Research Article

Deep Mantle Cycling of Oceanic Crust: Evidence from Diamonds and Their Mineral Inclusions

See allHide authors and affiliations

Science  07 Oct 2011:
Vol. 334, Issue 6052, pp. 54-57
DOI: 10.1126/science.1209300


A primary consequence of plate tectonics is that basaltic oceanic crust subducts with lithospheric slabs into the mantle. Seismological studies extend this process to the lower mantle, and geochemical observations indicate return of oceanic crust to the upper mantle in plumes. There has been no direct petrologic evidence, however, of the return of subducted oceanic crustal components from the lower mantle. We analyzed superdeep diamonds from Juina-5 kimberlite, Brazil, which host inclusions with compositions comprising the entire phase assemblage expected to crystallize from basalt under lower-mantle conditions. The inclusion mineralogies require exhumation from the lower to upper mantle. Because the diamond hosts have carbon isotope signatures consistent with surface-derived carbon, we conclude that the deep carbon cycle extends into the lower mantle.

Diamonds and the mineral inclusions that they trap during growth provide samples of materials from deep within Earth. On the basis of inclusion mineralogy, most diamonds sampled at the surface originated in continental lithospheric mantle at depths of <200 km (1). Several localities, however, yield rare “superdeep” diamonds with inclusion compositions that require a sublithospheric origin in the deep upper mantle and even the lower mantle (1, 2). Inclusions of majorite garnet that formed in the deep upper mantle (~200 to 500 km) commonly have compositions linking them to basaltic oceanic crust (18), and aluminous inclusions have been identified with compositions indicative of siliceous sediments (3). The diamonds that host these inclusions have carbon isotopic compositions that are atypical of normal mantle (δ13C ≈ –5‰), instead displaying a large isotopic range (~ −1 to –24‰) with a clear tendency toward isotopically “light” (< –10‰) compositions (13, 9). Although there is debate regarding the origin of light carbon in diamonds (10), a leading hypothesis is the subduction of the isotopically light organic carbon fraction of altered oceanic crust.

The rarest diamonds are those containing inclusions with compositions indicating an origin in the lower mantle (>660 km). Inclusions interpreted as representing the lower-mantle phases Mg-perovskite and Ca-perovskite have major element compositions that indicate an origin in mantle peridotite (2, 1113). No lower-mantle inclusions have previously been identified with major element compositions consistent with an origin in subducted basalt. Furthermore, the carbon isotopic compositions of diamonds with lower-mantle inclusions are typically all mantle-like (~ –4 to –6‰) (2), which suggests that surface-derived carbon may not survive into the lower mantle. Oceanic lithosphere clearly enters the lower mantle (14, 15), so the rarity of lower-mantle diamonds with inclusions of high-pressure phases that would occur in subducted basalt suggests that once oceanic crust enters the lower mantle, it usually remains there, possibly as a result of intrinsic high density and negative buoyancy (1618).

Here, we describe a suite of mineral inclusions in low-nitrogen (type IIa) diamonds from the Juina-5 kimberlite pipe in the Juina kimberlite field (92 to 95 million years old) (19) located in the Proterozoic Rio Negro–Juruena Mobile Belt southwest of the Amazon Craton, Brazil (20). Inclusions in sublithospheric diamonds commonly show mineralogical evidence of exsolution from originally homogeneous silicate phases into composite assemblages, and these are interpreted to have formed during ascent in the mantle unrelated to kimberlite eruption (25, 8, 12, 21, 22). Inclusion unmixing provides compelling evidence that some superdeep diamonds were transported upward by hundreds of kilometers in the upper mantle, presumably by upwelling of solid material (3, 4, 13). For example, the bulk compositions of composite garnet plus clinopyroxene inclusions in diamonds from the Juina region indicate a deep upper-mantle origin as majorite, with inclusion unmixing to garnet plus clinopyroxene occurring during transport to shallower levels beneath the lithosphere (24). Each of the inclusions presented here is composed of a multiphase mineral assemblage (Fig. 1). We interpret the composite silicate and oxide phases as exsolution products from originally homogeneous silicate phases that were trapped during diamond growth.

Fig. 1

Backscattered electron micrographs showing composite inclusions in diamonds from Juina-5. (A) An inclusion in diamond Ju5-20 composed of a mixture of spinel (Mg,Fe)Al2O4 (Sp) and nepheline NaAlSiO4 (Ne) (fig. S1 and table S5), together with a small sulfide (Sf) in one corner that we interpret as an originally distinct phase from the composite silicate; sulfide can participate in diamond crystallization reactions as a melt phase that is immiscible in silicate (33). (B) An inclusion in diamond Ju5-67 that is composed of phases with the compositions of spinel and a nepheline-kalsilite (Ka) phase, (Na,K)AlSiO4 (table S5). (C) An inclusion in diamond Ju5-89 containing spinel and a mixture of micrometer-sized Na-rich (Na) and K-rich (K) silicate regions, with a bulk composition similar to Ju5-67 (fig. S2 and table S5). (D) An inclusion in diamond Ju5-47 that consists of orthopyroxene (Opx), ulvospinel (Ulv), and olivine (Ol) (fig. S3 and table S5). (E) An inclusion in diamond Ju5-43 that consists of a complex mixture of orthopyroxene and a Ti-, Al-, and Fe-rich phase similar to tetragonal almandine pyrope phase (TAPP) (table S5). (F) An inclusion in diamond Ju5-104 composed of CaSiO3 plus micrometer-sized Ti-rich phases (e.g., CaTiO3) and a small sulfide (table S5).

Inclusion compositions and the depth of formation. The inclusions range from about 15 to 40 μm in their longest dimension. Individual phases within the composite inclusions were identified by Raman spectroscopy and spot electron microprobe analysis, and homogeneous bulk inclusion compositions were obtained using wide-beam microprobe analysis (Fig. 1, Table 1, and tables S1 to S5) (20). There is close correspondence between the inclusion bulk compositions and the individual phases that coexist in experiments on basaltic compositions under lower-mantle conditions. We ascribe the composite inclusions to the following lower-mantle phases: (i) “calcium ferrite” (CF) phase (Ju5-20); (ii) “new aluminum silicate” (NAL) phase, also known as the hexagonal phase (Ju5-67 and Ju5-89); (iii) Al-, Ti-, and Fe-rich Mg-perovskite (Ju5-43 and Ju5-47); and (iv) Ti-rich Ca-perovskite (Ju5-104). The CF and NAL phases, as well as Al-, Ti-, and Fe-rich Mg-perovskite compositions, have not been identified before as inclusions in diamond, but have previously been observed in experiments on basaltic bulk compositions at lower-mantle pressures and temperatures (1618). The sums of the cations and cation fractions in the bulk inclusion compositions closely correspond to the ideal stoichiometries of the experimentally synthesized phases (Table 1 and tables S1 to S4). It is unlikely that randomly trapped mineral aggregates at upper-mantle pressures could, by coincidence, have the correct stoichiometry of all four lower-mantle phases.

Each of the Juina-5 composite inclusions has a bulk composition that can be linked to a specific phase that would crystallize in a basaltic composition only in the lower mantle (Fig. 2). The composite inclusion in diamond Ju5-20 would form a homogeneous CF phase, and inclusions Ju5-67 and Ju5-89 would form the NAL phase at lower-mantle conditions (Fig. 2A). Although these two phases are compositionally similar to each other, a distinguishing feature is the considerable potassium solubility in experimental NAL phases, whereas experimental CF phases are potassium-free (Table 1 and tables S1 and S2). The bulk compositions of the composite inclusions in diamonds Ju5-43 and Ju5-47 are very similar to Mg-perovskites produced in experiments on basaltic compositions (Fig. 2B, Table 1, and table S3). They are much richer in Ti, Al, and Fe than is Mg-perovskite that forms in experiments on mantle peridotite (23) or in previously reported Mg-perovskite inclusions in diamonds (2, 11). The composite inclusion in diamond Ju5-104 has the bulk composition and stoichiometry of a CaSiO3 phase, but with a moderate Ti component [~3 weight % (wt %) TiO2] not previously observed in other CaSiO3 inclusions in sublithospheric diamonds. The inclusion composition is a close match to Ca-perovskite that coexists with Mg-perovskite in experiments on basaltic compositions (Fig. 2C, Table 1, and table S4).

Fig. 2

Portions of three ternary diagrams (cation fraction) illustrating the compositional correspondence between bulk composite mineral inclusions in diamonds from Juina-5 and phases produced in experiments at lower-mantle pressures. (A) Shaded fields represent CF and NAL phases synthesized in experiments on basaltic bulk compositions (tables S1 and S2). (B) Shaded fields represent Mg-perovskite phases synthesized in experiments on basaltic (table S3) and peridotitic bulk compositions (23). (C) The shaded field represents the compositions of Ca-perovskites that coexist with Mg-perovskite, as synthesized in experiments on basaltic bulk compositions (table S4). In (A) and (B), Fe3+ and Fe2+ are calculated from mineral formulae to satisfy site occupancy constraints (tables S1 to S3).

Experimentally estimated changes in mineralogy with depth for typical mid-ocean ridge basalts (17, 18) indicate that, as a group, the inclusions constitute a phase assemblage that coexists in basaltic compositions at depths between about 700 and 1400 km (Fig. 3A); an SiO2 inclusion has been identified in our Juina-5 collection as well, which likely originated as stishovite. We suggest that the inclusions must have originated when diamond-forming fluids incorporated basaltic components from oceanic lithosphere subducted into the lower mantle. To trap these specific inclusion compositions as homogeneous phases, the diamonds must have grown in the upper part of the lower mantle, and cannot simply be diamonds derived from shallower depths in the upper mantle but subducted into the lower mantle. For example, there are no known phases stable in the upper mantle with the bulk compositions of CF and NAL phases, and majorite garnets included in diamond in the deep upper mantle have far more calcium (e.g., 5 to 15 wt % CaO) (3, 6) and less titanium than observed in Mg-perovskite synthesized experimentally in basaltic compositions.

Fig. 3

(A) Estimated modal mineralogy in subducted basaltic oceanic crust as a function of depth in the mantle (17, 18). MgPv, Mg-perovskite; CaPv, Ca-perovskite; CF, CF phase; NAL, NAL phase; St, stishovite; Gt, garnet; Cpx, clinopyroxene. The inclusion mineralogy in diamonds from Juina-5, including MgPv, CaPv, CF phase, NAL phase, and stishovite, is stable at depths of ~700 to 1400 km in the lower mantle. (B) A schematic model for diamond formation and ascent beneath the Brazilian lithosphere. We suggest that the diamonds and inclusions initially formed from subducted oceanic crustal components in the upper part of the lower mantle and were transported in an upwelling plume to the upper mantle, where they unmixed into composite inclusions according to lower-pressure phase relations.

Diamond isotopic signatures and the origin of carbon. If the above hypothesis is correct, then the carbon from which the diamonds formed may have been deposited originally within oceanic crust at the seafloor. We measured the carbon isotopic composition (δ13C values) of the Juina-5 diamonds and found a range from about –1 to –24‰, with four of the six diamonds having values less than –15‰ (Fig. 4 and Table 1). These “light” isotopic values possibly indicate a recycled organic source of carbon (10). In contrast, all previously analyzed diamonds hosting ultramafic inclusions of lower-mantle origin have heavier, typical “mantle” carbon isotopic compositions around –5‰ (2).

Fig. 4

Carbon isotopic compositions of diamonds in this study compared to some possible carbon sources. White rectangles represent the range observed in each diamond on the basis of multiple spot measurements in different growth zones revealed by cathodoluminescence images (20). The isotopic compositions of several possible carbon sources are drawn schematically on the basis of ranges given in (10). Organic carbon denotes either biogenic or abiogenic noncarbonate carbon originating in surface or near-surface environments; mantle carbon denotes a carbon component that is typical of primitive ultramafic mantle rocks and peridotitic lithospheric diamonds; carbonate denotes surface-derived carbon deposited as carbonate from seawater.

The origin of light carbon isotopic values (< ~ –10‰) in mantle-derived samples is a matter of ongoing debate, with plausible explanations including intramantle isotopic fractionation, a primordial carbon reservoir with a light component, and subducted organic carbon (10, 24). Rayleigh fractionation in an open system involving phase separation can generate considerable isotopic fractionation and may account for much of the variation in δ13C in the range of –8 to –2‰ seen in lithospheric diamonds (10). However, it has yet to be demonstrated that this is a viable process for producing the isotopically light signatures (~ –10 to –25‰) commonly associated with sublithospheric diamonds of basaltic affinity, and an explanation would be needed for the correspondence of isotopically light carbon signatures with specific inclusion mineralogies. A primordial light carbon reservoir also appears unlikely, as the light component must have survived vigorous mantle mixing in an early magma ocean and billions of years of solid-state mantle convection to appear in the composition of these diamonds (24). Juina superdeep diamonds are likely only about 100 million years old, on the basis of a dated sublithospheric inclusion from the Collier-4 kimberlite 30 km to the north of Juina-5 (3), and so these young diamonds are very unlikely to have formed from an ancient, isotopically light primordial carbon component.

In contrast, a burgeoning body of evidence supports a subducted carbon source for many superdeep diamonds. The Juina-5 composite inclusions and many other inclusions in sublithospheric diamonds require an origin involving oceanic crust and sediments, and these commonly have light carbon isotopic compositions (2, 3, 6, 7, 9). Recent measurements of the carbon isotopic compositions in altered oceanic crust as deep as 2 km beneath the seafloor indicate mixing between an organic component (δ13C ≈ –27‰) and a carbonate component (δ13C ≈ 0‰) (25). As a group, sublithospheric diamonds with inclusions showing affinity with subducted oceanic crustal materials have carbon isotopic compositions that effectively span the entire isotopic range measured in altered oceanic crust. Our results suggest that subducted organic carbon can retain its isotopic signature even into the lower mantle. Experimental data indicate that subducted carbon, regardless of its original form in carbonates or organic compounds, can become fixed as either elemental carbon (graphite or diamond) or carbonate at high pressures in oceanic crust, depending on the redox state (26). Because the inclusions require that the diamonds grew in the lower mantle, we suggest that carbon was transported as carbonate, some of which would have been isotopically light, having originated as organic carbon (Fig. 4).

Implications for a deep carbon cycle. The diamonds and their inclusions may have grown when subducted lithosphere entered the shallow lower mantle and stagnated because of density inversion and increased mantle viscosity (14, 27) (Fig. 3B). If heated to ambient mantle temperatures, carbonated basaltic lithologies form carbonated melts, which can then be reduced to diamond during reactions with surrounding mantle (8, 28). Our results also indicate that the diamonds were transported by convection from the lower to the upper mantle, where the originally homogeneous inclusions unmixed. For example, phase relations along the NaAlSiO4-MgAl2O4 boundary (29) indicate that the bulk composition of inclusion Ju5-20 would yield the observed assemblage of nepheline plus spinel (Fig. 1A and fig. S1B) at depths of ~150 km; other inclusions in diamonds from the Juina region (3, 4, 8) also suggest equilibration near the base of the Brazilian lithosphere (~150 to 200 km). Thus, the diamonds record a history of upward transport on the order of 500 to 1000 km or more before being sampled by a Cretaceous kimberlite and brought to the surface.

On the basis of seismological and petrological evidence, previous workers have argued for a mantle plume beneath Brazil during the Cretaceous (30, 31). Furthermore, paleo-plate reconstructions show that the Juina region of Brazil was located at the margin of the African large low shear velocity provinces during the Cretaceous, which may be indicative of the presence of deep mantle plumes (32). We suggest that some portion of stagnated subducted lithosphere in which the diamonds grew was transported from the lower mantle to the base of the Brazilian lithosphere in a rising mantle plume (Fig. 3B). The Juina-5 diamonds and their inclusions provide compelling evidence for deep cycling of oceanic crust and surface carbon into the lower mantle and, ultimately, exhumation back to the upper mantle and Earth’s surface.

Table 1

Summary of the bulk major element chemistry of Juina-5 mineral inclusions and phases produced in high-PT experiments on basaltic compositions (cation fraction per formula unit), with carbon isotopic compositions of host diamonds. Isotopic values represent range observed in three or more analyses in different growth zones, as revealed in CL images. See tables S1 to S4 for experimental data and references.

View this table:

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S3

Tables S1 to S5

References (3445)

References and Notes

  1. See supporting material on Science Online.
  2. Acknowledgments: We thank I. Buisman and S. Kearns for assisting in the collection of electron microprobe data and L. Gobbo on behalf of Rio Tinto for providing samples. Supported by UK Natural Environment Research Council grant NE/H011242/1 (M.J.W.) and NSF grant EAR-1049992 (S.B.S. and J.W.). See (20) for additional compositional data on inclusion phases, experimental run products, and Raman spectroscopy.
View Abstract

Stay Connected to Science

Navigate This Article