A candidate which would come to my mind would be the baseline subtraction. Especially when graphitic carbon is involved, the peak asymmetry may lead to incorrect automatic substraction so that you may accidentally cut away a large portion of the peak.

Just to be sure: all single peaks were recorded with the same pass energy and other detector settings?

Jürgen Weippert Thank you for your comment.

From the survey spectra I quantified peaks of C1s, O1s, N1s, and Si2p. The high resolution spectra of these peaks were taken at the same pass energy, energy step and source.

I am using a Shirley background in each high resolution peak. I applied the component peaks in the high solution spectrum of C1s, the other elements I just applied the Shirley background. I didn't add any components.

The discrepancy in carbon atomic percentage (at%) between the survey spectrum and the C1s high-resolution peak fit suggests that there could be an issue with either the fitting parameters or the quantification approach. Several factors may contribute to this inconsistency. First, the fitting parameters for each C1s component, such as binding energies, peak shapes, and constraints, must accurately reflect the chemical environment in polyurethane; poor resolution of peaks for functional groups like C–C, C–O, or C=O can lead to inaccurate C1s quantification. Background subtraction choice is also crucial, as different methods like Shirley or Tougaard impact peak areas, so consistency across both survey and high-resolution scans is essential. Additionally, CasaXPS uses relative sensitivity factors (RSFs) for element quantification; any RSF discrepancies between survey and high-resolution scans could affect the C1s concentration. Overlap between elements or matrix effects can further influence the C1s at% in the high-resolution fit, especially with high nitrogen or oxygen concentrations. To resolve this, refine CasaXPS peak parameters, confirm consistent RSF use, and verify background subtraction methods. If the discrepancy remains, cross-verifying with other quantification software may provide additional insight.

In my opinion, i had also find CasaXPS not so user friendly, you better try in origin, it is slightly hectic in origin, but you will get better results in origin

You could also try "not fitting" the peaks, just integrating the whole area of it.

Jürgen Weippert Which ones please?

I carried out peak fitting for the C1s. For N1s, O1s, and Si2p I only added a Shirley background without any fitting/component peaks.

OK, so you fitted the C1s and set it in ratio to the others which were simply integrated? I would also simply integrate the C1s and compare the results. Be careful not to cut the asymmetry tail, if it's there. C1s fitting including graphitic carbon is a quite controversial topic in the XPS community.

Hello, Maria Magro!

In your case, the only parameter that changes when moving from the overview spectrum to the high-resolution spectrum is the duration of the measurement. Maybe during the long continuous time of recording the overview spectrum, something happens to the channeltron or amplifier. For example, maybe they heat up and the signal changes. As for me, I constantly observe how the XPS signal slightly decreases over time during long-term recording. In contrast, high-resolution spectra are recorded separately.