Hi everybody,

Thanks for all contributions,

I am back here just to make a sort of conclusion out of all the discussions that might be useful for all. Through a series of experiments (with XPS of Pt and carbon mainly) from which the original question came out and getting engaged with the interpretation of the results, and also using the comments I received here and keeping an eye on the literature, I could refine a sort of “procedure” according to which a reasonable processing of the data and interpretation becomes to a good extent trustworthy. Although all of the points mentioned in the following are not really constrains, but including them in the “procedure” is critical and helps a lot in a clean elaboration of the data. Obviously, this procedure is just a simplification of the encounter with XPS data and there are many delicate and narrow points to be considered in any individual case. The other thing is that for those using CasaXPS, all these things described here may sound more familiar, simply because I have no experience with other software and some terms used here might be limited to this software.

-------------

i. Survey scan (the procedure below describes the condition where no overlapping of different transitions (and therefore, no need to their deconvolution in narrow scans) that are relied on for survey quantification exists. If such overlapping exists, first, the elaboration of the narrow scans of the related transitions should be carried out as follows in the “ii. Narrow scans” part, and after the step 12 of the narrow scan procedure, we should return here!):

1- Calibration of the entire spectra based on C 1s at 284.8 (this step is better to be adjusted based on the narrow scan of C 1s).

2- Generalizing the calibration of the survey to all narrow scans.

3- Peaks identification.

4- Background selection (mainly but not exclusively shirley).

5- Selecting an Av Width (usually 1 or 2) for the shirley background borders to stabilize the background shape with respect to noise in the data.

6- Elemental composition from survey spectra.

7- Adding standard deviation to the calculated concentrations.

******************************

ii. Narrow scans:

1- Defining the region borders.

2- Background selection (mainly but not exclusively shirley).

3- Selecting an Av Width (usually 1 or 2) for the shirley background borders to stabilize the background shape with respect to noise in the data.

4- Selecting the proper line shape based on the symmetry and geometry of the peak. The most commonly used synthetic line shapes are

• Product Gaussian-Lorentzian GL(m),
• Sum Gaussian-Lorentzian SGL(m),

where m=0 is a pure Gaussian and m=100 is a pure Lorentzian shape (e.g., GL(30) that one of the mostly used ones is 70% Gaussian & 30% Lorentzian),

• The line-shapes LA(α, β, m) and LF (α, β, w, m) offer asymmetric line-shapes based on the Lorentzian functional form convoluted with a Gaussian. This type of line-shapes are absolutely important for doublet transitions that are intrinsically asymmetric.

5- Creating components.

6- Adjusting the synthetic model with putting constrains on “peaks position”, “peaks area” and “FWHM”.

7- For doublets, a physically meaningful solution involves ensuring the ratio of the peak areas based on the spin-orbit coupling (as in the following) that should be dictated to the components as a constrain:

• Doublet peaks resulting from the emission of a p-orbital appear with intensities in the ratio 2(1/2)+1:2(3/2)+1 or 1:2. (example: Cr 2p)
• Doublet peaks resulting from the emission of a d-orbital appear with intensities in the ratio 2(3/2)+1:2(5/2)+1 or 2:3. (example: Ce 3d)
• Doublet peaks resulting from the emission of a f-orbital appear with intensities in the ratio 2(5/2)+1:2(7/2)+1 or 3:4. (example: Pt 4f)

8- For different chemical states of an element in a certain transition, there are exact binding energies in data base. However, application of such exact numbers as constrains on peak positions is not recommended. Rather, it is useful to define an interval of 1-2 eV, the center of which is based on the binding energies found in the data base and leave the mathematical model free to decide where in this interval can be the best for the least square based calculations to minimize the model error.

9- Putting constrains on FWHM is sort of “dangerous” as long as a deep knowledge of the physics of the system is not available. So, I do not give any comment on it. Just one thing, very large FWHM are not desirable. So, ending up with FWHMs more than 5 eV, could recommend a revision of the entire data processing.

10- Fitting components.

11- Using the offset in energy between the doublet peaks (known from a database like NIST) as a checkpoint for controlling the success of the model. If the energy separation of the peaks (pair of the doublet at zero chemical state and not those of other chemical states) was not correct and could not be modified, this offset can also be enforced to the model as a constrain.

12- Transferring the results of peak model of the narrow scan to the survey spectrum in order to complete the quantification of the entire spectra.

Hi Manzar, Thanks for your answer.

The point you mentioned could be a good checkpoint for controlling the appropriateness of a fitting. But my question refers to one step before this, said that, how to select those constrains (if really needed) to have a better and more meaningful fit.

Two possible constrains are the Lorentzian to Gauss ratio and the FWHM of your peaks. To know meaningful values here, it is good to know your XPS system and have some experience. Usually, if I use the same measurement parameters, the L:G ratio stays more or less the same, I could fix them in the fitting. The FWHM is more difficult, since peaks broaden with increasing binding energy. So again if you know your experimental setup you can make an educated guess. Furthermore, all the peaks you fit, e.g. in the Pt 4f, should have the same values of FWHM and L:G. At least that is a good start, charged specimen could show a slight broadening.

Regarding constraints on position: if you find a publication that reports these values (Pt, Pt+, Pt2+,...) you could use the difference between them as restrains. However, comparing values from different publications is often not helpful, since the calibration of a lot of XPS setups are slightly different, leading to different values. 

These few articles may help answer your question:

https://www.researchgate.net/publication/263965482_The_Gaussian-Lorentzian_Sum_Product_and_Convolution_%28Voigt%29_Functions_Used_in_Peak_Fitting_XPS_Narrow_Scans_and_an_Introduction_to_the_Impulse_Function_

https://www.researchgate.net/publication/263162061_An_Introduction_to_Convolution_with_a_Few_Comments_Beforehand_on_XPS

https://www.researchgate.net/publication/263891984_The_Equivalent_Width_as_a_Figure_of_Merit_for_XPS_Narrow_Scans?ev=prf_pub

Article The Gaussian-Lorentzian Sum, Product, and Convolution (Voigt...

Article An Introduction to Convolution with a Few Comments Beforehand on XPS

Article The Equivalent Width as a Figure of Merit for XPS Narrow Scans

Hi Selina,

Thanks for your helpful comment.

About the L:G ratio, I usually use GL(30) according to some recommendations from the XPS technician. 100% Gaussian is another option based on the literature that I have not used yet.

About the FWHM, I usually prefer to stick to the ratio of the peaks if I know anything about it (e.g., Pt 4f 7/2 peak should have an area about 4/3 larger than the Pt 4f 5/2 peak as I mentioned in the question) and let the software to adjust the FWHM (putting the constrains on FWHM could be more difficult as far as I know, as in this case, I just know that a typical pair of 4f peaks results in a greater FWHM for the 4f 5/2 peak compared to the 4f 7/2 peak, but nothing more specific. I should remind that I’m quite unexperienced with XPS in the meanwhile). Usually, it comes out to be reasonable this way (FWHM for the 4f 5/2peak came out to be larger than that of 4f 7/2 peak for my Pt doublet after putting the constrain on the peaks area).

The biggest trouble in fact, is that of the positions as you also implied. There are reports in the literature, but an agreement on the positions can hardly be found if any.

Hi David,

Thanks for the papers. I'll try to check them and may return to you later.

Hi Justin,

Thanks for your comment. Through my searches for a database, I had come across the NIST XPS database. One can find it here: http://srdata.nist.gov/xps/

The thing is that when one does the deconvolution and finds let’s say three components under a single peak, apart from one that is referring to neutral state of that element, the other two are of two other chemical states of that element. For an element with a couple of such chemical states (almost all of elements are more or less this way), it wouldn’t be easy to say which ones are those two, even with access to a database since the position of the peaks in XPS is not absolutely fixed, but changes slightly experiment to experiment and instrument to instrument. But anyways, I think you are right and as far as a database is concerned, NIST is a good one. Sad but true (for a novice), it seems that in this story, practice only and experience, provides the confidence.

About the energy reference and calibration I use C 1s adjusted at 284.8 eV and in Casa as you told it is done easily.

Thanks again for your help.

Dear Mazdak:

As a first and rough approximation to obtain the parameters  for curve fitting of XPS spectra, we use the he evolution with sputtering time of the high resolution XPS spectrum obtained on our specimen.

In general, the spectra of the non-sputtered surface is dominated by the presence of the element in the form of higher oxidised state, increasing the contribution  from reduced or metallic components to the XPS spectra with the sputtering time.

I  hope it helps.

Kind Regards.

S.

Dear Sebastián,

Thanks for your comment. If I understood well, you meant using the so-called etching technique. A sputter depth profile involves removing material with an ion-gun interleaved with acquisition cycles where spectra are recorded from each new surface uncovered by the etching process. This can help to remove in a step by step way the topmost layers of the material and acquire the spectra between the etching steps. I think it can definitely be helpful in the sense that for subjects of study such as Pt catalyst that we expect a degree of oxidation on the surface, this technique shows us the point where that oxide layer is turning to a minimal level (ideally, zero) and we can assume the spectra acquired from that point onward as a representative of metallic Pt. Then, having this spectrum as a “standard” for metallic Pt, it will be much easier to say how different were the primary spectra from this standard and how much impurities including oxidized states of the matter were present there.

I think notwithstanding the usefulness of this technique, it cannot be considered as a constraint in the usual sense in the XPS field. Said that, those constraints are usually referred to curve fitting criteria considered in building a synthetic line or model to fit the acquired spectra. Components of this synthetic model will be representative of different transitions and chemical states. Therefore, those constraints include limitations enforced on the position, area and FWHM of the components (obviously, resting on some physically meaningful grounds) to address both a good mathematical match of the synthetic model and the experimental data and the physical meaning of the resultant components. Having a clear idea and good hand on these constraints helps the interpretation of the results a lot, especially when there are limitations on access to auxiliary tools such as ion gun for depth profiling.

Regards,

Mazdak

Dear Mazdak:

Thank you for your answer.  You are right, as you said"...it cannot be considered as a constraint in the usual sense in the XPS field...".

However and from a practical point of view is an useful approximation because the usual constraints ( position, area and FWHM of the components) may depend on the characteristics of the specific spectrometer used on the XPS analysis and on the characteristics of the specimens studied (for example charging or topographic properties) .     

Kind Regards.

S.

Hi everybody,

Thanks for all contributions,

I am back here just to make a sort of conclusion out of all the discussions that might be useful for all. Through a series of experiments (with XPS of Pt and carbon mainly) from which the original question came out and getting engaged with the interpretation of the results, and also using the comments I received here and keeping an eye on the literature, I could refine a sort of “procedure” according to which a reasonable processing of the data and interpretation becomes to a good extent trustworthy. Although all of the points mentioned in the following are not really constrains, but including them in the “procedure” is critical and helps a lot in a clean elaboration of the data. Obviously, this procedure is just a simplification of the encounter with XPS data and there are many delicate and narrow points to be considered in any individual case. The other thing is that for those using CasaXPS, all these things described here may sound more familiar, simply because I have no experience with other software and some terms used here might be limited to this software.

-------------

i. Survey scan (the procedure below describes the condition where no overlapping of different transitions (and therefore, no need to their deconvolution in narrow scans) that are relied on for survey quantification exists. If such overlapping exists, first, the elaboration of the narrow scans of the related transitions should be carried out as follows in the “ii. Narrow scans” part, and after the step 12 of the narrow scan procedure, we should return here!):

1- Calibration of the entire spectra based on C 1s at 284.8 (this step is better to be adjusted based on the narrow scan of C 1s).

2- Generalizing the calibration of the survey to all narrow scans.

3- Peaks identification.

4- Background selection (mainly but not exclusively shirley).

5- Selecting an Av Width (usually 1 or 2) for the shirley background borders to stabilize the background shape with respect to noise in the data.

6- Elemental composition from survey spectra.

7- Adding standard deviation to the calculated concentrations.

******************************

ii. Narrow scans:

1- Defining the region borders.

2- Background selection (mainly but not exclusively shirley).

3- Selecting an Av Width (usually 1 or 2) for the shirley background borders to stabilize the background shape with respect to noise in the data.

4- Selecting the proper line shape based on the symmetry and geometry of the peak. The most commonly used synthetic line shapes are

• Product Gaussian-Lorentzian GL(m),
• Sum Gaussian-Lorentzian SGL(m),

where m=0 is a pure Gaussian and m=100 is a pure Lorentzian shape (e.g., GL(30) that one of the mostly used ones is 70% Gaussian & 30% Lorentzian),

• The line-shapes LA(α, β, m) and LF (α, β, w, m) offer asymmetric line-shapes based on the Lorentzian functional form convoluted with a Gaussian. This type of line-shapes are absolutely important for doublet transitions that are intrinsically asymmetric.

5- Creating components.

6- Adjusting the synthetic model with putting constrains on “peaks position”, “peaks area” and “FWHM”.

7- For doublets, a physically meaningful solution involves ensuring the ratio of the peak areas based on the spin-orbit coupling (as in the following) that should be dictated to the components as a constrain:

• Doublet peaks resulting from the emission of a p-orbital appear with intensities in the ratio 2(1/2)+1:2(3/2)+1 or 1:2. (example: Cr 2p)
• Doublet peaks resulting from the emission of a d-orbital appear with intensities in the ratio 2(3/2)+1:2(5/2)+1 or 2:3. (example: Ce 3d)
• Doublet peaks resulting from the emission of a f-orbital appear with intensities in the ratio 2(5/2)+1:2(7/2)+1 or 3:4. (example: Pt 4f)

8- For different chemical states of an element in a certain transition, there are exact binding energies in data base. However, application of such exact numbers as constrains on peak positions is not recommended. Rather, it is useful to define an interval of 1-2 eV, the center of which is based on the binding energies found in the data base and leave the mathematical model free to decide where in this interval can be the best for the least square based calculations to minimize the model error.

9- Putting constrains on FWHM is sort of “dangerous” as long as a deep knowledge of the physics of the system is not available. So, I do not give any comment on it. Just one thing, very large FWHM are not desirable. So, ending up with FWHMs more than 5 eV, could recommend a revision of the entire data processing.

10- Fitting components.

11- Using the offset in energy between the doublet peaks (known from a database like NIST) as a checkpoint for controlling the success of the model. If the energy separation of the peaks (pair of the doublet at zero chemical state and not those of other chemical states) was not correct and could not be modified, this offset can also be enforced to the model as a constrain.

12- Transferring the results of peak model of the narrow scan to the survey spectrum in order to complete the quantification of the entire spectra.