How to implement a BSE calculation with double grid method

[jasonhan0710]

Dear all,

I have noticed that yambo can use a dense k grid in BSE calculation with interpolation.

In Claudio Attaccalite's Homepage (http://www.attaccalite.com/speed-up-die ... ith-yambo/), in step 2,

Perform a second DFT calculation with a dense k-point grid for example 40x40x1 shifted grid and the number of bands you need in optics. Notice that the problem of k-points convergence is due only to the bands close to the gap, so you do not need the dense k-point grid for all conductions bands.

Does it mean that I can use only, for example, 200 bands in the dense k point nscf calculation, which is 400 in the first nscf calculation? Will it affect the convergence of BSE calculation?

In step 6,

Now repeat the dielectric constant calculation of point 1. Yambo will average all denominators appearing in the epsilon calculation using the energies calculated on the dense grid.

Should I recalculate the static screening and BS kernel? Or just reimplement the BSE diagonalisation? How about the truncated coulomb potential for 2D system, should I recalculate it with the dense k grid basis?

Best,
Jason

[claudio]

Dear Jason

in general in the double grid you need less bands, I will say equal to two times
the number of valence bands is enough.
This because for higher bands there are less oscillations in the denominator with the k-grid.

best
Claudio

[jasonhan0710]

Dear Claudio,

Thank you for your reply.

Is it possible to get the exciton wave function in BSE calculation with double grid inversion solver?

Best,
Jason

[Bruno]

Dear Jason,

Unfortunately it's not possible. To obtain the wavefunctions you'll need to perform a full diagonalization and post-process it with ypp.

Regards,