Title Two-layered structure in the lowermost mantle beneath thesouthwestern Pacific( Dissertation_全文 )

Author(s) Yamada, Akira

Citation Kyoto University (京都大学)

Issue Date 1997-03-24

URL https://doi.org/10.11501/3123273

Right

Type Thesis or Dissertation

Textversion author

Kyoto University

学位申請論文

印何回山

Two-layered structure in t h e lowern1ost n1ant le b eneath t h e southwester n P acific

r\kira Yamada

Department of Geophysics. I\yoto Uni\·ersity. T-.:yoto, .Japan

Abstract . The heterogeneities in the D" layer beneath the southwestern Pacific are investigated by using the

combination of earthquakes around the Fiji Islands and receivers in the .Japanese Islands. The data of short-

period P wave are processed by the double-array stacking, a method to stack simultaneously the records from

different events. The estimated depth of the D'' reflector varies from 270 km to 170 km in the southwestern

Pacific. The numerical experiments using various P-velocity models are performed and the rapid velocity

increase well explains the waveform of PdP. This increase may be at the top surface of the high-velocity layer

(or high-velocity lamella).

The data of broad-band transverse component are also examined. The D'' reflection (SdS) is clearly seen

in some traces. The polarities of SdS are almost the same as those of the direct S wa\·es, which supports the

D" t·cflcctor \\'ith the rapid S-wa.\'e velocity increase with depth, as well as for P wave. \\'hile the residuals of

difl'erentia l travel time of ScS-S show fairly large variations, they show systematic large delays of ScS travel

times. These delays may be caused by the existence of an S-wa\·e low velocity layer right abo\·e the (;f.!B.

A probable cause of both P- and S-waves high-velocit.y anomaly is considered to be subducted slab material.

The low S-velocity layer underlying the high-velocity anomaly might form a shape like a dome, which could be

a seed of mantle plume.

I nt r o duct io n

The lowennost manti<'. the f)" layer, i;; Oil<' of tlw 1110:<1 cu riou:; n·gion 111 t lw l';nl h. ht'Cilll"'' of i Is st ron.:;

lwterogt>lteities. its nature a:< a 1 hc•rnml boundary layt'r het 1\'<'<'11 liu' liquid iron-alloyed corl' and t lw ;-;olid silicate

mantle <11HI its i111portant role on tlu' earth's d,Y IIalllic:; [e.g .. rc·vi,~ll' hy Loper and Lay. 19!).->]. Sn-erallaypotlae,.;e,;

ill't' ())'0()0SI'd Oil thl' ori•·in or the• D" lan•r: tlw ''J'il\'f'\'ard or''"' :'llhdurted ,.;]ah lllillc'ri;d [•' .. :!: .. ll ol'lll:lll ll and 0 . 0 .. ·"

\\'hit<'. l!lt':.!]. tlw <tCCUIJJidation uf tla·· producls rau ... ,•d hy th<' rlat'IJJic;d rc'<trliull bf'tl\'<'<'11 tlw lo\\'<'1'1110,..1 IJJilltlk

<111<1 0111 <'I' con· [.J •·a nloz. I !l!lO] and n•r,·nlly. t II·· arc n11111l:H iun of 1 t.,, •'X t r;1ct ion I·~ 1 II·· ··r~ ,I all iza 1 ion <'f 1 lw

inn•·r cor<' [Ito et al .. I!JD:>]. Tlw corn'la tion lwt\\'t'•·n 1lw so•isn10logical llett'rog<'rlt'ity in til·· do•ep nwntl t> and

tlw pnst subduction zorws is discu~st'd by Richards and I::ng1•hrPison [199:.!]. The lwtt>rogo•iwitiP~ in D" hy 1 lw

chemical reaction could he rdated with the tPmperatun• varia! ion in D".

The ,·ariations in the wa\'elcngt h of the heterogeneit it::> have a range from a few thousands kilometers. inferred

from the global tomographic technique [e.g., Sengupta and Toksoz, 1976: Inoue et al., 1990: Suet al., 199-1]. to

a few hundreds kilometers [Haddon and Cleary, 197•1; \\'eher and Da.\'iS, 1990; \'idalc and Benz , 1993; 1\:riiger

et al., 1995; Tono and Yomogida, 1996]. The great s ize of ·•avalanches" (subclucted slab metlerial) [e.g .. Tackley

et al., 1993] and the rather small size of '' inclusions" by th<' chemical reaction could bt• a.-.;sociated with the scale

variations of the heterogeneous structure in D".

The aims of this study arc to inq~stigate the shorr.-scale D" heterogeneities and to infer the physical and

chemical images of the D" layer. \\'e present the resu lts of hot.h data analysis and nulllerical experiment.. \\'e

analyse the tra\'el times and the \\'aveforms of direct P and S waves, D" reflect ions (PdP and SciS) and C~l8

reflections (PeP and ScS). Figure 1 shows the raypaths of the::w waves. Throughout this study, the differential

travel times between each reflection (PdP. SdS, PeP and ScS) and the direct wa,·e (P and S) are used to

compensate the anomalies caused by the heterogPneitio·s bt•neath the recei\'crs and the epicenters, because

the raypaths used here are almost identical to each other in th<' upper mantle and 1he crust, where lateral

heterogeneities may be strong. The anomalies in the din·,·rcntial travel times will be assumed to be caused by

the heterogeneities in the lowt•rmost mantle or departures from reference earth model.

The region sampkd in 1his :;tudy i~ located heneath tilt' ~oulllw.-:::tt'rn Pacific (Figun· :?). The global studi•·,;

\'docity n10dels for this region itrl' propo,.,cd [i~•·nd:dl <~nd ~llt'an·r, I!Hl:i; \\'eher and 1\orn•~. l!)~l:!: Sltihuta111 •'t

al., 19!):3 ;1nd Y;11nada <llld ~<lkanishi, 109Cia <11HI hJ. Tlw low \'l'iority lay"r with tlu• 111irkiwss of IU '"" [.\ Iori

Data Analys is

'''rlrral-conJporlt'lll rt'rord:-; fro111 1 hn••• arrays \\·ir h dianJt'l•·r sil.t's of a ff"w hundn·d kilolllo'tPrs. (:.!) short-pnic>d

\'t>rliral-conrpont'IH n•cord..; from an array with of It'll kilon~t'l••rs and (:3) broad-hand horiwnral-comporwur

H'Cords by STS I >'eismonwter:>. Tht' station location,; ar•' indicatt>cl in Figun• :l.

Three large arrays in Figure ;j are to observe the micro earthquakt>s under the national project of earthquakt>

prediction. which are opt'ratt•d hy 1\yoto University {Disaster Prevention Research In,;titute). the (.;niver,;ity of

Tokyo (Earthquake H.e,;earch Institute) and Hokkaido University. respectively (hereaft.cr L,\1, Lr\2 and L,\:3).

ThC' smal l array located in central .Japan is operated hy the .Japan ~lt'teorological Agency {Matsushiro Sei,-rnic

Array System, \ISAS her••after). The data analysis by u::oing L:\1, 2 and :) is recently pre:>cnll'U Ill Yamada and

:\akanishi (1996a and b] and by u:.ing \ISAS in Yamada ct al. (1996]. The results of tlwsc data analyses arc

used in this paper to obtain dt•arcr image of the D .. layer. In addition to the above four arrnys for the P-wnw

analy:.es, the broad-band records (1313 in Figure :3) a rc analpwd to st.ud_v S-wave structure of the IY layer.

\Vt• usc earthquakes with det'p focus(> :300 km) and l aq~t' magnitude (111b > 5.5) which occurred around the

Fiji Islands. The details about tht' event selection are found in Ynmada and Xakani~hi [I99Ga and b]. The event

parameters used in this ~tudy are summarized in Tahl" I. Table 2 shows the range of the epicentral distance

and the number of records used in this study.

Figure ·1 shows the close-up of tlw bounce region {bcrwat h southwestern Pacific). \\'e project the distribution

of 1 hcse bou nee points and the Fn•snel I' ll i pses on the ••a rt. h '$ su rfacc. The lengt h sea lc at l he depth of t.lw C.\ 113

is about half compared with that on the earth's surfacP.

S hOI·t-p<•t·iod P wav<'. Tlw llt '<"•·~siry of u,;in~ lit•' -.larkin;!. lt'dtniqn•· for llw analyst'~ of rlw shorr-p .. riod I'

\\'i\\'t' roiJH'S front t h,• following rl'a~ull. Tlte irnp"danr•· ron I r;tsl al llw D .. rt'llt'rlor 11111sl h·· "'"all, \\'hirh '"

inft'ITI'd froll1 the fad that 110 01' r .. w ohst'l'\"i1l ions of llw sltorl l'•'riod I' r.-llt•cl ions frO Ill I h·· lwl t'I'O~I'no'Oih I )"

has h,•,•n doll\' i 11 a ~i n~l·· ro·rord. To sf ack ro•rords frn111 d i lft•ro'nl •'\'t'lll ~ ,,.,. ""'' a don hlt•-a rray sr ack 111 ~ nwlltod

(Yantatln <Ill I '\-,k:lllishi . (!l'l(i;J ollld l·jntslt'ad of f t\IJ\t'lll ional ~r;,, · kin~ lllt 'l hod, t.y "'ill:!, I J., , ,Jr,\\'flt ·,;,; "" I h··

unk11n\\'11 paranwlt'r.

\'

/).-1(/.rl) = .~· t ct·;.r;(t + ('!;''- '!'() + (~7i(d)- l"TJ(d))). •=I

wlu•r•· I indicates the Ira1·.-1 tinH'. rl th·· distance of the n•fkctor from tlu> ('~lB. i iud••x lltlllllwr forth~ t'l't•nt-

rPceiiW pairs (i = 1, .\'. th~ nu111her of.\' for each region i" li:-.ted in Tablt> 2). i =ltht• pi1·ot pair. !Ci amplitudt•

normalization factor . .t•i{/) th~ ohs,•rvt>d seismogram. 1t onset time of P wave of the i-th pair, and l'J;(d)

theoretical differential traveltime between P and PdP of the i-th pair for t.he reflector at d km from the C~ t n.

In this study, the range of the d-\'aluc for the detection oft lw D" reflection is from 0 km to 400 km with an

increnlt'nl of 10 km . The amplitudes of all traces arc normnlized by the maximum peak-to-peak amplitude.

1;" is read from the original ::.cismogram. The correction for the polarity re\'ersal due to source mechanism is

included in u,. ~lore det<~il:; about the procedure and the accuracy estimation of tl11s method should be refern•d

to Yamada and ~akanishi [l99Ga and b].

Dro ad-band S wave . Two horizontal components of broad-hand data arc rotated lo t.he radial and th••

trans\'crse components. \\'e rind that the signal-to-noisl' ratios of the radiC'II components are low in the data

ust>d here. ~loreo\'er radial-polarizl•d SKS is expected to arri\'e in the time window bctwePn the direct S wwC'

and the C\IB reflection (ScS) and may contaminal<' the I)" n•fl,•ction. Therefore we analyse only tranS\'CI'Sl'

components in this study.

The one of the ai tns of t.his st.udy is the in\'estigation of the shorl- ll' <l\'elengt h (a few hundred kilometers}

ht•lt>rogeneities in 0". To slack n·cord::; from widt' region may avN<~ge the ~lrongly hl'lt•rogeneous D" and may

hroad-h<tnd st•i:<Jnograt>h,.. u,.. •. <l h••n· l···rall:'•' of tl~t· \l'icl•· di,..Jrihul ion of "t<ttiOJb ( l'•;.;un· :1) and hou11C•' poiul"

(l:i!-\un· I).

In ronlrast 10 tlu• shorl-p•·riod I' 1\':t\'t•, 1he I)" rd l··rlion a11cl n• l'r:H· Iion of' llw S wa1·•· :tn' ohst·n·c·d in tht•

,;int;lc• traCt' from the Ions l"'rind d:tla [<'.,!.!, .. c:alwrl~ and Lay. l~l!l:.! and 1\··ndall illltl Sllt'art·r. I!Jtl.i]. \h· t'Xilllllllt'

Shod-pt'r iod P wavt!. Tht> dt'lnilt>d Pxplanntion" for tlw r••=-u lt:; ;ue• descrihPd in our ll.'rt'lll papo•rs [Yamada

ami :\akanishi, 1996a and h for thro•c largf' arrays ( L \ t. :.! <111d :)) and Yamaeh "' al.. IU!Hj for a :>mall arra~

(:\IS.-\ <;J. :land :3)). llert' we• pre,.,ent an outline of these n~:-.ults.

The st·1ckt>d traces obtained from the abow f'quation arc converted to the enwlopt·s by u:-;.ing the Hilbert

transform. Figure 5 shows the ~ D images of Lhe e1l\·elopes for the regions LA I, ~ and :~. The en \·elope amplitude;;

a rc plotted as a function of t lw difl'er·Pntial t ra \·eltinw with rl'spcct. to the d irect. P wave and the distance of tlw

D .. reflector from the C:\IB. It is found that the estimated de~pths of the D .. reflectors vary from 210 km (LA I)

to ITO km (L:\:3). The ratht•r small amplitude of the o·· refle•ction from LA2 can be intt•rpreted as Lhe region

L.-\:.! 1s located in the transit ion from L\ 1 to LA:).

The ::-mall extent of the ~IS.-\ S arrny (Figure :3) leads to that the change of the d vahu• in the above equation

from 0 km to 400 km docs not makc significant changes in waveforms of -11 stacked traces fo r nlSAS l , 2 and

:3. Tht'rcfore we cannot de'krmine the rt'fiector depth fr0111 t hi' 2 f) map as is done for [. ,\ I, 2 and :3. ll owt>w•r

sl'\e-ral features can be find a~ follow~ .

figure 6 shows the stacked traces for ~ISASL Tlw () .. rdlo•ction preceding to PeP can f>,• found in the filtered

t race•;; \\ rlh pass-band!' of 0-0.:) li z and 0.:3-0.8 II z. It is worthy I o note that tlw polarity of I his PdP is re\·cr:-t>d

to that of PeP, which is found hy taking the corrt>lation COI'flic i••Jit between Pd P a nd PeP. Th is indicate that

the• I)" reflector beneath ~ I SASI has a negali \·c im p••danc•• co111 rast. The d imini,.;hing of PdP with the incn•a,.;o•

of fn•quo•ncy may h·· iudicat i\·1' of llw fin itt' 1 hickrw:-.:; of tlw !) .. r••llo•t:tor I hilt canst's this ano1nalous PdP. Tlw

d··pth of tho• rP!kctor ist•-.tilllato•d tol~t• at :}0-IJO k111 ahm•·tlro ('\ Ill infoorrcd fro111 tll<' diffo•ro'ntialtra\o•ltilll•'

of Pd 1'- P.

'J lw :-ic,nals of I'd !' aro· de•<~r in tho• s tarkcd t ran·s fro111 .\I S,\ S:! (110t pn•se·nlt'd lll'rt'). Tl1o' rt'(:!,IOII \I S.\ S:!

O\·e·rlnps with IIH! n•010 11 L \ I (1-'i)!,llre• 1). wlwre• tho• ]) .. r<'lle•rlur ;rt :!Ill k111 aiHWI' t l11• ('~Ill is <'sl il ll'lll'd. ' I h··

tr<~\e • l tinH·;o; of PdP from tit•· n•;:,ion \IS,\ S:? <trt' con,.,isro·ut with tho·() .. :-trurtun· at :!TO kn1 <~I>U\1' the C\IB

·)

r

:->hown in Ft:;ur•· 7 •n·· th t• s~nth••tic:-> calcuhtt>d J.y tlw rdl••rtivity nwtlwd an<ll•y th••l \C:..,f'•ll tnod.•l fl~•· nn••tt.

1991]. \\",,find cl"tr SdC:.., wnh larg~ amplitudes. which is not :->•'t•n in '-YIItlwtics. ,\lso found an• ,..i!!,nificalll tim••

dt'lay,.. of sec.;.

The Jiffen•ntial trawl times of SdS and ScS with respect to tht· direct S w<W«' arc measured by using the

cros:::-correlation. Tlw polarities ofSdS show almost the sante as t.hat ofS, which indtcatf' tlH' positive imped;tnce

contrast at the f)" reflector. The residuals of ScS-S, (1s<S - Ts )ob• - ('J~c;,s- T8 )c.,1 . arc plot ted in Figure 8fl.

The tlworcticallra\'t~ l tirMs ofScS and S a re calculated by using the 1,\ SPUL. The residuals show the po::itivc

values in the ran~c of 0.0:)-7.52 sec, which are explained by tlw S-v«•locity n•duct.ion of a few percents (0-2<7c or

more) in o··. Thc 0" thickness is estimated from the differential tnm:·l times of SdS-S (Figure 8b). assuming

that the S-vdocity :->t ruct nrc is the same as that of I ASP9l in I he depth rang•' fro111 t lw t•arth 's surface to the o··

reflector. Tlw estimated depths of the D" reflector range from 70 km to 400 km above the C:\113. The estimation

of depths of tlw I)'' reflectors depends on the S-\·elocity model. If we tncn·a~t> (dt>cn•ast>) tlw S velocity by :3Y(

with respect to I \SP!)I in the lowermost mantle. the e;-timatcd depth of lit•· o·· n•flector deepens (shallow:-)

by about 50 knt. \\'e arc confronted with the same problem of model dept'IHlencc wlwn tran:::lating the positive

ScS-S residuals into the \·alue~ of the S-n'locity reduction in t lw lowcrrno:;t mantI e. Therefore the quanti tat i\·e

estimat ions may involve <t great uncertainty and we discuss in lalt>r section t.ht• qualitati\·e fraturcs of D"; the

positive cont rast at. Ill!' rdlt•ctor and the S-veloc ity n•ducl ion.

N ume rical exp eriments

examitJt•d hy 11'irt:.:, llt•' syrttlwtir "''1'-lllO!,!,rilllt.; calntlott••d hy tlw r•·flt>ctivity lltt'lhnd . ' I Itt -.:lltlwtirs for rhis '''"t

art• rakulat.-d for tilt' t'\"t•rtt-ro•rt>in·r ;;•'Oillt'tri.·s or tilt' array L.\:1 Th·· C(Jil\"<lllltinn .,r tlw irt,-1 rtllllt'rttal rt""l'""''"

of tlw :>t•i,.tliO!!,r;tplts of this array (-.ltort-p••riod \·,·rtical cotnpottt•ut , fn·•·-p••nod of 1.0 ,..,.c aut! datnpiu;; fartor

ofO.I) i~ appli··d to tlw syntlwtir .... Tho• dotllillalll fn•qut•nri··,; or th·· synth<'lir,- rauy,·· frOIII 0. 1 liz 10 1.0 liz .

li

1\lodl•l,.. of vc•locity cltau gl'

Tho• P w locity modo·(~ ltl r:1lrnlat•• tit~· ,.:yn thdic,; are gronpo•d into fl\·e typo•,.:_ Tlwy an• ( I) a lanwlla with

hig,h P wlocity. (:n a lamella wtth low P \·elocity. (:q a transitional increa~P in P wlocity. (I) a tran,-itional

decrt><\:;c in P velocity and (5) a stack of lanwllae. \\"t•ber [Hl~J I] presPnts thl' n~,.;ults of n pro•liminary analy,.ts of

travd time and slowne,.;s for the models of high- and low-\docity lamella. For the ca:-;e, of (I )-( •1). thre'' ,-altJt·;:-;

of 10, :wand 50 km for the thickness of the lamella or tht' tran,.;ition zone art' tested. l"or li te stack of lamellae.

we usf' three values of 10, :30 and 50 km as the intervals that St'p<Hnte each lamell a of :30 km thick with high P

velocity. In all the models the cont ra:<t of the P velocity is set to he :3<;( and the depth of the top surfact> of tlw

vdocity change is located at 2600 km. \\"e ha,·e already estimated the impedance contra,.:t and the depth of tilt'

reflector in our recent paper [Yamada and :'\akanishi. l996h]. However. the change' of either contrasL or d••pth

affects the amplitude or the tr<IH'I time of PdP, respectivdy, and does not alter the w;wt'form itself. The ,-alue,;

of :3% a nd 2G05 km arc taken from P\\'DJ\ model in Weber and Davis [1990] as an example. The P-veloci ty

model of P\\"01\. itself is also u::;Pd as an example for the I)"" structure• with a discontinuous (abrupt) incn'ast' in

the P velocity. The IASP91 includ~?:-. :o;rwral discontinuitil's, ~loho. upper mantle discontinuities at the depth~

of 210. •110 and 660 km and has the change in the \"elocity gradit•nt in the lowermo:-.t tllatllh; (2140 km depth).

The secondary wa\"eS caused hy t l~t• reflt•rt ion and the conn.'r»ion at t hcse discont inuitit•s may rontami nate the

time window followed by the direct P ,,.a,·c, where Pd I' is expected Lo arrive. The sy nt hctic calculation and

dat.a analysis by using IASP9 1 model arc made to exatniut• this possibility.

0 In all the model:<. the \":till•'"' of hoth density and quality f;Jrtor ;JI"t' takt>n from PBI:\1 [Dzic•11·on:<ki and

,\ ndcr:-.on, I \131]. Tlw S- \'l'loctl ~ and do•tbi t y pro Ilks i 11 t lw lo>W<'rtlln"l ma ntl<> arc· p••rt 11 rh··d in t Itt' sa nit' tnann••r

t ht• folluw i ug.

(p<•lanr~) II"' rhe dirt•c r I' WCI\•' (to t•) \\·,.lind rh:\1 au iHP'l t'rrcl··d h~ rip• hi~h··"'r ··r•nl<•llr lrll t' l~ •r I'd I' t 'XI• 'IId-.

in rlrt• \t' rlicnl (d••prh ) rill!!!,•' of .-.hout <-0 km. \\·,.abo find rlu• deviarion of tlw ,·,..tinmr.•d , ·alu•• of :!IU km

(indicat••d h~ a hrok••fl lirlt' in seCOIHI par!t•l) from tilt' •'Xact Out• of:!'-- I km (:!fj(J:j krn d··pth from th•? o•anh·,..

surfan•) . The llatno•,..,; of the high PdP energy and tiH' dt>viation from tho• corrt•ct dPpth of riH' D .. rt>llcctor

suggest that the resolution of the determination of the D" reAl'ctor depth is a ft•w lens kilometer,;.

Lam ella wit h hig h velocity. Among three results from thret• high-velocity lamell a models whose differences

arc in t lw thickrwss of the lamella, Figure 9b shows the result frorn t.he lamella with the t.hickne:::s of :30 km.

The wawform of PdP (th ird) is distorted wi t.h respect lo t.hat of thr direct P wave (top), due to the interference

of rrllection,.; from the top surface (normal polarity) and the bottom surface (n.•vNsl'd polarity) of the lamella.

Tht• second upward :;wing in PdP waveform corresponds to bottom-side reflt>ction, which is not seen in PdP

from the P\ \' D I~ modt>l (Figure 9a). Comparing three n•sults oht airwd from thn.•e thick nes..;es of the lamella. it

is found that the increase of the thickness increases the amplit udc of PdP. This is because a long wavelength

wan• is not scnsrti\C to thin lamella. The amplitudP. of PdP for tht• :30 km-thick model is comparable to that

for tht> P\\'DK model. The thicknt>:;s of 30 km is an approximat" \\·;wekngth of '2 ,.;t>c P \\·a,·e in the lowermo"t

manti••.

Lamclln with low velocity. TIH' interference of two n•flo•ctions from both surfaces of lamella is also found

for t.his model. Figure 9c shows the result for Lhe low-veloc ity lanwlla wit h t.hP thickness of :30 km. In contrast

wi th l lw previous high veloc i Ly lanwlla. the \\'a\'d rai n of Pd P consists of t lu' fi rst-nrri ,·al rPversed reflect ion (top

surface) and so•condnr) arrival nor111al reflection (bottom surface). Tl 11~ st•pamt ion bt'I\\'Pen two reflect ions is

olH'ious for tho• lanwlla \\ith thirkno•:<,.; !!,rt'ater than 10 knt. Two lw;•ks of PdP an• ,.;t't>n in tlw contour nwp of

l·'ignn· !Jr.

Tlw rl'llt-rr ion frt>lll Ill>' lop ,.;nrf;Ko' rnay 1 .. ~ qualit :11 iVt·ly "'nail. ,..i11ro• I h·· :-••i:-rni•· PTit'r~~ I har arriVt•,. at I It•:

hountl:~r~ wath llf'.!!,ilti\·o· P-\'o•locity ronrra.;t tran:<mirs downwanlna<~n· ,.tron;;ly than rhat ar rlw boundary wirh

jllbiriw 1'-\•'llll'ity rourra"t \lort'll\'o•r rlw incid··nt :11agl•· of tlw Wa\o' :tt rlw hortoan ,.,mfar•' of rlw lanwlla i"

ri\tlwr :-1•'•'1' in lo\\ \t•luciry lant• •lla :1nd thi,.; abo l··ads to tlw .. na:tllo•r rdl•·•·tion rn<'lli··i··alt at tlw hortont :-nrhr•·.

'l'ho•rd(,n· tlw :11nplitnol•·s of tlto •,..,• I\\'<) rdJ,·rtion,. ar•· :<llt:tlln tlt;lll tltal f<>r lti;.;lt-\t ·l .. rrt~ l:un.·II:L

incn•a,;t• in p VPiocir: llllh IIJt• rhickrw,; .... or :3o kill. In lhi,; '""''·tilt' CIIIIJ>Iitudt• and lilt' 11<1\t•forru of Pdl' illt'

Hlrno,;t idt•nticalto that for tilt' J>\\"DI( model. 'l'lw ro1npari,;on of three rt',;ult,; by changin~ tilt' thickness frn111

lU km to .)0 km ,.how" that th•· n111plirude of PdP dPnt!<ht> with the incren"" of tlw tran,..ition zone thickne .... s.

The amplitude decrea...;;t•s to a half with the increase of the thickness from lO krn to .)0 krn. Thi::; is becatb<' til•'

high-frequency content of till' incident 11·ave become~ in:o:ensitive to the transition I.OIIC as it b~>comes widt•. \\'t•

also find a cont.inuous trrnd of delay of the tra\'1'1 Lime of PdP, correlating with thr increase of the thickne,;s.

However the time delay of PdP accompanied with the changt' of the thickness frorn 10 km to 50 km is "rnall.

about 1 sec.

Transitional decrease in P velocity. The result for the model of tlw transit tOnal decrea..o;r in P velocity. with

t ht• thick ness of 10 k m. is shown in Figure 9e. The signal of PJ P is characterized by (I) rat her small ampli t nd··

due to the small reflection coeffirienL, (2) reversed polarity with respect to that of the direct P wave and (:3)

rapid diminishing of amplitude with the increase of tlw thickness of the transition zone. The energy of PdP is

beyond the level of a cut-off for plotting the contour map for the model with thf' thickne,;s greater than :30 km.

Stack of lamellae . The model of the stack of lamellae ha:; a shape like 'teeth'. The contour maps for tlw~··

show complicated distribution of PdP energy. Tlw rt•flt•ctions at both surfaces of f'ach high-velocity lamella and

re\'erbcrations inside each lam(•lla and between lamt.>llae l'or111 continuous arri\'als of sei:;mic enrrgy (PdP-coda).

The difference~ in 1 he w;wel'orm of PdP wavetmin, cau~Pd by the change of iniN\'als from 10 km to 50 km,

are as follows. (1) 10 km interval (Figure 9f). Tlw first arrival is isolated and its ~hapl' is almost the ,;aJIIt'

a,; that. for tlw P\\'1)1\ n1od .. J. 'J'Iw PdP-coda rs nol l<>ll!!, in tint•' aud not IM<;~' i11 arnplitutl.-. Tlw thin inl•·n·ab

pt•rtlll'hation and til•• r•'"IHlll"" of 1l1i,. ,;lruclllrt' to tlw inritl••nl 11<11'<' i,; \IO:<•' to tilt' 1'\\'J)J, lltodPI. Tlw 1ahlt'

or 10 kill COITt'spond~ to a w:m·l··n~IIJ of l .:l ll ;,. (2) :3() ktu illtt•rval (Figun· Dg). 'J'Iw \l'ill't'lrain ha:; a lon:.?,

tail and I ht• fir;:;t arri\·al ha,; a l:tr~" :1111pli1ud,• ro111p:H•'d with 1 hat of I ht• pn•vio11, Ollt' . l11 llw rout011r 111ap

,;t'\'t'ral pt•aks follow 1 ht• lir,.t arri\'al. Tilt' con1pli.-a1t'd sy:<t<'Jil of rdlt·rt ion:' of I hi:-: ... 1 rnct un· 111akr'" "" dilfkulr

!1

t Ia i rd pant•l:::. \\"hi le t !at• wawform from ••a cit ·tooth· i~ almo:-t id<'lll ic a I I o that oht ai nt>d frorn till' mod••l of ont•

hi;;h-wlocity lamella in Figure !lh. th•• amplitude i~ rt•duced in ord••r of the dt"pth wlwn• t•ach lamdla is loc<Ht•d .

The crwrgy from the deepest 'tooth' cannot he found and may b.• mixed with that of PeP.

IASP91. The result for IASP91 is shown in Figure 9i. The largP amplitude of a...;sembly of the direct P ,,.a,·e

is ~hown on the left in contour nwp. Since the double-army stacking is to enhance Lhe signal from the reflector

with common dept.h, the direct P wave, whose turning depths are diffNeiH each other. cannot lllake the peak

in the contour map. The right-side ent>rgy correspond;; lo PeP. There is no signal in the time winJow between

P and PeP. This shows that any srgnals. for example, the reflection and conversion from the upper mantle

di:-;cont inuities, cannot be enhanced hy this method. Thert>fon• we can regard the signals hl•t ween P and Pc P

in the observed traces as l he reflections from the o·· reflect or:->.

Discussion

figure 10 shows three pairs of the direct P and PdP waveforms ohsct\·ed from three rl'gions Lr\1, 2 and :3.

where the PdP traces are t ht• ~anll' as in Figure 5. The wawoform:-; of PdP are clear for LA 1 and :3 but not for the

LA2. These wa\·eforms are comJHHt'd with the synthetics obtairu•d in the above section. The .-igniAcant fer~ture

of th<' observed PdP waveforms i$ the upll'ard sll' ing of the :second peak, ll'hich is seen in syn~hetics for tlw

high-v1•lority lamella (Figure 9b), hut not for the simple di:-;conl inuous model ( P\\'01\:, figure 9a) . llowever.

tht> f'stimation of tlrf' thicktw~s of lanwlla hy u.-ing the• arnplitnd•• of PdP i;; not ciPI<'rtllin••d uniqtwly. becau~f'

th•• a111plitndc depl'nd:s on both tlw rhrrkrw,..,.. ofbnwlla nnd ~·ontrast of I' \'<'locity \\'hrl·· the :stack oflanwlla··

with inl•·n·;ds lar;!;t'r than :10 krr1 prodnc·,.,.. thr• ~:111h' :-••rond ""'Ill!!,. tilt' Jar!!,•' and long t:til of PdP coda, app:ll'•'lll

in th•• synthetic arc 110t st't'll ill th•• o l l~t'!'\'<ttions: rht•ir ll':l\'t•lt•ls <tl'<' rallwr i~ol;dt•d. Th•• IIHHic•ls with rwgati\c'

contrast at the rdl<'dor. irtdudint; rran,.:itiortal d•·rn·:r~··· carlltOI n;plairt tlr•' Pd l' wa\·t'fonn .

This W;l\·t•forrn ll'ith tilt' :O.t'('OIHI ~Will!!, is the ('Oil~··qnc'ltft' or tilt' illkrf··rt'll('t' of lilt' normal and rt'\'t•r,..···l

l">larizc•d \\':1\'<'l•·ts 1 h:tt rorn·spt'lld tn til•' Inp- :uHI IH>I torn ·:-llrfa···· p•ll••t'll lltS of I he• lantr•lla lloWt'\'•'r. Briilrl •'I

10

I r I ilL· ro•llt rl or Ita" a fi IIi I,. ,..i /.t' iII ill•r i%1011 I al d i n>c I i· >11. Til.·~ ,;hr '"" Ill tlwi r t'X I ..... i 111•'111,.. I hilt tin• r..tlo•rt i· •II \\'it \(·I· I

with ro'\•'r:-;o•tl pohrit: "Jll"'ar:- followi11g tltt• Jtortnal polarizt•d rl'll,•nioll du<' to tlw <'\l"ll'ltrP of tilt' ··,·d~,:· <)f

Tho• tinw lag betwt't'n tht•st> two n•flt•ctions and tlu' shap<' of the combined ,,·awfonn..; d"P''IHI on tht> sit<' of

the reflector and on the frequency content::. of the incidt:-nt wa,·c. The time lag of T/'2 (T tlw period of incident

wave) may lead to the large amplitude of the combined reflecttons for the finite-size rdlrctor, and this time lag

is tlw definition of the first l·'resnd zone. As shown in Figure •I, the major axes of regions Lr\ I and LA3 ha\'C

t.he sizt•s almost comparable with the s ize of the Fresnel ellipse for P wa\'e. Tf there are two reflectors beneath

the regions LAl and Lr\3 with different depths (210 km and ITO km). the wa\'eforms of the obser\'ed PdP can

be also explained as the effect of the finite-size reflector, without the lamella structure.

'1 he increase of the seismic vt>locity at. the discontinuity is also suggested by the normal polarities of the SdS

waveforms. The second swing, which is an e\'idencr for the lamella but against the one-sided discontinuity

(e.g., P\\'01(-like model) , is found in scvl'ral traces, e.g., the observations on the left sidt• in Figure 7. We haw

confirnwd by using the S synthetics (not presented here) t!Htt the lamella produces the second swing for long­

period S wa,·e, but that the am pi it ude of second swing is ra 1 h<'r weak. Therefore the fact of small amplitude of

the SdS second swing in Figurl' 7 cannot preclude the lamella structure for Swan>.

E:i titer the high-velocity I amelia or the d isconl in u i ty with fi 11 i le-size is possible and the combination of the

two is also <1cceptable for both P nnd S wa\·C'. \\.hile W<' cannot rcsoi\'C' this non-uniqutu•:-:s, tltt• po;;it.i\'e contrast

at the lop of the reflector lll<IY lw inft•rred from the data analy:-;t>d lwrc.

OtH' possibility of tlw hic;h-wlocity structure'::: trtll' chara•·to•r 111i~ht lw tlw suhdnrlt•d on•anic crust. wlndt

is proposo•d hy \\'o>ht~r [I ~l!l · l]. It i,.. ttol rlo•ar wlwl hl'r tlw suht!ttflt•d ... ]ah ;trri\'o•s at tilt' lol\'t'I'IIH):-;1 maul I<' ;111d

wlwtho•r thr• ,..Jt;lJH' of it IS tnaintai••··d throng,h the pt'lll'lr;ttiou proro•:-;;-, or it ,..lag11atll>' <tl (j(j() k111 disro11tinnit;.

and !'ails i11to tit,• lowNtno:-;1 tllillttlr• as "a,·;llanrlll's .. [L<ty. l!l!l I and l.op•·r ilttd l,ay. I!Hl;>]. It ro11ld lw a plau,.;ihlr·

int••rpro•tation that tltt.' sltlodnrlt'd slat. lllillt'rial o'XIo•nd..; horizontal!;. with tilt' rOII\t'l'fi\o' 111otio11 in tlw low••r

111<1111lt• ;111d fort II" l he hi!!,h-wlorit: l:ntwlla

II

for tlw Pxistt•nce of the low vdocity layl'r in the lowt•rrnost IIHIIltk henf'ilth the ,;ontllll'•'siPI"Il l'ncific [.\Iori a11d

lldmh .. rgPr, 199:) and C:anwro and lfdmlwrger. l9%J. Tlwy ill"o suggest thnt it migiH ht• not a global hut a

regional feature. The inclusion:; into the lowermost mantle by the chemical reaction may be of great contf'nt,. of

iron compared with that of mantle [.Jeanloz. 1990 and It o ct al., I995j. The great change of the geotherm at tlw

C'i\If3 may lead to the exist.enct• of tht:>rmal boundary layt>r with a finite thickness [LopN, l991 and Gaherty and

lay, 1992J. It could be possibll' that the accumulation of the products by the chemical reaction, whose sci:>mic

velocity may be lower than that of ambient mantle due to the iron-rich products. acts a role as the thermal

boundary layer and can he setsmologically found as a low veloctty layer.

The two-layered structure inferred from the above description is schematically illustratt•d tn Figure J I. \\'l•

indicate in this sketch the values of the P-velocity increase within the high-velocity layer (or lamella), which arl'

inferred from the preliminary estimates of the impedance contrast obtained in Yamada nnd Nakanishi [19D6bj.

\\'t' assume when inferring the P-velocity increase that both the density and the P vdocity increase at the sanw

rate. The stack of the low and the> lugh velocity structure, postulated here, changes its total thickness from 210

km (LA I) to l 10 km ( Lr\:3). Assuming that the thickness of the high velocity layer is unchanged throughout

the region sampled here, the low \'t•locity layer thickens in the region LAl. Lt. is one plausible, not sulllcicntly

confident., speculation that. this thickened low vdocity layer in LA! is the plume ht•ad, which is now proc,'s:-:ing

the upward motion as a "manllf1 plume" [.\Lori and lldmbcrgt'r, 1995). Unfortunately we can only search onl'

direction among the rt•!!,ion :-:urrountlin~ LA I at JHt>.·-.•nl. The furtlwr examination of I)" structure in the n'giOn

lwighhored on L:\ I \\'til JHO\ 1tl•· "" wtt h tltt: lllOrt' dt'<ll' t•\·id.·nn·s oft his donH' and with it,.: horizontal sizt'.

:\t tilt' t•nd ofthi,.: st•rtion, 1\'o• "hould lw\'t' aslalt'ttlt'nt ahoul the variations in SrS-S n•sidu;lis a11d the dq>th,;

of I he I)'' s rdlcctor , shown in Fi~tll't' ~. TIH'SI' lar~·· \ill'ial ions (ll't' ilnOtll<dons wl~t•n lakin_!!, tlw ··xt••nt or

lht• 1-'n•stwll'llipsl' (Fi~urt• ·I) i11lt> arcount and 1ni~-;ht l>t• iltl<'I'IHI'II'd as tlw I'Xistt•nn• of IIIOrt' sllorl-\\'i\\'t•l•·tq,;th

(pt•rhnps. l••ss than ont' hlllltln•d kilontt'lns) lw~t-rn;;•·n• • iti•·s . Th·· StiS :unplitudl' (Fi~ur·· I) i..; also :lllontaloth

1:.!

l'n•:-;tJt•l dlip-:,•. Such au auottwlou:< :-1 rttc!ur<! cau ttot h,• t'.\1 ract•·d fro tn tlw l'-wa\<' rP:<tdls prt•:-;t•tttt'd ahovt'. Tlw

itllpNiauct• contra;:;( of P wa\t' al tht• n•Ot>ctor hettt~nth L:\ l.!. and :3 may be up to tPn pt'I'Ct'llh. if \\'1' a:<:'Uill<'

hori?.011tal rt>flector. Tht• d••tt•ction of the clear PdP :;ignab (t'.\CPpt for LA:?) indicat••;-; that the wa\dength

of heterogeneities may be a fpw hundred kilometers. This difl'crences of scale lt>ngth of heterogeneities or tht'

impedance contrast betwePn P and S waves might sugge,.;t the anomalous D" layer, where the rcspon;;e to tht>

incidence is different betwet•n p and S waves.

Conclusion

The analyses of travel tim•' and w;weform are 11pplied to th•• data of both short-pNiod P wa\·e and long-p••riod

S wave. \\'e can suggest the following image for the ()" layer hcneat h the southwestern Pacific sampled here.

Comparing the observation;; with the numerical expcrinwnts, the high-\·elocity lamella is the preferred modt•l

rather than the first-order discontinuity to exp lai n the wavcforn1 of PdP. However the discontinuity model cou ld

st ill b!' a possible candidate if wP take the finite size of the 1) .. n•flector into account. The low-velocity layer.

coated with this high-velocity layer, is inferred from the large delay of travel times of ScS. The total thickneN'

of the:se two layers changes from :270 km to 110 km. Thi,.; he1crogcneous structure might be interpreted <t." the

low-velocity dome (a seed of mantle plume) underlying the high-q~locity suhductcd nHit••rial.

Acknowledgments. Throughout this work I would lik•• to thank I. Naknnishi fo r 111nny fruitfu l suggestions

a.nd comnwnts. I al,;o thank 1\ . Oikc for his supports. Tltt• rdl••cli\·ity code for synthetic c:dculations was

providt'd by courtt>;:;y of .\1. \\'••lwr

R e ferences

Briiltl , l\1. , (;. \'t'riiH'<'r. and ~ 1 . 1\i t'l lll , Fn':<ll<'l wn•·s fo r ln·uadi>and data. (,'u,,d,y.,u·~. fil, (iUO-GU I. l!l!H) .

l)zit•wonski, ,\. ~1. . and D. L. ,\titl l' r,.;on, f>r,•litttina ry rd~·r··nr<' t'Mi h lttodd . l'hlf' · l~'urlh !'laud. 1111< r .. .!.).

:l!ll :l;'iti. I !l'\ I.

of t.lw nwntle lwtH' <llh the central-Pacific. C:wphy~. Nc.>. Let .. ..!.J. ~J77-9~0. 1!:!%.

Haddon, 1\. A. \\· .. and .J. H. Cl••ary, Evidence for ~cattcring of seisntic PJ\P 11':1\'c>s Mar the corc-mantlt'

boundary. Phys. Ear·th Planet. Inter .. 8. 2 11 -:2:3-l. 1974 .

Hofmann, A. \\'., and \\'. 1\l. \\'hite, i\Iantle plumes from ancient oceanic crust, Ear·th Plauet :)ci. Letl., .57.

421-436, 1982.

Inoue, H., Y. Fukao, K. Tanabe, and Y. Ogata, Whole mantle P-wave travel time tomography, Phys. Earth

Planet. Inter. 59, 294-328, 1990.

Ito, E., I\. ~forooka, 0. Uj ike, and T. Katsura, Reactions between molten iron and s ilicat.e melts at. high

pressure: Impli cations for the chemical evolution of earth's core, J. Geophys. Res., 100, 5901-5910, 1995.

J ean loz, R., The nature of the Earth's core, Annu. Re11. Earth Planet. Sci., 18, :357-386, 1990.

Kendall , .J.-1\L, and P. t\1. Shearer, On the structu re of the lowermost mantle beneath the south west PaciRc,

southeast Asia and Australasia, Phys. Earth Planet. Inter., 9:2, 85-98, 1995.

Kennett, B. L. N., eel., 1ASPE1 1991 Seismological Tables, Pub/. Australian J\.af. Univ., 1991.

Kruger, F., 1\L \\'eber, F. Scherbaum, and J. Schlittenhardt, Ev idence for normal and inhomogeneous lowermost

mantle and core-mantle boundary sl ru ctu re under the Arctic and northern Canada, Geophys. J. !Ill., 1 :'!2,

637-657, 1995.

Lay, T., The fate of descending slabs, Aunu. Rev. Earth Planet. Sci .. '!!2, :3:3-()1, 199·1.

0. Loper, 0 . E., 1\lantle plumes, Tecto nophysics, 187, 373-384, 1991.

Loper, 0 . E., and T . Lay, The corc-nmnt lc bonndary region, ./. C:cophJ;s. Rc.~ .. 100, 6:391-6-120. 199'>.

!\Iori, .J ., and D. V. ll elmhP rgl'l'. Locali;r,Pd hout1dary layer lwlow 1111' tnid-Pacili<: \'l'IO\ity anot tlaly idc•nt iliPd

from a Pel) prc\ ur,;or, .J. C:cot'h!J'· Rc.L 100, :w:3:>U-10:H5!), J!J\J.").

Ri chard,;, :0.1. ,\ .. and 1). c.:. Engd)rt•t,;otl. Larg<•-:o;call' lll illtllt• ('()11\'t'Ctiou ilt ld tIll' hi,tory or :::uhducl iou, .\'oll/1'(,

Sengupta,.\!. 1\., and \1. .~. Tobo;r,. Tltrt'r ditll•'ntionaltnod<'l of:wi,.; tn ic vPiority v:H iatiott iullw l'<trth',.; tnan1lc.

(:wtlir.rt~- /(, '· /,({{ .. .J. ~- 1 -0(). l!l7(i.

1-1

<., lilut;llli. r. :\. Tauakl \I ,,;,to. and 1.-: lliraharil. \ -.luoly or 1'-1111\1' \O'IOf'IIY di--CC>IIIIIIIIIIY Ill[) .. lay. •r ll'llh

.J-;uray Ht'rord"': Pn·liuuna r~ IT"'ult"' . ./. C:comag. (:wcl<clt ... f.), l:!i:i -1:!~:). Hl!l:t

Su. \\' .- .1 ., B. L \\' :lodward. and:\. \1. DziPII'On::;k i. Dq;n't' 12 modt•l of ,..h,·ar wlocity lwtProg••neity in th••

mantlt> . ./ C:rophiJ~. Rc~ .. 99. 6915-69~0. Hl9-L

Tack ley, P .. J.. D. J. Stevenson, G. G latzmaier, and (;. Schubert, Efrrcts of an en dol hermic phase l ransil ion at

670 km dt•pth on spherical mantle convection, .\'afure, .%1, 6~)9-70• 1 , H)9:3.

Tono, Y. , and f{ . Yomogida, Origin of short-period P-diiTracLcd ll'aves: :\ ca::;t• study oft he 1994 Bolivian deep

earthquakc>, submttlcd to Phys. Ear/It Planet. Inter., 1996.

Vi dale, .J. E., and II. ~I. Benz, Sei:::mological mapping of flnc structure near the base of the earth's mantle .

. Vature, .361, 529-5:32, 1993.

\ \'cbcr, ~I., and .J. P. Davis. Evidence of a laterally variable lowt>r mantle sl ructure from P- and S-wa\·es .

Geophys. J. Int .. 10:!, 231-255, 1990.

\\'eber, ;\I., and .\1. 1\ornig, r\ search for anomalies in the lowermost mantle using seismic bulletins, Phys. Earth

Planet. Inter., 7.], 1-28, Hl92.

\\'eber, ;\!. , Lamc>llae in D'"? An alternative model for lower mantle anomalies, Gcophys. Res. Letl .. 21, 25:31-

253·1' 199·1.

Yamada, A., and I. Nakanishi, Det~clion of P-11·aw• rcflecLor in I)'' beneath t.hc south-western Pacific using

doub le-arr;ty stacking, Gcopltys. Res. Lei/., 2.3, 155:3-1556, l9!)6a.

Yamada, A., and I. Nakanishi, Short-wavelength variation of P-w;wc rcOcctor in 0" beneath the south wcstcru

Pacific. 111 prrparatton. 19!l6h.

Yamada,,\ . ,\ . 1\ohayashi. and G .. \oki. Dt'lt'ctiou oftht• 1)"-r.-Jl.,(·tor h~ usiu!!, a small apPrture array. 'J'rr/,,r,/

Htporf, of .1/tlf,u,ltim Su,moloytntl QIJ,trraloriJ. J.\1.1. I I.,, lut"'"' ''· J . ~. l!l!lCi .

,.-,

Fi~un· c;1pr ron-.

Figure 1. Hayp;~lh" <)f ,,.n,·e.' u,;t•d h••rt•. :\ squar•' ;-;how;-; tlw hypocelll••r al 1lw d•·pth of .JUU kill a11d 1h••

recei1·Pr is indin1t••d hy <1 crrcl<': the epicl'ntral di;-;tance is 10° 111 this <'Xnrupl••.

Figure 2. The g~'ographical distribution of C\'ents (circle::.) and rt'CI'I\·••rs (,-quare:-} utilizNI here. Al:::o shown

are great circle paths connec ting event to rccei 1·er. The sampling rt•gion is lwnl'al h the southwestern Pacific

(rectanguln.r n•gion). The close-up of this region ca n be found in Figure I.

F igure 3. Rt'ct•ivcr distribution in .Japan. Large arrays with it.s ext.t'nts of several hundred kilometer;: arl'

indicated by blue syrnbols; Kyoto Uni\'ersit.y (LA l , circles), the Univt•rsily o l' Tokyo (LA2, triangles) and

llokkaiclo UnivNsity ( Lr\:3, squares). A green circle shows the location of a small array operated by .J ap;111

;\lrteorological Agency {:llatsushiro Seismic Array System. "lSAS). The panel on top lt>ft shows the close-up of

the 1\!S,\S. !led circlt~s indicate the distribution of the recei,·ers of till' broad-band seismographs (BB).

F ig ure 4. The bounce point. region sampled here (rectangular region in F'igurc 2). Colored ellipses show the

distribution of PeP bounce points, corresponding to arrays in FigurP :~. 1 hey are LA I (green). L:\2 (blue), LA:3

(gray), .\ISASl (yellow), ~ISAS:l (orange) and ~ISAS:3 (purple), respecttvely. Solid circles indicate ScS bounce

poi nts from broad-band records ( 1313 ). Two open ellipses on t ht' bottom lt~ft show the first Fresnel ellipses for P

wave (0.5 liz) and S wave (0. 1 liz) at the C~ l B. Here we plot a half size of llrt• first Fresnel Zone according to

Weber and Davis ( 1 990].

Figure 5. T lw 2 I) colored present.at ion of resu lt;; by the douhlt• array slacking (tiH' same figure as Figure :3 in

Yamada and i'/aknni~hi [ 19Hijl>]). il) result for n•gion [,,\I , h) L,\2 and c) Li\:3 . In <'ilch panPI. a signnl. t.hat is

inll'rprt'l• ·d as a candida l• ' of PdP. is indicated hy an arrow. ,\ t ran• on ••arlt nrap show,. <1 stncked tract'. who""

cnwlupe is a <'to,..s·s••c lioll of •'ach nrap al rh•· dt·plh of lh•• nraxi11111111 arnpliludt• of PdP. PdP is indicntcd h,1 a

dot in,•adr trctt'<' .

F ig Ht'(' G. llw :-lacked rrac••;-; from rl'.:;ion .\IS:\Sl. wi1lr th.: appliralion of;-;••wral band-pass fjJt,•rs (pa . ..;s-l•and

i:, indicalt•d on lop ri:;ht in t•aclr lral't'). Solid cirdt•s show rlw anotnalous l'dl', whost• polaritic•,; an' opposil•· 10

1 hosl' of J>r J>

low-pa:<,; fillt'n'd with tilf' Cllt-olr frt'ljllf'II('Y or U.l II~. Tlw ronliiiiiOll:' lirw j,; tlw oh,;t•nariou and th·· hrok··n

line the synthetics calculated by using 1:\SI'\!1. The epicenlral di,.,laure i, shown in hot tom l<'rt in ea ch tracl'.

Figure 8. a) \'ariat ion in I he residual,- of ScS-S. The ,-ize of circle i::- proportional to t lw a mount oft he rP;:;idnal.

Colored ellipses sho11· the bounce regions of P waves as the same in Figure -1. b) \'ariation in the thickness

of D'', that is obtained from the differential tra\·el time of SdS-S. Here we assume that the depth of the D"

renector is t.he depth of the top surface of the D'' layer. The size of circle is proportional to the thickness of the

D" layer. P-wave bounce points are also shown.

Figure 9. In all figures from a) to i), (top panel) the stacked synthetic trace of the direct P wave, (second)

2-D contour map wit.h the contour intervals of 1 dB and a signal of PdP is indicated, (third) the st.acked tract'

of PdP, whose envelope is the cross-section at the depth indicated by broken line in second panel and (bottom)

the P-velocity structu re used to the calculation of the synthetics.

a) result of PWDK, b) high-velocity lamella with the thickness of :30 km , c) low-velocity lamella with the

thickness of 30 km, d) transition with increase in P velocity by 30 km, e) transition with decrease in P velocity

by 10 km, f) st.ack of lamellae with intervals, that separate each lamella, of 10 km, g) stack of lamel lae with

intervals of 30 km and h) stack of lamellae with intervals of 50 km. i) result for IASP9l. Because of the lack of

any s ignal between P and PeP, no stacked trace (correspond ing to the third panel in figures a-h) is presented.

Figure 10. Comparison of the waveforms beLwePn the direct P wave and PdP. a) stacked traces of both wa\·es

from region Lr\1, b) Lr\2 and c) Lr\:3. The signals of PdP arc indicated by horizontal continuous bar::-, except

for LA:2 (broken bar), which shows low signn l-to-noise rat.io compared to PJPs for Lr\1 and LA:3.

Figure 11. Schematic sketch of the I)'' I<Lycr which is obta ined in th is st.udy. The total t.hickrwss of I)"

\·aries from 210 kIll (LA I ) to I 70 kIll ( Lr\ :3). r\ t. the lop of D", t.IH' rn pid i llcrt'il:>•' of J> <~nd S vdoci t i,,,_ i:-:

inferred. Tht' contra:-:!:-: of tire P n·locit.y aero:-:,; t lwst' rellt•ctor <Hl' ;rl:;o indicated. wltirlt are inl'••rred fro111

the prelirninary c:-;tirnill ion,: in Y;ullada anti i\akanislri [IU!l()bj. lnft•rn·d frOIII llw rornparison of oh:-:<'1'\'('d arHI

:>ynthct ic Sl'ismogranr:-:, lllt' hic;h-n'locity larlll'ilit ronld be one intt'l')ll' l'lation for t.his higlt-\·Piority byt'r. Bt•rwatll

this layl'l', tiH'I'l' may lw a low S-velority lii_\'c'r, while tlli,.; i:; indiralt'cl lwnt·at.lr 1, ,\ I :11HI L:\2 ;1nd i:-: nnknol\'11

Table 1. Event. paramdl'l'=' wwd in lhi" s tudy. Tlit' nol;llion,.; o f the columu ·· [)ala Se·t'· an' tlw ,.;anw <1:' in

Figure:~ \otc that \ISAS is dividf'd into three rq;ion" (\ ISr\SI . 2 and :3) accord1u~ to tlw locnt10ns of bounc•·

points.

Table 2. The range of epiccnlral distance and the numhcr of traces for each data set lh<'d here. The notation,;

o f data ::;et. a re the same as in Figure :3 and Table I.

Tahl1· 1.

Orit;tll ltnu•( (;\(1) L<11. Lou . D· ·l •l" ()·,ta ~ •• , .... ~I f) II ~I s (0) (0) (kIll)

lll b

i!T'J .\lnr. 1·1 II :31) :30.9 :?O.O&S 118 01\\" ~,;o -:[~T\:3 .. \J<.. ,\S:!-J!l ,'i I .lui o:~ 1 :~ ·12 00.1:':3 11.1:r>S 110::.$-11\\' ::dG :; ,; ~IS.-\S1 191:' I :\uc;. ::!6 5 00 15.6 2:3.59S 119.01[ :)GO :)H L .. \3 HJSI si'P 28 :3 :3 ·16.8 21.51S 111.79\\' :H) I 58 L,\:U1SA c;:z 198;) ~Jar. 15 00 16 02.15 20.6825 118.218\\' 5•15 5.1 ~IS AS::! 191:);') Aug. 28 ::!0 50 48.:3 21.01S 118.98\\' 625 6.1 LA:UISAS2 1986 :\ pr. 01 10 13 40.14 18.03/S 118.5:31\\' !>10 5.8 .\ ISASI 1987 Fe h. 10 00 59 28.51 19.4895 111.456\V :H):) ().2 !IISAS2 1981 Feb. ltl 13 :38 22 .11 11.9265 178.632\V 566 5.1 !11SAS1 1987 Apr. 29 1•1 27 :35.7 19.015 177.73\\ ' :385 5.~) LA3 1981 Aug. 26 6 56 46.2 20.745 118.46\V 56!) !>.7 LA:3 1988 .) an. 21 16 00 01.55 17.7635 178.737\ \' 566 5.1 l\ ISA51 1988 l\ov. 16 05 53 20.29 21.7685 179.426\V 582 5.8 !11SAS2 1989 Apr. 18 11 33 52.16 2:3.83·15 179.9·14£ 521 5.8 ~ I S:\S3 1989 :\lay 21 21 56 48.6:3 11.9525 178.593\\' 581 5.7 i\ ISAS1 1989 ::\o\'. 29 0:) 48 59.84 25.3745 119.6298 187 5.1 i\IS:\5:3 1990 :\lay 17 15 59 56.53 25.:3985 178.101 E 611 5.8 ,\JS:\53 1990 i\la~ 28 11 28 41.69 20.8145 171.981\\ ' •186 5.9 .\ ISAS2 1990 .Jun 26 12 08 29.39 22.0155 119.-1 t:3 \\' .')$1 6.0 .\15.-\52 1990 .Jun. 29 o:3 53 28.16 21.5525 179.3:32\\' 616 57 .\ISA52 1990 Aug 10 II II 36.78 J 9.8055 177.385\\' :!1:3 60 .\15A52 1990 Oct. 10 0:) 5-I 5:3.54 2:3.4915 179.0298 519 GO .\!S,\5:3 1991 :\ pr. 18 9 ll 20.1 22.925 179.:31\\' •Ill 5.1 L:\2 1991 Aug. 28 21 :32 35.4 22.2:35 179.6·1E .')99 55 LA1 1991 Srp. :30 0 21 46.4 20.8185 118.59\\' 566 6.3 013 1992 .Jan. I :3 9 :n 4:3.1 20.9:35 118. 72 \\' 515 5.7 L:\1 1992 Apr. ·I 1 11 12.:3 11.955 118.:31\\' 511 5.6 L:\ 1 1992 Aug. :30 20 9 5.8 17.925 118.71 \\' 565 5.8 LA1.2,1313 1992 Nov. 12 22 ::!8 51.5 22.4015 l/8.10 \ \' :360 5.9 BB 199:3 .\I a r. 21 5 'I 59.2 18.0..t5 I 78 .5:3\V 58$1 6.1 LA I .~ I SAS 1 ,f313 199:3 ,\ pr. 16 1·1 8 :38 .9 11.178$ 178.86 \V !)()!) G.O Bl3 19!'):3 r\ p r. 20 I (j 26 19.5 20.88S 118.70\\' !)!)::! 5.G LA 1.2 199:3 .) ul. !) 1:) :n 5:3.1 19.785 171.49\\' :398 6.0 LA1 199:3 r\ug. i 11 53 2·1. 2 23.8665 119.858 !)2:3 6.0 13B 199:3 Ort. 11 I :3 I 29.6 17.855 178.1:) \\' ::··

·)').) 58 L,\ 2,~ 1 S,\ Sl.l313

19!1 I .\Jar. H :.!:3 28 6.1 18.0:395 118. 11 \\' ;j():l () (j 1313 I!)!) I ~Jar. :31 '2'2 10 :):!.1 :-, 12.051S 11!).:):!:3 \\ ' 51:'0 6. 1 ~I SAS2,BB I !l!) I Ort .,-_, :?-2 :.W ~----> :l5.118S li!Ul l: ;"',( q .-,,!) 1313 I !l!):) .Jan . II lfi 51 1:?.0 :?U.I:'IOS 119.~1\\' fi:ll fi () BB

Tab I<· 2 .

Data So•l ~ \o Of (0) I ran·,; -L.\ I HI 1-i ll.ti .)I

L:\2 (i(i.() 10. 1 :11 (,,\:1 lll.li II I :II

.\IS:\S I Iii . I ( i1' (I I~

:-. IS.\ S:.! fi!l.li lll.'i II .\ IS.\S:I II () 1:.!.!1 .,-_,

l l l l .-, 'i I i 1.!1 'iO

)

Surface

Solid Mantle hypocenter

P.S PdP,SdS

Liquid Core

1· ~~5~,\~ I

120" 140" 160" 180"

40" 40"

20" 20"

o· o·

-20" -20"

120" 140" 160" 180"

40

30

120

• • • • •

130

• •

·~~~ • LAl

• LA2

• LA3 ~~~-

• MSAS

:~'% BB

140 150

15510~0

------~~==~~-----=====~~----.c======~------~ 15° 15° ~ 155° 160° ) 165°

{*-~~ LAl

~8 LA2

~~) LA3

MSASl

1 oo ~ f1 MSAS2

50 ~

e MSAS3

eBB

~-----

oo ~ ~~__/

150° 155°

()

• • •

)

-1 oo

~

• • • ~ 50

• .. -• ·' •

• ~ oo

160° 165° F·a.,~ ~ +

400

300

200

100

400

300

200

100

..-... E 4oo

.:::£ ..._... 0 300

+oJ

~ 200 -"+-

Q) 100 a: .... .... 0

0 10 20

0 10 20

0 10 20

Differential Travel Time(sec)

MSASl original

• z

z

I

-1 0 0 1 0 20 30 40

Differential Travel Time (sec)

Obs. / Syn. 66:~f42° _________ ___ " ,"' ..... ~.,-·-------- ..

'I

-6--6-------------- / "v -- ----- , ' v o 64 2 0 - I ,'•·'' ',,,.. ,'\ -

\ ,' __ , \ /,., __ , _________ _

I I •

1)

tl I I

I I 66:24_6_0

_____________ , \ ,' I I ,,

6 6: 3-~ii 0----------- --~'

66:;tti6_<> __________ ,

s SdS ScS

)

19920830 mb5.8 ________ _____ ,. d=565km 58.619° -

19930321 mb 6.1 ___________ ,_,_, d=589km 58.825°

19930416 mb 6.0 _____________ , d=565km 58.420° -

19931011 " '

,' ... _,, ..... _ --- ... --------- ... __ .,,,-\, I I ,,,

'~ -~~· .... _________________ ,, \ , ...... _,- ... _____ ___ . I I

v

,, / I , ... , __ ....... _____________ , \ ,'

1,,

, ......... _____ _

m 58 ..,, ., b • I IV " ,-, d=555km ss:s·s-io'-' \ ('•' .. ________________ _, \, ,- ... ___ _____ _ ,

19940309 mb 6.6 ______________ _, d=563km 58.898°

s

I I 1,1

,-, ... ,,.......... ......... ,-... ,' \ I - .,..., ........................... \ , ... .,

I I 1

l I \ ,' \ I \ J ,I

SdS ScS

100 sec

~ - ... ... ,----

F·~v.~~ ?

15" ~--=====--=======---=====---=+=1

1 o·

5"

0

ScS-S Residual (sec) 0 oof\

2.0 4.0 6.0 Si

0

150" 155" 160"

a)

0

0

0

165"

15 ° ~-r::::::::====--=======---=====---==+=1

D" thickness from SdS (km) b)

0 0 0 100 200 300

1 o·

5 LAl

0 LA2 ;·~·· ,· LA3

MSASl o· MSAS2

e MSAS3

150" 155"

0

160" 165"

a) b)

ll 0 10 20 0 10 20

4()()

300 300

200 200

100 100

0 0 10 20 0 10 20

15

]j ·10 ·15

0 10 20 0 10 20

Differential Travel Time

2400 2500 2600 2700 2800 2900

Differential Travel Time

~"S--- I ->~1Jt= I I~ 2300 2400 2500 2600 2700 2800 2900

Depth(km) Depth(km)

c) d)

~~ 0 10 20

400

300 300

200 200 Pdl~""- -- -Z31J J{,ii

100 100

0 0 10 20 0 10 20

15

Jl ·15

Figure g

e) D

0 10 20 0 10 20 400~----~~---------------~--~

300 ~JP<rrr- -----. I< in

200

100

0 10 20 0 10 20

:ll-~--~~ .:ll---~F£03 0 10 20 0 10 20

J IJ=:=:~~r 2500 2600 2700 2800 2900 2300 2400 2500 2600 2700 2800 2900

1:: b : Differ.ential Travel Time

> 2300 2400

Depth(km) Depth(km)

g) h)

0 10 20 0 10 20

0 10 20 0 10 20

.:1+1--:\f\V:f.::~l ·15 .l...---------------',.!!.!-----------'--

0 10 20 0 10 20

1:: ~-~::er~n~:~:;~~ > 2300 2400 2500 2600 2700 2800 2900

Differential Travel Time

ul·~---------~.----~------~r· t I -----.---~'ULC~lf > 2300 2400 2500 2600 2700 2800 2900

Depth(km) Depth(km)

Figure 9 (Continued)

i)

0 10 20

0 10 20

Differential Travel Time u ,4 r' --~--~--~--~~--~ th : · : . --~ r > 2300 2400 2500 2600 2700 2800 2900

Depth(km)

Figure 9 (Continued)

60 a) LAl Direct P 40

20 0

-20 -40 -98

0

-10

0 10 20

20 Direct P 10

0 -10 -~§ P P 280 km

0

-10 0 10 20

4 c) LA3 Direct P

0

1:~ 0.5 0.0 -

-0.5 -1.0

0 10 20

L /0 fl ~~,I rQ

)

LAl LA2 LA3

270km

~ .· High Vp (1.4%) HighVp(2~

High ':,~.~~ :. .. .. .. .... .. ......... High V s(?) ...___ir-: --H-ig-h-=-v=-. p:( 2=-. :9 ~~o) .............. .... : . ..

1111111

? : v· s(?) ,,,, . - . ,,, .

,,,,,,,,,,,,, ................ ir. ....•............................ .... , 170km

LowVs LowVs Vs?

CMB

southwest 500km northeast

F·~IA\I'Q. I I

(e.g., the Indian Ocean), but is constant beneath mid ocean

ridge (e.g., the Mid Atlantic Ridge).

Introduction

The mid ocean ridges (spreading regions) have been considered

to play a passive role in the force balance of plate tectonics5 and

only shallow structure has been studied in detai 16 . Seismic tomogra­

phy using surface waves7•8 has a limited resolution in both horizon­

tal and vertical direction, especially poor resolution in the depths

and contrasts of the upper mantle discontinuities.

\Ve examine the results of previous studies1-

4 by analysing the

data from recent large deep earthquakes in Fiji and Bolivia, whose

bounce points of P'P' are located near the Mid Atlantic Ridge and

far from the Mid Indian Ocean Ridge, respectively. We show the

regional difference of the underside reflections of P'P' from the 420-

and the 660-km discontinuities (referred to as P'420P' and P'66oP'

here after) and attempt to interpret the differences in terms of

chemical composition with an assumption of pyrolite mantle.

D ata

The Fiji earthquake of !\larch 9, 1994 (mb= 6.6) and the Boli­

vian earthquake of June 9, 1994 (mb= 7.0) provide us with a rare

chance for the analysis of P'420 P' and P'66oP'. We analyse the data

of short-period vertical-component seismograms from the Fiji and

Bolivian earthquakes recorded by the micro earthquake network

2

of Tohoku University and Southern California USGS-Caltech net­

work. \Ve use the hypocenter parameters li sted in mon thly PDE

(Preliminary Determination of Epicenters) report to calculate the

travel times and bounce points of P'P', P'42oP' and P'660P ' . Figure

1 shows that the bounce points for the Fiji earthquake are located

close to the Mid Atlant ic Ridge and that those for the Bolivian

earthquake far from the Mid Indian Ocean R idge. This diffe rence

of the bounce point locations encourages us to study structural dif­

ference of mantle beneath the mid ocean ridge and normal ocean.

Strict inspection of all data by display of original traces and fil­

tered traces obtained from several pass bands leads to 20 traces for

the Fiji earthquake and 82 traces for the Bolivian earthquake. \Ve

use the short-period vertical-component seismograms filtered with

a pass band of 0.2-0.5 Hz (Fig. 2). The epicentral distance ranges

from 67.1 o to 70.2 o for the Fiji eart hquake and 64.4 o to 70.5 o fo r

t he Bolivian earthquake.

Analys is

The slowness is one of important parameters to identify the un­

derside reflections of P 'P' from the upper mantle discontinuities.

The slant-stacking of the filtered seismograms in Fig. 2 is made

with a slowness interval of 0.1 sec/deg to measure the travel times

and slownesses of significant signals in the data. The comparison of

the calculated t ravel t imes and slownesses with those of clear peaks

3

in the contour maps indicates that P'660 P' from the Fiji earthquake

(Fig. 3a), and P'ssoP' and P'42oP' from the Bolivian earthquake

(Fig. 3b) can be identified without ambiguity. The fact that the

signals of P'660P' and P'42oP' consist of three dominant peaks (Fig.

3b) is a strong evidence for the signals from the great Bolivian

earthquake, because the waveforms of P waves recorded by broad­

band instruments suggest that three dominant shocks occurred dur­

ing Lhe rupture process of the earthquake. Another apparent signal

with slowness of -4.3 sec/deg (Fiji) and -4.3 sec/deg (Bolivia) is not

interpreted as P'dP' ( d < 420-km), but as SI<I<I<P9.

Figure 4a and 4b show the result of the slant-stacking with a

slowness of -2.7 secjdeg for the Fiji and Bolivian earthquake, respec­

tively. The signals of P'660P' clearly show up for both earthquakes.

The amplitude of P'420P' is almost similar to that of P'ssoP' for the

Bolivian earthquake, but is buried in ambient noises in the case of

the Fiji earthquake. These results in Figs. 3 and 4 arc consistent

with the previous array studies1-

4• This is the first clear evidence

for Lhe regional variation of underside reflection of P'P' from the

420-km discontinuity since the pioneering studies 10-

12 of precur­

sors to P'P'. The seismological discovery of the present study is

that there exists the difference of the reflectivity of the 420-km dis­

continuity between the area close (about 2 o) to the Mid Atlantic

Ridge and the area far away (about 20 o) from Lhc Mid Indian

4

Ocean Ridge. The data sets analysed in the present study does not

show such a clear regional variation for the 660-km discontinuity.

However, one of the array studies of P'dP'2 suggests the possibility

of a weak reflectivity of the 660-km discontinuity beneath the ~lid

Atlantic Ridge.

Implications on chemical composition

Advances in high-pressure experiments have reached a depth of

about 800 km in the Earth's mantle13-

15. It may be possible for us

to interpret the seismological results of the present study in terms

of mineralogical phase relation. We assume Mg2Si04-Fe2Si04 sys­

tem in the following discussion.

420-km discontinuity

Recent high pressure experiments16'17 of olivine (a) to modified

spinel ( (3) transformation, which is considered to correspond to the

420-km discontinuity, suggest the transformation width of 11 to 19

km in temperature range of 1400 to 1750 ·c for the composition of

(Mgo.sgFeo.11 )Si0416

, and the decrease of the width with tempera­

ture increase16•18

. Our observational result from the Fiji earthquake

is consistent with this experimental result. However, the clear re­

flection from the Bolivian earthquake seems inconsistent with the

width range. An easy way to overcome this inconsistency is to

change the Mg-Fe ratio from (Mg0 .89 Fe0 . 11 )Si04 beneath the dis­

continuity. They 16 infer that Mg content is grcaLcr than 95% or

5

smaller than 80%. We can suggest that the increase of Fe content

to 17% beneath the discontinuity is plausible. The effect of H2 0 19

on the transformation should be investigated with careful experi-

ment.

660-km discontinuity

Recent high pressure experiment 13 shows that the transforma­

tion of spinel to perovskite + magnesiowi.istitc take places at a

pressure of 23 GPa with a pressure interval less than 0.15 GPa at

a temperature of 1600 °C. These pressure values are converted to a

depth of 653 km and a depth interval less than 5 km. These values

are consistent with the results of the present study for both the Fiji

and Bolivian earthquakes.

One of the recent array studies2 suggests the possibility of the

reflectivity decrease with decreasing distance of the bounce points

to the Mid Atlantic Ridge. This might require a change of chemical

composition near the 660-km discontinuity across the Mid Atlantic

Ridge.

Conclusion

An unambiguous regional variation of the reflectivity of 420-km

discontinuity is observed between the bounce points close to the

Mid Atlantic Ridge and those far away from the Mid Indian Ocean

Ridge. The reflections from the 660-km discontinuity arc clearly

observed from the both bounce points. Recent array stucly2 by one

6

of the authors of the present study finds an evidence for the change

of reflectivity of the 660-km discontinuity beneath the !\lid Atlantic

Ridge.

Interpretation of these results in terms of chemical composition

is not unique. However, the change of chemistry near the disconti­

nuities is a simple but most plausible interpretation. Interpretation

by temperature variation seems difficult. The focussing and defo­

cussing due to heterogeneities in the Earth cannot be ruled out.

We must point out that the bounce points for the Fiji earthquake

are close to a hot spot20 (Fig. l(a)). There found no hot spot near

the bounce points for the Bolivian earthquake.

Acknowledgmen ts.

The data analysed in this study were provided by courtesy of To­

hoku University and California Institute of Technology.

7

References

1. .:'-Iakanishi, I., Seismic reflections from the upper mantle disconti­

nuities beneath the ~lid-Atlantic Ridge observed by a seismic array

in Ilokkaido region, Japan, Geophys. Res. Lett., 13, 1458-1461,

1986

2. Nakanish i, I., Reflections of P'P' from upper mantle discontinu­

ities beneath the Mid-Atlantic Ridge, Geophys. J. R. Astron. Soc.,

93, 335-346, 1988

3. Nakanishi, I., A search for topography of the mantle discontinu­

ities from precursors to P'P', J. Phys. Earth, 37, 297-301, 1989

4. Benz, H. M. and J. E. Vidale, Sharpness of upper-mantle discon­

tinuities determined from high-frequency reflections, Nature, 365,

147-150, 1993

5. Forsyth, D. and S. Uyeda, On the relative importance of the

driving forces of plate motion, Geophys. J. R. Astron. Soc., 43,

163-200, 1975

6. Detrick, R. S., P. Buhl, E. Vera, J. i\Iuttcr, J. Orcutt, J. Madsen,

and T. Brecher, Multi-channel seismic imaging of a crustal magma

chamber along the East Pacific Rise, ~ature, 326, 35-41, 1987

7. Nakanishi, I. and D. L. Anderson, 1\[easurements of mantle wave

velocities and inversion for lateral heterogeneity and anisotropy­

fl. Analysis by the single-station method, Geophys. J. R. Astron.

Soc., 78, 573-617, 1984

8

8. Woodhouse, J. H. and A. M. Dziewonski, Mapping the upper

mantle: Three-dimensional modeling of Earth structure by im·er­

sion of seismic waveforms, J. Geophys. Res., 89, 5953-5986, 1984

9. Engdahl, E. R. and E. A. Flinn, Remarks on the paper "Early

reflections of P'P' as an indication of upper mantle structure," by

R. D. Adams, Bull. Seism. Soc. Am., 59, 1415-1417, 1969

10. Engdahl , E. R. and E. A. Flinn , Seismic waves reflected from

discontinuities within Earth's upper mantle, Science, 163, 177-179,

1969

11. Whitcomb, J. H. and D. L. Anderson, Reflection of P'P' seis­

mic waves from discontinuities in the mantle, J. Geophys. Res., 75,

5713-5728, 1970

12. Adams, R. D., Reflections from discontinuities beneath Antarc­

tica, Bull. Scism. Soc. Am., 61, 1441-1451, 1971

13. Ito, E. and E. Takahashi, Postspinel transformat ions in the sys­

tem Mg2 Si 0 4 -Fe2Si04 and some geophysical implications, J. Geo-

phys. Res., 94, 10637-10646, 1989

14. Anderson, D. L., Theory of the Earth, Blackwell Scientific Pub­

lications, Boston, 1989

15. l rifune, T., Phase transformations in Lhe earth's mantle and

subducting slabs: Implications for their compositions, seismic ve­

locity and density structures and dynamics, The Island Arc, 2, 55-

71, 1993

9

16. Katsura, T. and E. Ito, The system I\1g2Si04-Fe2Si04 at high

pressures and temperatures: Precise determination of stabilities of

olivine, modified spinel, and spinel, J. Geophys. Res., 94, 15663-

15670, 1989

17. Akaogi, M., E. Ito, and A. Navrotsky, Olivine-modified spinel­

spinel transitions in the system Mg2Si04-Fe2Si04 : Calorimetric

measurements, thermochemical calculat ion, and geophysical appli-

cation, J. Geophys. Res., 94, 15671-15685, 1989

18. Ilelffrich, G. and C. R. Bina, Frequency dependence of the vis­

ibility and depths of mantle seismic discontinuities, Geophys. Res.

Lett., 21, 2613-2616, 1994

19. Wood, B. J., The effect of H20 (water) on the 410-kilometer

seismic discontinuity, Science, 268, 74-76, 1995

20. Turcotte, D. L. and G. Schubert, Geodynamics: Applications

of continuum physics to geological problems, John Wi ley & Sons,

450pp, 1982

10

r

Figure Captions

Figure 1. Surface bounce points (solid circle) of P'P'. The lo­

cations of hot spots20 (solid stars) are also shown. (a)Fiji-Japan.

(b )Bolivia-Southern Carifornia.

Figure 2. Short-period vertical-component seismograms filtered

with a pass band of 0.2-0.5Hz. (a)Fiji-Japan. (b)Bolivia-Southern

Carifornia.

Figure 3. Contour map showing the result of slant-stacking

of the filtered seismograms in Fig. 2. Also shown are the locations

of P'P', P'42oP', P'66oP', and SI<KI<P. (a)Fiji-Japan. (b)Bolivia­

Southcrn Carifornia.

Figure 4. Cross-section of Fig. 3 for a constant slowness of

-2.7 sec/deg. Also shown are the locations of P'P', P'42oP', P'660P'.

(a)Fiji-Japan. (b)Bolivia-Southern Carifornia.

11

. 0

....... . .

L_~ I

~ 'v->!v

r

. 0 -· 7' I \

)

-

- ) . <0 :.;:t_ ('I) 7 \

~ ~-.~

l\

J . (\J v ('I)

. 0 ('I)

~ ('I)

L_ ) ~ '-

P" I

~· ,.1

. 0

. 0 ('I)

I

-;::;z ~

. . 0 (\J r- r-

I I

~ . <0

1-..q--('I)

~ v •• v

~~1-('I)

••••• . (\J v ~

('I) . . 0 (\J r- r-

I I

~ ~

,

I l

. 0 ('I)

I

. 0

0

0 ('I) ('I)

60. go· 120.

-3o·

g2·

go· 120.

r

0 Vl

0 :::> ("") ('l)

><

Time (sec) (VM OFF) X 10+2

YMZ.uth

'J.XPOOC~~NfYtA uiJIIII'o'.oii~SNR.ulh

HOJ.uth HSK.ulh

-TBS.uth .~~NIB.ulh

~~--w~-OGA.ulh fVV'IW~,.,.,.,._.1-TMR. uhs

--t"-------~~.-vv-~----~~--.......,......<V--~~IIlll\.tW-I\JIINifii',;-V>MAJ.Mv>tv ... N'-IINI~-1-8NJ. uhs

·Jv'---A-.~~-+-MMA. uhs

Fi/j . 2(ct)

><

0 + ('...)

X 1012 Time (sec) [VM OFF)

f;J _ 2 ( hJ

0 ....--.... 0) (1) -1

"'0 --as -2 (/) ....._.... (/) -3 (/)

~ -4

$ 0 -5

(f) -6

I

- ~ -

-P' 66oP' ~

I

-

-

I

Fiji-Japan I I I

~ ~J' I .. 4 ,,

t I

I , -

I~~~~~~' I ~ 1-

I P'P' . ~ I._

~ ~~ l~ l/1

ij -

SKKKP I ~ 1

,v l' t ~ I

VI~ ~~~ 1 R ~ 1-

I ,~ ' I I I

I I I I

2100 2150 2200 2250 2300 2350 2400

Travel Time(sec)

Bolivia-US o~~~~~~~~~~~~~~~~~+

....--.... (J) Q) -1

""0 -­() -2 Q) (/) .....__. (/) -3 -(/)

Q) -4-

~ . t ~ ~ 0 -5 .

(J) .

P' 66oP '

r I

-6+·~~~~~~~~~~~~~~~~~

2100 2150 2200 2250 2300 2350 2400

Travel Time(sec)

Fiji-Japan

500

400 P'P'

300

200

100

0 2100 2150 2200 2250 2300 2350 2400

Travel Time(sec)

Bolivia-US

40

30 P'P'

20

10

o~~~~~~~~~~~~~~~~~+

2100 2150 2200 2250 2300 2350 2400

Travel Time(sec)