[CP2K-user] [CP2K:20828] AIMD with NPT (NPT_I?) gets quickly stuck

[Pierre Beaujean]

Hello,

I have an issue while running AIMD with NPT (here with NPT_I): on some 
systems, after a few steps (about ~200, but that can vary), the calculation 
just stop doing anything useful (output does not change, nor the .dcd or 
.ener files). However, when SSHing the calculation nodes, CPU usage is 
still high. So maybe it is stuck in a loop somewhere.

More information:

   - Here is an example input: input.txt 
   <https://gist.github.com/pierre-24/03841f80a23dcc303befc8527577f5e3#file-md-inp>. 
   I would rather not share the geometry publicly, but lets say that it is a 
   polymer chain with a few molecules of organic solvent (ethylene carbonate, 
   in fact), and it contains nothing exotic (but what is exotic and what is 
   not is probably a matter of taste :p ). The corresponding output is there 
   <https://gist.github.com/pierre-24/03841f80a23dcc303befc8527577f5e3#file-md-out>
   - The system was previously relaxed with at 10 ps NVT that ran smoothly. 
   I first tried to restart the calculation from the last NVT step, adding the 
   barostat, same result.
   - Some systems of similar size and composition (actually, with other 
   solvents) runs smoothly.
   - I add already post this problem on Github (see 
   https://github.com/cp2k/cp2k/issues/3361), but it was occuring with NPT_V, 
   while NPT_I was working. My criterion for the convergence were indeed a bit 
   loose back then, so this time, I reduced them by two order of magnitude 
   (EPS_SCF is now 1e-8 and EPS_DEFAULT is 1e-12), but ... No luck there :(
   - I'm currently using issue with 9.1 (changing the version during a 
   project is not always a good idea), but I also find this error with 2023.1 
   as well (I could try with the latest version, but I need to go into the 
   trouble of making a new EasyBuild file). I use cp2k.psmp on 256 cores 
   with MPI (OMP_NUM_THREADS is set to 1).
   - I tried on another supercomputer, same problem.
   - It *always* stop at the same point (see the output there 
   <https://gist.github.com/pierre-24/03841f80a23dcc303befc8527577f5e3#file-md-out>), 
   right before a new MD step: the "Total energy" line is always the last 
   thing printed (this was already the case in my Github message).

If you have any insight, thanks :)

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/e02be921-0c06-45b0-9564-a4e6e4ff1964n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20241028/9e6cbd01/attachment-0001.htm>