Core-level shifts - frozen or relaxed core

Message

Wasabi · #1 Post by **Wasabi** » Tue Jan 04, 2011 4:57 pm

Hi everyone,

I'm calculating core-level energy shifts and have a simple question:

In the paper of KÃ¶hler and Kresse (DOI:10.1103/PhysRevB.70.165405) core-level shifts are calculated in the final state approximation with frozen core electrons: "Screening by the core electrons is not taken into account(i.e., the other core electrons are kept frozen in the configuration for which the PAW potential was generated)".
There are several studies where people have calculated core-level shifts with a frozen core using the VASP code.

Now there is a paper of Marsman et al. (DOI: 10.1063/1.2338035) where a relaxed core PAW method is introduced. Zeng et al. (DOI: 10.1103/PhysRevB.70.165405) calculate core-level shifts in VASP using the method introduced by Marsman "... and remaining core states are relaxed along of the self-consistent calculation."

My question is: If I do the two-step calculation procedure, where in the second calculation I define the final-state approximation with the appropriate tags, is the core relaxed or not? If not, is there a way to relax the core?

TY and best regards,

wasabi

alex · #2 Post by **alex** » Wed Jan 05, 2011 7:21 am

Hello Wasabi,

relaxed core PAW are not yet available, if I'm not completely wrong.

Cheers,

alex

Wasabi · #3 Post by **Wasabi** » Wed Jan 05, 2011 9:30 am

Hi alex,

thx for the reply!

I just realized that a file named core_rel.F is part of the VASP code. The header of the first subroutine 'SPAWN_PP' says: 'This subroutine determines the set of ions (on the local node) that are unrelated by symmetry. It takes only those types into account that are set to be treated with the relaxed core method, i.e., for which INCAR flag LRCTYPE is set to true. (default LRCTYPE=.TRUE. for all)

Also, there is is a quote 'See Eq. (X) in M.Marsman, G.Kresse, to be published' in this file.

So maybe that file is related to relaxed core PAW?

Wasabi

alex · #4 Post by **alex** » Thu Jan 06, 2011 10:38 am

Well, I'd prefer some well tested and documented routines. It's hard to do some back engineering if the theory isn't published yet. Or just ask Marsman directly.

Cheers,

alex