Dear Sir,

in principle your idea sounds but I have some doubts whether this ansatz provides reliable results since your method is based on an unambiguous  relation between C-1s binding energy and the carbon-oxygen configuration of a specific carbon oxgen species. Unfortunately, there are carbon oxygen species with different C: O ratio but nearly the same chemical shift of the C-1s binding energy, i.e. differences in chemical shifts are smaller than the energy resoluion of a usual XPS set-up. See, e,g, Appendix 1 in  

High Resolution XPS of Organic Polymers - The Scienta ESCA300 Database by G. Beamson and D. Briggs

For cross-checking the reliability of your method let me suggest the following complementary procedure, at least as a first order approximation (i.e. when neglecting that the C and O amount is also affected by adsorbates)

Calculate the overall C : O : Si ratio in at-%, i.e.

n[C] : n[O] : n[Si].

This is equal to

n[C] : ( n[OC]+n[OSi] ) : n[Si].

Since n[Osi] = 2n[Si] for SiO2, you get 

n[OC] = n[O]-2n[Si] and finally

n[OC] : n[C] = ( n[O]-2n[Si] ) : n[C]

Please note: Since the kinetic energies of the O-1s, Si-2p and C-1s electrons are strongly different, calculating the atomic ratios just via the peak intensities (i.e. peak areas) and the photoemission cross sections may give only a rough estimate. For refinement, please take also into account that the electron mean free path as well as the transmission of your setup depend on kinetic energy.

Regards, FM

My personal experience on various coated polymer samples tells me that the caclulated atomic compositions (even after taking into account specific reponses) can be used rather for comparative purposes than for establishing exact chemical structures. In several cases I could not match the expected structure with the observed compositions.

The equation you provide assumes that each C-oxygen species contributes to the O content according to its stoichiometry. The equation provided by S. Müller takes the approach that each Si-oxygen species contributes to the O content according to its stoichiometry. In theory, the answers should be identical. In practice, the two equations will differ and could be used to give a bounds or first estimate.

You might also consider doing analysis at two different take-off angles, normal and glancing. With a film, the O(C)/C should approach the true O(C)/C as you go to glancing angle because the O(Si) will decrease in contribution.

Dear Professors,

Thank you for your answering the question.

As I have replied Prof. Dr. Frank Müller, I have read some papers describing the complicated oxygen-containing groups corresponding to different binding energy in the XPS spectra.

[Ref. Yamada, Yasuhiro et al., Analysis of heat-treated graphite oxide by X-ray photoelectron spectroscopy, Journal of Materials Science, 2013(48), 8171-8198]

I understand that the unambiguous correlation between the binding energy and the functional groups will lead the calculation from peak fitting results not very precise. However, simply fit the C1s spectra of graphene oxide to C=C, C-C, C-OH, epoxide, C=O and C(=O)O, has been used in a lot of papers in this field. Therefore, I think stepping forward to use the calculation from peak fitting results to semi-quantitively / roughly measure the oxygen/carbon ratio is reasonable.

[Ref. Pei, Songfeng st al. The reduction of graphene oxide, Carbon, 2012(50), 3210-3228]

I have tried Dr. Frank Müller's advice to use the oxygen, carbon, silicon atomic percentage to calculate the oxygen/carbon ratio. The result showed correlation with the result from the peak fitting method  but not precisely matched as attached. The problem is that, the x value for SiOx is not equal to 2. This can be solved by measuring the SiOx substrate. However, another difficulty is that, during the experiment, I found the x value may have changed due to the UV light irradiation. This has also been observed during our other researches.

Based on these reasons, I feel that the equation I described in the question could be a better choice for quantifying the O/C ratio in a graphene oxide covered Silicon substrate.

Would you agree with me? If you have further advices, please contact with me. Thank you very much again for your kindly replying.

All my best wishes,

Yudi

As you have done with the oxygenated C species in the first equation, can you extract the SiOx and SiO2 peaks by peak fitting? This could improve the agreement with the first and second equation ...

RO/C = (n[O] - 2n[SiO2] - n[SiOx]) / n[C]

The SiOx peak is offset by a definitive amount from the SiO2 and shows as a shoulder on what otherwise is a Gaussian peak. I suggest, with this change, that results from both methods should be presented because they provide upper/lower bounds on the value. The equations are derived based on assumptions; they are not derived from first principles. You would need to peak fit (not deconvolve) the O 1s to its O-C, O-Si, and O-other peaks to get closer to a first principles answer.

@Y.T.:

Of course, both methods have some drawbacks. Therefore, refering to both methods gives you an idea about typical error bars. Within this context, i.e. when considering typical XPS-related error bars, the result you have provided is not too bad!

In addition, it may be useful to follow the suggestion by J.J.W., namely to probe the stoichiometry of the samples for the most surface sensitive case, i.e. for large polar angle to eliminate most contributions from the silicon oxide. However, please ensure that the analyzer is focused on the sample exclusively. For large polar angles there might be the danger that also a part of the sample holder (with different C:O ratio) is probed.

@ J.J.W.

I think it has to read

RO/C = (n[O] - 2n[SiO2] - xn[SiOx]) / n[C] ???

Regards, FM