Dear Bruce： I am really sorry that I lost this mail you send,because the mechanism of my mail address. Thanks a lot for your patience to give me detailed advise and strategy about my problems.I will take a step back and gain a deeper understanding of my sample and method itself. Thanks Sincerely, zhanfei

-----原始邮件----- 发件人: "Bruce Ravel" 发送时间: 2014年11月11日 星期二 收件人: "XAFS Analysis using Ifeffit" 抄送: 主题: Re: [Ifeffit] Fwd: Re: Re: problem about negative

On 11/11/2014 10:19 AM, Bruce Ravel wrote:

...

Rbkg>1.1 can avoid the first peak,but the remaining curve is somewhat not smooth.How to distinguish whether a long wavelength oscillation appear as a real peak or false one?I am confused about it. I find when I avoid the first scattering peak after white line(kmin=4.8 maybe too large),the first peak of R space disappeared.Because usually the first scatering peak is related to the first coordiantion shell,can I think the first peak in R space less 1A is a real peak instead of noise？

Zhanfei,

As I said yesterday, I don't know anything about your sample, so it is hard for me to know what the "right" answer is to your problem. But I can make some observations and some suggestions.

* Your data range is really short. As you seem to have noticed (judging from the very short k-range in your project file), your data has serious systematic problems starting at 9 or 10 inverse Angstroms. I cannot quite tell, but your data appear to be transmission data. Well it is possible that you have some instability or nonlinearity at the beamline, I would guess that your sample is not very homogeneous. It is possible that more disciplined sample preparation might help extend that range of the interpretable data.

* You are right that slightly increasing the Rbkg value seems to make a positive difference in the extracted chi(k) data. Your data seem to be an example of the sort of data I was alluding to yesterday. The background subtraction is difficult because it is difficult to distinguish the Fourier components of the background function from the Fourier components of the data. I think that the background subtraction with Rbkg=1.1 looks much better than with Rbkg=1.0, but given how different they are, you have to be concerned about the correlations between the background and the parameters of the data.

It is clear that your data are of the sort for which background subtraction is difficult. So how do you know what is an acceptable background subtraction?

Well, using only Fourier methods, I think we have demonstrated that you cannot know the right answer without some kind of prior knowledge.

So, how do you get that prior knowledge?

Well, my advice is to first solve some simpler problems. Measure your standards.

Measure the common forms of moly oxide and moly sulfide. Measure moly metal. Analyze all of them. The advantage of the standards is that you know what the answer should be. Do the data processing and data analysis. Make sure that, when you do the analysis, you get the right answers.

Having done that exercise, you will then have a lot more knowledge about what the various forms of moly oxide look like and what the challenges are when doing the data processing and data analysis.

Some questions:

- Are there any forms of moly oxide for which the bond length is as short as 1.6? If so, do the conditions of formation exist in your system? Is your sample of a valence consistent with the valence of the Mo when it has such a short bond?

- If you convince yourself that it is possible for moly to have an oxygen atom at 1.6 or 1.7, what did you have to do with your standard to get a sensible analysis of the EXAFS data? Hopefully, that will guide you to doing a sensible analysis of your unknown sample.

- If none of your moly oxide standards have an O atom at 1.6 or 1.7, why do you think it is chemically reasonable for your unknown sample to have such a short bond? If, in fact, that short bond doesn't exist elsewhere in nature, why would your sample magically have such a short bond?

- If you do not believe in such a short bond, then does your experience with the standards give you the confidence to increase Rbkg such that the low-R signal is removed by the background function?

I completely understand that you have a compelling reason to understand your unknown sample and that I am suggesting that you spend a good chunk of time measuring and analyzing a bunch of standards that are not your actual research project. That might seem like time that takes you away from your real goal, but you have already demonstrated that you are stumped by your real goal. I am saying the only way to get over your current hurdle is to take a step back and gain a deeper understanding of the data, the physics of EXAFS, the methods of EXAFS analysis, and the intricacies of your current problem.

B

-- Bruce Ravel ------------------------------------ bravel@bnl.gov

National Institute of Standards and Technology Synchrotron Science Group at NSLS --- Beamlines U7A, X24A, X23A2 Building 535A Upton NY, 11973

Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/ _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit