Electron Probe Microanalysis of Trace Sulfur in Basaltic Glasses and Silicate Minerals

Abstract

Abstract Sulfur (S) in the mantle is conventionally assumed to be exclusively stored in accessory sulfide phases, but recent work shows that the major silicate minerals that comprise >99% of the mantle could be capable of hosting trace amounts of S. Assessing the incorporation of trace S in nominally S-free mantle minerals and determining equilibrium S partitioning between these minerals and basaltic melt requires analyzing small experimental phases with low S contents. Here, we develop a protocol for EPMA analysis of the trace levels of S in silicate phases. We use a suite of natural and experimental basaltic glass primary and secondary standards with S contents ranging from 44 ppm to 1.5 wt.%. The effects of beam current and counting time are assessed by applying currents ranging from 50 to 200 nA and total counting times between 200 and 300 seconds at 15 kV accelerating voltage. We find that the combination of 200 nA beam current with a 200 second counting time (80 second peak, 60 seconds each for upper and lower background, respectively) achieves precise yet cost-effective measurements of S down to a calculated detection limit of ~5 ppm and a blank-derived, effective detection limit of ~17 ppm. Close monitoring of the S peak intensity and position throughout the duration of each spot also shows that high currents and extended dwell times do not compromise the accuracy of measurements, and even low S contents of 44 ppm can be reproduced to within one standard deviation. Using our developed recipe, we analyzed a small suite of experimental clinopyroxenes (cpx) and garnets (gt) from assemblages of silicate partial melt + cpx ± gt ± sulfide, generated at 1.5 to 3.0 GPa and 1200 to 1300 °C. We find S contents of up to 71 ± 35 ppm in cpx and 63 ± 28 ppm in gt, and calculate mineral-melt partition coefficients (Dsmin/melt) of up to 0.095 ± 0.064 and 0.110 ± 0.064 for Dscpx/melt and Dsgt/melt, respectively. The sulfur capacity and mineral-partitioning for cpx are in good agreement with SXRF measurements in a prior study by Callegaro et al. (2020), serving as an independent validation of our EPMA analytical protocol.