[Wannier] DOS calculation with full and IBZ giving different results

Sat Dec 13 09:40:21 CET 2014

Hi Raul,
thank you for your detailed answer.
It seems to me that even the energies may not be the same in the points that you expect to be equivalent. A couple of further questions then, sorry if they may seem trivial…

1. Did you check that the structure has exactly the exact symmetry that you expect?
1bis. Can you check with the ab-initio code to check the eigenenergies at the points that should be degenerate to see if the energies are the same?
2. Did you check that the Wannier functions are well converged?
2bis. Can you check that the interpolated bandstructure with Wannier functions satisfy the symmetry that you expect, e.g. doing a band plot along two equivalent bands?

Thanks,
Giovanni

--
Giovanni Pizzi
Post-doctoral Research Scientist
EPFL STI IMX THEOS
MXC 340 (Bâtiment MXC)
Station 12
CH-1015 Lausanne (Switzerland)
Phone: +41 21 69 31124

On 13 Dec 2014, at 00:25, Raul Laasner wrote:

Thanks for responding. Yes, also without adaptive smearing the results differ a little bit. However, the difference decreases with k-mesh size. With dos_kmesh 10 the difference is very small and with lower values the results are identical.

For debugging purposes, I tried dos_kmesh 2 with adaptive smearing. If I put

  8
  0.00000000E+00 0.00000000E+00 0.00000000E+00 0.12500
  5.00000000E-01 0.00000000E+00 0.00000000E+00 0.12500
  0.00000000E+00 5.00000000E-01 0.00000000E+00 0.12500
  5.00000000E-01 5.00000000E-01 0.00000000E+00 0.12500
  0.00000000E+00 0.00000000E+00 5.00000000E-01 0.12500
  5.00000000E-01 0.00000000E+00 5.00000000E-01 0.12500
  0.00000000E+00 5.00000000E-01 5.00000000E-01 0.12500
  5.00000000E-01 5.00000000E-01 5.00000000E-01 0.12500

into my kpoint.dat, then I get the correct result. However, with

  3
  0.00000000E+00  0.00000000E+00  0.00000000E+00 0.12500
  5.00000000E-01  0.00000000E+00  0.00000000E+00 0.50000
  5.00000000E-01  5.00000000E-01  0.00000000E+00 0.37500

the results again differ. (I forgot to say that NaI has the rock-salt structure.)

I had a look at the loop that performs the main calculation in dos_main and inserted

  write(*,'(*(f7.5,1x))') kpt, dos_k(452,1)*kweight

after the 'dos_all=dos_all+dos_k*kweight' line to see how each k-point contributes to the sum. Here the index 452 refers to some energy level that I randomly picked. The result is

  0.00000 0.00000 0.00000 0.22149
  0.00000 0.00000 0.50000 4.22194
  0.00000 0.50000 0.00000 3.32614
  0.00000 0.50000 0.50000 0.00000
  0.50000 0.00000 0.00000 1.11124
  0.50000 0.00000 0.50000 0.00000
  0.50000 0.50000 0.00000 0.00000
  0.50000 0.50000 0.50000 0.45465

where the last column is the corresponding density. By symmetry, I would expect only 3 numbers to differ in this column, but this is not the case. Have I misunderstood something?

As another check, before the main loop, I put

  kpt = [0.5,0.0,0.0]
  call get_eig_deleig(kpt, eig, del_eig, HH, delHH, UU)
  print*, norm2(del_eig(1,:))
  kpt = [0.0,0.5,0.0]
  call get_eig_deleig(kpt, eig, del_eig, HH, delHH, UU)
  print*, norm2(del_eig(1,:))

which prints the magnitude of the band derivative of the first band at these two k-points, with the result

  0.44616527636129560
  3.1348138571982227E-003

Shouldn't these numbers be the same by symmetry?

Finally, I'm not sure what you mean by 'units for the kpoints' as I'm only entering the reduced coordinates.

Raul

On Fri, Dec 12, 2014 at 5:38 PM, Giovanni Pizzi <giovanni.pizzi at epfl.ch<mailto:giovanni.pizzi at epfl.ch>> wrote:
Dear Raul,
in order to understand if the problem is in the dos routine, or in the routine that generated the IBZ points, could you run the same thing (i.e. DOS, with and without wanint_kpoint_file), but using dos_adpt_smr=False, and instead a fixed-width smearing? (e.g. a gaussian smearing)
And then compare the results, and report them here?

As a further detail, could you also check if the units for the kpoints that you are passing in the wanint file are the one expected by the code?

Thanks,
Giovanni

On 12/12/2014 11:09 AM, Raul Laasner wrote:
Dear all,

When I calculate the density of the lowest conduction states of NaI (see attachment), I get different results for using k-points from the full and irreducible Brillouin zones. The difference is smaller when the k-mesh is allowed to use less symmetries, e.g. only the time reversal symmetry. There is no difference if the 'kpoint.dat' file contains the full BZ. This suggests the code I use for generating IBZ points might be in error, but I get the same results with both abinit and elk (they deliver different, but equivalent IBZ points). Could it be that the DOS calculation is very sensitive to small numerical inaccuracies and this leads to slightly different results for full and irreducible BZs? The DOS related part of my input file is the following:

dos true
dos_kmesh 20
dos_adpt_smr true
wanint_kpoint_file false # true for the second run

The difference is also present with dos_kmesh 50. Please ask for other details I'm not showing. Any suggestions are welcome.

Thanks,
Raul Laasner

--
Raul Laasner
Institute of Physics
University of Tartu
Ravila 14c, 50411, Estonia
e-mail: raullaasner at gmail.com<mailto:raullaasner at gmail.com>
cell: (+372)5182268<tel:%28%2B372%295182268>

_______________________________________________
Wannier mailing list
Wannier at quantum-espresso.org<mailto:Wannier at quantum-espresso.org>
http://mailman.qe-forge.org/cgi-bin/mailman/listinfo/wannier

--
Giovanni Pizzi
Post-doctoral Research Scientist
EPFL STI IMX THEOS
MXC 340 (Bâtiment MXC)
Station 12
CH-1015 Lausanne (Switzerland)
Phone: +41 21 69 31124

_______________________________________________
Wannier mailing list
Wannier at quantum-espresso.org<mailto:Wannier at quantum-espresso.org>
http://mailman.qe-forge.org/cgi-bin/mailman/listinfo/wannier

--
Raul Laasner
Institute of Physics
University of Tartu
Ravila 14c, 50411, Estonia
e-mail: raullaasner at gmail.com<mailto:raullaasner at gmail.com>
cell: (+372)5182268

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20141213/db65d0bd/attachment.html>