It has long been known that the measured kappa (Th-232/U-238, Or kappa) Of mid-ocean ridge basalts (MORB) and, by inference, the upper mantle, is much lower than the time-integrated kappa recorded by Pb isotope ratios (kappa(Pb)). We examine models that can reconcile this kappa conundrum by in situ decay of U and Th in the upper mantle. Monte Carlo simulations reveal a restricted range of permissible 'in situ' paths of MORE mantle evolution. These solutions require a roughly constant Th-232/U-238 ratio during early Earth history, followed by a period of steadily decreasing Th-232/U-238 from the end of the Archean to the present. These model criteria make good geological sense in terms of post-Archean recycling of crustal uranium back into the mantle, Preferential recycling of uranium, relative to thorium, can result from the high aqueous mobility of uranium in the oxidising environment at the Earth's surface. Soluble uranium is transported from continents to the altered oceanic crust and ultimately, by subduction, back into the mantle. In contrast, insoluble thorium remains in the weathered continental residue. This process is only likely to have become important after the marked increase in atmospheric oxygen fugacity at similar to 2.2 Ga which led to a change in the predominant surface oxidation state of uranium. The uranium fluxes required in this Post-Archean Uranium Recycling (PURE) model are compatible with estimates derived from present-day fluxes of 'excess' continental uranium returned to the mantle by subduction, integrated over some 2 Ga. Further modelling of Earth evolution in the context of differentiation into crust, depleted mantle and recycled plume reservoirs demonstrates the viability of this scenario in explaining modem-day lead isotopic signatures of both MORE and 'HIMU' ocean island basalts. We emphasise that resolution of the kappa conundrum does not require a steady state upper mantle with its lead isotope ratios buffered by entrainment of material from another, deeper reservoir. (C) 1999 Elsevier Science B.V. All rights reserved.
197YATimes Cited:33Cited References Count:60