Issuu

Geological Society of Australia

ABSTRACTS Number 45

Island Arc Magma Genesis Workshop February 2 - 7 , 1997

Geological Society of Australia Abstracts Number 45

State of the Arc '97 Island arc magma genesis workshop February 2 - 7, 1997 Mawson Laboratories The Department of Geology and Geophysics The University of Adelaide South Australia

Convenors: Dr Simon Turner Dr John Foden Prof Chris Hawkesworth

Sponsors:

THE UNIVERSITY OF ADELAIDE

SPECIALIST GROUP IN i ) GEOCHEMISTRY, MINERALOGY AND PETROLOGY

List of Abstracts Charlotte M. Allen Can island arc tectonic models be applied to major continental marginal batholiths?

Richard J. Arculus Geochemical evolution of island arcs - emphasis on the western Pacific

CJ. Bryant, R.J. Arculus and S.M. Eggins Geochemical evolution of the Izu-Bonin-Mariana arcs tracked by laser ablation ICP-MS analysis of individual glass shards

A.J. Crawford, A. Verbeeten, S. Eggins, L.V. Danyushevsky, P. Maillet, I.A. Sigurdsson and M. Monzier Hot subduction: Magmatism along the Hunter Ridge, SW Pacific

Jon. P. Davidson and Matthew F. Thirlwall Magmagenesis in the Lesser Antilles: A synthesis

J. Huw Da vies and Andrea Rowland Importance of temperature dependent viscosity and hydraulic fracture on physical models of subduction zone magmatism

Marlina A. Elburg and John Foden Magmatism in West Sulawesi, Indonesia: Geochemical characteristics and economic implications

John Foden Arc magmatic chemistry: the source versus process debate

John Gamble, Peter Wood, Richard Price, Tod Waight, Ian Smith and Mitsuhiro Nakagawa Major, trace element and isotope geochemistry of historic (1945-1996) eruptions from Ruapehu volcano, New Zealand with implications for open system magmatic processes in arc volcanoes

Trevor H. Green Application of mineral/fluid/melt trace element partitioning data to models of arc magma genesis

Chris Hawkesworth, Simon Turner, Frank McDermott, David Peate and Peter van Calsteren U-series isotopes in arc magmas

Janet Hergt Mantle dynamics during arc rupture and back-arc opening: evidence from Tonga, the Lau Basin and the Lau Islands

Pavel Kepezhinskas and Marc J. Defant Slab melt-mantle interactions: imphcations for arc magmatism and mantle evolution

Malcolm T. McCulloch Temporal changes in island-arc recycling: Pb isotopic evidence for post-Archean uranium recycing (PURE)

M.D. Murphy, J. Barclay, R, Macdonald, RJ.S. Sparks and M.R. Carroll Amphibole stability and its role in influencing intermediate fractionation trends in subduction-related magma series

Ian Nicholls, Danilo Vukadinovic and Jane Barling Isotope geochemistry of lavas of central Java, Indonesia and Heard Island, Indian Ocean - evidence for sedimentary and mande plume source components

Geoffrey T. Nichols, Trevor H. Green and Norman J. Pearson In search of fluid - melt partitioning values; development of the laser ablation ICP-MS method

Marc D. Norman Lithospheric mantle source for continental arc crust? Trace element compositions of diopside in spinel Iherzolites

I.J. Parkinson and RJ. Arculus Ground truthing arc melting models: peridotite geochemistry from forearc to backarc

Julian A. Pearce Trace element behaviour in arc systems

David W. Peate, Chris J. Hawkesworth and Julian A. Pearce Radiogenic and U-series isotope variations in Vanuatu arc lavas: fluid and sediment addition to an isotopically heterogeneous mantie wedge

Terry Plank and Marie Johnson Do subducted sediments melt beneath arc volcanoes? Some experimental answers

Richard C. Price, Tod E. Waight, John R. Chapman, Eloise E. Beyer, Ian E.M. Smith and Robert B. Stewart The geochemical evolution of arc magmas in a continental setting: evidence from detailed chemo-stratigraphy at Ruapehu, New Zealand

Bruce F. Schaefer, J.D. Hoek and John Foden Physical and chemical characteristics of the Lincoln batholith: apsects of Palaeoproterozoic magmatism and potential analogues

R. Stalder, G.A. Jenner, S.F. Foley, I. Horn and G.P. Brey The role of aqueous fluids in trace element fractionation in subduction zone processes: evidence from experiments at 3 to 5.5 GPa and 1000° C

Bob SteM^art, Richard Price and Ian Smith Fractionation of oxygen isotopes amongst melt, minerals and hydrous fluids, Egmont Volcano, New Zealand

Yoshihiko Tamura Island arc magma generation from thermally zoned mantle diapirs

Simon Turner, Chris Hawkesworth, Nick Rogers, Jessica Bartlett, Ian Smith and Tim Worthington U-series disequilibria, magma petrogenesis and flux rates along the depleted

Tonga-Kermadec island arc

M.C. van Soest, D.R. Hilton and R. Kreulen

Rick Varne and Massimo Gasparon

103

Using helium and carbon (CO2) isotopes to distinguish between source and crustal contribution in arc magmas. A case study from the Lesser Antilles island arc The relative importance of crustal contamination and subducted sediment components in west Sunda arc magmatism

Jon Woodhead, Steve Eggins and Wally Johnson

104

Tim J. Worthington, Richard C. Price and Ian Smith

108

Ming Zang, Jon Stephenson, Suzanne Y. O'Reilly, Marc Norman and Malcolm T. McCulloch

110

Chemical dynamics in the New Britain volcanic arc

Towards decoding source and process variations at arc volcanoes: trace elements and isotope analyses from Raoul (Tonga-Kermadec arc) and Ruapehu (Taupo volcanic zone)

Geochemistry and petrogenesis of basaltic rocks from north Queensland: has subduction-modified mantle played a role?

Can Island Arc Tectonic Models be applied to Major Continental Marginal Batholiths? Charlotte M. Allen Department of Geology, Key Centre for the Geochemical Evolution and Metallogeny of Continents, The Australian National University, Canberra ACT 0200, charlotte.allen@anu.edu.au

The lack of correspondence of the average compositions of continental marginal batholiths and island arc volcanic rocks has been noted for some time {cf Silver and Chappell, 1988; Table 1). Batholiths are, on the whole, much more "continental" with higher average contents of elements such as Si and incompatible trace elements. I am interested in exploring the reasons for the lack of correspondence, to learn about island arcs, and to understand just what range of stress regimes can be categorised as "island arc" or directly subductionzone-related. Is the major difiference between batholith and island arc composition simply the presence of thick continental crust that acts as a physical barrier to mafic magmas, which therefore melts and adds to batholith construction? Hildreth and Moorbath (1981) argued, based on research in the central Andes, that crustal thickness was the cause of observed systematic changes along the length of the Quaternary continental marginal arc. They suggested that increased crustal thickness leads to increased participation of crustal partial melts in MASH processes. The product of these interactions as sampled in Quaternary central volcanoes, however, is still significantly more mafic than that of the Sierra Nevada or Peninsular Ranges batholiths (Table 1). On the other hand Ward (1995) suggests that major batholiths are the products of a different tectonic setting not easily related to orderly chains of volcanoes referred to as arcs. He argued that for the North American Cordillera, major batholiths are much more easily related to broad fields of disorganised, calderaforming silicic magmatism than to dominantly andesitic volcanism in places like the Cascades. Furthermore, he compiled crystallisation ages that support the idea that silicic magmatism and correlated batholith-building events follow "arc" formation by 20-30 Ma in regular tectonic cycles that can be encapsulated in three stages: 1) rapid orthogonal subduction and arc generation, 2) clogging of the subduction zone, slowed or stopped subduction, extension and batholith formation, and 3) crustal thinning, mafic magmatism, and chaotic plate boundary reorganisation. This model has several appealing features if applied to the Carboniferous plutons of the northern New England Fold Belt, Australia. Chappell and Stephens (1988) argued that only several generations of melting ("remagmatisation") can differentiate a basaltic parent to give the large volumes of silicic plutons that average granodiorite in composition. In their model, M-type plutons are directly related to mantle derived magmas and these are found primarily in island arcs. I-types, though, are reprocessed crust. It is the mechanism of this reprocessing that interests me and whether M- and I-types are truly related. A problem with these comparisons, that cannot be ignored, is that sampling may have biased the averages, not the actual field collection but the obvious difference of level of intrusion between batholithic and volcanic rocks (ala Bruce e t a l , 1989). REFERENCES Bruce, R.M., Nelson, E.P & Weaver, S.G., 1989. Effects of synchronous uplift and intrusion during magmatic arc construction. Tectonophysics 7-329. Chappell, B.W. & Stephens, W.E., 1988. Origin of infracrustal (I-type) granite magmas. Transactions of the Royal Society of Edinburgh: Earth Sciences 79, 71-86. Ewart, A., 1982. The mineralogy and petrology of Tertiary-Recent orogenic volcanic rocks: with special reference to the andesitic-basaltic compositional range. In Thorpe, R.S. (ed) Andesites, p. 25-87, Wiley. Fierstein, J, Bruggman, P, Schwarz, et al. (1989). Chemical analyses of rocks and sediments from central Chile. U.S.GX Open File Report 89-78, Hildreth, W. and Moorbath S. (1988). Crustal contributions to arc magmatism in the Andes of central Chile. Contributions to Mineralogy and Petrology, 98, 455-489. Silver, L.T. and Chappell, B.W., 1988. The Peninsular Ranges batholith: an insight into the evolution of the Cordilleran batholiths of southwestern North America Transactions of the Royal Society of Edinburgh: Earth Sciences 79, 105-121.

Table 1 Average Compositions SW wPRB'' Pacific^ 174 number 57.77 63.41 Si02 0.73 0.64 Ti02 16.32 15.70 AI2O3 7.32 5.00 FeO 0.14 0.09 MnO 2.71 4.62 MgO 7.75 5.70 CaO 3.20 3.45 NazO 1.64 1.69 K2O 0.23 0.10 P2O5 99.72 98.49 total Cr Ba Rb Sr Y Th U

130 462 40 523 24 nd nd 22

67 451 49 268 24 6.0 1.4 12

149 66.05 0.65 16.23 3.41 0.06 1.49 4.41 3.82 2.26 0.15 98.53

Central Chile= 104 60.33 0.77 16.89 5.18 0.15 3.09 5.45 4.07 2.33 0.22 98.48

24 863 73 501 12 8.6 1.7 20

50 487 77 510 22 9 3 25

ePRB''

La Ewart (1982) PRB=Peninsular Ranges batholith, Chappell and Silver (1988) ^ Fierstein et al (1989)

Geochemical Evolution of Island Arcs - emphasis on the western Pacific Richard J. Arculus National Key Centre for Teaching and Research Geochemical Evolution and Metallogenesis of Continents Department of Geology, Australian National University, Canberra, ACT 0200, Australia Most models of arc petrogenesis are dynamic to varying extents, invoking relative movement of source components, and complex passages of magmas from source(s) to the crust and Earth's surface. The potential for temporal variation of magmatic chemical composition in these types of systems is clearly large. Considerable effort has been expended over the past 40 years in attempts to detect temporal variations in arc magmas, and to understand the (in)stability of the systems giving rise to the types of secular patterns observed. One of the prime reasons why we should be interested in these particular aspects of arc behaviour relates to the issue of global geochemical fluxes, especially those concemed with recycling of elements from the exosphere [atmosphere-hydrosphere-biosphere (-lithosphere)] back into the Earth's interior. Another reason for detailed study of temporal trends is the possibility of relating (in)stabilities of erupted compositions to length scales of heterogeneity in the various source components involved Attempts at reconciliation of subduction zone inputs V5. outputs over periods of time that represent essentially instantaneous sampling of these complex systems are unlikely to be convincing unless supported by more extended time studies. For example, our understanding of the recycling of elements like the alkalis, alkaline earths and Pb is limited for extensive time periods. And for elements like C and the chalcophiles (e.g., Cu, Zn, and many of the platinum group elements), we have meagre knowledge of recycling through subduction zone systems at all time scales, despite the clear anthropogenic need for such information. Terrestrial studies of the long-term geochemical evolution of arc systems are hampered by the problems of inadequate exposure and weathering of samples. In contrast, while deeply drilled recovery of pyroclastic deposits in the marine realm has demonstrably overcome these problems, the selectivity (weighted to compositions generally more felsic than basaltic andesite) of sampling of the magmatic systems involved, and the difficulty of source assignments means a combination of terrestrial and submarine approaches is appropriate. Commencing some 30 years ago, observations that magmas become progressively more alkalic with time for any given arc were enthusiastically and variously adopted for hypotheses of crustal growth in the overriding plate, increased degrees of crustal interaction via assimilationfractional crystallisation, staged ponding of magmas at increased depths in the crust, and variation of subduction zone parameters. Some of these ideas stemmed appropriately from the "comparative subductology approach espoused by Seiya Uyeda. For example, geochemical comparisons of magmas erupted through different crustal thicknesses, and at varying distances from a neighbouring trench and depths to subducted lithosphere have been particularly important in this regard. However, a growing awareness of the global variation in the nature of the prime source of most arc magmas - the mantle wedge overriding the subducted plate - has also prompted studies of the histoiy of the wedge through prior melt extraction and metasomatic (melt ± supercritical fluids) imprint events as a complicating factor. The trade-offs between melt extraction, subducted plate-derived inputs, and advective cycling of mantle into the region(s) of melt generation are some of the parameters sought through inversion of erutped magma characteristics. A major advantage of terrestrial studies is the opportunity for identification of eruptive source loci. For example, investigations of major arc systems such as Kamchatka and northern Honshu have shown marked migration in time and space of the volcanic front, and assumed variation in the detailed nature of the subjacent mantle wedge. In the case of Kamchatka, the alkali-rich Sredinny Range (SR) located --400 km west of the present trench was highly active in the late Miocene-Pliocene; at present, activity in the SR is confined to a single volcano at the southern end of the range, and the dominant Pleistocene-Recent activity (generally lower-K)

has been in the Central Kamchatka Depression and southeastern Kamchatka (--150 km from the trench). These on-land studies have been substantiated by analysis of ashes recovered by ODP Leg 145 in the northwest Pacific. During the past 25 Ma in northern Honshu, the volcanic front has migrated eastward and then westward, reaching its present position at --3-5 Ma. A complicating factor in the evolution of the northern Honshu arc has been the development of the Sea of Japan, with a prolonged period of rifting/stretching culminating in spreading at --15 Ma. Numerous studies have documented the ingress of Indian Ocean (lO)-type asthensophere beneath the Sea of Japan, transporting fragments of East Asian lithosphere with EMI & 2 characteristics eastwards in the Japanese archipelago. It appears that both lO-type MORB and various proportions of EM lithospheric components continue to be significant arc source components to the present of Honshu. In comparison with Kamchatka and Japan, subaerial exposures of the intraoceanic arcs of the western Pacific are limited, but comprehensive marine studies have unravelled the broad outlines of the tectonic evolution of these systems. Complications that have arisen with many of these systems include one or more episodes of back-arc rifting, and in the case of the Melanesian arcs of (parts of) Papua New Guinea, Solomons and Vanuatu, a post-Middle Miocene subduction polarity reversal. A particularly significant observation stenmiing from a combination of marine geophysical and analytical (ashes) studies is the volumetric importance of dacite-rhyolite compositions, particularly in the Izu-Bonin (IB) system, erupted from submarine calderas. In contrast, the supermarine edifices are dominated at surface by more mafic compositions. The presence of a significant crustal thickness {-1 km of a total ~ 20 km) of what appear to be tonalitic compositions in the IB crust is reconcilable with these surface observations. Important conclusions that can be drawn from these data however, given the probable fractional crystalUsation derivation of felsic-from-mafic magmas, are:-l. a substantial mafic-ultramafic cumulate complement should be present below the Moho of the IB system; 2. this complement may be transient and incorporated and recycled into the advecting wedge; 3. the bulk composition of the geophysically-defined IB crust is probably andesitic. Notable results from several workers concerning the evolution of the IB-Mariana (M) systems over '-48Ma of eruptive history since arc inception are:- 1. absence of a consistent and monotonic increase in alkalinity overall; 2. an apparent steady increase in the minimum K content of the Mariana magmas; 3. a change in comparative fertility (based on systematics of elemental ratios such as Nb/Ta, ZrAf, and Ught rare earth element (REE)/heavy REE) of mantle wedge sources in the IB arc from relatively fertile to refractory subsequent to backarc (Shikoku Basin) formation; 4. the absence of such an effect in the M arc, where more extensive backarc spreading has occurred; 5. the appearance of alkaline compositions in the M arc, related to one or more of: i. arc renaissance subsequent to initial stages of backarc rifitng; ii, tapping of a long-lived wedge heterogeneity along arc strike; UL first subduction of distinctive high-ji sediments derived from the Magellanic Seamounts; 6. the overwhelming predominance of tholeiitic magmas in the IBM systems - based on analyses of glass shards only. Note that these glasses are also persistently lower in A1 than contemporaneous arc rocks suggesting ubiquitous plagioclase accumulation. Given the compositional stability (including Sr - Nd isotopes) of the IBM systems over time periods of -- 10 Ma, we can conclude that the mantle wedge must be advectively transported (i.e. refreshed) through zones of slab-derived fluid/melt ingress and melt triggering, and at maximum slab-wedge coupling rates of ~ 10 cm a" the scale of across-strike wedge homogeneity must be of the order of ~ 1000 km. However, detailed studies (ash layer-bylayer) might reveal more finely resolved temporal variations that could be anticipated on the basis of features developed in the various backarc basins - e.g., arc-parallel spreading strike (Parece Vela, Shikoku, and Mariana Trough) vs. arc-normal spreading strike (West Philippine Basin).

Geochemical evolution of the Izu-Bonin-Mariana arcs tracked by laser ablation ICP-MS analysis of individual glass shards C J . Bryant, RJ. Arculus, and S.M. Eggins National Key Centre for Teaching and Research Geochemical Evolution and Metallogenesis of Continents Department of Geology, Australian National University, Canberra, ACT 0200, Australia Introduction. The plinian ash record is particularly valuable for understanding arc volcanism in that the ash layers form isochronous markers that enable the accurate geochemical characterisation of liquid compositions at a specific point in the (temporally extensive) subaerial volcanic activity of an individual arc system. However, until comparatively recently, there have been insufficient trace element data sets for glass shards to examine the temporal variations, primarily due to analytical difficulties. Here we present a summary of the compositional systematics revealed by -260 laser ICP-MS trace element analyses of -60 ash layers from the Izu-Bonin-Mariana (IBM) arcs. The intraoceanic IBM arc-backarc systems are an excellent testing ground for this type of study because (relative to many other arc systems) the tectonics and geology of the region are well understood through a combination of marine geological/geophysical/drilling and on-land studies. The essentials of the geologic history are: 1. Late Cretaceous-early Eocene spreading of the West Philippine Basin behind an arc system comprising the Amami Plateau-Daito Ridges; 2. proto-IBM arc inception on the northern (now eastern, because of the -90^ clockwise rotation of the Philippine Sea Plate) flanks of this system at -48 Ma with formation of the Palau-Kyushu Ridge; 3. initiation of backarc spreading in the south (Parece-Vela Basin) at -30 Ma behind an eastward migrating active arc of the West Mariana Ridge, propagating north to commence backarc spreading in the Shikoku Basin at 25 Ma; 4. cessation of spreading in the Parece Vela and Shikoku Basins at -15 Ma; initiation of backarc spreading in the Mariana Trough at - 8 - 6 Ma; initiation of backarc spreading in the IB system at - 2 Ma. The analysed M ashes cover the last 30 Ma history of the arc, whereas the IB tephras span only the last 16.5 Ma history of the arc. In both cases, the ashes range from basalt to rhyolite. Mariana On the basis of geochemical characteristics and in broad correspondence with the tectonic evolution of the M arc, the ashes have been divided into 4 groups: (i) >30 Ma, which in this study only includes one boninitic and one rhyolitic ash horizon; (ii) 30-13 Ma; (iii) 13-6 Ma; and (iv) <6 Ma. The 30-13 Ma ashes are low-K (Fig. 1) and tholeiitic, containing low abundances of high field strength elements (HFSE- Nb, Ta, Zr, Hf), alkalis (Rb, K Cs), other large ion lithophile elements (LILE-Sr, Ba), light rare earth elements (LREE) and other magmaphile (with respect to mantle assemblages) elements such as P, Th and U. Slightly more alkali-rich ash layers occur at -22 Ma and 16-17 Ma; however, the concentrations of incompatible elements remain typically below those observed in the <13 Ma ash horizons. The 30-13 Ma group has relatively flat REE patterns (LaNArbN = 0.5-1.5), although higher values are recorded in ash horizons at -22 and 27.5 Ma. The 34.9 Ma rhyolite horizon, is geochemically equivalent to rhyolites of the 3013 Ma group. The boninitic glasses overlap with the 30-13 Ma group for many elements at equivalent S i 0 2 contents, but contain significantly lower abundances of HFSE (except Nb), Y, Sr, and higher Rb and Cs. The 13-6 Ma group are intermediate to high-K tholeiites (Fig. 1), although they are more calc-alkaline in character than other IBM ashes, particularly at higher Si02. At >56% Si02, this group evolves toward high AI2O3, slightly higher Na20, and lower total FeO and CaO contents, despite having equivalent abundances of these elements at low Si02. Such characteristics may be explained by higher H2O activities with concomitant expansion of the primary liquidus field of clinopyroxene and suppression of plagioclase. The high alkali contents of these ashes are accompanied by high LILE, and HFSE, as well as P, B and higher LREE/HREE ratios (LaN/YbN typically = 2-4). In the <6 Ma group, there is a substantial decrease in the incompatible element abundances; however, the abundances never drop to the levels observed for the 30-13 Ma group (Fig. 1), typically overlaping the lower end of the abundance ranges of the 13-6 Ma ashes at equivalent Si02. Izu-Bonin Excluding one tephra horizon that is likely to be derived from Iwo-Jima or a neighbouring volcano (high LILE, LREE, HFSE and LaN/YbN = 5.7-6), ashes recovered from the IB forearc have generally lower abundances of alkalis, LILE, and HFSE than M ashes. The generally more incompatible element-depleted nature of the ashes is also reflected in low LREE/HREE ratios (average LaN/YbN = 0.58). With the exception of the more fractionated samples (>65% Si02) where accessory phases are more likely to influence Zr, HREE and other trace element abundances, the increase in incompatible element concentrations observed for the IBM arcs (<16.5 Ma IB - > 30-13 Ma (M) - > 0-6 Ma (M) - > 13-6 Ma (M), is accompanied by concomitant increase in Zr/Y, Zr/Yb, Nb/Ta, Zr/Hf and Zr/Nb ratios. We note that systematic spatial and temporal variations of Zr/Y are characteristic of the back-arc basins. Specifically, the West Philippine Basin has very

low (-2) Zr/Y ratios. This ratio increases in the Parece Vela Basin (~3) and then again in the Mariana Trough (>~3). Discussion It is generally agreed that HFSE elements and HREE (and by analogy Y) are relatively immobile in slab-derived hydrous fluids (see Pearce & Peate, 1995; for review). Consequently, variations in Zr/Y, Zr/Yb, Ti/Zr can be used as monitors of relative enrichment/depletion characteristics of the asthenospheric mantle wedge (Fig. 2). Of particular importance in the IBM system are the positive correlations between Nb/Ta, Zr/Hf, and Zr/Y, clearly reflecting variations in the degree of magmaphile element depletions in the mantle wedge sources. It has been argued from the temporal chemical changes of IB arc-derived volcaniclastic turbidites that the post-Shikoku Basin increase in refractoriness reflects prior melt extraction in the backarc basin melting event and consequent depletion (lowering of Zr/Y) of the wedge source entrained in arc magma generation (Gill et al., 1994). The converse is seen in the M arc where post-Parece Vela Basin arc-derived ashes (13-6 Ma group) have the highest Zr/Y of the entire history of the M arc. The lack of generalisable chemical systematics of the IBM system as a function of tectonic evolution was pointed out by Arculus et al. (1995). Remarkably, the increase in Zr/Y ratios in IBM ashes (excluding boninites) is accompanied by an increase in the incompatible element concentrations - that are conventionally regarded as being overwhelmingly slab-derived such as the alkalis and other LILE, Th and U. Furthermore, there is a positive correlation between Zr/Y and Rb/Sr, Ba/Sr (Fig. 3), R b / K 2 0 , Th/Yb and Th/U. Notably, the increase in these ratios is not correlated with La/Ba. For a flxed volumetric addition of LILE-bearing, slab-derived fluid to mantle wedge sources of varying refractoriness, we might expect a lower degree of partial melting for the more refractory compositions, and consequently higher LILE/HFSE or Rb/Sr, Ba/Sr ratios. Globally of course, the elevated Ba/La, Ba/Nb and Sr/Nd regarded as diagnostic of island arc compared with ocean ridge or hot spot magmas are best developed in arcs with low overall HFSE and REE abundances such as the IB, Tonga or New Britain arcs. This feature has been explained as resulting from generally similar slab-derived fluid fluxes triggering melting of variably depleted (and hence low-to-high intrinsic HFSE and REE abundances) mantle wedge sources. Positive correlations between Zr/Y and Rb/Sr, Ba/Sr, Th/U, are contrary to what one might expect given fluid-fluxed melting of a variably depleted source. The positive correlation observed for magmaphile elements generally in the IBM systems is anticipated if wedge fertility (monitored by Zr/Y for example) is coupled with higher intrinsic abundances of the LILE and a major fraction of the LILE in the arc magmas is wedge- rather than slab-derived. There has been a minority view (e.g., Arculus, 1981; Arculus & Powell, 1986; Hawkesworth et al., 1991) that the mantle wedge plays a significant budgetary role for what are typically viewed as purely slab-derived components such as Sr, and Pb, primarily based on the lack of correspondence between abundances and isotopic ratios calculated for slab contributions. Hawkesworth et al. (1991) argue that slab-derived aqueous fluids are able to leach elements (>-80% of the arc magmatic flux in the case of Sr) from the mantle en-route to the melting zone. However, aqueous fluids appear to be incapable of significantly altering Zr/Y ratios. Alternatively, Hickey-Vargas (1992) has argued that Pb isotopic signatures for the IBM arc specifically can be explained by the migration of small volume asthenospheric melts of HIMU (high U/Pb) character or melts of subducted Magellanic Seamounts HIMU volcaniclastics. A melt from such an enriched mantle source will plausibly also be characterised by high Zr/Y ratios. On the other hand, the observed correlations between Zr/Y and Rb/Sr etc may simply reflect the relatively fertility of the mantle. Arculus et al. (1991) argue that it is mantle compositions rather a slab-derived component that best explains chemical changes along the IzuHonshu transect. In the case of the IBM, the changes in wedge fertility might occur consequent to the injection of a more enriched mantle source, or incorporation of enriched lithosphere. Notably, the beginning and end of the high-K episode occur at 13 and 6 Ma respectively. Stern et al. (1988) argue that the high-K event is a precursor to backarc rifting. However, it is interesting to note that the high-K (and accompanying geochemical characteristics) ashes appear some 3-5 Ma (depending on the actual initiation age) before the Mariana Trough rifting event. No such major pulse incidentally is observed prior to the opening of the nascent backarc basins of the IB arc. Additionally, it is interesting to note that both the beginning and end of the high-K event are offset from the end of back arc rifting in the Parece Vela Basin, and beginning of back arc rifting in the Mariana trough respectively by about 3 Ma. Consequently, the high-K event may not be a response to backarc rifting but an "unrelated" event, and that the geochemical response time of volcanic magmas to a change in tectonic processes is (at least in the case of the M system) about 3 Ma. Such a feature could possibly relate to mantle advection rates in the wedge, and the time required to "clear out" the wedge. A major global geochemical issue is balancing the mass flux in-and-out of subduction zone systems. We stand to make to the best progress in studies where along - and across-strike variation in subducted slab inputs (e.g.. Plank and Langmuir, 1993) can be tested against along- and across-strike variations in the overriding mantle wedge and lithosphere, coupled with temporal studies. Of specific interest in terms of an IB vs. M comparison is the persistent enrichment of the alkalis and other LDLE in the latter compared with the former.

Following the success of complete sedimentary penetration into Jurassic crust at ODP Site 801 outboard of the M arc, it is clear that a similar reference site is critically needed outboard of the IB system to determine the role of slab input vs. other possible parameters in the control of these chemical characteristics. References Arculus, R.J. 1981. Island arc magmatism in relation to the evolution of the crust and mantle. Tectonophysics 75,113-133. Arculus, R.J. & Powell, R. 1986. Source component mixing in the regions of arc magma generation. Journal of Geophysical Research 91, 5913-5926. Arculus, R.J., Gust, D.A. & Kushiro, K. 1991. Fuji and Hakone. National Geographic Research and Exploration 7, 276-309. Arculus, R.J., Gill, J.B., Cambray, H., Chen, W. & Stern, R.J. 1995. Geochemical evolution of arc systems in the western Pacific: the ash and turbidite record recovered by drilling. In Active Margins and Marginal Basins of the Western Pacific (B. Taylor & J. Natland, eds.), AGU Geophysical Monograph 88, 45-65. Gill, J.B., Hiscott, R.N. & Vidal, Ph. 1994. Turbidite geochemistry and evolution of the Izu-Bonin arc and continents. Lithos 23, 135-168. Hawkesworth, C.J., Hergt, J.M., Ellam, R.M. & McDermott, F. 1991. Elemental fluxes associated with subduction zone magmatism. Philosophical Transactions of the Royal Society of London, Series A 335, 393-405. Hickey-Vargas, R. 1992. A refractory HIMU component in the sources of island-arc magma. Nature 360, 5759. Pearce, J. A. & Peate, D.W. 1995. Tectonic implications of the composition of volcanic arc magmas. Annual Reviews in Earth and Planetary Sciences 23, 251-285. Plank, T. & Langmuir, C.H. 1993. Tracing trace elements from sediment input to volcanic output at subduction zones. Nature, 362, 739-742. Stern, R.J., Bloomer, S.H., Lin, P.N., Ito, E. & Morris, J. 1988. Shoshonitic magmas in nascent arcs: new evidence from submarine volcanoes in the northern Marianas. Geology 16,426-430.

wt % S i 0 2 Figure 1: Harker diagram for K2O illustrating the temporal variation in the Izu-BoninMariana arcs. Symbols as in Figure 2 with filled circles = >30 Ma Mariana ashes.

lOOOj X

M<6Ma

M 6-13 Ma

M 13-30 Ma

• •

100:

MORE array

Boninites IB

N-MORB

NbA'b Figure 2: Scatterplot of Zr/Yb versus Nb/Yb (after Pearce & Peate, 1995) for the IzuBonin-Marianas ashes, illustrating the apparent variations in mantle fertility. 2.5 X

M<6Ma

M 6-13 Ma

M 13-30 Ma

•

Boninites

•

1.0 • CQ

fertility

0.2 Zr/Y Figure 3: Plot of Ba/Sr against Zr/Y for the different Mariana subgroups, and the Izu • Bonin ashes. Arrow indicates the direction of apparent increase in mantle fertility.

Hot Subduction: Magmatism along the Hunter Ridge, SW Pacific A. J. Crawford^. A.Ve^beeten^ S. Eggins^, L.V. Danyushevsky^ P. Maillet^ , 1. A. Sigurdsson^ , M. Monzier"^ 1: SRC for Ore Deposit Research, UTasmania, Hobart, Tas., 2: Dept of Geology ANU, Canberra ACT. 3: ORSTOM Center de Brest, France 4: ORSTOM Center, Ecuador

The Hunter Ridge is a largely submarine intra-oceanic island arc extending from the southerrunost Vanuatu arc to the Koro Sea region of Fiji, 400 km to the northeast (Fig. la). Although the few available bathymetric transects suggested a ridge-trench morphology, this Hunter 'Fracture Zone' has generally been regarded as a transform plate boundary linking the oppositely dipping Tongan and Vanuatu subduction systems. The Hunter Ridge arc was conceived when a major reorganization of spreading ridges in the N Fiji Basin produced an E-W orientated spreading system (Fig. lb) across this backarc basin, probably commencing around 7Ma (Auzende et al., 1996). Another reorganization of spreading ridges around 3Ma (Fig. lb) led to the demise of the Hunter arc, and produced a N-S orientated spreading ridge system in the N Fiji Basin, which is still active, and may be propagating southward into the young arc lithosphere beneath the central Hunter Ridge (MaiHet et al., 1989).

NFB lyiuivtei^SFB -10''$

170®

180°E

Figure 1: (a) Location of the Hunter Ridge, (b) tectonic evolution of the Hunter Ridge since TMa (modified after Auzende et al., 1996). Dredging along the Hunter Ridge and sampling of its northernmost extent, exposed as the island of Kadavu in Fiji, has yielded a diversity of magmatic suites, including arc tholeiites and high-Ca boninites, high-Mg lavas with some affinities to boninites and some affinities to adakites, and true adakitic lavas associated with remarkable low-Fe, high-Na basalts with 8-16 ppm Nb (herein high-Nb basalts). Lavas which show clear evidence of slab melt involvement in their petrogenesis occur at either end of the Hunter Ridge, whereas the arc tholeiites and high-Ca boninites appear to be restricted to the south central part of the ridge. Below, mineralogical and wholerock geochemical data for each of these suites are summarized, and a tectono-magmatic model for their genesis and distribution is suggested.

1: High-Mg Andesites: At the southernmost tip of the Vanuatu arc where it swings into the Hunter Ridge, high-Mg andesites (HMA) have been dredged from several submarine volcanoes, and sampled on the tiny islets of Matthew and Hunter; the latter are located --SSkm above the subducted slab (Monzier et a l , 1993). These HMA share some affinities with both boninites and adakitic lavas, and are substantially different from basalts in the southern Vanuatu arc (e.g. Tanna, Anatom), which are mainly low-K arc tholeiites (our unpubl. data). The HMA lavas are very low-Fe magmas (Fig. 2) with compositional features (e.g., abnormally high N a 2 0 / C a 0 and Zr/Sm) suggesting that a significant slab melt component was involved in their petrogenesis. For lavas with 5-9% MgO, Sc (22-34ppm) and Y (12-20ppm) contents are low and Sr (300-500ppm) contents are high relative to Vanuatu arc tholeiites. However abundances of these elements are not as low, and HREE depletions (La/Yb]sj=4.16.8) are less pronounced than in the true adakitic lavas from the NE end of the Hunter Ridge on Kadavu. Thus slab melts involved in the petrogenesis of these HMA may have included partial melts of both amphibolitic and amphibole eclogite source rocks in the subducted slab. However, common olivine phenocrysts with Fo88-94 and Cr-rich chromites (Cr# =71-82) indicate either that extensive interaction with the mantle wedge occurred during their ascent, or alternatively, that these magmas were produced by partial melting of mantle wedge peridotite that had been significantly modified by reaction with slab melts. Radiogenic isotope data for 4 HMA lavas {^'^Sr/^^Si (0.70292-0.70322), 143Nd/144Nd (0.51302-0.51305) and Pb isotopes (206/204pb=i8.823-18.934,207/204pb= 15.527-15.560) and ^^^/^^^Pb=38.328-38.447)) are almost identical to data for adakitic lavas and high-Nb basalts from Kadavu and the northernmost Hunter Ridge, both groups plotting on the Northern Hemishpere Reference Line (NHRL), and precluding involvement of subducted pelagic sediment.

70TSiO2% Hunter

S W Hunter Ridae o

High-Mg Andesites

Central Hunter Ridae •

High-Ca Boninites

Kadavu ^NE Hunter Ridae) •

Adakites

+ Ngaloa G p

Figure 2: Compositional features of the high-Mg andesites, adakitic lavas and high-Ca boninites from the Hunter Ridge, showing compositional fields for arc tholeiites from the southern Vanuatu arc and the south-central Hunter Ridge. 2: Primitive Arc Tholeiites. Further northeast along the Hunter Ridge, particularly in the region where the southernmost spreading centre of the N Fiji Basin is believed to be propagating into and through the Hunter Ridge (Maillet et al., 1989), dredged lavas include a continuum of compositions between very depleted arc tholeiites and lavas transitional to high-Ca boninites (Sigurdsson et al., 1994; our unpubl. data). The lavas with boninitic affinities in terms of Si02 (>53%) and MgO (>5%) contents have remarkably low Ti02 (mainly <0.3%) and Zr (<20ppm) abundances, and have notably lower Na20 and total Fe than the associated arc tholeiites (Fig. 2). The boninitic lavas were probably generated from a mantle source residual after extraction of the arc tholeiitic parent magmas, but at shallower levels, as indicated by the lower Fe and higher Si02 contents of the boninitic lavas. Both

the arc tholeiites and the boninites along the south central Hunter Ridge may have been produced when subduction commenced 5-3Ma beneath the first spreading system in the N Fiji Basin (Fig. lb), leading to abnormally shallow dehydration of subducting S Fiji Basin oceanic crust into hot lithosphere of the N Fiji Basin created only a short time earlier. Alternatively, they may have been generated when the southernmost spreading centre in the current N Fiji Basin propagated into the Hunter Ridge arc probably during the last 2 m.y.; dating in progress may confirm which of these models is appropriate. 3: The northern submarine section of the submarine Hunter Ridge is composed of both arc tholeiitic and primitive calc-alkaline basalts and their fractionation products. The latter show some compositional affinities to the HMA suite from further SW on the Hunter Ridge, have trace element signatures suggestive of involvement of slab melt in their petrogenesis (Verbeeten, 1996), and have Sr and Nd isotopic ratios that fall within the high-87/86Sr, low-143/144Nd part of the Pacific MORB field, and Pb isotopic ratios that fall on or just below the NHRL at 206/204Pb = 18.646-18.679. 4: The Hunter Ridge emerges at its northern end as the Kadavu island group in Fiji. Lavas <3 m.y. old in this group include a massive volume of adakitic andesites and dacites, and limited amoimts of unusual low-Fe, high-Na basalts (<0.5Ma) that form the Ngaloa Group, also with an adakitic signature, but having abundant Fo89-91 olivine phenocrysts and extremely oxidized Cr-spinels (Cr#=70-75). Elsewhere in Fiji at this time (2-OMa), magmatism is characterized by large shield volcanoes composed of OIB-like magmas reflecting the fact that Fiji had rotated away from the Tonga (Vitiaz) subduction system to an essentially intraplate setting (Gill and Whelan, 1989). The adakitic andesites and dacites have been subdivided into two suites with slightly different major and trace element compositions (Verbeeten, 1996); Si02 contents range from 56-66% and both suites have highK 'calc-alkaline' compositions, but lavas of the younger, more adakitic Western Kadavu suite are notable for their very low FeO'^ (3-6%), high Sr (800-1600ppm), low Y (mairJiy 10-15ppm) and Sc (816ppm) contents, and significant HREE depletion (La/Yb]sj= 13-23). The high-Nb basalts have been dated at 0.36±0.05Ma (Gill and Whelan, 1989); at 7-8% MgO they have 50-51.5% Si02,1.4-1.7% Ti02, 6.1-6.8% TeO*, 3.7-4% Na20,1.65-1.7% K2O and 0.7-0.8% P2O5,. Trace element features of the highNb basalts include remarkably high Sr (2000-3000ppm), high Zr (210-255ppm) and Nb (7.4-16.7ppm), lowY (15-22ppm) and Sc (13-23ppm), and strong HREE depletion (La/YbN=18-27). It is considered therefore that the Ngaloa Group high-Nb basalts were produced by small degrees of partial melting at low pressures (<15kbar) of a peridotitic mantle that had been modified by interaction with eclogitederived slab melts, although it cannot be ruled out that these basalts formed by extensive reaction of adakitic slab melts with mantle wedge peridotite during ascent to eruption. Sr, Nd and Pb isotopic compositions of the Western Kadavu adakitic andesites and dacites, and the Ngaloa Group high-Nb basalts are essentially identical, and show no involvement of subducted pelagic sediment, with Pb isotopes plotting on or just below the NHRL at 206/2^4pb=18.76-18.83. In summary, magmatism during the last 3 m.y. along the Hunter Ridge is dominated by low-Fe basalts, high-Mg andesites and adakitic andesites and dacites, demanding a role for slab melts in their petrogenesis, and implying in turn an abnormally hot mantle wedge or subducting slab. The preferred model for the production of these imusual magmas is initiation of subduction 7-5Ma on the Himter Ridge, with the downgoing S Fiji Basin slab being subducted into very young (<5m.y. old), hot backarc lithosphere of the N Fiji Basin. Radiogenic isotope data for these Hunter Ridge lavas indicate compositions analogous to Pacific MORB. Limited occurrences of arc tholeiites and lavas transitional to high-Ca boninite compositions are restricted to the south-central part of the Hunter Ridge, and may reflect propagation of the N Fiji Basin spreading ridge into this area, leading to dehydration and subsequent magma genesis at abnormally shallow levels. The OIB basalts erupting in the Koro Sea region of Fiji since 3 Ma show some important compositional differences from typical ocean island intraplate basalts, especially their lower FeO* and Ti02 contents. We suggest that these basalts reflect asthenospheric upwelling associated with the actively spreading Lau Basin to the east and the easterrunost spreading centre in the N Fiji Basin to the west, with upwelling being focussed into the Koro Sea region at the termination of the subducting slab beneath the Hunter Ridge. It is possible that this asthenosphere upwelled over the subducted slab and provided some of the heat required to induce partial melting and production of the extensive adakitic magmas in this region.

ACKNOWLEDGEMENTS We thank Jon Woodhead for providing the Pb-Nd-Sr isotopic analyses for 4 HMA lavas. REFERENCES Auzende, J.M., Pelletier, B. & Eissen, J-P., 1996. The North Fiji Basin: geology, structure and geodynamic evolution. In: Taylor, B. (ed.) Backarc basins: tectonics and magmatism, Plenum Press, New York, 139-175. Gill, J.B. & Whelan, P., 1989. Post-subduction ocean island alkali basalts in Fiji. Journal of Geophysical Research, 94,4561-4578. Maillet, P., Monzier, M., Eissen, J-P. & Lovat, R., 1989. Geodynamics of an arc-ridge junction: the case of the New Hebrides arc - North Fiji Basin. Tectonophysics, 165,251-268. Monzier, M., Danyushevsky, L., Crawford, A.J., Bellon, H. and Gotten, J., 1993. High-Mg andesites from the southern termination of the New Hebrides island arc (SW Pacific). Journal of Volcanology and Geothermal Research, 57,193-217. Sigurdsson, LA., Kamenetsky, V.S., Crawford, A.J., Eggins, S.M. and Zlobin, S.K., 1993. Primitive island arc and oceanic lavas from the Himter Ridge-Hunter Fracture Zone. Evidence from glass, olivine and spinel compositions. Mineralogy and Petrology, 47,149-169. Verbeeten, A., 1996. Petrology, geochemistry and tectoruc implications of magmatism on the northern Hunter Ridge and Kadavu Island Group, Fiji. PhD thesis (unpubl.). University of Tasmania, 190pp.

MAGMAGENESIS IN THE LESSER ANTILLES: A SYNTHESIS Jon P. Davidson^ Matthew, F. Thirlwall^ 1. Department of Earth and Space Sciences, UCLA, Los Angeles, CA 90095-1567, USA 2. Department of Geology, Royal Holloway and Bedford New College, University of London, Egham, Surrey, TW20 OEX, UK

The Lesser Antilles (Figure 1) is one of only two island arcs in the Atlantic Ocean Basin, and is formed by the westward subduction of Cretaceous/Jurassic Atlantic lithosphere beneath the Caribbean Plate. The islands form a simple chain with no significant back-arc or fore-arc volcanism. The only complexity is produced by a westward shift in the location of the northern part of the arc which probably occurred between 27 and 10 Ma. As a result, the majority of the arc mass in the north does not lie beneath the active volcanic islands. To the south the currently active and older arcs are superimposed, so that the central islands represent a more protracted period of volcanism - extending back to the Eocene - and have developed a more substantial arc platform. Q

\ ^ ^ The Qmll

ySaba ^ ^ ^ ^ ^ \ Statia ^ ) St. Kitt?^t ^ A Redonda \ ^ VMonserrat ^ Guadeloupe

150

le^N-

' M t Pelee Martinique

St Lucia' ^^StVincentlf ^ Y Grenadines •

DSDP Hole 543

Barbados

§ ( o rGrenada^ e

12'N .

Older arc Currently active arc 60^W

.Venezuela

Figure 1. Map of the Lesser Antilles GENERAL CHARACTERISTICS OF LESSER ANTILLES MAGMAS There is a general impression that the chemistry of erupted magmas varies from tholeiitic through calc alkaline to alkaline in passing from north to south along the arc (Brown et al., 1977; Turner et al., 1996). In fact chemical variations are more complex, specifically; 1. The "tholeiitic" affinity of rocks from the northern islands is, at best, weak. On standard discrimination diagrams most would be considered calc alkaline, although the relatively flat REE patterns might be described as island arc tholeiitic. The northernmost island of Saba bears little resemblance to any definition of "tholeiitic". Furthermore, both tholeiitic and calc alkaline suites are encountered on the older islands, such as St. Martin, in the northern part of the arc (Davidson et al., 1993). 2. The diversity of magma chemical and isotopic compositions in the central part of the arc (Martinique and St. Lucia) is far greater than in the recently active islands to the north. Almost the entire spectrum of chemical and isotopic compositions from the central and northern islands is represented in a single island, and does not appear to follow systematic changes through time. 13

3. The island of St. Vincent is characterized by isotopic and incompatible element compositions similar to those encountered in the northern part of the arc (although St. Vincent magmas are much more primitive) refuting any claim that a simple systematic along-arc chemical variation exists (Thirlwall et al., 1986). 4. Grenada has long been notorious in producing high MgO and mildly alkaline basalts. With differentiation, however, a fairly unremarkable calc alkaline suite is developed. Parts of the Grenada volcanic suite (e.g. the Sr- and LREE-enriched "C-series") are unique to that part of the arc. Nevertheless picritic magmas with relatively flat REE patterns do exist in Grenada and St. Vincent, which may represent potential general parent magmas for differentiates erupted elsewhere along the arc. DIFFERENTIATION OF LESSER ANTILLES MAGMAS In cases where sufficient data exist from a single volcanic edifice, such as The Quill on Statia or Mt. Pelee on Martinique, fractional crystallization can account for most of the chemical variation in major elements. Least squares models for both volcanoes yield best fit solutions involving subequal amounts of plagioclase and amphibole, along with minor oxide. This phase assemblage is inferred to have crystallized at crustal levels, and is consistent with trace element observations, such as decreasing Sr contents and increasing La/Sm with differentiation. The observed phenocryst phase assemblage is typically amphibole free until dacitic compositions are produced, although amphibole-plagioclase cumulate blocks are common. This suggests a polybaric differentiation history with chemical signatures largely inherited in response to plag-amphibole crystallization, followed by ascent into a shallow level subvolcanic chamber, above the amphibole stability depth, in which plag and pyroxene crystallized immediately prior to eruption. T

I ' ' ' I ' ' ' I • '

^ DSDP sediments:

• Central L. Antilles ° Northern L. Antilles O The Quill (Statia) • Mt Pelee (Martinique) Surface sediments

iMantle (N

18.9 50

SiO.

. I . .

18.4 18.6 18.8 19.0 19.2 19.4 19.6 19.8 20.0 2 0 6 p b / 2 0 4 p b Figure 2. ^^'^Pb/^^'^Pb - 206pb/204pb isotopes, showing Lesser Antilles volcanic rocks (excluding Grenada) and local sediments. MORB-like mantle is indicated, consistent with MORE analyzed from a DSDP hole to the west of the arc. The Quill and Mt. Pelee data are emphasized, and the inset indicates that these trends are the result of open system differentiation. Back extrapolation of the differentiation trends does not converge on a reasonable common parent. Sediment-mantle mixtures are unable to produce the most radiogenic volcanic Pb isotope compositions. Furthermore, in order to produce parental Lesser Antilles lavas, the sediment component must vary in composition along the arc, becoming less radiogenic northward. When data from the arc are considered as a whole, diffuse correlations exist between degree of differentiation and isotopic composition (Sr, Nd, Pb, O), inasmuch as the greatest range of isotope ratios is encountered in the most differentiated rocks. In the most isotopically extreme samples, radiogenic isotope ratios can be produced by mixing of sediment into the source, but very high values (confirmed by new laser fluorination mineral analyses) indicate crustal contamination. Closer examination of cogenetic suites such as the Quill and Mt. Pelee indicates that strong correlations actually exist, underscoring the importance of open system processes in modifying chemical compositions during differentiation. Even the Grenada picritic lavas, which are considerably 14

more primitive than rocks from the central and northern islands, show strong evidence for assimilation-fractional crystallization when examined in detail (Thirlwall et al., in press). Can all central and northern Lesser Antilles magmas have evolved through variable open system differentiation from a common primary magma? Two arguments suggest that this is unlikely; 1. Back extrapolation of differentiation trends - particularly Pb-Pb isotope trends which are constrained to be linear - do not converge on realistic parental compositions (Figure 2). 2. Incompatible element ratios such as Ba/La are constant during differentitation but distinct between volcanoes. SOURCE

CONTRIBUTIONS

Despite the demonstrable effects of open system differentiation in the arc crust, it is still possible to evaluate the relative contributions of slab (fluid ± sediment) and mantle to primary Lesser Antilles magmas. The contribution of sediments is best constrained using sediment-dominated elements such as Pb and LREE, while that of the fluid is reflected in the systematics of the most fluid-mobile elements such as B and LILE. The nature of the mantle wedge is not well constrained. The distribution of HFSE, which are not expected to be significantly modified by either fluid or sediment contributions, suggests that it is broadly comparable with MORB source (Thiriwall et al., 1994). Lower abundances of HREE and Zr in least evolved Lesser Antilles rocks compared with N-MORB suggest that the source may have been melted more extensively than at ridges, or may have undergone additional depletion, perhaps through melt extraction in the Grenada Basin to the backarc. The subducted sediment component at the Lesser Antilles is better constrained than at many other arcs by samples from DSDP holes just outboard of the trench, and by extensive piston coring of forearc surface sediment. Simple mantle-sediment mixing models cannot 1. Reproduce radiogenic isotope trends 2. Produce relative enrichments in Ba, K and Sr (e.g. high Ba/La) 3. Produce the observed Pb isotope characteristics (Figure 2) 4. Produce observed B isotope systematics The addition of a fluid derived from the subducted slab (in large part the variably altered ocean crust) is implicated by 1. LILE excesses that do not appear to be correlated with particularly high Pb or Sr isotope ratios 2. Heavy B-isotope compositions that are unlikely to be derived from sediment (subducted or crustal), but may well be derived via fluids from the altered oceanic crust. For those volcanic centers where the effects of differentiation are reasonably well-constrained, Ba/La and, to a lesser extent, La/Zr are nearly constant. They can be used to illustrate source differences (Figure 3). Note that none of the Lesser Antilles data fall between mantle and sediment compositions, requiring preferential Ba enrichment, suggested due to a slab derived fluid. Addition of sediment to a fluid-enriched source would lower Ba/La, 6B and 143Nd/l^Nd, and increase La/Zr, 87sr/86sr and 206pb/204pb, 207pb/204pb, 208pb/204pb. As yet unresolved is the mechanism of sediment and fluid enrichment. The latter appears to be characterized by Th-isotope disequilibrium, indicating that it took place shortly before magma generation (probably < 100,000 years; Turner et al., 1996). Notwithstanding the complexities introduced by crustal contamination, the subducted sediment contribution appears to be greatest in Martinique, St Lucia and Grenada. This produces an unsettling observation that those samples with the greatest inferred source sediment component are also subjected to more extensive crustal contamination. The primary magmas from the central part of the arc, by virtue of greater sediment source enrichment and arguably lower degrees of partial melting due to a relatively lower fluid component might be expected to be less sensitive to crustal contamination.

• Central L. Antilles o Northern L. Antilles OThe Quill (Statia) • Mt Pelee (Martinique) T

(Turner e/a/. 1996) The Quill (& N. Antilles)

35 Fluid? 30

Mt Pelee

e^W^Th)

Sediment

10 5

increasing LREE increasing K, Rb decreasing '^^Nd/'^Nd

I •••• I

•

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40

La/Zr Figure 3. Ba/La vs. La/Zr, illustrating the roles of slab-derived fluid and sediment added to the mantle source. Neither ratio is significantly affected by intra-crustal differentiation at The Quill and Mt Pelee (highlighted). Samples from the northern islands showing the most significant effects of fluid enrichment are also characterized by Th-isotope disequilibrium (inset cartoon after Turner et al., 1996), indicating that fluid fluxing occurred recently and probably acted as the trigger for partial melting. REFERENCES Brown, G.M., Holland, J.G., Sigurdsson, H., Tomblin, J.F. and Arculus, R.J., 1977, Geochemistry of the Lesser Antilles Volcanic Island Aic.Geochim. Cosmochim. Acta, 41, 785-801. Davidson, J.P., Boghossian, N. D. and Wilson, B.M., 1993. The Geochemistry of the Igneous Rock Suite of St Martin, Northern Lesser Antilles. J. Petrol, 34, 839-866 Thirlwall, M.F., Smith, T.E., Graham, A.M., Theodorou, N., Rollings, P., Davidson, J.P. and Arculus, R.J., 1994, High Field Strength Element Anomalies in Arc Lavas: Source or Process? J. Petrol, 35, 819-838 Thirlwall, M.F., Graham, A.M., Arculus, R.J., Harmon, R.S. and MacPherson, C.M., (in press) Resolution of the Effects of Crustal Assimilation, Sediment Subduction and Fluid Transport in Island Arc Magmas: PbSr-Nd-0 Isotope Geochemistry of Grenada, Lesser Antilles. Geochim. Cosmochim. Acta Thirlwall, M.F., Davidson, J.P. Graham, A.M., Smith, T.E., and Harmon, R.S., 1986. Nature and origin of inter-island chemical and isotopic variation in the Lesser Antilles. Terra Cognita, 6, 1986. Turner, S., Hawkesworth, C.J., van Calsteren, P., Heath, E., Macdonald, R. and Black, S., 1996. U-Series isotopes and Destructive Plate Margin Magma Genesis in the Lesser Antuilles. Earth Planet. Scl Lett., 142, 191-207.

IMPORTANCE OF TEMPERATURE DEPENDENT VISCOSITY AND HYDRAULIC FRACTURE ON PHYSICAL MODELS OF SUBDUCTION ZONE MAGMATISM J. Huw Davies and Andrea Rowland Department of Earth Sciences, The University of Liverpool, Liverpool, L69 3BX, U. K.

INTRODUCTION It is now accepted that water plays a critical role in generating magmatism at cold subduction zones. Detailed understanding of genesis of subduction zone magmatism is still very limited. For example it is unclear (i) how water leaves the slab, (ii) if the melting the result of compression melting, decompression melting or hydrous fluxing? This paper will concentrate initially on the thermal models that play a significant role in constraining models of subduction zone magmatism; before advancing to consider the transport mechanism of water. The thermal structures of the subducting slab and overlying mantle wedge are largely controlled by different parameters. The temperature of the subducting slab is largely controlled by the normal convergence velocity, the age of the subducting lithosphere and the level of shear heating; while, the thermal structure of the mantle wedge is controlled by the rheology and dynamics of the mantle wedge, including the shape of the mantle wedge (including the dip of the underlying slab and shape of the overriding mechanical lithosphere), and the velocity in the mantle wedge. In this paper we will emphasise (i) the ability to prepare thermal models for all subducting slabs, (ii) the importance of temperature dependent rheology in understanding mantle wedge temperatures, and (iii) to introduce the need to consider faults and fractures in understanding the transport of water out into the mantle wedge. SUBDUCTING SLAB TEMPERATURE The temperature of the subducting oceanic crust is primarily controlled by the age of the subducting lithosphere, the trench normal component of the convergence velocity, and the level of shear heating. Younger oceanic plates have higher temperatures since they are initially hotter. Lower convergence velocities normal to the trench lead to higher temperatures, due to the longer time for a material particle to reach any given depth, and greater time for conductive heating. Both these two parameters are fairly well constrained. Shear heating is less tightly constrained but most evidence points towards a low value. Roy den [1993a,b] has developed an analytic model which can allow one to incorporate all the above variables to model the temperature of a subducting slab while the overriding plate is rigid. While an extension of the analytic model of Davies and Stevenson [1992] allows the modeling of the temperature in the slab and an overriding mantle wedge with induced slab parallel flow. The combination of the above two inexpensive models allows one to model slab temperatures of all the world's subducting plates, taking into account the variable down dip shape. For example one finds high temperatures in the Western Aleutians, because the convergence velocity vector is obhque and hence the normal convergence velocity is very low. One also finds high temperatures in S. Andes because the age of the subducting plate is very young. It has been argued that the geochemistry of both of these regions reflects a significant contribution of melt from the subducting oceanic crust to the the overriding magmatism. Clearly the high modelled temperatures support the possibility of melting of the oceanic crust. The subducting plate beneath Adak though in the Central Aleutians, the volcano after which supposed slab melts have been named, is not as hot, and it does seem to be underlain by seismicity which is suggestive of rather cool temperatures in the slab. Clearly the lack of melting right at the Taitao peninsula, and on Solomon Islands, where ridges are subducting, probably reflects the fact that (I) the subducting crust is still not hot enough to melt when dry, but only when hydrated, and (II) the subducted lithosphere beneath these regions was never hydrated. The large majority of subducting oceanic plates though are not hot enough to melt, and hence the melting in those cases must be in the mantle wedge. MANTLE WEDGE TEMPERATURE The temperature in the mantle wedge is controlled primarily by the velocity of the matrix in the mantle wedge. This need not be the same as the convergence velocity. The high matrix velocity can result from high absolute plate velocities, high slab roll-back velocities, as well as high convergence velocities. The shape of the subducting slab, and the thickness of the overriding lithosphere also control the temperature in the mantie wedge. Simple comer flow models show that thick overriding lithospheres push hot temperatures deeper, while very shallow dips lead to cooler wedges. Finite element models allow realistic slab shapes to be modelled. We have modelled the case of the flat subduction beneath the Andes (see Fig. 1). The restricted height of the mantle wedge limits the induced wedge velocity and leads to the overriding wedge being very cool. This explains the lack of magmatism above such regions. Another aspect which has not been investigated in much detail for the thermal structure of the mantie wedge is

the influence of the wedge rheology, especially the influence of stress and temperature dependence on the mantle wedge viscosity. Studies of the influence of stress dependence; i.e. power law rheology, using corrected analytic solutions from Tovish et al [1978] for the velocity, suggest that a power law rheology will have a minimal effect on the thermal field. In contrast, numerical studies of the influence of temperature dependent rheology show that it has a very significant effect (see Fig. 2). The first consequence of introducing temperature dependent rheology, is that one can allow the thickness of the overriding lithosphere to be determined self-consistently through its thermal structure. Since a corner flow has a small net upwards component in the wedge velocity, it advects some heat upwards and makes flow upwards easier. This process leads to erosion and entrainment of the shallow overriding lithosphere. This can effectively lead to the mantle wedge extending all the way to the surface. This extreme case is not observed at subduction zones. This is probably because the buoyancy of the overriding crust prevents it from being entrained into the descending flow. This prevents the ascending flow reaching all the way to the surface. If this interpretation is correct, then the thickness of the overriding mechanical lithosphere above subduction zone mantle wedge will be the same as the thickness of the overriding crust. Plank and Langmuir [1988] made this identification implicitly. The sensitivity of the mantle wedge temperature to the convergence velocity decreases, compared to a constant viscosity wedge case. Since the matrix flow has a significant upward component to its flow this allows the possibility for some degree of decompression melting. Hydrous fluxing though must generally dominate, since the thermal structures would lead to little or no melting if the peridotite was dry. Therefore the critical question is, how does the water leave the slab and enter the mantle wedge? W A T E R TRANSPORT FROM THE SLAB Watson and Brenan's [1987] work suggests that hydrous fluids cannot cross cold olivine dominated wedge by porous flow. As pointed out by Marsh, diffusion is far too slow. So the question still remains, how does water leave the slab and cross the cold part of the mantle wedge (Davies, 1994)? The answer I speculate might be related to the answer of another subduction zone puzzle. What is the cause of intermediate depth earthquakes? Earthquakes occuring beneath around 60km depth are a mystery because of the high confining pressure. The rocks are not strong enough to sustain the shear stress required to overcome the high firictional forces that the high pressure/normal stresses entail. The deepest earthquakes in subduction zones; those occuring beneath 300km depth are believed to result from transformational faulting in the kinetically delayed phase change fom a to y olivine. Earthquakes from 50 to 300km depth need to be explained by another mechanism. One of the preferred suggestions is that this is the result of high pore fluid pressure, which acts to reduce the normal stress and reduces the friction, allowing reasonable shear stresses to lead to faulting [Green and Houston, 1995]. If this is the case then there are two consequences, (i) water is required to be present in the slab down to 300km depth, and (ii) that water is present around active faults. Earthquakes occur at all magnitudes, suggesting an interacting network of faults. The vast majority of their focal mechanisms have their most compressive, or least compressive axis in the down-dip direction. This requires the fault plane to be at an oblique angle to the slab surface. Before faults fail they dilate and fill with the surrounding pore water. Hence faults will locally collect pore water, and following faulting will channelize the water upwards. Faults will interact and form a network, hence water can be carried to, and collected near the slab surface. If sufficient water collects, then the water itself can cause hydraulic fractures, which will propagate perpendicular to the least compressive stress, which will be in a direction out into the wedge [Davies and Stevenson, 1992]. These fractures will propagate a distance which will depend upon the volume of water collected. The more water collected the further the cracks can propagate, since they can maintain the critical driving pressure for longer. The driving pressure reduces with propagation, since some of the water will be lost from the fracture, due to physical escape and chemical reactions. Since the initiation of hydraulic fractures is dependent on local conditions at the point of initiation, hydraulic fractures with different volumes can be expected to leave the same slab. Some might be small, and effectively stop in the sediment, leading to melting of the sediment. Others might stop in the sub-solidus mantle wedge, and react to give hydrous minerals. This water can continue its journey out to the mantle wedge as a result of combined free flow and matrix flow fixed in the mantle minerals [Davies and Stevenson, 1992]. The biggest hydraulic fractures could continue until they reach the hottest part of the mantle wedge; where they stop and induce massive melting directly and quickly. This melt could also quickly segregate; hence the time scale from release of water from the slab, to the emplacement of melt in the crust could in those cases be very short. The proportion of water that reaches the source region by the various routes is unfortunately poorly predicted by this paradigm since fracture initiation at depth is poorly constrained. The number/energy of earthquakes drops exponentially with depth, with a minimum around 300 - 400km depth. Similarily magmatism decreases exponentially towards the back arc, away from the volcanic front. The

hydraulic fractures following the collection of sufficient water near the slab surface could be the seismic sources that are observed at around 100km depth in some subduction zones with non double-couple mechanisms and lead to the b value of the Gutenberg-Richter relation being different from its typical value of =1. This mechanism has many implications. When the water is transported rapidly to the source region, any partitioning of short-lived isotopes will lead to radioactive disequilibria in the source region. Rapid transport to the source region would also lead to the melting occuring essentially as a batch melting experiment; i.e. the peridotite source around the water rich dyke melts extensively and there is a limited contribution from small degrees of melting; this could partly explain the depleted nature of many island arc magmas. The probably more common slow transport of water would lead to the system returning to radioactive equilibrium following any disequilibria due to dehydration. Any hydraulic fracture stopping in the sediment would lead to water saturated melting of the sediment, while it might undergo dehydration melting if there was no excess water. It is therefore unclear what thermal constraint the geochemical identification of a sediment melt component in the source would imply. Does it imply that the slab temperature was greater than the dehydration melting solidus or the water saturated solidus? The mechanism could also lead to very heterogeneous melting, with some regions being melted extensively with other regions suffering little or no melting. If the slab is hotter, then the water could leave the slab as a melt, and cross the mande wedge by porous flow, since melts can form interconnected networks. The melt is less likely to form fractures if it can continue to propagate by porous flow. Since the temperatures will continue to increase as the melt moves through the upper slab and mantle wedge, it will not freeze but will react with the surrounding material. There is therefore a strong possibility that the temperature of the subducting slab, can play a large role in the nature and path of the phase that carries water out into the hot parts of the mantle wedge. This can be significant in controlling the nature of the geochemical slab signature in the observed magmatism. This proposal that water could be leaving the slab either in hydraulic fractures of various size when cold, and in melts when hot; should be tested/investigated further. It would suggest that there should be a change in the slab component, especially when slabs gets so cool that they cannot melt.

DYNAMICS, DEPLETION AND ENRICHMENT

The geochemistry of island arc magmatism frequently suggests a depleted source, while continental arcs frequently suggest an enriched source, relative to the N-MORB source. Depletion in oceanic settings could result from previous melting at ridge or back-arc basin. Continental lithosphere in contrast could be enriched because they have been around a long time and collected small degree melts at their base. For this enrichment to enter the source requires the lithosphere to be entrained into the source region. This effect of entraining lithosphere is observed during simulations utilising temperature dependent rheology. The entrained lithosphere could also come from the back arc lithosphere in addition to the lithosphere directly overriding. Flow reversal; i.e. where the matrix flows in the opposite direction to the flow induced by the slab; can lead to the depletion of mantle source through multiple episodes of melting. This mechanism of depletion does not require recourse to other external mechanisms for source depletion. Factors that control flow reversal include local sources of buoyancy such as hydration, melt and residue; and degree of coupling between wedge and slab. Flow reversal cannot dominate the system though, since it leads to closed convection cells in the mantle wedge corner. These closed cells get ever cooler with time, as they are passed beneath by ever greater amounts of cold subducting slab. The system would eventually become too cool to generate melt. Most subduction zones have continuously active magmatism on a geological time scale, hence do not allow significant periods of magmatic quiescence. This suggests that flow reversal is likely to be only a weak modulating effect.

REFERENCES

Davies, J. H., and D. J. Stevenson, Physical model of source region of subduction zone volcanics, J. Geophys. Res,, 97, 2037-2070, 1992. Davies, J. H., Lateral water transport across a dynamic mantle wedge : A model for subduction zone magmatism, in "Magmatic, Systems", edited by M. P. Ryan, 197-221, 1994. Green II, H. W., and H. Houston, The mechanics of deep earthquakes, Ann. Rev. Earth Planet. ScL, 23, 169213, 1995. Plank, T., and C. H. Langmuir, An evaluation of the global variations in the major element chemistry of arc basalts. Earth Planet. Sci. Lett., 90, 349-370, 1988. Royden, L. H., The steady state thermal structure of eroding orogenic belts and accretionary prisms, J. Geophys. Res., 98, 4487-4507, 1993a. Royden, L., Correction to "The steady state thermal structure of eroding orogenic belts and accretionary prisms", J. Geophys. Res., 98, 20039, 1993b.

Tovish, A., G. Schubert, and B. P. Luyendyk, Mantle flow pressure and the angle of subduction : NonNewtonian comer flows, J. Geophys. Res., 83, 5892-5898, 1978. Watson, B. E., and J. M. Brenan, Fluids in the Lithosphere, Part 1 : Experimentally-determined wetting characteristics of CO2 - HjO fluids and their implications for fluid transport, host-rock physical properties and fluid inclusion formation. Earth Planet. Set. Lett., 85, 497-515, 1987. FIGURE 1 Steady state thermal structure of the subducting slab and mande wedge in Peru. Depth of whole model is 200km. The contours are labelled in terms of non-dimensional temperature; and need to be multiplied by your preference for ambient mantle potential temperature. The contours are every 0.1. Notice that the part of the mantle wedge above die flat subducting slab is very cool, due to the limited flow of asthenosphere in this region. Therefore no magamtism would be expected above the flat part of the subducting plate, as observed.

FIGURE 2 Steady state thermal structure of a constant dip subduction zone with a convergence velocity of 9cm/yr, and a 7km thick rigid overriding layer (to mimic an unsubductable oceanic crust), with a temperature dependent mantle wedge rheology. The activation energy is based on experimental values for olivine. The thermal lithosphere is 100km thick on both sides. Notice how the flow removes the thermal lithosphere in the wedge comer, leading to a hot wedge rising up towards the trench. The contours are in non-dimensional temperature, and are spaced 0.1 apart. To convert to mantle temperature in "C, multiply by your preferred ambient mantle potential temperature in "C.

Acknowledgments A. Rowland acknowledges the support of an NERC studentship.

Magmatism in West Sulawesi, Indonesia: Geochemical characteristics and economic implications. Marlina A. Elburg and John Foden, Department of Geology and Geophysics, Adelaide University, Adelaide SA5005, Australia The western part of the island of Sulawesi (Indonesia) consists of an arc which was active from the Paleogene to Early Miocene; the magmatic activity is thought to have been caused by west-directed subduction. The basement underneath the arc changes from continental in the south to oceanic in the north (Fig.l). The microcontinents of Sula and Buton collided with the arc in the Middle to Late Miocene, throughout which period regional magmatism continued. The setting and tectonic history of the area gives us an unique opportunity to study both temporal and lateral variations in magma geochemistry. This enables us to better understand how magma geochemistry refelects the contrasting role of source elements (including oceanic crust versus continental crust or lithospheric mantle) compared to the influence of changes in tectonic setting (subduction, collision and post-collision). Fig, 1: Schematic map of the island of Sulawesi (Indonesia). The diagonally striped pattern indicates which areas are underlain by continental crust. The crust underlying southwest Sulawesi belongs to Sundaland, whereas Buton and Banggai-Sula are supposed to belong to New Guinea (Vroon et ai, 1996).

The oldest rocks exposed in the southern and central parts of the arc are Jurassic to Cretaceous sediments; those in southern Sulawesi have Pb and Nd isotopic signatures unlike sediments from Australia or New Guinea (Vroon et al., 1996), suggesting that this part of Sulawesi belongs to Sundaland. The basement of the north arm is composed of Eocene to Oligocene submarine volcanics and sediments (Kavalieris et al., 1992).

• X O A

X X

Quatemary S.Sulawesi Miocene S. Sulawesi Fre-Miocene S. Sulawesi N. Sulawesi

•

Fig. 2: Si02 versus Kp diagram of samples from S. and N. Sulawesi. Rocks from N. Sulawesi have invariably lower K^O contents than those from S. Sulawesi.

•

O * *

O A

fA 45

Si02

Miocene to Quaternary igneous rocks in south, central and northern Sulawesi show marked geochemical distinctions. Those in northem Sulawesi have an oceanic arc signature, with mainly intermediate compositions and low potassium content (Fig. 2). Extrusive and intrusive rocks in central Sulawesi are bimodal with resepect to silica. Potassium contents are invariably high. The rocks from southern Sulawesi have geochemical characteristics which are intermediate between those from northem and central Sulawesi. The high-K signature is accompanied by marked enrichments in incompatible elements such as Ba, Nb, U, Th and Pb (Fig. 3), but also in the compatible elements Cr and Ni. 21

The rocks from central Sulawesi are also extremely enriched in radiogenic Sr and unradiogenic Nd (Fig. 4), with isotopic signatures much more extreme than those of sedimentary rocks in the area. Crustal contamination is unlikely to be the cause of the enriched signatures, as many of the rocks are very mafic (up to 900 ppm Cr) and have high Sr and Nd contents. South Sulawesi only shows moderate enrichments in isotopic compositions, which are less extreme than those observed in e.g. the Banda Arc. lOOO.OOr Quatemary S. Sulawesi Miocene S. Sulawesi Pre-Miocene S. Sulawesi Miocene C. Sulawesi Miocene N. Sulawesi

lOO.OOT O

s 10.00+ £

Fig. 3: N-MORB normalised trace element patterns for rocks from south, central and north Sulawesi. Regional as well as temporal shifts in trace element geochemistry can be observed

1.00" 0.10 Rb Ba Th U Nb K La Ce Pb Sr P Nd Zr Ti Y

Temporal variations in magma geochemistry are superimposed on these regional variations. In southem Sulawesi, pre-Miocene rocks have a calc-alkaline signature, which evolves to a high-K calc-alkaline signature in Miocene times, whereas Quatemary volcanics have lower potassium contents (Fig. 2). These variations are matched by changes in radiogenic isotopes, with the Miocene rocks having the most enriched Sr and Nd isotopic signature, the pre-Miocene rocks the least enriched signature (Fig. 4). Trace element behaviour does not follow this pattern. The Quatemary rocks show the greatest enrichments in incompatible elements such as Nb, Zr and Hf (approximately 4 times higher than NMORB) whereas the Miocene and pre-Miocene rocks have levels of these elements similar to NMORB. 0.51320.5130 • 0.5128

•

051265 _

Z 1 0.5126

0.7055

T3 Z

0.7065

2 0.5124

Fig. 4: Initial Sr and Nd isotopic compositions of samples from central and southem Sulawesi. Data from central Sulawesi from Bergman et al (1996). The extreme enrichments observed in the rocks from cetral Sulawesi may be related to the presence of ancient lithosphere underneath this area.

0.5122 0.5120 0.5118 •

0.70

1—

0.71

0.72

0.73

87Sr/8^Sr

0.74

The cause of the spatial and temporal geochemical variations is presently under investigation. The oceanic signature in the north and the continental signature in the central parts of Sulawesi is likely to reflect the presence of oceanic and continental basement (cmst or mantle) respectively. However, it is as yet unclear why central Sulawesi shows these unusually enriched isotopic compositions compared to the southem part of the island. The southem 'isotopic boundary' appears to lie near the city of ParePare, but no major faults or cmstal discontinuities have been reported in this area. Several different scenarios could explain the observed varations. Given their very large isotopic and geochemical, south and central Sulawesi must reflect contrasting cmstal terranes with verv significant aee differences. Altemativelv.

the geochemical signature of the Miocene and Quaternary volcanics may reflect the composition of the material subducted prior to and during continental collision. Finally, the 'continental' signature in central Sulawesi could reflect accretion of continental material during a previous event, either in the early Miocene (Bergman et al., 1996) or mid Cretaceous (Wakita et al., 1996). These options will be tested by sampling of pre-Miocene rocks in central Sulawesi, as well as by Pb isotopic analysis of the samples. The fact that the trace element pattems of the southern Sulawesi samples appear to become more enriched after the collision event may point towards a greater contribution of the continental lithosphere to magma genesis. It is most likely that this is related to post-collisional extension within the former arc, but an alternative option is that continental lithosphere underlying Buton may have been subducted. Finally, the possibility that we just see a greater contribution of OIB-like mande towards magma genesis a ^ r cessation of subduction and potential slab break-off cannot be discounted either. North Sulawesi contains well-known mineralised areas, with Cu-Au porhpyries and epithermal gold as the most important types. It is as yet uncertain whether the mineralization is related to subduction, or to intra-arc extension (Kavalieris et al., 1992), as both the age of the mineralization and the age of the associated rocks is often poorly known. However, the geochemical signature of the Miocene andesites which host epithermal gold in the Ratatotok district, is typical of a subduction setting (Fig. 3). Cu-Au porphyries in the Tombulilato area are associated with 2-4 Ma volcanics (Kavalieris et al., 1992), but the exact geochemical composition of these volcanics is not known. In contrast to the oceanic arc of northern Sulawesi, only few mineral deposits are known in the continental arc. A porphyry molybdenum deposit is present at Malala, in the 'neck' of Sulawesi. It is associated with granite and quartz monzonite porphyries, which are approximately 4 Ma old. The relatively high initial ^^Sr/^^Sr ratio of these granitoids reflects the continental basement. A small porphyry copper deposit is present at Sasak, in central Sualwesi, which is associated with alkaline mafic rocks. Although Solomon et al. (1990) have suggested that gold mineralisation is associated with reversal of subduction and melting of more refractory mantle materials, there is litde evidence that this setting accounts for gold mineralisation in N. Sulawesi. Levels of incompatible elements in the igneous rocks rather appear to increase than decrease with time, which is contrary to the scenario proposed by Solomon et al. Few deposits have been discovered in central and south Sulawesi, but prospects are thought to be good within the alkaline central sector, as high-quality Cu-Au deposits are often associated with alkaline (potassic) magmatism (e.g. Porgera and Ok Tedi in Papua New Guinea). The presence of Cu-Au deposits is generally dependent upon the availability of hydrothermal fluids of the appropiate composition (with high levels of salinity), and investigation are currently under way to determine chlorine levels of the central Sulawesi igneous activity. Bergman, S. C., D. Q. Coffield, J. P. Talbot and R. A. Garrard. (1996). Tertiary tectonic and magmatic evolution of westem Sulawesi and the Makassar Strait, Indonesia: evidence for a Miocene continent-continent collision. Tectonic evolution of Southeast Asia. London, Geological Society of London, 391-430. Kavalieris, L, T. M. Van Leeuwen and M. Wilson. (1992). "Geological setting and styles of mineralization, north arm of Sulawesi, Indonesia." J. Southeast Asian E. Sci. 23(113-129): Solomon, M. (1990). "Subduction, arc reverals, and the origin of porhyry copper-gold deposits in island arcs." Geology. 18: 630-633. Vroon, P. Z., M. J. Van Bergen and E. J. Forde. (1996). Pb and Nd isotope constraints on the provenance of tectonically dispersed continental fragments in east Indonesia. Tectonic evolution of Southeast Asia. London, Geological Society of London, 445-454. Wakita, K., J. Sopaheluwakan, K. Miyazaki, 1. Zulkamain and Munasri (1996). Tectonic evolution of the Bantirnala Complex, South Sulawesi, Indonesia. Tectonic evolution of Southeast Asia. London, Geological Society of London, 353-364.

ARC MAGMATIC CHEMISTRY: THE SOURCE VERSUS PROCESS DEBATE J.Foden Department of Geology and Geophysics, University of Adelaide, Adelaide, SA 5005 e .mail: jfoden @ geol ogy .adelaide .edu .au

ARC MAGMA AS GEOCHEMICAL MIXTURES Geochemical studies of arcs world-wide, reveal complexities in the signatures of island arc magmas that require the involvement of three or more source components. If the true role of arc magmatism in the evolution of the geochemical character of the continental crust is to realistically assessed, then the arc magma's sources need to be accurately determined. This isolation of the individual influence of each of these has long been a challenge. Most studies easily demonstrate mixing trends in isotope-isotope space that originate with a MORE component (see figure below). Some workers also debate that this may have more depleted character than N-MORB (Woodhead et al, 1993). A slab-derived and possible hydrous fluid-dominated component is also well defined in some studies. This is best observed in arcs in which the magmas are otherwise highly lithophile-element poor, such as the Tonga-Kermadec arc. This component is only moderately LIL element-enriched with very high ^Sr/^Sr and ratios, low Ce/Pb and low ^^^Nd/^Nd. It is also characterised by high U/Th and with U-series disequilibrium to the right of the equiline. A third and important geochemical vector is one particularly recognised in arcs in complex tectonic settings in the Mediterranean, the Aegean and in Indonesia (see figures below) and is one which leads to the development of K20-rich suites. Features include silica-undersaturation, very high K/Nb, Ba/Nb and Ce/Y with ^Sr/^Sr values which are moderately low at moderately low ^^^Nd/^^Nd values. On 87Sr/86Sr versus ^'^^Nd/^'^Nd diagrams these show mixing lines towards a significantly less radiogenic Sr endmember than the MORB-sediment mixture. Magmas enriched in this component don't tend to show U-series disequilibrium and have low ^^h/^-Th and ^^U/^'Th values. They appear to have k^h values that reflect a source with a long term history of high Th/U. The Nd- and Sr-isotopic characteristics imply a source with a moderately long history of LREE enrichment at moderate Rb/Sr and U/Pb This may be an OIB component, though it seems to differ from any of the known EM- or HIMU OIB endmembers defined from Ocean Island studies, in having some extreme incompatible element characteristics including high K/Na, K/Nb, K/La and Ba/Nb (see figure below). These may of course be due fractionation processes unique to arc setting. The coupled isotopic and trace element characteristics of this endmember distinguish it clearly from those of a melt component derived from the subducted sediment- and alteration- enriched upper slab. In its K-rich and LREE-rich character this endmember does closely resemble melts of enriched sub-continental lithospheric mantle, an observation supported by the settings of subduction provinces which host these suites in the vicinity of fragmented continental lithosphere (the Aegean, Indonesia, Italy). MORB

0.5132 T

Ba/Nb

/Mixing trend towards high-K , end member EASTERN 3UNDA ARC

RINDJANI

2 0.5128 9 2 pores aUotr^DoK

BULK EARTH BAN DA ARC

LEUCITITES^ High-K ^ __ Source Region

-4-

Tsf 10 15 20 25 30 35 40

0.5120-

Zr/Nb

0.700

Eastern Indonesian Sunda Arc Volcanics

0.705

0.710

-I

0.715

Sr / Sr

0.720

THE IDENTinCATION OF THE ENDMEMBERS AND THEIR ORIGINS AND ROLES

Special sampling and analytical strategies are needed to identify and resolve separate endmembers in arc systems. These will tend to dwell on temporal or spatial variation in arcs with changing or evolving tectonic situations. 1. In the case of normal , orthogonal subduction at constant rate, the geochemical influence of the slab including its sediment input, should provide a constant background to the geochemical budget of the contemporary arc magmas. Thus if an arc shows significant along-arc variation (as does the eastern Sunda Arc), then this variation should be imposed by other factors. These could include heterogeneity in the arc crust and lithosphere, or the introduction of a "foreign" source via the subduction zone. As an example , there is astrong posibility that due to convergence of the Indo-Australian with the southern and eastern Indonesian arc systems, dispersed fragments (micro-continents) of Gondwanan continental lithosphere are being swept into this subduction zone. The resultant collisions lead to deformation and uplift and variable subduction of continental crustal and subcontinental lithospheric mantle. Because of crustal buoyancy, the continental lithospheric fragments may become delaminated, the sub-continental mantle lithosphere component becoming isolated and being carried into the mantle under the arc. Because this lithospheric mantle is potentially relatively old, it may be the site of geochemical enrichment and anomalous Sr-, Nd- and Pb -isotopic signatures, thus providing potential K-rich magma sources of both continental crustal and mantle origin. 2. Out board migration/ slab role-back and skipping of the subduction zone often leads volcanic provinces which at first were on the axis of the arc to effectively move into the back arc. For instance, in the Aegean, the active arc has migrated southwards during the Neogene extension of the Aegean lithosphere. This has led volcanic provinces in western Anatolia (Turkey) to evolve from positions on the eastern end of the main Cycladean arc, to the back arc (Robert et al, 1992). Similar effects have occurred in Indonesia, effectively removing Kalimantan , southern Sulawesi and possibly south eastern Sumatra from positions on the active arcs in the early Miocene, to present back arc positions. In each of the above examples this effective migration has lead to the loss of some arc-type geochemical characteristics, most notably decreased Nb-anomalies, in otherwise alkalic, often potassic magmas. In south Sulawesi and in Kalimantan, significant loss of arc geochemical characteristics, such as low Nb/Ce has taken place in time scales of the order of 5-10 Ma. This relative enrichment of Nb in magmas transitional to back arc settings is also mirrored in some granites provinces and prompts the question of whether Nb-anomalies are in some way a function of magmatism in compressional stress fields versus extensional settings rather than specifically arc settings. 3. The complete manifestation of backarc extension leads to the development of refractory, MORB -like compositions often with negligible subduction geochemical fingerprint (eg Woodhead et al, 1993).

DO ARCS REALLY CONTRIBUTE TO CONTINENTAL GROWTH?

The rate, mechanisms and sources of elemental fluxes to the crust in modem arcs remains a source of contention. Uncritically many workers in the geochemistry of arcs justify a place for their work in the grand scheme of things by the tacit or explicit claim that the object of this research is part of a process central to the growth and geochemical evolution of continental crust. However ifii can be shown that arcs do little more than re-cycle lithophile element budgets from the continental lithosphere back to itself, in the process diluting these elements in a dominant mass of MORB-like melt derived from the depleted mantle wedge, then it is questionable if this process does indeed represent continental crustal growth. Such a process may add new mass to the continents (perhaps at a rate of %5-10 that of MOR production of oceanic crust) but will only dilute the lithophile element budget and geochemical character of that crust, making it on average less continental. Of course if sediment recycling is the main source of lithophile elements in arc magmas then the process of continent growth by arc addition does not represent a new net gain to the lithophile element budget and character of the continents, but a clear loss. The "sediment loop" recycles eroded upper continental crustal material on a time scale of < 10® years, but as shown by sever^ studies based on flux estimates to arc magmas (Hawkesworth et al, 1991) and on the isotope and trace element chemistry of both OIB and MORB, a larger major of the sediment (>%80) is recycled to the deeper mantle with the slab, where it participates in mantle source evolution on a time scale o f 1 0 ^ years (Woodhead et al, 1993) The "andesite model" (McLennan and Taylor, 1982) which is the formalised basis for the connection of modern arc activity to continental growth has long since been shown to be fundamentally (Foden, 1983).

It assumes that andesite is a primary melt and that net crustal addition to arcs (and thence to the continental margins) is andesite, thereby matching the supposed average continental crustal composition. In reality, it is clear that modern arcs are fed by hydrous, basaltic primary magmas whose fractionation at crustal depths leads to their division into complementary lower crustal ultramafic cumulates and upper crustal differentiated melts. Under these circumstances the average bulk composition of island arc crust is mafic and differs only from crust generated at MOR or in the back arc basins by the addition of one or more of the lithophile element-rich components in the arc. O'Nions and McKenzie (1988) and others have suggested that substantial incompatible element enrichment of the continental crust may occur through anorogenic, intra-cratonic basal flux of small melts to the sub-continental lithospheric mantle. These signatures may be transported to the upper crust during phases of anorogenic mafic magmatism, but they could also participate in subduction regimes where the enriched mantle may provide a key incompatible element rich source component. The subjection of fragments of ancient continental lithosphere to fluid and magmatic flux by subduction zones may be much more significant part of the plate tectonic paradigm than previously recognised. Like sediment subduction, this geochemical return to the continents through the arc magmatic cycle may be viewed as a recycling of geochemical characteristics, not a primary generation of these. REFERENCES Foden, J., 1983. The petrology of the calcalkaline lavas of Rindjani volcano, East Sunda Arc: amodel for island arc petrogenesis. J. Petrologyy 24, 98-130 Gasparon, M., 1993. Origin and evolution of mafic volcanics of Sumatra (Indonesia): their mantle sources, and the roles of subducted oceanic sediments and crustal contamination. PhD Thesis^ University of Tasmania, Hawkesworth, C J., Hergt, J.M., McDermott, F., and Ellam, R., 1991. Destructive margin magmatism and the contributions from wedge and subducted crust. Aust. J of Earth Sciences, 38, 577-594 McLennan, S., Taylor, S.R., 1982. Geochemical constraints on the growth of the continental crust. J. of Geology, 90,347-361. O'Nions, R.K., McKenzie,D., 1988. Melting and continent generation. Earth Planet, Sci. Letts,, 90,449456 Plank, T and Langmuir, C.H., 1993. Trace elements from sediment input to volcanic output at subduction zones. Nature, 362,739-743. Robert,U., Foden, J., Vame, R. 1992. The Dodecanese Province, SE Aegean : A model for tectonic control on potassic magmatism. Lithos, 28,241-260. Woodhead, J., Greenwood, P, Harmon,R and Stoffers, P., 1993. Oxygen isotope evidence fro recycling crust in the source of EM-type ocean island basalts. Nature, 362,809-813 Woodhead, J., Eggins, S,. Gamble, J., 1993. High field strength and transition element systematics in island arc and back-arc basin basalts: evidence for multi-phase melt extraction and a depleted mantie wedge. Earth Planet, Sci, Letts,, 114,491-504

MAJOR, TRACE ELEMENT AND ISOTOPE GEOCHEMISTRY OF HISTORIC (1945 -1996) ERUPTIONS FROM RUAPEHU VOLCANO, NEW ZEALAND WITH IMPLICATIONS FOR OPEN SYSTEM MAGMATIC PROCESSES IN ARC VOLCANOES. Tohn Gamble!. Peter Wood^, Richard PriceS Tod WaightS, Ian Smith^, Mitsuhiro Nakagawa^ 1. Dept. of Geology and RSES, Victoria University of Wellington, New Zealand. 2. Taupo Volcano Observatory, Institute of Geological and Nuclear Sciences, Wairakei, New Zealand. 3. Dept. of Geology, La Trobe University and VIEPS, Bundoora, Victoria, Australia. 4. Dept. of Geology, University of Auckland, Auckland, New Zealand On 18th September 1995 Ruapehu Volcano, the largest and southernmost of the active andesite - dacite composite volcanoes in the Taupo Volcanic Zone, New Zealand, burst into eruption through a series of vents located beneath the summit Crater Lake. The spectacular explosion in the late aftemoon of 23rd September 1995, widely reported by the world news media, led to closure of ski-fields on the moxmtain, ejected Crater Lake water, tephra and scoria bombs onto the upper slopes of the volcano and sent lahars down a number of river catchments (Naim et al., 1996). Phreatomagmatic eruptions, involving interaction between magma and Crater Lake water continued with decreasing vigour until 27th September 1996. Renewed vigorous eruptions between 1114th October 1995 were ash dominated, signalling exhaustion of the Crater Lake waters. Between October and January 1996 a tholoid or spine grew in the vent region and on 17 - 18th June 1996 spectacular strombolian eruptions from the open vent system showed that the eruptive cycle was unfinished. The 1995 - 96 eruptives range in composition from medium-K andesite to dacite and span the entire range of lavas over the previous 50 years (1945 -1995), (Figure 1). All samples are porphyritic, with phenocrysts of plagioclase, orthopyroxene and clinopyroxene, typical of most Ruapehu lavas (Graham and Hackett, 1987) 1 '• • 'I ''' 'I ''• 'I ''''I ''''I ' '

7.50

7.00

1.80

6.50

1995

•

1996

1945

1966

•

1969

•

1971

1977

1975

6.00

5.50

SiO^

Figure 1. Major element compositions of the 1945 -1996 eruptives from Ruapehu Volcano, New Zealand.

In detail, the bombs associated with the initial 18th September 1995 event, contained --58% Si02, bombs erupted during the more or less continuous activity between 23rd and 25th September 1995 ranged between 58% and 63% Si02 and those erupted over the period, October 11th - 14th, and June 17th 18th 1996 eruptions contain less than 58% Si02. Oxide vs oxide plots (Figure 1) show strongly correlated, more or less linear variations, although some scatter is apparent in the K2O vs Si02 plot. Trace elements also correlate well with major elements and other trace elements and isotopes (Figure 2) overlap and extend the field of TVZ basahs (Gamble et al., 1996) to more radiogenic compositions. 0.51310 n-rj•U^l -TT-I T 1 1 M 1 1 1M" 1•1 11 11 1 1 • 1 l-l T t 1 1 I 1 1

0.51295 - r -r -7 r I > . • < 1 >. . 1 1 . 1 . . 1 1 . . 1 . o : 0.51290 0.51285 0.51280 0.51275

^ 0.51290 h

0.51300

TVZ Basalts (Gamble et. al. 19%)

Mo ^^^ 1995 i%9

0.51270

" -

0.51280 h

Z 0.51270 h

mMm 1995 • 1945 ^

Kermadec Arc

o 0 O

• s

TVZ Basalts (Gamble et. al. 19%) Ruapehu (1945 -19%) Torlesse metasedimenis j

0.51260 h

0.51265

0.51250 h

. . . . 1. . . . 1. . . . 1 . . . . ! . . . . 0.51260 0.7055 0.706 0.7045 0.705 0.7035 0.704

0.51240

^^SrrSr

. . .. 1 . . R . . . 1 B . . ^ MIlllMlhlll 0.7030.7040.705 0.7060.7070.7080.709 0.71 0.711

''strsx

Figure 2. 87Sr/ 86Sr vs 143Nd/ 144Nd covariation in andesites ft-om Ruapehu Volcano. Data for TVZ basalts (Gamble et al, 1996) and Torlesse metasediments are shown for comparison. Detailed evaluation of the isotope data reveals the 1945 sample as the most radiogenic, with samples from the 1995-96 eruptions overlapping the range of all intervening eruptives (Figure 2). We suggest that the observed chemical and petrographic variations are best explained by injection of a fresh batch of andesitic magma into the volcano superstructure. This magma batch interacts and mixes with stagnated magma batches from previous events, and leads to abrupt chemical zonation (including reverse zoning) in the phenocryst assemblages (Nakagawa et al, in prep). The 1995 - 96 eruptives appear to represent a new cycle of activity, suggesting that there should be no complacency on the monitoring front. References Gamble, J.A.; Woodhead, J.D.; Smith I.E.M. & Wright, LC., 1996: Basalt and sediment geochemistry and magma petrogenesis in a transect from Oceanic Island Arc to Rifted Continental Margin Arc: The Kermadec - Hikurangi Margin subduction system. Joumal of Petrology, v 37, no. 6,. Graham, I.J. & Hackett, W.R., 1987: Petrology of calc-alkaline lavas from Ruapehu Volcano and related vents, Taupo Volcanic Zone, New Zealand. Joumal of Petrology, 28,531 - 567.

Nakagawa, M.,Wada, K., Wood, C.P., Thordarson, T., Gamble, J. A., (in prep) Intermittant discharge from discrete, small magma pockets: the 1995 eruptions of Ruapehu volcano. New Zealand. Nairn, LA., 1996: Volcanic eruption at a New Zealand ski resort prompts reevaluation of hazards. EOS, Transactions, American Geophysical Union, vol 77, no. 20th May 14th 1996,189 - 191.

APPLICATION OF MINERAL/FLUID/MELT TRACE ELEMENT PARTITIONING DATA TO MODELS OF ARC MAGMA GENESIS Trevor H. Green GEMOC, School of Earth Sciences, Macquarie University, NSW, 2109

It is widely accepted that most island arc basalts (lAB) come from a peridotitic source in the mantle wedge overlying the subduction zone (SZ). Geochemists commonly compare lAB with mid-ocean ridge basalts (MORB) because both are argued to represent relatively large degrees of melting of a mantle source, and both provide significant contributions to the earth's crust from the mantle. Striking chemical contrasts between lAB and MORB include an overall enrichment in lAB of Si02 and large ion lithophile elements (LILE), typified by Ba, Rb, Sr, U, Th and Pb, and depletion of high-field-strength elements (HFSE), typified by Nb, Ta and to a lesser extend Zr and Hf Models attempting to explam these chemical differences have usually proposed a critical contribution of LILE (but not HFSE) to the mantle wedge via a fluid phase from the subducted slab. This H20-rich fluid also has the effect of depressing the mantle solidus and enhancing the field of crystallization of olivine, yielding relatively Si02-rich basaltic melts. Alternatively, a contribution from a high-Si02 hydrous melt from the subducted oceanic crust, or from subducted pelagic sediments, interacts with and modifies the mantle wedge to give it the distinctive source characteristics needed for lAB. Another suggestion has been the chemical modification of the wedge by an upwelling carbonatitic melt. High pressure experiments have provided phase and major element constraints on these models, and importantly have outlined the restricted conditions where key accessory minerals (e.g. rutile) may have a critical role in controlling trace element (especially HFSE) distribution in derivative melts. High solubility of rutile in basaltic magmas at high pressure (P) is well established, so that rutile is not a residual phase to these magmas in their source regions. However, the marked decrease in solubility of rutile with decreasing temperature (T) and mcreasing Si02 dictates that rutile will be residual to silicic magma derivation from melting of subducted crust, and so may control the trace element content of the silicic melts that subsequently modify the composition of the peridotitic mantle wedge. Similarly, low rutile solubility in aqueous fluids at appropriate slab P and T suggests that rutile could be unportant in controlling the HFSE content of fluids entering the mantle wedge. New experimentally-obtained trace element partition coefficients between minerals and melt or fluid allow further constraints on arc petrogenetic models, and provide the possibility of distinguishmg between a dominant aqueous fluid, silicate melt or carbonatitic melt role in causing the distinctive trace element characteristics of the lAB source region. In particular, careful determination of and ^^ V ratios in island arc volcanics may provide pointers to the relative importance of these different trace element enrichment agents. Recent high precision results for arc magmas indicate variation o f ^ / j ^ from 11 (in the most Nb-depleted lAB) to 20 (in less Nb-depleted lAB) (Eggins et al, 1996) to 33 in high-K IA volcanics (Stolz et al, 1996) (compared with mantle = 17), whereas /^f varies from 30 to 48 (compared with mantle ^Vnf = 36). GEOCHEMICAL EVIDENCE FROM TRACE ELEMENT CHARACTERISTICS OF FLUIDS IN SUBDUCTION ZONES Using different approaches, several published estimates of fluid trace element content are presented in Fig. 1, normalized to N-MORB and to Sr = 1 to allow clearer relative comparison. There is remarkable consistency in the patterns, pointing to enrichment of the fluid in LILE (Cs to U and Pb). La, Ce and Sr are generally suggested to be slightly enriched, but the HFSE, Y and REE (Nd to Lu) are relatively depleted in the fluid. Unfortunately the data do not allow evaluation of ^ / j a or ^VHf ratios. FLUID/SILICATE MELT PARTITIONING DATA Determinations of trace element partitioning between aqueous fluid and silicate melts (see Table) ranging from basaltic to silicic are plotted in Fig. 2. All pairs show relative enrichment of Rb in the fluid, and relatively flat patterns for most other elements except for depletion of Th relative to U, Pb is enriched in fluid relative to andesitic melt (Keppler, 1996). Overall, elements partition much more strongly into fluids coexisting with silicic melts (where in fact the fluids approach the melt in composition) than into fluids coexisting with basaltic melts. Apart from the preceding points concerning Rb and Pb, fluids do not generally favour trace elements relative to melt, but Fig. 2 suggests that fluids may cause changes in element ratios, such as an increase in and possibly ^Vsp

TABLE: Major element contents of silicate melt starting compositions used in fluid/melt partitioning experiments. Si02 TiOa AI2O3 FeOxoT MgO CaO Na20 K2O P2O5 T.E.

60.1-64.5

62.4

12.5-13.2

18.5

4.1-4.2 8.2-8.6 5.7-5.9 0.3-1.1

5.6 6.2 4.3 1.2

68.8 0.3 16.4 3.0 0.5 4.1 4.7 0.5

6.1-2.3

47.2 2.1 14.8 11.3 8.9 8.8 3.5 1.5 0.6 ~1

T.E. denotes sum of trace elements added 1.

Ayers & Eggler, 1995 (NaCl - H2O fluids)

Keppler, 1996 (H.O or (Na, K)C1 - H2O fluids)

Adam et al, 1996 (trondhjemite) (H2O or H2O-F, H2O-CI fluids) Adam et al, 1996 (basanite) (H2O fluid)

MINERAL/FLUID OR MELT PARTITIONING DATA Partition coefficient (D) data for clinopyroxene, amphibole, garnet and rutile/fluid or melt pairs are given in Figs. 3-6, in order to evaluate any behaviour contrasts between minerals and variously fluid, silicate melt or carbonate melt. In general, mineral/fluid Ds are higher than mineral/melt values. However there are some significant exceptions and points of different behaviour, detailed as follows. For clinopyroxene (cpx) (Fig. 3) mineral/fluid Ds for Pb and Ba are similar to mineral/melt values. Cpx/fluid fractionates ^/xh more strongly than cpx/melt. Cpx/silicate and cpx/carbonate melt Ds appear similar, except for Zr and Hf, which are fractionated in opposite direction. For amphibole (amph) (Fig. 4) Rb and Pb mineral/fluid and mineral/melt Ds are close in value, but ^Vga behaviour is distinctly different for amph/fluid (<1) and each melt (>1 for carbonate melt, --1 for silicate melt). Also ^/ja is fractionated more by amph/fluid and amph/carbonate melt than by amph/silicate melt, and "^^^/REE is higher for amph/carbonate pairs than for either amph/silicate or amph/fluid pairs. For garnet (gt) (Fig. 5) mineral/fluid Ds for Ba and Sr are lower than mineral/melt values, and and will decrease in fluids but will increase in melts through gt fractionation. Gt/silicate or carbonate melt Ds show very similar behaviour. Although only a small number of rutile/fluid and melt D values is available, the very high D values are striking, so that a relatively small volume of rutile may have a significant effect on trace element behaviour. Rutile/fluid fractionates in the opposite direction to rutile/melt and to cpx, amph or gt/melt. Thus rutile/fluid fractionation will show a decrease in /j^ compared with an increase in ^/ja for all the mineral/melt fractionating cases. Also rutile/fluid D^ » Dj^, that is opposite to cpx/fluid, but sunilar to (though much higher than) gt/fluid. CONSTRAINTS ON FLUID VS MELT ROLE IN SUBDUCTION ZONE PROCESSES The cpx, amph, gt/fluid or melt D data indicate that for potential fluids or melts that could affect the peridotitic mantle wedge source region for SZ volcanics, relatively lower ^ ^ "^^^/i^e ^ the SZ volcanics point to a carbonatitic melt modifying role, whereas higher and ^Vga and lower suggests a fluid role. The similarity of trace element behaviour in fluid and trondhjemite indicates that there will be relatively little difference discernible between the role of high-pressure aqueous fluid and a low-degree trondhjemitic highpressure melt, m terms of modifying the trace element composition of the lAB source region. This generalization does not hold if trace element-enriched accessory minerals (e.g. rutile) remain residual during the derivation of the fluid or melt. The relatively lower T of sub-solidus fluid-related processes, compared with melt-related processes, enhances the likely role of accessory minerals in the source regions for the fluids. This 31

contrasts with the higher T melting situation where the greater solubility of the accessory minerals may strictly limit their potential for affecting the trace element contents of the derived melts. If rutile/fluid partitioning behaviour exerts an important control on the geochemistry of the lAB or SZ volcanics source region, then derived magmas may have < model mantle, whereas if rutile/melt control (together with cpx, amph or gt) is more significant than will be > model mantle. Thus the recently obtained ^/^a data for lAB of 11 to 33 may reflect this contrasting rutile/fluid or melt Nb and Ta partitioning behaviour. An important corollary is that a model continental crustal value of^/j^ ~ 11 would suggest that any major contribution to the growth of continental crust from SZ volcanism should come from magmas derived from a rutile/fluid affected source region. Additional eviden^ for a fluid rather than a silicate melt role may come from careful assessment of ^^La, ^' Th ^Vsr relative to ^/-ra. The fluid/melt partitioning data summarized here suggest that a negative correlation of these ratios would confirm that fluid-linked trace element behaviour was the controlling factor. REFERENCES Adam, J., Green , T.H., Sie, S.H., & Ryan, C.G., 1996. Trace element partitioning between aqueous fluids, silicate melts and minerals. EurJ.Mineral (submitted). Ayers, J.C., & Eggler, D.H., 1995. Partitioning of elements between silicate melt and HsO-NaCl fluids at 1.5 and 2.0 GPa pressure: Implications for mantle metasomatism Geochim.Cosmochim. Acta, 59, 4237-4246. Brenan, J.M., Shaw, H.F., Phinney, D.L., & Ryerson, F.J., 1994. Rutile-aqueous fluid partitioning of Nb, Ta, Hf, Zr, U and Th: implications for high field strength element depletions in island-arc basalts. Earth Planet. Sci. Lett. 128, 327-339. Brenan, J.M., Shaw, H.F. Ryerson, F.J., & Phinney, D.L., 1995. Mineral-aqueous fluid partitioning of trace elements at 9 0 0 a n d 2.0GPa: Constraints on the trace element chemistry of mantle and deep crustal fluids. Geochim.Cosmochim. Acta 59, 3331-3350. Eggins, S.M., Woodhead, J.D., Kinsley, L., Sylvester, P., McCulloch, M.T., Hergt, J., & Handler, M., 1996. A simple method for the precise determination of 40 or more trace elements in geological samples by ICP-MS using enriched isotope internal standardisation. Chem.Geol. (inpress). Green, T.H., 1994. Experimental studies of trace-element partitioning applicable to igneous petrogenesis — Sedona 16 years later. Chem.Geol. 117, 1-36. Hawkesworth, C.J., Hergt, J.M., Ellam, R.M., & McDermott, F., 1991. Element fluxes associated with subduction relatedmagmatism. Phil.Trans.KSoc.Lond. A. 335, 393-405. Jenner, G.A., Foley, S.F., Jackson, S.E., Green, T.H., Fryer, B.J., & Longerich, H.P., 1993. Determination of partition coefficients for trace elements in high pressure-temperature experimental run products by LAM-ICPMS. Geochim.Cosmochim. Acta 57, 5099-5103. Keppler, H., 1996. Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380, 237-240. McCulloch, M.T., & Gamble, J.A., 1991. Geochemical and geodynamical constraints on subduction zone magmatism. Earth Planet.Sci.Lett. 102, 358-374. Maury, R.C., Defant, M.J., & Joron. J-L, 1992. Metasomatism of the sub-arc mantle inferred from trace elements in Philippine xenoliths. Nature 360, 661-663. Pearce, J.A., Baker, P.E., Harvey, P.K., & Luff, I.W., 1995. Geochemical evidence for subduction fluxes, mantle melting and fractional crystallization beneath the South Sandwich Island arc. J. Petrol. 36, 1073-1109. Stolper, E., & Newman, S., 1994. The role of water in the petrogenesis of Mariana trough magmas. Earth Planet.Sci.Lett. 121, 293-325. Stolz, A.J., Jochum, K.P., Spettel, B., & Hofrnann, A.W., 1996. Fluid-and melt-related enrichment in the subarc mantle: evidence from variation in island-arc basalts. Geology,24, 587-590. Acknowledgments: The high-pressure experimental research involving determination of partition coefficients between minerals and melts or fluids has been supported by research grants from the Australian Research Council and Macquarie University. All of the data from Macquarie University used in this review has been obtained in collaboration with Drs. J. Adam, A. Chekhmir, A. Fujinawa, G. Nichols, N. Pearson, C. Ryan, S. Sie and E. Vicenzi and their contribution and interest is gratefully acknowledged.

0.01

Cs Ba U Ta Ce Sr Zr Sm Ho Tm Lu Rb Th Nb La Pb Nd Hf Tb Y Yb Maury

- Pearce

- H-worth

0.01

Cs Ba U Ta Ce Sr Zr Sm Ho Tm Lu Rb Th Nb La Pb Nd Hf Tb Y Yb - Keppler

- Ayers&E

• MoC&C-- S&N

Fig. 1 Inferred geochemical characteristics of sutxjuction zone fluids, nomraiized to N-MORB and Sr=1. Maury=Maury et al.1992; Pearce=Pearce et ai.1995; H^A«Drth=Hawkeswo^th et al. 1991; McC&G=McCulioch& Gamble, 1991; S&N=Stolper&Newman, 1994.

• Adam(bas) -W- Adam(tr)

Fig.2 Experimentally-determined fluid/melt partition coefficients at -0.3-2.0 GPa, 900-1000 C. Data sources and melt compositions: Ayers&E=Ayers& Eggler (1995)(andesite); Keppler=Keppier(1996)(andeslte); Adam(bas) or (tr) Adam et al.(1996){basanite) or (trondhjemlte).

1000

•

100

0.1

0.001 n nnni

3 k

0.01

•

td o

• ^

Clinopyroxene/Fluld or Mett j QQOI

-O- Ruid(Brenan ra.) • Fluid(Brenan ra.) -Ar Fluid(Adam) Silic. mett(various) Garb. melt(various) Fig. 3 Comparison of measured clinopyroxene/fluid or silicate(sihc.) melt or carbonatltic(cart).) melt partition coefficients at -2-2.5GPa. 900-1100 C. Brenan=Brenan et al. 1995; Adam=Adam et al. 1996; various=data from Green(1994), updated where appropriate.

-dk- Ruid(Adam) Garb. melt(various)

Fig.4 Comparison of measured amphibole/fluid or siiicate<silic.) melt or carbonatitic(carb.) melt partition coefficients at ~2-2.5GPa. 900-1100 G. Data sources as for Fig. 3.

Cs Ba U Ta Ce Sr Zr Sm Ho Tm Lu Rb Th Nb La Pb Nd Hf Tb Y Yb ^

Silic. melt(various)

Garb. melt(various) Fig. 5 Gomparison of measured garnet/fluid or silicate(silic.) melt or carbonatitic(carb.) melt partition coefficients at ~2-2.5GPa. 900-1100 G. Data sources as for Fig. 3.

Fluid(Brenan) Silic. melt(various)

Cs Ba U Ta Ce Sr Zr Sm Ho Tm Lu Rb Th Nb La Pb Nd Hf Tb Y Yb - e - FIuid(Brenan)

- - - - - - -

Cs Ba U Ta Ce Sr Zr Sm Ho Tm Lu Rb Th Nb La Pb Nd Hf Tb Y Yb

Fluid(Brenan)

Silicic Melt(Jenner)

Carbonatitic Melt(caic.) Rg.6 Gomparison of measured or calculated(caic.) rutile/fluid or silicate (silic.) melt or carbonatitic(carb.) melt partition coefficients at ~1-3.5GPa. 900-1100 G. Data sources ; Brenan=Brenan et al. 1994 and Jenner= Jenneretal. 1993.

U-SERIES ISOTOPES IN ARC MAGMAS Chris Hawkesworth^ Simon Turner\ Frank McDermott^, David Peate^and Peter van Calsteren^ ^Department of Earth Sciences, The Open University, Milton Keynes, MK7 6AA ^Department of Geology, University College, Dublin, Eire

Studies of short lived isotopes in igneous rocks and minerals have been influential in the development of recent ideas about melt generation models, and the rates at which melts are extracted and transported to the surface. ^^^U decays to stable ^^^Pb via a chain of short-lived isotopes which includes ^^^Th and ^^^Ra. The half-life for ^^^U is 4.47 x 10^ years, and 206p5/204p|3 variations in igneous rocks are therefore widely used to evaluate differences in source components that may have evolved over hundreds of millions of years (e.g. Zindler & Hart, 1986). In contrast, the half lives of 230Th and ^^^Ra are 75,380 and 1600 years, so they offer unique opportunities to investigate changes that take place on timescales of tens of thousands to a few thousand years. Since melt generation rates beneath mid-ocean ridges may be of the order of 10-^ kg.m-3.y"^ and residence times in shallow level magma chambers are typically <50,000 years, short-lived are important in attempts to develop physically realistic models for magmatic processes. Destructive plate margins are widely regarded as major sites of generation of continental crust, and of recycling of crustal material back into the upper mantle. Island arc magmas are generated in areas where the mantle is relatively cold, because of the subduction of oceanic crust, and partial melting is triggered by the introduction of hydrous fluids from the subducted crust. Overall, the average rate of crust generation per km above subduction zones is perhaps 5-10% that along mid-ocean ridges. In arc magmas some elements are preferentially enriched, and others are depleted relative to magmas generated along mid-ocean ridges and in intra-plate settings (e.g. Gill, 1981). It is inferred that certain elements are fractionated from one another in ways that are unique to destructive plate margins and, for example, that some elements are more readily mobilised in fluids from the subducted slab than others. It follows that the proportions of recycled and new, mantle-derived, material in arc magmas varies from element to element, and it appears that U and Th are readily fractionated in the processes responsible for island arc magmatism. Overall, the isotope compositions of U and Th therefore offer particular insights that range from the amounts and rates of transfer of material from the subducted slab, to the residence times of crystals and magmas within the crust prior to eruption (e.g. Gill & Williams, 1990; McDermott & Hawkesworth, 1991; Sigmarsson et aL, 1990; Gill et al., 1990; Condomines & Sigmarsson, 1993; Turner et al, 1996; Hawkesworth et aL, 1997). U AND TH IN ARC ROCKS U is expected to behave similarly to K, and to be relatively mobile in fluids released from the subducted slab. However, the behaviour of Th in the generation of arc magmas is more controversial, as is the extent to which Ta, Nb, Zr and Ti (as well as Th) are derived from variably enriched and depleted source regions in the mantle wedge (McCulloch & Gamble, 1991; Hawkesworth et aL, 1993; Pearce & Peate, 1995), or pre-existing continental crust in the form of subducted sediments (McDermott et aL, 1993; Plank & Langmuir, 1993). The distinctive high ratios of large ion lithophile (LIL) to high field strength (HFS) elements in arc rocks tend to be best developed in the more depleted, low-K rocks (Gill, 1981; Hawkesworth et aL, 1993; Pearce & Peate, 1995). A plot of U/Th-Th shows that high U/Th ratios are also best developed in the more depleted (low Th abundances) arc rocks (Fig. la). It is inferred that in arc magmas U behaves similarly to LILE such as Ba and K, whereas Th is much less mobile and behaves similarly to an HFSE. The obvious consequence is that U and Th are highly fractionated by the processes associated with destructive plate margin magmatism, particularly in the more depleted arcs. 1200

1000

Island Arcs

800 600 400 200

OT—

0.702 0

Th (ppm)

fluids •

•

Island Arcs

t-, •4.

•1.

crust/sediments? 0.706

0.708

87Sr/86Sr

Figure 1 Plots of U/Th-Th and Sr/Th-^'^Sr/^^Sr for basalts and andesites from selected arc suites (data sources in Hawkesworth et aL, 1997).

Sr/Th behaves similarly to Ba^Th and U/Th in island arc rocks, and it is plotted against ^^Sr/^^Sr in Fig. lb, in part because it also constrains the mixing arrays for Sr and Th isotopes. Sr/Th-^^Sr/^^Sr reveals a striking hyperbola that emphasises (i) that the distinctive high LIL/HFSE ratios are better developed in the more depleted arc rocks (low ^'^Sr/^^Sr), and (ii) that rocks with low Sr/Th, Ba/Th, and high Th tend to have high ^'^Sr/^^Sr. The restricted, and relatively unradiogenic Sr isotope ratios of samples with high LIL/HFSE suggests that the fluid component has relatively low ^'^Sr/^^Sr of -0.704. The rocks with higher Th and other incompatible element abundances tend to have more enriched isotope ratios (e.g. higher ^'^Sr/^^Sr, Fig. lb), and in principle they may be due to old trace element enriched material in the mantle wedge, and/or subducted sediments. However, a plot of Th/Ce vs i^SNd/l^^^d highlights the behaviour of Th in island arc rocks, and constrains the nature of the enriched component (i.e. with high Th, and ^^Sr/^^Sr). If Th was a relatively mobile element in this tectonic setting, high Th/Ce ratios should be best developed in the more depleted (high j-^cks. Strikingly, however, there is a broad negative array between Th/Ce and I'^^Nd/^'^'^Nd with the more depleted rocks being characterised by low, rather than high Th/Ce (Fig. 2). It is inferred that Th is not preferentially mobilised in the fluid component in arc rocks, consistent with the available experimental evidence (Keppler, 1996). The shift to low l43Nd/i44Nd is accompanied by marked increases in Th/Ce. High Th/Ce ratios are rare in mantle-derived rocks (Th/Ce = 0.016 and 0.11 in average MORB and OIB), but they are a feature of sediments with a significant continental component (Th/Ce = 0.22 in average post-Archaean shale). The low rocks also have low U/Th ratios, and it is concluded that the shift to high Th/Ce, low U/Th, high Th contents, and more enriched isotope ratios in these arc suites is primarily due to the contribution from subducted sediments (see also. Plank & Langmuir, 1993).

Figure 2 i43Nd/i44Nd-Th/Ce, for selected arc suites. Data sources as for Fig. 1, and as discussed in the text.

0.3

Th/Ce

•

0.2 ..

Island Arcs

0.5121

0.5123

•••

0.5125

0.5127

0.5129

0.5131

143Nd/ 144Nd In summary, the isotope and trace element compositions of arc magmas tend to require contributions from three components, viz. a (hydrous) fluid component and subducted sediments, both from the subducted slab, and the mantle wedge, and in principle these may be distinguished on the basis of selected minor and trace element ratios. In detail, the picture is complex because, for example, the isotope signature of a sediment contribution is in some cases associated with the high LIL/HFSE 'fluid' component (Morris et a l , 1990), rather than that with low LIL/HFSE ratios and high Th/Ce. In such cases, it is therefore inferred that the fluid component contains a contribution from subducted sediment. U AND TH ISOTOPES Many destructive plate margin rocks, and particularly those with low (^^Oxh/^^^xh), exhibit httle or no (238u/230xh) disequilibria (Fig. 3). In contrast, the rocks that do exhibit (238u/230xh) disequilibria tend to have high Th isotope and U/Th ratios, and hence lower Th, and other incompatible element abundances (compare Figs. 1 and 3). This indicates a link between the preservation of such short-lived disequilibria and U and Th abundances, which is why the U-series disequilibria in arc rocks are widely attributed to the addition of material that was itself out of isotopic equilibrium. A number of authors have further argued that the data are consistent with a broadly constant flux of U from the subducted slab, as that contribution is simply less significant in the rocks with higher U and Th contents (e.g. McDermott & Hawkesworth, 1991; Condomines & Sigmarsson, 1993).

Figure 3 e^^hfi^^Th) vs. ^^^UP-^^Th) illustrating the fields for recent MORE, OIB and island arc rocks. (Data sources in Hawkesworth et al, 1997).

(238u/232Th) Many of the arc rocks which are in or close to isotope equilibrium are those with lower Th isotope and U/Th ratios (Fig. 1) and higher incompatible element abundances. Arguments based on element ratios such as Th/Ce, Sr/Th and U/Th (Figs. 1 and 2), and those linking the fractionation-corrected abundances of Th with the regional subducted sediment flux (Plank & Langmuir, 1993), indicate that the low Th isotope, low U/Th component is primarily derived from subducted sediments. It is noteworthy that although few studies have measured ^^Be and Th isotopes on the same samples, the available data tend to exhibit positive rather than negative correlations between ^^Be and (^^^Th/^^^Th) (Sigmarsson et al, 1990; Gill et al, 1990; Reagan et al, 1994). This suggests that the sediment signature as evidenced by ^^Be is characterised by high (^^^Th/^^^Th), which is the opposite of that inferred from trace element arguments. However, the high i^Be material tends also to have high B/Be ratios (Morris et a/., 1990; Ryan & Langmuir, 1988); B is highly mobile, and so high B/Be ratios are a feature of the fluid component that also has high U/Th (Ryan & Langmuir, 1988, 1993) (Figs 1 and 3). Thus, it is the high LIL/HFSE fluid component which tends to have high ^^Be and low (^^^Th/^^Û), whereas the low Th isotope, low U/Th sediment component identified on the basis of trace element and Sr and Nd-isotope ratios, appears to be characterised by low ^^Be/^Be and (^^^Th/^^Û) isotope equilibrium. Whereas the best estimates are that the fluid component is transferred from the slab to the surface in 100,000-50,000 years (Turner et aL, 1996; Elliott et al., 1996), it would appear to have taken significantly longer for the low (^Ôxh/^^^Th) sediment component (perhaps >8 my; Hawkesworth et al, 1997).

SUMMARY STATEMENTS

1. Th is not readily mobilised in the fluid component along destructive plate margins. U is mobilised, and the resultant fractionation in U/Th can be used to estimate the rates of transfer slab derived components through the mantle wedge. The fluid component contains U, but little if any Th, and the resultant (^^^Th/^^^U) disequilibria indicate typical transfer times of 100-50 ky. In contrast, rocks with a greater contribution from subducted sediments (high Th/Ce and low I'^^Nd/^'^'^Nd) tend to have (230xh/238u) and low ^^Be, suggesting transfer times of several million years. 2. The variations in Th/Yb, and by implication in the fractionation-corrected Th abundances of arc magmas largely depend on the contributions from subducted sediments (Fig. 4). It is inferred that the distinctive high Th/Ta ratios of subduction related magmas primarily reflect the Th/Ta ratios of the subducted sediments, and that such high Th/Ta ratios are generated by processes other than those associated with recent subduction-related magmatism. Island Arcs

ThA^b

0.1

0.01

^^shale

Figure 4 A plot of Th/Yb-Ta/Yb, illustrating that in general island arc data can be broadly modelled by sediment-wedge mixing, since two component mixing is curved on log-log plots. Trace element ratios are used to minimise the possible effects of fractional crystallisation, and the PAAS is Post-Archaean Average Shale from (Taylor & McLennan, 1985). In such a simple two component model, the mantle wedge has lower Ta/Yb and Th/Yb than average MORB, and is therefore inferred to be more depleted than the source of such MORB (McCulloch & Gamble, 1991; Pearce & Peate, 1995).

3 The observation that the distinctive high LIL/HFSE ratios are best developed in the more depleted arc rocks is consistent with a broadly similar flux of the more mobile elements in the fluid component m magmas from different plate margins (McDermott & Hawkesworth, 1991; Hawkesworth et al, 1993; Condomines & Sigmarsson, 1993; Plank & Langmuir, 1993). 4. U and Th isotopes have also been used to evaluate magma residence times within the crust (e.g. Bourdon et al, 1994). Thus, separated minerals and groundmass from six rocks erupted in the last 4,000 years from Soufriere on St. Vincent in the Lesser Antilles, scatter about a 50,000 year errorchron on the U-Th equiline diagram (Heath et a/., 1977). Models are currently being developed to investigate how such apparent ages may relate to calculated replenishment times in steady state systems. 5 Bulk continental crust has a lower U/Th ratio (0.25) than at least some estimates for the bulk Earth (0 26) and the depleted upper mantle (0.39). However, the island arc rocks with low U/Th ratios appear to have inherited those from subducted sediments, and arc rocks with a low sediment contribution have significantly higher U/Th (Fig. 1). Thus, the U/Th ratios of new crustal material generated along de'structive plate margins are significantly higher than those of bulk continental crust. Rather the low averacre U/Th of bulk crust may be primarily due to different crust generation processes in the Archaean, when^lJ would be less mobile because conditions were less oxidising, and when residual garnet may have had more of a role in crust generation processes. REFERENCES

Bourdon, B., Zindler, A. & Womer, G., 1994. Evolution of the Laacher See magma chamber: Evidence from SIMS and TIMS measurements of U-Th disequilibria in minerals and glasses. Earth Planet. Sci. Utts. 126, 75-90. Condomines, M. & Sigmarsson, O., 1993. Why are so many magmas close to ^^^V-^^^Th radioactive equilibrium? Geochim. Cosmochim. Acta 57, 4491^97. Elliott, T., et al, in press. Element transport from subducted slab to juvenile crust at the Mariana Arc. J. Geophys. Res. Gill, J'B. '& Williams, R.W., 1990. Th isotopes and U-series studies of subduction-related volcanic rocks. Geochim. Cosmochim. Acta 54, 1427-1442. Gill, J.B., 1981. Orogenic andesites and plate tectonics. Springer-Verlag, Berlin. Gill', J.B.', Morris, J.D. & Johnson, R.W., 1990. Timescale for producing the geochemical signature of island arc magmas: U-Th-Po and Be-B systematics in recent Papua New Guinea lavas. Geochim. Cosmochim. Acta 57, 4269-4283. Hawkesworth, C.J., et al, 1993. Mantle and slab contributions in arc magmas. Ann. Revs. Earth Planet. Sci. 21, 175-204. Hawkesworth, C.J., et al, 1997. Subduction components in island arc magmas; constraints on flux estimates from U-series isotopes. Chem. Geol. (in press). Heath, E., et al, 1997. Magmagenesis at Soufriere volcano, St. Vincent, Lesser Antilles arc. J. Petrol (in prep). Keppler, H., 1996. Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380, lil-lAO. McCulloch, M. T. & Gamble, J.A., 1991. Geochemical and geodynamical constraints on subduction zone magmatism. Earth Planet Sci Lett. 102, 358-374. McDermott, F. & Hawkesworth, C.J., 1991. Th, Pb and Sr isotope variations in young island arc volcanics and oceanic sediments. Earth Planet. F., ScLDefant, Utts. 104, McDermott, M.J.,1-15. Hawkesworth, C.J. & Maury, R.C., 1993. Isotope and trace element evidence for three component mixmg in the genesis of the North Luzon arc lavas (Philippines). Contrib. Mineral Petrol 113,9-23. Morris, J.D., Leeman, W. P. & Tera, F., 1990. The subducted component in island arc lavas: constraints from Be isotopes and B-Be systematics. Nature 344,31-36. Pearce, J.A. & Peate, D.W., 1995. Tectonic implications of the composition of volcanic arc magmas. Ann. Revs. Earth Planet. ScL 23, Plank,251-285. T. & Langmuir, C. H., 1993. Tracing trace elements from sediment input to volcanic output at subduction zones. Nature 362, 739743. ^ , Reagan, M.K., Morris, J.D., Herrstrom, E.A. & Murrell, M.T., 1994. Uranium series and beryllium isotope evidence from an extended history of subduction modification of the mantle below Nicaragua. Geochim. Cosmochim. Acta 58, 4199^212. Ryan, J.G. & Langmuir, C.H., 1988. Beryllium systematics in young volcanic rocks: Implications for ^^Be. Geochim. Cosmochim. Acta 52, 237-244. Ryan, J.G. & Langmuir, C.H., 1993. The systematics of boron abundances in young volcanic rocks. Geochim. Cosmochim. Acta 57, 1489-1498. Sigmarsson, O, Condomines, M., Morris, J.D. & Harmon, R.S., 1990. Uranium and ^^Be enrichments by fluids in Andean arc magmas. Nature 346, 163-165. Taylor, S.R. & McLennan, S.M., 1985. The Continental Crust: its composition and evolution. Blackwell Scientific Publications, 312 pp. Turner, S.,et aL 1996. U-series isotopes and destructive plate margin magma genesis in the Lesser Antilles. Earth Planet. ScL Letts. 142, 191-207. Zindler, A. & Hart, S. 1986. Chemical Geodynamics. Ann. Rev. Earth Planet. Sci. 14, 493-571.

MANTLE DYNAMICS DURING ARC RUPTURE AND BACK-ARC OPENING: EVIDENCE FROM TONGA, THE LAU BASIN AND THE LAU ISLANDS. Janet Hergt

School of Earth Sciences, The University of Melbourne, Parkville, VIC; 3052, Australia The geodynamic interplay between a sub -arc mantle with Pacific and/or Indian MORJB-like affinities was first described by Hergt and Hawkesworth in 1994. Following this, a number of workers have recognised similar occurrences of these two domains in other arc/back-arc systems of the western Pacific (e.g., Hickey-Vargas et al, 1995; Crawford et aL, 1995). Controversy exists concerning the origin of these two mantle components and the extent to which they are geographically distributed. In order to test the geodynamic model proposed by Hergt and Hawkesworth (1994) for Tonga and the Lau Basin, additional samples from this region have been studied. These samples, from the Lau Islands, represent a record of magmatism beginning prior to arc rupture (i.e., prior to - 6 Ma) and extending through to 0.3 Ma ago (Whelan et aL, 1985). BACKGROUND The Lau back-arc basin has been the site of numerous petrological and geochemical studies, with interest stemming from the young age of opening (less than -6-7 Ma, Parson and Hawkins, 1994), and the orientation of the spreading axes relative to the active arc. Jenner et al (1987) measured samples from the Valu Fa Ridge, which is that part of the back-arc spreading ridge system closest to the active arc. It was noted that the compositions of the back-arc basalts are profoundly influenced by the close proximity of the arc (Ata island in particular), and subsequent studies have further compared samples from the Valu Fa Ridge with those erupted at sites more distant from the arc (e.g. Boespflug et al, 1990; Loock et al, 1990). The consensus is that at the Valu Fa Ridge, arc magmatism or a slab-derived flux of trace elements, produces melts with strongly arc-like isotopic and trace element signatures. This has been interpreted by Boespflug et al (1990) as evidence either for migration of the arc volcanic front into the back-arc domain, or for sea-floor spreading in its earliest stages of initiation at the Valu Fa Ridge. In contrast, magmas generated during sea-floor spreading away from the arc are close in composition to enriched MORB, and may show litde or no influence from the down-going plate. While described as 'enriched MORB' by most authors (Gill, 1976; Carlson et al, 1978; Hawkins and Melchior, 1985; Volpe et al, 1988), Loock et al (1990) used the Pb isotope compositions of their samples from the central basin to argue that the underlying mantle was in fact similar to Indian Ocean athenosphere, and that the convection cell beneath the Indian Ocean plate extends to this region. Differences between Pb isotope compositions in their samples from the central Lau Basin, and those from the Valu Fa Ridge led Loock et al (1990) to conclude that the mantle source closer to the arc contained a contribution from Pacific MORB-like material and Pacific sediments originating from a slabderived flux. Pb isotope compositions of basalts from the back-arc sites drilled during ODP Leg 135 provide additional insights to the study of this region and require the involvement of two mantle source regions—similar to those producing basalts at the Pacific and Indian mid-ocean ridges (Hergt and Hawkesworth, 1994). Figure 1 illustrates the Pb isotope compositions of the samples from these sites relative to a number of other fields for comparison. On a broad scale, the data clearly define two distinct trends which converge at high values of ^^^Pb/^^Pb. One of these extends into the field for Pacific MORB, while the other trend lies within the field defined by MORB from the Indian ocean (and data from rocks dredged from the modem spreading centres). Samples from the 'Pacific' trend are generally higher in ^^^Pb/^^'^Pb compared with most rocks from the modem Tofua arc, the exception being that they are lower when compared with limited data from the islands of Tafahi and Niuatoputapu (Ewart and Hawkesworth, 1987). In contrast, samples defining the 'Indian' trend plot within, or at lower ^^^Pb/^^'^Pb, compared with the Tofua samples. The Pb isotope variations are best explained as the product of mixing between an asthenospheric 'MORB' endmember, and an arc-like component (Hergt and Nilsson, 1994; Hergt and Hawkesworth, 1994). These data also demonstrate that the MORB endmember changed from one similar to Pacific asthenosphere, to one with a Pb isotope composition more like Indian Ocean asthenosphere during the evolution of the Lau Basin. Recently, a new model has been proposed describing the tectonic history involved in the opening of the Lau Basin (Parson and Hawkins, 1994), and, using this information to re-examine the Pb isotope variations, a clear pattern emerges. In short, the initial stages of back-arc formation involved horst and graben style extension of the arc crust, and some magmatism occurred within graben stmctures. True sea-floor spreading was initiated at -5.5 Ma and the tip of the propagating ridge migrated southward from the Peggy Ridge along the westem side of what was to become the Tonga platform. Axial spreading resulted in a fan-shaped opening in the northeastern part of the basin, while horst-and-graben style extension continued in the southwest.

Comparison between Pb-isotope variations and tectonic history, reveals that the basement from sites defining the 'Pacific' trend occurs entirely within the extended older arc crust forming the horst and graben fabric of the western Lau Basin, while samples defining the 'Indian' trend are from sites located within the new crust generated by true seafloor spreading. According to Hergt and Hawkesworth (1994) this correlation between the migration of the propagator, and a change in the asthenospheric endmember has had a number of important consequences, both for the evolution of the mantle beneath the Lau Basin, and the compositional evolution of the arc itself. The interpretation of Pb-isotope variations presented by Hergt and Hawkesworth (1994) is that roll-back of the subducting slab induced eastward asthenospheric advection into the sub-arc/basin system, and that this carried with it the boundary between Pacific mantle underlying the original arc, and the convecting Indian Ocean asthenosphere further to the west. This compositional interface between Indian and Pacific mantle, may have been similar to that described by Klein et al (1988) for the Australian-Antarctic discordance. Early rupture of the arc through block-faulting and horst and graben style extension was insufficient to upwell the Indian mantle source, and early melting in this region was confined to Pacific MORB and arc endmembers. Increased advection associated with continued rollback resulted in easward migration of the boundary between Indian and Pacific-like mantle. Consequently, mantle upwelling associated with the initiation of sea-floor spreading, succeeded in drawing up the Indian Ocean-like mantle that had advected into the underiying region. In this way, older Pacific Ocean asthenosphere has been swept aside, effectively being replaced during the southward migration of the propagating ridge tip. According to the tectonic model of Parson and Hawkins (1994), the ridge formed by the propagator is still in evidence as the Eastern Lau Spreading centre, and the propagator tip itself is now located at the Valu Fa ridge. Hergt and Hawkesworth (1994) noted that the latter is the only segment of the modem Lau Basin spreading systems producing magmas with influences from both Pacific and Indian Ocean Pb isotopic signatures. Figure 1 shows the Pb composition of the Valu Fa basalts relative to the two Tongan fields and the data for the Lau Basin—the Valu Fa data plot between the two back-arc arrays. Hergt and Hawkesworth (1994) suggest that this segment of the ridge is the site where the propagator tip is actively facilitating replacement of the Pacific mantle source with Indian MORB at the present day, and that the Valu Fa magmas represent mixtures that would have plotted on the 'Pacific' trend had it not been for the influence of the Indian ocean mantle. INFLUENCES ON ARC MAGMATISM—FURTHER SUPPORT The effective relacement of Pacific mantle with an Indian MORB source beneath the back-arc basin, must also have a profound influence on the Pb isotope composition of the arc magmatism. Hergt and Hawkesworth (1994) proposed that the underiying Pacific MORB mantle beneath the Tonga-Lau Basin system has been systematically replaced by Indian M O ^ owing to the migration of the propagating spreading centre down the westward side of the Tongan Ridge. This has not only changed the back-arc basin magmas from those lying on a Pacific MORB arc mixing trend, to those on an Indian MORB - arc mixing trend, but has also resulted in the replacement of the source of the arc magmas themselves. It is envisaged that the substitution of the wedge occurs as the propagator migrates southwards, and the Tonga Ridge swings away from its original position in a clockwise rotation. The only parts of the wedge to avoid this are those located either to the north or south of the locus of the propagating ridge tip. In order to test their model, Hergt and Hawkesworth (1994) demonstrated that the Pb isotope compositions of young islands located awayft-omthe influence of the propagator, are similar to the values obtained for the older manifestations of arc magmatism. For example, 'Eua, the only Tongan island to have exposed remnants of the original Eocene arc, has long been recognized as having Pb isotope compositions close to modem day magmas from Tafahi (e.g. Ewart and Hawkesworth, 1987). The Pb isotope compositions of Tafahi, Niuatoputapu, Ata, 'Eua and Sites 839 and 841 (from the Lau Basin) are all similar, and might define a broad field for the composition of original arc magmatism. Thus, there would appear to be two groups of island arc magmas; those which belong to the original arc (or currently erupt with these features), and the other, which typifies the majority of the modem Tofua arc (ie. excluding Ata, Tafahi and Niuatoputapu). A further test of this model is to examine the isotopic compositions of samples from the Lau Islands which are the abandoned remnants of the original arc (of which 'Eua was a part). Magmas preserved in the Lau Islands represent volcanism (i) within the ancient arc (LVG), (ii) during the earliest stages of rifting and back-arc basin formation (KVG), and (iii) continuing well after rifting and back-arc basin formation was established (MVG). According to the proposal of Hergt and Hawkesworth (1994), the earliest products of magmatism (arc rocks) would be expected to share the same 'Pacific' arc signatures as the ancient rocks preserved on 'Eua. Subsequent magmatism may also show evidence for an 'Indian' mantle influence if upwelling was sufficient to tap deq^er mantle source regions.

New Pb isotope data are shown for 'Eua and the same samples spanning the three magmatic groups of the Lau Islands reported by Cole etal (1990). The Lau Volcanic Group (LVG) include arc magmas preceding the onset of back-arc basin formation and plot together with data from 'Eua at the radiogenic end of the 'Pacific' array. This is precisely where the Pb isotope compositions of such magmas would be predicted to lie according to the model of Hergt and Hawkesworth (1994). While some samples from the Korobasaga Volcanic Group (KVG) and Mago Volcanic Group (KVG) lie in this same region, other data plot at significantly higher ^^^Pb/^^'^Pb compositions (KVG) and lower ^^Pb/^^Pb values (MVG). The influences on the petrogenesis of these magmas are clearly more complicated than so far described, however it is possible that the less radiogenic compositions of the MVG rocks relate to incorporation of Indian-type mantle material.

15.6

15.55

15.5

15.45

15.4 17.5

Figure 1. Variation diagram illustrating the new Pb isotope compositions of 'Eua and the Lau Islands (LVG, KVG, MVG) relative to MORB and other rocks from the area. Data sources are Hergt and Hawkesworth (1994) for the Lau Basin data, MORB data are compiled from the literature; Tongan arc data are from Ewart and Hawkesworth (1987) and references therein; the field for the modem Lau Basin is based on Hergt (unpublished); and the Valu Fa results are from Jenner et al (1987).

REFERENCES Boespflug, X., Dosso, L., Bougault, H., & Joron, J.L., 1990. Trace element and isotopic (Sr and Nd) geochemistry of volcanic rocks from the Lau Basin. Geol. Jahrb., 92, 503-516. Carlson, R.W., Macdougall, J.D., & Lugmair, G.W., 1978. Differential Sm/Nd evolution in oceanic basalts. Geophysical Research Letters, 5, 229-232. Cole, J.W., Graham, I.J., & Gibson, I.L., 1990. Magmatic evolution of Late Cenozoic volcanic rocks of the Lau Ridge, Fiji. Contributions to Mineralogy and Petrology, 104, 540-554. Crawford, A.J., Briqueue, L., Laporte, C., & Hasenaka, T., 1995. Coexistence of Indian and Pacific oceanic upper mantle reservoirs beneath the Central New Hebrides island arc. In: Active Margins and Marginal Basins of the Western Pacific. B. Taylor & J. Natland (eds). Geophysical Monograph 88., 199-217. Ewart, A.W., & Hawkesworth, C.J., 1987. The Pleistocene-Recent Tonga-Kermadec arc lavas: interpretation of new isotopic and rare earth element data in terms of a depleted mantle source model. Journal of Petrology, 28, 495-530. Gill, J.B., 1976. Composition and age of Lau Basin and Lau Ridge rocks: implications for evolution of an interarc basin and remnant arc. Geological Society of America Bulletin, 87, 1384-1395. Hawkins, J W., & Melchior, J.T., 1985. Petrology of Mariana Trough and Lau Basin basalts. Journal of Geophysical Research, 90, 11431-11468. Hergt, J.M., & Hawkesworth, C.J., 1994. Pb-, Sr-, and Nd-isotopic evolution of the Lau Basin: implications for mantle dynamics during backarc opening. In: Proceedings of the Ocean Drilling Program, Scientific Results, 135, J.W. Hawkins, L.M. Parson, J.R Allan, et al, Ocean Drilling Program, College Station, TX, 505-517. Hergt, J.M., & Nilsson-Farley, K., 1994. Major, trace element and isotope (Pb, Sr and Nd) variations in Site 834 basalts: implications for the initiation of back-arc opening. In: Proceedings of the Ocean Drilling Program, Scientific Results, 135, J.W. Hawkins, L.M. Parson, J.F. Allan, et aL, Ocean Drilling Program, College Station, TX, 471-485. Hickey-Vargas, R., Hergt, J.M., & Spadea, P., 1995. The Indian ocean-type isotopic signature in Westem Pacific marginal basins: origin and significance. In: Active Margins and Marginal Basins of the Westem Pacific. B. Taylor & J. Natland (eds). Geophysical Monograph 88., 175-197. Jenner, G.A., Cawood, P.A., Rautenschlein, M., & White,W.M., 1987. Composition of back-arc basin volcanics, Valu Fa Ridge Lau Basin: Evidence for a slab-derived component in their mantle source. Journal of Volcanology and Geothermal Research, 92, 209-222. Klein, E.M., Langmuir, C.H., Zindler, A., Staudigel, H., & Hamelin, B., 1988., Isotope evidence for a mantle convection boundary at the Australian-Antarctic Discordance. Nature 333, 623-629. Loock, G., McDonough, W.F., Goldstein, S.L., & Hofmann, A.W., 1990. Isotopic compositions of volcanic glasses from the Lau Basin. Marine Mining 9, 235-245. Parson, L.M., & Hawkins, J., 1994. Two-stage ridge propagation and the geological history of the Lau back-arc basin. In: Proceedings of the Ocean Drilling Program,Scientific Results, 135, J.W. Hawkins, L.M. Parson, J.F. Allan, et aL, Ocean Drilling Program, College Station, TX, 819-828. Volpe, A.M., MacDougall, J.D., & Hawkins, J.W., 1988. Lau Basin Basalts (LBB): trace element and Sr-Nd isotope evidence for heterogeneity in backarc basin mantle. Earth and Planetary Science Letters, 90, 174-186. Whelan, P.M., Gill, J.B., Kollman, E., Duncan, R., & Drake, R.E., 1985. Radiometric dating of magmatic stages in Fiji. In: Geology and Offshore Resources of Pacific Island Arcs - Tonga Region., D. Scholl & T. Vallier (eds)., Circum-Pacific Council for Energy and Mineral Resources, 415-440. ACKNOWLEDGMENTS The author would like to express her sincere gratitude to Tony Ewart and Jim Cole for their kind permission to work on these key samples. Analytical expenses are supported by the Australian Research Council.

SLAB MELT-MANTLE INTERACTION: IMPLICATIONS FOR ARC MAGMATISM AND MANTLE EVOLUTION Pavel Kepezhinskas and Marc J.Defant Department of Geology, University of South Florida, Tampa, FL, 33620, USA

INTRODUCTION Generation of siliceous Sr and LREE-enriched, water-saturated melts via partial melting of subducted slab under amphibolite to eclogite facies (adaldtes) is important in hot subduction zones (ridge subduction, subduction of young and hot oceanic crust, initiation of subduction, or oblique convergence) or in Archean convergent margins (Drunmiond et al., 1996). Slab-derived melts vigorously interact with the overlying subarc mantle wedge during their ascent and become progressively enriched in Mg, Cr and Ni resulting in generation of a high-Mg andesite or "transitional adaldte" (Kepezhinskas et al., 1996; Drummond et al., 1996). Melting of slab melt-metasomatiized source yields Nb-enriched alkaline basaltic melts (NEABs) wifli otherwise arc-like geochemical characteristics. The adakite-mantle interaction was even more profound in Archean subduction zones due to the higher tiiermal gradient in the early Earth. Large-scale derivation of tonalite-trondhjemite-granite (TTG) slab melts in Archean was synchronous with formation of diamondiferous cratonic roots via accretion of ultra-depleted (relative to MORB) island-arc mande at the base of continental lithosphere (Kepezhinskas et al., 1996). The TTG melt-mantle interaction may also have been responsible for generation of Archean high-Mg granodiorites and high-Nb lamprophyres and shoshonites. SLAB MELTS IN SUB-ARC MANTLE WEDGE Metasomatic veins of dacitic composition were documented in ultramafic mantie-derived xenoliths from Kamchatka (Kepezhinskas et al., 1996; Kepezhinskas and Defant, 1996). These veins are depleted in Ti, Zr and Y and are enriched in Al, Sr and display Na/K ratios of 3-7 (Figure 1). Typical siliceous glasses from mantle xenoliths in alkaline intraplate basalts have Na/K ratios less than 2 and high Ti contents resembling Ti02 concentrations in their host alkaline basalts. Glasses in Kamchatica mantle xenoliths are compositionally Figure. 1. Chemical composition of glasses in sub-arc mande xenoliths from Kamchatka (Kepezhinskas et al., 1996), Batan (Schiano et al., 1995) and Kuriles (Volynets et al., 1990) compared with expermental slab melts (Rapp and Watson, 1995), Kamchatica NEABs (Kepezhinskas et al., 1996) and Kamchatka pyroxenite (Kam. pyr.) and harzburgite (Kam. harz.) xenoliths (Kepezhinskas and Defant, 1996).

I I 55

Si02 (Wt.%)

similar to the experimental melts derived via partial melting of metabasalt at 15-32 kbars (Figure 1). Ion-probe analyses of Kamchatka mantie glasses reveal high Sr (1040-1736 ppm). La (66-125 ppm) and low Y (6-17 ppm) contents coupled witii high Sr/Y (60-285), La/Yb (20-30), Sr/Sm (46-209), Nb/La (0.198-0.280) and low Ti/Eu (71.4-128.3) and Y/Er (2.88-3.48) ratios. Typical arc andesites and dacites have Sr/Y ratios < 20-40 and La/Yb ratios < 20, while slab melts have Sr/Y ratios > 30 and La/Yb ratios > 20 (Drummond et al., 1996). Kamchatka siliceous glassesgenerally display high Cr contents (50-348 ppm) suggesting diat some reaction with peridotite took place prior to transport of die slab meltmetasomatized xenolith by an ascending basaltic magma.

Siliceous glass inclusions from Batan Island (Philippines) mantle xenoliths have high AI2O3 contents (17.2-19.7 wt. %) along with high Sr (279-810 ppm), La (31.2-69.2 ppm) and low Y (2.2-12.1 ppm) concentrations and high Sr/Y (41-238) and La/Yb (24-100) ratios (Schiano et al., 1995). These glasses have variable Na/K ratios of 0.62.05 and plot slighdy below the compositional field of experimental slab melts potentially reflecting various degrees of interaction with ultramafic material (Figure 1). Siliceous glasses in ultramafic xenoliths from Kurile arc have high Al and low Ti contents and variable Na/K ratios of 0.45-1.2 (Volynets et al., 1990). Kurile glasses display a decreasing Na/K ratio at increasing Si02 content which is typical of calc-alkaline arc magmas (Figure 1). We interpret these glasses as a result of partial melting of fluid-fluxed (abundant amphibole and phlogopite in glass-bearing xenoliths) sub-arc mande. Slab melts are found in the mantle wedge overlying both hot (northern) and cold (southern) subduction zone environments within the Kamchatka convergent margin (Kepezhinskas and Defant, 1996). This suggests that slab melt may be a ubiquitous component of sub-arc mantle even if current subduction conditions do not favor slab melting. This may very well explain afrequentoccurence of high-Mg andesites, transitional adakites and NEABs in die "not-so-hot" subduction zones (e.g., Aleutians, Andes, Philippines) which may be linked together via the process of interaction between adakite magmas and sub-arc mantle wedge. CHEMICAL SIGNATURES OF SLAB MELT-MANTLE WEDGE DVTERACTION Slab melts ascend into and react with the manfle wedge and become enriched in MgO, Cr and Ni while retaining their slab melt geochemical signature (e.g., high Sr/Y, La/Yb, Zr/Sm and low Y and Yb). During the course of this mteraction, slab melt becomes progressively saturated in in mantle components and starts to precipitate Al-Narich augite, garnet, Na-plagioclase and Al-rich spinel as observed in die ultramafic xenoliflis from Kamchatka (Kepezhinskas et al., 1996). High-Mg andesites (transitional adakites) in northern Kamchadca show high MgO and Cr contents and relatively low Sr/Y and La/Yb ratios which can be modelled by addition of peridotite to a pristine slab melt (Kepezhinskas et al., 1996). Kamchatkan transitional adakites contain olivme (F089-90), clinopyroxene (Mg# =85-88) and Cr-spinel (Cr/(Cr+Al) > 70) xenocrysts indicating that diey reacted with mantle peridotite on die way to the surface. Slab melt-contaminated mantle xenoliths from the Kamchatka arc show enrichments in Sr, La and Ta which are correlated widi modal contents of metasomatic disseminated phases (Al-Na-augite clinopyroxenes, amphiboles, and Al-rich spinels) which can be used as a rough estimate of die extent of slab melt-mantle wedge interaction (Figure 2). This enrichment is coupled with Nb enrichments detected in siliceous glasses from Kamchatka ultramafic xenolidis. This suggests tiiat extensive slab melt-mande interaction can produce Figure. 2. La vs. Ta variation in variably Slab melt-mantle metasomatized ultramafic xenoliths from - interaction the Kamchatica arc. Percentages reflect extent of slab melt-induced metasomatism estimated from petrographic studies (modal content of metasomatic Al-Ti-Na-augites a and Al-Fe-Mg-spinels) and ion-microprobe data on mantie clinopyroxenes (Kepezhinskas et al., 1996). UDM- ultradepleted unmetasomatized mantle xenolidis from Kamchatka. Kamchatka xpnoliths

HFSE (Nb and Ta) enrichment in both metasomatizing slab melt and die sub-arc mantie wedge. Slab melt-mantie interaction Ta (ppm) will also cause LILE (Sr) and LREE (La) and Na enrichment in the sub-arc mantie (Figures 1 and 2). Unmetasomatized to slighdy metasomatized harzburgite xenolidis in Kamchadca have relatively

0.05

0.1

0.15

low Na/K ratios while metasomatic wehrlites and pyroxenites exhibit elevated Na/K ratios of 3-11 due to the reaction with the Na-rich slab melt (Figure 1). Sub-arc wedge metasomatism by slab melt-mantie interaction produces a veined manfle source (matrix- harzburgites, veins- pyroxenites) enriched in HFSE (Nb and Ta), LILE (Sr) and LREE (La) capable of arc-related alkaline (high-Na) magma generation (Kepezhinskas et al., 1996). IMPLICATIONS FOR ARC MAGMATISM IN HOT SUBDUCTION ZONES Slab melt-mantle interaction in convergent margins associated with ridge subduction, oblique convergence or initiation of subduction results in hybridization of both ascending slab melts and sub-arc mantie sources. Major element relationships between adakites, transitional adakites, high-Mg andesites and boninites suggest conmion petrogenetic link (Drummond et al., 1996). NEABs which are commonly spatially and temporally associated with adakites and transitional adakites have typical arc depletions in HFSE even though they have high absolute concentrations in these elements. The NEABs are enriched in Nb relative to island-arc basalts and are lower in HFSE compared with oceanic-island basalts (Kepezhinskas et al., 1996). Based on the correlated Sr, La and Ta enrichments in slab melt-metasomatized mantie wedge xenoliths in Kamchatica, we suggest that progressive slab melt-mantie interaction in hot subduction zones will produce a hybrid source witii selective vein enrichment in Nb, Ta, Sr and LREEs (Figures 1 and 2). Melting of this slab melt-metasomatized source will generate arc magmas with "OIB-like" signatures (HFSE enrichment) coupled with typical arc signatures (Sr and LREE enrichment) without actual involvement of OIB-type mantie sources. We suggest tiiat a litiiologic association of adakite- transitional adakite-high-Mg andesite-NEAB exists in arc tectonic settings associated with ridge subduction, subduction of young and hot oceanic crust, initiation of subduction, or oblique convergence or in Archean convergent margins (Kepezhinskas et al., 1996; Drunmiond et al., 1996). SLAB MELT-METASOMATIZED MANTLE WEDGE: A PROTOLITH FOR THE CRATONIC LITHOSPHERE Kamchatica harzburgite xenoliths exhibit mantie olivine (F091.93), spinel -Cr/(Cr-hAl) of 0.65 to 0.81,and AI2O3 in opx - 0.4-1.6 wt. %. The modal opx content ranges from 10 to 25%. These harzburgites exhibit extremely low Ti02 (0.01-0.03 wt.%) and AI2O3 (0.4-0.6 wt.%) and high Cr (2635-4360 ppm), Ni (2361-2600 ppm) and refractory PGEs (Os- 3-6 ppb, Ir- 3.4-6.8 ppb, Ru- 10-28 ppb). Some harzburgite xenolitiis contain graphitecoated diamonds witii d^^C values of -33.4 and -35.1 VQO indicating a potential slab sedimentary origin for carbon.Kamchatica harzburgite xenoliths are systematically more depleted than the MORB mantie in both mineral and bulk compositions (Figure 3). The Kaapvaal-type cratonic peridotites have notable compositional Figure. 3. AI2O3 in orthopyroxene vs. Cr203 in co7 existing spinel for Kamchatica pre-metasomatized CCJ 6 mantie peridotites (Kam. PMM) and slab meltX N . Mantle metasomatized peridotites and pyroxenites (Kam. 2 5 \^epletion liiiiiM SMM) compared with oceanic peridotites (Bonatti and Michael, 1989), mclusions in diamond (Gumey I 4 and Zweistra, 1995), and cratonic peridotites C 3 — Kam. (Rudnicketal., 1994). Kam^ SMM PMM in A sunilarities to the Kamchatica peridotites: low Al i ' content of co-existing orthopyroxene and spinel, s 1 Slab melt high olivine Fo content, high modal orthopyroxene jc 1 i ^ Crato^c peridc(tites p ^ ^ < 0 abundance. Botii Kaapvaal and Kamchatka 70 peridotites contain orthopyroxene-rich veins 60 10 20 30 40 50 interpreted as a result of slab melt-mantle C12O3 (wt.%) in spinel interaction (Rudnick et al., 1994; Kepezhinskas and Defant, 1996). We propose a three-stage melt extraction model to achieve the depleted (relative to MORB manfle) chemistry of sub-arc mantie wedge (Kam.

PMM in Figure 3), cratonic peridotites and peridotitic inclusions in diamonds: 1) extraction of MORE melt at midocean ridges; 2) extraction of back-arc basaltic melt in back-arc spreading centers and 3) hydrous partial melting in a sub-arc mantle wedge to produce arc magmas. This backarc-island arc re-melting of the MORB-type mantie in Archean was probably accompanied by voluminous derivation of TTG suites via partial melting of young and hot oceanic crust and introduction of carbon necessary for diamond formation into the overlying depleted mantie wedge. This is consistent with an apparent synchronicity of TTG crustal production (3.5-2.1 Ga), Sm-Nd ages for diamonds (3.4-1.9 Ga), Os model ages for cratonic lithospheric peridotites (3.5-2.7 Ga) and kimberlite-bome eclogites (3.4-2.6 Ga) (Pearson et al., 1995). This slab melt-metasomatized mantle was subsequently carried below cratons and underplated at the base of growing continental lithosphere.

REFERENCES Bonatti, E., 8l Michael, P.J., 1989. Mantie peridotites from continental rifts to oceanic basins to subduction zones. Earth and Planetary Science Letters, 91, 297-311. Drummond, M.S., Defant, M.J., & Kepezhinskas, P.K., 1996. Petrogenesis of slab-derived trondhjemite-tonalitedacite/adakite magmas. Transactions (fthe Royal Society of Edinburgh, Earth Sciences, 87, 205-215. Kepezhinskas, P.K., Defant, M.J., & Drummond, M.S., 1996. Progressive enrichment of island arc mantle by melt-peridotite interaction inferred from Kamchatica xenoliths. Geochindca et Cosmochirmca Acta, 60, 1217-1229. Kepezhinskas, P.K., & Defant, M.J., 1996. Contrasting styles of mantie metasomatism above subduction zones: constraints from ultramafic xenoliths in Kamchatka. In: Bebout, G.E., Scholl, D.W., Kirby, S., and Piatt, P.J, eds., Dynamics of Subduction, AGU Monograph. Gumey, J.J., & Zweistra, P., 1995. The interpretation of tiie major element compositions of mantle minerals in dhmond tx^loiMon. Journal of Geochendcal Exploration, 53, 293-309. Pearson, D.G., Carlson, R.W., Shirey, S.B., Boyd, F.R., & Nixon, P.H., 1995. Stabilization of Archean lithospheric mantie: a Re-Os isotope study of peridotite xenoliths from tiie Kaapvaal craton. Earth and Planetary Science Letters, 134, 341-357. Rapp, R.P., & Watson, E.B., 1995. Dehydration melting of metabasalt at 8-32 kbar: continental growth and crust-mantie recycling, Journal of Petrology, 36, 891-931.

Implications

for

Rudnick, R.L., McDonough, W.F., & Orpin, A., 1994. Northern Tanzanian peridotite xenoliths: A comparison with Kaapvaal peridotites and mferences on metasomatic reactions. In: Meyer, H.O.A., and Leonardos, 0., eds.. Proceedings of the 5th International Kiwberlite Conference, CPRM, Brasilia. Schiano, P., Clocchiatti, R., Shimizu, N., Maury, R.C., Jochum, K.P., & Hofrnann, A.W., 1995. Hydrous, silica-rich melts in the sub-arc mantie and their relationship witii erupted arc lavas. Nature, 377, 595-600. Volynets, O.N., Avdeiko, G.P., Tsveticov, A.A., Ananyev, V.V., Antonov, A.Yu., Gladkov, N.G., & Markov, LA., 1990. Peridotite inclusions in Quaternary lavas from tiie Kuriles Island Arc. Transactions of the USSR Academy of Sciences, Geological Series, 3, 43-57 (in Russian).

TEMPORAL CHANGES IN ISLAND-ARC RECYCLING: Pb ISOTOPIC EVIDENCE FOR POST-ARCHEAN URANIUM RECYCLING (PURE) Malcolm T. McCulloch Research School of Earth Sciences The Australian National University, Canberra, ACT, 0200 INTRODUCTION The eruption of basalt at mid-ocean ridge spreading centres, their cooling and hydrothermal alteration by seawater, transit to subduction zones and finally recycling into the mantle are the central elements of the platetectonic paradigm. This cycle is important not only as the major mechanism for heat loss from the Earth, but also because it provides the geochemical pathway for the interaction that occurs between the Earth's mantle, hydrosphere and upper continental crust. Island-arcs are one of the key components of this cycle as they act as a selective 'mantle filter' for the recycled slabs as well as the active site for the generation of 'new' continental crust. In this study the key role that long-term recycling of hydrothermally altered oceanic basalt has played in the Pb isotopic evolution of the upper mantle is discussed. In particular it will be argued that the Pb isotopic composition of MORB's provides evidence for a major temporal change in the nature of the island-arc recycling process. This temporal change probably occurred during the late-Archean, early Proterozoic, with the transition from partial melting to dehydration of subducted slabs, together with a change in the oxidation state of the uppermost continental crust. It will be shown that post-Archean U-recycling has been the single most important process in determining the Pb isotopic evolution of MORB's (McCulloch, 1993). THE Pb ISOTOPIC COMPOSITION OF MORB'S AND THE T b PARADOX' The Pb isotopic composition of MORB's (and OIB's) is characterised by high 206p5/204pb relative to a single-stage Pb evolution for the Earth (Fig. la). The relatively high 206p5/204pt> ratios in MORB implies a post-Archean increase in the U/Pb ratio in the sources of these magmas as a result of the longer half-life of ^^^U compared to The Pb isotopic composition of MORB's therefore poses a paradox as it is now well established that during partial melting U is more incompatible than Pb. As a consequence melts have high U/Pb and the residual source should have low U/Pb. Thus, with continual extraction of magmas from the upper mantle, the residual MORB source should become progressively depleted in U relative to Pb and therefore have low 206pb/204p5 ratios that lie to the left, rather than to the right of a single stage geochron. A number of mechanisms have been proposed to explain the 'Pb paradox'. One possibility is that the Earth is younger than the 4.55 Ga age of meteorites. Based on the Pb isotopic composition of early Archean galenas from Isua (Richards and Appel, 1987), an age for the Earth of -4.52 Ga to 4.46 Ga has been proposed (Galer and Goldstein, 1996, and McCulloch, 1996). The effect of a younger age for the Earth is to shift the singlestage geochron to the right and in Fig. la it can be seen that some Indian Ocean MORB's now lie on or near to the younger 4.50 Ga geochron, but the vast majority of Atlantic and Pacific ocean MORB's still show large excesses in ^^^Pb. Regardless of the position of the geochron, the MORB Pb isotopic array still has a much

15.7 _ . . . . 1 . . . . 1 1 1r 1J • . • 1 1 1 1 1 1 1 1 1 1 40 •, MORB a : T = 4.50Ga 15.65 7 o Pacific 39.5 - o I • Atlantic j - • • ° • Indian / 15.6 ^ • =9^ ^ ; / 39 : T = 4.55Ga/ ^ o t S ; ^ T=1.8Ga J 15.55 a j ^ Q : 38.5 / • r f© f 15.5 I"o^ • : 38 ® J 15.45 15.4

^ • :

I 37.5

1 \i = S.5

15.35 " 1 . I 1 1' . '. 11 1A 11 11 I 1 . I 1 1 . I 1 1 1 i J. 17.0

17.5

18.0

18.5

19.0

19.5

20.0

206pb/204p^

37 17.0

1 1 1 1 •1 1 1 1 1 1 1 1 1 1 • 'b:>

MORB Pacific Atlantic Indian

Th/U = 3.9

0-i

• •

Th/U = 3.7 1 1 1 1 1 1 1 1 1 1 1t I1 1i 11 1 1 17.5

18.0

18.5

19.0

206pbye04pb

Figure 1. Pb isotopic composition of MORB's

y/n

19.5

20.0

shallower slope (Fig. la), indicating a multi-stage history, which if given strict age significance implies a major U/Pb fractionation event in the early-Proterozoic. Other explanations of the Tb paradox' include latestage growth of the Earth's core and recycling of radiogenic Pb from the continental crust into the mantle. As already discussed, early Archean Pb isotopic compositions imply that the Earth's core was probably formed by 4.50 Ga, which is also consistent with the relatively high abundances of the platinum group elements in the mantle which has been attributed to a 'late stage bombardment'. An alternate and now 'conventional' hypothesis, first proposed by Armstrong, (1981), is the recycling of radiogenic Pb from the continental crust. This hypothesis is unsatisfactory for several reasons. Firstly typical continental crust as represented for example by marine sediments has much higher 207pb/204p5 ^nd 208pb/204p5 j-^tios compared to MORB's. This is a consequence of the Pb isotopic composition of continental crust reflecting its Archean pre-history of high U/Pb and Th/U. For this reason the recycling of continental sediments into the MORB source would be expected to result in much steeper arrays in the Pb-Pb plots than what is observed. This problem is exacerbated if older sediments are recycled as they have even higher ^^'^Pb/^^'^Pb. The second key argument against sediment recycling is that both MORB's and OIB's have very similar Nb/U and Pb/Ce ratios which are distinct from the continental crust (Hofmann et al., 1986). This implies that if recycling of continental crust has occurred, both MORB's and OIB's have been differentially modified by the same degree, which is highly unlikely. Thus, the constraints from Pb isotopes together with the uniform elemental ratios such as Nb/U and Pb/Ce imply at most only minimal amounts of recycling of 'typical' continental crust. ThOJ RATIOS OF MORB s Galer and O'Nions, (1985) pointed out that modern MORB's have much lower Th/U ratios (-2.5) than the long-term average value (-3.7) inferred from their Pb isotopic composition (Fig. lb). Apparent inconsistencies between observed parent/daughter ratios and long-term 'mean' ratios are not however unique to the Th-U-Pb system and have been observed in the Lu-Hf and Sm-Nd systematics of MORB's. There are a number of possible explanations. The most obvious is the fractionation of parent/daughter elements during the partial melting-fractionation processes associated with MORB generation. For Th/U this would produce a trend that is opposite to what is observed (shown as recent fractionation trend in Fig. 3) as both the melts and source, (the latter inferred from U-Th disequilibrium) have low Th/U (<3) ratios (Fig. 3). For this reason Galer and O'Nions argued that the upper mantle was essentially an 'open system' with the residence time for Th<U«Pb. On this basis a very short a residence time of -600 Ma was proposed for the highly incompatible element Th. The time-scale depends on the initial or bulk Earth Th/U ratio and if a higher value of Th/U -4.2 (Fig. 2) is used instead of 3.9 as assumed by Galer and O'Nions, the residence time is increased to - 1 Ga for Pb. There are still a number of major problems with the model of Galer and O'Nions (1985). Firstly the residence time of -600 Ma proposed for Th is extremely short compared to the time-scales proposed for platescale motions in the upper mantle. As discussed, this problem has been emolliented somewhat by assuming a higher Th/U ratio for the bulk silicate Earth, but still remains at - 1 Ga. The most serious problem with this model is that it is inconsistent with the Pb isotopic composition of MORB's (Fig. la). As already noted the Pb isotopic composition of MORB's requires a recent increase in the U/Pb ratio of the MORB source whereas the Galer and O'Nions model based on the relative incompatibility of U>Pb implies the opposite. Clearly a self-consistent explanation is required for both the 'Pb paradox' and 'Th/U dilemma'. An obvious quandary for the 'open system' model is also the observation by Hofmann et al., (1986) that both MORB's and OIB's have similar Nb/U ratios. This requires that U (and Nb) have the same residence time in both the upper and lower 32

r-T

: T = 4 . 4 9 Ma 31.8 7

. . .

=9.0

Th/U = 4.3 Th/U = 4.2

31.6

31.4

-' '

• Isua galenas

11.00

- ' ^3500 Ma ~

. - ' 3600 Ma

31.2

30.8 "

' ' •

Amitsoq feldspars

Big Stubby

^^^^^^ ' ' ^ ^ - ' 3700 Ma

Th/U = 4.0

- ' 3800 Ma 1

1 1 11.20

1 11.40

1 11.60

1 11.80

. " 12.00

Figure 2. Pb isotopic compositions of early Archean galenas, from Richards and Appel (1987). These give a bulk Earth Th/U ~ 4.3.

mantle, the presumed sources of MORB's and OIB's respectively. Considering the different geometry and transfer processes this is highly unlikely. Implicit in the 'open system' model of the upper mantle is that the transfer of elements from the mantle to the continental crust is only controlled by solid-melt equilibria. In island-arc subduction zones, the major site of present-day magmatic activity, both Nb/U and ThAJ ratios are low, implying that other factors such as fluid-rock partitioning are important. POST-ARCHEAN URANIUM RECYCLING (PURE) Seafloor hydrothermal alteration is a pervasive process, and evidence from oxygen isotope studies indicates that seawater circulation can penetrate to depths within the oceanic crustal section of at least 2 to 5 kilometers. Although the effects of both low and high temperature hydrothermal alteration are complex, it is clear that the oceanic crust acts as a sink for U, as well as playing an important role in buffering the isotopic composition of Sr and oxygen in seawater. The possibility that recycling of altered U-enriched oceanic crust could play an important role in OIB's as well as MORB's has therefore been widely accepted (e.g. Hofmann et al., 1986^ A key question however is whether the recycling of oceanic crust has been an ongoing 'steady-state' process or has there been temporal changes? Archean greenstones show that seawater hydrothermal alteration of basaltic crust was certainly of similar if not greater importance than today. In addition it is also likely that throughout the Archean larger volumes of basalt as well as higher temperature magmas such as komatiites were erupted at oceanic spreading centres, although plume magmatism rather than passive upwelling may have been more dominant. Thus it is likely that in the Archean there were large volumes of altered basaltic crust undergoing seafloor alteration and ultimately recycling. Is there any evidence remaining of the large-scale recycling of altered Archean oceanic crust? Despite comprehensive analyses of OIB's, which are presumed to incorporate a significant component of recycled oceanic crust in their source, there is a complete absence of Archean signatures in their Pb isotopic compositions. This is apparent from the analysis by Chase (1981), who showed that OIB Pb's have a two-stage history, with the last-stage being U enrichment occurring from -2.47 Ga (Reunion) to 0.94 Ga (Hawaii). This conclusion remains valid even if the OIB Pb arrays are in part due to mixing, rather than aging of variably altered oceanic crust. In this respect the key evidence is the lack of high 207p|3/204p|5 ji^ either OIB's or MORB's. The absence of Pb isotopic evidence in OIB's for recycled Uenriched Archean crust is surprising. Even with a relatively short 'half-life' of 1 Ga, for a constant production ratio of oceanic crust, then -25% of the recycled oceanic crust should still be of Archean age. If, as already noted, the rate of production of altered Archean crust was significantly higher in the Archean (ie >x2), then the present-day mantle source for OIB's should still contain a significant component of old Archean recycled oceanic crust! Does this indicate a temporal change in the rate of U-recycling at island-arc subduction zones? There are several factors that may have induced changes in the rate of U-recycling. Firstly, the mobility of U is closely related to its oxidation state. In the Archean, conditions were much more reducing, and hence in the uppermost continental crust U was in its immobile U^"'" form. It is thus likely that during Archean times, U was retained in the upper continental crust and the U flux to the oceans was low. This is consistent with U deposits being mainly of late-Archean or early-Proterozoic age. The uppermost continental crust (and atmosphere) is today much more oxidising and during weathering U is oxidised to its highly mobile uranyl form. As a result the present-day flux of U from the continental crust to the oceans is high, with the U budget of modern seawater being dominated by continental (high 234u/238u) inputs. During seafloor hydrothermal alteration conditions are again reducing, resulting in U being fixed into altered oceanic crust. If , . . • 1 .-r-,-r-r-T-i-i

5 4.5 4 3.5 E 3

•

depleted mantle

: r :

U-recycling

O• " ° •

MORB source

<•/

1.5 1

- , , , , 1 , , . . 1 . . . . 1 0.0

5.0

results in both lower ThAJ and U/Pb ratios. Recycling of altered U-enriched oceanic crust increases the U/Pb ratio of the MORE ~ source. :

0 -

° /

10.0

15.0

238u/204pb

Figure 3. Plot of measured ThAJ versus

White (1993). Extraction I of continentalfrom crust (depleted mantle trend)

• ^/ . •(3>

2.5

. i |1 1 1 1 1 1 1 . . 1 _

Bulk silicate ~ • Earth : continental crust | ^ extraction /

° :

Recent fractionation trend

MORB's O

Pacific

•

Atlantic -

•

Indian

•

1 1 1 1 i 1 1 > 1 .1,.' 20.0

25.0

as is likely, Archean seawater had significantly lower concentrations of U, then altered Archean oceanic crust would not be enriched in U. The second possibility is that the behaviour of recycled slabs at island-arc subduction zones may have changed with time. Higher Archean geotherms may have resulted in partial melting rather than the present-day situation of dehydration of subducted slabs. During partial melting it is likely that melt/solid equilibria will dominate, resulting in the highly incompatible elements such as Th, U and Pb being effectively stripped from the downgoing slab and quantitatively transferred to the continental crust. Thus, there would be no record in the Pb isotopes of Archean ocean crust recycling, apart from the upper mantle having progressively lower Th/U and U/Pb ratios (Fig. 3). In the late-Archean early-Proterozoic, slab dehydration probably became the dominant regime, resulting in a significant change in the efficiency of the 'island-arc filter'. In particular, U-enriched oceanic slabs (with U now in its immobile reduced sate) would now be able to pass through the island-arc filter and be recycled into the deep mantle. This would result in an increased flux of U into the post-Archean mantle. Post-Archean slabs would thus be characterised by high U/Pb and somewhat lower ThAJ ratio (Th is also immobile in subduction zones). This is the basis of the post-Archean U-recycling model of McCulloch (1993) which has been given the acronym PURE by Zindler et al., (1994), indicating that U rather than Pb recycling, is the important process. The PURE model also has the important attribute that a single mechanism (U-recycling) provides a self-consistent explanation for both the 'Pb paradox' as well as the 'Th/U dilemma'. The high 206pb/204pt> ratios in MORB's (and OIB') are thus a direct outcome of post-Archean recycling of U-enriched altered oceanic crust with the 1.8 Ga to 2.0 Ga MORE Pb array also being a direct outcome of this process. Post-Archean U-recycling also contributes to the further 'recent' lowering of the Th/U ratio in the MORB source, which together with the progressive depletion of the upper mantle due to continental crust extraction, accounts for the difference between the average <Th/U> ratio (estimated from Pb isotopes) versus the presentday ratios. It should also be noted that the U-recycling process may not be ubiquitous, as some Indian Ocean MORB's have relatively low 206pb/204pb ratios, that lie to the left of the single stage geochron (Fig. la) together with relatively high Th/U, <Th/U> and low U/Pb (Figs 1&3). There are several testable predictions of the PURE model, especially in Pacific and Atlantic MORB's that exhibit high 206pb/204p5 (>18.5). Firstly a general correlation between MORB samples having both high 206pb/204pi5 and high U/Pb would be expected, at least on a local-scale. There is a trade-off however between the age of the recycled slab and the degree of U-enrichment. Recent fractionation from partial-melting and sulphide partitioning also needs to be taken into account (Fig. 3). Another interesting aspect is the effect of U-recycling on the Nb/U ratios of MORB's and OIB's. U-recycling would be expected to result in recycled oceanic crust having progressively lower Nb/U and hence OIB's should 'lead' MORB's by having lower ratios. This is observed for example in some OIB's (e.g. Pitcaim Islands). Nevertheless, the overall similarity of Nb/U ratios for OIB's and MORB's may indicate a large pulse of U-recycling at the Archean-Proterozoic boundary due to the combined effects of large-scale U mobilisation from the continents together with the initiation of slab dehydration. The Nb/U ratio (-50) of this pulse would thus dominate both OIB's and MORB's and could also account for the early-Proterozoic Pb-Pb age of MORB's. REFERENCES Armstrong, R.L., 1981. Radiogenic isotopes: the case for crustal recycling on a near-steady-state nocontinental-growth Earth. Phil Trans. Roy. Soc. Land. A301, 443-472. Galer, S.J.G. and O'Nions, R.K., 1985. Residence time of thorium, uranium and lead in the mantle with implications for mantle convection. Nature 316, 778-782. Galer, S.J.G., & Goldstein S.L., 1996. Influence of accretion on lead in the Earth, in Earth Processes Reading the Isotopic Code. Geophysical Monograph 95, 75-98. Chase, C.G., 1981. Oceanic island Pb: two-stage histories and mantle evolution. Earth and Planetary Science Letters 52, 277-284. Hofmann, A.W., Jochum, K.P., Seufert, M., & White, W.M., 1986. Nb and Pb in oceanic basalts: new constraints on mantle evolution. Earth and Planetary Science Letters 79, 33-45. McCulloch, M.T., 1993. The role of subducted slabs in an evolving earth. Earth and Planetary Science Letters 115, 89-100. McCulloch, M.T., 1996. Isotopic constraints on the age and early differentiation of the Earth. Journal of the Royal Society of Western Australia 79, 131-139. Richards, J.R., & Appel, P.W.U., 1987. Age of the "least radiogenic" galenas at Isua, West Greenland. Chemical Geology 66, 181-191. White, W. M., 1993. 238u/204p5 in MORB and open system evolution of the depleted mantle. Earth and Planetary Science Letters 115, 211-226. Zindler, A., Bourdon, B., & Elliott, T., 1994. Debunking the K conundrum, abstr. 8th ICOG, USGS circ. 1107, 371.

AMPHIBOLE STABILITY AND ITS ROLE IN INFLUENCING INTERMEDIATE FRACTIONATION TRENDS IN SUBDUCTION-RELATED MAGMA SERIES. Murphy, M.D.^ Barclay, J.\ Macdonald, R?, Sparks, R.SJ.^ and CarroU, M.R.^ 1. Geology Department, Bristol University, Bristol, UK. 2. Environmental Science Division, Lancaster University, Lancaster, UK.

The role of amphibole fractionation in influencing intermediate calc-alkaline differentiation trends in subductionrelated magmas has been debated by several authors (e.g. Gill, 1981, Romnick et al., 1992 and refs. therein). In some arcs, amphibole is an abundant phenocryst and its role in fractionation is readily tested. However the rarity or absence of amphibole in many arc volcanics suggests a very limited role for amphibole as an important fractionating phase (Gill, 1981). In some cases, cryptic amphibole fractionation has been invoked to account for geochemical variations among oogenetic magmas (Romnick et al., 1992). The latter proposal is based on the pressure-dependent stability of amphibole. At total pressures less than about 1.5-2 kbar, amphibole becomes unstable so that its absence in erupted magmas is attributed to breakdown during ascent by reaction with the melt to form pyroxene, plagioclase and titanomagnetite. The pressure dependence is ultimately related to melt water content as amphibole requires a minimum of about 4% H2O dissolved in the melt (e.g. Merzbacher and Eggler, 1984). Loss of water during magma ascent results in amphibole becoming unstable. Magma temperature is also an important constraint on amphibole stability. The maximum thermal stability limit of amphibole in calc-alkaline magmas at low pressure (2 kbar) is about 980°C (Sisson and Grove, 1993) and at high pressures (10 kbar) the limit is about 1050°C (Foden and Green, 1992). The discussion below considers the occurrence, stability and role of amphibole in producing geochemical variations at two well-characterised volcanoes in two different subduction settings: Sollipulli volcano in the Southern Volcanic Zone (SVZ) of the Andes and Montserrat volcano in the northern Lesser Antilles island arc. Simple chemical criteria for recognising amphibole fractionation signatures are discussed. The island of Montserrat is situated in the northern part of the Lesser Antilles island arc. The Soufnere Hills volcano, Montserrat, is currently active and is erupting a crystal-rich viscous andesite (56-59% Si02) lava dome and associated pyroclastics. Although the early products of the Soufriere Hills centre were predominantly two-pyroxene andesites, all of the more recent products, while similar in bulk composition to the older andesites, contain an assemblage of plagioclase, amphibole, orthopyroxene and titanomagnetite with minor quartz, apatite and ilmenite. Very minor clinopyroxene occurs only as microphenocrysts and microlites which have probably crystallised at a late stage during magma ascent. The range in whole-rock Si02 contents of the Soufriere Hills rocks is narrow (-56-63% Si02) but mafic magmatic inclusions ranging from high-Al basalt (6% MgO) to basaltic andesite (4% MgO) are ubiquitous in all of the andesitic lavas and the pyroclastics. Whole-rock compositions of the inclusions are very similar to those of basic lavas erupted from the slightly older but partly coeval South Soufriere Hills centre. The mafic inclusions are mostly elliptical with sharp or crenulated margins. Iliey contain phenocrysts of plagioclase only but have diktytaxitic groundmasses with quench-textured plagioclase, amphibole, clinopyroxene, orthopyroxene and titanomagnetite. Evidence for incorporation in a molten state includes the quench textures, partially to completely incorporated crystals derived from the andesite, the presence of chilled margins, crenulated contacts and the elliptical shapes (c.f. Bacon, 1986). Plagioclase of andesitic origin is commonly resorbed and reverse zoned from cores An48_55 to rims up to about Ango- Quench-textured amphibole has high AI2O3 (13-15%) distinct from amphibole of andesitic origin (68% AI2O3). The mafic lavas of the South Soufriere Hills have anhydrous assemblages of olivine, plagioclase, clinopyroxene and minor titanomagnetite. Amphibole-plagioclase and pyroxene thermometry give temperatures between 800-875'^C for the Montserrat andesite. Calculated relative oxygen fugacities are about 1 log unit above the NNO buffer curve (ANNO+1). Ongoing experiments at Bristol on a natural starting composition reproduce the phase assemblage and compositions at 1.5-2 kbar, H20-saturated conditions between 825-875°C at relative oxygen fugacities of ANNO+1. Estimated temperatures of the basic magmas are about 1050®C. Sollipulli volcano is situated at 39°S in the SVZ of the Andes. The volcano has erupted a wide range of magmas from high-Mg high-alumina basalt (9% MgO, 18% AI2O3) to rhyolite (74% Si02). Lavas are predominant but the most recent major eruption at 2,900 BP was a moderate volume explosive event producing a high-Si dacite (67-69%

Si02) pumice fall and ignimbrite. Very minor amounts of banded mafic pumice were also erupted. Mineral assemblages at Sollipulli are predominantly anhydrous, typical of other volcanoes in the southern SVZ. Olivine and Cr-spinel occur in the most primitive rocks, followed by plagioclase and clinopyroxene in magnesian basaltic andesites. Orthopyroxene and titanomagnetite occur in evolved basaltic andesites replacing olivine and Cr-spinel. Andesites and dacites contain plagioclase, two pyroxenes, titanomagnetite, ilmenite and apatite. Magma temperatures calculated using ohvine-liquid and pyroxene thermometry range from about 1180°C in magnesian basalts to about 1120°C in basaltic andesites. Andesites and low-Si dacites range in temperature from about 980-1050°C and high Sidacites range between about 900-1030°C. There is a good negative correlation between crystallinity and magma temperature, the more crystal-poor rocks recording the highest temperatures. The cooler more crystal-rich magmas also show evidence for significant amounts of crustal assimilation during fractionation (AFC). Two-oxide oxybarometry suggests that all magmas have evolved at relative oxygen fiigacities close to the NNO buffer curve. Amphibole is very rare at Sollipulli, occurring as a minor phase (<1% modal volume) in the predominantly dacitic products of two mixed magma eruptions. One sequence of dacitic lavas contains abundant dilaytaxitic-textured basaltic magmatic inclusions (9% MgO) which contain olivine (F087.5) and Cr-spinel only. Some more evolved inclusions (4.5% MgO) contain clinopyroxene, plagioclase and more Fe-rich olivine (F077). Minor amphibole (1113% AI2O3) occurs in the evolved inclusions only. Evidence for incorporation in a molten state is similar to that described above in the Montserrat lavas. Minor amphibole of identical composition to that in the inclusions also occurs in the dacite. Amphibole also occurs in minor amounts in mixed banded and dacite pumice of the 2,900 BP eruption. The banded pumice is formed by very late-stage magma mixing between basaltic andesite and high-Si dacite. In most cases, there is evidence that amphibole has grown by replacement of pre-existing clinopyroxene. In both eruptions, chemical evidence based on Fe/Mg, Al/Si and Ti partitioning between amphibole and melt suggests that the amphibole has grown in evolved basaltic andesite to andesite magmas rather than in the coexisting dacitic magmas. Petrological evidence is also consistent with an origin in magmas less evolved than the high-Si dacite, as amphibole is always associated with relatively calcic plagioclase and magnesian orthopyroxene, compositionally distinct from minerals of dacite origin. Important differences between the Montserrat and Sollipulli systems are summarised as follows. Amphibole is abundant in andesitic magmas at Montserrat whereas clinopyroxene is the dominant calcic mafic phase at Sollipulli. There are large differences in calculated magma temperature at similar Si02 between the two volcanoes. At Montserrat, basaltic andesites and andesites are 100-150°C cooler than equivalent magmas at Sollipulli. The Montserrat magmas are also more oxidised by about 1 log unit relative oxygen fugacity. H2O contents of the Montserrat magmas may also be higher but this is presently unconstrained. Temperature appears to be the most significant difference between the Montserrat and Sollipulli andesite magmas in determining whether amphibole is stable or not. All Sollipulli andesites and many dacites give temperatures which are above the maximum low-pressure thermal stability limit (~980°C) of amphibole whereas the Montserrat andesite magma is well within the amphibole thermal stability field. If the thermometry is accurate, amphibole would not be expected to play a role in low-pressure fractionation at Sollipulli. Even at high pressures, the Sollipulli basalts and basaltic andesites exceed the maximum temperature limit of amphibole (~1050°C). The occurrence of amphibole as a diktytaxitic matrix phase in mafic inclusions at Montserrat is believed to be related to rapid quenching of hydrous mafic magma in a cooler more silicic host liquid. The amphibole is not a phenocryst phase in the mafic magma because temperatures are too high, but forms subsequent to injection of mafic magma as blebs or dykes into cooler crystal-rich more silicic magma. At Sollipulli, major element partitioning criteria suggest that amphibole has formed in intermediate magmas derived byft^actionationof primitive hot (-IIBO'^C) basaltic magma which has intruded a cooler (~900°C) silicic magma chamber. Density contrasts between the basalt and resident dacite initially prevent mixing. The basalt initially crystallises an anhydrous assemblage forming a fractionated hydrous basaltic andesite layer at the interface with the resident dacite. Cooling of this layer to temperatures within the amphibole stability field results in growth of amphibole at the expense of clinopyroxene. Mingling and hybridisation may occur during eruption. Evolution of the Montserrat magmas can be modelled by fractional crystallisation of the observed phenocryst assemblage with early crystallisation of olivine, plagioclase, clinopyroxene followed by plagioclase, orthopyroxene, amphibole and titanomagnetite. The Sollipulli liquid line of descent can also be modelled by fractionation of the

observed assemblage and amphibole plays no role. Amphibole occurs at Sollipulli only in very special circumstances where mafic magmas pond below cooler silicic magmas. Ti is a useful indicator of amphibole fractionation as it is highly compatible with a partition coefficient of about 3-4 in intermediate to silicic melts so that if amphibole is an important phase, particularly with titanomagnetite, Ti will always decrease with fractionation. Analyses of recent (<4000 year) amphibole-bearing Montserrat lavas and pyroclastics, as well as some mafic inclusions, are plotted with two lava sequences from Sollipulli on Figure 1. The Montserrat sequence can be modelled as a cogenetic fractionation sequence with amphibole replacing clinopyroxene as a phenocryst at about 56% Si02. The modelled proportions of amphibole to titanomagnetite are about 6:1. The amount of mafic material (<l-2%) in the andesites has had a very minor effect on their whole-rock compositions. The Sollipulli-1 sequence is a series of mingled dacite lavas and mafic inclusions. The mixing end-members are basalt and high-Si dacite. The amount of mafic material is significant (up to 30%) producing mixing trends on variation diagrams. The Sollipulli-2 sequence is a group of cogenetic lavas which appear to have evolved predominantly by fractional crystallisation.

1.2 - -

O 0.8 0.6-• 0.4--

(To • +

- ^

Sollipulli-1 Sollipum-2 Montserrat

0.2

Si02 Figure l:a: Ti02 against Si02 plot. The Sollipulli-1 samples are a sequence of mixed magmas. The Sollipulli-2 samples represent a sequence of cogenetic lavas which have been generated by fractionation of an anhydrous assemblage. The Montserrat trend shows a more rapid decrease in Ti02 due to the influence of amphibole. Figure l:b: FeO*/MgO plot of the same samples. The Sollipulli-1 mixed rocks form a hyberbolic mixing trend whereas the Sollipulli-2 and Montserrat rocks plot are similar, plotting close to the calc-alkaline-tholeiite boundary field of Miyashiro (1974).

The Montserrat rocks show an early flat trend in Ti02, followed by a continuous decrease with increasing Si02, corresponding to the interval where amphibole is an important fractionating phase. The Sollipulli-2 rocks show an increase followed by a decrease in Ti02 when titanomagnetite and ilmenite become important phases. The Sollipulli-1 mingled rocks form a binary mixing line with a small negative slope, because the end-members have low Ti. The Montserrat amphibole fractionation trend is difficult to distinguish from the Sollipulli-1 mixing trend on the Ti plot although it is clearly distinguished from the Sollipuni-2 fractionation trend. Figure lb clearly distinguishes the Montserrat fractionation trend from the Sollipulli-1 mixing trend. The latter forms a hyperbolic mixing curve whereas the Montserrat and Sollipulli-2 fractionation trends plot at higher FeO*/MgO. These diagrams together are therefore useful for distinguishing between amphibole-dominated from clinopyroxene-dominated fractionation trends and for distinguishing fractionation from mixing trends if the mixing end-members are sufficiently different in composition. Both amphibole and clinopyroxene fractionation produce very similar calcalkaline fractionation trends on the FeO*/MgO plot. These criteria may be particularly useful for distinguishing the potential role of cryptic amphibole fractionation where petrographic evidence for amphibole crystallisation is absent. However the likelihood that cryptic amphibole fractionation is generally important in shallow-level differentiation is questionable. Pressure-dependent breakdown of amphibole produces a characteristic texture where orthopyroxene, clinopyroxene, titanomagnetite and plagioclase

form a very fine-grained reaction rim on the amphibole (e.g. Rutherford and Hill, 1993). If cryptic amphibole fractionation plays an important role in generating shallow-level intermediate differentiation trends, this reaction must go to completion and all trace of the original amphibole must be obliterated. This reaction is time-dependent and requires about 20 days for even a small 100 ]im crystal to completely break down. Very slow magma ascent rates, from depths of about 5-7 km within the amphibole stability field, would be required to permit complete breakdown of larger amphibole crystals. In the Montserrat dome andesite, amphibole shows various degrees of reaction from narrow rims to complete replacement, forming amphibole pseudomorphs consisting mainly of very fine grained intergrowths of pyroxene, plagioclase and titanomagnetite. The pseudomorphs are readily identified by their morphology and by the fine-grained reaction textures. It is difficult to envisage a mechanism by which amphibole can play a major role in high-level fractionation and subsequently break down at shallow depths during ascent while leaving no trace of reaction texture. In summary, the major control on whether amphibole plays an important role in fractionation in subduction-related volcanic sequences is magma temperature, assuming that pressure and melt water contents are sufficient to stabilise amphibole. Magmas erupted in the SVZ of the Andes record high temperatures, in many cases above the maximum temperature at which amphibole is stable. Magmas erupted at Montserrat in the Lesser Antilles are significantly cooler and amphibole is a stable phase in andesites. Ti is highly compatible in amphibole and is a useful indicator of amphibole fractionation as Ti should always decrease if amphibole fractionation is important. Mixing trends may be similar to amphibole fractionation trends but may be distinguished on Fe/Mg against Si02 plots if the mixing endmembers are sufficiently different in composition. Magma temperatures may be controlled by numerous factors such as mantle wedge composition, degree of melting, proportion of water at the site of melt generation ultimately related to slab-flux, extent of fractionation in the arc crust and the amount of crustal material assimilated during differentiation. REFERENCES Bacon, C.R., 1986. Magmatic inclusions in silicic and intermediate volcanic rocks. J. Geophys. Res. 91B, 60916112. Foden, J.D. and Green, D.H., 1992. Possible role of amphibole in the origin of andesite: some experimental and natural evidence. Contrib. Mineral. Petrol. 109, 419-493. Gill, J.B., 1981. Orogenic andesites and plate tectonics. Springer-Verlag, Berlin Heidelberg New York. Merzbacher, C. and Eggler, D.H., 1984. A magmatic geohygrometer: application to Mount St. Helens and other dacitic magmas. Geology 12, 587-590. Romnick, J.D., Mahlburg Kay, S. and Kay, R.W., 1992. The influence of amphibole fractionation on the evolution of calc-alkaline andesite and dacite tephra from the central Aleutians, Alaska. Contrib. Mineral. Petrol. 112, 101118.

Rutherford, M.J. and Hill, P.M., 1993. Magma ascent rates from amphibole breakdown: an experimental study apphed to the 1980-1986 Mount St. Helens eruptions. J. Geophys. Res. 98, 19665-19685. Sisson, T.W. and Grove, T.L., 1993b. Experimental investigations of the role of H2O in calc-alkaline differentiation and subduction zone magmatism. Contrib. Mineral. Petrol. 113, 143-166. ACKNOWLEDGEMENTS The research on Montserrat is supported by NERC urgency grant GR3/10679. The research on Sollipulli forms part of the first author's PhD work supported by EC researcher mobility grant CEG 910688 EPOCH.

ISOTOPE GEOCHEMISTRY OF LAVAS OF CENTRAL JAVA, INDONESIA AND HEARD ISLAND, INDIAN OCEAN - EVIDENCE FOR SEDIMENTARY AND MANTLE PLUME SOURCE COMPONENTS Ian Nicholls^ Danilo Vukadinovic^Jane Barling^ 1 Department of Earth Sciences, Monash University, Qayton, Victoria 3168, Australia 2 6526-94 Street, Edmonton, Alberta T6E 3C8, Canada 3 Geological Institute, University of Copenhagen, DK-1350 Copenhagen K, Denmark

INTRODUCTION A number of active and Pleistocene-Recent volcanic centes of north-central Java, Indonesia, have been the subject of detailed geochemical studies in recent years, and their products have provided information on the nature of magma sources and magma genesis processes in the Java-Bali arc sector and in the arc environment in general. This region of Java is of interest because in addition to "typical" calc-alkaline volcanoes dominated by basaltic andesites and andesites it includes a major stratovolcano, Gunung Slamet, which is dominated by basalts and basaltic andesites with high field strength element (HFSE) abundances (most obviously Ti) which are unusually high by island arc basalt (lAB)) standards (Whitford, et al, 1979; Vukadinovic & Nicholls, 1989; Vukadinovic, 1995), several complexes with high-K calc-alkaline mafic-intermediate products (G. Ungaran - Whitford, et al, 1979; theDieng Plateau group of centres - Sukhyar, 1989) and two extinct Pleistocene complexes (G. Muriah; G. Ringgit-Beser) with shoshonitic to high-K alkaline (including leucite-bearing) products (Nicholls & Whitford, 1983; Edwards, et al, 1991,1994). At much the same time as the recent work on north-central Java, a detailed study of the Heard Island oceanic island volcano, southern Indian Ocean, was under way, and the results have recently been presented (Barling & Goldstein, 1990; Barling, et al, 1994). Heard Island is made up mainly of the products of two volcanic complexes - the major Big Ben active cone and the Laurens Peninsula group of centres. The products, particularly those of the Big Ben cone, are mostly highly mafic oceanic island basalts (OIB), basanites and titanaugite-rich ankaramites (perhaps "limburgites") which are unusual in both major and trace element chemistry, notably in their very high abundances of Ti (up to >6% Ti02) and other HFSE and their extreme Ti02/Al203 ratios (up to --0.6 in primitive basanites with --15% MgO). The Heard Island lavas are also notable in that they show probably the greatest known range of Sr-Nd-Pb isotopic characteristics for products of a single oceanic island volcano (e.g. ^'^ST/^^ST = 0.7047-0.7080; = 0.51274-0.51239), and Sr-, Nd- and Pb-isotope ratios show well-defined linear or hyperbolic correlations. These correlations allowed Barling & Goldstein (1990) to calculate extreme end-member compositions of an apparent two-component isotopic mixing array, corresponding to asymptote values for Sr-Pb and Nd-Pb hyperbolae. These extreme end-members have subsequently been refined and less extreme "preferred" end-members derived, using constraints from trace element abundances, by Barling, et al (1994). Of significant interest is the observation that the "low 87/86" Heard Island end-member, identified with a major component of a past or present mantle plume, is isotopically very similar to the K-rich alkaline lavas of G. Muriah and G. Ringgit-Beser, Java, and also to the less radiogenic Slamet basalts. This observation suggests the existence of a common mantle component (presumably of ultimate asthenospheric origin) in the source regions for the strongly contrasted magmas of north-central Java and Heard Island. Trace element and Sr-Nd-Pb isotopic data for Slamet basalts-andesites indicate that their parental magmas (particularly those related to the older andesites) are affected by small degrees of crustal contamination and that the geochemistry of their mantle sources has been variably influenced by enrichment in LILE and LREE derived from a sedimentary component of the underlying subducted lithosphere (Vukadinovic & Nicholls, 1989; Vukadinovic, et al, in prep.). This contribution presents some of the elemental and isotopic data (including new Pb- isotope data) upon which this interpretation is based, and compares the isotopic compositions of the Slamet lavas with estimates of Heard Island "low 87/86" end-members and also lavas of Muriah and Ringgit-Beser. GEOCHEMICAL FEATURES OF THE JAVA-BALI ARC SECTOR Previous studies of the Sunda arc Java-Bali sector have shown diat ^'^Sr/^^Sr for tholeiitic and calc-alkaline lavas decreases from West Java to Bali. Since the crust thins and changes to oceanic in the same direction, a degree of crustal contamination of mantle-derived magmas is implied. However, when lavas showing clear evidence for contamination are excluded, a weak positive correlation between ^^Sr/^^Sr and depth to Benioff Zone remains (Whitford, 1975). Similarly, weak correlations are present between ^^Sr/^^Sr and K and other elements which increase in abundance with increasing depth to Benioff zone. Whitford et al (1979) proposed a general model which involved, with increasing depth to Benioff Zone, progressive deaease in the degree of melting of mantle 54

wedge material increasingly enriched in LILE (including Rb) and other incompatible elements due to metasomatism by small melt fractions from the basalt/sediment crustal component of the subducting lithosphere. Recent trace-element interpretations of basaltic rocks of Java have also suggested compositional zoning (especially in HFSE) within the mantle wedge beneath Java, attributable to metasomatic events independent of those associated with the subducting lithosphere (Vukadinovic, 1995; Edwards et al, 1991,1994). Early Sr-Nd isotopic data placed the Java-Bali arc sector within or to the high ^^Sxl^^Sr side of the "mantle array" close to bulk Earth, suggesting that on average the mantle wedge beneath that sector is less depleted than typical MORB-source mantle. This suggestion was reinforced by a subsequent detailed study of Slamet lavas, which indicated a mantle source with average E-MORB characteristics (Vukadinovic & Nicholls, 1989). GEOCHEMISTRY OF SLAMET LAVAS Abundant relatively primitive Slamet basalts provide a "window" into the mantle beneath the Java-Bali arc sector. Slamet calc-alkaline to high-K calc-alkaline lavas can be divided into two groups on the basis of relative incompatible element abundances of inferred parental magmas - the "low-abundance magmas" (LAM) group, dominant in the Old Slamet cone, and the "high-abundance magmas" (HAM) group of the active cone. Hie LAM group contains a higher proportion of andesitic rocks; its basalts have higher Zr/Nb and ^^Sx/^^Sr ratios, lower REE contents, and lower Nb/U and Zr/K ratios. But basalts of both groups contain up to 1.4-1.5% Ti02 (Vukadinovic & Nicholls, 1989; Vukadinovic & Sutawidjaja, 1995). 0.5134

0.5132

0.5130

0.5128

• G. Slamet Asymptote values from hyperbolic regressions Heard Island "low 87/86 end member ® Preferred value - Model 1 - Mode! 2

0.5126

0.5124

0.5122

0.5120 0.702

0.703

0.704

0.705

0.706

0.707

0708

0.709

0.710

87sr/86sr Fig. 1: ^'^Sr/^^Sr showing the isotopic compositions of Slamet lavas compared with those of Indian Ocean MORE, OIB and sediments and model Heard Island "low 87/86" end-members. Slamet lavas show significant variation in 87sr/86sr (0.70478-0.70629) and l^^d/l'^Nd (0.5126260.512886), with Nd-isotope ratios close to, but Sr-isotope ratios distinctly higher than those of most other Java groups (Fig. 1). The Sr-Nd data together define a diffusefieldwhich slightly overlaps fields for Indian MORB and OIB on their high ^'^ST/^^ST sides. Both the Slamet field and the high 8'7sr/86sr extension of the OIB field (defined mainly by data for Heard Island), trend toward thefieldof modem Indian ocean-floor sediments in Sr-Ndisotope space (but this is not the case for Heard Island in Pb-isotope space - see below). The most extreme (asymptotic) estimates of the Sr-Nd isotopic composition of the Heard Island "low 87/86" end-member, and a 55

"preferred" composition based upon a model constrained mainly by isotopic and trace element systematics for Heard Island Laurens Peninsula lavas (Model 1 - Barling, et al, 1994), plot very close to the low end of the Slamet Sr-Nd trend. A similarly-derived Heard Island end-member composition constrained mainly by data for the Big Ben cone lavas (Model 2) plots at almost identical to the lowest Slamet value, but also at significantly lower ^^Sr/^^Sr. Slamet lavas generally have high 207pb/204pb (15.66-15.77) and form a steep positive trend on a 207pb/204pb 206pb204p|3 diagram which coincides closely with the field for modern Indian Ocean sediments but overlaps minimally with Indian OIB and not at all with Indian MORB. In a 208pb/204pb v5. 206pb/2(Mpb plot, Slamet lavas, along with other Java groups, overlap with Indian MORB and OIB and also the higher 208pb/204pb part of the modern Indian Ocean sediments field. Hence the Pb-isotopic compositions of Slamet lavas (in fact most lavas of Java) concide closely with those of modem Indian Ocean sediments, suggesting that similar sediments, as a component of subducted oceanic crust, were involved in magma genesis. By contrast, all three Pb-isotopic ratios are significantly lower for Heard Island lavas, although the values overlap with the lower limits of the Slamet ranges (barely so in the case of 208pb/204pb) Barling & Goldstein (1990) and Barling, et al (1994) have defined a "high 87/86" end-member based upon the Heard Island isotopic arrays, and interpreted it as representing old (probably Proterozoic) sedimentary material involved in an ancient subduction episode. 0.710

0.709 0.708

G. Muriah G. Ringgit-Beser G. Siamet Heard Island "low 87/86" end member ® Model 1 s Model 2

Indian Sediment

0.707 CO (O 00

(f) 00

0.706 0.705 0.704 0.703 0.702 15.4

Indian MORB

15.5

15.6

15.7

15.8

207pb/204pb Fig. 2: ^^Sr/^^Sr V5'. ^O^Pb/^^'^Pb plot showing the compositions of Slamet lavas compared with those of Indian Ocean MORB, OIB and sediments, model Heard Island "low 87/86" end-members and lavas of the high-K alkaline volcanic complexes of central/east Java. Relationships between the isotopic compositions of the Slamet lavas and the Heard Island "low 87/86" endmember are best illustrated on a ^^Sv/^^Sr vs. 207pt>/204pb diagram (Fig. 2). Most Slamet lavas (which plot 56

with other calc-alkaline lavas of Java), fall at 207pb/204pb values higher than those of Indian MORE and also most Indian OIB. The Heard Island "low 87/86" end-member compositions have similar 87sr/86sr to Slamet lavas, and although they have distinctly lower 207pti/204pt>^ they still fall close to the least radiogenic Slamet examples. The Model 1 (Laurens Peninsula) end-member composition plots very close to lavas of the two highK alkaline complexes of north-central Java - G. Muriah and G. Ringgit-Beser - particularly the Muriah lavas, some of which (as recognised by Edwards, et al, 1991) have distinct "within plate" geochemical characteristics such as relatively high Nb abundances and small or absent negative Nb-anomalies. DISCUSSION The trace element and isotopic signatures of the Slamet lAB suite of Java and the Heard Island OIB suite indicate that their mande sources contain a component of sedimentary origin. In the case of Slamet, this component is probably related to ocean-floor sediment of Cenozoic age which has been carried as part of the oceanic crust into the Sunda Arc subduction system and dehydrated and/or melted, leading to LILE and LREE (but not HFSE) enrichment by fluids of the mantie wedge from which Slamet parental magmas are derived. In the case of Heard Island, a sedimentary component with characteristics of upper continental crustal material has been added, probably during an ancient subduction event, to either or both of the current mantle plume responsible for triggering Heard volcanism or an older plume which influenced the makeup of the lithosphere beneath Heard Island (Barling et al, 1994). When the "sediment" end-member signatures are subtracted from the source characteristics for Slamet and Heard Island magmas derived from lava compositions, the "mantle" end-member signatures obtained are surprisingly similar, at least in terms of isotopic ratios. This is also the case when the high-K alkaline suites of Muriah and Ringgit-Beser are compared with the Heard Island suite. This implies that a common mantle component may be present beneath north-central Java and Heard Island. Of the various model-dependent "low-87/86" end-members dmved for Heard Island by Barling, et al (1994), the one closest in isotopic characteristics to the lava suites of north-central Java is that d^ved from those authors' Model 1, constrained mainly by data from the Laurens Peninsula lavas. On the basis of trace element systematics, this end-member is believed to be associated with the lithospheric mantle beneath Heard Island, rather than a currendy active mantle plume. However, it may represent "fossil" plume material, added to the Indian Ocean lithospheric mantle during an older period of plume activity. If so, it is tempting to speculate that the plume responsible for this "fossil" component at Heard Island may also have generated a similar component in the mantle which now forms the "wedge" beneath the Sunda arc. Alternatively, a similar deep mantle source may have generated a number of plumes which were involved in various stages of evolution of the Indian Ocean lithospheric mantle prior to the formation of the modern Sunda Arc subduction system. REFERENCES Barling, J. & Goldstein, S.L., 1990. Extreme isotopic variations in Heard Island lavas and the nature of mantle reservoirs. Nature 348, 59-62. Barling, J., Goldstein, S.L. & Nicholls, LA., 1994. Geochemistry of Heard Island (Southern Indian Ocean): Characterization of an enriched mantle component and implications for enrichment of sub-Indian Ocean mande. J. Petrology 35, 1017-1053. Edwards, C.M.H, Menzies, M.A. & Thirlwall, M.F., 1991. Evidence from Muriah for the interplay of suprasubduction zone and intraplate processes in the genesis of potassic alkaline magmas. J. Petrology 32, 555-592. Edwards, C.M.H, Menzies, M.A., Thirlwall, M.F., Morris. J D , Leeman, W.P. & Harmon, R.S., 1994. The transition to potassic alkaline volcanism in island arcs: The Ringgit-Beser complex. East Java, Indonesia. J. Petrology 35, 1557-1595. Nicholls, LA. & Whitford, D.J., 1983. Potassium-rich volcanic rocks of the Muriah complex, Java, Indonesia: Products of multiple magma sources? J. Volcanol Geotherm, Res, 18, 337-359. Sukhyar, R., 1989. Geochemistry and petrogenesis of arc rocks from Dieng, Sundoro and Sumbing volcanic complexes. Central Java, Indonesia. PhD dissertation, Monash Univ., 319 pp. Vukadinovic, D., 1995. High-field-strength elements in Javanese arc basalts and chemical layering in the mande wedge. Mineral Petrol. 55, 293-308. Vukadinovic, D. & Nicholls, LA., 1989. The petrogenesis of island arc basalts from Gunung Slamet volcano, Indonesia: Trace element and ^'^Sr/^^Sr constraints. Geochim. Cosmochim, Acta 53, 2349-2363. Vukadinovic, D. & Sutawidjaja, L, 1995. Geology, mineralogy and magma evolution of Gunung Slamet volcano, Java, Indonesia. J, Southeast Asian Earth Sci. 11, 135-164. Vukadinovic, D., Gari6py, C., Nicholls, LA. & Whitford, D.J., Sr-Nd-Pb isotope composition of lavas from Gunung Slamet volcano, Indonesia (in prep., to be submitted to Volcanol Geotherm. Res,). Whitford, D.J., 1975. Strontium isotopic studies of the volcanic rocks of the Sunda arc, Indonesia and their petrogenetic implications. Geochim, Cosmochim. Acta 39, 1287-1302. Whitford, D.J., Nicholls, LA. & Taylor, S.R., 1979. Spatial variations in the geochemistry of Quaternary lavas acorss the Sunda arc in Java and Bali. Contrib. Mineral Petrol, 70, 341-356.

IN SEARCH OF FLUID - MELT PARTITIONING VALUES; DEVELOPMENT OF THE LASER ABLATION ICP M-S METHOD Geoffrey T. Nichols. Trevor H. Green, Norman J. Pearson GEMOC, School of Earth Sciences, Macquarie University, NSW 2109

INTRODUCTION Many geochemical constraints recently determined either from experimental studies or natural samples, indicate that the genesis of subduction-zone magmas, and particularly those of calc-alkaline compositions, is in some way related to the transport and interaction of source mantle with hydrous fluids. Such water-rich fluids, probably produced by the dehydration of minerals in the subducted slab, are identifiable because they impart specific trace-element signatures that are not produced by other transporting agents such as silicate melts. In some subduction zones for example, sediment-derived silicate melts contribute characteristic trace-element signals and their input is either inferred from experimental constraints (Nichols et al., 1994) or is evidenced by distinctive isotopic signatures (Kay etal., 1978). The geochemical attributes of hydrous fluids or vapour are poorly known, and it is therefore important to determine the partitioning behaviour of trace-elements between vapour and mantle melts, as well as between vapour and mantle minerals. In all but the most recent published work, mineral-melt partition coefficients and mineral-vapour values have been determined independently, and then the vapour-melt partition values have been calculated. In these experiments we have developed a new analytical technique Aat permits us to measure the trace element composition of quenched vapour, as well as quenched melt. EXPERIMENTAL METHODS Experiments were performed with a natural starting material of basanite-i-H20, at temperatures between 8251220°C, and at pressures of 20-30 kbar; most data however, were obtained from experiments at 1200°C and 25 kbar. The basanite from Mt Leura, Queensland, was enriched in the trace elements Rb, Y, Nb, Cs, La, Sm, Lu, Hf, Ta, Th, U which total 1.045 wt%. Experiments were conducted using an end-loaded 12.7 mm piston-cylinder apparatus with either AgsoPdso or AgyoPdso capsules. ANALYTICAL TECHNIQUES AND EQUIPMENT The first step in the analytical process involved the release and immediate analysis of the vapour phase using the Laser Ablation Microprobe (LAM) attached to an Inductively Coupled Plasma - Mass Spectrometer ICP-MS. A detailed description of the LAM ICP-MS instrumentation is given in Norman et al. (in press). Briefly, the laser is a Q-switched, frequency quadrupled Nd:YAG laser, operating at 266 nm (UV). All analyses were performed using a repetition rate of 4 Hz and energy of ~1 mJ/pulse. These conditions produced a sampling area -30 |Lim in diameter. The sample is housed in a sealed cell and the ablated material is transported in a stream of high purity Ar to the ICPMS. The sample cell is mounted on a petrographic microscope fitted with CCD-TV, enabling viewing of the ablation process. Data acquisition was monitored in a real-time graphics display. The laser was used to drill through the AgPd capsule to release the vapour ± liquid, at which time time the laser was turned off The vapour signals were transient, typically < 40 seconds duration. The following trace elements were analysed Be, Ti, Ni, Rb, Sr, Y, Zr, Nb, Ba, La, Ce, ^^'^Sm, ^^^Sm, Ho, Yb, Lu, Hf, Ta, Pb, Th, U with dwell times of 20 ms, using peak hopping mode and with one sweep across the mass-range per reading. After vapour analysis the metal capsules were sectioned and the experimental products exposed by polishing. Major element analyses of primary minerals, spherules and glassmatrix were determined using a Cameca SX50 electron microprobe. The trace element compositions of sufficiently large minerals, spherules, and glass matrix were analysed by LAM ICP-MS under the same conditions described for the vapour analysis, except that a longer dwell time of 50 ms was employed. In these determinations the signals were essentially steady-state, displaying nearly constant counts per second for analysis times between 120 to 185 seconds. In all cases raw counts were quantified using the NIST610 glass as an external calibration standard and "^Ca as an internal standard using the electron microprobe data for CaO. To successfully compare signals for the vapour -h H2O signal it was necessary to select a normalisation factor, as we were not able to provide an absolute concentration for any particular element. In this case we selected a somewhat arbitrary value of Sr=l(X) ppm, which was based on the concentration of Sr in spherules analysed using the proton-microprobe. Selecting other normalisation values, or other normalising elements, does not change the relative shape of the trace-element patterns, but translates the signal

either up or down such that the "absolute" concentrations vary. The accuracy and precision of the quantitative analysis by LAM ICP-MS is discussed in Norman et al. (in press). In this work as an example, we have propagated uncertainties on an individual analysis of a spherule (analysis sph29, run 1627), and as a relative % errors range from -14 to -20% (e.g., Lu, 1150 ± 170 ppm; Pb, 2.54 ± 0.50 ppm). These values include ICP-MS counting statistics as well as the error on the internal standard (CaO) determined by electron microprobe. In this example, the standard deviation on CaO (5.09 ± 0.71 wt%; 13.9% rsd) is the main contribution to the propagated uncertainty on the trace element concentrations. The magnitude of the standard deviation may reflect true compositional variation but may also in part be due to variable Na loss under the electron beam depending on the size and composition of individual spherules. EXPERIMENTAL PRODUCTS Depending on run temperature experiments produced some combination of a quenched-matrix and primary crystals of clinopyroxene, olivine, ± amphibole, in addition to vapour components. The components that represent the vapour, after experiments are quenched, are a complex mixture that include spheroidal shaped solids that are usually -20 |im in diameter (between 5 and 50 |im). We have called these "spherules" and other workers (e.g. Brenan et al, 1995) have observed similar quench products in crystal-vapour experiments and termed them "fish-roe". The vapour component also includes a water-rich solvent, gas vapour (probably a C-O-N mixture), and acicular micron-sized amphiboles or pyroxenes. Unfortunately, these vapour components are the modified products formed during the quenching processes, and so are not the original single vapour phase that was in equilibrium with the basanite during the experiment. The unmixing of experimental vapour ensures that the volumes or proportions of components are poorly constrained, and this provides one of the major complexities in the complete analysis of the equilibrium vapour (Veq) phase. The complexity of the vapour system can be considered using a mathematical representation of the interactions that occur both, during, and after an experiment. First, during an experiment, the total vapour (V) can be expressed as V = Veq + Vdis where Vgq is the equilibrium vapour phase, and V^js represents the vapour component dissolved in the melt. During quench Veq unmixes and V^js exsolves from the melt. The equilibrium vapour, Veq, unmixes into a number of components and may be written as, Veq = Sph -h H2O + gas + Xtal where Sph are spherules formed from quenched Veq, gas is a C-O-N vapour, and Xtal represents acicular microscale amphibole or pyroxene. The exsolution of vapour dissolved in the melt, V^ig, during quench, may be expressed as Vdis - Vex + SphD where Vex is the exsolved vapour component, and Sph^ are spherules formed from vapour that exsolves from the melt. Thus the total vapour in the system after an experiment is V = (Sph + H2O -H gas + Amph) -h (Vex + Sph^) DISCUSSION DECIPHERING THE VAPOUR COMPONENTS Our earlier experiments utilized the laser as a piercing-tool, and the ICP-MS as an accurate analytical device. The results indicated that the vapour removed from the capsules did not change composition significantly across the P-T range of the experiments. This result was unexpected given that some runs were either below or on the basanitesolidus, whereas other experiments were superliquidus. At that time we believed that because subsolidus runs had vapour signals similar to those of superliquidus experiments, we must be measuring a vapour signal dominated by the equilibrium vapour. This vapour therefore, was largely unaffected by the quenching process that could contribute to the measured signal as spherules formed, and as vapour exsolved from melt. Although we did not detect any compositional variation across the temperature range of our runs, we were able to observe a physical change in the quenched vapour. Subsolidus runs have rare spherules and acicular amphiboles, whereas superliquidus experiments have more spherules dispersed throughout the quenched melt. In one experiment, a sphere of melt was encapsulated by a 2 |im thick feldspathic rim, very close in composition to that of the spherules. Spherules occur adjacent to the edge of the 275 |Lim diameter ball of melt, but are predominantly clustered next to three radial cracks in the quenched melt-sphere. This evidence suggests that some spherules form directly from the vapour that exsolves from the melt. Although some spherules must also form entirely from vapour at subsolidus conditions, we have not measured any distinct compositional groups of spherules that may correspond to formation either directly from the vapour, or exsolved from the melt.

In order to understand the quenching processes further, and ultimately to constrain the equilibrium vapour composition, we made two modifications to our experiments. First, we added an oxide mix to the capsule as a layer, designed to produce Mg-phlogopite, and this was separated from the basanite+H20 by a layer of diamond aggregate. These runs were partly successful. They crystallized phlogopite as well as pyrope garnet, but the low viscosity of the basanite allowed it to pass through the permeable diamond aggregate and to mix with the phlogopite "trap". In these runs we were able to analyse the matrix and coarse phlogopites with the LAM ICP-MS and thereby determine basanite-phlogopite KdsIn other runs we added vitreous carbon spheres. They behave much like diamond aggregate, providing porosity between adjacent carbon-spheres, and like diamond are relatively unreactive. The vitreous carbon also has an advantage over diamond aggregate in that it absorbs vapour, and is easy to section or polish. In our initial runs, the analysed carbon-spheres show trace-element patterns almost identical to those of the spherules we analysed in separate runs using identical LAM ICP-MS methods. Because the trace-element patterns of spherules and carbonspheres are identical (Fig. 1), no matter which region of the carbon-spheres we analysed, we infer that this signal must represent that close to the equilibrium vapour, since it could not have formed and penetrated the carbonspheres during the available milliseconds of quench. We conclude that although some spherules must form during the quench process (based on textural evidence cited above), others must represent quenched solute components from primary equilibrium vapour. 105

1—I—I—\—I—I—I—I—I—I—I—I—I—I—I—I—\—I—!—: Spherules & carbon spheres E

104 1000 E Q_ Q-

•—•

E CO (O

100 10

0.1

—csphere — 0 — C sphere — • - - " C sphere • C sphere — ^ — Sph p-probe i —1—Sph V Sph

\\l 1 \ \ j V/ y a "

V7 yy Iff

1 : J

0.01

n ^ \ *!

R b B a T h U NbTa L a C e P b S r S m Z r Hf Ti Y Lu Ni Cr Be Figure L Displays trace element patterns of spherules (Sph) compared with those measured in carbon-spheres (C sphere), determined by LAM ICP-MS, but from different experiments run under similar conditions. The absolute trace element abundances are not well constrained for the carbon-spheres, because there is no reliable internal standard. Also shown is the trace element signal for spherules measured with the proton-microprobe (p-probe). These patterns are dominated by Pb and Ni depletions and relative enrichments in Ti, and Rb. Nb is weakly preferred over Ta, and Hf over Zr. In contrast, the signals we have measured from the vapour component (which equals quenched gas + H2O + Vgx), display almost mirror image trace-element signals in comparison with the spherule or carbon-sphere patterns (Fig. 2). The vapour signals are strongly enriched in Pb and Rb. Nb is variably enriched over Ta, and Zr over Hf. La, Sm, Ti, Y and Lu display variable depletions, and in some experiments were below analytical detection. Figure 2 compares vapour signals with those determined by Keppler (1996). Data from our experiments are broadly similar, showing both Pb and Rb enrichments however, our data have larger Pb/Rb than that of Keppler.

o--- 25825C 25102at; — 0 251100°C 251120°C+phlogmix — i l — 2512(xrc r — 1 — 301220°C 301220--C —

—

— O-

10-5

- KepD!erH20+Cl Kepp!erH20

RbBaTh U NbTa LaCePbSrSmZr Hf Ti Y Lu Ni

Figure 2. Displays vapour signals acquired by capsule-piercing. Key lists run pressure in kbar then run temperature. Vapour signals are normalized so that Sr=100, then divided by the starting composition. Keppler's (1996) data are plotted as signal only. The absolute values of the signals vary, but they have broadly similar profiles. SUMMARY

Our results indicate the following: (1) that the vapour component in equilibrium with melt does not quench as a single phase, but produces at least three discrete phases that include spherules, acicular micron-sized amphibole or pyroxene, and vapour+H20, (2) that spherules are produced in subsolidus as well as liquidus and superliquidus experiments, (3) that the composition of spherules and vitreous carbon spheres is essentially identical, suggesting that both spherules and the signal from carbon-spheres may closely represent the solute composition of equilibrium vapour. REFERENCES

Adam, J., Green, T.H., Sie, S.H., Ryan, C.G., Subm. Trace element partitioning between aqueous fluids, silicate melts and minerals. European Journal of Mineralogy. Brenan, J.M., Shaw, H.F., Ryerson, F.J., Phinney, D.L., 1995. Mineral-aqueous fluid partitioning of trace elements at 900°C and 2.0 GPa: Constraints on the trace element chemistry of mantle and deep crustal fluids. Geochimica et Cosmochimica Acta, 59, 3331-3350. Kay, R.W., Sun, S.-S., Lee-Hu, C.-N., 1978. Pb and Sr isotopes in volcanic rocks from the Aleutian Islands and Pribilof Islands, Alaska. Geochimica et Cosmochimica Acta, 42, 263-273. Keppler, H., 1996. Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature, 380, 237-240. Nichols, G.T., Green, T.H., Pearson, N., Sharma, A., 1996. A new analytical technique for measuring element partitioning between experimental vapour and melt. Geological Society of Australia, Abstracts 41, 317. Nichols, G.T., Wyllie, P.J., Stem, C.R., 1994. Subduction zone melting of pelagic sediments constrained by melting experiments. Nature, 371, 785-788. Norman, M.D., Pearson, N.J., Sharma, A., Griffm, W.L., 1996. Quantitative analysis of trace elements in geological materials by laser ablation ICPMS: instrumental operating conditions and calibration of NIST glasses. Geostandards Newletter (in press). Acknowledgements: We thank A. Sharma for her assistance in operating the Macquarie University LAM ICP-MS system; D. Draper for beneficial conmients on this abstract. 61

LiTHOsPHERic MANTLE SOURCE FOR CONTINENTAL ARC CRUST? TRACE ELEMENT COMPOSITIONS OF DIOPSIDE IN SPINEL LHERZOLITES

Marc D. Norman

GEMOCy School of Earth Sciences, Macquarie University, North Ryde NSW 2109 ABSTRACT. Depletions of high-field strength elements (HFSE) such as Ti, Nb, and Ta relative to other trace elements with supposedly similar incompatibilities in basaltic systems (e.g., REE, Y, and alkalies) are well known but poorly understood features of arc volcanics. Slab-derived fluids are most likely responsible for enrichments of mobile elements such B, As, Sb, Rb, and Cs in arc volcanics (Morris et al., 1990; You et al., 1996), but the HFSE and REE are essentially immobile in slab-derived fluids (Keppler, 1996, You et al., 1996), so that the fractionations of these elements observed in arc volcanics apparently suggest either a unique mantle source compared to ridges or plumes, different conditions of melting beneath arcs, or addition of a silicate melt component from the slab to the mantle wedge. Trace element compositions of diopside in spinel Iherzolite xenoliths from SE Australia also show striking depletions of Nb and Ti relative to Th, REE, and Y (Fig. 1), raising the possibility that the lithospheric mantle in this region may have been created or modified in an arc environment. Altematively, some of these xenoliths may represent appropriate mantle sources for arc magmas, or for the production of continental volcanic arc crust. SAMPLING AND METHODS. A small suite of spinel lherzolites from Mt. Shadwell, western Victoria, has been studied using electron microprobe and laser ablation ICPMS to determine the major and trace element compositions of their constituent minerals. These data provide useful constraints on the processes affecting the lithospheric mande in this region. For diopside, abundances of REE, Nb, Y, Zr, Hf, Th, U, Sr, Ga, Ti, Sc, V, Co, and Ni were determined by laser microprobe. Detection limits ranged from 10 ppb for U and Th, to 2 ppm for Ni. The NIST 610 glass was used for calibration of relative element sensitivities and each analysis was normalized using CaO values determined by electron microprobe as an intemal standard. Calibration values for the NIST 610 glass, and a description of laser operating conditions and error analysis are given by Norman et al. (1996). RESULTS. All of the xenoliths in this study are protogranular spinel lherzolites, derived from the lithosphere. Co-existing pyroxene and spinel compositions indicate equilibration temperatures of 850-950 ^C, corresponding to relatively shallow depths within the lithosphere (25-35 km) by reference to the SE Australian geotherm (O'Reilly and Griffin, 1985). Two groups of lherzolites were found: one which is relatively fertile in bulk composition (diopsiderich) and has trace element patterns indicating these xenoliths have escaped significant metasomatism, and another group which is depleted in bulk composition (diopside-poor) but with trace element pattems indicating cryptic metasomatic enrichment. The fertile lherzolites contain 12-13% modal diopside, corresponding to a bulk composition with 3% CaO. Olivines in these lherzolites are F089.3.90.6, and co-existing diopsides have Mg# of 91-92. Diopsides in these IherzoUtes have LREE-depleted pattems with striking depletions of Nb relative to La and Th, and less severe depletions of Ti and Zr relative to adjacent REE (Figs. 1, 2). These trace element pattems can be modelled by 2-5% batch melting of a primitive mantle source in the spinel facies (Fig. 2). The deep negative Nb anomalies observed in these diopsides appear to be a natural consequence of this melting, based on pubUshed distribution coefficients that indicate Nb is significandy more incompatible in diopside than either La or Th, and nearly as incompatible as Ba (Skulski et al., 1994; Hart and Dunn, 1993). In contrast, the observed depletions of Ti relative to the HREE in these diopsides cannot be modelled using published distribution coefficients (Fig. 2). Ti abundances show well-defined melting trends (e.g., decreasing Ti in diopside with increasing Cr in spinel), and probably cannot be explained by subsolidus equilibration between high-Ca and low-Ca pyroxene.

The modest degree of depletion that has affected these fertile Iherzohtes is similar to that required to produce the Depleted Mantle source of N-MORB, which can be modelled as the residue after extraction of 1-2% continental crust from the primitive upper mantle (Hofmann, 1988). A link between the compositions of these fertile spinel Iherzolite xenoliths and the processes that form continental crust is suggested by the trace element pattem of the melt that would be in equilibrium with these diopsides, calculated from the distribution coefficients. The pattem of this hypothetical melt bears a striking resemblance to that of bulk continental arc crust (Rudnick and Fountain, 1995), including enrichments in LREE/HREE, and a negative Nb anomaly (Fig. 3). In contrast, 2-5% batch melts produced from a primitive mantle source by the melting model described above would have a positive Nb anomalies due to the extreme incompatibility of Nb relative to La and Th in mantle diopside, producing a pattem unlike any^ known magma type. However, prior extraction of a only a very small degree partial melt (e.g. 0.1-0.5%) from the primitive source would be sufficient to impose a negative Nb anomaly on any subsequent magmas. Other highly incompatible elements such as the alkalies would also have been extracted efficiently into this initial small degree melt. This raises the question: how important are small degree partial melts in controlling mantle evolution and the formation of continental crast? Certainly, the negative Nb anomaly of the continental crast and the Nb/Th and Nb/LREE relations between depleted mantle and continental crast, would be difficult to understand if their compositions are controlled simply by percent-level degrees of partial melting, as suggested by a literal interpretation of trace element models for the continental crast and depleted mantle (Hofmann, 1988). If small degree melts are important, where are they in the geological record, and specifically, where is the material with positive Nb anomalies and accompanying enrichments of alkalies and other highly incompatible trace elements? A related question concems the mechanism of upper mantle depletion. Conventional wisdom holds that the Depleted Mantle is the complement to the continental crast. Generation of continental crast in volcanic arcs requires a multi-stage process involving extraction of a primitive basaltic crast from the mantle, segregation of the felsic continental component from this basaltic crast, and recycling and efficient mixing of the mafic residue back into the upper mantle. Recycling of subducted oceanic crast into the lower mantle is thought to be an important process for producing the source regions of mantle plumes, but this produces extreme compositional heterogeneity in plume basalts, in contrast to the global homogeneity of the Depleted Mantle endmember as sampled by N-MORB. The scale of heterogeneities in mantle plumes appears to be on the order of 1-10's of km, based on the compositional variations observed in plume basalts such as those from Hawaii (Lassiter et al. 1996). In contrast, the fertile spinel Iherzolites from Mt. Shadwell display compositional homogeneity among and within individual grains, apparently requiring the re-mixing of mafic residues after continental crast extraction back into the upper mantle on the scale of mm's to |im's, if this is the mechanism responsible for their depletion. While mixing is easier in the upper mantle compared to the lower mantle due to its lower viscosity, further consideration of the role of small degree partial melts in creating upper mantle depletions may be warranted (e.g., O'Nions and McKenzie, 1988; Galer and Goldstein, 1991). The second group of mantle xenoliths in this study are more depleted in bulk composition, with only 2-3% modal diopside, corresponding to < 1% CaO in the bulk rock. Trace element pattems of diopside in these rocks demonstrate a clear signature of cryptic metasomatic enrichment of LREE, Th, U, Sr, and Nb. Although the normalized trace element pattems of these diopsides retain a deep negative Nb anomaly, the absolute concentrations in these diopsides are 3-1 Ox greater than that measured in diopside from the fertile Iherzolites (Fig. 1). Mass balance models show that the metasomatism produced an absolute enrichment of Nb, Th, and LREE in the rock compared to the fertile Iherzolites, and that the high concentrations in the

metasomatized diopsides are not simply a dilution effect caused by the presence of less total diopside in the rock. Although a subduction-related origin for the metasomatism has been suggested for some amphibole-apatite Iherzolites of southeastem Australia (Griffin et al., 1988), Nb is virtually immobile in slab-derived fluids (Keppler, 1996, You et aL, 1996) and is essentially unaffected by subduction-related fluxes into arc magmas (Pearce et al., 1995). In contrast, the addition of Nb into the metasomatized IherzoUtes studied here shows that Nb was mobile during this metasomatism. Altematively, a plume source for the cryptic metasomatism present in these Mt. Shadwell Iherzolites may be possible. High Sm/Hf and Zr/Hf, and low Ti/Nb ratios in the metasomatized diopsides suggests that a carbonatitic fluid or melt may have been responsible for the trace element enriclmient, which may also be consistent with a plume source for the metasomatism. CONCLUSIONS. Trace element patterns of diopside in relatively fertile spinel Iherzolite xenoliths from Mt. Shadwell, western Victoria, are LREE-depleted and have striking depletions of Nb and Ti relative to the REE, raising the possibility that they may have evolved in an arc environment, or represent a suitable source for arc basalts. The Nb depletions can be accounted for by extraction of a relatively small degree (2-5%) batch melt, but the Ti depletions are difficult to explain by melting using published distribution coefficients. A melt in equilibrium with these diopsides would have a trace element pattem similar to that of bulk continental arc crust, although a prior episode of melting appears necessary to avoid Nb enrichments relative to Th and LREE in the melt. Cryptically metasomatized xenoliths from Mt. Shadwell are depleted in bulk compositions (i.e. diopside-poor). Diopsides from these xenoliths have trace element pattems indicating addition of Nb, LREE, Th, U, and Sr to these rocks. This may be difficult to account for by a subduction-related process because of the immobility of Nb in slab-derived fluids. Altematively, a carbonatitic fluid or melt derived from a plume source may have been responsible for the cryptic metasomatism observed in this group of xenoliths. REFERENCES Galer SJ.G. and Goldstein S.L., 1991, Early mantle differentiation and its thermal consequences. Geochim. Cosmochim. Acta 55, 227-239 Griffin W.L., O'Reilly S.Y., and Stabel A., 1988, Mantle metasomatism beneath western Victoria: 11. Isotopic geochemistry of Cr-diopside Iherzolites and Al-augite pyroxenites. Geochim. Cosmochim. Acta. 52, 449459 Hart S.R. and Dunn T., 1993, Experimental cpx/melt partitioning of 24 trace elements. Contrib. Mineral. Petrol. 113, 1-8 Hofmann A.W., 1988, Chemical differentiation of the Earth: the relationship between mantle, continental crust, and oceanic crust. Earth Planet Sci Lett 90, 297-314. Keppler H., 1996, Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380, 237-239. Lassiter J.C., DePaolo D.J., and Tatsumoto M., 1996. Isotopic evolution of Mauna Kea volcano: results from the initial phase of the Hawaii Scientific Drilling Project. J. Geophys. Res. 101, 11769-11780 Norman M.D., Pearson N.J., Sharma A., and Griffm W.L., 1996, Quantitative analysis of trace elements in geological materials by laser ablation ICPMS: instrumental operating conditions and calibration values of NIST glasses. Geostandards Newsletter, in press. O'Nions R.K. and McKenzie D.P., 1988, Melting and continent generation. Earth Planet Sci Lett 90, 449-456. O'Reilly S.Y. and Griffm W.L., 1985, A xenolith-derived geotherm for southeastern Australia and its geophysical implications. Tectonophysics 111, 41-63 Pearce J.A., Baker P.E., Harvey P.K., and Luff I.W., 1995, Geochemical evidence for subduction fluxes, mantle melting and fractional crystallization beneath the South Sandwich island arc. J. Petrol. 36, 1073-1109. Rudnick R.L. and Fountain D.M., 1995, Nature and composition of the continental crust: a lower crustal perspective. Reviews of Geophysics 33, 267-309 Skulski T., Minarik W., and Watson E.B., 1994, High-pressure experimental trace-element partitioning between clinopyroxene and basaltic melts. Chem. Geol. 117, \H-\A1 You C.-F., Castillo P.R., Gieskes J.M., Chan L.H., and Spivack A.J., 1996, Trace element behavior in hydrothermal experiments: implications for fluid processes at shallow depths in subduction zones. Earth Planet Sci Lett 140, 41-52.

100 <D

C E 10 o>

E CO < 0 0.1

1—I—I i I I I I i I I

1—r

Figure 1. Trace element patterns of diopside in mantle xenoliths from Mt. Shadwell, SE Australia. All data are by laser ablation ICPMS analyses. Two compositional groups are evident. Fertile Iherzolites have diopside with LREE-depleted patterns, whereas diopsides from depleted Iherzolites reveal cryptic metasomatic enrichment of incompatible trace elements. Note the depletions of Nb relative to Th and the LREE in both groups, and the absolute enrichment of Nb in the Th Nb La CeSrNdZr Hf SmEuTi Gd HoY Yb Lu metasomatized diopsides.

10.0

; •J

E <D

CD Q.

E CO CO

1 ' jr^g^^^A/ -

0.1

100 "D

i l l 1 1 1 1 1 1 11 1 1 11 1 1 1 1 TTi La Sr Zr Sm Ti Dy Er Yb Nb Ce Nd Hf Eu Gd Ho Y Lu

Figure 2. Compositions of diopside from fertile Mt. Shadwell Iherzolites (black) compared to melt compositions p loducedb y 1-5% batch melting of a primitive mantle source (dashed). Distribution coefficients after Hart and Dunn (1993) and Skulski etal. (1994). The measured pattems of the diopsides, including the Nb anomalies, are well matched by the melting model except for the relative depletion of Ti.

1—rn—n—i i i melt in equilibrium with cpx

0) "co oc 0 c

1 1 1 1:

1' ' . f ' ' •" iI*; '•/'r 5% • IM'' \l 1 llf' • |v/'

E V.

1 1 1i l l !

CO E

5 E

"i-

Th U Nb La Ce Pb Sr Nd Zr Hf Sm Eu Ti Dy Ho Y Er Yb Lu

Figure 3. A melt in equilibrium with diopside from the fertile Mt. Shadwell Iherzolites would have a trace element pattern similar to that of the continental arc crust (Rudnick and Fountain, 1995). Pb enrichment in the crust may indicate addition from a fluid phase.

Ground Truthing Arc Melting Models: Peridotite Geochemistry from Forearc to Backarc I, J, Parkin$on^ R.J. Arculus^ ^GEMOC, Dept. Geology, Australian National University, Canberra, ACT 0200, Australian

Melt composition and volumes have been successfully calculated in mid-ocean ridge (MOR) and intraplate settings using decompression melting models. Theoretical considerations of melting indicate that melt extraction occurs at small melt fractions (<1%) and that the process can be modelled by near-fractional melting. The trace-element content of residual peridotites is a sensitive indicator of both the type of melting (fractional versus batch), the total amount of melting and how much melting occurs within different peridotite facies and also whether melt/mantle interaction is an important process. Trace element geochemistry of abyssal peridotites, which are thought to be the residua to melting at MOR, and xenoliths from intraplate settings are consistent with theoretical models with melting being a polybaric fractional process. Therefore, residual peridotites are more sensitive recorders of melting processes than melts. In contrast, applying such models to melting in the mantle wedge above subduction zones is less easy because we lack knowledge in many of the important parameters for modelling. Currently three end member models exist for generation of melts within the mantle wedge: 1) Decompressional melting models similar to those beneath MOR but derived by melting of a hydrous mantle and with the thickness of the lithosphere being a control on the amount of melt generated (Plank and Langmuir, 1988; Pearce and Parkinson, 1993). 2) Fluidfluxing models with water derived from subducted slab lowering the melting point of the mantle and inducing melting (Davies and Bickle, 1991; Stolper and Newman, 1994). 3) Mantle/melt interaction models whereby the chemistry of the melts generated relates to the amount and style of melt interaction (Kelemen et al., 1990). These different models predict a variety of compositions for the residual peridotites, including variations in the total extent of melting, melting with residual amphibole and spinel and a variable amount of mantle/melt interaction. The only common denominator of these models is that subducted lithosphere ultimately plays an essential role in the generation of melts, and the major melt sources in the mantle wedge are likely to be more oxidised than in other tectonic settings. Given the success in using peridotite geochemistry in other tectonic settings, we have undertaken a study of a variety of suites of peridotites from several different island-arcs. We have analysed peridotites:- 1. drilled from the Izu-Bonin-Mariana forearc system; 2. exposed within the Solomon Islands; 3. mantle xenoliths from Grenada, (Lesser Antilles); 4. and xenoliths from Vanuatu. Our strategy involves major and trace element analysis of whole-rock samples coupled with standard electron microprobe analyses of minerals and trace element analyses of clinopyroxene, amphibole and orthopyroxene using Laser-Ablation-Inductively-Coupled-PlasmaMass-Spectrometry (LA-ICP-MS) techniques. Major element mineral chemistry is used to calculate temperatures and oxygen fugacities whereas trace element chemistry is used to elucidate melting histories of the peridotites. Very refractory harzburgites are exposed on the seafloor of the Izu-Bonin-Mariana forearc system by a series of serpentinite diapirs. Harzburgites from Torishima Forearc Seamount in the Izu-Bonin forearc are interpreted to be residues from melting within the mantle wedge. These peridotites are characterised by very refractory mineral compositions (Cr# spinel 0.50-0.70), low-clinopyroxene contents (<2%), minor amounts of modal amphibole and very low incompatible trace element concentrations. Rare earth element (REE) patterns for the clinopyroxenes have rather flat HREE-MREE patterns with variable LREE-enrichment. These data can be modelled as residues to 20-25% near-fractional melting within the spinel peridotite facies, possibly with some residual amphibole during the early stages of melting. The LREE-enrichment in these samples is thought to represent limited interaction with a LREE-enriched melt with the mantle wedge. The small amounts of amphibole within the peridotites are interpreted to have crystallised from a hydrous melt passing through the peridotite. Oxygen fugacities recorded by these peridotites are 0.8-1.8 log units above the fayalite-magnetitequartz (FMQ) buffer consistent with other subduction zone peridotites (Wood et al., 1990). Peridotites from Conical Seamount in the Mariana forearc are somewhat different in composition. The harzburgites are also very refractory, but clinopyroxenes have trace element signatures consistent with melting in the garnet field followed by melting in the spinel field. Furthermore, these peridotites record oxygen fugacities - 1.1 log units below to 0.4 log units above the FMQ buffer. These harzburgites are interpreted as residual mantle fragments of the Pacific plate which have been tectonically emplaced into the Mariana forearc. However, these harzburgites are cut by dunitic veins. These veins contain resorbed orthopyroxene and rare clinopyroxene and amphibole. Moreover, they record oxygen fugacities of 0.5 to 1.2 log units above the FMQ buffer. The dunites are interpreted as veins where oxidising subduction zone melt have reacted and then chanelled flow through the harzburgites. Similar dunitic veins has been reported in the Oman ophiolite by Kelemen et al. (1995), who argue that melt migration via channelised flow is dominant beneath MOR. Therefore it can be concluded that this type of process can also occur within the mantle wedge.

Grenada in the southern Lesser Antilles island arc dominantly erupts alkalic, low-silica, picritic lavas. Included in the M-series of Grenada are rare harzburgite, wehrlitic and dunite xenoliths. These xenoliths contain evidence for extensive melt/mantle interaction. The principal melt/mantle interaction is the reaction of orthopyroxene with relatively high pressure, low-silica melts to generate clinopyroxene, olivine and a high-silica melt. The melt is quenched as glass in some xenoliths and is characterised by Si02 contents of 58-60%, high alkali contents but is in equilibrium with mantle olivine. Where the melt interaction is pervasive, the effect is to transform harzburgites into wehrlites. The other effect of melt interaction in these samples is an increase in the oxygen fugacity from an already oxidised FMQ+1 log units to ~ FMQ+2 log units. These xenoliths provide clear evidence that migrating melts are effective oxidising agents. Although these xenoliths provide ample evidence for extensive mantle/melt interaction, it is clear that very few lavas with compositions of the highsilica glasses are erupted on Grenada. A more likely sequence of events is that generation of wehrlite protoliths within the lithospheric mantle is important in the genesis (through partial melting of wehrlite) of the very Carich nature of the GrenadaC-series (anakaramitic) lavas. Similar, wehrlitic rocks have been analysed from Vanuatu. Again this is an arc which erupts predominantly low-silica picritic lavas. Therefore the formation of wehrlites from lithospheric harzburgites may be a common feature of these types of arcs. The Solomon Islands represent one of the few arcs where relatively large terrestrial exposures (e.g., thrust sheets, serpentinite mounds) of mantle-derived peridotites occur. Two suites of arc-related peridotites have been discovered on the islands of Santa Isabel and San Jorge. On Santa Isabel, the peridotites include amphibolebearing harzburgites, harzburgites cut by diopside-pargasite veins and impregnated harzburgites containing plagioclase-clinopyroxene schlieren with pargasitic coronas. We interpret the petrography of these samples to indicate that they have interacted with melts at shallow mantle pressures. The spinel chemistry indicates that they original peridotites were refractory and had melted 20-25%. REE patterns for the clinopyroxenes are all rather flat to slightly LREE-depleted. HREE contents of the clinopyroxenes are consistent with high degrees of near fractional melting whereas the LREE contents indicate extensive interaction with a influxing melt. Oxygen fugacities of these peridotites are in excess of 1 log unit above FMQ. Peridotites from the island of San Jorge are exposed by an onland serpentinite diapir. Clasts of fresh harzburgites, orthopyroxenite and clinopyroxenites have been brought to the surface by the diapir. Trace element geochemistry of the the harzburgites indicates that they are residues to high degrees of partial melting which can be modelled by near fractional melting in the spinel peridotite field only. The pyroxenites are all LREE-depleted and were in equilibrium with LREE-depleted melts. The oxygen fugacities of the harzburgites and pyroxenites are from FMQ to FMQ+0.5 log units. On the basis of the geochemical data, we suggest that these peridotites may have originated as residues to back-arc basin melting. The slight oxidation, LREE-enrichment and high degrees of melting in the spinel field alone are all consistent with this type of origin. In support of this conclusion is the fact that these peridotites are spatially associated with transitional MORB to back-basin basalts and two-pyroxene gabbros. Some important conclusions can be drawn from the peridotite data. Firstly, many of the peridotite record high degrees of partial melting (>20%). In contrast to refractory peridotites from MOR and intraplate settings, where large amounts of melting are associated with high mantle potential temperatures and therefore some melting occurs in the garnet peridotite facies, subduction zone peridotites record melting in the spinel facies only. Some of the peridotites also have signatures consistent with melting with residual amphibole during the initial stages of melting. The large degrees of partial melting recorded by these peridotites are consistent with the addition of water to the mantle wedge which decreases the solidus of peridotite and initiates melting. The data are consistent with two possible models:- 1. a situation whereby melting is first induced by fluid-fluxing and is followed by melting of decompression (diapiric) melting of anhydrous peridotite (e.g., Pearce and Parkinson, 1993); 2. a fractional fluid-fluxing model where water is continuously fluxed through the melting region (e.g., Stolper and Newman (1994). Many of the peridotites are consistent with near fractional melting indicating that melt extraction processes within the mantle wedge are efficient and in many respects similar to those beneath MOR. Dunites recovered from the Mariana forearc provide evidence that melt migration is by chanelled flow rather than pervasive melt percolation. These observations are also consistent with U-Th isotope data which indicate rapid transit times of melts across the mantle wedge (Gill et al, 1993). Some subduction zone peridotites do record melt-mantle interaction ranging from rather limited interaction in the Solomon Island peridotites to extensive reaction in the Grenada peridotites. The amount of interaction reflects the thickness of the lithospheric mantle that the melts have to travel through, and whether the arc system is undergoing extension. Arcs built on thick lithosphere, such as continental arcs, are likely to be susceptible to more melt/mantle interaction. Intra-oceanic arcs such as the Izu-Bonin Mariana system, which has undergone periods of extension throughout its history, are least likely to have extensive melt/mantle interaction and melt extraction and migration will be similar to that beneath MOR. Finally, a ubiquitous feature of subduction-zone peridotites is their oxidised nature. All of the peridotites analysed in this study have oxygen fugacities in excess of 0.5 log units above the FMQ buffer and range up to

2.5 log units above the buffer. There is no obvious correlation between the amount of oxidation and the presence of hydrous phases such as amphibole or phlogopite. Simple partial melting and phase equilibria calculations indicate that the mantle at depth will have an oxygen fugacity more oxidised than FMQ and a bulk Fe203 greater than a primitive mantle value of 0.3%. The simplest explanation for the intrinsically oxidised nature of the mantle wedge is that water derived from the subducting slab is the ultimate origin of the oxygen. At higher pressures water dissociates to hydrogen and oxygen and is a more efficient oxidising agent. Whether the oxygen is carried in a water-rich supercritical fluid or as ferric iron in a melt is still debatable, although it is clear from magma/xenolith reactions that melts are efficient oxidising agents. REFERENCES Davies, J. H., & Bickle, M. J., 1991. A physical model for the volume and composition of melt produced by hydrous fluxing above subduction zones. Philosophical Transactions of the Royal Society of London A335, 355-364. Gill, J. B., Morris, J. D., and Johnson, R. W., 1993. Timescale for producing the geochemical signature of island arc magmas: U-Th-Po and Be-B systematics in recent Papua New Guinea lavas. Geochim. Cosmochim. Acta 57, 4269-4283. Kelemen, P. B., Kinzler, R. J., Johnson, K. T. M., & Irving, A. J., 1990. High field strength element depletions in arc basalts due to mantle-magma interaction. Nature 345, 521-524. Kelemen, P. B., Shimizu, N., & Salters, V. J. M., 1995. Extraction of mid-ocean-ridge basalt from the upwelling mantle by focused flow of melt in dunite channels. Nature 375, 747-753. Pearce, J. A., & Parkinson, I. J., 1993. Trace element models for mantle melting: application to volcanic arc petrogenesis. In Prichard, H. M., Alabaster, T., Harris, N. B. W., and Neary, C. R. (eds.) Magmatic Processes and Plate Tectonics. Geological Society of London Special Publication 76, 373-403. Plank, T., and Langmuir, C. H., 1988. An evaluation of the global variations in the major element chemistry of arc basalts. Earth and Planetary Science Letters 121, 349-370. Stolper, E. M., & Newman, S., 1994. The role of water in the petrogenesis of Mariana trough magmas. Earth and Planetary Science Letters 121, 293-325. Wood, B. J., Bryndzia, L. T., and Johnson, K. E., 1990. Mantle oxidation state and its relationship to tectonic environment and fluid speciation. Science 248, 337-345. Acknowledgements Steve Eggins and Les Kinsley are thanked for their continued assistance in aquiring high-quality trace element data using the LA-ICP-MS at RSES. Finacial support for this work was provided by the Australian Research Council (grant to RJA), and by the Natural Environment Research Council and The Royal Society (grants to UP).

TRACE ELEMENT BEHAVIOUR IN ARC SYSTEMS Julian A. Pearce J.A.Pearce@durhain.ac.uk Dept. of Geological Sciences, University of Durham, Durham DHl 3LE, UK INTRODUCTION Despite a rapidly-increasing database on the trace element contents of inputs (mantle wedge, subducted lithosphere and sediment) and outputs (arc magmas) to arc systems, and clear evidence for links between the two, the deep processes that fractionate trace elements remain diflBcult to define. These processes include: slab dehydration and melting; transport of the subduction component; melting column processes; mantle source processes; and lithosphere-melt interaction. The variables contributing to each component, and discussed in the text, are illustrated in Figure 1 below (from Pearce and Peate, 1995).

FIGURE 1. Schematic summary of the processes affecting arc magma composition. Key to numbered processes. Slab dehydration and melting: D fluid loss via accretionaiy prisms and seipentinite seamounts; 2) dehydration ± partial melting of subducting oceanic crust and sediment. Transport of Subduction Component: 3) hybridization of subduction component and down-dragged mantle lithosphere; 4) slab-induced downward drag of hybridized mantle; 5) re-release, and lateral migration of, aqueous fluids by amphibole breakdown at about 100km depth; 6)re-releaseof aqueous fluids by breakdown of other hydrous phases at greater depth; 7) initiation of hydrous melting of mantle at about lOOO^'C; 8) migration of small-volume hydrous melts through cross-fed mantle to the base of the melting column. Mantle Source Processes: 9) slab-driven 'comerflow'of mantle into the mantle wedge; 10) replenishment of the melting column by mantle advection; 11) mantle source depletion by small volume melt loss in bxk-arc region; 12) mande source enrichment by delamination of sub-continental hthosphere. Melting Column Pnx^esses: 13) buoyancy-driven mantle counter-flow; 14) decompression-melting of the mantle from about 60km depth; 15: separation of residual mantle from the melting column aided by slab- induced downward drag; 16) column depletion by imperfect separation, or re-incorporation, of residual mantle; 17) selective tapping of the melting column. Melt-Lithosphere Interactions: 18) melt segration at the base of the lithosphere; 19) interaction with, and crystallization within, mantle lithosphere; 20) magma-assimilation-storage-homogenization (MASH) at the base of the crust; 21) assimilation-fractional crystallization at shallower crustal levels.

DEHYDRATION AND MELTING OF THE SUBDUCTING SLAB Most empirical and theoretical studies indicate that the subducted oceanic crust and/or its sedimentary cover can dehydrate (produce hydrous fluids) or melt (produce siliceous fluids) according to the thermal state of the subduction system: dehydration is probably the 'norm' but crust and sediment melting can take place in association with ridge subduction or subduction beneath ridges and sediment melting may be common during arc-continent collision. Key points related to element behaviour include: * Experimental, empirical and theoretical studies indicate that this process plays a major role in the decoupling between large ion Uthophile elements (LILE) and high field strength elements (HFSE), LILE partition more strongly into the fluids while HFSE partition more strongly into the residual subducting crust and sediment. * Aqueous fluids decouple elements primarily on the basis of ionic potential and the nature of the fluid-forming reactions. In contrast, siUceous fluids decouple elements primarily on the basis of partitioning with respect to hydrous, acid magma, which is in turn controlled by the residual phase assemblage. In addition to LILE/HFSE ratios, a number of intra-LILE ratios (e.g. Ba/La and UyTh) and intra-HFSE ratios (e.g. ZxlY) are likely to be sensitive to whether the fluid is aqueous or siUceous * For slab melting (and even more for slab dehydration), precise modelling is difficult because of wide range of temperatures, fluid and melt compositions and residues, because of the uncertainties in whether melting is batch or fractional, and because of uncertainties in the extent of equilibration between the fluid phase and the residue. Recent experiments (collaboration with Rushmer, Ottolini and Bottazzi) and field studies indicate that there may be a strong kinetic control on the trace element composition of subduction-derived melts. •The behaviour of HFS elements may be just as important as that of LILE in understanding subduction systems: if they can be shown to be completely conservative (not added to the wedge) they can be used to 'see through' the subduction process to irrvestigate processes in the mantle wedge. A recently completed Hf-Nd isotope study (collaboration with Nowell, Kempton and Noble) demonstrates that Hf at least can be considered conservative when aqueous fluids are the dominant mode of transport from slab to wedge. However, Hf can be as, or more, mobile than Nd when siliceous melts are the transport medium. * The Hf isotope study supports experimental and empirical studies of slab melting which show that Hf can be more incompatible than Nd with an eclogite residue. The experiments also indicate that Zr, Nb, Ta and to a lesser extent Ti and Y are essentially conservative during dehydration but non-conservative to varying degrees during melting. TRANSPORT OF THE SUBDUCTION COMPONENT TO THE MELTING COLUMN. Most scientists support the two-stage process of Tatsumi et al. (1992) in which fluids first hydrate the mantle above the slab to form hornblende peridotite and are then re-released when the amphibole breaks down to garnet as the mantle is dragged to depths of 100 km. These fluids then migrate laterally to the zone of melting. Key points related to element behaviour include: * Transport of the subduction component theoretically provides an additional opportunity for trace element fractionation Once two phases are present, Navon and Stolper's 1987 model of a cross-fed chromatographic column may apply. Effectively there are three colimms to model: of aqueous fluids laterally through a gamet/homblende peridotite matrix; of meh laterally through a gamet/homblende peridotite matrix; and of hydrous melt sub-vertically through a garnet peridotite matrix. The general model is of element fractionation Unked to differencesin transport velocity which are in turn linked to diJBFerencesin bulk distribution coefficients and to mantle and fluid flow rates. * The big question is the extent to which this process modifies observed trace element ratios and this is not yet resolved. Strong evidence that it may take place comes from the 'lead paradox' presented by Hawkesworth et al. (1993): Pb/HFSE ratios show that the apparent % mantle contribution to the lead concentration of island arc basalts is typically 20% or less, yet these same lavas exhibit isotope ratios that require a much higher mantle contribution to its Pb. This paradox, they argue, can be resolved if mantle lead is exchanged with, or selectively extracted from, the mantle during transport. * One observation this process would resolve is that Y and the HREE are the most conservative of all incompatible trace elements, even when there is good evidence for slab melting: the transport equations strongly favourtheir retention in the mantle even if they are in the subduction component.

PROCESSES LINKED TO MANTLE MELTING. Experiments, geochemical modelling and studies of mantle peridotites all suggest that mantle melting beneath subduction zones is the product of two principal processes: addition offluids and decompression. Fluid addition is probably the principal process deeper than 60km and appears to cause some 10% of melting for every 0.2 wt.% of water added (an approximate arc average), and decompression is the principal process shallower than 60km, contributing about 1% of melting for every 4km of decompression within this depth range (Pearce & Parkinson, 1993). Key points related to element behaviour include: * The magma itself is unlikely to be hydrous enough to decouple the LILE from the HFSE. However, it is hydrous enough to affectoxygen fugacity (by some 2 log units) which in turn affects some transition elements. For example, the higher vanadium contents of island arcs conq)ared with MORB may be attributed to oxygen fiigacity diflferencES, as and V^ are more incompatible than * The melting colmnns beneath arcs should generate more melt in gamet facies than those beneath mid-ocean ridges because of the higher water content of the mantle. The 'gamet signature' (high MREE/HREE ratios) of many MORBs is, however, seldom observed in arc basalts, presumably because the higher degree of melting also means that the proportion of residual gamet is much smaller. The higher overall degree of melting compared with other environments is probably the principal cause of the lower absolute levels of HFSE in arcs compared with ridges and other settings. * Several arcs,such as the Scotia Arc, show trace and major element evidence for dynamic melting, in which some volcanoes (the Na-, K- and Fe-rich) represent the preferential pooling of deep melt fractions while others (the Na-, K- and Fe-poor) represent the preferential pooling of deep meh fractions. The role of the melting process in decoupling element ratios of LILE to HFSE of different incompatibilities (i.e Ba/Zr, though not Ba/Nb) is probably greatly underestimated. * Incompatible trace element depletion in arcs close to back-arc basins may be explained in terms of loss by the mantle of a small melt fraction (e.g., c. 2.5% in the Scotia system) before it reaches the sub-arc melting column. MELT-LITHOSPHERE INTERACTIONS These can play a significant role in modifying the composition of mantle-derived melts. Key points related to element behaviour include: * Studies of forearc peridotites (collaboration with Parkinson) provide clear evidence for chemical interactions between migrating melt and depleted mantle lithosphere. The process may lower the Cr content of some island arc lavas, although the effecton the more incompatible trace elements is less clear. There is better evidence that reactions with enriched mantle lithosphere may help explain trace element enrichments in some arc settings. * Geological and geochemical evidence both point to extensive interaction (MASH and AFC) between arc magma and crust, even in oceanic arcs. This process can contribute to the decoupling of LILE and HFSE. However, subduction-induced decoupling differs significantly from contamination-induced decoupling on other parameters, including oxygen isotopes. REFERENCES Navon, O. & Stolper, E. M., 1987. Geochemical consequences of melt percolation: the upper mantle as a chromatographic column. J. Geol 95, 285-307. Hawkesworth, C.J., Gallagher, K., Hergt, J.M. & McDermott, F., 1993. Trace element fractionation processes in the generation of island arc basalts. Phil Trans. R. Soc. Lond. A342, 171-19L Pearce, J.A. & Parkinson, I.J., 1993. Trace element models for mantie melting: application to volcanic arc petrogenesis. Geol. Soc. Lond. Spec. Publ. 76, 373-403. Pearce, J.A. & Peate, D.W., 1995. Tectonic implications of the composition of volcanic arc magmas. Annu. Rev. Earth Planet. Sci. 23, 251-285./ Tatsumi, Y., Murasaki, M., Arsadi, E.M. & Nohda, S., 1991. Geochemistry of Quaternary lavas from NE Sulawesi: transfer of subduction components into the mantie wedge. Contrib. Mineral. Petrol 107, 137-49. Acknowledgements. I am grateful to the many co-authors and co-workers who have contributed to this work but space precludes including ontiieautiior list. They include: Dave Peate, Ian Parkinson, Tracy Rushmer, Luisa Ottolini, Piero Bottazzi, GeoffNowell, Pamela Kempton, Steve Noble, Sarah Acland, Peter Baker, Phil Leat, Steve Edwards, Chris Ottiey and Richard Arculus.

RADIOGENIC AND U-SERIES ISOTOPE VARIATIONS IN VANUATU ARC LAVAS: FLUID AND SEDIMENT ADDITION TO AN ISOTOPICALLY HETEROGENEOUS MANTLE WEDGE David W. Peate^. Chris J. Hawkesworthl, and Julian A. Pearce^. d.w.peate@open.ac.uk: c.j.hawkeswoith@open.ac.uk: j.a.pearce@durham.ac.uk 1 Department of Earth Sciences, The Open University, Milton Keynes, MKT 6AA, United Kingdom. 2 Department of Geological Sciences, University of Durham, Durham DHl 3LE, United Kingdom. ISOTOPICALLY-DISTINCT DOMAINS WITHIN THE VANUATU MANTLE WEDGE An important influence on the recent tectonic evolution of the Vanuatu island arc in the south-west Pacific has been the collision of the D'Entrecasteaux Zone with the central region of the arc. The D'Entrecasteaux Zone is an EoceneOligocene island arc complex on the subducting plate, and the collision began between 1.5 & 3 Ma. Strong coupling between the subducting slab and the overriding arc has produced extensive tectonic disruption and uplift in the central Vanuatu arc. Intriguingly, certain compositional features of the arc lavas very systematically north-south along strike, and seem to be spatially correlated with the region affected by the D'Entrecasteaux Zone collision. Lavas in the north and south of the arc, away from the influence of the collision have Pb isotope characteristics of Pacific-MORB mantle, whereas lavas from the central islands have Pb isotope compositions that overlap with IndianMORB mantle (Figure la). Material recovered from the North Aoba Basin on ODP Leg 134 preserves a record of volcanic activity in central Vanuatu both before and after the D'Entrecasteaux Zone collision. Briqueu et al. (1994) showed that prior to the collision, the mantle source of arc volcanism in central Vanuatu was isotopically similar to that for the younger magmatism located away from the collision (Figure lb). The low 206pb/204pb (< ig.O) IndianMORB-like component is thus spatially and temporally linked to the D'Entrecasteaux Zone collision. However, as shown by Briqueu et al. (1994) and Peate et al. (in press), this component cannot originate from the subducting slab given the relatively high 206pb/204pb (18.5-18.8) for the D'Entrecasteaux basement and the subducting sediments of the North Loyalty Basin. Sr-Nd-Pb isotope data on samples from the southern spreading ridge in the North Fiji backarc basin by Peate et al. (in press) confirmed the suggestion by Crawford et al. (1995) that Indian-MORB mantle existed within the back-arc region.

Indian-MORB

local sediment / Pacific MORE central Vanuatu above DEZ collision N & S Vanuatu away from DEZ MereLava

post DEZ collision

pre DEZ collision Icentral Vanuatu: North Aoba Basin ODP Leg134 Sites 832 & 833 Briqueu etal. (1994)

18.0 18.2 18.4 18.6 18.8 18.0 18.2 18.4 18.6 18.8 19.0 206PDU b / ^/204DU ^^Pb FIGURE

Pb isotope compositions in Vanuatu arc lavas are dominated by three components: Tacific-MORB' & 'Indian-MORB' mantle, and a sediment-derived component (Peate et al. in press).

Thus, the major tectonic disruption of the Vanuatu arc caused by the D'Entrecasteaux collision appears to have produced an influx of new mantle material from the back-arc region into the source region of the central Vanuatu arc volcanism, replacing the existing mantle wedge. Hergt & Hawkesworth (1994) documented a similar temporal progression in samples from the back-arc Lau Basin and proposed that Pacific-MORB mantle beneath the Tonga Lau Basin system was systematically replaced by Indian-MORB mantle from the west due to slab rollback. However, the Vanuatu arc appears to be unique amongst island arcs in having both Pacific- and Indian-MORB mantle within the mantle wedge source for the arc volcanism. INFLUENCE OF SUBDUCTED MATERIAL The central Vanuatu lavas have lower 143Nd/144Nd than those to the north and south (Figure 2). In many island arcs, this difference is most readily explained by the addition of a subducted sediment component. However, in the case of Vanuatu, it is difficult to find any standard trace element indicator for sediment involment (e.g. Th/Ce) that varies systematically between the two groups on Figure 2. Hergt & Hawkesworth (1994) showed that the PacificMORB-like and Indian-MORB-like components in the Lau Basin had distinct Sr-Nd isotope compositions. It is possible to reproduce the Sr-Nd isotope composition of lavas in the north and south of the arc away from the D'Entrecasteaux Zone collision by addition of 1-3 % local sediment to a Pacific-MORB-like mantle wedge. Lavas in central Vanuatu can be modelled by addition of similar amounts of sediment but to an Indian-MORB-like mantle wedge. It seems that the amount of sediment required to model the Sr-Nd isotope compositions of the lavas remains relatively constant within the Vanuatu arc (1 to 3 %). Instead, it is the regional difference in 143Nd/144Nd within the mantle wedge, prior to sediment addition, that is the principal factor in controlling the Nd isotope composition of the lavas.

0.51320.51310.5130-

fluid

Pacific-MORB wedge c l Indian-MORB^^^ wedge

Vanuatu 0.5129- r• central above DEZ collision

0.5128-

N & S Vanuatu

away from DEZ

[x MereLava

0.702

0.703

sediment

0.704

0.705

0.706

87Sr/86Sr FIGURE

Lateral variations in Sr-Nd isotope composition along the Vanuatu arc can be explained by regional differences in the isotopic composition of the unmodified mantle wedge followed by addition of similar amounts (1-3 %) of local sediment. Merelava, in the rear-arc, is an exception and appears to be dominated by a fluid component derived from the altered basement of the D'Entrecasteaux Zone (data from Briqueu et al., 1994; Peate et al., in press).

The sediment column being subducted at the Vanuatu trench is distinctive compared to most other arcs in being dominated by volcaniclastic sediments. These sediments have a predominantly subduction-related andesite composition and appear to be detritus derived from the D'Entrecasteaux Zone. They have low trace element contents, and the contrast in Sr-Nd-Pb isotope composition with the active Vanuatu arc is limited. It is only the presence of

Fe-Mn micronodules in the upper horizons that causes the bulk sediment composition to be displaced to lower 143Nd/144Nd and higher 207pb/204pb and 208pb/204pb. The estimated Pb isotope composition of the bulk subducting sediment column lies close to the upper apex of the triangular field defined by the Vanuatu lavas on Figure la. This suggests that a subducted sediment component can be resolved from Pb isotope data for some Vanuatu lavas. It is notable that this component is most strongly expressed in those islands facing the North Loyalty Basin, where the greatest thickness of sediments is being subducted. The Vanuatu arc is characterised by a high K and Sr sediment flux and a very low Ba sediment flux, relative to other arcs (Plank & Langmuir et al., 1993; Peate et al., in press). The averaged composition of Vanuatu magmas, when viewed on the length scale of the whole arc south of the D'Entrecasteaux Zone, reflects the distinctive high K/Ba and Sr/Ba of the sediments being subducted. The central Vanuatu lavas form a linear mixing array on Figure 2, consistent with the variable addition of sediment-derived Pb to an Indian-MORB-like mantle wedge. The shift to higher 206pb/204pb is accompanied by an increase in Cs/Rb, Ba/La, Sr/Th and Pb/Nd to values higher than those found in the local subducting sediments. This might reflect the preferential partitioning of Cs, Ba, Sr and Pb into water-rich fluids released from the slab. Still unresolved though, is the relative extent to which the fractionation of these elements takes place during slab dehydration or during transport into the wedge where chromatographic effects could become important.

1.61

1.2 FIGURE 3

1.4

1.6

1.8

(238U/232Th)

2.2

230xh-238u disequilibrium data for recent (< 10 ka) Vanuatu arc lavas (Peate, unpubl. data, plus two Ambrym analyses from Condomines & Sigmarsson, 1993). The two southern islands (Matthew & Tanna) plot on a 60 ka isochron which is interpreted as the time since the slab-derived fluid, containing U, was added to the mantle source. Data from central Vanuatu are more difficult to interpret, given that on figure 1 Ambrym contains a sediment component (for Nd, and probably Th also) whereas Merelava is dominated by addition of a fluid derived from the altered basement to the D'Entrecasteaux Zone.

MERELAVA - AN ANOMALY IN THE REAR-ARC Merelava lies about 50 km behind the arc front islands of Gaua and Aoba, above the D'Entrecasteaux Zone collision. High values of trace element ratios such as Ba/La, B/Be & U/Th are generally interpreted as implying a large slab-

derived fluid component. In many arcs, moving from the arc front to the rear-arc, there is a decrease in these trace element ratios, but Merelava is anomalous in having the highest values of Ba/La (-70) & B/Be (-50) measured within the Vanuatu arc. It has been suggested that this is unrelated to the present subduction configuration but is due to a pre-existing subduction component inherited from an early part of the arc history when the subduction polarity was reversed. However, U-series data for Merelava (Figure 3) show a large degree of excess-238u {(238u/230Th) > 1.4} which indicates a recent fluid addition (< 60 ka). Merelava has a similar Sr & Pb isotope composition to the adjacent arc front islands, but at higher 143Nd/144Nd. This can be explained by addition of a high Sr/Nd fluid derived from the altered basement of the D'Entrecasteaux Zone, in contrast to the sediment component seen in the arc front lavas (Figure 2). EVIDENCE FROM U-SERIES DISEQUILIBRIUM STUDIES In an island arc environment, the systematics of 238u-230Th disequilibrium in magmas are dominated by the mobility of U relative to Th in fluids derived from the subducted slab. U-series data can provide information about the timing of U-enrichment and the composition of mantle wedge prior to fluid addition. Condomines & Sigmarsson (1993) presented U-series data on two samples from Ambrym in central Vanuatu that were close to 238u-230Th equilibrium, i.e. on the equiline on Figure 3. New mass spectrometric 238u-230Th data have been measured at the Open University on recent (< 10 ka) lavas from several islands within the Vanuatu arc, and these preliminary data are plotted on Figure 3. Many Vanuatu lavas plot to the right of the equiline with significant excess 2 3 8 u {(238u/230xh) > 1}, similar to other trace-element-depleted arcs worldwide (e.g. Tonga, Marianas). The two southern islands, Tanna and Matthew, have similar Sr-Nd-Pb isotope compositions and have high (230Th/232Th). They plot on a 60 ka isochron on Figure 3. This can be interpreted as the time elapsed since a U-rich slab fluid was added to the mantle wedge source that, based on radiogenic isotope data, comprised Pacific-MORB-like mantle modified by minor sediment addition. Ambrym has much lower (230Th/232Th), and this is consistent with its derivation from a sediment-modified Indian-MORB-like mantle wedge source. The high 208pb/204pb of Indian MORB relative to Pacific MORB indicates that it has a higher time-integrated ThAJ ratio, and this should be reflected in a lower (238u/232Th) and hence also a lower (230Th/232Th) for an Indian MORB source in secular equilibrium. Although the central islands (Ambrym and Merelava) plot on a 30 ka isochron, any interpretation is less straightforward given the difference in l^^Nd/l^^Nd. Ambrym contains a sediment component (for Nd, and probably Th also) that is not present in the mantle source for Merelava. Merelava is dominated by addition of a fluid derived from the altered D'Entrecasteaux Zone basement to a mantle source that probably had higher (230Th/232Th) than Ambrym. Thus it is possible that the U-rich fluid was added to the Merelava source more recently than 30 ka.

REFERENCES Briqueu, L., Laporte, C., Crawford, A.J., Hasenaka, T., Baker, P.E. & Coltorti, M., 1994. Temporal magmatic evolution of the Aoba Basin, central New Hebrides island arc: Pb, Sr, and Nd isotopic evidence for the coexistence of two mantle components beneath the arc. In: Greene, H.G., Collot, J-Y, Stokking, L.B. et al., Proceedings of the Ocean Drilling Program, Scientific Results 134, 393-401. Condomines, M., & Sigmarsson O., 1993. Why are so many magmas close to 238u-230u radioactive equilibrium? Geochimica Cosmochimica Acta, 57, 4491-4497. Crawford, A.J., Briqueu, L., Laporte, C. & Hasenaka, T., 1995. Coexistence of Indian and Pacific oceanic upper mantle reservoirs beneath the central New Hebrides island arc. In: Taylor, B. & Natland, J. (eds) Active Margins and Marginal Basins of the Western Pacific. AGU Geophysical Monograph 88, 199-217. Hergt, J.M., & Hawkesworth C.J., 1994. Pb-, Sr-, and Nd-isotopic evolution of the Lau Basin: implications for mantle dynamics during back-arc opening. In: Hawkins, J., Parsons, L, Allan, J. et al.. Proceedings of the Ocean Drilling Program, Scientific Results 135, 505-517. Peate, D.W., Pearce, J.A., Hawkesworth, C.J., Colley, H., Edwards, C.M.E., & Hirose, K., 1996. Geochemical variations in Vanuatu arc lavas: the role of subducted material and a variable mantle wedge composition. Journal of Petrology, (in press). Plank, T., & Langmuir, C.H., 1993. Tracing trace elements from sediment input to volcanic output at subduction zones. Nature, 363, 739-743.

DO SUBDUCTED SEDIMENTS MELT BENEATH ARC VOLCANOES? SOME EXPERIMENTAL ANSWERS Terry Plank- and Marie Johnson^ 1 Department of Geology, 120 Lindley Hall, University of Kansas, Lawrence, KS 66045 2 Department of Geography and Environmental Engineering, United States Military Academy, West Point, NY 10996

The return of continental material to the mantle at subduction zones is one of the remarkable outcomes of plate tectonics. Even more remarkable is the reemergence of subducted sediment in arc volcanoes. The geochemical tracers ^^Be and ^^^Pb bear witness to this sediment cycle. Although geochemical models have long included contributions from subducted sediment, oceanic crust and water to arc volcanism, the mechanisms of this recycling are still poorly known. What was once the simple breakdown of amphibole at - 100 km in the slab is now a plethora of possible hydrous phases (e.g., lawsonite, epidote, phengite, serpentine) which may supply water and trigger melting beneath arcs (Schmidt, 1996; Pawley and Holloway, 1993; Ulmer, 1995). \ ^ a t was once simple bulk sediment mixing is now a complex sediment fluid, generated by dehydration or melting or both (Morris, et al., 1990; Plank and Langmuir, 1992; Reagan et al., 1994; Elliott et al., 1996). The looming questions of element recycling at subduction zones involve not whether but how. What phases are involved? What reactions occur? We focus here on the question of whether subducted sediments melt in the slab beneath arc volcanoes. This question is important not only to the mechanisms of element recycling, but also to the thermal regime of subducting slabs. Forward models have become increasingly sophisticated (e.g.. Peacock, 1993), but uncertainties still exist over the values of various parameters (mantle viscosity, shear heating) and the dynamics of convection. Determining the extent to which sediments melt (or not) in subducting slabs can provide a tie point (the sediment solidus) to slab temperature, and so help refme thermal models of slab subduction. Inverse geochemical modeling of arc lavas has proceeded to where the sediment phase that recycles to the arc can be fairly uniquely identified. It has high ^^Be, and Be recycling from sediment to arc must be > 40% efficient (Zheng et al., 1994). Another key attribute is the high Th content of the sediment phase recycled to the arc. Based on a mass balance of sediment input and arc output fluxes, Plank and Langmuir (1992) concluded that subducted sediment loses 20-50% of its Th to the arc. Other studies have also found evidence in arc volcanics for a large sediment contribution to Th, based on Th-isotopes (Reagan, et al., 1994; Elliott et al., 1996) and Ta, based on Ta/Zr ratios (Turner et al, 1996). Correlations between Nd isotopes in the arc and the local sediments (White et al, 1985) also require efficient Nd recycling. All of these geochemical conclusions are noteworthy because neither Be, Th, Ta nor Nd are thought to be particularly mobile in aqueous fluids (Staudigel, et al., 1996; Tatsumi, 1986). This has led to the suggestion that sediments must melt in order to give up a significant portion of their Be, Th and Ta to the arc (Plank and Langmuir, 1992; Turner et al., 1996). Experimental confirmation of these geochemical effects, however, has yet to occur for appropriate sedimentary compositions and phases. Our approach to the problem is to provide some laboratory measurements of the trace element composition of sediment melts and dehydration fluids. In particular, can aqueous fluids in equilibrium with sediments at high pressure and temperature partition enough Be, Th and Nd to explain the enrichments observed in arc basalts? Or is a sediment melt required? While there has been a recent explosion of new partitioning studies of aqueous fluids in equilibrium with various mafic minerals (e.g., Brenan et al., 1995; Keppler, 1996), a similar approach to sediment systems is complicated by the large number minerals in equilibrium at high pressure: clinopyroxene, garnet, amphibole, muscovite, biotite, magnetite, coesite, kyanite, apatite (based on 30 kb phase equilibria in Johnson and Plank, 1993, and Nichols et al., 1994). The partitioning between most of these minerals and an aqueous fluid or silicate melt is-virtually unexplored. Thus, our goal is to combine experimental phase equilibria studies with geochemical studies of experimentally produced sediment melt and fluid compositions. We report here preliminary results obtained at 20 kb and 600-800°C with a diamond trap technique. In our experiments, a red clay (containing -15% combined H20"'" and H2O') is loaded into a Au capsule, covered with a layer of HNOs-rinsed diamonds (40-50 microns), and welded shut. As in Hirose and Kushiro (1993) and Baker and Stolper (1994), we use the diamond powder to trap fluids or melts generated during the experiment. Because of the high yield strength of diamonds, void spaces remain within the diamond layer even at high pressure. Ruids or melts are drawn into the void spaces initially due to their lower pressure; as the voids are filled the pressure of the entire charge reaches the load pressure, and the fluid and solid exchange to reach equilibrium. Experiments are conducted in a standard Boyd-England piston cylinder apparatus and held at pressure and temperature for generally >72 hours. The experiment is then quenched by shutting off the furnace power, the capsule is peeled open, and the sediment residue is carefully separated from the diamond powder. The two layers from the capsule are then dissolved separately in Savillex beakers with an HNO3-HF mixture. The 76

residue within the diamonds dissolves, while the diamonds themselves are impervious to the acid. The separate solutions for the sediment residue and diamond leach are then analyzed by ICP-MS, along with the original sediment loaded. Many advantages exist to recovering and analyzing both the fluid and solid residue from a high pressure experiment. One is that bulk partition coefficients (Csolid/Cfiuid) can be measured fairly directly (once the fluid fraction present in the diamonds is determined), without considering the individual contributions of the 7 or so stable mineral phases. Another, however, is that mass balance can be used to demonstrate that material was not lost from the charge, and that complete dissolution of phases was achieved for the ICP-MS analyses. This mass balance is illustrated in Figure 1, which shows results at 600 and 800 ""C and 20 kb. Both the sediment residue and diamond traps are normalized to the starting sediment composition, so it is immediately apparent that the fluid traps gain what the sediment loses. Thus, the two sides of the reaction plot as near mirror images of one another. Clear differences exist between the 600°C and 800°C experiments. At 600°C, only the alkalis, alkaline earths and Pb are transported to the diamond trap. This experiment should be below the sediment solidus, and thus the results should reflect aqueous fluid partitioning. The stronger partitioning into the fluid of the alkaline earths (Ba, Sr) relative to the alkalis (Rb, Cs) may result from the presence of micas in the residue. Four different experiments were carried out at 600°C for > 50 hours, and all show these same geochemical systematics. Thus we have some confidence that equilibrium was approached. With increasing temperature, the sediment loses more LREE, Th, and Nb. This relationship is illustrated in Figure 2, where Th in the diamond trap is shown to increase dramatically above 700°C. Based on our own experiments and others (Nichols et al., 1994), the sediment solidus should be between 600 and 700°C at 20 kb. We thus interpret the dramatic change in the behavior of Th to coincide with sediment melting, where the melts partition Th much more strongly than subsolidus fluids. Although these experimental results are preliminary, they demonstrate a viable technique for documenting geochemical changes to trace element partitioning as sediments cross their solidus. The 20 kb transect shows that mobilizing significant Th, Nd, and Ta from the sediments in the slab may require temperatures in excess of the sediment solidus, between 600 and 700°C. These temperatures are still greater than those predicted by current thermal models of normal slab subduction. Further experiments near the sediment solidus, combined with refinements to the thermal models should lead to an improved understanding of subduction reactions and their consequences. REFERENCES Baker, M. B. & E. M. Stolper, 1994. Determining the composition of high-pressure mantle melts using diamond aggregates. Geochem. Cosmochem, Acta 58, 2811-2827. Brenan, J.M., H.F. Shaw, FJ. Ryerson, & D.L. Phinney, 1995. Mineral-aqueous fluid partitioning of trace elements at 900°C and 2.0GPa: constraints on the trace element chemistry of mantle and deep crustal fluids, Geochim. Cosmochim, Acta 59, 3331-3350. Elliott, T., Plank, T., Zindler, A., White, W. & Bourdon, B., 1996. Element transport from subducted slab to volcanic front at the Mariana arc. Journal of Geophysical Research, in review. Hirose, K. & 1. Kushiro, 1993. Partial melting of dry peridotites at high pressures: Determination of compositions of melts segregated from peridotite using aggregates of diamonds. Earth Planet. ScL Lett. 114, 477-489. Johnson, M.C., & T. Plank, 1993. Experimental constraints on sediment melting during subduction, EOS 74,

680.

Keppler, H., 1996. Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380, 237-240. Morris, J. D., Leeman, W. P., & Tera, F., 1990. The subducted component in island arc lavas: constraints form Be isotopes and B-Be systematics. Nature 344, 31-36. Nichols, G.T., P.J. Wyllie, & C.R. Stem, 1994. Subduction zone melting of pelagic sediments constrained by melting experiments. Nature 371, 785-788. Pawley, A.R. & Holloway, J.R., 1993. Water sources for subduction zone volcanism: New experimental constraints. Science 260, 664-667. Peacock, S.M., 1993. Large-scale hydration of the lithosphere above subducting slabs. Chemical Geology 108, 49-59. Plank, T. & C.H. Langmuir, 1992. Sediments melt and basaltic crust dehydrates at subduction zones, EOS, 73, 637. Reagan, M., Morris, J. D., Herrstrom, E. A., & Murrell, M. T., 1994. U-Series and Be isotope evidence for an extended hsitory of subduction modification of the mantle below Nicaragua. Geochim. Cosmochim. Acta 58, 4199-4212. 77

Schmidt, M.W., 1996. Experimental constraints on recycling of potassium from subducted oceanic crust. Science 272, 1927-1930. Staudigel, H., Plank, T., White, W.M. & Schmincke, H., 1996. Geochemical fluxes during seafloor alteration of the upper oceanic crust: DSDP Sites 417 and 418, Bebout and Kirby, eds., SUBCON: Subduction From Top to Bottom, AGU Geophysical Monograph, in press. Tatsumi, Y., Hamilton, D.L. & Nesbitt, R.W., 1986. Chemical characteristics of fluid phase released from a subducted lithosphere and origin of arc magmas: evidence from high-pressure experiments and natural rocks. J. Vole. Geotherm. Res,, 29: 293-309. Turner, S.T., Hawkesworth, C., van Calsteran, P., Heath, E., Macdonald, R. & Black, S., 1996. U-series isotopes and destructive margin magma genesis in the Lesser Antilles. Earth & Planetary Science Letters 142, 191-207. Ulmer, P & Trommsdorff, 1995. Serpentine stability to mantle depths and subduction-related magmatism. Science 268, 858-861. White, W.M., Dupre, B. & Vidal, P., 1985. Isotope and trace element geochemistry of sediments from the Barbados Ridge - Demerara Plain region, Atlantic Ocean. Geochim. Cosmochim. Acta. 49, 1875-1886. Zheng, S.-H., Morris, J., Tera, F., Klein, J., & Middleton, R., 1994. Beryllium isotopic investigation of sedimentary columns outboard of subduction zones. ICOG abstracts, 8. Figure 1. Trace element concentrations (normalized to starting red clay) of diamond traps and sediment residues. Diamond traps contain fluid or melt phase. The near mirror image of the residue and diamond trap demonstrate mass balance within the experimental charge. Deficiencies in the HREE and U may reflect the inability to dissolve zircons within the sediment residue. Note dramatic differences in the LREE, Th and the HFSE between the 600°C and 800°C experiments. These differences are most likely due to the fact that the sediment solidus was crossed between the two experiments.

jdc/3^ a ^ a

J- P — x s ^ O r ^ i P D>HX)fC!

0.25

Figure 2. Concentration of Th in the diamond trap (normalized to the original sediment) as a function of temperature. The dramatic increase in Th lost from the sediment probably coincides with th onset of sediment melting. These experiments show that Th is more strongly partitioned and transported in a silicate melt than a aqueous fluid.

IC!I^ fl ifii -fi .2

H -5

700

Temp (^C)

THE GEOCHEMICAL EVOLUTION OF ARC MAGMAS IN A CONTINENTAL SETTING: EVIDENCE FROM DETAILED CHEMO-STRATIGRAPHY AT RUAPEHU, NEW ZEALAND. Richard C.Price^ Tod E. Waight^ John R.Chapman^ Eloise E. Beyer^ Ian E-MSmith^ and Robert B. Stewart^ ^ School of Earth Sciences, La Trobe University, Bundoora, Victoria, Australia, 3083. ^ Department of Geology, University of Auckland, Auckland, New Zealand. ^ Department of Soil Sciences, Massey University, Palmerston North, New Zealand.

INTRODUCTION The Mt Ruapehu is located at the southern end of the Taupe Volcanic Zone (TVZ); the principal focus of subduction related magmatism in New Zealand's North Island. The TVZ is the southern extension of magmatic activity associated with westward subduction of the Pacific plate beneath the Indo-Australian plate. To the north, along the Tonga-Kermadec island arc, oceanic crust is being subducted beneath oceanic crust and eruptives are dominantly basalt and basaltic andesite (Gamble et al., 1990). In the TVZ, where oceanic crust is being subducted beneath continental crust, volcanism is dominated by rhyolitic magmas and andesites and basalts are relatively uncommon (Graham et al., 1995). A well defined, westward dipping, Wadati-Benioff seismic zone lies at a depth of about 100 km beneath the TVZ and the zone is a region of thin (- 15 km) crust and exceptionally high heat flow (Stem and Davies, 1987; Hochstein, 1995). Ruapehu is the largest, active, andesitic volcano and the highest mountain (2797 m) in the North Island. The most recent eruption began in September 1995 with major magmatic and phreatomagmatic events occurring in September/November 1995 and June 1996. The magmatic history of the volcano is preserved in ring plain and tephra deposits (Donaghue et al., 1995) and lava flow sequences which Hackett (1985) subdivided into Te Herenga, Wahianoa, Mangawhero, and Whakapapa Formations. Recent detailed mapping and geochemical analysis of flows have illustrated that Hackett's formations contain quite complex stratigraphic sequences recording a complicated interplay of processes including crystal fractionation and cmstal assimilation as well as magma mixing. Geochemical composition strongly influences the morphology of lava flows, with lower silica eruptives forming relatively thinflowswith profiles indicating highfluidity,and higher silica flows having much thicker profiles. GEOCHEMICAL VARIATION IN FLOW SEQUENCES Information and interpretations presented here are based on new major and trace element and isotopic data for lava flow sequences exposed in the Whangaehu gorge and Mangatoetoenui valleys on the eastern side of the volcano. The sequences belong within Hackett's Wahianoa and Mangawhero Formations. If data for the Whangaehu gorge section are considered collectively, overall trends are similar to those observed in previously pubhshed data compilations for the whole volcano (Graham and Hackett, 1987). K2O abundances are broadly correlated with SiOi contents (Figure 1) and incompatible trace elements such as Ba, Rb, La, and Zr show similar correlations. Flows exposed in the section show considerable variability in isotopic composition (^'^Sr/^^Sr ratios range from 0.70496 to 0.70565) and ^'^Sr/^^Sr isotopic ratios show a rough positive correlation with Si02 abundance (Figure 1). u 07056

-T

O^ O

0.7054 00

0.7052 0.7050 0.80

55.0 55.5 56.0 56.5 57.0 57.5 58.0 58.5

On _J

o 8 o o I

55.0 55.5 56.0 56.5 57.0 57.5 58.0 58.5

SiO^ (wt%)

Figure 1: K2O versus SiOi and ^^Sr/^^Sr versus Si02 for all samples from the Whangaehu lava sequence, eastem Ruapehu.

On a plot of l^^Nd/^'^^Nd versus ^'^Sr/^^Sr the Whangaehu gorge section data show a pattern of behaviour similar to that observed for the whole volcano, with the two isotopic ratios being negatively correlated (Figure

2).

0.51295 0.51290 0.51285 0.51280

+ +

o 0+

0.51275 0.51270

4O o + ++ o^ + ^ +

-L. -L. -L 0.51265 0.7048 0.7050 0.7052 0.7054 0.7056 0.7058

Figure 2: ^^^Nd/^^^Nd versus ^^ST/^^ST diagram for all available Ruapehu data. Circles are samples from the Whangaehu gorge sectioa Graham and Hackett interpreted these patterns to reflect assimilation/ fractional crystallisation (AFC) processes operating in crustal magma chambers; the crustal contaminant was argued to be compositionally similar to basement greywacke (see also Graham et al., 1995). The conclusions were based on a consideration of a pooled data base for samples from the whole volcano. When data for the Whangaehu gorge section are considered collectively, the same patterns are observed as are observed for the whole volcano, but when the data for the Whangaehu sequence are examined in detail a much more complex pattern emerges. Sub-sequences, each containing three or four conformable flows, show a consistent cyclic pattern from relatively uruadiogenic Sr isotopic compositions to relatively higher ^'^Sr/^^Sr isotopic ratios (Figure 3). In the upper part of the sequence, the topmost two cycles show a concomitant increase in SiOi abundance; a pattern that would be consistent with an AFC process. U

0.7057 00 0.7056 U (/S 0.7055 00 0.7054 0.7053 0.7052 0.7051 0.7050 0.7049 cntN-HOOoor^vovor^oocor-vovnONO—< cs

58.5

mcN—'OOsoor^vnvor-oooor^vovnovo^cN CSCStNCN'-H'^^tNCNtNCN^^^^^^Q^SQJ 0âÔ^C^C>C>C>avaÔ^C>OÔÔÔÔÔâ^C^

Figure 3: ^^Sr/^^Sr and SiOi abundance versus stratigraphic position in Whangaehu gorge lava sequence. Individual conformableflowpackages representing sequentially empted lavas are connected by arrows. In the middle and lower part of the section, the patterns of the upper part are not repeated. Two lava packages show decreasing Si02 content through the cycle they preserve, but increasing ^"^Sr/^^Sr isotopic ratios and the lowermost cycle preserved shows relatively constant Si02 abundance but an increase in ^'^Sr/^^Sr. These patterns illustrate the complexity of andesite geochemistry and the dangers involved in constructing very precise quantitative models based on data for a relatively small number of samples from a large strato-volcano. They are not easily reconciled with a model involving a single magma chamber. 80

INTERPRETATION Flow sequences elsewhere on the volcano that can be interpreted to have erupted during a particular time interval, are characterised by variability similar to that observed in the Whangaehu gorge section. For example, a group of flows outcropping in a tributary of the Mangatoetoenui on the north eastern side of the volcano can be subdivided on field evidence and major element chemistry into three packages, each of two to three flows. Compositions range from low to high Si andesites. Geochemical variation within packages contrasts with the variation throughout the whole sequence. Each package of lava flows within related sequences such as the Whangaehu gorge section is interpreted to represent a batch of magma which evolved in and was erupted from a high level magma chamber. Progressive evolution observed within the whole sequence, or abrupt chiiges in overall chemistry between packages within a sequence, are interpreted as reflecting recharge from an evolving, deeper level, lower crustal reservoir. The data for the Whangaehu gorge lavas are consistent with a model whereby magmas evolve and erupt from a high level magma chamber, probably located immediately below the volcano. Eruption of discrete flow packages is possibly associated with period recharge of this high level chamber with fresh magma and the variation within the packages probably reflects both crystal fractionation and mixing between fresh, rechargmg magma and evolved magma remaining in the reservoir from previous recharge/fractionation events. The relative importance of each of these processes varies between packages. Fresh recharging magma batches vary considerably in composition, are all geochemically evolved, and show evidence for interaction with crust. It is suggested that recharging magma batches evolve in a deep magma reservoir, possibly at the crust/mantle boundary, where crustal assimilation is also taking place. Periodically batches of magma are fed upwards to recharge and evolve further in the high level magma reservoir. REFERENCES Donoghue, S.L., Neall, V.E., and Pahner, A.S., 1995. Stratigraphy and chronology of late Quaternary andesitic tephra deposits, Tongariro Volcanic Centre, New Zealand. Journal of the Royal Society of New Zealand 25, 115-206. Gamble, J.A., Smith, LE.M., Graham, IJ., Kokelaar, B.P., Cole, J.W., Houghton, B.F., and Wilson CJ.N., 1990. The petrology, phase relations and tectonic setting of basalts from the Taupo Volcanic Zone, New Zealand and the Kermadec island arc - Havre Trough, SW Pacific. Journal ofVolcanology andgeothermal Research 43, 235-270. Graham, I.J., Cole, J.W., Briggs, R.M., Gamble, J.A., and Smith, I.E.M., 1995. Petrology and petrogenesis of volcanic rocks from the Taupo Volcanic Zone: a review. Journal ofVolcanology and geothermal Research 68, 59-87. Graham, I.J. and Hackett, W.R., 1987. Petrology of calc-alkaline lavas from Ruapehu Volcano and related vents, Taupo Volcanic Zone, New Zealand. Journal of Petrology 28, 531-567. Hackett, W.R., 1985. Geology and petrology of Ruapehu Volcano and related vents. Unpublished PhD Thesis, Victoria University of Wellington, New Zealand. Hochstein, M.P., 1995. Crustal heat transfer in the Taupo Volcanic Zone (New Zealand): comparison with other volcanic arcs and explanatory heat source models. Journal ofVolcanology and geothermal Research 68, 117-151. Stem, T.A. and Davey, FJ., 1987. A seismic investigation of crustal and upper mantle structure within the Central Volcanic Region of New Zealand. New Zealand Journal of Geology and Geophysics 30, 217-231. Acknowledgements: This work was supported by a grant to R.C. Price from the Australian Research Council. The technical assistance of R. Maas, J. Metz, and I. McCabe is gratefully acknowledged.

PHYSICAL AND CHEMICAL CHARACTERISTICS OF THE LINCOLN BATHOLITH: ASPECTS OF PALAEOPROTEROZOIC MAGMATISM AND POTENTIAL ANALOGUES Bruce F Schaefer^ J.D Hoek^ and John Foden^ 1 Department of Geology and Geophysics, The University ofAdelaide, SA, 5005 2 School of Earth Sciences, University of Melbourne, Parkville, Vic, 3052

INTRODUCTION The Palaeoproterozoic Lincoln Batholith (Hoek and Schaefer, in press) of southern Eyre Peninsula on the Gawler Craton, South Australia, encompasses a voluminous suite of dominantly felsic magmas emplaced in the mid to upper crust during the period --1852-1840 Ma. The Lincoln Batholith preserves many physical features reminiscent of modem arc and plate margin magmatic settings, however chemically many characteristics of the intrusives have affinities with intraplate magmatism typical of the Australian Proterozoic (see Wybom et ah, 1992 and references therein). This contribution aims to investigate and detail the features of the Lincoln Batholith, with a view to constraining and commenting on processes involved, and the role and evolution cf magmatic source regions during this period of the Proterozoic. Ultimately an understanding of such processes is essential for comparative studies with contemporary plate margin magmatic settings and crustal growth mechanisms. THE LINCOLN BATHOLITH The term Lincoln Batholith has been recently coined (Hoek and Schaefer, in press) in an attempt to minimise the current level of confusion surrounding the nomenclature of Palaeoproterozoic magmatism on the Gawler Craton. The term is used to distinguish those aspects of magmatism associated with emplacement of the Donington Suite granitoids from other granitoids of the Palaeoproterozoic, previously grouped together as the Lincoln Complex. As such, the Lincoln Batholith mcludes the Donington and Colbert Suite granitoids and the synplutonic Jussieu Dykes, all of which are part of a single evolving magmatic system over the period 1852-1840 Ma. The type section of the Lincoln Batholith is confined to the eastem side of the Kalinjala Shear Zone on south eastem Eyre Peninsula, South Australia (Figure 1), with possible correlatives in the north eastem portion of Eyre Peninsula. The Lincoln Batholith is broadly bimodal, with by fer the majority of magmatism (>95%) being felsic. The mafic magmatism is confined to a suite of synplutonic mafic dykes, characterised by felsic selvedges generated by remelting and back-veining of the dykes by felsic host lithologies. Such host lithologies have been subdivided into the Donington and Colbert granitoid Suites by Mortimer (1984), Figure 1: Distribution of the Lincoln Batholith and associated regional ^ith the Colbert Suite restricted in distribution tO a cmstal rock tj^es on Eyre Peninsula South Australia, ^ e Kaliiy ala ^^^^ ^^^^^ ^ Donington. The Donington

Shear Zone is a crustal scale feature separating older granulite supracrustals of the Sleaford Complex and amphibolite facies Hutchison Group Metasediments from the Lincoln Batholith and younger volcano-sedimentary sequences in the east. Moody Suite Granites are syn-Kimban Orogenic (-1750-1700 Ma) intrusives in the Hutchison Group.

Suite COmpnses by far the bulk of the magmatism (f the Lincoln Batholith, and fiirther subdivisions within the Donington Suite are detailed below.

FELSIC MAGMATISM OF THE LINCOLN BATHOLITH The Donington Suite consists of quartz gabbronorite, hypersthene granite (Memory Cove Chamockite), alkali feldspar granites (rapakivi and even-grained varieties), and volumetrically minor highly evolved and fractionated

late stage melts (Carcase Rock Granite). Figure 2 summarises the trace element characteristics of three major units of the Donington Suite, all cf which preserve primary magmatic pyroxenes. The bulk of the Lincoln Batholith has been subject to extensive deformation and retrogression during the Kimban Orogeny (--1750-1700 ^ 100 Ma), making interpretation of whole rock geochemical data difficult. However, Figure 2 shows several a.Ba chemical features typical cf ® Proterozoic within plate aE magmatism, notably Rb, U, Th and LREE enriched values, significant Sr E depletion and elevated Y values a relative to modem arcs (Wybom et al, 1992 and refs. therein). Interaction with significant volumes Figure 2: Trace element variation diagram of volumetrically significant pyroxene bearing of crustal material is suggested by felsic phases in the Lincoln Batholith. Ranges of values are normalised to primitive distinct Nb, P, and Ti anomalies. mantle values of McDonough et al (1991). A general trend of increasing weight % SiOi Petrographic studies coupled with values is observedfromthe quartz gabbronorite through the Memory Cove Chamockite major and trace element modelling to the Megacrystic Chamockite. carried out by Mortimer (1984), suggest the quartz gabbronorite is a cumulus phase, with remaining Donington Suite granitoids being derived as products of continued fractional crystallisation. The TAS vs. Si02 diagram illustrated in Figure 3 serves to highlight both the subalkalic nature and the range and distribution of Si02 values for the Lincoln Batholith. Of significance is the presence of a 'silica gap' between approximately 58-62 weight % Si02, and the prevalence of magmas with > 68 weight % Si02. A subtle bimodality in Si02 distribution may be observed in this 12.0 range, with peaks occurring at -68ALKALINE 72 weight % and <--75-78 weight %. These features are accentuated in the comagmatic Colbert Suite (open squares in Figure 3), which also shows a number of other similar major element features to the Donington Suite. Trace element SUB ALKALINE features for both suites are similar to 0.0 those displayed in Figure 2, 45.0 65.0 55.0 75.0 85.0 distinguishing the felsic portions cf the Lincoln Batholith from S i 0 2 (wt%) contemporary island arc style Figure 3: Total Alkali vs. Silica (TAS) plot for the Lincoln Batholith. Filled circles = magmatism, particularly in terms cf Donington Suite, open squares = Colbert Suite, and filled triangle = Jussieu Dyke. pronounced Sr depletion and Boundary between alkaline and subalkaline series from Irvine and Baragar (1971). relatively elevated Y values. However, other features of the Donington Suite which are reminiscent of modem arcs include the presence of a Nb anomaly, general similarities in pattern (if not value) for intermediate rocks of the Donington Suite with a typical Andesite (Figure 2) and Ti depletion. MAFIC MAGMATISM OF THE LINCOLN BATHOLITH A volumetrically minor component of the Lincoln Batholith are the syn-plutonic mafic Jussieu Dykes; however these bodies potentially record the most important chemical data and field relationships for understanding the thermal nature of the lithosphere into which the Lincoln Batholith was emplaced. These dykes are characterised by the presence of felsic selvedges, and are clearly crosscut by dykes of the post-Lincoln Batholith, mafic Toumefbrt Dykes (see Figure 4). Besides felsic selvedges, the Jussieu Dykes show a range of features which indicate them to be synchronous with their host felsic plutons. For example, in regions of low strain, the preservation of the transition of emplacement styles of mafic magma from a typical planar dyke-like morphology to a series of disaggregated mafic enclaves 1000

entrained within felsic selvedge material (e.g. as illustrated in Figure 4). Such observations, coupled with further temporal constraints imposed by crosscutting relationships with both pre-Donington Suite basement rafts and + i-LJ. subsequent mafic dyke emplacement events Toumefort Dyke (the pre- Kimban Toumefort Dykes), are strongly indicative of a syn- Lincoln Batholith Selvedge origin for the Jussieu Dykes. A similar genesis Jussieu Dyke is implied for many of the mafic enclaves observed within the remainder of the Lincoln Donington Suite Batholith. Basement Syn-plutonic mafic dykes associated with volimiinous felsic intrusive magmatism is well documented fi-om a large number cf Palaeozoic plate margin, dominantly subduction related-terrains, some of the better documented including the Sierra Nevada and Klamath Mountains in the US, the Westem Cordillera in Peru, and the southern Malay Peninsula (see Pitcher, 1991 for further references) suggesting an important relationship between process, chemistry and tectonic setting. FOR TECTONIC Figure 4: Sketch of outcrop on northeastern point of Williams Island IMPLICATIONS displaying key physical characteristics of the Jussieu Dykes. The Jussieu SETTING Dyke is surrounded by a characteristic selvedge of felsic material. This particular dyke changes along strike length from a coherent planar body to a series of disaggregated mafic enclaves supported within the felsic selvedge. The dyke is interpreted to have been propagating through Donington Suite granitoids that were close to their solidus, with the associated thermal input causing secondary partial melting and hence forming the selvedge. Also of interest at this locality is the presence of rafts of pre- Donington basement material, and an unambiguous crosscut by a post- Lincoln Batholith mafic dyke belonging to the Toumefort Dykes.

In general, the development of granitoids is typically temporally and spatially associated with intrusive mafic magmas, however the differing relative proportions of magmas involved bears some relationship with specific tectonic scenarios. For example, syn-plutonic mafic dykes are most common in mantle derived batholiths of marginal and oceanic volcanic arcs, and are comparatively scarce in syn- to post- orogenic magmatic settings where granite source regions commonly have a larger crustal component. Intrusive end members of voluminous intra-plate A-type felsic magmatism (such as the Gawler Range-Hiltaba Suite volcano plutonic province) record no unambiguous syn-plutonic mafic dyke features, despite the proven existence cf large volumes of contemporaneous mafic and felsic magmas; possibly reflecting both the excessively thermally perturbed geotherm and greater volume of magmatic material in such settings. Whereas the felsic magmatism of the Lincoln Batholith shows chemical characteristics typical of many other within-plate Australian Proterozoic granites, and has been interpreted in the past to be at least in part associated with intra-plate processes, some chemical and particularly physical characteristics, such as the coexistence cf mafic and felsic magmas and the presence of syn-plutonic mafic dykes, are more typical of plate margin batholiths. Resolving the apparent ambiguity of such features occurring in a terrain preserving no contemporaneous subduction related features is crucial in understanding crustal evolution processes for the Palaeoproterozoic. Hence, until such tune as the processes controlling both the physical aspects and chemical evolution of magmatic systems in both plate margin and intra-plate settings are better understood , the use cf individual comparative criteria for modem plate tectonic analogues of ancient magmatic systems must be undertaken with caution. REFERENCES Hoek, J.D. and Schaefer, B.F. in press. The Palaeoproterozoic Kimban mobile belt, Eyre Peninsula: Timing and significance of felsic and mafic magmatism and deformation. Mortimer, G.E. 1984. Early to Middle Proterozoic granitoids, basahic dykes and associated layered rocks cf south eastern Eyre Peninsula, South Australia. Univ. Adel. PhD. Thesis. Unpubl Pitcher, W.S. 1991. Syn plutonic dykes and mafic enclaves. In: Didier, J. and Barbarin, B. Eds. Enclaves and granite petrology. Developments in Petrology 13, 383-391. Wybom, L.A.I., Wybom, D., Warren, R.G. and Drummond, B.J. 1992. Proterozoic granite types in Australia: implications for lower crustal composition, structure and evolution. Trans. Royal Soc. Edin. 83, 201-210.

THE ROLE OF AQUEOUS FLUIDS IN TRACE ELEMENT FRACTIONATION IN SUBDUCTION ZONE PROCESSES: EVIDENCE FROM EXPERIMENTS AT 3 TO 5.5 CPA AND lOOO' C R.Stalder\ G.A.JeIlne^^ S.F.Foley', LHom^ G.P.Brey' 1 Institut fuer Mineralogie, Universitat Frankfurt, Senckenberganlage 28, 60054 Frankfurt/Main, Germany 2 Department of Earth Sciences, Memorial University of Newfoundland, St. John's, NF AIB 3X5, Canada 3 Mineralogisch-Petrologisches Institut, Universitat Gottingen, Goldschmidtstrasse 1,37077 Gottingen, Germany

The most consistently observed geochemical difference between arc and non-arc volcanics is the depletion in high field strength elements (HFSE), especially Nb and Ta, relative to low field strength elements (LFSE) and light rare earth elements (LREE). The behaviour of all of these element groups, but particularly the HFSE, during island arc magma genesis processes is controversial. Possible mechanisms which have been suggested to decouple different groups of trace elements during subduction related enrichment processes or magma genesis fall into two major groups. In the first model, preferential enrichment in a hy^ous fluid of the LREE and LFSE relative to the HFSE exerts the most important control over trace element fractionation (e.g,, McCulloch and Gamble, 1991; Hawkesworth et al., 1994), although opinions differ as to the necessity of a residual titanite mineral in equilibrium with the fluid. In the second, partitioning of the HFSE into a titanite phase in a felsic melt residue during melting processes at the slab interface and/or in the mantle wedge effectively decouples the element groups (e.g., Foley and Wheller, 1990). To evaluate these models it is necessary to know tho partition coefficients, at appropriate conditions (i.e., pressures/depth of at least 3 GPa/110 km), between (1) silicate mande minerals and melt, (2) titanate phases and melt, (3) fluid and silicate mande minerals, and (4) fluid and titanate phases. In recent years much effort has been spent attempting to quantify trace element behaviour in tiie presence of aqueous fluids (e.g., Ayers et al. 1992; Brenan et al. 1994, 1995; Keppler, 1996). However, up to now tiiese studies have been limited to relatively low pressures (up to 2 GPa). To investigate trace element behaviour in aqueous fluids at higher pressures, we have utilized a belt apparatus, allowing us to expand the pressure range to 3.0 to 5.5 GPa 100 - 170 km). Coupled with this we have used an experimental charge assembly containing a diamond extraction trap, and laser ablation - inductively coupled plasma mass spectrometry (LAM-ICP-MS) to determine fluid/solid partition coefficients by direct analysis of residues and the fluid solute. In this study we present partition coefficients for fluid/pyrope-rich garnet, fluid/diopside and fluid/diopside+rutile, and for tiie elements - Ti, Zr, Hf, Nb, Ta, La, Ce, Sm, Tb, Yb, Sr, Ba andPb. Solid starting materials in the experiments consisted of doped, synthetic CMAS-glasses to ensure homogeneous distribution and shorter equilibration time. Glasses were ol-normative to keep the silica activity low. In tiie experiments involving rutUe, a finely ground natural ratile was added. As sample containers we used Au-capsules with an outer diameter of 3.0 mm and an inner diameter of 2.6 mm. In each capsule 50 % of the solids consisted of diamond crystals (grain size -50 jiim), and 50% the starting materials. The fluid was added directiy to the experimental charge as water or hydrochloric acid solution. The pore space between the diamonds is preserved during tiie experiments and the fluid is able to circulate throughout the capsule. On quenching tiie experiment, the fluid solute becomes trapped in the pore spaces between the diamonds. Data is reported for experiments run for 48h; after less than 14h glasses were completely crystallized. One reversal experiment showed that equilibrium is reached witiiin 2 days. The fluid trapped in the diamonds and crystalline residues were analysed with LAM-ICP-MS, using a frequency quadrupled (266 nm) Nd-YAG laser. Traps were analyzed with 0.6 mJ and defocused beam, so that pits are 50-200 |im in ^ameter; residues were analyzed with 0.4 mJ, pits usually are smaller than 30 |im. We chose Ni as the internal standard element in the trap since it is an element easily detectable in the synthetic diamonds and absent in starting materials. Intemal stand^dization for residues was not critical, since a high solid/fluid mass ratio was used in all experiments, and concentrations for major elements and most trace elements were not changed during the run significantiy. It should be noted that residues also may contain quenched fluid. Therefore calculated Dvalues for extremely incompatible elements are minimum values.

100-r

Cpx -•-CpxiRt (9:1) 0.1

0.01

Ba Sr Pb Nb Ta La Ce Zr Hf Sm Ti Tb Yb

Fig. 1: D (fluid/min) for diopside and a mixture of diopside and rutile at 5 GPa and 1000 Addition of only 10 % rutile lowers D-values for all HFSE by more than one order of magnitude.

100 T-

10-.

Gt1000.3 -•—GtlOOO, 5.5

0.1

0.01

0.001 - .

0.0001

Ba Sr Pb Nb Ta La Ce Zr Hf Sm Ti Tb Yb

Fig. 2: D (fluid/gt) at 1000 ''C and 3 (5.5) GPa. D sn/ D vb are around 100. At increasing pressure fractionation between compatible and incompatible trace elements becomes smaller.

Results for representative runs are illustrated in Figures 1 and 2. D(fluid/niineral) values for the elements determined are shown in the same element order used on primitive mantle-normalized diagrams, to facilitate comparison with trends in arc volcanics. The results for fluid/diopside (Fig. 1) show a slight, but relatively smooth fractionation of the LFSE, HFSE and REE; however there is no marked fractionation of HFSE from the REE or LFSE. Nb and Ta are soluble in the fluids, to a similar extent to those of LREE and some LFSE. The relative fractionation of the HFSE is markedly affected by the presence of rutile in the residue, which scavenges these elements from the fluid phase. D-values for Nb are always slightly higher than for Ta, but the significance of this is unclear. Fluid/garnet fractionation is much more pronounced than that for diopside/fluid (Fig. 2). For the REE this fractionation appears to be much stronger than that accomplished by silicate melt/garnet partitioning. For the fluid/garnet experiments, Ds^ / Dyb up to 100 could be determined. At 3 GPa, the overall fractionation of the elements is more pronounced than it is at 5.5 GPa. There may be some fractionation of the HFSE from REE and LFSE by garnet, particularly at 3 GPa; however, within the limits of our present results we are unsure, at present, if this is a significant feature. In addition to the results illustrated, we found that the addition of 1.5 M or 5 M HCl did not result in significant differences of trace element partitioning compared to those for pure water. The preliminary results obtained from these experiments suggest that: (1) LFSE and LREE may not be easily fractionated from the HFSE simply by calling on a role for fluids. In particular Nb and Ta seem to soluble in these high pressure fluids. (2) Only in the presence of residual titanite, will there be significant fractionation of the HFSE from the REE and LFSE. (3) Extreme REE-fractionation may be accomplished by afluidphase, as has been proposed by Stachel & Harris (1997). (4) In contrast to Keppler (1996), we do not find evidence to support a major effect of chloride on relative HFSE versus REE/LFSE partitioning. (5) Therefore it seems probable that HFSE depletion in island arc volcanics are mainly generated by specific retainment in a residual Ti-rich accessory phase. This mechanism will function for either dehydration or melting reactions.

References

Ayers J.C., Brenan J.M., Watson E.B., Wark D.A. & Minark W.G., 1992. A new capsule technique for hydrothennal experiments using the piston-cylinder ^paratus. American Mineralogist 77,1080-1086. Brenan J.M., Shaw H.F., Phinney D.L. & Ryerson F.J., 1994. Rutile-aqueousfluidpartitioning of Nb, Ta, Hf, Zr, U and Th: implications for high field strength element depletions in island-arc basalts. Earth Planetary Science Letters 128, 327-339. Brenan J.M, Shaw H.F.. Ryerson F.J. & Phinney D.L., 1995. Mineral-aqueousfluidpartitioning of trace elements at 900°C and 2.0 GPa: Constraints on the trace element chemistry of manUe and deep crustal fluids. Geochimica Cosmochimica Acta 59, 3331-3350. Foley S.F. & Wheller G.E., 1990. Parallels in the origin of the geochemical signatures of island arc volcanics and continental potassic igneous rocks: The role of residual titanates. Chemical Geology 85,1-18. Hawkesworth C.J., Gallegher K., Hergt J.M. & McDermott F., 1994. Destructive plate margin magmatism: Geochemistry and melt generation. Lithos 33,169-188. Keppler H., 1996. Constraints from partitioning experiments on the composition of subduction zonefluids.Nature 380,237-240. McCulloch M.T. & Gamble J.A., 1991. Geochemical and geodynamical constraints on subduction zone magmatism. Earth Planetary Science Utters 102, 358-374. Stachel, T. & Harris, J.W., 1997. Trace element geochemistry of silicate inclusions in diamond from Ghana - melt versus fluid involvement. Contributions to Mineralogy and Petrology (in press).

FRACTIONATION OF OXYGEN ISOTOPES AMONGST MELT, MINERALS AND HYDROUS FLUIDS , EGMONT VOLCANO, NEW ZEALAND Bob Stewart^ Richard Price^ and Ian Smith^ ^Department of Soil Science, Massey University, PO Box 11-222, Palmerston North, New Zealand ^School of Earth Sciences. LaTrobe University, Bundoora 3083, Australia ^Department of Geology, University of Auckland, PB 92019, Auckland, New Zealand

Fractionation of oxygen isotopes has been widely used in studies of hydrothermal systems and intrusive igneous bodies but there is a relatively small amount of data published on arc andesites. Much of the data on arc rocks is from whole rock analyses with only a few studies examining mineral-whole rock relationships, e.g. Harmon and Gerbe (1992). In our study we examined 13 samples from Egmont Volcano m Taranaki Province^ western North Island, New Zealand. The samples represent the four major stratigraphic groups we recognise on Egmont Volcano (Stewart et al, 1996), together with two older, olivine-bearing flows and a cumulate hornblende diorite. Samples were crushed and mafic minerals hand picked after partial concentration by magnetic separation. Titanomagnetite was hand picked with a magnetised dissection needle from a concentrate extracted with a hand magnet. Plagioclase was concentrated by high current magnetic separation. Whole rock (powder) and mineral separates were analysed for ^^O/^^O abundance at Oregon State University usmg a defocused CO2 laser in a CIF3 atmosphere. Oxygen was converted to CO2 for mass spectrometer measurement. NBS-30 biotite (5.10 % o ) was used as a standard and analyses of NBS-30 were 4.35+0.15 % o . All values were corrected to the accepted value. Precision on duplicate mineral analyses were; olivine ± 0.11 % o , clinopyroxene + 0.07 % o , amphibole + 0.2 /oo and plagioclase ± 0.32 % o . WHOLE ROCK DATA Whole rock ranges from 4.8 - 6.3 % o with a mean of 5.6 % o (Table 1) and is similar to the range and mean for MORB (Ito, et al, 1987) and oceanic arc basalts (e.g. Woodhead et al., 1987). This indicates that Egmont magmas did not interact with 5^^0-rich crust, just as they have not mteracted with crust containing radiogenic Pb, Sr or Nd (Price et al, 1992), even though the volcano sits on continental crust. Table 1 - Oxygen isotope data for whole rock and mineral separates from Egmont Volcano. Numbers are % o relative to Standard Mean Ocean Water (SMOW). olivuie

clinopyroxene

Sample

Whole R o a r

Cumulate

5.8

Burrell

5.4

5.8

5.7

BR6

5.5

5.6

Dawsons Falls

4.8

6.2

4.1

6.4

4.2

5.4

7.0

4.5

6.6

4.1

ESS

5.3

5.7

Tholoid

5.5

5.9

5.6

Sharks Tooth

5.5

South Flow

6.7

5.4

6.0

6.6 5.6

6.3

4.8

6.0

6.8

4.7

5.6

6.8

4.3 4.0

5.5

Ngarara Bluff

5.6

5.7

6.5

Staircase

5.8

5.5

6.3 6.3

5.3

Carapace

4.0

6.7

Rangitoto Flat

6.3

Turehu 6

6.1

5.9

5.1

6.2

5.3

4.3

Mean

5.6

6.0

5.6

5.5

5.6

6.5

0.4

0.2

0.3

0.2

0.1

0.3

Range

5.4-6.3

5.8-6.2

5.3-6.0

5.5-5.6

6.2-7.0

4.0-4.8

MINERAL DATA AND MINERAL-MELT FRACTIONATION Mineral ^^O/^^O compositions usually reflect the isotopic distribution between minerals and melt at the time of crystallisation and this is dependent upon both temperature and mineral-melt fractionation factors. There is little published data on the latter, the most widely quoted being that of Kalamarides

(1986) from the Kiglapait intrusion in Canada where a closed oxygen isotopic system is thought to have been maintained throughout crystallisation (Taylor and Sheppard, 1986). We derived melt ^Ô/^Ô compositions for our samples by using fractionation factors based on experimental data (Chiba, et al, 1989; Clayton, et al, 1989) and theoretical calculations by Kieffer (1982) and Zheng (1989; 1993a; 1993b). The two approaches are internally consistent. Mineral-melt fractionations are derived from the empu-ical calibration of Kalamarides (1986). Variability amongst individual analyses is less than that of the whole rock data which was also observed from Galanggung Volcano (Harmon and Gerbe, 1992). There is no evidence of stratigraphic variability in the Egmont data set. Data for mineral ^Ô/^Ô compositions is given m Table 1 and calculated equilibrium melt compositions in Table 2. Table 2. Melt in % o calculated from mineral-melt equilibria of Kalamarides, 1986. Stratigraphic groups are from Stewart, et al., (1996). Groundmass compositions are calculated by mass balance from measured data and modal analysis. Sample

Group

Whole Rock

Groundmass

Olivine-melt

Pyroxene-melt melt

-melt

6.1

Burrell

Summit

5.4

5.1

6.1

6.2

6.5

Tholoid

Summit

5.5

5.0

6.1

6.0

6.8

Sharks Tooth

Summit

5.5

4.6

6.3

6.5

6.7

Rangitoto Flat

Summit

6.3

6.7

Ngarara Bluff

Warwicks Castle

5.6

5.2

6.0

6.2

6.0

South Flow

Fanthams

5.0

5.9

6.6

6.3

ES8

Fanthams

6.1

6.5

6.6

6.0

5.9

6.3

6.1

Dawsons Falls

> 8000 yr BP

4.8

4.5

6.8

5.7

5.9

6.8

6.5

Turehu 6

> 8000 yr BP

6.1

6.9

6.5

5.4

5.8

5.9

Mean

5.7

5.5

6.6

5.9

6.0

6.4

Range

4.8-6.3

4.5-6.9

6.5-6.8

5.4-6.3

5.8-6.2

5.9-6.8

6.0-6.8

6.4

- S^^Ooi values are negative, rather than exhibiting the positive values expected if the two phases were in equilibrium., i.e. is higher than expected. The empirical mineral-melt fractionations suggest an equilibrium magma of 6.5 to 6.8 % o , higher than the other inferred magma compositions from Egmont and at or above the range for oceanic arcs which should most closely reflect wedge basalt compositions. Alternatively, the olivine may have interacted with a transitory, hydrous fluid (Rosenbaum, et al, 1994) in the magma source region which affected olivine only. Assuming equilibration, this fluid would have a value of 8.1 to 8.4 % o which falls in the range of "magmatic water" (6 - 9 % o ) and specifically "arc-type water" (9-11 % o ) postulated by Giggenbach (1992). In contrast to olivine, plagioclase-clinopyroxene, plagioclase-titanomagnetite and clinopyroxene-titanomagnetite plots exhibit tight arrays which parallel isotherms, indicating no substantial exchange with external fluids. EQUILIBRIUM MAGMA COMPOSITIONS Given that clinopyroxene is the most resistant to oxygen isotope exchange of the mmerals present, we can infer that the range of calculated melt compositions from clmopyroxene-melt fractionation represent that of magmas at or near the base of the crust. Thus melt compositions range from 5.4 - 6.3 % o with a mean of 5.9 % o (Table 2) which is similar to that quoted for MORB, oceanic arcs and Galanggung. Similar values are found for the other mineral phases (Table 2).

WHOLE ROCK AND GROUNDMASS ^W^O COMPOSITIONS Why does whole rock differfrommagma values estimated from calculated mineral-melt fractionations? Providing the assumptions underlying the calculations are sound, it suggests that the groundmass component is responsible for the variation. By combining the data for the individual minerals present with whole rock data and modal analyses, we calculated probable groundmass 5 O (Table 2). ESS and Rangitoto Flat have groundmass phases in equilibrium with the minerals except olivine and appear to have erupted with minimal oxygen isotope exchange during ascent. Turehu 6 has a groundmass significantly enriched in ^^O and appears to have interacted with ^^0-rich fluids in the mid to upper crust. The remaining lavas have groundmass compositions lower than their mineral-melt equilibrium liquid compositions and appear to have interacted with ^^0-depleted fluids. The most likely candidate is meteoric water from high altitude precipitation which in turn suggests exchange at shallow depths in the crust which only affected the groundmass. This also explains the greater variability of whole rock data compared with mineral data in the same rocks. REFERENCES Chiba, H., Chako, T., Clayton, R.N. & Goldsmith, J.R., 1989. Oxygen isotope fractionations involving diopside, forsterite, magnetite and calcite: applications to geothermometry. Geochimica et Cosmochimica Acta 53, 2985-2996. Giggenbach, W.F., 1992. Isotopic shifts in waters from geothermal and magmatic systems along convergent plate boundaries and their origin. Earth and Planetary Science Letters 113, 495-510. Harmon, R.S. & Gerbe, M., 1992. The 1982 - 83 Eruption at Galanggung Volcano, Java (Indonesia): oxygen isotope geochemistry of a chemically zoned magma chamber. Journal of Petrology 33, 585-609. Ito, E., White, W.M. & Gopel, C., 1987. The O, Sr, Nd and Pb isotope geochemistry of MORB. Chemical Geology 62, 157-176. Kalamarides, R.I., 1986. High-temperature oxygen isotope fractionation among the phases of the Kiglapait intrusion, Labrador, Canada. Chemical Geology 58, 303-310. Kieffer, S.W., 1982. Thermodynamics and lattice vibrations of minerals: 5. Applications to phase equilibria, isotopic fractionation, and high pressure thermodynamic properties. Reviews in Geophysics and Space Physics 20, 827-849. Price, R.C.,McCulloch, M.T., Smith, I.E.M. & Stewart, R.B., 1992. Pb-Nd-Sr isotopic compositions and trace element characteristics of young volcanic rocks from Egmont volcano and comparisons with basalts and andesites from the Taupo Volcanic Zone. Geochimica et Cosmochimica Acta 56, 941-953. Rosenbaum, J.M., Walker, D., & Keyser, T.K., 1994. Oxygen isotopic fractionation in the mantle. Geochimica et Cosmochimica Acta 54,4767-4777. Stewart, R.B., Price, R.C. & Smith, I.E.M., 1996. Evolution of high-K arc magma, Egmont volcano, Taranaki, New Zealand: evidence from mineral chemistry. Journal of Volcanology and Geothermal Research (in press). Woodhead, J.D., Harmon, R.S. & Fraser, D.G., 1987. O, S, Sr and Pb isotope variations in volcanic rocks from Northern Marianas Islands: implications for crustal recycling in intraoceanic arcs. Earth and Planetary Science Letters 83, 39-52. Zheng, Y., 1993. Calculation of oxygen isotope fractionation in anhydrous silicate minerals. Geochimica et Cosmochimica Acta 57, 1079-1091.

ISLAND ARC M A G M A G E N E R A T I O N FROM THERMALLY ZONED MANTLE DIAPIRS Yoshihiko Tamura Institute for Study of the Earth's Interior, Okayama University, Misasa, Tottori, 682-01, Japan

One of the characteristics of island arc volcanism is the generation of magmas with a wide range of Si02 content, typically ranging from basalt (Si02 < 53 wt. %) to rhyolite (Si02 > 70 wt. %). Generally, a continuous magma series comprising basalt, andesite, dacite, and rhyolite can be produced by fractional crystallization of mantle-derived basaltic magmas (e.g., Gill, 1981). However, a mystery in some arc volcanoes is that there exists voluminous andesite and dacite which cannot be generated by fractionation of basaltic magma, yet they coexist with basalt and it's fractionated melts (andesite, dacite) (e.g., Tamura, 1994). What is happening under these arc volcanoes to produce this bimodal volcanism? One explanation has been given from the point of view of experimental petrology. For example, Kushiro (e.g., 1974, 1990) stated that primary basaltic and andesitic magmas can be generated by .partial melting of upper mantle peridotite under anhydrous and hydrous conditions, respectively (Figure 1). The variation in Si02 in these primary magmas would be augmented by fractional crystallization (Kushiro, 1990). Most andesites and dacites of arc volcanoes, however, have lower MgO contents than the experimental melts, and cannot have equilibrated with mantle peridotite. To make matters worse, high-magnesian andesites are rare in arc volcanics, and have been deemed to be unusual among arc parental magmas (e.g., Crawford et al., 1989). Fig. 2 Fractional crystallization of basalt

fit

\ \

Fractional crystallization

Fractional crystallization of magnesian andesite

^ Magnesian ^ V^^desite m a g i m ^

Partial melting under anhydrous conditions

Partial melting under hydrous conditions

Calc-alkaline series

Andesite

Basalt 53

63 S i 0 2 wt %

Dacite

Rhyolite 70

F i g . 1. Schematic diagram illustrating the hypothesis that arc magmas result from parallel fractional crystallization of basalt and magnesian andesite, which are generated by partial melting of upper mantle peridotite under anhydrous and hydrous conditions, respectively. F i g . 2. Mantle-derived basalt and magnesian andesite would evolve in higher level magma chambers to produce two rock series. Those formed via fractionation from basalt through andesite to dacite are referred to as rocks belonging to the tholeiitic series: those formed via fractionation from magnesian andesite through andesite and dacite to rhyolite are referred to as rocks belonging to the calc-alkaline series. Most rocks of arc volcanoes do not represent mantle-derived primary magmas. Suppose, however, that basalt and magnesian andesite are generated simultaneously in the mantle wedge of the subduction zone. These two primary magmas would evolve into two distinct magma series through crystallization differentiation. One would be a magma series derived from basaltic magma (basalt, andesite, dacite, and rhyolite), and the other would be a magma series derived from magnesian andesitic magmas (andesite, dacite and rhyolite). Tamura (1994) showed that arc lavas of the Shirahama Group of south-central Honshu, Japan, were probably generated by mantle-derived bimodal magmatism of basalt and magnesian andesite through fractional crystallization. Those formed via fractionation from basalt through andesite to dacite are referred to as rocks belonging to the tholeiitic series; those formed via fractionation of magnesian andesite are referred to as rocks belonging to the calc-alkaline series (Figure 2). Magmatic temperatures inferred by the two-pyroxene thermometer show unambiguous differences between the tholeiitic series and calc-alkaline series of the Shirahama Group. Generally, temperatures of 9501100°C are obtained from tholeiitic samples, whereas 800-900°C is indicated for calc-alkaline samples (Tamura, 1994, 1995). At a given Mg# (e.g. augite, orthopyroxene), the composition of the groundmass in calc-alkaline series rocks is predominantly rhyolitic, whereas, in the tholeiitic series it is predominantly andesitic and dacitic (Tamura, 1995^

Tamura and Nakamura (1996) presented Sr-Nd isotopic data of the arc lavas of the Shirahama Group. They showed that only very small differences exist among the Shirahama tholeiitic series, calc-alkaline series, and mixed lavas; the Sr and Nd isotopic data cluster tightly within the mantle array, which preclude the involvement of a crustal component (Figures 3 and 4). These data require that the mantle source of both the tholeiitic series and the calc-alkaline series are isotopically identical. Fig. 3

Fig- 4 X O •

0.7040

0.5134

Mixed rocks Calc-alkaline series Tholeiitic series

0.5132

I ^

0.7035

0.5130

0.5128

0.7030

0.7025

0.5126 55 60 Si02 wt. <

0.7030

0.7040

0.7050

S^Sr/S^Sr

F i g . 3. ^'^Sr/^^Sr vs Si02 concentration for the Shirahama rocks. The tholeiitic series, calc-alkaline series, and mixed rocks have the same ^^Sr/^^Sr, irrespective of Si02 concentration. F i g . 4 . 1 "^^Nd/l ^"^Nd vs ^'^Sr/^^Sr for the Shirahama rocks. The data cluster tightly within the mantle array. Analyses from 38 other arc volcanoes have been compiled to investigate the intravolcano variability of ^'^Sr/^^Sr. Twelve of these display no intravolcano strontium isotopic variability (Figure 5), as is the case with the Shirahama Group, but others show significant variation of ^^Sr/^^Sr from individual volcanic centers, presumably reflecting crustal contamination. Interestingly, in most of these volcanoes, rocks that contain the least radiogenic strontium are andesites or dacites (Figure 6). The assimilation of crustal material by two primary Fig. 6 magmas, having the same ^'^Sr/^^Sr but different 0.7070 temperatures, can explain the observed variation of ^^Sr/^^SrNevados de Payachata Si02 (Figure 7) (Tamura and Nakamura, 1996).

0.7045

0.7040

0.7035

0.7030

0.7025 SiO^ wt.

F i g . 5. ^'^Sr/^^Sr vs Si02 content for rocks from arc volcanoes having a total isotopic variability < 0.0003 within each volcano. The lavas from the individual volcanic centers have a very narrow range of Sr isotopic composition from basalt to dacite or rhyolite. The lowest and the highest Si02 rocks are plotted and connected by lines. All of these volcanoes are underlain by thin continental crust (<-30 km). F i g . 6. ^^Sr/^^Sr vs Si02 content for rocks from arc volcanoes having a total isotopic variability > 0.0003. Arc volcanoes characterized by thicker continental crust show isotopic heterogeneity. The lowest Si02 rocks and the lowest ^^Sr/^^Sr rocks are plotted and connected. In most of these volcanoes, rocks that contain the least radiogenic strontium are andesites or dacites. Basalt from five volcanoes have the lowest 87Sr/86Sr (not shown), but the observed differences between these rocks and the lowest ^"^Sr/^^Sr andesites are < 0.0001, suggesting ^^Sr/^^Sr are similar in basalts and andesites in these volcanoes.

Fig. 7. Schematic evolution of hotter basalt and cooler magnesian andesite during ascent through continental crust and in a crustal magma chamber. Lines 1: mantle-derived basalt and magnesian andesite assimilate wallrocks during ascent through granitic crust (AEC process). The V00O amount of contamination is greater in hotter basaltic magma than in cooler magnesian 00 dL—i ^ andesite magma. Lines 2: fractional crystallization plays a dominant role when wallrocks have refractory compositions, such as gabbros, and/or magmas have lower temperature magnesian andesite than fusion temperatures of crustal rocks. Lines basalt magma magma 3: magmas in magma chambers evolve through simultaneous fractional crystallization and Si02 wt. % walhock assimilation (AFC process). Lines 4: magma mixing between parent and daughter magmas and/or between tholeiitic and calc-alkaline series magmas could produce magmas with intermediate compositions. The erupted magmas would have compositions shown by the shaded area; basalt has higher ^'^Sr/^^Sr than some coexisting andesite and rhyolite has lower ^^Sr/^^Sr than some coexisting andesite. C/D

Given that primary arc magmas are derived from mantle diapirs which have ascended through the mantle wedge of the subduction zone (Sakuyama, 1983), the following two constraints must be satisfied; (1) wet and cool peridotite and dry and hot peridotite are associated within single diapirs (Tamura, 1994), or, dry and hot diapirs intrude into wet and cool uppermost mantle (Kushiro, 1990), and (2) the mantle diapir as a whole has uniform isotopic chemistry (Tamura and Nakamura, 1996), or, the mantle diapir and the surrounding mantle have uniform isotopic chemistry in the case of Kushiro (1990). MORB and its cumulates

Hydrous iV?-^?-'?-'?'^ mantle diapir

diffusion of

H2O

Oceanic upper mantle

Residual peridotite (oceanic upper mantle)

Mantle diapir

Resldud"pe"ndotite (arc mantle diapir)

diffusion of heat

1000 °C Anhydrous cool mantle

Crust

basalt ^ * magnesian ^ andesite

mantle W W

V - ^

Anhydrous hot mantle

1400 'C

1300 "C dry \ 1 1 0 0 "C wet

Fig. 8. (a) Schematic cross-section of island arc following Tatsumi et al. (1983); b-d correspond to the three panels following, (b) Mantle diapir consisting of hydrous peridotite is formed in the lower part of the mantle wedge above the slab, (c) Diapir is heated during ascent through mantle wedge. H 2O diffuses to the surrounding dry mantle, (d) Heated diapir, which has wet and cool interior and dry and hot rind produces basalt and magnesian andesite, respectively, (e) One mantle diapir might produce one volcano (e.g., Sakuyama, 1983; Tatsumi et al.. 93

1983). Oceanic crust and oceanic uppermost mantle should intervene between the island arc volcano and the mantle diapir. (f) Postulated cross section from arc volcano to mantle diapir. Coexistence of MORB and arc rocks and their relevant cumulates would be expected. Figure 8 shows the model of Tamura (1994), which has advantages over the model of Kushiro (1990) for producing isotopically uniform magmas. A hydrous mantle diapir, which is formed from hydrated mantle peridotite above the subducting slab, would have uniform Sr-Nd isotopes in its infancy (Figure 8b). It would be heated and dehydrated during ascent through the mantle wedge, resulting in a compositionally and thermally zoned diapir (Figure 8c). The ascent rate is fast enough (Tamura, 1994) to keep the original isotopic ratios. Partial melting might occur during ascent, but magmas segregate from the partially melted diapir at shallower levels (e.g., Kushiro, 1983; Tatsumi et al., 1983). The diapir might be forced to stop ascending when the surrounding uppermost mantle comes to have higher viscosity according to the decrease of temperature, that is, the segregation depths of magmas might depend on the thermal structure of the uppermost mantle (Tatsumi et. al., 1983). The diapir would generate basalts from it's dry and hot rind, and magnesian andesites from it's cooler and wet interior (Figure 8d). Tholeiitic series (basalt, andesite, dacite, and rhyolite) and calc-alkaline series (andesite, dacite, and rhyolite) are produced through parallel fractional crystallization of basalt and magnesian andesite, respectively (Tamura, 1994). There is no consensus on the origin of ophiolitic complexes, as well as the origin of island arc magmas, but Osbom (1969) attempted to combine these two geological mysteries and solve them together. In the oceanic setting, island arc volcanoes would be constructed on oceanic crust and uppermost mantle (Figure 8e). The cross section of such a crust-mantle complex (Figure 8f) could show an apparently inconsistent coexistence of two different characteristics, one of which was derived from the island-arc, and the other from a mid oceanic ridge. Since Miyashiro (1973), it has been recognized that most ophiolites are accompanied by arc magmas, and they cannot be simple fragments of oceanic crust and underlying upper mantle. Ophiolitic complexes might be the best place to examine the mantle diapir model presented here. REFERENCES Crawford, A. J., Falloon, T. J., & Green, D. H., 1989. Classification, petrogenesis and tectonic setting of boninites. In: Crawford, A. J. {td) Boninites and related rocks. London: Unwin Hyman, 1-49. Gill, J. B., 1981. Orogenic andesites and plate tectonics. New York: Springer-Verlag. Kushiro, I., 1974. Melting of hydrous upper mantle and possible generation of andesitic magma: an approach from synthetic systems. Earth and Planetary Science Letters 22, 294-299. Kushiro, I., 1983. On the lateral variations in chemical composition and volume of Quaternary volcanic rocks across Japanese arc. Journal of Volcanology and Geothermal Research 18, 435-447. Kushiro, I., 1990. Partial melting of mantle wedge and evolution of island arc crust. Journal of Geophysical Research 95, 15929-15939. Miyashiro, A., 1973. The Troodos ophiolitic complex was probably formed in an island arc. Earth and Planetary Science Letters 19, 218-224. Osbom, E. F., 1969. The complementariness of orogenic andesite and alpine peridotite. Geochimica et Cosmochimica Acta 33, 307-324. Sakuyama, M., 1983. Petrology of arc volcanic rocks and their origin by mantle diapirs. Journal of Volcanology and Geothermal Research 18, 297-320. Tamura, Y., 1994. Genesis of island arc magmas by mantle-derived bimodal magmatism: evidence from the Shirahama Group, Japan. Journal of Petrology 35, 619-645. Tamura, Y., 1995. Liquid lines of descent of island arc magmas and genesis of rhyolites: evidence from the Shirahama Group, Japan. Journal of Petrology 36, 417-434. Tamura, Y., & Nakamura, E., 1996. The arc lavas of the Shirahama Group, Japan: Sr and Nd isotopic data indicate mantle-derived bimodal magmatism. Journal of Petrology 37, in press. Tatsumi, Y., Sakuyama, M., Fukuyama, H., & Kushiro, I., 1983. Generation of arc basalt magmas and thermal structure of the mantle wedge in subduction zones. Journal of Geophysical Research 88, 5815-5825.

U-SERIES DISEQUILIBRIA, MAGMA PETROGENESIS AND FLUX RATES ALONG THE DEPLETED TONGA-KERMADEC ISLAND ARC Simon Turner^, Chris Hawkesworth^ Nick Rogers^ Jessica Bartlett^ Ian Smith^ and Tim Worthington^ 1 Department of Earth Sciences, The Open University, Milton Keynes MK7 6AA, England 2 Department of Geology, University of Auckland, New Zealand The highly depleted intra-oceanic Tonga-Kermadec island arc forms an end-member of arc systems and a unique location in which to isolate the effects of the slab flux. High precision TIMS U, Th, Sr, Nd and Pb isotopes, along with complete major and trace element data, have been obtained on an extensive sample set comprising 58 lavas along the arc. The data provide information on both the petrogenesis of the lavas and also material transport rates and mantle wedge dynamics. Ca/Ti and Al/Ti ratios extend from values appropriate to a N-MORB source in the southern Kermadecs to very high ratios in Tonga (Figure 1) and these are interpteted to reflect increasing degrees of depletion of the mantle wedge due to backarc basalt extraction. This is mirrored in very low incompatible trace element concentrations.

Tonga Tafahi/Niuatoputapu • Fonualei/Late 50-Metis Shoal/Kao/ Tofua/Hunga Ha'apai 4 0 - - ^ 'Ata 60 --

Alg/Ti 8

depleted peridotite

Kermadec • Raoul O Macauley L'Esperance/ O Rumble seamounts/ Ngatoro Ridge =F=

30 • - 1+ Niuafo'ou 20 --

10 + 0

fertile peridotite — H

Cag/Tis FIGURE 1. Alg/Tig versus Cag/Tig (subscript refers to compositions at 8% MgO) illustrating the variable degrees of depletion of the mantle wedge (produced by basalt extraction at the backarc spreading centres) along the arc from highly depleted in Tonga (high Alg/Tig, Cag/Tig) to less depleted in the southern Kermadecs (similar to N-MORB or fertile peridotite melts). The compositions of experimental melts (also calculated at 8% MgO) from fertile (Hawaiian pyrolite) and depleted (Tinaquillo Iherzolite) (Falloon et al., 1988) are shown for comparision (outlined fields). Lava data from individual islands were extrapolated to 8% MgO by regression along linear or polynomial (in the case of inflected data arrays) fits to the data, while experimental data were extrapolated by removal of forsteritic olivine (see (Turner & Hawkesworth, 1995) for a fuller explanation). In Figures 2 and 3, the isotope data emphasise the need for at least three components in the petrogenesis of the lavas: (i) the mantle wedge similar to the source of the Lau Basin MORB; (ii) a component with elevated 207py204pt) towards which the Kermadec and southern Tongan lavas extend; (iii) a component characterised by high Ta/Nd and low ^^^Nd/^^Nd observed only in the northernmost Tongan islands of Tafahi and Niuatoputapu. Component (ii) is identified as average pelagic sediment on the downgoing Pacific plate and mass balance calculations indicate that less than 0.5% of this is recycled into the arc lavas; essentially all the subducted sediment is returned to the upper mantle (--0.03 km^ yr'^). Exceptionally low concentrations of Ta and Nb in the lavas require, either that the sediment addition takes place as very small partial melts in equilibrium with residual rutile, or that the addition of silicious melts and oxidising fluids stabilise rutile in the wedge. Component (iii) is identified as volcaniclastics from the Louisville Ridge. This unique sediment tracer takes - 4 Myr from the time of subduction to its first appearence in the arc lava signature.

15.70'

JO Q.

15.55'

15.50--

Louisville volcaniclastics

sediments

_Q 15.65 • CL g 15.60'

15.45' 18.0 18.2 18.4 18.6 18.8 19.0 19.2 19.4

206pb/204pt) FIGURE 2. 20'7pi;j/204p5.206p|3/204p|3 illustrating the generally restricted isotopic variation of the Tonga-Kermadec lavas. The Kermadec and southern Tonga lavas lie on a mixing array between mantle wedge (represented by the MORE arrays from the Lau Basin) and the average subducted sediments (ATS). Lavas from Tafahi and Niuatoputapu are strongly displaced to higher 206p5/204p5 towards the Louisville volcaniclastics. The backarc island of Niua fo'ou has higher 207p5/204pb but also 206pb/204pb. Lau Basin Indian and Pacific MORB fields from Hergt & Hawkesworth (1994). Average subducted sediment from Plank and Langmuir (1996), Louisville volcaniclastic-dominated sediments from the base of DSDP 204 from Hergt & Pearce (unpublished data). Symbols as in Figure 1.

0.06

Louisville volcaniclastics

0.05 0.04 •D 0.03 1-

0.02 0.01 0.00 0.5123

0.5125

0.5127

0.5129

0.5131

0.5133

143Nd/144Nc| FIGURE 3. Ta/Nd versus ^^^Nd/^^Nd showing the consistently low Ta/Nd of the Kermadec and southern Tongan rocks possibly decreasing with decreasing ^^^Nd/^^^Nd consistent with the addition of small amounts of pelagic sediment. In contrast, the Tafahi and Niuatoputapu lavas form an unusual array of increasing Ta/Nd coupled with decreasing ^^^Nd/^^Nd. As shown, this can be explained by contributions from a subduction component which is dominated by the Louisville volcaniclastics. Symbols and data sources as in Figures 1 & 2.

In addition to the wedge and sediment contributions, strong enrichments of Rb, Ba, U, K, Pb and Sr, relative to Th, Zr and the REE require addition of a component enriched in LILE. Recent experimental data show that hese elements are preferentially mobile in oxidising aqueous fluids (Brenan et al. 1995) and so the high LIL/HFSE ratios are interpreted as a fluid signal (see Figure 4). Because U is fluid mobile relative to Th, the large ^^^U excesses ((^^^Th/^^^U) = 0.8-0.5), especially well developed in the more depleted lavas (Figure 5) also implicates the role of a fluid component. 1200

fluidi

1000 +

Ba/Th

800 600 400 200 0

'•o

Louisville volcaniclastics

Lau Basin IMORB

0.702

0.704

0.706

sediments

0.708

0.710

FIGURE 4. Ba/Th versus^'^Sr/^^Sr showing that the Tonga-Kermadec lavas are characterised by high Ba/Th and low ^'^Sr/^^Sr. The interpretation is that the Tonga-Kermadec lavas are dominated by fluid-man tie-wedge interaction with only minor contribution from the subducted sediments. This also implies that the ^'^Sr/^^Sr ratio of the fluid is -0.7035 and thus likely to be derived from dehydration of the subducting altered oceanic crust. Symbols and data sources as in Figures 1 & 2.

0.2

0.6

1.0

1.4

1.8

2.2

2.6

(238u/232Th) FIGURE 5. (^^^Th/^^^Th) versus (^^^U/^^^Th) "equiline" diagram for the Tonga-Kermadec lavas showing the large ^^^U excesses in the more depleted Tonga lavas. A pseudo-isochron passing through the bottom of the data array indicates a minimum elapsed time of 50 000 yr between fluid addition from the slab and eruption. Symbols as in Fig. 1.

To summarise, any model for the petrogenesis of the Tonga-Kermadec lavas requires contributions from the mantle wedge, subducted sediment (pelagic for southern Tonga and the Kermadecs and a mixture of pelagic sediment with the Louisville volcaniclastics for Tafahi and Niuatoputapu) and also a fluid component inferred to be largely derived from the subducting altered oceanic crust. This multi-component model is illustrated in Figure 4. The Louisville sediment tracer constrains the transit time for the sediment component while U-series isotopes can be used to place constraints on other aspects of element transfer. The fluid contribution to the lava source has been calculated as ppm Rb, 10 ppm Ba, 0.02 ppm U, 600 ppm K, 0.2 ppm Pb and 30 ppm Sr. It has ^^ST/^^ST = 0.7035 and 206pb/204pb ^ ig 5 ^nd thus is inferred to be derived from dehydration of the subducting altered oceanic crust. U-Th isotope disequilibria reflect the time since fluid release from the subducting slab and a pseudo-isochron through the lowest (^^^h/^^^Th) lavas constrains this to be ~ 50 000 yr (Figure 5). Significantly, U-Th isotope data record similar timescales in the Lesser Antilles (-40 000 yr, Turner et al., 1996) and in the Marianas (30 000 yr, Elliott et al., 1996) which provides encouragement that these data reflect some general aspect of the flux rates beneath island arcs. Large ^^^Ra excesses have also been reported from Tonga-Kermadec ((226Ra/230Th) = 1.5-3.0, Gill & WilUams, 1990). Since 226Ra will return to secular equilibrium with 230Th ((226Ra/230xh) = 1) within 7500 yr of Ra/Th fractionation the 238u/230jh and ^^^Ra/^^^Th disequilibria are clearly decoupled (see also Turner et al., 1996). This is an unexpected result and clearly the 226Ra/230xh disequilibria must have developed after the process responsible for the major U/Th fractionation. It is suggested that Th-Ra isotope disequilibria record the time since partial melting and thus indicate rapid channelled magma ascent. Olivine gabbro xenoliths from Raoul are interpreted as cumulates to their host lavas with which they form zero age U-Th isochrons indicating that minimal time was spent in magma chambers. The subduction signature is not observed in lavas from the backarc island of Niuafo'ou and thus does not penetrate as far 200 km beyond the arc front volcanoes. These were derived from partial melting of fertile peridotite at 130-160 km depth with melt rates around 2 x 10"^ kg m"^ y r ^ possibly due to volatiles released from the breakdown of phengite and lawsonite in the underlying slab at 200 km depth. Fluid-induced melting rates beneath the arc are inferred to be greater than this. These results from the Tonga-Kermadec arc can be used to place constraints on recent dynamic models for island arcs (e.g. Davies & Stevenson, 1992). REFERENCES Brenan J.M., Shaw H.F., Ryerson F.J., & Phinney D.L., 1995. Mineral-aqueous fluid partitioning of trace elements at 900 "C and 2.0 Gpa: constraints on the trace element chemistry of mantle and deep crustal fluids. Geochim, Cosmochim. Acta 59, 3331-3350. Davies J.H. & Stevenson D.J., 1992. Physical model of source region of subduction zone volcanics. J. Geophys. Res. 97, 2037-2070. Elliott T., Plank T., Zindler A., White W., & Bourdon B., 1996. Element transport from subducted slab to juvenile crust at the Mariana arc. J. Geophys, Res. (submitted). Ewart A. & Hawkesworth CJ., 1987. The Pleistocene-Recent Tonga-Kermadec arc lavas: Interpretation of new isotopic and rare earth data in terms of a depleted mantle source model. J. Petrol. 28, 495-530. Falloon T.J., Green D.H., Hatton C.J., & Harris K.L., 1988. Anhydrous partial melting of a fertile and depleted peridotite from 2 to 30 kb and application to basalt petrogenesis. J. Petrol. 29, 1257-1282. Gill J.B. & Williams R.W., 1990. Th isotope and U-series studies of subduction-related volcanic rocks. Geochim. Cosmochim. Acta 54, 1427-1442. Hergt J.M. & Hawkesworth C.J., 1994. Pb-, Sr-, and Nd-isotopic evolution of the Lau Basin: implications for mantle dynamics during backarc opening. Proc. ODP Sci. Res. 135, 505-517. Pearce J.A., Ernewein M., Bloomer S.H., Parson L.M., Murton B.J., & Johnson L.E., 1995. Geochemistry of Lau Basin volcanic rocks: influence of ridge segmentation and arc proximity. In Volcanism Associated with extension at consuming plate margins, (ed. J.L. Smellie), pp.53-76. Geol. Soc. Lond. Spec. Publ. 81. Plank T. & Langmuir C.H., 1996. The geochemical composition of subducting sediment and its consequences for the crust and mantle. Chem. Geol. Spec. Publ. (in press). Turner S. & Hawkesworth C., 1995. The nature of the sub-continental mantle: constraints from the major element composition of continental flood basalts. Chem. Geol. 120, 295-314. Turner S., Hawkesworth C., van Calsteren P., Heath E., Macdonald R., & Black S., 1996. U-series isotopes and destructive plate margin magma genesis in the Lesser Antilles. Earth Planet. Sci. Lett. 142, 191-207.

USING HELIUM AND CARBONCCO^) ISOTOPES TO DISTINGUISH BETWEEN SOURCE AND CRUSTAL CONTRIBUTION IN ARC MAGMAS. A CASE STUDY FROM THE LESSER ANTILLES ISLAND ARC. M.C. van Soest^ D.R. Hilton^, R. Kreulen^ ^ Department of Earth Sciences, Vrije Universiteit Amsterdam, De Boelelaan 1085, 1081 HV Amsterdam, The Netherlands, email: soet@geo.vu.nl. 2 GRD, Scripps Institution of Oceanography, UCSD, La Jolla, San Diego, USA. ^ Faculteit der Geochemie, Instituut voor Aardwetenschappen, Universiteit Utrecht, Budapestlaan 4, 3505 TA Utrecht, The Netherlands.

Introduction A quantitative assessment of the fate of sediments at subduction zones is a vital component in understanding the cychng history between the mantle and the exosphere. The principal difficulty in making this assessment is determining the proportion of subducted sedimentary material that is recycled to the surface via arc-related volcanism versus the proportion that bypasses the zone of magma generation and is lost to the (deeper) mantle. The geochemistry of arc-related lavas provide an obvious means to tackle this problem and estimates of the contribution of subducted sediments to the source of such lavas have been made using a variety of geochemical techniques. Unfortunately, this approach has failed to resolve the issue completely due to the ambiguity of distinguishing 'sedimentary signatures' derived from subducted sediments from the effects of crustal contamination of mantle-derived magmas at shallow-levels in the crust. In theory, the use of helium isotopes (^He/'^He) offers the possibility of circumventing this ambiguity over 'sedimentary signatures' due to its enhanced sensitivity for tracing crustal contamination effects close to the surface. In subduction zones where oceanic crust and marine sediments are being subducted, helium isotope variations invariably overlap with or fall very close to the mid-ocean ridge basalt (MORB) value of SR^ (where RA = the atmospheric value) (Poreda and Craig, 1989). This observation is consistent with a helium provenance in the mantle wedge and not in subducting sediments: indeed it has been argued that sediments are a highly unlikely vehicle to transport heUum into the mantle due to its high diffusivity in sedimentary materials (Hilton et al., 1992). Only in regions such as the Andes where volcanism occurs through thickened uppermost crust are significant deviations from MORB-like ^He/^He ratios observed (Hilton et al., 1993): the implication being that crustal (radiogenic) helium can contaminate mantle melts in such circumstances. These characteristics make helium isotopes an excellent tracer for crustal contamination effects and any isotopic signature significantly lower than the mantle ratio of SR^ signifies addition of '^He after the magma has left the mantle. In addition, when a MORB-like ratio is measured for helium isotopes any 'sedimentary signal' indicated by other tracers must be derived by a sediment contribution from the subducting slab. Coupling helium isotope investigations with a tracer of sediment type - such as carbon isotopes - offers the additional possibility of identifying the nature of the subducted sedimentary component (marine versus terrigenous sediments for example). To investigate these possibilities, we have undertaken a regional survey of the helium and carbon isotope geochemistry of the Lesser Antilles volcanic arc (Figure la). This arc was chosen because (a) the presence and accessibility of currentiy and recentiy active volcanoes along the whole strike of the arc - approximately 850 km from Grenada to Saba; (b) it has been well studied for major and trace element chemistry and Sr, Nd, and Pb isotope chemistry (e.g. White and Dupre, 1986); and (c) the character and amount of the sediments in front of the trench changes along strike from mainly terrigenous and > 4km. thick in the south (due to the proximity of South America, Orinoco Delta, Guyana shield - Westbrook et al., 1984) to marine/pelagic and < 400m. thick in the north. We utilised both geothermal fluids (fumaroles and hot springs) and young phenocryst-bearing lavas to establish helium and carbon relations along the arc. Results Helium isotopes:

There is a predominant mantie He contribution throughout the arc (Figure lb). Highest ^He/'^He ratios are coincident with MORB helium (8RA) and occur towards the north of the arc - on the islands of Martinique, Dominica, Guadeloupe and Montserrat. In the south, from Grenada to St. Lucia, ^He/'^He ratios are lower and lie between 5.2-6.5Ra. 99

(R/R

Figure la

vs. Latitude Saba

18^

Saba Nevis® Montserrafl^

Nevis

Atlantic Ocean

Montserrat Guadeloupe

Guadeloup

16'_ Dominica

Dominica^ Martinique! St. Lucia (

Martinique

14' St. Lucia

St. Vincent #

dA-

Grenada J"

St. Vincent

12: Grenada

J2L

d^^C vs. Latitude

(FVR )c \

He \ts. Latitude

Nevis

Montserrat Guadeloupe

Dominica

Martinique

St. Lucia St. Vincent

10; . • ^

Fumaroles Dissolved Water Phase Hot Bubbling Acid Gases Olivine Ph>enocrysts

CO y^He

Figure 1: a) Location map of the Lesser Antilles islands, b) helium isotope ratios vs latitude The dashed line connect the samples with the highest ratio in a geothermal sample per island, c) 3l3C isotopes vs. latitude, the dashed line connects the samples with the highest helium isotopes, d) C02/3He vs. latitude, with again the dashed line to connect the samples with the highest helium ratios. In all 3 graphs the MORB average is also included. Discussion in the text.

Carbon isotopes: The regional variations established by helium isotopes can also be observed in the carbon isotopes (Figure Ic). The northern portion of the arc (Martinique-Montserrat) is characterised by values between -2 and -4 %o vs. PDB. These values are clearly higher than typical MORB values ( -5 to -7 %o vs. PDB) and would seem to

100

represent mixtures between MORE and (heavy) carbonate carbon. The south of the arc (Grenada-St. Lucia) is characterised by MORB-like carbon isotope ratios (values around -6 %o). CO/He: There is a strong negative correlation between ^He/'^He and C02/^He ratios in hydrothermal fluids (Figure Id). At Montserrat (8.1RA), C02/^He ratios fall in the range 10^-10^®; however most of the MORB-like ^He/^He ratios of the northerly islands have C02/^He ratios in the range 1-5 x 10^®. In contrast, the southern islands characterised by lower ^He/'^He ratios - have distinctly higher C O / H t values of Compared to MORB (C02/^He --2x10^) all islands appear to have undergone (some) addition of CO2. Discussion Data Selection As most of the analyses produced to date were made on fluid samples, a caveat regarding the integrity of such samples appears in order. For the most part, the possibihty of crustal additions unrelated to magma effects but due to fluid flow is assumed to be non-existent when the travel paths between the source of the fluids and the point where they reach the surface is short. As travel paths become longer and (fluid) temperature decreases, the possibility of high-level interaction between fluids and country rock increases. To avoid including data points compromised in such a manner, we have excluded any fluid samples greater than 3km. from the closest volcanic edifice from further consideration. Possible end-member compositions Assuming the mantle wedge in the Lesser Antilles has a MORB composition (^He/^He: 8 R^, d^^C: -5 to -7%o and C02/^He: 2 x 10^) our data points can be explained by mixing of the MORB source with one or more other sources. There are two obvious candidates: (1) A subducted slab/carbonate source. Compared to the MORB source this component is expected to be extremely depleted in helium (assuming He is not subducted) but is likely enriched in CO2. Isotopic ratios for carbon would be that of carbonates (0 to -2%o), and (2) An upper crustal reservoir. This reservoir would consist of the older arc rocks and old oceanic basement. Compared to the MORB source, this reservoir would be enriched in '^He through radiogenic production and will therefore have a low ^He/'^He ratio. This source will be depleted in CO2 compared to the MORB source and could have a more organic (lighter) carbon isotope signature. Mixing Processes Mixing curves between MORB and the two above crustal reservoirs have been plotted in an isotope-isotope diagram (Figure 2a). The K-values represent the (He/C) ratio of the MORB-reservoir divided by the same ratio of one of the crustal reservoirs. Values > 1 indicate the dominance of MORB helium and/or crustal carbon, while low values (<1) mean dominance of MORB carbon and/or crustal helium. Figure 2a shows that both the northerly and southerly datasets can be explained by binary mixing processes. For the northern group, mixing takes place in the mantle between the MORB source and a subducted slab source, while the southern group is best explained by mixing the MORB source with crustal material at shallow levels. In Figure 2b we plot mixing curves between the same endmember compositions but on a carbon isotope versus C02/^He plot. We note that the scenario of simple binary mixing between these sources is not valid as much more CO2 appears to have been added to the MORB source in comparison to the carbon isotope shift given by the calculated mixing curve. This case of 'excess' CO2 implies that we are dealing with either 3 component mixing between our sources or the possibility that the composition of the crustal sources differs from the compositions we have used in our calculations. For the northern group, a wider range in carbon isotopic composition of the subducted source is a distinct possibility: carbon isotope analyses on sediments from two DSDP cores in front of that portion of the arc have so far given values of -0.18 and -1.83 %c vs. PDB. According to the initial reports of the DSDP shipboard party (Westbrook et al., 1984, Moore et al., 1982), subduction of these particular sediments is a realistic possibility. For the southern group, three component mixing appears to be a more viable explanation although in this case it is fortuitous that three different islands have MORB-like values for carbon isotopes while experiencing very different amounts of CO2 addition. An alternative scenario would involve the southern group resulting from a binary mixture involving a reservoir with MORB-like carbon isotopes and radiogenic helium. Perhaps this would be a more realistic description of the upper crustal reservoir?

101

a"Cvs.(R/R J c

-10

-8

Figure 2a

-6

vs. CO j / ^ e

-2

-4

a"cvs.PDB

•

Northern islands

•

Southern Islands

Figure 2b

Binary Mixing Curves

Figure 2: Our data plotted together with mixing curves between MORB and Subducted slab and MORB and Crust. Discussion in the text. The question that remains is what causes the change from predominantly upper crustal contamination in the south of the arc to the dominance of source effects for the rest of the arc. In this respect, it is noteworthy that the change between St. Lucia and Martinique coincides approximately with a ridge (on oceanic basement) dividing the southern thick stack of mainly old terrigenous sediments and the northern region of marine/pelagic sediments. The role of this stack of old terrigenous sediments in the system is presently unclear. Does it obstruct sediment subduction therefore allowing an enhancement of the upper crustal contamination signal along the southern part of the arc, or does it represent a second subducted reservoir (and possibly a source of subducted radiogenic "^He as in the Banda arc - Hilton and Craig, 1989). This huge stack of sediments is likely to cause over-pressuring and expulsion of some fluids in the forearc. Such localities are presently under consideration for their helium and carbon characteristics. CONCLUSIONS From our study it has become clear that utilisation of helium and carbon isotopes to study arc magmatic processes can be effective in discriminating between source 'sediment signature' and crustal contamination. In the Lesser Antilles, we have observed regional variations in the helium and carbon isotope systematics. The northern group appears to be characterised by mixing of MORB with a subducted slab component while the southern group has a clear signature of crustal contamination: however, in this case, the C02/^He ratios indicate a possible third component. Presently, we are carrying out studies to characterise the composition of the sediments in front of the trench as well as a more detailed survey of forearc fluids adjacent to the southern part of the arc to test this possibility. REFERENCES * Poreda, R. & Craig, H., 1989. Helium isotope ratios in circum-Pacific volcanic arcs. Nature 338, 473-478. * Hilton, D.R., Hoogewerff, J.A., van Bergen, M.J., & Hammerschmidt, K., 1992. Mapping magma sources in the east Sunda-Banda arcs, Indonesia: Constraints from helium isotopes, Geochim. Cosmochim. Acta 56. * Hilton, D.R., Hammerschmidt, K., Teufel, S., & Friedrichsen, H., 1993. Helium isotope characteristics of Andean geothermal fluids and lavas, Earth Planet. Sci. Lett. 120, 265-282. * W.M. White, W.M., & Dupre, B., 1986. Sediment subduction and magma genesis in the Lesser Antilles: Isotopic and trace element constraints, J. Geophys. Res. 91 B6, 5927-5941. * Westbrook, G.K., Mascle, A., & Biju-Duval, B., 1984. Geophysics and the structure of the Lesser Antilles forearc. In: Initial Reports of the Deep Sea Drilling Project, Vol. 78A & 78B, NSF 1984, 23-38* Moore, J.C., * Biju-Duval, B., & others, 1982. Offscraping and underthrusting of sediment at the deformation front of the Barbados Ridge: Deep Sea Drilling Project Leg 78A, Bull. Geol. Soc. Am. 93, 1065-1077. * Hilton, D.R., & Craig, H., 1989. A helium isotope transect along the Indonesian archipelago, Nature 342, 906-908.

102

THE RELATIVE IMPORTANCE OF CRUSTAL CONTAMINATION AND SUBDUCTED SEDIMENT COMPONENTS IN WEST SUNDA ARC MAGMATISM Rick Vame^, Massimo Gasparon^

^ Department of Earth Sciences, University of Queensland, Brisbane, Qld 4072 2 Department of Geology, University of Tasmania, GPO Box 252-79, Hobart, Tas 7005 New geochemical and Sr, Nd, and Pb isotopic analyses of Quaternary to Cretaceous sediments from the Northeastern Indian Ocean are used to estimate the composition of the sedimentary material subducted along the Sunda Trench. Two sediment endmember components are identified: siliceous-clastic and calcareous-organogenic. Siliceous-clastic sediments are characterised by a component isotopically and geochemically similar to the compositions of "typical" average upper crust, and Sumatran granitoids (Gasparon & Vame, 1995). Calcareousorganogenic sediments can be regarded as siliceous-clastic sediments strongly diluted by organogenic material, mainly CaCOs. Pre-Miocene sediments differ isotopically from post-Miocene sediments: as a group their 206pb/204pb values are lower, but for a given 206pb/204pb value, their 207p|3/204pb ^nd 208pb/204p5 values are higher. Geophysical evidence suggests that post-Miocene sediments are largely accreted rather than subducted. Low abundances of ^^Be also imply that post-Miocene sediments are not being recycled into present-day volcanism. In addition, the post-Miocene siliceous-clastic sediments sampled in the vicinity of the Sunda arc are mostly derived from the arc itself and have a large volcanogenic component. These observations should be taken into account when sediment data are used to evaluate the extent of sediment recycling into the source of Indonesian arc volcanics. At individual volcanic centre scale, Sr, Nd and Pb isotopic ratios, LILE and LREE concentrations and LILE/LILE and LILE/REE values for west Sunda volcanics cannot easily be used to distinguish components which might have been contributed to ascending magmas by assimilation of arc crust from those due to source contamination during subduction. However, crustal contamination unrelated to subduction is clearly displayed in He^/He^ values in mafic phenocrysts of Sunda volcanics (Gasparon et al., 1994), and in positive B/Be-Si02 and B/Ber correlations in calcalkaline Sunda arc volcanics. When data are viewed on the regional scale, the spatial distribution of along-arc variations in Sr, Nd, and Pb isotope values in arc volcanics, from north Sumatra to Lombok, seems better related to along-arc variations in crustal thickness and composition, than to along-arc variations in chemical fluxes into the subduction zone. Highest sediment-derived fluxes occur in the eastern section of the West Sunda arc, under Bali and Lombok, where the arc volcanics show least geochemical and isotopic evidence of the involvement of "average upper crust" and where the arc crust is thin and young. In Sumatra, where volcanics show the greatest effects of "average upper crust" involvement, the crust is thicker and older, and chemical flux into the subduction zone seems relatively small. We conclude that simple calculations of geochemical fluxes into the Sunda subduction zone involving materials derived from Indian Ocean sediments provide a poor match for the range and spatial distribution of key geochemical and isotopic characteristics, such as LILE abundances, and He, Be, Sr, Nd, and Pb isotopic values, in the west Sunda volcanic rocks. We need to take crustal contamination properly into account when "inputs" are being related to "outputs" in subduction-related magmatism, not just in the Sunda arc but probably in all arcs. REFERENCES Gasparon, M., Hilton, D.R., & Vame, R., 1994. Crustal contamination processes traced by helium isotopes: Examples from the Sunda arc, Indonesia. Earth and Planetary Science Letters, 126,15-22. Gasparon, M., & Vame, R., 1995. Sumatran granitoids and their relationship to Southeast Asian terranes. Tectonophysics, 251, 277-299.

103

CHEMICAL DYNAMICS IN THE NEW BRITAIN VOLCANIC ARC Jon Woodhead\ Steve Eggins~ & Wally Johnson^ ^School of Earth Sciences, University of Melbourne, Parkville, VIC 3052 ^Dept. of Geology, Australian National University, Canberra, ACT 0200 ^Australian Geological Survey Organisation, Box 378, Canberra, ACT 2601 INTRODUCTION The New Britain arc constitutes a unique location in the study of subduction zone magmatism for at least two reasons: (i) in the central region of the arc, volcanism takes place over a remarkably wide depth range to the WadatiBenioff zone—from about 100 km at the 'volcanic front' down to almost 600 km in the north beneath the Witu islands. This situation is, as far as we know, unique and while the reasons for its occurrence are not well understood, isotope and elemental variations are systematic across the transect. We believe that these variations provide unparalleled constraints on the geochemical architecture of a subduction zone system. (ii) tholeiites from the New Britain volcanic front represent some of the most (if not the most) HFSE 'depleted' arc rocks yet observed and thus should provide valuable constraints on the origin of source depletion in arc lavas. Furthermore, we might expect them to record most faithfully the composition of any slab-derived 'components'. THE MANTLE SOURCE Numerous studies have suggested a major role for slab-derived elements in arc petrogenesis. However, it is becoming clear that a factor of equal importance is the nature of the mantle wedge prior to additions from the subducting slab. There are a number of aspects to this problem, the first of which relates to the nature of the pre-subduction mantle wedge. Hergt and Hawkesworth (1994) detailed compelling evidence for two different mantle components reflected in the compositions of volcanic rocks from the Lau Basin spreading centre, one with affinity to Indian Ocean and another to Pacific MORB type asthenosphere. This is an important theme (also repeated on the S.E. Indian Ridge—the Australian-Antarctic discordance—and in many of die island arcs of the SW Pacific e.g. Woodhead & Johnson, 1993), since the regional distribution of these magma types may allow reconstruction of mantle dynamics in the west Pacific over the past 40 My. As noted by Hergt and Hawkesworth (1994), who recognised Indian-type mantle in the Lau Spreading centres and in the associated Tofua arc, in the case of the Bismarck arc, we see Indian-type mantle material in both the arc (New Britain) and associated back-arc (Manus Basin). Prior-depletions: In an attempt to see-through the slab signature and investigate the nature of the presubduction mantle wedge, many researchers have turned to the high field strength elements (HFSE)—those with high ionic charge/radius ratios which are widely believed to be insoluble in aqueous fluids derived from the slab. Unfortunately the use of Nb, Ta and Hf has, until very recently, been limited by extreme analytical difficulties associated with their accurate measurement at the levels commonly encountered in subduction-related rocks, leading to some confusion over whether or not island arc lavas are indeed depleted relative to typical N-MORB. For this purpose it is clearly necessary to compare local MORB such as might be found in an associated back arc, with the arc lavas concerned. Woodhead et al (1993) attempted to use ratios of the more easily measured HFSE, Ti, Zr, and V to demonstrate that arc magma sources are indeed depleted relative to their associated back arc basins. Now, however, with the advent of highly accurate low-level analysis by ICPMS we can extend these studies to the elements Nb and Ta. New data clearly demonstrate that the New Britain arc front lavas are among the most elementally depleted arc rocks yet measured with Nb and Ta contents ranging from 140-400ppb and 10-26ppb respectively in the volcanic front. These are an order of magnitude more depleted than lavas of the Manus Basin. The critical question then arises as to the origin of such depletions. At least three hypotheses are current: a) that HFSE depletion is a result of partitioning into HFSE-bearing phases which are residual to either dehydration/melting of the slab or melting of the wedge, b) that such depletions result from mantle-melt, chromatographic-type interactions during magma ascent, and c) that they result from depletion due to prior melt extraction events. 104

Pearce and Parkinson (1993) considered the theoretical behaviour of mantle-derived trace elements during partial melting. Data for samples from the New Britain arc front and MORB-type basalts from the Manus Basin, conform to Pearce and Parkinson's (1993) Pattern Type 6 in which the very highly incompatible elements (Nb, Zr) have lower abundances than the highly incompatible elements (Ti, Y, Yb), which are in turn depleted relative to the moderately incompatible elements (Ca, Al, Ga, V, Sc). Such patterns are readily reproduced by melting of a source, itself residual from previous melt extraction episodes. Note, however, that basalts from the Manus Spreading Centre show limited depletion in the VHI elements. Irrespective of whether HFSE are transported from the subducting slab or not therefore, the observation remains that arc lavas are often more depleted in HFSE than their associated back-arc basin counterparts (not simply relative to LIL enrichments, but in absolute terms). It is difficult to imagine how this phenomenon can be related directly to fluids emanating from the slab. In studying the effect of elemental depletion in arc lavas, one feature quickly becomes apparent—a link with the occurrence and/or rate of back-arc spreading. For example, arcs such as the Antilles, Aleutian, and Sunda,which do not have associated back-arc spreading zones, never possess depleted HFSE compositions. At the other end of the spectrum. New Britain, which we have demonstrated above has probably the most depleted compositions known, is adjacent to the Manus spreading centre, the BAB with fastest known spreading rate. The only one of the above hypotheses which can produce any obvious correlation between arc chemistry and the rate of back-arc spreading seems to be that of prior source depletion. We would argue therefore (as have previous authors) that extraction of melts at the back-arc (where present) and subsequent entrainment of this depleted mantle into the convective cycle, draws already depleted material into the zone of arc magmagenesis. Residual phases in the slab may certainly help to limit subsequent re-enrichment of HFSE, but it is likely that they are simply acting on mantle which has already been 'processed' in the back-arc. Palaeo-enrichments: The Melanesian arcs and marginal basins represent an area of complex interaction between the major Indo-Australian and Pacific lithospheric plates and several microplates. However, this complexity does provide some insight into processes which could have occured in other, apparently more simple, subduction systems. In particular, the whole area has experienced at least one prior episode of subduction-related volcanism since initial southwesterly subduction of the Pacific plate beneath the region began in EoceneOligocene times. This period of subduction was terminated abruptly in the lower Miocene, apparently coincident with the arrival of the Ontong-Java Plateau—a vast area of anomalously thick oceanic crust (Coleman and Kroenke, 1981)—at the Vitiaz-West Melanesian trench. These rather unusual circumstances provide an opportunity to observe phenomena which may well affect the geochemistry of large domains of the sub-oceanic mantle but, under normal circumstances, cannot be directly linked to the subduction processes. It is apparent from studies of arc and back-arc volcanic rocks throughout this region that many geochemical signatures observed at the present day are inherited from this early subduction event i.e. they are fossil enrichments. For example, volcanic rocks from the New Georgia Group of the Solomon Islands are, in terms of trace element and isotope geochemistry, typical of subduction-related samples. However this signature does not appear to have any chemical relationship with the downgoing Woodlark Basin crust—the sediment cover on this young ridge is negligible and, furthermore, at least two volcanoes (Simbo Island and Kana Keoki seamount) occur on the southwest side of the subduction zone i.e., on the subducting oceanic plate. Furthermore, subduction-related signatures have been noted in the basalts of the Woodlark Spreading centre itself (Johnson et al., 1987; Perfit et al., 1987), adjacent to the New Georgia Group, with a marked zonation away from the old Vitiaz trench. One implication of these observations is that subductioninfluenced isotope and trace element signatures may be stored in the sub-oceanic mantle for long periods (8 Ma in this case) and subsequently tapped. The question then arises—could such palaeo-enrichments contribute to the geochemical fingerprints of the New Britain volcanic rocks? A number of observations suggest that this may be so. (i) A particularly strange feature of the isotopic composition (in particular Pb) of New Britain lavas is that they indicate little or no contribution from the present day compositions of both sediment and basalt on the downgoing Solomon Sea plate—^rather a more appropriate mixing end member has an isotopic composition similar to that of volcanics from Rabaul caldera, at the eastern end of the New Britain archipelago. Moreover, there is a general trend of increasingly radiogenic isotopes in passing eastwards from the islands of the western Bismarck arc through mainland New Britain to Rabaul i.e., perpendicular to the old Vitiaz-West Melanesian trench (Figure. 1). Could this reflect a similar process to the trends observed in the Woodlark Basin/New Georgia region i.e., an isotopic zonation related to the pre-Miocene subduction episode? Considering that there is little evidence of a correlation between present day slab sources and the composition of the adjacent volcanics, we believe that this is the most likely explanation and suggest that a Pb isotope study of the Woodlark Basin glasses might provide additional support for this hypothesis.

105

Palaeo-subduction signatures J

18.85

18.80-

Rabaul

-Q CL rr o

JD Q.

18.75-

18.70-

•

New Britain volcanic front

•

Eastern Bismarck arc

18.65600

550

500

450

400

350

300

"T 250

200

Distance in km from old Vitiaz West Melanesian Trench

Figure. 1. Pb isotopes vs. distance form the old Vitiaz-West Melanesian trench. (ii) Sinton et al. (in prep.) recognise at least three different magma compositions within the Manus Basin, termed 'arc', 'BABB' and 'MORB'types. Although the reasons for the occurrence of arc-like lavas v^ithin the marginal basin are undoubtedly complex, we note that lavas of arc-like composition (in terms of trace elements) are predominantly sited in the eastern rifts part of the Basin (i.e., nearest the palaeo-trench) and that a similar, but more extreme, gradation of isotopic compositions is seen in the Manus lavas as in New Britain with, once more, the same mixing end member. If these tentative suggestions prove viable, the implications are clear—that geochemical signatures produced as a result of subduction zone magmatism, may persist in the mantle for large periods of time, these 'enriched' mantle sources then being remobilised in later magmatic episodes. SLAB

INFLUENCES

The most remarkable feature of the New Britain arc lies in its almost complete record of across-arc volcanism, providing considerable insight into the nature of the subduction process. Woodhead and Johnson (1993) considered across-arc variations in chemistry using available data. These observations can now be extended using a new and complete trace element database obtained by ICPMS techniques. The data record the gradual decline in the influence of slab-derived fluids in passing from the volcanic front into the back-arc region, a fluid which appears to be largely carrying a basaltic slab and not a sedimentary signature. Note however that these slab influences appear to be superimposed upon the effects noted above. There are, however, two issues which remain largely unresolved. The first is that estimates of the size of the subduction component derived from trace elements are always larger than those permissible from isotopic observations (e.g., Hawkesworth et al., 1993). The second is that, despite often quite large variations in the isotopic and trace element composition of the sediments on the subducting slab (e.g., along the length of an arc system), the composition of the slab-derived component, as interpreted from the geochemistry of arc volcanics, remains remarkably constant.

106

In the traditional reasoning, the composition of the slab-derived component is controlled largely by the bulk composition of the slab. This is certainly clear when considering the 'global picture' in terms of both isotopes (e.g. Woodhead, 1989) and also some trace elements (Planck & Langmuir, 1993). However, it is possible that, on the scale of an individual arc segment, the trace element composition of the slab derived component is controlled largely by the nature of the fluid itself (pH, oxygen fugacity, etc.) i.e., elemental partitioning into the fluid is buffered, despite isotopically quite variable slab inputs—in other words the fluid will have a finite solute capacity irrespective of the composition of the slab. This may be an important factor in accounting for the disparity between isotopic and trace element methods of determining slab contributions, as noted by Hawkesworth et al. (1993). REFERENCES Coleman, P.J., and Kroenke, L.W., 1981. Subduction without volcanism in the Solomons island arc. GeoMarine Letters, 1, 129-134. Hawkesworth, C.J., Gallagher, K., Hergt, J.M., and McDermott, F., 1993. Mantle and slab contributions in arc magmas. Annual Reviews in the Earth and Planetary Sciences. 21, 175-204. Hergt, J.M., and Hawkesworth, C.J., 1994. Pb, Sr, and Nd isotopic evolution of the Lau Basin: imphcations for mantle dynamics during back-arc opening. In: Proceedings of the Ocean Drilling Program, Scientific Results, 135, J.W. Hawkins, L.M. Parson, J.F. Allan, et al., Ocean Drilling Program, College Station, TX, 505-518. Johnson, R.W., Jaques, A.L., Langmuir, C.H., Perfit, M.R., Staudigel, H., Dunkley, P.N., Chappell, B.W., Taylor, S.R., and Baekisapa, M., 1987. Ridge subduction and forearc volcanism: petrology and geochemistry of rocks dredged from the western Solomon arc and Woodlark basin. In Marine geology, geophysics, and geochemistry of the Woodleark Basin-Solomon islands (eds B. Taylor & N.F. Exon) Vol.7 pp. 155-226, CPCEMR Earth Science Series. Pearce, J.A.and Parkinson, I.J., 1993. Trace element models for mantle melting: application to volcanic arc petrogenesis. In Magmatic Processes and Plate Tectonics (eds H.M. Prichard et al.) Geological Society of London Special Publication 76, 373-403. Perfit, M.R., Langmuir, C.L., Baekisapa, M., Chappell, B., Johnson, R.W., Staudigel, H., and Taylor, S.R., 1987. Geochemistry and petrology of volcanic rocks from the Woodlark Basin: addressing questions of ridge subduction. In Marine geology, geophysics, and geochemistry of the Woodleark Basin-Solomon islands (eds B. Taylor & N.F. Exon) Vol.7 pp. 113-154, CPCEMR Earth Science Series. Planck, T, and Langmuir, H., 1993. An evaluation of the global variations in the major element geochemistry of arc basalts. Earth and Planetary Science Letters, 90, 349-370. Woodhead, J.D., 1989. Geochemistry of the Mariana arc: source composition and processes. Chemical 76, 1-124.

Geology,

Woodhead, J.D., and Johnson, R.W., 1993. Isotopic and trace element profiles across the New Britain island arc, Papua New Guinea. Contributions to Mineralogy and Petrology, 113, 479-491. Woodhead, J.D., Eggins, S., and Gamble, J., 1993. High field strength and transition element systematics in island arc and back-arc basin basalts: evidence for multi-phase melt extraction and a depleted mantle wedge. Earth and Planetary Science Letters, 114, 491-504.

107

TOWARDS DECODING SOURCE AND PROCESS VARIATIONS AT ARC VOLCANOES: TRACE ELEMENT AND ISOTOPE ANALYSES FROM RAOUL (TONGA-KERMADEC ARC) AND RUAPEHU (TAUPO VOLCANIC ZONE) Tim L Wo^thiDgton^ Richard C. Price^ Ian E.M. Smith^ ^ Dept of Geology, The University of Auckland, Auckland, New Zealand ^ Research School of Earth Sciences, La Trobe University, Melbourne, Australia

Island arcs that are far removed from the influence of continental lithosphere represent the simplest locations at which to investigate subduction-related magma genesis. The Tonga-Kermadec arc is an example of such an arc, and can be subdivided into the Tofua, the northern Kermadec, and the southern Kermadec segments; further south, the arc impinges upon the continental lithosphere of New Zealand and continues as the Taupo Volcanic Zone (TVZ). Plate reconstructions suggest that Louisville Ridge-derived material (GIB) is present beneath the Tofua segment, and an additional complication in this segment may be the progressive replacement of Pacifictype asthenosphere by Indian-type asthenosphere since 5 Ma. Similarly, the southern Kermadec segment may be influenced by the subduction of continent-derived sediment, and this segment is also significantly younger than the rest of the Tonga-Kermadec arc. In contrast, the northern Kermadec segment is free from obvious complexities. In this contribution, we present new high quality trace element and Sr-Nd-Pb isotope analyses from Raoul Island, which is located in the centre of the northern Kermadec segment. The 21 analysed lavas were carefully chosen to be representative of more than 200 lavas analysed by routine techniques. Later, we contrast these analyses with new high quality analyses from Ruapehu, at the southern end of the TVZ, to investigate the effects of continent-derived sediment subduction and continental lithosphere upon arc magmas.

RAOUL ISLAND Raoul is the only northern Kermadec volcano with substantial subaerial exposure (subaerial volume 10 km^ volume of edifice 300 km^). Sampled Raoul lavas are predominantly porphyritic plagioclase » augite > olivine « orthopyroxene basalts and basaltic andesites, interbedded with a subordinate suite of aphyric basaltic andesites and andesites. However, most eruptions since 3.7 ka have been of aphyric dacitic pyroclastic rocks, and these bear gabbroic cumulates that are inferred to represent the solidification zone of the magma body. All Raoul lavas can be classified as low-K, and the lava suites follow tholeiitic fractionation trends. Chondrite normalised REE patterns are LREE-depleted, with (LsJYh)^ « 0.7; the REE concentrations mcrease with indices of fractionation (e.g., Si02, Zr), and a negative Eu anomaly develops, but (LaAT))N does not change. N-MORB normalised patterns display the classic features of arc magmatism (LDLE enrichment, flat HFSE), together with a negative Nb anomaly. Low and variable values of NbAT) (0.2-0.6) are interpreted to reflect dynamic melting of a mantle wedge component more depleted than the N-MORB source, but the unusually large range (for a single volcano) may also be attributed in part to our more complete sampling of Raoul than has been the case at other island arc volcanoes. When compared to the N-MORB array, ratios among the incompatible elements indicate that the slab-related flux of Nb, Zr and Yb is negligible, whereas more than 80% of the Sr, Th, U, K, Rb, Pb, Ba, and Cs is slab-related. The slab-related proportion of most incompatible elements in Raoul lavas is similar to that described for lavas from the South Sandwich arc; however, the proportion of slab-related Ba is higher for Raoul, whereas for Rb and Th this proportion is less. Significantly, pelagic sediments from the Pacific Plate are Ba-rich relative to those from the South Atlantic. Raoul lavas exhibit variable radiogenic isotope compositions, bounded by ^Sr/^Sr = 0.7034-0.7036, ^ ^ W ^ N d = 0.51303-0.51307, ^^Pb/'^Pb = 18.64-18.69, ' ^ W ^ P b = 15.55-15.58, and ^''Pb/'^Pb = 38.2738.39. These ranges are greater than the respective analytical precision of the measurements, for example the variation in ^'^^Nd/^'^Nd is equivalent to 9a, but there is no consistent correlation between isotopic composition and either stratigraphic position or fractionation indices. Isotopically, Raoul is similar to 'Ata (southernmost Tofiia segment) but distina from the other Tofua segment volcanoes and those of the southern Kermadec segment; in our opinion, this reflects the simpler environment of Raoul outlined in the introduction.

108

RUAPEHU In contrast to the simple setting of Raoul, Ruapehu is situated to the west of the sediment-flooded Ifikurangi Trough and is built upon 15 km thick attenuated continental crust at the southern end of the TVZ. Ruapehu lavas are porphyritic plagioclase-rich two-pyroxene bearing basalts, andesites and dacites diat can be classified as medium-K and follow the calc-alkaline fractionaticxi trend. Chondrite normalised REE patterns are LREEenriched and (La/Yb)N increases from 1.9 in the basalts to 6.5 in the dacites. Large variations in many incompatible element ratios (e.g., Rb/Zr) can be correlated withfractionationindices, and there is a progressive increase in the concentration of most incompatible elements at a given Si02 content with time. Our new radiogenic isotope compositions span the basalt to dadte spectrum and are bounded by ^Sr/^^Sr = 0.70470.7057, '^'Nd/'^^m = 0.51291-0.51270, '^Pb/'^Pb = 18.80-18.85, ' " W ^ P b = 15.61-15.66, and ' ' W ^ P b = 38.64-38.81; in general, ^Sr/^^Sr and the Pb ratios increase with indices of fractionation, while ^^^Nd/^^Nd decreases. These geochemical trends are interpreted to reflect the operation of an open-system magma reservoir in which crustal assimilation is coupled with fractional crystallisation. A comparison of the incompatible element ratios of Ruapehu lavas with those from Raoul and the N-MORB array is revealing. Ruapehu basalts have HFSE ratios that fall within the N-MORB array and, as at Raoul, suggest that a negligible proportion of their Nb, Zr and Yb is slab-related; in addition, their mantle wedge component is required to be enriched relative to the N-MORB source (Nb/Yb « 1.4). However, the basaltic andesite to dacite lavas from Ruapehu form an assimilation trend that veers away from the N-MORB array, consistent with the suggestion fliat only these more fractionated Ruapehu lavas have assimilated significant quantities of crustal material. Considering now only those Ruapehu basalts that lack evidence of crustal assimilation, our results indicate that the slab-related proportion of Nd, Ce, La, Sr, Th, U, K, Rb, Pb, Ba, and Cs in these lavas is indistinguishablefromthat in Raoul lavas. THE DSFLUENCE OF SUBDUCTED SEDIMENT Our comparison of Raoul and Ruapehu lavas leaves us with a paradox. Small amounts of pelagic and arcderived sediment are subducted at the northern Kerniadec Trench, whereas large quantities of continent-derived sediment are subducted at the Hikurangi Trough. Despite this, each of the LILE and HFSE appears to be derived from slab-related and mantle wedge sources in the same proportions for Raoul basalt as iey are for Ruapehu basalt. On the other hand, the more evolved isotope composition of Ruapehu basalt is consistent with the change in subducted sediment. These contradicticms are highlighted by the small but noticeable difference in the proportion of slab-related Ba between Raoul and the South Sandwich lavas. One resolution to this paradox could be that flushing of the LILE from the subducted sediment occurs during dewatering at shallow depth, such that a large proportion of each LILE in this sediment is transferred to and incarcerated within lithospheric mantle under the forearc. In this model, the contribution of each LILEfromthe slab to asthenospheric mantle at greater depths, and thereby to arc magma, would be dominated by the LILE stored within altered oceanic crust, and this might reasonably be expeaed to show greater consistency along and between subduction zones than would sediment-derived LILE. However, because of its more extreme isotopic composition, any small accompanying residual flux of the LILEfromthe subducted sediment would be able to greatly influence the isotopic composition of arc magmas. SUMMARY Incompatible element variations in basalts from Raoul and Ruapehu reflect a change from a depleted mande wedge component at Raoul to an enriched wedge component at Ruapehu. The influence of sediment subduction is seen primarily in the more evolved isotope composition of Ruapehu basalt, and little effect is seen on the proportion of each LILE and HFSE that is slab-related as opposed to wedge-derived. Earlyflushingof the LILEfromsubducted sediments may explain the limited influence of the sediments. 109

GEOCHEMISTRY AND PETROGENESIS OF BASALTIC ROCKS FROM NORTH QUEENSLAND: HAS SUBDUCTION-MODIFIED MANTLE PLAYED A ROLE? Ming Zhang\ Jon Stephenson^ Suzanne Y 0'Reilly\ Marc Norman^ Malcolm T McCulloch^ 1, GEMOC, Macquarie University, Sydney, NSW 2109, Australia 2, Dept of Earth Sciences, James Cook University, Townsville, Qld 4811, Australia 3, Research School of Earth Sciences, Australian National University, Canberra, ACT 0200, Australia

INTRODUCTION Cenozoic basaltic volcanism is widespread in North Queensland (NQld) as a part of the eastern Australian volcanic zone. Basaltic rocks from 7 volcanic provinces, including Atherton, McBride, Chudleigh, Nulla and Mingela at Townsville-Caims area and McLean and Piebald near Cooktown, were chosen for this study. The basalts are located cross the major tectonic boundaries between the Mesoproterozoic Georgetown Inlier and the Phanerozoic Hodgkinson and Thomson fold belts. The McBride Province is located mostly inside the Georgetown Inlier, whereas the Mingela, Atherton, and Cooktown Provinces are in the Phanerozoic fold belts. The Chudleigh and Nulla volcanoes straddle the tectonic boundary. Most of the basalts occurring as volcanic cones and long lava flows were erupted during a period of 8.0-0.01 Ma, with an exception of the early Tertiary (44-31 Ma) plugs and dykes at Mingela (Stephenson, 1989). All these basaltic provinces are recognised as lava field provinces. Their age distribution does not show any relationship with the northward motion of the IndoAustralian plate over a mantle plume that resulted in the central volcano volcanism in eastern Australia during at least the last 35 Ma. In the following text, the term "Cairns" is used for Atherton, McBride, Chudleigh and Nulla Provinces, and "Cooktown" for McLean and Piebald. PETROLOGY AND GEOCHEMISTRY Sixty-one basalt samples from North Queensland were analysed for major (by XRF) and 40 trace elements (by XRF and ICPMS); up to 41 of them were also analysed for Sr-Nd-Pb isotopes (Pb isotope data not available for Chudleigh and Mingela Provinces). Based on a CIPW-norm classification (O'Reilly & Zhang, 1995), 57 samples are alkaline basalts, ranging from nephelinite (5%) and basanite (16%) to alkali olivine basalt (5%) and hawaiite (67%), and the rest are olivine tholeiite (7%). Mg numbers (Mg#=Mg/Mg+Fe"'') range 0.70-0.49, with an average of 0.64 (MgO=12.8-4.3 wt%), and average Ni content is 185 ppm (±60 ppm). The majority of these basaltic rocks (including some hawaiites) contain abundant mantle xenoliths and high-pressure megacrysts, indicating their primitive or near-primitive nature. Major and trace elements: In general, basalts from Cairns area and Mingela are higher in SiOj, AI2O3, but lower in Ti02, Sc, Y, Nb and La at a given Mg# than the lava field basalts from New South Wales (NSW) and the Victorian Newer Basalts in eastern Australia (Table 1). They also have CaO/AlsOs (0.56±0.08) lower than the NSW basaltic rocks (0.69±0.12), but similar to the Newer Basalts (0.57±0.11). In contrast, the Cooktown basalts are similar to many strongly alkaline basalts in NSW in their Si02, AI2O3 and Ti02 contents, but have higher Ca0/Al203 (0.82±0.20) than the later. Y abundances of the Cooktown samples are within the lower range for the NSW and Victorian basalts, but marginally higher than the other NQld basalts, whereas Nb, Th and LREE abundances are among the highest of all the basaltic rocks discussed here. No well-defined correlations between Mg# and minor and most incompatible trace elements are present for the NQld basalts. Incompatible element patterns: The primitive NQld basalts exhibit three different types of incompatible element patterns (Fig. 1). The majority of basalts from Cairns area are moderately silica-undersaturated (eg alkali olivine basalts and hawaiites). They display identical incompatible element patterns with slight to moderate enrichments in K, Sr and P over Nb, Ta, La, Zr and Hf, reflecting a lesser degree of enrichments in both HFSE and LREE. On the other hand, strongly alkaline basalts (eg /c-normative nephelinites), the predominant rock type in Cooktown area and the subordinate in Cairns, show fractionated incompatible element patterns similar to many NSW nephelinites, basanites, and ne-hawaiites (eg Barrington, Kandos, and some from Southern Highlands, Monaro and Dubbo; O'Reilly & Zhang, 1995), with strong enrichments in Nb, Ta and LREE and relative depletions to variable degree in Rb, K, Zr, Hf, and Ti. One moderately evolved hawaiite and one ol-tholeiite from Cooktown are similar to the majority of the Cairns basalts. The relative K-depletion in the Cooktown basalts becomes more significant (ie decreasing K/Nb) with increasing Si02-undersaturation. The incompatible element patterns of the early Tertiary Mingela basalts are similar to the typical OIB, the Newer Basalts, and some NSW lava field basalts (eg Blue Mts, Oberon, Grabben Gullen, and some Dubbo tholeiites). They exhibit a peak at Nb and Ta and a gradual decrease in mantle-normalised abundances from K to Yb, except for the presence of a slight Sr-enrichment, which is almost ubiquitous in the NQld basalts. The differences in the incompatible element patterns reflect the variations in some incompatible element ratios (eg K/Nb=340±80 for Cairns vs 1 110

180±80 for NSW; Table 1) and can be further illustrated using element-ratio plots such as Rb/Sr vs K/Nb (Fig. 2) and Sr/La vs K/Nb. For example, three distinct areas can be recognised on the Figure 2. The Cooktown strongly alkaline rocks fall in the area for the Barrington, Kandos and Dubbo alkaline basalts, forming a trend pointing to the region defined by amphibole- and apatite-bearing spinel peridotite xenoliths hosted by the Victorian Newer Basalts. The Mingela samples plot close to the Newer Basalts, Dubbo tholeiites and average OIB composition. The Cairns basalts mostly plot outside the region for the other AustraHan basalts and extend toward the N-MORB and the Kermadec-Tonga island arc basalts (Ewart & Hawkesworth, 1987) due to their high K/Nb and low Rb/Sr ratios. Table 1 Selected element abundances and element ratios for the Cenozoic basaltic rocks from lava-field provinces in eastern Australia Caims Mingela Cooktown NSW Vic Kermadecf PM* OB* MORB* Sample No. <42 <11 <8 <162 <23 <17 SiOj 48.2±1.7 48.510.9 45.314.6 45.611.8 47.611.7 50.911.7 44.8 15.3+0.8 16.110.4 13.0+1.8 14.311.3 14.211.4 16.5+1.6 4.45 AlA TiOj 1.94+0.21 1.9410.36 2.3710.45 2.48+0.43 2.5210.37 0.70+0.22 0.21 2.9 1.3 Sc 16.9±2.7 16.813.5 15.112.6 23.1+5.6 2012 4014 17.1 Y 27+3 19±3 2111.5 2714 2413 1616 4.55 29 28 Nb 46±19 54120 90155 78130 64117 0.6010.11 0.713 48.0 2.3 La 26.8111.5 35.2112.8 51.9126.0 41.2114.7 46112 3.912.3 0.687 37.0 2.50 K/Nb 339177 268134 1901124 176177 283152 384011440 350 250 258 Zr/Nb 4.110.6 4.310.5 3.011.2 3.311.5 5.110.5 52124 15 5.8 32 Sr/La 29.514.2 28.814.7 21.815.0 23.115.1 18.913.0 85179 31 18 36 (LaAnD)„§ 13.115.3 14.014.2 24.8111.7 16.3+5.7 18.2+5.9 1.6+1.1 1.39 12.3 0.59 $, chondrite-normalised ratios; % basaltic rocks from Kermadec-Tonga arc (Ewart & Hawkesworth, 1987); *, estimated compositions of primitive mantle, ave. OIB, and ave. normal MORB (Sun & McDonough, 1989)

0 100 c <n E

.1 'Z 10 o. o o cc

Atherton (LM192) Chudleigh (CHD9) McLean (CK40) Mingela (86) Rb Th Nb K Ce P Zr Sm Ti Y Ba U Ta La Sr Nd Hf Eu Tb Yb

Figure 1 Incompatible element patterns for representative basaltic rocks in North Queensland. Atherton and Chudleigh from Cairns area and McLean from Cooktown area Sr-Nd-Pb isotopic systematics: The NQld basaltic rocks have ^"^Sr/^^Sr of 0.7034-0.7048 and ^"^^Nd/^^Nd of 0.51302-0.51279 (eNd=+7.5—f-3.0) The Sr and Nd isotopic ratios correlate with neither their parent/daughter element ratios nor any fractionation or crustal contamination indicators such as Mg# and Si02. Although the Sr and Nd isotopic ratios of the Cairns and Cooktown basalts are within the ranges of the NSW lava field basalts and the Victorian Newer Basalts, they differ in their high ^"^Sr/^^Sr at a given ^'^^Nd/^'^Nd, thus forming a high ^"^Sr/^^Sr trend above the NSW and Victorian trend (Fig. 3). Coincident with this trend, spinel peridotite xenoliths found from the Cairns area are characterised by even higher ^"^Sr/^^Sr at a given ^'^^Nd/^'^Nd than their hosts (O'Reilly & Zhang, 1995). The most depleted NQld basalts are similar to relatively enriched Indian MORB, whereas the most enriched trend towards the EM2 component. In contrast to the high ^"^Sr/^^Sr trend of the younger NQld basalts, the early Tertiary Mingela basalts plot with in the compositional range for the NSW and Victorian basalts.

2 111

Pbisotopic ratios of the NQld samples range in of 17.90-18.66, in ^ " W ^ P b of 15.56-15.63, and in ^''^Pb/^'^Pb of 37.72-39.24, displaying Dupal signatures with A7/4Pb = +3.3-+12.5 and A8/4Pb = +32+106. They differ from the NSW lava field basalts in their generally low but variable ^'^Pb/^'^Pb and high A7/4Pb (18.60-19.14 and -4.4-+6.7, respectively, for NSW basalts). Therefore, on a vs plot (Fig. 4), the NQld basalts form a subparallel trend below the NSW basalts, indicating the presence of a source component with relatively low Th/U. The Dupal-type Pb isotopic signature and the correlations between ^'Sr/^'Sr and ^'^Pb/^'^Pb and between "''Nd/"^Nd and ^'^Pb/^'^Pb also require contributions from mantle reservoirs of both an Indian-MORB type and an EM2 type. 500 Kermadecf arc basalts 400 -

• •

300 ^

•

••

^ ^ i J .sediments / .

•.<> •

NQld

Jl O Vf AAj* * ^ ^vX ^

jl^

A X

• • A

A <

NSW

•

• +

• 200

o /

Victorian Newer basalts

100

+o+

°° Y / J amp-c& apperidotito

+ +

• #

Barrington & Kandos Dubbo other NSW Ave. OIB MORE source

0.10

0.05

0.01

0.005

Cairns Cooktown Mingela

Rb/Sr Figure 2 Rb/Sr vs K/Nb for the lava field basalts in eastern Australia, amp- & ap-peridotites, amphibole- and apatite-bearing spinel peridotite xenoliths in the Victorian Newer Basalts (O'Reilly & griffin, 1988); Newer Basalts, after McDonough et al. (1985); Kermadec basalts, after Ewart & Hawkesworth (1987) ' Indian

• • • T A + X

0.5130 I S S S L ^ - ^ - C ^ - M O R B • 5

0.5128

•

/ Pacific MORB

^ +

^ T-

•

0.5126

Atherton McBride Nulla/ChD" Cooktown ' Mingela NSW Victoria

+ + +

EMiV 0.5124 0.7025

0.7035

0.7045 87sr/86sr

0.7055

Figure 3 ^^Sr/^^Sr vs ^'^^Nd/^^Nd for the lava field basalts in eastern Australia. Pacific and Indian MORB after Mahoney et al. (1995). CHARACTERISTICS OF THE MANTLE SOURCES Three distinct mantle sources can be inferred from the Sr-Nd-Pb isotope data. One is represented by the early Tertiary Mingela basalts. It has both isotope and element signatures likely derived from an OB-type sublithospheric mantle source, possibly resulting from interaction of an enriched deep mantle with the overlying depleted asthenosphere of the Pacific-MORB type. A similar source has been considered as one of the dominant

3 112

mantle reservoirs for some basalts in NSW and Victoria (O'Reilly & Zhang, 1995). 39.5

39.0 -

38.5 -

38.0 -

37.5 15.4

15.5

15.6

207 pj^/204 p j ^

15.7

Figure 4 vs ^^^Pb/^^Pb for the lava field basalts in eastern Australia. The isotopic trends for the other NQld basalts can be simply interpreted as mixing of two source components. One is likely to be an asthenospheric Indian-MORB source that has been widely recognised in the late-Tertiary to Present basalts from island arc and backarc basin settings in the SW Pacific (eg the Kermadec-Tonga arc and the Lau basin, Hawkins, 1995). This Indian-MORB-type asthenosphere is considered to have at least partially replaced the previous Pacific-MORB-type asthenosphere beneath the SW Pacific margins. The other is a component with EM2 geochemical signatures, likely hosted in the subcontinental lithospheric mantle (SCLM). However, this SCLM component must be heterogeneous and is likely to have multiple origins if the differences in element abundances and ratios betv/een the Cairns and Cooktown basalts are taken into account. The Caims basalts are characterised by relatively high K/Nb, Sr/La, Sr/Nd, but low LaA^b. This, in connection with their relatively high Si02, AI2O3, but low Ti02, Y, HFSE and LREE abundances, indicates that the SCLM beneath the Caims area may have been heterogeneously modified by subduction-related processes, probably during the Paleozoic - early Mesozoic times when the Tasman Fold Belts were formed by westward subduction of the protoPacific plate beneath the eastern margin of the Australian continent. Isotopic signatures of mantle xenoliths in the same region support this conclusion. On the other hand, the incompatible element patterns of the Cooktown alkaline basalts may require contributions from a metasomatised SCLM source containing amphibole (± apatite), similar to that for the Barrington nephelinites (O'Reilly and Zhang, 1995). REFERENCES Ewart, A, & Hawkesworth, C.J., 1987. The Pleistocene-Recent Tonga-Kermadec arc lavas: Interpretation of new isotopic and rare earth data in terms of a depleted mantle source model. J. Petrol, 28, 495-530. Ewart, A, Chappell, B.W., & Menzies, MA., 1988. An overview of the geochemical and isotopic characteristics of the eastern Australian Cainozoic volcanic provinces. J. Petrol, Spec. Vol., 225-274. Hawkins, J.W., 1995. Evolution of the Lau Basin-Insight from GDP Leg 135, in Active Margins and Marginal basins of the Western pacific, Taylor, B., and Natland, J. (editors), AGU Geophysical Monograph 88, pp 125-173, McDonough, W.F., McCulloch, M.T., Sun, S.-s., 1985. Isotopic and geochemical systematics in Tertiary - Recent basalts from southeastem Australia and implications for the evolution of the sub-continental lithosphere. Geochim. Cosmochim. Acta 49, 2051-2067 Mahoney, J.J., Lones, W.B., Frey, F.A., Salters, V.J.M., Pyle, D.G., & Davies, H.L., 1995. Geochemical characteristics of lavas from Broken Ridge, the Naturaliste Plateau and southemmost Kerguelen Plateau: Cretaceous plateau volcanism in the southeast Indian Ocean. Chem. Geol, 120, 315-345. O'Reilly, S.Y., & Griffin, W.L., 1988. Mantle metasomatism beneath westem Victoria, Australia, I: Metasomatic processes in Cr-diopside Iherzolites. Geochim. Cosmochim. Acta 52, 433-447 O'Reilly, S.Y., & Zhang, M., 1995. Geochemical characteristics of lava-field basalts from eastem Australia and inferred sources: connections with the subcontinental lithospheric mantle. Contrib. Mineral Petrol, 121, 148-170. Stephenson, P.J., 1989. Northern Queensland. In Intraplate volcanism in eastem Australia and New Zealand, Johnson, R.W. (editor), Cambridge Univ. Press, 89-97. Sun, S.-S. & McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes, in Magmatism in the ocean basins, Saunders, A.D. & Norry, M.J. (editors). Spec. Pub. Geol Soc. London, 42: 313-346. Blackwell Scientific Publications.

113