|Home | About | Journals | Submit | Contact Us | Français|
Precise dating of diamond growth is required to understand the interior workings of the early Earth and the deep carbon cycle. Here we report Sm-Nd isotope data from 26 individual garnet inclusions from 26 harzburgitic diamonds from Venetia, South Africa. Garnet inclusions and host diamonds comprise two compositional suites formed under markedly different conditions and define two isochrons, one Archaean (2.95Ga) and one Proterozoic (1.15Ga). The Archaean diamond suite formed from relatively cool fluid-dominated metasomatism during rifting of the southern shelf of the Zimbabwe Craton. The 1.8 billion years younger Proterozoic diamond suite formed by melt-dominated metasomatism related to the 1.1Ga Umkondo Large Igneous Province. The results demonstrate that resolving the time of diamond growth events requires dating of individual inclusions, and that there was a major change in the magmatic processes responsible for harzburgitic diamond formation beneath Venetia from the Archaean to the Proterozoic.
The apparent antiquity of many diamond suites means that they represent a window to both the Earth’s early geological record and deep carbon cycle1. Precise dating of diamond growth events is a prerequisite for such studies but represents a challenge owing to the complex growth history of many gem quality diamonds2. Inclusion-bearing diamonds derived from the sub-continental lithospheric mantle (SCLM), preserve evidence of tectono-thermal events such as continental assembly and mantle melting1, 3–5. Diamonds crystallise from metasomatic reactions involving C-H-O-S-rich supercritical-fluids and/or silicate melts and on rare occasions include minerals that carry the imprint of the environment of growth2. Diamonds, transported from the SCLM to the surface as xenocrysts in kimberlitic or related magma types, protect the mineral inclusions from secondary processes such as later mantle metasomatism and re-equilibration with the host magma6. Radiogenic isotope studies of inclusions in combination with compositional data can constrain the conditions and timing of diamond growth, and provide fundamental information about the tectono-magmatic processes that led to the formation and modification of the lithospheric keels that underlie the oldest parts of the Earth’s continents1. Inclusions in diamonds derived from the sub-continental lithosphere are typically subdivided in three principal suites that characterise their mantle source rocks; peridotitic, eclogitic, and websteritic (see ref. 2 for an extensive review). The peridotitic suite is further subdivided into harzburgitic, lherzolitic, and wehrlitic parageneses based on Ca and Cr contents of garnet inclusions. Thus far, with one exception7, the low trace element concentrations of peridotitic inclusions in diamond and their small size has necessitated pooling tens to hundreds of inclusions to obtain Sm-Nd isochron ages8–13. In consequence, fundamental concerns have been raised about common inclusion parentage and the validity of these ages as records of actual diamond growth events14–16.
A previous study of inclusions from the Venetia diamond mine located within the Limpopo Mobile Belt in South Africa, examined 400 garnet inclusions of harzburgitic composition of which 140 were divided into four compositional groups for Nd and Sr isotope analysis13. Three of the pooled groups defined a nominal Sm-Nd isochron age of 2.30±0.04Ga (2σ) with an unradiogenic initial ratio (εNd=−8). The authors recognised that the results potentially recorded mixing of different components and presented a model wherein Venetia harzburgitic diamonds crystallised at ca. 2Ga associated with modification of a>3Ga Archaean harzburgitic SCLM by Bushveld-related magmas. The implication of this interpretation is that harzburgitic garnets of highly variable composition were formed during a single diamond-forming event in the mid-Proterozoic.
To provide constraints on how the tectono-magmatic conditions responsible for harzburgitic diamond formation may have evolved over time, and assess the possible temporal evolution of the Earth’s carbon cycle, we present coupled major and trace element and Sm-Nd isotope data for 26 individual garnet inclusions extracted from 26 individual peridotitic diamonds from Venetia, which were also measured for their carbon isotope compositions. The data allow the unravelling of the mixed signature recorded in previous pooled inclusion data, and produce new accurate ages that date regional, but contrasting, styles of magmatism that affected the mantle beneath the Limpopo Mobile Belt in the Archaean and Proterozoic. The data provide a tantalising glimpse of how processes associated with diamond formation may have changed over time.
The relationship between diamond and its inclusions is potentially key in interpreting absolute age information obtained from the inclusions. There has been significant recent debate as to whether inclusions are formed simultaneously with diamond (syngenesis), or are pre-existing objects incorporated during diamond growth (protogenesis)17. In addition, the term ‘synchronous’ has been proposed, describing protogenetic inclusions that owing to chemical equilibrium with the diamond-forming medium record the time of diamond growth16.
All inclusions studied in this work show cubo-octahedral diamond imposed morphology and some show diamond inherited surface morphology (Fig. 1). The occurrence of imposed cubo-octahedral morphology on monoclinic (pyroxenes), hexagonal (monosulphides), orthorhombic (olivines), and cubic (chromites or garnets) inclusions (“negative crystals”) has conventionally been taken as evidence that the two phases interacted during growth with the strong bonding of diamond imposing the cubo-octahedral morphology. Hence, inclusions with imposed morphology are considered to have formed syngenetically with diamond18, 19. Moreover, epitaxial relationships between inclusions and diamond hosts have also been thought as strong evidence of the control by diamond on the crystallisation of the inclusion18–21. Recent studies from Siberian diamonds reported that epitaxy played both a strong22, 23 and limited control24 on inclusion formation. Theoretical considerations of adhesion energy and interface energy, however, suggest that there is no energetic benefit for olivine-diamond interfaces to form an epitaxial relationship25. An X-ray topographic study of an olivine inclusion-bearing diamond reported an absence of volume distortion in diamond around the inclusions17. In addition, the surface morphology of olivine inclusions had stepped morphology. Both observations are consistent with growth or dissolution processes contributing to the final inclusion morphology. The recent report of epitaxy between clinopyroxene inside and outside a diamond in a lherzolite xenolith adds to the discussion16. These authors propose that the use of the term syngenetic may be inappropriate to describe inclusions. They argue that the term ‘synchronous’ is more applicable to explain isotopic data when minerals were formed prior to diamond growth (protogenetic), but owing to chemical equilibrium with the diamond-forming fluid minerals will record the time of encapsulation in the diamond16. Experimental studies have produced syngenetic growth of inclusions from hydrous fluids containing salts, silicates, and carbonates under upper mantle conditions. Such experiments provide unequivocal evidence that diamonds and silicates can form syngenetically, producing mineral assemblages comparable to those found in inclusion suites in natural diamonds26.
Taylor et al.27, however, argued that the sinusoidal rare earth element (REE) patterns that characterise harzburgitic garnet inclusions in diamonds worldwide are indicative of a complex genetic history for the garnets prior to diamond encapsulation. The above authors argued that the sinusoidal REE pattern with low absolute heavy rare earth element (HREE) implies an extensive melt depletion phase in the absence of residual garnet was followed by a metasomatic event that involved a highly fractionated light rare earth element (LREE) component, suggesting that the inclusion was protogenetic27. An equally plausible argument, however, is that the diamond-forming fluid interacted with the melt-depleted host peridotite to produce an environment with a sinusoidal REE pattern.
The interpretation of the above data remains a matter of debate. It appears definite that the external regions of inclusions with imposed morphology are syngenetic with diamond growth but that does not unambiguously prove that the bulk of the inclusions formed syngenetically17. Irrespective of the origin of the inclusions, a key aspect in dating a mineral inclusion is the extent to which the inclusion reached chemical equilibration with the diamond forming media, that is, do protogenetic grains record synchronous age information. The closure temperature of the Sm-Nd system in garnets in the crust and mantle varies depending on elemental diffusivity, mineral size, composition and cooling rate but is considered to be typically between 750 and 900°C28, 29. In the case of a garnet in the mantle in contact with a metasomatic fluid/melt precipitating diamond, free diffusion of REE is thus expected at typical mantle temperatures (1000–1400°C). Under such conditions, the time for a garnet of protogenetic origin to reach full Nd isotope equilibration with the fluid/melt will depend on the effective garnet grain size and elemental diffusivity. An upper limit of equilibration time can be determined by assuming a defect-free gem quality mineral under anhydrous conditions. Based on the typical size of our garnet inclusions (100μm radius), full chemical equilibration of REE (Sm, Dy, Yb) with a diamond-forming melt/fluid would occur within less than 0.6kyr at 1400°C (diffusion data from ref. 29). At a lower temperature of diamond formation (1000°C) and again assuming a defect-free 100µm radius garnet, it is possible that full chemical equilibration with the diamond forming melt/fluid could take up to 250kyr. Hence under such circumstances, instantaneous diamond formation could potentially entrain minerals partially recording a protogenetic age. Given the diamond imposed morphology and growth features seen on the crystal faces of analysed inclusions (Fig. 1), we have, however, evidence that at least the outer portion of the garnet inclusions grew simultaneously with the diamond from the same fluid. Interaction and equilibration during dissolution/precipitation of garnet with the fluid is instantaneous, resulting in a syngenetic relationship30. Furthermore, experimental studies have shown that under hydrous mantle conditions, typical for diamond growth, garnet becomes more susceptible to dislocation creep, likely resulting in a marked increase in diffusion rates31. Hence, we conclude that retaining a protogenetic age, even in the mineral core, seems highly unlikely.
All inclusions analysed are harzburgitic garnets, 0.1– 0.4mm in size, with weights between 20 and 230μg. Major element compositions of the inclusions were determined on unpolished, but flat and levelled crystal surfaces before trace element and isotopic analysis (see Methods section). CaO and Cr2O3 contents are variable, between 0.5 and 4.3wt.%, and 5.1 and 17.0wt.%, respectively (Supplementary Fig. 1 and Supplementary Data 1). The subcalcic compositions in equilibrium with a harzburgitic protolith are in agreement with previous work32. Garnet compositions record a broad negative correlation between CaO and Mg# that suggests that more melt-depleted garnets have lower CaO (Supplementary Fig. 1). The 143Nd/144Nd ratios (0.5097-0.5233) and Nd (0.9-26ppm) and Sm (0.01–5.8ppm) concentrations, used to calculate depleted mantle model ages (T DM), are presented in Supplementary Table 2. Within a garnet harzburgite assemblage, the vast majority of the REE will be contained in garnet (>90%), hence we can expect low Ca garnets to record geologically meaningful Nd model ages. This situation is in marked contrast to lherzolitic or eclogitic inclusion assemblages where element partitioning between clinopyroxene, garnet and the precipitating fluid/melt, will lead to fractionation of Sm/Nd ratios and most probably mineral model ages that can only be used for indicative purposes. Hence, the two populations of model ages derived from the harzburgitic garnets that can be distinguished in a probability density diagram are considered significant (Fig. 2a). An average T DM age of 3.0±0.1Ga is defined by nine samples and an age of 1.1±0.2Ga (1SD) is recorded by 10 samples. The 3.0Ga population generally has CaO contents lower than 2.5wt.%, whereas the 1.1Ga age population has CaO contents mostly higher than 2.5wt.%. We use 2.5wt.% CaO as a threshold value to further define the two groups, now with eight samples each (Fig. 2b). Significantly, the carbon isotope compositions of the host diamonds for the two age populations are also distinct; the older diamonds are depleted in 13C with δ 13C<−5 ‰ (−8.9 to −5.5 ‰) and the 1.1Ga age group has δ 13C>−5 ‰ (−4.9 to −3.6 ‰, Fig. 2c, Supplementary Table 1). Chondrite-normalised REE patterns of garnets from the old and young groups also have distinct geometries. The older group of garnets has strong “sinusoidal” patterns with generally higher LREE abundances that increase from LaN to NdN, has low Sm/Nd and concave down MREE-HREE. In contrast, the 1.1Ga group has lower LREE abundances that increase from LaN to SmN, has higher Sm/Nd and higher MREE-HREE abundances that define less-fractionated REE patterns. The minimum in the HREE is at Dy for the 1.1Ga group and at Er for the 3.0Ga group (Fig. 3).
Strikingly, the two inclusion populations define separate isochrons yielding ages of 2.95±0.07Ga and 1.15±0.11Ga. Both isochrons have initial ratios within error or close to chondritic mantle (Fig. 4, εNd of −2.8 and+3.5, respectively). No clear relationship that would indicate simple two-component mixing exists between 143Nd/144Nd and reciprocal Nd (Supplementary Fig. 1d). The realistic initial ratios of the isochrons and the good agreement of the isochron ages with the depleted mantle model ages support the age significance of the isochrons. Seven additional inclusions give T DMs that are both older (4.1Ga) and younger (1.3–2.3Ga) than the 3.0Ga age group (Fig. 2a). Even though some of these samples fall onto or close to the two well-defined isochrons (Fig. 4), their chemical characteristics do not justify including them into the isochron populations. For example, two samples with T DM of 1.3Ga (V306 and V471) plot on the isochron of 1.15Ga at low 147Sm/144Sm (Supplementary Table 2), however, their similar trace element patterns are distinct from the 1.15Ga sample population (steep HREE, see Supplementary Data 1 and Supplementary Fig. 2), suggesting growth from a different growth medium. In addition, the two samples with T DM of 4.1Ga have both low and high CaO (0.7; 2.7 wt%) but have relatively flat HREE patterns. They plot just below the 2.95Ga isochron (Fig. 4), but again there is no justification for including them in the isochron.
Our data establish that it is vital to utilise single inclusions to obtain accurate age constraints on the time of diamond growth and that pooling of inclusions may mix populations of different age, resulting in ages with diminished geological meaning. The data reported by Richardson et al.13 on combined harzburgitic garnet inclusions (32, 33, 34, and 42 specimens in each group) show a significantly smaller range in 147Sm/144Nd and 143Nd/144Nd compared with the individual inclusions, demonstrating that the pooling masked the original heterogeneity. More than 25% of the individual inclusions analysed here have higher Sm/Nd and 143Nd/144Nd than the composite inclusion data, but because of low Nd contents their presence in a pooled population is obscured by inclusions with higher Nd contents. The average of the T DM’s for individual inclusions (Supplementary Table 2) is identical to that obtained from the isochron of pooled garnet inclusions13. Richardson et al. concluded that harzburgitic diamond growth beneath Venetia was related to regional scale ‘basaltic’ magmatism of the Bushveld event at ~2Ga. The younger diamond growth event at 1.1Ga is characterised by garnets with relatively high Ca and Sm/Nd and hence is compatible with formation associated with large-scale regional basaltic magmatism, but significantly younger than inferred by the pooled inclusion data. The involvement of Archaean lithospheric mantle was inferred from the pooled inclusion data, however, the second harzburgitic diamond formation event at 2.95Ga that produced a population of garnets with low Ca and low Sm/Nd remained unrecognised. These conclusions mean that Proterozoic ages (~2Ga) obtained from pooled peridotitic garnet inclusions from Premier (Cullinan), SA12 and Udachnaya, RU33 may also be coincidental. Based on Re-Os dates from sulphide inclusions at Udachnaya, it has already been shown that old and young peridotitic diamond populations are recovered from these mines34, further stressing the need to re-examine published diamond ages obtained from pooled diamonds.
Below, we consider the potential coupling between diamond growth within the lithosphere and large scale magmatic events expressed by the presence of crustal intrusions in the area with similar ages as the diamonds. A correlation is made between diamond formation and large tectono-magmatic events due to the large amount of carbon recovered in the form of diamond from the Venetia mine. The Venetia kimberlite cluster comprises a surface area of only 28ha but since its opening in 1992 has produced in the order of 100M carats and comparable resources remain to be exploited (Anglo American Annual Reports and35). This equates to>40 tons of C moved from the mantle to the upper few km of the crust with a far larger mass of C stored in the kimberlites at depth and lost due to erosion of the upper levels of the kimberlite.
Kimberlites represent small degree melts that entrained material from only limited lateral extent within the SCLM. The majority of inclusions at Venetia are harzburgitic and derived from over a depth range of ~100km12. Hence we conclude that diamond formation must be related to large scale tectono-magmatic events.
The different major and trace element geochemistry of the garnet inclusions studied here establishes diamond growth under distinct conditions separated by almost two billion years. Fluid and melt dominated metasomatism result in distinctive trace element signatures with highly to mildly incompatible element contents (for example, LREE/HREE) decreasing strongly from fluids to melts36–38. In a diagram that quantifies the sinuosity of the REE pattern (for example, NdN/ErN) the two garnet inclusion groups show distinct behaviour related to fluid or melt dominated metasomatic diamond growth; (Fig. 5). The older harzburgitic garnets with low CaO, low HREE and sinusoidal REE patterns can be related to low-T fluid dominated metasomatism. Given the inclusion compositions, the metasomatic fluid was likely low in Ca, Fe, Ti, and Al, a characteristic of melt-depleted residua. The LREE enrichment in the garnet inclusions, however, indicates involvement of a C-H-O medium enriched in trace elements that reintroduced LREE to the depleted SCLM. In contrast, the younger garnet inclusions have elevated CaO, FeO and HREE. This type of metasomatism is typically associated with high-T silicate melt-metasomatism that forms lherzolitic type garnet inclusions when pervasive36 and is interpreted to be related to melts derived from the asthenosphere.
Based on the regional geological evolution of the Zimbabwe Craton, we infer that the older, 2.95Ga diamonds formed in a highly melt-depleted mantle residue from fluids mobilised by passive asthenospheric decompression during crust-forming magmatism (Fig. 6). The diamond growth event predates the Limpopo Mobile Belt that formed from collision of the Archaean Kaapvaal and Zimbabwe Cratons at ~2.65Ga39. Key to linking the diamond-formation event to magmatism in the region is reconstruction of the crust-mantle architecture prior to assembly of the two cratons. The origin of the Central Zone of the Limpopo Belt, the location of the Venetia diamond mine, is debated40. De Wit et al.41, for example, argue using geophysical data that the SCLM beneath the Central Zone of the Limpopo Belt was part of the Zimbabwe Craton with the overlying crust belonging to the Kaapvaal Craton or to an allochtonous block that over-thrusted towards the north39, 40, 42, 43. The low temperature peridotites from Venetia, however, show marked Si-enrichment more characteristic of the Kaapvaal SCLM, in contrast to the melt depleted diamond facies samples44. These data suggest potential decoupling of shallow and deep peridotite suites, a conclusion compatible with mantle-crust decoupling, which is supported by lead isotope data and the inferred derivation of crustal rocks from a common source with a long-lived high μ (U/Pb) value in the Central Zone and on the Zimbabwe Craton. This Pb isotope signature differs from the inferred low μ source below the Southern Limpopo Belt and the Kaapvaal Craton45. Based on the suggested decoupling, we expect the crustal rocks associated with the 2.95Ga diamond growth event at Venetia to be located to the north in the Zimbabwe Craton. The southern part of this Craton contains multiple crustal units dated at 2.9–3.0Ga46–48. The Chingezi Tonalites in the Belingwe greenstone belt, for example, yielded whole rock Sm-Nd ages between 2.95 and 3.05Ga with an average initial 143Nd/144Nd of 0.5088447, within error of our 2.95 isochron initial ratio (0.50867±13). Similarly, zircons from the Mashaba Tonalite, part of the Early Archaean Tokwe region (Fig. 6), yielded crystallisation ages for major crustal accretion between 2.8 and 3.05Ga49. Both groups of intrusions are inferred to represent the plutonic equivalents of contemporaneous felsic volcanic rocks in the Belingwe and Buhwa greenstone belts50, 51. The tectonic regime at this time is interpreted to have changed from a continental magmatic arc (Tokwe Terrain) to a rifting passive continental shelf associated with asthenospheric decompression and melting of the mantle beneath the lithosphere51. The resulting magmatism across the region is inferred to have driven the fluid-dominated metasomatism at depth within the depleted lithosphere that was responsible for diamond growth at 2.95Ga.
Importantly, the host diamonds of the Archaean inclusions generally have δ13C lower than −5‰ with a significant spread (average=−7.1±2.7‰ vs. PDB, 2SD). The isotopic variation defines a broad positive relation with CaO in the garnet inclusions (Supplementary Fig. 1c). The most subcalcic garnets have diamond hosts with the lightest carbon and vice versa. Notably, the garnets do not have typical subduction signatures, with Nb/La>1 (Supplementary Data 1 and ref. 38). Thus, these δ 13C values are unlikely to involve a recycled organic carbon component52, 53. The variable 13C depletion of this fluid event is considered to reflect Raleigh-style fractionation suggesting diamond formation from relatively small volumes of super-critical fluids. The range in δ 13C values implies diamond precipitation involving oxidation of a reduced fluid (CH4) causing isotopic fractionation to lower δ 13C54, 55. The observed relation between CaO and δ 13C possibly reflects the effect of decreasing solubility in reduced C-saturated fluids moving upwards along a geothermal gradient56.
The 1.15±0.11Ga diamond growth event can be linked to the Umkondo Large Igneous Province (LIP) magmatism that occurred between 1.106 and 1.112Ga over an area of 2 million km2 across the Kaapvaal and Zimbabwe Cratons57 (Fig. 6). Compositions of remnant sills and dykes are homogenous over the region suggesting formation from a large degree melting event in the asthenosphere58. The major and trace element compositions of the younger garnet inclusions imply an origin related to high-T melt metasomatism38 (that is, high HREE are only recorded in samples with inferred temperatures>1190°C). This observation is consistent with melt infiltration into the depleted lithosphere caused by impingement of the thermal anomaly on the base of the SCLM that led to the Umkondo LIP. The δ 13C values of the 1.15Ga diamonds record a restricted range with an average of −4.6 ‰, indistinguishable from asthenospheric carbon1. In addition, the initial ratio of the 1.15Ga isochron (εNd+3.5±5.6) is in agreement with derivation from the convective depleted mantle.
The Venetia data indicate that Proterozoic diamond formation was associated with melt metasomatism resulting from a major thermal perturbation of the SCLM by active asthenospheric upwelling related to a LIP event, whereas Archaean harzburgitic diamond formation can be related to relatively cool fluid-dominated metasomatism caused by rift-related magmatism. Resolving whether the observed change in the tectonic processes responsible for harzburgitic diamond growth is a consequence of planetary-wide temporal evolution requires detailed dating of more diamond inclusion suites. Data presented here demonstrate that dating by pooling of garnet inclusions in diamond may result in averaging of ages from multiple diamond generations and potentially obscure the actual geological events responsible for diamond formation. This conclusion means that ages obtained from pooled peridotitic garnet inclusions from Premier (Cullinan), SA12 and Udachnaya, RU33 may also be incorrect, bringing into question an apparent increase in worldwide peridotitic diamond formation at ~2.0Ga2. Our data provides convincing evidence for at least two distinct diamond growth episodes beneath Venetia. Seven samples with T DM both younger and older than the 3.0Ga age group may reflect additional diamond growth episodes. Further diamond and mineral inclusion investigations are required to assess whether the observed change in the tectonic processes responsible for the distinct harzburgitic diamond growth events at Venetia, is a continent- or even a planetary-wide occurrence.
The studied garnet inclusions and host diamonds were donated to J.W.H. by the Diamond Trading Company, a member of the DeBeers Group. The inclusions had been liberated from their host diamonds by J.W.H and co-workers in the 90s59. The harzburgitic garnet inclusions studied here were selected primarily on the basis of their relative large size to allow precise and accurate Sm-Nd isotope analyses.
Major element compositions were determined on flat unpolished but levelled garnet crystal faces by electron microprobe (EMPA, JEOL JXA 8530F Field Emission Probe Microanalyser, at Utrecht University) following the methods of Timmerman et al.60. After EMPA analyses, carbon coating was removed by washing in ultrapure methanol and milli-Q water. The inclusions were weighed on a Mettler 7 decimal balance with a sensitivity of 0.1µg and spiked with a 149Sm-150Nd mix aiming for a Nd sample/spike ratio of 20 with Nd concentrations estimated based on the Cr2O3-CaO compositions13. Samples were dissolved in a concentrated HF-HNO3 mix on a hotplate at 140°C for a minimum of 3 days. Chemical separation of Nd and Sm from the dissolved matrix was performed in a two column procedure using 0.16ml TRU-resin and 0.7ml LN resin61. Matrix fractions were collected and combined for trace element analyses by inductively coupled plasma mass spectrometry (Thermo X-Series II) at the Vrije Universiteit (VU) Amsterdam. BHVO-2 (n=5) was used to correct for the chemistry yields for each element. Three inclusions, too small to date, were dissolved and analysed for trace elements directly by ICPMS. Data from these extra inclusions overlap both the SIMS data for garnet inclusions from Stachel et al.38 and the dated inclusions processed through chemistry, thus confirming the yield corrections to be appropriate (Supplementary Fig. 2).
Sm and Nd were analysed by Triton Plus using four 1013 Ohm amplifiers installed in the instrument since 201262. Faraday-amplifier gains were determined using the La Jolla Nd standard following the method of Timmerman et al.63. Beams on 1013 Ohm amplifiers where kept below 3×10−13 A. Samples were run to exhaustion. Instrumental reproducibility for 100pg samples is determined by routine analysis of an in-house Nd standard (CIGO) that yields 0.511332±67 (n=49). Long-term reproducibility of 100ng loads of this standard measured on 1011 Ohm amplifiers, yields 0.511334±10 (n=28), which equates to 0.511834 for La Jolla. Isotope composition and isotope dilution Nd and Sm concentrations were determined using an iterative approach to strip the spike and perform an exponential instrumental mass fractionation correction (146Nd/144Nd of 0.721903 for Nd and 147Sm/152Sm of 0.56081 for Sm). Small USGS BHVO aliquots (~1mg) processed along with the inclusions yielded average Sm and Nd concentrations of 6.026±0.038 and 24.34±0.66ppm, respectively, and 143Nd/144Nd=0.512956±0.000074 (2SD, n=7). Total procedural Sm and Nd blanks averaged 0.4 and 0.3pg (n=5). Blank corrections using the isotopic composition of lab solutes (0.511856±90) have a negligible effect on isotope ratios and calculated T DM, and were not performed. Depleted mantle model ages were calculated with the parameters of Michard et al.64.
Carbon isotope ratios were determined on diamond fragments using a Carlo Erba NC 2500 element analyser coupled to a Thermo Finnigan Delta Plus isotope ratio mass spectrometer at VU, Amsterdam. Potential drift was monitored every six samples using pairs of internal standards (WK synthetic diamond powder, −7.16‰ and natural diamond powder supplied by P. Cartigny, −8.22‰). External standards USGS24 (−16.05‰) and VICS (+1.35‰) were measured in triplicate with each batch of 30 samples. Fragments of host diamonds were selected based on morphology, (external surfaces and interiors) and in four cases the specific growth zone that housed the inclusion was identified. Interiors and external surfaces of diamonds have consistent δ 13C except for one diamond (see Supplementary Table 1). Further details of the stable and radiogenic isotope analytical protocols can be found in Timmerman et al., EPSL 201763.
The research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement no. 319209. J.M.K. received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 654208 (Europlanet 2020 RI). J.W.H. thanks the Diamond Trading Company (a member of the De Beers Group of Companies) for the donation of the diamonds used in this study. We thank Richard Smeets, Bas van de Wagt and Sergei Matveev for support in the VU and UU laboratories and Suzette Timmerman for data discussion.
J.M.K. performed the majority of the lab work and wrote the first version of the paper. All authors were involved in extensive discussions of data interpretation and edited the manuscript. M.U.G. and G.R.D. performed part of the major element and carbon isotope analyses. J.W.H. selected and liberated the inclusions and provided the samples.
The authors declare no competing financial interests.
Electronic supplementary material
Supplementary Information accompanies this paper at doi:10.1038/s41467-017-00564-x.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.