EPW Forum

Dear Experts,
When I run the mpirun-np 64 epw.x-npool 64 < epw.in > epw.out command to calculate the phonon spectrum data, this error occurs

[iap01:2753851:0:2753851] Caught signal 8 (Floating point exception: integer divide by zero)
==== backtrace (tid:2753851) ====
0 /lib64/libucs.so.0(ucs_handle_error+0x2a4) [0x145da37ba144]
1 /lib64/libucs.so.0(+0x2231c) [0x145da37ba31c]
2 /lib64/libucs.so.0(+0x225ca) [0x145da37ba5ca]
3 epw.x() [0x462700]
4 epw.x() [0x408632]
5 epw.x() [0x407d62]
6 /lib64/libc.so.6(__libc_start_main+0xf3) [0x14606996c493]
7 epw.x() [0x407c6e]
=================================
forrtl: severe (71): integer divide by zero
Image PC Routine Line Source
epw.x 00000000011C9D6B Unknown Unknown Unknown
libpthread-2.28.s 000014606A4A8B20 Unknown Unknown Unknown
epw.x 0000000000462700 epw_readin_ 830 epw_readin.f90
epw.x 0000000000408632 MAIN__ 95 epw.f90
epw.x 0000000000407D62 Unknown Unknown Unknown
libc-2.28.so 000014606996C493 __libc_start_main Unknown Unknown
epw.x 0000000000407C6E Unknown Unknown Unknown

And the contents of the epw.out file appear as follows

Parallel version (MPI & OpenMP), running on 64 processor cores
Number of MPI processes: 64
Threads/MPI process: 1

MPI processes distributed on 1 nodes
K-points division: npool = 64
227434 MiB available memory on the printing compute node when the environment starts

Waiting for input...
Reading input from standard input

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 1 PID 2753852 RUNNING AT iap01
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

Attached are my epw.in file and path2.dat file.I tried ephwrite=.false., but the same error occurred.

Epw works fine when I use nkf and nqf instead of filkf and filqf. However, the files generated by nkf and nqf have bands data, dos data and phdos data but no phonon spectrum data.What should I do to get the phonon spectrum data?

Best regards,
fale li

Hi Fale li,


Your path2.dat file which you have attached has no points and is empty. That results in divide by 0 error since nktotf is 0. Please provide a correct path2.dat file.

Best regards,
Sabya.

stiwari wrote: ↑Sat Mar 01, 2025 11:20 pm Hi Fale li,


Your path2.dat file which you have attached has no points and is empty. That results in divide by 0 error since nktotf is 0. Please provide a correct path2.dat file.

Best regards,
Sabya.

Dear Sabya,

Is the path2.dat file you see empty? But the path2.dat file I used for calculation is not empty. The file content is as follows

65 crystal
0.000000 0.000000 0.000000 0.076923
0.041667 -0.041667 0.041667 0.076923
0.083333 -0.083333 0.083333 0.076923
0.125000 -0.125000 0.125000 0.076923
0.166667 -0.166667 0.166667 0.076923
0.208333 -0.208333 0.208333 0.076923
0.250000 -0.250000 0.250000 0.076923
0.291667 -0.291667 0.291667 0.076923
0.333333 -0.333333 0.333333 0.076923
0.375000 -0.375000 0.375000 0.076923
0.416667 -0.416667 0.416667 0.076923
0.458333 -0.458333 0.458333 0.076923
0.500000 -0.500000 0.500000 0.076923
0.500000 -0.500000 0.500000 0.076923
0.458333 -0.458333 0.500000 0.076923
0.416667 -0.416667 0.500000 0.076923
0.375000 -0.375000 0.500000 0.076923
0.333333 -0.333333 0.500000 0.076923
0.291667 -0.291667 0.500000 0.076923
0.250000 -0.250000 0.500000 0.076923
0.208333 -0.208333 0.500000 0.076923
0.166667 -0.166667 0.500000 0.076923
0.125000 -0.125000 0.500000 0.076923
0.083333 -0.083333 0.500000 0.076923
0.041667 -0.041667 0.500000 0.076923
0.000000 0.000000 0.500000 0.076923
0.000000 0.000000 0.500000 0.076923
0.020833 0.020833 0.479167 0.076923
0.041667 0.041667 0.458333 0.076923
0.062500 0.062500 0.437500 0.076923
0.083333 0.083333 0.416667 0.076923
0.104167 0.104167 0.395833 0.076923
0.125000 0.125000 0.375000 0.076923
0.145833 0.145833 0.354167 0.076923
0.166667 0.166667 0.333333 0.076923
0.187500 0.187500 0.312500 0.076923
0.208333 0.208333 0.291667 0.076923
0.229167 0.229167 0.270833 0.076923
0.250000 0.250000 0.250000 0.076923
0.250000 0.250000 0.250000 0.076923
0.229167 0.229167 0.229167 0.076923
0.208333 0.208333 0.208333 0.076923
0.187500 0.187500 0.187500 0.076923
0.166667 0.166667 0.166667 0.076923
0.145833 0.145833 0.145833 0.076923
0.125000 0.125000 0.125000 0.076923
0.104167 0.104167 0.104167 0.076923
0.083333 0.083333 0.083333 0.076923
0.062500 0.062500 0.062500 0.076923
0.041667 0.041667 0.041667 0.076923
0.020833 0.020833 0.020833 0.076923
0.000000 0.000000 0.000000 0.076923
0.000000 0.000000 0.000000 0.076923
0.000000 0.000000 0.041667 0.076923
0.000000 0.000000 0.083333 0.076923
0.000000 0.000000 0.125000 0.076923
0.000000 0.000000 0.166667 0.076923
0.000000 0.000000 0.208333 0.076923
0.000000 0.000000 0.250000 0.076923
0.000000 0.000000 0.291667 0.076923
0.000000 0.000000 0.333333 0.076923
0.000000 0.000000 0.375000 0.076923
0.000000 0.000000 0.416667 0.076923
0.000000 0.000000 0.458333 0.076923
0.000000 0.000000 0.500000 0.076923

Best regards,
fale li

Hi Fale Li,

There are a few issues with your input parameters that I would like to clarify. The setting ephwrite=.true. is not compatible with filkf=.true. and filqf=.true. since it requires uniform k and q grids. Additionally, you are using phonselfen=.true. along with eliashberg=.true., which are not compatible with each other.

If you are interested in running superconductivity calculations, then only use ephwrite=.true., eliashberg=.true., liso=.true., limag=.true. along with appropriate temperature values. Otherwise, these tags should be set to their default FALSE values.

If your primary goal is to compute the phonon spectrum, you can use band_plot=.true., filqf=.true., and filkf=.true. I also noticed that your input includes a2f=.true., phonselfen=.true.. If you need phonon self-energy calculations then you must use a uniform k-mesh (nkf) and ensure that filqf is correctly set along the path.

Let me know if this resolves your issues.

Best regards,
Shashi

Hi Shashi，

Thanks for your suggestion, I set ephwrite=.false., eliashberg=.false., liso=.false., limag=.false in epw.in. With filkf and filqf set, epw can operate normally.
But there is still no phonon spectrum data in the output file.

Best regards,
fale li

Hi Fale Li,

Are you interested in the phonon dispersion plot? If yes, please add band_plot = .true. and phonselfen=.false.

Please let me know if you still have issues.

Regards,
Shashi

Hi Shashi，

I do want to get a phonon dispersion plot, and I have band_plot =.true in my epw.in file. phonselfen=.false., but the output file has no phonon dispersion data. Could the reason be that I'm using a virtual crystal approximation to generate the pseudopotential?

Regards,
Fale Li

Hi Fale Li,

I suspect that Fermi energy is not correctly calculated because you commented out fermi_energy in your input file. When you interpolate band structure, you use points along a path, so the code will not calculate the Fermi energy correctly. When you provide uniform k-mesh, the code computes Fermi energy correctly.

Also, you have set fsthick window to be 0.4 eV. Depending on the not-so-accurate Fermi energy estimation from the k-path, the code might have placed the Fermi window incorrectly. EPW uses seleceq window to estimate the number of q-points that fall within the window. If your system is an insulator you might not even get any q-points that would lie within your smaller fsthick window. So I would suggest you remove the fsthick or set it to a very high value along with the correct fermi energy using efermi_read=.true and fermi_energy tags.

It was not clear from your attached input file, which tags you were using. To simplify the setup for phonon dispersion, I have attached an input file for you to try. Please test it, and if you face any issues, send me your input and output files, along with the Wannierization output file, so that I can provide a more accurate suggestion.

--
&inputepw
prefix = 'X',
amass(1) = 118.9024
outdir = './'
dvscf_dir = '../phonon-0Gpa/save'

ep_coupling = .true.
elph = .true.
epbwrite = .true.
epbread = .false.
epwwrite = .true.
epwread = .false.

nbndsub = 6
bands_skipped = 'exclude_bands = 1:4'
efermi_read=.true.
fermi_energy=16.527850 ! give accurate fermi level from scf output

wannierize = .true.
num_iter = 500
iprint = 2
dis_froz_min= 10
dis_froz_max= 21
proj(1) = 'X:s,d'
dis_num_iter=1000
wdata(1) = 'bands_plot = .true.'
wdata(2) = 'begin kpoint_path'
wdata(3) = 'G 0.00 0.00 0.00 H 0.50 -0.50 0.50'
wdata(4) = 'H 0.50 -0.50 0.50 N 0.00 0.00 0.50'
wdata(5) = 'N 0.00 0.00 0.50 P 0.25 0.25 0.25'
wdata(6) = 'P 0.25 0.25 0.25 G 0.00 0.00 0.00'
wdata(7) = 'G 0.00 0.00 0.00 N 0.00 0.00 0.50'
wdata(8) = 'end kpoint_path'
wdata(9) = 'bands_plot_format = gnuplot'

lifc = .true.
asr_typ = crystal

band_plot = .true.
filkf = 'path2.dat'
filqf = 'path2.dat'

nk1 = 6
nk2 = 6
nk3 = 6

nq1 = 6
nq2 = 6
nq3 = 6
/

Regards,
Shashi

Hi Shashi,

Thank you very much for your help! As you can tell, when filkf = 'path2.dat', the Fermi energy calculation is wrong.I used your input file, annotated out dis_num_iter=1000, and successfully obtained the phonon spectrum data.

Regards,
Fale Li