Problem with GW+BSE

[liusjtc]

Dear Yambo developers,

I have see many topics but I still have some problems about GW+bse :<
I do the following steps:
1.a2y
2.yambo
3.yambo -i -V RL
4.yambo -d -k hartree -g n -p p
5.yambo -b
6.yambo -o b -k sex -y h -V qp (change the KfnQPdb=“none” to KfnQPdb=“E<./SAVE/ndb.QP”)

I find the result I get is the same without interpolate the QP .
Then I change the KfnQP_E=|0|1.00|1.00| to |1.00|1.00|1.00|,and I get the result seems interpolating the QP.(the result stay still when I change the E=|0|1.00|1.00| to |3.00|1.00|1.00|)

When I change the BLongdir to 010,I fail again. The consequence is same without the QP.
I don’t understand the meaning of KfnQP_E well
I would be thankful if tell me about the changes do I need for to calculate the GW+BSE

Best
Liu Shijie
Shanghai JiaoTong University

[Daniele Varsano]

Dear liusjtc,
please fill your signature with your affiliation: this is a rule of the forum.

Next honestly it is not very clear what is the problem you are facing:
KfnQP_E is a three value variable: the first one is the scissor operator, ie the qp correction you want to assign to all conduction bands. The other two permits you to assign a stretching as it often happens that the qp correction can be fitted in a linear form.

If something went wrong when reading QP databases or in assigning the scissor, please post the report file as without inspecting them it is impossible to understand your problem.

Best,
Daniele

[liusjtc]

Dear Dniele Varsano

Thank you
Please tell me if there is something wrong when reading QP databases

[Daniele Varsano]

Dear Liu,
The QP correction introduced by the scissor are correctly read as you can see from the report of the 3rd run:

Code: Select all

[06] External QP corrections (K)
 ================================

 -- [QP@K] Linear/Quadratic extrapolation -------------------
 === QP/DB derived ===
 === Additional ===
  Energies 0th order c/v [ev]: 3.000000  0.000000
  Energies 1st order         : 1.000000  1.000000
 ------------------------------------------------------------
 [QP] Fermi level variation [ev]: 0.000000
 [QP] Last Filled/Metallic band :  32   32
 [QP] Ind. Gap Correction  [ev]: 3.000000

The QP database it is not read, instead. I assume you have the ndb.QP file in the SAVE directory (so you did not use the -J option in the GW calculation).
May be you need to separate with a space the description of the variable:

Code: Select all

E < ./SAVE/ndb.QP

Anyway consider the following:
1) QP database and scissor operator are somehow complementary, you want to apply QP correction and you can by inserting the QP values previously calculated or alternatively use a scissor approximation when it applies. I cannot see the point to introduce both.

2) Next from your report I can see that you are not calculating the BSE hamiltonian in the different case, but reading a previous one or restarting an Haydock iteration:

Code: Select all

[RD./SAVE//ndb.BS_Q1]---

Code: Select all

[RD./SAVE//ndb.Haydock_restart]--------------

pay attention to that. If you want to do a *new* calculation, either move/rename the ndb.BS_Q1 databases or use the -J dirname in order to collect databases in different directories.

Best,
Daniele

[liusjtc]

Dear Daniele Varsano

Thanks for your patient. I have solved my problems successfully.
But I still get confused again with my results.

I know that GW expend the band gap with slightly changing the shape of energy bands.
But my results seem a little different:
1.The refractive index of GW+Bse curve is lower than the Bse curve
2.The 010 direction and the 001 direction expend less than the 100 direction. (LongDrXs=111,BLongDir=100,010,001)

Best,
Liu

[Daniele Varsano]

Dear Liu,
I do not know the details of your system and of your calculations.
Anyway:

1.The refractive index of GW+Bse curve is lower than the Bse curve

This is totally consistent, GW open the gap so the spectra are blueshifted with respect using KS energies in the BS equation.

2.The 010 direction and the 001 direction expend less than the 100 direction. (LongDrXs=111,BLongDir=100,010,001)

This is indication that your system is anysotropic, it is not the case?, why do you expect similar response?

Best,
Daniele