Discovery of a heavy silicon isotope mantle reservoir

Abstract

Silicon cycling between Earth's reservoirs provides critical insights into how the Earth operates. While average crust and the bulk silicate Earth (BSE) share similar silicon isotope (δ³⁰Si) compositions, some mantle-derived magmas exhibit lower δ³⁰Si values than the BSE, implying the existence of an unidentified mantle reservoir with complementary higher δ³⁰Si values. We present silicon isotope data from Cenozoic lamproites and their hosted mantle pyroxenite xenoliths from the Himalaya-Tibet orogen. These mantle-derived rocks have higher δ³⁰Si values than the BSE, which resulted from reaction between mantle peridotite and ³⁰Si-rich silicate melts from subducted Indian continental crust. Our results demonstrate that slab melting can produce high-δ³⁰Si melts and complementary low-δ³⁰Si residues. These products are unevenly distributed in the mantle, with high-δ³⁰Si melts stored as metasomatic veins in the lithospheric mantle while low-δ³⁰Si residues are recycled into the deep mantle. This study provides evidence that mantle metasomatism by high-δ³⁰Si slab melts creates heavy silicon reservoirs in the lithospheric mantle above continental slabs or cool mantle wedges above oceanic slabs.

Keywords: Si isotope; lamproites; mantle heterogeneity; silicate melt metasomatism; slab melts.

Figure 1.

Global compilation of silicon isotopic compositions for mantle rocks and their derived magmas. The shaded box represents the estimate for the Si isotopic composition of the BSE [22]. The bulk composition of continental crust (red dashed line box) is derived from the silicon isotopic compositions for the upper, middle and lower crust as estimated by Savage et al. [39] and Savage et al. [40], and subsequently mixed in the proportions determined by the global compilation of Rudnick and Gao [71]: 31.7% for the upper crust, 29.6% for the middle crust and 38.8% for the lower crust. Notably, the silicon isotopic composition of the bulk continental crust closely resembles that of the BSE. Data sources: ultramafic xenoliths [24,72], mid-ocean ridge basalt [24,29,72], komatiite [23,30], island arc basalts [24], continental arc [25], ocean island basalts [26,72].

Figure 2.

δ³⁰Si vs SiO₂ for the lamproites and Himalayan metamorphic rocks analyzed in this study. Uncertainties are expressed as two standard errors (2SE). The igneous array (solid line) and its 2SE envelope (dashed lines), as defined by Savage et al. [18], are shown for comparison. Notably, the δ³⁰Si values of the lamproites are consistently higher than the predicted values of the terrestrial igneous array at a given SiO₂ content. For comparison, published data for terrestrial igneous bulk rock samples are also presented. The field of granite data are from Savage et al. [20] and Gajos et al. [43]; shale data are from Savage et al. [40]; clay mineral δ³⁰Si data are from Opfergelt et al. [16]; other data sources are the same as in Fig. 1.

Figure 3.

Si isotope compositions of lamproites and associated mantle pyroxenite xenoliths. (a) In situ Si isotope analysis of olivine phenocrysts from lamproites exhibits the most distinctive high δ³⁰Si signatures yet analyzed. Data sources of olivine Si isotope values are from Georg et al. [2], Savage et al. [18] and Armytage et al. [29]. (b) Opx-rich mantle websterite xenoliths from the lamproites have higher δ³⁰Si values than the BSE, whereas Cpx-rich xenoliths show values similar to those of the BSE. (c) Modelling results illustrating the effects of crustal assimilation and fractional crystallization on the lamproites. The wallrock composition (upper continental crust) is defined by SiO₂ = 66.6 wt.%, MgO = 2.48 wt.% with δ³⁰Si = −0.25 ± 0.16‰, representing the end-member compositions for crustal assimilation (from Savage et al. [39] and Rudnick and Gao [71]). The yellow shaded region delineates the range of magma δ³⁰Si values when the uncertainties (±0.16‰) of the wallrock composition are considered. For comparison, data from Hekla volcano in Iceland [18] are included, where significant Si isotope fractionation was observed during magma differentiation. To assess the impact of different silicate melt compositions, we model the δ³⁰Si/²⁸Si_melt trajectories for basalt (red curve) and basanite (blue curve), utilizing the respective silicon β-factors for each melt type [44]. (d) δ³⁰Si values of hand-picked mineral separates from the lamproites. For comparison, silicon isotope data from mineral separates of the Skaergaard Intrusion [18] and Cameroon line spinel lherzolites [2] are also shown.

Figure 4.

Si isotope fractionation during slab melting and its role in forming a heavy silicon isotope mantle reservoir. (a) Variations of Δ^30/28Si_{melt–protolith} (the difference in δ³⁰Si between the partial melt and the protolith) with melting temperature. The data clearly show that, regardless of the melting temperature, the degree of silicon isotope fractionation during metabasite melting is consistently greater than that during metasediment melting. The uncertainty on the mineral–melt Si isotope fractionation factor was taken as the source of uncertainty in the melting models. (b) Relationship between Δ^30/28Si_{melt–protolith} and the proportion of garnet + Cpx in the melting residues. The data suggest that the presence of these minerals (which is enriched in light Si isotopes) in the residual phases significantly influences silicon isotope fractionation, with higher proportions leading to greater fractionation. (c and d) Three end-member mixing models describing the effect of slab melts on Si isotopes of the mantle wedge. The solid curves represent the calculated mixing lines between the depleted MORB mantle (DMM), melts derived from the Indian continental upper crust, and melts derived from the Indian continental lower crust. The percentage values marked on the tick marks of these curves correspond to the proportion of melt originating from the Indian continental crust. The detailed calculations are outlined in the Supplementary data.