[Ifeffit] how to use the data in the XAN library to estimate the transmission of an x-ray filter

[Matt Newville]

Hi Nino,


On Tue, Apr 12, 2016 at 9:08 PM, pereira <ninorpereira at gmail.com> wrote:

> Hi All,
>
> new to this list and to the field, so please be patient and kind
> and helpful in any response you may have. So far everyone I've
> reached on this topic has been extraordinary good. I still have to
> learn the lore though.
>
> I'd like to compute the x-ray transmission as function of the x-ray
> energy, for a particular mass per unit area of a substance, say Zn.
> In the XAS data base I find Zn foil at room temperature, recently measured
> and even more recently put into the data base just in time for
> my purpose. Thank you!  The question is:  is there a ready-made program
> to use those data? I see that your community has a whole slew of
> programs that do the opposite, go from the attenuation and/or fluorescence
> measured on a beam line to the attenuation coefficient, apparently without
> having to know the mass per unit area of the filter. I see Demeter,
> Athena, Larch, and various others but they all seem to go the wrong
> way for (and much further than) what I need.
>
> Instead, I'd like to pick up the data you guys measured on the synchrotron,
> and convert those into the X-ray transmission of a filter for which I
> select
> the mass per unit area.
>
> In trying to figure out how to do this I look at the graphs in the
> database,
> and the numbers behind Zn foil.xdi.
>
> The graph on the top left side gives the "Raw XAFS" (y, say) as function
> of energy.
> I interpret that as y = - ln (itrans/i0), where i0 and itrans are in
>  Zn foil.xdi
> as the intensity of the incoming x-ray beam, and the beam behind the
> particular filter. Far from the edge, at 9600 eV, I see i0=93769.049842
> and itrans=56429.849906, so that y=0.508. I can't tell from the graph
> whether this
> is indeed the case, so I go to the NIST tables. They give 35.05 cm^2/g on
> the
> low side of the edge. Then, the mass per unit area of this particular foil
> must have been 14.5 mg/cm^2.
>
> At 10000 eV the same thing gives y=3.603. This is pretty close to what I
> see in the graph, so I'm tempted to think that guessing "Raw XAFS"=
> ln(i0/itrans).
> However, for (mu/rho) NIST gives 233.1 cm^2/g, so I get for the foil's mass
> per unit area 16 mg/cm^2. Not quite right.
>
> Any comments on what I might have done wrong?
>
> It seems to me that by measuring the transmission of a particular filter on
> some X-ray source we might have available would then in effect measure
> the mass per unit area, so that this can be taken into account by simply
> exponentiating itrans/i0 with the right exponent ( = multiplying the
> logs with the right factor).
>
> Is something like this already implemented in one of your programs? If so,
> which one?
>
>
For XANES and EXAFS we (most of us, anyway) tend to not care about the
value of the mass attenuation in cm^2/gr, though we do pay attention to
total  attenuation in cm^-1, at least insofar as we get the sample
thickness correct.     The measurements we report are rarely in true units,
as I0 is typically a sampled current from an upstream ion chamber and I1 is
a sampled current from a downstream ion chamber -- we don't bother
converting these numbers to flux just before and just after the sample --
we know there are arbitrary scale factors, but they tend to cancel out.

The step in the reported -log(I1/I0) should be a good measure of the edge
step in mu*x, and we typically aim for value around 1 (so that going across
the edge changes the attenuation by 1/e).


I think what you want are good values for mass attenuation for the Zn K
edge.  These are tabulated and available in the Hephaestus program (that is
part of the Demeter package), and also in Larch.    For example, in Larch,
you could do
     larch> en = linspace(9500, 10000, 501)
     larch> mu = mu_elam(30, en)
     larch> plot(en, mu)

Here, mu (in cm^2/gr) will go from about 35 below the edge to about 250
above the edge.    Of course, these tables are atomic, and have no chemical
or structural effects included.   But, a simple offset and scaling of the
measured mu to match these atomic values should be pretty good.

Hephaestus also has a nice calculator for the 1/e thickness of a sample (in
your case, filter) of a given composition and density, at a particular
energy.

Hope that helps!

--Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20160413/3cb32852/attachment-0001.html>