Geoscience Frontiers 6 (2015) 79e93

H O S T E D BY Contents lists available at ScienceDirect

China University of Geosciences (Beijing)

Geoscience Frontiers
journal homepage: www.elsevier.com/locate/gsf

Research paper

Conjecture with water and rheological control for subducting slab in

the mantle transition zone
Fumiko Tajima a, *, Masaki Yoshida b, Eiji Ohtani c
a
University of California at Irvine, 4164 Frederick Reines Hall, Irvine, CA 92697, USA
b
Institute for Research on Earth Evolution (IFREE), Japan Agency for Marine-Earth Science and Technology (JAMSTEC), 2-15, Natsushima-cho, Yokosuka,
Kanagawa 237-0061, Japan
c
Department of Earth and Materials Science, Graduate School of Science, Tohoku University, Sendai 980-8578, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Seismic observations have shown structural variation near the base of the mantle transition zone (MTZ)
Received 21 October 2013 where subducted cold slabs, as visualized with high seismic speed anomalies (HSSAs), atten to form
Received in revised form stagnant slabs or sink further into the lower mantle. The different slab behaviors were also accompanied
19 December 2013
by variation of the 660 km discontinuity depths and low viscosity layers (LVLs) beneath the MTZ that
Accepted 21 December 2013
Available online 2 January 2014
are suggested by geoid inversion studies. We address that deep water transport by subducted slabs and
dehydration from hydrous slabs could affect the physical properties of mantle minerals and govern slab
dynamics. A systematic series of three-dimensional numerical simulation has been conducted to
Keywords:
Stagnant slab
examine the effects of viscosity reduction or contrast between slab materials on slab behaviors near the
Discontinuity depth variation base of the MTZ. We found that the viscosity reduction of subducted crustal material leads to a sepa-
Deep water transport by subduction ration of crustal material from the slab main body and its transient stagnation in the MTZ. The once
Crustal separation trapped crustal materials in the MTZ eventually sink into the lower mantle within 20e30 My from the
Crustal recycling start of the plate subduction. The results suggest crustal material recycle in the whole mantle that is
consistent with evidence from mantle geochemistry as opposed to a two-layer mantle convection model.
Because of the smaller capacity of water content in lower mantle minerals than in MTZ minerals,
dehydration should occur at the phase transformation depth, w660 km. The variation of the disconti-
nuity depths and highly localized low seismic speed anomaly (LSSA) zones observed from seismic P
waveforms in a relatively high frequency band (w1 Hz) support the hypothesis of dehydration from
hydrous slabs at the phase boundary. The LSSAs which correspond to dehydration induced uids are
likely to be very local, given very small hydrogen (H) diffusivity associated with subducted slabs. The
image of such local LSSA zones embedded in HSSAs may not be necessarily captured in tomography
studies. The high electrical conductivity in the MTZ beneath the northwestern Pacic subduction zone
does not necessarily require a broad range of high water content homogeneously.
2014, China University of Geosciences (Beijing) and Peking University. Production and hosting by
Elsevier B.V. All rights reserved.

1. Introduction
In the past three decades many seismic tomography models
have visualized common long-wavelength features of high seismic
speed anomalies (HSSA) associated with subducted cold plates. The
tomography models captured images of different behaviors of
* Corresponding author. Tel.: 1 562 343 4148; fax: 1 949 830 2938.
E-mail addresses: tajimafumiko741@gmail.com, ftajima@uci.edu (F. Tajima).
Peer-review under responsibility of China University of Geosciences (Beijing)
subducted cold plates from region to region as they sink into the
mantle transition zone (MTZ), some atten to form stagnant slabs
with a lateral extent of over a few thousand kilometers, or others
penetrate into the lower mantle (e.g., van der Hilst et al., 1991, 1997;
Widiyantoro et al., 1999; Fukao et al., 2001; Grand, 2002; Zhao,
2004).
The tomography studies led numerical studies to produce a new
class of global mantle convection models with a link to seismic
models (e.g., Bunge and Richards, 1996; Bunge et al., 1998; Becker
and Boschi, 2002; Schuberth et al., 2009a). Simultaneous joint in-
versions of global seismic and geodynamic data (e.g., Simmons
et al., 2006) have shown greatly improved ts to the global

1674-9871/$ e see front matter 2014, China University of Geosciences (Beijing) and Peking University. Production and hosting by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.gsf.2013.12.005
80 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93

convection-related data using the mantle ow theory and provided

supports to reconcile tomography models with geodynamics of
large-scale mantle heterogeneity (Forte, 2007). The large-scale
heterogeneity is generated by the mantle convection process
that is driven by thermal anomalies and agrees well with seismic
tomography images except in the MTZ (Schuberth et al., 2009b;
see Fig. 1).
On the other hand there are still substantial differences among
the published models in their relatively short wavelength features,
in particular, in the MTZ and the upper-most lower mantle where
the major phase transformations of subducted slab minerals and
associated dehydrations occur. Shorter wavelength features in this
depth range should have clues to probe micro processes that
govern macro physics in mantle dynamics. Nonetheless, in-
terpretations of anomaly features are still non-unique and can be
equivocal in terms of effects of temperature, water and geochemical
properties. Different interpretations of anomaly features could lead
only to diverse modeling and hypotheses.
The separation of the crustal layer from the main body of sub-
ducting lithosphere in the MTZ has long been debated (e.g.,
Anderson, 1979; Ringwood and Irifune, 1988; Ringwood, 1994).
Numerical studies have tested mantle convection models with
physical parameters that were measured under high pressure and
temperature conditions. Only a few of them considered models that
are composed of the crust and mantle components with different
densities but the same homogeneous viscosity in two-dimensional
(2D) convection simulation (e.g., Richards and Davies, 1989;
Gaherty and Hager, 1994). These studies did not produce separa-
tion between the two components and concluded that a separation
of the crustal component from the subducting lithosphere should
not occur at the 660 km phase boundary. van Keken et al. (1996)
have shown a simplied sandwich model, in which a thin, weak
layer between the crust and the cool slab interior can effectively
decouple the crust from the slab main body, and the lighter crust
can rise, leading to a garnet enriched MTZ.
A number of researchers carried out seismic waveform
modeling with a focus on the MTZ structure or the topography of
the seismic discontinuity depths that bound the MTZ using rela-
tively short-wavelength body waves in an attempt to supplement
long-wavelength seismic tomography studies (e.g., Tajima and
Grand, 1995, 1998; Brudzinski et al., 1997; Tajima et al., 1998).
These studies resolved the ner structure beneath the attened
slabs in the northwestern Pacic (NWP) subduction zones with
discontinuity depth variations. The result is an important implica-
tion for mantle dynamics. However, the seismic rays used in these
studies were too sparse to constrain the anomaly structure of the
attened subducted slabs.
More recent studies attempted to determine the MTZ structure
associated with stagnant slabs using broadband data recorded by
the dense seismic networks in China (Wang and Chen, 2009; Wang
and Niu, 2010, 2011; Ye et al., 2011; Li et al., 2013). Nonetheless, the
methods adopted for the waveform modeling are similar to pre-
vious one-dimensional (1D) studies and did not fully utilize the
advantage of data from the dense networks. There are also studies Figure 1. Radial mantle ow magnitudes based on the convection model by Schuberth
et al. (2009a) are shown in the upper mantle (w340 km), mid-mantle (w1450 km),
that determined discontinuity depth variations using long-period
and near the base of lower mantle (w2800 km). Coherent patterns were not obtained
precursors to SS and PP (e.g., Flanagan and Shearer, 1998), short- in the MTZ. Note that the radial ows (orange for downward up to 5 cm/y and navy
period precursors (Collier et al., 2001), or differential travel-times for upward up to 5 cm/y) concentrate around the plate boundaries in the upper
between sScS and sScSSdS (reections from the discontinuities) mantle but diffuse (up to 2 cm/y) in the mid-mantle. The original gures were
provided by B. Schuberth.
recorded in the dense seismic networks in Japan (Tono et al., 2005).
Recent laboratory experiments determined the phase relations
in the peridotite-water system up to the top of the lower mantle
where dehydration may occur from subducting slabs. Dehydration difference in the Clapeyron slope for synthetic mantle minerals
is presumed because of the much smaller capacity of hydration of under dry (dehydrated) and wet (hydrous) conditions, and sug-
lower mantle minerals than minerals in the MTZ (e.g., Ohtani et al., gested a possibility of wet condition for subducted slabs in the
2004). Further, Ohtani and Litasov (2006) have shown clear regions where the seismically observed depression of the 660 km
 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93
discontinuity depth is large. Such a depression could not be
explained with cold temperature anomaly alone. On the other
hand, under a wet condition the majorite garnet transformation to
a high-pressure phase takes place at a lower pressure (Sano et al.,
2006). These changes of phase transformation pressures may
correspond to the seismic observations of discontinuity depth
variation between 660 and 690 km as described above (e.g., Tajima
et al., 2009). A possibility of viscosity reduction of garnet-rich
crustal material relative to the surrounding olivine-rich mantle is
also suggested under a wet condition in the MTZ (Katayama and
Karato, 2008). The rheological contrast between the crustal mate-
rial and the slab main body (peridotite) could take an effective role
to control slab dynamics.
Recently, Yoshida et al. (2012) developed a new three-
dimensional (3D) simulation model using a code with nite-
volume discretization on a staggered grid (Yoshida, 2008a, b), and
performed a series of numerical experiments focusing on viscosity
and density variations around the MTZ. These variations also pre-
sume to represent different geochemical properties. The numerical
experiments indicate substantial difference in the behavior of a
subducting plate between models under wet and dry conditions,
i.e., stagnant slab formation or slab penetration, and the distinct
role of the subducted crust in the MTZ. The subsequent study
expanded the simulation modeling by tracing crustal material
around the MTZ more quantitatively (Yoshida and Tajima, 2013).
In this paper we discuss the effect of deep water transport based
on the 3D numerical simulation that tested the implications or
models postulated for subducted slabs from geophysical observa-
tions and high pressure experiments.
2. Observations and experiments
2.1. Seismic properties associated with subducted plates
Seismic tomography studies have shown various behaviors of
subducted plates as they enter the MTZ (e.g., Fukao et al., 1992,
2001; Grand et al., 1997; Widiyantoro et al., 1999; Grand, 2002;
Ritsema et al., 2004). The slabs are lying at or piling up (forma-
tion of stagnant slabs) in the MTZ in a fairly broad region of the
NWP subduction zones while the slabs penetrate through the 660-
km discontinuity into the lower mantle in the Indonesian, Kerma-
dec and South American subduction zones. The subducting plates
are colder than the ambient mantle, and this thermal anomaly is
considered as the main driving force of subduction. When the
phase transformation prole of olivine-rich peridotite is referred to,
the phase transformation depths (i.e., seismic speed discontinu-
ities) within the subducted slabs should be different from those in
the surrounding mantle due to the thermal anomalies. The
discontinuity depths may be shallower than 410 km at the top of
the MTZ, and deeper than 660 km at the base of the MTZ.
A number of studies have investigated the global topography of
discontinuity depths or discontinuity depth variation associated
with stagnant slabs using various methods and data such as long-
period precursors to SS and PP (Flanagan and Shearer, 1998;
Shearer et al., 1999), travel times of short-period precursors to pP
and sP (Collier et al., 2001), or receiver function imaging with short-
period data from the high density seismic networks in Japan (Li
et al., 2000; Niu et al., 2005; Tonegawa et al., 2008). These
studies have shown that the discontinuity depths uctuate around
410 and 660 km. Collier et al. (2001) summarized that the wide
variability of the measured reectivities from 5% to 13%, which
implies the discontinuity thickness as small as 2e5 km. This
implication has caused problems when interpreting them in terms
of mantle composition, mineralogy and existing global seismic
speed models which suggest larger transition zone thickness or
81
smaller speed contrasts. One possible explanation is that the
strongest reections occur only under special conditions, possibly
where small-scale topography causes local focusing of seismic
waves.
Tono et al. (2005) used waveform data recorded by
w500 tiltmeters of the dense high-sensitivity seismograph
network, Hi-net and w60 F-net broadband stations in Japan (Okada
et al., 2004) and determined the discontinuity depth variation
beneath the Japan Islands in a unprecedented spatial resolution.
They analyzed the differential travel times between sScS and
sScSSdS (reections from the discontinuity (d) at 410 or 660 km)
and have shown that the 660-km discontinuity is dominantly
depressed beneath the dense seismic stations in southwestern
Japan. Beneath the northern part of the Japan Islands, where the
tomography does not show stagnant slab image, the topography
pattern is a mixture of depressed, normal, or elevated discontinuity
depths. The distribution of discontinuity depths suggests that the
transition from normal (w660 km) to depressed levels (w690 km)
takes place within a horizontal distance of w200 km. Assuming a
Clapeyron slope of 2.5 to 3 MPa/K (Ito and Takahashi, 1989), the
temperature beneath the attened slab is w200 K colder than the
surrounding mantle. This means that there is a lateral temperature
gradient larger than w1 K/km beneath the stagnant slabs. The rapid
lateral transition of the phase transformation depths may require
more reason than just the temperature anomaly disturbed by
subducted cold slabs.
There are studies that have been carried out to model triplicated
broadband body waveform data (0.03e1 Hz) recorded at regional
stations in the NWP rim (Tajima and Grand, 1995, 1998; Brudzinski
et al., 1997; Tajima et al., 1998; Wang and Chen, 2009; Wang and
Niu, 2010, 2011; Ye et al., 2011). Here, the waveform data were
recorded at regional distances (w14 to 30 ), turned in the MTZ and
thus strongly sampled the structure associated with the attened
HSSAs in the MTZ. The waveform modeling has resolving power of
attened structure and ability to identify speed anomalies distrib-
uted locally near the discontinuity depths that is in contrast to
long-wavelength tomography modeling alone (Tajima et al., 1998).
With body waveform data of more deep focus events
(magnitude  5.5; the occurrence period for 1990e2007), the MTZ
structure with stagnant slabs has been analyzed in detail in the
NWP subduction zones (Tajima et al., 2009). Results of these studies
show structural variations beneath the attened HSSAs, which are
largely represented by different layered models with HSSAs in the
MTZ with or without a depression of the 660 km discontinuity, in
spite of the 3D nature of subduction.
Fig. 2a shows the summary of structural variation along seismic
rays which are delineated by model M3.11, model M2.0 or other
(see Tajima et al., 2009). Here M3.11 structure is characterized with
HSSA of up to 3% relative to a standard model, iasp91 (Kennett
and Engdahl, 1991) in the deeper part of the MTZ (w525 to
660 km), and a discontinuity depth depression to 690 km (Fig. 2b).
This structure is suitable to represent a attened cold slab (olivine-
rich peridotite) predicted by mineral physics studies in the context
of phase transformation of ringwoodite into a higher-pressure
phase (perovskite magnesiowstite) under cold slab geotherm
(Ringwood and Irifune, 1988; Irifune and Ringwood, 1993;
Ringwood, 1994). The HSSA gradually decreases below 600 km
and disappears at w660 km. There is no HSSA between 660 and
690 km. The structure of model M2.0 also has HSSA like M3.11 in
the MTZ but is not accompanied by a discontinuity depth depres-
sion (see Tajima et al., 2009). Taken the prole of ringwoodite, the
lack of discontinuity depth depression in model M2.0 may be
interpreted with the temperature which is normal beneath the
attened slab. Other indicates the structure which is not repre-
sented by a layered model.
82 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93

Figure 2. Seismic observation associated with stagnant slabs in the NW Pacic subduction zone. (a) Summary of structural variations along seismic rays between deep focus events
and seismic stations in northeastern China. The structures along the rays are delineated by layered model M3.11 (solid navy) or M2.0 (dotted navy), or other which is not rep-
resented by a layered structure (solid black) (Tajima et al., 2009); (b) Models, M3.11, M2.0 and iasp91. The stagnant slab distribution in the deeper part of the MTZ is illustrated in the
inset (gray shading) approximately after the tomography model of Fukao et al. (2001). The red rectangles indicate segmented slabs, each of which has a slightly different subduction
direction from others. The portions drawn by dotted lines indicate the stagnated slabs. Model M2.0 has the same HSSA in the MTZ as Model M3.11 but the discontinuity depth is
normal, 660 km.

The subducted Pacic plate appears to segment when it enters
the MTZ (Tajima et al., 1998, 2009). The red rectangles denoted as
S1, S2, S3 and S4 in Fig. 2a illustrate the segmented slabs. The
subduction directions of these slabs rotate from northwest beneath
Sakhalin (S4) to nearly west beneath southwestern Japan (S1). The
seismic rays from the events in the Kurile and Izu-Bonin subduction
zones propagated to stations in northeastern China sampling these
slabs strongly. While most of the rays sampled the slabs obliquely
relative to the subduction directions, the rays from Izu-Bonin to
station BJT are subparallel to the subduction direction. These rays
sample the center of slab S1 and are clearly delineated with M3.11
structure. The M3.11 model data are rimmed by M2.0 or other
model data. The ray samplings may be too spotty to extract sys-
tematic patterns of discontinuity depth variation. Even so, the lo-
cations of these depressed discontinuity depths are compatible
with those determined beneath southwestern Japan by Tono et al.
(2005). The database and method of analysis used by Tono et al.
(2005) do not have resolution for the discontinuity features asso-
ciated with slabs S2, S3 and S4.
The small ellipses along the seismic rays in Fig. 2a show the
approximate locations of ray turning where seismic waves are most
sensitive to the structure. Light blue and navy ellipses indicate
different discontinuity depths of normal (660 km) and depressed
levels (w690 km), respectively. As noted above already, the rapid
lateral transition of the discontinuity depths from depression to
normal may require more reason than just the temperature change.
Some P waveforms of deep focus events show blurred triplication
arrivals or substantial later arrivals at certain stations even though
the source-time processes are simple. The black solid lines indicate
seismic rays which conveyed such anomalous P waveforms
(denoted as other model data) and propagated close to the M2.0
or M3.11 data (also Tajima and Nakagawa, 2006). The small gray
ellipses indicate the turning locations of rays. Fig. 3 shows some
examples of waveform data that strongly sampled the structure
beneath southwestern Japan and are modeled with (a) M3.11 or (b)
M2.0. P waveforms which propagated close to data in (a) or (b) but
could not be modeled with any layered model are shown in (c).
Fig. 4 illustrates how the triplicated seismic waveforms are
sensitive to the structure near the turning depths of the ray paths:
(a) ray paths from a deep event (depth w450 km) at distances of
w18 (D1) and w24 (D2); and (b) the calculated travel times for
iasp91, M2.0 and M3.11. The arrows indicate the distances for D1
and D2. At D1 the rst arriving phase turned in the MTZ, and the
secondary phase in the upper-most lower mantle, followed by the
waves that reected at the 660 km discontinuity. At D2 the rst
arriving phase refracted in the upper-most lower mantle, and the
secondary phase in the MTZ, followed by the waves that reected
at the 660 km discontinuity. Fig. 4b shows that the travel times
and time intervals between different phases are sensitive to the
models.
The blurred (fuzzy) or broadened waveform data of triplicated
phase arrivals (see the corresponding rays denoted as data of model
other in Fig. 2a, and the waveforms in Fig. 3c) may have been
produced by multiple path arrivals of the rst of the triplicated
phases at very local zones of low seismic speed anomaly (LSSA)
(illustrated with a water-thin red ellipse). The rst arrivals of these
 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93 83

Figure 3. Observed seismic waveforms for deep focus events in the Izu-Bonin subduction zone (top traces) that strongly sampled the structure in segment S1 are compared with
synthetics which were calculated with an optimum model (second from top) and other models (lower traces): (a) Data recorded at MDJ or BJT. The optimum model for these data is
M3.11; (b) Data recorded at MDJ or HIA. The optimum model for these data is M2.0; (c) Data recorded at HIA or BJT and do not agree with any synthetics calculated with a layered
model. The rays are denoted as other. Note that the time intervals of the rst and secondary arriving phases in the observed waveform agree well with those of the synthetic
waveforms calculated with the optimum models in (a) and (b). The triplication is not obvious in the waveforms of the event B19 and B26 data recorded at BJT. But the time intervals
between the rst and secondary arrivals in the synthetics calculated with model M3.11 agree with the widths of the wavelets.

event data (B1, B3 and B19 data recorded at HIA, and B19 and B26
data recorded at BJT) were refracted below the 660 km disconti-
nuity (branch CD). There are arrivals between branches CD and AB
suggesting multiple arrivals of waves at branch CD due to LSSA.
These data have not been modeled well but should not be dumped
into a group of poor-quality data as they contain information of
great account for tracing uids. We will discuss this issue later
in Discussion for conjecture of deep water transport, together with
dehydration induced uids at the base of the MTZ.
2.2. Implications from high pressure experiments
The compositional properties in the MTZ estimated by seismic
studies could be checked and validated through synthetic experi-
ments under high pressure (P) and temperature (T) conditions.
Recent studies suggest that inclusion of hydroxyls into solid
mineral phases, coexisting hydrous uids and dehydration induced
melts may affect various aspects in chemical and physical proper-
ties in different scales (e.g., Inoue et al., 1998; Jacobson et al., 2004;
Irifune et al., 2008), phase transformation kinetics (Ohtani and
Litasov, 2006), and rheology (Katayama and Karato, 2008).
Experimental results on high water solubility in wadsleyite and
ringwoodite up to 2e3 wt.% H2O suggest that the MTZ can be an
important water reservoir of the Earth (Inoue et al., 1995; Bercovici
and Karato, 2003; Ohtani et al., 2004; Ohtani, 2005; Litasov et al.,
2006; Ohtani and Litasov, 2006). Ohtanis group carried out a sys-
tematic series of high pressure experiments to determine the phase
relations of hydrous mantle minerals up to the uppermost lower
mantle conditions (Ohtani et al., 2004), and suggests four potential
sites of dehydration along the subducting plate, i.e., the mantle
wedge of the subduction zone, the MTZ, the uppermost lower
mantle and in the lower mantle around 1200e1500 km depths. The
84 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93

Figure 4. (a) Illustration of triplicated waves from a deep focus event that propagate to stations D1 (<20 ) and D2 (>20 ). The water-thin red ellipse near the 660 (or 690) km
discontinuity illustrates the LSSA zone captured by the waveforms. See the main text for details. (b) Reduced P-wave travel times from a deep event (depth w450 km) as a function
of propagation distance calculated for models M3.11 (solid navy), M2.0 (dotted navy) and iasp91 (solid green). The arrows 1 and 2 point the locations of D1 and D2 in the travel time
curves. Branch AeB represents the travel time of waves from MTZ, branch B-C the travel time of waves reected at the discontinuity, and branch CeD the travel time refracted
beneath the discontinuity. At D1 the rst arriving waves are from MTZ, the secondary arrivals from beneath the discontinuity, and the third ones reected at the discontinuity if
model M3.11 or M2.0 is adopted. At D2 the order of specic phase arrivals is reversed. Thus, the triplicated waves are very sensitive to the structure near the discontinuity depth, and
can capture associated anomalies well.

phase diagram of peridotite with 2 wt.% water indicates that there
may be dehydration induced uids or melts at the base of MTZ
during the phase transformation to higher pressure phases.
The phase diagrams derived for peridotite (olivine rich) and
basalt (garnet rich) have shown a clear difference in the Clapeyron
slopes between dry and wet conditions, and the stability elds of
wadsleyite (Wd) and ringwoodite (Rw) (high P phases of olivine)
expand under wet and low temperature conditions (Ohtani and
Litasov, 2006). The Clapeyron slope of Rw becomes steeper under
wet condition, and its capacity of water content increases under
colder temperature. The results suggest that the seismic observa-
tions regarding the depressed 660 km discontinuity, such as those
observed in the NWP subduction zones, cannot be explained with
cold temperature alone under dry condition but indicate a wet
condition. In contrast the post garnet (majorite) transformation to
perovskite shifts to lower pressures under wet condition than the
transformation pressure under dry condition (Litasov et al., 2005;
Ohtani and Litasov, 2006; Sano et al., 2006).
These are important implications that provide support for a
hypothesis of variable geochemical properties at the MTZ base by
many seismic studies. There is also controversy regarding the vis-
cosity variation in the MTZ and the uppermost lower mantle.
However, the behaviors of subducting slabs of different
geochemical properties and/or in the surrounding mantle of vis-
cosity variation have not been understood well in the scheme of
plate subduction dynamics. Katayama and Karato (2008) carried
out a high pressure experiment to compare the rheological prop-
erties between olivine and garnet with water effects, and suggested
weaker garnet than olivine under wet conditions in the MTZ. The
weaker garnet which is also denser than olivine in the MTZ
(Ringwood and Irifune, 1988) can ow and descend faster than
olivine-rich material. Accordingly zones of M2.0 may be formed
adjacent to zones of M3.11 structure at the base of MTZ (Tajima
et al., 2009). The lateral heterogeneity with these zones was
observed from seismic waveforms.
2.3. Other geophysical observations
Global scale observations show varying features of electrical
conductivities from region to region associated with stagnant slabs
in the MTZ (Kelbert et al., 2009; Utada et al., 2009). The conduc-
tivities determined for the MTZ in the NWP subduction zones are
high under water-rich condition caused by the deep water trans-
port with cold slabs while the conductivities beneath Europe are
relatively low due to dry stagnant slabs. These studies have pre-
sented intriguing but somewhat vague interpretations in terms of
 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93 85

temperature and water content because the spatial resolutions are
limited. Both of the regions show cold stagnant slabs (HSSAs) in the
MTZ in seismic tomography models. Much of the hazy interpreta-
tion arises from our poor understanding of water (or uids) dis-
tribution in the MTZ and the mechanisms of hydration and
dehydration associated with subduction process.
Here, the higher conductivities determined for the MTZ in the
NWP subduction zones than those beneath Europe (Kelbert et al.,
2009; Utada et al., 2009) do not necessarily require homogeneous
distribution of water in a broad region of MTZ. Instead, the con-
ductivity of the LSSA zone that has a width of less than w50 km of
uids could be w15 times as high as the conductivity of the sur-
rounding mantle with stagnant slab. The uids in the narrow zone
can increase the conductivity for a broad region with a width of
over 500 km in the MTZ. This situation may represent the con-
ductivity in the NWP subduction zone in comparison with that
beneath Europe. Here, w50 km is comparable to the diameter of
Fresnel zone for a P wave propagation which bottoms at the base of
the MTZ at w1 Hz.
The inference of mantle viscosity from surface observations
has been based on a number of inversion studies of geodynamic
data (e.g., Forte and Peltier, 1991; Ricard and Wuming, 1991;
King and Masters, 1992; Forte et al., 2010). The viscosity pro-
les determined by inversions are all characterized by strong
overall increases in viscosity with depth although there are
signicant differences in the detailed characters of these vis-
cosity models. Postglacial rebound studies have suggested that
no signicant viscosity decrease occurs at the 410-km seismic
discontinuity. Rather, there may be a viscosity increase with
depth in the MTZ, likely due to pressure dependence (Peltier,
1998). On the other hand, a low-viscosity layer (LVL) just
beneath the MTZ has been suggested from geoid inversion

studies (Cadek 
et al., 1997; Kido and Cadek, 1997; Kido and Yuen,

2000). Kido and Cadek (1997) admitted that the resolution of
inversion suffers from severe non-uniqueness and was not suf-
cient to determine the depth and absolute value of the low
viscosity channel. The LVL was considered to be of thermal ori-
ginda thermal barrier at the 660-km boundary arising from the
negative Clapeyron slope induces a high-temperature thermal
boundary layer beneath the MTZ.
3. Numerical simulation of slab dynamics
There are some pioneering studies in numerical simulation that
focused on the interactions between subducting plates and the
mantle phase transition boundaries (Christensen and Yuen, 1985;
Davies, 1995; Zhong and Gurnis, 1995; Christensen, 1996). Recent
studies have carried out simulation of plate interactions with the
phase boundaries in a 2D model space and examined conditions to

produce stagnant slabs in the MTZ (e.g., C zkov et al., 2002, 2007;
Tagawa et al., 2007a, b; Torii and Yoshioka, 2007; Yoshioka and
Naganoda, 2010). As mentioned above, the existence and inu-
ence of an LVL beneath the MTZ has been also suggested by several
studies. Yoshioka and Naganoda (2010) concluded that a subduct-
ing slab tends to stagnate if an LVL exists just below the 660-km
boundary because viscous resistance force between the slab and
the LVL is reduced there. Some studies have suggested that the
secondary upwelling plumes originated from the MTZ can be
attributed to the LVL (i.e., second asthenosphere) and the inu-
ence of the second asthenosphere is important in inducing con-
vection layering at 660 km and the mid-mantle plumes without a
root in the deep lower mantle (Cserepes and Yuen, 1997, 2000;
Cserepes et al., 2000). These studies have indicated that viscosity
(rheological) heterogeneity in the MTZ or contrast between the
subducting plates and the surrounding mantle may have a primary
effect to control the behavior of subducting plates.
More recently, Yoshida et al. (2012) carried out a systematic
study of 3D numerical simulation and have shown a substantial
difference in the behaviors of the subducting plate and the trace
of crustal material between the models under wet (hydrous) and
dry (dehydrated) conditions in the MTZ. Yoshida and Tajima
(2013) demonstrated the different behaviors of the subducting
plate and the trace of crustal material in the numerical experi-
ments between the models with and without an underlain LVL
beneath the MTZ. They found that a large amount of the crustal
material is temporarily stored in the MTZ, but begins to penetrate
into the lower mantle within 10 My after the beginning of
stagnation. The volume of trapped crustal material relative to the
total subducted crust and the stagnation time in the MTZ were
examined quantitatively to see if the crustal material could
accumulate to make a garnet enriched MTZ or may be recycled in

Figure 5. (a) Overview of the computational domain. The subducting plate is shown with an overlain oceanic crustal layer (green) and harzburgite layer (orange). The velocities of
subducting and overriding plates are assumed to be Vss 8 cm y1 and Vos 3 cm y1, respectively, and the speed of trench retreat is 3 cm y1 (Yoshida and Tajima, 2013; Yoshida,
2013). The light blue isosurface shows the temperature of 1000  C. The transparent light gray isosurfaces marked as p1 to p3 indicate the depths of the olivine to wadsleyite
(Ol / Wd), wadsleyite to ringwoodite (Wd / Rw) phase transition boundaries, and the ringwoodite to perovskite magnesiowstite (Rw / Pv Mw) phase decomposition
boundary, respectively. The boundary conditions for the top, bottom, and side boundaries of the computational domain and the details of the boundary region are described in the
previous paper by Yoshida and Tajima (2013). (b) Proles of the densities of the mantle (rm, solid line) and the oceanic crust (rc, dotted line) used in the present model.
86 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93

mantle convection. These issues are important to settle the vT

debate regarding the fate of subducted oceanic crustal material v$VT  V2 T 0 (3)
vt
(e.g., Ringwood and Irifune, 1988).
vCj
3.1. Model setup v$VCj 0 (4)
vt

The computational domain of mantle convection, which is where v represents the velocity; p, the dynamic pressure; h, the
numerically modeled as an incompressible Boussinesq uid with viscosity; t, the time; T, the temperature; Gi, the phase function
an innite Prandtl number, is conned to a 3D regional spherical- (0  Gi  1); Cj, the composition (0  Cj  1); i, the index of each
shell in the spherical polar coordinates (r, q, 4) with a thickness of phase in the mantle; j, the index of each compositionally different
1000 km and a lateral extent of 10  60 in the q and 4 directions material from the mantle; and er, the unit vector in the radial di-
(Fig. 5a). The ConvRS, a numerical simulation code based on a rection. The dimensionless parameters are the thermal Rayleigh
staggered grid, particle-in-cell, nite-volume (FV) method (e.g., number Ra, the phase Rayleigh number Raph(i), the compositional
Yoshida, 2010; Yoshida et al., 2012), is used to simulate mantle Rayleigh number Rach(j), and the mantle/shell radius ratio z:
convection. A semi-dynamic model for the subduction zone is
r0 a0 DTgb3 Drphi Drchj b
applied to realize plate subduction. The numbers of FVs used are Rah ; Raphi h Ra; Rachj h Ra; zh
128 (in r)  128 (in q)  768 (in 4). The model setup in this study is h0 k0 r0 a0 DT r0 a0 DT r1
essentially the same as that used in a previous study (Yoshida, (5)
2013; Yoshida and Tajima, 2013), and hence, it is only described
where the meanings and values of the other symbols are listed in
briey below.
Table 1.
The dimensionless conservation equations for mass, mo-
The olivine to wadsleyite (Ol / Wd) phase transformation at
mentum, and energy, which govern mantle convection under the
the reference depth of 410 km, the wadsleyite to ringwoodite
Boussinesq approximation, and the advection equation for the
(Wd / Rw) transformation at 520 km, and the Rw / Pv Mw
composition are expressed respectively as:
phase decomposition at 690 km are imposed in the model. The
V$v 0 (1) garnet in the oceanic crust is assumed to be 200 kg m3 denser than
that of the depleted peridotite (i.e., surrounding mantle material).
2 The density increase due to the garnet to post-garnet phase

  X
Nph
transformation is assumed to occur at a depth of 660 km under
Vp V$ h Vv Vv tr
4RaaT  Raphi Gi
hydrous conditions (Fig. 5b). In contrast, the density of harzburgite
i1
3 layer is assumed to be the same as that of the surrounding mantle.
X
Nch
3
The dimensionless coefcient of thermal expansion of the mantle is
Rachj Cj 5z er 0 2 considered to be depth-dependent (see Fig. 2c in Yoshida and
j1
Tajima, 2013), and is given by

Table 1
Parameters used in the numerical modeling.

Symbol Denition Value Units

g Gravitational acceleration 9.8 m s2
r0 Reference density of the mantle 3300 kg m3
a0 Reference thermal expansivity of the mantle 3.0  105 K1
Tm Reference mantle temperature 1573 K
DT Temperature difference across the model 1300 K
r1 Earths radius 6371 km
b Thickness of the model domain 1000 km
k0 Reference thermal diffusivity of the mantle 106 m2 s1
h0 Reference viscosity of the upper mantle 1020 Pa s
cp0 Reference specic heat at constant pressure of the mantle 1250 J kg1 K1
dT AndersoneGrneisen parameter for MgSiO3 perovskite 6.5a) e
gi Clapeyron slope of the Ol / Wd (i 1) and Wd / Rw (i 2) phase transitions and the 3.9b), 5.0c), 2.0d) MPa K1
Rw / Pv Mw (i 3) phase decomposition
Drph(i) Density increases at the Ol / Wd (i 1), Wd / Rw (i 2), and Rw / Mw Pv (i 3) 231, 99, 330e) kg m3
phase boundaries
Drch(j) Density difference between the crustal material (j 1) and the surrounding mantle, and the See Fig. 5b kg m3
harzburgite layer (j 2) and the surrounding mantle
Dhlm Viscosity contrast between the Pv Mw phase and the reference upper mantle 100 e
Dhoc_wet Viscosity ratio between the crustal material and the surrounding mantle below 103 e
the Ol / Wd phase boundary
Dhhrz Viscosity ratio between the harzburgite layer and the surrounding mantle 300
DhLVL Viscosity ratio between the LVL and the surrounding mantle 101
Dimensionless parameters
Ra Thermal Rayleigh number 1.26  107 e
Raph(i) Phase Rayleigh numbers for the Ol / Wd (i 1) and Wd / Rw (i 2) phase transitions 2.26  107, 9.70  106, 3.23  107 e
and the Rw / Pv Mw (i 3) phase decomposition
Rach(j) Compositional Rayleigh number for crustal material (j 1) and the harzburgite layer (j 2). Refer to Eq. (5) and Fig. 5b e
z Mantle/shell radius ratio 0.16 e
E Activation parameter 11.5 e

References: a) Katsura et al. (2009), b) Litasov et al. (2006), c) Katsura and Ito (1989), d) Litasov et al. (2005, 2006), e) Schubert et al. (2001).
 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93 87

Figure 6. Time evolution of the temperature eld (color contours) and velocity eld (arrows), and the distribution of crustal (green above Ol / Wd phase boundary to purple under
the boundary) and harzburgite (orange) layers along the cross-section q 86.25 (see Fig. 5a) for models without the low-viscosity layer (LVL) beneath the mantle transition zone.
Only the part of the model between q 160 and 200 is displayed. The black line contour intervals for temperature are 200  C. Three yellow lines in each panel show the Ol / Wd,
Wd / Rw, and Rw / Pv Mw phase transition/decomposition boundaries at depths of 410 km, 520 km, and 690 km, respectively. The dashed black lines in each panel show the
garnet to post-garnet phase transition boundary at a depth of 660 km.

 dT where Tm (1) is the dimensionless reference mantle temperature,

rm d
ad (6) E is the dimensionless activation parameter that controls the
r0 magnitude of the viscosity contrast in the modeled mantle, Gi is the
where rm(d) is the density as a function of depth associated with dimensionless phase function (0  Gi  1), and DhGi is the viscosity
the mantle phase transformations/decomposition (see Fig. 5b), and ratio between each phase and the reference upper mantle (olivine
dT is the AndersoneGrneisen parameter (see Table 1). phase).
For the mantle, a dimensionless Arrhenius form is used for the In the present study, E is xed at ln(105) 11.5, which permits a
temperature-dependent and phase-dependent viscosity: ve order viscosity contrast between the top surface boundary and
  the mantle (Yoshioka and Naganoda, 2010), DhG(i1,2) is 1 for the
  2E E wadsleyite and ringwoodite phases and DhG(i3) is set to be Dhlm for
hT; Gi exp Gi $ln DhGi $exp  (7)
T Tm Tm the Pv Mw phase (i.e., lower mantle), where Dhlm is the factor of
88 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93
viscosity increase at the Rw / Pv Mw phase boundary and is
xed as 100 (Yoshida, 2013; Yoshida and Tajima, 2013).
In some models the LVL just beneath the MTZ is considered. The
viscosity in the LVL with a thickness of hLVL is replaced by:
h DhLVL $hT; Gi (8)
where the viscosity contrast between the LVL and the surrounding
mantle DhLVL is xed to be 101 (Cserepes and Yuen, 1997, 2000;
Cserepes et al., 2000), and hLVL is taken as a free parameter
(50e100 km; see Section 3.2).
The subducting plate in the present model is composed of three
layers: oceanic crust, harzburgite, and depleted peridotite (Fig. 5a).
The oceanic crust has an initial thickness of w7.8 km, and its initial
viscosity is dened by Eq. (7). On the other hand, variations in water
content can have considerable effects on the rheological properties
of garnet-rich crustal material (Katayama and Karato, 2008). To
account for the effects of viscosity reduction under hydrous con-
ditions (Bercovici and Karato, 2003; Ohtani et al., 2004), the vis-
cosity of the oceanic crust below the Ol / Wd phase boundary is
replaced by
hT; Gi ; C1 DhCoc_wet
1
$hT; Gi (9)
where the ratio of viscosity contrast between the oceanic crustal
layer and the surrounding mantle, Dhoc_wet, is xed at 103 (Tajima
et al., 2009; Yoshida and Tajima, 2013), and C1 represents a
dimensionless composition for the oceanic crustal material
(0  C1  1).
The viscosity in the harzburgite layer with an initial thickness of
w31.3 km is replaced by
hT; Gi ; C2 DhChrz
2
$hT; Gi (10)
where the viscosity ratio between the harzburgite layer and the
surrounding mantle, Dhhrz, is taken as a free parameter that varies
from 1 to 300, and C2 represents the composition for the harz-
burgite material (0  C2  1). From the previous work by Yoshida
(2013), we found that for a larger Dhhrz, the harzburgite layer of
high viscosity makes the subduction angle to decrease because of
its strength, and thus, the subducting plate tends to easily stagnate
in the MTZ. In the present study, a xed value of 300 is used for
Dhhrz.
3.2. Results of numerical experiments
First the temporal evolution of temperature and velocity elds
along the cross-section of q 86.25 (see Fig. 5a) is shown for the
model without an LVL beneath the MTZ, i.e., DhLVL 1 in Fig. 6. The
green tracers indicate the initial crustal layer above the Ol / Wd
phase boundary at 410 km, and the purple tracers the distribution
of the crustal layer below the 410-km phase boundary where the
viscosity reduction takes place under a hydrous condition (Eq. (9)).
The orange tracers show the harzburgite layer.
Fig. 6 shows that the subducting plate begins to stagnate in the
MTZ within an elapsed time of 10 My from the start of simulation.
The lateral extent of the stagnated plate at the base of the MTZ
is  500 km (Fig. 6b). This conguration of the subducting plate is
similar to the stagnant slabs which were observed seismically
beneath the NWP subduction zones (e.g., van der Hilst et al., 1991,
1997; Widiyantoro et al., 1999; Fukao et al., 2001; Grand, 2002;
Zhao, 2004).
However, this phenomenon is rather transient, and the stagnant
slab begins to penetrate into the lower mantle within 10 Myr after
the formation of the stagnant slab (Fig. 6c). A substantial amount of
crustal material starts sinking into the lower mantle along with the
subducting slab main body in less than 20 Myr. A buckled crust
coupled with the harzburgite layer is observed at the heel of
stagnant slab (Fig. 6c and d). This buckling was caused by the
viscous resistance force which the harzburgite layer felt at the top
boundary of the high-viscosity lower mantle. The deformation of
the crustal layer produced here may correspond to seismological
observations which suggest remnant subducted crustal materials in
the uppermost lower mantle beneath the Mariana subduction zone
(Niu et al., 2003). A small amount of crustal and harzburgite ma-
terials is trapped in the MTZ for at least over several My which is
still shorter than the transit time of mantle convection (w200 Myr).
The transiently accumulated crustal materials at the base of MTZ
may affect the thermal and compositional circulation in the Earths
mantle.
Next, we investigate the effects of the LVL beneath the
Rw / Pv Mw phase boundary on the subducted slab behavior
and crustal material traces. Fig. 7 shows the temporal evolution of
temperature and velocity elds for the model with hLVL 100 km
along the same cross section as Fig. 6. As suggested in our previous
2D and 3D models (Yoshioka and Naganoda, 2010; Yoshida and
Tajima, 2013), subducting slabs tend to laterally spread in the
MTZ because the resistance force between the slab and the sur-
rounding mantle is reduced in the LVL. After the stagnation starts,
the subducting slab behavior and the crustal material traces are
largely different from the model without an LVL beneath the phase
boundary depth (cf. Figs. 6 and 7). The lateral extent of the attened
plate in the MTZ (Fig. 7c) is more than twice than in the model
without the LVL (Fig. 6c). Nonetheless, a signicant amount of
crustal material starts sinking into the lower mantle along with the
subducting slab main body within an elapsed time of 20e30 Myr,
and a buckling is also observed (Fig. 7d) as was in the model
without the LVL (Fig. 6d).
On the other hand seismic waveform analyses suggest that the
LSSA zones may be very local and thin, much thinner than 100 km.
Thus, we tested another model with a thinner LVL, hLVL 50 km.
Fig. 8 shows that the subducting slab still extends and spreads
laterally and a large amount of crustal and harzburgite materials is
transiently trapped in the MTZ (Fig. 8c). Then, crustal materials
start sinking into the lower mantle within an elapsed time of
20e30 Myr (Fig. 8d) as was observed in the model with hLVL of
100 km (Fig. 7). This result implies that the LVL beneath the MTZ
drastically changes the morphology of subducting slabs and the
traces of crustal materials no matter how thick it is. However, after
the stagnant slab is formed, its behavior seems to be quite sensitive
to the thickness of an LVL (cf. Figs. 7 and 8). The smaller hLVL is, the
closer the slab behavior becomes to the model without an LVL
(Fig. 6).
In summary the results of this numerical study suggest that the
LVL beneath the MTZ has a great impact on the morphology of
stagnant slab and the behavior of the crustal materials in the MTZ.
However, a large amount of crustal materials penetrates into the
lower mantle within 20e30 Myr after the start of the plate sub-
duction and the buckled crustal layer coupled with the harzburgite
layer is observed at the heel of stagnant slab regardless of the ex-
istence of an LVL.
4. Discussion for deep water transport
The new class of mantle convection models of pyrolite compo-
sition can be comparable with long-wavelength seismic tomogra-
phy models. The long-wavelength features (over 1000 km) in
seismic tomography models are consistent with the accounts of
large-scale mantle convection, in which the primary driving force is
the gravitational instability caused by thermal anomalies
(Schuberth et al., 2009a). In relatively short-wavelength features,
 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93 89

Figure 7. Same as Fig. 6 for the model with a 100 km thick LVL beneath the MTZ.

there is still substantial variation among published models, espe-
cially in the MTZ (e.g., Becker and Boschi, 2002; Grand, 2002;
Ritsema et al., 2004). The intricate characteristics in the MTZ,
which may affect mantle dynamics, cannot be explained with
thermal anomalies alone (e.g., Schuberth et al., 2009b). The fate of
subducted crustal material in the MTZ and deeper mantle is still a
matter of debate and the separation of crustal materials from the
main body of subducting lithosphere that could cause garnet
enriched MTZ has long been debated (e.g., Anderson, 1979;
Ringwood and Irifune, 1988; Ringwood, 1994).
The less coherent images of short-wavelength features in to-
mography models lead to a variety of implications and dynamic
models. There are various factors which may explain the less satis-
factory agreement between 3D seismic models of mantle structure
and the reconstructions of mantle heterogeneity in terms of sub-
ducted plates (Forte, 2007). First, the evolution of the slab trajectories
was not determined in a uid mechanically consistent manner. Sec-
ond, the slab models in numerical experiments assume that ther-
mally generated heterogeneities dominate the mantle, which may
not be applicable in regions of the mantle with signicant composi-
tional heterogeneity (e.g., Forte and Mitrovica, 2001; McNamara and
Zhong, 2005). These correspond to our argument that the viscosity
contrast between the crust and slab main body under a wet (hydrous)
condition had not been considered in numerical experiments.
90 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93

Figure 8. Same as Fig. 6 for the model with a 50 km thick LVL beneath the MTZ.

Using seismological observations Kawakatsu and Watada (2007)
determined a channel of hydrated mantle material on top of the
subducting plate in northeastern Japan up to a depth of w170 km.
They suggested that it is a pathway for water released from hydrous
oceanic crust and reported no such feature is observed below the
depth associated with the subducting slab. The hydration along the
subducting plate at this stage should be the result of dehydration
mostly from grain boundaries of slab minerals. Deeper in the MTZ
the dehydration may be mainly from mantle mineral crystals.
Recent P travel-time tomography models for East Asia addressed
the origin of intraplate volcanic activities with a link to the stagnant
Pacic slab that transported water deep into the MTZ (Wei et al.,
2012; Zhao and Tian, 2013).
The other seismological studies observed variation of the
660 km discontinuity depth and some anomalous waveforms
which sampled the structure near the base of the MTZ and postu-
lated a hypothesis of variation of geochemical compositions and
highly localized uids near the phase transformation depth (e.g.,
Tajima et al., 2009). These can be the consequence of deep water-
transport by subducted slabs and dehydration at the phase trans-
formation depth. At depths below the MTZ (approximately 800 to
1200 km) some studies found anomalous reectors/scatterers of
seismic waves beneath the western Pacic and Indonesian sub-
duction zones, and discussed that the scatterers would be most
likely basalt fragments in the subducted oceanic crust (Kaneshima
and Helffrich, 2003, 2010; Niu et al., 2003; Kaneshima, 2009).
 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93
Niu et al. (2003) also found seismic properties which may indicate
buckling of crustal material behind the root of stagnated slab in the
uppermost lower mantle. A recent study provided evidence for the
buckling of subducted oceanic lithosphere through kinematic
thermal structure analysis and mineral physical calculations (Zhang
et al., 2013).
Our previous numerical studies attempted to test the hypothesis
using a 3D model without a high-viscosity harzburgite layer and
showed that bending of the crustal layer occurs only when an LVL
beneath the MTZ is considered (Yoshida and Tajima, 2013). A recent
high pressure experiment derived models of velocity and density
signatures for the stagnant slab under China using the phase re-
lations measured for harzburgite combined with seismic models,
and found that a buckled slab model yields velocity anomalies (1
to 2% for Vp) (Zhang et al., 2013). The present study used a model
with rheological contrasts between the crust, a high-viscosity
harzburgite layer, the lithosphere main body, and surrounding
mantle assuming a wet condition in the MTZ. The result shows that
the crustal layer bends in the model even without an LVL beneath
the MTZ. It also shows that when such an LVL is not considered
beneath the MTZ, only a small amount of subducted crustal mate-
rial is trapped in the MTZ. But when the LVL is considered, a sig-
nicant volume of crustal material is transiently trapped in the
MTZ. However, the once trapped crustal materials sink into the
lower mantle within 20e30 Myr from the start of the plate sub-
duction, a relatively short time as compared with the whole mantle
convection cycle (up to w200 Myr). The results of this study sug-
gest recycling of crustal material in the whole mantle, which is
consistent with the evidence from mantle geochemistry (e.g.,
Albarde, 1998; Coltice and Ricard, 1999) as opposed to two-layer
mantle convection models (e.g., ONions and Oxburgh, 1983;
Allgre et al., 1986; Allgre, 1987) (see reviews by Kaneoka, 1995,
2008; Hofmann, 1997; Tackley, 2000, 2007).
High pressure experiments have determined high water ca-
pacity (w2 wt.%) of olivine in the MTZ and suggested that the MTZ
can serve as a water absorber for the mantle scale water circulation
system (e.g., Inoue et al., 1995; Bercovici and Karato, 2003; Ohtani
et al., 2004). Recently many hydrous minerals, such as dense hy-
drous magnesium silicates, have been synthesized at high P and
high T lower mantle conditions and tested for the stability elds.
Sano et al. (2008) reported their discovery of the stability eld of
the aluminous hydrous mineral d-ALOOH at pressures from 33 to
134 GPa at 1350 to 2300 K. This implies that this hydrous mineral in
a subducting slab could transport hydrogen deep into the lower
mantle. The experimental results favor our simulation results that
the crustal materials descend into the lower mantle in a relatively
short time after stagnation in the MTZ.
On the other hand, the hydrogen (H) diffusivities measured for
wadsleyite and ringwoodite are very small, i.e., one billion years are
necessary to diffuse H 10s km (Hae et al., 2006; Ohtani and Zhao,
2009). This implies that the distributions of uids associated with
hydration/dehydration of stagnant slabs are highly localized, possibly
with a dimension of less than 1 km because the duration time of slab
stagnation in the MTZ is typically less than 20 Myr. Accordingly the
affected zones from hydration/dehydration that surround stagnant
slabs may be thin. Such local zones of LSSA embedded in HSSAs may
not be necessarily captured in tomography studies.
In the numerical simulation the viscosity reduction is applied to
the entire crustal layer because the spatial discretization is set to be
w8 km. But the actual zone of viscosity reduction due to dehy-
dration/hydration process can be much smaller because of the very
small diffusivity rate of wadsleyite and ringwoodite. Even if water is
transported deep into the mantle by various hydrous minerals in
the subducting plate, water dehydrated from the stagnant slab
seems to be distributed in a very local zone. Nonetheless the thin
91
coat of reduced viscosity for the subducted slab should effectively
lubricate the slab stagnation and subduction as was tested for
shallow subduction process (Tagawa et al., 2007b).
Our previous 3D numerical experiments produced a model of a
subducted plate (stagnant slab) which consists of the slab main body
(peridotite) abutted by subducted crustal materials near the base of
the MTZ. Under wet conditions the subducting plate tends to stagnate
with a maximum lateral extent of over several hundred kilometers.
The weaker and denser crustal materials (garnetite) sink faster than
the plate main body (peridotite) that leads to plate segmentation in
the MTZ (Tajima et al., 2009) (see Fig. 3 in Yoshida et al., 2012). The
widths of segmented stagnant slabs are several hundred kilometers
as are observed for the NWP subduction zones (see Fig. 2a).
The higher electrical conductivities determined for the MTZ
associated with the NWP subduction zones than those beneath
Europe (Kelbert et al., 2009; Utada et al., 2009) do not necessarily
require homogeneous distribution of uids in the broad region of
MTZ. Instead, the LSSA zone (i.e., with uids) of a width (W) of less
than 50 km could raise the conductivity of the MTZ determined for
a broad region (W w several hundred kilometers of a segmented
slab) beneath the NWP rim by more than an order of magnitude.
Here, the diameter of Fresnel zone along the rays (w1 Hz) from
deep events in the Kuriles or Izu-Bonin to BJI or HIA in northeast
China is about 50 km.
5. Conclusions
We carried out a series of 3D numerical experiments to test the
hypothesis of deep water transport by a subducted plate and the
fate of crustal materials based on seismological observations from
the NWP subduction zones. In the present study we used a model
with a high-viscosity harzburgite layer in addition to the rheolog-
ical contrasts between the crust, the lithosphere main body and
surrounding mantle assuming a wet condition.
(1) The crustal layer bends to stagnate in the MTZ even in the
model which does not include an LVL beneath the MTZ.
(2) When an LVL is not considered beneath the MTZ, only a small
amount of subducted crustal materials is trapped in the MTZ.
When the LVL is considered, a signicant volume of crustal
materials is transiently trapped in the MTZ.
(3) The once trapped crustal materials sink into the lower mantle
within 20e30 Myr from the start of the plate subduction, a
relatively short time as compared with the whole mantle
convection cycle (up to w200 Myr).
(4) The results suggest the recycling of crustal materials in the
whole mantle, which is consistent with the evidence from
mantle geochemistry as opposed to two-layer mantle convec-
tion models.
(5) The distributions of uids associated with hydration/dehydra-
tion of stagnant slabs in the MTZ should be highly localized
because the H diffusivity is very small and the duration time
of slab stagnation is typically less than 20 Myr.
(6) Such local zones of LSSA embedded in HSSAs associated with
stagnant slabs may not be necessarily captured in tomography
studies.
Acknowledgments
We appreciate B. Schuberth for providing us with the original
gures of his mantle convection model for Fig. 1, and I. Katayama for
discussion at an early stage of this study. We also thank M. Santosh,
Co Editor-in-Chief, and the anonymous reviewer for helpful com-
ments on the manuscript. Some gures were produced using
Generic Mapping Tools (GMT) (Wessel and Smith, 1998). The
92 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93
calculations presented herein were performed on the supercom-
puter facilities (SGI ICE X) at JAMSTEC. M.Y. was supported partly by
a Grant-in-Aid for Scientic Research (B) (Grant Number 23340132)
from the Ministry of Education, Culture, Sports, Science and Tech-
nology (MEXT), Japan.
References
Albarde, F., 1998. Time-dependent models of UeTheHe and KeAr evolution and
the layering of mantle convection. Chemical Geology 145 (3e4), 413e429.
http://dx.doi.org/10.1016/S0009-2541(97).
Allgre, C.J., 1987. Isotope geodynamics. Earth and Planetary Science Letters 86
(2e4), 175e203. http://dx.doi.org/10.1016/0012-821X(87)90220-2.
Allgre, C.J., Staudacher, Y., Sarda, P., 1986. Rare gas systematics: formation of the
atmosphere, evolution and structure of the mantle. Earth and Planetary Science
Letters 81 (2e3), 127e150. http://dx.doi.org/10.1016/0012-821X(87)90151-8.
Anderson, D.L., 1979. The upper mantle transition region: Eclogite? Geophysical
Research Letters 6, 433e436.
Becker, T.W., Boschi, L., 2002. A comparison of tomographic and geodynamic mantle
models. Geochemistry, Geophysics, Geosystems 3. http://dx.doi.org/10.129/
2001GC000168.
Bercovici, D., Karato, S., 2003. Whole-mantle convection and transition-zone water
lter. Nature 425, 39e44.
Brudzinski, M.R., Chen, W.-P., Nowack, R.I., Huang, B.-S., 1997. Variations of P wave
speeds in the mantle transition zone beneath the northern Philippine Sea.
Journal of Geophysical Research 102, 11815e11827.
Bunge, H.P., Richards, M., 1996. The origin of large-scale structure in mantle con-
vection: effects of plate motions and viscosity stratication. Geophysical
Research Letters 23, 2987e2990.
Bunge, H.P., Richards, M., Lithgow-Bertelloni, C., Baumgardner, J., Grand, S.,
Romanowicz, B., 1998. Time scales and heterogeneous structure in geodynamic
Earth model. Science 280, 91e95. http://dx.doi.org/10.1126/science.280.5360.91.

Cadek, 
O., Czkov, H., Yuen, D.A., 1997. Can long-wavelength dynamical signatures
be compatible with layered mantle convection? Geophysical Research Letters
24 (16), 2091e2094. http://dx.doi.org/10.1029/97GL02054.

Czkov, H., van Hunen, J., van den Berg, A., 2007. Stress distribution within sub-
ducting slabs and their deformation in the transition zone. Physics of the Earth
and Planetary Interiors 161 (3e4), 202e214. http://dx.doi.org/10.1016/
j.pepi.2007.02.002.

Czkov, H., van Hunen, J., van den Berg, A.P., Vlaar, N.J., 2002. The inuence of
rheological weakening and yield stress on the interaction of slabs with the
670 km discontinuity. Earth and Planetary Science Letters 199 (3e4), 447e457.
http://dx.doi.org/10.1016/S0012-821X(02)00586-1.
Christensen, U.R., 1996. The inuence of trench migration on slab penetration into
the lower mantle. Earth and Planetary Science Letters 140 (1e4), 27e39. http://
dx.doi.org/10.1016/0012-821X(96)00023-4.
Christensen, U.R., Yuen, D.A., 1985. Layered convection induced by phase transi-
tions. Journal of Geophysical Research 90 (B12), 10291e10300. http://
dx.doi.org/10.1029/JB090iB12p10291.
Collier, J.D., Helffrich, G.R., Wood, B.J., 2001. Seismic discontinuities and subduction
zones. Physics of the Earth and Planetary Interiors 127, 35e49.
Coltice, N., Ricard, Y., 1999. Geochemical observations and one layer mantle con-
vection. Earth and Planetary Science Letters 174 (1e2), 125e137. http://
dx.doi.org/10.1016/S0012-821X(99).
Cserepes, L., Yuen, D.A., 1997. Dynamical consequences of mid-mantle viscosity
stratication on mantle ows with an endothermic phase transition.
Geophysical Research Letters 24, 181e184.
Cserepes, L., Yuen, D.A., 2000. On the possibility of a second kind of mantle plume.
Earth and Planetary Science Letters 183 (1e2), 61e71. http://dx.doi.org/10.1016/
S0012-821X(00)00265-X.
Cserepes, L., Yuen, D.A., Schroeder, B.A., 2000. Effect of the mid-mantle viscosity and
phase-transition structure on 3D mantle convection. Earth and Planetary Sci-
ence Letters 118, 135e148. http://dx.doi.org/10.1016/S0031-9201(99)00140-5.
Davies, G.F., 1995. Penetration of plates and plumes through the mantle transition
zone. Earth and Planetary Science Letters 133 (3e4), 507e516. http://
dx.doi.org/10.1016/0012-821X(95)00039-F.
Flanagan, M.P., Shearer, P.M., 1998. Global mapping of topography on transition
zone velocity discontinuities by stacking SS precursors. Journal of Geophysical
Research 103, 2673e2692.
Forte, A.M., 2007. Constraints on seismic models from other disciplines e impli-
cations for mantle dynamics and composition. In: Romanowicz, B.,
Dziewonski, A.M. (Eds.), Treatise on Geophysics, vol. 1, pp. 805e854.
Forte, A.M., Mitrovica, J.X., 2001. Deep-mantle high-viscosity ow and thermo-
chemical structure inferred from seismic and geodynamic data. Nature 410,
1049e1056.
Forte, A.M., Peltier, W.R., 1991. Viscous ow models of global geophysical observables.
Part 1: Forward problems. Journal of Geophysical Research 96, 20131e20159.
Forte, A.M., Qur, S., Moucha, R., Simmons, N.A., Grand, S.P., Mitrovica, J.X.,
Rowley, D.B., 2010. Joint seismic-geodynamic-mineral physical modeling of
African geodynamics: a reconstruction of deep-mantle convection with surface
geophysical constraints. Earth and Planetary Science Letters 295, 329e341.
Fukao, Y., Obayashi, M., Inoue, H., Nenbai, M., 1992. Subducting slabs stagnant in the
mantle transition zone. Journal of Geophysical Research 97, 4809e4822.
Fukao, Y., Widiyantoro, S., Obayashi, M., 2001. Stagnant slabs in the upper and lower
mantle transition region. Reviews of Geophysics 39, 291e323.
Gaherty, J.B., Hager, B.H., 1994. Compositional vs. thermal buoyancy and the evo-
lution of subducted lithosphere. Geophysical Research Letters 21, 141e144.
Grand, S.P., 2002. Mantle shear-wave tomography and the fate of subducted slabs.
Philosophical Transactions of the Royal Society A 360, 2475e2491.
Grand, S.P., van der Hilst, R.D., Widiyantro, S., 1997. Global seismic tomography: a
snapshot of convection in the earth. GSA Today 7 (4), 1e7.
Hae, R., Ohtani, E., Kubo, T., Koyama, T., Utada, H., 2006. Hydrogen diffusivity in
wadsleyite and water distribution in the mantle transition zone. Earth and
Planetary Science Letters 243, 141e148.
Hofmann, A.W., 1997. Mantle geochemistry: the message from oceanic volcanism.
Nature 385, 219e229. http://dx.doi.org/10.1038/385219a0.
Inoue, T., Yurimoto, H., Kudoh, Y., 1995. Hydrous modied spinel, Mg1.75SiH0.5O4: a
new water reservoir in the mantle transition region. Geophysical Research
Letters 22, 117e120.
Inoue, T., Weidner, D.J., Northrup, P.A., Parise, J.B., 1998. Elastic properties of hydrous
ringwoodite (g-phase) in Mg2SiO4. Earth and Planetary Science Letters 160,107e113.
Irifune, T., Ringwood, A.E., 1993. Phase transformations in subducted oceanic crust
and buoyancy relationships at depths of 600e800 km, in the mantle. Earth and
Planetary Science Letters 117, 101e110.
Irifune, T., Higo, Y., Inoue, T., Kono, Y., Ohfuji, H., Funakoshi, K., 2008. Sound ve-
locities of majorite garnet and the composition of the mantle transition region.
Nature 451, 814e817. http://dx.doi.org/10.1038/nature06551.
Ito, E., Takahashi, E., 1989. Postspinel transformations in the system Mg2SiO4-
Fe2SiO4 and some geophysical implications. Journal of Geophysical Research 94,
10,637e10,646.
Jacobson, S.D., Smyth, J.R., Spetzler, H., Holl, C.M., 2004. Sound velocities and elastic
constants of iron bearing hydrous ringwoodite. Physics of the Earth and Plan-
etary Interiors 143e144, 47e56.
Kaneoka, I., 1995. Noble gas constraints on the plume sources in the Earths deep
interior. In: Yukutake, T. (Ed.), The Earths Central Part: Its Structure and Dy-
namics. TERRAPUB, Tokyo, pp. 343e355.
Kaneoka, I., 2008. On the degassing state and the chemical structure of the Earths
interior inferred from noble gas isotopes e past and recent views. Geochemical
Journal 42 (1), 3e20.
Kaneshima, S., Helffrich, G., 2003. Subparallel dipping heterogeneities in the mid-
lower mantle. Journal of Geophysical Research 108 (B5), 2272. http://
dx.doi.org/10.1029/2001JB001596.
Kaneshima, S., 2009. Seismic scatterers at the shallowest lower mantle beneath
subducted slabs. Earth and Planetary Science Letters 286 (1e2), 304e315.
http://dx.doi.org/10.1016/j.epsl.2009.06.044.
Kaneshima, S., Helffrich, G., 2010. Small scale heterogeneity in the mid-lower
mantle beneath the circum-Pacic area. Physics of the Earth and Planetary
Interiors 183 (1e2), 91e103. http://dx.doi.org/10.1016/j.pepi.2010.03.011.
Katayama, I., Karato, S., 2008. The effect of water and iron content on the rheo-
logical contrast between garnet and olivine. Physics of the Earth and Planetary
Interiors 166, 57e66.
Katsura, T., Ito, E., 1989. The system Mg2SiO4-Fe2SiO4 at high pressures and tem-
peratures: precise determination of stabilities of olivine, modied spinel, and
spinel. Journal of Geophysical Research 94, 15663e15670. http://dx.doi.org/
10.1029/JB094iB11p15663.
Katsura, T., et al., 2009. P-V-T relations of MgSiO3 perovskite determined by in situ
X-ray diffraction using a large-volume high-pressure apparatus. Geophysical
Research Letters 36, L01305. http://dx.doi.org/10.1029/2008GL035658.
Kawakatsu, H., Watada, S., 2007. Seismic evidence for deep-water transportation in
the mantle. Science 316, 1468e1471.
Kelbert, A., Schultz, A., Egbert, G., 2009. Global electromagnetic induction con-
straints on transition-zone water content variations. Nature 460, 1003e1006.
Kennett, B.L.N., Engdahl, E.R., 1991. Traveltimes for global earthquake location and
phase identication. Geophysical Journal International 105, 429e465.

Kido, M., Cadek, O., 1997. Inferences of viscosity from the oceanic geoid: indication of a
low viscosity zone below the 660-km discontinuity. Earth and Planetary Science
Letters 151 (3e4), 125e137. http://dx.doi.org/10.1016/S0012-821X(97)81843-2.
Kido, M., Yuen, D.A., 2000. The role played by a low viscosity zone under a 660 km
discontinuity in regional mantle layering. Earth and Planetary Science Letters
181 (4), 573e583. http://dx.doi.org/10.1016/S0012-821X(00)00217-X.
King, S., Masters, G., 1992. An inversion for the radial viscosity structure using
seismic tomography. Geophysical Research Letters 19, 1551e1554.
Li, X., Sobolev, R., Kind, R., Yuan, X., Estrabrook, C., 2000. A detailed receiver func-
tion image of the upper mantle discontinuities in the Japan subduction zone.
Earth and Planetary Science Letters 183, 527e541.
Li, Juan, Wang, X., Wang, X., Yuen, D.A., 2013. P and SH velocity structure in the
upper mantle beneath Northeast China: evidence for a stagnant slab in hydrous
mantle transition zone. Earth and Planetary Science Letters 367, 71e81. http://
dx.doi.org/10.1016/j.epsl.2013.02.026.
Litasov, K.D., Ohtani, E., Sano, A., 2006. Inuence of water on major phase transitions in
the Earths mantle. In: Monograph, Earths Deep Water Cycle. AGU Geophysical
Monograph, Series, vol. 168, pp. 95e112. http://dx.doi.org/10.1029/168GM08.
Litasov, K.D., Ohtani, E., Sano, A., Suzuki, A., Funakoshi, K., 2005. Wet subduction
versus cold subduction. Geophysical Research Letters 32, L13312. http://
dx.doi.org/10.1029/2005GL022921.
 F. Tajima et al. / Geoscience Frontiers 6 (2015) 79e93
McNamara, A.K., Zhong, S.-J., 2005. Thermochemical structures beneath Africa and
the Pacic Ocean. Nature 437, 1136e1139.
Niu, F., Kawakatsu, H., Fukao, Y., 2003. Seismic evidence for a chemical heteroge-
neity in the midmantle: a strong and slightly dipping seismic reector beneath
the Mariana subduction zone. Journal of Geophysical Research 108 (B9), 2419.
http://dx.doi.org/10.1029/2002JB002384.
Niu, F., Levander, A., Ham, S., Obayashi, M., 2005. Mapping the subducting Pacic
slab beneath southwest Japan with Hi-net receiver functions. Earth and Plan-
etary Science Letters 239, 9e17. http://dx.doi.org/10.1016/j.epsl.2005.08.009.
Ohtani, E., 2005. Water in the mantle. Elements 1, 25e30.
Ohtani, E., Litasov, K., 2006. The effect of water on mantle phase transitions. Re-
views in Mineralogy and Geochemistry 62, 397e420.
Ohtani, E., Litasov, K., Hosoya, T., Kubo, T., Kondo, T., 2004. Water transport into the
deep mantle and formation of a hydrous transition zone. Physics of the Earth
and Planetary Interiors 143e144, 255e269.
Ohtani, E., Zhao, D., 2009. The role of water in the deep upper mantle and transition
zone: dehydration of stagnant slabs and its effects on the big mantle wedge.
Russian Journal of Geology and Geophysics 50 (12), 1073e1078.
Okada, Y., Kasahara, K., Hori, S., Obara, K., Sekiguchi, S., Fujiwara, H., Yamamoto, A.,
2004. Recent progress of seismic observation networks in Japan e Hi-net, F-net,
K-NET and KiK-net. Earth, Planets and Space 56, xvexxviii.
ONions, R.K., Oxburgh, E.R., 1983. Heat and helium in the earth. Nature 306,
429e431. http://dx.doi.org/10.1038/306429a0.
Peltier, W.R., 1998. Postglacial variations in the level of the sea: implications for
climate dynamics and solid-Earth geophysics. Reviews of Geophysics 36 (4),
603e689. http://dx.doi.org/10.1029/98RG02638.
Ricard, Y., Wuming, B., 1991. Inferring viscosity and the 3-D density structure of the
mantle from geoid, topography and plate velocities. Geophysical Journal In-
ternational 105, 561e572.
Richards, M.A., Davies, G.F., 1989. On the separation of relatively buoyant compo-
nents from subducted lithosphere. Geophysical Research Letters 16 (8),
831e834. http://dx.doi.org/10.1029/GL016i008p00831.
Ringwood, A.E., 1994. Role of the transition zone and 660 km discontinuity in
mantle dynamics. Physics of the Earth and Planetary Interiors 86, 5e24.
Ringwood, A.E., Irifune, T., 1988. Nature of the 650-km seismic discontinuity: im-
plications for mantle dynamics and differentiation. Nature 331, 131e136.
Ritsema, J., van Heijst, H.J., Woodhouse, J., 2004. Global transition zone tomography.
Journal of Geophysical Research 109, B02302. http://dx.doi.org/10.1029/
2003JB002610.
Sano, A., Ohtani, E., Litasov, K., Kubo, T., Hosoya, T., Funakoshi, K., Kikegawa, T., 2006. In
situ X-ray diffraction study of the effect of water on the garnet-perovskite trans-
formation in MORB and implications for penetration of oceanic crust into the lower
mantle. Physics of the Earth and Planetary Interiors 159, 118e126.
Sano, A., Ohtani, E., Kondo, T., Hirano, N., Sakai, T., Sata, N., Ohishi, Y., Kikegawa, T.,
2008. Aluminous hydrous mineral d-AlOOH as a carrier of hydrogen into the
core-mantle boundary. Geophysical Research Letters 35, L03303. http://
dx.doi.org/10.1029/2007GL031718.
Schubert, G., Turcotte, D.L., Olson, P., 2001. Mantle Convection in the Earth and
Planets. Cambridge Univ. Press, 956 pp.
Schuberth, B.S.A., Bunge, H.-P., Steinle-Neumann, G., Moder, C., Oeser, J., 2009a.
Thermal versus elastic heterogeneity in high-resolution mantle circulation
models with pyrolite composition: high plume excess temperatures in the
lowermost mantle. Geochemistry, Geophysics, Geosystems 10, Q01W01. http://
dx.doi.org/10.1029/2008GC002235.
Schuberth, B.S.A., Bunge, H.-P., Ritsema, J., 2009b. Tomographic ltering of high-
resolution mantle circulation models: can seismic heterogeneity be explained
by temperature alone. Geochemistry, Geophysics, Geosystems 10, Q05W03.
http://dx.doi.org/10.1029/2009GC002401.
Shearer, P.M., Flanagan, M.P., Hedlin, M.A.H., 1999. Experiments in migration pro-
cessing of SS precursor data to image upper mantle discontinuity structure.
Journal of Geophysical Research 104, 7229e7242.
Simmons, N.A., Forte, A.M., Grand, S.P., 2006. Constraining mantle ow with seismic
and geodynamic data: a joint approach. Earth and Planetary Science Letters
246, 109e124.
Tackley, P., 2000. Mantle convection and plate tectonics: toward an integrated
physical and chemical theory. Science 288 (5743), 2002e2007. http://
dx.doi.org/10.1126/science.288.5473.2002.
Tackley, P.J., 2007. Mantle geochemical geodynamics. In: Bercovici, D., Schubert, G.
(Eds.), Treatise on Geophysics, Mantle Dynamics, vol. 7. Elsevier, pp. 437e505.
Tagawa, M., Nakakuki, T., Tajima, F., 2007a. Dynamical modeling of trench retreat
driven by the slab interaction with the mantle transition zone. Earth, Planets
and Space 59 (2), 65e74.
Tagawa, M., Nakakuki, T., Kameyama, M., Tajima, F., 2007b. The role of history-
dependent rheology in plate boundary lubrication for generating one-sided
subduction. Pure and Applied Geophysics 164, 1e29. http://dx.doi.org/
10.1007/s00024-007-0197-4.
Tajima, F., Fukao, Y., Obayashi, M., Sakurai, T., 1998. Evaluation of slab images in the
northwestern Pacic. Earth, Planets and Space 50, 953e964.
Tajima, F., Grand, S.P., 1995. Evidence of high velocity anomalies in the transition
zone associated with southern Kurile subduction zone. Geophysical Research
Letters 22, 3139e3142.
Tajima, F., Grand, S.P., 1998. Variation of transition zone high velocity anomalies and
depression of the 660 km discontinuity associated with subduction zones from
the southern Kuriles to Izu-Bonin. Journal of Geophysical Research 103 (B7),
15015e15036.
93
Tajima, F., Nakagawa, T., 2006. Implications of seismic waveforms: complex physical
properties associated with stagnant slab. Geophysical Research Letters 33,
L03311. http://dx.doi.org/10.1029/2005GL024314.
Tajima, F., Katayama, I., Nakagawa, T., 2009. Variable seismic structure near the
660 km discontinuity associated with stagnant slabs and geochemical impli-
cations. Physics of the Earth and Planetary Interiors 172, 183e198. http://
dx.doi.org/10.1016/j-pepi, 2008.09.013.
Tonegawa, T., Hirahara, K., Shibutani, T., Iwamori, H., Kanamori, H., Shiomi, K., 2008.
Water ow to the mantle transition zone inferred from a receiver function
image of the Pacic slab. Earth and Planetary Science Letters 274, 346e354.
http://dx.doi.org/10.1016/j.epsl.2008.07.046.
Tono, Y., Kunugi, T., Fukao, Y., Tsuboi, S., Kanjo, K., Kasahara, K., 2005. Mapping of
the 410- and 660-km discontinuities beneath the Japanese islands. Journal of
Geophysical Research 110, B03307. http://dx.doi.org/10.1029/2004JB003266.
Torii, Y., Yoshioka, S., 2007. Physical conditions producing slab stagnation: con-
straints of the Clapeyron slope, trench retreat, and dip angles. Tectonophysics
445 (3e4), 200e209. http://dx.doi.org/10.1016/j.tecto.2007.08.003.
Utada, H., Koyama, T., Obayashi, M., Fukao, Y., 2009. A joint interpretation of
electromagnetic and seismic tomography models suggests the mantle transi-
tion zone below Europe is dry. Earth and Planetary Science Letters 281,
249e257.
van der Hilst, R.D., Engdahl, R., Sparkman, W., Nolet, G., 1991. Tomographic imaging
of subducted lithosphere below northwest Pacic island arc. Nature 353,
37e43. http://dx.doi.org/10.1038/353037a0.
van der Hilst, R.D., Widiyantoro, S., Engdahl, E.R., 1997. Evidence for deep mantle
circulation from global tomography. Nature 386, 578e584.
van Keken, P.E., Karato, S., Yuen, D.A., 1996. Rheological control of oceanic crust
separation in the transition zone. Geophysical Research Letters 23 (14),
1821e1824. http://dx.doi.org/10.1029/96GL01594.
Wang, T., Chen, L., 2009. Distinct velocity variations around the base of the upper
mantle beneath northeast Asia. Physics of the Earth and Planetary Interiors 172,
241e256. http://dx.doi.org/10.1016/j-pepi.2008.09.021.
Wang, B., Niu, F., 2010. A broad 660-km discontinuity beneath northeast China
revealed by dense regional seismic networks in China. Journal of Geophysical
Research 115 (B6). http://dx.doi.org/10.1029/2009JB006608.
Wang, B., Niu, F., 2011. Spatial variations of the 660-km discontinuity in the western
Pacic subduction zones observed from CEArray triplication data. Earthquake
Science 24, 77e85. http://dx.doi.org/10.1007/s11589-011-0771-9.
Wei, W., Xu, J., Zhao, D., Shi, Y., 2012. Journal of Asian Earth Sciences 60, 88e103.
http://dx.doi.org/10.1016/j.jseaes.2012.08.001.
Wessel, P., Smith, W.H.F., 1998. New, improved version of the Generic Mapping
Tools released. Eos, Transactions, American Geophysical Union 79 (47), 579.
http://dx.doi.org/10.1029/98EO00426.
Widiyantoro, S., Kennett, B.L.N., van der Hilst, R.D., 1999. Seismic tomography with P
and S data reveals lateral variations in the rigidity of deep slabs. Earth and
Planetary Science Letters 173, 91e100.
Ye, L., Li, J., Tseng, T., Yao, Z., 2011. A stagnant slab in a water-bearing mantle
transition zone beneath northeast China: implications from regional SH
waveform modeling. Geophysical Journal International. http://dx.doi.org/
10.1111/j.1365-246X.2011.05063.x.
Yoshida, M., 2008a. Core-mantle boundary topography estimated from numerical
simulations of instantaneous mantle ow. Geochemistry, Geophysics, Geo-
systems 9 (7), Q07002. http://dx.doi.org/10.1029/2008GC002008.
Yoshida, M., 2008b. Mantle convection with longest-wavelength thermal hetero-
geneity in a 3-D spherical model: degree one or two? Geophysical Research
Letters 35, L23302. http://dx.doi.org/10.1029/2008GL036059.
Yoshida, M., 2010. Preliminary three-dimensional model of mantle convection with
deformable, mobile continental lithosphere. Earth and Planetary Science Letters
295 (1e2), 205e218. http://dx.doi.org/10.1016/j.epsl.2010.04.001.
Yoshida, M., 2013. The role of harzburgite layers in the morphology of subducting
plates and the behavior of oceanic crustal layers. Geophysical Research Letters
40, 1e6. http://dx.doi.org/10.1002/2013GL057578.
Yoshida, M., Tajima, F., Honda, S., Morishige, M., 2012. The 3D numerical modeling
of subduction dynamics: plate stagnation and segmentation, and crustal
advection in the wet mantle transition zone. Journal of Geophysical Research
117 (B4), B04104. http://dx.doi.org/10.1029/2011JB008989.
Yoshida, M., Tajima, F., 2013. On the possibility of a folded crustal layer stored in the
hydrous mantle transition zone. Physics of the Earth and Planetary Interiors
219, 34e48. http://dx.doi.org/10.1016/j.pepi.2013.03004.
Yoshioka, S., Naganoda, A., 2010. Effects of trench migration on fall of stagnant slabs
into the lower mantle. Physics of the Earth and Planetary Interiors 183 (1e2),
321e329. http://dx.doi.org/10.1016/j.pepi.2010.09.002.
Zhang, Y., Wang, Y., Wu, Y., Bina, C.R., Jin, Z., Dong, S., 2013. Phase transition of
harzburgite and buckled slab under eastern China. Geochemistry, Geophysics,
Geosystems 14. http://dx.doi.org/10.1002/ggge.20069.
Zhao, D., 2004. Global tomographic images of mantle plumes and subducting slabs:
insight into deep Earth dynamics. Physics of the Earth and Planetary Interiors
146, 3e34. http://dx.doi.org/10.1016/j.pepi.2003.07.032.
Zhao, D., Tian, Y., 2013. Changbai intraplate volcanism and deep earthquakes in East
Asia: a possible link? Geophysical Journal International 195, 706e724. http://
dx.doi.org/10.1093/gji/ggt289.
Zhong, S., Gurnis, M., 1995. Mantle convection with plates and mobile, faulted plate
margins. Science 267 (5199), 838e843. http://dx.doi.org/10.1126/
science.267.5199.838.