Al band structures by real-axis GW are not smooth.

[emerarud]

Dear All,

I have been trying to calculate band strcutures of Al with real-axis GW.

However, the conduction bands calculated by real-axis GW seem to be strange (not smooth).

The band structure is attached (Al_bands_Real_axis_GW.png).

I tried to change some paramters (k-points, NGsBlkXd, BndsRnXd, ETStpsXd and so on),
but the results was not improved.

Are there any inappropriate parametes in my input files?

All the input and output files are attached.

Any suggestions are appreciated.

Note that:

The initial wave functions were generated by Quantum Espresso ver.5.1.1.

Yambo was compiled with intel compiler version 11.1.

Both of OpenMP and OpenMPI swicthes were turned on.

Drude term (DrudeWXd), which was calculated by Wien2k, was included.

I did not use "noinv = true" in QE because there are spatial inversions in Al.

Best regards

Kosuke Nakano.

[Daniele Varsano]

Dear Kosuke,

Most probably there is not any problem with your GW calculations, but it is the interpolator which it happens it has problem.
You can try to play a bit with the parameters of interpolator, but not guaranteed it will solve the problem.
You can try the following:
See if it changes something by setting:

Code: Select all

GfnQPdb= "E < Al_GW/ndb.QP" uki

or

Code: Select all

GfnQPdb= "E < Al_GW/ndb.QP" uei

other thing to try is to use more interpolation neighbours, more than 1

Code: Select all

GfnQP_N= 1   

try to raise the Shellfac.

Try to use a regualr mesh: see this post:
viewtopic.php?f=9&t=898&p=4003&hilit=INTERPGrid#p4003

Hope it helps,

Daniele

[emerarud]

Dear Daniele

Thank you for your prompt reply!

Code: Select all

GfnQPdb= "E < Al_GW/ndb.QP" uki

Code: Select all

GfnQPdb= "E < Al_GW/ndb.QP" uei

Code: Select all

GfnQP_N= more than 1   

Code: Select all

Shellfac=1000.00 

Unfortunately, None of these options improved the band structures ...

Try to use a regular mesh: see this post:
viewtopic.php?f=9&t=898&p=4003&hilit=INTERPGrid#p4003

I assumed that regular mesh was necessary only for DOS calculation.
What is appropriate values in my case?

I also searched for "lpfac" in src/modules/mod_interpolate.F, but I could not find the valuable.
It has been already obsolete value, hasn't it?

Best regards.

Kosuke Nakano

[Daniele Varsano]

Dear Kosuke,

sorry, in the previous post the syntax was wrong:
The correct one is:

Code: Select all

GfnQPdb= "E < Al_GW/ndb.QP uki"

or

Code: Select all

GfnQPdb= "E < Al_GW/ndb.QP uei"

Try to see if this improve the band structure.

For the regular mesh, you need to try different values of the grid.

As a last option, it is possible to use the WanT code for band interpolation,
http://www.wannier-transport.org/wiki/i ... /Main_Page
as it can read the QP outputs produced by Yambo.
May be Andrea can add here some hint on how to do that in the case it is not documented in the website.

lpfac is defined here:
ypp/init_ypp/ypp_init.F

Best,
Daniele

[andrea.ferretti]

Dear Kosuke,

just a quick follow up on the possibility to use WanT to interpolate yambo bands.
The interface has been coded in the development version of WanT but not yet released.
In case, I can provide a pre-release version of the code, though the next official release should be coming soon.

take care
Andrea

[emerarud]

Dear Daniele,

Thank you for your prompt reply!

I tried the options:

Code: Select all

 GfnQPdb= "E < Al_GW/ndb.QP uki" 

and

Code: Select all

 GfnQPdb= "E < Al_GW/ndb.QP uei" 

Unfortunately, they did not improve the band structures as attached files.

The results did not change when I set:

Code: Select all

 GfnQP_N= 50 

Code: Select all

 Shellfac=1000.00 

lpfac is defined here:
ypp/init_ypp/ypp_init.F

Thanks. I noticed that lpfac has already been defined as 20_SP.
So, I did not change the value.

Best regards.

Kousuke Nakano.

------------------------------------------------------------------

Dear Andrea,

Thank you for your suggestion!

just a quick follow up on the possibility to use WanT to interpolate yambo bands.
The interface has been coded in the development version of WanT but not yet released.
In case, I can provide a pre-release version of the code, though the next official release should be coming soon.

As mentioned above, I could not improve the band structures.

Could I use the interpolation tools of WanT?

Best regards.

Kousuke Nakano.