The Himalaya in 3D: Slab dynamics controlled mountain building and monsoon intensification Open Access

What do we know, and what remains to be discovered, about Himalayan geology? The primary answers are clear: the growth of Earth’s highest mountains results from the ongoing collision between the Indian and Eurasian continents. Continuing exploration across a range of geologic length and time scales is motivated by many questions, prominently including the following. (1) What are the initial and boundary conditions, key physical parameters, and idiosyncratic versus exportable characteristics of mountain-building from this leading natural laboratory for collisional tectonics and continental subduction? (2) In what ways are Himalayan lithospheric processes interacting with atmospheric, biotic, surface, and oceanic processes? (3) How can we understand and mitigate hazards (e.g., earthquakes, landslides, floods) across these mountains spanning many populous nations?

In this contribution we first review models for Himalayan tectonics across million year time scales. By exploring key data and interpretations, we highlight the need for three-dimensional evolutionary models. We then offer an example of such a model: along-strike changes in Himalayan mountain-building could have resulted from the along-strike migration of mid-Cenozoic slab detachment (i.e., slab breakoff). According to this hypothesis, slab evolution and resulting orogenic wedge changes are further speculated to have increased the elevation of the Himalaya and modified the force balance at the plate boundary, in turn yielding (1) increased South Asian monsoon strength through topographic growth, (2) decreased rates of India-Asia convergence by changing the forcing applied to the collisional boundary, and (3) Himalayan asymmetric arc curvature.

Long before the plate tectonic revolution, Argand (1924) presciently described the Himalayan mountains as consequences of the Indian lithosphere underthrusting beneath Eurasia (Fig. 1A). Later plate tectonic models maintained this basic framework (Fig. 1B) (Dewey and Bird, 1970; Powell and Conaghan, 1973).

In the 1980s, new models were proposed in response to the discovery of the South Tibet fault by Caby et al. (1983) and Burg et al. (1984). The South Tibet fault (also called the South Tibet detachment and the South Tibet fault system) was first recognized as a north-dipping, top-to-the-north shear zone, and/or a series of closely spaced top-to-the-north faults. This shear zone extends along the crest of the Himalayan mountains and separates the high-grade crystalline orogenic core to the south from a fold-thrust belt dominated by Paleozoic–Mesozoic Tethyan passive margin strata of northern India to the north (Fig. 2) (e.g., Burchfiel et al., 1992; Burg et al., 1984; Caby et al., 1983; Herren, 1987). The main characteristics of these South Tibet fault exposures, i.e., northerly dip, top-to-the-north shear records, and juxtaposition of lower amphibolite and lesser grade rocks on top of upper amphibolite and higher grade rocks, caused it to be quickly interpreted as a normal fault.

The conceptual challenge posed by the early South Tibet fault interpretations, i.e., why a large normal fault system would span the length of Earth’s highest contractional mountain chain, became the focal point of modeling efforts. Many kinematic models envision the South Tibet fault as a normal fault on top of an extruding wedge of high-grade crystalline material (Fig. 1C). In this context, proposed driving mechanisms for South Tibet fault motion include gravitational sliding along a tilted contact plane (Burg et al., 1984), rotation of principal stresses to near-vertical orientations due to the sharp topographic transition across the Himalayan mountains (Burchfiel and Royden, 1985), subhorizontal shearing below the Himalaya (Yin, 1989), accommodation of the buoyant rise of a partially subducted upper continental crustal slice, possibly triggered by slab detachment (Fig. 1D) (Chemenda et al., 1995, 2000), and a response to gravitational potential energy changes within the context of critical taper orogen models (e.g., DeCelles et al., 2001; Zhang et al., 2011). Alternatively, the South Tibet fault has been interpreted as the upper shear zone on top of a channel flow of high-grade material driven southward by the high gravitational potential of the Tibetan Plateau (Fig. 1E) (Nelson et al., 1996). Furthermore, such channel rocks may have been extruded to the steep, high topographic front of the range between the South Tibet fault and a basal thrust shear zone (termed the Main Central thrust) as a result of an early and/or middle Miocene climate shift that enhanced orographically focused precipitation and resultant erosional exhumation (Fig. 1E) (Beaumont et al., 2001; Hodges et al., 2001).

Recognition that the South Tibet fault may be a backthrust led to tectonic wedging models, in which the crystalline core was emplaced at depth (Fig. 1F) (Webb et al., 2007; Yin, 2006). Because these models do not include normal faulting, they do not require mechanical considerations beyond contractional boundary conditions.

In light of increasing evidence that the crystalline core of the orogen was built in the Oligocene and Miocene by southward-propagating thrust stacking (e.g., Ambrose et al., 2015; Carosi et al., 2010; Corrie and Kohn, 2011; Imayama et al., 2010; Montomoli et al., 2013; Reddy et al., 1993), tectonic wedging models have been superseded by duplexing models. Duplexing models posit that the crystalline core of the Himalayan orogen was built via thrust horse accretion, with the South Tibet fault as the active roof backthrust of a middle-to-deep crustal duplex system (Fig. 1G) (He et al., 2015; Larson et al., 2015).

In contrast to the two-dimensional tectonic models, Himalayan geology has significant arc-parallel variability. The South Tibet fault, i.e., the central structure of Himalayan tectonic models since the 1980s, ceases motion at different times along the length of the range (Leloup et al., 2010). This observation is not commonly cited, but can be noted by analysis of existing data. Likewise, patterns of decompression and cooling of the crystalline core of the orogen (Warren et al., 2014) indicate variable timing of major processes along strike. We outline these key data sets and discuss their kinematic interpretation in the following.

South Tibet fault timing data are shown in map view and plotted by longitude against age in Figures 2 and 3, respectively (see also Data Repository Table DR1¹). We show three categories of age data, labeled pre-/syn-motion, post-motion, and ⁴⁰Ar/³⁹Ar muscovite. Pre-/syn-motion data are U-Pb and Th-Pb dates of accessory phase crystallization and/or recrystallization (e.g., zircon, monazite) in deformed portions of the shear zone. Dated minerals are generally from deformed leucogranites because these are the youngest deformed rocks, and thus provide the tightest constraint on fault motion. The category post-motion data also refers to U-Pb and Th-Pb dates of accessory phases, in this case from undeformed leucogranites crosscutting shear zone fabrics. The ⁴⁰Ar/³⁹Ar muscovite ages are from rocks immediately above the shear zone, within the shear zone, and <3 km structurally below the shear zone, and ideally record the timing of cooling below an approximate closure temperature of 425 °C (Harrison et al., 2009). Most researchers interpret cessation of motion along the subhorizontal shear zone of the South Tibet fault prior to cooling of the shear zone and its local footwall below this temperature range (e.g., Kellett and Grujic, 2012; cf. Cooper et al., 2013).

Interpretation of pre-/syn-motion age data along the South Tibet fault is straightforward, with the exception of standard geochronological challenges that can affect any dating effort (e.g., Pb loss and/or metamict damage to zircon). There are two systematic challenges for post-motion ages, and both challenges indicate that these dates might not constrain the termination age of fault motion. (1) Although the dated leucogranite dikes crosscut shear zone fabrics, no one has identified and dated a dike that crosscuts the entire shear zone of the South Tibet fault. Therefore no one datum can preclude continued motion along the layers of the shear zone that are not crosscut. (2) Dated accessory minerals may be inherited (particularly zircon). If so, the crystallization age could predate the crystallization of the dike and therefore could predate the cessation of shearing along the crosscut shear zone layers. The ⁴⁰Ar/³⁹Ar muscovite ages are known to be affected by excess Ar across vast swaths of the Himalaya (Herman et al., 2010; Webb et al., 2011), which likewise produces excess ages. Furthermore, robust cooling ages should postdate South Tibet fault motion if the shear zone was subhorizontal during activity (e.g., Kellett and Grujic, 2012), because observations at nearly all fault localities suggest that the deformation temperatures exceeded the temperature range of Ar closure in muscovite. However, if the shear zone was north dipping during the primary phase of motion (as many argue, e.g., Burchfiel et al., 1992), then cooling may coincide with fault motion and these constraints would not constrain cessation of South Tibet fault activity.

Our interpretation of the along-strike variations in South Tibet fault cessation timing is denoted by a gray band in Figure 3B. This band generally traces the younger limit of the pre-/syn-motion ages, because these ages are commonly reliable, whereas attempts to date the post-shearing period may regularly yield pre- and/or syn-shearing ages, as discussed here. The interpolation utilizes only post-motion and muscovite ages that are consistent with the younger limits of pre-/syn-motion ages. There are two exceptions: two pre-/syn-motion ages in the east-central Himalaya are sufficiently young relative to the dominant pattern of post-fault motion ages that we assume they are problematic, so we exclude these two ages from the interpolation. The interpreted range of fault cessation timing narrows where data are plentiful (e.g., the central Himalaya), and broadens where there are few published constraints. The interpolated cessation of motion along the South Tibet fault is progressively younger from the western Himalaya (ca. 24–20 Ma) (e.g., Dézes et al., 1999; Vance et al., 1998) to the east-central Himalaya (ca. 13–11 Ma) (e.g., Kellett et al., 2009; Wu et al., 1998) (see Table DR1). Less well resolved is a possible reversal in pattern in the easternmost Himalaya, where sparse data show a sharp spatial transition to older ages (ca. 24–20 Ma) at the eastern end of the range (e.g., Yan et al., 2012). Some have suggested that the dominant pattern of eastward younging in fault cessation timing and a similar pattern in leucogranite crystallization ages may be related to motion along the Karakoram fault (Leech, 2008; Leloup et al., 2010).

The onset of South Tibet fault motion has been speculated to coincide with a metamorphic transition within the Himalayan crystalline core (termed the Greater Himalayan crystalline duplex; Fig. 2) ca. 27–26 Ma (Fig. 3; Table DR2; e.g., Stübner et al., 2014). Geochemical changes in dated monazite and zircon crystals (e.g., variations through time in heavy REE concentrations; Rubatto et al., 2013) suggest that the Greater Himalayan crystalline duplex underwent exclusively prograde metamorphism prior to 27 Ma, whereas some of these rocks record prograde metamorphism and other parts of this rock package record retrograde metamorphism after 26 Ma. Structurally higher rocks record the earliest retrograde metamorphism, and prograde to retrograde pressure-temperature paths are generally younger with increasing structural depth within the unit (Corrie and Kohn, 2011; Rubatto et al., 2013). Sparse data suggest that the metamorphic transition occurs at the same time along the strike of the Himalaya (Fig. 3B).

To reconstruct decompression-time paths and cooling histories, we compile existing data from sites where multiple pressure conditions have been identified and dated, and from sites where multiple temperature conditions have been identified and dated. Such findings from structurally high portions of the Greater Himalayan crystalline duplex are presented in Figures 3C and 3D as plots of pressure and temperature versus time, with the site longitude denoted via color coding (see Table DR3). Furthermore, along-strike cooling patterns are informed via detrital thermochronological dating results from the Himalayan foreland basin compiled in Figure 4 and Figure DR1 (see Table DR4).

Temperature-time constraints from the structurally high portions of the Himalayan crystalline core show that most range sectors cooled from ∼750–550 °C ca. 26–22 Ma to 200–100 °C by ca. 15–10 Ma, and that cooling paths varied systematically along the length of the orogen. Specifically, cooling from the highest temperatures through muscovite closure is progressively younger from the western Himalaya to the east-central Himalaya. Sparse pressure-time constraints might be interpreted to match this eastward younging trend, with the proviso that data of the far eastern Himalaya (from the syntaxial region) do not follow this trend. Instead, these rocks decompressed from ∼1.6 GPa ca. 24 Ma to ∼0.5 GPa ca. 17 Ma (Xu et al., 2010). (For further aspects of decompression from high pressures across the east-central Himalaya, see the Discussion.)

Detrital thermochronology data from foreland basin rocks provide an approximation of the cooling of adjacent Himalayan hinterland regions. A general trend appears in our compilations of ⁴⁰Ar/³⁹Ar muscovite and fission track zircon data: peaks in the cooling age populations appear younger to the east from 25 to 20 Ma to 10–8 Ma (Figs. 4A, 4B) (e.g., Bernet et al., 2006; Chirouze et al., 2012; Jain et al., 2009; Najman et al., 2003). This is consistent with an eastward-migrating pulse of hinterland cooling during this period. The trend is alternately amplified and diminished by along-strike variations in the depositional ages of samples. For example, central Himalaya samples deposited before 15 Ma cannot show cooling pulses younger than 15 Ma, and thus visually weight Figure 4A toward older central Himalayan ages. Similarly, all zircon fission track samples deposited after 10 Ma are from the central and eastern Himalaya, so all cooling younger than 10 Ma plotted in Figure 4B is visually weighted to these regions. Parsing by 5 m.y. increments of depositional age helps to see through these visual effects, as in Figure 4B.ii., which highlights zircon fission track samples deposited from 15 to 10 Ma. This plot is consistent with the general suggestion that a cooling pulse migrated eastward during the early and middle Miocene. Further subplots of this type are presented and explored in Figure DR1, and broadly confirm the trend.

Signals in the Himalayan foreland basin can be complicated by river sediment transport along the range trend (because not all river systems transport sediment perpendicularly away from the mountains) and thus might not only represent cooling and exhumation over the limited extent of the range immediately adjacent to the sampling location. Nonetheless, the observed trends of decompression and cooling are approximately synchronous with progressive early and middle Miocene cessation of South Tibet fault motion along the length of the range (Fig. 3). As with the South Tibet fault cessation, a pulse of decompression and cooling migrates from the western to the east-central Himalaya. In both cases, sparse data suggest that the easternmost Himalaya features a sharp reversal in this trend.

Because existing Himalayan tectonic models are both two dimensional and dominantly limited to crustal processes, lithospheric-scale processes might help explain the along-strike timing variations noted here. Recent work is promising in this respect, showing that the subducted Indian plate became anchored in the mantle during ongoing collision and then detached from the continental lithosphere via tears that initiated at the ends of the Himalaya and propagated inward during late Oligocene–middle Miocene time (Leary et al., 2016; Replumaz et al., 2010).

India indented Eurasia and moved northward over the anchored Indian slab from ca. 30 to ca. 25 Ma, as evidenced across southern Tibet by a southward migration of magmatism (DeCelles et al., 2011; Guo et al., 2013), and possibly by the development of the Kailas Basin (Carrapa et al., 2014; DeCelles et al., 2011; Leary et al., 2016) (Fig. 2). Detachment of the Indian slab at 25–15 Ma has been interpreted on the bases of (1) metamorphic and melting records indicative of a crustal heating event (Rolland et al., 2001; Stearns et al., 2013) (Table DR5), (2) changes in patterns of foreland sedimentation (e.g., Mugnier and Huyghe, 2006) (Table DR5), and (3) seismic tomographic images of the mantle below India that show a seismically fast region interpreted as detached Indian lithosphere (Replumaz et al., 2010). To explain an eastward decrease in the distance between the detached slab and the contiguous Indian craton, Replumaz et al. (2010) proposed that detachment of the slab began in the west ca. 25 Ma and migrated to the east-central Himalaya ca. 15 Ma. Similarly, magmatic records from the western to the east-central Himalaya show a west to east younging trend, which is consistent with eastward propagation of slab detachment (Guo et al., 2015) (see Figs. 2 and 3; Table DR6). For the eastern Himalaya, east to west younging of magmatic rocks from ca. 30–25 Ma at the eastern end to ca. 15–8 Ma in the east-central Himalaya (∼90°E) has been interpreted as a product of east to west lateral migration of slab detachment (Pan et al., 2012; Zhang et al., 2014) (see Figs. 2 and 3; Table DR6).

These findings indicate (1) northward underthrusting of the Indian slab prior to ca. 30 Ma, (2) slab anchoring and steepening from 30 to 25 Ma, (3) slab tearing leading to slab detachment initiating at both ends of the Himalaya ca. 25 Ma then migrating toward the central Himalaya, and (4) final Indian slab break-off occurring in the east-central Himalaya broadly from 15 to 8 Ma. It is intriguing that the lateral migration of slab detachment along the Himalaya corresponds in time and space with the cessation of motion along the South Tibet fault and the pulse of cooling and decompression along the Himalayan arc described here.

The spatiotemporal correlation of lateral migration of slab detachment with the Himalayan faulting, decompression, and cooling suggests systematic linking among these processes. We propose a model in which slab detachment and overall subduction dynamics instigated a series of coupled events, as described here and detailed in Figures 5 and 6.

In this model, slab anchoring (akin to rollback below the northward advance of India) steepened the sole thrust underlying the Himalaya. Such changes in sole thrust geometry are known to change the mechanical equilibrium and deformation kinematics of the orogenic wedge (Davis et al., 1983). In response to this change, the orogenic wedge thickened and shortened internally, initiating the main development of the Greater Himalayan crystalline duplex. Significant volumes of new material were accreted from the subducting Indian plate not only at the front of the orogenic wedge, but also at depth via duplexing. The South Tibet fault initiated as a major backthrust, and functioned as the active roof thrust to the underlying duplex. Slab detachment propagated from the ends of the Himalaya toward the east-central Himalaya as the deformation front moved northward over the anchored slab. As slab detachment propagated, the detached slab portions gradually sank deeper in the mantle and were overridden by the northward-moving Indian continent (Husson et al., 2014; Replumaz et al., 2010). Corresponding southward offset of the vertical traction caused by the weight of the subducted slab rezoned the dynamic deflections of the surface topography (Husson et al., 2014). Initially, the Indian plate subducted underneath the Himalaya, and the associated dynamic topography maintained the elevation of the Himalaya ∼1000–1500 m lower than their plain isostatic elevation. When the subducting slab anchored into the mantle, it moved southward relative to the Indian continent and the Himalaya, and the dynamic deflection was relocated farther south toward the foreland basin. Corresponding shallowing of the Himalayan sole thrust changed the deformation kinematics of the orogenic wedge again (see Dahlen, 1984; Davis et al., 1983), shutting off deep duplexing and backthrusting (see note 12 in Fig. 6B). The deep duplexing that thickened the crystalline core persisted for the longest period in the east-central Himalaya (i.e., the region where the final slab detachment occurred), creating a relatively thick crystalline stack there. We interpret the final cessation of South Tibet fault motion in the east-central Himalaya ca. 13–11 Ma (Figs. 2 and 3) as a gross estimate of final slab detachment timing. A contemporaneous extruding wedge system documented across that region, manifested by an out-of-sequence thrust fault below and a steep normal fault above (Kellett and Grujic, 2012), may be a structural response to final slab detachment.

Model results suggest that the dynamic deflection over active subducting slabs typically is ∼1000 m (e.g., Gurnis, 1992; Husson et al., 2012). In Husson et al. (2014) it was stated that the increase in elevation accompanying the demise of the slab into the mantle should be ∼1 km. Therefore, a topographic rise of this magnitude should start at the two ends of the range (where the horizontal slab tears initiate) ca. 25 Ma and migrate from east and west to the east-central Himalaya at 13–11 Ma (Figs. 5 and 6).

The high topographic barrier of the Himalaya is a key factor in controlling regional atmospheric flow patterns and thus in generating the South Asian monsoon (Boos and Kuang, 2010). Modeling by Ma et al. (2014) indicated that increases in this topography result in increases of monsoon intensity with a roughly linear relationship. Multiproxy records of monsoon intensity indicate that the summer monsoon rains were weak prior to ca. 24 Ma but became progressively stronger, to a peak period from ca. 15 to 11 Ma (Clift et al., 2008; DeCelles et al., 2007; Sun and Wang, 2005; Tada et al., 2016; Wan et al., 2009). We propose that this temporal correlation between the predicted topographic growth and the monsoon intensity reflects Miocene strengthening of the South Asian monsoon as ultimately a product of subduction dynamics.

In each released orogenic region where the sole thrust shallowed after slab detachment, the force balance adjusted accordingly in the Himalayan range. The vertical traction that slab remnants exerted underneath the Himalaya was accompanied by sublithospheric shear tractions. These shear tractions contributed to sustain the convergence of India toward Eurasia. When the Indian slab detached, the force switched from a subduction regime where slab pull dominates to a regime where slab suction dominates (Conrad and Lithgow-Bertelloni, 2004). The northward shear force transmitted by the mantle to the Indian plate declined during this transition after the slab detached, and gradually vanished as the slab remnant sank into the mantle. The gradual demise of the Indian slab load as it sank into the mantle modified the convection pattern and deprived the India-Eurasia convergence of one of its prominent driving forces, and thereby decreased compressive forces at the plate boundary. It follows that convergence rates are predicted to decline during this period. This prediction is broadly consistent with findings from plate circuit reconstructions (e.g., Copley et al., 2010; Iaffaldano et al., 2013; Molnar and Stock, 2009). Such studies show that India-Asia convergence rates quickly dropped after the collision of India ca. 50 Ma and decreased further until ca. 13–11 Ma, after which convergence rates stabilized or modestly increased. The possibility that slab detachment ca. 13–11 Ma produced a change in convergence rates provides an alternative to models in which convergence slowdown results from viscous resistance of intact Tibetan mantle lithosphere (Clark, 2012).

Longitudinal propagation of slab detachment can account for the curvature of the Himalayan mountain belt (see also Capitanio and Replumaz, 2013). The speed of lateral propagation of slab detachment produces variations in orogenic belt curvature: faster propagation produces less curvature and slower propagation produces more curvature. Indentation is faster when it is not lowered by a component of slab pull (which tends to make the trench retreat, whereas indentation makes it advancing), so regions where the longitudinal propagation of slab detachment juxtaposes orogen segments with and without attached slabs are torqued, in a mode similar to retreating subduction zones (Wortel and Spakman, 2000). The degree of bending depends upon the speed of lateral propagation of slab detachment: faster propagation allows less time for regional bending, and thus less arc curvature, and vice versa (Wortel and Spakman, 2000). Because we propose slab detachment propagation across ∼2000 km from the west and across ∼600 km from the east during the same ca. 14–12 m.y. period, our model predicts relatively open arc curvature west of 90°E, and tighter arc curvature to the east of 90°E. West of 90°E, the Himalaya mountain range is renowned for its nearly perfect arc, with a ∼2000 km radius of curvature (Bendick and Bilham, 2001) (Fig. 2). In contrast, farther east the range has tighter curvature (radius of curvature of ∼1200 km) (Fig. 2).

A variety of data sets indicate that major phases of Himalayan tectonic development from late Oligocene through middle Miocene time occurred asynchronously along the strike of the orogen. Such data include constraints on the cessation of motion along the South Tibet fault, cooling and decompression records, seismic tomography of the detached Indian continental slab, and distribution of volcanic rocks across southern Tibet. These findings show the need for time-dependent three-dimensional models. Because most current models are two dimensional, we attempt to create a model including the along-strike dimension. Our model shows major phases of Himalayan construction and uplift controlled by changes in dynamics of the subducting slab. Rollback of the Indian slab relative to the Himalaya initiated development of the Greater Himalayan crystalline duplex and its roof fault, the South Tibet fault, by altering the balance of forces applied to the orogenic wedge. Lateral migration of slab detachment shut off this structural system progressively along the length of the orogen and released the dynamic deflection of the topography that increased the elevation and strengthened the South Asian monsoon. Simultaneously, the release of dynamic traction from sublithospheric mantle flow after slab detachment may also have been responsible for an observed convergence slowdown.

In the following we first discuss how the slab dynamics model relates to and incorporates aspects of published two-dimensional models. We then explore key issues related to the new model, highlighting the timing of high-pressure metamorphism in the east-central Himalaya; the state of knowledge of the topography, monsoon, and exhumation across the system; and the post-slab detachment Himalayan development.

The model presented herein is new in that it explores the consequences of the subducting slab evolution for the crustal dynamics of the Himalayan orogenic wedge and South Asian monsoon evolution. However, the modeled lithospheric-scale evolution largely follows prior work. Underthrusting of Eurasia by India is well established (since the pioneering work of Argand, 1924), and cycles of rollback, lateral migration of slab detachment, and underthrusting with corresponding topographic effects have previously been explored in this region (e.g., Replumaz et al., 2010; DeCelles et al., 2011; Husson et al., 2014; Leary et al., 2016).

The development of the Greater Himalayan crystalline duplex and the corresponding motion along the South Tibet fault in the new model are generally consistent with the duplexing model presented by He et al. (2015), as well as many aspects of the duplexing model of Larson et al. (2015). As in the duplexing models, the new model shows the development of the Greater Himalayan crystalline duplex at depth, as a thrust duplex with a roof backthrust (the South Tibet fault) and the slip distance per accreted horse roughly equivalent to horse length (Fig. 6). Also similar to the duplexing models and the earlier tectonic wedging models, the slab dynamics model involves late (post–10 Ma) exposure of the main body of the Greater Himalayan crystalline duplex rocks. This is controversial in that the foreland detrital record is commonly interpreted to indicate early Miocene erosion of these rocks (e.g., DeCelles et al., 1998). However, prior analyses suggest that such detrital records could be produced by erosion of other Himalayan units in combination with isolated exposures of the Greater Himalayan crystalline duplex rocks by ca. 11 Ma (possibly along east-west extensional core complex systems in the Himalayan hinterland) followed by widespread exposure by ca. 5 Ma (see Yin, 2006; Webb, 2013).

Incorporation of the slab dynamics history enriches our understanding of proposed duplexing of He et al. (2015) by adding a series of detailed predictions that compare favorably with the geological record. The slab dynamics model offers rationales for why Greater Himalayan crystalline duplex growth and South Tibet fault motion start and finish. Namely, slab anchoring should steepen the Himalayan sole thrust, whereas slab detachment should allow rebound and shallowing of the sole thrust, and such changes to the sole thrust geometry are well understood to start and stop thickening of orogenic wedges (Dahlen, 1984). The model suggests that duplex growth and South Tibet fault motion should initiate after the ca. 30 Ma start of slab anchoring and before the ca. 25 Ma start of slab detachment. The ca. 27–26 Ma metamorphic transition from exclusively prograde to mixed prograde and retrograde metamorphism (Fig. 3; Table DR2) may signal this onset, with the retrograde metamorphism reflecting exhumation in response to thrust horse stacking. As for cessation timing, the main correlations that led to the model construction are the along-strike correspondence of cooling, decompression, and South Tibet fault cessation with slab detachment migration inferred from seismic tomography and southern Tibetan volcanism.

The slab dynamics model includes a late out-of-sequence extruding wedge system in the east-central Himalaya (Fig. 6) that has some commonalities with wedge extrusion models (e.g., Burchfiel and Royden, 1985; Chemenda et al., 1995, 2000). Wedge extrusion occurs as a response to deep burial of light crustal materials and also potentially to slab detachment in the models of Chemenda et al. (1995, 2000). The wedge extrusion of our model is localized to the east-central Himalaya, where a north-dipping brittle normal fault that crops out along the range crest accomplished rapid footwall cooling ca. 16 to ca. 12 Ma (e.g., Carrapa et al., 2016; Kellett et al., 2013), and to the south a contemporaneous out-of-sequence thrust system occurs (e.g., Grujic et al., 2011; Larson et al., 2016). The apparently restricted range of these systems could indicate that they respond to the relatively large magnitude burial and subsequent uplift associated with the final slab detachment, as in the Chemenda group modeling. Localized normal faulting associated with such wedge extrusion can help resolve confusion over South Tibet fault kinematics (i.e., the decade-old debate over whether it is a thrust or a normal fault). In this region (specifically, from eastern Nepal through the Bhutan Himalaya), many exposures of the South Tibet fault along the Himalayan range crest are spatially associated with the north-dipping brittle normal fault (e.g., Carrapa et al., 2016; Kellett et al., 2013). This region hosted much of the early work along the South Tibet fault (e.g., Burg et al., 1984; Burchfiel et al., 1992), and therefore the brittle fault is commonly interpreted as the last phase of South Tibet fault motion. However, this brittle fault is not seen in other sectors of the Himalaya, where structural geometry and cooling histories across the South Tibet fault suggest motion along a subhorizontal structure (e.g., Vannay et al., 2004; Webb et al., 2013). Subhorizontal ductile shear dominates South Tibet fault evolution, whereas late brittle normal faulting may cut this shear zone only where a late wedge extrusion system responded to final slab detachment.

The slab dynamics model makes specific claims about the nature and timing of high-pressure metamorphism in the east-central Himalaya. Specifically, in the model this metamorphism reflects the steepening and deepening of the orogenic wedge here as the slab steepened. The region would have experienced pressures that were anomalously high, perhaps to eclogite facies conditions, because it was the last region to undergo slab detachment. Deep duplexing and localization of slab weight would have persisted longest here. Because the slab was already detached both to east and west of this region prior to final slab detachment, this region would have supported some fraction of the neighboring detached slab weight both to east and west, approximately doubling this effect of excess adjacent weight in the few million years prior to final slab detachment. It follows that the high-pressure metamorphism of the lower orogenic wedge should have occurred only in the few million years immediately prior to the ca. 13–11 Ma final slab detachment. This model prediction is consistent with direct U-Pb dating of zircon in the east-central Himalaya. In combination with geochemical and textural analyses, U-Pb zircon geochronology yields eclogite facies metamorphic periods of 15.3 ± 0.3 to 14.4 ± 0.3 Ma (Grujic et al., 2011) and 14.9 ± 0.7 to 13.9 ± 1.2 Ma (Wang et al., 2017). However, the model prediction does not appear consistent with (1) published interpretations of Lu-Hf dating of high-pressure garnet (Corrie et al., 2010; Kellett et al., 2014) and (2) a study by (Regis et al., 2014) that links monazite geochronology with metamorphism after the high-pressure period. These studies argued that high-pressure metamorphism in this region occurred as early as ca. 38 Ma and locally persisted until ca. 15–13 Ma. We review the latter data sets and their context, show that alternative interpretations are compatible with the slab dynamics model, and discuss broader implications of this analysis.

Isotope geochronology on metamorphic minerals can be used to temporally constrain different parts of the pressure-temperature evolution. For example, Lu-Hf and Sm-Nd geochronology data are commonly interpreted to date early and late stages of metamorphic garnet growth, respectively. This interpretation is based on the fact that garnet preferentially incorporates heavy REEs, resulting in high Lu concentration in garnet cores, whereas Sm is rather homogeneously distributed (e.g., Kohn, 2009; Lapen et al., 2003). Commonly, the growth of garnet can be related to high-pressure conditions, and thus application of Lu-Hf garnet geochronology has been used to constrain high-pressure metamorphism in the east-central Himalaya to ca. 26–23 Ma (Corrie et al., 2010), or even to as old as ca. 38–34 Ma (Kellett et al., 2014).

However, the interpretation of Lu-Hf age data has been challenged based on evidence for different mechanisms that may modify the extracted age. Skora et al. (2006) presented a model for diffusion-limited uptake of REEs in garnet that would create local depletion of the REEs around the garnet accompanying the crystal growth and prevent equilibration with the bulk matrix. Sousa et al. (2013) used a mass-balance model to show that garnet isotope composition may not equilibrate with the bulk matrix, and thus reactivity and modes of reactant minerals govern the local effective bulk composition and will determine the initial Rb/Sr and/or Lu/Hf during garnet growth; their modeling suggests significant modification, to several tens of millions of years, in the extracted age for the case of Rb-Sr age data. The case of Lu-Hf is not as straightforward because the source of Lu prior to garnet growth remains elusive and the reactivity of zircon as source of matrix Hf is also unclear. However, a recent study using a large data set of detrital zircons from the Himalaya revealed large variation in the ε_Hf value between -24 to +3, which suggests that zircon may be actively contributing to modification of the matrix composition during metamorphism (Ravikant et al., 2011). Sousa et al. (2013) used this range in ε_Hf values to predict the apparent age error using ¹⁷⁶Lu/¹⁷⁷Hf in garnet; their calculations indicate that Lu-Hf data may be inaccurate by several million years depending on different shifts of the matrix Hf isotope composition caused by zircon recrystallization.

A different mechanism that may alter recorded Lu-Hf ages results from subtle differences in the diffusivity of parent and daughter isotope. Lu diffusion in garnet may be faster than its radiogenic daughter Hf (Mueller et al., 2010; Skora et al., 2006) based on comparison to REE + Hf diffusion data in zircon (Cherniak et al., 1997a, 1997b). This assumption has been also experimentally verified (Bloch et al., 2015). Therefore, Lu-Hf is different from other geochronology systems in that its parent isotope, and not the radiogenic daughter, may be preferentially lost from the crystal at sufficiently high temperatures (i.e., above the nominal closure temperature). This may not be problematic if Lu preferentially migrates into (or stays within) the garnet. However, the preferred partitioning of Lu into garnet decreases with increasing temperature, making matrix minerals such as clinopyroxene suitable hosts for Lu (Van Orman et al., 2001). Therefore, Lu potentially leaves the garnet at higher rates compared to its radiogenic daughter Hf and may accumulate in grain boundaries (Hiraga et al., 2004) or may be incorporated into matrix minerals or accessory phases. As a result, a lower Lu/Hf ratio is recorded in the garnet that translates into an apparent older age.

These processes may shift the extracted Lu-Hf data toward older ages by as much as tens of millions of years. We therefore interpret previously extracted Lu-Hf data to be potentially modified, and hence high-pressure conditions indicated by garnet growth may represent exclusively middle Miocene metamorphism.

Regis et al. (2014) explored the Jomolhari massif of northwest Bhutan and used a different suite of data to argue for eclogitic metamorphism in the east-central Himalaya prior to ca. 36 Ma. Prior work shows that a mafic eclogite from the northern end of the Jomolhari massif yields a U-Pb titanite cooling age of 14.6 ± 1.2 Ma (mean square of weighted deviates = 0.2, closure temperature estimated as between ∼700 and 500 °C) (Warren et al., 2012). Regis et al. (2014) used monazite petrochronology to show that metasedimentary rocks in the central and southern Jomolhari massif underwent granulite facies metamorphic conditions of 0.85 GPa and 800 °C ca. 36 Ma, and remained at high temperatures until at least ca. 18 Ma; they used the assumption that the Jomolhari massif represents a coherent rock body to then infer that the high-pressure metamorphism (recorded by the mafic eclogite) preceded the granulite facies metamorphism. In this interpretation, the eclogite-facies metamorphism must be older than ca. 36 Ma. However, if the Jomolhari massif did not evolve as a coherent rock body, then the northern eclogites may be structurally separated from the southern granulites. For example, the eclogites could be in the hanging wall of an out-of-sequence fault (possibly the Kakhtang thrust of Grujic et al., 2011), and the granulites may be in the footwall. In such cases, eclogitic metamorphism here may have occurred as late as ca. 17–13 Ma and predate structural juxtaposition with the granulitic rocks.

In summary, although prior interpretations of eclogite facies metamorphism timing across the east-central Himalaya appear inconsistent with the slab dynamics model, viable alternative interpretations of all constraints allow that this metamorphism may have occurred in middle Miocene time, which is consistent with the model. Of particular interest for geochronological study are the alternative interpretations of Lu-Hf garnet dates that suggest that these dates are not accurate in that they are much older than the actual timing of the eclogite facies metamorphism. Further exploration of the prograde to peak metamorphic timing here may confirm long-standing hypotheses that Lu-Hf garnet dates could greatly exceed the geological ages of dated events (e.g., Skora et al., 2006).

Construction of mountain chains and elevated plateaus is understood to be strongly influenced by climate-modulated erosion (Beaumont et al., 2001; Konstantinovskaia and Malavieille, 2005; Montgomery et al., 2001), and by subducting plate (or slab) dynamics (Carrapa et al., 2014; Fox et al., 2015; Replumaz et al., 2010; Wortel and Spakman, 2000). However, how climate and slab dynamics affect each other during mountain building remains poorly understood (Iaffaldano et al., 2011; Lamb and Davis, 2003). Various chicken versus egg interpretative challenges further limit our ability to decipher climate-erosion-tectonics interactions (Clift et al., 2008; Molnar and England, 1990). For example, for many mountain belts it is unclear whether tectonic shifts forced climatic changes, or climatic shifts generated new tectonic regimes. These issues are well illustrated in studies of the Himalaya, where subduction dynamics is recognized to uplift the range (Husson et al., 2014) and deform the Tibetan Plateau (DeCelles et al., 2011; Replumaz et al., 2014), but models of the kinematic evolution of the Himalayan mountains feature static subduction zone geometries (e.g., Beaumont et al., 2001; Herman et al., 2010; Webb, 2013) with only few exceptions (Carrapa et al., 2014; King et al., 2011). The rise of the Himalayan mountains is thought to explain the development of the South Asian monsoon (Boos and Kuang, 2010), yet the only published model with a significant role for climate, the channel flow model, shows major rock uplift and exhumation as being triggered by enhanced erosion resulting from the onset of the monsoon (Beaumont et al., 2001; Clift et al., 2008). The new model suggests instead that slab dynamics triggered a phase of Himalayan uplift, which in turn caused the intensification of the South Asian monsoon. Therefore, because subduction dynamics remains a priori unaffected by vagaries in climate, the problem is in principle no longer a chicken or egg issue, but instead a univocal relationship: there is a trigger and a target. Nevertheless, climate-induced erosion can modulate mantle convection and therefore tectonic velocities (Iaffaldano et al., 2011), so if future work can demonstrate that slab anchoring and detachment may be induced by climatic changes, then the feedback loop will close again.

In the context of the slab dynamics model, slab detachment would produce a shallowing of the Himalayan sole thrust, a maximum height in topography, and arc curvature (Figs. 5 and 6). We speculate that these factors could have had a broad range of consequences for post-slab detachment tectonics.

The high topography might have created a positive feedback between climate and tectonics, thereby maintaining the high topography: by intensifying the monsoon, erosion increased, leading to structural changes making shallow to middle crustal duplexing more vertically directed (i.e., antiformal stack development; see Konstantinovskaia and Malavieille, 2005), thereby providing the uplift necessary to maintain high topography and, in turn, the strong monsoon.

Normal fault systems accomplishing orogen-parallel extension across the northern Himalaya are thought to result from orogen-perpendicular thrusting along the Himalayan arc, because as rock packages are thrust forward they must span arc segments of increasing length (Murphy et al., 2009). If arc curvature controls these systems, then proposed progressive development of arc curvature in response to the lateral migration of slab detachment would predict that these systems initiated at different times along the length of the arc. Sparse data support this possibility, as the Leo Pargil extensional system of the western Himalaya may have developed ca. 23 Ma, ∼8 m.y. prior to the development of similar systems in the east-central Himalaya (Langille et al., 2012).

Some have coupled thermochronological data with balanced palinspastic reconstructions to argue for variations in Himalayan shortening rates of as much as an order of magnitude over the past ∼20 m.y. (Long et al., 2012; McQuarrie and Ehlers, 2015; Robinson and McQuarrie, 2012; Tobgay et al., 2012). These reconstructions have not considered slab dynamics impacts on the crustal kinematics and cooling histories. The along-strike temporal correlation between cooling pulses and slab detachment suggests that these reconstructions would benefit from reevaluation.

Many explorations of Himalayan tectonics in recent years have focused on along-strike changes in tectonic processes, and these focus almost entirely on post-Miocene processes (e.g., Cannon and Murphy, 2014; Copeland et al., 2015; Grujic et al., 2006; Van der Beek et al., 2016). In this work we show along-strike timing variations in Oligocene–Miocene Himalayan tectonic processes and relate these to a model in which the deformation of the Himalayan orogenic wedge was largely governed by slab dynamics processes. The model suggests that the along-strike timing variations were controlled by lateral migration of slab detachment. Some exciting outcomes of the model are new explanations for the intensification of the South Asian monsoon, the Miocene slowing of India-Eurasia convergence, and the development of asymmetric Himalayan arc curvature.

The proposed slab dynamics model also changes our understanding of Miocene Himalayan development within the broader context of East Asia collisional tectonics. Slab detachment is thought to initiate motion on major strike-slip faults within East Asia (Replumaz et al., 2014), suggesting that strong links between collision frontal and intracontinental deformation are controlled by slab dynamics. Recognition of climatic-tectonic links during Miocene slab detachment may be combined with knowledge of earlier slab anchoring-detachment-underthrusting cycles along the southern margin of Asia (DeCelles et al., 2011; Husson et al., 2014; Kapp et al., 2007; Replumaz et al., 2010) and elsewhere to explore how slab dynamics may have modulated climate throughout Earth’s plate tectonic history.

Thus far, community responses as we attempt to introduce this work have focused on the question of whether the model is “right” or not. We suggest that it is more important to check for current viability, because all models eventually meet Ozymandian fates. Furthermore, it is critical to consider whether the compiled data truly require significant third-dimensional variability during the Oligocene–Miocene development of the Himalayan mountains; if so, then our model serves as an early attempt to grapple with this variability, and we anticipate better works in the future.

We thank Fabio A. Capitanio and two anonymous reviewers for feedback on a closely related rejected manuscript; they helped inform our approach to this work. Reviews from J. Matthew Cannon and two anonymous reviewers helped us to improve the present contribution. Discussions with Jason Ali, Jess King, and Ryan McKenzie helped us clarify our concepts and communication. Funding for this work comes from the U.S. National Science Foundation (grant EAR-1322033 to Webb) and the Chinese Natural Science Foundation (grant 41430212 to Xu, Cao, Wang, and Webb). Webb dedicates this work to Xi Chen.