[lammps-users] "Bond atoms ... missing ... " again

LAMMPS LAMMPS Mailing List Mirror
1

Thank you for the answer, Steve! I have understood the explanation you gave which must be the reason
causing the collapse. However, I still don't understand why a stable system can
become unstable after such a long simulation time. Especially, the system and parameters used
seems not inappropriate.

To my understanding, in a real atomic system the covalent bonding is almost the strongest
interaction, which means the absolution change in the bond distance should be very small.
And it should be fast converged. Even the nonbonded interaction should have converged as
monitored by the intra- and inter- molecular energies. How can "a bad physics" happen in the
middle of a simulation? I can show the thermo output if somebody is interested. It's really like
a comet hitting the earth, a peaceful world destroyed in a second.

Anyway, I will investigate this when I have time. I also noticed somebody reported this kind of
problem happens in a parallel job but not in a serial job. I give my report later. But right now I
urgently need some result, but I was hindered by this problem. Move to another program temporarily :frowning:
I'd also like to know if somebody has met a similar case, the simulation crashed not in the beginning but
after a long simulation time?

Thank you!

lin

Steve Plimpton wrote:

2

for certain rare events your simulation is not long.

your system may just have been trapped in
a meta-stable state. sometimes certain
fluctuations build in a system only after
a while and then push your system into a
different state. i've seen it over an over
again. sometimes after continuing a trajectory
for a bit after enough data from a normal
production run was accumulated and the run
only continued to use up the remaining hours
from a cpu time allocation...

cheers,
   axel.

3

Thanks, Axel!

I can understand the meta-stable state problem. This happens a lot during a protein folding for example.
However, I think this is not the case in my problem. Because I've figured out the crash in my simulation is
just related to several atoms which are interacting with relatively strong local interaction ,bonding, angle and
improper dihedral. Please have a lot of the movie in my last post. These high frequency motions are not expected
to be trapped in meta-stable state. My guess is the problem is caused by some special potential settings.

lin

Axel Kohlmeyer wrote: