V51G-01 INVITED
Water and Melting in Back-arc Basins: New perspectives from the Eastern Lau Spreading Center
Since the work of Stolper and Newman (EPSL, 1994) it has been well recognized that water and extent of melting correlate positively in back-arc basin basalts. Quantification of this effect has been used to determine the effect of water content in the source on extent of melting. The slope of the relationship δF/δH2Oo is linear, and varies from one back-arc basin to another. MELTS and other modeling (Hirschmann et al., J. Petrol., 1999; Gaetani and Grove, Contrib. Mineral. Pet. 1998; Geophys. Mon., 2003; Kelley et al., JGR, 2006) has led to the suggestion that the slope varies regularly with mantle temperature, and that water has a much larger effect on melting at higher compared to lower temperatures. This modeling has been done in the context of isothermal, isobaric addition of water. For back-arcs worldwide, a critical aspect of the data is that more hydrous basalts have very low Fe contents, even when corrected appropriately for hydrous fractionation. This leads to clear negative correlations between Fe and H2O corrected back to mantle values at Fo90. The 3 wt.% variations in Fe content are not compatible with isobaric models, and require very different melting conditions for hydrous basalts as compared to anhydrous back-arc basalts. Back-arc basin basalts also plot on the global correlations of axial depth and Na8.0, and this relationship has been used to estimate mantle temperatures in back-arc basins, which on this basis extend to very high values. New data on major elements, trace elements and water from the Eastern Lau Spreading Center (ELSC), along with a re-evaluation of global back-arc data and modeling of mantle melting in the context of a polybaric spreading center environment (Langmuir et al., Geophys. Mon., 2006) provide new perspectives on these issues. The ELSC1 segment has a lower δF/δH2Oo than both the Mariana and Manus Basins, which would suggest the lowest temperature. However, its extent of melting inferred from its "F" intercept (on a plot of F vs. water in the source) is similar to the Marianas, suggesting a similar temperature, and its Na contents are as low as Manus, suggesting a high temperature. These inconsistent results can be understood from quantitative models and a more realistic melting process beneath back-arc spreading centers. δF/δH2Oo does not change with mantle temperature. In the back-arc environment, there are two independent halves of the melting regime, the "dry side" and the "wet side." The dry side undergoes polybaric fractional melting like other ocean ridges. The wet side (somehow) produces low pressure equilibrium hydrous melts with high water and low Fe contents. Mixing between the two creates the back-arc arrays. Large variations of Fe and Ti that anti-correlate linearly with water reflect this two component mixing in the back-arc. Both Ti and Na are mobile in the back-arc mantle, and source depletion and enrichment is an essential factor for evaluation of mantle temperature variations. Despite the low Na contents in the Lau Basin does not appear to be particularly hot, and instead is derived from a depleted mantle with low Na contents at only modestly elevated potential temperatures of 1400 degrees.
V51G-02
Trace element constraints on the origin of subduction components in the Eastern Lau Back- arc Spreading Center
Back-arc basin basalts display a unique and wide spectrum of chemical compositions ranging from depleted to enriched mid-oceanic ridge basalts to island arc basalts. The general explanation has been to relate the arc component to the "subduction input" and mantle enrichments as observed in mid-oceanic ridges (E-MORB) to old, pre-existing mantle compositions that flow into the mantle wedge. Melting has also generally been evaluated using the assumption of immobility of high field strength elements (HFSE) in subduction components, from which mobile elements systematics have been evaluated. A new and comprehensive geochemical data set on closely spaced samples from the Eastern Lau Spreading Center (ESLC) provides new perspectives on these issues. Here we emphasize the northern ELSC, where samples with a large range of water contents and factor of five variations in trace element abundances occur over a along a single ridge segment. Based on the trace elements systematics, there are two well-defined enriched components that occupy two distinct portions of the segment along-axis. The southern magmatic province is influenced by a water-rich component (wet component) that has strong affinities with island arc-like volcanism. The most enriched and depleted samples occur on a single segment within hundreds of meters of each other, requiring their generation from a single melting regime. The northern magmatic province is influenced by a component that has intermediate water content (damp component) and geochemical affinities with both E-MORB and island arc basalts. Trace elements inversion from the wet component samples indicates that the water-rich subduction fluids significantly mobilize HFSE, light rare earth elements (REE) and Y. The least affected elements are the heavy REE, and lutetium provides the best estimates of extent of melting to evaluate source component compositions. Using the observed linear correlations of water vs. other elements in the source, chemical compositions of both damp and wet components can be determined assuming the water content. The chemical variations and modeling suggest systematic and coupled relationships between a water-rich subduction fluid and a low-degree melt component that is consistent with melting of recycled eclogite or mantle peridotite. Mixing processes between these two components account well for the chemical diversity of the subduction components (wet and damp) observed in N-ELSC. We propose a model where mixing processes occur at depth in the subduction zone and where both components are derived from the subduction process and added to the back arc region via mantle diapirism. This model may apply generally in back-arc settings, and explain the common occurrence of "E-MORB" like components in this environment.
V51G-03
Mantle Flow and Melting Processes Beneath Back-Arc Basins
The chemical systematics of back-arc basin basalts suggest that multiple mechanisms of melt generation and transport operate simultaneously beneath the back-arc, resulting in a continuum of melts ranging from a relatively dry, MORB-like end-member to a wet, slab-influenced end-member [e.g., Kelley et al., 2006; Langmuir et al., 2006]. Potential melting processes at work include adiabatic decompression melting akin to that at mid-ocean ridges, diapiric upwelling of hydrous and/or partially molten mantle from above the subducting lithospheric slab [e.g., Marsh, 1979; Hall and Kincaid, 2001; Gerya and Yuen, 2003], and melting of back-arc mantle due to a continuous flux of slab-derived hydrous fluid [Kelley et al., 2006]. In this study, we examine the potential for each of these melting mechanisms to contribute to the observed distribution of melts in back-arc basins within the context of upper mantle flow (driven by plate motions) beneath back-arcs, which ultimately controls temperatures within the melting region. Mantle velocities and temperatures are derived from numerical geodynamic models of subduction with back-arc spreading that explicitly include adiabatic decompression melting through a Lagrangian particle scheme and a parameterization of hydrous melting. Dynamical feedback from the melting process occurs through latent heating and viscosity increases related to dehydration. A range of parameters, including subduction rate and trench-back-arc separation distances, is explored. The thermal evolution of individual diapirs is modeled numerically as they traverse the mantle, from nucleation above the subducting slab to melting beneath the back-arc spreading center, and a range of diapir sizes and densities and considered.
V51G-04
How does recycling of sediment components in arc magmatism really work?
Past work indicates substantial recycling of a sediment component rich in LILE, Th, Sr, Pb and LREE in arcs. For example, in the relatively well-constrained case of Central America, Plank et al (Geology 02) estimate that 80% of subducted, sedimentary Th is recycled in arc magmas. To understand how such a component is transferred from subducted sediment to arc lava, we examined trace-element variation in (a) mid-crustal (0.4 GPa) contact metamorphic rocks (Austin & Kelemen, Fall 06 AGU) and (b) ultrahigh-pressure (UHP, > 3 GPa) metasediments. Most UHP samples were metamorphosed along subduction-zone geotherms (Hacker, Int Geol Rev 06), but some record substantially higher T (e.g., Erzgebirge & Kokchetav, Massone EPSL 03). Unmelted, mid-crustal metapelites are indistinguishable from pelitic sediments for the entire suite of elements analyzed by ICP-MS at WSU. Melt extraction from the mid-crustal metapelites led to systematic depletion of incompatible elements in high-grade hornfels. Depletion increases with decreasing distance to the contact with a mafic pluton, most clearly at peak T > 750°C. In contrast, although many UHP metapelites record PT above the aqueous fluid-saturated solidus, and have fluid inclusions and/or hydrous phases, compared to pelites they show no detectable depletion of "fluid-mobile" elements such as LILE (Cs, Rb, Ba, U, K), Sr and Pb, no depletion of "fluid-immobile, incompatible" elements such as Th and LREE, and no systematic change in key soluble/insoluble ratios such as Ba/Th or K/Zr up to ~1000 C. Mobility of incompatible elements is evident for T > 1000 C, well above PT for subduction-zone geotherms. Presumably, trace phases rich in LILE, Th and LREE persist to ~1050 C in metapelites at UHP conditions.How can our observations be reconciled with the recycled sediment component in arc lavas? Our preferred hypothesis is that low-density metasediments rise into the mantle wedge when heating yields viscosities low enough for density-driven instabilities (Ringwood JGSL 74; Marsh AJS 76; Gerya & Yuen EPSL 03; Kelemen et al, Treatise on Geochem 03). In the wedge, metasedimentary diapirs heat as they rise, and undergo large degrees of super-adiabatic partial melting which exhaust trace phases, releasing the sediment component observed in arcs.
V51G-05 INVITED
Implications for backarc basin dynamics including a low viscosity wedge in simple 2 and 2.5D numerical flow models
Flow in the asthenospheric mantle wedge associated with backarc basins is often assumed to be driven by motions of the overlying and adjacent plates, suggesting a roughly 2D corner flow pattern with mantle flow rates comparable to plate velocities. Geodynamic models have a notoriously difficult time capturing and explaining some first order observations such as intermediate dips of subducting plates, and initiating and sustaining backarc spreading. In addition, arc-parallel shear wave splitting and geochemical observations imply along- strike flow in the Lau, Mariana, and other backarc basins, presenting a further challenge to geodynamic models of plate-driven subduction systems, as plate-driven flow models predict arc-perpendicular fast direction with A-type mineral texturing. Although B-type mineral texturing could rotate the fast directions 90 degrees relative to A-type and likely occurs in the forearc mantle, such texturing is unlikely in the asthenosphere wedge with its higher temperatures and weaker rheology, requiring a different explanation. Some of these outstanding issues in backarc dynamics may be addressed by considering a low viscosity region in the upper corner of the wedge. Recent work demonstrates that the existence and growth of a low viscosity wedge can control slab dip [Manea, 2006]. Here, I examine the effect of a low-viscosity wedge on asthenospheric flow in the along arc dimension and possible controls on the localization of backarc spreading. Along-arc flow rates through a low viscosity channel may be significantly faster than plate rates. Constraining along-arc flow velocity in the Lau subarc mantle by tracking mineral texturing development within model flow fields with a low viscosity wedge, show rates upwards of 50 cm/yr. Models without a low-viscosity channel beneath the arc fail to simultaneously satisfy the shear wave splitting and geochemical constraints. A low viscosity wedge also modifies the dynamic stress field within the mantle wedge, resulting in horizontal tensional stresses at the base of the overlying plate directly over the low viscosity region. These results may have implications on the localization of rifting in the backarc and the development of backarc spreading. In addition to rates of enhancing mantle flow to rates substantially faster than plate rates which may have profound implications for plate boundary mantle convection and mixing of mantle reservoirs, low viscosity regions in the wedge may contribute to the deformation and development of the basins themselves.
V51G-06
High-Ca Boninites From the Modern Tonga Arc
High-Ca boninites are volcanic rocks with unusual compositions (SiO2>53 wt%, Mg#>0.6, CaO/Al2O3>0.75) found in forearcs and trenches, continental cratons, and ophiolites. Generation of high-Ca boninites requires a combination of refractory mantle sources, elevated mantle temperatures and the addition of hydrous fluids. To satisfy these conditions, petrogenetic models invoke unusual tectonic settings such as subduction initiation, ridge subduction, or mantle plume interaction. We have discovered high-Ca boninites from an active arc volcano, Volcano A, a submarine volcano in the Tonga arc dredged during the NoToVE cruise in Nov 2004. Multi-beam sonar images of two pristine volcanic cones and glassy samples lacking Mn coatings suggest that these edifices were formed by modern volcanism. The boninites are represented in both the whole rock and melt inclusion populations of a sample dredged from a ridge on the northern flank of the northern cone. Similarities in the major element compositions of the largely aphyric whole rock and the glassy melt inclusions support both as samples of true boninitic liquids (MgO>9 wt%). These liquids are related by coupled crystal fractionation (from Fo92 to Fo85 in olivine hosts) and degassing (from 4 to 1 wt% H2O in the melt inclusions). Three other dredges from Volc A include whole rocks, glass, and melt inclusions that are related to the boninites by crystal fractionation. Taken together, the samples from Volc A represent a suite of boninites and their differentiates, forming a coherent liquid line of descent with parallel whole rock REE patterns which become more enriched with decreasing Mg#. The REE patterns for Volc A whole rocks are depleted in LREE, however, in contrast to the characteristic U-shaped REE patterns of classic boninites. Volc A is only the second example of boninites being erupted in an active volcanic arc, the first being Bamus volcano in New Britain (Johnson et al., Geol. Rund., 1983). Volc A is not remarkable in its geographical setting, located in the central portion of the Tonga arc, at around 21°S, along-strike with the arc island of Hunga- Ha'apai, and east of the Eastern Lau Spreading Center. Although it contains volcanic rocks with the highest Ba/La (100-130), Si6.00 (56 wt%) and lowest Ti6.00 (0.44 wt%) in the Tonga arc, Volc A forms an end- member within a continuous spectrum of compositions, and much of the Tonga arc appears to have boninite- affinities. Of the three conditions that promote boninite production (mantle depletion, high mantle temperatures, and high mantle water contents), we can exclude mantle temperature as the primary cause of Volc A boninites. Olivine-liquid temperatures are 1348±13°C for Volc A liquids, calculated in equilibrium with Fo92 mantle olivine at 2 GPa (using thermometers in Sisson & Grove, CMP, 1993; Sugawara, JGR, 2000). Such temperatures are very similar to those calculated for Central American arc volcanoes (1315-1320°C; Plank et al., this meeting) where magmas with similar water contents erupt, but boninites do not. Volc A olivine-liquid temperatures also contrast with those recorded in the high-Ca boninites found in the Tonga forearc (~1480°C; Falloon and Danyushevsky, JPet, 2000). Thus, we attribute the presence of boninites in the Tonga arc to the remelting of refractory mantle that melted previously during rapid back-arc spreading in the nearby Lau Basin.
V51G-07
Plate-driven flow constraints on arc and back arc magma mixing: influence of plate boundary geometry in the Lau Basin
The Lau spreading centers, from the slower-rate Valu Fa ridge in the south to the faster-rate Central Lau rift valley axis in the north, show significant variations in geochemical and geophysical observables that may indicate variable interaction between the ridge system and the arc. In the conceptual model of Martinez and Taylor [2003], the Valu Fa ridge, which is nearest the arc, has increased magma flux and a more andesitic geochemistry due to interaction between the spreading center and the arc melt supplies. This interaction is predicted to decrease as the ridge-arc distance increases steadily to the north along the Eastern and Central Lau spreading centers. To test this conceptual model we develop a series of 2D finite element mantle flow models, for each of the Valu Fa, and Eastern Lau and Central Lau spreading centers. These models are designed to discern effects solely due to changes in the trench-perpendicular plate motion and boundary geometry. The numerical experiments simulate flow, temperature, melting, and depletion of the residual mantle. Preliminary models, with a simple rheology and dry melting, predict that Valu Fa would develop a melting region whose low-fraction (<1%) portion is conjoined with the arc melting region. For the Central Lau spreading center, the lower melt fraction regions are not predicted to be conjoined. The maximum melt fraction beneath the spreading center is greater for the Central Lau model than for models of the southern ridges. The Eastern Lau spreading, in between, is predicted to develop a conjoined melt region. The higher northern model melt fractions are distributed over broader regions than for Valu Fa, indicating that a greater volume of melt should be extracted from the Central and Eastern Lau spreading centers. While the dry melting case does not support the hypothesis of higher melt flux beneath the Valu Fa, we expect that the addition of wet melting in the simulation will produce a significant amount of melting above the slab, and this combined with the dry melting beneath the ridge may produce a greater quantity of melt beneath the Valu Fa. Furthermore, we will explore the effects of different mantle rheologies and different melt extraction assumptions on melt production beneath the ridge.
V51G-08
High Field Strength and Rare Earth Element Systematics of the Kurile arc.
The depletion of high field strength elements (HFSE) relative to other trace elements with similar mantle/melt partition coefficients (e.g., rare earth elements; REE) is characteristic of subduction-related volcanic rocks. Although this behavior is common to most western pacific arcs (e.g., Mariana, Izu, Japan, Kamchatka arcs) the exact cause of HFSE depletions is still debated. This debate centers on the relative importance of mass transfer processes from the subducting slab to the overlying mantle wedge: how important are fluids versus melts in slab-mantle mixing in arcs? New Hf and Nd isotope and HFSE concentration data for lavas erupted from the length of the Kurile arc, both at the volcanic front and within the backarc, are reported. Relative to the Mariana, Izu, and Kamchatka arcs, the Kurile arc is characterized by more radiogenic Nd and Hf isotope ratios at the volcanic front. Concentrations and ratios of HFSE and REE are similar to those reported for the Izu and much of the Kamchatka arcs. When corrected for the effect of slab contributions, the Kurile arc is isotopically "Indian". Across-arc variations are also similar, suggesting that the nature of slab contributions changes with depth. Lavas erupted in the backarc have lower Nd and slightly lower Hf isotope ratios, and lower Ba/Th and Hf/Sm, than those at the volcanic front. Slab-mantle mixing can account for these isotope and trace element observations if HFSE are effectively fractionated from REE. Attributing HFSE/REE depletions in arcs to the relative fluid-immobility of HFSE relative to REE cannot explain both the isotope and concentration data, because neither the HFSE nor the REE are fluid mobile enough to be added to the mantle wedge in any substantial quantity. Hf and Nd isotope ratios and HFSE/REE concentration data require a melt transport agent, in which case HFSE are fractionated from REE by the presence of trace quantities of residual accessory phases (rutile and zircon) in the subducting slab. Slab melt is, therefore, ubiquitous along and across the Kurile arc.