Zero and non-zero components of shg errors

[sunxl]

Dear all，

First of all, I calculate a two-dimensional system, and can correctly identify the space group to which the system belongs in the self-consistent and non-self-consistent calculation of qe. For this system, its shg-yxx is non-zero and shg-xxx is zero. When I calculate shg using the G0W0-IPA and G0W0-BSE methods, I find that shg-xxx is not zero, regardless of the IPA or BSE methods.

1、For 2D systems, I added cut-off according to the tutorial, and except for the collision calculation I used single-precision yambo_rt, the rest used double-precision ***_nl, set Efield1 to 1,0,0 in the fixsymm.in file, removed the time inversion symmetry. Why does the obtained shg-xxx still have a large value? Does this affect the accuracy of my non-zero shg-yxx calculation? Attachments for the relevant input and output files are attached.
2、[WARNING] Variables Field1_Freq and Field1_NFreqs ignored. Does this warning have any effect on the calculation of shg?

Thank you in advance for your help!

Best,
sunxl

[claudio]

Dear sunxl

are you sure the atoms in you cell are not rotated? try to make a test on hBN monolayer
you should find the correct SHG, then it is the same for all systems

let me know
best
Claudio

[sunxl]

Dear claudio，

Thank you for your prompt response.

1、I have confirmed that the atoms in my structure have not undergone rotation, and I have verified the alignment of the corresponding x-axis and y-axis.

2、My test：I conducted tests on the single-layer BN based on the tutorial. https://wiki.yambo-code.eu/wiki/index.p ... X,_TD-BSE) While we know that single-layer BN should theoretically have only one non-zero component (yyy), when setting the external field direction to (0,1,0), both yyy and xyy components were obtained simultaneously. For the yyy component, my test results showed similarities with those demonstrated in the tutorial but were not completely identical (results are attached). For both yyy and xyy component data, I converted their units to pm/V by multiplying the original data by 4.189e8, but did not perform effective layer thickness rescaling. I observed that the maximum value of the yyy component was approximately 200 pm/V, while the xyy component reached around 50 pm/V (results are attached). If we perform effective thickness scaling on the SHG results, the magnitude of the xyy component would become larger.

3、From another post, I saw your explanation that if the maximum value of a non-zero component is an order of magnitude larger than the maximum value of a zero component, the results can be considered reasonable. For example, if the non-zero component yxx has a maximum value of 2000 pm/V, while the zero component xxx shows a maximum of 200 pm/V, such results would be considered valid. Is this interpretation correct?

Best，
sunxl

[claudio]

Dear sunxl

I repeated the tutorial and I agree with you that the xyy componenent that shold be zero is different from zero, even if smaller than the yyy.
This can be due to the not converged k-points sampling, because dipoles are calculated by finite differences in k-space.
I will make some test and let you know.
You can fast check it at the IPA level without doing GW and collisions

best
Claudio