Finite-Frequency Simulations of Core-Reflected Seismic Waves to Assess Models of General Lower Mantle Anisotropy

The core-mantle boundary (CMB) region is the site of the largest change in properties in the Earth. Moreover, the lowermost mantle above it (known as D″) shows the largest lateral variations in seismic velocity and strength of seismic anisotropy below the upper mantle. It is therefore vital to be able to accurately forward model candidate structures in the lowermost mantle with realistic sensitivity to structure and at the same frequencies at which observations are made. We use the spectral finite-element method to produce synthetic seismograms of ScS waves traversing a model of D″ anisotropy derived from mineralogical texture calculations and show that the seismic discontinuity atop the lowermost mantle varies in character laterally purely as a function of the strength and orientation of anisotropy. The lowermost mantle is widely anisotropic, shown by numerous shear wave splitting studies using waves of dominant frequency ~0.2-1 Hz. Whilst methods exist to model the finite-frequency seismic response of the lowermost mantle, most make the problem computationally efficient by imposing a certain symmetry to the problem, and of those which do not, almost none allow for completely general elasticity. Where low frequencies are simulated to reduce computational cost, it is uncertain whether waves of that frequency have comparable sensitivity to D″ structure as those observed at shorter periods. Currently, therefore, these computational limitations precludes the ability to interpret our observations fully. We present recent developments in taking a general approach to forward-modelling waves in D″. We use a modified version of SPECFEM3D_GLOBE, which uses the spectral finite-element method to model seismic wave propagation in a fully generally-elastic (i.e., 3D-varying, arbitrarily anisotropic) Earth. The calculations are computationally challenging: to approach the frequency of the observations, up to 10,000 processor cores and up to 2 TB of memory are needed. The synthetic seismograms can be directly compared to observations of shear wave splitting or other seismic phenomena and utilise all information from the waveform to accurately interpret D″ structures and elasticity. Using a recent model of mineralogical texture in the lowermost mantle (imposing no symmetry on the type on anisotropy), we model ScS waves traversing D″ in various regions. In this case, no lateral variations in average isotropic velocity exist, though the orientation and strength of anisotropy changes over a range of lengthscales (spherical harmonic degrees ≤128). We note a change in the amplitude (sometimes 0) and polarity (positive to negative) of arrivals which are reflected from the top of D″ (an arrival known as SdS) at ~300 km above the core-mantle boundary, even though no lateral variation exists between the isotropic overlying lower mantle and the anisotropic lowermost mantle. Supported by previous studies, this shows that changes only in anisotropy could be responsible for observed variations in SdS across the globe. Our approach can potentially be used to further model general elasticity at short wavelengths in any region in the Earth.