Lithosphere Delamination in Continental Collisional Orogens: A

PUBLICATIONS

Journal of Geophysical Research: Solid Earth

RESEARCH ARTICLE Lithosphere delamination in continental collisional 10.1002/2016JB013106 orogens: A systematic numerical study Key Points: Zhong-Hai Li1, Mian Liu1,2, and Taras Gerya3 • Negative buoyancy from thickened lithosphere drives delamination 1Key Laboratory of Computational Geodynamics, College of Earth Sciences, University of Chinese Academy of Sciences, • Rheology, density, and convergence 2 3 rate control lithosphere delamination Beijing, China, Department of Geological Sciences, University of Missouri, Columbia, Missouri, USA, Institute of modes Geophysics, Department of Earth Sciences, ETH Zurich, Zurich, Switzerland • Initiation and migration styles of pro- plate versus retro-plate delamination Abstract Lithosphere delamination is believed to have played a major role in mountain building; however, Supporting Information: the mechanism and dynamics of delamination remain poorly understood. Using a 2-D high-resolution • Supporting Information S1 thermomechanical model, we systematically investigated the conditions for the initiation of lithosphere delamination during orogenesis of continental collision and explored the key factors that control the various Correspondence to: modes of delamination. Our results indicate that the negative buoyancy from lithosphere thickening during Z.-H. Li, [email protected] orogenesis could cause delamination, when the reference density of the lithospheric mantle is not lower than that of the asthenosphere. In these cases, compositional rejuvenation of depleted continental lithosphere by magmatic/metasomatic plume- and/or subduction-induced processes may play crucial roles for subsequent Citation: Li, Z.-H., M. Liu, and T. Gerya (2016), lithosphere delamination. If the reference density of the lithospheric mantle is less than that of the Lithosphere delamination in continental asthenosphere, additional promoting factors, such as lower crust eclogitization, are required for delamination. collisional orogens: A systematic Our numerical simulations predict three basic modes of lithosphere delamination: pro-plate delamination, numerical study, J. Geophys. Res. Solid Earth, 121, doi:10.1002/2016JB013106. retro-plate delamination, and a transitional double-plates (both the pro-plate and retro-plate) delamination. Pro-plate delamination is favored by low convergence rates, high lithospheric density, and relatively strong Received 21 APR 2016 retro-plate, whereas retro-plate delamination requires a weak retro-plate. The Northern Apennines and Central Accepted 14 JUN 2016 Northern Tibetan Plateau are possible geological analogs for the pro-plate and retro-plate delamination modes, Accepted article online 19 JUN 2016 respectively. Our model also shows signiﬁcant impact of delamination on the topographic evolution of orogens. Large-scale lithosphere delamination in continental collision zones would lead to wide and ﬂat plateaus, whereas relatively narrow and steep mountain belts are predicted in orogens without major delamination.

1. Introduction Continental lithosphere delamination refers to the detachment and sinking (foundering) of lithospheric mantle, with or without a certain portion of the lower crust, from the tectonic plate. The concept of lithosphere delamination was introduced by Bird [1978, 1979], who proposed that the dense lithospheric mantle might peel off from the crust and sink into the underlying asthenosphere, if with an elongated conduit connecting the hot asthenosphere with the base of continental crust. A similar mechanism is the convective removal or thinning of continental lithosphere by the Rayleigh-Taylor gravitational instabilities developed in a thickened lithospheric mantle [e.g., Houseman et al., 1981; England and Houseman, 1989; Houseman and Molnar, 1997; Liu et al., 2004; Gorczyk et al., 2012; Gorczyk and Vogt, 2013]. In both cases the mantle lithosphere is removed and sinks into the asthenosphere because of its negative buoyancy. Delamination is used to explain the thinned or absent mantle lithosphere under orogens where the same processes that thickened the crust should have also thickened the mantle lithosphere. The examples include many orogens in the Alpine-Himalayan collisional belt such as the Tibetan Plateau [Bird, 1978; Jimenez-Munt et al., 2008], the Anatolia [Keskin, 2003; Gogus and Pysklywec, 2008a], the Apennines [Giacomuzzi et al., 2011; Chiarabba et al., 2014], and the Alboran Sea [Seber et al., 1996; Timoulalia et al., 2014]. Other orogenic belts with proposed delamination include the Cordillera in western America (both North and South), including the Altiplano-Puna Plateau [Kay and Kay, 1993; Beck and Zandt, 2002], the Sierra Nevada [Ducea and Saleeby, 1998; Zandt et al., 2004], the Colorado Plateau [Bird,1979;Levander et al., 2011], and the Canadian Cordillera [Bao et al., 2015]. Delamination has also been proposed for both ancient cratons such as the North China craton [Gao et al., 2009; Zhu et al., 2012] and young orogens such as the New Guinea [Cloos et al., 2005].

©2016. American Geophysical Union. In collisional orogens, lithosphere delamination may occur in either the overriding plate or the subducting All Rights Reserved. plate, or both, with variable spatial patterns. In the Himalayan-Tibetan system, delamination, or convective

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 1 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

thinning, is proposed [England, 1993; Molnar et al., 1993] to have occurred under Central Northern Tibetan Plateau, where the lithospheric mantle is thought to be abnormally thin or totally missing [Owens and Zandt, 1997; Tilmann et al., 2003]. In contrast, delamination could occur in the subducting continental plate, with the mantle lithosphere peeling off from the overlying crust. For example, the Northern Apennines is a fold-and-thrust belt that formed in response to the postcollisional retreating of the subducted Adria microcontinental lithosphere that is delaminated from the overlying crust [Faccenna et al., 2001; Faccenda et al., 2009; Chiarabba and Chiodini, 2013; Chiarabba et al., 2014]. Both analog and numerical models have been used to study the dynamics of delamination [e.g., Morency and Doin,2004;Valera et al., 2008, 2011; Gogus et al., 2011; Bajolet et al., 2012; Francois et al., 2014]. In previous numerical studies, lithosphere delamination was usually induced by prescribed weak zones in the initial model conﬁguration [e.g., Bird, 1979; Schott and Schmeling, 1998; Gogus and Pysklywec, 2008a, 2008b; Bajolet et al., 2012; Wang and Currie, 2015]. These models focus on the consequences of lithosphere delamination on the topography as well as the mechanical and thermal evolution of the overlying continent. In addition, some models simulate delamination induced by sinking and retreating of high-density subducting lithosphere decoupled from its overlying low-density crust [e.g., Beaumont et al., 2006; Faccenda et al., 2009; Gray and Pysklywec,2012; Ueda et al., 2012]. Alternatively, other models simulate various delamination processes induced by eclogitization of the thickened crust under modern [e.g., Krystopowicz and Currie, 2013] or Archean mantle temperature conditions [e.g., Johnson et al., 2014; Sizova et al., 2015; Fischer and Gerya, 2016]. In this study, we constructed two-dimensional (2-D) high-resolution thermomechanical models to systematically investigate the initiation and evolution of continental lithosphere delamination during collisional orogenesis. We found that the modes of delamination depend on the rheology of the collisional plates, the density contrast between lithospheric mantle and the asthenosphere, the eclogitization of the thickened crust, and the convergence rates.

2. Numerical Methodology 2.1. Governing Equations The lithospheric deformation and delamination are simulated numerically by solving the governing equations of conservation of mass, momentum, and energy and the constitutive equations with the 2-D ﬁnite difference and marker-in-cell methods in an irregularly spaced staggered grid in Eulerian coordinates [Gerya and Yuen, 2003, 2007; Gerya, 2010; Li, 2014]. 1. Mass conservation (Continuity equation) The conservation of mass is approximated by the incompressible continuity equation (extended Boussinesq approximation): ∂v ∂v x þ y ¼ 0 (1) ∂x ∂y

where vx and vy are velocities in the x and y directions, respectively. 2. Momentum conservation (Stokes equations)

∂σ′ ∂σ′ ∂ xx þ xy ¼ P ∂x ∂y ∂x ′ ′ (2) ∂σ ∂σ ∂P yx þ yy ¼ gρðÞC; P; T ∂x ∂y ∂y

where x and y denote the horizontal and vertical coordinates; P is the dynamic pressure; g is the gravitational σ′ σ′ σ′ acceleration; xx, xy,and yy are the deviatoric stress tensor components, with constitutive relations as follows: : : : σ′ ¼ η ε ; σ′ ¼ η ε ; σ′ ¼ η ε xx 2 eff xx xy 2 eff xyyy 2 eff yy : ∂ : ∂ : ∂ ∂ ε ¼ vx; ε ¼ vy; ε ¼ 1 vx þ vy (3) xx ∂x yy ∂y xy 2 ∂y ∂x η where: : eff is: the effective viscosity that depends on the composition, pressure, temperature, and strain rate; εxx, εyy, and εxy are the deviatoric strain rate components.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 2 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Table 1. Viscous Flow Laws Used in the Numerical Experimentsa 1 1 1 n 1 ID Symbol Flow Law E (kJ mol ) V (J MPa mol ) nAD (MPa s ) η0 (Pa s) RA Wet quartzite 154 8 2.3 3.2 × 10 4 1.97 × 1017 4 22 RB Plagioclase An75 238 8 3.2 3.3 × 10 4.80 × 10 RC Dry olivine 532 8 3.5 2.5 × 104 3.98 × 1016 RD Wet olivine 470 8 4.0 2.0 × 103 5.01 × 1020 a (6n) η0 is the reference viscosity, which is calculated by η0 = (1/AD)×10 . References are from Kirby [1983], Kirby and Kronenberg [1987], Ranalli and Murphy [1987], Ji and Zhao [1993], and Ranalli [1995].

For a speciﬁc rock type, i.e., composition (C), its density ρ depends explicitly on pressure (P)andtemperature(T): ρ ¼ ρ ½ αðÞ ½þ βðÞ P;T 0 1 T T 0 1 P P0 (4)

where the reference density ρ0 is the density under the conditions of P0 = 0.1 MPa and T0 = 298 K (i.e., pressure and temperature at Earth’s surface); α and β are the thermal expansion coefﬁcient and the compressibility coefﬁcient, respectively. 3. Energy conservation (Energy equation) DT ∂q ∂q ρC ¼ x y þ H þ H þ H p Dt ∂x ∂y r a S ∂T ∂T q ¼kTðÞ; P; C ; q ¼kTðÞ; P; C x ∂x y ∂y (5) dP H ¼ Tα ; a d:t : : ¼ σ′ ε þ σ′ ε þ σ′ ε HS xx xx yy yy 2 xy xy

where T represents temperature; Cp is the effective isobaric heat capacity; DT/Dt is the substantive (material) time derivative of temperature; qx and qy are the thermal heat flux components; k is the thermal conductivity, which depends on temperature (T), pressure (P), and composition (C); Hr, Ha, and HS denote the radioactive heat production, adiabatic heat production, and shear heating, respectively. 2.2. Viscoplastic Rheology The viscosity of rocks deforming by dislocation creep is defined as : 1n 1 E þ PV η ¼ ðÞε n ðÞA nexp (6) ductile II D nRT ÀÁ : : : 1=2 where εII ¼ 0:5εij εij is the second invariant of the strain rate tensor; AD (preexponential factor), E (activation energy), V (activation volume), and n (creep exponent) are experimentally determined flow law parameters (Table 1). The parameter R is the gas constant. We combine the viscous rheology with an extended Drucker-Prager yield criterion [e.g., Ranalli, 1995], which is equivalent to Mohr-Coulomb criterion in 2-D plane strain approximation, to simulate the viscoplastic beha- vior of the rocks: σ η ¼ yield: plastic ε 2 II (7) σ ¼ þ φ λ yield C0 Psin dry

where σyield is the yield stress; P is the dynamic pressure; C0 is the residual rock strength at P =0;φdry is the internal frictional angle of dry rocks; λ is the pore fluid/melt coefficient that controls the brittle strength of fl λ ¼ Pfluid=melt uid/melt containing porous or fractured media, which can be interpreted as 1 P . Because the fluid/melt parameter Pfluid/melt is not directly calculated in the model, we use a range of “sin(φdry)λ” values to represent the effective friction coefficient, based on previous systematic investigations [e.g., Gerya and Meilick, 2011; Vogt et al., 2012; Gerya et al., 2015]. The minimum value of the viscous or plastic viscosity defines the effective composite viscosity in the model [Ranalli, 1995]: η ¼ η ; η eff min ductile plastic (8)

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 3 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 1

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 4 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Table 2. Material Properties Used in the Numerical Experimentsa b 3 c 1 1 3 d e e Section Material ρ0 (kg m ) k (W m K ) Hr (μWm ) Viscous Flow Law Plastic C0 (MPa) Plastic sin(ϕdry)λ

All plates Sediment (1, 2) 2700 K1 1 RA 10 0.15 Weak-zone mantle (9) 3300 K3 0.022 RD 10 0.006 Asthenosphere (10) 3300 K3 0.022 RC 10 0.6 Pro-plate and end-plate Upper crust (3) 2700 K1 1 RA 10 0.15 Lower crust (4) 3000 K2 1 RB 10 0.15 Lithospheric mantle (7) 3300 K3 0.022 RC 10 0.6 Retro-plate Upper crust (5) 2700 K1 1 RA 10 0.15 Lower crust (6) 3000 K2 1 RB/RA 10 0.15 Lithospheric mantle (8) 3300 K3 0.022 RC/RD 10 0.6/0.006 Referencesf 1, 2 3 1 4 5 5 a 1 1 5 1 5 1 The isobaric heat capacity Cp = 1000 J kg K , thermal expansion coefficient α =3×10 K , and the compressibility coefficient β =1×10 MPa are used for all rock types. bNumbers of materials correspond to Figure 1c. c K1 = [0.64 + 807/(TK + 77)] exp(0.00004PMPa); K2 = [1.18 + 474/(TK + 77)] exp(0.00004PMPa); K3 = [0.73 + 1293/(TK + 77)] exp(0.00004PMPa). dParameters of viscous flow laws are shown in Table 1. e Strain weakening effect is applied for plastic rheology, in which both cohesion (C0) and effective friction coefficient (sin(ϕdry)λ) decrease with larger strain rate. fReferences 1–5 are Turcotte and Schubert [1982], Bittner and Schmeling [1995], Clauser and Huenges [1995], Ranalli [1995], and Vogt et al. [2012], respectively.

3. Initial Model Configuration and Boundary Conditions We used a large-scale model domain (4000 × 400 km) to study the dynamics of continental collision and lithosphere delamination. Using a nonuniform rectangular numerical grid, we represent the collision zone with a 1 × 1 km high resolution while using a 5 × 1 km for the rest of the model domain. A dense grid with more than 15 million of active Lagrangian markers is used to trace and mark internal lithological boundaries, material properties, and temperature. Although the model length is 4000 km, the deformation localizes in a relatively narrow (ca. 1000 km) region of interest. Effects of the Earth’s curvature are therefore neglected in our simplified Cartesian model. In the initial model, the continental plate includes a pro-plate, a retro-plate, and an end-plate (Figure 1b). An initial weak zone is placed between the pro-plate and retro-plate, which is applied to locate the initial convergence, not for initiating delamination. The initial continental lithosphere includes a 20 km thick upper crust, a 15 km thick lower crust, a 65 km thick lithospheric mantle, and the subjacent asthenosphere. Properties of variable rock types (Figure 1c) are summarized in Tables 1 and 2. The initial thermal structure of the lithosphere (white lines in Figure 1b) is laterally uniform with a linear gradient defined by 0°C at the surface and 1350°C at the bottom of the lithospheric mantle [e.g., Turcotte and Schubert, 2002]. The initial thermal gradient in the asthenosphere is 0.5°C/km. The top surface of the lithosphere is calculated dynamically as an essentially internal free surface by using a buffer layer of “sticky air” [Gerya and Yuen, 2003; Schmeling et al., 2008; Crameri et al., 2012]. It is an initially 10 km thick layer with a low viscosity (1018 Pa s) above the upper crust. Crameri et al. [2012] found that the 3 sticky air approach is adequate as long as the term (ηst/ηch)/(hst/L) is small, where ηst and hst are the viscosity

Figure 1. Initial model configuration. (a) Topography. (b) Composition and temperature. Enlargement (1700 × 400 km) of the numerical box (4000 × 400 km) is shown. White lines are isotherms in °C. Colors indicate the rock types, specified by the (c) colorgrid: 1,2, sediment; 3,4, continental upper and lower crust of the pro-plate and end-plate, respectively; 5,6, continental upper and lower crust of the retro-plate, respectively; 7, lithospheric mantle of the pro-plate and end-plate; 8, lithospheric mantle of the retro-plate; 9, initial weak zone mantle; 10, asthenosphere. (d)–(f) Three different configurations fi ρLith of initial density pro le, with changing reference density contrast between the lithospheric mantle ( 0 ) and astheno- ρAsth fi sphere ( 0 ) as shown in the gures. The density property of the initial model is homogeneous in the horizontal x direction. (g)–(j) Four different configurations of the initial rheological properties, with changing effective viscosity of the retro-plate lower crust and/or lithospheric mantle. The inset figures demonstrate the vertical cross sections, i.e., black solid lines, in the retro-plate. Pro-plate, retro-plate, and end-plate with the same rheological layering, i.e., homogeneous layering model (Figure 1g). Weak lower crust of the retro-plate (Figure 1h). Weak lithospheric mantle of the retro-plate (Figure 1i). Both weak lower crust and weak lithospheric mantle of the retro-plate (Figure 1j). The flow laws used in the numerical models can be found in the text and Tables 1 and 2. It is worth noting that the rheological configurations of pro-plate and end-plate are kept unchanged. The strain rate dependence of rheological properties (equation 1) may result in differences of the pro-plate effective viscosity in different models (cf. Figure 1j and Figures 1g–1i).

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 5 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

and thickness of the sticky air layer, and ηch and L are the characteristic viscosity and length scale of the model, respectively. According to this criterion, the quality of the internal free-surface condition in our models is rather moderate. Further decrease in the sticky air viscosity is precluded by numerical limitations. Indeed, the large viscosity contrast caused by this low-viscosity boundary layer minimizes shear stresses at the top of the solid portion of the model [Burg and Gerya, 2005]. The interface between this weak layer and the underlying crust is treated as an internal erosion/sedimentation surface [e.g., Burg and Gerya, 2005]. Large-scale erosion and sedimentation rates in our models are assumed to be negligible, but surface

slopes are limited by prescribing the maximum stable slope angle “θmax”. In the numerical models, tan(θmax) = 0.1 is applied, which is maintained at each time step by computing an instantaneous mass redistribution in the regions of unstable surface slopes. Our parameter studies indicate that variability of the prescribed maximum slope angle within a realistic range does not affect significantly the large-scale model evolution (Figure S1 in the supporting information). One of the most important driving forces for lithosphere delamination is the positive density contrast between the lithospheric mantle and the asthenosphere. However, the reference density, i.e., at the same pressure and temperature, of the melt-depleted continental lithospheric mantle is generally lower than that of more fertile asthenosphere [Djomani et al., 2001; Schutt and Lesher, 2006]. The lithospheric density may, however, be increased by eclogitized magmatic intrusions (dykes) from (ancient) plumes [Sobolev et al., 2011], which could result in compositionally dense lithospheric mantle (e.g., pyroxenite), compared to the underlain peridotite mantle [Jull and Kelemen, 2001; Lee et al., 2006, 2011]. In this aspect, compositional rejuvenation of depleted continental lithosphere by magmatic/metasomatic plume- and/or subduction-induced processes may result in positive density contrast between the lithospheric mantle and asthenosphere. In addition, thermal contraction can also contribute to high density of the lithosphere, because it is cooler than the asthenosphere [Houseman and Molnar, 1997]. In this study, we construct three sets of models with, respectively, neutral (Figure 1d), negative (Figure 1e), and positive (Figure 1f) reference density contrast between the lithospheric mantle and asthenosphere, to explore how the compositional density contrast affects the continental collision and delamination. Besides the density contrast, rheological properties also play significant roles in lithosphere delamination [e.g., Krystopowicz and Currie, 2013]. However, there has been much debate concerning the long-term (i.e., >1 Ma) strength of continental lithosphere. In one model, dubbed “jelly sandwich,” the strength resides in the crust and the uppermost mantle; whereas in another model, dubbed “creme brulee,” the mantle is weak and the strength is limited to the crust [Kohlstedt et al., 1995; Turcotte and Schubert, 2002; Burov and Watts, 2006]. Therefore, we have designed four different rheological profiles of the retro-plate lithosphere (Figures 1g–1j). In the first model, the same rheological profile is applied for the pro-plate, retro-plate, and end-plate (Figure 1g), in which the flow laws of “wet quartzite” is used for the upper continental crust,

“Plagioclase An75” for the lower continental crust, and “dry Olivine” for the lithospheric mantle and the asthenosphere (Tables 1 and 2), as generally used for a strong continental lithosphere. In the second model, the effective viscosity of the retro-plate lower crust is decreased by using the rheology of wet quartzite, same as in the upper crust (Figure 1h). This retro-plate rheology has a jelly sandwich style with strong upper crust and lithospheric mantle but weak lower crust. In the ﬁrst two models, where dry olivine rheology is used for all sections of the lithospheric mantle (Figures 1g and 1h), the ﬂuid/melt weakening effects are neglected.

Thereby, high plastic effective friction coefficient (sin(φdry)λ = 0.6) is used for the dry and strong lithospheric mantle (equation (7)). However, in the third model, we assume that the lithospheric mantle of the retro-plate experienced potential fluid/melt weakening during the previous oceanic subduction or terrane accretion, which may lead to a creme brulee rheological profile. Consequently, the wet olivine rheology is used for the retro-plate lithospheric mantle [Ranalli, 1995]. In addition, low plastic effective friction coefficient (sin

(φdry)λ = 0.006) is used for the weak lithospheric mantle (Figure 1i), based on previous systematic investigations of the ﬂuid/melt effect during subduction and/or plume-lithosphere interactions [e.g., Gerya and Meilick, 2011; Vogt et al., 2012; Gerya et al., 2015]. The fourth model combines the weak lower continental crust (Figure 1h) and weak lithospheric mantle (Figure 1i) of the retro-plate, which, thereby, is the end- member scenario of the weakest retro-plate lithosphere (Figure 1j). The velocity boundary conditions are free slip for the left, right, and top boundaries. The lower boundary is permeable, along which a mass-conservative permeable boundary condition is imposed [e.g., Burg and Gerya, 2005]. This inﬁnity-like external free slip condition along the lower boundary implies free slip condition

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 6 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

to be satisﬁed at about 100 km below the base of the model (external lower boundary as indicated in Figure 1b). As for the usual free slip condition, external free slip allows global conservation of mass in the computational domain and is implemented by using the following limitation for velocity components at the lower boundary:

∂vx/∂y = 0 and ∂vy/∂y = vy/Δyexternal,whereΔyexternal is the vertical distance from the lower boundary to the external boundary where free slip (∂vx/∂y =0,vy =0)issatisﬁed. It is worth noting that the external region, i.e., dashed rectangle below the permeable lower boundary and above the external boundary, is not modeled directly, but its presence is implied by the permeable boundary condition. The pro-plate is pushed rightward

at a constant convergence rate (Vx) imposed to a small internal domain in the left part of the pro-plate that remains fixed with respect to the Eulerian coordinate (Figure 1b). No convergent condition is applied to the right side of the mode, i.e., to the end-plate. The thermal boundary conditions have a fixed value (0°C) for the upper boundary and zero horizontal heat flux across the vertical boundaries. For the lower thermal boundary, an infinity-like external constant temperature condition is imposed (e.g., Figure 1b), which allows both temperatures and vertical heat fluxes to vary along the permeable lower boundary [e.g., Burg and Gerya, 2005], implying that the constant temperature condition is satisfied at about 1000 km below the bottom of the model (Figure 1b). This condition is

implemented by requiring ∂T/∂y =(Texternal T)/Δyexternal, where Texternal is the temperature at the external boundary and Δyexternal is the vertical distance from the lower boundary to the external boundary. This condition allows the temperature and heat ﬂux at the bottom boundary to be adjusted dynamically during the simulation. It is worth noting that the external boundary for temperature is fully independent from the external velocity boundary.

4. Model Results With three different density proﬁles (Figures 1d–1f) and four different rheological conﬁgurations (Figures 1g–1j), we constructed 12 numerical experiments, which demonstrate variable modes of continental collision and lithosphere delamination.

ρLith ¼ ρAsth 4.1. Neutral Reference Density of the Lithospheric Mantle and Asthenosphere ( 0 0 ) 4.1.1. A Model of Homogeneous Layering To illustrate the effects of rheological structure of the colliding continents on delamination, we started with a reference model in which the rheological properties for the pro-plate, retro-plate, and end-plate are the same (Figure 1g). In this model, the pro-plate is pushed from the left side at 5 cm/yr, causing it to subduct continuously into the asthenosphere. In the process its upper-middle crust, being relatively weak, is scraped off (Figure 2a). The detached crustal materials are accumulated in the collision zone and accreted to the overriding retro-plate. During this process, local erosion and sedimentation take place, which is mainly controlled by the prescribed maximum surface slope angle. When the slope is steep, the materials in the higher side are eroded and then deposited in the neighboring lower side as sediments. Consequently, a belt with series of thrust faults are formed, in which sediments become incorporated in the upper crust as dipping wedges (Figures 2a–2c). The upper crust of the retro-plate is also detached from the lower crust and thickened by the collision (Figures 2b and 2c). In this case, the surface location of the suture between pro-plate and retro-plate is different from that at the depth of lithospheric mantle (Figure 2c). During the convergence, the collision zone uplifts as the upper crust thickens there, reaching as high as 8 km from the original elevation of 1 km (Figures 2c and 1a). The effective viscosity of the middle crust of the thickened collisional orogen is particularly low (Figure 2f) because of crustal thickening and temperature increase during collision. In the sublithospheric depth, a certain portion of the subducted lower crust may detach from the slab, intruding into the mantle wedge (Figures 2b and 2c) and forming a compositionally buoyant (Figures 2d and 2e) and rheologically weak (Figure 2f) thermal-chemical crustal plume (diapir), the dynamics of which is systematically investigated in Currie et al. [2007] and Li and Gerya [2009]. The growth and migration of the sublithospheric crustal plume, as well as the induced small-scale mantle ﬂow (Figure 2f), leads to bottom erosion of the overriding lithosphere, which further results in a limited, small-scale delamination of the retro- plate lithosphere (Figure 2c). However, no major lithosphere delamination is predicted in this model of homogeneous rheological layering.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 7 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 2. Results of the model with laterally homogeneous layering (Figure 1g) and neutral reference density contrast between the lithospheric mantle and the asthenosphere (Figure 1d). (a)–(c) Composition and temperature evolution. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel. The black triangle in Figure 2c indicates the crustal suture zone between the pro-plate and retro-plate. The (d) density field, (e) density anomaly field, and (f) effective viscosity field with velocity vectors are shown for the last time step of 20.0 Ma (Figure 2c).

4.1.2. Model With a Weak Lower Crust of the Retro-plate In the second model, we test the effects of a weak lower crust in the retro-plate on potential development of delamination. In this case, the rheological property of the lower crust is changed from that of a relatively

strong Plagioclase An75 to that of a wet quartzite (Figure 1h and Table 1). The model results are generally similar to those of the reference model (cf. Figures 2c and 3a, both at 20.0 Ma). The weak lower crust in the retro-plate alone is insufﬁcient to induce major delamination. The plate convergence is still dominated by subduction of the pro-plate.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 8 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 3. Results of the model with weak lower crust of the retro-plate (Figure 1h) and neutral reference density contrast between the lithospheric mantle and the asthenosphere (Figure 1d). (a) Composition and temperature. Colors of rock types are as in Figure 1c. The black triangle in Figure 3a indicates the crustal suture zone between the pro-plate and retro-plate. (b) Density field. (c) Density anomaly field. (d) Effective viscosity field with velocity vectors. Time (Ma) of shortening is 20.0 Ma, as shown in the panels.

4.1.3. Model With a Weak Lithospheric Mantle of the Retro-plate In the third model, we show the effects of a weak lithospheric mantle in the retro-plate on delamination. Here we replace the rheological properties of the mantle lithosphere of the retro-plate with those of a weak wet oli-

vine (Table 1) and decreased the effective friction coefficient sin(φdry)λ for a lower plastic yield stress (Figure 1i). The results are significantly different from the previous two cases, with plate convergence mainly accommodated by thickening of the whole retro-plate lithosphere (Figure 4a), which produces a broad plateau over the entire retro-plate. Continuous convergence results in major lithosphere thickening. Some of the mantle lithospheric material is dragged into the asthenosphere by the subducting plate (Figure 4a), whereas lithospheric dripping developed at the far end of the weak retro-plate, close to the end-plate (Figures 4b and 4c), which eventually leads to a large-scale delamination of the lithospheric mantle (Figure 4d). The development of delamination in this case may be better understood by examining the density anomaly and effective viscosity fields (Figures 4c′ and 4c″). Positive density anomaly is resulted from the thickening of the cold lithospheric mantle of the retro-plate (Figure 4c′), due to temperature dependence of the effective density (equation (4)). In addition, the effective viscosity field indicates the weakness of the lithospheric mantle of retro-plate (Figure 4c″). Both conditions contribute to a major lithosphere delamination (Figure 4d). The density anomaly below the suture zone (Figure 4c′) does not lead to major delamination (Figure 4d), perhaps because of the relatively high effective viscosity there (Figure 4c″). 4.1.4. Model With a Weak Lower Crust and a Weak Lithospheric Mantle of the Retro-plate In the fourth model, the retro-plate has both a weak lower crust and a weak lithospheric mantle (Figure 1j). In this case, lithosphere delamination occurs in the retro-plate but faster than in the previous model. Small lithospheric blobs start to delaminate in the collision zone a few mega annum after collision started

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 9 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 4. Results of the model with weak lithospheric mantle of the retro-plate (Figure 1i) and neutral reference density contrast between the lithospheric mantle and the asthenosphere (Figure 1d). (a)–(d) Composition and temperature evolution. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel. The (c′) density anomaly ﬁeld and (c″) effective viscosity ﬁeld with velocity vectors are shown for the time step of 31.0 Ma, with delamination initiation.

(Figure 5a). The broad crustal thickening of the retro-plate produces broad high topography over the entire retro-plate. Lithospheric thickening eventually leads to major lithosphere delamination in the far end of the retro-plate (Figure 5b). Further convergence results in another round of thickening and delamination (Figures 5c and 5d). Finally, the lithospheric mantle of the retro-plate is almost completely delaminated, forming a wide and relatively ﬂat plateau with thickened continental crust of 60–70 km and a high topography (Figure 5e). The end-plate is underthrusted beneath the plateau in this case (Figure 5e).

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 10 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 5. Results of the model with weak lower crust and weak lithospheric mantle of the retro-plate (Figure 1j), as well as neutral reference density contrast between the lithospheric mantle and the asthenosphere (Figure 1d). (a)–(e) Composition and temperature evolution. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel.

Besides the positive density anomaly and low effective viscosity of the thickened lithospheric mantle, delamination is facilitated by the weak lower crust of the retro-plate (Figure 6), which allows the delaminating lithospheric mantle to be decoupled from the crust. This is one of the most favorable conditions for the delamination of lithospheric mantle as originally proposed by Bird [1979]. It is worth noting that the plate deformation and delamination in our models are mainly induced by lithosphere thickening from plate convergence, not by the underlying asthenospheric mantle flow. Moreover, lithosphere delamination in our model drives significant asthenospheric mantle flow (Figure 6), which, in turn, may further contribute to the delamination. The main factors for large-scale delamination are density increase in the thickened lithosphere and rheological weakness of the lithosphere (Figure 6).

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 11 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 6. (a–e) Effective viscosity ﬁeld with velocity vectors are shown for the same model as in Figure 5. (c′) and (d′)Density anomaly ﬁelds for two time steps 17.6 Ma (Figure 6c) and 18.1 Ma (Figure 6d), respectively. Time (Ma) of shortening is given in each panel.

4.2. Negative Reference Density Contrast Between Lithospheric Mantle and Asthenosphere ρLith ρAsth 3 ( 0 0 = 30 kg/m ) In the model with negative reference density contrast between the lithospheric mantle and the asthenosphere (Figure 1e), no major delamination is observed in the experiments with variable rheological structures (Figure 7). When the retro-plate is strong, the convergence is mainly accommodated by underthrusting of the pro-plate, with its upper-middle crust scraped off (Figures 7a and 7b). In contrast, a relatively weak retro-plate results in pure-shear thickening of the whole retro-plate lithosphere, with signiﬁcant high topography (Figures 7c and 7d). However, delamination is still prevented by the negative reference density contrast between the lithospheric mantle and the asthenosphere.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 12 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

ρLith ρAsth 3 Figure 7. Numerical models with negative reference density contrast ( 0 0 = 30 kg/m ) between the lithospheric mantle and the asthenosphere as shown in Figure 1e. (a)–(d) Models with rheological proﬁles as in Figures 1g–1j, respectively, where LC represents lower crust and LM for lithospheric mantle. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel.

ρLith ρAsth 4.3. Positive Reference Density Contrast Between Lithospheric Mantle and Asthenosphere ( 0 0 = 30 kg/m3) 4.3.1. Model of Homogeneous Layering In this model, the pro-plate subducts continuously into the asthenosphere, with its upper-middle crust scraped off and accreted to the overriding plate (Figure 8a). No obvious lithosphere delamination is predicted. The results are generally similar to the reference model with neutral reference density of the lithospheric mantle and the asthenosphere (Figure 2), except for the absence of sublithospheric crustal plume in this model (Figure 8a). 4.3.2. Model With a Weak Lower Crust of the Retro-plate In this model, the plate convergence is still dominated by subduction of the pro-plate, with its upper-middle crust scraped off and accumulated in the collision zone (Figure 8b). A piece of lithospheric mantle is detached from the bottom of retro-plate due to the erosion of sublithospheric crustal plume; however, the weak lower crust alone is insufﬁcient to induce major delamination. 4.3.3. Model With a Weak Lithospheric Mantle of the Retro-plate In this case, lithosphere initially thickens in the collision zone, and small blocks of the thickened lithosphere delaminate (Figure 9a). Further convergence thickens the lithosphere across the retro-plate, especially in the

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 13 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 8. Results of the model (a) with laterally homogeneous layering as in Figure 1g or (b) with a weak retro-plate lower crust as in Figure 1h. Positive reference density contrast between the lithospheric mantle and the asthenosphere is applied, ρLith ρAsth 3 0 0 = 30 kg/m , as in Figure 1f. Colors of rock types are as in Figure 1c. Time (Ma) of shortening for either model is 21.0 Ma as shown in the panels. The black triangles indicate the crustal suture zone between the pro-plate and retro-plate.

far-end region close to the end-plate, resulting in drippings at the bottom of the lithospheric mantle (Figures 9b and 9c). The broad crustal thickening of the retro-plate produces broad high topography over the entire retro-plate. Lithosphere thickening eventually leads to a large-scale delamination of the lithospheric mantle, starting from the far end of the retro-plate and migrating toward the suture zone (Figures 9d and 9e), eventually removing the entire lithospheric mantle of the retro-plate and forming a wide and relatively ﬂat plateau with thickened continental crust of 60–70 km (Figure 9f). The large-scale delamination caused rapid uplift of the topography (cf. Figures 9c–9e). At the end of lithosphere delamination, the elevation of the plateau can reach 6–7 km. 4.3.4. Model With a Weak Lower Crust and a Weak Lithospheric Mantle of the Retro-plate Not surprisingly, adding a weak lower crust to the previous case further enhances delamination. Small lithospheric blobs start to delaminate a few mega annum after the continental collision started (Figures 10a and 10b). Further convergence thickens the whole lithosphere of the retro-plate, especially the far-end region close to the end-plate (Figures 10c and 10d), which is similar to the previous model (Figures 9c and 9d). In about 15 Ma the entire mantle lithosphere of the retro-plate has delaminated (Figures 10e and 10f). A wide and ﬂat plateau is formed, with altitude of about 5–6 km (Figure 10f).

5. Effects of Other Factors Other factors may affect the delamination dynamics in collisional orogens. Here we discuss the effects of possible eclogitization of the lower crust and the convergence rates.

5.1. Lower Crust Eclogitization Eclogitization, a phase transformation of mafic crustal materials to denser eclogite, has been suggested as a major cause of delamination [e.g., Leech, 2001; Krystopowicz and Currie, 2013]. Lower crust eclogitization can increase density by 10% [e.g., Austrheim, 1987; Jolivet et al., 2005] or 300–600 kg/m3 [e.g., Doin and Henry, 2001]. To illustrate its first-order impact, we simplified eclogitization in the model by increasing the lower crustal density by 300 kg/m3 once the pressure is greater than 1.5 GPa (about 50 km in depth). ρLith ρAsth 3 First, we applied eclogitization to the models with negative reference density contrast ( 0 0 = 30 kg/m ) between the lithospheric mantle and the asthenosphere, which predict no major delamination without eclogitization (Figure 7). Having lower crust eclogitization significantly changes the model evolution. When the retro- plate lithospheric mantle is strong, steep subduction of pro-plate lithosphere occurs (Figures 11a and 11b);

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 14 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 9. Results of the model with weak lithospheric mantle of the retro-plate as in Figure 1i. Positive reference density ρLith ρAsth 3 – contrast between the lithospheric mantle and the asthenosphere is applied, 0 0 = 30 kg/m ,asinFigure1f.(a) (f) Composition and temperature evolution. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel.

whereas ﬂat subduction (underthrusting) is predicted in the models without eclogitization (Figures 7a and 7b). With a weak retro-plate lithospheric mantle and strong lower crust, retro-plate thickening dominates but no delamination occurs (Figure 11c), similar to the comparable cases but without eclogitization (Figure 7c). When both the lower crust and mantle lithosphere in the retro-plate are weak, eclogitization is shown to lead

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 15 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 10. Results of the model with both weak lower crust and weak lithospheric mantle of the retro-plate as in Figure 1j. ρLith ρAsth 3 Positive reference density contrast between the lithospheric mantle and the asthenosphere is applied, 0 0 = 30 kg/m , as in Figure 1f. (a)–(f) Composition and temperature evolution. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 16 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

ρLith ρAsth Figure 11. Effects of lower crust eclogitization for the models with negative reference density contrast ( 0 0 = 30 kg/m3). The density of the lower crust is increased by 300 kg/m3 when the pressure is greater than 1.5 GPa, i.e., eclogitization occurs. (a–d) Models with rheological proﬁles as in Figures 1g–1j, respectively, where LC represents lower crust and LM for lithospheric mantle. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel.

to major delamination in the retro-plate near the collision zone (Figure 11d). This could not occur without eclogitization (Figure 7d), because of the lower reference density of the mantle lithosphere. ρLith ¼ ρAsth Eclogitization is also applied to the reference models with neutral reference density contrast ( 0 0 ) between the lithospheric mantle and the asthenosphere. Delamination occurs mainly in the pro-plate when the retro-plate is not very weak (Figures 12a–12c). In these cases, eclogitization promotes the sinking of the pro-plate. Delamination, on the other hand, would occur mainly in the retro-plate when it is weakest (Figure 12d). The retro-plate delamination is similar to the comparable model without eclogitization in Figure 5. In both models, major delamination initiates in the far end of the retro-plate. However, the predicted delamination initiates earlier and moves faster with eclogitization. ρLith ρAsth 3 When eclogitization is applied to the models with positive density contrast ( 0 0 = 30 kg/m ), pro-plate delamination is similar to the previous models (Figure 12) but migrates faster, when the retro-plate is not very weak (Figures 13a–13c). In contrast, delamination occurs in retro-plate and migrates from the suture zone to the far end when the retro-plate is weakest (Figure 13d). The migration of retro-plate delamination differs signiﬁcantly from the comparable model without eclogitization (Figure 10), in which opposite migration

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 17 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

ρLith ρAsth 3 Figure 12. Effects of lower crust eclogitization for the models with neutral reference density contrast ( 0 0 =0kg/m ). The density of lower crust is increased by 300 kg/m3 when the pressure is greater than 1.5 GPa, i.e., eclogitization occurs. (a–d) Models with rheological proﬁles as in Figures 1g–1j, respectively, where LC represents lower crust and LM for lithospheric mantle. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel.

direction of retro-plate delamination is observed. These results indicate that lower crust eclogitization could play a major role in initiating delamination in the collision zone, where lithosphere thickening starts.

5.2. Convergence Rate The convergence rate plays a signiﬁcant role in the dynamics of continental collision and mountain building [e.g., Faccenda et al., 2008; Li et al., 2011, 2013, 2015]; it may also inﬂuence delamination. In the models

presented so far, a constant convergence rate of Vx = 5 cm/yr is applied. We conducted additional models with lower (Vx = 2 cm/yr) and higher (Vx = 8 cm/yr) convergence rates (Figures S2–S7), in order to show how variations of Vx affect the model results. ρLith ρAsth When the reference density of the mantle lithosphere is lower than that of the asthenosphere ( 0 0 = 30 kg/m3), no delamination is predicted regardless of the convergence rates (Figures S2 and S3). The main

difference is that lower convergence rate of Vx = 2 cm/yr leads to steep subduction in the models with relatively strong retro-plate (Figure S2), whereas ﬂat subduction, or underthrusting, is predicted with higher

(Vx = 5 cm/yr, Figure 7, and Vx = 8 cm/yr, Figure S3) convergence rates.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 18 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

ρLith ρAsth 3 Figure 13. Effects of lower crust eclogitization for the models with positive reference density contrast ( 0 0 = 30 kg/m ). The density of the lower crust is increased by 300 kg/m3 when the pressure is greater than 1.5 GPa, i.e., eclogitization occurs. (a–d) Models with rheological proﬁles as in Figures 1g–1j, respectively, where LC represents lower crust and LM for lithospheric mantle. Colors of rock types are as in Figure 1c. Time (Ma) of shortening is given in each panel.

ρLith ¼ ρAsth When the reference density is the same for the lithospheric mantle and the asthenosphere ( 0 0 ), lower convergence rate (Vx = 2 cm/yr) results in pro-plate delamination, when the retro-plate is not very weak (Figures S4a–S4c). It indicates that the prescribed convergence rate is lower than the local sinking velocity driven by the negative buoyancy of the subducting slab with the crustal materials detached. Double-plates delamination is predicted when the retro-plate is weakest (Figure S4d), in which pro-plate delamination occurs following a major retro-plate delamination near the suture zone. The faster convergence rate

(Vx = 8 cm/yr) does not change the general delamination mode in comparison with the reference models with Vx = 5 cm/yr, although details of the collisional styles are different (cf. Figures S5 and 2–5). ρLith ρAsth 3 In the models of positive density contrast ( 0 0 = 30 kg/m ), a lower convergence rate of Vx = 2 cm/yr causes different styles of delamination from the reference models (Figure S6). If the retro-plate lithospheric mantle is relatively strong, delamination occurs in the pro-plate (Figures S6a and S6b). When the retro-plate has a weak lithospheric mantle but a strong lower crust, slow convergence leads to delamination in both the pro-plate and retro-plate (Figure S6c). In contrast, delamination occurs dominantly in the retro-plate when

both its lower crust and lithospheric mantle are weak (Figure S6d). With faster convergence (Vx = 8 cm/yr),

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 19 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

the pro-plate subduction occurs if the lithospheric mantle of the retro-plate is relatively strong (Figures S7a and S7b). In contrast, delamination occurs dominantly in the retro-plate when its lithospheric mantle is weak

(Figures S7c and S7d). In these cases, the results are similar to those of the reference models with Vx = 5 cm/yr (Figures 9 and 10).

6. Discussion 6.1. Delamination Mode Selection Our numerical simulations show three basic modes of lithosphere delamination in collisional orogens: pro- plate delamination, retro-plate delamination, and the transitional mode of double-plates delamination (both pro-plate and retro-plate). Based on the numerical results, we constructed a 3-D regime diagram (Figures 14a–14c) to illustrate how the delamination mode depends on the convergence rate, retro-plate rheological properties, and the density contrast between lithospheric mantle and asthenosphere. Another regime diagram is constructed in order to show the effects of lower crust eclogitization on the delamination mode selection (Figure 14d). The regime diagram shows the absence of delamination in models when the reference density of the lithospheric mantle is lower than that of the asthenosphere (Figure 14a). The buoyant-converging lithospheres prevent any types of delamination, except when the lower crust eclogitization is considered (Figure 14d, top). The density increase due to eclogitization may result in lithosphere delamination in the rather weak retro-plate. With neutral or positive density contrast between the lithospheric mantle and the asthenosphere, delamination occurs in different modes under various conditions (Figures 14b and 14c). The prerequisite for delamination of the retro-plate is its weakness (Figure 14); otherwise, it would prevent major deformation and delamination of the retro-plate. The promoting factors of retro-plate delamination include faster conver- ρLith ρAsth gence and a higher value of 0 0 . Pro-plate delamination is favored by lower convergence rate, relatively strong retro-plate, and the lower crust eclogitization (Figure 14). In these cases, plate convergence is accommodated mainly by sinking of the pro- plate lithospheric mantle with (part of) the lower crust. The local velocity of slab sinking is larger than

horizontal plate convergence rate (Vx), leading to the retreat of the sinking slab that detaches from the upper crust. Delamination of both the pro-plate and retro-plate could occur simultaneously, but only in the models with low convergence rates and within a narrow range of parameters (Figures 14b and 14c). This mode of delamination can be regarded as a transition between the pro-plate and retro-plate delamination modes.

6.2. Comparison With Previous Models The dynamics of lithosphere delamination has been studied in Morency and Doin [2004] based on a mantle convection model with a stagnant lid. They found that lithosphere delamination initiates systematically when the lowermost crust and the top lithospheric mantle are sufficiently soft. They further constructed a 2-D regime diagram with the two axes of crust viscosity and mantle viscosity, which was separated into two domains: one with lower viscosity values, where delamination initiates; another one with higher viscosity values, where no delamination occurs. In addition, Schott and Schmeling [1998] proposed that delamination and detachment of the mantle lithosphere are only possible for a thickened lithosphere, which has been significantly weakened, e.g., by water. Our models also indicate that the weakness of the retro-plate is a prerequisite for its delamination during continental collision. With respect to the influence of rheological strength on lithosphere delamination, our results are thus consistent with previous numerical experiments, although our plate tectonics-like numerical setup is significantly different. Our systematic numerical experiments predict three basic modes of delamination: pro-plate, retro-plate, and double-plates delamination. Each of them has been reported in previous models, but separately. The pro- plate delamination has been simulated in models with a sinking and retreating subducting lithosphere peeling off from its overlying crust [e.g., Faccenda et al., 2009; Ueda et al., 2012]. These numerical models assume a strong retro-plate, so lithosphere delamination occurs only in the pro-plate. In addition, the prescribed convergence in Ueda et al. [2012] is turned off after a certain time of plate subduction. The resulted

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 20 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 14. Delamination regime diagram. (a)–(c) Dependence of delamination modes on the density contrast between the lithospheric mantle and the asthenosphere, and between the retro-plate rheology and the convergence rate. For the rheological properties, “reference,”“weak LC,”“weak LM,” and “weak LC + LM” represent rheological proﬁles as in Figures 1g–1j, respectively, where LC represents lower crust and LM for lithospheric mantle. The 3-D regime diagram is visualized in 2-D for clarity. (d) Comparisons of delamination mode selection for the models with or without lower crust eclogitization, in which constant convergence rate is applied.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 21 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

slow convergence by solely slab pull contributed to the retreating of the pro-plate. These results are generally consistent with our predictions with low convergence rate and relatively strong retro-plate as favorable conditions for pro-plate delamination (Figure 14). Retro-plate delamination has been simulated by Krystopowicz and Currie [2013] (“K-C models” henceforth). In their models, one weak plate is located between two strong plates. In the K-C models crustal eclogitization is critical to lithospheric delamination. In addition, a weak lower crust is generally required for the full detachment of the sinking mantle lithosphere. In our models, a weak retro-plate is necessary for delamination of its lithospheric mantle, and a weak lower crust promotes delamination. These results agree with the K-C models. However, in our models the effects of lower crust eclogitization depend on the density contrast between the lithospheric mantle and the asthenosphere, which play significant roles in the delamination processes. When the reference density of the continental lithospheric mantle is the same or greater than that of the asthenosphere, lithosphere delamination can occur during the retro-plate shortening without eclogitization (Figures 14b and 14c). Otherwise, lower crust eclogitization is needed for delamination (Figures 14a and 14d). The double-plates delamination predicted in our model has also been observed in the numerical models of Gray and Pysklywec [2012]. Their models used a higher reference density contrast (50 kg/m3) between the lithospheric mantle and the asthenosphere, as well as extremely high radioactive heating of a wide retro- plate margin that may decrease its rheological strength to initiate the retro-plate delamination. Finally, a mode with delamination on both the pro-plate and retro-plate is observed. Their results are consistent with our models in terms of needing a relatively large density contrast and weak retro-plate (Figure 14). Houseman and Molnar [1997] drew a useful distinction between the concept of delamination as proposed by Bird [1979] and that of convective thinning. In delamination, the locus of downwelling migrates back- ward as the delaminating plate sinks downward. In convective thinning, the loci of downwelling are stationary as the denser layer flows toward them and then vertically downward. In our experiments, however, it is rather difficult to clearly distinguish the convective thinning mode from the delamination mode. In several experiments (e.g., Figures 9 and 10), the detachment and sinking of lithospheric mantle initially occur in a stationary locus (i.e., similar to the convective removal), but then the downwelling may migrate to its sides (i.e, similar to the delamination mode). Therefore, we prefer using the term “delamination” in this study to describe a general (rather complex) process of lithosphere removal and discriminate among different “delamination modes.”

6.3. Delamination-Induced Topographic Uplift Topographic uplift is a direct and most signiﬁcant response to continental collision. Our numerical simulations indicate that a relatively narrow mountain belt, i.e., around 500 km in width, is resulted during continental collision without delamination (Figures 2 and 3). The averaged elevation is around 4 km, with the peak of up to 8 km (e.g., Figure 3). In contrast, delamination of the mantle lithosphere tends to form a broad plateau (e.g., Figures 4 and 5, as well as other similar cases).

6.4. Geological Implications The numerical models presented here are necessarily simpliﬁed, yet the results may provide some useful insight into mountain building in many continental collision zones that may have involved delamination. The pro-plate delamination mode implies that the mantle lithosphere of the subducting continental plate could delaminate from the overlying crust and sinks (e.g., Figures 12a–12c and 13a–13c). A possible geological analog is the Northern Apennines (Figure 15a), which is a fold-and-thrust belt that formed in response to the eastward retreating of the subducted Adria microcontinental lithosphere delaminating from the overlying continental crust [Faccenna et al., 2001, 2009; Chiarabba and Chiodini, 2013; Chiarabba et al., 2014]. Geological evidence indicates that the Northern Apennines underwent two phases of eastward migrating deformation that caused early compressional structures being dissected by later extensional deformation since the late Oligocene-Miocene [Elter et al., 1975; Faccenda et al., 2009]. Extension, due to pro-plate lithosphere delamination, produced progressive younging toward the east syntectonic extensional basins and associated magmatic activity. The Himalayan-Tibetan system may provide an example of retro-plate delamination (Figure 15b) [e.g., England, 1993; Owens and Zandt, 1997]. The collision between the Indian and Eurasian plates at about

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 22 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Figure 15. (a) Simplified tectonic profile of the Northern Apennines (modified from Barchi et al. [1998], Mele and Sandvol [2003], and Carminati et al. [2004]). (b) Characteristic profile of the Tibetan plateau from SW to NE (modified from Owens and Zandt [1997]).

5 cm/yr in the past 50 Ma has nearly doubled the thickness of the Tibetan crust, and should have done the same for mantle lithosphere. Although some surface wave inversions indicate a thick lithosphere beneath the entire Tibetan plateau [McKenzie and Priestley, 2008; Priestley and McKenzie, 2013], most geophysical observations suggest relatively thin, or even the absence of, lithospheric mantle in much of the central and northern Tibetan plateau [McNamara et al., 1995; Owens and Zandt, 1997; Tilmann et al., 2003]. This thin lithosphere is attributed to delamination or convective thinning [Molnar et al., 1993]. Our model results suggest that retro-plate delamination mode (e.g., Figures 4 and 5) is favored when the mantle lithosphere, perhaps the crust as well, is rheologically weak. This is reasonable, because the Tibetan lithosphere had experienced a series of (paleo)-Tethyan subduction and the collision of multiple continental terranes before the final collision with the Indian Plate [Yin and Harrison, 2000]. Thus, the Tibetan lithosphere may be weakened by significant hydration and partial melting. For the Himalayan-Tibetan system, the 3-D process with variation in the third dimension and significant lateral extrusion may influence the extent of deformation in the retro-plate [England et al., 1985; Tapponnier et al., 2001; Royden et al., 2008; Li et al., 2013; Capitanio et al., 2015]. It may further affect the lithosphere delamination in a certain degree, which is currently unclear. As a first inference, the lateral extrusion of retro-plate may postpone its delamination, because more convergence is needed to thicken the lithosphere and build enough negative buoyancy. On the other hand, the lateral extrusion, along large-scale strike-slip faults, may decrease the strain-dependent rheological strength of the faults and the retro-plate, which will facilitate the delamination. The effects of 3-D process, e.g., lateral extrusion, on lithosphere delamination in collisional orogens need further investigation with 3-D models. It is worth emphasizing that the goal of this work is not to reproduce the tectonic evolution of specific orogens with possible delamination but to illustrate the basic physics behind the delamination and orogenesis.

7. Conclusions We have used high-resolution thermomechanical numerical models to systematically investigate the dynamics of lithosphere delamination during continental collision. The main conclusions we may draw from this study include the following: 1. Lithosphere thickening from continental collision can sufﬁciently increase the negative buoyancy to drive lithosphere delamination, when the reference density of the continental lithospheric mantle is the same or greater than that of the asthenosphere. However, if the reference density of the lithospheric mantle is

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 23 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

less than that of the asthenosphere, additional promoting factors, e.g., lower crust eclogitization, are required for delamination. 2. Lithosphere delamination can occur in pro-plate, retro-plate, or both during continental collision, depend- ing on the density anomaly, the convergence rate, as well as the relative rheological strength of the pro- plate and retro-plate. 3. The pro-plate delamination is favored by lower convergence rate, relatively strong retro-plate, and the lower crust eclogitization. The Northern Apennines may provide as a geological analog for the pro-plate delamination mode. 4. Retro-plate delamination requires weak rheology of the retro-plate, and is favored by faster convergence and larger density contrast between the retro-plate lithospheric mantle and the asthenosphere. This process may have contributed to the interpreted thin lithosphere under Central Northern Tibetan Plateau. 5. Lithosphere delamination during continental collision generally results in a wide and ﬂat mountain belt or plateau. In collisional orogens without major delamination, relatively narrow and steep mountain belts are predicted.

Acknowledgments References This work was supported by the 973 fl – Project of China (2015CB856106), the Austrheim, H. (1987), Eclogitization of lower crustal granulites by uid migration through shear zones, Earth Planet. Sci. Lett., 81, 221 232. Major Research Program of NSFC Bajolet, F., J. Galeano, F. Funiciello, M. Moroni, A. M. Nagredo, and C. Faccenna (2012), Continental delamination: Insights from laboratory (41590860), the Strategic Priority models, Geochem. Geophys. Geosyst., 13, Q02009, doi:10.1029/2011GC003896. Research Program (B) of CAS Bao, X. W., D. Eaton, and B. Guest (2015), Plateau uplift in western Canada caused by lithosphere delamination along a craton edge, Nat. – (XDB18020104), the NSFC project for Geosci., 7, 830 833. fi young scholars (41304071), and the Barchi, M., G. Minelli, and G. Pialli (1998), The Crop 03 pro le: A synthesis of results on deep structures of the Northern Apennines, Mem. Soc. – National “Qian-Ren” Program to Z. H. Li. Geol. Ital., 52, 383 400. fl M. Liu acknowledges support from the Beaumont, C., M. H. Nguyen, R. A. Jamieson, and S. Ellis (2006), Crustal ow modes in large hot orogens, in Channel Flow, Ductile Extrusion, and University of Chinese Academy of Exhumation of Lower-Mid Crust in Continental Collision Zones, edited by R. D. Law, M. P. Searle, and L. Godin, Geo. Soc. London Spec. Publ, – Sciences. Numerical simulations were 268,91 145. run with the clusters of National Beck, S. L., and G. Zandt (2002), The nature of orogenic crust in the central Andes, J. Geophys. Res., 107(B10, 2230), doi:10.1029/2000JB000124. – Supercomputer Center in Guangzhou Bird, P. (1978), Initiation of intracontinental subduction in the Himalaya, J. Geophys. Res., 83(B10), 4975 4987, doi:10.1029/JB083iB10p04975. – (Tianhe-II). The data for this paper Bird, P. (1979), Continental delamination and the Colorado Plateau, J. Geophys. Res., 84(B13), 7561 7571, doi:10.1029/JB084iB13p07561. are available by contacting the Bittner, D., and H. Schmeling (1995), Numerical modeling of melting processes and induced diapirism in the lower crust, Geophys. J. Int., 123, – corresponding author (li.zhonghai@u- 59 70. cas.ac.cn). Thorough and constructive Burg, J.-P., and T. V. Gerya (2005), The role of viscous heating in Barrovian metamorphism of collisional orogens: Thermomechanical models – reviews by two anonymous reviewers and application to the Lepontine Dome in the Central Alps, J. Metamorph. Geol., 23,75 95. “ ” “ ” and the Associate Editor improved Burov, E. B., and A. B. Watts (2006), The long-term strength of continental lithosphere: Jelly sandwich or creme brulee ?, GSA Today, 16, – the paper. 4 10. Capitanio, F. A., A. Replumaz, and N. Riel (2015), Reconciling subduction dynamics during Tethys closure with large-scale Asian tectonics: Insights from numerical modeling, Geochem. Geophys. Geosyst., 16, 962–982, doi:10.1002/2014GC005660. Carminati, E., C. Doglioni and D. Scrocca (2004), Alps and Apennines, Spec. Vol. Italian Geo. Soc. for IGC 32, 141–151. Chiarabba, C., and G. Chiodini (2013), Continental delamination and mantle dynamics drive topography, extension and fluid discharge in the Apennines, Geology, 41, 715–718. Chiarabba, C., G. Giacomuzzia, I. Bianchi, N. P. Agostinetti, and J. Park (2014), From underplating to delamination-retreat in the northern Apennines, Earth Planet. Sci. Lett., 403, 108–116. Clauser, C., and E. Huenges (1995), Thermal conductivity of rocks and minerals, in Rock Physics and Phase Relations, edited by T. J. Ahrens, pp. 105–126, AGU, Washington D. C., Reference Shelf 3. Cloos, M., B. Sappie, A. Q. van Ufford, R. J. Weiland, P. Q. Warren, and T. McMahon (2005), Collisional delamination in New-Guinea: The geotectonics of subducting slab breakoff, Geol. Soc. Am. Spec., 400, 51. Crameri, F., H. Schmeling, G. J. Golabek, T. Duretz, R. Orendt, S. Buiter, D. May, B. Kaus, T. Gerya, and P. Tackley (2012), A comparison of numerical surface topography calculations in geodynamic modelling: An evaluation of the “sticky air” method, Geophys. J. Int., 189,38–54. Currie, C. A., C. Beaumont, and R. S. Huismans (2007), The fate of subducted sediments: A case for backarc intrusion and underplating, Geology, 35(12), 1111–1114. Djomani, Y., S. O’Reilly, W. L. Griffin, and P. Morgan (2001), The density structure of subcontinental lithosphere through time, Earth Planet. Sci. Lett., 184, 605–621. Doin, M.-P., and P. Henry (2001), Subduction initiation and continental crust recycling: The roles of rheology and eclogitization, Tectonophysics, 342, 163–191. Ducea, M., and J. Saleeby (1998), A case for delamination of the deep batholithic crust beneath the Sierra Nevada, California, Int. Geol. Rev., 40(1), 78–93. Elter, P., G. Giglia, M. Tongiorgi, and L. Trevisan (1975), Tensional and compressional areas in the recent (Tortonian to present) evolution of the Northern Apennines, Boll. Geofis. Teor. Appl., 17,3–18. England, P. (1993), Convective removal of thermal boundary layer of thickened continental lithosphere: A brief summary of causes and consequences with special reference to the Cenozoic tectonics of the Tibetan Plateau and surrounding regions, Tectonophysics, 223, 67–73. England, P., and G. Houseman (1989), Extension during continental convergence, with application to the Tibetan Plateau, J. Geophys. Res., 94(B12), 17,561–17,579, doi:10.1029/JB094iB12p17561. England, P., G. Houseman, and L. Sonder (1985), Length scales for continental deformation in convergent, divergent, and strike-slip environments: Analytical and approximate solutions for a thin viscous sheet model, J. Geophys. Res., 90(B5), 3551–3557, doi:10.1029/ JB090iB05p03551.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 24 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Faccenda, M., T. V. Gerya, and S. Chakraborty (2008), Styles of post-subduction collisional orogeny: Influence of convergence velocity, crustal rheology and radiogenic heat production, Lithos, 103, 257–287. Faccenda, M., G. Minelli, and T. V. Gerya (2009), Coupled and decoupled regimes of continental collision: Numerical modeling, Earth Planet. Sci. Lett., 278, 337–349. Faccenna, C., T. Becker, F. Lucente, L. Jolivet, and F. Rossetti (2001), History of subduction and back-arc extension in the central Mediterranean, Geophys. J. Int., 145, 809–820. Fischer, R., and T. Gerya (2016), Early Earth plume-lid tectonics: A high-resolution 3D numerical modelling approach, J. Geodyn., doi:10.1016/ j.jog.2016.03.004. Francois, T., E. Burov, P. Agard, and B. Meyer (2014), Buildup of a dynamically supported orogenic plateau: Numerical modeling of the Zagros/Central Iran case study, Geochem. Geophys. Geosyst., 15, 2632–2654, doi:10.1002/2013GC005223. Gao, S., J. F. Zhang, W. L. Xu, and Y. S. Liu (2009), Delamination and destruction of the North China Craton, Chin. Sci. Bull., 54, 3367–3378. Gerya, T. V. (2010), Introduction to Numerical Geodynamic Modelling, Cambridge Univ. Press, Cambridge. Gerya, T. V., and F. I. Meilick (2011), Geodynamic regimes of subduction under an active margin: Effects of rheological weakening by fluids and melts, J. Metamorph. Geol., 29,7–31. Gerya, T. V., and D. A. Yuen (2003), Characteristics-based marker-in-cell method with conservative finite-differences schemes for modeling geological flows with strongly variable transport properties, Phys. Earth Planet. Inter., 140, 293–318. Gerya, T. V., and D. A. Yuen (2007), Robust characteristics method for modelling multiphase visco-elasto-plastic thermo-mechanical problems, Phys. Earth Planet. Inter., 163,83–105. Gerya, T. V., R. J. Stern, M. Baes, S. Sobolev, and S. A. Whattam (2015), Plate tectonics on the Earth triggered by plume-induced subduction initiation, Nature, 527, 221–225. Giacomuzzi, G., C. Chiarabba, and P. De Gori (2011), Linking the Alps and Apennines subduction systems: New constraints revealed by high-resolution teleseismic tomography, Earth Planet. Sci. Lett., 301, 531–543. Gogus, O. H., and R. N. Pysklywec (2008a), Mantle lithosphere delamination driving plateau uplift and synconvergent extension in eastern Anatolia, Geology, 36(9), 723–726. Gogus, O. H., and R. N. Pysklywec (2008b), Near-surface diagnostics of dripping or delaminating lithosphere, J. Geophys. Res., 113, B11404, doi:10.1029/2007JB005123. Gogus, O. H., R. N. Pysklywec, F. Corbi, and C. Faccenna (2011), The surface tectonics of mantle lithosphere delamination following ocean lithosphere subduction: Insights from physical-scaled analogue experiments, Geochem. Geophys. Geosyst., 12, Q05004, doi:10.1029/ 2010GC003430. Gorczyk, W., and K. Vogt (2013), Tectonics and melting in intra-continental settings, Gondwana Res., 27(1), 196–208. Gorczyk, W., B. Hobbs, and T. Gerya (2012), Initiation of Rayleigh-Taylor instabilities in intra-cratonic settings, Tectonophysics, 514, 146–155. Gray, R., and R. N. Pysklywec (2012), Geodynamic models of mature continental collision: Evolution of anorogen from lithospheric subduction to continental retreat/ delamination, J. Geophys. Res., 117, B03408, doi:10.1029/2011JB008692. Houseman, G., D. McKenzie, and P. Molnar (1981), Convective instability of a thickened boundary layer and its relevance for the thermal evolution of continental convergent belts, J. Geophys. Res., 86(B7), 6115–6132, doi:10.1029/JB086iB07p06115. Houseman, G. A., and P. Molnar (1997), Gravitational (Rayleigh-Taylor) instability of a layer with non-linear viscosity and convective thinning of continental lithosphere, Geophys. J. Int., 128, 125–150. Ji, S., and P. Zhao (1993), Flow laws of multiphase rocks calculated from experimental data on the constituent phases, Earth Planet. Sci. Lett., 117, 181–187. Jimenez-Munt, I., M. Fernandez, J. Verges, and J. P. Platt (2008), Lithosphere structure underneath the Tibetan Plateau inferred from elevation, gravity and geoid anomalies, Earth Planet. Sci. Lett., 267, 276–289. Johnson, T. E., M. Brown, B. Kaus, and J. A. Van Tongeren (2014), Delamination and recycling of Archaean crust caused by gravitational instabilities, Nat. Geosci., 7,47–52. Jolivet, L., H. Raimbourg, L. Labrousse, D. Avigad, Y. Leroy, H. Austrheim, and T. Andersen (2005), Softening triggered by eclogitization, the first step toward exhumation during continental subduction, Earth Planet. Sci. Lett., 237, 532–547. Jull, M., and P. B. Kelemen (2001), On the conditions for lower crustal convective instability, J. Geophys. Res., 106(B4), 6423–6446, doi:10.1029/ 2000JB900357. Kay, R. W., and S. M. Kay (1993), Delamination and delamination magmatism, Tectonophysics, 219, 177–189. Keskin, M. (2003), Magma generation by slab steepening and break off beneath a subduction accretion complex: An alternative model for collision-related volcanism in Eastern Anatolia, Turkey, Geophys. Res. Lett., 30(24), 8046, doi:10.1029/2003GL018019. Kirby, S. H. (1983), Rheology of the lithosphere, Rev. Geophys. Space Phys., 21, 1458–1487. Kirby, S. H., and A. K. Kronenberg (1987), Rheology of the lithosphere: Selected topics, Rev. Geophys., 25, 1219–1244, doi:10.1029/ RG025i006p01219. Kohlstedt, D. L., B. Evans, and S. J. Mackwell (1995), Strength of the lithosphere: Constraints imposed by laboratory experiments, J. Geophys. Res., 100(B9), 17,587–17,602, doi:10.1029/95JB01460. Krystopowicz, N., and C. Currie (2013), Crustal eclogitization and lithosphere delamination in orogens, Earth Planet. Sci. Lett., 361, 195–207. Lee, C. T. A., X. Cheng, and U. Horodyskyj (2006), The development and refinement of continental arcs by primary basaltic magmatism, garnet pyroxenite accumulation, basaltic recharge and delamination: Insights from the Sierra Nevada, California, Contrib. Mineral. Petrol., 151(2), 222–242. Lee, C. T. A., P. Luffi, and E. J. Chin (2011), Building and destroying continental mantle, Annu. Rev. Earth Planet. Sci., 39,59–90. Leech, M. L. (2001), Arrested orogenic development: Eclogitization, delamination, and tectonic collapse, Earth Planet. Sci. Lett., 185, 149–159. Levander, A., B. Schmandt, M. S. Miller, K. Liu, K. E. Karlstrom, R. S. Crow, C. Lee, and E. Humphreys (2011), Continuing Colorado plateau uplift by delamination style convective lithospheric downwelling, Nature, 472, 461–465. Li, Z. H. (2014), A review on the numerical geodynamic modeling of continental subduction, collision and exhumation, Sci. China Earth Sci., 57,47–69. Li, Z. H., and T. V. Gerya (2009), Polyphase formation and exhumation of high- to ultrahigh-pressure rocks in continental subduction zone: Numerical modeling and application to the Sulu ultrahigh-pressure terrane in eastern China, J. Geophys. Res., 114, B09406, doi:10.1029/ 2008JB005935. Li, Z. H., Z. Q. Xu, and T. V. Gerya (2011), Flat versus steep subduction: Contrasting modes for the formation and exhumation of high- to ultrahigh-pressure rocks in continental collision zones, Earth Planet. Sci. Lett., 301,65–77. Li, Z. H., Z. Q. Xu, T. V. Gerya, and J. P. Burg (2013), Collision of continental corner from 3-D numerical modeling, Earth Planet. Sci. Lett., 380, 98–111.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 25 Journal of Geophysical Research: Solid Earth 10.1002/2016JB013106

Li, Z. H., M. Q. Liu, and T. Gerya (2015), Material transportation and fluid-melt activity in the subduction channel: Numerical modeling, Sci. China Earth Sci., 58, 1251–1268. Liu, M., X. Cui, and F. Liu (2004), Cenozoic rifting and volcanism in eastern China: A mantle dynamic link to the Indo-Asian collision?, Tectonophysics, 393,29–42. McKenzie, D., and K. Priestley (2008), The influence of lithospheric thickness variations on continental evolution, Lithos, 102,1–11. McNamara, D., T. Owens, and W. Walter (1995), Observations of regional phase propagation across the Tibetan Plateau, J. Geophys. Res., 100(B11), 22,215–22,229, doi:10.1029/95JB01863. Mele, G., and E. Sandvol (2003), Deep crustal roots beneath the northern Apennines inferred from teleseismic receiver functions, Earth Planet. Sci. Lett., 211,69–78. Molnar, P., P. England, and J. Martinod (1993), Mantle dynamics, uplift of the Tibetan plateau, and the Indian Monsoon, Rev. Geophys., 31, 357–396, doi:10.1029/93RG02030. Morency, C., and M. P. Doin (2004), Numerical simulations of the mantle lithosphere delamination, J. Geophys. Res., 109, B03410, doi:10.1029/ 2003JB002414. Owens, T. J., and G. Zandt (1997), Implications of crustal property variations for models of Tibetan plateau evolution, Nature, 387,37–43. Priestley, K., and D. McKenzie (2013), The relationship between shear wave velocity, temperature, attenuation and viscosity in the shallow part of the mantle, Earth Planet. Sci. Lett., 381,78–91. Ranalli, G. (1995), Rheology of the Earth, Deformation and Flow Processes in Geophysics and Geodynamics, 2nd ed., Chapman and Hall, London. Ranalli, G., and D. C. Murphy (1987), Rheology stratification of the lithosphere, Tectonophysics, 132, 281–295. Royden, L. H., B. C. Burchfiel, and R. D. van der Hilst (2008), The geological evolution of the Tibetan Plateau, Science, 321, 1054–1058. Schmeling, H., et al. (2008), A benchmark comparison of spontaneous subduction models—Towards a free surface, Phys. Earth Planet. Inter., 171, 198–223. Schott, B., and H. Schmeling (1998), Delamination and detachment of a lithospheric root, Tectonophysics, 296, 225–247. Schutt, D. L., and C. E. Lesher (2006), Effects of melt depletion on the density and seismic velocity of garnet and spinel lherzolite, J. Geophys. Res., 111, B05401, doi:10.1029/2003JB002950. Seber, D., M. Barazangi, A. Ibenbrahim, and A. Demnati (1996), Geophysical evidence for lithosphere delamination beneath the Alboran Sea and Rif-Betic mountains, Nature, 379, 785–790. Sizova, E., T. Gerya, K. Stuewe, and M. Brown (2015), Generation of felsic crust in the Archean: A geodynamic modeling perspective, Precambrian Res., 271, 198–224. Sobolev, S., A. Sobolev, D. Kuzmin, N. Krivolutskaya, A. Petrunin, N. Arndt, V. Radko, and Y. Vasiliev (2011), Linking mantle plumes, large igneous provinces and environmental catastrophes, Nature, 477, 312–316. Tapponnier, P., Z. Xu, F. Roger, B. Meyer, N. Arnaud, G. Wittlinger, and J. Yang (2001), Oblique stepwise rise and growth of the Tibet Plateau, Science, 294, 1671–1677. Tilmann, F., J. Ni, and INDEPTH III Seismic Team (2003), Seismic imaging of the downwelling Indian lithosphere beneath central Tibet, Science, 300, 1424–1427. Timoulalia, Y., H. Djellit, Y. Hahou, N. Jabour, and R. Merrouch (2014), New evidence of delamination in the Western Alboran Sea: Geodynamic evolution of the Alboran domain and its margins, J. Geodyn., 77, 206–216. Turcotte, D. L., and G. Schubert (1982), Geodynamics: Application of Continuum Physics to Geological Problems, pp. 450, John Wiley, New York. Turcotte, D. L., and G. Schubert (2002), Geodynamics, 2nd ed., pp. 456, Cambridge Univ. Press, New York. Ueda, K., T. V. Gerya, and J. P. Burg (2012), Delamination in collisional orogens: Thermomechanical modeling, J. Geophys. Res., 117, B08202, doi:10.1029/2012JB009144. Valera, J. L., A. M. Negredo, and A. Villasenor (2008), Asymmetric delamination and convective removal numerical modeling: Comparison with evolutionary models for the Alboran Sea Region, Pure Appl. Geophys., 165(8), 1683–1706. Valera, J. L., A. M. Negredo, and I. Jimenez-Munt (2011), Deep and near-surface consequences of root removal by asymmetric continental delamination, Tectonophysics, 502(1–2), 257–265. Vogt, K., T. Gerya, and A. Castro (2012), Crustal growth at active continental margins: Numerical modeling, Phys. Earth Planet. Inter., 192–193, 1–20. Wang, H., and C. Currie (2015), Magmatic expressions of continental lithosphere removal, J. Geophys. Res. Solid Earth, 120, 7239–7260, doi:10.1002/2015JB012112. Yin, A., and T. M. Harrison (2000), Geologic evolution of the Himalayan-Tibetan orogen, Annu. Rev. Earth Planet. Sci., 28, 211–280. Zandt, G., H. Gilbert, T. J. Owens, M. Ducea, J. Saleeby, and C. Jones (2004), Active foundering of a continental arc root beneath the southern Sierra Nevada in California, Nature, 431,41–46. Zhu, R. X., Y. G. Xu, G. Zhu, H. F. Zhang, Q. K. Xia, and T. Y. Zheng (2012), Destruction of the North China Craton, Sci. China Earth Sci., 55, 1565–1587.

LI ET AL. DYNAMICS OF LITHOSPHERE DELAMINATION 26