Hi Matthew, Jana, I think the Chantler values, especially in Hephaestus, are not particularly robust at the Ca L edges. To be clear, Elam gives L3, L2, and L1 energies as 346.2, 349.7, and 438.4 eV, and the edge jumps as 5.8, 1.4, and 1.1. I believe those edge jumps may have originated from Shaltout -- maybe Bruce can clarify that. The Chantler data from the NIST FFast web page (and in Hephaestus) are quite sparse. This is a definitely a problem for using the anomalous scattering factors near edges. I've talked with Chris Chantler about this a few times over the years. Not too long ago, he sent me data on a finer grid -- but he also told be recently that he hoped to have even better data he could send to me soon (all time-scales here on months-to-years here). I've included the finer data I have from Chantler into Larch. But the results for the Ca L edges are still not encouraging. The attached figure and ASCII data file give the results for mu(E) (gr/cm^2) from Elam and from Chantler. It's hard to see an L2 edge in either, and Chantler does not show an L1 edge. FWIW, the script to generate this is: #################### energies = linspace(300, 500, 101) muca_chantler = mu_chantler('Ca', energies) muca_elam = mu_elam('Ca', energies) newplot(energies, muca_chantler, ymax = 50000, label='Chantler') plot(energies, muca_elam, label='Elam') info_head = 'Ca edge Energy(eV) Fyield EdgeJump' info_l3 = ' L3 %.1f %.5f %.2f' % xray_edge('Ca', 'L3') info_l2 = ' L2 %.1f %.5f %.2f' % xray_edge('Ca', 'L2') info_l1 = ' L1 %.1f %.5f %.2f' % xray_edge('Ca', 'L1') write_ascii('CaMu.dat', energies, muca_elam, muca_chantler, info_head, info_l3, info_l2, info_l1, label='Energy MuCa_Elam MuCa_Chantler') ######################## I'm not sure that gives a lot of insight except that not trusting Chantler's values for these values might be reasonable. On Thu, Sep 24, 2015 at 10:16 AM, Matthew Marcus wrote:

I'm not after absolute data, just the edge-jump ratio. This would have to be extracted by peak+arctan fitting because any spectra will have peaks and a very limited range between edges. If the Chantler numbers are incorrect, then perhaps the edge-jump ratio is really 2.

Do you have a reference which can be cited?

I'll try the CXRO tool next, since CXRO specializes in soft X-rays. mam

On 9/23/2015 11:49 PM, Jana Padeznik Gomilsek wrote:

...

It is very hard to measure or to calculate absolute absorption data, especially in the vicinity of the absorption edges and especially in the soft x-ray region. Therefore there are significant differences between the tables and I think nobody knows which are better. Chantler, for example, says the expected uncertainties of the tables in your region are 50 % to 100 % ( http://physics.nist.gov/PhysRefData/FFast/Text2000/sec06.html#tab2). I would doubt the Chantler's L3+.1 number, all other numbers look ok - this is what you can get.

jana padeznik gomilsek

Message: 3

...

Date: Wed, 23 Sep 2015 18:02:08 -0700 From: Matthew Marcus To: XAFS Analysis using Ifeffit Subject: [Ifeffit] Problem with Hephaestus at Ca L-edges Message-ID:<56034B90.70405@lbl.gov> Content-Type: text/plain; charset=utf-8; format=flowed

I wanted to work out the edge-jump ratio between the L3 and L2 edges of Ca using Hephaestus. I ran into two problems:

1. The ratio implied by what it says for the unit-edge-step thickness does not agree with that derived by computing the absorption (cm^2/gm) above and below each edge and dividing the difference (L3+ - L3-)/(L2+ - L2-).

2. The results differ wildly depending on which resource I use:

L3-.1 L3+.1 L2-.1 L2+.1 (L1+ - L1-)/(L2+ - L2-) Elam 4759.796 27837.796 27478.018 38434.277 2.106375908 Chantler 4322.6 6547.121 32827.61 35436.543 0.852655473 Cromer-Leiberman 4288.524 33471.375 32786.294 47072.991 2.042659055

The Henke table doesn't yield an L2 edge jump at all, while the Shaltout yields the same results as Cromer-Leiberman. Which one should I trust and why?

This is old-style H. (V0.18), not Demeter. mam

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

--Matt

On 09/24/2015 12:47 PM, Matthew Marcus wrote:

What about that seeming inconsistency between the values H. gives for the thickness needed for unit edge step and the cm^2/gm values reported? That seems to be a bug indpependent of the tables used.

I don't know what you're referring to, but I am happy to accept patches or explanations of what to fix. Here's the relevant code: https://github.com/bruceravel/demeter/blob/master/lib/Demeter/UI/Hephaestus/... B -- Bruce Ravel ------------------------------------ bravel@bnl.gov National Institute of Standards and Technology Synchrotron Science Group at NSLS-II Building 535A Upton NY, 11973 Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/

On 09/24/2015 01:43 PM, Matthew Marcus wrote:

See the attached ppt for the demo. I think this gives the whole repeat-by and interpretation. Am I just missing something? That would not be totally surprising.

The algorithms in Hephaestus are pretty dim-witted and tuned (such as they are) to getting the calculations more or less right in the hard X-rays. I don't even try to do calculations in the soft x-ray range correctly. For example, the edge step calculations do the calculation at something like +/-50, which is a very stupid choice for soft x-ray energies. That should explain why L2:l3 ratio is 1. Presumably also why the L1 step is negative! See, e.g. https://github.com/bruceravel/demeter/blob/master/lib/Demeter/UI/Hephaestus/... You cannot trust anything Hephaestus has to say about edge jumps below 3-ish keV. Since the +/-50 is hard wired in, the jump ratios will be nonsense for any element for which the energy difference between the L2 and L3 edges is less than 100 eV. The Delta mu/rho numbers that you cite are interpolated from the Elam (usually) tables and should reliable to whatever extent you consider the table reliable. B -- Bruce Ravel ------------------------------------ bravel@bnl.gov National Institute of Standards and Technology Synchrotron Science Group at NSLS-II Building 535A Upton NY, 11973 Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/

On 09/24/2015 03:35 PM, Matthew Marcus wrote:

OK. Now I get it. Why use 50eV? Since lifetime and other broadening is not taken into acocunt in any of these tables, why not use a very small number?

Who knows? I wrote that bit of code a long time ago. Whatever the reason was that I chose +/-50 eV -- I am sure it was brilliant! The next release of Demeter will do this much more sensibly, as seen in the attached screenshot. Thanks for bringing this up. B

mam

On 9/24/2015 11:07 AM, Bruce Ravel wrote:

...

On 09/24/2015 01:43 PM, Matthew Marcus wrote:

...

See the attached ppt for the demo. I think this gives the whole repeat-by and interpretation. Am I just missing something? That would not be totally surprising.

The algorithms in Hephaestus are pretty dim-witted and tuned (such as they are) to getting the calculations more or less right in the hard X-rays. I don't even try to do calculations in the soft x-ray range correctly. For example, the edge step calculations do the calculation at something like +/-50, which is a very stupid choice for soft x-ray energies. That should explain why L2:l3 ratio is 1. Presumably also why the L1 step is negative!

See, e.g. https://github.com/bruceravel/demeter/blob/master/lib/Demeter/UI/Hephaestus/...

You cannot trust anything Hephaestus has to say about edge jumps below 3-ish keV. Since the +/-50 is hard wired in, the jump ratios will be nonsense for any element for which the energy difference between the L2 and L3 edges is less than 100 eV.

The Delta mu/rho numbers that you cite are interpolated from the Elam (usually) tables and should reliable to whatever extent you consider the table reliable.

B

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

-- Bruce Ravel ------------------------------------ bravel@bnl.gov National Institute of Standards and Technology Synchrotron Science Group at NSLS-II Building 535A Upton NY, 11973 Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/

OK, at least we've come to convergence about what 'edge jump' means in this context. Do you think the Elam value for (mu(+)-mu(-))_L3/(mu(+)-mu(-))_L2 is correct? mam On 9/24/2015 10:36 AM, Matt Newville wrote:

Hi Matthew,

On Thu, Sep 24, 2015 at 11:47 AM, Matthew Marcus mailto:mamarcus@lbl.gov> wrote:

What I was after was L3/L2, so I don't care about L1. The attached plot doesn't show the jumps for both L2 and L3.

When you say that the edge jumps are 5.8,1.4 and 1.1, are you referring to mu(+)/mu(-)? I was looking for (mu(+)-mu(-))_L3/(mu(+)-mu(-))_L2.

Sorry, finer data (0.20 eV steps) across the L3 and L2 edge attached. Chantler's data clearly do not give good values for the Ca L2 edge.

The edge jumps I reported are those reported in the Elam tables, which are described there (see https://github.com/XraySpectroscopy/XrayDB/blob/master/src/elam.dat#L79) as

mu(+) DIVIDED BY mu(-).

That clarifies those values for me, anyway! If I do (mu(+)-mu(-))_L3/(mu(+)-mu(-))_L2 with the Elam values for mu, I get 2.1.

Revised script:

############################# info_head = 'Ca edge Energy(eV) Fyield EdgeJump'

info_l3 = ' L3 %.1f %.5f %.2f' % xray_edge('Ca', 'L3') info_l2 = ' L2 %.1f %.5f %.2f' % xray_edge('Ca', 'L2') info_l1 = ' L1 %.1f %.5f %.2f' % xray_edge('Ca', 'L1')

l3_en = xray_edge('Ca', 'L3')[0] l2_en = xray_edge('Ca', 'L2')[0] l3_jump = mu_elam('Ca', l3_en+1) - mu_elam('Ca', l3_en-1) l2_jump = mu_elam('Ca', l2_en+1) - mu_elam('Ca', l2_en-1)

info_jump = 'Jumps: L3=%10.1f, L2=%10.1f, L3/L2=%10.3f' info_jump = info_jump % (l3_jump, l2_jump, l3_jump/l2_jump)

energies = linspace(340, 360, 101) muca_chantler = mu_chantler('Ca', energies) muca_elam = mu_elam('Ca', energies)

newplot(energies, muca_chantler, ymax = 50000, label='Chantler', xlabel=r'$E\rm\, (eV)$', ylabel=r'$\mu\rm\, (cm^2/gr)$', show_legend=True, title='Ca L3, L2 edges') plot(energies, muca_elam, label='Elam')

write_ascii('CaMu_fine.dat', energies, muca_elam, muca_chantler, info_head, info_l3, info_l2, info_l1, info_jump, label='Energy MuCa_Elam MuCa_Chantler') #############################

What about that seeming inconsistency between the values H. gives for the thickness needed for unit edge step and the cm^2/gm values reported? That seems to be a bug indpependent of the tables used.

Not sure.

--Matt

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit