Mineral-aqueous fluid partitioning of trace elements at 900°C and 2.0 GPa: Constraints on the trace element chemistry of mantle and deep crustal fluids

Abstract

To constrain the trace element composition of aqueous fluids in the deep crust and upper mantle, mineral-aqueous fluid partition coefficients ( D min/fluid) for U, Th, Pb, Nb, Ba, and Sr have been measured for clinopyroxene, garnet, amphibole, and olivine in experiments at 2.0 GPa and 900°C. Clinopyroxene-and garnet-fluid partition coefficients are similar for Nb (0.01-0.7) and Ba (∼10 −4-10 −5 ), whereas values of D cpx/fluid for Sr (0.5−4), Th (0.6–9), and Pb (0.04–0.09) are ∼1Ox (Th, Pb) to ∼1000x (Sr) higher than D garnet/fluid . At the same f O 2 (FMQ + 1), garnet-fluid partition coefficients for U are ∼10x higher than those for clinopyroxene. Amphibole-fluid partition coefficients are uniformly high (-I) for all elements studied, and, with the exception of Ba, interelement fractionations are similar to clinopyroxene. The olivine-fluid partition coefficient for Nb is similar to values measured for the other silicates, whereas D olivine/fluid for U, Th, Pb, Sr, and Ba are significantly lower. Clinopyroxene and garnet partition coefficients follow Henry's Law up to ∼300 ppm of either Ba, Pb, or Sr in the fluid. Both the major-element chemistry of clinopyroxene and fluid have some influence on partitioning, with the magnitude of these effects varying according to element type. Although clinopyroxene concentrations of Pb, Ba, and Sr were found to be homogeneous, core-to-rim decreases in wt% Al 2O 3 were found to correlate with reductions in the concentrations of Nb, U and Th, and hence D cpx/fluid. Both increases in solute content and the addition of NaCl to fluids lower the measured partition coefficients. A decrease in experiment f O 2 reduces D Th/D U for clinopyroxene, which is consistent with the compatibility of U 4+ relative to U 6+ in the clinopyroxene structure. Comparison of mineral/ fluid partition coefficients with mineral /basaltic melt values from the literature reveal notable distinctions in partitioning behavior for fluids vs. melts. Mineral-melt and mineral-fluid partitioning for elements such as Ba, Pb, and Sr are similar, but in contrast, U, Th, and Nb are more strongly partitioned into silicate melts than aqueous fluids. Such differences may provide a means of discerning the products of melt- vs. fluid-mediated metasomatism. Bulk eclogite- and lherzolite-aqueous fluid partition coefficients, calculated from mineral/aqueous fluid values, are used to illustrate how partitioning data can constrain (1) the trace element composition of fluids that may be a product of dehydration of basaltic oceanic crust and (2) the effect of the subarc mantle on trace element fractionation processes. The silicate assemblage produced during basalt dehydration (garnet + cpx ± amphibole) does not selectively deplete the coexisting fluid in Nb relative to the other elements studied, nor is Nb preferentially withdrawn from the fluid by passage through an amphibole lherzolite mantle. Results, therefore, reaffirm the notion that residual rutile is necessary to selectively deplete slab-derived fluids in high field strength elements. Calculations also indicate that fluids with excess [ 238U ] relative to [ 230Th] may be produced during dehydration of basaltic oceanic crust, and such excesses are retained or enhanced during transit through the mantle wedge, provided that mildly oxidizing conditions prevail. Slab-derived fluids can therefore produce the requisite low ratios of high-field-strength/large-ion-lithophile elements (such as Nb/Th) and [ 238U]/[ 230Th] > 1 in the source regions of island arc basalts by metasomatism of the mantle wedge. In addition to constraints on the composition of the fluid liberated during slab dehydration, our data allow us to estimate the trace element composition of the material returned to the deep mantle during subduction. Calculations indicate that, following dehydration, the U/Pb ratio in basaltic crust is increased and Rb/Sr is likely to be dramatically reduced. Subduction and prolonged aging of this material produces an isotopic reservoir with the characteristics of the HIMU component sampled by some oceanic island basalts.