DI31C-1803
Mixed Spin State of Iron in a New Post-Perovskite Silicate at Earth's Core-Mantle Boundary
Earth's lower mantle is mainly composed of (Mg,Fe)SiO3 perovskite. Previous discovery of a high- pressure transition in end-member composition MgSiO3 led to a flood of interpretations for the deepest region of the mantle based on the post-perovskite (ppv) phase. We present high pressure experimental results on a composition of Fe0.4Mg0.6SiO3. After conversion to ppv, Fe K/beta x-ray emission spectroscopy (XES) of the new phase shows a mixed spin structure with 70% of the Fe2+ in low-spin state (3d electron spin paired) and 30% remaining in high-spin state (spin parallel). Rietveld refinement of the x-ray diffraction (XRD) pattern indicates two non-equivalent Fe/Mg sites and a Pmma space group symmetry which is different from the Cmcm space group reported for MgSiO3 ppv. The Fe2+ partitions unequally between the two sites with 70% in one site and 30% in another site, coinciding with the XES finding. The new mixed-spin ppv may help to address a number of apparent inconsistencies that have surrounded the Fe-bearing ppv.
DI31C-1804
Thickness of the Post-Perovskite Boundary in Fe2+ and Fe3+ Bearing Systems
Combined with seismic observations of the D" region, the perovskite (Pv) to postperovskite (PPv) transition provides new opportunities to constrain important parameters at the core-mantle boundary, such as heat flux from the core through the observation of the double crossing between the geotherm and the PPv transition. In the mantle, the PPv boundary should have a finite thickness. We have determined the PPv boundary thickness in 0.9MgSiO3-0.1FeSiO3 and 0.9MgSiO3-0.1Fe2O3 in the laser-heated diamond cell combined with in situ X-ray diffraction. In order to reduce thermal and pressure gradients, we use argon as a medium. In order to reduce the kinetic effect, we measured the boundary along both forward (from Pv to PPv) and reverse (from PPv to Pv) paths. Our data tightly constrain the transition depth and Clapeyron slope (+6.7±0.5 MPa/K in Fe2+ and +8.4±1.0 MPa/K in Fe3+) and the results are consistent with seismological observations. However, 10 mol% Fe2+ increases the thickness to 450±50 km which is significantly larger than the thickness of the D" discontinuity (≤80 km), whereas 10 mol% Fe3+ increases the thickness to only about 100±50 km. The discrepancy with the seismic observation for Fe2+ can be reconciled by assuming strong partitioning of Fe to ferropericlase (Fp) such that PPv is much more depleted in Fe than Pv. However, the partitioning of Fe among Pv, PPv, and Fp is highly controversial. An Fe3+-enriched lower mantle provides a natural explanation for the thickness of the D" discontinuity. These two models can be further examined for the high electrical conductivity required by the observation of the Earth's nutation. The low Fe2+ in Pv and PPv and high Fe2+ in Fp would result in very low electrical conductivity. However, an Fe3+-enriched lower mantle would have high electrical conductivity because Fe3+--O bonding becomes metallic in the PPv structure and the coexistence of Fe2+ and Fe3+ will enhance electron hopping between these two. Therefore, our thickness measurements combined with seismic and geodetic observations indicate that the lowermost mantle may be chemically distinct from the rest of the mantle.
DI31C-1805
Partitioning of Ni and Co between liquid metal and ferropericlase at high pressures
A partitioning behavior of siderophile elements between liquid metal and lower mantle minerals are important to understand the processes occurred in the early Earth. A partitioning behavior between liquid metal and silicate melt or between liquid metal and mantle minerals has been studied to date. However, most of the previous studies were conducted at pressures below 30 GPa, which is lower than the proposed pressure, and those have experimentally ascertained that the partition coefficient varies with pressure of a core-mantle equilibrium in a deep magma ocean (30-60 GPa, >2000 K). Although Bouhifd & Jephcoat (2003) performed partitioning experiments for Ni and Co between liquid metal and silicate melt up to 42 GPa and 2500 K, their results have large uncertainties because of very low contents of Ni and Co in the silicate phase. Ferropericlase which is a major lower mantle phase is considered to be a reservoir of Ni and Co. In this study, we performed partitioning experiments of Ni and Co between liquid metal and ferropericlase at 36 and 40 GPa at 2700-2800 K in order to examine pressure dependence of partition coefficients of Ni and Co. The experiments were performed using a double sided laser heated diamond anvil cell. The starting material was a Fe54Ni29Co17 alloy foils (in wt %) sandwiched between (Mg0.9, Fe0.1)O ferropericlase powder. .The sample was insulated from the diamond surfaces by corundum disks. Textural observation and chemical analysis of the recovered samples were performed by a field emission-SEM EDS. Our results revealed that the exchange partition coefficients of Ni and Fe (KDNi- Fe), and Co and Fe (KDCo-Fe), between liquid metal and ferropericlase are 3.0 (7) and 2.3 (5) at 36 GPa and 2800 K and 4.0 (13) and 2.8 (10) at 40 GPa and 2700 K, respectively. These results revealed that there is no remarkable decrease of the exchange partition coefficients of Ni and Co at pressures between 36 GPa and 40 GPa.
DI31C-1806
Implications of Partially Molten Ultralow-Velocity Zones for Outermost Core Stratification
The low shear modulus and high temperature of ULVZ implies they may be partly molten. To maintain a dynamo, Earth's ancient core was much hotter than at present and the region will have been more extensively melted, forming a dense basal magma ocean (BMO) in Earth's earlier history. The presence of a BMO will have enhanced core-mantle chemical interactions relative to a solid lowermost mantle, however, recent experimental constraints imply that silicates with Fe#=[Fe]/([Fe]+[Mg])~0.1 or 0.2 cannot be in equilibrium with a core containing only 10% light elements: in this case, the core is under-saturated in O and possibly Si. At the same time, ULVZ sit atop the core and their density implies a Fe# much greater than the ambient mantle, which further exacerbates the magnitude of disequilibrium between core and mantle. This apparent paradox is readily avoided because light elements such as O dissolved into the top of Earth's core are many orders of magnitude too buoyant to be entrained and mixed into the bulk of the outer core, and a stably stratified layer must inevitably have formed at the top of the core that shields the interior and accounts for gradients in chemical potential. Radial transport of light elements in this case is limited to diffusion only, and various evolution scenarios provide new tests for the notion that ULVZ are partially molten, since such a feature appears to be an inevitable outcome of this hypothesis. Such a layer carries observational consequences for the seismic structure of the outermost core and the secular variation of the magnetic field, and while these hypothesis tests are compatible with present data more tests will be needed in the future.
DI31C-1807
Thermal and magnetic consequences of a stagnant layer at the base of the mantle that becomes entrained in mantle circulation
The presence of a dense layer at the base of Earth's mantle that is not participating in the mantle general circulation would have significant consequences for the thermal evolution of the lower mantle and core and for the Earth's magnetic history. Such a layer reduces the efficiency of heat transfer between the core and mantle and hence the cooling rate of the core leading to estimates of an older inner core and diminished energy available to drive the geodynamo. In this contribution, we present a suite of numerical models of the thermal evolution of the mantle that are coupled with a parameterized thermal and magnetic model for the core. We consider, in particular, scenarios where a dense layer persists at the bottom of the mantle which becomes entrained in the mantle general circulation mid-way through Earth evolution. We find that such scenarios produce young inner cores due to the high core heat flow that occurs once the layer has become entrained while the temperature at the CMB remains reasonably high due to the insulating effect of the layer at early times. Any enrichment of the layer in radio-active elements causes sufficiently low core-mantle heat flows before the layer is entrained that the geodynamo ceases to operate, however.
DI31C-1808
Tail's structure of mantle plume: implication for episodic eruption of flood basalts
According to the classical mantle plume model, Large Igneous Provinces (LIPs) are formed in a single eruption that lasts a short period of time. Recently an increasing number of isotopic age studies indicate that some plume eruptions are intermittent and the intervals between eruptions are in the range of 20 - 40 Ma. We believe that the key to understand episodic eruptions lies in the energy/matter supply from plume tail, and Markers-In-Cell (MIC) technique is used to monitor the fine structure in plume tails. The results clearly show that the controlling factors for episodic eruptions are ascending velocity and head expanding speed which are determined by Ra number and viscosity contrast ratio. On condition of high Ra number and strong viscosity contrast the plume tail will become unstable and takes distinct structure evolving with time, growing- shrinking-resuming-shrinking, and this process can repeat. If such a single cycle represents one upwelling or basalt eruption, the repeating events may correspond to the episodic eruptions. The interval between eruptions (20 - 30Ma) and total eruptive volume are consistent with the geological evidence supports our speculation. In addition, our simulations also provide a new clue to understand the plume dying mechanism, that is, the plume may diminish because of the lack of supply in the tail.
DI31C-1809
Mixing and Entrainment in Mantle Plumes: A 3D Experimental Investigation
Significant differences exist between isotopic signatures of typical mid-ocean ridge basalts (MORB) and those associated with many ocean islands, with ocean island basalts (OIB) generally exhibiting more variability in trace element concentrations and a bias towards enrichment in more primitive isotopes as well in some cases. Such observations coupled with other geophysical evidence have been used to suggest that OIB's are surface manifestations of upwellings originating in the deep interior near the core-mantle boundary that interact with distinct, heterogeneous reservoirs as material is transported from the Earth's interior to the surface. Although many have studied the chemistry and dynamics of these mantle plumes, fundamental questions remain. Such questions can be grouped into two general issues: a) Plume structure and dynamical interaction with the surrounding mantle, b) The degree of entrainment and mixing in mantle plumes of chemically distinct material from the deep mantle. We address these fundamental questions by performing detailed fluid dynamical experiments to determine the structure, temperature, velocity, and degree of entrainment in thermal plumes. Heat is used as the driving convective mechanism to form a single thermal plume. The experiments are conducted in a Plexiglas tank (inner dimensions of 26.5×26.5×26.5 cm). A small heater of 2.0 cm diameter and centered in the tank bottom is connected to programmable power supply. By varying voltage settings we can simulate varying heat fluxes in the deep mantle. Our experiments utilize Stereoscopic Particle Image Velocimetry (SPIV) and Thermochromic Liquid Crystals (TLC's) to reconstruct the 3D flow and temperature fields within the tank. Penetration height and plume head size are related to the varying buoyancy flux. In addition, velocity and vorticity fields determined using SPIV provide insight into the plume structure and the nature of the entrainment process.
DI31C-1810
Evidence for small-scale thermal convection near the CMB from joint multi-scale analysis of seismic images and finite element modelling
The discovery of the post-perovskite phase near the CMB has stirred a debate about the relative effects of thermo-chemical changes and distortion of the phase boundary by thermal convection. Previous work on chemical buoyancy in the D" layer considered relatively long wavelength solutions with horizontal wavelengths greater than 600 km. We consider horizontal (vertical) wavelengths as short as 10 km (4 km) to determine if thermal convection with a post-perovskite transition is compatible with recent multi-scale seismic imaging with ScS and SKKS precursors and coda, or if composition must be considered. We investigate the interaction of subducting lithospheric slabs with the D" region, accounting for the post-perovskite phase transition, by numerical simulation with a cylindrical finite element convection model. The nodal point resolution of the model is about 3 km (10 km) in the vertical (horizontal) direction (which is slightly smaller than the spatial resolution afforded by ScS and SKKS scattering). The slope and CMB intercept of the ppv Clapeyron curve are taken to be 11 MPa/K and 3500 K, respectively, while the CMB temperature is set to 4000 K. Phase-dependent composite non-linear rheology is used with a greater propensity for non-linear creep in ppv regions. Our results show ppv lens-like structures of ~250 km thickness and > 1000 km lateral extent near remnants of cold downwellings; mantle regions where hot plumes emerge from the CMB are pv- rich. The cold ppv regions deform according to non-Newtionian flow and have lower viscosity than the hotter pv plumes, which have linear (Newtonian) rheology. In the numerical models, the base of the ppv lenses is inside the thermal boundary layer near the CMB and is marked by a stronger S velocity gradient than the top. The evidence is built by subjecting the finite element modelling - upon subtracting a tomographic S velocity background model - to a decomposition into wave packets, restricted to (about 6) scales and orientations proven to be resolved by the multi-scale imaging of large sets of global network data. This resolution is derived from a matrix representation of the generalized Radon transform, making use of the concentration of wave packets, and opens new pathways for joint seismic and geodynamic analysis of the D" region.
DI31C-1811
ULVZ – Now you see it … now you don't
Recent years have witnessed significant improvements in our understanding of ultra-low velocity zones (ULVZs) - thin layers of strongly reduced seismic velocity and possibly increased density at the core mantle boundary (CMB). However, there appear to be little, if any, constraints on the true nature of ULVZs, their formation, evolution, and dynamics. Our knowledge of ULVZs is challenged by only a limited geographical sampling of Earth's CMB. Fundamental questions persist, such as: Does a ubiquitous ULVZ blanket the CMB? Or, rather, are ULVZs regional features that depend on specific chemical and thermal conditions of core and mantle? Limited sampling of the CMB is primarily due to uneven earthquake source and receiver distribution, combined with ULVZ probes having specific epicentral distances of utility. Of the 40 percent or so of the CMB probed for ULVZ structure, only a small fraction shows evidence for ULVZ structure. This may imply either that (a) ULVZs are indeed absent in most areas, or (b) ULVZs are everywhere, but too mild in properties (i.e. too thin or too weak) to generate the necessary waveform variations for positive identification. Improved seismic resolution and CMB coverage through new ULVZ probes and new source-receiver combinations are desired to move towards answering these questions. Here we present results from two previously unstudied regions: we find evidence for the absence of ULVZ structure in the northeast Pacific using a very dense network of seismic stations in the western US, and find an extremely thin ULVZ in the southwestern Pacific. For the latter region, we employ a new deconvolution approach that enables detection of thinner ULVZ layering. These patches are outside and inside of the large low-shear velocity province of the Pacific and geodynamic simulations indicate a possible connection with the boundaries of this structure and ULVZ distribution.
DI31C-1812 INVITED
Deep-earth diffracted-wave tomography: A multiple-frequency high-resolution approach
We present a method to address the seismic forward and inverse problem at the global scale, with a specific focus on waves diffracted in the lowermost mantle. This spectral-element approach is developed in spherically symmetric background models, offering full-wave sensitivity up to 1 Hz. In order to get a direct view of the interconnection between surface displacements and earth structure, we examine, for a given source-receiver pair, the time-dependent sensitivity of the seismic signal to perturbations in the target region of interest. That analysis suggests data selection criteria to incorporate into future inversions. It is furthermore conceivable to assess different 1D background models for a specific D" region in question which may be undertaken with very few preliminary test simulations and comparison to data. We measure and model our static observables (traveltimes, amplitudes) in multiple-frequency passbands, which together span the usable broadband range. This increases robustness and resolution by multiplying the number of constraints on the inverse problem. It also allows us to selectively draw only upon frequency bands with high signal-to-noise ratio (e.g., to exclude the microseismic noise bands for some locations). We discuss coverage maps and target regions for which diffracted-wave tomography promises the greatest improvements in resolution and potential geodynamic interpretations.
DI31C-1813
Exploring Earth's Lowermost Mantle With Core-Diffracted Waves From Linear Arrays
We investigate lowermost mantle structure using the slowness and decay-rate frequency dependence of core diffracted waves from large earthquakes (mb>5.5) recorded by two broadband seismometer arrays, MOMA and FLED, to provide insights on processes occurring at the base of the mantle. MOMA (Missouri to Massachusetts) and FLED (Florida to Edmonton) were linear transects deployed in 1995-96 and 2001-02, respectively, with the common goal of obtaining seismic images of the Earth's deep interior. Exploiting core diffracted wave observables with the geometry of these arrays aids in establishing strong constraints on the radial velocity structure above the core-mantle interface in a manner analogous to surface wave observables constraining uppermost mantle structure. To estimate the frequency dependence of the slowness and decay rate caused by the diffraction process, we measure the relative arrival time and amplitude of both Pdiff and Sdiff arrivals for a series of frequency bands between 0.017 and 0.17 Hz and make corrections for bias introduced by ellipticity and velocity structure above the lowermost mantle. We present a new semi- automated technique using cluster analysis to remove low-quality waveforms and to accurately determine the slowness values. Comparison with 1D reflectivity and 3D SEM synthetics facilitates quantitative interpretation. Preliminary results confirm that there is measurable frequency dependence in the slowness and decay rate caused by diffraction for both Pdiff and Sdiff, indicative of significant radial velocity gradients. Variations in the frequency dependence are also observed with both geographic location and between Pdiff and Sdiff along the same azimuth suggesting lateral changes in Vp, Vs and Vp/Vs ratio and thus the physical state at the lowermost mantle.
DI31C-1814
Large Scale Exploration of the Core Mantle Boundary Region (D") with Broadband SKKS Data
We have combined concepts from inverse scattering (generalized Radon transform, or GRT) and statistics into an approach toward imaging structure near Earth's core-mantle boundary with large amounts of broad- band ScS and SKKS seismograms acquired by global seismograph networks. The GRT requires few restrictive a priori assumptions about the structures of interest, which makes it complementary to forward modeling approaches. Automation of the GRT and statistical modeling enables the imaging of D" structure over very large geographical regions. In previous studies we focused on the use of the ScS phase, which probes the D" region from the top [Wang et al., JGR, 2006; Ma et al., JGR, 2007;Van der Hilst et al., SCIENCE. 2007]. More recently, we have demonstrated that SKKS can also be used to constrain structure of the lowermost mantle [Wang et al., GJI, 2008]. Since SKKS samples the CMB region from below it is complementary to ScS. Moreover, the excellent (geographical) data coverage by SKKS, which in most regions is much better than that of ScS, paves the way for high-resolution D" imaging on a sub-global scale. With the GRT of subsets of a total of ~4 million radial component, broadband SKKS waveforms (from ~10,000 earthquakes recorded at one or more of ~4,000 seismograph stations) we aim to produce a hemisphere-scale map of D" structure from Asia, across the Pacific, to the Americas. We will present validation test with synthetic data and preliminary results of this ambitious mapping project.
DI31C-1815
Generalized Radon Transform Imaging of the Lowermost Mantle Beneath East Asia With ScS and SKKS Waveform Data
In earlier studies we developed generalized Radon transforms of ScS and SKKS data (separately) to image the lowermost mantle beneath Central America (Wang et al., JGR, 2006; Van der Hilst et al., Science, 2007; Wang et al., GJI, 2008). These GRTs map broad-band seismogram windows -- comprising the main arrivals and their coda and precursors -- into images that reveal multiple, piece-wise continuous (and statistically significant) interfaces in the lowermost mantle. Since the quality of the GRT imaging operator depends on the range of scatter angles sampled by the data, joint inversion of top-side (ScS) and under- side (SKKS) data would vastly increase the resolution of scatter interfaces. Therefore, as a logical follow- up of our earlier work, we developed a joint generalized Radon transform (GRT) of ScS and SKKS data for high resolution D" imaging in CMB regions well sampled by both ScS and SKKS. Once such region is Central America, but we have also started exploration of the core-mantle boundary region beneath East Asia. Global S and P wave (travel time) tomography (e.g. Li et al, GGG, 2008) reveals a large (seismically) faster-than-average structure in the lowermost mantle beneath East Asia. This structure, which is probably related to ancient episodes of subduction of oceanic lithosphere beneath mainland Asia, the western Pacific island arcs, and Indochina, is densely sampled both by ScS and SKKS waves. Based on data coverage and the tomographically inferred structures, we selected data associated with CMB reflection points in the region from 20°-60°N and 60°-130°E. We process ~380,000 ScS and ~260,000 SKKS data from ~9,000 earthquakes (mb > 5.0) recorded at one or more of a total of ~4,500 receivers. We will discuss the joint inversion and present preliminary 2D (and perhaps 3D) images of the D" region beneath Asia, which reveal very strong D" heterogeneity.
DI31C-1816
Detailed structure of the transition from the "Pacific Anomaly" to the surrounding mantle
Seismic tomography have revealed a broad, seismically low velocity anomaly in the lower mantle beneath the Pacific (we term it the "Pacific Anomaly"), surrounded by the circum-Pacific high velocity zone. Here, we constrain the detailed structure of the transition from the Pacific Anomaly to the surrounding high velocity zone along three cross sections on the basis of forward travel time and waveform modeling of the seismic data. Three cross sections include two great arcs across the Anomaly from New Zealand to Alaska sampling the transition in north and one from Tonga to China sampling the transition in northwest. Our collected dataset consists of direct S, Sdiff and ScS phases recorded in the Global Seismographic Network, the Canadian National Seismography Network, the Alaska Regional Network, the Chinese Capital Seismic Network and the Northern China Interior Seismic Project, from earthquakes occurring in the Samoa Islands, the Tonga-Fiji Islands and south of Fiji Islands. After corrected for the effects of earthquake mislocation and the seismic heterogeneities outside the Pacific Anomaly, seismic observations suggest that the transitions from the Pacific Anomaly to the surrounding region exhibit similarities and differences along the three cross sections. In both directions, the transition is characterized by a change of velocity structure from a low velocity area with a negative shear velocity gradient to a high velocity area with a shear velocity increase of about 2.0-2.5% at about 180-220 km above the CMB followed by a negative velocity gradient toward the CMB. Both transitional regions exhibit a 100-km low-velocity basal layer extending from the Pacific Anomaly to beneath the high-velocity region. In north, the low-velocity region has a negative shear velocity gradient with velocity reductions of 0% at 200 km above the core-mantle boundary (CMB) to -4.0% at the CMB; the high velocity area has velocity increases of 2.5% at 180 km above the CMB to 1.0% at the bottom; and the basal layer has a velocity reduction of -4.0%. In north, the low-velocity region has a negative shear velocity gradient from -3.0% at the top (730 km above the CMB) to -3.5% at 100 km above the CMB and an average shear velocity reduction of -5.0% in the bottom 100 km of the mantle, the high-velocity region has velocity increases of 2.0% at 220 km above the CMB to 1.0% at the bottom; and the basal layer beneath the high velocity region has a velocity reduction of -10.0%. The structural features and shear velocity structures of the transitions from the Pacific Anomaly to the surrounding region suggest complex interaction between the Anomaly and the surrounding mantle.
DI31C-1817
Investigating the Edges of the Large Low Shear Velocity Province in the Lowermost Mantle Beneath the Pacific Ocean
Whether the large low shear velocity province (LLSVP) in the lowermost mantle beneath the Pacific Ocean has a purely thermal origin (e.g., related to a superplume) or is a chemically distinct structure (e.g., a thermochemical pile) is still under active investigation. A number of seismic investigations have documented a sharp transition between the Pacific and African LLSVPs and surrounding lower mantle material, which argues for a chemically distinct origin. We previously used USArray recordings of Fiji-Tonga and Kemadec events to study the northern portion of the Pacific LLSVP, and found wave broadening consistent with multi- pathing along the LLSVP margin. In this study, we additionally compare the P-wave times to the SH times. All timing measurements were recalculated with a new travel time picking algorithm, which reduced scatter in our earlier estimates. Rapidly changing arrival times are observed, and are consistent with a sharp LLSVP boundary. We also define a waveform misfit parameter that quantifies each observation's deviation from a mean source shape on an event by event basis. Here we investigate a data set with direct S and P waves that bottom above and within the Pacific LLSVP, covering a significant portion of the central Pacific. A double array stacking method is employed to search for wave broadening as well as post-cursors associated with a possible top of the LLSVP. Some waveform complexities are observed, but vary geographically. These data will be discussed in regards to the possible nature of the LLSVP, its top, and the chemistry and dynamics of the deep mantle.
DI31C-1818
Fine-scale lateral variation in the physical properties of the ultra-low velocity zone beneath Philippine Islands
I report the lateral velocity and density changes of the ultra-low velocity zone (ULVZ) at the base of the mantle beneath the East of Philippine Islands through the analyses of ScP produced by intermediate and deep earthquakes in the southern part of the Philippine subduction zone and recorded at the Japanese short-period seismic networks of Hi-net and J-array. Large and very dense station distribution of both arrays enabled to observe the fine-scale variation of the ULVZ properties and the lateral transition to the normal mantle precisely. A clear postcursor was observed in the stacked ScP waveform sampled at the specific CMB region in the range of ~124 - 127°E and ~12 - 15°N. Several ScP stacks show a subtle signal 3 to 7 seconds preceding the ScP arrival. The relative timing of the precursor arrival with the ScP is in good agreement with the predicted SdP arrival when ~20 km thick ULVZ is assumed. The thickness of the layer decreases gradually from beneath Philippine Islands (~20 km) to eastern offshore region (~10 km). No pre- and postcursors were observed in ScP waveforms reflected from the surrounding CMB region. The discussion will be focused on the strength of the velocity and density heterogeneities within the layer, as well as the spatial extent of the ULVZ in the study area.
DI31C-1819
Evidence for a low velocity zone at the core-mantle boundary beneath the Gulf of Mexico
We observed a clear phase like arrival prior to the PKIKP wave at a broadband seismic array in eastern Tibet from an intermediate depth earthquake occurring in Guatemala. The measured incident angle and back azimuth of this phase indicate that it is originated from scattering near the core-mantle boundary (CMB) of the source side. This phase, however, was not observed from another earthquake that is only 60 km away, suggesting that scattering depends strongly on the angles of the incident waves. Ray tracing and diffraction migration indicates that the precursor is a large-angle reflection from a dipping structure in the lowermost mantle east of Mexico. The seismic reflector dips northward by ~50° and is centered at ~95.6° W and 25.3° N with an east-west extension of ~40km. A decrease of P-wave velocity by a few to ten percent is required to explain the amplitude and polarity of the phase. It is unlikely to explain the large P-wave velocity contrast and the large dipping feature with the post perovskite phase transition. The reflector is located in a region of the core-mantle region that is marked by a high velocity anomaly related to the subducted Farallon slab. Numerical modeling suggested that a substantial amount of hot mantle can be trapped beneath a slab over long periods of time, leading to formation of a mega-plume. Thus, the observed sharp dipping boundary here might be corresponding to the edge of a low velocity zone that has been interpreted as evidence for the presence of trapped mantle in the core mantle boundary.