[Ifeffit] large Eo's and neg. D-W factors

[Richard Mayes]

Scott,

I think I understand now.  By restraining the negative Debye-Waller 
factor, I may have been treating the symptom instead of the cause.  I'll 
go back and check my model.

Also, should I be worried if my Debye-Waller factors get small 
(0.0005-0.001) if I optimize my model in FEFF to a point that my delr's 
are on the order of 0.005-0.01 Angstroms?  This has been a concern of 
ours since we started doing EXAFS, especially since all of our 
measurements are at room temp.  We haven't seen this as much with our 
vanadium on silica samples, but it seems to be true with all of our 
titanosilicate samples

Rich

Scott Calvin wrote:
> Hi Rich,
> 
> Oh--I think I should have been more specific. It sounds like you're 
> restraining the sigma2's. In my experience, that is not as effective 
> in the cases of switching paths as restraining the delr's. After all, 
> the sigma2's are just a symptom; the delr's are the direct measure of 
> switching.
> 
> As for Anatoly's observation, I'd qualify it. If it's a matter of 
> trying to pound a square peg into a round hole, then yeah, that's 
> dubious. I often have found, though, that I get lousy fit after lousy 
> fit for a long time. Then I make some key change to how I'm 
> =physically= modelling the system...allowing vacancies, or putting in 
> a minority phase, or changing the morphology, or allowing for local 
> distortion, or whatever, and suddenly the fits get much much better. 
> I often have pretty high confidence in the results when that happens. 
> It suggests that in that system I can't get a good fit with the wrong 
> model just by tinkering. In the other extreme, if I have the kind of 
> combination of data and model that seems to give a "good" fit no 
> matter what I do, but from one fit to the next the parameters jump 
> all over the place, then I have low confidence in the results. In 
> other words, I want systems that generate lousy fits unless I get the 
> structure right. :)
> 
> --Scott Calvin
> Sarah Lawrence College
> 
> At 08:55 AM 5/2/2007, you wrote:
>> Scott,
>>
>> Thanks for the clues!  The restraints seem to be working to get the DW
>> factors positive.  I'm still battling Ifeffit reversing the paths, but
>> things are getting slowly better.
>>
>> Rich
>>
>>> Hi Rich,
>>>
>>> I know you posted this a while ago, so you may have gotten other
>>> answers, but I don't see them in my overflowing inbox. :)
>>>
>>> Years ago, I did some work with V2O3, and it was a very stubborn
>>> system to get a good fit on. Other people I knew who were working on
>>> the system have also had some trouble. This and other evidence led to
>>> some people publishing speculations on local distortions that get
>>> averaged out in XRD but show up in EXAFS. I also recall people
>>> talking about similar difficulties with other vanadium oxide systems.
>>> (Sorry; I don't have those references handy anymore, but searches
>>> should locate them).
>>>
>>> As far as the second issue you mention, with Ifeffit reversing the
>>> position of two atoms of the same species, it does happen some times.
>>> Why not? A V-O path is basically a V-O path to FEFF; it doesn't know
>>> that one's a single bond and one's a double bond. But since moving a
>>> path by more than about 0.1 Angstrom in Ifeffit starts to reduce the
>>> accuracy of the FEFF calculation, it's something you do want to try
>>> to prevent. One way to do that is with restraints; it shouldn't take
>>> much incentive for Ifeffit to refrain from switching the paths.
>>>
>>> --Scott Calvin
>>> Sarah Lawrence College
>>>
>>>
>>> At 05:02 PM 4/18/2007, you wrote:
>>>> Greetings!
>>>>
>>>> I have inherited a vanadium EXAFS data set (due to a group member
>>>> graduating - lucky dog!) that is giving me major fits.  I am getting
>>>> consistent numbers with all my fits which visually look fantastic,
>>>> that is until I look at the numbers...one Debye-Waller factor is
>>>> negative (a V-O path) and the Eo's are large, the V-Cl path is on
>>>> the order of -20 eV, the V=O path is on the order of 17 eV, and the
>>>> V-O path falling at -11 eV.  ARe there any "tricks" outside of
>>>> restraints or constraints that I can attempt to get more realistic
>>>> values?  (I understand that a large Eo can be realistic, but a
>>>> negative Debye-Waller factor doesn't make sense.)  I can restrain
>>>> the V-O Debye-Waller factor and get a positive value, but then the
>>>> V=O D-W factor is negative.  If I restrain both D-W factors, the fit
>>>> falls apart, suggesting something else is wrong that I just can't find.
>>>>
>>>> Another issue I'm having involves something in Artemis that's
>>>> confusing me.  Sometimes after a fit, the Artemis Pallet shows the
>>>> V=O feature as the one at 1.8 Angstroms and the V-O feature at 1.6
>>>> Angstroms, completely reversed to the FEFF model.  I have seen this
>>>> when I separated the chlorides as well (I attempted to model the Cl
>>>> at different distances as reported by Deguns, et. al. (attached) due
>>>> to the immediate environment around the V in our sample being
>>>> similar to the one they report in scheme 1-(I), p5008).  Attached is
>>>> a text file that shows an Artemis Pallet output illustrating what
>>>> I'm trying to describe for the V=O and V-O paths.  The Artemis
>>>> project that generated the output is also attached.
>>>
>>> _______________________________________________
>>> Ifeffit mailing list
>>> Ifeffit at millenia.cars.aps.anl.gov
>>> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
>>>
>>
>> _______________________________________________
>> Ifeffit mailing list
>> Ifeffit at millenia.cars.aps.anl.gov
>> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
> 
> _______________________________________________
> Ifeffit mailing list
> Ifeffit at millenia.cars.aps.anl.gov
> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit