The influence of Hainan mantle plume and subducting recycled oceanic crust beneath the spreading ridge of the South China Sea (SCS) have been widely proposed recently, but still controversial and ambiguous. Here, we present seismic tomographic evidence, new major and trace element, and Pb isotopic compositions of volcanic glasses from one International Ocean Drilling Program drill core (Site U1434) in the SCS spreading ridge. The volcanic glasses are relatively enriched in alkalis and light rare earth elements (LREEs) and depleted in heavy REEs (HREEs), exhibit slightly positive anomalies in Nb, Ta, Zr, and Hf as well as a positive Nb relative to La and Th, and show relatively high 207Pb/206Pb and 208Pb/206Pb isotopic ratios, suggesting ocean island basalt- (OIB-) type and enriched mantle 2- (EM2-) type geochemical features likely related to a mantle plume. These geochemical features are consistent with those of late Cenozoic volcanic rocks in Hainan and surrounding areas associated with a mantle plume, likely providing the influence of Hainan mantle plume beneath the spreading ridge of the SCS. The SCS primary-melt and volcanic glasses indicate that the source mantle involved 18.5% eclogite (dense, recycled oceanic crust from the stagnant subducted slab) and 46.1% garnet pyroxenite (produced by the reaction between the peridotite melt and recycled oceanic crust). The existence of Hainan mantle plume and stagnant subducted slab is further supported by geophysical evidence from a recent three-dimensional P-wave seismic tomographic model.
The South China Sea (SCS) is a western Pacific Cenozoic marginal sea surrounded and affected by motions of the Eurasian, Indian, and Philippine/Pacific plates, which has been commonly divided into eastern and southwestern subbasins taking the Zhongnan Fault as the boundary (Figure 1). The results of microfossils in the interflow claystone  and deep-tow magnetic anomaly lineations [2–4] indicate that SCS continental-margin breakup and seafloor spreading occurred at 32–33 Ma in the northeastern SCS [1, 4], this spreading ridge jumps 20 km southward in the East Subbasin which occurred at ~25 Ma , and the seafloor spreading in the Southwest Subbasin propagated 400 km southwestward during the period of 23.6-21.5 Ma . The terminal age of seafloor spreading is 15 Ma in the East Subbasin and 16 Ma in the Southwest Subbasin . Subsequently, the eastern subbasin of the SCS subducted eastward along the Manila Trench . The collision between the northern Luzon Arc and the South China continent occurred at approximately 6.5 Ma . Collectively, the SCS records a diverse array of spatially and temporally complex tectonic processes, including continental rifting, seafloor spreading, subduction, and terrane collision, which could be termed a complete Wilson cycle . However, the tectonic dynamics of the SCS seafloor opening, the existence of the Hainan mantle plume, and the subducting recycled oceanic crust beneath the SCS spreading ridge remain controversial and ambiguous.
Previous comprehensive studies have identified the Hainan mantle plume based on seismic tomography of a plume-like mantle low-velocity structure [7–10], petrologic and geochemical features of volcanic rocks [11, 12], and a high mantle potential temperature developed beneath the Hainan (1440–1550°C; [11, 13]). Furthermore, Lei et al.  proposed that the Hainan plume likely originated from the lower mantle [15, 16]. The late Cenozoic ocean island basalt- (OIB-) type Hainan flood basalts [11–13, 17] and the widespread late Paleocene OIB-type intraplate volcanism of the South China continental margin (Figure 1; [11, 12, 18–21]) have all been correlated to the Hainan plume. Consequently, the late Cenozoic OIB-type volcanism in the SCS has been correspondingly thought to be related to the Hainan plume (Figure 1; [11, 12, 20]). Xu et al.  first proposed the subridge mantle beneath the SCS spreading center through mantle plume-ridge interactions. Yu et al.  reported that the middle Miocene SCS midocean ridge basalts (MORBs) record progressive mantle enrichment and reflect the contribution of the Hainan plume. Relatively young OIB-type volcanic clasts (<8.3 Ma) are found at the Hole U1431 , which are explained to be converted by the carbonated silicate melts and to be formed by the reactions with the lithospheric mantle. Both normal- (N-) MORB-type and enriched- (E-) MORB-type basalts in the eastern subbasin are explained to be influenced by the Hainan mantle plume ; the E-MORB-type basalts in the southwestern subbasin are affected by the recycling of lower continental crust . Zhang et al.  proposed that a Hainan hot spot with pyroxenite-rich components could have affected the dying spreading ridge of the SCS through hot spot-ridge interactions. Yang et al.  calculated the crystallization temperature of primitive olivine and manifested the mantle plume-ridge interaction during the last spreading of the SCS. Yang et al.  determined the SCS magma dynamics associated with oceanic crustal accretion at slow spreading ridges on the basis of the SCS plagioclase core-rim zonation.
The existence of the Hainan mantle plume and subduction-related recycled materials beneath the spreading ridge of the SCS has been proven by the above-mentioned related studies, but still been challenged and debated. To thoroughly comprehend, we utilize the latest seismic tomographic results and collected representative volcanic glasses from Site U1434 in the southwestern subbasin (Table S1; Fig. S1) as part of the International Ocean Drilling Program (IODP) Expedition 349 which was completed near the relict spreading center in the SCS. In this study, we present a comprehensive set of chemical data of volcanic glasses to elucidate the existence of the Hainan mantle plume beneath the SCS spreading center, calculate the primary-melt compositions to evaluate the influence of recycled oceanic crust from the stagnant subducted slab, and finally present a three-dimensional P-wave seismic tomographic model to verify our conclusions.
2. Sampling Setting and Petrography
The SCS was a western Pacific marginal sea at the junction of the Eurasian, Indo-Australian, and Philippine-Pacific plates (Figure 1) and was surrounded by the South China fold belt to the north, a subduction zone to the east, the Borneo Trough to the south, and the East Vietnam fault to the west (Figure 1). Volcanic glasses recovered during IODP Expedition 349 at Site U1434 in the southwestern subbasin (Table S1; Fig. S1) are used in this study. Samples from IODP Site U1434, which are situated immediately south of a large seamount near the relict spreading center in the southwestern subbasin and approximately 40 km northwest of IODP Site U1433 (Figure 1), consist mainly of seven igneous units spanning a total depth interval of 30.38 m (in the range 278.27–308.65 mbsf) and grouped as lithostratigraphic Unit IV with a succession of minor pillow basalt flows. The igneous basement rocks are overlain by hemipelagic claystone (Unit III) (Fig. S1).
Petrographically, volcanic glasses recovered from Site U1434 volcanic samples appear to be relatively fresh and have a massive structure (Figures 2(a) and 2(b)) with an aphanitic texture with cryptocrystalline groundmass (50 vol.%) and consist mainly of volcanic glasses (50 vol.%) (Figures 2(c) and 2(d)). These volcanic glasses show larger particles (0.3–0.6 mm), irregular shapes, and a uniform surface composition (Figures 2(e) and 2(f)).
3. Analytical Methods
Mineral analyses were performed using electron probe microanalyzer (EPMA) with four wavelength-dispersive spectrometers (WDS) by JEOL JXA-8230 electron probe microanalyzers at the Center for Global Tectonics, School of Earth Sciences, China University of Geosciences (Wuhan). The operating conditions were described in Wang et al.  in detail. 15 kV accelerating voltage, 20 nA probe current, and a 5-micron beam diameter had been used. Dwell times were 10 s on element peaks and half that on background locations adjacent to peaks. Raw X-ray intensities were corrected using a ZAF (atomic number, absorption, and fluorescence) correction procedure. A series of natural and synthetic SPI standards were utilized and changed based on the analyzing minerals. The standard samples have been analyzed five times to get the average values. The following standards were used: Sanidine (K), Pyrope Garnet (Fe, Al), Diopside (Ca, Mg), Jadeite (Na), Rhodonite (Mn), Olivine (Si), and Rutile (Ti). The major elements of volcanic glass are presented in Table 1.
Trace element analyses of volcanic glasses were conducted using Agilent 7900 Laser-ablation Inductively Coupled Plasma Mass Spectrometer (LA-ICP-MS) at the Wuhan SampleSolution Analytical Technology Co., Ltd., Wuhan, China. Detailed operating conditions for the laser ablation system and the ICP-MS instrument and data reduction are the same as description by Zong et al. . Laser sampling was performed using a GeoLas HD laser ablation system that consists of a COMPexPro 102 ArF excimer laser (wavelength of 193 nm and maximum energy of 200 mJ) and a MicroLas optical system. An Agilent 7900 ICP-MS instrument was used to acquire ion-signal intensities. The spot size and frequency of the laser were set to 44 μm and 5 Hz, respectively. Each analysis incorporated a background acquisition of approximately 20–30 s followed by 50 s of data acquisition from the sample. The internal standard used was 29Si determined by EPMA analysis. Three geochemically distinct reference glasses (BCR-2G, BIR-1G, and BHVO-2G) were used to cover the possible geochemical spectrum. An Excel-based software ICPMSDataCal10.8 was used to perform off-line selection and integration of background and analyzed signals, time-drift correction, and quantitative calibration for trace element analysis . Relative standard deviations (% RSD) of LA-ICP-MS analyses are less than 10% for all trace elements. The trace elements of volcanic glasses are presented in Table 2.
In situ lead isotope analyses in volcanic glasses were performed on a Neptune Plus MC-ICP-MS (Thermo Scientific), coupled with a RESOlution M-50 193 nm laser ablation system (Resonetics), which are hosted at the State Key Laboratory of Isotope Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIG-CAS). An X skimmer cone and Jet sample cone in the interface were used to improve the instrumental sensitivity. All isotope signals are detected with ion counters under static mode. The laser parameters were set as follows: beam diameter, 45 μm; repetition rate, 4 Hz; and energy density, ~4 J cm-2. Helium was chosen as the carrier gas (800 ml min-1). Each analysis consisted of 250 cycles with an integration time of 0.262 s per cycle. The first 28 s was used to detect the gas blank with the laser beam off, followed by 30 s laser ablation for sample signal collection with laser beam on. The mass bias and instrumental drift were corrected by using the standard-sample-bracketing method (SSB). NKT-1G (a basaltic glass, Elburg et al.  was chosen as the external standard. The detailed analytical procedure is reported in Zhang et al. . 30 analyses of an international basaltic glass (BHVO-2G) during the course of this study yielded a weighted mean of (2SD) and (2SD) (Table 3), which is consistent within errors with the reported value in Weis et al. .
4.1. Major and Trace Elements of Hole U1434A Volcanic Glasses
The studied volcanic glasses are contained in the Hole U1434A samples, and their geochemical features are distinctly different from those of the bulk-rock samples (Supplementary 1; Figure 3). The volcanic glasses are characterized by relatively high total alkali contents () and K2O contents (2.56–2.85 wt.%) (Table 1; Figure 3) and plot mainly within the field of trachybasalts in the TAS classification diagram (Figure 3). By contrast, the bulk-rock samples have relatively low total alkali content, mainly belonging to tholeiitic basalts (Supplementary 1; Figure 3). The volcanic glasses possess relatively low values of Mg# ranging from 37 to 40 and MgO (4.79–5.35 wt.%) over a limited range of SiO2 content from 48.05 to 49.59 wt.% (Table 1). The volcanic glasses exhibit relatively low SiO2, CaO, FeOT, and Na2O contents, a low CaO/Al2O3 ratio, slightly high Al2O3, TiO2, and K2O contents (Figure 4) in comparison with the SCS bulk-rock samples (Supplementary 1).
The chondrite-normalized rare earth element (REE) plots for the volcanic glasses are characterized by enrichment in the light REEs (LREEs; ) and relatively high total REE contents () and show no Eu anomalies () (Table 2; Figure 5(c)). Their REE patterns are similar to typical ocean island basalt (OIB) , but their REE contents are slightly higher than those of OIB (Figure 5(c)). The primitive mantle-normalized multielement patterns for the volcanic glasses show enrichment in the large ion lithophile elements (LILEs), similar to the pattern of OIB but slightly higher than the LILE contents of OIB (Figure 5(d)). The volcanic glasses have positive Nb, Ta, Zr, and Hf anomalies (Figure 5(d)), and relatively high ratios of (Nb/La)pm and (Nb/Th)pm (; ) (Table 2), suggesting the OIB-type geochemical characteristics associated with mantle plumes . The contents of Th in the volcanic glasses are approximately 5.70–8.06 ppm in average close to that of the average OIB (4 ppm) and that of average calc-alkaline basalts (8.4 ppm; ).
4.2. Pb Isotopic Compositions of Hole U1434A Volcanic Glasses
SCS Hole U1434A volcanic glasses have relatively low 208Pb/206Pb ratios ranging from 2.117668 to 2.158083 and high 207Pb/206Pb ratios ranging from 0.847677 to 0.872731 relative to those of enriched mantle 2 (EM2) (Table 3). These U1434 volcanic glasses exhibit a mixing of depleted mantle (DMM) and EM2 components in their mantle source (Figure 6), plot close to the EM2 end-members (Figure 6; Zindler and Hart, 1986; Hofmann, 1988). Volcanic rocks of eastern and southwestern SCS subbasins show significantly different Pb isotopic compositions. Site U1433 and U1434 basalts and volcanic glasses from southwestern SCS subbasin have relatively high 208Pb/206Pb and 207Pb/206Pb ratios (2.117668–2.158083; 0.847677–0.876637; Figure 6; ), whereas Site U1431 basalts from eastern SCS subbasin show relatively low 208Pb/206Pb and 207Pb/206Pb ratios (2.068157–2.092284; 0.836147–0.849262; ), which are more similar to those of Hainan island and the surrounding region (Figure 6) related to the existence of mantle plume. Our studied U1434 volcanic glasses and previous studied bulk-rocks are within the field of Indian-type MORB with the significant Dupal anomaly (Figure 6). The studied U1434 volcanic glasses have relatively higher 208Pb/206Pb and 207Pb/206Pb isotopic ratios than those of samples from the Philippine Sea plate (Figure 6), further indicating the strong Dupal anomaly.
4.3. Seismic Tomography Results
The present tomographic result is part of the result from a unpublished work, which is obtained by inverting a great number of local and teleseismic P-wave travel time data (Fig. S7) recorded by seismic networks deployed in Southeast Asia and its adjacent areas (Fig. S8; ), using a grid-discontinuity tomographic method . The result of the checkerboard resolution test is shown in Fig. S9, which shows the majority part of the study region can be well imaged using the seismic data, but beneath the SCS expansion center the resolution is low because of the lack of seismic rays there (Fig. S9). We present the isotropic P-wave velocity perturbations (dVp) with respect to the lateral average of the tomographic model. The areas with a negative dVp (i.e., the P-wave velocity is smaller than lateral average at the same depth) are shown in red colors whereas those areas with a positive dVp are shown in blue colors. Seven cross-sectional slices of isotropic P-wave velocity perturbations beneath the South China Sea and Hainan island are presented in Figure 7. Firstly, two-way subduction zone on the eastern edge of the SCS is observed (C–C’ section in Figure 7(c)). One is the young SCS oceanic slab subducting eastward to the Manila Trench; the other is the ancient SCS oceanic slab (or paleo-Pacific oceanic slab) subducting westward beneath the South China Sea (C–C’ section in Figure 7(c)). As the westwardly subducting oceanic slabs broke and fell to the mantle transition zone, these subduction-faulted oceanic slabs were detained in the mantle transition zone and migrated beneath the SCS and its western edge (D–D’ section, E–E’ section, and F–F’ section in Figures 7(d)–7(f)). Accordingly, the positive velocity anomaly is clearly observed near the mantle transition zone (Figures 7(a), 7(b), and 7(d)–7(f)). Secondly, the negative velocity anomaly beneath the Hainan island (G-G’ section in Figure 7(g)) and the northern margin of the SCS is obviously observed (A–A’ section and B–B’ section in Figures 7(a)–7(b)), which likely represents a Hainan mantle plume as indicated by Lei et al. . This negative velocity anomaly could extend from the lower mantle (Figures 7(a), 7(b), and 7(g)), likely suggesting that the Hainan plume originated from the lower mantle [14–16]. Cross-sectional slices of isotropic P-wave velocity perturbations show that a significantly negative velocity anomaly exists only beneath the Hainan island and the northern margin of the SCS (Figures 7(a), 7(b), and 7(g)). Because of the lack of seismic information of the SCS expansion center, it is difficult to accurately identify whether this negative velocity anomaly has extended to the SCS expansion center.
Depth slices across the South China Sea and Hainan island are shown in Figure 8. On the one hand, the positive velocity anomaly in the depth of 500–660 km beneath the SCS is observed (Figures 8(f)–8(h)), likely indicating the stagnant subducted slab in the mantle transition zone. On the other hand, the negative velocity anomaly in the depth of 250 km, 300 km, and 350 km likely represents an upwelling hot mantle plume as indicated by Wei and Chen . The negative velocity anomaly proliferates around the Hainan island, the north margin of the SCS, and the SCS spreading center (Figures 8(a)–8(c)), likely reflecting the influence area of Hainan mantle plume at the depth of 250–350 km. However, due to the lack of seismic data in the expansion center of the SCS, we cannot obtain seismic tomography with a depth of less than 250 km in the SCS expansion center; thus, we cannot exactly judge whether the Hainan mantle plume is flowing towards the SCS expansion center, nor can we observe the shallow spreading ridge magma flow.
5.1. Fractional Crystallization
The SCS volcanic samples exhibit large variations in Cr and Ni contents ranging from 210 to 665 ppm and from 65 to 537 ppm, respectively (Table S2, S3: Supplementary 1), indicating that they are not primary magma, but have likely undergone the process of magma evolution or fractional crystallization . Extremely complex correlations between the major-element compositions and MgO are observed (Figure 4), especially for the complex behaviors of SiO2, CaO, and Al2O3 (Figures 4(a), 4(c), and 4(f)). The constant FeOt contents (Figure 4(d)) and CaO/Al2O3 ratios (Figure 4(i)) with increasing MgO likely suggest fractional crystallization of olivine [11, 44]. The slightly decreasing Na2O and K2O contents (Figures 4(g) and 4(h)) and the constant TiO2 contents (Figure 4(b)) with increasing MgO likely indicates crystallization of clinopyroxene and olivine. The absence of the negative correlation between TiO2 and MgO (Figure 4(b)) excludes the fractional crystallization of titanium-containing minerals. The absence of correlation between Al2O3 and MgO (Figure 4(f)) and the slightly positive (or nonexistent) Eu anomalies (Figure 5(a)) likely preclude plagioclase fractionation or accumulation.
To quantify fractionation history, we applied MELTS_Excel (; http://melts.ofm-research.org) to place constraints on the crystallization history of SCS Site U1431, U1433, and U1434 lavas. We used the sample of SCS lavas with the value of MgO (8.43 wt.%) instead of the highest MgO (17.20 wt.%) to represent the parental magma, because this high-Mg sample () exhibits a cumulate texture and has large amounts of olivine phenocrysts, which is likely attributed to olivine accumulation (Supplementary 1 and Figs. S2A-C). We set the pressure to be 3.2 kbar which is the crystallization pressure based on the clinopyroxene-liquid compositions (Supplementary 1; ) and opened the system to oxygen exchange at the quartz-fayalite-magnetite (QFM) buffer based on the lithologic character and tectonic environment as Carmichael and Ghiorso  suggested. The detailed initial MELTS_Excel calculation results are presented in Appendix A. The liquidus temperature of our starting composition is selected at 1200°C (Appendix A). We then decreased the temperature in steps of 2°C. At approximately 1198°C, the olivine phase was shown to begin to crystallize. At >1140°C, olivine was the only phase crystallized and was completely removed from the system (Appendix A). Below 1140°C at MgO of ~6.36 wt.%, plagioclase phases crystallized (Figure 4; Appendix A). When the temperature dropped to 1102°C and 1074°C, the clinopyroxene and spinel, respectively, begin to crystallize (Figure 4; Appendix A). We used the final results from the MELTS_Excel calculations to model SCS fractional crystallization (the thick light gray lines in Figures 4, 9, and 10). Most bulk-rock major elements are consistent with the MELTS_Excel calculation results, except for the large variations of SiO2, CaO, and Al2O3 contents (Figures 4(a), 4(c), and 4(f)) which are likely caused by other geological process (e.g., magma mixing). Additionally, for comparison, Appendix B displays the MELTS_Excel calculation results of the system with the pressure of 3.2 kbar and the oxygen fugacity of nickel-nickel oxide (NNO) buffer.
5.2. Primary-Melt Compositions Calculated by Simulation Using the PRIMELT3 and MELTS_Excel Software
Our studied samples are the products of fractional crystallization of the primary magma produced by partial melting, which could be used to represent the chemical composition of the melt at different stages of magma evolution but not of the primary magma. As minerals continuously crystallize and separate from the residual melt as the magma cools, the chemical composition of that residual melt changes. Therefore, the evolving chemical composition of the residual melt corresponding to the precipitation of minerals along a crystallization path serves to constraining the composition of the primary magma [11, 46–48].
The fractional crystallization of minerals during magmatic differentiation could be determined by forward or inversion models [40, 49]. As discussed above, the MELTS_Excel software relies on a crystallization forward model by which one can control changes in temperature and pressure during the crystallization process through petrological simulations to analyze the composition of precipitated minerals and residual melt . The PRIMELT3 software uses a crystallization inversion model, whereby the magma composition gradually approaches that of the primary melt by adding olivines to the evolved magma .
The results in Figure 4, Appendix A, and MELTS_Excel have proved that the crystallization of clinopyroxene and plagioclase starts at <1140°C with MgO of ~6.36 wt.%. To minimize the effect of clinopyroxene and plagioclase fractionation, only samples with were chosen as starting samples. Based on the iteration method of Wang et al. , primitive melt compositions were obtained by adding olivines for those samples that are presumed to have experienced only olivine fractionation. The most-magnesian olivine phenocrysts in our studied SCS volcanic rocks have values as high as Fo89.2 (Table S7). Accordingly, we repeated the calculations of olivine and basaltic compositions until the equilibrium olivines reached Fo89.2. It is noteworthy that the SCS sample with the value of MgO (8.43 wt.%) instead of the highest MgO (17.20 wt.%) was selected to represent the parental magma, because this high-Mg sample was likely caused by olivine accumulation (Figs. S2A-C). The amount of olivine addition required to achieve liquid Mg# values consistent with a Fo89.2 source is typically in the range 1%–5% for the SCS basalts (Appendix C). The estimated primary-melt composition results are presented in Appendix C.
5.3. Lithology of the SCS Mantle Source
Although Zhang et al.  have confirmed that the postspreading volcanism at Site U1431 was formed by melting of a recycled oceanic crust-derived eclogite/pyroxenite-rich mantle source rising from a deeper mantle plume, this study utilized the compositions of volcanic glasses and calculated primary melts to further confirm the lithology of the SCS mantle source using the supplementary identification methods.
Liu et al.  determined that garnet pyroxenite, peridotite+CO2, eclogite, and hornblendite sources have distinctive major- and trace-element geochemical characteristics compared to those of peridotite sources (Figure 9; ). Comparisons between the calculated SCS primary magmas, SCS Site U1434 volcanic glasses, and types of experimental partial melts  have shown that relatively low TiO2 contents (1.15–1.61 wt.%) and relatively high Na2O/TiO2 (1.78–2.60) and CaO/Al2O3 (0.54–0.68) ratios of the calculated SCS primary melts are indicative of the lithology of garnet pyroxenite, eclogite, and peridotite melts (Figure 9), and relatively high TiO2 contents (2.04–2.43 wt.%) and slightly low FeOT contents (7.60–8.30 wt.%), Na2O/TiO2 (1.15–1.41), and CaO/Al2O3 (0.49–0.51) ratios of the SCS Site U1434 volcanic glasses are consistent with the involvement of eclogite melts (Figure 9).
The FC3MS (FeO/CaO–3MgO/SiO2) parameter proposed by Yang and Zhou  has been determined to be among the best indicators to discriminate peridotite and pyroxenite magma sources [50, 51]. The conducted melting experiments inferred that the FC3MS ratios of basaltic samples from peridotite sources are less than 0.65 while those from pyroxenite melts yield a minimum ratio of 0.65 (; Figure 10). The FC3MS ratios of the calculated primary-melt compositions and Site U1434 volcanic glasses range from 0.42 to 1.04, indicating that the SCS samples were derived mainly from pyroxenite melts but with the involvement of peridotite melts.
Humayun et al.  proposed that Fe/Mn ratios represent more intensive fractionation when garnet and clinopyroxene are the dominant phases relative to the olivine and orthopyroxene involved. The calculated SCS primary melts are characterized by relatively high Fe/Mn ratios (60.31–77.57) that are slightly higher than the ratio ranges for peridotite melts (Figure 11(a)), indicating that the studied primary melts were derived mainly from pyroxenite partial melts  with limited involvement of mantle peridotite  (Figure 11(a)). Also, olivine phenocrysts in the SCS volcanic rocks show relatively lower Mn/Fe and Ca/Fe ratios (Figure 11(b)), further testifying the significant influence of the pyroxenite addition. A pyroxenite magma source with the low olivine abundance and large amounts of pyroxene would bring about a higher bulk partition coefficients for Mn and Ca , ultimately resulting in the SCS melts with low Mn and Ca contents .
5.4. Estimation of Proportion of Pyroxenite and Recycled Oceanic Crust in SCS Basalts.
We have confirmed that the calculated SCS primary melts dominantly origin from garnet pyroxenite melts but with the involvement of eclogite and peridotite melts . Many previous studies had shown that pyroxene and garnet may be produced by the reaction of silica-oversaturated eclogite and olivine from peridotite in the convecting mantle [58, 59], finally resulting in a refertilized peridotite enriched in garnet pyroxene [60, 61]. That widely shared interpretation is consistent with our findings. Next, we calculated the variable mixing proportions of individual components, including the pyroxenite, eclogite, and peridotite that are presumed to have contributed to the SCS primary melts.
First, we calculated the proportion of the pyroxenitic end-member (), which is closely related to the calculated Mn/Fe ratios of equilibrium olivines, by applying the formula of . The calculated 100Mn/Fe ratio of the most-magnesian phenocrystic olivine is approximately 1.46, thus getting a value of 46.1% for using this equation. Next, we calculated the proportions of recycled oceanic crust (), which are closely related to the prior-determined proportions of pyroxenite (), the degree of melting of eclogite (), the proportions of initial-reaction eclogite-derived melt (), and the degrees of melting of peridotite () and pyroxenite (), via the formula of proposed by Sobolev et al. . In this formula, the proportions of produced pyroxenite () equal those of initial-reaction primary (eclogite-derived) melt () , the assumed extent of melting of eclogites () is approximately 50%, the degree of batch melting of pyroxenite () is lower than 60% at low pressure (1–2 GPa), and the degree of melt fraction of peridotite () is approximately 20% for OIB with thin lithosphere [59, 62]. Finally our calculation resulted in a value for of ~18.5%, which is also supported by geochemical features, as further discussed below.
Sobolev et al.  estimated the proportions of melt derived from pyroxenite, upwelling peridotite, and recycled oceanic crust for each parental magma from Siberian LIP, Hawaii island, Detroit Smt, Iceland island, and MORB. These proportions are closely related to the amount of recycled oceanic crust in the upwelling mantle, the thickness of lithosphere, and the potential mantle temperature . Our calculated and values (46.1% and 18.5%) and maximum mantle potential temperature (; Supplementary 1; ) of SCS spreading center are significantly higher than those of normal MORB (; ; and ), but similar to those of Detroit Smt (; ; and ) and Hawaii (; ; and ) , likely indicating the existence of upwelling mantle plumes carrying a significant amount of recycled oceanic crust from the deep mantle .
The calculated SCS primary-melt and volcanic glasses are characterized by a relatively large variation in Ce/Yb, Th/Nb, and Ba/La ratios (Table 2; Appendix C), likely indicating the influence of the significant involvement of fluid and bulk sediment/silicate melt derived from recycled oceanic crust. Because LILEs (e.g., K, Pb, and Ba) are more soluble and susceptible to fluid than HFSEs (e.g., REE, Zr, Hf), and bulk sediment has large variations of Ce and Th contents. In addition, our studied Site U1434 volcanic glasses exhibit Nb, Ta, and Ti enrichment (Figure 5), which are iconic features of stagnant subduction zones (Figures 7 and 8), because Nb, Ta, and Ti have a relatively high distribution coefficient and can be preferentially preserved in rutile minerals , which were commonly thought to be a residual phase during subduction. During the subduction of the oceanic crust, Nb, Ta, and Ti were preferentially detained by a residual phase in the form of refractory dense eclogite within the subducting slab. Thus, island-arc basalts are commonly enriched in LILEs, Pb, and LREEs but depleted in Nb, Ta, and Ti in the form of enriched peridotite. The continued accumulation of these refractory dense eclogites, enriched in Nb, Ta, and Ti and detained at depth in the mantle, produced a large-volume Nb, Ta, and Ti reservoir. Thus, to summarize, the studied SCS volcanic glasses likely exhibit the features of dense eclogite, attributable mainly to the involvement of subducting recycled oceanic crust .
5.5. The Influence of Hainan Mantle Plume beneath the Spreading Ridge of the SCS Inferred from the Geochemistry of Volcanic Glasses
Because volcanic glasses generally form under rapid cooling of magma, their components commonly record information about the magma at the time of solidification without the influence of magma evolution. In contrast, the components of bulk-rock samples keep an account of the complete process of magma evolution. The studied bulk-rock samples are tholeiitic basalts with relatively low total-alkali and K2O contents, which are characterized by trace-element distribution patterns typical of chemically enriched MORB (; Figures 3 and 5; Supplementary 1). However, Site U1434 volcanic glasses have much higher total-alkali and K2O contents and plot within the field of trachybasalts, whose REE and incompatible-element distribution patterns are similar to those of OIB (Figures 3 and 5). The trace-element ratios of similarly incompatible pairs (e.g., Th/U, Nb/U, Nb/La, Ba/Th, and Sr/Nd) are commonly used to constrain the characteristics of mantle sources because they undergo relatively limited fractionation during partial melting (Figure 12). The Nb/La and Sr/Nd ratios of the SCS bulk-rock tholeiites are closer to those of relatively depleted MORB (Figure 12; ), whereas our studied Site U1434 volcanic glasses show relatively enriched incompatible-element ratios (; ) relative to the bulk-rock samples and thus more closely resemble those of enriched OIB (Figure 12).
Zhang et al.  have shown that the basalts of the two subbasins in the SCS are characterized by different isotopic compositions, especially their different Pb isotopes. Our studied U1434 volcanic glasses further confirm the contrasted Pb isotope compositions of two SCS subbasins. Eastern SCS subbasin basalts at Site U1431 show relatively low 208Pb/206Pb and 207Pb/206Pb ratios, but southwestern subbasin basalts and volcanic glasses at Sites U1433 and U1434 display relatively high 208Pb/206Pb and 207Pb/206Pb ratios, suggesting compositional heterogeneity and different mantle evolution histories of these two subbasins . Our studied U1434 volcanic glasses are characterized by more enriched 208Pb/206Pb and 207Pb/206Pb ratios relative to those of volcanic samples from the Philippine Sea plate spreading centers (Figure 6). It indicates that, although the SCS is adjacent to the Philippine Sea plate and their formation is similar due to plate extension, there must be some involvement of significant enriched components in the SCS mantle source, which is definitely different from the Philippine Sea Plate. These enriched components are likely the involvement of the Hainan plume components, due to the occurrence of OIB-type volcanic glasses (Figure 5) with EM2-like mantle end-member radiogenic Pb compositions (Figure 6). Previous studies [11, 14, 17] have shown that EM2-like mantle end-member of volcanic activities is generally closely related to the Hainan OIBs. Furthermore, Zhang et al.  also confirmed that eastern and southwestern subbasin basalts at Sites U1431, U1433, and U1434 show geochemical trends that are consistent with the influence of Hainan plume, although the formation of U1434 and U1433 volcanic rocks is related to the recycling of lower continental crust. Zhang et al.  also verified that the Sr–Nd–Pb–Hf isotopic compositions of Site U1431 volcanic rocks varying from carbonatite to alkali basalt are similar to those of Hainan OIBs [24, 25].
In all derived geochemical diagrams (Figures 5 and 6 and S6 and 12), the geochemical features of the SCS Site U1434 volcanic glasses extremely resemble those of mantle-plume-related volcanic rocks from Hainan and its surrounding areas [11, 17, 38, 39]. In detail, (1) Site U1434 volcanic glasses show slightly positive anomalies in Nb, Ta, Zr, and Hf as well as a positive Nb relative to La and Th (; ; Figure 5(d)), suggesting OIB-type geochemical features associated with a mantle plume. Their OIB-type incompatible-element distribution patterns are similar to those of Hainan volcanic rocks (Figures 5(c) and 5(d)). (2) High olivine Ca and Mn contents, and low olivine Ni contents of SCS volcanic samples from the spreading ridge, resemble those of Hainan olivines (Fig. S6). (3) Relatively elevated Sr/Nd and Nb/La ratios of Site U1434 volcanic glasses are similar to those of Hainan volcanic rocks with OIB-type incompatible-element ratios (Figure 12). (4) U1434 volcanic glasses show relatively high 208Pb/206Pb and 207Pb/206Pb isotopic ratios plotting close to the EM2 end-member (Figure 6), which is generally thought to be closely related to the Hainan OIBs. In summary, our studied Site U1434 volcanic glasses likely provide the more direct evidence for the existence of Hainan mantle plume beneath the spreading ridge of the SCS.
5.6. Coexistence of Hainan Plume and Stagnant Slab: Evidence from a Three-Dimensional P-Wave Tomographic Model
Our three-dimensional P-wave tomography (Figures 7 and 8) further demonstrates the above geochemical discussions, which likely prove the existence of upwelling mantle plumes carrying significant amounts of subducting recycled oceanic crust from the stagnant subduction slab. On the one hand, seven cross-sectional slices of isotropic P-wave velocity perturbations beneath the SCS and Hainan island (Figure 7) significantly show a positive velocity anomaly at the depth of mantle transition zone (410–660 km). Depth slices across the SCS and Hainan island (Figure 8) also clearly display a positive velocity anomaly within the depth range of 400–600 km . The P-wave tomography model from Lei and Zhao  and the P-wave and S-wave tomography models from Montelli et al.  also show a positive velocity anomaly within the mantle transition zone. This positive velocity anomaly is likely to be stagnant subducting slab in the mantle transition zone which carrying a large amount of recycled oceanic crust [14, 64–66]. On the other hand, a negative velocity anomaly beneath the SCS and Hainan island is clearly observed at a depth of approximately 250 km in the cross-sectional slices (Figure 7) and depth slices (Figure 8), likely indicating an upwelling hot mantle around the SCS. This upwelling hot mantle may be Hainan mantle plume which likely originated from the lower mantle (Figure 7(g); ; Monetelli et al., 2006; ). Depth slices of 250 km and 300 km (Figures 8(a) and 8(b)) can clearly show that a deep hot asthenosphere mantle plume extends to the northern margin of the SCS, the SCS expansion center, Hainan Island, and even wider areas.
However, it should be noted that due to the lack of seismic data in the SCS expansion center, P-wave velocity structure at shallow depths (<250 km) cannot be obtained; thus, it is difficult to accurately identify whether the Hainan mantle plume has extended to the SCS expansion center, whether the shallow spreading magma flow existed within the depth of 250 km (Figures 7 and 8). From the present three-dimensional P-wave seismic tomographic images (Figures 7 and 8), we roughly infer that the SCS and Hainan island are likely to be affected by the deep upwelling hot mantle. In conclusion, our P-wave seismic tomographic results are nearly consistent with the geochemical data of bulk-rock and volcanic glasses, likely confirming that the existence of Hainan mantle plume and stagnant subduction slab beneath the spreading ridge of the SCS.
This study presents seismic tomographic evidence, new major and trace elements, and Pb isotopic compositions of volcanic glasses from one IODP drill cores (Site U1434) in the SCS spreading ridge. The volcanic glasses belong to relatively alkali-rich trachybasalts and show the enrichment of LREE, the depletion of HREE, slightly positive anomalies in Nb, Ta, Zr, and Hf as well as a positive Nb relative to La and Th, relatively high 207Pb/206Pb and 208Pb/206Pb isotopic ratios, which are similar to the OIB-type incompatible elements, and EM2-type isotopic features likely related to a mantle plume. The SCS primary melts and volcanic glasses indicate that mantle peridotite cannot serve as the sole source composition; apparently, 18.5% eclogite and 46.1% garnet pyroxenite melts were involved. Large Ce/Yb and Th/Nb ratio variations and Nb, Ta, and Ti positive anomalies of the SCS volcanic glasses further support our conclusion that the 18.5% recycled oceanic crust in the form of dense eclogite from the stagnant subduction slab contributed to the source of the SCS. Thus, the SCS source composition probably was garnet pyroxenite produced by the reaction between peridotite melt and recycled oceanic crust. The existence of Hainan mantle plume and subducting recycled oceanic slab beneath the SCS spreading ridge is further supported by our three-dimensional P-wave tomographic model.
A. Appendix A.
The detailed initial MELTS_Excel calculation results of the system with the pressure of 3.2 kbar and the oxygen fugacity of quartz-fayalite-magnetite (QFM) buffer.
B. Appendix B.
The MELTS_Excel calculation results of the system with the pressure of 3.2 kbar and the oxygen fugacity of nickel-nickel oxide (NNO) buffer.
C. Appendix C.
Estimated primary melt compositions for Site 1431, Site 1433, and Site 1434 volcanic rocks from the South China Sea.
All data in this study are analyzed by the authors. Associated tests were performed at the Center for Global Tectonics, School of Earth Sciences, China University of Geosciences (Wuhan); the Wuhan SampleSolution Analytical Technology Co., Ltd., Wuhan, China; and the State key Laboratory of Isotope Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIG-CAS). We guarantee the authenticity and unrepeated use of data.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
This work was supported by the program for Guangdong Introducing Innovative and Entrepreneurial Teams, Southern Marine Science and Engineering Guangdong Laboratory (311020018), Zhujiang Talent Project Foundation of Guangdong Province (grant no.: 2017ZT07Z066), Major Projects of the National Natural Science Foundation of China (41590863), Fundamental Research Funds for Young Teacher Development Project (32110-31610351), and Double First-Class Guidance Project (32110-18841213). Wang, Z.W., is supported by the Key Special Project for Introduced Talents Team of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou) (GML2019ZD0203) and National Natural Science Foundation of China (No. 41790465, 41904045). Hou, T., is supported by the Fundamental Research Funds for the Central Universities (2652015054) and National Natural Science Foundation of China (grant no.: 41922912, 41761134086, 42150008, and 2016YFC0600502).