The South China Sea (SCS), located to the south of the South China Block (SCB), is composed of the northern continental margin, a deep-water basin, and the southern continental margin (Yan et al., 2010). Currently, the SCS, interpreted as a plate-edge basin, is surrounded by four distinct tectonic zones, including a subduction zone in the east, a strike–slip fault zone in the west, a collisional thrust belt in the south, and a rifted passive margin in the north (Li et al., 2021). The deep-water basin of the SCS can be further divided into three sub-basins, including the East sub-basin, the Southwest sub-basin, and the North-west sub-basin (Figure 1). The extensively thinned continental crust varies from ∼6 km to ∼30 km thick around the deep-water basin (Hayes and Nissen, 2005). Unlike open oceans and other small back-arc basins in the West Pacific, the SCS deposited thick terrigenous sediments on the continental shelf without large abyssal fans (Huang and Wang, 2006). The northern continental margin of the SCS is characterized by the widely distributed (over 250 km) detachment faults, widespread high-velocity lower crust (HVLC) and magmatic intrusion below thinned crust, a large number of landward dipping fault systems, and the fast transition from continental margin to seafloor spreading (Sun et al., 2019).

Recent U-Pb age determinations for borehole-penetrated plutonic rocks imply a Late Jurassic amphibole plagiogneiss basement for the northern continental margin of the SCS, which was later intruded by Early Cretaceous plutons (Zhu et al., 2017). Previously sampled rocks indicate that the southern margin of the SCS, including the Nansha Block and the Reed Bank-Northeastern Palawan Block area, was once a part of the South China continent (Kudrass et al., 1986; Yan et al., 2010; Miao et al., 2021). Since the Mesozoic, the realm of the SCS has experienced the subduction of the paleo-Pacific (Taylor and Hayes, 1983). After a Cretaceous to Paleogene period of extension and rifting, which affected tectonic deformation in the upper crust (Savva et al., 2014) and the rheological properties of the lower crust (Ros et al., 2017) and resulted in the formation of numerous NE–SW-trending half-grabens (Tang et al., 2014), the continental fragments, including the Nansha, Reed Bank, Palawan, and Mindoro Blocks, were finally separated from the South China continent and drifted southward to their current position as the transition from plate subduction to seafloor spreading occurred during the Cenozoic (Kudrass et al., 1986).

The borehole NK-1 drilled on the Meiji Atoll of the southern SCS continental margin penetrated the volcanic basement that consists of volumes of dacitic, interbedded with basaltic and andesitic rocks from 1010 to 2020 mbsf (meters below sea floor). The studied andesitic rocks were interbedded with the dacites and basalts. The depth of the andesitic rocks collected from the borehole NK-1 is 1319.06–1446.09 mbsf, and the thickness of the andesitic rocks ranges from 0.05 m to 0.2 m. The andesitic rocks are dark gray in color with massive structures, commonly with porphyritic texture consisting of subhedral phenocrysts or microphenocrysts including plagioclases and pale green-brown amphiboles (Figure 2). These andesites also show calcite, sericite, chlorite, epidote, and iron oxides as secondary phases. The groundmass is characterized by randomly distributed acicular plagioclase laths and amphibole crystallites embedded in cryptocrystalline materials.

Zircons were separated from four andesite samples, using conventional hydroseparation and magnetic methods, followed by handpicking under a binocular microscope. The zircon grains were mounted in an epoxy resin disk, polished, and coated with gold. These grains were examined and imaged before analysis using a transmitted and reflected optical microscope and cathodoluminescence to determine crystal shapes and internal structures and to pick positions for spot analyses. Zircon U-Pb dating was undertaken by a CAMECA IMS-1280HR ion microprobe. The analytical procedures for zircon U-Pb geochronology were similar to those of Li X.-H. et al. (2009). Uncertainties of isotope ratios were at the 1σ level, and the calculated ages are reported for 2σ confidence. Data reduction was carried out using Isoplot 3.0 (Ludwig, 2000). The Qinghu zircon standard was used as an unknown sample to check the stability and accuracy of the measurement process. The operating condition was described by Li X. et al. (2009). Oxygen isotopes were examined using the CAMECA IMS-1280 ion microprobe, following the description of Li X.-H. et al. (2009) and Li et al. (2010). After U-Pb dating and O isotopic analysis of zircons, in situ zircon Lu-Hf isotope analyses were performed using a Neptune MC-ICP-MS equipped with a 193-nm laser ablation system. The detailed analytical procedure and correction for interference are similar to those of Wu et al. (2006). The obtained weighted mean ¹⁷⁶Hf/¹⁷⁷Hf ratios of the zircon standard Mud Tank and SA01 agree with the recommended values (Woodhead and Hergt, 2007).

The altered or weathered portion of the andesites was excised, and the remaining parts of the blocks were cut into small pieces. Half of these pieces were polished into thin sections, and the remaining rock chips were crushed and then powdered to a grain size of 200 mesh in an agate mill for bulk-rock analyses. Bulk-rock major element compositions were determined by the AXIOS Minerals X-ray fluorescence (XRF) spectrometry on fused lithium-tetraborate glass beads. The analytical precision for major elements ranges from 1% to 3%. Bulk-rock trace elements (including rare earth elements) were analyzed by an Agilent 7500a inductively coupled plasma mass spectrometry (ICP-MS), following the description of Yang J. H. et al. (2012). The powdered samples (approximately 50 mg) were dissolved in high-pressure Teflon vessels using a highly purified HF + HNO₃ mixture for 5 days at 150 °C. Rhodium (Rh) was added as an internal standard to monitor signal drift during analysis. Standard materials BIR-1a, RGM-2, GSR-3, and BHVO-2, which were analyzed along with samples, were used for analytical quality control, indicating the analytical errors were better than 5%.

Major element compositions of amphibole from andesites using a JEOL JXA-8100 electron probe microanalyzer (EPMA) operated in wavelength-dispersive spectrometer (WDS) mode. Operating conditions included an accelerating voltage of 15 kV, a beam current of 20 nA, a beam size of 5 μm, and counting times of 20–60 s. Energy-dispersive spectrometry (EDS) in combination with backscattered electron scanning (BSE) imaging was used to select spot locations for quantitative in situ analysis. The amphiboles were analyzed for Na, Cr, Mg, Si, Mn, K, Al, Ti, Fe, Ca, F, Cl, and Ni. The standard minerals and synthetic oxides used in these analyses for calibration included albite (Na), chromium oxide (Cr), olivine (Mg), bustamite (Mn), K-feldspar (K), jadeite (Al), rutile (Ti), magnetite (Fe), riebeckite (F), tugtupite (Cl), and diopside (Ca and Si). Atomic number-absorption-fluorescence (ZAF) correction routines were used to process the raw analytical results. The accuracy was better than 1%–5% for major elements (≥1wt%), and that of the minor element (<1 wt%) analyses was better than 10%. The precision was better than 1 wt% for element oxides. The analytical results of major and minor elements for amphibole are presented in Supplementary Table S6.

In situ trace element contents of amphibole and zircon were determined by laser ablation-inductively coupled plasma mass spectrometry (LA-ICP-MS) employing an Agilent 7500a Q-ICP-MS instrument (Agilent Technologies, United States) coupled to a 193 nm ArF excimer laser system (Geolas HD, Lambda Physik, Göttingen, Germany). The approach is similar to those outlined in Xie et al. (2008) and Wu et al. (2018), with isotopes measured using a peak-hopping mode with a laser beam diameter of ca. 60 μm and a 6 Hz repetition rate. The laser energy density is ∼4.0 J/cm². Helium gas was employed as the carrier gas to improve the transportation efficiency of ablated aerosols, and nitrogen gas was used as an additional gas to enhance sensitivity. Before analyzing the amphibole and zircon, the LA-ICP-MS system was optimized using NIST SRM 610 reference glass as the calibration material to achieve the maximum signal intensity and low oxide rates. Glass reference materials ARM-1 (Wu et al., 2019) and BCR-2G were also analyzed for data quality control. Multiple or single LA analyses for Amp were undertaken in conjunction with single-electron microprobe analyses during the experiment. Silicium (²⁹Si) content values previously obtained by electron probe microanalysis were used as an internal standard to quantify elemental concentrations of amphibole and zircon. Off-line selection and integration of background and analyte signals, time-drift corrections, and quantitative calibrations were performed using the GLITTER program (Griffin et al., 2008) for data reduction. The measured values of reference materials were in agreement with the recommended values. For most trace elements (>0.10 μg/g), the accuracy is better than ±10% with analytical precision (1 RSD) of ±10%. The analytical results of trace element compositions for amphibole and zircon are presented in Supplementary Table S5, S7, respectively.

Measurements of the Sr-Nd-Hf isotopic ratios were performed using a Neptune Plus multi-collector inductively coupled plasma mass spectrometer following the description of Yang et al. (2010). The powdered samples (about 100 mg) were dissolved in high-pressure Teflon vessels using a highly purified, concentrated HF + HNO₃+HClO₄ (2 mL+1 mL + 0.2 mL) mixture for 7 days at 150 °C. The solutions were then evaporated to fume HClO₄. A 1-mL aliquot of 6 M HCl was added to the Teflon vessel and dried, and then this procedure was repeated once. After it cooled, 5 mL of 3 M HCl was added to the residue. The Teflon vessel was placed on a hot plate at 100 °C for 12 h, and then the sample solution was loaded onto pre-conditioned 2 mL Ln Spec resin.

Light and middle rare earth elements (LMREE) were eluted first, and then the Lu (+Yb) fraction was eluted. After separating the Ti from Hf, Hf (+Zr) was extracted from the column with HF. Then, the solution of Hf (+Zr) was evaporated to dryness, and 1 mL of 2% HNO3 was added to the residue. The solution was ready for Hf analysis. The solution of LMREE was dried and re-dissolved in 1.5 mL of 2.5 M HCl. After loading the solution into a quartz column packed with AG50W-X12, Rb and Sr were eluted. Then, the LMREE fraction was stripped out with 6 mL of 6 M HCl. After the LMREE collections were dried and re-dissolved in 0.5 mL of 0.25 M HCl, the solutions were loaded onto Ln Spec resin. Then, the Nd and Sm were eluted by an appropriate amount of acid.

The measured ⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd and ¹⁷⁶Hf/¹⁷⁷Hf ratios were corrected for mass fractionation using ⁸⁷Sr/⁸⁶Sr = 0.1194, ¹⁴³Nd/¹⁴⁴Nd = 0.7219, and ¹⁷⁶Hf/¹⁷⁷Hf = 0.7325, respectively. During the analytical sessions, the measured values for the NBS987 Sr standard, the JNdi-1 Nd standard, and the Alfa Hf standard were 0.710277 ± 24 (2σ, n = 10) for ⁸⁷Sr/⁸⁶Sr, 0.512107 ± 10 (2σ, n = 10) for ¹⁴³Nd/¹⁴⁴Nd, and 0.282192 ± 14 (2σ, n = 12) for ¹⁷⁶Hf/¹⁷⁷Hf, respectively. BHVO-2 and BIR-1 were measured along with the samples and gave rise to the ⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, and ¹⁷⁶Hf/¹⁷⁷Hf ratios of 0.703481 ± 12 and 0.703101 ± 11, 0.512987 ± 12 and 0.513124 ± 13, and 0.283087 ± 10 and 0.283247 ± 20, respectively, which are in accordance with the reference values (GeoREM, http://georem.mpch-mainz.gwdg.de/). The total procedure blanks for the Sr isotope, Nd isotope, and Hf isotope were 240 pg, 70 pg, and 50 pg, respectively.

Measurement of the Pb isotopic ratio was performed using a Thermofisher Triton Plus multi-collector thermal ionization mass spectrometer following the description of Li et al. (2015). The powdered samples (approximately 100–120 mg) were completely dissolved in a Savillex Teflon screw-cap beaker using a highly purified concentrate HF + HNO₃ mixture for 7 days at 160 °C. The Pb fraction with high purity was separated from the rock matrix using a HBr-HCl elution procedure on Teflon columns containing ∼0.15 mL anion resin (AG1-X8, 100–200 mesh). The whole chemical procedural blank of Pb was approximately 200 pg. The Pb fraction with high purity was re-dissolved using a mixture solution of silica gel and H₃PO₄ and loaded onto a single-Re filament. International standard NIST 981 was used to monitor instrument stability during the period of data collection and correct mass fractionation. Measured Pb isotopic ratios were corrected for instrumental mass fractionation of 1.2‰ per atomic mass unit by reference to repeated analyses of the NIST 981 Pb standard.

The elected zircons extracted from four andesite samples are mostly euhedral to subhedral in morphology, transparent, and light brown in color. Cathodoluminescence (CL) images show numerous blurry concentric zoning with low to variable luminescence (Figure 3), with variable Th (5.0–784 ppm), U (22.0–930 ppm), and higher Th/U ratios (0.24–1.95) than metamorphic-derived zircons (<0.1), revealing a magmatic origin. Some zircons from the sample NK-1-V2-91 show inherited cores with growth rims, implying the occurrence of inherited zircon crystals. The size of the zircons is from 40 μm to 100 μm in width, with the length to width ratios from 3:2 to 3:1. Four of 12 zircon grains from the sample NK-1-V2-91 yield concordant ²⁰⁶Pb/²³⁸U ages of 221.7 ± 3.6 Ma (MSWD = 0.12, n = 4) with a weighted mean age of 221.5 ± 7.9 Ma (MSWD = 0.39, n = 4), which represent the crystallization age of andesitic magma. The remaining eight ancient zircons, with ²⁰⁷Pb/²⁰⁶Pb ages between 1898 and 2467 Ma, represent the crystallization ages of inherited zircons entrained by the andesitic magma.

All 24 U-Pb analyses for NK-1-V2-139 yield ²⁰⁶Pb/²³⁸U ages between 207 Ma and 227 Ma with a weighted mean age of 217.7 ± 2.1 Ma (MSWD = 1.8, n = 24), representing the crystallization age of NK-1-V2-91. Zircons from NK-1-V2-166 yield a lower intercepted age of 222.8 ± 2.4 Ma, which is identical to their weighted average ²⁰⁷Pb-corrected ²⁰⁶Pb/²³⁸U ages within uncertainties, that is, 223.1 ± 2.0 Ma. Zircons from NK-1-V2-171 yield concordant ²⁰⁶Pb/²³⁸U ages of 225.6 ± 3.0 Ma (MSWD = 0.38, n = 12) with a weighted mean age of 225.7 ± 3.8 Ma (MSWD = 2.8, n = 12), interpreted as the crystallization age of the sample (Figure 4).

The whole-rock geochemistry results for four andesites from the Meiji Atoll are listed in Supplementary Table S2. In the plots of SiO₂ versus Nb/Y diagram (Winchester and Floyd, 1977) (Figure 5a), the andesitic rocks plot predominantly within the field of andesite. In the plots of Th versus Co diagram (Hastie et al., 2007) (Figure 5b), the andesites predominantly fall in the fields of high-K calc-alkaline and/or shoshonite series. After normalizing all major element compositions to 100% anhydrous, major element data on the andesites show limited variation in SiO₂ content (57.8–61.7 wt%), Al₂O₃ content (14.8–18.7 wt%), CaO content (3.61–7.48 wt%), MgO content (1.37–2.91 wt%), TFe₂O₃ content (7.40–9.48 wt%), TiO₂ contents (0.71–1.68 wt%), and LOI (loss on ignition) content (0.80–1.89 wt%). The content of total alkali (Na₂O+ K₂O) ranges from 3.82 wt% to 9.83 wt%, and the Mg^# also shows a wide variation with 28.57–43.19 (Figure 6a). Furthermore, most of the major elements for andesites show relatively linear trends on Harker diagrams (Figures 6b-d), such as Al₂O₃ (Figure 6b).

Samples have ΣREE concentrations ranging from 74 ppm to 386 ppm, with Ni ranging from 2.32 ppm to 9.10 ppm, Cr ranging from 18.4 ppm to 64.7 ppm, and Co ranging from 9.11 ppm to 23.8 ppm for the transitional trace element compositions. Four andesites show subparallel trace element patterns resembling those of average continental arc andesites on the chondrite-normalized rare earth element (REE) patterns (Figure 7a). On the primitive mantle-normalized trace element diagrams (Figure 7b), they are characterized by variable enrichment in light rare earth elements (LREEs) and large ion lithophile elements (LILEs, e.g., Rb, K, U, and Pb) and depletion of high-field strength elements (HFSEs, particularly Nb, Ta, and Ti) compared to normal mid-ocean ridge basalts (N-MORB), with (La/Yb)_N ratios of 4.9–8.6. Weak negative Eu anomalies are observed in the andesitic samples (Eu/Eu* = 0.6–0.9).

Whole-rock Sr-Nd-Hf-Pb isotopic data of the andesites are listed in Supplementary Table S2. The initial ⁸⁷Sr/⁸⁶Sr, ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb, ²⁰⁸Pb/²⁰⁴Pb ratios and ε_Nd(t) and ε_Hf(t) values were recalculated based on the corresponding zircon U-Pb ages. The andesites have relatively high initial ⁸⁷Sr/⁸⁶Sr ratios (0.7083–0.7113) and display relatively limited variation in ε_Nd(t) values of −6.07 to −4.91 and ε_Hf(t) values of −7.70 to −2.16, with old Nd (1477–1675 Ma) and Hf (1172–1918 Ma) model ages. On the plot of ε_Hf(t) versus ε_Nd(t) isotopic diagram (Figure 8a), these samples define a subparallel linear trend resembling the Terrestrial Array (ε_Hf(t) = 1.33 × ε_Nd(t) + 3.19 (Vervoort et al., 2011)), without significant Nd-Hf isotopic decoupling. The initial ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb, and ²⁰⁸Pb/²⁰⁴Pb ratios of the calc-alkaline andesites are generally higher than those of interbedded and intercalated basaltic rocks but lower than the volumes of dacites drilled from Meiji Atoll. All samples define the positive correlations in (²⁰⁷Pb/²⁰⁴Pb)_i vs. (²⁰⁶Pb/²⁰⁴Pb)_i and (²⁰⁸Pb/²⁰⁴Pb)_i vs. (²⁰⁶Pb/²⁰⁴Pb)_i diagrams (Figures 8c,d), subparallel to the Northern Hemisphere Reference Line (NHRL). In the plot of (²⁰⁷Pb/²⁰⁴Pb)_i vs. (²⁰⁶Pb/²⁰⁴Pb)_i diagram (Figure 8c), the andesites plot significantly above the NHRL and to the right side of the 4.55 Ga geochron line. They also show similar (²⁰⁶Pb/²⁰⁴Pb)_i, (²⁰⁷Pb/²⁰⁴Pb)_i, and (²⁰⁸Pb/²⁰⁴Pb)_i ratios to the enriched mantle 2 (EM−Ⅱ) components, Middle to Late Jurassic basaltic rocks, and crust-derived suites in the Cathaysia Block.

The amphibole phenocrysts within the host andesites have been selected for in situ microprobe and laser ablation studies. Major and trace elements of amphiboles are listed in Supplementary Tables S6, S7, respectively. According to the nomenclature of Leake et al. (1997), the amphiboles from the Meiji andesites are mainly magnesio-hornblende, pargasite (^VIAl > Fe³⁺), and magnesio-hastingsite (^VIAl < Fe³⁺), with slight variations to actinolite, ferro-hornblende, and ferro-edenite (Figures 9a,b). These amphiboles are characterized by highly variable Mg^# values ranging from 0.34 to 0.78 for NK-1-V2-91, 0.48 to 0.83 for NK-1-V2-139, 0.65 to 0.96 for NK-1-V2-166, and 0.59 to 0.84 for NK-1-V2-171. The silica content ranges from 5.89 a.p.f.u to 7.55 a.p.f.u for NK-1-V2-91, 5.87 a.p.f.u to 7.56 a.p.f.u for NK-1-V2-139, 5.79 a.p.f.u to 7.52 a.p.f.u for NK-1-V2-166, and 5.86 a.p.f.u to 7.47 a.p.f.u for NK-1-V2-171. The alkali content (Na + K) is up to 0.70 a.p.f.u, and the Ti content is as high as 0.53 a.p.f.u. They show large variations in Cr (4.28–1290 ppm) and Ni (0.52–20.3 ppm) concentrations. The chondrite-normalized pattern of REE (Figures 7c-j) in amphibole phenocrysts is slightly convex upward with variable enrichment in middle-REEs (MREEs) relative to LREEs and HREEs (La_N/Sm_N = 0.16–0.83 and Gd_N/Yb_N = 0.85–2.04). Slightly to significantly negative Eu anomalies (Eu/Eu* = 0.45–1.01) are also observed in these amphibole grains. The incompatible trace element diagram reveals variable depletion in Th, U, and P and enrichment in Rb, Ba, and K relative to the neighboring immobile elements (Figure 7).

The trace element compositions of dating zircon grains were applied to monitor the geochemical signatures of their host magmas and elucidate the tectonic setting of drilled andesites (Supplementary Table S5). Chondrite-normalized rare earth element (REE) patterns for most of the analyzed zircon grains display significant positive Ce anomalies and negative Eu anomalies (Figure 3e), consistent with their magmatic origin. They are also characterized by the enrichment of HREEs and depletion of LREEs. The Yb and Hf contents of the analyzed zircons show relatively wide ranges of 13.78–851 ppm and 164–13,228 ppm, respectively. Nb and Y concentrations are also variable, with ranges of 0.88–9.0 ppm and 99.4–2033 ppm, respectively.

The results of in situ zircon Lu-Hf and O isotopic compositions from four andesite samples are listed in Supplementary Table S3. For the sample NK-1-V2-91, the initial ¹⁷⁶Hf/¹⁷⁷Hf ratios of Late Triassic zircons vary from 0.282555 to 0.282578, and the calculated ɛ_Hf(t) values range from −2.93 to −1.99 (Figure 8e), with a T_DM1(Hf) from 954 Ma to 979 Ma and T_DM2(Hf) from 1377 Ma to 1433 Ma. Meanwhile, the initial ¹⁷⁶Hf/¹⁷⁷Hf ratios of the remaining eight inherited zircons from NK-1-V2-91 vary from 0.281308 to 0.281446, and the calculated ɛ_Hf(t) values range from −6.26 to 2.85, with a T_DM1(Hf) from 2499 Ma to 2699 Ma and T_DM2(Hf) from 2641 Ma to 3060 Ma. For the samples NK-1-V2-139, 166, and 171, the initial ¹⁷⁶Hf/¹⁷⁷Hf ratios vary from 0.282197 to 0.282680, and the calculated ɛ_Hf(t) values range from −5.67 to 1.31, with a T_DM1(Hf) from 810 Ma to 1067 Ma and T_DM2(Hf) from 1157 Ma to 1588 Ma. Furthermore, these andesites also show large variations in zircon δ¹⁸O values from 4.0‰ to 9.0‰ (Supplementary Table S3; Figure 8b).

Previous geochronological studies have focused mainly on the volumes of dacites that are closely intercalated with andesites. In this study, a limited range of zircon U-Pb ages, 217.9–219.1 Ma, from dacites was believed to represent the eruption age of the volcanism (Miao et al., 2021; Wei et al., 2021). Nevertheless, the SIMS zircon U-Pb dating results from this study provide new constraints on the timing of intermediate volcanism erupted in Meiji Atoll, SCS. The older concordant ²⁰⁶Pb/²³⁸U zircon ages (225.6 ± 3.0 Ma), ²⁰⁷Pb-corrected ²⁰⁶Pb/²³⁸U ages (223.1 ± 2.0 Ma), similar weighted mean ²⁰⁶Pb/²³⁸U ages (217.7 ± 2.1 Ma), and concordant ²⁰⁶Pb/²³⁸U ages (221.7 ± 3.6 Ma) for the formation age of andesites record a protracted volcanic history that is slightly longer than previously thought. Therefore, we consider that the volcanic rocks in the Meiji Atoll were erupted during the Late Triassic. Furthermore, the ancient inherited zircons (1898–2509 Ma) carried by the andesitic magma (sample NK-1-V2-91) shed new light on the Precambrain continental basement component within blocks in the SCS.

Thermobarometers based on Amp–melt compositions can be used to estimate the crystallization temperature and pressure of Amp in arc magmatic systems. The whole-rock compositions of andesite were assumed to be in equilibrium with the core of the Amp phenocrysts. Core compositions of Amp were used as tests for equilibrium by comparing the host-rock SiO₂ content with the result calculated for silicate liquid in equilibrium with Amp (Putirka, 2016). The Fe-Mg exchange coefficients between Amp and melt [K_d(Fe-Mg)^Amp-liq] were employed as an additional test for the Amp–liquid pairs.

For the application of thermobarometry, we consider as suitable only core compositions from Amp crystals whose calculated melt SiO₂ differs by no more than 4.0 wt% from the SiO₂ content of the host rock (Maro et al., 2020) and the K_d(Fe-Mg)^Amp-liq values in the range 0.28 ± 0.11 (Putirka, 2016). In the Meiji andesites, most of the Amp is in equilibrium with the respective host rocks. There are also cases where the predicted melt SiO₂ for Amp is higher or lower by 4–16.8 wt% than the host rock composition and close to that of the basaltic and dacitic lavas (Figure 12a). This may imply a probable interaction between the basaltic, dacitic, and andesitic magma during the major magma recharge event of compositionally different magma. A minimum value of 3.5 wt% for melt H₂O concentration was assumed based on the stability of magnesio-hastingsite during the P-T calculations (Ridolfi et al., 2010).

We chose the Amp-melt thermometer [Equation 5 of Putirka (2016)] to estimate the crystallization temperature of the Amp phenocrysts in this study. All suitable Amp grains from different samples yield relatively low and consistent crystallization temperatures in the range from 898 °C to 977 °C with an uncertainty of ±30 °C (Putirka, 2016). In view of the dependence of pressure on water concentrations in the melt and the poor sensitivity of Al concentrations to pressure change (Putirka, 2016; Maro et al., 2020), the pressure estimates based on amphibole equilibria should be treated with caution. Here, we use multiple barometric calibrations to place constraints on the formation of the Amp phenocrysts. We compared the results obtained by using Equation 7a of Putirka (2016), Equations 1b–d of Ridolfi and Renzulli (2012), and the formula of Ridolfi et al. (2010). The calibration of Ridolfi et al. (2010) yields pressures that correlate well with those of Ridolfi and Renzulli (2012) (Figures 12g,h), even though both the numerical agreement and the correlation of the values are worse between Equation 7a of Putirka (2016), the formula of Ridolfi et al. (2010), and Equations 1b–d of Ridolfi and Renzulli (2012) (Figures 12c–f). We consider that the Ridolfi et al. (2010) and Ridolfi and Renzulli (2012) calibration itself does not incorporate melt H₂O concentration as an explicit parameter in their calibration equation, resulting in systematic discrepancies with the H₂O-dependent thermobarometric model of Putirka (2016) in our study. Although the Amp phenocrysts in the Meiji andesites exhibit discrepancies among different barometers (Supplementary Table S6), we emphasize that these data collectively denote relatively low crystallization pressures (2–6 kbar) within the mid-upper crust levels.

The values of oxygen fugacity (fO₂) estimates following Ridolfi et al. (2010) indicate relatively oxidized magma (0.77–2.47 log units above the nickel–nickel oxide (NNO) buffer for Amp in the Meiji andesites; Figure 12b). The positive correlation between the crystallization pressure and temperature (Figures 13a,b) using different Amp–melt equilibrium geothermobarometry, integrated with relatively high fO₂, may imply a continuous and cryptic Amp fractionation during the evolution of arc magmas despite the rare occurrence of Amp as a major mineral phase in erupted volcanic rocks (Davidson et al., 2007).

The identification of sources for the observed xenocrystic/inherited zircons from sample NK-1-V2-91 plays a key role in interpreting their age and exploring their geological implications. In general, various forming mechanism and sources were proposed to explain the occurrence of these xenocrystic and inherited zircons: 1) xenocrystic zircons captured by assimilation and contamination of sedimentary or igneous wall rocks during magma ascent and emplacement (Wang et al., 2020); 2) inherited zircons typically derived from intermediate and felsic host igneous rocks with sedimentary or igneous origins (e.g., S-type or I-type granites); 3) inherited zircons typically derived from SCLM modified by subducted oceanic crust-derived siliceous melts or aqueous fluids (Katayama et al., 2003; Jiang et al., 2020). Nine igneous zircons with high Th/U ratios (0.24–1.95) display ²⁰⁷Pb/²⁰⁶Pb ages much older than the magma crystallization age (218–226 Ma), indicating that they are not formed in the andesitic magma itself. Most of these old zircons are subhedral or euhedral with prismatic morphologies and fine in size (Figure 4), precluding the possibility of long-distance transport (if it happened). They also exhibit fine rhythmic zoning, indicative of zircons crystallized from intermediate-acid magmas; hence, they do not originate from a metasomatized SCLM.

The andesitic arc magma entraining these zircons was hydrous as indicated by a large amount of Amp phenocrysts, which is also not consistent with the preservation of residual zircons from mantle sources (Harrison and Watson, 1983). Exposed Archean-Proterozoic basement rocks and mantle xenoliths carrying the underlying Precambrian lithospheric relics in eastern South China require a significant amount of ancient continental crust beneath the SCB (Wang et al., 2020), which is consistent with the age patterns of the Precambrian xenocrystic/inherited zircons from SCB (Figure 8f). Therefore, the old zircon xenocrysts observed in Meiji andesites should be originated from assimilation or contamination of wall rocks during magma ascent and emplacement, as demonstrated by the result of modeling calculations for zircon δ¹⁸O and zircon ε_Hf(t) values (Figure 8f) that the Hf-O isotopic compositions of most Meiji andesitic zircons can be reproduced by mixing between a depleted mantle-derived melt (ε_Hf = + 12, δ¹⁸O = 5.3‰) and <10% crustal components (ε_Hf = - 12, δ¹⁸O = 10‰). They are considered advantageous for revealing the composition of the assimilated crust beneath the Nansha Block.

The Nd model ages (1776–928 Ma) for the granitic rocks from two dredge stations in the Nansha Block have been interpreted to represent the existence of a Mid-Proterozoic crystalline basement, even though the corresponding Precambrian rocks are rarely exposed in the blocks dispersed in the SCS (Yan et al., 2010). The Meiji dacites and andesites also display similar Nd model ages (1390–1537 Ma and 1477–1675 Ma) to those of Nansha granitic rocks, suggesting that the basement of the Nansha Block may preserve a residual amount of ancient crustal materials. A significant inherited zircons with ages of 1898–2509 Ma generally show negative ε_Hf(t) (−6.3 to −1.6) except for two, 2.8 and 8.6, and a two-stage Hf model with a concentration of ages around 2623–3059 Ma, which verifies that there may have been initial crust in the Paleoproterozoic in the source area of the Nansha Block igneous rocks. These features, in combination with the extensive evidence for Precambrian basement rocks, occur in samples from other blocks of SCS that are believed to share an evolutionary history with the Nansha Block (Holloway, 1982; Kudrass et al., 1986; Hall et al., 2008; Yan et al., 2010). The identification of the oldest Cathaysia crystalline basement (the Badu Complex), including the high-grade metasedimentary rocks with a 2.5 Ga protolith (Yu et al., 2012) or a 1.9–1.8 Ga protolith (Zhao et al., 2015), and the associated 1.9–1.7 Ga granitoids and metabasic rocks (Liu et al., 2014), provides additional and compelling evidence for the existence of Paleoproterozoic crystalline basement within the Nansha Block similar to the Cathaysia Block, in view of the overlapping of the inherited zircons’ U-Pb age between the Meiji andesites in this study and the igneous rocks from the eastern SCB (Figure 8f).

It is well known that calc-alkaline continental arc andesites distributed at convergent margins possess a similar composition in both major and trace elements to continental crust on average (Hans Wedepohl, 1995; Rudnick and Fountain, 1995). Hence, andesitic magmatism above the oceanic subduction zone is considered to have played a significant role in the generation of continental crust along convergent plate boundaries (Kelemen, 1995; Taylor and McLennan, 1995; Grove et al., 2012; Kelemen and Behn, 2016). Andean andesites in the western margin of South America were widely accepted as the typical products of long-lived oceanic–continental convergent margins commonly named “Andean-type margins” (Hildreth and Moorbath, 1988; Davidson et al., 1990; Stern, 2004). Similarly, active continental margin and subduction-related geodynamic settings have also been inferred for the eastern and southeastern Asian margin since the late Paleozoic and early Mesozoic (Jahn et al., 1990; Zhou and Li, 2000; Li and Li, 2007; Li X.-H. et al., 2012; Metcalfe, 2013; Qiu et al., 2022). The widespread Mesozoic magmatism observed along the northern and southern (e.g., Meiji Atoll) SCS margins and the southeastern SCB demonstrates tectonic continuity between the SCS continental margin and SCB during the Mesozoic era (Qiu et al., 2014; Xu et al., 2017; Xu et al., 2022; Cui et al., 2021; Miao et al., 2021).

The andesitic arc volcanism at Meiji Atoll cannot be attributed to paleo-Tethys ocean subduction and closure, given that the Indo-China Block collided with the South China Block at 247–237 Ma, as constrained by geochronological data from the Ailaoshan–Songma suture zone rocks (Wang et al., 2013). The inferred collision (247–237 Ma) is unlikely to account for andesitic magmatism in this study as these andesites are significantly younger and were located east or southeast of the Ailaoshan–Songma suture prior to SCS basin opening. This spatial–temporal discrepancy precludes a paleo-Tethyan origin. The pre-rift paleo-position of Meiji Atoll along the SCB’s coastal margin aligns with the inferred Mesozoic arc belt (Li and Li, 2007; Wei et al., 2021). In addition, paleomagnetic results revealed a >60 ° clockwise rotation of the South China Block following its Early Triassic collision with the North China Block (Zhao and Coe, 1987). This implies the Nansha Block (including Meiji Atoll), located along the SCB’s southeastern coast pre-rifting, occupied an eastern position facing the paleo-Pacific during the Late Triassic.

The Late Triassic upper-section basalts and trachytes (980–1196 mbsf) from Meiji Atoll exhibit geochemical affinities with modern Mariana and Japan arc systems, providing compelling evidence for paleo-Pacific oceanic slab subduction beneath the South China margin as well (Xu et al., 2022). Consequently, the pre-rift southern SCS continental margin (including Nansha Block) was positioned along the southeastern SCB coast, dominantly controlled by paleo-Pacific plate subduction during early Mesozoic time. Once again, the drilled continental arc andesites in this study strengthen the conception of a now-vanished Andean-type convergent plate margin (Figure 14), genetically associated with subduction of the paleo-Pacific plate during the early Mesozoic and followed by growth and closure of marginal seas along the East Asia margin. Therefore, the identification of Meiji arc andesites penetrated by a borehole at the southern South China Sea continental margin shed new light on the lateral distribution and evolution of the arc system during the Late Triassic.

The buildup of this magmatic arc has been proposed to signals a major geodynamic switch from passive-margin to active-margin tectonics along the SE periphery of continental Asia, as indicated by the occurrence of 270–264 Ma ash-fall tuff layers in the Permian depositional record of SE China and widespread 267–262 Ma granitoid rocks on the island of Hainan (Li et al., 2006; Zhang et al., 2019). Additional, the remarkable change in regional palaeogeography from a carbonate platform in Carboniferous-earliest Permian to an elongated uplift with onset of clastic deposition along the southeastern coastal regions in mid-Early Permian is also coincident with the occurrence of a prominent zircon population in the Permian sandstone and siltstone units that crop out in E-SE China (Li X.-H. et al., 2012; Hu et al., 2015; Zhang et al., 2019), implying a significant change in the geodynamic evolution along the coast of SE China. The identification of continental sliver or ribbon continent with Cathaysia-type sources in the aspect of zircon xenocrysts beneath the Gagua Ridge of Philippine Sea and eastern Taiwan (Shao et al., 2015; Qian et al., 2021), together with previously reported borehole samples in the northern SCS margin and our newly dated ca. 218–226 Ma arc andesites in the southern SCS margin (Xu et al., 2017; Cui et al., 2021; Miao et al., 2021; Wei et al., 2021), highlighting the protracted Andean-type active-margin orogenic belt over space. This Andean-type active continental margin has been proposed to experience flat-slab subduction during the Triassic (ca. 250–190 Ma), as indicated by a predicted ∼1300 km-wide coastal magmatic gap (Li and Li, 2007; Li Z.-X. et al., 2012).

Following the end of the flat-slab subduction phase, a new continental arc was initiated after ca. 190 Ma, interpreted to represent the resumption of normal subduction along the coastal region (Li Z.-X. et al., 2012). In contrast, a certain number of approximately 200 Ma detrital zircons have been reported from metasedimentary borehole samples of the northern SCS basin recently, which is interpreted to reflect the close affinity of these sites to Hainan Island, which is outside of the proposed flat-slab subduction segment (Cui et al., 2021). In addition, the drilled volcanic rock series from Meiji Atoll of southern SCS was interpreted to reflect a spatial difference in distribution of arc magmatism along the SCB and possibly reveal the size of the subducted oceanic plateau (Wei et al., 2021). In combination with the Late Triassic andesites in this study, we cannot exclude the possibility that Meiji volcanic rocks, once distributed along the southern margin of the SCB, represent the earliest re-initiation of continental arc magmatism after a phase of flat-slab subduction.