Oct 10, 1983 - inversion program written by Brian Mitchell. The final focal mechanisms adopted, which are all consistent with both. P wave first motions and S.
JOURNALOF GEOPHYSICALRESEARCH,VOL. 88, NO. BlO, PAGES 8160-8170,
TIlE
SHEAR-WAVE
VELOCITY
GRADIENT
AT
THE BASE
OF TIlE
OCTOBER10, 1983
MANTLE
Thorne Lay and Donald V. Helmberger
Seismological
Laboratory,
California
Institute
of Technology
Abstract. The relative amplitudes and travel times of ScS and S phases are utilized to place constraints on the shear-wave velocity gradient above the core-mantle boundary. A previously
have been directed toward obtaining global averages, and the degree of lateral variation in D" properties remains an open question. A conflicting result was found by Mitchell and
reported
Helmberger [1973],
long-period
ScSH/SH
amplitude
ratio
who utilized
the
relative
minimumin the distance range 65ø to 70ø is shown
amplitudes and timing of long-period ScS and S
to be a localized
by
phases
S phase, and
in D".
an amplitude
feature,
apparently
produced
anomaly in the direct
therefore need not reflect the velocity
gradient
200
km
of
the mantle.
ScSV arrivals
sensitive to the shear velocity above the core-mantle boundary.
The apparent arrival
time of the peak of ScSV is
7.6 to
as much as 4 s •reater than that of ScSHin the distance range 75v to
events
recorded
in
80ø for
North
Sea of
America.
Okhotsk
This
can be
localized
high
velocity
S-wave velocity
gradient
minimum in
the ScSH/SH
a
7.8
km/s.
These models can explain the
observedamplituderatio behavior, as well as an
apparent difference observed in the arrival of
explained by interference effects produced by a S wave velocity
the
found
low amplitudes of the ScS arrivals. Unable to explain this feature by models with negative or near-zero shear velocity gradients in D", they proposed models with positive S-wave velocity gradients above the CMB. These positive gradients extended over 40 to 70 km above the core, reaching velocities at the CMB as high as
calculations for the JB model or models with mild positive or negative velocity gradients in the lowermost
They
amplitude ratio near 68ø, which was attributed to
at the base of the mantle. The amplitude ratios that are free of this anomaly are consistent with
are particularly structure just
to constrain
transversely
and
radially
times
polarized
ScS.
Mitchell and Helmbergeralso proposed a low Q$
layer or strong positive
zone in
gradient in the lowermost 20 km
D",
or
finite
explain the baseline
outer core rigidity,
of
the
to
ScSH/SH amplitude
of the mantle. A velocity increase of about 5% is required to explain the observed shift between ScSV and ScSH. This thin, high velocity layer varies laterally, as it is not observed in similar data from Argentine events. Refined estimates of the outermost core P velocity structure are obtained by modeling SKS signals in
ratios. While the majority of their data was for deep South American events recorded in North America, they did analyze one deep Sea of Okhotsk event for which the radial and transverse ScS arrival times were not different, which suggested lateral variations in the D" velocity structure. In this paper we extend the analysis of ScS
the distance range 75ø to 85øß
and S phasesusing an enlarged data set in order to understand
Introduction
The
nature
of
diffracted
the
shear-wave
the base of the mantle has been a subject for many years. Gross
earth
derived
models
from
travel
times
and
discrepancy
studies
and
and Helmberger [1973]. ScSH/SH amplitude ratio
velocity
structure at controversial
the
S
the
between
results
of
A reinterpretation of the minimum indicates that
this feature is more clearly associated localized amplitude anomaly in the direct
free
rather
than
with
the
Mitchell
D" structure.
We find
with a S waves that
in
oscillations generally indicate smooth velocity gradients in the lower mantle, with slightly diminished gradients in the lowermost 200 km (D"
general the ScSH amplitude behavior is consistent with the JB model for two distinct lower mantle regions and do not find evidence for a major low
region).
Q$ zone at the base of the mantle. ScSVand ScSH
However, these studies have little
resolution
of the detailed
structure
of
the
D"
do show
region. Early investigations of diffracted SH waves, relying on classical ray theory interpretations, indicated very low S-wave velocities at the core-mantle boundary (CMB), and attendant strong negative velocity gradients within D" [Cleary et al., 1967; Cleary, 1969; Bolt
et al.,
1970;
Robinson and Kovach, diffracted
diffraction
SH
Hales
incorporating
theory
and
1972]. and
Roberts,
systematic
differences
for
one of
1970;
Recent studies of more
synthetic
timing
the regions, which can be well modeled by introducing a thin high velocity layer or strong positive velocity gradient above the core, but this thin zone varies laterally, and cannot be detected by using the ScSH/SH amplitude ratios alone.
Amplitude Data
complete
modeling
The S and ScS data analyzed in this paper
are
capabilities have proposed milder negative shear velocity gradients in D" [Mondt, 1977; Doornbos
from seven deep focus earthquakes in Argentina and 10 intermediate and deep focus events in the
and Mondt, 1979] and near-zero or slightly positive gradients [Okal and Geller, 1979; Mula and MUller, 1980; Mula, 1981]. These studies
Sea of Okhotsk, recorded at long-period WWSSN and Canadian Seismic Network stations in North America. stations
Copyright
1983 by the American Geophysical
Union.
1 shows the
for
both
locations
source
regions
of
of the Argentine
events.
The
parameters
for
are given
in Table
All
0148-0227/83/003B-1110505.00
impulsive
the
events
of the events were selected for their waveforms
and
for
their
the
and the
epicenters
Paper number 3Bll10.
8160
Figure used
source 1.
simple,
stable
SH
Lay and Helmberger: Shear Velocity
at the Base of the Mantle
polarizations. long period one
MBC
BLCm
E•M ßFFC ß SES
LHC
BOZ
ß
SCB
ON•
for
•RCDAAM•
DUG•
the
A representative profile of the tangential component seismograms for Sea
of
Okhotsk
events
is
shown
in
Figure 2. For these simple, impulsive waveforms the relative amplitudes and travel times can be accurately measured. Numerous other profiles of the SH seismograms are presented in Lay and Helmberger [1983].
CMC
YKCm
of
8161
•GOLFLO•
D
In order to correct the radiation pattern,
WES
determined
SCP
polarizations
,•BEC
'L
from
P wave
were
amplitude or
to
ratio
motions
extracted
or newly determined.
I '150W NOow
first
observed focal
and
S
wave
from the literature
Then the long-period
was used to refine
select
amplitudes mechanisms
between
various
SV/SH
the mechanisms
mechanisms
in
the
literature proposed for a given event. This was done by using a modification of a least squares inversion program written by Brian Mitchell. The final focal mechanisms adopted, which are all
N45*W
consistent
with
both
P wave
first
motions
and
S
wave amplitudes, are listed in Table 2, For the event of July 25, 1969, we could not determine a consistent mechanism. In only four of the 17 ß
WWSSN
ß
CSN
cases did
Stations
we
find
solutions
that
significantly
improved the SV/SH amplitude agreement over
Stotions
X Deep Argentine
Events
for
the
starting
mechanisms,
and
cases involved only 5ø changesin dip. Fig.
1.
Azimuthal
showing the
equidistance
location
of
epicenters and North
deep
projection
Argentine
American stations.
mainly due to the large scatter
event
that
two of these
This is
in the amplitude
ratios.
GSC,
For the
Argentine
events,
the
radiation
RCD,and SCHare approximately80ø from the
pattern corrections applied to the observed
Argentine source region. The hatchured region is the map projection of the deep mantle low
ScSH/SHamplitude ratios are all less than 12%, which reflects the stability of the SH radiation
velocity anomalyproposedby Lay [1983].
patterns to North Americanstations.
Since these
corrections are small, we include the uncoTrected ratios for the event of July 25, 1969, below.
radiation patterns
to
the North Americanarray.
Four of the Argentine events in our data set were
The horizontal components in the time interval containing the S and ScSphasesof all stations in the distancerange40ø to 80ø were digitized
used by Mitchell and Helmberger [1973]. They applied radiation pattern corrections to their ScSH/SHamplitude ratios that were generally
and
greater
rotated
into
radial
and
TABLE 1.
Date
transverse
Source Parameters
Origin Time
for
04:37:25.7 12:53:45.9 13:57:02.4
Events
Latitude
ñ 0.08 ñ 0.21
20% (B.
Used in this
J.
Mitchell,
Depth, km
Reference
52.56 ø ñ 0.022øN 153.67 ø ñ 0.030øE 52.15 ø ñ 0.018øN 152.57 ø ñ 0.025øE 49.5øN 154.4øE
424 ñ 4.2 466 ñ 2.7 136
ISC ISC NOAA
583
of
Okhotsk
Sept. 5, 1970
07:52:32.4
52.32øN
Jan. 29, 1971
21:58:06.7
51.72 ø ñ 0.032øN 151.04 ø ñ 0.024øE 540 ñ 5.7
May 27, 1972 Aug. 21, 1972 July 28, 1973 Sept. 21, 1974 July 10, 1976
04:06:49.6 06:23:48.6 20:06:35.4 15:54:59.1 11:37:14.0
ñ ñ ñ ñ ñ
0.25 0.16 0.15 0.37 0.14
personal
Study
Longitude
Sea
March 18, 1964 Oct. 12, 1967 Dec. 1, 1967
than
151.46øE
54.97 ø ñ 0.013øN 49.47 ø ñ 0.012øN 50.45 ñ 0.013øN 52.19 ñ 0.016øN 47.31 ñ 0.011øN
156.33 ø ñ 0.020øE 147.08 ñ 0.019øE 148.92 ñ 0.022øE 157.44 ñ 0.023øE 145.75 ñ 0.018øE
397 573 585 119 402
Strelitz ñ ñ ñ ñ ñ
2.8 2.2 2.1 3.5 1.7
ISC ISC ISC ISC ISC
Argentina
Dec. Mar. Dec. Jan. Sept. July Jan.
9, 5, 20, 17, 9, 25, 3,
1964 1965 1966 1967 1967 1969 1973
13:35:42.4 14:32:19.2 12:26:54.6 01:07:54.3 10:06:44.1 06:06:42.4 02:58:16.7
27.5øS 27.0øS 26.1øS 27.4øS 27.7øS 25.6øS 27.7øS
63.2øW 63.3øW 63.2øW 63.3øW 63.1øW 63.3øW 63.3øW
586 573 586 588 578 579 563
[1975]
Veith [1974]
NOAA NOAA NOAA NOAA NOAA NOAA NOAA
8162
Lay and Helmberger' Shear Velocity
at the Base of the Mantle
TABLE
communication, 1982). The focal mechanisms they used were also determined using SV/SH amplitudes but in P-wave
some first
cases were inconsistent motions. The scatter
with in
the the
amplitudes larger size
is substantial, and probably the of our data sets for each event more reliable mechanisms.
provides
The long-period peak-to-peak ScSH/SHamplitude ratios for the Argentine data are shown in Figure 3. Radiation pattern corrections have been
2.
Fault
Date Dec. March Dec. Jan. Sept. Jan.
Plane
Strike,degree
9, 1964 5, 1965 20, 1966 17, 1967 9, 1967 3, 1973
Orientations
Dip,degree
Rake,degree
78
-90
26
-68
43
-42
30
-44
171 12 30 28 3 357
19
-78
28
-83
84
-76
75
-52
applied. In the distance range 55ø to 75ø the S and ScS arrivals the amplitudes
are distinct phases can be accurately
Beyond75ø, ScS interferes
with
for which measured.
the instrument
42i I I I I I I I I I I I I I I I I 46
March 18, 1964 Oct. 12, 1967 Dec. 1, 1967 Sept. 5, 1970 Jan. 29, 1971 May 27, 1972 Aug. 21, 1972 July 28, 1973 Sept. 21, 1974 July 10, 1976
overshoot
50-
of
48 30 50 12 40 25 18 51 205 40
the
109
74
-77
77
-119
82
-93
19
44
76
-107
79
80
81
direct
apparent peak-to-peak
S
-87
arrival,
amplitude
and
There is a factor of 3 scatter at
distance,
which complicates
ratios.
This scatter
Figure 3 also shows the
58 •FFc
of
is primarily complexity
theoretical
ScSH/SH
amplitude ratio for a JB earth model. The curve was determined from long-period synthetics computed by using the Cagniard de Hoop
._IFBC '• 62DUG . ..•
generalized ray theory technique [•ee Lay and Helmberger,1983]. The effective t8 is the same for
degGsc
S and ScS for
calculations
_
this
and
presented
all
other
here.
synthetic
For a lower mantle
with a constantQ8 = 312 (e.g., the PREM modelof D•iewonskiandAnderson[1981]) the difference in t8 for ScSand S varies from0.2 to 0.0 s in the distance range 55ø to 80ø, which producesan
66GOL LHC _
•CH
insignificant
_
TUC
ratios.
70- ALO
effect
on the long-period
amplitude
However, for a model with a very low Q8
distribution near the CMB such as model SL• [Andersonand Hart, 1978], the difference in t8
_
increases 74- FLO
from
0.25
to
1.0
s in
the
same distance
range, which predicts a more rapid decay in the ScSH/SH amplitude ratios with distance than is apparent in the theoretical curve in Figure 3. The data in Figure 3 can be compared with that
_
MNT _
SCP
in Figure 6 of Mitchell
OGD
one omits their data points for Peruvian events and the Sea of Okhotsk event. Figure 3 has twice as many data points for the Argentine source
78-
GEO _
BLA
SHA
cluster
of low amplitude
JB model I
I
I
406.4
I
I
I
I 5216.41 I I 556,4
466.4
T-A'8.3,
2.
and Helmberger [1973]
if
region. In the range 65ø to 70ø there is a
ATL
82-
each
the interpretation
due to source and receiver structure as well as deep mantle structure.
_
the
of ScS diminishes
rapidly.
the amplitude
Fig.
87
Profile
of
North American stations
observations
646.4
s
tangential for
the
calculations,
components at September 5,
in this
ratios, but
well
there
range consistent
below
the
are numerous
with
the JB
model. At distances greater than 75ø the observed ratios drop off rapidly due to the interference
between S and ScS,
and
the
1970, Sea of Okhotskevent (d = 583 km). Direct
theoretical curve does also since the long-period
S is the first large arrival in each trace with ScS arriving around 580-600 s. Station JB travel
synthetics have similar interference. As is shownlater, similar data from the Sea of 0khotsk
time
anomalies
have
been
removed,
and
the
source region
ß The arrows indicate amplitudes are normalized
near67ø,
the arrival
of
detail
triplication
discussed by Lay and
[1983].
an ScS precursor produced by the
Helmberger
so
below.
do this
not
show an amplitude
anomaly
is
minimum
investigated
in
It is also important to note that
the amplitude ratios in Figure 3 scatter
in
range 0.2 to 0.75, which is significantly
shifted
the
Lay and Helmberger:
Shear Velocity
at the Base of the Mantle
I0 I I I I
i
8163
i
12/9/64
9/09/67
3/5/65 12/20/66 n'075
-
x
+
•P 7/25/69 X 1/03/73
1/17/67
_
+
l-
-050-
251-
01 55
•'•'^ •' '•,
i
i 65
i
•
i 75
-
X•F (Z) •,•FX •FX
u3025
x
l
85
-50
&, deg
-40
-30
-20
-I0
0
I0
20
Fig. 3. The long-periodpeak-to-peak ScSH/SH Azimuth, deg amplitude ratios for the Argentine events Fig. 4. Thesame datain Figure3 plottedas a recordedin North America. Radiation pattern correctionsfor the focal mechanisms given in
Table2 havebeenapplied.Thetriangles arefor
data recorded at azimuths to the east of N15øW from the source region. The curve shows the theoretical synthetics
ratios measured the JB model.
for
from
functionof azimuth fromthe source. Different symbols are usedfor eachevent. Only data in
thedistancerange55ø to 75ø areshown, because
the expected distance dependenceis small in this range.
long-period Geometric synthetics
relative to the range 0.1 to 0.5 spanned by the data in Mitchell and Helmberger [1973]. This shift, which apparently results from the difference in radiation pattern corrections applied,
is
baseline
of the data in
important
because
the
the
low average
earlier
study
was
cited as evidencefor a low Q8region at the base of the mantle. A close inspection
observations
that
of
define
the
the
individual
amplitude
ratio
spreading for a JB
corrections determined from mantle have been applied,
along with radiation pattern and event size corrections. The S waves show an azimuthal pattern, with relatively high amplitudes recorded at east coast stations. There is no corresponding
of 2
trend
in the ScS data.
east can account for the
ScSH/SH amplitude
for
of
the
Argentine
and
Sea
regions. The Argentine data
Mitchell
relative
of
the
stations
involved
[1973] reveals that all lie
on the
east
coast
of
North America. In particular, SCP, OGD, BLA, GE0, LND, MNT, OTT, and SFA (Figure 1) repeatedly have low amplitude ratios. Stations at comparable distance such as OXF, FLO, and LUB do not show low amplitude ratios. This indicates an azimuthally restricted anomaly, which probably does
not
Figure
reflect
3 all
radial
earth
of the observations
structure.
factor
in the
anomaly. Lay [1983] compared the long-period SH amplitude anomalies at North American stations
minimum near 67ø in our Figure 3 and Figure 6 of and Helmberger
The
to 3 enhancement of the S amplitudes
coast
Okhotsk
source
generally
show
enhancement of the SH amplitudes at east
stations,
which
indicates
that
the
trend
in
Figure 5 is not due to receiver structure. Further evidence that the SH waves from Argentina are anomalous is given by Lay [1983], who->studied the travel times from this data set. He concluded that the SH waves are 2 to 5 s late at the east coast stations and that this is
In
at azimuths
from
the source region east of N15øW are plotted with triangles. There is relatively little overlap between the two populations for this azimuthal separation, and all of the anomalously low ratios are
isolated
to
The Argentine
eastern
ScS
observations.
+
ScSH/SH amplitude
ratios
are
,•x
plotted as a function of azimuth in Figure 4. -The sharp separation of the low ratios along an azimuth of N15øW indicates the localized nature of the amplitude anomaly. The event of July 25, 1969, shows some relatively large amplitude ratios at eastern stations, but these may be erroneous because radiation pattern corrections were not applied for that event. The data at an
XX
+
-50
-•
-30
-20
-I0
0
I0
20•0
Azimuth, deg.
-•
•0
-20
-I0
0
I0
•
Azimuth, deg.
azimuth of 0ø are for station BEC, which is near 55ø distance. This station is isolated from the
Fig. 5. The long period first peak SH (left) and ScSH (right) amplitudes from the Argentine
east coast (Figure 1) and appears to be free of the azimuthal anomaly. The amplitude ratio minimum could result from either an S or ScS amplitude anomaly. Mitchell and Helmberger [1973] inferred that the ScS
events plotted as a function of azimuth from the source. Radiation pattern and geometrical spreading corrections have been applied as well as event size corrections. Note the relatively high S amplitudes recorded at east coast station,
phases were responsible based on the distance behavior of S and ScS amplitudes. In Figure 5 the zero line-to-first peak amplitudes for S and
whereas the ScS amplitudes at these stations are not enhanced. The symbols are the same as in Figure 4. ScS amplitudes at distances greater
ScS are
than 75ø are not included.
plotted
as a
function
of
azimuth.
8164
Lay and Helmberger:
because velocity 1700
to
are
they region 2700
not
km.
The
ScS
times
nor are
at
the
these
data
recorded
at
stations
S or ScS times
BEC. The map projection of this anomaly is shown in Figure 1. the
at the Base of the Mantle
encounter an anomalously low in the lower mantle at depths of
anomalous,
Since
Shear Velocity
lower
east
coast
.o 0.75
at
mantle stations
--= 0.50
appears to be contaminated by an anomaly in the direct S phase, we have removed the east coast observations
from
the
Argentine
data
set
E
'•'•0.25
in
Figure 6. The observations at BEC were retained since they are free of any obvious travel time or amplitude anomaly. As was discussed in Lay [1983], there may be an additional S wave amplitude anomaly in the midwestern and southern
9O
stations, with diminished Samplitudes producingFig. 7. Thelong-period peak-to-peak ScS/S largeScS/Sratios. This is
not as well
establishedas the east coast anomaly, but it
should bekeptin mindthat someof the larger valuesin Figure6 maybe dueto structurealong
theS-wave path. TheJB modelprovidesa reasonable fit to the average amplitude ratio
amplitude ratio for Seaof Okhotsk observations in NorthAmerica.Different symbolscorrespond
to different events. Thesolidsymbols givethe
meanandstandard error of the observations in
each5ø increment of distance.At distances greater than 75ø the amplitude ratio is
behavior,throughout the range55ø to 75ø and contaminated by interferencebetween S andScS. there is no clear fine structurerequiringlower Thelabeled curves are theoretical ratios mantle complexity. The average observed measured fromsynthetics for themodels discussed
amplitude ratio level is generally compatibleenvelope in the text. Thedash-dot the of theoretical ratios curves for indicate the models the same for S and ScS,whichindicates thatnolowQ8 withmild positive andnegative lineargradients
with the calculations for whicht8 is
zoneat the baseof the mantle is required by
this
data.
amplitude
Note
that
even if
some of
anomaly at the east coast
the
stations
due to the ScS phases, despite presented above to the contrary,
is
the evidence it is still
clear that the anomaly is azimuthally restricted, and no radial earth structure can account for it.
We have computed the
theoretical
ScSH/SH
amplitude ratios as a function of distance for modified JB models with positive and negative linear velocity profiles in D". The dashed lines in Figure 6 indicate the envelope of the
theoretical ratios for all models with
constant
gradients over 20 to 200 km thick zones with velocities at the CMBranging from 7.0 to 7.6 km/s (7.3 km/s for the JB model). Models with
stronger velocity
.o_
increases produce large
transition 7.0 km/s
zones reaching velocities less than produce a precursor to ScS which is not
observed, so these models can also be ruled
out.
The individual models produce fine structure in the ScSH/SH amplitude ratios not apparent in the JB model calculations, but the data scatter too much to resolve any such features. At distances
beyond75ø the
theoretical computationsscatter
more because of the variable interference between S and ScS, and little constraint on the velocity structure can be inferred.
Lay and Helmberger [1983] have shown that
an
S-wave triplication in the Argentine and Sea of Okhotsk data indicates the presence of a 2.75% shear-velocity discontinuity 250 to 280 km above the CMB. The presence of this structure does not
I0 l',,,, []
amplitudes around 75ø which are inconsistent with the data. Thin (