Yambo Community Forum

Dear developers, I would like to ask if it is possible to obtain the full dielectric tensor tensor (or the its imaginary part) after having performed three calculations with independent polarizations. Thank you!

Dear Michele,
I think it should be possible by using yambo_kerr.
Maybe others can tell you more on that.
Daniele

Dear Michele,
the standard version of yambo can compute the diagonal elements of the dielectric tensors. Playing with them, i.e. selecting different directions in input (you would need 6 in total), there are ways to reconstruct also the off-diagonal components of (the symmetric part of) the dielectric tensor. You can check in the literature for equations.

For the off diagonal components you can also use yambo_kerr, as mentioned by Daniele. Chosen a direction "d" in the "xy" plane in input, it will give epsilon_dd(w) and epsilon_{dd'}(w) where d' is a direction rotated by 90 degrees with respect to the original chosen. However it works only in the xy plane. Thus you need to rotate your sample (or change the code) to get also off diagonal terms which involve "z".
Notice that this approach, at variance with the previous one, would give also the anti-symmetric part of the dielectric tensor if the material has any.

Hoep it helps.
Kind Regards,
Davide

Hello,

Which equation is implemented to calculate the off-diagonal dielectric tensor for semiconductors? Eq. 22 or 23 of PRB 86, 125139 (2012)?

Best,
Xiaoming

If you select eq. (23) is used.

Instead with eq.(22) is used


Best,
D.

Thanks Davide,

For the velocity gauge, it seems the velocity operator need to be renormalized as explained in your paper PRB 95, 155203 (2017). I'm wondering why the position operator need not to be corrected?

Best,
Xiaoming

Not to necro this thread to much, but as my question is on this topic it would appear correct to extend this thread rather than make a new one.

With regards to reconstructing the dielectric tensor, I imagine that the pertinent relation is for (in standard Yambo) a input polarization (in xy) basis of 1/sqrt(2)*(1,1), that the dielectric response is (eps_xx+eps_yy)/2+eps_xy, in comparison to a response of eps_xx or eps_yy for (1,0), and (0,1), respectively? Just confirming my understanding of the previous posts.

Secondly, consider the case of a calculation with many steps such as BSE (e.g. GW correction calc., screening, BSE setup and solution). Would it be best practice to perform all calculations for each target polarization (so, for 6 tensor components, with standard Yambo this would be 24 total), or would it be reasonable to do some such as GW and screening in some basis such as (1,0,0) or (1,1,1) and then only do the BSE for all components? My current understanding is that the later is incorrect on fundamental grounds as the direction of screening certainly matters (after all, isn't this why people are so excited about some 2D materials?).

Dear andrewsalij ,

please sign your post with your name and affiliation, this is a rule of the forum and you can do by filling the signature in your user profile.

By default, BSEprop="abs", and in this case yambo calculate diagonal components only, if you need off-diagonal components you need to set BSEprop="dich".
A polarization along "xy" with BSEprop="abs", will give you eps_xx+eps_yy.

Regarding the GW calculation the effect of the polarization is milder than in absorption, in a 2D system, usually you can calculate the screening along one direction in the plane and then explore anisotropy in the BSE calculation. Of course, if the system is highly anisotropic also the direction in the screening calculation can have effects and you need to explore it.

Best,
Daniele

Thank you for your reply. If such a thing exists what would be the variable to adjust when doing ALDA or LRC calculations? I've read through the documentation to little avail. When I calculated with E field with (1,0,0), (0,1,0), and (1,1,0), the last one seems to just be an average of the first two.

Would it be necessary to redo all calculations including Quantum Espresso but rotated at 45 degrees?

I realize now that I was looking to find certain diagonal elements which were effectively zero.
For the sake of future reference, for elements which cannot be diagonalized, either perturbing the system with the field (1,1,0) or rotating the entire system by 45 degrees and determining the response from (1,0,0) yield the same results for ALDA, albeit with some finite differences due to numerical truncation.