Introduction

Much of the Earth’s intraplate volcanism is linked to deep-seated thermochemical plumes of upwelling hot mantle that originate at the core-mantle boundary1,2,3,4. They may originate from the margins of large low shear-velocity provinces (LLSVP) in the lower mantle5,6,7,8 that are interpreted to be a consequence of large-scale mantle convection driven by slab subduction to the lower mantle9,10,11,12. Lavas from the major intraplate volcanic provinces tend to have higher 3He/4He ratios than lavas from mid-ocean ridges that originate in the more vigorously convecting upper mantle13,14, while helium concentrations measured in the former (2–17 × 10−7 cm3 STP/g) are notably lower than in the latter (0.4–2.4 × 10−5 cm3 STP/g)15. The plumes with the highest flux rates, such as Iceland, Hawaii and Galapagos, have the highest 3He/4He13,16,17. Values of up to 65 Ra (where Ra is the present-day value of atmosphere of 1.34 × 10−6)18 have been recorded in early Iceland plume basalts19,20. This ratio is notably higher than that of the convecting upper mantle as recorded by depleted upper mantle-derived mid-ocean ridge basalts (MORB) (8 ± 1 Ra)21. The relative enrichment of primordial 3He in plume basalts reflects the limited degassing of the deep mantle as a consequence of mantle processing by partial melting. This evidence is corroborated by the presence of primitive Ne isotopes in high 3He/4He lavas from the high flux mantle plumes22,23. While the evidence for primordial noble gases in the deep mantle is undeniable, there is no consensus on the location of the high 3He/4He reservoir; explanations range from incompletely outgassed ancient mantle domains24, the core25 or remnants of magma ocean26.

The prevailing models of Earth’s evolution and structure require that the He concentration and 3He/20Ne ratio are higher in the deep mantle than in the more degassed upper mantle13,14,27,28,29. However, the absolute He concentration in high-3He/4He lavas is typically lower than in MORB basalts that are derived from the convecting upper mantle15. This long-standing helium paradox can be resolved if the deep mantle is more degassed than the upper mantle but requires that U is more incompatible in mantle melts than He30,31. However, the experimental determination of the partitioning of helium during melting of peridotite appears to rule out this mechanism32. This, along with the high partition coefficient and solubilities of the light noble gases in bridgmanite33 and ferropericlase34, suggests that the lower mantle should be less degassed than the upper mantle35. This helium paradox can also be explained if the more volatile-rich parental magmas routinely undergo disequilibrium degassing36. However, the extent to which this affects deep mantle-derived melts is difficult to determine. The absence of reliable estimates of the concentration of primordial volatiles in the mantle remains a major hindrance to the development of models of Earth’s evolution37.

The modern Afar mantle plume is a unique laboratory for studying the volatile inventory of the deep mantle. The Sr-Nd-Pb isotope and incompatible trace element (ITE) systematics of modern Afar mantle plume-derived basalts are well established as a mixture of three components; depleted MORB mantle, young HIMU-like plume mantle and Pan-African continental lithosphere38,39. Here, we report new analysis of basaltic glasses from a transect from the Afar mantle plume axis near the Gulf of Tadjoura along the Red Sea (Supplementary Fig. 1). They reveal coherent mixing trends between 3He/4He and ITEs. As the concentrations of trace elements in the main mantle reservoirs are well established, the mixing relationships can be used to determine the relative concentration of primordial He in the upwelling Afar mantle plume and the ambient upper mantle. The data require that the high 3He/4He mantle plume has notably lower He concentration than the upper mantle, an observation that is difficult to square with the consensus view that the deep mantle is a repository of primordial volatiles. We show how He-depleted mantle plumes can form by incorporation of subduction oceanic crust and discuss the implications.

Afar mantle plume

The earliest manifestation of the Afar mantle plume is the two-km thick sequence of continental flood basalts that covers ~400,000 km2 in Ethiopia and Yemen formed at around 30 Ma40. Plume arrival triggered continental breakup and generated the Afar triple junction, resulting in the opening of Red Sea, Gulf of Aden and ongoing rifting and magmatism along the Main Ethiopian Rift (MER)41,42,43,44,45. The eruption of mid-ocean ridge basalts and the shallow Moho beneath the Red Sea and Gulf of Aden ridges imply that oceanic crust is now forming46,47,48. The absence of lithospheric mantle beneath Red Sea-Gulf of Tadjoura tends to rule it out as a contributor to the basalt chemistry and may explain why the modern Afar plume basalts have lower TiO2 concentrations than the high 3He/4He HT2 continental flood basalts49. While seismic studies reveal mantle upwelling from the LLSVP at the core-mantle boundary beneath southern Africa50,51,52 there is a strong low-velocity anomaly beneath the Afar region that appears to be rooted in the mantle transition zone53,54,55. The influence of the modern Afar mantle plume is evident from high 3He/4He (up to 16 Ra) and primordial Ne in basalts from Afar, the MER, southern Red Sea and Gulf of Tadjoura23,42,56,57,58,59 and high mantle potential temperature (Tp) determined from inversion of REE concentrations (1370–1490 °C)60,61. The similarity of ITE and radiogenic isotope composition of the high-3He/4He MER-Afar-Gulf of Tadjoura basalts with the earliest high Ti (HT2) lavas of the Afar CFB province (up to 21 Ra)42,57,58,62,63 implies that the region records 30 Myr of plume-derived volcanism.

Results and discussion

Coherent 3He/4He-trace element variation in Red Sea and Gulf of Tadjoura basalts

The 3He/4He of fresh basaltic glasses dredged from the Red Sea (16-26°N) and the Gulf of Tadjoura (see methods section for sample locations) show a progressive southward increase from 8.4 to 14.4 Ra (Fig. 1 and Supplementary Table 1). The 3He/4He from the northern Red Sea (8.47−8.52 Ra) overlap global MORB values21, while high values from the Gulf of Tadjoura basalts (13.61 – 14.39 Ra) approach the highest values reported for modern basalts in the MER and Afar (15Ra)42,57,64. The Gulf of Tadjoura lavas have ITE and Sr, Nd and O isotopic ratios that overlap MER-Afar basalts65,66,67,68. Two enriched mantle components have been previously identified in Gulf of Tadjoura basalts; Tadjoura Enriched Component (TEC), which is dominated by Pan-African continental lithosphere-related, and Ramad Enriched Component (REC) that appears to be HIMU mantle thought to dominate the Afar plume38,39. All the glasses used in this study have Sr-Nd isotopes and trace element ratios indicating the presence of REC component38. Further, the trace element ratios (e.g. U/Pb, K/Nb) of the Gulf of Tadjoura basalts indicate a strong HIMU mantle signature with no evidence of notable lithosphere contribution69. It should be noted that the variation in Sr-Nd-Pb isotopes and the mantle heterogeneity beneath Gulf of Tadjoura revealed by previous studies39 is not observed in basaltic glasses used in this study38. The Pb isotope composition of the Afar basalts is notably less radiogenic than canonical HIMU values70, implying that the subducted oceanic crust that is present in the plume was recycled in the last few hundred Myr, a so-called “young HIMU” component39,71,72,73.

Fig. 1: New 3He/4He ratios of 15 basaltic glass samples along the Red Sea ridge and Gulf of Tadjoura.
figure 1

The data can be found in Supplementary Table 1. Red circles represent E-MORB, whereas blue circles are N-MORB (see text for grouping details). Smaller grey circles represent previous data23,56,120. MORB 3He/4He range is 8 ± 1 Ra21. Error bars in the y-axis represent 1σ.

Full size image

The Red Sea basalts are characterised by a range of La/Sm that is indicative of two mantle components in the upper mantle north of the upwelling Afar plume38,74. The high [La/Sm]n (>1) (where La/Sm is normalised for primitive mantle75) basalts from the southern Red Sea and Gulf of Tadjoura have an enriched MORB (E-MORB) type source, while low [La/Sm]n (<1) basalts from mid- and northern Red Sea are typical of normal MORB (N-MORB)76 (Supplementary Table 1). The heterogeneity of the mantle beneath the Red Sea is supported by the wide range in Sr-Nd-Pb isotopic composition of 13-25°N basalts (87Sr/86Sr = 0.70240-0.70396, 143Nd/144Nd = 0.512951-0.513194, 206Pb/204Pb = 18.040-19.608)38,77,78,79,80,81.

The Red Sea-Gulf of Tadjoura basalts appear to define strongly hyperbolic mixing trends when 3He/4He is plotted against a variety of ITE ratios (Fig. 2a–d). The low 3He/4He of the northern Red Sea basalts is typical of the upper asthenosphere mantle, which can be ascribed as a mixture of N-MORB and E-MORB38, while the high 3He/4He Afar plume component defined by the Gulf of Tadjoura basalts has a strong young HIMU affinity69.

Fig. 2: Helium isotope and trace element ratio mixing plots between Afar plume and heterogeneous upper mantle endmembers.
figure 2

a–d In the mixing plots, Afar plume and heterogeneous upper mantle endmembers are identified by Gulf of Tadjoura and Red Sea basalts. Trace element concentrations in the mantle endmembers are defined as; HIMU for Afar plume mantle from ref. 75; E-MORB (red circles) and N-MORB (blue circles) for the northern Red Sea82. The numbers above each mixing line correspond to the relative He concentrations of the upper mantle (UM) and Afar plume (AP), i.e. [He]UM/AP. Gulf of Tadjoura samples that were previously analysed23,56 are also shown in these plots for illustration, they were not included in the He concentration calculations (Supplementary Table 3). Error bars in the y-axis represent 1σ.

Full size image

The concentration of trace elements in young HIMU, E-MORB and N-MORB mantle end-members are determined from a generation of studies of mantle-derived basalts75,82. We use these constraints to infer the He concentration of the Afar plume (AP) mantle relative to the upper mantle (UM) in the Red Sea basalts. In Fig. 2, mixing lines between the high 3He/4He-HIMU Afar plume mantle and the two upper mantle components using different relative helium concentrations in UM and AP ([He]UM/AP) are shown based on the established end-member trace element concentrations (Supplementary Table 2) (see Methods section for details). The data do not define unique mixing lines; for N-MORB [He]UM/AP values range from 1 to 4 and for E-MORB mixing lines [He]UM/AP values vary from 4 to 20 (Supplementary Table 3).

The mean value determined for each sample using the four highlighted trace element ratios in N-MORB and E-MORB suggests that the Afar plume has 10-25% of the helium concentration of the local upper mantle. This is in stark contrast to the prevailing models of Earth evolution, which require that deep mantle, and therefore mantle plumes, are enriched in primordial helium relative to the degassed upper mantle14,29,36,37,83. The consistency of the [He]UM/AP values, as defined by the range of trace elements, implies that the conclusion is robust.

Helium-depleted mantle in the Afar plume

Assuming that depleted upper mantle has a maximum He concentration ([He]UM) of 4 ×1014 atoms/g28 and [He]UM/AP of 4-10, the upwelling Afar plume mantle has a He concentration of 0.4-1.0 ×1014 atoms/g. The strong young HIMU trace element signature of the Afar plume basalts implies a contribution from recycled oceanic crust (ROC)84,85, providing a possible source of the He-depleted mantle. Within the existing paradigm, it is possible to generate a low [He]-high 3He/4He mantle by mixing primordial He-rich high 3He/4He deep mantle with He-depleted low 3He/4He recycled oceanic crust (Fig. 3). Using deep mantle 3He/4He of 65 Ra20 and He concentration of 1.1 ×1015 atoms/g28,42, the He concentration of the ROC is in the range of 1.8−8.6 × 1013 atoms/g for [He]UM/AP of 4-10 (see the example using [He]UM/AP = 4 in Fig. 3). In this case, the He-depleted slab material dominates the Afar plume mantle (>95%). If the upwelling deep mantle in the proto-Afar plume is itself a mixture of high 3He/4He deep mantle and depleted upper mantle prior to mixing with the slab, the modern Afar plume requires a lower proportion of ROC (>85%) (Fig. 3).

Fig. 3: A schematic diagram of mantle He concentration and 3He/4He (Ra) illustrating how to generate the Afar plume 3He/4He signature.
figure 3

It displays the mixing between He-rich deep mantle and He-depleted mantle reservoirs using [He]UM/AP = 4. Deep mantle He concentration and depleted mantle He concentration data are from refs. 28,42. The yellow rectangle represents the Afar plume generated by [He]UM/AP = 4. The mixed deep mantle (25 Ra) can be generated by a simple mixing between the depleted mantle and deep mantle at the top of the lower mantle, represented by dashed black lines.

Full size image

Assuming that the slab was fully degassed during subduction86,87 and it contains U and Th concentrations of ROC88, the maximum modelled He concentration ([He]max) of 8.6 ×1013 atoms/g implies that the slab was subducted within the last 80 Myr (Supplementary Fig. 2a). This is consistent with the unradiogenic Pb isotope composition of modern basalts, which contain contributions from the Afar plume (206Pb/204Pb = 19.5)60,89. Further, it shows that until ~1 Gyr after subduction, a downgoing slab contains less He than the deep mantle; therefore, in simple binary mixing, the deep-mantle He isotope composition will dominate (Supplementary Fig. 2b).

The best candidate for a young slab in the Afar plume is subducted Tethyan oceanic crust that has been imaged beneath the region51,90. Seismic tomography observations for the Afar triple junction show slow seismic velocities indicating hot upwelling mantle originating from the 660–1000 km mantle transition zone (MTZ) (SGLOBE-rani91 in the SubMachine portal92; Supplementary Fig. 3). These profiles also track fast seismic anomalies between 440–660 km depth underlying the Zagros mountains to the north-east of the upwelling Afar plume. These fast anomalies represent the subducted Zagros-Makran slab that initiated at ~60 Ma93,94. Assuming the density contrast between the upper mantle and oceanic lithosphere (80 kg/m3)10,95,96, the mantle viscosity ratio between oceanic lithosphere to the upper mantle (50)96,97, and the thickness of the Makran slab (70 km)95,98, we calculate a sinking speed of 1.36 cm/year in the upper mantle for the Zagros-Makran slab (based on a simple Stokes sinker calculation – see ”Methods” section). This is consistent with the global mean slab sinking rate of 1.2–1.3 cm/yr for the entire mantle99,100,101. Assuming a plume upwelling rate of 50 cm/yr102, the slab could have subducted to no more than 1,100 km before being incorporated into the upwelling Afar plume. This implies that the Afar plume acquired its chemical and isotopic fingerprint from the Tethyan Zagros-Makran slab during large-scale mixing at the MTZ instead of the CMB. Slab-plume mixing within the MTZ and at upper lower mantle depths beneath Afar could be further facilitated by the presence of a hydrated MTZ103,104 resulting from the continuous subduction of Tethyan slab material since Pangea break-up51,90.

The dominance of ROC in the Afar plume mantle is consistent with the previous observation of limited deep mantle contribution (<5%) in modern Afar plume based on basalt chemistry42. This is difficult to reconcile with the high Tp recorded by MER-Afar basalts as slabs are likely to be notably colder than the deep mantle into which they are subducted105. It can be resolved if the dominant source of ROC in the plume is heated slab edge material. There is evidence to suggest that slab edges have a higher shear wave velocity compared to the centre of the slabs106,107. Slab edges are heated up to ambient mantle temperatures by the time they reach MTZ, allowing for the generation of high Tp readings in MER/Afar basalts105,106.

Implications

This study presents a reliable estimate of the absolute He content of upwelling deep mantle. In contrast to the prevailing paradigm, we find that the Afar mantle plume is depleted in primordial He relative to the convecting upper mantle. The low He content can be reconciled with the trace element and the isotopic composition of the plume lavas and can be plausibly explained by the incorporation of subducted oceanic crust in the last 80 Myr. This process generates domains of primordial volatile-depleted material. Until ~1 Gyr after subduction, downgoing slabs contain less He than the deep mantle (Supplementary Fig. 2). On this timescale, slabs can penetrate the deep mantle. Where these He-depleted mantle domains are incorporated into upwelling plumes, the bulk composition of the resulting intraplate volcanism can be dominated by the recycled slab, yet the He inventory will be dominated by the deep mantle contribution. This explains why high 3He/4He are recorded by many OIBs that display enriched geochemical signatures13 and requires care should be exercised when using moderately high 3He/4He mantle plumes to extrapolate the bulk composition of deep Earth.

Methods

Sample collection

This study uses fresh basaltic glass samples from two young Mid-ocean ridges: Red Sea (from 26-16oN) and Gulf of Tadjoura, Republic of Djibouti (Supplementary Fig. 1). Helium isotope analysis was performed on basaltic glasses dredged from the Red Sea during R/V Poseion P408-1 (the FS Poseidon Fahrtbericht/Cruise Report P408 [POS408] from ref. 108) and R/V Pelagia 64/PE350/351 (RV PELAGIA Cruise Report 64PE350/64PE351 from ref. 109) expeditions and cruises M31/2 from ref. 110, and SO29 from refs. 23,111. The location and trace element composition of the Red Sea basalts can be found in ref. 74. The location of basalt samples dredged from the Gulf of Tadjoura can be found in ref. 38. The trace element composition of these samples were reported using the method of ref. 38 (Supplementary Table 4).

Analysis procedures: helium isotope measurements

The helium isotope composition of fifteen Red Sea-Gulf of Tadjoura basaltic glasses were measured in the SUERC noble gas laboratory (Supplementary Table 1). Ten samples were analysed using a MAP-215-50 noble gas mass spectrometer using procedures reported by ref. 112. Five samples were analysed using Helix SFT noble gas mass spectrometer following procedures reported by ref. 113. In all cases the volatiles were extracted from vesicles in basaltic glasses by in vacuo crushing using a multi-sample hydraulic crusher apparatus. Helium blanks during both analytical sessions were less than 1% of measured 4He signals. Mass spectrometers were calibrated using HESJ He isotope standard.

Calculating relative mantle helium concentrations

In Fig. 2 we plot the He isotope composition of Red Sea basalts against trace element ratios. Mixing lines are plotted between Afar mantle plume and two upper mantle (N-MORB and E-MORB) components using the mass balance equations of Langmuir et al. 114. The helium isotope ratios composition of the Afar plume (AP) is 16 Ra42,59,60,68 and 8 Ra21 for the upper mantle (UM) end-members. There is less clarity on the ITE ratios of the upper mantle as they reflect the source and degree of enrichment84,115,116. In order to accommodate the established heterogeneity in shallow mantle beneath Red Sea38, we define end-member values for N-MORB and E-MORB mantle domains using the highest ratio recorded in samples along Red Sea74. For the Afar plume end-member we use the lowest ratio recorded in Gulf of Tadjoura samples according to the [La/Sm]n grouping (Supplementary Table 1). The mixing lines developed in Fig. 2 are set for [He]UM/AP of 0.5, 1, 5 and 20. The relative proportion of He in the mantle domains ([He]UM/AP) has been calculated for each sample and all trace element ratios (Rb/La, Rb/U, K/La, and K/Nb) (Fig. 2 and Supplementary Table 3) using the best fit of mixing lines for each sample by changing the [He]UM/AP value. The mean of the samples [He]UM/AP value is 7 ± 3 (1σ).

Slab sinking speed V stoke calculation of Makran Slab

We use Stokes law117 to calculate the slab sinking speed VStokes, of the Zagros-Makran subduction. This method assumes a simple Stokes sinker11,118,119, which accepts a higher-density blob (slab) sinking through a viscous fluid (mantle) and calculates the speed of this sphere (Stokes sphere) to calculate the speed that the Zagros-Makran slab has been sinking;

$${V}_{{Stokes}}=C\frac{\Delta {{{\rho }}}{{g}}{\alpha }^{2}}{{{{{\eta }}}}_{{{{\rm{m}}}}}}{{{\rm{;where}}}}$$
(1)
$$C=\,\frac{2+2{{{{\eta }}}}^{{\prime} }}{6+\,9{{{{\eta }}}}^{{\prime} }}{{{\rm{;where}}}}$$
(2)
$${{{{\eta }}}}^{{\prime} }=\frac{{{{{\eta }}}}_{s}}{{{{{\eta }}}}_{m}}$$
(3)

Where Δρ is the density contrast between slab and mantle and has a value of 80 kg/m3 from refs. 10,95,96, η′ is the mantle viscosity ratio between oceanic lithosphere to upper mantle with a value of 50 from refs. 96,97, α is the radius of the slablet with a value of 35,000 m after the thickness of the Makran slab taken to be 70 km from refs. 95,98. ηm is the upper mantle viscosity with a value of 5 × 1020 Pa s from ref. 97 and ηs as the oceanic lithosphere viscosity of 2.5 ×1022 Pa s from ref. 96 and g is the gravitational force taken as 9.81 m/s2.