Origins of the terrestrial Hf-Nd mantle array: Evidence from a combined geodynamical-geochemical approach

摘要

The formation and segregation of oceanic and continental crust from the mantle, and its return to the mantle via subduction and/or delamination, leads to the development of distinct geochemical reservoirs in the terrestrial mantle. Fundamental questions remain regarding the location, nature, and residence time of these reservoirs, as well as the respective roles of oceanic and continental crust in the development of the mantle's geochemical endmembers. The Lu-Hf and Sm-Nd isotope systems behave similarly in magmatic systems and together form the terrestrial mantle Hf-Nd isotopic array. Here we combine a geodynamic model of mantle convection with isotope and trace element (TE) geochemistry to investigate the evolution of the Hf-Nd mantle array. This study examines the sensitivity to: TE partition coefficients used in the formation of oceanic crust; density contrasts between subducting oceanic crust and the mantle; and the formation and recycling of continental crust. We show that the fractionation between the parent (Lu and Sm) and daughter (Hf and Nd) species needs to be higher than is indicated by partition coefficients determined from the present-day melting environment. This is consistent with the suggestion of deeper mantle melting earlier in Earth history and an increased role for residual garnet. Subduction and accumulation of dense oceanic crust produces a large mass of incompatible TE enriched material in the deep mantle. This deep mantle enrichment appears to play a more significant role than the extraction and recycling of continental crust in developing the Hf and Nd isotope and TE compositions of the mid-ocean ridge mantle source. The corollary of this result is that the formation of the continental crust plays a secondary role, contrary to the currently accepted paradigm. Nevertheless, the inclusion of continental crust formation and recycling produces a broader model mantle array, which better reproduces the spread in the natural data set. This model a