My Community

Dear vasp community,

I am calculating band structure of TiSe2 with SCAN functional. The calculation itself was okay, but it crashed at the end.

I use one-shot band structure calculation with KPOINTS_OPT. I also used LORBIT=14 for projection.

This is my log file:
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[acn21:3253866:0:3253866] Caught signal 11 (Segmentation fault: invalid permissions for mapped object at address 0x1492880e5000)
==== backtrace (tid:3253866) ====
0 0x0000000000012d10 __funlockfile() :0
1 0x00000000004a1d7b vaspxml_mp_xml_procar_() :0
2 0x0000000001cd3123 mlr_main_mp_interpolate_optics_dense_() :0
3 0x0000000001e8591d MAIN__() :0
4 0x000000000040dee2 main() :0
5 0x000000000003a7e5 __libc_start_main() :0
6 0x000000000040dde9 _start() :0
=================================
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
vasp_std 000000000201683A Unknown Unknown Unknown
libpthread-2.28.s 0000149401449D10 Unknown Unknown Unknown
vasp_std 00000000004A1D7B Unknown Unknown Unknown
vasp_std 0000000001CD3123 Unknown Unknown Unknown
vasp_std 0000000001E8591D Unknown Unknown Unknown
vasp_std 000000000040DEE2 Unknown Unknown Unknown
libc-2.28.so 0000149400D197E5 __libc_start_main Unknown Unknown
vasp_std 000000000040DDE9 Unknown Unknown Unknown

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 1 PID 3253867 RUNNING AT acn21
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

This is my INCAR:
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
#start parameters
NWRITE = 1
PREC = Accurate #precision mode
#electronic optimization
ENCUT = 650 #cutoff energy
EDIFF = 1.0e-6 #breakout condition for SC loop
NELMIN = 6 #minimum number of electronic SCF steps
NELM = 200
#ionic relaxation
IBRION = -1 #relaxation method
NBANDS = 128
#DOS-related
ISMEAR = 0 #determines how the partial occupancies are set for each orbial
SIGMA = 0.005
#Write flags
LORBIT = 14
LWAVE = T
#Exchange correlation treatment
METAGGA = SCAN
LASPH = .TRUE
ALGO = D
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Can't I use metaGGA, LORBIT=14, KPOINTS_OPT simultaneously on vasp 6.4.1?

I am sorry. This issue would be closer to a bug report

Dear hyeonseo_park,

Thank you for this bug report.

So far I'm not able to reproduce this issue, but I had to reduce the number of k-points and cutoff because your calculation is too large. For me the calculation exits normally at the end. Could you please check if you can reproduce this issue in a smaller calculation? For example, do you still get this error if you reduce the number of k-points, the energy cutoff and use fewer MPI ranks?

From the provided log it seems that the calculation crashed inside routine XML_PROCAR. Which suggests that the projections were calculated correctly, but the crash occurs when the code was writing them into vasprun.xml.

1 0x00000000004a1d7b vaspxml_mp_xml_procar_()

Best wishes,
Alexey

Dear Alexey,

Thank you for your confirmation.

I also checked this in a smaller calculation and it worked with the following changes.

1. Even though I already specified "ulimit -s unlimited" in .bashrc, I specified it in a job submission script "submit.sh" before running vasp.
2. After running vasp, I had the following code to remove unnecessary files:
mpirun -np $SLURM_NTASKS vasp_std >& log2

wait

#rm WAVECAR
#rm CHG
#rm vasprun.xml
#rm slurm-*

exit

Based on your reply, I think the line "rm vasprun.xml" may affect the process. This line might delete vasprun.xml before writing the PROCAR_OPT. I removed this for now.

Because it worked in a smaller calculation, I am calculating the bigger one with these changes.

Sincerely,
Hyeonseo

In a smaller calculation, it worked without any issues.

However, in the larger calculation, the problem persists. The calculation does not complete as expected. I have attached the vasprun.xml file here for your reference. It appears that the writing of projection information in the vasprun.xml file stopped midway.

I will try the same calculations with smaller k points (20 per path) and small number of processors.

By the way, does the fewer MPI ranks mean the smaller number of processors?

I ran your calculation with VASP 6.4.1 and VASP 6.5.0. I can reproduce this crash in VASP 6.4.1, but it doesn't occur in VASP 6.5.0.

Do you have access to VASP 6.5.0? Can you run your calculation with the newest version?

Thank you, Alexey.

I will discuss the software update with our institution.

In the meantime, I will proceed with the calculations using 15-20 k-points per path. I found that this scale works without decreasing quality of results.

Sincerely,
Hyeonseo