Introduction

The petrographic texture and genetic implications of rapakivi granite are unique. Sederholm (1891) first defined the texture of typical rapakivi as plagioclase-mantled alkali feldspar ovoids in granitoid rocks, and this texture has since been observed in numerous granitic systems worldwide (Vorma 1976; Haapala and Rämö 1992, 1999 and Haapala 1995, 2005). Rapakivi granite is of considerable geological interest as it provides information about tectonic and crustal evolution (Larin et al. 2021; Prasanth et al. 2023). Most typical Proterozoic rapakivi granites are found in cratons, with an A-type affinity that is commonly associated with continental rifts (Anderson 1980; Haapala and Rämö 1992, 1999 and Haapala 1995; Haapala et al. 2005). However, some rapakivi-textured granites occur in orogenic belts, including the Tataleng pluton in the Northern Qaidam Orogen (Lu et al. 2009; Liu et al. 2013a; Solomovich and Trifonov 2014; Sun 2017; Chen et al. 2019b), their crystallization process and tectonic setting remain a subject of debate. The main point of contention is whether they were formed during a transitional stage from compression to extension, marking the end of the main orogenic collision and the beginning of intracontinental orogeny (Lu et al. 2009; Larin 2009; Solomovich and Trifonov 2014; Sun 2017). Studies on rapakivi-textured feldspars have provided important records of plutonic processes (Sederholm 1891; Nekvasil 1991 and Haapala 1995; Haapala et al. 2005; Larin et al. 2021); however, the crystallization process of these feldspars is still controversially discussed (Hibbard 1981; Nekvasil 1991; Bussy 1990; Dempster et al. 1994; O’Brien et al. 2019; Alves et al. 2021).

Fig. 1
figure 1

Global distribution of rapakivi granites and the locations of several UHP terranes (modified and adapted from Rämö and Haapala (1995), Liou et al. (2009), Zhang et al. (2011b), and Solomovich and Trifonov (2014); Dupangling pluton after Zhao et al. (2023)). Numbers indicating age are in Ma

Full size image

Voluminous Paleozoic granites in the North Qinling Belt (NQB) preserve important information regarding tectonic evolution (Wang et al. 2013; Wang et al. 2015).

Analytical methods

All analyses were conducted at the State Key Laboratory of Continental Dynamics, Northwest University, **’an, China.

Sample collection and specimen preparation for petrographic analysis

Fieldwork involved collecting 7 kg of fresh samples from every pluton location (Fig. 2c), followed by indoor washing with clear water to eliminate surface residues for subsequent petrographic, geochemical, and U–Pb age studies. Petrography and electron probe microanalysis (EPMA) involved the preparation of six polished thick Sect. (50 μm) and four thin Sect. (30 μm). Additionally, three samples from each location were retained for geochemical analysis.

Analysis of mineral composition

The predominant mineral components were quantified using a JXA-8230 electron probe microanalyzer (JEOL, Tokyo, Japan), operated at an acceleration voltage of 15 kV, beam current of 1 × 10− 8 A, and beam spot size of 1 μm, under chamber vacuum pressure conditions of approximately 2.7 × 10− 4 Pa. Microstructural analysis was complemented with back-scattered electron images using EPMA, maintaining a working distance of 10.0 mm. Analysis targeted crystals using thallium acid phthalate (TAP: Si, Al, Mg, and Na), lithium fluoride (LiF: Fe and Mn), and pentaerythritol (PET: K, Ca, Ti, and P) as standards. Natural standards included Si (quartz), Al (jadeite), Mg (olivine), Mn (rhodonite), Fe (hemat ite), K (sanidine), Na (jadeite), Ca (diopside), and Ti (rutile), with continuum intensity total Z (atomic number) absorption fluorescence (CITZAF) corrections applied. EPMA examination encompassed various zoning bands and inclusion minerals within phenocrysts and major minerals within the matrix. FeOT represents total iron (oxide) content. The numbers of cations in feldspars were calculated after Ribbe (1983), in amphiboles after Hawthorne et al. (2012), and for biotite mica after Rieder et al. (1998), which all calculated based on the oxygen atom in a molecule.

U–Pb age dating of zircon

Up to 1500 zircons were extracted from approximately 8 kg of fresh granite samples using heavy liquid and magnetic separation techniques. These zircons were then meticulously examined under an optical microscope (Nikon SMZ800N, Shanghai, China) and embedded in epoxy resin. Following polishing with a corundum polishing liquid and removal of surface residues using an ultrasonic cleaner, the zircon grains underwent observation via optical (transmitted and reflected light) and cathodoluminescence (CL) techniques to reveal their internal texture.

Subsequently, the zircon grains underwent analysis for U–Th–Pb isotopes using laser ablation inductively coupled plasma mass spectrometry (LA–ICP–MS). CL imaging of the selected zircons was conducted using a Quanta 400 FEG environmental scanning electron microscope equipped with an Oxford energy-dispersive spectroscopy system and a Gatan CL3 + detector.

For zircon trace element analysis and U–Pb age dating, an Agilent 7500a ICP–MS from Hewlett Packard, the ComPex102 Excimer laser (working gas ArF, wavelength 193 nm) from Lambda Physik (Germany), and a GeoLas 200 M optical system from MicroLas were employed. The laser beam possessed a spot diameter of 33 μm, and the laser ablation depth on the sample was 40 μm.

Isotope ratios and trace element concentrations were calculated offline using GLITTER (Ver. 4.0), with standard zircon 91500 utilized as an external standard for isotope ratio correction. NIST 610 served as the external standard, and 29Si as the internal standard for elemental concentration calculations. Zircon U–Th–Pb age calculations and Concordia diagram plotting were performed using Isoplot (Ver 4.16, Ludwig 2003).

Lu–Hf isotope analyses of zircon

In situ Lu–Hf isotope analysis of zircon was conducted using a Nu PlasmaII MC–ICP–MS, with a laser spot size of 44 μm. Isotope fractionation correction for Lu–Hf was computed using an exponential law. To ensure precise 176Hf values, corrections were applied for the isobaric interference with 176Hf caused by 176Lu and 176Yb, with respective ratios of 176Lu/175Lu = 0.02656 and 176Yb/173Yb = 0.78696. Normalization of Yb and Hf isotope ratios was carried out against 173Yb/171Yb = 1.12346 and 179Hf/177Hf = 0.7325, respectively, to rectify instrumental mass bias.

Calibration of isotope fractionation utilized two external standards, zircon standard 91,500 and Mud Tank, with analyses performed once every 12 analyses. Data acquisition employed the time-resolved analysis mode with an integration time of 0.2 s. Background acquisition time was set at 30 s, while sample integration time was 50 s.

Calculation of εHf values employed a decay constant of 1.867 × 10− 11 yr− 1 for 176Lu, with chondritic ratios of 176Hf/177Hf = 0.282772 and 176Lu/177Hf = 0.0332 (Blichert-Toft and Albarède 1997), and a U–Pb crystallization age of 447 Ma. Single-stage model ages (TDM1) were determined with respect to the depleted mantle, characterized by a present-day 176Hf/177Hf ratio of 0.28325 and a 176Lu/177Hf ratio of 0.0384 (Griffin et al. 2000). Two-stage model ages (TDM2) were calculated by projecting the initial 176Lu/177Hf values of zircon back to the depleted mantle growth curve using 176Lu/177Hf = 0.015 to represent the average continental crust.

Bulk whole-rock main and trace element geochemistry

All samples designated for whole-rock chemistry underwent a process of pulverization and crushing to achieve a particle size of 5–10 mm, accomplished using both a jaw crusher and a tungsten-carbide shatter box with a 200 mesh sieve. Given the substantial size of rapakivi granite minerals, the samples underwent multiple rounds of mixing via the reduction method to ensure the homogeneity of the powdered composition. In the course of sample preparation, 0.70 g of dried sample powder was meticulously weighed and blended with LiF (0.40 g), Li2B4O7 (5.20 g), and NH4NO3 (0.30 g). Subsequently, the mixture was subjected to heating for 8 min at 1200 °C in an electric high-frequency furnace, resulting in the formation of fused bead sheets suitable for testing.

Major element analysis was conducted using an X-ray fluorescence spectrometer (Rigaku RIX 2100) equipped with an automated sample introduction system. Standards BCR-2 and GBM07105 were utilized, with duplicates analyzed every 10 samples. The acquired data exhibited a relative standard deviation of approximately 2%.

Trace and rare earth elements (REE) were analyzed employing an Agilent 7500a ICP–MS, with BHVO-2 and AGV-2 serving as standard reference materials. Blanks were analyzed for every 10 samples. Approximately 50 mg of powder underwent dissolution in an HF-HNO3-HClO4 mixture via a series of heating and evaporation procedures to prepare the sample. The resultant solution was then diluted with deionized water to a volume of 80 g in a PET bottle, with Rh employed as the internal standard. Notably, the relative error (RE) for the trace element test data predominantly exceeded 10% (Yuan et al. 2008).

Results

Petrographic observations

This study investigated grayish-white Niujiaoshan granite with a rapakivi texture. The studied granite predominantly appeared massive (Fig. 3a), with localized occurrences of dark enclaves (Fig. 3b). Feldspar phenocrysts in the granite ranged in size from approximately 0.2 × 0.2 to 1.3 × 1.5 cm, constituting approximately 55% of the total volume. Phenocrysts consisted mainly of ovoid K-feldspars, some of which were surrounded by irregular white plagioclase rims of different widths (Fig. 3c, d). The matrix was primarily composed of plagioclase, K-feldspar, quartz, a small quantity of biotite, and amphibole (Fig. 3e). Additionally, a mafic enclave composed of biotite, amphibole, plagioclase, and needle apatites was discovered (Fig. 3f). The phenocryst cores consisted of single crystal or intergrown K-feldspars composed of multiple crystals, several of which were perthites (Fig. 4a), and these were rich in microscopic inclusions (such as plagioclase, quartz, biotite, and amphibole) (Fig. 4b). The mantle was composed of oligoclase with a few quartz inclusions, and the boundary between the ovoid K-feldspar and mantle was irregular (Fig. 4c, d).

Fig. 3
figure 3

(a and b) Field occurrences and (c-f) microtextures of Niujiaoshan rapakivi-textured granite. (a) Rapakivi-textured granite. (b) Mafic enclave with plagioclase phenocrysts in the host rapakivi-textured granite. (c) and (d) Rapakivi texture feldspar in granite. (e) Mineral assemblage in matrix. (f) Mafic enclave mineral assemblage. (Kfs: K-feldspar, Pl: plagioclase, Qz: quartz, Amp: amphibole, Bt: biotite, mineral abbreviations after Whitney and Evans 2010)

Full size image
Fig. 4
figure 4

Micrograph and back-scattered electron (BSE) images of rapakivi-textured granite. (a) Exsolution albite of porphyritic core. (b) Inclusion assemblage in the nucleus of phenocryst K-feldspar. (c) Mineral assemblage in the matrix and rapakivi-textured phenocrysts. (d) Irregular boundary between the phenocryst K-feldspar and mantle oligoclase (Ab: albite; Ep: epidote; mineral abbreviations after Whitney and Evans 2010)

Full size image

Mineral chemical composition

Through EPMA analysis, we determined the chemical composition of the various zoning bands and inclusion minerals in the phenocrysts, as well as the major minerals in the matrix. The results are summarized in Table 1 and S1.

Table 1 Main element oxide contents in minerals from the Niujiaoshan rapakivi-textured granite by electronprobe microanalysis (EPMA), in weight% (wt%). FeOT represents total iron (oxide) content. Numbers of cations in feldspars calculated as previously described by Ribbe (1983), in amphiboles by Hawthorne et al. (2012), and for biotite mica cf. Rieder et al. (1998)
Full size table

Alkali feldspar

The alkali feldspar in the rock primarily consisted of K-feldspar in the phenocryst core, a small amount of perthite, and a small amount of K-feldspar in the matrix. The EPMA results showed that the composition of the K-feldspar in the phenocryst core and matrix was similar. The K2O content ranged from 16.09 to 16.66%, with end-member compositions of Or93.10–94.63Ab5.37–6.90 and Or92.83–97.09Ab2.75–7.17. The exsolution lamellae of the K-feldspar were composed of albite (with end-member compositions of Ab97.41–98.86Or1.07–1.41; Fig. 5a).

Fig. 5
figure 5

(a) Plagioclase classification diagram (Smith and Brown 1988). (b) BSE image and composition profile of plagioclase in the matrix

Full size image

Plagioclase (Pl)

The plagioclase was divided into four subcategories based on texture: inclusions in the phenocryst cores (Pl1), coarse-grained plagioclase in the matrix (Pl2), plagioclase mantle surrounding the K-feldspar phenocryst (Pl3), and fine-grained plagioclase feldspar in the matrix (Pl4). Most of the plagioclase grains had an oligoclase composition, with only a small number of andesine grains in the matrix (Fig. 5a). The end-member compositions of the different occurrences of plagioclase were as follows: Pl1: Ab72.98–77.42An21.80–25.62, Pl2: Ab68.62–80.14An18.84–32.09, Pl3: Ab76.00–84.20An14.92–23.31, and Pl4: Ab72.05–73.50An25.68–26.68. There were no marked variations among the plagioclase components of the different occurrences, but the An component of the rims of the coarse-grained plagioclase (Pl2) and An-rich oligoclase (Pl3) in the ring plagioclase were relatively higher (Fig. 4b). Variations in the composition of coarse-grained plagioclase (Pl2) in the matrix were observed, showing an increasing trend in the An component from the core to the rim (Fig. 5b).

Amphibole (Amp)

Amphibole grains were found inside the feldspar phenocrysts and also within the matrix. However, the compositions of both were not significantly different, falling within the range of magnesio-hornblende in the A(Na + K + 2Ca) versus C(Al + Fe3++2Ti) amphibole classification diagram, except for one tremolite amphibole in the matrix. The only difference was the wide variation in the amphibole composition of the matrix (Fig. 6a).

Fig. 6
figure 6

(a) Diagram of A(Na + K + 2Ca) versus C(Al + Fe3++2Ti) for amphibole (from Hawthorne et al. 2012). (b) Diagram of Mg-(AlVI+Fe3++Ti)- (Fe2++Mn) for biotite (from Foster 1960). Data sources: orogenic rapakivi-textured granites (Wang et al. 2003; Hu et al. 2008; Zhou et al. 2008, Liu et al. 2013); rapakivi granites (Rämö and Haapala 1995; Elliott 2001; Scandolara et al. 2013; Wang et al. 2018; Larin et al. 2021)

Full size image

Biotite (Bt)

Two types of biotite were identified: inclusions in the feldspar phenocrysts and matrix minerals. In the classification diagram of biotite based on chemical composition [Mg-(AlVI+Fe3++Ti)-(Fe2++Mn)], both fell within the magnesian biotite field, but the biotite in the matrix was relatively rich in Mg (Fig. 6b).

U–Pb dating of zircon

The zircons from the samples were typically transparent, colorless, automorphic crystals. The grain size of the zircon ranged from 50 to 300 μm. Most grains exhibited completely polycrystalline shapes, but a few particles were spherical in nature (Fig. 7a). Most zircon grains showed core–rim structures. The cores showed clear oscillatory zoning; their U and Th contents and Th/U values ranged 92.00–2904, 49.40–734, and 0.25–1.07 ppm, respectively (Table 2). The normalized REE patterns showed light REE (LREE) depletion, heavy REE (HREE) enrichment, positive Ce anomalies, and negative Eu anomalies, characteristic of magmatic zircons (Fig. 7b). The rims of the zircons were gray–black and showed poor luminescence in the CL images. Their U and Th contents and Th/U values ranged 189–945, 71.04–282, and 0.20–0.65 ppm, respectively. The zircon rims were also typical of magmatic zircons, except that the rim Th and U contents showed slightly decreasing trends. Few zircons exhibited inherited residual cores with better luminescence, weak oscillatory zoning, and the characteristic REE distribution pattern of magmatic zircon. The LA–ICP–MS dating results (Table 3) indicated no significant difference in age between the core and rim zoning of the zircon crystals. The 206Pb/238U ages of the core ranged from 443 to 452 Ma, whereas those of the rim varied from 443 to 453 Ma. These ages collectively formed a concentrated age cluster in the U–Pb Concordia diagram. The weighted average age of the core was 447 ± 7 Ma (mean squared weighted deviation [MSWD] = 0.089, n = 14). However, because of the loss of Pb in several inherited core regions, accurate ages could not be obtained for these specific data points. Individual analyses of these points yielded concordant 207Pb/235U ages of 1391 ± 11 Ma (Fig. 7c, d).

Fig. 7
figure 7

(a) Cathodoluminescence (CL) images, U–Pb dating (white circles), and εHf(t) values (yellow circles) of zircons from rapakivi-textured granite. (b) Chondrite-normalized REE patterns of zircons from rapakivi-textured granite. (c) U–Pb Concordia diagrams. (d) Enlarged image of U–Pb Concordia diagrams and weighted average age

Full size image
Table 2 Laser ablation inductively coupled plasma mass spectrometry (LA–ICP–MS) trace element analyses of zircon (×10− 6)
Full size table
Table 3 Zircon U–Pb isotope data obtained using LA–ICP–MS
Full size table

Lu–Hf isotopic composition of zircon

To investigate the origin of the granite, Lu–Hf isotope analyses were conducted on the core and rim of zircon crystals based on U–Pb dating. The results revealed that the 176Hf/177Hf values in the core ranged from 0.282439 to 0.282602, whereas those in the rim ranged from 0.282478 to 0.282556. Based on the calculated crystallization age of the rapakivi-textured granite, which is 447 ± 7 Ma, the corresponding εHf (t) values of the core and rim ranged from − 2.04 to + 3.63 (with an average of − 0.36) and − 0.79 to + 2.11 (with an average of 0.47), respectively. Except for one measurement point, most rim data points had slightly higher εHf (t) values than those of the core regions. The TDM2 ages corresponding to the core region and rim were 1195–1553 and 1290–1474 Ma, respectively (Table 4). In the εHf (t) versus t diagram, both were located near the evolution line of chondrite (Fig. 8), indicating that the source rock was mainly composed of ancient crustal material with a small amount of mantle-derived material.

Table 4 Zircon Lu–Hf isotope data obtained using LA–ICP–MS
Full size table
Fig. 8
figure 8

εHf(t) versus t diagram of zircons from the Niujiaoshan rapakivi-textured granite. The shaded part shows the ~ 450-Ma granites in the QOB, which are after Wang et al. (2009), Liu (2014), and Zhang et al. (2013)

Full size image

Geochemistry

Major elements

The SiO2 content of the rapakivi granite ranged from 67.54 wt% to 67.71 wt%, and the Al2O3 content varied between 15.09 wt% and 15.22 wt%, and was rich in alkalis (Na2O + K2O = 7.51–7.76 wt%). The Na2O/K2O value ranged from 1.09 to 1.19, the CaO content ranged from 3.1 wt% to 3.21 wt%, the Fe2O3T content varied between 2.7 wt% and 2.8 wt%, and the MgO and MnO contents ranged from 1.74 wt% to 1.79 wt% and 0.04 wt% to 0.05 wt%, respectively (Table 5). The rock had high Mg# (Mg# = Mg2+/[Mg2+ + Fe2+] = 55.7–56.1) and an A/CNK (molecular Al2O3/[Na2O + K2O + CaO]) value ranging from 0.92 to 0.94, indicating that it belonged to a peraluminous rock series (Fig. 9a). In the Na2O versus K2O granite-type discrimination diagram, the samples were classified as I-type granites (Fig. 9b, d). In the K2O versus SiO2 magma series discrimination diagram, they fell within the high-potassium calc-alkaline region (Fig. 9c).

Table 5 Major (wt%) and trace (×10− 6) element compositions of samples
Full size table
Fig. 9
figure 9

Classificatory diagrams of granite. (a) A/CNK versus A/NK diagram (from Maniar and Piccoli 1989). (b) K2O versus Na2O diagram (from Collins et al. 1982). (c) SiO2 versus K2O diagram (from Peccerillo and Taylor 1976). (d) An-Ab-Or diagram (from Barker 1979). Data sources: ~450 Ma granites in the QOB (same as in Fig. 7); orogenic rapakivi-textured granite and rapakivi granites (same as in Fig. 5)

Full size image

Trace elements

The total REE contents were relatively low in all analyzed samples (ΣREE = 98.70–109 ppm) with little variation. Clear fractionation was observed between LREE and HREE, with (La/Yb)N ranging from 24.93 to 26.69. The overall pattern was characterized by LREE enrichment and relative HREE depletion, displaying a rightward-inclined chondrite-normalized REE distribution curve, and featuring a weak negative Eu anomaly (δEu = 0.85–0.86). The primitive mantle-normalized spider diagram showed an overall depletion of high field strength elements (HFSEs), such as Nb, Ta, and Ti, slight enrichment of large-ion lithophile elements (LILEs), such as Ba, Th, and K, and significant Pb enrichment (Fig. 10).

Fig. 10
figure 10

(a) Chondrite-normalized REE diagram. (b) Primitive mantle-normalized multi-element spider diagram (normalizing data from Sun and McDonough 1989). Data sources align with those in Fig. 8

Full size image

Discussion

Emplacement age of Niujiaoshan rapakivi-textured granite

Most zircons in the Niujiaoshan rapakivi-textured granite were found to have evident core–rim structures, and both the core and rim comprised distinct or weak magmatic oscillatory zone structures. However, the rims showed decreasing Th contents and Th/U trends. The LA–ICP–MS dating results indicated that with the exception of one inherited core for which an older inherited age of 1391 ± 11 Ma was obtained, the 206Pb/238U ages of all other core and rim measurement points varied from 440 to 453 Ma, and the results were consistent within an error range. The Th/U value of all zircons was > 0.2, indicating that the zircons exhibited magmatic origin characteristics, which include HREE enrichment, LREE depletion, and distinct negative Eu anomalies (Hoskin and Schaltegger 2003). By combining these results with the petrographic characteristics, we inferred that the zircons underwent two growth stages or late-stage modifications after core formation. Therefore, we obtained a weighted mean age of 447 ± 7 Ma (MSWD = 0.089, n = 14) from 14 points in the zircon core, representing the formation age of the granite (Fig. 7).

Origin of rapakivi-textured feldspar

The petrographic features and mineral chemistry analysis results indicated that the minerals in the Niujiaoshan rapakivi-textured granite underwent at least two stages of secondary growth. In the early stages, the representative minerals comprised K-feldspar in the core of the porphyritic crystals, enclosed biotite (Bt1), amphibole (Amp1), plagioclase (Pl1), and quartz. In the late stages, the representative minerals comprised biotite (Bt2), amphibole (Amp2), K-feldspar, coarse-grained plagioclase (Pl2), quartz in the matrix, and oligoclase (Pl3) surrounding the rims of the K-feldspar porphyritic crystals. There are multiple opinions about the formation mechanism of the rapakivi-textured feldspar. However, three widely accepted explanations are (1) the mixing of mafic and granitic magmas (Hibbard 1981; Bussy 1990; O’Brien et al. 2019; Alves et al. 2021), (2) the formation of plagioclase along the mantle due to sub-isothermal decompression during the ascent of granitic magma (Nekvasil 1991; Eklund and Shebanov 1999), and (3) the formation of rapakivi texture through the subsolidus exsolution of plagioclase in the alkali feldspars (Dempster et al. 1994). Our results support the first explanation based on the following evidence: first, the mineralogical chemical analysis indicates that the compositions of biotite and amphibole fall within the region of crust–mantle magma mixing on the discrimination diagram (Fig. 11) (Ding 1988; **e and Zhang 1990; Yang et al. 2017). Second, the amphibole and biotite in the matrix are relatively enriched in Mg. The An component at the edge of the coarse plagioclase in the matrix exhibits reverse zonation. This suggests that the Ca and Mg components in the magma were relatively high at the time when the minerals in the matrix crystallized, and this corresponds to the crystallization process during the mixing of late basic magma (Wang and Wang 2009); however, this supposition would contradict the latter two viewpoints, as they propose a closed system through sub-isothermal decompression or subsolidus exsolution. Third, the εHf (t) values of the zircon cores range from − 2.04 to 3.63, with an average of − 0.36, and those of the zircon rims range from − 0.79 to 2.11, with an average of 0.47. These positive and negative εHf (t) values indicate the addition of mantle-derived material to the magma source region, which is consistent with the chemical characteristics of the mineral (Wu et al. 2007). Fourth, the presence of a small number of dark inclusions within the granite suggests that magma mixing occurred. The presence of needle apatite in these dark inclusions also serves as evidence for magma mixing (Fig. 3f; Wang et al. 2003; Lux et al. 2007). Lastly, contemporaneous intrusions of mafic dikes, such as the Sifangtai pluton, were derived from a mantle source (Chen et al. 2006; Liu et al. 2008). The granites also exhibit higher Mg# (55.7–56.1) values than melts that were solely formed by the melting of the lower crust (Rapp and Watson 1995). These characteristics indicate the presence of mantle components in the source region of the granites.

Fig. 11
figure 11

(a) Diagram of MgO versus FeOT/(FeOT+MgO) for biotite (from Wang and Wang 2009); (b) Diagram of Al2O3 versus TiO2 for amphibole (from Jiang and An 1984)

Full size image

In general, the mixing of mafic and acidic magmas alters crystallization conditions and composition, thereby causing instability of the previously crystallized alkali feldspar. This leads to the resorption and formation of irregular boundaries (Fig. 4c, d; Wang and Wang 2009). Under these conditions of disequilibrium, plagioclase tends to nucleate on the surface of preexisting K-feldspar crystals and rapidly crystallizes, forming plagioclase mantles instead of new crystals (Hibbard 1981; O’Brien et al. 2019). During rapid cooling, alkali feldspar reacts with calcium from the mantle-derived melt to generate an oligoclase mantle and quartz blebs in the mantle (Fig. 4d); the alkali feldspar then undergoes exsolution, causing perthitic intergrowth (Fig. 4a; Parsons et al. 2015). Therefore, the plagioclase mantles were formed through rapid condensation and crystallization when mafic magma originating from the mantle was introduced into granitic magma.

Comparison between rapakivi and rapakivi-textured granite

Typical rapakivi granites, such as the Finnish pluton (Sederholm 1891), Korosten pluton, Salmi pluton (Rämö and Haapala 1995), Miyun rapakivi granite (Yang et al. 2005), and Kodar Complex (Larin et al. 2021) as well as orogenic rapakivi-textured granites such as the Jangart Complex (Konopelko et al. 2007; Solomovich and Trifonov 2014), Itu batholith (Pereira et al. 2023), Deer Isle pluton (Lux et al. 2007; O’Brien et al. 2019), Shahewan pluton (Wang et al. 2003; Zhou et al. 2008), and Tataleng pluton (Lu et al. 2007a; Hu et al. 2008), have a common rapakivi texture characterized by ovoid alkali feldspar porphyry mantled by oligoclase. Despite their remarkable petrological similarity, the mafic mineral chemistry (Fig. 6) and geochemistry (Figs. 9 and 10) of the Phanerozoic orogenic rapakivi-textured granites typically differ.

The Early Paleozoic Niujiaoshan rapakivi-textured granite is an I-type granite whereas typical rapakivi granites differ (Figs. 9 and 10), suggesting that they were emplaced in different settings. The I-type granites in the QOB region exhibit post-orogenic features (Wang et al. 2003, 2005; Zhou et al. 2008). Additionally, there are post-collisional rapakivi-textured granites with an A2-type affinity (Solomovich and Trifonov 2014; Pereira et al. 2023). A2-type granites are generally compatible with post-orogenic tectonic settings and crustal orogeny (Eby 1990). Their trace elements also differ: typical rapakivi granite exhibits evident negative Eu anomalies, whereas rapakivi-textured granites have weak negative Eu anomalies. Eu anomalies in melt and anatectic systems are controlled by the bulk rock composition, the phase assemblage (including plagioclase, alkali feldspar, and other minerals), and oxygen fugacity (Trail et al. 2012; Yakymchuk et al. 2023). The weak negative Eu anomalies observed in rapakivi-textured granites may be due to higher pressure (Deng et al. 2004) and oxygen fugacity (Trail et al. 2012; Yakymchuk et al. 2023). Biotite and amphibole in rapakivi-textured granites are rich in Mg (Fig. 6) rather than Fe, which distinguishes them from most typical rapakivi granites. These differences are possibly attributed to the differences in the tectonic setting in which they were formed.

Rapakivi-textured granite is found within collisional orogens that also contain UHP metamorphic rocks (Fig. 1); however, the granite was formed after the UHP metamorphism, indicating that it appeared after the orogen collision. However, UHP orogens are more common than rapakivi-textured granite (Liou et al. 2009; Solomovich and Trifonov 2014), indicating that rapakivi-textured granite was formed under more restricted conditions. Indeed, rapakivi and rapakivi-textured granites are generally emplaced in large-scale extensional settings, such as an anorogenic or post-orogenic environment.

Geological significance: implications of crust–mantle interactions

Zircon U–Pb dating revealed that the Niujiaoshan rapakivi-textured granite was formed at approximately 447 ± 7 Ma. It was contemporaneous with the development of second-stage granites (~ 450 Ma) in the QOB (Wang et al. 2009, 2015; Zhang et al. 2013). Comparisons of their geochemical characteristics suggest that the Niujiaoshan rapakivi-textured granite is similar to that of ~ 450 Ma granites in the QOB, suggesting a comparable tectonic setting (Figs. 9 and 10). There are two different opinions redarding the formation of this granite stage. One opinion associates it with the continental deep subduction of the NQB, and this is supported by the proximity of the intrusion age and the retrograde metamorphic age of high pressure to UHP metamorphic rocks in the NQB (Zhang et al. 2013; Liu et al. 2016). Considering the field occurrence and geochemical characteristics of the granite rocks, the NQB metamorphic rocks are believed to be the magmatic response of the subducted slab rollback process (Zhang et al. 2013; Chen et al. 2019a). In contrast, other researchers believe that these granites are island arc magmas formed during the northward subduction process of the Shangdan Ocean, as they are positioned on the northern side of the Shangdan suture zone and lack clear metamorphism (Wu and Zheng 2013; Dong and Santosh 2016).

The Al2O3 and K2O + Na2O contents of the Niujiaoshan rapakivi-textured granite are high, with LREE and LILE enrichment and HFSE depletion in elements such as Nb, Ta, and Ti. In the structural environment discrimination diagram (Fig. 12), the rapakivi-textured granite falls within the collision uplift zone, indicating that it was formed during the post-collisional decompression stage (Küster and Harms 1998). Zircons found in the granite exhibit two-stage model ages ranging from 1.20 to 1.55 Ga, which aligns with the formation age of the Neoproterozoic metasediments in the Qinling Group (Gong 2017; Shi et al. 2018). These findings suggest that the granite may have originated from the melting of Neoproterozoic metasediments in the Qinling Group. Additionally, the granites exhibit high Mg# values and are located in the crust–mantle origin field on the biotite and amphibole plot (Fig. 11). This suggests that the sources of this magmatism include both crustal materials from early subduction and collision stages, as well as a considerable amount of upwelling mantle materials (Barbarin 1999).

Fig. 12
figure 12

Tectonic discriminant diagrams for granite (from Pearce et al. (1984); Batchelor and Bowden (1985)). Data sources align with those in Fig. 8

Full size image

The detailed petrographic and mineralogical analyses of this study indicate that the rapakivi structure of feldspar reflects the characteristics of crust–mantle magma mixing. Additionally, contemporaneous mafic dikes have developed within the QOB (Chen et al. 2006; Liu et al. 2008), representing extensional and depressurization events. Based on the geochemical characteristics, the granite during this period more likely formed in a decompressional environment associated with slab rollback and uplift, rather than a compressional environment associated with the sustained subduction of the Shangdan Ocean. Large amounts of HP/UHP metamorphic rocks have been found in the Qinling Complex. These rocks underwent peak metamorphism at ~ 500 Ma followed by heating and decompression at ~ 450 Ma, resulting in the retrograde metamorphism of the granulite facies (Liu et al. 2013b; Wang et al. 2013, 2014; Wu and Zheng 2013; Gong et al. 2016; Liao et al. 2016). This corresponds to an environment of deep subduction and slab detachment associated with rollback. In the post-collision stage, when the continental lithosphere converged from compression to relaxation, the lower ocean crust of the subducting plate generated a downward drag force, whereas the continental crust exhibited upward buoyancy owing to the low overall density. The existence of these two opposite forces inevitably led to the break-off of the lower part of the subducted plate at an appropriate time and the retraction of the upper continental crustal plate (Boutelier and Chemenda 2008). The tectonic environment changed from compression to extension, which inevitably resulted in the upwelling of a large amount of mantle materials and the partial melting of a large range of crustal materials. Thus, a large amount of high-potassium calc-alkaline magmatism occurred (Zhang et al. 2013).

In localized areas, crust–mantle interactions occurred at a rapid pace, with mafic magma intruding into partially crystallized granite and causing rapid changes in the composition and temperature of the magma system. Under such conditions, plagioclase crystals tended to nucleate and crystallize on already crystallized K-feldspar surfaces, forming rapakivi-textured feldspar (Hibbard 1981; O’Brien et al. 2019). Therefore, we believe that the Niujiaoshan rapakivi-textured granites of ~ 450 Ma, as well as other widely distributed granites and a small number of mafic rocks in the Qinling Complex, were formed as a response to this tectonic transitional event.

Conclusions

(i) The Niujiaoshan rapakivi-textured granite was formed at 447 ± 7 Ma. Granites of the same age during the Paleozoic in the QOB were likely formed in a tectonic environment in which the Qinling Group UHP terrane underwent exhumation, transitioning from compression to extension. (ii) The rapakivi texture feldspars were formed by magma mixing during crust–mantle interactions. (iii) Typical rapakivi and orogenic rapakivi-textured granites are generally emplaced in large-scale extensional settings, such as an anorogenic or post-orogenic environment. (iv) A limitation of this study is the uncertainty associated with the model of magma mixing and how this might have influenced the formation of rapakivi texture. Further investigations on rapakivi-textured granite from orogens should be conducted to elucidate its formation mechanism, which will assist in better understanding orogen evolution and crust–mantle interactions.