So many spectra

So I'm well and truly in data processing mode. My post-excavation job is to process, number-crunch, and interpret all the pXRF data that I collected in the Caribbean. In the first instance I have to write a report on the ceramic material from the La Poterie excavation which I analysed with pXRF. Then, I'll have to do the same for the St. Vincent material. Hopefully (fingers crossed and a prevailing wind) the data will tell us something really interesting and useful, which I can then use as a basis for a journal publication or a conference paper. Ultimately, my interpretation of the data will feed back into the larger project. At the same time as I am doing this work, other people are working on collating and interpreting the other data from the excavation. All of these smaller, specialist, interpretations will be gathered together and the dig directors will use all the information to recreate the story of the site.

What do I actually have to do? First I have to check all the spectra and make sure that the measurements were good. While I do that to certain extent in the field (since I no longer have access to the fragments, I cannot redo any measurements), it is possible that a measurement which looks good in the field, is bad when compared to other measurements of the same sample. The image below gives an example of this. These are all measurements of the same piece of glass (you can see the long list of measurements on the left). However, the red spectrum (indicated by the blue arrow) has shifted. So, I need to go back through my field notes and see if there is any reason why this measurement may be different. Did I have an air gap between the sample and the instrument? Did someone jog the table during the measurement? Was the temperature/humidity of the room noticeably different?

I then need to decide if this measurement is still usable. Can I correct the spectrum or not? I need to do this for every measurement I made, and while at La Poterie, I made nearly 700 measurements. Once I'm happy that my measurements are good, I then need to group them based on the elements present in the spectra. That means I need to identify every peak in a spectrum. The example here is a green glass, and you can see the big silica (Si) peak, calcium (Ca) peak, and iron (Fe) peak. Silica (sand) is the main ingredient in glass but this has a very high melting temperature, so an alkali flux (usually sodium or potassium) is added to lower the melting temperature. However, the pXRF is not sensitive enough to measure sodium, so I cannot see this in the spectrum, but it is still in the glass. Calcium is added to glass to stabilise it and to stop it dissolving in water. The iron is usually an impurity in the sand (it's iron that makes the glass green). The reason the iron peak is so big, is not because there is a lot of iron in the glass, but because the Si matrix is, chemically speaking, very light and iron is a heavy element relative to silicon. So, the Fe fluorescence finds it really easy to push through the matrix and reach the detector, hence it registers more. But more on interpreting the spectra later.

So, I have to identify all the peaks, then I need to tell the software all the elements present in every group of measurements. The software will mathematically calculate the area of each peak, and it will spit this information out as an excel file. The next step is to work with the peak area data, but that I will save for another blog...