The gap breaks down after GW approximation

[y.mostafa]

Dear Yambo Developers,

I'm currently trying to get the band structure of a silicon structure using yambo, but after adding the GW approximation, the gap breaks down. I am sending the figure of the band structure to make it clear what I mean.

Figure_1.png

In the figure, the black color corresponds to GW calculation and the blue color corresponds to DFT. I have also sent the input files in the attachment.

test3.zip

Please help me in this field.
Thank you very much.
Mostafa Yaghoobi,
PhD student,
Isfahan University of Technology,

[Daniele Varsano]

Dear Y. Mostafa,

please sign your post with your name and affiliation. This is a rule of the forum and you can do once for all by filling your signature in the user profile.

Can you please post also your report file?

Please note that your calculating corrections for 200 bands, which can be a waste of time as usually one is interested in few bands near the gap.

Code: Select all

%QPkrange                        # [GW] QP generalized Kpoint/Band indices
1|57|1|200|
%

Best,
Daniele

[y.mostafa]

Dear Daniele,

Thanks for your quick reply. I have sent the report files (r_electron_bands, r_setup, o.bandss_interpolated,r-yambo_out_HF_and_locXC_gw0_dyson_em1d_ppa) in the attachment. The size of the r-yambo_out_HF_and_locXC_gw0_dyson_em1d_ppa_01 file is 499.7 kB, so I couldn't upload it here.
I appreciate your guidance in advance.
Mostafa yaghoobi,
Isfahan University of Technology

[Daniele Varsano]

Dear Mostafa,
please sign your post with your name and affiliation. This is a rule of the forum and you can do once for all by filling your signature in the user profile.

What I need to check your calculation is the report file of the GW calculation and the one you sent is truncated (unfinished job).
In order to make think easier, can you repeat a GW calculation just for few bands around the gap, e.g

Code: Select all

%QPkrange                        # [GW] QP generalized Kpoint/Band indices
1|57|23|26|
%

when doing this, please check that convergence is reached wrt bands and dimension of the screening matrix, and then, if the problem persists send the report file?

Best,
Daniele

[y.mostafa]

Dear Daniele,

Unfortunately, I still could not solve the problem. Although the band structure changed a little.

I have sent r_yambo..., l_yambo..., ndb.QP, and o-yambo_out.qp in the attachment.
I hope that you mean these files by the report files. Thank you for your help.

ndb.QP.zip

r_yambo_out.zip

o-yambo_out.qp.zip

Mostafa Yaghoobi,
PhD student,
Isfahan University of Technology.

[Daniele Varsano]

Dear Mostafa,

I cannot see errors in your calculation.
I do not know the system you are studying (a sort of porous silicon crystal?).
What I can see is that in the region of the LDA gap there is a sort of Mexican hat, so a situation of inverted bands.

In order to understand what is happening here, I can suggest you to perform an HF (even out-of-convergence) calculation with QE and see if the LDA ordering and shape of the bands persist or it changes qualitatively.

Best,
Daniele

[Davide Sangalli]

I agree with Daniele.

It looks like HF would like to invert the bands ordering, compared to LDA, at K[1] and K[7], and the same remains true for GW.

First, one should check carefully the convergence parameters / approximations (e.g. PPA).
Maybe this band inversion goes away if the parameters are pushed (?)
If not, GW as formulated in the yambo code is not suppose to work fine in these situations.

An alternative could be a better DFT starting point, probably using some hybrid functional.

Best,
D.

[y.mostafa]

Dear Daniele and Davide

Thank you very much for your advice.
It seems that there has been an inverted band here.

Best regards,

Mostafa yaghoobi,
Isfahan University of Technology, Iran.