Well, there is quite some division in the community, but:

1) Baseline: I think Shirley is the most accepted background since it has the smallest "arbitrariness" going along with it. Tougaard is more versatile but it can be tuned into weirdness which is why it is rarer used.

2) Peak width: that is an apparative constant that can be determined experimentally. For the O1s peak width in this work:

Article Formation of Sublimable Nanographene Oxides by Reacting Coro...

the fitting coauthor used the width of O1s on atomically oxidized HOPG for which it is known from the cited literature that there is only one component. When continuing, the most important for a good fitting scheme is that you act self-consistently, arbitrary changes from spectrum to spectrum within the same paper are the one thing noone should do.

3) Line shape: yes, especially with sp2-C1s that is not exactly a fun topic. Some years ago a significant fraction of the community had agreed on Doniach-Sunjic but it seems to me this "semi-agreement" has faded away again.

I agree with Jürgen Weippert on all points.

Considering the line-shape choice, I would avoid it because of troubles in integration, that is in peak area quantification. DS lineshape integral from minus infinity to plus infinity is "infinite", and this would cause the integration window choice (that is used to adjust parameters during the fit and evaluate the at% at the end) too much arbitrary and hard to handle.

I suggest using a lorentzian lineshape or a Voigt lineshape, the latter being a convolution of the lorentzian which intrinsecally represent the photoelectron energy distribution, and a gaussian which accounts for the deltaE contribution due to the instrument.

If an asymmetry of the peak is forecasted or even visually evident in the spectrum, I would add a second lorentzian peak just by the primary one, with proper limits to position and FWHM, and carry on as many fitting trials as needed to obtain a result that makes sense... and this would call for experience.

Regarding the XPS fitting, CasaXPS is a really good software, which could be used. For XPS and other types of photoelectron spectroscopy, Shirley background is the most widely accepted function to make baseline. Many people avoid such background and make baselines in origin, which is not appreciated.

The lineshapes are best represented by pseudo-Voigt function, which is a convolution of gaussian and lorentzian function, and is in general symmetric. Generally in Rietveld's method for XRD and XPS-fittings mostly peaks are symmetric; and this function is advised. Although in special cases, BWF function could also be used to quantify asymmetry.

Finally, the peaks should be deconvoluted with a number of small peaks, which are justifiable and the goodness of fit should be less than 1. Then, one may claim that, the fitting has been fine enough. The process must be mentioned in the methods-section of the article.

Thank you all for your valuable feedback. It helped me picture better the situation and how peak fittings are evaluated/reviewed by others.

However, my concern still remains in my case, because in peer reviewed papers the background and type of line-shape/s used are not always mentioned, or explained into much detail. In addition, for the same material (say material X has a C 1s peak that has to be fitted), different authors of peer reviewed papers use different backgrounds and line-shapes despite the fact that the C 1s peak of material X is 'the same' in terms of shape!

Alas, since the community isn't unified on this topic, this will remain an issue and you will have to figure out what to do and what you consider serious yourself.

Yes, that is what I was thinking. Thanks a lot for your feedback! Much appreciated!

Can anyone remind me the physical meaning of Shirley background. In my knowledge, Shirley background represents secondary electron. Due to that reason, it is just valid for metal and should be avoided for organic elements XPS characterisation.

You have secondary electrons for every compound, not just for metals.

However, it is correct that the physical reasoning behind the Shirley baseline is actually rather vague, you can see for yourself in the original work:

Article High-Resolution X-Ray Photoemission Spectrum of the Valence ...

Honestly, all baselines in my opinion have a rather "handwaving" reason behind them but Shirley is probably best reproduced if different operators evaluate the same spectrum which makes it the good option for reproducible data that it is.

Globally, we can conclude this issue with a comment from the CasaXPS manual in the Tougaard section: " While conceptually appealing, the universal cross-section was designed as a practical background for general use and as such is almost certainly as incorrect as the Shirley and linear background types " (Source: http://www.casaxps.com/help_manual/manual_updates/peak_fitting_in_xps.pdf)

Thank you everyone! I think this discussion has explained my maximum inquiries, i have just one question regarding goodness of fit parameters. I am getting it like 5. is it acceptable?

Maria Magro Deconvolution of a composite peak into its individual peaks plays an important role in the interpretation of many types of graphs including XRD, XPS, FTIR, and PL etc. In this video, I have discussed how to deconvolute simple combined peaks, composite peaks and how to correct missing data in a given peak with the help of deconvolution. In the case you want to further ask about it, please do comment on the specific video, I'll respond to it shortly. I have provided the practice filee (OriginLab) here. Thanks

https://youtu.be/WX_X3XlbKRA

Thank you so much for the insightful contributions. As a beginner in XPS analysis, the discussion has helped me a lot.

However I have one question on positioning the peaks using C1s peak. Assuming you have two samples; un-doped and a doped one. While I now the doping of the sample might have an effect on the BE of the elements in the compound, and I think to get the correct BE of the elements, C1s is used to give the position of the peaks. So, my question is, while positioning the peaks for the two samples, should I use the C1s for the un-doped sample to position the two samples, or should I use the C1s for each sample separately? Thank you.

You would need to use the C1s of each sample individually. However, you should keep in mind that the C1s alignment method is not very reliable and should only be used to roughly adjust the binding energy scale e.g. in the case of insulating samples. To compare your two samples directly with one another it is not really suitable.

The details of why it's not very accurate and should be used with caution can be found here:

https://www.researchgate.net/project/Why-you-should-not-use-the-C-1s-peak-for-referencing-XPS-spectra

Well noted Thorsten Schultz , thank you so much for your guiding remarks and the reference. They are quite helpful. Thank you once again.