Hi Jerzy,

ni matter what is the scale size of the particles, scale size gives the % ratio (In EDX), so concentration can be determined. Secondly accuracy is approx. 95% while using EDX as i experienced. But for more accuracy, you can use RBS (Rutherford back scattering) and Pixi Analysis.

Thanks

Dear Jerzy, from my experience with EDX quantification on an FEI Titan equipped with ChemiSTEM technology, I can tell you that the size of particles matters. The smaller the particle, the longer the acquisition time to have sufficient counts to quantify the concentration. Tipically with a SuperX detector you can quantify without problems particles with a lowest diameter around 20-10 nm (10/5mins maps). Smaller particles will require ~>20 min acquisitions which results in the particles melting since they are really small. At least that is my experience. Furthermore the quantification accuracy strongly depends on your quantification method. Generally you won't achieve a real accuracy better than a few percents, so forget about 0.something percents detection, the signal won't be enough to raise above the background radiation. Regarding the quantification, the most common method is the Cliff-Lorimer, you can use that one, but beware that if you don't measure the k-factors from standards, and use calculated ones, you may end up with an error of 15%-20%. I would suggest you to read the following review for further information on the topic:

http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2818.2006.01549.x/full

http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2818.2006.01549.x/full

As a pontual analysis, the results are pontual. If you have enough sample, a XRF analysis could help you.

Thank you for sharing your thoughts. Let me be more specific about my sample:

It is basically zirconium but It contains 2.5 wt% Nb and 0.1 wt% Fe. After long heat treatment at 580oC the particles are formed in the size range of 10 nanometers.

They are expected to contain 85 wt% Nb, 10 wt% Zr and 5 wt% Fe and the Zr matrix is expected to contain 0.3 wt %Nb in solution and no Fe.

My questions are: Can one determine the composition of such particles, and especially the percentage of Fe in them by analytical EDX? Can one firmly say that Fe is or is not in the particles?  

In my view the difficulties are:

1. large matrix burden because particles are so small, 2. Fe is a low-Z element in high-Z surrounding, 3. Fe is only a small fraction 1/1000 of the total mass.

Can anyone run ZAF correction procedure in reverse?

To simulate how the EDX spectrum would look like for particles as described above?

Dear Jerzy,

what system are you using?

Keep in mind that it might be difficult to obtain reliable data on Fe concentrations due to spurios peaks generated by electrons scattered on the pole pieces. I don't think the Z number would be a problem since the fluorescence yields do not depend on Z.

Furthermore, if I understand correctly you also have another main problem. Since the particles are embedded in a matrix, you will have zirconium over and below your particles. In a TEM you are only able to get 2D projections and without electron tomography I am afraid you won't be able to perform the analysis you want. The concentrations are relative to all the interaction volume of your beam with the sample. When you scan over a particle you collect EDX signal also from the Zirconium above and below the particle. Thus when performing quantification you will obtain an overestimation of the concentration of Zr in the particles and underestimation of all other elements. The concentrations will then depend on the thickness of the Zr matrix.

To obtain the composition of the particles you would need to perform an EDX tomography experiment. To that purpose you would need a ChemiSTEM system and dedicated holders.

Alternatively you may be able to estimate the thickness of your sample with EELS and draw some conclusions on the real concentrations, assuming that you know the size of your particle, but I guess the best you could obtain in this case would be a rough estimation of the Nb w%.

Dear Daniele and other responders:

I'm trying to understand the EDX spectrum in Fig. 2(c) and conclusion in the paper

file:///Users/jerzysawicki/Desktop/The%20role%20of%20Fe%20on%20the%20solubility%20of%20Nb%20in%20α-Zr.html

that "there is not Fe in b-Nb particles".

My Mossbauer data in these alloys suggest that there is Fe in b-Nb particles.

As far as I understand X-ray generation by an electron beam the fluorescence yield plays a role, especially when atomic numbers of analyzed elements much differ.

The ratio of K-shell fluorescence yields of Nb to Fe is equal 2.3. This, in addition to heavy matrix burden, further suppressed the Fe signal in their Fig. 2(c). I attach my reinterpretation of this figure where the tiny peak at 6.4 keV has been multiplied by 2.3. It suggests that intensities of 6.4 keV line is about 1/20 of the intensity of Ka Nb line. It just would give about %Fe in b-Nb(85%Nb, 10%Zr, 5%Fe) particles.

Thank you for your interest and help.

Dear Jerzy,

I gave a loot at the article you cite.

From what I see the particles analyzed are in the hundreds of nanometers in size. From what I read the authors use TEM-EDX for qualitative purposes, just to see if they have Fe or not in their sample. They don't use EDX to quantify the concentrations, indeed it is not possible when you a matrix and your particles also contain the matrix element. Quantification in EDX is done through a ratio technique so either you have no matrix or at least you already know that your particles do not contain the matrix element, so you can exclude that from the analysis.

In the work I suppose they use Electron microprobe diffraction to estimate the lattice parameters of the particles and from those, the concentrations of Nb. In the caption of Table 1 you can read as they refer to beta-Nb "bcc phase with lattice paramenter of 0.33nm, corresponding to 90%wt Nb".

EDX can't yield concentrations as reported in Table 1, accuracy is at best of the order of few percents, not less, when analyzing nanometer sized objects.

In your case if you have 10 nm particles, I think you would need to nanoprobe diffraction, which usually requires a microscope with proper condenser apertures such to achieve a nanometer scale parallel beam.

If you can find isolated particles you may as well be able to just acquire a diffraction pattern from normal diffraction mode, estimate the lattice parameters and then compare to those of the pure phase or those taken before the annealing.

So to summarize, you don't need to perform EDX quantification because in this case it is not the technique able to yield the information that you are looking for.

Hope this will help! Good luck!

Daniele

I agree with Daniele.

In TEM microscopwill be easier

D.

Please, I have a question:

Does the Energy-dispersive X-ray spectroscopy (EDX or EDS) provide the chemical composition of the material as metal or metal oxide? i.e the results obtained from the EDX are in percent for metal or for metal oxide?

I found in different published papers that the data obtained from EDX are in % of metals and other papers as % metal oxide. When we use it as metals and when as metal oxide?

Is the following sentence right or false:  "EDX determines the chemical composition of materials as metals and exposes them as oxides but it doesn't mean that they are oxides". Many thanks for your reply.

Both EDX and EDS provide the percentages of elements detected in a given sample, not as in the form of oxides or other forms, if the proper software and calibrations are employed to perform all corrections - such as ZAF correction usually, or more advanced ones provided with the new microscopes.

These are percentages normalized to 100% as the sum of percentages for elements detected and quantified. Not all elements, such as very light ones are detected and therefore are not included in the 100%. Also, elements present in a small quantity, even such as iron, etc, below about 0.3 wt% cannot be quantified, or are quantified with big uncertainty. Another feature of these techniques is that one has to record quite many spectra from various areas of the sample to increase the accuracy of determining average composition, for instance, in alloys with precipitates. But, we are still lucky to have these techniques available in the electron microscopy.

X-ray fluorescence, using X-rays or gamma ray sources, if done carefully with proper calibrations and knowhow, can be somewhat more accurate as it can analyze much larger specimens

I may also add that listing the results of elemental analysis in a form of oxides percentages seems to come from chemical wet analysis traditions. I was often given by the analysts such data even if I knew that my solid samples, such as alloys, didn't contain iron oxide or other metal oxides.

Dear Dr Metwally,

Please try search in Google: Quantitative EDS X-ray microanalysis using SEM

There is a good tutorial available under this title and other SEM topics there.