band interpolation

[damao4361556]

Hi Everyone

I am sorry for so many posts by me.

Actually, for the band interpolation method if I understood correctly, the procedures are:

1. generate the QP energy, ndb.QP (GW-PPA is enough?)
2. use the post processing tool, ypp -s b -V, generate ypp.in
3. type "ypp"

But as I went through the tutorials, there is also other way to interpolate the band structure, for example:
http://www.yambo-code.org/tutorials/GaSb/bands.php

At the end of this tutorial, there is a picture
which describe the band structure of DFT, DFT(exact), GW, DFT+shissor

What does these differ from the last band interpolation method?

Sometimes I am confused on the four methods (DFT, DFT(exact), GW, DFT+shissor)

thanks

[Daniele Varsano]

Dear Eric,

1. generate the QP energy, ndb.QP (GW-PPA is enough?)

Of course It depends on your system, GW-PPA is usually a good approximation for semiconductors.

3. type "ypp"

By default ypp will generate the DFT band structure, if you want to interpolate the GW band structure you need to include this line in your input:

Code: Select all

GfnQPdb= "E < SAVE/ndb.QP"

Sometimes I am confused on the four methods (DFT, DFT(exact), GW, DFT+shissor)

They are not different interpolation method, but interpolation of bands structures calculated at different levels of theory.
DFT and GW do not need any explication.
DFT (exact) means that it is not interpolated along a path but directly calculated along that path.
DFT+scissor is the so called scissor operator approximation. In order to include quasi particle corrections, when the effect of the GW calculation is to rigidly open the gap and eventually stretch the conduction and valence band you can calculate the GW corrections for few k points and then apply the shift to all the other points. As you can see DFT+sciss and DFT just differ by a constant and they are not too different from the GW bands. Of course if this rude appriximation is valid depend on your system uden study and has to be validated.

Best,
Daniele

[damao4361556]

Hi Daniele

DFT (exact) means that it is not interpolated along a path but directly calculated along that path.

The DFT codes, such as PWSCF, can calculate the band structure along the given path. This is the DFT band structure.
DFT(exact) band structure is also generated by the DFT codes, for a given path?

DFT+scissor is the so called scissor operator approximation. In order to include quasi particle corrections, when the effect of the GW calculation is to rigidly open the gap and eventually stretch the conduction and valence band you can calculate the GW corrections for few k points and then apply the shift to all the other points. As you can see DFT+sciss and DFT just differ by a constant and they are not too different from the GW bands. Of course if this rude appriximation is valid depend on your system uden study and has to be validated.

It is not clear how is this method implemented exactly in Yambo code.
In this post, viewtopic.php?f=14&t=1122&p=5224#p5224, you pointed out that the scissor operator was introduced by variable

Code: Select all

 KfnQP_E

.
In order to do the DFT+scissor band structure, the variable

Code: Select all

 KfnQP_E

is also needed to be set, right?

I am sorry for so many question. But I just want to repeat the tutorials.

thanks

[Daniele Varsano]

Dear Eric,

DFT (exact ) is done with DFT code (nscf along a path). DFT is on a given uniform grid and interpolated along a path.
KfnQP_E is the variable governing the scissor.

Code: Select all

% KfnQP_E
 3.200000 | 1.00000 | 1.00000 |        # [EXTQP BSK BSS] E parameters  (c/v) eV|adim|adim
%

Forst number is the scissor value, the other two gives a linear stretching/shrinking if set bigger/smalle than one of the conduction and valence bands.

Best,
Daniele

[damao4361556]

Hi Daniele

With the GfnQPdb= "none" during the band interpolation step, we will get the LDA band structure, right? This band will be similar to the band obtained from the DFT codes, such as Pwscf?

By the way, in the tutorials, the LDA was always used. The GGA could also be okay to produce the input for Yambo

thanks

[Daniele Varsano]

Dear Eric,

This band will be similar to the band obtained from the DFT codes, such as Pwscf?

Yes, similar and not equal as Yambo perform an interpolation starting from the neregies calculated by pwscf.

By the way, in the tutorials, the LDA was always used. The GGA could also be okay to produce the input for Yambo

Yes you can use GGA as well.

Best,
Daniele