missing atoms in dump file

Hi,
i am dealing with a dramatic problem. I made several
simulations, where i checked the output log file only, which always
looked quite normal. now i was looking at the dump files and saw that
there where only 259200 atoms of type 1 (Si) written to the
dump file. but the output file says

Hi,
i am dealing with a dramatic problem. I made several
simulations, where i checked the output log file only, which always
looked quite normal. now i was looking at the dump files and saw that
there where only 259200 atoms of type 1 (Si) written to the
dump file. but the output file says

--
region 1 block INF INF INF INF INF 1.25
group lower region 1
12800 atoms in group lower
group ind region indenter
146433 atoms in group ind
group mobile subtract all lower ind
310400 atoms in group mobile
--

the dumpfile is written by command
---
dump 1 all atom 1000 wiggle_A40_T3.5nvt300k.dump
---
how could that happen?

hard to say without knowing what happened
between the two sets of commands.

axel.

I rerun the same simulation on one and 8 (1 node) processors and
everything seems to be ok. So it fails for # processors > 1 node on this
system. So its parallelization issue.
Jörg

I rerun the same simulation on one and 8 (1 node) processors and
everything seems to be ok. So it fails for # processors > 1 node on this
system. So its parallelization issue.

no. it also can be a "crappy input issue".
quite a few GI-GO problems are only revealed
when running in parallel.

regardless, since you are not providing enough
information to back up your claims, nobody will
be able to help you.

axel.

here is my input file:

thermo_modify lost ignore

Try removing that line, and see if you are losing
atoms before you dump them out.

Steve

thermo_modify lost ignore

Try removing that line, and see if you are losing
atoms before you dump them out.

he does. lots of them.
the problem is in using shrinkwrap boundary
conditions while defining the box in such a way
that there is a large vacuum area in z-direction.
so on the first re-neighboring, LAMMPS "loses"
the atoms because it can only communicate
atoms up to a certain distance.

the solution is to either define the simulation
box more carefully to match the initial shrinkwrap
box dimensions or use "m" boundary conditions
in z-direction. so the box doesn't shrink below
the initial dimensions.

axel.