full GW+BSE calculation

[cantele]

Hi all,

my issue is in running a full GW+BSE calculation. I want the BSE calculation to reade the quasiparticle corrections
computed within GW approximation, to check the validity of the "scissor operator" shift of the BSE spectrum for
a system under study.

The error I get is as follows:

[ERROR] STOP signal received while in :[07] Bethe-Salpeter Kernel
[ERROR] Null BSE kernel block dimension(s) found. Increase e/h range

I guess I didn't calculate quasiparticle corrections for all the states for which they are needed.

Can you give me some advice on how to perform this kind of test?

The input for the GW calculations is the following:

gw0 # [R GW] GoWo Quasiparticle energy levels
ppa # [R Xp] Plasmon Pole Approximation
HF_and_locXC # [R XX] Hartree-Fock Self-energy and Vxc
em1d # [R Xd] Dynamical Inverse Dielectric Matrix
EXXRLvcs= 2999 RL # [XX] Exchange RL components
% QpntsRXp
1 | 36 | # [Xp] Transferred momenta
%
% BndsRnXp
1 | 348 | # [Xp] Polarization function bands
%
NGsBlkXp= 250 RL # [Xp] Response block size
% LongDrXp
1 | 0 | 0 | # [Xp] [cc] Electric Field
%
PPAPntXp= 27.21138 eV # [Xp] PPA imaginary energy
% GbndRnge
1 | 348 | # [GW] G[W] bands range
%
GDamping= 0.1 eV # [GW] G[W] damping
dScStep= 0.10000 eV # [GW] Energy step to evalute Z factors
DysSolver= "n" # [GW] Dyson Equation solver (`n`,`s`,`g`)
%QPkrange # [GW] QP generalized Kpoint/Band indices
1 | 36 | 160 | 189 |
%
#%QPerange # [GW] QP generalized Kpoint/Energy indices
# 1| 36| 0.0|-1.0|
#%

whereas the input of the BSE calculation which is reading from the SAVE directory where the
GW run wrote is:

optics # [R OPT] Optics
bse # [R BSK] Bethe Salpeter Equation.
em1s # [R Xs] Static Inverse Dielectric Matrix
bss # [R BSS] Bethe Salpeter Equation solver
em1d # [R Xd] Dynamical Inverse Dielectric Matrix
ppa # [R Xp] Plasmon Pole Approximation
BSresKmod= "xc" # [BSK] Resonant Kernel mode. (`x`;`c`;`d`)
BScplKmod= "none" # [BSK] Coupling Kernel mode. (`x`;`c`;`d`;`u`)
% BSEBands
160 | 189 | # [BSK] Bands range
%
BSENGBlk= 250 RL # [BSK] Screened interaction block size
BSENGexx= 2999 RL # [BSK] Exchange components
% QpntsRXs
1 | 36 | # [Xs] Transferred momenta
%
% BndsRnXs
1 | 348 | # [Xs] Polarization function bands
%
NGsBlkXs= 250 RL # [Xs] Response block size
% LongDrXs
1.000000 | 0.000000 | 0.000000 | # [Xs] [cc] Electric Field
%
% QpntsRXp
1 | 36 | # [Xp] Transferred momenta
%
% BndsRnXp
1 | 348 | # [Xp] Polarization function bands
%
NGsBlkXp= 253 RL # [Xp] Response block size
% LongDrXp
1.000000 | 0.000000 | 0.000000 | # [Xp] [cc] Electric Field
%
PPAPntXp= 27.21138 eV # [Xp] PPA imaginary energy
BSSmod= "d" # [BSS] Solvers `h/d/i/t`
% BEnRange
0.00000 | 8.00000 | eV # [BSS] Energy range
%
% BDmRange
0.10000 | 0.10000 | eV # [BSS] Damping range
%
BEnSteps= 500 # [BSS] Energy steps
% BLongDir
1.000000 | 0.000000 | 0.000000 | # [BSS] [cc] Electric Field
%
KfnQPdb= "E < SAVE/ndb.QP"

[Daniele Varsano]

Dear Giovanni,

it looks that the bands range for calculating the BSE gives you 0 dimension in the kernel,
as you are not including unoccupied states. Have a look at the band structure after
QP calculations. May be something went wrong in the GW step.

Just an hypothesis,
Cheers,

Daniele

[cantele]

Because the error message says "Increase e/h range" I checked within r_optics_bse_em1s_em1d_ppa_bss_02 and found the following line:

X e/h E range [ev]:-1.000000 -1.000000

Has it to do with the error message?

Is it possible that the problem rises from the fact I used, for the quasiparticle calculation

%QPkrange # [GW] QP generalized Kpoint/Band indices
1 | 36 | 160 | 189 |
%

whereas the variable QPerange is left commented?

What is the variable controlling the e/h range?

[Daniele Varsano]

Hi Giovanni,

that's not should be problem,
QPkrange or QPerange are different way to choose the number of bands you want
to correct, by bands indexes or by energy range.
The e/h range is just the min. et max. differences between you occupied and
unoccupied states you are including in the calculation, so you can control it
by changing the BSEBands.
The error you get looks like
you are not including unoccupied states, so you have a zero dimension kernel.
It's quite strange what is happening.
Try to post the report file.

Cheers,

Daniele

[cantele]

that's not should be problem,
QPkrange or QPerange are different way to choose the number of bands you want
to correct, by bands indexes or by energy range.

I understand this, I was just wondering whether if the latter variable is not found
for some reason some variable is not updated in one of the databases.

The e/h range is just the min. et max. differences between you occupied and
unoccupied states you are including in the calculation, so you can control it
by changing the BSEBands.

OK, so I do calculate QP corrections to the SAME bands specified in BSEBands.

The error you get looks like
you are not including unoccupied states, so you have a zero dimension kernel.
It's quite strange what is happening.
Try to post the report file.

in attach the r_* and l_* for both the GW and the BSE calculation

[Daniele Varsano]

Dear Giovanni,
I had a look at the report file you post.
I can see that the problem is here. Once Yambo read the QP file,
and recalculate the Fermi level, you can see in the last lines of
the BSE report file:

Code: Select all

[QP] Fermi level variation [ev]: 5.176658
[QP] Last Filled/Metallic band : 192  196

so it looks it founds all the bands you included in the BSE calculation as occupied.
Now, I don't' know if Yambo is reading incorrectly the QP file, or it fails in the calculation
of the Fermi level, anyway I suggest you to plot the band structure, DFT vs GW, and look
if you have big variation in the band shape, causing swaps in the band indexes.
May be from that we can realize where the code is failing.

Cheers,

Daniele

[Khaled]

Hello developers,

I am using yambo 3.4.1 Revision 3187 and my system is organic-semiconductor, I have done BSE calculation by using BSEmod="casual" and BSKmod= "SEX" and everything went correctly without any errors, but after I changed BSEmod="coupling" and uncomment the flag WehCpl I got error message

Code: Select all

 [QP] Kpts covered exactly  [o/o]: 100.0000
 [QP] Fermi level variation [ev]:-0.472822
 [QP] Last Filled/Metallic band :  80   80
 [QP] Ind. Gap Correction  [ev]: 0.576690

  [08.01] Main loop
  =================


  [ERROR] STOP signal received while in :[08.01] Main loop

  [ERROR] Null BSE kernel block dimension(s) found. Increase e/h range

I have reduced the value of BSEEhEny , then EhEngyXs and then EhEngyXp and at each time I got the same error and as you said in the previous discussion in this page I should increase unoccupied bands of the parameter BSEBands, I did and I got the same error.

I have tried the same scenario with BSEmod="resonant" with commenting the flag WehCpl I also the same error message. I have attached the report files and input files for both GW and BSE resonant and coupling part. please I need your help.

I am waiting for your available suggestion. Thank you in advance.

Best regards,

[Khaled]

I have tried the same scenario with BSEmod="resonant" with commenting the flag WehCpl

I meant uncommenting the flag WehCpl. Sorry for that.

[Daniele Varsano]

Dear Khaled,
well this should not happen. Of course the variable to raise for increasing the e/h range is the BSE bands.
What I can see in your input is that you are assigning both

Code: Select all

% BSEBands
    40 |  120 |               # [BSK] Bands range
% BSEBands

and

Code: Select all

% BSEEhEny
0.000000 |5.000000 | eV    # [BSK] Electron-hole energy range
%

which are redundant, it is needed to specify just one of them. Now my suspicious is that the code get confused when adding the coupling because the in the coupling block the e-h energy is negative. So I would try to leave that variable as default: (-1,-1) so it will be ignored, all the Bands indicated in the BSEBands will be used (eventually you can lower them if you have memory problem) and see what happen. If the problem persist post it again.

Next, resonant mode and WehCpl does not make much sense, as it applies in the coupling part which is neglected in the resonant approximation.

Best,
Daniele

[Khaled]

Dr Daniele Varsano,

Thank you a lot. your suggestion works with me, as you said, I should set BSEEhEny = (-1,-1) to make the code ignore negative e\h in coupling block and also I have set BSEBands= 60 | 100 instead of 40| 120. but you know the usual problem with the diagnolization mod( BSSmod= "d") always about the dimension of the matrix and the main parameters which they are responsible for the dimension are BSEEhEny and BSEBands. I got the Dimension of the matrix =26880 and for that reason my calculation stopped at that stage. what should I do with respect to these two BSEEhEny in order to continue my calculation and without getting error due to dimension.

N.B : I do suspect Dr Daniele Varsano the problem now due to the dimension of the matrix, but I prefer to see the report file, may be something else cause this problem.

Thank you a lot Dr Daniele Varsano, I am waiting for your available suggestion.

Best regards,