Lithogeochemistry of the Mid-Ocean Ridge Basalts near the Fossil Ridge of the Southwest Sub-Basin, South China Sea

: Mid-ocean ridge basalts (MORB) in the South China Sea (SCS) record deep crust-mantle processes during seaﬂoor spreading. We conducted a petrological and geochemical study on the MORBs obtained from the southwest sub-basin of the SCS at site U1433 and U1434 of the International Ocean Discovery Program (IODP) Expedition 349. Results show that MORBs at IODP site U1433 and U1434 are una ﬀ ected by seawater alteration, and all U1433 and the bulk of U1434 rocks belong to the sub-alkaline low-potassium tholeiitic basalt series. Samples collected from site U1433 and U1434 are enriched mid-ocean ridge basalts (E-MORBs), and the U1434 basalts are more enriched in incompatible elements than the U1433 samples. The SCS MORBs have mainly undergone the fractional crystallization of olivine, accompanied by the relatively weak fractional crystallization of plagioclase and clinopyroxene during magma evolution. The magma of both sites might be mainly produced by the high-degree partial melting of spinel peridotite at low pressures. The degree of partial melting at site U1434 was lower than at U1433, ascribed to the relatively lower spreading rate. The magmatic source of the southwest sub-basin basalts may be contaminated by lower continental crust and contributed by recycled oceanic crust component during the opening of the SCS. sub-basin of normal (N)-MORB-type and (E)-MORB-type basalts. modeling the of mantle continental on the in the the


Introduction
The mid-ocean ridge marks the spreading center of Earth's plate tectonics. Due to the depressurization and subsequent partial melting of the ascending mantle materials, basaltic magma emanates from the mid-ocean ridge. The magma is then cooled by seawater to form mid-ocean ridge basalt (MORB). MORB carries important information about the composition of the upper mantle [1]. In addition, basaltic magma may experience magmatic mixing and fractional crystallization during ascending and migration, information of which is also recorded in the MORBs [2][3][4]. Therefore, MORBs are key to understanding the magma evolution in spreading centers [5]. The South China Sea (SCS) is one of the most important marginal seas in the Western Pacific. Recovering MORBs from the SCS was a challenge [6][7][8][9] because thick sediments cover the seafloor of the SCS basin [10]. In 2014, the International Ocean Discovery Program (IODP) Expedition 349 drilled through the overlying sediment and successfully retrieved SCS MORBs from the site U1431 (eastern sub-basin), U1433 and U1434 (southwest sub-basin) for the first time [11].
These SCS MORBs offered scientists a great opportunity to study the geochemical nature of the sub-ridge mantle and the evolution of the SCS. Most research focused on the MORBs recovered from the Site U1431 located at the east sub-basin of the SCS [12,13]. Zhang et al. [14] analyzed the major elements, trace elements and Sr-Nd-Pb-Hf isotopes of U1431 basalts, and figured that the east sub-basin consists of both normal (N)-MORB-type and enriched (E)-MORB-type basalts. Through modeling the influences of Hainan mantle plume and lower continental crust based on Sr-Nd-Pb-Hf isotope compositions, the origin of Indian-type basalts in the East sub-basin can be explained by the involvement of Hainan plume in the sub-ridge mantle during seafloor spreading. A lithogeochemical study on the U1431 basalts [15] revealed that an eclogite-/pyroxenite-rich component, possibly originated from the Hainan hotspot, played a fundamental role in the transition from spreading to intraplate volcanism of the dying spreading ridge of the SCS. In addition, the Al-in-olivine thermometer [16] demonstrated that anomalously high crystallization temperature of primitive olivine in the SCS MORB, corresponding to the high mantle potential temperature, is consistent with the plume-ridge interaction in the SCS.
Few studies investigated the MORBs from the site U1433 and U1434. Both U1433 and U1434 situate near the fossil spreading center of the Southwest Sub-Basin. U1434 is about 40 km to the northwest of U1433 (Figure 1). The MORBs of both sites are formed in the late stage of the seafloor spreading of the SCS. The age of the U1434 (~16.3Ma) MORB is slightly younger than the age of the U1433 MORB (~17.3 Ma) [12]. Both site U1433 and U1434 MORBs in the Southwest sub-basin are enriched (E)-MORB type basalts based on trace element compositions and are Indian-type MORBs based on Sr-Nd-Pb-Hf isotope compositions [14]. The different trace element patterns, La/Sm ratios and Sr-Nd-Pb-Hf isotopes indicated geochemical changes of magma before the cessation of seafloor spreading [14]. Based on Sr-Nd-Pb-Hf isotope compositions, Zhang et al. [14] suggested that the sub-ridge mantle of the southwest sub-basin was influenced by the Hainan plume and contaminated by lower continental crust during the continent rifting and opening of the South China Sea. Nevertheless, a comprehensive understanding of the seawater alteration, fractional crystallization, and petrological and mineralogical characteristics of the SCS basalts still requires in-depth investigation.
Minerals 2020, 10, x FOR PEER REVIEW 2 of 18 sub-basin consists of both normal (N)-MORB-type and enriched (E)-MORB-type basalts. Through modeling the influences of Hainan mantle plume and lower continental crust based on Sr-Nd-Pb-Hf isotope compositions, the origin of Indian-type basalts in the East sub-basin can be explained by the involvement of Hainan plume in the sub-ridge mantle during seafloor spreading. A lithogeochemical study on the U1431 basalts [15] revealed that an eclogite-/pyroxenite-rich component, possibly originated from the Hainan hotspot, played a fundamental role in the transition from spreading to intraplate volcanism of the dying spreading ridge of the SCS. In addition, the Al-in-olivine thermometer [16] demonstrated that anomalously high crystallization temperature of primitive olivine in the SCS MORB, corresponding to the high mantle potential temperature, is consistent with the plume-ridge interaction in the SCS. Few studies investigated the MORBs from the site U1433 and U1434. Both U1433 and U1434 situate near the fossil spreading center of the Southwest Sub-Basin. U1434 is about 40 km to the northwest of U1433 (Figure 1). The MORBs of both sites are formed in the late stage of the seafloor spreading of the SCS. The age of the U1434 (~16.3Ma) MORB is slightly younger than the age of the U1433 MORB (~17.3 Ma) [12]. Both site U1433 and U1434 MORBs in the Southwest sub-basin are enriched (E)-MORB type basalts based on trace element compositions and are Indian-type MORBs based on Sr-Nd-Pb-Hf isotope compositions [14]. The different trace element patterns, La/Sm ratios and Sr-Nd-Pb-Hf isotopes indicated geochemical changes of magma before the cessation of seafloor spreading [14]. Based on Sr-Nd-Pb-Hf isotope compositions, Zhang et al. [14] suggested that the sub-ridge mantle of the southwest sub-basin was influenced by the Hainan plume and contaminated by lower continental crust during the continent rifting and opening of the South China Sea. Nevertheless, a comprehensive understanding of the seawater alteration, fractional crystallization, and petrological and mineralogical characteristics of the SCS basalts still requires in-depth investigation.   [17,18].
In this study, we determined the mineral phases, chemical compositions, and optical microscopy of U1433 and U1434 MORBs using petrographic and microscopic observation, X-ray diffraction determination, and inductively coupled plasma mass spectrometry (ICP-MS) analysis. We seek to systematically study the magma formation and evolution processes in the southwest sub-basin of the SCS, including seawater alteration, fractional crystallization, mantle melting, and magmatic source, and also the possible influence of the Hainan plume to the magmatic activity of the mid-ocean ridge.

Geological Setting
The SCS is one of the world's largest marginal seas. It is located at the junction among the Indo-Australian plate, the Pacific plate, and the Eurasian plate ( Figure 1) [19]. Despite its relatively young age and small size, the SCS experienced almost the entire Wilson cycle, from continental breakup to seafloor spreading and followed by subduction [11,20]. The eastern margin of the SCS is bounded by the Manila Trench, and the western margin is marked by the Red River fault zone [21,22]. The SCS basin can be divided into three sub-basins based on structural variations: the northwest sub-basin, the east sub-basin, and the southwest sub-basin [23]. The Zhongnan fault extending from north to south ( Figure 1) separates the SCS into an east sub-basin and a southwest sub-basin [24]. Seafloor magnetic anomaly data indicated that the spreading time was 33-15.5 Ma at the east sub-basin and 24-16 Ma at the southwest sub-basin [24]. During the Cenozoic, the SCS underwent three episodes of seafloor spreading, and left three fossil spreading centers within the basin: the first one is oriented E-W in the northwest (NW) sub-basin (near the present 18 • N); the second one is oriented E-W in the east sub-basin (near the present 17 • N); and the third one (near the present 13.5 • -15.5 • N) ( Figure 1) extending southwestward from the east sub-basin into the southwest (SW) sub-basin when the spreading finally stopped at~16 Ma [9].

Sampling
The IODP Expedition 349 drilled two sites (U1433 with 2 holes and U1434 with 1 hole) in the Southwest Sub-Basin of the SCS during January-March 2014 [11]. At site U1433, Hole U1433A was cored using the advanced piston corer (APC) to refusal at 188.3 mbsf (meter below seafloor), Hole U1433B was drilled to 186.1 mbsf and then cored using the rotary core barrel (RCB). At site U1434, Hole U1434A was drilled to 197.0 mbsf and then cored with the rotary core barrel (RCB). MORB, dominantly pillow and massive basalt flows, was recovered from the site U1433 at depths of 796.67-857.48 mbsf ( Figure 2). The age was determined as~17.3 Ma using Ar-Ar dating [12]. Pillow basalt interposed with hyaloclastite breccia occurred at a depth of 278.27-308.65 mbsf from the site U1434. Ar-Ar dating indicated an age of~16.3 Ma [12].
Basaltic samples of both sites were collected from the IODP sample repositories at Kochi core center (Japan), under the guidance of local curators. We obtained 11 samples from Hole U1433B and 6 samples from Hole U1434A (Figure 2). Most samples are visually fresh, and only several samples experienced slightly alteration. All the samples were preserved in polyethylene bags for subsequent analyses.

Analytical Methods
All the rock samples were one-side polished to make thin sections with a thickness of 0.03 mm. Petrographic and microscopic observation on the thin sections of the SCS basalts were conducted on a Leica DM4500 P LED Polarization Microscope at Ocean College, Zhejiang University.
The major and trace element compositions of the basaltic samples were determined by at the Solid Earth Geochemical Laboratory, Harvard University. The samples were dried and ground to about 200 mesh with an agate mortar. Then, about ~50 mg of each sample was digested in an HF:HNO3 mixture. The trace elemental concentrations were subsequently analyzed by Thermo's XSeries inductively coupled plasma mass spectrometry (ICP-MS). A matrix 8N HNO3 solution with Ge (10 ppb), In (3 ppb), Tm (3 ppb), and Bi (3 ppb) was used as the internal standard. The major elemental contents were determined by ICP-MS equipped with a laser ablation system, following the method of Chen and Langmuir [25]. The measurement uncertainty was less than 2% via tests of blank samples and external standards (BCR-2, BHVO-2, DNC-1, JB-2, W-2a, and VE-32), and duplicate analyses.
The crystal structures of minerals in the samples were determined by X-ray powder diffraction using a PANalytical X'Pert3 Powder XRD at Ocean College, Zhejiang University. The 2θ ranged from 5° to 70° with a step of 0.02° min −1 . The operation voltage and current were maintained at 40 kV and 40 mA, respectively. The crystalline phases were identified from the XRD patterns using MDI Jade (version 6.5) software.

Petrographic and Mineralogical Characteristics
All samples have a porphyritic texture and the phenocrysts are mainly plagioclase with a grain size of 0.4 to 1.2 mm. Most of them are mostly subhedral with short columns and tabular in shape. The phenocrysts also contain a small number of olivine and most of them are anhedrals with grain sizes of 0.05 to 0.4 mm. Clinopyroxene is rare. The matrix of samples shows intergranular and intersertal texture. It consists mainly of plagioclase and clinopyroxene, most of which are needle-like

Analytical Methods
All the rock samples were one-side polished to make thin sections with a thickness of 0.03 mm. Petrographic and microscopic observation on the thin sections of the SCS basalts were conducted on a Leica DM4500 P LED Polarization Microscope at Ocean College, Zhejiang University.
The major and trace element compositions of the basaltic samples were determined by at the Solid Earth Geochemical Laboratory, Harvard University. The samples were dried and ground to about 200 mesh with an agate mortar. Then, about~50 mg of each sample was digested in an HF:HNO 3 mixture. The trace elemental concentrations were subsequently analyzed by Thermo's XSeries inductively coupled plasma mass spectrometry (ICP-MS). A matrix 8N HNO 3 solution with Ge (10 ppb), In (3 ppb), Tm (3 ppb), and Bi (3 ppb) was used as the internal standard. The major elemental contents were determined by ICP-MS equipped with a laser ablation system, following the method of Chen and Langmuir [25]. The measurement uncertainty was less than 2% via tests of blank samples and external standards (BCR-2, BHVO-2, DNC-1, JB-2, W-2a, and VE-32), and duplicate analyses.
The crystal structures of minerals in the samples were determined by X-ray powder diffraction using a PANalytical X'Pert3 Powder XRD at Ocean College, Zhejiang University. The 2θ ranged from 5 • to 70 • with a step of 0.02 • min −1 . The operation voltage and current were maintained at 40 kV and 40 mA, respectively. The crystalline phases were identified from the XRD patterns using MDI Jade (version 6.5) software.

Petrographic and Mineralogical Characteristics
All samples have a porphyritic texture and the phenocrysts are mainly plagioclase with a grain size of 0.4 to 1.2 mm. Most of them are mostly subhedral with short columns and tabular in shape. The phenocrysts also contain a small number of olivine and most of them are anhedrals with grain sizes of 0.05 to 0.4 mm. Clinopyroxene is rare. The matrix of samples shows intergranular and intersertal texture. It consists mainly of plagioclase and clinopyroxene, most of which are needle-like and slender columnar with disordered distribution, and the gap is filled with dark cryptocrystalline (Figure 3a-d). All samples are fresh, and the vesicular structure is not developed.
Minerals 2020, 10, x FOR PEER REVIEW 5 of 18 and slender columnar with disordered distribution, and the gap is filled with dark cryptocrystalline (Figure 3a-d). All samples are fresh, and the vesicular structure is not developed. XRD patterns are used to determine the minerals of the rock samples [26]. The crystal structures of the basaltic samples are indexed by the XRD patterns ( Figure 4). The crystalline minerals are dominated by labradorite, andesine, diopside, and forsterite in the U1433 basalt (Figure 4a  XRD patterns are used to determine the minerals of the rock samples [26]. The crystal structures of the basaltic samples are indexed by the XRD patterns ( Figure 4). The crystalline minerals are dominated by labradorite, andesine, diopside, and forsterite in the U1433 basalt (Figure 4a

Major Elements
The SiO2 contents of U1433 basalt fall in the range of 48.29% to 52.09% with an average value of 50.17%. The U1434 basalt exhibits slightly higher SiO2 concentrations of 49.46%-52.10% (average 51.03%). The MgO contents of U1433 and U1434 basalts are 5.03%-7.65% and 5.14%-7.20%, respectively. The samples from both the site U1433 and U1434 exhibit relatively high CaO and Al2O3 contents, with values of 10.13%-12.46% and 14.12%-18.80%, respectively. According to the major and trace elemental concentrations of basaltic samples, the samples are classified as basalt and basaltic andesite (Figure 5a), belong to the sub-alkaline series. Based on the K2O-SiO2 diagram (Figure 5b), all the samples of Site U1433 can be further determined as low-K tholeiitic series, while the bulk of the samples of Site U1434 plot in the fields of the low-K tholeiitic rock series, and a small number plot within the field for the calc-alkaline series.

Trace Elements
The trace elemental compositions of basaltic samples from U1433 and U1434 are summarized in Table 1. The primitive mantle-normalized trace-element distribution pattern ( Figure 6) is generally

Major Elements
The SiO2 contents of U1433 basalt fall in the range of 48.29% to 52.09% with an average value of 50.17%. The U1434 basalt exhibits slightly higher SiO2 concentrations of 49.46%-52.10% (average 51.03%). The MgO contents of U1433 and U1434 basalts are 5.03%-7.65% and 5.14%-7.20%, respectively. The samples from both the site U1433 and U1434 exhibit relatively high CaO and Al2O3 contents, with values of 10.13%-12.46% and 14.12%-18.80%, respectively. According to the major and trace elemental concentrations of basaltic samples, the samples are classified as basalt and basaltic andesite (Figure 5a), belong to the sub-alkaline series. Based on the K2O-SiO2 diagram (Figure 5b), all the samples of Site U1433 can be further determined as low-K tholeiitic series, while the bulk of the samples of Site U1434 plot in the fields of the low-K tholeiitic rock series, and a small number plot within the field for the calc-alkaline series.

Trace Elements
The trace elemental compositions of basaltic samples from U1433 and U1434 are summarized in Table 1. The primitive mantle-normalized trace-element distribution pattern ( Figure 6) is generally

Trace Elements
The trace elemental compositions of basaltic samples from U1433 and U1434 are summarized in Table 1. The primitive mantle-normalized trace-element distribution pattern ( Figure 6) is generally similar to that of E-MORB. Several samples presented positive anomalies of U and Sr.  [27]. Based on the La/Nb ratios, oceanic basalts can be divided into N-MORB with La/Nb > 1.0, E-MORB with La/Nb < 1.0, and HIMU Oceanic Island Basalt (OIB) with La/Nb of between 0.6 and 0.7 [27]. The La/Nb ratios of basaltic samples of U1433 and U1434 range from 0.73 to 1.01 (average 0.82) and from 0.71 to 0.86 (average 0.78), respectively. Combined with the primitive mantle-normalized trace-element distribution patterns, it is suggested that both the U1433 and U1434 basalts belong to E-MORB.

Seawater Alteration
In this paper, our purpose was to systematically study the magma formation and evolution processes in the southwest sub-basin of the SCS, and also the possible influence of the Hainan plume on the magmatic activity of the mid-ocean ridge. The fresh basalt samples were necessary. If the samples were not modified by seawater alteration, then they can be used to constrain the mantle properties and deep geodynamic processes [28]. In the primitive mantle-normalized trace-element diagrams (Figure 6), some of the samples in both sites show positive U anomalies. Uranium is a water-soluble element, whereas Th is a water-insoluble element [29], so the interaction of seawater with basalt will cause the altered sample enriched in U, but has little effect on Th. When the seawater interacts with the basalt, the Th/U ratio will deviate from the chondrite line. In the Th-U diagram (Figure 7), we also added the unaltered MORB glass data [30]. It shows that the Th/U ratio of most samples at site U1433 and U1434 does not deviate from the chondrite line, and only a few samples deviate from the chondrite line. Those samples also have positive U anomalies in the primitive mantle-normalized trace-element distribution diagram ( Figure 6). In addition, no significant Ce anomalies were observed in our samples, which suggests that the effect of subsea weathering on basaltic samples can be negligible [1]; therefore, we excluded that these samples were significantly affected by seawater alteration, and are all fresh. We used them to constrain the magma formation and evolution processes and mantle properties, discussed in the following part.

Seawater Alteration
In this paper, our purpose was to systematically study the magma formation and evolution processes in the southwest sub-basin of the SCS, and also the possible influence of the Hainan plume on the magmatic activity of the mid-ocean ridge. The fresh basalt samples were necessary. If the samples were not modified by seawater alteration, then they can be used to constrain the mantle properties and deep geodynamic processes [28]. In the primitive mantle-normalized trace-element diagrams (Figure 6), some of the samples in both sites show positive U anomalies. Uranium is a water-soluble element, whereas Th is a water-insoluble element [29], so the interaction of seawater with basalt will cause the altered sample enriched in U, but has little effect on Th. When the seawater interacts with the basalt, the Th/U ratio will deviate from the chondrite line. In the Th-U diagram (Figure 7), we also added the unaltered MORB glass data [30]. It shows that the Th/U ratio of most samples at site U1433 and U1434 does not deviate from the chondrite line, and only a few samples deviate from the chondrite line. Those samples also have positive U anomalies in the primitive mantle-normalized trace-element distribution diagram ( Figure 6). In addition, no significant Ce anomalies were observed in our samples, which suggests that the effect of subsea weathering on basaltic samples can be negligible [1]; therefore, we excluded that these samples were significantly affected by seawater alteration, and are all fresh. We used them to constrain the magma formation and evolution processes and mantle properties, discussed in the following part.

Fractional Crystallization
The basaltic primary magma would undergo fractional crystallization under different temperature and pressure conditions during its ascent and migration after its formation, which is an important part of the magma evolution process [31]. The magnesium number (Mg# = (Mg 2+ /(Mg 2+ + Fe 2+ ))) can be used to trace the fractional crystallization of mid-ocean ridge basalts [32]. The Mg# of our samples ranges from 51 to 62 (average 57) and 49 to 59 (average 57), respectively. This suggests that the primitive magma of the mid-ocean ridge basalts at the two sites experienced fractional crystallization during their upward migration [33]. There is a good linear relationship between MgO and other oxides. For example, SiO2, K2O, Ni, and Cr decreased with decreasing MgO, whereas the CaO, Al2O3, Na2O, TiO2, TFeO, Sr, and Y increased; however, the CaO/Al2O3 ratios remained

Fractional Crystallization
The basaltic primary magma would undergo fractional crystallization under different temperature and pressure conditions during its ascent and migration after its formation, which is an important part of the magma evolution process [31]. The magnesium number (Mg# = (Mg 2+ /(Mg 2+ + Fe 2+ ))) can be used to trace the fractional crystallization of mid-ocean ridge basalts [32]. The Mg# of our samples ranges from 51 to 62 (average 57) and 49 to 59 (average 57), respectively. This suggests that the primitive magma of the mid-ocean ridge basalts at the two sites experienced fractional crystallization during their upward migration [33]. There is a good linear relationship between MgO and other oxides. For example, SiO 2 , K 2 O, Ni, and Cr decreased with decreasing MgO, whereas the CaO, Al 2 O 3 , Na 2 O, TiO 2 , TFeO, Sr, and Y increased; however, the CaO/Al 2 O 3 ratios remained unchanged. In Figure 8, these typical variation trend characteristics may be due to the fractional crystallization of olivine, plagioclase, and pyroxene during magma evolution [34]. In this study, no obvious Eu anomalies were observed, indicating that the crystallization of plagioclase is insignificant [33]. The variation between TFeO, CaO, CaO/Al 2 O 3, and MgO can be explained by the fractional crystallization of pyroxene during magma evolution [34]; however, as shown in Figure 8, the clinopyroxene may not be the main fractional phase. The primary magma of site U1433 and U1434 have mainly undergone the fractional crystallization of olivine, accompanied by the relatively weak fractional crystallization of plagioclase and clinopyroxene.
Minerals 2020, 10, x FOR PEER REVIEW 11 of 18 unchanged. In Figure 8, these typical variation trend characteristics may be due to the fractional crystallization of olivine, plagioclase, and pyroxene during magma evolution [34]. In this study, no obvious Eu anomalies were observed, indicating that the crystallization of plagioclase is insignificant [33]. The variation between TFeO, CaO, CaO/Al2O3, and MgO can be explained by the fractional crystallization of pyroxene during magma evolution [34]; however, as shown in Figure 8, the clinopyroxene may not be the main fractional phase. The primary magma of site U1433 and U1434 have mainly undergone the fractional crystallization of olivine, accompanied by the relatively weak fractional crystallization of plagioclase and clinopyroxene.

Constrain the Source Mantle
The lherzolites, which make up the mantle, will partially melt under the decompression caused by the expansion of the mid-ocean ridge, thus forming magma [31,33]. The types of lherzolites vary with the depth of mantle. In general, the upper mantle, from the shallow to the deep, is the plagioclase peridotite region (the mantle depth < 30 km), the spinel peridotite region (the mantle depth 30-80 km) and the garnet peridotite region (the mantle depth > 80 km) [33]. As is seen in the chondrite normalized (Tb/Yb)N-(La/Sm)N diagram (Figure 9a), the (Tb/Yb)N values of the basaltic samples at the two sites in this study are all less than 1.8, all of which fall within the spinel peridotite region, indicating that the magma melting depth of the two sites is less than 80 km. The Nb/Y-SiO2 diagram can be used to study the melting conditions of the magma source of the mid-ocean ridge basalt [35]. In Figure 9b, all samples plot into the low-pressure and highly partially melted tholeiitic

Constrain the Source Mantle
The lherzolites, which make up the mantle, will partially melt under the decompression caused by the expansion of the mid-ocean ridge, thus forming magma [31,33]. The types of lherzolites vary with the depth of mantle. In general, the upper mantle, from the shallow to the deep, is the plagioclase peridotite region (the mantle depth < 30 km), the spinel peridotite region (the mantle depth 30-80 km) and the garnet peridotite region (the mantle depth > 80 km) [33]. As is seen in the chondrite normalized (Tb/Yb) N -(La/Sm) N diagram (Figure 9a), the (Tb/Yb) N values of the basaltic samples at the two sites in this study are all less than 1.8, all of which fall within the spinel peridotite region, indicating that the magma melting depth of the two sites is less than 80 km. The Nb/Y-SiO 2 diagram can be used to study the melting conditions of the magma source of the mid-ocean ridge basalt [35]. In Figure 9b, all samples plot into the low-pressure and highly partially melted tholeiitic magma area, indicating that the magma of the two sites is formed by high partial melting. Thus, the basaltic magmas at site U1433 and U1434 in the southwest sub-basin of the SCS might be mainly formed by low-pressure partial melting of spinel lherzolite, with the majority of melting occurred shallower than the garnet stability field. magma area, indicating that the magma of the two sites is formed by high partial melting. Thus, the basaltic magmas at site U1433 and U1434 in the southwest sub-basin of the SCS might be mainly formed by low-pressure partial melting of spinel lherzolite, with the majority of melting occurred shallower than the garnet stability field. Figure 9. Diagrams of (Tb/Yb)N versus (La/Sm)N: (a) (after ref. [36]) and Nb/Y versus SiO2; (b) (after ref. [35]) for basaltic samples at Sites U1433 and U1434. Red dashed line in (a) represents the conversion line, and it corresponds to a depth of 80 km.

Mantle Melting
Basaltic magma is formed by the partial melting of mantle lherzolite, and the mantle melting has a significant effect on the geochemical composition of basalt [31]. Previous studies have shown that the change range of compatible elements concentration is small, whereas the ratios of strong incompatible elements to weak incompatible elements will increase with the increase of the partial melting degree [37]. With the crystallization degree increasing, the concentration of compatible elements in the melt will decrease sharply; whereas those of incompatible elements increased, and the increase is not related to the incompatibility of the elements, thus the ratios of the strong incompatible elements/weak incompatible elements does not change significantly [37]. The compatible elements (Ni, Cr, and Co) concentration of site U1433 samples vary greatly, and the ratio of incompatible elements (Sr/Ba) varies indistinctively, which means those samples are related by fractional crystallization, but also influenced by different degree of partial melting. The compatible elements concentration of Site U1434 samples varies greatly, whereas the ratio of incompatible elements remain almost constant, which reflects the magma of Site U1434 is formed by fractional crystallization. In addition, the characteristic parameter values of rare earth elements can also be used to constrain the genesis process of rock [33]. According to the La/Sm-La diagram, it can be used to discriminated whether the magma is formed by partial melting or fractional crystallization. During partial melting, the La/Sm and La of the melt increased, whereas, in the process of fractional crystallization, the La/Sm remains constant, but La increased [33]. Figure 10a shows a non-significant positive correlation for most Site U1433 samples, and near-horizontal distribution trend for Site U1434 samples, also a non-significant positive correlation between two sites. The results mean that the diversity of U1433 basalt components is due to fractional crystallization and difference partial melting degree, and those of Site U1434 is due to fractional crystallization only. Combining with the ΣREE-MgO diagram (Figure 10b), the ΣREE value of different sites shows a significant difference, that is, the ΣREE value of site U1434 samples is generally higher than site U1433, suggesting partial melting may be the dominate process during the magma evolution of our basalts, and there is a difference in melt fraction between the two sites.

Mantle Melting
Basaltic magma is formed by the partial melting of mantle lherzolite, and the mantle melting has a significant effect on the geochemical composition of basalt [31]. Previous studies have shown that the change range of compatible elements concentration is small, whereas the ratios of strong incompatible elements to weak incompatible elements will increase with the increase of the partial melting degree [37]. With the crystallization degree increasing, the concentration of compatible elements in the melt will decrease sharply; whereas those of incompatible elements increased, and the increase is not related to the incompatibility of the elements, thus the ratios of the strong incompatible elements/weak incompatible elements does not change significantly [37]. The compatible elements (Ni, Cr, and Co) concentration of site U1433 samples vary greatly, and the ratio of incompatible elements (Sr/Ba) varies indistinctively, which means those samples are related by fractional crystallization, but also influenced by different degree of partial melting. The compatible elements concentration of Site U1434 samples varies greatly, whereas the ratio of incompatible elements remain almost constant, which reflects the magma of Site U1434 is formed by fractional crystallization. In addition, the characteristic parameter values of rare earth elements can also be used to constrain the genesis process of rock [33]. According to the La/Sm-La diagram, it can be used to discriminated whether the magma is formed by partial melting or fractional crystallization. During partial melting, the La/Sm and La of the melt increased, whereas, in the process of fractional crystallization, the La/Sm remains constant, but La increased [33]. Figure 10a shows a non-significant positive correlation for most Site U1433 samples, and near-horizontal distribution trend for Site U1434 samples, also a non-significant positive correlation between two sites. The results mean that the diversity of U1433 basalt components is due to fractional crystallization and difference partial melting degree, and those of Site U1434 is due to fractional crystallization only. Combining with the ΣREE-MgO diagram (Figure 10b), the ΣREE value of different sites shows a significant difference, that is, the ΣREE value of site U1434 samples is generally higher than site U1433, suggesting partial melting may be the dominate process during the magma evolution of our basalts, and there is a difference in melt fraction between the two sites. In order to clarify the difference in melt fraction between two sites, we quantitatively calculated the partial melting degree (F) of magma. By summarizing the experimental results of a large number of MORBs, Niu and Batiza [38] obtained an empirical formula to correct the sample composition: where Oxide (8) is the value when the oxide is corrected to MgO = 8%, Oxide(data) is the measured content value of the oxide (except MgO), mn is the empirical coefficient, n is the regression number. The empirical coefficients were corrected by Niu et al. [39] ( Table 2). The data of this empirical formula are all from MORBs, and it has a good effect on the study of MORBs [31]. According to the relationship among Na8, Ca8, and Al8, Niu and Batiza [38] put forward an empirical formula for quantitative calculation of partial melting degree (F): F(wt.) = 19.202 − 5.175 × Na8 + 15.537 × Ca8/Al8. According to the above formula, the partial melting degree of site U1433 and U1434 is 14.06%-19.40% and 16.35%-18.47%, respectively. In general, the partial melting degree of site U1433 is slightly higher than that of Site U1434.
Considering Figure 10 and the Sr-Nd-Pb-Hf isotopes characters of Sites U1433 and U1434 of Zhang et al. [14], we suggest that the geochemical composition of basalts between Sites U1433 and U1434 is mainly controlled by different degrees of partial melting. Research has shown that a spreading ridge would decrease in the degree of melting before the cessation of spreading because of decreasing spreading rate [40,41]. The partial melting degree of the magma source area of Site 1434 is lower than that of Site U1433; therefore, we proposed that as the ridge ceased in spreading, the mid-ocean ridge of the southwest sub-basin of the SCS has decreased in mantle melting degree. In order to clarify the difference in melt fraction between two sites, we quantitatively calculated the partial melting degree (F) of magma. By summarizing the experimental results of a large number of MORBs, Niu and Batiza [38] obtained an empirical formula to correct the sample composition: where Oxide (8) is the value when the oxide is corrected to MgO = 8%, Oxide (data) is the measured content value of the oxide (except MgO), m n is the empirical coefficient, n is the regression number. The empirical coefficients were corrected by Niu et al. [39] ( Table 2). The data of this empirical formula are all from MORBs, and it has a good effect on the study of MORBs [31]. According to the relationship among Na8, Ca8, and Al8, Niu and Batiza [38] put forward an empirical formula for quantitative calculation of partial melting degree (F): According to the above formula, the partial melting degree of site U1433 and U1434 is 14.06%-19.40% and 16.35%-18.47%, respectively. In general, the partial melting degree of site U1433 is slightly higher than that of Site U1434.
Considering Figure 10 and the Sr-Nd-Pb-Hf isotopes characters of Sites U1433 and U1434 of Zhang et al. [14], we suggest that the geochemical composition of basalts between Sites U1433 and U1434 is mainly controlled by different degrees of partial melting. Research has shown that a spreading ridge would decrease in the degree of melting before the cessation of spreading because of decreasing spreading rate [40,41]. The partial melting degree of the magma source area of Site 1434 is lower than that of Site U1433; therefore, we proposed that as the ridge ceased in spreading, the mid-ocean ridge of the southwest sub-basin of the SCS has decreased in mantle melting degree.

Magmatic Source
The SCS surrounded by subduction zones is an oceanic basin formed by the splitting of the eastern Eurasian continent, and there may be a Hainan plume beneath Leizhou-Hainan region which might influence the mid-ocean ridge magmatic activity (Figure 1). Given its relatively young age and small size, the magmatic source of the MORBs in the SCS may contain these complex tectonic information. Based on all the discussion above, we suggest that the MORBs at Sites U1433 and U1434 have undergone fractional crystallization and cannot be directly used to constrain the characteristics of the magma source; however, some trace element ratios of basalt can still reflect the material composition characteristics of its source area [42][43][44][45]. For example, the average Nb/Ta and Zr/Hf ratios of Site U1433 are 16.15 and 39.74, respectively, and of site, U1434 are 16.38 and 39.68, respectively. Those samples are similar to the primitive mantle, whose Nb/Ta and Zr/Hf values are (17.5 ± 0.5) and 36.27 [33], respectively. The Ce/Pb ratio is an effective tool for studying geochemical properties of basalt source areas [31]. The average Ce/Pb ratios of Sites U1433 and U1434 are 18.55 and 21.16, respectively, and are between the Ce/Pb values of the continental crust (~4) and the typical MORB and OIB (25 ± 5) [45]. The average Nb/U ratios of Sites U1433 and U1434 are 37.76 and 43.38, respectively, slightly lower than the Nb/U values of MORB and OIB (47 ± 10), and even some samples are lower than the Nb/U values of the primitive mantle (~30). The basaltic samples of the two sites are fresh, and the influence of seawater alteration to U anomalies is insignificant, thus the Nb/U values characteristics can suggest that the source area may contain enriched components (may be crust material) [43,44].
Basalt Nb/Th and Ta/U values can be used to reveal the origin of the mantle, and this is because the Nb/Th and Ta/U ratios were not sensitive to fractional crystallization, whereas the relatively water-soluble Th and U in the descending oceanic lithosphere are carried away by a series of subduction dehydration, leaving relatively water-insoluble Nb and Ta [46]; therefore, when the residual lithosphere is added to the ocean basalt source, it will lead to the enrichment of Nb and Ta in these basalts, thereby increasing the Nb/Th and Ta/U values. It can be seen in Figure 11a that the linear correlation between the Nb/Th and Th values of Sites U1433 and U1434 samples shows a certain negative correlation. At the same time, the Nb/Th values of the two sites samples are higher than those of the primitive mantle, and closer to that of E-MORB, indicating that the magma source composition of the samples is likely to be affected by the enriched components, which is consistent with the conclusion drawn from the study of Nb/U values of the two sites. In Figure 11b, the distribution trend of the sample points of the two sites is relatively closer to the lower continental crust (LCC), suggesting that the enrichment components in magma source of the MORBs at the two sites are probably containing the lower crust. In addition, because the partition coefficients of the four elements Nb, Th, Ta, and U show the characteristics of D Nb ≈ D Th < D Ta ≈ D U , when the residual lithosphere is added to the oceanic basalt source, it will lead to excess Nb* and Ta*, so the Nb* and Ta* values of basalts can also be used to reveal the magma source characteristics of basalts [43]. The values of Nb* and Ta* of samples at Site U1433 range from 1.21 to 1.96 (average 1.56) and from 0.45 to 1.73 (average 1.20), respectively, and the values of Nb* and Ta* of samples at Site U1434 range from 1.36 to 1.50 (average 1.40) and from 0.61 to 1.67 (average 1.36), respectively. That is to say, the values of Nb* of all samples at both sites are greater than 1, and the values of Ta* of most samples at the two sites are greater than 1, indicating that the magma source of the mid-ocean ridge basalts at Sites U1433 and U1434 may be influenced by the enriched components of the recycled lithosphere. The magmatic source of both two sites may contain enriched components of the lower crust and/or the recycled lithosphere. It is necessary to distinguish whether these enriched components are due to high-level crustal contamination, and/or a source feature. Zhang et al. [14] simulated the influence of the Hainan mantle plume and the continental lower crust on the composition of the depleted upper mantle, and found that the composition of the mantle in the southwestern sub-basin of the SCS (such as the characteristics of depleted Pb isotopes) reflects the existence of the continental lower crust mixing in the SCS mantle during the process of SCS seafloor spreading, and propose that the Hainan plume might have not only contaminated the depleted mantle under the South China Sea, but also promoted the effective mixing of the lower continental crust into the asthenosphere; however, the study of Yang et al. [16] shows that the Hainan mantle plume might contain large amounts of recycled oceanic crust components, which are ultimately derived from the Indian and Pacific oceanic subduction. If so, the Hainan plume may take recycled oceanic crust components into the magmatic source of MORBs. Based on the previous research results, we suggest that the enriched components in the magmatic source of southwestern sub-basin during the late period of SCS spreading are the lower continental crust and the recycled oceanic crust. They are probably the result of the lower crust components mixed into the asthenosphere promoted by the Hainan plume during the seafloor spreading of the South China Sea and the recycled oceanic crust components formed by the Indian and Pacific oceanic subduction, respectively.

Conclusions
This paper studies the geochemical characteristics of the mid-ocean ridge basalts at Sites U1433 and U1434 of the International Ocean Discovery Program (IODP) Expedition 349 in the southwest sub-basin of the SCS, and obtained the following points: (1) All Site U1433 MORBs and the bulk of Site U1434 MORBs belong to the sub-alkaline low-potassium tholeiitic basalt series. MORBs at IODP Sites U1433 and U1434 are all E-MORB type basalts, but Site U1434 is more enriched with incompatible elements than Site U1433.
(2) MORBs at IODP site U1433 and U1434 are basically unaffected by seawater alteration, and the primary magma of both sites mainly undergone the fractional crystallization of olivine, accompanied by the relatively weak fractional crystallization of plagioclase and clinopyroxene. The different magma evolution between the two sites is mainly controlled by different degrees of partial melting at different spreading rates. The degree of melting of Site U1434 is lower than that of Site U1433. The magmatic source of both two sites may contain enriched components of the lower crust and/or the recycled lithosphere. It is necessary to distinguish whether these enriched components are due to high-level crustal contamination, and/or a source feature. Zhang et al. [14] simulated the influence of the Hainan mantle plume and the continental lower crust on the composition of the depleted upper mantle, and found that the composition of the mantle in the southwestern sub-basin of the SCS (such as the characteristics of depleted Pb isotopes) reflects the existence of the continental lower crust mixing in the SCS mantle during the process of SCS seafloor spreading, and propose that the Hainan plume might have not only contaminated the depleted mantle under the South China Sea, but also promoted the effective mixing of the lower continental crust into the asthenosphere; however, the study of Yang et al. [16] shows that the Hainan mantle plume might contain large amounts of recycled oceanic crust components, which are ultimately derived from the Indian and Pacific oceanic subduction. If so, the Hainan plume may take recycled oceanic crust components into the magmatic source of MORBs. Based on the previous research results, we suggest that the enriched components in the magmatic source of southwestern sub-basin during the late period of SCS spreading are the lower continental crust and the recycled oceanic crust. They are probably the result of the lower crust components mixed into the asthenosphere promoted by the Hainan plume during the seafloor spreading of the South China Sea and the recycled oceanic crust components formed by the Indian and Pacific oceanic subduction, respectively.

Conclusions
This paper studies the geochemical characteristics of the mid-ocean ridge basalts at Sites U1433 and U1434 of the International Ocean Discovery Program (IODP) Expedition 349 in the southwest sub-basin of the SCS, and obtained the following points: (1) All Site U1433 MORBs and the bulk of Site U1434 MORBs belong to the sub-alkaline low-potassium tholeiitic basalt series. MORBs at IODP Sites U1433 and U1434 are all E-MORB type basalts, but Site U1434 is more enriched with incompatible elements than Site U1433. (2) MORBs at IODP site U1433 and U1434 are basically unaffected by seawater alteration, and the primary magma of both sites mainly undergone the fractional crystallization of olivine, accompanied by the relatively weak fractional crystallization of plagioclase and clinopyroxene. The different magma evolution between the two sites is mainly controlled by different degrees of partial melting at different spreading rates. The degree of melting of Site U1434 is lower than that of Site U1433. (3) The trace element characteristics of the MORBs at IODP Sites U1433 and U1434 show that the basaltic magmas might be mainly formed by low-pressure and high-level partial melting of spinel lherzolite, with the majority of the melting occurred shallower than the garnet stability field. (4) MORBs at Sites U1433 and U1434 may involve magma sources containing crust materials, and the magma source of the southwest sub-basin basalts may have been contaminated by lower continental crust and contributed by recycled oceanic crust components during the opening of the SCS.