Geochemistry of serpentinized peridotites from ODP Holes 207-1272A and 207-1274A

Despite the key importance of altered oceanic mantle as a repository and carrier of light elements (B, Li, and Be) to depth, its inventory of these elements has hardly been explored and quantified. In order to constrain the systematics and budget of these elements we have studied samples of highly serpentinized (>50%) spinel harzburgite drilled at the Mid-Atlantic Ridge (Fifteen-Twenty Fracture zone, ODP Leg 209, Sites 1272A and 1274A). In-situ analysis by secondary ion mass spectrometry reveals that the B, Li and Be contents of mantle minerals (olivine, orthopyroxene, and clinopyroxene) remain unchanged during serpentinization. B and Li abundances largely correspond to those of unaltered mantle minerals whereas Be is close to the detection limit. The Li contents of clinopyroxene are slightly higher (0.44-2.8 µg/g) compared to unaltered mantle clinopyroxene, and olivine and clinopyroxene show an inverse Li partitioning compared to literature data. These findings along with textural observations and major element composition obtained from microprobe analysis suggest reaction of the peridotites with a mafic silicate melt before serpentinization. Serpentine minerals are enriched in B (most values between 10 and 100 µg/g), depleted in Li (most values below 1 µg/g) compared to the primary phases, with considerable variation within and between samples. Be is at the detection limit. Analysis of whole rock samples by prompt gamma activation shows that serpentinization tends to increase B (10.4-65.0 µg/g), H2O and Cl contents and to lower Li contents (0.07-3.37 µg/g) of peridotites, implying that-contrary to alteration of oceanic crust-B is fractionated from Li and that the B and Li inventory should depend essentially on rock-water ratios. Based on our results and on literature data, we calculate the inventory of B and Li contained in the oceanic lithosphere, and its partitioning between crust and mantle as a function of plate characteristics. We model four cases, an ODP Leg 209-type lithosphere with almost no igneous crust, and a Semail-type lithosphere with a thick igneous crust, both at 1 and 75 Ma, respectively. The results show that the Li contents of the oceanic lithosphere are highly variable (17-307 kg in a column of 1 m * 1 m * thickness of the lithosphere (kg/col)). They are controlled by the primary mantle phases and by altered crust, whereas the B contents (25-904 kg/col) depend entirely on serpentinization. In all cases, large quantities of B reside in the uppermost part of the plate and could hence be easily liberated during slab dehydration. The most prominent input of Li into subduction zones is to be expected from Semail-type lithosphere because most of the Li is stored at shallow levels in the plate. Subducting an ODP Leg 209-type lithosphere would mean only very little Li contribution from the slab. Serpentinized mantle thus plays an important role in B recycling in subduction zones, but it is of lesser importance for Li.

Bulk and clay mineralogy of ODP Leg 110 holes

The mineralogy of both bulk- and clay-sized fractions of samples from Sites 671, 672, and 674 of ODP Leg 110 was determined by X-ray diffraction. The major minerals include quartz, calcite, plagioclase feldspar, and the clay minerals smectite, illite, and kaolinite. The smectite is a dioctahedral montmorillonite and is derived primarily from degradation of volcanic ash. Percentage of smectite varies with sediment age; Miocene and Eocene sediments are the most smectite-rich. High smectite content tends to correlate with elevated porosity, presumably because of the ability of smectite clays to absorb significant amounts of interlayer water. Because of a change in physical properties, the decollement zone at Site 671 formed in sediments immediately subjacent to a section of smectite-rich, high-porosity, Miocene-age sediments. Sediments above the decollement at Site 671, as well as all sediments analyzed from Sites 672 and 674, contain nearly pure smectite characteristic of the alteration of volcanic ash. Within the decollement zone and underthrust sequence, however, the smectite contains up to 65% illite interlayers. Although the illite/smectite could be interpreted as detrital clay derived from South America, its absence in the sediments stratigraphically equivalent to the decollement and underthrust sequences at Sites 672 and 674 favors the interpretation that it originated by diagenetic alteration of pre-existing smectite similar to that in the overlying sediments. A significant percentage of the freshening of the pore waters observed in these zones could be due to the water released during smectite dehydration.

Amino acid in Deep Sea Drilling Project cores

Several amino acid diagenetic reactions, which take place in the deep-sea sedimentary environment, were investigated, using various Deep Sea Drilling Project (DSDP) cores. Initially it was found that essentially all the amino acids in sediments are bound in peptide linkages; but, with increasing age, the peptide bonds undergo slow hydrolysis that results in an increasingly larger fraction of amino acids in the free state. The hydrolysis half-life in calcareous sediments was estimated to be ~1-2 million years, while in non-carbonate sediment the hydrolysis rate may be considerably slower. The amino acid compositions and the extent of racemization of several amino acids were determined in various fractions isolated from the sediments. These analyses demonstrated that the mechanism, kinetics, and rate of amino acid diagenesis are highly dependent upon the physical state (i.e., free, bound, etc.) in which the amino acids exist in the sedimentary environment. In the free state, serine and threonine were found to decompose primarily by a dehydration reaction, while in the bound state (residue or HCl-insoluble fraction) a reversible aldol-cleavage reaction is the main decomposition pathway of these amino acids. The change in amino acid composition of the residue fraction with time was suggested to be due to the hydrolysis of peptide bonds, while in foraminiferal tests the compositional changes over geological time are the result of various decomposition reactions. Reversible first-order racemization kinetics are not observed for free amino acids in sediments. The explanation for these anomalous kinetics involves a complex reaction series which includes the hydrolysis of peptide bonds and the very rapid racemization of free amino acids. The racemization rates of free amino acids in sediments were found to be many orders of magnitude faster than those predicted from elevated temperature experiments using free amino acids in aqueous solution. The racemization rate enhancement of free amino acids in sediments may be due to the catalysis of the reaction by trace metals. Reversible first-order kinetics are followed for amino acids in the residue fraction isolated from sediments; the rate of racemization in this fraction is slower than that predicted for protein-bound amino acids. Various applications of amino acid diagenetic reactions are discussed. Racemization and the decomposition reaction of serine and threonine can both be used, with certain limitations, to make rough age estimates of deep-sea sediments back to several million years. The extent of racemization in foraminiferal tests which have been dated by some other independent technique can be used to estimate geothermal gradients, and thus heat flows, and to evaluate the bottom water temperature history in certain oceanic areas.

Geochemistry of Walvis Ridge basalts

The proposed origins for the Enriched Mantle I component are many and various and some require an arbitrary addition of an exotic component, be it pure sediment or an enriched melt from the subcontinental lithosphere. With Pitcairn, Walvis Ridge is the 'type-locality' for the Enriched Mantle I (EMI) component. We analyzed basalts from DSDP Site 525A, Site 527 and Site 528 on the Walvis Ridge with the aim to constrain the history of its source. The isotopic compositions we measured for the three sites overlap with the values obtained by Richardson et al. (1982a) and extend towards less radiogenic Sr and more radiogenic Pb and Nd isotopic compositions. We used our new trace element and radiogenic isotope (Hf, Nd, Pb and Sr) characterization in combination with the literature data to produce the simplest possible model that satisfies the trace element and isotopic constraints. Although the elevated 207Pb/204Pb with respect to 206Pb/204Pb predicts an ancient origin for EMI, none of the proposed origins had modeled it as such. The data is consistent with the EMI composition being formed by the addition of a melt to a mantle with bulk Earth-like composition followed by melt extraction of a low degree melt. The timing of these two events is such that the metasomatism has to have taken place prior to 4 Ga and the subsequent melt removal before 3.5 Ga. This confirms the expectation of an ancient character for the EMI component. The Walvis Ridge data shows two distinct two component mixing trends: one formed by the less enriched Site 527 and Site 528 basalts and one formed by the Site 525A basalts. The two trends have the EMI endmember in common. The less depleted end of the Site 527-Site 528 basalts is FOZO-like and can be explained by the addition of a recycled component (basaltic oceanic crust plus sediment). This recycled component was altered during subduction. The sense and magnitude of the chemical fractionation resulting from the subduction alteration are in agreement with dehydration experiments on basalts and sediment. Compared to other EMI like basalts the Walvis Ridge basalts have flatter REE patterns and show less fractionation between large ion lithophile and heavy REE elements. Using the isotopic compositions as constrains for the parent-daughter ratios we were able to model the trace element patterns of the basalts as melting between 5 and 10% for Site 525A and between 10 and 15% for the depleted end of the Site 528-Site 527 array. In all cases a significant portion of melting takes place in the garnet stability field.

Stable chlorine isotopes in pore waters from DSDP Hole 79-546 and ODP Hole 131-808C

Stable Cl isotope ratios, measured in marine pore waters associated with the Barbados and Nankai subduction zones, extend significantly (to ~-8 per mil) the range of d37Cl values reported for natural waters. These relatively large negative values, together with geologic and chemical evidence from Barbados and Nankai and recent laboratory data showing that hydrous silicate minerals (i.e., those with structural OH sites) are enriched up to 7.5 per mil in 37Cl relative to seawater, strongly suggest that the isotopic composition of Cl in pore waters from subduction zones reflects diagenetic and metamorphic dehydration and transformation reactions. These reactions involve clays and/or other hydrous silicate phases at depth in the fluid source regions. Chlorine therefore cannot be considered geochemically conservative in these systems. The uptake of Cl by hydrous phases provides a mechanism by which Cl can be cycled into the mantle through subduction zones. Thus, stable Cl isotopes should help in determining the extent to which Cl and companion excess volatiles like H2O and CO2 cycle between the crust and mantle.

Geochemistry of subducted volcaniclastic sediments from Mariana arc magmas

The phase relations of natural volcaniclastic sediments from the west Pacific Ocean were investigated experimentally at conditions of 3-6 GPa and 800-900 °C with 10 wt.% added H2O (in addition to ~ 10 wt.% structurally-bound H2O) to induce hydrous melting. Volcaniclastic sediments are shown to produce a sub-solidus assemblage of garnet, clinopyroxene, biotite, quartz/coesite and the accessory phases rutile ± Fe-Ti oxide ± apatite ± monazite ± zircon. Hydrous melt appears at temperatures exceeding 800-850 °C, irrespective of pressure. The melt-producing reaction consumes clinopyroxene, biotite and quartz/coesite and produces orthopyroxene. These phase relations differ from those of pelagic clays and K-bearing mid ocean ridge basalts (e.g. altered oceanic crust) that contain phengite, rather than biotite, as a sub-solidus phase. Despite their relatively high melt productivity, the wet solidus for volcaniclastic sediments is found to be higher (825-850 °C) than other marine sediments (700-750 °C) at 3 GPa. This trend is reversed at high-pressure conditions (6 GPa) where the biotite melting reaction occurs at lower temperatures (800-850 °C) than the phengite melting reaction (900-1000 °C). Trace element data was obtained from the 3 GPa run products, showing that partial melts are depleted in heavy rare earth elements (REE) and high field strength elements (HFSE), due to the presence of residual garnet and rutile, and are enriched in large ion lithophile elements (LILE), except for Sr and Ba. This is in contrast to previous experimental studies on pelagic sediments at sub-arc depths, where Sr and Ba are among the most enriched trace elements in glasses. This behavior can be partly attributed to the presence of residual apatite, which also host some light REE in our supra-solidus residues. Our new experimental results account for a wide range of trace element and U-series geochemical features of the sedimentary component of the Mariana arc magmas, including imparting a substantial Nb anomaly to melts from an anomaly-free protolith.

(Table 1) Geochemical and geothermometry data from fibrous calcite vein fills in the Dabashan Fold Belt, China

Fibrous calcite veins with organic inclusions have been widely considered as indicators of oil and gas generation and migration under overpressure. Abundant fibrous calcite veins containing organic-bearing inclusions occur in faulted Lower Paleozoic through Triassic hydrocarbon source rocks in the Dabashan Foreland Belt (DBF). d13CPDB and d18OPDB values of the fibrous calcite range from - 4.8 to -1.9 to per mil and - 12.8 to - 8.4 per mil respectively, which is lighter than that of associated carbonate host rocks ranging from - 1.7 to + 3.1 per mil and - 8.7 to - 4.5 per mil. A linear relationship between d13CPDB and d18OPDB indicates that the calcite veins were precipitated from a mixture of basinal and surface fluids. The fibrous calcite contains a variety of inclusions, such as solid bitumen, methane bearing all-liquid inclusions, and vapor-liquid aqueous inclusions. Homogenization temperatures of aqueous inclusions range from 140 to 196° with an average of 179°. Salinities of aqueous inclusions average 9.7 wt% NaCl. Independent temperatures from bitumen reflectance and inclusion phase relationships of aqueous and methane inclusions were used to determine fluid pressures. Results indicate high pressures, elevated above typical lithostatic confining pressure, from 150 to 200 MPa. The elevated salinity and high temperature and pressure conditions of the fibrous calcite veins argue against an origin solely from burial overpressure resulting from clay transformation and dehydration reactions. Instead fluid inclusion P-T data and geochemistry results and regional geology indicate abnormally high pressures during fluid migration. These findings indicate that tectonic stress generated fracture and fault fluid pathways and caused migration of organic bearing fluids from the DBF during the Yanshan orogeny.

Geochemistry of serpentenites from DSDP Hole 84-566C

We studied a unique chrysotile-antigorite serpentinite, drilled on Deep Sea Drilling Project Leg 84 (Site 566) in the Guatemala forearc. Our in situ major and trace element data provide new constraints on possible reactions and associated trace element mobilisation during shallow serpentinite subduction. Chrysotile of the studied serpentinite, formed by the hydration of an upper mantle peridotite precursor, is partially replaced by antigorite (alone) which also occurs in 0.5 mm wide unoriented veins crosscutting the rock. Based on textural relationships and the P-T-X stability of the rock forming phases, the replacement of chrysotile by antigorite occurred at T < 300 °C, due to interaction between the chrysotile-serpentinite and an aqueous fluid. A comparison of the chemical compositions of reactant and product phases reveals that about 90% of the Cl, more than 80% of the B and about 50% of the Sr hosted originally by chrysotile was lost during fluid-assisted chrysotile-to-antigorite transformation and accompanying partial dehydration, and documents the much lower affinity of antigorite for trace element uptake than that of chrysotile. The fluid-assisted chrysotile-to-antigorite transformation and associated trace element loss documented here can occur in the shallow (< 30 km) region of subduction zones. This transformation decreases notably the Cl and B inventory of subducting serpentinites, which are regarded as one of the most important carriers of these elements into subduction zones. The evolution of serpentinites during initial subduction stages thus appears to be critical in the recycling of specific trace elements such as B or Cl from forearc to subarc depths.

Geochemistry of ODP Hole 135-841B

Two igneous rock units were recovered at Site 841. More than 200 m of island-arc rhyolites, rhyolitic tuffs, lapilli tuffs, and pumice breccias, divided into five units, compose the basement at the site. These rhyolitic volcanics are late middle Eocene or older and formed part of a subaerial rhyolitic volcano. These low-K rhyolites were produced by fractional crystallization of a more mafic arc-tholeiitic lava or by dehydration melting of lower crustal arc tholeiites. The Site 841 basement rocks are similar in composition to high-SiO2 lavas in the Eocene basement on 'Eua and crystallized from depleted island-arc-tholeiitic basalts like those exposed on 'Eua. No evidence is present in the rhyolites, or in the clasts enclosed within them, for boninite series magmas at Site 841. The Site 841 rhyolitic complex bears no resemblance to Cretaceous rhyolites from the Lord Howe Rise, which are enriched in K and incompatible elements. The volcanic rocks at Site 841 are part of a widely distributed Eocene volcanic episode that marked the earliest phases of subduction in the Tonga region; they are not part of an older crustal fragment. The second igneous sequence is a series of basaltic dikes and sills that intruded Miocene sediments. These basalts have trace element abundances and ratios identical to upper Miocene lavas from the Lau Ridge. The Site 841 basalts do not have any geochemical characteristics that suggest they were generated by unusual thermal conditions in the shallow sub-forearc mantle. They are most reasonably interpreted as intrusions fed by basement dikes propagated from the associated active arc. No evidence for local serpentinite exposures, like those that are common in the Mariana forearc, was found at Site 841. The results from Site 841 provide strong support for hypotheses of forearc evolution that have been advanced for the Izu-Bonin-Mariana system.

Geochemistry of pore fluids and bulk solids of ODP Site 195-1200, South Chamorro Seamount

At the South Chamorro Seamount in the Mariana subduction zone, geochemical data of pore fluids recovered from Ocean Drilling Program Leg 195 Site 1200 indicate that these fluids evolved from dehydration of the underthrusting Pacific plate and upwelling of fluids to the surface through serpentinite mud volcanoes as cold springs at their summits. Physical conditions of the fluid source at 27 km were inferred to be at 100°-250°C and 0.8 GPa. The upwelling of fluid is more active near the spring in Holes 1200E and 1200A and becomes less so with increasing distance toward Hole 1200D. These pore fluids are depleted in Cl and Br, enriched in F (except in Hole 1200D) and B (up to 3500 µM), have low 11B (16-21), and have lower than seawater Br/Cl ratios. The mixing ratios between seawater and pore fluids is calculated to be ~2:1 at shallow depth. The F, Cl, and Br concentrations, together with B concentrations and B isotope ratios in the serpentinized igneous rocks and serpentine muds that include ultramafic clasts from Holes 1200A, 1200B, 1200D, 1200E, and 1200F, support the conclusion that the fluids involved in serpentinization originated from great depths; the dehydration of sediments and altered basalt at the top of the subducting Pacific plate released Cl, H2O, and B with enriched 10B. Calculation from B concentrations and upwelling rates indicate that B is efficiently recycled through this nonaccretionary subduction zone, as through others, and may contribute the critical missing B of the oceanic cycle.

Geochemistry of oceanic crust beneath Costa Rica

Resolving flow geometry in the mantle wedge is central to understanding the thermal and chemical structure of subduction zones, subducting plate dehydration, and melting that leads to arc volcanism, which can threaten large populations and alter climate through gas and particle emission. Here we show that isotope geochemistry and seismic velocity anisotropy provide strong evidence for trench-parallel flow in the mantle wedge beneath Costa Rica and Nicaragua. This finding contradicts classical models, which predict trench-normal flow owing to the overlying wedge mantle being dragged downwards by the subducting plate. The isotopic signature of central Costa Rican volcanic rocks is not consistent with its derivation from the mantle wedge (Feigenson et al., 2004, doi:10.1029/2003GC000621; Herrstom et al., 1995, doi:10.1130/0091-7613(1995)023<0617:VILCAW>2.3.CO;2; Abratis and Woerner, 2001) or eroded fore-arc complexes (Goss and Kay, 2006, doi:10.1029/2005GC001163) but instead from seamounts of the Galapagos hotspot track on the subducting Cocos plate. This isotopic signature decreases continuously from central Costa Rica to northwestern Nicaragua. As the age of the isotopic signature beneath Costa Rica can be constrained and its transport distance is known, minimum northwestward flow rates can be estimated (~63-190 mm/yr) and are comparable to the magnitude of subducting Cocos plate motion (approx85 mm/yr). Trench-parallel flow needs to be taken into account in models evaluating thermal and chemical structure and melt generation in subduction zones.

Chemistry of interstitial waters of ODP Leg 112 samples

Two distinct hydrogeochemical regimes currently dominate the Peruvian continental margin. One, in shallower water (150-450 m) shelf to upper-slope regions, is characterized by interstitial waters with strong positive chloride gradients with depth. The maximum measured value of 1043 mM chloride at Site 680 at ITS corresponds to a degree of seawater evaporation of ~2 times. Major ion chemistry and strontioum isotopic composition of the interstitial waters suggest that a subsurface brine that has a marine origin and is of pre-early Miocene "age," profoundly influences the chemistry and diagenesis of this shelf environment. Site 684 at ~9°S must be closest to the source of this brine, which becomes diluted with seawater and/or interstitial water as it flows southward toward Site 686 at ~13?S (and probably beyond) at a rate of approximately 3 to 4 cm/yr, since early Miocene time. The other regime, in deep water (3000-5000 m) middle to lower-slope regions, is characterized by interstitial waters with steep negative and nonsteady-state chloride gradients with depth. The minimum measured value of 454 mM chloride, at Site 683 at ITS, corresponds to ~20% dilution of seawater chloride The most probably sources of these low-chloride fluids are gas hydrate dissociation and mineral (particularly clay) dehydration reactions. Fluid advection is consistent with (1) the extent of dilution shown in the chloride profiles, (2) the striking nonsteady-state depth profiles of chlorides at Sites 683 and 688 and of 87Sr/86Sr ratios at Site 685, and (3) the temperatures resulting from an average geothermal gradient of 50°C/km and required for clay mineral dehydration reactions. Strontium isotope data reveal two separate fluid regimes in this slope region: a more northerly one at Sites 683 and 685 that is influenced by fluids with a radiogenic continental strontium signature, and a southerly one at Sites 682 and 688 that is influenced by fluids with a nonradiogenic oceanic signatures. Stratigraphically controlled fluid migration seems to prevail in this margin. Because of its special tectonic setting, Site 679 at ITS is geochemically distinct. The interstitial waters are characterized by seawater chloride concentrations to -200 mbsf and deeper by a significantly lower chloride concentration of about two-thirds of the value in seawater, suggesting mixing with a meteoric water source. Regardless of the hydrogeochemical regime, the chemistry and isotopic compositions of the interstitial waters at all sites are markedly modified by diagenesis, particularly by calcite and dolomite crystallization.

Carbon, boron and oxygen isotope ratios of sediments from the Moresby Seamount

We report results from boron, carbon and oxygen stable isotope analyses of faulted and veined rocks recovered by scientific ocean drilling during ODP Leg 180 in the western Woodlark Basin, off Papua New Guinea. In this area of active continental extension, crustal break-up and incipient seafloor spreading, a shallow-dipping, seismically active detachment fault accommodates strain, defining a zone of mylonites and cataclasites, vein formation and fluid infiltration. Syntectonic microstructures and vein-fill mineralogy suggest frictional heating during slip during extension and exhumation of Moresby Seamount. Low carbon and oxygen isotope ratios of calcite veins indicate precipitation from hydrothermal fluids (delta13C PDB down to -17?; delta18O PDB down to -22?) formed by both dehydration and decarbonation. Boron contents are low (<7 ppm), indicating high-grade metamorphic source rock for the fluids. Some of the delta11B signatures (17-35?; parent solutions to calcite vein fills) are low when compared to deep-seated waters in other tectonic environments, likely reflecting preferential loss of 11B during low-grade metamorphism at depth. Pervasive devolatilization and flux of CO2-rich fluids are evident from similar vein cement geochemistry in the detachment fault zone and splays further updip. Multiple rupture-and-healing history of the veins suggests that precipitation may be an important player in fluid pressure evolution and, hence, seismogenic fault movement.

Bromine and iodine concentrations in sediments and pore fluids of ODP Leg 112

At the Peruvian convergent margin, two distinct pore fluid regimes are recognized from differences in their Cl- concentrations. The slope pore fluids are characterized by low Cl- concentrations, but elevated Br- and I- concentrations due to biogenic production. The shelf pore fluids exhibit elevated Cl- and Br- concentrations due to diffusive mixing with an evaporitic brine. In the slope pore fluids, the Br-, I-, and NH4+ concentrations are elevated following bacterial decomposition of organic matter, but the I- concentrations are in excess of those expected based on mass balance calculations using NH4+ and Br- concentrations. The slope sediment organic matter, which is enriched in iodine from oxidationreduction processes at the oxygenated sediment-water interface, is responsible for this enrichment. The increases in dissolved I- and the I- enrichments relative to NH4+ and Br- correlate well with sedimentation rates because of differential trapping following regeneration. The pore-fluid I-/Br- ratios suggest that membrane ion fiitration is not a major cause of the decreases in Cl- concentrations. Other possible sources for low Cl- water, including meteoric water, clathrate dissociation, and/or mineral dehydration reactions, imply that the diluting component of the slope low-Cl- fluids has flowed at least 1 km through the sediment. The low bottom-water oxygenation in the shelf is responsible for the low (if any) enrichment of iodine in the shelf sediments. Fluctuations in bottom-water oxygen concentrations in the past, however, may be responsible for the observed variations in the sediment I/Br ratios. Comparison of Na+/Cl- and Br-/Cl- molar ratios in the pore fluids shows that the shelf high-Cl- fluid formed from mixing with a brine that formed from seawater concentrated by twelve to nineteen times and probably was modified by halite dissolution. This dense brine, located below the sediment sections drilled, appears to have flowed a distance >500 km through the sediment.

Geochemistry of sediments beneath the Tonga-Kermadec arc

The fate of subducted sediment and the extent to which it is dehydrated and/or melted before incorporation into arc lavas has profound implications for the thermo-mechanical nature of the mantle wedge and models for crustal evolution. In order to address these issues, we have undertaken the first measurements of 10Be and light elements in lavas from the Tonga-Kermadec arc and the sediment profile at DSDP site 204 outboard of the trench. The 10Be/9Be ratios in the Tonga lavas are lower than predicted from flux models but can be explained if (a) previously estimated sediment contributions are too high by a factor of 2-10, (b) the top 1-22 m of the incoming sediment is accreted, (c) large amounts of sediment erosion are proposed, or (d) the sediment component takes several Myr longer than the subducting plate to reach the magma source region beneath Tonga. The lavas form negative Th/Be-Li/Be arrays that extend from a depleted mantle source composition to lower Th/Be and Li/Be ratios than that of the bulk sediment. Thus, these arrays are not easily explained by bulk sediment addition and, using partition coefficients derived from experiments on the in-coming sediment, we show that they are also unlikely to result from fluid released during dehydration of the sediment (or altered oceanic crust). However, partial melts of the dehydrated sediment residue formed at ~800 °C during the breakdown of amphibole +/- plagioclase and in the absence of cordierite have significantly lowered Th/Be ratios. The lava arrays can be successfully modelled as 10-15% partial melts of depleted mantle after it has been enriched by the addition of 0.2-2% of these partial melts. Phase relations suggest that this requires that the top of the subducting crust reaches temperatures of ~800 °C by the time it attains ~ 80 km depth which is in excellent agreement with the results of recent numerical models incorporating a temperature-dependent mantle viscosity. Under these conditions the wet basalt solidus is also crossed yet there is no recognisable eclogitic signal in the lavas suggesting that on-going dehydration or strong thermal gradients in the upper part of the subducting plate inhibit partialmelting of the altered oceanic crust.