Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Geophys. J . Int.

(1994) 119, 195-218

Shear-wave splitting in the mantle of the Pacific


Vkronique Farra’ and Lev Vinnik2
‘Laboratoire de Sismologie, Institut de Physique du Globe de Paris, 4 Place Jussieu, F-75005 Paris, France
‘Institute of Physics of the Earth, Moscow 123810, Russia

Accepted 1994 March 12. Received 1994 February 13; in original form 1993 July 24

SUMMARY

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


The bulk of previously known data on seismic anisotropy in the suboceanic mantle
was obtained from observations of azimuthal variations of the surface-wave and P,
velocities. Our study of azimuthal anisotropy in the mantle of the Pacific is based on
observations of shear-wave splitting. The idea of our method of measurement
(Vinnik & Farra 1992) is to analyse evolution of the waveform of the long-period
multiple reflected ScS and SSCSphases as a function of the order of reflection. The
estimates of anisotropy correspond to circular regions with a radius of several
hundred kilometres around the surface bounce points. The technique was applied to
the digital recordings of deep and intermediate events from the circum-Pacific
seismic belt at seismograph stations in and around the Pacific. The estimates of
polarization direction of the fast split wave and time lag of the slow wave are
obtained for a few regions of the Pacific basin, marginal basins and the Andes.
In the back-arc extensional basins whose spreading has continued to the present
time or ended later than about 15Ma, the inferred direction of extension in the
mantle is close to that imprinted in the crust (Lau basin, Japan Sea). In the back-arc
basins whose active development ended 15-20 Ma or earlier (South Fiji and Parece
Vela basins), the inferred directions of extension in the mantle are in strong
disagreement with the crustal data. In the Central Andes, the region of plate
collision and low-angle subduction, the inferred direction of flow in the mantle is
perpendicular to the plate boundary, contrary to the regularity previously found in
several other collisional belts. The fast direction of anisotropy in the mantle beneath
the southwestern part of the 100Ma old Pacific plate is very close to the absolute
plate motion direction; the magnitude of anisotropy corresponds to the lowest
values found in the continental mantle and is much lower than in most of the
marginal basins and the Andes. The weakness of this, presumably asthenospheric,
anisotropy in comparison with asthenospheric anisotropy under some old continen-
tal plates is probably related to a difference in upper mantle viscosities.
Key words: Andes, anisotropy, mantle, Pacific region, shear-wave splitting.

1 INTRODUCTION [1001 is horizontal, its direction coincides with the fast


direction of propagation of the Rayleigh and P, waves and
Seismic anisotropy in the mantle is of interest for the with polarization direction of the fast split wave in the
geoscientist because it is related to the lattice preferred vertical propagating S wave.
orientation of olivine which, in its turn, depends on finite Previous estimates of azimuthal anisotropy in the
strain. Axes [loo], [OlO] and [OOl] in olivine tend to become uppermost part of the oceanic mantle were obtained mainly
aligned with the longest, shortest and intermediate axes of from the P, data (e.g. Hess 1964; Bibee & Shor 1976;
strain ellipsoid, respectively (McKenzie 1979; Ribe 1992). Shimamura 1984; Shearer & Orcutt 1985, 1986) and, for the
Axis [lo01 is aligned with the direction of mantle flow if that deeper part of the mantle, from the phase velocities of
is in the form of progressive simple shear. This kind of flow long-period surface waves (e.g., Forsyth 1975; Mitchell &
is expected to arise below the lithosphere at a fast spreading Yu 1980; Tanimoto & Anderson 1984, 1985; Cara &
ridge, above a subducted slab, and in the convective Leveque 1988; Nishimura & Forsyth 1988, 1989; Montagner
boundary layer under the lithosphere (Ribe 1989). If axis & Tanimoto 1990, 1991). Most of these estimates

195
196 V . Farra and L. Vinnik

correspond to the mantle of the Pacific. They indicate that


the fast direction at the top of the mantle underneath the
old oceanic crust aligns with the fossil spreading direction in
the lithosphere. In the depth range around 200 km the fast
direction tends to be orthogonal to the mid-oceanic ridges
and thus is related to the present-day plate tectonics. The
drawback of the surface-wave data is not only their
relatively low lateral resolution (in the range of a few
thousand km) but also the trade-off between azimuthal
anisotropy and lateral heterogeneity of the isotropic
medium (see e.g. Babuska & Cara 1991).
An unambiguous manifestation of azimuthal anisotropy is
shear-wave splitting. Observations of shear-wave splitting in
the short period (1-2s) ScS phases (S reflections from the

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


core-mantle boundary) are known for some stations
surrounding the Pacific (Ando 1984; Fukao 1984). The most
useful method for measuring shear-wave splitting in the
mantle is based on the analysis of intermediate-period CORE
recordings of the SKS and SKKS seismic phases (Vinnik,
Kosarev & Makeyera 1984). Ellipticity of particle motion in Figure 1. Ray paths of ScS, (dash line) and ScS, (solid line).
the horizontal plane in these phases is determined
practically only by anisotropy underneath the receiver which
is generally not true for the other shear phases. The SKS The SV, and SH, motion components are determined by
techniques were successfully applied in recent continental the source radiation pattern and also by anisotropy and
studies (for reviews see, e.g. Silver & Chan 1991; Vinnik et heterogeneity in the source region; we assume that these
al. 1992) but they are less useful in the oceans because the components are similar for ScS, and ScS,. K, is the
oceanic stations are few and, usually, noisy, in the period reflection coefficient for S V at the core-mantle boundary; it
range of interest, between 3 and 15 s. depends on the angle of incidence and, consequently, on the
In this study we use the method which has much in order of reflection. F, is the matrix of response for the
common with the SKS techniques but is based on the anisotropic zone in the bounce point and F, is the similar
analysis of evolution of particle motion in the family of matrix for the anisotropic zone at the receiver. For
ScS,, multiple shear-wave reflections from the core-mantle simplicity we neglect the amplitude difference between ScS,
boundary and the Earth’s surface (Vinnik & Farra 1992). In and ScS, due to spreading of the wavefront and anelastic
the period range longer than 20s these phases are attenuation, since this difference should not affect the shape
sometimes well recorded even in a noisy near-shore of the particle motion. We also neglect the difference
environment. The technique allows the mapping of between the coefficients of reflection of S V and SH at the
azimuthal anisotropy in the region between the source and free surface. Similar expressions can be written for any
‘the receiver with a lateral resolution intermediate between other pair of the reflections. Matrices F, and F, can be
those of SKS (100 km) and of the long-period surface waves written as:
(a few thousand kilometres). This method was applied by
Vinnik & Farra (1992) to measure shear-wave splitting in
the mantle of the Fiji region in the south-west Pacific, by
using the records of station AFI. Now we describe the
results for other stations and other regions of the Pacific. On the assumption of transverse isotropy with the
horizontal symmetry axis, the approximate expressions for
2 MULTIPLE ScS T E C H N I Q U E the coefficients of the matrices can be written (Vinnik et al.
1989c) as
The idea of the method is to analyse the evolution of
particle motion in the records of long-period multiple ScS, SV, = cos’ /3 + sin’ /3 exp (-just),
and sSCS, at the same station as a function of the order of s < h = SH, = 0.5 sin 2/3[1- exp ( - i o b t ) ] , (3)
reflection n. The differences in the angles of incidence of the
two successive ScS, are in the range of several degrees, and SH,, = sin’ /3 + cos’ /3 exp (-iwbt),
it may be assumed for practical purposes that the S-wave where /3 and 6t are the angle between the symmetry axis and
radiation in the two directions is the same. Considering the the back azimuth of the event and time lag of the slow wave,
pair ScS,, ScS, (Fig. 1) and assuming some simplifications respectively. From eq. (1) we obtain:
we may write for the spectra of the corresponding particle
motions in the horizontal plane the following expressions:

If the parameters of matrix F, are known, eq. (4) can be


Shear-wave splitting in the Pacific mantle 197

used to find the parameters of matrix F,. Unfortunately, this where /3 is the angle between the fast direction of anisotropy
is not the case for most of our stations. In principle, the and the principal direction of the particle motion,
parameters of both, F, and F,, can be found from the &(an) = 26n6t, 6t is time lag of the slow wave for one path
records of a group of events, but the required amount of in the mantle and 6n is the difference in the orders of
records is too large to be practical. For these reasons, we reflections in the pair. The optimum pair of parameters of
use a different technique that has much in common with the anisotropy in the bounce points corresponds to the
SKS technique (Vinnik, Farra & Romanowicz 1889a). First, minimum of the penalty function E:
we assume that the reflection coefficient K , is equal to 1,
which is correct for observations at very small epicentral ~
( [ O ( t )- O*(t, (Y, st)]’ dt
distances. Moreover, assuming that w6t << 1, the matrix
I
E(cY,6 t ) = - C f

product F,F, can be reversed, so that: N


1 M 2 ( t ) dt

[”,I
SH,
= F,[ ”I1].
SH,
where (Y is azimuth of polarization of the fast wave, O* is
the theoretical 0 component obtained via eq. (6) and N is

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


Similar expressions can be written for any other pair of the number of pairs of the reflections.
reflections. If we consider two reflections ScS,. and ScS,, To estimate the accuracy of this technique in the general
n ‘ > n , one can write case we performed numerical experiments. The initial
motion was modelled by a cycle of sinusoid with 20 s period,
and the theoretical particle motions for ScS, and ScS, were
calculated using eq. 1. Theoretical seismograms were
processed using the simplified technique described above.
where 6n = n ’ - n is the difference in the orders of Table 1 summarizes the results for the most general case
reflections in the pair and F P is given by eqs ( 2 ) and (3) when anisotropy is present in the source, receiver and
with 6t replaced by 6n6t. bounce point regions. 6t for anisotropies in all regions are
Let us now start from the simplest situation: particle taken the same (0.6s for one path in the mantle) whereas
motion in the source is linear (there is no phase shift the directions of the axes of symmetry are present in all
between SV, and SH,),anisotropy in the receiver region is possible combinations, with 45 degree increments. To take
missing and the r,eflection coefficient K , is equal to 1. These into account the effect of inequality of the reflection
assumptions imply that the particle motion in ScS, is linear coefficients for S V and SH at the core-mantle boundary, the
(see eq. 1) and the principal motion direction (PMD) in ScS, seismograms are calculated for the angle of 45 degrees
at long periods is similar to that in ScS,. This situation, in between PMD and the radial direction. The adopted
principle, is similar to that with SKS provided that the radial epicentral distance is 20 degrees where the values of K , for
(R) and transverse (T) components in the expression for ScS, and ScS, are close to 0.7 and 0.9, respectively. Table 1
SKS are replaced by the principal component ( M ) of the indicates that in the vast majority of cases the proposed
particle motion in ScS, and the orthogonal horizontal technique provides reasonably accurate estimates of the
component (0),respectively. In practice, anisotropy in the parameters (the errors are not more than 20 degrees for (Y
source and the receiver regions as well as source complexity and 0 . 2 s for s t ) . Accuracy can further be enhanced by
may contribute to the 0 component of ScS,,. A contribution applying this technique to a group of records with differing
from them should be present in both reflections in the pair. PMD. Thus, our approach allows us to distinguish the effect
To reduce the disturbing effect of these contributions we of anisotropy in the bounce point from other sources of
apply to both reflections the same filter which transforms ellipticity of the ScS, particle motion. This is hardly possible
particle motion in the first reflection into a linear one. The with the observations of ScS, alone (Ansel & Nataf 1990).
filter for linearizing particle motion in the first reflection is We note that in the case of a malfunction of the recording
synthesized on the assumption that ellipticity of the first instruments which results in a phase shift between the
reflection is caused by anisotropy. In some cases this records of the horizontal components, this shift is eliminated
procedure is very efficient. If, for instance, ellipticities of by the filtering procedure described above.
waveforms in ScS, and ScS, are similar, this means that We neglect a possibility that the lower mantle may
anisotropy at the bounce point is missing, or PMD is close contribute to splitting in ScS,, in spite of indications that
to the polarization direction of either fast or slow split wave. anisotropy can be present in the D” region (Vinnik, Farra &
Generally, however, as can be shown by simple matrix Romanowicz 1989b; Lay & Young 1991). Possible evidence
manipulations, this procedure does not remove the of anisotropy in D has been found in the observations of
disturbance completely. Unfortunately, analytical evaluation anomalous polarization of the S-wave diffracted along the
of the residue is difficult. core-mantle boundary. It is possible that the anisotropy in
After applying the filtering process, we can calculate from D is very weak, and the observable effect can be
the M component of the second reflection in the pair, the accumulated only if the wavepath in D” is a few thousand
theoretical 0 component denoted by O* kilometres, which is not the case for ScS,. Moreover, the
O*=F*M anomalous polarization was documented so far for only a
(6) few paths in D”. In any case, the effect of possible
where, in frequency domain, the filter F is anisotropy in D upon splitting in ScS, is practically similar
to that in SKS; in numerous studies of upper mantle
1- exp [-iwSt(Gn)] anisotropy with the SKS techniques, this effect so far has
f ( w ) = 0.5 sin 28
cos2B + sin2B exp [-iwSt(Sn)]’ (7) been neglected.
198 V . Farra and L. Vinnik

Table 1. Results of testing the inversion procedure. Table (Continued.)

ab az Qr 6tl a1 6t2 Q2 ab az a r St, a1 6t2 a2


deg deg deg sec deg sec deg deg deg deg sec deg sec deg
0. 0. 0. 0.6 170. 135. 0. 45. - 140. - 50.
0. 0. 45. 0.6 10. 135. 0. 90. - 140. - 50.
0. 0. 90. 0.6 0. 135. 0. 135. - 140. - 50.
0. 0. 135. 0.8 160. 135. 45. 0. 0.6 150.
0. '0.
45. 0.2 10. 135. 45. 45. 0.6 140.
0. 45.45. 0.4 10. 135. 45. 90. 0.6 130.
0. 45.90. 0.2 10. 0.2 90. 135. 45. 135. 0.6 150.
0. 45.135. 0.4 10. 0.4 90. 135. 90. 0. - 140. - 50.
0. 1.2 40. -

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


90.0. 135. 90. 45. 140. - 50.
0. 90.45. 1.0 40. 135. 90. 90. - 140. - 50.
0. 90.90. 0.6 20. 135. 90. 135. - 140. - 50.
0. 90.135. 0.8 170. 135. 135. 0. 0.6 140.
0. 135.
0. 0.4 10. 0.4 90. 135. 135. 45. 0.6 130.
0. 45.
135. 0.2 10. 0.2 90. 135. 135. 90. 0.8 160.
0. 90.
135. 0.2 0. 135. 135. 135. 0.6 130.
0. 135.
135. 0.2 10. ( Y and
~ (Y, are the values of (Y at the bounce and receiver points of

45. 0. 0. 0.2 60. the model, respectively; (Y in the source is fixed at 0 degrees; az is
back azimuth of the S wave; bf in the receiver, source and bounce
45. 0. 45. 0.4 60. points of the model are fixed at 0.6s, at, and a,are the results of
45. 0. 90. 0.2 60. the inversion for the parameters of anisotropy in the bounce point
of the model; at, and is the second pair of values of the
45. 0. 135. 0.2 60. parameters which was obtained in some experiments.
45. 45. 0. 1.0 20.
45. 45. 45. 0.6 40. The size of the region responsible for the observed effect
45. 45. 90. 0.6 50. can be estimated as the first Fresnel zone around the surface
bounce point. The traveltimes of the rays which connect the
45. 45. 135. 0.6 50.
points within this zone with the source and the receiver
45. 90. 0. 0.2 70. 0.2 130. differ from that for the bounce point by less than a
45. 90. 45. 0.4 60. half-period (Kravtsov & Orlov 1990). A smaller region
45. 90. 90. 0.4 60. where the major portion of wave energy is reflected can be
45. 90. 135. 0.2 60. 0.2 130. defined by the contour with the quarter-period phase
45. 135. 0. difference. The quarter-period zones can be approximated
0.6 40.
by circular areas around the bounce points. For the 25s
45. 135. 45. 0.6 60. period, the radius of the circle is around 6 degrees (ScS,) or
45. 135. 90. 0.6 30. 7 degrees (ScS,).
45. 135. 135. 0.6 40.
90. 0. 0. 0.6 90. 3 RESULTS OF MEASUREMENTS
90. 0. 45. 0.6 90.
In the literature, there are many examples of applications of
90. 0. 90. 0.6 110.
observations of ScS, for various purposes (e.g. Sipkin &
90. 0. 135. 0.6 110. Jordan 1980; Chan & Der 1988), but these are usually
90. 45. 0. 0.4 100. 0.4 0. analyses of the SH component of selected events with a high
90. 45. 45. 0.4 100. signal-to-noise ratio. Our study is complicated by a necessity
90. 45. 90. 0.4 100. 0.4 0. to analyse the 0 component where the signal-to-noise ratio
is by an order of magnitude lower than in the principal
90. 45. 135. 0.2 100. 0.2 0. component. An additional requirement is that the noise
90. 90. 0. 0.6 90. level in the 0 component should be low in, at least, two
90. 90. 45. 0.8 70. reflections of different order. The other, though not so
90. 90. 90. 0.6 90. restrictive requirement, is the similarity of the PMD in both
90. 90. 135. 0.6 110. reflections. By inspecting practically all available digital
90. 135. 0. 0.2 100. 0.2 0. recordings of strong intermediate and deep events at
stations in and around the Pacific we have found 45 records
90. 135. 45. 0.4 100. 0.4 0. of the required quality (Table 2). PMD was determined as
90. 135. 90. 0.2 120. 0.8 0. the eigenvector of the corresponding covariance matrix. The
90. 135. 135. 0.4 100. filter for linearizing particle motion in the first reflection was
135. 0. 0. - 140. - 50. synthesized on the assumption that ellipticity of the first
Shear-wave splitting in the Pacific mantle 199

Table 2. List of records. Az is principal directions of ScS, particle motion in the horizontal plane.
No Date Lon Lat Dist Phase Az
deg deg deg deg
A FI
1 83 01 26 -179.21 -30.52 238 17.8 scs,,ScS2 158,164
2 84 06 15 -174.87 -15.79 24 7 3.2 scs,JCS2 75,76
3 84 10 15 -173.79 -15.74 129 2.66 ScS, ,ScS2,Sc& 75,76,76
4 84 10 30 -174 11 -17.12 141 3.90 scs,,ScS2,sScS,,
sscs2 95,93,101,101
5 84 11 17 -178.09 -18.74 45 1 7.73 sscs,,sScS2 160,157
6 86 10 30 - 176.68 -21.69 194 9.04 scs,,scs2 126,113
7 91 06 09 -176.27 -20.16 278 7.56 scs,,scs2,scs3 126,116,119
sscs,,sscs2 118,126
PPT

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


1 8 7 0 2 10 182.54 -19.49 395 26.49 SCSl ,ScS2 92,95
2 87 03 19 183.87 -20.40 214 25.25 SCS,,scs2,sscs,
,sscsz 105,102,111,97
3 8 8 0 1 15 184.00 -20.79 214 25.14 scs,,scs2 120,114
4 88 03 10 181.35 -20.92 623 27.62 scs,,scs2 145,146
5 89 03 11 185.24 -17.77 230 23.99 scs,,ScS2 105,103
CTAO
1 77 07 06 181.42 -21.0 597 32.9 scs,,scs3 142,138
2 80 04 13 183.30 -23.74 166 34.5 scs2.scs3 159,150
3 81 09 28 181.35 -29.27 309 33.1 scs2,scs3 169,175
4 81 10 07 181.00 -20.52 625 32.6 scs2,scs3 160,157
5 84 08 26 179.07 -23.59 560 30.6 scs2JCS3 165,154
6 84 11 17 181.97 -18.79 45 1 33.7 scs2,scs3,scs, 3,173,174
7 85 08 28 181.02 -21.01 625 32.5 scs2.scs3,scs, 155,155,134
SNZO
1 81 04 28 180.07 -23.62 553 17.63 SCS,,scs2 78,76
2 82 09 17 179.83 -23.15 562 18.04 SCS,,ScS2 84,80
3 84 08 26 179.07 -23.59 560 17.46 scs,JCS2 138.140

GUM0
group 1
1 84 03 06 139.11 29.60 90,89
2 88 09 07 137.43 30.24 66,78
group 2
3 84 02 01 146.31 49.10 131,143
4 90 05 12 141.88 49.04 165.174
MAJO
group 1
1 79 08 16 131.74 4 1.97 95,100,100,105
2 81 11 27 131.07 42.913 75,71
2
81 01 23 142.07 42.2 121 6.39 SCS,,SCS~ 141,142
84 04 20 148.77 50.19 593 SCS,
15 62,SCS~
,SCS~ 123,131,131
86 07 19 151.32 47.05 153 14.32 SCS~,SCS~ 117,130
87 05 18 147.693 49.28 542 14.49 SCSI,SCS~,SSCS,
,sSCS~126,133,123,126
90 05 12 141.88 49.04 611 12.78 SCS,,SCS~ 154,163
3
82 07 04 136.48 27.92 552 8.73 SCS,,SCS~ 169,179
85 04 03 139.6 28.4 455 8.21 SCS~,SCS~ 69,74
88 09 07 137.431 30.245 485 6.32 SCS~,SCS~ 44,49
91 05 03 139.570 28.061 459 8.55 scs,,scs2 79,82
group 4
12 78 05 23 130.45 31.0 176 8.49 ScSi,sCsz 62,63
13 81 01 02 128.38 29.03 217 11.15 SCS,,SCSZ 119.128
14 82 07 05 130.47 30.77 119 8.64 SCSI,SCSZ 127,138
15 87 07 03 130.322 31.196 168 8.45 SCS,,SCSZ 98,88
200 V. Farra and L. Vinnik
Table 2. (Continued.)
N~ Date Lon Lat Dep Dist Phase AZ

deg deg kin drg d%


INU
10 88 09 07 137.431 30.245 485 6.32 SCS~,SCS~ 44,49
KIP
1 9 0 0 5 12 141.88 49.04 611 5458 SCSZ,SC& 37,33
ZOBO
1 77 02 04 296.61 -22.66 5.55 8.1 ScSl,ScSz,ScS3 68,61,71
2 87 10 27 297.07 -28.676 605 13 7 SCS~,SCS~,SCS~ 90,90,87

reflection is caused by anisotropy. The record of each records of multiple ScS. We have no idea about the origins
reflection was individually filtered to attain the highest of these phases, but explain them tentatively by scattering at
possible signal-to-noise ratio in the 0 component. Only the lateral heterogeneities. The fast direction obtained for the

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


initial part of the signal is used because the later part can be whole group of the records of PPT (100 degrees) is well
contaminated by unwanted waves whose arrivals are constrained and is similar to that found from the data of
sometimes well seen in the 0 component. In the rest of this AFI. The estimate of 6t (around 0.6 s) is similar, as well.
section the data are described in detail. Positions of the At CTAO, there are seven records of events in the
sources, receivers, quarter-period zones and the resulting Fiji-Tonga zone with epicentral distances 30-35 degrees. At
fast directions of anisotropy are summarized in Fig. 2. The these distances we can not use the records of ScS, , and most
results of measurements are also summarized in Table 3. of the estimates are obtained from the records of ScS, and
ScS,. The estimates of the parameters obtained for the
3.1 Southwestern Pacific different pairs are in a good agreement with each other. The
The data for the southwestern part of the Pacific are final estimates (170 degrees for (Y and 1.0 s for 6t) are well
obtained at seismograph stations AFI, PPT, CTAO and constrained.
SNZO; the sources are intermediate and deep events in the At station SNZO there are records of three ScS,, ScS,
Fiji-Tonga subduction zone. The regions sampled by the pairs suitable for the analysis of splitting, all of them from
data are mainly a part of the Pacific basin with old (around deep events at epicentral distances near 18 degrees. The
100 Ma) lithosphere and marginal basins between the records and the corresponding plots of E are shown in Figs 5
Fiji-Tonga island arc and Australia. and 6. In the record of event 1 with PMD near 77 degrees
The data of GDSN station AFI are described by Vinnik & the linear m o t h in ScS, changes to elliptic in Sc&; the
Farra (1992) and, for this reason, are only mentioned here. inferred direction of fast velocity is 20 degrees. In the record
A high signal-to-noise ratio was found in the seismograms of of event 2 with PMD around 82 degrees, both particle
seven events. Sometimes we could use three successive motions are elliptical. After filtering which corrects for
phases (ScS,, ScS,, ScS,) and depth phases sSCS,and sSCS,. ellipticity of ScS, , particle motion in ScS, becomes linear, as
The solutions for different records are compatible with each well. The corresponding fast direction is in the range from 0
other, and the eryors of the final estimates of the fast to 80 degrees. ScS, of event 3 also needs correction for
ellipticity of ScS,. The solution for the fast direction in this
direction (Y are unlikely to exceed 10 degrees. They are
practically similar (100-110 degrees) for the Pacific basin pair is 40 degrees, close to the value inferred from record of
and the Lau basin. The estimates of 6t are around 0.5 s for event 1. The penalty function E for the whole group has
one path in the mantle of the Pacific basin and around 1 s clear minimum at a near 45 degrees; the preferred value of
for the Lau basin. 6t is around 1s.
At Geoscope station PPT we could use the records of five
events, all of them in the region of Fiji. Examples of the
3.2 Northwestern Pacific
records and the contour plots of the penalty function E are
shown in Figs 3 and 4, respectively. In the records of event 1 The data for anisotropy in the northwestern Pacific are
with the PMD near 94 degrees and event 3 with the PMD obtained at stations GUMO, MAJO, INU and KIP. The
near 117 degrees the linear motion in ScS, evolves into epicentres of events recorded at GUMO form two separate
elliptic in ScS,. Particle motion directions in ScS, are groups and the results correspond to two different regions.
clockwise (event 3) and counterclockwise (event 1). This The records of the first group (events 1 and 2) and the
relationship suggests that the polarization of the fast wave is corresponding penalty functions are shown in Figs 7 and 8.
between these two PMD (near 105 degrees). This conclusion In the record of event 1 the particle motions of both, sSCS,
is compatible with the results for the other records. In the and sScS, are elliptical, but sSCS,corrected for ellipticity of
record of event 5 with the PMD near 104 degrees, the linear sSCS, is linear. The resulting fast direction is either 0 or 90
motion is observed in both, ScS, and ScS,. This is possible degrees. In the record of event 2 the linear motion in ScS,
when the PMD is close or perpendicular to the direction of transforms into elliptical in ScS, with the fast direction near
the axis of symmetry. A similar relation between ScS, and 30 degrees. The difference between the solutions for the two
ScS, is found in the record of event 2 with a similar PMD. A events (30 degrees) is in the range of the possible errors of
notable feature of the record of ScS, of event 5 is an arrival measurement. The combined solution for the two events is
in the 0 component with a delay around 1 min with respect close to 0 degrees. The solutions for the events of the
to ScS,. Similar arrivals can be recognized in some other second group (events 3 and 4) are also compatible with each
Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022
202 V . Farra and L. Vinnik

55.

50.

45.

40.

35.

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


30.

25.
120. 126. 132. 138. 143. 149. 155.

60.

50.

40.

30.

20.

lo* I
0. --
140.
1--
- r - - - - w
152. 163. 175. 167. 198. 210.
..
275. 280. 285. $90. 295. 300. 305.

Figure 2. (Continued.)

and the final solution is well constrained. The optimum of a! is compatible with those for the other records. The
values of the parameters are 160 degrees for a! and 1.4 s for estimate of a for the group (170 degrees) is reasonably well
at. constrained while the estimate of 6t is determined
The records of the second group consisting of events 3-7 practically by one record and should be regarded as less
characterize a region around Hokkaido. The positions of the reliable. In the records of the fourth group (events 12-15),
minimum values of E for individual pairs of this group vary whose quarter-period zone corresponds to the region around
in a broad range but the regions of low values overlap. The southern Japan, the PMD vary between 60 and 140 degrees.
preferred value of a! for this group is near 120 degrees; the The effect of anisotropy in all records of this group is
estimate of 6t is uncertain. The data of the third group negligible.
(events 8-11) characterize a region to the south of Japan. In Figs 11 and 12 we present data for the Sakhalin event
The results for event 10 are different at stations MAJO and recorded at KIP. The linear motion in ScS, is transformed
INU. We use the record of INU because the resulting value into elliptical in ScS,. The resulting estimates of a! and 6t
Shear-wave splitting in the Pacific mantle 203
Table 3. Summary splitting scarcely mimic systematic effects of azimuthal anisotropy in
parameters. a group of records with differing PMD. In Section 3 we
discussed extensively the compatibility of the results within
Station a 61 Event each group of records. In some cases the compatibility is
almost perfect implying that the unwanted effects are
d% sec no
negligible. In other cases such effects are present, but they
AFI 100 1.1 1 are probably suppressed by combined processing of several
records.
AFI 107 0.5 1-7 An additional way to judge the accuracy of our
PPT 100 0.6 1-5 measurements is provided by the observations of splitting in
SKS at the stations within the regions sampled by ScS, or in
CTAO 170 1.o 1-7 the neighbourhood of these regions. The relevant data on
SNZO 45 1.o 1-3 splitting in SKS are summarized by Vinnik et al. (1992). The
value of a at SNZO (20 degrees) is close to that for the
GUM0 0 2. 12

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


neighbouring South Fiji basin (Fig. 2d): the difference is less
GUM0 150 1. than 30 degrees. Similarly, the values of a for INU and
3,4
MAJO (170 and 130 degrees, respectively) are close to those
MAJO 160 1.4 132 for the neighbouring regions in Figs 2(e) and 2(f), and the
value of a at ZOBO (120 degrees) is close to that for the
MAJO 120 1.4 3-7
neighbouring region in Fig. 2(h). These are the only
MAJO 170 2. 8-11 available data that can be compared; the difference between
the values of a inferred from SKS and ScS, is always within
MAJO 90 0.3 12-15
30 degrees.
KIP 80 0.9 1 Interpretation of splitting in the seismic phases with a
steep incidence, like ScS, or SKS, is usually complicated by
ZOBO 90 1.5 192 their low depth resolution. The assumption about a
negligible effect of crustal anisotropy, which is common in
are 8 0 f 3 0 degrees and around l s , respectively. These the continental studies based on the SKS technique, is even
estimates correspond to a large part of the northwestern more appropriate for the ocean crust, owing to its reduced
Pacific. thickness. Discrimination between fossil anisotropy in the
subcrustal lithosphere and recently formed anisotropy in the
3.3 Andes asthenosphere is, generally, easier for the oceans than for
the continents, owing to the younger age, smaller thickness
Anisotropy in a central part of Andes is inferred from the and better understood history and structure of the oceanic
records of two events at station ZOBO. The data are lithosphere.
displayed in Figs 13 and 14. In the ScS,, ScS, pair of the first Various relations between mantle anisotropy and plate
event both particle motions are elliptic but their ellipticities motions are already established for some of the Earth’s
are strongly different. A good signal-to-noise ratio is also regions (e.g. Vinnik et al. 1992), and in the rest of this
observed in ScS,, but the result for the pair Sch, ScS, section we try to do this for the regions of our study. The
deviates from that for the first pair. We attribute the regions sampled by our data fell into three distinctly
difference to a larger area sampled by ScS, and do not use different groups: (1) the Pacific basin with the old (around
the second pair. In the record of event 2, particle motions in 100 Ma) lithosphere, (2) back-arc basins of various age, and
ScS,, Sch and ScS, are linear. This result is compatible with (3) the Andes.
the solution for event 1. The result of combined processing
is 90 degrees for a and 1.5 s for dt.
4.1 Pacific basin
4 DISCUSSION A N D CONCLUSIONS
The data for the Pacific basin are obtained from the records
Prior to discussing the results, one should consider a of AFI, PFT, GUMO and KIP (Fig. 2). Unfortunately, the
possible trade-off between the effects of anisotropy and data of GUMO (second group) sample not only the Pacific
lateral heterogeneity in our data. The ray paths of the basin but also the neighbouring back-arc basins, and the
multiple ScS in the source and the receiver regions are obtained estimates of anisotropy present a combined effect
similar, and the disturbing effects of the corresponding of a few different tectonic units. The data of KIP are less
heterogeneities are minimized by our technique of accurate than the others and deserve separate discussion.
differential measurements. The other heterogeneities may The data of AFI and PPT correspond to the old oceanic
act as random scatterers, and scattered waves may lithosphere in the south-west Pacific with the age of 100 Ma
contaminate the records. In the 0 component we often (Sclater, Parsons & Jaupart 1981). Our estimates of the
observed disturbances arriving approximately 1 min later parameters of anisotropy for this region are reliable and
than multiple ScS. We tentatively explain them as scattered remarkably consistent in the two data sets. In both cases
waves and analyse only the initial parts of the ScS records. azimuth of polarization of the fast wave (a)is around
Of course, scattered energy might remain unnoticed and 100-105 degrees and time lag of the slow wave ( d t ) is close
contaminate the early parts of the records, but it could to 0.5 s. In both cases the present-day absolute plate motion
(a) P (b) P
5 0- 50-
0 s o ?
0
U
0
-50- -50-

2 i- - 25 -40

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


1. 13.40.00 11 . 11 1. 2 8 . 30.00 1 . 1
14 15 16 17 29 31

SCSl PPT Event 1 sc s 2

(' r
PPT Event 1

. . _ M
- - - I

10-

-10-

-50
8. 55.0.00 9. 10. 0 . 0 0 I I .
lb 11 I 13

PPT Eveht 3 SCSl PPT Event 3 sc s2


(el 0
I
I

10-
0 :
- -10-

80-
M ' P ;
0
U

-80-

5. 19- 0.00 5. 34.40.00


1 1

PPT Event 5 SCSl PPT Event 5 sc s2


Figure 3. Examples of records at PFT. Each record is represented by the M and 0 components of motion and particle motion plot. Time
marks are minutes; time interval used for the analysis is shown by ticks at the time axis. The beginning and end of the particle motion plot are
marked by circle and triangle, respectively. Event numbers correspond to the list of events in Table 2. (a) ScS, of event 1, (b) ScS, of event 1,
(c) ScS, of event 3, (d) ScS, of event 3, (e) ScS, of event 5 , ( f ) ScS, of event 5.
Shear-wave splitting in the Pacific mantle 205

9 (a) ? lhl
cu ru

> >
a a
-J 1
W w
0 0

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


0
0

0 .o FINCL E 180.0 0 .o FlNGLE 180 - 0

0 lC\

*
a
-J
w w
0 0

0 0
0 0

0 .o 180.0 0 .o
RNGLE RNGLE 180.0

Figure 4. Plots of E ( a , st) for PPT. (a) ScS,, ScS, of event 1, (b) ScS,, ScS, of event 2, (c) sSCS,, sSCS, of event 2, (d) ScS,, ScS, of event 3,
(e) ScS,, ScS, of event 4, ( f ) ScS,, ScS, of event 5, (8) all pairs.

direction (around 290 degrees after Minster and Jordan of the fast direction of anisotropy with the present-day
1978) is very close to the value of a (105 or 285 degrees). absolute plate motion direction was previously found for
The fast direction in the subcrustal lithosphere which is some old continental plates (Vinnik et al. 1992) and
presumably aligned with the fossil spreading direction is explained by resistive drag at the base of the plates. Our
around 330 degrees in the region sampled by the data of observations in the Pacific can be explained by a similar
AFI and around 260 degrees in the region sampled by the process. The time lag of 0.5 s in the mantle underneath the
data of PPT (for the fossil spreading directions see e.g. southwestern part of the Pacific plate is much lower than the
Nishimura & Forsyth 1989). The influence of the large values representative of the continental mantle. The
changes in the direction of fossil anisotropy on splitting in difference can be explained by a lower viscosity of the
ScS, is, clearly, small, and our data characterize mainly the oceanic upper mantle. The weakness of asthenospheric
present-day flow in the asthenosphere. A similar alignment anisotropy underneath the oldest (older than 80 Ma) parts of
206 V. Farra and L. Vinnik

cu

>
a
*
a
_J J
w w
0 0

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


0 0
0 0

0 .o 180.0 0 .o 180.0
RNGLE RNGLE

-
0
la)
1.7,

0 .O 180 .o
RNGLE
Figure 4. (Continued.)

the Pacific plate is noted also in the surface-wave data degrees, Vinnik et al. 1989a; 75 degrees, Kuo & Forsyth
(Nishimura & Forsyth 1988, 1989). 1992). Under plausible assumptions about magnitude of
The data of KIP suggest a value around 80 (260) degrees frozen anisotropy, it might account for a part of the effect in
for (Y in the northwestern Pacific. This value deviates from ScS. We conclude that anisotropy inferred from the data of
the present-day absolute plate motion direction (around 300 KIP most likely presents a combined effect of frozen
degrees) but the difference of 40 degrees is comparable with anisotropy in the subcrustal lithosphere and of recent flow in
the possible error of this measurement. According to the asthenosphere. More definitive judgment is hampered
Nishimura & Forsyth (1988) the direction of frozen by the relatively low accuracy of this measurement.
anisotropy in the subcrustal lithosphere of that region is
around 45 degrees, whereas asthenospheric anisotropy is too
4.2 Back-arc basins
weak to be observed. The direction of anisotropy at 200 km
depth inferred from the surface-wave data by Montagner & The data for the back-arc basins can be divided into two
Tanimoto (1991) is consistent with ours (Fig. 15). The fast groups. In several instances (Lau, South Fiji, Parece Vela
direction different from the absolute plate motion direction basins and Japan Sea) the size of the basin is comparable
is also found underneath KIP, using SKS techniques (45 with the size of the quarter-period zone, and the parameters
Shear-wave splitting in the Pac$c mantle 207

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


2 1 . 2 7 . 0.00 21. 43. 0.00 .
43 44
I
I
I
4p 46

SNZO Event 1 sc SNZO Event 1 scs2

5 0-

0 2
I

m
-50-

-10

41 . 42 43 I 44 57 Sp 1 59 60

1 3 . 5 6 . 0.00 5. 1 3 . 0.00
13
. I4
1
I
.
15
1
I 16

SNZO Event 2 sc s2 SNZO Event 3 SC s 1

Figure 5. The same as in Fig. 3 for SNZO. (a) ScS, of event 1 , (b) ScS, of event 1 , (c) ScS, of event 2, (d) ScS, of event 2 (e) ScS, of event 2
corrected for ellipticity of ScS,, (f) ScS, of event 3, (9) ScS, of event 3, (h) ScS, of event 3 corrected for ellipticity of ScS,.
208 V . Farra and L . Vinnik

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


5. 2 9 . 0.00 I I 5. 2 9 . 0.00 . I I .
29 3p I 31 32 29 3p I 31 32

SNZO Event 3 s C s2 SNZO Event 3 sc s2


Figure 5. (Continued.)

of anisotropy can be attributed to the upper mantle The Parece Vela basin within the Philippine Sea plate is
underneath the basin. This group of data is the most characterized by magnetic anomalies oriented roughly
informative, and it is discussed first. The data of the second north-south (Jolivet, Huchon & Rangin 1989). Spreading
group are more ambiguous because the corresponding took place from 30 to 17 Ma. To the east of the Parece Vela
quarter-period zones sample a few tectonic units. basin, within the same quarter-period zone, there is a
The fast direction for the Lau basin is around 100 degrees, narrow zone (Manana trough) where spreading in the
almost perpendicular to the plate boundary, as marked by east-west direction commenced 8 Ma ago (Jolivet et al.
the Tonga islands. Our estimate is broadly consistent with 1989). The size of the rough is small in comparison with the
the fast direction beneath station LAK on the Lau ridge size of the quarter-period zone. Our estimate of the fast
reported by Bowman & Ando (1987). Magnetic anomalies direction (0 degrees) and the corresponding direction of
indicate that the Lau basin is formed by spreading in the extension in the mantle are practically perpendicular to the
east-west direction during the last 3 Ma (Malahoff, Feden & direction of crustal spreading in the Parece Vela basin and
Fleming 1982). The fast direction of anisotropy in the the Mariana trough.
mantle is close to the direction of spreading indicated by The Sea of Japan opened as a distributed right-lateral
magnetic anomalies, implying that the direction of spreading shear zone several hundred km wide with the direction of
in the crust coincides with the principal direction of extension near 140 degrees (e.g. Jolivet et al. 1991; Tamaki
extension or shear flow in the mantle. The data suggest that et al. 1992). Opening of the sea commenced 32Ma and
anisotropy in the Lau basin is stronger than in the continued until 10 Ma. Later, NW-SE extension was placed
neighbouring region of the Pacific. by an east-west compression which has been continued until
The South Fiji basin is a marginal basin several hundred the present day. P-wave azimuthal anisotropy at the top of
kilometres wide produced by spreading which was active in the mantle in the Yamato basin (eastern part of the Japan
the period between 36 and 25Ma (Malahoff et al. 1982). Sea) is one of the largest in the world (9.1 per cent) with the
Magnetic anomalies are oriented predominantly parallel to fast direction near 140 degrees (Chung 1992). The direction
the Kermadek trench which indicates spreading in the of extension in the mantle of the Japan Sea inferred from
direction perpendicular to the trench. The quarter-period our splitting measurements is 160 degrees between the
zone for events recorded at SNZO also samples the Havre direction of extension imprinted in the crust (140 degrees)
trough, a 3 Ma old zone of back-arc spreading to the east of and that corresponding to the axis of extension of the
the South Fiji basin. This zone is very narrow in comparison present-day deformation (180 degrees). Taking into account
with the diameter of the quarter-period zone and, for this the large magnitude of the observed P-wave anisotropy and
reason, probably is unable to contribute much to the effect assuming the largest possible difference between velocities
in ScS,,. The direction of spreading in this zone, like in the of the split waves in the ultramafic rocks (around 4.5 per
South Fiji basin, is perpendicular to the Kermadek trench. cent, Christensen 1984), the thickness of the anisotropic
It is clear that the direction of extension in the mantle layer can be estimated as 150 km.
inferred from the data at SNZO is roughly perpendicular to The data for the above-mentioned basins can be
the directions of extension in the crust of the South Fiji generalized as follows. In the back-arc basins with the latest
basin and the Havre trough. This direction of anisotropy is episodes of extension not older than circa lOMa, the
consistent with compression perpendicular to the plate direction of extension in the mantle is close to that
boundary. A similar direction of anisotropy is found imprinted in the crust (Lau basin, Sea of Japan). In the
underneath SNZO, using the SKS technique (Vinnik et al. back-arc extensional basins whose active development
1992). ended more than 15Ma ago, the direction of extension in
Shear-wave splitting in the Pacific mantle 209

N N

>-
a
J
W W
0 0

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


0 0

0 0

0 .o 180 - 0 0 .o RNGLE 180 .O


RNGLE

> >-
a a
_I
-J
W W
0 0

0 .o 180.0 0 .o
RNGLE RNGLE
Figure 6. The same as in Fig. 4 for SNZO. (a) ScS,, ScS, of event 1 , (b) ScS,, ScS, of event 2, (c) ScS,, ScSz of event 3, (d) all pairs

the mantle inferred from the seismic data is parallel to the anisotropy for this group is 130 degrees practically similar to
strike of the island arc rather than to the palaeodirection of the opening direction of the Kuril basin. The quarter-period
spreading. Apparently, the subcrustal lithosphere, where zone of the third group of events recorded at MAJO
fossil anisotropy can be preserved, in the back-arc basins is includes a part of the Shikoku basin, whose opening in the
too thin to affect our measurements, and the deformations east-west direction took place from 25 to 12 Ma ago (Jolivet
inferred from the seismic data are those recently formed in et al. 1989). The fast direction of anisotropy (170 degrees) is
the asthenosphere. They are usually not much older then strongly different from the direction of spreading, though
several Ma. the basin is not much older than the Sea of Japan. It is
The data of the second group are more ambiguous, but, possible that the part of the basin within the quarter-period
generally, they do not contradict the data of the first group. zone is too small to affect the results of measurements. The
The quarter-period zone representative of the second group region of the quarter-period zone of the fourth group at
of events recorded at MAJO includes the region of the Kuril MAJO is too heterogeneous to discuss the reasons for the
basin whose opening was contemporaneous with the weakness of anisotropy. The data obtained at CTAO
opening of the Sea of Japan and probably was related to the correspond mainly to the southern part of the Coral sea,
same process (Jolivet el al. 1989). The fast direction of New Caledonia and Fiji plateau. The plateau is a 3-5Ma
210 V . Farra and L. Vinnik

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


V
2. 4 9 . 10.00 I 1 3. 4. 40.00 I
50 I 151 52 5 1 6 1 7 8
GUMO Event 1 sscs2 GUMO Event 1 s c s3

3. 4. 40.00 1 . 1 12. 6. 10.00 1


5 ’ 1 6 1 7 8 7 1 8 9

GUMOEvent 1 SsCS3 GUMO Event 2 SCSl


(4 0

1 o-’

0 -
s
n

-10-
25-

M s
n
-25-

12. 22. 0.00


22 28 I 24 25

GUMO Event 2 scsz


Figure 7. The same as in Fig. 3 for the first group of events at GUMO. (a) sScS, of event 1, (b) sSCS, of event 1, (c) sScS, of event 1 corrected
from ellipticity of sSCS,, (d) ScS, of event 2, (e) ScS, of event 2.
Shear-wave splitting in the Pacific mantle 211

0 (b)

> 2-
a a
J -1
W w
0 0

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


0 0

0 0 1
I

o .o 180.0 0 .o RNGLE 180.0


FlNGLE
0
cu

>-
a
-1
w
0

0
0

Figure 8. The same as in Fig. 4 for the first group of events at GUMO. (a) sScS,, sSCS, of event 1, (b) ScS,, ScS, of event 2, (c) both pairs.

old basin with evidence of east-west spreading. Apparently, zone are Altiplano and Puna, a plateau 4000 m high. Within
it is too small in comparison with the whole region sampled the quarter-period zone, the Nasca plate is subducted under
by the data to affect the results of measurements. In the rest the continental plate at a low angle. For a description of
of this region active deformations, like the overthrusting tectonics and dynamics of the region see, e.g. Isaacs (1988).
phase of the New Caledonia ultramafic rocks, took place Previous studies indicate that in most cases the fast
roughly in the same period of time as spreading in the South polarization direction in the presently active collisional belt
Fiji basin (Malahoff et al. 1982). Anisotropy with the fast aligns with the plate boundary (Vinnik et af. 1992). The fast
direction of 170 degrees is probably related to the direction of anisotropy in the study region is perpendicular
present-day flow in the mantle rather than to these relatively to the plate boundary, as in the young back-arc basins. A
old deformations. similar relationship was found in some other zones of
convergence (see, e.g. Vinnik, Makeyeva & Roecker 1992),
4.3 Andes but very seldom. In our case this direction can be explained
as a combined effect of shear strain and small-scale
Anisotropy in the upper mantle of the presently active convection in the mantle wedge above the subducted plate,
collisional belt of the Andes is inferred from the data of frozen anisotropy in the subducted lithosphere, and either
ZOBO. The main tectonic units within the quarter-period active flow or resistive drag below the plate. The cumulative
212 V . Farra and L. Vinnik

10-
M
-Z

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


ui
- 10-

21. 44. 0.00 2 1 . 5 9 . 0.00


44 4p I 46 41 59 60 I 61 I 62

MAJO Event 1 sc MAJO Event 1 sc s2

25-
0 :-
ui
-25-

-40

22. 3. 0.00
3 4 ( 6 1 6

MAJO Event 1 sscs2

4 00-

0
-400-
50-

M ; I-
- 50-

17. 3 4 . 0.00 I 1 17. 5 0 . 0.00 . I I


34 35 I 3k 31 50 5p I 52 53

MAJO Event 2 SCSl MAJO Event 2 SC S2

Figure 9. The same as in Fig. 3 for the first group of events at MAJO. (a) ScS, of event 1 , (b) ScS, of event 1 , ( c ) sSCS, of event 1 , (d) sSCS, of
event 1 , (e) ScS, of event 2, (f) ScS2 of event 2.
Shear-wave splitting in the PaciJic mantle 213

0 0
@J @J

>- t
a c
-1 1

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


W W
0 0

0 0
0 0

0 .o RNGLE 180.0 0 .o RNGLE 180 .o

-
0
N
fc)
N

2- >-
a a
-1 --I
W W
0 0

0 0
0 0

0 .o 180.0 0 .o QNGLE 180.0


QNGLE

Figure 10. The same as in Fig. 4 for the first group of events at MAJO. (a) ScS,, ScS, of event 1 , (b) sScS,, sSCS, of event 1, (c) ScS, , ScS, of
event 2, (d) all pairs.
214 V . Farra and L. Vinnik

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


5. 1 9 . 30.00 1 I I . I
20 2p I 22 5. 3 4 . 30.00 35 I36 37

KIP Event 1 sc s2 KIP Event 1 s cs3

Figure 11. The same as in Fig. 3 for KIP. (a) ScS, of event 1, (b) ScS, of event 1.

0
N

>-
a
w
n

0
0

0 .o 180 .o
RNGLE

Figure U.Plot of E ( a , s t ) for ScS,, ScS, of event 1 at KIP.


Shear-wave splitting in the Pacific mantle 215

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


7. ‘59. 0.00 1 I I 8. 1 5 . 0 . 0 0 I I .
59 60 I 6p 62 15 IF I 17 18

ZOBO Event 1 SCSl ZOBO Event 1 sc s 2


(a 0 (40
I I

8. 15. 0.00
I5 .’
I
IF
II 17
. 16
2 2 . 10. 0.00
Ib II
1I 12
. 1I 13

ZOBO Event 1 S CS 2 ZOBO Event 2 SCSl


(e) 0

L, b-----r?- -
t -
M
-250
216 V. Farra and L. Vinnik
9 (4 0

N c\I

>- >
a a
3 J
W w
0 0

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


0 0
- 0.26
0
I i
.o FlNGLE 180.0 0 .o 180 .o
0 FlNGLE
0
cu

>-
a
_J

w
0

0
0
0.16 > 1
0 .o RNGLE 180.0

Figure 14. The same as in Fig. 4 for ZOBO. (a) ScS,, ScS, of event 1, (b) ScS,, ScS, of event 2, (c) both pairs.

0
9 120.00
IT UDE
LONG180.00 240.00 * 300.00
m+-' 7

\ - .\

l i I I 1

Figure 15. Directions of polarization of the fast split wave (arrows) superimposed on the map of S-wave azimuthal anisotropy at 200 km depth
derived from the surface wave data (modified from Montagner & Tanimoto 1991).
Shear-wave splitting in the Pacific mantle 217

effect of these anisotropies is large (around 1.5 s), probably Western Pacific marginal basins, Tectonophysics, 160, 23-47.
owing t o their similar orientations. Jolivet, L., Huchon, P., Brunm, J.P., Le Pichon, X.,
All o u r estimates of polarization of the fast split wave a r e Chamrot-Rooke, N. & Thomas, J.C., 1991. Arc deformation
summarized in Fig. 15 where they are superimposed on the and marginal basin opening: Japan Sea as a case study, 1.
geophys. Res., %, 4367-4384.
fast directions at 200 km depth inferred from the phase
Kravtsov, Yu. A. & Orlov, Yu. I., 1990. Geometrical Optics of
velocities of the long-period surface waves (Montagner & Inhomogeneous Media, Springer Verlag, Berlin.
Tanimoto 1991). Between the two kinds of data there are Kuo, B.-Y. & Forsyth, D.W., 1992. A search for split SKS
discrepancies, especially pronounced in the south-west of waveforms in the North Atlantic, Geophys. J. Int., 108,
the study region. Among the possible reasons for the 557-574.
discrepancies, the difference between t h e corresponding Lay, T. & Young, C.J., 1991. Analysis of seismic SV waves in the
lateral resolutions is probably the most important. Further core’s penumbra, Geophys. Res. Lett., 18, 1373-1376.
improvements in the data and techniques are necessary t o Makeyeva, L.I., Vinnik, L.P. & Roecker, S.W., 1992. Shear-wave
resolve more details of t h e mantle flow within and around splitting and small-scale convection in the continental upper
the Pacific. mantle, Nature, 358, 144-146.
Malahoff, A , , Feden, R.H. & Fleming, H.S., 1982. Magnetic

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022


anomalies and tectonic fabrics of marginal basins North of New
ACKNOWLEDGMENT Zealand, J. geophys. Res., 87, 4109-4125.
McKenzie, D., 1979. Finite deformation during fluid flow, Geophys.
The study was carried out during the visit of one of the J . R. as@. SOC., 58,689-715.
authors (L.V.) to Institut d e Physique d u Globe d e Paris Minster, J.B. & Jordan, T.H., 1978. Present-day plate motions, J.
and he thanks R. Madariaga and his staff for their geophys. Res., 83, 5331-5354.
hospitality. T h e authors appreciate help from J.-P. Mitchell, B.J. & Yu, G., 1980. Surface wave dispersion,
regionalized velocity models and anisotropy of the Pacific crust
Montagner and A. Deschamps. Many records used in the
and upper mantle, Geophys. J. R. astr. Soc., 63, 497-514.
study were obtained from NEIC and IRIS data centres, and Montagner, J.P. & Tanimoto, T., 1990. Global anisotropy in the
the authors thank M. Zirbes and T. Ahern for their upper mantle inferred from the regionalization of the phase
cooperation. velocities, J . geophys. Res., 95, 4797-4819.
Montagner, J.P. & Tanimoto, T., 1991. Global upper mantle
tomography of seismic velocities and anisotropies, J. geophys.
REFERENCES Res., %, 20 337-20 351.
Nishimura, C. & Forsyth, D.W., 1988. Rayleigh wave phase
Ando, M., 1984. ScS polarization anisotropy around the Pacific velocities in the Pacific with implications for azimuthal
ocean, 1. Phys. Earth, 32, 179-195. anisotropy and lateral heterogeneities, Geophys. J. R. astr.
Ansel, V. & Nataf, H.C., 1990. On the use of Sc& phases for SOC.,94,479-501.
studying mantle anisotropy, Ann. Geophys., Special Issue Nishimura, C. & Forsyth, D.W., 1989. The anisotropic structure of
(abstract), presented at the EGS 25th General Assembly in the upper mantle in the Pacific, Geophys. J. lnt., %, 203-229.
Copenhagen, 1990 April 23-27. Ribe, N.M., 1989. Seismic anisotropy and mantle flow, J. geophys.
Babuska, V. & Cara, M., 1991. Seismic Anisotropy in the Earth, Res., 94, 4213-4223.
Kluwer, Dordrecht. Ribe, N.M., 1992. On the relation between seismic anisotropy and
Bibee, L. & Shor, G.. Jr, 1976. Compressional wave anisotropy in finite strain, J . geophys. Res., 97, 8737-8747.
the crust and upper mantle, Geophys. Res. Lett., 3, 639-642. Sclater, J.G., Parsons, B. & Jaupart, C., 1981. Ocean and
Bowman, J.R. & Ando, M., 1987. Shear-wave splitting in the continents: similarities and differences in the mechanisms of
upper-mantle wedge above the Tonga subduction zone, heat loss, J. geophys. Res., 86, 11535-11 552.
Geophys. J . R. astr. SOC., 88, 25-41. Shearer, P. & Orcutt, J., 1985. Anisotropy in the oceanic
Cara, M. & Leveque, J.J., 1988. Anisotropy in the asthenosphere: lithosphere-theory and observations from the Ngendei seismic
the higher mode data of the Pacific revisited, Geophys. Res. refraction experiment in the south-west Pacific, Geophys. J. R.
Lett., 11, 633-636. mtr. SOC., 80, 493-526.
Chan, W.W. & Der, Z.A., 1988. Attenuation of multiple ScS in Shearer, P. & Orcutt, J.A., 1986. Compressional and shear wave
various parts of the world, Geophys. J. Int., 92, 303-314. anisotropy in the oceanic lithosphere-the Ngendei seismic
Christensen, N.I., 1984. The magnitude, symmetry and origin of refraction experiment, Geophys. J. R. astr. SOC., 87, 967-1003.
upper mantle anisotropy based on fabric analyses of ultramafic Shimamura, H., 1984. Anisotropy in the oceanic lithosphere of the
tectonites, Geophys. J. R. astr. Soc., 76, 89-111. northwestern Pacific basin, Geophys. J . R. astr. SOC., 76,
Chung, T.W., 1992. A quantitative study of seismic anisotropy in 253-260.
the Yamato Basin, the southeastern Japan sea, from refraction Silver, P.G. & Chan, W.W., 1991. Shear wave splitting and
data collected by an ocean bottom seismographic array, subcontinental mantle deformation, J. geophys. Res., 96, 16
Geophys. J. Int., 109, 620-638. 429-16 454.
Forsyth, D.W., 1975. The early structural evolution and anisotropy Sipkin, S.A. & Jordan, T.H., 1980. Regional variations of Qscs,
of the oceanic upper mantle, Geophys. J . R. astr. SOC.,43, BUN. seism. SOC. Am., 70, 1071-1102.
103-162. Tamaki, K., Suyehiro, K., Allan, J., Ingle, J.C. & Pisciotto, K.A.,
Fukao, Y., 1984. ScS evidence for anisotropy in the Earth’s mantle, 1992. Tectonic synthesis and implications of Japan sea drilling,
Nature, 309, 695-698. Proc. ODP, Sci. Results, 127/128, College Station, TX.
Hess, H., 1964. Seismic anisotropy of the uppermost mantle under Tanimoto, T. & Anderson, D.L., 1984. Mapping convection in the
oceans, Nature, 203, 629-631. mantle, Geophys. Res. Lett., 11, 287-290.
Isaccs, B.L., 1988. Uplift of the central Andean plateau and Tanimoto, T. & Anderson, D.L., 1985. Lateral heterogeneity and
bending of the Bolivian orocline, J. geophys Res., 93, azimuthal anisotropy in the upper mantle: Love and Rayleigh
3211-3231. waves 100-250s, J. geophys. Res., 90, 1842-1858.
Jolivet, L., Huchon, P. & Rangin, C., 1989. Tectonic setting of Vinnik, L.P. & Farra, V., 1992. Multiple-ScS technique for
218 V . Furra and L . Vinnik

measuring anisotropy in the mantle, Geophys. Res. Lett., 19, Vinnik, L.P., Farra, V. & Romanowicz, B., 1989b. Observational
489-492. evidence for diffracted SV on the shadow of the Earth’s core,
Vinnik, L.P., Kosarev, G.L. & Makeyeva, L.I., 1984. Anisotropy Geophys. Res. Lett., 16, 519-522.
in the lithosphere from the observations of SKS and SKKS, Vinnik, L.P., Kind, R., Kosarev, G.L. & Makeyeva, L.I., 1989c.
Dokl. Acud. Nuuk SSSR, 278, 1335-1339 (in Russian). Azimuthal anisotropy in the lithosphere from observations of
Vinnik, L.P., Farra, V. & Romanowicz, B., 1989a. Azimuthal long-period S waves, Geophys. J. Int., 99, 549-559.
anisotropy in the Earth from observations of SKS at Geoscope Vinnik, L.P., Makeyeva, L.I., Milev, A. & Usenko, A. Yu., 1992.
and NARS broadband stations, Bull. seism. SOC. A m . , 79, Global patterns of azimuthal anisotropy and deformations in
1542- 1558. the continental mantle, Geophys. J. Int., 111, 433-447.

Downloaded from https://academic.oup.com/gji/article/119/1/195/567244 by guest on 12 September 2022

You might also like