Résumé
We report solution-ICPMS analyses of Rb, Ba, Th, U, Nb, Ta, REE, Sr, Zr and Hf for acid-leached minerals of anhydrous spinel peridotites and websterites from the Ronda peridotite (S. Spain). The same elements were also analyzed by LA-ICPMS in the silicates of three peridotites. The results obtained by solution-ICPMS and LA-ICPMS are similar for the less (HREE) and the most incompatible (Rb–Ba) elements, and provide comparable inter-element distribution coefficients (
D
xt/cpx) for these elements. However, moderately incompatible elements (typically LREE) show significant discrepancies between solution and in situ analyses.
D
opx/cpx and
D
ol/cpx for these elements are generally lower for solution than for in situ analyses.
D
xt/cpx for MREE, HREE, Zr and Hf are consistent with experimental values. In contrast,
D
xt/cpx for highly incompatible elements and LREE are higher than expected from available experimental data and/or crystal-chemical considerations. The observed
D
xt/cpx for the most incompatible trace elements may be explained by very small amounts of melt/fluid, or solid, inclusions trapped in these minerals. Inclusions would affect both solution- and LA-ICPMS data, but their proportion would be less important for LA-ICPMS analyses. We show with a mixing model that an extremely small amount of equilibrium partial melt (typically 0.01–0.1%) trapped in minerals is sufficient to increase the
D
opx/cpx for HIE and LREE by a factor of 5–20 and the
D
ol/cpx by two or three orders of magnitude. Similar effects may be produced by sub-percent amounts of HIE-rich fluids of solid microphases. Such very small volumes of inclusions may pass unnoticed during mineral handpicking and LA-ICPMS analysis. Hence,
D
xt/cpx for HIE and LREE should be considered cautiously when mineral analyses are used to constrain melt processes and mantle composition. Mass balance calculations were performed for a nominally anhydrous spinel harzburgite sample. Similar to previous studies, the mass balance indicates important discrepancies for HIE between peridotite composition reconstructed from mineral analyses (bulk and in situ) and whole rock composition. The major silicate minerals are the main repositories for REE, Zr and Hf (>75% of the whole rock budget), and also host ≥65% of Th and U. In contrast, more than 80% of the budget of Rb, Ba and Nb, and about 60% of Ta and Sr, is hosted by micro-components in grain boundaries (GBC) or trapped in minerals (inclusions). Alone, the GBC accounts for 50% of the budget of Nb and Ta. The inclusions are an important repository for Rb (39%), Nb (40%) and Sr (49%). The GBC and inclusion repositories display very similar trace element signatures, suggesting that they were once a single repository (<1 wt%) now re-distributed in different textural components. This repository could be a combination of hydrous phases and/or Ti oxides, and/or melt/fluid inclusions of mantle origin.