Subducted oceanic slab break-off in a post-collisional setting: Constraints from petrogenesis of Late Carboniferous dykes in central West Junggar, Xinjiang, NW China

Abstract Numerous Late Carboniferous – Early Permian dykes are found in West Junggar and represent an important part of the Central Asian Orogenic Belt. In this contribution, we use these dykes to assess the tectonic regime and stress state in the Late Carboniferous – Early Permian. The West Junggar dykes are mainly diorite/dioritic porphyrite with minor diabase and were formed in 324–310 Ma. They have been divided into two groups based on their orientation, petrology and geochronology. Group 1 dykes mostly comprise WNW-striking dioritic porphyrite and NE-striking diorite with minor diabase and resemble the Karamay-Baogutu sanukitoid. They were probably formed from depleted mantle at a relatively high temperature and pressure with the addition of 1–2% sediment/sedimental partial melt and 0–5% trapped oceanic crust-derived melts. Group 2 dykes are ENE-striking and are similar to sanukite in the Setouchi Volcanic Belt. These dykes were also derived from depleted mantle at a shallow depth but high temperature with the addition of 2–3.5% sediment/sedimental partial melt. Magma banding and injection folds in dykes and host granitoids indicate magma flow. Paleostress analysis reveals that both groups of dykes were formed in a tensile stress field. Their emplacement is favoured by presence of pre-existing joints or fractures in the host granitoids and strata. We conclude that large-scale asthenosphere mantle upwelling induced by trapped oceanic slab-off can explain the magmatism and significant continental crustal growth of West Junggar during Late Carboniferous to Early Permian.

Intermediate-mafic dykes are widespread in the southern West Junggar (Fig. 1), especially in the Karamay, Xiaerpu, Bieluagaxi and Liushugou areas (Fig. 2), which provide an excellent opportunity to understand the geological evolution of the West Junggar.The dykes in West Junggar are referred to as high-magnesium andesites (HMAs) including sanukitoids and adakitic-like rocks (Yin et al. 2010(Yin et al. , 2013;;Duan et al. 2019), which are distinct from typical island arc andesites due to their elevated SiO 2 , MgO and Mg # (Mg# = Mg 2þ /(Mg 2þ þFe 2þ )), Cr and Ni contents (Tatsumi, 2001).Their geochemical and isotopic features reveal an affinity of both mantle and crust and an uncommon tectonic setting in association with subduction zones or post-collisional  environment (Qian & Hermann, 2010).Three major alternative models have been proposed to account for the geochemical characteristics of these West Junggar dykes: (a) partial melting of peridotites metasomatized by slab-derived melts/fluids (Yin et al. 2010(Yin et al. , 2013(Yin et al. , 2015a(Yin et al. , 2015b;;Tang et al. 2012aTang et al. , 2012b)); or (b) sedimentderived melts at a ridge subduction environment (Duan et al. 2019) or (c) partial melting of remnant oceanic crust formed in early Palaeozoic during post-collisional processes (Xu et al. 2013).Thus, detailed petrological, geochemical and geochronological studies of the dykes in this area could help to resolve which model is most appropriate.Intermediate-mafic dykes usually occur in tensional tectonic regimes, and record information on the timing and processes of continental break-up (Strvastava, 2011).They also can be used to reconstruct the paleostress field and provide information on the genesis and tectonic setting (Ma et al. 2012).
In this paper, we present zircon U-Pb ages, elemental and Sr-Nd isotopic data for the Carboniferous dykes in West Junggar (the Karamay, Xiaerpu, Bieluagaxi and Liushugou dykes) to explore their petrogenesis and tectonic implications during the Late Palaeozoic.Based on our new data, we propose that Carboniferous dykes in West Junggar are similar to HMAs and were likely formed by partial melting of peridotites metasomatised by remnant oceanic crust-derived melts or sediment-derived melts in a postcollisional setting.

Geological background
The CAOB (Fig. 1a) is located between the Siberian Craton to the north and the Tarim and North China Cratons to the south.It extends from the Urals in the west, through Uzbekistan, Tajikistan, Kazakhstan, NW China, Mongolia and NE China to the Okhotsk Sea in the Eastern Russia (Sengör et al. 1993;Jahn et al. 2000;Windley, et al. 2007).The CAOB is generally thought to be related to the closure of the Palaeo-Asian Ocean because it is a composite terrane made up of diverse fragments including ancient microcontinents, arc terranes, oceanic volcanic islands, volcanic plateaus, oceanic crust and passive continental margin sequences (Coleman, 1989;Buckman & Aitchison, 2004;Windley et al. 2007;Choulet et al. 2012).
West Junggar is located in the middle of the Kazakhstan orocline (Tang et al. 2010;Ma et al. 2012) and is surrounded by the Altai orogen to the north, the Tianshan orogen to the south, the Kazakhstan plate to the west, and the Junggar basin to the east (Fig. 1a).Numerous Cambrian to Carboniferous ophiolitic mélanges have been reported from West Junggar (Feng et al. 1989;He et al. 2007;Yang et al. 2012aYang et al. , 2012b;;Zhu et al. 2015).Ordovician-Quaternary sedimentary strata are widely distributed in this region, with Devonian-Carboniferous volcano-sedimentary rocks being particularly abundant (BGMRX, 1993).West Junggar can be divided into northern and southern parts by the Xiemisitai Fault (Xu et al. 2013;Yin et al. 2015a), and some have further divided the southern part of West Junggar into central and southern domains, which recorded different subduction-accretion history (Xu et al. 2013;Chen et al. 2015).Relatively continuous Carboniferous strata covered by the Permian molasse and continental volcanic rocks with an angular unconformity are dominant in central and southern West Junggar (Zong et al. 2016).
Abundant widespread Palaeozoic magmatism and evidence for a prolonged orogenic event are notable characteristics of the West Junggar terrane.Three magmatic pulses have been recognised (Han et al. 2006;Chen et al. 2010;Gao et al. 2014).The oldest Late Devonian -Early Silurian pulse comprises calc-alkaline plutons and volcanic rocks (Chen et al. 2010(Chen et al. , 2015;;Choulet et al. 2012).The middle pulse is composed of Early Carboniferous granitoids, which are characterised by I-type plutons and adakitic rocks and Nb-enriched basalts.The youngest pulse, mostly formed between 310 Ma and 290 Ma, is dominated by A-type granites and sanukitoids, including monzogranite, alkali granite and charnockite (Han et al. 2006;Geng et al. 2009;Chen et al. 2010;Tang et al. 2012b).
The GPS and detailed observations of dykes in central West Junggar are listed in Table S1 and Fig. 4, respectively.

4.a Zircon U-Pb geochronology
Four samples from the dykes in West Junggar were selected for U-Pb zircon dating, and the results are listed in Table 1.The descriptions of the analytical methods used are given in Appendix 1.The zircon CL images and the concordia diagrams are presented in Figs. 5 and 6, respectively.All confidence intervals are reported at 95%.

4.b Whole-rock geochemistry
Analytical methods are described in Appendix 1 and the whole-rock elemental data for the dykes in the West Junggar area are shown in Table 2 and S4.A few samples have relatively high LOI values of 3.00-5.37wt.%, however, most samples have LOI values of 1.88-2.99wt.%, (Table 2), supporting a low degree of alteration or low-grade metamorphism.The major oxides and trace elements do not display any obvious correlations with LOI values (Table S2 and Fig. S1), further suggesting that their abundances are unlikely to have been modified by the low-grade metamorphism or alteration.
Subducted oceanic slab break-off in a post-collisional setting Based on the distribution according to their orientation, petrology and geochronology of dykes in West Junggar, they can be divided into three groups (Table S1): the Early stage is represented by ENE-trending Bieluagaxi dykes, which were generated at 324 Ma and include diorite, dioritic porphyrite, with minor diabase; the Late 1 stage contains WNW-trending dykes formed between 314 Ma and 310 Ma, including dioritic porphyrites in Karamay, Xiaerpu and Liushugou areas; the last stage is referred as Late 2 stage marked by relatively younger NE-trending diorite with minor diabase dykes (<310 Ma) in Karamay and Liushugou areas.
In view of the similar geochemical characteristics of the Late 1 and 2 stage dykes (Figs.6-8, Table 2 and S4), we divide dykes in West Junggar into two groups.Karamay, Xiaerpu and Liushugou dykes (Group 1: Late 1 and 2 stage) and Bieluagaxi dykes (Group 2: Early stage).
In summary, all studied dykes resemble HMAs and are characterised by intermediate SiO 2 , high MgO, high Cr and Ni contents, marked Nb and Ta depletions, high Sr, Ba and highly fractionated REE (Martin et al. 2005;Yin et al. 2010).

5.a Age correlation between dykes and host rocks
The zircons from the dykes in West Junggar have high Th/U ratios (Table 1) with oscillatory zoning shown by the CL images (Fig. 5), suggesting magmatic origins.However, most of the zircons in the dykes show typical alternating light and dark oscillatory zoning to those in the felsic magmatic rocks.The ages of zircons from the dykes are similar to the host granitic rocks, and so it is possible that the dated zircons in dykes were captured by the Carboniferous granitic host rocks.This conclusion is supported by the geological intercalated relationship between the dykes and the host granitic rocks.Both the dykes and the host granitic rocks underwent plastic interactions, indicating that the host rocks were still semicrystalline, and even hot, when the dykes were emplaced (Ma et al. 2012).Therefore, based on the weighted average 206 Pb/ 238 U age of zircons (Karamay dykes, 311 Ma; Xiaerpu dykes, 310 Ma; Bieluagaxi dykes, 324 Ma; Liushugou dykes, 314 Ma), and the field relationship between the dykes and the host granitic rocks, the dykes were formed almost simultaneously or slightly later than the host granitic rocks in the Late Carboniferous.
All studied dykes intrude into their corresponding host plutons except for the Liushugou dykes, which intrude into Devonian to Carboniferous strata (Fig. 2d).The Late Carboniferous -Early Permian host rocks (e.g., Karamay, Xiaerpu, Hongshan and Bieluagaxi granitoid complexes) in West Junggar are well-dated (Table S3) and consist of syenogranite þ alkali-feldspar granite ± monzogranite ± granodiorite ± diorite.These granitoids have emplacement ages between 276 Ma and 319 Ma (Table S3).Based on reported data (Chen & Jahn, 2004;Han et al. 2006; Geng et al,  The dykes mostly extend linearly (Fig. 3a, c, e) but some of them are curved or locally deformed (Fig. 3b, d, h).An extensional stress regime is evident due to the irregular jagged contact boundary between dykes and country rocks during the intrusion of these dykes into the pre-existing fractures (Fig. 3b, d, f, h).Thus, based on field observations and petrographic analysis along with geochronological data, we propose that all the dykes were formed after the intrusion of host granitoids, whereas the interval between them is short.When the dykes intruded into the host granitoids, the latter has not been completely cooled and consolidated.This is supported by the proposition that the emplacement interval between the Karamay granitoid and the dykes may have been about one million years (Ma et al. 2012).
Using data from He (2012) and Chen (2015), along with our new data, Group 2 dykes have higher MgO contents (4.42-8.41,mean value of 7.09, Table 1 and S4) than those of Group 1 (2.56-6.13,mean value of 4.61, Table 1 and S4), and majority of both groups plot in the high-Mg andesite area with a few samples in the Karamay -Baogutu sanukitoids area (Fig. 9b) except one sample from Karamay area (H8552-1, Table S4).Compared with

5.b.1 Crustal contamination and fractional crystallisation
Despite some compositional differences observed among various dykes in both groups of samples, they share relatively similar low initial ( 87 Sr/ 86 Sr) i ratios and high positive εNd(t) values (Table 3; Fig. 11a).The Nb/La and ( 87 Sr/ 86 Sr) i ratios are essentially constant with increasing Mg # and SiO 2 (Fig. 12a and b), which rules out significant crustal contamination during magma ascent and emplacement.This interpretation is also consistent with the lack of significant correlation between Nb/La and Sm/Nd (Figure not shown).
The negative correlations of TiO 2 , FeO T , MgO and CaO with SiO 2 , for the two groups of dykes (Fig. 9a, b, d and e), indicate fractional crystallisation of olivine and pyroxene.There is no obvious correlation between Al 2 O 3 , total alkalis and silica (Fig. 9f).The trace element variation in these dykes, when combined with data from other reported dykes in West Junggar, also suggests that fractional crystallisation of olivine and orthopyroxene has taken place (Fig. 12d), which is consistent with the presence of hornblende þ biotite as the dominant mafic minerals/phenocrysts in the rocks (Fig. 4a, c, e, f and h).The Cr and Ni abundances of Group 1 dykes can be quantitatively explained by the 0-18% fractionation of 95% orthopyroxene and 5% olivine (Fig. 12d) starting from the assumed parent magma composition represented by the relatively least fractionated sample H8747-5 from Group 1 samples.The quantitative simulation of the Cr and Ni abundances from Group 2 samples could be derived from 0 to 10% fractionation of 75% orthopyroxene and 25% olivine starting from the assumed parent magma composition represented by the least fractionated sample BLG1505h (Fig. 12d).The noticeable positive Sr anomalies and negligible negative Eu anomalies demonstrate that plagioclase was not a major fractionating phase in either group (Fig. 10).The positive linear correlation in the (La/Yb) N ratios vs. La diagram (Fig. 12e) for the Liushugou and Xiaerpu dykes suggests that the 0-0.02% fractional crystallisation of apatite could explain the composition of Group 1 samples (Fig. 12e, f).Modelling suggests that the Karamay samples in Group 1 experienced 0-0.02% fractional crystallisation of zircon (Fig. 12e, and g).Group 2 samples experienced fractional crystallisation of 0-0.05% monazite starting from the assumed parent magma composition represented by the least fractionated sample BLG1505h.

5.b.2 Petrogenesis of Group 1 dykes
The Group 1 dykes in West Junggar also have high Mg # values, Ni and Cr concentrations, indicating that they are likely to be melts derived from the mantle (Rapp & Watson, 1995).They have relatively low ( 87 Sr/ 86 Sr) i (0.7035-0.7044) and slightly high εNd(t) values (þ4.5-þ7.8),which are similar to the granitoids and dykes, which are derived from partial melting of the basement (i.e., subducted and trapped juvenile oceanic crust, Xu et al. 2013;Gao et al. 2014;Zhang et al. 2018;Zheng et al. 2020) in West Junggar (Fig. 11a; Table S5).The Group 1 dykes have high Th/La and low Nb/Th ratios that plot in the compositional spectrum of forearc basalts, indicating an oceanic slab-related origin (Fig. 14e) and this is also consistent with high Th/Yb ratios relative to Mid Ocean Ridge Basalt (MORB) (Fig. 14f).All of these features indicate an oceanic crust source and their young Nd model ages (T DM ) prove that their source is juvenile.
High magnesium magmas can be generated by the partial melting of hydrous peridotite mantles (Hirose, 1997), particularly when hydrous felsic liquids from the subducting oceanic slab or sediment-derived melts trigger partial melting (Mukasa et al. 2007).In a subduction zone, Sr and Ba are mobile in aqueous fluids, while LREE and Th are preferentially transferred from the slab when sediment-derived melts are involved.Therefore, ratios such as Ba/La, Th/La, Th/Yb and Sr/Yb can be used to assess the influence of aqueous fluids and sediment-derived melts (Johnson & Plank, 1999).Apart from Liushugou dykes with variable Ba/La ratios suggesting that their generation was related to the oceanicslab-derived fluid, Group 1 dykes have high Th/Yb but low Ba/La ratios (e.g., Fig. 14a), indicating that sediment-derived melt plays a dominant role in the formation of most dykes in West Junggar.However, the magmatic zircons in both groups of dykes have moderate U/Yb ratios and Y contents (Fig. 12c), which are consistent with both continental zircon and ocean crust zircon (Grimes et al. 2007).In Fig. 14b, c and d, the samples plot on a trend between GLOSS or sediment and MORB, indicating interaction between peridotite mantle and sediment-derived melt.
The retention of zircons in the sedimentary residue would result in significant depletion of Zr and Hf in the resultant magmas (Woodland et al. 2018).Both groups of samples do not show negative Zr-Hf anomalies on primitive mantle-normalised diagrams (Fig. 10b, d, f and h), indicating no or few residual zircons and a high degree of sediments melting.Furthermore, the presence of broad or weak zoning in the CL images of both group dykes indicates that the zircons in these rocks crystallised from high-temperature melts (Fig. 5; Wu & Zheng., 2004), which is also confirmed by the crystallisation temperatures (Table S6) of the magmatic zircons (estimated by using the Ti-in-zircon thermometry, Ferry & Watson, 2007).Karamay, Xiaerpu and Liushugou dykes (Group 1) have crystallisation temperatures of 789 °C-940 °C, 804 °C-1006 °C and 772 °C-1112 °C.Yin et al. (2010Yin et al. ( , 2012Yin et al. ( ,2012Yin et al. ( , 2013Yin et al. ( , 2015aYin et al. ( , 2015b)), Tang et al. (2010Tang et al. ( , 2012aTang et al. ( , 2012b)), Ma et al. (2012), Xu et al. (2013), He et al. (2015), Li et al. (2015), Zhan et al. (2015), Duan et al. (2019) and Ma et al. (2020).The Sanukitoid data of the Karamay-Baogutu area are from Tang et al. (2012a), Ma et al. (2012) and Yin et al. (2010Yin et al. ( , 2013)).In (d) and (e), the number represents the fractionated number.In (f)-(h), F is fractionated number.Partition coefficients are from Rollinson (1993), Green and Pearson (1986), Mahood and Hildreth (1983) and Yurimoto et al. (1990).Abbreviations: Ol = olivine, Opx = orthopyroxene, Cpx = clinopyroxene, Hb = hornblende, Ap = apatite, Mon = monazite, Allan = allanite.Several models have been proposed to explain the relatively high Sr/Y ratios and Mg # values of the HMAs, including (1) magma mixing with subsequent contamination (Guo et al. 2007); (2) fractional crystallisation of mantle-derived basaltic magma (Li et al. 2009(Li et al. , 2013)); and (3) melt-mantle interaction (Chung et al. 2003;Gao et al. 2004;Martin et al. 2005).Group 1 dykes have relatively similar low initial ( 87 Sr/ 86 Sr) i ratios and high positive εNd(t) values (Table 3; Fig. 11a), which precludes the possibility of crustal contamination.Geochemical modelling of batch partial melting, fractional crystallisation and magma mixing can be used to investigate the processes that exert an influence on the traceelement variations of cogenetic magmas (Schiano et al. 2010).Modelling of partial melting and fractional crystallisation would produce a straight line with a slope and a horizontal line, respectively, in plots of two incompatible elements with different partition coefficients (e.g., Th/Nd vs. Th, La/Sm vs. La; Fig. 9g and  h) (Treuil and Joron, 1975;Schiano et al. 2010).These plots of the Group 1 dykes show that they could have been generated by partial melting rather than fractional crystallisation or magma mixing (Fig. 9g and h).Hence, the model ( 1) and ( 2) can be ruled out.This conclusion is also supported by geological and geochemical observations, including (1) the general absence of mafic magmatism in central West Junggar during the Late Carboniferous -Early Permian.Therefore, fractional crystallisation cannot account for the widespread coeval intermediate to felsic magmas; (2) the lack of disequilibrium texture of phenocrysts and the presence of compositional ortho-zonal structure in plagioclase phenocrysts in the Group 1 dykes suggest that there has not been extensive magma mixing; (3) The Mg # of these high-Mg rocks are comparable to those of crust-derived melts that underwent assimilation with peridotite in the mantle (Fig. 13b; Rapp et al. 1999).( 4) The dykes coexist with Nb-enriched-like rocks in central West Junggar, which are generally considered to be produced by the melting of mantle metasomatized by slab-derived melts (Yin et al. 2013).
The presence of residual accessory phases profoundly controls the trace-element composition of the adakitic melts.Therefore several characteristic trace-element concentrations and ratios are commonly regarded as proxies for residual accessory phases in magma sources and so can be used to evaluate the pressure and temperature conditions of melting.The Sr/Y and (La/Yb) N ratios of intermediate-felsic magmas are essentially controlled by the mineral phases (e.g., garnet, rutile, amphibole and plagioclase) in the residue or by fractionation crystallisation of parent magma.Group 1 dykes have relatively steep HREE patterns ((Gd/Yb) N = 1.24-2.26mean value of 1.66) with high Sr/Y (31.68-82.85,mean value of 56.26) ratios, which are similar to the Karamay-Baogutu sanukitoids ((Gd/Yb) N = 1.11-2.56;Sr/Y = 26.18-129.53mean value of 52) (Table S4), which were derived from a source containing some residual garnet and rutile (Yin et al. 2010(Yin et al. , 2013;;Ma et al. 2012;Tang et al. 2012a).However, Group 1 dykes show Nb-Ta negative anomalies but without distinct Ti depletion on a primitive mantle-normalised multi-element plot (Fig. 10b, d and h), suggesting that rutile is absent in the magma source.Ratios such as K/Yb and Dy/Yb can be used to distinguish between melting in the spinel and garnet stability field of phlogopite-and/or amphibolebearing lherzolite (Duggen et al. 2005) because melting in the spinel and garnet stability field would produce melts with low Dy/Yb ratios (<1.5) and high Dy/Yb ratios (>2.5), respectively.Group 1 samples have Dy/Yb ratios of 1.60-2.24,indicating that they were probably derived from melting of a source in the spinel-garnet transition zone.As plagioclase is not a major fractionation phase and garnet is a residual in the source with fractionation of orthopyroxene and olive (Fig. 12d), Group 1 dykes have relatively high Sr/Y ratios.However, compared with typical adakitic rocks, the Group 1 dykes have low (La/Yb) N ratios, which was most likely caused by partial melting of peridotite mantle metasomatized by oceanic slab-derived adakitic melts (Ma et al. 2012;Yin et al. 2013).This indicates that the meltmantle interaction may have diluted their adakitic features (Fan et al. 2020).Based on a minimum pressure (1.5GPa) for rutile appearance, the depth for Group 1 sample production via melting of subducted oceanic crust cannot exceed 50 km (Xiong et al. 2005;Xiong, 2006).Using MORB as a source rock, phase-equilibrium modelling indicates that adakites are generated under P-T conditions of 1.0-1.6GPa and 800-1000 °C (Wang et al, 2020).Thus, it is clear that Group 1 dykes were formed at relatively high-temperatures (800°C-1050°C) and relatively high-pressure conditions (<1.5 Gpa) (Defant & Drummond, 1990;Xiong, 2006;Yin et al. 2013).
In relation to model (1), Qian and Hermann (2010) proposed that intermediate-felsic magmas were formed in an extensional setting by partial melting of lower crust followed by reaction of this melt with peridotite at middle or upper crustal levels.Petrographic observations and geochemical compositions of Group 2 samples (Figs. 4 and 9) differ from those HMAs generated by model 1, which usually comprise a certain amount of olivine xenocrysts or dunite xenoliths and display adakitic-like compositions with high Sr, Ba but strong Nb-Ta depletions (Qian & Hermann, 2010).Furthermore, there is no evidence in the area for the emplacement of peridotite at middle or upper crustal levels, and thus this model does not represent a feasible mechanism for the generation of the West Junggar dykes.
We propose that Group 2 dykes were derived from remelting of mantle-wedge peridotite modified by interaction with subducting sediments rather than aqueous fluids based on following evidence: (1) They possess high MgO, Mg # , Ni and Cr contents than typical island-arc andesites (Tatsumi, 2006).( 2) On the MgO vs. SiO 2 diagram (Fig. 8a), Group 2 samples show similarities to sanukitic HMAs in the Setouchi volcanic belt (Tatsumi, 2001;Yin et al. 2015), not only reflecting the differences in initial compositions compared with Group 1 samples but also suggesting a similar formation mechanism to Setouchi sanukitoids, which were generated by interaction between the peridotite mantle and sediment-derived melt in a young, or hot, subduction zone setting (Furukawa & Tatsumi, 1999).(3) The concentration of Th in marine sediments is more than two orders of magnitude higher than that of the mantle, and therefore the addition of sediment to the mantle will lead to increasing Th/La, Th/Sm and Th/Y ratios and decreasing (Nb/Th) ratios but will have little effect on Ta/Yb ratios (Plank & Langmuir, 1998;Liu et al. 2014).Group 2 have high Th/Yb and low Ba/La ratios, suggesting that they were generated by partial melting of the peridotite mantle metasomatized by sediment-derived melts, which is supported by the fact that they plot along a mixing line between the mantle and sediments on a Th/La vs. Sm/La, Th/La vs. Th and Th/Yb vs. Th/Sm and (Th/La) N-MORB vs. (Nb/Th) N-MORB diagrams, further suggesting significant interaction between subducting sediments and peridotite mantle (Fig. 14b, c and d).This depleted source had possibly undergone metasomatic enrichment by sediment-derived melt, resulting in an enriched-LILE and depleted-HFSE signature (Fig. 10e, f).(4) Fig. 15 indicates that addition of 2%-3.5% sediment/sediment partial melt to depleted mantle can model the composition of sanukitic Group 2 dykes in central West Junggar (Johnson & Plank, 1999;Zhang et al. 2021b).
Group 2 samples have Dy/Yb ratios of 1.61-2.08,indicating that like Group 1 dykes they were probably derived from melting of a source in the spinel-garnet transition zone.In Fig. 9d, Group 2 samples show no obvious Ti depletion indicating that rutile is not the residual phase.They have relatively flat HREE patterns ((Gd/ Yb) N = 1.19 -1.50 with mean value of 1.47 (Fig. 10e), low Sr/Y ratios of 18.67 -33.9 (mean value of 24.07) and relatively low (La/ Yb) N ratios, suggesting that there was little residual garnet in their source region.Moderate Y/Yb and (Ho/Yb) N ratios indicate that the residual phase in the source region of Group 2 samples is mainly amphibole rather than garnet (Wu et al. 2002).They display noticeable positive Sr anomalies and negligible negative Eu anomalies demonstrating that plagioclase was a residue in the source region (Fig. 10e, f).The phase-equilibrium between the melts and the plagioclase and amphibole during the partial melting of the basalts constrained by experimental petrology indicates that they were formed at pressures lower than 0.8 GPa (Ge et al. 2002;Wu et al. 2002).The Bieluagaxi dykes (Group 2) give Ti-in-zircon crystallisation temperatures between 783°C and 1013°C (Ferry & Watson, 2007), which is similar to the crystallisation temperatures of the Group 1 samples.Thus, Group 2 dykes were most likely derived from partial melting of peridotite mantle metasomatized by sediment-derived melts at a shallow depth but relatively high temperatures.
Various tectonic settings of West Junggar in Late Carboniferous have been proposed from post-collisional (Chen & Arakawa, 2005;Han et al. 2006;Chen et al. 2010;Xu et al. 2013;Gao et al. 2014;Zheng et al. 2020), to subduction-accretionary (Windley et al. 2007;Xiao et al. 2008) to ridge subduction settings (Geng et al. 2009;Tang et al. 2010Tang et al. , 2012a;;Ma et al. 2012;Yin et al. 2010Yin et al. , 2013Yin et al. , 2015a)).Arc magmatism ceased at the beginning of Late Carboniferous and the wide spatial distribution of granitoids formed in West Junggar during Late Carboniferous -Early Permian does not fit with the narrow and linear distribution pattern of magmatism that would be expected if slab break-off had occurred (Whalen et al. 2006;Li et al. 2015).Both group dykes are characterised by enrichment in LILEs and depletion in HFSEs and significant negative Nb anomaly showed by the two groups of dykes (Fig. 9) reveals that they may have been formed in a subduction-related setting (Rudnick & Gao, 2003) (Fig. 10).However, no record of oceanic crust and accretion-collision processes or (ultra-)high-pressure metamorphism younger than 300 Ma have been reported, indicating that the subduction-related setting is not a good choice (Liu et al. 2019).Thus, based on following pieces of evidence, we propose a post-collisional setting in West Junggar since Late Carboniferous: (1) Choulet et al. (2012) argued that accretion and subduction in the West Junggar must have ceased during the Late Carboniferous with transition from an accretionary to a postaccretionary setting between 305 Ma -300 Ma, which is consistent with the lack of evidence for Permian subduction in the sedimentary or tectonic records (Buckman & Aitchison, 2004;Choulet et al. 2012).( 2) Early Carboniferous magmatism has been ascribed to this subduction-accretion event, which then developed into an intra-continental orogenic event both in the northern and southern West Junggar region during the late Early Carboniferous to the early Late Carboniferous (Duan et al. 2019).( 3) Late Carboniferous collisional events between different tectonic blocks (e.g., Yili-Chinese central Tianshan block and Junggar terrane along the Chinese north Tianshan suture zone) (Zhang et al. 2016;Han & Zhao, 2018) in northern Xinjiang, led to the closure of the West Junggar remnant ocean basin and other sub-oceans of the Palaeo-Asian Ocean (Han & Zhao, 2018).The final closure of the Junggar Ocean likely occurred at the end of the Late Carboniferous in response to regional amalgamation events in the southwestern CAOB, representing the final assembly of the Kazakhstan Orocline (Zhang et al. 2021a).( 4) Undeformed maficintermediate dykes dissect the deformation structures including folds and faults, which were correlated with subduction-accretionary process in Early Carboniferous, present that the cessation of subduction-accretionary process when the abundances of dykes were emplaced (Zhan et al. 2015).Because these dykes could be deemed as 'stitching rocks', they were formed after the deformation caused by subduction-accretionary process in Early Carboniferous.( 5) Stress measurements on the studied dykes (172 in total) were obtained during field work and calculated by method from Delvaux and Spencer (2003).All dykes were formed in an extensional stress regime including pure, oblique and strike-slip extensional regimes (Table S8).This is consistent with the conclusion that the maficintermediate dykes in West Junggar formed in within-plate extensional setting, coeval with the widespread lithospheric extensional events in the Northern Xinjiang (Coleman, 1989;Jahn et al. 2000;Han et al. 2006;Zhan et al. 2015).( 6) Widespread deposition of Permian continental molasse deposits followed the formation of post-collisional magmatic rocks (Zong et al. 2016).The Gzhelian Molaoba Formation consisting of non-marine fluvial sediments represents the central West Junggar back-arc basin that finally closed by the Late Carboniferous (Liu et al. 2017).( 7) Abundances of Late Carboniferous -Early Permian undeformed A-type granitic intrusions in West Junggar (emplacement ages between 305 Ma and 280 Ma) have highlydepleted Hf-Nd isotopic affinities.They have been proposed to be derived from depleted mantle or juvenile basaltic lower crust that was generated in an extensional post-collisional setting (e.g., Xu et al. 2013;Gao et al. 2014;Zheng et al. 2020).Minor contemporaneous I-type granites were derived from the differentiated products of the basaltic melts or a mixture between basaltic melts and juvenile lower crust (Chen et al. 2010).The Late Carboniferous -Permian granitoids in West Junggar (Table S7) are plotted in the VAG and Post-COLG settings, which show transitional affinities (Fig. 16).They were probably formed in a transitional background from subductional to postcollisional settings.Additionally, the dykes were generated a little later than the host granitoids, indicating that they were probably formed in a post-collisional setting.

5.d.1 Magmatic flow of dykes
Syn-tectonic magmatic flow is defined as deformation by melt displacement with consequent rigid-body crystal rotation, but without sufficient interference between crystals to cause plastic deformation (Paterson et al. 1989).Although flow structures in dykes and plutons are generally seldom observed at outcrop or under the microscope, they include alignment of bubbles or phenocrysts in dykes, orientation and strain of enclaves, hot striations on the contact, magmatic foliation or lineation in plutons, rarely ropy flow structure and other structures (Baer, 1987;Tian & Shan, 2011;Zhou & Wang, 2012).All above are useful to determine the flow direction of magma, thus helping understand the processes responsible for magma mixing or segregation during flow, nucleation and propagation of dykes.
Flow structures observed in Bieluagaxi dykes rather than other dykes in West Junggar are mainly magmatic bands and folds.These structures were produced in the magma body as a result of flow-induced shearing within the dykes.In outcrop, the magmatic bands appear dark grey or light white in colour, with a variable amount of euhedral or subhedral hornblende and plagioclase  (Whalen et al. 1987).ORG, ocean-ridge granites; post-COLG, post-collision granites; syn-COLG, syn-collision granites; VAG, volcanic arc granites; WPG, within-plate granites.All the data of Late Carboniferous to Early Permian granitoids are listed in Table S7 from Gao et al. (2014) and references therein.
and/or quartz crystals of several millimetres in diameter.They are generally parallel or subparallel to the boundaries of dykes (Fig. 3i and j) but are not always parallel to each other, and they sometimes may make up rootless isocline folds (Fig. 3j).The contact between the bands and their ambient rocks is straight, slightly undulating or occasionally zigzag (Fig. 3i and j).As distinguishing the folding mechanisms is difficult or even impossible due to the insufficient observations at most outcrops, we just describe two interesting types: injection folds and disharmonic folds.The former appears at the blunt end of the dyke (Fig. 3j).Generally, they are isoclinal with the axial plane oriented parallel or subparallel to the sides of dykes.However, in the Fig. 3j, this injection fold occurs in local part of the dyke.The disharmonic fold has a variable size and thickness with respect to the limbs (Fig. 3h).Neither foliation nor cleavage is visible in the cores of this tight fold, suggesting the disturbance of flow-induced shearing by continuous laminar flow after magmatic bands formed in dykes.
In this study, the magmatic banding is generally considered a product of magma mingling due to the fact that most of the bands are parallel or subparallel to the contact.The megacryst K-feldspar has poikilotopic texture that euhedral amphiboles and acicular apatites embedded in the plagioclase (Fig. 4g), which is also supported by the plagioclase xenocrysts (Fig. 3g), dioritic xenolith (Fig. 3h) and round-to-elongated-shaped granitic enclaves (Fig. 3k) in dykes, indicating that the magma of the dioritic dyke was initially quenched into the unconsolidated host granitic magma and mingled with it to a certain extent.

5.d.2 Emplacement of dykes
The strike and dip orientation analysis of the dyke swarm confirms the prevalence of sub-vertical and sub-horizontal WNW-trending dykes (Late 1 stage) with minor NE-(Late 2 stage) and ENEtrending (Early stage) dykes throughout the entire study area (Fig. 2, S2; Table S1).The WNW-and NE-trending sub-vertical and sub-horizontal dyke orientation and ENE-trending (Early stage) sub-vertical dyke orientation appear to coincide with one of the two main fracture sets (NE-SW and WNW-ENE) in the host rocks (Fig. 3a, c, S3c, e, i and j) in Karamay, Xiaerpu and Bieluagaxi areas, and with localised NE-ENE-trending fractures or faults in Liushugou area (Fig. 3e, S3k), respectively, a fact that is compatible with the possibilities that most dykes could have exploited preexisting joints/faults or that dykes and dyke/parallel joints/faults are contemporaneous.
Variations in strike and/or dip have also been observed along individual dykes, revealing a continuous zigzagging stepped geometry (Fig. 3a, S3c, e, g-i) and are probably equally related to the exploitation of joints (Martinez-Poza & Druguet, 2016).The joints/faults within the host granitic rocks and surrounding Upper Carboniferous strata were generated in an extensional regime in Late Carboniferous.The joints/faults show tensile characteristics in the field (Fig. S3b, f, g, h and j).Thus, the tensile joints/fractures of the host granitic rocks and surrounding strata were formed in that extensional setting in Late Carboniferous, immediately or in a very short time after being filled by magma propagation.As discussed in section 5.d.1, the magmas of dioritic dyke were initially quenched into the unconsolidated host granitic magma, indicating that the emplacement interval between the dykes and host granites may have been very short, which is also indicated by their similar ages.In other words, these pre-existing joints/fractures were formed after the emplacement of granitoids in a tensional regime in Late Carboniferous, and the time interval between these preexisting joints/fractures and dykes may be probably only a few million years.Thus, these joints/fractures could be deemed as the 'contemporaneous' products with dykes in central West Junggar.A plausible model is that 'contemporaneous' joints in the granitoids were exploited during dyke emplacement and subsequently reactivated during cooling and post-cooling events, implying active tectonics during and after the intrusion of the dyke swarm (Martinez-Poza & Druguet, 2016).
The dykes are parallel and they were formed in response to the regional tensile stress field (Hou, 2012).Our paleostress analysis gives an estimation of the orientation and relative magnitude of the tectonic principal stress axes operating at the time of intrusion (Fig. S2).To Karamay and Bieluagaxi dykes, the σ1 was subvertical, and σ2 and σ3 sub-horizontal, whereas the σ2 of the dykes in Xiaerpu and Liushugou areas were sub-vertical with σ1 and σ3 sub-horizontal (Table S8; Fig. S2).There is a good correspondence between these paleostress values and field structures observed between dyke patterns in horizontal and vertical sections.Emplacement of the dykes took place in the form of extension fractures during a regional extensional tectonic regime with strikeslip, as inferred from the sub-vertical and sub-horizontal principal stress axes (σ1).Dykes preferentially opened perpendicular to their walls.The estimated average extension (T) accommodated by dyke intrusion is considerable as 2.7% (Li et al. 2005).Numerous dyke swarms, found in West Junggar, show similar geochemical and isotopic affinities of a post-collisional setting in Late Carboniferous.It is therefore likely that there is a magma chamber of relevant components in the deep part of the dyke.
Thus, the dyke intrusion extensional seems to have been favoured by the presence of 'contemporaneous' joints or fractures/ faults in the host granitic rocks or strata.The final pattern of dykes likely resulted from the interaction between these fractures and the external tectonic stress tensor active during dyke emplacement.

5.e Tectonic implications
In West Junggar, a Late Carboniferous magmatic association of various volcanic and intrusive rocks is extensively developed and is mostly derived from a mantle magmatic source (Chen & Jahn, 2004;Geng et al. 2009;Chen et al. 2010;Yin et al. 2010;Zheng et al. 2020).Late Carboniferous (306 Ma -315 Ma) N-MORB to OIBlike tholeiitic and alkaline basalts in Yeyagou, Maliya and Hatu were discovered and interpreted as derived from various enriched mantle sources (Zhang et al, 2011a(Zhang et al, , 2011b) ) or from a mixed mantle source consisting of subducted depleted oceanic lithosphere and enriched upwelling asthenospheric mantle associated with ridgetrench interaction (Tang et al. 2012c).However, recent work has shown that the Junggar basin is composed of (1) an assemblage of Late Palaeozoic arc fragments, accretionary complexes and trapped oceanic crust (Li et al. 2015), or (2) oceanic crust (Xu et al. 2013) consistent with the idea that the basement of the West Junggar region is mainly composed of slices of oceanic crust (Zheng et al. 2007;Zhang et al. 2018) with minor Precambrian basement (Hu et al. 2000).Existing geophysical data show that the Junggar Basin is characterised by high velocity, density and magnetisation (Zhao et al. 2008).A preserved Palaeo-oceanic lithosphere, which is becoming a piece of Eurasian continent beneath the Junggar region, has been presented by seismic evidence (Wu et al. 2023), which suggests that the basement of the Junggar Basin is oceanic crust.
Post-collisional magmatism was triggered by asthenosphere upwelling resulting in the widespread formation of Late Carboniferous and Permian granitoids and dykes in West Junggar and adjacent regions, e.g., Altay, Tianshan and North Tianshan (Han et al. 2006;Zheng et al. 2020).This interpretation is supported by Sr-Nd isotopic signatures of the magmatic rocks in West Junggar, which show widespread development of juvenile crust with minimal contribution of Precambrian rocks that preclude the occurrence of TTG (Zheng et al. 2020).The generation of the HMAs needs high-temperature conditions to trigger partial melting of the subducted slab and sediments (Tatsumi, 2008).Considering the occurrence of high Sr/Y type HMAs and sanukitic dykes with those tholeiitic and alkaline basalts in West Junggar, they are not only the indicator of an extensional setting but also a hightemperature geothermal gradient.Unlike Archean HMAs, commonly considered to be produced at a very high geothermal gradient (Martin et al. 2005), the Phanerozoic HMAs such as the Late Carboniferous sanukitic dykes in West Junggar require special dynamic processes to elevate the geothermal gradient of the post-Archean upper mantle (Shimoda et al. 1998).
There are two possible tectonic scenarios for raising the geothermal gradient, i.e., lithospheric delamination or subducted slab break-off (Liu et al. 2017).Both processes may result in asthenosphere upwelling and coeval basaltic underplating, leading to the partial melting of oceanic crust.Delamination may produce non-linear diffuse distribution of magmatism in a large area, whereas the magmatism caused by slab break-off will be confined to a narrow linear zone with significant crustal uplift (Whalen et al. 2006;Liu et al. 2017).The spatial-temporal pattern of Carboniferous magmatism in West Junggar shows a tendency to southward-younging in West Junggar (Zhang et al. 2021a), which makes slab break-off model more feasible.This suggestion is also consistent with the uplift of the Junggar terrane remnant oceanic crust during its final development between 335 Ma and 333 Ma (Zhu et al. 2015).The Late Carboniferous adakitic rocks and HMA-like rocks from the Karamay and Baogutu region could also be produced by partial melting of the broken-off oceanic crust during the initial phase of orogenic collapse (Ren, 2013).
In the Early Carboniferous, the back-arc basin, which resulted from northwest-ward intra-oceanic subduction of the Junggar Ocean in the Zhongguai area between the Junggar complex basement and the Yili block (Zhang et al. 2018), started to subduct with generation of arc-related I-type granites.Numerous volcanoclastic and tuffaceous materials from Carboniferous arc-related rocks were deposited quickly in this back-arc basin (Zhang et al. 2021a).During subduction, the oceanic slab underwent break-off resulting in asthenosphere upwelling and the tectonic regime changed from compression to an extensional post-collisional setting in the early Late Carboniferous.The trapped oceanic-crustlike basement of West Junggar was heated by the asthenospheric mantle and began to melt.Thus, Bieluagaxi (324 Ma) dykes were derived from the reaction between the peridotite mantle and the sediment-derived melts at a shallow depth but low temperature.
Magmatic activity mostly occurred between ca.332 Ma and 287 Ma based on the systematic combining and summarising of available zircon U-Pb chronological data of granitoids in West Junggar (Duan et al. 2019).During this period, the magmatism was successive and showed two important phases of magmatic activities with peak ages of ca.314 Ma and ca. 302 Ma.During this period, many magmatic rocks including A-type granites, adakitic-like and sanukitic-like rocks with minor tholeiitic, alkaline and Nb-enriched-like rocks were produced during the mantlecrust interaction process in the post-collisional regime.In Late Carboniferous, the slab-derived melt with adakitic affinity interacted with peridotite mantle and sediment-derived melts resulting in the formation of Group 1 dykes along the joints and fractures to transport and propagate.The Carboniferous strata were then folded during the tectonic extrusion of the Daerbute and Karamay ophiolitic mélanges because the central West Junggar terrane, which was situated in the core area of the Kazakhstan Orocline, underwent tectonic uplift and extensive deformation (Zhang et al. 2018).

5.f Phanerozoic Continental crustal growth
The Canadian -Alaskan Cordillera of North America and the CAOB are two large Phanerozoic orogenic belts that contain significant accounts of juvenile materials (Patchett & Samson, 2003).The CAOB was built by the subduction and closure of the Palaeo-Asian Ocean and is surrounded by the Siberian craton to the north, and the Tarim and North China cratons to the south and the European craton to the west (Zheng et al. 2020).The West Junggar orogenic belt is correlated with arc-related magmatism and accretion of oceanic rocks from latest Ediacaran to Early Carboniferous and was associated with asthenosphere upwellingrelated magmatism in post-collisional setting during Late Carboniferous to Permian (Gao et al. 2014;Zheng et al. 2020).
In West Junggar, widespread Late Carboniferous -Early Permian magmatism is mostly characterised by primitive εHf(t) and εNd(t) values and Palaeozoic model ages, which indicate massive continental crust growth by recycling of oceanic crust or direct differentiation of basaltic underplated material that originated from melting of a metasomatized lithospheric mantle (Jahn et al. 2000;Han et al. 2006;Chen et al. 2015).It has been estimated that nearly 50% of the continental crust in the CAOB is juvenile and the granitoids (Table S7) with high positive εNd(t) values contain more than 50% mantle-derived or juvenile materials (Sengör et al. 1993;Jahn et al. 2000).The estimate of the proportion of depleted mantle in the mixture can be semiquantified by using a simple two-end member mixing model (DePaolo et al. 1991).The majority of granitoids, dykes and enclaves in West Junggar plot between the crust and depleted mantle, and rocks with similar ages show approximately vertical trends, suggesting variable mixing of the crust and the depleted mantle (Fig. 11a and b) with large proportions of mantle components (Fig. 11c and d).The mantle components in Late Carboniferous igneous rocks including granitoids, dykes and enclaves in central West Junggar vary from 41.2% to 99.7% with average value of 84.6% (Fig. 9c and d).These post-collisional igneous rocks including studied samples with Sr-Nd isotopes characterised by young T DM and high positive εNd(t) values suggest considerable continental crustal growth in a postcollisional setting during Late Carboniferous to Early Permian.

Conclusions
(1) LA-ICP-MS zircon U-Pb dating results reveal that the dykes in central West Junggar were formed between 309.7±2.4Ma and 324.3±3.6 Ma and show coeval relationships with their corresponding host rocks.(2) The dykes are divided into two groups based on Sr/Y ratios: Group 1 including Karamay, Xiaerpu and Liushugou dykes with high Sr/Y ratios of 41-83, which are similar to the adakitic HMA; whereas Group 2 includes Bieluagaxi dykes with low Sr/Y ratios of 19-34, which resemble the sanukitic HMA.

Figure 1 .
Figure 1.(Colour online) (a) Simplified tectonic map of the Central Asian Orogenic Belt (Sengör et al. 1993).(b) Simplified geological map of the West Junggar (modified after Feng et al. 1989), inset shows the position of Xinjiang in NW China.

Figure 2 .
Figure 2. (Colour online) (a) Geological map of the study area (modified after Ma et al. 2012).Positions of Fig. 2b-d are marked.(b) Simplified geological map of the Karamay and Xiaerpu dykes (modified after Duan et al. 2019).(c) Simplified geological map of the Liushugou dykes (modified after Yin et al. 2010).(d) Simplified geological map of the Bieluagaxi dykes (modified after Yin et al. 2015a).

Figure 3 .
Figure 3. (Colour online) Outcrop photos of dykes and their host plutons.(a) Dykes intruded into the Karamay pluton.(b) Baked and chilled margins on the contact boundary between the dyke and the host rocks.(c) Dyke swarms intruded into Xiaerpu pluton.(d) Granitic as enclaves surrounded by the dyke.(e) Dykes intruded Lower Carboniferous strata (abbreviated and marked as C strata in the photo) in Liushugou area.(f) Intrusive contact between a NE-SW dyke and the Lower Carboniferous strata.(g) Bieluagaxi diorite intruded by the dyke with plagioclase xenocrysts on the edge of the dyke.(h) The dykes and the diorite interpenetrated with each other, and dioritic enclave and disharmonic fold.(i) The straight magmatic bands as orientation of aligned hornblende showing magma flow in the host Bieluagaxi granitic rocks.(j) The straight and curved magmatic bands with injection folds and granitic enclaves in the host Bieluagaxi granitic rocks.

Figure 4 .
Figure 4. (Colour online) Microscopic photos of the dykes in the West Junggar.(a) Typical diabasic texture that xenomorphic clinopyroxenes filling in the spaces between the semi-euhedral acerose plagioclases, in Karamay dolerite dyke under the cross-polarised light.(b) Plagioclase, chloritized Hb with Ap, Sph and Mt in the Karamay dioritic dyke under the single-polarised light.(c) Porphyritic texture that phenocryst mainly consists of idiomorphic Hb crystals of the Xiaerpu dioritic porphyrite in the single-polarised light.(d) Mineral assemblages of the Xiaerpu dioritic porphyritic dyke in the cross-polarized light.Phenocryst consists of Pl and Hb, and the matrix shows microcrystalline texture with numerous acicular plagioclases and mafic minerals.(e) Hypidiomorphic sheet biotite as the phenocryst in dioritic porphyrite in Liushugou dioritic porphyrite in the singlepolarised light.(f) Fine-grained granular texture with mineral assemblages of Pl, Hb and Opx in the Liushugou dioritic dyke under the cross-polarised light.(g) Long columnar Hb and plate-shaped Pl in the Bieluagaxi dioritic dyke under the single-polarised light.The euhedral amphiboles and acicular apatites embedded in the plagioclase indicate that the magma of dioritic dyke was initially quenched into the unconsolidated host granitic magma and then slowly cooled with it.(h) Porphyritic texture with twining Hb as phenocryst in the Bieluagaxi dioritic porphyritic dyke in the cross-polarised light.Pl = plagioclase, Opx = orthopyroxene, Cpx = Clinopyroxene, Hb = hornblende, Bi = biotite, Ap = apatite, Chl = chlorite, Sph = sphene, Mt = magnetite.

Figure 6 .
Figure 6.(Colour online) U-Pb Concordia diagrams showing zircon ages obtained by LA-ICP-MS for dykes in West Junggar.
1732 R Gao et al.A plot of Th/Nd ratio vs. ( 143 Nd/ 144 Nd) i /εNd(t) (Johnson & Plank, 1999; Zhang et al. 2021b) can be used to model the contribution of sediment-derived melt and three-component mixing models (DePaolo et al. 1991) for the dykes are shown Fig.

Table 1 .
Whole rock major (wt.%) and trace (ppm) element of Late Carboniferous dykes in West Junggar

Table 3 .
Sr-Nd isotopic data of the Late Carboniferous dykes in West Junggar