TDDFT Drude term questions

[hongtang2023]

Dear Dario,

Thank you so much for helping!

1. I did try " GrFnTpXd= "T" " or " GrFnTpXd= "r" " , but those "NaN" are still there. And the strange thing is that with Drude term, even I put " GrFnTpXd= "T" " or " GrFnTpXd= "r" " , but after the runs, it automatically became " GrFnTpXd= "R" ".

I will try to recompile the code with double precision and test it again.


2. If I use the recipe in 10.1103/PhysRevB.107.155130 to estimate the intraband frequency omega_A, can I do it with TDDFT IP/RPA, RPA/HARTREE, or ALDA to get the dielectric function or EELS data (for q=0 and q not zero)? then to get the Y_E(w) and Y(w).

I am not sure how you get the Y_E(w) and Y(w) in 10.1103/PhysRevB.107.155130. It seems that you did GW related calculations to get the dielectric function, right?


3. Another important question.
In TDDFT calculation, say TDDFT ALDA, to get the dielectric function or EELS data,
for q=0, I need to set "
% LongDrXd
1.000000 | 0.000000 | 0.000000 | # [Xd] [cc] Electric Field
%
"
here, I set the electric field direction 1 0 0.

However, for q not zero, for example, the second q point in my system is " Q [ 2]: 0.000000 0.000000 0.100000 [rlu] " , it seems that this q is pointing to the z axis (or vector c). but I am not sure what the direction of E field is for this second q point.

So, my question is that are the directions of E fields for q=0 case and q-not-zero-case required to be the same? Is " LongDrXd " only working for q=0 case? How to set the E field direction for q not zero case?


4. If I do TDDFT ALDA to get the dielectric function and EELS data (for q=0 and q not zero), then to get the Y_E(w) and Y(w), is any requirement for the size of the smallest non zero q? for example, the second q is " Q [ 2]: 0.000000 0.000000 0.100000 [rlu] " ,
is 0.1 rlu enough?


I appreciate your help very much!

Best,
Hong

[DarioALV]

Dear Hong,

You do not need GW to calculate the intraband frequency, only optical calculations at a level of theory of your choice and without any Drude term. You can use for instance independent particle.

Regarding the direction of the external field, you can set the same for all the q values including q=0. This is not related to the index of the q points. You can see the list of k and q points in the r_setup file. If you have more than 50 you may want to use <yambo -fatlog> to see them all, this also report two more units besides rlu.

If your system is cubic all the directions x, y, z will be equivalent, otherwise you may compute the intraband frequency in the direction of larger variation. You can find the index of the q you want in the r_setup file.

For the calculation Y_E(q=0) is the response function that accounts only for interband contributions (without Drude term)
Y accounts for both intra and interband contributions and needs to be computed in the q-> limit, it is then approximated as Y(q->0) = Y(q=q_min), where q_min is the smallest finite q of your grid in the direction of your choice
The intraband frequency, Omega_A^2, is given by the integral of w dw (Y-Y_E), see Eq. (13) of the paper.

Of course, you will need to set a frequency interval [0,w_max], where w_max depends on the number of bands and converge Omega_A with respect to q_min (I do not know if 0.1 rlu is enough for your system).

After you obtain Omega_A you can then do new calculations setting the Drude term with this frequency.

Best regards,
Dario

[hongtang2023]

Thanks so much! Dario


Best,
Hong