T22A-01 INVITED
Pulse-Like Ruptures due to Rate-Weakening Friction and Low Prestress in Earthquake Models and Laboratory Experiments
Laboratory experiments and theories of how fault materials behave suggest that the constitutive response of faults is far from simple. For slow slip rates, lab-derived rate and state friction formulations incorporate small, less than 10%, variations in frictional strength about a representative value which is about 100 MPa at the representative seismic depth of 8 km. For fast sliding velocities and large slips, additional weakening mechanisms are activated that can result in much lower frictional resistance during dynamic sliding. Hence we need to build earthquake models that would account for both high static strength and low dynamic strength of faults. The challenge lies in combining them in such a way that static stress drops are of the order of 1-10 MPa as typically observed. This can be accomplished by incorporating small defect regions that nucleate ruptures while the average stress on the fault is still low compared to its static strength (e.g., Lapusta and Rice, AGU, 2004). Simulations of fault response, in terms of earthquake sequences in the framework of a 2D depth-averaged elastic model of a faulted crustal plate subjected to slow tectonic loading, show that the fault would then operate with reasonable static stress drops, low shear stress, and low heat generation. The fault is governed by rate and state friction laws modified to include significant rate weakening at seismic slip rates. With such strongly rate-dependent law and low values of prestress before dynamic events, simulated dynamic ruptures propagate as short-duration pulses of slip, consistently with seismic inversions and theoretical studies (Heaton, 1990; Zheng and Rice, 1998). Other mechanisms that can lead to pulse-like ruptures include arrest waves due to local heterogeneities or normal stress variation for ruptures on bi-material interfaces. The possibility of obtaining pulses as the result of rate-weakening interfaces and low prestress has been explored in laboratory experiments (Lu, Lapusta, Rosakis, 2007; Lu, Rosakis, Lapusta, 2008). We have indeed experimentally observed spontaneous pulse-like ruptures on a Homalite interface prestressed both in compression and in shear, similarly to faults in the Earth's crust. Our results confirm that pulse-like ruptures can exist in the absence of a bi-material effect or local heterogeneities. For a set of experiments with increasing ratio of shear to normal prestress, which is achieved by increasing the inclination angle of the interface, we observe systematic variation in rupture modes from pulse-like to crack-like consistent with the theoretical and numerical studies of velocity-weakening interfaces (Zheng and Rice, 1998; Lapusta and Rice, 2008). The qualitative agreement between our experimental observations and models of rate-weakening faults provides indirect evidence for rate-weakening friction on Homalite interfaces and suggests that rate- weakening friction plays an important role in dynamic rupture behavior.
T22A-02
Comparison of Repeating Magnitude 2 Earthquakes Near the SAFOD Site, California, With Similar-Magnitude Mining-Induced Earthquakes in South Africa
A small patch of the San Andreas fault at a depth of about 2.7 km near the site of SAFOD (San Andreas Fault Observatory at Depth) produces magnitude 2 earthquakes that repeat at intervals of 2.89 years (Nadeau and Johnson, Bull. Seism. Soc. Am., 1998; Dreger et al., Geophys. Res. Lett., 2007). A repeat occurred on 20 October 2003 and was recorded by the SAFOD Pilot Hole Array, a vertical network of 3- component seismometers at 32 levels extending from 856 to 2096 m below the local ground surface (Imanishi and Ellsworth, AGU Geophysical Monograph 170, 2006). The ground velocity signals from this repeating earthquake, recorded in the Pilot Hole at hypocentral distances of several km, are remarkably similar to those recorded using a four-station network of IRIS/PASSCAL broadband recording units with accelerometers, deployed underground at depths between 2 and 3.5 km, within two of the deepest and most seismically active mines in South Africa. During our one-week deployment, four earthquakes, of seismic moment and hypocentral depth much the same as those of the repeating earthquakes near SAFOD, were recorded and analyzed to investigate their source processes. In addition to determining the traditional source parameters, including seismic moment, we used distance-corrected peak ground velocities to estimate maximum seismic slips within the four rupture zones, ranging from 4 to 27 mm. (Although our method of doing this involves several model assumptions, underground observations of maximum slip due to mining-induced earthquakes indicate that our approach yields realistic estimates.) These maximum slips, in conjunction with data from laboratory stick-slip friction experiments, were used to estimate maximum slip rates that fell in the range 2 to 7 m/s, typical for earthquakes in the continental crust. Applying the same methods to the 20 October 2003 earthquake data, as recorded in the Pilot Hole, revealed a maximum slip of about 17 mm and a maximum slip rate of 4.7 m/s. That is, our analysis has not indicated any aspect of this repeating earthquake that is out of the ordinary for crustal earthquakes of the same magnitude (McGarr and Fletcher, Bull. Seism. Soc. Am., 2003, 2007).
T22A-03 INVITED
Near-Source Observations of Earthquakes: Implications for Earthquake Rupture and Fault Mechanics
The deployment of instrumentation within seismically active crustal rocks in deep boreholes and mines has opened a new window for the study of the earthquake source. Advances in both sensors and high- temperature electronics enable recording of seismic waves over a very broad frequency spectrum (D.C. to several KHz) and amplitude spectrum (Earth noise floor to several g acceleration) in these challenging underground environments. By reducing the distance between source and receiver to a few hundred meters or less, it becomes possible to observe dynamic processes on space and time scales that approach those of laboratory experiments. Key questions that can be addressed in the near-source region include scaling of apparent stress and static stress drop, the minimum size of earthquakes, and the time, length and displacement scales of frictional evolution during nucleation. Earthquake source parameters determined using a variety of methods indicate that there is no breakdown in apparent stress or stress drop scaling for Mw > 0. Within any magnitude band, stress drops range between 0.1 and 100 MPa. The highest values are comparable to the laboratory-derived frictional strength of faults. Using deep borehole seismometers in the main hole of the San Andreas Fault Observatory at Depth (SAFOD) near Parkfield, CA, and in the Long Valley Exploratory Well (LVEW) in the center of Long Valley Caldera, CA, we have observed earthquakes at the lowest limit of magnitude detection (currently Mw -3.5 at SAFOD). The smallest events have source dimensions < 1 m, indicating that if there is a minimum earthquake size, it must lie at lower magnitude and spatial scales. Mean displacements in the smallest events are on the order of 100 microns, suggesting that the displacement weakening distance is smaller still. The rate of fault weakening can be studied using the earliest part of the P-wave arrival. We apply Kostrov's 1964 model for self-similar crack growth to determine the dynamic stress drop. For earthquake sources at distances of less than 1 km, we find no evidence for a slow initiation process. Instead, these earthquakes begin abruptly with the dynamic stress drop typically reaching 5 MPa within the first few milliseconds of rupture. These observations suggest a very small displacement weakening distance, comparable to values measured in the laboratory.
T22A-04
Numerical Simulation of Slip Localization in Mechanically Heterogeneous Fault Zones
Slip along faults is frequently localized within thin zones of the fault core that is surrounded by damaged rock with damage decreasing away from the fault core. Mechanisms of this slip localization are explored here by finite element simulations of mechanically heterogeneous fault zones. Our 2D finite element model includes an idealized fault zone embedded in a host rock medium and subjected to quasi-static boundary displacements. We employed material properties that are based on rock-mechanics tests. The increase in damage (micro-cracks) intensity towards the fault core is represented by a gradational increase of strain hardening towards the fault-zone. We apply a linear 3-node triangular mesh, gradually refined towards the center of the fault zone. The model is applied to two cases of recent earthquakes: The Pretorius fault in Tautona mine, South Africa, and the Nojima fault, Japan. The Pretorius fault has been inactive for at least 2.0 Ga and underwent healing processes during this period. Mapping of the rupture zone of a 2004 M=2.2 earthquake along this fault indicated reactivation and slip localization along the contact of the damaged host rock (quartzite) and the well-cemented quartzitic cataclasite. Rock mechanics experiments revealed that the host quartzite shows significant damage-controlled strain hardening, whereas the cataclasite behaves brittle with no damage. Our finite-element model shows an abrupt increase of the shear stress at the contact between the model fault- zone and the host rock, and a corresponding increase in the plastic shear strain. This result suggests that the shear stress gradient develops due to the contrast in plastic behavior between the host rock and fault- zone that likely leads to slip localization along the contact. The second case includes the active Nojima fault (Awaji Island, Japan) that slipped during the 1995 Kobe earthquake. Drilling across the Nojima fault indicated that the earthquake slip was localized within a few millimeter thick zone surrounded by weakly pulverized and altered granodiorite. Our model includes the distribution of mechanical properties across the Nojima fault derived from published rock mechanics experiments, with gradually decreasing elastic properties towards the fault core. The fault core (clay-rich gouge) is modeled by a low cohesive, weak and dominantly plastic material. Our finite-element model predicts the locations of slip localization.
T22A-05
Tensile cracks: a new link between geological observations of faults and seismological models of earthquake dynamics
Motivated by the theoretical prediction that pseudotachylyte injection veins may be used to interpret rupture velocity and directivity of ancient earthquake rupture, we studied the nucleation and growth of tensile microcracks in Homalite-100 near the tips of shear ruptures (Laboratory Earthquakes) propagating along an interface with frictional and cohesive strength. The samples were compressed under a static uniaxial load, P, and rupture was induced along an inclined interface by an exploding wire. Rupture velocities during the experiments ranged between 70% and 85% of the shear wave speed of Homalite (cs ~ 1255 m/s). Opening microcracks were produced only along one side of the interface where they were associated with the transient tensile stress perturbation of the cohesive end zone at the shear rupture tip. The orientation of microcracks varies with the rupture velocity and the ambient static stress field. The results of this study provide a rationale for interpreting observations of tensile fractures along exposed faults and create diagnostic criteria for interpreting the velocity, directivity, and static prestress state associated with past earthquake ruptures.
T22A-06
Foreshock And Aftershock Sequences: Insight From Laboratory Cyclic Pore Pressure Experiments
Fluid and fracture interaction are a key mechanism in the Earth's crust, and are not yet fully understood. The effects of pore pressure on pre-existing faults are twofold: Firstly, an increase in pore pressure reduces the effective normal stress that holds the fault locked; giving rise to stick-slip failure. Secondly, pressurized pore fluids may act over time to reduce the strength of the rock through mechanisms such as stress corrosion and static fatigue (creep). The former effect is well-known as the law of effective stress, and is often cited as the reason for the rapid response of seismicity due to fluid pressure changes. However, this law does not account for the permeability change with time, and hence, cannot fully explain the cause of protracted seismicity. The effect of cyclic pore pressure has been speculated as one of the factors that causes protracted seismicity. However, whilst numerous theoretical solutions have been developed to account for the diffusivity and hydraulic conductivity of the fractured zone, a severe paucity of laboratory data exists with which to test such hypotheses. We report new cyclic pore pressure experiments carried out on porous sandstone samples (Fontainebleau), which were pre-fractured using constant strain rate of 2E-6 in a conventional triaxial cell at a confining pressure of 25 MPa and a pore pressure of 5 MPa. Subsequently, different combinations of cyclic pore pressure amplitudes and frequencies were applied; with seismicity response (acoustic emission) measured continuously using a specifically developed continuous data acquisition unit. By recording 16 transducers continuously at 10 MHz sampling frequency to hard disk, we are thus able to reconstruct the temporal and spatial distribution of hypocentres without loss of information due to either the rapid fracture formation or the subtle changes of pore oscillation induced seismicity. Our results show that when the amplitude of pore pressure increase is small (less than 0.5 MPa), a large number of pore pressure cycles are required to induce seismicity on the pre-existing faults. Aseismic slips built up gradually followed by larger magnitude foreshocks and mainshocks along the faults. Conversely, if the pore pressure increase is large (greater than 0.5 MPa), most of the seismic events occur during the first pore pressure cycle, and are subsequently followed by an exponential reduction with successive pore pressure cycles. Our experiments provide a proxy to the seismicity (both initial and protracted) which is observed in field scale reservoir-induced seismicity / aftershock sequences. Furthermore, our results indicate that seismicity can be intensified (i.e. increase in the number of events as well as magnitudes) when the applied pore pressure magnitude exceeds the previous maximum that has been experienced by the sample.
T22A-07
Initiation and Activation of Faults in Dry and Wet Rock by Fluid Injection
We studied fracturing of rock samples induced by water injection in axial compression tests on cylindrical specimens of Flechtingen sandstone and Aue granite of 50 mm diameter and 105-125 mm length. Samples were intact solid rock cylinders and cylinders with central boreholes of 5 mm diameter and 52 mm length or through-boreholes of 2.5 mm diameter. To monitor acoustic emissions (AE) and ultrasonic velocities, twelve P-wave and six polarized S-wave sensors were glued to the cylindrical surface of the rock. Full waveforms were stored in a 12 channel transient recording system (PROEKEL, Germany). Polarity of AE first motion was used to discriminate source types associated with tensile, shear and pore-collapse cracking. To monitor strain, two pairs of orthogonally oriented strain-gages were glued onto the specimen surface. Samples were deformed in two consecutive loading steps: 1) Initial triaxial loading was performed at 20-50 MPa confining pressure on dry (under vacuum) or fully saturated samples until the yield point was reached. 2) In a second stage distilled water was injected into the samples with pore pressure increasing up to 20 MPa. For saturated samples the pore pressure was increased in steps and in periodic pulses. Injection of water into dry porous sandstone resulted in propagation of an AE hypocenter cloud closely linked to propagation of the water front. Position of the migrating water front was estimated from ultrasonic velocity measurements and measurements of the injected water volume. Propagation rate of AE-induced cloud parallel to bedding was higher than perpendicular to bedding, possibly related to permeability anisotropy. Nucleation of a brittle shear fault occurred at a critical pore pressure level with a nucleation patch located at the central borehole. Micro-structural analysis of fractured samples shows excellent agreement between location of AE hypocenters and macroscopic faults.
T22A-08
The Frictional Properties of Phyllosilicates at Earthquake Slip Speeds
Most mature natural faults contain a significant component of sheet silicate minerals within their core. In order to elucidate the ease which earthquake ruptures may propagate along such faults, we conducted a series of high velocity (1.3 m/s) laboratory friction experiments on synthetic fault zones containing pure kaolinite, sericite, illite, talc and montmorillonite under dry and wet conditions. The normal stress in the experiments was varied between 0.8 to 2.45 MPa. Under dry conditions, peak friction was reached during acceleration of the fault zones to the steady-state velocity. At the highest normal stress it varied between 0.7 and 0.4. The peak friction for each of the sheet silicates correlates well with the corresponding mineral electrostatic separation energy. The peak friction rapidly decreases to steady state values, typically ~0.2 for all the minerals tested, over slip weakening distances of between 1 to 3 m. Under wet conditions the peak friction reduces considerably or disappears. Thus steady-state friction is established almost immediately and has values corresponding to those under dry conditions. Correspondingly, the slip weakening distances in these wet experiments are very small. The results suggest that it will be energetically very easy for earthquake ruptures to propagate through wet, sheet silicate-rich fault zones.