Introduction

Andesitic arc magmatism at active continental margins is one of the most pronounced manifestations of crust–mantle interactions in an oceanic subduction channel (e.g., Hildreth and Moorbath 1988; Peacock 1993; Stern 2002; Eiler 2003; Spandler and Pirard 2013). A wealth of studies have revealed that the andesitic rocks with typical arc-like geochemical signatures are produced by melting of the overlying mantle wedge that was metasomatized by fluid/melt released from the subducted oceanic crust (e.g., Hawkesworth et al. 1991). By contrast, due to limited fluid availability, syn-subduction arc magmatism is absent during continental subduction (e.g., Rumble et al. 2003). However, the widespread occurrence of post-orogenic magmatic rocks in continental collisional orogens indicates that subcontinental lithospheric mantle (SCLM) on the top of deeply subducted continental crust could also be modified at the slab–mantle interface (e.g., Schaltegger 1997; Altherr et al. 2000; Gerdes et al. 2000; Zheng 2012). Therefore, these rocks preserve key information for the fertilization of the SCLM wedge and thus provide an important approach to decipher when and how did the crust–mantle interactions occur in the context of plate subduction. In the last three decades, although numerous studies have been performed on the generation and exhumation of ultrahigh-pressure (UHP) metamorphic rocks and attendant fluid/melt flow during continental collision (e.g., Chopin 2003; Hermann and Rubatto 2014; Cao et al. 2017), relatively less attention has been paid to encode the petrogenesis of post-orogenic magmatic rocks. What factors caused the extensive magmatism within collisional orogenic belts and the adjacent continental margins? How did the diverse magmatic rocks with mafic to felsic compositions form? What geodynamic mechanisms provided the required heat? Insights to these questions require detailed petrological, geochemical and geochronological constraints on targeted samples from representative regions.

The Dabie–Sulu orogen in central–eastern China has long been recognized as one of the largest UHP metamorphic belts in the world (Fig. 1; Hacker et al. 2000). This orogen underwent two vital stages of plate convergence during the Mesozoic, including the northward direct collision and movement of the Yangtze Craton beneath the North China Craton since the Late Triassic (e.g., Hacker et al. 2006), and the westward subduction of the paleo-Pacific plate in eastern Asia during the Jurassic to Cretaceous (Maruyama et al. 1997; Zhao and Ohtani 2009; Kusky et al. 2014), both contributing to significant modification and reworking of the ancient SCLM beneath the North China Craton. As a result, Late Mesozoic igneous rocks are widely exposed throughout the Dabie–Sulu orogen and the adjacent continental crust of the North China Craton (e.g., Jahn et al. 1999; Guo et al. 2004; Zhao et al. 2013; Zheng et al. 2018), gaining insight into the process of crustal recycling in mantle depths in a complex subduction-zone environment, and contributing to evaluate the geodynamic influences of the paleo-Pacific subduction in eastern Asian continent. Previous studies mainly concentrated on mafic igneous rocks and coeval granitoids in this region (e.g., Zheng et al. 2018), whereas the minor portion of post-orogenic andesitic rock suites has attracted less attention. The petrogenesis of andesite is controlled by multiple factors, such as chemical compositions of the melting source, the degree of flux/melt–peridotite reaction, source or magma mixing, fractional crystallization and crustal contamination (Gómez-Tuena et al. 2014). Until now, it is still controversial whether andesites that occur at both oceanic and continental subduction zones were crystallized from primary andesitic magmas or from various types of differentiated basaltic magmas (e.g., Kuno 1968; Hildreth and Moorbath 1988; Annen et al. 2005; Reubi and Blundy 2009; Straub et al. 2011; Meng et al. 2018). It has to be considered that they may also have experienced crustal contamination (e.g., Gill 1981; Gencalioglu Kuscu and Geneli 2010).

Fig. 1
figure 1

Geological sketch map of the Sulu orogenic belt (after Yoshida et al. 2004); the insert shows the tectonic location of the Sulu belt in China; WYF: Wulian–Yantai Fault; JXF: Jiashan–Xiangshui Fault; the grey framed area at General’s Hill was in part shown in Fig. 2 of Wang et al. (2017) and also in Fig. S1 of this study

Full size image

Here we present the results of petrography, mineral and whole-rock major and trace element and Sr–Nd isotope geochemistry, zircon U–Pb dating and in-situ Lu–Hf isotope studies of andesite dykes in the central Sulu belt to provide a comprehensive understanding of their petrogenesis in the context of continental/oceanic subduction. Our study is of help to illustrate on crust–mantle interactions and reworking of ancient SCLM in other collisional orogens worldwide, such as the Himalaya–Tibet in South Asia, the North Qaidam and South Altyn in China and the Western Gneiss Region in Norway.

Geological setting and sampling

The Sulu belt in Shandong peninsula represents the eastern segment of the Dabie–Sulu orogen that was produced from the northward collision and subduction of the Yangtze Craton beneath the North China Craton during the Triassic (e.g., Hacker et al. 2000; Ernst et al. 2007; Ni et al. 2013, 2016; Wang et al. 2018). It is bounded by the Wulian–Yantai fault to the north and the Jiashan–Xiangshui fault to the south and was offset northward by ~500 km from the Dabie belt along the NNE-trending sinistral Tan-Lu fault during the Cretaceous (Fig. 1; Xu and Zhu 1994). Further north beyond the Wulian–Yantai fault the Jiaobei terrane is located which is predominated by Neoarchean tonalite–trondhjemite–granodiorite gneisses and Paleoproterozoic metasedimentary sequences, belonging to the easternmost part of the North China Craton (Tang et al. 2008). Based on integrated field, petrological and geochemical studies, the Sulu belt is subdivided into a southern HP and a northern UHP metamorphic zone (e.g., Li et al. 2014). The UHP zone is primarily composed of gneisses, with minor amounts of eclogite, garnet peridotite, quartzite and marble. Coesite was recognized in eclogite and as inclusions in zircon from both the eclogite and the country rocks such as gneiss and marble, indicating that all these units underwent a joint in-situ UHP metamorphism at mantle depths >90 km (e.g., Liu and Liou 2011). The protoliths of these UHP rocks were mostly produced in a continental rift setting during the Neoproterozoic, probably in relation to the breakup of the supercontinent Rodinia (e.g., Zheng et al. 2005).

Of special interest in recent years is the partial melting these UHP rocks experienced during the latest stage of subduction to subsequent exhumation (e.g., Wang et al. 2014; Xia et al. 2018). Abundant studies have shown that different types of UHP rocks, including gneiss, eclogite and kyanite quartzite, were partially melted mainly induced by the release of water from nominally anhydrous minerals and breakdown of hydrous minerals including phengite/zoisite, leaving micro-scale multiphase solid inclusions to macro-scale granitic dykes and plutons as products (e.g., Hermann and Rubatto 2014; Wang et al. 2016, 2017). According to integrated geochemical and geochronological studies on mantle-derived garnet peridotites and syn-exhumation magmatic rocks (c.212–201 Ma) of this region, some researchers argued that the crustal-derived melts were transported to the overlying mantle wedge and reacted with the lithospheric mantle peridotite of the North China Craton (e.g., Zhao et al. 2012; Li et al. 2016). P–T–t conditions for the partial melting were constrained to be generally at 3.5–2.0 GPa and ~800 °C at 230–220 Ma (e.g., Wang et al. 2014, 2017).

After suturing of the two Cratons during the Triassic, the Sulu belt and central to eastern parts of the North China Craton were reactivated from c. 180 Ma due to shallow-angle subduction of the paleo-Pacific oceanic plate (Engebretson et al. 1985; Maruyama et al. 1997; Mao et al. 2011; Van der Meer et al. 2012). Subsequently in the Late Jurassic to Early Cretaceous (c. 145 Ma), the paleo-Pacific plate changed its velocity vector to the northeast, with the convergence direction subparallel to the entire continental margin of eastern Asia (Engebretson et al. 1985; Maruyama et al. 1997). The northern extension of this convergent plate boundary caused large-scale sinistral strike-slip structures in eastern China, such as the Tan-Lu fault system (Xu and Zhu 1994). In addition, extensive magmatism and gold mineralization were developed in eastern China, including Shandong, Hebei and Liaoning provinces, during the period of plate subduction and subsequent slab rollback (e.g., Wang et al. 1998; Yang et al. 2003; Zheng et al. 2018), indicating a possible petrogenetic relationship between these tectono-magmatic thermal events and the convergence of the paleo-Pacific plate. Numerous studies have established the geochronological framework of these Late Mesozoic intrusions and volcanic rocks, which were mainly produced during two episodes: Late Jurassic (c. 160–145 Ma) and Early Cretaceous (c. 140–110 Ma; Zhao et al. 2013). The Late Jurassic magmatic rocks are dominantly granites with minor amounts of mafic to andesitic rocks that mainly occur in the eastern part of the Sulu belt, Jiaobei terrane and Liaodong peninsula (e.g., Wu et al. 2005). In comparison, the Early Cretaceous magmatic rocks show greater variation in rock types and crop out throughout the entire Sulu belt and the central to eastern parts of the North China Craton (e.g., Zhao et al. 2013; Zheng et al. 2018).

This study focuses on a suite of andesite dykes that occurs at General’s Hill within the central Sulu UHP belt (Fig. S1; e.g., the eastern part of one dyke is also shown in Fig. 2 of Wang et al. 2017). The outcrop at General’s Hill extends along the coast for more than 1 km with a width of 50–100 m, where it basically consists of strongly foliated, tight to isoclinally folded migmatitic UHP eclogite, surrounded by granite gneiss (the northern part of the outcrop was shown in Wang et al. 2014, 2017). All the units are cut by Laoshan granite plutons and minor lamprophyre, andesite and granite porphyry dykes. The andesite dykes are ~2–20 m in width with exposed lengths of ~5–30 m (Fig. 2a–e). They are fresh and show clear chilled margins with the host eclogite; occasionally mafic microgranular enclaves are present within the andesite (Fig. 2f). The andesite has typical porphyritic textures; the phenocrysts (~30–40 vol.%) are mainly composed of plagioclase, hornblende and clinopyroxene, with minor amounts of finer-grained magnetite (Fig. 3). The plagioclase phenocrysts are lath-shaped and form euhedral–subhedral grains with long axis sizes up to ~1 cm across (Fig. 3a–d). Some plagioclase grains have abundant cracks, and are commonly altered by epidote, wairakite and clay minerals such as kaolinite and sericite (Fig. 3b and c). The hornblende forms subhedral to anhedral phenocrysts up to 1 mm in diameter; it commonly contains fine-grained apatite, titanite and magnetite as inclusions (Fig. 3e). The clinopyroxene phenocrysts with abundant cracks are euhedral–anhedral with long axis sizes up to ~3 mm, and commonly contain inclusions of magnetite and titanite (Fig. 3f–h). In most cases, they are surrounded by thin rims of hornblende (Fig. 3f–h). The matrix (~60–70 vol.%) mainly consists of fine-grained quartz and K-feldspar, with minor amounts of biotite, magnetite, titanite, apatite and zircon. Since all the andesites show similar features in the field, twelve unaltered andesite samples from different dykes were chosen for this study (Figs. S1 and 2). Using the petrological, geochemical and geochronological dataset created from these samples, we investigate the petrogenesis of these andesites, and discuss their implications for recycling of crustal materials into the ancient SCLM during continental/oceanic subduction.

Fig. 2
figure 2

Field relations of the andesite dykes with sample locations marked by orange colored circles; MME: mafic microgranular enclave. a–e occurrence of the four representative fresh andesite dykes that hosted by migmatitic eclogite; f oval MME in andesite

Full size image
Fig. 3
figure 3

Photomicrographs of major minerals and the microstructures of andesite. Cross-polarized light (a, b) and back-scattered electron (BSE; c, d) images to show the microstructure of plagioclase phenocrysts that are lath-shaped and in parts show altered cores (c), and which are commonly surrounded by clay minerals (b). The plagioclase phenocrysts often show clear zoning (d). BSE image of a hornblende phenocryst (e) that contains various inclusions such as Ap, Ttn and magnetite. Cross-polarized light (f) and BSE images (g, h) of clinopyroxene phenocrysts altered by variable amounts of hornblende. Abbreviations of mineral names are after Whitney and Evans (2010)

Full size image

Analytical methods

Mineral major element compositions

Major element compositions of minerals were measured using the JEOL JXA-8230 electron microprobe analyzer (EPMA) at the Center for Global Tectonics at School of Earth Sciences, China University of Geosciences (CUG), Wuhan. Operating conditions were as follows: 15 kV accelerating voltage, 20 nA cup current, 1 μm beam diameter, and 10 s counting time on peak and 5 s on background. A series of natural and synthetic standards from SPI (Structure Probe, Inc.) company was used, including: orthoclase for K; diopside for Ca; magnetite for Fe; jadeite for Na; pyrope for Mg; Yttrium Al garnet for Al; albite for Si; rutile for Ti; Apatite for P; pentilandite for Ni; rhodonite for Mn; chromite for Cr. Na, K and P were analyzed first to reduce electron beam-induced loss. Raw data were corrected using a ZAF algorithm (where Z = element atom number, A = X-ray absorption, F = X-ray fluorescence; Heinrich et al. 1986).

Mineral trace element compositions

Trace element compositions of major phenocryst phases including plagioclase, hornblende and clinopyroxene were measured using the laser ablation inductively coupled plasma-mass spectrometry (LA-ICP-MS) at the Wuhan Sample Solution Analytical Technology Co., Ltd., Wuhan, China. Detailed operating conditions for the laser ablation system and the ICP-MS instrument and data reduction are similar to those described by Zong et al. (2017). Laser sampling was performed using a GeolasPro laser ablation system that consists of a COMPexPro 102 ArF excimer laser (wavelength of 193 nm and maximum energy of 200 mJ) and a MicroLas optical system. An Agilent 7700e ICP-MS instrument was used to acquire ion-signal intensities. Helium was applied as a carrier gas. Argon was used as the make-up gas and mixed with the carrier gas via a T-connector before entering the ICP. A “wire” signal smoothing device is included in this laser ablation system (Hu et al. 2015). The spot size and frequency of the laser were set to 44 μm and 5 Hz, respectively. Trace element compositions of minerals were calibrated against various reference materials (BHVO-2G, BCR-2G, BIR-1G and NIST SRM610) without using an internal standard (Liu et al. 2008a). The measured and recommended values of the standards are provided in Table S1. Measuring conditions of each analysis were approximately 20 s of background acquisition, followed by 50 s data acquisition from the sample. An Excel-based software ICPMSDataCal was used to perform off-line selection and integration of background and analyzed signals, time-drift correction and quantitative calibration for trace element analysis (Liu et al. 2008a).

Whole-rock major and trace element analysis

Fresh rock samples free of visible alteration were crushed in a corundum jaw crusher to ˂60 mesh. For each sample, approximately 70 g was further powdered in an agate ring mill to ˂200 mesh for whole-rock major oxide, trace element and Sr–Nd isotope analyses. Major oxides were determined at the Comprehensive Rock and Mineral Test Center, Wuhan. For the major elements, H2O+ and CO2 were determined by gravimetry, while other oxides were measured using conventional X-ray fluorescence (XRF) spectrometry (Shimadzu XRF-1800). The analytical uncertainty is generally ˂5 %. Trace element concentrations were analyzed at the State Key Laboratory of Geological Processes and Mineral Resources (GPMR), CUG, Wuhan using an Agilent 7500a ICP-MS. International standards including AGV-2, BHVO-2, BCR-2 and RGM-2 were used as reference materials (associated results of the standards and recommended values are summarized in Table S2). The detailed sample digestion procedure for ICP-MS analysis, and the analytical precision and accuracy for trace elements follow those described by Liu et al. (2008b).

Whole-rock Sr–Nd isotope analysis

Whole-rock Sr and Nd isotope compositions were also acquired at the State Key Laboratory of GPMR, CUG, Wuhan, using a Neptune Plus multi-collector (MC) ICP-MS. 87Rb/86Sr and 147Sm/144Nd ratios were calculated from Rb–Sr and Sm–Nd concentrations measured by ICP-MS, respectively. The measured Sr and Nd isotope ratios were normalized to 86Sr/88Sr = 0.1194 (Nier 1938) and 146Nd/144Nd = 0.7219 (O'Nions et al. 1977), respectively. During the measurement, the NBS987 and Chinese GBW04411 standards gave an average 87Sr/86Sr value of 0.710274 ± 0.000008 (2σ) and 0.759909 ± 0.000006 (2σ), respectively, while the BCR-2 and JNdi-1 standards yielded an average 143Nd/144Nd of 0.512592 ± 0.000003 (2σ) and 0.512119 ± 0.000008 (2σ), respectively. Detailed sample preparation, chemical separation and analytical procedures for Sr–Nd isotope analysis are similar to those given by Gao et al. (2004).

Zircon U–Pb dating, trace element and Lu–Hf isotope analysis

Zircon was separated from samples following conventional crushing, sieving, heavy liquid and magnetic separation procedures. Individual zircon grains for analysis were separated from any remaining contaminant minerals under a binocular microscope. For each sample, a selection of the zircon separate was mounted in an epoxy resin block and polished to approximately half thickness for analysis. In order to monitor the surface and internal structures of the zircons, back-scattered electron and cathodoluminescence (CL) images were acquired using a FEI Quanta 450 field emission gun scanning electron microscope equipped with a Gatan Mono CL4+ CL system at the State Key Laboratory of GPMR, CUG, Wuhan. The operating conditions were 10 kV acceleration voltages and a spot size of 5 μm with a working distance of 13.9–14.1 mm.

Zircon U–Pb isotope and trace element compositions were determined synchronously by LA-ICP-MS at the State Key Laboratory of GPMR, CUG, Wuhan; working conditions and the data reduction protocol follow those described by Liu et al. (2008a). Samples were ablated using a GeoLas 2005 system (spot size of 32 μm), which was connected to an Agilent 7500a ICP-MS instrument to acquire ion-signal intensities. Each analysis was subjected to a background acquisition of approximately 20 s followed by 50 s data acquisition from the unknowns. Zircon 91500 was used as external standard for U–Pb dating, and was analyzed twice bracketing each batch of 6 unknown samples. U–Th–Pb isotopic ratios used for zircon 91500 during the measurement are from Wiedenbeck et al. (1995). Uncertainty of the preferred values was propagated to the ultimate results of the samples. Trace element compositions of zircons were calibrated against multiple-reference materials (BCR-2G and BIR-1G) without applying internal standardization (Liu et al. 2008a). The preferred values of element concentrations for the USGS reference glasses are from the GeoReM database (http://georem.mpch-mainz.gwdg.de/). Off-line selection and integration of background and analytical signals, and time-drift correction and quantitative calibration for U–Pb dating and trace element analysis were performed by ICPMSDataCal (Liu et al. 2008a). The data were processed using the program ISOPLOT/Ex_ver3 developed by Ludwig (2003).

After zircon U–Pb and trace element analysis, in-situ Lu–Hf isotope analysis of zircon was carried out using a Neptune Plus MC-ICP-MS in combination with a Geolas 2005 excimer ArF laser ablation system at the State Key Laboratory of GPMR, CUG, Wuhan. All Lu–Hf isotope analyses were acquired in single spot ablation mode with a spot size of 44 μm at the location of the same spots used for dating. Each measurement was done using a background acquisition of 20 s, followed by 50 s of signal acquisition from the unknowns. Zircon 91500, GJ-1 and Temora were used as reference standards, the results of which are given in Table S3. Detailed operating conditions for the laser ablation system and the MC-ICP-MS instrument, and the analytical method are the same as used by Hu et al. (2012). Mass bias of Hf and Yb were calculated using an exponential correction law and normalized values of 179Hf/177Hf = 0.7325 and 173Yb/171Yb = 1.1248 (Blichert-Toft et al. 1997), respectively. Interference of 176Yb on 176Hf and 176Lu on 176Hf were corrected by measuring the interference-free 173Yb and 175Lu isotopes and using the recommended ratios of 176Yb/173Yb = 0.7876 (McCulloch et al. 1977) and 176Lu/175Lu = 0.02656 (Blichert-Toft et al. 1997), respectively. A decay constant of 1.867 × 10−11 year−1 for 176Lu was chosen for data calculation (Söderlund et al. 2004). εHf(t) values were calculated relative to a chondritic reservoir with a 176Hf/177Hf ratio of 0.282772 and 176Lu/177Hf of 0.0332 (Blichert-Toft and Albarède 1997). Single-stage Hf model ages were calculated relative to a model depleted mantle with a present day 176Hf/177Hf ratio of 0.28325 and a 176Lu/177Hf ratio of 0.0384 (Vervoort and Blichert-Toft 1999). Two-stage Hf model ages were calculated by using a mean 176Lu/177Hf ratio of 0.015 for the average continental crust as suggested by Griffin et al. (2002).

Analytical results

Mineral major and trace element compositions

Representative major element compositions of phenocrysts (including plagioclase, hornblende and clinopyroxene) and matrix minerals (K-feldspar, biotite and titanite) are given in Table S4. In Table S5, trace element compositions of plagioclase, hornblende and clinopyroxene are summarized.

Plagioclase

Plagioclase is the most abundant phenocryst in the andesite. It contains low amounts of K-feldspar component (An28–42Ab57–69Kfs1–6). In the An–Ab–Kfs diagram, plagioclase mainly plots in the andesine field of the classical plagioclase nomenclature (Fig. 4a). In general plagioclase phenocrysts are weakly zoned (e.g., Fig. 3d); two representative grains from sample GH177–2 were selected to check their major and trace element variations (Fig. S2). As illustrated in Fig. S2a–d, the plagioclases show a zoning as expected from magmatic crystallization; CaO decreases and Na2O increases from core to rim, corresponding to ~An40 in the cores and ~An30 in the rims. However, the REE contents don’t show considerable variations within a single phenocryst and also between the two grains studied (∑REE =47.42–63.17 ppm vs. 43.18–58.31 ppm). Chondrite-normalized plots show consistent REE patterns that are characterized by an enrichment in LREE [(La/Sm)N = 21.66–52.35] but deleption in H- (Heavy-) REE [(Gd/Yb)N = 0.82–14.91] with pronounced positive Eu anomalies (Eu/Eu* = 9.91–24.44; Fig. S2e, f).

Fig. 4
figure 4

a An–Ab–Kfs triangular diagram (modified after Deer et al. 1963) showing the compositions of plagioclase phenocrysts (blue squares) and K-feldspar in the matrix (red dots); b classification of calcic hornblende phenocrysts (blue diamonds) and alteration rims (orange dots) of amphibole surrounding clinopyroxene after Leake et al. (1997); c classification of clinopyroxene phenocrysts after Morimoto (1988)

Full size image

Hornblende and actinolite

Hornblende occurs as phenocryst or alteration rim surrounding clinopyroxene. The hornblende phenocrysts contain variable amounts of SiO2 (50.33–56.68 wt.%), FeOt (10.29–14.34 wt.%), CaO (11.66–12.38 wt.%) and MgO (12.99–17.24 wt.%) (Table S4). According to the Hawthorne et al. (2012) nomenclature, the hornblende phenocrysts belong to the subgroup of calcium amphiboles and for convenience were displayed in the Leake et al. (1997) diagram, which show that the respective analyses plot in the magnesiohornblende to actinolite fields (Fig. 4b). On the other hand, alteration rims of amphibole surrounding clinopyroxene have similar SiO2 (52.84–55.77 wt.%) and CaO (11.93–12.41 wt.%), and relatively lower FeOt (8.51–11.08 wt.%) but higher MgO (16.10–18.35 wt.%) contents in comparison with those of the phenocrysts (Table S4). Based on the classification of Leake et al. (1997), the compositions of these late amphiboles plot in the actinolite field (Fig. 4b). Two single hornblende phenocrysts from sample GH177–2 were chosen to study major and trace element profiles, in order to check if there are any compositional variations present. As illustrated in Fig. S3, the oxides of SiO2, FeO, MgO and CaO however show very flat profiles.

For trace element compositions, the analyzed two hornblende phenocrysts have equally low total REE concentrations (∑REE = 5.10–10.30 ppm, except for two outliers at 21.69 ppm and 44.57 ppm, respectively) compared to the actinolite rims (∑REE = 20.53–41.52 ppm) surrounding clinopyroxene (Table S5). The hornblende phenocrysts show consistent chondrite-normalized REE patterns that are depleted in LREE [(La/Sm)N = 0.15–0.59, with two outliers at 1.96 and 2.40] and flat to slightly enriched in HREE [(Gd/Yb)N = 0.48–1.10, with two outliers at 2.06 and 1.24] generally with negative Eu anomalies (Eu/Eu* = 0.41–1.15; Fig. S4a–d), except for two LREE-enriched spots (i.e. the two outliers as described above), possibly resulting from contamination of tiny LREE-rich mineral inclusions. In primitive mantle-normalized multi-element diagrams (Fig. S4b, d), the both hornblende phenocrysts show pronounced positive anomalies of Th, U, Nd and Pb. Compared to the phenocrysts, the alteration rims of actinolite show depletion both in LREE and HREE [(La/Sm)N = 0.14–0.54; (Gd/Yb)N = 1.78–2.06] with negligible to negative Eu anomalies (Eu/Eu* = 0.27–0.87; Fig. S4e). They show positive anomalies of Th, U, Pb and Nd but negative anomalies of Ta and Sr in a primitive mantle-normalized multi-element diagram (Fig. S4f).

Clinopyroxene

The clinopyroxene phenocrysts are relatively high in SiO2 (53.47–55.02 wt.%), FeOt (6.06–8.59 wt.%) and CaO (20.47–23.63 wt.%), but low in MgO (13.83–14.91 wt.%), Al2O3 (0.23–1.34 wt.%) and TiO2 (from below detection limit to 0.23 wt.%) with high Mg# values [Mg# = 100 × Mg/(Mg + Fe2+) molar] of 74–81. Moreover, the clinopyroxene is homogeneous in major element compositions within a single grain (Fig. S5a, b). In the enstatite–ferrosilite–diopside–hedenbergite quadrilateral diagram after Morimoto (1988), the compositions plot in the diopside (Wo45–49En39–43Fs10–14) and augite (Wo43–44En42–43Fs13–14) fields (Fig. 4c). The clinopyroxene phenocrysts have moderate REE abundances (∑REE = 151.86–264.13 ppm) and exhibit flat chondrite-normalized LREE [(La/Sm)N = 0.72–1.06] and depleted HREE [(Gd/Yb)N = 1.69–2.25] patterns with conspicuous negative Eu anomalies (Eu/Eu* = 0.44–0.64; Fig. S5c). In a primitive mantle-normalized spider diagram, they are enriched in Th and U but depleted in Nb, Ta, Zr and Hf (Fig. S5d).

Matrix minerals

K-feldspar in the matrix is mainly composed of SiO2 (64.66–66.03 wt.%), Al2O3 (17.36–18.16 wt.%) and K2O (14.38–15.99 wt.%), with minor amounts of FeOt (0.04–0.17 wt.%), CaO (from below detection limit to 0.04 wt.%) and Na2O (0.68–1.24 wt.%); they possess low albite content (Kfs88–94Ab6–11; Fig. 4a). In the An–Ab–Kfs diagram, they mainly plot in the sanidine field (Fig. 4a). Biotite has relatively consistent compositions of SiO2 (36.00–37.91 wt.%), FeOt (18.45–19.29 wt.%), Al2O3 (12.42–13.62 wt.%), MgO (11.12–12.53 wt.%) and K2O (8.92–10.06 wt.%), with minor TiO2 (0.04–0.17 wt.%) and Na2O (0.68–1.24 wt.%). Titanite mainly comprises of SiO2 (30.62–31.11 wt.%), TiO2 (37.55–38.62 wt.%) and CaO (27.22–28.74 wt.%) with a chemical formula of Ca0.970–1.016Ti0.931–0.965Si1.011–1.021O5.

Whole-rock major oxides, trace elements and Sr–Nd isotope compositions

Twelve andesite samples were analyzed for major and trace element compositions, which are summarized in Table S6 and presented in Figs. S6, 5 and 6. They possess moderate concentrations of SiO2 (54.97–62.24 wt.%), Na2O + K2O (6.35–7.24 wt.%), Al2O3 (13.32–14.70 wt.%), MgO (3.37–7.12 wt.%), Ni (37–107 ppm) and Cr (124–360 ppm) with high Mg# values of 54–64 (Fig. S6). They mainly fall in the field of andesite on the Nb/Y vs. Zr/TiO2 ratio diagram (Fig. 5a; Winchester and Floyd 1977). In the total alkalis vs. silica diagram (TAS), the samples plot in the field of trachyandesite, with one exception that plots in the field of basaltic trachyandesite (Fig. 5b) and on a diagram of K2O vs. SiO2 (Fig. 5c), the andesites fall in the field of high-K calc-alkaline to shoshonite series.

Fig. 5
figure 5

Zr/TiO2 × 0.0001 vs. Nb/Y chemical classification after Winchester and Floyd (1977) (a), Na2O + K2O vs. silica (TAS, b) diagram after Le Maitre (2002), and SiO2 vs. K2O diagram (c; after Morrison 1980; Peccerillo and Taylor 1976) for the andesites at General’s Hill

Full size image
Fig. 6
figure 6

Whole-rock chondrite-normalized rare earth element patterns (a), and primitive mantle-normalized trace element patterns (b) of the andesites at General’s Hill. Data from contemporaneously crystallized lithospheric mantle-derived mafic–andesitic igneous rocks in the Sulu belt and the interiors of North China Craton are shown for comparison. Data sources: Fan et al. 2001; Zhang et al. 2002, 2012; Guo et al. 2004, 2005, 2014a, b; Xu et al. 2004b; Yang et al. 2005, 2008, 2012; Liu et al. 2006, 2008, 2009, 2012; Yang and Li 2008; Tang et al. 2009; Pei et al. 2011; Ma et al. 2014a, b; Cai et al. 2015; Deng et al. 2017. Normalization values are from Sun and McDonough (1989)

Full size image

On a chondrite-normalized diagram (Fig. 6a), the andesites are characterized by enriched LREE patterns relative to HREE [(La/Yb)N = 16.72–19.91], with weak to negligible Eu anomalies (Eu/Eu* = 0.83–0.97). In a primitive mantle-normalized multi-element diagram (Fig. 6b), they show pronounced enrichment in LILE, but depletion in HFSE. Both diagrams prove nearly identical patterns of all studied andesite samples.

Nine andesite samples (GH177–1 to GH177–9) were further analyzed for whole-rock Sr–Nd isotope compositions, as listed in Table S7. Initial 87Sr/86Sr ratios and ɛNd(t) values were corrected to t = 120 Ma, according to the results of zircon U–Pb dating reported below. The andesites exhibit variably high initial 87Sr/86Sr ratios of 0.7073–0.7086 and negative ɛNd(t) values of −15.7 to −14.4 (Fig. 7), yielding two-stage Nd model ages of 2088–2188 Ma.

Fig. 7
figure 7

ɛNd(t) vs. initial 87Sr/86Sr plot at t = 120 Ma for the andesites (cyan dots) at General’s Hill. Sr–Nd compositions of UHP eclogite (light blue area) and gneiss (grey area) in the Sulu belt, and contemporaneous lithospheric mantle-derived mafic–andesitic igneous rocks in the Sulu belt (light red area) and in the North China Craton (light green area) are also plotted for comparison (also calculated back to t = 120 Ma). Data sources for the UHP eclogite and gneiss are: Chen et al. 2002, 2014b; Zhao et al. 2007; Tang et al. 2008; Wang et al. 2017; Data sources for the andesitic–mafic igneous rocks are the same as those in Fig. 6; lower crust of the Yangtze Craton, lower and upper crust of the NCC (North China Craton) are from Jahn et al. (1999)

Full size image

Zircon U–Pb dating and trace elements

Zircon from four andesite samples (GH177–1, −3, −8 and − 12) was analyzed for U–Pb dating and trace element compositions, as summarized in Tables S8 and S9, respectively. The zircon grains, most of which have isometric to long prismatic forms, are generally euhedral, colorless and transparent in plane light. They have grain lengths ranging from 50 to 300 μm, with length to width ratios of 1:1–5:1. In CL images, most grains show oscillatory or banded zoning with moderate luminescence, but in some cases inherited zircons can be observed (Fig. 8). Some grains are partially resorbed or recrystallized and show thin metamorphic rims (Fig. 8a), suggesting modification by post-magmatic high-temperature hydrothermal alteration (cf. Liati et al. 2002; Zhang et al. 2012).

Fig. 8
figure 8

Cathodoluminescence images of representative zircon grains from four andesite samples. Circles mark the locations of LA-ICP-MS U–Pb (red) and Hf isotope (blue) analyses, respectively, together with corresponding 206U/238Pb dates and ɛHf(t) values

Full size image

Thirty-six analyses were performed on 36 grains for sample GH177–1. One analysis on an inherited zircon gives a Neoproterozoic 206Pb/238U age of 645 ± 18 Ma, while the other 35 spots on newly crystallized magmatic zircon domains with high Th/U ratios (0.85–2.71) yield concordant 206Pb/238U ages between 122 ± 6 and 111 ± 7 Ma with a weighted mean age of 116 ± 1 Ma (2σ, MSWD = 1.9; Fig. 9a). Twenty-seven spots were analyzed for sample GH177–3, among which 26 analyses show Th/U ratios of 0.79–1.51 and give 206Pb/238U ages from 130 ± 6 to 108 ± 6 Ma yielding a weighted mean age of 119 ± 2 Ma (2σ, MSWD = 5.8; Fig. 9b). One analysis on a relict zircon gives an Archean 207Pb/206Pb age of 3292 ± 48 Ma (Fig. 9b). Similarly, twenty-seven analyses of sample GH177–8 were studied, and 25 spots yield a weighted mean age of 119 ± 2 Ma (2σ, MSWD = 4.9); the other 2 spots give Paleoproterozoic and Neoproterozoic ages of 2354 ± 76 and 780 ± 19 Ma, respectively (Fig. 9c). Seventeen analyses performed for sample GH177–12 give concordant 206Pb/238U ages from 133 ± 4 to 118 ± 3 Ma, with a weighted mean age of 124 ± 2 Ma (2σ, MSWD = 4.4; Fig. 9d).

Fig. 9
figure 9

U–Pb concordia plots (a–d) and chondrite-normalized REE patterns (e–h) of zircons from the andesites. The insets in a–c are U–Pb concordia diagrams for concordant new zircon domains; the mean 206Pb/238U ages are reported with 2σ uncertainty. Normalization values are from Sun and McDonough (1989)

Full size image

Taken together, the analyses on newly crystallized magmatic zircon domains of all the andesite samples studied yield consistent Early Cretaceous U–Pb ages of 124 ± 2 to 116 ± 1 Ma, whereas the relict zircons give variable Neoproterozoic (780 ± 19 and 645 ± 18 Ma), Paleoproterozoic (2377 ± 46) and Archean (3165 ± 38 Ma) ages.

As illustrated in Fig. 9e–h, both magmatic and relict zircon domains from all the four samples possess similar trace element compositions. They have variably high total REE concentrations (∑REE = 444–4182 ppm, mostly >1500 ppm) with variable but in general low Hf/Y ratios (3.43–22.75). In chondrite-normalized REE plots they show steep HREE patterns [(Dy/Yb)N = 0.10–0.23] with pronounced negative Eu anomalies (Eu/Eu* = 0.11–0. 77, mostly <0.50).

Zircon Lu–Hf isotope compositions

A total of 32 analyses on both magmatic and relict zircon domains from three samples (GH177–1, −3 and − 8) was performed to determine Lu–Hf isotope compositions. The results are listed in Table S10 and graphically processed in Fig. 10. The ɛHf(t) values and Hf model ages are corrected to t = 120 Ma for magmatic zircon and to the apparent ages for relict zircon, based on the U–Pb dating results as reported above.

Fig. 10
figure 10

Histograms of zircon ɛHf(t) values (a, c, e) and two-stage Hf model ages (b, d, f) for the andesites at General’s Hill

Full size image

For sample GH177–1, the magmatic zircon domains (n = 7) have restricted Hf isotopic compositions with 176Hf/177Hf ratios of 0.281940–0.282021 and ɛHf(t) values of −26.9 to −24.0, corresponding to two-stage Hf model ages of 2681–2863 Ma; one analysis on a relict zircon (645 Ma) yields an ɛHf(t) value of −17.7 and a two-stage Hf model age of 2681 Ma (Fig. 10a and b). For sample GH177–3, the magmatic zircons (n = 11) exhibit ɛHf(t) values of −26.3 to −17.6, corresponding to two-stage Hf model ages of 2282–2838 Ma, while one available analysis on a relict zircon (3292 Ma) yields a less negative ɛHf(t) value of −4.2 and an older two-stage Hf model age of 3860 Ma (Fig. 10c and d). Similarly, the magmatic zircon domains (n = 10) in sample GH177–8 have negative ɛHf(t) values of −27.0 to −23.8, yielding two-stage Hf model ages of 2121–2871 Ma; relict zircons at apparent ages of 2354 and 780 Ma exhibit less negative ɛHf(t) values of −8.1 and − 7.1, respectively, corresponding to two-stage Hf model ages of 3380 and 2121 Ma (Fig. 10e and f).

Discussion

Timing of crystallization

Most zircon grains in the andesites show clear oscillatory zoning, variably high Th/U ratios (0.50–2.71, with a median value of 1.14), and steep HREE patterns with obvious negative Eu anomalies (Figs. 8 and 9; Table S8 and S9), typical of magmatic origin (cf. Pidgeon 1992; Hoskin and Ireland 2000; Du et al. 2017a, b; Rubatto 2017). The LA-ICP-MS U–Pb dating on these magmatic zircons yielded consistent ages of 124 ± 2 to 116 ± 1 Ma (Fig. 9a–d). These ages are in line with data of the timing of extensive magmatism in the Dabie–Sulu orogen and in the adjacent continental margin of the North China Craton during the Early Cretaceous (cf. Liu et al. 2009; Zhao et al. 2013). Therefore, we interpret the ages of 124–116 Ma as crystallization ages, thus dating the eruption of andesitic lava at General’s Hill.

Source nature of the andesite

As summarized in Table S6, the andesites have intermediate concentrations of MgO (3.37–7.12 wt.%), Ni (37–107 ppm) and Cr (124–360 ppm) with high Mg# values of 54–64, different from those crystallized from crustal-derived melts (Cr = 4.49–21.1 ppm; Mg# values = 15–54, with two exceptions of 56 and 58, respectively; cf. Patiño Douce and Beard 1995; Rapp and Watson 1995; Qian and Hermann 2013). In addition, the clinopyroxene phenocrysts exhibit high Mg# values of 74–81 (Table S4), also indicating crystallization from a mantle source (cf. Liang et al. 2018). Accordingly, they are interpreted to be generated from mantle-derived magmas of ultramafic to mafic lithologies. However, except for Cr and Ni contents, the studied andesites exhibit distinct features of trace elements and radiogenic isotopes that resemble those of the continental crust (Taylor and McLennan 1995), including: (1) arc-like trace element patterns with enrichment of LILE and LREE but depletion of HFSE (Fig. 6); (2) high initial 87Sr/86Sr ratios and pronounced negative whole-rock ɛNd(t) and zircon ɛHf(t) values (Figs. 7 and 10); and (3) enrichment of Th and U but depletion of Nb, Ta, Zr and Hf for the hornblende and clinopyroxene phenocrysts (Figs. S4 and S5). These features indicate that the mantle source can neither be the normal asthenospheric mantle nor a mid-ocean ridge-type mantle that are isotopically depleted (e.g., Hofmann 1988; Song et al. 2018). In this regard, in view of the tectonic setting where the andesite occurs and the evolutional history of this region, the most appropriate candidate for the mantle source is the SCLM of the North China Craton, which was fertile in compositions due to incorporation of crustal components. The andesites are enriched in LILE and contain relatively high amounts of K2O (2.81–4.52 wt.%), implying a potassium-bearing (with minerals such as hornblende and phlogopite) mantle source for their origin (Foley et al. 1996; Ionov et al. 1997). These rocks have Rb/Sr and Ba/Rb ratios of 0.09–0.16 and 16.25–21.77, respectively, which plot in the phlogopite field in the Rb/Sr vs. Ba/Rb diagram after Furman and Graham (1999) (Fig. S7), indicating a predominance of phlogopite over hornblende in the melting source.

How did the crustal materials be incorporated into the mantle-derived magmas?

When and how did the crustal material become incorporated into the mantle source generating andesitic magmatism? Answers to these questions are essential to decipher the petrogenesis of the andesites. In general, during ascent of mantle-derived magmas en route to the surface through the continental crust, they might be contaminated by crustal materials (e.g., AFC process). However, as demonstrated in Figs. 7 and 10, the andesites have a limited range of whole-rock initial 87Sr/86Sr and ɛNd(t) as well as zircon ɛHf(t) values. This argues against the possibility of significant crustal contamination during ascent of the mantle-derived magmas through the continental crust. On the other hand, although the andesites exhibit a relatively narrow range of MgO concentrations, Al2O3, Na2O + K2O and SiO2 decrease but FeOt and CaO increase with the increase of MgO (Fig. S6). This indicates that slight fractional crystallization of clinopyroxene and/or hornblende could have occurred during magma ascent. Moreover, plagioclase phenocrysts exhibit pronounced positive Eu anomalies, while the hornblende and clinopyroxene exhibit weak to moderate negative Eu anomalies (Figs. S2, S4 and S5), suggesting that plagioclase fractionation did occur during the evolution of the parent magma. However, consistently low contents of TiO2 (0.64–0.83 wt.%) in these rocks do not match fractionation of Fe–Ti oxides. The weak normal zoning of major elements of plagioclase may largely reflect simple growth history during cooling. Nevertheless, fractional crystallization cannot contribute to enriched signatures of isotope compositions if the andesites were derived from a depleted mantle source. Therefore, it is inferred that the andesites inherited their geochemical features from the melting source that was metasomatized by crustal materials through source mixing prior to the onset of andesitic magmatism (cf. Zhao et al. 2013; Chen et al. 2014a). The inferred existence of phlogopite and/or hornblende in the melting source also calls for a fertile mantle that was metasomatized by fluids/melts.

With respect to the process of source mixing for the Mesozoic igneous rocks in the Dabie–Sulu orogen, the prevailing model considered is that, which involves deep continental subduction of the Yangtze Craton and subsequent crust–mantle interactions in a continental subduction channel (e.g., Zheng 2012; Zhao et al. 2013; Dai et al. 2016). This model was proposed mainly based on the following observations (cf. Zhao et al. 2013; Dai et al. 2016; Li et al. 2016): (1) the igneous rocks, including basalt, hornblendite, lamprophyre, diabase and andesites, commonly contain relict zircons of Neoproterozoic and Triassic ages, resembling those of the UHP metamorphic rocks from the Yangtze Craton; (2) the igneous rocks have enriched isotope compositions (including Sr, Nd, Pb, Hf and O) that are similar to the host UHP gneiss and eclogite in this region; (3) mantle-derived orogenic peridotites in the Dabie–Sulu orogen have crust-like elemental and isotopic signatures; they contain metasomatic zircons also of Neoproterozoic and Triassic ages. The andesites at General’s Hill have relict zircons that yield Neoproterozic, Paleoproterozoic and Archean U–Pb ages (Fig. 9a–c), similar to the prominent ages dated from the UHP metamorphic rocks in the Dabie–Sulu orogen and the crustal rocks of the North China Craton, respectively (Zheng et al. 2005; Hacker et al. 2006; Liu and Liou 2011). Since crustal contamination was negligible in this case, these relict zircons should not be physically incorporated into the magma during its ascent. In this regard, the above mentioned geochemical and geochronological features indicate involvement of crustal rocks from both the Yangtze and North China Cratons into the SCLM by source mixing during the Triassic orogeny. On the other hand, as summarized in Table S11 and shown in Fig. 11, contemporaneously formed mafic–andesitic igneous rocks (140–110 Ma) with similar geochemical features not only crop out in the Dabie–Sulu orogen, they also occur in western Shandong, Liaoning peninsula and even at several localities in the central–western domains of the North China Craton, including the Taihang mountain, Beijing region and Jining area of Inner Mongolia (cf. Chen et al. 2003; Guo et al. 2014b; Zhang et al. 2014; Zheng et al. 2018). The Triassic continental subduction may not count for the fertilization of the SCLM and generation of the Early Cretaceous igneous rocks in the interiors of the North China Craton. As a consequence, a more common geodynamic mechanism is further called for to decode the Mesozoic magmatism in the vast regions of eastern China.

Fig. 11
figure 11

Spatial distribution of representative lithospheric mantle-derived mafic–andesitic igneous rocks of Early Cretaceous ages (c. 140–110 Ma) in eastern China, including the Sulu orogenic belt and the interiors of the North China Craton (modified after Liu et al. 2006). For the igneous rocks exposed in Shandong peninsula, spatial and chronological distribution information is shown in enlarged Fig. 11b. The yellow stars denote the location of each study, with specification of age information and literature source, while the red star represents the results of this study. TLF: Tan-Lu Fault; Lam: Lamprophyre; Dio: Diorite; Dol: Dolerite; Dia: Diabase; Gbr: Gabbro; And: Andesite; Pl-bearing Hbt: Plagioclase-bearing Hornblendite; Bas: Basalt; Data sources: Zhang et al. 2002, 2012; Chen et al. 2003; Guo et al. 2004, 2014a, b; Peng 2004; Xu et al. 2004a, b; Liu et al. 2006, 2008, 2009, 2012; Yang et al. 2008; Yang and Li 2008; Tang et al. 2009; Pei et al. 2011; Ma et al. 2014a, b; Cai et al. 2015; Deng et al. 2017

Full size image

The andesites at General’s Hill and lithospheric mantle-derived mafic–andesitic rocks of Early Cretaceous ages in eastern China share similar arc-like geochemical features (Fig. 6). They share similarities with typical island-arc basalts and andesites above oceanic subduction zones, such as the classical Andean subductioin zone and the Japanese Islands (Perfit et al. 1980; Stern and Kilian 1996; Kelemen et al. 2003a, b). These characteristics point towards a similar process of formation. Geophysical observations and large-scale numerical modeling data demonstrate that subduction of the paleo-Pacific plate played a dominant role in the mantle dynamics of eastern China during the Late Mesozoic (c. 150–90 Ma; Niu 2005; Seton et al. 2012; Kusky et al. 2014). Therefore, in addition to the metasomatic process during the Triassic orogeny, the subduction of paleo-Pacific plate along the eastern margin of Asian continent could be considered as another appropriate candidate for the fertilization of the SCLM beneath the North China Craton. Note that the andesites contain high initial 87Sr/86Sr ratios, significantly negative ɛNd(t) values and zircon ɛHf(t) values, which imply that the metasomatic fluids/melts were mainly released from marine sediments that were located on top of the subducted oceanic crust.

Petrogenesis of the andesite

Although more and more studies have confirmed the genetic link of andesite formation with plate subduction and crust–mantle interactions in a subduction channel, there are still open questions related to its genesis (e.g., as reviewed by Gόmez-Tuena et al., 2014). Two contrasting models, namely the basalt-input model and the andesite-input model, were proposed to the andesite formation (e.g., Hildreth and Moorbath 1988; Gómez-Tuena et al. 2014). The major characteristics of the alternative models refer to either a crystallization of primary andesitic melts from slab and/or mantle materials, or to a derivative andesitic product of a basaltic magma that differentiated in the continental crust. Below, we firstly evaluate the possibility of a differentiation from basaltic melts to form the andesite and then we focus on questions related to its petrogenesis.

As illustrated in Fig. 11, contemporaneous mafic–andesitic igneous rocks, including basalt, lamprophyre, spessartite, kersantite, gabbro, dolerite, hornblendite and monzodiorite, with similar elemental and isotopic features (Figs. 6 and 7), are widely exposed in the Sulu belt and the adjacent continental crust of the North China Craton (e.g., Guo et al. 2004; Wu et al. 2005; Zhao et al. 2013; Ma et al. 2014a, b; Cai et al. 2015; Deng et al. 2017; Zheng et al. 2018). This suggests an interrelation between these rocks and indicates the probability that the andesites represent evolved products of melts that originally yielded a basaltic composition. However, as shown in Fig. S6, there are no clear common features related to major and trace elements vs. MgO interrelation between these rocks and the studied andesites, indicating that fractional crystallization of basaltic magmas did not play a significant role in the andesite formation. In fact, the andesite at General’s Hill is crosscut by lamprophyre dykes (Fig. S1), which also suggests that it was not derived from the parent basaltic magmas that crystallized the lamprophyre. Therefore, the vast occurrences of Early Cretaceous mafic to andesitic igneous rocks in the Sulu belt and in the interiors of the North China Craton are characterized to originate from heterogeneous sources due to variable degrees of crustal metasomatism, and the andesites essentially crystallized from primary andesitic melts.

Based on integrated geochemical and geochronological studies of andesite from the Luzong basin in the Middle–Lower Yangtze River Belt of southern China, Chen et al. (2014a) proposed a two-stage process to account for the genesis of orogenic andesites that occur above oceanic subduction zones. These authors argue that mantle-wedge peridotite firstly reacted with slab-derived melts/fluids during the subduction of an oceanic crust, giving rise to fertile and enriched metasomatites of ultramafic to mafic compositions; the metasomatites were subsequently melted to generate andesitic magmatism. In view of the similarity in geochemical characteristics between these rocks and the studied andesites and the tectonic evolution of this region, a similar scenario is also applicable to illustrate the generation of andesitic magmas at General’s Hill, as illustrated in the geological sketch drawings of Fig. 12. In detail, the SCLM wedge beneath the North China Craton was firstly modified by hydrous melts/fluids mostly released from the subducted continental crust of the Yangtze Craton during the Triassic (Fig. 12a; c. 240–220 Ma). The subducted slab might also have brought crustal fragments from the North China Craton into the subduction zone, resulting in the existence of relict zircons in the andesites with Neoproterozoic, Paleoproterozoic and Neooarchean ages. Since the Jurassic (Fig. 12b; c. 180 Ma), paleo-Pacific plate started to subduct beneath the Eurasian continent at a shallow angle (e.g., Engebretson et al. 1985; Maruyama et al. 1997). The slab-derived fluids/melts, mostly from marine sediments, were transported to the overlying mantle wedge and further reacted with the SCLM peridotite to form metasomatic rocks of ultramafic to mafic compositions (cf. Kelemen et al. 2003a, b). The subducted paleo-Pacific slab then progressively rolled back from c. 145 to 120 Ma, which would not only cause lithospheric extension, but also result in flux of asthenospheric mantle materials to the ‘gap’ between old and new slab positions (Niu 2005; Kusky et al. 2014; Zheng et al. 2018). This process would provide the needed heat for partial melting of the overlying enriched mantle and the formation of andesitic magmas (Fig. 12c).

Fig. 12
figure 12

Geological sketch diagrams to illustrate the chronological development of subduction and the related genesis of arc-like andesites and contemporaneous mafic rocks in eastern China (modified from Deng et al. 2017). a Deep continental subduction of the Yangtze Craton beneath the North China Craton during the Triassic, generating ultrahigh-pressure metamorphic rocks. During the subduction and exhumation, melts and fluids were generated and penetrated into the overlying SCLM; b, c Subdcution of the paleo-Pacific plate further metasomatized the SCLM that was then melted due to slab rollback, resulting in the development of the andesitic magmatism. See text for further interpretation. SCLM: subcontinental lithospheric mantle; NCC: North China Craton; YC: Yangtze Craton; SOB: Sulu Orogenic Belt; CC: Continental Crust; TLF: Tan-Lu Fault

Full size image

Conclusions

This study of Early Cretaceous andesites in the Sulu UHP belt provides new insights that contribute to our understanding of how an ancient SCLM was metasomatized and reactivated in a complicated terrane that underwent both continental and oceanic subduction. The crystallization age of the andesites at General’s Hill which erupted at 124 to 116 Ma, is coeval with the large-scale mafic magmatism in the Dabie–Sulu orogen and in the interiors of the North China Craton. The andesites contain intermediate amounts of SiO2, MgO, Ni and Cr with high Mg# values, and are characterized by arc-like trace element patterns with enriched whole-rock Sr–Nd and zircon Hf isotope compositions. The hornblende and clinopyroxene phenocrysts show enrichment of Th and U but a pronounced depletion of HFSE, such as Nb and Zr. These features indicate that they were originated from an enriched mantle source of mafic–ultramafic compositions. Relict zircons of Neoproterozoic, Paleoproterozoic and Archean ages and two-stage Nd and zircon Hf model ages of Paleoproterozoic to Neoarchean Era may indicate involvement of crustal rocks from both the Yangtze and North China Cratons into the SCLM by source mixing during the Triassic orogeny. However, a comprehensive lateral comparison of contemporaneously formed mafic–andesitic rocks with the studied andesites reveals that the source rocks were further metasomatized by reactions of SCLM-wedge peridotite with fluids/melts released from subducted marine sediments at the slab–mantle interface during the paleo-Pacific plate subduction in the Jurassic to Cretaceous. Slab rollback and accompanied asthenospheric mantle upwelling are considered as the favored geodynamic mechanisms for the generation of extensive mafic to andesitic magmatism in the Sulu belt of eastern China.