Initially, the rock samples underwent crushing and washing, and then electromagnetic and heavy liquid separation were applied to concentrate zircon grains. Using a binocular microscope, a selection of well-shaped, larger zircon grains was made and then set in epoxy resin, followed by polishing to prepare analytical mounts. For these polished zircon grains, the most suitable candidates—those exhibiting clear internal zoning without noticeable cracks or bubbles—were selected based on cathodoluminescence (CL), transmission, and reflection imaging. These zircons were then analyzed using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) on an Agilent 7500a ICP-MS system, using a 30 μm laser spot diameter with He as the carrier gas. The reference standards NIST 91500 and NIST 610 were employed for instrumental calibration. Finally, data processing and plotting were conducted using the Glitter and Isoplot software packages.

Major element concentrations were determined using a Leeman Prodigy inductively coupled plasma optical emission spectrometer (ICP-OES). Before analysis, the samples were cleaned, dried, and powdered, then subjected to high-temperature ignition at 1,000°C for 2 h in a crucible. The mass difference before and after heating was recorded to calculate the loss on ignition (LOI). The samples were subsequently fused with lithium metaborate (LiBO₂), and the resulting solutions were transferred into volumetric flasks, diluted to a predetermined volume, and weighed for subsequent analysis. For trace element analysis, a high-temperature and high-pressure closed digestion method was employed for sample preparation, followed by analysis using an Agilent 7,700X inductively coupled plasma mass spectrometer (ICP-MS). Whole-rock Sr-Nd isotopic measurements were performed using a Nu Instruments Nu Plasma II MC–ICP–MS device. BCR-2 and AGV-2 were used as external standards, while GSB was employed to monitor Nd isotope measurements.

The zircon U-Pb dating results for the samples from the Xietongmen and Longger granites are presented in Table 1.

The zircon grains extracted from the Xietongmen granite are highly euhedral and predominantly prismatic, with grain sizes ranging from 30 to 150 μm and aspect ratios of 1:1 to 3:1. They show clear magmatic oscillatory zoning in the Cathodoluminescence (CL) images (Figure 3A). The Th/U ratios range from 0.67 to 1.27, aligning with the characteristics of magmatic zircons (Hoskin and Ireland, 2000). Among the 20 zircon spots analyzed for U-Pb isotopes, spots 3 (103 ± 1 Ma) and 8 (107 ± 1 Ma) yielded older ages, suggesting that they are inherited zircons. The remaining 18 zircon spots exhibit a relatively concentrated age distribution on the U-Pb concordia diagram (Figure 4A). The weighted mean ²⁰⁶Pb/²³⁸U age of these zircons is 95.9 ± 0.6 Ma (MSWD = 1.3), representing the crystallization age of the Xietongmen granite formed in the Late Cretaceous.

Similar to the Xietongmen granite, the Longger granite contains highly euhedral zircon grains, mostly prismatic, with grain sizes ranging from 50 to 150 μm and aspect ratios of 1:1 to 3:1. Their CL images display clear magmatic oscillatory zoning (Figure 3B). The Th/U ratios range from 0.36 to 2.34, consistent with the characteristics of magmatic zircons (Hoskin and Ireland, 2000). Among the 30 zircon spots analyzed, spots 1, 7, 19, and 22 yielded older ages, suggesting that they are inherited zircons. The remaining 26 spots exhibit a relatively concentrated age distribution on the U-Pb concordia diagram (Figure 4B). The weighted mean ²⁰⁶Pb/²³⁸U age of these zircons is 79.5 ± 0.4 Ma (MSWD = 1.04), representing the crystallization age of the Late Cretaceous Longger granite.

Whole-rock major and trace-element compositions of the Xietongmen and Longger granites are presented in Table 2.

All samples from the Xietongmen granite exhibit comparatively low LOIs (0.92–2.09 wt%). They are characterized by high SiO₂ (68.16–74.17 wt%), Al₂O₃ (12.84–16.24 wt%), K₂O (2.83–4.65 wt%), Na₂O (2.77–4.33 wt%) contents, and low CaO (1.79–2.65 wt%), TiO₂ (0.45–0.60 wt%), P₂O₅ (0.08–0.15 wt%), MgO (0.76–1.34 wt%) contents. The total alkali (Na₂O + K₂O) contents vary between 7.08 wt% and 8.40 wt%, and the Rittmann indices (σ = 1.69–2.73) remain below 3.3.

Longger granite samples have low LOIs (0.23–0.80 wt%), high SiO₂ (66.29–72.58 wt%), K₂O (3.36–6.49 wt%), Na₂O (3.23–4.85 wt%), Al₂O₃ (14.64–16.27 wt%) contents, and low Fe₂O₃ ^T (0.58–2.88 wt%), MgO (0.17–1.07 wt%, Mg^# = 27.88–47.42) contents. Total alkali contents range from 8.00 to 10.37 wt%, and Rittmann indices vary between 2.49 and 4.11.

In the K₂O + Na₂O versus SiO₂ diagram (Figure 5A), all Xietongmen granite samples plot within the granite field, and most Longger granite samples fall into the quartz monzonite field. In the K₂O versus SiO₂ diagram (Figure 5B), most Xietongmen and Longger granites plot within the high-K calc-alkaline field. In the A/NK versus A/CNK diagram (Figure 5C), most Xietongmen samples fall into the weakly peraluminous field, whereas Longger samples are predominantly metaluminous. Thus, the Xietongmen granite exhibits high-K calc-alkaline and weakly peraluminous geochemical characteristics, while the Longger granite is characterized by high-K calc-alkaline and metaluminous characteristics.

Regarding trace elements, the total rare earth element (∑REE) concentrations of the Xietongmen granite samples range from 16.76 ppm to 128.20 ppm, indicating a low concentration. The light REE to heavy REE (∑LREE/∑HREE) ratios vary from 2.43 to 13.88, while the chondrite-normalized (La/Yb)_N ratios range from 2.10 to 16.90, indicating significant enrichment of LREE relative to HREE and notable fractionation between the two rare earth groups.

The Longger granite samples show ∑REE concentrations between 141.72 ppm and 296.98 ppm. The LREE/HREE ratios vary between 5.53 and 27.72, and the (La/Yb)_N values vary between 5.60 and 58.29, showing strong LREE enrichment and fractionation between LREE and HREE.

Both the Xietongmen granite and Longger granite exhibit similar chondrite-normalized rare earth element (REE) patterns (Figures 6A,C), characterized by significant enrichment in LREEs and strong depletion in HREEs, with a rightward-sloping trend in the chondrite-normalized REE pattern. The primitive mantle-normalized multi-element spider (Figures 6B,D) diagram demonstrates that the Xietongmen and Longger granite samples exhibit enrichment in K, Rb, Th, U, and Pb, while showing depletion in Nb, Ta, Ti, and P.

Whole-rock Sr–Nd isotopic compositions for the Xietongmen granite samples are listed in Table 3. Based on the granite’s crystallization age of 96 Ma, the ⁸⁷Sr/⁸⁶Sr ratios range from 0.706045 to 0.706255, and initial ⁸⁷Sr/⁸⁶Sr ratios [(⁸⁷Sr/⁸⁶Sr)_i] vary between 0.7042 and 0.7044. The ¹⁴³Nd/¹⁴⁴Nd values range from 0.512744 to 0.512781, and initial ¹⁴³Nd/¹⁴⁴Nd ratios [(¹⁴³Nd/¹⁴⁴Nd)_i] range from 0.512657 to 0.512707. The ε_Nd (t = 96 Ma) values vary between +2.78 and +3.76, and all of the samples have relatively young two-stage Nd model ages (T _DM2) that vary between 591 and 670 Ma. The five granite samples analyzed exhibit positive ε_Nd(t) values, with a narrow range of variation (within one ε unit), indicating a relatively homogeneous magma source.

The Xietongmen granite samples are characterized by high SiO₂ (68.16–74.17 wt%), Al₂O₃ (12.82–16.24 wt%), Sr (226–364 ppm), low MgO (0.76–1.34 wt%), Y (2.93–12.28 ppm), and Yb (0.40–1.47 ppm) contents, and corresponding high Sr/Y (24.59–77.13) and La/Yb (2.10–16.90) ratios. All samples display LREE enrichment and depletion in HREEs and HFSEs (such as Nb and Ta), with slightly positive Eu anomalies. These geochemical features are comparable to those of adakitic rocks (Defant and Drummond, 1990; Castillo, 2006), which are generally defined by Al₂O₃ ≥ 15 wt%, MgO ≤3 wt% (rarely exceeding 6 wt%), Y < 18 ppm, Yb ≤ 1.9 ppm, Sr > 400 ppm, Sr/Y > 20, with slightly positive Eu anomalies (Defant and Drummond, 1990; Castillo, 2006).In the Sr/Y versus Y (Figure 7A) and (La/Yb)_N versus Yb_N (Figure 7B) diagrams, the Xietongmen granite samples are positioned within the adakite field, showing that they have an adakitic geochemical signature.

Adakite was initially regarded as derived from the partial melting of a young subducted oceanic slab (Defant and Drummond, 1990). However, recent studies propose several possible formation mechanisms for adakite, as follows: (1) partial melting of subducted oceanic crust (Rapp et al., 1999; Zhu et al., 2009b); (2) partial melting of thickened lower continental crust (Atherton and Petford, 1993; Hou et al., 2013); (3) partial melting of subducted continental crust (Wang et al., 2008); (4) fractional crystallization of primary basaltic magma (Macpherson et al., 2006).

Adakite derived from the fractional crystallization of primary basaltic magma typically leads to a distinct negative Eu anomaly and positive correlations between Sr/Y, Dy/Yb, La/Yb, and SiO₂ (Macpherson et al., 2006). However, the Xietongmen adakitic rock has no such characteristics. In the La versus La/Yb plot (Figure 7C), its geochemical signature is more consistent with partial melting than fractional crystallization. Therefore, the formation of the Xietongmen adakitic rock cannot be attributed to fractional crystallization.

Adakitic rocks formed by partial melting of subducted continental crust typically exhibit higher Th/Ba and Rb/Ba ratios, and negative ε_Nd(t) values (Lai and Qin, 2013). However, Xietongmen adakitic rock samples have low Th/Ba (0.03–0.07), Rb/Ba (0.24–0.48) ratios, and positive ε_Nd(t) values (2.78–3.76). Moreover, the Xietongmen adakitic rock formed at 95.9 Ma during the northward subduction of the Neo-Tethyan oceanic lithosphere beneath the Eurasian continent, rather than during the subduction of continental crust. Therefore, this interpretation is also ruled out.

Adakitic rocks derived from partial melting of a subducted oceanic slab typically have low K₂O content (Defant and Drummond, 1990). The samples of Xietongmen adakitic rock have comparatively high K₂O (2.83–4.65 wt%), similar to the K₂O contents (2.9–4.1 wt%) of adakitic rocks derived from lower crustal melting (Defant and Drummond, 1990). Their Nb/Ta ratios (7.71–15.43) are also close to those of continental crust (11–14) (Taylor and McLennan, 1995). Furthermore, all samples of the Xietongmen adakitic rock have low MgO (0.76–1.34 wt%), Cr (13.17–21.16 ppm), and Ni (5.53–11.55 ppm) contents. In the Ni versus Cr (Figure 7D), the Cr versus SiO₂ (Figure 7E), and MgO versus SiO₂ (Figure 7F) diagrams, most samples fall within the lower crust partial melting field. In the ε_Nd(t) versus ⁸⁷Sr/⁸⁶Sr plot (Figure 8), the specimens fall within the adakitic field associated with the thickened magnesian lower crust of southern Tibet.

The Xietongmen adakite samples exhibit enrichment in LILEs and LREEs, showing depletion in HFSEs. These samples exhibit relatively low initial ⁸⁷Sr/⁸⁶Sr isotopic values (0.7042–0.7044) and positive ε_Nd(t) values (+2.78 to +3.76), with two-stage Nd model ages of 591–670 Ma, indicating that the pluton likely derived from the partial melting of a juvenile lower crust. Additionally, Y/Yb ratios (7.42–10.14; ≈10) and (Ho/Yb)_N ratios (0.73–0.93; ≈1) suggest that amphibole was the principal residual phase in the magma source, with some residual garnet also present (Hu et al., 2014). The high Ba and Sr contents and the slightly positive Eu anomalies suggest that plagioclase did not remain as a residual phase in the source region. The relatively high Mg^# values (> 40) of some samples suggest minor involvement of mantle-derived material during magma evolution. The specimens plot within the crust and mantle field in the (La/Yb) N versus δEu diagram (Figure 9). Therefore, the magma evolution of the Xietongmen adakitic rock was primarily controlled by partial melting, with a residual mineral assemblage dominated by amphibole and garnet, and lacking plagioclase. The geochemical features suggest that the Xietongmen adakite was likely generated through the partial melting of a juvenile thickened lower crust, with a slight input from mantle-sourced components.

Granites can be categorized into I-, S-, A-, and M-type according to their origin and geological setting (Champion and Chappell, 1992; Chappell and White, 1992). M-type granites formed by fractional crystallization of mantle-derived magma and typically exhibit geochemical characteristics of high Mg^# (50–60) values and low SiO₂ (<65 wt%) contents (Bowden et al., 1979), whereas the Longger granite displays properties of high SiO₂ (66.29–72.58 wt%) contents and low Mg^# (27.88–47.42) values, thus excluding its classification as an M-type granite. The samples exhibit Zr + Nb + Ce + Y concentrations ranging from 180 ppm to 245 ppm and an alkalinity index (AI) of 0.50–0.70, which do not correspond to A-type granites (typically Zr + Nb + Ce + Y ≥ 350 ppm, AI > 0.85) (Whalen et al., 1987; Eby, 1990), thus excluding an A-type granite classification. In the (Na₂O + K₂O)/CaO versus (Zr + Nb + Ce + Y) (Figure 10A) and the Fe₂O₃ ^T/Mg versus (Zr + Nb + Ce + Y) (Figure 10B) diagrams, the Longger granite plots within the OTG (undifferentiated I, S, M-type granites) granite fields, suggesting that it could be either I-type or S-type granite. The Longger granite is also metaluminous, with an A/CNK ratio of 0.89–1.01. This ratio is below the threshold of 1.1 that distinguishes S-type granite (A/CNK >1.1) and is more consistent with I-type granite (Whalen et al., 1987; Chappell and White, 1992).

Previous research has indicated that P₂O₅, Th, Y, Rb, and other major and trace elements can effectively differentiate I-type and S-type granites (Whalen et al., 1987; Wolf and London, 1994). In S-type granites, P₂O₅ content increases or remains constant with increasing SiO₂ (Whalen et al., 1987; Wolf and London, 1994), whereas in I-type granites, it decreases as SiO₂ increases. Furthermore, I-type granites typically exhibit high concentrations of Th and Y, with Th showing a positive correlation with Rb (Chappell, 1999). The Longger granite samples exhibit a clear negative correlation in the P₂O₅ versus SiO₂ scatter plot, indicating an evolutionary trend characteristic of I-type granites (Figure 10C). Petrographic analysis shows that hornblende, a characteristic mineral of I-type granite, is also present in these samples. The Longger granite is classified as an I-type granite based on these geochemical and mineralogical features.

I-type granite has been widely regarded as formed through three genetic processes: (1) fractional crystallization of mantle-derived basaltic magma (Cawthorn and Brown, 1976; Wyborn et al., 1987); (2) fractional crystallization of crust-mantle hybrid magma (Turpin et al., 1990; Barbarin, 1996; Chappell et al., 2012); (3) partial melting of the lower crust heated by mantle magma underplating (Castro et al., 1991). I-type granites formed by the first way typically exhibit large-scale coeval basic rocks in their vicinity. However, no extensive basic rocks have been found in the study area. In the La-La/Yb diagram (Figure 7C), the Longger granite exhibits a partial melting trend rather than a fractional crystallization trend. Therefore, the Longger granite was not likely formed by the first one. For the second process, granites formed by crust-mantle mixing typically contain mafic microgranular enclaves, whereas no such dark enclaves have been found in the Longger granite (Turpin et al., 1990; Barbarin, 1996; Chappell et al., 2012). Additionally, the Mg^# values of the Longger granite samples range from 27.88% to 47.42% (mean 38.05%), which do not exhibit the high Mg^# values indicative of interaction with the mantle (Rapp et al., 1999; Hu et al., 2012). Therefore, the Longger granite was not formed in the second way.

The Longger granite samples exhibit high SiO₂ (66.29%–72.58 wt%) and Al₂O₃ (14.64 wt.%–16.27 wt%) contents. They are enriched in K, Rb, Th, U, and Pb, while displaying depletion in Nb, Ta, Ti, and P, indicating a dominantly continental crustal source (Hu et al., 2017). The Th/U ratios (2.3–11.2, mean 8.5) are comparable to those of the lower crust (6.0) (Rudnick and Gao, 2003). As two high-field-strength elements (HFSEs) with similar incompatibility, Nb and Ta remain relatively stable during magmatic evolution, making them effective indicators of magmatic source characteristics and plutonic evolution (Pfänder et al., 2007). All samples of the Longger granite Nb/Ta ratios (5.06–11.06) are lower than those of the depleted mantle (∼17) but comparable to those of the continental lower crust (∼11) (Pfänder et al., 2007). In the (La/Yb)_N versus δEu diagram (Figure 9), most rock specimens fall within the crustal-type granite region, further supporting a lower crustal origin (Rapp and Watson, 1995; Rudnick and Fountain, 1995). Additionally, the La versus La/Yb (Figure 7C) diagram indicates a distinct partial melting trend, and the Ni versus Cr (Figure 7D), Ni versus SiO₂ (Figure 7E), and MgO versus SiO₂ (Figure 7F) diagrams suggest derivation from lower crustal partial melting. The granite exhibits negative Eu anomalies (δEu = 0.23–0.78) and Ba, Sr depletion characteristics, indicating that it may have undergone plagioclase fractional crystallization (Chappell and White, 1992). In summary, the Longger granite was formed by partial melting of the lower crust. During its magmatic evolution and pluton emplacement, it underwent plagioclase fractional crystallization.

Further insights into the tectonic setting can be derived using granite tectonic classification diagrams (Pearce et al., 1984). In the Fe₂O₃ ^T/(Fe₂O₃ ^T + MgO) versus SiO₂ diagram (Figure 11A), the specimens plot in the island arc granite, continental arc granite, and collisional granite fields. The Rb versus Y + Nb diagram (Figure 11B) places all specimens in the volcanic arc granite field, while the Nb/Zr versus Zr diagram (Figure 11C) indicates a subduction-related setting. Additionally, most samples plot within the active continental margin field in the Th/Yb versus Ta/Yb diagram (Figure 11D). Combining these geochemical characteristics with geochronological data and the contemporaneous tectonic background of Late Cretaceous magmatism, it is evident that both the Xietongmen and Longger granites were emplaced in a subduction-related environment, likely linked to the northward subduction of the Neo-Tethyan oceanic plate beneath the Eurasian continent.

The Late Cretaceous igneous rocks within the Gangdese Tectonic Belt are interpreted as the result of large-scale magmatic activity triggered by the subduction of the Neo-Tethyan oceanic plate beneath the Lhasa Block (Murphy, 2019; Tassara et al., 2020; van Hinsbergen et al., 2021; Moyen et al., 2021; Luffi and Ducea, 2022). However, the specific subduction dynamics of the Neo-Tethyan Ocean during this period remain debated, with three prevailing models: (1) flat-slab subduction of the Neo-Tethyan Ocean (Coulon et al., 1986; Wen et al., 2008a; Wen et al., 2008b; Zheng et al., 2014); (2) slab rollback of the Neo-Tethyan oceanic lithosphere (Chung et al., 2005; Ma et al., 2013a; Ma et al., 2013b); (3) ridge subduction of the Neo-Tethyan oceanic lithosphere (Zhang et al., 2010; Zhu et al., 2013).

Wen et al. (2008a), Wen et al. (2008b) collected 25 samples from different regions of the Gangdese tectonic belt, including diorite, granodiorite, gabbro, and granite. Samples emplaced between 103 and 80 Ma were selected for systematic geochemical analysis, and data previously published by Quidellieur et al. (1997) were compiled. The results show that most Late Cretaceous samples exhibit adakitic geochemical characteristics. However, adakitic magmatism likely had a short duration, occurring only over a limited period in a specific tectonic domain. This indicates that the Neo-Tethyan Ocean underwent flat-slab subduction during the Late Cretaceous, followed by slab rollback. In general, flat subduction leads to the expulsion of mantle wedge materials due to compressive forces from the subducting slab, resulting in reduced mantle wedge materials in the subduction zone and thereby inhibiting the formation of mantle-derived mafic magmas. This mechanism is inconsistent with the widespread mafic magmatism reported in the early Late Cretaceous (Gutscher and Peacock, 2003; Kapp et al., 2005; Kapp et al., 2007; Kay et al., 2005; Xu et al., 2015; Zhu et al., 2018).

Ma et al. (2013a), Ma et al. (2014) discovered norites and hypersthene-bearing hornblendites emplaced at ∼93 Ma in the Milin area of the southern Gangdese tectonic belt. Geochemical and isotopic data for both rock types indicate that their parental magmas were likely derived from the interaction between upwelling asthenospheric mantle and metasomatized lithospheric mantle (Ma et al., 2013a; Ma et al., 2014). Therefore, they proposed that the early Late Cretaceous magmatic “flare-up” event was triggered by asthenospheric mantle upwelling caused by rollback of the subducted Neo-Tethyan oceanic slab. If the slab rollback mechanism is valid, the oceanic plate should retreat in the direction opposite to subduction. Consequently, magmatic rocks formed during this period should exhibit a progressively younger age trend from north to south. However, such a trend is not observed in the Gangdese Tectonic Belt (Zhu et al., 2009b; Zhu et al., 2018; Zheng et al., 2014).

Zhang et al. (2010) reported high-temperature charnockites formed at 86–90 Ma in the eastern Gangdese Tectonic Belt. Combining coeval calc-alkaline magmatism with the crystallization temperature (900°C) and pressure (1.0 GPa) of the high-temperature adakitic charnockites, they proposed that these high-temperature and low-H₂O activity charnockites formed during the subduction of the Neo-Tethyan mid-ocean ridge. Subsequent studies have further supported this model, offering plausible explanations for various geological observations (Zhu et al., 2013; Xu et al., 2015).

Guo et al. (2011), Guo et al. (2013) divided magmatic events in the Gangdese Belt into five stages through zircon U-Pb dating, among which the mid-ocean ridge subduction stage occurred at 89–80 Ma. Based on fieldwork in the Medog-Bomi area, Pan et al. (2014) proposed that the subduction of the mid-ocean ridge beneath the Lhasa Terrane occurred at ∼ 95–80 Ma. Zheng et al. (2014) argued that the subduction of the Neo-Tethyan mid-ocean ridge beneath the southern Lhasa Terrane occurred at 105–76 Ma, based on the presence of high-temperature granulite-facies metamorphism and Late Cretaceous basalt basement in the Langxian area, and the resultant heat accumulation caused melting of the overlying crust, leading to a magmatic peak. Ma et al. discovered hypersthene-bearing hornblendites formed at ∼93 Ma in the Milin area, whose pressure-temperature (P-T) conditions are consistent with the mid-ocean ridge subduction model. Therefore, for the subduction model of the Neo-Tethyan Ocean during the Late Cretaceous, it can be inferred that a northward mid-ocean ridge subduction regime operated in the early Late Cretaceous.

Most of the above studies also support the southward retreat of the subducting Neo-Tethyan slab beneath the southern Lhasa Terrane during the late Late Cretaceous. However, discrepancies remain regarding the timing of slab rollback, primarily whether it occurred before ∼70 Ma (Chung et al., 2005; Zhang et al., 2010; Ma et al., 2013a; Ma et al., 2013c) or after 70 Ma (Lee et al., 2009; Lee et al., 2012). Therefore, the approximate timing of Neo-Tethyan slab retreat remains unclear (Zhu et al., 2009a; Zhu et al., 2018; Zheng et al., 2014).

These studies show that during the mid-ocean ridge subduction of the Neo-Tethyan Ocean, high-temperature asthenospheric material upwelled through slab windows within the subducted oceanic ridge. These thermal anomalies transferred heat to the lithospheric mantle, raising its temperature. The heated lithospheric mantle then transferred heat to the overlying thickened lower crust, inducing partial melting and ultimately triggering large-scale magmatism. The Xietongmen adakitic rock was likely generated as a product of ridge subduction during this period (Figure 12A). It originated through the partial melting of a thickened lower crust and did not experience significant fractional crystallization. Therefore, its geochemical signatures can serve as an indicator for estimating crustal thickness. According to the crustal thickness estimation formula proposed by Chapman et al. (2015): H = 1.11 Sr / Y + 8.05 (1)

The Sr/Y values of the selected samples were used to estimate the corresponding crustal thickness, adhering to the applicable range of the formula and the data exclusion criteria. The results show that the median Sr/Y value of the Xietongmen adakitic rock samples is 45.47, yielding an estimated crustal thickness of 49.87 km. Additionally, using the formula proposed by Hu et al. (2017): H = 0.67 Sr / Y + 28.21 (2)

The estimated crustal thickness is 54.10 km. The consistency between these two estimates suggests that the crustal thickness of the Gangdese Tectonic Belt reached approximately 50 km in the early Late Cretaceous. This calculation result aligns with the conclusion of Zhu et al. (2023) based on a comprehensive analysis of petrological, geochemical, and geochronological data, which states that “the Gangdese crust experienced local thickening at ∼90 Ma.”

Calculating with Equations 1, 2, the estimated crustal thickness for forming the Longger I-type granite is 33.74 km and 43.86 km. Both values are lower than those for the Xietongmen adakitic rock, suggesting that the Longger I-type granite formed in a relatively extensional setting with a thinner crust, in contrast to the thickened crustal environment of the Xietongmen adakitic rock.

Thus, the Xietongmen adakitic rock, developed in the early Late Cretaceous of the Gangdese Belt, formed under a crustal thickening regime, while the Longger I-type granite, emplaced in the late Late Cretaceous, formed under conditions of crustal thinning. Ridge subduction likely occurred in the initial phase of the Late Cretaceous, followed by slab rollback in the late Late Cretaceous. The slab rollback facilitated extensive asthenospheric mantle upwelling and underplating, which thermally perturbed the lithospheric mantle. The heated lithospheric mantle subsequently transferred heat to the lower crust, causing partial melting and the formation of Longger I-type granite, which is primarily crustal-derived (Figure 12B). This interpretation is consistent with previous studies on crustal recycling in the Gangdese Belt (Chung et al., 2005; Zhang et al., 2010; Ma et al., 2013a; 2013b). However, this hypothesis requires further testing.