| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology Volume 42 Number 8 Pages 1449-1470 2001
© Oxford University Press 2001
U-series Isotope Data on Lau Basin Glasses: the Role of Subduction-related Fluids during Melt Generation in Back-arc Basins
1DEPARTMENT OF EARTH SCIENCES, THE OPEN UNIVERSITY, WALTON HALL, MILTON KEYNES MK7 6AA, UK
2SCHOOL OF EARTH SCIENCES, THE UNIVERSITY OF MELBOURNE, MELBOURNE, VIC. 3010, AUSTRALIA
3DEPARTMENT OF EARTH SCIENCES, CARDIFF UNIVERSITY, MAIN COLLEGE, PARK PLACE, P.O. BOX 914, CARDIFF CF1 3YE, UK
Received April 28, 2000; Revised typescript accepted February 5, 2001
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONTHE LAU BASIN REGION...RESULTSDISCUSSIONSUMMARYREFERENCES |
|---|
New thermal ionization mass spectrometry U-series disequilibrium data are presented for 24 basaltic to dacitic glasses from active spreading centres in the back-arc Lau Basin (SW Pacific), together with additional inductively coupled plasma mass spectrometry trace element analyses and Sr–Nd–Pb isotope data. Valu Fa Ridge samples, adjacent to the arc front, have high U/Th and (230Th/238U) <1, implying a recent (<<350 ka) addition of a U-rich slab-derived fluid. The Valu Fa data can be combined with existing 230Th–238U data for the Central Tonga arc to infer a fluid addition event at
50 ka. The similar sources and time scales for fluid transfer beneath the Valu Fa Ridge and beneath the arc itself suggest that the Valu Fa Ridge is propagating into the arc-front region. Central Lau Basin samples, further behind the arc, have lower U/Th and (230Th/238U) 
1, similar to typical mid-ocean ridge basalts (MORB). Within the Central Lau Basin, a water-rich subduction component is seen only in samples closest to the arc, and this fluid does not have the high-U/Th composition of the fluid at Valu Fa. Melt generation in the Central Lau Basin appears to be dominated by normal ridge-type processes, but the relatively low (230Th/238U) for these shallow ridges compared with global MORB could be a consequence of increased melt productivity as a result of the elevated water contents. The transition from 238U to 230Th excesses within the back-arc basin is not a smooth function of decreasing addition of a U-rich fluid moving away from the arc front, but also reflects the complex dynamics between two major mantle domains within the mantle wedge (‘Pacific’ beneath Valu Fa Ridge, ‘Indian’ beneath the Central Lau Basin). KEY WORDS: back-arc basins; U-series disequilibrium; fluids; subduction zones; melting
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONTHE LAU BASIN REGION...RESULTSDISCUSSIONSUMMARYREFERENCES |
|---|
A perplexing aspect of back-arc magmatism is the exact role of the adjacent subduction zone in influencing the nature of melt generation. It is not clear to what extent the compositions of magmas erupted at back-arc spreading centres are dominated by processes common to typical mid-ocean ridges, and what additional complexities are brought by the supra-subduction zone setting. The wide compositional range observed in back-arc magmas, from mid-ocean ridge basalt (MORB)-like to arc-like, is generally attributed to progressive re-enrichment of a variably depleted mantle wedge source by slab-derived fluids (e.g. Saunders & Tarney, 1979
; Pearce et al., 1984; Sinton & Fryer, 1987; Stolper & Newman, 1994; Pearce et al., 1995). Stolper & Newman (1994) further argued that the fluxing of the mantle source by H2O-rich fluids has an important control on the degree of melting in back-arc magmas. Important questions remain concerning the nature and transport of the slab-derived fluid, in particular about how quickly and how far it pervades the mantle wedge into the back-arc region, and how much it equilibrates with the mantle wedge during transport. It is also not clear how the processes of melt generation in back-arcs compare with those at other spreading centres such as mid-ocean ridges. U-series isotope data offer a unique opportunity to address these questions, as 238U–230Th isotope disequilibria record the effects of recent (<350 ka) U–Th fractionation and are sensitive to recent fluid inputs, mantle mineralogy, and the dynamics of mantle melting.
The processes that control 238U–230Th disequilibria in magmas from island-arc environments appear to be different from those in mid-ocean ridge settings (e.g. Newman et al., 1984; Gill & Williams, 1990; McDermott & Hawkesworth, 1991; Hawkesworth et al., 1997a). MORB magmas predominantly erupt with (230Th/238U) >1, and the excess 230Th is generally attributed to in-growth during partial melting initiated within the garnet stability field (e.g. Goldstein et al., 1991; Beattie, 1993; LaTourette et al., 1993; Lundstrom et al., 1995, 1998b; Bourdon et al., 1996b; Elliott, 1997). For island-arc magmas, 238U–230Th disequilibria are more common in rocks depleted in highly incompatible elements, and these samples mainly show (230Th/238U) <1. Such behaviour is consistent with recent addition of a U-rich slab-derived fluid to the mantle wedge source (e.g. Newman et al., 1984; Gill & Williams, 1990; McDermott & Hawkesworth, 1991; Condomines & Sigmarsson, 1993; Elliott et al., 1997; Hawkesworth et al., 1997a; Turner et al., 1997). It is not clear which of these two competing effects (i.e. recent fluid input vs source or melting processes) will dominate to control the sense of 238U–230Th disequilibrium in melts erupted at back-arc spreading centres. Presumably this will vary on both a local and a global scale, reflecting a variable influence from the subduction zone. There will be a complex interplay between several factors: the composition, amount, and timing of slab-fluid addition to the back-arc mantle, the fertility of the back-arc mantle, and the dynamics of mantle upwelling beneath the spreading centre.
The central Lau Basin is an ideal back-arc setting in which to investigate this problem because the active spreading centres are at varying distances from the arc front and trench (Fig. 1). Furthermore, geochemical studies have suggested that the added subduction component in the mantle source changes progressively with arc proximity (e.g. Pearce et al., 1995). In this paper, we present the first detailed study of U-series disequilibria in back-arc magmas, with new analyses on dredged glasses from the active spreading centres of the Lau Basin, to assess the timing and influence of variable addition of a volatile-rich subduction component on melting processes in the back-arc.
|
, recent submarine arc-front volcanoes. Grey shading indicates areas with water depths <2 km. The Kermadec arc is the lateral continuation of the Tonga arc, south of 23°S. •, location of drill sites from the ODP Leg 135. Horizontally striped shaded regions represent new crust formed by true sea-floor spreading on the ELSC and CLSC. The two boxes indicate the locations of the studied samples. |
THE LAU BASIN REGION AND SAMPLE DETAILS |
|---|
|
TOPABSTRACTINTRODUCTIONTHE LAU BASIN REGION...RESULTSDISCUSSIONSUMMARYREFERENCES |
|---|
The Lau Basin is a triangular-shaped, actively spreading back-arc basin, situated behind the Tonga arc in the SW Pacific (Fig. 1). Two comprehensive up-to-date reviews of the geology of the Lau Basin have been published recently that incorporate new results from extensive geophysical surveys, dredged material and Ocean Drilling Program (ODP) drilling (Hawkins, 1995a
, 1995b), and the following summary is largely based on these reviews. Samples selected for this study come from the central and southern parts of the Lau Basin, which have a relatively simple tectonic arrangement. Here, the basin initially opened solely by crustal extension and rifting. The first new oceanic crust formed at 5·5–5·0 Ma as sea-floor spreading began on a southward-propagating rift, the Eastern Lau Spreading Centre (ELSC). A younger rift, the Central Lau Spreading Centre (CLSC), was also initiated at the Peggy Ridge transform fault but further from the arc at 1·5–1·2 Ma, and it is propagating southwards at the expense of the ELSC. The two rifts overlap at 19·3°S and there is a small ‘relay’ spreading segment, the Intermediate Lau Spreading Centre (ILSC), between them. Thus, the active spreading centres are offset and become closer to the active Tonga arc-front volcanoes towards the south: the southern end of the ELSC, also known as the Valu Fa Ridge (VFR), is only 40 km from Ata volcano. Sea-floor spreading rates within the Lau Basin increase towards the north, consistent with the inverted triangular shape of the basin, and estimates derived from interpretation of magnetic lineations are 65 mm/yr at 21°S and 90 mm/yr at 18°S (full spreading rate: Taylor et al., 1996). Spreading rate estimates based on global positioning satellite (GPS) geodetic measurements (Bevis et al., 1995) are about twice these values, but this apparent discrepancy has recently been resolved in favour of the lower values by the recognition of a Niuafo’ou microplate between the Australia and Tonga plates north of 19°20'S (Zellmer & Taylor, 1999).
The northern part of the basin (north of 17°N) was not included in the present study for two reasons: (1) it has a complicated tectonic configuration (Fig. 1); (2) many of the lavas in this region, including the nearby arc-front islands (Tafahi and Niuatoputapu), show evidence for the presence of additional compositional components derived from the Samoan plume to the NE and also mobilized from volcaniclastic sediments on the subducting slab associated with the Louisville seamounts (Volpe et al., 1988; Hawkins, 1995a; Regelous et al., 1997; Turner & Hawkesworth, 1997, 1998; Turner et al., 1997; Wendt et al., 1997; Ewart et al., 1998). These enriched components are not evident in magmas erupted further south (i.e. the Central Lau Basin, Valu Fa Ridge and the main Tonga arc), or in the northern rear-arc island of Niuafo’ou.
A distinctive feature of the composition of the lavas from the Lau spreading centres is the presence, in addition to basalts, of highly fractionated samples (ferrobasalts, andesites: e.g. Vallier et al., 1991; Pearce et al., 1995), which we have also analysed. These evolved rocks are found near the propagating rift tips of both the ELSC–VFR and the CLSC, and similar rocks are often found on propagating segments of the global mid-ocean ridge system (Christie & Sinton, 1981).
The selection of samples for U–Th–Ra analysis was based both on the availability of suitable quantities of fresh glass, hopefully of ‘zero-age’ (<10 ka), dredged from the active spreading axes, and on obtaining a reasonable geographical coverage from the different spreading centres. Samples from the central part of the Lau Basin (CLSC, ILSC and ELSC) came from the R.S.S. Charles Darwin (CD33) cruise that dredged the neovolcanic zone (identified from GLORIA images) between 18°50' and 20°35'S (Pearce et al., 1995), with a few additional samples from the northern CLSC (18°34'–18°44'S) from the R.V. Sonne 1987 (SO48) cruise (Sunkel, 1990). Samples from the Valu Fa Ridge came from dredged material from the R.V. Lee 1984 cruise near 22°20'S (Jenner et al., 1987; Vallier et al., 1991) and from the R.V. Sonne 1984 (SO35) and 1987 (SO48) cruises (Boespflug et al., 1990; Sunkel, 1990; Bach & Niedermann, 1998; Bach et al., 1998). The locations of the studied samples (latitude, longitude, depth) are given in Table 1. Additional trace element and radiogenic (Sr–Nd–Pb) isotope analyses were carried out (Tables 1 and 2) to complete the existing compositional data available on these samples (Jenner et al., 1987; Loock et al., 1990; Sunkel, 1990; Vallier et al., 1991; Macpherson & Mattey, 1994, 1998; Pearce et al., 1995; Bach & Niedermann, 1998; Bach et al., 1998).
|
|
The samples were glass chips, hand-picked under a binocular microscope to avoid fragments with any visible Mn-oxide or alteration coatings, that were then leached using a H2O2–HCl mixture before digestion (Bourdon et al., 1996a). The procedures for sample digestion and chemical extraction of U, Th and Ra as followed at the Open University have been given by Turner et al. (1997, 2000). All isotopic measurements were performed on a Finnigan MAT262 mass spectrometer equipped with a retarding potential quadrupole (RPQ-II) to provide the high abundance sensitivity capability necessary for high-precision measurements of 230Th. This system and the mass spectrometric procedures for U and Th analysis have been described by van Calsteren & Schwieters (1995). The external reproducibility on 230Th/232Th is 1·0% (2 SD), based on repeated measurements of an ‘in-house’ Th standard solution (Th‘U’std) that has 230Th/232Th of about 6 x 10-6 (van Calsteren & Schwieters, 1995), similar to that of the measured Lau Basin glasses. The procedure used for measuring Ra is similar to that outlined by Lundstrom et al. (1998a), with particular care taken to monitor organic interferences (Turner et al., 2000).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONTHE LAU BASIN REGION...RESULTSDISCUSSIONSUMMARYREFERENCES |
|---|
U, Th and Ra abundances and isotope ratios for 24 Lau Basin samples are reported in Table 3. All samples were dredged from the inferred axial spreading ridges, except for 114KD and 61GC, which are from near-axis seamounts on the Valu Fa Ridge (Sunkel, 1990
). The relatively high spreading rates in the Lau Basin (65–90 mm/yr: Taylor et al., 1996) make it likely that the on-axis samples are younger than 10 ka such that no correction is required for post-eruptive decay of 230Th. This was checked for most samples by using the 226Ra–230Th decay scheme. The short half-life of 226Ra (1600 yr) means that any sample with 226Ra–230Th disequilibrium must be <8 ka. With a few exceptions, the samples have (234U/238U) values within 1% of secular equilibrium, indicating no post-eruption seawater alteration. The U–Th isotope results are presented on a (230Th/232Th) vs (238U/232Th) equiline diagram in Fig. 2. There is a clear contrast between the samples from the Valu Fa Ridge and those from the Central Lau Basin (ELSC, ILSC and CLSC) to the north, and so they will be described separately. The Valu Fa Ridge samples all have relatively high (238U/232Th) (1·28–1·37) and ‘arc-like’ (230Th/238U) <1, whereas the Central Lau Basin samples all have lower (238U/232Th) (0·98–1·15) and ‘MORB-like’ (230Th/238U) 1.
|
|
Figure 2 also includes published U–Th analyses from the Tonga arc region for comparison (Regelous et al., 1997; Turner et al., 1997). Samples from the main Tonga arc (18°–21°S: Fig. 1) have very high (238U/232Th) of 2·2–3·2, and significant 238U excesses with (230Th/238U) of 0·55–0·80. Data from Ata island also plot to the right of the equiline, with (230Th/238U) of 0·81–0·95, and they have slightly higher (238U/232Th) than the adjacent Valu Fa Ridge (1·38–1·63 vs 1·28–1·37). The eruption ages for the Ata lavas are poorly constrained, and so the fact that samples with (238U/232Th) similar to the Valu Fa lavas show less 230Th–238U disequilibrium could simply be due to post-eruption decay (Ata lavas have 226Ra–230Th in equilibrium, implying an age >>8 ka: Turner et al., 2000). The northern Tonga arc islands show less U enrichment than the main arc, with (238U/232Th) of 1·3–2·0, and (230Th/238U) of 0·69–0·95. The majority of samples analysed from the northern back-arc island of Niuafo’ou show 230Th excesses with (230Th/238U) of 0·94–1·41, at similar (238U/232Th) of 0·84–1·15 to the Central Lau Basin rift magmas (Regelous et al., 1997; Turner et al., 1997).
Valu Fa Ridge
Samples from the Valu Fa spreading centre have a relatively limited variation in both (238U/232Th) (1·34–1·37) and (230Th/232Th) (1·17–1·21), with the exception of the northernmost sample (55GC), which has a lower (238U/232Th) of 1·28 and (230Th/232Th) of 1·13. Sample 61GC, from an off-axis seamount, has similar (238U/232Th) to the adjacent spreading centre, but with lower (230Th/232Th) of 1·15. All samples show significant excess 238U with (230Th/238U) of 0·85–0·90. The L-1 and L-2 samples are from two closely spaced dredge hauls, and major and trace element data indicate that the samples within each dredge are compositionally very similar, perhaps fragments of the same flow (Vallier et al., 1991). This similarity is also apparent from the U-series results: (238U/232Th) and (230Th/232Th) are constant, within analytical error, at each dredge site, with the L-2 samples having slightly higher (238U/232Th) than those at L-1 (1·37 vs 1·35). These results provide a good estimate of the reproducibility of the data. The higher Th and U contents of the L-1 samples relative to those from L-2 (401 ppb vs 350 ppb, and 178 ppb vs 158 ppb, respectively) are consistent with their more fractionated major element compositions (see Table 1; Vallier et al., 1991). Most Valu Fa samples show significant 226Ra–230Th disequilibria (up to 100% excess 226Ra), confirming their relatively young eruption ages (<8 ka). The presence of young lava flows, many with evolved compositions, is consistent with the geophysical observations of a shallow magma chamber beneath the Valu Fa Ridge (Collier & Sinha, 1990).
Three of the samples (L-1-5, L-2-5, L-2-20) have previously been analysed for U-series nuclides by 
-counting methods (Vallier et al., 1991). This earlier study similarly found significant excess 238U and excess 226Ra [(230Th/238U) 0·88–0·92, (226Ra/230Th) 1·4–2·6], and further showed that the samples were more than 100 years old because 210Pb and 226Ra were in secular equilibrium. However, there are significant discrepancies between these data and the new mass spectrometric data on the same samples, which are clearly illustrated in Fig. 2: the -counting data show much higher (238U/232Th) and (230Th/232Th), 1·5 vs 1·2 and 1·7 vs 1·35, respectively. The U and Th concentrations determined by -counting are also not consistent either with the new U and Th concentrations measured both by isotope dilution and by inductively coupled plasma mass spectrometry (ICP-MS) or with other trace element data on these samples (Tables 1 and 2).
Central Lau Basin (ELSC, ILSC and CLSC)
Samples from the Central Lau Basin have (238U/232Th) between 0·98 and 1·15, which is significantly lower than in Valu Fa Ridge samples (Fig. 2). All samples show either 230Th–238U equilibrium or 230Th excesses of up to 11%, in marked contrast to the Valu Fa Ridge samples from further south. The highest (230Th/238U) values are found in two samples from the ELSC and ILSC. Five samples have measured 230Th–238U equilibrium and although only two of these samples have been analysed for 226Ra, both show significant 226Ra–230Th disequilibrium, indicating eruption ages of <8 ka, and therefore that the 230Th–238U equilibrium is not due to post-eruption 230Th decay. One of the evolved andesitic samples (13-2) shows secular equilibrium for 226Ra–230Th. The maximum observed (226Ra/230Th) disequilibrium (2·0) is similar to that found at the Valu Fa Ridge.
Regional variation of 230Th–238U disequilibrium
Within the Tonga arc–back-arc system, there is a systematic difference in the sense of 230Th–238U disequilibrium away from the trench. This regional variation is summarized in Fig. 3, with (230Th/238U) plotted against distance behind the trench. It is clear, however, that there is not a progressive trend of 230Th–238U disequilibrium with distance, but rather an abrupt change in behaviour. Arc and back-arc lavas erupted <250 km from the trench have 238U excesses, whereas those from >250 km from the trench have 230Th excesses or 230Th–238U equilibrium. It should be noted that even though the CLSC is 100 km further from the trench than the ELSC, samples from both spreading centres have broadly similar extents of 230Th–238U disequilibrium, although the highest values are found closer to the arc in samples from the ELSC and ILSC.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONTHE LAU BASIN REGION...RESULTSDISCUSSIONSUMMARYREFERENCES |
|---|
Regional overview
A particular feature of the Tonga–Kermadec–Lau Basin system is the regional variation in the rates of back-arc spreading, in the compositions of mantle segments above the subduction zones, and in the components mobilized from the subducted oceanic crust. These provide exceptional opportunities to isolate contributions from well-characterized source materials, and to evaluate how the melt generation processes vary with distance from the trench, the time scales of the transfer of material from the subducted ocean crust, and whether there is any change in the hydrous fluid component with distance into the mantle wedge.
Identification of mantle wedge and slab-derived components in the Tonga region
Previous studies have established that the Tonga–Kermadec–Lau Basin system includes a number of compositionally distinct components and these need to be identified first. Most discussions of different components in the Tonga–Kermadec and Lau Basin rocks have relied on variations in Pb isotopes, and these are summarized in Fig. 4. Hergt & Hawkesworth (1994) documented two distinct trends in the Pb isotope data from rocks drilled at six sites during ODP Leg 135 in the older parts of the Lau Basin (Fig. 1). One trend includes samples from Sites 834 and 839 and projects back into the field for Pacific Ocean MORB (Fig. 4a), whereas the other trend includes samples from Sites 836 and 837 and projects into the field for Indian Ocean MORB. This was interpreted in terms of southward displacement of mantle similar to the source of Pacific MORB by Indian Ocean MORB mantle as a result of slab rollback, and accompanied by the southward migration of the propagating ridge tip into the extended crust (Hergt & Hawkesworth, 1994). As predicted by that model, the new data on the younger Central Lau Basin rocks have unradiogenic Pb isotope ratios that plot well within the field for Indian MORB (see also Loock et al., 1990), and the data for Ata plot in the array for Pacific MORB (Fig. 4a). The Valu Fa analyses plot close to where the two trends intercept, and so the significance of their Pb isotope ratios in the context of the material in the mantle wedge is more ambiguous. Bach et al. (1998) argued that instead of having two compositionally distinct mantle domains within the mantle wedge, the Pacific MORB isotopic signature can be explained simply by addition of fluids released by dehydration of the subducting Pacific plate. In the case of the Valu Fa Ridge, this fluid addition might swamp any contribution from the wedge, making it difficult to identify the isotopic provenance of the mantle wedge beneath this region.
|
By looking at the Pb isotope data in conjunction with trace element data, it is possible to distinguish between these two models, as well as to identify different components from the subducted slab. Ce/Pb is a particularly useful ratio in this regard. Ce and Pb have broadly similar incompatibility during mantle melting, and it has been argued that the mantle has a broadly constant Ce/Pb value of 25 ± 5 (Hofmann et al., 1986
; Sims & DePaolo, 1997). In contrast, Pb is highly mobile in fluids relative to Ce (e.g. Miller et al., 1994). Arc rocks have relatively high Pb contents (and hence low Ce/Pb ratios), which are attributed to the preferential addition of Pb during subduction (e.g. Miller et al., 1994). The Pb isotope ratios of most arc rocks are therefore dominated by the contributions from subducted sediment and altered basalts, with little or no detectable contribution from the mantle wedge. Turner et al. (1997) used high Ba/Th ratios as a fingerprint of the fluid component in the Tonga arc lavas, and inferred that its 206Pb/204Pb ratio was 18·4, in contrast to the subducted sediments (low Ba/Th), which had 206Pb/204Pb >18·8. The variations in Ce/Pb ratios with 206Pb/204Pb are summarized in Fig. 4b, and it is important to note that binary mixing is linear in such a plot. The Central Lau Basin rocks scatter down to lower Ce/Pb ratios, consistent with a minor contribution from a Pb-rich fluid with 206Pb/204Pb = 18·4. The main array of the Tonga–Kermadec arc rocks reflects varying contributions from a similar fluid and a sediment component with 206Pb/204Pb >18·8. Evidence for the presence of mantle wedge material with a ‘Pacific’ MORB Pb isotope composition comes from some of the lavas from ODP Sites 834 and 839, which have high Ce/Pb values typical of MORB mantle unaffected by subduction fluids. The displacement of the Valu Fa Ridge rocks to higher Ce/Pb ratios relative to the main arc suggests less of a contribution from the subducted slab. It is consistent with the notion of the mantle wedge here having a ‘Pacific’ MORB composition similar to that tapped by lavas from ODP Sites 834 and 839, as opposed to the ‘Indian’ MORB mantle beneath the Central Lau Basin.
Fluid influences on melting and melt composition
A central issue is the amount of slab-released fluid present in the mantle where melting takes place, and whether the amount of fluid, particularly water, varies with the degree of melting and with distance from the trench. There are a few water analyses on Lau Basin rocks (Jambon & Zimmermann, 1990; Danushevsky et al., 1993; Kamenetsky et al., 1997; Bach et al., 1998; S. Newman, unpublished data, 1999), and these data indicate that the lavas are in general derived from mantle sources enriched in water relative to the average N-MORB source (e.g. Michael, 1995). In Fig. 5, enrichment in H2O relative to Ce (an element of similar incompatibility to water during mantle melting: Michael, 1995) is used to evaluate the variations in U/Th ratios. Ba/Th ratios are widely used to indicate the relative contribution of subduction-related fluids in arc rocks, but the plots of (238U/232Th) vs H2O/Ce and Ba/Th (Fig. 5) illustrate that two components may be identified in these rocks, in addition to MORB-type mantle in the wedge. One trend is to high (238U/232Th) with increasing Ba/Th and H2O/Ce (although the data for the latter are sparse) as defined by samples from Ata, which is the arc volcano furthest from the trench, and from Valu Fa, which is the back-arc section closest to the arc (Fig. 1). This is the generally accepted shift to high U/Th ratios attributed to the recent (<100 ky) introduction of fluids from the subducted crust. More surprisingly, the ELSC and CLSC rocks, which are further from the trench than those of the Valu Fa Ridge, define a second trend of slightly decreasing (238U/232Th) with increasing H2O/Ce accompanied by slight increases in Ba/Th (Fig. 5). They have higher H2O/Ce than average N-MORB (Michael, 1995), consistent with their more vesicular textures, and yet they are displaced to excess 230Th rather than excess 238U (Fig. 2). Thus, the water-rich component present in the mantle wedge in this region appears to be compositionally different from that found closer to the arc front. This could result from: (1) a similar slab-fluid, which had been compositionally modified through greater interaction with mantle wedge material (e.g. Stern et al., 1991; Hawkesworth et al., 1993; Stolper & Newman, 1994); (2) fluid dominantly derived from a different source, such as subducted sediments, which is inferred to have a longer residence time in the mantle wedge (e.g. Elliott et al., 1997; Turner & Hawkesworth, 1997; Class et al., 2000); and/or (3) it might be an inherent feature of the slightly trace element enriched Indian MORB source. These possibilities will be discussed in a later section.
|
, Seamount samples adjacent to the Valu Fa Ridge (Kamenetsky et al., 1997
Stolper & Newman (1994) reported a good positive correlation between H2O and U contents in glasses from the Mariana Trough, and a similar correlation is observed between U and H2O for the Lau Basin samples for which water data are available (Fig. 6a). This indicates that, to a first approximation, U can be used as a proxy for H2O, although we note that some H2O is likely to have been lost by degassing particularly in the evolved Valu Fa Ridge lavas and and that subduction fluids might have somewhat variable U/H2O values. Yb is plotted against MgO for the rocks of the Lau Basin and the Tonga arc in Fig. 6b, to emphasize that most suites include rocks with 8% MgO and that there are significant differences in the Yb contents of the less evolved rocks. The plot of U vs Yb (Fig. 6c) shows a series of steep arrays for the different rock suites, primarily reflecting within-suite differentiation. However, the low ends of these arrays tend to be in rocks with 8% MgO (Fig. 6b), and in those least evolved rocks there is a progressive shift to decreasing Yb contents with increasing U. As U reflects the fluid contribution, and may be a rough proxy for H2O in these rocks, and Yb varies with the degree of melting (see also Fig. 8, below), the arrays in Fig. 6c strongly suggest that the degree of melting decreases with decreasing fluid (water) contribution from the subducted slab with distance away from the trench. In general, however, within arc rocks, the relative contribution from the fluid component tends to be high (e.g. high Ba/Th, U/Th) in rocks derived from more depleted sources (e.g. high Al/Ti, low Na/Ta) (e.g. Ewart & Hawkesworth, 1987; Hawkesworth et al., 1991, 1997b; Woodhead et al., 1993; Turner et al., 1997), but there is less evidence that the degree of melting varies solely with the amount of added fluid. Global inter-arc correlations of major elements with lithospheric thickness (Plank & Langmuir, 1988), theoretical modelling of melting of hydrated peridotite (Hirschmann et al., 1999), differences in fO2 between primitive arc lavas and arc peridotites (Parkinson & Arculus, 1999), and the existence of primitive low-H2O basalts in some arc-front volcanoes (Sisson & Bronto, 1998) all suggest a component of decompression melting beneath arcs. Thus, the degree of melting beneath arcs is likely to be largely determined by a combination of volatile addition and decompression melting (e.g. Pearce & Parkinson, 1993; Pearce & Peate, 1995).
|
The Valu Fa Ridge
Fluid transfer rates at the Valu Fa Ridge
The Valu Fa Ridge lies in close proximity to the Tonga arc front (40 km from Ata) and it is only 150 km above the Benioff zone. Previous studies (e.g. Vallier et al., 1991; Bach et al., 1998) have shown that the Valu Fa magmas have pronounced arc-like compositions. This is illustrated both in the MORB-normalized trace element plot in Fig. 7 (e.g. Pearce & Peate, 1995), where the contrast to the MORB-like samples from the CLSC and ELSC is clear, and by the higher U/Th of the Valu Fa magmas seen in Fig. 2. At issue is the origin of this strong subduction influence in a back-arc rift. Boespflug et al. (1990) suggested that this may be a result of the migration of the arc volcanic front into the back-arc region. The Valu Fa Ridge is propagating southwards into older crust, but the age and composition of that crust are essentially unknown (Hawkins, 1995a). Most workers have suggested that the Valu Fa magmas tap a depleted MORB mantle source plus a slab-derived component (Jenner et al., 1987; Boespflug et al., 1990; Loock et al., 1990; Vallier et al., 1991; Bach & Niedermann, 1998; Bach et al., 1998), but some have also raised the possibility that some of the subduction characteristics are acquired through interaction with this older crust during rifting (Vallier et al., 1991; Hilton et al., 1993; Hawkins, 1995a). Hilton et al. (1993) showed that 3He/4He of Valu Fa magmas is negatively correlated with SiO2 and that the more evolved samples have extremely low 3He/4He (R/RA 1): in detail, samples with 53–55% SiO2 had 6–8 R/RA, and it was only samples with >58% SiO2 that had <3 R/RA. Hilton et al. suggested that these He results could be explained by assimilation of Lau crust by previously degassed magmas, and this model requires that parts of this basement crust are old enough to have in-grown sufficient radiogenic 4He. However, such shallow-level assimilation does not appear to have markedly influenced the U–Th data. This can be seen from comparing the most primitive (84KD, SiO2 51%) and most evolved (133GA, SiO2 57%) samples from the Valu Fa spreading centre: both samples have identical (238U/232Th) and (230Th/232Th), despite a three-fold enrichment in Th content. Furthermore, the significant 238U–230Th disequilibrium in all Valu Fa magmas is not consistent with any model in which the subduction characteristics of the Valu Fa magmas are derived from shallow-level bulk assimilation of arc-related crust, as this crust would be expected to be old enough to have 238U–230Th in equilibrium.
|
Excesses of 238U over 230Th that are characteristic of many arc rocks are generally attributed to the recent addition of a U-rich fluid (e.g. Newman et al., 1984; Gill & Williams, 1990; Sigmarsson et al., 1990; McDermott & Hawkesworth, 1991; Condomines & Sigmarsson, 1993; Elliott et al., 1997; Hawkesworth et al., 1997a; Turner et al., 1997). In detail, suites of well-dated samples from individual arcs often define linear trends on an equiline diagram, which, if interpreted as isochrons, yield ages of between 20 and 70 ka (e.g. Sigmarsson et al., 1990; Elliott et al., 1997; Hawkesworth et al., 1997a; Turner et al., 1997, 1999). The U/Th ratios, and hence the arrays on the U–Th equiline diagrams, are interpreted as mixtures between wedge + sediment (low U/Th) and fluid from the subducted slab (high U/Th). As the fluid is inferred to contain little or no Th (e.g. Hawkesworth et al., 1997b), the mixing arrays are horizontal at the time of fluid addition, and the slope after subsequent 230Th in-growth is inferred to indicate the age since fluid addition. The low-U/Th end-member may be regionally variable as a result of differences in composition of mantle wedge and proportions of added sediment material. Some of the scatter in the 238U–230Th data from the Tonga arc reflects variable sediment contributions, and the scatter is significantly reduced by considering only those samples with similar Nd isotope ratios, as Nd is not thought to be present in significant quantities in the fluid component (e.g. Hawkesworth et al., 1997b; see below).
Lavas from the Central Tonga arc all have high (238U/232Th) and significant 230Th–238U disequilibrium (Regelous et al., 1997; Turner et al., 1997), but it is difficult to constrain the slope of an array through the data, because of the relatively restricted range in U/Th. The northern Tonga lavas have lower U/Th (Fig. 2), but their relationship to the Central Tonga lavas is complicated by the presence of additional components in their mantle source (Regelous et al., 1997; Turner & Hawkesworth, 1997, 1998; Turner et al., 1997; Wendt et al., 1997; Ewart et al., 1998). Lavas from Ata also have lower U/Th, but their age is uncertain so that the data may require significant age corrections on the equiline diagram. The Valu Fa magmas are another potential low-U/Th candidate, and it is therefore important to assess their relationship to the arc-front magmatism.
Pearce & Parkinson (1993) demonstrated that a plot of Nb vs Yb, using analyses of primitive basalts, can be effective in resolving partial melting differences from source depletion effects. Valu Fa basalts fall on the same trend as that defined by lavas from Ata and most of the Central Tonga arc (except Tofua), and this trend can be explained by different degrees of melting of a similar, depleted source (Fig. 8). Thus, it seems reasonable to infer that the mantle source at the Valu Fa Ridge is broadly similar to that for the Central Tonga arc-front volcanoes and that the Valu Fa magmas represent the best estimate of the low-U/Th, fluid-poor, end-member. Therefore, they can be used to anchor a trend line through data from the Central Tonga arc on the equiline diagram. To minimize the effects of variable sediment addition, the Central Tonga arc data are restricted to samples with similar Nd isotope compositions to the Valu Fa magmas (143Nd/144Nd 0·51302–0·51307: Turner et al., 1997; Bach et al., 1998; Ewart et al., 1998). The resulting trend line (Fig. 2) can be interpreted as an isochron, indicating an age of 50 ka. This is within error of the age inferred from 231Pa–235U systematics of samples from just the Tonga arc (Bourdon et al., 1999). Both trends can be most simply explained by the addition of a U-rich, slab-derived fluid to the mantle source at 50–60 ka. Thus, the time scales of fluid transfer from the subducted slab appear to be similar beneath the Valu Fa back-arc ridge and beneath the arc itself. This is consistent with a tectonic model in which the Valu Fa Ridge is propagating into the arc-front region.
Melt generation processes at the Valu Fa Ridge
Previous studies have shown that the Valu Fa basalts sample a refractory mantle wedge, as indicated by the presence of Mg-rich olivines and Cr-rich spinels (Sunkel, 1990; Kamenetsky et al., 1997; Bach et al., 1998). High field strength element ratios such as Zr/Nb and Ti/Zr indicate that this mantle is more depleted than an N-MORB source. A recent study on melt inclusions in olivines from near-axis seamount lavas on both sides of the Valu Fa Ridge spreading centre has highlighted the complexities of melt generation processes and the mineralogy and trace element composition of the diverse mantle sources beneath the Valu Fa Ridge–Ata region (Kamenetsky et al., 1997). Major element compositions of the melt inclusions suggest a refractory, harzburgitic, hydrated sub-arc lithosphere source for the seamount lavas, variably veined with clinopyroxene-rich dykes closer to the arc. Trace element data on the melt inclusions, coupled with whole-rock analyses of lavas from the Valu Fa Ridge itself and Ata, indicate the presence of three distinct components within this region. The seamount lavas and their melt inclusions, plus Ata, fall between two components on many trace element plots: a fluid-rich component with high Ba/Th and low Ce/Pb, dominant in the eastern seamounts, and a ‘boninitic’ component with high La/Yb and low Ba/Th, dominant in the western seamounts. The Valu Fa Ridge samples are displaced from these trends towards unmodified MORB-type mantle wedge compositions.
The seamount magmas have been interpreted as the products of melting of shallow, hydrated sub-arc lithosphere as a result of conductive heating and decompression caused by entrainment into upwelling MORB-source mantle of the developing Valu Fa Ridge (Kamenetsky et al., 1997). In this model, the decrease in ‘subduction’ signature from east to west would be due to a decrease in the extent of ‘fluid’ addition to the lithosphere before rifting. However, conductive heating is likely to take longer than fluid-induced melting, which in turn suggests that any excess 238U as a result of fluid addition will have decayed back to isotope equilibrium. Only one seamount sample has been analysed (61GC), and this has excess 238U similar to the other Valu Fa rocks, consistent with fluid-induced melting this close to the arc front. Although it is likely that hydrated asthenosphere can decompress (and melt) to shallower depths beneath the Valu Fa spreading centre than beneath the Ata volcano, the degree of melting is apparently higher at Ata than at the Valu Fa Ridge, as is indicated by Fig. 8. Figure 6 suggests that the additional melting beneath Ata can be linked to a higher fluid flux.
|
The Central Lau Basin
Characterization of mantle sources beneath the Central Lau Basin
The Central Lau Basin spreading centres are further behind the trench than the Valu Fa Ridge, and the subduction influence on magma compositions is correspondingly diminished (e.g. Hawkins, 1995a; Pearce et al., 1995). All the lavas are broadly N-MORB in composition, as can be seen in the MORB-normalized trace element diagram (Fig. 7). It is important to clarify whether the CLSC lavas have any compositional features to distinguish them from N-MORB, which might indicate whether any slab-derived fluid related to active subduction has reached that far behind the arc. Major element trends shown by CLSC lavas are similar to MORB (Hawkins, 1995a; Pearce et al., 1995), but there are significant minor differences in selected trace element ratios relative to average N-MORB values: e.g. primitive CLSC basalts have higher Ba/Th, Ba/Nb and K/Nb, and lower Ce/Pb (Figs 4, 5 and 9). The highly fractionated CLSC lavas are excluded from these discussions because they have low Ba/Th and high Ce/Pb as a result of extensive plagioclase fractionation (Pearce et al., 1995): both ratios show good correlations with Eu/Eu* (e.g. Fig. 10b). These trace element characteristics indicate the influence of a subduction-related component, and at first sight it is tempting to attribute them to the muted effect of material added by contemporaneous subduction processes. However, it is important to note that average N-MORB values (e.g. GERM Web page) are defined based primarily on samples of Atlantic and Pacific MORB. The Pb isotope data summarized earlier clearly show that the CLSC lavas are derived from a source similar to that of Indian MORB (Loock et al., 1990; Hergt & Hawkesworth, 1994). Recent studies have indicated that the Indian MORB source is distinct from the Atlantic and Pacific MORB sources not only in terms of isotope composition but in trace element composition as well, with Indian MORB characterized by enrichments in Ba and Pb and depletions in Nb relative to Atlantic and Pacific MORB (e.g. Rehkämper & Hofmann, 1997). Figure 9 indicates that the CLSC basalts have Ba/Nb ratios and Sr isotope compositions that overlap with the more enriched end of the Indian MORB compositional array. In terms of volatiles, the CLSC lavas are enriched in H2O (H2O/Ce = 345 ± 69 for CLSC (Jambon & Zimmermann, 1990; Danushevsky et al., 1993; S. Newman, unpublished data, 1999) relative to average N-MORB (183 ± 33 for N-MORB: Michael, 1995) and although water data are sparse on Indian MORB samples, preliminary data suggest that the Indian MORB source might also be characterized by relatively high H2O/Ce (Michael, 1995). Two CLSC samples measured for Li isotopes have higher 
6Li (-0·5
and -1·7) than the apparent range for Atlantic and Pacific MORB (-3·4 to -4·7) (Chan et al., 1999), but no data are yet available to verify whether this is also a feature of the Indian MORB source. Major element systematics also indicate that the primitive CLSC lavas were formed by melting of an Indian MORB mantle source through processes typical of a mid-ocean ridge spreading centre. Basaltic lavas from the CLSC plot on the broad correlations found between axial depth, Na8·0 and Fe8·0 in MORB globally (the subscript denotes that the composition is fractionation corrected to an MgO of 8 wt %: Klein & Langmuir, 1987). Furthermore, in detail, Indian MORB basalts have systematically lower Fe8·0 than Atlantic MORB basalts at a given depth (Langmuir et al., 1992), and this is also shown by the CLSC basalts (data in Pearce et al., 1995). Thus, it appears that the distinctive compositional features of the CLSC mantle source owe their origin primarily to ancient enrichments that are characteristic of the Indian MORB mantle source rather than being related to contemporaneous subduction processes (see also Pearce et al., 1995).
|
|
The evolved rocks of the CLSC have undergone extensive crystallization of a plagioclase-dominated assemblage from parental basaltic magmas as a result of a high rate of cooling relative to fresh magma supply associated with the propagating rift tip (Pearce et al., 1995). Oxygen isotope data suggest that the precursor magmas to these highly fractionated lavas had assimilated hydrothermally altered oceanic crust material (Macpherson & Mattey, 1998). The slightly elevated 87Sr/86Sr but similar 143Nd/144Nd of the evolved rocks relative to the basalts on the CLSC (Fig. 10a) is consistent with such a model. The few Ra data available show a positive correlation between (226Ra/230Th) and Eu/Eu* (Fig. 10c), which could be explained by extensive plagioclase fractionation (see also Turner et al., 2000). However, it is difficult to evaluate the effects of radioactive decay for the evolved samples in the absence of independent eruption age information. There is no clear relationship, though, between (238U/232Th) or (230Th/238U) with either a monitor of crystallization such as Eu/Eu* or a monitor of assimilation such as 87Sr/86Sr that might otherwise indicate an important influence on the 238U–230Th disequilibria data.
The ELSC and ILSC lie closer to the arc than the CLSC, and so any compositional differences between lavas from these two areas might reflect processes associated with the present subduction setting. Major element features of the ELSC lavas, such as lower Na2O and Fe2O3 and higher SiO2 at a given MgO, coupled with an earlier onset of oxide and apatite saturation, all indicate that the mantle beneath the ELSC is, in general, more hydrous and more depleted than the CLSC mantle (Hawkins, 1995a; Pearce et al., 1995). This is consistent with the observation that ELSC and ILSC glasses have much higher H2O/Ce (788–1266) than CLSC samples (224–470) (Jambon & Zimmermann, 1990; Danushevsky et al., 1993; S. Newman, unpublished data, 1999). The ELSC and ILSC basalts included in this study have slightly higher Ba/Th (Fig. 5) and higher 206Pb/204Pb (Fig. 4) than the CLSC basalts. ELSC samples dredged just to the south of the studied lavas are more highly vesicular, despite similar ridge depths, which suggests an even greater volatile content, and they also have higher Ba/Th (Pearce et al., 1995). The elevated H2O/Ce ratios for the ELSC and ILSC relative to the CLSC are also much greater than found in any MORB globally (<380: Michael, 1995) and the fact that these increase with arc proximity, along with Ba/Th, suggests that this reflects an additional fluid component related to the active subduction environment, rather than being a more extreme variety of Indian-MORB-type mantle (Fig. 5).
In the northern ELSC and ILSC, only increased water contents are detectable, whereas closer to the arc, south of 20°S, Ba abundances are also clearly elevated, indicating that the composition of the subduction component varies systematically towards the arc, as suggested by Pearce et al. (1995). Furthermore, the fluid component in the ELSC and ILSC mantle source is apparently different in composition from that found near the arc front, as discussed above, and one of the principal differences is that it does not have a high U/Th ratio (see Fig. 5). One plausible model to account for this regional variation in fluid composition is a chromatographic model (e.g. Stern et al., 1991; Hawkesworth et al., 1993; Stolper & Newman, 1994), in which the slab-derived fluid progressively equilibrates with mantle material during transport through the wedge. The fluid gradually loses its subduction signature depending on the length of the fluid path through the mantle and the fluid–mantle partition coefficients: only those elements with low fluid–mantle partition coefficients (e.g. Ba, K) will retain evidence of their slab origin at significant distances behind the trench. The calculated subduction fluid composition for the Mariana Trough back-arc basalts also had a relatively low (238U/232Th) of 1·19 ± 0·07 (Stolper & Newman, 1994).
Melt generation processes in the Central Lau Basin
The Central Lau Basin lavas are all characterized by having (230Th/238U) 1, which clearly distinguishes them from the arc-like Valu Fa samples with (230Th/238U) <1 (Fig. 2). The variations shown by the Central Lau Basin samples on the equiline diagram cannot be explained by addition of a high-U/Th fluid similar to that at Valu Fa to a melt with low U/Th and high (230Th/238U), as it is the two ‘fluid-rich’ ELSC and ILSC samples that have the lowest U/Th and highest (230Th/238U). Instead, <