U21A-0001
Seismic Evidence of Subducting Slab for Whole Mantle Convection Versus Layered Convection
Recent seismic tomography models provide variable implications for mantle convection, i.e., the subducted slab flattens in the upper mantle and broadens that was visualized as high velocity anomaly (HVA) with a lateral extent of over a few thousand km beneath the northwestern Pacific ocean while a HVA zone appears to penetrate into the lower mantle associated with the Java subduction zone. Our analysis of broadband body waveforms has shown fairly broad regions where subducting plate is lying flat or piling up within the transition zone in the northwestern Pacific that is in agreement with tomography studies. However, we also found that the total volume of stagnant slab is considerably less than the image in the tomography model and much smaller than that of subducted slab during the known subduction history. The results imply that the stagnant slab moves somewhere else out of the transition zone or further descends into the lower mantle. Thus, the production of stagnant slab does not preclude the possibility of whole mantle convection. Features implied from stagnant slab along with other seismic evidence for subducted oceanic plates in the lower mantle suggest that convection is taking place in the whole mantle. On the other hand variation of the discontinuity depth at the bottom of the transition zone has been interpreted as variable distribution of geochemical properties and fluids associated with stagnant slab. Mantle materials descend into the lower mantle through the 660 km phase boundary only given gravitational instability due to thermo-chemical anomalies presumably associated with cold slab. The descending subducted slab into the lower mantle does not sufficiently support the convection patterns that are taking place in the entire mantle. The volume that involves in whole mantle convection may be a fraction of the entire mantle. Otherwise the layered structure is distinct in the mantle and there may be separate convection cells in the upper and lower mantle.
U21A-0002
Interpreting Seismic Constraints on 1-D Thermo-Chemical Structure of the Mantle Transition Zone: Implications for Mantle Dynamics
One-dimensional seismic reference models such as PREM and AK135 form the initial and background models for almost all inversions for three-dimensional seismic Earth structure. For quantitative interpretation in terms of physical parameters, i.e. temperature and composition, it is generally assumed that the mantle's seismic reference corresponds to an isochemical adiabatic structure produced by whole-mantle convection. However, tests have shown that the simplest mantle model - adiabatic pyrolite with a (MORB-formation consistent) potential temperature of 1300C - is not compatible with global seismic data sets. The discrepancy between the model and the seismic data could have resulted either from inappropriate use of mineral physics to generate the model, or because real average Earth structure is significantly different from this simple thermo-chemical structure. We test both these possibilities by generating a set of alternative 1-D thermal and chemical mantle models, and incorporating the effects of large uncertainties in both the elastic/anelastic parameters of the constituent minerals, and the thermodynamic procedures for calculating seismic velocities, into our computations. Although mineral physics uncertainties translate into substantial variation in seismic behaviour, there is subtle evidence to suggest that there are alternative thermo-chemical models which, either via a temporary shift to lower temperatures, and/or a change to a seismically faster chemical composition, in the transition zone and uppermost lower mantle, provide a significantly better fit to the seismic data than adiabatic pyrolite. This is compatible with average structures produced by thermo-chemical whole-mantle convection models from Tackley et al. (2005), with offset phase transition depths for the olivine and garnet components of the mantle's chemistry. In such models, average thermal structure is close to adiabatic, but average chemical structure departs substantially from pyrolite, as (seismically fast) basalt pools above 660 and (also seismically fast) harzburgite is enriched below. Our data show that such a complex physical background structure is likely. This needs to be considered when quantitatively interpreting models of seismic anomalies, especially when anomalies of different wave speeds are combined.
U21A-0003
Constraints on Upper Mantle Temperature from Seismic Attenuation
Imperfections in the crystalline structure of any mineral cause non-elastic effects, which in turn lead to dissipation and dispersion of wave propagation at seismic frequencies. The mechanisms are analogous to the creep (or deformation) mechanisms occurring at much lower frequency and are thermally activated. Owing to the advances in experimental techniques, it is now possible to have reliable laboratory measurements of seismic attenuation (Q) at temperatures (T) and pressures (P) corresponding to the uppermost mantle. The available data can be represented with appropriate P, T and grain-size (GS) dependent models. Extrapolation to high pressure is uncertain and relies on mostly unconstrained values of activation volume or on an empirical homologous temperature approach. We test the experimental Q models against available compilations of attenuation measurements. These include surface waves and free oscillations fundamental modes, free oscillation overtones and longer period fundamental measurements. Despite the uncertainties in seismic observations (e.g., the deviation between surface waves and normal mode measurements) and in the mineral physics parameters (e.g., the pressure dependence), available knowledge of anelastic behavior already puts a tight constraint on absolute average temperature (or grain size) and its gradient with depth in the first 400 km of the mantle. The same experimental P, T and GS dependent Q model, coupled with elastic data of mantle minerals, are used for the interpretation of long-period seismic waveforms. In a previous study, we found that long period seismic waveforms require, globally, an increase in shear velocity between 250km and 350km that is compatible with either a gradual enrichment in garnet with depth, or a negative thermal gradient, or an increase in grain size. Although a negative thermal gradient or an increase in grain size (or a combination of both) can explain the isotropic features without invoking a compositional change, both factors affect the Q model significantly. The resulting Q profiles obtained with all available combinations of elastic and anelastic models do not fit the attenuation data well, while observations are generally consistent with an adiabatic thermal gradient and a constant grain size within this depth range. This supports the possibility of a compositional change with depth throughout the upper mantle, which can explain the isotropic profile without affecting the Q profile.
U21A-0004
Influence of Rheology on the Lengthscale of Mantle Convection
In the debate of whole vs. layered mantle convection much attention has been paid to the vertical extent of convection planforms. Closely connected and potentially equally important in this discussion is the horizontal lengthscale of mantle convection. Studies that have addressed the horizontal lengthscale of mantle convection have found that prescribing a shallow low viscosity channel in convection models can be crucial for promoting long wavelength flow. However, work involving self-consistent mantle models, in which viscosity is a function of temperature, depth, and yield stress, have not readily observed long horizontal wavelengths. In this work, rather than prescribing a low viscosity channel, we generate one in numerical mantle convection models by using a temperature and pressure dependent rheology. We explore which rheologic ingredients are necessary for a shallow low viscosity channel to develop, and which degree of "channeliness" facilitates long wavelength convection and plate-like motion. Our numerical experiments may help constrain which mode of mantle convection is dominant in Earth's mantle.
U21A-0005
Whole Mantle Thermal Convection At The High Rayleigh Number
Rayleigh number for mantle approximately in 10000 times exceeds critical value, the order 103. Hence, according to the theory convection, the mantle not simply is in a condition convection instability, convection in it has intensive character. Chaotic time dependence is easily raised in a liquid heated up from within as it is necessary to bear heat from bowels of cells. The similar behaviour also takes place at heating from below, at enough high Rayleigh numbers. Transition from convection with chaotic time history, but stationary enough cells (so-called poorly turbulent mode) to convection with not organized picture of cells (so-called strongly turbulent mode) takes place at infinite Prandtle number with heating from below at Ra=107 or at lower Rayleigh number if takes place internal heating. Rayleigh numbers for mantle convection in a fluid layer heated from below Raq=4.8×106 and in a layer, heated from within RaH=3×109. These values in 104 and 106 times more critical Rayleigh number. Rayleigh number for single-layered mantle convection based on superadiabatic differences of temperatures between a surface and core-mantle boundary has the order 107, approximately in 104 times more critical. In the geological past, in particular in early Archean, Rayleigh number for the Earth was probable even more as the mantle was more hot, and, hence, less viscous, than now. For Rayleigh number corresponding single-layered whole mantle convection at heating from below the initial stage of development non-stationary mantle convection is investigated. As algorithm it was applied so-called Semi-Implicit Method for Pressure-Linked Equations (SIMPLE). Calculations were spent in square and rectangular areas. For the account of real parities of the vertical and horizontal sizes of the mantle, and also for exception of influence of lateral borders, modelling was spent in rectangular area with the aspect ratio 10:1 on a grid 502×52 points. Bottom and lateral borders rely impenetrable without sticking. The top border is considered free. Temperatures on the top and bottom border are accepted fixed and equal, accordingly, T0 and T1, and on lateral borders equality is necessary to zero of a thermal stream. The bottom border corresponds to a surface of the core. The top border are associated with oceanic lithosphere. As initial temperature distribution it is accepted corresponding weak one-cellular convection. Modelling covers an interval of dimensional time from 0 up to 670 million years. Was it is calculated 2500 planforms convection (on number of steps on time). As a result of the lead researches it is established, that in the given system arises unstable convection therefore collapse convection cells and on their place there are numerous ascending and descending plumes which, reaching a surface, spread in a horizontal direction. The descending and ascending jets of streams arising as a drop, move horizontally, and some nearby streams are united in one, more powerful. Two kinds of plumes it was observed in model: developed plumes stretched through all depth of the layer, having rather steady position, and fine mobile plumes drawn to greater subvertical plumes. The heat transfer strongly depends on the wavelength of convection. Not looking at that movement takes the form of casually arising plumes, inversion of temperature arising at the first stage (owing to a choice of entry conditions) in many respects defines character of process at an initial stage. The benthonic part well transfers possible character of the processes occuring on core-mantle boundary. Presence of numerous descending streams, apparently, is caused by isoviscous conditions.
U21A-0006
Illuminating Slab Remnants in the Lower Mantle Using PKP Precursors
High quality recordings at the Yellowknife seismic array (YK) provide a good opportunity to locate the scatterers responsible for some of the observed PKP precursors, specifically those that appear well separated on the seismograms. Using data from high quality earthquake doublets, we have shown (Cao and Romanowicz, EPSL, 2007) that it is possible to precisely locate the scatterers responsible for these precursors, and determine whether they originate from the PKPbc or the PKPab phase, and on the source or receiver side. We have now expanded this dataset to non-doublet events. The most reliably measured individual scatterers delineate a region in the lowermost mantle where tomographic models indicate the presence of a transition from fast to slow velocities, likely associated with subducting slabs. To produce the scatterers, this transition should be sharper than seen in large scale tomographic models. This provides evidence for the penetration of slabs deep into the lower mantle beneath western North America and for the persistence of heterogeneity in composition during the slab's slow descent to the core-mantle boundary.
U21A-0007
Formation of the Oceanic Lithosphere from the Upper Asthenosphere
In a global examination of the chemistry of MORBs, we find that Na8-Fe8-axial depth data do not support large variations in the temperature and pressure of MORB extraction from the mantle. Instead, the complete absence of high-pressure (> ~1.5 GPa) olivine-controlled crystallization of MORBs combined with solidus phase relations in the CaO-MgO-Al2O3-SiO2-Na2O-FeO system indicate that the inverse and positive Na8-Fe8 variations are produced from a heterogeneous source by melt extraction over a very narrow range of P and T (~1.2-1.5 GPa and 1250-1280°C) at the plagioclase-spinel lherzolite transition. This is inconsistent with the existence of hot mantle plumes (Easter, Galapagos, Iceland, Azores, St. Helena, Tristan, Afar) on or close to ridges. However, it is consistent with the very flat 410 km discontinuity beneath the East Pacific Rise, which does not permit the existence of even a single hot plume (Easter) beneath the ridge (Melbourne and Helmberger, 2002, JGR, 107, doi:10.1029/2001B000332). The global absence of MORBs with a high-pressure major-element signature implies that the isolation of the East Pacific Rise from the deeper mantle applies to all ridges. A new model is developed (Presnall and Gudfinnsson, in press, Origin of the Oceanic Lithosphere, J. Petrol.) that explains the formation of new seismic lithosphere (~70 km thickness) by lateral and upward migration of the slightly melted upper part (~70-140 km depth) of the low-velocity zone toward the ridge. Although decompression melting occurs over a large pressure range, melt extraction is constrained to the very narrow P-T range given above by the maximum T in the mantle at which CO2 vapor can be extracted. This condition occurs at a pressure just below that of the abrupt 280°C temperature decrease of the carbonated lherzolite solidus at the base of the seismic lithosphere. The constant association of strombolian and effusive eruptions at ridges (Clague, 2007, Geophys. Res. Abstr., 9, EUG, 02096) supports the view that explosive escape of CO2 from the upper part of the seismic low-velocity zone is the vehicle that facilitates transport of MORB melts to the surface. In this model, a new ridge is born when stresses on the lithosphere cause fracturing that penetrates the entire lithosphere. This allows explosive flashing of carbonate-rich melt to CO2 vapor, which escapes to the surface. Erupted melts carried with the CO2 vapor are initially low- volume carbonatitic melts that probably exist on the ocean floor ahead of propagating rift tips. These melts change progressively to alkalic basalt, and finally to MORB tholeiite as the ridge matures. The depleted upper asthenosphere rises and becomes new lower lithosphere that moves away from the ridge axis with the new crust. When changing lithospheric stresses shift fracture formation to a different locality, melts change back through alkalic basalt to carbonatite, venting of CO2 shuts down, and the ridge dies. In support of this modeling, alkalic lavas are found at the southern tip of the southward-propagating Eastern Volcanic Zone in Iceland. Carbonatitic melts farther south are unknown but would be hidden from view on the ocean floor. http://petrology.oxfordjournals.org/papbysection.dtl
U21A-0008
Partially layered mantle convection and Earth's thermal history
Recent seismic tomographic images indicate that slabs penetrate through the 660-km depth phase boundary in some locations while they are deflected in their trajectory in others, suggesting that convection in Earth's mantle is currently partially layered at this depth. The efficiency of heat transport by convection decreases with increasing layering as heat must be carried by conduction across an internal thermal boundary layer. There is a long-standing problem concerning Earth's thermal history in that the heat flow at Earth's surface is significantly greater than what is thought to be the internal heat generation by radioactive decay. It has been suggested that layered convection may allow significant quantities of thermal energy to be stored in the lower mantle to be released later as the degree of mantle layering decreases over time which could allow for the observed high surface heat flow with relatively low internal heating. In this contribution, we simulate the thermal history of the Earth using a spherical axisymmetric numerical model of convection in Earth's mantle coupled with a parameterized model for the thermal evolution of the core. Layering is induced in the model by the effects of an endothermic phase transition at 660-km depth with varying magnitudes of the Clapeyron slope. We find that although the mean degree of layering increases with an increasing magnitude of the Clapeyron slope of the endothermic phase transition, the time-averaged thermal evolution is very similar to a whole-mantle model unless the magnitude of the Clapeyron slope of the phase transition is unreasonably large.
U21A-0009
Thermo-Petrologic Evolution of Slabs in the Transition Zone Controls Deep Subduction
The interaction between subducted lithosphere and the transition zone of the mantle is a key issue in assessing the scale of mantle convection. However, interpretations of seismic tomography seem to indicate a puzzling mix of slab-like anomalies in the lower mantle as well as large-scale remnants of slab within the transition zone. High- resolution studies of the Tonga subduction zone, where the strongest thermal anomaly associated with subduction is expected, reveal that an anisotropic petrologic anomaly exists at the deep leading-edge of the Wadati-Benioff zone (WBZ) over distances of about 300 km. Strong radial anisotropy (~1%) and lower than expected seismic wave speeds (~3%) are evident from analysis of triplicate P and S-waveforms recorded by the CAVASCOPE broadband array. Similar evidence for a petrologic anomaly also exists over a length-scale of about 1,000 km in a seismically active, detached remnant of slab immediately to the west of the WBZ. Surrounding the anisotropic region of the remnant slab, elevated seismic wave speeds gradually diminish laterally, indicating that the petrologic anomaly is the center of a large, diffuse thermal anomaly. The only candidate for the petrologic anomaly that satisfies all available observations is metastable olivine -- a buoyant material when present in the transition zone and therefore acting as a barrier to slab penetration. On a global scale, similarly rapid subduction of cold lithosphere is pervasive in the western Pacific where large-scale remnants of slabs are common in the transition zone. We interpret these observations as a natural consequence of how lithosphere sinking into the transition zone generates a buoyant thermo-petrologic anomaly that evolves over time into a negatively-buoyant, pure thermal anomaly. As the slabs evolve, the effect of buoyancy from the petrologic anomaly would be important for retaining subducted material as subhorizontal remnants which, in turn, would continue to serve as an effective heat sink, lowering temperature over broad regions
U21A-0010
Tungsten abundance and isotopic compositions of Ocean Island Basalts, an Oceanic Plateau, MORB samples: in search for a core-mantle interaction
Core-mantle interaction has been investigated using several geochemical tracers, which can put strong constraints on mantle convection of the Earth. To elucidate such interactions, siderophile elements that are highly concentrated in the metallic core rather than in the silicate mantle are the most favorable candidates to provide geochemical evidence. Platinum-group element (PGE) abundance and Os isotopes of oceanic basalts can work to deconvolve the contribution of core material. Hf-W system is another tracer to elucidate core-mantle interaction. Tungsten is a moderately siderophile, incompatible, and refractory element that partitions preferentially into the Earth's core during core formation. Collerson et al.(2002) reported a negative anomaly of the W isotope ratio in kimberlite of South Africa, and suggested the possibility of a core-mantle interaction. However, Scherstén et al.(2004) retested the samples of kimberlite of South Africa and Hawaiian picrites, observing no anomalies in W isotope ratio. Whether the inconsistent results were caused by different sample treatments or analytical difficulties in W purification remains unclear. In this study, we will report W abundances with isotopic compositions of OIBs, Ontong Java LIP and MORB. The W abundances in the OIBs analyzed in this study show a large variation: OIBs(South Polynesia Island, Kilauea, Loihi)=200-900ppb, OJPs=20-500ppb, MORB(Indian and EPR)=20-90ppb. Tungsten isotopic compositions were measured using multi-collector ICP mass-spectrometry. We examined the reproducibility of W isotopic analyses from 3–4 repeated measurements of four samples of OIBs and estimate the external reproducibility of isotope analyses as 0.1-0.27 epsilon unit. The analyzed OIBs have high tungsten abundance, but their tungsten isotopic composition shows no negative anomalies as indicates core-mantle interaction. The samples with high W abundance from OJPs do not have negative tungsten anomalies, either. These results suggest four possibilities about mantle convection and core-mantle interaction. 1) These sources of OIBs or LIPs analyzed, do not originate from CMB. 2) Negative W isotope ratio anomaly of the core has been later diluted with the entrainment of mantle material with normal W isotope composition. 3) W does not diffuse from outer core into mantle as has been pointed out by Humayun et al. (2004).
U21A-0011
Errors in modeling contraction, isostatic compensation and rock-mass balance in conductive cooling models of the oceanic lithosphere: Implications on Earth's power, radioactive contents, and convective style
High values of global power (44 TW) are derived from cooling models of the oceanic floor, which find support through their alleged prediction of seafloor depths. However, a factor of 3 error exists in equations used to predict depth as a function of seafloor age. These models allow for contraction only in z, which is governed by linear thermal expansivity (αl), yet volumetric expansivity (αv= 3αl) has been erroneously used. Many additional problems exist in half-space and plate models: (1) Isostacy was considered, but rock-mass was not conserved, which means that subsidence as modeled, involves changing rock to water. (2) Contraction is based on infinitely high columns in the half-space model. (3) The half-space is assumed to start at z = dw, which means that T(z) should depend on dw but all such terms have been dropped. (4) Constant surface and constant basal temperatures are assumed, so the average temperature of the lithosphere cannot possibly change: thus, cooling or contraction is not permitted by model assumptions: the calculated contraction instead results from all these errors. These faulty 1-d models fit globally averaged depths only for seafloor ages < 60 My and only if basal temperature is ca 4000 K using reasonable values for the various physical parameters. This value is incompatible with mantle petrology, and predicts an excessive global power of >130 TW. Neither seafloor depths nor heat flux are represented by 1-d conductive cooling models. Instead of originating in contraction, subsidence actually results because mid-ocean ridges overlie hot mantle upwellings at the edge of convection cells, whereas old oceanic floor overlies colder mantle over the flanks of the convection cells. We show that upper mantle temperatures govern seafloor subsidence, suggesting layered convection.. Moreover, recent assessment of heat flow data, corrected for duplications and errors in location, limits global power to between 31+/-3 TW. This value is consistent with an enstatite chondrite model of the bulk silicate Earth, and does not require K or other substantial heat sources in the core. Known sequestering of most of the radioactives in the crust efficiently expels heat. Quasi-steady state is consistent with stratified convection.