Error : Lost Atoms while working in parallel but not in serial

Hi,

I am using an older version of LAMMPS (2007) because it has some modifications to pairstyles and diagnostics. I simulate shear flow of soft particles. My script runs fine in serial mode but I lose atoms in MPI or parallel mode. I have periodic boundary conditions. Is there something that i have to take care of when running in parallel ?

Thanks,
Lavanya

If an atom moves further than a processor's domain size
before reneighboring you will "lose it". In other words
you are doing bad dynamics. Try reneighboring more
often or visualizing your system or monitoring the thermo
output on a short timescale to diagnose and fix your
problem.

Steve

Thank you.I really appreciate your quick reply.

I have another question about the box bounds. It is a random packing and I know the box in which the centres of particles lie (say lowedge and highedge). So for the the data file in lammps, my bounds should be the lowedge + Radius and highedge+radius…Am I right ?

Thanks,
Lavanya

In the current version of LAMMPS, atoms in the data file
that are outside a periodic box are wrapped back into the
box. Don't know if that holds for the 2007 version, but
probably.

Steve