Electron phonon coupling convergence

[eliephys78]

Dear all,

I am interested n performing electron phonon calculations using EPW. How can I know that the value I got has indeed converged? In relation to what parameters IN THE NEW VERSION OF EPW (EPW 4.1) shall I check convergence.

Thank you in advance

University of Rondonia
Brazil

[sponce]

Dear eliephys78,

This depend a bit on the properties you want to study. In general you should look for:

- ecut
- increased k & q coarse grid (with associated phonon calculation)
- increased k & q fine grid (although it depend on the properties. If you want linewiths along a BZ path, you do not need to converge k-point off course)
- decreasing broadening parameter: degaussw
- effect of the PSP

In addition you should make sure that the Wannierization is correct.

If you do all of these checks/convergence study, you should have very solid numbers.

Best,

Samuel

[eliephys78]

Thanks for your reply.

Concerning the wannerization, is it sufficient to check that the electronic bands obtained using the Wannier functions coincide with those obtained via DFT? Is this the criterion to check that the wannerization is correct? If not, what is the criteria?

Thanks again

[sponce]

You should check that the Wannierized electronic band given by EPW is the same as DFT but also check that the Fourier phonon bandstructure from EPW is the same as with matdyn.

Its usually also good to check spread and similar info in the output of Wannier90.

[dinesh169]

Hello Everyone,

I also want to ask about wannier convergence. In most of the cases I can successfully match the DFT and interpolated wannier electronic band structures. But my concern is about the wannier function spreads. I can just go up to 0.001-0.00001 A^2 only.
In general spread should also be converged highly or matching between the DFT and wannier band structure is sufficient as said earlier?


best wishes,
Dinesh

[sponce]

Hello,

The spread will never go to 0.

If its "small enough", and you will know with experience, and both el and ph interpolations are fine, then it should be ok.

Best,

Samuel

[dinesh169]

Hello Everyone,

I am working on electron-phonon scattering rates in graphene. I am trying to converge calculation wrt k/q grids. I see some strange behaviors when increasing k/q very high.

I am using EPW v.4.1.0 and here is my input file

1;2305;0c--
&inputepw
prefix = 'gr'
amass(1) = 12.01078
outdir = './epw_output/'
iverbosity = 3

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

etf_mem = .true.

epwwrite = .false.
epwread = .true.

nbndsub = 8
nbndskip = 0
kmaps = .false.

wannierize = .false.
num_iter = 20000
iprint = 2
dis_win_max = 30
dis_froz_max= 6.4

proj(1)= 'c= 0.50024 , 0.28901, 4.86961 : s'
proj(2)= 'c= 0.78758 , 0.44657, 3.86235 : s'
proj(3)= 'c= 1.00000 , 0.86602, 4.87412 : s'
proj(4)= 'c= -0.30138 , 0.50097, 2.23760 : s'
proj(5)= 'c= 0.74999 , 0.43301, 4.87354 : s'
proj(6)= 'c= -0.00056 , 0.57678, -0.00770 : s'
proj(7)= 'c= 0.24998 , 0.43301, 0.00135 : s'
proj(8)= 'c= 0.99939 ,-0.00132, 5.68430 : s'


wdata(1) = 'bands_plot = true'
wdata(2) = 'begin kpoint_path'
wdata(3) = 'G 0.0000000 0.0000000 0.0000000 K 0.6666667 -0.3333333 0.000000'
wdata(4) = 'K 0.6666667 -0.3333333 0.0000000 M 0.5000000 -0.5000000 0.000000'
wdata(5) = 'M 0.5000000 -0.5000000 0.0000000 G 0.0000000 0.0000000 0.000000'
wdata(6) = 'end kpoint_path'
wdata(7) = 'conv_window = 2'
wdata(8) = 'num_print_cycles = 1'
wdata(9) = 'conv_tol = 1.0E-6'
wdata(10)= 'dis_conv_tol = 1.0E-16'
wdata(11)= 'dis_mix_ratio = 0.5'
wdata(12)= 'dis_num_iter = 20000'
wdata(13)= 'guiding_centres = T'


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

parallel_k = .true.
parallel_q = .false.

fsthick = 150 ! eV
eptemp = 300 ! K
degaussw = 0.025 ! eV
dvscf_dir = './save'

system_2d = .true.
! rand_q = .true.
! rand_nq = 10D5
! rand_k = .true.
! rand_nk = 30000
! restart = .true.

filukk = './gr.ukk'
! filqf = './qpath.out'

nk1 = 12
nk2 = 12
nk3 = 1

nq1 = 6
nq2 = 6
nq3 = 1
nkf1 = 300
nkf2 = 300
nkf3 = 1

nqf1 = 300
nqf2 = 300
nqf3 = 1


/
7 cartesian
0.0000000 0.0000000 0.0000000 0.0555556
0.0000000 0.1924501 0.0000000 0.3333333
0.0000000 0.3849002 0.0000000 0.3333333
0.0000000 -0.5773503 0.0000000 0.1666667
0.1666667 0.2886751 0.0000000 0.3333333
0.1666667 0.4811252 0.0000000 0.6666667
0.3333333 0.5773503 0.0000000 0.1111111

Wannier band structure is exactly same as that of dft.
I change fine k/q grids in calculations and nqf = nkf (both grids are same).

Here is the link for the results.
https://www.dropbox.com/s/psg9q4rqb2lbd68/drawing2.pdf?dl=0

As you can see there are humps around +- 1 eV at 300x300 and 500x500 grid. I think its not because of smearing. DO we need to increase the coarse meshes while increasing the fine meshes? or do you have any idea what is the reason for the humps?

And one more question asr_type = 'crystal' is not implemented? I tried to use and code says keyword mismatch (some error like that).

Looking forward for your answers.

Best wishes,

[dinesh169]

Hello Everyone,

Another quick question that imode follows the same order as quantumespresso?

best wishes,

[vinpandey]

hi users,
I am new to this calculation. can you tell me, how do you decide values of wmax_specfun , wmin_specfun , nw_specfun , nbndskip, nbndsub, Dis_win_max and Dis_froj_min ?


thanks

[sponce]

Hello,
@dinesh169: Yes in principle it should follow the same order
@vinpandey: The description of the input variables are given in http://epw.org.uk/Documentation/Inputs

The first 3 you asked are about the spectral function and specify the frequency range you want.
The last 4 are related with Wannier: number of bands from the first one that you skip if you want to Wannierzied, nb of Wannierized abands and Windows of energy in which you want to Wannierze. More info can also be found on the Wannier90 software webpage.

Best,
Samuel