Microcontinent subduction and S-type volcanism prior to India–Asia collision

Continental crust has long been considered too buoyant to be subducted beneath another continent, although geophysical evidence in collision zones predict continental crust subduction. This is particularly significant where upper continental crust is detached allowing the lower continental crust to subduct, albeit the mechanism of such subduction and recycling of the upper continental crust remain poorly understood. Here, we investigate Paleocene S-type magmatic and volcanic rocks from the Linzizong volcanic succession in the southern Lhasa block of Tibet. These rocks exhibit highly enriched 87Sr/86Sr, 207Pb/206Pb and 208Pb/206Pb together with depleted 143Nd/144Nd isotope ratios. The geochemical and isotopic features of these rocks are consistent with those of modern upper continental crust. We conclude that these Paleocene S-type volcanic and magmatic rocks originated from the melting of the upper continental crust from microcontinent subduction during the late stage of India–Asia convergence.


Geochronology and geochemistry
In the granitoid samples the U-Pb data on 26 zircon grains from sample T473 and 13 from F15-1 yielded concordant 206 Pb/ 238 U ages of 60.8 ± 0.4 Ma (  Table S1), suggesting magmatic crystallization. In addition, inherited zircons are not present in both granitoid and volcanic rocks 21 .

Alteration, assimilation and fractional crystallization
Our petrographic observation shows that the rocks underwent low degree sericite alteration (Fig. 3B). The loss-on-ignition (LOI) values of the samples are uniform and vary from 2.04 to 4.18 wt% (average = 2.99 wt%), implying slight alteration. Here, we argue that the geochemical composition can still be employed as a tracer to discuss the magma origin because the content of oxides does not show significant geochemical variation. Besides, the samples are characterized by sub-parallel patterns of REE and trace element concentrations (Fig. 5C,D), indicating they still preserve their original geochemical signatures 30 . Both the granitoids and volcanic rocks have high SiO 2 , which chemically correspond to the high silica granites (HSGs) with SiO 2 > 70 wt%. The HSGs are considered can be formed by highly fractional crystallization of the mafic parents 31 . The important question is whether the granitoids and volcanic rocks inherited their peraluminous geochemistry from the source region or evolved from a relatively mafic parent 32 by assimilation and fractional crystallization (AFC). Previous studies consider the Dianzhong volcanic rocks as I-type rocks originated from mantle wedge that containing some crustal component followed by the AFC processes [5][6][7]9 . However, these S-type volcanic and magmatic rocks have extremely low Na 2 O, CaO , MgO, FeO T , Cr (5.20-15.42), Ni (1.09-3.57), and Co (0.60-1.54) contents with radiogenic Sr and Pb isotope composition compared with Dianzhong I-type series 5,6,8,9 , which indicate that they are not likely originated from partial melting of the mantle wedge or the mafic lower continental crust (LCC). Besides, the initial 87 Sr/ 86 Sr ratios and ε Nd (t) of the granitoids and volcanic rocks exhibit inconsistent relations rather than any correlation with SiO 2 contents (Fig. 6A,B), and the volcanic rocks and granitoids also yield relatively uniform Pb isotope compositions (Supplementary Table S3), thus precluding significant crustal assimilation. With respect to the fractional crystallization, we apply high-field-strength elements Nb, Ta and Zr to evaluate the degree of fractional crystallization 25,33 . The Nb and Ta concentrations show a positive correlation with the degree of fractional crystallization, while the Nb/Ta and Zr concentration progressively decrease 33 . The isotope compositions indicate that these magmatic rocks were derived from the www.nature.com/scientificreports/ upper continental crust. Their Nb/Ta ratios (9.12-10.33) are slightly lower than the upper continental crust (11.36) and different from the highly fractionated Himalaya leucogranites (< 5) (Fig. 5E,F). The low Nb, Ta and high Nb/Ta ratios indicate the granitoids and volcanic rocks from the Sinongduo area could not have experienced strongly fractional crystallization and the geochemistry of these magmatic rocks reflects the composition of the source region.

Magma source and petrogenesis
The granitoids and volcanic rocks from the Sinongduo area are strongly peraluminous and plot in the S-type magma field (Figs. 5B, 7), in contrast to the I-type Gangdese batholith and coeval volcanic rocks elsewhere in the Lhasa block. The Lhasa terrane was not under a post-collisional setting in the Paleocene, and therefore these rocks cannot be correlated to post-collisional strongly peraluminous granites formed in such settings 34 . Even  27 . Data for the Himalayan highly fractionated leucogranite are from Liu et al. 28 , and Lin et al. 29 . Data for the Xuru Tso granitoids and volcanic rocks are from Lee et al. 9 , and Gao et al. 10 . Other data for the Sinongduo granitoids and volcanic rocks are from Yang 11 , Zhou et al. 12 , and Ding et al. 21 .  (3) having restricted to high SiO 2 compositions with relatively low Na 2 O (generally < 3.2%) but high K 2 O (~ 5%) contents. As mentioned above, the characteristics of high SiO 2 and K 2 O, low Na 2 O contents, and ASI values of 1.64-2.10 imply that these granitoids and volcanic rocks have affinity to S-type granite. The normative corundum (Fig. 3E,F) contents of 3.78-6.39% based on the CIPW calculation is even more enriched than the peraluminous felsic S-type series in the Lachlan fold belt and contrasts with the depleted I-type series 24 . The classification of the Sinongduo magmatic rocks as S-type is also supported by their whole-rock initial Sr 87 /Sr 86 isotopes (0.707-0.732, except ZK0602-2), as Sr is more radiogenic in S-type granites with initial Sr 87 /Sr 86 > 0.708 24 .
Given that the geochemical characteristics are similar to those of upper crustal sedimentary rocks, S-type magma is considered to form by melting of a sedimentary UCC source that had experienced at least one cycle of weathering 24 rather than from the more mafic LCC 27 or oceanic crust. A long-term weathering process can remove Na and Ca and maintain high K in the abundant clay minerals 39 , which leads to the high ASI in sediments. Therefore, the key factor for the generation of S-type granitic magma is a sedimentary source 40 . Since such sediments are deposited only in the Earth's upper continental crust, S-type granitic magma is considered to be product of anatexis 40,41 .
As seen in the chondrite-normalized REE patterns (Fig. 5C), all the granitoids in our study display nearly overlapping patterns with that of the UCC, indicating more enriched LREE compositions than for the LCC and the Indian Ocean sediments. In addition, both granitoids and volcanic rocks from our study have significant negative Eu anomalies, which are very similar to those of the UCC (average Eu/Eu* = 0.65 27 ) and contrast with   27 ). Although these negative Eu anomalies are generally interpreted as the residues of plagioclase feldspar, a more direct cause is assumed here that reflects the Eu-depleted primary magma because of the low degree of fractional crystallization. In the 87 Sr/ 86 Sr-143 Nd/ 144 Nd isotope space (Fig. 8), it is seen that compared with other magmatic rocks, including other Linzizong volcanic rocks, the Paleocene S-type magmatism shows an obvious increase in 87 Sr/ 86 Sr, confirming UCC affinity. The same conclusion also can be reached from the radiogenic Pb isotopes (Supplementary Table S3). Two potential scenarios can be envisaged for formation of the Sinongduo S-type granitic rocks: (1) crustal anatexis 40,41 ; (2) extraction from shallow andesitic magma reservoir by fractional crystallization 12 . For the crustal anatexis hypothesis, a deep buried metasedimentary source is required for the generation of S-type magma 47 . Exhumation of the overthickened crust is a widely accepted mechanism for the formation of strongly peraluminous S-type granites in post-collisional setting 34,47 . However, this model is inapplicable for the S-type granites formed non-collisional setting. A two-stage rollback of the subducted oceanic crust that caused crustal extension is proposed to account for the S-type granites in the Phanerozoic circum-Pacific orogenic belts 40 . S-type granites formed by this mechanism are consistently associated with high-temperature-low-pressure (HTLP) metamorphic complexes, or even some core complexes which were exhumed during continental extension 48,49 . A thinned lithosphere is also required for a substantial transient heat flux to the crust 40 . Overall, the Sinongduo S-type magmatic rocks cannot be correlated to both these models because HTLP metamorphic complex is absent and continental collision had not yet occurred at that time. Besides, the S-type magma cannot be extracted from shallow andesitic magma reservoir because of the low degree fractional crystallization. This hypothesis is also limited in explaining the large outcrop of these volcanic and intrusive rocks that cover more than 300 km 2 in the study area ( Fig. 2A,B).
As the Gangdese batholith and Linzizong volcanic rocks are treated as products of the northward subduction of the Neo-Tethys oceanic crust 6,9 , we propose an alternative scenario in which the Linzizong S-type volcanic rocks and granitoids are derived from subducted materials of UCC geochemical affinity. Zircon saturation temperatures (Tzr) of the Sinongduo granitoids are in the range of 768-804 °C (mean = 793 °C) based on the method of Boehnke et al. 50 . The absence of inherited zircons in the granitoids indicates that partial melting took place at high-temperature conditions and the initial magmas were undersaturated in zirconium. In this case, the calculated zircon saturation temperatures provide minimum estimates of melting temperature 51 . Besides, microscopic observation indicates that the granitoids contain fewer zircon crystals. These features, together with the wide distribution of the volcanic rocks suggest that the melting temperature exceeded 804 °C, comparable

Implications for the India-Asia convergence
Our first assumption is that the subducted UCC represents a portion of the northern edge of the Indian plate, in which case the initial India-Asia collision started at approximately 65 Ma, which is in accordance with the popular geophysical collision model that shows continental lithosphere beneath southern Tibet 52,53 . However, it conflicts with most published studies that support collision ages to range from Late Cretaceous to Oligocene 54 , especially the recently accepted 55 ± 5 Ma age for the onset of collision [54][55][56][57] . On the other hand, Andean-type calc-alkaline magmatism with consistent isotopic compositions from the Early Jurassic to the middle Eocene 6,9 naturally implies the persistence of Neo-Tethys oceanic crust subduction beyond a collisional background and suggests that the 65 Ma age for the India-Asia collision should be excluded 56 . Hence, the subducted UCC cannot be the northern edge of the Indian plate, and should be attributed to other continental subduction systems.
An alternate possibility would be the subduction of microcontinents (microplates). Fragmentation and microcontinent formation are common phenomena in many continental margins 57 . The opening of the embryonic Neo-Tethys may have caused rifting of the northern margin of the Indian plate under a divergent tectonic setting and formed a cluster of microcontinents (Fig. 9) between the Indian plate and the Asia plate 58 . Besides, thermomechanical study reveals that the thermal and buoyancy effects of mantle plume impingement on the bottom of the continental part of a subducting plate can also induce the separation of microcontinents from the main body of the continent 59 . With the initiation of the northward subduction of the Neo-Tethys crust in early Mesozoic 60 , possibly induced by the impingement of mantle plume at the transition zone between oceanic and continental lithosphere 61 , these microcontinents are thought to have accreted to the Lhasa block along the IYSZ in the early Cenozoic, similar to the Burma terrane 62 and the Oaxaquia in North America 63 . However, no geological records have been reported to document the accretion in the IYSZ.
In general, strong rheological coupling of UCC and LCC can separate the continental crust from the downgoing mantle lithosphere, while low rheological coupling of the UCC and LCC allows the LCC to sink into the mantle, resulting in continental subduction 64 . Numerical models indicate subduction of the Indian LCC 65 , which is also confirmed from geophysical evidence 52,53 . However, the possibility and mechanism of the subduction and www.nature.com/scientificreports/ recycling of the UCC remain poorly understood. Numerical modelling studies indicate that the UCC can subduct to a great depth if the continental crust is strongly coupled with the mantle lithosphere under a relatively low Moho temperature [66][67][68] . Another critical factor controlling the accretion and scale of continental subduction is the continental mass 69 , and compared with the short-subducted lengths of intact continental crust, microcontinents can be entirely subducted 70 . The coherence may play an important role in the subducting plate if the detachment did not occur 71 . A so called "crustal pocket" can subduct to a depth of 50-120 km 67 . Microcontinents might undergo subduction process comparable to that of the "crustal pocket" in the subduction zone. The melts derived from the subducted UCC are immediately transported to the surface across the rheologically coupled lithosphere 72 along deep-seated lithospheric-scale faults that formed in the overriding plate 73 . We consider this mechanism for the S-type magmatic rocks that are distributed along the regional Luobadui-Milashan fault (LMF) as clusters ( Fig. 2A). We thus propose that the microcontinents rifted from the India plate could have been wholly subducted without any accretion to the collision boundary. If the whole UCC was subducted, then little or possibly no trace of accretion may be recorded in the suture zone, as these were entirely erased. Only the melting of UCC components can induce secondary geochemical consequences which are traceable from magmatic signatures 42 . After the subduction of the UCC, a significant period of oceanic crust subduction must have followed prior to the continental collision (Fig. 9). Our model has the advantage of reconciling the heterogeneity of the Linzizong volcanic rocks and coeval batholiths and can also support the subsequent India-Asia collision geodynamics.
Previous studies considered S-type magma to be a product of crustal anatexis 40 . However, our proposal opens up alternate possibilities of partial melting and recycling of subducted UCC, particularly involving microcontinents.

Methods
Zircon U-Pb dating. The granitoid samples for LA-ICP-MS zircon U-Pb dating were crushed, and zircons were separated using conventional heavy liquid and magnetic separation techniques. Zircons were handpicked under a microscope, mounted in epoxy resin, polished to approximately half their original thickness, and studied under both reflected and transmitted light. To examine the internal structure, cathodoluminescence (CL) images of zircon grains were obtained using a JSM6510 scanning electron microscope. Zircon U-Pb dating was performed using a Neptune MC-ICP-MS coupled with a New Wave UP213 laser ablation system at the Institute of Mineral Resources, Chinese Academy of Geological Sciences. The operating conditions and detailed analytical procedures followed those described by Hou et al. 74 . The U-Pb ages of the zircons were calculated and plotted using the Isoplot3 software (Ludwig, 2003). Individual analyses are presented with 1σ error, whereas age uncertainties are quoted at the 95% level (2σ).
Whole-rock major and trace element analyses. Whole-rock major element analyses and trace element analyses were conducted with an X-ray fluorescence (XRF) spectrometer (Primus II, Rigaku, Japan) and an Agilent 7700e ICP-MS system, respectively, at the Wuhan Sample Solution Analytical Technology Co., Ltd., Wuhan, China. The sample powder was accurately weighed and mixed with the cosolvent (Li 2 B 4 O 7 : LiBO 2 : LiF = 9:2:1) and oxidant (NH 4 NO 3 ) in a Pt crucible, which was then placed in the furnace at 1150 °C for 14 min. Next, the melted sample was saturated with air for 1 min to produce flat disks on the firebrick for the XRF analyses. Approximately 1 ml of HNO 3 and 1 ml of HF were slowly added to a 50 mg sample of powder in a Teflon bomb, which was placed in a stainless-steel pressure jacket and heated to 190 °C in an oven for > 24 h. After evaporating the sample to dryness twice, 1 ml of HNO 3 , 1 ml of Milli-Q (MQ) water, and 1 ml of 1 ppm internal standard solution were added, and the Teflon bomb was resealed and placed in the oven at 190 °C for > 12 h. The final solution was transferred to a polyethylene bottle and diluted to 100 g by the addition of 2% HNO 3 for ICP-MS analysis. Analyses of international rock standards (AGV-2, BHVO-2 and BCR-2) indicated that the precision and accuracy of the results were better than 5%.
Whole-rock Sr-Nd-Pb isotope analyses. The Sr-Nd-Pb isotope ratios were determined by using a Finnigan Triton thermal ionization mass spectrometer (TIMS) at the State Key Laboratory for Mineral Deposits Research, Nanjing University. The measured 87 Sr/ 86 Sr and 143 Nd/ 144 Nd isotope ratios were normalized to 86 Sr/ 88 Sr = 0.1194 and 146 Nd/ 144 Nd = 0.7219, respectively, for mass fractionation correction. During the period of data acquisition, the mean 87 Sr/ 86 Sr ratio of the Sr standard (NBS987) and the 143 Nd/ 144 Nd ratio of the Nd standard (JNDi-1) were 0.710261 ± 0.00006 and 0.512128 ± 0.00004 (2σ), respectively. Further, the measured 87 Sr/ 86 Sr and 143 Nd/ 144 Nd values for the standard BCR-2 were 0.705018 ± 0.000003 and 0.512624 ± 0.000004, respectively (the equivalent reference values obtained from Weis et al. 75 are 0.705019 ± 0.000016 and 0.512634 ± 0.000012). The analytical procedures for Sr and Nd isotopes followed those described in detail by Pu et al. 76 . The value of ε Nd (t) was calculated with reference to the chondritic uniform reservoir (CHUR) 143 Nd/ 144 Nd ratio of 0.512638.
For Pb isotope analysis, sample powders were weighed into a Teflon bomb and dissolved by a combination of purified HNO 3 and HF at 190 °C for 48 h. Pb was separated and purified on ion exchange columns with diluted HBr as the eluant. Finally, the Pb fraction was eluted using 6.0 M HCl and gently evaporated to dryness prior to mass spectrometric measurement. The detailed procedures used in measuring Pb isotopes can be found in White et al. 77 . The measured Pb isotopic ratios were rectified for instrumental mass fractionation by performing replicate analyses of the standard NIST-981. The standard NIST-981 yielded 206 Pb/ 204 Pb = 16.9318 ± 0.0003, 207 Pb/ 204 Pb = 15.4858 ± 0.0003, and 208 Pb/ 204 Pb = 36.6819 ± 0.0008. In addition, the international standard BCR-2 was verified using an unknown sample by employing this method. The measured values for the BCR-2 Pb standard were 18

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).