Oxygen Isotope Variations in Lau Basin Lavas

Chemical Geology 144Ž. 1998 177±194

Oxygen isotope variations in Lau Basin lavas

Colin G. Macpherson ), David P. Mattey Department of Geology, Royal Holloway UniÕersity of London, Egham, Surrey TW20 0EX, UK

Abstract

New laser-fluorination oxygen isotope data are presented for volcanic glasses and phenocrysts from Lau and North Fiji Basin lavas. The low oxygen blank of the technique allows accurate analysis of these often scarce igneous phases. d 18O values of volcanic glass range from 5.53½ to 6.06½. Oxygen isotope ratios of ferromagnesian and plagioclase phenocrysts are lower and higher than their host glasses, respectively, and coexisting phases appear to represent oxygen isotope equilibrium at magmatic temperatures. Evolved lavas from the propagating and dying rift tips in the Central Lau Basin display significant 18O-enrichment with magmatic evolution due to FeTi-oxide crystallisation and require that precursor melts had relatively low d 18O values immediately prior to oxide saturation. Development of these precursor melts from basaltic parents involved assimilation of hydrothermally altered sheeted dykes or gabbros in recently accreted oceanic crust accompanied by crystallisation of a plagioclase-rich assemblage. The stable isotope geochemistry and petrology of Central Lau Basin lavas constrain magmatic evolution to depths between 2.5 and 5 km beneath the seafloor. A dacite from Valu Fa Ridge is slightly 18O-enriched relative to Valu Fa basalts, consistent with crystallisation of a mineral assemblage similar to mid-ocean ridge basalts. 18Or 16O ratios of Lau Basin basaltic glasses decrease with Mg Number; this observation could be reconciled with the assimilation and fractional crystallisation model proposed for initial differentiation of the evolved lavas. However, oxygen isotope data for Central Lau Basin basalts correlate positively with BarLa while basaltic glasses from 18 throughout the region display a negative correlation between d O and Na8.0 suggesting that recycling of subducted oxygen andror mantle fertility may also exert an influence on 18Or 16O of primary melts. Presently, the lack of quantitative control over oxygen isotope fractionation in basaltic systems prohibits resolution of these effects. Detailed investigation of less complex suites of lavas, employing the precision offered by laser fluorination, may provide new insights into the behaviour of oxygen at magmatic temperatures. q 1998 Elsevier Science B.V.

Keywords: Lau Basin; Oxygen isotopes; Magmatism; Subduction; Melting; Assimilation

1. Introduction

Unaltered lavas erupted at mid-ocean spreading centres display limited18 Or 16O variationŽ 5.7" 0.3½; Ito et al., 1987. . This is thought to result from ) Corresponding author. Present address: Geosciences Research melting of an upper mantle reservoir that is uniform Division ± 0220, Scripps Institution of Oceanography, 9500 with respect to oxygen isotopes, followed by limited Gilman Drive, La Jolla, CA 92093-0220, USA. Tel.: q1Ž. 619 534-3260; fax: q 1Ž. 619 534-0784; e-mail: degrees of chemical evolution in sub-ridge magma [email protected] chambersŽ Muehlenbachs and Clayton, 1972; Taylor

0009-2541r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved. PII S0009-2541Ž. 97 00130-7 178 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 and Sheppard, 1986, and references therein. . Pro- oxygen from various reservoirs into primary Lau tracted fractional crystallisation can produce oxygen Basin melts. isotope variations in cogenetic lava suites but this is rare at mid-ocean spreading centres, having been recognised only at the 958W propagating rift tip of 2. Samples and technique the Galapagos Spreading CentreŽ Muehlenbachs and Byerly, 1982; Ito et al., 1987. . Oxygen isotope ratios The Lau Basin is a triangular-shaped back-arc are frequently employed in petrogenetic studies of basin situated west of the Tonga RidgeŽ. Fig. 1a , subduction zone magmatism to discriminate between where the Pacific Plate is being subducted beneath source and crustal contamination of the magmatic the Australian Plate at the Tonga TrenchŽ Karig, systemŽ. James, 1981 . While the effects of both 1970. . The triangular shape results from rapid, crystallisation and open system behaviour have been clockwise rotation of the Tonga Ridge relative to the observed in some tholeiitic lava series, the magni- remnant Lau Ridge, and the southward propagation tude of variations that may be produced by these of two phases of ocean spreadingŽ Parson and processes are small: F0.5½Ž e.g. Taylor, 1968; Hawkins, 1994; Bevis et al., 1995. . At around 5.5 Anderson et al., 1971; Matsuhisa et al., 1973; Shep- Ma the first phase of spreading was initiated, south- pard and Harris, 1985; Woodhead et al., 1987; Ellam east of what is now Peggy Ridge, producing oceanic and Harmon, 1990. . Furthermore, identification of crust at the Eastern Lau Spreading CentreŽ. ELSC original magmatic 18 Or 16 O variation in submarine and the Valu Fa RidgeŽ. Fig. 1a . These spreading lavas is often obscured by the effects of post-erup- centres propagated southwards through extended arc tive alterationŽ Muehlenbachs and Clayton, 1972; crust between the Tonga and Lau ridges. Then, Pineau et al., 1976; Taylor and Sheppard, 1986. . around 1.5 Ma, the Central Lau Spreading Centre Phenocrysts or fresh volcanic glass can be analysed Ž.CLSC began to propagate through crust formed at to avoid alteration products but these phases are the ELSCŽ. Fig. 1a . commonly too scarce in seafloor dredge hauls to Parson et al.Ž. 1990 identified three active spread- allow conventional oxygen isotope analysis of entire ing segments in the Central Lau Basin, representing suites of lavas. propagating and dying rifts with an intermediate This study presents oxygen isotope data obtained `relay' sectionŽ. Fig. 1b . The crests of these ridges by laser-fluorinationŽ. LF for fresh lavas collected were dredged by Cruise 33 of the R.V. Charles from three sites of active volcanism in the Lau Darwin Ž.Pearce et al., 1994 . Eruptive products from Basin. Laser ablation of bulk samples vastly reduces the ELSC and Intermediate Lau Spreading Centre oxygen blanks compared to the classic `bomb' fluo- Ž.ILSC are dominantly basaltic with Mg No.)50, rination method and thus allows routine determina- although glasses from the northern ELSCŽ. site 20 tion of18 Or 16O ratios in igneous phasesŽ Sharp, are more evolvedŽ. Table 1 . Ferrobasalts and an- 1990; Mattey and Macpherson, 1993. . Furthermore, desites erupted at the propagating rift tip of the highly refractory phases can be analysed with high CLSC display evidence of extreme chemical frac- precision. Intensive sampling of a propagating rift tip tionation and volatile degassing, compared to other and associated dying rift in the Central Lau Basin Central Lau Basin lavas and mid-ocean ridge basalt has allowed examination of oxygen isotope systemat- ŽMORB; Pearce et al., 1994; Macpherson and Mat- ics through a diverse compositional spectrum of tey, 1994; Farley and Newman, 1994. . Plagioclase lavas generated in an oceanic setting. In an effort to dominated the crystallising assemblage in melts with detect subduction-related spatial controls on the 4 wt.%-MgO-7.5 wt.% subduing Sr enrichment 18 Or 16 O ratios of lavas, oxygen isotope data are and producing negative Eu anomalies during differ- also reported for lavas from the northern and south- entiation. Strong FeO, TiO2 and V depletions in ern extremities of the basin. These results are com- evolved lavas are the result of FeTi-oxide crystallisa- pared with other geochemical properties of the lavas tionŽ. Pearce et al., 1994 . Elevated concentrations of and conventionally obtained 18 Or 16 O data from incompatible alkali and alkali earth elements and nearby localities in an effort to constrain the flux of greater vesicularity in ELSC lavas, relative to MORB, C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 179

Fig. 1.Ž. a Map of the Lau Basin and Eastern part of the North Fiji Basin modified from Parson et al.Ž. 1994 . A heavy box outlines the propagating rift system of the Central Lau Basin shown inŽ. b , and the dashed line marks the approximate extent of oceanic crust formed at the Valu Fa Ridge and East Lau Spreading CentreŽ. ELSC . KK16 was recovered approximately 300 km west of Fiji. SO indicates the location of the Valu Fa samples for which detailed localities are presented by Frenzel et al.Ž.Ž. 1990 . b Geometry of the Central Lau Basin neovolcanic zonesŽ. stippled identified by Parson et al. Ž. 1990 , and dredge sites from Cruise 33 of the R.V. Charles Darwin. CLSCsCentral Lau Spreading Centre; ILSCsIntermediate Lau Spreading Centre; ELSCsEastern Lau Spreading Centre. The CLSC is propagating south at the expense of the ELSC. The ILSC is a `relay' zoneŽ. Parson et al., 1990 . 180 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 may reflect a greater flux of subducted material followed by oxide saturationŽ. Frenzel et al., 1990 . being recycling close to the arcŽ. Pearce et al., 1994 . Magma chambers have been seismically imaged, at The southern end of the Valu Fa Ridge marks the 1±4 km depth, along the entire length of the spread- limit of ocean spreading in the Lau BasinŽ Parson ing segment from which these samples were col- and Hawkins, 1994. . Collection sites and petrology lectedŽ. Collier and Sinha, 1992 . Radiogenic isotope of the three glasses analysed from this ridge are and trace element ratios suggest that a small but described by Frenzel et al.Ž. 1990 . Major element significant slab flux is present in Valu Fa lavas concentrations of the basaltic andesites are similar to ŽJenner et al., 1987; Sunkel, 1990; Loock et al., MORB and the daciteŽ. SO35-77KD2 was produced 1990.Ž. ; however, Hilton et al. 1993 showed that He by `normal basaltic differentiation' of basic melts isotopic variations are decoupled from Sr, Nd and Pb isotopes as a result of combined degassing and assimilation. Table 1 The Northern Lau and North Fiji Basin glasses Laser fluorination oxygen isotope data for Lau Basin and North Fiji Basin glasses were collected from disparate sitesŽ. Fig. 1a . Sample KK5 was recovered from the western end of Peggy Sampleab Type Mg No. c18dd OŽ. ½ s.d. n Yield % e Ridge while KK16-12 was recovered from the cen- Central Lau Spreading Centre Ž. LB10-1-1 Feb 27 5.77 0.06 2 94.6 tral part of the North Fiji Basin Sinton et al., 1993 . LB10-1-3 Feb 27 5.82 0.00 2 94.1 Chemically the KK lavas resemble MORB with the 87 LB12-5-1 a 31 5.53 99.2 exception of slightly elevated Srr 86 Sr ratios that LB12-5-2 a 31 5.64 0.03 2 94.8 are found in many tholeiitic lavas from the Lau and LB13-4 a 19 6.03 0.10 2 98.8 Fiji basins and FijiŽ. Sinton et al., 1993 . Sample LB13-5-1 a 19 6.00 100.5 LB13-5-2 a 19 5.92 101.2 231KD is a ferrobasalt from the northern edge of the LB15-1-2 b 59 5.72 0.04 2 102.7 North Fiji BasinŽ. Johnson and Sinton, 1990 , where LB15-1-4 b 59 5.80 103.9 other ferrobasalts have been recovered and magnetic anomalies and oblique magnetic fabric are observed East Lau Spreading Centre Ž. LB20-5-1 a 30 5.73 102.5 Sinton et al., 1993; Brocher, 1985 . Therefore, LB20-5-2 a 30 5.72 0.07 2 103.2 231KD probably evolved from MORB-like magma LB22-6-1 b 56 5.73 0.05 2 101.1 at a propagating rift tipŽ. Johnson and Sinton, 1990 . LB23-8-1 b 65 5.80 0.01 2 102.5 LB23-8-2 b 65 5.90 102.0 LB24-1-1 b 60 5.87 0.06 2 103.5

Intermediate Lau Spreading Centre Notes to Table 1: LB18-1-2 b 68 5.54 0.07 3 102.5 a Three numbers of LB samples represent, respectively, dredge LB41-2-1 b 66 5.89 101.6 siteŽ. within site , petrographic type and random fragment number. LB41-3-2 b 71 5.83 0.01 2 101.0 Full chemical descriptions of each, unique, glass can be found in Pearce et al.Ž.Ž. 1994 CLB , Johnson and Sinton Ž. 1990 , Sinton et Northern Lau and Fiji basins al.Ž.Ž. 1993 NLB and Frenzel et al. Ž.Ž. 1990 Valu Fa . KK5-15 b 55 5.69 0.15 2 101.8 b Lava types based on SiO content: bsbasalt; basbasaltic KK16-12 b 60 5.64 0.02 2 100.8 2 andesite; FebsFeTi-basalt; asandesite; dsdacite. 231KD Feb 36 5.91 0.11 2 96.7 c = r q 2q 100 Mg ŽMg Fe. assuming FeO:Fe23 O of 85:15. Simi- Valu Fa Ridge() Southern Lau Basin larity of Mg No. within each Central Lau Basin site is a function SO35- ba 66 5.88 99.9 of the samples available and not the overall chemistry of each site Ž.see Pearce et al., 1994 . 124KD1 d SO35- ba 45 5.94 0.06 2 100.2 Where more than one determination was made the mean is Ž. Ž. 91KD1 reported with standard deviation s.d. and number of analyses n listed. SO35- d 34 6.06 95.6 e 77KD2 Percent of oxygen expected are based on major element analyses of each glass. MORB Averagef 5.79 0.30 123 f Compilation by MacphersonŽ. 1994 . NLB Averageg 5.82 0.18 12 g Northern Lau Basin lavasŽ Pineau et al., 1976; Lichtenstein and Loock, 1993. . C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 181

Oxygen isotope ratios of volcanic glass and phe- ratios calculated for the feldspar are similar to con- nocrysts were analysed using the Royal Holloway ventional determinationsŽ. Macpherson, 1994 . Like University of London laser-fluorinationŽ. LF system clinopyroxene, glass reacts vigorously during the Ž.Mattey and Macpherson, 1993 . This is similar to early stages of fluorination and care must be taken to conventional oxygen extraction linesŽ e.g. Clayton prevent sputtering of molten silicate away from the and Mayeda, 1963. with the exception that, rather reaction site. Subsequently, glass melts into a pool, than fluorination of powdered samples in externally at the bottom of the sample hole, and gently reacts heated nickel `bombs', a Nd : YAG laser is used to until oxygen extraction is complete. Over the period directly heat uncrushed samples in the presence of a of this study NBS-30 biotite gave d 18 O of 5.03½ fluorinating agentŽ Mattey and Macpherson, 1993; Ž."0.08, ns80 , and all data are normalised to the Macpherson, 1994. . This has the advantage of reduc- recommended NBS-30 value of 5.1½. Reproducibil- ing background contributions, thus permitting routine ity for various phases can be compared with the analysis of samples in the 0.02±2 mg rangeŽ Mattey Royal Holloway secondary standards San Carlos and Macpherson, 1993. . Samples were ultrasonically olivine and GP101 clinopyroxene that gave 4.88½ washed in distilled water and again in acetone prior Ž.Ž."0.09, ns26 and 8.25½ "0.06, ns10 , re- to weighingŽ. to "1 mg , then 0.4 to 2.5 mg of spectivelyŽ. cf. Mattey and Macpherson, 1993 . There material was selected under a binocular microscope. are no internationally recognised feldspar standards Care was taken to avoid visible alteration and phe- for 18 Or 16 O so analyses of NBS-28 were also con- nocrysts in glass fragments. Oxygen released from ducted by mixing with olivine giving a d 18 Oof " s silicates was converted to CO2 over hot graphite for 9.58½ Ž.0.17, n 8, recommended 9.66½ . isotopic analysis on a VG PRISM mass spectrome- Therefore greater uncertainty Ž."0.17½ is attached ter. Fractionation of oxygen isotopes during gas han- to feldspar analysesŽ. Macpherson, 1994 . Yields for dling was avoided through selecting suitable sample all standards were 100% Ž."2% . sizes and employing graphite that had previously performed many conversionsŽ Mattey and Macpher- son, 1993. . Isotope ratios are reported as per mil 3. Results deviation from SMOW in the standard d notation. The difference between d 18 O values of any two Results of LF analyses are listed in Tables 1 and 18 18 materials i and j; D Oiyj , is assumed to be propor- 2. The total range of d O values in fresh glasses is tional to the isotopic fractionation factor 5.53½ to 6.06½ with replicate analyses generally f 18 " aŽ1000 ln aiyjid O.yj . agreeing to better than 0.1½Ž. 1s , Table 1 . The The behaviour of ferromagnesian silicates during three fresh Northern Lau and Fiji Basin glasses range Nd : YAG laser ablation is described by Mattey and from 5.64½ to 5.91½, similar to bomb fluorination MacphersonŽ. 1993 . Feldspar is transparent to 1064 analyses of unaltered whole rock from the Northern nm radiation; however, analyses of infra-red trans- Lau BasinŽ. Pineau et al., 1976 , and a marginal parent phases can be achieved by mixing with an increase in d 18 O values is observed with degree of olivine or clinopyroxene of known isotopic composi- chemical evolution, as indicated by decreasing Mg tion. Such mixtures fluorinate readily and 18 Or 16 O NumberŽ. Table 1 . The Valu Fa glasses also display

Table 2 Oxygen isotope ratiosŽ. ½ of phenocrysts from Central Lau Basin lavas Sample Olivine Yield %aaa Clinopyroxene Yield % Plagioclase Yield % LB15-1-2 4.82 103.3 LB20-5-2 6.05 101.2 LB23-8-1 5.15 102.2 LB24-1-1 4.87 98.7 5.28 98.4 6.12 100.8 a Calculated assuming wt.% oxygen contents of 44.0% in olivine, 45.5% in clinopyroxene, 48.0% in plagioclase. 182 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 a minor increase in d 18 O values with decreasing Mg mineral pairs involving plagioclase fall within the Number, from 5.88½ to 6.06½. For a given Mg ranges obtained conventionally for basic rocks Number, our analyses of Valu Fa glasses fall at the ŽAnderson et al., 1971; Friedrichsen and Hoerns, 18 lower end of the range of southern Lau Basin whole 1979. . Values for D Omineralyglass obtained for rock d 18 O values obtained by bomb fluorination LB24-1-1 are similar to conventionally determined 18 Ž.Vallier et al., 1991; Lichtenstein and Loock, 1993 . groundmass±phenocrysts pairs, with D Oolyglass d 18 y 18 y O values of glasses from the Central Lau Ž.1.0½ and D Ocpxyglass Ž.0.59½ values falling Basin range from 5.53½ to 6.03½. The oxygen in the middle of the previously determined range 18 isotope ratio of LB18-1-2 is significantly lower than while D Oplagyglass lies at the higher end of the other Central Lau Basin basaltsŽ. Table 1 . This is an conventional data rangeŽ Macpherson, 1994, and ref- olivine±phyric lava that has the highest MgO and Ni erences therein. . contents of the Central Lau Basin suite, features that have been attributed to accumulation or resorption of olivine phenocrysts or xenocrystsŽ Pearce et al., 4. Discussion 1994. . Olivine is depleted in18 O by approximately Exposure of submarine lavas to seawater after 1½ relative to molten basaltŽ Anderson et al., 1971; eruption and cooling can result in growth of sec- Muehlenbachs and Clayton, 1972; LB24-1-1. ; there- ondary minerals that can significantly enrich the 18 O fore, olivine resorption should lower d 18 O of a melt, content of the rockŽ e.g. Muehlenbachs and Clayton, in addition to enriching Mg and Ni. While it may be 1972; Pineau et al., 1976. . Volcanic glass is particu- unreasonable to advocate olivine accumulation alone larly susceptible to such alteration and caution must to explain the entire 0.28½ difference between this be exercised in attributing d 18 O variations to mag- glass and the mean value of the remaining basalts, matic processes until the extent of post-eruptive al- the consistency of oxygen isotope ratio, petrography teration can be judged. However, correlations be- and chemistry in this sample are noteworthy. In tween oxygen isotope ratios and water concentra- addition to oxygen isotopes, several other geochemi- tions, that would be expected if alteration were a cal properties of melts may be influenced by phe- major controlŽ Muehlenbachs and Clayton, 1972; nocryst accumulation, therefore LB18-1-2 is omitted Pineau et al., 1976. , are not observed in the Lau from discussion of magmatic 18 Or 16 O systematics. Basin glassesŽ. Fig. 2 . Phenocrysts are less prone to The remaining basic Central Lau Basin lavasŽ Mg alteration than volcanic glass so extents of alteration No.)50. have an average d 18 O value ofŽ 5.82½, can also be estimated by examining D18 O values of "0.07½. that is identical, within error, to both the coexisting glass and phenocrystsŽ Muehlenbachs and MORB average and the average of whole rock Clayton, 1972; Kyser, 1986. . As noted above, iso- 18 Or 16O ratios of Northern Lau Basin lavasŽ Pineau topic differences between mineral separates and glass et al., 1976; Ito et al., 1987; Lichtenstein and Loock, appear to be representative of magmatic values and 1993; Table 1. . The d 18 O values of ferrobasalts and in particular we note the small, positive D18 O andesitesŽ that have Mg No.-40; sites 10±13 and plagyglass values obtained for all plagioclase separates. This 20. encompass the entire d 18 O range of the Central value would probably be reversed by even small Lau Basin suite and higher values are associated degrees of alteration to glass. These observations with lower Mg NumbersŽ. Table 1 . suggest that d 18 O values of the glasses have experi- Oxygen isotope ratios of phenocrysts are listed in enced negligible modification since eruption and that Table 2. With respect to host glasses, d 18 O values of 18 Or 16 O ratios in coexisting phases of Lau Basin olivine and clinopyroxene phenocrysts are low, while lavas represent equilibrium at magmatic tempera- d 18 O values of plagioclase are high. The D18 O cpxyol tures. value of 0.41½ in LB24-1-1 is similar to LF analyses of mantle nodules and to values determined by 4.1. Oxygen isotope composition of eÕolÕed laÕas conventional analysis of basic terrestrial lavas ŽAnderson et al., 1971; Friedrichsen and Hoerns, A strong negative correlation between d 18 O and 1979; Mattey et al., 1994. . The d 18 O values for Mg Number Ž.Rs0.90 exists for those Central Lau C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 183

Fig. 2. Relationship between water content and d 18 O values of Lau Basin lavas. Large open symbols, this work; small closed symbols, whole rock bomb fluorination data of Pineau et al.Ž. 1976 , Vallier et al. Ž. 1991 and Lichtenstein and Loock Ž. 1993 . Additional water contents from Aggrey et al.Ž. 1988 and Farley and Newman Ž. 1994 . Abbreviations: ELSCsEast Lau Spreading Centre; ILSCsIntermediate Lau Spreading Centre; CLSCsCentral Lau Spreading Centre; NFLsNorthern Lau and Fiji basins. Error bars omitted for clarity. The diagonal line represents the effect of post-eruption hydration on whole rock lava samples from the Northern Lau Basin and the shaded field represents bulk water contentsŽ. Pineau et al., 1976 . Water concentrations of Central Lau Basin glasses determined by Fourier Transform Infra-Red SpectroscopyŽ. FTIR that detects only water dissolved in the glass Ž Farley and Newman, 1994 . .

Basin lavas with Mg No.-40 erupted close to the for lowering melt d 18 O values. However, the Cen- terminations of the propagating and dying rift tips tral Lau Basin intermediate and evolved lavas ap- Ž.Fig. 3a . A similar relationship, attributed to mag- proximate a liquid line of descent characteristic of netite crystallisation, also exists for evolved lavas basalt±ferrobasalt±andesite series. Chemical evi- from the Galapagos Spreading Centre propagating dence for phenocryst accumulation, which is ob- rift tip at 958WŽ. Muehlenbachs and Byerly, 1982 . served in nearby seamount lavas, is not apparent in Oxides possess low d 18 O values, relative to basaltic the majority of the spreading centre lavasŽ Pearce et melt, at magmatic temperatures so oxide crystallisa- al., 1994. . Differences between the inferred d 18 Oof tion, even in limited volumes, can produce signifi- intermediate melts and local basalts are not largeŽ i.e. cant18 O enrichment of meltsŽ Friedrichsen and Ho- 0.25½ or more. and are of a similar magnitude to erns, 1979; Muehlenbachs and Byerly, 1982. . The variations observed in individual phases from mantle coincidence of 18 O enrichment with Fe, Ti and V peridotitesŽ. Mattey et al., 1994 . While it could depletion in the Central Lau Basin suite is strong therefore be inferred that the intermediate melts were evidence for operation of a similar processŽ. Fig. 3 . themselves derived from low-18 O basaltic parents, 18 18 In addition, we note that D Oplagymelt tends toward there is no evidence for low- O basaltic glasses of negative values in evolved meltsŽ Taylor and Shep- liquidus composition. Chemical evolution has ex- pard, 1986. , also enhancing18 O-enrichment during erted significant control on the major and trace ele- fractional crystallisation. ment chemistry of the entire compositional spectrum If fractional crystallisation of a magnetite-bearing of Central Lau Basin lavasŽ. Pearce et al., 1994 , and assemblage is responsible for 18 O-enrichment of the therefore we believe that differentiation from basaltic evolved Central Lau Basin lavas, the Mg No.±d 18 O compositions to the point of oxide saturation in relationship suggests that precursor intermediate propagating rift tip magma chambers involved a melts were18 O-poor Žd 18 OF5.50½. relative to decrease in melt d 18 O values. Mechanisms that could basalts erupted nearbyŽ. Fig. 3a . Olivine resorption lower 18 Or 16 O during evolution towards lower MgO 18 has been mentioned above as a possible mechanism contents are:Ž. 1 a positive value for D Oassemymelt 184 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194

clase was the most abundant phase crystallising from Central Lau Basin basaltic melts including: negligible Sr enrichment, development of negative Eu anomalies, and enrichment of some mildly compatible trace elements, such as V, Co and Sc, as evolution proceedsŽ. Fig. 3b . Therefore, the crystallising assemblage may remove 18 O from basaltic melts in the Central Lau Basin more efficiently than in other suites of basalts. The relative abundances of plagioclase and ferromagnesian phases in the crystallising assemblage can be estimated since the Sr concentrations in the Cen- tral Lau Basin lavas are relatively constant. This suggests that the bulk partition coefficient for Sr was approximately unity. Using reasonable estimates of Sr distribution coefficients between melt and the observed phenocrysts requires crystallisation of olivine : clinopyroxenes :plagioclase in the propor- 18 tions 1 : 1 : 3. Applying the D Omineralyglass values we have measured in LB24-1-1Ž. Tables 1 and 2 to these phenocryst proportions produces an estimate 18 y for D Oofassemymelt 0.17½. Proportions of 1 :1:6 are required to produce a zero value for 18 D Oassemymelt , but would lead to significant Sr de- Fig. 3.Ž. a Mg Number versus d 18 O for Lau Basin glasses. Error pletion in more evolved lavas without depleting 18 O. bars are "1 s.d. for replicated analyses or "0.08½ otherwise. Therefore, while plagioclase crystallisation was un- Symbols and data sources as in Fig. 2. Stippled box represents the doubtedly important during evolution of Central Lau Ž.Ž. average reported for mid-ocean ridge basalts see Table 1 . b Mg D18 Number versus VŽ. ppm for Lau Basin lavas Ž Frenzel et al., 1990; Basin melts, and Oassemymelt was probably more Johnson and Sinton, 1990; Sunkel, 1990; Pearce et al., 1994. . In positive than is true of many cogenetic suites of the Central Lau Basin suite the onset of V depletion is coincident lavas, fractional crystallisation of the observed as- with Fe and Ti depletion inferring oxide saturation in the melts. semblage cannot produce melts that are 18 O-depleted, relative to their basaltic precursors. Addi- tional processes must be considered to account for Ð the difference between the isotope ratio of the the low d 18 O values inferred for propagating rift tip crystallising assemblage and the melt;Ž. 2 addition of lavas immediately prior to oxide saturation. a18 O-poor componentŽ.Ž. Taylor, 1987 ; or 3 some combination ofŽ. 1 and Ž. 2 . 4.1.2. Assimilation and the depth of magma cham- 4.1.1. Fractional crystallisation bers Previous studies and the data presented here sug- Assimilation of 18 O-poor country rock has been 18 18 gest that D Oplagymelt , is positive for basaltic melt at proposed to explain low d O values in lavas erupted magmatic temperaturesŽ Muehlenbachs and Clayton, through thick, oceanicŽ. hot-spot crust in Iceland and 1972; Woodhead et al., 1987, and references therein; Hawaii, and island arc crust in IndonesiaŽ Muehlen- 18 Tables 1 and 2. . Therefore, a positive D Oassemymelt bachs et al., 1974; Hattori and Muehlenbachs, 1982; value could be achieved if plagioclase crystallisation Condomines et al., 1983; Taylor, 1987; Harmon and sufficiently exceeded the combined crystallisation of Gerbe, 1992; Eiler et al., 1996; Thirlwall et al., olivine and clinopyroxene. Pearce et al.Ž. 1994 pro- 1996. . Rocks with low d 18 O, relative to average pose several lines of evidence to suggest that plagio- MORB, occur at depths of 2±5 km beneath the C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 185

Ž.palaeo- ocean floor in the lower sheeted dykes and modern ocean crust sequences typically range from layered gabbros of both modern ocean crust and 3.5½ to 4.0½Ž Muehlenbachs, 1986, and references ophiolitesŽ Gregory and Taylor, 1981; Ito and Clay- therein. . Assuming initial CLSC basalts have d 18 O ton, 1983; Muehlenbachs, 1986; Stakes et al., 1991; values of 5.75½Ž. site 15, 7.5 wt.% MgO , then 10% Kempton et al., 1991; Lecoyer and Reynard, 1996. . and 15% assimilation of crust with oxygen isotope Low d 18 O values develop through interaction be- ratios of 3.5½ and 4.0½, respectively, is required to 18 tween crust and seawater at relatively high tempera- produce the lowest d Oglass value. Spulber and tures Ž.)3008C in the hydrothermal regime associ- RutherfordŽ. 1983 showed that the point of oxide ated with ocean ridge magmatismŽ Gregory and Tay- saturation in propagating rift tholeiites depends on lor, 1981. . Mineralogical changes associated with f of the melt but that at 1 kbar the fraction of melt O 2 this process are the alteration of feldspar and growth remaining, F, would be approximately 0.3. There- of hydrous phases such as amphibole, chlorite and fore, if the lowest oxygen isotope ratio occurred epidoteŽ. Ito and Clayton, 1983; Stakes et al., 1991 . immediately prior to oxide saturation, we can esti- It is these altered and hydrous, secondary phases at mate that r, the ratio of mass assimilated to mass intermediate depths that represent the major 18 O-poor crystallised, is approximately 0.2Ž i.e. ; reservoir in oceanic crust. Intrusion of evolved melts 12.5%r70%. . These estimates of mass assimilated into brecciated regions at mid-crustal levels has been and r should be regarded as maxima for three 18 reported in the Oman ophiolite and the South West reasons. As noted above, the D Oassemymelt value in Indian Ocean RidgeŽ Gregory and Taylor, 1981; Central Lau Basin basaltic melts was probably less Stakes et al., 1991. . Such associations could provide negative than is regarded typical of most spreading an ideal situation for melt±crust interaction and may centre melts; therefore, fractional crystallisation provide an explanation for low d 18 O values in inter- would be less efficient in enriching the 18 O content mediate Central Lau Basin melts. of the melts. Second, the least refractory phases in Evidence for melt stagnation at suitable depths the wall rock are the hydrous and altered minerals beneath the ocean-floor exists both for propagating formed by hydrothermal processes. Compared to bulk rift tip lavas in general, and specifically for the contamination, selective assimilation of a smaller Central Lau Basin suite. Macpherson and Mattey mass of these 18 O-poor phases could produce similar

Ž.1994 modelled CO2 degassing of the Central Lau isotopic shifts in the melts. Finally, a smaller quan- Basin glasses analysed in this work and proposed tity of contaminant would also be required if the that volatiles and melt equilibrated at a minimum melts interacted with very low18 Or 16O crust Ž- depth of 2.5 km beneath the seafloor. The absence of 3½.Ž , such as found in the Hess Deep Lecoyer and amphibole, particularly in evolved, water-rich lavas Reynard, 1996. . such as those of the CLSC, suggests a maximum Therefore, we suggest that the melts that were depth for differentiation of 5 kmŽ Spulber and precursors to evolved Central Lau Basin lavas assim- Rutherford, 1983; Farley and Newman, 1994. . In ilated -15% of their own mass of d 18 O-depleted addition, experiments on oceanic tholeiite from the oceanic crust during crystallisation of a plagioclase- Galapagos Spreading Centre suggest that evolution dominated assemblage. A combination of these pro- of tholeiitic melt to ferrobasalt and more evolved cesses produced low d 18 O intermediate melts. As- CLSC compositions is most likely to occur at depths similation of such quantities of wall rock is easily of 3±6 kmŽ. Spulber and Rutherford, 1983 . There- reconciled with the limited size of magma chambers fore, both the petrology and carbon isotope geochem- beneath propagating rift tips where waxing magma- istry of Central Lau Basin lavas point to stagnation tism would be accompanied by relatively large tem- of melts at depths of 2.5±5 km beneath the seafloor. perature contrasts between magma and wall rock Simple mass balance calculations can be used to Ž.Christie and Sinton, 1981; Perfit et al., 1983 . Oxide estimate the volume of oceanic crust that must be saturation, at Mg No.f45, had a major influence on assimilated to generate the lowest oxygen isotope oxygen distribution due to the large, negative value 18 ratios observed from melts similar to the local basalts. of D Ooxideymelt causing further crystallisation to Minimum whole rock d 18 O values in ophiolite and enrich melts in18 OŽ Muehlenbachs and Byerly, 186 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194

18 1982. . Reversal of D Oplagymelt in evolved melts generated at the ELSC approximately 1 MaŽ Parson may also have contributed to reducing the d 18 O et al., 1994. . Since there has been little time for in value of the crystallising assemblage with respect to situ growth of 4 He, assimilation of young crust the meltsŽ. Taylor and Sheppard, 1986 . It is possible should be expected to have little effect on He isotope that the most evolved melts may have continued to ratios in CLSC meltsŽ Kurz and Jenkins, 1981; Hilton assimilate 18 O-poor crust, but the effects have been et al., 1993. . This is confirmed by a3 Her 4 He ratio masked by oxide crystallisation. Clearly, the inter- of 8.6 Ra, indistinguishable from the MORB aver- play of crystallisation and assimilation may generate age, measured in a CLSC rift tip glass despite a very a wide range of evolutionary vectors in Mg No.± low He concentration that would have made the melt d 18 O space, depending on the mineralogy and iso- highly susceptible to contaminationŽ Poreda and topic composition of the crust and the value of Craig, 1992. . Low noble gas concentrations are also 18 D Oassemymelt during differentiation. observed in glasses from the Valu Fa Ridge, that are Trace element concentrations in CLSC lavas are thought to result from extensive degassing of noble similar to MORB except for slight enrichment of the gases prior to assimilationŽ. Hilton et al., 1993 . large ion lithophile elementsŽ. LILE relative to high However, carbon isotope systematics suggest that field strength elementsŽ. HFSE and light rare earth d 13 C values of Central Lau Basin glasses were elementsŽ. LREE with similar compatibilities. Pearce largely controlled by eruption that must occur after et al.Ž. 1994 suggest that these features could result assimilation of magma chamber wall rocksŽ Mac- from contribution of either modern or ancient sub- pherson and Mattey, 1994. . Therefore, the chemical duction components. However, we note that evolved effects of assimilation may provide a useful refer- lavas from the Galapagos propagating rift tips, re- ence point in resolving pre- and syn-eruptive de- mote from any subduction zone, also display trace gassing events affecting ocean ridge magmatism. element enrichments that exceed the predictions of closed-system crystallisation modelsŽ Mattey and Muir, 1980; Perfit et al., 1983. . O'Hara and Matthews 4.2. EÕolution of Valu Fa, Northern Lau Basin and Ž.1981 showed that cannibalisation of previous erup- Fiji Basin laÕas tive products by a magma chamber will enrich incompatible element concentrations of melts while the The Valu Fa glasses analysed in this work are effects on compatible element concentrations would representative of a larger suite collected from a small be negligible. Crust beneath the CLSC rift tip is very area Ž.-30 km along the ridge . Chemical composi- similar to presently erupting CLSC melts in terms of tions of the entire suite approximate a single liquid major and trace elements, but with slightly higher line of descent involving olivineqplagioclaseq LILE concentration due to their origin closer to the clinopyroxene crystallisation, with an oxide joining Tonga ArcŽ e.g. DSDP Site 836, Ewart et al., 1994; the assemblage at a late stageŽ. Frenzel et al., 1990 . Parson et al., 1994. . Therefore, assimilation of Increasing d 18 O values with decreasing Mg Num- oceanic crust may be, at least partially, responsible bers are consistent with fractional crystallisation of for the small LILE enrichment observed in CLSC the observed phenocrysts and resemble 18 O-enrich- lavas, either through the LILE-rich nature of the ments reported from many cogenetic suites of lavas crust andror LILE concentration during differentia- ŽMatsuhisa et al., 1973; Muehlenbachs and Byerly, tionŽ. O'Hara and Matthews, 1981 . 1982; Sheppard and Harris, 1985; Woodhead et al., Crustal assimilation has also been advocated by 1987. . The highest d 18 O value in this study was Hilton et al.Ž. 1993 to explain radiogenic He and obtained for the Valu Fa dacite glass and is consis- atmospheric Ar isotopic ratios in evolved spreading tent with oxide saturation during its differentiation. centre lavas from Valu Fa Ridge. The Valu Fa Therefore, d 18 O values of Valu Fa glasses presented spreading centre propagated through attenuated arc here may reflect fractional crystallisation of a melt crust that is the inferred source for the radiogenic with d 18 O similar to SO35-124KD1. heliumŽ. Hilton et al., 1993 . However, the CLSC Oxygen isotope ratios obtained by Lichtenstein propagated through relatively young back-arc crust and LoockŽ. 1993 for Valu Fa Ridge lavas display C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 187 the greatest variation yet observed for fresh Lau topes of some Valu Fa lavasŽ. Hilton et al., 1993 . Basin lavas and show considerably greater variation The arc crust through which the Valu Fa Ridge has than our analyses of glasses from this spreading propagated may possess considerable 18 Or 16 O het- centreŽ. Fig. 3a . The lower end of the range observed erogeneity since it has been constructed by succes- by these workers coincides with our data so may also sive cycles of arc magmatism, probably including the result from differentiation of a MORB-like melt. The development of hydrothermal systems, and sedimen- higher d 18 O values, at a similar Mg No., require tation related to both the Lau and Tonga arcsŽ Parson addition of an 18 O-rich component. While Sr, Nd and Hawkins, 1994, and references therein. . There- and Pb isotope ratios of some of the 18 O-rich lavas fore, rift propagation through rifted arc crust may be provide evidence for recycling of subducted material even more conducive to crustal assimilation than is Ž.Loock et al., 1990 , coexistence of lavas with vari- the case for propagation through oceanic crust, and able d 18 O on the same spreading segments suggests may be an important process in the development of either that oxygen isotope heterogeneity exists on a the total d 18 O variation observed in Valu Fa Ridge very short length scale in the mantle beneath the lavasŽ. Fig. 3a . Further investigation of d 18 O varia- Valu Fa Ridge or that melts have interacted with tions in Valu Fa lavas and attenuated Southern Lau 18 O-rich crust to varying degrees. Establishment of Basin crust are required to constrain the roles of new magmatic plumbing systems associated with source and magma chamber processes on what is one ridge propagation provides an ideal environment for of the world's most petrologically diverse suites of crustal assimilation to occur, as has been convinc- spreading centre lavas. ingly demonstrated for radiogenic and rare gas iso- The Northern Lau and Fiji Basin glasses were

Fig. 4. Relationship between d 18 O values and BarLa in Central Lau Basin glasses. Error bars as in Fig. 3. Three binary mixing curves for slab-derived hydrous fluid and mantle peridotite are shown: Mscontamination of fertile MORB source by fluid derived from altered oceanic crust and sediment; Dscontamination of depleted MORB source by fluid derived from altered oceanic crust and sediment; Pscontamination of fertile MORB source by fluid derived from sediment. Tick marks indicate the fraction of fluid in the source. Element abundances for mantle sources and fluids from Ewart and HawkesworthŽ. 1987 and Hawkesworth et al. Ž. 1991 , respectively. An oxygen ratio of 2 : 1 was assumed forŽ. hydrous fluid : peridotite. Oxygen isotopic compositions are "5.75½ for melt from uncontaminated peridotite and "19½ for Pacific sedimentŽ. Savin and Epstein, 1970 . 188 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 collected over a wide area and are unlikely to represent a single liquid line of descentŽ. Fig. 1a . While 18 O-enrichment with decreasing Mg No. is observed, and again appears to be concordant with whole rock powder analysesŽ. Fig. 3a , the d 18 O variation cannot be unequivocally attributed to fractional crystallisation.

4.3. Source of oxygen in Lau Basin basalts

Oxygen isotope compositions of basaltic glasses Ž.Mg No.)50 from the Central Lau Basin display a very weak positive correlation between Mg Number 18 18 18 and d O that may reflect evolution towards O-de- Fig. 5. Relationship between Na8.0 and d O values in Central pleted compositions by the processes outlined in the Lau Basin, NFL and Valu Fa basaltsŽ. 5 wt.%-MgO-8.5 wt.% . previous sectionŽ. Fig. 3a . Assimilation of altered Data are site averaged where more than one sample is available or collection sites are less than 10 km apart. Symbols and d 18 O ocean crust has recently been advocated to explain sources as in Fig. 2. Na of Central Lau Basin lavas from Pearce 18 8.0 low d O values in ocean island basalts from Hawaii et al.Ž. 1994 and calculated from the algorithm of Klein and and Gran CanariaŽ Eiler et al., 1996; Thirlwall et al., LangmuirŽ. 1987 for other sites. Error bars for d 18 O are "1s of 1996. . We also note that cogenetic suites from As- averages, "1s of analytical error for individual LF analyses or " cension Island and Chichijima Island each display a 0.2½ for individual bomb fluorination analyses. Error bars for Na are "1s of averages or "0.1. d 18 O minimum at intermediate compositionsŽ Shep- 8.0 pard and Harris, 1985; Dobson and O'Neil, 1987. . Therefore, assimilation of hydrothermally altered " crust Ž.enhanced plagioclase crystallisation may correlation between Na8.0 and Fe 8.0 among the basalts play a more important role in the evolution of oceanic may result from variable mantle fertility. Residual basalts than previously realised. However, the major sources produce basaltic melts with lower sodium and trace element chemistry of the Central Lau Basin concentrations, at a given MgO content, possibly due basalts reveals that recycling of subducted water and to previous extraction of a relatively Na-rich melt mantle source fertility influence their chemistry Ž.Pearce et al., 1994 . The most depleted mantle Ž.Pearce et al., 1994 . Correlations between oxygen signatures are found in ELSC lavas close to the arc isotope ratios and chemistry of the basaltic glasses suggesting that slab-derived fluid may encourage that are described in this section suggest that these melting of residual peridotite. process also influence 18 Or 16 O. Figs. 4 and 5 explore the possibility that recycling Pearce et al.Ž. 1994 have demonstrated that water and source fertility may influence oxygen isotope may be recycled into the source of some ELSC and ratios of basaltic lavas erupted in the Lau Basin. Fig. ILSC basalts from the down-going slab. The pres- 4 displays a positive correlation between d 18 O and ence of water is manifested in several ways. First, BarLa in Central Lau Basin basalts that extends enrichment of some LILE elements, such as Ba and from MORB-like values along a limited trajectory Rb, in ELSC lavas most probably results from the towards possible recycled fluid compositions. Simple high mobility of these elements in hydrous fluids models of mantle wedge contamination by fluids released from subducted materialŽ Tatsumi et al., derived from altered MORBqsedimentŽ curves M 1986; You et al., 1996. . Second, hydrous melting of and D.Ž. and sediment alone curve P have been peridotite close to the Tonga Arc is inferred from calculated for fertile and depleted peridotiteŽ see high Si8.0 values of ELSC lavas that are thought to caption to Fig. 4 for details. . The best fit to the reflect enhanced orthopyroxene melting at high P Central Lau Basin data is obtained for models in- HO2 Ž.cf. Jaques and Green, 1980 . Finally, a positive volving a sediment-derived fluid andror a depleted C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 189 mantle source. Fluid fluxes required to produce the Fig. 5 could be interpreted as a result of Lau Basin d 18 O variation among the ELSC and ILSC lavas are basalts being extracted from mantle that had previ- F0.5% and cannot be resolved on plots involving Sr ous lost varying fractions of a relatively Na-rich, and Nd isotopesŽ. Hergt et al., 1991 . 18 O-poor melt. However, we believe that this expla- Our data for Central Lau Basin basaltic glasses nation cannot be reconciled with the mass balance of display a positive correlation between Na8.0 and oxygen. Generation of tholeiitic melt from peridotite d 18 O, and several bomb fluorination analyses of requires 10±30% partial meltingŽ Jaques and Green, Northern and Central Lau Basin lavas also fall within 1980. and, since oxygen abundances are similar in error of this trendŽ. Fig. 5 . Note that the most molten and crystalline silicate, oxygen isotope distri- significant departures from this trend are due to high bution during peridotite melting should be controlled d 18 O values of some Valu Fa lavas that may reflect by the residue. Therefore, melting of peridotite with 18 18 the presence of a recycled or assimilated O-rich positive D Oresidueymelt values would produce pri- componentŽ. see above . Similar relationships can be mary tholeiites with d 18 O values lower than bulk demonstrated between d 18 O and other geochemical peridotiteŽ. i.e. -5.5½, Mattey et al., 1994 . Frac- properties that reflect variations in source fertility, tional crystallisation is unable to enrich melts by r r r e.g. Ti Zr, Al23 O TiO 2 and CaO TiO 2Ž Ewart and more than a few hundredths of a per mil without Hawkesworth, 1987; Woodhead et al., 1993. . These driving them to SiO2 -rich compositions, outside the correlations clearly suggest that a link exists between basalt fieldŽ Muehlenbachs and Byerly, 1982; Shep- fertility of the Lau Basin mantle and oxygen isotope pard and Harris, 1985. ; therefore low d 18 O basalts ratios of extracted melts. Two origins can be postu- would be commonplace on the Earth's surface. The lated for this relationship:Ž. 1 an indirect link as a reverse is true: MORB, the most voluminous molten result of recycling processes, orŽ. 2 direct control of extract of the mantle, have an average d 18 O value 18 Or 16 O by mineralogy and melting. greater than the peridotite meanŽ Table 1; Ito et al., An indirect link could arise from a greater fluid 1987. . Thus, we are not convinced that 18 flux being required to induce melting in residual D Oresidueymelt values are positive and do not be- mantleŽ. Pearce et al., 1994 . In this case, a larger lieve that 18 O-rich melts from residual peridotite are proportion of 18 O-rich, slab-derived fluid would be generated due to previous extraction of a low-18 O associated with melting of the least fertile portions of melt. the wedgeŽ. Fig. 5 . Although both Ž18 O-rich. fluid A direct link between 18 Or 16 O and mantle fertil- and depletedŽ. low-Na8.0 peridotite would be re- ity can be explained if isotopic fractionation factors quired to produce the observed melts, the fertility of during melting are similar to values observed and the source would not directly cause the high d 18 O inferred from natural lavas and peridotitesŽ i.e. 0) 18 ) 18 ) 18 values of melts. D OenstymeltD O cpxymeltD O olymelt ; cf. Mat- Alternatively, mantle fertility may directly pro- tey et al., 1994. . Under equilibrium conditions induce the oxygen isotope heterogeneity of Lau Basin congruent melting of orthopyroxene will increase the basalts. Preliminary isotopic fractionation data for modal abundance of olivine in peridotite residue 18 the systems enstatite±basaltic meltŽ Muehlenbachs producing a largerŽ. more negative D Oresidueymelt and Kushiro, 1974. and olivine±tholeiitic melt value. Since a greater volume of oxygen is present in Ž.Kyser, 1990 suggest that melts may be18 O-de- the residue than in melts, this increase in 18 18 pleted relative to their peridotite residues at very D Oresidueymelt will tend to enrich melts in O. 18 high melting temperatures, i.e. D Oenstymelt and Orthopyroxene melting is enhanced during melting 18 D Oolymelt are positive. Based on then available of residual peridotiteŽ. Jaques and Green, 1980 ; peridotite data, Kyser et al.Ž. 1982 proposed this as a therefore, increasingly residual peridotite will yield means to enrich the 18 O content of peridotite residues increasingly 18 O-rich equilibrium melts. High-18 O during melting and to explain global oxygen isotope melts could also be generated from residual peri- differences between different types of basalt ob- dotite by disequilibrium melting. Orthopyroxene served by Kyser et al.Ž. 1981 . Employing these consistently has the highest d 18 O value among the experimental fractionation factors the relationship in major peridotite phasesŽ. Mattey et al., 1994 . There- 190 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 fore, since orthopyroxene contributes a relatively water concentrations in lavas throughout the range of large fraction to the melt during melting of residual compositions. However, this observation does prompt peridotiteŽ. Jaques and Green, 1980 , melts produced the following important comment. Water contents from more harzburgitic sources would be relatively are frequently used to correct oxygen isotope analy- 18 O-rich. We note that melts from residual sources ses of whole-rock powders for the effects of post- tend to have relatively high values of Mg Number eruptive alteration by assuming that all alteration Ž.Jaques and Green, 1980 and this could explain the produces simultaneous hydration and 18 O enrichment weak Mg No.±d 18 O correlation in the basaltsŽ Fig. Ž.e.g. Ferrara et al., 1985; Harmon et al., 1987 . The 3a. . A larger recycled fluid-flux is still inferred to Lau Basin data show that evolution of oceanic tholei- produce melting of residual peridotiteŽ Pearce et al., ites through combined assimilation and fractional 1994. , thus producing the BarLa±d 18 O correlation crystallisation may elevate water concentrations 18 of Fig. 4. However, in the scenarios outlined in this without modifying d O and that magmatic H2 O paragraph the fluid does not contribute excess 18 Oto concentrations can display an order of magnitude source prior to melting. variation at near-constant d 18 OŽ. Fig. 2 . This sug- At present it is not possible to distinguish whether gests that it may be invalid to correct oxygen isotope the correlation between Na8.0 and oxygen isotope ratios of lava suites by assuming a constant initial ratios in Lau Basin basalts results from addition of water concentration. an 18 O-rich fluid or from melting effects driven by In summary, the lack of quantitative isotope frac- the presence of such a fluid in variably depleted tionation data and complementary studies on petro- peridotite. Both situations are consistent with the genetically simpler regions prevents resolution be- highest fluid flux and most residual peridotite occur- tween three models for generation of oxygen isotope ring in ELSC lavas close to the Tonga Arc. If mantle variations in Lau Basin basalts. A weak, negative fertility does exert a strong control on oxygen iso- correlation between Mg No. and d 18 O may reflect tope ratios then similar correlations may be expected differentiation of basaltic melts towards 18 O-poor for lavas within and between individual segments of intermediate compositions similar to those postulated spreading centres worldwide. as precursors to the most evolved lavas. Alterna- A final point worthy of discussion is the lack of tively, recycling of a hydrous fluid from the sub- correlation between oxygen isotope ratios and water ducted slab may produce 18 O-rich lavas either by content in the Lau Basin glassesŽ. Fig. 2 . In the elevating d 18 O of the mantle source, or by enhanc- preceding paragraphs we have outlined three possi- ing orthopyroxene melting in residual peridotite. Fur- ble origins for d 18 O variation in Lau Basin basalt, ther investigation of oxygen isotope variation in each of which require introduction of a hydrous lavas from individual segments of spreading centres, component to the mantle source. From the arguments remote from subduction zone influences, may help outlined in this work and by Pearce et al.Ž. 1994 , a clarify the role of melting in generating 18 Or 16 O 18 positive correlation between d O and H2 O concen- heterogeneity in basaltic lavas worldwide. At that tration may be expected through either addition of an time the relative importance of source fertility and 18 O-rich hydrous fluid to peridotite or enhanced recycling of oxygen in subduction related magma- orthopyroxene melting resulting from high fluid tism may become clearer. fluxes. On the other hand, water is an incompatible element so its concentration in melts should decrease as the degree of melting increases. At crustal levels, 5. Conclusions fractional crystallisation will tend to increase both 18 d O and H2 O in residual melts while assimilation Ž.1 Magma chamber processes exert strong influ- of 18 O-poor oceanic crust would lead to a negative ences on oxygen isotope ratios in evolved Lau Basin correlation between these properties. Given the pos- lavas. Evolution of melts at the Central Lau Basin sibility of interplay between these processes we are propagating and dying rift tips initially involved not surprised at the absence of a single, significant assimilation of a modest amount of hydrothermally correlation between the oxygen isotope ratios and altered oceanic crust Ž.-15% and crystallisation of C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 191 a plagioclase-dominated assemblage. This produced Acknowledgements 18 O-depleted intermediate melts and may also have elevated LILE abundances. Enrichment of 18 Oin This research was funded by a grant to CGM melts commenced at the point of oxide saturation. from Fisons InstrumentsŽ. VG Isotech Ltd. Samples Oxygen and carbon isotope ratios of propagating and were provided by Julian Pearce, Richard Muhe and dying rift tip glasses are consistent with differentia- John Sinton. We are grateful to Sally Newman for tion and degassing in magma chambers between 2.5 sharing FTIR results with us prior to publication. and 5 km below the seafloorŽ Macpherson and Mat- Dave Alderton, Kevin Ansdell, Jim Hawkins, David tey, 1994; this study. . Hilton, Matthew Thirlwall and Derek Wyman are Ž.2 The Central Lau Basin data suggest that crustal thanked for discussion and comments on earlier contamination cannot be discounted merely on the manuscripts. Constructive reviews by Jon Davidson strength of values falling in the range of d 18 O values and two anonymous reviewers were of great help in typical of oceanic basalts. The nature of country focusing the scope of this paper. rocks must be carefully considered before such state- ments are made. Oceanic basalts may be susceptible to both elevation and reduction of 18 Or 16 O through References interaction with hydrothermally altered crust. This conclusion is supported by LF oxygen isotope evi- Aggrey, K.E., Muenow, D.W., Sinton, J.M., 1988. Volatile abun- dence from ocean islandsŽ Eiler et al., 1996; Thirl- dances in submarine glasses from the North Fiji and Lau back-arc basins. Geochim. Cosmochim. Acta 52, 2501±2506. wall et al., 1996. . 18 Anderson, A.T., Clayton, R.N., Mayeda, T.K., 1971. Oxygen Ž.3 Valu Fa Ridge glasses display a small O-en- isotope thermometry of mafic igneous rocks. J. Geol. 79, richment with chemical evolution consistent with 715±729. fractional crystallisation of a MORB-like phenocryst Bevis, M., Taylor, F.W., Schutz, B.E., Recy, J., Isacks, B.L., assemblage. A significant part of whole rock, bomb Helu, S., Singh, R., Kendrick, E., Stowell, J., Taylor, B., Calmant, S., 1995. Geodetic observations of very rapid con- fluorination data for lavas from Valu Fa Ridge is vergence and back-arc extension at the Tonga arc. Nature 374, 18 coincident with the new data. However, higher d O 249±251. values in some, conventionally analysed, lavas with Brocher, T.M., 1985. On the formation of the Viatiaz Trench similar values of Mg No. may reflect assimilation of lineament and Northern Fiji Basin. In: Brocher, T.M.Ž. Ed. , old arc crust or recycling of subducted sediment Geological Investigations of the Northern Melanesian Border- Ž.land. Earth Science Series, Vol. 3, Circum-Pacific Council for Lichtenstein and Loock, 1993 . Sampling sites in Energy and Mineral Resources, Houston, TX, pp. 13±34. the Northern Lau Basin are too disparate to allow Christie, D.M., Sinton, J.M., 1981. Evolution of abyssal lavas firm conclusions to be drawn regarding the role of along propagating segments of the Galapagos Spreading Cen- differentiation in oxygen isotope systematics. tre. Earth Planet. Sci. Lett. 56, 321±335. Ž.4 Basaltic glasses from throughout the Lau Basin Clayton, R.N., Mayeda, T.K., 1963. The use of bromine pentaflu- d 18 oride in the extraction of oxygen from oxides and silicates for display correlations between O and geochemical isotope analysis. Geochim. Cosmochim. Acta 27, 43±52. parameters that reflect source fertility. It is unclear Collier, J.S., Sinha, M.C., 1992. The Valu Fa Ridge: the pattern of whether this is the result of a single process or a volcanic activity at a back-arc spreading centre. Mar. Geol. more complex melting history requiring higher 104, 243±263. Ž18 O-rich. fluid fluxes to melt increasingly residual Condomines, M., Gronvold, K., Hooker, P.J., Muehlenbachs, K., O'Nions, R.K., Oskarsson, N., Oxburgh, E.R., 1983. Helium, peridotite. Insufficient control on oxygen isotope oxygen, strontium and neodymium isotopic relations in Ice- partitioning is presently available from experimental land volcanics. Earth Planet. Sci. Lett. 66, 125±136. studies or petrogenetically less complex suites of Dobson, P.F., O'Neil, J.R., 1987. Stable isotope compositions and basalts to make qualitative assessments of these rela- water contents of boninite series volcanic rocks from Chichi- tionships. In particular, further investigation of ex- jima, Bonin Islands, Japan. Earth Planet. Sci. Lett. 82, 75±86. Eiler, J.M., Farley, K.A., Valley, J.W., Hofmann, A.W., Stolper, perimental and natural glasses is required to clarify E.M., 1996. Oxygen isotope constraints on the sources of 18 the role of orthopyroxene melting in generating d O Hawaiian volcanism. Earth Planet. Sci. Lett. 144, 453±468. variations in cogenetic suites of lavas. Ellam, R.M., Harmon, R.S., 1990. Oxygen isotope constraints on 192 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194

the crustal contribution to the subduction-related magmatism indicators of crustal contamination. Annu. Rev. Earth Planet. of the Aeolian Islands, southern Italy. J. Volcanol. Geothermal Sci. 9, 311±344. Res. 44, 105±122. Jaques, A.L., Green, D.H., 1980. Anhydrous melting of peridotite Ewart, A., Hawkesworth, C.J., 1987. The Pleistocene±Recent at 0±15 kbar pressure and the genesis of tholeiitic basalts. Tonga±Kermadec Arc lavas: interpretation of new isotopic Contrib. Mineral. Petrol. 73, 287±310. and rare earth element data in terms of a depleted mantle Jenner, G.A., Cawood, P.A., Rautenschlein, M., White, W.M., source model. J. Petrol. 28, 495±530. 1987. Composition of back-arc basin volcanics, Valu Fa Ridge, Ewart, A., Bryan, W.B., Chappell, B.W., Rudnick, R.L., 1994. Lau Basin: Evidence for a slab-derived component in their Regional geochemistry of the Lau±Tonga arc and backarc mantle source. J. Volcanol. Geothermal Res. 32, 209±222. systems. Proc. ODP, Sci. Results 135, 557±564. Johnson, K.T.M., Sinton, J.M., 1990. Petrology, tectonic setting Farley, K.N., Newman, S., 1994. H O, CO , sulfur and f and the formation of back-arc basin basalts in the North Fiji 22 O2 systematics in basalts from the ancient and modern Lau backarc Basin. Geol. Jahrb. Hannover D92, 517±545. Basin, SW Pacific. EOS 75, 731. Karig, D.E., 1970. Ridges and basins of the Tonga±Kermadec Ferrara, G., Laurenzi, M.A., Taylor, H.P., Tonarini, S., Turi, B., island arc system. J. Geophys. Res. 75Ž. 2 , 239±254. 1985. Oxygen and strontium isotope studies of K-rich volcanic Kempton, P.D., Hawkesworth, C.J., Fowler, M., 1991. Geochem- rocks from the Alban Hills, Italy. Earth Planet. Sci. Lett. 75, istry and isotopic composition of gabbros from layer 3 of the 13±28. Indian Ocean crust, hole 735B. Proc. ODP, Sci. Results 118, Frenzel, G., Muhe, R., Stoffers, P., 1990. Petrology of the vol- 127±143. canic rocks from the Lau Basin, southwest Pacific. Geol. Klein, E.M., Langmuir, C.H., 1987. Global correlations of ocean Jahrb. Hannover D92, 395±479. ridge basalt chemistry with axial depth and crustal thickness. Friedrichsen, H., Hoerns, S., 1979. Oxygen and hydrogen isotope J. Geophys. Res. 92, 8089±8115. exchange reactions between sea water and oceanic basalts Kurz, M.D., Jenkins, W.J., 1981. The distribution of helium in from Legs 51 through 53. Init. Rep. DSDP, Vols. 51, 52, 53 oceanic basaltic glasses. Earth Planet. Sci. Lett. 53, 41±54. Part 2, pp. 1177±1182. Kyser, T.K., 1986. Stable isotope variations in the mantle. In: Gregory, R.T., Taylor, H.P., 1981. An oxygen isotope profile in a Valley, J.W., Taylor, H.P., O'Neil, J.R.Ž. Eds. , Stable Isotopes section of Cretaceous oceanic crust, Samail ophiolite, Oman: in High Temperature Processes. Rev. Miner. Min. Soc. Am. evidence for d 18 O buffering of the oceans by deep Ž.)5km 16, 141±164. seawater±hydrothermal circulation at mid-ocean ridges. J. Kyser, T.K., 1990. Stable isotopes in the continental lithospheric Geophys. Res. 86Ž. B4 , 2737±2755. mantle. In: Menzies, M.A.Ž. Ed. , The Continental Lithosphere. Harmon, R.S., Gerbe, M., 1992. The 1982±1983 eruption at Oxford University Press, pp. 127±156. Galunggung volcano, JavaŽ. Indonesia : oxygen isotope geo- Kyser, T.K., O'Neil, J.R., Carmichael, I.S.E., 1981. Oxygen chemistry of a chemically zoned magma chamber. J. Petrol. isotope thermometry of basaltic lavas and nodules. Contrib. 33, 585±609. Mineral. Petrol. 77, 11±23. Harmon, R.S., Hoefs, J., Wedepohl, K.H., 1987. Stable isotope Kyser, T.K., O'Neil, J.R., Carmichael, I.S.E., 1982. Genetic rela- Ž.O, H, S relationships in Tertiary basalts and their mantle tions among basic lavas and ultramafic nodules: evidence from xenoliths from the northern Hessian depression, W. Germany. oxygen isotope compositions. Contrib. Mineral. Petrol. 81, Contrib. Mineral. Petrol. 95, 350±369. 88±102. Hattori, K., Muehlenbachs, K., 1982. Oxygen isotope ratios of the Lecoyer, C., Reynard, B., 1996. High-temperature alteration of Icelandic crust. J. Geophys. Res. 87, 6559±6565. oceanic gabbros by seawaterŽ Hess Deep, Ocean Drilling Hawkesworth, C.J., Hergt, J.M., Ellam, R.M., McDermott, F., Program Leg 147. : Evidence from oxygen isotopes and ele- 1991. Element fluxes associated with subduction related mag- ment fluxes. J. Geophys. Res. 101, 15883±15897. matism. Philos. Trans. R. Soc. London A335, 393±405. Lichtenstein, U., Loock, G., 1993. Oxygen isotope compositions Hergt, J.M., Hawkesworth, C.J., McDermott, F., Pearce, J.A., of volcanic glasses from the Lau Basin. Terra Abstr. 5, 1991. A new estimate of the slab derived fluid in arc magmas: 547±548. implications for the Tonga±Lau Basin system. Terra Abstr. 3, Loock, G., McDonough, W.F., Goldstein, S.L., Hofmann, A.W., 45. 1990. Isotopic compositions of volcanic glasses from the Lau Hilton, D.R., Hammerschmidt, K., Loock, G., Friedrichsen, H., Basin. Mar. Min. 9, 235±245. 1993. Helium and argon isotope systematics of the Central Macpherson, C.G., 1994. New Analytical Approaches to the Lau Basin and Valu Fa Ridge: evidence of crustrmantle Oxygen and Carbon Stable Isotope Geochemistry of Some interactions in a back-arc basin. Geochim. Cosmochim. Acta Subduction-Related Lavas. Unpublished Ph.D. Thesis, Univ. 57, 2819±2841. of London, 206 pp. Ito, E., Clayton, R.N., 1983. Submarine metamorphism of gabbros Macpherson, C.G., Mattey, D.P., 1994. Carbon isotope variations

from the MidCaymen Rise: an oxygen isotope study. Geochim. of CO2 in Central Lau Basin basalts and ferrobasalts. Earth Cosmochim. Acta 47, 535±546. Planet. Sci. Lett. 121, 263±276. Ito, E., White, W.M., Gopel, C., 1987. The O, Sr, Nd and Pb Matsuhisa, Y., Matasubaya, O., Sakai, H., 1973. Oxygen isotope isotope geochemistry of MORB. Chem. Geol. 62, 157±176. variations in magmatic differentiation processes of the vol- James, D.E., 1981. The combined use of oxygen and radiogenic canic rocks in Japan. Contrib. Mineral. Petrol. 39, 277±288. C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194 193

Mattey, D.P., Macpherson, C.G., 1993. High-precision oxygen geochemistry of ocean sediments and shales. Geochim. Cos- isotope microanalysis of ferromagnesian minerals by laser mochim. Acta 34, 43±63. fluorination. Chem. Geol.Ž. Isot. Geosci. Sect. 105, 305±318. Sharp, Z.D., 1990. A laser based microanalytical method for the Mattey, D.P., Muir, I.D., 1980. Geochemistry and mineralogy of in situ determination of oxygen isotope ratios of silicates and basalts from the Galapagos Spreading Centre, Deep Sea oxides. Geochim. Cosmochim. Acta 54, 1353±1357. Drilling Project Leg 54. Init. Rep. DSDP 54, 755±771. Sheppard, S.M.F., Harris, C., 1985. Hydrogen and oxygen isotope Mattey, D.P., Lowry, D., Macpherson, C.G., 1994. Oxygen iso- geochemistry of Ascension Island lavas and granites: variation tope composition of the mantle. Earth Planet. Sci. Lett. 128, with fractional crystallisation and interaction with sea water. 231±241. Contrib. Mineral. Petrol. 91, 74±81. Muehlenbachs, K., 1986. Alteration of the oceanic crust and the Sinton, J.M., Price, R.C., Johnson, K.T.M., Staudigel, H., Zindler, 18 O history of seawater. In: Valley, J.W., Taylor, H.P., O'Neil, A., 1993. Petrology and geochemistry of submarine lavas from J.R.Ž. Eds. , Stable Isotopes in High Temperature Processes. the Lau and North Fiji back-arc basins. In: Kroenke, L.W., Rev. Miner. Min. Soc. Am. 16, 425±444. Eade, J.V.Ž. Eds. , Basin Formation, Ridge Crest Processes and Muehlenbachs, K., Byerly, G., 1982. 18 O-enrichment of silicic Metallogenesis in the North Fiji Basin. Circum-Pacific Coun- magmas caused by crystal fractionation at the Galapagos cil for Energy and Mineral Resources, Houston, TX, Earth spreading centre. Contrib. Mineral. Petrol. 79, 76±79. Science Series, Vol. 15, Springer Verlag, New York, pp. Muehlenbachs, K., Clayton, R.N., 1972. Oxygen isotope studies 119±135. of fresh and weathered submarine basalts. Can. J. Earth Sci. 9, Spulber, S.D., Rutherford, M.J., 1983. The origin of rhyolite and 172±184. plagiogranite in oceanic crust: an experimental study. J. Petrol. Muehlenbachs, K., Kushiro, I., 1974. Oxygen isotope exchange 24, 1±25.

and equilibrium of silicates with CO22 or O . Carnegie Inst. Stakes, D., Mevel, C., Cannat, M., Chaput, T., 1991. Metamor- Washington, Yearb. 73, 232±236. phic stratigraphy of hole 735B. Proc. ODP, Sci. Results 118, Muehlenbachs, K., Anderson, A.T., Sigvaldson, G.E., 1974. Low- 153±180. 18 O basalts from Iceland. Geochim. Cosmochim. Acta 38, Sunkel, G., 1990. Origin of petrological and geochemical varia- 577±588. tions of Lau Basin lavasŽ. SW Pacific . Mar. Min. 9, 205±234. O'Hara, M.J., Matthews, R.E., 1981. Geochemical evolution in an Tatsumi, Y., Hamilton, D.L., Nesbitt, R.W., 1986. Chemical advancing, periodically replenished, periodically tapped, con- characteristics of fluid phase released from a subducted litho- tinuously fractionated magma chamber. J. Geol. Soc. London sphere and origin of arc magmas: evidence from high-pressure 138, 237±278. experiments and natural rocks. J. Volcanol. Geothermal Res. Parson, L.M., Hawkins, J.W., 1994. Two stage ridge propagation 29, 293±309. and the geological history of the Lau backarc basin. Proc. Taylor, H.P., 1968. The oxygen isotope geochemistry of igneous ODP, Sci. Results 135, 819±828. rocks. Contrib. Mineral. Petrol. 19, 1±71. Parson, L.M., Pearce, J.A., Murton, B.J., Hodkinson, R.A., 1990. Taylor, H.P., 1987. Comparison of hydrothermal systems in lay- Role of ridge jumps and ridge propagation in the tectonic ered gabbros and granites, and the origin of low-18 O magmas. evolution of the Lau back-arc basin, southwest Pacific. Geol- In: Mysen, B.O.Ž. Ed. , Magmatic Processes: Physiochemical ogy 18, 470±473. Principles. Geochem. Soc. Spec. Publ. 1, 337±357. Parson, L.M., Rothwell, R.G., MacLeod, C.J., 1994. Tectonics Taylor, H.P., Sheppard, S.M.F., 1986. Igneous rocks, I. Processes and sedimentation in the Lau BasinŽ. Southwest Pacific . Proc. of isotopic fractionation and isotope systematics. In: Valley, ODP, Sci. Results 135, 9±21. J.W., Taylor, H.P., O'Neil, J.R.Ž. Eds. , Stable Isotopes in High Pearce, J.A., Ernewein, M., Bloomer, S.H., Parson, L.M., Murton, Temperature Processes. Rev. Miner. Min. Soc. Am. 16, 141± B.J., Johnson, L.E., 1994. Geochemistry of Lau Basin vol- 164. canic rocks: influence of ridge segmentation and arc proxim- Thirlwall, M.F., Jenkins, C., Vroon, P.Z., Mattey, D.P., 1996. ity. In: Smellie, J.L.Ž. Ed. , Volcanism Associated with Exten- Crustal interaction during construction of oceanic islands: sion at Consuming Plate Margins. Geol. Soc. Spec. Publ. 81, Pb±Sr±Nd±O isotope stratigraphy of the shield basalts of 53±75. Gran Canaria. Chem. Geol. 135, 233±262. Perfit, M.R., Fornari, D.J., Malahoff, A., Embley, R.W., 1983. Vallier, T.L., Jenner, G.A., Frey, F.A., Gill, J.B., Davis, A.S., Geochemical studies of abyssal lavas recovered by DSDV Volpe, A.M., Hawkins, J.W., Morris, J.D., Cawood, P.A., Alvin from Eastern Galapagos Rift, Inca Transform and Morton, J.L., Scholl, D.W., Rautenschlein, M., White, W.M., Ecuador Rift, Part 3. Trace element abundances and petrogen- Williams, R.W., Stevenson, A.J., White, L.D., 1991. Subalka- esis. J. Geophys. Res. 88, 10551±10572. line andesite from Valu Fa Ridge, a back-arc spreading centre Pineau, F., Javoy, M., Hawkins, J.W., Craig, H., 1976. Oxygen in southern Lau Basin: petrogenesis, comparative chemistry isotope variations in marginal basin and ocean-ridge basalts. and tectonic implications. Chem. Geol. 91, 227±256. Earth Planet. Sci. Lett. 28, 299±307. Woodhead, J.D., Harmon, R.S., Fraser, D.G., 1987. O, S, Sr and Poreda, R., Craig, H., 1992. He and Sr isotopes in the Lau Basin Pb isotope variations in volcanic rocks from the northern mantle: depleted and primitive mantle components. Earth Mariana Islands: implications for crustal recycling in intra-oc- Planet. Sci. Lett. 113, 487±493. eanic arcs. Earth Planet. Sci. Lett. 83, 39±52. Savin, S.M., Epstein, S., 1970. The oxygen and hydrogen isotope Woodhead, J.D., Eggins, S., Gamble, J.A., 1993. High field 194 C.G. Macpherson, D.P. MatteyrChemical Geology 144() 1998 177±194

strength and transition element systematics in island arc and You, C.-F., Castillo, P.R., Gieskes, J.M., Chan, L.H., Spivak, back-arc basin basalts: evidence for multi-phase melt extrac- A.J., 1996. Trace element behavior in hydrothermal systems: tions and a depleted mantle wedge. Earth Planet. Sci. Lett. implications for fluid processes at shallow depths in subduc- 114, 491±504. tion zones. Earth Planet. Sci. Lett. 140, 41±52.