bond (angle, dihedral) atom %d %d missing on proc #

Dear lammps users:

I wonder if any of you who does polymer stuff has come across such problem: The MD runs perfectly okay with temperature, energy and pressure looking stable, then at a random step the simulation gets terminated with errors saying: bond (angle, dihedral) atoms %d %d missing on proc #. However, if you save restart file periodically and restart the calculation from the restart file right before the break point with reassigned thermal velocity, you can pass the break point safely.

What is the cause of such problem? Any solution to solve it? Thanks a lot.

Shaorui

This error occurs in parallel, b/c atoms move too far between
reneighboring. If you ask the code to reneighbor every step (if needed)
and it still happens, then you are running bad dynamics. So the problem
is your model. Fix it and the problem wiill go away.

Steve

Hi Steve:

Is “neigh_modify delay 0 every 1 check yes” the safest setting?

Shaorui

yes - that insures neighbor lists are built as soon
as necessary, but not more often than necessary.

However, it won't help if your timestep is too
big or atoms still move long distances in a single
step (e.g. due to overlaps).

Steve