Band interpolation wiggling

[Fabiof]

Dear all,

I am trying to plot the bandstructure of the InSe material. But I am getting some wiggling.
I found this topic viewtopic.php?f=9&t=898&p=4002#p4002 and I changed the parameter

lpfac = 20.0_SP to
lpfac = 40.0_SP

in the file ypp/modules/YPP_SET_defaults.F

But it made no difference. Is that normal?


Fabio

[Daniele Varsano]

Dear Fabio,
a possibility is that your k point sampling is not dense enough and ypp makes confusion in the interoplation.
This happens also when trying to interpolate KS band structure?
If the problem arises for the QP bands only there is a way to read PW band structure and interpolate QP correction only.
Let me know. This is not yet documented but I will try to sketch here the needed procedure.

Otherwise you can interpolate using the Wannier90 interface:
In this post you can find how to do that with also a small tutorial thanks to Dr. Antimo Marrazzo.

viewtopic.php?f=9&t=1358&p=6528&hilit=Wannier#p6494

Best,

Daniele

[Fabiof]

Dear Daniele,

It only happens for the QP bandstructure. The KS looks good.
I will try Wannier90 interface to see if the problem is corrected.

Thank you,

Fabio

[Daniele Varsano]

Dear Fabio,

Ok, I think that passing through the Wannier 90 interface is the best option.
Other option, as commented above, is to interpolate QP corrections instead of QP bands.
This requires a multiple step procedure:

1) run a bands calculation in QE: this will be in a separate directory named PW_SAVE
2) do a setup calculation in this directory in order to have SAVE/ns.db1 in this directory
3) use ypp to define the path:
e.g. 01_ypp :

Code: Select all

bzgrids                        # [R] BZ Grid
High_Symm                      # [R] High_Symmetry points and paths
cooOut= "cc"                  # Points coordinates (out) cc/rlu/iku/alat
PtsPath= "K M G K H A L G"     # Series of Points for path generation
NPtsPath= 50                   # Number of Points for each segment of the path

4) then run ypp pointing to the PW_SAVE directory for the path and to the QP database for the corrections:
02_ypp

Code: Select all

electrons                      # [R] Electrons (and holes)
bnds                           # [R] Bands
cooIn= "rlu"                   # Points coordinates (in) cc/rlu/iku/alat
% DOSERange
 1.000000 |-1.000000 | eV      # Energy range
%
DOSESteps=  500                # Energy steps
DOS_broad=  0.10000    eV      # Broadening of the DOS
% INTERP_Grid
 24 | 24 | 12 |                          # Interpolation BZ Grid
%
INTERP_Shell_Fac= 20.00000     # The bigger it is a higher number of shells is used
GfnQPdb=  "E < QP/ndb.QP"
CIRCUIT_E_DB_path= "PW_SAVE"   # SAVE obtained from the QE `bands` run (alternative to %BANDS_kpts)
BANDS_steps=  50               # Number of divisions
% BANDS_bands
  1 |  5 |                     # Number of bands
%

Note the GfnQPdb and CIRCUIT_E_DB_path variable.

In this way corrections instead of QP energy will be interpolated: usually, this ensures a better quality of the final bands.

Best,

Daniele

[Fabiof]

Dear Daniele

Thank you for posting the procedure of the other option.

How and where did you generate the 01_ypp. Is it generated inside the output folder of bands calculation where you do the setup calculation?

Fabio

[Daniele Varsano]

Dear Fabio,
It is generated by typing

Code: Select all

>ypp -k h 

but you are right, I checked and it seems that this feature has been moved to the gpl version only partially, we will fix it soon.
So, I suggest you go through the Wannier interface that should also be more accurate.

Thanks for pointing it out.

Best,
Daniele

[Fabiof]

Dear Daniele,

Thank you for the information.

Fabio

[Fabiof]

Dear Daniele,

Is it possible to have a file with the QP energies for the k-points when symmetries are removed?
For example, I have the QP for 91 k-points. When I remove all symmetries I have the report file with the 324 kpoints with the LDA energies.
Is there a way that I can have the QP for these kpoints without doing another yambo calculation?

[Daniele Varsano]

Dear Fabio,
unfortunately it is not possible, at least in an automatic way.
You can build a QP database by using the ypp -q g, but you need to build the input file by hand,
so assigning the QP correction for each band/kpoint that turns out to be quite anjoying for many k/bands point.

Best,
Daniele

[Fabiof]

Dear Daniele,

Ok, thank you for your answer.

Fabio