GW band structure exciton binding energy

[damao4361556]

Hi Everyone

I am new to Yambo.
I want to calculate GW band structure and exciton binding energy, could anyone suggest me the general procedures to do such calculations?

Sometimes I am confused on runlevels of Yambo.

thanks

--
Eric Lee
University of Southern California

[Daniele Varsano]

Dear Eric,
the best way to warm up with yambo is to start with the tutorials.
You can find here a step by step example for GW calculations:
http://www.yambo-code.org/tutorials/GW/index.php

Bethe Salpeter are in different tutorial, but you can start here:
http://www.yambo-code.org/tutorials/fan ... ns/LiF.php

If you encounter problems or something is not clear you can post here in the forum.

Best,
Daniele

[damao4361556]

Hi Daniele

Thanks for your kind reply.
I am going through the tutorials. But I am unclear about the following points:

1) When we calculated the QP energy with the single-shot G0W0 approxmation, we need to do the test of convergence with respect to
K-point, EXXRLvcs, BndsRnXs, and NGsBlkXs.
But when we turn to the optical properties within the regime of Bethe-Salpeter equation, do we need to test the convergence again, regarding the variables of [05] and [06] steps (i.e., BSEBands) in this tutorials http://www.yambo-code.org/tutorials/fan ... ns/LiF.php

2) I am interested in the GW band structure. From tutorial 1, I know that the QP energy correction at every k point and corresponding bands can be extracted with respect to the LDA energy, i.e.
# K-point Band Eo E-Eo Sc(Eo)
1.00000 1.00000 -12.07289 -1.45848 4.03804
1.000000 2.000000 0.000000 0.663829 2.859195
1.000000 3.000000 0.000000 0.663829 2.859195

At here E-E0 is the QP energy correction, this value will be further used for the calculation of BSE (obtained the optical spectrum), right? This is analogous to the scissor operator approach (by setting the XfnQP_E variable), and in yambo we can only implement this technology to open the LDA gap or shift the onset of the absorption peak?

3) The exciton binding energy is defined by QP band gap (Eq) minus the optical band gap (Eo). Here Eq is the direct or indirect band gap after GW correction? Eo is the energy of the lowest energy absorption or strongest absorption peak?

I am sorry I have so many stupid questions. I have these questions after I tried the examples in the tutorials and went through the Yambo form. Your answer will help me get the correct results with the usage of YAMBO code.

thanks in adavnce

Best

Dear Eric,
the best way to warm up with yambo is to start with the tutorials.
You can find here a step by step example for GW calculations:
http://www.yambo-code.org/tutorials/GW/index.php

Bethe Salpeter are in different tutorial, but you can start here:
http://www.yambo-code.org/tutorials/fan ... ns/LiF.php

If you encounter problems or something is not clear you can post here in the forum.

Best,
Daniele

[Daniele Varsano]

Dear Eric,

1) Yes, you need to perform convergence test on the bands to be included in the BSE, from that number depend the size of the BSE matrix so usually one starts with few conduction and valence bands and see when the spectrum converges by adding bands. Also you need to converge BSENGexx (this is not expensive) and BSENGBlk (this governs the screening and usually one has an idea from the previous GW calculation).

2)

At here E-E0 is the QP energy correction, this value will be further used for the calculation of BSE (obtained the optical spectrum), right? This is analogous to the scissor operator approach (by setting the XfnQP_E variable), and in yambo we can only implement this technology to open the LDA gap or shift the onset of the absorption peak?

Not exactly:
E-E0 is the QP energy correction. In the BSE you can either use the scissor operator by using KfnQP_E, please note the K and not the X, in order to apply to the kernel of the BSE and not in the response function for the calculation of the screening, but you can insert the QP energies previously calculated by using the variable:
KfnQPdb setting it as:

Code: Select all

KfnQPdb= " E < ./SAVE/ndb.QP"   

in this way yambo will consider the QP energies stored in the ndb.QP file calculated previously (the directory SAVE is the default, but it is different if you used the -J option). If you are interested to see directly the QP energy and not the correction they are also reported in the report file.

3)

Here Eq is the direct or indirect band gap after GW correction?

When talking about optics (q~0) I would say the direct gap, while in EELS experiment I would refer to the transmitted q. Anyway this is question of definition and you can find different way to define it in literature.

Eo is the energy of the lowest energy absorption or strongest absorption peak?

Each excitations has is binding energy, so the lowest one will have the higher binding energy, than it can be bright or "dark" (very small or zero intensity). The strongest absorption then we will have its own binding energy.


Best,

Daniele

[damao4361556]

Hi Daniele

A small question about the exciton binding energy (Eb).

Eb = Eq - Eo

where Eq is the GW corrected band gap, Eo is the optical band gap.

After BSE calculation, we could get the list of the energies correspond to each absorption peak, so we can use the excitation energies from the list as the Eo, right?

I just want to be for sure this point.

thanks

[Daniele Varsano]

Dear Eric,

, but what is the difference of col. 2-3 and col 4-5 and col 6-7.

Yes. In order to have that list you need to diagonalize the BSE hamiltonian (-y d) and then use the ypp tool.

Daniele

[damao4361556]

Dear Daniele

Regarding the optical gap - if we want to use it to calculate the exciton binding energy - we should use the excitation energy after the QP correction during the BSE spectrum step (i.e., apply the scissor operator as shown in the tutorial), right? So that the excitation energy can match the experimental value as much as possible.

I am sorry but I just want to make sure this point.

thanks

[myrta gruning]

Dear Eric

For the exciton binding energy you can plot the BS spectrum (col 1:2) versus the QP-RPA spectrum (col 1:4 same file).
The latter has the offset at the fundamental (direct) band gap as no excitonic effects are included. comparing the offset of the two spectra give an estimate of the exciton dbinding energy. In general you should compare the position of the bound exciton peak (from the spectrum you compute within GW+BSE) with the (direct) fundamental band gap (that you compute e.g. within GWA).

Hope this helps,
Myrta

[deliciousnai]

Dear All

I tried to calculate GW on GaN with yambo -g n -p p (I used yambo ver 4.1.3), sadly I always got error message in my terminal:

Code: Select all

--------------------------------------------------------------------------
mpirun noticed that process rank 2 with PID 11143 on node xxx exited on signal 9 (Killed).
--------------------------------------------------------------------------

and sometimes i got error like this in log file:

Code: Select all

 <05s> [LA] SERIAL linear algebra
[ERROR] STOP signal received while in :[05] Dynamic Dielectric Matrix (PPA)
[ERROR] Error allocating WFs

or:

Code: Select all

 <36s> [DIP] Checking dipoles header
[ERROR] STOP signal received while in :[05] Dynamic Dielectric Matrix (PPA)
[ERROR][NetCDF] NetCDF: Variable not found

I tried to change EXXRLvcs, FFTGvecs and %QPkrange but nothing workout, I wonder if there's something wrong with my pwscf or my yambo input?
I attached my input and output bellow.

Best,

Naima

[Daniele Varsano]

Dear Naima,
most probably your error are due to the fact you are running out-of-memory.
In your input I can see that you are calculating a lot of quasiparticle corrections in a single run (7x64=448):

Code: Select all

%QPkrange                    # [GW] QP generalized Kpoint/Band indices
  1|7|  1| 64|
%

This variable control the bands and k points you need to evaluate the GW corrections and usually people is interested in bands around the Fermi level. Are you sure you are interested in the GW corrections of the first deep bands and high energy bands?
I suggest you to reduce to few bands around the Fermi level, e.g. :
%QPkrange # [GW] QP generalized Kpoint/Band indices
1|7| 31| 34|
%
for calculating two occupied and two empty bands.
In case you need all the bands you can calulate them in different runs.

Beside that, please note:

Code: Select all

NGsBlkXp= 1 RL 

this is a parameter that need to be converged, you are running with just 1 RL which is totally unconverged.

Code: Select all

% BndsRnXp
   1 |  64 |                 # [Xp] Polarization function bands
%

64 bands over 32 occupied could be not at convergence.
I invite you to take your time and follow the tutorial in order to perform a menaningful calculation. A new step-by-step tutorial on GW can be found here:

http://www.yambo-code.org/wiki/index.ph ... rial:_h-BN

Now you are running in serial, if you have the possibility to run in parallel, beside the reduction of the time-to-solution you will benefit also from memory distribution.

Best,

Daniele