Early Paleozoic Continental Arc Mafic Magmatism in the North Qaidam Tectonic Belt: Implications for Subduction of the Proto-Tethyan Oceanic Lithosphere

Magmatism produced by partial melting of crust or mantle occurs during subduction, collision, and postcollisional processes. Magmatism at different tectonic stages from oceanic subduction to continental subduction/collision carries crucial information about the nature of crust and/or mantle and provides indirect evidence of crust-mantle interaction. Subduction zones are the locus for recycling crustal rocks into the mantle and a principal driver of crust-mantle interaction. Continental subduction/collision is accompanied by crustal thickening and generation of crust-derived granitic magma with minor involvement of mantle materials [1–3]. By contrast, subduction of oceanic crust is the primary process that recycles crustal material into the mantle and plays a crucial role in developing heterogeneous mantle reservoirs [4–7]. Arc magmatism—especially mafic arc magmas generated during oceanic subduction at convergent plate margins—is characterized by distinctive trace element features such as enrichment in large-ion lithophile elements (LILE) and light rare-earth elements (LREE) but depletion of the high field-strength elements (HFSE); these magmas are generally sourced from a mantle wedge that was metasomatized by subducted oceanic slab-derived fluid and/or melt [8–11]. These arc-related magmas are considered as an important indicator of oceanic subduction [10, 12] and represent a crucial mode of crust-mantle material exchange [13, 14]. Continental arcs are thought to be the dominant location for long-term continental crustal growth [15–17]. Thus, continental arc magmas provide a window into the nature of the subarc mantle and the processes of crust-mantle interaction and crustal growth and can be used to track the closure of oceanic basins and recycling of subducted oceanic crust [11, 18, 19].

The North Qaidam tectonic belt (NQTB) is an early Paleozoic subduction-collision belt that records the opening of Proto-Tethyan Ocean, followed by oceanic subduction and closure, continental subduction/collision, and the final orogenic collapse [20–27]. Previous studies have mostly focused on high-pressure to ultrahigh-pressure metamorphic rocks and granitic rocks that are related to continental subduction/collision [27–31]. Some studies propose that ocean basin closure and continental collision in the NQTB occurred before 500 Ma [32], although most studies suggest that continental collision in the NQTB was no earlier than 460–450 Ma [21, 25, 28, 33, 34]. The geochemical signature and nature of the source of contemporaneous mafic arc magmatism during oceanic subduction have important implications for resolving this controversy. However, syn-subduction magmatism, particularly mafic magmatism, has received little attention, and thus, the nature of their source and the tectonic implications are poorly constrained. Although minor continental arc mafic rocks of early Paleozoic age have been reported in the NQTB [35, 36], it remains uncertain if they represent a continental arc-related magmatic system in the northern Oulongbuluke microcontinent. However, identification and reconstruction of continental arc-related magmatic systems are of great significance for constraining the timing of the Proto-Tethyan Ocean closure and continental collision and for understanding subduction of the Proto-Tethyan Ocean and the tectonic evolution of the NQTB.

In this contribution, we present an integrated study of petrology, whole-rock geochemistry (major trace elements and Sr–Nd isotopes), zircon U–Pb geochronology, and zircon Lu–Hf isotope analysis of recently identified continental arc mafic rocks in the North Wulan metamorphic complex to decipher the origin of these mafic rocks, to elucidate the nature of mantle-crust interaction during subduction of the Proto-Tethyan Ocean, and to reveal their geodynamic significance. Early Paleozoic continental arc mafic magmas in the NQTB originated from the subcontinental lithospheric mantle (SCLM) that was metasomatized by the altered basaltic oceanic crust-derived fluids and/or sediment-derived melts during the subduction of the Proto-Tethyan oceanic lithosphere. Our results provide new constraints on the early Paleozoic tectonic process operated in the NQTB and on the subduction and closure of the Proto-Tethyan Ocean.

The North Qaidam tectonic belt is located in the south of the South Qilian tectonic belt along the northeastern margin of the Tibetan Plateau (Figures 1(a) and 1(b)) and belongs to the Proto-Tethyan tectonic domain. It records early Paleozoic subduction and closure of the Proto-Tethyan Ocean and subsequent continental collision/subduction [20–23, 27, 28, 37]. It is composed of two lithotectonic segments, including the Oulongbuluke microcontinent in the north and the North Qaidam subduction-collision complex belt in the south (Figure 1(b)), which are separated by the roughly east–west-trending Wulan–Yuka strike-slip fault [38]. It is bounded to the north by the South Qilian tectonic belt and to the south by the Qaidam Block (Figure 1(b)). The Altyn-Tagh fault and Wahongshan-Wenquan fault define the western and eastern boundaries of the NQTB.

The North Qaidam subduction-collision complex belt, situated in the southern segment of the NQTB, contains a HP-UHP metamorphic complex and the Tanjianshan Group. The HP-UHP metamorphic complex is dominated by the high-pressure orthogneiss and paragneiss with minor eclogite, garnet peridotite, and mafic granulite [22, 25, 27–29, 31, 39–48]. The occurrence of continental-type eclogite and coesite-bearing paragneiss suggests that crustal components of the Qaidam block were subducted to mantle depths at 450–420 Ma [22, 27, 29, 31, 41, 44]. Some eclogites have N-MORB or E-MORB trace element compositions with inferred protolith ages of 540–500 Ma and metamorphic ages of 460–440 Ma; these were interpreted to represent components of early Paleozoic metamorphic ophiolites [28–30, 47]. These features suggest that the oceanic slab was subducted to mantle depths and experienced HP-UHP metamorphism prior to continental subduction. Garnet peridotite occurs as lenses in HP-UHP metamorphic granitic and pelitic gneiss. Diamond was identified in zircon from garnet peridotite in the Lüliangshan area, which suggests that garnet peridotite could have exhumed from depths greater than 100–200 km [43, 44]. Recent studies propose that these garnet peridotites originated from an ancient subcontinental lithospheric mantle wedge above an early Paleozoic subduction zone and record metasomatism of mantle wedge peridotite by hydrous fluids/melts derived from the subducted slab [39, 46].

The Tanjianshan Group is composed of greenschist- to amphibolite-facies volcano-sedimentary rocks, including basalt, andesite, dacite, gabbro, diabase, volcaniclastic rocks, and limestone [49, 50]. Intermediate to mafic volcanic rocks have geochemical compositions similar to those from island arcs or MORB and have zircon U–Pb ages of 535–486 Ma [20, 51–53]. Fu et al. [20] divided the Tanjianshan Group into three tectonic units, including an accretionary complex, forearc basin, and island arc. Together with the forearc ophiolite, these units were interpreted as products of a short-lived intraoceanic arc-trench system that developed during subduction of the Proto-Tethyan oceanic slab [20].

The Oulongbuluke microcontinent is situated at the northern extent of the NQTB and was interpreted as a Precambrian continental fragment that was detached from the northwestern Tarim Block [54–56]. The basement of Oulongbuluke microcontinent was subdivided into three groups based on the ages, rock assemblages, and metamorphic grade. From oldest to youngest, they are the Delingha–Mohe gneiss complex, the Dakendaban Group, and the Wandonggou Group [54, 56]. The Delingha–Mohe gneiss complex yields zircon U–Pb ages of 2.48–2.24 Ga [54, 57, 58], the Dakendaban Group paragneiss has zircon U–Pb ages of ~2.20–1.95 Ga [54, 59], and the Wandonggou Group greenschist-facies volcanic-sedimentary rocks yield an Rb–Sr isochron age of c. 1022 Ma [60]. The Quanji Group is the dominant unit of the sedimentary cover and is separated from the basement by an angular unconformity; this group records the Late Neoproterozoic breakup of Rodinia [61]. The Oulongbuluke microcontinent was variably displaced and reworked by deformation, metamorphism, and magmatism during tectonic events from the Neoproterozoic to the early Paleozoic [21, 33, 62–66]. At least two generations of Phanerozoic mafic intrusions are found in the North Qaidam tectonic belt: early Paleozoic gabbro and diabase [34, 36, 52, 63] and Mesozoic gabbro [67, 68].

The North Wulan metamorphic complex is located in the northeastern part of the Oulongbuluke microcontinent (Figure 1(b)). It records multiple magmatic and metamorphic events in the Mesoproterozoic, early Neoproterozoic, early Paleozoic, and early Mesozoic [21, 33, 34, 55, 63–66, 68, 69]. Metamorphic P–T estimates and zircon geochronology from mafic and pelitic granulites reveal that the North Wulan metamorphic complex experienced early Paleozoic upper amphibolite to low-pressure granulite facies metamorphism with near-peak metamorphic conditions of ~800–900°C at 5.5–7 kbar from ~480 to 450 Ma [21, 64, 69]. Mafic rocks are widely distributed in the North Wulan metamorphic complex as plutons or dikes (Figure 1(c)). The OIB-like mafic rocks have intrusive ages of 490–485 Ma in the North Wulan metamorphic complex and were interpreted as the product of partial melting of primitive asthenosphere during the subduction of an oceanic slab [63]. E-MORB-like and arc-like mafic rocks with zircon U–Pb ages of 440–430 Ma were identified in the Chahanhe area from the center of the North Wulan metamorphic complex; these mafic rocks were interpreted to be generated from partial melting of asthenospheric mantle and preexisting enriched lithospheric mantle during the initial stage of continental subduction [34].

This study focuses on mafic rocks (gabbro and hornblende gabbro) exposed in the Shailekeguolai and Halihatu areas (Figure 2). Bodies of these mafic rocks generally strike NW-SE, which is roughly parallel to the orientation of the gneissosity in the surrounding rocks (Figures 2(a) and 2(b)). Mafic rocks in the Shailekeguolai area occur as dikes hundreds of meters in width and intrude into paragneiss (Figures 2(a) and 3(a)). In the Halihatu area, mafic rocks are separated from quartzite by thrust faults (Figures 2(b) and 3(c)) or intrude into garnet-bearing pelitic granulite and metamorphic sandstone as dikes hundreds of meters in width (Figures 2(b) and 3(e)). The mafic rocks are generally undeformed and have undergone minor alteration. Fourteen representative mafic rock (gabbro and hornblende gabbro) samples were collected for analysis. Mafic rocks have medium- to coarse-grained gabbroic textures with no evidence of significant deformation (Figures 3(b), 3(d), and 3(f)). They consist of clinopyroxene (35–40 vol%), plagioclase (40–45 vol%), and hornblende (15–20 vol%). Accessory minerals include zircon, apatite, magnetite, and ilmenite. Clinopyroxene is subhedral and can contain plagioclase inclusions (Figure 3(f)). Clinopyroxene phenocrysts can contain irregular margins, which is inferred to represent replacement by amphibole, epidote, and chlorite (Figure 3(d)). Plagioclase is subhedral to anhedral grain, has grain sizes of 0.5–2.0 mm, and is locally sericitized (Figures 3(b), 3(d), and 3(f)). Hornblende is subhedral to euhedral and locally contains plagioclase inclusions.

Eleven fresh mafic rock samples were crushed to 200 mesh using a jaw crusher. Whole-rock major and trace element analyses were performed using X-ray fluorescence (XRF) spectrometry on fused Li-borate glass beads with an Agilent 7500a inductively coupled plasma mass spectrometer (ICP-MS) after sample powder digestion by HNO₃ and HF in Teflon bombs at the State Key Laboratory of Continental Dynamics, Northwest University (Xi’an, China). The analytical precision of the XRF analyses for major elements is better than 2%. Standard sample BCR-2G and GBW07105W were measured to evaluate the standardization and performance stability during analysis. For the trace elements analyses, the analytical precision for REE, Zn, Ga, Rb, Co, Ni, Y, Zr, Hf, Nb, and Ta is better than ±5% and for others is better than ±5~10%. Two randomly selected repeat samples and the USGS standards (including BHVO-2, AGV-2, BCR-2, and GSP-2) were used to monitor analytical reproducibility and quality.

Whole-rock Sr and Nd isotopic analyses for six samples (14RLD113, 14RLD114, 15TL08, 15TL09, 15HT08, and 15HT09) were conducted using a Finnigan MAT-262 thermal ionization mass spectrometer (TIMS) at Laboratory for Radiogenic Isotope Geochemistry, University of Science and Technology of China. Chemical separation and purification were completed by conventional ion-exchange techniques. About 100–150 mg of sample powder was completely decomposed in Teflon beakers with a mixture of $HF+HClO4$⁠. Quartz columns equipped with a 5 ml resin bed of AG 50 W-X12 are used to separate Rb and Sr. Quartz columns with 1.7 ml Teflon powder are used to separate Sm and Nd. ¹⁴³Nd/¹⁴⁴Nd ratios and ⁸⁷Sr/⁸⁶Sr ratios were, respectively, normalized to $146Nd/144Nd=0.7219$ and $86Sr/88Sr=0.1194$⁠. Further details on column chromatography and analytical methods are given in Chen et al. [70].

Whole-rock Sr–Nd isotopic analyses for five additional samples (14RLD115, 14RLD116, 15TL10, 15HT10, and 15HT11) were conducted using a Thermo-Scientific Neptune MC-ICPMS at CAS Key Laboratory of Crust-Mantle Materials and Environments in University of Science and Technology of China. About 100–150 mg of sample powder was dissolved in steel-jacketed Teflon bombs with a mixture of $HF+HNO3+HClO4$⁠. The Eichrom DGA resin (50–100 μm, 2 ml) was used to separate Sr and Nd. Strontium was further purified by Sr-specific resin prior to measurement. Detailed analytical methods are found in Ma et al. [71].

Zircon was separated from crushed whole-rock samples using standard density and magnetic techniques and was further selected by handpicking under a binocular microscope. Zircon grains were mounted in epoxy resin mounts, polished, and imaged using a Tescan MIRA3 instrument equipped with a cathodoluminescence (CL) detector to determine the locations of U–Pb isotope and trace element analyses. Three mafic intrusions from the North Wulan metamorphic complex (Figure 2) were selected for zircon U–Pb isotope analysis at the LA-ICP-MS laboratory of the School of Resource and Environmental Engineering, Hefei University of Technology. The measurements of U–Pb isotope ratios were conducted by using the LA-ICP-MS system using an Agilent 7500a ICP-MS coupled with a COMPex PRO 102 ArF-Excimer laser source (⁠$λ=193nm$⁠). The analyses were performed using a laser energy of 80 mJ, a repetition rate of 6 Hz, and a spot beam diameter of 32 μm. Zircon 91500 was selected as an external calibration standard and was analyzed twice every five analyses. Plešovice was used as a secondary standard and was analyzed twice every ten analyses. Twenty-four analyses for Plešovice yielded a weighted ²⁰⁶Pb/²³⁸U age of $338.1±3.7Ma$ (⁠$MSWD=0.82$⁠), which is consistent with the recommended value of 337 Ma [72]. The zircon U–Pb isotope data were reduced using the program ICPMSDataCal 9.6 [73], and common Pb corrections were conducted with the method of Andersen [74]. Only zircon $analyses>90%$ concordant were used to calculate weighted mean ages. Isoplot v.3.23 was used to calculate concordia plots and weighted mean ²⁰⁶Pb/²³⁸U ages [75], and the uncertainties of weighted mean ages are reported at 95% confidence limits.

Zircon Lu–Hf isotope analysis was carried out at the Institute of Geology and Geophysics, Chinese Academy of Sciences in Beijing, using a Neptune multicollector inductively coupled plasma mass spectrometer (MC-ICPMS) in combination with a Geolas-193 laser-ablation system. Detailed instrument conditions and methods are described by Wu et al. [76]. The location of the zircon Lu–Hf isotope analysis within single zircon grains is the same as the location of zircon U–Pb analysis. Typically, ablation was conducted using a laser spot diameter of ca. 44 μm with 26 s ablation time, a 10 Hz repetition rate, and 10 J/cm² energy density. Reference zircons (mud tank, Plešovice, and Qinghu) were analyzed to monitor the instrument status and data quality. The weighted mean ¹⁷⁶Hf/¹⁷⁷Hf ratios of reference zircons mud tank and Plešovice are $0.282510±0.000003$ (⁠$2σ$⁠, $MSWD=0.73$⁠) and $0.282486±0.000004$ (⁠$2σ$⁠, $MSWD=0.77$⁠), respectively, which agrees with the values reported by Woodhead and Hergt [77] and Sláma et al. [72]. Reference zircon Qinghu yields weighted mean ¹⁷⁶Hf/¹⁷⁷Hf ratios of $0.282998±0.000004$ (⁠$2σ$⁠, $MSWD=0.86$⁠), which is similar to the value recommended by Li et al. [78]. The ¹⁷⁶Lu decay constant of $1.865×10−11/year$ recommended by Scherer et al. [79] is employed in this study, and initial $εHft$ values are calculated based on the chondritic reservoir values of $176Hf/177Hf=0.282772$ and $176Lu/177Hf=0.0332$ [80]. Single-stage Hf model ages (⁠$TDM1$⁠) are calculated relative to the depleted mantle with present-day ¹⁷⁶Hf/¹⁷⁷Hf of 0.28325 and ¹⁷⁶Lu/¹⁷⁷Hf of 0.0384 [81]. Two-stage Hf model ages (⁠$TDM2$⁠) are calculated relative to the average continental crust with an assumed mean ¹⁷⁶Hf/¹⁷⁷Hf value of 0.015 [82].

The whole-rock major and trace element compositions of eleven mafic rock samples from the North Wulan metamorphic complex are listed in Electronic Appendix 1. Major elements are normalized to volatile-free values before plotting. These mafic rock samples contain Na₂O and K₂O contents ranging from 1.64 to 3.09 wt% and from 1.27 to 2.17 wt% and are dominantly classified as gabbro on a $K2O+Na2O$ versus SiO₂ diagram and belong to the subalkaline series (Figure 4(a)). They have sodic compositions on a K₂O versus Na₂O diagram (Figure 4(b)). These mafic rocks have SiO₂ contents of 46.75–50.72 wt%, CaO contents of 8.14–9.99 wt%, Al₂O₃ contents of 13.04–17.02 wt%, MgO contents of 5.89–8.30 wt%, and P₂O₅ contents of 0.12–0.29 wt% (Figure 5). The CaO, Na₂O, SiO₂, Al₂O₃, and CaO/Al₂O₃ contents show no clear correlation with increasing MgO contents (Figures 5(a)–5(d) and 5(g)). The Fe₂O₃^T and TiO₂ contents of these mafic rocks are variable (8.18–15.2 wt% for Fe₂O₃^T and 0.71–1.73 wt% for TiO₂) and generally decrease with increasing MgO (Figures 5(e) and 5(f)). They have variable concentrations of Cr (19.1–51.2 ppm) and Ni (18.4–42.2 ppm). Mg^# values range from 43.5 to 66.1. Their chondrite-normalized rare earth element (REE) patterns are highly fractionated (⁠$La/YbN=7.55$–10.1 and $Dy/YbN=1.22$–1.27); they are enriched in the LREE and have minor negative Eu anomalies (⁠$Eu/Eu∗=0.79$–0.95) (Figure 6(a)). Primitive mantle-normalized trace element patterns show strong enrichment in Pb, Nd, and the large ion lithophile elements (LILEs), such as Cs, Rb, K, and Ba; they are depleted in P and high field-strength elements (HFSEs), such as Nb, Ta, Ti, Zr, and Hf (Figure 6(b)).

LA-ICP-MS zircon U–Pb geochronology was performed for three mafic rock samples, and the zircon U–Pb isotopic data are presented in Electronic Appendix 2. Representative CL images and zircon U–Pb concordia diagrams are illustrated in Figure 7. Zircon from the three samples is subhedral to euhedral, prismatic, and 100–350 μm in size and has length/width ratios of 1 : 1 to 3 : 1. Zircon shows oscillatory or planar zoning in CL without residual or inherited cores (Figures 7(a), 7(c), and 7(e)). Zircon in samples 15TL07 and 15HT07 has broader oscillatory zoning in CL than zircon in sample 14RLD112 (Figures 7(a), 7(c), and 7(e)). Twenty-five analyses of zircons from sample 15TL07 yield a range of ²⁰⁶Pb/²³⁸U age from 489 to 451 Ma with Th/U ratios of 0.09–0.97 and give a weighted mean age of $472±4.9Ma$ (⁠$MSWD=0.81$⁠) (Figure 7(b)). Twenty-eight analyses of zircon from sample 15HT07 yield high Th/U ratios of 0.79–1.4 and ²⁰⁶Pb/²³⁸U ages ranging from 497 to 466 Ma with a weighted mean age of $480±4.8Ma$ (⁠$MSWD=0.31$⁠) (Figure 7(d)). Thirty analyses of zircon from sample 14RLD112 show relatively high Th/U ratios of 0.58–1.17 and yield ²⁰⁶Pb/²³⁸U ages varying from 505 to 464 Ma with a weighted mean age of $483.2±4.6Ma$ (⁠$MSWD=0.43$⁠) (Figure 7(f)).

Zircon Lu–Hf isotopic data for the studied mafic rock samples are reported in Electronic Appendix 3 and are shown in Figure 8. Twenty-one Lu–Hf analyses from zircon grains in sample 15TL07 yield ¹⁷⁶Lu/¹⁷⁷Hf ratios of 0.000221–0.003496 (Figure 8(a)) and ¹⁷⁶Hf/¹⁷⁷Hf ratios of 0.282433–0.282585. At an assumed age of 472 Ma, the initial ¹⁷⁶Hf/¹⁷⁷Hf ratios are 0.282422 to 0.282557 and $εHft$ values are -2.0 to +2.8 (Figures 8(a)–8(c)), corresponding to single-stage Hf model ages (⁠$TDM1$⁠) of 1.2–1.0 Ga and two-stage Hf model ages (⁠$TDM2$⁠) of 1.6–1.3 Ma.

Twenty-two analyses on zircons from sample 15HT07 show ¹⁷⁶Lu/¹⁷⁷Hf ratios and ¹⁷⁶Hf/¹⁷⁷Hf ratios ranging from 0.000491 to 0.001869 and from 0.282252 to 0.282323, respectively, with initial ¹⁷⁶Hf/¹⁷⁷Hf ratios of 0.282246–0.282318. At 480 Ma, these ratios yield $εHft$ values of -8.0 to -5.5 (Figures 8(a)–8(c)). $TDM1$ model ages vary from 1.4 to 1.3 Ga, and $TDM2$ model ages are 2.0–1.8 Ga.

Twenty analyses on zircon grains from sample 14RLD112 yield ¹⁷⁶Lu/¹⁷⁷Hf ratios of 0.000386–0.001258 (Figure 8(a)) and ¹⁷⁶Hf/¹⁷⁷Hf ratios of 0.282426–0.282502. At 483 Ma, the initial ¹⁷⁶Hf/¹⁷⁷Hf ratios vary from 0.282419 to 0.282497 and $εHft$ values range from -1.9 to +0.9 (Figures 8(a)–8(c)), corresponding to $TDM1$ model ages of 1.2–1.0 Ga and $TDM2$ model ages of 1.6–1.4 Ga.

Whole-rock Sr–Nd isotope compositions of mafic rocks from the North Wulan metamorphic complex are given in Electronic Appendix 4 and summarized in Figure 9. The isotopic data for garnet peridotite that originated from SCLM [83] and eclogites with oceanic affinities [30, 47] in the NQTB are also plotted in Figure 9 for comparison. Initial ⁸⁷Sr/⁸⁶Sr and $εNdt$ values were calculated at their individual weighted mean age of zircon U–Pb dating. These mafic rock samples have high initial ⁸⁷Sr/⁸⁶Sr ratios of 0.710363 to 0.719404 and negative $εNdt$ values of −7.77 to −2.30, corresponding to large variations in one-stage model ages (1.9−1.3 Ga).

The North Wulan metamorphic complex experienced early Paleozoic high dT/dP metamorphism and was interpreted to be in the upper plate of an inferred subduction system in the NQTB [21, 64]. Zircon in mafic rock samples 15TL07 and 15HT07 from the North Wulan metamorphic complex is subhedral to euhedral and prismatic; it has broad oscillatory or planar zoning in cathodoluminescence (Figures 7(a) and 7(c)) and relatively high Th/U ratios of 0.09–1.4 (Appendix 2), which are characteristic of magmatic zircon [84]. Zircon in mafic rock sample 14RLD112 is euhedral and prismatic, contains relatively narrow oscillatory zoning in cathodoluminescence (Figure 7(e)), and has high Th/U ratios of 0.58–1.17 (Appendix 2). Although growth banding of zircon from mafic rocks is generally broader than that of zircon from granites, zircon from mafic rocks can have fine oscillatory zonation in CL due to a lower contrast in U content between the zones [85]. Therefore, zircon in sample 14RLD112 is also interpreted to have crystallized from mafic magma. Zircon U–Pb geochronology of mafic rock samples yields ²⁰⁶Pb/²³⁸U weighted mean ages of $472±5Ma$⁠, $480±5Ma$⁠, and $484±5Ma$⁠, respectively (Figures 7(b), 7(d), and 7(f)), which are interpreted as crystallization ages. These are early Ordovician ages and indicate that mafic magmas were emplaced in the upper plate of an early Paleozoic subduction system. Regionally, this period of mafic magmatism was also identified in the Wanggaxiu gabbro complex (468 Ma, [36]) and in the Hudesheng mafic-ultramafic intrusions (465–455 Ma, [35]). These crystallization ages are older than the timing of metamorphism of oceanic-type and continental-type eclogites from the HP-UHP metamorphic complexes in the NQTB (ca. 460–420 Ma, [22, 27, 28, 30, 31, 41]) but are consistent with the timing of high-temperature low-pressure metamorphism of granulites found in the upper plate of the former oceanic subduction zone [21, 64, 69]. Thus, these mafic intrusions from the North Wulan metamorphic complex represent syn-subduction mafic magmatism during subduction of the Proto-Tethyan oceanic lithosphere.

Before characterizing the origin and nature of the magma source, the effect of fractional crystallization, crustal contamination, and magma mixing on magma compositions needs to be evaluated. Mafic rocks from the North Wulan metamorphic complex have lower SiO₂ (46.75–50.72 wt%), Cr (19.1–50 ppm), and Ni (18.4–42.2 ppm) and relatively lower MgO contents (5.89–8.30 wt%) than those of primary magmas in equilibrium with mantle peridotite [86]; this suggests that mafic magmas from the North Wulan metamorphic complex may have undergone some fractional crystallization during magma ascent and emplacement. Mafic rocks from the North Wulan metamorphic complex show increasing Cr with increasing MgO (Figure 10(a)), which may reflect the process of fractional crystallization of clinopyroxene and/or olivine. However, there is no systematic covariation between CaO/Al₂O₃ ratios, Ni and SiO₂ concentrations, and MgO contents (Figures 5(a), 5(b), and 10(b)), suggesting that fractional crystallization of clinopyroxene and/or olivine was minimal. Slightly negative Eu anomalies (Figure 6(a)) may reflect minor fractionation of plagioclase, but no Sr anomalies are observed in primitive mantle-normalized trace element patterns, and there is no covariation between MgO and Eu/Eu$∗$ (Figure 10(c)). No covariation of Al₂O₃ and MgO (Figure 5(d)) also indicates that there was minimal fractional crystallization of plagioclase. The mafic rocks have negative correlations between TiO₂, $FeOT$⁠, and MgO (Figures 5(e) and 5(f)), suggesting crystal accumulation/fractional crystallization of Fe-Ti-bearing minerals (such as magnetite and/or ilmenite). Thus, these characteristics suggest that the fractionation of Fe–Ti oxides rather than plagioclase, olivine, and clinopyroxene controlled magma compositions of the studied mafic intrusions. However, most major elements do not covary with MgO (Figure 5), and the mafic rocks have La/Sm versus La trends that are more consistent with partial melting rather than fractional crystallization (Figure 10(d)). Therefore, the effects of fractional crystallization on magma compositions were probably negligible.

Mafic rocks in the North Wulan metamorphic complex may have been contaminated by continental crust during migration and emplacement. Continental crust generally has high SiO₂ and K₂O contents, radiogenic ⁸⁷Sr/⁸⁶Sr ratios, nonradiogenic ¹⁴³Nd/¹⁴⁴Nd isotope ratios, low Ti, Nb, and Ta concentrations, and high Pb, Zr, and Hf concentrations [87]. The whole-rock initial ⁸⁷Sr/⁸⁶Sr ratios are expected to increase, and the $εNdt$ values are expected to decrease with increasing SiO₂ contents and K₂O contents if crustal contamination occurs. However, $87Sr/86Sri$ ratios show no correlation with SiO₂ and K₂O (Figures 11(a) and 11(c)), and $εNdt$ is positively correlated with SiO₂ and K₂O (Figures 11(b) and 11(d)); such trends are not consistent with crustal contamination. The Nb/U ratios remain constant irrespective of SiO₂ content (Figure 11(e)), which is also inconsistent with crustal contamination. The mafic rock samples show positive correlations between SiO₂ and Ce/Pb (Figure 11(f)), which is also not consistent with the crustal contamination. In addition, no inherited zircon cores or zircon xenocrysts were observed in these samples, which suggests that continental crustal contamination was insignificant in the petrogenesis of mafic rocks in the North Wulan metamorphic complex. However, considering that the mafic rocks have high initial ⁸⁷Sr/⁸⁶Sr ratios of 0.710363 to 0.719404, low $εNdt$ values of -7.77 to -2.30, and variable zircon $εHft$ values ranging from -8 to +2.8, we cannot completely exclude the possibility of crustal contamination, although we have no other evidence to support significant crustal contamination.

Mixing of mantle-derived magmas with crustal-derived magmas during transport and emplacement may also change the composition of primary mafic magmas. Magma mixing commonly results in field evidence of this process, such as enclaves of various rock types (e.g., [82]), but this has not been observed in mafic rocks from the North Wulan metamorphic complex (Figures 3(a), 3(c), and 3(e)). There are also no mineral disequilibrium textures that typically accompany magma mixing, such as rapakivi textures, complex zoning of plagioclase, and corrosion microstructures [88–90]. Zircon core-rim structures can also be suggestive of magma mixing (e.g., [91]), but such features are not observed in zircon from mafic rocks from the North Wulan metamorphic complex (Figures 7(a), 7(c), and 7(e)). Furthermore, there is no significant variation in the zircon Hf isotopic composition for each sample (Figure 8), which also does not support magma mixing and/or crustal contamination.

In summary, although the mafic rocks may have undergone minor crystal accumulation/fractional crystallization of Fe–Ti oxides, fractional crystallization of other minerals appears to have been negligible and would have had only a minor effect on trace element compositions of the magma. Furthermore, there is no evidence of substantial crustal assimilation and magma mixing. Therefore, we interpret the trace element concentrations and isotope ratios of mafic rocks in the North Wulan metamorphic complex to reflect the compositions of their mantle sources rather than significant crustal assimilation and fractional crystallization.

Mafic igneous rocks are crucial windows into the nature of their mantle sources. The mafic rocks from the North Wulan metamorphic complex studied here have high MgO (5.89–8.30 wt%), Mg^# values of 43.5 to 66.1, and low SiO₂ (46.75–50.72 wt%) contents (Figures 5(b) and 5(h)), suggesting that they were derived from partial melting of ultramafic mantle sources [92]. Partial melting of normal depleted asthenospheric mantle is expected to form normal middle-ocean-ridge basalt (N-MORB) characterized by the depletion of LILE and LREE as well as radiogenic Nd isotope ratios and nonradiogenic Sr isotope ratios [93]. Mafic rocks in the North Wulan metamorphic complex show enrichment in the LILE and LREE (Figures 6(a) and 6(b)), which suggests that their mantle source was modified by the LILE-enriched and LREE-enriched materials. The high initial ⁸⁷Sr/⁸⁶Sr ratios of 0.710363 to 0.719404, negative $εNdt$ values of −7.77 to −2.30, and variable zircon $εHft$ values of -8 to +2.8 for our mafic rock samples (Figure 9) also suggest the involvement of the crustal components in their mantle sources. Therefore, the enriched LILE, LREE, and Sr–Nd–Hf isotopes in mafic rocks from the North Wulan metamorphic complex are inferred to represent the involvement of the crustal components into their mantle sources.

The involvement of oceanic crustal carbonatite into mantle-wedge peridotite would result in the enrichment of most incompatible trace elements in the mantle source and the relative depletion of K, Zr, Hf, and Ti [94, 95]; these features are inconsistent with the trace element characteristics of mafic rocks in the North Wulan metamorphic complex (Figure 6(b)). Subduction of an oceanic slab could imprint crustal signatures into mantle sources through slab-derived felsic melt/fluid-peridotite reaction at the slab-mantle interface in the subduction channel, which would generate ultramafic metasomatites enriched in the LILE and LREE but depleted in the HFSE (Nb, Ta, Ti, and Zr). Generally, subduction zone fluids are regarded as the crucial media to transfer arc-like geochemical signatures from subducting oceanic slab to the mantle wedge [10, 96, 97]. The abundance of hornblende—inferred to be magmatic and not metamorphic—in mafic rocks from the North Wulan metamorphic complex (Figures 3(b) and 3(d)) suggests that hydrous fluids were involved in their petrogenesis. Thus, the arc-like trace element features for mafic rocks in the North Wulan metamorphic complex can be explained as the addition of subducted oceanic slab-derived hydrous fluids into their mantle source.

Subducted materials generally consist of the oceanic lithosphere and overlying sediments, and both of these can contribute material to the overlying mantle wedge. Altered basaltic oceanic crust (AOC) generally has a wide range of ⁸⁷Sr/⁸⁶Sr ratios [98]; thus, the wide range of initial ⁸⁷Sr/⁸⁶Sr ratios (0.710363 to 0.719404) in mafic rocks from the North Wulan metamorphic complex (Figure 9) suggests that altered oceanic crust-derived fluids and/or melts may have been incorporated into their mantle source. However, subducted basaltic oceanic crust-derived fluids and/melt cannot lead to the significant enrichment of Nd–Hf isotopes of their mantle source because oceanic crust is generally derived from depleted mantle [93]. Thus, fluids and melt generated from basaltic oceanic crust cannot create a mantle source with Nd–Hf geochemical features that are substantially different from the depleted mantle. The subduction of oceanic sediment into the mantle can also produce arc-like trace element features of mafic rocks [99, 100]. In addition, the relatively enriched radiogenic Sr–Nd–Hf isotopic compositions, Proterozoic Nd–Hf model ages (Figures 8 and 9; Appendices 3 and 4), and positive Pb anomalies (Figure 6(b)) also support the contribution of oceanic sediment materials with ancient terrestrial origins into their mantle sources. Ancient crust-derived rocks with radiogenic Sr and nonradiogenic Nd–Hf isotope signatures on continental margins can be weathered and transported into the subduction trench and further transported into the mantle during subduction [11, 101]. Partial melting of this ancient terrigenous detritus could have generated hydrous felsic melts with enriched radiogenic isotopic signatures, which are subsequently transported into the mantle via melt-peridotite reactions [11, 102]. Therefore, the addition of seafloor sediments with ancient terrestrial origin to the mantle source can explain the arc-like trace element features as well as the relatively radiogenic Sr isotope and nonradiogenic Nd–Hf isotope compositions of mafic rocks in the North Wulan metamorphic complex.

Mafic rocks with arc-like trace element signatures in the North Wulan metamorphic complex exhibit similar Nd isotopic compositions but more radiogenic Sr isotopic compositions compared to the SCLM in the North Qaidam tectonic belt (Figure 9); we interpret these differences to reflect the incorporation of seafloor sediments that originated from the weathering of the ancient continental crust into their mantle sources during the early Paleozoic subduction of the Proto-Tethyan oceanic lithosphere. These mafic rocks have apparent high Ba/Th but low $La/YbN$ and $La/SmN$ ratios (Figures 12(a) and 12(b)), which indicates that they were derived from a mantle source that interacted with the subducted slab-derived fluid. However, these rocks also show low U/Th and Sr/La ratios but high Th concentrations and Th/Yb ratios (Figures 12(c) and 12(d)), which suggests the involvement of sediment-derived melt into the mantle wedge. The high Th/Nb and Th/Yb, and high Ba/Th and Ba/La ratios (Figures 12(e) and 12(f)) also support the joint involvement of slab-derived fluid and sediment-derived melt into their mantle source. Therefore, both slab-derived fluid and sediment-derived melt could have contributed to the mantle source of mafic rocks in the North Wulan metamorphic complex.

Arc-like geochemical signatures of mafic rocks can originate from either the newly metasomatized asthenospheric mantle [103, 104] or ancient lithospheric mantle [18, 96, 97, 105, 106] in the absence of significant crustal contamination. Petrological forward modeling demonstrates that most of the release of fluids from the subducted AOC typically occurs at lithospheric depths; thus, the overlying lithospheric mantle would be metasomatized [96]. The high initial ⁸⁷Sr/⁸⁶Sr ratios of 0.710363 to 0.719404 and negative $εNdt$ values of −7.77 to −2.30 of the mafic rocks from the North Wulan metamorphic complex indicate that the mantle source was enriched; thus, these mafic rocks most likely originated from an ancient lithospheric source. The negative zircon $εHft$ values (-8 to +2.8) of the mafic rocks (Figure 8) suggest that they were derived from the enriched lithospheric mantle rather than depleted asthenosphere mantle. The older single-stage Hf model ages (1.4–1.0 Ga, Appendix 3) also suggest that the subcontinental lithosphere mantle beneath the Oulongbuluke microcontinent was metasomatized by subducting slab-derived fluids and sediment-derived melts. Previous studies suggested that the North Qaidam area experienced multiple Proterozoic mantle metasomatism events during the evolution of the Columbia and Rodinia supercontinents [66, 83], which resulted in fertilization and enrichment of the lithospheric mantle. At least two enrichment events—one during the Proterozoic and one during the early Paleozoic—may have generated the enriched isotopic composition of the SCLM in the North Qaidam area. The variations of zircon Hf isotopes in the mafic rock imply that their magma source was relatively heterogeneous in terms of Hf isotopes (Figures 8(a)–8(c)). These geochemical characteristics can be attributed to the heterogeneous mantle source caused by variable (i.e., patchy) metasomatism with subducting slab-derived fluids and sediment-derived melts.

Garnet peridotites in the NQTB originated from the SCLM and represent the components of the mantle wedge beneath the early Paleozoic continental arc [39, 46, 83]. These orogenic peridotites have arc-like trace element signatures and enriched Sr–Nd isotope values [46], which were interpreted to result from fluid-peridotite interaction in an early Paleozoic subduction channel [39]. This crust-mantle interaction incorporated the subducting oceanic crust-derived geochemical signatures into the wedge peridotite; subsequent partial melting of metasomatized wedge peridotite (pyroxenite and/or hornblendite) transferred these features to continental arc mafic magmas. Therefore, these orogenic peridotites could be compatible mantle sources for arc-like mafic rocks in the North Wulan metamorphic complex.

The low Nb/U and TiO₂/Al₂O₃ ratios (Figure 13(a)) of mafic rocks from the North Wulan complex suggest that their source was affected by dehydration melting of the subducted oceanic terrigenous metasediment within the rutile stability field [107]. Depletion of Nb and Ta in the arc-like mafic rocks (Figure 6(b)) also indicates that rutile was stable and retained Nb, Ta, and Ti in the source during the partial melting of the subducted oceanic slab at subarc depths [107, 108]. Mantle metasomatism usually takes place through fluid/melt-peridotite reactions and produces mantle metasomatites that can contain hornblende, clinopyroxene, garnet, and phlogopite [109–111]. Mafic rocks from the North Wulan metamorphic complex have $10000×Zn/Fe$ values of 8–12 (Appendix 1), which is slightly higher than those of peridotite (6–11, [112]), suggesting that they are derived from a pyroxenite-bearing mantle source. The mafic rocks also have Fe/Mn ratios of 52–98 (Appendix 1), which is higher than those of peridotite that have Fe/Mn ratios of 50–60 [113]. Thus, the mafic rocks from the North Wulan complex were most likely derived from a pyroxenite-bearing mantle source. In addition, the mafic rocks have low Ba/Rb of 6.35–12.87 and high Rb/Sr of 0.12–0.20 (Figure 13(b)), which suggests that they are derived from a phlogopite-bearing mantle source rather than hornblende-bearing mantle sources (e.g., [114]). Partial melting of peridotite in the spinel versus garnet stability fields is generally pressure dependent and can be distinguished using Sm/Yb–La/Sm and Dy/Yb–La/Yb plots (Figures 13(c) and 13(d)) [115, 116]. Melting in the spinel stability field is expected to yield relatively low Sm/Yb and Dy/Yb values, which are consistent with those of mafic rocks in the North Wulan metamorphic complex (Figures 13(c) and 13(d)). Thus, mafic rocks were derived from the partial melting of a phlogopite-bearing and pyroxenite-bearing mantle source in the spinel stability field.

In summary, mafic rocks from the North Wulan metamorphic complex originated from partial melting of an enriched mantle source in the SCLM of the Oulongbuluke microcontinent. The mafic rocks have relatively enriched Sr–Nd isotope values and elevated concentrations of the LILE and the LREE; this suggests a crustal contribution to their mantle sources. The mantle source was enriched and metasomatized by the input of both subducted slab-derived (i.e., AOC and overlying sediments) fluid and sediment-derived melt. Based on trace element characteristics, the mantle source probably contained phlogopite, pyroxene, and spinel.

Mafic magmas form in distinct tectonic settings from variably modified mantle compositions; these variables can be deciphered using the trace element and isotope ratios of mafic rocks. Arc magmatism above subduction zones is relatively rare during continental subduction [117, 118] due to the low thermal gradients of continental subduction zones and/or the absence of aqueous fluid released from the subducted continental slab [117, 119, 120]. By contrast, syn-subduction arc magmatism during subduction of an oceanic slab is common at convergent plate boundaries [11]. Mafic rock samples in the North Wulan metamorphic complex show enrichment in the LILEs (Cs, Rb, K, and Ba) and depletion in HFSEs (Nb, Ta, Ti, and Zr) with pronounced negative Nb–Ta and Zr–Hf–Ti anomalies (Figure 6(b)). These trace element compositions suggest that the mafic magmas formed in the oceanic subduction-related settings. The Zr/Nb ratios vary from 8.4 to 12.8 and Nb/La from 0.31 to 0.39, which is similar to those of arc magmas [121, 122]. The mafic rocks have Nb/La ratios of 0.31–0.39, Nb/U ratios of 6.21–14.47, and Nb concentrations of 7.39–15.90 ppm, which are also similar to those of island arc basalts [123]. On a Th–Hf/3–Ta ternary diagram, the compositions of our mafic rock samples are plotted into the field of continental arc basalt (Figure 14(a)). Mafic rocks also have Th/Yb and Nb/Yb ratios that are similar to those of continental arc basalts (Figure 14(b)). Thus, the trace element and isotopic compositions of mafic rocks in the North Wulan metamorphic complex suggest a continental arc setting for their petrogenesis.

A continental arc setting for mafic rocks in the north Wulan metamorphic complex is also supported by eclogites and associated HP-UHP gneiss with ages of metamorphism between 460 and 420 Ma in the NQTB [25, 27–29, 31, 40–42, 47, 48]; these eclogites formed during oceanic subduction to continental collision/subduction. HP-UHP metamorphic rocks in the NQTB underwent peak metamorphism at 460–420 Ma; of these rocks, some oceanic-type eclogites with MORB chemical signatures (metamorphosed ophiolite) have metamorphic ages of c. 460–440 Ma [22, 28–30, 124]. Metamorphism of oceanic-type eclogites was slightly later than the period of mafic magmatism in the North Wulan metamorphic complex. Thus, this mafic magmatism could represent syn-subduction arc magmatism. In addition, the ancient crustal basement of the North Wulan metamorphic complex mainly consists of Paleoproterozoic to Neoproterozoic paragneiss, granitic gneiss, marble, and schist with minor amphibolite [55, 65, 66]; mafic rocks intruded into this ancient crustal basement of the North Wulan metamorphic complex. These features suggest that these arc magmas were generated in a continental arc setting, but not an oceanic island arc setting.

Early Paleozoic subduction-related mafic magmas were also reported to the south of the North Wulan metamorphic complex [35, 36]. These include the Hudeshen mafic-ultramafic rocks (465–455 Ma) that intrude the Oulongbuluke microcontinent and show Alaskan-type mafic complex affinities; these were interpreted to originate from the partial melting of hydrous mantle metasomatized by hydrous fluids in a continental arc setting [35]. The Wanggaxiu gabbro complex with an age of 468 Ma in the Dulan area is characterized by enrichment in LILE and depletion in HFSE, which are interpreted to be derived from subduction-metasomatized mantle wedge [36]. The 473–465 Ma granitic intrusions in the NQTB are also considered the products of arc-related magmatism in an active continental margin setting [125]. Thus, a continental arc is inferred to have existed along the northern margin of the Qaidam block during early Paleozoic subduction of the Proto-Tethyan oceanic lithosphere (Figure 15). Although andesite is considered to be the dominant rock in continental arc settings, mafic igneous rocks with arc-like geochemical signatures in continental arc settings have also been widely reported in many orogenic belts, such as the Famatinian arc in the Sierra de Valle Fértil (southern segment of the Andes) [126], the Gangdese magmatic arc in southern Tibet [127], and Fiordland arc in New Zealand [128].

Mantle-derived mafic magmas generated by hydrous partial melting of lithospheric mantle in the continental arc played a crucial role in crustal growth of the NQTB. Thus, lithospheric mantle-derived mafic magmas may have represented a significant contribution to the growth of juvenile continental crust. In addition, the protoliths of migmatitic amphibole-biotite gneiss in the North Wulan metamorphic complex have continental arc affinities and yield crystallization ages of 506–494 Ma; these rocks record metamorphism and anatexis at ca. 465 to 450 Ma, which reflects the reworking of former arc mafic rocks during long-lived oceanic subduction [33]. Thus, continental arcs are key locations on earth for the growth and maturation of continental crust.

Mafic rocks with arc-like geochemical signatures in the North Wulan metamorphic complex originated from a subcontinental lithospheric mantle source with the involvement of melt and fluids derived from subducted basaltic oceanic crust and metasedimentary rocks. Previous studies suggest that the North Wulan metamorphic complex underwent LP-HT amphibolite- to granulite-facies metamorphism with peak metamorphic conditions of ~800–900°C at 5.5–7 kbar at 474–446 Ma [21, 64]. Zircon U–Pb geochronology shows that the crystallization ages of 483–472 Ma for these continental arc mafic rocks (Figure 7) are similar to the early stages of LP-HT peak metamorphism (474–446 Ma) in the North Wulan metamorphic complex. Thus, mafic magmatism and LP-HT granulite-facies metamorphism reflect a conjugate arc-back-arc system that operated during the early Paleozoic northward subduction of the Proto-Tethyan oceanic lithosphere. Therefore, the timing of Proto-Tethyan oceanic subduction beneath the Oulongbuluke microcontinent should be earlier than 483 Ma.

There are no observed chilled margins on the mafic rocks in the North Wulan metamorphic complex or contact metamorphic aureoles in the country rocks (LP-HT granulites); this suggests that metamorphism of the granulites and emplacement of the mafic magmas were at similar temperatures. Thus, the North Wulan metamorphic complex represents an exposure of deep continental arc-back-arc crust. OIB-like mafic rocks with intrusive ages of 490–485 Ma are identified in the North Wulan metamorphic complex and are considered the product of partial melting of the primitive asthenosphere without melt extraction [63]. These two types of early Paleozoic mafic rocks with different affinities in the North Wulan metamorphic complex indicate substantial mantle heterogeneity beneath the Oulongbuluke microcontinent.

The Tanjianshan arc complex in the southern part of the NQTB was interpreted to have formed in an intraoceanic arc setting (Figure 15, [20]). These intraoceanic arc magmas are distinctly different from the continental arc magmatism reported in this study (Figure 15). This suggests that the NQTB witnessed a collage of continental arc constructed on the Oulongbuluke microcontinent and intraoceanic arcs during the closure of the Proto-Tethyan Ocean.

The Proto-Tethyan Ocean in the NQTB separated the Oulongbuluke microcontinent from the Qaidam Block in the early Paleozoic. Its closure resulted in accretion and collage of intraoceanic arc and finally continental collision between the Oulongbuluke microcontinent and Qaidam Block [20, 21, 27, 34, 37, 129]. Previous studies of the Oulongbuluke microcontinent have mostly focused on the affinities with surrounding continental terranes and its relationships with the assembly and rifting of the Rodinia and Columbia supercontinents [54–56, 58, 59, 65]. However, the role of the Oulongbuluke microcontinent in the process of early Paleozoic North Qaidam orogenesis is still unclear. The identification of continental arc mafic rocks in this study and LP–HT metamorphic rocks (474–446 Ma, [21]) in the North Wulan metamorphic complex indicates that the Oulongbuluke microcontinent participated in the early Paleozoic orogenesis as the basement of a continental arc. This suggests that early Paleozoic subduction of oceanic crust occurred in the North Qaidam prior to continental subduction and had a northward subduction polarity. Huang et al. [32] proposed that continental collision in the North Qaidam occurred at ~500 Ma based on the presence of syn-collisional granitoids in the Qilian Block. However, the presence of 483–455 Ma mafic rocks in a continental arc setting ([35, 36]; this study) indicates that there was indeed an early Paleozoic ocean in the North Qaidam area, and this ocean did not close completely before 455 Ma, which further suggests that the timing of continental collision/subduction in the NQTB should not be earlier than 455 Ma. Continental arc magmatism in the North Wulan metamorphic complex in the north, the intraoceanic arc-trench system represented by the Tanjianshan Group in the middle, and the HP-UHP metamorphic complex in the south together record the whole evolution of the North Qaidam tectonic belt. Thus, the NQTB is an independent orogenic belt and records the orogenic process from ocean opening, oceanic crust subduction, and basin closure to terminal continental collision/subduction.

A schematic evolution diagram of the NQTB with the role of continental arc mafic magmatism is shown in Figure 15. The Proto-Tethyan oceanic crust was subjected to seawater alteration in the oceanic basin, resulting in changes in its Sr isotope composition (Figure 15(a)). With the northward subduction of the Proto-Tethyan oceanic lithosphere, the subducted terrigenous sediment on the ocean floor suffered dehydration at shallow forearc depths, resulting in the loss of some fluid-mobile elements. The Proto-Tethyan oceanic plate was further subducted to greater subarc depths, and then, the basaltic oceanic crust underwent dehydration and the (meta) sedimentary rocks partially melted (Figure 15(a)). The fluids derived from dehydration of altered oceanic basalts and melt from subducted metasedimentary rocks migrated upward and variably metasomatized the overlying subarc mantle to form the mantle metasomatites and resulted in the input of the crustal components into the subarc mantle. This metasomatism process led to the compositional heterogeneity of the mantle wedge beneath the Oulongbuluke microcontinent. The ongoing subduction of the Proto-Tethyan oceanic lithosphere beneath the Oulongbuluke microcontinent leads to oceanic slab retreat and rollback (Figure 15(b)). The heat from asthenosphere upwelling caused partial melting of metasomatized subcontinental lithospheric mantle (metasomatites) to form the mantle-derived continental arc magma in the North Wulan metamorphic complex (Figure 15(b)). At the same time, high temperatures in middle to lower crust of this continental arc-back-arc setting generated upper amphibolite-facies to granulite-facies metamorphism in the North Wulan metamorphic complex [21, 64, 69]. As the ocean basin closed and continental subduction initiated, intraoceanic arcs and continental arcs were welded together, and some continental or oceanic components that were subducted to the mantle depth were exhumed to the surface as HP-UHP metamorphic rocks (Figure 15(c)). In summary, mafic rocks in the North Wulan metamorphic complex witnessed crust-mantle interaction in an oceanic subduction channel and also recorded the subduction of Proto-Tethyan oceanic crust.

• (1)

  Mafic rocks in the North Wulan metamorphic complex were emplaced during 483–472 Ma, which is nearly coeval with the metamorphism of the surrounding LP-HT metamorphic rocks (480–450 Ma) and is earlier than HP-UHP eclogite-facies metamorphism (460–420 Ma).

• (2)

  The mafic rocks are characterized by geochemical signatures similar to arc-related rocks with enrichment of the LILE, depletion of the HFSE, radiogenic Sr isotope values, and nonradiogenic Nd–Hf ratios; these are all features consistent with continental arc basalt affinities.

• (3)

  These mafic magmas were derived from the SCLM of the Oulongbuluke microcontinent that was metasomatized by the altered oceanic crust-derived fluids and/or sediment-derived melts in different proportions. The mantle contained a phlogopite and pyroxene mineral assemblage and melted in the spinel stability field.

• (4)

  Based on our new results and those from previous studies, we conclude that the retreat and rollback of the Proto-Tethyan oceanic slab triggered asthenosphere upwelling and the partial melting of metasomatized lithospheric mantle to form the mafic magmas. The North Wulan metamorphic complex records early Paleozoic tectonic evolution of a continental arc-back-arc system and the subduction of the Proto-Tethyan oceanic crust.

All data supporting the results of this study can be found in the figures and supplementary material submitted to the journal.

The authors declare that they have no conflicts of interest.

This study was financially cosupported by the National Natural Science Foundation of China (41902235, 41272221), the Natural Science Foundation of Anhui Province, China (2008085QD172), and the Fundamental Research Funds for the Central Universities (JZ2021HGTB0109, JZ2020HGQA0194). We thank Yu Han, Liangchao Da, and Qiqi Zhao for their assistance with field work, Dr. Quanzhong Li for LA-ICP-MS zircon U–Pb dating, Dr. Yueheng Yang for their help with zircon Lu–Hf isotope analyses, Chunlei Zong for her assistance with whole-rock major and trace element analysis, and Prof. Fukun Chen and Prof. Liqun Dai for Sr–Nd isotopic analyses.