Regarding cartesian points in epw.in for MgB2

[skumarsingh]

Dear EPW users,

I was wondering about the cartesian points in epw.in file for MgB2. Below are the last part of epw.in file:

28 cartesian
0.000000000 0.000000000 0.000000000 0.004629630
0.000000000 0.000000000 0.145933920 0.009259259
0.000000000 0.000000000 0.291867841 0.009259259
0.000000000 0.000000000 -0.437801761 0.004629630
0.000000000 0.192450090 0.000000000 0.027777778
0.000000000 0.192450090 0.145933920 0.055555556
0.000000000 0.192450090 0.291867841 0.055555556
0.000000000 0.192450090 -0.437801761 0.027777778
0.000000000 0.384900179 0.000000000 0.027777778
0.000000000 0.384900179 0.145933920 0.055555556
0.000000000 0.384900179 0.291867841 0.055555556
0.000000000 0.384900179 -0.437801761 0.027777778
0.000000000 -0.577350269 0.000000000 0.013888889
0.000000000 -0.577350269 0.145933920 0.027777778
0.000000000 -0.577350269 0.291867841 0.027777778
0.000000000 -0.577350269 -0.437801761 0.013888889
0.166666667 0.288675135 0.000000000 0.027777778
0.166666667 0.288675135 0.145933920 0.055555556
0.166666667 0.288675135 0.291867841 0.055555556
0.166666667 0.288675135 -0.437801761 0.027777778
0.166666667 0.481125224 0.000000000 0.055555556
0.166666667 0.481125224 0.145933920 0.111111111
0.166666667 0.481125224 0.291867841 0.111111111
0.166666667 0.481125224 -0.437801761 0.055555556
0.333333333 0.577350269 0.000000000 0.009259259
0.333333333 0.577350269 0.145933920 0.018518519
0.333333333 0.577350269 0.291867841 0.018518519
0.333333333 0.577350269 -0.437801761 0.009259259


I know the first three column are taken from the output of phonon calculations (Dynamical matrices for ( 6, 6, 6) uniform grid of q-points
( 28 q-points):) but I didn't get about 4th column. Can someone let me know how to get fourth column?

Thanks and Regards

Sandeep

[roxana]

Hi Sandeep,

The 4th column gives the weight of each irreducible q-point.

Best,
Roxana

[skumarsingh]

Dear Prof. Roxana,

Thanks for your quick response. Is it possible to get weight of each irreducible q-point using kpoints.x in QE or some other approach?

regards

Sandeep

[hlee]

Dear Sandeep:

If you are using EPW v5.1 (included in QE v6.4.1 package) or EPW v5.2 (included in QE v6.5 package) [I don't know about the older versions], you can ignore the fourth column, that is, you can remove the 4th column and you just need the q-point coordinates.

See the relevant part of the subroutine of elphon_shuffle_wrap (excerpt from elphon_shuffle_wrap.f90 in EPW v5.1):

Code: Select all

  !                                                                                                                                                                                                         
  ! Read qpoint list from stdin                                                                                                                                                                             
  !                                                                                                                                                                                                         
  IF (meta_ionode) READ(5,*) nqc_irr
  CALL mp_bcast(nqc_irr, meta_ionode_id, world_comm)
  ALLOCATE( xqc_irr(3,nqc_irr) )
  ALLOCATE( xqc(3,nq1*nq2*nq3), wqlist(nq1*nq2*nq3) )
  xqc_irr(:,:) = zero
  xqc(:,:) = zero
  wqlist(:) = zero
  !                                                                                                                                                                                                         
  IF (meta_ionode) THEN
    DO iq_irr = 1, nqc_irr
      READ(5,*) xqc_irr(:,iq_irr)
    ENDDO
  ENDIF
  CALL mp_bcast(xqc_irr, meta_ionode_id, world_comm)

If you use the development version or the upcoming EPW v5.3 (included in QE v6.6), you don't need to include the q-point list any more in epw.in.

Sincerely,

H. Lee

[skumarsingh]

Dear Lee,

Thank you very much for prompt reply. I am using EPW v5.1 (included in QE v6.4.1 package). It means I do not need 4th column.

regards

Sandeep

[skumarsingh]

Dear Lee,

Do we need weighted kpoints in nscf calculations? Can we also drop in nscf calculations?

thanks and regards

Sandeep

[hlee]

Dear Sandeep:

In this case, you can't remove the weight, but you can include any value, for example, 1.0 .

See the last paragraph below:
(excerpt from the page, https://www.quantum-espresso.org/Doc/IN ... ml#idm1099)

xk_x, xk_y, xk_z, wk REAL
Special k-points (xk_x/y/z) in the irreducible Brillouin Zone
(IBZ) of the lattice (with all symmetries) and weights (wk)
See the literature for lists of special points and
the corresponding weights.

If the symmetry is lower than the full symmetry
of the lattice, additional points with appropriate
weights are generated. Notice that such procedure
assumes that ONLY k-points in the IBZ are provided in input

In a non-scf calculation, weights do not affect the results.
If you just need eigenvalues and eigenvectors (for instance,
for a band-structure plot), weights can be set to any value
(for instance all equal to 1).

However, if you need to run nscf calculations for EPW, I would suggest you to add the k-point list from kmesh.pl in Wannier90 (located in the directory of utility under the root directory of Wannier90).

Sincerely,

H. Lee

[skumarsingh]

Dear Lee,

Thanks for reply in details. It works well.

Could you please explain below terms?

dis_froz_max= 8.8
proj(1) = 'B:pz'
proj(2) = 'f=0.5,1.0,0.5:s'
proj(3) = 'f=0.0,0.5,0.5:s'
proj(4) = 'f=0.5,0.5,0.5:s'

I think f belongs to fractional coordinates here . Why we have projection for B not Mg? What type of projections we need (I mean random or spinor) in epw calculations?

regards

Sandeep

[hlee]

Dear Sandeep:

The choice of initial projections for Wannier functions (WFs) is not unique and it depends on the material of interest.
The choice for MgB2 you mentioned is an optimum one for 5 WFs and it originates from the band and bond character, etc.

Instead, you can try another choice, for example, B s and p orbitals and Mg s orbital [See the Methods section in https://www.nature.com/articles/s41524-019-0191-2], etc; but, in this case, you have a total of 9 WFs.

Usually, we choose the atom-centered projections. However, there is no general recipe applicable to all materials for the choice of initial projections; you can find more information in the literature on Wannier functions.

Sincerely,

H. Lee