Disentanglement convergence criteria not satisfied

[lei2016]

Dear All,

I have been trying to use EPW to calculate electron-phonon interactions in single layer MoS2. However, I always got "Disentanglement convergence criteria not satisfied" even if I have changed DIS_NUM_ITER to 1000 by modifying the source code of EPW. The Delta (frac.) term always oscillates between 1e-3 to 1e-5.

Could you please give me any suggestion on achieving convergence of disentanglement in Wannier?
My input file epw.in and part of nscf_epw.in are shown below.

Thanks,
Lei

Code: Select all

    
nscf_epw.in 
ibrav=  0, nat= 3, ntyp= 2,nbnd=18,
    ecutwfc = 100.0
    !occupations='smearing', smearing = 'mv', degauss = 0.002

Code: Select all

--
&inputepw
  prefix      = 'MoS2'
  amass(1)    = 95.94
  amass(2)    = 32.065
  outdir      = './'

  iverbosity  = 0

  elph        = .true.
  epbwrite    = .true.
  epbread     = .false.

  etf_mem     = .false.

  epwwrite    = .true.
  epwread     = .false.

  lpolar      = .true.

  nbndsub     =  11
  nbndskip    =  0

  wannierize  = .true.
  num_iter    = 500
  iprint      = 2
  !dis_win_max = 180.0
  !dis_win_min = -120.0
  dis_froz_max= -1.8d0
  dis_froz_min = -1.9d0
  !proj(1)     = 'random'   
  proj(1) = 'Mo:d'
  proj(2) = 'S:p'

  elinterp    = .true.
  phinterp    = .true.

  tshuffle2   = .true.
  tphases     = .false.

  elecselfen  = .true.
  phonselfen  = .true.
  a2f         = .false.

  parallel_k  = .true.
  parallel_q  = .false.

  fsthick     = 30.0 ! eV 
  eptemp      = 300 ! K 
  degaussw    = 0.05 ! eV  
  efermi_read = .true.
  fermi_energy= -1.806  
  dvscf_dir   = './save/'
  filukk      = './MoS2.ukk'

  nkf1        = 160
  nkf2        = 160
  nkf3        = 1

  nqf1        = 160
  nqf2        = 160
  nqf3        = 1
  
  nk1         = 16
  nk2         = 16
  nk3         = 1

  nq1         = 8
  nq2         = 8
  nq3         = 1
 /
      10 cartesian
       0.000000000   0.000000000   0.0000000 0.0312500
       0.000000000   0.144337567   0.0000000 0.1875000
       0.000000000   0.288675135   0.0000000 0.1875000
       0.000000000   0.433012702   0.0000000 0.1875000
       0.000000000  -0.577350269   0.0000000 0.0937500
       0.125000000   0.216506351   0.0000000 0.1875000
       0.125000000   0.360843918   0.0000000 0.3750000
       0.125000000   0.505181486   0.0000000 0.3750000
       0.250000000   0.433012702   0.0000000 0.1875000
       0.250000000   0.577350269   0.0000000 0.1875000

[Vahid]

Hi Lei,

For wannierization, you are asking for 11 Wannier functions in a frozen window of 0.1eV wide. Are you sure that all the 11 bands fit into this window?

BTW, to change DIS_NUM_ITER, you can add the following line to your input:

Code: Select all

wdata(1)    = 'dis_num_iter= 1000'

Vahid

Vahid Askarpour
Department of Physics and Atmospheric Science
Dalhousie University,
Halifax, NS, Canada

[lei2016]

Dear Vahid,

Thank you for your reply. Indeed I tried different combinations of the size of frozen window (including the default value), however, none of them worked. I am wondering if there is a general guidance on what values I should use for nbndsub, dis_froz_max/min, dis_win_max/min, and any other relevant parameters. Many thanks!
Lei

Hi Lei,

For wannierization, you are asking for 11 Wannier functions in a frozen window of 0.1eV wide. Are you sure that all the 11 bands fit into this window (including the default size) as well as nbndsub, but none of them generated converged disentanglement calculation. I am wondering if there is any general guidance on choosing the value of nbndsub, dis_froz_max/min, dis_win_max/win, as well as any other relevant parameters. Many thanks!
Lei

BTW, to change DIS_NUM_ITER, you can add the following line to your input:

Code: Select all

wdata(1)    = 'dis_num_iter= 1000'

Vahid

Vahid Askarpour
Department of Physics and Atmospheric Science
Dalhousie University,
Halifax, NS, Canada

[Vahid]

Dear Lei,

You are not skipping any bands (nbndskip=0), and you are looking for 11 Wannier functions, then you need 11 electronic states plus other higher conduction states that are mixed with these 11 states. This determines how many bands to include in the QE calculations (nbnd). You may freeze the 11 states by ensuring that they are inside the dis_froz_min and dis_froz_max window and generate the 11 Wannier functions. If 8 of the states (say valence states) are separated from the other three, e.g. px,py,pz states, by a gap, you can also freeze the 8 states and put all eleven states in a window defined by dis_win_min and dis_win_max. The 3 conduction states are usually mixed with other higher conduction states and so the Wannier code will automatically disentangle the three from the other higher states.

This all depends on the chemistry of the system and the type of bonding. For the well-documented silicon case, you can get 4 wannier functions for the 4 valence bands or 8 functions for 4 valence+4 conduction bands. The bonding is sp3 so you would use sp3 projection in the Wannier input file (proj in EPW input). You may view these functions with XCrysDen or VMD to see if they make sense and check the Wannier centers to see if they are consistent with the chemistry of the material. I always prefer to do the Wannier analysis first and then include the optimized Wannier parameters in EPW.

BTW, the Wannier windows are defined in eV whereas the band structure of QE uses Rydberg.

I am sure the moderators can do a better job in explaining this and I look forward to seeing any corrections to my comments.

Best,
Vahid