| In this thesis I study the velocity and attenuation structure of the North American mantle using multiple-frequency shear-wave and Love-wave measurements, together with finite-frequency sensitivity kernels.The software for dynamic ray tracing and fast computation of body-wave finite-frequency sensitivity kernels is described and extensively validated and tested for accuracy. The program works for arbitrarily defined phases and one-dimensional background models. In kinematic and dynamic ray tracing, an integration step size of about 20 km is needed to produce travel-time errors under 0.1 s for the most common seismic phases. In kernel computation, a minimum integration step size of 10--30 km is sufficient to obtain numerical errors of the kernel's spatial quadrature below observational uncertainties. Larger errors may occur for long-period minimax phases such as SS . The paraxial approximation fails and errors become intolerable at epicentral distances larger than 140°.A global data set is built to contain multiple-frequency SH-wave travel-time and amplitude anomalies and SS-wave differential delays, estimated by band-pass filtering and cross-correlation. Most of the data are recorded at USArray stations. Frequency dependence is observed for all three types of data, and is strongest for amplitudes. The shallow structure is constrained by the addition of Love-wave phase delays.Velocity and attenuation heterogeneities are simultaneously estimated by allowing for focusing. The velocity model shows evidence of heavy fragmentation of the Farallon slab, including two separate subduction systems under western and eastern North America respectively, trench-perpendicular slab tears, and blob-like slab fragments in the lower mantle. The velocity model reveals a lower-mantle plume originating at about 1500 km depth beneath the Yellowstone area and tilting about 40° from vertical. Complex interaction between the plume and slab fragments is observed. High correlation coefficients between velocity and attenuation heterogeneities beneath the Central and Eastern U.S. suggest one physical source, most likely temperature, dominant variations. The smaller correlation coefficients and larger deltalnQS-deltaln VS slopes under the Western U.S. suggest an influence of non-thermal factors such as the existence of water and partial melt.The benefits of the methodological improvements are investigated. Amplitude data help to sharpen the edges of narrow velocity heterogeneities in the shallow upper mantle. The focusing effect dominates over the attenuation effect in interpreting amplitude anomalies. The addition of Love-wave phase delays helps to improve the resolution of both velocity and attenuation, and the effect is noticeable even in the lower mantle. |
