Depth profile of frictional properties in the inner Nankai accretionary prism using cuttings from IODP Site C0002

Abstract

We conduct frictional experiments using cuttings collected at Nankai Trough IODP Site C0002 over 980.5–3262.5 mbsf (meters below seafloor) depth interval to better understand the frictional properties through the accretionary prism. The experiments are conducted at the in situ effective normal stresses (9–37 MPa) under brine-saturated conditions, and the slip velocity is abruptly changed in a stepwise manner to either of 0.3, 3, or 33 µm/s after the steady-state friction is reached. The friction coefficient (μ) of the cuttings samples ranges from 0.45 to 0.60, with a slight increase in μ with increasing depth, related to decreasing smectite content. The velocity dependence of friction (a − b) is positive at all depths and ranges from 0.001 to 0.006, which indicates a velocity-strengthening behavior; these values are consistent with relatively homogeneous deformation microstructures. The critical slip distance (D_c) ranges from 0.5 to 123 μm, with relatively large values obtained for the smectite-rich samples. The changes in both the friction coefficient and rate- and state-friction parameters are likely associated with mineralogical change and consolidation with increasing depth. Although all of the cuttings samples collected from Site C0002 exhibit a velocity-strengthening behavior, a slight decreasing trend in a − b with increasing depth indicates either a nearly neutral velocity dependence or a possible transition to velocity-weakening behavior at greater depths, which may be attributed to the occurrence of slow earthquakes in the Nankai accretionary prism.

1 Introduction

Nankai megathrust earthquakes have occurred repeatedly at approximately 100–150-year intervals and have historically been devastating events (Ando 1975; Hori et al. 2004). The 1944 Showa-Tonankai Earthquake and 1946 Showa-Nankai Earthquake were the most recent major earthquakes that occurred in the Nankai Trough. Slow earthquakes, including slow-slip events (SSEs) and very low-frequency earthquakes (VLFEs), have also been observed in shallow parts of the Nankai Trough accretionary prism (e.g., Ito and Obara 2006; Sugioka et al. 2012; Araki et al. 2017; Yokota and Ishikawa 2020; Shiraishi et al. 2020). Since slow earthquakes are sensitive to small changes in stress such as tidal stress, they can potentially serve as an indicator of the stress level and contribute to our understanding of megathrust earthquake processes (Obara and Kato 2020). The Nankai Trough Seismogenic Experiment (NanTroSEIZE) was conducted in 2007–2019 as part of the Integrated Ocean Drilling Program/International Ocean Discovery Program (IODP) to better understand the seismogenesis of Nankai Trough earthquakes. In this project, the D/V Chikyu drilled toward the plate boundary fault, where great earthquakes and tsunamis have been repeatedly generated, and collected sediment samples from the accretionary prism, measured　various physical properties, and installed sensors to monitor conditions at the drilled sites.

Core samples and cuttings were collected at various locations across the Nankai Trough during the NanTroSEIZE expeditions. Site C0002, the centerpiece of the broader NanTroSEIZE Kumano transect, reached the maximum depth of 3262.5 m below the seafloor (mbsf) in the Nankai accretionary prism (Tobin et al. 2020). A series of core samples were successfully recovered at Site C0002; however, coring was limited to the 475–1057, 2163–2218.5, and 2816.5–2848.5 mbsf intervals (Tobin et al. 2015, 2020). Takahashi et al. (2014) conducted triaxial deformation experiments using the core samples collected from the shallow interval at Site C0002 (< 1049 mbsf) and demonstrated that the friction coefficient was closely related to the clay mineral content. Okuda et al. (2021) found a velocity dependence on the frictional behavior of a deep core sample (2840.5 mbsf) from Site C0002, with velocity-strengthening behavior at slip velocities of > 10 μm/s, whereas velocity-weakening behavior is dominant at lower slip velocities. Although these experimental data are vital in advancing our understanding of the slip behavior in the Nankai accretionary prism, the limited coring intervals do not provide the depth variation of frictional properties through the accretionary prism. However, cuttings samples were continuously collected from the circulating drilling mud over the 875–3262.5 mbsf interval at Site C0002 during the riser operations for Expeditions 338, 348, and 358. In this study, we use cuttings samples from every 50 m depth interval over the 980.5–3262.5 mbsf range to establish a continuous depth profile of the frictional properties in the inner Nankai accretionary prism. We conducted laboratory experiments at in situ effective normal stresses under brine-saturated conditions, and we use the rate- and state-dependent constitutive friction law to analyze the frictional behavior. We then discuss the slip behavior in the accretionary prism and its spatial variations in the Nankai Trough.

2 Samples and methods

2.1 Geological setting

The Nankai Trough is a convergent plate margin where the Philippine Sea Plate subducts beneath the Eurasian Plate at a subducting rate of 4–6.5 cm/year (Heki and Miyazaki 2001; Loveless and Mead 2010). Site C0002 is located off the Kii Peninsula (Fig. 1), within the Kumano Basin and inner Nankai accretionary prism. The lithologies observed at this site are divided into five major units: Units I–III are in the Kumano forearc basin (above 975.5 mbsf), and Units IV and V are in the inner accretionary prism (e.g., Tobin et al. 2015). In this study, we used the cuttings samples collected from Units IV and V in the accretionary prism. Unit IV is dominated by silty claystone, with rare interbedded sandstone, and is interpreted as an older accretionary prism that consists of accreted Shikoku Basin hemipelagic and/or trench sediments. Unit V is dominated by silty claystone and is considered to be Shikoku Basin hemipelagic and/or trench sediments that were deposited during the Miocene (Tobin et al. 2015).

Fig. 1

Map of the NanTroSEIZE drilling sites. Site C0002 is indicated by the red square, and white circles mark the other NanTroSEIZE sites (after Tobin et al. 2015). Yellow arrows denote the estimated far-field vectors between the Philippine Sea Plate and Japan (Heki and Miyazaki 2001). Stars mark the hypocenters of the 1944 and 1946 great earthquakes

2.2 Sample descriptions

The cuttings samples used in this study were selected at every 50 m from the depth interval over the 980.5–3262.5 mbsf range at Holes C0002N, C0002P, and C0002Q (Fig. 2). The cuttings were washed with natural seawater at the drilling site to remove any residual drilling mud, and intact cuttings were then hand-picked to avoid any altered samples from their original state, following the approach outlined in Kitamura et al. (2021). The cuttings were dried in an oven at 40 °C for at least 24 h, then crushed, and sieved to produce gouge powder with a < 106 µm particle size. The total clay mineral, smectite, and illite contents of the cuttings samples from Holes C0002N and C0002P were taken from the X-ray diffraction data reported by Underwood (2017a, b), and the clay mineral content of the cuttings from Hole C0002Q was taken from Kitajima et al. (2020). The total clay mineral content is the sum of the smectite, illite, kaolinite, and chlorite contents. The smectite content decreased from Units IV to V and ranged of 6 to 35 wt% range. Conversely, the illite content exhibited relatively large variations and spanned the 13–36 wt% range, with an increase in illite content with increasing depth below 2500 mbsf (Fig. 2).

Fig. 2

Pressure, clay mineral content, and unit classification profiles for Site C0002 (after Tobin et al. 2015, 2020; Underwood 2017a, b; Kitajima et al. 2020). The lithospheric pressure is calculated from the overburden stress, which is derived from the rock density, and the hydrostatic pore pressure is calculated using a fluid density of 1.024 g/cm³

2.3 Experimental method

Frictional experiments were conducted using the biaxial frictional machine at Hiroshima University, Japan, under fluid-saturated conditions using an NaCl solution that simulated seawater (0.5 mol/L) at room temperature. The details of the experimental procedures can be found in Noda and Shimamoto (2009) and Katayama et al. (2015). The experiments were conducted at in situ effective normal stresses corresponding to the sample depths ($\sigma_{{{\text{eff}}}}$ = 9–37 MPa), which were calculated from the difference between the lithostatic overburden pressure and hydrostatic pore pressure (Fig. 2). The overburden pressure and hydrostatic pressures were obtained by integrating the bulk densities of the collected sediments and seawater, respectively. The gouge sample was sandwiched between gabbro blocks in a double-direct shear configuration, in which the thickness of the simulated gouge layer was initially ~ 0.5 mm. The assembly was placed in the water tank filled with brine and pre-compacted for 1 h at 2.5 MPa, and then normal stress was increased to the target value and kept for another 1 h before deformation (Fig. 3). The two-stage compaction was carried to prevent an excess pore pressure in the gauge layer. In this apparatus, pore fluid pressure cannot be controlled, such that we applied the in situ effective normal stresses as the normal stress under fluid-saturated conditions. The shear stress was applied downward from the upper piston to shear the simulated fault surfaces. After steady-state slip behavior was achieved at 8 mm displacement using a constant slip velocity of 3 µm/s, the slip velocity was changed abruptly in a stepwise manner among 0.3, 3, and 33 µm/s to analyze the rate- and state-dependent frictional behavior (velocity step test). The mechanical data were recorded using a KYOWA data recorder (EDX-100A) at 10 Hz sampling rate. The axial displacement was monitored by the electromagnetic transducer and corrected by the machine stiffness ($4.4 \times 10^{8} \; {\text{N}}/{\text{m}}$). The steady-state friction coefficient (μ) and the velocity-dependent parameters (a − b and D_c), which are defined by the rate- and state-dependent friction law (e.g., Dieterich 1979; Ruina 1983), were examined during these experiments.

Fig. 3

Photographs of the cutting samples, preparation of the simulated gouges, and experimental apparatus. The experiments were conducted under fluid-saturated conditions using a 0.5 mol/L NaCl solution

The friction coefficient μ can be calculated as the shear stress $\tau$ divided by the effective normal stress $\sigma_{{{\text{eff}}}}$:

$$\mu = \frac{\tau }{{\sigma_{{{\text{eff}}}} }}.$$

(1)

The rate- and state-dependent friction law used in this study is expressed as:

$$\mu \left( {V,\theta } \right) = \mu_{0} + a\ln \left( {\frac{V}{{V_{0} }}} \right) + b\ln \left( {\frac{{V_{0} \theta }}{{D_{{\text{c}}} }}} \right) ,$$

(2)

where $\mu_{0}$ is the initial friction coefficient before the velocity step; V₀ and V are the slip velocities before and after changes in slip velocity; a and b are nondimensional parameters that define the velocity dependence of the friction coefficient; D_c is the critical slip distance; and θ is a time-dependent state variable (e.g., Dieterich 1979; Ruina 1983). We used the one state variable version of the rate- and state-dependent friction law in this study. We used the aging law for the time evolution of θ, which is expressed as:

$$\frac{{{\text{d}}\theta }}{{{\text{d}}t}} = 1 - \frac{V\theta }{{D_{{\text{c}}} }}.$$

(3)

When the steady state is reached after the velocity step from V₀ to V (i.e., dθ/dt = 0, such that $\theta = D_{c} /V$), the steady-state friction coefficient at slip velocity V is expressed as:

$$\mu_{{{\text{ss}}}} = \mu_{0} + \left( {a - b} \right)\ln \left( {\frac{V}{{V_{0} }}} \right).$$

(4)

The velocity dependence of a − b is then obtained as:

$$a - b = \frac{{{\Delta }\mu_{ss} }}{{{\Delta }lnV}},$$

(5)

where Δμ_ss = μ_ss − μ₀ and ΔlnV = ln(V/V₀). a − b in an indicator of whether the frictional sliding is stable or unstable; a − b < 0 indicates velocity-weakening and potentially unstable slip that can evolve into an earthquake, whereas a − b > 0 indicates velocity-strengthening and stable slip. We obtained the velocity-dependent parameters by analyzing the data from the velocity upstep using the RSFit3000 software (Skarbek and Savage 2019). The analyzed frictional properties were compared with the clay contents reported by Underwood (2017a, b) and Kitajima et al. (2020). If mineral data were not available at the same depth, we used the closest data to our samples when the depth difference was < 5 m.

3 Results

3.1 Friction coefficient (μ)

We conducted 46 experiments using the cuttings obtained from Holes C0002N, C0002P, and C0002Q (Table 1). Most experiments exhibited slip-hardening, such that the friction coefficient increases slightly with displacement (Fig. 4). Friction coefficient for each sample was calculated at a displacement of 8 mm and was in the 0.45–0.60 range, with a slight increase in μ with increasing depth (Fig. 5). We obtained μ values of 0.45–0.54, 0.46–0.60, and 0.48–0.54 for Holes C0002N, C0002P, and C0002Q, respectively. There were relatively small variations in μ throughout Holes C0002N and C0002Q, whereas the samples from Hole C0002P exhibited relatively large variations in μ, with a general increase in μ with increasing depth. Previous experiments that analyzed the Site C0002 core samples reported μ values of 0.53–0.55 in the shallow depth interval (Takahashi et al. 2014; Okuda et al. 2021), which is consistent with our data; however, the results at the deeper intervals yielded a lower value (μ = 0.41) than those obtained in this study. Bedford et al. (2021) have recently reported μ values of 0.37–0.45 using the cutting samples collected from the 3212.5–3217.5 mbsf interval. The differences in friction coefficient among these different studies for the deeper intervals of Site C0002 are likely caused by slip-hardening behavior, because the samples exhibited a slight increase in μ as the displacement increased. We showed that the variation of friction coefficient calculated at different displacements (Fig. 5), indicating that those calculated at a similar small displacement are mostly consistent with previous experiments.

Table 1 Summary of the experimental results

Fig. 4

Results of the frictional experiments of the cuttings samples collected from Holes C0002N (980.5–2240.5 mbsf), C0002P (2260.5–3050.5 mbsf), and C0002Q (3102.5–3262.5 mbsf). The experiments were run at a slip velocity of 3 µm/s for the first 8 mm of displacement, followed by stepwise changes of 0.3, 3, and 33 µm/s to test the velocity dependence of friction

Fig. 5

Depth profile of the friction coefficient obtained via analysis of the cuttings at Site C0002. Friction coefficient obtained at a displacement of 8 mm is shown by orange filled circles, and the lower and upper data (orange open circle) are obtained at displacements of 2 mm and 16 mm, respectively. The black open circles, squares, and diamond indicate data from previous studies at Site C0002 (Takahashi et al. 2014; Okuda et al. 2021; Bedford et al. 2021)

3.2 Velocity dependence parameters (a − b and D _c)

The velocity steps yielded positive a − b values for all of cuttings samples at Site C0002, with a − b in the 0.001–0.006 range (Fig. 6A). There was no systematic difference in a − b among the different hole, with a − b values in the 0.002–0.006, 0.001–0.005, and 0.002–0.005 for Holes C0002N, C0002P, and C0002Q, respectively. The resultant a − b for the upstep velocity to 33 μm/s was slightly larger than that for the upstep velocity to 3 μm/s. There was a slight decrease in a − b with increasing depth toward a neutral value, especially in the case of the upstep velocity to 3 μm/s. These positive a − b observations are consistent with previous studies at Site C0002 (Takahashi et al. 2014; Okuda et al. 2021; Bedford et al. 2021), although Okuda et al. (2021) reported that a − b can be negative at relatively low slip velocities (≤ 1 µm/s). Our experiments were conducted under brine-saturated condition, with no control on the pore (atmospheric) pressure; however, Bedford et al. (2021) tested a range of pore pressures and found a slight increase in a − b with increasing pore fluid pressure.

Fig. 6

Depth profiles of A the velocity dependence of friction a − b and B critical slip distance D_c at Site C0002. Red and blue circles indicate the results for velocity change from 3 to 33 μm/s and from 0.3 to 3 μm/s, respectively. The error bars are estimated from the variations in the repeated tests. Open circles, squares, and diamond indicate data from previous studies at Site C0002 (Takahashi et al. 2014; Okuda et al. 2021; Bedford et al. 2021)

The critical slip distance (D_c) was between 0.5 and 123 μm and generally decreased with depth (Fig. 6B). D_c tended to be larger for higher slip velocity. These values largely agree with those obtained by previous studies at Site C0002 (Okuda et al. 2021; Bedford et al. 2021). Bedford et al. (2021) reported that D_c was relatively insensitive to the pore pressure but increased with increasing effective normal stress. However, our depth profile exhibited a nearly constant D_c below ~ 2000 mbsf, with effective normal stresses in the 20–37 MPa range, which may be influenced by clay mineral variations.

3.3 Microstructure observations

We observed the microstructures of the C0002N-141SW (1510.5 mbsf, Hole C0002N) and C0002P-266SW (2560.5 mbsf, Hole C0002P) samples after the experiments (Fig. 7). The entire gouge layer of each sample was deformed and possessed abundant Riedel shear bands. Both samples exhibited a positive velocity dependence, and there was no systematic difference in the frictional behaviors of these two samples, even though they are from different depths. Homogeneous deformation microstructures, like those observed in Fig. 7, are typical for velocity-strengthening materials, whereas highly localized deformation is more common in velocity-weakening materials (e.g., Beeler et al. 1996).

Fig. 7

Back-scattered electron images of the recovered samples A C0002N-141SW and B C0002P-266SW. The gouge layer is sandwiched between the gabbro forcing blocks, and the arrows in the blocks indicate slip directions. Riedel shears are widely developed in these samples. These images were taken using an electron-probe microanalyzer with a 15-kV accelerating voltage and 2.5 nA probe current

4 Discussion

4.1 Relationship between the frictional properties and clay mineral content

The frictional properties that were determined using our experimental data were compared with the clay mineral contents of the cuttings samples reported by Underwood et al. (2017a; b) and Kitajima et al. (2020). Figure 8 shows the relationship between the friction coefficients and the total clay mineral and smectite contents, including previous experimental data from samples collected along the Kumano transect in the Nankai Trough. Our results indicate no systematic correlation between μ and the total clay mineral content; however, there was a weak negative correlation between μ and smectite content (Fig. 8). This decrease in μ with increasing smectite content has also been reported in previous studies (e.g., Takahashi et al. 2014; Ikari et al. 2013a, 2018; Okuda et al. 2021; Oohashi et al. 2015). It is widely accepted that smectite is characterized by a significantly lower μ value than those for other clay minerals under fluid-saturated conditions (e.g., Moore and Lockner 2007; Ikari et al. 2007; Behnsen and Faulkner 2012; Tetsuka et al. 2018; Morrow et al. 2017). Tembe et al. (2010) reported a nearly linear decreasing trend of the frictional strength with increasing smectite content in the quartz-mixed gouge materials. Our samples possessed small μ variation for smectite contents up to 34%; however, studies that have analyzed Nankai samples with higher smectite contents (Takahashi et al. 2014; Ikari et al. 2018) have revealed that μ decreases to as low as ~ 0.1 when the samples contain ~ 50% smectite, which is close to the end-member μ value of fluid-saturated smectite. The overall frictional strength is therefore controlled by the presence of weak materials, such as smectite, in the matrix of sample. Such a lithological dependency of the friction coefficient has previously been reported in various clay-bearing materials that have been collected from the IODP projects (e.g., Deng and Underwood 2001; Brown et al. 2003; Kopf and Brown 2003; Ikari et al. 2013b, 2018; Boulton et al. 2019). The Japan Trench Fast Drilling Project (JFAST) samples that were collected near the fault zone possessed μ values as low as ~ 0.2, which was significantly lower than those of the wall rocks (µ > 0.5), with these variations attributed to different smectite contents (Ikari et al. 2015). The presence of such weak materials is thought to be crucial for the shallow trenchward slip of the 2011 Tohoku Earthquake (e.g., Remitti et al. 2015). Sedimentary samples that were collected from the Cocos Plate, Costa Rica, possessed large variation of friction coefficient, ranging from 0.2 to 0.8 (e.g., Kopf 2013; Namiki et al. 2014; Kurzawski et al. 2016). These variations have also been attributed to the different clay mineral contents and lithologies, which likely influenced the spatial variations in slip behavior in the Costa Rica subduction system.

Fig. 8

Relationship between the friction coefficient and A total clay mineral content and B smectite content. Data obtained in this study are shown by orange circles, and open circles, triangles, squares, and diamond indicate data from previous studies of the Nankai samples collected along the Kumano transect (Takahashi et al. 2014; Ikari et al. 2013a; Okuda et al. 2021; Bedford et al. 2021)

Figure 9 shows the relationship between the velocity dependence of friction (a − b) and the total clay mineral and smectite contents. All of the samples exhibited a velocity-strengthening behavior (a − b > 0) and no systematic correlation with the total clay mineral content, but a − b increased slightly with increasing smectite content. Clay-rich gouges commonly exhibit a positive velocity dependence (e.g., Morrow et al. 1992; Ikari et al. 2007; Tembe et al. 2010), which is in agreement with our results. This is likely due to the low or negative values of the rate parameter b in the case of clay-rich materials (Saffer and Marone. 2003). However, a negative velocity dependence of friction was reported for several Nankai sediment samples under low effective stresses (Tsutsumi et al. 2011) and at ultralow slip velocities (Ikari and Kopf 2017). Okuda et al. (2021) reported a velocity-weakening behavior for the samples collected from the Nankai Trough at slip velocities similar to our experiments, but their samples consisted of intact samples with slightly lower smectite contents.

Fig. 9

Relationship between a − b and A total clay mineral content and B smectite content. Red and blue circles indicate the results for velocity change from 3 to 33 μm/s and from 0.3 to 3 μm/s, respectively. Open circles, squares, and diamond indicate data from previous studies of the Nankai samples along the Kumano transect (Takahashi et al. 2014; Okuda et al. 2021; Bedford et al. 2021)

Figure 10 shows the relationship between the critical slip distance (D_c) and clay mineral content. Although there were only limited total clay and smectite content variation in our samples, D_c tended to be larger for the sample with higher smectite contents. D_c is considered to be controlled primarily by the real contact area of the material surface (e.g., Dieterich and Kilgore 1994). D_c could be related to the rate parameter b, whereby both represent a change in the real contact area after the velocity step. The nearly maximized real contacts for weak and platy materials, such as clay minerals, may explain the relatively large D_c and the positive a − b in the smectite-rich samples. Frictional healing experiments have shown a slower recovery rate and larger transient slip distance for smectite compared to the other minerals (Katayama et al. 2015). This is possibly due the weak creep strength of smectite, indicating a rapid growth of creeping and hence a nearly saturation in contact area after a short period. In contrast, the contact area for illite may be far from being saturated as suggested by the higher recovery rate of illite. Consequently, the slight decrease in D_c with increasing depth is likely attributed to the change in mineralogy from smectite-rich sediment to illite-rich sediment.

Fig. 10

Relationship between D_c and A total clay mineral content and B smectite content. Red and blue circles indicate the results for velocity change from 3 to 33 μm/s and from 0.3 to 3 μm/s, respectively. Open squares and diamond indicate data from previous studies of the Nankai samples along the Kumano transect (Okuda et al. 2021; Bedford et al. 2021)

4.2 Implications for fault slip behavior in the Nankai Trough

A comparison of our results with those obtained at various locations along at the Kumano transect in the Nankai accretionary prism (e.g., Ikari and Saffer 2011; Takahashi et al. 2014; Ikari et al. 2013a; Okuda et al. 2021) is shown in Fig. 11. There are large variations in μ values of the Nankai samples, even at the same drilling sites. We show that our depth profile of the friction coefficient at Site C0002, where μ ranges from 0.45 at shallow depths to 0.6 at greater depths, possesses values that are slightly lower than those at Site C0009, which is located a few tens of kilometers landward in the accretionary prism (Takahashi et al. 2014). This difference could be attributed to the consolidation and lower abundance of smectite at C0009 due to diagenesis, as the mechanical strength of the sedimentary samples is highly sensitive to the degree of cementation (e.g., Schnaid et al. 2001; Ikari and Hüpers 2021). Samples from Site C0004/10, which is located in the outer prism where abundant splay faults are developed, possess lower μ values (0.22–0.48), with the intact samples possessing slightly higher μ values compared to the powder samples (Roesner et al. 2020). In these sites, the samples recovered from the fracture zones yielded markedly lower friction coefficient than those from the wall rocks (Ikari and Saffer 2011). Tsutsumi et al. (2011) reported relatively low μ values (0.28–0.49) at low slip velocities < 26 mm/s at Sites C0001 and C0004, although there was an abrupt drop in μ at slip velocities ≥ 26 mm/s. Conversely, the samples from the prism toe at Site C0006, and the incoming Philippine Sea Plate at Sites C0011 and C0012 possessed highly variable μ values, with 0.32–0.83 for the prism toe and 0.15–0.72 for the incoming sediments, which are likely due to the large lithological variations around these sites (Stipp et al. 2013; Ikari et al. 2013a, 2018; Okuda et al. 2021).

Fig. 11

Cross section of the Kumano transect across the Nankai accretionary prism (modified after Tobin et al. 2020), constructed using data from various drilling sites during the NanTroSEIZE expeditions, with the frictional properties at Sites C0002 (this study, Takahashi et al. 2014), C0009 (Takahashi et al. 2014), C0004/10 (Roesner et al. 2020), C0006 and C0011 (Okuda et al. 2021), and C0012 (Ikari et al. 2013a)

The velocity dependence of friction is also spatially variable in the Nankai accretionary prism. We show that all of the samples (down to 3262.5 mbsf) exhibit a velocity-strengthening behavior at relatively low slip velocities (up to 33 μm/s). This is in agreement with previous Site C0002 experimental results and with those from Site C0009, which is more landward in the inner accretionary prism (Takahashi et al. 2014). Therefore, the inner accretionary prism seems to be a region of stable slip, whereas the outer prism and further seaward regions exhibit a highly variable velocity dependence (Fig. 11). The fault zone materials collected at Site C0004 and C0010 possessed a − b values in the -0.008 to 0.006 range, with the intact samples possessing a negative velocity dependence (Roesner et al. 2020). A velocity-weakening behavior is more common in the regions close to the trench at Site C0006 and further seaward at Site C0011 (Okuda et al. 2021). There are several factors that control the velocity dependence of friction, including lithification and consolidation (Moore and Saffer 2001; Ikari and Hüpers 2021), the effective normal stress (Moore and Saffer 2001; Bedford et al. 2021), shear localization (Saffer and Marone 2003; Ikari et al. 2011), and small-scale material heterogeneities (Bedford et al. 2022). We showed that a − b decreased slightly with depth at Site C0002, which is in agreement with the lithification and consolidation model, although such a trend may also be attributed to a decrease in the smectite content and increase in the effective pressure with increasing depth. The velocity-weakening behavior is more widespread at shallow depths at Sites C0004, C0006, and C0011, such that various factors are needed to account for the complexity of these velocity dependences. Note that velocity dependence of friction can also be influenced by the sample preparation. The intact samples collected from the Nankai Trough exhibited velocity-weakening behavior in contrast to the powdered samples that favor velocity-neutral to velocity-strengthening behavior (Roesner et al. 2020). Although the intact materials are relevant for an incipient fault zone, the powered samples as those used in our experiments represent slip behavior of the mature fault zones where fault movements have created natural gouge.

Slow slip events (SSEs) and very low-frequency earthquakes (VLFEs) have been observed in the accretionary prism and décollement zone in the Nankai Trough (e.g., Ito and Obara 2006; Sugioka et al. 2012; Araki et al. 2017; Shiraishi et al. 2020). The slip rates of these slow earthquakes (~ 1 m/day) are similar to those imposed in laboratory experiments, but the hypocentral depths are much deeper than the samples collected by the IODP projects. We show that a − b tends to decrease with increasing depth (Fig. 6), with consolidation and mineralogical changes likely contributing to this decrease. Although our samples are limited to depths ≤ 3262.5 mbsf, a slight decrease in a − b with depth may result in a neutral or negative value at greater depths. However, a − b is dependent on various physico-chemical properties including slip velocity, pore fluid pressure, temperature, structural developments, mineral assemblages, and their abundances (e.g., Marone 1998). In particular, temperature likely plays a key role in controlling the aseismic to seismic transition as temperature increases with depth in the accretionary prism. den Hartog and Spiers (2013) have shown that the transition from velocity strengthening to velocity weakening occurred at temperature of ~ 250 °C for illite-rich gouges under hydrothermal conditions. The transitional temperature is higher than that expected for the updip limit of seismogenic zone; however, relatively low nucleation velocities in nature compared to laboratory velocities may attribute to velocity weakening behavior at lower temperatures (den Hartog and Spiers 2013). In addition to velocity dependence of friction, fault stability is also influenced by critical slip distance, being more stable with a larger D_c (e.g., Scholz 2002). We show that D_c decreases slightly with depth, indicating that the shallow accretionary prism is relevant for stable region, whereas faults could become unstable at deeper portions due to the possible transition to negative velocity dependence with a small critical slip distance. Numerical modeling studies have favored such transitional slip behavior for the occurrence of slow earthquakes (e.g., Shibazaki and Shimamoto 2007; Rubin 2008). However, slow earthquakes may not be solely related to the rate-dependent friction, as the pore fluid pressure and associated effective normal stress on the fault zones are considered to be additional key factors for the slip instability in the Nankai accretionary prism (e.g., Kodaira et al. 2004; Hirose et al. 2021).

5 Conclusions

We conducted frictional experiments using cuttings samples that were obtained from 980.5 to 3262.5 mbsf at Nankai Trough Site C0002. The measured friction coefficients tended to increase slightly with increasing depth, which is likely controlled by the clay mineralogy, especially smectite content. The obtained velocity dependence of friction (a − b) indicated a velocity-strengthening behavior at all depths, which suggests that Site C0002 is a stable slip region, at least for the laboratory range of slip velocities tested here (0.3 to 33 μm/s). a − b decreased slightly with increasing depth owing to consolidation and mineralogical changes, and the transition to velocity weakening may occur at greater depths. The occurrence of slow earthquakes in the Nankai accretionary prism may be associated with such transitional slip behavior.