The stabilizing effect of high pore-fluid pressure along subduction

17 Pore-fluid pressure is an important parameter in controlling fault mechanics as it lowers the effective 18 normal stress, allowing fault slip at lower shear stress. It is also thought to influence the nature of fault 19 slip, particularly in subduction zones where areas of slow slip have been linked to regions of elevated 20 pore-fluid pressure. Despite the importance of pore-fluid pressure on fault mechanics, its role on 21 controlling fault stability, which is determined by the friction rate parameter (a − b), is poorly 22 constrained, particularly for fault materials from subduction zones. In the winter of 2018-19 the 23 accretionary complex overlying Nankai Trough subduction zone (SW Japan) was drilled as part of 24 Integrated Ocean Drilling Program (IODP) Expedition 358. Here we test the frictional stability of the 25 accretionary sediments recovered during the expedition by performing a series of velocity-stepping 26 experiments on powdered samples (to simulate fault gouge) while systematically varying the pore-fluid 27 pressure and effective normal stress conditions. The Nankai gouges, despite only containing 25% 28 phyllosilicates, are strongly rate-strengthening and exhibit negative values for the rate-and-state 29 parameter b. We find that for experiments where the effective normal stress is held constant and the 30 pore-fluid pressure is increased the Nankai gouges become more rate-strengthening, and thus more 31 stable. In contrast, when the pore-fluid pressure is held constant and the effective normal stress is varied, 32 there is minimal effect on the frictional stability of the gouge. The increase in frictional stability of the 33 gouge at elevated pore-fluid pressure is caused by an evolution in the rate-and-state parameter b, which 34 becomes more negative at high pore-fluid pressure. These results have important implications for 35 understanding the nature of slip in subduction zones and suggest the stabilizing effect of pore-fluid 36 pressure could promote slow slip or aseismic creep on areas of the subduction interface that might 37 otherwise experience earthquake rupture. 38

pressure and effective normal stress conditions. The Nankai gouges, despite only containing 25% 28 phyllosilicates, are strongly rate-strengthening and exhibit negative values for the rate-and-state 29 parameter . We find that for experiments where the effective normal stress is held constant and the 30 pore-fluid pressure is increased the Nankai gouges become more rate-strengthening, and thus more 31 stable. In contrast, when the pore-fluid pressure is held constant and the effective normal stress is varied, 32 there is minimal effect on the frictional stability of the gouge. The increase in frictional stability of the 33 gouge at elevated pore-fluid pressure is caused by an evolution in the rate-and-state parameter , which 34 becomes more negative at high pore-fluid pressure. These results have important implications for 35 understanding the nature of slip in subduction zones and suggest the stabilizing effect of pore-fluid 36 positive relationship between ̅̅̅ and ( − ), many other phyllosilicate minerals show no relationship 130 (Table 1). 131    Smith and Faulkner (2010), 154 [9] Mair and Marone (1999), [10] Marone et al., (1990) Okuda et al., (2021). 159 160

Previous investigations into the frictional behaviour of Nankai sediments 161
To investigate the roles of effective normal stress and pore-fluid pressure on ( − ) we use samples 162  Table 2). 179

Sample preparation 182
Drill cuttings recovered from a depth interval of 3212.5-3217.5 mbsf were used for experiments. 183 First, the cuttings were washed to remove any residue drilling mud before being left to dry in an oven 184 at 60°C for 24 hours. Cuttings were then crushed and sieved to form a simulated gouge powder with a 185 grain size of <125 µm, similar to sample preparation methodologies used in previous studies (e.g. direction, to ensure that shear occurs within the layer itself and not between the edges of the gouge and 206 the forcing blocks. Once the layer is prepared the direct-shear assembly is wrapped in a low-friction 207 PTFE sleeve (0.25 mm thickness) before being inserted into a 3 mm thick PVC jacket. The PTFE sleeve 208 is used to minimize friction between the jacket and the direct-shear assembly in the vicinity of the layer. 209 The jacketed direct-shear assembly is then positioned between the platens of the sample assembly and 210 inserted into the pressure vessel of the triaxial apparatus. Normal stress is applied to the layer by the 211 confining pressure, and pore-fluid pressure is introduced via three porous disks on each forcing block, 212 spaced to ensure an even distribution of fluid (Fig. 2). Deionized water was used as the pore fluid in 213 this study. Both the confining and pore-fluid pressures are held constant during an experiment by servo-214 controlled pumps attached to each pressure system, with a resolution of better than 0.05 MPa. The gouge 215 layer is sheared by the axial piston and the applied force is measured via an internal force gauge with a 216 measurement resolution of better than 0.05 kN. In this setup a maximum load-point displacement of 8.5 217 mm can be achieved, which equates to a shear strain (γ) of ~10, given the final layer thickness of ~0.85 218 mm. 219 study. The rate-and-state parameters, and , were determined by processing the velocity steps using 233 the RSFit3000 program (Skarbek and Savage, 2019) which applies an inverse modelling technique with 234 an iterative least-squares fit. The program also solves for Dc (reported in Supplementary Tables 1 and  235 2) and treats the stiffness as a fitting parameter. 236 At the end of each experiment the permeability of the gouge was measured using the transient pulse 237 decay method (see Brace et al., 1968). This involves abruptly increasing by approximately 0.5 MPa 238 at the upstream end of the sample, producing a pressure differential across the gouge layer. This pressure 239 differential then decays with time as the pore-fluid dissipates through the sample allowing for the 240 permeability to be calculated.

Frictional strength and behaviour 247
An example of a typical frictional sliding test is shown in Figure 3a. The gouge samples initially 248 undergo quasi-elastic loading, shown by the steep increase in coefficient of friction, before yielding and 249 the initiation of steady-state sliding at approximately 1 mm of load point displacement. The friction 250 coefficient of the Nankai gouge at steady-state sliding is between 0.37-0.45 for all tests, with negligible 251 cohesion (Fig. 3c). Note that the reported shear stress values in Figure 3c  MPa later in the paper. It should also be noted that there is no obvious dependence of characteristic slip 293 weakening distance, Dc, with either ̅̅̅ or . All of the rate-and-state parameters ( , and Dc) for each 294 velocity step are reported in Supplementary Tables 1 and 2. 295 As we have tested the rate-dependence of friction, ( − ), of the Nankai gouge over a range of 296 pore-fluid pressure and normal stress conditions, the data can also be plotted as a pore-fluid factor, λ 297 (where λ = / ). Although there is a positive relationship between ( − ) and λ (i.e. ( − ) 298 increases as λ increases), it is more difficult to separate the individual contributions of and ̅̅̅ when 299 the data are plotted in this way, therefore we have chosen to present this data in the supplementary 300 material (see Supplementary Figure 1). In contrast, the results in Fig. 4 clearly show that the main 301 control over ( − ) is provided by , with ̅̅̅ having minimal effect on the frictional stability of the 302

gouge. 303
The pore-pressure dependence on ( − ) that we observe for tests conducted at ̅̅̅ ≥ 25 MPa on 304 Nankai gouge is similar to that observed by Xing  To elucidate further the cause of the pore pressure dependence observed in Figure 4, the average 319 up-step values for the individual friction rate parameters and are plotted in Figure 5. The friction 320 rate parameter is always higher than , leading to the rate-strengthening behaviour observed for 321 Nankai gouge. The data show that the friction rate parameter is largely independent of the pore-fluid 322 pressure (Fig. 5a), with values between 0.006-0.0076 for the entire range of pore-fluid pressures 323 investigated. However, the friction rate parameter shows a negative dependence on pore-fluid 324 pressure, decreasing from ~0 at = 5 MPa, to -0.0032 at = 75 MPa (Fig. 5b), highlighting that 325 changes in with pore-fluid pressure are responsible for the increased frictional stability.

Gouge permeability 331
The permeability of the Nankai gouge measured at the end of each experiment is low, with values 332 in the range of 10 -21 to 10 -22 m 2 (Fig. 6). The permeability is dependent on the effective normal stress, 333 with the lowest values occurring at high ̅̅̅. There does not appear to be any pore-fluid pressure 334 dependence on the measured permeability values (Fig. 6). The slow run-in rates at which the samples 335 were initially displaced, coupled with the permeability of the gouge suggest that excess pore-fluid 336 pressures did not develop within the gouge and adversely affect the experimental results presented (see 337 25 MPa (Fig. 4a), with pore-fluid pressure exerting the dominant control on the frictional stability at 370 these conditions (Fig. 4b). 371 The results in Figure 5 show that the increase in ( − ) with is caused by a decrease the friction 372 rate parameter , which becomes more negative at elevated (Fig. 5b). The cause of negative -values 373 is still not fully understood but they have been widely reported in previous investigations on predominantly controlled by ̅̅̅ not (Fig. 6). In contrast the negative -values are largely independent 387 of ̅̅̅ and are controlled by (Fig. 5b). This suggests that although transient pore-pressure variations 388 may affect the frictional response of the gouge they cannot fully explain the cause of the negative -389 values and why they become more negative with increasing in our experiments. This is further 390 evidenced by previous experimental work where positive -values have been reported for gouges with 391 similarly low permeabilities to the Nankai gouge tested here (e.g. Morrow et al., 2017). Therefore the 392 trends we observe in the velocity dependence of the Nankai gouge, where ( − ) increases with as 393 the -values decrease, suggest that this is primarily caused by the inherent frictional properties of the 394 gouge itself; any transient pore pressure effects that result from the low permeability nature of the gouge 395 will likely only have a secondary effect on the bulk frictional behaviour (Ikari et al., 2009a). 396 The rate-parameter is often termed the evolution effect and is classically thought to represent a 397 change in the asperity contact area after a velocity step (Dieterich and Kilgore, 1994; Marone, 1998). 398 However, there has been debate in the literature as to whether the contact area hypothesis is the whole 399 story, or whether the contact 'quality' (theory of adhesion; (Bowden and Tabor, 1950)) also affects the 400 frictional properties. Another fundamental manifestation of the evolution effect is the time-dependent 401 increase in frictional strength that occurs when rocks/gouge are held in stationary contact (often termed "frictional aging"), which has also traditionally been attributed to an increase in contact area as a result