V33E-01 INVITED
The Oxidation State of Global Subduction Zone Basalts and its Relationship to Volatiles, Magmatic Processes, and Source Composition
Oxidation state is a central variable in magmatic systems. In subduction zones, the mantle wedge is exposed to hydrous fluids from an oxidized subducting plate, potentially driving a fundamental shift in the oxidation states of arc and back-arc basin magmas and their sources. Despite its importance, however, magmatic oxidation state and its relationship to conditions in the mantle source can be difficult to constrain. Here, we present new, in-situ μ-XANES analyses of Fe+3/ΣFe ratios, as an indicator of melt oxidation state, in natural, primitive pillow glasses from the Mariana, Lau, and Manus back-arc basins (MgO>6 wt.%; n=31) and a global suite of olivine-hosted arc melt inclusions (MI; MgO>4 wt.%; n=16). These new data show that back-arc basin basalts preserve Fe+3/ΣFe ratios of 0.14-0.21, more oxidized than MORB (Fe+3/ΣFe=0.11-0.17), and arc basalts range to even higher ratios of 0.17-0.36. Analysis of MI equilibrium with host olivine compositions indicates that either post-entrapment crystallization or outward Fe+2 diffusion may have occurred in the MI's studied, but the magnitude of these effects is small (9±5% change in FeO; see also Cottrell & Kelley, this mtg.). Coupled with new and existing major element, volatile (H2O±CO2, S, Cl, F), and trace element data, we also test the variation of melt oxidation state with indicators of extent of crystal fractionation and of mantle source composition. The arc and back-arc glasses capture a full range of natural, undegassed magmatic H2O concentrations (0.1-5.3 wt.%), and show a general, global increase in Fe+3/ΣFe with increasing H2O content, although the Mariana trough defines a trend distinct from the Manus and Lau basins. The Fe+3/ΣFe ratio does not correlate with Mg#, suggesting that the melt oxidation states are not controlled by the extent of crystal fractionation. In the Mariana trough, Fe+3/ΣFe does increase with increasing Ba and Sr concentrations, suggesting a direct link between melt oxidation state and the magnitude of addition of a slab-derived component rich in H2O, Ba, and Sr to the mantle source beneath the back-arc basin.
V33E-02
Water/Cerium as a Proxy for Slab Fluid Temperature
Numerical models and laboratory experiments predict widely varying temperatures of fluid generation in subducting slabs beneath volcanic arcs, from <600° C to >900° C. Moreover, slab thermal structures are expected to vary regionally, due to variations in mantle temperature, slab coupling, slab age and convergence rate. In order to provide constraints for such predictions, we develop here H2O/Ce as a potential slab fluid thermometer. Unlike incompatible trace elements, where concentrations in the fluid are governed by source abundances and partition coefficients, Ce fluid concentrations will be governed by the strongly temperature-dependent solubility of allanite and monazite, common REE-phases in sediment and basalt lithologies [1-4]. At low temperatures (<600° C), fluids saturated in allanite or monazite will have very low abundances of REE (<10 ppm) and low solute contents (H2O >90 wt%). At high temperatures (>900° C) fluids or melts will have high abundances of REE (>100 ppm) and low H2O contents (<10 wt%). Thus H2O/Ce in fluids will decrease by several orders of magnitude as temperature decreases across this relevant range, from >10,000 to <1000. Such variation is in stark contrast to the very small H2O/Ce variations expected during mantle melting, and observed in oceanic basalts far from subduction zones (150-250 [5]). We have examined H2O/Ce variations in volcanoes from several arcs, using H2O measurements from least degassed, primarily olivine-hosted melt inclusions and REE measurements from the same inclusions or whole rocks. H2O/Ce varies dramatically from <400 at Irazu in Costa Rica [6] and Tuxtepec in Mexico [7] to ~20,000 at submarine Volcano A in Tonga [8]. H2O/Ce in the fluid component for each arc can be estimated from linear mixing of Nb/Ce- H2O/Ce, between mantle, arc and a fluid end-member with zero Nb/Ce. Arc fluid end-members decrease in H2O/Ce systematically from Tonga and the Marianas (>5,000) to the Aleutians and Central America (<5,000) to Mexico (<1,500), consistent with a greater proportion of low temperature fluids (<700° C) supplying Tonga and Marianas volcanoes, and a greater proportion of high temperature fluids or melts (>850° C) supplying Mexican volcanoes. These inferred differences in the mean fluid temperatures appear to correlate with the slab thermal parameter (the product of age, convergence rate and sine of the slab dip angle), even though most dynamical models predict that other factors, such as slab coupling, mantle temperature, fluid transport dynamics and depth to the slab, are more likely to control slab surface temperatures beneath the arc. 1. Rapp and Watson, 1986, ContribMineralPetrol; 2. Montel, 1993, ChemGeol; 3. Hermann, 2002, ChemGeol; 4. Klimm et al., 2008, JPet; 5. Dixon et al., 2002, Nature; 6. Benjamin et al., 2007, JVGR; 7. Cervantes and Wallace, 2002, Geology; 8. Cooper et al., 2007, EosTransAGU.
V33E-03
Heterogeneous δ18O in the mantle wedge beneath Medicine Lake and Mt. Shasta volcanoes (California): ancient or modern subduction signature?
This study presents new analyses of O-isotopes in olivine phenocrysts from most primitive high alumina olivine tholeiite (HAOT) from Medicine Lake volcano (MLV; California) with MgO > 8%. The measured δ18OOl-values range from 4.69‰ to 5.49‰, with an average of 5.07‰ (n = 12), which is low relatively to the mantle olivine values (5.2‰±±0.2‰). We compare these data to O-isotopes measured in olivine phenocrysts from the most primitive lavas from Mount Shasta, which show high δ18OOl-values relatively to the olivine mantle value, 5.89-6.08‰ in HAOT (n = 2), 5.31-5.81‰ in basaltic andesite (BA; n = 7) and 5.54-5.85‰ in primitive magnesian andesite (PMA; n = 5). The primitive crystal poor nature of these lavas, Mg# > 0.65, and the fact that we analyzed olivine, the first mineral to crystallize in these lavas, allow us a good assessment to mantle-derived magmas. The HOAT are known to be generated by nearly anhydrous melting of spinel peridotite, which makes them a good indicator of the composition of the mantle unaffected by the present day subduction fluids. Therefore it appears that the mantle beneath MLV has a low- to mantle-like δ18OOl with variation of up to 0.8‰. However, beneath Mt Shasta the mantle has a relatively high δ18OOl and more homogeneous (based on two samples). Overall, it appears that the arc mantle of the South Cascade segment is heterogeneous with more than 1.3‰ variation in δ18OOl. The question that we address here is: Is the high-δ18O signature measured in olivine phenocrysts from BA and PMA from Mt Shasta come from the preexisting mantle source itself or from the present subduction fluids? If we consider BA and PMA to be generated by high-δ18O fluids flux melting in the present subduction environment, how can we explain high-δ18O values measured in HAOT? The heterogeneous character of the mantle in the South Cascades could be due to ancient subduction fluids more and more depleted during the slab dehydration that fluxed the mantle wedge in order to generate high δ18O mantle-like beneath Mt Shasta and low δ18O mantle-like beneath MLV. Therefore BA and PMA in Mt. Shasta could be generated by low-δ18O fluids flux melting in the present subduction environment.
V33E-04
New Constraints on Fluid Addition Beneath the Tonga Arc: Reconciliation of U-Th-Ra Disequilibria in a Single-Stage Fluid Addition Model
Previous U-series isotope studies of Tongan lavas yielded inclined arrays on a U-Th equiline diagram. Taken together, these along-arc data were inferred to record timescales of ~ 50 kyr since fluid addition of U and large Ra excesses were reconciled using second-stage (or continuous) fluid addition models. However, Ata island displayed a clear horizontal trend on the U-Th equiline diagram suggesting that here, addition of U could have occurred only a few kyr ago. New high precision U-Th disequilibrium data have been obtained by solution MC-ICP-MS analysis for a new suite of Tongan rocks. Samples from the islands of Tofua, Kao and Late all form horizontal arrays on the U-Th equiline diagram. Thus is appears that the 50 kyr array in earlier studies actually represents a series of horizontal en-echelon island trends, each recording recent addition of slab derived fluid to the mantle wedge. Thus, the need for Th addition via the fluid is negated, in good agreement with its fluid immobile character. Moreover, this new insight enables reconciliation of the observed positive correlation between (238U/230Th) and (226Ra/230Th) that, until now, required a two-stage (or continuous) fluid addition model in order to account for the persistence of both U and Ra excesses in the same samples. The very large (up to 600%) Ra excesses in many rocks cannot be produced by in-growth melting models. Identification of positive correlations between (226Ra/230Th) and widely accepted indices of fluid addition (e.g. Ba/Th, Sr/Th) support the conclusion that metasomatism of the Tongan sub-arc mantle wedge was brought about by rapid single-stage fluid addition from the down-going subducted Pacific plate.
V33E-05
Mantle Melting beneath Central Japan with Overlapping Subducting Plates and Enhanced Fluid Flux
Arc magmas are thought to be originated through the interaction among the subducting slab, the slab- derived fluid (slab-fluid), and the mantle wedge. There are two important processes, except for crustal processes, concerning generation of arc magmas: one is the fluid process involving slab-fluid which metasomatizes the source region of mantle melting, and another is the melting process of the source mantle.. We aim to reveal (1) how the two processes are combined and can consistently explain the volcanic rock compositions as the products of the processes, and (2) how the variations of fluid flux, especially an enhanced fluid flux from the slabs beneath central Japan (Nakamura et al., 2008), affects the melting conditions. Nakamura et al. (2008) have identified the two slab-fluids from the two subducting plates beneath central Japan, and have found that the fluid fluxes are sensitive and vary with the geometry and configuration of the subducting slabs, including an enhanced fluid flux where the two slabs clearly overlap. Here, we examine the melting process in the mantle wedge above the overlapping slabs, which are then compared with the melting process above the single slab, e.g., beneath the Izu-arc. The amounts and chemical compositions of the slab-fluids that have been added to the source region of melting are well documented by Nakamura et al. (2008), based on which the melting conditions can be tightly constrained using the chemical compositions of most undifferentiated rocks. The melting process is described by two parameters, melting degree and the proportion of garnet lherzolite involved in melting as Bgt = garnet lherzolite/(garnet lherzolite + spinel lherzolite). Major and trace elements on 417 samples, including REEs, and Sr-Nd-Pb isotope ratios from 28 volcanoes have been analyzed for the Quaternary volcanic rocks in central Japan. The trace elements composition of primary magma is estimated by correction for fractionated phases to the most undifferentiated rock for each volcano. The melting degree and Bgt are optimized to fit the concentrations of Nb and heavy rare earth elements to those in the estimated primary magmas, and then test the reproducibility of other incompatible elements. As a result, the melting process in central Japan is characterized by relatively low melting degree and high Bgt compared to the Izu-arc, which implies that the melting occurred at deeper depths. Therefore, the overall geochemical feature of the volcanic rocks in central Japan can be attributable to the combination of enhanced fluid flux and low melting degree at relatively deep depths due to the cold environment associated with the overlapping subduction.
V33E-06
Origin of fluids in hydrothermal systems at the Kusatsu-Shirane Volcano, Japan: Halogen and I-129 results
Kusatsu-Shirane Volcano is located at the volcanic front on Honshu Island, Japan, has a crater lake (Yugama) at the summit and is surrounded by numerous hot springs. We used halogen concentrations in combination with the I-129 system in order to investigate the relation of fluids to processes at this active subduction zone. Because of the biophilic nature of I and the presence of a cosmogenic isotope, I-129, the iodine system together with halogen systematics has been used for the determination of source formations for hot springs and fluids in a variety of settings. We report here halogen data from 56 lake water samples and from 41 hot spring samples, collected from 1987 to 2008, and I-129/I ratios from a subset of these samples. Concentrations of iodine, bromine and chlorine in Yugama and hot spring waters ranged between 0.001 - 0.060 mM, 0.01 - 0.012 mM and 45 - 100mM, respectively. While all the iodine concentrations were higher than those in seawater by factors up to 100, Cl concentrations typically are below that of seawater. Positive correlations between I, Br and Cl were observed, specifically the pattern of secular variations of iodine and bromine is similar to that of chloride in Yugama water. Relationships between the increase of halogen concentrations and the volcanic activity (e.g. frequency of earthquakes) of Mt. Kusatsu-Shirane were observed in the Crater Lake, in which iodine showed the highest increase among the three halogens. While I- 129/I isotope ratios in hot springs were generally below the pre-anthropogenic ratio (129)I/I = 1500x10-15), all samples collected from the Yugama crater lake have ratios above that value, indicating the presence of anthropogenic I-129 in that open system. The I-129/I ratios in hot springs are compatible with the estimated ratio for iodine derived from subducted marine sediments in this region. These results suggest that recycling of subducted sediments is an active source of iodine in the Kusatsu-Shirane hydrothermal system, in good agreement with the interpretation of earlier results for Cl concentrations.
V33E-07
Flux-melting across the entire Izu backarc
It is widely accepted that arc volcanism is the result of flux melting of the mantle wedge. The movement of material from slab to mantle wedge plays an important role in mediating mantle melting and elemental cycling during subduction, although the exact nature of the flux is more controversial. The differences between slab- melts and slab-fluids, and the question of whether these are derived from subducted sediment or basalt, have important implications for the thermal structure of the slab and the fate of subducted elements beyond the subduction zone. We use the composition of 48 mafic basalts (> 6 % MgO) ~200 km across the Izu arc and backarc, between latitudes 30-32 °N, to calculate the composition of primary mantle melts in equilibrium with Fo90 olivine. The degree of melting required to produce each calculated primary mantle melt (F) is then estimated from inversions of trace element data (Kelley et al., 2006), and F is then used to calculate the abundance of trace elements in the mantle source for each sample. As with other arcs and backarc basins, there is a clear linear relation between water content of the mantle source (calculated from melt inclusions) and inferred degree of melting. This suggests that mantle water contents control melting across the entire extent of the Izu back arc. However, commonly used proxies of slab-derived fluids (Ba/La, Pb/Ce) do not correlate with F, even though H2O does. Instead, commonly used proxies for sediment and sediment melts (Th/La, La/Yb) do correlate with F, at the 99 % confidence level. Nb/Zr and Pb isotope ratios also correlate with F. Pb-Pb and Nd-Hf isotope data for Izu backarc basalts are consistent with the addition of a flux derived from both altered oceanic crust (90 %-95 %) and sediment (5 %-10 %). The amount of sediment increases across the backarc and decreases with time. HFSE/LREE ratios (e.g., Nb/La, Hf/Sm) require partial melting of the subducting slab in the presence of residual rutile (1 %) and zircon (< 0.01 %). Hydrous partial melts of the slab likely carry H2O as a dissolved constituent to the mantle source of backarc volcanism. Implications of these findings are 3-fold: 1) partial melts of the slab, not aqueous fluids, cause flux melting of the Izu backarc mantle wedge across the entire region of observed recent magmatism; 2) the top of the slab must be hot enough to melt (> 850 °C); and 3) accessory phases (rutile and zircon) that remain residual after partial melting of the slab may contribute to the HFSE budget of OIB.
V33E-08
Implications Of Light And Trace Elements Signatures in Melt Inclusions Of St Vincent And Grenada Island On The Lesser Antilles Arc Behavior
St. Vincent and Grenada islands are located in the south part of the Lesser Antilles arc, generated by the
subduction of the Atlantic plate beneath the Caribbean plate. In the both islands, the erupted high-MgO
basalts (MgO > 10.0 wt%) are thought to be representative of the primary magmas and to be
generated by the melting of a MORB mantle source enriched by fluids derived from the subducted slab
[1].
We present here trace element compositions determined by ion probe in melt inclusions (M.I.)
trapped in olivines (Fo84-91) from magnesian scoriae, for which light and volatile elements, as
stable isotopes were previously measured [2-3]. Their major elements compositions point out a broad
variability, but show the primitive character of these M.I. (SiO2 < 50.0 wt%). Grenada M.I. are
enriched in K2O (0.4-2.5 wt%) and MgO (up to 12.5 wt%) compared to St. Vincent M.I.
(up to 0.85 wt% and 10.0 wt%, respectively). Their trace element patterns encompass those of
whole rocks. Compared to St. Vincent, Grenada M.I. recorded more variable trace element compositions. All
M.I. patterns are characteristic of subduction zones, with Ba and Sr enrichments associated with pronounced
negative Nb anomalies, implying slab-fluids influence, as also demonstrated by high Cl/F ratios (up tp
10.9) [4]. As a whole, REE patterns are enriched and poorly fractionated compared to MORB. The
trace element patterns, combined with light element and isotopic compositions, are interpreted in term of
variations of degree of mantle partial melting and slab influence, with variable contributions of aqueous fluids
released from altered oceanic crust and from sediments, and of sediment melt. Both St. Vincent and Grenada
primary magmas record the influence of aqueous fluids, whereas the addition of sediment melt is only
identified in Grenada M.I., as Zr positive anomalies (Zr contents up to 1200 ppm), a rare feature in
basaltic melts. Although light trace elements and stable isotopes illustrated mostly aqueous fluids from AOC
and sediments dehydrations [2-3], trace elements highlight the influence of silicate melts. This
comparison underlines the contribution of several different sources in the arc melt formation.
[1]
Macdonald et al., 2000
[2] Bouvier et al., 2008
[3] Bouvier et al., submitted
[4] Straub
and Layne, 2003