Element zoning trends in olivine phenocrysts from a supposed primary high-magnesian andesite: an electron- and ion-microprobe study

Abstract

Electron and ion-probe microanalysis have been used to obtain zoning profiles for major and trace elements in olivine phenocrysts from a high-magnesian andesite from Shodo-Shima island, southwest Japan. This rock was previously thought to represent undifferentiated, primary magma. Some crystals have unzoned cores, while others show cores which are reversely zoned with respect to Mg/ (Mg+Fe), Ni, Mn and Cr. In addition, some Ni profiles show a normally zoned “hump” at the most central portions of the reversely zoned crystals. All crystals show normally zoned rims. The Li concentrations are constant throughout the cores of all crystals studied, but rise sharply, by a factor of up to at least six, in the rims. The Ca and Co concentrations are essentially constant throughout all the crystals. Mechanisms for producing the observed zoning profiles are discussed, and it is concluded that the reverse zoning was produced by the introduction of crystals into a less differentiated magma than that in which they grew. The reversely zoned crystals could therefore represent xenocrysts which were introduced into an undifferentiated magma, or phenocrysts introduced into a more primitive magma by a magma mixing process. The Ni profiles are used to estimate the residence time of these crystals in the more primitive magma. The following trace element partition co-efficients have been estimated for the olivine-groundmass system in this rock: DNi=16; DMn=1.1; DCo=4.2; DCa =0.02; DTi=0.005; DV=0.05; DSc=0.2; DNa=0.0002. Studies of trace element zoning will become increasingly important as the new generation of trace element microprobes become available but a larger database of experimentally determined values for trace element partition coefficients and diffusion coefficients in crystals and magmas, and a better understanding of other disequilibrium processes are required to fully exploit the new data.