1 Introduction

As the largest continental marginal sea along the East Asian continental margin, the South China Sea (SCS) is the key region of interaction between the Pacific and Tethyan Tectonic Domains (Guo et al., 1983; Sun et al., 2005; Zhu et al., 2015; Wang et al., 2020). Three major hypotheses have been proposed to explain the rift initiating mechanism of the SCS: 1) Tectonic extrusion related to the sinistral motion on the Red River Fault (RRF) (Tapponnier et al., 1986); 2) slab-pull arising from subduction of the proto-SCS (Taylor and Hayes, 1983; Hall, 2002), and 3) the opening of the SCS initiated by the Hainan plume (Xu et al., 2012). It is generally accepted that the adjacent areas of the SCS could be divided into collision- extrusion and proto-SCS slab-pull tectonic provinces (Ren and Lei, 2011). Situated in the northwestern SCS continental margin at the intersection of two tectonic provinces, the Qiongdongnan Basin (QDNB) shows obvious variations in the west – east crustal extension, and preserves a wealth of information for understanding the lithospheric extension and breakup mechanism of the SCS (Fig.1; Zhang et al., 2009; Zhang et al., 2019). The difference in the west – east lithospheric extension of the QDNB appears to be related to lateral variation in the gigantic mantle low-velocity zone along the RRF, a lower crustal high-velocity layer (HVL), and detachment faults extending deep into the lower crust (Shi and Zhou, 2000; Yao et al., 2005; Zhang et al., 2007; Franke, 2013; Qiu et al., 2019). The Changchang Sag of the eastern QDNB features an extremely thin crystalline crust, an HVL occurring at the crustal base of the central sag, and uncorrelated magmatic activities with the rifting of SCS suggested by deep reflection seismic profiles. Accordingly, the Changchang Sag is generally regarded as part of the ocean-continent transition zone along the northwestern margin of the SCS, which has been identified as a non-volcanic or magmapoor passive margin (Wu et al., 2001; Qiu et al., 2003; Gao, 2015; Zhou, 2018). However, mantle serpentinization, which is usually regarded as one of the typical features occurring at magma-poor passive margins, is not observed in the northwestern margin of the SCS (Dean et al., 2000; Li, 2011; Ren et al., 2015). The lower crustal HVLs, which are also widely observed beneath the eastern part of the northwestern margin of SCS, are often recognized as the makeable feature of volcanic margins (Gao et al., 2016). Numerous studies have been conducted on the origin of the lower crustal HVLs and their relationship with the process of margin rifting (Nissen et al., 1995; Ding et al., 2016, 2018; Lei and Ren, 2016; Wang, 2020). In addition, the tectonic setting of the western QDNB is also undefined. Although its direction of lithospheric expansion is similar to that of the eastern QDNB, its rifting stages have been significantly remolded by the boundary fault of the RRF (Yuan et al., 2008; Luan and Zhang, 2009; Liu et al., 2015). Research on the difference in the west – east lithospheric extension of the QDNB could provide additional information toward an understanding of the passive continental margin rifting process.

Lithospheric rheology plays a key role in controlling the behavior of multiple-phase rifting of passive margins (He et al., 2001; Liu et al., 2006; Peron-Pinvidic and Manatschal, 2019). According to the different rheological structures of the lithosphere, the non-volcanic passive margins have been divided into two end-members known as the crust-mantle coupled model and the crust-mantle decoupled model (Huismans and Beaumont, 2011; Ren et al., 2015). The former proposes one hard, brittle layer of crust and upper mantle and the breakup of the crust before the mantle lithosphere with exposure of serpentinized mantle lithosphere. The latter considers a sandwich form with a soft, ductile lower crust occurring between two hard layers, leading to non-uniform extension with significant delays in the crustal breakup and HVL and magmatic underplating (Huismans and Beaumont, 2011; Ren et al., 2015). In addition, varying amounts of ductile behavior also result in differences in the extent of lithospheric thinning (Zhang et al., 2007; Tetreault and Buiter, 2018). Therefore, the construction of rheological structures is effective for exploring the dynamic deformation occurring at the north-western margin of the SCS.

Owing to difficulties associated with deep-water exploration, most studies on the lithospheric thermo-rheological structure focus on the shallow-water areas of the Pearl River Mouth Basin (PRMB), Xisha Trough, or Yinggehai Basin (YGHB) (Zhang and Wang, 2000; Shi et al., 2002; Tang et al., 2018). Reliable thermo-rheological structures across the QDNB have been modeled on the basis of abundant crustal-scale structural models constructed from high resolution seismic reflection in addition to heat flows calculated from deep-well drilling (Qiu et al., 2014, 2019; Shi et al., 2017). In this study, the results of previous research combined with the two thermo-rheological profiles calculated across the western and eastern QDNB are used to evaluate the continental margin rifting process of the SCS.

2 Geological Setting

Located in the northwestern passive rifted continental margin of the SCS, the QDNB is surrounded by PRMB extension structures to the east, the YGHB strike-slip system to the west, and the Xisha Trough to the southeast (Fig.1; Li et al., 2012; Wang et al., 2014). With the combined control of the two tectonic provinces, the QDNB fault system is divided into two parts by the accommodation of a strike-slip zone, NE-trending normal faults to the west, and nearly W–E-trending detachment faults to the east (Zhang et al., 2019). Controlled by the stretching stress introduced by the rollback of the paleo-Pacific plate and the slab pull of the paleo-SCS, the QDNB entered the syn-rift stage during the late Cretaceous (Doust and Summner, 2007; Yin, 2010). With the 500 – 700 km sinistral slip of the RRF and the seafloor spreading of the SCS at nearly 34 – 32 Myr, the QDNB exhibits a different W–E tectonic rifting process that activated the central accommodation zone (Barckhausen et al., 2014; Xie et al., 2015). In response to the ridge jump of the SCS at 23 Myr, the strike-slip inversion of the RRF and the SCS spreading ended at about 17 Myr, and the western and eastern QDNB successively entered the post-rift stage (Wang et al., 2006; Sun et al., 2009; Xie et al., 2015). The reactive 5.3 – 57 km dextral slip of the RRF at 5 Myr contributed to the rapid subsidence of the western QDNB (Guo et al., 2001; Yuan et al., 2008; Cai et al., 2010; Wang et al., 2011). Therefore, the different rifting processes occurring in the western and eastern QDNB show genetic connection with the RRF strike-slip system and extensional movement related to the SCS spreading (Li et al., 2012; Gong et al., 2017).

For the lithospheric structure, deep seismic reflection profiles suggest a roughly southeastward decreasing trend from 30 km at Hainan Island to 13 km at Changchang Sag with a Moho upheaval of 4 – 6 km at the central depression zone (Zhang et al., 2007; Tong et al., 2009; Zhao et al., 2011). Similarly, the entire crustal extension factor of the central depression zone is 1.9 – 3.2, which is higher than that in the surrounding areas showing a roughly west-to-east decrease (Zhang et al., 2010). Notably, depth-dependence extension is also obvious throughout the QDNB, and the contribution of the lower crust is significantly greater than that of the upper crust. This suggests that the plastic flow of the lower crust is related to mantle exhumation (Zhang et al., 2007; Tong et al., 2009; Zhang et al., 2010). Furthermore, igneous petrology and geophysical data indicate that the continental marginal rifting and mantle exhumation were derived likely from the extensional collapse of the Cathaysian continental marginal orogenic belt (Zou, 2011). However, the seismic tomography and analog modeling indicate a NW-trending gigantic mantle low-velocity zone 200 km wide from the RRF to the SCS, suggesting another mechanism driving the lithospheric rifting and mantle exhumation (Sun et al., 2005; Yao et al., 2005). In contrast to the widespread HVL in the northern margin of the SCS, a 2 – 6 km-thick anomalous HVL has been investigated only beneath the central depression zone of the QDNB. This layer could be considered as mantle underplating or the residual mafic layer of the extended Cathaysian lower crust (Zhang et al., 2009; Qiu et al., 2014). The rheological properties of the HVL remain unclear. It has been defined as a strong, brittle layer occurring at the crustal base of Xisha Trough (Shi et al., 2002) and as a weak, ductile layer at the PRMB (Hu et al., 2020). Therefore, the construction of thermo-rheological structures in this study will provide a better understanding of the origin of the HLV and the continental margin rifting of the SCS.

3 Data and Methods 3.1 Data

The data used to construct the lithospheric thermo-rheological model in this study are based on surface heat flow, P-wave velocity from seismic reflection profiles, and rock density structures inversed by gravity anomalies.

The heat flow data include the heat flow probe and oil-gas drilling data of 154 observations reported previously (Zhang and Wang, 2000; Mi et al., 2009; Li et al., 2010; Wang et al., 2014; Shi et al., 2017). The heat flow map of the QDNB and its adjacent area was compiled with 0.5˚× 0.5˚ grid accuracy (Fig.2). Significant seaward variations include a low heat flow of 50 – 60 mW m⁻² in the northern depression and uplift zone to a high heat flow of > 70 mW m⁻² in the central depression zone and southern uplift zone. Different from the roughly seaward increasing trend, the Xisha Uplift shows a lower heat flow of 50 – 70 mW m⁻², which is similar to that in the northern continental shelf. Along the NE-trending high heat flow near the central depression zone, the highest heat flow range of 85 – 105 mW m⁻² was observed in the eastern QDNB near the central accommodation zone, Shenhu Uplift, and Xisha Trough.

For the P-wave velocity, two deep seismic reflection profiles across the western and eastern QDNB were used to constrain the structure, velocity, and composition of the crust and upper mantle (L1 and L2 in Figs.1 and 2; Wang et al., 2013; Qiu et al., 2014). According to the initial crustal structure model taken from seismic refraction profiles, well constraints from standard velocity-density relationships, and high resolution free-air gravity anomalies, the density model was inversed along profiles L1 and L2 (Figs.3 and 4; Zhang et al., 2009; Qiu et al., 2014; Gao, 2015). Corresponding to the lower crustal HVLs > 7.0 km s⁻¹, significant density anomalies > 2.80 g cm⁻³ were also observed at the crustal base of the central depression zone. This layer is strongly thinned and was likely underplated by magma during margin rifting (Fan et al., 2019). Notably, the scope of the HVLs is more widespread than that of the high-density bodies owing to the limited resolution of the density inversion.

3.2 Thermal Model

We adopted a four-layer model consisting of the sedimentary, upper crust, lower crust, and upper mantle layers (Ma et al., 2012; Zhang et al., 2017) based on interpreted seismic reflection profiles. The two-dimensional (2D) steady heat conduction equation was used to calculate the temperature given by Zhang et al. (2016) as

$ \frac{\partial }{{\partial x}}(K\frac{{\partial T}}{{\partial x}}) + \frac{\partial }{{\partial z}}(K\frac{{\partial T}}{{\partial z}}) + A = 0, $

(1)

where T is temperature (K), z is depth (km), K is thermal conductivity (W mK⁻¹), and A is heat production (μW m⁻³). Furthermore,

$ \left\{ \begin{array}{l} T(x, {z_0}) = {T_0}(x) \\ \frac{{\partial T}}{{\partial x}}({x_0}, z) = \frac{{\partial T}}{{\partial x}}({x_1}, z) = 0 \\ - K(x, {z_m})\frac{{\partial T}}{{\partial x}}(x, {z_m}) = {Q_m}(x) \\ \end{array} \right., $

(2)

where z₀ is the depth of the seabed, z_m is the depth of bottom boundary, at 50 km in the models, x₀ is the left boundary, x₁ is the right boundary, T₀ is the surface temperature at the seabed, Q₀ is the surface heat flow (mW m⁻²), and Q_m is the bottom heat flow (mW m⁻²) at the depth of z_m. The 2D calculation of temperature distribution was performed using COMSOL Multiphysics 5.5 software and was solved by optimizing Q_m to fit the surface heat flow Q₀. The root-mean-square error (RMSE) between inversion surface heat flow and observation surface heat flow is only 1.74 and 3.18 along profiles L1 and L2, respectively (Figs.5a and 6a). If z_m is equal to the depth of Moho, the Q_m is defined as the mantle heat flow. The crustal heat flow (Q_c) could be obtained by simple subtraction of surface heat flow (Q₀) from mantle heat flow (Q_m). The ratio between them (Q_c / Q_m) is used to describe the vertical distribution of the heat flow within the lithosphere and to estimate the main heat flow source.

The thermal conductivity K and the heat production rate A are the basic thermophysical parameters of the lithospheric thermal structure (He, 2014). Because of the thick sediments in the basin, we assumed that K and A, acquired from well tests, remained constant in the sedimentary layer (Cai et al., 2019). For the deeper crust, K is a function of temperature T; the temperature effects are listed in Table 1 (Shi et al., 2002). The rate of heat production, A, is also inhomogeneous within the lithosphere. The most widely accepted model for heat production within the crust is the exponential decrease model (Cermák, 1989):

$ A(z) = {A_0}{{\text{e}}^{(- z/D)}}, $

(3)

where D is the characteristic depth within which most of the radiogenic elements are distributed (Zhou et al., 2017). Owing to its poor applicability, the exponential decrease model was used only for the top 10 km of the upper crust, referred to as the UTK layer; the heat production for the middle and lower crust was estimated according to the crustal velocity (Tang et al., 2018). These values are also listed in Table 1.

3.3 Rheological Structure

Yield stress envelopes (YSEs) are widely used to characterize the rheology of the lithosphere (Jackson, 2002; Burov and Watts, 2006; Burov and Diament, 2011; Zhang et al., 2013b). The rheological structure can be divided into upper brittle and lower ductile parts; the brittle regime for most rocks can be adequately described by Byerlee's law (Ranalli, 1991) in the calculation of lithospheric strength:

$ {\sigma _{{\text{brittle}}}} = {\sigma _1} - {\sigma _3} = \alpha \rho gz(1 - \lambda), $

(4)

where σ₁ and σ₃ are the maximum and minimum principal stresses, respectively; ρ is the average density set by the density model (kg m⁻³); g is acceleration owing to gravity (m s⁻²); λ is the pore fluid factor; and α is a parameter depending on the type of faulting, with values of 3, 1.2, and 0.75 used for thrust, strike-slip, and normal faults, respectively.

The study area is clearly controlled by normal faulting; hence, α was set to 0.75. The acceleration owing to gravity g and the pore fluid factor λ were set to typical values of 9.81 and 0.37, respectively (Wang, 2001). The creep strength for plastic deformation in the crust is dominated by the power-creep law (Kirby, 1983):

$ {\sigma _{{\text{creep}}}} = {\sigma _1} - {\sigma _3} = {(\frac{\varepsilon }{C})^n}\exp (\frac{E}{{nRT}}), $

(3)

where ε is the strain rate depending on geological data; T is temperature; R is the gas constant, set as 8.13447 J mol⁻¹ K⁻¹; and n, C, and E are material rheological parameters determined by the laboratory (Wang, 2001).

The total strength of the lithosphere with depth-dependent rheology is estimated by integrating the yield envelope (Williams et al., 2011):

$ s = \int_0^l {\min (} {\sigma _{{\text{brittle}}}}, {\sigma _{{\text{creep}}}}){\text{d}}z, $

(6)

where s is the lithospheric strength per unit width and l is the thickness. Therefore, we could obtain the curve sharp of YSEs by calculating the value of lithospheric strength (s).

Owing to the clear inhomogeneous deformation in the study area, the strain rate was set roughly as 10⁻¹⁶, following Hu et al. (2020). The creep parameters were selected by careful testing considering the significant function of HVLs for the lower crust strength. Compared with the mafic lower crust in the eastern part of the northern margin of the SCS, the lower crust beneath the QDNB is generally described as an intermediate composition (Shi et al., 2002; Wang, 2007). The values of the rheological parameters used in the construction of the rheological model are given in Table 2.

4 Results 4.1 Temperature Distribution Within the Lithosphere

The calculated temperature distribution of the lithosphere along profiles L1 and L2 with surface heat flow, heat flow from the deep mantle, and the ratio between them are shown in Figs.5 and 6. Obvious lateral changes in a temperature controlled by the surface heat flow ranged from the northern continental shelf to the southern continental slope, as presented along profile L1. Although the Moho temperature beneath the southern uplift zone exceeded 600℃, that beneath the northern depression and uplift zone is only 360℃, with a sharp temperature gradient occurring at the central depression zone. The vertical distribution of the heat flow within the lithosphere is described by the ratio of crust/mantle heat flow (Q_c / Q_m), which showed a trend approximately opposite that of the surface heat flow (Fig.5). The lower Q_c / Q_m of 0.20 – 0.60 in the southern uplift zone and Xisha Uplift contrasts with the higher values of 1.00 – 1.20 in the northern depression and uplift zone, which indicates more heat from the shallow crust. Therefore, the central depression zone of the QDNB can be regarded as a transition region from the cold continental shelf to the warm deep-water ocean. These conclusions are supported by previous studies conducted in the PRMB (Tang et al., 2018; Hu et al., 2020).

A more complex but gentle lateral temperature change is exhibited along profile L2, which extends almost to the high heat flow areas (Figs.2 and 6). With Moho temperatures of about 400℃, the southern uplift and central depression zones are colder than the Xisha Uplift, northern depression, and uplift zone. However, the Moho temperature distribution is not in perfect agreement with that at depths of 40 – 50 km. Compared with the lower Moho temperature of the central depression zone, the highest temperature at depths of 40 – 50 km was observed at the northern part of the central depression zone, which could be related to mantle magma upwelling (Fig.6). These features are slightly different from the temperature distribution in the Xisha Trough conducted by Shi et al. (2002). In the previous research, the highest temperature is right in the central part of Xisha Trough (Shi et al., 2002). In contrast to the lateral variation of Q_c / Q_m along profile L1, little change was observed along profile L2, with all values below 1, at 0.28 – 0.49. These results suggest more heat from the deep mantle. Considering the strongly thinned lower crust in its central zone, the Xisha Trough is usually regarded as the result of a depth-dependent lithospheric scale thinning, and its rifting is usually triggered by mantle flow (Lei and Ren, 2016).

4.2 Rheological Structure

The geodynamic processes of lithospheric deformation can be evaluated by considering the flux of material in the deeper crust, which can be inferred from the rheological structure of the lithosphere (Panza and Raykova, 2008; Brandmayr et al., 2011; Zhang et al., 2013b). The degree of brittle deformation shows an obvious decrease from the northern depression and uplift to the Xisha Uplift along profile L1, particularly at the upper mantle (Fig.7). The relatively strong central depression zone along profile L1 is divided into four load-bearing layers, including a weak lower crustal ductile layer sandwiched between two hard, brittle layers of the upper crust and upper mantle (P2 in Fig.7c). The HVL is recognized as the ductile layer, which is consistent with that reported by Hu et al. (2020). A stiffer lithosphere is present at the northern depression and uplift zone of the QDNB, with peak strength exceeding 800 MPa at the topmost part of the upper mantle; however, a relatively weak ductile layer is present at the base of the lower crust (P1 in Fig.7c). In contrast to that in the northern part of the QDNB showing limited, thinned continental crust (P1 and P2 in Fig.7c), the southern uplift zone and Xisha Uplift (P3 and P4 in Fig.7c) exhibit significant weakness in the lower crust and upper mantle, with a higher heat flow of 65.0 mW m⁻². This heat originated from the mantle and is likely related to mantle upwelling. The southern uplift zone and Xisha Uplift also exhibit characteristics of a strongly thinned crust, with brittle deformation of P3 and P4 restricted to the topmost 15 km of the upper crust and more ductile extension deformation occurring in the lower crust and upper mantle. These features appear to be consistent with the crustal extension factor discussed by Zhang et al. (2007) and Zhang et al. (2010).

Obvious lateral variations are also presented along profile L2, although the overall rheological strength is lower (Fig.8). The central depression zone shows a significantly higher percentage of brittle deformation than that occurring in the other regions. The northern part of the central depression zone (P6 in Fig.8c) is also divided into four load-bearing layers with strong rheological anomalies, with a significant lower crustal ductile layer occurring at the location of the HVL. In particular, the southern part of the central depression zone (P7 in Fig.8c) is characterized by a very stiff lithosphere with brittle deformation occurring in almost the entire crust and upper mantle; these features are similar to the typical rigid and cold characteristics of oceanic crust (Hu et al., 2020). Similar to P3 and P4 along profile L1, the brittle load-bearing layers of the Xisha Uplift, northern depression, and uplift zone are located in the topmost 15 km of the upper crust, and ductile deformation occurs in the lower crust and upper mantle (P5 and P8 in Fig.8c).

5 Discussion 5.1 East–West Differential Thermo-Rheological Structure of the QDNB

The geodynamic processes and patterns of lithospheric deformation can be evaluated by the thermo-rheological structure (Burov and Watts, 2006). Two main rheological regimes are generally applied to the lithosphere. The first is the jelly sandwich (JS) regime, which suggests that the continental lithosphere consists of a weak lower crust sandwiched between the relatively strong upper crust and strong upper mantle layers (Burov and Diament, 1995; Zhang et al., 2013b). The second is the crème brûlée (CB) regime, which suggests that the strength of the continental lithosphere is contained entirely in the crust and that the upper mantle is significantly weaker owing to high temperatures and weakening by water (Jackson, 2002; Zhang et al., 2013b). The JS and CB regimes can be further divided into two sub-groups of JS-1 and CB-1, characterized by a weak lower crust and JS-2 and CB-2 regimes with relatively strong lower crusts (Burov and Watts, 2006; Zhang et al., 2013b). A further JS regime, JS-2, which is also known as the Christmas Tree regime, contains only one hard, load-bearing layer overlying a ductile mantle (Jackson et al., 2008). The rigid oceanic lithosphere is usually regarded as a Christmas tree regime with uniform deformation (Hu et al., 2020).

From the calculated YSEs along profiles L1 and L2, three rheological regimes are recognized in this study (Figs.7, 8, and 9). The first is a typical JS-1 regime in the continental shelf and in most parts of the central depression zone corresponding to typical continental crust or limited thinned crust, with Moho temperatures falling 400℃ (P1, P2, and P6 in Figs.6 and 7). Unlike the dominant heat flow originating from the shallow crust at P1 of the continental shelf, more heat flow is derived from the deep mantle at P2 and P6 of the central depression zone, suggesting continental-oceanic transition characteristics with a weaker lower crust and upper mantle. In contrast, the second group, which includes the southern uplift zone, Xisha Uplift, and some part of the northern uplift zone (P3 – P5 and P8 in Figs.7 and 8), falls within the CB-1 regime with high Moho temperatures > 600℃, significantly weak lower crust, and weak upper mantle. For the CB-1 regime, the competent layers are mechanically decoupled by weak, ductile layers; thus, the contribution of a weak lower crust to the lithospheric strength is negligible. The third group is a JS-2 regime of typical oceanic crust with only one hard, load-bearing layer of whole crust and topmost upper mantle (P7 in Fig.8). The ductile layers at P7 of the eastern QDNB are sufficiently thin to be ignored in the rheological structure, suggesting a mechanically coupled relationship between the crust and the mantle. Furthermore, the degree of extension in the eastern QDNB is higher than that in the western QDNB, as suggested by the cold and rigid ocean crust in P7.

5.2 Implications for Rifting of Northwestern Margin of SCS

Rifted margins are the result of an extension within a continent, which leads to rupturing of the continental lithosphere and the subsequent formation of a new oceanic crust (Peron-Pinvidic et al., 2019). Thermomechanical numerical experiments conducted in previous studies highlight thermo-rheological structures as key parameters controlling the rifting process in the passive margin, particularly during the initial stages of extension (Tetreault and Buiter, 2018). Generally regarded as the 'small Atlantic', the SCS had a non-volcanic or magma-poor passive continental margin during its seafloor spreading at 32 – 17 Myr and experienced thermal subsidence after continental rifting (Li, 2011). However, significant discrepancies have been observed at the northwestern margin of SCS that remain enigmatic. These differences include widespread HVL, unclear evidence of exposed mantle lithosphere, and active magmatic events (Wu et al., 2001; Wang, 2020).

In their study of the whole crustal extension factor, Zhang et al. (2010) reported limited thinned crust at the continental shelf of the QDNB with a factor of 1.0 – 1.2. The JS-1 regime, with a weak lower crust at P1 of the western QDNB, could be estimated as close to the initial nonrifting continental crust. The QDNB experienced only limited upper crustal extension (1.45 – 2.2), which is significantly smaller than the whole crustal extension (1.5 – 3.2) (Zhang et al., 2010). In addition, considering the decoupled relationship between the upper crust and the mantle filling by the lower crustal viscous layer from the initial crustal model of JS-1 regime, the non-uniform extension is an acceptable continental margin rifting model (Huismans and Beaumont, 2011). The model has been suggested by many scholars (Zhang et al., 2007; Zhang et al., 2010; Gao et al., 2016; Lei and Ren, 2016). The characteristic result of this model is the breakup of the lower crust and mantle lithosphere before that of the upper crust (Fig.9; Huismans and Beaumont, 2011).

At the early stage of rifting, the strength of the lower crust and upper mantle was significantly reduced before the breakup of the upper crust (Figs.7, 8, and 9). According to the seismic profiles, the initial rifting almost occurred at Eocene, with high angle normal faults only rooted in the brittle upper crust. Almost all of the typical locations of P1 – P6 and P8 are in this stage with different degrees of extension and are still defined as thinning continental crust before the establishment of oceanic crust (Fig.9). The location of P2 shows a similar lithospheric rheological structure with the initial crustal model of P1, which suggests the relatively lower extension. However, the locations of P3 – P5 and P8 fall into the CB-1 regime, showing a greater degree of lower crustal extension and more heat flow from the deep mantle than those at the locations of P1 and P2 featured by the JS-1 regime.

With the increasing extension, the stage of extreme localized crustal thinning mostly occurred at 38 – 32 Myr, which is mainly recognized at the Xisha Trough of eastern QDNB (Fig.9; Lei and Ren, 2016). The crustal extension factor of the eastern QDNB is significantly higher than that of the western QDNB, with the highest value of > 3.2 and the lowest value of 2.8 km for lower crust thinness observed at Changchang Sag (Zhang et al., 2010; Wang et al., 2013). The rheology of the heavily thinned lower crust becomes more brittle (Zhou, 2018). The breakup features are defined by the major basin-forming conjugate faults extending deep into the lower crust along profile L2 of the eastern QDNB (Qiu et al., 2013; Zhang et al., 2019). Similar composite, symmetric horst and graben faults have been observed in Xisha Trough (Zhang et al., 2013a). When the basin entered post-rift subsidence with cooling of the hot 'oceanic' lithosphere, the central depression zone (P7) of the eastern QDNB assumed the JS-2 regime, showing features close to that of a rigid and cold oceanic crust (Fig.9). Notably, the Xisha Trough mostly shows the highest lower crustal extension before the breakup, which is featured by the lower crustal HVL related to the mantle underplating (Lei and Ren, 2016).

In response to the seafloor spreading of the SCS from east to west at 32 – 16 Myr, the accommodation zones in the central QDNB were activated along the pre-existing northwest-trending faults, and the crustal extension was only restricted in the Xisha Trough (Figs.1 and 9; Gao et al., 2016; Sibuet et al., 2016; Zhang et al., 2019). With the rupturing of the mantle lithosphere and necking of the lower crust, the upper crust of northwest sub-basin began to break up in contact with the hot mantle upwelling (Sun et al., 2009; Barckhausen et al., 2014). However, the HVL beneath the PRMB at the northern margin is likely the result of SCS spreading at this stage, which is only observed at the base of the lower crust.

According to the extension model of the passive margin as suggested by the thermo-rheological structure, the HVL is generally considered to be the result of mantle underplating, which could have been derived from the pressure-driven lateral flow of mantle (Huismans and Beaumont, 2011). The pressure-driven lateral mantle flow is also the key process facilitated by the lower crust thinning with creep deformation (Clift et al., 2002; Huismans and Beaumont, 2011). However, whether the beginning of the lateral flow of the mantle facilitated the breakup of the continental crust remains controversial. A variety of causes have been proposed in previous studies, such as the mantle low-velocity zone related to the RRF (Yao et al., 2005), the collapse of the Cathaysian continental marginal orogenic belt (Zou, 2001), the continuous asthenosphere low-velocity zone from the Yangtze craton (Huang and Xu, 2011), and mantle convection owing to the subduction of the Pacific Plate (Zhang et al., 2010; Li et al., 2012). The time of the east–west differential tectonic evolution of the QDNB shows a strong correlation with the RRF, particularly at the end of the syn-rift stage (Yuan et al., 2008; Gong et al., 2017). The east – west differential thermo-rheological structure and the degree of continental margin extension in the QDNB are also presented in a smallscale range. Therefore, the influence of the mantle low-velocity zone related to the RRF is a crucial factor in the east–west differential margin rifting process of the QDNB. Thus, the east–west differential thermo-rheological structures of the QDNB presented in this study provide a basic foundation for further relevant studies.

6 Conclusions

1) In response to the east–west differential patterns of shallow, brittle deformation in the QDNB, we present different east–west thermo-rheological structures from the continental shelf to the deep-water zone. The rheological structure of the western QDNB changes from the JS-1 regime of slightly thinned continental crust to the CB-1 regime of strongly thinned continental crust. The central depression of the eastern QDNB is defined as the JS-2 regime, and shows the characteristics of cold and rigid oceanic crust; this suggests a higher degree of extension.

2) The HVL of the QDNB resulted from the non-uniform extension with mantle underplating, which was facilitated by the lower crustal ductile layer and was derived by lateral mantle flow.

3) The influence of the mantle low-velocity zone related to the RRF is determined to be a crucial factor of the east–west differential margin rifting process occurring in the QDNB.

Acknowledgements

We are extremely grateful to the anonymous referees for their insightful comments, which have improved this paper. This work was supported by the National Natural Science Foundation of China (Nos. 41530963, 41176038 and 91858215).