Problems while integrating MBX with the LAMMPS PIMD functionality

[Kehan Cai]

Dear MBX developers,

Hope this message finds you well!

This is a new MBX user. Recently we have been working on integrating the MBX with the LAMMPS PIMD functionality. We found it worked well while running test tasks, e.g., PIMD simulations for approximately 1 ps with a step size of 0.5 fs. However, when we performed simulations for a longer time, e.g., 100 ps, the task was terminated at around 50 ps, accompanied by an error message saying:

gammq: x = -nan, a = 0.75

lmp_mpi_mbx: potential/electrostatics/gammq.cpp:156: double elec::gammq(double, double): Assertion `x >= 0.0 && a > 0.0' failed.

After checking the source code of MBX, it seemed to me that the problem arose from the expansion of the Gamma function, but I could not diagnose it more clearly. I will appreciate it if you may help me solve the problem.

FYI, the Slurm output file is attached to the email, which includes full error messages. 

The LAMMPS version is 29 Aug 2024, and the MBX is updated.

The initial configuration used here was copied from the example folder of MBX:

MBX/plugins/lammps/examples/512h2o/initial.data 

Please let me know if you need any other information or files.

Thank you for your assistance in advance!

Best regards,

Kehan

[Richa Rashmi]

Dear Kehan,

Thank you for your email. Could you please also send us the LAMMPS input file (lammps.in) that you used for this simulation?

Thank you.
Best regards,
MBX team

[Kehan Cai]

Hi all,

Attached please find the LAMMPS input script for running PIMD. The key command that differs from the conventional classical MD is

    fix pimd/langevin ...

which is introduced in the website of

    https://docs.lammps.org/fix_pimd.html

Please let me know if you have any questions.

Thank you!

Best regards,

Kehan

--
You received this message because you are subscribed to the Google Groups "MBX-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mbx-users+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/mbx-users/faa0fa73-8e55-4277-b965-58d5403174afn%40googlegroups.com.

[Richa Rashmi]

Dear Kehan,

Thank you. Could you please also share with us your LAMMPS trajectory file for this simulation?

Best regards,

MBX Team

[Kehan Cai]

Hi all,

Attached please find a compressed file containing trajectory files of all beads/replicas, where I used 32 beads in this simulation so there are 32 trajectory files.

 traj.tar.gz

Best regards,

Kehan

To view this discussion visit https://groups.google.com/d/msgid/mbx-users/eab9586e-874c-4c8d-9019-7f6bbf1aac93n%40googlegroups.com.

[Richa Rashmi]

Dear Kehan,

1. Have you tried the simulations with a smaller timestep, say 0.1 fs? Do you still see a crash? Please refer to https://groups.google.com/u/1/g/mbx-users/c/GhcmUDy7pcY/m/gVZIZQXFBAAJ.
2. Could you please change the restart frequency to 10 when doing the 0.1 fs tests, and if it crashes please send us the restart files and other input files?

Thank you.
Best regards,
MBX Team

[Kehan Cai]

Hi All,

Thanks for the suggestion! I will try a simulation with a step size of 0.1 fs, and I will update the result if there is any anomalous issue raised.

Best regards,

Kehan

To view this discussion visit https://groups.google.com/d/msgid/mbx-users/5c74903b-ce28-4bad-994d-74df08678cb1n%40googlegroups.com.