The presence of anisotropy in the inner core seems today widely accepted,
although its exact strength and distribution remains the subject of debate.
Unfortunately, due to imperfect path coverage, a large portion of the inner
core remains unsampled by quasi-polar PKPdf, which are used
in body-wave studies, and models of anisotropy often rely on a small number of
measurements for rays quasi-parallel to the earth's spin axis.
In the light of our work on the presence of large heterogeneity
at the base of the mantle (*Bréger and Romanowicz,* 1998),
we recently argued that the contamination of differential residuals by
deep mantle structure for those paths could have been underestimated,
yielding unrealistically large magnitude of inner core anisotropy
(*Bréger et al.* [1999,2000]). It has become critical to seek new
observables that will provide additional constrains on the seismic
structure of the deepest shell of the earth, and here, we present some
preliminary results obtained using P'P' waves.

Because of its two legs in the inner core, the P'P'df phase is very sensitive to structure below the Inner Core Boundary (ICB). Earthquakes in the Aleutians recorded
at the NORSAR array in Norway, and in the South Sandwich Islands recorded
at station CASY, correspond to paths for which the epicentral distance is between
50 and 65^{o}, which is an appropriate distance to observe P'P', and
for which the average angle
between the two legs of P'P'df in the inner core and the earth's spin axis
is on the order of 16^{o}. Models of inner core anisotropy derived from travel times
of body waves have been so far based on datasets with very few observations
for
smaller than about 25^{o}. This particular source geometry
is therefore well suited to provide critical additional constrains on the inner core
anisotropic structure.

In Figure 28.1, we present an example of short-period vertical component data
for an earthquake near Kodiak Island recorded at 23 stations of the NORSAR array.
Three distinct arrivals are clearly visible around the predicted arrival times of P'P'df, bc, and ab and can be attributed to P'P'df, bc, and ab for several reasons.
There is no other large enough teleseismic event that could be responsible for these arrivals. There is very little seismicity around the NORSAR array, and a small local event would generate
waves with very different slownesses. The average observed amplitude ratio between P'P'bc and
P'P'df and P'P'bc and P'P'ab are compatible with theoretical values, and, as expected, P'P'df and P'P'bc on the one hand, and P'P'ab and the Hilbert transform of P'P'df on the other hand, have

similar waveforms (Figure 28.1b-c). For an event of this type (Mb=6.1 Depth=17.3 km), the combination of a source duration of several seconds and of depth phases is expected to lead to a complex P coda. P'P'df should not critically affect P'P'bc because of the large P'P'bc/P'P'df ratio, which implies that the first arrival of P'P'bc has an amplitude which is well above that of the late P'P'df coda. However, P'P'bc could potentially perturb the P'P'ab waveform, yielding larger uncertainties in the measurements of P'P'ab travel times. We also verified that arrival slownesses measured using cross-correlation were consistent with theoretical values.

Measuring travel time residuals on individuals
seismograms with an accuracy less than 1s can be problematic, because
P'P' first motions are usually hidden by noise. In order to reduce as much as possible the uncertainty on travel time measurements, we used a standard *N*^{th} root stacking
method with N=3.

We were also able to measure P'P'df and P'P'bc - P'P'df residuals for an event
in the South Sandwich Islands recorded at station CASY (-66.28^{o}N,+110.53^{o}E),
which corresponds to an average angle
of about 14.6^{o}.

Our P'P' observations corresponds to PKPdf legs that regions of the inner core previously identified as anisotropic, and in Figure 28.2 we compare our observations with two simple axisymmetric models of inner core anisotropy, with strength of 1.5 and 3.5%, respectively. Both models clearly overpredict the P'P'bc - P'P'df observations.

The apparent discrepancy between P'P' and earlier observations could stem from an exceptionally
complex inner core structure, with very sharp anisotropic gradients
and lateral contrasts (*Bréger et al.*, 1999), and possibly produced by a complex
pattern of convection associated with a strong intrinsic
anisotropy. Following some of our earlier studies, we believe that a strong
contribution of D" could also explain this discrepancy.

To conclude, P'P' seems a very promising tool to investigate
the structure of the inner core, and preliminary results
confirm the results that we obtained using PKP(bc-df) differential
travel times (*Bréger et al.*, 1999).

We are very grateful to the NORSAR team, and in particular to Drs Johannes Schweitzer and Joergen Torstveit for providing us with their data. We also thank Johannes Schweitzer and two anonymous reviewers for helpful comments. This work was partially supported by IGPP/LLNL grants # 99-GS013 and 00-GS010. It is BSL contribution 00-05.

Bréger, L., and B. Romanowicz, Three-dimensional structure at the base
of the mantle beneath the central Pacific, *Science, 282*, 718, 1998.

Bréger, L., B. Romanowicz, B, and H. Tkalcic, PKP(BC-DF) travel time residuals
and short scale heterogeneity in the deep earth, *Geophys. Res. Lett.*, *26*, 3169, 1999.

Bréger, L., H. Tkalcic, and B. Romanowicz, The effect of D" on PKP(AB-DF) travel time residuals and possible implications for inner core structure, *Earth Planet. Sci. Lett.*, *175*, 133, 2000.

Kennett, B.L.N., E.R. Engdahl, and R. Buland, Constraints on seismic
velocities in the Earth from travel times. *Geophys. J. Int.*, *122*,
108, 1995.

The Berkeley Seismological Laboratory, 202 McCone Hall, UC Berkeley, Berkeley CA 94720

Questions and comments to www@seismo.berkeley.edu

Copyright 2000, The Regents of the University of California.