The evolution of fault strength and behavior during the initial stages of slip plays an important role in driving the onset of instability and fault weakening. Using small-displacement triaxial experiments on quartz sandstone, this study highlights the rapid onset of microstructural change on fault interfaces and identifies new evidence for an evolution in physical processes with increasing slip and velocity. Pre-ground fault surfaces have been slipped over a range of velocities (0.36 µm s–1 to 18 cm s–1) and at normal stresses comparable to upper- to mid-crustal conditions (92–287 MPa). Microstructural analysis of the fault interfaces reveals the formation of amorphous material at displacements <170 µm and slip durations < 1 ms. Mechanical and microstructural observations have been combined with numerical modeling to present the first documented examples of a transition from mechanical amorphization to flash heating, then frictional melting, with changes in slip conditions. The sequence of processes activated during the initial stages of fault movement may provide new insights into factors that influence the onset of slip in the seismogenic crust.
During slip, extreme transient changes in the thermal and mechanical conditions on the fault surface induce microstructural and phase changes at the sliding interface (Pec et al., 2012; Spray, 1992; Yund et al., 1990). In recent years, attention has focused on the use of experiments to identify mechanisms that induce fault weakening during earthquake-producing slip (Di Toro et al., 2011). Most experiments have used high-velocity rotary shear apparatus to impose seismic slip velocities on a variety of rock types, but at lower normal stress (σn) and larger displacements than are typical of most earthquakes in the mid- to lower continental seismogenic regime. However, these experiments have revealed a complex evolution of fault strength at high slip velocities, leading to the identification of a number of mechanisms that may contribute to fault weakening. Mechanisms identified in dry quartz-rich samples include the formation of silica gels (Di Toro et al., 2004), flash heating at asperity contacts (Rice, 2006; Goldsby and Tullis, 2011), lubrication by very fine-grained wear products (Reches and Lockner, 2010), and frictional melting (Di Toro et al., 2006). Despite the advances in understanding high-velocity frictional sliding, the physical processes that induce the initial stages of fault weakening remain poorly constrained by rotary-shear experiments due to the technical difficulty of achieving high slip velocities at realistic σn over displacements <1 mm (Niemeijer et al., 2012).
A limited number of studies have attempted to explore the early stages of fault rupture using a traditional triaxial arrangement to initiate high-velocity sliding during stick slip (Koizumi et al., 2004; Passelègue et al., 2013; Brantut et al., 2016). In this study, we use a novel experimental setup and high-resolution microstructural analysis to investigate the behavior and physical development of quartz interfaces during the early stages of slip. These experiments were performed at both high confining pressures (Pc = 50–200 MPa) and elevated ambient temperatures (400–900 °C), providing access to both stable sliding and stick-slip behaviors in our samples.
Deformation experiments were conducted using a Paterson-type gas medium apparatus (see Item DR1 in the GSA Data Repository1). The sample assembly consists of cylinders of Fontainebleau sandstone ground with a fault oriented at 30° to the shortening direction and contained within a thin-walled steel jacket. Thirty-two experiments were conducted (see Table DR2 in the Data Repository) using a constant axial displacement rate between 0.36 and 0.72 µm s–1. Mechanical data were recorded using a digital data acquisition system at a frequency of 1 Hz. A new system of high-speed (megahertz frequency) data acquisition was developed for measuring displacement during rapid slip events. This system uses a fiber-based interferometric sensor to provide non-continuous, triggered data acquisition. High-resolution displacement and slip velocity measurements were obtained during four experiments with a total of 21 triggered rapid slip events (see Table DR3).
Experiments were halted at varying displacements to allow microstructural analysis of the evolving fault surfaces. Slip surfaces were examined using secondary electron (SE) imaging in a Zeiss UltraPlus field emission scanning electron microscope (FE-SEM). Samples for cathodoluminescence (CL) analysis were prepared by impregnating the jacketed samples with epoxy before producing a longitudinal section perpendicular to the fault plane. Imaging and spectral analysis were undertaken using a FEI Verios FE-SEM fitted with a Gatan MonoCL4 Elite cathodoluminescence system. Selected areas of the fault surface were prepared for transmission electron microscopy (TEM) analysis using a Helios NanoLab 600 Dualbeam FIB-SEM. TEM was undertaken using a Philips CM300 TEM operating at 300kV.
Maximum frictional heating is estimated using a flash heating model (Rice, 2006; Brantut and Platt, 2016) that assumes that the local temperature increase at stressed asperity contacts exceeds the bulk temperature rise for the slip surface. Assuming that the maximum temperature (Tmax) is achieved if asperities are in contact for the duration of the slip event (t), the asperity temperature is given by:where Tamb is the ambient temperature, τc is asperity strength, estimated to be 10% of the shear modulus (∼3.1 GPa; Rice, 2006), V is the experimentally derived velocity, r is the material density (2650 kg m–3), c is heat capacity (1082 J kg K–1), and k is the thermal diffusivity (2.23 mm2 s–1). All parameters are given for quartz (see Item DR1 for model sensitivity to parameter variability).
Experiments were characterized by an initial period of elastic loading, before reaching a poorly defined yield point and subsequent period of strain hardening. Peak shear stresses between 71 and 228 MPa were followed by a transition into either (1) stable sliding or (2) episodic rapid slip with stress drops of 19–127 MPa and slips d = 15–158 μm (Fig. 1A). The high-frequency displacement data further define a subdivision of the stick-slip experiments into low-velocity, small-amplitude slip events and high-velocity, large-amplitude slip events (Fig. 1B). A conservative estimate of the maximum slip velocity has been made by taking the derivative of a logistic function fitted to the displacement data (see inset, Fig. 1B; Item DR1). Using this approach, slip velocities can be grouped according to the different slip regimes: (1) stable sliding, with slip velocities V = 0.36–0.72 µm s–1; (2) episodic low-velocity slip, V » 1–10 cm s–1; and (3) episodic, large-amplitude slip events, V » 10–18 cm s–1. These behaviors were observed over a range of Tamb and Pc. Up to ∼600 °C, slip behavior is characterized by stable sliding in agreement with previous experimental studies made at moderate to high σn (Friedman et al., 1974; Paterson and Wong, 2005). Above Tamb = 600 °C, the fault interfaces develop stick-slip behavior. The comparatively low peak σn (93–193 MPa) associated with high Tamb and low Pc produces low-velocity slip, whereas high-velocity slip occurs when faults have higher σn (178–287 MPa) at high Tamb and high Pc (Fig. 1C).
The three slip behaviors identified by the mechanical data are each defined by a distinctive microstructural style. For experiments characterized by stable sliding, fault surfaces are dominated by fracturing and the formation of copious gouge particles, ranging in size from <20 nm to 20 μm. In localized regions, the fault surface contains apparently non-crystalline material interspersed between 10 and 200 nm quartz clasts, forming a film-like matrix (Figs. 2A and 2B). This material is similar in appearance to the gels reportedly produced during high-velocity sliding experiments on quartz interfaces (Di Toro et al., 2004). For experiments experiencing episodic low-velocity slip events, the interface is characterized by gouge-covered surfaces and the formation of debris-free patches that are conchoidally fractured sub-parallel to the fault surface; the patches have rare lobate edges (Fig. 2C). The patches are generally <10 µm in diameter, and many are bordered by closely spaced fractures formed perpendicular to the slip direction. In contrast, episodic high-velocity slip events produce textures that are recognizable as glass and consist of stretched ribbon-like filaments and flow textures that form by the interaction of asperities with a viscous fluid film (Fig. 2D). The glass regions form a discontinuous layer up to 2 µm thick and cover between 10% and 60% of the fault surface. Continuous lineations formed in the melt surface extend for >150 µm, indicating that melting commenced within the first tens of microns of slip.
Structural characterization of the distinctive microstructures identified using SEM was undertaken on TEM using FIB milled foils. The gel-like matrix formed during stable sliding cannot be definitively shown to be amorphous due to interspersed crystalline quartz nanoparticles with diameters considerably smaller than the ∼100 nm thickness of FIB section. However, dark-field imaging highlights zones of increased crystallinity, with the relative abundance of the apparently non-crystalline matrix increasing toward the slip surface, signifying a correlation between possible amorphization and increasing shear strain and slip localization. TEM investigations of the patches formed during episodic, low-velocity slip reveal ∼1-μm-thick lenses of non-porous amorphous material on the slip surface (Fig. 2E). These zones contain common entrained crystalline clasts ranging in size from 10 to 200 nm.
The glass textures on high-velocity slip surfaces are confirmed as being amorphous using TEM imaging and selected area diffraction (Fig. 2F). The glassy amorphous material contains only rare crystalline clasts and is compositionally indistinguishable from the quartz substrate. The presence of occasional elongate vesicles, possibly derived from decrepitated fluid inclusions, indicate shearing. However, the apparent structure of the glassy amorphous material, as shown by diffraction patterns, is essentially the same as the amorphous lenses formed during episodic low-velocity slip.
Further exploration of the amorphous material was undertaken using CL-SEM. The emission intensity and wavelength of the CL spectra associated with the glassy amorphous material formed during episodic rapid slip (both high and low velocity) are clearly distinguishable from those of adjacent quartz grains (Fig. 3A; for additional images and spectra see Item DR4). Color banding is evident and represents repeated melting events during multiple slip events. The matrix formed during stable sliding similarly shares the same distinctive spectral signature, although at a much lower intensity, and also shows internal banding. However, unlike the glassy amorphous material formed during stick slip, the banding formed during stable sliding has diffuse edges and is distributed throughout the width of the fault core, forming textures that suggest a fluid-like flow (Fig. 3B).
Cataclastic processes alone do not cause the observed changes in the CL signature of the matrix. Gouge clasts with the same luminescence as the adjoining quartz are abundant in the pores adjacent to the fault and are even dispersed within the amorphous material. Likewise, amorphous fragments are intermingled with the gouge, suggesting cataclastic reworking of previously amorphized material.
Estimating Temperature at Asperity Contacts
For the largest slip velocities, ∼18 cm s–1, the maximum asperity temperature (Tasp) is estimated as ∼2620 °C (Fig. 4), far exceeding the equilibrium melting temperature of quartz (∼1700 °C; Navrotsky, 1994). This estimate is consistent with the occurrence of glass-like textures on the interfaces. In comparison, for the fractured amorphous patches formed during episodic low-velocity slip (V < 10 cm s–1), temperatures are estimated to be in the range Tamb + 100 °C < Tasp < 1700 °C. These results suggest that slip rates in most of the low-velocity regime are insufficient to melt asperity contacts, even assuming metastable melting temperatures (∼1426 °C; Navrotsky, 1994). At the slow slip rates (<1 µm s–1) associated with stable sliding, maximum temperatures are estimated to be only marginally higher than Tamb.
DISCUSSION AND CONCLUSIONS
From the microstructural observations and numerical modeling, the different morphologies of amorphous material on slip surfaces are interpreted to reflect the onset of different damage processes that are strongly dependent on normal stress, slip velocity, and displacement. During high-velocity slip events (V >10 cm s–1), asperity contacts rapidly generate sufficient heat to commence melting the fault interface. The abundance of glass indicates that melt processes have advanced beyond merely flash heating and are transitional to the development a continuous melt film (Hirose and Shimamoto, 2005). A threshold value of V > 10 cm s–1 is identified as the minimum velocity required to frictionally melt quartz asperities at Pc = 100MPa and is consistent with theoretically estimated critical velocities for fault weakening by flash heating (Rice, 2006).
A key point to emerge from the episodic low-velocity slip events is that the amorphous material formed in localized patches occurs even when modeled temperatures are significantly lower than the melting point of quartz. It is suggested that while flash heating certainly occurs, other processes such as mechanical (Tkáčová et al., 1993) and shear amorphization (Pec et al., 2012; Wolf et al., 1990) could be critical to the initial loss of crystalline order. Contemporaneous or subsequent frictional heating would allow minor viscous flow of the amorphized asperity contacts if they passed through the glass transition (∼1200 °C).
In contrast to the rapid-slip events, little frictional heating occurs during the stable sliding experiments. In these experiments, the amorphous matrix is interpreted to result from intense comminution of gouge particles resulting in a loss of crystalline structure (Yund et al., 1990). Importantly, at normal stresses comparable to crustal faulting conditions, the formation of this film-like matrix occurs over displacements that are several orders of magnitude less than have been observed in previous high-velocity experiments (Di Toro et al., 2006; Yund et al., 1990) and under drained, nominally dry conditions.
These experiments provide powerful insights into the dramatic structural changes that occur on fault interfaces during the initial milliseconds of fault slip. The ease with which amorphization occurs during small-displacement slip events indicates that amorphous material (whether mechanically induced or frictionally melted) should form during most earthquake ruptures or slow-slip events nucleating at moderate to high σn, corresponding to depths equivalent to the mid- to lower continental seismogenic regime. However, the small volumes and thermodynamic metastability of the amorphous material suggest that it may not be preserved over geological time scales. A greater understanding is needed of both the processes inducing amorphization within fault zones and also the conditions and time periods over which the amorphous material is preserved. Both factors likely have significant implications for the behavior of faults during slip and subsequent reactivation, and also for interpreting microstructures and processes occurring on natural faults.
A. Schubnel generously supplied the Fontainebleau sandstone used in this study. H. Miller, H. Kokkonen, and L. Li are thanked for technical support. FIB- and CL-SEM were undertaken at the Australian National Fabrication Facility. FE-SEM and TEM were completed at the Australian National University Centre for Advanced Microscopy. This study was supported by Australian Research Council grants DP130102687 and FT130100329 (to Slagmolen). We thank G. Di Toro, N. De Paola, and N. Brantut for very helpful reviews.
↵1GSA Data Repository item 2016348, Item DR1 (extended experimental methods), Item DR2 (experimental details), Item DR3 (experimental details of rapid-slip events captured using interferometer), and Item DR4 (additional microstructural images), is available online at www.geosociety.org/pubs/ft2016.htm or on request from .
- Received 15 June 2016.
- Revision received 28 September 2016.
- Accepted 1 October 2016.
- © 2016 Geological Society of America