The Dulate arc, located in East Junggar (NW China) in the southern Central Asian orogenic belt, records a Devonian magmatic arc evolution, offering a window to understanding the orogenic processes of the Central Asian orogenic belt. Here we present new geochemical and isotopic data for Late Devonian high-Mg andesite (HMA) and Nb-enriched basalt (NEB) suites from the Qiakuerte area, East Junggar. The HMA samples are typical subduction-related volcanic rocks. They have SiO2 contents ranging from 53.30 to 54.59 wt%, high MgO (5.0–5.26 wt%), and high Mg# values (~55) and show enrichments in large ion lithophile elements (LILEs) and depletions in high field strength elements (HFSEs). The HMA samples have high (La/Yb)N ratios and Sr/Y (~6.5 and 50–59, respectively) with no Eu anomalies. The HMA samples have high Na2O (~3.3 wt%) and low K2O (~2.5 wt%) and Th (~2.4 ppm) contents, combined with positive εNd(t) and low (87Sr/86Sr)i values. These characteristics suggest that the samples were formed mainly through interactions between subducted oceanic melts and mantle peridotites. Compared to normal arc basalts, the NEB samples have higher concentrations of Nb (~20 ppm), higher primitive mantle–normalized Nb/La (0.50–0.58), and higher ratios of Nb/U (9.4–14.6). The NEB samples also have positive εNd(t) and low (87Sr/86Sr)i values, indicating that their source was mantle wedge that had been metasomatized by slab melt. Considering the widespread presence of A-type granites, the abnormally high heat flow, and the tectonic characteristics of East Junggar, we conclude that a slab window created by the subduction of an ocean ridge was responsible for the melting of slab and the formation of the NEB-HMA suites. These processes may have also played a key role in the tectonic evolution processes of East Junggar during the Late Devonian.

The Central Asian orogenic belt represents one of the most important sites of juvenile crustal growth during the Paleozoic and is characterized by accretion of various terranes, such as island arc, accretionary prisms, seamounts, Proterozoic to Paleozoic mid-ocean-ridge basalt (MORB)– and supra-subduction zone–type ophiolites, and microcontinents (Kepezhinskas, 1986; Şengör et al., 1993; Jahn et al., 2000; Safonova, 2009; Furnes and Safonova, 2019). Previous studies have shown that at least half of the crustal material was due to the addition of juvenile material during the Neoproterozoic and Paleozoic (Şengör et al., 1993; Windley et al., 2002). Syngenetic deformation and sedimentary records between subduction and orogeny are obscured by later geological events, which makes it difficult to identify such processes. In these situations, identifying magmatism related to subduction of active spreading ridges may be an effective tool for defining the corresponding process. Generally, island-arc magmas are probably derived from the partial melting of a mantle wedge with calc-alkaline characteristics (Defant et al., 1991; Wortel and Spakman, 2000). But ridge subduction may cause calc-alkaline-type to oceanic island basalt (OIB)–type volcanism (Hole et al., 1991; Kinoshita, 1995; Guivel et al., 1999; Cole and Stewart, 2009; Thorkelson et al., 2011; Zhang et al., 2018b). In this case, a series of special magmas such as high-Mg andesite (HMA), Nb-enriched basalt (NEB) adakite and A-type granite and associated mineralization (such as large-scale porphyry) may occur on active continental margins (D’Orazio et al., 2001; Wang et al., 2007; Sun et al., 2009; Castillo, 2012). HMA is usually related to a subduction-related environment (Defant and Drummond, 1993; Sajona et al., 1993; Bourgois et al., 1996). NEB was originally observed from arc basalts, which are more enriched in high field strength elements (HFSEs), particularly Nb (compared to <10 ppm in typical arc basalts; Defant et al., 1992; Calmus et al., 2003; Kepezhinskas et al., 2020, 2022). NEB could be produced through melting of the mantle metasomatized through interaction with slab-derived melts (Sajona et al., 1993, 1994). Thus, the NEB-HMA suite is important for understanding the nature of mantle and tectonic processes such as crustal growth, oceanic crust subduction, and melt-rock interaction (Sajona et al., 1993; Wang et al., 2007; Yin et al., 2017; Zhang et al., 2020; Jing et al., 2022).

As a typical accretionary orogen, the Central Asian orogenic belt is situated between the North China and Tarim cratons to the south and the Siberian craton to the north (Fig. 1A). The East Junggar terrane, located in the southwestern part of the Central Asian orogenic belt, records the collisional processes between the peri-Siberian and Kazakhstan orogenic systems during the closure of the Paleo-Asian Ocean (Xiao et al., 2015). A striking feature of East Junggar is that it experienced subduction and collision processes due to successive consumption of the Paleo-Asian Ocean during the Paleozoic (Xiao et al., 2004, 2018; Scheltens et al., 2015). These tectonic processes resulted in the formation of the Dulate arc, the Yemaquan arc, and the Jiangjunmiao accretionary complex. Thus, East Junggar is an excellent natural laboratory that allows us to discover the subduction processes and melt-rock interaction that occurred during the late Paleozoic.

In this study, we present new whole-rock geochemical compositions and Sr-Nd isotopic data for seven basalt and andesite samples from the Dulate arc in East Junggar. These data provides insight into the geodynamic processes that influenced volcanism during the evolution of the Dulate arc.

The East Junggar terrane is located in the southern Central Asian orogenic belt (Fig. 1A). From south to north, East Junggar includes the Jiangjunmiao accretionary complex, Yemaquan arc, and Dulate arc, which are separated by the Kalamaili and Zhaheba-Armantai ophiolite belts (Fig. 1B) (Xiao et al., 2004, 2010). These ophiolites were formed during the Paleozoic progressive consumption of the southern branches of the Paleo-Asian Ocean (Coleman, 1989).

The Dulate arc extends along the northern side of the Zhaheba-Armantai ophiolite belt and is the northernmost part of the East Junggar terrane (Fig. 1B). It consists mainly of Devonian–Carboniferous volcanic and sedimentary rocks, which previous studies interpreted to have formed in an island arc setting by the northward subduction system (Zhang et al., 2008b, 2009; Tang et al., 2020). A late Carboniferous post-collisional setting is supported by the A2-type post-collisional granitoids in this area (Liu and Liu, 2014). The Yemaquan arc is located south of the Zhaheba-Armantai ophiolite belt and is dominated by Devonian to Carboniferous volcanic and volcanoclastic rocks (Xiao et al., 2009). In the Dulate and Yemaquan arcs, the granitic rocks were mostly emplaced at 270–320 Ma (Han et al., 2006; Zhang et al., 2006, 2018a; Mao et al., 2008; Shao et al., 2019; Xu et al., 2020a). The southernmost part of the East Junggar terrane is the Jiangjunmiao accretionary complex, which consists mainly of Devonian to Permian island arc rocks with high positive values of εNd(t) and zircon εHf(t) (Xiao et al., 2009; Su et al., 2012; Zhang et al., 2015).

The Zhaheba-Armantai ophiolite belt is situated between the Yemaquan arc to the south and Dulate arc to the north (Fig. 1B). Zircon U-Pb dating on rocks from the Zhaheba-Armantai ophiolite belt yielded early Paleozoic ages of 495.1 ± 3.5 Ma, 421.5 ± 4.1 Ma, and 503 ± 7 Ma for plagiogranite and 406 ± 4 Ma for gabbro (Jian et al., 2003; Xiao et al., 2009; Luo et al., 2017). Long et al. (2012) revealed that the early Paleozoic sediments in the Chinese Altai and the graywackes in the southern Yemaquan arc have similar detrital zircon age spectra. This evidence suggests the Zhaheba-Armantai ophiolite belt formed in the early Paleozoic and was emplaced before the Early Devonian (Long et al., 2012; Zhang et al., 2015). The age of the Kalamaili ophiolite has been determined by U-Pb zircon dating, and the ages include 373 Ma for plagiogranites (Tang et al., 2007) and 417 Ma and 330 Ma for mafic rocks (Wang et al., 2009a; Huang et al., 2012; Hu et al., 2014; Xu et al., 2015; Liu et al., 2017), in addition to an Devonian and early Carboniferous age for radiolarian chert (Shu and Wang, 2003; Liu et al., 2007; Shen et al., 2011). Previous studies have shown that the forearc features of these mafic rocks from the Kalamaili ophiolite range from island-arc tholeiite–like to MORB and OIB–like (Bu et al., 2005; Liu et al., 2007, 2017; Ma et al., 2007). Liu et al. (2017) proposed that the magma of the Kalamaili ophiolite, characterized by high temperatures and low pressures, was likely generated through ridge subduction.

We conducted a detailed study in the Qiakuerte area in the Dulate arc, located on the southern edge of the Dulate island arc belt (Figs. 1C and 1D). This region comprises mainly Middle Devonian–Carboniferous volcanic and volcaniclastic rocks and clastic rocks (Long et al., 2012; Tao et al., 2018). The Middle–Late Devonian strata in the study region are divided into two formations: the Tuolanggekuduke Formation (D2t) and Kaxiweng Formation (D3k) (Fig. 1D). Unit D2t is dominated by volcanic rocks (dacite, tuff, and andesite), sandstone, limestone, and tuff (Zhang et al., 2008a). Unit D3k is in angular-unconformity contact with the underlying D2t. Unit D3k comprises basaltic lava, andesite, tuff, and dacitic lava intercalated with sandstone. The Carboniferous strata consist of basalts and basaltic andesites with minor interbedded andesites, rhyolites, tuffs, siltstones, and carbonaceous shales (Xu et al., 2020b).

The suite of studied volcanic rocks belongs to the Kaxiweng Formation. (Figs. 1C and 1D) (XBGMR, 1993). The basaltic lava is ~220 m thick and covered by basaltic andesite. Fresh volcanic samples including four basalts and three basaltic andesites were collected from the section (Fig. 1D).

The basalts are dark gray, massive, and fine grained (Figs. 2A and 2C) and composed mainly of plagioclase (70–80 vol%) and clinopyroxene (10–20 vol%) with minor opaque minerals. Plagioclase (0.5–1 mm) is elongated and euhedral, while clinopyroxene is subhedral to anhedral and fills the spaces between plagioclase crystals (Fig. 2E). The basaltic andesite has a typical porphyritic texture (Figs. 2B and 2D) and consists of plagioclase (30 vol%, 3 mm) and clinopyroxene phenocrysts (40 vol%, 3–5 mm) within a groundmass of short and elongate plagioclase and minor clinopyroxene, opaque minerals (Fig. 2F).

Fresh specimens of whole-rock samples were powdered to <200 mesh using a tungsten carbide ball mill. Loss on ignition (LOI) was obtained by weighing after 1 h of calcinations at 1100 °C. Then, rock powders (~0.6 g) were then dissolved with Li2B4O7 (6 g) in a TR-1000S automatic bead fusion furnace at 1100 °C for 20 min. Major element abundances (in weight percent) were determined on fused glass beads by X-ray fluorescence (Shimadzu XRF-1500) at the Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing. The precision for all major elements is estimated to be within 5% with an accuracy of better than 5%. For trace element abundance analyses, rock powders (50 mg) were dissolved using a mixed acid (HF:HClO4 = 3:1) in capped Savillex Teflon beakers at 120 °C for six days and subsequently dried to wet salt and redissolved in 0.5 ml HClO4, dried, and then diluted to 100 ml for analysis. Trace element concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS; Nu Instruments AttoM) at the Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences. Indium was used as an internal standard to correct for matrix effects and instrument drift. We used pure elemental standards for external calibration. Granite GSR1 and basalt GRS3 were used as reference materials for all analyses. The precision for most trace elements is estimated to be at 5% for most elements according to the analyses of the GSR1 and GSR3 standards.

The Sr-Nd isotopic compositions of the volcanic rocks were determined by Thermo Fisher Scientific NEPTUNE Plus multi-collector ICP-MS. About 150 mg of sample powders were first spiked with mixed isotope tracers and then dissolved in a solution of HNO3 and HF in Teflon capsules before Rb-Sr and Sm-Nd isotopic analyses. The sample solutions were separated using a conventional cation-exchange technique. 87Sr/86Sr ratios were corrected for instrumental mass fractionation using the exponential fractionation law and assuming 88Sr/86Sr = 8.375209. 143Nd/144Nd ratios were corrected for instrumental mass fractionation using the exponential fractionation law and assuming 146Nd/144Nd = 0.7219. The NBS 607 and BCR-1 standards, with 87Sr/86Sr = 1.20032 ± 28 (2σ) and 143Nd/144Nd = 0.512626 ± 9 (2σ), respectively, were used to assess analytical precision, which was ~1% for 87Rb/86Sr and ~0.5% for 147Sm/144Nd. The Nd-depleted mantle model ages (TDM) were calculated using 143Nd/144Nd ratio of 0.513151 and a 147Sm/144Nd ratio of 0.21357 for the present-day depleted mantle and an average crustal 147Sm/144Nd ratio of 0.118 (Jahn and Condie, 1995).

Whole-rock major and trace elemental data for seven samples are listed in Table 1. Based on petrographic and geochemical features, the studied samples can be divided into two groups: basalt and basaltic andesite. Because the samples exhibit high LOI values (1.60–3.26) and the presence of secondary minerals, the effect of alteration should be evaluated before using the geochemistry of basalts and basaltic andesites to constrain their petrogenesis and origin. Firstly, most analyzed samples have LOI values <2 wt% and Ce anomalies of 0.97–1.03, which suggest negligible fluid-assisted alteration and metamorphism (Polat and Kerrich, 2002). Zirconium is an HFSE that remains immobile during post-magmatic alteration, while large ion lithophile elements (LILEs) can easily migrate (Petersen, 1983). The samples display positive correlations between fluid-mobile elements (e.g., K, Rb, Th, and La) and Zr (Verma, 1981; Polat and Hofmann, 2003) (Fig. S1 in the Supplemental Material1). Additionally, the uniform Sr-Nd isotope compositions (Table 2) suggest that these isotopes were not affected by alteration and may represent the initial values (Verma, 1981). This evidence indicates that the variation in elemental content of the samples is likely due to magmatic processes rather than post-magmatic alteration. Therefore, the effect of alteration on the geochemistry of these intrusive rocks can be considered negligible. In the following plots and discussion, all oxide contents of the samples have been recalculated to 100% on a volatile-free basis with all Fe as Fe2O3T.

Basalt

The studied basaltic samples have SiO2 contents of 49.41–49.57 wt% and display high total Fe2O3 (12.81–12.93 wt%) and MgO (4.93–4.98 wt%), Mg# (100 × [atomic MgO/(MgO + FeO)]) values = 47.1–47.4) contents. All the basalts are tholeiitic basalts on SiO2 versus Zr/TiO2 and FeOT/MgO (T is all Fe as FeOT) versus SiO2 diagrams (Figs. 3A and 3B) (Miyashiro, 1974; Winchester and Floyd, 1977). In addition, these basalts possess relatively high TiO2 (2.73–2.78 wt%) and P2O5 (~1.1 wt%) contents and high HFSE contents (Nb = 19–20 ppm, Zr = 431–447 ppm, [Nb/La]PM [PM is primitive-mantle normalized] = 0.49–0.55, and Nb/U = 10–15) compared to normal arc volcanic rocks, similar to those of typical NEBs ([Nb/La]PM >0.5, Nb >10 ppm, and TiO2 >1 wt%; Sajona et al., 1994; Kepezhinskas et al., 1996; Figs. 4A4C). They all plot in the NEB field in the discrimination diagrams of P2O5 versus TiO2 (Fig. 4A), Nb/U versus Nb (Fig. 4B), and Nb/La versus MgO (Fig. 4C). Besides, these samples are enriched in light rare earth elements (LREEs) ([La/Yb]N [N is chondrite normalized] = 5.53–6.22) with slightly negative Eu anomalies (Fig. 5A; Eu/Eu* (Eu/Eu* =EuN/[SmN × GdN]0.5 = 0.91–0.96). In a primitive mantle–normalized spider diagram, the samples show remarkable enrichment of LILEs (e.g., Ba and U) relative to HFSEs with negative Nb-Ta-Ti anomalies (Fig. 5B), consistent with the geochemical characteristics of subduction-related magmas.

Basaltic Andesite

The Qiakuerte basaltic andesite samples contain intermediate SiO2 (53.30–54.59 wt%), Al2O3 (~16.30 wt%) and high TiO2 (~1.05 wt%) and Fe2O3T (9.03–9.86 wt%). They exhibit higher MgO (5.00–5.26 wt%), Cr (80.0–95.62 ppm), Ni (37.2–40.4 ppm), and Mg# (55–57) values and lower FeOT/MgO ratios (~1.6) and CaO contents (7.25–7.47 wt%) than typical volcanic rocks (Tatsumi and Ishizaka, 1982). These features are consistent with typical HMA (Fig. 4D) (Kelemen, 1995; Tatsumi, 2001; Polat and Kerrich, 2002). Meanwhile, the basaltic andesite samples have extremely high Sr (983.0–1063.7 ppm) and low Y (19.7–21.2 ppm) and Yb (1.6–2.0) contents, resulting in high Sr/Y (45.1–51.4) and uniform (La/Yb)N (~6.5) ratios. In the SiO2 versus FeOT/MgO diagram, the samples exhibit tholeiitic characteristics (Fig. 3B). Furthermore, they are characterized by fractionated REE patterns (Fig. 5C) and strong depletion in heavy REEs (HREEs) and Y, with negligible Eu anomalies (Eu/Eu* = 0.93–0.98). These samples are enriched in LILEs (e.g., Rb, Sr, and Ba) and depleted in HFSEs (e.g., Nb, Ta, and Ti) in the trace element diagram (Fig. 5D).

The whole-rock Sr-Nd isotope analytical data for the basalt and basaltic andesite are listed in Table 2. Here, we use an age of 377 Ma to calculate initial Sr-Nd isotopic compositions (as discussed in the age of Qiakuerte volcanic rocks of the Discussion section). The measured 87Sr/86Sr and 143Nd/144Nd ratios vary from 0.704039 to 0.704084 and 0.512819 to 0.512827, respectively, for the basalt, and for the basaltic andesite, they vary from 0.704945 to 0.704995 and 0.512762 to 0.512763, respectively. The calculated initial 87Sr/86Sr for basalt and basaltic andesite range from 0.703627 to 0.703737 and 0.704104 to 0.704152, respectively. The calculated εNd(t) varies from 6.7 to 6.9 and from 5.0 to 5.5 for the basalt and basaltic andesite, respectively (Fig. 6).

Age of Qiakuerte Volcanic Rocks

Although zircon U-Pb isotopes of basaltic andesite have been determined in this study, unfortunately, no effective age has been obtained, probably because Zr abundance is very low (~100 ppm; Table 1); alternatively, contemporaneous zircons may be too small for extraction by conventional methods.

The age of the Kaxiweng Formation has been determined by a series of studies (Jia et al., 2019; Ren, 2019; Liu et al., 2021; Wang, 2021). Jia et al. (2019) reported Late Devonian iconic fossils (ammonoids Kosmoclymenia cf. tabulata Sun et Shen). Additionally, Liu et al. (2021) recorded Late Devonian plant fossil from the Kaxiweng Formation adjacent to our study area (Fig. 1C). And, palynological study also has determined the age of the Kaxiweng Formation to be early Late Devonian (Frasnian) (Jia et al., 2019). In addition, a series of high-precision zircon U-Pb isotope analyses also shows the same results (An, 2021; Ren, 2019; Xu, 2020). The Kaxiweng Formation exhibits an unconformable contact with the Middle Devonian Tuolanggekuduke Formation (386 Ma), and the zircon U-Pb ages (ca. 354 Ma) of the overlying Nanmingshui Formation volcanic rocks were determined by An (2021). Besides, a reliable zircon U-Pb age of 376.0 ± 8.3 Ma was obtained by laser ablation ICP-MS analyses from the andesites of the Kaxiweng Formation (Ren, 2019). Recently, Xu (2020) also reported a similar zircon U-Pb age of 377.0 ± 5.1 Ma from the andesitic porphyrite at the base of the Kaxiweng Formation, which has similar petrographic features and geochemical characteristics to the studied andesites. In short, based on abundant evidence, the eruption age of the NEB and HMA suite at the base of Kaxiweng Formation in this study is ca. 377 Ma.

Petrogenesis of the HMA and NEB Suites

Effects of Crustal Contamination

All geochemical data from basalts and basaltic andesites suggest that crustal assimilation was insignificant to the petrogenesis of the rocks. Firstly, the analyzed samples have higher Ti/Y ratios (300–342) but much lower Th/La (0.04–0.15) and Th/Ce (0.02–0.07) ratios than continental crust (Taylor and McLennan, 1995; Plank, 2005). Meanwhile, if mantle-derived magma were affected by crustal contamination, samples would give rise to negative Nb-Ta anomalies but positive Zr-Hf anomalies (Rudnick and Gao, 2003) with a high (Th/Nb)N value (>>1). This is commonly regarded as a trace element index of crustal contamination (Saunders et al., 1992; Rudnick and Fountain, 1995). The geochemical features of the samples are inconsistent with crustal contamination, with low (Th/Nb)N (0.5–1.7) and slightly negative Zr-Hf anomalies (Figs. 5B and 5D). Therefore, crustal contamination is negligible in these HMA and NEB samples.

Petrogenesis of HMAs

Four models have been proposed to account for the origin of specific occurrences of HMAs. These include: (1) partial melting of enriched mantle under hydrous conditions (Stern and Hanson, 1991; Moyen et al., 2001; Martin et al., 2005), (2) melting of delaminated lower crust (Kelemen et al., 1998; Gao et al., 2004), (3) interaction between subduction slab–derived melts or oceanic sediments and mantle wedge peridotites (Kelemen, 1995; Yogodzinski et al., 1995; Tatsumi, 2006), and (4) mixing between crustal melt and magma from metasomatized mantle (Streck et al., 2007). Hirose (1997) showed that the melt composition under the enriched mantle condition (model 1) exhibits Al2O3 contents ranging from 17.24 wt% to 21.70 wt% and CaO contents ranging from 8.53 wt% to 9.99 wt% through melting experiments, which is not consistent with Qiakuerte HMAs (Al2O3 = 16.29–16.43 wt%; CaO = 7.25–7.47 wt%). Further, the characteristics of Sr-Nd isotopes (εNd(t) >0) of the studied HMAs are different from those of the enriched mantle. These signatures suggest that they cannot have formed by partial melting of enriched mantle. For model 2, Qiakuerte HMAs possess very low Th (~2.4 ppm) and Th/Ce (~0.06) which are different from values in melts of partial melting of delaminated lower crust with high contents of incompatible elements (Th = 10–20 ppm; Plank, 2005; Wang et al., 2006). Likewise, experimental petrology has further clarified that melts originated from meta-basaltic rocks and eclogites at 1.2–4 GPa commonly show low Mg# values and MgO contents regardless of the degree of melting (Rapp et al., 1999). The HMAs in the Qiakuerte region possess high MgO (5.00–5.26 wt%) contents and Mg# values (~55), also indicating that they cannot have formed by partial melting of thickened lower crust. In the case of model 3, HMA magma is usually considered to be primitive and to represent a nearly primary magma originating in the mantle (Kelemen et al., 2004). However, studied HMAs are notable enrichments in SiO2 (53.30–54.59 wt%), Al2O3 (~16.30 wt%), LREEs, LILEs, and the depletions in HREEs and HFSEs.

We suggest that the HMAs were most probably generated by mixing between crustal melts and magma from the metasomatized mantle (model 4), based on the following lines of evidence: (1) The HMAs exhibit Mg# ranging from 51 to 55, which is significantly higher than those of melt derived directly from partial melting of the mafic lower crust (Mg# <45, as established by Rapp and Watson [1995]) and typical adakites (~50, as described by Martin [1999] and Richards and Kerrich [2007]). The high Mg# values along with the enhanced concentrations of Cr, Ni, and V in our samples imply the involvement of mantle-derived peridotite. Meanwhile, the HMAs exhibit positive εNd(t) values ranging from +4.97 to +5.52 and initial (87Sr/86Sr)i ratios within the range of 0.704104–0.704152. This geochemical signature is similar to that of coeval ophiolites in East Junggar, suggesting that the mantle component contributed to the genesis of the HMAs (Fig. 6). (2) The enrichments of SiO2 (53.30–54.59 wt%); Al2O3 (~16.30 wt%), LREEs and LILEs, the depletions in HREEs and HFSEs, and low to moderate Ni (37–40 ppm) and Cr (80–96ppm) contents for the HMAs of this study indicate the involvement of crustal materials in their petrogenesis. (3) The HMAs have relatively high Sr/Yr ratios (45.1–51.4) and TiO2 contents (~1.05 wt%), which are similar to those of bajaitic HMAs when plotted on Sr/Y versus Y and TiO2 versus Mg# diagrams (Figs. 7A and 7B). Bajaitic HMAs are widely considered to be generated by the interaction between slab-related melts and magma from the metasomatized mantle (Kay, 1978; Kelemen, 1995; Yogodzinski et al., 1995; Shimoda et al., 1998; Tatsumi, 2001; Kamei et al., 2004). In summary, we argue that the generation of HMAs in the Qiakuerte region was most likely linked to mixing of crustal melts and depleted mantle peridotites.

Petrogenesis of NEBs

Previous studies have proposed two petrogenetic models to interpret the origin and distinctive geochemical characteristics of NEBs in convergent margins: (1) OIB mantle as the source for NEB (Reagan and Gill, 1989; Kepezhinskas et al., 1996, 2022) and (2) mantle wedge interaction with adakitic melts (Kepezhinskas et al., 1996; Sajona et al., 1996). OIB mantle as the source for Qiakuerte NEB is unreasonable. In this model, NEBs would be generated from an enriched mantle source, showing positive anomalies of Nb, Ta, and Ti on the trace element distribution diagram (Reagan and Gill, 1989; Sun and McDonough, 1989). This is inconsistent with the distinctive negative Nb, Ta, and Ti anomalies observed in the studied NEBs (Fig. 5D). Furthermore, the average values of Nb/U (10–14) and Ce/Pb (~13) for representative NEB samples are much lower relative to those of OIBs (47 ± 10 and 25 ± 5, respectively; Hofmann et al., 1986), which does not support that our samples originated from the mantle source of OIB. Instead, their positive Ba, Sr, and U anomalies and negative Nb, Ta, and Ti anomalies probably reflect a mantle source that was previously modified by subduction-related materials (Pearce and Peate, 1995). Although rutile as a residual might lead to Nb-Ti depletion, the Nb and Ti content in slab melt is much higher than in slab-derived fluid because Nb and Ti preferentially reside in melt (Keppler, 1996). Combined with depleted Nd isotopes (εNd(t) = 6.7–6.9), the evidence from trace elements suggests that the NEBs likely originated from a depleted mantle source metasomatized by slab melts. Contemporaneous (378 Ma) adakitic rocks associated with the melting of subducted oceanic crust also occur in the Dulate arc region (Zhou et al., 2019). Slab-melt metasomatism in the mantle source of NEBs is also supported by the Nb/Yb versus Nb diagram (Fig. 8A). In the diagram, the research NEBs all fall in the field of NEBs originated from mantle wedge metasomatized by slab melts instead of the field of slab window basalts originated from asthenospheric mantle melting without reaction with the overlying mantle wedge. The involvement of slab components can be further demonstrated by high La/Nb (1.74–1.99) but low La/Ba (0.08–0.09) ratios, as illustrated on the La/Nb versus La/Ba diagram (Fig. 8B), in which all Kaxiweng Formation rocks are typical features of a subduction-modified lithospheric mantle source.

Because trace elements with similar partition coefficients hardly change during magmatic process, they, especially REEs, can provide an independent means of assessing the depth and extent of partial melting (Aldanmaz et al., 2000). Thus, REEs are used to trace the origin of the magmas and depth and degree of mantle melting. The NEBs have relatively low Ce/Y ratios (<3), indicating that they were generated within the spinel-garnet stability field at a depth of <80 km (McKenzie and O’Nions, 1991). The low (Tb/Yb)N ratios (<2) further suggest that the NEBs derived from spinel-bearing peridotites (Wang et al., 2002). In the Sm/Yb versus Sm and Dy/Yb versus La/Yb diagrams (Figs. 9A and 9B). They are plotted around the garnet-spinel lherzolite melting curve, implying a mantle source consisting of spinel-garnet lherzolite, and are modeled as low-degree (1%–5%) partial melting of such a mantle source (Figs. 9A and 9B). So, the most probable petrogenesis for NEBs is the partial melting of slab melt–metasomatized mantle wedge.

Geodynamic Processes

East Junggar has traditionally been thought to have resulted from the northward subduction of the Paleozoic Paleo-Asian Ocean crust beneath the Siberian plate (Şengör et al., 1993; Xiao et al., 2004, 2014, 2015; Liu et al., 2017; Xu et al., 2020a). The tectonic evolution in the Dulate arc and even the whole East Junggar during the Devonian still remains debated. Models for the formation of the Dulate arc include: (1) The Dulate arc was formed in the Paleozoic due to the southward subduction of the Paleo-Asian Ocean (Zhang et al., 2004, 2018a; Wang et al., 2020), (2) the Dulate arc was formed by the northward subduction of the Armantai Ocean (Xu et al., 2020a), and (3) the northward subduction of the Kalamaili Ocean is the main reason for the formation of the Dulate arc.

The tectonic model invoking northward subduction of the Kalamaili Ocean is favored here because it can explain most of the geological features of East Junggar. Firstly, based on southward-younging structures associated with the Dulate arc, northward subduction of the oceanic crust is advocated by Xiao et al. (2009). Secondly, the presence of Devonian boninite, adakite, HMA, NEB, and picrite indicates a subduction-related tectonic setting in the Dulate arc (Niu et al., 2006; Zhang et al., 2008b). Thirdly, the Chinese Altai to the north of the Dulate arc has the characteristics of having been a back-arc basin setting during the late Paleozoic (Xu et al., 2003; Zhang et al., 2003; Shen et al., 2018). The Kalamaili and Armantai ophiolites lie to the south of Dalute Arc. In addition, the following data support the idea that the Dulate arc was generated by northward subduction of the Kalamaili Ocean: (1) Researchers argue that available zircon U-Pb ages reveal that the Armantai ophiolite belt has ages between 406 Ma and 503 Ma (Jian et al., 2003; Xiao et al., 2008), which was before the Early Devonian (Long et al., 2012; Zhang et al., 2015; Zhao et al., 2022). Also, massive Early–Middle Devonian picrites (Mao et al., 2008), boninites (Qin et al., 2002), HMAs and NEBs (Zhang et al., 2005; Niu et al., 2006), and adakitic rocks (Zhang et al., 2009; Li et al., 2014) are distributed in the northern side of the Kalamaili ophiolite. This evidence suggests that subduction events continued to occur in a forearc setting in the Dulate arc until the latest Middle Devonian. (2) Meanwhile, the age of the Kalamaili ophiolite has been determined by U-Pb zircon dating, including 373 Ma (Tang et al., 2007) for the plagiogranites and 417 Ma (Huang et al., 2012; Hu et al., 2014), 330 Ma (Wang et al., 2009a), and 371 Ma (Xu et al., 2015; Liu et al., 2017) for the mafic rocks in addition to a Devonian and early Carboniferous radiolarian chert age (Shu and Wang, 2003; Liu et al., 2007). These data for the Kalamaili ophiolite belt indicate that subduction of oceanic crust occurred from the early Paleozoic to the early Carboniferous. Specifically, during the Late Devonian, the dynamics and evolution of the Dulate arc were associated with the subduction of the Kalamaili oceanic slab.

Magmatic suites with specific rock assemblages and compositions are considered to be an important key to understanding tectonic-magmatic history. The NEB-HMA suite in this study can help to refine the Devonian subduction process in East Junggar. As mentioned in the Results section, both NEBs and HMAs have been identified in the Late Devonian Dulate arc. Petrogenetic studies show that these rock types are the result of partial melting of a peridotitic mantle wedge that had undergone previous metasomatism by melts derived from the subducted slab. It is difficult for subducting oceanic crust to undergo large-scale partial melting because melting is always confined to an “abnormally hot” environment. Researchers have summarized two reasons for abnormally hot petrogenesis: (1) the initiation of subduction with young oceanic crust (Defant et al., 1992; Sajona et al., 1993; Peacock et al., 1994) and (2) a slab window related to slab break-off (Defant and Drummond, 1990; Yogodzinski et al., 1995; Calmus et al., 2003; Wang et al., 2007; Castillo, 2012) or ridge subduction (Kinoshita, 1995; Aguillón-Robles et al., 2001; Cole and Stewart, 2009; Windley and Xiao, 2018). Given that arc-related magmatism began no earlier than the late Cambrian (Luo et al., 2017) and subsequently lasted until the Late Devonian (Zhang et al., 2008b; Wan et al., 2011; Wang et al., 2021). Thus, considering the timing of subduction initiation, the Dulate region was an island arc during the Late Cambrian (Liu et al., 2016). This also indirectly rules out a slab break-off model because such break-off commonly occurs in the initial stage of continent-continent collision (van de Zedde and Wortel, 2001; Coulon et al., 2002; Macera et al., 2008; Chen et al., 2014).

Numerous studies have shown that slab windows can be formed as a consequence of ridge subduction, leading to anomalous thermal and chemical conditions (Thorkelson, 1996; Sisson et al., 2003; Groome and Thorkelson, 2009). In East Junggar, these in turn have resulted in anomalous and widespread regional magmatism that has produced different rock types. This proposal is further supported by the following evidence. Firstly, the occurrence of anomalous thermal conditions during the Devonian period is supported by the Middle Devonian HMAs and picrites found in the Dulate arc (Niu et al., 2006; Zhang et al., 2008b). The formation of HMAs and picrites necessitates the occurrence of notably elevated temperatures to melt the plate or associated sediments (Kelemen, 1995). Based on the clinopyroxene-liquid thermobarometer, research shows that picrites formed at high temperature (1335–1356 °C) (Zhang et al., 2008b). Furthermore, Niu et al. (2006) reported boninite juxtaposed with adakite in the Devonian rocks in the Dulate arc and proposed high thermal gradients based on the petrogenesis. Meanwhile, Late Devonian A-type granites with high εNd(t) (+6.7 to +7.7) in the Dulate arc (Liu and Yuan, 1996; Jian et al., 2003; Tong et al., 2006; Zhang et al., 2006; Shen et al., 2011) further support the model that the NEB-HMA suite was generated in an extensional setting and induced by a slab window caused by ridge subduction (Liu and Yuan, 1996; Jian et al., 2003; Tong et al., 2006; Zhang et al., 2006; Shen et al., 2011).

As mentioned in the petrogenesis of the HMA and NEB suites of the Discussion section, the Late Devonian Qiakuerte HMAs and contemporaneous NEBs were generated by partial melting of peridotitic mantle, which was previously metasomatized by slab melts. Late Devonian (378 Ma) diorites are observed in the Dulate arc, which have also been suggested to be related to slab-derived melts (Zhou et al., 2019). Meanwhile, the Late Devonian rocks with a 371 Ma age from the Kalamaili ophiolite have also been suggested to be related to ridge subduction (Liu et al., 2017). All these suggest that the ridge subduction of the Kalamaili Ocean is a plausible explanation for the diverse magmatic rocks (NEB, HMA, adakite, and A-type granite) in East Junggar. From the point of view of kinematics, a ridge subduction slab window can be formed when the ridge is orthogonal, oblique, or near parallel with the trench, as exemplified by those from Central America, Patagonia, West Junggar, and Northern Cordillera (Thorkelson, 1996; Yogodzinski et al., 2001; Thorkelson et al., 2011; Windley and Xiao, 2018). The spatial distribution of associated rocks (adakite, NEB, HMA, A-type granite, and picrite) juxtaposed with Kalamaili ophiolite (Fig. 1B) and geochronological ages of formations suggest a southward younging in the Dulate arc (Niu et al., 2006; Xiao et al., 2009). Moreover, Wong et al. (2010) found that 439 Ma Alegedayi ophiolitic mélanges (located in the northern part of the Chinese Altai) reflected an early Silurian ridge subduction. Meanwhile, Sun et al. (2009) found that ridge subduction occurred in the Chinese Altai at 420 Ma. The Late Devonian (377 Ma) in the Dulate region of this study is also characterized by ridge subduction. Liu et al. (2017) also suggest that the magma of the Kalamaili ophiolite (the southernmost part of East Junggar) showing low pressure and high temperature may have experienced ridge subduction during 371 Ma. All this evidence indicates a triple junction that has a southward movement.

In summary, our new results combined with previous studies suggest that the ridge subduction of the Kalamaili Ocean plausibly explains the diverse magmatism (NEBs, HMAs, picrites, and A-type granites) in East Junggar. The proposed model is shown in Figure 10. During the Late Devonian, the formation of a slab window followed ridge subduction with an extension effect. The upwelling of sub-oceanic asthenosphere through the slab windows provided anomalous heat, which induced partial melting of the mantle wedge and slab edges (Thorkelson, 1996). This slab windows model explains the generation of the studied HMA and NEB suite.

The NEB-HMA suite of basalts and contemporaneous basaltic andesites is interpreted as being the product of the partial melting of mantle peridotite that was metasomatized by slab-derived melts and mixing between crustal melts and magma from the metasomatized mantle. This rock association requires melting of subducting oceanic lithosphere and associated melt–mantle wedge interactions, suggesting an environment of abnormal high heat flow in the Late Devonian. Based on a compilation and evaluation of data on the widely developed coeval A-type granites and tectonic characteristics, a slab window triggered by ocean ridge subduction was identified in the Dulate arc region during the Late Devonian.

1Supplemental Material. Incompatible elements (K O, Rb, Th, and La) versus Zr plots of the volcanic rocks of Kaxiweng Formation, in order to distinguish the alteration effects. Please visit https://doi.org/10.1130/GEOS.S.25009121 to access the supplemental material, and contact editing@geosociety.org with any questions.
Science Editor: Andrea Hampel
Associate Editor: Wei Wang

Constructive reviews and suggestions by Pavel Kepezhinskas and another anonymous reviewer and editorial suggestions helped to improve the revised version. We thank Qingchen Wang and Xiaofa Yang for their help in fieldwork. We appreciate the Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, for assistance in major-trace element and Sr-Nd isotopic analyses. This study was financially supported by the National Natural Science Fund of China (grant 41502110).

Gold Open Access: This paper is published under the terms of the CC-BY-NC license.