Gravity-Seismic Joint Inversion of Lithospheric Density Structure in the Qiongdongnan Basin, Northwest South China Sea

Qiongdongnan Basin (QDNB), located at the northwestern corner of the South China Sea (SCS), is a key juncture between the extensional tectonic regime in the northern continental margin and the shear tectonic regime in the western continental margin. Analyzing the crustal density structure and tracking the thermodynamic controlling factors are effective approaches to reveal the nonuniform breakup process of the northwestern SCS. Herein, focusing on the obvious tectonic deformation with distinct eastern and western parts in the QDNB, we present the crustal density structures of five profiles and identify the high-density anomaly related to the synrifting mantle underplating and postrifting magmatic intrusions. The crustal density model was constructed from the Bouguer gravity anomaly, ocean bottom seismic profiles, and multichannel seismic reflection profiles. The northern part of QDNB, with normal crustal density, lower surface heat flow of <55 mW/m2, and limited extension factor of 1.25–1.70, is recognized as the initial nonuniform extension continental crust. The mantle underplating beneath the QDNB is identified as a high mantle density of 3.30–3.40 g/cm3 and a high lower crustal density of 2.92–2.96 g/cm3, which is usually recognized by the high-velocity layers in the northeastern margin of SCS. The magmatic intrusions are identified as the high-density bodies ranging from 3.26 g/cm3 at the base to 2.64 g/cm3 at the top, which become stronger from the west to east. The central part of Xisha Trough is featured by the cooling of the heavily thinned lower crust in the final continental rifting stage, which is close to the cold and rigid oceanic crust. Lateral variations in the deep magmatic anomaly should be the crucial factor for the nonuniform breakup process in the northwestern margin of SCS.As the largest Cenozoic marginal basin located in the western Pacific region, the South China Sea (SCS) was formed in a complex tectonic setting due to the strong interaction among the Indo-Australian, Eurasian, and Pacific plates [1, 2] (Figure 1). Because of the stretching stress introduced by the rollback of the paleo-Pacific plate and the slab pull of the paleo-SCS, the SCS shows the ununiform continental rifting and progressive seafloor spreading from east to west, featuring highly inhomogeneous crustal structure and asymmetric magmatism beneath the marginal basins [3-5]. It is generally accepted that the SCS might represent a “plate-edge or Pacific-type” extensional basin, and the passive upwelling of the fertile asthenospheric mantle induced by the surrounding subductions could be primarily responsible for these inhomogeneous features [6, 7]. Some scholars even believed that the SCS had experienced a “magma-rich”-type breakup process in its middle-eastern part, but a “magma-poor” one is observed in the west [4, 8].In the case of the northern continental margin of SCS, the compositional or structural east–west heterogeneity has been observed by various geological and geophysical data, especially the prominent detachment faults, magmatic activity, middle crustal low-velocity layer, lower crustal high-velocity layers (HVLs), and partial serpentinization [1, 7, 9-11]. Compared with the large amount of Ocean Bottom Seismograph (OBS) profiles and the International Ocean Discovery Program (IODP) seven sites programmed in the middle-east part of the northern continental margin, there are only a few OBS profiles carried out in the western part with limited resolution and reliability, including the Qiongdongnan Basin (QDNB) [12-14]. In spite of the well constraints for the sedimentary cover and fracture systems provided by the multichannel seismic profiles of petroleum companies, the study on deep crustal composition is still controversial, especially the magmatism anomaly [15-18].As the typical “plated-edge and plated-interior” continental rifting model proposed from the new IODP research, the magmatic activity and lower crustal HVLs should form in a uniform tectonic environment controlled by the mantle upwelling, underplating, diking, and eruption [6, 7, 11, 19]. Furthermore, more detailed research identified the QDNB in the middle-western continental margin as the “magma-intermediate”-type continental margin with a considerable scale synrifting magma activity [20]. The lower crustal HVLs from OBS profiles are generally considered the most direct and important approaches to estimate the intensity of synrifting magmatic underplating [21-23]. However, probably limited by a small number of OBS profiles, there is still no conclusive evidence to support the lower crustal HVL beneath the QDNB for the present [13, 14, 24]. However, probably limited by a small number of OBS profiles, there is still no conclusive evidence that has yet been obtained to support the lower crustal HVL beneath the QDNB for the present, which could be attributed to the small number of OBS profiles. The only related evidence is the P-wave stack velocity calculated from the multichannel seismic profiles with limited reliability, which suggests a thin lower crustal HVL beneath the QDNB [25-27]. The subsidence and dynamic uplift estimated from the multichannel seismic profiles confirm the stronger mantle upwelling beneath the eastern QDNB at synrifting stage [7]. The synrifting magma evidenced by multichannel seismic profiles seems to be very limited in extent and distribution, which is mostly observed beneath the eastern QDNB [21, 22]. Meanwhile, unlike the limited synrifting magmatism anomaly, multichannel seismic profiles show that the postrifting igneous rocks are observed almost beneath the whole QDNB, significantly in the Xisha Trough of eastern QDNB [21, 28].Thus, the existence of lower crustal HVL, the scale of mantle underplating, or the lateral variation of deep magmatism anomaly is still unclear, which could need more evidence. The gravity-seismic joint inversion of lithospheric density structure has been proven to be a very effective means to recognize the magmatic anomaly, especially in the “magma-rich” region, such as Pearl River Mouth Basin and Taixinan Basin [6, 9, 29]. However, there are few studies on the lithospheric density structure in “magma-intermediate” region due to the higher difficulty of the recognition of limited-scale synrifting magma activities [26, 28, 30]. Limited by poor resolution of early data, the studies of gravity-seismic joint inversion only confirm the existence of lower crustal high-density layer beneath the QDNB but failed to describe the detail of space variation [26, 30]. Even the latest research from Gao [28] denies the existence of lower crustal high-density anomaly. To solve this problem, we performed the combined reflection seismic and gravity modeling and published five density structural profiles across the QDNB, which highlights the details of magmatism anomaly. Constrained by the latest high-precision data multichannel seismic profiles, the lithospheric density structure should be an effective method to explore the magmatism anomaly and provide more deep information.The QDNB, located at the western corner of the northern continental margin, is situated at the intersection of two tectonic regimes: the S-N extension of the Northwest Subbasin to the east and the sinistral/dextral transtension of the Red River fault to the west [25, 28, 31]. Although disagreements have always occurred, most scholars suggest that the seafloor spreading of the SCS occurred from ~32 to 15.5 Ma, including at least one ridge jump at approximately 23 Ma [32-34]. Especially under the dominance of the SCS breakup in the Early Oligocene, the accommodation zone in the QDNB was activated along the preexisting NW-trending basal faults and uplift, which divided the QDNB into two parts: the NE-trending western QDNB featuring asymmetric grabens, high-angle normal faults, and lower heat flow; and the EW-trending eastern QDNB featuring composite symmetric grabens, highly developed detachment faults extending deep into lower crust, and widespread postrift magmatic activity [15, 21, 35, 36]. In response to the multistage of SCS spreading, the tectonic activities of the western QDNB were strongly weakened at about 23 Ma, but the detachment faults of Xisha Trough located in the eastern QDNB are still active until 23–16 Ma [25, 37]. Characterized by detachment faults extending deep into the lower crust, the Xisha Trough of the eastern QDNB is usually regarded as a typical failed rifting basin [27, 28, 31]. For the postrift stage, subsidence and magmatic activity occurred in the whole QNDB, especially the accelerated subsidence from 10.5 to 5.5 Ma and stronger magmatic intrusion in the Xisha Trough [28, 38, 39]. Therefore, some scholars believe that the long-existing magma anomaly (weak body) beneath Xisha Trough may be the key factor for the differences between the western and eastern QDNB [39].The lithospheric structures of the QDNB have been studied in detail via multichannel seismic reflection profiles, OBS profiles, gravity inversion, seismic tomography, and surface heat flow [13, 14, 24, 26, 28, 38, 40-42]. Numerous multichannel seismic reflection profiles revealed a gradually decreasing trend in the crust thickness from >22 km on the northern and southern parts to <10 km on the central depression, with two extremely thinned domains of <4 km [28, 31, 36, 41]. These two extremely thinned domains, located in the Ledong Sag of western QDNB, and another in the Baodao Sag and Changchang Sag of eastern QDNB, show much higher extension factors of >6 than the others of 1.5–3.2 [41, 43-45]. In addition, the multichannel seismic profiles also suggest several 2 to 6-km-thick HVLs of >7.0 km/s beneath the two extremely thinned domains, which are almost widely distributed in the entire portion of the lower crust [26, 30]. Confusingly, similar high-velocity anomalies are still not observed in the OBS profiles beneath QDNB, which are widely observed beneath the eastern part of the northwestern SCS, including the Pearl River Mouth Basin and Taixinan Basin [13, 14, 24, 46, 47]. HVLs are usually regarded as the result of partial melting caused by a synrift upwelling asthenosphere, as indicated by the mafic composition and high density of 2.90–3.05 g/cm3 [6, 9, 16, 29]. Follow-up research also has found that the HVLs were usually defined as a weak ductile layer in the Pearl River Mouth Basin but a strong, brittle layer occurring at the crustal base of the Xisha Trough [48, 49].Unlike the limited evidence for HVLs, the multichannel seismic reflection profiles show the wide distribution of postrifting magmatism on the middle-eastern QDNB, including the igneous diapirs, lava flows, sills, and dykes, especially on the hyperextended crust of Xisha Trough [18, 31, 50, 51]. Some igneous diapirs cut through the Moho reflectors and overlying sediment layers, and even were extruded into the seabed of hyperextended crust [21, 31]. Almost the entire magmatism was widely active from the middle Miocene, which was believed to be a continuous effect of sysrifting deep mantle upwelling [6, 20], or direct influence related to Hainan mantle plume [14]. Recent research shows that the deep mantle upwelling and related dynamic uplift could be traced back to 28.4 Ma [7]. Therefore, considering the difficulty of identifying HVLs and magmatic activities, the construction of the lithospheric density structure undertaken in this study provides additional constraints for the space variation of deep magmatism anomaly beneath the QDNB.Data for inverting the lithospheric density structures are composed of Bouguer gravity anomaly, multichannel seismic reflection profiles, and OBS profiles, which have been published by previous studies (OBS2014, OBS2011-2, OBS2013-1, OBS1996, and L1–L5 in Figures 1-3).The Bouguer gravity anomaly data derived from the CryoSat-2 and Jason-1 satellites were obtained from the latest “gravity model V32.1” released by the Sandwell with 1.0’ × 1.0’ resolution (https://topex.ucsd.edu/index.html). Bounded by the tectonic transition zone in the central QDNB, the Bouguer gravity anomaly in the research area was segmented into two anomalous areas (Figure 1(b)). More precisely, the western anomaly area exhibited a gently seaward-increasing trend from a low gravity anomaly (~20 mGal) at the continental shelf to a high gravity anomaly (~60 mGal) at the Ledong Sag and Lingshui Sag. For the eastern QDNB, a low gravity anomaly (<40 mGal) and a high gravity anomaly (>100 mGal) can also be observed at the continental shelf and Xisha Trough, respectively. Moreover, it is worth noting that an analogous Bouguer gravity anomaly (< 130 mGal) is also found in the Northwest basin.Five multichannel seismic reflection profiles published in previous studies with record lengths of 12 seconds (two-way travel time) were used to constrain the structure of the crust and upper mantle [26, 28, 31, 36]. According to the thorough study by Zhao [52], the time-depth conversion of the power function (D = 924.2t1.368 + 103.5) was made to estimate the thickness of sediment layers. As well, combined with the research about the velocity in the upper and lower crust, the average P-velocity value of 6.5 km/s was assumed to calculate the thickness of the crystalline crust [13, 14, 24, 40, 41, 53]. To ensure the accuracy of the crustal structures, the OBS profiles of OBS2014, OBS2011-2, OBS2013-1, and OBS1996 were also used to constraint the detail of Conrad discontinuity and Moho depth [13, 14, 24, 53]. Considering the uncertainty for deeper structure, the P-wave stack velocity calculated from the multichannel seismic profiles of L2 and L5 was only used to provide the general trend of the crustal velocity [26, 27].The Bouguer gravity anomalies are related to the shape and position of the polygon [54]. Using the two-dimensional (2D) P-wave velocity structure and faults estimated by the joint analyses of OBS profiles and multichannel seismic profiles [13, 14, 24, 53] (Figure 1(a)), several initial polygons have been constructed to represent the density structure along the profiles. The initial polygons were also adjusted based on the previous crustal density research by Qiu [26], Gao [28], and Zhang [30], especially the scope of high-density bodies in the lower crust and upper mantle. In addition, we try our best to describe the shape details of igneous rocks, benefiting from the high-precision multichannel seismic profiles.From Hinze [55], the Bouguer anomaly of the polygonal prism at the original point (Figure S1) can be written as:whereG is the gravitational constant, ρ is the density contrast, N is the polygonal node, Y = 0.5×(Y1-Y2), x1, x2, z2, and z2 are the coordinates for a corner point of the polygonal prism in the (x, y, and z)-coordinate system, and u1, u2, ω1, and ω2 are the coordinates for a corner point of the polygonal prism in the new coordinate system. For the 2.5D case, DLRY and DLRU represent the differential logarithm of radial distances in terms of the (x, z)- or (u, w)-coordinates of the vertices, and DATY is the differential arctangent. Rij2 = Ri2 + Yj2, R202 = R22 + Y02, and R102 = R12 + Y02, (Y0 = O). R2 is the distance from the point (x2, O, and z2) to the origin O of the coordinate system. R1 is the distance from the point (x1, O, and z1) to the origin O of the coordinate system, and S and C are the slope of the line between points (x2, y1, and z2) and (x1, y1, and z1), which are used as coordinate system conversion coefficients. For the details, please refer to Zhao [54] and Hinze [55].To ensure the rationality of inversion results, the initial estimate of crustal density was derived from the P-wave velocity reported by Karplus [56] using velocity density scaling [57]:The P-wave velocity was mainly derived from the OBS profiles of OBS2014, OBS2011-2, OBS2013-1, and OBS1996 [13, 14, 24, 53]. The P-wave stack velocity calculated from the multichannel seismic profiles also provides an additional constraint [26, 27]. Combined with these previous studies of the geophysical model, we presented the initial crustal density range for the gravity anomaly inversion (Table 1). The gravity anomaly inversion was performed using the method of artificial bee colony algorithm [58] with at least 2500 iterations in the inversion process of each profile. The root-mean-square error (RMSE) between fitting gravity anomaly and actual gravity anomaly is 4.80, 1.38, 1.24, and 5.13 along profiles L1, L2, L3, and L4 (Figures 4 and 5). The RMSE of profile L5 will be discussed in detail in the following sections.According to the results of density inversion, the five-layer crustal model is developed along these five profiles, including the sea water, sediments, upper crust, lower crust, and upper mantle (Figures 4 and 5). The Cenozoic sediments could be divided into the rifting layer and postrifting layer, with the density scope of 2.20–2.40 and 2.40–2.60 g/cm3. Considering the stronger stratigraphic compaction in deeper depth, the thicker sediment layer in the central basin usually suggests a higher density, including the Ledong Sag, Lingshui Sag, and Xisha Trough. The results of sedimentary density are roughly consistent with joint analyses of density log data and seismic reflection profiles from previous studies but provided more details [26-28]. The upper crust also shows significant variation of lateral density along profiles but is almost controlled by the degree of upper crustal thinning. The central part of QDNB is strongly stretched to 3–5 km, even 0 km beneath the Xisha Trough, with a higher density of 2.71–2.78 g/cm3 in the upper crust. The northwestern part of QDNB shows a lower density of 2.60–2.69 g/cm3 in the upper crust, which is closer to the normal continental crust with low-magnitude extension (Yebei Sag and Songxi Sag in Figure 4). Additionally, a certain increase in density was observed in other parts, especially the base part of the upper crust, demonstrating high densities of 2.65–2.78 g/cm3.The lower crust is different from the upper crust, as manifested by two layers; the upper layer has a lower density of 2.80–2.90 g/cm3, and the lower layer has a higher density of 2.85–2.96 g/cm3. Moreover, the central part of QDNB also shows a higher density of 2.92–2.96 g/cm3 in the upper crust, which seems to be related to mantle upwelling or underplating. The lower crust is almost thinned to 0 km beneath the Xisha Trough along profile L4, with the igneous diapirs. These areas show the features of mafic composition, which is closer to the oceanic crust. Similarly, the density of the upper mantle exhibits a larger increase from the north or south shoulders of 3.00–3.21 g/cm3 to the central basin of 3.30–3.40 g/cm3, which is the most obvious indicator of deep magmatism anomaly. The igneous diapirs were also recognized as high-density bodies, ranging from 3.26 g/cm3 at the base to 2.64 g/cm3 at the top. The igneous rocks probably have been mixed with the low-density bodies from the upper crust or sediments. In addition, the surface heat flow data are also projected onto the profiles to help identify the scope of the magmatism anomaly [59].The western QDNB features by asymmetric grabens, high-angle normal faults in the upper crust, lower Bouguer gravity anomaly, limited low-angle detachment, and minor postrift magmatic activity [15, 21, 35, 36, 60] (Figures 1 and 2). The crustal density shows a significant variation along profiles L1–L2 due to the different degrees of crustal thinning. The intense crustal extension is observed in the Ledong Sag, Lingshui Sag, and the south part of profile L1, with the mantle uplift of 3–5 km, higher heat flow of 65–80 mW/m2, and higher extension factor of >3.00 [36, 41] (Figure 4). The mantle high-density bodies of 3.40 g/cm3 are conveniently beneath the intense crustal extension areas, which could be related to the deep magma upwelling. A similar high density of 2.92–2.96 g/cm3 is also recognized at the base of the lower crust, which could be related to the mantle underplating during the synrift phases. These lower crustal high-density bodies are usually regarded as the typical features of HVLs widely observed beneath the Pearl River Mouth Basin and Taixinan Basin [6, 9]. The Yabei Sag and Yanan Sag in the north part of profile L1 and the Xisha Uplift in the south part of profile L2 still keep the normal density of continental crust, with a limited extension factor of 1.25–1.70 [35, 36]. However, little igneous diapirs or large-scale magmatic intrusions were observed along profiles L1 and L2, suggesting a decrease in magmatic activity intensity from the synrifting phase to the postrifting phase.The profiles L3–L5 across the eastern QDNB exhibit an obviously different crustal structure, with the composite symmetric grabens, highly developed detachment faults extending deep into the lower crust, and widespread postrift magmatic activity [15, 21, 35, 36, 60] (Figures 1 and 3). The crustal extension and mantle underplating at eastern QDNB seem to be stronger, with the high-density bodies of 2.92–2.96 and 3.30–3.40 g/cm3 widespread beneath the most areas of lower crust and upper mantle (Figure 4). The thickness of the hyperextended crust is even close to 0 km in the central part of Xisha Trough along profile L4. The mantle uplift and heat flow are up to 10 km and 120 mW/m2 in the eastern QDNB, which is clearly higher than that identified in the western QDNB. The normal density of continental crust is only observed in the north part of profile L4 and Shenhu Uplift of profile L5. The most remarkable feature of eastern QDNB is the large-scale postrifting igneous diapirs, with the density of 2.63–3.26 g/cm3. The igneous intrusions along profile L5 are more intense than those observed along profile L3–L4, with the higher density of 2.94–3.26 g/cm3. Besides, it is worth noting the large error between the calculated anomalies and measured anomalies at the Xisha Uplift along profile L5 despite the density of the upper limit set in each layer. The boundary effect was also carefully considered and avoided in the inversion process. Considering the wide distribution of postrifting magmatic structures recognized in Xisha Uplift by previous studies, we suggest another igneous diaper beneath the southern end of profile L5.Magma plays a key role in the rifting and breakup process of the passive continental margin. The lower crustal HVLs are usually regarded as the results of mantle underplating at synrifting stage, which have developed up to 10 km in the northeastern margin of SCS [7, 20, 61]. The lower crustal HVLs are featured by Vp > 7.2 km/s (or Vp > 7.0 km/s) and mafic composition [10, 20, 62]. Therefore, the HVLs should be recognized as the high-density bodies of 2.92–3.05 g/cm3, which is the material mixing of the lower crust and upper mantle [6, 9, 16, 26]. In addition, the mantle upwelling will generate a lateral pressure in the lower crust due to the lithospheric isostatic balance [36]. The lower crust will be highly stretched when the mantle upwelling is strong enough to cause the lower crustal flow under high-temperature condition [63, 64]. Recent research suggested that the QDNB belongs to the “magma-intermediate”-type continental margin with a considerable scale synrifting magmatic activity [20]. However, there are no OBS profiles and multichannel seismic profiles to provide reasonable evidence for the lower crustal HVLs beneath QDNB, which probably is too small to be recognized at the current resolution [13, 14, 40, 60].Given the above discussion, we processed the joint interpretation of lower crustal density, lower crustal thickness, upper mantle density, and crustal extension factors across the QDNB to identify the scope of mantle underplating. The polygon shape of density inversion probably could not provide the accurate vertical thickness of HVLs, which include some subjective errors. Nonetheless, the bouguer gravity anomaly still provides information of high-density bodies beneath QDNB, which could give the lateral scope of mantle underplating. The high mantle density bodies of 3.30–3.40 g/cm3 and high lower crustal density bodies of 2.92–2.96 g/cm3 related to the mantle underplating were recognized beneath the most part of QDNB, especially the central part with high crustal extension factors of >3.00 (Figures 4 and 5). According to the results of dynamic uplift calculated by Zhao [7] and rapid subsidence predicted by Shi [38], the deep mantle upwelling beneath eastern QDNB could be stronger than the western QDNB.The identification of magmatic intrusions in the OBS profiles and multichannel seismic profiles is usually a formidable task due to the deep burial depths, which requires the clear images of 3D seismic data [19, 65]. The bouguer gravity anomaly also gives the additional information for the identification of large-scale igneous rocks, which could be ignored in the interpretation of seismic profiles. The recent research suggested that the postrifting magmatic intrusions could be the continuity of synrifting mantle underplating [6]. The magmatic diapiric structures exited the Moho and penetrated into the extremely thinned continental crust along the axis of Xisha Trough [31]. The magmatic intrusions are identified as the high-density bodies ranging from 3.26 g/cm3 at the base to 2.64 g/cm3 at the top. The large-scale igneous rocks were only observed beneath the eastern QDNB, even notably stronger at the eastern end of Xisha Trough with two higher density bodies of 2.94–3.26 g/cm3. Spatially, postrifting magmatic activities become stronger from the west to east, which is also supported by the surface heat flow data (Figures 1 and 5).Above all, the identification of high lower crustal density bodies of 2.92–2.96 g/cm3 and magmatic intrusions of 2.64–3.26 g/cm3 has been also supported by the previous studies of gravity-seismic joint inversion [26, 30]. Compared with the low-resolution of previous studies, we have presented the details of density spatial variation and defined the scope of mantle underplating beneath QDNB.The lithosphere is usually characterized by rheological stratification. The upper crust dominated by brittle deformation releases strains by means of fracture, while the lower crust dominated by plastic deformation releases strain energy via plastic flow releases upon the increments in temperature and depth (Figure 5(a)) [66]. Magmatic systems and detachment systems are the key processes involving the nonuniform breakup of continental margin, which appear to strongly control the lithospheric rheology [11]. The prominent detachment faults usually occur either between the brittle upper crust and the ductile lower crust or shallower levels within the brittle crust [11]. The magmatic additions at the base of the lower crust could complicate the composition, temperature, and rheology [48, 49]. According to the 2D steady-state heat conduction equation [67], Byerlee law [68], and creep law of rocks [69], the rheological structures of the lithosphere beneath the QDNB have been estimated by previous study [70] (Figure 5). The crème brûlée (CB-1), jelly sandwich-1 (JS-1), and jelly sandwich-1 (JS-2) rheological regimes were identified in the QDNB, which shows the noticeable correlativity with the results of density inversion and the degree of crustal extension [70] (Figure 5). These rheological regimes of CB-1, JS-1, and JS-2 are usually used to describe different tectonic backgrounds. The CB-1 regime is featured by the relatively strong crust and significantly weaker upper mantle due to the weakening by the high temperature of the active tectonic zone [71]. The JS-1 regime consists of relatively strong, possibly brittle upper crust and upper mantle layers separated by a weak, possibly ductile lower crust layer, which is close to the normal continental crust [72]. The JS-2 regime is usually used to describe the cooled oceanic lithosphere, which is also known as the Christmas tree regime with only one load-bearing layer [49]. For the details of the calculation process, please refer to Li [66].The northern part of QDNB is featured by a normal density continental crust, a slightly thinner crust of >22 km, a lower surface heat flow of <55 mW/m2, and a limited extension factor of 1.25–1.70, which usually estimated as being closer to the initial nonrifting continental crust [35, 41, 43]. The rheological structure of the initial continental crust falls into the JS-1 regime with a lower crustal viscous layer, as has been previously reported [43, 73]. Due to the decoupled crust–mantle relationship caused by a weak lower crust, the continental margin lithosphere rifting follows a nonuniform extension model [63]. The lateral variation in the deep magmatism anomaly further exacerbated the nonuniform breakup process in the northwestern margin of SCS.In the early stage of continental margin rifting, the magmatic upwelling caused early stretching to occur in a wide region, with low-grade crustal necking, and the mantle lithosphere began to break up before the crust (Figure 5(b)). The mantle high-density bodies of 3.30–3.40 g/cm3 were still preserved beneath the central and south parts of QDNB.In the middle stage, considering the decoupled crust–mantle relationship, most of the continental crustal extension occurred in the lower crust, which is heavily thinned and modified by the mantle underplating (Figure 5(b)). The high-density bodies of 2.92–2.96 g/cm3 at the base of the lower crust are widely observed beneath the hyperextended crust, especially the central part of QDNB with a crustal thickness of <10 km. However, the mantle underplating seems to be stronger at the eastern QDNB. If the continental crustal rifting stops at this stage, the lower crust still maintains considerable thickness or temperature and shows the weak rheology of ductile deformation. The western QDNB is currently in this stage and falls into the JS-1 regime, with limited detachment faults in the upper brittle crust (P1–P2 in Figure 5(a)). The lower crustal flow is possible in this stage if the mantle upwelling is strong enough to produce lateral pressure (Figure 5(b)).In the final stage, with the breakup of the brittle upper crust, the hot magma cools to form the new oceanic crust, which falls into the JS-2 regime with only one brittle load layer, including the extreme crustal thinning areas very closer to oceanic crust (Figure 5(b)). The central Xisha Trough is currently close to this stage, with the detachment faults extending deep into the lower brittle crust (P6 in Figure 5(a)). However, the postrifting magmatism is still very active in the Xisha Trough with high temperature, which makes the rheology of most parts fall into the JS-1 regime or CB-1 regime (P3–P4, P5, and P7 in Figure 5(a)). Thus, the lateral variation of deep magmatic anomaly should be the crucial factor for the nonuniform breakup process in the northwestern margin of SCS. Combined with plate reconstructions, the lateral variation of deep magmatic anomaly induced by subduction of the proto-SCS and the following SCS is inferred to be the cause of mantle upwelling in the northwest SCS margin [6, 36].The identification of deep magmatic anomaly is usually a formidable task due to the limited information from few OBS profiles, especially in “magma-intermediate” region. Constrained by the latest high-precision multichannel seismic profiles, we performed the combined reflection seismic and gravity modeling and presented lithospheric density structure profiles to recognize the scale of mantle underplating and magmatic intrusions beneath the QDNB.The mantle underplating beneath the QDNB is identified as having high mantle density of 3.30–3.40 g/cm3 and a high lower crustal density of 2.92–2.96 g/cm3, which is usually recognized by the HVLs in the northeastern margin of the SCS.The postrifting magmatic intrusions became stronger from the west to east, with the density ranging from 3.26 g/cm3 at the base to 2.64 g/cm3 at the top.The eastern QDNB shows a stronger crustal extension and deep magmatic anomaly than the western QDNB. The central part of Xisha Trough falls into the JS-2 regime because of the cooling of the heavily thinned lower crust in the final continental rifting stage, which is close to the cold and rigid oceanic crust.The mantle underplating beneath the QDNB is identified as having high mantle density of 3.30–3.40 g/cm3 and a high lower crustal density of 2.92–2.96 g/cm3, which is usually recognized by the HVLs in the northeastern margin of the SCS.The postrifting magmatic intrusions became stronger from the west to east, with the density ranging from 3.26 g/cm3 at the base to 2.64 g/cm3 at the top.The eastern QDNB shows a stronger crustal extension and deep magmatic anomaly than the western QDNB. The central part of Xisha Trough falls into the JS-2 regime because of the cooling of the heavily thinned lower crust in the final continental rifting stage, which is close to the cold and rigid oceanic crust.All data are available in the manuscript and supplementary material.The authors declare that there are no conflicts of interest regarding the publication of this article.We are extremely grateful to the editor and anonymous referees for their insightful comments, which have improved this paper. This work was supported by the Shandong Provincial Natural Science Foundation, China [Grant No. ZR2022QD087], the National Natural Science Foundation of China [Grant No. 92058213], the Open Fund of the Key Laboratory of Marine Geology and Environment, Chinese Academy of Sciences [Grants No. MGE2022KG2], and the Open Fund of the Key Laboratory of Submarine Geosciences, Ministry of Natural Resources [Grants No. KLSG2207].