Early Devonian Arc-Related Volcanic Rocks in the Haerdaban, North Margin of the Yili Block: Constraint on the Southward Subduction of the Junggar Ocean

The origin and tectonic implication of Early–Middle Devonian magmatism in the northern margin of YB (Yili Block) remain enigmatic and are important for understanding Late Paleozoic evolution of the Junggar Ocean and southern Kazakhstan Orocline. Here, we present the systematic study of whole-rock geochemical and Sr–Nd isotope features as well as U–Pb–Hf isotope characteristics of zircon crystals for newly identified Early Devonian volcanic rocks from the northern margin of YB. The volcanic rocks are composed of rhyolite, rhyolite porphyry, and rhyolitic tuff. Zircon U-Pb age dating indicates they were formed at ca. 407~418 Ma. They have high SiO2 (70.16–77.52 wt.%) and alkali (5.10–9.56 wt.%) contents, and high Zr + Nb + Ce + Y content (~456 ppm), indicative of A-type magma. Their relative depletion of Nb, Ta, and Ti, and enrichment of LILEs show arc affinity. Their low initial 87Sr/86Sr ratios (0.699708–0.709822) and negative εNd(t) values (−1.8 to −4.0) indicate a mainly continental magma source and their positive εHf(t)values (+6.13 to +14.81) are possibly due to the garnet effect. All these above reveal that volcanic rocks were generated by re-melting of lower crust under a high temperature condition, which was induced by long-lived heat accumulation with no or minimal basalt flux. Combined with active continental margin inference evidenced by contemporaneous sedimentary rocks, we attribute the generation of the volcanic rocks to a continental arc setting related to the southward subduction of Junggar oceanic crust. Thus, we infer the Early–Middle Devonian arc-related magmatic rocks in the northern margin of YB are eastward counterparts of the southern limb of the Devonian Volcanic Belt, which resulted from a relatively steady-state southward subduction.


Introduction
The Kazakhstan Orocline is one of the major oroclines in the Central Asian Orogenic Belt (CAOB), which was formed by the bending of the originally linear Kazakhstan composite arc in the Carboniferous to Early Permian and further tightened to form the present orocline in the Late Permian to Early Triassic [1]. It contains a collage of Precambrian continental fragments and several Paleozoic island arcs [2,3]. The major Cambrian and Early Ordovician arc systems distribute along the northern limb [3], and the Late Paleozoic was formed by the southward subduction of the North Tianshan Ocean in the Late Paleozoic [19,[27][28][29]. The STAC was formed by the northward subduction of the South Tianshan Ocean in the Paleozoic and it mainly composed of lower Cambrian to Carboniferous sedimentary rocks, ophiolites and metamorphic rocks [28,30,31]. KYB and CTB are two major microcontinents with Precambrian metamorphic basements in the westernmost of the Chinese Tianshan.  [32]). (b) Simplified geological map of the Chinese Tianshan belt (modified after [33]). (c) Geological map of the Biezhentao Mountain area and location of the study area (modified after [34][35][36]).
The YB, an important Precambrian continental fragment in the southern limb of the Kazakhstan collage system, is a triangular area that extends westward into Kazakhstan, and it is bordered by the North Tianshan Suture to the north and North Nalati Suture to the south (Figure 1b). The Precambrian rocks are mainly exposed in the north and south margin of YB, and unconformably overlain by the Paleozoic sedimentary and volcanic rocks, which are intruded by the voluminous Ordovician-Permian granitoids [11,28,31,34,35,37,38]. The Precambrian rocks mainly outcrop in its northern and southern margins discontinuously, for example, the amphibolite, migmatite, quartzite, marble, and various schist in the Wenquan and Sayram areas [34]. The Cambrian-Silurian rocks consist mainly of cherts, siltstones, carbonates, felsic pyroclastic rocks, and normal clastic sedimentary rocks, and outcrop mainly in the Boluokenu Mountain and the sedimentary sequences are complete, although the exposed area of Cambrian-Ordovician strata is limited. Late Paleozoic to Mesozoic strata have wide distribution in the YB and is characterized by sandstone, siltstone, and shale interbedded with volcanic rocks [35]. As mentioned above, volumes of Paleozoic magmatic rocks including both volcanic and intrusive rocks distribute along the north margin of YB, but the Late Silurian to Early Devonian magmatism is poorly reported.
The study area is located in southern region of the Boertala River, Wenquan County. The Proterozoic strata are widely distributed along the Biezhentao Mountain, and the Neoproterozoic gneissic granites were recognized in the Wenquan Complex (Figure 1c). The Precambrian strata are overlain unconformably by the Paleozoic rocks and they are in-  [32]). (b) Simplified geological map of the Chinese Tianshan belt (modified after [33]). (c) Geological map of the Biezhentao Mountain area and location of the study area (modified after [34][35][36]).
The YB, an important Precambrian continental fragment in the southern limb of the Kazakhstan collage system, is a triangular area that extends westward into Kazakhstan, and it is bordered by the North Tianshan Suture to the north and North Nalati Suture to the south ( Figure 1b). The Precambrian rocks are mainly exposed in the north and south margin of YB, and unconformably overlain by the Paleozoic sedimentary and volcanic rocks, which are intruded by the voluminous Ordovician-Permian granitoids [11,28,31,34,35,37,38]. The Precambrian rocks mainly outcrop in its northern and southern margins discontinuously, for example, the amphibolite, migmatite, quartzite, marble, and various schist in the Wenquan and Sayram areas [34]. The Cambrian-Silurian rocks consist mainly of cherts, siltstones, carbonates, felsic pyroclastic rocks, and normal clastic sedimentary rocks, and outcrop mainly in the Boluokenu Mountain and the sedimentary sequences are complete, although the exposed area of Cambrian-Ordovician strata is limited. Late Paleozoic to Mesozoic strata have wide distribution in the YB and is characterized by sandstone, siltstone, and shale interbedded with volcanic rocks [35]. As mentioned above, volumes of Paleozoic magmatic rocks including both volcanic and intrusive rocks distribute along the north margin of YB, but the Late Silurian to Early Devonian magmatism is poorly reported.
The study area is located in southern region of the Boertala River, Wenquan County. The Proterozoic strata are widely distributed along the Biezhentao Mountain, and the Neoproterozoic gneissic granites were recognized in the Wenquan Complex ( Figure 1c). The Precambrian strata are overlain unconformably by the Paleozoic rocks and they are intruded by the Paleozoic plutons. As shown in the geological map of Baskan Pass-Huocheng Sheet, the Late Paleozoic strata distribute in the Haerdaban area and mainly comprise sandstone, conglomerate, and limestone, which was originally divided into Upper Devonian and Low Carboniferous [34]. Recently, our investigation found that volcanic rocks can be recognized in the Agutu area ( Figure 2a) and Buergasite area (Figure 2b), which includes rhyolite, rhyolite porphyry, and rhyolitic tuffs. The outcrops of volcanic rock are mostly domed and a few are intercalated (Figure 3a). The lava and clasolite are mostly in gradual contact and it is difficult to draw a boundary. truded by the Paleozoic plutons. As shown in the geological map of Baskan Pass-Huocheng Sheet, the Late Paleozoic strata distribute in the Haerdaban area and mainly comprise sandstone, conglomerate, and limestone, which was originally divided into Upper Devonian and Low Carboniferous [34]. Recently, our investigation found that volcanic rocks can be recognized in the Agutu area ( Figure 2a) and Buergasite area (Figure 2b), which includes rhyolite, rhyolite porphyry, and rhyolitic tuffs. The outcrops of volcanic rock are mostly domed and a few are intercalated (Figure 3a). The lava and clasolite are mostly in gradual contact and it is difficult to draw a boundary.   truded by the Paleozoic plutons. As shown in the geological map of Baskan Pass-Huocheng Sheet, the Late Paleozoic strata distribute in the Haerdaban area and mainly comprise sandstone, conglomerate, and limestone, which was originally divided into Upper Devonian and Low Carboniferous [34]. Recently, our investigation found that volcanic rocks can be recognized in the Agutu area ( Figure 2a) and Buergasite area (Figure 2b), which includes rhyolite, rhyolite porphyry, and rhyolitic tuffs. The outcrops of volcanic rock are mostly domed and a few are intercalated (Figure 3a). The lava and clasolite are mostly in gradual contact and it is difficult to draw a boundary.   Fifteen samples of volcanic rocks from two areas were collected, including rhyolite, rhyolite porphyry, and rhyolitic tuffs (Figure 3a,b). Rhyolite samples are gray to dark gray, porphyritic, and have typical flow structures (Figure 3b). The phenocrysts (~15%) are mainly composed of quartz and K-feldspar and range in size from 0.2 to 1.5 mm (Figure 3e). The groundmass is mainly composed of aphanitic feldspar and quartz, and minor amounts of biotite and sericite (Figure 3f). The rhyolitic tuffs samples are gray to tawny and contain~45% volcanic ash,~35% rhyolite detritus, and~25% crystal fragments (Figure 3c,d,f). Sometimes, the augen or mylonite structure develops in the volcanic rocks after the experience of later deformation (Figure 3d).

Analytical Methods
The detailed analytical methods involved in this study including zircon U-Pb dating, whole-rock major and trace element analyses, zircon Lu-Hf, and Sr-Nd isotopes analyses.

Zircon U-Pb and Hf Isotope Analyses
Zircon crystals were selected from rhyolitic samples and embedded in epoxy resin, and then polished down to expose the interior of the crystals. Cathodoluminescence (CL) images were taken and used to demonstrate the internal texture of the zircons and to select optimum spot locations during the laser-ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). Zircon U-Pb isotopic data were collected from 32 µm diameter regions of single grains by use of ICP-MS (Agilent 7700) and a laser-ablation system at the Key Laboratory for the Study of Focused Magmatism and Giant Ore Deposits, MNR, Xi'an. The adopted laser-ablation system is a GeoLas 200M equipped with a 193 nm laser. Harvard zircon 91,500 was used as the external standard for isotopic fractionation, NIST SRM610 glass was used as the external calibration standard, and 29 Si was used as the internal standard for the elemental analyses. Isotopic ratios and element concentrations for zircons were calculated with the software Glitter 4.0 (Macquarie University). The ages were calculated with the program Isoplot (ver. 2.94), an add-in for Microsoft Excel [39]. The detailed analytical procedures for age and trace element determinations of zircons were similar to those described in the references [40].
Zircon Hf isotope compositions were examined simultaneously from the same spot. The 193 nm laser ablation system attached to a Neptune Plasma Multicollector-ICP-MS at the same laboratory. The analysis was undertaken under a laser repetition rate of 10 Hz at 200 mJ and an ablated spot size of 60 µm. The measured 176 Lu/ 177 Hf ratios and the 176 Lu decay constant of 1.867 × 10 −11 yr −1 were used to calculate initial 176 Hf/ 177 Hf ratios [41]. The present-day 176 Hf/ 177 Hf = 0.28325 and 176 Lu/ 177 Hf = 0.0384 values were used to calculate the depleted mantle model age (T DM ) [42]. The chondritic values of 176 Hf/ 177 Hf = 0.0332 and 176 Lu/ 177 Hf = 0.282772 were used for the calculation of ε Hf values [43]. The detailed analytical procedures were described by reference [44].

Whole-Rock Major and Trace Element Analyses
Samples were analyzed for whole-rock chemical composition at the Key Laboratory for the Study of Focused Magmatism and Giant Ore Deposits, MNR, Xi'an. The major elements were analyzed by wet chemical techniques and X-ray fluorescence spectroscopy (XRF-1500). The analytic precision of major elements ranged from 1% to 5%. Trace and rare earth elements (REEs) were determined by ICP-MS (Element II) and the analytic precision was generally better than 5%.

Whole-Rock Sr and Nd Isotopes Analyses
The Rb-Sr and Sm-Nd isotopic analysis followed procedures similar to those described by reference [45]. Whole-rock powders for Sr and Nd isotopic analyses were dissolved in Savillex Teflon screw-top capsule after being spiked with the mixed 87 Rb-84 Sr and 149 Sm-150 Nd tracers prior to HF + HNO 3 + HClO 4 dissolution. Rb, Sr, Sm, and Nd were separated by use of the classical two-step ion exchange chromatographic method and the isotopic compositions were obtained using a Thermo Fisher Scientific Triton Plus multi-collector thermal ionization mass spectrometer at IGGCAS (State Key Laboratory of Lithospheric Evolution, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China). The whole procedure blank was lower than 250 pg for Rb-Sr and 100 pg for Sm-Nd. The isotopic ratios were corrected for mass fractionation by normalizing to 88 Sr/ 86 Sr = 8.375209 and 146 Nd/ 144 Nd = 0.7219, respectively. The international standard samples, NBS-987 and JNdi-1, were employed to evaluate instrument stability during the period of data collection. The measured values for the NBS-987 Sr standard and JNdi-1 Nd standard were 87 Sr/ 86 Sr = 0.710264 ± 0.000012 (n = 4, 2SD) and 143 Nd/ 144 Nd = 0.512114 ± 0.000009 (n = 3, 2SD), respectively. USGS reference material BCR-2 was measured to monitor the accuracy of the analytical procedures, with the following results: 87 Sr/ 86 Sr = 0.705022 ± 0.000011 and 143 Nd/ 144 Nd = 0.512632 ± 0.000010. The 87 Sr/ 86 Sr and 143 Nd/ 144 Nd data of BCR-2 show good agreement with previously published data by TIMS technique [45].

Results
The analytical results are listed in the Supplementary Tables S1-S4, respectively.

Zircon U-Pb Dating and Hf Isotopic Composition
The sampling locations for U-Pb and Hf isotope analyses are shown in Figure 2.  Figure 4h) are interpreted as the crystallization age of the rhyolite porphyry. They yield high 176 Hf/ 177 Hf ratios (ranging from 0.282692 to 0.282911), and high ε Hf (t) value (from +6.13 to +12.99) with Hf model ages ranging from 554 to 1015 Ma.

Whole-Rock Geochemical Composition
Fifteen samples of volcanic rocks, including thirteen rhyolite and rhyolite porphyry samples and two rhyolitic tuff samples from the Haerdaban area, were selected to analyze the whole-rock major and trace elemental compositions. Results are listed in Supplementary  Table S3

Whole-Rock Geochemical Composition
Fifteen samples of volcanic rocks, including thirteen rhyolite and rhyolite porphyry samples and two rhyolitic tuff samples from the Haerdaban area, were selected to analyze the whole-rock major and trace elemental compositions. Results are listed in Supplementary Table S3   Six rhyolite/rhyolite porphyry samples were selected to analyze the whole-rock Sr and Nd isotopic compositions. The results are listed in the Supplementary Table S4. The samples have relatively low initial 87 Sr/ 86 Sr ratios (0.699708 to 0.709822), and show lower initial 87 Sr/ 86 Sr ratios comparing with the Middle to Late Devonian granite in the study area (0.711603-0.712615) [11]. They have initial 143 Nd/ 144 Nd ratios from 0.511909 to 0.512008 and negative εNd(t) values (−1.8 to −4.0) (Figure 5b), with two-stage Nd model ages ranging from 1305 to 1476 Ma.

Petrogenesis
The Early Devonian volcanic rocks from the Haerdaban area in the northern margin of YB are characterized by high SiO2 (70.16 to 77.52 wt.%), Na2O+K2O (5.10 to 9.56 wt.%), and Zr (156-411 ppm) contents, as well as low concentrations of Al2O3 (11.27 to 16.36 wt.%), CaO (0.21 to 2.57 wt.%), and MgO (0.1 to 1.22 wt.%). On the primitive mantle-normalized spider diagram (Figure 7b), these samples show the strong depletion of Ba, Sr, P, and Ti, and the enrichment of Th, U, and K indicates that they have experienced a notably fractional crystallization. The high differentiation index (DI = 82.6-96.8) and the Zr/Hf ratios (27.1-36.5) suggest the moderately evolved granites attribute [52][53][54]. Whole-rock Zr saturation temperatures (TZr) were calculated and the results have a range from 783 °C to 952 °C (average TZr = 860 °C). It is similar to those of typical A-type granites (800-900 °C) [55], but lower than those plume-related A-type granites (930-1050 °C,) [56] and higher than I-type granites and S-type granites [57]. These geochemical signatures, including high SiO2 and Zr contents, as well as the high zircon saturation temperatures, are consistent with their classification as A-type granite [52,54,58]. In the discrimination diagrams (Figure 8a,b), the rocks fall into the field of A-type granites. Sometimes, the high-Si A-type granites may show the similar geochemical characteristics with the high fractionated Itype granites [58]. The high Fe # values (Fe # = FeO T /(FeO T + MgO), from 0.67 to 0.93) and high Zr + Nb + Ce + Y (average 456 ppm) content are obviously different from those high fractionated I-type granites [48,59,60].
The petrogenesis of A-type alkaline magmas is still a matter of debate. Commonly, the A-type granites can be produced by the fractionation of mantle-derived magma, crust anatexis, or mixing of mantle-derived magma and crustal rock [58,61,62]. Fractionation of mantle-derived magma or re-melting of under-plating basaltic rocks generally produces small amounts of silicic magma [63]. Therefore, a significant quantity of mafic magma and mafic residues were required in such a model [64,65]. The metaluminous to weakly peraluminous composition of the samples are inconsistent with the peralkaline attribute of fractionation mantle-derived magma (Figure 6d) [53,57]. The lack of mafic rocks, high SiO2  [52][53][54]. Whole-rock Zr saturation temperatures (T Zr ) were calculated and the results have a range from 783 • C to 952 • C (average T Zr = 860 • C). It is similar to those of typical A-type granites (800-900 • C) [55], but lower than those plume-related A-type granites (930-1050 • C,) [56] and higher than I-type granites and S-type granites [57]. These geochemical signatures, including high SiO 2 and Zr contents, as well as the high zircon saturation temperatures, are consistent with their classification as A-type granite [52,54,58]. In the discrimination diagrams (Figure 8a,b), the rocks fall into the field of A-type granites. Sometimes, the high-Si A-type granites may show the similar geochemical characteristics with the high fractionated I-type granites [58]. The high Fe # values (Fe # = FeO T /(FeO T + MgO), from 0.67 to 0.93) and high Zr + Nb + Ce + Y (average 456 ppm) content are obviously different from those high fractionated I-type granites [48,59,60]. crust [67]. As shown in previous studies, there is little or no magmatic event from Early Silurian to Late Silurian (440 Ma to 420 Ma) in the northern margin of YB [9,18], representing a long-time relatively stable condition. Despite no or little basalt flux during the Early Devonian, the long-lived heat accumulation in a thick crust could induce a magmatic burst after more than 20 Ma [67,68]. Therefore, the Haerdaban volcanic rocks were mainly derived from the partial melting of felsic crustal rocks induced by the long-lived heat accumulation with no or minimal basalt flux. Figure 8. Plot of (Na2O + K2O)/CaO versus Zr + Nb + Ce + Y, FeOT/MgO versus Zr + Nb + Ce + Y discrimination diagrams for the Early Devonian volcanic rocks from Haerdaban area in the northern margin of Yili Block (a,b, reference area after [58]).
The low Sr (41.2 to 120 ppm) and high Rb (56.2 to 372 ppm) concentrations of the samples imply that the fractional crystallization may be involved in the production of the A-type volcanics. The depletion of Ba, Sr, P, Ti, and Eu anomalies in the trace element patterns, as well as the high differentiation index also supported the fractional crystallization. The fractional crystallization of plagioclase and K-feldspar minerals can account for the depletion of Ba, Sr, and Eu. The P and Ti anomalies can be attributed to the fractionation of apatite and Fe-Ti oxides. It is noted that the εNd(t) values of samples range from −1.8 to −4.0, but εHf(t) values of zircon range from +6.13 to +14.8, showing that Nd and Hf isotopic compositions are decoupled. The Nd-Hf isotopic decoupling may be related to the elevated Lu/Hf of the rocks, which could be attributed to the removal of zircon with low Lu/Hf, or the involvement of garnet with high Lu/Hf during the re-melting of the crust [69][70][71]. The characteristic of positive Zr and Hf anomalies, crust-like Lu/Hf ratios (0.06 to 0.16), do not support the removal of zircon during the re-melting of the crust. The HREE of samples much higher than the upper continental crust (UCC) and lower continental crust (LCC) (Figure 7a), combined with the high zircon saturation temperatures (TZr = 783 °C to 952 °C), showing that the breakdown of garnet may have been involved during crustal re-melting. The breakdown of garnet would likely affect the Lu-Hf rather than the Sm-Nd isotopic systematics, so the Nd isotopic compositions may be more consistent with the source rocks [71]. The oldest inherited zircon age (1465 Ma, sample ZJ-TW1) is in agreement with the two-stage Nd model ages (1305-1476 Ma), which further The petrogenesis of A-type alkaline magmas is still a matter of debate. Commonly, the A-type granites can be produced by the fractionation of mantle-derived magma, crust anatexis, or mixing of mantle-derived magma and crustal rock [58,61,62]. Fractionation of mantle-derived magma or re-melting of under-plating basaltic rocks generally produces small amounts of silicic magma [63]. Therefore, a significant quantity of mafic magma and mafic residues were required in such a model [64,65]. The metaluminous to weakly peraluminous composition of the samples are inconsistent with the peralkaline attribute of fractionation mantle-derived magma (Figure 6d) [53,57]. The lack of mafic rocks, high SiO 2 content with low Ni and Co of samples is indicative of the minor contribution from mantle-derived material in generation. Therefore, we can predict that the volcanic rocks did not directly generate by the fractionation mantle-derived magmas or mixing origin. The low Al 2 O 3 /TiO 2 ratios (36-96) and high CaO/Na 2 O ratios (0.07-1.37), as well as the negative whole-rock ε Nd (t) values (−1.8 to −4.0) indicate a lower crustal source [66]. Moyen [67] proposed that the pure crust can melt and form the granite induced by longlived heat accumulation with no or minimal basalt flux and the heat source chiefly from the radioactive and basal heat flux on the basis of heat budget calculation. In the heat balance models, more than 20 Ma is necessary to accumulate heat and induce the crust melting at a thick crust [67]. As shown in previous studies, there is little or no magmatic event from Early Silurian to Late Silurian (440 Ma to 420 Ma) in the northern margin of YB [9,18], representing a long-time relatively stable condition. Despite no or little basalt flux during the Early Devonian, the long-lived heat accumulation in a thick crust could induce a magmatic burst after more than 20 Ma [67,68]. Therefore, the Haerdaban volcanic rocks were mainly derived from the partial melting of felsic crustal rocks induced by the long-lived heat accumulation with no or minimal basalt flux.
The low Sr (41.2 to 120 ppm) and high Rb (56.2 to 372 ppm) concentrations of the samples imply that the fractional crystallization may be involved in the production of the A-type volcanics. The depletion of Ba, Sr, P, Ti, and Eu anomalies in the trace element patterns, as well as the high differentiation index also supported the fractional crystallization. The fractional crystallization of plagioclase and K-feldspar minerals can account for the depletion of Ba, Sr, and Eu. The P and Ti anomalies can be attributed to the fractionation of apatite and Fe-Ti oxides. It is noted that the ε Nd (t) values of samples range from −1.8 to −4.0, but ε Hf (t) values of zircon range from +6.13 to +14.8, showing that Nd and Hf isotopic compositions are decoupled. The Nd-Hf isotopic decoupling may be related to the elevated Lu/Hf of the rocks, which could be attributed to the removal of zircon with low Lu/Hf, or the involvement of garnet with high Lu/Hf during the re-melting of the crust [69][70][71]. The characteristic of positive Zr and Hf anomalies, crust-like Lu/Hf ratios (0.06 to 0.16), do not support the removal of zircon during the re-melting of the crust. The HREE of samples much higher than the upper continental crust (UCC) and lower continental crust (LCC) (Figure 7a), combined with the high zircon saturation temperatures (T Zr = 783 • C to 952 • C), showing that the breakdown of garnet may have been involved during crustal re-melting. The breakdown of garnet would likely affect the Lu-Hf rather than the Sm-Nd isotopic systematics, so the Nd isotopic compositions may be more consistent with the source rocks [71]. The oldest inherited zircon age (1465 Ma, sample ZJ-TW1) is in agreement with the two-stage Nd model ages (1305-1476 Ma), which further emphasizes that the breakdown of garnet during crustal re-melting results in the elevated HREE and ε Hf (t) values of the melts.
In summary, the Early Devonian volcanic rocks from Haerdaban area are considered to be generated by re-melting of lower crust induced by the long-lived heat accumulation with no or minimal basalt flux and subsequently fractional crystallization is significant.

Tectonic Setting and Implications
The Cambrian to Early Ordovicican strata distributed in the northern margin of YB reveal a passive continental margin environment [27,72]. The Middle to Late Ordovician arc-related plutons were reported in the northern margin of YB and indicated that the Junggar Ocean is likely to have begun southwards subduction at least in the Middle Ordovician [4,5,18,72]. The Middle Devonian to Carboniferous arc-related volcanic and intrusive rocks produced by the southward subduction of the Junggar Ocean combined with Late Paleozoic accretionary complexes in the northern margin of YB indicate an active continental margin [37,[73][74][75].
The Early Devonian volcanic rocks from Haerdaban area show relative depletion of Nb, Ta, and Ti, and enrichment of LILEs, which are similar to the arc magmas typically generated in subduction-related settings [76][77][78]. In the Nb versus Y and Rb versus Y + Nb tectonic discrimination diagrams, these samples plot in the volcanic arc granite (VAG) and within plate granite (WPG) fields ( Figure 9). Besides, the samples have relatively high HREE concentrations and low Ti contents, unlikely formed in the riftor plume-related settings. The Middle-Late Devonian arc-related magmatism are exposed along the northern margin of YB, which were generated in a continental arc setting, indicating the continuous southward subduction of Junggar oceanic crust (or North Tianshan oceanic crust) beneath the Northwestern Chinese Tianshan [11]. This is also supported by the sedimentary evidence. Our investigation found that the Late Paleozoic sedimentary strata in the Biezhentao mountain area, including Lower Devonian Dahalajunshan Formation, Middle Devonian Hanjiga Formation, and Upper Devonian Tuosikuertawu Formation, which are characterized by near-source deposition and contain amounts of arc-related volcanic debris, suggest an active continental margin [79]. Additionally, the Devonian Tulasu Formation in the Tulasu basin also reflected an active margin [14]. Thus, we infer that the northern margin of the YB is an active continental margin in the Early Devonian.
Minerals 2021, 11, x FOR PEER REVIEW 11 of 16 emphasizes that the breakdown of garnet during crustal re-melting results in the elevated HREE and εHf (t) values of the melts. In summary, the Early Devonian volcanic rocks from Haerdaban area are considered to be generated by re-melting of lower crust induced by the long-lived heat accumulation with no or minimal basalt flux and subsequently fractional crystallization is significant.

Tectonic Setting and Implications
The Cambrian to Early Ordovicican strata distributed in the northern margin of YB reveal a passive continental margin environment [27,72]. The Middle to Late Ordovician arc-related plutons were reported in the northern margin of YB and indicated that the Junggar Ocean is likely to have begun southwards subduction at least in the Middle Ordovician [4,5,18,72]. The Middle Devonian to Carboniferous arc-related volcanic and intrusive rocks produced by the southward subduction of the Junggar Ocean combined with Late Paleozoic accretionary complexes in the northern margin of YB indicate an active continental margin [37,[73][74][75].
The Early Devonian volcanic rocks from Haerdaban area show relative depletion of Nb, Ta, and Ti, and enrichment of LILEs, which are similar to the arc magmas typically generated in subduction-related settings [76][77][78]. In the Nb versus Y and Rb versus Y + Nb tectonic discrimination diagrams, these samples plot in the volcanic arc granite (VAG) and within plate granite (WPG) fields ( Figure 9). Besides, the samples have relatively high HREE concentrations and low Ti contents, unlikely formed in the rift-or plume-related settings. The Middle-Late Devonian arc-related magmatism are exposed along the northern margin of YB, which were generated in a continental arc setting, indicating the continuous southward subduction of Junggar oceanic crust (or North Tianshan oceanic crust) beneath the Northwestern Chinese Tianshan [11]. This is also supported by the sedimentary evidence. Our investigation found that the Late Paleozoic sedimentary strata in the Biezhentao mountain area, including Lower Devonian Dahalajunshan Formation, Middle Devonian Hanjiga Formation, and Upper Devonian Tuosikuertawu Formation, which are characterized by near-source deposition and contain amounts of arc-related volcanic debris, suggest an active continental margin [79]. Additionally, the Devonian Tulasu Formation in the Tulasu basin also reflected an active margin [14]. Thus, we infer that the northern margin of the YB is an active continental margin in the Early Devonian.  [76,77]). Syn-COLG: syn-collision granites; WPG: within plate granites; VAG: volcanic arc granites; and ORG: ocean ridge granites.
The Early Devonian granites (our unpublished data), Early Devonian volcanic rocks (this study), and previous reported magmatism [11,13,19] have similar geochemical characteristics (such as enrichment in LILEs relative to HFSEs, and depletion of Nb, Ti, and P) to that of continental arc magma, which can be related to the southward subduction of the  [76,77]). Syn-COLG: syn-collision granites; WPG: within plate granites; VAG: volcanic arc granites; and ORG: ocean ridge granites.
The Early Devonian granites (our unpublished data), Early Devonian volcanic rocks (this study), and previous reported magmatism [11,13,19] have similar geochemical characteristics (such as enrichment in LILEs relative to HFSEs, and depletion of Nb, Ti, and P) to that of continental arc magma, which can be related to the southward subduction of the Junggar Ocean. The Early-Middle Devonian magmatism suggests a relatively extensive magmatism in the northern margin of YB [11], which is the eastward counterpart of southern limb of the DVB. The DVB and BY magmatic belts were superimposed together in the northern margin of YB, which are different from the spatial-temporal distributions in the western and northern margins, implying a continuous arc-related magmatism. The ε Nd (t) and ε Hf (t) values of these Early Devonian to Permian arc magmatic rocks increase steadily, which are different from that of magmatic rocks generated during the slab retreating (Figure 10a,b) [80][81][82]. Therefore, the~E-W linear trending of arc magmatisms in the northern margin of YB combined with the geochemical and Nd-Hf isotope data in this study and previous studies indicate that they were generated in a relatively steady-stated subduction setting without significant slab retreat or advance. The Early Devonian volcanic rocks in this study formed in a continental arc setting related to the southward subduction of Junggar oceanic crust ( Figure 11).
Minerals 2021, 11, x FOR PEER REVIEW 12 of 16 magmatism in the northern margin of YB [11], which is the eastward counterpart of southern limb of the DVB. The DVB and BY magmatic belts were superimposed together in the northern margin of YB, which are different from the spatial-temporal distributions in the western and northern margins, implying a continuous arc-related magmatism. The εNd(t) and εHf(t) values of these Early Devonian to Permian arc magmatic rocks increase steadily, which are different from that of magmatic rocks generated during the slab retreating (Figure 10a,b) [80][81][82]. Therefore, the ~E-W linear trending of arc magmatisms in the northern margin of YB combined with the geochemical and Nd-Hf isotope data in this study and previous studies indicate that they were generated in a relatively steady-stated subduction setting without significant slab retreat or advance. The Early Devonian volcanic rocks in this study formed in a continental arc setting related to the southward subduction of Junggar oceanic crust ( Figure 11).   magmatism in the northern margin of YB [11], which is the eastward counterpart of southern limb of the DVB. The DVB and BY magmatic belts were superimposed together in the northern margin of YB, which are different from the spatial-temporal distributions in the western and northern margins, implying a continuous arc-related magmatism. The εNd(t) and εHf(t) values of these Early Devonian to Permian arc magmatic rocks increase steadily, which are different from that of magmatic rocks generated during the slab retreating (Figure 10a,b) [80][81][82]. Therefore, the ~E-W linear trending of arc magmatisms in the northern margin of YB combined with the geochemical and Nd-Hf isotope data in this study and previous studies indicate that they were generated in a relatively steady-stated subduction setting without significant slab retreat or advance. The Early Devonian volcanic rocks in this study formed in a continental arc setting related to the southward subduction of Junggar oceanic crust ( Figure 11).