tdep-developers/tdep

[phonon_dispersion_relations] ERROR exit code 4: symmetry error

Closed this issue · 8 comments

Hello,
I'm trying to compute phonons of bulk MoS2 and when I use --dos and/or --dumpgrid, I get the following error:

ERROR
 exit code 4: symmetry error
 
 Unexpected tetrahedron weight
 
     occurs in file: type_qpointmesh_gridgeneration.f90
     occurs on line: 1283
        active rank: 3 out of 16

Here is the full command I am using:

mpirun phonon_dispersion_relations -qg 16 16 16 --dos --unit thz -nq 100
Please find attached the infiles:
- infile.forceconstants
- infile.forceconstant_thirdorder
- infile.forces
- infile.meta
- infile.positions
- infile.ssposcar
- infile.ucposcar
- infile.stat
and the log file I got from the execution of the program.

files.zip

If anyone could help with this it would really be helpful, thank you in advance.

It seems to work for odd numbers of qpoints (say 15x15x15), and without MPI. Can you confirm?

It seems to work for odd numbers of qpoints (say 15x15x15), and without MPI. Can you confirm?

Indeed on only 1 core it works fine; on multiple cores the odd number thing you suggest works, otherwise that error is raised.
A bit strange, do you have any idea of why this occurs?

Not yet. Can you check if the result with and without MPI are identical? Same for odd/even grids (for sufficient grid density)

Not yet. Can you check if the result with and without MPI are identical? Same for odd/even grids (for sufficient grid density)

Done. I did serial & parallel with 5x5x5 and the results are identical (I checked eigenvalues and eigenvectors), and serial with 63x63x63 and 64x64x64 and again identical.

Thanks for checking! We'll keep the issue open as a reminder to look into this at some point

@chewingram can you check with the latest commit a470018 ?

Hi! I just tried on a 16x16x16 grid with 8 cores before and after pulling, and I confirm now it works.
Thanks!

Thanks for checking!