Reduced version of the composite image
of the Dhofar A-287 meteorite sample,
produced during the feature analysis run
In particle analysis it is a frequent requirement to find a small and rare mineral grain containing a specific element inside a large sample. The challenge here is to find as many as possible (if not all) of these grains, and to do it in a reasonable time. Manual methods are tedious and time consuming, while an automated feature analysis program – like ESPRIT Feature – can be very fast.
Feature analysis is the study of shapes, the calculation of its morphological parameters and positions and the tabulation of these results. For further information depth, the chemical composition can also be measured and particles can be classified subsequently by applying user defined rules to the obtained data (e.g. must contain Zr). With respect to analysis speed, the compromise lays in the resolution and magnification of the recorded and analyzed images. The number of phases that are analyzed by the system, the counts per spectrum as well as the count rate affect the speed.
In this case the analyzed sample was a lunar basaltic meteorite Dhofar287-A thick section hosting small grains containing baddeleyite (zirconium oxide mineral ZrO2 or zirconia with a stoichiometric ratio of Zr: 74.03 wt. % to O: 25.97 wt. %). The expected size of these grains can be as small as 3 μm or as big as more than 60 μm. All these grains needed to be found to measure them with radiometric dating methods such as a U-Pb using an ion probe. Grains under 3–6 μm would not be measurable with this method, therefore size filtering was a necessity, too.
The particles were initially detected by their morphological parameters and their brightness. As zirconium is a rather heavy element it can expected that baddeleyite particles will appear bright in the BSE images acquired for the search, binarization thresholds for image analysis were set accordingly. In the end this procedure found 997 possible baddeleyite particles in 90 fields of view and produced a 70 megapixel composite image of the 9 x 5 mm² sample for particle revisiting. EDS analysis of the initially found particles resulted in a list of only 11 particles that had the right size and composition for subsequent task of dating the sample.
The whole procedure took less than 1 hour 30 minutes, whilst a manual search would have not only taken many hours but would have included the risk of missing particles. This shows that using an particle analysis program like ESPRIT Feature is actually the only feasible approach for small particle search in large samples.