GW+BSE optics

[burkzdemir]

Dear Yambo developers,

I runned GW+BSE calculation with "yambo -o b -k sex -y d -p c". However, calculation did not return dielectric function? What should be the run level for the dielectric function of GW+BSE?

[Daniele Varsano]

Dear Burak ,
please post your input and report/output files otherwise it is practically impossible to help you.
By the way, the correct command line to build up a BSE input is for instance:
yambo -o b -k sex -y d

while for a GW calculation is
yambo -p p -g n

Best,

Daniele

[burkzdemir]

Results are in the attachments. My input is;

cohsex # [R Xp] COlumb Hole Screened EXchange
gw0 # [R GW] GoWo Quasiparticle energy levels
HF_and_locXC # [R XX] Hartree-Fock Self-energy and Vxc
em1s # [R Xs] Static Inverse Dielectric Matrix
EXXRLvcs= 33721 RL # [XX] Exchange RL components
Chimod= "hartree" # [X] IP/Hartree/ALDA/LRC/PF/BSfxc
% BndsRnXs
13 | 23 | # [Xs] Polarization function bands
%
NGsBlkXs= 1 RL # [Xs] Response block size
% DmRngeXs
0.150000 | 0.150000 | eV # [Xs] Damping range
%
% LongDrXs
0.000000 | 1.000000 | 0.000000 | # [Xs] [cc] Electric Field
%
%QPkrange # [GW] QP generalized Kpoint/Band indices
1|10| 13| 23|
%

[Daniele Varsano]

Dear Burak,
this is totally fine.
You have quasi particle calculation in the cohsex approximation and you have the qp results in the output.
Now, if you want to calculate a BSE spectra you need to prepare your input (yambo -o b -k sex -y d) and run it.
You cannot do a GW and BSE calculation using the same input. For a correct use of the code I invite you to follow the tutorials you can find here:
http://www.yambo-code.org/wiki/index.ph ... =Tutorials

Best,
Daniele

[burkzdemir]

I runned "yambo -o b -k sex -y d" but the code did not return the dielectric function. I attached the setup and log file.

[Daniele Varsano]

Dera Burak,
the code stopped. I do not know if for memory reason or something else.
Try to see if you have an error message in one of your logs file. I cannot say more as the input file is not present.

Best,
Daniele

[burkzdemir]

I am getting segmentation fault error (core dumped) when I run on single processor.

[Daniele Varsano]

Dear Burak,
ok, this is very generic,
but at what point of the calculation?
What's your input file?
If you do not provide all the information that can help to spot the problem it is practically impossible to help you.

Best,

Daniele

[burkzdemir]

I am using yambo 4.3.1. I just runned "yambo -o b" and I got segmentation error again. Following is my input;

optics # [R OPT] Optics
bse # [R BSE] Bethe Salpeter Equation.
BSEmod= "retarded" # [BSE] resonant/retarded/coupling
BSKmod= "IP" # [BSE] IP/Hartree/HF/ALDA/SEX
% BEnRange
1.00000 | 5.00000 | eV # [BSS] Energy range
%
% BDmRange
0.150000 | 0.150000 | eV # [BSS] Damping range
%
BEnSteps=0 # [BSS] Energy steps
% BLongDir
0.000000 | 1.000000 | 0.000000 | # [BSS] [cc] Electric Field
%
% BSEBands
13 | 23 | # [BSK] Bands range
%

And I get the error at the following stage;

<---> [01] CPU structure, Files & I/O Directories
<---> [02] CORE Variables Setup
<---> [02.01] Unit cells
<---> [02.02] Symmetries
<---> [02.03] RL shells
<---> [02.04] K-grid lattice
<---> [02.05] Energies [ev] & Occupations
<01s> [03] Transferred momenta grid
<01s> [RL indx] Q-grid:User defined / from ndb.kindx.
<43s> X indexes |########################################| [100%] 41s(E) 41s(X)
<47s> SE indexes |########################################| [100%] 03s(E) 03s(X)
<48s> [04] Response Functions in Transition spaceSegmentation fault (core dumped)

Is there something wrong with my compilation?

[Daniele Varsano]

Dear Burak,
I cannot say much, but it could be a memory issue. You have 3456 k points in the full BZ

Code: Select all

  X BZ K-points :  3456

and you are using 6 valence and 5 conduction bands, so the BSE matrix it consists of Ndim=NK x Nc x Nv= 103680.
It is a quite large matrix as it consists of Nd(Nd+1)/2 elements, consisting roughly in 80GB of memory (if I did the math right).
Try to see if this is the case by reducing the % BSEBands range. If this is the case next you can choose a parallelization strategy
to distribute the memory load.

Best,
Daniele