Here we perform an extensive search of temporal change through the seismic data and report temporal changes of seismic data sampling the lowermost mantle between a pair of nearly co-located earthquakes occurring on 2000/07/07 and 2009/12/17. We infer temporal changes of seismic properties in the lowermost mantle based on waveform modeling of the seismic data and discuss implications of these discovered temporal changes of seismic properties to the dynamics and mineral physics in the lowermost mantle.
We search possible seismic signal associated with temporal change of seismic structure in the Earth's lowermost mantle through the seismic data of global earthquake doublets (Methods). An earthquake doublet is two earthquakes occurring at close locations (<1 km) but at different times. The relative seismic signal changes at a station between a doublet are only sensitive to the relative doublet event parameters and temporal change of seismic structure along the seismic ray paths between the occurring times of the doublet. As the seismic effects of relative doublet event parameters can be accurately accounted for, doublet analysis has been an important tool for detecting temporal changes of seismic structures near the earthquake sources and in the Earth's inner core. We find no credible temporal change of seismic waves sampling the lowermost mantle in the seismic data of all identified doublets spanning from 1993 to 2023 (Methods), except in one doublet AI-0009 occurring on 2000/07/07 (event 2000) and 2009/12/17 (event 2009) beneath the Aleutian Islands (i.e., doublet AI-D7 in the Supplementary Data).
Evident temporal changes are observed between the doublet AI-0009 data of SKS and S-Scd-ScS phases at an epicentral distance of 94.9° recorded at seismic station IU.SDV in Santo Domingo, Venezuela (Fig. 1). SKS is a seismic wave traveling as a shear wave in the mantle and a compressional wave in the Earth's outer core, while S-Scd-ScS phases are shear waves traveling through the lowermost mantle with S and Scd turning above and in the D'' region respectively and ScS reflected at the CMB (Fig. 1b). We illustrate those temporal changes by superimposing the seismic data between the doublet after corrections for the effect of relative doublet location and origin time. Any mismatch of relative arrival time and shape in those superimposed doublet waveforms reflects the effect of temporal change of seismic structure between the occurring times of the doublet. Superimposed SKS waveform pair recorded by IU.SDV exhibits evident differences in both waveform shape and travel time, with a broader wavelet in event 2009 than in event 2000 and the phase arriving 160 ms earlier in event 2009 than in event 2000 (left panel, Fig. 1c). Superimposed S-Scd-ScS waveform pair recorded by IU.SDV exhibits a misalignment of relative travel time in the radial component with the seismic signals in event 2009 changing from 80 ms earlier in the beginning portion of the wavelet to 30 ms earlier in the middle of the wavelet to no discernable change in the later portion of the wavelet. In contrast, no difference is observed in the transverse components of the doublet data in either waveform shape or travel time (right panel, Fig. 1c). These temporal changes of SKS and S-Scd-ScS phases between the doublet are consistently observed at the frequency bands of good data quality from 0.4 Hz to 2.0 Hz. No other discernible temporal changes of S-wave waveform and travel time are observed between the doublet that can be confidently attributed to temporal change of seismic structure in the lowermost mantle (Fig. 2, Supplementary Information).
The waveform changes observed in the IU.SDV SKS observations and the relative travel time difference observed in the radial S-Scd-ScS waveforms cannot be explained by the relative doublet location and origin time, as those relative differences of source parameters cannot generate different SKS waveform shapes between the events and a relative time difference in the radial components of S-Scd-ScS phases accompanying with no relative travel time difference in the transverse components of the waveforms between the doublet. The observed temporal changes of the seismic data cannot be explained by possible differences of source focal mechanisms between the doublet, as no waveform shape changes are observed in S-Scd-ScS phases. Because SKS and S-Scd-ScS phases have almost identical take-off angles from the seismic sources, any focal mechanism difference between the doublet would generate similar changes of waveshapes between the two phases. Synthetic tests also indicate all possible source mechanisms within the uncertainties of the GCMT solution produce little waveform and travel time differences of the seismic phases between the doublet, further excluding the earthquake sources as the origin of the observed SKS and S-Scd-ScS changes between the doublet (Supplementary Fig. 1). The observed temporal travel time changes of the IU.SDV SKS and S-Scd-ScS phases cannot be caused by clock errors of the station, as no travel time differences are observed in other seismic phases of the station, including the subsequent pulses after SKS arrivals (the phases in a time window of 1435-1441 s on the left panel in Fig. 1c) and the transverse S-Scd-ScS phases. And, the observed temporal changes of IU.SDV observations are not the result of seismic noise. Synthetic analysis shows that random noises in the seismic data at station IU.SDV would generate negligible differences in the waveform and relative travel time of both SKS and radial S-Scd-ScS phases between the doublet (Supplementary Fig. 2). We conclude that the observed IU.SDV waveform and travel time differences between the doublet are caused by temporal changes of seismic structure between the occurring times of the doublet.
The observed IU.SDV temporal change of seismic data cannot be explained by a change of shallow seismic structure between the doublet. Because S-Scd-ScS phases have similar take-off angles from the earthquake source and similar incident angles to the station, a change of shallow structure beneath the station would cause a nearly constant travel time shift in the seismic signals of S-Scd-ScS phases, different from the observed varying arrival time changes from 80 ms in the beginning portion of the wavelet to no discernable change in the later portion of the wavelet. Changes of shallow structure also cannot explain the travel time shift of SKS, as the subsequent energy pulse after SKS exhibits no temporal change between the doublet.
The temporal changes of SKS travel time and waveform at IU.SDV are the result of structural changes in the lowermost mantle between the occurring times of the doublet. In the entry point of the IU.SDV SKS phases at the CMB, several studies have suggested possible presence of ULVZs in the lowermost mantle. ULVZs are small-scale structures of tens of kilometers high and hundreds to thousands of kilometers wide with very low seismic velocities. ULVZs are proposed to be the regions of partial melt and are the most likely candidate structures in the lowermost mantle that could experience seismically detectable changes in a decadal time scale. We thus suggest a temporal change of a ULVZ near the SKS entry point at the CMB for the explanation of the observed temporal changes of SKS phases in IU.SDV. We should point out that a change of a ULVZ structure near the SKS exit point at the CMB could explain the seismic data equally well, and the modeling results would equally apply to a changing ULVZ near the SKS exit point at the CMB. Because the shape of ULVZ cannot be well constrained with the limited seismic data, we test ULVZs with a simple edge and explore how the change of the simple edge explains the observed temporal changes of SKS waveform and travel time. Our results would depend on the non-uniqueness of the ULVZ geometry we assume, but modeling with this simple geometry allows us to explore the first-order feature and length scale of morphological change of the ULVZ, as those parameters are not significantly dependent on the actual geometry of the ULVZ. Detailed synthetic modeling (Methods) indicates that both a ULVZ with its edge shrinking and a ULVZ with its edge laterally moving could explain the seismic data (Fig. 3a). The length scale of lateral change of ULVZ edge is 16-45 km for a shrinking ULVZ and is 21-27 km for a moving ULVZ. The shrinking ULVZ model could explain the seismic observations slightly better than the moving ULVZ model, but both types of models are acceptable in their synthetic fitting to the doublet observations (Fig. 3a, Supplementary Fig. 3). ULVZ shrinking and moving in both sides of the SKS CMB entry point could explain the seismic data equally well.
The observed temporal change in the radial components of S-Scd-ScS waveforms at station IU.SDV is the result of temporal change of SV-wave velocity in the D'' region between the occurring times of the doublet. As no discernable changes are observed in the transverse components of the S-Scd-ScS phases, the SH-wave velocity should have no changes between the doublet. We thus search the best-fit changed SV-wave velocity model in the lowermost mantle that would explain the observed temporal change in the radial components of the S-Scd-ScS waveforms. Note that the observed early arrival of the SV energy in event 2009 in the beginning portion of the S-Scd-ScS waveforms would require an SV-wave velocity increase in some depths above the CMB (i.e., the top portion of the D'' region), while the no temporal change of ScS phase in the latter portion of the S-Scd-ScS waveforms would require a decrease of SV-wave velocity close to the CMB (i.e., the bottom portion of the D'' region) to offset the effects of the SV-wave velocity increase required by the S-Scd data in the top portion of the D'' region. Detailed synthetic modeling (Methods) shows that the observed temporal change of the radial components of the S-Scd-ScS waveforms can be best explained by an SV-wave velocity change from an increase of 0.10%-0.13% at the D'' discontinuity to 0% at the middle depth of the D'' layer to a decrease of 0.13%-0.18% at the base of the mantle between the occurring times of the doublet (Fig. 3b and Supplementary Fig. 4).