Introduction

Since the chemical variability of ocean island basalts (OIB) was described in the 1980s, it has become clear that the Earth's mantle is heterogeneous on small and large scales. In order to explain the isotopic variability observed in OIB, several distinct mantle "end-members" have been proposed (e.g. DM, HIMU, EM1, EM2, FOZO and C) [White, 1985; Zindler & Hart, 1986]. The term "end-member" (or "component") implies a mantle reservoir or reservoirs with specific isotopic composition, and implies two things:

1. Fractionation of the radioactive parent/daughter ratio for different isotopic systems (e.g. U/Pb, Th/Pb, Sm/Nd, Rb/Sr, Re/Os etc.). Because of the extremely slow diffusion rate in the deep mantle [Hofmann & Hart, 1978], this fractionation can happen only at shallow depth, i.e., in the upper mantle or the crust.
2. Isolation of the "reservoir" for a long period of time (1- 2 Ga). This is necessary in order for the different isotopic systems to evolve to the specific signatures observed in the "end members".

In addition, basalt generation from the mantle source implies:

1. The source material must yield, by partial melting, the major- and trace-element chemical composition observed in OIB. This limits possible parent/daughter isotope-ratio fractionation in the source and, as a result, the possible isolation time interval required to generate the different isotopic signatures.
2. Pure end-member compositions are practically never observed in oceanic islands, suggesting that several end-members must be present in the source in order to generate the isotopic composition range observed in the OIB of a single island.

1. Hypotheses for the origin of OIB

1.1 Ancient recycled subducted oceanic crust model

The proposal made more than a decade ago, that the variation of trace element and isotopic ratios in OIB is related to the recycling of ancient oceanic crust [Hofmann & White, 1982] and pelagic or terrigenous sediment assimilation [Weaver, 1991, Chauvel et al., 1992], is now widely accepted. This model is popular because:

1. The recycling of oceanic crust through the mantle is a undeniable fact;
2. The generation of arc volcanism from the subduction and dehydration of oceanic crust suggests major fractionation of trace elements in the residual subducted oceanic crust;
3. Positive Nb and negative Pb anomalies are observed both in OIB and in MORB. This suggests a link in the generation of these two types of basalt;
4. The addition of pelagic and terrigenous sediments to subducted oceanic crust could explain the chemical and isotopic differences observed between different end-members signatures.

Nevertheless, several chemical and physical arguments suggest that subducted oceanic crust is not the source of OIB [e.g. Sun & McDonough, 1989; Halliday et al., 1995. Niu & O'Hara [2003] state, provocatively:

"There is no obvious association between ancient recycled oceanic crust and OIB (mantle plumes) in terms of petrology, geochemistry, and mineral physics".

This claim is based on the following arguments:

1. Melting of oceanic crust with basaltic or picritic composition cannot produce the high-magnesian lavas that are parental to most OIB. Primitive OIB melts (>15% MgO) are probably more magnesian than bulk oceanic crust (<13% MgO).
2. Oceanic crusts subjected to subduction zone dehydration reactions must be depleted in water-soluble incompatible elements such as Ba, Rb, Cs, Th, U, K, Sr and Pb relative to water-insoluble incompatible elements such as Nb, Ta, Zr, Hf and Ti, etc. Residual crust with such trace-element systematics is unlikely to be the fertile source of OIB.
3. Ancient oceanic crust is isotopically too depleted to produce the isotopic signatures of most OIB. OIB Sr-Nd-Hf isotopes preserve no signature that indicates previous subduction zone dehydration histories.
4. Subducted oceanic crust at shallow lower-mantle conditions form mineral assemblages that are much (>2.3%) denser than the ambient peridotitic mantle. If subducted crust melts in the deep lower mantle, this melt, depending upon its composition, may have still greater (up to 15%) density than solid peridotitic mantle. Therefore ancient oceanic crust that has subducted into the deep lower mantle cannot return in bulk to the upper mantle in either a solid or molten state.

1.2 The problem of major element composition

In addition to the problems highlighted by Niu & O'Hara [2003], the question of the silica-deficient composition of OIB was pointed out many years ago [Green et al., 1967]. Many ocean island rocks are characterized by silica undersaturated compositions (low silica and high alkali contents), whereas the partial melts of oceanic-crust lithologies produce silica saturated liquids [Green et al., 1967]. Partial melting of a mixture of subducted crust and peridotite produced by mechanical desegregation of subducted eclogite [Kogiso et al., 1998], or by hybridization of peridotite by partial melts of eclogite [Yaxley & Green, 1998], generate alkali glasses. But these glasses contain too high Al₂O₃ and too low CaO to represent a potential source of OIB. Experimental data indicate that partial melting of peridotite in the presence of carbonate [Hirose, 1997] or silica-deficient pyroxenite [Hirschmann et al., 2003; Kogiso & Hirschmann, 2003; Keshava et al., 2004] produces liquids which are close to the nephelinite, basanite or alkali-basalt major element compositions observed in OIB. Conversion of silica-saturated MORB to silica-deficient pyroxenite is possible, but requires either hybridization with peridotite, metasomatism by silica-deficient fluid such as carbonatite, or removal of low-pressure partial melts or fluids in the presence of olivine or spinel [Hirschmann et al., 2003, Lundstrom et al., 2000]. However, the effect of these different processes on the trace element budget of pyroxenites is not known.

1.3 An alternative model for the source of OIB

Silica deficient pyroxenite veins could be produced within continental or oceanic lithospheric mantle by percolation of magmatic liquids generated at low degrees of partial melt (low F) in upper-mantle rocks enriched in volatiles, alkali and incompatible trace elements at the base of the lithosphere [Sun & McDonough, 1989; Halliday et al., 1995; Niu et al., 1996; 1997; 2002; 2003]. As suggested by Niu & O'Hara [2003], this metasomatic process could occur as oceanic lithosphere cools and is transported away from the mid-ocean ridge. Subducted oceanic basal lithosphere enriched by metasomatic veins provides an alternative for the source of OIB. A similar hypothesis for the source of OIB, based on a comparison of OIB and MORB from the central Atlantic, was proposed by Halliday et al. [1995]. They postulated that the source region of OIB is enriched by small-degree partials melt soon after formation of the oceanic lithosphere at the periphery of plumes or at mid-ocean ridges. Metasomatism of the uppermost mantle would generate a "near-surface fractionated" (NSF) source with high U/Pb, Ce/Pb whilst maintaining Ce/U relatively unfractionated. The similarity of composition between basalt produced by modest extension in continental regions and OIB led McKenzie & O'Nions [1995] to propose that the source of OIB could be delaminated, metasomatised, continental lithosphere. As with models for metasomatism of ocean lithosphere, metasomatism of continental lithosphere implies the percolation of metasomatic melts rich in volatiles and incompatible elements into lithospheric mantle previously depleted by melt extraction [McKenzie & O'Nions,1995].

The problem with these various alternative models for the source of OIB is that they do not explain:

1. Why there are systematic isotopic and trace element variations in the source of basalt from a single island, and
2. How metasomatism can produce the apparent end-member mantle sources of OIB.

Moreover, the different mantle end-members (HIMU, EM1 and EM2), which are defined by their isotopic compositions, are also characterized by different trace-element ratios such Nb/Th, Ba/Nb, Th/La [Weaver, 1991; Chauvel et al., 1992]. In the recycled, ancient oceanic crust model [Hofmann & White, 1982], these elemental variations are related largely to the addition and assimilation of recycled sediment to the recycled oceanic crust [Weaver, 1991; Chauvel et al., 1992]. In contrast, no clear model for the correlation between trace-element and isotopic variations has been developed for different metasomatic alternative models.

The Cantal basalts from the Massif Central, France, have been interpreted as the product of metasomatism within the lithospheric mantle. This work provides new information that can constrain the formation of distinct compositions in lherzolitic OIB mantle sources.

2. A source for Cantal basalt involving veins plus enclosing lithospheric mantle

2.1 Trace element variations in the Cantal basalt source

The basalts erupted between 13 and 3 Ma ago in the Cantal alkali massif are similar to OIB in major and trace elements composition (Figures 1 & 2) [Wilson & Downes, 1991; Pilet et al., 2002]. They present a substantial composition range, from alkali basalts to basanites, which can be explained by a variation in the degree of partial melting of a common mantle source.

With respect to trace element compositions, these basalts show an unusual increase of Nb/Th and Ce/Pb ratios from 10.5 and 23 respectively during the first period of basalt eruption (13-9 Ma) to 18.7 and 44 during the last period (9-3 Ma) (Figure 3). The ranges in Nb/Th and Ce/Pb encompass practically all the variation observed in the major oceanic islands [Halliday et al., 1995; Jochum et al., 2001]. Moreover, Figure 3 indicates that the first erupted Cantal basalts have similar Nb/Th and Ce/Pb ratios as are observed in EM1, while the last are similar to HIMU.

Variations in Ce/Pb and Nb/Th in mafic rocks are generally attributed to sediment assimilation in the mantle source [Weaver, 1991; Chauvel et al., 1992] or during passage through the crust. However, the basalts are homogeneous with respect to their Sr, Nd and Pb isotopic compositions (Figure 4) and do not display any correlation between Ce/Pb or Nb/Th and the Sr, Nd or Pb isotopic ratios [Pilet et al., 2002; 2004].

Figure 4: 207Pb/204Pb vs 206Pb/204Pb diagram for Cantal basalts together with a selection of oceanic basalts. Dark grey field: MORB isotopic variation; light grey fields: isotopic variations observed in different oceanic islands [Zindler & Hart, 1986].

This implies that:

1. Variable sediment contamination of their mantle sources or crustal assimilation can be excluded by the observed Nb/Th and Ce/Pb variations, because a correlation between isotope and trace element variation would then be seen, which is not the case, and
2. The variation in trace elements in the source reflects "short-term" mineralogical control of trace-element ratios in a vein-plus-enclosing-mantle source rather than "long term" extraction of continental crust from the mantle, sediment recycling and incomplete homogenization of ancient heterogeneities in the mantle.

This mineralogical modification of the mantle source may result from the metasomatic process "percolative fractional crystallization" (PclFC) [Harte et al., 1993], which produces veins in the upper mantle.

2.2 Mineralogical evidence for a PclFC mechanism in the Cantal basalt source

The PclFC mechanism results in clinopyroxene (CPX) xenocrysts in the Cantal basalts [Pilet et al., 2002; 2004]. These xenocrysts have a compositional range from Ti-augite with Mg# > 70 (Mg# = Mg/[Mg+Fe²⁺]) to green aegirine-augite with Mg# < 50 (hereafter GCPX) (Figure 5).

The GCPX are interpreted as mantle xenocrysts because:

1. CPX with compositions similar to Cantal GCPX is observed in mantle veins from La Palmas [Wulff-Perderson et al., 1999, Pilet et al., 2002];
2. A fluid-inclusion study performed on GCPX from Hungarian basalts [Szabo & Bodnar, 1999] points to an upper-mantle origin;
3. Sr, Nd, Pb and O isotopic compositions of GCPX in continental flood basalts from Yemen [Baker et al., 2000] indicate the absence of crustal contamination during GCPX crystallization and a direct relationship to an asthenosphere-derived melts;
4. The observation of mantle xenoliths in Cantal rock where GCPX are observed excludes practically all possibility of mixing in the lower crust to explain the presence of the GCPX.

2.3 Influence of the enclosing lithosphere over the metasomatic process

The mineralogical and chemical constraints described above indicate that continental basalts with compositions similar to OIB could be generated by melting veins plus enclosing continental lithospheric mantle. But does this model apply to metasomatic vein formation within the oceanic lithosphere? Several lines of evidence suggest that it does:

1. The model suggests that the trace-element composition of basalts is mainly controlled by the composition of veins in the source [Pilet et al., 2004]. This is because of the difference in the degree of trace-element enrichment in veins compared to the surrounding mantle. For example, concentrations of incompatible trace elements are 40-80 times higher in vein minerals than in lithospheric spinel peridotite minerals (e.g., GCPX have 12-40 ppm of La whereas clinopyroxene from depleted peridotite xenoliths from the Massif Central contain only 0.3-0.6 ppm of La [Zangana et al., 1999]). Thus the nature of the enclosing lithosphere does not play an essential role in the generation of OIB trace-element compositions;
2. Cantal lithosphere has a depleted composition, which has been attributed to the extraction of basaltic liquid from an initial near-chondritic mantle source [Zangana et al., 1999; Lenoir et al., 2000; Downes et al., 2003]. Thus, before Neogene volcanic activity, Cantal lithospheric mantle was similar to typical unmetasomatised oceanic lithosphere;
3. GCPX xenocrysts in alkali basalt are described from both oceanic and continental settings [Pilet et al., 2002]. This suggests that the PclFC process is global.

Furthermore, the PclFC process observed in Cantal basalt sources can be linked with the low-F, volatile- and incompatible-element-enriched, metasomatic agent that seems to affect the oceanic basal lithosphere [Halliday et al., 1995; Niu et al., 1996; 1997; 2002; 2003]. The similarities in the two processes are:

1. Initial liquids derived from the PclFC mechanism must be produced at low F, i.e., the melt fractions must be sufficiently large to segregate from the peridotitic source but not so large that they are transported out of the upper mantle into the crust.
2. The initial PclFC model liquid must have high-H₂O and CO₂ contents, as demonstrated by amphibole inclusions in CPX xenocrysts, carbonate in association with GCPX in xenoliths from Australian basalt [Wass, 1979] and CO₂ fluid inclusions in GCPX from Hungarian basalts [Szabo & Bodnar, 1999].
3. In both the continental and oceanic environments, the initial metasomatic model melt is generated in the low velocity zone at the base of the lithosphere [Pilet et al., 2004, Hoernle et al., 1995, Granet et al., 1995, Niu et al., 2002; 2003].

3. Evolution of metasomatised heterogeneous lithosphere

3.1 Trace element variation within metasomatised heterogeneous lithosphere

The compositional differences between Cantal basalts erupted during the two periods of volcanism provide an unusual opportunity to characterize compositional variations introduced by metasomatic processes in lithospheric mantle. As indicated in Figure 6b, which is a schematic model of metasomatized lithosphere, we distinguish upper and lower parts of this hypothetical lithosphere. The composition of the upper part is constrained by the composition of first-erupted Cantal basalt while the lower part is related to the composition of last-erupted basalt. As shown in Figure 8, there are striking differences between the two parts.

Figure 8: Trace element variations observed between the first (13-9 Ma) and the last (9-3 Ma) emitted basalts in the Cantal Massif. Diagram normalised to the average composition of the 9-3 Ma basalts, measured at 9 ppm of Th, which accounts for differences in degrees of partial melting. Light blue lines: minimum and maximum normalised value of 9-3 Ma basalt measured at 9 ppm of Th. Red line: average and confidence interval of 13-9 Ma basalt. The range of µ ratios measured in 13-9 Ma (30.3 - 42.8) and 9-3 Ma (38.8 - 53.0) basalts. Cantal basalts may be used to calculate µ ratios of 24 and 31.2 in their respective sources.

These differences are particularly important for Ba, Nb, Ta, and Pb, yielding an increase in U/Pb and Th/Pb. Smaller variations observed for Sm and Nd are associated with a decrease in Sm/Nd for early basalts relative to late ones. There is also probably a decrease in Rb/Sr from early to late basalts but this is less safe since Rb was mobile during alteration of the 13-9 Ma Cantal basalts. The decrease in Rb/Sr is inferred on the basis of (1) a decrease in Ba/Sr between the two periods of basalt emission (Ba is generally correlated with Rb in basaltic rocks [Hofmann & White, 1983]) and (2) a decrease in Sr and an increase in Rb content in the most differentiated veins. The µ ratio average is estimated to be 31.2 in the source of the 9-3 Ma basalt (with extreme values up to 37.1) and 24.0 in the source of  the 13-9 Ma basalt (with extreme values up to 23.1). Analogous estimates for ¹⁴⁷Sm/¹⁴⁴Nd indicate ¹⁴⁷Sm/¹⁴⁴Nd values of 0.144 and 0.153 for the first and last erupted basalt respectively.

3.2 Temporal evolution of the metasomatised heterogeneous lithosphere

Metasomatised lithosphere could be entrained in asthenosphere flow during subduction of the oceanic lithosphere or delamination of the continental lithosphere. It would then become a long-term heterogeneity producing isotopically enriched mantle. We can model the temporal evolution of isotopic systems in the enriched mantle.

The subduction process generates a metamorphic transformation of vein mineralogy and probably hybridization within the enclosing mantle. However, because of the very low diffusion rate at high pressure [Hofmann & Hart, 1978; Kogiso et al., 2004], chemical heterogeneity will be preserved within the subducted mantle lithosphere. Nevertheless, during these processes, the enriched component will be greatly diluted in a manner similar to that proposed for enrichment by subduction of oceanic crust [Chauvel et al., 1992], erosion/delamination of sub-continental lithosphere [McKenzie & O'Nions, 1995] or model NSF mantle recycling [Halliday et al., 1995]. It is extremely difficult to estimate the dilution factor, but the average of the enriched component calculated previously and the depleted mantle ( µ _DM = 10 and ¹⁴⁷Sm/¹⁴⁴Nd _DM = 0.21 [Rehkämper & Hofmann, 1997]) may be used for model evolution calculations.

Because metasomatised lithosphere would be heterogeneous with variable U/Pb, Th/Pb and Sm/Nd ratios, different Pb and Nd isotopic ratios will be produced in the upper and lower parts of the lithosphere. The µ ratio of the basal lithosphere would, after 1.77 Ga, have a ²⁰⁶Pb/²⁰⁴Pb isotopic ratio similar to that of the sources of HIMU basalts (Figure 9a). The absolute value of this isotopic ratio is highly dependant of the estimate of the µ ratio in the source, isolation time, dilution factor and choice of initial isotopic composition which is taken here to be similar to depleted mantle at 1.77 Ga or 1 Ga. Conversely, the difference in µ calculated between the lower and upper lithosphere (and in the corresponding last and first emitted Cantal basalt) is better constrained and varies linearly with the dilution factor or with the method used to estimate the µ value of the sources. The different values of µ produce ²⁰⁶Pb/²⁰⁴Pb variations of ~1.2 over 1.77 Ga (Figure 9a) using the average µ values, and ~2.2 using the extreme µ values for the first (16.55, average of the value calculated from Cantal basalts and µ value of DM) and last (23.55) emitted basalts. For an isolation time of 1 Ga, the ²⁰⁶Pb/²⁰⁴Pb variation range calculated is somewhat smaller than, though similar in scale to, that seen in individual oceanic islands (~0.6 for average µ values and ~1.9 for extreme µ values) (Figure 9a).

The combination of µ and Sm/Nd heterogeneities in Cantal basalt sources produces variations in ¹⁴³Nd/¹⁴⁴Nd – ²⁰⁶Pb/²⁰⁴Pb that simulate mixing between HIMU and EM components reported for OIB worldwide (Figure 9b). The inferred high Rb/Sr in the upper part of the hypothesised lithosphere relative to the lower part would produce significant ⁸⁷Sr/⁸⁶Sr variations with time in the derived basalts. This estimate of ⁸⁷Sr/⁸⁶Sr variation is in agreement with the apparent mixing trends between HIMU and EM components inferred from ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb and ¹⁴³Nd/¹⁴⁴Nd variations calculated from the upper and lower parts of the heterogeneous lithosphere.

Isotopic trends of Nd-Sr-Pb in oceanic basalts are correlated with trace-element ratios that are used to infer elemental signatures of the hypothetical EM and HIMU end-member mantle components [Weaver, 1991; Chauvel et al., 1992] (Figure 3, 10a, 10b). As indicated in Figure 10, the variations in trace element ratios from the first to the last period of Cantal basalt emission define a trend that mimics mixing between HIMU and an EM component. Figure 10: a) Ba/Nb vs La/Nb and b) Th/Nb vs Th/La diagrams for Cantal basalts; 13-9 Ma (open red circles) and 9-3 Ma (filled red circles) from the upper and lower parts of metasomatised heterogeneous lithosphere. HIMU (triangles), EM1 (filled squares) and EM2 (open squares) compositions are from Weaver [1991]. These data demonstrate that fractionation of trace-element ratios (U/Pb, Th/Pb, Rb/Sr, Sm/Nd, Nb/La etc.), necessary to generate EM and HIMU isotopic characteristics over time, can be simulated by the evolution of a metasomatic agent within the lithosphere. These observations are compelling evidence that variable isotopic and trace-element-ratio signatures in the mantle sources of single oceanic islands can be produced by metasomatised and subducted oceanic lithosphere.

4. Interpretation of end-member compositions

4.1 Example from HIMU – EM variation in Polynesia basalts

Mangaia (Cook Islands) and Tahaa (Society Islands) basaltic rocks are considered to be derived from the partial melting of nearly "pure" HIMU and EM2 mantle components respectively [Hart et al., 1992]. Trace-element variations observed between basalts from these two islands (Figure 11a) are remarkably similar to the variations observed between the two periods of Cantal basalt emission (Figure 8). So, by analogy with the model proposed for Cantal basalt formation, the Mangaia and Tahaa mantle sources may have been produced during a metasomatic event within the oceanic lithosphere. In this hypothesis, Mangaia and Tahaa source compositions correspond to a heterogeneous lithosphere containing mafic-to-intermediate and differentiated veins respectively.

Data from olivine-hosted melt inclusions analyzed in both Mangaia and Tahaa basalts [Saal et al., 1998] strongly support this idea. Melt inclusions from Mangaia alone have ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁶ and ²⁰⁸Pb/²⁰⁶Pb variations which cover much of the HIMU to EM2 lead isotope composition range (Figure 9a). The lead isotope range of Mangaia melt inclusions is slightly larger than that deduced for the Cantal basalts from the variation of µ after an isolation time of 1.77 Ga. However, there is no reason to suggest that the lithospheric sources of Cantal basalt were the most affected by the postulated metasomatism. Tahaa melt inclusions are similar to Mangaia melt inclusions, but show a smaller Pb isotopic range corresponding to EM2 isotopic compositions.

The progressive variation from HIMU to EM2 isotopic compositions in melt inclusions is associated with an increase in Pb content and a decrease of Cr in CPX crystallized within the inclusions [Saal et al., 1998]. These relationships indicate that a differentiation process was involved in the generation of the mantle reservoirs prior to long-term evolution of the Pb isotopic system [Saal et al., 1998]. These variations are in perfect agreement with the hypothesis proposed here. The differentiation process which generates the different vein composition within the lithosphere leads to a decrease of U/Pb, Th/Pb and Cr and an increase of Pb from the bottom to the top of the oceanic lithosphere prior to subduction (Figures 6 and 8). The continuous variation of U/Pb and Th/Pb from the bottom to the top of the lithosphere will generate, after sufficient isolation time, a decrease in ²⁰⁶Pb /²⁰⁴Pb associated with an increase of ²⁰⁷Pb/²⁰⁶ and ²⁰⁸Pb/²⁰⁶Pb, which are correlated to a decrease of Cr and an increase of Pb.

This differentiation process can explain the different correlations and variations observed in Mangaia and Tahaa melt inclusions and whole rocks. This suggests that HIMU and EM2 end-member isotopic compositions correspond to the extreme trace-element fractionation generated by metasomatic processes within the lithosphere, as illustrated at Cantal. A similar metasomatic origin for the EM2 signature has been proposed previously by Workman et al. [2004] based on a study of the Samoan volcanic chain. The metasomatic interpretation of the EM2 component is also in agreement with the hypothesis of Eiler et al. [1997] to explain the origin of high δO¹⁸ observed in some EM2 rocks. Following these authors, the high δO¹⁸ values in EM2 could be linked either to addition of ~2-6% of terrigenous sediment in the mantle source or with a metasomatic process within mantle peridotite [Workman et al., 2004; Eiler et al., 1997].

Rarotonga basalts do not present the low ¹⁴³Nd/¹⁴⁴Nd observed in Pitcairn basalts which are considered to be the melt products of a near "pure" EM1 source. A hypothesis for the decrease of Sm/Nd in the source of Pitcairn basalts is the fractionation of garnet during the metasomatic process which generates the differentiated veins (Kd^{Garnet/liquid}_Sm > Kd^{Garnet/liquid}_Nd). Phlogopite and garnet represent high pressure phases in alkaline rocks. Thus, we propose that metasomatic formation of the EM1 isotopic composition, rather than EM2, is linked to the thickness of the lithosphere – thick for the formation of EM1 (continental lithosphere?) and thin for EM2 (oceanic lithosphere?). However, more studies are necessary to investigate further these last suggestions.

4.2 Variation of high field strength elements in HIMU and non-HIMU ocean island basalt

In the subducted oceanic crust hypothesis for the source of OIB, enrichments in Ba, Rb, Th, U and La relative to Nb and Ta in the EM1 and EM2 end-members, compared to the HIMU source (Figures 10a and 10b), is ascribed to recycling of pelagic (EM1) and terrigenous (EM2) sediments into the mantle [Weaver, 1991; Chauvel et al., 1992]. Following Kogiso et al. [1997], this hypothesis is problematic for HIMU and other OIBs in southern Polynesia. The major element characteristics of those OIB magmas do not support significant involvement of basaltic components in non-HIMU sources; and the Nb enrichment of HIMU basalts relative to other incompatible elements is difficult to explain simply by the addition of a crustal component to the source of HIMU rocks [Kogiso et al., 1997]. These inconsistencies are solved by the metasomatic hypothesis presented here, where the high Nb/Th in HIMU is explained by the presence of a small amount of Ti-rich oxide in metasomatic veins. Ti-oxides have large distribution coefficients (Kd) for HFSEs. Thus, low La/Nb, Ba/Nb, Th/Nb and high Hf/Lu in the HIMU source is explained by precipitation of Ti-oxide during modal metasomatism, preferentially enriching Nb and Hf relative to other incompatible elements in the lithospheric source. Hf and Nd isotope ratios in non-HIMU OIB exhibit a strong positive correlation (Figure 13), indicating a control by magmatic processes [Salter & White, 1998; Niu & O'Hara, 2003]. HIMU basalts differ from other basalts falling on the main OIB array in that they have a low ¹⁷⁶Hf/¹⁷⁷Hf value for a given ¹⁴³Nd/¹⁴⁴Nd, reflecting a larger Hf/Lu ratio in their source. Even if recycled subducted oceanic crust can duplicate the Hf-Nd isotopic composition observed in HIMU basalts, mixing the subducted residuum with sediments in the mantle does not produce the main ¹⁷⁶Hf/¹⁷⁷Hf - ¹⁴³Nd/¹⁴⁴Nd array for non-HIMU OIBs [Salter & White, 1998]. Enrichment in Hf relative to Lu in the HIMU source by modal metasomatism before the evolution of the isotopic system explains the Hf-Nd isotopic variation in all OIB sources more simply (Figure 13).

5. Summary

5.1 General

1. Variations in the chemical compositions of Cantal basalts are interpreted as the result of a metasomatic agent within the lithosphere. The Cantal example demonstrates that basalts characterized by OIB chemical compositions could be generated by partial melting of metasomatic veins plus enclosing lithospheric mantle. Such a heterogeneous source may form within either the continental or oceanic lithosphere.
2. The evolution of the metasomatic agent could generate trace element variations (U/Pb, Th/Pb, Rb/Sr, Sm/Nd, Nb/La etc.) in different parts of the lithosphere. Isolation of this metasomatic lithosphere would generate isotopic and trace element ratio variations that are similar to the variations observed in oceanic islands. These observations are compelling evidence that variable isotopic and trace element ratio signatures in the mantle sources of single oceanic islands can be produced by subduction of metasomatised oceanic lithosphere or delamination of metasomatised continental lithosphere.
3. The similarity between the trace element variation observed in Cantal basalt and the variation observe between Mangaia (HIMU) and Tahaa (EM2) - Rarotonga (EM1) islands suggest that model HIMU and EM end-member isotopic and chemical compositions could correspond to the extreme trace-element fractionation generated by metasomatic process within the lithosphere.

5.2 Implications for the existence of plumes

1. The new interpretation of OIB sources proposed here requires re-evaluation of the processes that control the chemical evolution of Earth's mantle reservoirs and their chemical and isotopic indicators.
2. The metasomatic hypothesis is entirely consistent with the isotopic heterogeneity observed in the source of oceanic basalts and experimental data which indicate that the most plausible source of OIB material are pyroxenites [Hirschmann et al., 2003; Kogiso et al., 2003]. This model is particularly interesting for the formation of high-silica undersaturated rocks as nephelinite or basanite where the common interpretation as recycled oceanic crust clearly fails to explain their major element composition.
3. Recycled metasomatised lherzolite may be the major mantle source sampled by OIBs; recycled oceanic crust and sediment may be less common in OIB sources than is usually assumed.
4. The addition of volatiles, in particular of CO₂, to lherzolite significantly decreases the solidus temperature. The metasomatic model for the origin of OIB suggests that OIB could be produce at a lower temperature than is commonly assumed in the recycled oceanic crust model.

Acknowledgements

We are grateful for long discussions with Mike O'Hara and Yaoling Niu. We also thank D. Fontignie, A. Potrel, M. Poujol, J. Kramers, B. Villemant, M. Tubrett and L. Hewa for help with TIMS, INAA, MC-ICP-MS and laser ablation ICP-MS analyses.

References

• F. Albarede, How deep do common basaltic magmas form and differentiate?, J. Geophys. Res., 97 (1992) 10,997-11,009
• J.A. Baker, C.G. Macpherson, M.A. Menzies, M.F. Thirlwall, M. Al-Kadazi, D.P. Mattey, Resolving crustal and mantle contributions to continental flood volcanism, Yemen: Constraints from mineral oxygen isotope data, J. Pet., 41 (2000) 1805-1820
• R.M. Bedini, J.L. Bodinier, J.M. Dautria and L. Morten, Evolution of LILE-enriched small melt fractions in the lithospheric mantle: A case study from the East African rift: Earth Planet. Sci. Lett., 153 (1997) 67-83
• J. Blichert-Toft, F. Albarede, The Lu-Hf isotope geochemistry of chondrites and evolution of the mantle-crust system, Earth Planet. Sci. Lett., 148 (1997) 243-258
• C. Chauvel, A.W. Hofmann, P. Vidal, HIMU-EM; the French Polynesian connection, Earth Planet. Sci. Lett., 110 (1992) 99-119
• G. Dobosi, G.A. Jenner, Petrologic implications of trace element variation in clinopyroxene megacrysts from Nógrád volcanic province, north Hungary: a study by laser ablation microprobe-inductively coupled plasma-mass spectrometry, Lithos, 46 (1999) 731-749
• J. Dostal, B.L. Cousens, C. Dupuy, The incompatible element characteristics of an ancient subducted sedimentary component in ocean island basalt from French Polynesia, J. Pet., 39 (1998) 937-952
• H. Downes, M.K. Reichow, P.R.D. Mason, A.D. Beard, M.F. Thirlwall, Mantle domains in the lithosphere beneath the French Massif Central: trace element and isotopic evidence from mantle clinopyroxenes, Chem. Geol., 200 (2003) 71-87
• C. Dupuy, H.G. Barsczus, J. Dostal, P. Vidal, J.-M. Liotard, Subducted and recycled lithosphere as mantle source of ocean island basalts from southern Polynesia, central Pacific, Chem. Geol., 77 (1989) 1-18
• J.M. Eiler, K.A. Farley, J.W. Valley, E. Hauri, H. Craig, S.R. Hart, E. Stolper, Oxygen isotope variations in ocean island basalt phenocrysts, Geochim. Cosmochim. Acta,, 61 (1997) 2281-2293
• M. Granet, M. Wilson, U. Achauer, Imaging a mantle plume beneath the French Massif Central, Earth Planet. Sci. Lett., 136 (1995) 281-296.
• T.H. Green, D.H. Green, A.E. Ringwood, 1967, The origin of high-alumina basalts and their relationships to quartz tholeiites and alkali basalts, Earth Planet. Sci. Lett., 2 (1967) 41-51
• A.N. Halliday, et al., Incompatible trace elements in OIB and MORB and source enrichment in the sub-oceanic mantle, Earth Planet. Sci. Lett., 133 (1995) 379-395.
• S.R. Hart, E.H. Hauri, L.A. Oschmann, J.A. Whitehead, Mantle plumes and entrainment; isotopic evidence, Science, 256 (1992) 517-520
• B. Harte, R.H. Hunter, P.D. Kinny, Melt geometry, movement and crystallization, in relation to mantle dykes, veins and metasomatism, Phil. Trans. Royal Soc. London, Series A, 342 (1993) 1-21
• K. Hirose, Partial melt compositions of carbonated peridotite at 3 GPa and genesis of alkali basalt magmas, Geophys. Res. Lett., 24 (1997) 2837-2840
• E.H. Hauri, S.R. Hart, Re-Os isotope systematics of HIMU and EMII oceanic island basalts from the south Pacific ocean, Earth Planet. Sci. Lett., 114 (1993) 353-371
• M.M. Hirschmann, T. Kogiso, M.B. Baker, E.M. Stolper, 2003, Alkalic magmas generated by partial melting of garnet pyroxenite, Geology, 31 (2003) 481-484
• A.W Hofmann and S.R. Hart, An assessment of local and regional isotopic equilibrium in the mantle Earth Planet. Sci. Lett., 38 (1978) 44-6
• A.W. Hofmann, W.M. White, Mantle plumes from ancient oceanic crust, Earth Planet. Sci. Lett., 57 (1982) 421-436.
• A.W. Hofmann, W.M. White, Ba, Rb and Cs in the Earth's Mantle, Zeitschrift für Naturforschung, 38a (1983) 256-266
• K. Hoernle, Y.S. Shang, D. Graham, Seismic and geochemical evidence for large-scale mantle upwelling beneath the eastern Atlantic and Western and Central Europe, Nature, 374 (1995) 34-30
• K.P. Jochum, B. Sarbas, A.W. Hofmann, Characterizing OIB mantle sources by trace element ratios using the GEOROC Database, GERM, workshop 2001, La Jolla, Abstract, (2001) B22-B24
• F. Kalfoun, D. Ionov, C. Merlet, 2002, HFSE residence and Nb/Ta ratios in metasomatised, rutile-bearing mantle peridotites, Earth Planet. Sci. Lett., 199 (2002) 49-65
• S. Keshava, G. H. Gudfinnsson, G. Sen, Y. Fei, High-pressure melting experiments on garnet clinopyroxenite and the alkalic to tholeiitic transition in ocean-island basalts, Earth Planet. Sci. Lett., 223 (2004) 365-379
• T. Kogiso, Y. Tatsumi, G. Shimoda, H.G. Barsczus, High m (HIMU) ocean island basalts in southern Polynesia: new evidence for whole mantle scale recycling of subducted oceanic crust, J. Geophys. Res., 102 (1997) 8085-8103
• T. Kogiso, K. Hirose, E. Takahashi, Melting experiments on homogeneous mixtures of peridotite and basalt: application to the genesis of ocean island basalts, Earth Planet. Sci. Lett., 162 (1998) 45-61.
• T. Kogiso, M. M. Hirschmann, D. J. Frost, High-pressure partial melting of garnet pyroxenite: possible mafic lithologies in the source of ocean island basalts, Earth Planet. Sci. Lett., 216 (2003) 603-617.
• T. Kogiso, M.M. Hirschmann, P.W. Reiners, Length scales of mantle heterogeneities and their relationship to ocean island basalt geochemistry, Geochim. Cosmochim. Acta, 68 (2004) 345-360
• X. Lenoir, C.J. Garrido, J.-L. Bodinier, J.-M. Dautria, Contrasting lithospheric mantle domains beneath the Massif Central (France) revealed by geochemistry of peridotite xenoliths, Earth Planet. Sci. Lett., 181 (2000) 359-375
• X. Lenoir, C.J. Garrido, J.L. Bodinier, J.M. Dautria and F. Gervilla, The recrystallization front of the Ronda Peridotite: Evidence for melting and thermal erosion of subcontinental lithospheric mantle beneath the Alboran Basin, J. Pet., 42 (2001) 141-158
• C.C. Lundstrom, J. Gill, Q. Williams, A geochemically consistent hypothesis for MORB generation, Chem. Geol., 162 (2000) 105-126
• D. McKenzie, K.R. O'Nions, The source regions of ocean island basalts, J. Pet., 36 (1995) 133-159
• Y. Nakamura, M. Tatsumoto, Pb, Nd and Sr isotopic evidence for a multicomponent source for rocks of Cook-Austral islands and heterogeneities of mantle plumes, Geochim. Cosmochim. Acta, 52 (1988) 2909-2924
• Y. Niu, D.G. Waggoner, J.M. Sinton, J.J. Mahoney, Mantle source heterogeneity and melting processes beneath seafloor spreading centres: the East Pacific Rise, 18¾ – 19¾S, J. Geophys. Res., 101 (1996) 27711-27733
• Y. Niu, R. Batiza, Trace element evidence from seamounts for recycled oceanic crust in the eastern Pacific mantle, Earth Planet. Sci. Lett., 148 (1997) 471-483
• Y. Niu, M. Regelous, I.J. Wendt, R. Batiza, M.J. O'Hara, 2002, Geochemistry of near-EPR seamounts: importance of source vs. process and the origin of enriched mantle component, Earth Planet. Sci. Lett., 199 (2002) 327-345
• Y. Nui, and M.J. O'Hara, Origin of ocean island basalts: A new perspective from petrology, geochemistry, and mineral physics considerations, J. Geophys. Res., 108, (2003) B4, 2209
• S. Pilet, J. Hernandez, B. Villemant, Evidence for high silicic melt circulation and metasomatic events in the mantle beneath alkaline provinces: the Na-Fe-augitic green-core pyroxenes in the Tertiary alkali basalts of the Cantal massif (French Massif Central), Min. Pet., 76 (2002) 39-62
• S. Pilet, J. Hernandez, F. Bussy, P.J. Sylvester, Short-term metasomatic control of Nb/Th ratios in the mantle sources of intra-plate basalts, Geology, 32 (2004) 113-116
• M. Rehkämper, A.W. Hofmann, Recycled ocean crust and sediment in Indian Ocean MORB, Earth Planet. Sci. Lett., 147 (1997) 93-106
• A.E. Saal, S.R. Hart, N. Shimizu, E.H. Hauri, G.D. Layne, Pb isotopic variability in melt inclusions from oceanic island basalts, Polynesia, Science, 282 (1998) 1481-1484
• V.J.M. Salters, W.M. White, Hf isotope constraints on mantle evolution, Chem. Geol., 145 (1998) 447-460
• S.S. Sun, W.F. McDonough, Chemical and isotopic systematics of oceanic basalts; implications for mantle composition and processes, in A.D. Saunders and M.J. Norry, eds., Magmatism in the oceanic basins. Geological Society of London Spec. Pub. 42 (1989) 313-345
• Cs. Szabo, R.J. Bodnar, Fluid-inclusion evidence for an upper-mantle origin for green clinopyroxenes in late Cenozoic basanites from the Nógrad-Gömör volcanic field, northern Hungary, southern Slovakia, in Snyder, G.A., et al., Planetary petrology and geochemistry, The Lawrence A. Taylor 60th birthday volume, Columbia, Bellwether Publishing (1999) 83-91.
• G.M. Thompson, I.E.M. Smith, J.G. Malpas, Origin of oceanic phonolites by crystal fractionation and the problem of the Daly gap: an example from Rarotonga, Contrib. Min. Pet., 142 (2001) 336-346
• S.Y. Wass, Fractional crystallization in the mantle of late-stage kimberlitic liquids – Evidence in xenoliths from Kiama area, N.S.W., Australia, in Boyd, F.R., and Meyer, H.O.A., eds., The mantle sample: Inclusion in kimberlites and other volcanics: Washington, Am. Geophys. Union. (1979) 366-373
• B.L. Weaver, The origin of ocean island end-member compositions: trace element and isotopic constraints, Earth Planet. Sci. Lett., 104 (1991) 381-397
• W.M. White, R.A. Duncan, Geochemistry and geochronology of the Society Islands: new evidence for deep mantle recycling, in A. Basu and S.R. Hart, eds, Earth processes, reading the isotopic code. AGU, Washington DC: Geophysical monograph, 95 (1996) 183-206
• W.M. White, Sources of oceanic basalts: Radiogenic isotopic evidence, Geology, 13 (1985) 115-118
• R.K. Workman, et al., Recycled metasomatized lithosphere as the origin of the Enriched Mantle II (EM2) end-member: evidence from the Samoan volcanic chain, Geochem. Geophys. Geosysems, 5 (2004) 2003GC000623
• M. Wilson, H. Downes, Tertiary-quaternary extension-related alkaline magmatism in western and central Europe, J. Pet., 32 (1991) 811-849
• E. Wulff-Pederson, E.-P. Neumann, R. Vannucci, P. Bottazzi, L. Ottolini, Silicic melts produced by reaction between peridotite and infiltrating basaltic melts: ion probe data on glasses and minerals in veined xenoliths from La Palma, Canary Islands, Cont. Min. Pet., 137 (1999) 59-82.
• G.M. Yaxley, D.H. Green, Reactions between eclogite and peridotite: mantle refertilisation by subduction of oceanic crust, Schweitz. Mineral. Petrogr. Mitt., 78 (1998) 243-255
• N.A. Zangana, H. Downes, M.F. Thirlwall, G.F. Marriner, F. Bea, Geochemical variation in peridotite xenoliths and their constituent clinopyroxenes from Ray Pic (French Massif Central); implications for the composition of the shallow lithospheric mantle, Chem. Geol., 153 (1999) 11-35
• A. Zindler, S.R. Hart, Chemical geodynamics, Ann. Rev. Earth Planet. Sci., 14 (1986) 493- 571.

last updated 20th February, 2005