Of course I tried PCA!  Over a certain range of temperatures, one component looks tolerably like magnetite.  The other is sort of,
but not exactly, maghemite.  However, if I go below that restricted range, it's not pseudobinary anymore.  Also, I still don't know
what the "maghemite-like" end-member actually is.  Still, for the purposes of the ALS user-meeting poster I have to hang in a week or
so, I'll probably take the PCA analysis plus a FEFF fit to the one end-member, and try to do a more thorough job when I have time.
No, it doesn't fit well to a combo of the references I have, which does include maghemite.
 
To be fair, what I have is not an isothermal series but an isochronal series at different temperatures, so there's less reason for
expecting simple kinetics than in an isothermal series.
    mam
----- Original Message -----
From: Anatoly Frenkel
To: MAMarcus@lbl.gov
Sent: Monday, September 24, 2007 6:59 PM
Subject: RE: [Ifeffit] Artemis fitting using FEFF paths plus experimentalreferences

Matthew - please forward my reply to the list if you consider it appropriate. I am spammed out. Anatoly
 
*************************
 
Matthew,
 
May be it is not relevant, since I do not know if you have a series of EXAFS data at different stages of the oxidation, or just the initial and the final stages. If you do have several stages, including the starting (unoxidized) and final (oxidized, or a sum of oxidized and unoxidized, depending where you stopped at) ones, the principal component analysis seems to match your needs well.  I assume you also have access to EXAFS data in bulk magnetite and maghemite references.
 
PCA analysis may tell you (if you are lucky and your system is similar to those where such information was previously obtained):
 
1) The volume fractions of intermediate species at each stage of the oxidation:
2) Identities of each intermediate species
 
The first part is entirely model-independent. The second part is not, but a reasonable selection of standard compounds (e.g., magnetite or maghemite) may help you isolating the intermediates. The good thing is that you do not need to fit a Fe-Fe and Fe-O bonds to the spectrum of the entire heterogeneous sample, where these contributions may be present in different species, and you have a large number of fitting parameters that describe such a model. Instead, you may end up obtaining that your species are identical with their bulk reference compounds (and thus no fit is required since the only thing you need to know is the volume fraction of this species) or, you may have to fit a couple of FEFF paths to that unknown species, after you deconvolute it from the mixture knowing what the rest of the phases are (it is doable in the case of one intermediate, but can be done in the case of a larger number of intermediates as well).
 
Anatoly

-----Original Message-----
From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov]On Behalf Of Matthew
Sent: Monday, September 24, 2007 7:05 PM
To: XAFS Analysis using Ifeffit
Subject: [Ifeffit] Artemis fitting using FEFF paths plus experimentalreferences

I have a system in which Fe:C alloy nanoparticles are being oxidized.  I have the EXAFS of the unoxidized nanoparticles, and I'd
like to use that as a reference
for fitting as in:

Fe_EXAFS(oxidation_stage) = x*(unox_particles)+(1-x)*(Fe->O paths + Fe->Fe paths +...)

where the Fe->O and Fe->Fe are based on magnetite and maghemite, which seem to be products of the oxidation.  It's not just a linear
fit to
unoxidized+magnetite+maghemite because the oxide is somewhat disordered. If I do Artemis fits to the shells from the two oxides
alone, I find I lose
some Fe atoms (sum of fractions < 1), a result which is echoed in linear-combo fits to the oxides alone.  Thus, I really want to do
a hybrid fit.  The unoxidized
particles are highly disordered to the point where they don't fit well with reasonable shells, so I'd rather not have to represent
them as FEFF shells.

How can I do this hybrid fit?
    mam

_______________________________________________
Ifeffit mailing list
Ifeffit@millenia.cars.aps.anl.gov
http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit