[Ifeffit] Using 3 or more CIF file for fitting in Artemis

[Matt Newville]

Hi Esmael,

On Fri, Dec 16, 2022 at 3:36 AM Esmael Balaghi <
esmael.balaghi at fit.uni-freiburg.de> wrote:

> Dear all,
>
> Following the previous discussion (
> https://www.mail-archive.com/ifeffit@millenia.cars.aps.anl.gov/msg07118.html
> ) on using more than one CIF file for fitting, as Scott Calvin and Matthew
> Marcus suggested, we can fold the molar ratio into the amp parameter for
> each CIF. Does it mean that the overall S02 for all phases (all used CIF
> files in the fitting) is "set" to 1?
>

I would recommend thinking of S02 as a value that belongs to all paths of a
data set, and is part of - but not the whole of - the amplitude for any
path.

Typically, for the regular fitting with one model (cif file or any
> simulated feff file), we take the S02 value from the known metal foil that
> has been measured simultaneously with the unknown sample.
>
> So if we used the mentioned method and folded the molar ratio into the
> amp, does it mean in Artemis we use the following terms?
>
> "in Artemis: S02= abs(amp) for one phase and for another one: S02=
> abs(1-abs(amp)" and leave “amp” parameter to “guess”
>

When mixing structures from different CIF files there can be a slightly
tricky issue that each path file has a coordination number built in, and
this might be different for different structures. Like, for a perfect
octahedron, this "degen" would be 6, but if another structure with the same
neighboring atom has a distorted octahedron, some paths might have a degen
of "1" or "4" or something else.

If that is the case, I would suggest forcing all the "degen" values to 1,
and then using "S02 * N1"  for the amplitude term of path 1 and "S02 * N2"
for the amplitude term of path 2, etc, and then define/constrain N2 to be
"6- N1" (or whatever you think is appropriate total coordination is).

On the other hand, if you are mixing different phases, and you expect (for
example) there is some Pt metal that is well-defined by a single structure
("CIF") and also some Pt oxide (also with a well-defined structure), and
the question is "how much oxide is there", then you might leave the path
degeneracies as calculated from the structures and use

"S02 * Fraction_Ptmetal" for the Pt-Pt Path from the Pt structure, and "S02
* Fraction_PtOxide" for the Pt-O Path from that structure, and then
define/constrain "Fraction_PtOxide = 1 - Fraction_Ptmetal".

Either way, you can set "S02" as appropriate (say, found using a spectrum
of a standard with well-known coordination", and you have easy access to
trying different combinations of coordination numbers or compositional
fractions.


Also if we are using X, Y, Z,.... CIF files, does it mean the summation of
> all S02 for all X,Y,Z,... cif models would be equal to 1? Im not sure, but
> I guess in this work (
> https://www.sciencedirect.com/science/article/pii/S0016703702008888 ),
> the authors used this way and fixed the overall S02 for all phases to 1.
>
I would appreciate having your comment on this method or if you have any
> suggestion on how to deal with the S02 value in Artemis when we are using
> more than 2 or 3 phases (cif files) for the fitting process.
>

If the idea is that you have a mixture of Pt metal, PtO, and PtCl, you
could use path amplitudes of
"S02 * Fraction_Ptmetal"
"S02 * Fraction_Ptchloride"
"S02 * Fraction_Ptoxide"

let the first two fractions vary and define/constrain "Fraction_PtOxide = 1
- (Fraction_Ptmetal + Fraction_Ptchloride)".

--Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20221216/d378fd15/attachment.htm>