See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/256660535

From basalts to boninites: The geodynamics of volcanic expression during

induced subduction initiation

Article  in  Lithosphere · October 2012

DOI: 10.1130/L215.1

CITATIONS READS
27 136

3 authors, including:

Paul Asimow Wei Leng

California Institute of Technology University of Science and Technology of China
294 PUBLICATIONS   6,794 CITATIONS    42 PUBLICATIONS   726 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Geochemistry of some Late Neoproterozoic Rocks,Eastern Desert,Egypt. View project

The evolution of late Neoproterozoic rocks in southern Sinai, Egypt View project

All content following this page was uploaded by Paul Asimow on 19 May 2014.

The user has requested enhancement of the downloaded file.

 RESEARCH

From basalts to boninites: The geodynamics of volcanic

expression during induced subduction initiation
Wei Leng1,*, Michael Gurnis1, and Paul Asimow2
1
SEISMOLOGICAL LABORATORY, CALIFORNIA INSTITUTE OF TECHNOLOGY, PASADENA, CALIFORNIA 91125, USA
2
DIVISION OF GEOLOGICAL AND PLANETARY SCIENCES, CALIFORNIA INSTITUTE OF TECHNOLOGY, PASADENA, CALIFORNIA 91125, USA

ABSTRACT

Subduction initiation may unfold via different pathways in response to plate strength, plate age, and driving mechanism. Such pathways
influence volcanism on the overriding plate and may be preserved in the sequence of erupted volcanic products. Here, we parameterize
melting in a mechanical model to determine the volcanic products that form in response to different subduction initiation modes. We find
that with a mode of continuous initiation with infant-arc spreading, the foundering of the subducting slab and water release from the
slab govern a succession from basalts with compositions similar to mid-ocean-ridge basalts (MORB) to boninites. The modeled transition
from MORB-like to boninite composition typically occurs within a few million years. When plate strength is reduced, the subducting slab
tends to segment, with extensive melting occurring when the slab breaks; most melting occurs close to the trench. When plate strength
increases, subduction initiation becomes continuous without infant-arc spreading; such a mode leads to a limited, very low degree of
melting occurring during a long interval of plate convergence before subduction initiation starts, although extensive melting near the
trench is still possible when subduction initiation starts after a protracted period of plate convergence (~10 m.y.). If the subduction initia-
tion is driven by constant stresses, such as through ridge push, the slab subducts rapidly in response to continuous acceleration of the
plate under action of the far-field push; significant melting, including boninite eruption, can be generated within a few million years with
no trench migration. Based on the tectonic and volcanic evolution, these different modes may be applicable to the initiation of the Izu-
Bonin-Mariana arc (infant-arc spreading and a sequence from MORB-like to boninites), the New Hebrides arc (slab segments in the upper
mantle), the Puysegur Trench in New Zealand (scarce distribution of volcanism and no infant-arc spreading), and the Aleutian Trench
(strong volcanism and no infant-arc spreading).

LITHOSPHERE; v. 4; no. 6; p. 511–523 | Published online 26 October 2012 doi: 10.1130/L215.1

INTRODUCTION in response to induced subduction initiation by but no infant-arc spreading is observed. For
external forces (Toth and Gurnis, 1998; Hall et the Puysegur subduction zone, i.e., the north-
Subduction initiation occurs at the inter- al., 2003; Gurnis et al., 2004; Thielmann and ern segment of the Macquarie Ridge complex
faces of adjacent plates, such as transform faults Kaus, 2011; Leng and Gurnis, 2011). However, between the Australian and Pacific plates, plate
between two oceanic plates, possibly induced the volcanic expression on the overriding plate convergence has been under way since 15 Ma,
by a change in plate motions (McKenzie, 1977; in response to unfolding subduction initiation and the downdip tip of slab has reached 150 km
Gurnis et al., 2004; Stern, 2004). The detailed remains unexplored. depth (Sutherland et al., 2009), but the evidence
initiation processes can be deciphered from stud- The volcanic expression on the overriding for volcanism is scarce, with only small isolated
ies of sedimentary and volcanic rock sequences plate shows distinctive patterns for the several seamounts found on the Pacific plate, notably
of specific nucleating margins (House et al., known subduction initiation events. Extensive Solander Island (Reay and Parkinson, 1997).
2002; Sutherland et al., 2009; Reagan et al., magmatism characterized by depleted (tho- These different volcanic patterns may have
2010), and then placed within the context of the leiitic) and ultradepleted (boninitic) magmas been caused by divergent evolutionary pathways
regional kinematics of the major plates (e.g., occurred during the initiation of the Izu-Bonin- in response to different mechanical conditions.
Sharp and Clague, 2006). Computational mod- Mariana subduction zone (Stern and Bloomer, Driven by an imposed constant velocity, sub-
els have been used to investigate the mechan- 1992). Detailed, in situ sampling of the Mariana duction initiation may occur catastrophically
ics of initiation, including the tectonic evolu- forearc near Guam has revealed a volcanic stra- after a finite amount of plate convergence (Hall
tion of the overriding and subducting plates, tigraphy in which basalts similar to mid-ocean- et al., 2003; Gurnis et al., 2004). In this case, the
ridge basalts (MORB) were the first to erupt in older and denser plate founders with the devel-
the nascent arc and were quickly followed by opment of infant-arc spreading and rapid trench
boninites and then by normal arc lavas within rollback, all occurring within a few million
*E-mail: [email protected]. several million years (Reagan et al., 2010). The years. Once the foundering starts, the theoretical
Editor’s note: This article is part of a special issue ti- same volcanic sequence is found ~1500 km to model of induced subduction is similar to Stern
tled “Initiation and Termination of Subduction: Rock Re- the north near the Bonin Islands (Ishizuka et al., and Bloomer’s (1992) conceptual model of self-
cord, Geodynamic Models, Modern Plate Boundaries,”
edited by John Shervais and John Wakabayashi. The
2011). In contrast, the formation of the Aleu- nucleating subduction. The difference between
full issue can be found at http://lithosphere.gsapubs tian island arc has been accompanied by a high the two is the prediction of compression and
.org/content/4/6.toc. rate of magma production (Jicha et al., 2006), rapid uplift of the overriding plate in the induced

LITHOSPHERE
For | Volume
permission to 4 | Number
copy, contact 6 | www.gsapubs.org
[email protected] | © 2012 Geological Society of America 511
LENG ET AL.

model just before slab foundering, while the experiments on melting. From the parameterized Kelvin) (Table 1). In addition, the maximum and
self-nucleating model has no uplift and com- approach, extent of melting (Katz et al., 2003), minimum viscosities are limited with ηmax and
pression. Gurnis et al. (2004) argued that relic major-element composition of melts for both ηmin, respectively (Table 1). We use a Drucker-
structures on the overriding plate showed uplift lherzolite (Kinzler and Grove, 1992a, 1992b) Prager yielding model for the plasticity:
just before the Izu-Bonin-Mariana arc formed, and harzburgite melting (Parman and Grove,
but this possibility needs further testing. The 2004), and trace-element compositions can be τ y = μ P + C, (3)
mechanical conditions under which this mode determined. Although the parameterizations are
will occur have been clarified by Leng and Gur- largely empirical and may not reflect the exact where τy, μ, P, and C are the yield stress, coef-
nis (2011). When we systematically increase the physical and chemical process of melting, they ficient of friction, pressure, and cohesion.
coefficient of internal friction or decrease the are simple and computationally efficient. More Cohesion is the strength of the material at zero
rate of plastic weakening of the plates, the mode importantly, they are able to largely reproduce normal stress. The values of μ and C are lin-
of subduction initiation induced by imposed the melting results of laboratory experiments early reduced with accumulated plastic strain
velocity is found to change from (1) segmented and provide us with a first-order understanding to reproduce the weakening of mantle material
initiation to (2) continuous initiation with infant- of melting processes during subduction initia- with increasing plastic strain (Buck and Polia-
arc spreading and eventually to (3) continuous tion. Here, we collect the relevant parameter- kov, 1998; Poliakov and Buck, 1998):
initiation without infant-arc spreading (Leng izations from the literature, build a geodynamic
and Gurnis, 2011). Here, the terms “segmented” melting model, and obtain a complete picture of
⎛ ε ⎞
versus “continuous” describe whether or not the way in which subduction initiation evolves μ = μ 0 1 − εp , (4)
⎝ f ⎠
the slab breaks at depth during initiation. These with time and how this process can influence
three different modes have respectively been volcanic sequences.
⎛ ε⎞
applied to the initiation of the New Hebrides, ( ⎝
)
C = C f + C 0 − C f 1 − εp ,
f⎠
(5)
Izu-Bonin-Mariana, and Puysegur subduction A COUPLED GEODYNAMIC MELTING
zones (Leng and Gurnis, 2011). Subduction ini- MODEL FOR SUBDUCTION INITIATION
tiation can also be induced by a constant stress, where μ0, εp, εf, C0, and Cf are the initial coef-
such as by a ridge push force. The constant Geodynamic Model ficient of friction, accumulated plastic strain,
stress case may apply to the Aleutian subduc- reference plastic strain, initial cohesion, and
tion zone, initiated by subduction of the small We used the two-dimensional (2-D) finite- minimum cohesion. Here, εp represents the total
Kula plate (Seton et al., 2012). In this case, the element model Ellipsis (Moresi et al., 2003) to accumulated plastic strain after the yield stress
mode of subduction initiation is similar to con- study the visco-elasto-plastic behavior of the lith- is reached, and εf represents a threshold value
tinuous initiation without infant-arc spreading, osphere and mantle during subduction initiation. of plastic strain. When the accumulated plastic
but it requires a much shorter plate convergence Ellipsis uses Lagrangian particles to accurately strain is larger than εf, no further weakening
time (Leng and Gurnis, 2011). track material properties and deformational his- is accounted for. Table 1 shows the invariant
The relationship between volcanism and tory. Mechanically, the geodynamic model is model parameters used. It has been suggested
the evolutionary pathway of a nascent subduc- the same as the subduction initiation model in that μ0 and εp determine the mode of subduc-
tion zone can be addressed through coupling Leng and Gurnis (2011), but here we add melt- tion initiation (Leng and Gurnis, 2011). We vary
between a quantitative melting model and a ing. Governing equations and model details can these two parameters herein to produce different
mechanical model; such models can provide be found in Moresi et al. (2003) and Leng and types of subduction initiation.
insights into the connections between the chem- Gurnis (2011). We only provide an overview of The computational domain is 1050 km in
ical composition of magmatic products and the mantle rheology and the model domain. width by 300 km in depth (Fig. 1). A 50-km-
evolution of temperature, pressure, mantle water We used an incompressible Maxwell mate- thick compressible “sticky air layer” is inserted
contents, and surface plate kinematics. Using rial for visco-elasticity. The deformation rate is above the normal solid domain to simulate a free
thermodynamic principles, petrological models the sum of an elastic part and a viscous part: surface, which may be important for subduction
such as pHMELTS (Asimow et al., 2004) have (e.g., Crameri et al., 2012) (Fig. 1). The air layer
1 1
been developed to compute the extent of melt- ε ij = sij + sij, (1) has zero density but has a viscosity ηmin. In nom-
2G 2η
ing and major- and trace-element compositions inal computations, 128 × 64 elements are used
of mantle-derived melts. This approach may where εij , sij , sij , G, and η are the strain rate, time
not be efficient for application to large-scale rate of change of deviatoric stress, deviatoric
geodynamics models (Hebert et al., 2009a), as stress, shear modulus, and viscosity, respectively,
TABLE 1. MODEL CONSTANTS
pHMELTS needs to compute melt extent and and i and j are spatial indices. The viscosity is
composition for tens of thousands of points for non-Newtonian and temperature dependent: Symbols and definition Values
every time step in a geodynamic model, which G shear modulus 30 GPa
⎛1 ⎞
typically runs for tens of thousands of time steps ⎛ε ⎞ ⎝n −1⎠ ⎡ E ⎛ 1 1 ⎞⎤ η0 reference viscosity 5 × 1019 Pa s
at significant computational cost. η = η0 ⎜ II ⎟ exp ⎢ ⎜ − ⎟⎥ , (2) ε0 reference strain rate 10–15 s–1

⎝ ε0 ⎠ nR
Alternatively, computations of extent of ⎣ ⎝T T0⎠⎦ n strain exponent 3.05
E activation energy 540 kJ mol–1
melting and magma compositions can be param- R gas constant 8.31 J mol–1
eterized (Katz et al., 2003; Kinzler and Grove, where η0, εII, ε0, n, E, R, T, and T0 are the refer- T0 reference temperature 1355 °C
1992a, 1992b; Parman and Grove, 2004), with ence viscosity, the second invariant of the devia- ηmax maximum viscosity 1025 Pa s
several controlling parameters constructed to toric strain rate tensor, reference strain rate, ηmin minimum viscosity 1019 Pa s
describe complicated melting processes. These strain exponent, activation energy, gas constant, C0 initial cohesion 44 MPa
Cf minimum cohesion 4.4 MPa
parameters are calibrated to match laboratory and absolute and reference temperatures (both in

512 www.gsapubs.org | Volume 4 | Number 6 | LITHOSPHERE

 Basalts to boninites: Geodynamics of volcanism during subduction initiation | RESEARCH
Figure 1. The geodynamic Weak zone and 525 km in horizontal direction (Fig. 1),
model domain for study- Overriding plate Subducting plate
we uniformly place 40 sampling points, which
ing subduction initiation. store the thickness and compositions of erupted
When the imposed veloc- Sampling points Air layer 50 km
volcanic rocks and subsequently translate them
ity boundary condition
with the surface (plate) velocity. The effective

350 km
(BC) is replaced by an

300 km
imposed stress boundary widths of these points are the same as the ele-
condition, we replace the Stress position for ment widths beneath the sampling point. These
imposed stress BC Weak zone for
mechanical condition of imposed stress BC sampling points move in response to plate
the right boundary from motion and stretching. If the distance between
flow-through to free slip 1050 km two adjacent sampling points becomes too
and put another weak zone
(the dashed rectangle, 50 by 100 km) at the right end of the subducting plate. large, >87.5 km, 10 more sampling points are
uniformly added between these two. For the
newly generated melts, the trace-element com-
positions are computed from constant partition
with 16 tracking particles initially inserted uni- depleted MORB mantle (DMM) composition coefficients for each trace element (Table 3)
formly in each element; additional higher reso- (Workman and Hart, 2005). This DMM com- (Workman and Hart, 2005) with the batch melt-
lutions were used by Leng and Gurnis (2011), position is slightly depleted with respect to bulk ing equation (Schilling, 1966):
and the resulting dynamics were resolved with silicate earth, as required by the complemen-
this resolution. The subducting and overriding tary trace elements and isotopic signatures of
CL 1
plates have different initial thermal ages and are continental crust and the residual upper-mantle = , (6)
divided at the box center with a weak zone (17.5 MORB source, but it is actually a fertile lherzo- C B D + F ∗ (1 − D )
by 17.5 km) (Fig. 1). The overriding plate age is lite composition. We only consider eight major
2 m.y., and the subducting plate age tp is varied elements and replace others with SiO2 (Table 2). where CL, CB, D, and F are the element concen-
according to different tectonic settings. There is The initial mantle phase partitions are (in weight tration in liquid, bulk composition of the source
no heat flux at the left and right boundaries. The percent): olivine 57%, orthopyroxene 28%, mantle (updated in time, so this formulation in
initial temperature field is computed from the clinopyroxene 13%, and spinel 2% (Workman fact describes near-fractional melting), partition
half-space cooling model with a fixed reference and Hart, 2005). Each time a particle moves to coefficient, and melt extent. A nonmodal trace-
temperature T0 at the bottom boundary. The top, a new position, we compute the extent of melt- element partitioning model would be superior,
bottom, and left boundaries are free slip. We ing from the temperature, pressure, water con- but for the present exercise, this simple model
use two different boundary conditions to induce tent, and weight percent of modal clinopyroxene is sufficient; note that the bulk D values used
subduction initiation. The first is an imposed based on the parameterized method of Katz et al. are appropriate to describe the average melting
constant velocity at the right boundary. The (2003). If the extent of melting is larger than that process observed in mid-ocean-ridge settings.
vertical velocity is zero, and an imposed inward tracked by the particle, new melt is generated The major-element compositions are computed
velocity Vx is applied from 0 to 100 km depth. and extracted. The extent of melting and residual with either the lherzolite multiple saturation
To conserve mass, an outward Vx is applied from composition of the particles are then updated. surface (when clinopyroxene is present) based
200 to 300 km depth. We use a tanh function We consider that newly generated melts are on the parameterized method of Kinzler and
from 100 to 200 km depth for velocity change instantaneously extracted from the mantle and Grove (1992a, 1992b) or with the harzburgite
from inward to outward. The second condition vertically transferred to the surface to form vol- multiple saturation surface (when clinopyrox-
uses imposed stresses in the subducting plate. canic rocks. At the plate surface, between 175 ene has been completely consumed) based on
We put a constant stress at a fixed position, 700
km from the left boundary, from the surface
down to 50 km depth (Fig. 1). With this stress TABLE 2. MAJOR-ELEMENT COMPOSITIONS (IN WEIGHT PERCENT) OF MELTS WITH
boundary condition, the right boundary is set PRESSURE AS 1.0 GPa AND TEMPERATURE T INCREASING FROM 1500 TO 1680 K
to be free slip, and another weak zone (dashed T (K) Melt extent Melt type* SiO2 Al2O3 MgO FeO TiO2 Na2O K2O CaO
rectangle in Fig. 1, 50 by 100 km) is placed at
DMM† 45.524 3.98 38.73 8.18 0.13 0.28 0.006 3.17
the right end of the subducting plate to detach it
1500 0.008 L 55.25 16.76 7.48 5.98 0.92 5.24 0.63 7.74
from the boundary.
1515 0.024 L 52.57 16.73 8.96 7.06 0.91 4.33 0.06 9.38
1530 0.046 L 50.41 16.56 10.31 7.95 0.84 3.11 0 10.81
Parameterized Model of Mantle Melting 1545 0.072 L 49.79 15.95 11.31 8.50 0.75 1.93 0 11.77
Process 1560 0.101 L 49.37 15.47 12.13 8.85 0.65 1.04 0 12.48
1575 0.134 L 49.12 15.22 12.70 8.97 0.55 0.52 0 12.92
We constructed a parameterized melting 1590 0.169 H 52.70 13.56 15.05 8.37 0.25 0.15 0 9.92
1605 0.207 H 55.43 11.02 17.00 8.50 0.15 0.06 0 7.84
model that can be conveniently implemented
1620 0.226 H 56.39 9.43 18.07 8.82 0.14 0.05 0 7.11
into Ellipsis. Each of the 16 tracking particles 1635 0.233 H 56.72 8.64 18.70 9.02 0.14 0.04 0 6.74
initially inserted in each element represents a 1650 0.242 H 56.76 8.23 19.12 9.16 0.13 0.04 0 6.56
mantle parcel and tracks the extent of melting, 1665 0.253 H 56.89 7.70 19.63 9.32 0.13 0.03 0 6.29
phase proportions, and major- and trace-element 1680 0.266 H 57.03 7.14 20.21 9.48 0.13 0.03 0 5.99
compositions of the parcel. The initial major- *L—lherzolite melting, H—harzburgite melting.
†
and trace-element compositions (in weight per- DMM represents the depleted mid-ocean-ridge basalt (MORB) mantle source (Workman and Hart, 2005).
We only consider eight major elements here and replace others with SiO2.
cent) for these particles are obtained from the

LITHOSPHERE | Volume 4 | Number 6 | www.gsapubs.org 513

LENG ET AL.

TABLE 3. TRACE-ELEMENT COMPOSITIONS (IN PPM) FOR DEPLETED MORB MANTLE (DMM), this method, we obtain a wet mantle zone that
PRIMITIVE MANTLE (PM), AND SLAB-DERIVED FLUID AND PARTITION COEFFICIENTS D*
moves with the subducting plate and has the
Elements Rb Ba Th U Nb Ta La Ce chemical signature of subduction fluid.
D 0.00001 0.00012 0.001 0.0011 0.0034 0.0034 0.01 0.022
DMM 0.05 0.563 0.0079 0.0032 0.1485 0.0096 0.192 0.550 VOLCANIC SEQUENCES ASSOCIATED
PM 0.635 6.989 0.085 0.021 0.713 0.041 0.687 1.775 WITH DIFFERENT SUBDUCTION
Slab fluid 14.7 137.1 1.12 0.44 0.747 0.26 9.9 20.36
INITIATION MODES
Elements Pb Pr Nd Sr Zr Hf Sm Eu
D 0.014 0.027 0.031 0.025 0.033 0.035 0.045 0.050
Mode of Continuous Subduction
DMM 0.018 0.107 0.581 7.664 5.082 0.157 0.239 0.096
PM 0.185 0.276 1.354 21.1 11.2 0.309 0.444 0.168
Initiation with Infant-Arc Spreading
Slab fluid 1.69 0.871 1.47 468.1 98.5 2.55 3.45 0.0
Elements Gd Tb Dy Ho Y Er Yb Lu We start with case A01, for which we set the
D 0.056 0.068 0.079 0.084 0.088 0.097 0.115 0.120 subduction plate age tp and the imposed veloc-
DMM 0.358 0.070 0.505 0.115 3.328 0.348 0.365 0.058 ity Vx to be 82 m.y. and 4 cm/yr, respectively.
PM 0.596 0.108 0.737 0.164 4.55 0.48 0.493 0.074 With plate strength parameters μ0 = 0.2 and εp
Slab fluid 0.0 0.0 0.0 0.0 29.4 0.0 0.0 0.0 = 0.5, we obtain a subduction initiation mode of
Note: We compiled slab-derived fluid composition based on results from Stolper and Newman (1994). continuous initiation with infant-arc spreading
MORB—mid-ocean-ridge basalt. (Leng and Gurnis, 2011). It takes several mil-
*Partition coefficients and depleted MORB mantle (DMM) compositions come from Workman and Hart (2005);
PM composition comes from Sun and McDonough (1989). lion years to underthrust the old plate beneath
the young one (Fig. 2A). After the negative
buoyancy of the subducting plate overcomes
the resistance from plate bending and friction, it
the parameterized method of Parman and Grove For our melting model, the reference tem- begins to founder rapidly, leading to significant
(2004). Therefore, the thickness and composi- perature T0 plays a key role for computing the asthenospheric upwellings (up to several tens of
tions of the surface volcanic rocks associated extent of melting and compositions. We deter- centimeters per year), strong infant-arc spread-
with mantle melting processes can be computed mined T0 using the following approach. Since ing, and trench retreat (Fig. 2B).
from our model. We output these results every the age of the overriding plate in our model is To illustrate the effects of water on the vol-
0.25 m.y. After the melt extraction, the phase just 2 m.y., strong melting and a layer of sur- canism, we first computed melting without
proportions of the source mantle are also com- face volcanic rock are generated at the first time the inclusion of water. During the first several
puted and updated according to lherzolite melt- step when we start the model. We consider that million years of plate convergence, there is no
ing (Kinzler and Grove, 1992a, 1992b) or harz- this layer resembles the oceanic crust formed melt generated, and the mantle melting region
burgite melting (Parman and Grove, 2004). from MORB. When T0 is 1355 °C, the thickness is confined to the initial melting region induced
We computed a test of the major-element of this volcanic layer is 7 km, the same as the below the thin overriding plate (Figs. 2A and
compositions for different extents of melting. typical thickness of oceanic crust (Orcutt et al., 2D). When the slab founders and subduction
With pressure at 1.0 GPa and no water content, 1984). We therefore use this T0 in our melting begins, infant-arc spreading causes adiabatic
we increased the temperature from 1500 to 1680 model. Note that in the following results and melting beneath the spreading center and forms
K with an interval of 15 K. New melts were discussion, this initial non-subduction-related volcanic rocks (Figs. 2B and 2E). Initially,
extracted each time temperature increased. The basement volcanic layer is removed from the these rocks are mostly within group B, the low-
composition of melts varied when the extent of melting results. degree-of-melting (normal) MORB-like rocks,
melting increased from 0.8% to 26.6% (Table The subduction of oceanic crust brings water which move from the spreading center toward
2). MgO continuously increased with extent of into the mantle wedge, which significantly alters the trench with plate motion (Fig. 2E). As the
melting. When the extent of melting was larger the melting process. The modeling of chemical slab continues to founder, the spreading center
than 13%, all clinopyroxene was consumed, and signatures of subduction fluid and water trans- moves with trench retreat, and more extensive
the melting process changed from lherzolite to port in the mantle wedge can be complicated melting generates rocks of group C, high-
harzburgite melting (Table 2). For lherzolite (e.g., Iwamori, 1998; Hebert et al., 2009b). Here, degree-of-melting MORB-like rocks (Figs. 2C
melting, SiO2 and Al2O3 decreased while CaO we use a simple method to model the effects of and 2F). The accumulated thickness of the vol-
increased with extent of melting. However, for water and slab-derived fluids on volcanism asso- canic rocks reaches ~8 km.
harzburgite melting, CaO and Al2O3 decreased ciated with subduction initiation. We consider When the effects of water are included, the
while SiO2 increased with extent of melting. that there is a thin wet layer at the top of the sub- melting extent is enhanced in the mantle wedge
We used the oxide weight percents of SiO2 and ducting plate. The mantle zone extending verti- due to water injection (Figs. 3A, 3B, and 3C).
MgO to divide the resulting volcanic rocks into cally 60 km above the subducting plate is con- The wet mantle zone with 0.3% of water con-
four compositional groups: Group A represents sidered “wet,” with a water content of 0.3%. To tent occurs above the slab, extending 60 km
the very low-degree-of-melting rocks with SiO2 simulate refertilization of the mantle wedge by upward (Fig. 3D). The enhanced melting leads
> 52% and MgO < 10%; group B is the low- slab-derived fluids, the concentrations of trace to much larger thicknesses of volcanic rocks
degree-of-melting (normal) MORB-like rocks elements in this wet mantle zone are updated above the wet zone, up to 15 km (Fig. 3E). The
with SiO2 < 52% and MgO < 13%; group C is at each computation time step to a mixture of horizontal distribution of erupted compositional
the high-degree-of-melting MORB-like rocks 99.7% residual source composition and 0.3% of groups at different time indicates that all the
with SiO2 < 52% and MgO > 13%; and finally, slab-derived fluid composition. The slab-derived eruptions start at ~4.2 m.y. when slab begins to
group D represents boninites with SiO2 > 52% fluid composition (Table 3) was compiled from founder (Fig. 3F). Notably, the mantle source
and MgO > 13%. the results of Stolper and Newman (1994). With entering the mantle wet zone is residual to the

514 www.gsapubs.org | Volume 4 | Number 6 | LITHOSPHERE

 Basalts to boninites: Geodynamics of volcanism during subduction initiation | RESEARCH

A 3.8 m.y. 10 cm/yr D

0 8

Thickness (km)
6
Depth (km)

100

200
2

300 0
200 300 400 500 600 200 300 400 500 600

B 4.6 m.y. E
0 8

Thickness (km)
6
Depth (km)

100

200
2

300 0
200 300 400 500 600 200 300 400 500 600

5.7 m.y. F
C 0 8
Thickness (km)

6
Depth (km)

100

200
2

300 0
200 300 400 500 600 200 300 400 500 600

Horizontal distance (km) Horizontal position (km)

0 200 400 600 800 1000 1200 1400 A B C D

Temperature (°C) Compositional group
Figure 2. (A–C) The temperature and velocity field for case A01, the mode of continuous subduction initiation with infant-arc spreading, at
3.8, 4.6, and 5.7 m.y. after plate convergence starts. The white lines are the contours of extent of melting without consideration of water. The
contour interval is 5%. (D–F) The surface volcanic rock sequences corresponding to A, B, and C. Colors show different compositional groups
(see text for classification scheme). The horizontal bars in each volcanic column show the time interval for every 0.25 m.y.

LITHOSPHERE | Volume 4 | Number 6 | www.gsapubs.org 515

LENG ET AL.

Temperature Temperature
A 3.8 m.y. 10 cm/yr B 4.7 m.y. °C
0 1400
1200
Depth (km)

1000
100
800
600
200 400
200
300 0
200 300 400 500 600 200 300 400 500 600
Horizontal distance (km) Horizontal distance (km)
Temperature Water content
C 5.8 m.y. D 5.8 m.y. (%)
0 0.4
Depth (km)

0.3
100
0.2

200
0.1

300 0.0
200 300 400 500 600 200 300 400 500 600
Horizontal distance (km) Horizontal distance (km)

E F
16
Eruption time (m.y.)

14 6
Thickness (km)

12
10
8 5
6
4
2
0 4
200 300 400 500 600 200 300 400 500 600
Horizontal position (km) Horizontal position (km)

A B C D A B C D
Compositional group Compositional group
Figure 3. (A–C) The temperature and velocity field for case A01 at 3.8, 4.7, and 5.8 m.y. with consideration of water effects. The white lines are
the contours of extent of melting with interval of 5%. (D) The water contents due to the fluid released from the slab at 5.8 m.y. for case A01. The
green color shows the thin wet layer at the top of the slab. (E) The surface volcanic rock sequences corresponding to C. Legends are the same as
Figure 2F. (F) The horizontal distribution of erupted compositional groups at different times (every rectangle occupies a time interval of 0.25 m.y.).

516 www.gsapubs.org | Volume 4 | Number 6 | LITHOSPHERE

 Basalts to boninites: Geodynamics of volcanism during subduction initiation | RESEARCH
melt extraction beneath the infant-arc spreading Figure 4. (A) The compositional
center and is remelted due to water injection into A Compositional group groups and thicknesses for the
this zone. This process characteristically gener- 16 sequence shown at the sam-
ates boninites, which are erupted at the surface D pling point at 500 km horizontal
14 Contours: 0.25 m.y.
distance in Figure 3E. The hori-

Thickness (km)
near the trench (Figs. 3E and 3F). The sites with 12 zontal contours in each column
boninite eruption move with trench retreat. On C 10 show the time interval for every
closer inspection of the rock sequence at the 8 0.25 m.y. (B) The major-element
sampling point that ends up at position 500 km B 6 compositions and Mg# for dif-
in Figure 3E (between the spreading center and ferent compositional groups in
4 A. There are small differences
the trench), we find a prominent stratigraphic A 2 of compositions for different
sequence for the volcanic rocks, changing from
0 time intervals in each group.
group B to C and then to D within a few million Here, we show the composi-
years (Fig. 4A). The major-element composi-
tions of these rocks show the transition from B Major element tions for the first time interval
from the bottom in each group.
normal MORB-like rocks to boninites (Fig. 4B). (in wt%) SiO2 Al2O3 MgO FeO TiO2 Na2O K2O CaO (Mg#) We also show typical major-
The Mg# increases with this transition from element compositions for
Group A: 54.79 17.62 7.78 6.23 0.91 5.10 1.16 6.42 ( 69.2) compositional group A, which
70.2 to 76.5 (Fig. 4B). The trace-element com-
Group B: 49.35 17.28 11.06 8.46 0.82 3.04 0.05 9.95 ( 70.2) does not appear in A. (C) The
positions for group B, which are normalized by trace-elements concentrations
primitive mantle (PM) compositions (Sun and Group C: 48.41 15.60 14.02 9.37 0.56 1.16 0.01 10.87 ( 72.9)
(normalized by primitive man-
McDonough, 1989), show similar patterns to Group D: 52.57 12.16 16.17 8.94 0.27 0.18 0.00 9.70 ( 76.5) tle [PM] concentrations) for the
the normal MORB rocks (Fig. 4C; compared compositional groups shown in
with Hofmann, 1997). For groups C and D, the A. The colored circles, squares,
trace elements are more depleted for compatible C Trace element and triangles represent groups
B, C, and D, respectively. The
elements but are enriched for elements that are black circles and triangles rep-
rich in slab-derived fluid (Fig. 4C). 100
(PM normalized)

resent the averaged mid-ocean-

Concentrations

Starting with case A01 and holding plate ridge basalt (MORB) composi-
strength parameters μ0 and εp constant, we vary 10 tion from Hofmann (1997) and
the subduction plate age tp between 82 and 42 the boninite composition at
1
m.y. and the imposed velocity Vx between 4 the Izu-Bonin-Mariana forearc
0.1 region from Reagan et al. (2010),
and 2 cm/yr and obtain eight additional cases.
Rb Ba Th U Nb Ta La Ce Pb Pr Nd Sr Zr Hf Sm Eu Gd Tb Dy Ho Y Er Yb Lu respectively.
The results show that the subduction initiation
modes for all nine cases are continuous initiation
with infant-arc spreading. The surface volcanic
sequences typically include boninite and show
the transition between different groups (Fig. 5).
However, two cases with Vx = 2 cm/yr and tp = 62 Contours: 0.25 m.y.

and 82 m.y. generate no boninites. To investigate

the missing boninite formation, we place a track- 4.0
ing particle initially at 35 km in depth and 350 cm/yr
Scale
km in horizontal distance for both case A01 and
4 Figure 5. The typical surface volcanic
case A01a, which has Vx = 2 cm/yr and tp = 62 km sequences including boninite (if any)
m.y. The paths of the tracking particle during the generated from cases similar to case
whole process for these two cases (black curves A01 but with different subduction plate
in Figs. 6A and 6B) show that mantle material age tp and imposed velocity Vx. The
beneath the spreading center moves toward the horizontal contours in each column
3.0 show the time interval for every 0.25
slab when subduction initiates with slab founder- cm/yr m.y. For compositional group A, the
ing and trench retreat. For case A01, the depleted
contours are too crowded to show, so
material beneath the spreading center reaches we show the total time for obtaining
the surface of the slab. The injection of water the whole thickness of group A instead
from the slab remelts the depleted material, and of showing contours for every 0.25 m.y.
boninite is generated. For case A01a, however, 2.0
the combined effects of decreased plate veloc- cm/yr
ity and increased plate age delay the starting
of infant-arc spreading. The subducting plate 3.75 m.y. 2.25 m.y.
reaches the bottom of the model domain shortly
Vx
after infant-arc spreading starts. The depleted
material beneath the spreading center does not 42 m.y. 62 m.y. 82 m.y.
reach the wet zone above the slab before the slab tp
is stopped by the bottom boundary. Therefore, A B C D
no boninites are observed (Fig. 5). Compositional group

LITHOSPHERE | Volume 4 | Number 6 | www.gsapubs.org 517

LENG ET AL.

Segmented Subduction Initiation Mode

Temperature
With plate strength parameters μ0 = 0.1 and A case A01 at 5.7 m.y. °C
εp = 0.2, we have a weak plate, which leads to a 0 1400
segmented mode of subduction initiation (Leng
and Gurnis, 2011). We set the subduction plate 1200
age tp and the imposed velocity Vx to be 42 m.y.

Depth (km)
1000
and 2 cm/yr, respectively (case A02). After 100
several million years of plate convergence, the 800
subducting plate bends and infant-arc spreading 600
occurs (Fig. 7A). However, unlike in the mode 200
of continuous subduction initiation with infant- 400
arc spreading, the weak plate cannot sustain 200
the slab pull forces, and the subducting plate
breaks to form a slab segment (Fig. 7A). As 300 0
plate convergence continues, another slab seg- 200 300 400 500 600
ment breaks from the plate (Fig. 7B). During the Horizontal distance (km)
whole process, significant melting occurs, and
large volumes of volcanic rocks form. However,
the weak plate leads to rapid vertical descent B case A01a at 9.5 m.y. °C
of the broken slab segments, and the wet zone 0 1400
due to water release from the slab segments
only exists in a small region (Figs. 7A and 7B). 1200
Therefore, extensive melting and thick volcanic
Depth (km)

1000
sequences including boninites only occur close 100
to the trench; this zone moves with the trench 800
as it rolls back (Figs. 7C and 7D). Starting with 600
case A02 and holding plate strength parameters 200
μ0 and εp constant, we vary the subduction plate 400
age tp between 82 and 42 m.y. and the imposed 200
velocity Vx between 4 and 2 cm/yr, and the
resulting subduction initiation modes are all 300 0
segmented, with similar volcanic expression as 200 300 400 500 600
shown for case A02.
Horizontal distance (km)
Mode of Continuous Subduction Figure 6. The temperature field and melting extent contour for case A01 at
5.7 m.y. (A) and case A01a at 9.5 m.y. (B). The contour interval is 5%. The
Initiation Without Infant-Arc Spreading
black curves show the paths of the tracking particle from 0 m.y. to the final
time. Initially, the tracking particle is located at 35 km depth and 350 km in
High plate strength leads to continuous horizontal distance. Case A01a is similar to case A01, except that we change
subduction initiation with neither infant-arc subduction plate age tp = 62 m.y. and imposed velocity Vx = 2 cm/yr.
spreading nor trench retreat (Leng and Gurnis,
2011). With tp = 42 m.y. and Vx = 2 cm/yr, we
increase plate strength to μ0 = 0.6 and εp = 0.4
(case A03). Until 6.8 m.y. after plate conver- volcanic rocks at this site after eruption starts The subduction initiation in cases A01
gence begins (Fig. 8A), very few volcanic rocks (Fig. 8D). We also hold plate strength param- through A03 is induced by an imposed constant
form (less than 0.5 km in thickness from 0 to eters μ0 and εp constant for case A03 and vary velocity on the subducting plate. Subduction
6.8 m.y.) (Fig. 8C). Most of these rocks are of the subduction plate age tp between 82 and 42 initiation may also be induced by a constant
group A, very low-degree-of-melting rock (Fig. m.y. and the imposed velocity Vx between 4 and stress applied on the subducting plate (Toth and
8C). The typical major-element concentrations 2 cm/yr. For the case with tp = 42 m.y. and Vx = 4 Gurnis, 1998; Leng and Gurnis, 2011). With tp =
of group A are characterized by high SiO2 and cm/yr, the resulting subduction initiation mode 42 m.y., and the same plate strength parameters
Al2O3, but low MgO (Fig. 4B). After several is the same as case A03, which leads to similar as we chose for case A01, μ0 = 0.2 and εp = 0.5,
more million years of convergence, the subduct- volcanic expression, except that the correspond- we change from the imposed velocity boundary
ing plate bends, with relatively weak astheno- ing time scales are approximately halved due condition to the imposed stress boundary condi-
spheric upwelling beneath the overriding plate to increased plate velocity. For the cases with tion (Fig. 1) for case A04. A tectonic stress of
(Fig. 8B). The upwelling causes melting, but tp = 82 m.y. and Vx = 2 and 4 cm/yr, due to the 80 MPa is applied within the old plate from the
the melting region is confined to a small mantle increased subducting plate age, the subduction surface down to 50 km depth (Fig. 1), which is
region where water is released from the slab sur- initiation modes change to continuous with equivalent to a tectonic force of 4 × 1012 N/m.
face (Fig. 8B). Since there is no trench migra- infant-arc spreading (Leng and Gurnis, 2011) This stress boundary condition leads to a sub-
tion, the site of extensive volcanic eruption does and have similar volcanic expression as shown duction initiation mode similar to case A03.
not move, which causes thick accumulation of for case A01. With such a force, the subducting plate accel-

518 www.gsapubs.org | Volume 4 | Number 6 | LITHOSPHERE

 Basalts to boninites: Geodynamics of volcanism during subduction initiation | RESEARCH

5.2 m.y. 10 cm/yr

A 0
C 16
14

Thickness (km)
12
Depth (km)

100
10
8
6
200
4
2
300 0
200 300 400 500 600 200 300 400 500 600

B 5.8 m.y. D
0 16
14

Thickness (km)
12
Depth (km)

100
10
8
6
200
4
2
300 0
200 300 400 500 600 200 300 400 500 600

Horizontal distance (km) Horizontal distance (km)

°C
0 200 400 600 800 1000 1200 1400 A B C D
Temperature Compositional group
Figure 7. (A–B) The temperature, velocity, and melting extent contours at 5.2 and 5.8 m.y. for case A02, the segment subduction initiation
mode. Legends are the same as Figure 2C. (C–D) The surface volcanic rock sequences corresponding to A and B. Legends are the same as
Figure 2F.

erates during plate convergence. It only takes the water released from the slab causes large model domain becomes slightly longer, ~3.0
1.5 m.y. for the plate to reach the bottom of the melting extents close to the trench, which can m.y. If this force is applied uniformly between
model domain (Figs. 9A and 9B). Note that here remelt the depleted mantle due to the initially 0 and 100 km depths, the resulting stress of 40
the mantle melting extent due to the initially thin thin overriding plate (cf. Figs. 9A and 9B for the MPa is smaller than the initial cohesion C0 and
overriding plate is larger than in previous cases, melting contour). Therefore, a large volume of cannot induce subduction initiation.
because the mantle has less time to cool. There boninite is generated.
is no infant-arc spreading associated with this The typical ridge push force is estimated to GEOLOGICAL APPLICATIONS OF
subduction initiation process. The surface vol- be ~4 × 1012 N/m (McKenzie, 1977). Here, we DIFFERENT MODES OF SUBDUCTION
canic rock patterns are similar to case A03 (Figs. apply such a force uniformly between 0 and 50 INITIATION
9C and 9D). Moderate melting occurs beneath km depths, which leads to a constant stress of
the overriding plate, while extensive melting 80 MPa. Applying this force uniformly between The Izu-Bonin-Mariana subduction zone
and boninites form close to the trench (Fig. 9D). 0 and 67 km depths leads to a stress of 60 MPa. initiated at ca. 50 Ma with strong infant-arc
There is little horizontal transport of depleted This stress causes subduction initiation with spreading and voluminous boninite erup-
mantle beneath the overriding plate for this case. volcanism similar to case A04, except that the tions (Stern and Bloomer, 1992; Reagan et al.,
However, since the mantle has less time to cool, time for the slab to reach the bottom of the 2010) and may be associated with the change

LITHOSPHERE | Volume 4 | Number 6 | www.gsapubs.org 519

LENG ET AL.

A 6.8 m.y. 10 cm/yr C

0 2

Thickness (km)
Depth (km)

100
1
200

300 0
200 300 400 500 600 200 300 400 500 600

10.7 m.y.
B 0 D 12
10

Thickness (km)
Depth (km)

100 8
6
200 4
2
300 0
200 300 400 500 600 200 300 400 500 600
Horizontal distance (km) Horizontal distance (km)
°C
0 200 400 600 800 1000 1200 1400 A B C D
Temperature Compositional group
Figure 8. (A–B) The temperature, velocity, and melting extent contours at 6.8 and 10.7 m.y. for case A03, the mode of continuous
subduction initiation without infant-arc spreading. Legends are the same as Figure 2C. (C–D) The surface volcanic rock sequences
corresponding to A and B. Legends are the same as Figure 2F.

in direction of the Pacific plate from northward observed at Izu-Bonin-Mariana forearc region Bonin Islands (Ishizuka et al., 2011), suggesting
to westward (Sharp and Clague, 2006). Case (Fig. 4C), indicating that mantle source in Izu- that a significant portion (more than 1500 km
A01, having a mode of continuous subduction Bonin-Mariana region may be more depleted in width) of Pacific plate foundered during Izu-
initiation with infant-arc spreading, provides an than the averaged mantle composition consid- Bonin-Mariana subduction initiation and caused
evolutionary pathway similar to the sequence ered in our model. In the volcanic sequences infant-arc spreading and strong volcanism along
of tectonism and volcanism of the juvenile Izu- obtained (Figs. 4 and 5), the rock composition the whole Izu-Bonin-Mariana forearc region.
Bonin-Mariana subduction zone. After several typically changes from MORB-like to boninite One limitation is that the age of the modeled
million years of plate convergence, the subduct- (if any) within a few million years. The thick- overriding plate is only a few million years,
ing plate founders and causes strong infant-arc ness of these sequences varies from several to which would be consistent with oceanic litho-
spreading and volcanic eruptions. In particular, more than 10 km. We consider such volcanic sphere near a spreading center. However, the
this mode of subduction initiation produces sequences to be characteristic of the volcanic Mariana and Bonin sections are about ~1500
volcanic rocks that are similar to the observed expression related to the mode of continuous km from each other, and there is no evidence
sequences, including the order, thicknesses, and subduction initiation with infant-arc spread- that both sections are adjacent to a spreading
major- and trace-element compositions (Fig. 4) ing. They indicate the occurrence of slab foun- center perpendicular to a nascent trench.
(Reagan et al., 2010; for observed units, see fig. dering, infant-arc spreading, and remelting of The New Hebrides subduction zone may
4 in Ishizuka et al., 2011). The concentrations of depleted mantle due to water release from the have initiated with a segmented subduction
highly incompatible trace elements in boninite slab. Such sequences are observed at both the initiation mode similar to case A02. The New
obtained from our model are larger than those Mariana forearc (Reagan et al., 2010) and the Hebrides subduction zone initiated through a

520 www.gsapubs.org | Volume 4 | Number 6 | LITHOSPHERE

 Basalts to boninites: Geodynamics of volcanism during subduction initiation | RESEARCH

A 1.0 m.y. 10 cm/yr C

0
14

Thickness (km)
12
Depth (km)

100 10
8
6
200
4
2
300 0
200 300 400 500 600 200 300 400 500 600

1.5 m.y.
B 0 D
14

Thickness (km)
12
Depth (km)

100 10
8
6
200
4
2
300 0
200 300 400 500 600 200 300 400 500 600

Horizontal distance (km) Horizontal distance (km)

°C
0 200 400 600 800 1000 1200 1400
A B C D
Temperature Compositional group
Figure 9. (A–B) The temperature, velocity, and melting extent contours at 1.0 and 1.5 m.y. for case A04, subduction initiation
induced by an imposed constant stress. Legends are the same as Figure 2C. (C–D) The surface volcanic rock sequences corre-
sponding to A and B. Legends are the same as Figure 2F.

polarity reversal at ca. 10–12 Ma (Greene et al., Plate convergence has been occurring con- which produces rocks mostly in group A, very
1994), but a slab segment is found beneath the tinuously since ca. 15 Ma at the Puysegur sub- low-degree-of-melting rocks, for the first sev-
North Fiji Basin (Hamburger and Isacks, 1987; duction zone south of New Zealand, which nev- eral million years. Nevertheless, the volcanism
Richards et al., 2011). Richards et al. (2011) ertheless shows neither infant-arc spreading nor in case A03 will eventually be extensive after
suggested that this slab segment occurred by significant volcanism. Therefore, this subduc- ~10 m.y. of plate convergence. We thus suspect
the tearing away of the deeper portion from tion zone may have been initiated with the mode that this indicates possibly strong volcanic erup-
the shallower portion of the slab at ca. 5 Ma, of continuous subduction initiation without tions in the Puysegur subduction zone in the
or ~5–7 m.y. after the subduction initiation infant-arc spreading, as in case A03. This mode future, although the region of eruption would be
event. The subducted plate at the New Hebri- can explain the observed scarce distribution confined to a narrow region close to the trench.
des Trench is continuous with the crust of the of volcanism in the Puysegur subduction zone The Aleutian subduction zone initiated at
South Fiji Basin (Seton et al., 2012). Therefore, so far. Moreover, the volcanic rocks found in approximately the same time as Izu-Bonin-
it is possible that this plate formed within a Solander Island in this zone contain mostly ada- Mariana, ca. 50 Ma. The subducting plate in the
backarc basin and remains mechanically weak kite and andesite (Reay and Parkinson, 1997). Aleutians is the Kula plate, a small plate con-
due to high volatile concentrations. Our model No basalts are reported. There is little to no gar- fined by the Kula-Pacific ridge a few thousand
suggests that if the New Hebrides subduction net involved in the signature. Therefore, these kilometers to the south (Seton et al., 2012). This
initiation is in the segmented mode, we expect rocks are probably derived from a low degree of plate geometry can lead to a constant ridge push
to observe thick volcanic products close to the partial melt from the wedge (Foley et al., 2011). force from the Kula-Pacific ridge during the for-
trench that possibly include boninites. This is consistent with results from case A03, mation of Aleutian subduction zone. Therefore,

LITHOSPHERE | Volume 4 | Number 6 | www.gsapubs.org 521

LENG ET AL.

Aleutian subduction initiation may be driven by spreading; such a mode leads to limited, very thermodynamics: Earth and Planetary Science Letters,
v. 278, p. 243–256, doi:10.1016/j.epsl.2008.12.013.
a constant stress from mid-ocean-ridge push. low degrees of melting occurring during a long Hebert, L.B., Asimow, P., and Antoshechkina, P., 2009b, Fluid
Relatively strong volcanic eruptions have been interval of plate convergence before subduction source-based modeling of melt initiation within the
reported during the formation of the Aleutian initiation starts, although extensive melting near subduction zone mantle wedge: Implications for geo-
chemical trends in arc lavas: Chemical Geology, v. 266,
subduction zone (Jicha et al., 2006), while no the trench is still possible when subduction initi- p. 297–310, doi:10.1016/j.chemgeo.2009.06.017.
infant-arc spreading has been inferred for the ation starts after a protracted period of plate con- Hofmann, A.W., 1997, Mantle geochemistry: The message
from oceanic volcanism: Nature, v. 385, p. 219–229,
Bering Sea. There are no reports of boninite vergence (~10 m.y.). If the subduction initiation doi:10.1038/385219a0.
along the Aleutian arc. However, if case A04 is is driven by constant stresses such as through House, M.A., Gurnis, M., Kamp, P.J.J., and Sutherland, R.,
applicable, this suggests that a large volume of ridge push, the slab subducts rapidly in response 2002, Uplift in the Fiordland region, New Zealand:
Implications for incipient subduction: Science, v. 297,
boninite may exist in the forearc, on the overrid- to continuous acceleration of the plate under p. 2038–2041, doi:10.1126/science.1075328.
ing plate close to the Aleutian trench, and might action of the far-field push; significant melting, Ishizuka, O., Tani, K., Reagan, M.K., Kanayama, K., Umino, S.,
be found by drilling in this area. With neither Harigane, Y., Sakamoto, I., Miyajima, Y., Yuasa, M., and
including boninite eruption, can be generated
Dunkley, D.J., 2011, The timescales of subduction initia-
slab foundering nor infant-arc spreading, a vol- within a few million years with no trench migra- tion and subsequent evolution of an oceanic island arc:
canic sequence that changes from MORB-like tion. Based on the tectonic and volcanic evolu- Earth and Planetary Science Letters, v. 306, p. 229–240,
to boninite is not found for this mode (Fig. 9D). doi:10.1016/j.epsl.2011.04.006.
tion, these different modes may be applicable to Iwamori, H., 1998, Transportation of H2O and melting in sub-
the initiation of the Izu-Bonin-Mariana arc, the duction zones: Earth and Planetary Science Letters,
DISCUSSION AND CONCLUSION New Hebrides arc, the Puysegur Trench in New v. 160, p. 65–80, doi:10.1016/S0012-821X(98)00080-6.
Jicha, B.R., Scholl, D.W., Singer, B.S., and Yogodzinski, G.M.,
Zealand, and the Aleutian Trench, respectively. 2006, Revised age of Aleutian island arc formation
Our model indicates that the initial coeffi- implies high rate of magma production: Geology, v. 34,
cient of friction plays an important role in deter- ACKNOWLEDGMENTS p. 661–664, doi:10.1130/G22433.1.
Katz, R.F., Spiegelman, M., and Langmuir, C.H., 2003, A new
mining the mode of subduction initiation. The parameterization of hydrous mantle melting: Geo-
initial coefficient of friction may not vary much We thank Bob Stern and two anonymous chemistry Geophysics Geosystems, v. 4, no. 9, 1073,
for different rocks. However, in Equation 3 for doi:10.1029/2002GC000433.
reviewers, whose comments help to improve the Kinzler, R.J., and Grove, T.L., 1992a, Primary magmas of
yielding, we did not consider the effect of pore manuscript. Leng was supported by the O.K. mid-ocean ridge basalts: 1. Experiments and methods:
fluid pressure, which can significantly reduce Earl Fellowship at Caltech. Additional support Journal of Geophysical Research, v. 97, p. 6885–6906,
doi:10.1029/91JB02840.
the yield stress. Therefore, the coefficient of was provided by the National Science Founda- Kinzler, R.J., and Grove, T.L., 1992b, Primary magmas of mid-
friction in our model represents an effective tion (EAR-0810303) and the Caltech Tectonics ocean ridge basalts: 2. Applications: Journal of Geophys-
value that could vary significantly between 0.0 Observatory (by the Gordon and Betty Moore ical Research, v. 97, p. 6907–6926, doi:10.1029/91JB02841.
Leng, W., and Gurnis, M., 2011, Dynamics of subduction
and 0.6 for different locations due to the pos- Foundation) with contribution number TO 209. initiation with different evolutionary pathways: Geo-
sible existence of different pore fluid pressures. chemistry Geophysics Geosystems, v. 12, Q12018,
Water transport, melt generation, and melt doi:10.1029/2011GC003877.
REFERENCES CITED McKenzie, D.P., 1977, The initiation of trenches: A finite ampli-
migration in subduction zones can each be tude instability, inTalwani, M., and Pitman, W.C., III, eds.,
complicated (e.g., Iwamori, 1998; Hebert et al., Asimow, P.D., Dixon, J.E., and Langmuir, C.H., 2004, A Island Arcs, Deep Sea Trenches, and Back-Arc Basins:
hydrous melting and fractionation model for mid-ocean Washington, D.C., American Geophysical Union, Mau-
2009b). Here, we consider a wet zone with uni- ridge basalts: Application to the Mid-Atlantic Ridge near rice Ewing Series, v. 1, p. 57–61.
formly distributed water content above the slab the Azores: Geochemistry Geophysics Geosystems, Moresi, L., Dufour, F., and Muhlhaus, H.B., 2003, A Lagrang-
v. 5, Q01E16, doi:10.1029/2003GC000568.
and simplify the treatment of slab-derived trace ian integration point finite element method for large
Buck, W.R., and Poliakov, A.N.B., 1998, Abyssal hills formed deformation modeling of viscoelastic geomaterials:
elements in the melting region. We ignore the by stretching oceanic lithosphere: Nature, v. 392, Journal of Computational Physics, v. 184, p. 476–497,
consequences of melting for the energy budget, p. 272–275, doi:10.1038/32636. doi:10.1016/S0021-9991(02)00031-1.
Crameri, F., Tackley, P.J., Meilick, I., Gerya, T.V., and Kaus, B.J.P.,
the effect of latent heat on the thermal field, and 2012, A free plate surface and weak oceanic crust produce
Orcutt, J.A., McClain, J.S., and Burnett, M., 1984, Evolution
of the ocean crust: Results from recent seismic experi-
the feedback to melt generation. Also, we con- single-sided subduction on Earth: Geophysical Research ments, in Gass, I.G., Lippard, G.J., and Shelton, A.W.,
sider that melts are extracted vertically from the Letters, v. 39, L03306, doi:10.1029/2011GL050046. eds., Ophiolites and Oceanic Lithosphere: Geological
Foley, F.V., Pearson, N.J., Rushmer, T.A., Adam, J., and Turner, Society of London Special Publication 13, p. 7–16.
mantle to the surface immediately after they are S., Petrogenetic constraints on the origin and evolution Parman, S.W., and Grove, T.L., 2004, Harzburgite melting
generated. In short, the detailed migration and of the volcanic rocks on Solander Island, New Zea- with and without H2O: Experimental data and predic-
fractionation processes of melts and water trans- land: American Geophysical Union, Fall Meeting 2011, tive modeling: Journal of Geophysical Research, v. 109,
abstract V43A-2553. B02201, doi:10.1029/2003JB002566.
port in the mantle wedge are ignored. So, our Greene, H.G., Collot, J.Y., Fisher, M.A., and Crawford, A.J., Poliakov, A.N.B., and Buck, W.R., 1998, Mechanics of stretch-
results only reflect a first-order approximation of 1994, Neogene tectonic evolution of the New Hebrides ing elastic-plastic-viscous layers: Applications to slow-
magma compositions, spatial distribution, and island arc: A review incorporating ODP drilling results, spreading mid-ocean ridges, in Buck, W.R., et al., eds.,
in Proceedings of the Ocean Drilling Program, Scientific Faulting and Magmatism at Mid-Ocean Ridges: Ameri-
eruption timing during subduction initiation. Results, Volume 134: College Station, Texas, Ocean Drill- can Geophysical Union Geophysical Monograph 106,
In summary, we find that with a mode of ing Program, p. 19–46. p. 305–325.
Gurnis, M., Hall, C., and Lavier, L., 2004, Evolving force balance Reagan, M.K., and 15 others, 2010, Fore-arc basalts and
continuous initiation with infant-arc spread-
during incipient subduction: Geochemistry Geophysics subduction initiation in the Izu-Bonin-Mariana system:
ing, the foundering of the subducting slab, and Geosystems, v. 5, Q07001, doi:10.1029/2003GC000681. Geochemistry Geophysics Geosystems, v. 11, Q03X12,
water release from the slab govern a succession Hall, C.E., Gurnis, M., Sdrolias, M., Lavier, L.L., and Muller, doi:10.1029/2009GC002871.
R.D., 2003, Catastrophic initiation of subduction follow- Reay, A., and Parkinson, D., 1997, Adakites from Solander
from MORB-like composition to boninites. The ing forced convergence across fracture zones: Earth and Island, New Zealand: New Zealand Journal of Geology
modeled composition transition typically occurs Planetary Science Letters, v. 212, p. 15–30, doi:10.1016 and Geophysics, v. 40, p. 121–126, doi:10.1080/0028830
within a few million years. When plate strength /S0012-821X(03)00242-5. 6.1997.9514746.
Hamburger, M.W., and Isacks, B.L., 1987, Deep earthquakes Richards, S., Holm, R., and Barber, G., 2011, When slabs col-
is reduced, the subducting slab tends to segment, in the southwest Pacific: A tectonic interpretation: Jour- lide: A tectonic assessment of deep earthquakes in
with extensive melting occurring when the slab nal of Geophysical Research, v. 92, p. 13,841–13,854, the Tonga-Vanuatu region: Geology, v. 39, p. 787–790,
breaks; most melting occurs close to the trench. doi:10.1029/JB092iB13p13841. doi:10.1130/G31937.1.
Hebert, L.B., Antoshechkina, P., Asimow, P., and Gurnis, M., Schilling, J.-G., 1966, Rare earth fractionation in Hawaii volca-
When plate strength increases, subduction ini- 2009a, Emergence of a low-viscosity channel in sub- nic rocks [Ph.D. thesis]: Boston, Massachusetts Institute
tiation becomes continuous without infant-arc duction zones through the coupling of mantle flow and of Technology, 390 p.

522 www.gsapubs.org | Volume 4 | Number 6 | LITHOSPHERE

 Basalts to boninites: Geodynamics of volcanism during subduction initiation | RESEARCH
Seton, M., Muller, R.D., Zahirovic, S., Gaina, C., Torsvik, T., Stolper, E., and Newman, S., 1994, The role of water in the osphere Dynamics: Dollnsee, Germany, EGU General
Shephard, G., Talsma, A., Gurnis, M., Turner, M., Maus, petrogenesis of Mariana Trough magmas: Earth and Assembly 2010, p. 2322.
S., and Chandler, M., 2012, Global continental and ocean Planetary Science Letters, v. 121, p. 293–325, doi:10.1016 Toth, J., and Gurnis, M., 1998, Dynamics of subduction initia-
basin reconstructions since 200 Ma: Earth-Science /0012-821X(94)90074-4. tion at pre-existing fault zones: Journal of Geophysical
Reviews, v. 113, p. 212–270, doi:10.1016/j.earscirev Sun, S.S., and McDonough, W.F., 1989, Chemical and isotopic Research, v. 103, p. 18,053–18,067, doi:10.1029/98JB01076.
.2012.03.002. systematics of oceanic basalts: Implications for mantle Workman, R.K., and Hart, S.R., 2005, Major and trace element
Sharp, W.D., and Clague, D.A., 2006, 50-Ma initiation of composition and processes, in Saunders, A.D., ed., composition of the depleted MORB mantle (DMM):
Hawaiian-Emperor bend records major change in Pacific Magmatism in the Ocean Basins: Geological Society of Earth and Planetary Science Letters, v. 231, p. 53–72,
plate motion: Science, v. 313, p. 1281–1284, doi:10.1126 London Special Publication 42, p. 313–345. doi:10.1016/j.epsl.2004.12.005.
/science.1128489. Sutherland, R., Gurnis, M., Kamp, P.J.J., and House, M.A.,
Stern, R.J., 2004, Subduction initiation: Spontaneous and 2009, Regional exhumation history of brittle crust dur-
induced: Earth and Planetary Science Letters, v. 226, ing subduction initiation, Fiordland, southwest New
p. 275–292. Zealand, and implications for thermochronologic sam-
Stern, R.J., and Bloomer, S.H., 1992, Subduction zone infancy: pling and analysis strategies: Geosphere, v. 5, p. 409– MANUSCRIPT RECEIVED 29 MARCH 2012
Examples from the Eocene Izu-Bonin-Mariana and 425, doi:10.1130/GES00225.1. REVISED MANUSCRIPT RECEIVED 20 AUGUST 2012
Jurassic California arcs: Geological Society of America Thielmann, M., and Kaus, B.J.P., 2011, Shear heating and sub- MANUSCRIPT ACCEPTED 28 SEPTEMBER 2012
Bulletin, v. 104, p. 1621–1636, doi:10.1130/0016-7606 duction initiation, in Abstracts for the 12th International
(1992)104<1621:SZIEFT>2.3.CO;2. Workshop on Modeling of Mantle Convection and Lith- Printed in the USA

LITHOSPHERE | Volume 4 | Number 6 | www.gsapubs.org 523

View publication stats