Nd isotopic evolution of the upper mantle during the Precambrian: models, data and the uncertainty of both

This paper presents a dual approach to resolve the terrestrial evolution of the Sm-Nd system: a critically reviewed compilation of Precambrian Nd initial isotope ratios and, independently, a forward modelling effort of the Nd isotopic development of the upper mantle, The transport balance model used is identical to a previously published one which allows to reconcile calculated and observed parameters of the Th-U-Pb system. In order to express compositional and age heterogeneity, the continental crust is divided into four domains: upper and lower as well as younger and older. The mantle is represented by an upper and a lower reservoir. Model versions successful for Th-U-Pb and Sm-Nd produce a very narrow range of epsilon(Nd) development curves for the upper mantle. All these scenarios show very little recycling of continental crust into the upper mantle before 2-1.6 Ga ago. The data as well as the model show a steady increase of the epsilon(Nd)[T] value of the upper mantle from ca 3 Ga on to a present day value of + 10. In the Early Archean, a quasi constant epsilon(Nd)[T] Nd value is seen. The model predicts that the upper mantle was not significantly different from bulk silicate Earth (BSE) in this early period. The data, however are generally one epsilon(Nd) unit above the accepted Chondritic Uniform Reservoir (CHUR) evolution. Lunar rocks also show this offset and, in particular, appear to indicate its existence since >4.5 Ga ago. This implies that the initial BSE and bulk silicate Moon are apparently slightly different from the conventional initial CHUR. Possible explanations for this apparent offset are some uncertainty of the CHUR model, analytical artefacts and interlaboratory bias. Published deviations from the accepted CHUR values and the effects caused by different fractionation correction procedures are all in the same direction, and make an apparent offset of ca + 1 epsilon(Nd) unit of terrestrial and lunar data relative to CHUR plausible. Applying an apparent BSE evolution with an initial at one epsilon(Nd) unit above the CHUR model brings terrestrial modelling of the upper mantle and data into good alignment. As the model describes internally fully homogeneous reservoirs, a scatter of real data around the model development curves is expected as a result of imperfect internal mixing. A third order polynomial fit to epsilon(Nd) evolution curves generated by the model for the depleted upper mantle is given as a new reference for Nd model age calculations, via the relation: epsilon(Nd)[T](sample) = 0.164T(3) - 0.566T(2)-2.79T + 10.4. (C) 1998 Elsevier Science B.V.