EDX basically gives you elemental information. The maximum energy of the X-ray is obviously limited by the maximum energetic electrons (20-40Kev normally). Now, for quantitative analysis, you really need to collect all the possible X-rays emitted from an element from K series upto absorption edge. This is not possible and even major K-alpha line can not be seen for high Z value elements. Oxygen however will give always K-alpha line for having low Z value.

The point is from EDX you can find out that there is Fe, O in the sample but accurately ca not determine whether this is FeO, Fe2O3, or Fe3O4. Even if you calculate it from the normalized atomic concentrations it will be erroneous.

But if you keep the EDX experimental condition same that is hit different samples with same energetic beam and keep the working distance same, then for different samples you can qualitatively have an idea about different elements from the intensities of their peaks.

Even if the used setup could detect the oxygen then you'd not be able to perceive the type of oxide phase. Each element has a unique atomic structure allowing unique set of peaks on its X-ray spectrum. The incident beam (electron beam or X-ray) will provoke the atom and give rise X-ray. The process relies on the uniqueness of characteristic X-rays of each element. By the intensity and the pertinent FWHM you can find the compositional percentage. To study the type of oxides I'd strongly recommend XPS.

EDX basically gives you elemental information. The maximum energy of the X-ray is obviously limited by the maximum energetic electrons (20-40Kev normally). Now, for quantitative analysis, you really need to collect all the possible X-rays emitted from an element from K series upto absorption edge. This is not possible and even major K-alpha line can not be seen for high Z value elements. Oxygen however will give always K-alpha line for having low Z value.

The point is from EDX you can find out that there is Fe, O in the sample but accurately ca not determine whether this is FeO, Fe2O3, or Fe3O4. Even if you calculate it from the normalized atomic concentrations it will be erroneous.

But if you keep the EDX experimental condition same that is hit different samples with same energetic beam and keep the working distance same, then for different samples you can qualitatively have an idea about different elements from the intensities of their peaks.

Both answers above are correct i.e.: EDX cannot provide the oxydation state of an element for lack of energy resolution of the X-ray generated by the incident e-beam. Furthermore, the original question referred to the "elemental analysis of a surface". As a matter of fact EDX is NOT a surface analysis method in the sense that it provides the elemental analysis of the interaction volume in the sample. The size of such volume depends on the enrgy of the primary beam of electrons and on the energy of the element being investigated (low atomic number elements emit low energy X-rays, hence are absorbed more within the interaction volume, etc. etc.).

Perhaps a more specific question e.g.: which element(s) is being investigated and at which depth within the sample (in micrometers or fraction thereof) one expects its presence, could yield a more precise answer.

Thank you for your answers, but my point is that I have seen some papers, put chemical composition, such as different oxides, by the use of SEM/EDX, any ideas about this situation?

Dear Seyyed, It will be helpful if you can provide at least one such paper on that context. So that we can understand from their point of view.

As from my experience, I have seen that people sometimes can provide chemical composition from from EDX provided all the K-shell X-rays are collected from all the elements. Say, if you have MgO, then you are pretty sure that at 20 KeV beam energy all the K-shell Xrays are collected for both the Mg and O. But, same is true true for Y2O3. Y being of higher Z, only L shell X rays will be collected. So, finding 2:3 ratio will be erroneous from EDX. Again, those people who quote this kind of numbers normally support it from XPS data. EDX can probe much deeper in the bulk compared to XPS. So, people sometimes use these two techniques simultaneously to have a detailed idea about their sample.

But bottomline is , chemical composition can never be determined accurately from EDX.

I agree with Rajib Kar's comments about the need of more specific information in order to solve the problem at hand and I agree about the caution needed to handle the data obtained by EDX.

Regarding the bottom line I would rather say that the accuracy of an EDX analysis is always a function of the analytical parameters, including instrumental settings and nature of the sample under investigation (including chemical composition, surface and bulk properties of the material; i.e. density, roughness, conductivity, state of hydration, etc. etc.). For instance, the determination of a metal alloy, such as Au-Ag-Cu on a polished and clean sample, can be determined with a very high degree of accuracy (at the right instrumental settings, of course!).

I also agree with Rajib Kar's about the inaccuracy of elemental analysis using EDX. Whats more if the difference of Z between two oxides is greater. Nonetheless, one can find it useful to estimate the phase occurrence, including oxides of course, especially when dealing with the stoichiometry of the materials. When the data is normalized, a neat calculation will give us a very close result with the one we would expected. Special attention has to be given when dealing with the volatile element such as Cl since some of them may easily oxidized due to electronegativity factor. Running the analysis at different point, would give us an average value which is more convincing that only at one point.