[gpaw-users] Local Green's functions from GPAW for MND calculation?

[Thomas Olsen]

Hi Toma



That is certainly possible. To get the contribution from atom ia and kpt ik at \omega=5.0 just do



calc = GPAW('some_calculation.gpw')

P_ni = calc.wfs.kpt_u[ik].P_ani[ia]

e_n = calc.get_eigenvalues(kpt=ik)

e_nn = np.diagflat(1 / (5.0 - e_n +0.1j ))

Gak_ii = P_ni.T.conj().dot(e_nn).dot(P_ni)



... or something like that - I did not check that it works. The result is a matrix in the projector indices. The i indices are combined n, l and m (the real part of spherical harmonics), so you will have to check what they correspond to for a given atom.



/Thomas

________________________________
Fra: Toma Susi [toma.susi at univie.ac.at]
Sendt: 2. marts 2015 09:39
Til: Thomas Olsen
Cc: gpaw-users; Jan Rusz
Emne: Re: [gpaw-users] Local Green's functions from GPAW for MND calculation?

Dear Thomas,

That’s right, those are the indices. The valence overlaps should be sufficient at least to start with.

However, referring to the PDOS Wiki page, for our purposes we would need to extend the formula for \rho_i^a to non-diagonal elements, i.e., \rho_{i,i'}^a. We can stay diagonal in terms of atom index "a", but need to consider off-diagonal terms in i/=i’. Is this possible?

Thank you,
Toma


On 26 Feb 2015, at 09:36, Thomas Olsen <tolsen at fysik.dtu.dk<mailto:tolsen at fysik.dtu.dk>> wrote:

Hi Toma

What do the indices here refer to? I am guessing that (nk) are Kohn-Sham state indices and (lms) are angular momentum indices for a given atom. In that case it is trivial since <lms|nk> are well approximated by the projector overlaps, which are computed in any standard GPAW calculation. You can also calculate the all-electron overlaps if you like, but that is a bit more involved. I can give you a quick example of how to get them if you like. Or you can have a look at the PDOS page in the GPAW documentation.

/Thomas

________________________________
Fra: gpaw-users-bounces at listserv.fysik.dtu.dk<mailto:gpaw-users-bounces at listserv.fysik.dtu.dk> [gpaw-users-bounces at listserv.fysik.dtu.dk<mailto:gpaw-users-bounces at listserv.fysik.dtu.dk>] på vegne af Toma Susi [toma.susi at univie.ac.at<mailto:toma.susi at univie.ac.at>]
Sendt: 25. februar 2015 14:56
Til: gpaw-users
Cc: Jan Rusz
Emne: [gpaw-users] Local Green's functions from GPAW for MND calculation?

Greetings,

I had the pleasure of meeting Jan Rusz from Uppsala recently, and heard of a nice implementation of the dynamical core-hole screening of Mahan, Nozieres and De Dominicis that he has been using. The calculations make use of initial and final state local Green’s functions:

G_lms,l'm's'(E) = \sum_kn <lms|kn><kn|l'm's'> / (E_k - E - i*eps)

To my understanding of these things, it does sound this is designed with plane-wave codes in mind. Jan has written a small code for WIEN2k for calculating these, and there exists a MATLAB script from VASP outputs. The real computational task is apparently getting the converged final state for each atom of interest.

Since I know GPAW is rather nice for final state calculations, I was wondering if it would be possible to get the right kind of output from GPAW? Jan thinks we need the code to be able to provide <kn|lms> projections from the Bloch states, but perhaps something equivalent would be available in GPAW? However, I wouldn’t want to use the PW mode, since the main interest would be to calculate rather extended defects.

Either a quick “no, cannot be done” or a “yes, trivial to implement, let me help” answer would be ideal! :-)

Many thanks,
Toma

________________________________
Dr. Toma Susi
FWF Lise Meitner fellow
University of Vienna, Austria

http://mostlyphysics.wordpress.com<http://mostlyphysics.wordpress.com/>

"Love is the only emotion that enhances our intelligence."
-Humberto Maturana

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20150302/b6683f82/attachment.html>