[lammps-users] Error: Lost atoms

Hello,

I have a 2300 LJ particle simulation box that is roughly 10 by 10 by 22. My box is periodic in x and y, but not z. I occasionally run into the Lost atoms: original .15g current .15g error in LAMMPS. My current strategy of “dealing” with this is to write restart files, but is there something else can I do to alleviate this? Is there something wrong with my code or computer? I have looked through the past sourceforge entries (not all of them) and couldn’t find anything.

I hope some of my following code will help, however, I’m not sure what would be useful in solving this problem of mine. I’ll respond with any additional information that may be relevant.

pair_style hybrid/overlay lj/cut/opt 2.5 lj/cut/coul/cut 2.5 9.5

neighbor 0.5 nsq (I needed to do this after changing my z dimension from 20 to 22 in length, not sure if this was the best thing to do)
neigh_modify every 1 delay 10 check yes page 500000 one 10000

fix wall1 wall1 setforce 0.0 0.0 0.0 # Fix walls in place
fix wall2 wall 2 setforce 0.0 0.0 0.0 # Fix walls in place
fix thermostat liquid langevin 1.0 1.0 75.0 3
fix timeintegration liquid nve

Thanks,

Brian

I should also say that I can get as high as 430000 time steps (with the default dt for LJ units) before I get this error.

You can always set thermo_modify lost ignore, and LAMMPS will
continue even if you lose atoms. But the real question is why
are you losing atoms. Its usually b/c something is bad in your
model and one or more atoms is being blown out of the box
in one or a few timesteps and LAMMPS can't track it. If you
monitor thermo and dump output and dump lots of snapshots near when it
is happening and viz them, you should be able to figure out
what is going wrong.

Steve

I’m fairly sure I fixed the problem by decreasing my timestep to a lower value. With too high a dt, I think particles started to overlap with my wall particles, shooting them out of the box.

Thanks for the help!