Disequilibrium partial melting of metasediments in subduction zones: Evidence from O-Nd-Hf isotopes and trace elements in S-type granites of the Chinese Altai

As two key processes in crustal evolution, crustal melting and granitic magma formation both significantly facilitate the transfer of heat and volatile elements in the crust, contributing to the mechanically weakened crust throughout orogenic episodes. It is critical to properly decipher the origin of granites in order to better understand continental differentiation and their genetic link to contemporaneous mineralization (e.g., Brown, 2013; Kemp et al., 2007; Kemp and Hawkesworth, 2003). Well-known S-type granites are generally characterized by a large molar ratio of Al₂O₃/(CaO + Na₂O + K₂O) (i.e., A/CNK > 1.1), high SiO₂ content, and low sodium content (Na₂O < 3.2 wt%; Searle et al., 1997; Chappell and White, 2001; Gou et al., 2015). Previous studies attributed the formation of S-type granites to the partial melting of sedimentary rocks (Chappell and Simpson, 1984; Chappell and White, 2001), which generally occurred in post- or syncollisional settings (Sylvester, 1998; Barbarin, 1999). Examples of these include the Himalayas (Searle et al., 1997, 2009), European Hercynides (Finger et al., 1997; Bea et al., 1999; Poller et al., 2002), and Caledonides (Sylvester, 1998, and references therein). However, the past decades have witnessed the emergence of the argument that S-type peraluminous granites could form in various geodynamic processes, including ridge subduction (Cai et al., 2011a. 2011b), mantle plume formation (Li et al., 2003), collisional orogenic climaxing (Barbarin, 1999; Nabelek and Liu, 2004), and back-arc extension (Collins and Richards, 2008; Kemp et al., 2009).

As a ubiquitous, Hf-enriched, highly refractory accessory mineral in S-type granites (Tang et al., 2014), zircon is favored for analysis because of its capacity to preserve the isotopic compositions of parent magma during complex magmatic processes (X.L. Wang et al., 2014). Recent studies have reported the decoupling of the Hf isotope system from other radiogenic systems in several scenarios, including that in granites (Zheng et al., 2008), that in metasomatized mantle materials (Choi and Mukasa, 2012), and that in subduction zone magmas (Hoffmann et al., 2011; Su et al., 2015). Mixing of magmas from different sources was argued to play a role in such decoupling (Kemp et al., 2007; Nebel et al., 2007), and recently more attention has been paid to disequilibrium melting, which might contribute to this decoupling (Tang et al., 2014; Yu et al., 2017). In this context, the zircon dissolution rate is a critical value deserving comprehensive evaluation.

Situated among the Siberian, East European, Tarim, and North China cratons, the Central Asian orogenic belt, also referred to as the Altaids, is the largest Phanerozoic accretionary orogenic belt in the world (Fig. 1A; Sengör and Natal’in, 1996; Badarch et al., 2002; Jahn et al., 2004; Safonova et al., 2004; Yuan et al., 2007; Xiao et al., 2008; Liu et al., 2017, 2018a, 2018b). Its evolution and development have been widely attributed to the episodic accretion of island arcs, ophiolites, accretionary complexes, seamounts, and microcontinental blocks during the Paleozoic (Chen and Jahn, 2004; Xiao and Kusky, 2009a, 2010; Biske and Seltmann, 2010; Wong et al., 2010; Yang et al., 2015; Windley and Xiao, 2018). As the key part of Central Asian orogenic belt, the Chinese Altai, herein referring to the part located in Russia, northeastern Kazakhstan, northwestern China, and southwestern Mongolia (Chen et al., 2002), records the accretion of the peri-Siberian orogenic system in the early Paleozoic, as well as collision of the Kazakhstan–South Mongolian orogenic system. During this process, large volumes of Paleozoic granites associated with subduction and postcollisional settings were emplaced in the Chinese Altai, in which extensive Nd-Hf isotopic decoupling has been observed (Cai et al., 2011a, 2011b; Yu et al., 2017; Zhang et al., 2017). In order to better understand the driver of such Nd-Hf isotopic decoupling, this study implemented comprehensive analyses, including new zircon U-Pb dating, Nd-Hf-O isotopic composition measurements, and major- and trace-element geochemical analysis, on representative samples from five arc-related granites in the Chinese Altai.

The Altai orogenic belt in northern Xinjiang, China, lithologically consists of Cambrian–Devonian clastic sedimentary and volcanic rocks and their metamorphic equivalents, and it is bounded by Mongolia in the east and Kazakhstan and Russia in the west (Fig. 1B; Windley et al., 2002, 2007; Xiao et al., 2004; Long et al., 2007, 2008a, 2008b; Sun et al., 2008; Zhang et al., 2017). Geologically, the whole Altai orogenic belt is thought to be composed of several parts, including the Gorny Altai (the Russian part), the Rudny Altai (vicinity of the Russia-Kazakhstan boundary), the East Kazakhstan or Kalba area (the Kazakhstan part), the Chinese Altai (the Chinese part), and the Mongolian Altai (the Mongolian part; Buslov et al., 2004). Geographically in the center of the Altai orogenic belt, the Chinese Altai has been highlighted due to its close tectonic interrelations with the Junggar continental microblock (Long et al., 2010; Sun et al., 2009). Five fault-bounded terranes can be further divided within the Chinese Altai, based on consideration of stratigraphy, metamorphism, deformation pattern, magmatic activity, and geochronology, and these terranes are separated by four large faults, namely, the Hongshanzui fault, the Kalaxianger fault, the Abagong-Kurit fault, and the Maerkakuli fault (Fig. 1B; Sengör et al., 1993; Long et al., 2007; Yuan et al., 2007; Sun et al., 2008; Cai et al., 2011a, 2011b). Specifically, terrane 1 primarily consists of Late Devonian to early Carboniferous clastic sediments, limestones, and some minor island-arc volcanic rocks metamorphosed at a low greenschist facies; terrane 2 is dominantly composed of a Middle Ordovician turbidite sequence of lower greenschist facies; terrane 3 is the largest one and is prevalently composed of early Paleozoic sediments metamorphosed at medium to high grades; terrane 4 principally consists of Devonian turbiditic sandstones, pillow basalts, and some siliceous volcanic rocks; and terrane 5 is mainly composed of Devonian fossiliferous successions that are, in turn, overlain by the late Carboniferous formations.

Zircon U-Pb dating in the igneous rocks suggests widespread ages of magmatism in the Altai orogenic belt, especially in the Chinese Altai, continuously from the early Paleozoic to the early Mesozoic (Fig. 1; Han et al., 1997; Chen and Arakawa, 2005; Zhu et al., 2006; Briggs et al., 2007; Long et al., 2007; Yuan et al., 2007; Sun et al., 2008; Cai et al., 2011a, 2011b, 2012; Ren et al., 2011; Lv et al., 2012; Zhang et al., 2017). Previous studies also pinpointed the sequential emplacement of early Paleozoic I-type, S-type, and volcanic-arc granites in the Chinese Altai, which reached a peak in the Devonian.

The samples for this study were collected from five freshly exposed outcrops with granite intrusions in the southern Altai domain, with sampling locations shown in Figure 1B. The granites are generally light gray in color, with massive structures and granitic textures (Fig. 2). The Tuerhong granites consist of (in descending order) plagioclase (35–40 vol%), K-feldspar (25–35 vol%), quartz (25–30 vol%), biotite (5–7 vol%), muscovite (5 vol%), and a very small amount of epidote as the accessory mineral (Figs. 2A and 3A). The Keshikusite granites are composed of plagioclase (40–45 vol%), K-feldspar (20–30 vol%), quartz (15–20 vol%), muscovite (5–10 vol%), and biotite (4–5 vol%; Figs. 2B and 3B). Granites from the Aleretuobie pluton consist of quartz (25–30 vol%), plagioclase (20–25 vol%), K-feldspar (15–20 vol%), muscovite (10 vol%), and biotite (2 vol%), and a very small amount of garnet as the accessory mineral (Figs. 2C and 3C). The Tuoputieke granites are composed of plagioclase (35–40 vol%), quartz (both 25–30 vol%), K-feldspar (both 25–30 vol%), muscovite (10–15 vol%), and biotite (1–3 vol%), and a very small amount of sillimanite as the accessory mineral (Figs. 2D and 3D). The Tasitake granites consist of quartz (35–40 vol%), plagioclase (25–30 vol%), K-feldspar (10–15 vol%), biotite (3–6 vol%), and muscovite (4–5 vol%), and a very small amount of sillimanite as the accessory mineral (Figs. 2E and 3E).

Comprehensive major- and trace-element measurements were conducted at the State Key Laboratory of Ore Deposit Geochemistry, Institute of Geochemistry, Chinese Academy of Sciences (IGCAS). Specifically, major elements were characterized through inductive coupling of a Leeman Prodigy spectrometer and a plasma-optical emission spectrometer with high-dispersion Echelle optics. It was assumed that analytical uncertainties of most oxides of corresponding major elements should be smaller than 1%, except for the oxides of Ti and P (specifically, uncertainty of TiO₂ is ∼1.5%, while that of P₂O₅ is ∼2.0%), according to the U.S. Geological Survey (USGS) Rock Standards BCR-1 and AVG-2, as well as the Chinese National Rock Standard GSR-3. As for trace-element analyses, an Agilent 7500a inductively coupled plasma–mass spectrometer (ICP-MS) was employed, the data quality of which was determined through analysis of USGS rock reference materials BCR-1 and BHVO-1. Compared to the uncertainties of major-element measurements, those of most trace-element uncertainties were larger, generally above 5%.

Zircon U-Pb dating was implemented at the State Key Laboratory of Isotopic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIGCAS). First, heavy-liquid and magnetic methods were used to concentrate zircon grains that were collected from Chinese Altai granites, followed by handpicking separation under a binocular microscope. Then, zircon grains were mounted in epoxy and polished down to approximate half sections. Cathodoluminescence (CL) and optical microscopy were used to locate those zircon grains that had clearer and less-fractured rims, which were thought to be good candidates for laser-ablation analyses. Then, placed in the two-volume sample cell flushed with Ar and He, samples were analyzed by laser ablation with a constant energy of 80 mJ, constant frequency of 8 Hz, and a spot diameter of 31 μm. The ablated material was carried by He gas to the Agilent 7500a ICP-MS for U-Pb isotopic analysis. Mass bias drift was corrected with reference to the standard glass NIST 610 (Pearce et al., 1997). Temora was used as the age standard (²⁰⁶Pb/²³⁸U = 416.8 Ma; Black et al., 2003). Concentrations of U, Th, and Pb were obtained by normalizing their count rates to that of Si, assuming that SiO₂ was stoichiometric in zircon (Tu et al., 2015). Finally, the Isoplot program was used to calculate apparent and discordia U-Pb ages (Ludwig, 2003). Individual analysis uncertainty is reported with 1σ, and the weighted mean ²⁰⁶Pb/²³⁸U age was computed at the 2σ level.

In order to determine the Nd isotopic features, mixed isotope tracers were used to spike the sample powders, and then Teflon capsules containing HF and HNO₃ acids were introduced for sample dissolution. Nd isotopes were then separated from rare earth element (REE) fractions on HDEHP (di-(2-ethyl hexyl)phosphoric acid) columns with a 0.18 N HCl elutant. Then, isotopic compositions were measured on a thermal ionization mass spectrometer (TIMS) at the Experimental Test Center, Tianjin Institute of Geology and Mineral Resources, Tianjin, China. According to this study, ¹⁴³Nd/¹⁴⁴Nd ratios were determined to be 0.710240 ± 15 (2σ) and 1.20032 ± 30 (2σ) for the BCR-1 Nd isotopic standard and 0.512663 ± 9 (2σ) and 0.511862 ± 7 (2σ) for the La Jolla Nd isotopic standard. The workflow by Li et al. (2015) was followed to prepare samples and analyze results.

In situ zircon Hf isotopic analysis was conducted at the State Key Laboratory of Environmental Geochemistry, IGCAS, with the attachment of a 193 nm ArF excimer laser-ablation system to a Neptune Plasma multicollector ICP-MS. The laser repetition rate was set to be 10 Hz at 200 mJ, with an ablated spot size of 63 μm. For the specific workflow, please refer to Xu et al. (2004). The ¹⁷⁶Hf/¹⁷⁷Hf ratio as obtained by this study was 0.282307 ± 24 (2σ, N = 11), which is consistent with that by Woodhead et al. (2004), who implemented the solution method on zircon standard 91500, which was 0.282302 ± 8 (2σ).

Zircon oxygen isotopic measurement was conducted using a Cameca 1280 secondary ion mass spectrometer (SIMS) at IGCAS, following similar analytical procedures to those proposed by Tang et al. (2015). Given the aberration reduction that can be achieved by a Gaussian distributed beam with a primary aperture of 200 m, acceleration of the Cs+ primary ion beam was carried out at 10 kV, with an intensity of ∼2 nA, which was then rasterized over a 10 m × 10 m area. In order to compensate for sample charging during homogeneous electron density analysis over a 100 m oval area, this study introduced a normal incidence electron flood gun. Negative secondary ions were extracted with a −10 kV potential. In order to more effectively measure oxygen isotopes, multicollection mode was used on two off-axis Faraday cups. Vienna standard mean ocean water composition (VSMOW, ¹⁸O/¹⁶O = 0.0020052) was used to normalize measured ¹⁸O/¹⁶O ratios, and the Penglai standard zircon with an ¹⁸O value of 5.31‰ was referenced for correction of the instrumental mass fractionation factor (IMF; Li et al., 2010). Based on 20 cycles of measurements, the internal precision of ¹⁸O/¹⁶O ratios was 0.2‰ (2 standard error [SE]). Moreover, repeated measurements of Penglai demonstrated the external reproducibility of ¹⁸O/¹⁶O ratios to be higher than 0.40‰ (2 standard deviation [SD]). Meanwhile, the Qinghu zircon standard was also measured as an unknown for consistency verification. The results validated that the weighted mean ¹⁸O value of 48 measurements of the Qinghu zircon, 5.39‰ ± 0.37‰ (2 SD), was comparable to that measured by Li et al. (2013), which was 5.4‰ ± 0.2‰.

Zircon grains in granite samples of this study are typically pale yellow, transparent, subhedral to euhedral crystals, with fine grain sizes (80–100 μm). All zircons exhibit bright CL, with clear concentric oscillatory zones (Fig. 4). The Th concentrations of all the analyzed zircons vary from 32 to 1926 ppm, while the U concentrations range from 74 to 2725 ppm, yielding relatively high Th/U ratios (0.13–1.36). Zircon xenocrysts are common in the CL images, with the Th/U ratios ranging from 0.36 to 0.73, suggesting their magmatic origins. After elimination of discordant ages, the LA-ICP-MS U-Pb dating of the zircons yielded ages of 397.7 ± 5.7 Ma for the Tuerhong granites, 425.0 ± 8.0 Ma for the Keshikusite granites, 445.1 ± 7.4 Ma for the Aleretuobie granites, 453.5 ± 6.1 Ma for the Tuoputieke granites, and 497.8 ± 3.9 Ma for the Tasitake granites (Fig. 5). Meanwhile, U-Pb dating of the xenocrysts yielded a wider range of ages (3318–1785 Ma), indicating zircon crystallization in the source magma (GSA Data Repository Table DR1¹).

The granites from the Chinese Altai have a moderate to high content of SiO₂ (67.51–74.19 wt%), a low content of total alkalis (Na₂O + K₂O = 5.81–8.77 wt%), and a high content of Al₂O₃ (13.23–17.44 wt%). They also have low CaO (0.62–1.25 wt%), P₂O₅ (0.04–0.39 wt%), and MgO (0.21–2.23 wt%) contents, with low Mg–numbers (100 × [Mg²⁺/{Mg²⁺ + Fe²⁺}], 0.09–0.28) indicating the generation of porphyry from a highly evolved magma (Goodenough et al., 2000). Samples from the Chinese Altai fall in the areas indicating “subalkalic granite” in the (K₂O + Na₂O) versus SiO₂ diagram (Fig. 6A), “high-K calc-alkaline series” in the K₂O versus SiO₂ diagram (Fig. 6B), and weakly “peraluminous” in the A/NK versus A/CNK diagram, with A/CNK ratios varying from 1.12 to 1.31 (Fig. 6C).

Trace-element concentrations of Chinese Altai granites are shown in GSA Data Repository Table DR2. In the primitive mantle–normalized spider diagram, the granite is characterized by the enrichment of large ion lithophile elements (LILEs), such as Rb, Pb, Th, and K, and the pronounced depletion of Ba, Sr, and Eu (Fig. 7A). The contents of total rare earth elements (REEs), excluding Y, of the studied granites vary from 206.56 to 480.96 ppm. Chondrite-normalized REE patterns show light REE (LREEs) enrichment relative to the heavy REEs (Fig. 7B). The strong negative Eu anomalies might be attributable to the significant fractional crystallization of feldspar and/or late-stage fluid-melt interactions (Chen and Jahn, 2004).

Whole-rock Nd isotopic data from the five studied granites are presented in Table 1. The ε_Nd(t) values and Nd model ages were calculated based on the zircon U-Pb ages for each of the granites determined in this study. The ε_Nd(t) values of the granites are relatively uniform, ranging from −3.8 to −1.5 (Fig. 8A). The depleted mantle Nd model ages (T_DM2) are quite old, ranging from 1340 to 1180 Ma (Table 1).

The ε_Hf(t) values of the Chinese Altai granites are highly variable, ranging from −11.17 to +13.27 (Fig. 8B). In accordance with the zircon U-Pb ages (3318–1785 Ma), the ε_Hf(t) values and T_DM2 ages of the zircon xenocrysts range from −11.17 to −8.75 and from 1839 to 1678 Ma (Table 2), respectively, indicating reworked ancient crustal materials in the magma sources.

The zircons from the Chinese Altai granite samples show high δ¹⁸O values (from +9.34‰ to +13.82‰; Fig. 9; Table 3), which are equivalent to those of Precambrian (U-Pb dating) residual zircons (9.41‰–12.58‰).

According to the new age data from this study for the Tuerhong, Keshikusite, Aleretuobie, Tuoputieke, and Tasitake plutons in the southern Chinese Altai, the five granitic intrusions were emplaced from 497.8 ± 3.9 Ma to 397.7 ± 5.7 Ma. This conclusion is comparable with previous studies on ages of peraluminous and S-type granites based on zircon U-Pb dating (507–392 Ma; Tong et al., 2007; Cai et al., 2011a, 2011b; Zhang et al., 2017). In addition, synchronous felsic volcanism is also apparent in the form of peraluminous rhyolites in the Chinese Altai, with ages ranging from 412 to 406 Ma. All these ages indicate the occurrence of peraluminous granitic magmatism over a long period in the Chinese Altai, spanning from 507 to 327 Ma.

In various chemical classification diagrams of (K₂O + Na₂O) versus 10,000 Ga/Al, and Zr versus 10,000 Ga/Al, all the studied S-type granites fall in the areas indicating “I-, S-, and M-type” granites, without any falling in fields that indicate A-type granites (Figs. 10A–10B; Whalen et al., 1987; Eby, 1990). As demonstrated by the evident crustal affinity between zircon O (from +9.34‰ to +13.82‰) and whole-rock Nd isotopes (from –3.84 to –1.54), the studied granites should not be M type. In the A (Al₂O₃-Na₂O-K₂O) versus C (CaO) versus F (FeOt) ternary diagram (Fig. 10D), all granites from the two plutons plot in the “S-type” field rather than the “I-type” field. Moreover, the studied granites have high A/CNK values (from 1.12 to 1.31) and low Fe, Mg, Ca, and Na contents, consistent with the S-type granites. In addition, the studied granites also show slight decreases of Y with increasing Rb (Fig. 10C). All these lines of evidence indicate the studied granites are S-type granites. This is further supported by the presence of aluminum-saturated minerals, such as garnet and sillimanite (Fig. 3).

It is generally accepted that S-type granite could be produced by partial melting of metasedimentary protoliths (Koester et al., 2002; Jiang et al., 2011). The U-Pb ages of the zircon xenocrysts of the studied S-type granites are unlikely to be exotic because they are consistent with those of detrital zircons from the regionally widespread Habahe and Kulumuti Groups and their high-grade equivalents (Long et al., 2007, 2010; Sun et al., 2008). Moreover, previous studies reported ε_Nd(t) values of the metasediments in the Chinese Altai to range from –4.3 to –0.2. These points indicate that the studied S-type granites are most likely metasedimentary rocks from the Habahe and Kulumuti Groups.

The CaO/Na₂O ratio (0.3) was experimentally found to be sourced from plagioclase-rich, clay-poor psammitic and meta-igneous rocks (Sylvester, 1998). The low CaO/Na₂O ratios of S-type granites (0.24–0.30) imply their derivation from the partial melting of metapelite (Fig. 11A). In the experimental field diagram of melts from mafic and felsic pelites and graywackes (Fig. 11B), the studied S-type granites fall in the region with low (Na₂O + K₂O)/(MgO + FeO + TiO₂) ratios (0.71–3.56), suggesting a mafic-metapelitic source. Moreover, their CaO/(MgO + FeOt) ratios (0.08–0.45) are similar to those of metasedimentary-derived melts (Fig. 11C). Therefore, the studied S-type granites most likely were generated by the partial melting of an ancient mafic and metapelitic source. The linear compositional trends in the studied granites imply that fractional crystallization was the dominant process during differentiation (Fig. 12). The S-type granites show a clear trend on the Ba versus Eu/Eu* diagram, indicating fractional crystallization of plagioclase (Fig. 13A). The negative Sr and Eu anomalies also suggest that fractional crystallization of plagioclase played a major role. In addition, the studied S-type granites show fractionation trends for clinopyroxene and hornblende in the Rb/Sr versus Sr diagram (Fig. 13B).

Significant Nd-Hf and O-Hf decoupling was observed in all samples of this study. This is demonstrated by whole-rock Nd and zircon Hf isotopes, which define a negative trend displaced above and orthogonal to the terrestrial array in ε_Nd(t)-ε_Hf(t) space. There is a clear positive correlation between ε_Nd(t) values and MgO contents (Fig. 14A), and no correlation between ε_Hf(t) values and MgO contents (Fig. 14B), suggesting anomalous behavior of Hf isotopes in causing Nd-Hf isotopic decoupling. This is further supported by the distinct ε_Hf(t) values as compared to typical S-type granites (Appleby et al., 2010; Jiao et al., 2015).

One of the fundamental assumptions that enable the use of Hf isotopic tracers to infer source composition is that melt reaches isotopic equilibrium with protoliths during partial melting (Tang et al., 2014). To satisfy this assumption, all phases in the source must break down simultaneously and be melted proportionally, which is generally associated with zircon dissolution (Watson and Harrison, 1983; Tole, 1985; Baker et al., 2002; Harrison et al., 2007; Boehnke et al., 2013). The varying zircon dissolution rate at a single magma source may result in varying Hf isotope compositions between different batches of melts and may give rise to the decoupling of the Hf isotope system from other radiogenic isotope systems (Flowerdew et al., 2006). Zircon xenocrysts can be frequently observed in the CL images of the studied granites, suggesting that some residual zircons from the source were entrained by melt during melt loss. In addition, the high Nd/Hf ratios (16.62–58.83) and low Hf concentrations (0.6–3.1 ppm) of the studied granites also can be explained by the retention of Hf in residual zircon in the source, which lowered the concentrations of Hf relative to Nd in the produced melts. All these lines of evidence suggest that the “zircon effect” could be responsible for the Nd-Hf isotopic decoupling of the studied granites. In this scenario, undissolved zircons may retain a significant amount of ¹⁷⁷Hf at the source, elevating the ¹⁷⁷Hf/¹⁷⁶Hf ratios of the melts and decoupling them from the ¹⁴³Nd/¹⁴⁴Nd ratios. Unlike ¹⁷⁷Hf concentrations, Nd and O concentrations cannot be altered by retention in residual zircons; melts therefore retain the Nd and O isotope ratios of the source material.

Several factors have been identified to control the zircon dissolution rate, including temperature, pressure, zircon crystal size, and zirconium concentrations at the source (Baker et al., 2002; Boehnke et al., 2013; Harrison et al., 2007; Tang et al., 2014; Tole, 1985; Watson and Harrison, 1983). The varying zircon dissolution rates at a single magma source may result in varying Hf isotopic compositions among different batches of melts (Tang et al., 2014). The crystallization temperature of the Late Silurian to Early Devonian peraluminous granitic magmas was constrained by Cai et al. (2011a, 2011b), who made an estimate of 764–857 °C. In addition, petrological studies on Ordovician to Silurian high-grade rocks in the Chinese Altai show high-temperature and low-pressure metamorphism with temperature estimated by the Ti-in-zircon thermometer (Jiang et al., 2010). All these suggest that the Chinese Altai was under a high-temperature regime. Zhang et al. (2017) proposed the capacity of zircon to survive in temperatures lower than 780 °C, indicating the less important role of temperature in determining the appearance of residual zircons and the nature of isotope decoupling. Previous studies suggested comparable compositions of the peraluminous granites in the Chinese Altai with those of melts produced by the melting of various kinds of metasediments at low pressures (P ≤ 5 kbar), suggesting the derivation of their precursor magma from shallow crustal levels. It is generally accepted that large mineral grains are dissolved more slowly than small grains (Liu et al., 2008; Yakymchuk, 2017). Such a “size effect” may be important during fast melting at the source. Moreover, metasediments from the Chinese Altai usually host fine zircon crystals with the sizes from 80 to 150 μm, similar to regular zircons in metasediments (Tang et al., 2014).

When the initial zirconium concentration at the source is sufficiently high (e.g., >100 ppm), the interbatch Hf becomes more concentrated and less radiogenic than the bulk protolith. This may lead to a melt that is extracted from a single source that evolves from a mantle-like source at the early stage to a crustal-like source after extensive melting. Zirconium is abundant in the continental crust, with average concentrations ranging from 68 ppm in the lower crust to 193 ppm in the upper crust (Rudnick and Gao, 2003). However, Y.J. Wang et al. (2014) reported that the metasedimentary rocks in the central Chinese Altai have extremely high zirconium contents (126–377 ppm). The metasedimentary rocks from the Kangbutiebao and Aletai Groups are also characterized by extremely high zirconium contents (115–424 ppm; Long et al., 2008a, 2008b). This means that the zircon effect, as a result of the slow disequilibrium melting of zircons, may be common during crustal melting in the Chinese Altai. Under any of the circumstances considered here, zircons cannot be completely dissolved before melt loss begins (Rosenberg and Handy, 2005). If the initial zirconium concentration at the source is above 200 ppm, even the smallest zircons can survive 40% melting of the source (Tang et al., 2014). Therefore, this study indicates that isotope decoupling and the appearance of residual zircons might be attributed to the extremely high zirconium content of the source along with disequilibrium melting processes that retained large amounts of Hf in residual zircons in the source.

Sylvester (1998) suggested that peraluminous or S-type granites in “high-temperature” collisional orogens have larger volumes and lower Al₂O₃/TiO₂ ratios than those in “high-pressure” collisional orogens. Figure 15 shows the fields of granites from high-temperature and high-pressure collisional orogens defined by Sylvester (1998) and Peng et al. (2012). All the studied S-type granites fall into the high-temperature field rather than the high-pressure field, indicating that granitic magmatism should not be dominated by the syncollisional setting.

A postcollisional setting might be a viable model for the studied S-type granites. However, recent zircon U-Pb dating results for igneous rocks demonstrated continuously widespread magmatism over the entire Altai orogenic belt from the early Paleozoic to the early Mesozoic (Zhu et al., 2006; Briggs et al., 2007; Long et al., 2007; Yuan et al., 2007; Sun et al., 2008; Cai et al., 2011a, 2011b, 2012; Lv et al., 2012). Remarkably, the emplacement of early Paleozoic subduction-related I-type granites was nearly continuous and reached a peak in the Devonian, and postcollisional A-type granites mainly appeared after ca. 310 Ma (Long et al., 2007; Yuan et al., 2007; Sun et al., 2008; Cai et al., 2011a, 2011b, 2012). Thus, it is inferred that a tectonic transition occurred at ca. 310 Ma in the Altai orogenic belt (from subduction to postcollision). During the late Paleozoic, continuous subduction magmatism was common, produced by the northward subduction of Paleo-Asian oceanic lithosphere in the Altai orogenic belt (Long et al., 2007; Sun et al., 2008; Cai et al., 2011a, 2011b, 2012; Wang et al., 2011). The subduction-related granites emplaced in the Altai orogenic belt have been dated, with ages of ca. 507 to ca. 313 Ma (Cai et al., 2011b; Wang et al., 2011; Lv et al., 2012; Zhang et al., 2017), indicating the probable life span of this subduction from the middle Cambrian to the late Carboniferous. In particular, postcollisional magmatism mainly began after ca. 310 Ma, with representative examples of the Suoerbulake (281 ± 4 Ma), Bukesala (277 ± 2 Ma), and Lamazhao (276 ± 9 Ma) granites (Long et al., 2007; Yuan et al., 2007; Sun et al., 2008; Cai et al., 2011a, 2011b, 2012; Wang et al., 2006). As argued by studies on the ultrahigh-temperature (UHT) granulite-facies rocks in the Altai orogenic belt, there should be a counterclockwise pressure-temperature path, suggesting a depression and cooling event at 280–277 Ma associated with underplating and heating of mantle-derived magma in an extensional tectonic setting (Li et al., 2014; Tong et al., 2014). Therefore, subduction-related magmatism in the Altai orogenic belt possibly shifted to postcollisional magmatism during the late Carboniferous and Early Permian (Long et al., 2007; Yuan et al., 2007; Sun et al., 2008; Cai et al., 2011a, 2011b, 2012; Zhu et al., 2006; Lv et al., 2012). All these lines of evidence indicate that the studied S-type granites (from 497 Ma to 397 Ma) were not formed in a postcollisional setting.

In general, the genesis of arc magmas is predominantly controlled by the introduction of aqueous fluids (Kushiro, 1969) rather than by heat or decompression melting, as normal subduction is a relatively “cold” and “high-pressure” tectonic process. Therefore, the normal subduction setting might not be responsible for the studied S-type granites. This study indicates that a slab window model is viable for the Chinese Altai, where the slab window refers to the gap in the subducted plates that allowed the upwelling of asthenospheric mantle, which has been documented during spreading ridge–trench interactions (Thorkelson, 1996; Sisson et al., 2003; Cole et al., 2006; Tang et al., 2010; Yin et al., 2010). During the late Cambrian to Early Devonian, the Irtysh-Zaysan ocean ridge subducted beneath the Chinese Altai. As the slab window opened, the upwelling asthenosphere heated and then melted the overlying lower–middle crusts that consisted of zircon-rich metasediments (Li et al., 2014; Tong et al., 2014), leaving a zircon residue to generate S-type granites. The doleritic dikes in the Chinese Altai were derived from the partial melting of a refractory mantle with previous magma extraction. In addition, the gabbroic dikes in the Chinese Altai also suggest upwelling of hot asthenosphere in a subduction setting, which indicates that the Chinese Altai was likely under a subduction-related high-temperature regime in the Devonian. This is also supported by a prominent ca. 390 Ma high-temperature metamorphic event defined by the zircon overgrowth rims of high-grade gneisses (Jiang et al., 2010; Cai et al., 2011a, 2011b). Therefore, the formation of this extensive high-temperature magmatic association implies the occurrence of ridge subduction associated with slab window opening between 497 and 397 Ma in the Chinese Altai.

Based on the petrological, geochemical, and geochronological data present in this study, we draw the following general conclusions.

• (1) The five studied granites in the Chinese Altai are classified as S-type granites and were emplaced between 497 and 397 Ma. Their parental magmas might be derived from the disequilibrium partial melting of metasediments in the Chinese Altai.

• (2) Decoupling of the O-Nd-Hf isotopes resulted from disequilibrium melting processes, and zircons from metasedimentary sources may survive during crustal melting, as indicated by the relatively high zirconium concentration (>100 ppm) of the source.

• (3) Ridge subduction might have played an important role in the Chinese Altai during the late Cambrian to Early Devonian.

We would like to express our great gratitude to Matthew Mayne and an anonymous referee for their critical insight and constructive comments, which have greatly improved this manuscript. Many thanks are given to Xin Zhang of the Institute of Geochemistry, Chinese Academy of Sciences (IGCAS), for his help with geochemical analyses. This work was financially supported by grants from the National Natural Science Foundation of China (no. 41502209), the National Key R&D Program of China (no. 2015CB250901), and the National Science and Technology Major Project (nos. 2016ZX05034–001, 2017ZX05035–002, 2017ZX05039–004).