We have measured the effect of alkalis on S concentration at sulfide saturation (SCSS) in an under-explored compositional space of natural hydrous arc melts (basaltic andesite to shoshonite) at 1270 °C and 1 GPa. At an oxygen fugacity approximately 2.5 log units below the fayalite-magnetite-quartz (FMQ) buffer, SCSS increases with Na2O (562 ppm S/wt% Na2O), K2O (98 ppm S/wt% K2O), and total alkalis (88 ppm S/wt% Na2O+K2O) over the compositional range we have studied (1.6–3.1 wt% Na2O; 0–6.5 wt% K2O; 1.9–6.3 wt% FeOtot). Experiments with ∼1.3 wt% H2O show approximately half the increase in SCSS with alkalinity compared to those with ∼3.0 wt% H2O. Our results show a possible limit to the increase in SCSS solely by increasing alkali concentration at ∼7.5 wt% total alkali concentration. Using our results and published data, we retrained earlier SCSS models to provide a better fit to test data. We also developed a new empirical model using theoretical optical basicity as a compositional parameter that predicts SCSS in the overall data set with slightly better accuracy compared to previous models:
with temperature (T) in Kelvin, pressure (P) in GPa, the optical basicity (Λ) and mole fractions (X) of FeO (calculated from Kress and Carmichael 1991), and H2O in the melt. The discrepancies between observed and predicted SCSS for our experiments of varying alkalinity reflects the heavy bias toward anhydrous, alkali-poor basalt compositions in the underlying data sets on which most models are developed.