Scott Calvin's XAFS Divination Set
Abstract
This is a set of actual XAFS data gathered especially as a "test set" for analysis. Why do that? Several reasons are suggested below, under "Uses."
Details
The set is focused primarily on iron oxides, but also includes metallic iron and a few iron salts with organic anions. The data was collected on beamline X-11B of the NSLS, and is of moderate to good quality, depending on the sample. Multiple scans of each sample are included, and they are cut off before the cobalt edge. An iron foil was used in the reference channel.
Important: To keep this from being easily solved by linear combination methods alone, and to simulate the common situation encountered in XAFS analysis of having a sample that is "almost" like a mixture of bulk materials, every sample was collected at a random temperature between 303 and 403 K. That temperature is not reported in the data set below, in order to simulate the kind of differences between samples and standards that are not known a priori.
Standards
Standards collected on the same beamline, but not necessarily at the same temperatures.
Known Constituents
In this set, the identity of the constituents are known, but the fraction of each that is present in the sample is not known (to you, that is). This is the easiest set to analyze (after the standards), and is good for beginners to use to test their proficiency and get practice at the technique.
Unknown Constituents from a Known List
In this set, 2-3 constituents from the list of standards were mixed in an unknown (to you) ratio. This is a somewhat more difficult set, but should be analyzable by any seasoned XAFS researcher.
Known Constituents and one Mystery Constituent
Here the identity of all but one of the constituents are given to you...but the last constituent is an unknown simple organic compound (a salt?) involving iron. If I'm a little vague, it's because I haven't attempted the analysis yet, so only the colleague who ordered the material and the undergraduate who made the random assignments and prepared the samples knows what it is. I expect this to be an interesting challenge for seasoned XAFS researchers. The task is comparable to that often faced in published analyses.
Unknown Constituents
This is the tough one. One or two of the constituents is from the list of standards (but you don't know which ones), and the other is the mystery constituent from the above set.
Uses
I can think of a number of uses for this set.
- As practice exercises for those learning XAFS analysis
- As a proficiency test: new research assistants could be asked to achieve a certain accuracy with some of these sets before being trusted with important analyses
- For fun. Just once, wouldn't you like a XAFS analysis problem where you could check yourself afterward?
- To compare the accuracy of different techniques, approaches, software...
- For My Nefarious Plan (see below)
My Nefarious Plan
My primary purpose for creating this set is to get some baseline figures on the accuracy of real-life XAFS analysis. There have been attempts in the literature to estimate this from the bottom up, that is, by estimating the uncertainty in each step of the analysis. My project is to try to estimate this from the top down: I'll get actual practitioners to do an analysis on part (or all) of this set and then report to me their results. This is then a double-blind study of the accuracy of XAFS analysis as it is practiced. I plan to then publish the results on accuracy, creating a publication we can all point to ("In similar systems, XAFS analysis has been shown to be able to identify the proportion of phases present with an uncertainty of ##%"). The publication will contain no identifying information as to who performed which fits. Participants will be included in the acknowledgments if they choose.
Getting the Answers
To find out the correct answers, and incidentally participate in my nefarious plan, send an email to "SCalvin at mailaps dot org." Include the results you have gotten, as much detail as you'd like on how you got them, what you think your level of expertise is, and about how long you worked on the problem. I (or my undergraduate, if I have not completed the analysis myself) will then email back the correct percentages and phases for the set(s) you've specified. Although the minimum requirement is phase identification and fraction of each phase, nearest-neighbor bond lengths and would also be useful, since these samples and standards were collected at various temperatures.
Example: You could send "I am a graduate student learning EXAFS analysis. I've found that sample B5 consists of 37 +/- 4 % Fe, 41 +/- 8% Fe2O3, and 22 +/-8% alpha FeOOH. The nearest neighbor bond distances in the , Fe, Fe2O3, and FeOOH is 2.53 +/- 0.02 angstroms. The nearest-neighbor distances in the oxidized compounds appeared reasonably consistent with the crystallographic values, but I could not determine them individually due to the complications of multiple phases. I used linear combination of XANES to get the phase percentages and phase identifications, and then fitted to feff calculations to get the bond lengths and confirm the XANES results. The software used was Athena and Artemis. This took about six hours of my life, which I'd like back, but my Ph. D. adviser says it's good for my soul." I'd then send back the actual percentages and temperatures.