The Effect of Clay Content on the Dilatancy and Frictional Properties of Fault Gouge

Mature fault cores are comprised of extremely fine, low permeability, clay‐bearing gouges. Saturated granular fault materials are known to dilate in response to increases in sliding velocity, resulting in significant pore pressure drops that can suppress instability. Up to now, dilatancy has been measured predominantly in clay‐poor gouges. Clay minerals have low frictional strengths and, in previous experiments, even small proportions of clay minerals were shown to affect the frictional properties of a fault. It is important, therefore, to document in detail the impact of the proportion of clay on the frictional behavior and dilatancy of fault rocks. In this work, a suite of triaxial deformation experiments elucidated the frictional behavior of saturated, synthetic quartz‐clay (kaolinite) fault gouges at effective normal stresses of 60, 25, and 10 MPa. Upon a 10‐fold velocity increase, gouges of all clay‐quartz contents displayed measurable dilatancy with clay‐poor samples yielding comparable changes to previous studies. Peak dilation did not occur in the pure quartz gouges, but rather in gouges containing 10 to 40 wt% clay. The clay content of the simulated gouges was found to control the gouge frictional strength and the stability of slip. A transition occurred at ∼40 wt% clay from strong, unstably sliding quartz‐dominated gouges to weak but stably sliding clay‐dominated gouges. These results indicate that in a low permeability, clay‐rich fault zone, the increases in pore volume could generate pore‐fluid pressure transients, contributing to the arrest of earthquake nucleation or potentially the promotion of sustained slow slip.


of 16
The role of pore fluid pressure in determining fault shear resistance (τ) is described using the effective pressure law: where μ is the coefficient of friction, σ n is normal stress and p is pore pressure (von Terzhagi, 1936). Thus, by increasing the pore pressure, the apparent strength of a brittle fault decreases (Ougier-Simonin & Zhu, 2013), or vice-versa. Given the low compressibility of aqueous fluid, even small changes in the pore volume of a shearing fault gouge during changes in slip velocity will produce large fluid pressure variations. (Morrow & Byerlee, 1989;Rudnicki & Chen, 1988). The dilation and compaction of granular materials is accredited to changes in the packing arrangement of the grains (Mead, 1925;Reynolds, 1885). Slip velocity increases are accompanied by dilation of the sample material and a pore pressure drop, whereas slip velocity decreases show compaction of material and pore pressure increases (Brantut, 2020;Carpenter et al., 2016;Lockner & Byerlee, 1994;Morrow & Byerlee, 1989;Proctor et al., 2020;Rathbun & Marone, 2013;Samuelson et al., 2009). The timescale of pore pressure changes due to dilatancy or compaction depends on the competition between the changes in pore pressure and the drainage of pore fluid away from sites of excess pressure (Faulkner et al., 2018).
Previous friction experiments focused on dilatancy have typically used quartz-rich sands and powders (Lockner & Byerlee, 1994;Rathbun & Marone, 2013;Samuelson et al., 2009) or studied the initiation of a shear fracture in intact material and subsequent slip (Brantut, 2020;Proctor et al., 2020). In contrast, mature fault cores consist of extremely fine-grained fault gouges that are typically rich in clay minerals and have very low permeabilities, from 10 −18 to 10 −24 m 2 (Behnsen & Faulkner, 2011;Chester et al., 1993;Crawford et al., 2008;Faulkner et al., 2003;Faulkner & Rutter, 2000;Ikari et al., 2009;Morrow et al., 1984). Low permeabilities in clay-rich fault materials increases the timescales of the pore pressure changes and their significance in terms of fault shear strength and behavior.
Dilation of fault material could arrest a nucleating rupture due to a drop in pore pressure, which would increase the apparent shear strength of the fault. Dilatancy in response to slip velocity changes could also suppress fully dynamic behavior leading to the propagation of slow slip earthquakes (SSE), which can last periods of days to weeks (Frank et al., 2015;Liu, 2013;Rubin, 2008;Segall et al., 2010). SSE are inferred to occur in regions of elevated pore pressures at close to lithostatic conditions, including plate boundary settings such as the subduction interface or transform faults (e.g., Audet et al., 2009;Frank et al., 2015;Kodaira et al., 2004;Peng & Gomberg, 2010;Saffer & Wallace, 2015). Hence, fault networks that host SSE could be highly sensitive to pressure fluctuations and stable fault slip may be promoted on otherwise unstable faults (Obara & Kato, 2016;Ougier-Simonin & Zhu, 2013;Peng & Gomberg, 2010). SSE-hosting fault rocks are often rich in clay minerals (e.g., Audet et al., 2009;Frank et al., 2015;Ikari, 2019;Ikari et al., 2015), thus measuring dilatancy in clay-rich gouges is vital to understand whether pore pressure fluctuations will impact fault seismicity.
While the focus in this work is on dilatancy, intrinsic changes in the frictional strength with changes in slip velocity, such as in earthquake nucleation or propagation, are also investigated. Rate and state friction describes the response to slip velocity variations and is quantified by the constitutive law: where μ 0 is the initial steady state friction coefficient at slip velocity V 0 (ms −1 ), the velocity after the step change is V (ms −1 ), a is the direct effect, b is the evolution effect, D C is the slip weakening distance and θ is the state variable with units of time (Dieterich, 1979). If the friction stability parameter (a-b) is greater than zero in response to a velocity step increase, the material is designated as velocity strengthening. Velocity strengthening materials cannot nucleate seismic slip because with accelerating slip the material strengthens and so prevents instability. Conversely, a friction stability parameter of less than zero indicates that the material is velocity weakening. Velocity weakening is a prerequisite for unstable slip due to the material's overall weakening response to slip acceleration. Previous measurements of the stability of clay minerals (Giorgetti et al., 2015;Ikari et al., 2009;Logan & Rauenzahn, 1987;Ruggieri et al., 2021;Saffer & Marone, 2003;Tembe et al., 2010;Zhang et al., 2020) have shown them to be velocity strengthening, although there is a paucity of systematic measurements of how clay content affects stability compared to measurements of the bulk strength. The compilation by Ikari et al. (2011) indicates that there is broad correlation between clay content and higher (a-b) values. In contrast, more recent measurements by Ikari (2019) show that natural, clay-bearing fault gouges host SSE at slip rates simulating tectonic driving rates (cm/yr). Hence, it is important to document the effect of gouge composition on the friction parameter (a-b), alongside the dilatancy, in order to understand the relative contributions of intrinsic material effects and pore fluid pressure changes on shear resistance.
In this work, the materials and methods used to perform the experiments are first described. A series of synthetic quartz-kaolinite gouges were produced that cover kaolinite contents from 0 to 100 wt% and the synthetic gouges were sheared in a conventional triaxial deformation apparatus. The experimental results first describe the pore volume changes that occur in the synthetic fault gouges as a function of the controlled clay content, both during a bulk compaction phase and also in response to steps in the sliding velocity. The changes in the bulk strength and variation of the friction parameter (a-b) as a function of clay content are then reported. In the discussion, all the results are compared to existing data from the literature and the controls on dilatancy, gouge strength and stability are considered. Finally, the implications of the dilatancy for natural fault slip for the range of fault materials tested are discussed.

Materials and Methods
The materials used for the quartz-kaolinite synthetic fault gouges were Min-U-Sil 15 silica powder from the US Silica Company and KGa-1B kaolinite source clay from the Clay Minerals Society. Kaolinite was the chosen clay mineral for this study as it is a non-swelling clay and it is frictionally weaker than framework silicates in a similar way to other clay species (Behnsen & Faulkner, 2012;Moore & Lockner, 2004). Behnsen and Faulkner (2012) found that the water-saturated shear stress at yield for kaolinite was similar to chlorite, illite and lizardite. Hence, kaolinite acts as a useful proxy in this study for other species of non-swelling phyllosilicates that are common to fault gouges. Kaolinite is of common occurrence in fault rocks-examples of kaolinite-bearing faults that host both aseismic creep and seismic slip include the Median Tectonic Line in Japan, the Longitudinal Valley Fault and the Chelungpu Fault in Taiwan, at up to 19% (Anko section), 10%-12% and 10%-20% kaolinite respectively (den Hartog et al., 2021;Ishikawa et al., 2014;Kuo et al., 2009). Consequently, it is not only a good proxy for non-swelling clays, but is a commonly found constituent of many natural fault zones.
Min-U-Sil 15 was chosen for its high purity at >99% silica and the fine grain size sieved to <15 μm. KGa-1B also has a purity of >99% (Vogt et al., 2002) and was sieved during manufacturing to <44 μm to produce extremely fine grain sizes, with 57 wt% at <2 μm (Pruett & Webb, 1993;Vogt et al., 2002). Thus, the grain sizes in the quartz-kaolinite mixtures was comparable with fault gouges that have grain sizes down to the sub-micron scale.
As quartz and kaolinite have very similar densities (2.65 g/cm 3 ), the synthetic fault gouges were mixed by weight, which was also representative of the relative volumes. Kaolinite composition was varied by 10 wt% increments to produce 11 samples from 0 to 100 wt% kaolinite. With the aim of producing a well-mixed starting material, 20 g of the samples were tumbled as dry powders for a 3-hr period at 68 rpm in a 575 cm 3 volume container. Dry tumbling was chosen because the similar grain sizes and densities of the kaolinite and quartz powders promotes mixing over segregation (Williams, 1968). Microstructural observations of pre-shear gouge samples show a good degree of mixing, with only minor clumping of clay due to a build-up of static charge in intermediate mixtures.
All deformation experiments were conducted using a direct shear configuration in a triaxial deformation apparatus ( Figure 1). Using servo-controlled pumps, the apparatus can apply and maintain confining pressures up to 250 MPa and pore fluid pressures up to 200 MPa to an accuracy of 0.007 MPa. The confining and pore fluids used in the experiments were silicon oil and deionized water, respectively. An internal force gauge measures the axial force at a resolution better than 0.05 kN (Mitchell & Faulkner, 2008).
In the experiments, samples of fault gouge were deformed in direct shear sliders (Faulkner et al., 2018;Sánchez-Roa et al., 2016) shown in Figure 1. The dimensions of the direct shear sliders created a sample area of 720 mm 2 and two rubber spacers allowed a maximum load point displacement of 6 mm. Air dry synthetic gouge samples of 1.6 g were pre-compacted by loading using a uniaxial press oriented perpendicular to the layer at 10 MPa for the 10-25 MPa effective normal stress tests, or at 50 MPa for the 60 MPa effective normal stress tests. This produced samples with initial thicknesses of ∼1.3 mm before pressurization, improved sample cohesion for assembly, and reduced the initial porosity of the gouge. Porous stainless steel disks with permeabilities of 10 −13 m 2 in the direct shear sliders allowed transfer of pore fluid between the sample and the servo-controlled pore pressure pump ( Figure 1).
The synthetic gouge samples were sheared in velocity-step tests at pore pressures held at 40 MPa and effective normal stresses (σ n eff ) of 60, 25, and 10 MPa to represent increasingly overpressured conditions. Samples were initially run at an axial loading rate of 0.3 μm/s for a displacement of 2 mm to bring the sample past yield, where the shear strength of the sample reached steady state. Following the initial yield phase, the sliding velocity was stepped between 0.3 and 3.0 μm/s every 0.5 mm of load point displacement to a maximum of 5.5 mm. The displacement on the friction curves has not been corrected for machine compliance, which was 75 kN/mm. Permeabilities of the gouge samples were measured at the end of shearing using the pore pressure oscillation technique (Bernabé et al., 2006;Faulkner & Rutter, 2000;Kranz et al., 1990), which returned final permeabilities of 10 −17 m 2 for quartz-rich gouges to 10 −20 m 2 for kaolinite-rich gouges. This concurs with previous permeability measurements of phyllosilicate powders, including muscovite, pyrophyllite, biotite, phlogopite, illite and talc (Behnsen & Faulkner, 2011), and natural fault gouges (Faulkner & Rutter, 2000;Faulkner et al., 2003Faulkner et al., & 2018Wibberley & Shimamoto, 2003;Wu et al., 2020), which show ranges in permeabilities from 10 −17 to 10 −24 m 2 . The permeability was not monitored throughout the test as this reduced the capacity to monitor very small changes in pore volume during velocity steps, the technique for which is described next.
Pore volume changes during a test were monitored using the pore pressure control system, shown in Figure 1. The servo-controlled 5,000 mm 3 pump adjusted the volume of fluid in the system to maintain the pressure at 40 MPa from measurements in the sample; a pore pressure increase required pore fluid to be extracted into the pump, whereas a pressure decrease required pore fluid to be added to the sample. By monitoring the displacement of the piston in the pore fluid control pump using a high-resolution linear variable differential transformer (LVDT), the equivalent changes in pore volume in the sample were measured ( Figure 2). The maximum volume change that the LVDT could measure with one stroke was 200 mm 3 and the measurements were resolvable to 0.01 mm 3 . This pore volume monitoring technique (volumetery) included a time-dependent component because the control pump could only respond to pore pressure changes that had diffused out of the gouge sample. In order to assess the potential effect of these transients, the characteristic time (t) for equilibration of pore pressure across the sample was calculated using the thickness of the sample (l) and the hydraulic diffusivity (κ, see Supporting Info for calculations): (3) This returned characteristic times across the range of gouge compositions of 10 −3 to 1 s. The synthetic gouges therefore had sufficiently high permeabilities for any pore pressure transients to equilibrate rapidly with the control pump, and thus gave reliable measurements of pore volume changes produced during the velocity steps (180-1,800 s).
To determine the volume changes in response to velocity steps in a reproducible way, the volume curve for an individual velocity step was incrementally adjusted to produce a least-squares fit to the overall compaction trend (see Figure 2 and Supporting Info for calculations). In addition to reporting the measured volume changes, a dimensionless dilatancy parameter ε is reported for comparison between literature dilatancy data (Samuelson et al., 2009;Segall et al., 2010). ε normalizes the measured volume change (Δvol) to both the total sample volume (vol) and the step change from initial slip velocity (V 0 ) to the increased slip velocity (V): (4) This removes the effects of sample volume and the scale of velocity step change from ε and allows direct comparison between published datasets.
Rate and state frictional parameters of the gouges were determined using the program RSFit3000 (Skarbek & Savage, 2019) that uses a non-linear least squares routine (Reinen & Weeks, 1993) to fit the experimental data to obtain the direct effect a, evolution effect b, and the slip weakening distance D c from imposed velocity step changes. The stiffness (k) of the apparatus is also a fitting parameter (Noda & Shimamoto, 2009). In RSFit3000, the state evolution was described using the aging law (Ruina, 1983): (5) This experimental framework that measures the gouge frictional strength, stability and dilatancy has been developed so that it can be applied to a wider range of fault gouge compositions.

Results
Prior to the onset of sliding, the samples were pressurized, which resulted in them compacting hydrostatically. Upon commencement of sliding, a phase of shear enhanced compaction occurred (Figure 3). The majority of shear enhanced compaction occurred prior to the yield phase of the tests. The kaolinite-rich gouges saw less compaction prior to the yield phase than the kaolinite-poor gouges, which is in line with previous studies that showed that more pre-shear hydrostatic compaction occurred in the clay-rich gouges (e.g., Crawford et al., 2008).  All gouges continued to compact after yield, although at a slower rate. The total amount of compaction, shown in Figure 4, was sensitive to the kaolinite content of the gouge. Increasing the kaolinite content led to a decrease in overall compaction, such as at 60 MPa σ n eff the 0 wt% kaolinite gouge experienced 106 mm 3 of compaction compared to 79 mm 3 of compaction for the 100 wt% kaolinite gouge. Increasing σ n eff decreased the overall shear enhanced compaction only in the clay-rich samples. The 100 wt% kaolinite gouge underwent 96, 87 and 40-70 mm 3 pore volume reduction across the increasing effective normal stresses (Figure 4).
Dilation and compaction in response to velocity step changes that deviated from the overall compaction trend were observed in all the quartz-kaolinite gouges ( Figure 5). Dilation always followed an increase in slip velocity and compaction always followed a decrease in slip velocity. The compaction following a velocity decrease was greatest in the 30 wt% kaolinite (70 wt% quartz) gouge at 10 MPa σ n eff , but compaction was fairly consistent across the gouge compositions. The dilation was greatest at kaolinite contents of 10-40 wt% with a decreasing trend toward the two end-member gouge compositions. Unlike the shear enhanced compaction results, there was no effect from varying the σ n eff on the dilatancy behavior. The dilatancy was asymmetric as pore volume increases after velocity upsteps were always greater than pore volume decreases following velocity downsteps. For example, at 25 MPa σ n eff the 10 wt% kaolinite had a mean dilation of 0.88 mm 3 and a mean compaction of 0.25 mm 3 .
All of the synthetic gouges reached a steady state yield strength following the initial 2 mm of load point displacement ( Figure 3). The frictional strength of the synthetic gouges showed a strong dependence on the kaolinite content ( Figure 6). The frictional strength of the gouges at the end of the tests, as shown in Figure 6, decreased with increasing kaolinite content from 0.689 for 0 wt% kaolinite to 0.275 for 100 wt% kaolinite at 60 MPa σ n eff . This negative trend was non-linear-at 60 MPa σ n eff frictional strength decreased steeply to 40wt% kaolinite, beyond which the decreasing trend shallows. At the lower σ n eff conditions, the transition from a shallowly decreasing trend to a steeper decreasing trend in frictional strength occurs at ∼30 wt% kaolinite.  The frictional parameter (a-b) was also strongly dependent on the kaolinite content (see Figure 7). Only gouges with 0 wt% kaolinite displayed velocity weakening behavior in response to velocity steps, with the test at the highest effective normal stress (60 MPa) producing stick-slip instabilities at the end of the test (Figure 3). All gouges that contained any amount of kaolinite showed velocity strengthening behavior. As with the gouge strength, the (a-b) results showed a change in sensitivity to gouge clay content. Below 40 wt% kaolinite, (a-b) increased with increasing kaolinite content but beyond 40 wt% kaolinite (a-b) was less dependent on the kaolinite content. The stability of the synthetic gouges was also strongly affected by the effective normal stress, as shown in Figure 7. Decreasing σ n eff decreased b to negative values, leading to larger (a-b) values and greater stability of sliding in the gouges. For example, in the 100 wt% kaolinite gouge (a-b) increased from 0.003 at 60 MPa σ n eff to 0.025 at 10 MPa σ n eff . Samples of the synthetic gouges were recovered post-shearing for observations of the microstructures using a scanning electron microscope (Figure 8). The gouge microstructures showed a change with increasing kaolinite content from localized shear zones to more diffuse deformation across the gouge layer. Fractures were classified as shear fractures only if grain size reduction via cataclasis could be identified. In the clay-poor (0-10 wt% kaolinite) gouge samples, there was evidence of grain comminution and cracking in localized shear zones, as   shown by the significant decrease in quartz grain size (Figures 8a and 8b). Shear across the layer had been accommodated by a series of R 1 Riedel and shear-direction parallel Y-shears. The Y-shears were located on the periphery of the gouge layer, close to the interface with the direct shear slider surface (sometimes called boundary or B shears). As the kaolinite content increased, the quartz grains became increasingly more isolated and the microstructure changed from a clast-supported to a matrix-supported framework. From intermediate kaolinite contents (>30 wt%), Riedel shears and evidence of grain comminution became increasingly rare. The lack of a throughgoing Y-shear was shown by the 80 wt% kaolinite sample in Figure 8g, as the sawtooth interface between the shear slider and the gouge had been preserved.

Controls on Dilatancy
Previous measurements of dilatancy in granular materials have focused on quartz-rich materials. The data set presented in this work greatly expands the compositional range of measurements to include clay-rich and clayonly materials (Figure 9). The results in Figure 9 show that gouge composition is a factor on the magnitude of dilatancy. Figure 9 also shows that maximum ε occurs not in the quartz-only gouges, but in the quartz-clay mixtures with 10-40 wt% clay. Changes in composition of a gouge incorporates multiple factors that may affect the scale of dilatancy, such as grain size, grain shape (aspect ratio) and roughness, although the scale of dilatancy in this study is comparable to measurements by Zhang et al. (2020) on smectite-quartz mixtures. The dilatancy of the very fine-grained (<2 μm) clay material was very similar or exceeded the dilatancy in the coarser-grained quartz. The maximum dilatancy at intermediate clay proportions likely results from changes in the mechanisms that accommodate shear, as compared to the end member gouges. First, the smearing of a weak phase along grain contacts reduces the shear strength of a granular material, even at small proportions (5 wt%) of the weak phase (Rutter et al., 2013). The addition of very fine-grained clay may act to disrupt quartz-quartz contacts through grain sliding on the clays. The mechanisms of grain rolling and dilation would be promoted over cracking of the quartz grains. Second, Rathbun and Marone (2010) proposed dilatancy increases with the width of the zone of active shear. The proportion of kaolinite to quartz increases the width of the active shear zone. The Min-U-Sil quartz powder forms highly localized R 1 Riedel shears in triaxial deformation experiments (e.g., , however, with increasing kaolinite content, the gouges deform in broader, less localized shear zones and localized shears become rare (Figure 8). Dilatancy may be greater in the kaolinite-rich gouge mixtures due to the more distributed shear that occurs across the gouge layer in comparison to the quartz-only gouge.
The dilatancy results also showed a distinct asymmetry between the scale of dilation and compaction for velocity up-steps versus down-steps ( Figure 5). The quartz-only gouges showed equivalent compaction and dilation in response to velocity step changes, whereas all kaolinite-bearing gouges showed greater dilation than compaction. The asymmetry in the dilatancy results suggests that the strain rate history of the gouge affects the overall pore volume change in a sample. A velocity-stepping experiment produced less overall compaction than a constant velocity experiment, due to this asymmetry. In the measurements by Rathbun and Marone (2013) on quartz powders and clay-bearing glacial tills, both symmetry and asymmetry in the frictional curves and volume changes were observed. The asymmetry in their data is the opposite to that seen in this study, with compaction greater than dilation. Rathbun and Marone (2013) posit that dilation occurs only in the active slip zone and requires both shear strain and interparticle slip, but compaction occurs across the whole layer and only requires a change in stress to occur. The broad and diffuse microstructures in the kaolinite-bearing gouges of this study (Figure 8) may incorporate a higher proportion of the sample that dilates at a slip velocity step increase. The inversion of the asymmetry may result from the higher clay content and different grain size ranges in the gouges of this study, as this subsequently created differences in the gouge microstructures to the experiments done by Rathbun and Marone (2013).
Grain packing models (e.g., Lupini et al., 1981;Marion et al., 1992;Takahashi et al., 2007) can be used to understand the changes in frictional strength of a gouge due to the gouge's composition. For gouges with low clay contents, an interconnected framework of quartz grains supports the applied stresses and governs the gouge deformation (Ikari et al., 2007;Logan & Rauenzahn, 1987;Takahashi et al., 2007). With increasing clay content, the available pore space within the quartz grain-supported framework will be filled by the finer-grained clay (see Figure 8). Under zero pressure the porosity of the quartz is ∼40% and the clay ∼70%. Hence at ∼40 wt% clay, the clay grains fill the available pore space and the gouge strength cannot be supported by quartz grain contacts alone (Takahashi et al., 2007). At clay contents of >40 wt%, quartz-quartz contacts become rare and the applied stresses are supported by an interconnected clay matrix (see Figure 8). When pressure is applied both end members compact, so that the porosity minimum migrates from 40 wt% clay to lower values (Crawford et al., 2008). A 3 phase decrease in frictional strength observed for smectite mixtures was explained using a transitional regime between the quartz-and clay-dominated regimes, which was not observed here (Lupini et al., 1981;Tembe et al., 2010). However, the porosity measurements by Marion et al. (1992) show that porosity is at a minimum between 20 to 40 wt% clay in a quartz sand-kaolinite mixture at effective stresses of 10-50 MPa. This broadly fits with the transition in the frictional data from the grain-supported framework of quartz to the matrix-supported framework of clay at ∼30-40 wt% clay in this study.
Across various literature sources, apparatus and experimental parameters, there is consensus that clay minerals typically display velocity strengthening behavior (Figure 11). The consistency of the (a-b) parameter beyond ∼40 wt% kaolinite indicates that the frictional response to a velocity-step change is strongly influenced by the kaolinite (Figure 7) and is supported by the microstructural transition from a localized quartz-dominated fabric to a distributed clay-dominated fabric (Figure 8). Quartz is a typically velocity weakening material but the addition of only 10 wt% kaolinite, a typically velocity strengthening material, is enough to change the gouge to velocity strengthening behavior. In contrast to the frictional strength results, the (a-b) parameter shows not only a strong positive dependence on kaolinite proportion but also on σ n eff .  report a similar dependence of (a-b) on σ n eff in quartz powders as is observed in this study. As σ n eff increases, the friction parameters become more velocity weakening and stick slip behavior becomes prevalent, although this is not consistent across all reported data. Saffer and Marone (2003) and Zhang et al. (2020) observed negative correlation of (a-b) with σ n eff in smectite-quartz mixtures (but not in illite-quartz mixtures), although these tests were conducted at room humidity without pore pressure.  demonstrated that changing σ n eff had a less significant impact on (a-b) than changing the pore fluid pressure in clay-bearing fault gouges. By increasing the pore fluid pressure, (a-b) also increased-at pore fluid pressures above 25 MPa, changes in σ n eff had little effect on (a-b) . The change in phyllosilicate behavior between dry and wet conditions, which is significant in kaolinite, lizardite and chlorite, has been well documented and may account for the opposing trends in the literature Behnsen & Faulkner, 2012;Han et al., 2020;Moore & Lockner, 2004;Saffer & Marone, 2003).
In this study, the direct effect (a) remained constant at ∼0.005 at clay proportions >40 wt% across all σ n eff . The increased (a-b) parameter at lower σ n eff was due to changes in the evolution effect (b), which decreased to strongly negative values, with a minimum of −0.011. Negative b values do not conform to the interpretation of the evolution effect deriving from changes in the contact area (Dieterich, 1979;Li et al., 2011). If the evolution effect is  Moore and Lockner (2011), and [11] Giorgetti et al. (2015). a result of changes in the contact area, contact saturation would occur at b = 0 so that a negative b value would imply that the contact area of the grains exceeds saturation (Ikari et al., 2009). Negative b values therefore lend weight to the argument that the frictional aging of grain contacts, through processes such as chemical bonding, controls the friction evolution effect in clay-rich gouges (Li et al., 2011). Mineral surfaces of phyllosilicates commonly have a high surface charge, and so can produce grain contacts with variations in strength dependent on the presence or absence of absorbed water (Han et al., 2020).

Implications of Gouge Dilatancy
Several studies have incorporated dilatancy into models for earthquake nucleation and propagation (e.g., Dal Zilio et al., 2020;Liu, 2013) and rate and state friction frameworks (Ferdowsi & Rubin, 2020;Heimisson et al., 2021;Rathbun & Marone, 2013;Segall & Rice, 1995). In the results presented here, only the quartz gouges showed velocity-weakening behavior and could therefore nucleate earthquakes. Consequently, for earthquake nucleation, the pore pressure changes in clay-poor (quartz-rich) gouges are more significant. The dilation in response to acceleration of slip would cause a pore pressure drop and any potential instability would be damped. Dilatancy in clay-rich gouges would lead to more prolonged pore pressure changes than in clay-poor gouges, because low permeabilities in clay-rich gouges tend toward undrained conditions. However, the weak, velocity strengthening clay-rich gouges will not nucleate instabilities. Instead, dilatancy increases the capacity of more clay-rich regions of faults to arrest growing slip patches during nucleation and earthquake propagation. Current models of fault seismicity have incorporated dilatancy parameters on the order of 10 −4 to investigate the role of dilatancy in the seismic cycle (Dal Zilio et al., 2020;Heimisson et al., 2021;Liu, 2013;Segall et al., 2010). This study has measured dilatancy in the range of 2 × 10 −4 to 1.2 × 10 −3 across the full compositional range of quartz to clay content.
In previous investigations (Rathbun & Marone, 2013;Segall & Rice, 1995), attempts have been made to describe the stability of fault slip using both the rate and state friction parameters and the dilatancy parameter. An example from Segall and Rice (1995) defines the stability parameter (E): where β is the compressibility of the material. If E < 1 − a/b it indicates that, under the set conditions, the fault has the potential to host an earthquake, whereas a value of E > 1 − a/b indicates that quasi-static slip behavior would occur (Segall & Rice, 1995). An alternative method to calculate the effect of dilatancy on fault slip behavior is to relate the change of steady state shear stress (τ ss ) with the natural log of slip velocity (v) to the steady state frictional strength of the material (μ ss ) (Segall & Rice, 1995): In order to produce instability, this parameter would need to be negative and bring the curve below a critical stiffness criterion (k crit ) that is defined for an undrained system as: Use of these parameters is based on the assumption of velocity weakening materials. Hence the approach has limited applicability for the majority of this data set, which is dominated by velocity strengthening characteristics. However, the velocity strengthening behavior and lack of unstable slip may be, in part, an effect of the low temperature conditions of the tests. The aseismic creep observed for clay-bearing material in rock deformation experiments is at odds with the observations that most mature (and hence likely clay-bearing) fault zones are also seismogenic (e.g., Chester et al., 1993;Jefferies et al., 2006;Logan & Chester, 1987;Toy et al., 2015;Wibberley & Shimamoto, 2003). A limited quantity of experimental data indicate that clay-bearing gouges become less stable with increasing temperature (Boulton et al., 2014;den Hartog et al., 2013;Verberne et al., 2010).
The potential effect of compaction weakening is not considered in most stability analyses. Faulkner et al. (2018) showed that compaction of gouge in experiments can produce pore pressure transients that affect the gouge mechanical strength and also the determination of the rate and state parameters. In nature, if porosity is 'reset' due to fluidization in earthquakes, there is potential for compaction to occur with fault slip during earthquake nucleation. The dilatancy effects described in this work, produced by slip velocity changes, would be superimposed on these slip-dependent porosity changes. Consequently, the interplay of compaction and dilatancy has the potential to produce quite complex pore pressure development during accelerating fault slip.

Conclusions
A series of velocity-step deformation experiments were conducted on synthetic quartz-kaolinite fault gouges to investigate the controls on pore volume changes in fault gouges. The kaolinite content in the gouges was controlled at 10 wt% increments to test the effect of composition on the frictional strength, stability and dilatancy of the gouges. The pressure conditions were set at 60, 25, and 10 MPa σ n eff to simulate a fault system at hydrostatic to overpressured conditions. The experimental results include that: • All of the quartz-kaolinite gouges experienced dilation in response to an increase in slip velocity and compaction in response to a decrease in slip velocity. • The scale of dilatancy in response to velocity-step changes was affected by the gouge kaolinite content but not by the effective stress conditions. • Increasing kaolinite content in the synthetic gouges decreased the gouge frictional strength and increased (a-b) to promote stable sliding rather than unstable, earthquake slip. • Increasing effective normal stress increased the likelihood of nucleation of unstable slip by decreasing the rate and state friction parameter (a-b).