately want to restart these simulation.

Any advises will be very helpful.

Dong Woo Kang

2

This is not likely a round-off issue, but - as the warning says - due to lack of image flag information. With only coordinates, you will create bond between the wrong periodic copies of atoms. Compare the Molecular energy before and after the crash.

To answer your question: yes, it is possible, but only under certain circumstances.

Axel

3

Dear axel,

So it means that, as an example of a water molecule which one hydrogen atom lie over the periodic boundary, it will cause problem with bonds when I restart simulation with only coordinate information. Right?

If this is right, can I fix this problem when I adjust original coordinates with periodic boundary(e.g. plus or minus box length), so particles are set to be unwrapped.

Dong Woo Kang

2019년 2월 19일 (화) 오전 12:01, Axel Kohlmeyer <[email protected]>님이 작성:

4

Dear axel,

So it means that, as an example of a water molecule which one hydrogen atom lie over the periodic boundary, it will cause problem with bonds when I restart simulation with only coordinate information. Right?

assuming, that your dump file only outputs the x, y, z coordinate and
neither xu, yu, zu, or ix, iy, iz, yes. that would be what is causing
the "inconsistent image flag" message.

If this is right, can I fix this problem when I adjust original coordinates with periodic boundary(e.g. plus or minus box length), so particles are set to be unwrapped.

theoretically, yes.

axel.

5

>
> Dear axel,
>
> So it means that, as an example of a water molecule which one hydrogen atom lie over the periodic boundary, it will cause problem with bonds when I restart simulation with only coordinate information. Right?

assuming, that your dump file only outputs the x, y, z coordinate and
neither xu, yu, zu, or ix, iy, iz, yes. that would be what is causing
the "inconsistent image flag" message.

> If this is right, can I fix this problem when I adjust original coordinates with periodic boundary(e.g. plus or minus box length), so particles are set to be unwrapped.

theoretically, yes.

There was a recent discussion of how to use VMD to fix image flag
problems in LAMMPS data files (which can be generated from dump files
using the "read_dump" and "write_data" commands in lammps).
https://sourceforge.net/p/lammps/mailman/message/36532759/
lammps.sandia.gov/doc/read_dump.html
lammps.sandia.gov/doc/write_data.html

Incidentally, for what it's worth, here is the "dump" command I use to
insure that the ix,iy,iz and mol attributes of every atom are saved:

dump 1 all custom 100000 traj.lammpstrj id mol type x y z ix iy iz

(of course, you will want to modify the dump-ID, "1", or interval
"100000" to suit your needs)
I suggest using a similar dump command when running future simulations.

Andrew

P.S. If your dump file contains the "mol" attribute (molecule-ID
number), you can bypass VMD and write a script which will make sure
that none of the atoms in the same molecule (ie., with the same "mol"
attribute) cross the periodic boundary conditions. If your molecule
is small compared to the periodic boundary conditions, then you can
simply add multiples of the periodic box dimensions to the x,y,z
coordinates of each atom until the distance from that atom to the
center of mass of the other atoms in the molecule is as small as
possible. Do this starting from one atom in the molecule, and keep
adding new atoms to thismolecule until all the atoms in that molecule
are included, updating the center of mass as you go along. (This way
you don't have to worry about which atoms are bonded together which is
a pain.)