Mapping olivine composition in the lithospheric mantle

The major-element composition of a grain of peridotitic Cr-pyrope garnet, coupled with a temperature estimate (T-Ni) derived from its Ni content, can be used to calculate the forsterite (Fo) content (100Mg/(Mg+Fe)) of the coexisting olivine, using an inversion of the olivine-garnet Mg-Fe exchange geothermometer. This calculation reproduces Fo(oliv) in garnet peridotite xenoliths within +/- 0.5%. Because the T-Ni of garnet xenocrysts in volcanic rocks can be projected to a paleogeotherm to derive a depth estimate, the application of the technique applied to suites of such xenocrysts provides a means of mapping the average Fo(oliv) of the mantle as a function of depth. Application of the technique to garnet concentrates from kimberlites and other volcanic rocks from Siberia, Australia, Canada and South Africa shows a decrease in mean Fo(oliv). with depth in most cratonic lithospheric sections, Near the base of most sections, Fo(oliv) approaches the value (Fo(89-90)) expected for asthenospheric mantle. The increase in Fo(oliv) toward the top of the sections is consistent with the higher degrees of partial melting at shallow depth predicted by models of melting during adiabatic decompression. However, trace-element data on the garnets suggest that these trends probably reflect more extensive melt-related metasomatism toward the base of the lithosphere. Lithospheric sections from some localities known to host lower-mantle diamonds (Slave Craton, South Australia) show that the lithospheric mantle consists of at least two distinct layers, which may reflect diapiric addition to the lower lithosphere. In the upper parts of these sections, the mean Fo(oliv) of the lithospheric mantle decreases from Archean (Fo(93-94)) through Proterozoic to Phanerozoic (Fo(90-91.5)), reflecting a secular decrease in the average degree of melt extraction from newly created lithospheric mantle. These variations in Fo(oliv) significantly affect lithosphere density and should be considered in modelling gravity and seismic data. (C) 2000 Elsevier Science B.V. All rights reserved.