Relaxing a Transition State

Joshua,

In case you did not notice, you only replied to me instead of to the mailing list.

Back to the reply. It seems there is a problem with your compute cluster and your PBS scheduler - you should probably bring this to the attention of your Admins. I have never seen such a problem on any of the supercomputers I run on, which includes bg/q, cray and intel machines… Please note that not every scheduler uses variables such as mem, pvmem, and vmem.

Ray

Josh,
If your code is calling lammps as a library, until very recently there was a memory leak in the fix qeq/reaxc that made lammps to take over large memory chunks as the driver code ran. Perhaps this is the problem you are facing because the cluster version predates the one that has such a problem fixed.
Carlos

Thanks for the great advice. As it turns out, I went with relaxing the transition state while holding the bonds constrained and this gets me very close to the transition state. Nudged elastic band is too complicated to use for the optimization of the force field for a complex molecular system.

I have been using neb in order to check to see if everything makes sense.

I believe I may have found a bug in the python script neb_final.py. Currently, I am using 16 images along the MEP and I noticed that after executing the script to make dump.neb.final, that the images appear to be out of order (just a guess). So the ordering of the images looks like 1, 10, 11, 12, 13, 14, 15, 16, 2, 3, 4, 5, 6, 7, 9. I am using bash to execute the script if that makes any difference.

Thanks for the great advice. As it turns out, I went with relaxing the
transition state while holding the bonds constrained and this gets me very
close to the transition state. Nudged elastic band is too complicated to use
for the optimization of the force field for a complex molecular system.

I have been using neb in order to check to see if everything makes sense.

I believe I may have found a bug in the python script neb_final.py.
Currently, I am using 16 images along the MEP and I noticed that after
executing the script to make dump.neb.final, that the images appear to be
out of order (just a guess). So the ordering of the images looks like 1, 10,
11, 12, 13, 14, 15, 16, 2, 3, 4, 5, 6, 7, 9. I am using bash to execute the
script if that makes any difference.

have a look at the script itself; there it says:

# -r dump1 dump2 ... = replica dump files of NEB atoms
# must be in correct sequence

so this is not a bug of the script, but how you pass the files.
probably you are using a shell globbing pattern (e.g. -r prefix*.dump)
which sorts by string value and not by numerical value. try using: -r
prefix?.dump prefix??.dump instead.

axel.