Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/222315417

ReconsideringtheoriginsofisotopicvariationsinOceanIslandBasalts:Insightsfromfine-scalestudyofSãoJorgeIsland,Azoresarchipelago

ARTICLEinCHEMICALGEOLOGY·JULY2009

ImpactFactor:3.52·DOI:10.1016/j.chemgeo.2009.04.005

CITATIONS

17

READS

88

4AUTHORS:

Marc-AlbanMillet

DurhamUniversity

34PUBLICATIONS288CITATIONS

SEEPROFILE

RégisDoucelance

UniversitéBlaisePascal-Clermont-FerrandII

45PUBLICATIONS645CITATIONS

SEEPROFILE

JoelBaker

UniversityofAuckland

179PUBLICATIONS4,843CITATIONS

SEEPROFILE

P.Schiano

UniversitéBlaisePascal-Clermont-FerrandII

122PUBLICATIONS3,467CITATIONS

SEEPROFILE

Allin-textreferencesunderlinedinbluearelinkedtopublicationsonResearchGate,

lettingyouaccessandreadthemimmediately.

Availablefrom:P.Schiano

Retrievedon:05February2016

Chemical Geology 265 (2009) 289–302

Contents lists available at ScienceDirect

Chemical Geology

j ourna l homepage: www.e lsev ie r.com/ locate /chemgeo

Reconsidering the origins of isotopic variations in Ocean Island Basalts: Insights fromfine-scale study of São Jorge Island, Azores archipelago

Marc-Alban Millet a,⁎, Régis Doucelance a, Joel A. Baker b, Pierre Schiano a

a Laboratoire Magmas et Volcans, Université Blaise Pascal, OPGC, CNRS, IRD, 5, rue Kessler, 63038 Clermont-Ferrand Cedex, Franceb School of Geography, Environment and Earth Sciences, Victoria University of Wellington, PO Box 600, Wellington, New Zealand

Shallow-level interaction

⁎ Corresponding author. Present address: School ofEarth Sciences, Victoria University of Wellington, PO BoxTel.: +64 4 463 5391; fax: +64 4 463 5186.

E-mail address: marc-alban.millet@vuw.ac.nz (M.-A.

0009-2541/$ – see front matter © 2009 Elsevier B.V. Adoi:10.1016/j.chemgeo.2009.04.005

a b s t r a c t

a r t i c l e i n f o

Article history:Received 26 September 2008Received in revised form 30 March 2009Accepted 4 April 2009

Editor: R.L. Rudnick

Keywords:OIBIsotopeMantle plumeMixing trendsAzores archipelago

New major and trace element and Sr–Nd–Pb isotope data have been determined for 21 basaltic samples fromSão Jorge Island, Azores archipelago. Samples can be separated into two groups best identified in a plot of208Pb/204Pb versus 206Pb/204Pb where they form two sub-parallel mixing arrays. Lavas from the old (Topo)formation have lower 208Pb/204Pb for a given 206Pb/204Pb and more radiogenic Sr than samples fromintermediate (Rosais) and young (Manadas) formations. Topo samples also tend to have higher MgOcontents and lower incompatible trace element concentrations. Both Pb mixing arrays can be related tomixing of plume melts having a HIMU-like Pb isotope signature with two depleted components. Onedepleted component is best seen in analyses of Topo samples and is interpreted to represent upper mantlematerial from the nearby Mid-Atlantic ridge. The second depleted component has Sr–Nd–Pb isotopiccharacteristics similar to E-MORB and resides in the oceanic crust basement under São Jorge. Pb isotopeanalyses of lavas from São Jorge make it possible to re-define the Terceira end-member of the Azoresarchipelago, moving its composition to more radiogenic Pb with 206Pb/204Pb~20.51, 207Pb/204Pb~15.67 and208Pb/204Pb~39.56, at 143Nd/144Nd~0.51295 and 87Sr/86Sr~0.70375. Extending the interpretations made forthe origins of the São Jorge isotopic mixing arrays to other islands from the Azores archipelago (Pico, Faial,Terceira and São Miguel), we show that most of the isotopic variability recorded by Azores magmas can berelated to mixing of plume melts with two distinct and homogeneous signatures as well as several othercomponents dispersed in the shallow mantle/lithosphere under the Azores. This illustrates how caution isrequired when interpreting ocean island basalt isotopic data as reflecting their deep mantle plume source,and subsequently for constraining mantle topology.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

Extensive measurements of radiogenic isotope ratios (Sr, Nd, Pband, to a lesser extent, Hf and Os) in oceanic basalts have providedmajor constraints on mantle geodynamic models. Based on theassumption that these isotope ratios are representative of their deepmantle sources, studies of Ocean Island Basalts (OIB) and Mid-OceanRidge Basalts (MORB) have helped define the nature of mantle end-members and length-scales of mantle heterogeneities. On a globalscale, this has led to the identification of mantle end-members withextreme compositions termed DMM, EM1, EM2 and HIMU (Zindleret al., 1982; Allègre and Turcotte, 1985; Zindler and Hart, 1986; Allègreet al., 1986/1987). Apart from the Depleted MORB Mantle (DMM)component, all other end-members have been defined solely on thebasis of OIB data. EM1 has the isotopic composition with the mostextreme (highest) 87Sr/86Sr and lowest 206Pb/204Pb defined by analy-

Geography, Environment and600, Wellington, New Zealand.

Millet).

ll rights reserved.

ses of Pitcairn seamounts (Woodhead and Devey, 1993; Eisele et al.,2002), EM2 shows the most radiogenic Sr at intermediate Pbmeasured in samples from the Samoan volcanic chain (Jacksonet al., 2007a), and HIMU corresponds to samples with the mostradiogenic Pb analysed from Mangaïa Island (Woodhead, 1996). Anadditional end-member, corresponding to the intercept of convergingmixing trends in Sr–Nd–Pb isotope variation diagrams and thusdisplaying intermediate 87Sr/86Sr, 143Nd/144Nd and 206Pb/204Pb ratios,is also currently proposed (FOZO: Hart et al., 1992; Farley et al., 1992;Hauri et al., 1994; C: Hanan and Graham, 1996; Jackson et al., 2007b;PHEM). It also has high 3He/4He ratios and is thought to be commonto all oceanic basalts. If the existence of the four extreme end–members (DMM, EM1, EM2, HIMU) is now accepted (althoughthe nature of their origin and the existence of additional extremeend-members is still a matter of debate), the idea of a common end-member remains questionable. Indeed, mixing trends observed forOIB at local (island) and/or regional scales (archipelago) may eitherreflect different proportions of various components present inupwelling mantle plumes, or shallow-level interaction processesduring magma ascent. Consequently, deciphering the exact origin ofisotopic trends in OIB is required before isotopic variations can be

Fig. 1. Map of the Azores archipelago with the age of the oldest subaerial volcanicactivity on each island shown.

290 M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

interpreted as deep mantle heterogeneities occurring on a globalscale.

Shallow-level magma interactions can include several processesoccurring at different levels within the oceanic lithosphere. Fromsurface to source, the geochemistry of plume magmas can be alteredby post-eruption seawater and/or meteoric water interaction at thesurface or near-surface, by Assimilation and Fractional Crystallization(AFC) during storage in magma chambers at crustal depths, or viaPlume–Lithosphere Interactions (PLI) during melt ascent through thelocal lithosphere. Identification of isotopic variations related toshallow-level interactions in Comores (Späth et al., 1996; Class andGoldstein, 1997), Cape Verde archipelago (Doucelance et al., 2003;Escrig et al., 2005; Millet et al., 2008) and Canary Islands (Hoernleet al., 1991; Thirlwall et al., 1997; Widom et al., 1999) show that suchprocesses are common when volcanic activity is situated on very oldoceanic floor (N120 Ma).

The objective of this work is to examine potential shallow-levelmagma interactions that can be observed in a different geodynamicalsetting compared to Comores, Cape Verdes or Canary Islands. Thefocus of this paper is on São Jorge Island, Azores archipelago, locatedon young oceanic lithosphere and close to the mid-Atlantic ridge. We

Fig. 2. Detailed map of São Jorge Island annotated with sampling locations. The island can bvolcanic activity: (1) Topo: westernmost and oldest formation; (2) Rosais: easternmost andvolcanic activity.

present major and trace element data and Sr–Nd–Pb isotope ratios for21 basaltic samples. We first consider geochemical variations that areobserved at the scale of the island in an attempt to resolve informationrelated to deep source heterogeneities versus those due to shallow-level interactions. Then, we speculate about the implications of theSão Jorge Island dataset on the processes responsible for thegeochemical variability in other islands from the Azores archipelagoin order to: (a) assess the origin of the fine-scale variability of volcanicrocks from other Azores islands and (b) define the true isotopicfingerprint of the Azores mantle plume.

2. Geographic location and previous studies

The Azores archipelago (Fig. 1) comprises nine islands of volcanicorigin, representing the emerged part of a large oceanic plateau thatextends on both sides of the Mid-Atlantic Ridge (MAR). The islandscan be divided in three geographical groups: (1) the Occidental Group,located to the west of the MAR, that comprises Corvo and FloresIslands; (2) the Central Group, to the east of the MAR, formed byTerceira, Graciosa, Pico, Faial and São Jorge Islands; and (3) theOriental Group, to the east of the Central Group, with São Migueland Santa Maria. Subaerial volcanic activity started on Santa Maria~8 Myr ago (Abdel-Monem et al., 1975) and five islands have recordedhistorical eruptions, the last one occurring on Faial in 1957. Recenttomographic studies have proposed that the volcanic activity isassociated with the presence of a mantle plume starting from thecore-mantle boundary (Montelli et al., 2004; Montelli et al., 2006).Using a similar method, Yang et al. (2006) also observed a plumeshaped thermal anomaly in the upper mantle beneath the Azores.Nonetheless it should be noted that Van der Hilst and De Hoop(2005) have questioned the theoretical background of finite frequencytomography.

São Jorge is located in the middle of the Central Group, betweenTerceira (to the northeast) and Pico/Faial (to the south), and is ~200 kmeast of the MAR. The island has an elongated shape, inherited fromfissural volcanism, sub-parallel to the Terceira rift (located~20 kmnorth), and sub-perpendicular to the MAR. São Jorge comprises threegeological formations characterizing different periods of volcanicactivity (Fig. 2). The Topo formation crops out in the east of the island,

e divided into three different geological complexes corresponding to distinct periods ofintermediate complex; and (3) Manadas: central and site of the most recent (historical)

291M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

and is the oldest formation with published ages ranging from ~550 kyrto 1.32Myr (K–Armethod, Feraud et al., 1980; Hildenbrand et al., 2008).Subsequent periods of volcanic activity are represented by the Rosais(intermediate — western part of the island) and Manadas (young —

central part of the island) complexes, and date from few 100 kyrto historical volcanic activity (last eruption in 1808). The lack ofsystematic age progression precludes lava emplacement to be solelycontrolled by movement of the underlying tectonic plate toward theeast.

Large geochemical variations have been recorded at both intra-island and archipelago scales (White et al., 1976; Hawkesworth et al.,1979; Dupré et al., 1982; Davies et al., 1989; Widom and Shirey, 1996;Turner et al., 1997; Widom et al., 1997; Moreira et al., 1999; Schaeferet al., 2002; Widom and Farquhar, 2003; Madureira et al., 2005; Beieret al., 2007; Elliott et al., 2007; Turner et al., 2007). From an isotopicperspective, Azores basalts appear to have contributions from 4 end-members (Fig. 3):

- The first end-member (SM) corresponds to the most radiogenic Pbmeasured in samples from São Miguel Island. It plots to the left ofthe Northern Hemisphere Reference Line (NHRL; Hart, 1984), anddisplays enriched Sr–Nd isotope ratios. The origin of this end-member has been the debate of many studies with proposedmodels ranging from shallow, delaminated metasomatized litho-spheric mantle (Widom et al., 1997; 2nd scenario from Moreiraet al., 1999; Widom and Farquhar, 2003) to recycled materialintrinsic to the Azores mantle plume (Turner et al., 1997; 1stscenario from Moreira et al., 1999; Beier et al., 2007; Elliott et al.,2007). The most recent studies also showed that neither delami-nated SCLM nor sediments display the adequate isotopic signatureto account for this end-member.

- The second end-member (T), defined by Terceira samples, also hasradiogenic Pb, but falls to the right of the NHRL. Sr and Nd isotoperatios are intermediate with regard to the OIB global range ofvariations (87Sr/86Sr~0.7037; 143Nd/144Nd~0.51295). NegativeΔ7/4 recorded for Terceira end-member (Dupré et al., 1982;Turner et al., 1997; Moreira et al., 1999) is usually considered anindicator of a young HIMU component (Thirlwall, 1997). However,coupled high 3He/4He (~11.5 RA) and excesses in 20Ne and 21Necompared to MORB values (Moreira et al., 1999; Madureira et al.,2005) in Terceira samples are unlikely for such a component andsuggest that this end-member is a more complex componentrather than the signature of a single, young HIMU, end-member.

- Samples from São Miguel and Terceira Islands define two specifictrends that join at depleted values of Sr–Nd–Pb isotopes, defining athird end-member (D), common to all islands. In detail, this end-member plots in the field of localMORB (31–41°N, data fromDossoet al., 1999). It has consequently been associated with the upper

Fig. 3. 87Sr/86Sr versus 206Pb/204Pb diagram illustrating previously published data on Pico/FF) used to describe the range of isotopic variations measured in basalts from the Azores arc

mantle sampled along the MAR at the latitude of the Azoresarchipelago (Turner et al., 1997, Moreira et al., 1999).

- Faial lavas define another trend starting from the Terceira end-member toward less radiogenic Pb (to the left of the NHRL) andslightly enriched Sr–Nd isotopes. This defines a fourth component(F), whose nature has still to be addressed. However, recent studieshave shown that Faïal lavas are associated with very unradiogenicOs (187Os/188Os~0.11) as well as the heaviest B isotope data(Schaefer et al., 2002; Turner et al., 2007), arguing for the involve-ment of an old fluid-and-melt-depleted component.

Samples from Pico show strong similarities with Faial lavas (apartfrom a couple of basalts that display intermediate characteristicsbetween Terceira and Faial), indicating that the two trends associatedwith these islands might be genetically linked. Basalts from otherlocalities (Corvo, Flores, Graciosa, and Santa Maria) also plot atintermediate positions between Terceira and Faial trends. However,the relatively small number of analysed samples prevents a definitiveconclusion about their geographical grouping.

It should be noted that 3 of the 4 end-members described abovebelong to the Central Group of islands, and therefore are onlyseparated spatially by a length-scale of 40 to 100 km. This suggeststhat their marked differences in isotopic composition may representsmall-scale heterogeneity of the Azores mantle plume. Conversely,this could also reflect the presence of shallow contaminants in theupper mantle underneath the Azores archipelago.

3. Analytical procedures

All samples were coarsely crushed with a hydraulic press.Millimetre-sized grains were then hand-picked under a binocularmicroscope, rinsed with triply-distilled water in an ultrasonic bath,before being gently crushed with an agate mortar and pestle.

Major and trace element contents were determined at the Serviced'Analyses des Roches et des Minéraux (SARM) at CRPG Nancy. Powder-rock samples were fused along with lithium borate before beingdissolved in dilute nitric acid/hydrogen peroxide. Solutions were thenaliquoted for measurements on ICP-AES for major elements and ICP-MS for trace elements. A detailed description of the procedure can befound in Carignan et al. (2001). Additional information aboutmethodsand analytical reproducibility can be also found at http://helium.crpg.cnrs-nancy.fr/SARM/analyses/roches.html, duplicates measurementsfor 2 samples can be found in appendices).

Sr and Nd chemical separations and isotope measurements werecarried at the Laboratoire Magmas et Volcans in Clermont–Ferrand(LMV). Approximately 100 mg of sample powder was leached withhot 6 M HCl (5 mL, 100 °C) during 3 h following the proceduresdescribed in Millet et al. (2008), acid-digested with HF-HNO3, and

aial, Terceira and São Miguel islands, together with isotopic end-members (SM, T, D andhipelago.

292 M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

passed through the “cascade” column protocol (Sr Spec, TRU Spec andLn Spec columns) described in Pin and Bassin (1992) and Pin et al.(1994). Total procedural blanks (including leaching and HF-HNO3

digestion) were b0.5 ng and b0.2 ng for Sr and Nd, respectively. Allisotope measurements were made in static mode on a Finnigan Tritonthermal ionization mass spectrometer (TIMS), utilising the virtualamplifier and loading Sr and Nd onto double W filaments. Sr isotoperatios were mass-fractionation-corrected to 86Sr/88Sr=0.1194 andnormalized to 87Sr/86Sr=0.71025 for the NIST SRM987 standard. Ndisotope ratios were mass-fractionation-corrected to 146Nd/144Nd=0.7219 and normalized to 143Nd/144Nd=0.51196 for theRennes-AMES standard. Repeated analyses of the two standardsduring the course of the study gave 87Sr/86Sr=0.710250±15 (2σ,n=16) and 143Nd/144Nd=0.511961±6 (2σ, n=8), in agreementwith 2008 long-term reproducibility in this laboratory i.e., 87Sr/86Sr=0.710245±6 (2σ, n=38) and 143Nd/144Nd=0.511959±6 (2σ,n=27). Measurements of the LaJolla Nd standard during the course ofthis study gave 0.511845±3 (2σ, n=4). All sample duplicatesreproduced within these estimates of external reproducibility (seeappendices).

Pb isotope measurements were made on whole-rock samplepowders (~100 mg) at the Geochemistry Laboratory of VictoriaUniversity of Wellington, New Zealand. Chemical separation wascarried out in cleaned pipette tip columns filled with ~0.2 mL of AG1-X8 resin (100–200 mesh). Sample loading and elution of matrix wascarried out in 0.8 M HBr, and Pb collection was undertaken in 2 mL of7MHCl. Total procedural Pb blanks (including acid digestion)were ca.10 pg. Pb isotope measurements were performed on a Nu Plasma MC-ICP-MS in static mode. Mass discriminationwas corrected by standardbracketing of samples with NBS-981 using values of Baker et al.(2004). Accuracy and reproducibility of Pb isotope measurementswere assessed by repeated measurements of JB2 standard (n=8) andyielded 18.342±170 ppm, 15.561±250 ppm and 38.275±290 ppmfor 206Pb/204Pb, 207Pb/204Pb and 208Pb/204Pb, respectively. All sampleduplicates fall inside these uncertainties.

4. Results

4.1. Major and trace element variations

Samples from São Jorge that were analyzed during the course ofthis study all have MgON4 wt.%, with SiO2 ranging from 44 to 48 wt.%(Table 1). The samples are all subalkaline basalts in the total alkalis —silica nomenclature, apart from the three most differentiated samples,which are hawaiites and basanites. All samples also show low LOI (losson ignition) contents b1 wt.%, except for SJ01 (~2%) and SJ05 (4.5%).SJ05 is considered too weathered to truly represent its initialcomposition and is not considered further in this study.

Major element variations as a function of MgO content are shownin Fig. 4. Topo samples (older) are on average more primitive(MgON8 wt.%) than samples from the intermediate and recent Rosaisand Manadas complexes (primitive to slightly differentiated withMgO=4 to 11 wt.%). This difference reflects the continuous fractionalcrystallization of principally olivine and clinopyroxene, as illustratedby the CaO decrease and increase of Al2O3/CaO ratio with decreasingMgO. In addition, lower TiO2 contents for samples with MgOb5 wt.%suggest the onset of titanomagnetite fractionation.

Trace element patterns normalized to primitive mantle (Hofmann,1988) show similar patterns (Fig. 5) to HIMU basalts (average ofMangaia samples: Woodhead, 1996), most notably depletions in themost incompatible elements, negative anomalies in K and Pb, andmoderate enrichments in Nb, Ba and Y. Patterns of selected samplesfrom each geological formation show a progressive, time-related,enrichment in highly incompatible elements, as shown by strongvariations of (La/Yb)N=5.5–13, and to a lesser extent La/Sm and Rb/Sr. Such enrichments are not observed for moderately incompatible

elements, as exemplified by the (Gd/Yb)N ratio which is nearlyconstant in all samples (=2.5 to 3.0).

Variations of major element contents as well as ratios of both traceand major elements are coupled with Th contents (Fig. 6). Notably,MgO and Th co-variations (Fig. 6a) indicate that Th is a robustdifferentiation index. This also suggests that variations of (La/Yb)Nratios (Fig. 6d) are, in part, related to olivine and clinopyroxenefractionation. Variations related to extent of melting, although lessimportant, are also observed as shown by the range of Th content atconstant MgO.

4.2. Sr–Nd–Pb isotope variations

Pb isotope ratios measured in São Jorge basalts (Table 1) showmarked variations in 206Pb/204Pb (19.347 to 20.511) with ratherlimited 207Pb/204Pb and 208Pb/204Pb variations (15.626 to 15.673 and39.050 to 39.564, respectively). Nearly all samples plot to the right ofthe NHRL in both 208Pb/204Pb versus 206Pb/204Pb (Fig. 7a) and 207Pb/204Pb versus 206Pb/204Pb (Fig. 7b) diagrams, overlapping the Terceiratrend, but also extending it to significantlymore radiogenic Pb (20.511against 20.027 for the 206Pb/204Pb ratio). In detail, most samples lie ontwo sub-parallel trends (Rosais/Manadas complexes; Topo samples)in the 208Pb/204Pb versus 206Pb/204Pb diagram, with the Rosais andManadas samples having higher 208Pb/204Pb for a given 206Pb/204Pb.Such a clear difference between these two trends, however, is notobserved when considering the 207Pb/204Pb, suggesting the differencebetween the two trends is related to a component with recentlyfractionated for the Th/U ratios. Samples SJ01 (Topo formation) andSJ11 (Rosais complex) do not fall on the São Jorge trends in the 207Pb/204Pb versus 206Pb/204Pb diagram (Fig. 7b), suggesting a different andpossibly more complex history for these two magmas.

ThePb isotopedifference betweenTopoandRosais/Manadas samplescan also be observed in the 143Nd/144Nd versus 87Sr/86Sr diagram, inwhich they form two sub-perpendicular trends (Fig. 8). Older samples(Topo) have higher and rather constant Sr isotope ratios (0.703705 to0.703765)withvariable 143Nd/144Nd(0.512900 to0.512983;notemostofthis variation is largely due to sample SJ01 (that is also anomalous for itshigh 207Pb/204Pb ratio compared to all other samples, see above andFig. 7b) andSJ04. Conversely, younger lavas (Manadas andRosais) displaylower and more variable 87Sr/86Sr (0.703399 to 0.703638) at almostconstant 143Nd/144Nd (0.512914 to 0.512940).

It should also be noted that samples from the Manadas formationare characterized by lower 87Sr/86Sr than Rosais basalts, indicating atemporal decrease of the 87Sr/86Sr of magmas erupted at Sao Jorge.

4.3. Comparison with previous data

Three historical lavas from the Manadas complex (one samplefrom 1808 and two samples from 1580) were previously analyzed formajor and trace elements and Sr–Nd–Pb isotopes by Turner et al.(1997). Major and trace element data from this previous study are ingood agreement with our measurements; the three historicalManadas basalts (Turner et al., 1997) plot on the differentiationtrend defined by our samples and have similar concentrations to thoseof our samples from the 1580 and 1808 lava flows. However, com-parison between isotopic data of the two datasets is more proble-matic. While Sr isotope ratios measured for the three historicalManadas samples (Turner et al., 1997) are broadly consistent with ourSr isotope dataset for Manadas lavas, previously published 143Nd/144Nd values are highly variable (0.51283; 0.51284; 0.51297) com-pared to the range recorded by our samples (0.512926±14; 2σ,n=7). In the case of Pb isotopes, the two analyses (Turner et al., 1997)of samples from the 1580 eruption display significantly differentvalues as compared to one another and with our analysis (sampleSJ15). Similarly to these observations, Elliott et al. (2007) noted somediscrepancies between Pb isotope data for Pico and São Miguel

Table 1Major-, trace-elements and isotopes ratios for 21 samples of the São Jorge island.

Sample SJ01 SJ03 SJ04 SJ05 SJ07 SJ08 SJ50 SJ51 SJ52 SJ10 SJ11 SJ12 SJ13 SJ14 SJ09 SJ15 SJ55 SJ56 SJ57 SJ58 SJ60

Formation Topo Topo Topo Topo Topo Topo Topo Topo Topo Rosais Rosais Rosais Rosais Rosais Manadas Manadas Manadas Manadas Manadas Manadas Manadas

Latitude(°N)

38.54 38.54 38.57 38.57 38.57 38.56 38.57 38.57 38.57 38.72 38.72 38.68 38.68 38.68 38.63 38.68 38.68 38.69 38.67 38.67 38.71

Longitude(°E)

−27.78 −27.76 −27.80 −27.82 −27.82 −27.82 −27.86 −27.86 −27.82 −28.23 −28.24 −28.20 −28.20 −28.20 −27.98 −28.20 −28.05 −28.11 −28.12 −28.16 −28.17

SiO2 45.07 45.37 45.66 42.46 44.12 44.87 45.10 45.29 44.35 46.32 45.15 45.33 45.74 44.38 45.29 45.24 44.80 46.12 48.64 44.90 45.65TiO2 3.25 3.26 3.33 3.35 3.76 3.66 3.21 2.71 3.78 3.46 3.06 3.08 3.60 3.91 3.64 3.32 3.78 3.88 2.97 3.97 3.15Al2O3 15.53 14.50 13.94 15.08 14.43 15.04 13.90 12.46 14.53 16.69 13.64 14.79 15.33 14.97 15.50 15.81 16.10 16.54 16.94 16.29 13.79Fe2O3 11.58 11.87 11.51 13.38 12.83 12.90 11.88 12.41 12.86 10.30 12.56 12.69 12.46 13.25 12.65 12.23 12.68 13.81 11.36 13.82 12.25MnO 0.17 0.17 0.17 0.18 0.18 0.18 0.19 0.16 0.18 0.14 0.17 0.17 0.17 0.18 0.18 0.17 0.17 0.20 0.19 0.18 0.17MgO 8.24 8.58 9.99 8.95 8.45 7.97 8.84 11.99 8.20 6.27 10.52 8.72 6.81 7.35 7.12 6.97 6.61 5.35 4.53 6.81 10.89CaO 9.32 10.39 10.21 8.31 10.38 9.96 10.29 11.50 10.41 10.66 11.03 10.38 9.47 10.15 9.93 10.39 9.27 8.48 8.05 9.42 11.58Na2O 2.23 3.18 2.91 2.10 2.98 2.96 3.06 2.19 3.21 2.94 2.76 2.81 3.67 3.19 3.10 3.42 3.58 3.87 4.56 3.59 2.53K2O 1.16 1.03 1.00 0.57 1.04 0.90 1.23 0.46 1.08 1.07 0.84 0.85 1.22 1.07 0.91 0.87 1.24 1.46 1.83 1.23 0.80P2O5 0.55 0.52 0.53 0.49 0.56 0.46 0.60 0.26 0.56 0.59 0.41 0.45 0.77 0.60 0.69 0.59 0.71 0.95 0.95 0.72 0.43LOI 2.23 −0.31 −0.02 4.48 0.25 −0.15 0.32 −0.53 −0.05 0.33 −0.59 −0.34 −0.65 −0.45 0.08 −0.30 −0.36 −0.44 −0.68 −0.82 −0.67Total 99.31 98.56 99.23 99.34 98.97 98.75 98.61 98.90 99.10 98.76 99.55 98.93 98.58 98.60 99.09 98.70 98.57 100.22 99.33 100.10 100.55Rb 21 22 22 9 20 16 26 8 21 17 17 17 26 21 15 16 26 31 40 23 16Ba 329 270 282 272 270 239 313 134 267 278 235 228 315 301 302 234 329 370 416 301 205Th 3.8 3 3.2 3.1 2.5 2.4 3.7 1.3 2.6 3.4 2.5 2.6 4.33 3.3 3.5 2.3 3.7 3.9 5.8 3.3 2.4Nb 46 38 46 40 42 34 48 19 41 45 35 33 55 46 51 34 54 62 75 51 34U 1.26 1.05 1.08 1.00 1.00 0.91 1.31 0.52 0.98 1.33 0.85 0.84 1.55 1.15 1.30 0.81 1.30 1.35 1.93 1.14 0.82La 38 30 33 34 30 26 36 15 29 33 27 26 43 35 40 25 39 46 55 36 25Ce 80 65 69 59 65 56 75 33 65 73 59 57 93 75 8 57 84 98 116 79 56Pb 2.24 2.13 1.93 1.64 1.95 1.52 3.54 0.99 1.90 1.72 1.48 1.33 2.25 1.82 1.80 1.55 2.03 2.14 2.95 1.95 1.22Pr 9.7 8.3 8.7 8.9 8.4 7.5 9.3 4.6 8.5 9.3 7.4 7.4 11.7 9.5 10.6 7.4 10.7 12.5 14.3 10.2 7.3Nd 40 36 36 39 36 33 38 21 36 39 31 32 49 40 45 32 45 52 59 44 31Sr 507 587 567 464 657 737 601 418 661 639 598 474 749 715 667 570 769 834 824 786 560Sm 8.2 8.1 7.7 8.8 8.3 7.6 8.2 5.4 8.4 8.5 6.9 7.2 10.5 8.7 9.5 7.4 9.5 10.9 11.7 9.4 7.0Hf 6.08 5.49 5.73 6 5.82 4.97 5.87 3.62 5.94 6.55 4.58 5.09 7.07 5.49 7.14 4.85 6.84 7.34 8.38 6.10 5.11Zr 260 231 240 255 262 210 262 141 260 293 195 209 324 232 317 210 305 357 413 279 225Eu 2.63 2.63 2.50 2.89 2.83 2.53 2.53 1.91 2.75 2.84 2.26 2.35 3.33 2.82 3.13 2.52 3.10 3.46 3.59 3.08 2.28Gd 7.36 7.50 7.01 8.38 8.05 7.20 7.20 5.59 7.90 7.64 6.40 6.68 9.15 7.84 8.66 7.03 8.49 9.55 10.10 8.39 6.45Tb 1.04 1.10 1.01 1.26 1.15 1.04 1.05 0.84 1.15 1.08 0.93 0.96 1.28 1.10 1.23 1.01 1.22 1.32 1.40 1.18 0.94Dy 5.57 5.92 5.50 6.80 6.31 5.83 5.72 4.73 6.41 5.97 5.03 5.45 6.80 5.94 6.69 5.63 6.57 7.04 7.40 6.28 4.96Ho 0.99 1.08 0.98 1.21 1.14 1.04 1.04 0.86 1.12 1.05 0.91 0.96 1.21 1.08 1.18 1.00 1.16 1.23 1.31 1.10 0.90Er 2.62 2.80 2.63 3.18 2.95 2.68 2.68 2.16 2.87 2.68 2.32 2.45 3.05 2.71 3.06 2.57 2.95 3.10 3.37 2.78 2.28Y 28 29 27 35 33 28 27 24 31 29 26 25 33 28 33 27 32 37 39 32 26Tm 0.36 0.38 0.34 0.43 0.39 0.35 0.35 0.29 0.38 0.37 0.31 0.33 0.42 0.35 0.40 0.34 0.39 0.42 0.46 0.37 0.31Yb 2.28 2.37 2.15 2.63 2.41 2.18 2.30 1.77 2.44 2.31 1.92 2.08 2.57 2.20 2.52 2.19 2.47 2.62 2.90 2.36 1.90Lu 0.33 0.35 0.33 0.40 0.35 0.32 0.35 0.26 0.36 0.34 0.28 0.31 0.38 0.33 0.38 0.32 0.36 0.39 0.43 0.34 0.28Cu 44 37 45 34 40 34 44 46 46 31 54 45 28 30 28 28 33 22 19 23 24Co 44 45 47 48 52 46 44 61 48 45 58 48 40 46 44 43 44 39 31 47 59Ni 180 147 218 169 131 108 142 234 118 100 202 166 87 79 82 82 71 27 27 65 186Cr 380 300 442 334 274 206 323 617 254 228 377 198 173 171 111 168 80 20 25 89 42687Sr/86Sr 0.703743 0.703733 0.703705 0.703715 0.703752 0.703765 0.703761 0.703742 0.703745 0.703426 0.703638 0.703562 0.703606 0.703632 0.703399 0.703421 0.703437 0.703412 0.703435 0.703438 0.703504143Nd/144Nd 0.512900 0.512963 0.512924 0.512983 0.512959 0.512976 0.512956 0.512980 0.512955 0.512923 0.512914 0.512927 0.512927 0.512924 0.512922 0.512940 0.512925 0.512924 0.512921 0.512921 0.512930206Pb/204Pb 19.347 19.861 19.800 20.047 20.511 20.276 20.177 20.127 20.507 19.972 19.776 19.793 20.134 20.186 19.896 19.993 20.001 19.959 19.959 19.850 19.874207Pb/204Pb 15.626 15.631 15.626 15.627 15.673 15.653 15.647 15.639 15.667 15.636 15.643 15.621 15.638 15.654 15.624 15.637 15.637 15.632 15.632 15.627 15.628208Pb/204Pb 39.059 39.084 39.196 39.121 39.564 39.402 39.364 39.194 39.556 39.414 39.274 39.266 39.481 39.496 39.334 39.364 39.450 39.408 39.406 39.320 39.310

293M.-A

.Millet

etal./

Chemical

Geology

265(2009)

289–302

Fig. 4. Variations of major elements with MgO contents in São Jorge lavas. Systematic decrease of CaO together with increasing Al2O3/CaO indicate fractional crystallization of olivineand clinopyroxene. Samples from Topo complex are more primitive on average than Rosais and Manadas lavas.

294 M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

samples. Given these inconsistencies, we decided not to include anydata measured by Turner and co-workers on any island. Nevertheless,their interpretations and conclusions will be discussed on the basis ofthe remaining data from the literature.

5. Intra-island variability

In this section, we describe the variations delineated by São Jorgebasalts in isotope variation diagrams. Notably, we identify local end-members and discuss them with respect to previous isotopic studiesof the Azores archipelago.

5.1. São Jorge end-members

The two trends defined by São Jorge samples in isotope variationdiagrams, as well as the distinct composition of SJ01 and SJ11 (seeSection 4.2 and Figs. 7a,b and 8) require at least four end-members.Their proposed compositions are shown in Fig. 9 and will be discussedin more detail in the following section. However, it is important tonote that these defined compositions are arbitrary; they could bemore extreme than shown in Fig. 9, provided they are on the samemixing trends.

In the 207Pb/204Pb versus 206Pb/204Pb diagram, all São Jorgesamples with the exception of SJ01 and SJ11 plot on a well-definedmixing trend that overlaps the Terceira trend and extend it to moreradiogenic Pb (206Pb/204Pb~20.51 and 207Pb/204Pb~15.67), definingend-member 1 (Fig. 9b). This end-member has a negative Δ7/4 valueof about −4.7, and a 87Sr/86Sr ratio close to 0.70375 (Fig. 9c), and inother ocean island settings has been interpreted as reflecting a youngHIMU-like mantle component (Thirlwall, 1997).

In the 208Pb/204Pb versus 206Pb/204Pb diagrams (Fig. 9a), São Jorgebasalts define two distinct mixing trends, perhaps eminating fromend-member 1. A detailed look at Fig. 9c shows the lack of any trendfor Rosais and Manadas lavas, which are rather shifted toward lessradiogenic Sr isotope ratios. Hence, the pseudo-alignement of Rosaisand Manadas samples in Fig. 9a is rather the consequence of a smallvertical scatter than the signature of an independent mixing process.In detail, samples from Topo plot on a trend from end-member 1 toless radiogenic values of 87Sr/86Sr~0.70345, 206Pb/204Pb~19.25 and208Pb/204Pb~38.75, similar to those of local MORB (31–41°N, datafrom Dosso et al., 1999) and defining end-member 3. Samples fromRosais and Manadas complexes are intermediate between end-member 1 and less radiogenic 206Pb/204Pb~19.25 and 208Pb/204Pb ~ 39.35 values coupled to unradiogenic Sr (~0.70330),

Fig. 5. Abundances of trace elements in representative São Jorge lavas normalized to primitivemantle values (Hofmann,1988). Inset on the top-right of the figure shows a comparisonbetween the general pattern of São Jorge lavas and patterns for extreme OIB end-members EM1, EM2 and HIMU.

295M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

characteristic of end-member 2. Closer examination of Fig. 9c revealsthat samples from the Manadas and Rosais formations may actuallydiverge from various proportions of the Topo mixing line (end-members 1+3) toward end-member 2. Isotopic ratios determined forend-member 2 resemble those measured in a MORB sample collectedon the Mid-Atlantic Ridge at 43°N (Kamenetsky et al., 1998).

As noted above, samples SJ01 and SJ11 plot outside the São Jorgemixing trend in the 207Pb/204Pb versus 206Pb/204Pb diagram (Fig. 9b).These two basalts have Sr–Nd–Pb isotopic compositions falling in thefield of Faïal volcanics (to the left of the NHRL). These samples requirea further end-member 4, the composition of which is similar to theunradiogenic end of Faial trend. The influence of end-member 4 in SaoJorge is limited to samples SJ01 and SJ11 only, hence discussion aboutits nature will be addressed in the section dealing with isotopicvariations at the archipelago scale.

5.2. Geochemical evolution of São Jorge volcanism

The main temporal geochemical change of São Jorge Islandvolcanism is the difference observed between basalts from Topo andyounger samples from Manadas and Rosais formations. One couldargue that this difference reflects melting of another component inplume source, potentially a pyroxenitic or eclogitic lithology thatcould also generate the observed enrichment in (La/Yb)N ratio.However, a garnet bearing lithology (i.e. eclogite) can be rejected dueto the lack of variation of the Gd/Yb ratio in São Jorge lavas. Inaddition, although a pyroxenitic component certainly plays a role inthe generation of São Jorge lavas, variable extent of melting of suchcomponent in the plume cannot be invoked as the reason for theobserved variations. If so, the most radiogenic samples for Pb isotopes(i.e. the most HIMU-like) would be expected to have the highest(La/Yb)N ratio, whereas they are amongst the lowest and no strongcorrelation is observed between 206Pb/204Pb and (La/Yb)N. Never-theless, because Rosais and Manadas samples have isotopic compo-

sitions that differ from Topo lavas, it is likely that they are mixedwith a component that also display (La/Yb)N potentially derivingfrom a pyroxenitic source.

The difference between samples from those complexes is bestexpressed in the 208Pb/204Pb versus 206Pb/204Pb diagram (Fig. 9a)where these two populations of basalts broadly delineate two sub-parallel trends as samples from Manadas/Rosais have higher 208Pb/204Pb for a given 206Pb/204Pb than the older Topo basalts. One way toquantify such an offset between the two sample populations is tocompute the shortest distance between each sample and the best-fitline for Topo samples (slope=0.6517, intercept=26.17, r2=0.84,with the exception of sample SJ01 as it does not plot with other Toposamples to the right of the NHRL, but falls in the field of Faïal lavas)in the 208Pb/204Pb versus 206Pb/204Pb plot. This offset, designatedhere as ΔSJ, is a linear combination of two Pb isotope ratios and con-sequently is also behaving as a source tracer. It quantifies the scatteringof any São Jorge lavas around themain alignment formedby the samplesin this diagram. Slight changes of the reference line slope do not affectstrongly observed covariations.

Fig. 10 shows variations of ΔSJ with 143Nd/144Nd, MgO, Al2O3,Al2O3/CaO. Samples from the Manadas and Rosais formations havepositive ΔSJ, associated with low MgO and 143Nd/144Nd, and highAl2O3. Samples from both groups are aligned on the same trend inplots ofΔSJ against these chemical and isotopic data. This suggests thatisotopic and chemical variations in São Jorge basalts are resulting fromprocesses involving: (i) mixing (attested by the Pb–Nd isotope var-iations) with a component that has an isotopic composition similar toE-MORB (end-member 2, see Section 5.1), and (ii) magmaticdifferentiation (attested by the major elements contents and ratio)that are related to olivine and clinopyroxene crystallization. Allsamples collected along the island seem to be affected as even Toposamples show important range of variation along ΔSJ, only youngestsamples seem to be shifted toward higher values of this parameter.Such a combination of processes is typical of magma chamber

Fig. 7. (a) 208Pb/204Pb versus 206Pb/204Pb and (b) 207Pb/204Pb versus 206Pb/204Pbdiagrams for Azores basalts and Atlantic MORB collected at latitudes 31°–41°N. Symbolsfor São Jorge volcanics are as in Fig. 4 and bigger than errorbars. Samples from the Topocomplex, on the one hand, and Rosais–Manadas formations, on the other hand, definetwo sub-parallel trends in the 208Pb/204Pb versus 206Pb/204Pb diagram, whereas nodifference can be observed in the 207Pb/204Pb versus 206Pb/204Pb plot. Note that twosamples (SJ01 and SJ11) fall outside of the São Jorge trend in this last diagram.

Fig. 8. Sr–Nd isotope compositions of Azores samples. Symbols are as in Fig. 6 and biggerthan errorbars. Samples from Rosais and Manadas complexes show variable 87Sr/86Srratios at relatively constant 143Nd/144Nd values. Conversely, samples from Topo havevariable Nd isotope compositions at near constant Sr isotope ratios.

Fig. 6. Variations of (a) MgO, (b) CaO, (c) Al2O3/CaO and (d) (La/Yb)N with Th contentsfor São Jorge lavas. Systematic relationships that are observed indicate that fractiona-tion of olivine and clinopyroxene are the main source of variation of the (La/Yb)N ratio.Symbols as in Fig. 4. Normalized values for La and Yb are taken from Hofmann (1988).

296 M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

processes where assimilation of wall-rock material occurs duringfractional crystallization (AFC). As a consequence, the geochemicalvariations between the different formations of São Jorge Island are notinterpreted to be related to heterogeneity of the Azores mantle plume,but rather to AFC processes in the oceanic crust basement beneath SãoJorge. Moreover, they take place all along the volcanic history of SaoJorge island as samples from all volcanic complexes show covariationsof trace elements and isotopes ratios with the ΔSJ parameter.

It may seem somewhat paradoxical that the basement of São Jorge(N10 Myr; Cannat et al., 1999), which is interpreted in this study to

being assimilated, shows an isotopic composition similar to that of anE-MORB, which reflects the actual interaction between the Azoresmantle plume and the Mid-Atlantic Ridge. However, the Azoresoceanic plateau started to be created around 50 Myr ago (Searle,1980), suggesting that MORB with similar compositions have beenproduced in the area over long time making this the most probablelocally available crustal assimilant.

5.3. Origin of the depleted end-member involved in Topo lavas

As discussed above, samples from the Topo complex appear to beless influenced by crustal contamination. Because their scatter around

Fig. 9. Locations of the four end-members needed to account for the isotopic variabilityof São Jorge basalts in (a) 208Pb/204Pb versus 206Pb/204Pb, (b) 207Pb/204Pb versus 206Pb/204Pb, and (c) 87Sr/86Sr versus 206Pb/204Pb isotope variation diagrams. Symbols are asin preceding figures and bigger than errorbars. The large shaded triangle represents thecomposition of an E-MORB sample collected in the latitude of the Azores archipelago(data from Kamenetsky et al., 1998).

Fig. 10. Variations of major and trace element contents as well as major, trace elementand isotope ratios with the ΔSJ parameter (see text for details). Co-variations recordedbetween (a) 143Nd/144Nd, (b) MgO, (c) Al2O3, (d) (La/Yb)N and the ΔSJ parameterindicate that magmatic differentiation and mixing occurs simultaneously, mostlyaffecting Rosais and Manadas samples. The component responsible for isotopicvariations is therefore most likely assimilated during an AFC process.

297M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

the best-fit line is rather small, mixing lines defined by those samplesin the different isotope variation diagrams reflect a process occurringbefore interaction with the oceanic basement. We now consider theorigin of the depleted end-member 3.

If we assume end-member 3 to correspond to a single component,two distinctmodels can be envisaged for its location in themantle and itsnature. As a plume-supplied component, it could correspond to recycledoceanic lithosphere, aswasproposedby Schaefer et al. (2002) andTurneret al. (2007) to explain isotopic compositions of volcanics from FaialIsland. Conversely this component could have a shallow origin resultingfrom the interaction of the Azores mantle plume with the Mid-Atlanticridge. From an isotopic perspective, these two models lead to distinctcompositions. Recycled oceanic lithosphere should indeed displayslightly more depleted isotopic compositions than material from theAtlantic ridge as incompatible elements have been segregated out of theformer for at least a few hundred millions of years. In any case, thecomposition of the depleted end-member lies on the Topo mixing line.Extrapolation of this line toward less radiogenic Pb in the 208Pb/204Pb

versus 206Pb/204Pbdiagram(Fig. 9a) intersects thefieldofMORBsamplescollected in the North Azores Fracture Zone (data fromDosso et al.,1999)at 206Pb/204Pb~19.2–19.3. As theNorthAzores Fracture Zone is located inthe extension of São Jorge Island at a distance of ~200 km (Fig. 1), thisstrongly favours the interaction between the Azores mantle plume andthe Mid-Atlantic Ridge for the origin of the depleted end-member 3.

Fig.12.Mixing relationships between São Jorge–Terceira samples and Pico–Faial lavas inthe Δ8/4 versus 87Sr/86Sr diagram. Symbols are as in preceding figures. The Faialalignment intersects the São Jorge–Terceira trend at intermediate locations thusindicating a late occurrence in the mixing chronology. Mixing curves are calculatedassuming a low degree partial melting of oceanic depleted lithosphere (for São Jorge–Terceira samples) and delaminated continental lithosphere (for Pico–Faial lavas) with acomposition similar to that of Appalachian Tholeiites (Pegram, 1990). Tick marks onmixing curves are every 10% if not specified. See text for details of mixing models.

298 M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

6. Reconsidering the isotopic variability of Azores basalts

In previous sections, we have shown that the isotopic variability ofSão Jorge samples is essentially controlled by two mixing processes:(1) all samples result from mixing between a young HIMU-like end-member and a depleted component sampled at the North AzoresFracture Zone; (2) some samples, in particular, intermediate and youngvolcanics of Rosais and Manadas formations are furthermore progres-sively contaminated by the oceanic crustal basement that has acomposition similar to that of E-MORB. The second mixing processdoes, however, generate quite small isotope variations compared to therange of isotopic variations that are observed for the whole archipelagoand subordinate to the more important process of interaction betweenthe Azores mantle plume and the MAR. Therefore, in the following wefocus on the identification of (i) the signature of the MAR in Azoresbasalts and (ii) the genetic links between the three regional isotopictrends defined by Terceira/São Jorge, Pico/Faial and São Miguel basalts,respectively (see Section 2).

6.1. Influence of the Mid-Atlantic Ridge

Samples fromSãoMiguel and Terceira/São Jorge are represented in aΔ8/4 versus 206Pb/204Pb diagram in Fig. 11. Regression lines computedfor the two groups intersect in the field of MORB sampled at the NorthAzores Fracture Zone, in agreement with observation made by Moreiraet al. (1999) using a He–Pb isotope correlations as well as the He–Neisotope study of Madureira et al. (2005). This suggests that lavas fromthose two islands share a commondepleted component that is related tothe MAR and, therefore, not intrinsic to the Azores mantle plume.

Alternatively, Beier et al. (2007) have proposed that lavas from theSete Cidades complex of São Miguel, which are the most isotopicallydepleted from this island, were linked with a FOZO-like component(Hart et al., 1992; Stracke et al., 2005). They also reject an interactionwith the mid-Atlantic ridge, arguing that MORB have generally toolow Th/U, U/Pb and Th/Pb ratios. Samples from this portion of theMAR have not been measured for U, Th and Pb concentrations (Dossoet al., 1999). However, they show enrichments in highly incompatibleelements that suggest plausible elevated (U, Th)/Pb ratios. Moreover,FOZO-like components should have unradiogenic He isotope ratios,which is not observed in this case as the two alignments join at aMORB-like 3He/4He values (Moreira et al., 1999).

Consequently, we favour the involvement of a depleted componentrelated to the MAR in the genesis of Azores basalts. This component istherefore located at a shallow level. Nevertheless it has to be notedthat this does not necessarily imply an interaction with the ridge

Fig. 11. Regression trends for São Jorge–Terceira basalts, on the one hand, and São Miguellavas, on theotherhand, inaΔ8/4 versus 206Pb/204Pbdiagram. Symbols are as inprecedingfigures. Intersection of the two regression lines in the field of MORB collected in the NorthAzores Fracture Zone suggests that the common component involved in the genesis ofbasalts from these three islands is related to the Mid-Atlantic ridge.

system. Indeed, compositional similarities of the depleted end-member observed in Terceira, São Jorge and São Miguel lavas withMORB from the North Azores Fracture Zone (located in the middle ofthe Azores archipelago) is more indicative of a genetic link rather thanidentical components. Another alternative scenario would then beinteraction with the underlying oceanic lithosphere, created somemillions years ago along the MAR, during plume ascent.

Hyperbola corresponding to the mixing between a pristine OIB melt(with an isotopic composition equal to that of the São Jorge radiogenicPb end-member) and small-degree lithospheric melts have beencomputed in order to test this last model. Satisfactory results areobtained for an oceanic lithosphere mineralogy of ol:0.65; opx:0.15;cpx:0.15; sp:0.05 (melting mode: 0.05; 0.05; 0.45; 0.45) containing 7.7,0.6 and 0.01 ppm of Sr, Nd and Pb, assuming an accumulated fractionalmeltingprocesswith ameltingdegree of 0.3% (Fig.12).Most isotopicallydepleted samples require ~40% assimilation of those melts to accountfor their composition. Variations of the degree of melting only slightlyaffect the hyperbola curvature in such a way that melting degreesranging from 0.1 to 1% allow reproducing the dispersion of data pointsaround the plotted 0.3% model. Such model nevertheless raises thequestion of the ability of the local oceanic lithosphere to contaminateplume melts. Assuming that the base of the lithosphere is locatedaround 50 km depth (~1.5 GPa), its solidus should be around 1340 °C(Gudfinnsson and Presnall, 2000). As the temperature excess of theAzores plume is around 100 °C (Yang et al., 2006), the base of thelithosphere should be no colder than 1240 °C in order for it to melt inanhydrous conditions. However, Bourdon et al. (2005) have proposedthat the Azores local lithosphere contains small proportions of volatilesthat could help to reduce this crude estimate.

6.2. Genesis of the Pico/Faial alignment

Previous studies have shown that the two trends defined by Pico/Faial samples, on the one hand, and Terceira samples, on the otherhand, shared a common end-member with moderately elevated Pbisotope ratios (206Pb/204Pb~20). Our results significantly shift thiscommon end-member toward more radiogenic, HIMU-like, composi-tions. Doing so, it modifies the relative arrangement of those twoalignments and provides further constraints on the nature of end-member 4 (see Fig. 9 for location).

Representation of samples from Pico/Faial and Terceira/São Jorge ina Δ8/4 versus 87Sr/86Sr diagram (Fig. 12) best expresses the newinsights our data provides with respect to the isotopic variability of the

299M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

Azores mantle plume. Samples from the Pico/Faial group define a trendthat originates from the Terceira/São Jorge alignment at radiogenic Pbwith Δ8/4~−40 toward positive Δ8/4 values and more radiogenic Sr.This argues for the following chronological evolution: (1) mixing invariable proportions of end-members 1 and 3 accounts for variationsmeasured in Terceira/São Jorge samples; (2) followed by a secondmixing process that involves end-member 4 that generates composi-tions determined for Pico/Faial lavas. Consequently, the new data fromSão Jorge Island presented here suggests that the component respon-sible for slightly enriched values of Pico/Faial is located at a shallow leveland is not intrinsic to the Azores mantle plume because addition of theenriched component occurs after interaction of the Azores mantleplume with depleted upper mantle. This model is different to previousmodels that proposed recycling of a fluid- and melt-depleted oceaniclithosphere to explain the isotopic variations of the Azores basalts(Schaefer et al., 2002; Turner et al., 2007).

A possible candidate for the shallow end-member 4, is subconti-nental lithospheric mantle material delaminated during the openingof the Atlantic Ocean. Such material generally has positive Δ8/4 andΔ7/4 values, as well as slightly enriched Sr–Nd isotopic signatures,which are all characteristics of Pico/Faial basalts. Such a signature hasbeen identified in the signature of many oceanic basalts sampled inthe Atlantic Ocean such as Walvis Ridge (Richardson et al., 1982;Gibson et al., 2005), Fernando de Noronha (Gerlach et al., 1987), CapeVerde (Hoernle et al., 1991, Doucelance et al., 2003; Escrig et al., 2005),North Oceanographer Transform Fault (Shirey et al., 1987) andGodzilla seamount (Geldmacher et al., 2008).

The nearest subcontinental material from the Azores is locatedunder the Appalachian chain, Northern America. Tholeiites from thisregion display an isotopic composition (Pegram, 1990) that iscompatible with end-member 4 (Fig. 12). Bulk assimilation of suchmaterial during magma ascent can be modelled assuming Sr, Nd andPb contents of 10, 2 and 0.4 ppm, respectively (mixed into a pristineOIB melt containing 465, 38 and 1.45 ppm of those elements). Suchvalues are close to the ones proposed by Zartman andHaines (1988) andimply ~75% of SCLM assimilation, which seems unreasonably elevated.However, it is plausible that small degree melting of the SCLM bodyoccurs during assimilation. A first-order model to this process can beconstructed with an enriched mantle mineralogy of ol:0.55; opx:0.25;cpx:0.15; sp:0.05 (partition coefficients from Halliday et al. (1995) andreferences therein) containing 10 ppm Sr, 1.7 ppm Nd and 0.18 ppm Pb.Assuming accumulated fractional modal melting (melting mode: 0.05;0.05; 0.45; 0.45), a 0.5% partial melt would be sufficiently enriched togenerate an adequate mixing hyperbola (OIB end-member identical tothe bulk assimilation model) and only 4% mass fraction would beenough to account for themost enriched isotopic composition recordedin Faial samples.

Pico and Faial lavas are also characterized by very unradiogenic Osisotope ratios, and d11B and d18O values higher and lower, respectively,than MORB, as well as slightly elevated Nb/B ratios (Schaefer et al.,2002; Turner et al., 2007). While subcontinental mantle is known tohave unradiogenic Os isotopes (Shirey and Walker, 1998), the oxygenand boron signatures are more problematic. Indeed, oxygen isotoperatios of subcontinental mantle are supposed to be identical to that ofthe MORB mantle (Mattey et al., 1994) and no data have yet beenreported for B isotopes. Nevertheless, a low d18O signature has beenmeasured in a São Jorge sample (Turner et al., 2007) and it is also acharacteristic of HIMU basalts (Eiler et al., 1997). Consequently, it islikely that the lower than MORB oxygen isotope signature measured byTurner et al. (2007) is related to the São Jorgemantle source rather thanto end-member 4. Finally, Nb/B ratios as elevated as the ones measuredin Faial have been reported in samples fromFernando deNoronha (Ryanet al., 1996). Although these samples have not been measured forradiogenic isotope ratios, lavas from this locality have been interpretedas being contaminated by delaminated subcontinental material,suggesting a similar process is possible under Pico and Faial Islands.

Given the enriched isotopic signature is mostly restricted to Picoand Faial Islands (undetected in Terceira basalts), the volume of thefragment of delaminated subcontinental material (end-member 4)can be estimated. If we assume that Pico and Faial subaerial volcanicsrepresent a volume of 300 km3 and that all this volume iscontaminated at 4% (worst case scenario), then the SCLM fragmentneeds to be ~4200 km3. This represent a box of 60 km length, 35 kmwidth and only 2 km height that therefore would easily fit at thebottom of the lithosphere in between the islands of São Jorge, Faialand Pico.

7. The dual signature of the Azores mantle plume

In the previous sections, we have considered the relative importanceof isotopic variations measured in Azores basalts that are related to theplume source compared to that which we interpret to representshallow-level interactions. As a result of this filtering, from the four end-members needed at the archipelago scale, only the São Jorge (end-member 1) and São Miguel (end-member SM) radiogenic Pb isotopecompositions are considered to be intrinsic to the Azoresmantle plume.Such a dual plume composition is unusual, especially as no apparentmixing lines seem to join those two plume-related radiogenic Pb end-members. In the following, we model the isotopic and trace-elementcomposition of São Jorge sourcemelts in order to investigate this featureand speculate about on recycling processes.

Recent studies have proposed that the São Miguel enriched end-member comprises recycled enriched basaltic material (Elliott et al.,2007; Beier et al., 2007), thus implying that São Miguel basalts with themost radiogenic Pb are sampling a pure component rather than a mix ofphysically distinct lithologies in the Azores mantle plume. On thecontrary, the nature of the São Jorge source is slightly more puzzling. ItsPb isotopic composition appears to indicate the influence of a young,HIMU-like, component (Thirlwall,1997: i.e., radiogenic Pb, negativeΔ8/4and Δ7/4 values), yet its Sr–Nd isotope ratios do not support such anorigin. For example, HIMU OIB with 206Pb/204Pb ratios around 20.5 suchas those foundat StHelena Islanddisplay 87Sr/86Sr=0.7028, significantlylower than the value of ~0.7037 recorded in Topo lavas. Such a differencecannot be accounted for by assuming a more extreme Pb isotopes ratiosfor the São Jorge end-member as the slope of the São Jorge trend in theSr–Pb diagram is sub-parallel to the x-axis. Therefore, the radiogenic Pband moderate Sr isotope signature is a characteristic of the São Jorgesource rather than the result of a sampling bias.

HIMU sources in oceanic volcanism are often linked to the recyclingof altered oceanic crust (Weaver, 1991; Chauvel et al., 1992; Hofmann,1997; Stracke et al., 2003). Alteration is invoked to raise the U content ofthe oceanic crust, as well as its Sr isotopic ratio, as seawater ischaracterized by a more elevated 87Sr/86Sr than mantle. Hence onecould imagine an extremely altered oceanic crust to account for the SãoJorge unusual HIMU signature for Sr isotopes. Nevertheless, if such asignature was related to extensive oceanic crust alteration prior tosubduction, it would be expected that other HIMU sources woulddisplay such enrichments. To the contrary, OIB end-members with206Pb/204PbN20.5 usually display 87Sr/86Srb0.703 (Stracke et al., 2005).Therefore, the São Jorge source is unlikely to result from the recyclingand ageing of a single package of altered oceanic crust and mixing ofdifferent lithologies present in the plume has to be considered.

The elevated Pb isotope signature of São Jorge basalts, as well asthe lower than MORB d18O value recorded by Turner et al. (2007)argue for high proportions of recycled altered oceanic crust in theirsource. The São Miguel component, also intrinsic to the Azores plume,displays enriched Sr–Nd isotope signatures and radiogenic Pb.Consequently, we propose that the São Jorge end-member resultsfrom mixing between recycled altered oceanic crust and material atthe origin of the São Miguel end-member. Using worksheets providedby Stracke et al. (2003) in order to compute chemical parameters ofthe altered oceanic crust, we find that mixing of (1) an igneous crust

300 M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

that was composed of 30% N-MORB (value from Sun and McDonough,1989), 20% altered MORB (Staudigel et al., 1996) and 50% gabbro(sample 735B, Hart et al., 1999), and recycled ~1.85 Gyr ago with(2) the São Miguel component (as defined by Beier et al., 2007) in themass proportions 0.89–0.11 makes it possible to model the São Jorgesource fingerprint (see appendix for details). Assuming this result, wealso find that ~1.7% accumulated fractional melting in the garnetstability field is needed in order to reproduce the trace-elementpattern of the most primitive São Jorge basalt by partial melting.

Doing so, we have assumed a near-perfect solid-state mixingbetween the recycled oceanic crustal material and the São Miguelcomponent. However, the process by which the two components aremixed is unconstrained. The São Jorge source could either be afragment of recycled oceanic crust that was slightly enriched before itssubduction, much like models by Beier et al. (2007) and Elliott et al.(2007) for the São Miguel component, or the mixing could take placein the deep source of the plume or even during upwelling. Numericalmodels from Farnetani et al. (2002) would tend to dismiss the laterhypothesis hence promoting the idea of an early enrichment.

Turning to the heterogeneity of the Azores mantle plume, thisresult suggests that the influence of the São Miguel component mightbe more regional than previously thought. Moreover, this would alsoshow that the Azores mantle plume is an example of varying degreesof enrichment of recycled oceanic crust in OIB sources.

8. Summary and conclusions

New major and trace element and Sr–Nd–Pb isotope ratio mea-surements have been carried out for 21 samples from São Jorge Island.Isotopic results show the extension of the previously measured rangeof variations for Terceira Island towardmore radiogenic Pb ratios, withnegative Δ7/4 and Δ8/4 values, implying the influence of a HIMU-likecomponent in the source of the Azores mantle plume.

At the island scale, samples from the oldest complex (Topo) define amixing line between the HIMU-like component and a depleted end-memberwhose composition corresponds to that ofMORB samples fromthemid-Atlantic ridge (MAR) collected along the North Azores FractureZone (NAFZ; the nearest fracture zone from São Jorge). This mixing isinterpreted as resulting from the interaction of plume-derived meltswith a shallow depleted component. Intermediate and recent geologicalformations (Rosais andManadas) have amore complex history, derivingfromvarious proportions of the HIMU-NAFZmixing line toward an end-member with the composition of an E-MORB. The influence of this lastend-member is linked to olivine and clinopyroxenedifferentiation and istherefore thought to occur at oceanic crustal level and be generatedduring AFC processes.

At the archipelago scale, lavas from São Jorge Island plot on thesame alignment as Terceira samples. Best-fit lines for these twoislands and São Miguel basalts intercept in the field of NAFZ MORB,suggesting the influence of a MAR-related depleted component is aregional characteristic of the Azores archipelago. Such a conclusion isin agreement with the He–Pb isotope study of Moreira et al. (1999).Our new values also allow us to address the issue of the small-scaleheterogeneity of the Azores mantle plume, illustrated by thecontrasting isotopic signatures of Pico/Faial and Terceira Islands.Extension of the Terceira trend toward more radiogenic Pb valuesimply that the Pico/Faial alignment do not sample the very radiogenicPb but rather originate from a proportion ofmixing of the Terceira/SãoJorge trend. Our preferred interpretation for this trend is thecontamination of plume-derived melts by a subcontinental mantlefragment, delaminated during the opening of the Atlantic Ocean.Influence of this component is mostly restricted to Pico and Faialislands but affects also two samples from São Jorge, suggesting alength scale of the order of the tens of kilometres.

Those results necessitate serious reconsideration of the Azoresplume compositional heterogeneity. Indeed, from the four end-

members needed to explain the regional-scale variations, only twoof them are likely to be deep-seated in the plume (São Miguelenriched component and São Jorge HIMU-like end-member). Spec-ulation about the Sr isotopic ratios of São Jorge mantle source showsthat it cannot be explained by the single involvement of recycledoceanic crust. On the contrary, mixing of such component with 11% inmass of the São Miguel component is required to account for traceelements and isotope ratios.

Two general conclusions can be drawn from this study. First, thepresent dataset highlights the influence of shallow-level interaction asamajor agent of isotopic variability in OIB, evenwhen volcanic activityis occurring over young lithosphere such as in the case of the Azoresarchipelago. Coupled to previous studies where the oceanic litho-sphere is much older and thicker (Canary Islands: Hoernle et al., 1991;Comores: Späth et al., 1996; Class and Goldstein, 1997; Thirlwall et al.,1997; Widom et al., 1999; Cape Verde archipelago: Doucelance et al.,2003; Escrig et al., 2005; Millet et al., 2008), this illustrates howcaution is required when interpreting the isotopic variations mea-sured in OIB in the framework of chemical geodynamics. Second,modeling of the São Jorge end-member composition suggests apossible genetic link between the São Miguel enriched compositionand the source of basalts from other Azores islands. This implies thatthe São Miguel composition might not be the consequence of apeculiar set of recycling parameters. In our view, the two isotopicfingerprints of the Azoresmantle plumewould represent the recyclingof variously enriched oceanic crust material. This would suggest thatrecycling of OIB melts (Beier et al., 2007) or underplated basalts(Elliott et al., 2007) together with MORB-like crust could be of widerinfluence in the genesis of HIMU-like sources than hitherto thought.

Acknowledgements

The authors are much indebted to Manuel Moreira who providedthe samples. We also would like to thank Monica Handler and IvanVlastelic for insightful discussions as well as Chantal Bosq and AlexMcCoy-West who helped us in the clean lab. Finally, careful editing byRoberta Rudnick and comments by two anonymous reviewers greatlyimproved the quality of the manuscript.

Appendix: modeling of São Jorge radiogenic Pb end-member

Major- and trace-elements data for duplicate measurements of some São Jorge samples.

SJ 11

SJ 58 A

SiO2

44.80 45.50 45.65 44.15 TiO2 3.03 3.09 3.80 4.14 Al2O3 13.59 13.69 16.42 16.15 Fe2O3 12.45 12.68 13.04 14.60 MnO 0.17 0.17 0.18 0.18 MgO 10.44 10.60 6.56 7.05 CaO 10.99 11.06 9.40 9.44 Na2O 2.74 2.77 3.73 3.46 K2O 0.84 0.85 1.28 1.18 P2O5 0.41 0.41 0.76 0.68 LOI −0.52 −0.66 −0.76 −0.87 Total 98.94 100.15 100.05 100.15 Rb 17 18 25 22 Ba 227 244 319 283 Th 2.5 2.5 3.5 3.1 Nb 34 36 53 48 U 0.85 0.85 1.21 1.06 La 27 28 39 33 Ce 57 60 83 75 Pb 1.5 1.5 2.0 1.9 Pr 7.4 7.5 10.8 9.6 Nd 30 33 46 41 Sr 566 630 822 749 Sm 6.7 7.1 9.9 8.9 Hf 4.5 4.7 6.4 5.8

301M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

(continued)Appendix (continued)

SJ 11

SJ 58 A

Zr

181 209 296 262 Eu 2.2 2.3 3.2 2.9 Gd 6.3 6.5 8.7 8.1 Tb 0.91 0.95 1.23 1.14 Dy 4.9 5.1 6.6 6.0 Ho 0.89 0.92 1.15 1.05 Er 2.3 2.4 2.9 2.7 Y 24 28 33 30 Tm 0.30 0.32 0.39 0.36 Yb 1.9 2.0 2.5 2.3 Lu 0.27 0.28 0.35 0.33 Cu 50 59 25 21 Co 52 65 46 48 Ni 181 223 63 66 Cr 351 403 86 92

Sr–Nd–Pb isotopes ratios of duplicates measurements for some São Jorge lavas.

Sample

143Nd/144Nd 87Sr/86Sr 206Pb/204Pb 207Pb/204Pb 208Pb/204Pb

SJ 11

0.512914±7 0.703638±11 19.777±3 15.645±4 39.280±11 0.512913±7 0.703644±11 19.776±3 15.643±4 39.274±11

SJ 51

0.512990±12 0.703742±9 20.124±3 15.641±4 39.198±11 0.512971±12 0.703743±12 20.127±3 15.639±4 39.194±11

SJ 58

0.512921±8 0.703439±11 19.858±3 15.628±4 39.322±11 0.512923±8 0.703436±12 19.842±3 15.625±4 39.318±11

All calculations have been performed using two Excel spreadsheetsprovided by Stracke et al. (2003) with their paper, i.e. ‘‘basalt +sediment recycling.xls’’ for isotopes and ‘‘TE_OIBmelts.xls’’ for trace-element compositions, respectively. Both spreadsheets computes thegeochemical signature of recycled oceanic crust by a two-stage model.Modeling stages account for the evolution of the geochemicalcharacteristics before (stage 1) and after (stage 2) subduction zoneprocessing. Details are given in the following.

Trace-element contents of altered oceanic crust (before subduction)were calculated as the sum of 30% N-MORB (value from Sun andMcDonough, 1989), 20% altered MORB (Staudigel et al., 1996) and 50%Gabbro (sample 735B, Hart et al., 1999). Subduction processing wasaccounted for by removing elements according to mobility coefficients(values from Kogiso et al., 1997). Recycled crust (after subduction) wasthen mixed in mass proportions 0.05–0.95 with depleted mantle(composition identical to that proposed by Workman and Hart, 2005).

Turning to the isotopic fingerprint of the subducted crust, Pb ratioshave been computed using radioactive decay equations between 4.55and 1.85 Ga (recycling time) with μ1=8.15, κ1=3.00 and CanyonDiablo (Tatsumoto et al., 1973) as the initial composition (μ=238U/204Pb; κ=232Th/238U). Regarding Sr–Nd isotope ratios, it is assumedthat the oceanic crust was generated 2 Ga ago by a mantle source thatwould produce MORB samples with 87Sr/86Sr=0.7025 and 143Nd/144Nd=0.51335 as present-day ratios. Seaweater alterationwas takeninto account by exchanging 5% of Sr with a composition equal to0.7047 (that of seawater at the time of recycling).

Evolution of recycled crust after subduction processing is entirelycontrolled by mobilities of elements. Hence, in order to fit measureddata, all variables have to be adapted for the first stage of evolution.This is especially true for the 232Th/238U ratio that is set to a low valueof 3.00 in the first stage. Such a value fails to reproduce the present-day 206Pb/204Pb ratio of MORB. As a comparison, Chauvel et al. (1992)modeled the HIMU-like, Mangaia end-member with a κ1 value of 3.8.On the contrary, the post-subduction κ2 value used in this study is setto 3.57, reflecting more efficient extraction of U than Th in during slabdehydration, whereas Chauvel et al. (1992) decrease this value downto 3.2. Those considerations outline the variability of chemicalfractionation that can occur during recycling processes and justifyour lower-than-expected κ1 value in order to compensate the rigidityof the post-subduction stage modeling.

The trace-element pattern of melt corresponding to São Jorgeradiogenic Pb end-member is finally computed as the 1.7% accumu-lated fractional melting of the recycled oceanic crust+ambientmantle source, mixed with 11% in mass of the São Miguel component(data from Beier et al., 2007). Partial melting was thought to occur inthe garnet stability field (ol:0.5; opx:0.26; cpx:0.2; gt:0.04) with thefollowing melting mode: ol:0.1; opx:0.15; cpx:0.3; gt:0.45. Partitioncoefficients used for melting are those proposed by Stracke et al.(2003) in their spreadsheets. Trace-element contents of the mostprimitive sample of São Jorge Island were reproduced with a toleranceof ±20% for all elements measured in our samples apart from Pb(+24%), Eu (−31%) and Gd (−28%).

References

Abdel-Monem, A.A., Fernandez, L.A., Boone, G.M.,1975. K–Ar ages from the eastern Azoresgroup (Santa Maria, São Miguel and the Formigas Islands). Lithos 8 (4), 247–254.

Allègre, C.J., Hamelin, B., Provost, A., Dupré, B., 1986/1987. Topology in isotopicmultispace and origin of mantle chemical heterogeneities. Earth Planet. Sci. Lett. 81,319–337.

Allègre, C.J., Turcotte, D.L., 1985. Geodynamical mixing in the mesosphere boundarylayer and the origin of oceanic islands. Geophys. Res. Lett. 12, 207–210.

Baker, J., Peate, D., Waight, T., Meyzen, C., 2004. Pb isotopic analysis of standards andsamples using a 207Pb–s204Pb double spike and thallium to correct for mass biaswith a double-focusing MC-ICP-MS. Chem. Geol. 211, 275–303.

Beier, C., Stracke, A., Haase, K.M., 2007. The peculiar geochemical signatures of SãoMiguel (Azores) lavas: metasomatised or recycled mantle sources? Earth Planet.Sci. Lett. 259 (1–2), 186–199.

Bourdon, B., Turner, S.P., Ribe, N.M., 2005. Partial melting and upwelling rates beneaththe Azores from a U-series isotope perspective. Earth Planet. Sci. Lett. 239, 42–56.

Carignan, J., Hild, P., Mevelle, G., Morel, J., Yeghicheyan, D., 2001. Routine analyses oftrace element in geological samples using flow injection and low pressure on-lineliquid chromatography coupled to ICP-MS: a study of geochemical referencematerials BR, DR-N, UB-N, AN-G and GH. Geostand. Newsl. 25, 187–198.

Cannat, M., et al., 1999. Mid-Atlantic Ridge-Azores hotspot interactions: along-axismigration of a hotspot-derived event of enhancedmagmatism 10 to 4Ma ago. EarthPlanet. Sci. Lett. 173 (3), 257–269.

Chauvel, C., Hofmann, A.W., Vidal, P., 1992. HIMU-EM: the French Polynesianconnection. Earth Planet. Sci. Lett. 110, 99–119.

Class, C., Goldstein, S.L., 1997. Plume-lithosphere interactions in the ocean basins:constraints from the source mineralogy. Earth Planet. Sci. Lett. 150, 245–260.

Davies, G.R., Norry, M., Gerlach, D.C., Cliff, R.A., 1989. A combined chemical and Pb–Sr–Nd isotope study of the Azores and Cape Verde hotspots: the geodynamicimplications. In: A.D. Saunders and M.J. Norry (Editors), magmatism in the oceanbasins. Geol. Soc. Spec. Publ. Lond. 231–255.

Dosso, L., et al., 1999. The age and distribution of mantle heterogeneity along the Mid-Atlantic Ridge (31–41°N). Earth Planet. Sci. Lett. 170 (3), 269–286.

Doucelance, R., Escrig, S., Moreira, M., Gariépy, C., Kurz, M.D., 2003. Pb–Sr–He isotopeand traces element geochemistry of Cape Verde archipelago. Geochim. Cosmochim.Acta 67, 3717–3733.

Dupré, B., Lambert, B., Allègre, C.J., 1982. Isotopic variations within a single island : theTerceira case. Nature 620–622.

Eiler, J.M., et al., 1997. Oxygen isotope variations in ocean island basalt phenocrysts.Geochim. Cosmochim. Acta 61 (11), 2281–2293.

Eisele, J., et al., 2002. The role of sediment recycling in EM-1 inferred fromOs, Pb, Hf, Nd,Sr isotope and trace element systematics of the Pitcairn hotspot. Earth Planet. Sci.Lett. 196, 197–212.

Elliott, T., Blichert-Toft, J., Heumann, A., Koetsier, G., Forjaz, V., 2007. The origin of enrichedmantle beneath São Miguel, Azores. Geochim. Cosmochim. Acta 71 (1), 219–240.

Escrig, S., Doucelance, R., Moreira, M., Allègre, C.J., 2005. Os isotope systematics in Fogobasalts: evidence for lower continental crust residing in the oceanic lithospherebeneath the Cape Verde Islands. Chem. Geol. 219, 93–113.

Farley, K.A., Natland, J.H., Craig, H., 1992. Binary mixing of enriched and undegassed(primitive?) mantle components (He, Sr, Nd, Pb) in Samoan lavas. Earth Planet. Sci.Lett. 111, 183–199.

Farnetani, C.G., Legras, B., Tackley, P.J., 2002. Mixing and deformations inmantle plumes.Earth Planet. Sci. Lett. 196, 1–15.

Feraud, G., Kaneoka, I., Allègre, C.J., 1980. K/Ar ages and stress pattern in the Azores:geodynamic implications. Earth Planet. Sci. Lett. 46, 275–286.

Geldmacher, J., Hoernle, K., Klugel, A., van denBogaard, P., Bindeman, I., 2008.Geochemistryof a newenrichedmantle type locality in thenorthernhemisphere: implications for theorigin of the EM-I source. Earth Planet. Sci. Lett. 265 (1–2), 167–182.

Gerlach, D.C., Stormer Jr, J.C., Mueller, P.A., 1987. Isotopic geochemistry of Fernando deNoronha. Earth Planet. Sci. Lett. 85 (1–3), 129–144.

Gibson, S.A., Thompson, R.N., Day, J.A., Humphris, S.E., Dickin, A.P., 2005. Melt-generation processes associated with the Tristan mantle plume: constraints on theorigin of EM-1. Earth Planet. Sci. Lett. 237 (3–4), 744–767.

Gudfinnsson, G.H., Presnall, D.C., 2000. Melting behaviour of model lherzolite in thesystem CaO–MgO–Al2O3–SiO2–FeO at 0.7–2.8 GPa. J. Petrol. 41 (8), 1241–1269.

Halliday, A.N., Lee, D.C., Tommasini, S., Davies, G.R., Paslick, C.R., Fitton, J.G., James, D.E.,1995. Incompatible trace elements in OIB and MORB and source enrichments in thesub-oceanic mantle. Earth Planet. Sci. Lett. 133, 379–395.

302 M.-A. Millet et al. / Chemical Geology 265 (2009) 289–302

Hanan, B.B., Graham, D.W.,1996. Lead and helium isotope evidence from oceanic basaltsfor a common deep source of mantle plumes. Science 272 (5264), 991–995.

Hart, S.R., 1984. Large-scale isotope anomaly in the Southern Hemisphere mantle.Nature 309, 753–757.

Hart, S.R., Blusztajn, J., Dick, H.J.B., Meyer, P.S., Muehlenbachs, K., 1999. The fingerprint ofseawater circulation in a 500-meter section of ocean crust gabbros. Geochim.Cosmochim. Acta 63 (23–24), 4059–4080.

Hart, S.R., Hauri, E.H., Oschmann, L.A., Whitehead, J.A., 1992. Mantle plumes andentrainment: isotopic evidence. Science 256, 517–520.

Hauri, E.H., Whitehead, J.A., Hart, S.R., 1994. Fluid dynamic and geochemical aspects ofentrainment in mantle plumes. J. Geophys. Res. 99, 24275–24300.

Hawkesworth, C.J., Norry, M.J., Roddick, J.C., Vollmer, R.,1979. 143Nd/144Nd and 87Sr/86Srratios from the Azores and their significance in LIL-element enriched mantle.Nature 280, 28–31.

Hildenbrand, A., Madureira, P., OrnelaMarques, F., Cruz, I., Henry, B., Silva, P., 2008. Multistage evolution of a sub-aerial volcanic ridge over the last 1.3 Myr: S. Jorge Island,Azores Triple Junction. Earth Planet. Sci. Lett. 273 (3–4), 289–298.

Hoernle, K., Tilton, G., Schmincke, H.-U., 1991. Sr–Nd–Pb isotopic evolution of GranCanaria: evidence for shallow enriched mantle beneath the Canary Islands. EarthPlanet. Sci. Lett. 106, 44–63.

Hofmann, A.W., 1988. Chemical differentiation of the Earth: the relationship betweenmantle, continental crust and oceanic crust. Earth Planet. Sci. Lett. 90, 297–314.

Hofmann, A.W., 1997. Mantle geochemistry: the message from oceanic volcanism.Nature 385, 219–229.

Jackson, M.G., et al., 2007a. The return of subducted continental crust in Samoan lavas.Nature 448 (7154), 684–687.

Jackson, M.G., Kurz, M.D., Hart, S.R., Workman, R.K., 2007b. New Samoan lavas from OfuIsland reveal a hemispherically heterogeneous high 3He/4He mantle. Earth Planet.Sci. Lett. 264 (3–4), 360–374.

Kamenetsky, V.S., et al., 1998. Calcic melt inclusions in primitive olivine at 43∞N MAR:evidence for melt-rock reaction/melting involving clinopyroxene-rich lithologiesduring MORB generation. Earth Planet. Sci. Lett. 160 (1–2), 115–132.

Kogiso, T., Tatsumi, Y., Nakano, S., 1997. Trace element transport during dehydrationprocesses in the subducted oceanic crust: 1. Experiments and implications for theorigin of ocean island basalts. Earth Planet. Sci. Lett. 148 (1–2), 193–205.

Madureira, P., Moreira, M., Mata, J., Allègre, C.J., 2005. Primitive neon isotopes in TerceiraIsland (Azores archipelago). Earth Planet. Sci. Lett. 233 (3–4), 429–440.

Mattey, D., Lowry, D., Macpherson, C., 1994. Oxygen isotope composition of mantleperidotite. Earth Planet. Sci. Lett. 128 (3–4), 231–241.

Millet, M.-A., Doucelance, R., Schiano, P., David, K., Bosq, C., 2008. Mantle plumeheterogeneity versus shallow-level interactions: a case study, the São NicolauIsland, Cape Verde archipelago. J. Volcanol. Geotherm. Res. 176 (2), 265–276.

Montelli, R., et al., 2004. Finite-frequency tomography reveals a variety of plumes in themantle. Science 303 (5656), 338–343.

Montelli, R., Nolet, G., Dahlen, F.A., Masters, G., 2006. A catalogue of deep mantleplumes: new results from finite-frequency tomography. Geochem. Geophys.Geosyst. 7, Q11007. doi:10.1029/2006GC001248.

Moreira, M., Doucelance, R., Kurz, M.D., Dupré, B., Allègre, C.J., 1999. Helium and leadisotope geochemistry of the Azores Archipelago. Earth Planet. Sci. Lett. 169,189–205.

Pegram, W., 1990. Development of continental lithospheric mantle as reflected in thechemistry of the Mesozoic Appalachian Tholeites. Earth Planet. Sci. Lett. 97, 316–331.

Pin, C., Bassin, C., 1992. Evaluation of a strontium-specific extraction chromatographicmethod for isotopic analysis in geological materials. Anal. Chim. Acta 269, 249–255.

Pin, C., Briot, D., Bassin, C., Poitrasson, F., 1994. Concomitant separation of strontium andsamarium-neodymium for isotopic analysis in silicate samples, based on specificextraction chromatography. Anal. Chim. Acta 298 (2), 209–217.

Richardson, S.H., Erlank, A.J., Duncan, A.R., Reid, D.L., 1982. Correlated Nd, Sr and Pbisotope variation in Walvis Ridge basalts and implications for the evolution of theirmantle source. Earth Planet. Sci. Lett. 59, 327–342.

Ryan, J.G., Leeman, W.P., Morris, J.D., Langmuir, C.H., 1996. The boron systematics ofintraplate lavas: implications for crust andmantle evolution. Geochim. Cosmochim.Acta 60 (3), 415–422.

Schaefer, B.F., Turner, S., Parkinson, I., Rogers, N., Hawkesworth, C., 2002. Evidence forrecycled Archaean oceanic mantle lithosphere in the Azores plume. Nature 420,304–307.

Searle, R., 1980. Tectonic pattern of the Azores spreading centre and triple junction.Earth Planet. Sci. Lett. 51 (2), 415–434.

Shirey, S.B., Bender, J.F., Langmuir, C.H., 1987. Three-component isotopic heterogeneitynear the oceanographer transform, Mid-Atlantic ridge. Nature 325, 217–223.

Shirey, S.B., Walker, R.J., 1998. The Re-Os isotope system in cosmochemistry and high-temperature geochemistry. Annu. Rev. Earth Planet. Sci. 26, 423–500.

Späth, A., Roex, A.P.L., Duncan, R.A., 1996. The geochemistry of lavas from the Comoresarchipelago, Western Indian ocean: petrogenesis and mantle source regioncharacteristics. J. Petrol. 37, 961–991.

Staudigel, H., Plank, T.,White,W.M., Schmincke, H.U.,1996. Geochemical fluxes duringseafloor alteration of the basaltic upper crust: DSDP sites 417 and 418. In: Bebout,D.W.S.G.E., Kirby, S.H., Platt, J.P. (Eds.), Subduction: Top to Bottom. Geophys. Monogr.Ser. AGU, Washington, D.C., pp. 19–38.

Stracke, A., Bizimis, M., Salters, V.J.M., 2003. Recycling oceanic crust: quantitativeconstraints. Geochem. Geophys. Geosys. 4.

Stracke, A., Hofmann, A.W., Hart, S.R., 2005. FOZO, HIMU and the rest of the mantle zoo.Geochem. Geophys. Geosys. 6.

Sun, S.s., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts:implications for mantle composition and processes. Special Publications, vol. 42(1).Geological Society, London, pp. 313–345.

Tatsumoto, M., Knight, R.J., Allegre, C.J., 1973. Time differences in the formation ofmeteorites as determined from the ratio of lead-207 to lead-206. Science 180 (4092),1279–1283.

Thirlwall, M.F., 1997. Pb isotopic and elemental evidence for OIB derivation from youngHIMU mantle. Chem. Geol. 139 (1–4), 51–74.

Thirlwall, M.F., Jenkins, C., Vroon, P.Z., Mattey, D.P., 1997. Crustal interaction duringconstruction of ocean islands: Pb–Sr–Nd–O isotope geochemistry of the shieldbasalts of Gran Canaria, Canary Islands. Chem. Geol. 135 (3–4), 233–262.

Turner, S., Hawkesworth, C., Rogers, N., King, P., 1997. U–Th isotope disequilibria andocean island basalt generation in the Azores. Chem. Geol. 139 (1–4), 145–164.

Turner, S., Tonarini, S., Bindeman, I., Leeman,W.P., Schaefer, B.F., 2007. Boron and oxygenisotope evidence for recycling of subducted components over the past 2.5 Gyr.Nature 447, 702–705.

Van der Hilst, R.D., De Hoop, M.V., 2005. Banana-doughnut kernels and mantletomography. Geophys. J. Int. 163, 956–961.

Weaver, B.L., 1991. The origin of ocean island basalt end-member compositions: traceelement and isotopic constraints. Earth Planet. Sci. Lett. 104 (2–4), 381–397.

White, W.M., Tapia, M.D.M., Schilling, J.-G., 1976. The petrology and geochemistry of theAzores Islands. Contrib. Mineral. Petrol. 69, 201–213.

Widom, E., Carlson, R.W., Gill, J.B., Schmincke, H.-U., 1997. Th–Sr–Nd–Pb isotope andtrace element evidence for the origin of the Sao Miguel, Azores, enriched mantlesource. Chem. Geol. 140 (1), 49–68.

Widom, E., Farquhar, J., 2003. Oxygen isotope signatures in olivines from São Miguel(Azores) basalts: implications for crustal andmantle processes. Chem.Geol.193 (3–4),237–255.

Widom, E., Hoernle, K.A., Shirey, S.B., Schmincke, H.-U., 1999. Os isotope systematics inthe Canary Islands and Madeira: lithospheric contamination and mantle plumesignatures. J. Petrol. 40, 279–296.

Widom, E., Shirey, S.B., 1996. Os isotope systematics in the Azores: implications formantle plume sources. Earth Planet. Sci. Lett. 142 (3–4), 451–465.

Woodhead, J.D., 1996. HIMU in an oceanic setting: the geochemistry of Mangaia Island(Polynesia) and temporal evolution of the Cook–Austral hotspot. J. Volcanol.Geotherm. Res. 72, 1–19.

Woodhead, J.D., Devey, C.W., 1993. Geochemistry of the Pitcairn seamount, 1: sourcecharacter and temporal trends. Earth Planet. Sci. Lett. 116, 81–99.

Workman, R.K., Hart, S.R., 2005. Major and trace element composition of the depletedMORB mantle (DMM). Earth Planet. Sci. Lett. 231 (1–2), 53–72.

Yang, T., Shen, Y., van der Lee, S., Solomon, S.C., Hung, S.-H., 2006. Upper mantlestructure beneath the Azores hotspot from finite-frequency seismic tomography.Earth Planet. Sci. Lett. 250, 11–26.

Zartman, R.E., Haines, S.M., 1988. The plumbotectonic model for Pb isotopic systematicsamong major terrestrial reservoirs—a case for bi-directional transport. Geochim.Cosmochim. Acta 52 (6), 1327–1339.

Zindler, A., Hart, S.R., 1986. Chemical geodynamics. Ann. Rev. Earth Planet. Sci. Lett. 14,493–571.

Zindler, A., Jagoutz, E., Goldstein, S.L., 1982. Nd, Sr, and Pb isotopic systematics in a threecomponent mantle: a new perspective. Nature 298, 519–523.