EPW crashes in cdiagh2 subroutine in SOC calculation

[tessw]

Dear EPW creators and users,

I want to calculate the spectral function a2F with EPW and have a working setup for calculations without spin-orbit coupling (SOC). Sadly EPW crashes every time I try to include SOC with the following error message:

Code: Select all

     ...
     Using random q-mesh:        50000
     Size of q point mesh for interpolation:      50000
     Using random k-mesh:       200000
     Size of k point mesh for interpolation:     400000
     Max number of k points per pool:             1390

     Fermi energy coarse grid =  10.770282 eV

     Skipping the first   40 bands:

     The Fermi level will be determined with   0.02725 electrons

     Fermi energy is calculated from the fine k-mesh: Ef =  10.747310 eV

     ===================================================================

              ibndmin =     1  ebndmin =     0.776
              ibndmax =     6  ebndmax =     1.114


     Number of ep-matrix elements per pool :       375300 ~=    2.86 Mb (@ 8 bytes/ DP)

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine cdiagh2 (14):
     info =/= 0
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

     stopping ...

Wannierization for both bands and phonons seems to work (use the new pp.py file from Gitlab after Nov 2018 for SOC calculations if you use QE 6.3 and EPW 5!), but here I am stuck. I already checked the source code at https://gitlab.com/QEF/q-e/blob/develop/PHonon/PH/rigid.f90, but that did not help as well.

My epw.in looks as follows:

Code: Select all

--
&inputepw
  prefix      = 'kto',
  outdir      = './'

  elph        = .true.

  epwwrite    = .true.
  epwread     = .false.
  kmaps       = .false.

  nbndsub     = 6
  nbndskip    = 40

  wannierize  = .true.
  dis_win_max  = 15.160
  dis_froz_max = 14.946
  dis_froz_min = 10.550
  dis_win_min  =  8.480

  proj(1)     = 'Ta:l=2,mr=2,3,5'

  wdata(1) = 'bands_plot = .true.'
  wdata(2) = 'begin kpoint_path'
  wdata(3) = 'G 0.00 0.00 0.00 X 0.50 0.00 0.00'
  wdata(4) = 'X 0.50 0.00 0.00 M 0.50 0.50 0.00'
  wdata(5) = 'M 0.50 0.50 0.00 G 0.00 0.00 0.00'
  wdata(6) = 'G 0.00 0.00 0.00 R 0.50 0.50 0.50'
  wdata(7) = 'R 0.50 0.50 0.50 X 0.50 0.00 0.00'
  wdata(8) = 'X 0.50 0.00 0.00 M 0.50 0.50 0.00'
  wdata(9) = 'M 0.50 0.50 0.00 R 0.50 0.50 0.50'
  wdata(10) = 'end kpoint_path'
  wdata(11) = 'bands_plot_format = gnuplot'

  iverbosity  = 0

  eptemp      = 0.005  ! K    (std: 300.d0)
  degaussw    = 0.025  ! eV   (std: 0.025)
  degaussq    = 0.050  ! meV  (std: 0.05)

  a2f         = .true.
  phonselfen  = .true.

!  lifc        = .true.

  dvscf_dir   = '../phonons/save'

  rand_k       = .true.
  rand_nk      = 200000

  rand_q       = .true.
  rand_nq      = 50000

  nk1         = 12
  nk2         = 12
  nk3         = 12

  nq1         = 4
  nq2         = 4
  nq3         = 4

 /
  10 cartesian
      0.000000000   0.000000000   0.000000000
      0.000000000   0.000000000   0.250000000
      0.000000000   0.000000000  -0.500000000
      0.000000000   0.250000000   0.250000000
      0.000000000   0.250000000  -0.500000000
      0.000000000  -0.500000000  -0.500000000
      0.250000000   0.250000000   0.250000000
      0.250000000   0.250000000  -0.500000000
      0.250000000  -0.500000000  -0.500000000
     -0.500000000  -0.500000000  -0.500000000

Am I doing something wrong here or is this some bug in EPW? As a note, I was able to calculate SOC structures with QE 6.2.1, albeit with a probably slightly different input file some months ago...

[tessw]

Has really no one experienced this error before or knows where to look further to solve this issue?

[sponce]

Dear tessw,

I'm guessing you are doing it correctly but just to be sure:
- you have to modify your scf, nscf.in input to include soc
- you have to double your nb of Wannier function (here I only see 6 bands being Wannierzied).
Does this mean you have 3 when you do without SOC ?
- are you using the same PSP (so a fully relativistic one) for both with and without SOC.
If not, try that.

If that does not help, please also provide your scf, nscf, ph input as well as all the other files without SOC that works.

Best wishes,
Samuel