Osmium isotopic constraints on the nature of the DUPAL anomaly from Indian mid-ocean-ridge basalts

The isotopic compositions of mid-ocean-ridge basalts (MORB) from the Indian Ocean have led to the identification of a large-scale isotopic anomaly relative to Pacific and Atlantic ocean MORB. Constraining the origin of this so-called DUPAL anomaly may lead to a better understanding of the genesis of upper-mantle heterogeneity. Previous isotopic studies have proposed recycling of ancient subcontinental lithospheric mantle or sediments with oceanic crust to be responsible for the DUPAL signature. Here we report Os, Pb, Sr and Nd isotopic compositions of Indian MORB from the Central Indian ridge, the Rodriguez triple junction and the South West Indian ridge. All measured samples have higher (187)Os/(188)Os ratios than the depleted upper-mantle value and Pb, Sr and Nd isotopic compositions that imply the involvement of at least two distinct enriched components in the Indian upper-mantle. Using isotopic and geodynamical arguments, we reject both subcontinental lithospheric mantle and recycled sediments with oceanic crust as the cause of the DUPAL anomaly. Instead, we argue that delamination of lower continental crust may explain the DUPAL isotopic signature of Indian MORB.     

Contrasting origins of the upper mantle revealed by hafnium and lead isotopes from the Southeast Indian Ridge

The origin of the isotopic signature of Indian mid-ocean ridge basalts has remained enigmatic, because the geochemical composition of these basalts is consistent either with pollution from recycled, ancient altered oceanic crust and sediments, or with ancient continental crust or lithosphere. The radiogenic isotopic signature may therefore be the result of contamination of the upper mantle by plumes containing recycled altered ancient oceanic crust and sediments, detachment and dispersal of continental material into the shallow mantle during rifting and breakup of Gondwana, or contamination of the upper mantle by ancient subduction processes. The identification of a process operating on a scale large enough to affect major portions of the Indian mid-ocean ridge basalt source region has been a long-standing problem. Here we present hafnium and lead isotope data from across the Indian-Pacific mantle boundary at the Australian-Antarctic discordance region of the Southeast Indian Ridge, which demonstrate that the Pacific and Indian upper mantle basalt source domains were each affected by different mechanisms. We infer that the Indian upper-mantle isotope signature in this region is affected mainly by lower continental crust entrained during Gondwana rifting, whereas the isotope signature of the Pacific upper mantle is influenced predominantly by ocean floor subduction-related processes.     

Recycled oceanic crust observed in 'ghost plagioclase' within the source of Mauna Loa lavas

The hypothesis that mantle plumes contain recycled oceanic crust is now widely accepted. Some specific source components of the Hawaiian plume have been inferred to represent recycled oceanic basalts, pelagic sediments or oceanic gabbros. Bulk lava compositions, however, retain the specific trace-element fingerprint of the original crustal component in only a highly attenuated form. Here we report the discovery of exotic, strontium-enriched melt inclusions in Mauna Loa olivines. Their complete trace-element patterns strongly resemble those of layered gabbros found in ophiolites, which are characterized by cumulus plagioclase with very high strontium abundances. The major-element compositions of these melts indicate that their composition cannot be the result of the assimilation of present-day oceanic crust through which the melts have travelled. Instead, the gabbro has been transformed into a (high-pressure) eclogite by subduction and recycling, and this eclogite has then been incorporated into the Hawaiian mantle plume. The trace-element signature of the original plagioclase is present only as a 'ghost' signature, which permits specific identification of the recycled rock type. The 'ghost plagioclase' trace-element signature demonstrates that the former gabbro can retain much of its original chemical identity through the convective cycle without completely mixing with other portions of the former oceanic crust.     

The return of subducted continental crust in Samoan lavas

Substantial quantities of terrigenous sediments are known to enter the mantle at subduction zones, but little is known about their fate in the mantle. Subducted sediment may be entrained in buoyantly upwelling plumes and returned to the Earth's surface at hotspots, but the proportion of recycled sediment in the mantle is small, and clear examples of recycled sediment in hotspot lavas are rare. Here we report remarkably enriched 87Sr/86Sr and 143Nd/144Nd isotope signatures in Samoan lavas from three dredge locations on the underwater flanks of Savai'i island, Western Samoa. The submarine Savai'i lavas represent the most extreme 87Sr/86Sr isotope compositions reported for ocean island basalts to date. The data are consistent with the presence of a recycled sediment component (with a composition similar to the upper continental crust) in the Samoan mantle. Trace-element data show affinities similar to those of the upper continental crust--including exceptionally low Ce/Pb and Nb/U ratios--that complement the enriched 87Sr/86Sr and 143Nd/144Nd isotope signatures. The geochemical evidence from these Samoan lavas significantly redefines the composition of the EM2 (enriched mantle 2; ref. 9) mantle endmember, and points to the presence of an ancient recycled upper continental crust component in the Samoan mantle plume.     

Some basic concepts and problems on the petrogenesis of intra-plate ocean island basalts

Basaltic magmatism that builds intra-plate ocean islands is often considered to be genetically associated with ＂hotspots＂ or ＂mantle plumes＂. While there have been many discussions on why ocean island basalts （OIB） are geochemically highly enriched as an integral part of the mantle plume hypothesis, our current understanding on the origin of OIB source material remains unsatisfactory, and some prevailing ideas need revision. One of the most popular views states that OIB source material is recycled oceanic crust （ROC）. Among many problems with the ROC model, the ocean crust is simply too depleted （e.g., [La/Sm]PM 〈1） to be source material for highly enriched （e.g., [La/Sm]pM 〉〉 1） OIB, Another popular view states that the enriched component of OIB comes from recycled continental crust （RCC, i.e.; terrigenous sediments）. While both CC and OIB are enriched in many incompatible elements （e.g., both have [La/Sm]PM 〉〉1）, the CC has characteristic enrichment in Pb and deletion in Nb, Ta, P and Ti. Such signature is too strong to be eliminated such that CC is unsuitable as source material for OIB. Plate tectonics and mantle circulation permit the presence of ROC and RCC materials in mantle source regions of basalts, but they must be volumetrically insignificant in contributing to basalt magmatism. The observation that OIB are not only enriched in incompatible elements, but also enriched in the progressively more incompatible elements indicates that the enriched component of OIB is of magmatic origin and most likely associated with low-degree melt metasomatism. H2O and CO2 rich incipient melt may form in the seismic low velocity zone （LVZ）. This melt will rise because of buoyancy and concentrate into a melt rich layer atop the LVZ to metasomatize the growing lithosphere, forming the metasomatic vein lithologies. Erupted OIB melts may have three components： （1） fertile OIB source material from depth that is dominant, （2） the melt layer, and （3） assimilation of the metasomatic vein lithologies formed earlier in the growing/grown lithosphere. It is probable that the fertile source material from depth may be （or contain） recycled ancient metasomatized deep portions of oceanic lithosphere. In any attempt to explain the origin of mantle isotopic end-members as revealed from global OIB data, we must （1） remember our original assumptions that the primitive mantle （PM） soon after the core separation was compositionally uniform/homogeneous with the core playing a limited or no role in causing mantle isotopic heterogeneity; （2） not use OIB isotopes to conclude about the nature and compositions of ultimate source materials without understanding geochemical consequences of subduction zone metamorphism; and （3） ensure that models and hypotheses are consistent with the basic petrology and major/trace element geochemistry.     

Evidence for recycled Archaean oceanic mantle lithosphere in the Azores plume

The compositional differences between mid-ocean-ridge and ocean-island basalts place important constraints on the form of mantle convection. Also, it is thought that the scale and nature of heterogeneities within plumes and the degree to which heterogeneous material endures within the mantle might be reflected in spatial variations of basalt composition observed at the Earth's surface. Here we report osmium isotope data on lavas from a transect across the Azores archipelago which vary in a symmetrical pattern across what is thought to be a mantle plume. Many of the lavas from the centre of the plume have lower 187Os/188Os ratios than most ocean-island basalts and some extend to subchondritic 187Os/188Os ratios-lower than any yet reported from ocean-island basalts. These low ratios require derivation from a depleted, harzburgitic mantle, consistent with the low-iron signature of the Azores plume. Rhenium-depletion model ages extend to 2.5 Gyr, and we infer that the osmium isotope signature is unlikely to be derived from Iberian subcontinental lithospheric mantle. Instead, we interpret the osmium isotope signature as having a deep origin and infer that it may be recycled, Archaean oceanic mantle lithosphere that has delaminated from its overlying oceanic crust. If correct, our data provide evidence for deep mantle subduction and storage of oceanic mantle lithosphere during the Archaean era.     

Evolution of helium isotopes in the Earth's mantle

Degassing of the Earth's mantle through magmatism results in the irreversible loss of helium to space, and high (3)He/(4)He ratios observed in oceanic basalts have been considered the main evidence for a 'primordial' undegassed deep mantle reservoir. Here we present a new global data compilation of ocean island basalts, representing upwelling 'plumes' from the deep mantle, and show that island groups with the highest primordial signal (high (3)He/(4)He ratios) have striking chemical and isotopic similarities to mid-ocean-ridge basalts. We interpret this as indicating a common history of mantle trace element depletion through magmatism. The high (3)He/(4)He in plumes may thus reflect incomplete degassing of the deep Earth during continent and ocean crust formation. We infer that differences between plumes and the upper-mantle source of ocean-ridge basalts reflect isolation of plume sources from the convecting mantle for approximately 1-2 Gyr. An undegassed, primordial reservoir in the mantle would therefore not be required, thus reconciling a long-standing contradiction in mantle dynamics.     

A young source for the Hawaiian plume

Recycling of oceanic crust through subduction, mantle upwelling, and remelting in mantle plumes is a widely accepted mechanism to explain ocean island volcanism. The timescale of this recycling is important to our understanding of mantle circulation rates. Correlations of uranogenic lead isotopes in lavas from ocean islands such as Hawaii or Iceland, when interpreted as model isochrons, have yielded source differentiation ages between 1 and 2.5?billion years (Gyr). However, if such correlations are produced by mixing of unrelated mantle components they will have no direct age significance. Re-Os decay model ages take into account the mixing of sources with different histories, but they depend on the assumed initial Re/Os ratio of the subducted crust, which is poorly constrained because of the high mobility of rhenium during subduction. Here we report the first data on (87)Sr/(86)Sr ratios for 138 melt inclusions in olivine phenocrysts from lavas of Mauna Loa shield volcano, Hawaii, indicating enormous mantle source heterogeneity. We show that highly radiogenic strontium in severely rubidium-depleted melt inclusions matches the isotopic composition of 200-650-Myr-old sea water. We infer that such sea water must have contaminated the Mauna Loa source rock, before subduction, imparting a unique 'time stamp' on this source. Small amounts of seawater-derived strontium in plume sources may be common but can be identified clearly only in ultra-depleted melts originating from generally highly (incompatible-element) depleted source components. The presence of 200-650-Myr-old oceanic crust in the source of Hawaiian lavas implies a timescale of general mantle circulation with an average rate of about 2 (±1)?cm?yr(-1), much faster than previously thought.     

Recycled dehydrated lithosphere observed in plume-influenced mid-ocean-ridge basalt

A substantial uncertainty in the Earth's global geochemical water cycle is the amount of water that enters the deep mantle through the subduction and recycling of hydrated oceanic lithosphere. Here we address the question of recycling of water into the deep mantle by characterizing the volatile contents of different mantle components as sampled by ocean island basalts and mid-ocean-ridge basalts. Although all mantle plume (ocean island) basalts seem to contain more water than mid-ocean-ridge basalts, we demonstrate that basalts associated with mantle plume components containing subducted lithosphere--'enriched-mantle' or 'EM-type' basalts--contain less water than those associated with a common mantle source. We interpret this depletion as indicating that water is extracted from the lithosphere during the subduction process, with greater than 92 per cent efficiency.     

Helium isotopic evidence for episodic mantle melting and crustal growth

The timing of formation of the Earth's continental crust is the subject of a long-standing debate, with models ranging from early formation with little subsequent growth, to pulsed growth, to steadily increasing growth. But most models do agree that the continental crust was extracted from the mantle by partial melting. If so, such crustal extraction should have left a chemical fingerprint in the isotopic composition of the mantle. The subduction of oceanic crust and subsequent convective mixing, however, seems to have largely erased this record in most mantle isotopic systems (for example, strontium, neodymium and lead). In contrast, helium is not recycled into the mantle because it is volatile and degasses from erupted oceanic basalts. Therefore helium isotopes may potentially preserve a clearer record of mantle depletion than recycled isotopes. Here I show that the spectrum of 4He/3He ratios in ocean island basalts appears to preserve the mantle's depletion history, correlating closely with the ages of proposed continental growth pulses. The correlation independently predicts both the dominant 4He/3He peak found in modern mid-ocean-ridge basalts, as well as estimates of the initial 4He/3He ratio of the Earth. The correspondence between the ages of mantle depletion events and pulses of crustal production implies that the formation of the continental crust was indeed episodic and punctuated by large, potentially global, melting events. The proposed helium isotopic evolution model does not require a primitive, undegassed mantle reservoir, and therefore is consistent with whole mantle convection.     

Tungsten isotope evidence that mantle plumes contain no contribution from the Earth's core

Osmium isotope ratios provide important constraints on the sources of ocean-island basalts, but two very different models have been put forward to explain such data. One model interprets (187)Os-enrichments in terms of a component of recycled oceanic crust within the source material. The other model infers that interaction of the mantle with the Earth's outer core produces the isotope anomalies and, as a result of coupled (186)Os-(187)Os anomalies, put time constraints on inner-core formation. Like osmium, tungsten is a siderophile ('iron-loving') element that preferentially partitioned into the Earth's core during core formation but is also 'incompatible' during mantle melting (it preferentially enters the melt phase), which makes it further depleted in the mantle. Tungsten should therefore be a sensitive tracer of core contributions in the source of mantle melts. Here we present high-precision tungsten isotope data from the same set of Hawaiian rocks used to establish the previously interpreted (186)Os-(187)Os anomalies and on selected South African rocks, which have also been proposed to contain a core contribution. None of the samples that we have analysed have a negative tungsten isotope value, as predicted from the core-contribution model. This rules out a simple core-mantle mixing scenario and suggests that the radiogenic osmium in ocean-island basalts can better be explained by the source of such basalts containing a component of recycled crust.     

High 3He/4He ratios in picritic basalts from Baffin Island and the role of a mixed reservoir in mantle plumes

The high 3He/4He ratio of volcanic rocks thought to be derived from mantle plumes is taken as evidence for the existence of a mantle reservoir that has remained largely undegassed since the Earth's accretion. The helium isotope composition of this reservoir places constraints on the origin of volatiles within the Earth and on the evolution and structure of the Earth's mantle. Here we show that olivine phenocrysts in picritic basalts presumably derived from the proto-Iceland plume at Baffin Island, Canada, have the highest magmatic 3He/4He ratios yet recorded. A strong correlation between 3He/4He and 87Sr/86Sr, 143Nd/144Nd and trace element ratios demonstrate that the 3He-rich end-member is present in basalts that are derived from large-volume melts of depleted upper-mantle rocks. This reservoir is consistent with the recharging of depleted upper-mantle rocks by small volumes of primordial volatile-rich lower-mantle material at a thermal boundary layer between convectively isolated reservoirs. The highest 3He/4He basalts from Hawaii and Iceland plot on the observed mixing trend. This indicates that a 3He-recharged depleted mantle (HRDM) reservoir may be the principal source of high 3He/4He in mantle plumes, and may explain why the helium concentration of the 'plume' component in ocean island basalts is lower than that predicted for a two-layer, steady-state model of mantle structure.     

An olivine-free mantle source of Hawaiian shield basalts

More than 50 per cent of the Earth's upper mantle consists of olivine and it is generally thought that mantle-derived melts are generated in equilibrium with this mineral. Here, however, we show that the unusually high nickel and silicon contents of most parental Hawaiian magmas are inconsistent with a deep olivine-bearing source, because this mineral together with pyroxene buffers both nickel and silicon at lower levels. This can be resolved if the olivine of the mantle peridotite is consumed by reaction with melts derived from recycled oceanic crust, to form a secondary pyroxenitic source. Our modelling shows that more than half of Hawaiian magmas formed during the past 1 Myr came from this source. In addition, we estimate that the proportion of recycled (oceanic) crust varies from 30 per cent near the plume centre to insignificant levels at the plume edge. These results are also consistent with volcano volumes, magma volume flux and seismological observations.     

Re-visiting barium isotope compositions of mid-ocean ridge basalts and the implications

Barium (Ba) isotopes can be used as potential tracers for crustal material recycling in the mantle. Determination of the Ba isotope composition of the depleted mantle is essential for such applications. However, Ba isotope data for mantle-derived basalts are still rare. In this study, we reported high-precision Ba isotope data of 30 oceanic basalts including 25 mid-ocean ridge basalts (MORBs) from geochemically and geologically diverse mid-ocean ridge segments and five back-arc basin basalts. The δ138/134Ba values of these samples varied from ?0.06‰ to +0.11‰, with no systematic cross-region variation. Together with published data, we constrained the average δ138/134Ba of global MORBs to +0.05‰±0.09‰ (2 standard deviation, n = 51). Based on depleted MORBs that have (La/Sm)_N < 0.8, low 87Sr/86Sr (< 0.70263), and low Ba/Th < 71.3, we estimated the average δ138/134Ba of the depleted MORB mantle (DMM) as + 0.05‰ ± 0.05‰ (2SD, n = 16) that is significantly lower than the DMM (≈ 0.14‰) reported previously. If a new estimation of the DMM is applied, it is unreasonable to infer that the Ba isotope signatures of the “enriched-type” MORBs (E-MORBs) could be attributed to pervasive sediment recycling in the upper mantle. We, therefore, conclude that the Ba isotope compositions of the E-MORBs could be sourced from the incorporation of subducted altered oceanic crust and/or sediments depending on the Ba isotope composition and other geochemical information of the local mantle.     

Geochemical characteristics and tectonic implications of HP-UHP eclogites and blueschists in southwestern Tianshan, China

Four rock assemblages in correspondence with two different tectonic settings have been recognized in the NEE-SWW extending HP-UHP metamorphic belt in southwestern Tianshan, northwest China. Eclogite assemblage EC1 is geochemically akin to alkaline within-plate oceanic island basalt (OIB). EC2 shows affinity to enriched mid-oceanic ridge basalt (EMORB). Rare earth element (REE) and other immobile trace element characteristics of blueschist assemblage BS1 resemble those of normal mid-oceanic ridge basalt (NMORB). These three assemblages are likely formed on a seamount setting, and the prevalent presence of carbonate minerals and omphacite quartzite stripes/gobbets suggests ancient pelagic sediments including marls are probably developed upon the basaltic seamount. Whereas the geochemical characteristics of BS2 assemblage are of volcanic arc basalt-type. The seamount with the pelagic sediments on it is brought into the subduction zone, and volcanic arc basalts formed on the active continental margin and trench sediments are eroded and enwrapped in the subducting mass, they are altogether subjected to high to ultrahigh pressure metamorphism and subsequent exhumation towards surface. The HP-UHP metamorphic belt is thus interpreted as a subduction-accretionary complex formed by tectonic juxtaposition and imbrication of seamount, seafloor, trench and volcanic arc sequences during oceanic crust subduction.     

Lithium isotope evidence for subduction-enriched mantle in the source of mid-ocean-ridge basalts

'Recycled' crustal materials, returned from the Earth's surface to the mantle by subduction, have long been invoked to explain compositional heterogeneity in the upper mantle. Yet increasingly, problems have been noted with this model. The debate can be definitively addressed using stable isotope ratios, which should only significantly vary in primitive, mantle-derived materials as a consequence of recycling. Here we present data showing a notable range in lithium isotope ratios in basalts from the East Pacific Rise, which correlate with traditional indices of mantle heterogeneity (for example, 143Nd/144Nd ratios). Such co-variations of stable and radiogenic isotopes in melts from a normal ridge segment provide critical evidence for the importance of recycled material in generating chemical heterogeneity in the upper mantle. Contrary to many models, however, the elevated lithium isotope ratios of the 'enriched' East Pacific Rise lavas imply that subducted ocean crust is not the agent of enrichment. Instead, we suggest that fluid-modified mantle, which is enriched during residency in a subduction zone, is mixed back into the upper mantle to cause compositional variability.     

Thallium isotopic evidence for ferromanganese sediments in the mantle source of Hawaiian basalts

Ocean island basalts are generally thought to be the surface expression of mantle plumes, but the nature of the components in the source regions of such mantle plumes is a subject of long-standing debate. The lavas erupted at Hawaii have attracted particular attention, as it has been proposed that coupled 186Os and 187Os anomalies reflect interaction with the Earth's metallic core. It has recently been suggested, however, that such variations could also result from addition of oceanic ferromanganese sediments to the mantle source of these lavas. Here we show that Hawaiian picrites with osmium isotope anomalies also exhibit pronounced thallium isotope variations, which are coupled with caesium/thallium ratios that extend to values much lower than commonly observed for mantle-derived rocks. This correlation cannot be created by admixing of core material, and is best explained by the addition of ferromanganese sediments into the Hawaii mantle source region. However, the lack of correlation between thallium and osmium isotopes and the high thallium/osmium ratios of ferromanganese sediments preclude a sedimentary origin for the osmium isotope anomalies, and leaves core-mantle interaction as a viable explanation for the osmium isotope variations of the Hawaiian picrites.     

Ancient, highly heterogeneous mantle beneath Gakkel ridge, Arctic Ocean

The Earth's mantle beneath ocean ridges is widely thought to be depleted by previous melt extraction, but well homogenized by convective stirring. This inference of homogeneity has been complicated by the occurrence of portions enriched in incompatible elements. Here we show that some refractory abyssal peridotites from the ultraslow-spreading Gakkel ridge (Arctic Ocean) have very depleted 187Os/188Os ratios with model ages up to 2 billion years, implying the long-term preservation of refractory domains in the asthenospheric mantle rather than their erasure by mantle convection. The refractory domains would not be sampled by mid-ocean-ridge basalts because they contribute little to the genesis of magmas. We thus suggest that the upwelling mantle beneath mid-ocean ridges is highly heterogeneous, which makes it difficult to constrain its composition by mid-ocean-ridge basalts alone. Furthermore, the existence of ancient domains in oceanic mantle suggests that using osmium model ages to constrain the evolution of continental lithosphere should be approached with caution.     

Early differentiation and volatile accretion recorded in deep-mantle neon and xenon

The isotopes (129)Xe, produced from the radioactive decay of extinct (129)I, and (136)Xe, produced from extinct (244)Pu and extant (238)U, have provided important constraints on early mantle outgassing and volatile loss from Earth. The low ratios of radiogenic to non-radiogenic xenon ((129)Xe/(130)Xe) in ocean island basalts (OIBs) compared with mid-ocean-ridge basalts (MORBs) have been used as evidence for the existence of a relatively undegassed primitive deep-mantle reservoir. However, the low (129)Xe/(130)Xe ratios in OIBs have also been attributed to mixing between subducted atmospheric Xe and MORB Xe, which obviates the need for a less degassed deep-mantle reservoir. Here I present new noble gas (He, Ne, Ar, Xe) measurements from an Icelandic OIB that reveal differences in elemental abundances and (20)Ne/(22)Ne ratios between the Iceland mantle plume and the MORB source. These observations show that the lower (129)Xe/(130)Xe ratios in OIBs are due to a lower I/Xe ratio in the OIB mantle source and cannot be explained solely by mixing atmospheric Xe with MORB-type Xe. Because (129)I became extinct about 100 million years after the formation of the Solar System, OIB and MORB mantle sources must have differentiated by 4.45 billion years ago and subsequent mixing must have been limited. The Iceland plume source also has a higher proportion of Pu- to U-derived fission Xe, requiring the plume source to be less degassed than MORBs, a conclusion that is independent of noble gas concentrations and the partitioning behaviour of the noble gases with respect to their radiogenic parents. Overall, these results show that Earth's mantle accreted volatiles from at least two separate sources and that neither the Moon-forming impact nor 4.45 billion years of mantle convection has erased the signature of Earth's heterogeneous accretion and early differentiation.     

Water-rich basalts at mid-ocean-ridge cold spots

Although water is only present in trace amounts in the suboceanic upper mantle, it is thought to play a significant role in affecting mantle viscosity, melting and the generation of crust at mid-ocean ridges. The concentration of water in oceanic basalts has been observed to stay below 0.2 wt%, except for water-rich basalts sampled near hotspots and generated by 'wet' mantle plumes. Here, however, we report unusually high water content in basaltic glasses from a cold region of the mid-ocean-ridge system in the equatorial Atlantic Ocean. These basalts are sodium-rich, having been generated by low degrees of melting of the mantle, and contain unusually high ratios of light versus heavy rare-earth elements, implying the presence of garnet in the melting region. We infer that water-rich basalts from such regions of thermal minima derive from low degrees of 'wet' melting greater than 60 km deep in the mantle, with minor dilution by melts produced by shallower 'dry' melting--a view supported by numerical modelling. We therefore conclude that oceanic basalts are water-rich not only near hotspots, but also at 'cold spots'.