Narrow, highly-comminuted shear localization features in faults, known as principal slip zones (PSZs), are commonly associated with large-offset seismogenic faults. In this study, laboratory friction experiments were performed using shale and slate gouges where deformation was encouraged to localize at the gouge–wall-rock boundary. The slate gouges develop a black, narrow PSZ composed of densely packed submicron particles that appear sintered while the spectator gouge remains largely undeformed. These PSZs form at subseismic slip velocities of ∼10−5 m/s and with a calculated temperature rise of ∼3 °C. Instances of velocity-weakening friction, which is necessary for unstable fault slip, are only observed for slate samples with a PSZ; shale gouges, however, do not develop a PSZ and exhibit only velocity-strengthening frictional behavior. The development of a PSZ may therefore be a prerequisite for future earthquake slip to occur, rather than unequivocal evidence of past earthquake slip.


A long-standing problem in earthquake science is conclusively identifying evidence of past coseismic slip in the rock record; currently, only the presence of preserved quenched rock melt, or pseudotachylyte, is accepted as such (Cowan, 1999; Di Toro et al., 2005). Other potential indicators of seismic slip are mostly geothermometers, including decarbonation reaction products (Han et al., 2007; De Paola et al., 2011), vitrinite reflectance (Sakaguchi et al., 2011), biomarker ratios (Savage et al., 2014), and mobile element geochemistry (Ishikawa et al., 2008). These indicators are promising, but must be identified in laboratory analyses, are not always present, and in some cases are still controversial (e.g., Fulton and Harris, 2012).

Highly localized, narrow comminuted zones of ultracataclasite termed principal slip zones (PSZs) are common features of major, large-offset faults inferred to have hosted coseismic slip (e.g., Chester et al., 1993; Sibson, 2003; Smith et al., 2011; Rowe et al., 2013). Friction experiments in which meter/second slip rates are employed also commonly show boundary shear localization features as PSZs (e.g., Ujiie and Tsutsumi, 2010; De Paola et al., 2011; Bullock et al., 2015). These studies suggested that certain characteristics that form at high velocity, such as black fault gouges (Balsamo et al., 2014), shiny mirror-like surfaces (Fondriest et al., 2013; Smith et al., 2013), silica gel (Di Toro et al., 2004), or clay-clast aggregates (Boutareaud et al., 2010), require coseismic slip velocities and are therefore unambiguous indicators of coseismic slip when observed in the field. However, it has also been demonstrated that shiny localization surfaces (Verberne et al., 2013, 2014) and clay-clast aggregates (Han and Hirose, 2012) can be reproduced at subseismic velocities, indicating that these are not reliable as coseismic slip indicators. If PSZs in general can be similarly formed at low slip velocities, they may also cause misinterpretation of fault zone earthquake history. Furthermore, many of these high-velocity features tend to be observed in quartz or calcite gouges rather than in phyllosilicate-rich material that is common in many fault zones. Causative relations, or whether the formation of PSZs precedes dynamic weakening and vice versa, are also uncertain. In this study, laboratory friction experiments were performed on shale and slate gouges sheared at low velocity and to high shear strains, in which boundary shear localization was encouraged by using a smooth rock substrate. Mechanical behavior of these laboratory faults was then analyzed by correlating the friction velocity dependence with microstructural features.


Laboratory friction experiments were conducted in a biaxial double-direct shear apparatus at room temperature and humidity to ensure successful sample recovery and imaging. The tested gouge samples were Pennsylvania (USA) slate and the Rochester Shale (New York, USA), powdered to a grain size of <125 µm. X-ray diffraction shows that the Pennsylvania slate is composed of 54% phyllosilicate minerals (40% illite, 14% chlorite), 31% quartz + feldspar (albite), and 15% calcite + dolomite. The Rochester Shale is composed of 68% phyllosilicate minerals (59% illite, 9% kaolinite and/or dickite) and 27% quartz + plagioclase. These particular materials were chosen to match the clay mineral content of natural faults (e.g., Vrolijk and van der Pluijm, 1999; Underwood, 2007); the slate represents a slightly higher metamorphic grade. Grooved steel was used for the two outer forcing blocks, and the center block was made of rock [Westerly Granite (eastern USA), Etna basalt (Italy), or Pennsylvania slate] polished with a 60 grit grinding wheel, in order to encourage shear localization at the gouge–center block interface (Fig. 1). The rocks used as center blocks were chosen to simulate appropriate crustal protoliths, and because clay-rich gouges can form even when the protolith is phyllosilicate poor (e.g., Chester et al., 1993; Jefferies et al., 2006). Gouge layers were prepared to a uniform thickness (3 mm prior to loading) and sheared under a normal stress of 20 MPa using a displacement rate boundary condition (11 µm/s) imposed at the edge of the gouge layer. The final shear strain range is γ = 92–102, where the bulk engineering shear strain γ = Σ(Δd/h); d is the displacement and h is the instantaneous layer thickness.

In each experiment the shear stress τ was continuously measured and used to calculate a sliding coefficient of friction µ from the ratio of τ to normal stress σn, assuming that cohesion is zero (Handin, 1969). At target shear strains of γ < 5 and ∼25, 45, 75, and 95, the velocity (or rate) dependence of friction was measured by employing steps of threefold increasing velocity in the range V = 1–300 μm/s (Fig. 1A). For each individual velocity step the friction rate parameter is calculated as a-b = Δµ/ΔlnV using inverse modeling techniques (Reinen and Weeks, 1993; Ikari et al., 2011). Values of a-b > 0 indicate velocity-strengthening behavior associated with stable fault creep, and a-b < 0 indicates velocity weakening, which is required for fault slip instability and earthquake nucleation (Dieterich, 1981; Marone, 1998; Scholz, 1998).


Steady-state residual friction coefficients for Rochester Shale range from µ = ∼0.48 to 0.55. Effects of the center block lithology are small, but on average samples sheared against Etna basalt are the weakest (µ = 0.50), while shearing against Pennsylvania slate (µ = 0.52) and Westerly Granite (µ = 0.54) results in slightly higher values. Pennsylvania slate gouges are stronger than the Rochester Shale, with values of µ ranging from ∼0.55 to 0.64. Similar to Rochester Shale, samples sheared against Etna basalt are slightly weaker (average µ = 0.58) compared to those sheared against Westerly Granite or Pennsylvania slate (average µ = 0.61 for both).

For Rochester Shale gouges, strictly velocity-strengthening frictional behavior is observed, with a-b ranging from 0.0010 to 0.0048 (Fig. 1B). In general, little dependence can be observed on shear strain or substrate (center block) lithology, although the lowest a-b values tend to be observed for sliding against Etna basalt. However, some velocity weakening is observed for Pennsylvania slate gouges (Fig. 1C). This occurs mostly for experiments where the substrate was Westerly Granite at shear strains γ > ∼20, exhibiting a-b = −0.0010 to 0.0027. Values of a-b for slate gouge sheared against Etna basalt range from −0.0004 to 0.0055, with the velocity weakening occurring at γ = ∼25. When slate is also the substrate, slate gouge is strictly velocity strengthening, with a-b ranging from 0.0005 to 0.0036. Both the Rochester Shale and Pennsylvania slate exhibit a general trend of increasing a-b with increasing sliding velocity. These results are consistent with previous experimental work using these specific gouges under similar conditions (Saffer and Marone, 2003; Ikari et al., 2011).


The most striking structural observation in this study is that for all experiments using Pennsylvania slate gouge, a narrow (≤∼50 µm thick) dark layer, which is interpreted here to be a PSZ, develops at the boundary between the bulk spectator gouge and the smooth center forcing block (Fig. 2). The PSZ recovered at the end of the experiment has a dark color and the presence of a small cohesion that clearly distinguish it from the spectator gouge; these features do not develop in any experiments with Rochester Shale. One sample of Pennsylvania slate was sheared against Westerly Granite at a constant 11 µm/s, without conducting velocity steps, specifically for imaging. The spectator gouge that makes up the bulk of the sample is relatively coarse grained with several tens of micron-sized grains surviving (Fig. 2B). Major structural features are absent [other than a few crosscutting R1 (Riedel) shears], but some grain size reduction can be seen at the lower boundary of the spectator gouge, where it was in contact with the PSZ. The PSZ, however, is extremely comminuted, so that no individual grains are visible at magnifications to 20× (Fig. 2C). Despite its narrow width the PSZ is clearly foliated and in some cases exhibits crosscutting R1 shears (Figs. 2C, 2D). This visual evidence confirms the PSZ as a shear localization feature in these experiments.

Scanning electron microscopy images under higher magnification show that both the Pennsylvania slate starting gouge material and sheared spectator material consist of infrequent, larger grain aggregates several tens of microns in size and, more commonly, micron-scale phyllosilicate flakes (Fig. 3). The PSZ is imaged here as a dense flake composed of closely packed, rounded grains of submicron size, some as small as tens of nanometers. In many places, the grains are so tightly packed that porosity is no longer visible and they appear to be fused together. In contrast, the spectator material contains randomly oriented larger surviving grains and angular flakes.


Although the highly comminuted boundary shear PSZs with sintered nanograin texture formed in these experiments have many similarities to PSZs on natural large-offset faults (e.g., Chester et al., 1993; Sibson, 2003; Smith et al., 2011; Rowe et al., 2013) and in shearing experiments at seismic slip rates (e.g., Ujiie and Tsutsumi, 2010; De Paola et al., 2011; Bullock et al., 2015), they are formed at slow slip rates (∼10 µm/s, or 10−5 m/s). This provides experimental support to the assessment of Cowan (1999), who suggested that such features cannot uniquely be interpreted as the product of coseismic slip. Furthermore, despite their sintered appearance, the PSZs in these experiments are not associated with large amounts of frictional heating, because low sliding velocities are not expected to generate significant temperatures (Mair and Marone, 2000). Following Cardwell et al. (1978), the temperature rise, ΔT, in the PSZ at the end of these experiments is calculated for an infinitely thin fault zone as : the shear stress τ = 12 MPa, displacement d = 0.08 m, and shearing duration t = 8000 s are taken from these experiments, and density ρ = 2500 kg/m3, heat capacity c = 850 J/kg °C, and thermal diffusivity κ = 1.1 × 10−6 m2/s are used as appropriate values for slate (Robertson, 1988). This ΔT = ∼3 °C; therefore, the grains are probably cold pressed into dense aggregates and are not welded by locally high temperatures. Sintering of nanoparticles in calcite gouge by solution-transfer mechanisms is possible at room temperature and with small amounts of adsorbed water (2–3 wt%) (Verberne et al., 2014), although it is unconfirmed whether this is also a viable mechanism for phyllosilicates.


The friction data in this study suggest that formation of a PSZ favors frictional instability, because velocity-weakening frictional behavior is only observed in cases where a PSZ has developed. Rochester Shale gouges that do not exhibit a PSZ only exhibit velocity strengthening. Furthermore, no velocity weakening is observed at low strain (γ < 5) for the Pennsylvania slate gouges. Correspondingly, there is no evidence of a black PSZ in extruded, low-strain gouge that is left on the trailing portion of the forcing blocks from the early part of the tests. In this case, deformation occurs by shearing on R1 shears, as observed in previous studies using these gouges (Ikari et al., 2011). Collectively, this suggests that the formation of a PSZ in phyllosilicate-rich gouge represents an intermediate step in the evolution toward frictional instability. The absence of a PSZ indicates a creeping fault, but the development of a PSZ, and associated velocity-weakening frictional behavior, is one factor (along with appropriate elastic properties and effective stress conditions; Scholz, 1998) necessary for coseismic slip in the future.

The inference that PSZs are a precursor to coseismic slip is consistent with previous experimental work demonstrating that a critical shear strain or displacement is required for velocity weakening to occur, and is associated with shear localization at the gouge-forcing block boundary (e.g., Logan et al., 1992; Beeler et al., 1996; Scruggs and Tullis, 1998). The results of this study are also similar to recent experiments on calcite gouge by Verberne et al. (2013, 2014), who demonstrated that the shiny surfaces in calcite gouge are composed of aggregates of nanoparticles generated at subseismic slip rates that sometimes (but not always) exhibited velocity-weakening friction. Other recent laboratory experiments show that shearing unconsolidated quartz + feldspar sand produces an extremely fine-grained black gouge at coseismic velocities (1 m/s), while lower rates (100 µm/s) produce a poorly developed cataclastic texture (Balsamo et al., 2014). For the high-velocity experiments in which black gouges were observed, dynamic weakening occurred after at least ∼20 cm displacement. In this study, the PSZ developed at total displacements of <6 cm and possibly as low as ∼2 cm; therefore, the black gouge PSZs observed by Balsamo et al. (2014) may have formed as an intermediate step before the onset of dynamic weakening, consistent with the idea that a highly comminuted boundary shear PSZ may be a necessary prerequisite for earthquake slip.

I gratefully acknowledge Chris Marone for providing support and helpful discussions, Brett Carpenter and Andre Hüpers for discussions and comments on an early version of this manuscript, and Petra Witte and Daniel Hepp for assistance with imaging. John Logan, Francesca Meneghini, and an anonymous reviewer provided helpful comments and suggestions. This work was partially supported by National Science Foundation grants OCE-0648331, EAR-0746192, and EAR-0950517.