At ALS, we have a box of foils from EXAFS Materials that seems to get 
passed around from beamline to beamline.  I ran absorption scans on 17 
edges from the metals in the box one day, and found that there was 
considerable scatter in the expected vs observed edge positions:
http://bl831.als.lbl.gov/~jamesh/pickup/mono_calib.png
Here I have plotted the "correct" position of each edge as determined by 
Bearden & Burr (1967), against the edge I determined using the criterion 
recommended in the "Reference Spectra" document in the 
exafsmaterials.com website: the first inflection point in the derivative 
spectrum.  I think a large amount of the scatter is because my mono 
(like many PX/MX beamlines) is Si(111) and not Si(220) like the one used 
to determine the reference spectra.  It is not hard to imagine how 
blurring the spectrum with a wider energy spread of the incident beam 
will shift the position of the "edge".  One could try to use the 
electron binding energy tabulated in the "little orange book":
 http://xdb.lbl.gov/Section1/Sec_1-1.html
but these do not always take into account the "near edge" features (like 
the white line from SeMet) which change depending on the chemical 
environment around the metal, radiation damage, etc.  It would be nice 
if someone could calibrate some standard reference materials using a 
Si(111) monochromator, but I don't know of anyone who has done this.


However, another way to get your x-ray wavelength is using Bragg's law:
lambda = 2*d*sin(theta)
and the d-spacing of silicon is known to be 5.43159 ± 0.00020 A, and 
NIST will sell you certified Si powder:
https://www-s.nist.gov/srmors/view_detail.cfm?srm=640d

The problem here is that although you know d very accurately, the error 
in lambda is dominated by sin(theta), or rather the uncertainty in your 
detector distance.  The pixel field on most detectors is actually quite 
accurate, as a NIST-traceable calibration is used to make the pinhole 
calibration mask, and the encoder on most detector distance stages is 
very accurate for relative moves (counting ticks on the encoder).  But 
there is always an offset from the "zero" position predicted by the 
encoder to the true center of rotation that is hard to know. 

Nevertheless, all you really want is for the d-spacing of silicon powder 
rings to be right at all detector distances. You can use the program 
FIT2D to refine the wavelength, distance, detector tilt, etc. or any 
combination thereof for a given image, but you will find that the 
repeatability of such a fit (using different starting parameters) is not 
great because the distance and wavelength are highly correlated.  
However, there is a way around this:

Since we know that a relative move of the distance will be accurate, 
there should be one and only one offset that you can add to the recorded 
value of distance of each image to make it the "right" distance.  You 
can define the "right" offset as the one where FIXing the resulting 
"right" distance in FIT2D and refining everything else gives you the 
same refined value for the wavelength from every image.  You need to 
manually "refine" this offset for a few rounds.  What you will generally 
see is that the graph of fitted wavelength vs the distance is a straight 
line, and you want to make the slope of this line to be zero.  
Eventually, you will arrive at some offset that gives you the smallest 
spread in refined wavelength values.  The average refined wavelength is 
then the "true" wavelength.  Should be able to get it within one or two 
eV.  Perhaps more if you take a lot of silicon powder images.

At ALS beamlines 8.3.1 and 12.3.1 I have done both kinds of calibration, 
and I am fairly certain I get the wavelength accurate to within 1 eV by 
calibrating the half-way-up point of an absorption scan of a ~122 micron 
thick copper metal foil to 8979.0 eV.

-James Holton
MAD Scientist


Richard Gillilan wrote:
> In the past we've used elemental foils from exafsmaterials.com for 
> energy calibration of our MAD beamline. These standards are for EXAFS 
> and XANES. Most are thin (5 micron) metal foils.
>
> Has anyone had experience with other sources of standards or other 
> forms (such as compounds rather than pure elements)?
>
> I notice that a number of companies offer XRF standard kits.
>
> Richard Gillilan
> MacCHESS