Heat- and melt-fluxed melting of lower continental crust: Insights from two types of subduction-related granitoids in northeastern China and the implications for crustal reworking and growth

Earth’s continental crust is composed predominantly of low-density, buoyant, granitoid rocks that ultimately originated from the mantle (Rudnick and Gao, 2003; Brown, 2013). The generation of granitoid magmas and their subsequent emplacement in the upper crust are a result of the ascent of crustal material from the lower part of the crust or evolved liquids from the upper mantle (Moyen, 2009). The anatexis of pre-existing continental crust may lead to intracrustal differentiation or crustal reworking, whereas if mantle components contribute to the granitoid melts, then the net growth of continental crust would occur (Annen et al., 2006). Active continental margins overriding modern and paleo-subduction zones are favorable sites for the formation of granitoids (Ernst, 2010), and therefore the magmatic processes of these intrusive rocks may provide valuable information regarding the differentiation and growth of continental crust.

Voluminous calc-alkaline granitoids are distributed along the eastern Jilin and Heilongjiang provinces of NE China and adjacent areas (e.g., the Sikhote-Alin orogenic belt of Russia, SAOB), constituting a major magmatic arc segment in the eastern Asian continental margin (Fig. 1A). Available geochronological and geochemical data have revealed that these granitoids were formed chiefly during the Late Triassic to Paleocene (213–56 Ma) and were related to western Pacific subduction systems (Wu et al., 2011; Xu et al., 2013; Grebennikov et al., 2016; Ma et al., 2017). Generally, granitic melts cannot be directly extracted from the mantle in subduction zones because the partial melting of a metasomatized mantle wedge generally leads to the formation of basaltic to magnesian andesitic magmas (Tatsumi, 1982; Grove et al., 2002; Lee and Anderson, 2015). Thus, how mantle-derived mafic magmas finally fractionate into granitic material and produce large granitoid batholiths in the overlying crust has long been a topic of close interest (e.g., Jagoutz et al., 2011). In this contribution, we present new petrological and geochemical data (including zircon U–Pb ages and Sr–Nd–Hf isotopes) for two Jurassic granitoid plutons in the Yanbian region, NE China. Our results reveal two distinct generation processes for subduction-related granitoids: (1) heat-fluxed melting of a preexisting lower arc crust by basaltic underplating, and (2) mixing of lithospheric-mantle-derived magmas and siliceous crustal melts. These two processes represent continental crustal reworking and growth in the eastern Asian continental margin, respectively.

NE China represents the easternmost part of the Central Asian orogenic belt (Fig. 1A) and is located between the tectonic domains of the Paleo-Asian Ocean and the Pacific Ocean (Wu et al., 2011; Xu et al., 2013). NE China is composed of several micro-blocks (Fig. 1B), including the Erguna, Xing’an, and Songliao blocks and the Jiamusi-Khanka massifs (Zhou and Wilde, 2013). From the Neoproterozoic to late Paleozoic, archipelago-type accretionary orogenesis dominated the area of NE China (Xiao et al., 2003; Ma et al., 2019). The amalgamation of multiple terranes (e.g., arc systems and microcontinents) led to various NE China blocks being combined (Zhou and Wilde, 2013). During the Mesozoic, NE China began to be affected by the subduction of the Paleo-Pacific Plate and became a part of a large-scale continental-margin arc. Massive granitoids, mostly calc-alkaline I-type granites and minor A-type granites, were formed throughout NE China during the late Paleozoic and Mesozoic (Wu et al., 2011).

The Yanbian region is located in the southeastern part of NE China (Fig. 1B). Tectonically, this region is transected by the Solonker–Xra Moron–Changchun–Yanji fault, which is regarded as the suture zone formed following the closure of the Paleo-Asian Ocean (Xiao et al., 2003). Basement rocks of the Yanbian region consist principally of metamorphosed Precambrian components (Zhou and Wilde, 2013). Mesozoic intrusions cover an area of more than 15,000 km² (Fig. 1C) and constitute ∼70% of the exposed rocks in the Yanbian region. These rocks consist predominantly of monzogranites, granodiorites, and diorites together with minor mafic intrusive rocks (e.g., gabbros) (Guo et al., 2015). According to previous geochronological studies (Wu et al., 2011), three stages of magmatic activities can be identified in the Yanbian region: the Early Triassic, the Late Triassic to Early Jurassic, and the Cretaceous. Geochemically, most granitoids show calc-alkaline characteristics of arc-type rocks and have a wide variation in ε_Nd(t) values and T_DM ages (depleted-mantle model age), suggesting complicated genetic processes (Wu et al., 2000; Ma et al., 2015).

Two types of granitoids, as represented by the Shiguo monzogranites and Gaoling granodiorites (Fig. 2), were chosen for field, petrological, and mineralogical studies. Both plutons are located in the eastern Helong County of the Yanbian region, NE China (Fig. 2). The Shiguo monzogranites crop out over an area of ∼30 km². These monzogranites intruded into late Permian granitoids and Precambrian metamorphic rocks and were subsequently fragmented by younger intrusions (Fig. 2). No mafic microgranular enclaves (MMEs) or xenoliths are observed in the Shiguo monzogranites (Figs. 3A and 3B). The Gaoling granodiorites have a larger exposed area of ∼280 km² and intrude into the late Permian granites, Early Jurassic gabbroic to dioritic rocks, and Precambrian metamorphic basement (Fig. 2). Locally, the Gaoling granodiorites are covered by Cretaceous strata and Pliocene volcanic rocks. Numerous MMEs can be found in the Gaoling pluton, showing elongated and rounded shapes and diffuse to sharp contacts with their host rocks in outcrops (Fig. 3C and 3D).

The Shiguo monzogranites are medium to coarse grained and have major mineral phases of alkali feldspar (∼35%), plagioclase (∼30%), quartz (∼30%), biotite (5%), and minor hornblende (<3%) (Fig. 3E). The accessory mineral phases are zircon, apatite, titanite, and Fe–Ti oxides. The Gaoling granodiorites show coarse-grained textures and have a mineral assemblage of plagioclase (∼40%), alkali feldspar (∼25%), quartz (∼20%), hornblende (5%–10%), and biotite (5%–10%) (Fig. 3F). The accessory minerals are zircon, apatite, titanite, and Fe–Ti oxides. The MMEs enclosed in the host granodiorites have typical igneous textures (including euhedral, equigranular, and fine grained) and are principally dioritic and gabbroic in composition. These MMEs consist mainly of hornblende (20%–30%), plagioclase (30%–45%), alkali feldspar (∼10%), quartz (5%–10%), and biotite (5%–10%) (Fig. 3G). The accessory minerals are acicular apatite (Fig. 3H), titanite, zircon, and Fe–Ti oxides.

Heavy liquid and magnetic methods were used to separate zircon grains from crushed fresh samples. Zircons were manually picked using a binocular microscope. A polished epoxy mount was used to hold the zircon grains and was covered by a 50 nm coating of high-purity gold in a vacuum environment. The internal structures of zircon grains for in situ isotopic analyses were ascertained by transmission and reflection photos and cathodoluminescence (CL) images. Laser-ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) was used to determine the in situ U–Pb ages of zircons. Age determinations were performed using a Thermo Fisher ELEMENT XR ICP-MS instrument equipped with an UP²¹³Nd: YAG laser (10 Hz) with a beam spot diameter of 30 μm at the National Research Center for Geoanalysis (NRCG), Beijing, China. Helium was used as a carrier gas to deliver the ablated aerosol to the torch. The instrument was optimized using NIST 610, which allowed the maximum signal intensity (²³⁸U and ²³²Th signal intensity of >20, 0000 counts per second) and low oxide production (ThO⁺/Th⁺ of <0.1%) to be obtained and isotopic signal ratios of ²³⁸Th/²³⁸U ≈1 to be achieved, thus reducing the effect of dynamic fractionation during analysis. The detection time for ²³²Th, ²³⁵U, and ²³⁸U was 2 ms and for ²⁰² Hg, ²⁰⁴Pb, ²⁰⁶Pb, ²⁰⁷Pb, and ²⁰⁸Pb was 3 ms.

Single-spot analytical data consisted of background signal collection for 20 s and sample signal collection for 40 s. The zircon standard GJ-1 was analyzed every 10 unknown sample spots, and the zircon standard Plesovice was analyzed twice every 10 samples. Six Plesovice determinations analyzed in this way yielded a concordia age of 337.2 ± 2.8 Ma. GJ-1 analytical spots gave a concordant ²⁰⁶Pb/²³⁸U age of 600.4 ± 5.2 Ma. The software programs Glitter (Version 4.0) and Isoplot were used for data reduction and to construct concordia diagrams. The results are presented in Table 1.

Major- and trace-element analyses were conducted at the Key Laboratory of Orogenic Belts and Crustal Evolution (KLOBCE), Peking University, China. X-ray fluorescence was used to determine abundances of the major elements, which were analyzed as fused glass discs using an ARL ADVANT’XP+ instrument with an accelerating voltage of 50 kV, an accelerating current of 50 mA, and a beam spot diameter of 2.9 mm. The analytical errors are less than 2%. ICP-MS was used to measure trace-element concentrations at the NRCG, Beijing, China. The analytical uncertainties of element abundances below 10 ppm and over 10 ppm are 10% and better than 5%, respectively. The results are presented in Table 2.

The extraction of strontium and neodymium was carried out at the KLOBCE, Peking University, China. HNO₃ + HF digestion was used to dissolve the samples, each of which was placed in a sealed Savillex beaker at a temperature of 80 °C. Rubidium, strontium, and light rare-earth elements (LREEs) were obtained by a cation-exchange column that was packed with Bio-Rad AG50Wx8 resin. The further purification of samarium and rubidium was achieved using a second cation-exchange column with dilute HCl.

Sr–Nd isotopic ratios were analyzed using a negative thermal ionization mass spectrometer by TRITON, at the Tianjin Institute of Geology and Mineral Resources, China Geological Survey. ⁸⁷Sr/⁸⁶Sr ratios and ¹⁴³Nd/¹⁴⁴Nd ratios were normalized to ⁸⁶Sr/⁸⁸Sr = 0.1194 and ¹⁴⁶Nd/¹⁴⁴Nd = 0.7219, respectively. NBS-987 SrCO₃ (⁸⁷Sr/⁸⁶Sr = 0.710250) and JMC Nd₂O₃ (¹⁴³Nd/¹⁴⁴Nd = 0.511122) were used to adjust the ⁸⁷Sr/⁸⁶Sr and ¹⁴³Nd/¹⁴⁴Nd ratios. The concentration uncertainty (2σ) introduced by isotope dilution during the measurement process was 1%–2% for rubidium, 0.5% for strontium, and 0.2%–0.5% (depending on concentration) for samarium and neodymium. The mean blanks during the procedure were Rb = 100 pg, Sr = 400 pg, Sm = 50 pg, and Nd = 50–100 pg. The decay constants used for ⁸⁷Rb and ¹⁴⁷Sm were 0.0142 Ga⁻¹ and 0.00654 Ga⁻¹, respectively. The depleted mantle is assumed to have an isotopic composition following a linear path with εNd(t) = 0 at 4.56 Ga and εNd(t) = +10 currently. The results are presented in Table 3.

In situ Lu–Hf isotopic analyses were performed using a Thermo Finnigan Neptune-plus MC-ICP-MS instrument, which was equipped with a J-100 femtosecond ablation system (Applied Spectra Inc.) housed at the NRCG, Beijing, China. Zircons were ablated with ∼30-μm-diameter ablation pits for 31 s with a repetition rate of 8 Hz at 16 J/cm². ¹⁷⁶Lu has a negligible effect on ¹⁷⁶Hf because of the low ¹⁷⁶Lu/¹⁷⁷Hf in zircon (mostly <0.002). The fractionation coefficient (βYb) was calculated with the mean ¹⁷³Yb/¹⁷²Yb value of single spots and was also used to register the influence of ¹⁷⁶Yb on ¹⁷⁶Hf. The calculation parameters are presented in Table 4 and are described in detail by Ma et al. (2013).

One zircon sample was selected from each of the Shiguo monzogranites, the Gaoling granodiorites, and the MMEs. The majority of zircons are transparent, with grain sizes of 100–200 μm, and some show inherited cores (e.g., in samples YJ-1 and YJ-8). As seen in the CL images of representative zircons (Fig. 4), most of the grains are euhedral and show oscillatory zones typical of a magmatic origin. The Th/U ratios of the three zircon samples are in the range of 0.4–1.2, 0.4–0.9, and 0.4–0.9, respectively, indicating a magmatic origin (e.g., Yang et al., 2007).

Thirty-six analytical spots for zircon sample YJ-1 (a Shiguo monzogranite) gave concordant ²⁰⁶Pb/²³⁸U ages of 181.1–201.7 Ma (Table 1). Several spots were analyzed overlapping different zones, which may give a mixed value of crystallization and inherited/metamorphic ages. Most spots yielded a mean weighted age of 188.9 ± 2.0 Ma (mean square of weighted deviates [MSWD] = 2.4), which is interpreted as the crystallization age of the monzogranites. Twenty-five analytical spots for zircon sample YJ-8 (a Gaoling granodiorite) yielded concordant ²⁰⁶Pb/²³⁸U ages between 169.3 and 173.9 Ma. The mean weighted age of this sample is 171.0 ± 1.4 Ma (MSWD = 0.17), and one spot of 270.7 Ma may be an inherited zircon age. Twenty analytical spots of zircon sample YJ-13 (MMEs of the Gaoling granodiorites) gave concordant ²⁰⁶Pb/²³⁸U ages between 168.4 and 175.8 Ma with a mean weighted age of 173.4 ± 1.2 Ma (MSWD = 0.47). Therefore, MMEs and their host granodiorites have identical crystallization ages within analytical errors, suggesting that their genesis was essentially coeval. In summary, the Shiguo monzogranites and the Gaoling granodiorites were formed during the Early Jurassic and Middle Jurassic, respectively, and have an age gap of ∼18 m.y.

The whole-rock geochemical data are presented in Table 2. The Shiguo monzogranites contain high silica (SiO₂ = 66.8–68.7 wt%) and total alkali (K₂O + Na₂O = 7.4–7.9 wt%), and relatively low MgO (1.3–2.2 wt%), Al₂O₃ (14.7–15.4 wt%), and CaO (2.7–3.1 wt%) contents. The monzogranites are metaluminous (A/CNK = 0.94–0.99) and show high-K calc-alkaline characteristics. As shown in Harker diagrams for major and trace elements versus SiO₂ content, these monzogranites exhibit a restricted range of variation (Fig. 5).

The total REE abundance of the Shiguo monzogranite samples is 124.4–182.8 ppm (Table 2). In the chondrite-normalized REE patterns (Fig. 6A), these monzogranites show enrichment in LREEs relative to heavy REEs (HREEs). The samples show moderate negative Eu anomalies (0.49–0.71) (Fig. 6A) and enrichment in large-ion lithophile elements (LILEs; e.g., Rb, Th, U, and Pb) and depletion in high-field-strength elements (HFSEs; e.g., Nb, Ta, and Ti) (Fig. 6B), with all of these chemical compositional characteristics being similar to those of the average upper continental crust (Rudnick and Gao, 2003). Moreover, the Shiguo monzogranites have high zircon saturation temperatures (T_Zr) (Watson and Harrison, 1983), ranging from 785 to 840 °C (with most >800 °C; Table 2).

Compared with Shiguo monzogranites, Gaoling granodiorites have relatively low SiO₂ (60.4–66.2 wt%) and total alkali (Na₂O + K₂O = 6.3–7.8 wt%) contents and high MgO (1.6–2.6 wt%), Al₂O₃ (15.7–16.6 wt%), and CaO (3.4–4.3 wt%) contents. These granodiorites are metaluminous, with A/CNK = 0.91–0.98 and A/NK = 1.56–1.76, and they exhibit medium- to high-K calc-alkaline characteristics. MMEs enclosed in the Gaoling granodiorites are dioritic to gabbroic in composition and, compared with the host granodiorites, have lower SiO₂ (51.1–60.1 wt%) contents and higher MgO (3.0–4.6 wt%), CaO (5.2–7.3 wt%), and Al₂O₃ (17.2–19.5 wt%) contents (Table 2). These MMEs have high Na₂O/K₂O ratios (1.9–3.5), with total alkali contents between 6.1 and 7.5 wt%. MMEs show wide ranges of chemical compositions and exhibit a reasonably linear relationship with their host rocks in the Harker diagrams (Fig. 5).

For trace-element compositions, the Gaoling granodiorites and associated MMEs have quite similar characteristics. In the diagram of chondrite-normalized REE patterns (Fig. 6C), these granodiorites and MMEs show enrichment in LREEs and moderate enrichment in medium REEs (MREEs) compared with HREEs and exhibit slight negative Eu anomalies (0.71–0.96) (Fig. 6C). Moreover, these rocks show enrichment in LILEs and depletion in HFSEs (Fig. 6D). It is worth noting that the Gaoling granodiorites have high Sr (535–654 ppm) but low Y (10–26 ppm) contents, and most of them show adakite-like signatures. In Sr/Y versus Y and (La/Yb)_N versus Yb_N diagrams (Fig. 7), some samples plot in the adakite field on account of their high Sr/Y (21–56) and La/Yb (21–33) ratios and low Y (mostly <18 ppm) and Yb (<1.9 ppm) abundances. Moreover, calculated T_Zr values of the Gaoling granodiorites lie between 748 and 799 °C (Table 2) and are generally lower than those of the Shiguo monzogranites.

Initial ⁸⁷Sr/⁸⁶Sr (I_Sr) ratios of the Shiguo monzogranites are between 0.70378 and 0.70413, and the ε_Nd (189 Ma) values range from +2.3 to +2.5 (Table 3; Fig. 8), with T_DM2 values of 770–780 Ma. The Gaoling granodiorites have different isotopic compositions from those of the Shiguo monzogranites, with I_Sr ratios between 0.70545 and 0.70564 and ε_Nd (171 Ma) values from −1.5 to −1.3. Calculated T_DM2 values for the Gaoling granodiorites lie between 1065 and 1081 Ma. The MMEs have I_Sr ratios of 0.70537–0.70583 and ε_Nd (173 Ma) values ranging from −2.1 to −0.5, which are almost indistinguishable from those of the host Gaoling granodiorites.

Both Shiguo and Gaoling samples plot in the field of isotopic compositions of Phanerozoic granitoids in NE China (Fig. 8), which are generally considered to have involved large volumes of juvenile components (Wu et al., 2000). Notably, the Shiguo monzogranites have more depleted isotopic compositions and younger T_DM2 ages compared with the Gaoling granodiorites, indicating that the two plutons probably originated from distinct sources or underwent different petrogenesis processes, as detailed in the discussion below.

Zircon Hf isotopic data are reported in Table 4 and presented in Figure 9. The Shiguo monzogranites show fairly homogeneous Hf isotopic compositions. The initial ¹⁷⁷Hf/¹⁷⁶Hf ratios range from 0.28286 to 0.28294, and the ε_Hf (189 Ma) ratios are between +7.3 and +10.2 (Fig. 9A), with young T_DM2 ages varying from 576 to 764 Ma (Fig. 9B). In contrast, the Gaoling granodiorites show wide variation in initial ¹⁷⁷Hf/¹⁷⁶Hf ratios (0.28257–0.28282) and ε_Hf (171 Ma) ratios (−3.5 to +5.6) (Fig. 9C), with T_DM2 ages varying between 859 and 1439 Ma (Fig. 9D). The initial ¹⁷⁷Hf/¹⁷⁶Hf ratios and ε_Hf (173 Ma) values of the MMEs range from 0.28264 to 0.28287 and from −1.0 to +7.1 (Fig. 9E), respectively. We also carried out in situ analyses in different sites on a single zircon grain for each rock type to gauge the Hf isotopic variation. The results reveal a wide variation in ε_Hf (t) values (8–10 ε units) from the core to the rim for a single zircon from each of the Gaoling granodiorites and MMEs. However, this wide variation was not detected for the zircon grain from the Shiguo monzogranites (Table 4).

In summary, the Shiguo monzogranites have positive and uniform ε_Hf (t) values (Fig. 9), indicating a depleted source dominated by juvenile material. In contrast, the Gaoling granodiorites have negative to positive ε_Hf (t) values (Fig. 9), suggesting derivation from an open magma system or from multiple sources, as detailed in the discussion below.

The Shiguo monzogranites host diagnostic hornblende (Fig. 3) and show metaluminous (A/CNK = 0.94–0.99) and calc-alkaline characteristics, indicating that they can be classified as I-type granitoids. Two main models for the genesis of I-type granitoids in NE China have been suggested (Wu et al., 2000): (1) the fractional crystallization of primary-mantle-derived magmas, and (2) the partial melting of previously accreted lower-crustal rocks. Experimental results of mantle-rock melting indicate that such melts are chiefly basaltic to magnesian andesitic in composition (Parman and Grove, 2004). However, the Shiguo monzogranites have high SiO₂ (66.8–68.7 wt%) contents, as well as low MgO (1.3–2.2 wt%), Cr (<18 ppm), and Ni (<8 ppm) contents, and exhibit characteristics typical of upper continental crust (e.g., enrichment in K, Rb, and Pb) (Fig. 8; Rudnick and Gao, 2003), thus precluding the possibility that they represent products directly derived from the mantle. Moreover, to produce granitic magmas with SiO₂ contents of ≥60 wt% through a fractionation process, at least 60% crystallization of complementary mafic rocks is required (Foden and Green, 1992). However, compared with the widespread occurrence of granites, the requisite volumes of mafic rocks are limited in the Yanbian region. In fact, only minor gabbroic plutons are sporadically developed (Guo et al., 2015). Furthermore, coeval intermediate rocks (e.g., diorites and andesites) are also generally lacking in this region. Therefore, the absence of coeval complementary mafic to intermediate rock types suggests that the Shiguo monzogranites are likely not residual melts formed through the fractionation of a mantle-derived mafic magma. Instead, their highly fractionated chemical features, together with the presence of inherited zircons in these rocks, lead us to infer that the Shiguo monzogranites were likely derived from the melting of preexisting crustal rocks.

Extensive investigations have demonstrated that Phanerozoic granitoids in the Central Asian orogenic belt typically exhibit low I_Sr values, positive ε_Nd(t) values, and young model ages (most <1000 Ma; Jahn et al., 2000; Kröner et al., 2008). Geochemically, most of these Phanerozoic granitoids show highly fractionated features, and some researchers have therefore suggested that they may have formed via a two-stage process (Wu et al., 2000): (1) the generation of a relatively juvenile lower crust by basaltic underplating; and (2) the subsequent remelting of newly accreted material. Previous studies have shown that late Paleozoic mafic intrusions (Guo et al., 2016) in the Yanbian area and the Mashan Group (Zhou and Wilde, 2013; Wilde, 2015) in adjacent areas may represent the components of local lower crust. The Shiguo monzogranites of the present study have highly depleted isotopic compositions, indicated by low I_Sr ratios (0.70378–0.70413) and positive ε_Nd (t) (+2.2 to +2.3) and ε_Hf (t) (+7.3 to +10.2) values, which suggest a relatively juvenile source. We further infer that the source was probably dominated by underplated basaltic rocks, as the Shiguo monzogranites have geochemical compositions (e.g., high silica and particularly low MgO) similar to those of experimental melts generated from the dehydration melting of mafic rocks (Rapp and Watson, 1995).

As discussed above, the Shiguo monzogranites are classified as I-type granites and show geochemical features analogous to those of typical arc magmatic rocks (Chappell and White, 1992). These observations, together with the depleted isotopic compositions, imply that these monzogranites probably have inherited arc affinities. Their source, therefore, was probably dominated by arc series that formed during the preceding subduction (Chen and Arakawa, 2005). In Sr/Y versus Y and (La/Yb)_N versus Yb_N diagrams (Fig. 7), the Shiguo monzogranites plot in the field of normal arc rocks because of their low Sr (249–287 ppm) and La (21–42 ppm) contents, indicating that the melting of the source rocks took place at relatively low pressure. Moreover, Miller et al. (2003) found that the dehydration melting of crustal rocks by the heat flux generated from underplating basaltic magmas could produce granitoids with high T_Zr values (>800 °C; mean 837 °C), whereas granitoids with low T_Zr values (<800 °C; mean 766 °C) probably require the influx of fluids and/or melts because the melting of hydrous minerals (e.g., hornblende and biotite) usually takes place at relatively low temperatures. The Shiguo monzogranites have relatively high T_Zr values (>800 °C with one sample at 785 °C), which is consistent with a process of dehydration melting of the lower crust. Therefore, the advective heat supplied from underplating basaltic magmas may have played a significant role in the genesis of the monzogranites.

In summary, it is inferred that the source of the Shiguo monzogranites was dominated by arc-type lower-crustal rocks. The underplating of basaltic magmas derived from lithospheric mantle wedge may have resulted in the heat-fluxed melting of juvenile arc-type crustal material in association with the subduction of the Paleo-Pacific Plate during the Jurassic (Fig. 10A).

In contrast to the Shiguo monzogranites, the Gaoling granodiorites contain abundant MMEs (Figs. 3C and 3D). Four genetic types have been proposed for the MMEs enclosed in felsic rocks: (1) mineral cumulates that formed via fractional crystallization (Collins et al., 2006;); (2) residual and refractory materials of the source rocks carried to shallow crustal levels (Chappell and White, 1992); (3) exotic xenoliths from wall rocks where magmas passed through or emplaced (Maas et al., 1997); and (4) the incomplete mixing of mafic magmas and felsic melts (Yang et al., 2007; Clemens and Stevens, 2012). Specifically for the MMEs in the Gaoling granodiorites, cumulate textures are lacking, which is at variance with the expected features of fractional crystallization. Residual phases after partial melting should have metamorphic textures. However, MMEs show typical igneous textures (Fig. 3G), and therefore the residual model is also not supported. If the MMEs were exotic xenoliths captured from wall rocks, they should exhibit solid-state deformation features and have a crystallization age older than that of the host rock, which is inconsistent with our data. Therefore, we favor the model postulating the MMEs being products of incomplete mixing (Clemens and Stevens, 2012). This is supported by the following lines of evidence: (1) The Gaoling MMEs are elongated or rounded in shape (Figs. 3C, 3D), indicating a partially crystallized, convective flowing magma; (2) the MMEs exhibit fine-grained and equigranular textures (Fig. 3G) unlike the coarser-grained host granodiorites (Fig. 3F), which, coupled with the presence of needle-like apatite in the MMEs (Fig. 3H), is suggestive of rapid crystallization when the mafic magmas were added into a cooler host felsic magma; (3) the MMEs have identical crystallization ages to those of the host granodiorites (Fig. 4), establishing that they were coeval; (4) the wide variation in whole-rock geochemistry and the linear relationship between MMEs and their host rocks in Harker diagrams (Fig. 5) indicate a binary mixture; (5) the MMEs have a wide range of zircon ε_Hf (t) values, including that shown by a single zircon grain from core to rim (>6 ε units), which together with the bimodal character of ε_Hf (t) values (Fig. 9E) leads us to infer that two magma end-members were involved in MME genesis.

It is clear that many intermediate to felsic magma chambers are not isolated and closed systems and may be fed by more mafic magmas with distinct chemical or isotopic compositions (Xu et al., 2004). MMEs can thus be regarded as products of incomplete mixing (Vernon, 1984). As discussed above, the MMEs hosted in the Gaoling granodiorites are magmatic enclaves (not xenoliths, cumulates, or refractory materials) and therefore likely represent mafic end-member components injected into the host granodiorite magmas (Fig. 10B). Low silica (mostly 51–54 wt%) and relatively high MgO contents, as well as high ε_Hf (t) values (up to +7.1), indicate an isotopically depleted mantle source. It should be noted that the MMEs and the host granodiorites share similar whole-rock Sr–Nd isotopic characteristics, whereas their zircon in situ Hf isotopic compositions clearly differ (Fig. 9). This may be explained by Sr–Nd isotopes reaching a homogeneous state in both the host magmas and the MMEs after mixing/mingling (Griffin et al., 2002). In contrast, zircon can preserve primary Lu–Hf isotopic signatures during crystallization processes, as it is resistant to thermal disturbances after crystallization (Blichert-Toft, 2001). The maximum ε_Hf (t) value of MMEs is +7.1, similar to that of the coeval Tumen mafic complex, which originated from a metasomatized mantle wedge in association with the subduction of the Paleo-Pacific Plate (Guo et al., 2015). Moreover, the occurrence of euhedral hornblendes (Fig. 3G) suggests a hydrous MME magma. This, along with the arc-type geochemical features of the MMEs, such as selective enrichment in LILEs and negative Nb and Ta anomalies (Fig. 6), indicates that the parental magma of MMEs was probably derived from metasomatized lithospheric mantle in a subduction setting.

Similarly, evidence for mixing can be observed in the host granodiorites. Such evidence includes the linear chemical variation trends with MMEs in Harker diagrams (Fig. 5) and the wide range of zircon ε_Hf (t) values (−3.5 to +5.6) (Fig. 9). As the Sr–Nd isotopic compositions of the host granodiorites show no correlation with silica content, crustal contamination is inferred to have been negligible (Liu et al., 2010). Both chemical and isotopic variation may result from the mixing process. In this case, the host granodiorites were formed by partial crustal felsic melts with the addition of mantle-derived material. However, we consider that the crustal source of the Gaoling granodiorites differed from that of the Shiguo monzogranites, based on the following evidence. First, as mentioned above, the Shiguo monzogranites have depleted Sr–Nd isotopic features with ε_Nd (t) = +2.3 to +2.5, in contrast to the Gaoling granodiorites (ε_Nd (t) = −1.5 to −1.3) (Fig. 8). Second, the minimum ε_Hf (t) value of the Gaoling granodiorites is −3.5 (Fig. 9), which is distinct from the uniform positive ε_Hf (t) values of the Shiguo monzogranites. Therefore, compared with the Shiguo monzogranites, the crustal source of the Gaoling granodiorites contains an amount of ancient components.

Some of the Gaoling granodiorite samples show adakite-like signatures (e.g., high Sr/Y and La/Yb ratios) (Fig. 7). Adakitic features may be acquired through source melting under eclogite-facies conditions (Defant and Drummond, 1990), and two models have been suggested for the genesis of rocks with such features: (1) the partial melting of young subducting oceanic crust (Defant and Drummond, 1990), and (2) the melting of thickened mafic lower crust (Atherton and Petford, 1993). In the case of the Gaoling granodiorites, the melting of subducting oceanic crust can be discounted because these granodiorites do not show the depleted isotopic compositions of mid-ocean-ridge basalt. The melting of a thickened lower crust is also not favored for the following reasons. First, the Gaoling granodiorites have Mg# values (>45) that are higher than those of experimental melts (<40) from basaltic rocks for the same silica content (Rapp and Watson, 1995). Second, melts generated in the thickened lower crust under eclogite-facies conditions display extremely depleted HREEs relative to MREEs (Xu et al., 2002; Chung et al., 2003). However, the Gaoling granodiorites have moderate MREE/HREE ratios (e.g., Sm/Yb = 3.5–4.1) and show concave REE patterns (Fig. 6C), indicating a garnet-free source (Moyen, 2009).

We therefore suggest that the adakite-like geochemical signatures of the Gaoling granodiorites could have been caused by a process of magma mixing and subsequent fractional crystallization, as proposed by Guo et al. (2007) and Chen et al. (2013). As represented by the MMEs, the melting of a metasomatized lithospheric mantle could produce melts with high contents of mobile elements (e.g., LILEs) (Gibson et al., 1995). The involvement of such melts in crustal felsic magmas would result in the formation of a hybrid parental magma with enriched LILEs and LREEs (Guo et al., 2007; Chen et al., 2013). Thus, the high Sr and La contents and high Sr/Y ratios of the host granodiorites may have resulted in part from such a mixing process. Moreover, some ferromagnesian phases (e.g., hornblende and clinopyroxene) and accessory minerals (e.g., titanite and apatite) have low D_Sr values and high D_Y values (Rollinson, 1993), and therefore the fractional crystallization of these minerals could have effectively magnified the adakitic signatures. To test this hypothesized process, we conducted modeling based on the Rayleigh Law, and the results of this modeling are shown in Figure 7A. We found that the magmas of granodiorites evolve following a curved trend of fractional crystallization with a mineral assemblage of hornblende (45%) + clinopyroxene (35%) + plagioclase (10%) + apatite (5%) + titanite (5%). Therefore, we infer that the Gaoling granodiorites were probably formed by magma mixing plus subsequent hornblende-dominated fractional crystallization rather than by the partial melting of subducting slab or thickened lower crust as previously thought (Samaniego et al., 2002). Similar petrogenetic processes have also been inferred for adakite-like rocks from the volcanoes of southern Colombian and the northern Andes of Ecuador in the eastern Pacific arcs (e.g., Calvache and Williams, 1997; Chiaradia, 2009).

Chemical differentiation of the continental crust can take place through a process of partial melting of the lower to middle crust and the subsequent transfer of those melts to the upper crust (Sawyer, 1998). Generally, such reworking is used to explain those granitoids formed by the partial melting of ancient metasedimentary rocks (Jayananda et al., 2006; Zheng and Zhang, 2007), whereas the influence of juvenile material is usually underestimated or disregarded. Here, the formation of the Shiguo monzogranites strongly suggests the reworking of juvenile metaigneous protoliths in the lower crust. Importantly, this process might be a common mechanism for the formation of voluminous positive ε_Nd (t) granitoids of NE China (Wu et al., 2000).

Numerous Mesozoic granitoids in NE China have positive ε_Nd (t) values (0 to +4) and young T_DM model ages (500–1000 Ma), as well as low I_Sr ratios (0.705 ± 0.001) (Jahn et al., 2000), indicating a significant contribution from juvenile material. The majority of these granitoids were formed by the melting of juvenile lower crust rather than by evolved products sourced directly from the mantle (Wu et al., 2000). In fact, as for the easternmost part of the Central Asian orogenic belt, the area of NE China has undergone a complex evolution within both the Paleo-Asian and Paleo-Pacific tectonic regimes (Wilde, 2015). Arc series components (mostly arc rocks and underplates) have significantly modified the constituents of the crust in NE China (Chen and Arakawa, 2005). Sr–Nd isotopic modeling shows that 55%–85% of the lower crust basement in NE China has been replaced by accreted basaltic material (Wu et al., 2000). Subduction of the Paleo-Pacific Plate triggered the partial melting of the lithospheric mantle and the underplating of basaltic magmas beneath the lower crust, which provided sufficient heat to partially melt earlier-accreted material. Although the occurrences of contemporaneous mafic intrusive complexes (e.g., the Tumen pluton; Guo et al., 2015) imply that minor basaltic magmas were emplaced in the shallow crust, significant portions of basaltic magmas were underplated beneath the lower crust (Wu et al., 2000). Accordingly, large parts of the Mesozoic granitoids, especially highly fractionated ones, were generated by the heat-fluxed melting of lower-crustal rocks, induced by new underplating of basaltic magmas (Fig. 10A). Such granitoids represent reworked juvenile continental crust. Importantly, this process could be a dominant mechanism for producing the constituents of upper continental crust, as these granitic rocks show geochemical characteristics typical of upper continental crust (Rudnick and Gao, 2003).

In addition to granitoids that resulted from the melting of predominantly lower crust, a number of granitoids in NE China are of hybrid origin, as represented by the Gaoling granodiorites of the present study. Recent studies have increasingly recognized such hybrid granitoids (Yang et al., 2007; Li et al., 2009; Chen et al., 2013). Unlike the reworking-type granitoids (e.g., the Shiguo monzogranites), which represent magmas from solely lower-crustal sources, mantle-derived components were involved in the genesis of the Gaoling granodiorites, which were formed as a result of a melt-fluxed melting process caused by the addition of mantle material (Miller et al., 2003). If so, the crustal growth could have occurred in two ways: (1) the direct intrusion of mantle-derived magmas after fractional crystallization (Grove et al., 2002), or (2) the mixing of mantle- and crust-derived magmas (Clemens and Stevens, 2012; Chen et al., 2013).

In the Yanbian region, the first of these two cases is represented by the Tumen mafic intrusive complexes, which originated from a metasomatized lithospheric mantle wedge (Guo et al., 2015). However, compared with granitoid rocks, the volumes of mafic and intermediate rocks formed in the middle to upper crust through fractional crystallization are relatively limited (Tatsumi, 1982). As discussed above, experimental studies of mantle melting indicate that underplated material comprises mainly basalts to magnesian andesites (Tatsumi, 1982; Lee and Anderson, 2015). However, the average composition of the bulk continental crust is andesitic. An effective way is to remove the more mafic cumulates or eclogite-phase rocks from the lower crust (Kay and Kay, 1993). Such removal would cause a compositional transition from basaltic to andesitic for the bulk continental crust as the mafic/ultramafic cumulates or residual material (e.g., dense eclogites) sink back to the mantle (Lee and Anderson, 2015).

The mixing of mantle- and crust-derived magmas, the second of the above two cases, has been suggested as an alternative process for the genesis of arc rocks in active continental margins and as a significant mechanism for continental crustal growth (Chiaradia, 2009). Mixing could take place either in the crust–mantle transition zone or in an upwelling magma chamber (Hildreth and Moorbath, 1988). On the basis of numerical modeling and experimental data, Annen et al. (2006) argued that the observed variety of arc rocks is controlled largely in the lower crust by the mixing of two end-members: (1) residual H₂O-rich melts after the fractional crystallization of mantle-derived basaltic magmas, and (2) siliceous melts formed by the partial melting of lower-crustal rocks. With regard to the Gaoling granodiorites of the present study, it is clear that mantle-derived mafic material was involved in the siliceous crustal magmas (Fig. 10B), with the former undergoing mechanical and/or chemical exchange with the latter. A calculation of the relative contribution made by each of these two end-members to the formation of Gaoling granodiorites (shown in the diagram of I_Sr versus ε_Nd(t); Fig. 8) indicates that mantle material accounts for ∼60%–70% of the hybrid magmas by volume. Importantly, the proposed mixing process (or melt-fluxed melting) is likely to be a significant mechanism for crustal growth in active continental-margin settings, as such a process tends to produce andesitic hybrid magmas with chemical compositions close to those of the bulk continental crust.

1. New zircon U–Pb results indicate that two subduction-related granitoids (enclave-free monzogranites and enclave-bearing granodiorites) in NE China were emplaced at 189 Ma and 171 Ma.

2. The monzogranites show characteristics typical of upper continental crust and have high T_Zr values, potentially formed by the heat-fluxed melting of previously accreted lower crust. The granodiorites have relatively low silica contents and low T_Zr values, possibly generated by mixing between mantle- and crust-derived magmas. The adakite-like signatures of the granodiorites may have resulted primarily from the involvement of enriched mantle material and subsequent fractional crystallization rather than from the melting of a subducting slab or thickened lower crust.

3. The two types of granitoids (monzogranites and granodiorites) represent crustal reworking and growth in the eastern Asian continental margin, respectively. This finding has significant implications for the understanding of the formation of silica-rich upper continental crust and andesitic bulk continental crust.

We thank Wengang Liu and Chao Li for their assistance in U–Pb and Sr–Nd–Hf isotope analyses. We gratefully acknowledge the constructive suggestions from Science Editor Kurt Stüwe and reviewers Feng Guo and Simon Schorn, which helped us improve the manuscript. This research was supported by Fundamental Research Funds for the Central Public Welfare Research Institutes (K1502), the Open Research Project from the State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences (GPMR201816), the National Natural Science Foundation of China (41572054, 41911530106), and the National Key R&D Plan (2018YFC0603801, 2017YFC0601403).