SCF Grid Size and Yambo Memory Consumption

[jmullen]

Greetings all,

Two quick questions:

1) Can I run a GW calculation (and import using p2y, of course) with the results of a band structure calculation in Quantum Espresso using 'nscf' and only manually selected k-points, or can I only use the grid results from 'scf'?

2) Is there a "back of the envelope" calculation I can use to estimate the memory required to run a GW correction to a QE bandstructure (to help define the Yambo input variables)? I cannot afford trial and error on computers which bank my usage time.

Regards
Jeff Mullen
Physics
NCSU

[claudio]

>1) Can I run a GW calculation (and import using p2y, of course) with the results of a band structure calculation in Quantum Espresso using 'nscf' and only >manually selected k-points, or can I only use the grid results from 'scf'?

you need the full regular grid in the calculation of the dielectric constant, but then you can select single k-point of the same grid to calculate
GW corrections. If you want GW corrections to other k-points not in the grid you have to shift the grid on these points in order to have the same q-points.

>2) Is there a "back of the envelope" calculation I can use to estimate the memory required to run a GW correction to a QE bandstructure (to help define the Yambo >input variables)? I cannot afford trial and error on computers which bank my usage time.

try small tests with few plane waves and bands

ciao
claudio

[Daniele Varsano]

Dear Jeff,

1) Can I run a GW calculation (and import using p2y, of course) with the results of a band structure calculation in Quantum Espresso using 'nscf' and only manually selected k-points, or can I only use the grid results from 'scf'?

There is not any problem to perform GW calculations with electronic structures coming from "nscf". The only
issue you have to take care is to have an uniform sampling of the Bz.

Daniele

PS: Claudio, please add your affiliation in the signature.