Run epsilon on a subset of bands

Submitted by manningyoung on Tue, 05/23/2017 - 21:53

Forums 

User questions

Hi all,

Is there a way to run the epsilon calculation on a specific subset of bands? Or should this be done via the WFN file? EDIT: I note siesta2bgw allows for an energy window, but pw2bgw does not. (I'm using ESPRESSO.)

The number_bands flag sets an upper bound, but I want to isolate a few near the Fermi energy.

Thanks in advance!

Best,
Manning

Submitted by babarker on Wed, 05/24/2017 - 13:18

Hello Manning,

What do you mean exactly by a "specified subset of bands"?

Best,
Brad

Submitted by manningyoung on Wed, 05/24/2017 - 21:24

Hi Brad,

Suppose my WFN file contains 15 bands. Is it possible to run the epsilon calculation only on bands 10-15?

I believe this can be done using siesta2bgw with
wfng_energy_flag = .true.
wfng_energy_min = [min (eV)]
wfng_energy_max = [max (eV)]
but I've found no such solution for ESPRESSO.

Best,
Manning

Submitted by babarker on Thu, 05/25/2017 - 11:46

Hello Manning,

The epsilon calculation will use all occupied bands supplied in the wavefunctions WFN and WFNq. (Really, just WFNq, but they files need to be consistent.)

You can regenerate wavefunction files with different pseudopotentials that exclude the states you do not believe you will need for epsilon. The only instance I can imagine that this might be appropriate, though, is in excluding the very low-lying (n-1)s and (n-1)p states that must accompany semi-core (n-1)d states to calculate accurate bare-exchange matrix elements. (See Murilo L. Tiago, Sohrab Ismail-Beigi, and Steven G. Louie, "Effect of Semicore Orbitals on the Electronic Band Gaps of Si, Ge, and GaAs Within the GW Approximation," Phys. Rev. B 69, 125212 (2004))

I am not sure why the developers supplied that option in siesta2bgw, though I am largely unfamiliar with Siesta.

Best,
Brad