Re: Question about cluster size in Atoms3.0

[Bruce Ravel <ravel@phys.washington.edu>]

Dr. Jian,

I hope you don't mind if I CC my response to your questions to the
feffusers mailing list.  I think that you questions are of sufficiently
general interest that other may appreciate seeing this.

On Wednesday 24 September 2003 07:59 am, Jian DING wrote:
	
> I have a question about the cluster size in TKAtoms 3.0 Obeta7. It
> is said in your documentation that, the cluster size defaults to the
> smaller of 7 A and 1.1 times the length of the largest lattice
> constant. However, when I tried to enter the cluster size smaller
> than the lattice constant, there is no error happened. For example,
> I enter the cluster size as 3.5A, or even 3A,althouth the lattce
> constant of SiC is 4.358A. When I run TKAtoms 3.0, The normal
> feff.inp files could be generated.

I suppose I should have said that the cluster radius defaults to 7 A
*or* 1.1 times the largest lattice constant, which ever is larger,
unless a cluster radius is specified, in which case that is used.  So,
if you specify a shorter cluster radius, that shorter value will be
used.  Indeed, whatever cluster radius you specify will be used.
The default only comes into play if you leave the little box for the
cluster radius empty.

> I am going to do some EXAFS calculation about dopants, for example,
> replacing several Si atoms by Al atoms in SiC. The atom number
> included in a cluster increases much more with the cluster size
> increasing.  Because I think it should be more effective to
> investigate the dopant effect in a cluster with fewer atoms, so I
> hope to calculate the smaller cluster if possible. I wonder if there
> is any effect on the calculation result if I use smaller cluster
> size mentioned above.

You have hit upon one of the essential issues in understanding how to
use feff8.  One wants to use as small a cluster as possible so the
calculation goes quickly.  On the other hand one wants to use a large
cluster so as not to neglect any important contributions to the
scattering.  For a material like SiC, I suspect that one will need a
rather large cluster because I suspect that the mean free path is
quite long, especially in the XANES region.

A good strategy for balancing the desire for a small cluster with the
need for a large cluster is run a sequence of feff calculations,
increasing the cluster size each time.  The best way to do this would
be to run Atoms with a large cluster radius.  Then set the radius
parameter to the SCF and FMS cards to successively larger values.
Eventually increasing the cluster size will not change the
calculation.  This tells you what size cluster you need to fully
represent the material.

The issue of introducing dopants into the feff calculation is a tricky
one.  I have written about this before.  Here is a link to a
discussion of this topic in the ifeffit mailing list archives.  It is
mostly about EXAFS, but I discuss XANES calculations on doped
materials at the end.
   http://millenia.cars.aps.anl.gov/pipermail/ifeffit/2002-July/000092.html

Note that convergence in cluster size may be different for a doped
cluster as for undoped SiC due to the introduced disorder.  Or it may
not be.  The only way to know is to develop your strategy for
computing the doped material and then investigating cluster size
convergence using that strategy.

   
I hope that helps.  If you have any further questions, your best bet
would be to subscribe to the FEFFUSERS mailing list.  See
   http://leonardo.phys.washington.edu/feff/Notes_comments/feffusers.html
and post your questions there.

Hope that helps,
Bruce


   
-- 
 Bruce Ravel  ----------------------------------- ravel@phys.washington.edu
 Code 6134, Building 3, Room 222
 Naval Research Laboratory                          phone: (1) 202 767 5947
 Washington DC 20375, USA                             fax: (1) 202 767 1697

 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/