Trace element partitioning between amphibole, phlogopite, and basanite melt

We have measured amphibole-melt and phlogopite-melt partition coefficients (D) for 22 trace elements in experimentally crystallized natural basanites with the ion microprobe. The synthesized phases display an exceptional degree of homogeneity for both major and trace elements, as demonstrated by the ratio of the standard deviation to the mean counting statistics uncertainty of the measurements. In pargasitic hornblende, actinides are highly incompatible (D = 0.001), LILE and HFSE are mildly incompatible (D = 0.04 - 0.2and0.1 - 0.2, respectively), and REE partition coefficients vary from 0.05 to 0.6, with a maximum near Ho. Except for the LILE (D = 0.1 - 3.7), phlogopite partition coefficients are generally lower, especially the REE (D ≈ 0.01). The partitioning results are consistent with a model in which the variation in partition coefficient with ionic radius results from the crystal lattice strain induced by the size misfit of the substituting trace element. This result predicts a decrease in Young's Modulus (E) with increasing size of the cation sites in the crystal lattice, and E derived for the largest site in both amphibole and phlogopite agree well with experimentally determined bulk mineral values. The ability to model partitioning with an elastic strain model provides an important link between trace element partitioning and the macroscopic properties of minerals. Relative to an anhydrous peridotite, partial melting of an amphibole or phlogopite bearing peridotite will result in no Th-U fractionation, slight LILE depletions, and, aside from Ti, no significant HFSE depletions. Thus, barring the addition of any slab components besides H₂O, partial melting of hydrated peridotite is not a plausible explanation for any of the geochemical features commonly associated with subduction zone magmas.