CBM & VBM

[xjxiao]

Dear all,
While reproducing the AlSb calculations from the literature, I noticed that Yambo might have swapped the conduction band minimum (CBM) and valence band maximum (VBM) at the Gamma point.
In PBE calculations, the system shows a negative bandgap, but Yambo assigned: The lower-energy state (and its GW correction) as the valence band and the higher-energy state (and its GW correction) as the conduction band. This leads to odd band structures when plotted with ypp

AlSb_mywork.png

, contradicting the literature.

AlSb_literature.png

Could there be an issue with my calculations? Thanks!
Xiao

[Daniele Varsano]

Dear Xiao,

1) What do you mean by negative gap? Is the system a metal? Given the PBE electronic structure, the assignment done by Yambo seems to me to be reasonable. If you think that the system should be a metal, is it possible that you used occupation fixed in your QE calculation? In thisc case consider adding a smearing.

2) I can see you have a 2D k point sampling, so you are dealing with a 2D system, right? Consider isolating the system using a truncated coulomb cutoff (yambo -r).
rim_cut
RandQpts= # [RIM] Number of random q-points in the BZ. --Usually few millions
RandGvec= 97 RL # [RIM] Coulomb interaction RS components --Usually few tenths
CUTGeo= "slab z"


You can also use acceleration algorithm with respect k points. (yambo -rw)
https://wiki.yambo-code.eu/wiki/index.p ... 2D_systems

3) As your band structure is rather delicate, try to understand what's happening without adding uncertainties due to the extrapolation of the bands. Consider also to interpolate the bands using yambopy.


Best,

Daniele

[xjxiao]

Dear Xiao,

1) What do you mean by negative gap? Is the system a metal? Given the PBE electronic structure, the assignment done by Yambo seems to me to be reasonable. If you think that the system should be a metal, is it possible that you used occupation fixed in your QE calculation? In thisc case consider adding a smearing.

2) I can see you have a 2D k point sampling, so you are dealing with a 2D system, right? Consider isolating the system using a truncated coulomb cutoff (yambo -r).
rim_cut
RandQpts= # [RIM] Number of random q-points in the BZ. --Usually few millions
RandGvec= 97 RL # [RIM] Coulomb interaction RS components --Usually few tenths
CUTGeo= "slab z"


You can also use acceleration algorithm with respect k points. (yambo -rw)
https://wiki.yambo-code.eu/wiki/index.p ... 2D_systems

3) As your band structure is rather delicate, try to understand what's happening without adding uncertainties due to the extrapolation of the bands. Consider also to interpolate the bands using yambopy.


Best,

Daniele

Dear Daniele,
Thanks a lot for your reply!
This is a 2D system, which is a direct-gap semimetal under PBE calculations. The above figure, AlSb_mywork, was plotted using ypp. If I directly plot the PBE band structure using QE, the result is as follows.

AlSb_QE.png

The PBE and GW energies at the Gamma point are shown below, and it can be observed that band inversion occurs after GW correction, indicating that it should be a direct-gap semiconductor.

Code: Select all

#    K-point            Band               Eo [eV]            E-Eo [eV]          Sc|Eo [eV]         E [eV]
        1                   52                0.000000          -0.149666          -2.161419          -0.14967
        1                   53                0.129740          -1.008197           0.542081          -0.87846

However, during the BSE calculation, when reading ndb.QP, Yambo assumes the original system was a semiconductor and that it becomes a semimetal after GW correction. Is this problematic? The optical absorption spectrum I calculated does not match the literature. The left figure below is from the literature, and the right figure is my result.

optical.png

Where might the issue lie?

P.S. The literature did not use the -r flag when calculating GW, so I also omitted it.

Best wishes!

[Daniele Varsano]

Dear Xiangjun,

from the report file, it seems that Yambo recognize the system as a metal both at PBE and GW level.
I suggest you to check the entire GW band structure and convince you if this is the case. Actually, from the first plot you sent this seems to be the case. If you expect that GW should open the gap I suggest you to check your structure and convergence parameters.

Also, I suggest you to set BSEmod to "retarted" to include the antiresonant part in the BSE.

Best,

Daniele

[xjxiao]

Dear Daniele,

from the report file, it seems that Yambo recognize the system as a metal both at PBE and GW level.
I suggest you to check the entire GW band structure and convince you if this is the case. Actually, from the first plot you sent this seems to be the case. If you expect that GW should open the gap I suggest you to check your structure and convergence parameters.

Thanks for your patient reply! My parameters are all adopted from the literature and should be well-converged.
Based on the current information, the system appears to be a semiconductor at PBE level.

[X] === Gaps and Widths ===
[X] Conduction Band Min : 0.129740 [eV]
[X] Valence Band Max : 0.000000 [eV]
[X] Filled Bands : 52
[X] Empty Bands : 53 720
[X] Direct Gap : 0.129740 [eV]
[X] Direct Gap localized at k : 1
[X] Indirect Gap : 0.129740 [eV]
[X] Indirect Gap between kpts : 1 1
[X] Last valence band width : 2.031530 [eV]
[X] 1st conduction band width : 1.830955 [eV]

When Yambo reads energy data, it seems to automatically treat the smaller energy value as the lower band, potentially misidentifying a semimetal as a semiconductor. However, after GW correction, the energy ordering of the conduction band minimum (CBM) and valence band maximum (VBM) at the Gamma point is reversed, indicating that the system has opened a bandgap.
Does Yambo distinguish bands solely based on energy values, or does it also consider wavefunction characteristics? (Note: Wannier interpolation can correctly reproduce the band structure.)

Also, I suggest you to set BSEmod to "retarted" to include the antiresonant part in the BSE.

Thanks for your suggestions. What factors were considered in making these modifications?

[Daniele Varsano]

Deaer Xiangjun,

you are right, at KS level yambo sees a gap. Please note, the KS energies are read from the QE calculation, and they should be the same besides an energy shift. Please check the QE output. If crossing are not seen and should be there, try to increase the k point sampling.

However, after GW correction, the energy ordering of the conduction band minimum (CBM) and valence band maximum (VBM) at the Gamma point is reversed, indicating that the system has opened a bandgap.

This does not seem to be the case, as the GW electronic structure is seen as a metal:

Code: Select all

  [X+QP] Filled Bands                               :   50
  [X+QP] Metallic Bands                             :   51   54
  [X+QP] Empty Bands                                :    55   720

Please, you are including the QP energy also in the screening, which is quite unusual unless you mean to do a sort of eigenvalue self-consistency,

Does Yambo distinguish bands solely based on energy values

Yes, you can do wannier interpolation, but not at this step of the calculation.

What factors were considered in making these modifications?

This is the default and assures the correct time-ordering.

Best,

Daniele