Dear Matthew, Thank you very much for your reply! I appreciate your explanation about experimental lineshapes. I also confirmed that the "check for updates" is now updated! I am not sure about Voigt functions having negative gamma values. A quick search online indicated to me that gamma and sigma have to be always positive. A peak having negative tails also does not seem to be physically reasonable. I have one more question/feedback regarding the pre-edge peak fitting in Larix. Is it possible to add arctangent functions (including linear + arctan) as options for baseline fitting? Someone else also brought it up before, and your answer said you found linear+Lorentzian to work better than the arctangent. However, for my data, the linear + Lorentzian slopes upward at the beginning of the fit range: [image: image.png] I haven't tried the arctangents yet, so it is possible that they actually do worse. And maybe it is ok to have the baseline slope up like that (it's a tiny bit, anyways). Still, it will be nice to have the arctan option available, as it seems to be the most common way in the literature, and people like myself might want to compare how different options perform. Lastly, I couldn't find out how people judge whether a given baseline function is good or not -- is it just from visual inspection, or are there some objective criteria? And relatedly, if possible, I would like to understand how you found the linear+Lorentz to work better than arctan. Thank you so much! Best, Soyoung On Tue, Jan 14, 2025 at 5:48 AM Matthew Newville wrote:

Hi Soyoung,

1. In Help > Check for updates, it still shows 2024.12.0 as the newest version.

Thanks, this is (or should be) fixed now.

2. When I tried the pre-edge peak fitting for one of my data with a Voigt profile, it resulted in a negative value of Gamma, which should be unphysical (to my mind at least). It turns out that my pre-edge peak can be fitted very well to a single Gaussian. Is it an error that can be fixed?

I think that gamma for a Voigt function can be negative. That fits with my experience too: when gamma goes to 0, a Voigt is exactly Gaussian. When gamma is negative, it puts slightly negative tails near the peak (as if subtracting a Lorenztian, though not exact).

Using Gaussians for pre-edge peaks is completely normal – I think most of the literature does this. The pre-edge peaks have broadened from both “core-level widths” (this will give a Lorentzian profile, in theory) and from the source and monochromator resolution. The source resolution is Gaussian, and the monochromator resolution is more complicated but Gaussian-ish (sharper tails than Lorentzian). Usually, the source and mono dominate.

For some beamlines with very low source divergence, the Lorentzian term can be noticeable.

In fact, when doing HERFD, it is the core-width that is suppressed, so the peaks become more Gaussian.

3. In pre-edge peak fitting again, I tried first fitting the baseline, then fixing the parameters for baseline with what was fitted, then fitting the peak. Somehow, the resulting fit shows the same baseline parameters, but the plotted baseline is shifted downwards.

Sorry that is is fussier than it should be. The baseline looks good in your plot with gamma ~=0.3. The “baseline” should account for a linear pre-edge and the main line.

But also, after it is determined, you can alter the values for these components. You can even remove these components and add your own fit components to account for the whole curve.

--Matt

*From: *Soyoung Kim via Ifeffit *Date: *Monday, January 13, 2025 at 5:07 PM *To: *XAFS Analysis using Ifeffit *Cc: *Soyoung Kim *Subject: *[Ifeffit] Re: Larch 2025.1.1

Dear Matthew, Thank you for the kind email! I always appreciate your hard work for the XAS community. I have a few small feedbacks: 1. In Help > Check for updates, it still shows 2024. 12. 0 as the newest version. 2. When I tried the pre-edge

ZjQcmQRYFpfptBannerStart

*This Message Is From an External Sender *

This message came from outside your organization.

ZjQcmQRYFpfptBannerEnd

Dear Matthew,

Thank you for the kind email! I always appreciate your hard work for the XAS community.

I have a few small feedbacks:

1. In Help > Check for updates, it still shows 2024.12.0 as the newest version.

2. When I tried the pre-edge peak fitting for one of my data with a Voigt profile, it resulted in a negative value of Gamma, which should be unphysical (to my mind at least). It turns out that my pre-edge peak can be fitted very well to a single Gaussian. Is it an error that can be fixed?

3. In pre-edge peak fitting again, I tried first fitting the baseline, then fixing the parameters for baseline with what was fitted, then fitting the peak. Somehow, the resulting fit shows the same baseline parameters, but the plotted baseline is shifted downwards.

For 2 & 3, I have attached screenshots as well as the raw data. If these are issues too minor/trivial, please feel free to ignore them.

I would also like to ask if I may be added to the zoom meeting. By the way, I am on the west coast, so only after 11am your time will work for me. But if 9 am your time is the best for other reasons, it is ok. Perhaps being able to access the recording would be enough for me.

Thank you very much!

Sincerely,

Soyoung

On Sun, Jan 12, 2025 at 6:55 PM Matthew Newville via Ifeffit < ifeffit@millenia.cars.aps.anl.gov> wrote:

Hi Folks, Larch version 2025. 1. 1 has been released. This fixes a serious issue for generating Feff data files, and a few other issues too. It had been more than a year since fresh binary installers were made, so those have been updated too.

ZjQcmQRYFpfptBannerStart

*This Message Is From an External Sender *

This message came from outside your organization.

ZjQcmQRYFpfptBannerEnd

Hi Folks,

Larch version 2025.1.1 has been released. This fixes a serious issue for generating Feff data files, and a few other issues too.

It had been more than a year since fresh binary installers were made, so those have been updated too.

As always, suggestions and feedback are welcome. Since this is the first release of the year, and since we have an XAFS Conference and Q2XAFS meeting coming up this summer in Chicago (July 14-20), I encourage people to think about what they would like to see for these and other tools for XAFS and related work. And also: think about coming to the conference ;).

For those interested, we’ll restart monthly meetings to discuss the development of these tools later this month (probably January 24, probably 9 AM Chicago time). This is not development-only, but often about feature requests, complaints about the GUIs, and ideas. If you are interested, let me know, and I’ll happily add you to the invite list for that Zoom call.

We would like to hear feedback on:

a. Should we have a monthly Zoom meeting (maybe alternating time of day) as “Office Hours” ?

b. Should we send a survey of user needs and desires for features, recommendations, etc?

If you have opinions on any of these topics, let me know either here or by direct email.

--Matt

ifeffit mailing list: https://millenia.cars.aps.anl.gov/mailman3/lists/ifeffit.millenia.cars.aps.a... to unsubscribe, send mail to ifeffit-leave@millenia.cars.aps.anl.gov