On mantle chemical and thermal heterogeneities and anisotropy as mapped by inversion of global surface wave data

We invert global observations of fundamental and higher-order Love and Rayleigh surface wave dispersion data jointly at selected locations for 1-D radial profiles of Earth's mantle composition, thermal state, and anisotropic structure using a stochastic sampling algorithm. Considering mantle compositions as equilibrium assemblages of basalt and harzburgite, we employ a self-consistent thermodynamic method to compute their phase equilibria and bulk physical properties (P, S wave velocity and density). Combining these with locally varying anisotropy profiles, we determine anisotropic P and S wave velocities to calculate dispersion curves for comparison with observations. Models fitting data within uncertainties provide us with a range of profiles of composition, temperature, and anisotropy. This methodology presents an important complement to conventional seismic tomography methods. Our results indicate radial and lateral gradients in basalt fraction, with basalt depletion in the upper and enrichment of the upper part of the lower mantle, in agreement with results from geodynamical calculations, melting processes at mid-ocean ridges, and subduction of chemically stratified lithosphere. Compared with preliminary reference Earth model (PREM) and seismic tomography models, our velocity models are generally faster in the upper transition zone (TZ) and slower in the lower TZ, implying a steeper velocity gradient. While less dense than PREM, density gradients in the TZ are also steeper. Mantle geotherms are generally adiabatic in the TZ, whereas in the upper part of the lower mantle, stronger lateral variations are observed. The retrieved anisotropy structure agrees with previous studies indicating positive as well as laterally varying upper mantle anisotropy, while there is little evidence for anisotropy in and below the TZ.