Q dependence of exciton eigenvalues

[DavidPolito93]

Dear Daniele,

coming back to one of your previous answer, you told me that in order to get the single value Kx or Kd at q=0 I should compute:

<Psi_lambda | Kx | Psi_lambda>=\sum_cvk \sum_c'vk'' A_lambda^cvk A*_lambda^c'v'k' <cvk|Kx|c'v'k'>.

On the other hand, for q different from zero I don't see where the difference is: I mean, the only thing is to consider k' = k-q isn't it?

Best,
Davide

[Daniele Varsano]

Dear Davide,

On the other hand, for q different from zero I don't see where the difference is: I mean, the only thing is to consider k' = k-q isn't it?

Yes, for finite q, this is your basis, and these are the elements written in the excitonic matrix.

Best,

Daniele

[DavidPolito93]

Dear All,

So I am trying to implement the feature to extract those informations from the ndb.BS_PAR* database. Just a few questions regarding that:

1) What is actually stored in BSE_RESONANT is the whole kernel (i.e. K = Kx + Kd) isn't it? So in order to get the two contributions separately I should do a singlet computation with BSENGexx different from zero from which I get the whole K and a triplet computation with BSENGexx = 0 so that I only get Kd. Then Kx = K - Kd. Is it right?

2) Concerning the size of the database, this is 2500. My computation has 5 valence and 5 conduction bands and 100k points in the whole BZ (which are reduced to 51k points by symmetry). So I assume the size is 5x5x100=2500 and there is no symmetry reduction in the database. Is it correct?

Best,

Davide

[Daniele Varsano]

Dear Davide,

1) You can do a triplet calculation (Kd only) and an Hartree calculation (Kx only).
2) Yes the matrix is stored in the full BZ.

Best,
Daniele

[DavidPolito93]

Dear Daniele,

Thank you for your answer. How do you do a Hartree only (Kx) calculation in BSE?

Best,

Davide

[Daniele Varsano]

Dear Davide,

just select:

Code: Select all

BSKmod= "Hartree" 

Best,

Daniele