Solid Inner Core of the Earth

Aimin Cao, Barbara Romanowicz, and Nozomu Takeuchi "An Observation of PKJKP: Inferences on Inner Core Shear Properties", Science, published online 14 April 2005



  The seismic phase PKJKP, which traverses the inner core as a shear wave, and would provide direct evidence for its solidity, has been difficult to detect. Using stacked broadband records from the Gra"fenberg array in Germany, we document a high signal to noise phase, whose arrival time and slowness agree with theoretical predictions for PKJKP. The back-azimuth of this arrival is also consistent with predictions for PKJKP as is the comparison with a pseudo-liquid inner core model. Envelope modeling of the PKJKP waveform implies a slightly larger shear velocity gradient with depth in the inner core than that in PREM model.

 Download pdf reprint

 Download pdf supplementary material


Press Coverage

  Discovery Channel: Earthquake Detector Finds Solid Earth Core

  NewScientist: Earth's Solid Centre Confirmed
  LiveScience: Finally, a Solid Look at Earth's Core
  MSNBC News: Finally, a Solid Look at Earth's Core

  SPIEGEL: Erde hat einen harten Kern

  DIE WELT: Der Kern der Erde ist fest

  Gazeta: Jak trzesie, to przeswietla

  Noorderlog: PKJKP!

  Deutsche Welle: Earth Has a Solid Heart

  KOMPAS_Cyber_Media: Inti Bumi Diperkirakan Padat

For all enquiries, please contact Aimin Cao:


or Barbara Romanowicz




(A) Ray paths of PKJKP and PKIKP. The star and square indicate the source and GRF array locations, respectively. (B) The theoretical amplitude ratio of PKJKP over PKIKP as a function of frequency based on the reference model PREM, after correcting for transmission and geometrical spreading. The reference epicentral distance is $138^o$. Given the dynamic range of present seismometers, it is unlikely that one could observe PKJKP (or pPKJKP) in the frequency range $ \sim 1.0$ Hz {\it (4)}. (C) Geographical setting of the event (dot) and GRF seismic array (square). The solid line is the ray path of PKIKP and the dashed line is the ray path of PKJKP projected on the earth's surface. The triangle marks the location of the bottoming point of PKJKP in the inner core. The upper-right inset shows the source time history of the event characterized by a P phase recorded at a broadband station (YAK, distance = $80.1^o$) of the Global Seismographic Network, located in a similar azimuth as GRF. The lower-left inset illustrates the P-wave radiation pattern in the vertical plane of the great circle. This event is exceptional: (i) the source duration is less than 9 seconds; (ii) the expected PKJKP is emitted from the top of the lobe of the P-wave radiation pattern; (iii) the potential interfering phases identified in previous studies {\it (4)(5)}, such as PcPPKIKP, pPcPPKIKP, sPcPPKIKP, and PKKPdf, are at least 17 seconds away from the predicted PKJKP arrival time (according to PREM).


(A) Observed vespagram for PKIKP+PKiKP and their depth phases (the energy level is amplified 1.6 times). The center of the energy maximum is for a slowness of $\sim 1.9 s/deg$, which is the average of slownesses of PKIKP (1.85 s/deg) and PKiKP (2.04 s/deg) predicted from PREM {\it (7)}. The following weaker energy maximum corresponds to pPKIKP+pPKiKP, and has the same slowness, as predicted from PREM. (B) Stacked waveforms for PKIKP+PKiKP and their depth phases for the energy maximum in (A). (C) Observed vespagram for the potential PKJKP (energy level is amplified 40 times). The slowness of the energy maximum is $\sim -1.6 s/deg$, close to the PREM prediction of -1.43 s/deg. The arrival time is also compatible with PREM (1695 sec for the maximum energy, compared to a prediction of 1690 sec for the high frequency onset of the pulse). (D) Stacked waveform corresponding to the energy maximum in (C). (E) Vespagram in the back-azimuth and travel time domain. This shows the direction of arrival of the detected energy, which we identify as PKJKP, in the negative slowness range of Fig. 2C. The estimated back-azimuth is $\sim 223^o$, which shows that the observed energy propagates along the major arc from the source (the expected back-azimuth of PKJKP is $218.^o$). This indicates that the observed phase is not a near-array scattered phase, and provides additional evidence for its identification as PKJKP.

Synthetic modeling. (A) Waveform modeling of PKIKP+PKiKP as well as pPKIKP+pPKiKP based on USGS PDE moment tensor. Both observed (dashed line) and synthetic (solid line) seismograms are normalized after applying the bandpass filter (0.06-0.1 Hz). Synthetics are obtained using DSM {\it (13)}. (B) Synthetic differential seismogram for the PREM model compared to a true liquid inner core, for which the shear wave velocity is equal to zero. A (PcPPKIKP), B (pPcPPKIKP+sPcPPKIKP), and C (PKKPdf) are artificially enhanced by the assumption of liquid inner core. (C) Synthetic differential seismogram based on the pseudo-liquid inner core used in this paper. We now can clearly see both PKJKP and pPKJKP phases. The amplitude of PKJKP is approximately 2.2 times larger than that of pPKJKP.

Synthetic vespagrams. (A) Pseudo-liquid inner core model. Time windows are identical to those in Figure 2C. Energy level is amplified 40 times, as in Fig. 2C. D, E, and F are crust, mantle, or outer core phases, and G is PcPPKIKP. See Fig. S3 for a plot with energy level amplified only 20 times to bring out the relative strength of these phases. (B) Solid inner core model, assuming $Q_\beta =300$. Because the strong mantle phase E in the synthetic model arrives at the same time as PKJKP, the dominant energy of phase E hides the much weaker PKJKP, which only slightly distorts the pattern of phase E. Likewise, pPKJKP slightly distorts the pattern of phase F. Phases E and F are not present in the observed stacks. Therefore, we cannot directly use the comparison of observed vespagram to that predicted by the solid inner core model, and instead, we use a differential seismogram modeling approach, in which the energy from phases E and F is removed. (C) Synthetic differential vespagram in the slowness-time domain. This vespagram is calculated for the solid inner core minus the pseudo-liquid inner core models. The time window is the same as that in Fig. 3C. The estimated slownesses of the energy maxima are both -1.4 s/deg, as are the predictions based on PREM. This identifies the two phases in the differential seismogram (Fig. 3C) as PKJKP and pPKJKP.