Hello Fouad,

As I can see in your sample you only have three elements (O, and Si with traces of Cu from the Tem holder as you comment), however, which is the objective of your work?  If it is only the identification of the element ratio or something else.

As this may follow which approach you use to obtain the best use of a tool like EDS. I have already measure many samples with an EDS-SEM and something that is always an absolute true is that samples do not lie so if you have the same result you that is the reality, you can used others instruments to corroborate this like XRD or NMR.

I attached to this message a paper where they used a EDS-SEM to qualitative determine which are the elements for their sample. Hope this will help you.

Dear Jose

Thank you for your response but I performed XRD, HRTEM, N2 sorption (BET, BJH,t-method, MP-method) and SEM, all of these analysis shown that the final material has been synthesized and I published my work. I'm not looking for any other method to figure out if the material has been synthesized or not.

My question is do you think the EDS is really can be used as a quantitative method or not? (I mentioned my job to show an example for inaccuracy of EDS based on my personal experience)

Hello Fouad,

I would use EDS quantitative but with reservations, as a method that support the results with other measurements (like the ones you comment on your last message).

Normally quantitative analysis depended much on the method and correction. For EDS method, the correction that must be done for quantitative analysis is ZAP correction.

I had a look on your results. Atomic ratio (A% Si:O=26:74) in TEM is much different with one in SEM (Si:O=41:58), right?

EDS could be used as a quantitative analysis, the best results are obtained for  "heavy" elements. The oxygen is not so "heavy" and its determination could be not precise. Have you tried to do EDS on different particles? I suggest you to use flat particles or at least without parts too much depth. I think that the 1-1.5 ratio is not so bad...considering the kind of your sample

I agree Marcella Bini and Bac Luong.

Also, I remember you that in EDS-SEM the intensity and the ratio is about Kalpha1 100% versus Kbeta1 is 20 %.

This is different by intensity and the ratio in Kalpha1 and Kbeta1 in EDS.

I don't know  EDS-TEM.

Please see you AXIL - QXAS X-RAY Analysis Software for model definition of source/detector/sample tipology/relative geometry and read Jenkins, R., De Vries, J. L., Practical X-ray Spectrometry, Springer-Verlag

Thank you all for your help.

I really don't know about correction in the EDS. Because I did not performed EDS analysis for samples by myself. (We have a EDS operator who does this work).

I did analyze other samples and heavier elements. (Zn/MCM-41, Fe2O3/MCM-41, Sn/MCM-41, SO3H(CH2)6/KIT-1, etc.) In all samples the ratio of metal loading was less than I expected it to be.

Prof. Bini gave very interesting answer. In all samples I have mesopores, could it be possible that some elements were buried in the pores and EDS could not detect them because it could not penetrate the sample that deep?

It is both qual and quant given certain assumptions:  a sample that is homogeneous within the interaction volume, perfectly flat and horizontal with respect to the incident beam and stable under the electron beam. You need to have performed suitable calibrations to obtain quant information. Your analyses are using standardless software, which is getting better for light elements, but such results are at best semi quants. If you have good material that fulfils the requirements above then yes you can get quant results. One of the problems with mesoporous silica, is its porosity, so the interaction volume is full of space rather than SiO2. This will affect the results. 

A comment. On EDX2 the peak resolution looks to be quite poor the oxygen is not well separated from the noise peak, The noise peak is not centred well on zero. You have very few counts which will adversely affect the counting statistics. 

Mesoporous silica is very susceptible to charging effects.

Put a solid quartz sample in both and see what the difference is there. You should be getting to within 2% of the correct weights. 

You have used standardless method of quantifucation, which is useless for light elements such as O. You need standards to make calibration. It is easy for your task - relatively pure SiO2 is easy to get. I have noticed also that accelerating voltage was not imputed for calculations - another mistake. And how you got that "ratio of Si:O;1:2 but EDS-SEM showed 1:1.5 and EDS-TEM showed 1:1 ?" From you tables these ratios are 1:2.8 and 1:1.4. 

So, again, your analysis without standard is useless for quantification. Good news are that you do not even need quantification since you want just  to confirm that your specimen is SiO2. Prepare (polished) specimen of SiO2, coat it with carbon sumiltaneously with your specimen (so that you'll have the same thickness on both pieces), acquire spectra and compare them. You may need to scale them to the same height if one of them is porous.

Thank you all for your help.

Stoichiometric compounds like SiO2 are beautiful samples to test the performance of EDS systems and the accuracy and the precision of quantification results. In the past, the quantification of light elements and low energy X-ray line were challenging, especially when using a standardless method. However, the hardware and software performance has improved significantly within the last decade.

I would like to refer to my article with comparison of standardless and standards-based quantification of a stoichiometric (Ga,Al)N semiconductor sample. It documents that a relative error of

Dear Prof. Salge thank you for sharing your knowledge. I really appreciate your help.

Heavy elements can be best analysed quantitatively by EDS. Light elements may give unexpected results.

It looks as if you used standard P/B-ZAF processing, which works best on heavier elements.  For lighter elements, they recommend Phi(Rho, Z) processing. Edit: I used both methods. Both failed on the light elements (using 15kV). I ended up using P/B-ZAF, which yielded excellent results for heavy elements in minerals of known composition. I compared samples using ratios of heavy elements.