[Ifeffit] thanks for your posting

[Matt Newville]

Hi Mustafa,

On Fri, Feb 22, 2008 at 1:13 PM, Mustafa Hussein
<mustafa.hussein86 at yahoo.com> wrote:
> Dear Prof Newville,
>
> Thanks for your reply. I have some questions, I am listing them below,

OK.  I'm going to forward with to the mailing list as well.

> I am simulating the excited state dynamics of two platinum atoms, can I
> specify two atoms as absorbing atoms? or I have to re-run the calculation
> again to average over the two sites of absorption?

You have to re-run the calculation again.

> I used the Atoms.inp Archive, but I did not arrive at a molecule that might
> be similar to mine.
> So I need to build an input file? it is okay, I think I can deal with shell
> scripting.

That's fine.  Using Pt metal might be a good place to start too.

> What is meant by HOLE? and the values 1 and 1.0 are the default or they can
> be changed? and at what basis?

In
   HOLE  1  1.0

the first '1' means "Use the K edge (1s electron)" which is one of the
most common edges used for EXAFS.  For Pt you may want to simulate the
more-commonly-measured L3 edge, which would use
  HOLE 4

The '1.0' means "multiply by a constant factor of 1.0", which is there
to allow one to put in a factor (usually called S02) that multiplies
the  calculated EXAFS.  For now, leave it as 1.0 -- it can be dealt
with later or ignored for a pure simulation.

There is some documentation for this on the Feff web page.

> What do the 0.00000 coordinates for the absorbing atom mean? I suppose they
> mean the atom is centered in a computational box or something, I do not know
> exactly!!

Feff  needs only atomic coordinates  of a "cluster of atoms" -- it
could be a molecule, part of a crystalline solid, or coordinates from
a simulation for a liqud, glass, etc.   It does not need or implicitly
impose periodic boundary conditions.

So the coordinates are all in Angstroms, but relative to some
arbitrary origin.  It's common to put the absorbing atom at 0,0,0, but
it is not necessary: "ipot = 0" specifies the absorbing atom.   If you
can generate x,y,z coordinates in Angstroms from your simulation, just
use those.

> You input six oxygen atoms, what are they? all bonded to two Mn atoms?? or
> they are pertinent to solvent molecules?

For the example I gave, those oxygens were bonded to the Mn.  For
EXAFS it generally does not matter a great deal whether the atoms are
bonded or not, though for a gas or liquid 'residence time' is a
factor, and the cluster should reflect a "snapshot" of the atomic
arrangement on the timescale of the EXAFS measurement (which is
femto-seconds).   In many cases, the solvent is visible, especially
for metals in water -- the hydration shell is seen.

> I quote "generally feff needs no more than a few hundred atoms", of the
> molecule in question, or including solvent molecules too? Actually, I am
> doing my simulation in the gas phase, would this be okay with the input
> requirements?

Oh, sure.   One potential (pun intended, sorry!) issue is that Feff
does a real space calculation of the atomic potentials for the atoms
in your cluster.  For a gas phase (or non-condensed molecule) the
'emptiness' on the outside of the cluster can make the calculation a
little less reliable.   In my experience,  I often put a few extra
waters (well, O atoms) around a small molecule just to help the
calculation.

Also, note that for Feff6 that comes with Ifeffit/Artemis, it's
recommended to take out all the H atoms -- they aren't see by the
x-rays anyway and make the potentials calculation harder.

I'm hope someone from the mailing list might be able to add more
insight on simulations for small molecules....

Hope that helps,

--Matt