I would not overinterpret this. The values seem to be very similar. Keep in mind that EDX analyses are only (semi-)quantiatative when you work on well-prepared polished sections. In case you used fractured surfaces for EDX, the values given above are more or less meaningless.

EDS analysis, when used properly, is pretty accurate for showing increase/decrease of some elements in series of specimens.

- how many spots for each concentration were analysed? (I hope at least 5)

- were specimens polished?

- was conductive coating (carbon) applied?

Please provide more information:

Are those chemical compositions quantitative or semiquantitative? (Did you use standards on the EDS/SEM?

Are those compositions from the anhydrous cement particles?

Are the measurements from spot or area analysis?

Are these the average of how many measurements ?

The samples was finished using 1um diamond polishing paste and covered by carbon in high-vacuum carbon coater. We used quantitative analysis, the device is regularly checked on standard specimens. The measurements was taken in wide area analysis (approx. 2x1,5mm) and they were repeatable for multiple measurements of each sample.

Lets see the analysis of different phases of such a modified cement paste: hydrated and unhydrated cement, pores and marble powder.

According to previous answers, It would be important to consider the detection limit of the equipment and standard deviation of the experiment before further conclusions.

@ Andrzej Żak

It looks like you are a coauthor of the question.

Since EDS analysis looks OK, I would check specimen preparation. Repeat with cement mixing, in addition use 20% MP, etc., to check what is happening.

Concerning the experiments - as mentioned above you should do EDX on well prepared polished sections and analyze >50 data points (spots) in the hydrate matrix. As the interaction volume of the electron beam (at least 2 µm) with your sample is larger than the size of your hydrate phases, you will collect mixed analyses. What helps in this case and enables you to deduce which pases are present and to derive e.g. the Ca/Si and Al/Si ratios in the C-S-H are atomic ratio plots. A plot Al /Ca vs. Si/Ca with all your datpoints is reommended. See e.g. Article Hydration of a Portland Cement with High Replacement by Sili...

@ Frank Winnefeld

Why " >50 data points (spots)"?

If each measurement for specimen A is higher in, say, Ca, than that of specimen B, then - I bet - 3-5 measurements will be enough for statistically significant difference.

Because we can learn much more when we determine composition and nature of the hydration products. The information that we have more Ca in the hydrates does not bring us further in knowledge. With the spot analysis and the atomic ratio plots you can determine composition of C-S-H and also reveal which phases are present besides C-S-H and CH, such as ettringite and AFm phases.

But the original question was about Ca, not about phases present. It's different fish. Why adding more Ca to mixture gives EDS weaker Ca signal? - that was the question. We cannot explain it even if we know about hydration products. So 5 measurements could be enough.

The real question should be what we learn from such area scans. IMO not much, so it is a waste of instrument time. Point analyses or element mappings provide much more useful information.

I enjoyed reading all useful answers

and thank you for sharing it with us

Please repeat EDS analysis and maybe you can get a result for %20 then you will be sure. Marble has more CaO and normally we can wait more in the result of analysis.