Paleocene (c. 62 Ma) Leucogranites in Southern Lhasa, Tibet: Products of Syn-collisional Crustal Anatexis during Slab Roll-back?

Abstract

Voluminous peraluminous leucogranites are common in large-scale orogenic belts and are crucial in gaining a fuller understanding of the related geodynamic process. However, the origin of such syn-collisional leucogranites remains highly controversial. In this contribution, we report petrological and geochemical data for Paleocene (c. 63 Ma) garnet-bearing, two-mica granites and associated biotite granites from the Gangdese batholith in southern Tibet. The Zhengga biotite granites have high SiO₂ (70–73 wt %) and low MgO (0·4–0·7 wt %) contents with initial ⁸⁷Sr/⁸⁶Sr ratios of 0·7049–0·7050, ε_Nd(t) values of +0·5 to +1·2 and zircon δ¹⁸O values of 5·6–6·9‰, similar to most early Paleocene granitoids in southern Lhasa. These geochemical characteristics suggest that the Zhengga biotite granites were derived from a crustal source that mixed with variable amounts of Gangdese juvenile lower crust and minor ancient crust-derived melts. The Zhengga peraluminous, garnet-bearing, two-mica granites have similar Sr–Nd–O isotope compositions to the biotite granites (0·7037–0·7050, +0·4 to +0·8, 5·5–7·3‰, respectively) as well as higher SiO₂ (73–76 wt %) and lower TiO₂ (<0·06 wt %), MgO (<0·3 wt %), Fe₂${\mathrm{O}}_3^{\mathrm{T}}$ (<2 wt %) and CaO (<0·7 wt %) contents. These most probably represent highly evolved biotite granite magmas that differentiated in the mid-crust. The first contact of India with Asia appears to have occurred in central Lhasa during the early Paleocene (65–63 Ma) and led to crustal thickening and cessation of magmatism. Early Paleocene slab roll-back would have significantly enhanced asthenospheric corner flow and supplied a long-lived heat source for coeval crustal anatexis and metamorphism in southern Lhasa during the early phase of continental collision. Similar interaction between continental collision and oceanic subduction may also occur in other large-scale convergence zones in which the lithosphere and crust are anomalously hot.

INTRODUCTION

Leucogranites are a class of granite, usually with low (<5 vol. %) mafic mineral contents, that have been conventionally interpreted as the products of crustal melting. These granites are well developed in collisional orogens and are characterized by high silica and aluminium contents, and consequently usually contain aluminium-bearing minerals, such as garnet, tourmaline, muscovite and cordierite (Le Fort et al., 1987; Inger & Harris, 1993; Harrison et al., 1999; Patiño Douce, 2005; Brown, 2007; Zeng et al., 2011). This magmatic activity has been linked to a variety of geodynamic processes (e.g. crustal thickening, lithospheric delamination, back-arc extension, channel flow and mantle plumes (Harrison et al., 1997; Sylvester, 1998; Xu et al., 2008; King et al., 2011; Zeng et al., 2011). The crustal granitic melts represented by leucogranites can be used to help unravel the physical and geochemical response of the mid- to lower crust to the changing tectonic regime during the evolution of a large orogen (e.g. Patiño Douce & Johnston, 1991; Harrison et al., 1997; Ding et al., 2005; Lee & Whitehouse, 2007; King et al., 2011; Zeng et al., 2011).

The Tibetan Plateau formed in response to the continental collision of India with Asia and arguably represents one of the most important tectonic events during the Cenozoic. In the Tibetan Plateau leucogranites have been conventionally interpreted as the result of crustal anatexis and are considered as petrological evidence for collisional crustal processes, such as crustal flow, extension and thickening (Inger & Harris, 1993; Harrison et al., 1997; Ding et al., 2005; Lee & Whitehouse, 2007; King et al., 2011; Zeng et al., 2011). Most research on leucogranites in the orogen has focused on the more southerly Himalaya Block (Fig. 1a). In comparison, relatively little work has been completed on the detailed syn-collisional geology along the southern Lhasa Block, which represents the southernmost part of the Asian continent. This southern Lhasa Block is separated from the Himalaya Block by the Indus–Yarlung Tsangpo suture (IYTS) (Fig. 1a). In this contribution we focus on Cenozoic garnet-bearing two-mica granites and associated biotite granites in southern Tibet that potentially provide important constraints on the tectono-magmatic history of the orogen (e.g. Harrison et al., 1999; King et al., 2011; Zeng et al., 2011; Zhang et al., 2013).

Fig. 1.

(a) Geological map of the Lhasa Block [modified from Chung et al. (2009)]. SNMZ, Shiquan River–Nam Tso Mélange Zone; LMF, Luobadui–Milashan Fault; BNS, Bangong–Nujiang suture; IYTS, Indus–Yarlung Tsangpo Suture; MCT, Main Central Thrust; MBT, Main Boundary Thrust; STDS, South Tibet Detachment System. (b) Detailed occurrences of Paleocene magmatic rocks in the Xigaze–Nyingchi section of the Gangdese batholith. Yellow ovals mark the locations of zircon U–Pb isotopic ages of the Paleocene magmatic rocks (including published data and from this study) and further details are listed in Supplementary Data File 1. (c) Geological map of the Sangri area showing the sampling locations in this study and the location of published zircon U–Pb isotopic ages of the Gangdese magmatic rocks. Details are listed in Supplementary Data File 1.

A wide range of studies (using different dating methods) have presented evidence that the initial collision between India and Asia occurred in the early–middle Paleocene (c. 65–58 Ma) (e.g. Klootwijk et al., 1992; Beck et al., 1995; Yin & Harrison, 2000; Ding et al., 2005; Leech et al., 2005; Cai et al., 2011; Yi et al., 2011; DeCelles et al., 2014; Wu et al., 2014; Hu et al., 2015). However, existing work on Early Paleocene isobaric high-temperature granulite-facies metamorphism and significant arc magmatism in southern Lhasa often does not support an early–middle Paleocene collision event (e.g. Chung et al., 2005; Guo et al., 2012; Zhang et al., 2013; Zhu et al., 2015). Thus, the detailed crustal evolution and related dynamic processes in southern Tibet during the initial collision stage (possibly in the early–middle Paleocene) are highly controversial. A significant amount of Paleocene magmatism has been discovered in southern Lhasa over the last 10 years (e.g. Wen et al., 2008; Ji et al., 2009, 2014; Lee et al., 2009; Guo et al., 2012; Zhang et al., 2013; Jiang et al., 2014). Consequently, an improved knowledge of the relationship between this Paleocene magmatism and metamorphic events during the Indo-Asian collision process will help our understanding of the origin of the magmatism and the processes involved in continental collision.

The Paleocene leucogranites found in the eastern syntaxis show evidence of regional crustal anatexis and associated coeval granulite-facies metamorphism (Fig. 1b) (Guo et al., 2012; Zhang et al., 2013). Here we present detailed petrography, geochronology, major and trace element, and Sr–Nd–Hf–O isotopic data for the Zhengga leucogranites (garnet-bearing, two-mica granites) and associated biotite granites from the eastern segment of the Gangdese batholith. These new data suggest that large-scale Paleocene crustal anatexis probably occurred throughout southern Lhasa. U–Pb zircon dating of these granites yields Paleocene (c. 63 Ma) ages and isotopic data show slightly depleted Sr–Nd isotopic compositions. These granites provide us with an excellent opportunity to constrain the origin of leucogranites and the role of Neo-Tethyan slab subduction and detailed syn-collisional crustal evolution processes during the onset of the India–Asia collision. A similar interplay between subduction and collision may be a common feature in large-scale continental collisional orogenic belts worldwide.

GEOLOGICAL BACKGROUND AND ROCK CHARACTERISTICS

The Himalayan orogen of South Tibet formed in response to the Indo-Asian collision and is composed essentially of Himalayan sequences and the south Lhasa Block. These are separated by the Indus–Yarlung Tsangpo suture (IYTS) zone, which contains mafic rocks representing the remnants of the Neo-Tethyan Ocean (Fig. 1a) (Yin & Harrison, 2000, and references therein). South of the IYTS zone, two subparallel belts of Cenozoic leucogranites have been recognized, referred to as the High Himalayan (HH) and Tethyan Himalayan (TH) leucogranites (Fig. 1a). The High Himalayan leucogranites in the south form sheets, dykes, sills and laccolithic bodies that were emplaced along the South Tibetan Detachment System (STDS) (Le Fort, 1975; Le Fort et al., 1987; Harrison et al., 1999; Guo & Wilson, 2012). The Tethyan Himalayan leucogranites are typically exposed in the cores of North Himalayan domes within the central Tethyan Himalayan physiographic region (∼80 km north of the High Himalayan leucogranite belt) (King et al., 2011; Zeng et al., 2011; Liu et al., 2014). Previous research has shown that the High Himalayan leucogranites were generally emplaced during the latest Eocene–Middle Miocene from 35 to 12 Ma (e.g. Coleman, 1998; Harrison et al., 1999; Guo & Wilson, 2012; Liu et al., 2014) whereas the Tethyan Himalayan leucogranites were primarily intruded between 45 and 7 Ma (e.g. Zhang et al., 2004; Ding et al., 2005; King et al., 2011; Zeng et al., 2011; Liu et al., 2014; Liu et al., 2016a, 2016b; Wu et al., 2017). The Eocene leucogranites were derived by melting of thickened crustal materials and are found predominantly in the Tethyan Himalayan belt (e.g. Zeng et al., 2011; Liu et al., 2014).

The Lhasa Block was the last of a series of continental fragments accreted onto southern Asia during the Phanerozoic, before the collision of India (Allègre et al., 1984; Dewey et al., 1988; Yin & Harrison, 2000). Along the southern margin of the Lhasa Block (representing the southernmost part of the Asian continent), Late Triassic–Paleocene subduction of the Neo-Tethyan oceanic slab formed the Late Triassic–Early Tertiary Gangdese magmatic arc, comprising the Gangdese batholith and the Paleocene–Eocene Linzizong volcanic succession (e.g. Chung et al., 2005; Chu et al., 2006; Wen et al., 2008; Ji et al., 2009; Lee et al., 2009). The southern Lhasa sub-block (i.e. the Gangdese area) also contains minor Triassic–Jurassic volcano-sedimentary rocks (Fig. 1a). Additionally, Late Devonian–Early Carboniferous granites have been recently discovered in the area. These have been interpreted to be derived from partial melting of ancient continental crust, and are characterized by negative ε_Hf(t) values of zircon with T_DM2 ages ranging from 1·90 to 1·40 Ga (Ji et al., 2012; Dong et al., 2014).

Significant Paleocene to mid-Eocene (65–41 Ma) magmatic activity with a major age peak at c. 53–46 Ma occurred in the Gangdese magmatic arc (e.g. Yin & Harrison, 2000; Chung et al., 2005; Mo et al., 2006; Wen et al., 2008; Ji et al., 2009; Lee et al., 2009). This Paleocene–Eocene magmatism resulted from the tectonic transition from oceanic subduction to continental collision and is proposed to have been the result of both asthenosphere melting and crustal reworking (e.g. Chung et al., 2005; Mo et al., 2007; Ji et al., 2009; Lee et al., 2012; Zhang et al., 2013; Zhu et al., 2015). Limited data indicate that Paleocene medium- to high-grade metamorphic rocks, with associated leucogranites, mainly occur in the southeastern part of the Lhasa Block, especially in the Eastern Himalayan syntaxis (Guo et al., 2012; Zhang et al., 2013, 2015).

The Zhengga area is located in the Sangri County of Tibet and is less than 10 km north of the Yarlung Zangbo suture zone (Fig. 1c). Large-scale granitic intrusive bodies, which are constituent parts of the Gangdese batholith, are found in the area (Fig. 1c). These granites were intruded into both the Lower Jurassic Yeba Formation and the Devonian plutons (Fig. 1c). The Paleocene granites consist of undeformed biotite granites and minor garnet-bearing two-mica granites (Fig. 2a and b).

Fig. 2.

Field geological characteristics and petrography of the Zhengga garnet-bearing two-mica granites and biotite granites: (a) field outcrop of granitic intrusive rocks; (b) massive garnet-bearing granites; (c) main mineral assemblage of the garnet-bearing granite; (d) garnets associated with plagioclase and muscovite; (e) coexisting garnet, muscovite, biotite and quartz; (f) intergrown muscovite, biotite and plagioclase; (g) polycrystalline garnet; (h) main mineral assemblage of the biotite granite; (i) two types of biotite in the biotite granite. Pl, plagioclase; Grt, garnet; Bi, biotite; Ms, muscovite; Qtz, quartz; Cal, calcite.

The garnet-bearing two-mica granites and biotite granites generally exhibit massive and medium- to fine-grained (0·2–5·0 mm) granular textures, respectively (Fig. 2a and b). All these garnet-bearing granites have similar mineral compositions, which comprise quartz (∼40–45 vol. %), K-feldspar (∼25–30 vol. %), plagioclase (∼15–20 vol. %), and subordinate garnet (∼5 vol. %), biotite (∼3–5 vol. %) and muscovite (∼2–3 vol. %), with minor zircon, magnetite and daphnite (Fig. 2c–e;Table 1). Garnet crystals in the Zhengga garnet-bearing granites are euhedral or subhedral (Fig. 2c–e and g) and are associated with feldspar, biotite, muscovite and quartz (Fig. 2c–e), suggesting that the compositions of garnet should be in equilibrium with these coexisting minerals. The intergrowth of plagioclase, biotite and muscovite also further supports this (Fig. 2f). The Zhengga biotite granite, like the coeval Gangdese biotite granite, consists of plagioclase (∼25–30 vol. %), alkali feldspar (∼20–25 vol. %), quartz (∼35–40 vol. %) and biotite (∼5–10 vol. %) with minor Fe-oxides and titanite (Fig. 2h and i;Table 1). Biotite crystals in the Zhengga biotite granite can be divided into two types: (1) chloritized biotite, which is dark green both in plane- and cross-polarized light; (2) biotite, which is light brown in plane-polarized light and green and yellow when viewed under cross-polarized light. Plagioclase crystals commonly exhibit typical regular concentric growth zoning (Fig. 2h), and some plagioclase crystals are embedded in larger quartz grains. Microcline and perthite can be observed in garnet-bearing two-mica granites.

Table 1:

Samples and brief description

Sample	Rock type	Latitude	Longitude	Mineralogy
09TB66	GBG	92°10'31''	29°15'56''	Pl + Qtz + Kf + Gt + Bt + Ms + Chl
11SR01-1	GBG	92°10'30·1''	29°15'56·5''	Pl + Qtz + Kf + Gt + Bt + Ms
11SR01-2	GBG	92°10'30·1''	29°15'56·5''	Pl + Qtz + Kf + Gt + Bt + Ms
11SR01-3	GBG	92°10'30·1''	29°15'56·5''	Pl + Qtz + Kf + Gt + Bt + Ms
11SR05-2	GBG	92°10'57''	29°15'20''	Pl + Qtz + Kf + Gt + Bt + Ms
11SR05-3	GBG	92°10'57''	29°15'20''	Pl + Qtz + Kf + Gt + Bt + Ms
16ML06-1	GBG	92°10'32·7''	29°15'54·1''	Pl + Qtz + Kf + Gt + Bt + Ms + Mt
16ML06-2	GBG	92°10'32·7''	29°15'54·1''	Pl + Qtz + Kf + Gt + Bt + Ms
16ML07-1	GBG	92°10'31·5''	29°15'54·8''	Pl + Qtz + Kf + Gt + Bt + Ms
16ML08-1	GBG	92°10'30·2''	29°15'56·1''	Pl + Qtz + Kf + Gt + Bt + Ms
16ML08-2	GBG	92°10'30·2''	29°15'56·1''	Pl + Qtz + Kf + Gt + Bt + Ms
16ML09-1	GBG	92°10'30·8''	29°15'55·9''	Pl + Qtz + Kf + Gt + Bt + Ms
11SR03	BG	92°10'35''	29°15'52''	Pl + Kf + Qtz + Bt + Ttn
11SR04	BG	92°10'54''	29°15'25''	Pl + Kf + Qtz + Bt + Ms (min)
11SR05-1	BG	92°10'57''	29°15'20''	Pl + Kf + Qtz + Bt + Ms (min)
11SR06	BG	92°10'33''	29°16'04''	Pl + Kf + Qtz + Bt
16ML02-1	BG	92°10'34·2''	29°15'52·1''	Pl + Kf + Qtz + Bt + Ms (min)
11SR02-1	TMG	92°10'33''	29°15'54''	Pl + Kf + Qtz + Bt + Ms + Gt (min)
11SR02-3	TMG	92°10'33''	29°15'54''	Pl + Kf + Qtz + Bt + Ms

GBG, garnet-bearing two-mica granite; BG, biotite granite; TMG, (Devonian) two-mica granite; Pl, plagioclase; Kf, potassium feldspar; Qtz, quartz; Gt, garnet; Bt, biotite; Ms, muscovite; Chl, chlorite; Ttn, titanite; Mt, Ti-bearing magnetite.

RESULTS

The least altered samples were selected for geochemical and isotopic analysis. Whole-rock major and trace element and Sr–Nd isotope data are given in Tables 2 and 3, respectively. Analytical methods, mineral composition data, laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) and secondary ion mass spectroscopy (SIMS) zircon U–Pb geochronology and zircon Hf–O isotopic data for the Zhengga garnet-bearing granite and biotite granite samples are given in Supplementary Data Files 2–5; Supplementary Data are available for downloading at http://www.petrology.oxfordjournals.org.

Table 2:

Major (wt %) and trace element (ppm) data for the Zhengga leucogranites and associated rocks

Sample no.:	09TB66	11SR01-1	11SR01-2	11SR01-3	11SR05-2	11SR05-3	16ML06-1	16ML06-2	16ML07-1
Rock type:	GBG	GBG	GBG	GBG	GBG	GBG	GBG	GBG	GBG
SiO2	75·01	75·64	74·92	75·11	73·99	73·63	72·70	73·05	73·62
TiO2	0·05	0·02	0·03	0·03	0·03	0·06	0·04	0·01	0·06
Al2O3	14·39	14·46	14·73	14·46	14·59	14·93	15·22	15·04	14·42
Fe2${\mathrm{O}}_3^{\mathrm{T}}$	0·85	0·98	0·94	1·07	0·73	0·78	1·19	1·47	1·65
MnO	0·13	0·19	0·18	0·13	0·15	0·12	0·18	0·23	0·21
MgO	0·19	0·18	0·18	0·21	0·21	0·21	0·23	0·21	0·22
CaO	0·58	0·48	0·54	0·64	0·65	0·67	0·65	0·45	0·57
Na2O	3·76	4·31	4·35	4·16	4·28	4·20	4·72	4·33	4·30
K2O	4·29	3·98	4·40	4·36	4·72	4·65	4·54	4·82	4·48
P2O5	0·04	0·04	0·04	0·03	0·05	0·05	0·06	0·05	0·06
LOI	0·58	0·54	0·49	0·58	0·38	0·48	0·43	0·34	0·39
Total	99·87	99·87	99·87	99·87	99·87	99·87	99·87	99·87	99·87
A/CNK	1·21	1·18	1·14	1·14	1·09	1·13	1·10	1·14	1·11
Sc	1·18	3·48	4·18	5·41	6·20	5·33	2·41	2·59	2·82
V	13·6	11·8	17·1	16·3	12·7	18·9	5·02	3·58	5·81
Cr	1·64	66·5	108	153	130	150	7·56	12·5	29·2
Ni	1·77	4·97	6·01	7·26	6·29	6·27	1·02	1·97	2·33
Rb	127	127	147	123	142	144	156	145	158
Sr	51·5	29·6	36·4	43·5	49·5	53·1	44·5	22·9	41·0
Y	23·3	25·1	26·9	21·6	30·8	25·8	18·7	22·9	20·3
Zr	36·7	55·8	35·8	39·7	42·1	43·4	37·0	32·1	39·1
Nb	8·32	6·61	6·86	5·49	7·08	8·94	11·6	6·09	13·7
Cs	4·08	6·16	6·49	4·55	4·00	3·90	7·36	6·02	7·49
Ba	72·0	42·0	50·4	68·0	99·2	104	58·2	35·2	74·9
La	7·70	6·71	8·27	7·31	7·07	8·99	7·33	6·23	7·83
Ce	19·0	15·4	18·9	16·7	16·7	21·5	17·0	14·6	19·1
Pr	2·36	1·91	2·37	2·05	2·17	2·85	2·05	1·72	2·44
Nd	8·30	7·02	8·56	7·71	8·06	10·4	7·32	6·04	8·86
Sm	2·43	1·94	2·51	2·22	2·43	2·89	2·06	1·62	2·48
Eu	0·09	0·06	0·09	0·09	0·12	0·13	0·09	0·04	0·09
Gd	2·72	2·19	2·80	2·57	2·80	2·94	2·09	1·77	2·47
Tb	0·58	0·49	0·59	0·52	0·64	0·59	0·45	0·42	0·51
Dy	3·74	3·71	4·14	3·43	4·84	4·16	2·96	3·38	3·32
Ho	0·73	0·86	0·85	0·70	0·99	0·80	0·58	0·76	0·63
Er	1·81	2·32	2·17	1·61	2·32	1·96	1·41	1·90	1·50
Tm	0·25	0·34	0·29	0·23	0·30	0·27	0·20	0·28	0·21
Yb	1·49	2·18	1·82	1·44	1·82	1·69	1·22	1·62	1·29
Lu	0·19	0·34	0·27	0·22	0·26	0·25	0·15	0·19	0·16
Hf	1·61	2·80	1·68	2·01	2·09	2·06	2·00	1·84	2·14
Ta	1·10	0·82	0·88	0·88	0·83	0·95	1·21	0·71	1·27
Th	4·81	4·26	4·85	4·72	4·75	6·07	4·47	4·11	5·66
U	1·56	1·35	1·14	1·97	1·52	1·73	0·91	0·91	1·39
Eu/Eu*	0·10	0·09	0·10	0·12	0·14	0·14	0·14	0·08	0·11
TZr (°C)	629	661	619	627	626	633	614	608	622

Sample no.:	16ML08-1	16ML08-2	16ML09-1	11SR03	11SR04	11SR05-1	11SR06	16ML02-1
Rock type:	GBG	GBG	GBG	BG	BG	BG	BG	BG
SiO2	74·08	72·85	73·14	70·57	71·87	73·16	70·36	69·67
TiO2	0·05	0·05	0·05	0·26	0·15	0·14	0·23	0·26
Al2O3	14·11	14·47	14·36	15·06	14·83	14·76	15·28	15·21
Fe2${\mathrm{O}}_3^{\mathrm{T}}$	1·42	1·62	1·87	2·72	1·67	1·48	2·57	3·26
MnO	0·18	0·19	0·20	0·09	0·07	0·09	0·08	0·13
MgO	0·22	0·26	0·25	0·76	0·46	0·41	0·67	0·81
CaO	0·63	0·71	0·67	2·68	1·75	1·63	2·61	2·75
Na2O	4·21	4·74	4·16	3·97	3·57	4·65	3·95	4·11
K2O	4·68	4·55	4·91	3·12	4·79	2·87	3·31	3·44
P2O5	0·05	0·06	0·05	0·09	0·06	0·06	0·08	0·09
LOI	0·34	0·47	0·31	0·28	0·40	0·55	0·43	0·25
Total	99·87	99·87	99·87	99·87	99·87	99·87	99·87	99·87
A/CNK	1·07	1·03	1·07	1·02	1·04	1·08	1·03	0·98
Sc	3·46	3·16	2·73	4·41	6·14	5·72	4·72	4·22
V	5·31	6·74	5·94	33·7	22·2	21·0	34·3	28·7
Cr	6·86	11·9	25·5	100	112	115	125	14·2
Ni	2·97	1·98	2·55	5·75	6·37	6·29	6·19	2·40
Rb	145	149	135	85·3	119	111	79·4	91·4
Sr	49·7	53·7	47·5	294	211	200	309	327
Y	17·8	26·6	19·7	17·1	12·3	16·2	12·8	20·1
Zr	29·9	49·7	33·1	75·5	53·4	59·7	67·3	56·9
Nb	11·8	11·2	7·84	6·13	5·26	13·0	5·32	9·12
Cs	5·60	4·46	4·95	3·85	2·65	4·03	7·60	3·70
Ba	75·6	70·6	75·3	484	390	430	547	415
La	6·91	8·11	8·06	20·9	15·8	13·8	13·0	11·6
Ce	15·7	19·2	18·6	42·0	30·6	27·2	25·3	23·6
Pr	1·91	2·42	2·30	5·07	3·65	3·27	3·08	2·95
Nd	6·86	8·63	8·22	19·3	12·8	11·6	11·5	11·6
Sm	1·95	2·53	2·36	3·90	2·48	2·57	2·27	2·80
Eu	0·10	0·11	0·10	0·67	0·51	0·44	0·66	0·65
Gd	1·96	2·87	2·43	3·42	2·37	2·58	2·21	2·68
Tb	0·41	0·63	0·50	0·51	0·37	0·45	0·34	0·49
Dy	2·78	4·22	3·20	2·91	2·10	2·81	1·99	3·12
Ho	0·56	0·82	0·62	0·60	0·43	0·50	0·42	0·69
Er	1·40	1·93	1·46	1·74	1·05	1·20	1·28	1·92
Tm	0·21	0·27	0·21	0·27	0·16	0·18	0·20	0·32
Yb	1·26	1·56	1·23	1·90	1·10	1·21	1·53	2·23
Lu	0·16	0·20	0·16	0·32	0·18	0·20	0·28	0·36
Hf	1·62	2·59	1·78	2·45	1·82	2·17	2·12	2·20
Ta	0·97	0·98	1·43	0·67	0·69	1·35	0·55	1·12
Th	3·81	5·38	5·40	5·73	5·33	6·41	5·02	4·59
U	0·92	2·59	3·74	4·95	1·47	2·79	1·44	1·99
Eu/Eu*	0·16	0·12	0·13	0·56	0·64	0·52	0·90	0·72
TZr (°C)	598	630	604	660	636	652	651	629

Fe₂${\mathrm{O}}_3^{\mathrm{T}}$⁠, total Fe₂O₃ content; Mg^# = molecular Mg²⁺/(Mg²⁺ + Fe²⁺)×100; A/CNK = [Al³⁺/(Ca²⁺ + K²⁺ + Na²⁺)]; Eu/Eu* = Eu_N/(Sm_CN×Gd_CN)^1/2. LOI, loss on ignition. Subscript CN indicates chondrite normalization; the chondrite normalization values are from Sun & McDonough (1989). T_Zr values are calculated after Boehnke et al. (2013). Abbreviations as in Table 1.

Table 3:

Sr and Nd isotope data for the Zhengga leucogranites and associated rocks

Sample	Rock type	Rb	Sr	Sm	Nd	87Rb/ 86Sr	87Sr/ 86Sr ± 2σ	(87Sr/ 86Sr)i	147Sm/ 144Nd	143Nd/ 144Nd ± 2σ	T2DM (Ga)	(143Nd/ 144Nd)i	εNd(t)
09TB66	GBG	127	51·5	2·43	8·30	7·16	0·711362 ± 2	0·704957	0·177	0·512652 ± 8	1·08	0·512579	+0·43
11SR01-1	GBG	127	29·6	1·94	7·02	12·4	0·714894 ± 5	0·703770	0·167	0·512663 ± 3	1·04	0·512594	+0·73
11SR01-2	GBG	147	36·4	2·51	8·56	11·7	0·714156 ± 5	0·703683	0·177	0·512660 ± 3	1·06	0·512587	+0·59
11SR01-3	GBG
11SR05-2	GBG	142	49·5	2·43	8·06	8·33	0·711200 ± 5	0·703845	0·183	0·512660 ± 3	1·07	0·512585	+0·54
11SR05-3	GBG
16ML06-1	GBG								0·171	0·512667 ± 5	1·04	0·512598	+0·77
16ML06-2	GBG
16ML07-1	GBG								0·169	0·512666 ± 4	1·04	0·512598	+0·77
16ML08-1	GBG
16ML08-2	GBG	149	53·7	2·53	8·63	8·03	0·711362 ± 8	0·704289	0·177	0·512666 ± 4	1·05	0·512594	+0·70
16ML09-1	GBG	135	47·5	2·36	8·22	8·21	0·711433 ± 13	0·704201	0·174	0·512653 ± 7	1·08	0·512582	+0·47
11SR03	BG
11SR04	BG	119	211	2·48	12·8	1·63	0·706379 ± 5	0·704918	0·118	0·512631 ± 4	1·07	0·512582	+0·49
11SR05-1	BG
11SR06	BG	79	309	2·27	11·5	0·742	0·705642 ± 4	0·704978	0·119	0·512633 ± 4	1·07	0·512584	+0·53
16ML02-1	BG	91	327	2·80	11·6	0·808	0·705625 ± 12	0·704914	0·146	0·512676 ± 4	0·99	0·512617	+1·15
11SR02-1	TMG	209	142	4·90	22·8	4·24	0·727320 ± 8	0·723584	0·130	0·512039 ± 3	1·78	0·511986	–11·1
11SR02-3	TMG	254	131	5·40	26·4	5·60	0·732303 ± 8	0·727373	0·124	0·512044 ± 3	1·77	0·511993	–11·0

⁸⁷Rb/⁸⁶Sr and ¹⁴⁷Sm/¹⁴⁴Nd are calculated using whole-rock Rb, Sr, Sm and Nd contents in Table 1. ε_Nd(t) = [(¹⁴³Nd/¹⁴⁴Nd)_s/(¹⁴³Nd/¹⁴⁴Nd)_CHUR − 1]×10000. T_2DM = ln{[ε_Nd(t) − ε^DM_Nd(t)]/[(f_C − f_DM)] + e^λt}/λ (Liew & Hofmann, 1988). In the calculation, (¹⁴³Nd/¹⁴⁴Nd)_CHUR = 0·512638, (¹⁴⁷Sm/¹⁴⁴Nd)_CHUR = 0·1967, (¹⁴³Nd/¹⁴⁴Nd)_DM = 0·513151, (¹⁴⁷Sm/¹⁴⁴Nd)_DM = 0·2135, (¹⁴⁷Sm/¹⁴⁴Nd)_CC = 0 ·12, λ_Sm = 6·54 × 10^–12 a^–1 and t = 63 Ma. Abbreviations as in Table 1.

Mineral geochemistry

Representative electron microprobe analyses of plagioclase, alkali feldspar, muscovite, biotite and garnet are given in Supplementary Data File 3. The Zhengga garnet-bearing two-mica granites and biotite granites are characterized by markedly different mineral compositions.

The Zhengga biotite granites show a wide compositional range of feldspars including andesine (An_30–48Ab_51–69Or_1–11), oligoclase (An_12–30Ab_79–87Or_1–13), albite (An_1–3Ab_97–98Or₁), K–Na-feldspar (An_0–4Ab_12–42Or_58–85) and K-feldspar (An₀Ab_5–7Or_94–95), whereas feldspars from the Zhengga garnet-bearing two-mica granites consist mainly of oligoclase (An_10–21Ab_78–89Or₁), albite (An_3–9Ab_89–97Or_1–2) and K-feldspar (An₀Ab_2–5Or_95–98) (Supplementary Data File 3). Some plagioclase grains in the Zhengga granites show oscillatory compositional zoning (Fig. 3). Feldspars in the biotite granites have a core of alkali feldspar (An_0–4Ab_12–42Or_58–85) or andesine (An_30–48Ab_51–69Or_0–2) with a mantle or rim of oligoclase (An_17–29Ab_70–82Or_0–1). In contrast, feldspars in the garnet-bearing granites consist of a core of oligoclase (An_13–21Ab_78–86Or₁) and a rim of albite (An_6–9Ab_90–93Or_1–2) (Figs 3 and 4). Alkali feldspar from the Zhengga biotite granites is K-feldspar with a composition of An_0–4Ab_3–42Or_58–96 (Supplementary Data File 3). All these features suggest a crystallization sequence of andesine + K–Na feldspar (early), oligoclase + K-feldspar (later) and albite + K-feldspar (latest). In addition, plagioclases from the Zhengga garnet-bearing granites have mostly lower abundances of modal An (anorthite) compared with the biotite granites, suggesting a more evolved magma.

Fig. 3.

(a) Back-scattered-electron (BSE) image; (b) Al, (c) Na and (d) Ca chemical mapping of plagioclase grains from a Zhengga biotite granites.

Fig. 4.

Variations and core–mantle–rim structure of the plagioclase grains in a Zhengga biotite granite (a–c) and garnet-bearing two-mica granite (d–f). Circles indicate the locations of analyzed sites. Ab, albite; An, anorthite.

Biotites in the garnet-bearing two-mica granites are classified as lepidomelane with low to moderate MgO (4·3–6·8 wt %) and TiO₂ (0·2–1·2 wt %) contents, whereas biotites from the biotite granites are magnesian biotites with high MgO (9·6–10·8 wt %) and TiO₂ (2·1–3·1 wt %) (Supplementary Data File 3). Most of the muscovites in the Zhengga garnet-bearing two-mica granites are primary and also have low TiO₂ contents (0–0·57 wt %, most <0·14 wt %; Supplementary Data File 3). All garnets in the Zhengga garnet-bearing, two-mica granites are magmatic pyralspite that consist of almandine and spessartine with minor pyrope and andradite (Alm_23–59Sps_36–64Pyr_1–5And_0–8; Supplementary Data File 3). The magmatic garnets from the garnet-bearing, two-mica granites exhibit similar compositional zoning and are characterized by increasing Mn, decreasing Fe and oscillatory Ca from core to rim (Fig. 5). In addition, all garnet grains have a narrow rim that is richer in Mn (Fig. 5).

Fig. 5.

(a–c) Mn, (d–f) Fe and (g–i) Ca chemical mapping of garnet grains from the Zhengga garnet-bearing two-mica granites.

Zircon U–Pb geochronology

Three garnet-bearing granite samples and three biotite granites were selected for zircon dating. Zircons in these samples have crystal lengths of ∼150–300 μm and length:width ratios from 2:1 to 3:1. Zircon U–Pb isotopic data are given in Supplementary Data File 4. Well-developed oscillatory zoning and high Th/U ratios of zircons from the Zhengga granites indicate a magmatic origin (Hoskin & Black, 2000). U–Pb spot analyses on samples 11SR03, 11SR04 and 11SR05-1 yielded ²⁰⁶Pb/²³⁸U ages of 61–66 Ma (SIMS), 61–64 Ma (LA-ICP-MS) and 58–65 Ma (SIMS), with weighted-mean ages of 62·6 ± 0·5 Ma (MSWD = 1·4), 62·7 ± 0·7 Ma (MSWD = 0·24) and 62·7 ± 0·8 Ma (MSWD = 2·0), respectively, for the three biotite granite samples (Fig. 6a–c;Supplementary Data File 4). LA-ICP-MS U–Pb spot analyses on samples 09TB66, 11SR01-2 and 11SR05-2 yielded ²⁰⁶Pb/²³⁸U data of 61–64 Ma, 55–63 Ma and 61–63 Ma, with weighted-mean ages of 62·2 ± 0·6 Ma (MSWD = 0·18), 57·4 ± 1·6 Ma (MSWD = 2·1) and 61·5 ± 0·6 Ma (MSWD = 0·22), respectively, for the three garnet-bearing two-mica granites samples (Fig. 6d–f;Supplementary Data File 4). With the exception of a younger age of 57·4 Ma for the garnet-bearing, two-mica granite sample 11SR01-2, the other garnet-bearing, two-mica granite and biotite granite samples show consistent zircon U–Pb ages ranging from 62·7 to 61·5 Ma, indicating that the Zhengga leucogranites and biotite granites were probably emplaced in the early Paleocene (c. 62 Ma). It should be noted that some of the data for sample 11SR01-2 were excluded from Supplementary Data File 4 owing to possible loss of radiogenic Pb.

Fig. 6.

LA-ICP-MS and SIMS zircon U–Pb concordia diagrams with CL images for biotite granite (a) 11SR05-1, (b) 11SR03 and (c) 11SR04, and garnet-bearing granites (d) 11SR01-2, (e) 09TB66 and (f) 11SR05-2.

Major and trace element geochemistry

The Zhengga granite samples all have high SiO₂ contents, with the garnet-bearing, two-mica granites having higher SiO₂ and lower TiO₂, Fe₂${\mathrm{O}}_3^{\mathrm{T}}$⁠, MgO and CaO contents than the biotite granites (Table 2). All the Zhengga garnet-bearing granites have high K₂O (4·0–4·9 wt %) contents with markedly peraluminous compositions [A/CNK = molecular Al₂O₃/(CaO + Na₂O + K₂O) = 1·03–1·21] and have similar major element compositions to Miocene High Himalayan leucogranites (Fig. 7). In contrast, the biotite granites are metaluminous to slightly peraluminous (A/CNK = 0·98–1·08) with relatively low K₂O (3·1–4·8 wt %) and high CaO (1·6–2·8 wt %).

Fig. 7.

(a) SiO₂ vs Na₂O + K₂O classification diagram [after Middlemost (1994)]; (b) SiO₂ vs K₂O [after Peccerillo & Taylor (1976)]; (c) A/CNK vs A/NK classification diagram [after Shand (1951)]; (d) normative minerals (An–Ab–Or) of the Himalayan leucogranite [modified from Wu et al. (2014)]. All plotted data were recalculated to 100 wt % on a volatile-free basis. Data sources: Jurassic central Lhasa (CL) leucogranites are from Liu et al. (2006); Paleocene southern Lhasa (SL) igneous rocks are from Wen (2007), Lee et al. (2012), Zhang et al. (2013) and Jiang et al. (2014); Eocene Tethyan Himalaya (TH) granites are from Zeng et al. (2011) and Liu et al. (2014); Miocene High Himalaya (HH) leucogranites are from Zhang et al. (2004) and Guo & Wilson (2012).

The Zhengga granite samples exhibit a range of total rare earth element (REE) contents and are characterized by slightly to moderately enriched light REE (LREE) with flat chondrite-normalized heavy REE (HREE) patterns and variable negative Eu anomalies [Eu/Eu* = Eu_CN/(Sm_CN + Gd_CN)^1/2] (Fig. 8a). In more detail, the Zhengga garnet-bearing, two-mica granites show slightly enriched LREE patterns ([La/Sm]_CN = 1·9‒2·5) with significant negative Eu anomalies (Eu/Eu* = 0·1–0·2). In contrast, the biotite granites have more enriched LREE patterns ([La/Sm]_CN = 2·7‒4·1) with moderate negative Eu anomalies (Eu/Eu* = 0·5–0·9) that are not as marked as those for the garnet-bearing granites (Fig. 8a, Table 2).

Fig. 8.

Primitive mantle-normalized trace element and chondrite-normalized REE patterns of the Zhengga garnet-bearing granites and biotite granites. Data sources: Jurassic central Lhasa (CL) leucogranites are from Liu et al. (2006); Paleocene southern Lhasa (SL) I-type granites are from Wen (2007), Lee et al. (2012), Zhang et al. (2013) and Jiang et al. (2014); Paleocene southern Lhasa (SL) S-type granites are from Zhang et al. (2013); Eocene Tethyan Himalaya (TH) granites are from Zeng et al. (2011); Miocene High Himalaya (HH) leucogranites are from Guo & Wilson (2012). Chondrite and primitive mantle normalization values are from Sun & McDonough (1989).

On primitive mantle-normalized plots (Fig. 8b), the Zhengga granites are enriched in large ion lithophile elements (LILE), such as Rb, Th and U, and have slightly depleted high field strength elements (HFSE), such as Nb, Zr and Ti. They have negative Ti anomalies [Ti/Ti* = Ti_PM×2/(Sm_PM + Tb_PM) = 0·02–0·22] and variable Sr anomalies [Sr/Sr* = Sr_PM×2/(Ce_PM + Nd_PM) = 0·17–1·42]. It should also be noted that the Zhengga garnet-bearing, two-mica granites have moderately negative Nb anomalies [Nb/Nb* = Nb_PM×2/(Th_PM + La_PM) = 0·23–0·60] but with relatively high [Nb/La]_PM (0·72–1·69, mean 1·13) ratios, which are different from those for arc magmatic rocks.

Sr‒Nd isotope geochemistry

Overall, the Zhengga granites have very similar initial ⁸⁷Sr/⁸⁶Sr ratios (0·7037–0·7050) and ε_Nd(t) values (+0·43 to +1·15) (Fig. 9a;Table 3). These isotopic signatures distinguish the Paleocene granites from the Devonian granites as the latter have more enriched Sr–Nd isotopic compositions ([⁸⁷Sr/⁸⁶Sr]_i = 0·7238–0·7470 and ε_Nd(t) = −11·1 to −10·0; Table 3). Because of the variable ¹⁴⁷Sm/¹⁴⁴Nd of the Zhengga granites, two-stage Nd model ages (T_2DM) are used in this study (Liew & Hofmann, 1988). The two-stage Nd model age is an age at which the isotopic composition of the sample was identical to that of a model reservoir (most often CHUR or depleted mantle) and compensates for the effects of possible secondary Sm/Nd fractionation as a result of intra-crustal partial melting (Liew & Hofmann, 1988). In terms of two-stage Nd model ages (T_2DM) (based on the depleted mantle), the garnet-bearing granites and biotite granites yield almost the same results (1·04–1·08 Ga and 0·99–1·07 Ga, respectively) (Table 3).

Fig. 9.

(a) ε_Nd(t) vs (⁸⁷Sr/⁸⁶Sr)_i diagram for the Zhengga garnet-bearing granites and biotite granites. Data sources are as in Fig. 7. Additional data sources: Yarlung Tsangpo ophiolites from Mahoney et al. (1998) and Zhang et al. (2005); Miocene HH leucogranites from Guo & Wilson (2012); Eocene TH granites from Zeng et al. (2011) and Liu et al. (2014); Eocene to Miocene SL adakites from Chung et al. (2003), Hou et al. (2004), Guo et al. (2007) and Jiang et al. (2014); Cretaceous SL magmatic rocks from Wen et al. (2008), Ma et al. (2013a, 2013b, 2013c, 2015). The mixing end-members are: SL juvenile lower crust [09TB79: ε_Nd(t) = +3·8, (⁸⁷Sr/⁸⁶Sr)_i = 0·7045, Nd = 6·16 ppm and Sr = 852·4 ppm] from Ma et al. (2013b); global subducting sediment [GLOSS: ε_Nd(t) = −8·6, (⁸⁷Sr/⁸⁶Sr)_i = 0·7170, Nd = 27 ppm, Sr = 327 ppm] from Plank & Langmuir (1998); Devonian granite (11SR02-3) representing the ancient Gangdese crust [ε_Nd(t) = −11·0, (⁸⁷Sr/⁸⁶Sr)_i = 0·7274, Nd = 26 ppm, Sr = 131 ppm; Table 3]; HH crystalline basement [NL07: ε_Nd(t) = −15·6, (⁸⁷Sr/⁸⁶Sr)_i = 0·7357, Nd = 11 ppm, Sr = 89 ppm] from Guo & Wilson (2012). (b) Zircon δ¹⁸O vs ε_Hf(t) diagram. Average δ¹⁸O value of mantle magmas (5·3 ± 0·3‰, 1SD) is from Valley et al. (1998). Data sources: Cretaceous Zhengga gabbros from Ma et al. (2013a); Paleocene Yajie gabbros from Wang et al. (2015). (c) Zircon ε_Hf(t) vs ²³⁸U/²⁰⁶Pb age diagram. Data sources: Gangdese batholith from Ji et al. (2009) and Zhu et al. (2011); I-type granites from Ji et al. (2014); Paleocene I-type granites from Zhang et al. (2013), Wang et al. (2015) and Ji et al. (2014); Paleocene S-type granites from Zhang et al. (2013). (d) Zircon δ¹⁸O vs Nb/U diagram.

Zircon Hf‒O isotopic geochemistry

In situ zircon Hf isotope data for sample 09TB66 and O isotope data for samples 09TB66, 11SR01-2, 11SR04 and 11SR05-1 are listed in Supplementary Data File 5. Zircons from sample 09TB66 (garnet-bearing, two-mica granite) have positive ε_Hf(t) values (+5·5 to +11·7), which are markedly higher than those of zircons from coeval S-type granites located near the Eastern Himalayan Syntaxis (Fig. 9c). In addition, zircons from sample 09TB66 also have young Hf model ages (0·27–0·52 Ga). All the Zhengga garnet-bearing granites and biotite granites have similar zircon δ¹⁸O values (5·5–7·3‰) (Fig. 9d), which are slightly higher than those (5·3 ± 0·3‰) of igneous zircons in equilibrium with mantle-derived magmas (Valley et al., 1998; Valley, 2003) (Fig. 9b).

DISCUSSION

Petrogenesis of the Zhengga granites

Biotite granites

The Zhengga biotite granites are characterized by high silica, moderate amounts of K₂O and low MgO contents with slightly peraluminous compositions and thus are unlikely to be derived from mantle-derived magma by fractional crystallization. Crust-derived granitic magmas are either supracrustal (formed from metasedimentary rocks) or infracrustal (formed from igneous rocks that solidified at depth) in character; this is reflected in the S- and I-type notation used to describe these granites (White & Chappell, 1977; Kemp et al., 2007). P₂O₅ is particularly effective in distinguishing between I- and S-type granitoids, as it decreases with fractional crystallization in I-type granites and increases with fractionation in S-type granites (Chappell & White, 1992). Decreasing P₂O₅ with increasing SiO₂ (Fig. 10c) suggests that both Zhengga granite types are I-type and not S-type granites.

Fig. 10.

SiO₂ vs (a) Al₂O₃, (b) Fe₂${\mathrm{O}}_3^{\mathrm{T}}$ (total Fe₂O₃ content), (c) P₂O₅, (d) Eu/Eu*, (e) Nb/La and (f) Sr/Y for the Paleocene Zhengga granites. The trends of I-type and S-type granites are from Chappell & White (1992).

In addition, although isotopic compositions overlap with I-types (Chappell & White, 1992), S-type granites have generally higher δ¹⁸O and ⁸⁷Sr/⁸⁶Sr values with lower ¹⁴³Nd/¹⁴⁴Nd isotopic compositions. Unlike whole-rock data, zircon is able to preserve the magmatic O isotope composition (Peck et al., 2003), and igneous zircons in equilibrium with mantle-derived magmas have an average δ¹⁸O value of 5·3 ± 0·3‰ (1SD) (Valley, 2003). This range is insensitive to magmatic differentiation, because the attendant rise in bulk-rock δ¹⁸O is compensated by an increase in zircon/liquid δ¹⁸O fractionation from +0·5‰ for mafic melts to +1·5‰ for silicic derivatives (Valley et al., 2005). Values of δ¹⁸O in zircon above 5·6‰ thus fingerprint an ¹⁸O-enriched supracrustal component in the magma from which the zircon crystallized, this being either sedimentary rock (10–30‰) or altered volcanic rock (up to 20‰) (Eiler, 2001, 2007). Given that intragrain heterogeneity of δ¹⁸O usually does not exceed 1‰ (Kemp et al., 2007), the relatively variable and low zircon δ¹⁸O (5·6–6·9‰) values of the Zhengga biotite granites therefore track the progressive interaction between two end-member components during zircon crystallization, these being parental low δ¹⁸O magmas and metasedimentary-derived materials.

The Latest Triassic to Cretaceous Gangdese granitoids have high and positive ε_Nd(t) (up to +5·5) and ε_Hf(t) (up to +16·5) values (Chu et al., 2006; Wen et al., 2008; Ji et al., 2009; Zhu et al., 2011; Ma et al., 2013b, and references therein), indicating a juvenile crust beneath the southern Lhasa terrane. Because Late Cretaceous Zhengga gabbros that were intruded into the lower crust are also exposed in the same region (Fig. 1c;Ma et al., 2013b), we propose that the Gangdese juvenile crust represented by the Zhengga gabbros is the most likely source of the low-δ¹⁸O end-member (Fig. 9b). However, it is significant that many magmatic rocks with relatively low ε_Nd(t) and ε_Hf(t) values were emplaced in the southern Lhasa terrane during the Paleocene (Fig. 9a and c). As mentioned above, owing to their relatively high and variable zircon O isotope and low ε_Nd(t) values, a contribution from more enriched components to the source of the Zhengga biotite granites also needs to be considered. Sr–Nd isotope modelling reveals that subducted pelagic sediments, Indian crust and ancient Lhasa basement all are likely potential enriched end-members (Fig. 9a).

The Zhengga granitic magmas are unlikely to have experienced significant crustal contamination for the following reasons: (1) ε_Nd(t) (+0·4 to +1·2) and (⁸⁶Sr/⁸⁷Sr)_i (0·7037–0·7050) values are homogeneous (Table 3), and there is a lack of correlation between ¹⁴⁷Sm/¹⁴⁴Nd and initial ¹⁴³Nd/¹⁴⁴Nd ratios (Supplementary Data File 6) (Vervoort & Blichert-Toft, 1999); (2) given that crustal components generally exhibit distinctly low ε_Nd(t), MgO, low Nb/La and Nb/Th values, and high ⁸⁷Sr/⁸⁶Sr ratios (Rudnick & Fountain, 1995), any crustal assimilation that occurred during magma ascent would have caused an increase in (⁸⁷Sr/⁸⁶Sr)_i and a decrease in ε_Nd(t) in the magma suites (e.g. Rogers et al., 2000). The (⁸⁷Sr/⁸⁶Sr)_i, (¹⁴³Nd/¹⁴⁴Nd)_i and Nb/La values of the Zhengga biotite granites show no significant correlation with increasing SiO₂ contents (Fig. 10e and Supplementary Data File 6), which is also inconsistent with significant crustal assimilation.

Previous studies have shown that melts derived from subducted oceanic sediment contain high concentrations of both Th and LREE (e.g. Hawkesworth et al., 1997). However, the lack of significantly higher Th and LREE contents and/or elevated Th/U ratios suggests insignificant involvement of oceanic sediment in the generation of the Zhengga biotite granites (Fig. 8; Table 2). The associated Devonian granites (Fig. 1c) are probably substantially derived from ancient Lhasa basement (Dong et al., 2014), and thus are characterized by similar Sr–Nd isotopic compositions (ε_Nd(t) = −11·1 to −11·0; (⁸⁷Sr/⁸⁶Sr)_i = 0·7236–0·7274) (Fig. 9a;Table 3) to the Paleocene S-type granites in the Eastern Himalayan Syntaxis (Zhang et al., 2013). We therefore favour the involvement of ancient Lhasa basement in the generation of the Zhengga biotite granites. Modelling of equilibrium (batch) melting also indicates that the high-degree (c. 10–50%) melting of 95% Zhengga gabbro (Ma et al., 2013b) plus 5% two-mica granite can produce melts compositionally similar to the Zhengga biotite granites (Fig. 11), further supporting this model. The details of partial melt modelling are listed in Supplementary Data File 7. Alternatively, it is also possible that another crustal component, such as Indian continental crust (Jiang et al., 2014), might have contributed to the source of these granites, although lack of other solid evidence to support direct contact of India with this region during the early Paleocene would seem to preclude this possibility.

Fig. 11.

Equilibrium (batch) melting modelling results for Zhengga biotite granites from southern Lhasa. Melting of 95% juvenile mafic lower crust (average of the Zhengga gabbros; Ma et al., 2013b) + 5% ancient crust (average of the Devonian granites given in Supplementary Data File 7). Patterns with symbols are modelling results (5% melting increments). The source composition is 40% plagioclase, 45% amphibole and 15% clinopyroxene. Partition coefficients of plagioclase amphibole and clinopyroxene are from McKenzie & O’Nions (1991). REE data for the initial material and partition coefficients are given in Supplementary File 7.

The Zhengga biotite granites also have moderately negative Eu anomalies (Eu/Eu* down to 0·5) indicating plagioclase and/or K-feldspar fractionation. Fractionation of plagioclase would result in negative Sr–Eu anomalies, and that of K-feldspar would produce negative Eu–Ba anomalies (Hanson, 1978). In log–log diagrams of Ba vs Sr and Ba/Sr vs Sr (Fig. 12), the Sr and Ba concentrations of the Zhengga biotite granites decrease moderately in the early stage of crystallization but decrease sharply in the later stages. These could be explained by the early fractionation of plagioclase, followed by the later fractionation of K-feldspar and plagioclase, rather than biotite (Fig. 12). Details of the fractional crystallization modelling are given in Supplementary Data File 7. Plagioclase crystals from the biotite granites show oscillatory compositional zoning with a decreasing An (anorthite) component from andesine cores to oligoclase mantles and rims (Figs 3 and 4). These complex zoning patterns are most probably related to long-lived magma recharge and fractionation. The early crystals have undergone resorption and overgrowth in later more evolved magma.

Fig. 12.

(a) Ba/Sr vs Sr and (b) Ba vs Sr. The grey dashed lines are composition trends showing that fractionation of plagioclase plays an important role in the earliest stage of differentiation; whereas in the later stage, separation of K-feldspar and minor plagioclase appears to have controlled the variation of these elements. Partition coefficients and calculated composition of K, Rb, S and Ba are listed in Supplementary Data File 7.

Calculation of zircon saturation temperatures (T_Zr) using the updated equation proposed by Boehnke et al. (2013) yields 629–660°C (Table 2). A study of granites from the Cordillera in the USA and from New Zealand indicate that their T_Zr are universally lower (<800°C) than those expected from dehydration melting of mafic crust (T >900°C) (Collins et al., 2016). Accordingly, Collins et al. (2016) proposed that silicic Cordilleran magmas form in magmatic arcs where hydrous basaltic magmas solidify in the arc root, producing a mafic underplate that emits aqueous fluids that transfer to the crust and promote water-fluxed partial melting at ambient temperature and pressure (∼750–800°C at 8 kbar) conditions. In addition, low Zr (53–76 ppm) contents and lack of inherited zircon grains in the Zhengga biotite granites are consistent with significant fractional crystallization of zircon or a source undersaturated in zircon (Table 2 and Supplementary Data File 4). These inheritance-poor granitoids are undersaturated in zircon and hence the calculated T_Zr (<800°C) is likely to reflect the minimum emplacement temperature, which is lower than initial generation temperature (Miller et al., 2003). Thus, we use alkali feldspar (An_0–4Ab_12–42Or_58–85) and andesine (An_30–48Ab_51–69Or_0–2) as combinations of early feldspars to estimate the formation temperature of the Zhengga biotite granites. Calculations using the two-feldspar (plagioclase and alkali feldspar) thermometer of Putirka (2008) indicate that the Zhengga biotite granites have high initial formation temperatures of 889–998°C, which are consistent with coeval high-T (>800°C) granulite-facies metamorphism in the Eastern Himalayan Syntaxis (Guo et al., 2012; Zhang et al., 2013).

Mineral assemblages in the crustal source of the felsic rocks can be further constrained by their geochemical characteristics. Given that plagioclase is enriched in Sr, and garnet is depleted in LREE and enriched in HREE and Y, the moderately negative to positive Sr (Sr*/Sr = 0·7–1·4) anomalies (Fig. 8b) and low La/Yb and Sr/Y ratios of the Zhengga biotite granites (Fig. 13a;Table 2) reflect a plagioclase-rich source containing little or no garnet (Rapp & Watson, 1995; Qian & Hermann, 2013). Pressure–temperature conditions for partial melting of crustal rocks based on experimental data are summarized in Fig. 13b. They indicate that garnet forms at pressures of 0·4–0·6 GPa and temperatures of 750–900°C during partial melting of metasedimentary rocks (Clemens & Wall, 1981; Stevens et al., 1997; Patiño Douce & Harris, 1998; Wang et al., 2012), and that the lower limit of garnet stability is 0·5 GPa (line 9 in Fig. 13b). Plagioclase is a common residual mineral during partial melting of metasedimentary and igneous rocks (such as tonalites and basalts), but it disappears at pressures >1·2–1·5 GPa (Rapp & Watson, 1995; Rapp et al., 2003; Patiño Douce, 2005; Xiong et al., 2005) (Fig. 13b). Thus, these features suggest that the formation of the Zhengga biotite granites occurred over a range of pressures from 0·5 to 1·2 GPa, consistent with magmatic epidote from the Zhengga gabbros (Ma et al., 2013b). In addition, most of the Paleocene magmatic rocks are also characterized by low La/Yb and Sr/Y ratios relative to the latest Cretaceous granites (Fig. 13a), suggesting a relatively thin crust in the southern Lhasa Block during the Paleocene, which is also supported by previous work (Ji et al., 2014).

Fig. 13.

(a) La/Yb vs Sr/Y [after Wang et al. (2016)]. This diagram indicates the effects of residual garnet (Grt) and plagioclase (Pl) during partial melting (for discussion see the text). Field 1: adakitic melts derived from eclogitic rocks in the stability field of garnet with little or no plagioclase; Field 2: crustal melts in the stability field of plagioclase and garnet; Field 3: crustal melts in the stability field of plagioclase with little or no garnet. (b) Pressure–temperature conditions for partial melting of crustal rocks. Curves or lines for melting or mineral stability: (1) H₂O-saturated or wet solidus curve of crustal rocks (Patiño Douce, 2005; Xiong et al., 2005; Qian & Hermann, 2013); (2) dehydration solidi curve for crustal rocks (17, 47); (3) dry solidus curve (Qian & Hermann, 2013); (4) garnet-in/out curve during dehydration melting of metabasalts (Rapp & Watson, 1995; Patiño Douce & Harris, 1998; Rapp et al., 2003; Patiño Douce, 2005; Xiong et al., 2005; Nair & Chacko, 2008); (5) garnet proportion contours (5 and 20 wt %) in the residue of metabasalt melting (Nair & Chacko, 2008); (6) rutile-in/out curve (Xiong et al., 2005); (7) plagioclase-in (beneath curve)/out (above curve) curve during dehydration melting (Rapp et al., 2003; Qian & Hermann, 2013); (8) orthopyroxene-in (beneath curve)/out (above curve) curve during dehydration melting (Qian & Hermann, 2013); (9) the boundary line for garnet-bearing (above line) and cordierite-rich (beneath line) melts during dehydration melting of metasedimentary rocks (Green, 1977; Clemens & Wall, 1981; Wang et al., 2012); (10) upper thermal stability of F-free mica in pelite, wacke, mica schist and tonalite (Patiño Douce & Harris, 1998; Patiño Douce, 2005; Hacker et al., 2014).

In summary, all these features support a model in which the Zhengga biotite granites were generated by high degrees of partial melting (10–50%) of Gangdese juvenile middle to lower crust with minor (∼5%) ancient Lhasa crustal basement (represented by two-mica granite) in the temperature range 889–998°C. These melts subsequently underwent magma recharge and fractional crystallization of K-feldspar and plagioclase in deep (15–40 km) crustal magma chambers.

Garnet-bearing two-mica granites

Most leucogranites are characterized by high silica and alumina, and usually contain aluminium-bearing minerals, such as garnet, tourmaline, muscovite and cordierite (Le Fort, 1975). Thus, leucogranite is recognized as being a typical product of partial melting of metasediments (e.g. Le Fort, 1981; Patiño Douce & Harris, 1998; Harrison et al., 1999; Searle et al., 2010; Guo & Wilson, 2012). This model is supported by Cenozoic Himalayan leucogranites that are characterized by high silica contents (>70 wt %), low TiO₂, MgO and total Fe₂O₃ (<1 wt %) contents, peraluminous compositions (A/CNK >1·1), high initial ⁸⁷Sr/⁸⁶Sr ratios of 0·73–0·78, and low ε_Nd(t) values of −10 to −15. It has been proposed that these granites are derived from the metapelites of the High Himalayan Crystalline Sequence (e.g. Inger & Harris, 1993; Harrison et al., 1999; Murphy, 2007; King et al., 2011; Guo & Wlison, 2012).

The Paleocene Zhengga garnet-bearing, two-mica granites are characterized by relatively depleted Sr–Nd isotopic composition [ε_Nd(t) = +0·43 to +0·77; (⁸⁷Sr/⁸⁶Sr)_i = 0·7037–0·7050], which clearly distinguishes them from the Devonian ancient crust-derived granites and the Miocene High Himalayan leucogranites (Fig. 9a). In addition, the Paleocene Zhengga garnet-bearing, two-mica granites have relatively low δ¹⁸O values (5·5–7·3‰) similar to the Zhengga biotite granites and gabbros (Fig. 9d;Supplementary Data File 5), which are inconsistent with a metasedimentary source. High ε_Hf(t) (+5·5–+11·7) values of the garnet-bearing granite sample 09TB66 also support this inference (Supplementary Data File 5). These features, combined with the nearly identical Sr and Nd isotopic compositions (Fig. 9a) of the garnet-bearing, two-mica granites and the coeval biotite granites, indicate that they are likely to be derived from a similar source.

Compared with the biotite granites, the Zhengga garnet-bearing granites have higher silica contents (72·7–75·6 wt %; Fig. 7a), markedly lower concentrations of other major oxides (such as TiO₂, MgO, Fe₂${\mathrm{O}}_3^{\mathrm{T}}$ and CaO), and display marked depletions in Ba, Sr, Ti and Eu (Sr/Sr* = 0·17–0·34; Ti/Ti* = 0·02–0·06; Eu/Eu* = 0·08–0·16; Table 2), suggesting that they have experienced significant fractionation. The biotite and muscovite in the garnet-bearing granites also both have low TiO₂ contents (0·04–1·21 wt % and 0·03–0·14 wt %, respectively) (Supplementary Data File 3), which further support extensive fractionation. Plagioclase from the garnet-bearing granites consists of a core of oligoclase (An_13–21Ab_78–86Or₁) and rim of albite (An_6–9Ab_90–93Or_1–2) (Fig. 4) and has mostly lower An (anorthite) contents compared with the biotite granites, also suggesting a more evolved magma.

Negative Ti anomalies indicate fractionation of Ti-bearing phases (e.g. titanite, ilmenite and rutile). Extensive fractionation of K-feldspar with minor plagioclase is responsible for marked Eu depletion in these rocks (Fig. 8). Given that biotite crystallization in peraluminous magmas is favoured by elevated Fe, Mg and Ti contents (Scaillet et al., 1995), the low Fe and Mg contents are interpreted to result from fractionation of biotite and amphibole. Zircon and apatite both have high partition coefficients for the REE, whereas garnet has very low partition coefficients for the LREE and large partition coefficients for the HREE. In previous studies it has been proposed that REE contents are mainly controlled by accessory minerals such as zircon, apatite, allanite and monazite (e.g. Wu et al., 2003; Chu et al., 2009). However, given the insignificant depletion of the HREE (Fig. 8a) in these rocks, we contend that these accessory minerals were not significant fractionating phases during the generation of the Zhengga garnet-bearing granites.

The Zhengga garnet-bearing, two-mica granites are characterized by abundant almandine-rich garnet and muscovite and are different from the biotite granites (Fig. 2; Table 1; Supplementary Data File 3). Previous experimental work has proposed two main models for the origin of garnet in peraluminous granitoids:

1. Muscovite + Biotite + 3Quartz = Garnet + 2K-feldspar + 2H₂O (e.g. Abbott & Clarke, 1979; Clemens & Wall, 1981);

2. Biotite + (Mn, Al)-rich liquid = Garnet + Muscovite (Miller & Stoddard, 1981).

Given the presence of garnet and muscovite (Table 1) and the decreasing abundance of biotite, we propose that the second model best explains the formation of garnet and muscovite in the Zhengga garnet-bearing two-mica granites. Plagioclase (mostly albite) in the garnet-bearing granites has lower An and higher Al₂O₃ contents compared with the biotite granites (Supplementary Data File 3) and all magmatic garnets show spessartine enrichment (Alm_23–59Sps_36–64Pyr_1–5And_0–8; MnO = 12·5–27·4 wt %; Supplementary Data File 3), consistent with an evolved (Mn, Al)-rich liquid that supports model 2 for the formation of garnet and muscovite.

Garnets from the Zhengga garnet-bearing granites have low pyrope and relatively high almandine and spessartine contents (Supplementary Data File 3). Pyrope-rich garnet generally coexists with high-temperature (>950°C) melts (Green, 1977). Thus, the presence of pyrope-poor garnet in the garnet-bearing granites suggests relatively low temperatures (<950°C), consistent with low zircon saturation temperatures (598–661°C; Table 2). Nevertheless, given relatively low Zr contents and negative Zr anomalies (Fig. 8b;Table 2), suggesting significant fractionation of zircon and undersaturation in Zr, the zircon saturation temperatures are likely to be underestimated (by up to +50°C) (Miller et al., 2003).

It has been proposed that Mn enrichment enhances the stability of garnet in magmas (Green, 1977; Clemens & Wall, 1981; Miller & Stoddard, 1981). The MnO and CaO contents of experimentally crystallized garnets show an antithetic relationship (Green, 1977), consistent with garnets in the Zhengga garnet-bearing granites (Fig. 5; Supplementary Data File 3). The grossular content of crystallizing garnets increases with increasing pressure (Green, 1977). Thus, the low grossular and high spessartine contents of the garnet suggest that the Zhengga garnet-bearing, two-mica granites were formed at relatively low pressure. Experimental data indicate that almandine-rich garnet is most stable in silicic melts with 3–5 wt % water at T <850°C and P >4–5 kbar (e.g. Clemens & Wall, 1981). Moreover, spessartine-rich garnets could be stable in silicic magmas to pressures as low as 3 kbar (Green, 1977). The magmatic garnets in the garnet-bearing, two-mica granites exhibit similar compositional zoning (i.e. Alm_54–59Sps_36–51) and are characterized by increasing Mn, decreasing Fe and oscillatory Ca from core to rim, with a narrow outermost rim that is richest in Mn (Fig. 5). These compositional features are most likely the result of crystallization from a highly evolved magma (Clemens & Wall, 1981) and/or at low pressure (Green, 1977).

In summary, the garnet-bearing, two-mica granites are likely to represent the products of advanced fractionation of the biotite-granite magmas at relatively high temperature (∼650–700°C) in magma chambers at mid-crust depths (∼12–18 km) during the latest stage of magma evolution. Their compositional features can be explained by fractionation of K-feldspar, biotite and titanite with minor plagioclase and accessory minerals, such as zircon, apatite, allanite and monazite. Garnet and muscovite were most probably formed by reaction between biotite and highly evolved (Mn, Al)-rich melt.

Paleocene magmatism and diachronous continental collision

The distribution of Paleocene magmatism within the Lhasa Block of South Tibet is summarized in Supplementary Data File 1. An early Paleocene linear magmatic and metamorphic belt stretches for over 1500 km to the north of the Indus–Yarlung Tsangpo suture; this comprises an eastern segment from Xigaze to Nyingchi (between 95 and 86°E) (Fig. 1b) and a western segment (between 82 and 80°E) with a quiescent magmatic area in the central Gangdese (between 86 and 82°E) (Supplementary Data File 1). Given the Paleocene magmatism and coeval magmatic quiet or quiescent area in the central Gangdese, we advocate a diachronous continent–continent collision model in which India collided first with the middle part of southern Tibet at ∼65–63 Ma, followed by suturing progressing both westward and eastward at ∼55–50 Ma (Fig. 14).

Fig. 14.

A model for interaction between diachronous continental collision and subducted oceanic slab roll-back causing crustal anatexis, crustal thickening, metamorphism and possible crustal flow along the southern margin of the Tibetan Plateau during the Early Paleocene (c. 63 Ma). The Paleocene biotite granite magmas were formed by anatexis in the middle to lower crust (∼40–15 km) and subsequently some upwelled to middle to upper crust levels (∼20–10 km), which resulted in widespread large-scale granitic chambers or magma reservoirs in lower to shallow crust levels (40–10 km). The Zhengga garnet-bearing two-mica granites represent highly evolved magmas that formed in shallow crustal chambers. Greater India boundary is after Hall (2002), Zhou & Murphy (2005) and Ali & Aitchison (2005).

In the early Paleocene (∼65–63 Ma), the northernmost part of the Greater Indian continent first collided with the Asian continent in central Lhasa (approximately between 86 and 82°E; Fig. 14). The resistance generated by this initial collision event between the Indian and Asian continents and buoyancy forces owing to the lower density of the continental crust are likely to have resulted in a change in subduction velocity and angle (Buck & Toksöz, 1983; Shellnutt et al., 2014). This change in subduction velocity is consistent with the abrupt deceleration of the convergence rate between India and Asia from 170 to 100 mm a^–1 at ∼60 Ma (e.g. Lee & Lawver, 1995). Collision led to crustal thickening and cessation of central Gangdese magmatism (Zhu et al., 2017). An initial collision event between Asian and Indian continental crustal blocks in the Early Paleocene (c. 65–63 Ma) is also supported by the following:

1. Hf isotope and geochronological data of detrital zircons from the Sangdanlin and Gyangze basins, located along the southern flank of the Indus–Yarlung suture zone, suggest that Asian-derived clastic sediments started contributing to sedimentation on the Indian continental margin at ∼60 Ma (DeCelles et al., 2014; Wu et al., 2014), hence there was direct contact between the continental blocks at this time;

2. Provenance analysis for upper Cretaceous strata in the Gyangze, Zhepure Mountain and Sangdanlin section of the Tethyan Himalaya indicates that the collision of India with Asia occurred before c. 65 and 60 Ma (Cai et al., 2011; DeCelles et al., 2014; Hu et al., 2015);

3. The oceanic rocks in ophiolite mélanges were obducted onto the northern margin of India during latest Cretaceous–earliest Tertiary time (c. 66–62 Ma) (Ding et al., 2005);

4. A paleomagnetic study of Paleocene marine sediments in the Gamba area of the Tethyan Himalaya in southern Tibet further constrains the timing of initial contact between India and Asia to before 60·5 ± 1·5 Ma (Yi et al., 2011).

Hu et al. (2017) defined diachronous collision as when the initial timings of two distant collision events on the same margin are more than 5 Myr apart. Thus, the following evidence for a later (c. 55–50 Ma) collision in the eastern and western Gangdese further supports the diachronous collision model:

1. Metamorphism of the subducted leading edge of the Indian continent (Tso Morari eclogite) and syn-collisional adakites in Ladakh of the western Himalaya took place at ∼53 to 47 Ma (Leech et al., 2005, 2007; Shellnutt et al., 2014);

2. A provenance study of the youngest detrital sedimentary rocks in the western Tethyan Himalaya of the Indian plate shows that zircons of Asian affinity were deposited on the Indian plate at 54 Ma and constrains terminal India–Asia collision to have occurred in the Western Himalaya by 54 Ma (Najman et al., 2017);

3. Detrital zircon geochronology for the Cenozoic foreland basin strata in the Hazara–Kashmir syntaxial region of northern Pakistan indicates that a provenance shift from Indian affinity to Asian affinity occurred at c. 56–55 Ma (Ding et al., 2016b);

4. Work on collision-related medium-pressure metamorphism in the eastern Himalayan Yardoi area of eastern Tibet suggests that the onset of the Asia–India collision had occurred by ∼50 Ma (Ding et al., 2016a).

Paleocene crustal anatexis and soft collision in southern Lhasa

We show in our study that the Paleocene (c. 63–57 Ma) Zhengga garnet-bearing, two-mica granites and biotite granites in the Gangdese batholith probably originated from a middle to lower crustal source, including juvenile and ancient materials. Existing data suggest the existence of a Paleocene crustal magmatic belt along the southern flank of the Indus–Yarlung suture zone (Fig. 1b). This Paleocene magmatism consists of adakites, I- and S-type granites, and minor gabbros and rhyolites and ignimbrites from the Dianzhong Formation; these have positive and negative ε_Nd(t) and zircon ε_Hf(t) values (Fig. 9), which indicate that most Paleocene granites in southern Lhasa were derived from the partial melting of crustal rocks (e.g. Guo et al., 2012; Lee et al. 2012; Zhang et al., 2013; Jiang et al., 2014; Wang et al., 2015; Zhu et al., 2015). Therefore, a significant crustal anatexis event is proposed to have occurred during the Paleocene (c. 63–57 Ma) in the Xigaze–Nyingchi area of the southern Lhasa sub-block (Fig. 1b).

Metamorphic phase equilibrium modelling shows that the pelitic schists from the Nyingchi complex in the eastern Gangdese underwent significant near-isobaric cooling in the granulite-facies field at high temperature of 700–830°C and nearly constant pressure of 10–11 kbar (Zhang et al., 2013). Our temperature calculations using the two-feldspar thermometer (Putirka, 2008) also suggest that the Zhengga biotite granites with low La/Yb and Sr/Y ratios were derived from a mid- to lower crustal source at high temperatures of 889–998°C. Similar Sr–Nd and zircon O isotope variations (Fig. 9) with increasing SiO₂ and decreasing Mg, Fe and Ti contents (Table 2) from the Zhengga biotite granites to garnet-bearing two-mica granites indicate that they were products of different evolutionary stages in mid- to lower crustal magma chambers during magma ascent and emplacement (Fig. 14). The compositional zoning of feldspar and garnet (Figs 4–6) and variable Ti contents of biotite (Supplementary Data File 3) also indicate that the Paleocene Zhengga granites have undergone magma recharge and fractionation at temperatures and pressures of ∼650–800°C and 1·2–0·4 GPa, respectively. Thus, high-temperature mid- to lower crustal magma chambers probably were widespread in southern Lhasa and also could explain the presence of coeval high-temperature syn-intrusion granulite rocks in the south Lhasa Block (Guo et al., 2012; Zhang et al., 2013) (Fig. 14).

In this study, plagioclase crystals from the Zhengga granites showing a broad range of compositions from andesine (An_30–48Ab_51–69Or_1–11) to albite (An_1–3Ab_97–98Or₁) (Fig. 4; Supplementary Data File 3) require a long-lived (>1 Myr) magma chamber (Karakas et al., 2017). However, sustaining such a long-term chamber is not easy. Numerical models indicate that an initial magma body is usually small and prone to frequent eruptions upon recharge, and also undergoes rapid (less than 10⁵ yr) cooling in a cold host crust (Degruyter & Huber, 2014; Karakas et al., 2017). As a consequence, the chamber will probably crystallize and become mechanically locked before it can grow by any appreciable amount. Both temperature and pressure decrease as a function of time (Liang, 2017). Thus, a low decompression rate, low cooling rate and high mass inflow rate could all potentially explain the growth of large silicic reservoirs in the shallow crust (Degruyter & Huber, 2014; Karakas et al., 2017; Liang, 2017).

As discussed above, a significant amount of evidence supports an initial Paleocene onset of continental collision. However, the convergence between two cold continents seems incompatible with the large-scale magmatism in the southern Lhasa terrane between c. 65 and 45 Ma (e.g. Chung et al., 2005; Wen et al., 2008; Ji et al., 2009; Lee et al., 2009, 2012; Zhu et al., 2015, 2017), which is also the main reason why magmatism cannot accurately constrain the timing of initial collision. Here, we propose a model of interaction between initial continental collision and oceanic slab roll-back to account for the above observations and other Paleocene geological features in the region (Fig. 14).

A slab roll-back model has been widely proposed to explain early Paleocene crustal anatexis and metamorphism in southern Lhasa (e.g. Chung et al., 2005; Kapp et al., 2005, 2007; Wen et al., 2008; Lee et al., 2009, 2012; Guo et al., 2012; Zhang et al., 2013; Zhu et al., 2013, 2017). Generally, slab roll-back triggers mantle corner flow and asthenospheric upwelling and results in significant partial melting of sub-continental lithospheric mantle and lower crust, similar to that proposed for the Late Cretaceous southern Lhasa terrane (Ma et al., 2013a, 2015) and the Eocene Ladakh Batholith (Shellnutt et al., 2014). Here, we propose that the initial collision in central Lhasa followed slab roll-back and resulted in cessation of central Gangdese magmatism and crustal thickening. Despite this, the early Paleocene crustal thickening was probably limited but the crustal thickening triggered by continental convergence probably suppressed decompression partial melting of sub-continental lithospheric mantle and generated a thermal boundary between crust and lithospheric mantle beneath southern Lhasa (Buck & Toksöz, 1983). Future work on Paleocene primitive mantle-derived rocks and lower crust-derived adakites in the region will yield more information on crustal thickness and the thermal state of the lithosphere, and may provide a test for the validity of this inference.

A sudden rise in convergence rate is identified prior to initial collision (Lee & Lawver, 1995; White & Lister, 2012), which probably correlates with roll-back of the Neo-Tethyan slab (Chung et al., 2005; Shellnutt et al., 2014). The roll-back would facilitate the dragging down of the attached Indian continental lithosphere to greater depths (Chemenda et al., 2000). This process explains why there was a general lack of contraction and shortening at upper crustal levels along the Gangdese arc and elsewhere in the Lhasa Block in the Paleogene (e.g. Chung et al., 2005, and references therein).

Subsequently, the resistance generated by the initial Indo-Asian collision and buoyancy forces owing to the lower density of the continental crust were likely to result in the abrupt deceleration of the convergence rate between India and Asia from 170 to 100 mm a^–1 at ∼60 Ma (e.g. Lee & Lawver, 1995). Slab roll-back may have continued as suturing progressed (von Blanckenburg & Davies, 1995). It would have been accompanied by a southward migration of asthenospheric convection beneath Tibet, which would have significantly enhanced the asthenospheric corner flow and changed the thermal structure of the mantle wedge (Chung et al., 2005). This process would supply a long-lived heat source for crustal anatexis by thermal diffusion. Moreover, increased pressure and crustal thickening resulting from progression of suturing and a reduced crustal cooling rate would have extended the lifetime of crustal magma reservoirs in southern Lhasa (Fig. 14).

Mid- to lower crustal flow has been widely recognized on a large scale in the Tibet–Himalaya Orogen and has played a role in crustal processes and thermal structure, such as melting, thinning, surface deformation and exchange or transportation of material (e.g. Nelson et al., 1996; Royden et al., 1997; Beaumont et al., 2001; Shapiro et al., 2004; King et al., 2011). Rosenberg & Handy (2005) proposed a melt fraction of >7 vol. % as the ‘melt connectivity transition’, marking an increase of melt interconnectivity that causes a dramatic drop in crustal strength. Our melt modelling suggests that the Paleocene Zhengga biotite granites were formed by high degrees (10–50 vol. %) of partial melting at 889–998°C (Fig. 12). This, combined with the extensive crustal anatexis in the western and eastern Gangdese, as discussed above, supports the likelihood of crustal flow in southern Tibet, similar that which occurred in the central and northern Tibet example (Wang et al., 2016). Further support for this model has been provided by Wu et al. (2017), who have recently proposed that granitic differentiation is dominated by flow segregation or dynamic sorting rather than crystal settling by gravitation. Thus, the Zhengga highly evolved garnet-bearing two-mica granites may also have experienced crustal flow, which would have contributed to crustal thickening (Zhu et al., 2017) (Fig. 14). An abrupt Paleogene shift in zircon Hf isotope compositions in Triassic–Eocene Gangdese granitoids and post-collisional adakites from Mesozoic depleted mantle-type signatures to more enriched crustal characteristics, prior to formation of subducted Indian crust-derived adakites, potentially reveals a hidden input from Indian continental crust (Chu et al., 2011). Middle- to lower crustal flow would provide an effective mechanism to supply an input from Indian crust before the full contact between the Indian–Asian continents.

Widespread high-temperature crustal anatexis and metamorphism resulted in hot and weakened crust in southern Lhasa. In addition, three-dimensional models indicate that slab roll-back is probably accompanied by trench migration (Stegman et al., 2006). Both the trench retreat and coeval Lhasa crustal anatexis would buffer the impact from the India continent, which is congruous with a soft collision in the Early Paleocene as advocated by Lee & Lawver (1995) and Chung et al. (2005). This large-scale interaction between continental collision and oceanic subduction not only has the potential to explain the evolution of the Paleocene Himalayan–Tibetan orogeny, but could also help explain the anomalously hot lithosphere and crust in some other large-scale convergence zones on Earth (Buck & Toksöz, 1983).

CONCLUSIONS

1. Garnet-bearing two-mica granites and associated biotite granites in the Sangri area, southern Tibet, were emplaced in the Paleocene (c. 63–57 Ma).

2. The Zhengga biotite granites were generated by high-degree (10–50%) partial melting of a mid- to lower crustal source comprising Gangdese juvenile crust with minor (∼5%) ancient Lhasa crust at high temperatures of 889–998°C.

3. The Zhengga garnet-bearing two-mica granites appear to be derived from advanced fractionation of biotite-granite magmas in relatively high-temperature (∼700–750°C) magma chambers at mid-crust depths (∼12–18 km).

4. A Paleocene magmatic and high-temperature metamorphic belt, extending linearly for more than 1500 km along the north of the Indus–Tsangpo Suture, suggests that a significant Early Paleocene crustal anatectic event occurred in the Xigaze–Nyingchi area of the southern Lhasa sub-block.

5. The interaction between continental collision and subducted slab roll-back may be responsible for the evolution of the Paleocene Himalayan–Tibetan orogeny; similar processes may have occurred in other large-scale convergence zones.

ACKNOWLEDGEMENTS

We are very grateful for the hard work of Georg Zellmer, Greg Shellnutt, Romain Tartèse, Mei-Fei Chu and another reviewer, whose constructive criticism and suggestions have led to clarification of various aspects of this paper. We appreciate the assistance of Yong-Sheng Liu, Yue-Heng Yang, Zhao-Chu Hu, Xi-Rong Liang, Xiang-Lin Tu, Jin-Long Ma, Qing Yang, Chang-ming Xing, Lie-Wen Xie, Wan-Feng Zhang, Bo-Qin Xiong, Guang-Qian Hu and Ying Liu for zircon age, whole-rock and mineral geochemical analyses.

FUNDING

Financial support for this research was provided by the DREAM Programs of China (2016YFC0600407 and 2016YFC0600309), the National Natural Science Foundation of China (Nos 41402048 and 41630208), the Key Program of the Chinese Academy of Sciences (QYZDJ-SSW-DQC026), talent project of Guangdong Province (2014TX01Z079), the State Scholarship Fund of China (No. 201604910124) and the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIGCAS 135 project 135TP201601). This is contribution No. IS-2489 from GIGCAS.

SUPPLEMENTARY DATA

Supplementary data for this paper are available at Journal of Petrology online.