Application of proton-microprobe data to trace-element partitioning in volcanic rocks

In situ proton microprobe analyses are used to calculate partition coefficients (D) for Mn, Fe, Cu, Zr, Ga, Rb, Sr, Y, Zr, Ba and Pb, and more restricted data for Ni, Ge, Nb and As, in pyroxenes, olivines, biotites, hornblendes, feldspars and Fe-Ti-oxides, from 32 volcanic rocks (all with glassy matrices). These range from leucitite and picritic basalt to high-silica and peralkaline rhyolites. Data for pigeonite, aenigmatite, allanite and leucite are included, where these occur as additional phases. Coefficients have also been Rayleigh fractionation corrected, this correction being significant for those elements with high coefficients in specific mineral phases, and where pronounced crystallisation has occurred. Partition coefficients are presented as plots vs. NBO/T ratios of coexisting glasses (=melts). Negative correlations are observed, based on log-transformed data, for Mn, Zn, Fe and Y (pyroxenes), Mn, Zn and Fe (olivines), Mn and Zn (Fe-Ti-oxides), Nb, Y, Sr and Zr (hornblendes), Sr (plagioclase), and Ba, Cu, Zn and Fe (feldspars). These element partition coefficients are not correlated with bulk mineral compositions. Zr and Ga are not correlated with either glass or mineral compositions (except Zr in hornblende). The correlations support the role of melt polymerisation in controlling element partitioning, at least for those elements listed, although possible effects of temperature cannot be uniquely isolated in the present data set. The coefficients presented here are generally consistent with published values for equivalent rock compositions, one difference being that some of the higher published coefficients (e.g., Sr in plagioclase) are not matched in the present data set. From a practical viewpoint, the relative regularity of the coefficient data presented does facilitate choice of relevant coefficient values that are required for petrogenetic modelling procedures.