pressure/temperature blowing

LAMMPS LAMMPS Mailing List Mirror
1

Hi All

I was doing NPT MD simulations and i observed that at some point temperature and pressure start blowing up and follow an upward curve (into thousands) with time, whereas there were another values fluctuating around expected means. There is a less probability that this is due to systemic issues. I have observed this kind of behavior earlier too. Can anyone explain please? I specifically wanted to know if this is any thermostat/barostat failure.

Thanks a lot!

regards
Chetan

2

Hi All
I was doing NPT MD simulations and i observed that at some point temperature
and pressure start blowing up and follow an upward curve (into thousands)
with time, whereas there were another values fluctuating around expected
means. There is a less probability that this is due to systemic issues. I

actually, i would exactly think this.

have observed this kind of behavior earlier too. Can anyone explain please?
I specifically wanted to know if this is any thermostat/barostat failure.

_very_ unlikely. without knowing any details about the siumelation
it is impossible to give a qualified answer.

axel

3

Thanks again, Axel. The reason i think it is not systemic is since i did 5 ns of equilibration run which ran perfectly ok (with respect to temperature and pressure etc)…then production run calculating diffusivities started. after some time was elapsed (say 0.4 ns), temperature and pressure started blowing up (both of which were perfectly ok for previous 5.4 ns). I have submitted the same run again. If it is systemic and it should show up again.

4

Thanks again, Axel. The reason i think it is not systemic is since i did 5
ns of equilibration run which ran perfectly ok (with respect to temperature
and pressure etc)..then production run calculating diffusivities started.
after some time was elapsed (say 0.4 ns), temperature and pressure started
blowing up (both of which were perfectly ok for previous 5.4 ns). I have

that doesn't mean anything. again, without know any details of the system,
it is impossible to give any specific recommendations. the first thing you
have to do is to visualize the trajectory and look for any specific events
or structural changes. without knowing anything, it could be caused by
something as trivial as a typo in the input. or if you have a marginally too
large time step, it may take that long to have a large enough fluctuation
that brings two atoms close enough together to make them behave
unphysical and blow up the calculations. there are many possibilities.

submitted the same run again. If it is systemic and it should show up again.

that is the worst thing to do. what if it keep going by chance and
you have an error in your simulation and find it only during final
analysis or after you published your results?

axel

5

Thanks again, Axel. I get your point and i will see how it can be tracked down. One more thing is that there is another similar job which ran perfectly and gave expected answers. The difference between that “perfect” job and this “not-working” job is just the difference in annealing.

Thanks