[Wien] Wien vs. FEFF

[Michel Jaouen]

	Dear Wien Users,

	I will add few comments to the following contributions.

>I'd like to make an addition to the letter of Kevin:
>FEFF is better for higher energies above the Fermi energy, as it is not
>linearized method (as WIEN), and the multiple-scattering methods are good for
>higher energies as a whole. The larger cluster you take in FEFF the closer
>you may come down to the E_F. If you are interested in more than 30-40 eV
>above E_F, you should use FEFF, in a closer region WIEN is better.

I quite generally agree with the above remark. Feff is mainly devoted 
to spectroscopies (xas, eels, xes) and in many cases it gives ldos 
that are very close to those obtained with Wien or any other band 
structure codes when the muffin-tin approximation can be considered 
as acceptable (metals or metallic alloys). It can also work for oxide 
or covalent materials in some case (I have made nice diamond or O-K 
edge for STO) but you then have to play with the cluster size and the 
angular momentum expansion (for instance for diamond you need to use 
490 atoms and an s-p-d basis set to obtain all the xanes fine 
structures excepted the exciton line). Futhermore you can use the 
tdlda that may be of help in some case (L edges for instance) even if 
it is not yet fully automated.

>  > There are also a lot of effects which neither of the programs get right :
>>  wien is ground-state, and feff (currently) only has one-electron Green's
>>  functions; that's not good enough to treat excited states accurately (think
>>  of core hole effects, etc.)
>
>In FEFF there is an account of the self-energy in the Hedin-Lundqvist
>approximation in order to go beyond the DFT theory, though when I did this
>for the NiO case I didn't obtain an improvement of the resulting EELS
>spectra.

You can also play with other exchange for screening (Dirac-Hara, 
ground state (that must be used to compare the ldos with those 
calculated from Wien for example, partially local, ..). You have also 
to take care of 2 points when comparing a feff calculation to eels 
measurements: i) the background is not the same in eels ans xas and 
you have to correct for, ii) quite generally elnes spectra are 
different than xanes ones due to the lower energy resolution in the 
first case (I have never seen a routine diamond elnes spectrum 
similar to the xanes one) and to the effect (not already  fully 
understood) of the swift electron in eels which "distort" the atomic 
wavefunctions (a kind of polarization effect) and thus modifies the 
shape of the spectrum at the edge jump.

>The FEFF accounts the dipole transitions only, that is, it is valid for XAFS
>(XANES), and not EELS; though very often the EELS spectra have rather low
>contribution from nondipole transitions.

I want to correct this which is not true: you can use the MULTIPOLE 
card. Default is only dipole, but you can calculate dipole and 
quadrupole or dipole and magnetic dipole transitions (only with 
feff8).

	To summarize, the main drawback of feff is that it is 
restricted to the muffin-tin approximation (a full potential version 
is under developpment), but it is a nice tool (fully automated) that 
is complementary to a code like Wien. It depends of which physics you 
are interested with: if it is band structure, use Wien, if it is 
spectroscopy, use feff.

	I hope it may help

	Best regards

	Michel Jaouen

-- 
__________________________________________

  Michel Jaouen
  Universite de Poitiers - UFR Sciences - SP2MI
  LMP UMR 6630-CNRS
  Boulevard Pierre et Marie Curie - Teleport 2
  BP 30179
  86962 Futuroscope - Chasseneuil Cedex
   France
  Tel : (33) 5 49 49 67 37
  Fax : (33) 5 49 49 66 92
  e-mail : Michel.Jaouen at univ-poitiers.fr
__________________________________________