S21B-0554
Unified Understanding of Dynamic Earthquake Rupture in Terms of Thermoporoelastic Effects
We theoretically study 2-D dynamic earthquake rupture taking account of thermoporoelastic effects including fluid flow and inelastic porosity change; the porosity on the fault is assumed to increase inelastically with increasing fault slip. We found in the analysis of a 1-D fault model that a single nondimensional parameter Su controls the system behavior [Suzuki and Yamashita, 2007]; the slip-weakening and -strengthening behavior appeared when Su is less and greater than a critical value, respectively. The case 0<Su<1 is found to be excluded from the consideration because rock melting is expected. Since the case Su=0 was already analyzed in Suzuki and Yamashita [2006], we assume Su>1 in this study. Since the reduction in stress drop is larger for a larger slip for Su>1 as found in Suzuki and Yamashita [2007], the stress drop reduction is larger at fault patch further behind the extending fault tips in 2-D fault model. This can successfully simulate a pulse-like fault slip [e.g., Heaton, 1990] as actually simulated in our 2-D analysis. In addition, the maximum slip velocity in our fault model does not change much with crack tip extension, which differs from expected from the classical Griffith crack model. It is also reported in some studies that the radiation efficiency estimated seismologically exceeds unity for some earthquakes, which contradicts its definition. This paradox can be solved if we assume a relatively large Su value for these earthquakes as suggested by Suzuki and Yamashita [2007], which is actually confirmed here by numerical simulation of 2-D fault model. Seismic moment in such case is smaller than that in the classical crack model. This occurs because the slip- strengthening becomes dominant some time after the slip onset. The intensity of the stress at the crack tips in our model is therefore smaller than that in the classical crack model. In addition, there is no strong dependence of the stress drop on the crack length. This suggests that the growth of crack is more easily arrested by a spatial heterogeneity of strength distribution when Su>1.
S21B-0555
Spontaneous rupture processes with thermal pressurization: Spatial variation of rupture and effect of shear zone thickness and fault shap
We show thermal pressurization (TP) can affect spatial variation in stress/slip evolution of earthquake rupture processes, depending on shear zone thickness and fault shape, based on 3-D numerical simulations for spontaneous ruptures with TP. In this study, a rectangular fault is placed in an infinite, homogenous and elastic medium. The length of the fault is 8km and the width is 3km. The numerical algorithm is based on the finite- difference method of Kase and Kuge (2001). Rupture is initiated in a small patch at the center by decreasing shear stress to dynamic frictional stress, and proceeds spontaneously, governed by a slip-weakening law with the Coulomb failure criteria. We allow effective normal stress to vary with TP by the formulation of Bizzarri and Cocco (2006). We examine drained strike-slip fault (DS), undrained one (US), drained dip-slip fault (DD), and undrained one (UD), with shear-zone thickness w of 20cm, 2cm, and 2mm. Our numerical simulations show stress and slip distributions on the faults are not the same and depend on fault shape and shear zone thickness, although the rupture processes with w of 2cm and 2mm differ only slightly. On drained faults, tractions drop with increasing slip in the same way everywhere. On undrained faults, on the other hand, traction curves as function of slip and amount of stress drop depend on location on the faults. As a result, rupture length for attaining super- shear rupture in US is shorter than that in DS for the same S-value (Andrews, 1976), and nucleated rupture on US for w of 2cm quickly becomes super shear. Spatial variation of stress characterizing US includes a trough of stress drop in the ruptured zone for w of 20cm and short-wavelength variation for w of 2cm. In the case of UD, the stress variation is different from that in US because of an early arrival of a healing phase, and two peaks in the final slip distribution can arise from the combination of large slip due to TP and the early arrival of the phase. For the other three models, in contrast, rupture propagation is crack-like, and slip decreases with distance from the nucleation point. Therefore, the results suggest numerical simulations with TP are important and the results cannot be obtained by those simulations without it.
S21B-0556
Effects of Non-linear Terms and Fault Width on Pore Fluid Pressurization
Faults generate heat due to friction while slipping in earthquakes. If there are pore fluids along the fault, they will be heated and expand. The pore fluids will have little effect on faults in high permeability settings, as they quickly escape. In a low permeability setting, the expanding pore fluids are not able to escape quickly, and thermal expansion of the fluids will increase the fluid pressure, lowering the effective normal stress (and thus frictional stress) along the fault. To investigate this process, we solve the non-linear equations presented in Mase and Smith (1985). These equations involve several non-linear terms that make it necessary to solve the equations iteratively. We have previously shown some results of this methodology for various permeability structures and slip rates. Here we focus on the importance of individual terms in the equations by running models with individual terms neglected. Among our results, we find that conduction significantly affects the temperature and pressure, while advection has a negligible effect on the solution. The implications may be important for researchers constructing simplified models of the pore fluid pressurization process. We also look at the effect of fault width (the width of the area that is shearing and producing heat). In particular, we are interested in the effects of the fault width on the maximum temperature reached, as well as the total amount of frictional heat generated. While a wider fault will tend to have a lower peak temperature because of the distributed slip, it can also result in a larger overall heat generation, because the average temperature over the fault width can be higher than for a narrow fault with a higher, but narrower temperature peak. In contrast, while the narrow faults initially have the highest pressures, the wider faults eventually surpass them both in maximum pressure, and in the amount of overall pressurization.
S21B-0557
Analysis of Different Frictional Laws and Their Implications in the Scaling and Mode of Earthquake Rupture Using a Dynamic Elasto-plastic Frictional Contact Model and the Finite Element Method.
Dynamic simulations of rupture propagation in crustal fault systems are presented. We demonstrate the applicability of our elasto-plastic fault model for modeling dynamic rupture and wave propagation in fault systems. Firstly, we demonstrate the rich array of dynamic properties produced by our elasto-plastic finite element fault model. These are governed by a number of model parameters including: the spatial and material heterogeneity of the fault, the loading strains applied, and not least of all the frictional law employed. Rupture propagation on a fault is controlled by the constitutive properties of the fault. A dynamic elasto-plastic constitutive law for the interface friction at the fault is formulated based on the Coulomb failure criterion and applied in a way analogous to non-associated elasto-plasticity. The penalty method is used to enforce the fault boundary conditions. We employ various slip weakening frictional laws to examine their effect on the resulting earthquake rupture speed, size and mode. We also provide benchmark tests of our method against other reported solutions in the literature. Secondly, we present simulations of multiple earthquake cycles. We propose a numerical method that can produce synthetic earthquake catalogues, which implements four distinct phases sequentially: loading of the next earthquake event, dynamic rupture of the fault, an absorbing wave phase, and a sub-cycle of any (dynamically determined) number of creep events between earthquakes.
S21B-0558
Dynamic ruptures along bimaterial interfaces in 3D
Large faults with a long slip history often separate rocks of different elastic properties. Such bimaterial interfaces have been reported to accommodate remarkable dynamic properties in the 2D in-plane case that may be relevant to many issues of earthquake rupture dynamics. Slip along a bimaterial interface generates dynamic changes of normal stress, modifying the local fault strength which, in principle, can generate a unilateral wrinkle-like pulse. This mechanism is neither present in the homogeneous case nor the 2D anti-plane case. Recently it has been shown that some of these properties stay valid also in the 3D case where there is a mixing of the in-plane and anti-plane case. However, the relevance of this mechanism for natural faulting is a subject of ongoing debate. Independently there are good examples for which the bimaterial mechanism seems to be necessary to properly interpret the observations: Asymmetric alongstrike distribution of aftershocks and asymmetric rock damage across faults of the San Andreas system have been reported. The uncertainty in the basic physical concepts and the constraints of their parameter ranges makes it difficult to come to definite conclusions. Here we present results of an ongoing numerical investigation of bimaterial ruptures in 3D extending the parameter range and resolution of previous studies.
S21B-0559
Attenuation of Radiated Ground Motion and Stresses from Three-Dimensional Supershear Ruptures
Radiating shear and Rayleigh waves from supershear ruptures form Mach fronts that transmit large-amplitude ground motion and stresses to locations far from the fault. We simulate bilateral ruptures on a finite-width vertical strike-slip fault (of width W and half-length L with L >> W) breaking the surface of an elastic half-space, and focus on the wavefield out to distances comparable to L. At distances much smaller than W, two- dimensional plane-strain slip-pulse models (i.e., models in which the lateral extent of the slip zone is unbounded) [Dunham, 2005; Bhat et al., 2007] accurately predict the subsurface wavefield. Amplitudes in the shear Mach wedges of those models are undiminished with distance from the fault. When viewed from distances far greater than W, the fault is accurately modeled as a line source that produces a shear Mach cone and, on the free surface, a Rayleigh Mach wedge. Geometrical spreading of the shear Mach cone occurs radially and amplitudes there decrease with the inverse square-root of distance [Ben-Menahem and Singh, 1987]. The transition between these two asymptotic limits occurs at distances comparable to W. Similar considerations suggest that Rayleigh Mach waves suffer no attenuation in the ideally elastic medium studied here. The rate at which fault strength weakens at the rupture front exerts a strong influence on the off-fault fields only in the immediate vicinity of the fault (for both sub-Rayleigh and supershear ruptures) and at the Mach fronts of supershear ruptures. More rapid weakening generates larger amplitudes at the Mach fronts. A paper has been prepared on this topic, with title the same as for this abstract, by E. M. Dunham and H. S. Bhat, submitted to \it{J. Geophys. Res.} http://people.deas.harvard.edu/~bhat/documents/DunhamBhatSupershear3dJGR052307.pdf
S21B-0560
Numerical Study of Frictional Properties and the Role of Cohesive End-Zones in Large Strike- Slip Earthquakes
Cohesive end-zones (CEZs) are regions of increased frictional strength and/or cohesion near the peripheries of faults that cause slip distributions to taper toward the fault-tip. Laboratory results, field observations, and theoretical models suggest an important role for CEZs in small-scale fractures and faults; however, their role in crustal-scale faulting and associated large earthquakes is less thoroughly understood. We present a numerical study of the potential role of CEZs on slip distributions in large, multi-segmented, strike-slip earthquake ruptures including the 1992 Landers Earthquake (Mw 7.2) and 1999 Hector Mine Earthquake (Mw 7.1). Displacement discontinuity is calculated using a quasi-static, 2D plane-strain boundary element (BEM) code for a homogeneous, isotropic, linear-elastic material. Friction is implemented by enforcing principles of complementarity. Model results with and without CEZs are compared with slip distributions measured by combined inversion of geodetic, strong ground motion, and teleseismic data. Stepwise and linear distributions of increasing frictional strength within CEZs are considered. The incorporation of CEZs in our model enables an improved match to slip distributions measured by inversion, suggesting that CEZs play a role in governing slip in large, strike-slip earthquakes. Additionally, we present a parametric study highlighting the very great sensitivity of modeled slip magnitude to small variations of the coefficient of friction. This result suggests that, provided a sufficiently well-constrained stress tensor and elastic moduli for the surrounding rock, relatively simple models could provide precise estimates of the magnitude of frictional strength. These results are verified by comparison with geometrically comparable finite element (FEM) models using the commercial code ABAQUS. In FEM models, friction is implemented by use of both Lagrange multipliers and penalty methods.
S21B-0561
Transition of Mode-II Cracks From Sub-Rayleigh to Supershear Speeds in the Presence of Favorable Heterogeneity
Understanding sub-Rayleigh-to-supershear transition of mode II cracks is a fundamental problem with important practical implications for earthquake dynamics and seismic radiation. In the Burridge-Andrews mechanism of supershear transition on interfaces with homogeneous friction properties and constant prestress, a daughter crack nucleates, for sufficiently high prestress, at a shear stress peak traveling with the shear wave speed in front of the main crack. The daughter crack is supershear from its very beginning. This transition mechanism requires high enough prestress and has transition distances that depend on the prestress level and friction properties of the interface through two nondimensional parameters. We perform detailed numerical simulations of spontaneous mode II rupture on an interface governed by a linear slip-weakening friction and find that sub-Rayleigh-to-supershear transition of mode II cracks occurs in a number of models that subject cracks to supershear loading fields. We consider a spontaneously expanding sub- Rayleigh crack (or main crack) which advances, along a planar interface with linear slip-weakening friction, towards a place of favorable heterogeneity, such as a preexisting subcritical crack, a small patch of higher prestress, or a small patch of lower static strength. For a range of model parameters, a secondary dynamic crack nucleates at the heterogeneity and acquires supershear speeds due to the supershear stress field propagating in front of the main crack. Transition to supershear speeds occurs directly at the tip of the secondary crack, with the tip accelerating rapidly to values numerically equal to the Rayleigh wave speed and then abruptly jumping to a supershear speed. Models with favorable heterogeneity achieve supershear transition and propagation for much lower prestress levels than the ones implied by the Burridge-Andrews mechanism and have transition distances that depend on the position of heterogeneity. Inferences of supershear earthquake propagation are sometimes interpreted using the Burridge-Andrews mechanism to constrain prestress or parameters of fault friction. If supershear transition is governed by presence of heterogeneity as considered here, such inferences may be misleading. We investigate the dependence of supershear transition and subsequent crack propagation on model parameters using the case with a patch of higher prestress. The dependence on patch size and prestress is far from trivial. For example, under certain conditions, a smaller patch results in sustained supershear propagation far beyond the patch location, while a larger patch causes only a short supershear burst, with the speeds returning to sub-Rayleigh beyond the patch. We will present the parameter study and our explanation for this and other phenomena. We also find that the procedure of initiating the main crack significantly affects subsequent crack propagation and hence the loading provided by the main crack to the location of favorable heterogeneity. That, in turn, affects the nucleation of the secondary crack and crack tip speed beyond the location of heterogeneity. This study indicates that a small preexisting crack or higher-stressed patch can completely change the failure process on the interface, perturbing a sub-Rayleigh crack into becoming supershear. Our preliminary 3D results suggest qualitatively similar but quantitatively different behavior.
S21B-0562
Spectral element modeling of dynamic rupture and long-term slip on rate and state faults
In this work, the spectral element method (SEM) is used to model earthquake rupture and longer-term slip on a vertical strike-slip fault governed by rate and state friction. Previous studies of long-term slip behavior on rate and state faults mostly used boundary integral methods (BIM) which cannot include, at least in their current implementation, complex crustal structures such as variable bulk properties, fault damage zones, and non-planar fault geometries. SEM approach will allow us to include those factors into models that simulate long histories of seismic and aseismic slip. We have extended SEM to dynamic rupture simulations on rate and state faults and validated it by comparison with BIM solutions for a 2D test problem. We use the 3D dynamic SEM formulation to study several problems, including the effect of a shallow steady-state velocity-strengthening fault region (or layer) on a single simulated earthquake. In the absence of the layer, the rupture speed becomes supershear near the free surface due to a phase conversion, as also observed on linear slip-weakening faults. In contrast, when a velocity-strengthening layer exits, the supershear pulse is suppressed, which could explain the lack of universally observed supershear rupture near the free surface. The addition of the shallow velocity-strengthening layer in the model also suppresses slip on the fault. The slip suppression is larger next to the free surface but there is an appreciable slip reduction over the entire fault. The resulting profile of slip with depth is consistent with observations of shallow co-seismic slip deficit. We are also advancing towards a SEM formulation for modeling long-term slip histories, by combining the dynamic SEM with a quasi-static SEM formulation that we developed for simulations of aseismic slip. We have used the combined formulation to simulate seismic cycles in the form of small repeating earthquakes in a 2D SEM model, and find that the results are in agreement with those of BIM. We will report our current effort to extend the combined SEM formulation to 3D.
S21B-0563
Experimental study on deformation field evolution in rock sample with en echelon faults using digital speckle correlation method
Digital speckle correlation method (DSCM) is one kind of photomechanical deformation measurement method. DSCM could obtain continuous deformation field contactlessly by just capturing speckle images from specimen surface. Therefore, it is suitable to observe high spatial resolution deformation field in tectonophysical experiment. However, in the general DSCM experiment, the inspected surface of specimen needs to be painted to bear speckle grains in order to obtain the high quality speckle image. This also affects the realization of other measurement techniques. In this study, an improved DSCM system is developed and utilized to measure deformation field of rock specimen without surface painting. The granodiorite with high contrast nature grains is chosen to manufacture the specimen, and a specially designed DSCM algorithm is developed to analyze this kind of nature speckle images. Verification and calibration experiments show that the system could inspect a continuous (about 15Hz) high resolution displacement field (with resolution of 5μm) and strain field (with resolution of 50με), dispensing with any preparation on rock specimen. Therefore, it could be conveniently utilized to study the failure of rock structure. Samples with compressive en echelon faults and extensional en echelon faults are studied on a two-direction servo-control test machine. The failure process of the samples is discussed based on the DSCM results. Experiment results show that: 1) The contours of displacement field could clearly indicate the activities of faults and new cracks. The displacement gradient adjacent to active faults and cracks is much greater than other areas. 2) Before failure of the samples, the mean strain of the jog area is largest for the compressive en echelon fault, while that is smallest for the extensional en echelon fault. This consists with the understanding that the jog area of compressive fault subjects to compression and that of extensional fault subjects to tension. 3) For the extensional en echelon sample, the dislocation across fault on load-driving end is greater than that cross fault on fixed end. Within the same fault, the dislocation across branch far from the jog area is greater than that across branch near the jog area. This indicates the restriction effect of jog area on the activity of fault. Moreover, the average dislocation across faults is much greater than that across the cracks. 4) For the compressive en echelon fault, the wing cracks initialized firstly and propagate outwards the jog area. Subsequently, a wedge strain concentration area is initialized and developed in the jog area because of the interaction of the two faults. Finally, the jog area failed when one crack propagates rapidly and connects the two ends of faults. The DSCM system used in this study could clearly show the deformation and failure process of the en echelon fault sample. The experiment using DSCM could be performed dispensing with any preparation on specimen and not affecting other inspection. Therefore, DSCM is expected to be a suitable tool for experimental study of fault samples in laboratory.
S21B-0564
Thermal Field Indicator for Identifying Active Faults and its Instability From Laboratory Experiments
The relationship between the thermal filed and strain field during deformation of faults is the physical basis to clarify whether satellite infrared information and the ground temperature field can be used to study fault activity. This study attempts to discuss these problems by experiments in the laboratory. The two-direction servo-control system was used to load on the samples with compressional and extensional en echelon faults. An infrared thermal image system and a contact-type thermometer recorded synchronously variations of the bright temperature field of infrared radiation and temperature field during deformation of the rock specimens. A digital CCD camera and a soft ware based on the digital speckle correlation method (DSCM) was utilized to capture images and to analyze them, yielding processes of displacement and strain fields. The experimental result shows as follows: 1 The temperature is highest at the jog area of the compressional en echelon faults, whereas that is lowest at the extensional en echelon faults prior to failure of the jog area. The record by DSCM displays that the mean strain of the jog area is largest for the compressional en echelon faults, while that is smallest for the extensional en echelon faults. These mean that the temperature field has clear responses to the opposite stress states at the jog areas of two kinds of en echelon faults, providing an indicator for determining whether the fault segment has slid. 2 The en echelon faults experience two deformation stages from stress building up and fault propagating at the jog area to unstable sliding along the fault. Correspondingly the mechanism of heating-up is turned from strain heating into frictional heating. Three kinds of phenomena have been observed at the jog area and its vicinity during the stage of transformation. They are temperature drop, fast fluctuation of temperature, and pulses of temperature rising, respectively. Mechanism of these phenomena is discussed. 3 These variations of the thermal field at the jog area are followed by swift rise of temperature along the fault. The onset of temperature rise along fault occurs 2-3 seconds prior to its unstable slip. However, the temperature drop of the jog area happens about 20 seconds before the unstable slip of the fault and the appearance of temperature rising pulses is 10-20 seconds earlier than that of the unstable slip. They are precursors to unstable slip of the fault. These experimental analyses demonstrate that observations and studies on the thermal variations at the sensitive portion of a faults comparing with other relevant data are of great importance for detecting precursors ahead of unstable slip of active fault.
S21B-0565
Smooth, Mature Faults Radiate More Energy than Rough, Immature Faults in Parkfield, CA
Laser-based observations of fault surfaces indicate that cumulative slip is inversely related to fault roughness in the direction of slip, suggesting that older, more mature faults are smoother than younger ones with less cumulative slip (Sagy, et al., 2006). Given a difference in roughness between old and young fault surfaces, it is conceivable that earthquakes on older, smoother faults might radiate different amounts of seismic energy than earthquakes of comparable size on younger, rougher faults. The total energy budget of an earthquake is limited by the release of elastic strain energy associated with faulting. If more energy is required to rupture asperities on a rougher fault, a smaller portion of the total budget is available for the radiation of seismic waves. Radiated seismic energy is the only directly observable parameter in the partitioning of the earthquake energy budget, and the ratio of radiated energy to moment therefore provides information about rupture velocity. If fault roughness affects radiated energy, then such geological observations could also provide information about rupture propagation. We compare spectral shapes of earthquakes on old and young faults in the Parkfield area to observe if differences in roughness affect radiated energy. We use earthquakes of comparable magnitude below Mw 3.7, and assume that secondary faults have less cumulative displacement than the San Andreas Fault. The spectra of earthquakes on secondary faults have high-frequency bands depleted in amplitude relative to earthquakes on the San Andreas Fault. Additionally, the cumulative spectral energy based on these spectra give on average three times higher ratios of radiated energy to moment for the earthquakes on the San Andreas Fault. This suggests that mature, i.e. smoother faults are more efficient radiators of seismic energy.
S21B-0566
Shear experiments of granular materials and implications for fault slip
Faults are known to slip irregularly in the form of stick-slip. There are various parameters which affect this motion (e.g., particle size, shear rate). But how each of these parameters affect the motion is unclear. Recently, Anthony and Marone (2005) performed laboratory experiments to investigate how the periodicity or slip are controlled by the particle properties of fault gouge. However, particle size was not varied much and the effect of interstitial fluid was unexplored. Fluid effects are considered to be important for understanding slip in subduction zones (Obara, 2002). Here we present the results of shear experiments of dry and liquid-saturated granular materials to understand how the properties of fault gouge and the imposed shear rate control the long-term statistics of fault slip. We shear sorted glass beads using a rotating viscometer and find that they exhibit stick-slip. We characterize the temporal variation of torque measurements using several scales, and analyze their statistical properties. Under a fixed rotation rate we find that as the particle size increases, the stress drop and the slip recurrence interval increases whereas the degree of creeping prior to the slip decreases. From analyzing the images of the upper surface of sheared granular materials, we find that the radial range where particles are mobile are approximately scaled as 10-15 particle size, and hence the number of particles consisting a force chain is approximately constant. Using these results, we calculate the minimum shear strain needed for a slip to occur and find that for a particle size of 0.196 mm it is 5 × 10-3 for a shear rate of 1 × 10-2 1/s and tends to decrease with increasing particle size. Our results indicate that the difference in the stick-slip behavior primarily arise from the difference in interparticle friction which increases with particle size. When the granular material is saturated with viscous fluid, we find that the interparticle friction is drastically reduced thus affecting the stress drop and degree of preslip creep. Although our granular model of fault slip is much simplified, it demonstrates that for the same far-field slip rate the difference of the effective particle size of fault gauge or the presence of interstial liquid can account for the variation of fault behavior from creeping to stick-slip.
S21B-0567
Hydrological and chemical monitoring during Fluid Injection Test in Taiwan Chelungpu-fault Drilling Project
Taiwan Chelungpu-fault Drilling Project drilled two research boreholes (Hole A and B; approximately 40 m of their distance) through the Chelungpu Fault in Da-Keng, which ruptured in the 1999 Mw 7.6 Chi-Chi earthquake, in 2004. A branched borehole was drilled from Hole B in 2005 (Hole C), and then both Hole A and Hole C were perforated at the depth of the fault zone. The depth of perforation is 1111 m in Hole A and 1137 m in Hole C. Between the two boreholes, Fluid Injection Test (FIT) was performed on from November 2006 to March 2007 to estimate permeability and to understand hydrological and chemical properties along Chelungpu fault. Water was injected four times from Hole C at constant pressure during this FIT (4 MPa on November 2006 and January 2007, 3 and 5 MPa on March 2007). The arrival of injected water was monitored by seismometers, manometers, a Quadrupole Mass Spectrometry and chemical sensors at Hole A. In this present, we will report the results of water quality, gas and water pressure monitoring at Hole A. During FIT, tap water was used for injected water, which was characterized by high Oxidation Reduction Potential (ORP; 250 mV) and high Dissolved Oxygen (DO; 5.6 mg/L). Because both ORP and DO of the well water at Hole A kept low (ORP; -350 - -150, DO; <0.5 mg/L) before FIT, the arrival of injected water can be found by rise of these values. 1st FIT was performed for approximately 100 hours from 22:00 on 7th to 8:30 on 12th November. As a result, the values of ORP and DO increased on 10th November, which is 3 days after the start of 1st FIT. Then, the flow rate at Hole A suddenly increased 7 days after the first chemical reaction on 10th, that is, 10 days after the start of 1st FIT. This suggests that the permeability is 10-16 m2 assuming that the width of a permeable zone is 1 m by the preliminary estimation of the permeability based on the model of Kitagawa et al. (2002).
S21B-0568
High magnetic susceptibility produced in high velocity frictional tests and heating tests on core samples from the Chelungpu fault in Taiwan
We carried out high-velocity frictional tests and simple heating tests on crushed fault gouge from core samples from Hole B of the Taiwan Chelungpu-fault Drilling Project to investigate the cause of high magnetic susceptibilities in the fault core. Black ultracataclasite resembling that observed in Hole B formed during the experiments, even under low axial stress of 0.5 to 1.5 MPa. The bulk magnetic susceptibility of the tested samples was proportional to the frictional work applied and increased as slip increased. Thermomagnetic analysis of the samples before frictional testing revealed that magnetization increased at temperatures above 400°C, probably because of thermal decomposition of paramagnetic minerals. The decomposition of siderite and pyrite that changes to magnetite can explain high magnetic susceptibility and observed thermomagnetic curves. Simple heating also increased magnetic susceptibility though the change is much smaller compared to that by frictional tests. Both the thermally and mechanically induced formation of ferrimagnetic minerals by high velocity friction might have caused a magnetic susceptibility anomaly. Our experimental results support the assumption that heat generation of short duration, even if it is below the melting point, can increase magnetic susceptibility.
S21B-0569
Hydraulic and frictional properties of natural clay-rich sediments from ODP Leg190 Nankai Trough and IODP Expedition 311 Cascadia Margin
Understanding mechanical and hydraulic properties of sediments in accretionary subduction zones undergoing progressive consolidation and localized shear under a wide range of loading rates is crucial to understand the mechanics of subduction earthquakes and accretion processes. We have initiated several suites of experiments to investigate the effects of rate and temperature on mechanical behavior during consolidation and sliding friction using sediment samples recovered from Nankai Trough and Cascadia Margin subduction zones during ODP Leg 190 and IODP Expedition 311, respectively. The experiments include 1) hydrostatic consolidation tests, 2) permeability measurements during consolidation, 3) low-speed friction experiments, and 4) high-speed friction experiments. Several hydrostatic consolidation experiments have been conducted on samples collected from across the proto- decollement zone at ODP Leg190 Site 1173 (360 and 450 mbsf), at elevated pressure and temperature up to 70MPa and 150°C, respectively. Volumetric strain (magnitude of consolidation) increases with the effective pressure and is much greater at higher temperature or after long times, and for the samples taken from above, relative to those from below, the proto-decollement. The pronounced temperature-time sensitivity observed reflects thermally-activated mechanisms of consolidation such as intergranular friction during grain rearrangement, mineral dehydration reactions including the smectite-illite transition, dissolution, and intracrystalline plasticity. Permeability was measured for sediment samples from Nankai and Cascadia under stepwise increasing and decreasing confining pressures up to 70 MPa with a steady-state flow method using Nitrogen gas and water. Permeability varies systematically by confining pressure, sediment type and the collected depth, and pore fluid type. Both low-speed (0.015-15 μm/s) and high-speed (0.1-1.3 m/s) friction experiments were conducted on disaggregated samples from the decollement zone at ODP Leg 190 Site 1174 (839 mbsf). For low-speed friction experiments, 0.2-mm-thick layers were sheared between gabbro blocks in a biaxial-shear apparatus at normal stress of 20-30 MPa. Throughout the experiments, displacement hardening behavior is observed with a friction coefficient ranging from 0.33 to 0.5. The velocity stepping tests show velocity strengthening behavior, with variable b-value; negative b-value is observed at 0.15 μm/s only after the sample experiences the higher slip rate of 15 μm/s. For the high-speed friction experiments, 1-mm-thick layers were sheared between gabbro cylinders in a rotary-shear apparatus at normal stress of 0.3-1.3 MPa. In these tests the initial friction coefficient is approximately 1 and subsequent dynamic weakening is observed only at the slip rate of 1.3 m/s where the friction coefficient drops over 10 m of slip to a steady state value of approximately 0.1. The friction behavior significantly varies with slip rate, and the dynamic weakening during high-speed frictional sliding may be caused by local increases in temperature from frictional heating. All results suggest that both temperature and rate have significant effects on the hydraulic and frictional properties of sediments in accretionary subduction zones.
S21B-0570
Dynamic Source Modeling of the Miyagi-oki Earthquakes
Historical and seismic records indicate that interplate earthquakes with magnitude of ~7.5 have occurred repeatedly in the Miyagi-oki region, northeastern Japan, with a recurrence interval of ~37 years. The 1978 Miyagi-oki earthquake occurred, with magnitude of 7.4, and 27 years later, the 2005 event occurred in this region with magnitude of 7.2. Wu et al. [submitted] estimated rupture processes of the two events using teleseismic and strong motion records. Both events ruptured from the same hypocenters. The 1978 event consists of southern two asperities near the hypocenter and northern large asperity, and the 2005 event repeatedly ruptured only the southern two asperities. Estimation of dynamic source parameters of the two events may offer a key to understanding of reccurrence and interaction between asperities of interplate earthquakes. In this study, we construct dynamic source models of the Miyagi-oki earthquakes and evaluate dynamic source parameters, especially fracture energy, GC, which would be able to be evaluated stably. The dynamic source models are constructed based on kinematic source models. We simulate the spontaneous dynamic rupture on a fault embedded in an infinite homogeneous elastic medium by using the forth-order finite difference method with 3D staggered grid. The grid interval and time-step size are 0.2 km and 0.01 sec, respectively. We assume the slip-weakening model with DC of 0.4 m as a constitutive law on the fault and modify distributions of static stress drop and strength excess by forward modeling. We could construct a dynamic source model of the 2005 event in which the rupture process agrees with kinematic source model in the first asperity, which is close to the rupture nucleation point. In this model, the value of GC is about 3 MJ/m2 in the first asperity. This value is comparable to those of large earthquakes estimated by previous studies [Beroza & Spudich, 1988; Ide, 2002]. However, significant rupture did not occur in the second asperity 40 km away from the rupture nucleation point in the down-dip direction, because the stress did not increase enough to exceed the assumed strength in the second asperity. We will consider conditions for rupture in the second asperity.
S21B-0571
Tensile Microcrack Formation During Experimental Dynamic Shear Rupture Under Uniaxial Loading
Motivated by the occurrence of high-angle pseudotachylyte injection veins along exhumed faults, we studied secondary tensile microcrack formation during dynamic shear rupture in the laboratory. Shear ruptures were induced by an exploding wire embedded along a frictionally held and glued interface in Homalite. During the experiments, the samples were held under a uniaxial load P applied at an angle α to the normal to the rupture interface. Test values of α varied between 30° and 70°, and P varied between 15MPa and 30MPa. The dynamic stress fields produced by the propagating shear rupture were recorded using photoelasticity and high-speed digital photography. Observed isochromatic fringe patterns were similar to contours of maximum shear stress evaluated using a solution for dynamic propagation of a mode II crack with a velocity-weakening endzone. Rupture velocities during experiments ranged between 60% and 90% of the shear wave velocity. Secondary microcracks were produced in the Homalite samples during rupture under multiple loading configurations. In all cases, the region of shear stress concentration was spread over a cohesive endzone (10-20 mm in length) behind the traveling rupture tip. In some cases, microcracks were observed to grow at a finite distance behind the shear rupture tip. We observed several interesting behaviors: (i) some tensile microcracks appear to form due to transient stress concentrations associated with the mode II rupture and are roughly periodic; (ii) some tensile microcracks are associated with rupture termination and are concentrated at the final rupture tips; and (iii) a correlation between rupture velocity and microcrack orientation appears to exist. Future work will address the correlation of secondary tensile fracture orientation and spacing with rupture parameters including velocity, directivity, and remote stress state.
S21B-0572
Dynamic Slip in 2D Discrete Element Method Numerical Simulations of a Rough Fault
2D discrete element method (DEM) simulations are used to investigate the properties of the dynamic rupture of a heterogeneous fault. The model consists of two rectangular blocks of fully bonded particles with a pre-existing fault between the blocks across which the particles are not bonded and interact only by frictional forces. An intrinsic small scale roughness of the fault surface is present due to the construction of the fault model from random spherical particles. Additionally, heterogeneity on a large length scale is introduced, generating asperity and non-asperity regions along the fault by varying the amount of small-scale surface roughness between these regions. Contact friction is defined using a Coulomb Law. The model evolves from a stress-free initial state into stick-slip behaviour while a constant normal stress and a constant shear velocity are applied to the edges of the model. The resulting slip events show a number of properties similar to real seismic events. We observe qualitatively realistic source-time functions, although the absolute slip velocities are too high, realistic stress drops and rupture velocities. The power spectral density (PSD) of the resulting slip distributions is consistent with a fractal distribution, as observed in nature. The results indicate that a simple friction law coupled with geometrical complexity yields many of the characteristic features seen in real rupture propagation.
S21B-0573
Simulations of Stress Drops for Evolving Seismicity on a Heterogeneous Fault in an Elastic Half-space
Understanding what limits the size of earthquake stress drops has important implications for estimating ground motion. Seismological derivations of stress drops are typically of order ~1-10 MPa, yet theoretical estimates based on laboratory friction data and measurements of stress in the crust are two orders of magnitude higher. The discrepancy may stem from the fact that earthquake stress drops are average values over a rupture area that may have highly heterogeneous initial stress, while the theoretical estimates assume an essentially homogeneous stress. We investigate properties of stress drops in simulations of evolving seismicity and stress field on a strongly heterogeneous fault. The model fault (Ben-Zion and Rice, 1993) consists of a set of inherently- discrete slip patches surrounded by a 3-D elastic half-space. The discrete slip patches provide a simple representation of heterogeneities associated with segmentation and other geometrical complexities. Previous studies have shown that the model produces many statistical features of seismicity compatible with observations including frequency-size and temporal event statistics, hypocenter distributions, intermittent criticality, and scaling of source-time functions. We apply similar analyses to understand how stress drops in an evolving heterogeneous stress field depend on rheological properties of the model (spatial distribution of static/kinetic friction values and coefficients of power law creep), hypocentral depth, earthquake magnitude and temporal position within large earthquake cycles.
S21B-0574
A Model for Aperiodicity in Earthquakes
Conditions under which a single oscillator model coupled with Dieterich- Ruina's rate and state dependent friction exhibits chaotic dynamics is stud- ied. Properties of spring-block models are discussed. The parameter values of the system are explored and the corresponding numerical solutions pre- sented. Bifurcation analysis is performed to determine the bifurcations and stability of stationary solutions and we find that the system undergoes a Hopf bifurcation to a periodic orbit. This periodic orbit then undergoes a period doubling cascade into a strange attractor, recognized as broadband noise in the power spectrum. The implications for earthquakes are discussed.
S21B-0575
Direct measurement of the breakdown slip from near-fault strong motion data
Obtaining reliable estimates of the frictional behaviour on earthquake faults is a fundamental task, particularly the breakdown slip Dc, which has an important role on rupture propagation through the earthquake energy budget. Several studies have attempted to estimate Dc indirectly from kinematical analysis of fault ruptures (e.g., Ide and Takeo, JGR, 1997). However, such estimates are complicated because of both the limited band-width of the observed seismograms used to image the rupture process and the rapid decay of high frequencies with distance from the fault. Mikumo et al. (BSSA, 2003) proposed a method to estimate Dc on the fault plane as the slip at the time of the peak sliprate function (Dc'). Fukuyama and Mikumo (GRL, 2007) proposed to extend this method beyond the fault plane, by estimating Dc as twice the rake-parallel particle displacement at the time of the peak particle velocity. The factor of two arises from an equal amount of opposite displacement on either side of the fault. They concluded that such method allows reliable Dc' estimates with negligible dependence on the perpendicular distance from the fault, and used it to obtain Dc' estimates for the 2000 M6.6 Tottori (0.3 m) and the 2002 M7.9 Denali (2.5 m) earthquakes. The study by Fukuyama and Mikumo was based on simple two-dimensional Green's functions in a homogeneous full space for an anti-plane kinematic crack, and suffers from three fundamental omissions: 1) the free surface and heterogeneous structure, 2) the finiteness of the rupture surface and 3) the dynamic rupture complexity of real 3D earthquakes. Here, we re-examine the methodology proposed by Fukuyama and Mikumo by means of a more realistic approach. We use spontaneous rupture propagation simulated by a recently developed and highly accurate approach, namely the staggered-grid split-node (SGSN) method in a fourth-order staggered- grid finite difference method (Dalguer and Day, JGR, 2007). We assume a vertical strike-slip fault governed by both linear and non-linear slip-weakening friction laws. Our results show that both the free surface and the stopping phases strongly affect Dc estimates. The particle motion recorded by surface instruments is amplified roughly by a factor of two due to the presence of the free surface. As a consequence, the method by Fukuyama and Mikumo over-estimates Dc when applied to strong motion data recorded on the earth's surface. Moreover, contrary to the results by Fukuyama and Mikumo, we observe a strong distance-dependence of the Dc estimates perpendicular to the fault. This variation includes a minimum near the fault, increasing up to about 140% of the target Dc value at a distance 2-3 km from the fault. At further distances from the fault the Dc estimate decreases to about 60% of the target value 10 km away. This distance dependence of the Dc estimate is presumably caused mainly by stopping phases propagating from the fault boundaries. Simulations in heterogeneous media including a low-velocity layer, intrinsic attenuation (Q) and stochastic initial stress conditions allow us to asses the reliability and uncertainty involved in the method proposed by Fukuyama and Mikumo. Dc estimates under these realistic conditions are important but remain below a factor of two in most of the cases we have analyzed. In summary, the accuracy of the method is strongly affected by the presence of the free surface, finite fault extent, and likely by complexity in the velocity structure and rupture propagation.
S21B-0576
3D Dynamic Crack Rupture by a Finite Volume Method
Dynamic rupture of a 3D spontaneous crack of arbitrary shape has been investigated using a Finite Volume (FV) approach. The full domain is decomposed in tetrahedra while the surface on which the rupture is supposed to take place is discretized with triangles which are faces of tetrahedra. Because of this meshing strategy, any shape of the rupture surface could be designed and is performed once before simulations start. First of all, the elastodynamic equations are described into a pseudo-conservative form for easy application of the FV discretisation. Explicit boundary conditions are given using criteria based on the conservation of discrete energy through the crack surface. Using a stress-threshold criterion, these conditions specify fluxes through those triangles which have suffered rupture. On these broken surfaces, stress follows A linear slip-weakening law although other friction laws can be implemented as well. Numerical solutions on a planar fault are achieved for the problem version 3 of the SCEC community dynamic-rupture benchmark exercise (Harris and Archuleta, 2004) and compared with those provided by a Finite Difference (FD) technique (Day et al, 2005). Another benchmark problem is also tackled involving a nonplanar curved fault (Cruz-Atienza et al, 2007). Solutions for this difficult exercise are compared with those computed with a Boundary Integral (BI) method (Aochi et al, 2000). In both benchmarck problems, comparisons show that rupture fronts are well modelled with a slight delay in time especially along the antiplane direction related to the low-order interpolation of the FV approach which requires further mesh refinement or/and an higher-order interpolation strategy as for Galerkin Discontinuous approach. Slip-rate and shear stress amplitudes are well modelled as well as stopping phases and stress overshoots. We expect this method, which is well adapted to multi-preocessor parallel computing to be competitive with others for solving large scale dynamic ruptures scenario of seismic sources in the near future. References : Aochi, H., E. Fukuyama and M. Matsuura, 2000. Spontaneous rupture propagation of a non-planar fault in 3D elastic medium, PAGEOPH, 157, 2003-2027. Cruz-Atienza, V.M., J. Virieux, J. and H. Aochi, 3D finite-difference dynamic-rupture modeling along nonplanar faults, Geophysics, 72, SM123-SM137. Day, S. M., L.A. Dalguer, N. Lapusta and Y. Liu, 2005, Comparison of finite difference and boundary integral solutions to three-dimensional spontaneous rupture: Journal of Geophysical Research, 110, B12307, http://dx.doi.org/10.1029/2005JB003813. Harris, R. A. and R. J. Archuleta, 2004, Earthquake rupture dynamics: Comparing the numerical simulation methods: EOS, 85, 321.
S21B-0577
Galerkin boundary integral equation method for spontaneous rupture propagation problems
We develop a Galerkin finite element boundary integral equation method (GaBIEM) for spontaneous rupture propagation problems for a planar fault embedded in a homogeneous full 2D space. A simple 2D anti plane rupture propagation problem, with a slip-weakening friction law, is simulated by the GaBIEM. This method allows one to separate explicitly the kernel into singular static and time-dependent parts, and a nonsingular dynamic component. The simulated results throw light into the performance of the GaBIEM and highlight differences with respect to that of the traditional, collocation, boundary integral equation method (BIEM). The rate of convergence of the GaBIEM, as measured from a root mean square (RMS) analysis of the difference of approximate solutions corresponding to increasingly finer element sizes is of a higher order than that of the BIEM. There is no restriction on the CFL stability number since an implicit, unconditionally stable method is used for the time integration. The error of the approximation increases with the time step, as expected, and it can remain below that of the BIEM.