1 INTRODUCTION

Polygonal faults (PFs) are pervasively developed in several subsided basins within the passive continental margin worldwide (Cartwright, 2011). They are resulted from layer-bound extensional faulting where the underlying and overlying strata are not disturbed, and thus they are also termed as layer-bounded faults (Cartwright and Dewhurst, 1998). They are also non-tectonic in origin with the polygonal geometry in plan view (Cartwright and Dewhurst, 1998), and primarily develop in the fine-grained sedimentary sequences that formed during the early burial history (Cartwright and Dewhurst, 1998). PFs serve as a pathway for hydrocarbon expulsion and provide a conduit for hydrocarbon migration (Alrefaee et al., 2018; Hoffmann et al., 2019). They are also used to assess the slope stability of the continental margins and the sealing sequences for petroleum, CO₂, and nuclear waste (Ireland et al., 2011). More interestingly, PFs are a challenge for classic soil mechanical theory, and they are created under the net-zero lateral strain, whereas they represent a shear failure in the early burial stage (King and Cartwright, 2020).

Although PFs generally show a classic polygonal geometry in map view, they also exhibit other geometries. PFs have a relationship in strictly orthogonal geometry with tectonic faults in a region where PFs and tectonic faults border as shown in the Sable Subbasin, Canadian Atlantic margin (Hansen et al., 2004). PFs deviate from their polygonal geometries to a remarkably radial geometry array around salt diapir, mud diapir, or gas chimney (Hansen et al., 2005; Stewart and Davies, 2006; Sun et al., 2010; Wang et al., 2010). Strikes of PFs parallel the strike of the slope, where PFs are well aligned and antithetic to the bedding dip in offshore Mauritania (Ireland et al., 2011). In addition, geometries of PFs change from linear for high slope, to rectangular for moderate slope, and to polygonal for low or no slope in the Great South Basin (Li et al., 2020). Under anisotropic stress conditions, PFs are preferentially oriented along the maximum horizontal stress (Ghalayini et al., 2017; Ho et al., 2018). Deep-water slope fan sandstones act as mechanical barriers and inhibit downward fault propagation thereby influencing fault height, interval space, and even strike in the North Sea Basin (Jackson et al., 2014) and offshore Uruguay (Turrini et al., 2017). PFs deform/fracture layers where coarsegrained channels underlie, and their strikes are always perpendicular (or parallel) to the axes of channels. (Victor and Moretti, 2006; Cartwright, 2011; Ireland et al., 2011). However, the differences of PFs between fine-grained and coarse-grained channels has not been ever further studied.

Numerous PFs are pervasively developed in finegrained sediments (vast thick mudstones) of the deep-water area of the northern South China Sea, including the Qiongdongnan basin (QDNB) and the Zhujiang (Pearl) River Mouth basin (PRMB) (Sun et al., 2014). In the western part of QDNB, PFs generally present typically polygonal geometry, and some are radial around mud diapirs and gas chimneys (Sun et al., 2010; Wang et al., 2010; Chen et al., 2011). These PFs are seen in two tiers within the upper Middle Miocene and Upper Miocene strata (Sun et al., 2010). Some of them are perpendicular to each other, resulting from the tensile stress field produced by differential settlement (Han et al., 2016). In the eastern part of QDNB, tectonic faults and a mud diapir severely impact the classic polygonal geometry of PFs; the former cause PFs to realign and are virtually orthogonal to the tectonic faults, and the latter induce PFs to exhibit a radial geometry around the mud diapir (Yin et al., 2010; Li et al., 2021a). Mudstone is the dominant lithology of polygonally faulted host strata in the Lower-Middle Miocene layers (Yin et al., 2010).

This study aims to understand the different geometries of the PFs related to differently lithological channels and their genetic mechanism. Different geometries of PFs are depicted using 3D seismic data from the Beijiao sag, middle QDNB, northern South China Sea. Differently lithological channels causing the differences of PFs are investigated in detail. Horizontal slices extracted on seismic coherence attributes are used to detect the different geometries of PFs resulting from the interplay of the presence of low-amplitude fine-grained and high-amplitude coarse-grained channels, and mounds/levees between channels. Root mean square (RMS) attribute is used to identify coarse-grained sandstones filled in highamplitude channels. This work aims at providing answers to the following questions:

a. What are the differences of (channel-segmenting and -bounding) PFs related to high-amplitude (coarsegrained) and low-amplitude (fine-grained) channels in the upper Miocene?

b. Can the differences of the PFs differentiate or determine the lithology of deep-water channels?

c. What is the genetic mechanism of the PFs related to the channels?

2 GEOLOGICAL SETTING

Qiongdongnan basin (QDNB) is located in the northwestern part of the South China Sea (SCS) (Fig. 1) and acts as one of the largest passive continental margins in the Western Pacific Ocean (Ru and Pigott, 1986). The QDNB is divided into four tectonic belts: northern depression belt, northern uplift belt, central depression belt, and southern uplift belt from north to south (Fig. 1) (Su et al., 2014). The central depression belt bears the Huaguang (HG), LedongLingshui, Yongle, Songnan-Baodao, Beijiao (BJ), and Changchang (CC) sags. The Beijiao (BJ) sag is bounded by the BJ uplift to the south, the Songnan low uplift to the north, Linnan low uplift to the west, and the CC sag to the east (Fig. 1).

The QDNB underwent three tectonic evolutionary stages mainly rifting followed by thermal subsidence and accelerated thermal subsidence (Table 1), which elevated the topography in the northern and southern part, whereas the topography in between the middle part is low (Xie et al., 2006, 2011; Tian et al., 2015). The QDNB covers approximately 82 000 km², with more than 60% of the present-day area under a deepwater environment. Basin fill, in response to tectonic movement, is composed of two super-sequences: the lower rift super-sequence and the upper post rift supersequence. These two super-sequences are separated by an angular unconformity corresponding to seismic reflection horizon T60 (23.3 Ma) (Sun et al., 2010; Tian et al., 2015). Coupled with the three tectonic evolutionary stages, the depositional environment transformed from alluvial to lacustrine, onshore to neritic, shelf-slope to abyssal environments from the Paleocene to Recent (Table 1; Tian et al., 2015). The deep-water environment in the central depression belt can be dated back to the early-to-late Miocene when there are deposits of vast thick fine-grained formation, where PFs are developed and distributed. In the western and eastern part of the QDNB, PFs are strictly confined in the upper Miocene (horizons T30– T40) (Han et al., 2016) and the middle-lower Miocene (T40–T60) (Li et al., 2021a), respectively. In the BJ of the middle QDNB, PFs are restricted within lowerto-upper Miocene strata (T30–T60), where there were the presence of deep-water (vast thick) fine-grained deposits (Li et al., 2021b, c).

3 DATA AND METHOD

The 3D seismic dataset has been offered by the China National Offshore Oil Corporation (CNOOC) and covers approximately 1 250 km². Three-kilometer long steamer with 240 channels was utilized to acquire the seismic data, and a total volume of 8×20 inch tuned air gun was used as the seismic source. The seismic data was processed by conventional techniques such as amplitude correction, band-pass filtering, and poststack time migration. The vertical sampling and shot intervals were 4 ms and 25 m, respectively. The bin size of the seismic volume is 12.5 m×25 m in the x-line and inline directions. The 3D seismic data have a dominant frequency of ca. 40 Hz. The vertical seismic resolution of the seismic data is ca. 15 m that equals 1/4 of wavelength at the dominant frequency.

Based on nannofossil biostratigraphy from limited drilling cuttings, isochronous seismic horizons for T70 (29.3 Ma), T60 (23.3 Ma), T50 (15.5 Ma), T40 (11.6 Ma), T30 (5.5 Ma), T20 (2.58 Ma), and their age significances are provided by CNOOC. The well YL19-1-1 is supplemented and used for age calibrations in the seismic-stratigraphic framework in the deep-water part.

Seismic attribute maps (such as RMS, coherence) were mapped to identify faults, high-amplitude channels. The time window of these attribute maps is selected as 40 ms along Horizon T40 (0 ms below and 40 ms above) because the upper Miocene channels are directly located on the Horizon T40. They were extracted from the 3D seismic data using OpendTect 6.6 (software) to show the presence of channels and faults. The RMS attribute highlights the variation of acoustic impedance within a selected time window (sample interval), and its map show where high amplitude turbidite channels with high values and low-amplitude muddy channels with low value are distributed. The higher RMS value corresponds to the greater variety of acoustic impedance, indicating the presence of coarse-grained sandstones encased by fine-grained mudstones. Herein RMS was used to detect the coarse-grained sandstones of channels. The seismic coherence attribute was used to interpret PFs (low coherent values) on selected horizons, and showed turbidite channels with high values. The blend map of RMS and coherence better displayed the relationship between channels and PFs. The width and height of channels and mounds were calculated using the average P-wave velocity of 2 400 m/s in the uppermiddle Miocene strata from the well YL19-1-1. Time structure map of Horizon T40 showed the distribution characteristics of (coarse-grained and fine-grained) channels between mounds.

4 RESULT 4.1 Characteristics of high-amplitude and-lowamplitude channels 4.1.1 High-amplitude and low-amplitude channels in plan view

3D seismic survey covers BJ uplift and BJ sag, where the (regional) slope dips towards the northwest-north (NWN) (Fig. 2a). In the BJ sag, channels alternatively and parallelly occur and are oblique (30°–40°) to the strike direction of the slope (Fig. 2a). They extend in the approximately east-west direction and are about 2-km wide and 5–15-km long. Interval spaces between each two channels are located in elongated levees or mounds. The scales of the channels in the western part of the study area are larger than those in the eastern part of the study area, and they gradually decrease towards the dip direction of the slope. Some channels locally merge and bifurcate.

In the western part of the study area, elongated high-amplitude (HA) channels also extend for the near east-west direction and are about 2-km wide and 5–15-km long (Fig. 2b). They alternatively and parallelly occur. Interval spaces between each two high amplitude channels are shown by low-amplitudes. High amplitude channels are predominantly characterized by high coherence values, and levees between channels are featured by low coherence values (Fig. 2c–d). In the eastern part of the study area, low-amplitudes dominate the area, where a well YL19-1-1 encountered fine-grained sediments in the upper Miocene channels (Li et al., 2021b, c). Low-amplitude channels are also predominantly characterized by high coherence values and levees are featured by low coherence values (Fig. 2c–d). Both high amplitude channels (Fig. 2e–g) and lowamplitude channels (Fig. 2h–j) are segmented by channel-segmenting PFs, and channel-bounding PFs are distributed along the boundaries of the channels. Both of the PFs are nearly perpendicular (Fig. 2k–l).

4.1.2 High-and-low-amplitude channels in seismic profiles

A seismic profile (Fig. 3) along the thalweg of a low-amplitude channel shows that sediments (above Horizon T40) filled in the channel are characterized by week amplitudes and seismic events within the channel are faulted. Additionally, a seismic profile (Fig. 4) along the thalweg of a high amplitude channel presents that sediments within the channel are characterized by high amplitudes. The thickness of high amplitudes of the channel gradually thins eastward, and seismic events within the channel are less faulted in contrast to counterparts within the low-amplitude channel. Additionally, several highamplitude channels occur in the middle Miocene strata, and they are the result of the interaction of turbidite flows and bottom currents (Li, 2019). A channel on horizon T20 is mainly characterized by chaotic reflectors. High amplitudes beneath the west limb of the channel may be caused by (shallow) free gas (Fig. 4b). The free gas has capacity to absorb the high-frequency component of the seismic signals, easily resulting in extremely weak (chaotic) reflectors in the underlying (reflector) strata. In this condition, the coherence attribute map from underlying Horizon T40 above (-) 40 ms shows that low coherence values occur in the northwest of the study area (Fig. 2c).

The seismic profile across several low-amplitude upper Miocene channels shows these channels exhibit a wave-shaped structure, and their scales are variable and tend to gradually decrease northward (Fig. 5a–b). Mounds (levees) between the channels are featured by low-moderate amplitude mounded reflectors. The seismic profile (Fig. 6a–b) across several high-amplitude channels presents that these channels also present a wave-shaped structure and their heights always surpass the heights of mounds/levees between these channels (Fig. 6a–b). Scales of these channels are variable and these channels are small in the middle (the numbers of "5, 6" in Fig. 6c) and large at both ends ("1–4" and "7, 8"). Mounds (levees) between channels are also featured by low-moderate amplitude mounded reflectors. Additionally, a channel on horizon T20 also is mainly characterized by chaotic reflectors (Fig. 6b). High amplitudes, with low frequency, beneath the north limb of the channel may be caused by shallow free gas, absorbing the high-frequency component and resulting in extremely weak (chaotic) reflectors with low coherence values in the northwest of the study area (Fig. 2c).

4.2 Characteristics of channel-segmenting and -bounding PFs 4.2.1 Channel-segmenting and -bounding PFs in plan view

Channel-segmenting and -bounding PFs are termed by (Victor and Moretti, 2006). Here we also adopt the terms to describe the features of the PFs. These PFs are imaged in Fig. 2c. They do not present typically polygonal geometry but show channel-segmenting and -bounding geometries. The channel-segmenting PFs and channel-bounding PFs are approximately perpendicular to each other. Additionally, the magnitude of the PFs in the east area where low-amplitude channels occur is larger than those in the western part of the area where high-amplitude channels occur. Low coherence values are distributed in the northwestern part of the area. The blend (Fig. 2d) of Fig. 2b–c was imaged to show that channel-segmenting PFs are always perpendicular to the axes of channels, and channel-bounding PFs are always located along the boundaries of channels and nearly parallel the axes of channels.

The differences of the PFs in the low-amplitude (east) and high-amplitude (west) channel distribution areas are evident. In the western part, channel-segmenting PFs cross the high-amplitude channels and they are strictly confined within the channels (Fig. 2e–g). Their lengths are 0.2–0.8 km and are always less than widths (approximately 2 km) of the channels. Although a few channel-segmenting PFs touch or cross the boundaries of the channels, the majority of them do not touch the boundaries. Most channel-bounding PFs always parallel the boundaries of the channels and extend for 0.2–2-km long. They present low coherence values, exhibiting elongated black belts, i.e. fault concentrated belts. Their combination consists of those faults concentrated belts. The lengths of these belts are the same as the lengths of channels.

In the eastern part, channel-segmenting PFs cross the low-amplitude channels, and they are not confined within the channels (Fig. 2h–j). Their lengths are 1.5–3.5 km, some of which are greater than widths (approximately 1.5 km) of the channels and even extend over (widths of) two/multi- channels. Most channel-bounding PFs always parallel the boundaries of the channels and extend for 0.5–3-km long. Their combination also consists of the fault concentrated belts, yet which are slightly wider than those of channel-bounding PFs related to high-amplitude channels.

Overall, the magnitude of the PFs related to highamplitude channels is smaller than those related to the low-amplitude channels. Although both are perpendicular to each other, the numbers of PFs in the western part are less than those in the eastern part (Fig. 2k vs. 2l).

4.2.2 Channel-segmenting and -bounding PFs in seismic profiles

The seismic profile (Fig. 3), along the thalweg of a low-amplitude channel, show that channelsegmenting PFs exhibit the feature of intraformational faults which are just distributed within lower-to-upper Miocene strata of BJ sag. They are strictly confined by Horizons T60 and T40 (Fig. 3a–b). Most of them nearly reach or near Horizons T40 and T60. The combinations of some PFs present Y-shape geometries and others exhibit the feature of gently listric geometries, as listric PFs described by Victor and Moretti (2006) and Cartwright (2011). These PFs seem to have approximately equal numbers of faults dipping in opposite directions.

The seismic profile along the thalweg of a high amplitude channel shows channel-segmenting PFs related to sandy channels (Fig. 4a–b) are shorter than those related to muddy channels (Fig. 3a–b). The combinations of some PFs also exhibit Y-shaped geometries. A few PFs just extend within the high amplitude channels. Some PFs extend downward into the middle Miocene strata, a few of which even extend into the lower Miocene. As the intensity and thickness of the high amplitudes within the channel laterally become low and thin from west to east, respectively, more and more PFs cross the channel. In brief, the upper Miocene high-amplitude channel shows that it inhibits the propagation and nucleation of channel-segmenting PFs, in contrast to those PFs associated with the low-amplitude channel.

The seismic profile across low-amplitude channels shows that the majority of channel-bounding PFs are densely situated on the mounds (Fig. 5c), a few of which extend downward into middle Miocene strata. Some of the PFs also show slightly Y-shaped geometries. Some minor PFs occur in the middlelower Miocene strata. Several tectonic faults occur below Horizon T60. There is a tendency that the larger the channels and mounds are (e.g., the numbers of channel ("1") and mound ("Ⅰ")), the longer the channel-bounding PFs are; and vice versa (e.g., the numbers of channel ("9") and mound ("Ⅸ")) (Table 2). Furthermore, the larger the channels and mounds are, the more the PFs are, and the wider the PFs (presenting fault concentrated belts) are, which can be seen in Fig. 5b.

Figure 6 shows that channel-bounding PFs also present on the mounds rather than within the channels, some of which extend downward into middle Miocene strata. A few of the PFs also present Y-shaped geometries. PFs are devoid above and below the high-amplitude channels, and some minor faults mainly occur below mounds. Tectonic faults extend upward into lower Miocene strata. There is an approximate tendency that the bigger the scale of the two channels (e.g., the numbers 2 and 3 of channels) at the two sides of a mound (2) is, the longer the channel-bounding PFs (on the mound 2) are; and vice versa (e.g., the numbers 6 and 7 of channels) (Table 3). The scale of mounds appears not to influence the scale of the PFs, which is impacted apparently by the scale of the channels. In brief, high-amplitude channels inhibit the propagation and nucleation of the PFs.

Two types of average channel-segmenting PFs related to the low-amplitude (Ⅰ) and high-amplitude (Ⅱ) channels and two average channel-bounding PFs related to the low-amplitude (Ⅲ) and highamplitude (Ⅳ) channels are plotted in the diagram of throw (T) vs. depth (z) (Fig. 6d). Both of them present C- or M-shaped geometries (Fig. 6d). They gradually decrease upward and downward, with a marked gradient in these throws. There is a clear trend that their vertical traced lengths are proportional to maximum displacements (Fig. 6d). In channel-segmenting PFs, the PF (Ⅰ) related to the low-amplitude channel is far longer than that (Ⅱ) related to the high-amplitude channel. The latter is clearly confined by the high-amplitude channel. In channel-bounding PFs, the PF (Ⅲ) related to the low-amplitude channel is slightly longer than that (Ⅳ) related to the high-amplitude channel. The latter also is clearly inhibited by the high-amplitude channel.

5 DISCUSSION 5.1 Lithology of low-amplitude and high-amplitude channels

PFs is predominantly developed in the deepwater basin in the passive continental margin, where simultaneously there is the pervasive development of sandy (e.g., turbidite) and muddy (e.g., contourite) channels, especially in the north SCS (Li et al., 2021b, c). Also, in the deep-water zone, it is well known that there are just few drilling wells due to the higher cost. It is very valuable if muddy and sandy channels can be differentiated by differences of the PFs without drilling wells in the early stage of deep-water petroleum exploration. To preferably summarize the rule of the differences of PFs corresponding to muddy and sandy channels, it is necessary to initially unravel lithology of herein low-amplitude and high-amplitude channels.

Numerous channels occur in the bottom of the upper Miocene strata in the entire study area (Fig. 2a). Low-amplitude (upper Miocene) channels are shown in the northeastern/eastern part of the study area and high-amplitude channels are clearly exhibited in the southwestern/western part (Fig. 2b). A low-amplitude upper Miocene channel was encountered by a drilling well YL 19-1-1 and was interpreted as a fine-grained muddy channel (see Fig. 14 in Li et al., 2021b). In addition, it is well known that low-amplitudes usually correspond to lower variety of acoustic impedance within strata (Brown, 2011), indicating the same lithology within the strata. Furthermore, bathyal depositional environment was widely developed in the Miocene in the study area (Table 1). It is reasonably inferred that low-amplitude channels in the eastern part of the study area are confirmed as fine-grained (muddy) channels.

A high-amplitude channel was also recognized as a turbidite channel by YL 19-1-1 (Li et al., 2021b) in the same study area. High amplitudes correspond to the greater variety of acoustic impedance, suggesting the existence of coarse-grained sandstones embedded by thick mudstones. In addition, high-amplitude channels acting as a barrier to the propagation of PFs are reckoned as coarse-grained sandstones, as described by Jackson et al. (2014). Therefore, it is reasonably inferred that high-amplitude channels are reckoned as coarse-grained (sandy) channels.

5.2 Differences of PFs corresponding to muddy and sandy channels

The geometries of channel-segmenting and -bounding PFs easily define and determine the presence of channels (Victor and Moretti, 2006; Cartwright, 2011). However, it is very difficult for petroleum geologists to determine and differentiate sandy channels from muddy channels without the drilling well dataset from the deep-water zone. Fortunately, not only sandy channels but also muddy channels were developed in this study (discussed above), and they yet offer an optimal opportunity to do it.

In the western part of the study area, high amplitude channels are mainly composed of sandy channels. Sandy channels, especially with the net-to-gross ratio of high proportion of sandstones in channels, always inhibit or prevent the propagation of PFs (Cartwright, 2011; Jackson et al., 2014). Thick sandstones (suggested by thick high amplitudes) in the sandy channel evidently prevent the nucleation of PFs in its interior (Fig. 4) and even inhibit the propagation of PFs near the sandy channels (Fig. 6). Additionally, channel-segmenting PFs related to sandy channels are strictly confined within channels and are shorter than the width of channels (Stewart and Davies, 2006; Cartwright, 2011; Jackson et al., 2014), as analogues shown in Fig. 2e in the western part of the study area. In the eastern part of the study area, low-amplitude channels mainly consist of mudstones which are confirmed by the drilling well YL19-1-1 (Li et al., 2021b, c). However, the relationship between PFs and muddy channels is not previously deciphered. There are several aspects for determining sandy channels from muddy channels by using the distinct features/differences of the PFs:

1) Compared to PFs related to sandy channels, PFs associated with muddy channels are no longer inhibited or prevented in their nucleation and propagation (Fig. 3b vs. Fig. 4b).

2) Interval spaces between channel-segmenting PFs in muddy channels are smaller (0.2–0.5 km) than those (0.5–1 km) in sandy channels (Fig. 2g vs. 2j). Additionally, the numbers of the PFs in muddy channels are more (average 2–3/km, Fig. 2j) than those (average 1–2/km, Fig. 2g) in sandy channels.

3) Channel-segmenting PFs related to muddy channels are longer and can cross multi- muddy channels (Fig. 2j), whereas channel-segmenting PFs associated with sandy channels are shorter and their lengths are virtually less than widths of sandy channels (Fig. 2g). This is one of the important features to determine sandy channels from muddy channels.

4) The widths of channel-bounding PFs (fault concentrated belts) on levees between muddy channels are slightly wider and longer than those on levees of sandy channels (Fig. 2g vs. 2j).

5) The longer and wider the channel-bounding PFs (corresponding to muddy channels) are, the larger the muddy channels are. Whereas, the shorter and narrower the PFs (corresponding to sandy channels) are, the larger the sandy channels are (Table 2 vs. Table 3; Fig. 2g vs. 2j). The longer and denser the channel-segmenting PFs (corresponding to muddy channels) are, the larger the muddy channels are. Whereas, the shorter and rarer the PFs (corresponding to sandy channels) are, the larger the sandy channels are (Fig. 2g vs. 2j; Fig. 3 vs. Fig. 4).

In brief, the scales of the channel-segmenting and -bounding PFs are proportional to the scales of the muddy channels, whereas the scales of the PFs are inversely proportional to the sandy channels. It is promising that these distinct features /differences (mentioned above) of the PFs can be widely employed to determine deep-water reservoirs of sandy channels from muddy channels without deep-water drilling wells in the early period of deep-water petroleum prospecting and exploration.

5.3 Combined geneses of channel-segmenting and -bounding PFs

The geneses of worldwide PFs, such as syneresis, overpressure hydrofracture, gravitational spreading, gravity sliding, volume contraction, density inversion, and diagenesis, remains controversial despite over decades of investigation (Cartwright, 2011; Ireland et al., 2011; Han et al., 2016; Li et al., 2020). Whereas gravitational spreading strongly supports one genesis of channel-segmenting and -bounding PFs. The occurrence prerequisite of gravitational spreading activity is the existence of a ductile basal (overpressured) layer of mudstones and brittle overburden sediments (Victor and Moretti, 2006). High-amplitude coarse-grained sandstones filled in channels naturally act as brittle sediments, as described by Victor and Moretti (2006). Additionally, as for lowamplitude muddy channels, smectite proportion in the upper Miocene mudstones is lower than counterpart in the middle Miocene mudstones from the Ocean Deep Project (ODP) drilling well in the deep-water north of SCS (Wan et al., 2008). It is generally inferred that mudstones containing more smectite tend to be more ductile and vice versa. The upper Miocene muddy channels overlying middle Miocene muddy strata also act as brittle sediments because they contain less smectite compared to the more smectite in underling middle Miocene mudstones.

During the process of the gravitational spreading of a channel, channel-segmenting PFs always perpendicularly cross the axis of the channel, and a few channel-bounding PFs approximately parallel the channel boundary (Fig. 7a). As time goes on, more PFs occur and channel-segmenting PFs always are strictly confined within (rather than across) the channels (Fig. 7b). These characters are also seen clearly in Fig. 2c–d. The cross sectional profile shows that channel-bounding PFs are intensively situated on the mounded levees or ridges (Fig. 7c), as shown in Figs. 5 and 6. Furthermore, gravitational spreading activity is always accompanied with listric PFs, as described by Victor and Moretti (2006). They are clearly seen in Fig. 3. Therefore, combined with the discussion above, the gravitational spreading is appropriately responsible for one genesis of PFs to date.

Although gravitational spreading responsible for the genesis of channel-segmenting and -bounding PFs is widely accepted, the different lithology of channels causing the different magnitude of PFs may reveal that there is yet another mechanism accounting for the genesis of PFs. Different lithology of channels encasing the PF tier does influence the formation of PFs (Fig. 2d), although Cartwright (2011) states "Lateral variation in single PF tiers can occasionally be attributed uniquely to some form of lithological control but this is not well understood at present". Compared to mud bodies, sand bodies generally have high permeability and porosity. The sand bodies (e.g., sandy channels) can accommodate the more fluids from mud bodies (surrounding strata) during the burial, probably reducing their own overpressure to be the absence of PFs around thick sandy channels (Figs. 4 & 6). Additionally, there is the presence of prevalent PFs far away from the sandy channels (Fig. 2d). It is inferred that the overpressure is closely associated with the formation of herein PFs influenced by lithology of channels. Overpressure hydrofracture is one genesis of PFs in the QDNB (Han et al., 2016), and Miocene (PF) host strata is in the overpressure condition in the QDNB (Wang et al., 2014). Due to the loss of overpressure, just a few PFs just occur around thick sandy channels. Overpressure hydrofracture difficultly occur in sandy channels and their surrounding strata where PFs are absent. Therefore, it is inferred that overpressure hydrofracture may be responsible for another genesis of herein PFs. There are still some reasons as follows:

1) The overpressured conditions revealed by the drill wells are widely achieved in the deep-water area in the Miocene of QDNB, especially in the BJ sag (Wang et al., 2014).

2) When pore fluid pressure reaches or surpasses 85% of net strata pressure, the strata easily tend to hydraulically fracture (Roberts and Nunn, 1995; Hao, 2005; Han et al., 2016). In the overpressured condition, the pore fluid pressure (P_f) enable the Morh circle to shift left, where P_f is equal to the displacement of the circle, and then the circle is easily tangent to the Morh-Coulomb failure envelope (Fig. 8) (Hao, 2005; Han et al., 2016). Thus, the strata fracture easily.

3) Some Y-shaped PFs are a result of the overpressure hydrofracture (see Fig. 14 in Han et al., 2016).Y-shaped PFs in seismic profiles are also exhibited in Figs. 3–4.

4) The coarse-grained sandstones in channels with well reservoir quality absorb fluids and then reduce the overpressure of surrounding mudstone strata, causing the absence of PFs (Dewhurst et al., 1999). Compared to the pervasive distribution of PFs in the lowamplitude channel (Figs. 3 & 5), the strata surrounding the high-amplitude coarse-grained channels are devoid of PFs in the surrounding strata (Figs. 4 & 6).

5) Overpressure hydrofracture has been reported to contribute to the formation of PFs in the QDNB (Wu et al., 2009; Han et al., 2016; Li et al., 2017), where BJ sag is situated.

If just only overpressure hydrofracture were the genesis of PFs, they would only present the polygonal geometry rather than channel-segmenting and -bounding PFs. Consequently, it is inferred that the mechanism of herein PFs is a result of combined geneses, i.e., gravitational spreading and overpressure hydrofracture.

6 CONCLUSION

The differences of the PFs related to coarse-grained and fine-grained channels are shown: the larger the finegrained channels are, the longer and wider the channelsegmenting PFs (presenting fault concentrated belts) are; whereas the larger the coarse-grained channels are, the shorter and narrower the PFs are. Coarse-grained channels act as a mechanical barrier to the propagation of PFs, whereas fine-grained channels are beneficial to the propagation and nucleation of PFs.

The scaled differences of channel-segmenting and -bounding PFs can be used to differentiate coarsegrained channels from fine-grained channels, which can be widely employed to determine deep-water reservoirs in the early period of deep-water petroleum prospecting and exploration.

The geometry of channel-segmenting and -bounding PFs are resulted from the activity of gravitational spreading. In addition, Y-shaped PFs also related to channels due to overpressure hydrofracture. It is inferred that the mechanism of herein PFs related to channels in the BJ sag is combined geneses of gravitational spreading and overpressure hydrofracture.

7 DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.