The ABINIT Users Mailing List ( CLOSED )

[marzari@mit.edu]

Dear Fred, Xavier,


I would also be very interested in having the max-loc Wannier
functions in Abinit.

The Wannier code is now available in the public domain
(http://www.wannier.org); this version includes both the
original algorithm (PRB 56, 12847, 1997) for composite groups of bands 
separated by a gap, and the disentanglement procedure (Souza, Marzari,
and Vanderbilt, PRB 65, 035109, 2002) to extract subspaces of interest 
from an entangled manifold (e.g. d bands in copper).

In addition, the code can work in two modes: either it reads the u_nk 
wavefunctions (in real or reciprocal space) as produced by an independent
plane-wave electronic structure code, or, if another basis set is used
(e.g. FLAPW) it reads directly the matrix elements <u_mk|u_nk'>,
and then performs all the disentanglements/localizations requested
(for an applications using FLAPW, see PRB 65, 184422, 2002 ).

Interfacing a generic, non-plane-wave code to the Wannier code would
require only coding the calculations of <u_mk|u_nk'> for values of
m, n, k and k' that are an explicit output of the Wannier code itself
(once it knows the gemoetry of the supercell). Interfacing a plane-wave
code requires only the writing of the wavefunctions in real space according
to some obvious criterion, or, with some more effort, the interfacing between
the internal G vector notation of the Wannier code (that follows that of
Castep ca. 1995) and that of the native code.

Bottom line: I would really encourage any effort towards ABINITization
(I'm sure you haven't heard this word before); we could also consider some 
initial visit to Boston, to get such a project started on the right foot.

In any case, I would really appreciate to be kept informed of any such
project; incorporating these developments in a coherent fashion in the
public code is an important priority.

All the best !

                      nicola