[Ifeffit] Re: strange distance obtained using atoms

Bruce Ravel ravel at phys.washington.edu
Mon Nov 15 16:01:25 CST 2004

On Thursday 11 November 2004 9:11 am, Valmor Mastelaro wrote:

> Bruce, I´m using the atoms program to obtain the feff.inp file for a PZT
> compound. when I introduced the crystallographic informations on the atoms,
> I obtained only one distance that is too short to be true, Ti-Zr = 0.483
> I verify all the crystallographic parameters in the literature and they
> seens to be correct. (Journal of Physics: condensed matter 10, 6251-6269
> (1998).

Hi Valmor.

I am taking the liberty of forwarding my answer to the
Ifeffit mailing list as well as the feffusers list.  

Here is the Atoms input file that Valmor sent:

   space = R 3 c
   a =    5.78510       c =      14.2880
   core =       Ti      edge =  K       rmax =    9.0
   ! elem   x          y          z     tag           occ.
      Pb    0.00000    0.00000    0.27170                1.00000
      Ti    0.00000    0.00000    0.03590                1.00000
      Zr    0.00000    0.00000    0.00210                1.00000
      O     0.16321    0.34281    0.08333                1.00000

Actually, the one very short atom was only the most obvious of a host
of problems with this input data.  Look also at the Zr/Ti coordination
shell.  There are far too many atoms and they are much too close
(about 0.483 AA apart -- imagine that!) to one another.

Reading the paper that he cited, I see that coordinates are given for
the anion and cation positions by parameterized deviations from the
symmetry positions of the space group.  It's really a very clever
paper -- the parameters of the structure are things like cation
displacement and octahedral tilt.  Very cool stuff, all of which is
explained abstractly in Table 1.  Then in Table 3, the authors give
the actual values of the parameters for each of the anions and
cations.  Among the data in this table are the fractional occupancies
of Zr and Ti along with separate displacement parameters for those two

Before I go on, you need to read this old entry to the Ifeffit mailing
That entry actually answers your question about why Atoms isn't
working as you anticipated, but I'll expand on a few points here.

Feff (and therefore Atoms) presumes that you have one entire atoms at
any given site.  Feff does not allow you to fractionally occupy a
location in space.  This may be a shortcoming of Feff, but there are
good reasons why Feff behaves that way.  Since Feff does not know how
to compute a cluster with fractional atoms, Atoms does attempt to
write fractional atoms to the cluster.

Thus, when you put *both* Ti and Zr in the atoms.inp file, Atoms put
six of each in the unit cell.  (This crystal has six formula units in
each unit cell.)  That's obviously incorrect.  The paper you cite
clearly states that the samples have between 13 and 40% Ti according
to the stoichiometry Pb Zr_{1-x} Ti_x O_3.  So it is simply incorrect
to have both the Ti and Zr entries in your atoms input file.

As I explain in the link above, you generally cannot solve a problem
like this with a single Feff calculation.  You simply must consider an
analysis strategy which involves at least 2 and possibly many more
Feff calculation.

Are you doing EXAFS analysis?  If so, then you should consider running
Feff once without the Ti site and using the parameter given in the
cited paper for the Zr site.  Then run Feff again with the Zr site and
using the parameters given for the Ti site.  Then, when you do your
analysis, you will need to multiply the amplitude terms for the paths
from the Zr calculation by (to use the example of one end member) 0.87
and multiply the amplitudes of the paths from the Ti calculation by

Or are you doing XANES calculations?  In that case, you will need to
run many Feff calculations with the Ti and Zr sites randomly populated
following the modle of the Au2O3 that I discuss in the link above.

In any case, the bottom line that you have to remember when using
Atoms and Feff is that these programs do not and cannot treat
occupancy in the same manner as is discussed in a paper on Reitveld
refinement.  (That is true at least in the context of running a
*single* Feff calculation.)  You have to understand and think hard
about how Atoms works and how Feff works and how they should be used
for your specific analysis problem -- this is particularly true in a
problem like this, which involves a rather complicated scheme for
treating the relationship between stoichiometry and structure.

Hope that helps,

 Bruce Ravel  ----------------------------------- ravel at phys.washington.edu
 Code 6134, Building 3, Room 405
 Naval Research Laboratory                          phone: (1) 202 767 2268
 Washington DC 20375, USA                             fax: (1) 202 767 4642

 NRL Synchrotron Radiation Consortium (NRL-SRC)
 Beamlines X11a, X11b, X23b
 National Synchrotron Light Source
 Brookhaven National Laboratory, Upton, NY 11973

 My homepage:    http://feff.phys.washington.edu/~ravel 
 EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/

More information about the Ifeffit mailing list