S31D-01 INVITED 08:00h
Gouge Formation by Dynamic Pulverization During Earthquakes
The formation of fine grain gouge along mature faults is usually attributed to quasi-static grain crushing, but there is no model for gouge formation that accounts for the dynamic effects of a fast propagating earthquake rupture. The present analysis examines the deformation field close to the tip of a mode II fracture by using the asymptotic solution for an in-plane propagating fracture in an isotropic elastic solid (Freund, 1990). In the solution, the propagation velocity may not exceed the Rayleigh wave velocity, Cr. The parameters that weakly depend on the fracture velocity (density, Poisson's ratio, shear modulus and elastic wave velocities), are assumed constant with values of moderately competent granite; the stress intensity factor, KII, that could vary with velocity is considered in its full range. The calculations show that when a fracture propagates at V $>$ 0.85 Cr, intense deformation conditions develop at a finite distance of 1-3 mm from the tip: At V of about 0.99 Cr, the tensile stresses approach 10 GPa within a 6 mm wide zone around the tip, and the volumetric dilation rates alternate between $10^5 s^-^1$ expansion and similar absolute value of contraction. Similar extreme conditions are known in nature only for the shock wave at impact sites. This model can be bounded with our recent results of the gouge texture from an earthquake rupture zone in a South African mine (Reches et al., this meeting, session T20). The surface area of gouge approaches $80 m^2/g$ that corresponds to a surface energy of 0.2-0.36 $MJ/m^2$ for a 1 mm thick gouge zone. Our mapping of the rupture zone reveals tens of subparallel fractures that are ~1 mm thick and filled with gouge; summation for 10-30 fractures yields surface energy of 3-10 $MJ/m^2$. By assuming that this surface energy is a lower limit on the fracture energy G, we estimate from the model that this earthquake propagated at V $>$ 0.9 Cr with maximum tensile stress exceeding 5 GPa, and dilation rate exceeding $5*10^5 s^-^1$. We propose that the extreme deformation conditions at the tip region of a fast propagating earthquake rupture pulverize the fault rocks by processes known from shock analyses. Further, while we analyzed three earthquake rupture zones in South African mines, only two of them display large amounts of pulverized gouge. According to our model, the third rupture zone propagated at V $<$ 0.85 Cr, corresponding to stresses and dilation rates that are too low to pulverize the rocks. We finally propose that the intensity of pulverization can serve as a paleo-velocity criterion of earthquake rupture.
S31D-02 08:15h
Pulverized Fault Zone Rocks Along the Mojave Section of the San Andreas Fault: Distribution and Mechanical Significance
We present field mapping of the San Andreas Fault zone along the Mojave section between Three Points and Pallet Creek (enclosing 60 km of the fault length), showing that pulverized crystalline rocks are a common component of the fault-zone structure, with asymmetric distribution across the fault. Studies of exhumed inactive fault zones show that faults have a damage zone with width of 100's of meters, and that shear is localized to a core fault zone layer typically only 10's of cm wide. Fracture density in the damage zone increases toward the fault core, in which highly deformed fault-rocks exhibit distinct textural and mineralogical characteristics (Schulz and Evans 2000; Chester and Chester, 1998). In addition, fault zone rocks along strands of the San Andreas Fault were found to be intensely pulverized (Brune, 2001). The grains of the pulverized rocks appear to have failed by tension and present extreme reduction in the original grain size. Shear in the pulverized rock is very limited and the texture of the protholite is preserved. Recent analysis of a 70 m wide belt of pulverized granite in Tejon Pass indicates fracturing in the sub-micron scale with non-fractal particle size distribution (Wilson et al., 2004). Our mapping shows that pulverization is restricted to crystalline rocks, and that these pulverized rocks are common within the fault zone, but their distribution across the fault is asymmetric. Crystalline rocks on the north east side of the fault are found to be pulverized wherever they are exposed, up to a distance of ~50 meters from the fault. This is followed by a transition zone ~10-20 meters wide of selective pulverization, beyond which the rocks are macroscopically fractured but not pulverized. Exposures on the south west side of the fault are much less common; where they are found, crystalline rocks adjacent to the fault show a mixed pattern that can range from full pulverization through selective pulverization and up to macroscopic fracturing. The observed pulverized crystalline rocks are likely to result from strong extensional dynamic strain associated with the passage of repeating earthquake ruptures. In particular, wrinkle-like ruptures along a material contrast provides a framework for strong dynamic extensional strain with preferred propagation direction (e.g., Weertman, 1980; Andrews and Ben-Zion, 1997; Ben-Zion and Huang, 2002) that may produce asymmetric distribution of tension-based rock damage across the fault. High resolution imaging of fault zone structures with seismic trapped waves (e.g., Ben-Zion et al., 2003; Peng et al., 2003; Lewis et al., 2004) show that the trapping structures of several major faults, associated with low velocity damaged rock, are limited to the upper 3-4 km. The lack of trapping structures below that depth suggests that the pulverized fault zone rocks may not extend below the top few kilometers.
S31D-03 08:30h
Structures Formed in Experimentally Sheared Artificial Fault Gouge: Precise Statistical Measurements
The physical parameters governing earthquakes change with the ongoing formation and evolution of structures, formed in the course of a single or multiple earthquakes, within a particular fault zone or in a broad volume containing interacting tectonic faults. Our precise knowledge of these complex phenomena is still elusive. Especially, works considering geometrical evolution of shear structures under controlled conditions are rare. In order to gain some insights we accomplished a set of 12 laboratory experiments using a servo-controlled direct-shear apparatus, under room temperature and without controlling the air humidity. Two fault gouge layers (industrially produced quartz powder, average particle size of 5 $\mu$m, and pre-shear thickness of 1.5, 2.0 and 3.0 mm,) were sandwiched between three granite blocks. The middle block was slid in order to create frictional structures within the simulated gouge. The total imposed shear strain varies between 0.14 and 11.80. The post-shear gouge layer thickness ranges from 0.99-2.11 mm. Each experiment was run under a constant normal stress (varying from 10-44 MPa through the experiments) and at a constant shear velocity (0.07, 0.7 and 7 $\mu$m/s, through the experiments). Later, in cross-sections of solidified by epoxy glue gouge (parallel to the shear direction, normal to the gouge walls,) we quantified the numerous R-shears, according to their density distribution, fracture thickness (measured perpendicularly to the fracture walls), fracture angle and morphology, and fracture length. In gouge views parallel to the sliding blocks, we measured fracture length and along-strike R-shear morphology. Although the latter data are with lower quality, both observational sets provide precise statistical fracture data as well snapshots of evolving 3D structures. We observe shear localization with decreasing gouge layer thickness and with increasing normal stress. The average density of major fractures increases from 2.83 to 3.67 [fracture/cm] for decrease of the post-shear gouge layer thickness. This is at the expense of a considerable decrease of visible more diffusive minor fractures. On the other hand, the fractures formed at lower normal stress are more irregular and show average fracture density of 4.48 [fracture/cm]. The latter decreases down to 3.64 at higher normal stress, as the fracture morphology becomes more regular. The fracture density increases abruptly from zero, after a small total shear strain (0.15-0.50), and later the change is slower or none with the increase of the total shear strain; the fractures are already localized and they accommodate most of the brittle deformation. Also we observe weak polarity in fracture development in accordance to the sliding sense, especially in the subset of fractures starting from the gouge wall and dying out within the gouge layer. More such fractures are developed along the leading part of the sliding blocks. Our results throw new light over the formation and development of fault-related structures and their dependency on the earthquake-governing physical parameters.
S31D-04 08:45h
Comparative Laboratory and Numerical Simulations of Shearing Granular Fault Gouge: Micromechanical Processes
Laboratory studies of granular shear zones have provided significant insight into fault zone processes and the mechanics of earthquakes. The micromechanisms of granular deformation are more difficult to ascertain, but have been hypothesized based on known variations in boundary conditions, particle properties and geometries, and mechanical behavior. Numerical simulations using particle dynamics methods (PDM) can offer unique views into deforming granular shear zones, revealing the precise details of granular microstructures, particle interactions, and packings, which can be correlated with macroscopic mechanical behavior. Here, we describe a collaborative program of comparative laboratory and numerical experiments of granular shear using idealized materials, i.e., glass beads, glass rods or pasta, and angular sand. Both sets of experiments are carried out under similar initial and boundary conditions in a non-fracturing stress regime. Phenomenologically, the results of the two sets of experiments are very similar. Peak friction values vary as a function of particle dimensionality (1-D vs. 2-D vs. 3-D), particle angularity, particle size and size distributions, boundary roughness, and shear zone thickness. Fluctuations in shear strength during an experiment, i.e., stick-slip events, can be correlated with distinct changes in the nature, geometries, and durability of grain bridges that support the shear zone walls. Inclined grain bridges are observed to form, and to support increasing loads, during gradual increases in assemblage strength. Collapse of an individual grain bridge leads to distinct localization of strain, generating a rapidly propagating shear surface that cuts across multiple grain bridges, accounting for the sudden drop in strength. The distribution of particle sizes within an assemblage, along with boundary roughness and its periodicity, influence the rate of formation and dissipation of grain bridges, thereby controlling friction variations during shear.
S31D-05 09:00h
Direct Observation of Depth Variation in Fault Zone Structure Through and Below the Seismogenic Crust: Preliminary Results From the SEMP Fault System in Austria
One of the most exciting and important frontiers in earthquake science is the linkage between the internal structure and the mechanical behavior of fault zones. In particular, little is known about how fault-zone structure varies as a function of depth, from near-surface conditions down through the seismogenic crust and into the ductile lower crust. Such understanding is vital if we are to understand the mechanical instabilities that control the nucleation and propagation of seismic ruptures. This imperative has led us to the Oligo-Miocene Salzach-Ennstal-Mariazell-Puchberg [SEMP] fault zone in Austria. The SEMP system is an extremely rare example of a major strike-slip fault that has been exhumed differentially such that it exposes a continuum of structural levels along strike. This exhumed fault system thus provides a unique opportunity to systematically examine depth-dependent changes in fault-zone geometry and structure along a single fault. Our ongoing field studies focus on structural transects across the SEMP fault zone at exhumation levels ranging from the near-surface at the eastern end of the fault (Vienna pull-apart basin), within the seismogenic crust (central Austria), and down into the ductile lower crust exposed in the Tauern window of western Austria. In addition to detailed field mapping of structural fabrics, fluid-rock interactions, relative timing relationships, and variations in fault geometry, we are also conducting detailed analyses of fault-zone rocks designed to explore deformation at a wide range of scales using petrographic microscopy, cathodoluminescence microscopy, fluid-inclusion studies, scanning-electron microscopy, and transmission/analytical-electron microscopy. Preliminary results from one of our first detailed study sites, at Gesäuse in central Austria, reveal strikingly asymmetric damage across the fault. The limestones exposed south of the fault are fractured, but relatively coherent to within a few meters of the main fault trace. In contrast, the dolomites exposed north of the fault are pervasively sheared and fractured to the cm scale for a distance of at least 200 m from the main fault trace. This extreme asymmetry in damage is consistent with recent models of dynamic rupture in which the rupture propagation direction is controlled by material contrasts across the fault. Detailed studies such as these will allow us to create a synoptic view of the SEMP fault zone from top to bottom - a view that describes how the fault zone varies in its characteristics at different depths.
S31D-06 09:15h
Transition from cataclastic flow to aseismic brittle failure in Carrarra marble
Interest in the brittle-ductile transition has increased considerably in recent years, in large part due to the fact that the maximum depth of seismicity corresponds to a transition in the crust and in the upper mantle from seismogenic brittle failure to aseismic cataclastic flow, i.e. from localized to homogeneous deformation. The mechanics of the transition depends both on some extrinsic variable (state of solid stress, pore pressure, temperature, fluid chemistry and strain rate) and intrinsic parameters (crack and dislocation density, modal composition of the rock or porosity for example). In the present study, two triaxial experiments were performed on Carrara marble at room temperature. The rock samples were first deformed in the cataclastic domain (up to $\sim5%$ axial strain) until they exhibited severe damage accumulation, i.e. wavespeed attenuation. Damaged samples were then brought back {\bf at constant differential stress} into the the brittle field by solely reducing the effective mean stress. Throughout both experiments, compressional wave velocities were measured along several raypaths. Acoustic Emissions, when any, were monitored and localized after testing. A complete 2 minutes failure recordings of failure (12 channels at 10MHz sampling frequency) was also obtained using ESG's Hyperion gigarecorder during one of the experiment. Our new set of data shows that during cataclastic deformation, elastic wave velocities show large variations, but only a small degree of elastic anisotropy when compared to what is generally observed in typical brittle materials such as granite or sandstones. After sufficient damage accumulation and when reducing the mean stress, both samples exhibited a fast acceleration in axial strain. Tertiary creep was followed by the nucleation of a brittle failure. Observed differential stress drops during rupture propagation were of the order of $150\,MPa$. Although failure occurred with large slip and stress drop, only very few AEs could be detected. The complete recording shows that rupture nucleated, initiated and propagated almost aseismically. Elastic properties, macroscopic strain, AE recordings and post-deformation microstructural analysis are here put together and show what we believe are the first experimental evidence of aseismic failure in the laboratory. Those results may have direct implications for the understanding of silent earthquakes and aseismic slips in the field. Limestones, because they can deform plastically at room temperature, may be good deep fault gouge analogs and thus their behavior could have direct consequences on the understanding of fault zones and the earthquake cycle.
http://www.lassondeinstitute.utoronto.ca/young/people/alex2.htm
S31D-07 09:30h
Repeated Rupture of Asperities on a Single Fault: Aseismic and Seismic Interactions. A Case Study From Induced Seismicity at Soultz (France)
The hydraulic stimulation of the granite rock mass around 3 km in depth at the geothermal site of Soultz-sous-Forets (Alsace, France), in 1993, has produced more than 11,000 earthquakes in 17 days, recorded on surface and borehole seismometers. Among them, more than 150 multiplets consist of at least five events, which were studied for constraining the related fracture orientations. The injection borehole is cut by several major fractures, among which a few have slipped by up to a few centimetres during the injection, as evidenced by the logging repeated at the end of the experiment. We present the results concerning the fracture which showed the largest slip in the borehole, 4.3 cm, with a normal faulting mechanism constrained by in situ measurement of the dip, strike, and slip angle. We selected the multiplets located close to the extrapolated plane, and showing a similar mechanism, assuming that they belong to the same fault surface. For each multiplet, we calculated the source radius from corner frequency. This dimension (typically 10 m) appears larger than the hypocenter distances (a few meters), which implies that they consist of the repetition of the rupture of a single asperity. We calculate the seismic moment to get the slip for each event, and, for each multiplet, we are thus able to deduce the cumulative slip of the related asperity. The latter shows that the slip rate progressively slows down after the first events, following an Omori law. A more detailed space-time analysis shows that the multiplet sequences are thus the repeated slip of asperities forced by this transient slip. Their Omori-type decay are thus not related to the standard model of accelerating slip deduced from rate-and-state velocity weakening friction law (Dieterich,1994), but to a decelerated slip due to velocity strengthening friction law. The asperities also interact between themselves, when their distances are smaller than about twice the source dimension, as shown by the statistics of interevent times. Some asperities also happen to trigger repeatedly within a few seconds, which we interpret as a retro-action process related to co-seismic fault compaction around the asperity involving delayed fluid pressure increase within the asperity.
S31D-08 09:45h
Repeating earthquakes may indicate a relation between fault healing and proximity to a mainshock asperity
We analyze time-dependent variations in recurrence interval (Tr) and seismic moment (Mo) of 92 repeating earthquake sequences on the Calaveras fault that were aftershocks of the 1984 M6.4 Morgan Hill, California, earthquake. The recurrence interval between successive repeating events is proportional to the time elapsed since the mainshock, and is compatible with the Omori law for aftershock decay. Our results agree with previous inferences that the repeating earthquakes are produced by asperities undergoing stick-slip failure and are reloaded by transient creep after the mainshock. Rates of deep repeating earthquakes diminish faster than those of the shallow events, suggesting a variation in the duration of the postseismic transient within the aftershock zone. We also examine relation between Mo and Tr for these repeating clusters, which we use as a proxy for fault-healing rate. The moment increases 26$\pm$24% per decade change in Tr for clusters located above 5.5 km depth, consistent with our previously published estimates. Most of the clusters that are located below 5.5 km have lower or even negative trends in moments (-3$\pm$11%). The deep repeating events appear to be just above a large asperity that broke during the Morgan Hill mainshock. The asperity is $\sim$10 km deep and has a radius of $\sim$5 km. We consider models in which fault healing is influenced by evolution of fluid pressures in the fault zone, differences in rock type, and the influences of temperature, shear and normal stress on healing, and how these properties may be related to the location of the mainshock asperity.