TDDFT on DFT+U

[nthiliniek]

Hi All,

I want to calculate the optical spectrum with TDDFT-LRC. But I have applied Hubbard correction for the pwscf calculation to correct the band gap of the system. Since GW on top of DFT+U is not implemented, I have already modified interfaces/p2y/mod_p2y.F as : if(pw_lda_plus_u) call warning(' LDA+U. Hubbard correction is not considered in yambo.') .

1)Apart from modifying that part, do I need to change/add any additional parameters in the input file of TDDFT-LRC calculation?

2) If I want to perform BSE calculation for the same system, do I need to change/add any additional parameters in the input file?

Thanks in advance!

[claudio]

Dear Niranji

the only important change you have to do in the input files is to use covariant dipoles.
In order to do so when you generate the input file add "-V resp" and change the line

DipApproach= "G-space v" # [DIP] [G-space v/R-space x/Covariant/Shifted grids]

in

DipApproach= "Covariant" # [DIP] [G-space v/R-space x/Covariant/Shifted grids]

and everything should be fine.
Covriant dipoles are defined in Appendix D of
https://hal.science/hal-02032065v1/file ... 325902.pdf

best
Claudio

[nthiliniek]

Hi Claudio,

Thanks so much for the reply!

Niranji

[nthiliniek]

Hi Claudio,

1)As you mentioned, when I generate an input file with "-V resp", it comes with different other parameters (NonPDirs, MolPos, DipComputed, ShiftedPaths, DipoleEtresh....etc) as well.
Is it okay to add only

DipApproach= "Covariant"

manually to the input file without generating an input file with -V resp, for simplicity.

2)I did a test run (IPA) with covariant dipoles. it seems like calculation is quite expensive for my system. My system is 2-D and it has 576 electrons. What do you think if I stick with normal dipoles (G-space v) ?

[nthiliniek]

Dear Niranji

the only important change you have to do in the input files is to use covariant dipoles.
In order to do so when you generate the input file add "-V resp" and change the line

DipApproach= "G-space v" # [DIP] [G-space v/R-space x/Covariant/Shifted grids]

in

DipApproach= "Covariant" # [DIP] [G-space v/R-space x/Covariant/Shifted grids]

and everything should be fine.
Covriant dipoles are defined in Appendix D of
https://hal.science/hal-02032065v1/file ... 325902.pdf

best
Claudio

Hi Claudio,

1)As you mentioned, when I generate an input file with "-V resp", it comes with different other parameters (NonPDirs, MolPos, DipComputed, ShiftedPaths, DipoleEtresh....etc) as well.
Is it okay to add only
DipApproach= "Covariant"
manually to the input file without generating an input file with -V resp, for simplicity.

2)I did a test run (IPA) with covariant dipoles. it seems like calculation is quite expensive for my system. Even with the smallest calculation, it clashes. My system is 2-D and it has 576 electrons. What do you think if I stick with normal dipoles (G-space v) ?

Thanks
Niranji

[Daniele Varsano]

Dear Niranji,

you can calculate dipoles using the G-space method, but be aware that in this way you are neglecting the non-local commutator term. If such term is important or not should be tested. You can try to compare the two approach in a non-converged calculations (few bands) to have an idea of the error and continue with the G-space method if you realize that the error is not dramatic.
Best,
Daniele