Hello,
I’m trying to replicate the results from a simulation I ran on an older LAMMPS version (21 May 2008) with a newer LAMMPS version (9 Mar 2011). I’m attaching three LAMMPS input files. The first input file is labelled input_for_old_LAMMPS. This is the input file I’m trying to replicate. The other two input files are labelled input_new_LAMMPS_1 and input_new_LAMMPS_2. They are my two attempts at replicating the simulation run on the old LAMMPS version. All three simulations are reading the same input configurational file (input.conf). The two simulations on the new LAMMPS blow up mid-run (the x-dimension gets huge) and do not do a good job maintaining the set system pressure.
I was wondering why this is the case. Is it possible that the old LAMMPS version had a bug in it, and actually the correct result is the one I am obtaining with the newer LAMMPS version? Or perhaps a new bug was generated in the newer version related to pressure coupling. I should mention that the results from the older LAMMPS version agree with physical results from laboratory experiments. So we are inclined to think that the older LAMMPS version did not have a bug in this issue, and I am either not doing a good enough job replicating the input file, or there might be a bug in the newer version.
If you would like to see the input.conf file as well, please let me know and I’d be happy to send it over. I could also send over the LOG files of the three simulations if need be.
Thanks a lot for your help,
Amit Shavit
input_new_LAMMPS_2 (526 Bytes)
input_new_LAMMPS_1 (536 Bytes)
input_for_old_LAMMPS (498 Bytes)
Hello,
I'm trying to replicate the results from a simulation I ran on an older
LAMMPS version (21 May 2008) with a newer LAMMPS version (9 Mar 2011). I'm
attaching three LAMMPS input files. The first input file is labelled
input_for_old_LAMMPS. This is the input file I'm trying to replicate. The
other two input files are labelled input_new_LAMMPS_1 and
input_new_LAMMPS_2. They are my two attempts at replicating the simulation
run on the old LAMMPS version. All three simulations are reading the *same*
input configurational file (input.conf). The two simulations on the new
LAMMPS blow up mid-run (the x-dimension gets huge) and do not do a good job
maintaining the set system pressure.
i don't think that this is a bug, but you have to keep in mind
that there is one important difference between the two lammps
versions that you are quoting: the nose-hoover thermostat
infrastructure has been completely rewritten to use nose-hoover
chains instead of a single chain by default.
you can get the old behavior with adding these flags to fix npt
mtk no pchain 0 tchain 1
I was wondering why this is the case. Is it possible that the old LAMMPS
version had a bug in it, and actually the correct result is the one I am
obtaining with the newer LAMMPS version? Or perhaps a new bug was generated
in the newer version related to pressure coupling. I should mention that
it is hard to tell with having experimented with input parameters for a bit,
but it looks like you simply got lucky without N-H chains. the chains are
the proper way to go. a single NH thermostat is not ergodic.
but the chains don't handle large fluctuations and aggressive
time steps as well as the single NH thermostat. sometimes
playing with the drag flag helps. but it is more likely that you
have borderline stable dynamics and switching to the new
code base pushed you over the border. 
it may still be worth updating to a more recent version of LAMMPS,
since bugs are found and fixed all the time and improvements added
at the same time.
cheers,
axel.