A Partial Record of Mixing of Mantle Melts Preserved in Icelandic Phenocrysts

Abstract

The record of mixing of mantle melts in magma chambers has previously been observed in the compositions of olivine-hosted melt inclusions from Borgarhraun, a primitive basalt flow from the Theistareykir volcanic system, northern Iceland. Borgarhraun also contains high Mg-number (85–92) clinopyroxenes, which exist in polycrystalline nodules and as phenocrysts. Coincident major and trace element analyses were made in compositional zones of these clinopyroxenes, and Ce/Yb ratios of the melts in chemical equilibrium with each of the clinopyroxene zones were calculated using carefully selected crystal–melt partition coefficients. These calculations allow direct comparison of clinopyroxene compositions with existing melt inclusion data. The range of Ce/Yb ratios in the crystals and in the equilibrium melts cannot be accounted for by crystallization alone, requiring simultaneous mixing and crystallization of compositionally variable mantle melts. However, the range in Ce/Yb for melts in equilibrium with these high Mg-number clinopyroxenes is smaller than that of melt inclusions hosted by olivines with equivalent Fo contents. Also, the mean composition of the melts from which clinopyroxene grew has significantly lower Ce/Yb than the olivine-hosted melt inclusions. The record of mantle melt variability in clinopyroxenes is thus biased towards more depleted (low Ce/Yb) melt compositions. This bias can be understood if the trace element variation in the Borgarhraun parental melts is coupled to major element variation, as expected from petrological parameterizations of mantle melting. The major element variation influences the phase relationships and controls the appearance of liquidus phases during fractional crystallization in near-Moho magma chambers. Small-degree, deep melts, formed in the presence of garnet, have high Ce/Yb ratios. On cooling, these melts have a longer olivine-only crystallization path than melts derived from the shallow mantle. When these deep-sourced melts eventually become clinopyroxene saturated, they have too low Mg-numbers to crystallize high Mg-number clinopyroxenes such as are found in Borgarhraun. In contrast, shallow, depleted melts saturate in clinopyroxene at high Mg-number. The delayed onset of clinopyroxene crystallization in the enriched melts, coupled with concurrent mixing and crystallization of melts generated at a range of depths in the mantle, can account for the difference in the distribution of the trace element composition of high Mg-number melts saturated in olivine and clinopyroxene. The trace element compositions of high Mg-number clinopyroxenes in Borgarhraun therefore provide only a partial and biased record of the mixing of mantle melts. As well as showing that melt mixing may be preserved in phenocryst compositions, the results illustrate that trace element disequilibrium between crystals and carrier melt can be a consequence of magma mixing, rather than necessitating a xenocrystic origin for the crystals. Furthermore, care must be taken when using clinopyroxene separates from primitive basalts to examine compositional heterogeneity, as they provide a record of the chemical evolution of the magmatic system that is biased towards depleted compositions and therefore incomplete.