MR31D-01 INVITED
Localization Mechanisms of Melting and Melt Migration in the Mantle Wedge at Convergent Plate Boundaries
Along-strike magmatic segmentation with a 50-100 km scale has been recognized for several convergent plate boundaries, including the Aleutians, the Cascades (Marsh, 1979), and Japan (Tamura et al., 2002). Mantle wedge seismic velocity structure beneath Japan is also segmented with low seismic velocity columns beneath volcanic segments. Identifying plausible buoyant flow mechanisms that may lead to magmatic segmentation is thus an important aspect of understanding melting and melt migration in these settings. Reaction infiltration instability during melting has been identified as a possible mechanism of melt channelization (Aharonov et al. 1995; Spiegelman et al., 2001), but the predicted scale of several compaction lengths (1-200 m) appears to be too small for this mechanism alone to explain the observed segmentation. Thermal convection in the mantle wedge (Honda and Yoshida, 2005) may be driven by cool downwellings ~100 km in scale that develop at the base of the overriding plate. Decompression melting is thus localized in upwelling regions between the cool downwelling sheets aligned with plate motion. Convective flow in the mantle wedge of an appropriate scale might also be driven by instability of thickened crust beneath the magmatic arc (Behn et al., 2007), but a direct connection of this mechanism with melt production has not yet been explored. Instability may also result from buoyancy that develops along the top of the downgoing plate. Marsh and Carmichael (1974) proposed that melt generated along the top of the downgoing plate ascended diapirically; however recent evidence does not generally favor melting of the downgoing plate. Dehydration of the plate however releases hydrous fluid that should rise buoyant into the overlying mantle. Buoyant solid flow may hence develop due to the presence of hydrous fluid or low density serpentine (and other hydrous phases) formed in cool mantle just above the top of the downgoing plate. Serpentine formation involves a significant volume increase and serpentenized mantle is therefore less dense than dry mantle containing an equivalent amount of free water. Exothermic serpentine formation is limited by the heat of reaction. However, decreasing pressure in buoyantly ascending mantle allows the continuing formation of serpentine, thus leading to buoyant instability. Ongoing theoretical studies are exploring the conditions required for instability, the wavelength at which it would occur, and the physical properties of mantle materials that would make it possible.
MR31D-02
Length scales of magma transport in reactive two-phase flow
During magma migration, both interfacial tension and mass exchange between the matrix and the melt play an important role in controlling the efficiency and rate of melt extraction and the chemical signature of the magma. In this work, we develop a new formulation governing the dynamics of a two-phase aggregate coupling effects of interfacial tension and mass exchange between the melt and the matrix by dissolution-precipitation. Dissolution-precipitation, which is limited by the rate and length scales of diffusive mass transport of ions, typically redistributes melt over small length scales. A process likely to dominate in the length scale of laboratory experiments. Rapid diffusive mass redistribution can also significantly reduce the rate of segregation of buoyant small-wavelength melt pockets. Growth or decay of large wavelength melt structures such as melt-rich layers in the Earth's lower mantle, blobs of core-forming material in the proto-Earth, and magma bodies beneath volcanic arcs and midoceanic ridges is dominated by a balance between interfacial tension, buoyancy, and viscous deformation of the matrix. Efficiency of buoyancy-driven extraction of large wavelength melt structures are strongly modulated by interfacial tension depending on the average grain size of the matrix and the state of disaggregation of the matrix.
MR31D-03 INVITED
Deep Upper Mantle Melting Detected with ScS Reverberations near Subduction Zones
Observations of a low-velocity layer atop the mantle transition zone are often interpreted as a layer of partial melt trapped within the deep upper mantle. Multiple ScS reverberations are sensitive to radial changes in impedance within the mantle and have been used to investigate the extent and thickness of these low-velocity layers. Revenaugh and Sipkin (1994) first reported evidence for a layer of dense, silicate melt resting atop the 410-km discontinuity beneath eastern China and the Sea of Japan. A recent study with a larger dataset has confirmed the presence of this low-velocity layer and examines its geographic extent on the arc-side of the subducting Pacific and Philippine slabs. Laterally extensive and geographically coherent observations have also been made using ScS reverberations in the southwest Pacific, beneath the Coral and Tasman Seas. In both of these regions, the thickness of the proposed melt layer is on the order of ~~50-70 km, an estimate that is at odds with predictions by dynamic and petrologic models. We propose that the layer is produced by water-induced melting initiated in the upper mantle (~~350 km depth) with a resulting melt that is negatively buoyant and percolates down to the 410-km discontinuity. If this melt has a zero-degree dihedral angle, then complete grain boundary wetting may preserve the signature of the melt, even for very low melt fractions. Excess water may be introduced into the region via storage in dense hydrous magnesium silicates and nominally anhydrous minerals within subducting slabs. ScS reverberations have also been used to detect a low-velocity layer atop the 410-km discontinuity on the oceanward side of the subducting slab at the western margin of the Pacific plate. Without the slab to transport excess water into this region, we suggest the combined effect of low water content with increased temperature induces melting in this situation. Revenaugh, J. and Sipkin, S. A., 1994. Seismic evidence for silicate melt atop the 410-km mantle discontinuity. Nature (London). 369, 474-476.
MR31D-04
High resolution mapping of Earth's deepest magma chamber: ultra-low velocity zones at the core mantle boundary
Ultra-Low Velocity Zones (ULVZ) are seismically imaged enigmatic features at the core-mantle boundary (CMB). They are localized features characterized as thin layers (<40 km), with strong reductions in seismic velocities (up to 40%). Several seismic probes for ULVZ structure have been exploited, revealing strong evidence for ULVZ in a few isolated regions. Only about half of the area of the CMB has been seismically probed and many areas do not show any evidence for the presence of ULVZ. Lack of evidence for ULVZ structure either indicates that it is too small/thin (or mild in properties) to be detected with current seismic probes or that it is absent. High resolution seismic array studies reveal small ULVZ patches with scale lengths of a few hundred km's or less. They also show a rapid lateral transition from non-ULVZ mantle to ULVZ material. The seismic properties of ULVZ appear most consistent with partially molten material. Recent high resolution waveform studies also find evidence that the ULVZ material is denser than the surrounding mantle. Geodynamical calculations suggest that the ULVZ material can remain in distinct pockets even with large density increases that can exceed 10%. However, ULVZ viscosity plays an important role: if the ULVZ viscosity is significantly lower than the surrounding mantle, other mechanisms may be needed to keep ULVZ material in isolated pockets, and preventing it from flattening out on the CMB. Using a multidisciplinary approach, we study the existence, structure, and stability of ULVZs. We will present recent seismological evidence for ULVZ with seismic properties in agreement with the existence of dense partially molten material. New deconvolution processing techniques allow us to increase our resolution capabilities: we can now detect ULVZ layering down to about 2 km thickness. Using this method, we will present evidence for a previously undetected ULVZ structure east of Australia. High resolution geodynamical modeling shows that dense thermo-chemical piles (such as found beneath the central Pacific and southern Africa) might play an important role in ULVZ dynamics, including stabilizing ULVZ into lenses or ridges towards the perimeter edges of the piles. Seismological, mineral-physical and geodynamical evidence all point to the possibility of ULVZ being the deepest and largest magma chambers on Earth.
MR31D-05
Keeping mush mushy at the core-mantle boundary: the role of internal convection and secular cooling
Williams and Garnero (1996) proposed that thin (5-40 km thick) patches of dramatically decreased seismic velocity above the core-mantle boundary (ultralow-velocity zones, or ULVZ) could best be explained by the presence of partial melt, and this remains the favored mechanism to explain the anomalous seismic properties. However, simple estimates for compaction and expulsion of melt from a porous solid on the order of 1 Gyr require an effective bulk viscosity that is probably much larger than realistic, and therefore melt should have separated long ago from the interstices it occupies in the matrix. In more detail, however, this layer is subject to a more complicated style of internal stirring governed by the combined influences of motions induced by flow in the overlying mantle and motions arising in response to the compaction-driven drainage of interstitial melt, which depend critically on the melt fraction. In the simplest scenario, analogous to the sedimentation of solids from a convecting slurry, this circulation may enhance expulsion of fluid from the mush. Additional important factors include the possibility of melting and freezing in different parts of the layer due, for example, to small thermal gradients, slow secular cooling at the top of the core, chemical flux to or from the core, and internal compositional stratification. We use numerical models of compaction and flow in a two-phase medium in equilibrium according to a simple binary phase diagram to better understand the evolution of porosity in a churning mush. Flow is driven by convection in the overlying mantle along with internal buoyancy forces due to phase and composition variations, with the former becoming more important when the tendency is toward a gravitationally stable stratification of the mush. Flux of light elements to or from the core is also studied by imposing composition at the lower boundary. A central aim of this work is to identify plausible sets of conditions in which thin partially molten layers can be maintained at the core-mantle boundary over time scales of order the age of the Earth.
MR31D-06 INVITED
Liquidus Temperatures in the Iron - Sulfur System and Melting of Fe3S at High Pressure
The details of binary iron-light element systems at pressures relevant to the core can be used to constrain core composition and temperature. We have conducted several experiments in the iron-sulfur system using the diamond anvil cell. Compositions of 10 and 16 weight percent sulfur were used with angle dispersive x-ray diffraction and double-sided laser heating at Sector 13 of the Advanced Photon Source. Melting was determined by the disappearance of diffraction lines associated with the liquidus phase upon raising the temperature above the liquidus, and the reappearance of those reflections upon lowering the temperature below the liquidus. Fe3S melts incongruently at 21 GPa and continues this behavior up to higher pressures. Liquidus temperatures at 10 wt. % sulfur have been determined up to 155 GPa.
MR31D-07 INVITED
Primary Ca-rich Carbonate Melts in the Transition Zone
We present new experimental and geochemical constraints on the origin of composite Ca(Ti,Si)O3 and Ca- rich majorite garnet diamond inclusions from Juina kimberlite, Brazil. The evidence reveals that the inclusions did not form as subsolidus minerals, but instead crystallized directly from calcium-rich carbonate melts during crystallization of the host diamond. Subsolidus Phase Relations. We interpret composite CaSiO3 + CaTiO3 inclusions as exsolution products from a single-phase perovskite (Pv) in the transition zone1. The MgSiO3 component in the bulk CaTiSi-Pv is exceedingly low (<0.2 mol%), unlike experimental observations of Ca-Pv coexisting with either majorite-garnet or Mg-Pv (3-7 mol%) in peridotite or eclogite2,3. Indeed, our new subsolidus phase relations show MgSiO3 increasing substantially in Ca-Pv with increasing CaTiO3- content (20-50 GPa, 2000 K). The Ca-content of the majoritic inclusions are exceptionally high (10-15 wt% CaO), also unlike in peridotite or eclogite (< 7%). Unless bizarre mantle lithologies are invoked, subsolidus paragenesis for these inclusions is effectively precluded. Melting Phase Relations. We present new experiments showing that at transition zone depths, primary melts from carbonated eclogite crystallize CaTi-rich perovskites with composition very like the inclusions, and with exceptionally low MgSiO3 (<0.2 mol%). Liquidus majorite is very calcic (10-20 wt% CaO), spanning the range of garnet inclusions. This evidence indicates that the mineral inclusions crystallized from Ca-rich carbonate melts4. Trace Element Modeling. The trace element chemistry of the inclusions as determined using SIMS techniques support a model in which the inclusions equilibrated with small-degree melts. Overall the inclusions are massively enriched in a range of incompatible trace elements, (e.g. 103 to 104 x CI in perovskite). Based on experimental mineral-melt partitioning data, calculated coexisting melts have features inherited from subducted oceanic crustal materials. We suggest a process where small-degree carbonate melts are derived from subducted oceanic crust that has foundered in the transition zone. Primary Ca-rich melts evolve through reduction of the carbonate component5,6 and crystallize diamond, which traps coexisting liquidus phases during growth. The carbonate melts are loaded with incompatible trace elements with crustally derived signatures, and may be agents of ancient, pervasive metasomatism in the transition zone and upper mantle. 1. Brenker et al., EPSL 236, 579-587 (2005). 2. Hirose et al, Nature 397, 53-56 (1999). 3. Irifune & Ringwood, EPSL 117, 101-110 (1993). Keshav et al,11th EMPG Abstracts, 36 (2006). 5. Gunn & Luth, Am. Min. 91, 1110- 1116 (2006). 6. Safonov et al, EPSL 253, 112-128 (2007).
MR31D-08
Role of Partial Melting in the Asthenosphere
The presence of partial melting in the asthenosphere has been supported by the occurrences of the seismic low velocity zone, the high electric conductive layer and the experimentally determined peridotite solidus in the presence of H2O and CO2 (e.g., Wyllie, 1988 JGR). More recently however, the significance of partial melting in the asthenosphere has been questioned based on laboratory measurements on Vs and Qs of mantle material as a function temperature and pressure (e.g. Faul and Jackson, 2005 EPSL). The recent discovery of young (ca.1Ma) alkali basalt magmatism on the 140 Ma subducting Pacific plate (Hirano et al., 2006 Science) revealed a strong evidence for the ubiquitous presence of partial melts under the oceanic lithosphere. Based on the petrologic study of the submarine alkali basalt described by Hirano et al. (2006), we estimated a primary magma composition for the young alkalic basalts near the Japan Trench (JPT-1CW; 44.3 wt% SiO2, 17.3 wt% MgO, 9.9 wt% FeO, 2.9% K2O, 1.6 wt% H2O and 1.9 wt% CO2). High-pressure melting experiments were carried out on JPT-1CW composition and we found that this magma can be in equilibrium with mantle peridotite at 3 GPa and 1400 degreeC (100 degreeC below the solidus of dry peridotite KLB-1, Takahashi, 1986 JGR). In order to study the role of partial melting under the oceanic lithosphere, experiments were carried out using a mixture of 97 wt% peridotite KLB-1 and 3 wt% JPT-1CW in the pressure range between 2.2 and 3.5 GPa and the temperature range between 1250 and 1500 degreeC. Typical run durations are 100 hrs for low- temperature and 20 hrs for high-temperature runs. Experiments were conducted using piston-cylinder apparatuses with talc-Pyrex-graphite assembly. Two types of sample containers (graphite-Pt and Re-Pt) were used and experimental results were duplicated for most P-T conditions. Three melting regimes were recognized in our experiments; 1) within the stability of phlogophite (lower than 1300 degC) neither signature of melting nor that for mobilization of elements were detected, 2) between 1300 and 1400 degC, almost all potassium has been lost from the peridotite matrix by migration of H2O-rich fluid, 3) at higher than 1400 degC, K-rich partial melts were formed and distributed throughout the peridotite matrix. Although the dry solidus for the peridotite KLB-1 varies from 1420 degC at 2.2 GPa to 1600 degC at 3.5 GPa (Takahashi, 1986 JGR), the three melting regimes for the wet peridotite are insensitive to temperature in the studied pressure range. Based on our experiments, we propose a model for the role of partial melting in the asthenosphere; 1) The upper bound of partially molten asthenosphere is most probably controlled by inflected peridotite solidus due to the stabilization of carbonate and breakdown of phlogophite at around 3GPa and 1300 degC. 2) In the asthenosphere, super-critical fluids promote segregation of alkali elements from deep mantle regime. 3) Oceanic plate would act as an impermeable lid for the asthenosphere and a thin melt-enriched layer would be produced beneath the oceanic lithosphere after the long term melt segregation. 4) This partial-melt layer may act as a slip-plane and would define the mechanical thickness of oceanic plate. 5) Ubiquitous occurrence of K-rich metasomatism at the keel of the lithosphere supports the existence of such partial-melt layer. Therefore, partial melting in the asthnosphere plays fundamentally important roles both in geochemical evolution and geodynamics of the Earth.