JOURNALOF GEOPHYSICALRESEARCH, VOL. 104,NO. B11,PAGES 25,479-25,491, NOVEMBER 10, 1999

Origin of ocean island basalts: A new model based on lead and helium isotope systematics

Balz S. Kamber and Kenneth D. Collerson

Departmentof Earth Sciences,University of Queensland,Brisbane, Australia

Abstract. Currentmodels of oceanisland basalt (OIB) Pb isotopesystematics based on long- term isolationof recycledoceanic crust (with or withoutsediment) are not supportedby solutions to bothterrestrial Pb paradoxes.It followsthat the lineararrays of OIB datain Pb isotopediagrams are mixing lines andhave no age significance.A new modelis presentedthat takesinto account currentsolutions to bothterrestrial Pb paradoxesand that explains combined Pb andHe isotope evidencein termsof binarymixing. The key featureof thismodel is a two-stageevolution: first, long-termseparation of depletedmantle from undepletedlowermost lower mantle. Mixing between thesetwo reservoirs results in thewide spread in 2ø7pb/2ø4pbratios and generally high (but variable)3He/4He ratios that typify enriched mantle 1 (EM1)OIBs. The second stage involves metasomatismof depletedupper mantle by EM1 type,lowermost mantle-derived melts. Evolution in the metasomatizedenvironment is characterizedby variablebut generallyhigh (Th+U)/(Pb+He) ratiothat leads to a rapidincrease in 2ø8pb/2ø4pband2ø6pb/2ø4pb ratios and decrease in 3He/4He. Mixing betweendepleted mantle and melts from metasomatizedmantle portions reproduces the characteristicsof high g (HIMU) OIBs. The Sr versusNd isotopearray is compatiblewith binary mixingbetween depleted mantle and near-chondritic lowermost mantle because of the large variationin Sr/Nd ratiosobserved in EM1 andHIMU OIBs. OIBs contaminatedby subcontinental lithosphericmantle (EM2) exhibitmore complex isotope systematics that mask their primary geochemicalevolution.

1. Introduction important to note that SCLM can remain isolated from the convective astenospherefor billions of years and hence The vast majority of oceanisland basalts(OIB) are not only evolveto isotopicsignatures that deviatestrongly from those enrichedin incompatibletrace elementsbut also carry isotopic of thedepleted mantle. SCLM can extend past the present-day signatures interpreted to reflect evolution in enriched mantle shorelineof continents,and OIBs that erupt in theirproximity (variable but high (U+Th)/Pb, high Rb/Sr, low Sm/Nd, and low may be contaminated. Furthermore, subcontinental Lu/Hf). A plethora of models have been proposedto explain lithosphericmantle can be entrainedinto the convecting these features, but no consensushas been reached in spite of oceanic astenosphereupon continentalbreakup. The more than three decades of geochemical research into OIBs. geodynamicsignificance of a SCLMOIB sourcecomponent is Among the debated issues two questionsstand out: first, can thereforevery different from that of recycledcomponents. the isotopic signaturesof OIBs be interpreted to reflect the The claim for the existenceof (subducted)recycled existence of chemically distinct domains in the mantle that componentsthat have remainedconvectively isolated over escaped convective homogenization; and second, if such billionsof yearsis problematicif these domains are envisaged domains exist, where in the mantle are they located, what is to residein the depletedmantle. One very strongargument their origin, and how have they withstoodhomogenization? against the persistence of such domains is the second One popularview is basedon the work by Zindler and Hart terrestrialPb paradox.It is now firmly establishedthat the [1986], who proposed that chemically distinct domains do seculardecrease in Th/U ratioof the depletedmantle reflects exist in the mantle and that over time such domains would increasedrecycling of cootinent-derivedU (and Pb) backinto evolve characteristicisotopic signatures.Apart from the mid- the mantleafter establishmentof a pandemicoxidizing ocean ridge basalt (MORB)-source-depleted mantle, the hydrosphereand atmosphere at circa 2.2. Ga [e.g.Kramers and following mantle reservoirshave been suggestedto contribute Tolstikhin,1997; Collerson and Kamber,1999; Elliott et al., to OIB magmatism: (1) recycled oceanic lithosphere, both 1999].On the basisof the homogeneityof the Pb isotope- ancient(up to circa 2 Ga) and young; (2) recycled sediment;(3) inferredTh/U ratio and the observedNb/Th and Nb/U ratiosin subcontinental lithospheric mantle (SCLM); and (4) an MORB it hasto be concludedthat recycled components are enigmatic "plume component" probably derived from the well mixed into the MORB sourcemantle and that substantial lower mantle. chemicaland isotopic heterogeneity in the depleted mantle is There is now strongisotopic and trace element geochemical relativelyshort-lived [Collerson and Kamber,1999]. The evidence that SCLM is a component in some OIBs. It is rangein 2ø7pb/2øapbratios of OIBs(excluding those with a SCLM component)requires, however, that sourcesremained Copyright1999 by theAmerican Geophysical Union. isolatedfor billionsof years.The consensusthat has been reachedon the significance of the second terrestrial Pb paradox Papernumber 1999JB00258. thereforenecessitates a reevaluation of OIB Pb isotope 0148-0227/99/1999JB00258509.00 systematics.

25,479 25,480 KAMBER AND COLLERSON:ORIGIN OF OCEAN ISLAND BASALTS

An independenttest of OIB Pb isotopemodels is possible ratio.It is concludedthat OIBs are a mixtureof onlytwo long- by combinedconsideration of Pb and He isotopesystematics lived components,one beingthe undegasseddeepest portion since these are coupled through their parent isotopes.At of the lower mantleand the otherbeing the depletedMORB present, the He isotope characteristicsof those OIBs with source mantle. We discussimplications of this model for radiogenicPb can only be explainedby opensystem behavior chemicalevolution of the mantleand convection dynamics. [Hanyu and Kaneoka,1998]. In thispaper we presenta reviewof OIB Pb andHe isotope 2. Selection of OIB Database systematicsand show that they requirea two-stageevolution. First, long-termseparation of primitive undegassedlowermost Like continentalcrust, its underlyingSCLM root remains mantlefrom depletedmantle leads to a significantdifference in isolatedfrom the astenospherefor substantialperiods of time the 2ø7pb/2ø4pbratio. Second,short-term modification of Pb and plays a key role in stabilizingand preservingcontinental and He isotope systematics in localized, metasomatized crust. Even though SCLM may be entrained into the portions of the upper mantle, characterizedby variable but astenosphereby delamination,this processhas no bearingon high (Th+U)/(Pb+He) ratio that leads to an increase in the Pb isotopecomposition of the oceanicmantle [Kramers 2ø8pb/2ø4pband 2ø6pb/2ø4pbratios and a decreasein 3He/4He and Tolstikhin,1997]. The effectof SCLM on OIB isotopeand trace elementgeochemistry is contaminationduring magma ascent.SCLM is thus not a sourcecomponent in the stricter sense,and we have excludedfrom the global OIB databaseall SCLM-contaminatedocean islands. Using the interpretationof isotopic and trace element data by original authorswe have omitted data from the following ocean islands: Canaries [Elliott, 1991; Hoernle and Schmincke, 1993]; Comores [Spath et al., 1996]; Heard [Barling et al., 1994]; and Inaccessible [Cliff et al., 1991]. The distinctionbetween subducted sediment component and SCLM is difficult. Figure la shows that the omitted OIBs define a field at high Th/Yb ratios for a given Nb/Yb ratio which extends over almost the entire range of enrichment indicated by the Nb/Yb ratio. This could either reflect i , i i I i ] , , i i , i , i , , , , i , , , i I i i i i contaminationwith sedimentor SCLM. Accordingto Hilton et .... i ' ' ' ' i ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' al. [1995] the HeardIsland OIBs werecontaminated with 4He at shallowlevels due to interactionwith buoyantSCLM. They arguedthat otherocean islands, whose sources were previously consideredto carry the signatureof subductedsediment, may only havebeen contaminated at very shallowlevels by SCLM. Theseconclusions have recently been supported by Os isotope :3GLOSS data from Kerguelen [Hassler and Shimizu, 1998] and the Canaries [Widom et al., 1999]. In fact, many oceanislands (Comores,Cape Verde, and CanaryIslands) that occurclose to the African continenthave geochemicalcharacteristics which can be interpretedto reflect involvementof SCLM [Sp•ithet al., 1996].As neitherAfrica nor Antarcticaare surroundedby 0 10 20 30 40 50 60 subductionzones which could inject sedimentinto the mantle, Nb/Yb the SCLM signatureis likely to be due to contaminationby remnantsof sub-Gondwanalithosphere. For these reasonswe Figure 1. Th/Yb versus Nb/Yb plot for global ocean island excludethe abovementioned data from our investigation. basalt (OIB) dataset. All samples have MgO > 6 wt % to minimize effects of fractional crystallization. (a) OIBs with subcontinental lithospheric mantle (SCLM) contamination 3. Pb Isotope Characteristicsof OIBs (solid diamonds)are Canaries[Elliott, 1991], Comores[Spath et al., 1996], Heard [Barling et al., 1994], and Inaccessible Pb isotope systematicsof OIBs show two prominent [Cliff et al., 1991]. On average, these clearly plot at high featuresin commonPb space(Figure 2a). First, all OIBs plot Th/Yb values for a given Nb/Yb ratio, irrespective of the to the right of the meteoritic isochron. Second,OIBs define a degree of melting. (b) Data sourcesfor OIBs without SCLM linear array that is emphasized by data from individual contamination (open circles) are given in caption to Figure 2. archipelagos. Plot includes additional data for Easter Island [Haase et al., In an important contribution,Chase [1981] concludedthat 1997], Hawaii [Fodor et al., 1992], R6union [Fisk et al., 1988; linear OIB Pb isotopearrays of individualoceanic islands were Albargde et al., 1997], Tristan da Cunha [Weaver et al., 1987], secondaryisochrons that date timing of melt extractionfrom a and Walvis Ridge [Humphris and Thompson,1983]. Only four out of 257 samplesused in the Pb isotopedatabase plot clearly common source. The increasein U/Pb and Th/Pb during above the regressionline (dotted line). SCLM contamination melting resulted in ¸IB Pb plotting to the right of the or a subductedsediment component can thereforebe excluded. Geochron. Chase [1981] interpretedthe commonsource of Average of global subductedsediment (GLOSS [Plank and OIBs to be oceaniccrust from ancientdepleted mantle, defined Langmuir, 1998]) and field defined by SCLM-contaminated by a single-stagePb isotopeevolution with a 238U/2ø4pbratio OIBs are plotted for comparativepurposes. (g) of 7.91+0.04.Figure 2a illustratesthat a linearregression KAMBER AND COLLERSON: ORIGIN OF OCEAN ISLAND BASALTS 25,481

magmasare derivedfrom ancient,recycled oceanic crust [e.g., 15.8 Hofmann, 1997]. However, this interpretationdepends a priori on the assumptionthat the Pb isotopeevolution of the MORB

15.6 sourcemantle can be modeled with a single-stageIt.

15.4 4. Pb Isotope Evolution of the MORB Source Upper Mantle

15.2 Models of the Pb isotope evolution of Earth's geochemical reservoirs must explain both terrestrial Pb paradoxes: (1) the

15.0 fact that modern MORB samples plot to the right of the meteoritic isochron, and (2) the discrepancy that exists between the modern MORB source mantle Th/U ratio of 2.6 15.8 compared to a time-integrated ratio of ~3.8 from the Pb isotope compositions(see Kramers and Tolstikhin [1997] for

15.6 a detailed review). The MORB source mantle Pb evolution scenario of Chase [1981] clearly offers no explanation for either of these paradoxes. All but the most simplistic Pb 15.4 isotope evolutionary models predict that the U/Th/Pb ratios of the MORB source mantle must have changed through 15.2 geologicalhistory. Some models, which try to solve the first terrestrial Pb paradox, proposedthat the MORB sourcemantle

15.0 It increasedwith time due to Pb depletion[e.g., Cummingand Richards, 1975; Chauvel et al., 1995]. However, these models

14 15 16 17 18 19 20 21 22 fail to explainthe secondterrestrial Pb paradox.Chauvel et al. 2O6pb/2O4pb [1995] developed a model to explain the present-day Pb isotopecomposition of MORB but did not solve its temporal Figure 2. The2ø7pb/2ø4pb versus 2ø6pb/2ø4pb isotope evolution.Their modelcan thusnot be usedto test the validity compositionsof ocean island basaltsfrom Azores [Turner et of Chase's [1981] assumptions.Models which solve both al., 1997; Widom et al., 1997], Balleny [K.D. Collersonet al., manuscript in preparation, 1999], Cameroon [Lee et al., paradoxeshave in common that MORB source mantle It 1994], Cook-Austral [Palacz and Saunders,1986; Chauvel et decreased from ~8.5 after accretion to ~6 in late Archaean al., 1992, 1997; Hdmond et al., 1994; Woodhead, 1996; times and increasedagain to ~11 at the present day [e.g., Kogisoet al., 1997], Galapagos[White et al., 1993;Reynolds Zartman and Haines, 1988; Kramers and Tolstikhin, 1997; and Geist, 1995], Hawaii [West et al., 1992; Frey et al., 1994; Collerson and Kamber, 1999]. The postcore formation Chen et al., 1996; Garcia et al., 1996], Iceland [Elliott, 1991; temporal topology of MORB source mantle It is influenced Hards et al., 1995], Pitcairn [Woodheadand McCulloch, 1989; mainly by the amountof continentalcrust present at any given Woodhead and Devey, 1993], St. Helena [Chaffey et al., time, the amount of continentalrecycling, and the timing of 1989], and Tristan da Cunha [le Roex et al., 1990]. All establishmentof a pandemicoxidizing atmosphere(in which sampleshave MgO > 6 wt % (to avoid effectsof fractionation; U becomesmobile duringweathering). The crustvolume versus seeFigure 1) and are from combinedtrace element and isotope data sets. (a) Comparisonof data with the single stage (it = time curve predictedby Pb isotopicconstraints by Kramersand 7.91) depletedmantle Pb isotopeevolution model by Chase Tolstikhin [1997] has recently been confirmed by an [1981]. The combined regressionline and the model curve independent estimate based on the temporal changes in interceptgive identicaldates of 1.7 Ga. (b) Comparisonof data Th/U/Nb systematicsof the MORB source mantle [Collerson with depletedmantle evolutioncurve (solid line) and "erosion and Kamber, 1999]. This modelis alsoin excellentagreement mix" curve (dottedline) by Kramers and Tolstikhin[1997]. No with the observed Nd isotope evolution of depleted mantle- intercept is obtained with the depleted mantle curve. Large derived rocks [Ntigler and Kramers, 1998] and geophysical open circles show the compositionsof depletedmantle and constraintson crustal growth versusdestruction [Reymer and "erosion mix" at 1.7 Ga. The intersection with the tie line Schubert, 1984]. (arrows) showsthat -45% of the Pb would have to be derived To evaluate whether modern OIB basalts could have been from a subductedsedimentary source. generated from ancient subducted oceanic crust, their Pb isotope ratios must be compared to an appropriate Pb evolution curve for the MORB source mantle (e.g., that proposed by Kramers and Tolstikhin [1997]). Figure 2b of the entire OIB Pb isotope database defines a secondary comparesthe OIB Pb isotope regressionto the MORB source isochronwith a slope equivalentto an age of circa 1.7 Ga. This evolution line of Kramers and Tolstikhin [1997]. This is identical to the single stage It (7.91) MORB sourcemantle evolution line differs significantlyfrom a single-stagecurve intercept. (Figure 2a) betweencirca 3 and 1.5 Ga becausethe depleted Two aspects of Chase's [1981] model have become mantle evolved with a strongly reduced It during that time paradigmsto explain the geochemistryof oceanic basalts:(1) period. The most important observationin Figure 2b is that the view that linear arrays of OIB Pb isotopes have age the OIB regression line fails to intersect a realistic MORB significance (i.e., dating the time of melt separation)and (2) source mantle evolution curve. Furthermore, regressions that the ultimate source for incompatible trace element through individual OIB data sets either do not intersect the enriched OIBs with radiogenic Pb (so-called HIMU = high It) mantle curve or intersectit at locii that correspondto much 25,482 KAMBER AND COLLERSON: ORIGIN OF OCEAN ISLAND BASALTS youngermodel ages than those derived from the slopeof the 2ø8pb/2ø6pband 2ø7pb/2ø6pb ratios of these melt inclusions regression.These relationships thus demonstrate that OIBs are which Saal et al. [1998] interpreted to reflect binary mixing. not derivedby meltingof ancientsubducted oceanic crust. At first glance, such mixing models would have to involve a This finding is in apparent conflict with interpretationof number of end-members(because the slopesof individual OIB Nb/U, Ta/Th, and Ce/Pb ratios of basalts.The similarity of the archipelagoscan vary substantially[cf. Chase, 1981]). averageof theseratios in present-dayMORB and OIB has been However, in the following, we show that it is not necessary used to suggest that most OIBs are derived by melting of to postulate a number of end-membersand that the entire ancient subductedoceanic crust [Hofmann, 1997]. However, variation in Pb isotope compositions can be explained by the reason that these ratios (and Ce/Pb) are nonchondritic mixing between only two long-lived end-member reservoirs reflects extraction (and recycling) of continental crust. The with substantiallydifferent 2ø7pb/2ø4pb ratios. The wide similarity of theseratios betweenpresent-day MORB and OIBs variationin the 2ø8pb/2ø4pband 2ø6pb/2ø4pb(but not in would therefore suggestthat OIBs are derived from young 2ø7pb/2ø4pb)ratios can be createdby short-termisolation of oceanic crust. Ancient (1-2 Ga) oceanic crust has ratios very high (Th + U)/Pb reservoirs [e.g., McKenzie and O'Nions, different from present-day MORB [Collerson and Kamber, 1998]. 1999]. The fact that averages of these ratios are identical The two long-lived end-membersin our model are MORB within error must require a different explanation. One (i.e., depletedmantle) and a sourceassociated with DupA1.The important observation is that there is considerably more composition of the MORB end-member is similar to that variation in these ratios in the OIB database than in MORB. defined by Kramers and Tolstikhin [1997], although it is Campbell [1998] has suggestedthat the linear array definedby important to note that it differs between individual ocean OIB data in a Nb/U versusNb/Th diagram is explainedby basins.The compositionof the DupA1 sourceis more difficult mixing melts derivedfrom chondriticand depletedsources. The to constrain.The DupA1 anomalyhas been defined as an array similarity of the averages with present-day MORB may of Pb isotopecompositions that have high 2ø7pb/2ø4pb for a therefore be fortuitous. Furthermore, Nb/U, Ta/Th, and Ce/Pb given2ø6pb/2ø4pb ratio [after Duprd and Allbgre, 1983]. It is ratios in OIBs are too scatteredto warrantusing their average typical of Indian MORB and Southern Hemisphere OIBs valuesto infer a unique sourcechemistry. [Castillo, 1988]. In our model, however, we postulate a putative end-membersource for the DupA1 anomaly which we call DupA1 source that represents a single Pb isotope 5. Role of Continental Sediment in OIB Sources composition rather than an array of compositions.The Pb The only scenarioin which the OIB Pb isotoperegression isotope composition of the DupA1 source is approximatedby could have circa 1.7 Ga age significancewould have to involve the composition of the Discovery Tablemount OIB which melting of a som'ce which represents a mixture of 1.7 Ga exhibits the strongest DupA1 anomaly at the lowest oceanic crust and a component that had evolved with a 2ø6pb/2ø4pb[Sun, 1980]. It is, however,important to keepin substantially higher g. In the framework of current OIB mind that the Discovery Tablemount itself may also have a models, the only plausible high-g componentis sediment MORB contribution and the true DupA1 sourcemay have an derived from continentalcrust. Figure 2b illustratesthat -45% evenhigher 2ø7pb/2ø6pb ratio. Nevertheless, the mainfeature of the original Pb would have to be derived from sediment of our model is that both end-members have a similar (approximated by the "erosion mix" in the model of Kramers 2ø6pb/2ø4pbratio, but the DupA1source has a significantly and Tolstikhin [1997]). The flux of sedimentwould dependon elevated2ø7pb/2ø6pb ratio (Figure 3). Mixingbetween these how much Pb from the original sedimenteventually reached two sourcesalone can thereforeexplain almostthe entire range the mantle (i.e., was not lost by metasomatictransfer during of 2ø7pb/2ø4pbratios of globalOIBs. The tie line between dehydrationin the subductingslab [cf. Chauvel et al., 1995]). present-dayMORB and DupA1defines the unradiogeniclimit of There is, however, strong trace element evidence that the global OIB data array shownin Figure 2a. continental contamination of the source of OIBs is not In contrastto the variation in 2ø7pb/2ø4pb,the entire permissible[e.g., Holmann, 1997]. This is illustrated(Figure spectrum of compositionsof OIBs toward more radiogenic lb) on a plot of the Nb/Yb ratio versusthe Th/Yb ratio JEwart 2ø6pb/2ø4pband 2ø8pb/2ø4pbratios could have evolved over et al., 1998]. With the exceptionof four analyses(out of 257), geologically short timescales (e.g., 100-200 Myr) in an all non-SCLM-contaminatedOIBs have relatively low Th/Yb environment with elevated U/Pb and Th/Pb. On this basis we ratios at a given Nb/Yb ratio. This indicates insignificant proposea two-stageevolution for OIB Pb isotopesystematics. contribution of continent-derivedmaterial (for comparison, In this model, portions of the depleted mantle or oceanic see the locus of global subductingsediment (GLOSS) [Plank lithospheric mantle (that later become OIB sources) are and Langmuir, 1998]) in the OIBs selectedfor the database. metasomatizedby a melt or fluid with DupA1 Pb isotope CombinedPb isotopeand trace element data thus show that the composition (we use the term metasomatismin its widest slopesof the OIB Pb isotopearrays have no age significance. possible sense,describing a processin which depleted upper mantle is refertilized in highly incompatible elements by addition of melts or fluids derived from the DupA1 source).In 6. A New Model for OIB Pb Isotope Systematics addition to inheriting the DupA1 source Pb isotope The linear Pb isotope arrays displayedby individual OIBs compositionwe also proposethat the metasomatizeddepleted must therefore reflect binary mixing. This conclusion is mantle becomes variably enriched in U and Th relative to Pb. strongly supportedby the recent discoveryof direct evidence Over the time that elapsesbetween metasomatismand melting for binary mixing in HIMU islands [Saal et al., 1998]. Saal et of OIB parentalmagma the 2ø6pb/2ø4pb and 2ø8pb/2ø4pb ratios al. [1998] found that the range of Pb isotopecompositions will increase strongly as a function of the g and Th/U ratio of recorded by individual melt inclusions in isotopically the metasomatizeddepleted mantle. The 2ø7pb/2ø4pbratio, homogenouslavas spans 50% of that of the worldwide OIB however, will only increase slightly. The uncertainty in the data. A very strong linear correlation exists between compositionof the true DupA1 sourcePb isotopecomposition KAMBER AND COLLERSON: ORIGIN OF OCEAN ISLAND BASALTS 25,483

15.80 2ø8pb/2ø4pbversus 2ø6pb/2ø4pb space (Figure 3b), but the ..'"4.56 Ga ,, slope defined by the metasomatizedsource is sensitive to the 15.75 / Geochron choice of the Th/U ratio. 15.70 / , This model can be tested by comparison with Pb isotope compositions of OIBs. For example, using Pb isotope data DU from two of the Galapagosislands (Floreana and St. Cruz; data 15,6o •15.65 / /p• from White et al. [1993]) which define distinctive and different 15.55 st. F•øreana linear arrays (Figures 3a and 3b) in common Pb space. The slopeof the St. Cruz data in the uranogenicdiagram (Figure 3a) 15.50 corresponds to an apparent age of circa 2.3 Ga, whereas Floreana data yield a slope equivalent to an apparent age of 15.4015.45 ...... '• ,.c' ...... ',...... MORB, , .... • .... • circa 2.0 Ga. These regression lines intersect present-day 17.0 17.5 18.0 18.5 19.0 19.5 20.0 20.5 21.0 MORB (Figure 3a) and the 150 Ma metasomatized-DupA1 41.0 ..... , .... , .... • .... , .... • .... , .... isochron at g correspondingto 55 and 100. The thorogenicPb can be modeled to produce a good linear correlation if the 40.5i B •2" metasomatized mantle evolved with a Th/U ratio of-2.5. This calculation clearly shows that the entire OIB Pb isotope array can be generated by the envisaged two-stage 4ø.øf Floreana evolution. This model can be used to constrain the nature of f the DupA1 source, and its validity can be assessedusing combined He and Pb isotope systematicsfrom Hawaii.

38.5•38.0• St.C•• 7. He Isotope Constraints on the Source of DupAl 37.537.0• ' •'•"; • ...... -"" MORB• .... • .... • .... [ .... 17.5 18.0 18.5 19.0 19.5 20.0 20.5 21.0 The significance of the DupA1 isotopic signature is 206pb/204pb enigmatic. The key question is whether DupA1 has a well- Figure 3. Two-sta•cPb evolutionmixin• modelshows that defined Pb isotope composition or whether it represents an the cntffc OiB Pb isotopea•a• (•ø7Pb/•ø4Pb versus array sub-parallel to Northern Hemisphere MORB Pb. In the aø•Pb/aø4PbonFi•u•c 3a, •08 Pb/•04] Pbversus •0• Pb/•04 Pb first case, DupA1 isotope compositioncould reflect that of the FJ•u•c3b) canbc •cncmtcdb• manticmctasomatJsm ]50 lower (primitive) mantle. In the latter case, DupA1 could be pdo• to tacit •cnc•atJon.The mctasomatJzJn•tacit (]abc]cd interpreted as long-lived (-1-2 Gyr) isotopically distinct DupA]source) has a compositionwith a hi•hc•fimc-intc•mtcd reservoirs in the upper mantle probably subducted oceanic thanmid-ocean dd•c basalt(•OEB) andis approximatedb• lithosphereand/or sedimentas suggestedby Rehkt•mper and the DJscovc• Tab]cmountocean is]and basalt bacE-ca]cu]atcd Hofmann [1997]. It is significant that many OIBs that plot ]50 •a (usin• • = 8.? and Th/U = •.?). The vccto• above the Northern Hemisphere Pb isotope reference line (and connccfin•the DupA] source with •a' (= 55)and •a" (= ]00) (a) an Jsochmnof ]50 •a (b) calculatedwith a Th/U milo hencecarry a so-calledDupA1 anomaly) have elevated 3He/4He •.5. •JxJn• between a ]50 •a mctasomatJzcdmantic with ratios.As practicallyno 3He is producedin the mantle variablepa (lines connccfin• DupA] source to •a) and [Tolstikhin,1975], elevated 3He/4He ratios are universally da• •OEB can •cnc•atcthe cntffc t•Jan•]cdefined b• the interpretedto reflect incorporationof primordial He in a source •]oba] oceanis]and basalts, shown in Figure•. Fo• •casons with ordinaryupper mantle He [e.g., Farley and Neroda, 1998]. clarity,on]• datafrom two Ga]apa•osis]ands (Santa C•z and The unavoidable implication for conventional geochemical F]o•cana [Wh•t• • •1., ]993]) a•c shownfo• comparison. mantle models is that Pb and He isotopesin OIBs would have •ixin• linesbetween •a' and•a" with •OEB succcssfu]]• to be at least partly decoupled,with Pb derived from a recycled predictthe a•a•s disp]a•cdb• SantaCruz (opendiamonds) and sourceand He being plume-derived [cf. Hanyu and Kaneoka, F]o•cana(cross-hatched squares), respectively. The slopes 1998] or degassed at shallow mantle levels [Hilton et al., thesemJxJ• ]JnCScorrespond to apparent•csc•voff a•cs of cffca•.0 and cffca•.3 Ga, •cspccfivc]•. 1995]. However, Eiler et al. [1998] have recently reported an intriguingcorrelation between He and Pb isotopesin Hawaiian lavas which can be used to test our OIB Pb isotopemodel. precludesdating of the time of evolutionin the metasomatized Assumingthat DupA1 is the lower mantle and has relatively environment. In our model this metasomaticstage lasts 150 constant Pb and He isotope compositions, a qualitative Myr which is similar to the age of the oceaniclithosphere prediction of the evolution of He isotope systematicsof our underlying many ocean islands. We derive the Pb isotope model is illustrated in Figure 4a. The DupA1 component compositionof the metasomatisingDupA1 melt by back- initiallyhas a primordial3He/4He ratio. The 3He/4He ratio will correctingthe DiscoveryTablemount OIB Pb to 150 Ma, using decrease and the 2ø8pb/2ø4pb and 2ø6pb/2ø4pb ratios will g = 8 and Th/U = 2.7. The following calculationsare not increase as a function of time and the ambient (U+Th)/He and sensitive to the exact choice of these parameters.Figure 3a (U+Th)/Pb ratios. Following melting of the metasomatized showsthe compositionsof such metasomatizedparts of the source,the 3He/4He ratio will be furtherchanged by mixing mantle after 150 Myr using g of 50 and 100. with a MORB melt (see Figures4a and 4b for full discussion). Melting of mixturesinvolving normal MORB sourcemantle According to our model, positive correlationsbetween Pb and metasomatizedmantle yields linear arrays in Pb isotope isotoperatios and 3He/4He ratios should result if lineararrays diagramswith slopescorresponding to apparentages in the exist in Pb isotopediagrams. Furthermore, our model predicts order of 1 to 2 Gyr (Figure 3a). A similar result is obtainedin that steep Pb isotope regressionlines should be associated 25,484 KAMBER AND COLLERSON: ORIGIN OF OCEAN ISLAND BASALTS

15.80 / 3He/4Hedecreases 15.75 ,,'f(tirne and[Th+U]/He) 100 15.70 DupAI• ..'"" ,', 15.65 2 .,.'"" o %- 15.60 -1- 60 Q- . • 15.55 X / / mixingb• -

.. 15.50 •[ j•'•/ MORB,3He../4He either I•1C31::1•• // somatize•ddecreasesm.(for antlesource and 1) 20 MORB %'"

15.4515.40 , , i, ,/•1 ...... i •or increases (for• ....source 2) 7 18 19 20 21 22 17.0 17.5 18.0 18.5 19.0 19.5 20.0 20.5 21.0 206pb/2ø4pb 2øSpbfø4pb Figure 4a. Qualitative predictionof He isotopeevolution Figure4c. 2ø6pb/2ø4pbversus 3He/4He (RA) ratios of Loa basedon Pb isotopemodel (Figure 3). The primordial,high and Loihi and Kea basalts(no combinedHe-Pb data available 3He/4He ratio of the DupA] sourcedecreases in the for Koolau).As predictedby thequalitative model (Figure 4a), metasomatizedmantle as a function of time and (U+Th)/He. In both groupsdefine linear trends.There is significantscatter the most extremecases (high-H oceanisland basalts), the ratio which could partly be due to poorerreproducibility of He will approximate0. Mixing with MORE will further lower the isotopemeasurements and due to the factthat He isotopesare 3He/4Heratio for the ]essradiogenic metasomatized mantle usually measuredon phenocrysts,whereas Pb isotopesare (source ]) or increase the ratio for those metasomatized mantle measured on rock powders. Nevertheless, the linear domains which evolved to the most radiogenic Pb regressionsintersect at a compositionnot dissimilarto Pacific compositions(source 2). The prediction of this qualitative MORE. The importantobservation is that the Kea samples mode]is that if a linear array in a Pb isotopeexists for a set of define a much shallower trend in accordance with the model OIEs, linear arrays should also be obtainedfor plots of Pb predictions(the Kea source(3 in Figure 4b) evolvedwith a isotoperatios versus the 3He/4Heratio. higher(Th+U)/He ratio than the Loa and Loihi source(2 in Figure4b)). The He isotopecomposition of the metasomatized mantle sourcescan be estimatedby extrapolatingthe with the primordial He isotope signatures(unless He was regressionlines to the Pb isotopecomposition determined in degassedduring metasomatism cf. [Hilton et al., 1995]). the model(Figure 4b). Note, however,that in reality,these mixing lines would probablybe curvesbecause of differences Unfortunately, a comprehensiveOIB Pb and He isotope in He and Pb concentrations between MORE and databaseonly exists for Hawaii [Eiler et al., 1998]. Koolau, metasomatizedmantle. Nevertheless, to a first approximation, the Kea metasomatizedmantle (3) is estimatedto have had a 3He/4He(RA) ratio of-35 whilethe Loa and Loihi mantle (2) 39.5 wasdistinctly more primordial with a 3He/4Hc(RA) ratio of -60. Simplelever rule allows to extrapolateto theoriginal He isotopecomposition of DupA1. The 3He/4He (RA) ratio of-90 39.0 31 obtainedby this methodis significantlyhigher than the commonlyassumed lower mantle ratio of-35 [Porcelli and 38.5 Wasserburg,1995].

38.0

37.5 Mauna Loa (plus Loihi) and Mauna Kea yield well-definedbut different arrays in thorogenicPb space (Figure 4b). Also shownin Figure4b is a vectorshowing the effectsof DupA1 37.0 17.0 17.5 18.0 18.5 19.0 19.5 20.0 20.5 metasomatism (150 Ma) which satisfies the data (same 206pb/204pb parametersas for the Galapagosislands except for a Th/U ratio of 1.0). Figure 4c showsthat positive correlationsare found Figure 4b. Thorogenic Pb isotope diagram showing compositionsof basalts from Koolau (open circles), Loa and betweenthe He and Pb isotopecompositions and that Mauna Loihi (cross-hatchedsquares) and Kea (solid circles). Data for Loa displaysa distinctivelysteeper array. The topologyof Koolau are from Roden et al. [1994] and Bennett et al. [1996] these mixing lines is likely to be nonlinear due to differences and for Loihi, Loa, and Kea are from compliationby Eiler et al. in Pb and He concentrations in the end-members. Linear [1998]. Note that the three groups define significantly regressionswere calculated (Figure 4c) as the data have different trends as shown by linear regressions. The insufficient spread to constraina mixing hyperbola. End- regressionsintersect narrowly at a compositioncorresponding member3He/4He ratios for MaunaLoa andMauna Kea canbe to the least radiogenicPacific MORB samples[Castillo et al., calculatedfrom the 150 Ma Pb isotopeDupA1 isochron (Figure 1998] except for the anomalous unradiogenic Garrett 4b). This extrapolation to metasomatized mantle before Transform basalts [J. I. Wendt et al., manuscript in mixing with MORB-like mantle shows(in agreementwith the preparation, 1999]. A mantle model was calculatedto predict predictionof the model)that the 3He/4Heratio of the Mauna the compositions of the metasomatized (150 Ma) mantle sourceswith the sameparameters as in Figure 3b, exceptthat a Keasource had been reduced to3He/4He (RA) of-35, whereas Th/U ratio of 1 was found to best fit the data. These end- thatof the of the Mauna Loa source was still -65 3He/4He(RA). member compositionsfor (1) Koolau, (2) Loa and Loihi, and Furthermore,these values can be usedto constrainthe 3He/4He (3) Kea were obtainedwith g of 10, 50, and 100, respectively. ratio of the original DupA1 source(prior to in situ U and Th KAMBER AND COLLERSON:ORIGIN OF OCEAN ISLAND BASALTS 25,485

3.0 -1 and4, the observedlinear relationship of thesedata (r • = (DupAIsource-MORB). .. 0.87) is predictedby our model. In other words, it must be 2.5 expectedthat HIMU islandsshow very little variationin their Samoa...... He isotope compositionswith variation in their Pb isotope 2.0 MaunaKea•,• r2= 0,87 compositions. Our combined Pb-He model explains both 1.5 MORB and OIB trendsand avoidsthe involvementof up to five mantle end-membersproposed by other authors [e.g., Hanan and Graham, 1996]. Oahu

0.5 -....'"?'...... • • Loihi 8. Geological Test of the Mantle Mixing Model :e• Mangaia Two features of our model break with conventional 0.0 i i i 0 50 100 150 200 geochemical views of OIB genesis and therefore must be (A3He/4HeRA)/(A 2ø•pb/2ø4pb) discussedin more detail. The first is our interpretation that the enriched reservoir (DupA1) is the undegassedlower mantle, the Figure 4d. Plot of the slope (m) defined in thorogenicPb second concernsthe inferred moderately high-•t metasomatism spaceversus the total rangein He isotopecomposition divided in the OIB melt sourceregion. by thetotal range in 2ø6pb/2ø4pbcomposition for severalOIB According to our model, OIBs are binary mixtures between data sets(Loa and Loihi and Kea from Eiler et al. [1998]; Oahu MORB sourceupper mantle and undegassedlower mantle. This from Roden et al. [1994] and Bennett et al. [1996]; Samoa is in effect a "rediscovery" of the model first proposed by from Farely et al. [1992]; Mangaiafrom Woodhead[1996] and Hanyu and Kaneoka[ 1997]). The graphshows that the various Morgan [1971] and Schilling [1973] but later rejected on the OIBs define a wide rangein their 2ø8pb/2ø4pbversus grounds that it did not account for the highly variable 2ø6pb/2ø4pbslopes, with Mangaia(HIMU) characterizedby geochemical and isotopic composition of OIBs [see, e.g., the shallowest slope. The total range in He isotope Hofmann, 1997]. However, there is growing geophysicaland compositions(A 3He/4He (R A) = 3He/4He(R A) maximum- geochemical evidence in support of this original "standard" 3He/4He(RA) minimum) divided by the totalrange in Pb model. Improved resolution of seismic tomographyhas shown isotope compositions(A 2ø6pb/2ø4pb= 2ø6pb/2ø4pb that mantle plumes beneath hotspotspreviously interpreted to maximum-2ø6pb/2ø4pb minimum) increases systematically be sampling upper mantle heterogeneities, in fact originate with increasing2ø8pb/2ø4pb versus 2ø6pb/2ø4pb slope. In from the core-mantle boundary and not from the 670-km other words, HIMU islands show very little variation in their discontinuity (e.g., Hawaii [Ji and Nataf, 1998] and Iceland He isotope compositionsfor a given variation in Pb isotope composition.This is predictedby the qualitativemodel (Figure [Shen et al., 1998]). While the tomographic data in 4a). The shallowest slopes in Pb isotope diagrams thus themselves do not necessitate the involvement of lowermost correspondtometasomatized mantle in whichthe 3He/4He (PA) mantle in the source of erupted OIBs, important recent Os ratio is much reduced, thereby strongly limiting the possible isotope data for basalts from Hawaii and the Azores show rangeof 3He/4He(R A) between 0 (HIMU) and -8.5 (MORB). involvement of a core or lower mantle-derived component The strongest variations in He isotope compositions are [Walker et al., 1995; Widomand Shirey, 1996; Brandonet al., expectedfor mixturesbetween pure DupA1 sourceand MORB 1998]. Furthermore, the magnitude of the decreasein both P (shown for comparison). and S wave velocities in the lowermost portion of the mantle is interpreted to reflect partial melting in the vicinity of the D" layer [Mori and Helmberger, 1995; Williams and Garnero, 1996; Revenaughand Meyer, 1997; Vidale and Hedlin, 1998; decay) by the simple lever rule. Using this analysiswe obtain Lay et al., 1998]. It is significant that these ultralow-velocity a valueof ca. 90 3He/4He (RA). zones appear to be spatially correlated with present-day This self-consistent model for Pb and He isotope hotspots [Williams et al., 1998]. This provides further systematicsconstrains the sourceof the DupA1 componentto evidence that plumes emanating from the core-mantle be the undegassed(primitive) portion of the lower mantle. An boundary could provide melts (with DupA1 isotope important feature of our model is the inference that all OIBs composition) that metasomatize the astenosphere or the (except for those with a SCLM component)carry a primordial oceanic lithospheric mantle. He isotopesignature. However, in the radiogenicOIBs (HIMU) These discoveriesnecessitate deep boundary layer processes theoriginally high 3He/4He ratio is reducedto a valueless than to be revaluated, in particular, the effects of volatiles on the MORB due to U and Th decay (possiblyalso by degassing).As solidus of the deep mantle [cf. Zerr et al., 1998]. It is clear a result,HIMU OIBs are characterized bylow (-6 3He/4He(RA)) from the thermal structure of Earth that lower mantle plumes but relatively uniform He isotope ratios in spite of a wide can only originate from the base of the lower mantle. range in Pb isotope ratios. This relationship is illustrated on Geochemicalevidence presented in this paper thus requiresthat Figure 4d in which the slope (m) defined in thorogenic Pb only lowermost lower mantle is undegassed and is spaceversus the total range in He isotope compositiondivided characterizedby near-bulk silicate earth isotope evolution. It by the totalrange in 2ø6pb/2ø4pbcomposition for Koolau, places no constraintson the degree of depletion of the lower Kea, Loihi, Samoa, and Mangaia is plotted. The total range in mantle above the low-velocity boundary layer. The rapidly He isotopecompositions (A 3He/4He (RA)= 3He/4He (RA) improvingpicture of lower mantle dynamicsand structurethus maximum- 3He/4He (RA) minimum) divided by the total range fully supportsour conclusionthat the DupA1 Pb, He (and, as we in Pb isotopecompositions (A 2ø6pb/2ø4pb= 2ø6pb/2ø4pb will show later, Sr and Nd) isotoperatios characterizea region maximum-2ø6pb/2ø4pb minimum) increases systematically in the lowermost mantle rather than ancient recycled oceanic with increasing2ø8pb/2ø4pb versus 2ø6pb/2ø4pb slope. If the lithosphere [Kellogg et al., 1999]. This confirms Castillo's Th/U ratios of natural metasomatized OIB sources is between [1988] suggestionregarding the geographical extent of the 25,486 KAMBER AND COLLERSON:ORIGIN OF OCEANISLAND BASALTS

DupA1 signaturein OIBs and Indian MORB which is closely Mid-AtlanticRidge whichis associatedwith DupA1Pb isotope associatedwith large-scale regions of low seismic velocity in systematics[Kamenetsky et al., 1998]. the lower mantle and that the DupA1 anomalytraces regions of 4. High U/Pb OIB archipelagosshow relatively high upwelling lower mantle. 2ø6pb/2ø4pbratios for a given2ø7pb/2ø4pb value. Halliday et The second feature of our model that needs geological al. [1995 p.133] concludedthat this "featureis mosteasily confirmation is the inferred moderately high-l.tDupAl-induced explained by enrichment of U relative to Pb about 100 Ma metasomatism in the OIB melt source region. Liquid priorto melting,a time similarto the ageof the lithosphere". immiscibili.ty of carbonate-silicate melts stands out among 5. McKenzie and O'Nions [1998] have suggestedthat the intramantle processes that are able to strongly fractionate traceelement geochemistry of manyOIBs requiresfertilization between different incompatible elements [e.g., Veksler et al., of "typical" upper mantle prior to OIB melting and that 1998]. Carbonatite metasomatismin oceanicupper mantle has subductedoceanic crust and/or sedimentare insufficiently been documented [Hauri et al., 1993; $chiano et al., 1994; enriched in the most incompatibleelements to accountfor Johnson et al., 1996; Baker et al., 1998]. It is also a common concentrationstypically found in OIBs with MgO>6 wt %. feature in continentallithospheric mantle [Green and Wallace, Clearly, more researchis neededto better characterizethe 1988; Yaxley et al., 1991; Bell and $irnonetti, 1996; Xu et al., combinedU/Th/Pb systematicsof metasomatizedmantle (and 1996; Ionov et al., 1997; Wiechert et al., 1997; Witt- oceaniclithosphere). On the basisof currentdata it is likely Eickschen et al., 1998]. In some cases(e.g., Cape Verde and that two processeshave the potentialto stronglyfractionate U Canary Islands), carbonatitesare directly associatedwith OIBs and Th from Pb: first, as mentionedabove, separation of a [Kokfelt et al., 1997; Kogarko et al., 1995]. The origin of carbonatemelt from a silicate melt, and second,exhaustion of carbonatite melts themselxresis controversial. Until recently, Ca-perovskite in lower mantle melting. Trace element it was believed (on isotope-geochemicalgrounds [e.g., Hauri distributionbetween lower mantlephases Mg-wustite, Mg- et al., 1993]) that carbonatites could originate from carbon perovskite, and CaSi-perovskite [Kesson et al., 1998] is which was recycledinto the upper mantle with subductedslabs. probably very extreme, with concentrations of most trace $asada et al. [ 1997] refutedthis view basedon the presenceof elementsin CaSiO:•being 3 ordersof magnitudehigher than in excess 129Xe in carbonatites that must be derived from a the otherminerals [Kato et al., 1988;Harte et al., 1994]. If U, reservoir which is less degassedthan MORB sourcemantle. Th, and Pb distributionin CaSi-perovskiteis similarto that of More recently, Marty et al. [1998] reported important, CaTi-perovskite(as indicatedby similar rare earth element irrefutable evidencethat a 380 Ma carbonatitecomplex from (REE) and high-field-strength-elementdistribution), there is a the Kola Peninsula contains contribution from primordial He possibilitythat meltsfrom the lower mantlemay havehighly (3He/4He(R A) upto 19.1)and Ne sources.In our view it is variable U/Pb and Th/Pb ratios, dependingon the residual significant that carbonatites and metasomatized mantle mineralogy. commonly occur in regions associated with DupA1 isotope characteristics[e.g., Toyoda et al., 1994]. Metasomatic melts derived from undegassedmantle therefore exist in nature. In 9. Implicationsfor Sr and Nd Isotope Systematicsof OIBs this respect, it has been suggested that carbonatitic metasomatism is caused by volatile flux associated with CombinedSr and Nd isotopesystematics of OIBs are less mantle plumes thereby fertilizing the upper mantle [Baker et complicatedthan those of Pb andHe. The negativecorrelation al., 1998]. trend between global isotoperatios, originally termed The remaining questionis whether the t.t of metasomatized "mantletrend", was interpretedto reflect mixing between mantle is high enoughto accountfor the observedincrease in depletedand undepleted magma sources. As isotopicdata for the2øspb/2ø4pb and 2ø6pb/2ø4pb ratios over a timeperiod of moreOIBs were obtained,it becameclear that a significant 100-200 Myr. To date, there exist no measurementsof U/Pb number plotted substantiallyaway from the "mantle trend" and Th/Pb ratios of metasomatizingmelt inclusions,but there (Figure 5a). It is importantto note that thoseOIBs that were is strong circumstantial evidence that the envisaged upper omitted from our database on the grounds of SCLM limit for t.t (between 50 and 100) is a realistic estimate for contaminationgenerally plot farthest away from a linear metasomatized mantle. Evidence for this is as follows: mixingtrend [Hofmann,1997 Figure 3]. It is necessaryto 1. Clinopyroxenesfrom metasomatizedmantle xenoliths considerwhether the deviationsfound in the remainingOIBs have la ranging from 12.5 to 162 and Th/U ratios between 1.4 couldbe producedby mixingof onlytwo end-members. and 5.5 [Hauri et al., 1993; Baker et al., 1998]. Silicates from Usingthe mixingequations by DePaoloand Wasserburg Brazilian carbonatites associated with the Parana basaltic [1979]a mixingenvelope was calculated that encompassed the province have an average • of 158_+66and an average Th/U entireOIB 87Sr/86Srversus 143Nd/144Nd isotope array. The ratio of 2.1_+0.8[Toyoda et al., 1994]. These carbonatitesare end-membersused were a typical MORB melt and an enriched from 80 and 130 Ma, and their silicates show the increase in melt with slightlysubchondritic 143Nd/144Nd and slightly 2ø8pb/2ø4pband 2ø6pb/2ø4pb ratios predicted by ourmodel. superchondritic87Sr/86Sr. In orderto encompassall the OIB 2. The He and Xe isotope composition of these Biazilian data (Sr/Nd)MORB/(Sr/Nd)enricheti(K) has to rangefrom 0.2 carbonatites[Sasada et al., 1997] show strongeffects of alpha (uppermixing curve on Figure5a) to 5.0 (lowermixing curve). decayof U and Th and 238U-fission,respectively. This This predictsa wide rangeof--7 to --40 in Sr/Nd ratiosin OIBs. underscoresthe validity of our combinedPb and He isotope Figure5b is a histogramof OIB Sr/Ndratios and clearly shows model. that the observed Sr/Nd variability permits significant 3. Ca-rich melt-inclusions with high U and Th deviationsfrom the "mantletrend". This rangein the Sr/Nd concentrationshave recently been reportedfrom 43øN on the ratio in globalOIBs is ratherunexpected since Sr andNd have KAMBER AND COLLERSON: ORIGIN OF OCEAN ISLAND BASALTS 25,487

near identical bulk distribution coefficients in MORB-style 0.5136 melting [e.g., Hofmann, 1997]. These data indicate that our 0.5134 binary mixing model is not in conflict with Sr and Nd isotope constraints. 0.5132 z '• 0.5130

10. Significanceof the New Model z 0.5128 for Mantle Isotope Taxonomy '- 0.5126 Conventionalmantle isotope taxonomyuses plots of Pb isotoperatios versusSr and Nd isotoperatios. It implicitly 0.5124 follows from our model that the use of such plots is I questionableand may lead to erroneousconclusions. This is 0'51•2.70i''(•.702 0.703 0.704 0.705 C•.706 becausethe Pb isotope systematicsreflect two evolutionary 87Sr/86Sr stages.First, the long timescaleseparation of upperand lower mantlewhich leads to a pronounceddifference in 2ø7pb/2ø4pb 5O ratios. Second, a short timescale evolution in an environment with highly variableU/Pb and Th/Pb. This leadsto pronounced 4O differencesin 2ø6pb/2ø4pband 2ø8pb/2ø4pbratios. Sr and Nd isotope systematics,however, mabdy reflect the long-term 3O separation.Sm/Nd fractionationin particularis limited by the similarity in chemical behavior of these elements. There is 2O potential for some modification of the Sr isotope signature

during the short-timeevolution if metasomatismwere to lead 10 to extreme Rb/Sr fractionation. However, as shown above, mixing calculations demonstratethat such a processis not [1-• i-i requiredto explain the Sr-Nd isotopearray. 0 20 30 40 50 Nevertheless, our model allows us to reinterpret the nature Sr/Nd of previouslypostulated mantle end-members: The enriched mantle 1 (EM1) end-member is characterized Figure 5. Sr-Nd isotopic and elemental systematics in global OIBs excluding those with a SCLM component(data by high2ø7pb/2ø4pb and relatively low 2ø6pb/2ø4pband sourcesas in Figures1 and2). (a) The 143Nd/144Ndversus 2øspb/2ø4pbratios and showsa strongaffinity with primordial 87Sr/86Srisotope plot highlightsthe scatterof OIB data noble gases. Nd and Sr isotopes plot close to chondritic around a linear negative slope (the "mantle array"). A mixing values. Therefore we postulate that the EM1 end-member envelope was calculated to encompassall the data. The representslower mantle with no significanteffects of short primitiveend-member was chosen to have143Nd/144Nd = timescale metasomatism. This could either be due to a lack of 0.5134, 87Sr/86Sr= 0.702, 11.18 ppm Nd, and 135 ppm Sr significant U/Pb and Th/Pb fractionation of the (typical depleted MORB). The enriched end-member has metasomatisingmelt or could indicate that the metasomatic slightlysubchondritic 143Nd/144Nd of 0.5124and slightly stage was very short. Mixing between lower mantle and superchondritic87Sr/86Sr of 0.7055.For the uppermixing curve, Nd and Sr concentrations in the enriched end-member depleted mantle results in the characteristically steep Pb were set to 5.5 ppm and 330 ppm, respectively. This isotope slopes, strongly variable He isotope ratios, and a corresponds to a K value of 0.2; where K is narrowrange in 2ø6pb/2ø4pband 2ø8pb/2ø4pb ratios. (Sr/Nd)moRB/(Sr/Nd)enriched. For thelower mixing curve, Nd The HIMU end-member is characterizedby very radiogenic and Sr concentrations in the enriched end-member were 2ø6pb/2ø4pband 2ø8pb/2ø4pbratios and is associatedwith He assumed to be 52 ppm and 125 ppm, respectively. This isotopesthat are significantlymore evolved than MORB. Nd correspondsto a K value of 5.0. The predictedSr/Nd ratiosof isotope ratios are typically midway between chondrite and the relevant mixtures are also plotted and range between 6.9 depleted mantle, and Sr isotope ratios are typically quite and 37. (b) Histogramof Sr/Nd ratiosin global OIBs. Note that primitive. We postulatethat the HIMU end-memberrepresents the range is significantly greater than that reported by metasomatizedupper mantle (or even oceanic lithospheric DePaolo and Wasserburg[1979] and exceedsthat requiredby mantle) that inherited the isotope composition of lower the mixing calculations (6.9 to 37). It is therefore possible that the entire143Nd/144Nd versus 87Sr/86Sr array of OIBs mantle (i.e., EM1). However, in contrast to EM1, the could be generatedby mixing of only two end-members. metasomaticstage of HIMU lastedsome 100-200 Myr and was characterizedby high U/Pb and Th/Pb ratios. Mixing between HIMU and depleted mantle results in the characteristically shallower (than EM1) Pb isotope slopesand the typically low 3He/4Heratios. The rather primitive Sr isotopecomposition of In our revised nomenclature EM2 is not an end-member but HIMU could reflect involvement of a metasomatic melt/fluid representsSCLM that contaminatesoceanic mantle derived with strongly fractionatedincompatible elements (i.e., low Sr melts. Because of the heterogeneityof SCLM (in terms of concentration, high U and Th concentrations). Obviously, compositionand age) no unique isotope compositioncan be EM1 and HIMU are genetically related in that they both start defined for EM2. Since SCLM-contaminated OIBs were omitted off with lower mantle isotope compositions.There is thus no from the database, our combined Pb-He model cannot be sharpdividing boundarybetween these two end-members. applied to EM2 melts. 25,488 KAMBER AND COLLERSON: ORIGIN OF OCEAN ISLAND BASALTS

The last end-member we require is depleted MORB source 39.0 mantle and we believe that all OIBs entrained some depleted mantle component. All other putative "end-members"are not 38.5 required and can be explained as mixtures of the above four 38.0 ©ADiscoveryTablemount• components. Iow•r=rr•a.•tle:I / • • 37.5 11. Implications for the Dynamics of Mantle Convection and the Chemical Evolution 37.0 of the Mantle 36.5 Our reinterpretation of OIB geochemistry places new constraints on the chemical structure of the mantle and the 36.0 MORBsourcem•,•••,,,,• ...... •tl,e, , dynamics of mantle convection. The most important 6.0 17.5 1 18.5 16.0•1 .... •16.5 .... • ....17.0 • .... • .... • ....8.0 • .... •,.,, 1,9.0 conclusions are as follow: • a 1.oGa .." 1. Apart from SCLM, there is no evidence for long-lived geochemical reservoirs in the oceanic upper mantle. This finding is supportedby considerationof the long-termisotope 15.0 '""' and geochemical history of this reservoir. In particular, the temporal evolutionsof Pb and Nd isotopesand U/Nb and Th/U ratios require that continentalmaterial has been recycledback 14.5•3.5G 3.0 Ga .."'"- into the upper mantle, especially after 2.0 Gyr [Kramers and Tolstikhin, 1997; Kramers et al., 1998; Nagler and Kramers, 1998; Collerson and Kamber, 1999; Elliott et al., 1999]. Yet [ / o;ocron._ = recycled continental material appearsto be well dispersedby 13.0 •/''' • .... •'' •;• '• • I •,,, t, • • I • • I• • • • vigorous convectionin the MORB sourcemantle as evidenced, 11 12 13 14 15 16 17 18 19 for example, by the constant MORB Nb/U ratio (44.5_+2.5). 2O6pb/2O4pb The most anomalousMORB isotopecompositions are found in Indian MORB. On the basis of our model we conclude that Figure 6. (a) Thorogenic Pb isotope evolution of lowermost mantle modeled to predict present-dayisotope compositionof these isotope anomalies reflect stronger leaking of lower Discovery Tablemountocean island basalt. Model parameters mantle material into the astenosphereunder the Indian ocean are initial Pb at 4.3 Ga (taking effects of core formation into (see Figure 6a). account);2ø6pb/2ø4pb = 9.818 and 2øspb/2ø4pb= 29.839 2. Our model requireslowermost mantle to be undegassed, [Kramers and Tolstikhin, 1997]; • = 8.9; Th/U = 4.21; and a relatively undepleted in incompatible elements and to have time of 4.3 Gyr. Also shown are most depletedMORB samples evolved to yield near-bulk-silicate-earth isotope ratios. from Garrett transform (solid diamonds; data from J. I. Wendt Seismicvelocity studiesindicate that deeply subductedoceanic et al., manuscript in preparation, 1999) to illustrate the lithosphere appears to disseminateat a depth of-1500 km observed range in MORB Pb isotope composition. The [van der Hilst et al., 1997; Kellogg et al., 1999]. From these anomalous isotope composition of Indian MORB (cross- observationsit is unlikely that the D" layer which is sampled hatchedsquares; data from Rehk•imperand Hofmann [1997]) is explained by admixture of lowermost mantle material to by OIBs is the repositoryfor subductedoceanic slabs [Kellogg normal MORB. This would indicate mass transfer between et al. , 1999]. deepest and shallowest mantle in an ocean basin that has 3. If our model is correct,the Pb isotopecomposition of the evolved during the last -100 Myr. (b) UranogenicPb isotope Discovery Tablemount OIB can be used to estimate the evolutionof lowermostmantle modeledto predictpresent-day composition of lowermost undegassedmantle. On Figure 6b isotope composition of Discovery Tablemount ocean island we show that the present-day lowermost mantle Pb isotope basalt. Model parameters are initial Pb at 4.3 Ga (taking composition can be modeled by a single-stageg (closed effectsof coreformation into account);2ø6pb/2ø4pb = 9.818 system)of 8.9 and a Th/U ratio of 4.21, startingwith an initial and2ø7pb/2ø4pb = 11.201 [Kramers and Tolstikhin, 1997]; • = compositionof 2ø6pb/2ø4pb= 9.818; 2ø7pb/2ø4pb = 11.201; 8.9; and a time of 4.3 Gyr. MORB sourcemantle and Geochron and2øspb/2ø4pb = 29.839 at 4.3 Ga ([Kramersand Tolstikhin are shown for comparison.Note that the MORB sourcemantle startsto deviate significantly from the lowermostmantle after 1997] taking effects of core formation into account). If the -3 Gyr due to crust extraction (which lowered •) and plots Discovery Tablemount OIB contains a (small) MORB farther to the right of the Geochron due to preferential component(as has to be suspected),the g of lowermostmantle recyclingof U after -2.0 Ga [Collersonand Kamber, 1999]. may by slightly higher, but the Th/U ratio is well within error of chondritic estimates (see Table 1 of Kramers and Tolstikhin [1997] for a compilationof estimates). 4. Our results also have implications for the initial He Porcelli and Wasserburg, 1995; O'Nions and Tolstikhin, isotope composition of Earth. The pertinent question is 1996]. However, terrestrial Ne and Ar isotope systematics whetherEarth's initial He isotopecomposition was solar (280 have recently been interpreted in terms of a solar initial rare 3He/4He(RA)) or "planetary" (-100 3He/4He (RA)). It has been gas isotopecomposition of Earth [Honda et al., 1993; Pepin, argued that a "planetary" compositionis more appropriate 1998]. Our model showsthat the He isotoperatios measuredin [Ozima and Podosek, 1983], and models attemptingto OIBs are loweredby admixtureof MORB-type He and by U and constrain mass transfer between lower and upper mantle Th decay during a metasomaticstage (Figure 4c). Taking these reservoirs have used this value as a starting point [e.g., effects into account, we calculate a present-day lowermost KAMBERAND COLLERSON: ORIGIN OF OCEAN ISLAND BASALTS 25,489 mantleratio of-90 3He/4He(RA). This is almostas high as the Castillo, P. R., J. H. Natland, Y. Niu, and P. F. Lonsdale, Sr, Nd and Pb isotopic variation along the Pacific-Antarctic risecrest, 53-57øS: "planetary"value and given >4 Ga nucleogenic 4Heproduction Implications for the compositionand dynamicsof the South Pacific in a near-chondritic lowermost mantle, the initial ratio of uppermantle, Earth Planet. Sci. Len., 154, 109-125, 1998. Earthmust have been significantly higher than the "planetary" Chaffey, D. J., R. A. Cliff, and B. M. Wilson, Characterisationof the St. value,possibly ashigh as the solar wind value (280 3He/4He Helena magma source,in Magmatismin the OceanBasins, edited by A.D. Saundersand M. J. Norry, Geol. Soc. Spec.Publ., 42, 257-276, (RA)). 1989. Chase, C. G., Oceanic island Pb: Two-stage histories and mantle 12. Summary evolution, Earth Planet. Sci. Lett., 52, 277-284, 1981. Chauvel, C., A. W. Hofmann, and P. Vidal, HIMU-EM: The French A comparisonof OIB Pb isotopesystematics with the Polynesianconnection, Earth Planet. Sci. Lett., 110, 99-119, 1992. Chauvel, C., S. L. Goldstein, and A. W. Hofmann, Hydration and temporalPb isotopeevolution of the depletedmantle dehydration of oceanic crust controls Pb evolution of the mantle, demonstratesthat linear arrays defined by the global OIB Chem. Geol., 126, 65-75, 1995. database and by individual archipelagoshave no age Chauvel, C., W. McDonough, G. Guille, R. Maury, and R. Duncan, significance.They must therefore be mixinglines. The range Contrastingold and young volcanismin Rurutu Island, Austral chain, in 2ø7pb/2ø4pbin OIBs is the result of mixing betweenonly Chem. Geol., 139, 125-144, 1997. Chen, C.-Y., F. A. Frey, J. M. Rhodes, and R. M. Easton, Temporal two reservoirs:the uppermantle and a near-bulk-silicate-earth geochemicalevolution of Kilauea volcano:Comparison of Hilina and g reservoir,which we propose to bethe lowermost mantle. The Punabasalt, in Earth Processes:Reading the IsotopicCode, Geophys. widesprehd in 2ø6pb/2ø4pband2ø8pb/2ø4pb of OIBs, however, Monogr. Ser., vol. 91 edited by A. Basu and S. Hart, pp. 161-181, is the result of short timescale evolution in localized AGU, Washington,D.C., 1996. Cliff, R. A., P. E. Baker, and N.J. Mateel, Geochemistryof Inaccessible environmentswith highlyvariable (Th+U)/Pb. The so-called island volcanics, Chem. Geol., 92, 251-260, 1991. EM1 end-member lacks effects of the short timescale Collerson, K. D., and B. S. Kamber, Evolution of the continents and the modificationwhich is most pronouncedin the HIMU end- atmosphere inferred from Th-U-Nb systematics of the depleted member.The two-stageevolution proposed for OIB Pb (and mantle, Science, 283, 1519-1522, 1999. He) isotopesystematics explains why thereis no correlation Cumming, G. L., and J. R. Richards, Ore lead isotope ratios in a continuouslychanging Earth, Earth Planet. Sci. Lett., 28, 155-171, betweenPb :_,sotopeswith otherradiogenic tracers and thereby 1975. providesan explanationfor the HIMU isotopeparadox DePaolo, D. J., and G. J. Wasserburg, Petrogeneticmixing models and [Ballentineet al., 1997].A two-stageevolution model also Nd-Sr isotopic patterns, Geochim. Cosmochim.Acta, 43, 615-627, obviates the need for decouplingof He and Pb in the 1979. Dupr6, B., and C. J. All•gre, Pb-Sr isotopevariations in Indian Ocean interpretationof OIB genesis[Hanyu and Kaneoka, 1998]. basaltsand mixing phenomena,Nature, 303, 142-146, 1983. Eiler, J. M., K. A. Farley, and E. M. Stolper,Correlated helium and lead Acknowledgments.We thankJan Kramers, James Brenan, and an isotopevariations in Hawaiian lavas, Geochim. Cosmochim.Acta, 62, anonymousreviewer for excellentreviews that led to considerable 1977-1984, 1998. improvementof the manuscript. We wouldalso like to thankMarcel Elliott, T., Element fractionationin the petrogenesisof ocean island Regelousfor compilingparts of the OIB databaseand Tony Ewart, basalts,Ph.D. thesis,Open Univ., Milton Keynes,England, 1991. MassimoGasparon, and Sue Kesson for useful discussions. This work Elliott, T., A. Zindler,and B. Bourdon,Exploring the kappaconundrum: hasbeen supported by SwissNational Science Foundation grant 8220- 050352 to B.S.K. andAustralian Research Council grants to K.D.C. The role of recyclingin the lead isotopeevolution of the mantle, Earth Planet. Sci. Lett., 169, 129-145, 1999. Ewart, A., K. D. Collerson,M. Regelous,J. I. Wendt, and Y. Niu, References Geochemicalevolution within the Tonga-Kermadec-Lauarc-back- arc systems:The role of varyingmantle wedge composition in space Albar•de,F., B. Luais,G. Fitton,M. Semet,E. Kaminski,B. G. J. Upton, and time, J. Petrol., 39, 331-368, 1998. P. Bach•lery,and J.-L. Chemin6e, The geochemical regimes of Piton Farley,K. A., andE. Neroda,Noble gasesin the Earth'smantle, Annu. de la Fournaisevolcano (R6union) during the last 530,000 years,J. Rev. Earth Planet. Sci., 26, 189-218, 1998. Petrol., 38, 171-201, 1997. Farley,K. A., J. H. Natland,and H. Craig,Binary mixing of enriched Baker,J., G. Chazot,M. Menzies,and M. Thirlwall,Metasomatism of andundegassed (primitive?) mantle components (He, Sr, Nd, Pb) in the shallowmantle beneath Yemen by the Afar plumemlmplications Samoanlavas, Earth Planet. Sci. Lett., 111, 183-199, 1992. for mantle plumes,flood volcanism,and intraplatevolcanism, Fisk, M. R., B. G. J. Upton,C. E. Ford,and W. M. White, Geochemical Geology,26, 431-434, 1998. and experimentalstudy of the genesisof ReunionIsland, Indian Ballentine,C. J., D.-C. Lee,and A. N. Halliday,Hafnium isotopic studies Ocean,J. Geophys.Res., 93, 4933-4950, 1988. of the Cameroonline and new HIMU paradoxes,Chem. Geol., 139, Fodor, R. V., F. A. Frey, G. R. Bauer, and D. A. Clague, Ages, rare- 111-124, 1997. earth-elementenrichment and petrogenesisof tholeiiticand alkalic Barling,J., S. L. Goldstein,and I. A. Nicholls,Geochemistry of Heard basalts from Kahoolawe Island, Hawaii, Contrib. Mineral. Petrol., Island (southernIndian Ocean): Characterizationof an enriched 110, 442-462, 1992. mantlecomponent and implications for enrichmentof the sub-Indian Ocean mantle, J. Petrol., 35, 1017-1053, 1994. Frey, F. A., M. O. Garcia, and M. F. Roden, Geochemical characteristics of Koolau Volcano: Implications of intershield Bell, K., and A. Simonetti,Carbonatite magmatism and plume activity: geochemicaldifferences among Hawaiian volcanoes,Geochim. Implicationsfrom the Nd, Pb andSr isotopesystematics of Oldinyo Cosmochim. Acta, 58, 1441-1462, 1994. Lengai,J. Petrol.,37, 1321-1339,1996. Bennett,V. C., T. M. East, and M. D. Norman, Two mantle-plume Garcia, M. O., J. M. Rhodes, F. A. Trusdell, and A. J. Pietruszka, componentsin Hawaiianpicrites inferred from correlatedOs-Pb Petrologyof lavasfrom the Puu Oo eruptionof Kilaueavolcano, III, isotopes,Nature, 381,221-224, 1996. The Kupaianahaepisode (1986-1992), Bull. Volcanol.,58, 359-379, Brandon,A.D., R. J. Walker, J. W. Morgan,M.D. Norman,and H. M. 1996. Prichard,Coupled Os-186 and Os-187 evidencefor core-mantle Green, D. H., and M. E. Wallace, Mantle metasomatismby ephemeral interaction,Science, 280, 1570-1573, 1998. carbonatite melts, Nature, 336, 459-462, 1988. Campbell,I. H., The mantle'schemical structure: Insights from the Haase,K. M., P. Stoffers,and C. D. Garbe-Sch6nberg,The petrogenetic meltingof productsof mantleplumes, in The Earth'sMantle: evolution of lavas from Easter Island and neighbouringseamounts, Composition,Structure, and Evolution, edited by I. Jackson,pp. 259- near-ridgehotspot volcanoes in the SE Pacific,J. Petrol., 38, 785- 301, CambridgeUniv. Press,New York, 1998. 813, 1997. Castillo, P., The Dupal anomalyas a trace of the upwellinglower Halliday,A. N., D.-C. Lee, S. Tommasini,G. R. Davies,C. R. Paslick,J. mantle, Nature, 336, 667-670, 1988. G. Fitton, and D. E. James,Incompatible trace elementsin OIB and 25,490 KAMBER AND COLLERSON: ORIG1N OF OCEAN ISLAND BASALTS

MORB sourceenrichment in the sub-oceanicmantle, Earth Planet. paradoxes,forward transportmodelling, core formationand the Sci. Lett., 133, 379-395, 1995. historyof the continentalcrust, Chem. Geol., 139, 75-110, 1997. Hanan,B. B., and D. W. Graham,Lead and heliumisotope evidence Kramers,J. D., T. F. N/igler,and I. N. Tolstikhin,Perspectives from from oceanicbasalts for a commondeep source of mantleplumes, globalmodeling of terrestrialPb and Nd isotopeson the history of the Science, 272, 991-995, 1996. continentalcrust, Schweiz. Mineral. Petrogr. Mitt., 78, 169-174,1998. Hanyu, T., and I. Kaneoka,The uniform and low 3He/4He ratiosof Lay, T., Q. Williams,and E. J. Garnero,The core-mantleboundary HIMU basaltsas evidencefor their origin as recycledmaterials, layer anddeep Earth dynamics,Nature, 392, 461-468, 1998. Nature, 390, 273-276, t 997. le Roex,A. P., R. A. Cliff, andB. J. I. Adair, Tristanda Cunha,South Hanyu, T., and I. Kaneoka,Open systembehaviour of helium in caseof Atlantic:Geochemistry and petrogenesis of a basanite-phonolitelava the HIMU sourcearea, Geophys. Res. Lett., 25, 687-690, 1998. series,J. Petrol., 31,779-812, 1990. Hat'ds,V. L., P. D. Kempton,and R. N. Thompson,The heterogeneous Lee, D.-C., A. N. Halliday,J. G. Fitton,and G. Poli, Isotopicvariation Iceland plume: new insightsfrom the alkaline basaltsof the Snaefell with distance and time in the oceanic sector of the Cameroon line: volcaniccentre, J. Geol. Soc.London, 152, 1003-1009, 1995. Evidencefor a mantleplume origin and rejuvenationof magma Harte, B., M. T. Hutchinson, and J. W. Harris, Trace element transportpaths, Earth Planet. Sci.Lett., 123, 119-138, 1994. characteristicsof the lower mantle:an ion probestudy of inclusions Marty,B., I. Tolstikhin,I. L. Kamensky,V. Nivin, E. Balaganskaya,and in diamondfrom S5o Luiz, Brazil, Mineral. Mag., 58A, 386-387, J. L. Zimmermann,Plume-derived rare gases in 380 Ma carbonatites 1994. fromthe Kola region(Russia) and the argonisotopic composition in Hassler,D. R., and N. Shimizu,Osmium isotopic evidence for ancient the deepmantle, Earth Planet.Sci. Lett., 164, 179-192, 1998. subcontinentallithospheric mantle beneath the KerguelenIslands, McKenzie, D., and R. K. O'Nions, Melt productionbeneath oceanic southernIndian Ocean, Science,280, 418-421, 1998. islands,Phys. Earth Planet.Inter., 107, 143-182, 1998. Hauri,E. H., N. Shimizu,J. J. Dieu,and S. R. Hart,Evidence for hospot- Morgan,W. J., Convectionplumes in the lower mantle,Nature, 230, 42- related carbonatitemetasomatism in the oceanicupper mantle, 43, 1971. Nature, 365, 221-227, 1993. Mori, J., and D. V. Helmberger,Localized boundary layer belowthe H6mond,C., C. W. Devey, and C. Chauvel,Source composition and mid-Pacificvelocity anomaly identified from a PcP precursor,J. meltingprocesses in the Societyand Australplumes (South Pacific Geophys.Res., 100, 20,359-20,365, 1995. Ocean):Element and isotope(Sr, Nd, Pb, Th) geochemistry,Chem. N/igler,T. F., andJ. D. Kramers,Nd isotopicevolution of the upper Geol., 115, 7-45, 1994. mantleduring the Precambrian:Models, data and the uncertaintyof Hilton, D. R., J. Barling, and G. E. Wheller, Effect of shallow-level both, PrecambrianRes., 91,233-252. 1998. contaminationon the helium isotope systematicsof ocean-island O'Nions, R. K., and I. N. Tolstikhin, Limits on the mass flux between lavas, Nature, 373, 330-333, 1995. lowerand upper mantle and stability of layering,Earth Planet. Sci. Hoernle,K., andH.U. Schmincke,The role of partialmelting in the 15- Lett., 139, 213-222, 1995. Ma geochemical evolution of Gran Canaria: A blob model for the Ozima,M., andF. A. Podosek,Noble Gas Geochemistry,Cambridge Canary hotspot,J. Petrol, 34, 599-626, 1993. Univ. Press,New York, 1983. Hofmann,A. W., Mantle geochemistry:the messagefrom oceanic Palacz,Z. A., andA.D. Saunders,Coupled trace element and isotope volcanism,Nature, 385, 219-229, 1997. enrichmentin the Cook-Austral-Samoaislands, southwest Pacific, Honda,M., I. McDougall,and D. Patterson,Solar noble gases in the Earth Planet. Sci. Lett., 79, 270-280, 1986. Earth: The systematicsof helium-neonisotopes in mantlederived Pepin, R. O., Isotopicevidence for a solar argon componentin the samples,Lithos, 30, 257-265, 1993. Earth's mantle, Nature, 394, 664-667, 1998. Humphris,S. E., and G. Thompson,Geochemistry of rare earthelements Plank,T., andC. H. Langmuir,The chemicalcomposition of subducting in basaltsfrom the Walvis Ridge; implicationsfor its origin and sedimentand its consequencesfor the crustand mantle, Chem. Geol., evolution, Earth Planet. Sci. Lett., 66, 223-242, 1983. 145, 325-394, 1998. Ionov, D. A., W. L. Griffin, and S. Y. O'Reilly, Volatile-bearing Porcelli, D., and G. J. Wasserburg,Mass transferof helium, neon, minerals and lithophile trace elementsin the upper mantle, Chem. argon, and xenon througha steady-stateupper mantle, Geochim. Geol., 141, 153-184, 1997. Cosmochim.Acta, 59, 4921-4937, 1995. Ji, Y., and H.-C. Nataf, Detectionof mantle plumesin the lower mantle Rehk/imper,M., and A. W. Hofmann, Recycled ocean crust and by diffractiontomography: Hawaii, Earth Planet. Sci. Lett., 159, 99- sedimentin Indian Ocean MORB, Earth Planet. Sci. Lett., 147, 93- 115, 1998. 106, 1997. Johnson,K. E., A.M. Davis, and L. T. Bryndzia,Contrasting styles of Revenaugh,J., and R. Meyer, Seismicevidence of partialmelt withina hydrous metasomatismin the upper mantle: An ion microprobe possiblyubiquitous low-velocity layer at the base of the mantle, investigation,Geochim. Cosmochim. Acta, 60, 1367-1385, 1996. Science, 277, 670-673, 1997. Kamenetsky, V. S., S. M. Eggins, A. J. Crawford, D. H. Green, M. Reymer, A., and G. Schubert, Phanerozoic addition rates to the Gasparon,and T. J. Falloon,Calcic melt inclusionsin primitiveolivine continentalcrust and crustal growth, Tectonics, 3, 63-77, 1984. at 43øN MAR: Evidencefor melt-rockreaction/melting involving Reynolds,R. W., andD. J. Geist,Petrology of lavasfrom Sierra Negra clinopyroxene-richlithologies during melt generation,Earth Planet. volcano,Isabela Island, Galfipagos archipelago, J. Geophys.Res., Sci. Lett., 160, 115-132, 1998. 100, 24,537-24,553, 1995. Kato, T., A. E. Ringwood,and T. Irifune, Experimentaldetermination of Roden,M. F., T. Trull, S. R. Hart, andF. A. Frey,New He, Nd, Pb,and elementpartitioning between silicate perovskites, garnets and liquids: Sr isotopicconstraints on the constitutionof the Hawaiianplume: Constraintson early differentiation of the mantle, Earth Planet. Sci. Results from Koolau Volcano, Oahu, Hawaii, USA, Geochim. Lett., 89, 123-145, 1988. Cosmochim.Acta, 58, 1431-1440, 1994. Kellogg, L. H., B.H. Hager, and R. D. van der Hilst, Compositional Saal,A. E., S. R. Hart, N. Shimizu,E. H. Hauri,and G.D. Layne,Pb stratificationin the deepmantle, Science, 283, 1881-1884, 1999. isotopicvariability in melt inclusionsfrom oceanicisland basalts, Kesson,S. E., J. D. Fitz Gerald, and J. M. Shelley, Mineralogyand Polynesia,Science, 282, 1481-1484, 1998. dynamicsof a pyrolite lower mantle,Nature, 393, 252-255, 1998. Sasada,T., H. Hiyagon, K. Bell, and M. Ebihara,Mantle-derived noble Kogarko,L. N., C. M. B. Henderson,and H. Pacheco,Primary Ca-rich gasesin carbonatites,Geochim. Cosmochim. Acta, 61, 4219-4228, carbonatite magma and carbonate-silicate-sulphide liquid 1997. immiscibilityin the uppermantle, Contrib. Mineral. Petrol., 121,267- Schiano,P., R. Clocchiatti,N. Shimizu,D. Weis, and N. Mattinelli, 274, 1995. Cogeneticsilica-rich and carbonate-richmelts trappedin mantle Kogiso, T., Y. Tatsumi, G. Shumoda,and H. G. Barsczus,High g minerals in Kerguelen ultramafic xenoliths: Implications for (HIMU) ocean island basaltsin southernPolynesia: New evidence metasomatismin the oceanicupper mantle, Earth Planet. Sci. Lett., for whole mantle scale recycling of subductedoceanic crust, J. 123, 167-178, 1994. Geophys.Res., 102, 8085-8103, 1997. Schilling,J. G., Icelandmantle plume Geochemical study of Reykjanes Kokfelt, T., P.M. Holm, C. J. Hawkesworth, D. W. Peate, and C. Ridge,Nature, 242, 565-571, 1973. H6mond, OIB sourceregions stabilisedat the time of continental Shen, Y., S.C. Solomon,I. T. Bjarnason,and C. J. Wolfe, Seismic break-up:HIMU basaltsfrom the Cape Verde IslandsEos Trans. evidencefor a lower-mantleorigin of the Icelandplume, Nature, AGU, 78 (46), Fall Meet. Supp.,F814, 1997. 395, 62-65, 1998. Kramers, J. D., and I. N. Tolstikhin, Two terrestrial lead isotope Sp/ith,A., A. Le Roex,and R. A. Duncan,The geochemistryof lavas KAMBER AND COLLERSON:ORIGIN OF OCEANISLAND BASALTS 25,491

from the ComoresArchipelago, western Indian Ocean: Petrogenesis Wiechert, U., D. A. Ionov, and K. H. Wedepohl, Spinel peridotite and mantle source region characteristics,J. Petrol., 37, 961-991, xenolithsfrom the Atsagin-Dushvolcano, Dariganga lava plateau, 1996. Mongolia:A recordof partialmelting and crypticmetasomatism in Sun, S.-S., Lead isotopicstudy of youngvolcanic rocks from mid-ocean theupper mantle, Contrib. Mineral. Petrol., 126, 345-364,1997. ridges,ocean islands and islandarcs, Philos. Trans.R. Soc.London, Williams,Q., andE. J. Gamero,Seismic evidence for partialmelt at the 297, 409-445, 1980. base of Earth's mantle, Science,273, 1528-1530, 1996. Tolstikhin, I. N., Helium isotopesin the Earth's interior and in the Williams, Q., J. Revenaugh,and E. J. Garnero,A correlationbetween atmosphere:A degassingmodel for the Earth, Earth Planet. Sci. Lett., ultralowvelocities in the mantleand hotspots, Science, 281,546-549, 26, 88-96, 1975. 1998. Toyoda,K., H. Horiuchi,and M. Tokonami,Dupal anomalyof Brazilian Witt-Eickschen,G., W. Kaminsky,U. Kramm,and B. Harte, The nature carbonatites:Geochemical correlations with hotspotsin the South of youngvein metasomatismin the lithosphereof the West Eifel Atlantic and implicationsfor the mantle source,Earth Planet. Sci. (Germany):Geochemical and isotopicconstraints from composite Lett., 126, 315-331, 1994. mantle xenoliths from the Meerfelder Maar, J. Petrol., 39, 155-185, Turner, S., C. Hawkesworth, N. Rogers, and P. King, U-Th isotope 1998. disequilibriaand oceanisland basalt generation in the Azores,Chem. Woodhead, J. D., Extreme HIMU in an oceanic setting: The Geol., 139, 145-164, 1997. geochemistryof MangaiaIsland (Polynesia), and temporal evolution van der Hilst, R. D., S. Widiyantoro, and E. R. Engdahl,Evidence for of the Cook-Australhotspot, J. Volcanol.Geotherm. Res., 72, 1-19, deep mantle circulationfrom global tomography,Nature, 386, 578- 1996. 584, 1997. Woodhead, J. D., and C. W. Devey, Geochemistryof the Pitcairn Veksler, I. V., C. Petibon, G. A. Jenner, A.M. Dorfman, and D. B. seamounts,I, Source characterand temporal trends, Earth Planet. Dingwell, Trace elementpartitioning in immisciblesilicate-carbonate Sci. Lett., 116, 81-99, 1993. liquid systems:An initial experimentalstudy using a centrifuge Woodhead, J. D., and M. T. McCulloch, Ancient seafloor signals in autoclave, J. Petrol., 39, 2095-2104, 1998. Pitcairn island lavas and evidencefor large amplitude,small-scale Vidale, J. E., and M. A. H. Hedlin, Evidencefor partial melt at the core- mantleheterogeneities, Earth Planet. Sci. Lett., 94, 257-273,1989. mantleboundary north of Tonga from the strongscattering of seismic Xu, Y., J.-C. C. Mercier, M. A. Menzies,J. V. Ross,B. Harte, C. Lin, waves, Nature, 391,682-685, 1998. and L. Shi, K-rich glass-bearingwherlite xenolithsfrom Yitong, Walker, R. J., J. W. Morgan, and M. F. Horan, Osmium-187enrichment northeasternChina: Petrologicaland chemicalevidence for mantle in someplumes: evidence for core-mantleinteraction?, Science, 269, metasomatism,Contrib. Mineral. Petrol., 125, 406-420, 1996. 819-822, 1995. Yaxley, G. M., A. J. Crawford, and D. H. Green, Evidencefor Weaver, B. L., D. A. Wood, J. Tarney, and J.-L. Joron,Geochemistry of carbonatitemetasomatism in spinelperidotite xenoliths from western ocean island basalts from the South Atlantic: Ascension, Bouvet, St. Victoria, Australia,Earth Planet. Sci. Lett., 107, 305-317, 1991. Helena, Gough and Tristan da Cunha, in Alkaline Igneous Rocks, Zartman,R. E., andS. Haines,The plumbotectonicmodel for Pb isotopic editedby J. G. Fitton, and B. G. J. Upton, Geol. Soc. Spec.Publ., 30, systematicsamong major terrestrialreservoirs--A case for bi- 253-267, 1987. directionaltransport, Geochim. Cosmochim. Acta, 52, 1327-1339, West,H. B., M. O. Garcia,D.C. Gedach,and J. Romano,Geochemistry 1988. of tholeiitesfrom Lanai, Contrib.Mineral. Petrol.,112, 520-542, Zerr, A., A. Diegeler,and R. Boehler,Solidus of Earth'sdeep mantle, 1992. Science, 281,243-246, 1998. White,W. M., A. R. McBirney,and R. A. Duncan,Petrology and Zindler, A., and S. Hart, Chemical geodynamics,Annu. Rev. Earth geochemistryof the Galgtpagos Islands: Portrait of a pathological Planet. Sci., 14, 493-571, 1986. mantleplume, J. Geophys.Res., 98, 19533-19563,1993. Widom,E., andS. B. Shirey,Os isotopesystematics in the Azores: Implicationsfor mantle plume sources, Earth Planet. Sci. Lett., 142, 451-465, 1996. Widom,E., R. W. Carlson,J. B. Gill,and H.-U. Schmicke, Th-Sr-Nd-Pb K.D. Collerson and B.S. Kamber, Department of Earth Sciences, isotopeand trace element evidence for the origin of the S•o Miguel, Universityof Queensland,Steele Building, Brisbane, Queensland 4072, Azores,enriched mantle source, Chem. Geol., 140, 49-68, 1997. Australia. ([email protected];[email protected]. Widom,E., K. A. Hoernle,S. B. Shirey,and H. U. Schmincke,Os edu.au) isotopesystematics in the Canary Islands and Madeira: Lithospheric contaminationand mantle plume signatures, J. Petrol., 40, 279-296, (ReceivedNovember 23, 1998; revisedJune 14, 1999; 1999. acceptedJuly 23, 1999.)