Late Cretaceous tectonothermal events of the Gangdese belt, southern Tibet

Orogenic belts are typically grouped into two broad categories: continent-continent collisional orogens and subduction-accretionary orogens (Uyeda and Kanamori, 1979; Stern, 2002; Grujic, 2006; Cawood and Buchan, 2007; Kellett et al., 2009). These two models are commonly end members, given that many orogens experience both continental collision and oceanic subduction during their evolution (Yin and Harrison, 2000; Jolivet et al., 2016; Kapp and DeCelles, 2019; van Hinsbergen et al., 2019). For example, the Tibetan-Himalayan orogen is typically considered a prototypical continent-continent collisional orogen and has received paramount attention within the geological community (Molnar et al., 1993; Tapponnier et al., 2001; Royden et al., 2008; Labrousse et al., 2010; Grujic et al., 2020; Carosi et al., 2013, 2018; Jamieson and Beaumont, 2013; Xu et al., 2013; Li et al., 2015a; Ding et al., 2016). However, during the Mesozoic, the southern margin of the Lhasa terrane was subject to a period of accretionary subduction prior to the Indo-Asian continental collision (Ratschbacher et al., 1992; Aitchison et al., 2000; Zhao et al., 2021), which may have resembled the presentday tectonic configuration of the South American Andes (Manea et al., 2017).

Subduction-accretionary orogens form curvilinear belts of magmatic and accreted material above sites of subducting oceanic lithosphere (Cawood and Buchan, 2007). Broadly speaking, subduction-accretionary orogens can also be sub-divided into two end-member categories: those in which the orogen is under compression and those in which it is under extension. These two end members have been termed “Chilean type” and “Mariana type,” respectively, by Uyeda and Kanamori (1979). “Chileantype” orogens tend to be characterized by widespread crustal shortening and uplift, including the development of fold-thrust systems in forearc and retro-arc regions, in conjunction with episodic arc magmatism (DeCelles et al., 2015; Ducea et al., 2015; Paterson and Ducea, 2015). Arc-continent accretion and arc magmatism are highly episodic, with magmatic activity commonly characterized by flareups and lulls (Attia et al., 2020; Ma et al., 2022). However, in the case of subduction-accretionary orogens, the driving mechanisms of formation are less obvious compared to continent-continent collisional orogens. Still debated are the crust-mantle–scale processes that drive the formation of accretionary orogens and the role of these processes in the episodic nature of deformation and magmatism (Paterson and Ducea, 2015).

The Lhasa terrane (southern Tibet), which was accreted to the southern margin of the Eurasian continent, not only records the final Indo-Asian collision processes but also preserves evidence for the subduction of the Neotethyan oceanic slab (Wen et al., 2008a; Kapp and DeCelles, 2019; Zhu et al., 2019; Ma et al., 2021a). However, much of the Mesozoic evolution remains poorly understood, in part due to structural and magmatic modification during the Cenozoic (Metcalf and Kapp, 2019; Sundell et al., 2021). The Gangdese arc magmatic belt, exposed along the southern margin of the Lhasa terrane (Fig. 1A), is inferred to have formed during the subduction of Neotethyan oceanic lithosphere beneath the Lhasa terrane, antecedent to the collision of the Indian and Asian continents at ca. 60–55 Ma (DeCelles et al., 2014; Hu et al., 2015).

The tectonic regime that prevailed during the Late Cretaceous along the southern margin of the Lhasa terrane is still debated. Some models suggest that Late Cretaceous to Paleogene shortening of the Lhasa terrane may have been punctuated by a 90–70 Ma phase of extension that led to the rifting of a southern portion of the Gangdese arc, with the opening of a back-arc ocean basin (Kapp and DeCelles, 2019; Sundell et al., 2021). Alternative models invoke Late Cretaceous crustal deformation in the back-arc and forearc basins of the Gangdese magmatic arc that is associated with the accretionary stage of the Tibetan-Himalayan orogeny (Kapp et al., 2007; Ma et al., 2017a). In other words, the Gangdese belt may have experienced a “Chilean-type” orogeny during the Late Cretaceous (Wen et al., 2008b; Ding et al., 2014; Dong et al., 2018; Zhu et al., 2019). The two competing end-member hypotheses for the Late Cretaceous tectonic evolution of the southern Lhasa terrane require reexamination. However, studies of the tectonic structures and their evolution are still scarce (Kapp et al., 2007; Ma et al., 2017a; Wang et al., 2017a). More importantly, the temporal relationship between deformation and the episodic magmatism of the Gangdese belt (namely the arc magmatic tempos) remains largely unexplored. These relationships are key for understanding tectonicscale driving forces and for modeling the thermal-structural evolution of orogens from early-stage accretion to late-stage continental collision.

The central part of the Gangdese belt, specifically the region between Lhasa and Mozhugongka, exposes Lower Jurassic to lower Eocene volcano-sedimentary sequences as well as voluminous intrusive rocks, including the Paleocene Quxu batholith (Figs. 1 and 2). Rocks exposed in this region therefore preserve a record of the structural evolution of the Late Cretaceous accretionary orogeny along with their relation to the magmatic tempos of the Gangdese batholith. Here, we present new structural observations from field outcrops and a drill core as well as geochronological data to characterize the tectonic framework and deformation of the Gangdese back-arc basin. We also explore the relationship between the episodic magmatism and tectonics during the accretionary orogenic phase. Our work sheds new light onto the processes and timing of accretionary orogenesis of the Lhasa terrane prior to the Indo-Asian collision and furthers our understanding of coupled continent-continent collisional and accretionary-subduction orogeny.

The Gangdese magmatic belt, located at the southern margin of the Lhasa terrane, extends for >1500 km along strike, with a width ranging between ~20 and 100 km (Fig. 1A). To the west, the Gangdese magmatic belt is linked to the Ladakh and Kohistan batholiths. To the east, it is linked to the Lohit batholith (Zhu et al., 2019). The Gangdese magmatic belt is composed primarily of largescale batholiths and corresponding volcanic sequences, preserving a record of Neotethyan oceanic slab subduction from the Mesozoic to Cenozoic (Yin and Harrison, 2000; Ji et al., 2009; Hu et al., 2015; Najman et al., 2017; Li and Mattern, 2021). Compositionally, the Gangdese batholith comprises gabbro, diorite, granodiorite, monzogranite, and granite as well as porphyry stocks, ranging in age from Middle Triassic (ca. 240 Ma) to late Miocene (ca. 9.5 Ma) (Laskowski et al., 2018; Meng et al., 2018; Ma et al., 2020). Two dominant age peaks are interpreted to correspond to two magmatic flareups at 90 ± 5 Ma and 50 ± 3 Ma (Zhu et al., 2019). Two contrasting tectonic models have been widely used to interpret the Late Cretaceous magmatic flareup: slab rollback and ridge subduction of the Neotethyan oceanic lithosphere beneath the Lhasa terrane (Ma et al., 2015; Zhu et al., 2019; Meng et al., 2021; Ding et al., 2022a). The second magmatic flareup that peaked ca. 50 Ma could have been caused by the slab breakoff of the subducted Neotethyan oceanic lithosphere (Zhu et al., 2015, 2019).

Volcano-sedimentary sequences that are spatio-temporally associated include the Lower to Middle Jurassic Bima and Yeba Formations, the Upper Jurassic Duodigou Formation, the Lower Cretaceous Linbuzong and Chumulong Formations, and the Upper Cretaceous Shexing Formation, among others (Figs. 1 and 3). The Bima Formation comprises basalt, basaltic andesite, andesite, rhyolite, limestone, and slate (Kang et al., 2014; Wang et al., 2016; Ma et al., 2017b). The Yeba Formation is characterized by bimodal volcanic rocks whose formation is attributed to back-arc rifting that was triggered by the northward subduction of the Neotethyan oceanic slab (Zhu et al., 2008a; Liu et al., 2018; Ma et al., 2019). The Duodigou Formation is dominantly limestone. The Lower Cretaceous Linbuzong and Chumulong Formations comprise chiefly slate and siltstone, and the Upper Cretaceous Shexing Formation comprises siltstone and sandstone. The variable lithologies imply a regressive depositional environment.

Tectonically, the Gangdese magmatic belt is bound to the north by the Luobadui-Milashan fault Yarlung Tsangpo suture zone to the south (Fig. 1A) (Pan et al., 2006; Zhu et al., 2008b). The Luobadui-Milashan fault zone exposes a stack of folded, north-directed thrust faults, ~15 km north of the Linzhou area, including the Gulu-Hamu retro-arc thrust, which records northward translation of the Gangdese belt over the Central Lhasa subterrane (Murphy et al., 1997; Kapp et al., 2007). The structurally highest of these thrusts, with Paleozoic strata in the hanging wall, is intruded by a suite of 57–50 Ma granitoids along strike to the west (Kapp et al., 2007). To the north of the Luobadui-Milashan fault zone, the Central Lhasa subterrane is characterized by Precambrian basement intruded by voluminous Jurassic–Cretaceous, collision-related S-type granites (Zhu et al., 2008b). In contrast, the Gangdese belt is characterized by juvenile crust, apparently lacking basement rocks (Ji et al., 2009). In addition, the Triassic–Jurassic igneous rocks of the Gangdese belt show highly depleted zircon Lu-Hf and whole-rock Sr-Nd isotopic signatures and arc-like geochemical composition indicative of continued subduction of the Neotethyan oceanic lithosphere (Zhu et al., 2008b; Wang et al., 2016; Ma et al., 2020).

The Xigaze forearc basin, immediately north of the Indus–Yarlung Tsangpo suture zone (Fig. 1B), is composed of the Xigaze Group, including the Ngamring, Padana, and Qubeiya Formations, deposited at ca. 107–65 Ma (Wu et al., 2010; Orme et al., 2015). To the north, the Xigaze forearc basin is bound by the south-dipping Great Counter thrust, which is of late Oligocene–early Miocene age (Fig. 1B) (Yin, 2006; Laskowski et al., 2017). To the south, the basin is defined by the north-dipping Gangdese thrust, which is thought to have initiated during the late Oligocene (Yin et al., 1994). Forearc and subsequent syncollisional sequences were deposited on the Xigaze ophiolitic basement (Huang et al., 2015; Orme and Laskowski, 2016; Wang et al., 2017b). Further to the south, a sedimentary-matrix mélange was thrust southward over the southern Tethyan Himalayan sequence along the east-west–striking Zhongba-Gyangze thrust at ca. 71–61 Ma (Ding et al., 2005; Wang et al., 2017a).

The Late Cretaceous Gangdese back-arc basin (~200 km in length and ~60–70 km in width) is located at the northern side of the Mesozoic Gangdese batholith, north of the Quxu batholith (Fig. 1B). The basin exposes marine sedimentary rocks of the Upper Jurassic Duodigou Formation and terrestrial sedimentary rocks of the Cretaceous Linbuzong, Chumulong, Takena, and Shexing Formations (Figs. 2 and 3) totaling a thickness of ~3 km. The basement of the basin consists of basic and intermediate volcanic rocks of the Yeba Formation with ages between 190 and 174 Ma (Fig. 3) (Zhu et al., 2008a). Back-arc sequences are overlain by the ca. 65–45 Ma Linzizong Formation volcanic succession along an angular unconformity (hereon referred to as the Gangdese angular unconformity) (Mo et al., 2007; Zhu et al., 2015). Below the unconformable contact, the Gangdese back-arc sequences are strongly folded, whereas the overlying Linzizong volcanic sequences remain weakly deformed (Kapp et al., 2007).

Based on detailed field investigation, we distinguish two episodes of overprinting deformation. The first deformation phase (D₁) is contractional, accommodated by folding and south-vergent thrusting. The second phase of deformation (D₂) is also contractional and expressed as regional doming and open folding of D₁ structures.

D₁ is defined by a series of folds and south-vergent thrusts developed in the Upper Jurassic to Lower Cretaceous strata of the Gangdese back-arc basin (Fig. 4). Fold axial traces strike eastwest and axial planes are characterized by upright orientations in the upper strata and by south-verging open folds in the stratigraphically lower strata (Fig. 4; Fig. S1 in the Supplemental Material¹). Formation of the south-vergent structures is inferred to be associated with a ~2-km-thick composite sole detachment zone, exposed between the underlying Middle Jurassic Yeba Formation sequence and the overlying Upper Jurassic–Cretaceous sediments. We refer to these structures as the Gangdese décollement (GDSD) (Figs. 2B and 4). The GDSD includes two décollement surfaces. The lower décollement surface (GDSD-1) includes a ~500-m- to ~1-km-thick mylonite zone developed within volcanic and sedimentary rocks of the underlying Middle Jurassic Yeba Formation and overlying Upper Jurassic lime-stone of the Duodigou Formation (Fig. 5). The upper décollement surface (GDSD-2) is composed of a 1.5–2-km-thick shear zone developed in slate and sandstone between the underlying Upper Jurassic Duodigou Formation limestone and the overlying Lower Cretaceous Linbuzong Formation slate and sandstone (Fig. 4). Despite shearing and inferred displacement along the décollement, the original stratigraphy remains intact. We suggest that the upper and lower contacts of the Upper Jurassic Duodigou Formation limestone localized strain, promoting formation of the two décollement surfaces.

The mylonitic sandstone and schist of GDSD-1 are characterized by north-dipping foliation (S₁) and north-south–trending stretching lineation (L₁) (Figs. 5A and 5B). The north-south–trending stretching lineation is defined by stretched volcanic clasts, felsic mineral clusters, and plagioclase clasts (Fig. 5D). S-C fabric orientation and asymmetric volcanic clasts within the mylonitized volcanic rocks indicate dominantly top- to- the- south and/or top- to- the- southeast sense of shear (Figs. 5A–5C). Top- to- the- south and/or top- to- the- southeast shearing of GDSD-1 is further evidenced by asymmetric tails on feldspar clasts, asymmetric calcite, S-C fabric orientation, and felsic boudins or lenses (Figs. 6 and 7).

Rocks deformed within GDSD-2 record similar structures and kinematics to those in GDSD-1. In outcrop, S-C fabrics, asymmetric porphyroclasts, and folds are consistent with top- to- the- south sense of shear (Ma et al., 2017a). In thin section, deformed rocks from the GDSD-2 zone show asymmetric feldspar, calcite, and biotite clasts (Fig. 6B), S-C fabrics (Fig. 6C), and sigmatype felsic lenses (Fig. 6D) also consistent with these kinematics.

Rocks deformed by GDSD-2 were found in drill core ZK2314, drill line 23, in the central segment of the Jiama mining district (Figs. 2 and 8), northeast of Lhasa. The ZK2314 drill core recovered 311.2 m, which included Lower Cretaceous Linbuzong sandstone and slate (core depths of 0–220 m), a porphyry zone (220–230 m), and Upper Jurassic marble or limestone of the Duodigou Formation (240–311.2 m) (Fig. 9). The porphyry Cu ore body intruded along the shear zone of GDSD-2 at 16–15 Ma (Fig. 8), indicating that deformation along the shear zone had ceased by that time (Yang et al., 2016).

Slate and sandstones of the Lower Cretaceous Linbuzong Formation (0–220 m of drill core ZK2314) record evidence for ductile deformation including foliation and cleavage development (S₁), grainsize reduction (which we interpret to result from dynamic recrystallization), shear-band development, and folding of quartz veins (Figs. 9 and 10). Microstructures recorded within the drill core either are consistent with top- to- the- south and/or top- to- the- southeast sense of shear or are ambiguous and do not contain convincing shear-sense indicators. In parts of the drill core, asymmetric folding could be suggestive of top- to- the- north sense of shear, opposite that of GDSD kinematics (Figs. 9 and 10). The Upper Jurassic limestone of the Duodigou Forma tion (240–311.2 m of ZK2314) is characterized by grainsize reduction, development of a gently north-dipping foliation, and a NNE-SSW–trending stretching lineation (Figs. 9 and 10).

We posit that the deformation of the Cretaceous Linbuzong and Shexing Formations (Fig. S1E), recorded by the south-vergent structures with upright folds in the top and tight to isoclinal folds with north-dipping axial planes in the lower part, is associated with the sole GDSD-2 décollement (Fig. 9). In contrast, the Paleogene Linzizong Forma tion volcanic sequence is weakly deformed and unconformably overlies on the intensely deformed Shexing Formation (Fig. S1E; Kapp et al., 2007).

The GDSD was folded into open, generally upright to slightly south-vergent folds and formed a kilometer-scale dome structure. Radiating orientations of the S₁ foliation and stretching lineation were observed in the Jurassic Yeba and Duodigou Formations in the Lhasa region, providing evidence for this separate deformational event (Fig. 2A). The northern and southern flanks of the anticline show steeply north-dipping and south-dipping foliation orientations, respectively, whereas the western flank records foliation orientations gently dipping to the west (10°–30° dip) accompanied by an east-west–trending lineation orientation (Figs. 2A and 2B). This domal structure was previously named the “Lhasa dome” (Ma et al., 2017a). The steep to nearly vertical dips recorded by the Yeba Formation may result from steepening of S₁ due to doming (Fig. 5). We tentatively propose that the cryptic, and in some cases ambiguous, top- to- the- north shear sense indicators recorded in parts of the drill core could be associated with formation of the Lhasa dome, given that timing estimates overlap.

To constrain the timing of the Gangdese accretionary orogenic phase, specifically the activity of the GDSD in the Gangdese back-arc basin, we used the ⁴⁰Ar-³⁹Ar system to date cooling ages of white mica, plagioclase, and K-feldspar derived from mylonitic rocks that record south-vergent shearing (Fig. 7). We also performed zircon U-Pb dating on sandstone samples from the top of the Upper Cretaceous Shexing Formation to constrain the maximum depositional age and on the granites that intruded the deformational strata (Fig. 2) to constrain the minimum age of deformation.

Six ⁴⁰Ar-³⁹Ar ages were determined in two laboratories. Four samples (XY8-1-7, XJ15-3, XY10-1-1, and XY11-2-1-6) were collected from the lower décollement layer of GDSD-1 (Fig. 2). Samples XY8-1-7 and XJ15-3 are classified as volcanic mylonite. Sample XY10-1-1 is a mica schist whose protolith was likely a tuffaceous volcanic rock (Fig. 7). Sample XY11-2-1-6 is a granitic stock intruded into the Upper Jurassic Duodigou Formation limestone. These four samples (XY8-1-7, XJ15-3, XY10-1-1, and XY11-2-1-6) were dated at the Noble Gas Geochronology Laboratory at the University of Melbourne, Australia. Another two samples (M16-11-6 and M16-11-8, both mylonitic sandstone) were dated at the Key Laboratory of Deep-Earth Dynamics, Ministry of Natural Resources, Institute of Geology, Chinese Academy of Geological Sciences. Detailed analytical methods of ⁴⁰Ar-³⁹Ar dating are presented in Text S1 (see ^{footnote 1}). Results are tabulated in Tables S1 and S2.

⁴⁰Ar-³⁹Ar results produced a mix of concordant and discordant spectra, although even discordant results can be interpreted to give approximate constraints on cooling age. The white mica of sample XY-8-1-7 yields a discordant age spectrum with a decrease in apparent ages from ca. 65 to ca. 40 Ma for the first 12% of ³⁹Ar release, followed by a gradual increase to ca. 58 Ma (Fig. S2A [see ^{footnote 1}]). The sericitized plagioclase of sample XJ15-3 yields a discordant age spectrum, with ages increasing from ca. 45 to ca. 61 Ma, with some anomalously old values at the high-temperature steps (Fig. S2B). The sericite of mica schist sample XY10-1-1 yields a discordant, “humpshaped” spectrum, with an age range of ca. 90–100 Ma (Fig. S2C). The K-feldspar of granite sample XY11-2-1-6 yields a welldefined plateau age of 58.91 ± 0.48 Ma (2σ) for the mid-to high-temperature steps (Fig. S2D).

The white mica of sample M16-11-6 yields a concordant age spectrum with a weighted plateau age of 89.04 ± 1.11 Ma, corresponding to a normal isochron age of 88.87 ± 2.03 Ma (Figs. 11A and 11B). Similarly, the white mica of sample M16-11-8 yields a weighted plateau age of 89.51 ± 1.13 Ma, indistinguishable from the normal isochron age of 88.62 ± 1.30 Ma within error (Figs. 11C and 11D).

Detailed analytical methods of zircon U-Pb dating are presented in Text S1 (see ^{footnote 1}). Sample Ct18 (sandstone) was collected from the top of the Shexing Formation, which is unconformably overlain by the 65–40 Ma volcano-sedimentary Linzizong Formation (Fig. 2; Fig. S1E). U-Pb dating of its detrital zircon grains yielded a large age range from Neoarchean to Late Cretaceous (ca. 2638–87 Ma), revealing complex provenance. The 11 youngest concordant zircon U-Pb ages yielded a weighted mean age of 91.1 ± 2.1 Ma (Fig. 12). This age is indistinguishable from the age of the youngest red beds of the Shexing Formation (ca. 90 Ma; Kapp et al., 2007; Li et al., 2015b; Wang et al., 2020; Wei et al., 2020). In Zhu et al. (2019), the youngest age group of detrital zircons from the uppermost Shexing Formation sandstone was dated at 85 ± 1 Ma, and they suggested that the strong folding and deformation of the Upper Cretaceous Shexing Formation at Maxiang took place between 85 Ma and 69 Ma. In other words, the maximum depositional age of the Shexing Formation could be no older than 85 Ma (Zhu et al., 2019), which is a little younger than the age of ca. 90 Ma from Kapp et al. (2007) and the present study. However, the calculation methods of both Zhu et al. (2019) and the present study have employed the weighted mean age of the youngest age group of detrital zircons, whose precision mainly relied on the age number, age distribution, and analytical errors. Thus, both results are reasonable and acceptable. More work is needed in the future. Tentatively, our new results, in combination with published age data, suggest that the maximum depositional age of the Shexing Formation is ca. 90 Ma. The analytical results are presented in Table S3.

Six samples (XZ2-8-1, XZ2-8-2, XZ2-6-1, XD4-6-1, XG8-7, and XG8-8) from granites in the studied deformed strata in the Lhasa region and adjacent areas (Fig. 2) were selected for zircon U-Pb dating. Cathodoluminescence images of zircon grains from these samples show euhedral morphologies and sharp oscillatory zoning. The Th/U ratios of the zircon grains from the six samples range 0.32–0.97, 0.61–1.48, 0.39–2.14, 0.63–2.06, 0.77–1.8, and 0.52–1.32, respectively, indicating a magmatic origin (Table S4 [see ^{footnote 1}]).

U-Pb compositions (Table S4) of the zircon grains yield concordia ages of 42.2 ± 2.8 Ma (mean square of weighted deviates [MSWD] = 0.44, n = 13, sample XZ2-8-1), 66.6 ± 1.7 Ma (MSWD = 0.10, n = 15, sample XZ2-8-2), 43.7 ± 5.6 Ma (MSWD = 0.74, n = 23, sample XZ2-6-1), 48.8 ± 0.7 Ma (MSWD = 1.7, n = 16, sample XD4-6-1), 52.7 ± 0.7 Ma (MSWD = 1.2, n = 20, sample XG8-7), and 61.6 ± 1.3 Ma (MSWD = 1.3, n = 20, sample XG8-8) (Fig. 13). These data indicate that the studied granites crystallized between ca. 67 and 42 Ma, which is consistent with previous age dating results (65.6–59.3 Ma) for granites around the Lhasa region (Wen et al., 2008a; Ji et al., 2009; Zhu et al., 2015; Ma et al., 2017a) (Fig. 2).

Based on our structural analysis, we suggest two stages of deformation: a first deformation stage (D₁) with top- to- the-south–directed shearing and a second deformation phase (D₂) of regional doming. D₁ records early deformation of GDSD-2, while D₂ represents regional doming, evidenced by radiating orientations of the foliation and stretching lineation (Fig. 2). As stated above, the décollements are interesting in that they do not preserve evidence for duplication of the stratigraphic section. Based on the results presented in the cross sections, we found that they do not juxtapose older rocks on younger rocks (Figs. 2 and 4). GDSD-1 is developed between the underlying volcano-sedimentary sequences of the Middle Jurassic Yeba Formation and the overlying Upper Jurassic limestone of the Duodigou Formation (Fig. 4). GDSD-2 is developed between the underlying Upper Jurassic Duodigou Formation limestone and the overlying Lower Cretaceous Linbuzong Formation slate and sandstone (Fig. 4).

D₂ open folding and dome structure may have been triggered by or otherwise related to the intrusion of the ca. 65–43 Ma granitic plutons during continuing contraction (Ma et al., 2017a), analogous to gneiss domes that are structural domes cored primarily by gneissic rocks and granite and mantled by highgrade schist and/or gneiss (Whitney et al., 2004) (Fig. 2B). Units that show D₁ structures were intruded by granites with ages of ca. 65–43 Ma (Fig. 2). The steep orientation of D₁ structures may result from steepening due to doming (Fig. 5).

The dated granitic plutons yielded crystallization ages of 65–43 Ma, falling into the interval of 65–40 Ma magmatic flareup in the Gangdese belt (Ji et al., 2009; Zhu et al., 2019; Ma et al., 2022), represented by the Linzizong Formation volcanics and the Quxu batholith. However, no precise or specific age of the dome could be distinguished from these Ar-Ar dating results (Fig. S2 [see ^{footnote 1}]). At present, we cannot identify the precise syntectonic plutons from these dated granites due to a lack of detailed studies on the steep fabrics and the transition from magmatic to solid-state deformation in the granite. A comprehensive study of magmatic flow, submagmatic flow, and solid-state deformation, in combination of precise age dating, is needed to identify the syntectonic plutons north of Lhasa and to better understand the proposed doming in the retro-arc region of the Gangdese belt.

The Gangdese décollements are located to the south of the Gangdese retro-arc fold-thrust system, in close association with the Emei La and the Gulu-Hamu thrusts (Figs. 1B and 2). Structurally, the Gangdese décollements record a similar, top- to- the- south sense of shear as the Emei La and Gulu-Hamu thrusts. Temporally, the Gangdese décollements were formed coeval with the Emei La thrust, both active within the time range of 90–69 Ma (Kapp et al., 2007). Thus, we suggest that the south-vergent structures of the Gangdese décollements may have been part of a major south-directed retro-arc fold-thrust belt in the southern Lhasa terrane during the Late Cretaceous.

The Gangdese unconformity, separating the overlying Paleogene–Eocene Linzizong Formation volcanic sequence and the underlying, folded Upper Jurassic to Cretaceous strata is a characteristic feature of the Gangdese back-arc basin (Fig. S1E [see ^{footnote 1}]). The Upper Jurassic–Cretaceous strata exhibit intense folding and thrusting associated with the sole décollement—the Gangdese décollement (GDSD) (Figs. 2 and 4)—while the Linzizong volcanic sequence is only weakly deformed (Fig. S1E) (Kapp et al., 2007). Our new zircon U-Pb age data from undeformed granites intruded into deformed Upper Jurassic–Cretaceous strata record crystallization at ca. 67–42 Ma (Figs. 2A and 13), indicating that the folding and thrusting (D₁) occurred before ca. 67 Ma. In addition, numerous reported ages of the Linzizong Formation volcanic rocks are equal to or younger than ca. 65 Ma (Ding et al., 2003; Mo et al., 2003; Lee et al., 2009), further supporting that the Gangdese angular unconformity predated ca. ~67–65 Ma.

The protolith ages of the volcanic mylonites of the Lower to Middle Jurassic Yeba Formation range between ca. 195 and 170 Ma (Zhu et al., 2008a; Wei et al., 2017; Liu et al., 2018; Ma et al., 2019). We propose that the ca. 100–90 Ma white mica ⁴⁰Ar-³⁹Ar age of mica schist sample XY10-1-1 from the top of the Yeba Formation of GDSD-1 provides an older age limit for initiation of GDSD-1 (Fig. S2). Two other mylonitic sandstone samples (M16-11-6 and M16-11-8) yielded consistent results, with plateau ages of 89.0 ± 1.1 Ma and 89.5 ± 1.1 Ma and corresponding normal isochron ages of 88.9 ± 2.0 Ma and 88.6 ± 1.3 Ma, respectively. These results further indicate that GDSD-1 was active prior to 89 Ma (Fig. 11). We interpret the younger ca. 65–40 Ma cooling ages of the two volcanic mylonitic samples (XY8-1-7 and XJ15-3) from beneath GDSD-1 have been reset by a thermal pulse related to the pluton emplacement or triggered by the continuing contraction due to the Indo-Asian collision (Ratschbacher et al., 1994).

An alternative explanation for the Late Cretaceous ⁴⁰Ar-³⁹Ar ages is that they record burial due to crustal thickening in the back-arc basin. In our opinion, however, this possibility should be ruled out based on the following: (1) the dated white micas come from mylonitic samples, whose protolith are Jurassic-aged volcano-sedimentary sequences (Figs. 5–7); (2) beginning at ca. 80 Ma, sediment transport to the forearc and trench decreased, which is inconsistent with significant crustal exhumation in the Gangdese batholith and the back-arc crust (Wu et al., 2010; Metcalf and Kapp, 2019); (3) our new ages of ca. 90–89 Ma are indistinguishable from previous results on the Late Cretaceous deformation of the Jurassic-aged Yeba Formation (Zhong et al., 2013; Feng et al., 2022); and (4) similarly, Ratschbacher et al. (1992) reported that a top- to- the- south to top- to- the- ESE displacement in the Gangdese belt is truncated by the Gangdese unconformity, and therefore they proposed that this deformation occurred between 90 and 60 Ma.

Based on fossil assemblage and ⁴⁰Ar-³⁹Ar dating, previous studies found that the base of the Shexing Formation (i.e., the folded sandstone beneath the >67–65 Ma unconformity) was deposited at ca. 110 Ma and that the youngest red sandstone was deposited at ca. 90 Ma (Kapp et al., 2007; Li et al., 2015b; Wang et al., 2020). Our new detrital zircon U-Pb age data for the Upper Cretaceous Shexing Formation red bed yielded a maximum depositional age of ca. 91 Ma, which agrees within uncertainty with a previously published ca. 90 Ma depositional age (Kapp et al., 2007; Leier et al., 2007; Li et al., 2015b; Wang et al., 2020; Wei et al., 2020; Xing et al., 2020). In contrast, this result is older than the K-Ar age of 68–75 Ma for basalt within the Shexing Formation (Cao et al., 2017a). However, new interpretations suggest that the dated basalt intruded as a sill within the Shexing Formation (Wang et al., 2020). Deformed rocks of the Upper Jurassic to Cretaceous sequence are intruded by undeformed granites at ca. 67–42 Ma and are unconformably overlain by the Linzizong Formation volcanic sequence (65–44 Ma) (He et al., 2007; Zhu et al., 2015). Considering the Gangdese angular unconformity formed between 85 and 69 Ma (Zhu et al., 2019), we therefore infer that the Gangdese accretionary orogenic phase initiated at ca. 90 Ma, culminating between ca. 85 and 69 Ma.

The largescale Gangdese angular unconformity has been interpreted to mark the initial collision between the Asian and Indian continents (Mo et al., 2008). Recently, an increasing number of studies have argued that the initial Indo-Asian collision occurred at 59–55 Ma (DeCelles et al., 2014; Hu et al., 2015; Zhu et al., 2019) or as late as 40–23 Ma (Aitchison et al., 2007; Ao et al., 2018; van Hinsbergen et al., 2019). Based on our structural analysis along with new and published age data, we argue that shortening along the GDSD initiated at ca. 90 Ma, followed by the formation of the ca. 85–67 Ma angular unconformity of the Lhasa terrane.

Numerous studies have found that magma addition and crustal growth have been episodic throughout Earth’s history, with prominent magmatic episodes at ca. 2100–1800 Ma, 1100–800 Ma, and 350–250 Ma, corresponding to the assembly of the Columbia, Rodinia, and Pangea supercontinents, respectively (Condie, 1998, 2007; Rogers and Santosh, 2003; Kröner and Stern, 2004; Cawood and Buchan, 2007; Linnemann et al., 2008; Cao et al., 2017b; Spencer et al., 2018). Recent studies have similarly demonstrated that the construction of Phanerozoic arc belts worldwide has also been episodic, with so-called magmatic flareups (voluminous magma additions) typically lasting ~10–30 m.y. and intermittent magmatic lulls (minor magma additions) (Ducea and Barton, 2007; Ducea et al., 2015; Paterson and Ducea, 2015; Cao et al., 2017b; Chapman et al., 2017; Ma et al., 2021b, 2022). These magmatic tempos are thought to be controlled by deep geodynamic processes, for example, by the angle of the subducting slab and the extent of mantle wedge partial melting (DeCelles et al., 2015; Chapman and Ducea, 2019; Martínez Ardila et al., 2019).

Magmatism in the Gangdese belt is clearly episodic with magmatic flareups at ca. 100–80 Ma and ca. 65–45 Ma and a pronounced arc magmatic lull between ca. 80 and 65 Ma (Figs. 14 and 15) (Ji et al., 2009; Zhu et al., 2019; Ma et al., 2022). The cause of the ca. 100–80 Ma magmatic flareup is debated, having been ascribed to either ridge subduction (Fig. 16A) (Zhang et al., 2010) or slab rollback (Ma et al., 2015). The ca. 65–45 Ma magmatic flareup, which was volumetrically more significant than the ca. 100–80 Ma flareup, has been attributed to slab rollback and subsequent slab breakoff, marking the transition from ocean-continent subduction to Indo-Asian continent-continent collision (Mo et al., 2003; Ma et al., 2017c). Studies documenting magmatic flareups of the Gangdese belt clearly document a magmatic lull between ca. 80 and 65 Ma (Figs. 14 and 15). Details of the geodynamic configuration during this magmatic lull, however, remain largely unresolved.

The Gangdese belt underwent several major evolutionary stages between the Late Cretaceous and the Eocene (Fig. 16). First, mid-ocean ridge subduction of the Neotethyan oceanic slab occurred beneath the southern Lhasa terrane between ca. 100 and 85 Ma, which triggered a magmatic flareup with formation of voluminous intrusions and high-temperature granulitefacies metamorphism (Fig. 16A) (Zhang et al., 2010; Kapp and DeCelles, 2019; Zhu et al., 2019; Ding et al., 2022b). Second, the Gangdese accretionary orogenic phase took place at ca. 85–65 Ma. Different dating campaigns (this study; Zhong et al., 2013; Ma et al., 2017a; Zhu et al., 2019) have, however, revealed that fold-thrust and sole décollement of the Gangdese back-arc basin took place between 90 and 67 Ma, commencing prior to but temporally overlapping with the arc magmatic lull in the Gangdese belt (Wen et al., 2008a). The Gangdese belt underwent crustal shortening, thickening, uplift, and erosion, as well as a magmatic lull, plausibly caused by flat-slab subduction of young, buoyant Neotethyan oceanic lithosphere (Fig. 16B) (Kapp et al., 2007; Zhong et al., 2013; Zheng et al., 2014; Ma et al., 2017a). This implies a close relationship between the magmatic lull and synchronous crustal deformation, which we explore further in the Geodynamics of the Gangdese Accretionary Orogenic Phase section below. Eventually, the Gangdese batholith entered a contractional regime at ca. 65–45 Ma. The contraction and intrusion of granitic plutons triggered the formation of the Lhasa dome structure (Ma et al., 2017a). In the meantime, the slab front of the Neotethyan oceanic lithosphere began to roll back, which was accompanied by extension within the Gangdese back-arc region. The continuing rollback induced voluminous volcanism and southward migration of the arc-type magmatism (Fig. 16C) (Ding et al., 2003; Mo et al., 2003; Ji et al., 2009; Lee et al., 2009; Zhu et al., 2015; Ma et al., 2017c).

The voluminous ca. 100–85 Ma arc magmatism in the Gangdese belt was most likely triggered by subduction of the mid-ocean ridge of the Neotethys beneath the southern Lhasa terrane rather than slab rollback (Fig. 16A). Support for the mid-ocean ridge subduction model comes from the occurrence of high-temperature granulite-facies metamorphism (Zhang et al., 2010, 2015; Guo et al., 2013; Qin et al., 2019) and contraction along the southern margin of the Lhasa terrane (Zhong et al., 2013; Ma et al., 2017a; Feng et al., 2022). Young, relatively buoyant and thick oceanic lithosphere near mid-ocean ridges has a greater effective elastic thickness and is therefore more likely to subduct at a gently dipping or flat-slab geometry (van Hunen et al., 2004; Huangfu et al., 2016; Manea et al., 2017). For example, the subhorizontal slab segments below western South America are suggested to result from the subduction of relatively buoyant oceanic plateau and aseismic ridges (Martinod et al., 2010). In the Gangdese belt, the flat-slab subduction started no later than ca. 83–80 Ma, evidenced by (1) the occurrence of adakitic granitoids ca. 83–80 Ma in the Gangdese belt (Wen et al., 2008a, 2008b; Guan et al., 2010; Zheng et al., 2014; Xu et al., 2015), (2) the waning of arc magmatism from ca. 83 to 65 Ma in the Gangdese belt (Ma et al., 2022), (3) the scarcity of widespread rift-related volcanic rocks between 90 and 65 Ma (Zhu et al., 2019), (4) the cessation of the Cretaceous Takena Formation deposition at ca. 90–80 Ma (Leier et al., 2007), (5) the 85–69 Ma regional unconformity between the underlying Upper Cretaceous Shexing Formation sandstone and the overlying Paleogene Dianzhong Formation volcanic rocks in the Gangdese belt (Zhu et al., 2019), and (6) the ca. 90–69 Ma forearc and retro-arc fold-thrust system (He et al., 2007; Kapp et al., 2007).

Flat-slab subduction, in turn, may exert a traction on the overlying lithosphere, leading to contraction (Fig. 16B), as is hypothesized in Cordilleran orogens (Manea et al., 2017; Gutscher, 2018). In South America, flat-slab subduction of the Nazca plate beneath the Central Andes during the Eocene–Oligocene and beneath northcentral Chile and Peru since the Pliocene has triggered uplift and deformation of the overlying crust, recorded by a series of thrusts in the forearc and back-arc regions (Ramos et al., 2002; Kay and Coira, 2009; Martinod et al., 2010; Chiarabba et al., 2016). The flattening of the subducting slab angle may thicken the crust and squeeze out the mantle wedge, a process which could initiate intrusion of adakites following the termination of arc magmatism (Fig. 16B) (Booker et al., 2004).

We show that the Gangdese belt, like the South American Cordilleran orogen (McQuarrie, 2002; Ramos, 2009), experienced folding and thrusting along the southern margin of the Lhasa terrane in the Late Cretaceous at ca. 90–65 Ma. In the forearc region, the Indus–Yarlung Tsangpo mélange was obducted southward along the south-directed thrust at ca. 63 Ma, in concert with the development of an angular unconformity within the Xigaze forearc basin (between late Maastrichtian time and ca. 62 Ma) (Ding et al., 2005). To the south of the Xigaze forearc basin, a series of top- to- the- south thrusts (e.g., the Zhongba-Gyangze thrust) were active between 71 and 61 Ma (Wang et al., 2017a). Similarly, the retro-arc foreland basin of the Gangdese belt developed a series of thrusts (e.g., the Emei La thrust and Gulu-Hamu thrust), which were active between 90 and 69 Ma (Kapp et al., 2007), as well as asymmetric, northvergent, mesoscale folding (Fig. 16B) (He et al., 2007).

The Cretaceous retro-arc basin to the north of Lhasa city was previously inferred to have been deformed from ca. 90 to 65 Ma (Ma et al., 2017a). Our new results agree with this interpretation, and we provide more constraints, including white mica ⁴⁰Ar-³⁹Ar dates inferred to constrain the approximate timing of deformation. We show that the angular unconformity between the overlying Linzizong Formation sequence and the underlying Shexing Formation in the Linzhou retro-arc basin developed between 90 and 67 Ma, in close agreement with constraints that suggested an unconformity age between ca. 85 and 69 Ma (Zhu et al., 2019). Our ⁴⁰Ar-³⁹Ar dating of white mica from mylonites along the GDSD-1 décollement yielded plateau ages of 89.5–89 Ma and corresponding 88.9–88.6 Ma isochron ages, which we interpret to indicate that top- to- the- south shearing was initiated at or before 89 Ma, immediately following (i.e., ≤1 m.y.) deposition of the Shexing Formation. The GDSD décollement is inferred to be associated with a ~2-km-thick composite sole detachment zone, exposed between the underlying Middle Jurassic Yeba Formation sequence and the overlying Upper Jurassic–Cretaceous sediments. The zones of deformation may have experienced significant slip between units. They seem more akin to detachment horizons within a fold belt rather than thrust shear zones that accommodated large amounts of shortening. The top- to- the- south GDSD décollement roots beneath the arc and carries arc rocks toward trench, which may be analogous to the West Andean thrust system, a series of structures that translate the arc westward toward the Chile trench (McQuarrie, 2002; Ramos, 2009; Díaz et al., 2014).

Subduction at low angles may induce an amagmatic regime in the overlying magmatic arc as the mantle wedge migrates in the same direction as the subducting plate (Gutscher et al., 2000; Booker et al., 2004). This plausibly occurred in the Gangdese belt, where magmatism waned from ca. 85 to 65 Ma (Fig. 14) (Wen et al., 2008a). This magmatic lull is consistent with hindered heating from the asthenospheric mantle and limited or no dehydration of the subducting slab. Accompanying the waning of the arc-type magmatism in the Gangdese belt, the Central and Northern Lhasa subterranes were gradually uplifted to the “Lhasaplano” (Kapp et al., 2007). The elevation of the “Lhasaplano” reached ~4500 m before ca. 60 Ma, corroborated by carbon and oxygen isotope data for the Nianbo Formation marls in the Linzhou Basin (Ding et al., 2014) and detrital zircon Eu/Eu* [asterisk (*) denotes square root of (Sm × Gd)] of sandstones from the Lhasa River (Tang et al., 2021).

Rollback of the Neotethyan oceanic slab front marked cessation of flat-slab subduction and induced voluminous volcanism (e.g., the Linzizong Formation volcanics) (Fig. 16C) (Wen et al., 2008a; Zhou et al., 2018). Due to slab rollback, the arc-type magmatism migrated more than ~100 km from north to south at ca. 65–50 Ma (Zhu et al., 2015, 2019), similar to migration observed in the postorogenic magmatic province in Mesozoic South China (Li and Li, 2007) and that of the Mesozoic–Cenozoic magmatism in the western United States (Humphreys, 2009; Ardill et al., 2018). Arc magmatism flared at ca. 50 Ma (Fig. 14), forming the largescale Quxu batholith (Fig. 16C), from which voluminous microgranular enclaves record significant input of relatively primitive magma, triggered by the subsequent slab breakoff (Dong et al., 2006; Mo et al., 2009; Ma et al., 2017c; Wang et al., 2019).

Alternatively, the relatively sparse geological observations available for the 90–70 Ma interval permit an alternative tectonic scenario along the southern Asian margin: rifting and subsequent opening of a back-arc ocean basin (Kapp and DeCelles, 2019). At present, the evidence for the rifting model comes mainly from the western Neotethyan realm, for example, the clockwise rotation and southward translation above a retreating subduction zone of the ca. 95–90 Ma Semail ophiolite of Oman prior to its obduction onto the Arabian continental margin at ca. 75 Ma (Searle et al., 2015), the generation of ca. 80–65 Ma ophiolites within suture zone between India and the Afghan block (Kakar et al., 2012), the ca. 90–80 Ma onset of southward rollback of a northward subducting slab beneath the Kohistan arc (Burg, 2011), and the ca. 90–80 Ma intraoceanic Spong arc in the northwestern Himalaya (Pedersen et al., 2001). Some workers suggest that the Late Cretaceous northdipping subduction zones probably extended eastward and transitioned into the retreating oceanic subduction zone beneath the Gangdese arc (Ma et al., 2013; Sundell et al., 2021). If this was the case, then the rollback of the northdipping subducting slab would have induced upperplate extension and led to an opening of a back-arc ocean basin, such as the proposed Xigaze back-arc basin to the north of the Xigaze arc (Kapp and DeCelles, 2019). However, there is no documented evidence for the existence of the so-called Xigaze arc in the eastern Neotethyan realm, let alone the Xigaze back-arc basin. More work is needed to examine the model of the Xigaze back-arc basin. To date, available data are most consistent with a Cordilleran-style orogenesis along the southern margin of the Lhasa terrane during the Late Cretaceous (Wen et al., 2008a; Metcalf and Kapp, 2019).

Based on our work in combination with previous studies, we suggest that the Gangdese belt could be classified as a Chileantype advancing accretionary orogen, formed by northward flat-slab subduction of the Neotethyan oceanic lithosphere beneath the southern margin of the Asian continent during the Late Cretaceous (ca. 85–65 Ma) (Fig. 16B).

Structural analysis, ⁴⁰Ar-³⁹Ar dating, and U-Pb zircon dating reveal that a top- to- the- south décollement and associated southvergent folds and thrusts accommodated crustal shortening, resulting in a thickened Gangdese belt during the Late Cretaceous. The timing of décollement movement, folding, and thrusting (ca. 90–67 Ma) overlaps with the formation of the Gangdese angular unconformity (ca. 85–69 Ma) developed between lower Upper Jurassic–Cretaceous Gangdese back-arc sequences and upper Paleogene–Eocene Linzizong Formation volcanics. Moreover, published structural, geochronological, geochemical, and sedimentary data indicate that the Gangdese belt experienced an arc magmatic lull between ca. 85 and 65 Ma. Therefore, we propose that the Gangdese belt (southern Lhasa terrane) underwent a Chileantype accretionary orogeny during the Late Cretaceous (ca. 85–65 Ma). We suggest that this accretionary orogeny was likely associated with northdirected flat-slab subduction of the Neotethyan oceanic lithosphere beneath the southern Asian continental margin.

Fruitful discussions with Wenrong Cao, Weiqiang Ji, Dicheng Zhu, Andrew Laskowski, and Kate Metcalf that shed light on the accretionary orogenic phase of the Gangdese belt are much appreciated. Yuan Ma, Zhongbao Zhao, Zhihui Cai, and Huaqi Li helped us with the interpretation of structural observations. We are indebted to Associate Editor Francesco Mazzarini, reviewer Rodolfo Carosi, and an anonymous reviewer for their perceptive comments that improved this manuscript greatly. Science Editor Andrea Hampel is thanked for efficient handling and positive consideration of our work. Erdmann gratefully acknowledges support from the VOLTAIRE project (ANR-10-LABX-100-01), funded by Agence Nationale de la Recherche through the French Programme d’Investissement d’Avenir (PIA). This study was co-supported by the National Natural Science Foundation of China (grants 42272267, 42172263), the second Tibetan Plateau Scientific Expedition and Research Program (STEP) Grant (2019QZKK0901, 2019QZKK0802), the Chinese National Key Research and Development Project “Key scientific issues of transformative technology” (2019YFA0708604), the Scientific Investigation on Basic Resources program of Ministry of Science and Technology of China (2021FY100101), and the Geological Survey of China (DD20221630). Data sets for this research are available online (https://doi.org/10.5281/zenodo.7638083).