QACD-quack

The experimental method that was used to determine the optimal conditions for map collection is provided in consideration that the optimal conditions for efficiently collecting element maps of sufficient quality for the QACD method will vary between SEMs and detectors. Carbon-coated samples were analysed using the Zeiss Sigma HD SEM outfitted with dual 150 mm2 active area Oxford Instruments Xmax SDD EDSs in the School of Earth and Ocean Sciences at Cardiff University. The SDDs were controlled and the output from the digital signal processor (DSP) integrated in the Oxford system was processed by Oxford Instruments Aztec Software package. Optimisation of the mapping procedure was carried out using the the Plagioclase, Olivine, and Diopside standards from Astimex (see supplementary Table 1 for compositions). These X-ray maps were acquired with a constant beam accelerating potential of 20 keV while systematically varying individual parameters in order to address the effect of each parameter on both the quality of the map output and the total live time needed to collect it. These parameters include pixel dwell time (i.e. 500, 1000, 5000, 10000, 20000, 50000, and 100000µs), and beam current (i.e. 1, 2.5, and 5 nA). A full write-up and the details of the experimental method for optimising element map collection is provided in the supplementary material. Although the beam current was varied as a parameter, it was kept within limits of that needed to generate adequate X-ray flux to reach a reasonable deadtime and a combined output count rate >200,000cps on a pure copper target as a chosen reference standard. The beam current was measured with a stage mounted Faraday cup. However, due to known difficulties in calibrating stage mounted faraday cups with known current sources, the absolute value of the beam current was not known, but the repeatability was estimated to be better than 0.2% relative based on the independent measurement by the spectrometer of the integrated X-ray counts from the Cu standard.

 

For the optimisation, the resulting maps were post-processed with a simplified version of the QACD method (see Section 3.1). The maps were imported and processed according to the QACD Initialize step (i.e., pixel noise and a 3-by-3 median filter). Once imported, each map is added together to produce a map of the total number of counts at each pixel. Each map is then divided by the total counts map as a means of normalising the data. This process of normalisation allows for spectra collected at various dwell times and beam currents to be directly compared. At this stage in the correction procedure there is still some variation between the spectra normalised peaks of various phases. This variation can be further reduced by calculating the mean atomic number (Z), mean atomic weight (A), and the h parameter (h = 1.2 x (A / Z 2)) of the absorption correction defined by Philibert (1963) for each pixel and multiplying the spectrum normalised elements by h.  For the purposes of ratio map calculation, the derivation and application of the h parameter to the spectrum-normalised elements has no effect on the resulting element ratios. Finally, the element ratio maps are corrected according to equations derived from a growing database of both analyses of standards and unknowns and theoretical EDS spectra generated using the DTSA-II software package from NIST (Ritchie et al., 2008; Ritchie, 2009)(Figure 1).

 

The resulting histograms of An in plagioclase (Ca / (Ca + Na + K)) and Mg# in pyroxene and olivine (Mg / (Mg + Fe)) were used to compare the quality (i.e. peak mean, median and 2σ error) of the resulting compositional peak for each histogram. For an accelerating potential of 20 keV on the Cardiff University SEM, a pixel dwell time of 20,000 µs, or 20 ms, and a minimum output count rate of 200,000 cps are required to produce element ratio peaks that fall within a 2σ error of the reported standard value (see Figure 2). For pixel dwell times above 20 ms there is no significant decrease in error, suggesting that there is no advantage to longer dwell times aside from the imaging of lower concentration elements.