U11B-01 INVITED
A two Layer Convecting Mantle With Exchange : A Unified Model Based on Geochemical, Seismic and Heat Flow Observations
The question of layered versus whole mantle convection has been pending since early models of mantle convection (Oxburgh and Turcotte, 1967; McKenzie and Richter, 1978). In a first phase, layered mantle models appeared to be the winners, because they explained most of geochemical observations. They of course also explained the seismic focal solutions (compressive/ non compressive) in subducting plate and the heat flow constraints (Richter, 1980). Later on, the discovery that slabs could penetrate the lower mantle was considered as proof of whole mantle convection. Most numerical experiments were then developed in this context, overlooking many geochemical observations such as rare gas isotopes or radioactive elements. In this presentation, we will examine the different data one by one. a) geochemical constraints include the budget equations for Sr, Nd, Hf, isotopes, the budget equation for He, Ne, Ar, the budget for heat producing elements U, Th, K (including Th/U and K/U ratios). (Allègre and al., 1979, 1982; De Paolo and Wasserburg, 1977; O'Nions and al.,1977). b) Constrains linking geochemical observations and geodynamics. How the so-called depleted mantle is generated ? Continuities and affinities between MORB and OIB, including the Dupal and Non Dupal provinces: The non-pristine source for OIB based on Pb isotope data. The marble cake structure for upper mantle. Continental crust recycling via sediments and delamination processes. In thise respect, we emphasize the difference in statistical variance of isotope or trace element ratios in the different/types of basalts reflecting the difference in stirring intensity in their sources. We also use information from extinct radioactivities 142Nd and 129Xe. In each case, we estimate the errors for measurements and models. c) The seismic evidence of slab penetration into the lower mantle are from Creager and Jordan(1984) to Van der Hilst et al.(1991), Sparkman and al.(1993), with the counter example of non-penetrating slabs as emphasized by Fukao and al.(2001) and the recent observation of the large energy spectrum differences at 670 km depth (Gu and al., 2006). We discuss the problem of return flow, which is crucial for both energy budget and convection regime. The recent work on plume by Montelli and al. (2004, 2006) shows the existence of broad plumes in the lower mantle and thin plumes in the upper mantle. d) The estimate of heat flow coming from the lower mantle of 35-32 TW. The work of Davies(1990) and Sleep(1992) shows clearly that this transfer is not the result of plumes reaching the surface, because they correspond at most to 3TW. At the reverse the estimated heat flow carried by the lower mantle plumes is much higher (Nolet and al., 2006). We also discuss the heat flow paradox to explain a Urey ratio of 0.4 with whole mantle convection. In conclusion, we propose mantle with two layers convecting separately but with some exchange of matter, this global exchange corresponding to 1.1024kg since 4.4 Gy. Plume genesis is a two-stage process. Lower mantle plumes heat the Mesosphere boundary layer generating second generation plumes which reach the surface (Allègre and Turcotte; 1983; Allègre, 1987). In the upper mantle itself, we have to distinguish between a vigorously convecting asthenosphere and a sluggish convecting transition zone, both convecting in same cells.
U11B-02 INVITED
Layers in the mantle - dynamics and structure
The answer to the question whether the Earth's mantle is dominated by whole or layered convection depends in large part on the interpretation of the signals that the observational sciences provide. Seismology provides a snapshot of the current structure with clear evidence for structural layering in the mantle. Geochemistry provides a long term integrated look into the composition of the Earth with strong suggestions for both significant mixing and long term sequestration of isotopic heterogeneity. Forward modeling of mantle convection can be used to test under which dynamical conditions we can recreate the present day structure and long term integrated views that seismology and geochemistry provide. In addition, basic physical observations such as heat flow, plate velocities, geoid and topography, provide alternative constraints on the structure of mantle convection. It appears that strict mantle layering at 660 km depth can be ruled out based on dynamical considerations alone, but the question remains how strong the mass fluxes are from the upper mantle into the lower mantle. I will use existing and new geodynamical models of mantle convection to estimate the extent to which layering can be generate by physical processes, such as compositional variations, phase changes and pressure- dependent rheology, that still provide reasonable predictions for surface velocities and heatflow. I will show that from a strictly dynamical point of view the preferred form of mantle convection remains a moderated form of whole mantle convection, in which the lower mantle moves more sluggishly than the upper mantle. This type of model can also explain many, but not all, of the seismological and geochemical observations.
U11B-03 INVITED
Seismic Tomography and Mantle Flow
This year marks the 30th anniversary of the beginning of global seismic tomography. The hope was that mapping velocity heterogeneities in the Earth interior would answer some of the fundamental questions of geodynamics. This has turned out more difficult than originally thought. Nevertheless, seismic tomography reveals several features that should lead to constraining the range of allowable geodynamic models. One of such findings is the "redness" of the power spectrum of lateral heterogeneity. It is dominated by large wavelength features in the boundary layers, which are most heterogeneous and where spectra show a rapid decrease after degree 6, or earlier, while in the regions that separate them the spectra are white and have relatively low amplitude.. Consequently, these large wavelength features can be resolved globally without the danger of a significant bias from a truncated shorter wavelength field. Few, if any, of the existing geodynamic models can reproduce the strength of the boundary layer spectra. One of the relatively recent findings is the boundary layer property of the transition zone. It is characterized by a strong degree-2 signal, absent in the upper mantle above 400 km depth and at the top of the lower mantle. This fact indicates that the whole mantle circulation must be severely impeded, leading to a drastic change of the pattern of lateral heterogeneity across the 650 km discontinuity. Only models derived using overtone data can effectively resolve this feature. The source of dominating degree 2 and 3 signal in the lowermost mantle, with the African and Pacific mega- plumes and the circum-Pacific ring of fast velocities, remains a mystery. The spectrum of density anomalies in the lowermost mantle predicted by subduction reconstructions (i.e. Ricard et al., 1994) is white and is very different from that of seismic velocities. However, they correlate well for degrees 2 and 3 (Richards and Engerbretsen, 1991); the implication is that the processes in the lowermost mantle – whatever is their cause – may control the overall distribution of the subduction zones at the surface.
U11B-04 INVITED
The Elephants' Graveyard: Constraints from Mantle Plumes on the Fate of Subducted Slabs and Implications for the Style of Mantle Convection
The style of mantle convection (e.g., layered- vs. whole-mantle convection) is one of the most hotly contested questions in the Geological Sciences. Geochemical arguments for and against mantle layering have largely focused on mass-balance evidence for the existence of "hidden" geochemical reservoirs. However, the size and location of such reservoirs are largely unconstrained, and most geochemical arguments for mantle layering are consistent with a depleted mantle comprising most of the mantle mass and a comparatively small volume of enriched, hidden material either within D" or within seismically anomalous "piles" beneath southern Africa and the South Pacific. The mass flux associated with subduction of oceanic lithosphere is large and plate subduction is an efficient driver of convective mixing in the mantle. Therefore, the depth to which oceanic lithosphere descends into the mantle is effectively the depth of the upper mantle in any layered mantle model. Numerous geochemical studies provide convincing evidence that many mantle plumes contain material which at one point resided close to the Earth's surface (e.g., recycled oceanic crust ± sediments, possibly subduction-modified mantle wedge material). Fluid dynamic models further reveal that only the central cores of mantle plumes are involved in melt generation. The presence of recycled material in the sources of many ocean island basalts therefore cannot be explained by entrainment of this material during plume ascent, but requires that recycled material resides within or immediately above the thermo-chemical boundary layer(s) that generates mantle plumes. More recent Os- isotope studies of mantle xenoliths from OIB settings reveal the presence not only of recycled crust in mantle plumes, but also ancient melt-depleted harzburgite interpreted to represent ancient recycled oceanic lithosphere [1]. Thus, there is increasing evidence that subducted slabs accumulate in the boundary layer(s) that provide the source of mantle plumes, as suggested 25 years ago by Hofmann & White [2]. Determination of the depth of origin of mantle plumes would provide a 1st-order constraint on the depth of plate subduction and the volume of the "upper" mantle. Improved seismic techniques and deployment of OBS arrays may soon allow robust imaging of mantle plumes in the deep mantle, although preliminary results are controversial [3]. Detection of a conclusive geochemical signature of core/mantle interaction would also provide strong evidence for a deep origin of mantle plumes, although there is considerable debate as to what such a signature would entail. In summary, determination of the depth of origin of mantle plumes may provide the key to deciphering the fate of subducted slabs and the overall style of mantle convection. Although this problem remains unresolved after several decades of work, recent developments in both geophysics and geochemistry provide hope for a final resolution within the next 10 years. [1] M Bizimis, M Griselin, JC Lassiter, VJM Salters, G Sen, EPSL 257, 259-293, 2007. [2] AW Hofmann, WM White, EPSL 57, 421-436, 1982. [3] R Montelli, G Nolet, F Dahlens, G Masters, E Engdahl, S-H Hung, Science 303, 338-343, 2004.
U11B-05
Whole Mantle Thermo-Chemical Convection Models With Realistic Mineral Physics Naturally Develop Chemical Stratification
Starting with [Christensen and Yuen, 1985 JGR], many isochemical convection models have demonstrated the existence of "intermittent" or "partial" layering enforced by the ringwoodite to perovskite+magnesiowustite phase transition over a certain range of Clapeyron slope values, which has often been cited as a possible mechanism for reconciling conflicting evidences for whole-mantle and layered convection. Current mineral physics constraints indicate, however, that the likely value of the Clapeyron slope is too low to enforce this mode, although studies have shown that a viscosity increase at 660 km depth might account for much of the observed variation in slab dynamics without appealing to a phase transition. When chemical variations are additionally taken into account, the dynamical effect of phase transitions can again become important. Firstly the additive effect of the '660' phase transition and chemical buoyancy can combine to keep denser than average material in the lower mantle and less dense than average material in the upper mantle, the so-called filter effect first identified by Weinstein [1992 EPSL]. Secondly, the pyroxene-garnet components transform to perovskite at a higher pressure than olivine components, giving positive buoyancy to MORB and negative buoyancy to harzburgite in the depth range 660-720 km, which has been shown to cause local chemical stratification around 660 km depth. Thirdly, MORB is likely denser than average mantle in the deep mantle, and some fraction of it settles into a layer above the CMB. These effects are here demonstrated and quantified in 3-D spherical convection calculations in which the mineralogy is calculated self-consistently as a function of temperature, pressure and composition (expressed as the ratios of 5 oxides) using free energy minimization. Compositional variations arise self-consistently from melting. These build on the earlier studies of Xie and Tackley [2004 PEPI, JGR], Nakagawa and Tackley [2005 Gcubed; 2006 GRL], and Tackley et al. [2005 AGU monograph]. In conclusion, as with most "great debates" the likely resolution lies inbetween the endmember scenarios, with circulation that extends through the entire mantle but a dynamically-maintained chemical stratification without a sharp boundary.
U11B-06
Global slab structure from (P-wave) travel time tomography: neither layered nor whole mantle convection
We comment on the fate of slabs of subducted lithosphere using a new global model of three dimensional (3-D) variations in mantle P-wave velocity. The model is parameterized by means of rectangular cells in latitude, longitude, and radius, the size of which adapts to sampling. The largest single data source is ISC-NEIC data reprocessed by Engdahl and co-workers, from which we use routinely picked, short period P, Pg, Pn, pP and pwP data (for earthquakes between 1964-2004). Resolution in the lowermost and uppermost mantle is improved by differential times of core phases (PKPDF - PKPAB, PKPBC - PKPAB, Pdiff - PKPDF) and surface reflected waves (PP-P), respectively. The low frequency differential times (Pdiff, PP) are measured by waveform cross-correlation. Approximate 3-D finite frequency kernels are used to integrate the long period data (Pdiff, PP) and short period (P, pP, PKP) data. Spatial resolution is ~100 km in best sampled upper mantle regions. Our model reveals in unprecedented detail the rich variation in style of subduction of lithospheric slabs into the mantle. The images confirm the structural complexity of downwellings in the transition zone discussed in previous papers (Van der Hilst et al., Nature, 1991, 1995, 1997). Slab deflection is apparent in the transition zone beneath back arc regions in the western Pacific and the Mediterranean (Fukao et al., Rev. Geophys., 2001), but deeper penetration seems to occur beneath many other convergent margins, in particular Indonesia and the eastern Pacific/Americas (e.g., Ren et al., JGR, 2007). Owing to added data from stations in China, our model reveals with more clarity the structure of slab fragments stagnant in the transition zone beneath East Asia. As we have suggested before, these results of variable depth subduction are not consistent with the canonical models of either strict layering at 660 km depth or unhindered whole mantle convection.
U11B-07
Multi-scale Convection in a Mantle with Strongly Temperature-dependent Rheology
Owing to the heterogeneity and complex rheology of mantle material, several scales of convection coexist in the Earth's mantle, producing cold linear slabs, mid-ocean ridges, 3D superswells and hot spots. However, if those features have individually been generated and studied in numerical and laboratory experiments, the exact conditions for their coexistence in a self-consistent convective model have remained elusive. We studied the characteristics of thermal instabilities developping when a layer of sugar syrup, a fluid with a strongly temperature-dependent viscosity and high Prandtl number (> 7000), is heated from below and cooled from above. A new visualization technique allowed to determine both the temperature and velocity fields inside the experimental tank. We focuss on high Rayleigh numbers (1.7x106 to 3.3x107) and intermediate viscosity ratios (7 to 4100). For this parameter range, "sluggish lid" convection occurs, in which three different scales of convection develop. Owing to the viscosity increase with decreasing temperature, the tank thermal structure becomes asymmetric: thermal boundary layer (TBL) instabilities, typical of high Rayleigh number convection, develop under the coldest, therefore most viscous, part of the upper thermal boundary layer which cannot move as fast as the less viscous fluid. The largest convective scale is therefore cellular, with cold downwelling sheets of viscous fluid encasing hotter parts of the tank. Within each of those cells develop several (typically 3 to 7) hot 3D upwelling plumes. Upon impinging under the cold TBL, each plume in turn generates locally a small ring of cold material which does not reach the bottom of the tank. The introduction of a denser layer at the bottom of the tank can vary the morphology of the hot instabilities but has no influence on the existence of the large-scale cold circulation. Hence high Rayleigh thermal convection in the sluggish lid regim can produce large-scale cells delimited by cold subducting slabs, within which several 3D plumes develop. On Earth, two of such cells exist, the Pacific and the Indo-Atlantic boxes. Our experiments further suggest that what has been named the two "hot superplumes", i.e. the two seismically slow regions encased within the subduction rings, are in fact each constitued of several hot instabilities.
U11B-08
Quantifying the Flux of Slab Material Through the Transition Zone and Beyond
While tomographic images are interpreted as the present day location of subducted slabs in the mantle, the variety of slab morphologies imaged suggests that slabs meet significant resistance at the upper part of the lower mantle. Unfortunately, these images only provide a qualitative estimate of the flux of material through the transition zone. It is known that the Clapeyron slope of the ringwoodite to perovskite plus ferropericlase phase transformation, the viscosity of the lower mantle, the effect of pressure on the coefficient of thermal expansion, the rate of trench migration, and the composition, thermal structure and rheology of the slab impact the dynamics of deeply subducted slabs. Previous attempts to quantify the flux of material through the transition zone have included following passive tracers, using two-point correlation functions, and the flow-based mass-flux diagnostic. Passive tracers are ideal for tracking the fate of slab; however it is difficult to use passive tracers to actually quantify the flux upper mantle material (if any) that is entrained with the slab into the lower mantle. The two-point correlation function is useful for identifying layering and can be related to the mass flux diagnostic, the most direct measure of the vertical flux of material in a fluid system. I will illustrate the mass flux diagnostic and two-point correlation function on some easy to interpret flows and then speculate on the two-point correlation function calculations from tomographic models.
U11B-09
Layered Convection in Earth's Mantle: Results from New Control Volume Based Analyses
The issue of the extent to which the mantle convective circulation may be layered by the influence of phase transitions and/or by the variation of physical properties remains unresolved. That the endothermic transition at 660 km depth coupled with an increase in viscosity across this horizon provides a significant impediment to radial mass flux is clear on the basis of recent seismic tomographic images from several geographic regions. In these regions, high resolution reconstructions of Benioff zone body-wave heterogeneity demonstrate that the downgoing slab appears to be "trapped" in the transition zone rather than continuing to descend through it.We will describe a series of new simulations of the convective mixing process based upon the application of a highly accurate control volume-based methodology. The model includes the influences of three major phase transformations (Olivine-Spinel, Spinel-Perovskite+Magnesiowustite, and Perovskite-post Perovskite) as well as accurate pressure and temperature dependences of thermal conductivity, coefficient of thermal expansion and viscosity. In this model the flow is characterized by significant layering, with radial mass flux strongly inhibited across the base of the transition zone. The "avalanche effect" previously identified by Solheim and Peltier (JGR 99, 1994) and Butler and Peltier (JGR 107, 2002) is found to be especially pronounced. Our models are consistent with the Urey ratio, global heat flow and CMB temperature constraints.