Yambo switching spin up and spin down for BSE

[kli103]

Dear developers,

I used qe 6.1.0 and yambo 4.4.0 to do GW+BSE@PBE0. When doing BSE calculation, I encountered the problem that BSE spectra differed a lot in the following two cases:

1. yambo reading ndb.QP for GW+BSE;
2. using scissor operator that is obtained by subtracting QP correction of VBM from QP correction of CBM, e.g., KfnQP_up_E = QPC(CBM_up) – QPC(VBM_up), KfnQP_dn_E = QPC(CBM_dn) – QPC(VBM_dn), where QPC stands for quasi-particle correction.

In looking for the possible issues that caused the problem, I found that yambo switched the spin up and spin down QP energies in BSE calculation from BSE report and BSE output.

For example, in terms of the triplet state system I am doing, from GW calculation output "o-all_Bz.qp" I get

Code: Select all

#    K-point      Band         Eo           E-Eo         Sc|Eo        Spin_Pol
      1.0000     145.0000      -3.8606       0.8036       2.2319       1.0000
      1.0000     145.0000      -3.7895       0.7925       2.2186      -1.0000
      1.0000     146.0000      -2.2941       1.1091       2.1889       1.0000
      1.0000     146.0000       2.4610       1.1303      -1.8933      -1.0000
      1.0000     147.0000      -2.0422       0.9186       1.7527       1.0000
      1.0000     147.0000       2.7222       1.5460      -2.2168      -1.0000
      1.0000     148.0000       2.0419       1.1566      -1.5933       1.0000
      1.0000     148.0000       3.0862       0.8122      -1.4614      -1.0000

where 148↑ is spin up CBM, 147↑ spin up VBM, 146↓ spin down CBM and 145↓ spin down VBM.

Code: Select all

Then KfnQP_up_E = 1.1566 - 0.9186 = 0.2380 eV and KfnQP_dn_E = 1.1303 - 0.7925 = 0.3378 eV. 

However, in case 1 where yambo read ndb.QP, BSE report "r-x_optics_bse_bsk_bss_rim_cut" shows

Code: Select all

[05.01] External/Internal QP corrections
========================================
...
  = FIT report for E<./all_Bz/ndb.QP[ E-E PPA E-E PPA: 27.21138 XG:  7591 Xb:    1  1000 Scb:    1  1000] =
  = First column (conduction), Second column (valence) =

  = Spin DN ============================================
   Gap correction (database) [ev]: 0.234857
               (FIT, actual) [ev]:  0.06039
   Energies 0th order        [ev]: 1.316240  1.020997
   Energies 1st order            : 0.881223  1.124789
                            Error:   0.0528    0.0071

  = Spin UP ============================================
   Gap correction (database) [ev]: 0.338145
               (FIT, actual) [ev]:  -0.0037
   Energies 0th order        [ev]: 1.164239  0.829788
   Energies 1st order            : 0.828994  1.561357
                            Error:  0.04216   0.05219

We can see that the gap correction for spin up and spin down are switched compared with the KfnQP_up_E and KfnQP_dn_E calculated above. And if looking at the BSE output file of spin up "o-x.-all_Bz-spin-up.data", we can find at line 57 and line 58,

Code: Select all

#    Eo [eV]      DeltaE [eV]
   -3.789545     0.792506
    2.460562     1.131596

These values are actually of VBM (145↓) and CBM (146↓) of spin down.

Please kindly take a look at this problem. Thank you.

Best regards,
Kejun

[Daniele Varsano]

Dear Kejun,
thanks for reporting, we will have a look at this issue as soon as possible and let you know.

Best,
Daniele

[Daniele Varsano]

Dear Kejun,
I just realized you are using the 4.4 release which is quite old.
Can you try to update to a newer release of Yambo and see if the problem persists. If so, we will check the problem in deep.
Best,
Daniele

[kli103]

Dear Daniele,

Yes. I will update yambo and try to do the same calculation again. I will let you know once I get results.

Best,
Kejun

[kli103]

Dear Daniele,

I used yambo 4.5 to do GW+BSE@PBE0 by reading ndb.QP. The issue still exists. Please see the details in the following.

In the GW output,

Code: Select all

#    K-point      Band         Eo           E-Eo         Sc|Eo        Spin_Pol
      1.0000     145.0000      -3.8606       0.7765       2.2027       1.0000
      1.0000     145.0000      -3.7895       0.7651       2.1892      -1.0000
      1.0000     146.0000      -2.2941       1.1091       2.1915       1.0000
      1.0000     146.0000       2.4610       1.1204      -1.9043      -1.0000
      1.0000     147.0000      -2.0422       0.9188       1.7670       1.0000
      1.0000     147.0000       2.7222       1.5279      -2.2299      -1.0000
      1.0000     148.0000       2.0419       1.1453      -1.6057       1.0000
      1.0000     148.0000       3.0862       0.7972      -1.4774      -1.0000

so KfnQP_up_E = 1.1453 - 0.9188 = 0.2265 eV and KfnQP_dn_E = 1.1204 - 0.7651 = 0.3553 eV.

However, in the BSE report "r-y_optics_bse_bss_dipoles_rim_cut",

Code: Select all

  [05.01] External/Internal QP corrections
  ========================================
  ...
  = Spin DN ============================================
   Gap correction (database) [ev]: 0.224908
               (FIT, actual) [ev]:  0.01708
   Energies 0th order        [ev]: 1.264105  1.022116
   Energies 1st order            : 0.844842  1.137899
                            Error:   0.0982    0.0072

  = Spin UP ============================================
   Gap correction (database) [ev]: 0.355695
               (FIT, actual) [ev]: 0.152204
   Energies 0th order        [ev]: 1.306532  0.798633
   Energies 1st order            : 0.469601  1.524406
                            Error:  0.08890   0.04830

In the BSE output "o-y.-all_Bz-spin-up.data", at line 58 and 59,

Code: Select all

#    Eo [eV]      DeltaE [eV]
   -3.789545     0.765113
    2.460562     1.121890

The results from BSE report and output are reverse of those from GW.

Best,
Kejun

[Daniele Varsano]

Dear Kejun,
thanks for reporting, we will inspect this soon.

Daniele

[Davide Sangalli]

Dear Kejun,
can you try the patch you find in this post for version 4.5 viewtopic.php?p=8565#p8565

For version 4.4 it is instead attached here.

Best,
D.

[kli103]

Dear Davide,

Yes, I can try the patch. Could you tell me the reason of trying the patch?

Best,
Kejun

[Daniele Varsano]

Dear Kejun,
the reason is that this could fix the problem, as there were some problems in spin-polarized calculations when reading the QP databases, which are solved applying this patch (see the thread indicated by Davide). If the problem still persists please let us know.

Daniele

[kli103]

Dear Daniele and Davide,

I modified yambo-4.4.0 codes and tested. The issue is still there. Spin up and spin down data switch. Corrected gap information in section [05.01] External/Internal QP corrections of BSE report is correct after code modification.

[K]Indirect Gaps [ev]: 4.317925 4.333777
[K] [spin up]: 4.317925 4.333777
[K] [spin dn]: 6.588544 6.822598
[K]Direct Gaps [ev]: 4.318357 4.338337
[K] [spin up]: 4.318357 4.338337
[K] [spin dn]: 6.588544 6.830115

The attached files contain diff between code before and after the modification.

I am testing yambo-4.5.0. I will tell result to you as soon as it is done.

Best,
Kejun