Dielectric Behavior in Varying Supercell Sizes

[y.mostafa]

Dear Yambo Developers,

I have a question regarding the dielectric behavior of a structure in relation to its size.

In my recent study, I acquired the dielectric function for a silicon structure. Subsequently, I expanded the silicon structure by doubling it on each side and performed BSE calculations again for the larger supercell.

Upon analysis of the imaginary part of the dielectric function diagrams for both the original silicon structure and the enlarged supercell, I observed a reduction in absorption for the larger supercell structure. It's worth noting that I used the same parameters for both calculations. I have attached the input files and report files for your reference.

I am curious to know if the absorption characteristics of materials are influenced by the number of atoms in their structures. Could you please provide insights or guidance on this matter?

Figure_1.png

Your assistance in clarifying this issue would be greatly appreciated.
I had mistakenly posted this post in the electron and phonon effect section before, but I could not find a way to delete that post.
Thank you for your time and support.

Sincerely,
Mostafa Yaghoobi

[Daniele Varsano]

Dear Mostafa,

as you argued, the result should be the same. Anyway, it seems that your two calculations are not equivalent.
The two cells have different size, so to have equivalent calculations the k point grid should be scaled accordingly.
Also, the number of bands and energy cutoff for the screening should be scaled accordingly, in the case of cutoff you provided it in energy unit so it is ok (different is you had given in number of reciprocal lattice vectors), the number of bands (both in the summation for the screening and in the BSE kernel) should be chosen accordingly to the electron-hole energy.

Best,
Daniele

[Davide Sangalli]

Let me add that, due to the folding, it is probably not possible to obtain the same spectra using BSEbands alone.
The variable BSEEhEny can be used tu select al the states within a give energy range..

Best,
D.

[y.mostafa]

Dear Friends
I have a few silicon structures that have direct band gap in the 1 to 1.3 eV range. The size of the structures ranges from 12 atoms to 40 atoms. Considering the amount of band gap energy and the directness of the energy gap of the structures, I predict that the optical absorption spectrum of these structures will show better absorption compared to the silicon diamond structure. But with the increase in the number of atoms, absorption decreases.
So I wanted to test to see if the absorption structure changes with increasing size (which, of course, should not change). The calculation I made gives me the perspective that the BSE calculation depends on the size of the structure. That is, I have to do the calculation based on the size of the structure.

To obtain the absorption spectrum of the structures, I performed three calculations according to the following commands.

Code: Select all

yambo -X s -F screening.in
yambo -F screening.in -J BSE_01
yambo -o b -k sex -F BSE_kernel.in -J BSE_01
yambo -F BSE_kernel.in -J BSE_01
cp BSE_kernel.in BSE_diag_solver.in
yambo -F BSE_diag_solver.in -y d -V qp -J BSE_01
yambo -F BSE_diag_solver.in -J BSE_01 

But in file screening.in, I changed the parameters as follows:

Code: Select all

NGsBlkXs= 8                Ry    # [Xs] Response block size
% LongDrXs
 1.000000 | 1.000000 | 1.000000 |        # [Xs] [cc] Electric Field
%

And in file BSE_kernel.in, I changed the parameters as follows :

Code: Select all

BSENGexx=  40           Ry    # [BSK] Exchange components
BSENGBlk=  8             Ry    # [BSK] Screened interaction block size [if -1 uses all the G-vectors of W(q,G,Gp)]

also I adjust BSEBands as follow:

Code: Select all

 
% BSEBands
    5 of valance bands |  5 of conduct bands |                       # [BSK] Bands range
%

, and in BSE_diag_solver.in i set KfnQPdb to add GW correction to calculation,

Code: Select all

KfnQPdb= "E < GW/ndb.QP"                  # [EXTQP BSK BSS] Database action

I also adjusted the number of K points according to the size of the structures (the k_points distance is 0.15 1/A for all structures).
Also I set number of bands to 300 for all of the structures in GW correction calculation and in screening.

Code: Select all

% BndsRnXs
    1 |  300 |                       # [Xs] Polarization function bands
%

Please let me know if I did something wrong.
Best,
Mostafa Yaghoobi.

[Daniele Varsano]

Dear Mostafa,

as stated above, also from Davide, the number of bands (both in BSE and X calculations) should be set accordingly to the electron-hole energy difference, i.e. they should span the same energy range. You can set this, looking at the eigenvalues present in the report or in the QE output.

Best,
Daniele

[y.mostafa]

Dear Friends,
Thank you for your cooperation and guidance.
I understand what you mean.
However, this seems impossible (considering the same energy range for structures).
For example, according to the band structure of silicon, if I want to choose the least number of bands for BSE calculations, I should choose at least 3 bands. Their energy range will be between 2 and 10 eV. Now if I want to choose the same range for another structure that I want to examine, the number of bands would be close to 80. It will be practically impossible to do this calculation.
Is there an alternative solution for this?
Thank you for helping me in this matter.
I appreciate your time and consideration.

[Daniele Varsano]

Dear Mostafa,

it depends on the energy range you are interested in studying the spectrum. If you are interested in low energy excitations, you need to converge them for each system. What was reported in the previous post from me and Davide, it was the condition to have two equivalent calculations when the system is modeled with a different unit cell.

Best,
Daniele

[y.mostafa]

Dear Daniele Varsano,
Thank you for your cooperation and support.