Anomalously low amplitude of S waves produced by the 3D structures in the lower mantle

•S and Sdiff waves with anomalously low amplitude are observed.•Defocusing effect of low Vs structures at the base of the mantle is likely the cause.•The previously proposed thin UVLZ partially explains the amplitude anomalies.•A large low Vs region placed adjacent to ULVZ contributes to explain observations.•Amplitude anomalies are useful measurements to constrain the 3D slow structure.

Direct S and Sdiff phases with anomalously low amplitudes are recorded for the earthquakes in Papua New Guinea by seismographs in northern America. According to the prediction by a standard 1D model, the amplitudes are the lowest at stations in southern California, at a distance and azimuth of around 95° and 55°, respectively, from the earthquake. The amplitude anomaly is more prominent at frequencies higher than 0.03 Hz. We checked and ruled out the possibility of the anomalies appearing because of the errors in the focal mechanism used in the reference synthetic waveform calculations. The observed anomaly distribution changes drastically with a relatively small shift in the location of the earthquake. The observations indicate that the amplitude reduction is likely due to the 3D shear velocity (Vs) structure, which deflects the wave energy away from the original ray paths. Moreover, some previous studies suggested that some of the S and Sdiff phases in our dataset are followed by a prominent postcursor and show a large travel time delay, which was explained by placing a large ultra-low velocity zone (ULVZ) located on the core-mantle boundary southwest of Hawaii. In this study, we evaluated the extent of amplitude anomalies that can be explained by the lower mantle structures in the existing models, including the previously proposed ULVZ. In addition, we modified and tested some models and searched for the possible causes of low amplitudes. Full 3D synthetic waveforms were calculated and compared with the observations. Our results show that while the existing models explain the trends of the observed amplitude anomalies, the size of such anomalies remain under-predicted especially at large distances. Adding a low velocity zone, which is spatially larger and has less Vs reduction than ULVZ, on the southwest side of ULVZ, contributes to explain the low amplitudes observed at distances larger than 100° from the earthquake. The newly proposed low velocity zone mostly overlaps with the northern part of the Pacific large low shear velocity province (LLSVP) revealed in tomographic models. Although the very low amplitudes observed at a distance of about 95° remain unexplained, our results indicate that the boundary of the Pacific LLSVP is sharp, and the amplitude of S waves at these large distances is lowered by strong vertical and/or lateral deflection at the boundary toward the interior of the low velocity province.