Zircon crushing, single mineral selection, target development, and U-Pb and cathodoluminescence (CL) image acquisition were performed at Beijing Yanduzhongshi Geological Analysis Laboratories, Ltd. Zircon U-Pb dating was performed using a laser ablation system with a UP-213 deep ultraviolet laser (213 nm), and a MicroLas optical system. Inductively coupled plasma mass spectrometry (ICP-MS) was performed using a Jena M90, with the international zircon standard 91500. The ordinary Pb calibration method was used to calibrate the measurement results. ICP-MS DataCal software was used for data processing (Liu et al., 2010). Zircon U-Pb concordia diagrams and the mean squared weighted deviation (MSWD) were obtained using Isoplot. The age value error of 1σ was used in determining the ²⁰⁶Pb/²³⁸U age, and a 95% confidence interval was utilized (Ludwig, 2003).

Whole-rock geochemical analysis was performed at the Beijing Yanduzhongshi Geological Analysis Laboratories, Ltd. Fourteen fresh samples (none of the materials showed were altered) were ultrasonically cleaned with deionized water, oven-dried at 60 °C for 24 h, and pulverized to <75 μm using an agate mortar prior to geochemical analysis. For the whole rock major element testing, the sample powder was weighed and mixed with lithium tetraborate (Li₂B₄O₇) as a flux in a ratio of 1:8. Ammonium nitrate (NH₄NO₃) was added as an oxidant, and lithium bromide (LiBr) was added as a flux. The sample was heated to 1,050 °C for 15 min using a fusion furnace to form a uniform glass sheet in a platinum crucible. Testing was conducted using wavelength dispersive X-ray fluorescence (XRF) (Zetium, Malvern PANalytical, Malvern, United Kingdom), with a relative standard deviation (RSD) of less than 2%. The sample powder was weighed, placed in a polytetrafluoroethylene dissolution tank, and HF + HNO₃ was added prior to whole-rock trace element analysis. The material was placed in a high-pressure digestion tank in a drying oven at 190 °C for 72 h, after which it was removed for acid flushing. The solution was diluted to a constant volume for machine testing. Testing was completed using the Agilent 7700 ICP-MS, and a testing accuracy of better than 10% was achieved.

Nd isotope analysis was performed using six sets of whole-rock samples at Beijing Yanduzhongshi Geological Analysis Laboratories Ltd. First, 0.25 g of the sample was accurately weighed and placed into a Teflon stew tank. HNO₃ and HF were added, and the sample was heated for 48 h digestion under closed conditions at 190 °C. The temperature was reduced to 160 °C under closed conditions to remove any HF. The HNO₃ solution was added, and the mixture was dissolved at a temperature of 150 °C over 6 h under sealed conditions until a constant volume of 25 g was achieved. An appropriate amount of the Nd solution was centrifuged. The obtained suspension supernatant was evaporated to dryness, and the solution was adjusted to a suitable pH. A special LN resin was used to separate the pure Nd and obtain a sample solution. Finally, multi-collector ICP-MS (MC-ICP-MS) (Neptune Plus, Thermo Fisher Scientific) was performed to determine the ¹⁴³Nd/¹⁴⁴Nd value for the sample solution. According to the exponential law for ¹⁴³Nd/¹⁴⁴Nd (0.7218), online mass fractionation correction was performed on the measured ¹⁴³Nd/¹⁴⁴Nd values with an error of 2σ (uncertainty in the mass spectrometry measurement).

Laser ablation-ICP-MS (LA-ICP-MS) zircon U-Pb dating was performed on three granite samples and one enclave sample from the Sabei, Huangyangshan, and Sujiquan plutons in the eastern KGB. Rectangular zircons with a complete crystal structure were selected under transmitted and reflected light, and CL images were obtained to perform zircon U-Pb dating (Figure 3). The data used for zircon U-Pb dating are presented in Table 1.

The results for the whole-rock major, trace, and rare earth elements (REEs) in the Sabei, Huangyangshan, and Sujiquan plutons and the MMEs are listed in Table 2. The three plutons have high contents of silicon (76.30–77.41 wt%), potassium (4.04–4.67 wt%), alkali (Na₂O + K₂O = 8.18–8.80 wt%), and Na₂O (3.98–4.40 wt%); the total alkali mass fraction exceeds 8%, indicating the rock is subalkaline. The Mg# value ranges from 8.6 to 35.3. The aluminum saturation index (A/CNK) is 0.94–0.99, and A/NK is 0.97–1.06.

The MMEs have lower contents of SiO₂ (55.37–67.33 wt%) and total alkali (Na₂O + K₂O) (5.89–7.00 wt%) than the host rock, indicating a diorite composition. The TFeO (3.47–7.57 wt%), MgO (1.27–5.26 wt%), and CaO (3.05–6.97 wt%) contents are significantly higher than those of the host granite, and the Al₂O₃ content is higher (14.93–16.26 wt%), indicating aluminum supersaturation (A/CNK = 1.26–1.49). However, the enclaves are relatively enriched in P₂O₅ (0.13–0.17 wt%) and Mg# (44.2–55.3). The values are slightly higher than those of the host rocks, suggesting a mantle melt composition.

All host rock samples from Sabei, Huangyangshan, and Sujiquan fall into the granite field of the total alkali-silica (TAS) diagram (Figure 4a), with MMEs falling into the granodiorite and monzodiorite areas. The host rock samples show weak peralkaline-metaluminous characteristics in the A/CNK-A/NK diagram (Figure 4b), whereas the MME samples are peraluminous. Most of the host rock samples and MMEs fall into the high-K calc-alkaline series of the SiO₂-K₂O diagram (Figure 4c).

The Harker diagram (Figures 5a–h) indicates that the host granite samples plot in a concentrated zone, indicating a linear relationship between host granites and MMEs. Al₂O₃, MgO, CaO, MnO, FeO, TiO₂, and TFe₂O₃ are negatively correlated with SiO₂; K₂O is positively correlated with SiO₂; TFeO is positively correlated with MgO; and Na₂O shows no linear relationship with SiO₂. Excellent linear correlations are observed for the dominant oxide ratios: Na₂O/CaO vs. Al₂O₃/CaO, Na₂O/CaO vs. SiO₂/CaO, SiO₂/MgO vs. Al₂O₃/MgO, and CaO/MnO vs. MgO/MnO (Figure 6).

The results of the rare earth and trace element analyses for the Sabei, Huangyangshan, and Sujiquan plutons and the Sabei MMEs in the eastern KGB are listed in Table 2. The granite samples from the Sabei, Huangyangshan, and Sujiquan host rocks in the Karamaili East section exhibit elevated ΣREE contents ranging from 121.78 to 261.32 ppm, with Σ light REEs (LREEs) ranging from 81.23 to 206.60 ppm and Σ heavy REEs (HREEs) from 32.94 to 64.90 ppm. The ΣLREE/ΣHREE ratio is 1.28–6.27.

The chondrite-normalized REE patterns for the granite samples from the Karamaili belt display a deep V-shaped distribution and a negative Eu anomaly (Eu/Eu* = 0.01–0.34) (Figure 7a). The host granites and MMEs exhibit the same signatures in the primitive mantle-normalized trace element spider diagram (Figure 7b), characterized by systematic incompatible element enrichment, particularly in high field strength elements (HFSEs). Negative anomalies are observed at Ba, Sr, P, and Ti, while positive anomalies occur at Rb, Th, K, and the Zr-Hf pair. All samples display systematic Nb-Ta depletion relative to adjacent elements (Th, K, La-Ce). A spider diagram of the trace elements normalized to the primitive mantle indicates substantial Ba, Sr, P, Eu, and Ti depletion and relative depletion in Tb and Tm. HFSEs (Ta, Th, U) and LILEs (Rb, K, Pb) are relatively enriched (Figure 7b). The granite zircon chondrite-normalized REE patterns reveal considerable LREE depletion and HREE enrichment, with La to Lu values showing a variation over five to six orders of magnitude, and a strong negative Eu anomaly (Figure 7c). In contrast, no LREE and HREE fractionation is observed in the zircons from the MME, with only a slight negative Eu anomaly (Figure 7d), indicative of plagioclase fractionation or retention during magma evolution, as Eu replaces Ca in plagioclase.

Four samples were collected from the KGB for whole-rock Nd isotope analysis (Table 3), with four samples from the Sabei pluton (SB-1, SB-3, SBMME-1, SBMME-2), one from the Huangyangshan pluton (HYS-1), and one from the Sujiquan pluton (SJ-1).

Table 3 shows that the Sm/Nd ratios for the Sabei, Huangyangshan, and Sujiquan plutons range from 0.135 to 0.156, with similar ratios for different rock types, implying a uniform source material and a relatively closed Sm-Nd system during petrogenesis. These values are lower than the chondritic ¹⁴⁷Sm/¹⁴⁴Nd ratio of 0.1967 (Hamilton et al., 1983). The ¹⁴³Nd/¹⁴⁴Nd ratios of 0.512787–0.512808 and εNd(t) values for the Sabei, Huangyangshan, and Sujiquan rocks fall in the range +4.0 to +4.5, with T_2DM values ranging from 608 to 651.

In contrast, the SBMMEs exhibit higher Sm/Nd ratios (0.196–0.230), ¹⁴³Nd/¹⁴⁴Nd ratios from 0.512921 to 0.512965, and εNd(t) values from +5.5 to +5.7. T_2DM values of 476–488 are obtained for the Sabei MMEs.

The Carboniferous was the critical stage of tectonic transformation in Eastern Junggar. Therefore, constraining the evolutionary period of siliceous magmatic activity in the Karamaili area is crucial (Chen et al., 2010; Zhang et al., 2010; Li et al., 2015). Previous studies have suggested that the emplacement time in the KGB was 328–283 Ma (Han et al., 2006; Lin et al., 2007; Su et al., 2008; Tang et al., 2009; Yang et al., 2009; Gan et al., 2010; Han et al., 2012). In this study, the LA-ICP-MS zircon U-Pb ages for the Sabei monzogranite, Huangyangshan biotite syenogranite, and Sujiquan biotite granite are 315.9 ± 2.4 Ma, 316.1 ± 2.2 Ma, and 319.4 ± 2.0 Ma, respectively. These ages indicate that the emplacement time of the plutons in the eastern section of the KGB can be constrained to 320–315 Ma. The LA-ICP-MS zircon U-Pb age for the MMEs of the Sabei pluton is 313.9 ± 3.5 Ma. Field investigations reveal no discernible intrusive contacts between host granites and MMEs. Consequently, the geochronological data indicate that both lithological units have similar emplacement ages within the analytical uncertainty. It has been suggested that MMEs and granitic plutons are products of the synchronous evolution of the magma.

The granites in the Sabei, Huangyangshan and Sujiquan plutons are characterized by high Si, K, and alkali contents. The CaO, MgO, Ta, Ba and TiO₂ contents are low, with 10,000 × Ga/Al = 3.06–4.45. These values are significantly higher than the average values for I-type and S-type granites (2.10 and 2.28) and the relatively enriched HFSEs (Ta, Th, U) and LILEs (Rb, K, Pb), indicating geochemical characteristics similar to those of A-type granite. The strong negative Eu anomalies (δEu = 0.01–0.34) and low Sr contents (3.2–37.5 ppm, mean = 9.2 ppm) differentiate these granites from highly fractionated I-type granite, They are important indicators of A-type granite (Xu et al., 2015). The whole-rock zircon saturation temperatures of the granites range from 735 °C to 895 °C, with most samples showing higher temperatures than typical zircon saturation temperatures of highly-fractionated I-type granite (average 764 °C) (Alberto and Patino, 1997; King et al., 1997). This finding indicates that the magma partially melted under water-deficient conditions at temperatures of >830 °C, which is a typical A-type granite formation condition.

The KGB samples fall in the A-type granite area of the TFeO/MgO vs. 10000 Ga/Al discrimination diagram (Figure 8a), indicating they differ from the I- and S-types. The geochemical characteristics are similar for A-type and highly fractionated I-type granites (King et al., 1997). Most of the Sabei, Huangyangshan, and Sujiquan pluton samples fall within the A-type granite area in the (K₂O + Na₂O)/CaO vs. Zr + Nb + Ce + Y diagram (Figure 8b). Thus, they differ from the samples derived from the Kamusite and Laoyaquan plutons in the western part of the KGB, which fall within the highly differentiated I-type granite. The samples from the KGB fall within the A2-type granite area in the Nb, Y, Ce, and Yb classification diagrams (Figures 8c,d), indicating orogenic A-type granite genesis. It differs from the non-orogenic A₁-type granite, and the magma source has an affinity with island-arc basalts (IABs). These results, in conjunction with the geological features, including the location of the KGB at the plate margin and the delayed timing of ocean basin closure and subduction, suggest that the Sabei, Huangyangshan, and Sujiquan granites are typical quasi-aluminous, alkaline-high-potassium calc-alkaline A₂-type granites.

MMEs with different numbers, sizes, and shapes are widely observed in the Sabei and Huangyangshan plutons. The genesis of these enclaves is generally determined using four approaches: refractory residues of granite protolith melting (Chappell et al., 1987; Chappell and White, 1991; White et al., 1999; Chappell and Wyborn, 2012), xenoliths from the wall rock (Maas et al., 1997), the early crystallized cumulates from cognate magma (Noyes et al., 1983; Dahlquist, 2002; Donaire et al., 2005), and the mixing/mingling of magma (Baxter and Feely, 2002; Barbarin, 2005; Słaby and Martin, 2008).

The contacts between MMEs and host granites typically exhibit sharp macroscopic boundaries, occasionally displaying diffuse transitions, with ubiquitous dark fine-grained chilled margins. These textural features preclude an origin as xenoliths from the wall-rock or refractory residues of granite protolith melting. The pluton contains abundant dioritic MMEs characterized by magmatic fine-grained hypidiomorphic granular textures and massive structures, indicative of rapid crystallization under disequilibrium conditions. Petrogenetic constraints suggest that supercooled mafic magmatic blobs intruded into cooler felsic host magma during magma mixing. This process induced metasomatic replacement of early-crystallized plagioclase, forming K-feldspar reaction rims. The ubiquitous acicular to columnar apatite inclusions in the plagioclase, displaying distinct crystallographic orientations from host rock apatite, represent quench-crystallized products (Barbarin, 1988; Barbarin and Didier, 1992). These textural indicators, particularly the diagnostic acicular apatite morphologies, suggest significant thermal contrast (>200 °C) between mafic enclave-forming magma and host felsic magma, with rapid heat dissipation during crystallization. However, the MMEs are randomly distributed in the host granitoids, and petrological analysis reveals that they have magmatic textures and numerous disequilibrium textures, which are contradictory to autoliths, suggesting they likely formed through magma mixing. Transitional zones at the interfaces reveal plagioclase phenocrysts containing euhedral oscillatory-zoned cores overgrown by multiple generations of anhedral rims, recording episodic growth under fluctuating physicochemical conditions. Sector-zoned textures in the plagioclase indicate dynamic crystallization environments. The presence of relatively coarse-grained amphibole clusters within the MMEs indicates precursory crystallization of mafic phases prior to magma interaction (Zorpi et al., 1991), whereas felsic mineral assemblages formed during later-stage co-crystallization of residual enclave magma and host granodioritic melt under comparable cooling rates (Vernon, 1984). Chemical hybridization through interdiffusion between mafic and granodioritic magma to achieve a local equilibrium resulted in mutual geochemical modification. This petrogenetic mechanism accounts for the observed geochemical similarity in the major/trace element compositions and mineral assemblage between host rocks and MMEs.

In the Harker diagrams of major oxides versus SiO₂ (Figure 5), the samples from this pluton exhibit well-defined negative linear correlations between Al₂O₃, CaO, FeO, K₂O, MgO, MnO, TiO₂, Na₂O, and TFeO with increasing SiO₂ content. These systematic trends strongly suggest magma mixing between host granites and MMEs. Fractional crystallization typically generates trends due to sequential mineral phase removal (e.g., plagioclase and amphibole), whereas mixing systems preserve linear arrays reflecting binary endmember blending. The TFeO-MgO correlation diagram (Figure 5i) supports this interpretation. All compositional data plot along established magma mixing trend lines, indicative of chemical hybridization between mafic and felsic melts. Figure 6 shows well-defined inverse linear correlations in key oxide ratios between MMEs and host granites. These trends reflect potential magma mixing evolutionary trends between the host granite and its enclaves. The primitive mantle-normalized spider diagrams (Figure 7b) reveal broadly parallel patterns between enclaves and host rocks, consistent with assimilation-fractional crystallization (AFC) during coeval magma evolution (Lei et al., 2019). The AFC accounts for the observed geochemical similarity between the dioritic enclaves and granitic hosts, particularly the depletion in Nb, Ta, P, and Ti. These anomalies primarily reflect fractionation of accessory phases (e.g., apatite and Ti-bearing oxides), coupled with crustal contamination. The chondrite-normalized REE patterns display analogous LREE-enriched profiles, yet exhibit HREE crossover features between host rocks and enclaves (Figure 7a). This finding suggests that the host granites and enclaves represent distinct but interacting magma batches.

The MMEs of the Sabei plutons exhibit the following features. (1) The zircon geochronology of the MMEs is 313.9 ± 3.5 Ma, whereas the age of the host granite is 315.9 ± 2.4 Ma, showing consistency in the formation chronology. This equivalence precludes the possibilities of refractory residues of granite protolith melting or xenoliths from the wall rock because both source materials would predate the host magma chronology (Brown and Fyfe, 1970; Kempton et al., 1987); (2) The host granite and MMEs exhibit different lithologies. Distinct boundaries exist between the round, oval, or droplet-shaped enclaves and the host rock, and plagioclase with a sieve structure is common (Figure 2e). Therefore, these observations preclude the potential origins of the detachment of the early-stage crystallized homologous magma (O’Hara, 1977). (3) Acicular apatite generally originates from rapidly cooling magma (Wyllie et al., 1962), i.e., its presence indicates rapid cooling. (4) The petrography of the MMEs indicates they have a magmatic origin and do not originate from the surrounding rock; i.e., they are not xenoliths. The similarity in the zircon U-Pb ages indicates that the MMEs and host rocks coexisted and were products of contemporaneous mafic and acidic magmas. (5) The TFeO vs. MgO diagram shows that the MMEs did not evolve along the crystallographic differentiation trend of mafic magma, but were distributed near the magma mixing trend line. This finding indicates an evolution toward the host granite composition (Figure 5i), suggesting that the MMEs are represent hybridized melt blobs left over from the siliceous magma due to magma mixing during the uplift of the mafic magma. The age of the Sabei MMEs is consistent with that of the host granite and the characteristics of the trace elements and REE are similar, indicating a dioritic magmatic enclave. The magma is derived from mantle material and is a product of magma mixing.

The Karamaili ophiolite mélange is considered a fragment of the eastern Karamaili Ocean in the Junggar Ocean (Xu et al., 2015). The expansion of the ocean basin likely occurred from the late Silurian to the late Devonian (Jian et al., 2003; Shu and Wang, 2003; Xu et al., 2011). Subsequently, the Junggar Ocean was subducted under the Siberian Plate at the edge of the continental block, generating accretionary wedges and continental margin volcanic arcs, such as the abundant late Devonian-Early Carboniferous island arcs and volcanic rocks observed in the area (Li, 1995; Zhang and Guo, 2010; Cai et al., 2012).

The Eastern Junggar exhibits the characteristics of a multi-island arc structure, with many island arc belts and fore-arc, and back-arc basins that formed during compression in the Early Carboniferous. Subsequently, the ocean basin closed, the island arcs merged, and the land mass was uplifted by extrusion in the middle Early Carboniferous. During the syn-collision orogenic phase, the subduction orogeny weakened, leading to the slab break-off of the subducted oceanic plate in the late Carboniferous. Subsequently, compression-extension tectonic transformation occurred, gradually transitioning into the post-collision extension stage and causing the underplating of large-scale mantle-derived materials. The basic magma intruded upward and partially melted the upper crustal material, resulting in the formation of large-scale siliceous magma, juvenile crustal growth, and the Late Carboniferous KGB (Han et al., 2006; Su et al., 2006; Gan et al., 2010; Li et al., 2015; Zhang et al., 2020).

As shown in the Y + Nb-Rb diagram, most of the KGB samples fall in the intraplate granite area, indicating a formation environment comprising ridge granite, volcanic arc granite, and syn-collision granite (Figure 9a). The R₁-R₂ and Al₂O₃-SiO₂ diagrams of the post-orogenic granites indicate that almost all KGB samples are post-orogenic, A-type granites (Figure 9b). This result shows that the section formed in an extensional environment and a post-collision setting; thus, the granites differed significantly from those granites related to rift and continental uplift (Figure 9c). In addition, the A-type granite in the KGB also falls into the post-orogenic (PA) field (Figure 9d); thus, it differs from anorogenic granites. In summary, the results suggest that the A₂-type granites in the Sabei, Huangyangshan, and Sujiquan plutons of the eastern KGB formed in an extensional environment and a post-collision geotectonic setting at the end of the orogenic event at around 320–315 Ma.

In accordance with the different granites observed in the easternmost and westernmost KGB, the latest report suggests that the area was not an extensional environment in the late Carboniferous but was in transition from compression to extension. The western KGB experienced compression while the eastern KGB underwent extension during the Late Carboniferous (Wang et al., 2025), These discrete structures imply a high probability of the proposed conditions (Figures 9a,c,d).

Magma reservoirs serve as critical loci for analyzing upper crustal magmatic processes. Crystal mush—formed through multiple magma intrusions (Bachmann and Bergantz, 2004; Coleman et al., 2004; Glazner et al., 2004; Lipman, 2007; Miller et al., 2009; Lipman and Bachmann, 2015; Blundy and Annen, 2016; Ma et al., 2020; Zhang et al., 2022)—exhibits high crystallinity, elevated viscosity, and limited mobility, inhibiting convection, fractionation, and mixing (Bachmann and Bergantz, 2004; Bachmann and Huber, 2016; Ma et al., 2020). However, large-scale upwelling of mantle-derived mafic magma thermally rejuvenates viscous crystal-rich mush, elevating temperatures, reducing viscosity, increasing melt fraction, and decreasing crystallinity. Injection of this high-temperature mafic magma into cooler silicic host magma facilitates mixing, driving rapid compositional hybridization. This process can transform mafic melts into intermediate compositions (e.g., basaltic-andesite to diorite) through combined assimilation and fractional crystallization (Cooper and Kent, 2014; Miller, 2016; Rubin et al., 2017; Jackson et al., 2018; Wilson et al., 2021). The relatively low solidus temperature of dioritic magma promotes early crystallization, forming cumulate textures. Chemical exchange between MMEs and host granite diminishes as magmas cool (Wang, 2000). During flow emplacement, large enclave swarms may fragment into smaller MMEs (Figure 2a). Buoyancy-driven ascent of crystal-rich mush through overlying strata may culminate in volcanic eruptions (Annen, 2009; Spera et al., 2016; Ma et al., 2020) (Figures 10, 11).

As magma ascends from the magma reservoir toward the Earth’s surface, mineral crystallization and fractionation can significantly alter its composition (Elburg, 1996). The MMEs have zircon U-Pb ages consistent with those of the host granites; however, compositional differences in major and trace elements are observed. As shown in Figure 5i, the linear relationship between MMEs and the host granite is consistent with the magma evolutionary trends (Langmuir et al., 1978), suggesting that the MMEs and host granite are products of magma mixing rather than crystallization and fractionation. The mafic enclaves in the Sabei Pluton are dioritic rather than gabbroic. The Mg^# value is higher than that of the host granite; however, the highest value 55.3 is smaller than the average value obtained for the primitive mantle (Mg^#>70) (Bloomer and Hawkins, 1987). The Mg^# value is commonly used to indicate the degree of partial mantle melting. The mantle peridotite with a high Mg^# value (>90) may have undergone a higher degree of partial melting, whereas the Mg^# value of the Sabei MMEs (maximum value of 55.3) suggests that the mantle-derived basic magma did not undergo extensive partial melting but was mixed with granitic magma during its ascent, leading to a simultaneous decrease in the Mg^# value and SiO₂ content (Figures 8b, 12). In addition, the MMEs in Sabei exhibit a mantle-crust evolutionary trend, which is characterized by a decreasing Sr content and an increasing Rb content which is not observed in the host granite. Unlike the Sabei granite, the MMEs have high Mg and Fe contents but low Si and Na contents. Moreover, the zircon U-Pb age is consistent with that of the host granite, recording syn-ascension mixing of mantle-derived mafic and crustal felsic magmas prior to Late Carboniferous emplacement.

The CAOB experienced Phanerozoic juvenile crustal growth, with estimates exceeding 50% (Şengör et al., 1993). The zircon εHf (t) value and whole rock Nd isotope εNd(t) of the granites in the East Junggar are significantly higher than zero, indicating substantial crustal growth in the region during the Late Paleozoic (Su et al., 2007; Tang et al., 2007; Yang et al., 2008; Wang et al., 2023). The zircon εHf (t) values of the granites in the KGB are high, ranging from +10.5 to +14.7, which is close to the value for the depleted mantle (Han et al., 1997; Chen and Arakawa, 2005; Chen et al., 2010). This result suggests that the source is juvenile crust (Wu et al., 2007; Wang and Hou, 2018). In contrast, the granites of Western Junggar are characterized by lower εHf(t) values (Tang et al., 2008; Geng et al., 2009). The whole-rock εNd (t) values are +4.0 to +4.5 (as are the values obtained for the Sabei and Sujiquan plutons); all values are positive (Figures 13a,b), indicating they may have been formed by the remelting of new crustal materials or the differentiation of mantle-derived magma (Whalen et al., 1987; Han et al., 1997; Wu et al., 2003; Bonin, 2007). The Sm/Nd-εNd(t) diagram (Figure 13a) indicates that the granites in the KGB are strongly influenced by continental contamination. The Sabei MMEs show a linear relationship with the host rock, reflecting the evolution of basic and siliceous magma. Additionally, the whole-rock Nd model age (T_DM2) of the KGB is in the range 600–650 Ma, suggesting its source area was the Neoproterozoic juvenile lower crust (Wang et al., 2009), which is consistent with the zircon Hf model age (T_DM2) (Zhang et al., 2021). The Karamaili granite formed by remelting of earlier-accreted juvenile crustal material, which originated from depleted mantle sources during prior subduction (Jahn et al., 2000). Its emplacement reflects crustal reworking rather than primary growth. Crucially, “vertical crustal growth” refers to magmatic underplating at the crust-mantle boundary (Rudnick and Taylor, 1987), where new mantle-derived melts crystallize beneath pre-existing crust, increasing net crustal thickness and mass. During ascent, these juvenile magmas assimilated upper-crustal components (e.g., metasedimentary rocks), evidenced by inherited Proterozoic zircon cores and elevated δ¹⁸O values. Concurrent magma mixing with mantle-derived melts generated MMEs, supported by disequilibrium textures and Hf-Nd isotope decoupling (Kemp et al., 2009). Spatial variations in the KGB manifest this process: shallow emplacement in the eastern Sabei and Sujiquan plutons preserved abundant MMEs due to limited residence time, whereas prolonged fractionation in western Kamusite pluton formed highly differentiated granites.

In summary, the oceanic crust broke off and caused unbalanced delamination during the subduction at the end of the Late Carboniferous (Zhang et al., 2021), causing the partial melting of large amounts of mantle-derived materials. This event led to underplating at the bottom of the lower crust, triggering upwelling of the basic magma and resulting in the recharge and strong activation of the siliceous magmatic reservoir in the upper crust. Small amounts of basic magma caused the strong activation of the Kamusite and Laoyaquan plutons in the western part of the KGB. Thus, the high-silica melt was extracted from reactivated crystal mush from the high-silicon melt, forming highly differentiated granites with minimal magma mixing signatures. A large amount of basic magma was exchanged with the magma of the Sabei, Huangyangshan, and Sujiquan plutons in the eastern section. Many MMEs were formed by the mixing of basic and siliceous magma, which is manifesting as local-scale magma mixing. A slightly later gabbro pluton (311–307 Ma) of non-ophiolitic origin in the KGB confirms the presence of basic magma in the Late Carboniferous (Wang et al., 2025). Additionally, a large number of diorite and gabbro veins formed in the Sujiquan and Huangyangshan granites in the eastern KGB, differing from the felsic veins that developed only in the Kamusite and Laoyaquan granites of the western KGB. Thus, the supply of basic magma was the dominant factor determining the final occurrence of siliceous magma, which implies both local-scale (mixing) and regional-scale (underplating) impacts of mafic magmatism. In other words, the geochemical linear evolution of the MMEs in the Sabei pluton shows that the injection of basic magma into siliceous magma is a continuous and multi-stage magmatic mixing (Figure 5a–i).