Age and anatomy of the Gongga Shan batholith, eastern Tibetan Plateau, and its relationship to the active Xianshui-he fault

The Tibetan Plateau (Fig. 1) is the world’s largest area of high elevation (∼5 km average) and thick crust (70–85 km thick), and the timing of its rise is important for tectonics and for understanding the influence of topography on climate and the erosional flux of sedimentary detritus into rivers and oceans. Geological evidence for crustal thickening and topographic uplift includes the timing of compressional deformation, regional metamorphism, and magmatism. Earlier notions of a late Cenozoic thickening and rise of the Tibetan Plateau (e.g., Molnar et al., 1993) following the early Eocene collision of India with Asia and the closing of the intervening Neotethys Ocean at 50.5 Ma (Green et al., 2008) have been challenged by recent geological investigations. In Chung et al. (2005), Kapp et al. (2007a, 2007b), and Searle et al. (2011), the widespread occurrence of Andean-type subduction-related Late Jurassic–early Eocene granites (Ladakh-Gangdese batholith) and calc-alkaline volcanic rocks across the Lhasa block was noted, strongly suggesting an Andean-type topography with similar crustal thickness during the period prior to the India-Asia collision. Soon after the India-Asia collision, calc-alkaline Gangdese-type magmatism ended (St-Onge et al., 2010) and volumetrically minor, sporadic, but widespread adakitic magmatism occurred across the plateau from 50 Ma (Chung et al., 2005).

The present-day structure of eastern Tibet is characterized by a high, flat plateau, with a few exceptional topographic anomalies, such as the Gongga Shan (7556 m) massif, and a steep eastern margin along the Longmen Shan, showing an abrupt shallowing of the Moho from depths of 60–80 km beneath the plateau to depths of 35–40 km beneath the Sichuan Basin (Z. Zhang et al., 2010). There is an almost complete lack of Cenozoic shortening structures, with the exception of steep west-dipping faults associated with the M7.9 Wenchuan earthquake along the Longmen Shan margin (Hubbard and Shaw, 2009). Most of the deformation in eastern Tibet is Indosinian (Triassic–Jurassic) in age (Harrowfield and Wilson, 2005; Wilson et al., 2006; Roger et al., 2003, 2004, 2010). The thick Triassic Songpan-Ganze flysch sedimentary rocks are tightly folded about upright fold axes and are above a major horizontal detachment above Paleozoic basement (Harrowfield and Wilson, 2005). Most granites that have been dated intruded during the period 220–188 Ma (Roger et al., 2004, 2010). A complete Barrovian-type metamorphic sequence is present in the structurally deeper Danba dome, where peak sillimanite-grade metamorphism has been dated as 179.4 ± 1.6 Ma using in situ U-Pb monazite analysis (Weller et al., 2013). There is no record of any Cenozoic metamorphism anywhere in north, east, or central Tibet. The old ages of deformation, metamorphism, and magmatism argue strongly against the Miocene–recent homogeneous crustal shortening models for Tibet (Dewey and Burke, 1973; Houseman et al., 1981).

Cenozoic structures in central and eastern Tibet are represented by large-scale strike-slip faults (Fig. 2). The Ganzi (Yushu) and Xianshui-he left-lateral strike-slip faults cut across all the geology of the eastern plateau and have diverted river courses in the upper Yangtse and Jinsha river systems. These faults curve around the eastern Himalayan syntaxis and are thought to be responsible for southeastward extrusion of the south Tibetan crust (Molnar and Tapponnier, 1975; Peltzer and Tapponnier, 1988; Peltzer et al., 1989; Tapponnier et al., 2001) and clockwise rotations caused by the northward indentation of India (England and Molnar, 1990).

Along the Xianshui-he fault a granitic batholith, the Gongga Shan massif, crops out for ∼200 km along the southwestern margin (Fig. 3). This forms a major topographic high, with the Gongga Shan massif reaching 7756 m. Roger et al. (1995) reported a granite emplacement U-Pb zircon age of 12.8 ± 1.4 Ma for one sample from the western margin of the batholith along the Kangding-Yadjang road; they also reported Rb/Sr cooling ages of 11.6 ± 0.4 Ma from a deformed granite and 12.8 ± 1.4 Ma and 9.9 ± 1.6 Ma from an undeformed granite. Liu et al. (2006) sampled the same granite and obtained a SHRIMP (sensitive high-resolution ion microprobe) U-Pb age of 18.0 ± 0.3 Ma. Li and Zhang (2013) reported SHRIMP U-Pb zircon ages of ca. 31.8 Ma and ca. 26.9 Ma for a leucosome and melanosome collected from the eastern margin near Kangding and ages of 17.4 Ma and 14.4 Ma from the main granite pluton; they interpreted the ages as resulting from a stage of metamorphism and migmatization at 32–27 Ma and magma intrusion at 18–12 Ma. The Gongga Shan granites are some of the rare examples of Cenozoic crustal melts exposed on the Tibetan Plateau, and therefore a study of their ages and origin is important in connection with the proposed crustal models, particularly the whole-scale underthrusting model of Argand (1924) and the lower crust flow model of Royden et al. (1997) and Clark and Royden (2000).

We studied three main transects across the Gongga Shan batholith along the Kangding, Yanzigou, and Hailugou valleys (Fig. 4) and collected samples for petrology and U-Pb geochronology. In this paper we first summarize the geology of eastern Tibet and discuss geophysical constraints on the structure of the deep crust, notably evidence for the timing of crustal thickening and Cenozoic crustal melting. We describe the Xianshui-he fault and present new field observations for the Gongga Shan batholith that provide constraints on the emplacement history of different plutonic phases, for which we provide new in situ U-Pb zircon and allanite age data. We use the timing constraints on the granite batholith to infer the age of initiation, and offset along, the Xianshui-he fault, and use our new data on the Gongga Shan granites and the Xianshui-he fault to discuss the models for the tectonic evolution of the Tibetan Plateau.

Widespread Neoproterozoic granitoids and gneisses record the cratonization of the Yangtse block across south China. The Kangding complex that crops out along the eastern margin of the Xianshui-he fault has SHRIMP U-Pb zircon ages of 797 ± 10 Ma to 795 ± 13 Ma, roughly contemporaneous with several other gneiss complexes along the western margin of the Yangtse block (Zhou et al., 2002). Rifting led to the opening of the Paleo-Tethys Ocean across North Asia and the initiation of several subduction zone systems along the KunLun-Anyemaqen terrane to the north (220–200 Ma granite batholiths), the Jinsha suture to the south, and the Yushu-Batang zone to the east (Roger et al., 2004, 2010). During the middle Permian the rifting of the Neotethys to the south isolated the Qiangtang block, which became one of the Cimmerian microcontinental plates within the Tethys (Sengör, 1985). The huge Emeishan continental flood basalt eruptions ca. 260 Ma led to extensional rifting and formation of the Songpan-Ganze basin (Xu et al., 2001; Song et al., 2004. The Songpan-Ganze terrane shows a thick sequence of Triassic sedimentary rocks in this branch of the Paleotethys. Convergence between the Qiangtang terrane and the north and south China blocks led to closure of the Songpan-Ganze ocean and the major Late Triassic–Early Jurassic Indosinian orogeny. This is evidenced by large-scale regional upper crustal shortening across the Songpan-Ganze terrane (Harrowfield and Wilson, 2005) and lower crustal regional metamorphism recorded from the Danba structural culmination (Weller et al., 2013). Eastern Tibet has been elevated above sea level since the Late Triassic–Jurassic, and there is no evidence of any later tectonic events until the late Cenozoic.

In eastern Tibet crustal thickness beneath the Songpan-Ganze and KunLun terranes is ∼70 km, shallowing to ∼54 km beneath the Tsaidam (Qaidam) basin to the north (Mechie and Kind, 2013). Moho depths decrease from ∼60 km beneath the eastern part of the plateau to 40–36 km beneath the Sichuan Basin (Zhang, M., et al., 2010). This corresponds to a very steep topographic boundary along the Longmen Shan, the eastern border of the plateau. The Qiangtang terrane crust south and west of the Xianshui-he fault is a zone of anomalous low velocities (<3.3 km/s), strong radial anisotropy (Huang et al., 2010), high Poisson’s ratios, and high electrical conductivity in the middle crust (Bai et al., 2010). These data imply that the lower 10–15 km of crust in eastern Tibet is strong and has no flow characteristics, whereas the middle crust is weak and may have interconnected fluids promoting some form of flow (Liu et al., 2014). The Songpan-Ganze terrane north of the Xianshui-he fault is a zone of higher crustal viscosity with less, or no, fluids in the middle crust. This area is characterized by regional Triassic–Jurassic Indosinian metamorphism (Weller et al., 2013) and has no evidence of Cenozoic crustal shortening, folding, or metamorphism.

The eastern margin of the Tibetan Plateau is marked by the Longmen Shan range, which shows steep west-dipping thrust faults, like that that ruptured during the 2008 Wenchuan earthquake (Hubbard and Shaw, 2009), exhuming old Precambrian and Paleozoic rocks along the hanging wall (Baoxing and Pengguan massifs). The eastern margin of Tibet is completely different from the southern Himalayan margin. In the Longmen Shan there is no regional Cenozoic metamorphism or crustal melting as seen along the Himalaya, there is no Main Central thrust equivalent structure, no South Tibetan detachment–type structure and thus no evidence for any kind of channel flow (Searle et al., 2011). A model of eastward-directed flow of the lower crust from beneath the Tibetan Plateau was proposed by Royden et al. (1997) and Clark and Royden (2000) to explain the tectonic features of eastern Tibet and Yunnan. Global positioning system (GPS) data record the eastward-directed motion of the upper crust of east Tibet with motion appearing to flow around the stable Sichuan Basin (Gan et al., 2007). However, these GPS data record present-day motion of the surface, not deep crust motion. However, recent seismic data support a strong lowermost crust under all of Tibet (Lhasa and Qiangtang terranes) and a weak middle crust (Tilmann et al., 2003; Mechie and Kind, 2013). Unlike the partially molten middle crust of southern Tibet, which is laterally connected with the Cenozoic metamorphism along the Himalaya, there is no such boundary along the eastern margin of Tibet in the Longmen Shan. The lower crustal flow models for eastern Tibet proposed by Royden et al. (1997) and Clark and Royden (2000) are not grounded in geological observations or data. If the lower crust flowed around the stable Sichuan block, this is not reflected in the mapped surface bedrock geology and there is no evidence to support contemporary upper crustal shortening in east Tibet, or documentation in the geology of Sichuan or Yunnan for Tibetan rocks flowing to the east and southeast.

There is some geological evidence for small-scale post–50 Ma localized partial melting of the Tibetan lower crust (forming adakites; Chung et al., 1998, 2003, 2005; Wang et al., 2010) and middle crust (forming leucogranites and rhyolites; Molnar et al., 1987; Wang et al., 2012) as well as some young (to ca. 8 Ma) leucogranitic magmatism along the NyenchenTanggla (Liu et al., 2004; Kapp et al., 2005; Weller et al., 2016). Although there is some geophysical (seismic and magnetotelluric data) evidence for present-day low-degree partial melting, there is no evidence for widespread mid-crust sillimanite-grade migmatites and leucogranites such as are along the Greater Himalaya (mid-crustal channel flow).

Strong evidence for Indosinian (Late Triassic–Jurassic) crustal thickening in eastern Tibet comes from the regional folding and thrusting in Songpan Ganze sedimentary units. Late Triassic sedimentary rocks between 5 and 15 km thick show tight upright folds and penetrative cleavage above a major regional detachment separating low-grade or unmetamorphosed sedimentary rocks above from Proterozoic basement rocks beneath (Harrowfield and Wilson, 2005). A lack of post-Triassic sedimentary rocks on the east Tibetan Plateau suggest that it may have been topographically above sea level since then. Late Triassic crustal thickening led to regional Barrovian metamorphism, which is exposed in the exhumed Danba antiform, where amphibolite facies kyanite and sillimanite-grade metamorphism has been dated by U-Pb monazite as 192–180 Ma (Weller et al., 2013).

There is evidence for ca. 200 Ma regional metamorphism in southern Tibet (Weller et al., 2015) followed by regional magmatic crustal thickening along southern Tibet (Gangdese ranges) during the Late Jurassic to early Eocene (ca. 188–45 Ma; Chu et al., 2006; Wen et al., 2008; Chiu et al., 2009; Chung et al., 2009). Intrusion of the Gangdese I–type subduction-related calc-alkaline batholith and eruption of andesites, dacites, and rhyolites occurred along the 2000 km length of the Trans-Himalayan batholith (Kohistan, Ladakh, and Gangdese granites and extrusives). The geology of these ranges suggests a topographic uplift during the Jurassic to early Eocene, similar to the modern-day Peru-Bolivian Andes. An intense magmatic flare-up ca. 50 Ma occurred along the Gangdese batholith; volcanic compositions ranged from calc-alkaline to shoshonitic and adakitic (Lee et al., 2009). Shoshonitic volcanics imply a deep, hot mantle and lower crust–derived adakites imply a thick continental crust. Abundant adakitic melts requiring a garnet-bearing amphibolite or eclogite lower crust source across Tibet occurred from 47 Ma in the Qiangtang terrane and from at least 30 Ma in the Lhasa terrane (Chung et al., 2005), implying that the whole of Tibet was crustally thickening and topographically high since the early-middle Eocene (Searle et al., 2011). Lower crustal felsic and mafic granulite xenoliths entrained in Cenozoic ultrapotassic shoshonites from the Lhasa and Qiangtang terranes also conclusively show that extreme crustal thickness must have been present across Tibet during the Miocene (Hacker et al., 2000; Chan et al., 2009).

Evidence for widespread plateau formation during the Late Cretaceous–Paleogene (pre–45 Ma) also comes from regional ⁴⁰Ar/³⁹Ar, fission track, and [U-Th]/He data (Kirby et al., 2002; Hetzel et al., 2002; Rohrmann et al., 2011), which show that large regions of the plateau underwent cooling and exhumation prior to 45 Ma coeval with as much as 50% upper crustal shortening (Kapp et al., 2005, 2007). The Eocene low-relief plateau surface shows that the plateau was high and dry with very little erosion, similar to the present-day situation, from 45 Ma (Hetzel et al., 2011). Significant high topography existed across the entire Tibetan Plateau before the India-Asia collision, with pulses of rapid exhumation at 30–25 Ma and 15–10 Ma recorded by low-temperature thermochronology).

There is no geological or geochronological evidence to suggest that the Tibetan Plateau was uplifted only in the past 7–8 m.y., as suggested by various lines of circumstantial evidence (e.g., Molnar et al., 1993). It may have undergone an increase in elevation during late Cenozoic time, but all lines of evidence point to the fact that the entire plateau was elevated above sea level since mid-Cretaceous time, attained at least Andean (Bolivian Altiplano) type elevations during the Cretaceous–Eocene, and was as thick as present day (∼65–75 km) during the Miocene, probably since the early Eocene.

The combined Ganzi-Yushu and Xianshui-he faults extend for ∼1200 km length from the central part of the Tibetan Plateau, curving around toward the southeast and ending in a series of splays in western Yunnan, north of the Red River fault (Burchfiel et al., 1989; Wilson et al., 2006; Wang et al., 2014). The Xianshui-he fault is one of the most seismically active strike-slip faults of the Tibetan Plateau region, with 9 major earthquakes of M7–M7.9 between A.D. 1725 and 1983 (Wang et al., 1998). Focal depths of earthquakes are as deep as 20 km, suggesting active sinistral slip in the upper crust. Active slip rates were estimated as 15 ± 5 mm/yr from dating fault offset material (Allen et al., 1991) and geodetic estimates of the current slip rate are estimated as 9–12 mm/yr from InSAR) (interferometric synthetic aperture radar; Wang et al., 2009). Measured GPS slip rates suggest that present-day slip may be ∼10–12 mm/yr (Zhang et al., 2004).

The main Xianshui-he fault extends along the eastern margin of the Gongga Shan massif, but a series of en echelon sinistral faults cut across the batholith, forming a classic left-lateral strike-slip duplex system (Fig. 4). From Chinese maps these fault strands each show only minor offsets of the granite margin between 5 and 10 km. One major fault strand cutting through Triassic metasediments to the west of the Gongga Shan shows a spectacular gouge zone that can be traced for >200 km from Barmie to Ganzi (Wilson et al., 2006). North of Kangding the Xianshui-he fault shows ductile fabrics as well as later brittle gouge zones. The ductile shearing fabrics die out to the west away from the fault, indicating that both ductile and brittle movement along the fault postdated granite emplacement. Around Kangding the fault shows a transpressional uplifted western margin (Gongga Shan granite) with a >2 km topographic difference from the batholith west of the fault to the Kangding Neoproterozoic complex east of the fault. South of Kangding the Xianshui-he fault cuts through metasediments and Proterozoic gneisses east of the batholith. Two major northwest-southeast–aligned fault splays cut the granite batholith, and field relationships clearly indicate that faulting came after granite emplacement (Fig. 4). Toward the Moxi township the fault cuts through Paleozoic metasedimentary rocks ∼12 km to the east of the eastern intrusive margin of the Gongga Shan granite batholith. The trace of the Xianshui-he fault trends south toward Kunming, where again it splays into several different strands aligned at right angle to the northwest-southeast–aligned Ailao Shan–Red River shear zone (Burchfiel and Chen, 2010).

Total left-lateral displacement has been suggested as ∼50–60 km based on dubious pinning points (e.g., Precambrian–Proterozoic unconformity on Chinese maps, and older faults that may not originally have been the same structure). From Chinese maps, offsets of the Gongga Shan granites may be only ∼15 km along the western margin of the batholith and to 60 km along the eastern margin. We report our new findings from the Gongga Shan batholith of eastern Tibet in the following.

The Gongga Shan batholith extends for more than 100 km in an arcuate trend across central east Tibet and is ∼10–15 km wide (Fig. 4). The batholith is cut by strands of the left-lateral Garze-Yushu and Xianshui-he strike-slip faults, which extend for >800 km from the central Tibetan Plateau east and southeast into Yunnan. The faults cut across regional geology and clearly offset the Gongga Shan granites, although the precise amount of offset is difficult to determine accurately (Roger et al., 1995). We studied three major valleys transecting the Gongga Shan batholith: the main Kangding road, the Yanzigou valley, which cuts westward into the batholith just north of the Gongga Shan (7556 m), and the Hailuogou valley, which extends west of Moxi town to the high peaks south of the Gongga Shan.

The main road from Barmie to Kangding cuts across the central part of the Gongga Shan batholith. The western part of the batholith contains two main phases of granite, an earlier granodiorite and a later biotite granite. Sample BO-59 (Fig. 5A) was collected from a location close to the sample collected by Roger et al. (1995) that they dated as 12.8 ± 1.4 Ma. The middle part of the batholith comprises a lithology: K-feldspar + quartz + plagioclase + biotite monzogranite that forms most of the batholith (sample BO-62; Fig. 5B). Later minor intrusions of biotite + tourmaline pegmatites and secondary muscovite ± garnet granite veins intrude the main phase. At the Shuguang bridge locality in the middle of the Kangding road section, one outcrop shows three distinct crosscutting relationships (Fig. 5C). An earlier biotite monzogranite (BO-57) has a weak foliation and was intruded by a second phase of more leucocratic biotite ± muscovite granite (BO-52) with migmatitic textures (schlieren of older melanosomes). Both lithologies are cut by a later undeformed pegmatite (BO-55). In places a younger phase of garnet leucogranite has intruded all previous lithologies (Fig. 5D) and complex intrusive relations have been mapped along the northeastern margin of the batholith north of Kangding (Fig. 5E). The youngest phase of intrusion in this transect is a fine-grained undeformed biotite microgranite (Fig. 5F) that cuts all previous lithologies.

The western margin of the Gongga Shan batholith appears to be a vertical intrusive contact (Fig. 6A). The granite along the western margin above the new Kangding airport south to the Zheduo Shan pass shows little or no fabric, and there does not appear to be a contact aureole in the country rock (Triassic shales). The granite along the eastern margin has a vertical contact, cutting both fabrics in the country rocks to the east (Kangding complex) and internal fabrics within the granite. Four phases of granite intrusions have been mapped at Kangding, from early biotite monzogranite to late garnet + biotite leucogranite cut by pegmatite dikes (Fig. 6B). Ductile foliations within the granites strike ∼028° north-northeast oblique to and truncated by the strike of the Xianshui-he fault. South of Kangding and across the Baihaizi pass the Conch gully area shows a mixture of biotite + hornblende granodiorites with igneous enclaves (Fig. 6C), magmatic mixtures of more enclave-rich granite intermingled with the granodiorite (Fig. 6D), and more evolved garnet leucogranite (Figs. 6E, 6F).

The Yanzigou valley cuts west directly into the heart of the granite batholith immediately to the north of Gongga Shan. East-verging recumbent folds in probable Triassic metasediments (Fig. 7A) are truncated abruptly at the margin of the batholith (Fig. 7B). The oldest intrusions in the area are series of foliated biotite + hornblende granodiorites (BO-76). Toward the west in the Swallow Cliffs area and around the snout of the north Gongga Shan, glacier leucogranites (BO-68) appear to be increasingly common, with evidence of partial melting in metasedimentary migmatites. The leucogranites are always the younger intrusive phase, intruding into and breaking up enclaves of the more mafic granites (Figs. 7C, 7D). Multiple phases of granitoids with clear crosscutting relationships are present in spectacular outcrops in the middle part of the batholith (Figs. 7E, 7F, and 8). The eastern margin of the batholith shows a prominent fabric striking 160° northwest-southeast and dipping 50° northeast. The fabric in the gneisses is abruptly truncated by the granite margin. The Xianshui-he fault in this profile is 11–12 km to the east of the granite contact and cuts through gneisses of the Kangding complex. The field relationships show that the granite batholith was not related to the strike-slip fault in any way.

The Hailuogou valley extends west of Moxi town into the highest peaks around the Gongga Shan (7556 m; Fig. 9A). Glaciers have carved a deep gorge into the high country around the western part of the batholith, but the Gongga Shan is snow covered and somewhat inaccessible. The granite contact along the eastern margin is vertical and clearly exposed along the northern rim of the Hailuogou valley (Fig. 9B). Foliations in the gneisses are near vertical at the contact but folded further to the east. Above the cable car station on the Gongga Shan glacier, the dominant lithology is a hornblende + biotite diorite with K-feldspar megacrystic granite also containing hornblende and biotite (Fig. 9C). Igneous diorite enclaves are common in the granite (Fig. 9D). Boulders in the glacier and streams above the cable car station suggest that the Gongga Shan peak is composed of granodiorite. There is no sedimentary or country rock talus, implying that the western margin of the batholith is west of the Gongga Shan summit. There is little evidence along the Hailuogou valley profile of the more garnet-bearing leucogranite or migmatite phases common in the Yanzigou valley to the north.

Zircon, allanite, and titanite were separated from seven granitic samples by standard techniques (crushing, milling, Rogers table, Frantz magnetic separation, and heavy liquids). Grains were picked under alcohol and mounted in 1 in (∼2.5 cm) epoxy mounts. In order to guide analytical spot placement, zircon grains were imaged using cathodoluminesce (CL). Backscattered electron (BSE) imaging was conducted for allanite and titanite grains, but revealed no systematic zoning.

U-Pb geochronology utilized a Nu Instruments Attom single collector–inductively coupled plasma mass–spectrometer (SC-ICP-MS) coupled to a New Wave Research 193FX Excimer laser ablation system with an in-house teardrop-shaped cell (Horstwood et al., 2003). The full method was described in Spencer et al. (2014). In brief, ablation used static spots ranging from 20 to 40 μm depending on the size of the growth zoning that the sample allowed. Ablation parameters were 5Hz at ∼2 J/cm² for 30 s, with 10 s wash-out. Standard sample bracketing utilized the average drift-corrected ratios for 91500 (Wiedenbeck et al., 1995), or GJ-1 (Jackson et al., 2004) and Plešovice (Sláma et al., 2008) for normalization of zircon; 40010 (Smye et al., 2014) for allanite; and Ontario-2 for titanite (Spencer et al., 2013). Data were reduced using an in-house spreadsheet, and Isoplot (Ludwig, 2003) was used for age calculation. Allanite age data were corrected for excess ²⁰⁶Pb due to initial ²³⁰Th disequilibrium, based on a whole-rock Th/U ratio of 3. This ratio is arbitrary and not sample based; therefore, young (younger than 50 Ma) ages may be biased toward older by several percent (Smye et al., 2014). Intercept ages for allanite (except the free regression of BO-76) use an upper intercept based on an assumed common lead component of 0.83 ± 0.2 (Stacey and Kramers, 1975).

All analyses are plotted and ages are quoted at 2σ. Imprecise ages based on intercepts, due to lead loss or mixing, or those with excess scatter based on MSWD (mean square of weighted deviates) values (and presumably due to geological variation such as inheritance and lead loss), are quoted as ca. xx Ma. Precise ages given with uncertainties, interpreted as individual crystallization events, are quoted as age ± α/β Ma, where α refers to the measurement and session-based uncertainty, and β is the total uncertainty after propagation of systematic uncertainties (decay constants, reference material age uncertainty, long-term reproducibility of the laboratory method; see Horstwood et al., 2016).

Sample BO-52 gave a moderate zircon yield. Grains were a mixture of sizes, mostly elongate, typically from 100 to 200 µm. The majority of grains show oscillatory zoning in CL, often with two apparent phases of growth; some have brighter cores and darker rims, and some have dark and recrystallized cores. Some grains have a thin rim, and some are fractured.

Analyses (n = 40) yield a spread of ages from 112 to 179 Ma, and 2 core analyses gave older ages of ca. 235 Ma (Fig. 10). Two of the younger analyses gave ages of ca. 35 Ma, and a third is slightly older, 51 Ma; these are from a CL dark overgrowth and recrystallized zones. There is no obvious single population from within the older ages, although the average of these is ca. 167 Ma (using TuffZirc; Ludwig, 2003). Overgrowths and thick rims on zircon are not exclusively related to younger ages; they extend to 167 Ma. Only a few allanite analyses were obtained, and these exhibit some scatter. A tight cluster of analyses gives a regression with an age of ca. 173 Ma; this is within uncertainty of the oldest of the zircon age populations. A second array through analyses with a high radiogenic component gives an age of ca. 16 Ma, suggesting a younger allanite crystallization or resetting event. The interpretation of the geochronological data is equivocal; however, we interpret the zircon and allanite data collectively as recording migmatization ca. 16 Ma.

Sample BO-55 gave a low zircon yield. The grains are typically 100–200 µm long, and some are smaller. In CL, oscillatory zoning is exhibited, and this has been disturbed in some grains. Some grains are fractured, and some show evidence of recrystallization. Thin rim overgrowths are not obvious, but a few grains show thicker zoned overgrowths that cut across internal zones.

The four oldest analyses overlap at 159 ± 2/4 Ma (MSWD = 0.34) (Fig. 11). A younger population of analyses overlaps ca. 41 Ma. Three analyses overlap at 37 Ma, and may represent another distinct age population. Twelve further analyses spread from 22 to 15 Ma, and may represent a single discordia that is recording lead loss from a ca. 41–37 Ma event, or from a ca. 159 Ma event. The youngest date recorded by a lead-loss trend is ca. 15 Ma. All of the younger populations were obtained on a mixture of recrystallized zones and zoned overgrowths, and there is no discernable change in Th/U ratio with age. Multiple overlapping ca. 40 Ma ages from recrystallized zones imply that this age reflects an actual crystallization event, and is not solely due to lead loss, as this would generally be variable within a single grain.

Allanite exhibits an array with high common lead content, and probable mixing between two events. The oldest analyses define an array with a lower intercept ca. 164 Ma; this is correlative to the oldest zircon ages. The youngest allanite grain records a lower intercept ca. 15 Ma, which can be interpreted as a maximum age for the younger allanite growth or resetting event. The age of pegmatite crystallization is interpreted to be ca. 15 Ma, with both older zircon and allanite ages reflecting inheritance from the protolith.

Sample BO-57 gave a moderate zircon yield. All grains are elongate igneous zoned zircons, 100–250 µm long, with planar or oscillatory zoning. There appears to be one main growth zone, although a few grains have brighter cores, and these are recrystallized in a few grains. Some outer rim regions exhibit resorption textures. Analyses (n = 32) define a population with an age of ca. 182 ± 1/4 Ma (MSWD = 1.4; Fig. 12). Two analyses are slightly older, 195 and 207 Ma, possibly reflecting inheritance, and one analysis is slightly younger at 166 Ma, probably representing lead loss. Three distinctly older grains give Proterozoic ages (795 Ma, 969 Ma, 2469 Ma). The crystallization age of the granite is interpreted to be 182 ± 4 Ma.

Sample BO-59 gave a very low zircon yield. There are mixed grain shapes and sizes (50–200 µm), all with relict igneous shape and zoning. There is planar and oscillatory zoning visible in most grains. Analyses (n = 7) form a subconcordant population at 166 ± 2/4 Ma (MSWD = 1.2) (Fig. 12). Two grains are older, ca. 206 Ma; these are not from obviously older cores, but still may represent inheritance.

Allanite exhibits an array with a high common lead content, and probable mixing between two events. A cluster of analyses at the oldest ages produces a regression with an age of ca. 166 Ma that is correlative to the oldest zircon population. Regression through the younger analysis gives a lower intercept of ca. 18 Ma, providing a maximum age for a young allanite growth or resetting. The age of granite crystallization is interpreted to be 166 ± 4 Ma, and ca. 18 Ma is interpreted as a tectonothermal event affecting this unit.

Sample BO-62 gave a moderate zircon yield (Fig. 13). There are mixed grain sizes and shapes from 50 to 200 µm. Most grains have oscillatory or planar igneous zoning, but this is disturbed or recrystallized in many cases. Inclusions are fairly common. Some grains have faint zoning that looks disturbed. Thick rim overgrowths occur on some grains. We obtained 137 analyses from 80 grains. The majority of the analyses are on a regression (mixing line) between ca. 800 Ma and a young (Neogene) lower intercept. Scrutiny of the younger ages and lower intercept reveals two distinct subconcordant populations. One is dated as ca. 14 Ma, obtained from 4 grains, and some with oscillatory zoning. Another is dated as ca. 6–5 Ma. Overlapping ages of ca. 6.3 Ma across one grain, 6.7 Ma across another, and 5.5–5 Ma in others, implies prolonged or multiple crystallization events. Discordance in many analyses can be attributed to a high common lead content for many of the young analyses. The youngest age population, with concordant (>90%) analyses ranging from 6.7 to 5.0 Ma, are derived from recrystallized zones, embayed but oscillatory zoned rims and overgrowths, and one primary oscillatory zoned crystal. In addition, multiple overlapping ages have been determined from single crystals. Collectively, the data imply that the young ages are attributable to crystallization in a melt, rather than being a result of age disturbance or resetting.

One allanite analysis with moderate radiogenic lead content gives an age of ca. 170 Ma (with an assumed common lead composition). Because this may be disturbed, no particular emphasis should be placed on this age; however, it is interesting to note that it correlates with zircon and allanite ages from other samples. A population of allanite analyses with very high common lead content define a poor regression to a young age; the lower intercept is 5 ± 1/1 Ma. This age and uncertainty should be given low confidence, because there is very little radiogenic component; however, it is also interesting to note that it correlates with the youngest zircon age domains ca. 5 Ma.

Sample BO-68 gave a moderate zircon yield. The sample comprises elongate 100–300-µm-long igneous grains with broken tips. All are oscillatory or planar zoned, but many are also recrystallized along internal zones. It appears that there is one growth phase, with no obvious rim overgrowths. A couple of presumably inherited grains give ages of ca. 531 and 685 Ma. The rest of the analyses cluster around a possible single population, with lead loss recorded in a few grains, and a couple of analyses possibly representing slightly older inheritance. Subconcordant analyses (n = 20) give an age of 204 ± 2/4 Ma (MSWD = 1.4), which is interpreted as the age of granite crystallization (Fig. 14). One concordant analysis at 177 Ma is from a bright rim, suggesting that a younger tectonothermal event at this age is recorded in this sample.

Four allanite analyses give an age of ca. 174 Ma; this correlates with the single rim age of 177 Ma, adding confidence to the interpretation of a separate event younger than 205 Ma.

Sample BO-76 gave a moderate zircon yield, with 100–300-µm-long elongate grains. Zoning is oscillatory, planar or sector, and is typically quite weak in contrast (in CL). Inclusions and fractures are common, and some grains appear to have resorption features. Rim-type overgrowths are not apparent, but some outer zones may be recrystallized. Despite the appearance of possible different growth zones, 32 analyses define a single population at 215 ± 2/5 Ma (MSWD = 1.3), which is interpreted as the age of granite crystallization (Fig. 15). The only exclusion is a single older, and presumably inherited, grain (discordant ²⁰⁷Pb/²⁰⁶Pb age of 790 Ma). A population of moderately radiogenic allanite analyses gives an apparent regression, with minimal scatter, of ca. 204 Ma. A population of moderately radiogenic titanite analyses gives a regression with an overlapping age of ca. 206 Ma; three analyses are younger, and may reflect open-system disturbance, lead loss, or younger titanite growth.

The age data from the seven samples described constrain two striking tectonic-plutonic events: a complex Triassic–Jurassic (Indosinian) plutonic record, and a possible Eocene to Miocene record of monzonite-leucogranite emplacement (Fig. 16). The Indosinian events are recorded in six of the seven samples. Discrete zircon populations in five of these samples, including those interpreted as inherited, probably document a series of igneous crystallization events ca. 215, 204, 182, 166, and 159 Ma. For those samples that have Indosinian crystallization ages, inherited zircons are few, with ages ranging from ca. 2469 to 200 Ma, but with the dominant subset being Neoproterozoic. Indosinian-aged inheritance, for example 206 Ma grains in a 166 Ma monzogranite, suggests reworking of young material during the complex Indosinian evolution. Allanite ages that are younger than the main zircon population, but also of Indosinian age, imply that younger Indosinian tectonothermal events or intrusions affected older Indosinian magmatic rocks. One sample (BO-62) lacks any Indosinian record in its zircon age data, but includes many ca. 800 Ma analyses, mostly from zircon cores. The lack of inherited zircons in most samples precludes a confident derivation of the magmatic source based on age characteristics. Neoproterozoic ages are common in the Triassic Songpan-Ganze sediments, which are likely source rocks for the (S-type) leucogranites. The ca. 800 Ma Kangding gneiss complex (Zhou et al., 2002) may have sourced the 800 Ma cores in BO-62, either directly (i.e., through melting), or indirectly via erosion of the complex and melting of the resulting sedimentary rocks.

Young melt events are recorded in three samples. BO-62 has zircon overgrowths on ca. 800 Ma cores, with concordant analyses implying two separate ages of new zircon growth, one ca. 14 Ma, and one ca. 5 Ma. A poor allanite regression giving a ca. 5 Ma age supports this younger age. Migmatization in one sample (BO-52) is presumed to be at least as young as the three youngest zircon ages, ca. 51 Ma and 35 Ma. Allanite ca. 16 Ma may represent the timing of migmatization. A crosscutting pegmatite (BO-55) has a small population of subconcordant ca. 41–37 Ma zircon analyses, suggesting a possible growth event at that time. Younger analyses are scattered, but imply pegmatite crystallization ca. 15 Ma, constrained by the youngest concordant analyses.

The U-Pb age data presented here lead to the following key question: how much young (i.e., Miocene) melting has occurred in the Gongga Shan batholith and is recorded in its composite tectonomagmatic history? Oscillatory zoned (in CL) overgrowths of variable thickness and with resorption textures suggest that both ca. 16–14 Ma and 5 Ma events involved significant melt at the grain scale. The sample that contains zircons of both these ages (BO-62) is an undeformed biotite monzogranite that, from our field observations and mapping, represents a volumetrically significant part of the batholith in its central region. This would imply that significant portions of the batholith are as young as 5 Ma. Although there is no outcrop evidence anywhere in east Tibet for any regional metamorphism or crustal thickening at that time, this may be a function of erosion and exposure level and it is possible that these rocks remain buried.

Previous U-Pb age constraints from different parts of the batholiths reveal ages from ca. 35 to 13 Ma (Roger et al., 1995; Li and Zhang, 2013). The 14 Ma age obtained here for zircon growth overlaps with these previous ages. Li and Zhang (2013) found ages of ca. 37–32 Ma in migmatitic rocks and interpreted them as a high-temperature event. The 41–37 Ma ages in BO-55 possibly record part of this same cryptic event.

The Gongga Shan massif is a composite granite batholith composed of three distinct parts. Much of the batholith appears to be of Indosinian origin. Triassic–Jurassic I-type granodiorite-biotite granite (215–159 Ma) formed in an Andean-type setting during closure of Paleotethys. Some volumetrically unknown component is composed of a garnet two-mica leucogranite of crustal melt origin but also of Indosinian age (204 ± 4 Ma). A third component is a multiple injection complex that exhibits ages from 41 to 15 Ma, with leucogranites and a pegmatite dike network that intruded between 15 and 5 Ma. Ductile fabrics within the granites and migmatites show flow folding related to middle or lower crust melting processes and thus are not related to the Xianshui-he strike-slip fault. Several samples show evidence of old Indosinian and young crustal melts in the same rock. The young leucogranites may have formed by remelting of Indosinian crust, possibly buried components of the Songpan-Ganze flysch. Because much of the batholith is inaccessible and remains undated, it is unclear precisely what proportion of the batholith is composed of young Miocene granite.

Only a few areas across Tibet show evidence for localized young crustal melts. Certainly the exposed geology of eastern Tibet consists of old, Precambrian basement complexes (e.g., Kangding complex), Paleozoic–early Mesozoic sedimentary and metasedimentary rocks, and Paleozoic–Triassic granites. The young granitic phase of the Gongga Shan batholith (Roger et al., 1995; Li and Zhang, 2013; our data) is the only evidence for late Miocene or younger crustal melting in this part of Tibet. Like the Himalayan leucogranites, these young S-type granites are derived by dominantly muscovite-dehydration reactions in pelitic protoliths in the middle crust. Similar young granites are reported from the Western Nyenchen Tanggla range of south Tibet, where U-Pb zircon ages of 25–8 Ma have been obtained (Liu et al., 2006; Kapp et al., 2005; Weller et al., 2016). These young melts are thought to represent an exhumed partial melt, similar to the small bright spots imaged in seismic and magnetotelluric studies (Brown et al., 1996; Nelson et al., 1996; Wei et al., 2001; Haines et al., 2003; Bai et al., 2010). We suggest that the young granite phases of the Gongga Shan may have a similar provenance. The limited amount of young crustal melt and the lack of regional Cenozoic metamorphism preclude any channel flow operating in eastern Tibet, unlike the south Tibet-Himalayan mid-crustal channel flow (Searle et al., 2010).

All the granites in the Gongga Shan batholith are cut abruptly by the left-lateral Xianshui-he fault. The northwest continuation of the Xianshui-he fault, called the Garze-Yushu fault, shows major offsets of tributary rivers of the Yangtse and has had numerous earthquakes in the past few hundred years, including the 2010 M_w (moment magnitude) 6.8 Yushu event (Li et al., 2011). The Jinsha River course appears to have been offset by ∼85 km and the course of the Yalong River offset by 35 km (Wang et al., 1998). The fault shows maximum sinistral offsets of ∼85 km in offset river courses of the Jinsha and upper Yangtze Rivers (Fig. 17). These fault strands each show only minor (5–10 km) offsets of the granite margin. A major transpressional component is present along the Kangding-Moxi segment, accounting for the 2–3 km of differential topography in the Gongga Shan region. The fault system also curves around from west-northwest–east-southeast (Ganzi-Yushu fault) to northwest-southeast (Xianshui-he fault) to north-south (Anninghe fault), showing almost 90° counterclockwise rotation of the Lhasa and Qiangtang blocks, associated with the northward indentation of India to the west. Our data suggest that the Xianshui-he fault was initiated after 5 Ma because it cuts granites of that age. It is theoretically possible that the fault has an earlier history, but there is no record of this in the present-day exposures. Ductile fabrics in parts of the injection complex along the eastern margin of the batholith are abruptly truncated by vertical brittle strike-slip fault strands along the Kangding road section. Farther south the Xianshui-he fault is ∼10 km or more east of the batholith. The granites are not connected to the strike-slip fault in any way and their ages relate to localized crustal melting, not to strike-slip faulting. Measured GPS slip rates suggest that present-day slip may be ∼10–12 mm/yr (Wang et al., 2001; Zhang et al., 2004). Despite being one of the most seismically active faults in Tibet, it shows only limited offset and therefore cannot have been responsible for large amounts of eastward lateral extrusion of the thickened crust.

Do the Gongga Shan granites represent melts derived from the lower crust of eastern Tibet? It is possible that the young crustal melts seen in Gongga Shan are representative of widespread young crustal melting, or that they are volumetrically minor melts sourced from the middle crust. The Longmen Shan and Lijiang faults mark a major transition from seismically fast (cratonic) mantle to the east (Sichuan Basin, southern Yunnan and Yangtze craton) and seismically low wave speeds to the west in the Tibetan Plateau (Liu et al., 2014). We suggest that this Longmen Shan–Lijiang fault system, in addition to bounding the topographically high plateau to the west, marks the eastern boundary of the indenting Indian plate lower crust. We suggest that the lower crust of the Lhasa and Qiangtang terranes are both underlain by strong Archean–Paleoproterozoic crust of India underthrusting the plateau north as far as the Xianshui-he fault. The crust to the southwest of the Xianshui-he fault shows low-velocity zones in the mid-crust, radial anisotropy and higher electrical conductivity consistent with localized pockets of melt (Huang et al., 2010; Bai et al., 2010). The crust north of the Xianshui-he fault is subtly different, showing higher average viscosity (Liu et al., 2014). The fact that Precambrian basement, Paleozoic cover rocks, and Indosinian metamorphic rocks (Danba area) are exposed at the surface in eastern Tibet above 65–70-km-thick crust, coupled with the lack of Cenozoic metamorphism or deformation, suggests that the Tibetan crust has been passively uplifted by underthrusting of Indian basement beneath (Argand, 1924; Searle et al., 2011).

Although some geophysical data point to a weak middle crust with pockets of interconnected fluids (Mechie and Kind, 2013), there is no geological evidence anywhere for lower crustal flow (Royden et al., 1997; Clark and Royden, 2000). Unlike south Tibet–Himalaya, the minor amount of fluids in the middle crust in eastern Tibet is not sufficient to form any sort of lateral channel-type flow, and there is no evidence along the Longmen Shan range of Cenozoic metamorphism or east-vergent ductile flow deformation. Along the Himalaya there is abundant evidence for mid-crustal channel flow during the Miocene [10–20 km thickness of partially molten middle crust migmatites and garnet, two-mica (±cordierite, andalusite, sillimanite) leucogranites bounded by a south-vergent thrust below and a north-dipping low-angle normal fault above; Searle et al., 2010], but along the eastern border of Tibet, the Longmen Shan, the geology and structure are completely different. Here, the deformation, metamorphism, and magmatism is almost entirely Triassic–Jurassic (Weller et al., 2013; our data) or older. There is no evidence for Cenozoic crustal shortening, and no evidence for east-directed thrusting or flow of any sort. If the lower crust beneath Tibet was flowing to the east, as suggested by Royden et al. (1997, 2008) and Clark and Royden (2000), then there should be a large amount of Cenozoic active shortening in the upper crust. There is none, only the steep west-dipping thrust faults associated with the M7.9 Wenchuan earthquake accommodating minor active shortening (Hubbard and Shaw, 2009). If lower crustal flow had occurred, the geology of southern Yunnan would be expected to show this. Instead the proposed flow directions cut across geological strike and structures throughout Yunnan. We propose, instead of outward flow, extensional tectonics, and lowering of surface elevation (Royden et al., 1997, 2008), that the plateau is maintaining or even increasing elevation, as evidenced by compressional tectonics along the southern margin (Himalaya–south Tibet) and compressional tectonics along the Longmen Shan (Wenchuan earthquake).

The more reasonable model to explain the geology of eastern Tibet is one of passive underthrusting of Indian lower crust toward the north-northeast all the way north as far as the Xianshui-he fault. At the time of India-Asia collision and closure of the Neotethys, ∼50 m.y. ago, the north Indian plate consisted of ∼7–8 km of Phanerozoic upper crustal sediments overlying 2–4 km thickness of Neoproterozoic low-grade sedimentary rocks (Haimanta and Cheka Groups), overlying 25–30 km of Archean–Paleoproterozoic granulite basement (Indian shield). The Himalaya are composed entirely of Neoproterozoic and younger rocks, mainly unmetamorphosed in the Tethyan Himalaya upper crust, and metamorphosed in the Greater Himalaya middle crust, structurally below the Tethyan Himalaya. At least 500 km, more probably 1000 km, of upper crustal shortening has occurred since collision in the Indian Himalaya. The Archean granulite lower crust that originally underlay these rocks was old, cold and dry, and thus unsubductable. The only place this Indian lower crust could have gone is northward, underthrusting the Asian plate beneath the Lhasa and Qiangtang blocks of Tibet in a process originally suggested by Argand (1924). We suggest that this underthrusted Indian lower crust has effectively doubled the Tibetan crust since 50 Ma and passively uplifted the rocks of the Tibetan Plateau. This alone can account for the almost complete lack of Cenozoic shortening and Cenozoic metamorphism across the Tibetan Plateau. Our late Miocene ages from the Gongga Shan granites are the only indication of post-collision Cenozoic metamorphism and magmatism in eastern Tibet, but are volumetrically minor.

We thank the Ministry of Science and Technology, Taiwan, and the National Natural Science Foundation of China for support and funding. Landsat-7 imagery and Shuttle Radar Topography Mission digital elevation models are courtesy of the National Aeronautics and Space Administration. Field work was funded mainly by the Institute of Geology, China Earthquake Administration (Beijing). Searle and Chung were partially supported by a Taiwan–UK Royal Society grant. U-Pb dating was carried out at the NERC (Natural Environment Research Council, UK) laboratory at the British Geological Survey, Keyworth. Elliott and Weller were supported by NERC grants. St-Onge was funded by the Earth Sciences Sector, Natural Resources Canada. We thank Nicole Rayner (GSC, Geological Survey of Canada), two anonymous reviewers, and associate editor Mike Taylor for detailed comments that significantly improved the manuscript. This is GSC Earth Sciences Sector contribution 20150332.