Why must use normconserving PP?

Submitted by enixchen on Wed, 02/24/2016 - 20:00

I'm wondering why BerkeleyGW can only use NCPP. I read from the BGW literature that "Note that norm-conserving pseudopotentials must be used, or else extra contributions would need to be added to our matrix elements." Could you please provide more reference explaining this? Thank you very much.

jornada's picture

Submitted by jornada on Mon, 02/29/2016 - 16:26

The main idea is the following:
When you relax the norm-conservation condition, you are essentially approximating the true quasiparticle wave function |mk⟩ as a sum of two terms: the pseudo-wavefunction |m`k⟩, which is expressed in terms of plane-wave coefficient, and a correction term |Bk⟩ that is only "on" near the core region. So, you cannot express matrix elements such as ⟨mk| e^{-i(q+G)r} |nk+q⟩ solely in terms of the pseudo-wavefunction coefficients ⟨m`k| e^{-i(q+G)r} |n`k+q⟩, and you'll end up extra terms that depends on these set of |Bk⟩. We don't currently support the computation of these extra terms in BerkeleyGW.

A good reference for these extra matrix elements is Blochl's original work which introduced the Projected Augmented Method: http://journals.aps.org/prb/abstract/10.1103/PhysRevB.50.17953


Submitted by babarker on Fri, 09/30/2016 - 12:34

Hello enixchen and BGW forum users,

While BerkeleyGW can not use wavefunctions generated from Vanderbilt pseudopotentials, the relatively new scheme by Prof. Hamann is pretty close to being a "best of both worlds" approach, requiring a reduced kinetic energy cutoff (compared to say, TM or RRKJ) while also conserving norms.

A good source of ONCVPSP files is listed at http://www.pseudo-dojo.org/ . (The usual caveats about manually verifying the quality of pseudopotentials applies.) The input parameters at pseudo-dojo appear to be very "stable" with respect to changes of various other input options, such as addition of further angular momentum channels, etc.

Bradford A. Barker