GEOPHYSICAL
RESEARCH
LETrERS,
VOL.18,NO.8,PAGES
1373-1376,
AUGUST
1991
ANALYSIS
OF SEISMIC
SV WAVES
IN THE CORE'S PENUMBRA
Thorne
Lay1andChristopher
J.Young
2
Data andAnalysis
Abstract.Vertically-polarized
S-waves(SV) attenuate
rapidly
as they diffract aroundthe Earth'score,whereashorizontallypolarizedS-waves(SH) propagateto large distancesin the
core shadowzone. The amplitudedecay of diffracted SV
signalsis so strongthat few studieshave beenmadeof these
phases,despitetheir acutesensitivityto velocitystructurejust
above the core-mantleboundary. We analyze SV signalsin
the penumbra of the core's shadow, finding systematic
waveform complexitiesindicative of local stratificationand
apparentanisotropyat the baseof themantle,with implications
for the dynamicprocesses
in thisinternalboundarylayer.
ObservedSV signalsfrom North American stationsfor a
deepfocusearthquakein the Izu slabare shownin Figure 1.
At distances
lessthan82ø thefirst strongarrivalis thedirect
SV wave, which has a turning point in the lower mantle
(belowAlaskain this case). The SKS phase(downgoingSV
convertedto P in the outercore,andthenbackto upgoingSV)
arrivesbeforethe core-reflectedScSV phase,andat a distance
of 82 ø, crosses over and arrives ahead of direct SV.
At
distancesgreaterthan 92ø, the SKKS phase(oncereflected
from the undersideof the CMB) separatesfrom the coda of
SKS andexhibitsa characteristic
phasedistortion.
Introduction
The 200 km thick D" region at the baseof the mantle is a
major thermal,andpossiblychemical,boundarylayer in the
05/13/77
mantleconvection
system
[Lay,1989]. Seismic
wavesthat
._%e
..... ' '
diffractalongthe core-mantleboundary(CMB) aresensitiveto
velocity structurein D", and have been extensivelystudied
[Young andLay, 1987]. Typically, diffractedwave analyses
usecompressional
waves(P) andhorizontally-polarized
shear
waves (SH). These wavesboth lose energyslowly as they
diffract andcan be observedthroughoutthe coreshadowzone
(angulardistances> 90ø). Vertically-polarizedshearwaves
(SV) havereceivedscantattentionbecausethey loseenergy
rapidly asthey diffract,andSV observations
beyond105ø are
rare.
Inefficient
SV
diffraction
is due
Paper
1991 by the American Geophysical
number 91GL01691
0094-8534/91/91GL-0169153.00
79'•KS
•lSO
84 •UG
Union.
.t•
;BC
.
--
m 89
interferencewith ScSV and the strongdensityincreaseat the
CMB [ChapmanandPhinney,1972].
Vinnik et al. [ 1989]recentlypresented
a few observations
of
SV arrivalsat largedistancesin the shadowzone(> 105ø) that
could be caused by splitting of diffracted SH due to
anisotropicstructurein D" or by strongnegativevelocity
gradientsin D" closeto the critical gradientwhich prevents
diffractionfrom occurring.Sincetherearemanyobservations
of rapid amplitudedecayof SV near the onsetof the core
shadowzone [e.g. Kind and Mtiller, 1977], the observations
of SV signalsat large distancesare probablyassociatedwith
laterallyheterogeneous
structurein D", but additionalanalysis
of SV wavesis clearlywarranted.
This study analyzesSV signalsnear the onsetof the core
shadowzone, termedthe penumbra,where diffractioncauses
a rapid decay of SV with increasingdistance. SV signals
providea differentsensitivityto D" velocitystructurethanP or
SH phases,and the subtledestructiveinterferenceeffectsthat
preventefficientdiffractioncanbe usedto detectbothinternal
stratification and anisotropyof D". Both features have
importantimplicationsfor processes
occuringnearthe CMB.
Copyright
74
;OR----•_•
,•V•
to destructive
1Institute
of Tectonics,
University
of California,
SantaCruz
2Sandia
National
Laboratory
d = 448 km
94C •- - -'•'-I
IJeT
.1•
pU-----•,
424
454
I
••
484
514
544
574
T - Delta x 9.0 (s)
Figure 1. Radial componentSV signalsrecordedby longperiod seismographicstationsin North America for a deep
(448 km) Izu eventof May 13, 1977. First arrivalsare aligned
on the superimposed
travel-timecurvesfor the PREM model,
andpeakamplitudes
in eachtraceareequalized.
SV diminishes rapidly beyond 94ø, the onset of the
penumbrafor these20 s periodarrivals,limitingthenumberof
high signal-to-noise
observations.This is not due to radiation
pattern,for the focal mechanismhas stableSV radiationover
the entiretake-offanglerangeencompassing
SV andSKS
(SKS is reversedin polarity relative to SV by the phase
conversions at the CMB). The diffracted SV (SVdiff)
waveformsnear 99ø have an extra wiggle. This featureis
observedfor othereventsandotherstations(Figure2) soit is
not a receivereffect. The SVdiff complexityis unlikely to
arisefrom sourcecomplexitygiventhe simplewaveformsof
1373
1374
LayandYoung:
Analysis
ofDiffracted
SVWaves
SKS•
MNTA=100.5
SKKS
It
thesedistances,the SV wavefronttraversesD" at grazing
anglesand any reflectionsfrom layering are post-critical.
Thus,thewavefieldis primarilysensitiveto theshearvelocity
structure,not to density or P velocity. The densityand P
velocitymodelshave 1.5% increases
at the samedepthsasthe
SVdiff
SCP
A-103.2
•
OGD
A=104.5
•
1/31/73
shearvelocitydiscontinuities.SmoothreferenceEarth model
PREM predictsa simpleamplitudedecay,with no waveform
1/31/73
I
I
I
0
40
80
complexity,while thediscontinuity
modelspredictcomplex
SV waveforms.The first arrivalin eachcasecorresponds
to
sec
Figure 2. ObservedSV signalsfrom Izu subductionzone
events. SV complexityassociated
with a D" discontinuityis
observed for different events and different stations.
the SKS and direct SV signals,which have take-off angles
bracketingthe SVdiff energy. This featureis stable,though
only a few signalsareuseabledueto low signallevels.
Complexityof SVdiff is predictedby modelsof velocity
structurein D" derived from SH-waves. Young and Lay
[1990] have analyzed SH signals also bottoming below
Alaska. They proposemodel SYLO, with a 2.8% S-wave
velocity increase243 km above the CMB, to explain an
apparenttriplicationin the SH wavefield. Model SYLO also
produces
an SV triplication,with extraarrivalsbetweendirect
SV andScSV at distances
lessthan82ø, andfollowingthefirst
SV arrivalat distances
beyond90ø. The closerdistance
range
is complicated
by thearrivalof SKS,sotheonlypracticalway
to lookfor theSV triplication
is at distances
beyond90ø.
Figure 3 presentssynthetic SV waveform profiles for
modelswith varying D" shearvelocity discontinuities.At
energy turning below the discontinuity,while the secondis
due to post-criticalenergyrefractingalongthe discontinuity.
Model SYLO predictsa strongdovetailin the SV waveforms
over a limited range, but the effect is more pronouncedand
shifted in distance relative to the data in Figure 1.
Discontinuitiesat depthsof 200 or 150 km above the core
produce better agreement with the observations,with a
convergence
of the two pulses.For deeperdiscontinuities
the
secondarrival is not observed.The slightnegativevelocity
gradientsin D" (taken from SH model SYLO) increasethe
amplitudesof SVdiff relative to PREM, but much more
pronounced
velocitydecreases
areneededto explainthelarge
diffracteddistanceobservationsof Vinnik et al. [ 1989].
The SV and SH waveformsin the core penumbrafor the
May 13, 1977 eventcan be modeled(Figure4) by a D" shear
velocitymodelwith a discontinuity
of 2.1% at a depth175km
abovethe CMB (Model SA in Figure3). Model SA provides
SCH
A=94.1
SYLO•.•.•J
SCP
A=102.9
..........SH
s^...........................
/ •\,.:::
................................................
't•.•.;
.........
85
•
SVdiff SyntheticProfiles
'
•
•
ß" t ,.o.
oO
OTI'
SYLO
A=99.8
BLA
"• ..........., ,•
'"'
A=104.7
.:..,_¾-v ......
....
.......SH
s^.................................
...................................
••-•,
,..
.........
'..'.:::::'2::::..":?'j•.".:."z..'"'•./.
;;..'.7;,'•::..15::::::::
\/....
SYLO/),,
._
•' .•..,.._
,,.......................
105
- -•-
110
•'--•
851
i
-
,•
s^.:;:
•...":::::.:::,::::;;;;;;:'.;2;;:..';;;;;;;;
sv:::::::::::::::::::::::::::::::::::::
t....."
•'
•.....-"
"
- •
,
i
I
•
I
oo
go•
Figure4. Observedand syntheticSH and SV signalsin the
core'spenumbrafor the event of May 13, 1977. Synthetics
areshownfor modelsSYLO andSA fromFigure3.
•
/
110
•
Time(s)
I
80
;t 8
00r
602
I
40
sec
i•
95
582
!
0
622 582
602
•me (s)
622 6.9
7.1
7.3
7.5
V (km/s)
Figure3. Syntheticradialcomponent
SV signalsfor a variety
of D" shearvelocity models(lower right). The reflectivity
methodwas usedto computethe seismograms,
andthe source
parameters
areappropriate
for theeventin Figure1.
a goodfit to theSV data,andan adequatefit to theSH signals.
The reflectivity synthetics include a fault mechanism
determinedby the HarvardCMT inversion.For the periods
being examined,the modelingcannotdistinguishbetweena
sharpdiscontinuityor a transitionzone with the velocity
increasedistributedover a depthrangeof up to 50 km. We
usesharpdiscontinuities
for simplicity,but this shouldis not
requiredby thedata. TheP velocityanddensitystructure
also
have2.1% discontinuities,
but thesynthetics
arenot sensitive
to those parameters. An outer core P velocity structure
proposedby Lay andYoung [ 1990] is usedin the calculation
of the SKS and SKKS arrivals.
This core model is identical to
LayandYoung:Analysis
of Diffracted
SV Waves
PREM exceptin the outer 100 km, wherethe P velocitiesare
smoothlyreducedto 7.9 km/sat the CMB.
The basic waveform
characteristic associated with the shear
velocity discontinuity in D" is the dovetail in the SV
waveformsbeyond94ø, andtheassociated
secondary
arrivalin
the SH signals. The timing and relative amplitudeof this
featureprovidesthe sensitivityto the depth and size of the
discontinuity. The differencesbetween the SYLO and SA
syntheticsare more pronouncedfor SV thanfor SH, thusthe
SV dataactuallyprovidebettersensitivity
to thestructure
than
the SH data. Destructive interference between direct S and
ScSV enhancesthe secondaryarrival refractedalong the
discontinuity.There are, however,limitationsin the sensitivity to the D" structuredue to the grazingtrajectoryof the
1375'
BLA
•
9/10/73
•
SH
A = 95.7
SV
WES
•_.••
8/13/67
•
A
= 98.9
I
0
I
I
40
80
sec
Figure5. Recordings
showingdelayof theSV onsetrelative
to SH. TheSKSphase(firstarrivalontheSV component)
is
cleanlyisolatedon SV, makingit unlikely that receiver
anisotropyis responsible.
entirewavefield.In particular,thedepthof thediscontinuity
ScSon the two components
do not accountfor this shiftof the
trades-offdirectlywith the velocitygradientbelow it, as this
controlstherelativetimingof thetwo arrivals.We havekept onsettime. We havebeenunableto explainthissplittingby
any isotropicstratifiedmodelfor D".
this gradientfixed to the preferredD" gradient from SH
To furtherexplorethecharacteristics
of theSV-SHsplitting,
modeling,but thismay notbe appropriate
for theseparticular
we examinedigitized,rotatedseismograms
from a total of 21
paths. The SV signalsat closerdistances(Figure 1) are in
eventsin the westernPacificwith pathssamplingthe deep
agreementwith thismodelas well, but are notdiagnosticwith
mantlebelowAlaska. Of theseevents,16 provideat leasta
respectto thepresenceof a triplication.
We have not exhaustively explored the model space few reliable observations of diffracted SV-SH differential
compatiblewith thesewaveforms,sincethe dataare sparse; onsettimes, with 6 eventsprovidingScSV- ScSH timesfor
closerdistances.Typically,SKS signalsfor theserecordings
however,the few clear SV signalssupportthe existenceof
were cleanlyrotatedontothe radial component,or the direct
localizedD" stratification.The precisedepthand size of the
SH andSV onsettimeswerenotclearlyshifted,suggesting
shearvelocity discontinuityin model SA differs from model
thatreceiversu'ucture
produces
onlyminoronsettimesplitting
SYLO, but the lattermodelis an averagestructurederivedfor
for the long-periodrecordings. This is consistentwith the
an extensive,1500 km scaleregion in D", which may well
small time shifts(usuallylessthan 1.5 s) observedfor even
havelateralvariationsacrossit. In addition,anisotropymay
the
strongestanisotropicreceivereffects.
exist in D" (see below), makingit impossibleto fit both SH
Under the assumption that deep mantle structure is
andSV with a singleisotropicmodel.
for the instanceswherewe do observesplitting,
The strongattenuationof SV signalsin thecore'spenumbra responsible
theobserveddifferentialtimeswereplottedat the bottoming
limits the number of additional observations available to
exploreD" stratification.Generally,theSV signalsarevery point for the associatedpath (Figure 6). SKS-SH and SKSScSH anomalies relative to PREM calibrate the extent of D"
small beyond105ø, on the order of 10% or lessof the SKS
heterogeneity. Filled symbolsindicate fast SH or ScSH
amplitude. In many cases,the amplitudesare too small to
arrivals.
The uniform distribution of fast SH arrivals is
evaluate whether a shear velocity discontinuityin D" is
consistent
with the fastD" layerin modelSYLO. Theremay
requiredor refuted by the data; however,when clear arrivals
are present,they tend to supportthe discontinuitymodel for
pathsunderAlaska. Thereis lessevidencefor SV complexity
for pathsunderthe centralPacific,which is consistentwith the
lack of evidencefor a D" discontinuityin thisregionin the SH
wavefield [Garnero and Helmberger, 1988]. Our SVdiff
observationssamplingunder the Indian Ocean and central
Eurasiaareinadequateto convincinglyarguefor or againstD"
stratificationin thoseregions. An effort to improve the data
qualityusingbroadband
digitalrecordings
is needed;however,
presentlyonly sparse,isolatedobservations
are available. SV
signals
nearthepenumbra
clearlyhavepotential
for extending
-5
s+5
si
our understanding
of D" stratification.
For the majorityof the usefulSV observations,
we observed
close agreementbetweenSV and SH arrival times, but in a
SV-SH: {•
few casesthe two components
appearto be significantly
shiftedin onsettime. Examplesof relativelydelayedSVdiff
arrivalsin the penumbraare shownin Figure5. The clean
rotationof the SKS arrivalsuggests
thatthisis not a resultof
anisotropic
shearwavesplittingunderthereceiver.Synthetic
SV and SH calculationsfor the corresponding
sourcesand
between diffracted SV and SH onset times and ScSV-ScSH
receivers show that differences in interference between S and
peaktimes.X's indicateno shiftin onsetor peaktime.
Figure6. Surfacecoordinates
of mid-points
for raypaths
for
which differential times have been measured. Symbols
indicate the magnitudeand sign of the differential time
anomalies. a)
SKS-SH and SKS-ScSH differential time
anomalies with respect to PREM.
b) Differential times
1376
LayandYoung:Analysis
of Diffracted
SV Waves
be a westward transition to slower structure near the area
sampledby the Izu events. Figure 6b showsthe diffracted
SV-SH and ScSV-ScSH anomalies. We relied on simple
measures
of splitting,involvingclearshiftsof onsettime for
diffracted arrivals and peak arrival times for ScS. The
differential ScS anomalieswere made relative to complete
synthetic waveforms computedfor PREM to avoid any
contaminationfrom SKS and SKKS interferencein the long-
periodsignals.Theeastern
portionof theregionhasrelatively
delayedSVdiff andScSV arrivals,with anomalies
aslargeas
4 s. The western-central
portionof theregion,includingthe
areasampledby pathsfrom the Izu sourceregionthatexhibit
complexityof diffractedSV, showsno consistentshift in
onsettime. In thisregionsplittinganomaliesarelessthan2 s,
andmay be zero.
Many of the split S arrivals in Figure 6b are ScS
observations,some being shown in Lay and Helmberger
[1983]. They modeledthe ScS waveformsby introducinga
strongpositivevelocitygradientabovetheCMB, whichshifts
observationthat the regionunderAlaska appearsto showboth
laterally varying, but extensive stratification, as well as
laterally varying S wave splitting indicates a complex
environment.If the splittingis attributedto anisotropy,the
lateral variation could be the result of gradientsin the flow
structurein the thermalboundarylayer;however,it may also
be possibleto explainsomeof the splittingby localizedstrong
positivegradientswithin the stratifiedzone,perhapsresulting
from additional chemical heterogeneity. More complete
imagingof D" structure
is neededto constraintheprocess.
Acknowledgments.
G. Mtiller andW. Schottprovidedthe
SV and SH reflectivityprogramsusedin this study. Two
anonymousreviewersprovidedconstructivecomments.This
work was supportedby NSF grantEAR-8451715 and the W.
M.
Keck
Foundation.
Doornbos et al. [1986] show that the effects on P and SH
diffraction parametersare relatively mild, but have not
number
116 of the
References
the ScSH and ScSV arrivals because of differences in how
energyreflectsfrom the gradient.However,synthetics
show
that a positiveD" gradientdoesnot producethe clear onset
shift of slightlydiffractedsignalslike thosein Figure5. An
alternative interpretation[Doornboset al., 1986] is that
anisotropy
in a stronglyshearedthermalboundarylayermay
causesplittingof phasesthat sampleD". This could arise
from intrinsic anisotropy of silicate-perovskite grains
accentuated
by flow fabricsandlayeringin theboundarylayer.
Contribution
Instituteof TectonicsandRichterSeismological
Laboratory.
Chapman, C. H. and R. A. Phinney, Diffracted seismic
signals and their numerical solution, in: Methods in
Computational Physics, B. A. Bolt, Ed., 12, 165-230,
1972.
Doornbos, D. J., S. Spilopoulos, and F. D. Stacey,
Seismological
propertiesof D" andthe structure
of a thermal
boundary layer, Phys. Earth Planet. lnt., 41, 225-239,
1986.
Garnero,E., D. Helmberger,andG. Engen,Lateralvariations
near the core-mantle boundary, Geophys.Res. Let., 15,
exploredtheSH-SV characteristics.
Vinnik et al. [1989] also
609-612, 1988.
suggest
thatazimuthalanisotropy
in D" couldbe responsible
Kind, R., and G. Miiller, The structureof the outer core from
for observations
of SV at largedistances
in theshadowzone.
SKS amplitudesand travel times, Bull. Seisin.Soc. Am.,
67, 1541-1554, 1977.
Discussion
Our observationssupportthe suggestionby Vinnik et al.
[1989] that SV in the core shadowmay reveal important
attributesof D" velocitystructure.In particular,whensuitable
signalqualityis availableSV signalsnearthecore'spenumbra
are particularlysensitiveto any stratificationin D". This is
becausethe energyturningjust abovethecoreis weakenedby
the onsetof diffraction, allowing post-criticalenergy grazing
alongany shallowerdiscontinuities
to be clearlyobserved.In
addition,the timingof diffractedSV andSH canpotentiallybe
usedto detectanisotropic
or othercomplexstructurein D".
Mostpreviousworkon modelingD" shearvelocitystructure
hasreliedon SH phases.It is encouraging
thatthe SV signals
exhibitcomplexitythatcanbe modeledby the sameclassof
modelsinferredfrom SH data. Simultaneous
modelingof SH
and SV for otherregionsusingbroadbanddata may provide
tightconstraints
on laterallyvaryingD" velocitystructure.
Lay [ 1989]hasdiscussed
theimplications
of a shearvelocity
discontinuity,or transitionzonenearthe top of the D" layer.
This may be mostreadily interpretedas a manifestation
of
chemicallayering, althoughthe structureis not laterally
continuous. Doornboset al. [1986] have discussedpossible
anisotropy
in D" in termsof strongshearflowsin a thermal
boundary
layerin thelowermost100km of themantle.Such
a thermalboundarylayeris requiredif recentestimates
of very
high temperaturesin the outermostcore are correct. The
Lay, T., Structure of the core-mantle transition zone: A
chemical and thermal boundarylayer, EOS, Trans. Am.
Geophys. Un., 70, 49, 54-55, 58-59, 1989.
Lay, T., and D. V. Helmberger, The shear-wavevelocity
gradientat the baseof the mantle,J. Geophys.Res., 88,
8160-8170, 1983.
Lay, T., and C. J. Young,The stably-stratified
outermostcore
revisited, Geophys.Res. Let., 17, 2001-2004, 1990.
Vinnik, L. P., V. Farra, and B. Romanowicz, Observational
evidence for diffracted SV in the shadowof the Earth'score,
Geophys.Res. Lett., 16, 519-522, 1989.
Young, C. J., and T. Lay, The core mantleboundary,Ann.
Rev. Earth Planet. Sci., 15, 25-46, 1987.
Young,C. J., andT. Lay, Multiplephaseanalysisof theshear
velocity structurein the D" region beneathAlaska, J.
Geophys.Res., 95, 17385-17402, 1990.
T. Lay, Institute of Tectonics and C. F. Richter
SeismologicalLaboratory, University of California, Santa
Cruz, CA 95064.
C. J. Young, SandiaNational Laboratory,Division 6231,
Albuquerque,NM 87185.
(ReceivedJanuary25, 1991;
revisedMay 15, 1991;
acceptedJune24, 1991)