Trace element variations in olivine phenocrysts from Ugandan potassic rocks as clues to the chemical characteristics of parental magmas

Abstract

Olivine phenocrysts in ugandite and leucite basanite from the western branch of the East African Rift have been analysed for up to 34 trace elements by Laser-ICP-MS with detection limits as low as 1 ppb. A combination of point analyses with varying ablation crater diameters and line scans allow the identification of subtle zonations from core to rim, as well as characterization of the chemical effects of contamination along cracks. Trace element concentrations are remarkably uniform between large and small phenocrysts; fractionated leucite basanites (Mg# 59) have higher DCa and DAl, and less fractionated LREE/HREE than MgO-rich ugandites (Mg# 75-80). Minor zonation is seen in elements with cation charges from 5+ to 2+ (P, Ti, Zr, Cr, Al, Sc, V, Cu, Mn, Ni) and show correlation between Ti and Al, but not P. Early phenocryst cores with high Li or Ni, low Mn, or enrichments in many trace elements can be identified, whereas xenocrysts have exceptionally low Na, Cr, Ti, V and Co. Partition coefficients for Ni are 31-35, less than in lamproites, with which they demonstrate an approximately linear correlation with K2O content, K2O/Al2O3 and K2O/Na2O of the melt, but none with SiO2 content or Mg#. D-values for Cr, Mn and Co overlap with those of basalts, whereas those for Sc (0.011-0.018), Zn (0.44-0.49) and Ga (0.006-0.007) are lower. DV of various potassic rocks (0.015 in the Ugandan rocks) confirms the dependence on fO2 calibrated by the Fe3+/(Fe3++Fe2+) of spinels; the Ugandan potassic rocks crystallized at fO2 = FMQ to FMQ + 1. The ugandite olivines have some trace element characteristics reminiscent of those in metasomatized Kaapvaal peridotites, but not ocean islands. Line scan analyses are contaminated in Al, Ca, Cu, Ga, Sr, Zr, Nb, La and Ce, elements that are also concentrated in microcracks between subgrains, indicating smearing out during polishing, and demonstrating that large spot analyses produce the best results.