From newville@cars1.uchicago.edu Tue Nov 20 15:07:22 2001
Date: Mon, 19 Jul 1999 10:34:17 -0500 (CDT)
From: Matt Newville <newville@cars1.uchicago.edu>
To: Anatoly Frenkel <frenkel@bnl.gov>
Cc: Alex Ankudinov <alex@leonardo.phys.washington.edu>
Subject: Re: using FEFFIT to fit experimental standards

Hi Anatoly,

Yes, I have seen cases where including H in Feff calculations screws up the
fit results.  My recollection was that both the phase and amplitude were
altered, but that the amplitude change was more significant.  It's possible
that such problems have been fixed with Feff8, but I haven't tried.

> Since I am analyzing some Cd in solutions data, with a mixture of Cd-O and
> Cd-S paths in the first shell, it would be odd to use for this purpose the
> Cd-O path obtained from FEFF for the bulk crystal model. The most
> appropriate theoretical model, which would better describe Cd-O in a
> molecule is therefore ambiguous, since depending on whether I include
> hydrogens in FEFF or not, I am getting the FEFF amplitude differed by ca. a
> factor of two.  I decided to bypass this ambiguity rather than dealing with
> it by simply extracting Ampl(k) and Phase(k) for Cd-O path from its
> experimental solution Chi(k) data, and prepare a feff0001.dat. It then can
> be used in FEFFIT in the same way as the feff0001.dat calculated with FEFF.
> 
> Have you done that before? I tried a simple test, by preparing a
> theoretical Chi(k) using FEFF and FEFFIT, back fourier transformed it,
> subtracted the 2kReff from its phase, and compared the result with the
> Phase(k) produced by FEFF for the same theory. The results were very
> different. According to your feffit manual, they should be the same. Have
> you done anything similar to what I am trying to achieve, so that I won't
> reinvent the wheel?

No, I've never tried that before.  It seems doable, but may not be
completely straightforward.  You probably got most of this, but I think
the way I'd do it would be: 

 1.  Make sure there's no k-weighting in the FT when extracting the
     amplitude and phase from the data (or, un-k-weight the amplitude).
 2.  The columns of feff0001.dat are:
         k   real[2*phc]   mag[feff]  phase[feff] red factor  lambda   real[p]@#
     I'd set them up like this:
         real[2*phc] = 0. 
         mag[feff]   = extracted amplitude
         phase[feff] = extracted phase-2kR
         red factor  = 1.
         lambda      = 10.e7  (ie, something huge)
         real[p]     = k
 3.  Just to be safe, I'd use the 'normal feff0001.dat' k-grid.
 4.  Setting up to 'nleg, deg, reff' line  and the atoms list should be
     fairly straightforward.   The degeneracy and reff are really used, but
     the rest of the numbers are only used in the Debye function.  

If you want to send me what you have for the feff->feffit->feff0001.dat,
I'll try it here and see what happens. 

--Matt

 |= Matthew Newville       email: newville@cars.uchicago.edu          =|
 |= GSECARS, Bldg 434A     voice: (630) 252-0431                      =|
 |= Argonne Natl Lab       fax:   (630) 252-0443                      =|
 |= 9700 South Cass Ave    www:   http://cars.uchicago.edu/~newville/ =|
 |= Argonne, IL 60439 USA                                             =|


