[Ifeffit] an interesting Atoms problem

[Bruce Ravel]

Hi!

This week I received an interesting question about Atoms whose answer will, I 
think, be of wide interest to the denizens of this list.

The fellow who started the discussion had an atoms.inp file for ammonium 
tetraoxorhenate, (NH4)ReO4.  Here is the input data:
    space = I 41/a 
    a =  5.995 c = 12.440
    rmax=10  core = Re1
    atoms
    # elem.     x         y         z         tag 
      Re     0.00000   0.00000   0.00000      Re1 
      O      0.20450   0.37860   0.17020      O1
      N      0.00000   0.00000   0.50000      N1  

These data ran atoms to completion, but then Feff stopped after issuing the 
statement about computing the "muffin tin radii and interstitial parameters" 
with this error message:
   WARNING: NO ATOMS CLOSE ENOUGH TO OVERLAP ATOM   10,  UNIQUE POT    3!!

My correspondant found this message a little cryptic and asked for advice.

You will notice that the input data for atoms did not specify the hydrogen 
positions.  Often it is fine with both Atoms and Feff to neglect the 
positions of the hydrogen atoms.  For example, in a mineral phase with water 
molecules in the crystal structure, one can often just specify the positions 
of the water oxygens and neglect the water hydrogens.  Doing so will still 
let both Atoms and Feff run to completion and leave you with a Feff 
calculation that is quite useful in analysis.

So what happened here?  Apparently my correspondent presumed that ammonium is 
like water in that specifying the locations of the hydrogens is not essential 
to the Feff calculation.  Like in water, the hydrogens are fairly tightly 
bound to the nitrogen -- less than an Angstrom away -- so it might be ok.

In this case, however, the hydrogens are essential.  Adding the hydrogen 
position
      H      0.37200   0.45600   0.03300      H1
avoids the problem during the Feff run.  Feff finishes and the analysis can 
proceed.

It turns out that, while the N and the H are, indeed, quite close together, 
the N is quite a long way away from its next nearest neighbor, which is 
oxygen.  The oxygen shell is split into two subshells, the closer of which is 
at about 2.85 Angstroms from the N.

When Feff goes to compute the muffin tin potential, a near neighbor separation 
of 2.85 Angstroms is too large and it is unable to determine the radii of the 
atoms.  That is the meaning of Feff's slightly cryptic error message.  With 
the intervening hydrogen atoms, Feff's algorithm for constructing the muffin 
tins is no longer ill-defined -- the separation between atoms is no longer 
too large.

To conclude, while it is probably unlikely that the hydrogen locations can be 
well measured by the EXAFS experiment, it is quite important to include the 
hydrogens in the theory.  The rule of thumb that says that hydrogens in 
hydrated crystal structures can be neglected is not a very good general rule.  
This is certainly a case where the hydrogens cannot be ignored.

Well, *I* thought it was interesting.  Hopefully you did too!

B


-- 
 Bruce Ravel  ----------------------------------- bravel at anl.gov  -or-
						  ravel at phys.washington.edu
 Environmental Research Division, Building 203, Room E-165
 Argonne National Laboratory                        phone: (1) 630 252 5033
 Argonne IL 60439, USA                                fax: (1) 630 252 9793

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