A geodynamic model of subduction evolution and slab detachment to explain Australian plate acceleration and deceleration during the latest Cretaceous–early Cenozoic

W.P. Schellart1,* 1DEPARTMENT OF EARTH SCIENCES, VRIJE UNIVERSITEIT AMSTERDAM, DE BOELELAAN 1085, 1081 HV AMSTERDAM, NETHERLANDS

ABSTRACT

During the latest Cretaceous–early Cenozoic, the northern margin of the Australian plate was characterized by a large (4000 km wide) north- to northeast-dipping subduction zone (New Guinea–Pocklington subduction zone) consuming a marginal basin. Geological and geophysical data imply that the subduction zone was active ca. 71–50 Ma, and suggest that it was responsible for plate acceleration from ~1.0 to ~7.3 cm/yr ca. 64–59 Ma, and plate deceleration from ~7.3 to ~0.3 cm/yr at 52–49 Ma. This paper presents a numerical model of buoyancy-driven subduction to test if the rates of Australian plate acceleration and deceleration can be ascribed to the progressive evolution of a subducting slab. The geodynamic model reproduces the first-order plate velocity evolution of the Australian plate, with a transient ~5 m.y. period of acceleration from 2 to 8 cm/yr during upper mantle slab lengthening, an ~5 m.y. period of rapid plate motion (~5–8 cm/yr), and a short, 3.9 m.y., period of plate deceleration, starting with a 2 cm/yr velocity drop during 3.1 m.y. of continental subduction and followed by ~0.8 m.y. of rapid deceleration (4 cm/yr velocity drop) during slab detachment. The geodynamic model demonstrates that plate velocity increases or decreases of ~4–6 cm/yr can occur over a period lasting <1 m.y. to a few million years, comparable to what is observed for the latest Cretaceous–early Cenozoic evolution of the Australian plate. Such rates of plate acceleration and deceleration could be tested against plate kinematic data for other subduction settings on Earth.

LITHOSPHERE; v. 9; no. 6; p. 976–986 | Published online 11 October 2017 https://​ doi​ .org​ /10​ .1130​ /L675​ .1​

INTRODUCTION 1983; Goes et al., 2008), slab width (Schellart et al., 2010), plate circum- ference attached to a slab (Forsyth and Uyeda, 1975; Gripp and Gordon, The Earth’s lithosphere is segmented into numerous larger and smaller 2002), double subduction (Jagoutz et al., 2015), and the transition from plates (Bird, 2003) that move at a wide variety of plate tectonic speeds, normal (i.e., oceanic) subduction to collision and/or continental subduc- to as much as ~10 cm/yr. The velocities of these plates have traditionally tion (e.g., Patriat and Achache, 1984; Klootwijk et al., 1992; Molnar and been calculated in hotspot reference frames, such as the Pacific (Minster Stock, 2009; van Hinsbergen et al., 2011; White and Lister, 2012; Zahi- and Jordan, 1978; Gripp and Gordon, 2002; Wessel and Kroenke, 2008), rovic et al., 2012; Capitanio and Replumaz, 2013), the models generally Indo-Atlantic (e.g., O’Neill et al., 2005), or global (e.g., Gordon and Jurdy, have the common element of subduction to explain rapid plate motion 1986; Doubrovine et al., 2012) hotspot reference frames, as well as no- (e.g., Conrad and Lithgow-Bertelloni, 2002). An important exception is net-rotation reference frames (e.g., Argus and Gordon, 1991; Kreemer the so-called plume push force, which has been proposed to explain the et al., 2003; Argus et al., 2011). More recently, other geological and extreme plate velocity of the Indian plate in the latest Cretaceous–earliest geophysical features have been used to constrain the absolute velocities Cenozoic (Cande and Stegman, 2011; van Hinsbergen et al., 2011). The of the tectonics plates, such as subducted slabs in the upper and lower ridge push force is generally thought to be an order of magnitude smaller mantle (e.g., van der Meer et al., 2010; Schellart, 2011; Butterworth et than the slab pull force (Forsyth and Uyeda, 1975). al., 2014), upper mantle seismic anisotropy (e.g., Long and Silver, 2009; The Australian plate is home to the Australian continent, which is cur- Kreemer, 2009; Conrad and Behn, 2010), and spreading ridges at the rently the fastest moving continent on Earth (Keep and Schellart, 2012) Earth’s surface (e.g., Becker et al., 2015; Wessel and Müller, 2016). A with north- to northeast-directed velocities as high as 6.5 cm/yr (Fig. 1). number of conceptual and geodynamic models have been put forward to These high velocities can be ascribed to the wide subduction zones that explain the variations in plate velocities. Although such models vary as are present along the northern boundary of the Australian plate, namely to what might be the main physical mechanism that explains variation in the north- to northeast-dipping Sunda subduction zone in the northwest absolute plate velocities, such as subducting plate age (e.g., Carlson et al., and the north- to northeast-dipping Melanesian subduction zone in the northeast. Over the past ~75 m.y., the Australian plate has generally under- *[email protected] gone rapid plate velocities, except for 2 periods, at 73–63 Ma and 49–40

976 © 2017 Geological Society of America.www.gsapubs.org For permission | toVolume copy, contact9 | Number [email protected] 6 | LITHOSPHERE Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 Australian plate motion change due to subduction evolution and slab detachment | RESEARCH

o 10 Spakman (2015), a 4000-km-wide (trench-parallel extent) subduction Pacific plate zone was identified in the New Guinea region active ca. 71–50 Ma, and Sunda plate NewGui ne fossil a o the phases of plate acceleration, rapid plate motion, and plate deceleration su - 0 b Po Pacific of the Australian plate were ascribed to the geodynamic evolution of this duc c Ocean klin tion gton zo subduction zone. The subduction zone consumed a small oceanic back- ne o 10 arc basin, the Emo backarc basin, with a maximum north-south extent of Indian Ocean ~1100 km. The phase of plate acceleration was ascribed (in Schellart and Spakman, 2015) to progressive upper mantle slab lengthening during the Australian plate o 5 cm/yr 20 initial transient subduction phase and the phase of plate deceleration was Incipient subduction ascribed to slab detachment during the final stage of subduction. Subduction Tasman I present here a geodynamic numerical model of buoyancy-driven Collision/ Sea obduction o Fossil Basin 30 subduction to test if the rates of plate acceleration and deceleration, and subduction Southern Ocean the high rate of Australian plate motion during the intervening period, can be ascribed to the progressive evolution of the subduction zone in the New o 40 Guinea region, from transient slab lengthening to the final stage of slab Antarctic plate detachment. The geodynamic model provides new insight into the rates ooo o o o o 80 100 120 140 1602180 00 of plate acceleration and deceleration during subduction of the edge of a Figure 1. Plate tectonic map of the central and eastern part of the Aus- plate that contains a large continent. tralian plate, showing the present-day plate boundaries and velocities (in cm/yr), as well as the fossil New Guinea–Pocklington subduction zone. METHODS The velocities are based on the relative plate motion model of DeMets et al. (1994) using the Indo-Atlantic moving hotspot reference frame from O’Neill et al. (2005). Numerical models are presented that have specifically been designed to investigate the subduction evolution of a small oceanic backarc basin (Emo backarc basin), with a particular focus on the latest stage of subduc- Ma, when plate velocities reached minima of 0.8–2.9 cm/yr and 0.3–1.6 tion when all oceanic lithosphere is consumed and continental passive cm/yr, respectively, in the Indo-Atlantic moving hotspot reference frame margin lithosphere approaches the trench. The models use the code Under- (Fig. 2). During the intervening period the velocities were much higher world (Stegman et al., 2006; Moresi et al., 2007, 2014), in which plate with a maximum of 7.3 cm/yr at 59–52 Ma. During the latest Cretaceous motion, subduction and mantle flow are modeled in a Cartesian domain (Maastrichtian) to early Eocene, the Australian plate was bordered to using compositional buoyancy contrasts in an incompressible Boussinesq the east, south, and west by spreading ridges in the Tasman Sea, South- fluid at a very low Reynolds number. Distinct volumes (e.g., continental ern Ocean, and Indian Ocean (e.g., Royer and Sandwell, 1989; Yan and upper crust, lower crust and lithospheric mantle, sublithospheric upper Kroenke, 1993; Gaina et al., 1998; Hall, 2012; Seton et al., 2012); this mantle, lower mantle) are represented by sets of Lagrangian particles cannot explain the period of rapid plate motion and the preceding and that are embedded within a standard Eulerian finite element mesh, which following phases of plate acceleration and deceleration. In Schellart and discretizes the problem to solve the governing equations. For additional

10 ] 10 A B 9 IAHS-2005 9 IAHS-2005 GHS-2012 GHS-2012 8 8 7 7 6 6 5 5 Australian plate [cm/yr ] 4 4 3 3 2 2 1 1

0 Northward velocity component [cm/yr 0 locity of point on 80 70 60 50 40 30 20 10 0 80 70 60 50 40 30 20 10 0 Ve Time [Ma] Time [Ma] Figure 2. Absolute velocity of the Australian plate at 74–0 Ma for a point currently located in the Gulf of Carpentaria (i.e., relatively close to the center of the New Guinea–Pocklington fossil subduction zone, circle with cross in Fig. 1). (A) Total velocity. (B) Northward velocity component (i.e., approximately perpendicular to the strike of the fossil subduction zone). Velocities were calculated in an Indo-Atlantic moving hotspot reference frame (O’Neill et al., 2005; IAHS-2005) and global moving hotspot reference frame (Doubrovine et al., 2012; GHS-2012) using the relative plate motion model of Müller et al. (2008). Note that the Indo-Atlantic hotspot reference frame is the preferred reference frame (see discussion of Plate Velocity Changes in Model and Nature). The velocities in the Indo-Atlantic frame are after Schellart and Spakman (2015).

LITHOSPHERE | Volume 9 | Number 6 | www.gsapubs.org 977 Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 SCHELLART

information on the numerical technique and the nondimensional equa- trailing, positively buoyant 3600-km-long continental lithosphere segment tions, see Moresi et al. (2007, 2014) and Stegman et al. (2006). Velocities representing the Australian continent (which includes a 100 km passive in the models are scaled following the scaling formulations presented in margin). The length of the oceanic lithosphere segment is based on the Schellart and Moresi (2013). reconstruction from Schellart and Spakman (2015), while the length of The modeling domain is 1100 km deep and 6200 km long and has the continental part is based on the north-south extent of the Australian free-slip boundary conditions along the top surface, bottom surface, and continent in the latest Cretaceous–earliest Cenozoic. lateral side walls (Fig. 3). Mesh resolution in the 6200 × 1100 km numeri- The oceanic part of the subducting plate is 80 km thick and con- cal domain is 1024 (length) by 256 (depth) elements. An adaptive spatial sists of 4 layers with their own thickness and rheology following earlier mesh refinement has been implemented to increase resolution in the region modeling approaches (e.g., Schellart et al., 2010; Stegman et al., 2010), of the subduction zone and close to the surface. The smallest cell dimen- except that the models presented here use viscoplastic rheologies with sions in this central region (rectangle with dotted outline in Fig. 3), 220 km a von Mises yield stress throughout the lithosphere to facilitate possible thick and 1860 km wide, are 2.79 km (length) by 2.15 km (depth). Initial detachment. The physical properties of the different layers are listed in particle distribution is 20 particles per cell (total of 5,242,880 particles). Table 1. The 80 km thickness of the oceanic lithosphere is based on the The reference model A involves a layered mantle volume incorporat- age of the Emo backarc basin, which is derived from the age of the pro- ing a free, four-layer subducting plate, a fixed, neutrally buoyant, non- toliths of the Emo metamorphics and Owen Stanly metamorphic rocks stratified overriding plate, a low-viscosity sublithospheric upper mantle, that are thought to originate from this basin. Early research on the meta- and a high-viscosity lower mantle. The top 100 km of the model domain morphics proposed a Mesozoic age (Davies and Warren, 1992) or Late consists of a low-viscosity, low-density air layer to provide a free surface Cretaceous age (Worthing and Crawford, 1996). In more recent work, to the tectonic plates (sticky air layer following Schmeling et al., 2008; the plate tectonic reconstruction from Hall (2012) implies that an ocean Crameri et al., 2012). The subduction zone interface is implemented in basin or backarc basin north of the New Guinea passive margin already the same way as in the numerical models in Schellart and Moresi (2013) existed in the Late Jurassic–Early Cretaceous. In other recent work, the and the laboratory models of Duarte et al. (2013, 2015), in which a weak age of the protolith of the metamorphics has been described as Cretaceous zone develops that consists of the weak materials from the top layer of (Baldwin et al., 2012), while Davies (2012) recorded a mid-Cretaceous the subducting plate, which allows for progressive one-sided subduction. age with radiometric dates of 120–107 Ma. This would imply an oceanic The subducting plate has a negatively buoyant, 1120-km-long oceanic backarc basin with an age range of ~36–70 million years at the time of lithosphere segment representing the Emo oceanic backarc basin and a subduction ca. 71–50 Ma. Considering this age range, an 80-km-thick

Forearc Emo backarc Passive oceanic margin Sticky air lithosphere Australian continental lithosphere Overriding plate Subducting plate

Sub-lithospheric upper mantle 1000 km 6200 km

Lower mantle

Trench Passive margin

Non-dimensional effective viscosity

-3 -2 -1 0 1 2 3 4 10 10 10 10 10 10 10 10 Figure 3. Model set-up of the numerical model of buoyancy-driven progressive subduction in a layered mantle in two-dimen- sional space with free-slip boundaries. In the model, the subducting plate consists of a 1120-km-long oceanic lithosphere segment (which includes the 280-km-long initial slab perturbation) representing the Emo oceanic backarc basin and a 3600-km-long continental lithosphere segment representing the Australian continental lithosphere. The oceanic and conti- nental lithosphere of the subducting plate have the same four-layer rheology, but a different density in which the oceanic lithosphere is negatively buoyant, while the continental lithosphere has a continental crust (top two layers) that is posi- tively buoyant. See Table 1 for details of the physical properties. The initial slab perturbation curves into the subduction zone with a maximum dip angle at the slab tip of 27.3°. Note that the top panel shows the different domains in the model, while the bottom panel shows the nondimensional effective viscosity. The rectangle with the dotted outline shows the region with maximum horizontal and vertical resolution. The lateral extent of the passive margin is 100 km. Note that the continental crustal thickness across the passive margin changes linearly (top panel), while the rheology across the passive margin is constant (bottom panel).

978 www.gsapubs.org | Volume 9 | Number 6 | LITHOSPHERE Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 Australian plate motion change due to subduction evolution and slab detachment | RESEARCH

TABLE 1. NUMERICAL MODEL PROPERTIES FOR REFERENCE MODEL A Domain Density RheologyNondimensional Yield stress Thickness (kg/m3) viscosity (MPa) (km) Subducting plate (oceanic part)

layer 1 (top) 3280 viscoplastic (von Mises) 100 ηUM-Max 5.88 10

layer 2 3280 viscoplastic (von Mises) 100 ηUM-Max 29.4 20

layer 3 3280 viscoplastic (von Mises) 1000 ηUM-Max 352.820

layer 4 (bottom) 3280 viscoplastic (von Mises) 50 ηUM-Max 29.4 30 Subducting plate (continental part)

top crust 2800 viscoplastic (von Mises) 100 ηUM-Max 5.88 10 bottom crust 2800 viscoplastic (von Mises) 100 ηUM-Max 29.4 20 top LithM 3200 viscoplastic (von Mises) 1000 ηUM-Max 352.8 20 bottom LithM 3200 viscoplastic (von Mises) 50 ηUM-Max 29.4 30

Overriding plate forearc 3200 linear viscous 1000 ηUM-Max 60

Overriding plate backarc 3200 linear viscous 50 ηUM-Max 60

Sublithospheric upper mantle 3200 nonlinear viscous 0.1–1 ηUM-Max 580–660

Lower mantle 3200 linear viscous 100 ηUM-Max 340

Air0linear viscous 0.1 ηUM-Max 100

Note: ηUM-Max is the maximum viscosity of the sublithospheric upper mantle, which has a nonlinear stress-dependent viscosity 20 with a stress exponent n = 3.5 (following Mackwell et al., 1990). The scaled viscosity ηUM-Max = 5 × 10 Pa∙s. The continental part of the subducting plate has two crustal layers and two lithospheric mantle (LithM) layers. Model B is the same as reference model A except that the yield stress of layer 3 and top LithM in the subducting plate is 382.2 MPa.

oceanic lithosphere (implying an average age of ~50 million years at the detachment (e.g., van Hunen and Allen, 2011; Capitanio and Replumaz, time of subduction) is justified. An average density contrast of 80 kg/m3 2013; Chertova et al., 2014; Capitanio et al., 2015), because such models between the oceanic slab and ambient sublithospheric mantle is adopted can take into account three-dimensional (3D) spatial effects, including in the models, as it is assumed that there is pervasive eclogitization of the lateral migration of slab detachment, tearing, and trench-parallel flow of basaltic crust into eclogite facies rocks (Cloos, 1993). mantle material. However, this was found to be unfeasible at this time, The continental part of the subducting plate, which includes the considering the very large size of the subduction zone (~4000 km), the 100-km-long passive margin, has the same rheological layering with the required resolution, and the required number of time steps (several thou- same thickness as the oceanic part (Fig. 3, bottom panel). The only differ- sand). These earlier works generally used lower spatial resolution and ence between the oceanic and continental domains is the density structure, modeled smaller plates and subduction zones. Furthermore, considering with a lighter, positively buoyant, continental lithosphere due to a 30 that the natural prototype is a wide subduction zone, its geodynamic km continental crust, a denser, negatively buoyant, oceanic lithosphere behavior, in particular that of its central portion, can be approximated (Table 1), and a passive margin across which the density changes linearly with a model using a 2D spatial set-up (Schellart and Moresi, 2013). due to a linear change in continental crustal thickness from 30 km at the continental side to 10 km at the oceanic side. This set-up was specifically MODEL RESULTS chosen to investigate where strain localization and slab detachment would occur in a scenario without any lateral rheological contrasts, but with only Reference model A shows a general style of subduction with progres- lateral density contrasts in the subducting plate. sive slab lengthening and an increase of the slab dip angle (averaged along The model overriding plate is neutrally buoyant, implying that it is the entire slab length) during the free sinking phase, from 21° at the start oceanic in nature with a density that is comparable to that of the sublitho- (Fig. 3) to 40° at an intermediate stage (Fig. 4A), to 46° at an advanced spheric mantle. This is most likely if one considers the tectonic setting stage (Fig. 4B). As the slab approaches the 660 discontinuity its tip is to the north of Australia in the latest Cretaceous–early Eocene, where slightly deflected to a lower slab dip angle. During the free sinking phase

reconstructions indicate that there was an ocean to the north (Hall, 2012). the trench-normal subducting plate velocity vSP⊥ increases rapidly to a max- Another reason for implementing a neutral buoyancy was to minimize imum of ~8 cm/yr, when the slab tip is located at ~578 km depth (82 km the effect of the overriding plate on the subduction process, such that the above the 660 km discontinuity) (Fig. 5A). The trench-normal trench

change from oceanic to continental subduction and the process of slab velocity (vT⊥) is significantly less than vSP⊥ and increases to a maximum of

detachment could be studied without other complicating factors. For the ~2.5 cm/yr, which is reached at a time before the maximum of vSP⊥. Dur-

rheology the simplest set-up was chosen (following Duarte et al., 2013; ing the slab tip–660 km discontinuity interaction phase (~8–12 m.y.), vSP⊥

Schellart and Moresi, 2013) that allows for continuous and progressive decreases from ~8 to 5–6 cm/yr, while vT⊥ remains relatively unchanged. subduction, because little is known about the properties and rheology of From ~12.2 to 15.3 m.y. continental subduction takes place (Figs. 6A,

the overriding plate. The model overriding plate has a forearc region with 6B), during which vSP⊥ decreases from ~5 to 3 cm/yr. This is followed by a higher viscosity than the backarc region (Table 1) (following Schellart a short phase of slab detachment from 15.3 to 16.1 m.y. (Figs. 6C–6E),

and Moresi, 2013), as it is generally thought that the forearc is relatively during which vSP⊥ decreases from ~3 to –1 cm/yr and then increases again cool and strong with depressed isotherms due to the cold subducting plate to ~0 cm/yr (Fig. 5A). During continental subduction and slab detachment,

diving below it, and therefore has a higher effective viscosity than the arc vT⊥ decreases from ~2 cm/yr to 0 cm/yr. After slab detachment, subduction

and backarc regions, where isotherms are elevated. stops, vSP⊥ and vT⊥ are close to zero, while the detached slab continues to Apart from reference model A, one other model is presented, model B, sink into the mantle (Figs. 4H, 4I, and 6F). with a higher yield stress compared to the reference model (Table 1). A During the phase of continental subduction, the continental lithosphere three-dimensional geometrical set-up would be ideal for the geodynamic is pulled into the mantle by the oceanic slab segment, and the maximum models, as presented in other modeling work of subduction and slab subduction depth of the continental crust (its base, originally at 30 km

LITHOSPHERE | Volume 9 | Number 6 | www.gsapubs.org 979 Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 SCHELLART

0 km cm/yr), continental subduction stops and the only remaining activity is A the very slow progressive steepening of the slab (Figs. 5B and 7). The 500 maximum subduction depth of the continental crust (its base, originally 4.2 m.y. 1000 at 30 km depth) at the end of the model run is 144 km. 0 km B DISCUSSION 500 7.4 m.y. 1000 Plate Velocity Changes in Model and Nature 0 km C 500 The subduction models A and B show a general style of subduction behavior during the oceanic subduction period that is comparable to previ- 12.2 m.y. 1000 ous buoyancy-driven subduction modeling studies, with progressive slab 0 km D steepening (e.g., Jacoby, 1973; Funiciello et al., 2004; Schellart, 2004a; 500 Guillaume et al., 2009) and an increasing trenchward subducting plate 13.9 m.y. velocity during the free sinking phase to a maximum velocity when the 1000 slab tip is just above (<~100 km from) the 660 km viscosity discontinu- 0 km E ity (e.g., Schellart, 2008; Capitanio et al., 2010; Schellart et al., 2011; 500 Meyer and Schellart, 2013; Chen et al., 2015, 2016) (Fig. 5). In the cur- 15.3 m.y. rent models it takes only ~5 m.y. to increase from v = ~2 cm/yr to ~8 1000 SP⊥ 0 km cm/yr. Such a velocity increase over such a time period is comparable F to that implied by the velocity curve for the Australian plate in the Indo- 500 Atlantic moving hotspot reference frame (Fig. 2), showing an increase 16.0 m.y. from ~1 to ~7.4 cm/yr between ca. 64 and ca. 59 Ma and an increase in 1000 0 km the northward velocity component, which is approximately perpendicular G to the strike of the fossil subduction zone, from ~0.2 to ~5.0 cm/yr. The 500 increase in subducting plate velocity in the models and for the Austra- 16.1 m.y. 1000 lian plate can be ascribed to the progressive lengthening of the slab and 0 km deepening of the slab tip (Figs. 4A, 4B, and 5), and thereby the increase H in slab negative buoyancy and net slab pull (Schellart, 2004b). These high 500 subducting plate velocities can be reached, despite the fact that the plate 16.4 m.y. 1000 is relatively long in the latest Cretaceous–early Cenozoic (~4000–5000 0 km km long Australian plate in the north-south direction). I The decrease in subducting plate velocity during the slab tip–660 500 km discontinuity interaction phase is generally also observed in earlier 17.0 m.y. 1000 works (e.g., Schellart, 2008; Capitanio et al., 2010; van Hunen and Allen, 0 km 1000 2000 3000 4000 5000 6000 2011; Schellart et al., 2011; Meyer and Schellart, 2013; Chen et al., 2015, Non-dimensional effective viscosity 2016). Reference model A further shows a period of decrease in vSP⊥ -3 -2 -1 0 1 2 3 4 from ~4.5 cm/yr to ~3 cm/yr during ~3.1 m.y. of continental subduction, 10 10 10 10 10 10 10 10 followed by an ~0.8 m.y. phase of slab detachment, during which vSP⊥ Figure 4. Model results showing the evolution of the reference numerical decreases from ~3 cm/yr to –1 cm/yr (Fig. 5A). The slowdown during subduction model (reference model A) at nine different stages. All images continental subduction can be explained by the decrease in buoyancy show the nondimensional effective viscosity field. (A) Early stage of the initial transient subduction phase (4.2 m.y.). (B) Late stage of the initial tran- force of the slab (i.e., becoming less negative) due to addition of positively sient subduction phase with maximum subducting plate velocity (7.4 m.y.). buoyant continental slab material (Figs. 6A–6C, images on left). The rapid (C) Start of continental subduction (12.2 m.y.). (D) Continental subduction slowdown during slab detachment can be explained by the rapid decrease phase (13.9 m.y.). (E) Start of slab detachment with necking of lithosphere in coupling between the detaching slab segment and the trailing slab seg- and formation of conjugate shear zones (15.3 m.y.). (F) Late stage of slab ment due to the rapid decrease in effective viscosity in the detachment detachment with necking almost complete (16.0 m.y.). (G) Completion of zone (Figs. 6B–6E, images on right). During the ~3.9 m.y. period of con- slab detachment (16.1 m.y.). (H) Early stage after slab detachment with sinking of detached slab (16.4 m.y.). (I) Late stage after slab detachment tinental subduction and slab detachment the velocity decreases ~5.5 cm/yr, with sinking of detached slab (17.0 m.y.). which is comparable to the ~3 m.y. period from ca. 52 Ma to ca. 49 Ma during which the Australian plate velocity decreased ~7 cm/yr (Fig. 2A) and the northward velocity component of the Australian plate decreased depth), just before slab detachment is complete, is 96 km. This indicates ~5 cm/yr (Fig. 2B) in the Indo-Atlantic moving hotspot reference frame. a depth increase of 66 km. The slab detachment process starts with the Subduction model A thereby demonstrates the physical viability of the formation of two conjugate shear zones (Fig. 6C) and occurs mostly by conceptual model in which slowdown of the Australian plate in the latter symmetrical necking (Figs. 6C–6E). Once slab detachment is complete, part of the early Eocene is ascribed to continental subduction and subse- the trailing slab segment slightly rebounds. quent slab detachment, as illustrated in Figures 5A and 6. The other model (model B with a stronger subducting plate than model In the preceding two paragraphs the subducting plate velocities from A) generally shows the same evolution as model A until ~12 m.y.; model the numerical models have been compared to the subducting Australian B starts continental subduction just after ~12 m.y., and this continues until plate velocities in the Indo-Atlantic moving hotspot reference frame from

~20 m.y. when vSP⊥ has decreased to zero, vT⊥ is slightly negative (~–0.5 O’Neill et al. (2005). As explained in detail elsewhere (e.g., Schellart

980 www.gsapubs.org | Volume 9 | Number 6 | LITHOSPHERE Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018

Australian plate motion change due to subduction evolution and slab detachment | RESEARCH

T 10 10 T

vSP model B

A T 610 vSP Reference model A B T 9 vT 9 v model B 578 T Start 8 continental 8 Start subduction continental 7 7 subduction Start slab 449 6 441 detachment 6 (conjugate 5 shears form) 5 Slab tip Slab tip 308 4 337 touches 660 km 4 touches discontinuity 660 km 218 Slab 224 discontinuity

locity [cm/yr] 3 detachment locity [cm/yr] 3 complete Ve 2 Ve 2 1 1 0 0 -1 -1 -2 -2 0 246810 12 14 16 18 20 22 0246810121416182022 Time [m.y.] Time [m.y.]

Figure 5. Diagrams showing the evolution of the trench-normal subducting plate velocity (vSP⊥, trenchward is positive) and trench-

normal trench velocity (vT⊥, retreat is positive). (A) Subducting plate velocity (black line) and trench velocity (gray line) for reference model A with a core layer with yield stress C = 352.8 MPa. (B) Subducting plate velocity (black line) and the trench velocity (gray line) for model B with a stronger core layer (C = 382.2 MPa). Note that only the reference model shows slab detachment (see Figures 4 and 6), while model B does not (Fig. 7). The gray arrows with the gray numbers indicate the depth of subduction (in km) of the slab

tip. Note that vSP⊥ is measured at the trench; vSP⊥ at the trailing edge of the subducting plate is less due to (unwanted) stretching of the subducting plate, which is a consequence of the great initial length (4720 km) and finite strength of the subducting plate.

0 km A 100 Start 200 continental 300 12.2 m.y. subduction

0 km B 100 200 300 13.9 m.y.

0 km C 100 Start slab detachment 200 (conjugate 300 15.3 m.y. shears form)

0 km D 100 200 300 16.0 m.y.

0 km E 100 Slab 200 detachment 300 16.1 m.y. complete

0 km F 100 200 300 16.4 m.y. 1000 1200 1400 1600 1800 2000 2200 2400 km 1000 1200 1400 1600 1800 2000 2200 2400 km 1000 1200 1400 1600 1800 2000 2200 2400 km Non-dimensional second invariant of the strain rate Non-dimensional effective viscosity

-7 -6 -5 -4 -3 -2 -1 -3 -2 -1 0 1 2 3 4 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 Figure 6. Zoom in of model results (reference model A) in the region of the subduction zone showing the evolution of the numerical subduction model at different stages during continental subduction, slab detachment, and after slab detachment. (A) Time at the start of continental subduction (12.2 m.y.). (B) Continental subduction phase (13.9 m.y.). (C) Start of slab detachment with necking of litho- sphere and formation of conjugate shear zones (15.3 m.y.). (D) Late stage of slab detachment with necking almost complete (16.0 m.y.). (E) Completion of slab detachment (16.1 m.y.). (F) Early stage after slab detachment with sinking of detached slab (16.4 m.y.). Images on the left show the different domains in the numerical model, images in the middle show the second invariant of the nondimensional strain rate, and images on the right show the nondimensional effective viscosity field.

LITHOSPHERE | Volume 9 | Number 6 | www.gsapubs.org 981 Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 SCHELLART

0 km A 500 4.3 m.y. 1000 0 km B 500 14.3 m.y. 1000 0 km C 500 20.3 m.y. 1000 0 km 1000 2000 3000 4000 5000 6000 0 km 1000 2000 3000 4000 5000 6000

Non-dimensional effective viscosity

-3 -2 -1 0 1 2 3 4 10 10 10 10 10 10 10 10 Figure 7. Model results showing three evolutionary stages of numerical subduction model B with a relatively strong core layer (C = 382.2 MPa), which does not develop slab detachment. (A) Early stage of the initial transient subduction phase (4.3 m.y.). (B) Early phase of continental subduction (14.3 m.y.). (C) Very late stage of continental subduction phase (20.3 m.y.), when subduct- ing plate velocity is close to zero. Images on the left show the different domains in the model, while images on the right show the nondimensional effective viscosity field.

et al., 2008; Schellart, 2011; Schellart and Spakman, 2012, 2015), this pull forces, much like the free slab sinking phase and subducting plate Indo-Atlantic hotspot reference frame is preferred over other reference acceleration shown in Figure 5, causing fast plate acceleration at 45–40 frames for calculating plate velocities and plate boundary velocities, which Ma and fast north- to northeast-directed plate velocities since ca. 40 Ma. is particularly evident for the western Pacific domain. The reasons are many, including minimization of global trench migration velocities and Continental Subduction and Slab Detachment rollback-induced toroidal volume fluxes in this reference frame (Schellart et al., 2008), optimal agreement between predicted and observed fossil Velocities and Duration of Continental Subduction and Slab slab locations in the southwest Pacific (Schellart and Spakman, 2012) Detachment and slab structures in the western Pacific (Schellart, 2011), and highest Reference model A with slab detachment shows a phase of continental coincidence with fossil slab sinking-induced dynamic topography in the subduction and slab detachment lasting only ~3.9 m.y., with a decrease

southwest Pacific (Schellart and Spakman, 2015). For comparison, the in vSP⊥ of ~5.5 cm/yr. Earlier works on slab detachment generally show velocities have also been plotted in the global moving hotspot reference comparable reductions in subducting plate velocity (e.g., Burkett and Bil- frame from Doubrovine et al. (2012). The velocity profile for the north- len, 2010; van Hunen and Allen, 2011; Capitanio and Replumaz, 2013; ward velocity component is roughly comparable in shape to the Indo- Capitanio, 2014). Normal (oceanic) subduction models from Burkett and

Atlantic velocity profile in the period 65–49 Ma (Fig. 2B), but the total Billen (2010) show comparable reductions in vSP⊥ of ~5–9 cm/yr during acceleration, peak velocity, and total deceleration are reduced. slab detachment, as do the continental subduction break-off models of In Schellart and Spakman (2015) geological and geophysical evi- Capitanio (2014) and Capitanio and Replumaz (2013) with a reduction

dence was presented to propose the existence and the timing of activity in vSP⊥ of ~5 cm/yr and a reduction in convergence velocity of 2–4 cm/yr, (ca. 71–50 Ma) of the fossil New Guinea–Pocklington subduction zone; it respectively. The continental subduction break-off model of van Hunen was argued that this subduction zone was largely responsible for velocity and Allen (2011) shows a higher reduction of ~14 cm/yr, but their model changes of the Australian plate, with an increase in speed from ~64 Ma to generally shows much faster subducting plate velocities, to ~23 cm/yr. ~59 Ma due to slab lengthening, rapid plate motion ca. 59–52 Ma during The duration of slab detachment for model A is relatively short (~0.8– mature subduction, and slowing ca. 52–49 Ma due to slab detachment. 1.0 m.y.), but similar short durations (~0.3–3.0 m.y.) have been reported The reference numerical model A is largely consistent with this conceptual in earlier works on 2D numerical modeling of subduction and slab detach- model, but also shows that part of the slowing (~27%) can be ascribed to ment (Duretz et al., 2012, 2014), as well as 2D and 3D numerical modeling the phase of continental subduction prior to slab detachment. of normal (oceanic) subduction (~1–3 m.y.) by Burkett and Billen (2010). The numerical geodynamic model can explain the subducting plate The 2D numerical models of subduction followed by continental subduc- velocity changes and velocity magnitudes in the period 65–45 Ma, but does tion of van Hunen and Allen (2011) report a long delay time between the not explain the significant north-directed plate acceleration that started start of continental subduction and slab detachment (in the range ~9–23 ca. 44 Ma. This new phase of plate acceleration has been ascribed to the m.y.), much longer than reported here (~3.9 m.y.), which is possibly due formation of new subduction zones along the northwestern, northern, and to the higher yield stress in their models (400 MPa) compared to the one northeastern boundaries of the Australian plate between ca. 50 and ca. 40 used here for reference model A (352.8 MPa). Model B, with a yield Ma (Schellart and Spakman, 2015), and this phase of renewed subduction stress of 382.2 MPa, did not show any yielding for the duration of the has not been included in the numerical model. Formation of the north- to simulation, which was allowed to continue for ~9.5 m.y. since the start northeast-dipping Sunda subduction zone ca. 45 Ma (Hall, 2012), North of continental subduction. Sulawesi–Halmahera subduction zone ca. 45 Ma (Hall, 2012), and north- The models presented here have a 2D spatial set-up, and thus represent east- to east-dipping New Caledonia–Northland subduction zone at 50–40 a subduction scenario in which slab detachment occurs contemporane- Ma (Schellart and Spakman, 2012) would have resulted in increasing slab ously along the entire lateral extent of the subduction zone. Earlier work,

982 www.gsapubs.org | Volume 9 | Number 6 | LITHOSPHERE Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 Australian plate motion change due to subduction evolution and slab detachment | RESEARCH

however, proposed that slab detachment and slab tears migrate subhori- margin size as here (100 km) has a very comparable continental subduc- zontally (e.g., Wortel and Spakman, 2000). Numerical subduction models tion period of ~10 m.y. with a 3D spatial set-up have reproduced this process, demonstrating that lateral tear migration can be very rapid, as much as ~80 cm/yr (van Hunen Forces During Continental Subduction and Slab Detachment and Allen, 2011). For a 4000-km-wide subduction zone this would imply In reference model A, the slab detaches in a late stage of continen- that slab detachment can take place in 2.5 m.y. if it initiated at one loca- tal subduction, while there is no detachment in model B. This can be tion in the center and migrated outward. It is plausible that along such a explained by the buoyancy forces that operate during continental subduc- wide subduction zone, slab detachment starts at multiple locations, and tion and the different yield stress in models A and B. Figure 8A shows

migrates laterally from these places, thereby reducing the time of detach- the evolution of the buoyancy force of the oceanic slab segment FBu-Oc ment along the entire extent of the subduction zone. (negative) (per meter length of the trench) and the buoyancy force of the

Earlier work has found that temperature effects on slab detachment continental slab segment FBu-Co (positive) (per meter length of the trench). duration are relatively small (decrease of ~8%–10%) (Gerya et al., 2004; These forces have been calculated with the following equations: Duretz et al., 2012), indicating that the duration of slab detachment for

the isothermal model A presented here is of the right order of magnitude FLBu-OcO=ρ∆ cOcOTgc , (1) and would only be marginally affected in case thermal gradients were and

incorporated. Instead, the variability in duration of slab detachment has FLBu-CoC=ρ∆ oCoCTgo , (2) been ascribed mainly to differences in rheology and yield stress of the

slab material (e.g., Andrews and Billen, 2009; Duretz et al., 2012). The where ΔρOc is the density contrast between ambient mantle and oceanic 3 yield stress of 352.8 MPa used here for model A with slab detachment is slab (ΔρOc = –80 kg/m ), LOc is the length of the oceanic slab segment that comparable to that reported by Andrews and Billen (2009; 300 MPa) to contributes to the buoyancy force (so excluding the segment subhorizon- allow for slab detachment. tally overlying the 660 km discontinuity; this parameter changes with

Model B without slab detachment shows a decrease in vSP⊥ from ~4.5 time), TOc is the thickness of the oceanic slab segment (changes with time 2 cm/yr to ~0 cm/yr during continental subduction over an ~8 m.y. period. and space), g is gravitational acceleration (9.8 m/s ), ΔρCo is the density 3 Laboratory models of subduction and late-stage continental subduction contrast between ambient mantle and continental crust (ΔρCo = 400 kg/m ),

from Edwards et al. (2015) also lack slab detachment, and show periods LCo is the length of the continental slab segment (changes with time), and

of continental subduction that generally last longer, ~10–30 m.y. This can TCo is the thickness of the continental crustal slab segment (changes with be partly ascribed to the neutral buoyancy of the continental lithospheric time and space; note that the continental lithospheric mantle is neutrally mantle in the current models compared to the negative buoyancy for the buoyant). As can be observed, subduction of oceanic lithosphere creates a models from Edwards et al. (2015), giving the continental lithosphere in negative (downward) buoyancy force with a minimum of ~–5.2 × 1013 N/m the current work a larger net positive buoyancy force, thereby reducing that is reached between the start of continental subduction and the start continental subduction. Most of the discrepancy, however, can be ascribed of slab detachment. Subduction of continental lithosphere creates a posi- to the fact that 14 out of 15 models in Edwards et al. (2015) have a larger tive (upward) buoyancy force with a maximum of ~6.4 × 1012 N/m that is passive margin (250–500 km). The one model with the same passive reached during slab detachment. The differential tensional buoyancy force

7 8 A FBu-Oc B FS-Max 6 FBu-Co FSP-Eff 7 5 FBu-Diff

13 4 12 6 3 Start slab detachment Bu-Diff 2 (conjugate 5 F Start slab shears form) 12% F Bu-Diff detachment 1 10% (conjugate 4 shears form) 0 -1 Slab 3 Start detachment Start Slab -2 continental complete continental detachment subduction subduction complete -3 2 -4 1 Force per meter trench [x 10 N/m] -5 Force per meter trench [x 10 N/m] -6 0 0 246810121416182022 0246810121416182022 Time [m.y.] Time [m.y.] Figure 8. Model results showing the evolution of the forces in reference model A. (A) Development of the buoyancy force of the oceanic

slab segment (FBu-Oc), the buoyancy force of the continental slab segment (FBu-Co), and the differential buoyancy force (FBu-Diff = FBu-Co – FBu-Oc)

with progressive time. (B) Development of the effective slab pull force (FSP-Eff) and the yield force in the subducted slab (FS-Max) with pro-

gressive time. Note that all the forces are per meter length of the subduction zone trench. Error bars are indicated for data points of FBu-Oc,

FBu-Co, FBu-Diff, and FS-Max (based on maximum errors in measuring TOc, LOc, TCo and LCo; see text for definitions). (For discussion seeForces During Continental Subduction and Slab Detachment.)

LITHOSPHERE | Volume 9 | Number 6 | www.gsapubs.org 983 Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 SCHELLART

(per meter trench length) between the continental slab segment and oceanic CONCLUSIONS

slab segment, with FBu-Diff = FBu-Co – FBu-Oc, is focused in the boundary region between the oceanic and continental slab segments and reaches a peak In this work a numerical geodynamic model of subduction, continental 13 at the onset of slab detachment with FBu-Diff = ~5.6 × 10 N/m (Fig. 8A). subduction, and slab detachment has been presented to simulate the sub- The strength of the slab is effectively determined by the yield stress duction evolution at the northern margin of the Australian plate during the of the strong core layer of the subducting plate (layer 3 and Top LithM in latest Cretaceous and early Cenozoic, and to quantify the velocity of the Table 1), which is 352.8 MPa. With such a yield stress and a 15–20-km- Australian plate. The numerical model produces a comparable subducting thick core layer in the vicinity of the ocean-continent transition during plate velocity evolution as documented for the Australian plate between the continental subduction phase, this gives a yield force, the maximum ca. 66 and 49 Ma, with a first phase of plate acceleration due to progres-

in-plane slab pull force (per meter trench length) in the slab, FS-Max, of sive upper mantle slab lengthening, which increases the slab pull force, 5.3–7.1 × 1012 N/m. The evolution of this yield force for model A is plot- a period of relatively rapid plate motion during mature subduction, and a

ted in Figure 8B. We can see that FS-Max is about an order of magnitude phase of plate deceleration due to continental subduction and subsequent

smaller than FBu-Diff, which might suggest at first that the slab is much slab detachment. Most (~73%) of the plate deceleration is caused during too weak to sustain such a large buoyancy force and that slab detachment the short (~0.8 m.y.) phase of slab detachment, while a smaller compo- should have occurred in a much earlier phase of subduction. A significant nent of plate deceleration (~27%) is caused by the preceding phase of part of the buoyancy force, however, is sustained elsewhere in the sys- continental subduction, which lasted ~3.1 m.y. A geodynamic subduction tem, in particular by viscous forces in the ambient mantle (Conrad and model that shows continental subduction but lacks a final phase of slab Lithgow-Bertelloni, 2002; Schellart, 2004b; Krien and Fleitout, 2008; detachment is characterized by a slower and more gradual plate decelera- Leng and Zhong, 2010). Earlier work showed that only ~10%–12% of the tion that lasts ~8 m.y., which is in conflict with observations of fast plate buoyancy forces is transmitted in the plane of the subducted slab toward deceleration for the Australian plate between ca. 52 and 49 Ma. It is thus the trailing plate (Schellart, 2004b; Sandiford et al., 2005). This would concluded that slab detachment is an essential element to explain the rapid

imply that the effective slab pull force (per meter trench length) FSP-Eff = plate deceleration of the Australian plate. This conclusion is consistent

0.10–0.12 × FBu-Diff. Figure 8B shows the evolution of FSP-Eff, demonstrat- with the evidence presented in Schellart and Spakman (2015) that the ing that a maximum effective slab pull force is reached at the start of slab fossil New Guinea–Pocklington slab detached in the early Eocene and 12 detachment, with FSP-Eff = 5.6–6.7 × 10 N/m. Figures 6C and 8 further is currently located in the upper part of the lower mantle below central show that slab detachment starts at a time when a significant section of and southeastern Australia.

continental lithosphere has been subducted (LCo ≈128 km) and when FS-Max More generally, I propose here that rapid plate deceleration (~3–5

≈0.10FBu-Diff, indicating that at this time the effective slab pull force is large cm/yr) over periods lasting not more than ~1 m.y. can be ascribed to the enough to start rapid shearing, necking, and yielding in the slab. Before process of slab detachment following continental subduction. It is thus slab detachment, the amount of subducted continental lithosphere was conceivable that rapid decelerations that have occurred in the geological

insufficient to create a FBu-Diff of sufficient magnitude, and the slab was past for other tectonic plates have a comparable underlying cause, namely

not sufficiently stretched yet to decrease FS-Max, giving FSP-Eff < FS-Max, and slab detachment. so slab detachment was not possible yet. Applying calculations as above to model B, for a late stage as depicted ACKNOWLEDGMENTS in Figure 7C, F = –3.40 × 1013 N/m (±9.5%), F = 1.07 × 1013 N/m I thank Anne Replumaz and two anonymous reviewers for their constructive and helpful com- Bu-Oc Bu-Co ments. This research was supported through a Vici Fellowship from the Dutch Science Founda- 13 (±17.2%), and FBu-Diff = 4.47 × 10 N/m (±26.6%). This would imply an tion (NWO) and has been supported by computational resources from the NCI National Facility in Australia through the National Computational Merit Allocation Scheme (projects ei8 and qk0). effective slab pull force (per meter trench length) FSP-Eff = 4.47–5.37 × 1012 N/m. This is less than the estimated yield force (per meter trench length) in the slab of F = 7.64 × 1012 N/m (±5%), thereby explaining REFERENCES CITED S-Max Andrews, E.R., and Billen, M.I., 2009, Rheologic controls on the dynamics of slab detachment: why slab detachment did not occur in model B. Tectonophysics, v. 464, p. 60–69, doi:​10​.1016​/j​.tecto​.2007​.09​.004. Argus, D.F., and Gordon, R.G., 1991, No-net-rotation model of current plate velocities incor- porating plate motion model Nuvel-1: Geophysical Research Letters, v. 18, p. 2039–2042, Continental Subduction Depth doi:​10​.1029​/91GL01532. Models A and B show different maximum depths of continental sub- Argus, D.F., Gordon, R.G., and Demets, C., 2011, Geologically current motion of 56 plates duction, 96 km and 144 km, respectively, which can simply be explained relative to the no-net-rotation reference frame: Geochemistry Geophysics Geosystems, v. 12, Q11001, doi:​10​.1029​/2011GC003751. by the weaker subducting plate in A compared to B. The weaker plate in Baldwin, S.L., Fitzgerald, P.G., and Webb, L.E., 2012, Tectonics of the New Guinea region: A resulted in slab detachment prior to the continental slab segment reach- Annual Review of Earth and Planetary Sciences, v. 40, p. 495–520, doi:10​ ​.1146​/annurev​ -earth​-040809​-152540. ing its maximum depth that could have been reached with the available Becker, T.W., Schaeffer, A.J., Lebedev, S., and Conrad, C.P., 2015, Toward a generalized plate negative buoyancy force of the oceanic slab segment. This resulted in a motion reference frame: Geophysical Research Letters, v. 42, p. 3188–3196, doi:​10​.1002​ shorter duration of slab pull during continental subduction and thus a /2015GL063695. Bird, P., 2003, An updated digital model of plate boundaries: Geochemistry, Geophysics, Geo- shallower depth of continental subduction. systems, v. 4, 1027, doi:​1010​.1029​/2001GC000252. The observed continental subduction depths reported here are compa- Burkett, E.R., and Billen, M.I., 2010, Three‐dimensionality of slab detachment due to ridge‐ trench collision: Laterally simultaneous boudinage versus tear propagation: Geochemistry rable to, although in the lower range of, depths reported in previous mod- Geophysics Geosystems, v. 11, Q11012, doi:10​ ​.1029​/2010GC003286. eling studies of buoyancy-driven subduction (e.g., van Hunen and Allen, Butterworth, N.P., Talsma, A.S., Müller, R.D., Seton, M., Bunge, H.-P., Schuberth, B.S.A., Sheph- ard, G.E., and Heine, C., 2014, Geological, tomographic, kinematic and geodynamic con- 2011; Sizova et al., 2012; Edwards et al., 2015). For example, Edwards straints on the dynamics of sinking slabs: Journal of Geodynamics, v. 73, p. 1–13, doi:​ et al. (2015) reported depths in the range 70–560 km for experiments with 10​.1016​/j​.jog​.2013​.10​.006. different passive margin width, continental crustal thickness, and oceanic Cande, S.C., and Stegman, D.R., 2011, Indian and African plate motions driven by the push force of the Reunion plume head: Nature, v. 475, p. 47–52, doi:​10​.1038​/nature10174. lithospheric thickness. Their experiment with a comparable passive margin Capitanio, F.A., 2014, The dynamics of extrusion tectonics: Insights from numerical modeling: width (100 km), continental crustal thickness (30 km), density contrast (366 Tectonics, v. 33, p. 2361–2381, doi:​10​.1002​/2014TC003688. 3 Capitanio, F.A., and Replumaz, A., 2013, Subduction and slab breakoff controls on Asian in- kg/m ), and oceanic lithospheric thickness (100 km) showed a low conti- dentation tectonics and Himalayan western syntaxis formation: Geochemistry Geophys- nental subduction depth (70 km), comparable to the work reported here. ics Geosystems, v. 14, p. 3515–3531, doi:​10​.1002​/ggge​.20171.

984 www.gsapubs.org | Volume 9 | Number 6 | LITHOSPHERE Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 Australian plate motion change due to subduction evolution and slab detachment | RESEARCH

Capitanio, F.A., Stegman, D.R., Moresi, L.N., and Sharples, W., 2010, Upper plate controls on Keep, M., and Schellart, W.P., 2012, Introduction to the thematic issue on the evolution and deep subduction, trench migrations and deformations at convergent margins: Tectono- dynamics of the Indo-Australian plate: Australian Journal of Earth Sciences, v. 59, p. 807– physics, v. 483, p. 80–92, doi:​10​.1016​/j​.tecto​.2009​.08​.020. 808, doi:​10​.1080​/08120099​.2012​.708360. Capitanio, F.A., Replumaz, A., and Riel, N., 2015, Reconciling subduction dynamics during Klootwijk, C.T., Gee, J.S., Peirce, J.W., Smith, G.M., and McFadden, P.L., 1992, An early India- Tethys closure with large-scale Asian tectonics: Insights from numerical modeling: Geo- Asia contact: Paleomagnetic constraints from Ninety east Ridge, ODP Leg 121: Geology, chemistry Geophysics Geosystems, v. 16, p. 962–982, doi:​10​.1002​/2014GC005660. v. 20, p. 395–398, doi:​10​.1130​/0091​-7613​(1992)020​<0395:​AEIACP>2​.3​.CO;2. Carlson, R.L., Hilde, T.W.C., and Uyeda, S., 1983, The driving mechanism of plate tecton- Kreemer, C., 2009, Absolute plate motions constrained by shear wave splitting orientations ics: Relation to age of the lithosphere at trenches: Geophysical Research Letters, v. 10, with implications for hot spot motions and mantle flow: Journal of Geophysical Research, p. 297–300, doi:​10​.1029​/GL010i004p00297. v. 114, B10405, doi:​10​.1029​/2009JB006416. Chen, Z., Schellart, W.P., and Duarte, J.C., 2015, Overriding plate deformation and variability Kreemer, C., Holt, W.E., and Haines, A.J., 2003, An integrated global model of present-day of fore-arc deformation during subduction: Insight from geodynamic models and ap- plate motions and plate boundary deformation: Geophysical Journal International, v. 154, plication to the Calabria subduction zone: Geochemistry Geophysics Geosystems, v. 16, p. 8–34, doi:​10​.1046​/j​.1365​-246X​.2003​.01917.x. p. 3697–3715, doi:​10​.1002​/2015GC005958. Krien, Y., and Fleitout, L., 2008, Gravity above subduction zones and forces controlling plate Chen, Z., Schellart, W.P., Strak, V., and Duarte, J.C., 2016, Does subduction-induced mantle motions: Journal of Geophysical Research, v. 113, B09407, doi:​10​.1029​/2007JB005270. flow drive backarc extension?: Earth and Planetary Science Letters, v. 441, p. 200–210, Leng, W., and Zhong, S., 2010, Constraints on viscous dissipation of plate bending from doi:​10​.1016​/j​.epsl​.2016​.02​.027. compressible mantle convection: Earth and Planetary Science Letters, v. 297, p. 154–164, Chertova, M.V., Spakman, W., Geenen, T., van den Berg, A.P., and van Hinsbergen, D.J.J., doi:​10​.1016​/j​.epsl​.2010​.06​.016. 2014, Underpinning tectonic reconstructions of the western Mediterranean region with Long, M.D., and Silver, P.G., 2009, Mantle flow in subduction systems: The subslab flow field dynamic slab evolution from 3-D numerical modeling: Journal of Geophysical Research, and implications for mantle dynamics: Journal of Geophysical Research, v. 114, B10312, v. 119, p. 5876–5902, doi:​10​.1002​/2014JB011150. doi:​10​.1029​/2008JB006200. Cloos, M., 1993, Lithospheric buoyancy and collisional orogenesis: Subduction of oceanic plateaus, Mackwell, S.J., Bai, Q., and Kohlstedt, D.L., 1990, Rheology of olivine and the strength of the continental margins, island arcs, spreading ridges and seamounts: Geological Society of lithosphere: Geophysical Research Letters, v. 17, p. 9–12, doi:​10​.1029​/GL017i001p00009. America Bulletin, v. 105, p. 715–737, doi:10​ .1130​ /0016​ -7606​ (1993)​ 105​ <0715:​ LBACOS​ >2​ .3​ .CO;2.​ Meyer, C., and Schellart, W.P., 2013, Three-dimensional dynamic models of subducting plate- Conrad, C.P., and Behn, M.D., 2010, Constraints on lithosphere net rotation and asthenospheric overriding plate-upper mantle interaction: Journal of Geophysical Research, v. 118, p. 775– viscosity from global mantle flow models and seismic anisotropy: Geochemistry Geo- 790, doi:​10​.1002​/jgrb​.50078. physics Geosystems, v. 11, Q05W05, doi:​10​.1029​/2009GC002970. Minster, J.B., and Jordan, T.H., 1978, Present-day plate motions: Journal of Geophysical Re- Conrad, C.P., and Lithgow-Bertelloni, C., 2002, How mantle slabs drive plate tectonics: Sci- search, v. 83, p. 5331–5354, doi:​10​.1029​/JB083iB11p05331. ence, v. 298, p. 207–209, doi:​10​.1126​/science​.1074161. Molnar, P., and Stock, J.M., 2009, Slowing of India’s convergence with Eurasia since 20 Ma Crameri, F.H., Schmeling, H., Golabek, G.J., Duretz, T., Orendt, R., Buiter, S.J.H., May, D.A., and its implications for Tibetan mantle dynamics: Tectonics, v. 28, TC3001, doi:​10​.1029​ Kaus, B.J.P., Gerya, T.V., and Tackley, P.J., 2012, A comparison of numerical surface topog- /2008TC002271. raphy calculations in geodynamic modelling: An evaluation of the “sticky air” method: Moresi, L., Quenette, S., Lemiale, V., Mériaux, C., Appelbe, B., and Mühlhaus, H.-B., 2007, Geophysical Journal International, v. 189, p. 38–54, doi:​10​.1111​/j​.1365​-246X​.2012​.05388​.x. Computational approaches to studying non-linear dynamics of the crust and mantle: Davies, H.L., 2012, The geology of New Guinea—The cordilleran margin of the Australian Physics of the Earth and Planetary Interiors, v. 163, p. 69–82, doi:​10.1016​ /j​ .pepi​ .2007​ .06​ .009.​ continent: Episodes, v. 35, p. 87–102. Moresi, L., Betts, P.G., Miller, M.S., and Cayley, R.A., 2014, Dynamics of continental accretion: Davies, H.L., and Warren, R.G., 1992, Eclogites of the D’Entrecasteaux Islands: Contributions Nature, v. 508, p. 245–248, doi:​10​.1038​/nature13033. to Mineralogy and Petrology, v. 112, p. 463–474, doi:​10​.1007​/BF00310778. Müller, R.D., Sdrolias, M., Gaina, C., and Roest, W.R., 2008, Age, spreading rates, and spread- DeMets, C., Gordon, R.G., Argus, D.F., and Stein, S., 1994, Effect of recent revisions to the ing asymmetry of the world’s ocean crust: Geochemistry, Geophysics, Geosystems, v. 9, geomagnetic reversal time scale on estimates of current plate motions: Geophysical Q04006, doi:​10​.1029​/2007GC001743. Research Letters, v. 21, p. 2191–2194, doi:​10​.1029​/94GL02118. O’Neill, C., Müller, D., and Steinberger, B., 2005, On the uncertainties in hot spot reconstruc- Doubrovine, P.V., Steinberger, B., and Torsvik, T.H., 2012, Absolute plate motions in a reference tions and the significance of moving hot spot reference frames: Geochemistry, Geophys- frame defined by moving hot spots in the Pacific, Atlantic, and Indian Oceans: Journal of ics, Geosystems, v. 6, Q04003, doi:​10​.1029​/2004GC000784. Geophysical Research, v. 117, p. B09101, doi:10​ ​.1029​/2011JB009072. Patriat, P., and Achache, J., 1984, India-Eurasia collision chronology has implications for Duarte, J.C., Schellart, W.P., and Cruden, A.R., 2013, Three-dimensional dynamic laboratory crustal shortening and driving mechanism of plates: Nature, v. 311, p. 615–621, doi:​10​ models of subduction with an overriding plate and variable interplate rheology: Geo- .1038​/311615a0. physical Journal International, v. 195, p. 47–66, doi:​10​.1093​/gji​/ggt257. Royer, J.-Y., and Sandwell, D.T., 1989, Evolution of the eastern Indian Ocean since the Late Duarte, J.C., Schellart, W.P., and Cruden, A.R., 2015, How weak is the subduction zone in- Cretaceous; constraints from Geosat altimetry: Journal of Geophysical Research, v. 94, terface?: Geophysical Research Letters, v. 42, p. 2664–2673, doi:​10​.1002​/2014GL062876. p. 13,755–13,782, doi:​10​.1029​/JB094iB10p13755. Duretz, T., Schmalholz, S.M., and Gerya, T.V., 2012, Dynamics of slab detachment: Geochem- Sandiford, M., Coblentz, D., and Schellart, W.P., 2005, Evaluating slab-plate coupling in the istry Geophysics Geosystems, v. 13, Q03020, doi:​10​.1029​/2011GC004024. Indo-Australian plate: Geology, v. 33, p. 113–116, doi:​10​.1130​/G20898​.1. Duretz, T., Gerya, T.V., and Spakman, W., 2014, Slab detachment in laterally varying subduc- Schellart, W.P., 2004a, Kinematics of subduction and subduction-induced flow in the upper tion zones: 3-D numerical modeling: Geophysical Research Letters, v. 41, p. 1951–1956, mantle: Journal of Geophysical Research, v. 109, B07401, doi:​10​.1029​/2004JB002970. doi:​10​.1002​/2014GL059472. Schellart, W.P., 2004b, Quantifying the net slab pull force as a driving mechanism for plate Edwards, S.J., Schellart, W.P., and Duarte, J.C., 2015, Geodynamic models of continental tectonics: Geophysical Research Letters, v. 31, L07611, doi:​10​.1029​/2004GL019528. subduction and obduction of overriding plate forearc oceanic lithosphere on top of con- Schellart, W.P., 2008, Kinematics and flow patterns in deep mantle and upper mantle subduc- tinental crust: Tectonics, v. 34, p. 1494–1515, doi:​10​.1002​/2015TC003884. tion models: Influence of the mantle depth and slab to mantle viscosity ratio: Geochem- Forsyth, D.W., and Uyeda, S., 1975, On the relative importance of the driving forces of plate istry Geophysics Geosystems, v. 9, Q03014, doi:10​ ​.1029​/2007GC001656. motion: Royal Astronomical Society Geophysical Journal, v. 43, p. 163–200, doi:10​ ​.1111​ Schellart, W.P., 2011, A subduction zone reference frame based on slab geometry and sub- /j​.1365​-246X​.1975​.tb00631​.x. duction partitioning of plate motion and trench migration: Geophysical Research Letters, Funiciello, F., Faccenna, C., and Giardini, D., 2004, Role of lateral mantle flow in the evolu- v. 38, L16317, doi:​10​.1029​/2011GL048197. tion of subduction systems: insights from laboratory experiments: Geophysical Journal Schellart, W.P., and Moresi, L., 2013, A new driving mechanism for backarc extension and International, v. 157, p. 1393–1406, doi:​10​.1111​/j​.1365​-246X​.2004​.02313​.x. backarc shortening through slab sinking induced toroidal and poloidal mantle flow: Re- Gaina, C., Müller, D.R., Royer, J.-Y., Stock, J., Hardebeck, J.L., and Symonds, P., 1998, The sults from dynamic subduction models with an overriding plate: Journal of Geophysical tectonic history of the Tasman Sea; a puzzle with 13 pieces: Journal of Geophysical Re- Research, v. 118, p. 3221–3248, doi:​10​.1002​/jgrb​.50173. search, v. 103, p. 12,413–12,433, doi:​10​.1029​/98JB00386. Schellart, W.P., and Spakman, W., 2012, Mantle constraints on the plate tectonic evolution Gerya, T.V., Yuen, D.A., and Maresch, W.V., 2004, Thermomechanical modelling of slab detach- of the Tonga-Kermadec-Hikurangi subduction zone and the South Fiji Basin: Australian ment: Earth and Planetary Science Letters, v. 226, p. 101–116, doi:10​ .1016​ /j​ .epsl​ .2004​ .07​ .022.​ Journal of Earth Sciences, v. 59, p. 933–952, doi:​10​.1080​/08120099​.2012​.679692. Goes, S., Capitanio, F.A., and Morra, G., 2008, Evidence of lower-mantle slab penetration Schellart, W.P., and Spakman, W., 2015, Australian plate motion and topography linked to phases in plate motions: Nature, v. 451, p. 981–984, doi:​10​.1038​/nature06691. fossil New Guinea slab below Lake Eyre: Earth and Planetary Science Letters, v. 421, Gordon, R.G., and Jurdy, D.M., 1986, Cenozoic global plate motions: Journal of Geophysical p. 107–116, doi:​10​.1016​/j​.epsl​.2015​.03​.036. Research, v. 91, p. 12,389–12,406, doi:​10​.1029​/JB091iB12p12389. Schellart, W.P., Stegman, D.R., and Freeman, J., 2008, Global trench migration velocities Gripp, A.E., and Gordon, R.G., 2002, Young tracks of hotspots and current plate velocities: and slab migration induced upper mantle volume fluxes: Constraints to find an Earth Geophysical Journal International, v. 150, p. 321–361, doi:10​ .1046​ /j​ .1365​ -246X​ .2002​ .01627​ .x.​ reference frame based on minimizing viscous dissipation: Earth-Science Reviews, v. 88, Guillaume, B., Martinod, J., and Espurt, N., 2009, Variations of slab dip and overriding plate p. 118–144, doi:​10​.1016​/j​.earscirev​.2008​.01​.005. tectonics during subduction: Insights from analogue modelling: Tectonophysics, v. 463, Schellart, W.P., Stegman, D.R., Farrington, R.J., Freeman, J., and Moresi, L., 2010, Cenozoic p. 167–174, doi:​10​.1016​/j​.tecto​.2008​.09​.043. tectonics of western North America controlled by evolving width of Farallon slab: Sci- Hall, R., 2012, Late Jurassic–Cenozoic reconstructions of the Indonesian region and the Indian ence, v. 329, p. 316–319, doi:​10​.1126​/science​.1190366. Ocean: Tectonophysics, v. 570–571, p. 1–41, doi:​10​.1016​/j​.tecto​.2012​.04​.021. Schellart, W.P., Stegman, D.R., Farrington, R.J., and Moresi, L., 2011, Influence of lateral slab Jacoby, W.R., 1973, Model experiment of plate movements: Nature. Physical Science (Lon- edge distance on plate velocity, trench velocity and subduction partitioning: Journal of don), v. 242, p. 130–134, doi:​10​.1038​/physci242130a0. Geophysical Research, v. 116, B10408, doi:​10​.1029​/2011JB008535. Jagoutz, O., Royden, L., Holt, A.F., and Becker, T.W., 2015, Anomalously fast convergence of Schmeling, H., et al., 2008, A benchmark comparison of spontaneous subduction models— India and Eurasia caused by double subduction: Nature Geoscience, v. 8, p. 475–478, Towards a free surface: Physics of the Earth and Planetary Interiors, v. 171, p. 198–223, doi:​10​.1038​/ngeo2418. doi:​10​.1016​/j​.pepi​.2008​.06​.028.

LITHOSPHERE | Volume 9 | Number 6 | www.gsapubs.org 985 Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018 SCHELLART

Seton, M., et al., 2012, Global continental and ocean basin reconstructions since 200 Ma: Wessel, P., and Müller, R.D., 2016, Ridge-spotting: A new test for Pacific absolute plate mo- Earth-Science Reviews, v. 113, p. 212–270, doi:​10​.1016​/j​.earscirev​.2012​.03​.002. tion models: Geochemistry Geophysics Geosystems, v. 17, p. 2408–2420, doi:​10​.1002​ Sizova, E., Gerya, T., and Brown, M., 2012, Exhumation mechanisms of melt-bearing ultrahigh /2016GC006404. pressure crustal rocks during collision of spontaneously moving plates: Journal of Meta- White, L.T., and Lister, G.S., 2012, The collision of India with Asia: Journal of Geodynamics, morphic Geology, v. 30, p. 927–955, doi:​10​.1111​/j​.1525​-1314​.2012​.01004​.x. v. 56–57, p. 7–17, doi:​10​.1016​/j​.jog​.2011​.06​.006. Stegman, D.R., Freeman, J., Schellart, W.P., Moresi, L., and May, D., 2006, Influence of trench Wortel, M.J.R., and Spakman, W., 2000, Subduction and slab detachment in the Mediterra- width on subduction hinge retreat rates in 3-D models of slab rollback: Geochemistry nean-Carpathian region: Science, v. 290, p. 1910–1917, doi:​10​.1126​/science​.290​.5498​.1910. Geophysics Geosystems, v. 7, Q03012, doi:​10​.1029​/2005GC001056. Worthing, M.A., and Crawford, A.J., 1996, The igneous geochemistry and tectonic setting Stegman, D.R., Farrington, R., Capitanio, F.A., and Schellart, W.P., 2010, A regime diagram for of metabasites from the Emo metamorphics, Papua New Guinea; a record of the evolu- subduction styles from 3-D numerical models of free subduction: Tectonophysics, v. 483, tion and destruction of a backarc basin: Mineralogy and Petrology, v. 58, p. 79–100, doi:​ p. 29–45, doi:​10​.1016​/j​.tecto​.2009​.08​.041. 10​.1007​/BF01165765. van der Meer, D.G., Spakman, W., van Hinsbergen, D.J.J., Amaru, M.L., and Torsvik, T.H., 2010, Yan, C.Y., and Kroenke, L.W., 1993, A plate tectonic reconstruction of the Southwest Pacific, Towards absolute plate motions constrained by lower-mantle slab remnants: Nature 0–100 Ma, in Berger, W.H., et al., Proceedings of the Ocean Drilling Program, Scientific Geoscience, v. 3, p. 36–40, doi:​10​.1038​/ngeo708. Results, Volume 130: College Station, Texas, Ocean Drilling Program, p. 697–709, doi:​10​ van Hinsbergen, D.J.J., Steinberger, B., Doubrovine, P.V., and Gassmöller, R., 2011, Accelera- .2973​/odp​.proc​.sr​.130​.055​.1993. tion and deceleration of India‐Asia convergence since the Cretaceous: Roles of mantle Zahirovic, S., Müller, R.D., Seton, M., Flament, N., Gurnis, M., and Whittaker, J., 2012, Insights plumes and continental collision: Journal of Geophysical Research, v. 116, B06101, doi:​ on the kinematics of the India-Eurasia collision from global geodynamic models: Geo- 10​.1029​/2010JB008051. chemistry Geophysics Geosystems, v. 13, Q04W11, doi:​10​.1029​/2011GC003883. van Hunen, J., and Allen, M.B., 2011, Continental collision and slab break-off: A comparison of 3-D numerical models with observations: Earth and Planetary Science Letters, v. 302, MANUSCRIPT RECEIVED 22 APRIL 2017 p. 27–37, doi:​10​.1016​/j​.epsl​.2010​.11​.035. REVISED MANUSCRIPT RECEIVED 17 JULY 2017 Wessel, P., and Kroenke, L.W., 2008, Pacific absolute plate motion since 145 Ma: An assess- MANUSCRIPT ACCEPTED 24 AUGUST 2017 ment of the fixed hot spot hypothesis: Journal of Geophysical Research, v. 113, B06101, doi:​10​.1029​/2007JB005499. Printed in the USA

986 www.gsapubs.org | Volume 9 | Number 6 | LITHOSPHERE Downloaded from https://pubs.geoscienceworld.org/gsa/lithosphere/article-pdf/9/6/976/3985681/976.pdf by guest on 18 September 2018