Selection of parameter values for calculating the collision before the SHG

[sunxl]

Dear all:

1. I have already completed the BSE calculation, and now I want to calculate SHG with exciton effects. When calculating BSE, I tested for convergence and set the BSEBands to 5 valence bands and 2 conduction bands. Should the COLLBands be set to the same as the BSEBands when calculating the SHG collision? Are there any connections between these two parameters?

2、Should each of the HARRLvcs, EXXRLvcs, and CORRLvcs start at 1 Ry and gradually increase to test for convergence? Is convergence obtained by calculating shg?

I appreciate any guidance and tutorial on that!

Best,
sunxl

[claudio]

Dear Sunxl

>1. I have already completed the BSE calculation, and now I want to calculate SHG with exciton effects. When calculating BSE, I tested for convergence and set the BSEBands to 5 valence bands and 2 conduction bands. >Should the COLLBands be set to the same as the BSEBands when calculating the SHG collision? Are there any connections between these two parameters?

yes COLLBands should be the same of BSEBands

> 2、Should each of the HARRLvcs, EXXRLvcs, and CORRLvcs start at 1 Ry and gradually increase to test for convergence? Is convergence obtained by calculating shg?

Start with small values 1000 mRy and then increase it 1500 mRy, 2000 mRy etc.... (you have to use the same value in the dyanmics)
Use yambo_rt for the calculation of collisions because it is faster and then yambo_nl for the non-linear response.

If the full calculation is too expensive different approximation are available for the collission
but you have to test if they work in your case

https://wiki.yambo-code.eu/wiki/index.p ... r_response

best
Claudio

[sunxl]

Dear Claudio,

Thank you for your patient explanation.

When performing IPA-SHG calculations, I noticed that there are three parameters: BSEBands, HARRLvcs, and EXXRLvcs. However, for BSE-SHG calculations, there are four parameters: BSEBands, HARRLvcs, EXXRLvcs, and CORRLvcs.
If I have already performed convergence tests for the three parameters in IPA calculations (e.g., BSEBands = 3-6 and HARRLvcs = EXXRLvcs = 2Ha), can I directly set BSEBands = 3-6 and HARRLvcs = EXXRLvcs = CORRLvcs = 2Ha for BSE-SHG calculations? Or is it necessary to perform separate convergence tests for all four parameters in BSE-SHG calculations?

Additionally, for HARRLvcs, EXXRLvcs, and CORRLvcs, is it necessary to keep these parameters equal to each other, or can they be set independently?

Best,
sunxl

[claudio]

Dear Sunxl

> If I have already performed convergence tests for the three parameters in IPA calculations (e.g., BSEBands = 3-6 and HARRLvcs = EXXRLvcs = 2Ha),
> can I directly set BSEBands = 3-6 and HARRLvcs = EXXRLvcs = CORRLvcs = 2Ha for BSE-SHG calculations?
> Or is it necessary to perform separate convergence tests for all four parameters in BSE-SHG calculations?

you can be I advise you to make some test in BSE for example using few k-points to have fast results
for example you can set

HARRLvcs = EXXRLvcs = CORRLvcs = 2500 mHa
HARRLvcs = EXXRLvcs = CORRLvcs = 3000 mHa
HARRLvcs = EXXRLvcs = CORRLvcs = 3500 mHa
HARRLvcs = EXXRLvcs = CORRLvcs = 4000 mHa

best
Claudio

[sunxl]

Dear Claudio,
Thank you for your patient explanation.

For small k points of 6*6*1, when I select HARRLvcs = EXXRLvcs = CORRLvcs =1000,2000,3000, 4000 and 5000mHa respectively, I calculated collision first, and then calculated shg. I found that with the increase of HARRLvcs = EXXRLvcs = CORRLvcs, the obtained shg value becomes smaller and smaller. Why does shg become smaller with the improvement of accuracy?

I put my calculation results and output files in the attachment, hoping that you can help me check the problem.

Best,
sunxl

[claudio]

Dear Sunxl

I don't have a general answer, in general we also found that local-field effects (RPA)
decrease the SHG response, see Fig. 1 top panel in

https://journals.aps.org/prb/pdf/10.110 ... .89.081102

that is in agreement with what you found, while excitonic effect tends to increase it.
Just a question, did you use the cutoff in the z-direction?

best
Claudio

[sunxl]

Dear Claudio,

Thank you for your patient explanation.

1、First, I confirm that I have added cut-off to the input files for both collision and SHG. The output files for these two steps are also included in the attachment. When I was reviewing the output file from my collision calculations, I noticed that a parameter in the final input parameters section, namely DipBands, was not set in the input file and was automatically assigned a value of 80. Is this value obtained by reading the ndb.QP file? (In the quasiparticle correction, I selected BndsRnXp=80). Is there any relationship between these two? Could you please help me check if there are any errors in the input parameter settings and the output files?


Perhaps my explanation was unclear. When I calculate SHG with excitonic effects by adjusting the three parameters HARRLvcs, EXXRLvcs, and CORRLvcs, regardless of how they are set, the resulting SHG with excitonic effects is always larger than the SHG obtained using the IPA method. In other words, compared to IPA, excitonic effects indeed enhance SHG.

2、I used a 6×6×1 k-point grid to test the three parameters HARRLvcs, EXXRLvcs, and CORRLvcs. As these three parameters were gradually increased from 1000 mHa to 7000 mHa, the SHG values steadily decreased, and the SHG values obtained at 6000 mHa and 7000 mHa began to stabilize (the figure is included in the attachment). Does this indicate that convergence is achieved at 6000 mHa? What confuses me is that as the parameter precision increases, SHG should increase, shouldn’t it?

3、Additionally, I have found that when other settings remain the same and only the number of k-points is changed, the SHG with excitonic effects decreases as the number of k-points increases. I have the same confusion regarding this as mentioned above.


To accurately describe my issue, I might be a bit verbose. Thank you again for your patient assistance.

Best,
sunxl

[claudio]

Dear

1、First, I confirm that I have added cut-off to the input files for both collision and SHG. The output files for these two steps are also included in the attachment. When I was reviewing the output file from my collision calculations, I noticed that a parameter in the final input parameters section, namely DipBands, was not set in the input file and was automatically assigned a value of 80. Is this value obtained by reading the ndb.QP file? (In the quasiparticle correction, I selected BndsRnXp=80). Is there any relationship between these two? Could you please help me check if there are any errors in the input parameter settings and the output files?

do not bother of this parameter the code will automatic set it

Perhaps my explanation was unclear. When I calculate SHG with excitonic effects by adjusting the three parameters HARRLvcs, EXXRLvcs, and CORRLvcs, regardless of how they are set, the resulting SHG with excitonic effects is always larger than the SHG obtained using the IPA method. In other words, compared to IPA, excitonic effects indeed enhance SHG.

this is good

2、I used a 6×6×1 k-point grid to test the three parameters HARRLvcs, EXXRLvcs, and CORRLvcs. As these three parameters were gradually increased from 1000 mHa to 7000 mHa, the SHG values steadily decreased, and the SHG values obtained at 6000 mHa and 7000 mHa began to stabilize (the figure is included in the attachment). Does this indicate that convergence is achieved at 6000 mHa? What confuses me is that as the parameter precision increases, SHG should increase, shouldn’t it?

I think you can consider 6000 as a converged result

3、Additionally, I have found that when other settings remain the same and only the number of k-points is changed, the SHG with excitonic effects decreases as the number of k-points increases. I have the same confusion regarding this as mentioned above.

in some sense this is normal. When you have few k-points all the spectral weight is concentrate in few transition at the given k-points, more k-points you add more the spectral weight is distributed among
the different k-points. This is true in independent particle but also when excitonic effects are present even if the behavior could be slightly different.

best
Claudio