Eocene deep crust at Ama Drime, Tibet: Early evolution of the Himalayan orogen

The Cenozoic Himalayan orogen presents a spectacular opportunity to study a complete section of the crust that was reworked during a major continent-continent collision. The south-facing orogenic front of the Himalaya exposes both sedimentary rocks of the Tethyan sedimentary sequence and the amphibolite- to granulite-facies Greater Himalayan sequence and associated high-metamorphic-grade rocks. Together these units provide a window into upper- and midcrustal metamorphism and deformation during the early stages of a continental collision (e.g., Godin et al., 2001; Aikman et al., 2008; Cottle et al., 2009; Carosi et al., 2010). However, the early evolution of the lower crust within this system remains enigmatic.

The Greater Himalayan sequence is interpreted to have formed a part of the northern passive margin of India that was buried, thickened, metamorphosed, and ultimately exhumed during the Cenozoic India-Asia collision (e.g., Hodges, 2000). Rocks of the Greater Himalayan sequence exhibit a pervasive imprint of Miocene high-temperature metamorphism associated with anatexis, and they preserve ductile fabrics that indicate Miocene subhorizontal, south-verging, pervasive, ductile flow (e.g., Grujic et al., 1996; Hodges, 2000). The temperatures of metamorphism and deformation of the Greater Himalayan sequence broadly increase from west to east, with granulite-facies rocks occurring in the eastern Himalaya, but not in the western Himalaya, except for the western syntaxis region. The Greater Himalayan sequence rocks in the eastern Himalaya thus reflect some of the deepest crustal levels during collision and provide an important view into the deep-crustal evolution of this orogenic belt.

Vestiges of high-pressure (HP) metamorphic rocks associated with the Greater Himalayan sequence in the eastern Himalayan regions of Ama Drime in Tibet, Arun Valley in Nepal, Sikkim in India, and NW Bhutan provide an opportunity to “sample” the lower crust from a continental collision (e.g., Groppo et al., 2007; Cottle et al., 2009; Corrie et al., 2010; Rolfo et al., 2008; Grujic et al., 2011; Fig. 1). The peak pressure conditions for these rocks are estimated to have reached >1.5 GPa at >580 °C (Ama Drime; Groppo et al., 2007) and >1.5 GPa at ∼670 °C (Arun Valley; Corrie et al., 2010), indicating they were at lower crustal depths of ∼60 km at the time of HP metamorphism. Although critically important for the construction of reliable geodynamic models, the timing of this metamorphism is poorly known. Through Lu-Hf garnet geochronology and new U-Pb data, this study reassesses this part of the history of the mid- to lower crust at Ama Drime, Tibet.

These new data yield the timing of prograde HP metamorphism, as well as granulite-facies overprinting, providing constraints on the early evolution of the lower crust during the Himalayan orogeny.

Eclogite-facies rocks were first discovered in the Himalayan orogen in the Kaghan Valley in Pakistan (Pognante and Spencer, 1991), where they occur along the Indus-Yarlung suture at the western syntaxis of the orogen (Fig. 1). Eclogites have also been identified at Tso Morari in NW India, east of Kaghan along the Indus-Yarlung suture (de Sigoyer et al., 1997; Guillot et al., 1997). Both eclogite occurrences comprise well-preserved ultrahigh-pressure (UHP) assemblages that record peak pressures of ≥2.7 GPa at ca. 55–46 Ma (de Sigoyer et al., 2000; O’Brien et al., 2001; Treloar et al., 2003; Leech et al., 2005; Parrish et al., 2006). Their timing of metamorphism and structural position suggest that these rocks were subducted with the leading edge of the Indian plate, then detached, and rapidly exhumed to a shallow crustal position shortly after collision of India with Asia.

More recently, granulitized eclogites were identified in the eastern Himalaya (e.g., Lombardo and Rolfo, 2000). In contrast to the UHP rocks of the western syntaxis region, retrogressed eclogites in the eastern Himalaya: (1) are not spatially associated with the Indus-Yarlung suture, but occur within or beneath the Greater Himalayan sequence; (2) do not preserve UHP indicators; and (3) are strongly overprinted by granulite- and amphibolite-facies events. All eclogite bodies occupy a similar structural position within the midcrustal package, relatively close to the South Tibetan detachment system (Fig. 1, inset). A sample from the Arun Valley is the possible exception; it occurs farther south and appears to lie structurally much lower in the exposed midcrust package (Corrie et al., 2010). A Lu-Hf date of 20.7 ± 0.4 Ma for a granulitized eclogite in Arun Valley (Corrie et al., 2010) and U-Pb dates of zircon in a similar rock from NW Bhutan of 16–14 Ma (Grujic et al., 2011) suggest that eclogites-facies metamorphism in the eastern Himalaya may be of late Oligocene or Miocene age.

The Ama Drime Massif consists of complexly intercalated paragneiss and orthogneiss units that dip to the north. These units either underlie, or are a deeper section of, the Greater Himalayan sequence and are bounded to the west and east by Miocene N-S–striking detachments, the Ama Drime and Nyönno Ri detachments, respectively (e.g., Jessup et al., 2008; Cottle et al., 2009; Fig. 1). Vestiges of HP metamorphism, as identified by Lombardo et al. (1998) and Lombardo and Rolfo (2000), are preserved in mafic pods hosted by orthogneiss. Zircon U-Pb analyses demonstrated that the igneous protoliths of the mafic pods and orthogneiss formed at ca. 990 and ca. 1800 Ma, respectively (Cottle et al., 2009). Four stages in the metamorphic history of the mafic rocks were identified: (M1) eclogite-facies metamorphism at >1.5 GPa, >580 °C; (M2) granulite-facies metamorphism at 0.8–1.0 GPa, >750 °C; (M3) granulite-facies overprint at ∼0.4 GPa, ∼750 °C; and (M4) amphibolite-facies retrogression at <0.4 GPa, ∼700 °C (Groppo et al., 2007). The age of the M3 stage was constrained to be younger than ca. 13 Ma using U-Th/Pb dating of monazite and xenotime (Cottle et al., 2009). Muscovite and biotite ⁴⁰Ar/³⁹Ar cooling ages of ca. 13–10 Ma (Kali et al., 2010) indicate rapid cooling from ∼750 °C to ∼300 °C during the mid-Miocene.

All the samples analyzed in this study were collected from the AD43 sample site of Cottle et al. (2009; 28.182265°N, 87.391009°E, see their fig. 3). Field relationships and petrographic descriptions of AD43 are summarized here, but see Cottle et al. (2009) for detailed descriptions. The mafic eclogite body from which AD43 was sampled is lenticular, with dimensions of ∼60 m × ∼20 m. The long axis of the lens is parallel to a dominant subhorizontal ductile fabric in the Ama Drime orthogneiss. The mafic body has been imbricated by brittle bookshelf sliding during overall top-to-the-W ductile shear of the Ama Drime orthogneiss, and these fractures have been filled by ca. 12 Ma leucogranite dikes (Cottle et al., 2009). AD43 is a dark-green, medium-grained, garnet-bearing mafic rock containing abundant clinopyroxene + plagioclase symplectite, orthopyroxene, and hornblende, which are interpreted to have replaced the eclogite mineral assemblage Ca-rich garnet + omphacite + amphibole (Fig. 2).

Small (5–50 µm) zircon grains are abundant in subsamples AD43A–AD43D. In general, zircon grains in these samples exhibit a rounded (e.g., Figs. 3E and 3G) to bead-like (Figs. 3B, 3C, 3F, and 3H) morphology. Grains are anhedral, and only a few grains exhibit an igneous-type internal zoning (Figs. 3A and 3J). Zircon is most commonly found at grain triple junctions (e.g., Figs. 3B, 3C, 3D, and 3F; Table 1) and associated with ilmenite (e.g., Figs. 3E, 3G, and 3I; Table 1).

Zircon in thin section was imaged under cathodoluminescence (CL) and backscattered-electrons (BSE) at the University of California–Santa Barbara (UCSB) using a scanning electron microscope (Fig. 3). U-Pb geochronology was conducted at UCSB by sampling zircon in situ in thin sections using a Photon Machines ArF Excimer coupled to a Nu Plasma multicollector–inductively coupled plasma–mass spectrometer (ICP-MS) at UCSB. Spot selection was guided by the CL and BSE images. Analytical details are outlined in the supplementary data,¹ and results are listed in Table 1 and plotted in Figure 4.

Each sample yielded two populations of U-Pb data: a spread of discordant data with a Neoproterozoic upper intercept (ca. 980 Ma) and a suite of data exhibiting high common Pb and yielding Miocene lower-intercept ages (Fig. 4). The lower-intercept ages for the latter data set are 14.2 ± 3.3 Ma, 16.8 ± 1.7 Ma, and 14.0 ± 0.3 Ma for AD43A, AD43B, and AD43C, respectively (see Tera-Wasserburg diagrams, Fig. 4). AD43D did not yield sufficient data to reliably constrain a lower-intercept age.

AD43A garnet grains have, on average, a diameter of 2.15 mm, with a rim diameter of ≤0.15 mm (Fig. 5A). Garnet grains in AD43B have an average diameter of 6 mm, with a rim diameter of ≤0.4 mm (Fig. 5B). AD43D garnet grains are ∼2.3 mm across with a rim diameter of ≤0.25 mm (Fig. 5C).

Whole-rock major-element and trace-element compositions were determined using X-ray fluorescence (XRF) and ICP-MS at Washington State University (Table 2). Garnet X-ray mapping and major-element analyses (Fig. 5) were collected by electron microprobe analysis (EMPA) on polished thin sections using a CAMECA SX-100 housed at UCSB. Trace-element concentrations in garnet (Fig. 5) were analyzed by laser-ablation single-collector ICP-MS using a Photon Machines ArF Excimer (λ = 193 nm) laser coupled to a Nu Instruments AttoM high-resolution ICP-MS, also housed at UCSB. All data and analytical details are compiled in the supplementary data (see ^{footnote 1}).

Lutetium and Hf were separated from elemental matrices using cation-exchange chromatography modified after Münker et al. (2001), and Lu and Hf analytes were analyzed using the Nu Plasma multicollector ICP-MS at UCSB. Apparent ages and uncertainties (2 s.d.) were calculated using Isoplot version 3.27 (Ludwig, 2003), applying a decay constant of 1.867 × 10⁻¹¹ yr^–1 for λ¹⁷⁶Lu (Scherer et al., 2001, 2003; Söderlund et al., 2004). Full technical details on sample preparation and Lu-Hf analysis are given in the supplementary data (see ^{footnote 1}).

Garnet is rich in almandine and typically concentrically zoned, with grains exhibiting Ca-rich cores that have a 29%–33% grossular component (Fig. 5). More complex zoning also occurs (Fig. 5B). Groppo et al. (2007) modeled analogous garnet core and rim compositions as having formed during M1 (eclogite-facies) and M2 (granulite-facies), respectively. Lutetium concentrations are highest in garnet from AD43A, averaging at ∼4 ppm in grain cores and 1 ppm in grain rims (Fig. 5A). Samples AD43B and AD43D both contain less Lu (<1.5 ppm) and do not display clear zoning patterns (Figs. 5B and 5C). However, Lu concentrations in AD43B cores are up to ∼1 ppm Lu, while rims are more typically 0.2–0.4 ppm. Lu in the cores of AD43D also reach ∼1 ppm, while rims are 0.2–0.4 ppm. Bulk samples consistently yielded Lu/Hf between 0.16 and 0.17. The samples provided apparent Lu-Hf ages of 33.9 ± 0.8 Ma, 36.0 ± 1.9 Ma, and 37.5 ± 0.9 Ma for AD43A, AD43B, and AD43D, respectively (Table 3; Fig. 6).

The Neoproterozoic ages indicated by upper intercepts in ²⁰⁶Pb/²³⁸U-²⁰⁷Pb/²³⁵U space (Figs. 4A, 4C, 4E, and 4G) agree with a previously determined igneous age of these samples of 987 ± 2 Ma (Cottle et al., 2009) and are reasonably interpreted to represent crystallization of the igneous mafic protolith. The few grains that preserve vestiges of igneous-type growth zoning (e.g., Figs. 3A and 3J) yielded the oldest ages and are interpreted to have undergone the least amount of recrystallization.

The young discordant grains commonly form unusual bead-like, anhedral textures (e.g., Fig. 3H). The smallest grains analyzed were at the size range of the laser spot size used (∼20 µm). While these analyses necessarily liberated a significant amount of common Pb from the surroundings of the zircon grains, the small zircon grains also appear to represent the most completely recrystallized grains, or possibly neocrystallized zircon. Lines of best fit from an assumed primordial common Pb consistently yield Miocene ages for recrystallization. In these samples, zircon is rarely included in the rock-forming minerals; it is most commonly located at grain boundaries, suggesting that its formation was related to late-stage metamorphic reactions. Zircon is also commonly located in, adjacent to, or near ilmenite, which is thought to have replaced M1 eclogite-facies rutile during M2 granulite-facies metamorphism (Groppo et al., 2007). The bead-like anhedral morphology, close association with ilmenite, and irregular internal zoning of these grains are all features that have previously been associated with crystallization during granulite-facies metamorphism (Corfu et al., 2003, and references therein). Thus, we interpret that zircon was recrystallized and possibly formed anew during granulite-facies metamorphism at ca. 15–13 Ma. This is consistent with a previous age estimate for granulite-facies metamorphism of <13.2 ± 1.4 Ma as determined from monazite and xenotime geochronology, and the occurrence of postkinematic and postmetamorphic dikes at 11.6 ± 0.4 Ma (Cottle et al., 2009).

Because of the strong partitioning of Lu into garnet, Lu tends to concentrate in garnet cores relative to rims and relative to the bulk rock (e.g., Kohn, 2009). This effect is observed in the three samples subjected to Lu-Hf geochronology, albeit to variable extents. For this reason, and the low diffusivity of Hf (Scherer et al., 2000; Smit et al., 2013), the Lu-Hf system in garnet typically yields an integrated age that represents, or is weighted toward, prograde garnet growth (e.g., Lapen et al., 2003). Our bulk-grain analyses incorporated components of both eclogitic (M1) and granulitic (M2) garnet, such that the resulting ages could be a mixed result between significantly different age components. A comparison of the Lu concentration in garnet core versus garnet rim of these samples demonstrates that garnet ages are biased toward the garnet core age, with only a small contribution from the garnet rim. The relative contribution of these zones to the bulk-grain Lu-Hf ages can be broadly estimated by means of a simple mass balance approach (e.g., Corrie et al., 2010). Basing our calculations on the average core and rim Lu concentrations, average garnet diameters, and thickest rim diameters (to minimize bias due to resorption, e.g., ∼0.15 mm for AD43A; Fig. 5A), we estimate that M1 garnet is responsible for roughly 87% of the total Lu budget in the three samples (see the supplementary data and Table A3 for details of the calculations [see ^{footnote 1}]). The bulk-grain Lu-Hf ages are thus strongly weighted toward the M1 garnet age component. This is consistent with the observation that the Lu concentrations determined by isotope dilution analysis closely approximate those determined for the cores of grains by laser-ablation ICP-MS (Table 3; Fig. 5). Assuming that the complementary ∼13% is represented by the age estimated here from zircon U-Pb and in Cottle et al. (2009; 15–13 Ma), then our Lu-Hf data yield an age of ca. 38 Ma for M1. While this is a coarse estimation, it illustrates that the Lu-Hf age results largely represent the age of garnet cores and likely closely approximate the nucleation of the crystals during ca. 38 Ma M1 metamorphism.

Resorption of garnet can result in alteration of garnet ages as parent Lu is preferentially incorporated in garnet rims while daughter Hf returns to the bulk rock (Kelly et al., 2011). While garnet rim growth in these samples was found typically to be concordant with euhedral garnet cores, all three samples have textures suggestive of rim corrosion or resorption, perhaps during M4 amphibolite-facies metamorphism. AD43B and AD43D both show high Lu concentrations at the garnet grain boundary, which support an interpretation of Lu reincorporation during resorption. These observations suggest that the spread of Lu-Hf ages to some extent reflects a variable amount of garnet resorption (greatest in AD43A and least in AD43D). As did the presence of rims, this inference indicates that ca. 38 Ma may be a minimum age for M1 metamorphism.

The prograde metamorphism recorded by AD43 is significantly older than the growth and/or recrystallization of zircon, xenotime, monazite, and uranothorite in the same outcrop (e.g., Cottle et al., 2009). The Miocene U(-Th)-Pb ages from this study are interpreted to represent either of the lower-pressure M2–M3 metamorphic events (Groppo et al., 2007). These Miocene results coincide in time with the apparent Lu-Hf ages of 15–14 Ma obtained for amphibolite-facies gneisses in Arun Valley by Corrie et al. (2010). This consistency suggests synchronous amphibolite-facies metamorphism in the Greater Himalayan sequence at Arun Valley and granulite-facies metamorphism at Ama Drime.

Interestingly, a mafic granulitized eclogite from Arun Valley, having similar mineral content, texture, and host rock to those described here, yielded a single Lu-Hf age of 20.7 ± 0.4 Ma (Corrie et al., 2010), ∼17 m.y. younger than the Lu-Hf ages from Ama Drime. This age is interpreted by Corrie et al., using a mass balance approach, to represent a mixed age of ∼65% older (26–23 Ma) eclogites-facies garnet and 35% younger granulite-facies garnet. Furthermore, zircon U-Pb and trace-element data from NW Bhutan indicate eclogite-facies metamorphism during ca. 15–14 Ma (Grujic et al., 2011). The differences in age between samples from Arun Valley, NW Bhutan, and our samples could indicate diachroneity in M1 metamorphism, occurring at least 10 m.y. later for rocks now exposed at Arun Valley, and at least 20 m.y. later for rocks now exposed in NW Bhutan. Such diachroneity would suggest that the granulitized eclogites in the eastern Himalaya do not represent a single metamorphic event, but rather record progressive conditions during a >20 m.y. history of HP metamorphism in the deep crust. Eclogitization of Indian lower crust is interpreted to be ongoing in the eastern Himalaya beneath the Tibetan Plateau, ∼100 km north of the Indus-Tsangpo suture, forming a layer of eclogitized Indian lower crust beneath Tibet (Hetényi et al., 2007). Taken together, we conclude that eclogitization beneath Tibet occurred as early as ca. 38 Ma and has been a continuous process ever since.

The implications of the new Lu-Hf ages for understanding the evolution of the Himalayan orogen depend upon the tectonic setting of the analyzed rocks at the time of their prograde metamorphism. The possible end-member tectonic scenarios include subduction-related metamorphism, as in the case of the northwestern Himalaya examples, or metamorphism as a result of crustal thickening by shortening or underthrusting. Following the first end-member, Corrie et al. (2010) considered the possibility that the Arun Valley and Ama Drime eclogites were part of a slab that was eclogitized during its subduction into the mantle and then experienced slab breakoff, similar to the northwestern Himalayan UHP rocks, resulting in buoyant ejection of the HP rocks into the hot middle crust (their models 2A and 2B). However, the Ama Drime rocks are not characterized by rapid and deep subduction (≥100 km), near-simultaneous peak temperatures and peak pressures, rapid cooling and exhumation following peak pressure, and juxtaposition along or near the orogenic suture like the UHP eclogites of the northwestern Himalaya (e.g., Parrish et al., 2006). Three key lines of evidence rather support the alternative end member of metamorphism during crustal thickening: (1) Maximum pressure did not necessarily exceed a typical thickness for orogenic crust (∼60 km) (Groppo et al., 2007); (2) peak temperature (M2–M3) postdates peak pressure (M1) by as much as 20 m.y., arguing against a rapid cycle of burial and exhumation; and (3) the rocks are not associated with the orogenic suture, as would be expected for rocks attached to a subducting slab. We conclude that the Ama Drime eclogites are best interpreted as having formed as part of an orogenic root during Eocene thickening and underthrusting of Indian crust. This implies that the Himalayan orogen had already achieved a crustal thickness of at least ∼60 km and an eclogitized lower crust by the late Eocene.

This result and interpretation can be directly compared with geodynamic models of the evolution of the Himalaya. There are several aspects of the rocks studied here that agree with predictions of the HT1 channel flow model, a geodynamic model of continental collision with parameters representative of the Himalayan orogen (Jamieson et al., 2004). The HT1 model predicts that by 21 m.y. after the collision is initiated (equivalent to 33 Ma using the HT1 assumption that collision initiated at 54 Ma; Jamieson et al., 2004), much of the lowermost crust is at eclogite-facies metamorphic conditions. It further predicts that high-pressure conditions may be maintained in lower-crustal rocks for over 20 m.y., that the timing of high-pressure metamorphism should be diachronous within the orogen as material is continuously transferred into the lower crust, that granulite-facies conditions will follow the long residence time at eclogite-facies conditions, and that granulite-facies metamorphism will involve near-isothermal decompression, in this case as a result of entrainment into an orogenic channel. Finally, since none of the model points in HT1 reaches the orogenic front by model end, Jamieson et al. (2004) predicted that granulitized eclogite now exposed at the surface in the Himalaya may have been sitting deeper and closer to the suture than the tracked model points of HT1, such that an even older age (than 33 Ma) for eclogite-facies metamorphism may be expected in nature than is predicted in the model. These similarities provide a mechanically feasible explanation for the burial and initial exhumation of the Ama Drime granulitized eclogites, with major N-S–striking normal faults that bound Ama Drime likely aiding final exhumation. As yet, it is unclear whether the rocks sat at eclogite-facies conditions for the duration of the >20 m.y. between prograde garnet growth and granulite-facies metamorphism, or whether they were exhumed and cooled before reheating to granulite-facies conditions at ca. 14 Ma. Further detail from the pressure-temperature path for the Ama Drime rocks, specifically focusing on the period between eclogite-facies metamorphism at ca. 38 Ma and granulite-facies metamorphism at ca. 14 Ma, would resolve this uncertainty and enable further testing of the applicability of the HT1 model to this part of the Himalaya.

Prograde HP metamorphism of Indian plate rocks at Ama Drime occurred at ca. 38 Ma, the oldest reported eclogite-facies metamorphic event in the central and eastern Himalaya. This metamorphism is interpreted to represent burial of rocks at the base of the crustal section during thickening of the Indian plate during collision with Asia. Unlike UHP eclogites near the western syntaxis of the Himalaya, the pressure-temperature-time history of the Ama Drime HP rocks does not reflect subduction, detachment from the downgoing plate, and rapid exhumation. Instead, these rocks reflect late Eocene crustal thickening to at least ∼60 km, followed by a more than 20 m.y. residence time in the lower to middle crust, before middle Miocene (ca. 15–13 Ma) granulite-facies metamorphism and rapid exhumation.

This research is based upon work supported by the National Science Foundation under grant EAR-1119380 awarded to J. Cottle and D. Kellett, with additional financial support from the Deutsche Forschungsgemeinschaft (grant SM 308/1–1) to M.A. Smit. Thanks go to A.R.C. Kylander-Clark, G. Seward, J.C. Vrijmoed, and J. Laforge for technical assistance. This manuscript is Geological Survey of Canada contribution 20120286.