Earth and Planetary Science Letters 286 (2009) 456–466

Contents lists available at ScienceDirect

Earth and Planetary Science Letters j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e p s l

Survival of lithium isotopic heterogeneities in the mantle supported by HIMU-lavas from Rurutu Island, Austral Chain I. Vlastélic a,⁎, K. Koga a, C. Chauvel b, G. Jacques a,1, P. Télouk c a

Laboratoire Magmas et Volcans, Observatoire de Physique du Globe de Clermont-Ferrand, Université Blaise Pascal, CNRS UMR 6524, 5 Rue Kessler, 63038 Clermont-Ferrand, France Laboratoire de Géodynamique des Chaînes Alpines, Observatoire des Sciences de l'Univers de Grenoble, Université Joseph Fourrier, CNRS UMR 5025, 1381 rue de la Piscine, 38041 Grenoble cedex 9, France c Laboratoire des Sciences de la Terre, Ecole Normale Supérieure de Lyon, CNRS UMR 5570, 46 Allée d'Italie, 69364 Lyon cedex 07 b

a r t i c l e

i n f o

Article history: Received 10 March 2009 Received in revised form 6 July 2009 Accepted 6 July 2009 Available online 18 August 2009 Editor: R.W. Carlson Keywords: lithium isotopes diffusion ocean island basalts HIMU Rurutu island

a b s t r a c t Thirty years ago, Hofmann and Hart [Hofmann, A.W., Hart, S.R., 1978. An assessment of local and regional isotopic equilibrium in the mantle, Earth Planet Sci. Lett. 38, 44–62] showed that local disequilibria of slowly diffusing radiogenic tracers (e.g. Sr) are preserved in the mantle over 1–2 Ga time scales. Recently, it was suggested that this is not the case for fast diffusing elements such as lithium [Halama, R., McDonough, W.F., Rudnick, R.L., Bell, K., 2008. Tracking the lithium isotopic evolution of the mantle using carbonatites, Earth Planet Sci. Lett. 265, 726–742], thus questioning the ability of lithium isotopes to constrain long-term effects of recycling of material with crustal signatures. A key issue in this debate is the identification in hotspot volcanism of Li isotopic fingerprint consistent with recycling. Previous studies proposed that HIMU type volcanism, which is thought to sample mantle domains that include subducted altered oceanic crust, has 7Li/ 6 Li distinctively higher than the fresh mid-ocean ridge basalts. This work focuses on Rurutu island, where both HIMU (20.88 b 206Pb/204Pb b 21.42) and non-HIMU (19.11 b 206Pb/204Pb b 20.45) lavas occur. When considering only the freshest and most primitive lavas, the lithium isotopic signatures of HIMU (+ 5.4 b δ7Li b + 7.9‰) and non-HIMU (+ 2.9 b δ7Li b + 4.8‰) lavas do not overlap, thus supporting the idea that HIMU mantle has distinctly elevated δ7Li. This result suggests that Li isotopic heterogeneities could survive diffusion over 1–2 billion years, the amount of time required to develop the highly radiogenic Pb signature of HIMU-lavas. Modeling lithium diffusion out of a lithium-rich, isotopically heavy altered oceanic crust reveals that isotopic disequilibrium persists over long time periods in comparison with the rapid decrease of chemical disequilibrium. For instance, a kilometer-thick altered oceanic crust loses most of its Li excess in a matter of a few tens of millions of years but could preserve 7Li/6Li distinctively higher than the ambient mantle over a time period in excess of 1.5 Ga. After 1.5 Ga, a kilometer-thick altered oceanic crust has heavy but uniform 7Li/ 6 Li, while large isotopic variations persist in the nearby mantle (b 20 km). Thus, decoupling between Li and slowly diffusing radiogenic isotopes can be predicted in the mantle nearby subducted crust. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Until very recently, lithium stable isotopes were considered as an ideal tracer of recycling processes in the mantle, because of the large variations (tens of permil) in their relative abundance in surface environments (e.g. Elliott et al., 2004), but this view was recently challenged for a number of reasons. First, the precise mantle signature is accessible only after accounting for the modification processes

⁎ Corresponding author. Laboratoire Magmas et Volcans, Observatoire de Physique du Globe de Clermont-Ferrand, UMR 5025, 5 Rue Kessler, 63038 Clermont-Ferrand. Tel.: +33 4 73 34 67 10; fax: +33 4 73 34 67 44. E-mail address: [email protected] (I. Vlastélic). 1 Now at IFM-GEOMAR, Wischhofstrasse 1–3, 24148 Kiel, Germany. 0012-821X/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2009.07.013

during magma storage and segregation, such as contamination (Tomascak et al., 2008; Chan et al., 2009), and diffusion-driven fractionation (Richter et al., 2003; Seitz et al., 2004; Lundstrom et al., 2005; Jeffcoate et al., 2007; Parkinson et al., 2007; Ionov and Seitz, 2008). Second, Li isotopes rarely correlate with long-lived radiogenic isotopes that are commonly used to identify mantle heterogeneities. Third, the range of isotopic variations thought to reflect mantle heterogeneity is unexpectingly narrow (a few permil) (e.g. Tomascak et al., 2008). The cause of these small variations has been ascribed to a) small compositional contrast between the average input at subduction zones and mantle (Marschall et al., 2007), b) isotopic homogenization during melting (Ryan and Kyle, 2004) and/or c) fast diffusion of lithium in the asthenospheric mantle (Halama et al., 2008). On the other hand, some observations support the existence of small, but real Li isotopic variations in the mantle. Along mid-oceanic ridges,

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

Li–Sr–Nd isotope co-variations were identified locally (Elliott et al., 2006; Simons et al., 2008). Among the ocean islands with a strong HIMU affinity (St Helena, Mangaia, Rurutu, Raivavae and Tubuai), lithium compositions are distinctly heavier (+5.0 N δ7LiN +7.4‰) than those of the fresh mid-ocean ridge basalts (+2.5b δ7Lib +5.5‰). This observation possibly reflects the existence of mantle domains with distinctly heavy lithium compositions (Ryan and Kyle, 2004; Nishio et al., 2005; Chan et al., 2009). However, it is difficult to evaluate whether the heavy signatures represent genuine mantle source values or values modified by near surface petrogenetic processes, primarily because of the small number of samples analyzed for each locality. The HIMU mantle end-member is an ideal target to test if lithium isotopes can constrain recycling processes. HIMU is a relatively pure mantle component probably derived from recycling of altered oceanic crust free of sediments known to have a highly heterogeneous Li signature (Chan et al., 2006). Thus, if the subduction processes do not significantly modify the Li isotopic composition of altered oceanic crust (Marschall et al., 2007), heavy lithium composition is expected in the HIMU component. In addition, the highly radiogenic Pb signature of HIMU requires storage in the mantle over a period of time exceeding 1.5 Ga (Chauvel et al., 1992). The preservation of heavy Li signature in HIMU mantle would indicate that isotopic heterogeneities in Li could survive diffusion over time scales much longer than those proposed by Halama et al. (2008). Chan et al. (2009), who recently investigated in detail Li isotopic variations in Cook–Austral islands (Raivavae, Rapa, Mangai, and Tubuai), concluded that the isotopic composition of the HIMU mantle is lighter than the previously proposed composition, but still distinctly heavier than the non-HIMU mantle. The present study focuses on Rurutu Island, where both moderately radiogenic Pb and typical HIMU signatures have been identified (Chauvel et al., 1997). Having distinguished secondary processes from the mantle signature, we confirm that high 7Li/6Li is a robust feature of the HIMU mantle component and, based on diffusion modeling, we discuss how Li isotopic heterogeneities can be preserved in the mantle over billion-year time scales. 2. Geologic, petrologic and geochemical background Rurutu island belongs to the Austral–Cook volcanic chain, which extends over 2000 km in the south Pacific Ocean, from the Macdonald seamount to the Palmerson atoll (Fig. 1a). As first shown by Duncan and McDougall (1976), Rurutu was built during two main volcanic stages.

457

The old lavas erupted ca. 12 Ma ago, first in submarine then in subaerial environments. The young lavas emplaced about 11 Ma later on top of the carbonate sediments that were deposited during the long period of quiescence. The old volcanism is generally associated with the hotspot track linking the active Mcdonald seamount to the 19 Ma old Mangaia island (Diraison, 1991; Chauvel et al., 1997), although a connection with a northern track extending from Raivavae (6.5 Ma) to Tubuai (9 Ma) and Rurutu islands has been recently suggested (Bonneville et al., 2006). The young Rurutu volcanism belongs to the hotspot track that extends from the 8 Ma old Atiu Island to the recently discovered, 0.2 Ma old Arago seamount, 130 km southeast of Rurutu island (Bonneville et al., 2002). Rurutu is made of aphyric (young volcanism) to olivine–clinopyroxene–plagioclase phyric lavas (old volcanism) (Guille et al., 1998). The old lavas consist mainly of alkali basalts whereas the young lavas are more differentiated, with composition ranging from basanite to hawaiite. As Mangaia and Tubuai volcanics, the old Rurutu lavas have a typical HIMU isotopic fingerprint (206Pb/204Pb from 20.88 to 21.42; 87Sr/86Sr from 0.702761 to 0.703003 and 143Nd/144Nd from 0.512852 to 0.512906) (Chauvel et al., 1997). The young lavas show much less HIMU affinity (with 206Pb/204Pb ranging from 19.11 to 20.45), possibly reflecting contamination of plume melts within the lithosphere (Chauvel et al., 1997) or at shallower level. Previous Li isotopic analyses of Rurutu include one sample (RRT-C10)from the old lava group with a composition (δ7Li = +5.4‰) similar to Tubuai (+4.9 b δ7Lib +5.5‰), but lighter than Mangaia (+5.5 bδ7Li b +7.8‰) compositions (Nishio et al., 2005; Chan et al., 2009). 3. Samples Samples listed on Table 1 were selected on the basis of freshness by previous geochemical and geochronological studies after inspection of thin sections and lost on ignition (LOI) measurements (Duncan and McDougall, 1976; Dupuy et al., 1988, 1989; Diraison, 1991; Chauvel et al., 1997; Guille et al., 1998).With the exception of samples from Dupuy et al. (1988, 1989), sampling locations are known (Fig. 1b). Miocene samples have slightly higher LOI (0.8 to 2.2% except for one sample with 2.9%) compared to the Pleistocene samples (LOI from −0.2 to 0.7% except for one sample with 1.6%). The negative LOI of sample 74-388, which is ascribed to oxidation of iron, is indicative of freshness. Guille et al. (1998) described traces of palagonitic alteration in the oldest basaltic samples, consistently with their emplacement in submarine environment. Using X-ray diffraction and scanning electron

Fig. 1. Location and map of Rurutu island. (a) The 4000-m depth contour as well as major islands (circles) and seamounts cited in text (triangles) are shown. (b) Location of samples is shown, when known (Duncan, 1975; Duncan and McDougall, 1976; Diraison, 1991; Guille et al., 1998).

458

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

microscopy, we identified smectite (montmorillonite) in samples RR01 and RR67. When chip material was available (Table 1), chips without trace of alteration were selected for the Li isotopes analysis. The distinctive isotopic signature of the Pleistocene sample 74-394 (e.g. 206Pb/204Pb of 19.11) led Chauvel et al. (1997) to question an Austral plume origin. This sample is considered in this study because its isotopic signature falls within the field of the Cook–Austral Islands. Guille et al. (1998) provided the most detailed mineralogical description of the Rurutu samples. Miocene alkali basalts contain phenocrysts (magnesian olivine, calcic clinopyroxene, and bytownite plagioclase) within a microcrystalline groundmass (olivine, clinopyroxene, labrador plagioclase, and titanomagnetite). Phenocrysts abundance was estimated in samples 74-386 (10% Pl, 5% Cpx, 5% Ol) and 74396 (15% Cpx, 10% Ol, 5% Pl) (Duncan, 1975). Distinctive features of the Miocene hawaiites include the presence of titanomagnetite and millimetric bytownite phenocrysts, and andesine and low magnesium olivine within the groundmass. Miocene tholeiites are characterized by the presence of clinopyroxene phenocrysts relatively poor in calcium. The Pleistocene samples consist mainly of a microcrystalline groundmass (olivine, calcic pyroxene, andesine plagioclase, titanomagnetite, plus nepheline in basanites and alkali feldspars in hawaiites) with rare phenocrysts (olivine and clinopyroxene in basanite plus titanomagnetite in hawaiites). 4. Analytical methods 4.1. Trace element concentrations Trace element concentrations were measured on acid-digested powders using a quadrupole ICP-MS (Agilent 7500, Laboratoire de Géodynamique des Chaînes Alpines). The reaction cell (He mode) was used to reduce mass interferences on Sc. Although of small magnitude, the instrumental drift was monitored and corrected using a multispike (Be–As–In–Tm–Bi). Signal was calibrated from the repeated analyses of the BR standard, using for reference the values recommended by Eggins et al. (1997). The external reproducibility of the method, as estimated by running repeatedly the BHVO-2 standard as an unknown sample is less than 5% (2σ error). 4.2. Lithium isotopic compositions 4.2.1. Leaching experiments Samples selected for this study are amongst the freshest available from Rurutu Island but were nevertheless leached before measuring Li isotopic compositions in order to remove the potential effects of alteration. One hour leaching experiments performed on BHVO-2 powder indicate moderate Li loss (≤ 10%) and small isotopic fractionation (≤0.5‰) in the leached powder as long as dilute acid (0.5 to 3N HCl) is used (Fig. 2). In detail, δ7Li of the leached powder seems to slightly decrease when leaching occurs at room temperature, while an opposite trend is suggested by an experiment conducted at higher temperature. The 0.5N HCl leaching yields δ7Li in the residue indistinguishable from the composition of the unleached powder. It was thus decided to apply this procedure to the Rurutu samples. For comparison, some of the Rurutu samples were also analyzed without leaching as well as after stronger leaching (3N HCl, ~60 °C) (Table 1).

4.2.2. Li purification Accurate and precise determination of the isotopic composition of lithium in basaltic samples requires total extraction and high-level purification of lithium. This is a difficult task particularly because of Na. It is present at percent level in basalts and has similar selectivity as Li for cation exchange resins when eluting with inorganic acids. Mixtures of inorganic acids with methanol (or ethanol) significantly improve Li–Na separation with undesirable consequences, such as the release of Na and resin degradation. Resin degradation not only releases organic material, which must be subsequently eliminated (Jeffcoate et al., 2004), but may also alter resin exchange capacity. The use of dilute HCl is a viable alternative (James and Palmer, 2000) despite the close selectivity of Li and Na for cation exchange resins. We followed this method and improved Na–Li separation by using highresolution columns (high height/diameter ratios). Rock powders or chips (50 mg) were gently leached (0.5N HCl for 1 h at room temperature) and digested in a mixture of concentrated HClO4 (0.3 ml) and HF (1 ml) for 24 h at 80 °C. Acid evaporation was conducted at 110° for two days. Conversion to chlorides was made with 1 ml of HCl 6N. Following evaporation, samples were dissolved in 1 ml of HCl 0.5N and total dissolution of samples was verified by centrifugation before samples were loaded on columns. Two types of columns filled with AG50W-X8 200–400 mesh cationic resin (capacity of 1.7 meq/ml) were used to purify lithium. The first column (diameter: 0.64 cm, height: 24 cm, and volume: 7.7 ml) has a capacity of 13.1 meq. The total cation loading does not exceed 10% of the resin exchange capacity, which is required for efficient separation. The second column (diameter: 0.4 cm, height: 15 cm, and volume: 1.88 ml) has a capacity of 3.2 meq. The first column was calibrated using rocks with variable Fe/Li and Mg/Li ratios (basalte, picrite and olivine). The second column was calibrated with a basaltic sample processed through the first column and with a synthetic solution to monitor the Na peak. Column calibrations were checked at the beginning and the end of the study, which lasted about one year. The chemistry blank measured by quadrupole ICP-MS is less than 0.1 ng. 4.2.3. Mass spectrometry Organic material and Cl left by chemical separation were eliminated with a few drops of concentrated HNO3. Purified lithium was then taken up by 0.05N HNO3 for 7Li/6Li measurement by MC-ICPMS (Nu 500, Ecole Normale Supérieure de Lyon). Samples were introduced through a desolvator (Nu DSN) at a rate of 100 μl min− 1, yielding a total Li beam of 4 to 6 V for 70 ng/g Li solutions. Mass discrimination was monitored externally with IRMM-016 standard using a sample-standard bracketing technique. Mass fractionation varied between 12 and 16% per amu over one year. The washout procedure (300 s with 0.65N HNO3 and 300 s with 0.05N HNO3) reduced the Li signal by a factor of 104. All samples were diluted to have the same concentration as the standard. The amount of Li extracted (N250 ng) allows multiple measurements of each sample. Repeated analysis of the USGS BHVO-2 standard (batch 759) yielded 7Li/6Li= +4.3‰ ± 0.4 (2σ, n = 8). 5. Results Lithium isotopic compositions and new trace element concentrations are reported in Table 1, along with the previously published data

Notes to Table 1: Trace element concentrations are μ/g. δ7Li = ((7Li/6Li)sample / (7Li/6Li )IRMM − 1) ⁎ 1000. Leaching procedure: (a) unleached, (b) 0.5N HCl, (c) 3N HCl, 60°C. δ7Li in italics: values thought to reflect secondary processes (see text). δ7Li in bold: working values. n: repeated analysis of the same dissolution. N: fully repeated analysis. dup: duplicate analysis. Samples ages, major element data, lost on ignition (LOI) and 206Pb/204Pb are from Duncan (1975), Duncan and McDougall (1976), Diraison (1991), Chauvel et al. (1997) and Guille et al. (1998).

Table 1 New trace element concentrations and Li isotopic compositions of Rurutu lavas. 74-386 Alk. basalt 8.4 Old

74-390 Alk. basalt 12.0 Old

74-396 Alk. basalt 12.0 Old

Li Sc V Rb Sr Y Zr Nb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Yb Lu Hf Ta Pb Th U LOI (wt.%) Mg# CaO/Al2O3 206 Pb/204Pb

6.43 34.0 387 15.9 462 25.3 168 30.7 0.11 143 23.4 48.5 6.29 26.5 5.98 1.86 5.74 0.85 4.78 0.88 2.28 1.70 0.24 3.93 1.91 1.71 2.53 0.66 1.58 51.9 0.87 21.01

9.13 24.9 369 17.0 556 29.2 223 45.5 0.33 283 33.1 67.1 8.59 34.6 7.37 2.35 6.79 1.00 5.54 1.02 2.58 1.95 0.29 4.99 2.80 2.57 3.86 1.03 1.35 48.0 0.77 21.09

7.08 26.9 300 19.4 592 25.7 215 45.0 0.16 220 32.8 68.2 8.60 34.5 7.20 2.21 6.34 0.92 5.02 0.9 2.23 1.69 0.23 4.68 2.75 2.74 3.64 0.98 1.06 55.7 0.73 21.42

Material (δ7Li)

Powder

Powder

Powder

RR01 Alk. basalt

RR67 Ol. Tholeiite 12.3 Old

RRT-32 Alk. basalt 14.2 Old

RRT-037 Alk. basalt 11.3 Old

122M Basanite 1.60 Young

74-388 Hawaiite 1.07 Young

74-392 Hawaiite 1.05 Young

74-394 Basanite 1.85 Young

RR20b Hawaiite

Old

RR03 Hawaiite 12.2 Old

Young

13.7 35.6 338 9.86 479 27.3 207 47.3 0.240 175 29.0 61.3 8.02 32.1 7.02 2.21 6.37 0.96 5.15 0.98 2.44 1.81 0.260 4.86 3.10 2.14 3.16 0.89 1.42 55.0 0.98 21.08

12.0 13.4 137 15.7 769 33.9 255 52.0 0.340 263 35.5 75.4 9.92 39.3 8.71 2.85 7.93 1.18 6.35 1.16 3.01 2.28 0.33 5.59 3.20 2.53 3.74 1.01 2.17 45.0 0.48 20.98

4.95 32.3 308 4.43 352 24.5 126 19.7 0.03 85.6 14.8 31.7 4.27 18.7 4.91 1.64 5.12 0.78 4.54 0.85 2.15 1.65 0.24 3.30 1.24 0.77 1.52 0.43 2.87 56.2 0.80 20.88

6.32 27.4 395 5.15 511 23.6 152 27.6 0.240 116 20.5 43.1 5.64 23.2 5.34 1.79 5.25 0.78 4.37 0.81 2.09 1.60 0.23 3.65 1.70 1.56 2.14 0.58 1.20 50.1 0.72 21.06

5.27 39.0 396 12.4 483 25.2 186 35.7 0.2 170 26.1 55.1 7.15 28.8 6.34 2.00 5.83 0.86 4.80 0.88 2.26 1.74 0.24 4.50 2.23 1.88 2.81 0.75 0.76 57.3 0.98 20.92

7.95 21.6 262 22.0 884 35.7 285 65.0 0.22 299 49.7 108 13.4 54.9 11.1 3.35 9.54 1.33 7.03 1.24 3.10 2.25 0.33 6.24 4.00 2.79 5.15 1.37 1.63 57.7 0.77 20.22

9.66 18.8 264 33.8 933 36.9 359 94.0 0.29 354 57.8 122 14.9 60.2 11.8 3.49 9.82 1.40 7.24 1.27 3.23 2.39 0.35 7.63 5.44 4.08 6.17 1.59 -0.24 50.3 0.58 20.35

11.3 14.3 155 30.5 1054 40.7 371 88.8 0.28 363 63.7 134 16.3 62.0 12.1 3.71 10.1 1.45 7.65 1.37 3.54 2.72 0.4 7.69 5.09 3.68 7.10 1.83 0.58 45.0 0.48 20.33

12.0 15.7 315 71.7 1770 43.9 563 155 0.690 1072 116 222 26.2 98.0 17.0 5.10 13.0 1.76 8.60 1.48 3.66 2.61 0.36 11.6 8.70 6.21 11.4 2.65 0.72 59.7 0.77 19.11

13.1 11.3 125 42.9 1295 46.4 480 120 0.48 482 91.9 187 22.4 85.6 15.8 4.69 12.5 1.78 8.85 1.60 3.90 2.93 0.45 9.60 6.64 5.11 9.46 2.55 0.37 39.9 0.51 20.28

chip

chip

Powder

Powder

Powder

Powder

Powder

Powder

Powder

chip

δ7Li (a) n sd δ7Li (b) n sd

RR27 Hawaiite

RR59 Basanite

Young

RR5 Hawaiite 1.11 Young

RRT-60 Hawaiite 1.16 Young

BHVO-2

Young

RRT-015 Nephel. 1.23 Young

15.2 13.2 135 41.5 1188 46.1 454 113 0.41 444 81.5 171 20.7 79.3 15.0 4.52 12.5 1.70 8.61 1.60 4.02 2.98 0.45 9.59 6.52 4.29 8.44 2.18 0.47 42.2 0.56 20.31

13.3 14.5 169 31.0 1256 48.7 412 99.6 0.4 456 81.3 172 21.4 84.2 16.0 4.77 13.1 1.81 9.22 1.62 4.01 2.90 0.42 8.70 5.74 4.07 7.95 2.09 0.66 48.2 0.57 20.29

10.3 15.1 212 35.9 1064 41.2 427 108 0.39 386 73.3 154 18.3 71.9 13.4 3.98 11.0 1.57 8.05 1.41 3.50 2.61 0.390 8.66 6.28 3.41 7.61 2.06 0.61 46.7 0.59 20.35

10.4 19.2 263 32.7 943 38.1 367 93.0 0.31 360 60.3 128 15.6 61.9 12.1 3.66 10.1 1.45 7.53 1.33 3.29 2.46 0.36 7.74 5.51 2.48 6.24 1.60 0.15 50.4 0.57 20.35

11.0 13.9 153 29.1 1105 40.2 360 85.5 0.31 342 61.8 130 15.8 60.6 11.7 3.63 10.0 1.40 7.62 1.37 3.47 2.69 0.4 7.64 4.95 3.34 6.75 1.82 0.70 46.5 0.49 20.31

4.98 32.6 337 9.29 398 28.5 182 18.5 0.1 126 15.0 36.2 5.31 24.3 6.14 1.98 5.97 0.902 5.22 0.98 2.55 1.94 0.280 4.33 1.18 1.46 1.22 0.41

chip

chip

Powder

Powder

Powder

Powder

5.2 2 0.0

3.2 2 0.1

4.3 N=8 0.2

4.6 4 0.2

2.9 3 0.2

4.1 N=2 0.1

9.0 2 0.1 5.7 2 0.2

6.8 3 0.1

5.6 2 0.3

7.7 2 0.1

δ7Li (b) dup n sd

7.5 3 0.3

δ7Li (c) n sd

4.7 3 0.0

6.8 4 0.2

4.0 2 0.1

6.5 2 0.1

8.0 2 0.0

4.8 2 0.1

3.3 3 0.0

3.9 3 0.2

3.6 2 0.1

6.3 4 0.1

7.0 4 0.1

6.6 3 0.0

6.8 3 0.2

7.8 3 0.0

5.9 3 0.1

4.4 3 0.2

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

Sample name Rock type Age (Ma) Group

7.8 2 0.1 6.6 3 0.1

10.1 2 0.1

4.8 n=2 0.2

459

460

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

Fig. 2. Leaching experiments. Fifty milligrams of the USGS BHVO-2 powder (batch 759) were leached with distilled HCl (0.5N to 3 N) for 1 h at room temperature. The amount of Li leached was quantified by ICP-MS assuming an initial concentration of 4.8 μg/g in the unleached powder. Fifty milligrams of BHVO powder was also leached with 3N HCl for 1 h in a warm (~60 °C) ultrasonic bath. For this experiment, the Li isotopic composition of both residue and leachate were analyzed and the amount of Li leached was estimated during isotopic measurement by comparing the intensity of the Li signal to that of the bracketing standard (IRMM-16). The isotopic composition of the unleached BHVO-2 powder acquired during this study is shown for comparison. Error bars (1σ) are inferred from repeated analysis of the same dissolution (not shown for BHVO-2 bulk composition).

on Rurutu samples (Chauvel et al., 1997). New trace element data compare within 10% (REE, Sr, Zr, Hf, and U), 20% (Rb, Y, Ba, Ta, Pb, and Th) and 30% (Li, Sc, and Nb) with the previously published data. Large differences (N50%) in Rb, Sr, Zr, and Nb concentrations are observed only for sample 74-394. Lithium concentration varies between 5.0 and 13.7 μg/g in old lavas and between 8.0 and 15.2 μg/g in young lavas. These values fall within the range of Polynesian basalts, most samples analyzed so far having between 3 and 17 μg/g of lithium (Dupuy et al., 1988, 1989; Dostal et al., 1996; Chan et al., 2009). Lithium abundance correlates negatively with indices of pyroxene fractionation (CaO/Al2O3 and Sc concentration) and shows strong positive correlations with heavy rare earth elements (such as Yb) within each lava group(Fig. 3). Considering the whole data set, there is a large isotopic overlap between HIMU (δ7Li from 4.0 to 7.9‰) and non-HIMU-lavas (δ7Li from 2.9 to 7.0‰) even if HIMU samples with δ7Li below 5‰ and non-HIMU samples with δ7Li above 5‰ are few (one out of eight samples and three out of ten samples, respectively). Increasing leaching strength indicates that the elevated δ7Li values (i.e. above the normal mantle signature of 3–5‰) measured in both HIMU and non-HIMU samples are robust features (sample RR01 being an exception). On the other hand, a detailed inspection of the data set reveals that two Miocene samples (RR01 and RR67) are unsuitable for a Li isotopic study of the magma sources. Sample RR01 has a large excess of lithium (13.7 μg/g whereas no more than ~ 6 μg/g is expected based on its high Sc content) and its isotopic composition shows the significant influence of leaching. The least fresh sample, sample RR67 with LOI of 2.9%, shows anomalous trace element ratios (e.g. Nb/Cs of 779 instead of less than 330) and distinctly light Li composition (+ 4.0‰) amongst the oldest lavas. Excluding samples RR01 and RR67, the Li isotopic signatures of old (+5.6 bδ7Li b +7.9‰) and young (+2.9b δ7Li b +7.0‰) lavas still overlap (Fig. 4), but the overlap is now essentially due to the elevated δ7Li of the most differentiated (Li-rich) Pleistocene samples. 6. Discussion 6.1. Lithium abundances The variations in lithium abundance (a factor of 3) in the Rurutu lavas are unexpectingly large for a moderately incompatible element.

Fig. 3. Li abundance systematics in Rurutu lavas. Li concentration is plotted versus (a) CaO/ AlO3, (b) Sc and (c) Yb concentrations. Analytical uncertainty is equal or smaller than the symbol size. (b) The well-developed Li-Sc correlation is thought to reflect clinopyroxene fractionation. This process has been modeled using a starting composition of 5 μg/g Li and cpx/liq 40 μg/g Sc, fractional crystallization, D Licpx/liq = 0.1 and DSc = 2. (c) The increase of Li/Yb with Yb concentration within each lava group (dashed lines) is consistent with highpressure crystallization of clinopyroxene. The difference in Li and Yb abundances between the two groups of lava, while Li/Yb remains on average unchanged (plain lines), may result from low-pressure fractionation (Ryan and Langmuir, 1987; Ryan and Kyle, 2004).

For instance, a similarly incompatible element such as Yb varies only by a factor of 2. As a consequence extended rare earth element patterns normalized to primitive mantle display both positive and negative Li anomalies (Fig. 5a and b) when Li is positioned between Er and Yb (i.e. at Tm position). The high variability of Li concentration in Rurutu lavas could reflect various processes ranging from post-

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

461

Fig. 4. Relationships between lithium isotopic composition and lithium content in Rurutu lavas. Samples showing evidence for surface or near-surface modification of Li content or Li isotopic composition (RR01 and RR67) are not shown in this plot. The composition of a liquid undergoing clinopyroxene fractionationis modeled assuming a starting composition of 5 μg/g of Li and δ7Li = 4.7‰, clinopyroxene with2 μg/g of Li and δ7Li = − 5.0‰, and partition coefficients as in Fig. 3b. Clinopyroxene has generally low Li content (≤2 μg/g) and light Li composition (δ7Li b + 3‰) (e.g. Seitz et al., 2004; Jeffcoate et al., 2007; Chan et al., 2009). The values used in the model are intended to estimate the maximal effect of clinopyroxene fractionation.

eruption alteration or weathering (e.g. Dostal et al., 1996) to the less well-known vapor transfers, which may influence the Li distribution in magma reservoirs shortly before eruption (Berlo et al., 2004) and across lava flows during post-eruptive degassing (Kuritani and Nakamura, 2006). With the exception of sample RR01, which has a large excess of Li, there is no evidence for such processes controlling Li variability in the Rurutu samples. Instead, amongst the trace elements, lithium best correlates (negatively) with Sc (Fig. 3b) suggesting that clinopyroxene fractionation plays a first order role in controlling Li abundance. It has been previously shown that the incompatibility of Li relative to Yb depends on the crystallizing assemblage (Ryan and Langmuir, 1987; Ryan and Kyle, 2004). In particular, because Li is more incompatible than Yb in clinopyroxene, but less incompatible in plagioclase, Li/Yb in lavas may increase with the crystallization depth. We suggest that the anomalous variability of Li in Rurutu lavas results from such process and that Li anomalies in trace element patterns reflect the imperfect positioning of Li, which is sometimes more (positive anomalies) and sometimes less (negative anomalies) incompatible than Yb. Perhaps fortuitously, the average patterns of each group of lava show no lithium anomaly (Fig. 5c). The trace element patterns of OIB generally show negative Li anomaly (Fig. 5c), possibly reflecting depletion of Li relative to HREE in plume sources (Dostal et al., 1996). At first glance, the absence of such lithium anomaly in the average trace element patterns of the Rurutu lavas suggests that the HIMU mantle reservoir is not depleted in lithium. However, as discussed above, the highest Li/Yb ratios of the Rurutu lavas likely reflect high-pressure fractionation of clinopyroxene. The less evolved sample (RRT-037, with 5.3 μg/g of Li) has a Li/Yb ratio of 3.0, which is only slightly higher than the average OIB (Li/ Yb = 2.6, Li = 5.6 μg/g) (Sun and McDonough, 1989). Thus, if the HIMU mantle reservoir is less depleted in Li compared to other OIB sources, the compositional difference must remain small.

Fig. 5. Extended rare Earth elements patterns. Concentrations are normalized to primitive mantle abundances (McDonough and Sun, 1995). (a) Old Rurutu lavas, (b) Young Rurutu lavas, (c) Average patterns of old and young Rurutu lavas compared to the average pattern of OIB (Sun and McDonough, 1989).

6.2. Lithium isotopic compositions Per-mil level fractionation of Li isotopes was shown to occur during low temperature alteration (Chan et al., 1992, 2002), weathering (Pistiner and Henderson, 2003; Huh et al., 2004) and diffusion in

magmatic systems (e.g. Richter et al., 2003; Lundstrom et al., 2005; Parkinson et al., 2007). The extent to which such shallow-level processes influence the Li isotopic variation in the Rurutu samples

462

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

must be thoroughly assessed before the mantle source heterogeneity can be investigated. The lack of co-variations between δ7Li and LOI or ratios of trace elements that are less versus more fluid–mobile (e.g. Nb/U, Nb/Rb, and Nb/Cs) rules out a first order influence of posteruption processes on the Li isotopic variations (sample RR67 being an exception). We previously noted a relationship between δ7Li and the extent of differentiation of Pleistocene lavas, the highest δ7Li (N + 5‰) characterizing the most evolved samples (those plotting at the highLi, low-Sc end of the Li-Sc correlation). Parkinson et al. (2007) who studied lavas from Salomon Islands suggested that whole rock isotopic compositions reflect a mixture of (1) Li-enriched matrix with elevated 7Li/6Li and (2) Li-depleted clinopyroxene phenocrysts with low 7Li/6Li. They estimated that because of the presence of 30% phenocrysts with Li composition 5 to 10‰ lighter than the matrix, whole-rocks must have 7Li/6Li 1 to 3‰ lower than their mantle source. Such an interpretation does not apply to the Rurutu samples because (a) major-trace element data suggest removal, and not accumulation of clinopyroxene, and (b) the anomalous 7Li/6Li ratios are higher, and not lower than the accepted mantle values (typically + 3 to + 5‰).On the other hand, removal of clinopyroxene containing isotopically light lithium could significantly increase 7Li/6Li in lavas (Fig. 4). Such a process could explain the elevated δ7Li of the most Li-rich (and Scpoor) samples from Rurutu Island (RR03, RR05, RR27, and RR20b). Samples affected neither by surface or near-surface Li mobilization nor by extensive clinopyroxene removal are thought to reliably record the mantle signature. In the following discussion only such samples are considered. Supporting previous suggestions (Ryan and Kyle, 2004; Nishio et al., 2005; Chan et al., 2009), the difference in Li isotopic composition between the young and the old Rurutu lavas is most easily explained if the HIMU mantle has distinctly high 7Li/6Li ratios (Fig. 6a). The composition of the most primitive and freshest HIMU-type sample analyzed (RRT-037) raises the possibility that HIMU mantle has δ7Li up to +7.9‰. Such a signature contrasts with the relatively light signature (+4.9 b δ7Li b 5.5‰) of Tubuai volcanism (Nishio et al., 2005; Chan et al., 2009; I. Vlastélic, unpublished data) and the moderate δ7Li (b + 6.2‰) recently estimated by Lassiter et al. (2008) for the true HIMU mantle on the basis of olivine phenocrysts analyses (Fig. 6b). The Li isotopic variability (+5.4 b δ7Li b +7.9‰) of the relatively fresh and unaltered HIMU-type lavas of Rurutu (this study and Nishio et al., 2005) is surprisingly large given the small range of Pb isotopic composition (20.88 b 206Pb/204Pb b 21.42). These ~2‰ variations of δ7Li could reflect the diffusion-driven isotopic fractionation in magmatic system, although the precise influence of such a process on the whole rock data is difficult to assess. Alternatively, lithium isotopic variations at nearly constant Pb composition could reflect the nearly linear arm of a mixing hyperbola between the depleted mantle and the pure HIMU mantle (Ryan and Kyle, 2004), although unrealistically low Li/Pb (≤ 1) is required in the HIMU mantle to satisfactorily explain our data. Another possibility is that 7Li/6Li is decoupled from slowly diffusing tracers because of diffusion-driven fractionation of Li isotopes occurring in the mantle. In the section below, we will quantitatively examine this third scenario.

6.3. Modeling Li isotope diffusion from subducted crust into the mantle Long-lived, heavy radiogenic isotopes diffusing slowly in the mantle proved suitable to detect small (centimeter scale) and old (1–2 Ga) mantle domains influenced by the subduction processes (Hofmann and Hart, 1978). Halama et al. (2008) recently pointed out that the situation is more problematic for a fast diffusing element, such as Li. Modeling Li diffusion out of a sphere, they found that Li heterogeneities less than 2 km in size are totally erased in less than 100 Ma, and they suggested that Li isotopic heterogeneities must be homogenized just as fast. If this reasoning was correct, it would

Fig. 6. Lithium isotopic composition versus 206Pb/204Pb. (a) Li–Pb isotopic relationship in Rurutu lavas. Samples whose Li isotopic compositions show evidence for surface or near-surface modification (RR01 and RR67) are not shown in this plot. (b) Li–Pb isotopic relationship in basalts from Cook–Austral islands. Only compositions thought to reflect the mantle signature are shown (see related papers). Ol: Li analyses performed on olivine separates (Chan et al., 2009). WR: Li compositions measured on whole-rock samples (Nishio et al., 2005; this study). The external error (2σ) on lithium isotopic composition is from this study (repeated analysis of BHVO-2). Uncertainty on Pb isotopic composition is less than the symbol size.

remain unclear how heavy lithium signatures could survive over 1– 2 Ga time scales in the source of the HIMU type volcanism. The conclusion reached by the Li diffusion modeling by Halama et al. (2008) is perhaps premature for two reasons: 1) the authors presented the most favorable circumstances to erase the heterogeneous initial conditions. First, the assumption of constant Li concentration at the sphere's surface implies that the surrounding mantle is strongly convecting and instantly homogenizes all lithium diffusing out of the sphere. Second, a sphere favors diffusive exchange more than the other commonly used symmetric geometries (i.e. slab, cylinder) because of its lowest possible surface to volume ratio among these. Conversely, the slab geometry and the continuous diffusion of Li in the mantle are equally realistic conditions, which are less efficient to erase the heterogeneities. 2) A significant isotopic fractionation can occur if diffusing isotopes have sufficiently different diffusivities (Richter et al., 2003). The model for the preservation of the Li isotope heterogeneity must therefore consider the diffusion of isotopic species independently. The diffusion of the Li isotopes out of a lithium-rich, isotopically heavy altered oceanic crust (approximated by a slab geometry) is modeled here using the equation of Fick's second law of diffusion: h i i A Li At

h i1 0 i A @ / A Li A DLi = Ax Ax

ð1Þ

where t is time, x the distance in one dimension, D the diffusion coefficient in a medium ϕ, [iLi] the concentration of Li with atomic

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

463

mass of i. The equation was solved for two sets of boundary conditions (Eqs. (2) and (3)) known as the diffusion in an infinite medium with a finite source with half thickness of δ: h

h

i

i

Li

Li

i

j

AOC

t = 0; − δbxbδ

i

j

t = 0;xb − δ

ð2Þ

= C0

=

h

i

Li

i

j

t = 0;x N δ

=

h

i

Li

i

j

Mantle

x = ∞;x = − ∞

= C0

ð3Þ

An analytical solution (Eq. (4)) was used for conditions where Mantle (D being constant over time and space) (Crank, 1975): DAOC Li = DLi h

i

Li

h

i

i x;t

=

Li

iAOC 0

2

0

1

x+δ x−δC B @erf qffiffiffiffiffiffiffiffiffi − erf qffiffiffiffiffiffiffiffiffiA / 2 DLi t 2 D/Li t

ð4Þ

For slightly more complex boundary conditions, where Li parti≠ DMantle , we solved tions between the mantle and the crust or DAOC Li Li numerically the equation by the Crank–Nicholson finite-difference method. The accuracy of the numerical solution was confirmed by comparing it to the analytical solution for the same boundary condition. The diffusion coefficient of Li in the mantle being poorly known, it is assumed to be in the order of 10− 10 m2/s (Coogan et al., 2005; Dohmen et al., 2008). Following the case of silicate melt, it is assumed that 6Li diffuses 3% more rapidly than7Li (Richter et al., 2003), although recently Dohmen et al. (2008) reported up to 6% difference in olivine. Unless otherwise stated, the initial compositions used are those of unprocessed altered oceanic crust ([Li] = 7.5μg/g, δ7Li = +10‰; Chan et al., 2002) and primitive mantle ([Li] = 1.7 μg/g, δ7Li = +3.2‰; Seitz et al., 2007). The thickness (2δ) of the altered oceanic crust is assumed to range between 700 and 1000 m (see Fig. 2 of Chan et al., 2002). In agreement with the model of Halama et al. (2008), we find that the chemical gradient between altered crust and surrounding mantle equilibrates in relatively short time periods. Note that the assumed value of DLi (5 × 10−10 m2/s) is comparable to the fastest diffusivity used by Halama et al. (2008). For instance, 90% of the initial abundance contrast between mean altered oceanic crust and mantle is erased after 20 to 500 Ma, depending on Li diffusivity and slab thickness (Fig. 7a). This result could explain the absence of significant Li enrichment in the HIMU mantle. Contrary to the conclusion of Halama et al. (2008), the isotopic disequilibrium persists over much longer time periods (Fig. 7b), because isotopic fractionation is created as long as Li diffuses. For an altered crust thickness (2δ) of 700 m and DLi = 5 × 10−10 m2/s, 40% of the initial isotopic disequilibrium persists after 1.5 Ga, while chemical equilibrium is nearly attained (N96%) (Fig. 8). Long-term preservation of the isotopic disequilibrium partly results from very early (b50 Ma), large isotopic fractionation, which raises the mean slab composition up to nearly two times the initial value. The cause of this early fractionation is that 6Li diffuses preferentially out of the slab as long as the concentration difference between the mantle and the altered crust exists. As a surprising consequence, the subducted crust returns to its initial isotopic composition (δ7Li = +10‰) after 160 Ma, while the Li concentration has dropped from 7.5 to 2.3 μg/g (Fig. 8). A consequence of the slow decrease of δ7Li is that the subducted crust could preserve a Li isotopic signature distinctly heavier than the mantle for a period of time in excess of one billion years provided that Li diffusivity is slower than ca. 5 × 10−10 m2/s. For example, for DLi = 5 × 10− 10 m2/s, the crust retains δ7Li N + 6.0‰ after 1.4 Ga, and δ7Li N + 5.5‰ after 2 Ga. It should be noted that lithium concentration and δ7Li may slightly decrease during dehydration of the oceanic crust (Marschall et al.,

Fig. 7. Modeling lithium diffusion from altered oceanic crust to mantle. The average composition of altered oceanic crust ([Li] = 7.5 μg/g, δ7Li = + 10‰ (Chan et al., 2002)) diffusively equilibrating with mantle ([Li] = 1.7 μg/g, δ7Li = + 3.2‰ (Seitz et al., 2007) is shown as a function of time. It is assumed that Li partitions equally between altered oceanic crust and mantle. The ratio of the diffusion coefficients of 7Li and 6Li (expressed as D7/D6 = (m6/m7)β) in the mantle being unknown, we use the mass fractionation factor (β = 0.215) measured in basaltic melts (Richter et al., 2003). Results are shown for four different sets of parameter (altered crust thickness (2δ), DLi), which are, from top curve to bottom curve: 1000 m, 5 × 10− 10 m2/s; 700 m, 5 × 10− 10 m2/s; 1000 m, 5 × 10− 9 m2/s; and 700 m, 5 × 10− 9 m2/s. (a) Mean Li concentration of altered oceanic crust versus time. (b) Concentration-weighted mean of Li isotopic composition of altered oceanic crust versus time. The inset shows the early 7Li/6Li fractionation, which results from 6Li diffusing faster than 7Li out of the subducted crust. See text for more details.

2007). In an extreme case where there is no abundance and no isotopic contrast between crust and mantle after dehydration, there is no effective diffusion. If there is isotopic contrast but no abundance contrast, there is no early increase of 7Li/6Li in the crust, and the isotopic disequilibrium only decreases through time. Note that the less the abundance contrast, the smaller the peak of early isotopic fractionation (see Fig. 9 of Richter et al. (2003)). Clearly, the initial abundance contrast is a critical parameter as well as the diffusivity. In the case of the composition of the dehydrated oceanic crust proposed by Marschall et al. (2007) (Li ~ 4.3 μg/g and δ7Li ~ +7‰), δ7Li of the mean crust increases by ~5‰ during the first tens of millions of years and preserves δ7LiN +5‰ over 1.5 Ga.

464

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

Fig. 8. The extent of isotopic disequilibrium between mean altered oceanic crust and mantle plotted versus chemical disequilibrium. References for chemical and isotopic disequilibria are 100% for the initial condition and 0% for complete homogenization. The curve is shown for a crust thickness of 700 m and DLi = 5 × 10− 10 m2/s. Time (in Ma) is indicated. Slower diffusivity and thicker crust would give older age at the indicated tick marks, while following the same trajectory. Varying altered oceanic crust thickness by a factor of five contributes less than 1% variation to the maximum isotopic composition; the shape of the curve remains essentially unchanged.

The thickness of the altered oceanic crust is another critical parameter that is poorly constrained. A seven kilometer-thick initial oceanic crust may stretch or accumulate at some depths in the mantle. In our calculation, we have assigned 1 km as a reference thickness for the altered crust. Certainly, anything thicker than 1 km would help preserve the heterogeneity. On the other hand, for a 200 m thick altered crust, approximately 15% of the initial isotopic contrast is preserved at 1.5 Ga (δ7Li = 4.2‰ for the initial conditions used here). For a 20 m thick altered crust for the same duration, only 1.5% of the initial isotopic contrast is preserved (δ7Li = 3.3‰). Hence, the survival of Li isotopic heterogeneities in the HIMU mantle is possible only if the size of the recycled domains is sufficiently large (i.e. N102 m).

We have so far emphasized the preservation of the Li isotopic heterogeneity represented by the mean of the altered crust composition. In detail, diffusion-driven fractionation generates the highest 7Li/ 6 Li within the altered crust, and the lowest 7Li/6Li within the mantle. For instance, after 1.5 Ga the difference between the center and the edge of the altered crust is less than 0.03‰ (1 km-thick altered crust, DLi = 5 × 10− 10 m2/s), but large differences in 7Li/6Li characterize the nearby mantle, from +7.4‰ near the interface to + 2.1‰ at 13 km from the crust (Fig. 9). Note that the value of + 2.1‰ is lower than the initial mantle value (+ 3.2‰). The isotopic variation within the altered crust reaches its maximum around 6 Ma (with 6.5‰ variation), and is less than 0.3‰ after 200 Ma. Thus, even if the old altered crust is most likely homogeneous with respect to Li isotopes, a considerable Li isotopic heterogeneity remains in the system made of crust and surrounding mantle. Because of diffusion, the Li isotopes display a complex spatial distribution near the subducted crust (Fig. 9), which is not expected in the case of the slowly diffusing and heavier isotopic tracers. Therefore, the lack of a robust correlation between Li and Pb isotopes (Fig. 6) is possibly inherited from the source mantle and may not necessarily result from near surface processes. The analytical solution used for Figs. 7 and 8 (Eq. (4)) is useful to illustrate the behavior of the Li isotopes but lacks some complexities that may influence the conclusion. For example, because of the difference in mineralogy, the Li diffusivities may be different in the crust and the mantle. For the same reason, it is expected to partition Li at the crust– mantle interface. For the case of DMantle = 0:1 × DAOC = 5 × 10 − 11 m2 s, Li Li the mean crust value is +12.1‰ at 1.5 Ga, and for the opposite case, DMantle = 10 × DAOC = 5 × 10 − 9 m2 = s while keeping all the initial Li Li conditions the same, the mean crust value is +4.7‰ for the same time period. When Li diffuses faster in the mantle than in the crust, the diffusion-driven isotopic fractionation diminishes (e.g. maximum contrast of 1.8‰ at 1.5 Ga for the conditions given above). The system behaves in an opposite manner when Li diffuses slower in the mantle (e.g. maximum contrast of 11.9‰ at 1.5 Ga). For a partition coefficient between   AOC=Mantle the crust and the mantle KLi of 2, the mean crust value is AOC=Mantle

= 0:5, it is +7.6‰ at 1.5 Ga. Because +6.6‰ at 1.5 Ga, and for KLi AOC=Mantle the mean crust value at 1.5 Ga is +7.4‰ for KLi = 1, small variation of the partition coefficient gives a minor influence to the AOC=Mantle preservation of the isotopically distinct crust. Our estimate of KLi ranges from 0.28 to 0.71, which are derived from a compilation of experimental and natural data accounting for variation of mineral modes (Ryan and Langmuir, 1987; Brenan et al., 1998; Seitz and Woodland, 2000; Marschall et al., 2006; Ottolini et al., 2009). Lithium partitioning at the crust–mantle interface would not change our conclusions. However, the contrast of diffusivity between the crust and the mantle has a significant consequence. Here, we reiterate our position that the diffusivity used here is considered the fastest, and a significant isotopic contrast of recycled altered oceanic crust is maintained for most of the cases considered here. We interpret that the preservation of the contrasted Li isotopic composition of the altered oceanic crust over 1.5 Ga is likely in the mantle. 7. Concluding remarks

Fig. 9. Li isotopic distribution in subducted altered oceanic crust and nearby mantle after 1.5 Ga. Calculations are done for 1 km-thick altered oceanic crust and DLi = 5 × 10− 10 m2/s. Other parameters are as in Fig. 7.

Recycling isotopically heavy Li in HIMU-type volcanism implies (1) the preservation of the heavy Li signature of the altered oceanic crust during subduction despite dehydration processes, an idea recently discussed by Marschall et al. (2007) and Chan et al. (2009). (2) Although Li diffuses much more rapidly than the heavy radiogenic tracers, the Li isotope heterogeneities must survive in the mantle over Ga time scales. We have shown that a kilometer-thick altered oceanic crust loses most its Li excess in a matter of a few tens of millions years (consistently with Halama et al. (2008)) but could preserve 7Li/6Li distinctively higher than the ambient mantle over a time period in excess of 1.5 Ga. Thus, the commonly used approximation that

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

diffusion erases the Li concentration contrast and the isotopic disequilibria over similar time-scales is incorrect. Our results also suggest that the decoupling between Li and radiogenic isotopes often observed in oceanic basalts could be a natural consequence of the subducted crust equilibrating with the mantle, and may not entirely result from near-surface processes. It should also be emphasized that Li isotopic heterogeneities in the mantle do not necessarily require input of material with distinct 7Li/6Li, but could simply develop because of chemical gradients. Lastly a significant uncertainty about the preservation of the altered oceanic crust Li hinges on the lack of accurate Li diffusion coefficients at the upper mantle conditions. Neodymium diffusion in clinopyroxene decreases with depth along the adiabatic geotherm (Van Orman et al., 2001), suggesting that tracer diffusion is slower at a greater depth. Then, the preservation of Li heterogeneities could be favored in the deep mantle. We hope this uncertainty is resolved by laboratory high-pressure experiments in the near future. Acknowledgements This paper largely benefited from the careful and constructive review by Horst Marschall. The authors are grateful to an anonymous reviewer for his positive evaluation and to Rick Carlson for editorial handling. Thanks also to C. Poggi (LGCA) for the trace element analysis, C. Bosq, K. David, J.-L. Piro, J.-L. Devidal (LMV, ClermontFerrand) and C. Douchet (LST, Lyon) for technical assistance in the lab. This study benefited from the financial support from the INSU program “Structure, Evolution et Dynamique de l'Intérieur de la Terre, AO2008”. References Berlo, K., Blundy, J., Turner, S., Cashman, K., Hawkesworth, C., Black, S., 2004. Geochemical precursors to volcanic activity at Mount St. Helen, USA. Science 306, 1167–1169. Bonneville, A., Le Suavé, R., Audin, L., Clouard, V., Dosso, L., Gillot, P.-Y., Janney, P., Jordahl, K., Maamaatuaiahutapu, K., 2002. Arago Seamount: the missing hotspot found in the Austral Islands. Geology 30, 1023–1026. Bonneville, A., Dosso, L., Hildenbrand, A., 2006. Temporal evolution and geochemical variability of the South Pacific superplume activity. Earth Planet. Sci. Lett. 244, 251–269. Brenan, J.M., Neroda, E., Lundstrom, C.C., Shaw, H.F., Ryerson, F.J., Phinney, D.L., 1998. Behavior of boron, beryllium, and lithium during melting and crystallization: constraints from mineral-melt partitioning experiments. Geochim. Cosmochim. Acta 62, 2129–2141. Chan, L.-H., Edmond, J.M., Thompson, G., Gillis, K., 1992. Lithium isotopic composition of submarine basalts: implications for the lithium cycle in the oceans. Earth Planet. Sci. Lett. 108, 151–160. Chan, L.-H., Alt, J.C., Teagle, D.A.H., 2002. Lithium and lithium isotope profiles through the upper oceanic crust: a study of seawater–basalt exchange at ODP Sites 504B and 896A. Earth Planet. Sci. Lett. 201, 187–201. Chan, L.-H., Leeman, P.W., Plank, T., 2006. Lithium isotopic composition of marine sediments. Geochem. Geophys. Geosys. 7, Q06005. doi:10.1029/2005GC001202. Chan, L.-H., Lassiter, J.C., Hauri, E.H., Hart, S.R., Blusztajn, J., 2009. Lithium isotope systematics of lavas from the Cook–Austral Islands: constraints on the origin of HIMU mantle. Earth Planet. Sci. Lett. 277, 433–442. Chauvel, C., Hofmann, A.W., Vidal, P., 1992. HIMU-EM: the French Polynesian connection. Earth Planet. Sci. Lett. 110, 99–119. Chauvel, C., McDonough, W., Guille, G., Maury, R., Duncan, R., 1997. Contrasting old and young volcanism in Rurutu Island, Austral chain. Chem. Geol. 139, 125–143. Coogan, L.A., Kasemann, S.A., Chakraborty, S., 2005. Rates of hydrothermal cooling of new oceanic upper crust derived from lithium-geospeedometry. Earth Planet. Sci. Lett. 240, 415–424. Crank, J., 1975. The Mathematics of Diffusion. Oxford University Press, Oxford. 414 pp. Diraison, C., 1991. Le volcanisme aérien des archipels polynésiens de la Société, des Marquises et des Australes–Cook. Téphrostratigraphie, datation isotopique et géochimie comparées. Contribution à l'étude des origines du volcanism intraplaque du Pacifique central. Ph. D. Thesis, Brest, France. Dohmen, R., Kasemann, S., Coogan, L., Chakraborty, S., 2008. Diffusion of Li in olivine: complex behavior arising from effects of Li concentration and defect chemistry. Eos Trans. AGU 89 (53) (Fall Meet. Suppl., Abstract V32A-04). Dostal, J., Dupuy, C., Dudoignon, P., 1996. Distribution of boron, lithium and beryllium in ocean basalts from French Polynesia: implications for the B/Be and Li/Be ratios as tracers of subducted components. Mineral. Mag. 60, 563–580. Duncan, R.A., 1975. Linear Volcanism in French Polynesia. Ph. D. Thesis, Canberra, Australia.

465

Duncan, R.A., McDougall, I., 1976. Linear volcanism in French Polynesia. J. Volcanol. Geotherm. Res. 1, 197–227. Dupuy, C., Barsczus, H.G., Liotard, J.-M., Dostal, J., 1988. Trace element evidence for the origin of ocean island basalts: an example from Austral Islands (French Polynesia). Contrib. Mineral. Petrol. 98, 293–302. Dupuy, C., Barsczus, H.G., Dostal, J., Vidal, P., Liotard, J.-M., 1989. Subducted and recycled lithosphere as the mantle source of ocean island basalts from southern Polynesia, central Pacific. Chem. Geol. 77, 1–18. Eggins, S.M., Woodhead, J.D., Kinsley, L.P.J., Mortimer, G.E., Sylvester, P., McCulloch, M.T., Hergt, J.M., Handler, M.R., 1997. A simple method for the precise determination of ≥40 trace elements in geological samples by ICPMS using enriched isotope internal standardization. Chem. Geol. 134, 311–326. Elliott, T., Jeffcoate, A., Bouman, C., 2004. The terrestrial Li isotope cycle: light-weight constraints on mantle convection. Earth Planet. Sci. Lett. 220, 231–245. Elliott, T., Thomas, A., Jeffcoate, A., Niu, Y., 2006. Lithium isotope evidence for subduction-enriched mantle in the source of mid-ocean-ridge basalts. Nature 443, 565–568. Guille, G., Guillou, H., Chauvel, C., Maury, R., Blais, S., Brousse, R., 1998. L'île de Rurutu (archipel des Australes, Polynésie française): une edification complexe liée au fonctionnement de deux points chauds. Géol. Fr. 3, 65–85. Halama, R., McDonough, W.F., Rudnick, R.L., Bell, K., 2008. Tracking the lithium isotopic evolution of the mantle. Earth Planet. Sci. Lett. 265, 726–742. Hofmann, A.W., Hart, S.R., 1978. An assessment of local and regional isotopic equilibrium in the mantle. Earth Planet. Sci. Lett. 38, 44–62. Huh, Y., Chan, L.-H., Chadwick, O.A., 2004. Behavior of lithium and its isotopes during weathering of Hawaiian basalt. Geochem. Geophys. Geosys. 5, Q09002. doi:10.1029/ 2004GC000729. Ionov, D.A., Seitz, H.-M., 2008. Lithium abundances and isotopic compositions in mantle xenoliths from subduction and intraplate settings: mantle source vs. eruption histories. Earth Planet. Sci. Lett. 266, 316–331. James, R.H., Palmer, M.R., 2000. The lithium isotope composition of international rock standards. Chem. Geol. 166, 319–326. Jeffcoate, A.B., Elliott, T., Thomas, A., Bouman, C., 2004. Precise, small sample size determination of lithium isotopic compositions of geological reference materials and modern seawater by MC-ICP-MS. Geostand. Geoanal. Res. 28, 161–172. Jeffcoate, A.B., Elliott, T., Kasemann, S.A., Ionov, D., Cooper, K., Brooker, R., 2007. Li isotope fractionation in peridotites and mafic melts. Geochim. Cosmochim. Acta 71, 202–218. Kuritani, T., Nakamura, E., 2006. Elemental fractionation in lavas during post-eruptive degassing: evidence from trachytic lavas, Rishiri Volcano, Japan. J. Volcanol. Geotherm. Res. 149, 124–138. Lassiter, J.C., Hauri, E.H., Hart, S.R., Blusztajn, J., Chan, L., 2008. Lithium isotope variations in lavas and olivine phenocrysts from the Cook–Austral Islands: constraints on sample alteration and the true Li-isotope signature of HIMU mantle. Eos Trans. AGU 89 (53) (Fall Meet. Suppl., Abstract V32A-02). Lundstrom, C.C., Chaussidon, M., Hsui, A.T., Kelemen, P., Zimmerman, M., 2005. Observations of Li isotopic variations in the Trinity ophiolite: evidence for isotopic fractionation by diffusion during mantle melting. Geochim. Cosmochim. Acta 69, 735–751. Marschall, H.R., Altherr, R., Ludwig, T., Kalt, A., Gméling, K., Kasztovszky, Z., 2006. Partitioning and budget of Li, Be and B in high-pressure metamorphic rocks. Geochim. Cosmochim. Acta 70, 4750–4769. Marschall, H.R., Pogge von Strandmann, P.A.E., Seitz, H.-M., Elliott, T., Niu, Y., 2007. The lithium isotopic composition of orogenic eclogites and deep subducted slabs. Earth Planet. Sci. Lett. 262, 563–580. McDonough, W.F., Sun, S.-s., 1995. The composition of the Earth. Chem. Geol. 120, 223–253. Nishio, Y., Nakai, S., Kogiso, T., Barsczus, H.G., 2005. Lithium, strontium, and neodymium isotopic compositions of oceanic island basalts in the Polynesian region: constraints on a Polynesian HIMU origin. Geochem. J. 39, 91–103. Ottolini, L., Laporte, D., Raffone, N., Devidal, J.-L., Le Fèvre, B., 2009. New experimental determination of Li and B partition coefficients during mantle partial melting. Contrib. Mineral. Petrol. 157, 313–325. Parkinson, I.J., Hammond, S.J., James, R.H., Rogers, N.W., 2007. High-temperature lithium isotope fractionation: insight from lithium isotope diffusion in magmatic systems. Earth Planet. Sci. Lett. 257, 609–621. Pistiner, J.S., Henderson, G.M., 2003. Lithium-isotope fractionation during continental weathering processes. Earth Planet. Sci. Lett. 214, 327–339. Richter, F.M., Davis, A.M., DePaolo, D.J., Watson, E.B., 2003. Isotope fractionation by chemical diffusion between molten basalt and rhyolite. Geochim. Cosmochim. Acta 67, 3905–3923. Ryan, J.G., Langmuir, C.H., 1987. The systematics of lithium abundances in young volcanic rocks. Geochim. Cosmochim. Acta 51, 1727–1741. Ryan, J.G., Kyle, P.R., 2004. Lithium abundance and lithium isotope variations in mantle sources: insights from intraplate volcanic rocks from Ross Island and Marie Byrd Land (Antarctica) and other oceanic islands. Chem. Geol. 212, 125–142. Seitz, H.-M., Woodland, A.B., 2000. The distribution of lithium in peridotitic and pyroxenitic mantle lithologies — an indicator of magmatic and metasomatic processes. Chem. Geol. 166, 47–64. Seitz, H.-M., Brey, G.P., Lahaye, Y., Durali, S., Weyer, S., 2004. Lithium isotopic signatures of peridotite xenoliths and isotopic fractionation at high temperature between olivine and pyroxenes. Chem. Geol. 212, 163–177. Seitz, H.-M., Brey, G.P., Zipfel, J., Ott, U., Weyer, S., Durali, S., Weinbruch, S., 2007. Lithium isotope composition of ordinary carbonaceous chondrites, and differentiated planetary bodies: bulk solar system and solar reservoirs. Earth Planet. Sci. Lett. 260, 582–596.

466

I. Vlastélic et al. / Earth and Planetary Science Letters 286 (2009) 456–466

Simons, K.K., Langmuir, C.L., Goldstein, S.L., Hemming, N.G., 2008. Lithium isotope systematics of the Azores Platform: insight into Li variability of mantle sources. Eos Trans. AGU 89 (53) (Fall Meet. Suppl., Abstract V43C-2176). Sun, S.S., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In: Saunders, A.D., Norry, M.J. (Eds.), Magmatism in the Ocean Basins: Geol. Soc. Spec. Publ., vol. 42, pp. 313–345.

Tomascak, P.B., Langmuir, C.H., le Roux, P.J., Shirey, S.B., 2008. Lithium isotopes in global mid-ocean ridge basalts. Geochim. Cosmochim. Acta 72, 1626–1637. Van Orman, J., Grove, T., Shimizu, N., 2001. Rare earth element diffusion in diopside: influence of temperature, pressure, and ionic radius, and an elastic model for diffusion in silicates. Contrib. Mineral. Petrol. 141, 687–703.