GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L04302, doi:10.

1029/2005GL025106, 2006

Observations of the mid-mantle discontinuity beneath Indonesia from

S to P converted waveforms
Elizabeth Vanacore,1 Fenglin Niu,1 and Hitoshi Kawakatsu2
Received 1 November 2005; revised 4 January 2006; accepted 17 January 2006; published 23 February 2006.

[1] S to P conversions were employed to derive a coherent Our previous studies [Kawakatsu and Niu, 1994; NK97]
discontinuity structure beneath Indonesia. Analysis of data employed short-period seismograms recorded by Jarray
recorded by three regional arrays from nine deep earthquakes [J-Array Group, 1993] to identify S‘1000’P. The major
not only confirmed the results of previous studies on the concern of using Jarray data is whether the complicated
existence of the mid-mantle discontinuity beneath the Java mantle structure beneath the array could affect the slowness
arc but also revealed its presence north to Kalimantan Island. measurement and phase identification. We accordingly
S to P waves converted at the discontinuity were observed on analyzed the Indonesia region earthquake data recorded
the stacked diagrams with a negative slowness relative to the by the Southern Africa Seismic Experiment broadband
P wave and a conversion depth ranging from
1080 km in array (Kaapvaal array) [Carlson et al., 1996].
the west to
930 km in the east. Citation: Vanacore, E., F.
Niu, and H. Kawakatsu (2006), Observations of the mid-mantle
discontinuity beneath Indonesia from S to P converted waveforms, 2. Data and Method
Geophys. Res. Lett., 33, L04302, doi:10.1029/2005GL025106.
[ 4 ] Nine regional events with depths greater than

500 km (Table 1) were studied in order to observe S to
1. Introduction P conversions; the events consist of eight newly analyzed
events and one data result from NK97 recorded by Jarray,
[2] Mapping seismic heterogeneities in the upper part of
Hinet, and the Kaapvaal array. Event data from the three
the lower mantle (mid-mantle) is important to the under-
networks available to us are before 1995 for Jarray,
standing of mantle dynamics and the distribution and the
after 2001 for Hinet, and 1997 – 1999 for Kaapvaal array.
nature of geochemical reservoirs [Wen and Anderson, 1997;
The nine deep events are selected from the Harvard CMT
Kellogg et al., 1999; van der Hilst and Karason, 1999;
catalog [Dziewonski et al., 1981] with a moderate size (5.5 <
Tackley, 2000; Helffrich and Wood, 2001]. Recent seismo-
Mw < 6.5) and a simple source-time function.
logical studies indicate seismic discontinuities [Kawakatsu
[5] Prior to stacking seismograms for the individual
and Niu, 1994; Niu and Kawakatsu, 1997; Vinnik et al.,
events, the raw data are preprocessed in three basic steps.
1998, 2001] and strong seismic reflectors/scatterers at
First, the signal to noise ratio (SNR) of the individual
mid-mantle depths in the western Pacific subduction
seismograms is calculated with based on amplitudes of the
regions [Kaneshima and Helffrich, 1999; Castle and
P wave and the background noise; seismograms with a
Creager, 1999; Kruger et al., 2001; Niu et al., 2003;
minimum ratio of
2 are retained for further analysis. For
Kaneshima and Helffrich, 2003; Kaneshima, 2003]. The
the individual events analyzed, the number of retained
mid-mantle discontinuities are absent in other regions
seismograms ranged from
50 to 600. In all but two events,
such as beneath the South American subduction zone
the number of retained traces was greater than 100. Sec-
[Castle and van der Hilst, 2003]. Most of the seismic
ondly, the remaining seismograms are low pass filtered with
observations are based on the detection of S to P
a corner frequency of 1 Hz. Thirdly, handpicked P-wave
conversion at the discontinuities.
first arrivals of the individual seismograms are aligned to 0s,
[3] This study’s goal is to conduct a search for seismic
to allow for stacking with reference to the P wave arrival. In
structure using S to P conversions over an expanded area
Figures 2a and 3a, we show two examples of the data after
compared to our previous study [Niu and Kawakatsu, 1997,
preprocessing. Here, the waveform data are recorded by the
hereinafter referred to as NK97] including the entire Kali-
borehole short-period Hinet (Figure 2a) and the broadband
mantan Island (Figure 1) for future comparison with SS
Kaapvaal array (Figure 3a). In both examples the S to P
precursor data. We chose this region as our testing example
converted phase is clearly visible before stacking in the
for two reasons: first, NK97 suggests that the mid-mantle
time window of 35 – 45s after the direct P wave. No
discontinuity has the largest S-wave velocity contrast in the
seismic waves are expected to arrive in this time window
region; the S to P conversion at the discontinuity (hereafter
based on 1D global reference velocity models, such as
S‘1000’P) is visible in most of the individual recordings, so
iasp91 [Kennett and Engdahl, 1991].
there is no issue of stacking artifacts; second, a SS precursor
[6] The preprocessed data are 4th-root stacked with
study by Deuss and Woodhouse [2002] found several
respect to the P wave arrival to highlight coherent later
negative/positive reflection peaks at same depth range.
arrivals [Muirhead, 1968; Kanasewich, 1973]. Details of the
1
stacking procedure are given by Kawakatsu and Niu [1994].
Department of Earth Science, Rice University, Houston, Texas, USA. In the stacked seismograms the SDP phase is characterized
2
Earthquake Research Institute, University of Tokyo, Tokyo, Japan.
by an arrival time between the direct P and pP phase arrivals
Copyright 2006 by the American Geophysical Union. and a negative slowness that separates these arrivals from P
0094-8276/06/2005GL025106 wave arrivals associated with aftershocks. Stacked results

L04302 1 of 4
L04302 VANACORE ET AL.: MID-MANTLE DISCONTINUITY BENEATH INDONESIA L04302

yield depths ranging from 1080 km to 930 km (Table 1).

The predicted slowness listed in the table is the relative
slowness calculated for an S wave converted from P at the
observed depths for each event. We assumed a horizontal
discontinuity in the calculation. In general, the observed
slowness matches the expected one quite well. Data from
two events, however, show a relatively large discrepancy,

0.1 s/deg. Many factors including anomalous structure
near the source and beneath the arrays could affect the
slowness measurement. Since the anomalous slowness of
the two events is measured from two different arrays, it is
easier to explain the discrepancy by a dipping structure,
although the uncertainty in the slowness measurement is
almost at the same level (Figures 2b and 3b). The major
Figure 1. Map view of the nine events and respective feature of the mid-mantle discontinuity shown here is that
S‘1000’P conversion points. P-wave tomographic model of its depth is shallow in the east; depth observations of less
Fukao et al. [2001] for the layer of 900 to 1000 km is than 1000 km are restricted to longitudes greater than
shown by the color contour. Line A-A0 is the cross section 113°E.
shown in Figure 4. [8] Conversion points of S‘1000’P for events #1, 2, 6, 7,
8 and 9 are located within the intense mid-mantle high
velocity anomaly (HVA) beneath the Indonesia region
(Figures 2b and 3b) are employed to estimate the conversion (Figure 1), which has been consistently imaged by most
depth of the observed SDP phases. Arrival times relative to tomographic studies using different data sets as summarized
the P-wave arrival are compared to corresponding theoret- by Fukao et al. [2001]. S‘1000’P amplitudes of these events
ical travel times estimated using the iasp91 model from are very large, indicating some sort of correlation between
which the conversion depths and the theoretical slowness the mid-mantle discontinuity and the HVA (Figure 4).
are derived. The conversion points within the mantle are [9] We noticed a series of arrivals between
30– 60 s
calculated for each earthquake-station pair for a discontinu- after the direct P, similar to observations at the Mariana arc
ity at 1000 km depth, the approximate average depth of the [Niu et al., 2003; Kaneshima and Helffrich, 2003;
discontinuity in the study region, shown as point clusters in Kaneshima, 2003]. Since we used a nonlinear stack scheme
Figure 1. and assumed horizontal discontinuities, scattering energies
from small-scale heterogeneities [Kaneshima and Helffrich,
2003] did not show up in the stacked traces. The coexis-
3. Results tence of these small-scale heterogeneities might be very
[7] 4th-root stacks of the nine events analyzed exhibit a important to understanding the nature of the mid-mantle
discontinuity throughout the region with the exception of discontinuity. Further investigations with more sophisticated
events three and four, located in the southeast corner of the seismic methods, such as migration, are preferred to image
study region. The lack of a S‘1000’P arrival for these two the small scale heterogeneities.
events does not necessarily imply a limit on the extent of the
discontinuity; NK97 contains data from two earthquakes in
the same sub-region occurring at (6.86°, 125.41°) in 1990 4. Discussion
and (7.26°, 122.57°) in 1991 with mid-mantle disconti- [10] Observations of mid-mantle discontinuities in the
nuities observed at 940 km and 945 km respectively. Indonesia region at depths ranging from
920 – 1080 km
Amplitudes of the S‘1000’P of these two events are about depths and dipping to the west are prevalent throughout the
half of the other events. The lack of a clear S‘1000’P arrival region as demonstrated in this study by S to P converted
after the stacking process may be due to unfavorable focal energy. The abruptness and geometry of this depth change
mechanisms associated with the two events from this study. is, however, not well constrained by our results. Given the
Conversion depths calculated by matching arrival times of consistency of the depth estimates from the events in the
the S‘1000’P arrivals based upon the IASP91 velocity model western portion of the region, numbered 6 – 9 (Figure 1 and

Table 1. Earthquakes Used in the Study and Observations of the S‘1000’P

Event # Event Date Tobs, s Pobs, s/deg Depth, km Pcal, s/deg Lat. Long. Event Depth, km Mag., Mw
1 12/21/02 34.0 0.24 930 0.32 4.97 123.12 596 5.6
2 07/01/03 38.0 0.30 970 0.35 4.57 122.55 595 6.0
3 12/17/02 — — — — 6.97 125.38 493 5.7
4 02/03/03 — — — — 7.07 123.68 585 5.8
5 02/16/01 40.5 0.23 960 0.26 7.05 117.50 533 6.0
6 04/27/98 41.2 0.16 1080 0.15 6.08 113.10 571 5.7
7 07/11/97 43.3 0.24 1080 0.15 5.70 110.80 574 6.0
8a 11/15/94 48.6 0.30 1080 0.32 5.61 110.20 570 6.2
9 05/13/95 49.6 0.30 1080 0.28 5.22 108.92 554 5.9
a
Indicates event from previous study [Niu and Kawakatsu, 1997].

2 of 4
L04302 VANACORE ET AL.: MID-MANTLE DISCONTINUITY BENEATH INDONESIA L04302

Figure 2. (a) A subset of Hinet seismograms plotted in the

order of epicentral distance for a deep event that occurred at
Figure 3. Another example. Seismograms are recorded by
the Celebes Sea in July 1, 2003. Arrows indicate a later
the Kaapvaal broadband seismic array, clear S to P
arrival shown in most of the seismograms. (b) Amplitudes
conversion at
1080 km depth is also observed here.
of the 4th-root stacked traces are shown in color code as a
function of slowness for the same events. Later arrivals are
identified by the observed time and ray parameter. Both are olating this correlation in combination with the correlation
shown with respect to the direct P wave. between the observed discontinuity and the regional HVA,
we associate our observations with slab material stagnated
Table 1) and the shallow discontinuity depths observed in at depth. A global search for the observed
1000 km
the east, it is reasonable to hypothesize that the discontinu- discontinuity is necessary to determine if this phenomena
ity has a large amount of topography, over 100 km in is directly associated with slab stagnation at mid-mantle
magnitude, between
113° and
120°. depths or is associated with a global discontinuity possibly
[11] Although there are some suggestions that the ob- due to a compositional boundary within the mantle. One
served mid-mantle reflectors/scatterers are associated with study by Shen et al. [2003] observed a P to S conversion at
subducted oceanic crust [Kaneshima and Helffrich, 1999;
1000 km depth beneath the Hawaii and Iceland hotspots
Niu et al., 2003], we still don’t have a conclusive explana- and suggested that the mid-mantle discontinuity might be
tion for the mid-mantle discontinuities beyond the fact that associated with a compositional boundary [Wen and
it is not caused by global phase transitions. Furthermore, it
has been recently proposed the mantle contains a chemical
boundary layer at
1000 km depth where one line of
evidence is the correlation between the surface location of
subduction in the Jurassic to Eocene and the Permian to
Triassic with the presence of a tomographic anomaly at 800
to 1000 km depth [Wen and Anderson, 1997; Anderson,
2002]. Our study area is within a correlated zone, suggest-
ing that the seismic discontinuities observed are associated
with slab material accumulated at mid-mantle depths. The
correlation between ancient subduction zones and the pres-
ence of HVA’s at depths of 500 to 1000 km has been Figure 4. Cross section of A-A0 labeled in Figure 1 from
previously examined in regions directly to the east and 9°S, 105°E to 10°N, 126°E with labeled conversion points
south of our study area [Hall and Spakman, 2002]. Extrap- for 5 events with tomography of Fukao et al. [2001].

3 of 4
L04302 VANACORE ET AL.: MID-MANTLE DISCONTINUITY BENEATH INDONESIA L04302

Anderson, 1997; Kellogg et al., 1999; van der Hilst and Dziewonski, A. M., T.-A. Chou, and J. H. Woodhouse (1981), Determina-
Karason, 1999; Tackley, 2000]. This single phase observation tion of earthquake source parameters from waveform data for studies of
global and regional seismicity, J. Geophys. Res., 86, 2825 – 2852.
occurring in a region without a recent history of subduction is Fukao, Y., S. Widiyantoro, and M. Obayashi (2001), Stagnant slabs in the
not sufficient evidence to argue for a global discontinuity at upper and lower mantle transition zone, Geophys. Rev., 39, 291 – 323.

1000 km depth due to a compositional change. Hall, R., and W. Spakman (2002), Subducted slabs beneath the eastern
Indonesia-Tonga region: Insights from tomography, Earth Planetary
[12] If the structure observed here is related to subducted Sci. Lett., 201, 321 – 336.
slabs, as indicated by the S‘1000’P observations, then we can Helffrich, G., and B. Wood (2001), The Earth’s mantle, Nature, 412, 501 –
speculate on two physical mechanisms to explain the 507.
J-Array Group (1993), The J-Array program: System and present status,
presence of this phase. One possible explanation is the J. Geomagn. Geoelectr., 45, 1265 – 1274.
breakdown of the dense hydrous magnesium silicates Kanasewich, E. R. (1973), Time Sequence Analysis in Geophysics, 364 pp.,
(DHMS) within the cold slabs [Shieh et al., 1998; Ohtani Univ. of Alberta Press, Edmonton, Alberta, Canada.
et al., 2001]. Such a breakdown may trigger chemical Kaneshima, S. (2003), Small scale heterogeneity at the top of the lower
mantle around the Mariana slab, Earth Planet. Sci. Lett., 209, 85 – 101.
segregations within slabs that forms the seismic disconti- Kaneshima, S., and G. Helffrich (1999), Dipping lower-velocity layer in the
nuity as observed in this study. Another possible origin is mid-lower mantle: Evidence for geochemical heterogeneity, Science, 283,
the post garnet phase transition of MORB, which is ob- 1888 – 1891.
Kaneshima, S., and G. Helffrich (2003), Subparallel dipping heterogeneities
served to occur at mid-mantle depths [Ono et al., 2001]. in the mid-lower mantle, J. Geophys. Res., 108(B5), 2272, doi:10.1029/
This interpretation has some difficulty in explaining the 2001JB001596.
relatively flat geometry of the structure, as the subducted Kawakatsu, H., and F. Niu (1994), Seismic evidence for a 920-km discon-
tinuity in the mantle, Nature, 371, 301 – 305.
oceanic crust probably undergoes significant deformation Kellogg, L. H., B. H. Hager, and R. D. van der Hilst (1999), Compositional
once it enters into lower mantle. Even if a simple flat stratification in the deep mantle, Science, 283, 1881 – 1884.
geometry is preserved then a low velocity layer instead of Kennett, B. L. N., and E. R. Engdahl (1991), Travel times for global
discontinuity is expected. A recent mineral physics study by earthquake location and phase identification, Geophys. J. Int., 105,
429 – 465.
Kubo et al. [2002] suggested that the post garnet phase Kruger, F., M. Banumann, F. Scherbaum, and M. Weber (2001), Mid-man-
transition might not occur since MORB garnet can survive tle scatterers near the Mariana slab detected with a double array method,
in metastable state on the order of ten million years when Geophys. Res. Lett., 28, 667 – 670.
Kubo, T., E. Ohtani, T. Kondo, T. Kato, M. Toma, T. Hosoya, A. Sano,
temperature is a couple hundred degrees Kelvin lower. This T. Kikegawa, and T. Nagase (2002), Metastable garnet in oceanic crust at
may lead to an accumulation of a metastable garnet layer at the top of the lower mantle, Nature, 420, 803 – 806.
the top of the lower mantle; the lower boundary of the layer Muirhead, K. J. (1968), Eliminating false alarms when detecting seismic
events automatically, Nature, 217, 533 – 534.
is a possible cause for the regionally observed discontinuity Niu, F., and H. Kawakatsu (1997), Depth variation of the mid-mantle
[Kubo et al., 2002]. Dynamically the metastable garnet seismic discontinuity, Geophys. Res. Lett, 24, 429 – 432.
model provides a mechanism explaining the stagnation of Niu, F., H. Kawakatsu, and Y. Fukao (2003), Seismic evidence for a che-
slab material in the region at depths of
800 to 1000 km mical heterogeneity in the midmantle: A strong and slightly dipping
seismic reflector at mid-mantle depth beneath the Mariana subduction
without invoking the presence of a stratified mantle with a zone, J. Geophys. Res., 108(B9), 2419, doi:10.1029/2002JB002384.
chemical boundary layer at similar depths as suggested by Ohtani, E., M. Toma, K. Litasov, T. Kubo, and A. Suzuki (2001), Stability
Wen and Anderson [1997] and Anderson [2002]. If the of dense hydrous magnesium silicate phases and water storage capacity in
the transition zone lower mantle, Phys. Earth Planet. Inter., 124, 105 – 117.
DHMS or metastable garnet scenario is valid and the Ono, S., E. Ito, and T. Katsura (2001), Mineralogy of subducted basaltic
observed discontinuity is related to subducted slab material, crust (MORB) from 25 to 37 GPa, and chemical heterogeneity of the
then further research is needed to determine the consistency lower mantle, Earth Planet. Sci. Lett., 19, 57 – 63.
Shen, Y., S. C. Solomon, and C. J. Wolfe (2003), Seismological evidence
of this feature within other subduction regions as well as for a mid-mantle discontinuity beneath Hawaii and Iceland, Earth Planet.
possible factors controlling the formation or lack thereof of Sci. Lett., 214, 143 – 151.
this discontinuity. Shieh, S. R., H. K. Mao, R. J. Hemley, and L. C. Ming (1998), Decom-
position of phase D in the lower mantle and the fate of dense hydrous
silicates in subducting slabs, Earth Planet. Sci. Lett., 159, 13 – 23.
[13] Acknowledgments. We thank IRIS, Jarray and Hinet for provid- Tackley, P. J. (2000), Mantle convection and plate tectonics: Toward and
ing data. Discussions with A. Levander and C.-T. Lee were helpful in integrated physical and chemical theory, Science, 288, 2002 – 2007.
preparing for the manuscript. We also thank A. Zollo and two anonymous van der Hilst, R. D., and H. Karason (1999), Compositional heterogeneity
reviewers for their constructive comments. in the bottom 1000 kilometers of Earth’s mantle: Toward a hybrid con-
vection model, Science, 283, 1885 – 1888.
References Vinnik, L., F. Niu, and H. Kawakatsu (1998), Broad-band converted
Anderson, D. (2002), The case for irreversible chemical stratification of the phases from midmantle discontinuities, Earth Planets Space, 50,
mantle, Int. Geol. Rev., 44, 97 – 116. 987 – 997.
Carlson, R. W., T. L. Grove, M. J. de Wit, and J. J. Gurney (1996), Program Vinnik, L., M. Kato, and H. Kawakatsu (2001), Search for seismic discon-
to study the crust and mantle of the Archean craton in southern Africa, tinuities in the lower mantle, Geophys. J. Int., 14, 41 – 56.
Eos Trans. AGU, 77, 273 – 277. Wen, L., and D. L. Anderson (1997), Layered mantle convection: A model
Castle, J. C., and K. C. Creager (1999), A steeply dipping discontinuity in for geoid and topography, Earth Planet. Sci. Lett., 146, 367 – 377.
the lower mantle beneath Izu-Bonin, J. Geophys. Res., 104, 7279 – 7292.
Castle, J. C., and R. D. van der Hilst (2003), Searching for seismic scatter-
ing off mantle interfaces between 800 km and 2000 km depth, J. Geo- 
phys. Res., 108(B2), 2095, doi:10.1029/2001JB000286. H. Kawakatsu, Earthquake Research Institute, University of Tokyo,
Deuss, A., and J. H. Woodhouse (2002), A systematic search for mantle Tokyo 113, Japan.
discontinuities using SS-precursors, Geophys. Res. Lett., 29(8), 1249, F. Niu and E. Vanacore, Department of Earth Science, MS-126, Rice
doi:10.1029/2002GL014768. University, 6100 Main St., Houston, TX 77005, USA. ([email protected])

4 of 4