Differential t* measurements via instantaneous frequency matching: Observations of lower mantle shear attenuation heterogeneity beneath western Central America

We infer shear attenuation in the lower mantle by using the method of instantaneous frequency matching to calculate differential t* between core-reflected ScS and direct S (δt* _ScS-S). The instantaneous frequency at the envelope peak of a seismic phase is related to the average Fourier spectral frequency of that phase. To estimate δt* _ScS-S for a given trace, we first calculate the instantaneous frequency at the envelope peak of S and ScS. The trace is then attenuated through convolution with a suite of t* operators until the instantaneous frequency at the envelope peak of the seismic phase with the initially larger instantaneous frequency matches the value of the smaller instantaneous frequency from the initial calculation. The differential t* operator required to accomplish the match is then δt* _ScS-S. We also calculate δt* _ScS-S from the slope of the spectral ratio of windowed ScS and S. Both the spectral ratio and instantaneous frequency methods produce consistent results for high signal-to-noise ratio synthetic waveforms with S and ScS well separated in time, and where there are no other interfering phases. The instantaneous frequency method gives more stable results for low signal-to-noise ratio waveforms, and where S and/or ScS are affected by other interfering seismic phases. The instantaneous frequency matching method is applied to broadband data from South American earthquakes recorded in California that sample the lower mantle beneath Central America and the Cocos plate. δt* _ScS-S ranges from approximately -4 to 2 s, but are predominately negative, suggesting S is more attenuated than ScS for these data. We estimate the possibly contaminating effects of 3-D velocity heterogeneity on δt* _ScS-S through analysis of synthetic seismograms computed for a cross-section through a tomographically derived model of global shear wave heterogeneity, using an axisymmetric finite difference algorithm. Synthetics for path geometries of our data predict a δt* _ScS-S of ∼0.2 s. We investigate the effect of seismic anisotropy by comparing δt* _ScS-S before and after a subset of the data were corrected using splitting parameters obtained by linearizing the particle motion of the S and ScS phases. The rms error of the residuals between the corrected and uncorrected δt* _ScS-S is ∼0.2 s. Neither of these efforts, however, match the large negative observed δt* _ScS-S values, suggesting the mid-mantle beneath western Central America is in fact much more attenuating than the lowermost mantle below it, or S may be broadened by out-of-plane propagation effects, involving the remains of the Farallon plate containing stronger velocity heterogeneity than is imaged by seismic tomography.