Interpolated QP Grid

[mdvorak]

Hello Developers,

I have a question regarding the interpolation of QP energies. In a BSE calculation, I know Yambo can interpolate QP energies to more k-points then are available in the QP database with the -ui option. I do not understand how to specify the interpolated k-points though. e.g. I calculate GW corrections on a 6x6x6 grid and want to interpolate to a denser grid like 10x10x10 or 12x12x12. Is there a way to specify the density of the interpolated grid? I don't see any variable to specify this.

Thank you,
Marc Dvorak

[myrta gruning]

Dear Marc

it should be automatic. The denser grid is the one you are using for the BSE calculation and the grid used for the QP corrections is recognized by yambo. What you can play with are the interpolation neighbours KfnQP_N: given P of the denser grid it said how many Q closes to P you want to consider to determine the value in P.

Regards,
m

[mdvorak]

Thanks......I thought Yambo must be capable of this, but I still don't completely understand how this works - 100% user error. Right now I am using the following:

KfnQPdb= "E < ./SAVE/ndb.QP ui" # [EXTQP BSK BSS] Database
KfnQP_N= 2 # [EXTQP BSK BSS] Interpolation neighbours

to read the QP database and activate the interpolation using 2 neighbors, and it is working. I think this uses the 2 nearest neighbors in every direction to interpolate the value onto the new k-point. I don't know what the new k-point is though or how many of them there are. Where do I specify the new k-point?
You said:

"interpolation neighbours KfnQP_N: given P of the denser grid it said how many Q closes to P you want to consider to determine the value in P."

As you have described above, what variable sets the value of P? My understanding is that KfnQP_N sets the value of ONLY Q. Also, are these interpolated k-points and their energies printed anywhere in the output? Thanks for the extra help,
Marc

[myrta gruning]

Hallo Marc

As you have described above, what variable sets the value of P? My understanding is that KfnQP_N sets the value of ONLY Q. Also, are these interpolated k-points and their energies printed anywhere in the output? Thanks for the extra help,

The P are the k points of the grid you are using in your present calculation, so you find the information over them in the report.
The Q are the k points of the grid you used to get the QP correction in a previous calculation.

In both cases the k point grid is not a parameter of yambo, but it is inherited by the DFT non-SCF calculation

From what you wrote I understand you are expecting something different, that is: you did a calculation for Qs, you specified the Ps and get the GW corrections there.
This is not how it is working.

Instead, let's say you did a GW calculation for system X using a NxNxN grid, for which you converged the results.
Next you want to calculate the optical absorption of X using BSE but to converge you need a MxMxM grid with M>N.
During this BSE calculation (in which you use the MxMxM grid) yambo allows you to use the ndb.QP database you generated using the smaller NxNxN grid.
Yambo gets the QP corrections for the MxMxM grid from the NxNxN grid using interpolation. The number of neighbors it used are determined by KfnQP_N.

I hope this may help you.

Regards,
m

[mdvorak]

Thank you for the help Myrta, I understand this interpolation now and have it working.

Marc