Underground diamonds reveal unknown Earth history

Adobestock photo of a diamond in the rough.

Editor's Introduction

Primordial and recycled helium isotope signatures in the mantle transition zone

annotated by

Have you ever wondered what our planet looks like from the inside? What happens to the rocks and minerals that experience the superhot temperatures and superhigh pressures in Earth’s mantle? Although there is consensus that Earth’s structure is made up of a solid core, viscous mantle, and a silicate crust, it has been extremely challenging to get information from the deepest regions of our planet. So far, scientists have identified several chemical reservoirs in the mantle region. This research article looks at how we can use diamonds as time capsules to extract information from the oldest and most undisturbed parts of Earth.

Paper Details

Original title
Primordial and recycled helium isotope signatures in the mantle transition zone
Original publication date
Vol. 365, Issue 6454, pp. 692-694
Issue name


Isotope compositions of basalts provide information about the chemical reservoirs in Earth’s interior and play a critical role in defining models of Earth’s structure. However, the helium isotope signature of the mantle below depths of a few hundred kilometers has been difficult to measure directly. This information is a vital baseline for understanding helium isotopes in erupted basalts. We measured He-Sr-Pb isotope ratios in superdeep diamond fluid inclusions from the transition zone (depth of 410 to 660 kilometers) unaffected by degassing and shallow crustal contamination. We found extreme He-C-Pb-Sr isotope variability, with high 3He/4He ratios related to higher helium concentrations. This indicates that a less degassed, high-3He/4He deep mantle source infiltrates the transition zone, where it interacts with recycled material, creating the diverse compositions recorded in ocean island basalts.


Wide ranges of Sr-Nd-Hf-Pb isotopic compositions for ocean island basalts (OIBs) and continental plume–related basalts indicate the existence of different geochemical reservoirs in the mantle (12). This is also evident from their widely variable helium (He) isotope composition of 4.3 to 50 (134), with 3He/4He ratios expressed as R/Ra, the ratio of 3He/4Hesample to the 3He/4Heair standard (1.4 × 10−6). The chemical evolution, nature, and scale of these different reservoirs remain problematic. By contrast, mid–ocean ridge basalts (MORBs) are formed by shallow melting of the upper mantle and show much more uniform helium isotope compositions [8 ± 1 R/Ra (SD)] (5) and more homogeneous Sr-Nd-Hf-Pb isotope compositions. The higher R/Ra values of plume-related basalts relative to MORBs are taken as key evidence of a primordial undegassed reservoir with high 3He/4He ratios present in Earth’s lower mantle (610). However, the preservation of such a reservoir over the Earth’s history has been questioned on the basis of geophysical evidence of slab subduction into the lower mantle and mantle convection models (11). An upper mantle location for the high-3He/4He reservoir has also been suggested on the basis of seismic anomalies, heterogeneities sampled by small degrees of melt, and modeled low–U-Th/3He domains formed through melt depletion (11217). Resolving the existence and location of a long-term preserved, primordial, undegassed, high-3He/4He reservoir, and where it interacts with other reservoirs in the mantle, is key to understanding the evolution of Earth and deep mantle convection.

Diamonds are physically and chemically robust, allowing retention of He isotope signatures that reflect their formation environment (1820). Here, we present helium isotope data of fluid inclusions from sublithospheric diamonds; these data provide an unparalleled perspective on He isotopes from transition zone depths (410 to 660 km). We studied 24 diamonds (1.3 to 6 mm in size) from the Juina-5 and Collier-4 kimberlites and São Luiz River (Juina, Brazil). The Juina area is renowned for the unusually predominantly sublithospheric origin of its diamonds (2123). Our diamonds have physical features and properties that are consistent with other diamonds from Earth’s transition zone (410 to 660 km depth), including dislocations or diffuse growth zones (database S1) and no detectable nitrogen (N) or fully aggregated N defects (database S2) (24). The diamonds contain mineral inclusions seen in other transition zone diamonds (212526), including breyite (database S3 and fig. S1, A and B), coesite, carbonates, sulfides, and amorphous silicate (fig. S1, C and D). Although only diamond C4-106 has a confirmed superdeep origin, on the basis of its breyite mineral inclusions, all diamonds show typical sublithospheric features. After establishing the sublithospheric origin for our diamonds, we measured helium isotopes of the fluid inclusions. We combined these data with picogram analyses of Pb-Sr isotopes of fluid inclusions, trace-element patterns of fluid inclusions, and carbon isotope values of the diamond hosts (24).


Fig. 1 Helium isotope compositions (R/Ra) of fluid inclusions and carbon isotope compositions (δ13C) of their diamond hosts. Two trends (A and B) are observed in the studied sublithospheric Brazilian diamonds. Values for São Luiz cloudy diamonds (31), OIBs from Hawaii (30), and MORBs (5) are shown for comparison. R/Ra values are shown with 1 SD uncertainty, and the δ13C values are averages ± 1 SD variations of multiple analyses across the diamond.

Figure 1 analyzes the change in the helium and carbon isotope compositions in diamond samples from various parts of the world. To quantify the analysis, the helium isotope compositions were measured using the expression R/Ra, where R represents the 3He/4Hesample and R represents the 3He/4Heair. Similarly, the carbon isotope compositions were calculated by expressing the ratios between the stable carbon isotopes as parts per thousand. Mapping the helium isotope compositions with that of carbon isotopes results in two general trends, represented in Figure 1 by the gray, shaded areas labeled A and B.

Area A

Area A, shown in gray on the right-hand side of Figure 1, depicts a trend in the compositions of helium and carbon isotopes. This region has widely varying helium isotope ratios (R/Ra). However, the carbon isotope signature (δ13C) remains mostly stable around –5 parts per thousand. Note the predominance of diamonds from the São Luiz region in this area, closely followed by the ocean island basalts from Hawaii and the mid-ocean ridge basalts.

Area B

Area B depicts a negative correlation between the compositions of helium and carbon, i.e., as the X-axis values (δ13C) increase the Y-axis values (R/Ra) decrease. Unlike Area A, this region covers widely varying isotope compositions of both helium and carbon. However, in trend B, the R/Ra values representing helium isotope concentrations are not as high as observed in trend A.

Helium is trapped in submicrometer fluid inclusions (database S1), as the Juina diamonds are relatively young, likely formed less than 500 million years ago (2327), and He diffusion through a diamond is extremely slow (19). Preservation of He and its isotope signatures in diamond is supported by He heterogeneities within individual diamonds (1820). We removed the outer rim of the diamond to eliminate any 4He implantation [up to 30 μm from the diamond surface (19)] from the surrounding matrix, either from the mantle or the kimberlite. The potential impact of production of radiogenic 4He within the diamond generally results in an R/Ra shift of <0.8, based on low U-Th-Sm concentrations measured on 10 of the studied diamonds (database S2) and assuming a diamond formation age of 500 million years ago (fig. S2). Cosmogenic production of 3He at Earth’s surface may also modify He isotope compositions, which is a concern for diamonds from alluvial sources (São Luiz River). The highest R/Ra value for our diamonds was from the Juina-5 kimberlite, and it coincides with the R/Ra of 49.8 from Baffin Island picrites (4), the highest observed in basalts. Diamonds recovered from kimberlite pipes are less likely to have been exposed to cosmic rays, but we cannot exclude exposure entirely, as the diamonds from Collier-4 and Juina-5 kimberlite pipes were likely recovered from the near-surface oxidized yellow ground kimberlite material. Thus, our studied diamonds provide the most direct and undegassed evidence of the variation in helium isotope compositions in Earth’s transition zone.

Previous studies of δ13C-δ18O diamond-mineral inclusion correlations (25) and major and trace-element abundances of mineral inclusions (2328) have established the presence of subducted material in the transition zone under the Juina region. The He isotopic signatures released from the sparse fluid inclusions vary from 0.7 to 49.9 R/Ra (Figs. 1 and 2). This range is larger than the variation found in plume-related basalts. Lead isotope ratios also bear similarities to plume-related basalts (Fig. 3, fig. S3, and database S2) and, along with trace-element patterns (Fig. 4 and fig. S4), provide evidence for the involvement of subducted material. Furthermore, the large range in 87Sr/86Sr ratios (0.7051 to 0.7260) (Fig. 3) suggests the involvement of old continental crust [enriched mantle (EMII) component]. Low Rb/Sr ratios that do not support the observed radiogenic Sr can be caused by subduction-related loss of Rb relative to Sr. A negative Nb anomaly characterizing the involvement of subducted material is present in all the trace-element patterns of our studied samples. This anomaly implies that a recycled crustal component, and not the ambient mantle, dominates the trace-element budget in the fluid inclusions. Furthermore, the negative Y/Ho, negative Sr, and positive Eu anomalies and the Zr-Hf depletion measured in our samples are all best explained by shallow (crustal) processes (24). Eu anomalies range from 0.01 to 3.2 Eu/Eu* and correlate positively with La/Nd ratios. We found no correlation of trace-element ratios with diamond host C or fluid He isotope compositions, indicating decoupling of volatile from lithophile elements.


Fig. 2 Helium isotopic composition of the fluid inclusions. (A and B) R/Ra versus He concentrations of the fluid inclusions in alluvial and Juina-5 sublithospheric diamonds from array A. (C and D) R/Ra versus He concentrations of fluid inclusions in Collier-4 and Juina-5 sublithospheric diamonds from array B. Helium data from lithospheric fibrous diamonds (34) with a plume component are given for comparison. As most of the noble gases are contained in fluid inclusions in diamonds, some of the variability in He concentrations can be attributed to differences in fluid inclusion population densities between diamonds. Errors are 1 SD. Helium concentrations are in cubic centimeters per gram (cm3/g).

Figure 2 compares the helium isotope compositions (R/Ra) and the individual helium-3 and helium-4 concentrations obtained from the fluid inclusions in the sublithospheric diamonds. The R/Ra values depicted in Figure 2 have been extracted from Figure 1. Panels A and B in Figure 2 show values from Area A in Figure 1 and panels C and D correspond to the Area B in Figure 1.

Panels A & B

In this plot, there is a positive correlation between the R/Ra values and the individual helium-3 and helium-4 compositions. As the R/Ra ratio increases, the helium-3 composition increases while the helium-4 composition decreases. As observed in Figure 1, the R/Ra values from Area A cover a wide range of values (from 0.7 to 49.9).

Panels C & D

In this plot, there is a negative correlation between the R/Ra values and the individual helium-3 and helium-4 compositions. As the R/Ra ratio increases, the helium-3 and helium-4 compositions both also increase. As in Figure 1, the R/Ra values from Area B are restricted to a narrower range of values (from a little less than 1 to 13).


Fig. 3 Strontium 87Sr/86Sr versus 206Pb/204Pb isotopic compositions of fluid inclusions. The data support the presence of enriched mantle (EMII component) in three of the diamonds and a HIMU component in one diamond. Strontium isotope compositions shown are for present day and with a correction for 500 million years of 87Sr ingrowth from 87Rb decay. Data for OIB and MORB fields are from (38). Errors are 2 SD.

Figure 3 shows a comparison between the strontium and lead isotope compositions. It is important to note that strontium-87 is produced by radioactive decay of the rubidium-87 isotope.

The carbon isotope compositions of the diamonds form two arrays with the R/Ra values (Fig. 1). The large range in R/Ra values (trend A, Fig. 1) could be the result of mixing of various mantle reservoirs and oceanic crust–lithosphere with different R/Ra ratios (29) but with typical mantle carbon isotope values centered around −5‰. Similar ranges in He isotope composition were observed in a compilation of MORBs (5), OIBs (30), and two cloudy São Luiz diamonds (31). We also found diamonds that presented a negative correlation between δ13C and R/Ra (trend B, Fig. 1). Although unexpected, a high-3He/4He source dominating the isotope ratio could explain why the R/Ra values are higher than those found for MORBs. This would be most visible in rocks that have low helium concentrations and low U-Th-Sm contents, such as recycled pelagic sediments strongly depleted in almost all their helium and U-Th during subduction; indeed, the low δ13C values of some Juina diamonds have previously been related to subducted pelagic sediments (23). The diamonds that showed mantle-like carbon and low R/Ra values may be related to oceanic lithosphere, which would likely be more retentive of U-Th than sediments would, given their lower water content and diminished propensity to form melt. Mixing between fluids from oceanic lithosphere (mantle δ13C, high U-Th/3He, low R/Ra) and pelagic sediments contaminated by high-3He/4He material (low δ13C, low U-Th, high R/Ra) could explain trend B. Helium is trapped in fluid inclusions during diamond precipitation, likely caused by an interaction between a low-degree oxidized melt from subducted material and ambient reduced mantle or plume material (32). Several additional lines of evidence support a high-3He/4He source, potentially delivered by a mantle plume that may have originated from the African Large Low Shear Velocity Province (33). A clear increase in 3He concentrations and, to a lesser extent, in 4He concentrations, is accompanied by higher 3He/4He ratios for the Collier-4 and Juina-5 diamonds (Fig. 2, C and D). This observation requires that the high-3He/4He source has higher He abundances than reservoirs with low 3He/4He ratios and thus supports the presence of a primordial 3He plume. A similar conclusion was reached for fibrous lithospheric diamonds from Russia (34). Further, a high-3He/4He component could be related to a Cretaceous mantle plume in the Juina area. The contemporaneous formation of alkaline rocks and the Trindade plume track (32) with the young diamond formation age of a sublithospheric Collier-4 diamond [101 ± 7 million years (23)] along with the kimberlite eruption ages around 93 million years ago in the Juina area (35) all provide evidence for a deep high-3He/4He source delivered by a plume. Transportation of superdeep diamonds to shallower depths also has been suggested to occur by a mantle plume (22). Evidence from seismic tomography indicates that this plume-related mantle remains coupled to the lithosphere beneath Brazil today (36).


Fig. 4 Trace-element patterns of fluid inclusions. Three different groups of patterns were observed, revealing anomalies consistent with the influence of recycled material. Individual patterns are shown in fig. S4.
Explanation of Y-axis

Figure 4 gives an estimate of the compositions of trace elements in the fluid inclusions of the sampled diamonds. Here the Y-axis represents the ratio between the sample compositions to that of Cl chondrite, which is very similar to the composition of a solar nebula. This process of normalization using Cl chondrite is a common method in petrology and is indicative of undisturbed solar matter. Hence, in Figure 4, the trace elements to Cl chondrite ratios are almost all less than 1.  With a ratio less than 1, this indicates that the samples under study have undergone considerable radioactive decay.

The complexity of isotope compositions observed in oceanic basalts has attracted development of a variety of models to explain them. Central to these models has been the problem of constraining the heterogeneous He isotope compositions of the OIB source, because these compositions seem to be largely decoupled from those of other radiogenic isotopes. Resolving this issue is critical, because He in particular has been used to define large-scale mantle structures (3) despite debate about the depth of the high-3He/4He source region tapped by OIB (1617). The He isotopic data for fluid inclusions in superdeep diamonds presented here resolve this issue by showing direct evidence that the high-3He/4He source must be present in the deep mantle, beneath a depth of 410 km. Further, the wide-ranging Sr-Pb-C isotope compositions of the superdeep diamond-forming fluids document the extreme variability in the Earth’s transition zone due to recycled crustal inputs. This recycled material is also recorded by C-N-O isotopes in other superdeep diamonds and their mineral inclusions (2537) and clearly has the potential to generate much of the isotopic variation found in OIBs. Our results show that the transition zone is an important heterogeneous reservoir sampled by ascending plumes, ultimately forming OIBs with less extreme isotope compositions due to mixing.




Materials and Methods

Figs. S1 to S5

Tables S1 to S4

Databases S1 to S4

References (3971)



1. C. Class, S. L. Goldstein, Nature 436, 1107–1112 (2005).

2. A. Hofmann, in Treatise on Geochemistry, vol. 3, R. W. Carlson, Ed. (Elsevier Science, ed. 2, 2014), pp. 67–101.

3. D. Hilton, D. Porcelli, in Treatise on Geochemistry, vol. 2, R. W. Carlson, Ed. (Elsevier, 2003), pp. 327–353.

4. F. M. Stuart, S. Lass-Evans, J. G. Fitton, R. M. Ellam, Nature 424, 57–59 (2003).

5. D. W. Graham, Rev. Mineral. Geochem. 47, 247–317 (2002).

6. M. G. Jackson, J. G. Konter, T. W. Becker, Nature 542, 340–343 (2017).

7. H. M. Gonnermann, S. Mukhopadhyay, Nature 459, 560–563 (2009).

8. H. M. Gonnermann, S. Mukhopadhyay, Nature 449, 1037–1040 (2007).

9. D. Porcelli, G. Wasserburg, Geochim. Cosmochim. Acta 59, 4921–4937 (1995).

10. L. Kellogg, G. Wasserburg, Earth Planet. Sci. Lett. 99, 276–289 (1990).

11. R. D. van der Hilst, S. Widiyantoro, E. Engdahl, Nature 386, 578–584 (1997).

12. S. W. Parman, Nature 446, 900–903 (2007).

13. A. Meibom et al., Earth Planet. Sci. Lett. 208, 197–204 (2003).

14. R. L. Christiansen, G. Foulger, J. R. Evans, Geol. Soc. Am. Bull. 114, 1245–1256 (2002).

15. G. Foulger et al., Geophys. J. Int. 146, 504–530 (2001).

16. G. Foulger, D. Pearson, Geophys. J. Int. 145, F1–F5 (2001).

17. D. L. Anderson, Proc. Natl. Acad. Sci. U.S.A. 95, 4822–4827 (1998).

18. S. Timmerman, M. Honda, D. Phillips, A. L. Jaques, J. W. Harris, Mineral. Petrol. 112 (suppl. 1), 181–195 (2018).

19. D. Shelkov, A. Verchovsky, H. Milledge, C. Pillinger, Chem. Geol. 149, 109–116 (1998).

20. M. D. Kurz, J. J. Gurney, W. J. Jenkins, D. E. Lott III, Earth Planet. Sci. Lett. 86, 57–68 (1987).

21. A. Thomson et al., Contrib. Mineral. Petrol. 168, 1081 (2014).

22. M. J. Walter et al., Science 334, 54–57 (2011).

23. G. P. Bulanova et al., Contrib. Mineral. Petrol. 160, 489–510 (2010).

24. Supplementary materials.

25. A. Burnham et al., Earth Planet. Sci. Lett. 432, 374–380 (2015).

26. F. Kaminsky, Earth Sci. Rev. 110, 127–147 (2012).

27. B. Harte, S. Richardson, Gondwana Res. 21, 236–245 (2012).

28. A. Thomson et al., Lithos 265, 108–124 (2016).

29. T. Hanyu, I. Kaneoka, Nature 390, 273–276 (1997).

30. D. R. Hilton, G. M. McMurtry, R. Kreulen, Geophys. Res. Lett. 24, 3065–3068 (1997).

31. R. Burgess, L. Johnson, D. Mattey, J. Harris, G. Turner, Chem. Geol. 146, 205–217 (1998).

32. A. R. Thomson, M. J. Walter, S. C. Kohn, R. A. Brooker, Nature 529, 76–79 (2016).

33. T. H. Torsvik, K. Burke, B. Steinberger, S. J. Webb, L. D. Ashwal, Nature 466, 352–355 (2010).

34. M. Broadley et al., Geochem. Perspect. Lett. 8, 26–30 (2018).

35. F. V. Kaminsky et al., Lithos 114, 16–29 (2010).

36. J. C. VanDecar, D. E. James, M. Assumpção, Nature 378, 25–31 (1995).

37. M. Palot, D. Pearson, R. Stern, T. Stachel, J. Harris, Geochim. Cosmochim. Acta 139, 26–46 (2014).

38. A. Stracke, M. Bizimis, V. J. Salters, Geochem. Geophys. Geosyst. 4, 8003 (2003).