A high field strength element perspective on early lunar differentiation

Abstract

Lunar rocks are inferred to tap the different fossil cumulate layers formed during crystallisation of a lunar magma ocean (LMO). A coherent dataset, including Zr isotope data and high precision HFSE (W, Nb, Ta, Zr, Hf) and REE (Nd, Sm, Lu) data, all obtained by isotope dilution, can now provide new insights into the processes active during LMO crystallisation and during the petrogenesis of lunar magmas. Measured 92Zr and 91Zr abundances agree with the terrestrial value within 0.2 ε-units. Incompatible-trace-element enriched rocks from the Procellarum KREEP Terrane (PKT) display Nb/Ta and Zr/Hf above the bulk lunar value (ca. 17), and mare basalts display lower ratios, generally confirming the presence of complementary enriched and depleted mantle reservoirs on the Moon. The full compositional spectrum of lunar basalts, however, also requires interaction with ilmenite-rich layers in the lunar mantle. Notably, the high-Ti mare basalts analysed display the lowest Nb/Ta and Zr/Hf of all lunar rocks, and also higher Sm/Nd at similar Lu/Hf than low-Ti basalts. The high-Ti basalts also exhibit higher and strongly correlated Ta/W (up to 25) and Hf/W (up to 140), at similar W contents, which is difficult to reconcile with ortho- and clinopyroxene-controlled melting. Altogether, these patterns can be explained via assimilation of up to ca. 20% of ilmenite- and clinopyroxene-rich LMO cumulates by more depleted melts from the lower lunar mantle. Direct melting of ilmenite-rich cumulates or the possible presence of residual metals in the lunar mantle both cannot easily account for the observed Ta/W and Hf/W patterns. Cumulate assimilation is also a viable mechanism that can partially buffer the Lu/Hf of mare basalts at relatively low values while generating variable Sm/Nd. Thus, the dichotomy between low Lu/Hf of lunar basalts and high time integrated source Lu/Hf as inferred from Hf isotope compositions can potentially be explained. The proposed assimilation model also has important implications for the short-lived nuclide chronology of the Earth-Moon system. The new Hf/W and Ta/W data, together with a compilation of existing W-Th-U data for lunar rocks, indicate that the terrestrial and lunar mantles are indistinguishable in their Hf/W. Virtually identical εW and Hf/W in the terrestrial and lunar mantle suggest a strong link between final core-mantle equilibration on Earth and the Moon forming giant impact. Previously, linear arrays of lunar samples in 182W vs. Hf/W and 142Nd vs. Sm/Nd spaces have been interpreted as isochrons, arguing for LMO crystallisation as late as 250 Myrs after solar system formation. Based on the proposed assimilation model, the 182W and 142Nd in many lunar magmas can be shown to be decoupled from their ambient Hf/W and Sm/Nd source compositions. As a consequence, the 182W vs. Hf/W and 142Nd vs. Sm/Nd arrays would constitute mixing lines rather than isochrons. Hence, the lunar 182Hf– 182W and 146Sm– 142Nd data would be fully consistent with an “early” crystallisation age of the LMO, even as early as 50 Myrs after solar system formation when the Moon was probably formed.