Multipeak Fitting 2

(Igor Pro 6.37)

"Auto-locate Peaks Now" is missing a few sidebands for my spectrum file. I know I can "Add or Edit Peaks" manually to add peaks/sidebands, but it's very difficult to find the best fit by doing it manually. Can I specify the total number of peaks and pick the peak positions, and then ask Igor Pro to do the fitting considering all peak positions chosen? The Origin demo below shows what I was describing:

https://www.youtube.com/watch?v=Yi0Pmc4v2ts&t=1m48s

Thank you.

It's kind of hard to tell exactly what the Origin demo is showing, but the background music is nice :)

It looks to me like that demo is showing how to add initial guesses for peaks to the set of peaks to be fit. If so, that is precisely what Add or Edit Peaks is for.

In reply to by johnweeks

In the Origin demo, looks like one can simply pick the locations of the peaks and Origin will do the rest. But in Igor Pro's "Add or Edit Peaks", I have to add and EDIT the peaks myself, adjusting the shapes of the peaks several times and still not getting a best fit. Or did I get it wrong? It seems I am missing something when Igor Pro doesn't do additional peak fitting for me.

You use Add or Edit peaks to adjust the initial guesses before doing a fit. The assumption is that if you are using Add or Edit Peaks, the auto-peak locator didn't work well, or missed one or more peaks. Since the auto-locator didn't work well, you are showing Multipeak Fit where you think the peak is, how wide it is, and what the amplitude is. After you click the Done button in the Add or Edit panel, you will see the peaks you added or edited on the main graph. At that point, you can click the Fit button to get Igor to do the fit using the collection of guessed peaks as the starting point.

Have you looked at the demo yet? File->Example Experiments->Curve Fitting->Multi-peak Fit 2 Demo. It includes a step-by-step guided tour; the third part of the guided tour includes the use of the Add or Edit panel.

Can you post an Igor experiment file with a data set in it?

@isend: I see three aspects of your post.

First, your question seems to be whether a user can do the same type of multiple peak fitting UI operations in Igor Pro as a user can do in Origin. This is suggested by your comments to the extent "one can *simply* ... and Origin will do the rest. But in Igor Pro ..., *I have to ... myself*." I think the answer here is that one UI design (Origin) will not be the same as the other (Igor Pro), but the operational functionalities are identical when the proper approach is taken in one or the other. In this case, you should ask specifically questions such as "How do I set the peak type in Igor Pro so that I don't have to select it each time"? or "How can I add a new peak in Igor Pro and have it duplicate the same peak settings as some other peak?".

Second, your question seems to be about the general approach to do peak fitting (whether in Origin or Igor Pro). This is suggested by your comment to the extent that "I have to ... *and still do not [get] a best fit*". I might ask at this point ... Do you know what it means to "get a **best** fit"? Do you get a best fit in Origin? What leads you to believe that Igor Pro does not give a **best fit** when you use exactly the same parameters as you use in Origin?

My final observations are from a review of the demo YouTube video. The fitting is on XPS spectra, a topic I know well. In this regard, I first have a hard rebuke to make. When you truly are fitting XPS spectra, please do NOT follow the fundamental methods of the tutorial. For the specific example it shows, the lessons it leaves are grossly incorrect. In XPS of peak fitting of metals and their oxides, one should **never** fit the pairs of doublets with four individual peaks. Each peak doublet (pure metal and oxide) must be fit as linked doublets. This was true even twenty years ago, let alone being true at the time the demo was made (three years ago). Secondly, when you are doing peak fitting of XPS spectra with Igor Pro, you might do well to browse these three packages.

https://www.wavemetrics.com/project/EccentricXPS

https://www.wavemetrics.com/project/XPStools

https://www.wavemetrics.com/project/FitXPS

They might do you better than working directly with the generic MultiPeak Fitting package.

Finally, I follow on the suggestion from John Weeks: Please post an example experiment and explain specifically where you are having problems by reference to that example.

 

In reply to by johnweeks

johnweeks wrote:

At that point, you can click the Fit button to get Igor to do the fit using the collection of guessed peaks as the starting point.

Thank you! I was missing the "Do Fit" button because I never thought of clicking OK to make the tutorial message go away (to unblock the view)! (screenshot attached)

Sorry to have made an ignorant mistake.

I was trying to fit a 2-peak curve (screenshot attached), but the range of Peak 0 it gave looked limited – its Gaussian shape did not extend and go to zero on either side/tail. What does this mean? Can I make it show the entire Gaussian curve for Peak 0? Thank you.

The heuristic used to figure out the width needed to display the peaks sometimes fails. But you can have MPF2 make the widths as wide as your graph: Click the disclosure triangle labelled "Options" to display some additional settings. Turn on the checkbox labelled "Display Peaks Full X Width of Graph"