Petrology and geochemistry of Crozet hotspot alkali basalts : evaluation of the source mineralogy

The isotopic and trace elements compositions of hotspot lavas have clearly shown the presence of distinct mantle reservoirs. However, the sources lithologies responsible for their major elements compositions still remain unclear. This study addresses this issue using a new sample collection from the Crozet hotspot in the Southern Indian Ocean (CAPGEOS field trips 2009-2010). Despite the scarcity of isotopic data for this hotspot (13 samples), previous studies have identified HIMU and/or EM-type end-members in the source of Crozet lavas. This poorly constrained result certainly reflects the lack of a complete geochemical data set. For the first time we report a new and comprehensive set of major and trace elements data for 54 lavas from Crozet hotspot (Possession island). Major elements variations, which define an alkali silica-saturated magmatic suite, are best explained by fractional crystallization, accumulation (ol+cpx), and to a lesser extent magma mixing processes. Primitive lavas (MgO>9 wt.%) display CaO/Al2O3 ratios >1 and low SiO2 contents (45 wt.%), suggesting that these compositions were produced by the partial melting of a garnet bearing peridotitic source or a pyroxenitic source. In order to provide additional constraints on phase equilibria and on the nature of the source, we propose to invert the REE concentrations of Crozet lavas following the method from Hofmann and Feigenson (1983). Variations in LaN/SmN ratio range from 2.5 to 4.7 (for samples with MgO>5 wt.%) suggesting subsequent heterogeneities within the source. We first select samples with MgO contents varying between 5 and 10 wt.% to minimize the effects of crystal fractionation and crystal accumulation. The cogenetic origin of the lavas was ensure by screening their chemical compositions for constant LaN/SmN (here, 3.4±0.4). Primary melts compositions are then calculated by removing the effect of fractional crystallization in two steps. First, by correcting our data set to a common value of 8.7 wt. % MgO using the observed LLD, then by incremental addition of olivine until reaching equilibrium with olivine Fo90. The results of our REE inversion indicate: (1) a source strongly enriched in light REE relative to heavy REE with LaN/SmN ratio of ~2.1, and (2) a low garnet contribution (less than 1%) to the melt chemistry. We propose also that the computed set of bulk partition coefficients for the source (D) and for the melt (P) are compatible with a peridotitic source mineralogy that has less than 1% garnet, 10-20% clinopyroxene, and 80-90% olivine +orthopyroxene. From these results we can also estimate that degree of partial melting for Crozet lavas to be around 4%. While the major elements compositions of Crozet lavas might call for a pyroxenitic source, the systematics of their REE argues in favor of a garnet free peridotitic source. In order to resolve this paradox we will: (1) propose a better evaluation of the problem of crystal accumulation in the estimation of primary melts compositions, and (2) consider a melting model involving carbonated melts and their relevant partition coefficients.