High content of potassium reported by spectrometer in silverware items

Question

We tried to analyze a few silver items (800/1000 fineness) using a hand-held ED-XRF spectrometer, which is normally used for analyzing industrial alloys etc. The operating element range of the device was Mg-U.

The most surprising result of the experiment was a very high reported content of potassium, often higher than silver. For example, one of results (listing just elements with content >= 1%):

| element                           | content |
|-----------------------------------|---------|
| K                                 |   36.0% |
| Ag                                |   27.5% |
| (sum of elements lighter than Mg) |   24.7% |
| Cu                                |    5.0% |
| Sn                                |    4.5% |

Such high potassium content was pretty consistent across all our silver samples, but not present in others (like stainless steel)

Clearly it's not possible that the alloy could contain that much potassium. Where could this result come from? Can potassium be somehow bound to the molecules at the thin surface layer? Or is there some explanation for this phenomenon?

Answer 1

To give some numbers to MaxW's comment, you can see on this webpage the x-ray absorption energies of basically all the elements.

For example, the K-edge of potassium is $3.61~\mathrm{keV}$. The L-II edge of silver is $3.52~\mathrm{keV}$ and the L-1 edge of silver is $3.81~\mathrm{keV}$. So, it seems like you are somehow automatically assigning the peaks and scanning over a very large range of energies. The problem is that this will run into problems if you assume that all transitions are out the $1s$ orbitals (K-edge) because there are L-edge transitions for heavier elements which are comparable in energy to those from lighter elements.

So, without more information about how the transitions are being assigned, it seems very likely to me that MaxW was right to say these are L-edge transitions from Ag.

Answer 2

To put the other answer in a graphical context, I took the X-ray emission spectra of K and Ag from some place online*, and plotted them together, crudely resized to fit the same scale. Some of the Ag L-lines (I think it's Lβ 1 and/or 2) are at exactly the same position as the K Kɑ line. So, the software sees a certain amount of counts at K, and identifies it. Good software knows that there's silver there and the K is spurious. Badly written software does not.

The solution is usually simple: just untick potassium in the element identification menu (or whatever it is called on your device).

^{*The specific source does not matter. Any web search for "silver/potassium EDS spectrum will look similar.}