Doping and substrate effect for static dielectric matrix

[clin]

Dear developers,

I wonder if there is any simple way in Yambo that I can rigidly shift the Fermi level of the system to model doping effects. By checking the supported input variables, "Nelectro" seems relevant. Is there a way to directly set the Fermi level?

Another question is if there is any input variable that I can use to include the effect of dielectric constant of substrate. For example, the bare Coulomb potential in Yambo is 4\pi/q^2. Instead, I would like the bare Coulomb potential to be 4\pi/q^2/\epsilon_r, where \epsilon_r is the relative dielectric constant of a substrate.

Thanks,
Changpeng

[Daniele Varsano]

Dear Changpeng,

yes, changing the Nelectro in the input file will do the job.
At the moment, there is no the possibility to include a relative dielectric constant in the potential.

Best,

Daniele

[clin]

Dear Daniele,

Many thanks. I have a further question about RIM potential and Coulomb potential. When looking into the code of calculating Coulomb integral:
https://github.com/yambo-code/yambo/blo ... ter_Gamp.F
it seems to me only G=G'=0 component of RIM-averaged potential (ndb.RIM) is used (assuming no truncation?), while for G\=0 component of Coulomb potential the truncated one stored in ndb.cutoff is used. Is this because the truncated potential is already regularized? If this is true, this means one doesn't necessarily to set RandGvec larger than 1 when Coulomb truncation is applied?

In such case, since with truncation the potential is already regularized, why won't to apply the RIM-averaged potential only for q=0, as this is the only point where the actual divergence occurs?

I also noticed for a calculation reading ndb.RIM from the disk there will be a WRN info that says the truncated is not applied in the database even the ndb.RIM is first calculated with Coulomb truncation:

Code: Select all

 [RD./static_screening//ndb.RIM]-------------------------------------------------
  Brillouin Zone Q/K grids (IBZ/BZ)                :   64   64   64   64
  *WRN* Coulomb cutoff potential                   : none
  Coulombian RL components                         :  109
  Coulombian diagonal components                   : yes
  RIM random points                                :  3000000
  RIM  RL volume             [a.u.]                :  0.565781
  Real RL volume             [a.u.]                :  0.565807
  Eps^-1 reference component                       : 0
  Eps^-1 components                                :  0.000000  0.000000  0.000000
  RIM anysotropy factor                            :  0.000000
 - S/N 006718 ---------------------------------------------- v.05.02.00 r.22427 -

Best,
Changpeng

[Daniele Varsano]

Dear Changpeng,

it depends on the used coulomb potential. If the "slab" is used, it is analytic and RIM is applied on top of the Vcutoff, otherwise e.g. box it is as you said.

In such case, since with truncation the potential is already regularized, why won't to apply the RIM-averaged potential only for q=0, as this is the only point where the actual divergence occurs?

Even if it is regularized, the RIM potential is more accurate than the simple summation of Vq*Omega_q where Omega_q is the Volbz/Nq.

About the WRN, if the box is used that's correct: the ndb:RIM contains integral of bare potential that it is used to build the "box" potential. If, instead the slab is used than there is a problem in the reporting to be fixed.

Best,
Daniele

[clin]

Dear Daniele,

I see. Thanks a lot.

For the WRN info, indeed I used the slab cutoff, so it is a bug?

Best,
Changpeng

[Daniele Varsano]

Dear Changpeng,

it is possible, I'll check it as soon as I have time.

Daniele