Early Permian Syn-Subduction Extension in the South Tianshan (NW China): Insights From A-Type Granitoids in the Southern Altaids

A-type granite is an important geodynamic indicator because it requires a high melting temperature that is commonly driven by extensional events. Here we report geochronology, whole-rock geochemistry, and zircon Lu-Hf isotopes of newly identified A-type granitic rocks from the South Tianshan in the southern Altaids. Zircon LA-ICP-MS ages indicate that the granitoids were emplaced at ca. 298–272 Ma. Geochemically, they are metaluminous to slightly peraluminous (A/CNK = 0.95–1.10), and belong to the high-K calc-alkaline to shoshonitic series. They are characterized by relatively high zircon saturation temperatures (824–875°C), K2O + Na2O contents (7.31–9.36%), high field strength elements (HFSE; Zr + Nb + Ce + Y = 365–802 ppm), and Ga/Al ratios (2.8–4.2), which all point to an A-type affinity. In addition, they have slightly enriched Hf isotope compositions (ε Hf(t) = −10.9 to + 0.6), and corresponding Mesoproterozoic (1,272–1759 Ma) crustal model ages, suggesting they were probably generated by partial melting of mature crust that contained minor mantle-derived magmatic material. The granitoids have distinctive subduction-related trace element signatures, with deep Nb and Ta troughs, elevated large ion lithosphere elements (LILEs), and flat HFSEs patterns, very similar to arc-derived granites in the Lachlan accretionary orogen. Integration of these new sedimentological, structural and geochronological results with relevant published information provides a new data-archive, which indicates that neither the Tarim mantle plume nor post-collisional extension can explain the genesis of these A-type granitoids. Instead, we propose a new more pertinent and robust model according to which they formed due to high temperature gradient in a subduction-related extensional setting probably triggered by southward rollback of the South Tianshan oceanic lithosphere, which caused upwelling of asthenospheric mantle combined with an increased temperature that led to large-scale crustal melting. This process gave rise to a broad magmatic arc in the southern active margin of the Yili-Central Tianshan. Our new data shed light on the retreating accretionary orogenesis of the southern Altaids in the Permian.

Permian magmatic rocks are widespread in the western Tianshan ( Figure 1B; Supplementary Table S1), petrogenesis of which has long been an issue of hot debate that has hampered a better understanding of the latest stages of evolution of the southern Altaids. The main models to explain formation of the Permian felsic magmatism in the South Tianshan include: 1. a post-collisional extensional setting related to pre-Permian closure of the South Tianshan ocean (Konopelko et al., 2007;Long et al., 2008;Seltmann et al., 2011;Huang et al., 2015a;Qin et al., 2021). 2. A result of the Permian Tarim mantle plume (e.g., Zhang and Zou, 2013). Although it is well accepted that the South Tianshan was still under a subduction regime in the Permian (Li et al., 2005;Xiao et al., 2013;Sang et al., 2018;Wen et al., 2020;Abuduxun et al., 2021a), less attention has been paid to the petrogenesis of the Permian granitoids in relevant tectonic models. Thus, these conflicting views have created an important, but unresolved question, which is critical for understanding the Permian accretionary architecture of the southern Altaids.
A-type granites generally require a high melting temperature, which is commonly driven by extensional events, such as continental rifting and/or post-collisional extension (Whalen et al., 1987;Eby, 1992). Moreover, several recent studies have pointed out that A-type granites can also occur in a subductionrelated extensional setting (e.g., Collins et al., 2019;Yin et al., 2021). Therefore, the A-type granites in ancient accretionary orogens are important petrogenetic indicators, which provide crucial information for unravelling their ambient geodynamic processes.
So far, Permian A-type granites have not been well documented in the Chinese part of the South Tianshan; the main examples are in Kyrgyzstan ( Figure 1B). Nevertheless, their spatiotemporal relations to coeval magmatism in adjacent tectonic units is not well constrained, although they are usually considered to be post-collisional (Konopelko et al., 2007;Konopelko et al., 2009).
In this study, we report zircon U-Pb geochronology, Lu-Hf isotopes and whole rock geochemistry of newly-identified A-type granitoids from key locations in the South Tianshan. By integration of current and previously published data, we propose that upper-plate extension triggered by slab rollback was the geodynamic process that was responsible for the spatially and temporally-related Early Permian granitoids in the South Tianshan. Our data provide new insights for the retreating accretionary orogenesis of the southern Altaids before its final amalgamation with the Tarim Craton.

GEOLOGICAL FRAMEWORK
Tectonic units referred in this study include from north to south: the Yili-Central Tianshan arc, the South Tianshan accretionary complex, and the northern Tarim Craton ( Figure 1B).
The South Tianshan accretionary complex, formed as a result of continuous northward subduction of the South Tianshan oceanic plate (Abuduxun et al., 2021a;Sang et al., 2018;Xiao et al., 2013), followed by final amalgamation of the Tarim Craton with the Yili-Central Tianshan arc (Alexeiev et al., 2015;Han et al., 2011) ( Figure 1B). Accordingly, it records the process of terminal suturing, in the southern Altaids, between the Yili-Central Tianshan arc and the northern passive margin of the Tarim Craton (Han et al., 2011;Xiao et al., 2013).
The principal strata in the South Tianshan are Paleozoic siliciclastic turbidites, limestones, cherts and schists (BGMRXUAR, 1993). They were previously considered to be passive margin sediments on the northern slope of the Tarim Craton (Han et al., 2011;Biske et al., 2018). Recently, however, fragmented and/or dismembered elements of ocean plate stratigraphy (OPS) were recognized in previously assigned Silurian to Carboniferous strata from several locations in the South Tianshan (Safonova et al., 2016;Sang et al., 2018;Abuduxun et al., 2021aAbuduxun et al., , 2021b. They are composed of closely associated MORB/OIB-type basalts, ribbon cherts and turbidites that occur in mélanges with a typical block-in-matrix structure or in coherent strata in thrust-repeated tectonic slices (Sang et al., 2018;Abuduxun et al., 2021b). The depositional ages of the Paleozoic strata, especially of metamorphosed sediments, are younger than previously estimated, as indicated by new ages of detrital zircons (Fu et al., 2018;Huo et al., 2019;Abuduxun et al., 2021a). Provenance analysis of the clastic sedimentary rocks indicates that they were mainly derived from the Yili-Central Tianshan arc (Sang et al., 2018;Huo et al., 2019;Abuduxun et al., 2021a).

SAMPLING AND BRIEF PETROGRAPHY
The regional distribution of the Permian granitoids in the South Tianshan is shown in Figure 1B. They mainly occur in stocks of variable size that have intruded pre-Permian rocks along the strike of the South Tianshan (BGMRXUAR, 1993). Samples for zircon U-Pb dating, Lu-Hf-isotope and whole rock geochemical analysis were collected from five intrusions in the South Tianshan ( Figures 1B, 2). The lithologies, petrological compositions and sampling coordinates of the investigated granitoids are briefly summarized in Table 1.
Sample 19ST17 was collected from a coarse-grained syenogranite ( Figure 3C) in the Laohutai area ( Figure 2A). As shown in Figure 3H, it consists mainly of K-feldspar (45-50%), plagioclase (10-15%), quartz (25-30%), and minor biotite (∼5%). Accessory minerals include zircon and apatite. The wall rock of the pluton was ascribed to the Mesoproterozoic Akesu Group according to previous investigations (BGMRXUAR, 1982;Wang, 2007a). However, it is worth noting that the depositional ages of these "old" rocks were commonly estimated by stratigraphic correlations and/or on their degree of metamorphism. Recent age data suggest these rocks are not older than Paleozoic (Huo et al., 2019).

Zircon U-Pb Geochronology
Zircon mounts and cathodoluminescence (CL) images were made at Beijing Zhongke Kuangyan Test Technology Co., Ltd. Each sample was crushed, and its zircon grains were separated using standard magnetic and high-density liquid techniques. The zircons were handpicked and cast in an epoxy mount, which was polished to expose their interiors for imaging by optical and CL techniques. The mounts were subsequently polished to EBSD standard using collodial silica and carbon coating. A Tescan MIRA 3 field emission scanning electron microscope (SEM) was used to collect CL images of individual zircon grains.
Zircon U-Pb analyses were made at the Beijing Quick-Thermo Science & Technology Co., Ltd, using an ESI New Wave NWR 193 UC (TwoVol2) laser ablation system connected to an Agilent 8900 ICP-QQQ. Individual zircon grains (mounted and polished in epoxy) were ablated in a constant stream of He that was mixed downstream with N 2 and Ar before entering the torch region of the ICP-QQQ. After warmup of the ICP-QQQ and connection with the laser ablation system, the ICPMS was first tuned for robust plasma conditions by optimizing laser and ICP-QQQ settings, and for monitoring 232 Th 16 O + / 232 Th + ratios (always ≤0.2%) and 238 U + / 232 Th + ratios (always between 0.95 and 1.05) while ablating NIST SRM 612 in line-scan mode. 91500zircon was used as a primary reference material for U-Pb age determinations, and PleŠovice zircon (Mean 337 ± 1.2 Ma; MSWD 0.95) as a secondary reference (Wiedenbeck et al., 1995). Background subtractions and corrections for laser downhole elemental fractionation were undertaken with the Iolite data reduction package within the Wavemetrics Igor Pro data analysis software (Paton et al., 2010). Concordia diagrams were processed using ISOPLOT 4.15 (Ludwig, 2003).

Hf Isotopes
Zircon Hf isotope analyses were carried out in situ using a Geolas HD excimer ArF laser ablation system attached to a Neptune Plus All data were acquired on zircons in single spot ablation mode at a spot size of 44 μm. Helium was used as a carrier gas to transport ablated samples from the laser-ablation cell to the ICP-MS torch by a mixing chamber merged with Argon The energy density of laser ablation used in the present study was ∼7.0 J cm −2 . Each measurement consisted of 18 s of acquisition of the background signal followed by 50 s of ablation signal acquisition. 176 Lu/ 175 Lu 0.02656 (Blichert-Toft et al., 1997) and 176 Yb/ 173 Yb 0.79639 (Fisher et al., 2014) ratios were determined in order to correct the isobaric interferences of 176 Lu and 176 Yb on 176 Hf. Hf and Yb isotope ratios were normalized to 179 Hf/ 177 Hf 0.7325 and 173 Yb/ 171 Yb 1.132685 (Fisher et al., 2014) using an exponential correction for mass bias. Zircon GJ1 was used as the reference standards during the routine analyses. Off-line selection and integration of analyzed signals, and mass bias calibrations were performed using ICPMSDataCal (Liu et al., 2010). The ε Hf (t) values were calculated by assuming chondritic values of 176 Lu/ 177 Hf 0.282785 and 176 Hf/ 177 Hf 0.0336 (Bouvier et al., 2008). The single-stage model ages (T DM1 ) were calculated relative to the depleted mantle with a presentday 176 Hf/ 177 Hf 0.28325 and 176 Lu/ 177 Hf 0.0384 (Griffin et al., 2000). The two-stage continental crust model ages (T DM2 ) were also calculated by plotting the initial 176 Hf/ 177 Hf of zircons back to the depleted mantle evolutionary curve using the value of 176 Lu/ 177 Hf (0.015) for the average continental crust (Griffin et al., 2000). The calculated ε Hf (t) values and two-stage model (T DM2 ) were compiled using the zircon 206 Pb/ 238 U ages.

Whole Rock Geochemistry
The geochemical analysis of the bulk-rock samples was performed at the ALS Chemex Co. Ltd in Guangzhou. All samples were crushed to less than 200-mesh after removal of weathered surfaces and then mixed with Li 2 B 4 O 7 and LiBO 2 to make homogeneous glass beans at 1,050-1,100°C. Major elements were determined by XRF-1500 Sequential X-ray Fluorescence Spectrometry on fused glass beads, the analytical precision of which ranged from ±1% to ±2% based on certified standards and duplicate analyses.
Trace elements and Rare Earth Elements (REE) were analyzed with an inductively-coupled Plasma Mass Spectrometer (ICP-MS). About 50 mg of powder for every sample were added to a lithium metaborate flux, mixed well and fused in a furnace at 1,000°C. The resulting melt was then cooled and dissolved in 100 ml of 4% HNO3 solution. The analytical precision of the ICP-MS data at the ppm level is better than 5%.

Zircon U-Pb Ages and Hf Isotopes
The results of zircon U-Pb dating are listed in Supplementary  Table S2, and are graphically illustrated in Figure 4. The results of in-situ Lu-Hf isotopes are listed in Supplementary Table S3, and are graphically visualized in Figures 5, 6. Representative cathodoluminescence (CL) images are also shown in Figure 4.
The CL images show that zircons separated from all dated samples are transparent to semitransparent, and mostly have euhedral and prismatic morphologies with clear oscillatory zones and aspect ratios of 1:2-1.3 in CL images (Figure 4), indicative of a magmatic origin. They mostly show high Th/U ratios (from 0.3 to 1.92), further indicating a magmatic origin (Corfu et al., 2003).
A total of 26 analyses were conducted of the granite porphyry sample 18ST60, among which 22 grains have concentrated 206 Pb/ 238 U ages between 287 and 298 Ma, yielding a weighted mean 206 Pb/ 238 U age of 292 ± 2.6 Ma (MSWD 0.23); this age is interpreted as the crystallization age of this sample. The remaining 4 grains yield concordant older ages of 314 Ma,421 Ma,424 Ma,and 436 Ma, which might be inherited zircons or were captured from the wall rock during magma ascent. Sixteen spots on the dated zircons were analyzed for their Hf isotopic compositions. Twelve grains with Permian ages display ε Hf (t) values ranging from −2.3 to 0.6 ( Figure 5), corresponding to (T DM2 ) ages between 1,272 and 1,461 Ma. An inherited 314 Ma zircon displays a similar ε Hf (t) value (+0.5) and corresponding (T DM2 ) age (1,303 Ma) as the Permian zircons. In contrast, the three other inherited zircons (421,424,436 Ma) show more negative ε Hf (t) values of −10.9 to −9.2 ( Figure 5), with corresponding older (T DM2 ) ages of 2010-2,106 Ma ( Figure 6).
A total of 23 analyses of the granite porphyry sample 18ST64 yield concordant ages between 286 and 308 Ma, and define a weighted mean 206 Pb/ 238 U age of 294 ± 2.4 Ma (MSWD 1.4). This age is considered as the crystallization age of sample 18ST64.
Out of the 17 analyses of the syenogranite sample 19ST17, one was performed on an inherited or captured zircon with a concordant older age of 314 Ma. The remaining 16 spots yield a weighted mean 206 Pb/ 238 U age of 286 ± 3.0 Ma (MSWD 6.4), which is considered to be the crystallization age of this sample; fourteen in-situ Lu-Hf analyses were obtained for this sample. The results display negative ε Hf (t) values of −6.8 to −1.6 ( Figure 5), corresponding to (T DM2 ) ages of 1,411-1759 Ma.
A total of 20 analyses of the biotite monzogranite sample 20ST78 comprise an age range of 264-277 Ma, with a weighted mean 206 Pb/ 238 U age of 272 ± 1.2 Ma (MSWD 0.99); this is interpreted as the crystallization age of the biotite monzogranite. All twenty spots were analyzed for their Hf isotopic compositions. The data show slightly negative ε Hf (t) values of −4.3 to −0.4 ( Figure 5), with corresponding (T DM2 ) ages between 1,324 and 1,579 Ma.
Out of the 15 analyses of the granodiorite sample 20ST93, two were performed on the inherited or captured zircons with prominent older ages of 2,298 Ma and 2,764 Ma. The  Table S3) and the Permian Tarim plume-related A-type granites that have apparent positive ε Hf (t) values (data sources are can be found in Su et al., 2019;Wei et al., 2019;Zhang et al., 2010).
Frontiers in Earth Science | www.frontiersin.org February 2022 | Volume 9 | Article 831677 remaining 13 analyses yield a weighted mean 206 Pb/ 238 U age of 298 ± 5.8 Ma (MSWD 0.50) representing the crystallization age of this sample. With the exception of two Precambrian inherited zircons, the remaining thirteen grains with Permian ages were analyzed for Hf isotopic compositions. The results display negative ε Hf (t) values between −9.6 and −3.5 ( Figure 5), with corresponding (T DM2 ) ages ranging from 1,548 to 1932 Ma.

Whole Rock Geochemistry
The results of major and trace element geochemical analyses of the Permian granitoids are listed in Supplementary Table S4. Samples 18ST60, 18ST64, 19ST17 and 20ST78 generally show SiO 2 contents ranging from 69.25 to 78.20 wt.% and K 2 O + Na 2 O contents between 7.95 and 9.36 wt.%, and they plot in the granite field in the total alkalis (K 2 O + Na 2 O) wt.% vs. silica (SiO 2 wt.%) classification diagram for plutonic rocks ( Figure 7A) (Cox et al., 1979). In addition, these samples are characterized by low to moderate concentrations of Al 2 O 3 (11.11-14.07 wt.%), TiO 2 (0.15-0.46 wt.%), CaO (0.22-1.70 wt.%), MgO (0.12-0.48 wt.%) and P 2 O 5 (0.12-0.15 wt.%). In comparison, sample 20ST93 exhibits relatively lower SiO 2 contents ranging from 65.71 to 66.81 wt.% and high K 2 O + Na 2 O contents between 7.31 and 7.85 wt.%, and they plot on the boundary between the granodiorite and granite fields in the classification diagram ( Figure 7A) (Cox et al., 1979). This sample is also characterized by relatively higher Based on the K 2 O (wt.%) vs. SiO 2 (wt.%) diagram ( Figure 7B) (Peccerillo and Taylor, 1976), the studied samples plot in the fields of the high-K calc-alkaline and shoshonitic series, with high K 2 O content ranging from 4.01 to 5.69 wt.% and K 2 O/Na 2 O ratios from 1.17 to 1.64 (Supplementary Table S4). All samples are metaluminous and/or slightly peraluminous granitoids as indicated by their A/CNK ratios (0.95-1.10), and mostly they plot on the left side of the I-S line (A/CNK 1.1) in the A/NK vs. A/CNK diagram (Figure 8) (Maniar and Piccoli, 1989).
One of the common features of the Permian granitoids is their significant negative Nb and Ta anomalies in the primitive  Supplementary Table S3 for data sources) have older T DM2 model ages concentrated at 1.3-1.7 Ga (older than 2.0 Ga for inherited zircons), compared with (B) the plume-related Permian A-type granitoids in the Tarim Craton (data sources can be found in Su et al., 2019;Wei et al., 2019;Zhang et al., 2010), which have younger T DM2 model ages mostly ranging from 540 Ma to 1,500 Ma.  Cox et al., 1979) showing that the samples plot in the fields of granite and granodiorite, which is consistent with their main mineral compositions shown in Figure 3; (B) K 2 O (%) vs. SiO 2 (%) diagram (after Peccerillo and Taylor, 1976) showing that the investigated granitoids belong to the high calc-alkaline to shoshonitic series.
Frontiers in Earth Science | www.frontiersin.org February 2022 | Volume 9 | Article 831677 mantle-normalized spider diagrams ( Figure 9A) (Sun and McDonough, 1989), which are also characterized by conspicuous enrichments of some LILEs such as Rb, Th, U, and evident depletion of Ba and Sr. The Permian granitoids have similar chondritenormalized rare earth element (REE) patterns as indicated by their significant enrichments in light rare earth elements (LREEs) relative to heavy rare earth element (HREE) with (La/Yb) N values between 7.8 and 37.9 ( Figure 9B; Supplementary Table S4). All samples demonstrate a significant fractionated REE pattern indicated by high (La/ Sm) N values between 3.3 and 5.2. Except for sample 20ST93 (0.73-0.79), the remaining five samples show distinct negative Eu anomalies (0.26-0.52) with a concave-upward shape ( Figure 9B).
The A-type affinity of the Permian granitoids is further demonstrated by their relatively high melting temperature (824-875°C) based on the Zr saturation in the magma system (Watson and Harrison, 1983). This temperature is consistent with that of typical A-type granites, but it is much higher than that of I-type granites (781°C and 764°C for unfractionated and fractionated I-type granites, respectively) (King et al., 1997). Moreover, all samples are characterized by high Zr/Hf ratios (35-43), indicating they are moderately fractionated compared with highly fractionated granitoids (Wu et al., 2017). Therefore,  Maniar and Piccoli, 1989) showing the slightly peraluminous to metaluminous nature of the studied Permian granitoids.  Sun and McDonough (1989 we consider that the Permian granitoids in this study are A-type granites.

Petrogenesis and Geodynamic Setting
As noted earlier, there is currently no consensus regarding the petrogenesis and geodynamic setting responsible for the Permian granitoids in the South Tianshan. This section integrates all available regional geological, geochronological and geochemical data in order to evaluate the existing controversies, after which we propose an alternative geodynamic scenario to explain the generation of the Permian granitoids.

Genetically Related to the Tarim Mantle Plume?
One major interpretation is that the Permian granitoids in the South Tianshan are genetically related to the mantle plume in the Tarim Craton (e.g., Zhang and Zou, 2013); this model is mainly based on the coincidence in time of the plume compared with that of the putative plume-induced A-type granites (Su et al., 2019;Wei et al., 2019). However, there are significant differences in the characteristics of the A-type granites in the two areas as follows: Firstly, the Permian A-type granites in the western Tarim Craton are spatially and temporally closely associated with voluminous plume-derived basalts, mafic-ultramafic intrusive complexes and abundant alkali mafic dykes (Zou et al., 2015). In contrast, in the South Tianshan the Permian magmatism lacks these hallmarks and instead is dominated by felsic intrusions. Furthermore, the well documented A-type granitoids in the South Tianshan (Huang et al., 2015a;Konopelko et al., 2007Konopelko et al., , 2009Long et al., 2008;Wang et al., 2007b) mostly have an A 2 -subtype affinity based on the ternary discrimination diagrams FIGURE 10 | Genetic discrimination diagrams (after Whalen et al., 1987) showing the A-type affinity of the Permian granitoids in this study.  Frontiers in Earth Science | www.frontiersin.org February 2022 | Volume 9 | Article 831677 ( Figure 12), and that is significantly different from the plumeinduced A 1 -subtype granites in the Tarim Craton (Yang et al., 2007;Zhang et al., 2008;Wei and Xu, 2011;Huang et al., 2012;Su et al., 2019;Wei et al., 2019). Secondly, the A-type granites in the South Tianshan have distinctive negative Nb and Ta anomalies ( Figure 9A), which are commonly indicative of a subduction-influenced origin, whereas those in the western Tarim have apparent positive Nb and Ta anomalies (Figure 3, Wei et al., 2019;Figure 7, Zhang and Zou, 2013; Figure 8B, Zou et al., 2015).
Thirdly, there is a general Hf isotopic difference between the plume-related A-type granites of Tarim and those in the South Tianshan. This not only demonstrates the distinctive Hf compositions of the studied granitoids, but also highlights the contrasting difference in their magma sources. As shown in Figures 5, 6, the A-type granites in the western Tarim are characterized by distinctive positive ε Hf (t) values and corresponding younger crustal model ages (Zou et al., 2015). The relatively depleted feature of the A-type granites in the Tarim Craton is interpreted to result from intensive fractionation of a plume-related OIB-like basaltic magma (Zou et al., 2015), or from partial melting of newly underplated rocks induced by the mantle plume, followed by extensive fractionation and minor crustal assimilation (Su et al., 2019). In contrast, the A-type granitoids in the South Tianshan display apparently negative to slightly positive ε Hf (t) values (−10.9 to +0.6) and older crustal model ages (1727( -1759, as compared with those in the western Tarim. This is consistent with the Nd isotope compositions (ε Nd (t) −1.6 to-6.9) of the A-type granites from Kyrgyzstan (Konopelko et al., 2007), indicating their derivation from very different magma sources (section 6.2.3). The molar oxide discrimination diagram of CaO/(MgO + FeO t ) vs. Al 2 O 3 / (MgO + FeO t ) (Altherr et al., 2000) indicates that they were mainly generated by partial melting of meta-greywackes (Figure 13), suggesting an arc-related tectonic setting (Yin et al., 2021), which is in accordance with the deep Nb and Ta troughs reflected by the multi-element spider diagram for the granitoids ( Figure 9A).
Last but not least, the difference is also shown by the lower zircon saturation temperatures (824-875°C) of the A-type granites in the South Tianshan, as compared with those in the western Tarim (890-1,010°C) (Su et al., 2019), which likely suggests the magmas were generated by fundamentally different crustal melting processes.
The above significant observations demonstrate the notable differences between the A-type granites in these two tectonic environments from which it is evidence that the geodynamic process for generation of the A-type granites in the South Tianshan was different from that responsible for those in the Tarim Craton (Wei et al., 2019). This reasoning is supported (Xiao et al., 2013;Abuduxun et al., 2021a) by the fact that the South Tianshan was still separated from the Tarim Craton in the Early Permian when the mantle plume was active. Consequently, we conclude that it is unlikely that the Tarim mantle plume was responsible for the A-type magmatism in the South Tianshan.

Products of Post-collisional Magmatism?
A second common interpretation for the petrogenesis of the Permian granitoids in the South Tianshan is that they formed in a FIGURE 12 | Discrimination diagrams (after Eby, 1992) showing the dominant A 2 -subtype affinity of the A-type granitoids in the South Tianshan (literature data sources: Konopelko et al., 2007Konopelko et al., , 2009Huang et al., 2015a;Long et al., 2008;Wang et al., 2007b), which is significantly different from the plume-related A 1 -subtype affinity of the granites in the Tarim craton (data sources: Huang et al., 2012;Su et al., 2019;Wei et al., 2019;Wei and Xu, 2011;Yang et al., 2007;Zhang et al., 2008).  Altherr et al., 2000) showing that the Permian granitic magma was mainly generated by the melting of meta-greywackes composed of arcrelated volcanogenic sedimentary rocks.
Given the widespread occurrence of geologically and structurally well investigated OPS mélanges, ophiolites and imbricated trench-filled turbidites, the South Tianshan orogen was considered to be an accretionary complex that formed on the southern active margin of the Yili-Central Tiranshan arc by continuous northward subduction of the South Tianshan oceanic lithosphere (Xiao et al., 2013;Sang et al., 2018;Abuduxun et al., 2021b). A further line of geochronological evidence indicates that the closure of the South Tianshan Ocean lasted to the End-Permian to Late Triassic (Li et al., 2005;Sang et al., 2017;Wen et al., 2020;Abuduxun et al., 2021a). That is to say, the South Tianshan was on the leading edge of the upper plate (the Yili-Central Tianshan arc), and was separated from the Keping passive margin by a wide ocean until the Late Permian to Late Triassic when the terminal suture formed along the base of the accretionary complex (Xiao et al., 2013;Abuduxun et al., 2021a). Accordingly, it is apparent that the Permian granitoids in the South Tianshan formed in the upper plate, rather than in the lower plate, because the terminal suture between the Yili-Central Tianshan arc and the Tarim Craton is roughly along the North Tarim Fault to the south (Xiao et al., 2013;Abuduxun et al., 2021a), and not along the Atbashi-Inylchek-South Nalati Fault to the north (e.g., Wang et al., 2018b).
Critically, as discussed in section 6.2.1, the Permian A-type granites in the Keping area, where the leading edge of the lower plate was located, were genetically related to a mantle plume, and bear no resemblance in either rock associations or elemental and isotopic geochemistry to those in the South Tianshan. The key implication from the above reasoning is that it is unlikely that the Permian plutons in the South Tianshan formed across the terminal suture, as previously considered (e.g., Han et al., 2011;Wang et al., 2018a), but rather that they were confined to the upper plate beneath which north-dipping subduction of the oceanic lithosphere was still active. Thus, the post-collision model is implausible.

An Alternative Petrogenetic-Tectonic Model
As demonstrated in section 6.1, a result of the genetic classification based on major and trace element geochemistry, integrated with high melting temperatures (824-875°C), is that the Permian granitoids in the South Tianshan have an A-type affinity.
There has long been a general consensus that A-type granites mostly formed in extensional tectonic environments regardless of the source of their magmas (Whalen et al., 1987;Wu et al., 2017).
All our studied samples belong to the A 2 -subtype ( Figure 12), which is widely considered to be associated with extension events in convergent margins (Eby, 1992). Nevertheless, a distinctive feature of our Permian A-type granitoids is that they have diagnostic subduction-related trace element signatures such as deep Nb and Ta troughs ( Figure 9A), elevated LILEs and flat HFSEs patterns ( Figure 9B), all very similar to those of granites in the Lachlan accretionary orogen in Australia, which robust analysis has convincingly demonstrated to have formed in a subductiongenerated arc (e.g., Collins et al., 2019). Accordingly, we integrated our new results with published, high-quality, comprehensive, sedimentological, structural and geochronological data (Xiao et al., 2013;Sang et al., 2018;Abuduxun et al., 2021a), which conformed that the Permian A-type granitoids in the South Tianshan formed in an extensional supra-subduction setting.
Corroborative evidence for such extension comes from the presence of Early Permian extensional faulted basins in the South Tianshan (Liu et al., 2013) and the ca. 272 Ma high-temperature granulite facies metamorphism in the Yushugou area (Zhang et al., 2019b). Apatite fission track (AFT) thermochronology results have also suggested that the southwestern Tianshan had no strong uplifting in the Permian (Dumitru et al., 2001).
In addition, of the 299-282 Ma rhyolitic and basaltic lavas from the Xiaotikanlike Formation in the Heiyingshan section of the South Tianshan , the basaltic lavas have trace and rare element patterns similar to those of oceanic island basalts (Huang et al., 2015b), and the rhyolites have relatively high Zr saturation temperatures (up to 824°C), which can be most likely attributed to an increased heat input from underplated mantlederived basaltic magmas (Cheng et al., 2017). Therefore, these volcanic rocks are consistent with and indeed point to formation in an Early Permian extensional setting in the South Tianshan.
A similar Early Permian extensional situation is also reported from the Yili-Central Tianshan. The geological structure profile based on seismic data in the Zhaosu-Tekesi Depression shows that the Lower Permian strata are characterized by titled rotation downward near the NE-trending normal fault, indicating a sedimentary filling pattern of the extensional faulted (or rifted) basin and related thermal subsidence in the southern Yili Block (Li D. et al., 2015). The Early Permian A-type granites in the Yili-Central Tianshan (Xu et al., 2013;Li N.-B. et al., 2015) also require an extensional environment for the high-temperature magmatism.
Taking into consideration the fact that formation of these Permian granitoids was structurally confined to a narrow linear belt along the strike of the orogen ( Figure 1B), we consider that the subduction-related extension was most likely triggered by slab rollback, which usually causes major contemporaneous thermal anomalies along the strike of a subduction zone. Again, we emphasize that neither the Tarim mantle plume model nor the post-collisional extension model can explain the arc-parallel linear distribution of the Permian granitoids.
Frontiers in Earth Science | www.frontiersin.org February 2022 | Volume 9 | Article 831677 Given the evidence presented above, we propose a new petrogenetic-tectonic model as visualized in Figure 14. The southward rollback of the northward subducting South Tianshan oceanic lithosphere triggered extension in the overring plate and associated upwelling of asthenospheric mantle in the Early Permian. The southward slab rollback consequently provided a suitable condition for magma underplating combined with increased temperature leading to large-scale crustal melting, which gave rise to a broad, linear belt of arc magmatism in the southern active margin of the Yili-Central Tianshan ( Figure 1B) where the A-type granitoids were produced as a result of the high thermal gradient.
The A-type granites have slightly enriched Hf isotopic compositions ( Figure 5; ε Hf (t) −10.9 to +0.6), indicating that neither mantle nor juvenile crustal material could be major contributors to their magma generation. On the other hand, this implies that the corresponding older crustal model ages, mainly concentrated at 1,272-1759 Ma, were generated by remelting of Mesoproterozoic recycled crustal components, and those samples that plot in the meta-greywacke field in the molar oxide discrimination diagram ( Figure 13) were no doubt mainly derived by partial melting of arc-related volcanogenic sedimentary rocks (Yin et al., 2021). Therefore, mature crustal material with minor mantle-derived magmatic inputs were the most feasible magma source for the investigated A-type granites.
In summary, subduction-related extension triggered by slab rollback was the most likely tectonic environment for generation of the A-type granitoids in the South Tianshan. The ambient processes predictably involved the upwelling of asthenosphere, decompression melting, and an associated rise in temperature, which all led to large-scale partial melting in the upper plate.

Implications for the Retreating Accretionary Orogenesis of the Southern Altaids
A specific comparison of the rock associations, Lu-Hf isotopic compositions, whole-rock geochemistry, and magma conditions highlights the contrasting differences between the A-type granites in the South Tianshan and those in the western Tarim Craton. Consequently, it is unlikely they were involved in the same geodynamic process. That is to say, the South Tianshan was not affected by the Permian Tarim mantle plume. This is supported by provenance analysis, which provided no evidence for the presence of voluminous plume-related detritus in the Permian-lower Triassic strata from the South Tianshan (Liu et al., 2013;Abuduxun et al., 2021a). Therefore, a robust analysis of all variable factors leads to the conclusion that the South Tianshan and the Tarim Craton was still separated by a wide ocean in the Permian, and accordingly this negates any idea of pre-Permian closure of the South Tianshan ocean (Han et al., 2011;Wang et al., 2018b;Alexeiev et al., 2019). Therefore, it is reasonable to suggest that the South Tianshan was unlikely a plume-modified collisional orogeny . A significant Late Carboniferous magmatic "flare-up"event has been well documented in the Yili-Central Tianshan arc, which is characterized by coeval mafic and felsic magmatic rocks, such as ca. 337-322 Ma bimodal volcanic rocks in the Wusun mountain (Su et al., 2021), ca. 317-310 Ma mafic dikegranitoid associations in the Zhongyangchang area (Tang et al., 2014), and ca. 314 Ma adakitic granodiorites in the Qiongkusitai area (Yin et al., 2016). Geochemically, they show a clear arcaffinity; moreover, the mafic rocks were likely derived from a depleted mantle source containing an asthenospheric component (Tang et al., 2014). Therefore, the magmatic "flare-up"has been interpreted to be formed in a subduction-related extensional setting triggered by southward rollback of the South Tianshan oceanic lithosphere (e.g., Tang et al., 2014;Yin et al., 2016). If integrated with the Permian A-type granitoids in the South Tianshan, these extension-related magmatic rocks show a trend of southward migration. This observation is also supported by the southward growth of the South Tianshan accretionary complex as indicated by the oceanward younging of the maximum depositional ages of trench-filled turbidites (Abuduxun et al., 2021a, b;Fu et al., 2018). Therefore, it is reasonable to propose that the Late Carboniferous-Early Permian broad magmatic arc with diverse extension-related FIGURE 14 | Schematic model illustrating the tectonic relations that were responsible for the Early Permian A-type granitoids in the South Tianshan. In the Early Permian, upper plate extension was probably related to southward slab rollback that induced upwelling of asthenospheric mantle, which led to large-scale crustal melting that gave rise to the broad granitic arc in the southern active margin of the Yili-Central Tianshan. Formation of the A-type granitoids was facilitated by the high heat gradient beneath the extensional upper plate.
Frontiers in Earth Science | www.frontiersin.org February 2022 | Volume 9 | Article 831677 magmatism that requires a high temperature in the southern active margin of the Yili-Central Tianshan was built probably as a result of prolonged southward rollback of the progressively aging South Tianshan oceanic lithosphere before its final closure in the End-Permian to Late Triassic (Xiao et al., 2013;Sang et al., 2018;Wen et al., 2020;Abuduxun et al., 2021a), and that led to the final amalgamation of the Yili-Central Tianshan arc with the Tarim Craton (Gao et al., 2011;Han et al., 2011;Xiao et al., 2015). The newly identified Permian A-type granites in this study share the same geochemical and isotopic characteristics with widespread coeval A-type granites in Kyrgyz South Tianshan (Konopelko et al., 2007(Konopelko et al., , 2009. Therefore, the retreating and extensional accretionary structure of the southern Altaids is well recorded by the linear distributed and spatiotemporal-orientated Late Carboniferous-Early Permian magmatic "flare-up" along the southern active margin of the Yili-Central Tianshan from Kyrgyzstan to NW China.

CONCLUDING REMARKS
LA-ICP-MS zircon U-Pb dating reveals that the newly identified A-type granitoids from the South Tianshan in the southern Altaids were emplaced at ca. 298-272 Ma. The whole-rock geochemistry and calculated zircon saturation temperatures indicate that these are metaluminous to slightly peraluminous A 2 -type granites. The zircon Hf isotopic compositions show that the granitoids were mainly generated by partial melting of mature crustal components with a minor contribution from a mantlederived magmatic source.
The A-type granitoids in the South Tianshan are incontrovertibly different from those in the western Tarim Craton in terms of mantle plume generation. Integration of the new and previous data leads to the conclusion that they were produced due to a high temperature gradient in an extensional, subduction-generated environment probably induced by southward slab rollback of the South Tianshan oceanic lithosphere, rather than due to post-collisional extension.
In the Permian the South Tianshan in the southern Altaids was a retreating extensional accretionary orogen rather than a plumemodified collisional orogen.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
NA provided initial data. WX designed the study. BW corrected and improved the writing of the manuscript. NA, PH, JG conducted the fieldwork. HY participated in analysis on petrology and photomicrographs. MS helped with the LA-CIP-MS zircon U-Pb dating. XL helped with the Hf isotopic analysis. All authors participated in discussion of the study and approved the submitted version.

ACKNOWLEDGMENTS
Careful reviews and insightful suggestions from the reviewers greatly improved the manuscript. Editorial handling by Dr. Xiubin Lin is highly appreciated.