Re: f' and f'' scale problem?

["John J. Rehr" <jjr@leonardo.phys.washington.edu>]

Dear Feffusers,

 Stephane Grenier writes:

>     i find an unreasonable  scale discrepancy between f' and f'' 
> calculated by feff (using XANES and DANES cards on a Mn oxyde cluster) 
> and those from the Cromer-Liberman tables (See picture). If f' might 
> still be reasonable, f'' is rather small. I suspect something wrong here 
> as f' and f" from feff don't seem to be scaled correctly neither 
> relatively to each other.

   The descrepancy is due to the units used by FEFF8 to present the
spectra:

   Presently the FEFF8 code produces normalized x-ray absorption cross
sections for a given edge

           mu_norm(omega)=mu(omega)/mu(E_0+50 eV),

so mu_norm asymptotes to unity at large omega.

   To get the cross-section mu(omega) in experimental units (angstroms**2),
one must multiply by the normalization constant given in the last line
of the header in xmu.dat, i.e.,

#  xsedge+ 50, used to normalize mu           3.4849E-04

   To get the contribution to f" from that edge in oscillator strength
units, one must multiply  by an additional factor (in atomic units)

    f"(omega)=omega c mu(omega)/[4 pi  bohr**2 hartree]

where omega is the x-ray energy in eV, c = 137, bohr = 0.529, and
hartree =27.2.

  To compare with Cromer-Liberman, one also has to sum over edges.

  The FEFF Project is currently working on extensions which automatically
sum over edges and carry out the extra arithmetic for various spectra
to simplify this task for users. The feff8 doc is being updated to
include this information as well.  We would appreciate hearing from users
who need these extensions.

  Sincerely,
  John Rehr