Band convergence strategy (IP RPA)

[muhammadhasan]

Hi,

Would you please guide me how should I select the bands to do the convergence test? Here is a part of the r_setup file of Gold (RPA)

Conduction Band Min: 20.64221 ev
Valance Band Max: 20.64221 ev
Filled Bands: 602
Metallic Bands: 603 616
Empty Bands: 617 800

Should I start with 616-617 and then 615-618, 614-619, 613-620,.... etc?
Or I should start from 1-617, 1-618, 1-619, 1-620,..... etc?

I am also confused with the same value of conduction Min and valance Band Max though I have used converged DFT result to generate the r_setup file.

Thank you so much for your time.

Best
Md J Hasan
PhD Student
Mechanical Engnieering
University of Maine

[Daniele Varsano]

Dear Md J Hasan,

as you are calculating IP spectrum, the more bands you include the more transitions are taken into account. You can add them progressively until the range of the spectrum you are interested in is converged. If you are not interested in the high energy part of the spectrum you can avoid including deep state that will contribute to the high energy part.

Please note you system is metallic, that's why you get the same VBM and CBM, note you have bands from 603 to 616 that are partially occupied. At least some bands fully occupied and completely empty should be considered.

Best,
Daniele

[muhammadhasan]

Hi Professor,

Thank you so much. Actually I am interested on dielectric function between 0 to 2 ev of energy. Then I think, 601-618, 600-619, 599-620 ...etc should be good for me based on your suggestion (avoid including deep state and fully occupied, completely empty considered)

Thanks!

Another question regarding that is,
I am doing convergence test now. The convergence of energy cut off is completed and I found 45 Ry (FFTGvecs). My input command was:

Code: Select all

 $ yambo -F yambo.in_IP -J 45Ry_fullbnd

where, I have used

Code: Select all

%BndsRnXd
1 | 800
%

In this run, the dipole matrix elements p (ndp.dipoles) is saved in 45Ry_fullbnd folder.
so next time, when I start band convergence, Can I use the same folder as follows?

Code: Select all

$ yambo -F band_convergence.in_IP -J 45Ry_fullbnd

then each time of band convergence calculation, yambo reads from the 45Ry_fullbnd, so it does not need to compute the dipole matrix elements again, and the calculation will be fast.
Or should I use different folder (-J) each time for different band range?

Thanks again.

Best
Md J Hasan
PhD Student
Mechanical Engineering
University of Maine

[Daniele Varsano]

Dear J Hasan,

usually, FFTGvecs can be left at its default. Of course, you can reduce them, but then you need to pay attention to the orthonormality of the wavefunction. In general this is not a parameter of convergence unless you have a memory problem.

Yes, using the -J option Yambo will write and read database from the SAVE directory and the one indicated in the -J option. If you stored there all the dipoles, yambo will read them.

Best,
Daniele

[muhammadhasan]

Hi Professor,

Thank you for your helpful comments.

You are right, due to memory problem I have decided to reduce it further. Primarily it was FFTGvecs= 99845 RL and then, I reduced it to 40 Ry. As you said we need to pay attention to the orthonormality of the wavefunction. I have followed https://www.yambo-code.eu/forum/viewtopic.php?t=1932, where you have shown how to check the orthonormality. In my case (40Ry) I got it as follows:

Code: Select all

 [WF-Oscillators/G space/Transverse up loader] Normalization (few states)  min/max  :  0.36356E-10  0.99997 

It is very close to 0,1. so, I decided to fix the value for the next runs.

Thanks.

Best
Md J Hasan
PhD Student
Mechanical Engineering
University of Maine