U14A-01
Mechanisms for 3-D Flow in the Mantle Wedge at Subduction Zones
Shear-wave splitting measurements from local events at volcanic arcs show strong variability in the orientation of seismic anisotropy, with certain regions characterized by trench-parallel anisotropy and other regions displaying more variable fast polarization directions. This pattern of anisotropy is inconsistent with simple models for corner flow and suggests a more three-dimensional (3-D) style of convection in the mantle wedge. We explore the origin of 3-D flow driven by gravitational instabilities arising in both lower arc crust and the subducting plate. At conditions appropriate for arc lower crust (800°C-1000°C; 1 GPa) many crustal assemblages are denser than the underlying mantle. Such high-density layers can become gravitationally unstable and sink into the mantle with a characteristic along-arc spacing of 30-50 km on timescales of ~106 years. 3-D finite element calculations show that "foundering" of this material can produce regions of trench-parallel flow in the mantle wedge, consistent with seismic observations and fabrics in residual peridotites at the base of the Jurassic Talkeetna arc section in south central Alaska. The initiation of instabilities will likely occur between active volcanoes where igneous crust cools to <850°C near the Moho. Instability growth in these locations will in turn drive upwelling and melting in the intervening regions. This process could persist through time, resulting in alternating periods of magmatic activity and quiescence at arc volcanoes. An analogous process that can produce similar flow patterns in the mantle wedge and localization of arc volcanism, is upward transport of low density diapirs composed of subducted sediment, serpentinite, and/or partial melt rising from the slab. In summary, the observation of trench parallel anisotropy at volcanic arcs, suggests that the composition of the mantle wedge may be modified by a combination of foundering of arc lower crust and hydrous diapirs rising from the subducting slab.
U14A-02
Constraints on Structure and Melting of Heterogeneous Plumes From Laboratory Experiments With Three Components
Many studies of chemical geodynamics consider the fate of a single, compositionally distinct layer at the base of the mantle, but subducted oceanic lithosphere introduces two distinct lithologies (higher-density eclogite and lower-density harzburgite) into the mantle (a third lithology, intermediate-density lherzolitic peridotite). To address the dynamic complexities of interactions between these materials, we conducted three-dimensional laboratory experiments that use glucose syrup (Rayleigh number: 106\)) to model the mantle and a two-layer subducted lithosphere. The viscosity and density of the syrup are controlled by its water content, which is varied to simulate the distinct physical properties of each of the three lithologies. Experiments were conducted in a 20cc tank, heated from below to create a basal thermal boundary layer (BTBL). The two-layer glucose slab was frozen and placed within the tank, where it sank into the BTBL. These experiments produced heterogeneous upwellings with temporal and spatial variations in both temperature and composition that are much more complex than predicted by classic plume theory. Temperature, composition, and distribution of material in the tank through space and time were recorded during each experiment. We scale these data to mantle-equivalent conditions and address the observational implication for melting such heterogeneous plumes, both within larger (200 - 600 km) plume heads and smaller (<100 km) trailing conduits. Results show length scales of chemical heterogeneity range from <10 km up to 300 km. Thermal heterogeneity was often correlated with composition, where the denser, eclogite analog had higher temperatures than the lighter, harzburgite analog. Distinct domains form within plumes and melting begins at different depths, dependent on the temperature and composition of each domain and the solidus of each composition (e.g. eclogite melts at lower temperatures than harzburgite). The combination of thermo-chemical variation and differences in solidi explains heterogeneous melting behavior. In many cases, whole portions of buoyant plume heads either never reach the dry solidus, or predict very small melt fractions, and primarily serve as pathways for the ascent of subsequent fusible material.
U14A-03
Flow and Melting in a Heterogeneous Mantle: A Case Study from the Galápagos Spreading Center
The Galápagos Spreading Center (GSC, ~83-98°W, 1-3°N) in the Eastern Pacific Ocean interacts with the nearby Galápagos hotspot (~92°W, ~0.5°S). This is manifested in both physical (e.g., axis morphology, crustal thickness) and chemical variations along the ridge. Lavas recovered along the ridge closest to the hotspot have geochemical characteristics distinct from more distal lavas, as well as from lavas of non-hotspot-influenced segments of other mid-ocean ridges. Within a 500 km radius of the hotspot (e.g., between 92°W and 95.5°W), regular patterns are observed in crustal thickness, axial morphology, and some incompatible trace element ratios, whereas radiogenic isotope compositions show weak to no correlation with increasing distance from the hotspot. The largest variability in several geochemical indicators along this portion of the ridge is observed over an along-axis distance of ~10 km near a propagating ridge tip (95.5°W). West of this ridge offset, evidence for any geochemical influence by the hotspot is completely absent. Here, we evaluate three possible models that may explain observed geochemical and geophysical variations along the ridge. These models include: 1) a ridge fed by a narrow, geochemically distinct mantle plume conduit that becomes progressively diluted by ambient mantle farther from the plume (e.g., Schilling et al. G-cubed 2003), 2) a narrow, geochemically distinct plume that does not mix with the ambient mantle over the distance it feeds the ridge (e.g., Ito et al. JGR 1997), and 3) a broad region of active upwelling of a uniformly heterogeneous mantle (Ito and Mahoney EPSL 2005a,b). The high geochemical and geophysical data density along the western GSC provide an opportunity for a robust case study of the predictions of these three models.
U14A-04
Mixing the mantle marble-cake: timescale constraints from Os isotopes
In their seminal paper, Allegre and Turcotte (1986) presented a model in which the upper mantle is a mixture of depleted, harzburgitic mantle and subducted basalt that has been mechanically mixed together, the mantle marble-cake. Since their publication, most studies of mantle heterogeneity have focused on the enriched components, which are equated with subducted basalt and/or sediments, and successfully explain OIB Sr-Nd-Pb isotopic systematics. In this talk, we will focus on a different part of the marble-cake, depleted (harzburgitic) heterogeneities. Though abundant in abyssal peridotites and ophiolites, these have been difficult to study geochemically because they have very low concentrations of typical trace elements and radiogenic isotopes, and are overprinted by any mixing with enriched mantle or melts. However, Os is compatible during mantle melting, is enriched in depleted mantle and thus is robust with respect to mixing with enriched components or metasomatism. Somewhat surprisingly, Os isotope studies of the convecting mantle show clear evidence for depleted heterogeities up to 2 billion years old, but the relative paucity of data (less than 100 analyses), makes it difficult to extract meaningful mixing information. Rapid analysis of osmiridium grains by laser-ablation inductively coupled multi-collector plasma mass spectrometry now allows large Os datasets to be acquired (100s of datapoints), which are suitable for statistical analyses (Meibom, 2002). Here we present new and published laser-ablation analyses of osmiridiums from a global collection. The data generally show an exponential decrease in heterogeneities with age, such that over 90% of heterogeneities are destroyed within 2 billion years, though rare heterogeneities as old as 2.7 Ga survive. The exponential decrease in survivorship is generally consistent with the mechanical mixing model of Allegre & Turcotte (1986). Subsequent 2-dimensional mixing models suggest that high-viscosity blobs can persist for much longer time-scales than 2 Ga (Manga, 1996). Thus the relatively fast mixing suggested by the Os data may imply that 1) the depleted heterogeneities have similar viscosities to their surrounding mantle, 2) 2D models overestimate mixing timescales or 3) other processes, such as melt infiltration, may destroy depleted heterogeneities.
U14A-05
Preservation of Fertile Mantle Components at Mid-Ocean Ridge
Recycled lithosphere is suspected to contribute to the geochemical enrichment not only of Ocean Island Basalts (OIB) but also exceptional Mid-Ocean Ridge Basalts collected at ultraslow ridges. In particular, the chemistry of basalts at volcanic centers along the Southwest Indian Ridge oblique supersegment of 9-16°E is best explained by an exceptionally strong contribution from an enriched mafic component mixed in the upwelling mantle. Why this component is so strong at these volcanic centers can be understood in at least two ways. 1) The mantle underneath each volcanic center is anomalous. Such an explanation is equivalent to appealing to small plume-like features underneath each volcanic center. 2) The fertile component is present everywhere but melt migration gathers the resulting magma toward the volcanic centers. In this contribution, we test the second hypothesis using a numerical melt migration model in which magma rises vertically until it encounter the base of the thermal lithosphere. In the SWIR 9-16°E area, variations in ridge axis azimuth produce a strong relief to the base of the lithosphere, which focuses magma towards the location of the observed volcanic centers. Magma produced off-axis, which is dominated by the fertile component, is focused even more strongly than near-axis magma, explaining the relative enrichment of the surface lava. We compare the expected enrichment pattern with the geochemistry of collected lava and show that, were the ridge straighter or spreading faster, this signal would be more difficult to observe.
U14A-06
Global Variations in Abyssal Peridotite Compositions I: Isotopic and Trace Element Characteristics of the MORB-Source Mantle
Abyssal peridotites are typically assumed to be the residues of near-fractional melting of an initially uniform source composition. We use a global compilation of abyssal peridotites to assess the extent to which these assumptions are true. After filtering the dataset for samples with characteristics of melt addition (e.g., veins, high spinel TiO2 concentrations), we observe considerable variation in nominally residual peridotites. As a function of spreading rate, peridotites at slow and ultra-slow spreading rates extend to more depleted and more enriched compositions than can be explained by a simple near-fractional melting model. Globally, trace elements in residual peridotite Cpx vary by 2-4 orders of magnitude over length-scales ranging from the grain-scale to the ridge segment-scale. Similarly, variations in spinel Cr# and peridotite modes occur at all length-scales. Variations at the sub-dredge scale cannot be the result of variations in the pressure-temperature regime during melting. We suggest instead that these variations reflect various melt-rock reaction processes during melt extraction. In addition, we suggest that some of the variation in peridotite composition reflects initial source heterogeneity. To constrain the role of source heterogeneity, we assess the isotopic range of abyssal peridotite Cpx. Using a global compilation (70 samples), we find that the average Nd isotopic composition of peridotite Cpx (0.513167) is similar to average MORB (0.51313; Su and Langmuir, 2003). Peridotites are observed to extend to similarly enriched isotopic compositions as those of normal-MORB, but to more depleted compositions than MORB. These depleted isotopic compositions should be accompanied by depleted elemental concentrations and thus be relatively undetectable in MORB. We interpret the range in peridotite isotopic compositions to reflect initial source chemical heterogeneity. Hence, the upper mantle is both initially heterogeneous before the onset of melting and undergoes processes that create further heterogeneities during melting and melt extraction.
U14A-07
Producing Ni-rich olivine phenocrysts by mixing partial melts of eclogite and peridotite: an alternative to an olivine-free source for Hawaiian shield basalts
It has been posited that presence of unusually Ni-rich (2500-4000 ppm) magnesian olivine phenocrysts in SiO2-enriched Hawaiian shield-building basalts, most notably the Koolau lavas, is inconsistent with a deep, olivine-bearing source rock. Instead, Sobolev et al. (2005) proposed that these lavas are generated by a multi- stage process in which partial melts of eclogite react with peridotite within the plume to form an olivine-free source rock with high Ni concentration. As the plume continues to ascend, partial melts of this "hybrid" pyroxenite mix with peridotite melts to produce SiO2- enriched Hawaiian shield-building lavas that crystallize high-Ni olivine. This model has also been used to argue for significant amounts of "hybrid" pyroxenite in the source regions of lavas from other ocean islands, continental basalts, and even MORB, implying that the upper mantle is highly heterogeneous (Sobolev et al., 2007). New experimental results demonstrate that Ni-rich magnesian olivine crystallizes from mixtures of peridotite partial melt and Ni-poor eclogite partial melt that have equilibrated with mantle olivine. This occurs because the concentration of Ni decreases linearly as eclogite partial melt is added to peridotite partial melt, whereas changing major element composition of the mixed melts causes DNi to increase hyperbolically. Experiments were conducted in which either (1) siliceous partial melt of eclogite or (2) primitive basalt was equilibrated with San Carlos olivines at 1 bar and 1201-1350°C. Experimental results demonstrate that eclogite partial melts in equilibrium with mantle olivine retain their high SiO2, low FeO and MgO characteristics. Theoretical modeling calibrated from these experimental results suggest that reaction of siliceous eclogite melt with mantle olivine at low pressure produces a melt containing ~300 ppm Ni. Despite its low Ni content, mixing of this melt with peridotite partial melt produces a high SiO2 melt that crystallizes Ni-rich, magnesian olivine. The dependence of olivine-melt partition coefficients on melt composition also explains the enrichment or depletion of other minor element in Koolau olivines (e.g., Ca and Mn). Our results obviate the need for a multi-step melt generation process in which reaction with large amounts of siliceous eclogite partial melt exhausts olivine from portions of mantle peridotite within the Hawaiian plume. As a result, the amount of eclogite required to explain the composition of the Koolau lavas is greatly reduced. Reference Sobolev, A.V. et al., (2005), Nature, 434, 590-597 Sobolev, A.V. et al., (2007), Science, 316, 412-417