Deep Mantle Seismic Modeling and Imaging

Detailed seismic modeling and imaging of Earth’s deep interior is providing key information about lower-mantle structures and processes, including heat flow across the core-mantle boundary, the configuration of mantle upwellings and downwellings, phase equilibria and transport properties of deep mantle materials, and mechanisms of core-mantle coupling. Multichannel seismicwave analysismethods that provide the highest-resolution deepmantle structural information include network waveformmodeling and stacking, array processing, and 3D migrations of Pand S-wave seismograms. These methods detect and identify weak signals from structures that cannot be resolved by global seismic tomography. Some methods are adapted from oil exploration seismology, but all are constrained by the source and receiver distributions, long travel paths, and strong attenuation experienced by seismic waves that penetrate to the deep mantle. Largeand small-scale structures, with velocity variations ranging from a fraction of a percent to tens of percent, have been detected and are guiding geophysicists to new perspectives of thermochemical mantle convection and evolution. 91 A nn u. R ev . E ar th P la ne t. Sc i. 20 11 .3 9: 91 -1 23 . D ow nl oa de d fr om w w w .a nn ua lr ev ie w s. or g by U ni ve rs ity o f C al if or ni a Sa nt a C ru z on 0 4/ 28 /1 1. F or p er so na l u se o nl y. EA39CH04-Lay ARI 23 March 2011 23:58 INTRODUCTION Seismic waves generated by earthquakes and controlled sources provide the highest-resolution probes of Earth’s interior structure below the ∼10-km maximum depth achieved by drilling. Elastic P and S waves spread outward in all directions from surface or underground sources of energy release, with velocities controlled by the incompressibility, rigidity, and density of minerals and fluids in the interior. These P and S waves reflect and refract from abrupt changes in material properties, eventually arriving at Earth’s surface, where they can be recorded by seismometers (ground motion–recording instruments). Seismic wave travel time accumulates along the entire path traversed from source to receiver, providing integral constraints on seismic velocities along the path, whereas the existence and amplitudes of reflected phases constrain contrasts in material properties localized at internal boundaries between rocks of different properties or between rocks and fluids. Seismological characterization of the deepEarth involves a combined analysis of seismic wave travel times and amplitudes for direct and scattered Pand S-wave arrivals. In general, seismic wave–travel time analyses of Earth’s lowermost mantle resolve smooth, large-scale structures, whereas scattered-phase analyses resolve rough, small-scale structures and boundaries. The accuracy and detail to which the interior can be characterized depends strongly on the bandwidth of the recorded seismograms, the station spacing relative to the direction of wave propagation, the frequency of the waves, and the ray-path coverage of the medium provided by the spatial distribution of sources and seismometers. In the case of detailed seismic imaging of shallow crustal environments, land and/or oceanic deployments of thousands of seismometers and hundreds of controlled sources in two-dimensional (2D) distributions are used. Procedures for resolving both smooth and rough structures at shallow depths have been extensively developed for these dense data sets and provide the remarkable 2D and 3D crustal structural images used in oil and resource exploration, with resolution that approaches the scale of geological formations. As the targeted imaging depth increases, structural resolution diminishes as a result of (a) loss of signal bandwidth due to anelastic attenuation of the seismic waves and (b) reduced ray-path spatial sampling of the medium of interest due to significant limitations of the configurations of sources (typically earthquakes) and stations (typically land-based). Seismic waves from shallow sources that penetrate ∼2,890 km deep to Earth’s core-mantle boundary (CMB) travel immense path lengths thousands of kilometers long through attenuating and heterogeneous rock layers before being recorded at the surface, and they require a more expanded recording aperture compared with that used in crustal investigations to accommodate the higher-wave velocities and larger structural targets present at greater depths. Although deep mantle imaging cannot achieve the quality and resolution typical in oil exploration work, recent expansion of many seismic networks and ease of data availability have resulted in increased use of more sophisticated data-processing methodologies. In this review, we consider advances in our understanding of deepmantle structure over the past decade (see Garnero 2000 for a review of earlier work), emphasizing developments from the continuum of high-resolution methods ranging from waveform modeling to stacking and migration, all of which exploit systematic behavior of wave interactions with deep velocity heterogeneity. There are many motivations for determining detailed seismic velocity structures of the deep mantle. Although smooth, large-scale (>500–1,000 km) global mantle structures continue to be refined through the use of seismic wave–travel time tomography (e.g., Ritsema & van Heijst 2000, Gu et al. 2001, Grand 2002, Antolik et al. 2003, Zhao 2004, Panning & Romanowicz 2006, Takeuchi 2007,Kustowski et al. 2008,Li et al. 2008) and some consistency is emerging, particularly among S-wave velocity models, there are intrinsic ambiguities in attributing the smooth components of seismic velocity variations to effects of temperature, composition, or anisotropic fabric. 92 Lay · Garnero A nn u. R ev . E ar th P la ne t. Sc i. 20 11 .3 9: 91 -1 23 . D ow nl oa de d fr om w w w .a nn ua lr ev ie w s. or g by U ni ve rs ity o f C al if or ni a Sa nt a C ru z on 0 4/ 28 /1 1. F or p er so na l u se o nl y. EA39CH04-Lay ARI 23 March 2011 23:58 Simultaneous inversion for Pand S-wave velocities—and if possible, density—can overcome some of the ambiguities, but essential information about the geological processes (melting, shearing, folding, stratification, phase transitions, transport properties) is more directly manifested in the rough, fine-scale structural information revealed by seismic wave reflections and scattering, similar to the information manifested in a heterogeneous crustal environment. Fundamental attributes of the boundary layer at the base of the mantle such as the distribution of chemical heterogeneities, flow structures, and zones of partial melting can be only marginally sensed, if at all, by methods that map large-scale mantle structure. Detailed information can be extracted through the direct modeling of visible secondary arrivals or waveform distortions of principle seismic phases, as discussed in the next section. However, the seismic wave manifestations of complexities in the deep mantle often involve subtle features that can be difficult to detect and quantify. This motivates the use of robust procedures to extract weak arrivals amid a background of ambient noise, similar to procedures used in shallow seismic exploration applications (e.g., Rost & Thomas 2002). DEEP MANTLE SEISMIC MODELING AND IMAGING CHALLENGES We focus on determinations of structures in the deep mantle, by which wemean the∼1,000 km of the lower mantle directly overlying the CMB. The lowermost ∼200–300 km of the lower mantle is formally designated the D′ ′ region (Bullen 1949) because of its long-recognized departures from the seismic homogeneity expected for uniform compositionmaterial under self-compression. The D′ ′ region has generally been associated with a thermal boundary layer at the base of the mantle produced by heat fluxing out of the much hotter core, and possibly with a concentration of dense materials from the mantle (or light materials from the core) that have segregated throughout Earth history (Lay & Garnero 2004, Labrosse et al. 2007, Garnero & McNamara 2008, Trønnes 2009). Approximately a decade ago, studies established that some deep mantle structures may extend upward∼800–1,000 km from the CMBwith a distinctly red (long wavelength–dominated) heterogeneity spectrum (e.g., Ritsema et al. 1998, Ishii & Tromp 1999, Kellogg et al. 1999, van derHilst &Kárason 1999, Garnero 2000); we thus define this thicker region to be the deepmantle of interest here. Although basic principles of determining seismic velocity structure that hold for shallow crustal structure also apply in deep mantle studies, many distinct challenges confront efforts to resolve detailed deep structure. Foremost is the need to use earthquakes as sources of seismic waves that can penetrate deeply into the planet; this requirement greatly limits the spatial configuration and increases the variability of the sources relative to shallow controlled-source applications. Corresponding constraints on receiver locations tend to result in irregular global-network observatory configurations that are largely confined to continents that have large interstation spatial separations. Inmany investigations, regional networks or dense arrays of stationswith smaller separations have been used to study deep structure, but typically these have a small spatial aperture or relatively short site occupancy (18–24 months) and commonly provide limited azimuthal sampling of localized deep mantle regions of interest. As noted above, the signals sampling the deep mantle must propagate thousands of kilometers downward and upward through shallower rock layers. These signals lose high-frequency energy owing to intrinsic attenuation, which generally limits the maximum signal frequencies to less than ∼0.3–1 Hz. In the deep mantle, this implies Pand S-wave signal wavelengths no shorter than ∼7–40 km, values that intrinsically delimit resolution of any fine structures that may be present. One of the key strategies of exploration seismology is to strive toward sampling each position in the imaged medium with waves of different incidence angles and azimuths; this is possible to a limited degree in only a few regions of the deep mantle owing to the constraints on source and receiver distributions. Multiple ray-path sampling of the www.annualreviews.org • Deep Mantle Modeling and Imaging 93 A nn u. R ev . E ar th P la ne t. Sc i. 20 11 .3 9: 91 -1 23 . D ow nl oa de d fr om w w w .a nn ua lr ev ie w s. or g by U ni ve rs ity o f C al if or ni a Sa nt a C ru z on 0 4/ 28 /1 1. F or p er so na l u se o nl y. EA39CH04-Lay ARI 23 March 2011 23:58 same region also requires use of many earthquakes with variable source signals that must be equalized in the modeling. The great depth of lower-mantle structures, the tendency of most seismic imaging methods to preferentially sense quasi-horizontal boundaries, and the small contrasts in material properties involved (typically only a few percent except just above the CMB)motivate the combination of information fromprecritical angles of incidence (involvingweak reflections of seismic waves mainly sensitive to impedance contrasts) and postcritical angles of incidence (involving large but grazing triplications with phase-shifted reflections mainly sensitive to velocity contrasts). Seismicwave reflection and conversion coefficients from seismic velocity gradients and discontinuities that depend on ray geometry, material density, and seismic velocities produce complex wave behavior, which complicates imaging applications. Fully elastic wave calculations are required in most cases, whereas shallow applications can often utilize acoustic wave approximations. To address some of these imaging difficulties, lower-mantle studies have employed a diversity of seismic phases (Figure 1) that sample deep structure as fully as possible, notably including conversions of P and S waves that are usually ignored or suppressed in shallow exploration applications. Only a handful of portable instrument deployments have been designed primarily for deep mantle studies. However, the open availability of continuous broadband seismic data from international seismic observatories and portable instrument deployments provided by the Incorporated Research Institutions for Seismology (IRIS) and International Federation of Digital Seismograph Networks (FDSN) data centers, as well as seismic data from regional earthquake monitoring networks, enables substantial data sets to be brought to bear on deep mantle investigation, irrespective of what originally motivated the instrument deployments. Large seismogram data sets (as in Figure 2) have been essential for advancing the determinations of deep mantle structure considered here. We begin by briefly considering recent travel time investigations because they commonly frame the detailed modeling and imaging applications for the deep mantle. MULTIPLE-ARRIVAL TRAVEL TIME APPROACHES Almost all seismic wave investigations of deep mantle structure utilize a reference 1D model of seismic wave velocities and density as a function of depth, such as the Preliminary Reference EarthModel (PREM) (Dziewonski & Anderson 1981). 1D reference models have smooth velocity increases with depth across the lower mantle and, commonly, reduced rates of increase in the lower few hundred kilometers of the mantle (or even slightly negative velocity gradients). These are followed by a large, discontinuous drop in P-wave velocity, a vanishing of S-wave velocity, and a large increase in density across the CMB near 2,900-km depth. Even a simple 1D model predicts substantial complexity of the wavefield owing to topand bottom-side reflections and conversions of wave energy at the sharp CMB (Figure 1), along with diffractions of grazing wave energy (e.g., Garnero 2000). These complexities are predictable through the use of 1D model ray tracing and synthetic waveform computation methods that were developed in the 1970s. The methods provide reference signal predictions relative to which actual data complexities (like those in Figure 2) can be juxtaposed, and localized 1D, 2D, or 3D models can be developed. Seismic tomography usually exploits the travel time fluctuations of the major phases predicted for a 1D model to invert for perturbations relative to the reference structure; it does so through a set of basis functions (e.g., polynomials, splines, volume elements) to represent the 3D velocity perturbations in the medium. The deep mantle may be characterized as part of a global mantle inversion or a separate target of tomography desensitized to shallower structure by use of differential arrival times between direct phases (e.g., P, S, PKP, SKS) and CMB-reflected phases (e.g., PcP, ScS, PKKP, SKKS). Depending on the source-receiver paths involved, the travel time imaging may involve localized deep regions (e.g., Bréger et al. 2001, Saltzer et al. 2001, Wysession et al. 94 Lay · Garnero A nn u. R ev . E ar th P la ne t. Sc i. 20 11 .3 9: 91 -1 23 . D ow nl oa de d fr om w w w .a nn ua lr ev ie w s. or g by U ni ve rs ity o f C al if or ni a Sa nt a C ru z on 0 4/ 28 /1 1. F or p er so na l u se o nl y. EA39CH04-Lay ARI 23 March 2011 23:58