V44B-01
Seismic Tomography and Structure of the Transition Zone
Most of the tomographic models employ data sets that had satisfactory resolution in a limited radial extent. For example, studies using teleseismic travel times do not have resolution in the upper mantle, even though using matrix conditioning methods it is possible to derive a "model" of the upper mantle. Similarly, some studies using surface waves presented detailed "models" at depths far exceeding resolution of the data used. A data set that fills in the gap created by the limitations of the surface wave dispersion and teleseismic travel times data consists of the higher modes. Only three modeling groups use the combination of all three data types and, therefore, can obtain the structure with a sufficient radial resolution to identify the properties just above, within and just below the transition zone. These three groups are Berkeley (non-linear asymptotic coupling theory; Li and Romanowicz, 1996) with the most recent model by Panning and Romanowicz (2006), Harvard (path average approximation; Woodhouse and Dziewonski, 1984) with the most recent model by Kustowski et al. (2007) and Caltech/Oxford (separation of overtones; van Heist and Woodhouse, 1997) as represented by a model of Ritsema et al. (1999). These global models have a nominal horizontal resolution of about 1,000 km and variable radial resolution, from about 50 km below the Moho, to 100-150 km in the asthenosphere and transition zone, and about 250 km in the middle and lowermost mantle. These models show that there are three boundary layers in the mantle. The one near the surface is dominated by spherical harmonics up to degree 8, with the peak at degree 5. It ends between 200 and 250 km depth, after which the power spectrum is low and tends to be white. The second boundary layer is in the transition zone, where overall power increases and is clearly dominated by degree 2. The spectrum changes abruptly at the top of lower mantle, where it decreases in amplitude and becomes white; this character of the spectrum continues to about 2000 km depth. The third boundary layer has the maximum power near the CMB, it is dominated by degrees 2 and 3. These great wavelength velocity anomalies are not limited to the lowermost mantle, but extend to about 1000 km above the CMB and, perhaps, with a significantly diminished amplitude all the way to the transition zone; this is further documented by consistently positive values of the radial correlation matrix in the lower mantle. The transition zone signature, both in wavenumber and space domains, implies impeded mass flux between the upper and lower mantle. There are three locations (Indonesia, Fiji-Tonga and South America) where spatially limited fast velocity anomalies underlie large scale fast velocity anomalies in the transition zone, thus indicating a possibility of some penetration of the subducted material into the lower mantle.
V44B-02
Are Subducting Slabs Wet or Dry in the Transition Zone? Comparisons of Seismicity and Model Predictions
We compare distribution of seismicity with what might be expected for dry and wet models of subducting slabs interacting with the mantle transition zone. For depths between about 100 to 250 km, earthquake distribution matches closely with where breakdowns of major hydrous minerals in the oceanic crust and mantle are expected -- a result consistent with dehydration embrittlement as the nucleation mechanism for intermediate-focus earthquakes. Below about 250 km, however, such mechanisms would have been exhausted; and correspondingly, at depths near 300 km, the frequency of earthquakes drops rapidly to a minimum of only one per year per 10 km interval in depth globally. In the "dry" model, disappearance of most H2O from the slab leads the way for transformation-induced faulting during the olivine-wadsleyite and olivine-ringwoodite reactions whose kinetics are temporally retarded by low temperatures in slabs. This scenario naturally explains abrupt termination of all deep earthquakes near the base of the transition zone, and is consistent with petrologic anomalies recently found in the transition zone beneath Tonga-Fiji and the Mariannas. In the "wet" model, earthquakes are essentially absent between depths of 250-350 km because H2O is bound into phase A and/or dissolved in olivine and pyroxene. Subsequently H2O is carried passively into the transition zone where it fluxes the olivine-wadsleyite reaction with little overstep of the reaction boundary and/or is passed from phase A to E and then to D. At greater depths, however, predictions from this model cannot reproduce observed patterns of deep earthquakes. As olivine reacts to form wadsleyite, the solubility of H2O becomes much higher, allowing for no free fluid and hence no earthquakes from dehydration. If transient release of H2O should occur, a flurry of earthquakes would be expected just below the "410 km" discontinuity (which occurs at depths of perhaps 300-380 km in slabs), at greater depths H2O would either remain dissolved in nominally anhydrous phases (hence not involved in triggering earthquakes at all), or it might conceivably generate some earthquakes but only at depths of ~500km and ~650 km as free H2O is transiently released during reactions to increasingly hydrous phases such as E and D. In places where some slabs may pass into the lower mantle (as in the Mariannas), breakdown of hydrous ringwoodite would be expected to release H2O, inducing a flurry of earthquakes in the uppermost lower mantle just below the 660-km discontinuity; or hydrous ringwoodite could be consumed in part by Phase D which would then be expected to release H2O and generate earthquakes at around depths of 900 km in the lower mantle. Either way, such predictions call for earthquakes to occur beneath the 660-km discontinuity where not a single earthquake has been observed.
V44B-03
Successful and Failing plumes in a Heterogeneous Mantle: the Icelandic Case
Although Iceland is always cited as an exemple of hot spot volcanism produced by a deep mantle plume, an increasing number of observations cannot be explained by the classical plume model of a mushroom-shaped plume out of a sustained localized heat source. Volcanic episodes with moderate temperature predate the major episode of mafic magma emplacement (~60Ma) containing hot picrite magma with strong rare gas anomalies. Present-day Iceland shows moderate temperatures, a strong rare gas anomaly, and an apparent disconnection between slow seismic anomalies in the upper and lower mantle. Noteworthy, the same mixture of geochemical ingredients are found in Icelandic lavas during its 80 Myr of activity. We present a new experimental study of the more realistic case of thermochemical convective instabilities developping out of a heterogeneous bottom hot thermal boundary layer. Depending on the buoyancy ratio B, two end-member regimes are observed. For large B, a thermal plume develops above the denser layer and only a small amount of denser fluid is entrained in the plume. For small B, the dense layer can be sufficently heated to become buoyant and rise: the thermo-chemical plume is therefore mainly constituted of material from the chemically denser layer. The fate of the heterogeneous material in the plume then depends on time since the instability cools as it ascends. As a result, the core of the plume head, which consists of initially hotter but chemically heavier material, can cool enough to become denser than the ambient fluid before reaching the surface of the tank: the heterogeneous material then sinks back and a new thermal plume with a lower temperature anomaly is generated from the top edge of the heavier collapsing blob. In this "failing-plume" mode, the thermo-chemical plume fails to deliver most of the chemical heterogeneity to the surface. Hence, the thermal and compositional structure of a thermo-chemical plume changes with time and is quite irregular. In particular, it is not because a region is hot that it is buoyant and rising. The interplay between thermal convection and at least 3 layers (or "reservoirs") of different densities could therefore well explain the time- dependence and morphology of the Icelandic melting anomaly. And in this framework, geochemical data can be explained if each reservoir is a mixture of the same mantle components, albeit in different proportions.
V44B-04
Stretching of passive tracers and implications for mantle mixing
Mid ocean ridge basalts(MORB) and ocean island basalts(OIB) have fundamentally different geochemical signatures. Understanding this difference requires a fundamental knowledge of the mixing processes that led to their formation. Quantitative methods used to assess mixing include examining the distribution of passive tracers, attaching time-evolution information to simulate decay of radioactive isotopes, and, for chaotic flows, calculating the Lyapunov exponent, which characterizes whether two nearby particles diverge at an exponential rate. Although effective, these methods are indirect measures of the two fundamental processes associated with mixing namely, stretching and folding. Building on work done by Kellogg and Turcotte, we present a method to compute the stretching and thinning of a passive, ellipsoidal tracer in three orthogonal directions in isoviscous, incompressible three dimensional flows. We also compute the Lyapunov exponents associated with the given system based on the quantitative measures of stretching and thinning. We test our method with two analytical and three numerical flow fields which exhibit Lagrangian turbulence. The ABC and STF class of analytical flows are a three and two parameter class of flows respectively and have been well studied for fast dynamo action. Since they generate both periodic and chaotic particle paths depending either on the starting point or on the choice of the parameters, they provide a good foundation to understand mixing. The numerical flow fields are similar to the geometries used by Ferrachat and Ricard (1998) and emulate a ridge - transform system. We also compute the stable and unstable manifolds associated with the numerical flow fields to illustrate the directions of rapid and slow mixing. We find that stretching in chaotic flow fields is significantly more effective than regular or periodic flow fields. Consequently, chaotic mixing is far more efficient than regular mixing. We also find that in the numerical flow field, there is a fundamental topological difference in the regions exhibiting slow or regular mixing for different model geometries.
V44B-05 INVITED
Depletion of the Continental Lithosphere
It has been well known for thirty years that the density of the mantle is controlled by composition as well as by temperature, and that mantle convection is therefore strictly double diffusive. The change in density of the mantle caused by the removal of 25% melt is the same as that produced by a change in temperature of 500°C. Compositional variations are therefore likely to have an important effect on mantle dynamics. But it is difficult to include compositional variations in numerical models of mantle convection, partly because they result from melt extraction and partly because the Lewis number, the ratio of the thermal to the compositional diffusivity, is so large (~ 1010). From an observational point of view there is a further problem. Though melt extraction has an important effect on density, it has almost no effect on the seismic velocities, at least in the upper mantle. Therefore compositional variations therefore cannot be mapped using seismic tomography. The only available method of doing so is by using geochemistry. Melts sample a larger region of the mantle than do nodules, and can therefore provide geodynamically more useful estimates of the mean density. Depletion by melt removal leaves a residue that is refractory. It is only likely to generate more melt if it is re-enriched by metasomatic melt. If it is then remelted, the resulting melts have easily recognisable signatures of the depletion and re-enrichment. The most enriched melts come from the most depleted residues and have a distinctive heavy rare earth signature. These effects can be used to estimate the composition and density of the source regions of the melt. Such studies show that the lithosphere beneath southern Africa, Tibet, and other regions of thick lithosphere have had their density decreased by the removal of ~ 25% melt, which has stabilised the lithosphere against convective instabilities. Similar effects may be important in the deeper parts of the mantle. However, because such regions do not generate melt, their composition is unknown.
V44B-06
Recycled Metasomatised Lithosphere as a Source of Alkaline OIBs
Ocean island basalts (OIBs) are generally thought to provide information on the chemistry of the deep mantle, but the sources of these lavas are still being actively debated. For example, although the trace-element and isotopic compositions of OIBs suggest the presence of recycled components in their source regions, the origin of these recyled components is controversial. Oceanic crust is one possible source of recycled material, and it may be important in the generation of tholeiitic magmas (hy- and qtz- normative magmas) from large oceanic islands and continental lava flows (1, 2); the fact that partial melts of oceanic crust are silica oversaturated (3) makes it difficult to envision a significant role for them in the generation of alkaline (i.e., ne-normative) magmas. Veined/metasomatized oceanic or continental lithosphere in the source regions of OIBs is an alternative source of recycled components in these alkaline magmas (4, 5). We conducted melting experiments on natural amphibole-rich veins and on their dehydrated equivalents at 1.5 and 2.5 GPa to test the metasomatic hypothesis for the origin of recycled components in OIB sources. These experiments demonstrate that melting of metasomatic veins can reproduce key features of the major and trace element compositions of many nephelinites and basanites from both oceanic and continental settings; moreover, these experiments show that reaction of partial melts of amphibole-rich veins (or their dehydrated equivalents) with surrounding lherzolite can explain the observed compositional trends from nephelinites to alkali basalts. These experiments suggest that melting of metasomatized lithosphere (either "in place" or after recycling into the convecting mantle) is a viable alternative to more widely invoked models of alkaline basalt formation by melting of recycled oceanic crust ± sediments. This reinterpretation of the origin of alkaline OIBs has implications for mantle chemistry: First, it suggests that recycled components in the sources of islands characterized by tholeiitic magmas (i.e., Hawaii or Iceland) and in the sources of islands where ne-normative compositions are dominant (i.e. Polynesia, island in the Atlantic Ocean, etc.) are distinct. Second, it suggests that alkaline rocks are produced by large degrees of melting of small volumes of trace-element and volatile-rich material (primarily amphibole-bearing veins) rather then by low degrees of melting of less-enriched material (such as recycled oceanic crust or peridotites). If so, these alkaline rocks do not carry as much chemical information on the large scale composition of the convecting mantle as is often presumed. Third, the range of isotopic composition observed in alkaline lavas from a single oceanic island does not necessarily imply the interaction of distinct mantle components (such HIMU, EM) but could reflect the range and time integrated history of trace element compositions in veins formed by percolative fractional crystallization of metasomatic agents within the lithosphere (5). (1) Sobolev et al. (2007) Science 316, 412-417. (2) Herzberg (2006) Nature 444, 605-609. (3) Green et al. (1967) Earth Planet. Sci. Lett. 2, 41-51. (4) Niu and O'Hara (2003) J. Geophys. Res. 108, 2209. (5) Pilet et al. (2005) EPSL, Earth Planet. Sci. Lett. 236, 148-166.
V44B-07
Oxygen isotopes in MORB revisited: insights into the nature of the Indian MORB source and implications for crustal recycling
Geochemical differences between Indian ocean MORB (Mid Ocean Ridge Basalt) and Pacific or Atlantic MORB have been attributed to addition of one or more of the following enriched materials to the depleted mantle beneath the Indian ridge: 1) local plume material; 2) recycled pelagic sediment; 3) subduction-modified mantle wedge material; or 4) recycled continental lithospheric mantle or lower crust. We present high-precision laser- fluorination analyses of δ18O in fresh MORB glasses from the Australian-Antarctic discordance (AAD) and use these data to evaluate models for the distinctive character of Indian MORB mantle. Our samples include both Indian- and Pacific-type AAD MORB, and also include samples representative of the most enriched and most depleted (including 'ultra-depleted') AAD MORB. Based on these data, we found that the addition of lower crust to the upper mantle is most consistent with the combination of δ18O and radiogenic-isotope data for the AAD Indian MORB. In addition, there is no significant difference between mean (5.49-5.52‰) or standard deviation (0.1‰) of δ18O of Indian and Pacific MORB within the AAD. More broadly, a compilation of high-precision MORB data which includes data for the 9-10° N region of the East Pacific Rise and previously- published data for MORB globally shows no distinction in range (~5.25-5.80) or mean (5.5) of δ18O between Indian, Pacific or Atlantic MORB globally. Although some of the range in data reflects scatter due to analytical uncertainty (0.05‰ 1 s.d.), the larger standard deviation of 0.1‰ for data for each ocean basin indicates that there is oxygen-isotope heterogeneity preserved in the upper mantle on the scale of sampling during MORB melting. In subsets of the global data (e.g., 'Normal' MORB: Eiler et al., 2000, Nature, v.403 p. 530; or data for the north Atlantic: Cooper et al., 2004, EPSL v. 220, p.297), this heterogeneity in oxygen correlates with trace-element and radiogenic isotopic signatures of enrichment. Furthermore, data for each ocean basin can be fit by a normal distribution. The final distribution of data reflects the average percentage and physical distribution of enriched material in the upper mantle, potentially modified by sampling during melting, and with some scatter introduced through measurement errors. However, Monte Carlo simulations show that if significant differences in the skewness or mean of δ18O sampled during melting (which would be expected if the percentage or distribution of enriched material sampled differed significantly between ocean basins), we would expect these differences to be apparent in the data sets even with the limited number of samples (134 total). Therefore, either the upper mantle globally contains a similar percentage of crustal material which has a consistent oxygen-isotope composition, or there is a fortuitous trade-off between the percentage and δ18O of enriched material in the different ocean basins. This in turn implies that crustal materials with the most extreme oxygen-isotope compositions (e.g., pelagic sediments) are present only in very minor abundances in the upper mantle. In addition, it also implies that differences in radiogenic-isotope data between the different ocean basins are due primarily to the age and nature (e.g., altered oceanic crust vs. lower continental crust) of the enriched component, rather than due to large differences in abundance of enriched material.
V44B-08
Improving the resolution of the mantle picture
Although the identification of the mantle components provided a simple framework into which to place the isotopic variation of MORB and OIB, in the modern day such a picture is too simplistic and the actual components present in the mantle can be quite different. Recent improvements in measurement techniques allow for more precise isotope analyses as well as higher sample throughput. Consequently, we are now at the beginning of an era of discovery that follows from obtaining a higher resolution images. For three decades Hf and Nd isotopic compositions in ocean island basalts have been very well correlated, but MORB have not show a good correlation between the two isotope systems. Is this a lack of correlation in MORB is related to sample density or are the Nd and Hf isotope systems are truly uncorrelated? We present detailed investigations of the variations in Nd and Hf isotopic compositions in MORBs at normal ridges and from restricted geographic areas to determine whether, on a local scale, Nd and Hf isotopic compositions are correlated. This requires high precision data as the isotopic variations in N-type MORB are relatively small. For example, at the Mid-Cayman Rise the Hf-isotopic composition varies by 1.2 epsilon units. Our data from the Pacific, Atlantic and Indian Ocean show that, on a local scale, despite the limited variations the Nd and Hf isotopic compositions are well correlated. On a Hf-Nd isotope correlation diagram MORB from ridges far removed from hot-spots form parallel "local" arrays with similar slope as MORB from hotspot-influenced ridges, indicating that the nature of the heterogeneity that causes the local Hf-Nd isotope variation can be related to contamination by "enriched" OIB-type mantle. The parallel "local" arrays (differing Hf-isotopic composition) suggest a second larger scale length heterogeneity most likely related varying amounts of recycled oceanic lithosphere, which has a radiogenic Hf-isotopic composition. We also resampled and reanalyzed a suite of basalts from Makapuu Head on Koolau volcano, the isotopically enriched endmember of Hawaiian lavas. In 208Pb/204Pb vs 206Pb/204Pb space our new data forms a steeper slope than the KSDP lavas, and extend towards more unradiogenic Pb ratios than previously published. In contrast to the KSDP lavas, the Makapuu lavas converge with the posterosional Honolulu volcanics and Salt Lake Crater pyroxenite compositions (our new data) in the unradiogenic Pb isotope end. In 208Pb/204Pb vs. Nd isotopes the Makapuu lavas also extent towards the isotopically depleted endmember of the Hawaiian plume compositions as this is defined by the Honolulu volcanics and the pyroxenites. These observations are consistent with the presence of an ancient depleted component within the enriched Koolau endmember. These data combined with other recent high precision isotopic data from other Hawaiian volcanoes reveals that each volcano requires distinct endmember isotopic compositions. These two examples show that detailed investigations at local scales are important in defining the components present in the mantle. In addition these local studies, done at high resolution and high precision, are now able to provide information on the length scales of the heterogeneities. It is expected that more of these high-resolution studies will redefine our view of the dynamics of the mantle.