Zircons from each sample were separated using conventional magnetic and heavy-liquid methods. After separation, zircon grains with crystal lengths >60–200 μm were linearly mounted with epoxy resin and then polished to expose the interior grains for U-Pb dating. Prior to U-Pb analysis, cathodoluminescence (CL) images were taken at the Institute of Earth Sciences, Academia Sinica, Taiwan. CL pictures help to locate analytical spots by examining the internal structures of individual zircon grains and avoiding inclusions during analysis. In-situ laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) U-Pb dating was performed using an Agilent 7500s quadrupole ICP-MS coupled with a 193 nm Analyte G2 excimer laser ablation system housed at the Department of Geosciences, National Taiwan University. The beam size of the laser ablation pits is ∼30 μm in diameter. Measured U-Th-Pb isotope ratios were calculated using the GLITTER 4.4 (GEMOC) software, and calibration was performed using the zircon standard GJ-1 aged at 608.5 ± 0.4 Ma (Jackson et al., 2004). Zircon 91,500 (1065 ± 0.4 Ma; Wiedenbeck et al., 1995) and Plešovice (337.1 ± 0.4 Ma; Sláma et al., 2008) were analyzed as secondary standards for data quality control in each analysis cycle. Operating conditions and analytical procedures followed (Chiu et al. 2009). Common lead in zircons was corrected with the lead correction function proposed by Andersen (2002), and weighted mean U-Pb ages and Concordia plots were then conducted using Isoplot 4.15. (Ludwig, 2008). For U-Pb ages <50 Ma, ²⁰⁶Pb/²³⁸U ages are chosen without consideration of concordance due to the low U concentration of mafic samples and the imprecision of ²⁰⁷Pb measurements for such young zircons using the LA-ICPMS method (Gehrels et al., 2008; Spencer et al., 2016).

For ⁴⁰Ar/³⁹Ar dating, samples and standards were cleaned with an ultrasonic bath in acetone, alcohol, and deionized water, respectively, and then dried. Approximately 250 mg of amphibole separated from amphibolite pebbles was packaged and irradiated along with standards in the VT-C position of the Tsing-Hua Open-Pool Reactor (THOR) at Tsing-Hua University, Taiwan. Fish Canyon Sanidine (FCs) standard of 28.10 ± 0.04 Ma (Spell and McDougall, 2003) was used to monitor the neutron flux, and the J-value of 0.0047937 ± 0.000025 at the 1-sigma level was obtained from the analyses of FCs adjacent to the sample. The sample was loaded (without any other sample or standard) into an ultra-high vacuum laser chamber and baked at 150°C for over 24 h to remove atmospheric contamination prior to analysis. Argon was extracted using the single-grain total fusion method with a Synrad 48–1 CO₂ laser fusion system attached to a VG3600 mass spectrometer at National Taiwan Normal University (NTNU). Before isotopic analysis, the sample gas was purified by a Ti-sponge for 3 min and then by two getters (one kept at 450 C and the other at room temperature) for 5 min to remove active gases (e.g., CO₂, H₂O, CH₄). Detailed instrumental and analytical procedures are outlined in Lo et al. (2002). Data processing, including peak intensity regression and corrections for blanks, mass spectrometer discrimination, isotopic decay, and neutron-induced interference reactions, was performed using the ArArCALC software of Koppers (2002). The atmospheric argon ratio of 298.56 ± 0.31 was suggested by Lee et al. (2006). Isochron and weighted mean age re-statistics were obtained using IsoplotR (Vermeesch, 2018). Uncertainties in ages were reported as one standard deviation, taking into account errors in J-values, standard mineral ages, and isotopic interference factors.

To determine how the ophiolite-dominated Shihmen Conglomerate was intercalated with pelagic turbidites from the South China Sea, the provenance properties of the pebbles have to be investigated. The data reported here can elucidate key provenance characteristics of the pebbles, such as the parent rock affinities, formation ages, and thermal-metamorphic history for amphibolitic clasts. Both plutonic gabbro and foliated amphibolite pebbles yielded similar U-Pb zircon crystallization ages of 22–24 Ma (Figures 4A,B), indicating that the dated gabbro was the protolith of the foliated amphibolites. The age results are consistent with geochemical analyses indicating a common ocean ridge tholeiite origin for both the gabbroic and amphibolite clasts (Page and Lan, 1983; Pelletier and Stephan, 1986). The gabbro crystallization age is similar to that reported by Wang et al. (2012) and falls within the time span of South China Sea spreading (∼37 - ∼15 Ma; Taylor and Hayes, 1983; Briais et al., 1993; Yeh et al., 2010), therefore the South China Sea crust should be the common source of these mafic materials since the oceanic crust of the westernmost Philippine Sea plate has been interpreted as either Eocene (∼50–∼40 Ma: Karig et al., 1975; Sibuet et al., 2002) or Cretaceous (∼130–∼120 Ma: Deschamps et al., 2000; Queano et al., 2007), ruling out the upper Manila Trench plate as a possible source. The dating results are also consistent with previous attempts to analyze these mafic pebbles (Pelletier and Stephan, 1986; Tian et al., 2019; Cui et al., 2021; Meng et al., 2021), and they also concur with the paleocurrent analyses pointing to Eurasian-side sources instead of the overriding plate (Chang et al., 2003).

The cause of metamorphism on the foliated amphibolite could be inferred from the ⁴⁰Ar/³⁹Ar amphibole ages. Since the turbidite sequence exposed on the Hengchun Peninsula is not metamorphosed with a maximum peak temperature <200°C as suggested by vitrinite reflectance constraints (Zhang et al., 2016), the K-Ar isotope system of amphibole, with much higher closure temperatures (∼490–578°C, Harrison, 1982), must have remained closed during subduction to accretionary wedge stages. The obtained isochron age of 13.3 ± 2.0 Ma, roughly 10 million years younger than the gabbro crystallization age from zircon U-Pb dating, also rules out sub-seafloor metamorphism as the main source, which typically occurs immediately after oceanic lithosphere formation along mid-ocean ridges with proximal heating and generally produces extensive veining and sulfide deposits indicative of intense fluid-rock interactions (Frost and Frost, 2014). The foliated amphibolite pebbles are mostly vein-free and exhibit significant ductile shearing as the aligned amphibole crystals and dynamically recrystallizing albite grains (Figures 3A,B), which is characteristic of tectonic deformation along the major shear zone rather than metasomatic seafloor metamorphism. The ⁴⁰Ar/³⁹Ar amphibole ages record either amphibole paragenesis of both metamorphism and shearing or cooling during the subsequent exhumation. These would indicate that part of the 22–24 Ma gabbroic oceanic crust that is the source of the Shihmen Conglomerate must have been dynamically metamorphosed at ∼13 Ma or slightly earlier.

The occurrence of both gabbro and foliated amphibolite as rounded pebbles up to boulder sizes, together with other mafic clasts, indicates that a portion of the South China Sea crust was not only subaerial but must have possessed high relief to generate the torrential dumping of these coarse deposits into the South China Sea basin (Pelletier and Stephan, 1986). Detrital zircon U-Pb ages from the concordant sandy interlayer (23.0 ± 0.3 Ma, Figure 4C) are almost identical to those from the gabbroic (22.8 ± 0.6 Ma, Figure 4B) and foliated amphibolite pebbles (24.0 ± 0.4 Ma, Figure 4A), demonstrating that this ophiolitic source of the South China Sea ages produced both coarse pebbles and sands. The mafic magmatic origin of the dated zircon crystals from both pebbles and sandstones is also indicated by the low U content, which is mostly less than 100 ppm or even less than 10 ppm. The age spectrum of the Shihmen sands contrasts dramatically with normal sandstones of the Mutan Formation (Figure 6A) which are dominated by various pre-Cenozoic zircon populations of the southeastern Chinese continent, including Yenshanian (200–65 Ma), Indosinian (250–200 Ma) and Luliangian (1900–1700 Ma) ages, with extremely rare Tertiary (<50 Ma) results (Zhang et al., 2014; Tsai et al., 2020). The turbidite layers adjacent to or in the periphery of the Shihmen deep-sea fan yield zircon populations (Figure 6B; Tsai et al., 2020; Meng et al., 2021) as diluted mixtures between Shihmen sands (Figure 6C) and normal Mutan sands (Figure 6A). Taking into account the inference from paleocurrent analyses that the sediment transport and source directions for the entire Mutan Formation were from the southeastern Chinese continent (Chang et al., 2003), the provenance of the Shihmen Conglomerate would be an ephemeral isolated rugged mountain composed exclusively of South China Sea ophiolite on the Eurasian continental margin, and the sediment routing system would be separated from the major rivers and their offshore extensions that drained the other common Mutan turbidites (Figure 7B). The tentatively postulated short-lived mountain building was most likely an obduction of the South China Sea crust onto the Eurasian continental margin at ∼13 Ma (Figure 7B); the lower-crustal gabbroic rocks along and near the basal thrust were dynamically recrystallized to foliated amphibolite in a metamorphic sole-like fashion (Searle et al., 2022; profile B-2 in Figure 7B) in the hanging-wall; both the upthrust gabbro and amphibolite, with minor other mafic lithologies, created a high-relief topography and delivered coarse detritus eastward to form deep-sea channel and fan turbidite deposits that are intercalated with normal Mutan sediments during the mid-Miocene at ∼11–13 Ma (Sung, 1991, profile B-2 in Figure 7B). Another possibility is that the amphibolite was generated during subduction metamorphism of the South China Sea slab, and a portion of it was extruded back into the shallow crust and amalgamated with the obducting ophiolite sequence. The predominance of gabbroic lithologies in the pebbles may be due to the slow-spreading nature of the South China Sea crust (Chung and Sun, 1992; Briais et al., 1993).

The mechanism of the brief South China Sea obduction proposed above can be elucidated by the temporal relationships with tectonic events in the Manila subduction system and the Taiwan collisional orogeny. Immediately after seafloor spreading, the South China Sea was subducted under the Philippine Sea Plate along the Manila Trench since at least 15 Ma, as evidenced by the earliest arc magmatism (Lai et al., 2018; Figure 7A). The first continental subduction at the Manila Trench, which probably took place near present-day northern Taiwan due to the oblique configuration between the trench and the COT (Suppe, 1988; Figure 7B), was estimated to be around 13 Ma (Chen et al., 2019) and was likely 11–9 Ma in central Taiwan, as inferred from the onset of blueschist retrograde cooling (Lo and Yui, 1996; Lo et al., 2020). The date of the obduction in question was close to these estimates of the initial impingement of the continent on the Manila Trench, which would induce major changes in the mechanical boundary conditions of the plate boundary (Malavieille and Trullenque, 2009) to cause the trench jump as the convergent strain was partitioned away from the trench itself (Chang et al., 2009b). Such mechanical transformation lies in the different frictional properties and seismic coupling of the subduction zone and is controlled by the roughness of the incoming seafloor (Lallemand et al., 2018), particularly the presence of cover-basement irregularities such as seamounts (Tan et al., 2022). Greater roughness is expected for the incorporation of continental crust with rugged topography and the cover-basement interface of the continental margin, causing strain partitioning from the subduction interface to a larger area, eventually leading to mountain building (Chang et al., 2009b; Malavieille et al., 2021). For an obliquely subducting oceanic slab such as the South China Sea, the initial arrival of the Eurasian continental margin at the Manila Trench in the north would locally increase plate coupling and interface friction (profile B-1 of Figure 7B); a sudden interruption of subduction would divert a portion of the plate convergence to the COT southwestward as one of the major structural weaknesses within the descending plate, resulting in the simultaneous obduction of the South China Sea onto the Eurasian continental margin (Figure 7B). The hypothesized obduction would be short-lived, as the strain reorganization described above is transient in nature; therefore, the Shihmen Conglomerate has a rather limited temporal occurrence, no later than ∼11 Ma. The Shihmen Conglomerate, as part of the South China Sea sedimentary succession (the Mutan Formation), was then incorporated into the accretionary wedge of the Manila subduction system and is now exhumed on the Hengchun Peninsula (Figure 7C).