Hi Todd,

I have some familiarity with the beamline where these data were collected so I will take a stab at what might have gone wrong and what you may be able to do to salvage some of the data.

Ideally, you want an Io(E) that is as flat as possible over the scan range, or at least over the region where the amplitude of your EXAFS signal starts rapidly declining, say above 8 A-1. We typically try to get a 10-20% smooth variation in Io intensity over the scan range in our measurements, which works for the setup and the systems we are interested in. The reason for this “requirement” is that while normalization of If or It to Io should take care of all variations in your incident photon intensity, your detectors respond differently to large changes in intensity (even if your detectors were exact replicates, which they are not, they are still using different gases in Io, It, and If). This non-linearity in the detector responses will result in any noise and EXAFS-like variations present in Io to also show up in your normalized If/Io or log(Io/It) spectrum and to interfere with the EXAFS signal you are trying to analyze. The problem will be exacerbated for elements like Pb at the L3 edge, where the EXAFS signal is of smaller amplitude to begin with so the noise and features resulting from non-linearity will be a bigger proportion of your measured signal.

What seems to have happened in your measurement is that vibrations or temperature drifts or other mechanical instability caused the second monochromator crystal to move away from the Bragg condition, so your Io was likely travelling randomly along the rocking curve, causing large changes in measured Io intensity and changes in Io(E) structure from scan to scan. This beamline has a feedback loop that attempts to keep the incident intensity at the top of the rocking curve, but it may not have been doing its job at the time of your measurement or it may have been turned off to reduce the noise in Io resulting from the induced oscillation in the second crystal that creates the difference signal used in the feedback loop. These problems in Io could then be transferred in your data due to non-linearities in your detectors’ response. It is also possible that the harmonic rejection mirror was not set up ideally and allowed harmonics in the incident beam, which would also result in non-linear response and spectral artifacts, particularly when your beam travels up and down the rocking curve.

I guess the bottom line is that you now have a data set with individual scans that appear different from each other. I took a look at the project file you sent and it seems that the data are relatively repeatable scan to scan when you normalize and extract the EXAFS, despite what you show as large variation in Io and in the un-normalized data in your email. For datasets like that, I would not use the scans that differ significantly from the others and the scans that show large noise at high k (e.g. 006, 008, 010). When you average the relatively repeatable remaining scans you should be able to get workable data up to 9-10 A-1 or so. Whether you can tackle the research question you set out to answer with data like those depends, of course, on the system and the question…

Hope that helps,

Max.

From: Ifeffit [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov] On Behalf Of Luxton, Todd
Sent: Thursday, March 31, 2016 9:28 AM
To: ifeffit@millenia.cars.aps.anl.gov
Subject: [Ifeffit] Question about Io during data collection

All:

Recently our group was at the APS collecting Pb L(III) spectra on an ID line using quick XAFS. Data was collected from -200 to +800 eV for Pb L(III) at 0.2 eV steps with a count rate of 0.025 seconds. Each scan took about 2.5 minutes to complete. During our measurements we began noting issues with the linearity of scans collected in the extended region of the same sample. After poking around in the data we noticed that Io was not linear throughout the measurement for a portion of the scans. This was not always the true (see Athena project attached to the email, and images attache to the email). I was always under the impression that Io should ideally remain linear throughout the energy region scanned, or at least remain unchanged between replicate scans. We worked with the beamline scientists to try alleviate/fix the issue, but it kept persisting. I realize this is not an IFEFFIT issue, but I was hoping someone might be able to help me understand what was going on and why this was happening? If I have not included enough information for the question please let me know.