Oxygen isotope heterogeneity of the mantle beneath the Canary Islands

2 downloads 4 Views 550KB Size Report
Canary Islands: a discussion of the paper of Gurenko et al. ... David Lowry • D. Graham Pearson ... Keywords Ocean island basalt 4 Canary Islands 4 Oxygen.

Contrib Mineral Petrol DOI 10.1007/s00410-012-0755-3


Oxygen isotope heterogeneity of the mantle beneath the Canary Islands: a discussion of the paper of Gurenko et al. James M. D. Day • Colin G. Macpherson David Lowry • D. Graham Pearson

Received: 27 October 2011 / Accepted: 1 March 2012 Ó Springer-Verlag 2012

Abstract Gurenko et al. (Contrib Mineral Petrol 162:349–363, 2011) report laser-assisted fluorination (LF) and secondary ionization mass spectrometry (SIMS) 18 O/16O datasets for olivine grains from the Canary Islands of Gran Canaria, Tenerife, La Gomera, La Palma and El Hierro. As with prior studies of oxygen isotopes in Canary Island lavas (e.g. Thirlwall et al. Chem Geol 135:233–262, 1997; Day et al. Geology 37:555–558, 2009, Geochim Cosmochim Acta 74:6565–6589, 2010), these authors find variations in d18Ool (*4.6–6.0 %) beyond that measured for mantle peridotite olivine (Mattey et al. Earth Planet Sci Lett 128:231–241, 1994) and interpret this variation to reflect contributions from pyroxenite-peridotite mantle sources. Furthermore, Gurenko et al. (Contrib Mineral Petrol 162:349–363, 2011) speculate that d18Ool values for Communicated by J. Hoefs. This comment refers to the article available at 10.1007/s00410-010-0600-5. An author‘s reply to this comment is available at 10.1007/s00410-012-0756-2. J. M. D. Day (&) Geosciences Research Division, Scripps Institution of Oceanography, La Jolla, CA 92093-0244, USA e-mail: [email protected] C. G. Macpherson Department of Earth Sciences, University of Durham, Durham DH1 3LE, UK D. Lowry Department of Earth Sciences, Royal Holloway, University of London, Surrey TW20 0EX, UK D. G. Pearson Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton T6G 2E3, Canada

La Palma olivine grains measured by LF (Day et al. Geology 37:555–558, 2009, Geochim Cosmochim Acta 74:6565–6589, 2010) may be biased to low values due to the presence of altered silicate, possibly serpentine. The range in d18Ool values for Canary Island lavas are of importance for constraining their origin. Gurenko et al. (Contrib Mineral Petrol 162:349–363, 2011) took a subset (39 SIMS analyses from 13 grains from a single El Hierro lava; EH4) of a more extensive dataset (321 SIMS analyses from 110 grains from 16 Canary Island lavas) to suggest that d18Ool is weakly correlated (R2 = 0.291) with the parameter used by Gurenko et al. (Earth Planet Sci Lett 277:514–524, 2009) to describe the estimated weight fraction of pyroxenite-derived melt (Xpx). With this relationship, end-member d18O values for HIMU-peridotite (d18O = 5.3 ± 0.3 %) and depleted pyroxenite (d18O = 5.9 ± 0.3 %) were defined. Although the model proposed by Gurenko et al. (Contrib Mineral Petrol 162:349–363, 2011) implicates similar pyroxenite-peridotite mantle sources to those proposed by Day et al. (Geology 37:555– 558, 2009, Geochim Cosmochim Acta 74:6565–6589, 2010) and Day and Hilton (Earth Planet Sci Lett 305:226–234, 2011), there are significant differences in the predicted d18O values of end member components in the two models. In particular, Day et al. (Geochim Cosmochim Acta 74:6565–6589, 2010) proposed a mantle source for La Palma lavas with low-d18O (\5 %), rather than higherd18O (c.f. the HIMU-peridotite composition of Gurenko et al. in Contrib Mineral Petrol 162:349–363, 2011). Here we question the approach of using weakly correlated variations in d18Ool and the Xpx parameter to define mantle source oxygen isotope compositions, and provide examples of why this approach appears flawed. We also provide reasons why the LF datasets previously published for Canary Island lavas remain robust and discuss why LF and


Contrib Mineral Petrol

SIMS data may provide complementary information on oxygen isotope variations in ocean island basalts (OIB), despite unresolved small-scale uncertainties associated with both techniques. Keywords Ocean island basalt  Canary Islands  Oxygen isotopes  Laser fluorination  SIMS  Mantle heterogeneity

Assessment of laser fluorination (LF) O isotope measurements in Canary Island lavas Of the existing techniques used to determine oxygen isotope ratios in olivine and pyroxene, LF has been demonstrated to be the method of choice due to its accuracy and precision (e.g. Sharp 1990; Mattey and Macpherson 1993; Mattey et al. 1994; Valley et al. 1995; Rumble et al. 2007). Although recent advances in SIMS have improved precision of d18O1 measurements to C0.3 % (all uncertainties for O isotopes reported in this article are 2r) for 10–20-lm spot sizes (Kita et al. 2009; Eiler et al. 2011), the SIMS technique still cannot match the long-term precision of LF (B0.2 %). For silicate minerals, such as olivine and pyroxene, the accuracy of LF, which employs larger absolute sample masses (typically 1–2 mg) than SIMS (\0.1 mg), relies on (1) careful and methodical hand-separation of mineral grains and fragments, which are repeatedly washed and purified to obtain clean separates free from inclusions (e.g. opaque oxides) and alteration; (2) pre-fluorination of the minerals and sample chamber with the fluorinating agent; (3) careful assessment that oxygen yields from fluorination are close to 100 %; and (4) regular assessment of isotopic fractionation and instrumental conditions by analysis of silicate mineral standards of variable d18O composition (Mattey and Macpherson 1993). These processes must be carefully observed, as emphasized from the analytical difficulties Gurenko et al. (2011) reported associated with insufficient cleaning and pre-treatment of olivine grains from some Canary Island lavas prior to LF analysis. These authors invoked alteration based upon poor reproducibility of results for their La Palma samples during an initial phase of LF analysis (LP1-Ol1 = 4.20 ± 0.63 %, n = 2; LP13-Ol1 = 4.24 ± 0.66 %, n = 2; LP48Ol1 = 4.27 ± 0.27 %, n = 2) versus a later phase where they treated olivine splits with 30 % HF or HBF4 prior to fluorination (LP1-Ol1 = 5.15 %, n = 1; LP13-Ol1 = 5.24 ± 0.36 %, n = 5; LP48-Ol1 = 4.89 %, n = 1). The d18O values they obtained from the initial analytical phase, employing techniques from the same laboratory previously used for other sample suites (e.g. Bindeman et al. 2008), were anomalously low, down to 3.82 % in a La Gomera olivine (LG35-Ol1) for which they obtained 4.70 % in later analyses. 1

(d18O = ([(18O/16O)sample/(18O/16O)VSMOW] -1) 9 1,000).


They considered the possibility that the low-d18O values obtained during their initial analytical phase might be due to the presence of chromite inclusions, but discounted this based on the mass of chromite required, a conclusion with which we agree. Instead, they argued that systematically low-d18O values obtained in their early analytical phase were caused by alteration products along cracks (e.g. serpentinite or iddingsite) that were not properly removed due to no pre-fluorination. They then speculate that d18Ool values reported by Day et al. (2009, 2010) for La Palma, which are within the range of, or lower than, those in their own La Palma data set, are also due to the presence of altered silicate and consequently surmise that: ‘we think that the remaining alteration of olivine may be a possible reason for the lower La Palma d18Oolivine reported by Day et al. (2009, 2010)’. This supposition was based on the comparisons of their analysis of La Palma olivine separates from three lavas by LF (5.17 ± 0.39 %; n = 7) and four lavas by SIMS (5.59 ± 0.67; n = 57 analyses of 20 olivine grains) versus LF data presented by Day et al. for three xenoliths and a total of 24 lavas (4.87 ± 0.36 %; n = 27). Whereas the initial phase of d18O data acquisition by Gurenko et al. (2011) may have incorporated altered materials, this is not the case for the data reported by Day et al. (2009, 2010) for four principal reasons: 1.


Analysis of mineral grains free from visible inclusions or alteration. Mineral grains were subjected to washing in ultrapure water and ethanol to remove any adhering particles prior to inspection using a 10–50 9 magnification binocular microscope to ensure that they were free of inclusions and showed no evidence of alteration (e.g. Fig. 1a). The analysed separates (typically 1.5–1.9 mg, composed of multiple fragments) were pre-fluorinated using BrF5 prior to extracting oxygen, and high oxygen yields (95 ± 5 % as calculated from chemical data for the olivine) indicate representative isotopic analysis (Mattey and Macpherson 1993). Careful standardisation was performed, with at least one standard analysed per four unknowns. No correlations with degree of alteration and d18O. Our data display no systematic relationships between d18O values of separated olivine and the degree of visible alteration and water/rock interaction of olivine in the rock. Seamount series picrite pillow lava LP03 is the most altered rock that we measured, displaying evidence for alteration of some olivine crystals to iddingsite and serpentine (supplementary information of Day et al. 2010). The fresh olivine core material from LP03 that we measured is at the higher end of the d18Ool range for La Palma (5.0 %) and is within (our) analytical uncertainty of the mean d18Ool for La Palma in the Gurenko data. We measured lower d18Ool values in unaltered cumulate xenoliths (e.g. LP13a = 4.36 %) and historic lavas (AD

Contrib Mineral Petrol

Fig. 1 a Olivine (Ol) and clinopyroxene (Cpx) grains from LP02 (d18Ool = 4.73 %; d18Oclinopyroxene = 5.01 %) immediately after separation (no cleaning or purification) to illustrate features related to identification of pure mineral separates appropriate for laserfluorination oxygen isotope analyses. Note dark inclusions and slight discoloration for some olivines; these grains were not used for O isotope measurements. For clinopyroxene, grains were fragmented until transparent slivers were obtained in order to identify inclusionand alteration-free material. The resultant pure mineral grain material

was thoroughly cleaned in ultrapure water and ethanol and prefluorinated prior to analysis. Back-scatter electron images of representative olivine grains from b LP 11, c EH 13 and d, e EH 18 that span a range in measured d18Ool from 4.71 to 5.26 %. The matrices in the lavas are a mixture of vesicles (dark regions), plagioclase, pyroxene and olivine and Fe-Ti oxides (bright regions). There is a lack of alteration in any of the studied olivines and evidence for Fe/ Mg zonation at olivine grain rims (e)

1646 to 1949 = 4.92–5.02 %) giving confidence that no altered material was included in mineral splits that were fluorinated. Equilibrium between olivine and clinopyroxene. In addition to La Palma olivine, Day et al. (2009, 2010) report d18O values for coexisting clinopyroxene and for both olivine and clinopyroxene from lavas on the neighbouring island of El Hierro. Values of D18Oolivine–clinopyroxene for El Hierro (-0.24 ± 0.18 %; n = 9; 2r) and La Palma (-0.35 ± 0.40 %; n = 16; 2r) are consistent with the values expected for equilibrium at high temperature (Fig. 2). Furthermore, the *0.3 % difference in d18Ool between El Hierro (5.17 ± 0.16 %; n = 14; 2r) and La Palma (4.87 ± 0.36 %; 2r) is supported by the clinopyroxene data (El Hierro = 5.42 ± 0.18 %; n = 9; La Palma = 5.20 ± 0.22 %; n = 16; 2r). If alteration was responsible for the difference in d18O values between the two islands, it would require identical amounts of alteration for these two phases, despite evidence for higher resistance to weathering of clinopyroxene (e.g. Goldich 1938; Hausrath et al. 2008). Our repeat analyses of individual olivine and clinopyroxene separates show little variation

beyond external analytical uncertainty (SCOL I and II olivine = ± 0.16 %; 2r; Day et al. 2010). Altered olivine or clinopyroxene should yield less reproducible results, akin to the uncertainties reported by Gurenko et al. (2011) for their initial phase of LF (0.27–0.66 %). Gurenko et al. (2011) did not measure clinopyroxene and did not measure olivine from El Hierro by LF, so direct comparisons with our data are not possible. Greenschist facies metamorphism in the presence of meteoric water and high water/rock ratios has lowered the d18O value of hypabyssal and plutonic mafic rocks from La Palma (Javoy et al. 1986; Deme´ny et al. 2008). If we accept d18Ool from the peridotite endmember advocated by Gurenko et al. (2011) represents the lowest magmatic d18Ool at La Palma (4.8 %) and take the most extreme chlorite/serpentinite separate measured from La Palma (?1 %, Deme´ny et al. 2008), this would still require in excess of 10 % alteration minerals to generate the lowest d18Ool measured by Day et al. (2009, 2010). We reiterate that even minor quantities of serpentine in olivine grains would be visible and would lead to low fluorination yields (\90 %) due to their removal




Contrib Mineral Petrol

Fig. 2 Plot of d18Oclinopyroxene versus d18Ool for La Palma and El Hierro lavas from Day et al. (2010). Dashed lines represent isotherms calculated in Zheng (1993). Mantle nodules analysed by laser fluorination are from Mattey et al. (1994). La Palma and El Hierro olivine–clinopyroxene pairs exhibit a similar range of D18O to mantle nodules, but extend to lower d18O values

during pre-fluorination. The lavas we analysed show no petrographic indications (e.g. secondary hornblende and biotite; Day et al. 2010) of having experienced alteration at the elevated temperatures required to generate low-d18O serpentine. Indeed, any serpentine present in the lavas is likely to have formed at relatively low post-emplacement alteration temperatures (\300 °C), which would result in an alteration product with high-d18O (Skelton and Valley 2000). High-resolution back-scatter electron images of three samples from El Hierro and La Palma, which span the range of d18Ool measured by Day et al. (2009, 2010), are shown in Fig. 1 and illustrate, with one exception (LP03, shown in the supplementary information of Day et al. 2010), the petrographic freshness of samples and absence of alteration on the outside of crystals, within cracks, or within olivine. High reflectance inclusions in these images are oxides that can be screened during visual inspection under a high-powered binocular microscope. In summary, we contend that the LF data presented by Day et al. (2009, 2010) represent true magmatic d18O values for Canary Island olivine and pyroxene grains. We therefore consider these data in the context of SIMS d18O values reported for Canary Island lavas.

Intracrystalline versus intercrystalline oxygen isotope variations Gurenko et al. (2011) performed SIMS and LF measurements of d18Ool in Canary Island lavas. The mean d18Ool


values of LF measurements obtained by these authors lie within the ‘mantle range’ for d18Ool (4.8–5.4 %) and are, therefore, more akin to the data of Day et al. (2009, 2010) than their SIMS measurements, which tend to systematically higher d18O values (Fig. 3a). Furthermore, Fig. 3b demonstrates that the combined LF d18O data set for La Palma forms a continuum versus olivine forsterite contents (molar Mg/[Mg ? Fe]) reflecting either low-d18O, high Fe/ Mg melts (Day et al. 2009), or contamination by crust as observed at Hawai’i (Wang and Eiler 2008). In contrast, SIMS d18O values for La Palma olivine are scattered, mainly to high values at similar Fo content, compared with the LF data (Fig. 3b). Eiler et al. (2011) recently concluded that SIMS analyses can yield O isotope compositions in close (*0.1–0.2 %) agreement with bulk separates measured by LF, provided samples are prepared and analysed following strict guidelines (also see Kita et al. 2009). These guidelines are important because of unresolved small-scale uncertainties associated with both the SIMS and LF O isotope techniques (Eiler et al. 2011). Eiler et al. (2011) also concluded that, with a few exceptions related to rapid crystal growth and associated assimilation of crust, oxygen isotope compositions of specific mineral phenocryst populations in mafic and ultramafic terrestrial rocks are generally homogeneous and well-suited to LF of mineral separates. Therefore, oxygen isotope compositions obtained by SIMS on intercrystalline variations in phenocrysts from OIB lavas may offer sensitive means for detecting meltcrust interactions. SIMS d18O measurements may be most useful in volcanic environments with large-volume magma production (e.g. high-degree tholeiitic partial melts) and greater exposure to low-d18O hydrothermally altered crust (e.g. Iceland, Yellowstone, Kamchatka; Eiler et al. 2011), but may also be useful for tracking magma mixing processes between melts and contaminants in smaller-volume (e.g. low-degree alkali basalt partial melts) magmatic systems, such as the majority of Canary Island eruptions, and many other OIB. Contamination of some Canary Island magmas can occur by interaction with underlying oceanic lithosphere and with edifice materials (e.g. Hoernle and Tilton 1991; Hoernle 1998; Hansteen and Troll 2003). With few exceptions (Thirlwall et al. 1997), the majority of studies of Canary Island lavas have discounted large amounts ([10 %) of crustal contamination based on multiisotopic and geochemical studies (e.g. Hoernle et al. 1991; Deme´ny et al. 2008; Day et al. 2009, 2010; Gurenko et al. 2009, 2011; Day and Hilton 2011). It is important to note, however, that melt–crust interaction can generate significantly more intracrystal and intercrystal d18O variation within a basalt sample than is apparent from its whole-rock or multigrain mineral analyses (e.g. Baker et al. 2000), which will yield weighted mean ratios of all the

Contrib Mineral Petrol b Fig. 3 a Oxygen isotopic composition of olivine grains analysed by laser-fluorination (LF) and SIMS by Day et al. (2009, 2010) and Gurenko et al. (2011). The shaded vertical band shows the upper mantle olivine range (Mattey et al. 1994). GC Gran Canaria, TF Tenerife, LG La Gomera, LP La Palma, EH El Hierro. b Relationships between d18O values and Fo content of olivine for LF/SIMS data of Gurenko et al. (2011) and average Fo contents for Canary Island (Day et al. 2009) and Hawai’i (Wang and Eiler 2008) olivines. Arrow denotes shallow-level contamination of Hawaiian magmas (Wang and Eiler 2008). Shading shows range of laser-fluorination data for El Hierro (red) and La Palma (beige). c Relationship between d18Ool and the parameter Xpx (the estimated amount of pyroxenite in the source of lavas using the relationship of Xpx = 1.341E-03 9 [Ni 9 FeO/ MgO] - 0.437; Gurenko et al. 2009) for Canary Island lavas. Shown is the data set of Day et al. (2010), measured by laser-fluorination, and the data set of Gurenko et al. (2011) measured by SIMS and by LF. Linear regressions and R2 values are shown for SIMS data for the sample EH4 and LG35 from Gurenko et al. (2011) and for the data set (WCI) of Day et al. (2010). Two sigma error bars are shown for data from Day et al. (2010)

et al. (2011), provide additional evidence for the likelihood of fine-scale heterogeneity in SIMS d18O values within individual olivine grains.

End-member mantle components beneath the Canary Islands

components present in the rock or phenocryst population. Contamination of Canary Island lavas has been advocated to explain the d18O values of clinopyroxenes in Gran Canaria shield lavas by mixing with mantle-like d18O and a high-d18O sediment component (Thirlwall et al. 1997), and so it should be no surprise to find an intra-flow record of these processes in SIMS d18Ool values of Canary Island alkali basalts. The varied mineral chemistry and evidence for fine-scale overgrowths of the rims of olivine grains (e.g. Fig. 1) from lavas from El Hierro and La Palma, along with samples from other Canary Islands presented by Gurenko

Gurenko et al. (2011) interpreted weakly correlated (R2 = 0.291) relationships between Ni 9 FeO/MgO and SIMS d18O values in olivine from a single El Hierro basanite lava (EH4) to estimate the source character of peridotite and pyroxenite end-member mantle feeding Canary Island parental magmas. They argued for a peridotite end-member mantle component with d18Ool of 4.8 ± 0.3 % and a pyroxenite end-member mantle component with d18Ool of 5.5 ± 0.3 %. To investigate the robustness of this approach, we calculated the Xpx parameter of Gurenko et al. (2009, 2011) for olivine and compare this to d18O for the full data sets from Day et al. (2009, 2010) and Gurenko et al. (2009) in Fig. 3c. The EH4 olivine data used by Gurenko et al. (2011) to estimate mantle d18O end-member compositions are identified. We also show the linear regression for a La Gomera sample (LG35) from Gurenko et al. (2011) measured by SIMS for d18Ool, which has a slightly better R2 value to EH4, but defines mantle endmember compositions of d18O * 4.2 % for peridotite, and d18O * 6.7 % for pyroxenite. We also performed a linear regression of data from Day et al. (2010), obtaining mantle end-member compositions of d18O * 4.4 % for peridotite and d18O * 6.0 % for pyroxenite. As pointed out by Day et al. (2010), assuming the model of Gurenko et al. (2009), the HIMU-peridotite end-member must have low d18O (\4.6 %) and would be a significant repository of low-d18O in the Canarian mantle. However, potential end-member


Contrib Mineral Petrol

d18O compositions do not appear tightly defined. This observation is based on both Xpx–d18O comparisons (Fig. 3c), which have low R2 values, and O–Os isotope variations observed for El Hierro and La Palma lavas (Day et al. 2009), which indicate a range in the pyroxenite endmember compositions. This latter conclusion is consistent with variable d18O values measured for exposed pyroxenite massif localities (Pearson et al. 1991). In fact, the range of estimated d18O values from comparison with Xpx for the peridotite (4.2–5.2 %) and pyroxenite (5.5–6.7 %) endmembers from the three Canary Island examples spans nearly the entire range of O isotope compositions measured by LF for uncontaminated mantle-derived OIB to date. Thus, we argue that the use of SIMS d18O values in olivine from a single sample should proceed with some caution, as oxygen isotope compositions in individual olivine crystals may be sensitive recorders of both mantle- and crustalderived contributions in some OIB (Thirlwall et al. 1997; Eiler et al. 2011) and because individual phenocrysts from different islands (e.g. EH4 and LG35) can yield highly variable results ([1 %) for establishing d18O variability in OIB mantle sources. Acknowledgments This work was performed with support from the National Science Foundation (EAR-1116089) and The San Diego Foundation (C-2011-00204). We thank C. Harris, two anonymous reviewers and Editor J. Hoefs for their constructive comments.

References Baker JA, Macpherson CG, Menzies MA, Thirlwall MF, Al-Kadasi M, Mattey DP (2000) Resolving crustal and mantle contributions to continental flood volcanism, Yemen; constraints from mineral oxygen isotope data. J Petrol 41:1805–1820 Bindeman I, Gurenko A, Sigmarsson O, Chaussidon M (2008) Oxygen isotope heterogeneity and disequilibria of olivine crystals in large volume Holocene basalts from Iceland: evidence for magmatic digestion and erosion of Pleistocene hyaloclastites. Geochim Cosmochim Acta 72:4397–4420 Day JMD, Hilton DR (2011) Origin of 3He/4He ratios in HIMU-type basalts constrained from Canary Island lavas. Earth Planet Sci Lett 305:226–234 Day JMD, Pearson DG, Macpherson CG, Lowry D, Carracedo JC (2009) Pyroxenite-rich mantle formed by recycled oceanic lithosphere: oxygen-osmium isotope evidence from Canary Island lavas. Geology 37:555–558 Day JMD, Pearson DG, Macpherson CG, Lowry D, Carracedo JC (2010) Evidence for distinct proportions of subducted oceanic crust and lithosphere in HIMU-type mantle beneath El Hierro and La Palma, Canary Islands. Geochim Cosmochim Acta 74:6565–6589 Deme´ny A, Cassillas R, Vennemann TW, Hegner E, Nagy G, Ahijado A, De La Nuez J, Sipos P, Pilet S, Milton J (2008) Plume-related stable isotope compositions and fluid–rock interaction processes in the Basal Complex of La Palma, Canary Islands, Spain. In: Coltorti M, Greogoire M (eds) Metasomatism in oceanic and continental lithospheric mantle. Special Publication vol 293, geological society, London, pp 155–175


Eiler JM, Stolper E, McCanta MC (2011) Intra- and intercrystalline oxygen isotope variations in minerals from basalts and peridotites. J Petrol 52:1393–1413 Goldich S (1938) A study of rock weathering. J Geol 46:17–58 Gurenko AA, Sobolev AV, Hoernle KA, Hauff F, Schmincke H-U (2009) Enriched, HIMU-type peridotite and depleted recycled pyroxenite in the Canary plume: a mixed-up mantle. Earth Planet Sci Lett 277:514–524 Gurenko AA, Bindeman IN, Chaussidon M (2011) Oxygen isotope heterogeneity of the mantle beneath the Canary Islands: insights from olivine phenocrysts. Contrib Mineral Petrol 162:349–363 Hansteen TH, Troll VR (2003) Oxygen isotope composition of xenoliths from the oceanic crust and volcanic edifice beneath Gran Canaria (Canary Islands): consequences for crustal contamination of ascending magmas. Chem Geol 193:181–193 Hausrath EM, Navarre-Sitchler AK, Sak PB, Steefel CI, Brantley SL (2008) Basalt weathering rates on Earth and the duration of liquid water on the plains of Gusev crater, Mars. Geology 36:67–70 Hoernle K (1998) Geochemistry of Jurassic oceanic crust beneath Gran Canaria (Canary Islands): implications for crustal recycling and assimilation. J Petrol 39:859–880 Hoernle K, Tilton G (1991) Sr–Nd–Pb isotope data for Fuerteventura (Canary Islands) basal complex and subaerial volcanics: application to magma genesis and evolution. Schweiz Mineral Petrogr Mitt 71:5–21 Hoernle K, Tilton G, Schmincke H-U (1991) Sr–Nd–Pb isotopic evolution of Gran Canaria: evidence for shallow enriched mantle beneath the Canary Islands. Earth Planet Sci Lett 106:44–63 Javoy M, Stillman CJ, Pineau F (1986) Oxygen and hydrogen isotope studies on the basal complexes of the Canary Islands: implications on the conditions of their genesis. Contrib Miner Petrol 92:225–235 Kita NT, Ushikubo T, Fu B, Valley JW (2009) High precision SIMS oxygen isotope analysis and the effect of sample topography. Chem Geol 264:43–57 Mattey D, Macpherson C (1993) High-precision oxygen isotope microanalysis of ferromagnesian minerals by laser-fluorination. Chem Geol 105:305–318 Mattey D, Lowry D, Macpherson C (1994) Oxygen isotope composition of mantle peridotite. Earth Planet Sci Lett 128:231–241 Pearson DG, Davies GR, Nixon PH, Greenwood PB, Mattey DP (1991) Oxygen isotope evidence for the origin of pyroxenites in the Beni Boussera peridotite massif, N. Morrocco: derivation from subducted oceanic lithosphere. Earth Planet Sci Lett 102:289–301 Rumble D, Miller MF, Franchi IA, Greenwood RC (2007) Oxygen three-isotope fractionation lines in terrestrial silicate minerals: an inter-laboratory comparison of hydrothermal quartz and eclogitic garnet. Geochim Cosmochim Acta 71:3592–3600 Sharp ZD (1990) A laser based microanalytical method for the in situ determination of oxygen isotope ratios of silicates and oxides. Geochim Cosmochim Acta 54:1353–1357 Skelton ADL, Valley JW (2000) The relative timing of serpentinisation and mantle exhumation at the ocean-continent transition, Iberia: constraints from oxygen isotopes. Earth Planet Sci Lett 178:327–338 Thirlwall MF, Jenkins C, Vroon PZ, Mattey DP (1997) Crustal interaction during construction of oceanic islands: Pb-Sr-Nd-O isotope geochemistry of the shield basalts of Gran Canaria, Canary Islands. Chem Geol 135:233–262 Valley JW, Kitchen N, Kohn MJ, Neindorf CR, Spicuzza MJ (1995) UWG-2, a garnet standard for oxygen isotope ratios: strategies for high precision and accuracy with laser heating. Geochim Cosmochim Acta 59:5223–5231

Contrib Mineral Petrol Wang Z, Eiler JM (2008) Insights into the origin of low-d18O basaltic magmas in Hawaii revealed from in situ measurements of oxygen isotope compositions of olivines. Earth Planet Sci Lett 269:376–386

Zheng Y-F (1993) Calculation of oxygen isotope fractionation in anhydrous silicate minerals. Geochim Cosmochim Acta 57:1079– 1091