T21D-01 INVITED
Slip Localization, Fault Weakening and Slow Earthquakes as a Consequence of Fault Gouge Strengthening in Hydrothermal Regimes – Insights From Laboratory Experiments
A laboratory study of simulated quartz gouges was conducted to investigate how solution transfer processes influence the evolution of the mechanical behaviour of fault wear products at high temperature, hydrothermal conditions. Experiments were performed under nominally dry conditions, as well as in the presence of an aqueous pore fluid, at elevated temperatures (500 to 927°C), and at an effective confining pressure of 100MPa to simulate, on a laboratory timescale, processes that may be important in fluid-active fault zones at depth in the continental crust. The mechanical data and microstructural analysis indicate that the kinetics of solution transfer processes can exert a fundamental control on the mechanical behaviour of fault wear products. It is found that, at nominally dry conditions, gouges deform by cataclastic creep and distributed shear, with strength and microstructures being relatively unaffected by temperature. However, at chemically reactive, hydrothermal conditions (600-927°C, small grain size, and slow deformation rate), rapid porosity reduction is accommodated by dissolution-precipitation processes. Deformation under such conditions results in a fast increase of grain contact area and the development of cohesive bonds between adjacent particles, which in turn inhibits cataclastic granular flow. With increasing displacement and compaction of the quartz gouge, there is a transition from distributed cataclastic flow, to slip localization at the interface between the gouge and one of the forcing blocks. This deformation mode switch is associated with dramatic weakening (up to 50% drop in shear resistance, and changes in the apparent coefficient of friction from 0.7 to approximately 0.4). Stress drop occurs over many minutes in the laboratory. It is speculated that solution-assisted gouge strengthening, and consequent slip localization with associated slow, yet dramatic stress drop, could provide a mechanism for the occurrence of slow earthquakes.
T21D-02
The Nature and Evolution of Fluid-Related Weakening Mechanisms Along a Continental Low-Angle Normal Fault: the Zuccale Fault, Elba Island, Italy
Despite extensive research concerning the mechanical significance and geometric evolution of low-angle normal faults, few studies have focussed on the importance of the fault rock material which is generated during prolonged slip and exhumation. The Zuccale fault on the Island of Elba is closely associated with syn- tectonic igneous intrusions, and it possesses a complex fault rock 'stratigraphy' which records the interaction between multiple deformation mechanisms and fluids derived from distinct crustal reservoirs. Optical- and scanning-electron microscopy, combined with XRD and stable isotope analyses, reveal systematic changes in fault rock chemistry and texture: 1) Cataclasis and dissolution-precipitation creep were the dominant deformation mechanisms during the early stages of fault activity. Cataclasis facilitated the influx of chemically active fluids, leading to widespread syn-tectonic growth of weak phyllosilicate minerals, including talc and chlorite. Crystal-plasticity was important within calcite-rich fault rocks which initially deformed at depths of ~6-8km. Calcite grains (~10μm in diameter) possess a strong C-axis preferred orientation, suggesting that they experienced dynamic recrystallisation by dislocation creep. These calcite-mylonites were crosscut by vein material which was progressively sheared and recrystallised, indicating cyclic brittle-ductile deformation. During the later stages of fault activity, granular flow became an important deformation mechanism. Rolling and sliding of grains past one another was accommodated along clay-lined grain boundaries; 2) During exhumation, dolomite was superseded by calcite as the dominant syn- tectonic fault cement. Dolomite veins within the fault core record transiently high fluid pressures followed by mineral sealing and fault strengthening. The δ13CV-PDB signature of such vein dolomite is strongly clustered around a mean value of -5.7‰, whilst δ18OV-SMOW varies between 10‰ to14‰. These values suggest an igneous fluid source, and in this case the carbon appears ultimately to be mantle-derived. Later calcite veins have δ13C V-PDB values of -6.5‰ to -10.5‰ and δ18OV-SMOW between 25‰ and 27‰, suggesting precipitation from a shallow, mixed fluid source containing local marine limestone and organic components. Potential fault zone weakening mechanisms which have been identified along the Zuccale fault include: (i) dissolution-precipitation creep within phyllosilicate-rich fault rocks, (ii) a switch to grain-size sensitive deformation mechanisms within calcite-mylonites, (iii) granular flow accommodated by fine-grained clay phases, and (iv) transiently high fluid pressures which probably occurred over short timescales and across localised fault patches. These weakening mechanisms were not mutually exclusive, but their relative importance must have varied as a function of fault zone structure and composition, local strain rates, and the availability of fluid before, during, and following igneous intrusion.
T21D-03
Fault Rock Hardening as a Precursor to Localized Rupture During Systemic Fault Zone Weakening
Mature fault zones commonly comprise thick zones of previously deformed material within which discrete, seismic discontinuities are localized. That the thickness, composition and accommodating deformation processes within the host zone vary with depth is well established; likewise, the temporal evolution of the fault zone produces a distinct history of behaviour whereby repeated seismic rupture produces comminution at a range of scales, with dilatant behaviour linked to the introduction of fluids and related introduction of new materials. The overall trend is one of macroscopic, distributed material softening, yet significant seismic ruptures continue to develop. Within the Minas fault zone, Nova Scotia, Canada, major transpressive displacement is registered over km-wide length scales, with intense shear strain component focussed within ~200 metre wide zones of broadly ductile deformation containing discrete primary fault rupture surfaces, with oblique secondary faults. Structures within the fault zone can be correlated with interseismic, preseismic and coseismic periods based on temporal and spatial relationships. Interseismic deformation registers as macroscopic ductile deformation of fine-grained sedimentary host rock and gouges developed during earlier fault cycles. Relatively modest, but typical bulk strain rates (e.g. 10-12s-1) are indicated by semi- brittle microstructures, particulate flow and pressure solution. Preseismic deformation is associated with formation of distinct, patterned strain partition; the patterning has been observed at all scales, from both upper and lower crustal material exposed within the core of the fault zone. The patterning comprises development of shear bands or Riedel shears, depending on material and depth, along which pre-seismic displacements are concentrated. Coseismic events witness dilation, fluid ingress and shear, including development of frictional melts. Depending on the depth of faulting, fluid throughput is evidenced by quartz and calcite veins, or partial melts. Dramatic dilation, seen as fragmentation, is largely constrained to the outer boundary of the fault zone core, suggesting strong focussing of fluids (fluid pressure) along the sharp transition from pure shear to simple shear components of the transpression. Overall, there is a requirement for strain hardening within the fault zone as the precursor to the softening concurrent with rupture nucleation, consistent with non-associative flow.
T21D-04 INVITED
Fault Heating and Lubrication During Earthquakes: Experimental Constraints
The determination of fault strength (rock friction sensu latu) at seismic slip rates (about 1 m/s), is of paramount importance in earthquake mechanics, as fault strength controls rupture properties, stress drop, radiated energy and heat produced during slip. Given the lack of determination through seismological methods, elucidating constraints arise from experimental studies. Here we show that a review of the experiments (~400) performed in rotary shear apparatuses at slip rates of 0.1 - 1.3 m/s indicate a significant decrease in friction (up to one order of magnitude) for cohesive (silicate-, quartz- and carbonate- built) and non-cohesive (clay-rich and dolomite gouges) rocks. Low friction is concurrent to an increase in temperature in the slipping zone which triggers thermally-activated physico-chemical processes responsible for fault lubrication (decarbonation and dehydration reactions, flash heating, melt lubrication, etc.). Extrapolation of experimental data to natural conditions, suggests large coseismic stress drops (> 70 MPa) at earthquake nucleation depths (7 - 10 km), irrespective of fault rock composition and of the specific weakening mechanism involved. Such large stress drop estimates are consistent with dynamic stress drops obtained from seismic inversion data and geological studies.
T21D-05
Friction at Seismic Slip Speeds: Experiments and Theory
We present new experimental data and theory that describe the thermal weakening of fine-grained gouges during earthquake slip. We postulate that particles in fine-grained gouges thermally soften due to an intrinsic decrease in the elastic shear modulus in response to rapid heating of the gouge layer. The temperature dependence of the asperity strength is assumed to obey a modified Watchman's equation. In our model, the velocity dependence of the effective coefficient of friction results from the temperature dependence of the theoretical yield strength of the contact asperities, rather than the sudden loss of the asperity strength at some critical temperature. Temperature of individual asperities depends on highly transient dissipation during asperity contacts ("flash heating") as well as the average temperature of the slip zone (which, unlike flash heating, monotonically increases with slip and depends on normal stress). Eventual contact melting can occur depending on the effective normal stress, slip rate, and total slip. Upon reaching the solidus, the residual contact strength is estimated assuming Couette flow for given particle size, slip rate, and melt rheology. We conducted a series of high-speed friction experiments to test the model predictions. Experimental data indicate that there is a systematic evolution of the friction coefficient from ~0.6-0.7 to as low as 0.2 as velocities increase from 0.03 m/s to 2.5 m/s. The inferred power-law exponent of the velocity dependence is ~(-0.4), and the critical weakening velocity appears to depend on the normal stress, consistent with the hypothesis that the observed velocity dependence of friction stems from thermal softening of the asperities. We infer the characteristic grain size using SEM images of the experimentally produced gouge. The grain sizes appear to be power law distributed with the majority of grains less than 1-5 μm in diameter. We calculate the temperature evolution within the gouge layer assuming 1-D non-steady heat conduction and complete conversion of mechanical work into heat. The predicted time history of temperature inside the sample is in excellent agreement with the experimentally measured temperature. We use the approach model to calculate the transient heating of individual asperities, and compare the calculated average strength of the asperities to the observed coefficient of friction (assuming that once in high speed motion changes in the true contact area are negligible). The model reasonably well reproduces the overall evolution of dynamic friction with slip velocity. Laboratory data do reveal an important (10-20%) initial weakening over slip distances of the order of 1 m that cannot be readily explained in terms of purely thermal effects. We argue that the observed initial weakening is probably mechanical in part, due to initial dilation of the gouge at slip initiation and progressive localization of slip within the gouge layer.
T21D-06
Strong Velocity Weakening in Fault Gouges: Results from Rock Analogue Experiments
Fluids are important in deformation processes in the upper- to middle crust where they exert strong influence on frictional behaviour of fault gouges via mechanical (pore fluid pressure) and chemical effects (solution- transfer processes). Despite the importance of hydromechanical effects, not much is known about the interplay of chemical and mechanical processes, in part because the required conditions are difficult to simulate in the laboratory (i.e. high temperature, low strain rate and high strain). We report results from an experimental study of simulated fault gouge composed of rock salt sheared under conditions where pressure solution is known to operate. The experiments extend previous work to higher sliding velocities and allow comparison with rotary shear tests. We find that steady state friction is very similar for both the direct shear and rotary shear configurations (for pure salt gouges in the presence of brine at a normal stress of 5 MPa, slip rates of 0.03-10 μ m/s and shear strains up to 15). However, at sliding velocities higher than previously obtained in the rotary shear configurations (i.e. > 10 μm/s) and high strains, we find that samples of rock salt weaken significantly and ultimately slide unstably (i.e. stick-slip). Sliding experiments on quartz at the same stress and temperature conditions, where chemical effects are muted, do not show this significant weakening. Rate and state frictional (RSF) parameters determined from velocity-stepping tests are large compared to values reported on other materials (a > 0.05 and b > 0.05). The mechanical data suggest that the gouges dilate significantly during sliding, with steady state porosity increasing with increasing sliding velocity. Microstructural observations show the presence of a zone of highly comminuted grains along the shear zone boundary, forming a through-going boundary-parallel Y-shear at high sliding velocities. In contrast, samples deformed at low sliding velocities do not show boundary parallel shear, but rather exhibit low porosity 'spectator' regions isolated by dilational zones in the Riedel shear orientation. In these gouges, the grain size does not appear to have been significantly reduced in comparison with the starting grain size. We posit that the significant strain weakening observed at high sliding velocities is caused by severe grain size reduction as shear localization develops, i.e. by frictional wear, ultimately leading to the development of a through-going boundary parallel Y-shear. Unstable slip is probably related to rupture on this Y-shear surface with intermittent healing of the asperities by pressure solution or plastic creep. Furthermore, the data show that the weakening and subsequent unstable slip can be delayed, (i.e. occur at higher strains) by lower sliding velocities, larger initial grain sizes, lower normal stresses and the presence of fluids. This suggests a competition between mechanical wear and solution-transfer aided healing and/or compaction. These data document the need to expand the range of conditions for detailed experiments on quartzose fault gouges to include the hydrothermal conditions expected in the upper- to middle crust.
T21D-07
Earthquake Ruptures with Thermal Weakening and the Operation of Faults at Low Overall Stress Levels
We have conducted rupture propagation simulations incorporating flash heating of microscopic asperity contacts and thermal pressurization of pore fluid [Noda, Dunham, and Rice, in preparation, 2007-08]. These are arguably the primary weakening mechanisms at coseismic slip rates, at least prior to large slip accumulation. Ruptures on strongly rate-weakening faults take the form of slip pulses or cracks, depending on the background stress level. Self-sustaining slip pulses exist only within a narrow range of stresses; below this range, artificially nucleated ruptures arrest, and above this range, ruptures are crack-like. Certain features of our simulations lend support to the idea that faults operate at the minimum critical level required for propagation, such that natural earthquakes take the form of slip pulses. Using flash heating parameters measured in recent laboratory experiments, the critical range occurs when the ratio of shear to effective normal stress on the fault is 0.2-0.3 (a range that is only mildly influenced by the choice of thermal pressurization parameters, at least within a reasonable range of uncertainty around laboratory-measured values). This level is consistent with the low stress inferred to be acting on the San Andreas fault (SAF); a ratio of shear to effective normal stress of 0.24 was measured at 2.1 km depth in the SAFOD pilot hole [Hickman and Zoback, 2004], adding further support to other measurements indicating that the maximum horizontal compressive stress is nearly perpendicular to the SAF. While the overall background stress level is quite small, stresses concentrated at the rupture front are consistent with typical static (and low velocity) friction coefficients of 0.6-0.9; this stress concentration is required to initiate slip. Growing slip pulses have stress drops close to 3 MPa and feature slip increasing with propagation distance at a rate of about 0.14 m/km. These values are consistent with seismic inferences of stress drop and field constraints on slip-length scaling. On the other hand, cracks have stress drops of over 20 MPa, and slip at the hypocenter increases with propagation distance at a rate of about 1 m/km.
T21D-08
Shear Heating-Induced Thermal Pressurization During the Nucleation of Earthquakes
Shear heating-induced thermal pressurization has long been posited as a weakening mechanism during earthquakes. It is often assumed that thermal pressurization does not become important until earthquakes become moderate to large in magnitude. Schmitt et al. [AGU, 2007] confirmed the estimate of Segall and Rice [JGR, 2006] that thermal pressurization becomes dominant during the quasi-static nucleation phase by conducting 2D numerical simulations that account for full thermomechanical coupling, with rate and state dependent friction. In that work, thermal pressurization becomes the dominant weakening mechanism at slip rates of 10-5 to 10-3 m/s, depending on the fault zone hydraulic diffusivity. Interestingly, the thermal pressurization process leads to a contraction of the nucleation zone, rather than the growing crack (aging law) or unidirectional slip pulse (slip law) associated with drained rate- and state-dependent frictional nucleation. The results of Schmitt et al. [AGU, 2007] had a shortcoming in that the principal slip surface was treated as a zero-width feature, while in reality it should be a finite-width shear zone. We address that shortcoming with a new set of numerical simulations. We assume a finite-width fault governed by rate and state friction with the radiation damping approximation to simulate inertial effects. Both thermal and hydraulic diffusion are computed via finite differences on separate, coupled grids that adaptively remesh to minimize computational expense while maintaining accuracy. New results suggest that the thermal pressurization effect is modestly reduced by including the finite thickness of the shear zone. Despite the reduction in the effect, the new results still indicate that (1) thermal pressurization is important before seismic slip and (2) thermal pressurization restricts growth of the nucleation zone.