Page 1 of 1
Band reversal in GW
Posted: Thu Jul 14, 2022 7:25 am
by sdwang
Dear all,
I find a band reversal in the GW results compared with the PBE ones. My system is 2D with a direct PBE gap of 0.008303 eV(at the k point of 49). And the GW results are as follows:
---
QP [eV] @ K [49] (iku): -0.500000 -0.500000 0.000000
B=60 Eo= -2.36 E= -2.53 E-Eo= -0.17 Re(Z)=0.75 Im(Z)=-0.214399E-2 nlXC= -22.5032 lXC= -18.0008 So= 4.27851
B=61 Eo= -2.34 E= -2.27 E-Eo= 0.07 Re(Z)=0.75 Im(Z)=-0.117164E-2 nlXC= -16.8008 lXC= -15.0595 So= 1.83684
B=62 Eo= -1.93 E= -1.90 E-Eo= 0.03 Re(Z)=0.75 Im(Z)=-0.205785E-2 nlXC= -17.4891 lXC= -15.2494 So= 2.27820
B=63 Eo= 0.00 E= 0.66 E-Eo= 0.66 Re(Z)=0.76 Im(Z)=-0.213259E-2 nlXC= -12.8546 lXC= -15.1054 So= -1.38438
B=64 Eo= 0.01 E= 0.46 E-Eo= 0.45 Re(Z)=0.76 Im(Z)=-0.204936E-2 nlXC= -12.0753 lXC= -14.2854 So= -1.62188
B=65 Eo= 0.05 E= 0.76 E-Eo= 0.71 Re(Z)=0.76 Im(Z)=-0.207785E-2 nlXC= -12.6966 lXC= -15.3515 So= -1.72120
B=66 Eo= 1.79 E= 2.72 E-Eo= 0.93 Re(Z)=0.76 Im(Z)=-0.147795E-2 nlXC= -8.64970 lXC= -13.6453 So= -3.77699
B=67 Eo= 2.03 E= 2.96 E-Eo= 0.93 Re(Z)=0.76 Im(Z)=-0.162771E-2 nlXC= -8.69498 lXC= -13.6545 So= -3.73556
---
I think the gap should be (0.66-0.46) but here the VBM is larger than the CBM. And consequently, the following BSE process reads the gap from ndb.QP will appear:
How should we fix the problem?
Thanks!
Shudong
Re: Band reversal in GW
Posted: Mon Jul 18, 2022 9:06 am
by Davide Sangalli
Dear Shudong,
the number of filled bands and the number of metallic bands is defined according to the bands ordering. This is why you get the message.
However this is not an issue in the simulations, just of the way yambo stores info about the two above mentioned number.
This means that you get a warning, but all the occupations are correctly computed.
Indeed, when yambo evaluate the Fermi function for the occupations it sorts all the eigenvalues with respect to their energy, regardless of the band, spin and k-point index.
As a consequence, despite the warning message, the BSE simulations should be fine.
Indeed, also at the BSE level, transition are selected depending on the occupations.
Best,
D.
Re: Band reversal in GW
Posted: Mon Jul 18, 2022 9:24 am
by sdwang
Dear Davide,
In the report file of bse calculation I can find:
Code: Select all
[02.05] Energies & Occupations
==============================
[X] === General ===
[X] Electronic Temperature : 0.000000 0.000000 [eV K]
[X] Bosonic Temperature : 0.000000 0.000000 [eV K]
[X] Finite Temperature mode : no
[X] El. density : 0.79702E+23 [cm-3]
[X] Fermi Level : -2.338813 [eV]
[X] === Gaps and Widths ===
[X] Conduction Band Min : 0.008303 [eV]
[X] Valence Band Max : 0.000000 [eV]
[X] Filled Bands : 63
[X] Empty Bands : 64 600
[X] Direct Gap : 0.008303 [eV] <-----
[X] Direct Gap localized at k : 49
[X] Indirect Gap : 0.008303 [eV]
[X] Indirect Gap between kpts : 49 49
[X] Last valence band width : 1.849351 [eV]
[X] 1st conduction band width : 1.311757 [eV]
and the QP report as:
Code: Select all
[06.01] QP corrections report
==============================
[K+QP] === General ===
[K+QP] Electronic Temperature : 0.000000 0.000000 [eV K]
[K+QP] Bosonic Temperature : 0.000000 0.000000 [eV K]
[K+QP] Finite Temperature mode : no
[K+QP] El. density : 0.79702E+23 [cm-3]
[K+QP] Fermi Level : 0.456814 [eV]
[K+QP] === Gaps and Widths ===
[K+QP] Conduction Band Min : 0.456814 [eV]
[K+QP] Valence Band Max : 0.456814 [eV]
[K+QP] Filled Bands : 62
[K+QP] Metallic Bands : 63 64 <--------
[K+QP] Empty Bands : 65 600
[K+QP] === Metallic Characters ===
[K+QP] N of el / N of met el : 126.0000 2.0000
[K+QP] Average metallic occ. : 0.250000
[WARNING][K+QP] Metallic system <-----
in which without the QP_gap and the system is metallic which is actually a semiconductor. And also in the abs. the whole spectrum moves to the lower energy range due to the metallic warning.
Thanks!
Shudong
Re: Band reversal in GW
Posted: Mon Jul 18, 2022 10:43 am
by Davide Sangalli
Yeah,
the report is as I would expect.
The first part refers to the KS energies used in building the screening.
The second part to the QP energies used to build the excitonic matrix where the message is wrong (and also the computed gap).
Can you attach the output of the bse simulation (.alpha file)?
Best,
D.
Re: Band reversal in GW
Posted: Tue Jul 19, 2022 8:35 am
by sdwang
Dear Davide,
Attached are the alpha files. The 01 is the result used ndb.QP energy to do the BSE calculation and for result 02, I inserted the energy difference between PBE and GW in the option "% KfnQP_E". The two results are totally different.
Thanks!
ps: I also found the 01 results the 2nd and the 3rd data are almost identical which should be different since the 2nd is the BSE while the 3rd is the GW+IP result, as shown in the 02 results.
Best,
Shudong
Re: Band reversal in GW
Posted: Tue Jul 19, 2022 10:29 am
by Davide Sangalli
This is what I get by plotting the results.
BSE_plots.png
It seems that the issue is due to the dipoles.
As you see overall the spectrum is reasonable.
However there is one dipole which is largely overestimated. Since top vance and bottom conduction are inverted, my best guess is that this activates a new possible transition (likely old bottom conduction -> conduction).
The energy difference at PBE level between these two states is 40 meV and this could give issue in the computed dipoles which result from "<v>/delta epsilon".
The energy difference at GW+PBE is at around 400 meV, which is more or less there the transition is from the plot.
I would have expected similar issues in the PBE-IP spectrum from (old top valence -> old bottom conduction), but likely such transition is dark and the 8 meV energy difference is not giving particular issues.
To confirm that this is the problem you can try the following:
(i) switch to the velocity gauge
(ii) Compute also the excitonic jdos
Anyway, I'd try to carefully understand what these states are, and why they are inverted at PBE level
Best,
D.