Normal Mode Constraints on Elastic Structures in the Earth's Mantle

Normal mode observations of elastic parameters provide constraints on the origin and nature of the two lower mantle Large Low Shear wave Velocity Provinces (LLSVPs). Normal modes are the only seismic data sensitive to density perturbations, in addition to variations in shear-wave (Vs), compressional-wave (Vp) and bulk sound (Vc) velocity. In contrast to body waves, normal modes do not suffer from differences in data coverage throughout the globe, making them well suited for global mantle studies. Modes are ideal for constraining the ratio R(S/P) = dln(Vs)/ dln(Vp), which is generally found to be greater than 3 in the lower mantle (e.g. Romanowicz, 2001; Koelemeijer et al., 2016), larger than expected for an isochemical mantle without phase transitions (Karato & Karki, 2001). Furthermore, anti-correlation of dln(Vs) and dln(Vc), observed by many lower mantle studies (e.g. Su & Dziewonski, 1997; Trampert et al., 2004), may be an indicator of chemical heterogeneity. Finally, 3D density variations allow to discriminate between dense stable mantle anchors and light buoyant superplumes. Denser than average LLSVPs have been observed previously (e.g. Ishii & Tromp, 2004; Lau et al., 2017), but observations of light LLSVPs exist as well (Koelemeijer et al., 2017). Here, we use normal mode data to develop tomographic models of 3D variations in Vs, Vp, Vc and density. Normal mode spectra can be inverted in two ways, using either 1) a direct spectrum one-step inversion or 2) a two-step inversion with splitting function measurements as intermediate step. So far, studies of R(S/P), the anti-correlation of Vs and Vc, and density have been performed with the two-step inversion. The one-step inversion performs better in terms of average spectral misfit in an inversion of only Vs (Jagt & Deuss, 2021). In our preliminary results, we observe a decrease in lower mantle R(S/P) for the one-step inversion with respect to the two-step inversion, dropping below values predicted for a chemically heterogeneous mantle. For density, we observe a dense basal layer that could reconcile previous observations of both light and dense LLSVPs. We will explore the effects of the chosen inversion method on the above-mentioned parameters and on the interpretations of chemical heterogeneity we draw from our models.