Quantitative XPS can be tedious. Not because it is an unreliable method but because one needs to understand quite a few things about how the signal is generated befor one can even try to decide how to attempt a quantification.

As so many times, comparisons between similar specimens under identical conditions are much easier to be done in a meaningful way than absolute quantifications. It may therefore help to have reference specimens with known compositions at hand so that unknown compositions can be found by comparison.

Also, one does need some knowledge on the specimens such as e.g. whether you can reasonably expect that the distribution of the elements is homogeneous. Many a times this is not the case. This will complicate the analysis.

Also, so-called sensitivity factors or ionization cross sections have been published (e.g. by Yeh, Yeh & Lindau) and may help greatly, nevertheless, peaks at strongly different photoelectron kinetic energies will have different probabilities to reach the detector (due to the energy dependent electron escape length).

Background analysis (which was mentioned before) can greatly help in this respect. Besides the literature already quoted, the work (and software) of S. Tougaard is worth mentioning. But it is not trivial and far from a I-push-the-button-and-the-computer-does-it-all-for-me kind of procedure. Thourough study of the XPS "problem" is required. It is also a rewarding exercise.

Generally most of the data acquisition software provided with the XPS gives the elemental composition based on the theoretical formulas. You can calculate it manually also but try to get it from the software.

There are software called Multipak you can used it

In a XPS spectrum peak intensities of different elements correspond to their atomic percentage present in a sample. However calculation of atomic percentage is not very straightforward. Without a proper background subtraction calculation of atomic percentage may be misleading. In case of non-monochromatic X-ray source satellite subtraction is also required. The software CasaXPS is versatile for these subtractions. But it is not a free software.

@M Aslam: Perhaps I am mistaken and I mean no disrespect but ...

You seem to lack even an introductory level of understanding about XPS to allow you to read the article you have quoted with any degree of clarity. Have you read any of the numerous books or review journal articles that cover all aspects of XPS starting from a very simple level? For such books, you should look for any of these authors: J. F. Watts, P van der Heide, T. Barr, D. Briggs, or J Wagner. For a broader scope of many measurement systems, you could look for the encyclopedia by E. Kaufmann on Materials Characterization. It has a chapter on XPS. Finally, when all else fails, your XPS instrument should have a handbook with introductory information.

You also seem to lack a basic level of understanding about how analytical methods are applied and how electronic instrumentation systems work. This information is typically taught in undergraduate analytical chemistry and physics instrumentation courses. I recommend as one step that you could find a good undergraduate level textbook on analytical chemistry and read the chapters on how to obtain quantitative composition from spectra. You could also search for such simple phrases as "how instruments measure composition" to begin your studies.

Quantitative XPS can be tedious. Not because it is an unreliable method but because one needs to understand quite a few things about how the signal is generated befor one can even try to decide how to attempt a quantification.

As so many times, comparisons between similar specimens under identical conditions are much easier to be done in a meaningful way than absolute quantifications. It may therefore help to have reference specimens with known compositions at hand so that unknown compositions can be found by comparison.

Also, one does need some knowledge on the specimens such as e.g. whether you can reasonably expect that the distribution of the elements is homogeneous. Many a times this is not the case. This will complicate the analysis.

Also, so-called sensitivity factors or ionization cross sections have been published (e.g. by Yeh, Yeh & Lindau) and may help greatly, nevertheless, peaks at strongly different photoelectron kinetic energies will have different probabilities to reach the detector (due to the energy dependent electron escape length).

Background analysis (which was mentioned before) can greatly help in this respect. Besides the literature already quoted, the work (and software) of S. Tougaard is worth mentioning. But it is not trivial and far from a I-push-the-button-and-the-computer-does-it-all-for-me kind of procedure. Thourough study of the XPS "problem" is required. It is also a rewarding exercise.

Dear Madam,

in addition to KF´s previous statement, there is still another source of "error" that

cannot be taken from literature values since it is related to the specific XPS setup.

Kai mentioned the energy dependence of the so-called electron mean

free path. When you compare photoemission intensities from orbitals with

large differences in binding energy the corresponding photoelectrons also differ strongly in kinetic energy.

In the typical range of kinetic energies in XPS (i.e. some 100´s

of eV) the electron mean free path that increases with kinetic energy

causes the faster electrons to probe a larger (i.e. deeper) volume of the bulk, i.e.,

in terms of stoichiometry, these faster electrons are oversetimated.

In addition to the energy dependence of the electron mean free path there

is also a energy dependence of the transmission of the specific spectrometer.

It depends on electron optics, type of analyzer, pass energy, combination of

slits, etc., i.e. it is very complicated.

This energy dependence of the transmission can partially cancel the energy dependence of the electron mean free path. For quantitative analysis, both

factors have to taken into account to get reliable results. Otherwise, try to

do the analysis for orbitals that are as close as possible, since for

Ekin1/Ekin2 ~ 1, it does not matter with which exponent A the kinetic energy

part of the intensity ratio, i.e. ~ (Ekin1/Ekin2)^A, is weighted.

Sincerely, FM

@K Roy: You absolutely must confirm whether any third-party software uses the sensitivity values for your XPS instrument to calculate elemental compositions. Otherwise you introduce calibration offsets (Type B uncertainties) in your results.

@Kai Fauth and FM Sir......thanks for the important information regarding the XPS data analysis. Thanks a lot.

Kai Fauth and J. Weimer adressed the main topics related to quantitative XPS. Unfortunately you did not give clear information on the sample morphology, as this is the most important issue. As overlayers on substrates are to be treated different as changes in composition in a deposited material. As a rule of thumb: an accuracy below 5% needs really a sophisticated experimental work and very careful data evaluation. You should examine the literature given by J. Weimer and take some weeks of practice in a group of experts.

@Kanak Roy: I have never used CASA. In fact, where necessary we have usually implemented the more advanced evaluation steps ourselves, e.g. as plug-ins to other pieces of software. But this was not necessary very often for the work I have participated in. So I cannot be of much help on this point.

This thread has evolved in to a discussion about what contributes to uncertainty in composition values with XPS analysis. The theme of uncertainty budget is an important one in experimental data analysis, and I would like to add a wider and deeper perspective in this regard.

Any numerical result must be reported with the value, its units, and a confidence level. The confidence level is a measure of the contributions of Type A and Type B uncertainties (see NIST and others for detailed discussions of these). Here are some contributing factors to uncertainties for any chemical spectroscopy and for XPS specifically.

Type A

* How many samples did you analyze?

* How many regions on a given sample did you analyze?

* What is the S/N in your data?

* What is the measurement precision of your analytical instrument?

* What is the measurement precision of your data analysis method?

Type B

* What is the calibration uncertainty for your instrument in the specific configuration that you used?

- When was your XPS system last calibrated (for BOTH measurement axes - energy and intensity)?

- To what extent did you compare your results from your samples against those from calibration standards?

- Did you use the "standard" geometry for analysis or did you change something, for example the take-off angle, that requires a different set of calibration parameters?

* What is the accuracy of your data analysis method?

- Does your baseline fit the Shirley, Tougaard, or other model?

- Are you using the proper peak parameters (shape, height, fwhm, and position)?

- Have you under-specified or over-constrained any of your fit parameters?

- Does the analysis method use the right sensitivity factors for the measurement configuration in your instrument?

- Is your sample truly homogeneous in composition and/or have you properly accounted for it not being this way?

Given the ongoing publicity around the theme of uncertainty budget, I am still amazed by professionals who publish analytical results without even a passing consideration to address the above questions. I am even more amazed when such results are propagated as gospel by researchers in the field. Perhaps this summary will motivate those who are new to the topic to recognize, the need to report the confidence level of your numerical results to the best possible degree is to be taken more seriously than might otherwise be implied.

@Kanak Roy:

So far, I didn´t use CASA. As I mentioned in my first answer there are some spectrometer specific features, as, e.g., the transmission of the analyzer, that

have some impact on quantitative analysis. Even in case that CASA (or any other software) provides libraries containing such spectrometer specific corrections

(for all configurations possible) it still remains a black box that does not know the physics of your sample.

An example:

As I mentioned before the intensity of a peak at a particular binding energy/kinetic

energy is affected by the energy dependence of the electron mean free path as well as by the energy dependence of the transmission.

In case that you probe the bulk of your sample both aspects have to be taken in to

account when comparing elemental composition.

In case that you probe a monolayer on a substrate (just think of epitaxial graphene on a transition metal surface) the intensity of the C1s electrons is of course affected by the energy dependence of the transmission. However, the C1s intensity IS NOT affected by the energy dependence of the electron mean free path since the C1s electrons are emiited only from the topmost layer. So, they do not have to travel trough a solid and no attenuation of intensity takes place.

In contrast, the intensity of, e.g., the TM-3d electrons from the substrate is affected by the transmissionthe as well as by electron mean free path since some of these electrons have to pass the solid before crossing the surface.

As you can see, the way of analysis strongly depends on the sample you

look at. I think software should just support your analysis but it cannot not do

the whole job of a solid state physicist dealing with surface analysis.

Sorry, to be more precisely: it must read as "software MUST NOT do the

whole job of a solid state physicist" since literature is full of examples when

people used such a black box software without any knowledge even of the XPS basics.

Very often, you can read "spectra have been deconvoluted to obtain the best fit". In these cases, people obviously give priority to numerical maths but they completely ignore the physics behind the photoemission process, such as the proper (angular momentum+1):(angular momentum) intensity ratio within a doublet. Of course, the whole data interpretation that is based on such "best fits" cannot not be beyond any doubt.

Sincerely, FM

It is quite interesting reading all the comments on this topic, so far. I should underline some point. The original question posted by Marheem Aslam calls for a deeper analysis of how anybody should approach to an analytical technique, above all when the same is not trivial, and XPS is not trivial at all. Examination and physico-chemical information extraction from spectra MUST derive from a competent work, where the same experiment (and the whole experimental setup) has been designed knowing where to go. Using a software can be an help, not a solution. In my experience, all the XPS work I have performed was based on:

1. proper choice of sample (e.g. preparation, cleaning)

2. proper choice of anti-SCR arrangements when needed

3. adaptation of spectrometer params to maximize the chance of getting reliable results for the case of interest (slits opening, pass energy, etc)

4. proper calibration (e.g. work function alignement)

5. repetaibility verification and statistical meaning of the data collected

... and much more.

XPS is for XPS experts, in my opinion, even because of all those strange and apparently interfering signals one can face that could easily lead to misinterpretetion (shake-up, replicas, Auger lines, etc.)

Frequently an experienced spectroscopist does not need a specific software... an ORIGIN shett or even EXCEL may be enaugh. But experience means, above all, recognizing the origin of a peak and the details of its lineshape before assigning it a label.

I hope that Marheen will forgive me if I suggest her, and anyone who is approaching XPS (but also any kind of spectroscopy), to consider the same as a matter of deeper study and not only a mere analytical technique, or he/she may ask specialized peoples or Labs to collaborate and help.

You can use CasaXPS to calculate atomic percent and many different parameters like FWHM, deconvolution for the peak.

Download and Guideline for CasaXPS: http://www.casaxps.com/

Follow this video to calculate atomic percent: http://bit.ly/2tui78b