Hard to say from your PDF... would not say so.

From what I can see, you worked at ca. 10 kV, which is why you go for the La line of Silver ? Plus, acquiring along a line is not the best way to do EDS when you're lacking signal.

I would suggest you do the analysis again, but at a higher magnification (if you can, of course) in order to try and get more counts, and do acquire an area on the nanofiber, not a line. Then, wait for enough counts to see if you have a Silver peak...

Plus, don't believe what the quantitative analysis tells you !

Thank you Pierre for your sharing. For your info, the energy is 15 kV.

Agree that scanning in line direction is not good. I have asked the testing service company for a re-test scanning across the nanofiber line (the test was not conducted myself).

It is known that SEM EDX mapping should not be used for quantitative analysis. Thank you for your reminder.

Plus, I would like to ask the meaning of Wt % Sigma if you may know. Thanks.

BRs,

David

Seems your subcontractor uses an Oxford Inst. system (AZtec software), from what I remember, Wt% Sigma is the statistical error for the Quant analysis (Oxford does not specify how they compute it). But this means that if this sigma is high, relevance for the wt% or at% is low.

All sigmas are high, even for Na which is a major element : your quant is worth nothing and I'm surprised your contractor even shows this...

Anyway, a few hundred counts as can be seen on the profile is nothing but noise.

Pierre Caulet Thank you for these info. The instrument is indeed Oxford. I have added three point scanning results for your info.

At this point, I really think you should go back to them and stress that you specifically want to tell if Silver is present or not. What I see in the three points analysis is that error is still very high. Not a surprise as they essentially analyze...Carbon from the sticky carbon pad below your sample.

At this magnification, the interaction of the electron beam with your sample is large : when you analyze a nanofiber, you mostly generate signal from a sphere, the volume of which is largely made of the carbon pad...

So, personally, I would first work at as high a Mag I could get way with, filling the screen with my fiber if possible, possibly limiting the accelerating voltage at 5 kV (this will still emit Silver La lines, but oveer a lower background), and I would run the analysis over a much longer time, checking to see if I finally generate an Ag peak on the spectrum or not.

Alternatively, they might try out at 30 kV to generate the Ag Ka line which would be very distinctive and would limit Na, F and C, but penetration into the stub would be much higher (in effect "diluting" the sample) and the interaction volume much bigger.

Or, even better, I would suggest they deposit the nanofibers onto a TEM grid (holes or lacey carbon), use a TEM holder for SEM/STEM, and do the EDS analysis where no supporting grid/carbon is present : you would not have the carbon background and the interaction volume would mostly include the fibers.

Not a routine analysis you have here !

I agree that scans were performed extremely poorly, by janitor, not by knowlegeable operator. I would not pay for such an "analysis".

Pierre Caulet Thank you for these professional suggestions. Very useful.

For EDS on fibres a starter could be:

J. L. Pouchou, X-ray Microanalysis of stratified specimens, Analytica Chemica Acta, 283, 81-97 (1993),

https://doi.org/10.1016/0003-2670(93)85212-3

Reliable EDS quantification would require the sample to be flat, homogeneous and thick.

In the paper STRATA (later STRATAGEM) is mentioned (commercial program).

A free program is BadgerFilm by Aurelien Moy

https://probesoftware.com/smf/index.php?topic=1373

I first thought that Vladimir's janitor comment was a little harsh, but then I looked at the results more closely.

Why aren't the C, O, and Au peaks labeled? They are significant features in the spectrum, but they are not quantified. I understand that Oxford (and others) will factor out the element from the coating, but it is very risky to ignore C and O when they are clearly present, even if they are from a carbon tab.

There are no clear peaks for F, Na, Al, Si, or Ag, but those elements are all quantified with some rather ridiculous results. The sigma values indicate that values are not to be trusted. Frankly, I think the values should have been higher. I cannot believe the F value is 80% with only 13% sigma.

I would like to see the fitted spectrum. I suspect that F is being used to account for the O peak since the O was not measured. Oxford would let them deconvolute the O peak then ignore it, at least in the full version of the software (not in Aztec One). I would also like to see the report of the net peak intensities so see if the numbers are reasonable.

Consider that the Au peak is probably from 10 nm of gold, but the signals from F, Na, Al, Si, and Ag are all MUCH less than the Au peak.

What is the fiber supposed to be made of? The image looks like SE. Was anything recorded in BSE. Is there reason to expect the heavier elements. From this evidence, I suppose the fiber is organic.

Were these spectra definitely from the fiber? Could the image have drifted? We have Oxford's AutoLock feature for tracking features in cases of image drift. Even then, we sometimes fail to track features and we end up with spectra from the surroundings. It would have been nice here if the operator collected spectra on and off the feature to see if there is any difference. They might also have scanned across the fiber, as you suggested.

Finally, lower voltage would have been better. I usually run at 10 kV for starters but might have tried 6 kV in this case. And include a BSE image.

In the previous post, I suggested that the F content was probably due to an O peak that was not accounted for. I have attached an example of a concrete specimen. When the normal elements are left out, there can be significant amounts of Cf and W measured - but they are in error. Notice the poor fit. Those elements drop to insignificant levels when the correct elements are included.

Warren Straszheim Thank you for your insightful response. I made the nanofiber by incorporating particles containing Ag into polymer containing F.

The reason I did not consider testing C and O is that the test was carried out by commercial lab and a maximum of five elements was allowed to be selected for point and line EDX spectra. Otherwise, attentional fee will be charged. (in hindsight I should have added C and O.)

I thought the F (from PVDF) could be seen as a basis, while the existing of other elements (Na, Al, Si and Ag) would prove the successful incorporation of the particles containing Ag.

The commercial lab have already re-tested the same sample for free, and please find below pdf. Original results are also attached for your reference.

Thank you again for the various points (conducting BSE, considering possible drifting, and testing at lower voltage.)

David Pp

My response is basically, the same. There is no evidence of F, let alone Ag in your sample.

I think it is unfortunate that you were limited in your choice of elements. Commercial labs can certainly choose to operate any way they please. I hope there was communication along the way to decide the elements. As an analyst, I would have been frustrated seeing that C, O, and Au were the main elements but not being "allowed" to work with them. I see no evidence of the 5 elements that you did specify. A form with checkboxes is too restrictive for this kind of work. Based on this, I would choose a different lab.

EDS does peak-fitting as part of the spectral analysis. If you do not allow O to match the peak at 0.54 keV, the software will do what it can with F and report some positive value (like 89%). I often do such a demonstration with the researchers here. S, Mo, and Au all have a peak around 2.3 keV. Any one of them can be used to account for a peak there if only one element was allowed. If you allow all three elements, the true match will probably reveal itself. The analyst must be allowed to check.

It is imperative to compare the final fitted spectrum to the raw data to see if the fit was reasonable. In this case, without C, O, and Au, the result would have been horrifying, and I chose that word carefully.

EDS can be very powerful, but it can also be quite wrong when done incorrectly.

David Pp

I loaded the .txt files (mas file format) into the free program NIST DTSA-II Neptune.

For the point spectra (using the files 谱图 1.txt, 谱图 2.txt, 谱图 3.txt, 谱图 4.txt; renamed) see: DTSA_Points.png

For the line spectra (using the files 线总谱图1.txt, 线总谱图2.txt; renamed) see: DTSA_Points.png

With my naked eye I would identify C, O, and some F.

The large peak could by Pt (sample coating??), no sign for Ag.

For an EDS line scan nowadays at every point a full spectrum is recorded, may be these line spectra are sum spectra along the line.

May be looking at single spectra could give insight of the fibre composition compared to the support.

Warren Straszheim I see your viewpoint about the EDS peak-fitting. Thanks a lot. :)

Stefan Baunack Thank you for your efforts to make the two pictures with the original data. Glad to see the the existing of F. It's good to consider EDS mapping instead of point and line spectra in the future. :)

David Pp

1. „I want my money back!” If the analysis was done on a commercial basis you should refuse to pay. If it would have been the 1st analysis of a student the next steps would be: give him/her something to read, do a discussion and afterward he/she would be allowed to do a 2nd try.

2. What is the analytical problem? The question was about silver, now you are lucky to see fluorine. I’m not familiar with the system, but beam damage (loss of F) should be considered. In this case point measurements would be the worst choice.

3. In the spectra generated from the text files the intensity is given in counts. EDS spectra are dominated by counting error. If you have N counts per channel, the statistical error will be sqrt(N). The overall intensity in the line scan spectra is lower, but the maximum intensity is similar, e.g. the time for every spectrum should be similar. If the line spectra are the sum over all spectra in the line scan, the aquisition time must be divided by the number of points (ca. 100), therefore the signal-to-noise ratio in a single measurement in a line scan much lower (worse). Therefore the line scans shown are likely useless.

4. Don’t try EDS mapping: the time per pixel is much lower, EDS maps are not suitable for quantification.

5. A suggestion would be: Define 3 areas of large length and small width (slit-like) and place the 1st over the fibre, the 2nd nearby and the 3rd some µm away (You can use beam scan rotation to align the image of the fibre). The large area will reduce beam effects. Use a sufficient time to acquire good spectra (30 sec).

6. After the measurement you should see the areas as contamination spots. If the SEM image shows that the sample (fibre) has moved, all efforts were in vain. If not: are the significant differences in the spectra? Evaluation of the results could start here.

Edit: 2024-01-05 16:23 typos corrected

I comment Stefan Baunack on his answers above. I have nothing to add to them but my voice to echo his thoughts.

Spectra give you the best statistics with about tens of seconds per spectrum. Maps give the least time per spectrum. We collect maps at 256x240 pixels for 10 minutes. That amounts to 10 ms of data per pixel. That is only a few hundred counts per spectrum, at best. It is dreadfully noisy. Thus, it is the least sensitive option. Line scans are better, but still fall short of Point&ID spectra.

I would have liked to reprocess the data myself as did Stefan, but I cannot yet. We have Aztec software, but apparently it is an older version. Oxford is notorious for its data not being backwards compatible. You have to have the same or newer software version to read the data.

I now see a small F signal in Stefan's plots. Those are displayed on a log intensity scale which shows the small intensities better. I could have done that within Aztec as well. DTSA (DeskTop Spectrum Analyzer) is a cross-platform tool.

I wanted to see what the peak fit would be like with and without the inclusion of C and O peaks. I hope to upgrade my Aztec version and see if I can then read your data.

I don't suppose you will get any money back, but I would look at a different arrangement or a different lab for future analyses. We charge based on the time involved rather than the number of elements. We will estimate a cost, but we do not use a "flat rate" for up to N elements. That is not well suited to SEM-EDS.